You are on page 1of 546

SYLLABUS

NUEN 601 Nuclear Reactor Theory


Spring 2017
Lecture MWF 11:30-12:20 AIEB M309

Catalog Description

Nuclear Reactor Theory. (3-0). Credit 3. Neutron-nucleus interactions; neutron energy spectra;
transport and diffusion theory; multigroup approximation; criticality calculations; cross-section
processing; buildup and depletion calculations; modern reactor analysis methods and codes.
Prerequisite: Approval of instructor.

Detailed Course Description

NUEN 601 is a 1st-year graduate-level course. Within the Nuclear Engineering Master of
Science degree curriculum, it is the detailed study of theory of nuclear reactor behavior.

Topics addressed in the course include:


1. Physics of nuclear fission, the nuclear chain reaction, and criticality
2. Neutron cross sections, cross section energy dependence, and cross section processing
3. Neutron transport and diffusion theory
4. Diffusion theory applied to bare homogeneous reactors and reflected reactors, one-
group solutions for bare and reflected homogeneous reactors
5. Neutron slowing down and energy spectra
6. Energy-dependent diffusion and the multigroup approximation to the diffusion equation,
two-group solutions for bare and reflected homogeneous reactors
7. Time-dependent diffusion solutions and kinetic behavior of reactor systems
8. Buildup and depletion calculations; fission-product poisoning
9. Nuclear fuels and material performance under irradiation, heat, and stress: basics
10. Fundamentals of heat transfer, fluid flow, and thermodynamics
11. Reactivity feedback
12. Nuclear fuel cycles: basics
13. Assembly and full core neutronic design and analysis: basics

Class Time and Location

NUEN 601 is a 3-hour lecture course. Lectures are scheduled for:

Mon., Wed., Fri., 11:30-12:20, CHEN (Brown Bldg @ University & Spence) Room 111

There will also be an optional but strongly recommended help session, approximately weekly, at
a time and place to be determined.

Prerequisites

Students entering this course should have successfully completed NUEN 604 or equivalent.

1
Students are expected to bring a working knowledge of many of the theoretical concepts
employed in this class. For example, students should have had some previous introduction to
the following topics:

1. Basic nuclear and atomic physics


2. Sources of radiation (alpha, beta, x-ray, gamma-ray, and neutron sources)
3. Interaction of radiation with matter (gamma-rays, charged-particle, and neutron
interactions)
4. Radiation detection (including the operation of ion chambers, gamma and alpha
spectroscopy systems, and basic neutron detection systems)
5. Counting statistics
6. Radiation dose, radiation exposure, and dosimetry
7. Basic concepts of radiation safety
8. Radiation shielding and buildup factors
9. Fundamentals of the Monte Carlo method for nuclear systems modeling
10. Ordinary differential equations (ODEs), partial differential equations (PDEs),
eigenvalues, eigenvectors, and systems of linear equations.
11. Scalar flux and cross sections

Course Learning Objectives

Students who successfully complete this course should be able to:

1. Demonstrate deep understanding of criticality and explain it to non-experts


2. Describe key features of cross sections of nuclides of major interest in nuclear reactor
design, and explain how these features enable reactors to function as they do
3. Describe delayed neutrons and explain their importance to nuclear reactors
4. Explain differences and similarities among time-dependent, steady-state, and eigenvalue
problems, and explain physical situations to which each is applicable
5. Compute reaction rates and leakage rates
6. Apply the diffusion approximation to neutrons in a reactor and demonstrate quantitative
understanding of how neutrons distribute themselves spatially
7. Apply slowing-down equations to neutrons in a reactor and demonstrate quantitative
understanding of how neutrons distribute themselves in energy
8. Apply time-dependent conservation equations, including the Point Reactor Kinetics
Equations (PRKEs), to neutrons in a reactor and demonstrate quantitative understanding
of how neutron populations change with time
9. Derive the neutron transport equation, and apply to it the diffusion approximation and/or
the multigroup approximation
10. Compute and explain the expected behavior of a nuclear reactor under changes to
reactivity due to control elements, fission products, and burnup
11. Present clear explanations of nuclear reactor phenomena in writing and speaking
12. Explain the basics of heat transfer and fluid flow, and describe reactivity feedback
13. Explain how fuel composition and physical properties change over time in a nuclear
reactor
14. Explain the basics of the fuel cycle as it relates to commercial reactors, from the mine to
ultimate disposition

2
Instructor Information

Name: Marvin L. Adams, Ph.D., P.E.


Titles: HTRI Professor of Nuclear Engineering
Director, Institute for National Security Education and Research
Deputy Director, Center for Exascale Radiation Transport
Assoc. Director, Center for Large-scale Scientific Simulation
Telephone Number: (979) 845-4198
Email address: mladams@tamu.edu
Office Hours: help session, time TBD, + by appointment
Office Location: AIEB, Room M303

We will find a time and place for a help session ~once/week. This will be devoted to working
problems in small groups and answering individual questions. In addition, I encourage you to
email me if you have questions that can be answered by email, or if you want an individual
appointment or small-group outside of the group help session.

Teaching Assistant (and Homework Grader)

Name: Christopher Fullerton


Email: cfullert@tamu.edu
Office Hours: TBD
Office Location: TBD

Textbooks and/or Resource Materials

Lecture Notes
The primary reference for this course is a set of lecture notes that will be provided via the
universitys eCampus system. Students should print these notes and bring them to class for
each lecture session. The notes include blank spaces that students are to fill in during lectures.

Required Text
The required text for this class is:
1. E.E. Lewis, Fundamentals of Nuclear Reactor Physics, Elsevier, Amsterdam (2008)
and ISBN 9780080560434, Academic Press, available for free download

Reading assignments will be provided for many lecture sessions. Students should read these
assignments prior to the lectures. This book is available for FREE from the librarys website.

Supplemental (or Optional) Texts


Due to the vast amount of information covered in this class, the student may find it useful to
obtain some of the following texts as supplemental information:
2. J. R. Lamarsh and A. Barratta, Introduction to Nuclear Engineering, 3rd Edition,
Addison-Welsey Publishing Company, Reading, PA (2001).
3. J.J. Duderstandt and L.J. Hamilton, Nuclear Reactor Analysis, John Wiley & Sons,
(1976).
4. W. M. Stacey, Nuclear Reactor Physics, Second Edition, ISBN 3527406794, John
Wiley & Sons, New York (2007).
http://onlinelibrary.wiley.com/book/10.1002/9783527611041

3
Specific reading assignments from these supplemental texts will not be given; however, reading
from these texts may prove useful if the student does not fully comprehend the information
provide in lecture or in the required text (Lewis).

Grading Policies

The students final grade will be determined using the following percentages. For each student
individually I will choose the weights within the given ranges to maximize the overall grade.
Thus, if you do poorly on one segment its weight will be at the low end of the range, but if you
do well on a segment its weight will be at the high end.

Homework: 25-35%
Regular Exams: 25-40%
Final Examination: 25-40%

Homework
Homework assignments will consist of problem sets designed to assist with mastery of the
course material. Homework will be assigned approximately every week or two.

I encourage you to work together on homework. Learn from each other discuss the problems,
but don't copy from each other. Copying will induce penalties! List the classmates you worked
with on the first page of each assignment you turn in. I may require that certain problems be
worked alone. If you have trouble with a problem, use your resources your book, your notes,
your fellow students, your TA, and me to get help.

Each homework submitted should include (1) the students name on the front page, (2) the
assignment number, (3) neat, legible work organized so the grader can follow it, (4) units for
each solution that should have them, (5) a staple in the upper left corner if there are multiple
pages.

Regular Examinations
Two written examinations will be conducted during the semester. These exams will be in-class,
closed-book, closed-notes, no-electronics exams and will cover all material up to the date of the
exam. Review sheets (study questions) will be provided at least one week in advance of each
exam.

Final Examination
A final examination for the class will be given according to the University Final Examination
Schedule (10:30 12:30 a.m., May 9, 2017). This exam will be comprehensive and cover all
information discussed in lectures, laboratory sessions, laboratory reports, and homework. A
review sheet (study questions) will be provided at least one week before the exam.

Late-HW Policy
Homework is due at the start of class on the due date. Penalties for late homework are:
until 5:00 on due date: 10% of points earned
until 5:00 next working day: 20%
until last class of semester: 30%
after last class of semester: no credit
Last class is at 11:30 a.m., May 1, 2017.

4
Other Pertinent Course Information

All of the material for this course will be maintained on the Universitys eCampus system. This
includes an electronic copy of this syllabus, lecture notes, assignments, and study material.
The instructor will use the university email system to communicate important messages to the
students. Students should check email often to stay updated. The eCampus system can be
accessed through http://eCampus.tamu.edu. If you are unfamiliar with this system, instruction will
be provided.

Americans with Disabilities Act (ADA)

The Americans with Disabilities Act (ADA) is a federal anti-discrimination statute that provides
comprehensive civil rights protection for persons with disabilities. Among other things, this
legislation requires that all students with disabilities be guaranteed a learning environment that
provides for reasonable accommodation of their disabilities. If you believe you have a disability
requiring an accommodation, please contact Disability Services, in Cain Hall, Room B118, or
call 845-1637. For additional information visit http://disability.tamu.edu

Religious Holidays

If you are a member of a religious faith that has one or more holidays which require you to be
absent from any class listed above, please tell your instructor at least two weeks in advance of
your absence and make arrangements to make-up the class.

Academic Integrity

All students at Texas A&M University are bound by the Aggie Honor Code:

An Aggie does not lie, cheat or steal, or tolerate those who do.

For more information, the student is referred to the Honor Council Rules and Procedures on the
web at http://www.tamu.edu/aggiehonor.

As commonly defined, plagiarism consists of passing off as one's own the ideas, work, writings,
etc., that belong to another. In accordance with this definition, you are committing plagiarism if
you copy the work of another person and turn it in as your own, even if you have the permission
of that person. Plagiarism is one of the worst academic sins, for the plagiarist destroys the trust
among colleagues without which research cannot be safely communicated. If you have
questions regarding plagiarism, please consult the latest issue of the Texas A&M University
Student Rules [http://student-rules.tamu.edu/], under the section Scholastic Dishonesty.

5
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Chapter 1

Nuclear Reactors: A Historical


Perspective

We are grateful to Dr. Jean Ragusa for assembling this chapter.

For additional information, the internet has much to offer

1.1 A Brief History

Before we mention the first human-made1 nuclear reactor, the Chicago Pile number 1 (CP-1), and
its reaching criticality in 1942, we will briefly look back over about 50 years of research that led to
the criticality of CP-1. It is customary to choose 1896 as the starting point of the nuclear era.

1.1.1 Before 1896

Elements are the building blocks of matter. The smallest particle of an element that still retains the
identity of that element is the atom. All atoms of a given element are identical to one another, but
differ from the atoms of other elements. Ancient Greeks (Democrite) first predicted the existence
of the atom around 500 BC. They named the predicted particle atomos, meaning indivisible.

In 1803, John Dalton (1766-1844) proposed a systematic set of postulates to describe the atom.
Daltons work paved the way for modern-day acceptance of the atom. Similarly, in 1808, Louis
Joseph Gay-Lussac (1778-1850) published his memoir on the combination of gaseous substances
1
Nature assembled critical nuclear reactors on earth long before humans did. Search the internet for the Oklo
reactors, for example. Fascinating story!

2
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

with each other. Even though his work supported Daltons atomic theory, Dalton didnt accept
Gay-Lussacs work. But the Italian Amedeo Avogadro (1776-1856) saw it as the key to a bet-
ter understanding of molecular constituency. Atomic theories of Dalton, Gay-Lussac and others
were the only theories that permitted people to understand the periodic table, proposed in 1869 by
Dmitri Mendeleiev (1834-1907). Nonetheless, many other scientists of the 18th and 19th centuries
considered the atom to be merely a subordinate player in chemical reactions, an uninteresting, ho-
mogeneous, positively charged glob that contained scattered electrons (the plum-pudding model
of the atom, aka Thomsons model). That premise remained unchallenged until the end of the 19th
century, when a series of brilliant discoveries opened the door to the atomic science of the twentieth
century. Working concurrently and often collaboratively, three pioneering scientists helped release
the genie of the atom.

1.1.2 Early Breakthroughs

1.1.2.1 The fortuitous discovery of radioactivity

Henri Becquerel (1852-1908), a French physicist, was the son and grandson of physicists. Becquerel
was familiar with the work of Wilhelm Conrad Roentgen on December 22 1895, photographed
his wifes hand, revealing the unmistakable image of her skeleton, complete with wedding ring.
Roentgens wife had placed her hand in the path of X-rays which Roentgen created by beaming
an electron ray energy source onto a cathode tube. Roentgens discovery of these mysterious
rays capable of producing an image on a photographic plate excited scientists of his day, including
Becquerel. Becquerel chose to study the related phenomena of fluorescence and phosphorescence.
In March of 1896, quite by accident, he made a remarkable discovery.

Becquerel found that, while the phenomena of fluorescence and phosphorescence had many simi-
larities to each other and to X-rays, they also had important differences. While fluorescence and
X-rays stopped when the initiating energy source was halted, phosphorescence continued to emit
rays some time after the initiating energy source was removed. However, in all three cases, the
energy was derived initially from an outside source.

In March of 1896, during a time of overcast weather, Becquerel found he couldnt use the sun as
an initiating energy source for his experiments. He put his wrapped photographic plates away in
a darkened drawer, along with some crystals containing uranium. Much to Becquerels surprise,
the plates were exposed during storage by invisible emanations from the uranium. The emanations
did not require the presence of an initiating energy sourcethe crystals emitted rays on their own!
Although Becquerel did not pursue his discovery of radioactivity, others did and, in so doing,
changed the face of both modern medicine and modern science.

1.1.2.2 Po and Ra: Marie Sklodowska-Curie (1867-1934) and Pierre Curie (1859-
1906)

Working in the Becquerel lab, Marie Curie and her husband, Pierre, began what became a
life-long study of radioactivity. It took fresh and open minds, along with much dedicated work, for

3
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

these scientists to establish the properties of radioactive matter. Marie Curie wrote, The subject
seemed to us very attractive and all the more so because the question was entirely new and nothing
yet had been written upon it.

Becquerel had already noted that uranium emanations could turn air into a conductor of electricity.
Using sensitive instruments invented by Pierre Curie and his brother, Pierre and Marie Curie
measured the ability of emanations from various elements to induce conductivity. On February 17,
1898, the Curies tested an ore of uranium, pitchblende, for its ability to turn air into a conductor
of electricity. The Curies found that the pitchblende produced a current 300 times stronger than
that produced by pure uranium. They tested and recalibrated their instruments, and yet they still
found the same puzzling results. The Curies reasoned that a very active unknown substance in
addition to the uranium must exist within the pitchblende. In the title of a paper describing this
hypothesized element (which they named polonium after Maries native Poland), they introduced
the new term: radio-active.

After much grueling work, the Curies were able to extract enough polonium and another radioactive
element, radium, to establish the chemical properties of these elements. Marie Curie, first with
her husband and then continuing after his death, established the first quantitative standards by
which the rate of radioactive emission of charged particles from elements could be measured and
compared. In addition, she found that there was a decrease in the rate of radioactive emissions
over time and that this decrease could be calculated and predicted. But perhaps Marie Curies
greatest and most unique achievement was her realization that radiation is an atomic property of
matter rather than a separate independent emanation.

Despite the giant step forward which science had now taken in its understanding of radioactivity,
scientists still understood little of the structure of the atom. This understanding awaited the work
of Ernest Rutherford.

1.1.2.3 Ernest Rutherfords model of the atom

In 1911, Rutherford (1871-1037) conducted a series of experiments in which he bombarded a


piece of gold foil with positively charged (alpha) particles emitted by radioactive material. Most of
the particles passed through the foil undisturbed, suggesting that the foil was made up mostly of
empty space rather than of a sheet of solid atoms. Some alpha particles, however, bounced back,
indicating the presence of solid matter. Atomic particles, Rutherfords work showed, consisted
primarily of empty space surrounding a well-defined central core called a nucleus. His atomic
model superseded quickly the plum-pudding model that had been proposed by J.J. Thomson (who
was his former boss!). (Lest you think Thomson was not an outstanding scientist, remember that
he discovered the electron.)

In a long and distinguished career, Rutherford laid the groundwork for the determination of atomic
structure. In addition to defining the planetary model of the atom, he showed that radioactive ele-
ments undergo a process of decay over time. And, in experiments which involved what newspapers
of his day called splitting the atom, Rutherford was the first to artificially transmute one element
into another.

4
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Rutherford eventually coined the terms for some of the most basic principles in the field: alpha,
beta, and gamma rays, the proton, the neutron, half-life, and daughter atoms. Several of the
centurys giants in physics studied under him, including Niels Bohr, James Chadwick, and Robert
Oppenheimer.

1.1.3 On the road to fission

1.1.3.1 The neutron

In 1932, James Chadwick (1891-1974) discovered the neutron.

1.1.3.2 Artificial radioactivity

In 1934, Frederic Joliot (1900-1958) and his wife Irene Joliot-Curie (1897-1956), discovered the
artificial radioactivity using an alpha-source. When bombarded with alpha particle, aluminum 27
(usual aluminum) produces one neutron and phosphorous 30. The latter beta-decays in 2.5 minutes.

1.1.3.3 Neutron induced reactions

Since the discovery of the neutron by Chadwick in 1932, nuclear physicists used neutrons to bom-
bard many elements. It was easy to get the neutron to the nuclei because of the neutral charge
of the neutron. If the neutron is captured, an isotope of the initial target nucleus is obtained.
The new isotope may be radioactive, in which case it decays, often by beta-decay. Enrico Fermi
(1901-1954), among others, performed research on neutron-induced reactions. Fermi was mostly
interested in discovering and filling the blanks in the periodic table. He bombarded uranium (which
at that time was the last known element) with neutrons. He also wanted to explore the domain past
uranium, the transuranics element. To his surprise, he obtained many more radioactive elements
than he was hoping for.

1.1.3.4 Fission

Fermis experiment was reproduced in many laboratories. It took 4 years for an answer to come. In
1938, Otto Hahn and Fritz Strassman showed the presence of barium as a result of bombardment
of uranium with neutrons. Since Ba is an element of intermediate mass, the conclusion is obvious:
uranium had fissioned after absorbing a neutron. Lisa Meitner calculated that fission released a
huge amount of energy.

Researchers around the world reproduced the fission experiments. Soon, it was discovered that
fission releases 3 secondary neutrons (a figure that was a little bit optimistic!), which led to the
idea of creating a self-sustained chain reaction.

Francis Perrin was the first to mention the idea of critical size (equivalent to critical mass).

5
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

1.1.4 WW II

The next parts of the story are well known:

the fear of Nazi Germany,

Szilards visit to Einstein and Einsteins letter to Pres. Roosevelt,

the Manhattan project (which we will discuss more in detail when we discuss the CP-1 reactor)

the uranium bomb, Hiroshima, 8/6/1945

the plutonium bomb, Nagasaki, 8/9/1945

1.2 The Chicago Pile -1

1.2.1 Details and pictures

Construction of CP-1, or Chicago Pile Number One, was done under the football stadium in an
abandoned squash court. On December 2, 1942, mankind first harnessed the energy of the atom.
Fermis pile produced only 1/2 watt of power. But that was all the power the United States needed
to start the next phase of the bombs development.

The pile contained 771,000 pounds of graphite, 80,590 pounds of uranium oxide and 12,400 pounds
of uranium metal when it went critical. It cost about $1 million to produce and build. The pile
took the form of a flattened ellipsoid which measured 25 feet wide and 20 feet high.

6
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Figure 1.1: The 7th layer of graphite and edges of 6th layer containing 3-inch pseudospheres of
black uranium oxide. Beginning with layer 6, alternate courses of graphite containing uranium
metal and/or uranium oxide fuel were separated by layers of solid graphite blocks.

7
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Figure 1.2: Tenth layer of graphite blocks containing pseudospheres of black and brown uranium
oxide. The brown briquets, slightly richer in uranium, were concentrated in the central area. In
the foreground and on either side are cavities filled with graphite, now presumed to have been an
expedient measure dictated by shortage of fuel and, possibly, a last minute change in the lattice
arrangement.

8
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Figure 1.3: Construction detail

9
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Figure 1.4: Drawing of CP-1

Figure 1.5: A painting of the CP-1 scene

10
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Figure 1.6: Log chart from CP-1 initial criticality

1.2.2 Already many lessons learned

Developments occurred with amazing speed in the early days of nuclear engineering. Recall that
the neutron was discovered in 1932, fission in 1938 (published in 1939), and then Fermi had a
critical reactor operating before the end of 1942. Equally amazing was how quickly the pioneers
developed an understanding of concepts and developed procedures for dealing with fission reactors.
For example:

1. Heterogeneous core configuration:


Fissions occur in the uranium, neutrons stream out into the graphite, where they lose energy
(slow down) by scattering and have a good chance of escaping parasitic absorption in 238 U
while they are slowing down. Most neutrons eventually reach thermal equilibrium (low en-
ergy), and then most of them migrate back into the uranium, where slow neutrons are likely
to cause fission. If the graphite and uranium were uniformly mixed, too many neutrons would
be parasitically absorbed in U-238 while they were slowing down, and even an infinitely large
reactor could not go critical unless the uranium were enriched in 235 U.

2. Instrumentation and control (I&C) systems were present:


One operator was keeping an eye on the console while the other one was in charge on the
Cadmium control rod. Cadmium is a neutron absorber. When the control rod is inserted,

11
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

more neutrons are absorbed, stopping the chain reaction; when it is withdrawn, more neu-
trons are available for additional fissions, enhancing the chain reaction. The communication
between instrumentation and control allows for the desired operation functions. (Here that
communication was verbal!)

3. Safety:
Besides I&C, additional safety measures are needed. At CP-1, in case of an emergency, a
safety rod, maintained outside of the core through a rope that ran up to the balcony, would
be released by cutting the rope with an ax. This rod would be released with the Safety
Control Rod Ax Man (SCRAM) cut the rope. The last line of defense consisted of a liquid-
control squad that stood on a platform, ready to flood the pile with a cadmium-salt solution.

4. Radiation protection:
The detector (galvanometer) suspended outside of the reactor would measure the ambient
radiation.

1.2.3 Detailed Story of The First Pile

By Corbin Allardice and Edward R. Trapnell

On December 2, 1942, man first initiated a self-sustaining nuclear chain reaction, and controlled it.

Beneath the West Stands of Stagg Field, Chicago, late in the afternoon of that day, a small group
of scientists witnessed the advent of a new era in science. History was made in what had been a
squash-rackets court.

Precisely at 3:25 p.m.,Chicago time, scientist George Weil withdrew the cadmium-plated control
rod and by his action man unleashed and controlled the energy of the atom.

As those who witnessed the experiment became aware of what had happened, smiles spread over
their faces and a quiet ripple of applause could be hear. It was a tribute to Enrico Fermi, Nobel
Prize winner, to whom, more than to any other person, the success of the experiment was due.

Fermi, born in Rome, Italy, on September 29, 1901, had been working with uranium for many
years. In 1934 he bombarded uranium with neutrons and produced what appeared to be element
93 (uranium is element 92) and element 94. However, after closer examination it seemed as if nature
had gone wild; several other elements were present, but none could be fitted into the periodic table
near uranium where Fermi knew they should have fitted if they had been the transuranic elements
92 and 94. It was not until five years later that anyone, Fermi included, realized he had actually
caused fission of the uranium and that these unexplained elements belonged back in the middle
part of the periodic table.

Fermi was awarded the Nobel Prize in 1938 for his work on transuranic elements. He and his
family went to Sweden to receive the prize. The Italian Fascist press severely criticized him for not

12
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

wearing a Fascist uniform and failing to give the Fascist salute when he received the award. The
Fermis never returned to Italy.

From Sweden, having taken most of his personal possessions with him, Fermi proceeded to London
and thence to America where he has remained ever since.

The modern Italian explorer of the unknown was in Chicago that cold December day in 1942. An
outsider looking into the squash court where Fermi was working would have been greeted by a
strange sight. In the center of the 30- by 60-foot room, shrouded on all but one side by a gray
balloon cloth envelope, was a pile of black bricks and wooden timbers, square at the bottom and a
flattened sphere on top. Up to half of its height, its sides were straight. The top half was domed,
like a beehive. During the construction of this crude appearing but complex pile (the name which
has since been applied to all such devices) the standing joke among the scientists working on it
was: If people could see what were doing with a million-and-a-half of their dollars, theyd think
we are crazy. If they knew why we are doing it, theyd know we are.

In relation to the fabulous atomic bomb program, of which the Chicago Pile experiment was a
key part, the successful result reported on December 2nd formed one more piece for the jigsaw
puzzle which was atomic energy. Confirmation of the chain reactor studies was an inspiration to
the leaders of the bomb project, and reassuring at the same time, because the Armys Manhattan
Engineer District had moved ahead on many fronts. Contract negotiations were under way to build
production-scale chain reactors, land had been acquired at Oak Ridge, Tennessee, and millions of
dollars had been obligated.

Three years before the December 2nd experiment, it had been discovered that when an atom
of uranium was bombarded by neutrons, the uranium atom sometimes was split, or fissioned.
Later, it had been found that when an atom of uranium fissioned, additional neurons were emitted
and became available for further reaction with other uranium atoms. These facts implied the
possibility of a chain reaction, similar in certain respects to the reaction which is the source of the
suns energy. The facts further indicated that if a sufficient quantity of uranium could be brought
together under the proper conditions, a self-sustaining chain reaction would result. This quantity
of uranium necessary for a chain reaction under given conditions is known as the critical mass, or
more commonly, the critical size of the particular pile.

For three years the problem of a self-sustaining chain reaction had been assiduously studied. Nearly
a year after Pearl Harbor, a pile of critical size was finally constructed. It worked. A self-sustaining
nuclear chain reaction was a reality.

Construction of the Pile

Construction of the main pile at Chicago started in November. The project gained momentum,
with machining of the graphite blocks, pressing of the uranium oxide pellets, and the design of
instruments. Fermis two construction crews, one under Zinn and the other under Anderson,
worked almost around the clock. V.C. Wilson headed up the instrument work.

13
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Original estimates as to the critical size of the pile were pessimistic. As a further precaution, it
was decided to enclose the pile in a balloon cloth bag which could be evacuated to remove the
neutron-capturing air.

This balloon cloth bag was constructed by Goodyear Tire and Rubber Company. Specialists in
designing gasbags for lighter-than-air craft, the companys engineers were a bit puzzled about the
aerodynamics of a square balloon. Security regulations forbade informing Goodyear of the purpose
of the envelope and so the Armys new square balloon was the butt of much joking.

The bag was hung with one side left open; in the center of the floor a circular layer of graphic
bricks was placed. This and each succeeding layer of the pile was braced by a wooden frame.
Alternate layers contained the uranium. By this layer-on-layer construction a roughly spherical
pile of uranium and graphite was formed.

Facilities for the machining of graphite bricks were installed in the West Stands. Week after week
this shop turned out graphite bricks. This work was done under the direction of Zinns group, by
skilled mechanics led by millwright August Knuth. In October, Anderson and his associates joined
Zinns men.

Describing this phase of the work, Albert Wattenberg, one of Zinns group, said, We found out
how coal miners feel. After eight hours of machining graphite, we looked as we were made up for
a minstrel. One shower would remove only the surface graphite dust. About a half-hour after the
first shower the dust in the pores of your skin would start oozing. Walking around the room where
we cut graphite was like walking on a dance floor. Graphite is a dry lubricant, you know, and the
cement floor covered with graphite dust was slippery.

Before the structure was half complete, measurements indicated that the critical size at which the
pile would become self-sustaining was somewhat less than had been anticipated in the design.

Computations Forecast Success

Day after day the pile grew toward its final shape. And as the size of the pile increased, so did
the nervous tension of the men working on it. Logically and scientifically they knew this pile
would become self-sustaining. It had to. All the measurements indicated that it would. But still
the demonstration had to be made. As the eagerly awaited moment drew nearer, the scientists
gave greater and greater attention to details, the accuracy of measurements, and exactness of their
construction work.

Guiding the entire pile construction and design was the nimble-brained Fermi, whose associated
described him as completely self-confident but wholly without conceit.

So exact were Fermis calculations, based on the measurements taken from the partially finished
pile, that days before its completion and demonstration on December 2, he was able to predict
almost to the exact brick the point at which the reactor would become self-sustaining.

But with all their care and confidence, few in the group knew the extent of the heavy bets being
placed on their success. In Washington, the Manhattan District had proceeded with negotiations

14
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

with I duPont de Nemours and Company to design, build, and operate a plant based on the
principles of the then unproved Chicago pile. The $350,000,000 Hanford Engineer Works10 at
Pasco, Washington, was to be the result.

At Chicago during the early afternoon of December 1st, tests indicated that critical size was rapidly
being approached. At 4:00 p.m. Zinns group was relieved by the men working under Anderson.
Shortly afterwards, the last layer of graphite and uranium bricks was placed on the pile. Zinn, who
remained, and Anderson made several measurements of the activity within the pile. They were
certain that when the control rods were withdrawn, the pile would become self-sustaining. Both
had agreed, however, that should measurements indicate the reaction would become self-sustaining
when the rods were withdrawn, they would not start the pile operating until Fermi and the rest
of the group could be present. Consequently, the control rods were locked and further work was
postponed until the following day.

That night the word was passed to the men who had worked on the pile.

Assembly for the Test

About 8:30 on the morning of Wednesday, December 2nd, the group began to assemble in the
squash court.

At the north end of the squash court was a balcony about ten feet above the floor of the court.
Fermi, Zinn, Anderson, and Compton were grouped around instruments at the east end of the
balcony. The remainder of the observers crowded the little balcony. R. G. Nobles, one of the young
scientists who worked on the pile, put it this way: The control cabinet was surrounded by the
big wheels; the little wheels had to stand back.

On the floor of the squash court, just beneath the balcony, stood George Weil, whose duty it was
to handle the final control rods. In the pile were three sets of control rods. One set was automatic
and could be controlled from the balcony. Another was an emergency safety rod. Attached to one
of this rod was a rope running through the pile and weighted heavily on the opposite end. The rod
was withdrawn from the pile and tied by another rope to the balcony. Hilberry was ready to cut
this rope with an axe should something unexpected happen, or in case the automatic safety rods
failed. The third rod, operated by Weil, was the one which actually held the reaction in check until
withdrawn the proper distance.

Since this demonstration was new and different from anything ever done before, complete reliance
was not placed on mechanically operated control rods. Therefore, a liquid-control squad, com-
posed of Harold Lichtenberger, W. Nyer, and A. C. Graves, stood on a platform above the pile.
They were prepared to flood the pile with cadmium-salt solution in case of mechanical failure of
the control rods.

Each group rehearsed its part of the experiment.

At 9:45 Fermi ordered the electrically operated control rods withdrawn. The man at the controls
threw the switch to withdraw them. A small motor whined. All eyes watched the lights which
indicated the rods position.

15
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

But quickly, the balcony group turned to watch the counters, whose clicking stepped up after the
rods were out. The indicators of these counters resembled the face of a clock, with hands to
indicate neutron clock. Nearby was a recorder, whose quivering pen traced the neutron activity
within the pile.

Shortly after ten oclock, Fermi ordered the emergency rod, called Zip, pulled out and tied.

Zip out, said Fermi. Zinn withdrew Zip by hand and tied it to the balcony rail. Weil stood
ready by the vernier control rod which was marked to show the number of feet and inches which
remained within the pile.

At 10:37 Fermi, without taking his eyes off the instruments, said quietly:

Pull it to 13 feet, George. The counters clicked faster. The graph pen moved up. All the
instruments were studied, and computations were made.

This is not it, said Fermi. The trace will go to this point and level off. He indicated a spot on
the graph. In a few minutes the pen came to the indicated point and did not go above that point.
Seven minutes later Fermi ordered the rod out another foot.

Again the counters stepped up their clicking, the graph pen edged upwards. But the clicking was
irregular. Soon it leveled off, as did the thin line of the pen. The pile was not self-sustainingyet.

At eleven oclock, the rod came out another six inches; the result was the same: an increase in rate,
followed by the leveling off.

Fifteen minutes later, the rod was further withdrawn and at 11:25 it was moved again. Each time
the counters speeded up, the pen climbed a few points. Fermi predicted correctly every movement
of the indicators. He knew the time was near. He wanted to check everything again. The automatic
control rod was reinserted without waiting for its automatic feature to operate. The graph line
took a drop, the counters slowed abruptly.

At 11:35, the automatic safety rod was withdrawn and set. The control rod was adjusted and Zip
was withdrawn. Up went the counters, clicking, clicking, faster and faster. It was the clickety-click
of a fast train over the rails. The graph pen started to climb. Tensely, the little group watched,
and waited, entranced by the climbing needle.

Whrrrump! As if by a thunder clap, the spell was broken. Every man frozethen breathed a sigh
of relief when he realized the automatic rod had slammed home. The safety point at which the rod
operated automatically had been set too low.

Im hungry, said Fermi. Lets go to lunch.

Time Out for Lunch

Perhaps, like a great coach, Fermi knew when his men needed a break.

16
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

It was a strange between halves respite. They got no pep talk. They talked about everything
else but the game. The redoubtable Fermi, who never says much, had even less to say. But he
appeared supremely confident. His team was back on the squash court at 2:00 p.m. Twenty
minutes later, the automatic rod was reset and Weil stood ready at the control rod.

All right, George, called Fermi, and Weil moved the rod to a predetermined point. The spectators
resumed their watching and waiting, watching the counters spin, watching the graph, waiting for
the settling down and computing the rate of rise of reaction from the indicators.

At 2:50 the control rod came out another foot. The counters nearly jammed, the pen headed off
the graph paper. But this was not it. Counting ratios and the graph scale had to be changed.

Move it six inches, said Fermi at 3:20. Again the changebut again the leveling off. Five minutes
later, Fermi called: Pull it out another foot.

Weil withdrew the rod.

This is going to do it, Fermi said to Compton, standing at his side. Now it will become self-
sustaining. The trace will climb and continue to climb. It will not level off.

Fermi computed the rate of rise of the neutron counts over a minute period. He silently, grim-faced,
ran through some calculations on his slide rule.

In about a minute he again computed the rate of rise. If the rate was constant and remained so,
he would know the reaction was self-sustaining. His fingers operated the slide rule with lightning
speed. Characteristically, he turned the rule over and jotted down some figures on its ivory back.

Three minutes later he again computed the rate of rise in neutron count. The group on the balcony
had by now crowded in to get an eye on the instruments, those behind craning their necks to be
sure they would know the very instant history was made. In the background could be heard Wilcox
Overbeck calling out the neutron count over an annunciator system. Leona Marshall (the only
female present), Anderson, and William Sturm were recording the readings from the instruments.
By this time the click of the counters was too fast for the human ear. The clickety-click was now
a steady brrrr. Fermi, unmoved, unruffled, continued his computations.

The Curve is Exponential

I couldnt see the instruments, said Weil. I had to watch Fermi every second, waiting for orders.
He face was motionless. His eyes darted from one dial to another. His expression was so calm it
was hard to read. But suddenly, his whole face broke into a broad smile.

Fermi closed his slide rule

The reaction is self-sustaining, he announced quietly, happily. The curve is exponential.

The group tensely watched for twenty-eight minutes while the worlds first nuclear chain reactor
operated.

17
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

The upward movement of the pen was leaving a straight line. There was no change in indicate a
leveling off. This was it.

O.K., Zip in, called Fermi to Zinn who controlled that rod. The time was 3:53 p.m. Abruptly,
the counters slowed down, the pen slid down across the paper. It was all over.

Man had initiated a self-sustaining nuclear reactionand then stopped it. He had released the
energy of the atoms nucleus and controlled that energy.

Right after Fermi ordered the reaction stopped, the Hungarian-born theoretical physicist Eugene
Wigner presented him with a bottle of Chianti wine. All through the experiment Wigner had kept
this wine hidden behind his back.

Fermi uncorked the wine bottle and sent out for paper cups so all could drink. He poured a little
wine in all the cups, and silently, solemnly, without toasts, the scientists raised the cups to their
lipsthe Canadian Zinn, the Hungarians Szilard and Wigner, the Italian Fermi, the Americans
Compton, Anderson, Hilberry, and a score of others. They drank to successand to the hope they
were the first to succeed.

A small crew was left to straighten up, lock controls, and check all apparatus. As the group filed
from the West Stands, one of the guards asked Zinn:

Whats going on, Doctor, something happen in there?

The guard did not hear the message which Arthur Compton was giving James B. Conant at Harvard,
by long-distance telephone. Their code was not prearranged.

The Italian navigator has landed in the New World, said Compton. How were the natives?
asked Conant. Very friendly.

1.3 World nuclear power reactors as of 2015

The following info was downloaded from http://www.wano.org.uk/.


WANO = World Association of Nuclear Operators.

18
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Figure 1.7: Symbols and abbreviations in the World lists of nuclear reactors.

19
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Figure 1.8: World list of nuclear reactors, part 1.

20
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Figure 1.9: World list of nuclear reactors, part 2 (with some overlap with part 1).

21
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Figure 1.10: Locations of nuclear power plants in the U.S. and Canada. Some locations have
multiple reactors.

Figure 1.11: Locations of nuclear power plants in Europe. Some locations have multiple reactors.

22
NUEN 601, M.L. Adams notes, 2017 Chapter 1. Nuclear Reactors: A Historical Perspective

Figure 1.12: Locations of nuclear power plants in SE Asia. Some locations have multiple reactors.

Figure 1.13: Locations of nuclear power plants in other parts of the world. Some locations have
multiple reactors.

23
NUEN 601, M.L. Adams notes, 2017

Part II

Fission, Chain Reactions, & Nuclear


Reactions

24
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

Chapter 2

Fission

2.1 Introduction

We know that fission of a heavy nucleus releases a relatively large quantity of energy. Oddly
enough, fusion of two light nuclei also releases a relatively large quantity of energy. Fissioning a
medium-mass nucleus or fusing two of them would, on the other hand, consume energy. Why is
this? In this chapter we will explore this question.

Nuclear fission reactors actually produce new fuel as they operate. If they produce more useful
fuel than they destroy, they are called breeder reactors. Otherwise, they are simply converters.
Commercial reactors in operation today are converters. In this chapter we will also briefly discuss
and quantify this.

2.2 Brief Look at Key Fission Physics

2.2.1 Binding energy

The mass of a nucleus is less than the sum of the masses of the neutrons and protons contained in
it. This mass difference is called the mass defect. The mass defect, m, is given by

m = Zmp + N mn Mnucleus (Z, A) , (2.1)

where Z is the number of protons in the nucleus, A is the number of nucleons (protons + neutrons)
in the nucleus, mp is the mass of the proton (mp = 1.007227 u1 ), mn is the mass of the neutron
(mn = 1.008665 u), and Mnucleus (Z, A) is the mass of the nucleus. A is often called the mass
number. When we express m in energy units, we have the nuclear binding energy,

BEnuclear = m c2 = [Zmp + N mn Mnucleus (Z, A)] c2 ,


1
u is the abbreviation for atomic mass unit, defined to be 1.6605402 1027 kg= 931.5016 MeV/c2

25
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

i.e., the energy that would be required to break the nucleus into its constitutive nucleons. This
energy is called the binding energy since it represents the energy that binds the nucleus together.
On the other hand, when Z protons and N neutrons are brought together to form a nucleus, an
energy of the amount m c2 is released. If m c2 is re-supplied, the Z protons and N neutrons
can be separated again.

Electron binding energy. When discussing nuclear (not chemical) reactions we usually neglect
the energy that binds the electrons to the nucleus, because this energy is many orders of magnitude
smaller than the energy with which nucleons are bound to each other. This small electron binding
energy is as follows for a nucleus of mass Mnucleus (Z, A) and its Z electrons of mass me :

BEe (Z, A) = [Zme + Mnucleus (Z, A) Ma (Z, A)] c2 , (2.2)

where Ma (Z, A) is the mass of the neutral atom including its Z bound electrons. The atoms total
binding energy is the sum of the nucleons binding energy and the electrons binding energy, which
can be expressed as:

BEatom (Z, A) = BEnucleus (Z, A) + BEelectrons (Z, A)


= mtotal c2
= [Z (mp + me ) + N mn Ma (Z, A)] c2 . (2.3)

But because BEelectrons (Z, A) << BEnucleus (Z, A) we often make the approximation

BEnucleus (Z, A) BEatom (Z, A) = [Z (mp + me ) + N mn Ma (Z, A)] c2 . (2.4)

This approximate formula is useful because tables of Ma (Z, A) are easily obtainable (charts of
nuclides, reference books, the internet, e.g., http://ie.lbl.gov/toimass.html and within it, the
nice search site http://ie.lbl.gov/toi2003/MassSearch.asp, ...).

The total binding energy is an increasing function of the atomic mass number A, but it does
not increase at a constant rate. A more interesting curve is binding energy per nucleon (i.e.
BEnuclear /A) as a function of A. Let us sketch this in Fig. 2.1.

For low As, there are a few deviations from the curve, which are linked to magic numbers of
nucleons that form unusually stable (tightly bound) nuclei. Examples are 4 He and 12 C.

At A 60, the curve reaches its maximum (Ni, Fe) and then for A & 60, the curve is smoothly
decreasing.

Nuclei with high average binding energy per nucleon are more tightly bound, so a larger amount
of energy must be used to break a nucleon away, whereas nuclei with low average binding energy
per nucleon are less tightly bound and can be broken apart more easily.

26
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

10

50 100 150 200



Figure 2.1: Binding energy per nucleon as a function of mass number. Nuclides with mass numbers
in the 50-70 range are the most tightly bound. 7.6 at A=12 and A=240. 1 at A=2. 8.8 at A=60.
7 at A=4.

Therefore, whenever a more tightly bound nucleus is formed by combining two less tightly bound
lighter nuclei, energy is released in the process. This describes the fusion principle. On the other
end of the curve, whenever a heavy nucleus can be split into two lighter parts, which are more
tightly bound, energy is released in the process. This describes the fission principle.

Since fission of heavy nuclei is favored energetically, we might ask why it does not occur sponta-
neously. First, because of short-range nuclear forces there is a

potential-energy barrier

(of several MeV) that must be overcome before a nucleus is free to fly apart. That is, the nucleus
must climb a several-MeV hill before it can fall off of a many-MeV cliff.

Second, spontaneous fission does occur, because of quantum-mechanical tunneling through the
potential barrier. This is a rare event in most heavy nuclides; for example, the half-life for spon-
taneous fission in 238 U is 6.5 1015 years. However, there are exceptions: Californium-252 (252 Cf)
has a half-life for spontaneous fission of only 66 years. Because of this, 252 Cf is often used as a
neutron source.

27
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

We must design our reactors so that heavy nuclei can routinely overcome the several-MeV fission
barrier. There are at least two ways to overcome the barrier:

1. slam the nucleus with a multi-MeV particle, using its kinetic energy

2. let the nucleus absorb a neutron, thereby releasing binding energy

We use the second approach in our reactors, partly because multi-MeV particles are not available
in sufficient quantity to use the first, and partly because neutrons are availablethey are emitted
from fission, which suggests the potential of a self-sustaining chain reaction.

We can compute the energy released by any particular fission event. For example,
1n + 235 U 95 Zr + 138 T e + 3 1n

BE/A [MeV/nucleon] () 7.5 8.6 8.3


BE [MeV] 1762 817 1146
() estimate the value of BE /A from Fig. 2.1
n

In this example, we have a net gain of 1146 + 817 1762 = 201 MeV per fission. This estimate,
which we obtained just by looking at a plot, is quite close the the precise value. If you want the
precise figures for this reaction, you must look up precise atomic-mass or binding-energy values.

2.3 Fissile, Fissionable, and Fertile Nuclides

If a nuclide is reasonably likely to fission after absorbing a very slow neutron, it is called

fissile.

Fissile nuclei surpass the fission barrier with only the

binding energy of another neutron.

Some fissile nuclei:

U-233, U-235, Pu-239, Pu-241 .

If a nuclide is reasonably likely to fission after absorbing a neutron with kinetic energy of an MeV
or two, but not after absorbing a slow neutron, it is called

fissionable.

28
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission
Fission Cross sections, ENDF-neutron-pointwise
104
U-233
U-235
2
10 U-238
Th-232
Cross Section (barns) Pu-239
Pu-240
100

10-2

10-4

10-6

10-8
10-4 10-3 10-2 10-1 100 101 102 103 104 105 106 107
Energy (eV)

Figure 2.2: Microscopic fission cross sections for several fissile and fissionable nuclides. Note that
the vertical log scale spans many orders of magnitude. Note further that although the 238 U fission
cross section becomes quite small for energies below 1 MeV, there is no threshold below which
it goes to zero. The same is true of 232 Th, although the vertical axis is not shown at low enough
energies to see this.

Fissionable nuclei need the binding energy of another neutron, plus

some more energy,

to make it over the fission barrier.

Some fissionable nuclei:

Th-232, U-234, U-236, U-238, Pu-240, Pu-242 .

Fission in fissionable nuclei is

not a true threshold reaction,

in the following sense: the cross section is not zero below any threshold neutron energy, because
fission (which you will remember is energetically favored!) can occur by quantum-mechanical
tunneling even when there is not enough energy to get over the barrier. Its just less likely. Figure
2.2 shows the fission cross sections for some fissile and fissionable nuclides.

29
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

Fertile nuclides are those that are non-fissile but can produce a fissile nuclide after absorbing a
neutron. Two important examples:
2.3 d 239
238 239 23 min 239
92 U(n, ) 92 U 93 Np 94 Pu ,
232 233 22 min 233 27 d 233
90 Th(n, ) 90 Th 91 Pa 92 U . (2.5)

2.4 Conversion and Breeding


Reactors that contain fertile material convert some of it to fissile material. They are called

converter reactors.
All reactors convert to some extent, and most commercial reactors convert a significant amount of
238 U to 239 Pu, as indicated in the first line of Eq. (2.5). Other actinides (elements with Z > 88,

some of which are fissile) are also produced. These new nuclides may produce a significant portion
of a reactors power, especially at the end of a fuel cycle. It is possible to extract this material from
spent reactor fuel and use it in later fuel cycles.

For any reactor, we define:


fissile atoms produced
conversion ratio = (2.6)
fissile atoms consumed

If this ratio is greater than unity (which is quite possible!), the reactor is called a

breeder

reactor. In this case the conversion ratio is called the

breeding ratio.
If we define fuel to mean fissile atoms, then breeder reactors produce more fuel than they
consume. This cannot go on indefinitely; at some point the worlds supply of fertile material would
run out. However, the worlds supply of fertile material is far, far greater than its supply of fissile
material. Because of this, breeder reactors could extend the lifetime of fission power into the distant
future (many, many centuries).

Breeding in a reactor is possible only if the following inequality is true.

reproduction factor

P# of fuel nuclides h i i i
# of neutrons produced i=1 f
= = P# of fuel nuclides >2. (2.7)
# of neutrons absorbed in fuel i=1 [ia ]

30
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

Note that will depend on

the energies of the neutrons getting absorbed.

In fact, tends to increase substantially at neutron energies above 100 keV (0.1 MeV) for most
fissile materials. For this reason, breeders are usually fast rather than thermal reactors. This
means their neutrons are mostly kept at high energies rather than being slowed down to thermal
(low) energies. However, it is possible to design a thermal breeder reactor using 232 Th as the fertile
fuel and 233 U as the fissile fuel. Figure 2.3 illustrates why this is possible.

Figure 2.3: Reproduction factor, , as a function of the energy of the incident neutron, for 233 U,
235 U, and 239 Pu.

We close this section with more detailed plots of binding energy per nucleon, shown in Fig. 2.4.

31
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

Figure 2.4: Binding energy per nucleon as a function of mass number. Top figure uses the most
stable isobar (most tightly bound nuclide) for each value of A and identifies a number of nuclides
of interest, with straight lines drawn between data points. Bottom figure shows values for more
nuclides, not just the most stable isobars, and identifies several of them, with no lines drawn
between data points.

32
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

2.5 Fission Process and Results

We have seen that the absorption of a neutron puts the resulting compound nucleus in an excited
state, mostly from the binding energy added by absorption of the neutron but also including the
center-of-mass-frame kinetic energy. We have also seen that sometimes this excitation energy is
enough to cause a heavy nucleus to fission. Fission is illustrated in Fig. 2.5.

Figure 2.5: Schematic of a fission event in 235 U.

The following are all part of the fission process. For each event we give the time scale and the
amount of energy that is released to the particles that are emitted.

1. scission (splitting into two or more fragments): 0.01 picoseconds (ps)

2. prompt neutrons emitted: 0.01 1 ps; 5 MeV

3. prompt gammas emitted: 0.01 1 ps; 7 MeV

4. fission fragments slow down: 10 ps; 165 170 MeV to matter

5. fission products and daughters decay:  1 s to  centuries; 25 30 MeV


Note that the 25 30 MeV from fission-product decay includes

12 15 MeV carried away by antineutrinos


7 8 MeV carried by s
6 7 MeV carried by s
small amount carried by delayed neutrons

33
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

One immediate result of fission (10-20 femtoseconds after scission) is

fission fragments (usually two)

that rush away from each other with total kinetic energy of roughly 165 MeV, which is most of the
energy that is ultimately released by the fission process (roughly 200 MeV). The fission fragments
give up their kinetic energy quickly in typical reactor materials:

They come to rest 10 picoseconds (ps) after scission.

They lose energy via

Coulomb interactions with electrons and surrounding nuclei,

thereby converting their kinetic energy to

heating of the material within a small distance from the fission event.

Note: Even if we restrict ourselves to a particular nuclide, such as 235 U, we can talk only about av-
erage values for the energy carried by different kinds of particles. The fission process is stochastic
there are lots of different ways that the compound nucleus can split, and lots of different paths that
the excited fission fragments can follow to de-excite.

Question: Does all of the energy released by fission go into heating the reactor material? That is,
is all of this energy recoverable? Explain.

Answer: No. Some of the particles escape, carrying away their kinetic energy:

The fission-fragment energy is deposited locally.

The betas are deposited locally.

In a large reactor, only a few percent of the neutrons and gammas escape.

The neutrinos escape.

Question: Suppose a reactor has been operating at 100% power for a long time. If we scram
the reactor and thus stop fissions from occurring, have we stopped the heating of the fuel and
structural materials? Explain.

Answer: No. The fission products continue to decay, releasing betas and gammas that deposit
energy in the reactor material.

34
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

Question: What fraction of reactor power comes from the decay of fission products?

Answer: ( 13 15 MeV)/( 200 MeV) = 6-7%

What does this tell you about cooling systems for reactors? Do we still need them after weve
scrammed the reactor? Yes. Decay heat is more than sufficient to melt the reactor.

Question: What other mechanisms (reactions) produce recoverable energy in reactors?

Answer: Radiative capturethe binding energy of the neutron is emitted in s

Neutron scattering also heats the reactor material, because the neutron transfers KE to material.
However, this is not additional energy on top of what is released from fissionit is part of the 5
MeV carried from the fission event by the prompt neutrons.

2.6 Fission-fragment mass distribution

One might guess that the most likely way a compound nucleus would split would be into two equal-
sized halves. It turns out that a 1/3-2/3 split is favored significantly over a half-and-half split, at
least for fission that is initiated mostly by the binding energy of a neutron. (If we cause fission
by bombarding heavy nuclei with sufficiently energetic particles, say hundreds of MeV, then the
half-and-half split becomes most favored.) For fissions in nuclear reactors we have approximately
the distribution of fission products given in Fig. 2.6 below.

The top left figure compares, on a logarithmic scale, the mass distribution of fission fragments from
fission of three different fissile nuclides when fission is caused by slow neutrons. Note that there are
two distinct high-probability humps with a much lower-probability valley in between. This is seen
more clearly on the linear scale in the figure in the top right. Note further that when a higher-mass
nuclide fissions, it is the low-mass hump that moves to the right (compare 233 U against 239 Pu). The
bottom figure shows how the valley gets filled in when fission is caused by higher-energy neutrons,
shown here for fission of 235 U induced by thermal neutrons and by 14-MeV neutrons.

35
NUEN 601, M.L. Adams notes, 2017 Chapter 2. Fission

Figure 2.6: Fission-fragment mass-number distribution. Top left: fission induced by absorption of
low-energy neutrons, for 233 U, 235 U, and 239 Pu. Top right: same thing for 235 U but on a linear
scale. Bottom: .

36
NUEN 601, M.L. Adams notes, 2017 Chapter 3. Chain Reactions

Chapter 3

Chain Reactions

3.1 Introduction

We know that neutron production and loss rates determine the behavior of a nuclear reactor. In
this chapter we introduce some terms that help us describe production rates, loss rates, and how
well they are balanced.

3.2 Multiplication Factor: Definitions

Suppose we could remove all the neutrons from a source-free reactor, then initiate exactly enough
fissions to create N fission neutrons. Let us define these to be a

generation of neutrons.
Some neutrons in any given generation will cause fission. The neutrons resulting from such fissions
are defined to consitute the next generation.

We introduce an often-seen definition of the multiplication factor:

k = multiplication factor

# of neutrons in a generation
= . (3.1)
# of neutrons in the previous generation
Then we see that:

k<1 neutron population is decreasing subcritical


k=1 neutron population is steady critical
k>1 neutron population is increasing supercritical

37
NUEN 601, M.L. Adams notes, 2017 Chapter 3. Chain Reactions

Another commonly cited definition of multiplication factor is:


production rate of neutrons
k =
loss rate of neutrons
P
= . (3.2)
L
Now that we have defined k, we note that

these definitions are wrong

unless the neutrons are in their fundamental mode (fundamental distribution) for the reactor
in question. In the case of a fundamental mode distribution, the two definitions are correct and
equivalent. Later we will discuss this fundamental mode (and other modes) and we will find that
the true definition of k is:

k = largest eigenvalue of [Loss]1 [Prod],

where [Loss] is the neutron loss operator and [Prod] is the neutron production (from fission) oper-
ator. (Note the similarity to P/L in Eq. (3.2).) For now, the main takeaway point is:

k is a property of the system being studied,


not of the neutrons in the system.

3.3 Multiplication Factor: Formulas

We shall develop here a formula for the multiplication factor, k, in a reactor. We do so by considering
the life history of a neutron from birth to death, describing each fork in the road with a mathematical
statement of probability. (We assume that the neutrons are in the fundamental-mode distribution.)
First we define:

38
NUEN 601, M.L. Adams notes, 2017 Chapter 3. Chain Reactions

PF N L = fast non-leakage probability


= probability that a fast neutron does not leak

PT N L = thermal non-leakage probability


= probability that a thermal neutron does not leak

p = resonance-escape probability
= probability that a fast neutron is not absorbed while slowing to thermal,
given that it didnt leak

f = thermal utilization
= probability that thermal neutron is absorbed in fuel, given that it was absorbed

uF = fast utilization
= probability that fast neutron is absorbed in fuel, given that it was absorbed while fast

PT AF = prob. that thermal absorption in fuel causes fission

PF AF = probability that fast absorption in fuel causes fission

T = average number of neutrons emitted per fission caused by a thermal neutron


i.e., per thermal fission

F = average number of neutrons emitted per fission caused by a fast neutron


i.e., per fast fission

What would PT AF be in terms of s? How about f ?

In Fig. 3.1 we sketch the possible neutron life histories in a reactor.

39
NUEN 601, M.L. Adams notes, 2017 Chapter 3. Chain Reactions


Figure 3.1: Possible histories of neutrons during their life cycle in a reactor. Manipulation of the
probabilities of each event produces the well-known six-factor formula. top // bottom; left , right:
fast neutron // PF N L , 1-PF N L // doesnt leak, leaks // p, 1-p // slows to thermal, absorbed while
fast // PT N L , 1-PT N L , 1-uF , uF // absorbed, leaks, abs. in junk, abs. in fuel // f, 1-f, 1-PF AF ,
PF AF // abs. in fuel, abs. in junk, captured, causes fission F // PT AF , 1- PT AF // T
causes fission, captured Comments: E (105 , 107 ) eV, so fast and thermal are
crude distinctions (like tall and short. In SS, prob. = ratio of rates. Calcg rates
is key focus of Rx analysis; depend on n distribn in ~r, E, .

It follows that the multiplication factor must be:

k = PF N L pPT N L f PT AF T + PF N L (1 p)uF PF AF F . (3.3)

40
NUEN 601, M.L. Adams notes, 2017 Chapter 3. Chain Reactions

It is customary to define

T = thermal reproduction factor


= expected # of ns produced per thermal absorption in fuel
= PTAF T
It is not customary, but we shall also define:

F = fast reproduction factor


= expected # of ns produced per fast absorption in fuel
= PFAF F
This gives us:

k = PFNL [ pPTNL f T + (1 p)uF F ] (3.4)

It is also customary to define:

 = fast-fission factor

expected # of ns from all fissions


=
expected # of ns from thermal fissions
PFNL pPTNL f PTAF T + PFNL (1 p)uF PFAF F
=
PFNL pPTNL f PTAF T
1 p uF PFAF F
= 1+
p PTNL f PTAF T
This gives us:

k = PFNL PTNL pf T . (3.5)

This is the well-known

six-factor formula.
In an infinite medium nothing leaks, and we have the four-factor formula:

k = pf T (3.6)
Beware: While p, f , and T depend only on material properties,  depends also on PTNL and thus
is different in an infinite medium than in a finite medium. See example next page!

41
NUEN 601, M.L. Adams notes, 2017 Chapter 3. Chain Reactions

Note that

k is an upper bound for k.

Example
Recall that a neutron in a reactor will ultimately either leak or get absorbed. Suppose that in a
certain reactor, for every 1000 neutrons that are born from fission, the following is how many leak
and how many get absorbed:

20 leak after they become thermal

80 leak while they are still fast

200 are absorbed in fuel while they are fast

400 are absorbed in fuel after they become thermal

100 are absorbed in non-fuel materials while fast

200 are absorbed in non-fuel materials while thermal

Suppose further that of every 400 thermal neutrons that are absorbed in fuel, 320 actually cause
fission, and that on average there are 2.5 neutrons emitted from these thermally-induced fissions.
Likewise, suppose that of every 200 fast neutrons absorbed in fuel, 60 actually cause fission, and
that on average there are 3.0 fast neutrons emitted from these fast fissions.

Question: What are the six factors in the six-factor formula, and what is the multiplication
factor?

Solution. First, forget all the formula stuff and use common sense to get the multiplication
factor. Then you can check later to make sure your factors give the right answer. The common-
sense approach to k goes like this: How many fission neutrons will result from 1000 initial fission
neutrons? Well, we know that there are 320 thermal fissions, with 2.5 neutrons emitted from each,
which gives us (2.5)(320) = 800 new neutrons. Also, there are 60 fast fissions, producing (3)(60)
= 180 more new neutrons. So the multiplication factor is (800+180)/1000 = 0.98.

We now calculate the six factors:

PF N L = (1000-80)/1000 = 920/1000
PT N L = (620-20)/620 = 600/620
p = (920-200-100)/920 = 620/920
f = 400/(400+200) = 400/600
T = 2.5 320/400 = 800/400

Thats five of them. For the fast-fission factor, , we first obtain some intermediate results:

42
NUEN 601, M.L. Adams notes, 2017 Chapter 3. Chain Reactions

1p = (920-620)/920 = 300/920
uF = 200/(200+100) = 200/300
F = 3.0 60/200 = 180/200

Now we can get the fast-fission factor:

PFNL pPTNL f PTAF T + PFNL (1 p)uF PFAF F


 =
PFNL pPTNL f PTAF T
800/1000 + 180/1000 180
= =1+
800/1000 800
980
=
800

It is easy to see that if we multiply the six factors, we obtain 980/1000, as we should. Note
that because the probabilities are conditional, the numerator of one factor typically cancels the
denominator of another!

Tricky example

Suppose the reactor in the previous example were made infinite, so that PFNL = PTNL = 1.
What is the new k, which is k ? To get the answer we might be tempted to blindly apply the
four-factor formula using the factors calculated above. This would produce:
tempting but wrong 980 800 400 620 980 620
k = T f p = = 1.1007 (3.7)
800 400 600 920 920 600

The problem is that changing the leakage changes (reduces, in fact) the value of the fast-fission
factor, , so we cannot use the value we had in the finite reactor. Recall that

PFNL pPTNL f PTAF T + PFNL (1 p)uF PFAF F


 =
PFNL pPTNL f PTAF T

reactor pf PTAF T + (1 p)uF PFAF F (1 p)uF F


=1+
pf PTAF T pf T
(300/920)(200/300)(180/200) 180 600 180 30
= 1+ =1+ =1+
(620/920)(400/600)(800/400) 800 620 800 31

Using this value of the fast-fission factor we find the correct k-infinity:

k = T f p = 1.0942 (3.8)

43
1.2 4.0E-07

3.5E-07
1

3.0E-07

0.8
2.5E-07

0.6 CDF 2.0E-07


Fission Spectrum, E_inc=1MeV

1.5E-07
0.4

1.0E-07

0.2
5.0E-08

0 0.0E+00
0.E+00 2.E+06 4.E+06 6.E+06 8.E+06 1.E+07
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Chapter 4

Nuclear Reactions

4.1 Introduction

In this chapter we describe the kinds of nuclear reactions that play key roles in nuclear reactors
and in the nuclear fuel cycle. We will not exhaustively study all possible nuclear reactions; for
example, we will not study reactions that can occur when extremely energetic particles collide with
one another, as they do in high-energy accelerators designed for studying particle physics. Instead,
we will focus on two kinds of nuclear reactions:

1. Radioactive decay

2. Interactions between one neutron and one nucleus

This subset of the broader category of nuclear reactions is rich and varied, and it gives rise to
a host of important phenomena in nuclear reactors and, more generally, in the nuclear fuel cycle
(which includes the handling and disposition of used fuel and nuclear waste).

In addition to describing nuclear reactions, we will learn how to quantify the rates at which they
take place.

4.2 General Principles

In a general nuclear reaction, we begin with some set of initial particles, each with its own internal
energy state and its own kinetic energy, and we end with some set of final (post-reaction) particles,
each with its own internal energy state and its own kinetic energy.

44
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

We observe that certain quantities are always conserved in the nuclear reactions of interest:

total energy, which includes rest-mass energy,

momentum

charge

number of nucleons
We can use these conservation laws to calculate a lot of things about nuclear reactions that we
need to know to achieve our nuclear-engineering goals. (Homework assignments will give you
opportunities to demonstrate this.)

An important general concept in nuclear reactions is the

Q value of a reaction

which is defined:

Q = (KE of final particles) KE of initial particles

= (rest-mass energy of initial particles) (rest-mass energy of final particles) (4.1)


X X
= c2 m0,i m0,f . (4.2)
i=initial particles f =final particles

Here KE means kinetic energy. You can think of the Q-value as the amount of energy released
by the reaction. It is the amount of energy that is converted from mass, which is also the total
amount of kinetic energy that is gained. Note that Q can be negative. Definitions:

If Q > 0, the reaction is exothermic it releases energy.

If Q < 0, the reaction is endothermic it consumes energy.

4.3 Radioactive Decay

In radioactive decay, a nucleus spontaneously emits one or more particles (for convenience in our dis-
cussions, we generalize our definition of particle to include photons), which emerge with nonzero

45
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

kinetic energy, and thereby converts itself into a different nucleus. Here different could refer to a
different isomeric state of the same nuclide, as when an excited nucleus emits one or more photons
to reach its ground state, or it could refer to a different nuclide, as is the case when there is emission
of any particle other than a photon. A summary of several important types of radioactive decay is
given in Table 4.1.

Table 4.1: Types of Radioactive Decay


Decay type Reaction Description
Excited nucleus decays to lower-energy state
gamma () AP AP + by emitting one or more photons
Z Z
Alpha particle (2 protons + 2 neutrons
AP A4
alpha () Z Z2 D + 42 He bound together, = 4 He nucleus) is emitted
A neutron in the nucleus converts to a proton,
beta ( ) AP
Z A D
Z+1 + + an electron, and an antineutrino. The electron
is ejected and is called a beta particle. The
antineutrino is also ejected. The nucleus
is usually left in an excited state
(which may decay quickly by photon emission).
A proton in the nucleus converts to a neutron,
positron ( + ) AP
Z A D
Z1 + + + a positron, and a neutrino. The
positron and neutrino are ejected.
An orbital electron is captured by the nucleus
electron and combined with a proton to form a neutron
capture (EC) AP
Z + e A D
Z1 + and a neutrino. The neutrino is ejected,
and the nucleus is usually left in an excited state
(which may decay quickly by photon emission).
AP A1 1H
proton (p) Z Z1 D + 1 A proton is ejected from the nucleus.
AP A1 1n
neutron (n) Z Z D+ 0 A neutron is ejected from the nucleus.
internal Excitation energy in nucleus is used to
conversion (IC) AP AP + e eject an orbital (usually K-shell) electron
Z Z

46
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

The current understanding of radioactive decay is that it is a random, no-memory process that is not
influenced by a nucleuss surroundings or by any external forces or fields. Given this understanding,
it follows that any given decay rate can be quantitatively described by a single number called the

decay constant

which is usually given the symbol


The decay constant has units of

inverse time

It has the following interpretation: given an infinitesimally small time interval, dt, the probability
that a particular nuclide will decay within that time interval by a particular decay process is

Probability of decay within dt = dt

where is the decay constant for that particular decay process for that particular nuclide.

Because the decay process is random, we cannot predict exactly when a given atom will decay, nor
can we predict exactly how many atoms in a particular sample will decay within a certain time
interval. We can, however, calculate

expected values

of such times and numbers of decays. For example, if a sample contains N radioactive atoms of a
given type (given nuclide in given internal energy state), then the expected number of decays in
the infinitesimal time increment dt is simply

expected number of decays in dt = N dt . (4.3)

Thus, the expected decay rate in a sample containing N (t) radioactive atoms at time t is:

expected decay rate at time t = N (t) . (4.4)

Note that the decay rate of a sample of material is also the activity of the sample:

decay rate at time t = activity at time t = A(t) . (4.5)

There are several common units for expressing activity:

1 Bq = 1 becquerel = 1 disintegration per second (1 dps) (4.6)


10
1 Ci = 1 curie = 3.7 10 dps (4.7)

The strange-looking definition of the Ci stems from its original definition as the activity of one gram
of radium-226, which was the natural radioactive substance studied by Marie and Pierre Curie.

47
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

4.4 Decay Dynamics

An expression of conservation (change rate = gain rate loss rate) of a given radioactive species
leads to a differential equation that can be solved for the species population as a function of time.
Consider, for example, a sample containing N0 atoms of a given species at time 0. Suppose that
nothing creates new atoms of this species and that the only way to lose a member of the species is
for it to decay, with decay constant . Then we have:

dN
= gain rate loss rate = 0 N (t) = N (t) . (4.8)
dt
The solution is easily found (for example by using an integrating factor):

N (t) = N0 et , N0 N (0) . (4.9)


From this it is not difficult to compute certain interesting characteristic times:

mean (average) lifetime = 1/


ln 2
half-life expected time for half of a population to decay =

Now consider a decay chain, in which the decay of nuclide A produces daughter nuclide B, which
decays to daughter C, etc. Suppose that nuclide A may be produced (for example by some neutron-
nucleus interaction) at some known constant rate. Call this rate R nuclei per unit time. Then our
conservation equations for the populations of nuclides A, B, and C are
dNA
= R A NA (t) (4.10)
dt

dNB
= A NA (t) B NB (t) (4.11)
dt

dNC
= B NB (t) C NC (t) (4.12)
dt
If daughter C is stable, then C = 0 and the equation is still valid. This is a set of coupled ordinary
differential equations (ODEs) with straightforward coupling. Note that we can solve first for NA (t),
then use this to quantify Bs gain term, then solve for NB (t), etc.

4.5 Neutron-Nucleus Interactions: Some Basics


When a neutron interacts with a nucleus, a variety of different reactions are possible. Some of the
more important ones in nuclear reactors are:
Fission:
1 A A+1 A1 A2
0n + ZX Z X Z1 X1 + Z2 X2 + [a few] 10 n + [ 200 MeV] (4.13)

48
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Radiative Capture:
1 A A+1 A+1
0n + ZX Z X Z X + [one or more] (4.14)

Potential Elastic Scattering:


1 A
0n + ZX 10 n + A
ZX (4.15)

Resonance Elastic Scattering:


1 A A+1
0n + ZX Z X 10 n + A
ZX (4.16)

Inelastic Scattering:
1 A A+1
0n + ZX Z X 10 n + A
ZX + [one or more] (4.17)

In radiative capture, a free neutron disappears (captured by a nucleus). Scattering does not change
the free-neutron population, but it does alter the velocity (energy and direction) of the neutron.
Elastic scattering alters the kinetic energy of the nucleus but not its internal energy level. Inelastic
scattering alters both the kinetic energy and internal energy of the nucleus, leaving it excited,
briefly, before it expels its excess energy in the form of one or more gammas.

The above is not an exhaustive list of neutron-nucleus interactions. For example, the following
reactions (and many more) also occur: (n,p), (n,deuteron), (n,alpha), (n,2n), (n,3n), . . . These
occur less frequently than scattering, capture, and fission reactions. However, over long time
periods they are important contributors to the production and depletion of nuclides in the reactor.
Thus, if we want to predict reactor behavior over long time periods (and we do), we must take into
account far more reactions than just fission, capture, and scattering.

We must know not only the kinds of important reactions that are possiblewe must calculate the
rates of each important reaction. For this we need to know interaction probabilities, which are
contained in cross sections. In the following sections we discuss cross sections, their properties, and
how to use them to compute neutron-nucleus reaction rates.

49
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

4.6 A Simple Experiment

Consider a beam of neutrons, all traveling at the same speed in the same direction, perpendicular
to a thin sheet of material. See Fig. 4.1. The material contains only a single nuclide. We assume
the sheet is so thin that all nuclei see the same beam intensityno nucleus is shielded by other
nuclei. We define:

I beam intensity
= neutron density in beam neutron speed = nv
 
neutrons
= rate per unit area at which neutrons reach target (4.18)
cm2 s
 
nuclei
N number density of target nuclei (4.19)
cm3
 2
A target beam interaction area cm (4.20)
x target thickness [cm] (4.21)


Figure 4.1: Mono-energetic mono-directional beam of particles incident on a thin target containing
a single nuclide, with all nuclei at rest (cold target). Write units on the figure: nuclei/cm3 ,
n/(cm2 -s), cm.

50
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

We reason that the rate at which neutrons collide with nuclei must be proportional to IA (the rate
at which neutrons strike) and to N x (the areal density of target nuclei). Putting this statement
into mathematical language, and using the symbol to represent the constant of proportionality,
we get the following

Rate = nvAN x (4.22)

or
cm2
      
colliding particles particles  2  nuclei
= cm [cm] (4.23)
s nucleus cm2 s cm3

4.6.1 Microscopic Cross Sections

The constant of proportionality, , is called the microscopic cross section. It has units of

area per nucleus


If neutrons were classical point particles and nuclei were classical spheres, would be just the
cross-sectional area of the nucleus. (Think about it.) In reality neutron-nuclear interactions obey
quantum-mechanical laws; nevertheless, you can think of as

the effective cross-sectional area

that a nucleus presents to a neutron.


When dealing with microscopic cross sections we use the following area unit:

1 barn 1024 cm2 (4.24)

We have described as a constant of proportionality. However, if we changed anything about our


simple experiment, we might expect our constant to change. This is, in fact, the case:

Microscopic cross sections depend on:


the relative speed between the neutron and nucleus
and on
the nuclide (type of nucleus)

Example Dont go overleave it for them.

Consider a neutron beam impinging on a thin sheet as shown in Fig. 4.2. The beam has an intensity
of nv = 1010 n/(cm2 -s). The neutrons all have the same speed; each has a kinetic energy of 1 MeV.
The beam completely covers the target area. The microscopic cross section for 1-MeV neutrons is
5 barns (per atom), the target area is 3 cm2 , the target thickness is 0.01 cm, and the atom density
is 41022 atoms/cm3 . At what rate do the incident neutrons collide with target atoms?

51
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions


Figure 4.2: Neutron beam on thin sheet.

Solution: By Eq. (4.22) we have

Rate = IAN x
  10   24
 4 1022 atoms cm2
   
5b 10 n  2 10
= 3cm [0.01cm]
atom cm2 s cm3 b
= 6 107 n/s

4.6.2 Microscopic Cross Sections for Different Reactions

In the preceding discussion we were looking at the total rate at which neutrons interact with nuclei,
without distinguishing among different kinds of interactions (elastic scattering, fission, inelastic
scattering, capture, etc.). The that we used above is called the microscopic total cross section
and is denoted t . It is the sum of microscopic cross sections for various specific interactions:

t s + a

s
a
z }| {
e z }| {
z }| {
= e,r + p +in + ( + f + n,2n + n,3n + n, + n,p + ) (4.25)

where

t = microscopic total cross section


s = microscopic scattering cross section
a = microscopic absorption cross section

e = microscopic elastic scattering cross section


e,r = microscopic resonance elastic scattering cross section
p = microscopic potential scattering cross section
in = microscopic inelastic scattering cross section (4.26)

52
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

or c = microscopic radiative capture cross section (gamma(s) emitted)


f = microscopic fission cross section
n,2n = microscopic cross section for 2 neutrons emitted
n,3n = microscopic cross section for 3 neutrons emitted
n, = microscopic cross section for alpha particle emitted
n,p = microscopic cross section for proton emitted

The microscopic cross section for a certain kind of reaction can be thought of as the effective
area of the nucleus for that particular kind of reaction. Also, the ratio

x /t = probability that, given a collision, a reaction of type x occurred. (4.27)

Example
In the previous example, the total microscopic cross section was 5 barns/atom. Suppose the
scattering cross section (s ) was 0.5 barns/atom. What was the absorption rate in the target?

Solution:
We know the collision rate was 6 107 n/s. (Do we really know this?) We also know that

a /t = (t s )/t = (5 0.5)/5 = 0.9

= 90% of collisions were absorptions = 5.4 107 abs/s (4.28)

Question: Did we miss anything?

scattering ...

discuss it ...

4.6.3 Macroscopic Cross Sections

We notice that the microscopic cross section for reaction x, x , is usually multiplied by the
number density, N , of the element in question. [See Eq. (4.22).] We therefore find it convenient to
give the product N x a name. If a material contains only one kind of nuclide, then

x = macroscopic cross section for reaction of type x

N x (4.29)

for each different type, x = t, a, s, , f, e, in, of interaction. Macroscopic cross sections have
units of inverse length:
[nuclei/cm3 ] [cm2 /nucleus] = cm1 .

53
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

4.6.4 Mixtures of Nuclides

In the preceding discussion we considered targets composed purely of a single nuclide. The
generalization to mixtures is straightforward.

Consider a material composed of two nuclides, Y and Z, as shown in Fig. 4.3. Let us pretend for a
moment that the Z nuclei are not present, as shown in Fig. 4.4. We know the reaction rate for
this caseit is the single-nuclide case we already analyzed:

Rate for case Y = nvAYt x , (4.30)


h
|i {z }
n
cm3
[ cm
s ]
[cm2 ][cm1 ][cm]

where
Yt N Y tY . (4.31)

Likewise, if we pretend now that the Y nuclei are not present, we find:
hni
Rate for case Z = nvAZ t x , (4.32)
s
where
Z Z Z
t N t . (4.33)

!
Figure 4.3: Thin target with two nuclides, Y and Z.

Now the important point to realize is that the two nuclides do not interfere with each others
reactions with neutrons. It follows that the total reaction rate is the sum of the Y and Z rates:
hni
Rate for mixture = nvAmixture
t x , (4.34)
s
where
mixture
t Z Y
t + t .

54
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Figure 4.4: Thin target with two nuclides, Y and Z, with only Y shown.

This was the two-nuclide case. It is easy to generalize:

I
X I
X
t it = N i ti , (4.35)
i=1 i=1

where i denotes a nuclide index and I is the number of different kinds of nuclides in the mixture.
Remember: When computing number densities in a mixture, each is number per unit volume of
mixture. Dont talk yourself into assigning part of the volume to each nuclide! Remember the two
pictures above, one with only Z and the other with only Y !

Example Skip, but tell them to make sure they understand


We uniformly mix 50 grams of helium (4 He) and 6 grams of hydrogen (1 H) in a 200-cm3 container.
What is the macroscopic total cross section at a neutron speed of 2200 m/s?
Data:

tHe = 0.8 b
tH = 38 b
atomic mass of He 4 amu = 4 g/mol
atomic mass of H 1 amu = 1 g/mol .

55
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Solution

1. We know that t = He H
t + t .

2. Compute t for each component:

He
t = N He tHe

10 24cm2
    
50g 1mol 0.6022nuclei 0.8b
=
200cm3 4g mol nucleus b
0.03cm1
H H H
t = N t

10 24cm2
    
6g 1mol 0.6022nuclei 38b
=
200cm3 1g mol nucleus b
0.69cm1

3. Thus t 0.72cm1 .

4.7 Reaction Rate Densities

Let us now go into detail about computing reaction rates. Recall our thin-target experiment, and
suppose that all the nuclei are at rest in the lab frame (which means a very cold target). Let

~v lab frame velocity of each neutron in the beam

~ lab frame velocity of a given nucleus in the target


V

= ~0 in this cold-target case

Then we have

vr relative speed between neutron and nucleus

~ | = |~v | = v = lab-frame neutron speed in this cold-target case .


= |~v V (4.36)

and the reaction-rate density (reactions per unit volume per unit time) is

RRD = nvN t,cold (v) in this cold-target case . (4.37)

Suppose we used our simple cold-target experiment to measure the cross section for a range of
neutron energies, with the result shown in Fig. 4.5. (We have chosen to plot t,cold as a function
of neutron kinetic energy, not speed. This is okay because there is a one-to-one relation between
neutron speed and neutron kinetic energy in any given reference frame.)

56
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Figure 4.5: Cold cross section, t,cold , as a function of neutron energy in the lab reference frame
(which is the same as target nucleuss frame). Vertical axis label t,cold (E); horizontal E = Elab .

4.7.1 Vibrating Nuclei

Now consider the same material with vibrating nuclei. This is the real-life case, of course. Now
there is a range of relative speeds seen by the incident neutrons, as indicated in Fig. 4.6.

neutron v nucleus V vrel = |v V|

Figure 4.6: Schematic of relative speed between incident neutrons and three nuclei selected from a
vibrating population.

To compute reaction-rate densities in this case we have to know something about

the distribution of nucleus velocities.

~ ):
Let us define this distribution, N(V
~ )dVx dVy dVz = nuclei/cm3 whose velocities are
N(V

~ .
in the velocity cube dVx dVy dVz at V (4.38)

57
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

(See Appendix B.) Now if we consider only the nuclei whose velocities are in the velocity cube
dVx dVy dVz , we see that the relative speed between the neutrons and those particular nuclei is:

~ .
vrel = ~v V

This relative speed matters in two ways:

1. It affects the cross section: the cross section for this set of nuclides is the same as the cold
cross section evaluated at a speed of vrel , not v.

2. It affects the effective beam intensity, because the rate at which neutrons impinge on this set
of target nuclei is determined by vrel , not v. This means the effective intensity is
vrel
Ieff = nvrel = I .
v

It follows that the reaction-rate density between the neutrons and the nuclei with this particular
range of velocities is

RRD with only the


n o  
nuclei in veloc. cube = n ~v V
~ )dVx dVy dVz t,cold ~v V
~ N(V ~ . (4.39)

The total reaction-rate density is the sum (integral) over all of the velocity cubes:
 
RRD = n ~ )t,cold ~v V
N(V ~ ~v V
~ dVx dVy dVz . (4.40)
nucleus velocities
show units on this line show units on this line

If we recognize that the neutron density in the beam (n) equals the beam intensity divided by the
neutron labframe speed (I/v), and we multiply and divide by the integral of N over all Vx Vy Vz ,
we can rewrite this as follows:
 
~ )t,cold ~v V
N(V ~ ~v V
~ dVx dVy dVz
~ )dVx dVy dVz .
nucl.vels.


RRD = I

N(V
v N(V ~ )dVx dVy dVz
nucl.vels. nucl.vels.

(4.41)


~ )dVx dVy dVz = N , the ordinary number density.
N(V (4.42)
nucl.vels.

58
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

The ratio in brackets defines a kind of averaged cross section, averaged over all nuclei using the
relative neutron/nucleus speed for each nucleus. More precisely, it is an averaged [relative speed
cross section] divided by lab-frame speed, with the N being the weight function used in the
averaging:
 
~ ) ~v V
N(V ~ t,cold ~v V
~ dVx dVy dVz
t (~v , N) nucl.vels. . (4.43)
v N(V~ )dVx dVy dVz
nucl.vels.

Make sure you understand: if you remove the v from the denominator, the remaining ratio is a
weighted average of the quantity vrel t,cold (vrel ), with weight function equal to N.


~ , is sometimes greater than and sometimes less
Notice something important: vrel , which is ~v V

than the neutron speed, v. This means the averaging process includes values of the cold cross
section from all speeds in the neighborhood of the neutrons lab-frame speed, v. To say it another
way, if we want a simple formula to give the correct reaction rate, the cross section we must use is

an average of the cold cross section

over a range of neutron speeds.

If the material is so cold and/or the neutron speed is so high that t,cold (|~v V ~ |) t,cold (v), then
nucleus velocities are unimportant and the averaged cross section becomes the cold cross section:

t,cold (|~v |) v ~ )dVx dVy dVz
N(V
cold limit nucl.vels.
t (~v , N) = t,cold (v) . (4.44)
v N(V ~ )dVx dVy dVz
nucl.vels.

But in general the correct averaged cross section depends on the distribution of nucleus velocities.

4.7.2 Isotropic Nucleus Motion

If the nuclei are vibrating isotropicallythat is, if there is no preferred direction along which the
nuclei are wigglingthen our averaged cross section cannot depend on the neutrons direction.
That is, in this case

the material looks the same from any direction,

and thus the reaction rate must be the same for a neutron entering the material from any direction.
So in this special case of isotropic nucleus motion, the average cross section depends on the
neutrons speed, not direction, and we can write it as t (v, N) instead of t (~v , N):
~)
isotropic N(V
t (~v , N) t (|~v | , N) t (v, N) . (4.45)

59
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

4.7.3 Maxwellian Nucleus Motion


How do we characterize the distribution of nucleus velocities? In many cases of practical interest,
including almost all cases of interest in nuclear-reactor analysis, the nucleus velocity distribution
is well approximated by a

Maxwellian

distribution:
2
~
M V
 3/2
 
~ Maxwellian M
N V N exp
2kT 2kT

3/2 " #
M Vx2 + Vy2 + Vz2

M
=N exp , (4.46)
2kT 2kT

where M is the mass of one molecule of the material, N is the molecular number density, k is
Boltzmanns constant, and T is the material temperature. We can see from this expression that if
N is a Maxwellian (often called a Maxwell-Boltzmann) distribution, then:

N does not depend on the direction of nucleus motion (i.e., it is an isotropic distribution),
which you can see because it depends only on the magnitude |V ~ |, and

N is completely determined by the materials temperature.

Remark: The Maxwellian distribution is just the product of Gaussian or Normal distributions
in each of the three velocity components. Each Gaussian has mean = zero and standard deviation
= kT /M .

To summarize: in many cases of practical interest, the nuclei vibrate isotropically, and their
distribution is well approximated by a Maxwellian, which depends only on temperature. In
such cases the microscopic cross section of Eq. (4.44) depends only on the neutron speed (not
direction) and the material temperature:
~)
Maxwellian N(V
t (~v , N) t (v, T ) , (4.47)

where T is the material temperature.

60
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

It follows that for our thin-target experiment in which the nuclei vibrate in a Maxwellian distribu-
tion, we recover our original thin-target formula as long as we use a microscopic cross section that
has been appropriately averaged over nucleus velocities:
~
RRD = I N(V ~ |) |~v V | dVx dVy dVz
~ )t,cold (|~v V
v
nucl.vels.

~ )t,cold (|~v V
~ |)|~v V
~ |dVx dVy dVz

N(V
= I nucl.vels.


~ )dVx dVy dVz
N(V
v ~ )dVx dVy dVz
N(V
nucl.vels. nucl.vels.

= I t (~v , N) N

~)
If Maxwellian N(V
I t (v, T ) N . (4.48)

This generalizes to a thin target composed of a mixture of nuclides:

~ i )} I
Maxwellian {Ni (V X
RRD I ti (v, T ) N i
i=1

= It (v, T ) = It (E, T ) , (4.49)

where t is the macroscopic cross section for the mixture. We have recognized that there is a
one-to-one relation between the neutrons lab-frame speed (v) and its lab-frame kinetic energy (E),
which means we can use either v or E to represent the dependence of the averaged cross section on
the neutrons lab-frame speed. It is common practice to use E.

It is also common practice to drop the T argument in the cross section. You are expected
to know from context whether or not the cross sections being used depend on the
material temperature!

Consider the averaged cross section for a given nuclide and recall that it is an average of the cold
cross section for that nuclide over a range of energies around the neutrons lab-frame energy. The
range over which the averaging takes place increases with increasing temperature, because higher
temperature produces a wider range of vibrational velocities of the nuclei. We illustrate this in Fig.
4.7, which illustrates how the averaging process smooths out the cross section more and more as
material temperature increases.

61
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions


Figure 4.7: Illustration of averaged cross section for three temperatures. (E, T ) vs. E for Tcold ,
T1 > Tcold , and T2 > T1 .

The most important consequence of this is

Doppler broadening

of cross-section resonances. This effect, pictured in Fig. 4.7, leads to increased parasitic capture
of neutrons as fuel temperature increases, at least in commercial reactors. [This is not obvious at
this point in our discussion, but we will discuss it later in more detail.] This important

negative-feedback

mechanism adds considerably to the safety of large power reactors.

Even away from resonances, the cross sections we use are

temperature-dependent,

for the same reason. The change in the averaged cross section with changing temperature is just
larger near a resonance.

4.7.4 Distribution of neutron velocities (not a beam)

Recall our expression for RRD given a beam neutrons of intensity I and lab-frame kinetic energy
E, incident perpendicularly on a very thin target whose nuclei are vibrating in a Maxwellian
distribution at temperature T :
RRD = I t (E, T ) . (4.50)

62
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

(Remember: for this equation to be correct, t must have been appropriately averaged over all
nucleus velocities.) If we recall the physical interpretation of t , then this equation implies a
certain physical interpretation for the beam intensity:
  
expected reactions neutron path length expected reactions
= . (4.51)
cm3 -s cm3 -s neutron path length

Does this interpretation make sense? That is, is beam intensity really the total distance traveled
by neutrons, per unit volume per unit time?

An easy way to test this is to first figure out the density of neutrons in the beam. Consider a
mono-velocity beam whose leading edge is just reaching a perpendicular surface of area A at time
t. At time t + t, the leading edge of the beam has gone a distance vt past the surface, where v
is the speed of each neutron in the beam. The number of neutrons that have crossed the surface
is IAt. All of these neutrons are in the volume Avt. So the neutron density in the beam is
number divided by volume:
number that cross plane in t IAt I
n = neutron density in beam = = = . (4.52)
swept-out volume in t Avt v
It is easy to compute path-length rate from the density. Consider some volume, V , in which the
density of neutrons is n and all neutrons are moving with speed v (possibly in different directions).
What is the total path length traveled by neutrons in the volume in time dt?

total path length = [path length per neutron] [number of neutrons]


= [v dt] [n V ] . (4.53)

To get path-length rate we divide by dt. To get path-length rate density, we also divide by the
volume, V . The result:
neutron path length total path length v dt n V
3
= = = nv . (4.54)
cm -s V dt V dt

This is an important, general result! We restate it:

path-length rate per unit volume = neutron density neutron speed. (4.55)

We give this quantity a name and its own symbol:

path-length rate per unit volume = nv = = scalar flux . (4.56)

We will constantly use this quantity, the scalar flux, to calculate reaction rates of interest in nuclear
reactors. Important: Eq. (4.56) is correct even if neutrons are all moving in different directions.

63
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

It follows from this discussion that:

If neutrons are moving in a mono-velocity beam, then = I = beam intensity.

What if neutrons have a distribution of velocities in the lab frame? What is the correct expression
for reaction-rate densities? To answer this, we first define a velocity-dependent neutron distribution
function, n, analogous to the nucleus distribution function, N, that we used previously:

n(~v )dvx dvy dvz = neutrons/cm3 whose velocities are

in the velocity cube dvx dvy dvz at ~v . (4.57)

Given only the neutrons with velocities in this range, what would be the RRD with nuclei whose
velocities are distributed according to N? If we apply what we have developed so far we find the
answer:

RRD with neutrons


 
in veloc. cube = n(~v )dvx dvy dvz ~ ) ~v V
N(V ~ dVx dVy dVz .
~ t,cold ~v V (4.58)
nucl. vels.
show units on this line show units on this line

The RRD is the sum (integral) over all neutron velocities:

 
RRD = dvx dvy dvz n(~v ) ~ ) ~v V
dVx dVy dVz N(V ~ t,cold ~v V
~ . (4.59)
neutron vels. nucl. vels.

This is the general expression for the neutron-nucleus interaction rate density (total
collision rate density). A similar expression holds for each particular kind of neutron-nucleus
interaction rate densityabsorption, capture, fission, elastic scattering, etc. This six-dimensional
integral simplifies quite a bit, though, if the nuclei are wiggling isotropically and we compute an
averaged cross section as detailed previously, as we now explore.

If the nuclei are in a Maxwellian distribution, then as before we can generate an averaged cross
section and use it to replace the second triple integral and thereby simplify the RRD expression:

RRD = dvx dvy dvz n(~v ) [N v t (v, T )] . (4.60)
neutron vels.

64
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

The term in brackets depends only on the neutron speed, not direction (a consequence of our
assumption that the nuclei are moving isotropically). Because of this, it is convenient to express
the neutrons velocity distribution in terms of neutron speed and direction. (See App. B for how
to change from velocity variables to speed and direction variables.) Then the three integrals over
vx , vy , and vz are replaced by an integral over speed (v) and two integrals over direction (polar
and azimuthal anglessee App. A). The direction integration can be carried out, leaving only an
integral over neutron speed:

RRD = dv nspeed (v)v t (v, T ) . (4.61)
0

Here the speed subscript specifies that nspeed is a neutron density function that is a density in
regular xyz space and also in speed. We recognize that speed times density is scalar flux:

RRD = dv speed (v) t (v, T ) . (4.62)
0

Usually we use neutron energy rather than speed:



RRD = dE (E) t (E, T ) . (4.63)
0

It should be no surprise that there is a similar expression for each different kind of reaction:

fission rate density = dE (E) f (E, T ) ,
0
absorption rate density = dE (E) a (E, T ) ,
0
scattering rate density = dE (E) s (E, T ) , (4.64)
0

etc.

Remember: The equations above are correct only if nuclei are vibrating in approximately a
Maxwellian distribution and each microscopic cross section has been appropriately averaged over
this distribution of nucleus velocities.

4.8 Temperature Dependence

To this point we have barely mentioned an important topic: the temperature dependence of cross
sections. If there were no such dependence, reactors would behave much differently than they do.
In fact, they would be

less safe!

65
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

This is because temperature dependence of cross sections provides an important

negative feedback

mechanism in most of the worlds reactors.


We stated previously that microscopic cross sections depend on the

relative speed between the neutron and nucleus.

Equivalently, we can say that microscopic cross sections depend on the total kinetic energy (neutron
plus nucleus) in the center-of-mass frame.

Let us tackle some simple test problems to make sure we understand this. Suppose that some
nuclide had the microscopic cross section shown Figure 4.8, as a function of the relative speed (vr )
between the neutron and nucleus. This is the same thing as the cold cross section as a function
of neutron lab-frame speed (v), because in the cold-target limit vr = v.

Figure 4.8: Microscopic cross section for fictitious example nuclide. COLD cross section

Re-consider the thin-target experiment depicted in Fig. 4.1. Recall that we have a beam of intensity
I = nv [n/(cm2 -s)] of mono-velocity neutrons perpendicularly incident upon a thin target of a single
isotope whose number density is N [nuclei/cm3 ].

Suppose the beam intensity is 108 n/cm2 -s, the target number density is 2 1022 nuclei/cm3 , the
target thickness is 0.01 cm, and the target area is 3 cm2 . Suppose further that the neutrons have
a speed of 9200 m/s in the laboratory reference frame. Note that since the beam intensity is speed
times density, the neutron density in the beam is 108 n/cm2 -s / 9200 m/s = 108.7 n/cm3 .

66
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Example 1: Cold target


In the cold-target limit, the motion of the nuclei is negligible, which means that the relative
speed between each neutron-nucleus pair is simply the lab-frame speed of the neutron. In this case,
what is the rate [collisions/s] at which neutrons collide with nuclei?

Answer 1: From the Fig. 4.8, the microscopic cross section, t , that we must use is the one that
is valid for vr = 9200 m/s. Thus:

rate = nvAxN tcold (vr )vr =9200m/s


1024 cm2
 
 n  nuclei
= 108 2 (3cm2 )(0.01cm) 2 1022 (250b)
cm .s cm3 b
n
= . . . = 1.5 107 .
s

Example 2: Simple nucleus motion


Now suppose the nuclei are vibrating, which is more realistic. To keep things simple we will pretend
in this example that the nuclei are moving as follows at any given time:

1/5 of them are moving toward the neutron beam with speed 1000 m/s ;

1/5 are moving away from the neutron beam with speed 1000 m/s;

The remaining 3/5 have negligible motion, such that the relative speed between them and
any incident neutron is approximately the neutrons lab-frame speed.

In this case, what is the rate [collisions/s] at which neutrons collide with nuclei?

67
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Answer 2: From the Fig. 4.8, the microscopic cross sections, t , that we must use are:

tcold (vr )vr =9200m/s = 250 b for 3/5 of the nuclei,

tcold (vr )vr =8200m/s = 30 b for 1/5 of the nuclei,

tcold (vr )vr =10,200m/s = 40 b for 1/5 of the nuclei,

Now we have
# of nuclide
Xvelocities
rate = ~j AxNj cold (vr )
n ~v V

~j |
t vr =|~v V
j=1

# of nuclide
Xvelocities ~j
~v V

N j
= nvAxN tcold (vr )vr =|~vV~j |
v N
j=1
 
 n
8

2 2 nuclei
= 10 (3cm )(0.01cm) 2 10
cm2 .s b.cm
 
9200 3 8200 1 10200 1
(250 b) + (30 b) + (40 b)
9200 5 9200 5 9200 5
| {z }
=164.2 b

n
= . . . = 9.84 106 .
s
In this case the target nuclei on average present a cross section of

164.2 b

to the incident neutrons.

Example 3: Simple nucleus motion again


Repeat the previous example but with neutron lab-frame speed = 8200 m/s instead of 9200 m/s.
We adjust the density of neutrons in the beam to keep the same intensity. In this case, what is the
rate [collisions/s] at which neutrons collide with nuclei?

68
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Answer 3: From the Fig. 4.8, the microscopic cross sections, t , that we must use are:

tcold (vr )vr =8200m/s = 30 b for 3/5 of the nuclei,

tcold (vr )vr =7200m/s = 40 b for 1/5 of the nuclei,

tcold (vr )vr =9200m/s = 250 b for 1/5 of the nuclei,

Now we have
# of nuclide
Xvelocities
rate = ~j AxNj cold (vr )
n ~v V

~j |
t vr =|~v V
j=1

# of nuclide
Xvelocities ~j
~v V

Nj
= nvAxN tcold (vr )vr =|~vV~j |
v N
j=1
 

8n 
2 2 nuclei
= 10 (3cm )(0.01cm) 2 10
cm2 .s b.cm
 
8200 3 7200 1 9200 1
(30b) + (40b) + (250b)
8200 5 8200 5 8200 5
| {z }
=81.1 b

n
= . . . = 4.56 106 .
s
In this case the target nuclei on average present a cross section of

81.1 b

to the incident neutrons.

Lessons Learned

Except at absolute zero T , there is a distribution of relative speeds between neutrons and
nuclei, even in the simple case of all neutrons moving at the same lab-frame velocity.

Even when there is a distribution of nucleus velocities, we can still use the concept of a
microscopic cross section that the nuclei present on average to a neutron of a given velocity.
(Note that velocity includes speed and direction.)

This averaged (or effective) microscopic cross section is obtained by averaging the actual
(cold) cross section over a range of relative speeds. (Technically it is an average of the
product vr (vr ) divided by lab-frame speed v.)

69
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

4.9 Reaction Rate Densities: Special Cases

We previously developed an expression for reaction-rate density (RRD) that is essentially exact for
our purposes. For reactions of type r, where r could be total, fission, scattering, absorption, etc.,
we have

 
RRDr = dvx dvy dvz n(~v ) ~ ) ~v V
dVx dVy dVz N(V ~ r,cold ~v V
~ . (4.65)
neutron vels. nucl. vels.

This expression is cumbersome, and in most cases of interest we can use simplified versions, which
we derive here for various situations that often arise in practice.

4.9.1 Neutron speed >> Nucleus speed

~ | is so close to v that (|~v V


If vr = |~v V ~ |) (|~v |), then the nucleus velocities do not matter,
and we obtain:
 
RRDr = dvx dvy dvz n(~v ) dVx dVy dVz N(V ~ ) ~v V
~ r,cold ~v V
~
neutron vels. nucl. vels.

~ |)r,cold (v)

vV and r,cold (|~v V
dvx dvy dvz n(~v ) ~ )vr,cold (v)
dVx dVy dVz N(V
neutron vels. nucl. vels.


= dvx dvy dvz n(~v )vr,cold (v)
neutron vels.


= dv dn(v, )vr,cold (v) (4.66)
0 4


= dv d(v, )r,cold (v) (4.67)
0 4


= dv(v)r,cold (v) (4.68)
0


= dE(E)r,cold (E) (4.69)
0

70
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Here we have employed several variable changes to facilitate the collapse of the six original integrals
down to a single integral. We have also defined the angular flux, , which is the neutron speed
times the direction-dependent neutron density, and we have recognized that the scalar flux is the
direction integral of . We will revisit these concepts in more detail later.

It is important to recognize that it is possible to meet the condition v  V without meeting


~ |) r,cold (v). This can happen near resonances, for example,
the second condition, r,cold (|~v V
because can change dramatically with only a small change in its argument.

4.9.2 1/v Cross Sections

In many cases of practical interest, for low neutron energies we find that many cross sections are
proportional to the inverse of the relative speed between the neutron and nucleus. That is, we often
encounter cross sections of the form:
constant x,cold (v0 )v0
x,cold (vrel ) = = (4.70)
vrel vrel

If this is the case, then our six-dimensional integral for reaction-rate density simplifies tremendously:

 
RRDr = dvx dvy dvz n(~v ) ~ ) ~v V
dVx dVy dVz N(V ~
~ r,cold ~v V
neutron vels. nucl. vels.


~ r,cold (v0 )v0
1/vrel

dvx dvy dvz n(~v ) ~ ) ~v V
dVx dVy dVz N(V
~v V
~
neutron vels. nucl. vels.

= r,cold (v0 )v0 dvx dvy dvz n(~v ) ~ )


dVx dVy dVz N(V
neutron vels. nucl. vels.

= r,cold (v0 )v0 ntot Ntot

= r,cold (v0 ) (v0 ntot ) (4.71)

71
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

This is an important and quite useful result.

1/v CROSS SECTIONS

If a microscopic cross section as a function of relative speed (vrel ) is proportional to 1/vrel for
some vr range, then the corresponding reaction rate density in that range is

independent of the velocity distribution of the nuclei,

independent of the velocity distribution of the neutrons, and

given by a simple expression such as the following, which uses absorption as an example:
#of X
nuclides
i
RRDr = ntot v0 Ntot ai (v0 ) (4.72)
i=1

Remark: often the term 0 ntot v0 is defined and called the 2200-m/s flux. It is simply an
algebraic convenience. It does not have the physical interpretation of an actual rate per unit
volume at which neutrons are making tracks.

4.9.3 Isotropic Nucleus Motion (the usual case)

If the nuclei are vibrating isotropicallythe same in all directionsthen the reaction rate density
will be the same for a neutron going in any direction. That is, the reaction probability will depend
only on the neutrons lab-frame speed, not its direction. The math works out as follows. We recall
the definition of our microscopic cross section averaged over nucleus velocities:
 
N(V ~ ~v V
~ )r,cold ~v V ~ dVx dVy dVz
1 nucl.vels.
t (~v , N) . (4.73)
v N(V ~ )dVx dVy dVz
nucl.vels.

We define the neutron and nucleus direction vectors:

n ~v /v (4.74)
nuc V~ /V (4.75)

72
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

We observe that
h i1/2
~ ~ ) (~v V
~v V = (~v V
~)

h i1/2
~ V
= ~v ~v + V ~ 2~v V
~

h i1/2
= v 2 + V 2 2vV n nuc (4.76)

We change variables to polar coordinates in velocity space, so the densities are in terms of speeds
and directions. If the nucleus motion is isotropic, then the nucleus speed- and direction-dependent
density can depend only on nucleus speed. We write the density function as N (V )/4, where N
has units of nuclei/(cm3 -cm/s):

2 1  
1
dV N (V ) 4 dnuc d

~v V~


~v ~
V

1 0 0 1 nuc r,cold
t (~v , N) = 2 1 (4.77)
v 1
dV N (V ) 4
0 0 dnuc 1 dnuc

(Here nuc cos(nuc ).) We are free to choose any coordinate axes to define the angles nuc
and nuc . We choose the polar axis to be along the neutron flight direction, n . This means
n nuc = nuc , because nuc is the angle between the two directions = angle between north
pole and nuc . Then:
2 1
1
~

~

dnuc dnuc ~v V r,cold ~v V

4 0 1
2 1
1 1/2  1/2 
dnuc v 2 + V 2 2vV nuc r,cold v 2 + V 2 2vV nuc

= dnuc
4 0 1

1 1 1/2  1/2 
dnuc v + V 2 2vV nuc
 2
= r,cold v 2 + V 2 2vV nuc (4.78)
2 1

This integral depends on the neutron speed, v, but it is the same no matter what the neutron
direction is. That is, it is independent of the neutron direction, n , as we knew from physical
considerations that it must be. We now have the math result that agrees with our physics-based
conclusion.

73
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

We conclude:

GIVEN ISOTROPIC NUCLEUS MOTION,


the cross section averaged over nucleus velocities does not depend on neutron direction:

isotropic nucleus motion


 
r (~v , N) r v, N

1 1
p p 
dV N (V ) d v 2 + V 2 2vV v 2 + V 2 2vV
1 0 2 1 r,cold
= (4.79)
v 0 dV N (V )

Further, recall that if the nuclei are in a known distribution, the distribution may be characterized
by a small set of parameters. For example, given the commonly found Maxwellian distribution,
then N (V ) is completely characterized by a single number: the material temperature. In this case
the averaged cross section for a given nuclide depends only on the neutrons lab-frame speed and
the material temperature.

Given the averaged cross section defined in Eq. (4.79), the RRD for reactions of type r becomes:
#nuclides
nuclei Maxwellian
X
RRDr dv d v n(v, ) Ni ri (v, T )
0 4 | {z }
i=1
| {z }
(v,) ir (v,T )
#nuclides
X
= dv d (v, ) ir (v, T )
0
| 4 {z }| i=1
{z }
(v)
r (v,T )


= dv (v)r (v, T )
0
= dE (E)r (E, T ) (4.80)
0

74
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

For completeness we write down the Maxwellian-averaged ri , explicitly showing that it depends
only on the neutron speed and material temperature:

GIVEN MAXWELLIAN NUCLEUS VELOCITY DISTRIBUTION,


the cross section averaged over nucleus velocities depends only on neutron speed and material
temperaturethe only variables that arent integrated out in the following:

Maxwellian nucleus motion


r (~v , N) r (v, T )
2e M V2
1 1
p p 
dV V 2kT d v 2 + V 2 2vV v 2 + V 2 2vV
1 0 2 1 r,cold
= MV 2
(4.81)
v dV V 2 e 2kT
0

Remember: if you use a simple one-dimensional integral to compute a reaction-rate density, then
you must use cross sections that have been properly averaged over nucleus motion. Otherwise the
RRD requires a six-dimensional integration.

4.10 Why Resonances?

Resonances occur in cross sections because nuclei can be very hungry for neutrons of certain energies,
and at the same time can find neutrons with other energies quite distasteful. This is understandable
in terms of the energy levels of a nucleus. Consider Figure 4.9, which schematically illustrates
resonances in the 238 U cross section.

Question: Why do cross sections have sharp peaks (resonances)?

Answer: If the binding energy of the neutron + the center-of-mass-frame kinetic energy would
put the compound nucleus in a preferred energy level, then formation of the compound nucleus
is quantum-mechanically favored and thus much more likely.

Thus, given a target nucleus at rest, the cross section for forming the compound nucleus can depend
strongly on neutron kinetic energy. When the energy is just right, the cross section will be high.
Otherwise, it will be much lower. The high part is called a

resonance

75
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Figure 4.9: Energy-level schematic explaining cross-section resonances: 238 U example. M238 c2
ground (bottom); M239 c2 ground (dash); (M238 + mn )c2 (offset line); arrows = mc2 , BE of n,
KECM238 , KE n ; top left: U-239 energy levels; horiz axis = (E
CM t CM )

The math:

A+1
If BEn + total KECM = energy level of Z X resonance in of A
ZX . (4.82)

where

BEn binding energy of the absorbed neutron


total KECM kinetic energy of neutron + kinetic energy of A
ZX
in center-of-mass frame

76
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

The cross section for any reaction involving a compound nucleus will exhibit resonances. The only
reaction that does not involve a compound nucleus is

potential scattering,

in which the neutron scatters off of the potential of the nucleus without ever penetrating it.

Under reasonable assumptions, it is possible to derive a formula for x around a resonance. See
any nuclear-reactor physics text for the Breit-Wigner formula, for example.

You are encouraged to peruse the cross-section plots in Appendix C, noting how many resonances
you can see, how narrow they are, and how tall they are.

4.11 Differential and Double-Differential Scattering Cross Sec-


tions

Often we care not only about the rate at which a certain reaction occurs, but also about the results
of the reaction. For example, suppose we know the rate at which 1-MeV neutrons scatter off of
hydrogen in our reactor. What happens to the scattered neutrons? At what energy or energies do
they emerge, and in what directions? We cannot get very far in reactor analysis without knowing
the answers. The necessary information is contained in differential scattering cross sections, which
combine an ordinary scattering cross section with a probability distribution function in outgoing
energy, direction, or both.

4.11.1 Differential scattering cross section in energy

For reactor analysis we require not only the cross section for scattering, but also the

distribution of scattered-neutron energies

that results from the scattering events. Let us define this distribution:

P (Ei Ef ) dEf = probability that a neutron scattering with initial energy Ei

emerges with energy in dEf about Ef .

We could measure P (Ei Ef ) by returning to our thin-target experiment and placing energy-
sensitive detectors all around the target. If we did this for neutrons of a certain energy Ei , we
might find something like the plot shown in Fig. 4.10.

The function P (Ei Ef ) is a probability density function. The area under the curve is 1, and
P has units of inverse energy (MeV1 , for example).

77
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions


Figure 4.10: Illustration of P (Ei Ef ). horiz azis Ef ; show Ef = Ei and Ef = Ef,min .

Exercise:
Assume that the function P (Ei Ef ) is known. If neutrons of energy Ei scatter, what fraction
of them will emerge in the interval E1 < Ef < E2 ?

Solution:
From the definition of P (Ei Ef ) it follows that
E2
fraction = dEf P (Ei Ef ) (4.83)
E1

Definition The scattering rate density for neutrons of energy Ei involves the scattering cross
section at Ei , s (Ei ). Thus, to get the rate at which neutrons scatter from energy Ei to an Ef
within some specified interval, we will need to multiply s (Ei ) by P (Ei Ef ). This happens so
often that we give the product its own name and symbol. Note that

s (Ei )P (Ei Ef ) dEf = microscopic cross section for scattering


from energy Ei into an interval dEf at Ef

Draw vertical E axis in margin, with Ei , Ef , and dEf .

We define:
s (Ei Ef ) s (Ei )P (Ei Ef ) (4.84)

We call this the microscopic differential scattering cross-section in energy. Its units are

area per nucleus per unit energy,

commonly expressed in barns/(MeV-nucleus). Naturally, we also define a macroscopic differential


cross section for nuclide j:

js (Ei Ef ) = N j sj (Ei Ef ) (4.85)

78
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

and for a mixture:


I
X
s (Ei Ef ) = N j sj (Ei Ef ) . (4.86)
j=1

The units of the macroscopic differential scattering cross section are inverse length per unit energy,
commonly expressed in cm1 MeV1 .
The name differential cross section makes some sense if you notice that the integral of the differ-
ential scattering cross section is the regular scattering cross section:

dEf s (Ei Ef ) = dEf s (Ei )P (Ei Ef )
0 0

= s (Ei ) dEf P (Ei Ef )
0

= s (Ei ) . (4.87)

4.11.2 Differential scattering cross section in direction (or in angle)

The discussion above defines the machinerythe differential scattering cross section in energy
for quantifying how scattering changes a neutrons energy. Scattering also changes a neutrons
direction. If we care about the directional distribution of neutrons, we must be able to quantify
such changes. For this we use the differential scattering cross section in direction, s (Ei , i f ),
which is defined such that:
 
s (Ei )P i f df = microscopic cross section for scattering from
direction i to a direction in the cone df around f

= s (Ei , i f )df (4.88)

Of course there is a corresponding macroscopic differential scattering cross section in direction,


obtained by multiplying the microscopic cross section by the nuclide number density.
Once again, the integral of the differential scattering cross section is the scattering cross section:
 
df s (Ei , i f ) = df s (Ei )P i f
4 4

 
= s (Ei ) df P i f
4

= s (Ei ) . (4.89)

79
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

We have noted that in general our differential scattering cross section, s (Ei , i f ) can depend
on neutron energy, target nuclide, pre-scatter direction i , and post-scatter direction f . Actually,
it is rare for the cross section to depend explicitly on i and f . Usually, it is a function only of
the angle between i and f :

s (Ei , i f ) = function of 0 , almost always , (4.90)

where

0 i f = cos ,

where is the scattering anglethe angle between i and f . This means the probability density
of the scattered-neutron direction is also just a function of 0 :
   
P Ei , i f = almost always a function of 0 only = P Ei , i f . (4.91)

CAVEAT!!!!
Beware: sometimes the differential scattering cross section has units of [cm2 / (nucleus-ster)], which
is what we have shown so far, but sometimes the units are: [cm2 / (nucleus-cosine)]. Always make
sure which one you are dealing with! The per-steradian quantity is a regular cross section times a
function whose average is 1/4. The per-cosine quantity is a regular cross section times a function
whose average is 1/2. The difference is a factor of 2. Unfortunately, the same symbol is used for
both functions.

I will follow the following convention (but other authors may not):

If either P or is written with the argument i f , then it is a per-steradian quantity.

If either P or is written with the argument 0 , then it is a per-unit-cosine quantity.

The relationship between the two is:

cm2 cm2
   
s (Ei , 0 ) = 2s (Ei , i f ) (4.92)
nucleus.cosine nucleus.ster

80
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

We can use our differential scattering cross section to determine, for example, the probability that
a neutron is scattered forward from a scattering collision:

prob. forward = df P (Ei , i f )
i f >0

s (Ei , i f )
= df
i f >0 s (Ei )

1
s (Ei , 0 )
= d0 (4.93)
0 s (Ei )

You must be able to perform manipulations like this!

4.11.3 Double-differential scattering cross section

When a neutron scatters off of a nucleus, its change in direction is usually correlated with its
change in energy. That is, we cannot treat directional changes and energy changes as if they were
independent. It follows that a complete description of the scattering process treats both together,
using s (Ei Ef , i f ), which is defined such that:
 
s (Ei )P Ei Ef , i f dEf df = microscopic cross section for scattering from
direction i to a direction in the cone df around f
and from energy Ei to an energy in dEf around Ef

= s (Ei Ef , i f )dEf df (4.94)

Once again, the integral of this differential scattering cross section is the scattering cross section:

dEf df s (Ei Ef , i f )
0 all directions
 
= dEf df s (Ei )P Ei Ef , i f
0 4

 
= s (Ei ) dEf df P Ei Ef , i f
0 4


= s (Ei ) dEf P (Ei Ef )
0

= s (Ei ) . (4.95)

81
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

We have the following identities:



dEf s (Ei Ef , i f ) = s (Ei , i f ) (4.96)
0 | {z } | {z }
cm2 /(nucleus-ster-MeV) cm2 /(nucleus-ster)


d s (Ei Ef , i f ) = s (Ei Ef ) (4.97)
4 | {z } | {z }
cm2 /(nucleus-ster-MeV) cm2 /(nucleus-MeV)


dEf d s (Ei Ef , i f ) = s (Ei ) (4.98)
0 4 | {z } | {z }
cm2 /(nucleus-ster-MeV) cm2 /nucleus

(4.99)

4.11.4 Calculation of s (Ei Ef ) for an important case

Consider elastic scattering, in which kinetic energy is conserved, for the case when the target
nucleus can be taken to be at rest in the lab frame (which is a good approximation if the neutron
speed is >> than the nucleus speed). We assume that our differential cross section depends only
on the scattering angle (L in the lab frame, C in the center-of-mass (COM) frame). Conservation
of energy and momentum leads to the following relationship between lab-frame neutron kinetic
energy and COM-frame scattering angle:

1
Ef = Ei [(1 + ) + (1 )0C ] [elastic only] (4.100)
2
where

Ef = final (post-scatter) neutron energy, lab frame, (4.101)


Ei = initial (pre-scatter) neutron energy, lab frame, (4.102)
2
= [(A 1)/(A + 1)] call attention to this (4.103)
A = ratio of (target mass)/(neutron mass) , (4.104)
0C = cosine of angle through which neutron scatters, COM frame. (4.105)
This equation is derived in Duderstadt & Hamilton, pp. 39-42, and in Stacey, section 1.6. D&H
also has a nice figure at the top of p. 40 depicting scattering in lab and COM frames, which Stacey
reproduces in his Fig. 1.29. What assumptions went into these figures?
Equation (4.100) shows a one-to-one relationship between energy change and scattering angle. (This
is what we expect from elastic scattering.) We will use this to derive an expression for ( Ei Ef ).
We first assume that we are given the differential scattering cross section in angle in the COM
frame:

s (Ei , 0C ) = s (Ei )P (0C ) = given (4.106)

82
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Given a one-to-one relation between scattering-angle cosine (0C ) and final energy (Ef ), which is
what Eq. (4.100) describes, we must have the following equality:

s (Ei Ef )dEf = s (Ei )P (Ei Ef )dEf = s (Ei )P (0C )d0C (4.107)

with Ef and 0C related as in Eq. (4.100). Note that s (Ei Ef ) 6= s (Ei , 0C ). Each is a density
to relate one to the other, we must include their respective phase-space volume elements, dEf and
d0C . This is described in Appendix B, and it is just like making a change of variable in an integral:

1
Let Ef = Ei [(1 + ) + (1 )0C ] .
2
1
Then dEf = Ei d0C , call attention
2
and therefore we must have
1
s (Ei )P (Ei Ef ) Ei d0C = s (Ei )P (0C )d0C
| 2 {z }
dEf
2P (0C )
P (Ei Ef ) = (4.108)
(1 )Ei

Note that there are limits on the Ef for which this is valid. The COM scattering angle, 0C , ranges
from 1 to +1, and if scattering is elastic, then Ef can take on only the corresponding values given
by Eq. (4.100). The limiting values are:

Ef = Ei when 0C = 1 , which corresponds to a backscatter, and


Ef = Ei when 0C = +1 , which corresponds to a grazing collision.

Thus, we finally have the following important result for elastic scattering from nuclei whose lab-
frame speeds are small relative to the neutrons speed:


2P (0C )
, Ef (Ei , Ei )


~|
elastic, |~v |>>|V

(1 )Ei
s (Ei Ef ) s (Ei ) (4.109)


0, otherwise .

If scattering is isotropic in the COM frame, which is often a very good approximation for elastic
scattering off of light isotopes, then the probability density function in 0C becomes simple:

isotropic scattering in COM 1


P (0C ) .
2
This important case is called

s-wave scattering.

83
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

In this case we have

1


, Ef (Ei , Ei )
~ |, swave
elastic, |~v |>>|V (1 )Ei

s (Ei Ef ) s (Ei ) (4.110)


0, otherwise .

and

1


, Ef (Ei , Ei )
~ |, swave
elastic, |~v |>>|V (1 )Ei

P (Ei Ef ) (4.111)


0, otherwise .

This says that if scattering is elastic and isotropic in the COM frame and (a common
situation), then neutrons of initial energy Ei emerge with the simple uniform distribution of post-
scatter energies shown in Fig. 4.11.


Figure 4.11: Illustration of P (Ei Ef ) for s-wave elastic scattering from stationary nuclei. horiz
axis Ef ; show Ef = Ei and Ef = Ei , with flat distribution between them of height 1/(1 )Ei ).
Look at plots in Appendix

Given the assumptions we have made (elastic scattering, neutron speed >> nucleus speed), note
that:

exiting neutron energy is always between Ei and Ei .

a backscatter (0C = 1) produces the largest energy loss .

84
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

If we add the assumption of isotropic scattering in the COM frame, we also have:

the exiting-neutron energy distribution is uniform within the allowed range of Ef

average exiting energy is Ei (1 + )/2.

In particular, note that a neutron loses

about half

of its energy, on average, when it scatters elastically off of a hydrogen nucleus. It can lose no more
than

about 2%

of its energy when it scatters elastically off of a uranium nucleus.

Appendix C provides plots of P (0C ) for various values of Ei , for several nuclides of interest in
reactor studies. Note that isotropic in the center-of-mass frame means that P (0C ) has a constant
value of 0.5.

4.12 Neutrons emerging from fission

Fission is also a reaction from which neutrons emerge. We could define a differential fission cross-
section much like the differential scattering cross section in energy:

f (Ei Ef ) = f (Ei )Pfission (Ei Ef ) (4.112)

In many practical cases, the energies of the neutrons that emerge from fission are almost independent
of the incident neutron energy, which means Pfission (Ei Ef ) is a function of only Ef . We given
this function its own symbol () and name:
usually
Pfission (Ei Ef ) (Ef ) = fission spectrum. (4.113)

The fission spectrum is plotted on linear-linear and log-log scales in Fig. 4.12. The peak value
occurs at 800 keV. The average energy of a neutron emerging promptly from fission is 2 MeV.
99% of the neutrons emerge with energies between 100 keV and 10 MeV.

Regarding the directional distribution of emerging neutrons, in nuclear-reactor analysis it is very


nearly

isotropic.

This means there is no need to define a function that quantifies the directional distribution of
neutrons that emerge from fission. Instead we have simply
usually 1
Pfission (i f ) (4.114)
4

85
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

0.40 1.E+00

0.35

0.30
Fission Spectrum (inverse MeV)

Fission Spectrum (inverse MeV)


1.E-01
0.25

0.20

0.15
1.E-02

0.10

0.05

0.00 1.E-03
0 1 2 3 4 5 6 7 8 9 10 1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01
Energy (MeV) Energy (MeV)

Figure 4.12: Fission spectrum, (E).

4.13 Interactions with low-energy neutrons

When low-energy neutrons scatter off of atoms that are bound to each other in molecules, lattices,
or polymers, the interactions can be substantially different that if the atoms were unbound (as in
a monatomic or free gas). The details can be quite complicated. The effects are most significant
for low-mass atoms. Here we summarize some high-level considerations and provide some plots to
illustrate some of the effects.

4.13.1 Scattering off of molecules

Consider first a gas or liquid composed of molecules, with H2 O being an excellent example. Both
the likelihood of a scattering event and results of the event are strongly influenced by the fact that
the atoms are bound in molecules. Your intuition may suggest that the cross section for scattering
off of an H2 O molecule should be:
intuitive guess
sH2 O (E) 2sH (E) + sO (E) . (4.115)

This is exactly correct for E >> BEmolecule , where BEmolecule is the molecular binding energy,
which is the energy required to break the molecular bonds. However, if E . BEmolecule , then there
is no simple formula for the molecular scattering cross section in terms of the cross sections for the
constituent atoms. In general, the effect of the binding is to increase the scattering cross section
for low-energy neutrons. Figure 4.13 illustrates this effect for H2 O.

We see that it is more likely for a low-energy neutron to scatter off of hydrogen if it is bound
in a molecule than if it is free. The results of the scattering event are different as well. A molecule

86
1.2 4.0E-07

3.5E-07
1

3.0E-07

0.8
2.5E-07

0.6 CDF 2.0E-07


Fission Spectrum, E_inc=1MeV

1.5E-07
0.4

1.0E-07

0.2
5.0E-08

0 0.0E+00
0.E+00 2.E+06 4.E+06 6.E+06 8.E+06 1.E+07
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Figure 4.13: Total cross section of a water molecule compared to the sum of the cross sections of
the individual atoms.

has many degrees of freedomrotational, vibrational, and torsional, for examplethat all affect
the kinematics of the scattering event. This can be calculated, and this is done in cross-section
processing codes, but it is not as simple as the two-body kinematics that governs higher-energy
interactions.

4.13.2 Scattering off of solids

In a solid that has no long-range order, such as a polymer or glass, atoms are bound to each other
but have no regular order. As is the case with molecules, the atom-to-atom binding significantly
changes the neutron scattering cross section relative to that of free atoms.

In a solid that has a lattice structure, atoms are not only bound to one another but also have a
regular order. In addition, they have collective vibrational modes, which are sometimes described in
terms of phonons. The regular order leads to diffraction effects for low-energy neutrons. Graphite
is an example of a material important to nuclear-engineering applications that exhibits these effects.
Scattering cross sections for free carbon and for carbon in a graphite lattice are shown in Fig. 4.14.

87
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

102
MT 221
MT 229
MT 230
101 MT 229 + 230
Cross section (b)

100

10-1 -5
10 10-4 10-3 10-2 10-1 100 101
Energy (eV)
Figure 4.14: Carbon scattering cross sections for low-energy neutrons. MT 221 (top curve) is the
scattering cross section for unbound 12 C atoms. MT 229 (smooth curve whose value is 4 b at 2 eV
and 0.2 b at 0.01 eV) is the incoherent elastic-scattering cross section. MT 230 (jagged curve)
is the coherent elastic-scattering cross section. The discontinuities arise from Bragg scattering.
MT230 goes to zero for E < 0.0017 eV. MT 229 + 230 is the total scattering cross section for
12 C in the form of graphite. Figure was generated using the NJOY cross-section processing code.

4.14 Attenuation, Mean Free Paths, . . .

4.14.1 Attenuation and the uncollided flux: homogeneous material

Consider a uniform beam of mono-energetic neutrons incident normally (perpendicularly) upon a


target, as in Fig. 4.1. Unlike our earlier thought experiment, this time we allow the target to be
arbitrarily thick. We define:
 
n
I(x) intensity of uncollided neutrons at distance x into target
cm2 s
= intensity of neutrons that reach x without interacting with nuclei. (4.116)

88
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Conservation of neutrons tells us:

beam beam intensity intensity


intensity = intensity + added between removed between
at x2 at x1 x1 and x2 x1 and x2

There is no mechanism for adding to the uncollided intensity, so the second term on the right-hand
side is zero. If we assume that x2 x1 is very small, with no hidden nuclei, then the intensity
removed is given by our Eq. (4.22)the thin-target collision ratedivided through by the area A:

t (x2 x1 )<<1 I(x1 )N t A [x2 x1 ]


Intensity removed between x1 and x2 (4.117)
A
This equation gives removal because

every collision removes a neutron from the uncollided beam.

We must divide the equation by A because intensity removal is removal rate per unit area.

Our conservation statement, written mathematically, is therefore:

I(x2 ) = I(x1 ) + 0 I(x1 ) t (x2 x1 ), x2 x1 small. (4.118)


or
I(x2 ) I(x1 )
= I(x1 )t , x2 x1 small. (4.119)
x2 x1

Taking the limit as x2 approaches x1 , we obtain a differential equation:


dI
= I(x)t . (4.120)
dx
We can easily solve this equation using an integrating factor. We obtain:

I(x) = I0 et x , (4.121)

where I0 is the uncollided intensity at x = 0. Thus, uncollided neutrons are

exponentially attenuated

as they try to pass through matter. This is analogous to the exponential decay of radioactive
nuclides as they try to pass through time.

4.14.2 Attenuation with non-constant t

If t is not constant throughout the material, then our differential equation is:

dI
= I(x)t (x) . (4.122)
dx

89
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

We can still solve this equation with an integrating factor, but now the factor has an integral in its
x
exponent. Multiply through by exp{ 0 dst (s)} to obtain:
 x   x 
dI
exp dst (s) + I(x)t (x) exp dst (s) = 0 . (4.123)
0 dx 0

Recognize that
x
d
dst (s) = t (x) , (4.124)
dx 0

which means
 x   x 
d
exp dst (s) = t (x) exp dst (s) . (4.125)
dx 0 0

Then we see that


 x   x 
dI
exp dst (s) + I(x)t (x) exp dst (s)
0 dx 0
  x 
d
= I(x) exp dst (s) . (4.126)
dx 0

It follows that the solution is


 x 
I(x) = I0 exp dst (s) . (4.127)
0

This is a generalization of the previous solution, which held only for constant t .

Example 1: Beam attenuation.


Consider three mono-velocity neutron beams incident on the left, top, and right faces of a brick-
shaped object, as shown in Fig. 4.15. What is the uncollided scalar flux, (x, y, z), as a function of
position inside the brick? You are given the following data:
n cm
n1 = 103 ~v1 = 105 ey
cm3 s
n cm
n2 = 2 103 3 ~v2 = 2 105 ey
cm s
n cm
n3 = 3 103 3 ~v3 = 4 104 ez
cm s
atoms b 2.2 105 cm/s
N = 0.02 t (v) = 5
b.cm atom v
dimensions = x y z = 2cm 5cm 3cm

90
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Figure 4.15: Three mono-velocity neutron beams normally incident on three faces of a rectangular
parallelepiped. Each beam covers the face on which it is incident. The z axis is upward, the y
axis is to the right, and the x axis points out of the page. The origin of the coordinate system,
x = y = z = 0, is at the back lower left corner of the object.

Solution of Example 1: We begin with several fundamental truths:

Scalar flux is neutron density times speed.

If neutrons are moving in beams, then scalar flux is the sum of contributions from each beam.

Density of uncollided neutrons in a beam is exponentially attenuated as a function of distance


traveled through the material.

Attenuation depends on total cross section, which in general depends on neutron speed.

Let us apply these principles to compute each beams contribution to the uncollided scalar flux
(which is the beam intensity, for we know that beam intensity is beam density beam speed).
First we determine the macroscopic total cross section that applies to each beam:

atoms b 2.2 105 cm/s


t,1 = N t (|~v1 |) = 0.02 5 = 0.22 cm1
b.cm atom 105 cm/s
atoms b 2.2 105 cm/s
t,2 = N t (|~v2 |) = 0.02 5 = 0.11 cm1
b.cm atom 2 105 cm/s
atoms b 2.2 105 cm/s
t,3 = N t (|~v3 |) = 0.02 5 = 0.055 cm1
b.cm atom 4 105 cm/s

91
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Then we recognize that the contribution of each beam to the uncollided scalar flux is the incident
beam intensity times an attenuation factor, which depends on the total cross section and the
distance the neutrons must travel from the incident face to a given point in the object:
h n cm i [ 0.22 y]
u,beam1 (x, y, z) = 103 3 105 e cm
cm s
h n cm i [ 0.1 (5 cmy)]
u,beam2 (x, y, z) = 2 103 3 2 105 e cm
cm s i
h n cm [ 0.055 (3 cmz)]
u,beam3 (x, y, z) = 3 103 3 4 105 e cm
cm s
Finally, we sum the contributions to get the complete uncollided scalar flux inside the object:
n 0.22 0.1 0.055
o n
u (x, y, z) = 108 e cm y + 4 108 e cm (5 cm y) + 1.2 109 e cm (3 cm z)
cm2 .s
Note:

In some cases the attenuation distance is not simply the x or y or z coordinate. Instead, for
example, the beam coming in from the top has to penetrate a distance (3 cm z), not just
z, to reach a point in the material that has coordinate z.

Units are important. x, y, and z contain length units. You cannot add (subtract) something
to (from) one of these variables unless that something also has length units. It does not
make sense to write (3 z), for example. It does make sense to write (3 cm z) or (3 z/[1
cm]), for example. Also note that exponents must be dimensionless. It would be incorrect to
write (0.22y) as an exponent, but ([0.22 cm1 ]y) is okay.

In this example, the scalar flux does not vary with x, so x does not appear on the right-hand
side of its equation. Dont let this bother you. It happens because all three of the incident
beams in this problem are uniform in x, and their attenuation distances dont vary with x.

Example 2: More complicated beam attenuation.


Consider a mono-velocity neutron beam diagonally incident on the left, top, and right faces of a
brick-shaped object, as shown in Figs. 4.16 and 4.17. What is the uncollided scalar flux, (x, y, z),
as a function of position inside the object? You are given the following data:

cube width = 4 cm
n
n = 103
cm3 " #
cm 2 2
~v = 4 104 ey ex
s 2 2
t = 1.5 cm1 at this neutron speed.

92
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

Figure 4.16: 3D view of a beam incident on a cube at a 45 angle relative to two of the faces. The
z axis is upward, the y axis is to the right, and the x axis points out of the page. The origin of the
coordinate system, x = y = z = 0, is at the back lower left corner of the object.

Figure 4.17: View looking down on a beam incident on a cube at a 45 angle relative to two of the
faces: the face at x = 3 cm and the face at y = 0.

Solution of Example 2:
We recognize that the uncollided scalar flux at a given point in the cube is the incident beam
intensity times an attenuation factor, which depends on the total cross section and the distance
the neutrons must travel through the material to reach the given point. The chief challenge in this
example is to compute the attenuation distance as a function of (x, y, z) inside the object. The
student is advised to derive the following expression for himself or herself:
y
, y (3 cm x) ,
cos 45


attenuation distance d(x, y, z) =
3 cm x , y (3 cm x) .


cos 45

93
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

The solution follows:


n
u (x, y, z) = 4 107 et d(x,y,z)
cm2 .s n o
y

exp 1.5 (1 cm) cos 45 , y (3 cm x) ,
7 n
= 4 10
cm2 .s n
cm x
o
exp 1.5 (1 3cm) , y (3 cm x) .


cos 45

Note that in half of the object (the upper-left triangular portion in the view of Fig. 4.17), the scalar
flux depends only on the distance from the left face, which is y. In the other half, the flux depends
only on the distance from the front face, which is 3 cm x.

Do not let simple geometry stop you from solving a problem!

4.14.3 Mean free paths

It is often useful to know the average path length that a neutron travels between collisions. We
give it a name:

mean free path average path length between collisions

Let us compute it. We first need the probability of traveling an arbitrary distance x before colliding
within dx about x:

p(x)dx probability that a neutrons first collision is in dx about x


= (first-collison rate in dx about x)/(incident rate)

t I(x)dx
=
I0

= t et x dx . (4.128)

Now we want the average (mean) of the distance x. Note that in general, the average of a function
f (x) over the interval (x1 , x2 ) is:

x2
w(x)f (x)dx
mean (average) value of f = hf i = x1
x2 , (4.129)
x1 w(x)dx

94
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

where w(x) is some given weight function. In our case the weight function is the probability
density function p(x); thus:

mean distance to first collision

= mean free path


xp(x)dx
= 0
0 p(x)dx


= xp(x)dx tell them denom = 1
0

1
= . (4.130)
t

A large cross section means a short mean free path, as one would expect. Of course, mean free
path has units of

length.

We can define mean free paths for various individual reactions as well. For example,

mean path length (not straight line distance) to absorption


= absorption mean free path

1
= . (4.131)
a
Another commonly-used mean free path is the scattering mean free path, 1/s .

Example.
How far do 2200-m/sec neutrons travel, on average, between collisions in a gas mixture with t =
0.72 cm1 ?

Solution
The question asks for the neutron mean free path. The answer is

mean free path = 1/t 1/(0.72cm1 ) 1.39cm . (4.132)

Note that the number of mean-free paths along the x axis between x1 and x2 is
x2
constant t
number of mean-free paths = dst (s) t (x2 x1 ) (4.133)
x1

95
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

4.15 Summary

We know that reactor behavior fundamentally depends on the production and loss rates of neu-
trons. Many production and loss mechanisms are neutron-nucleus reactions; thus, computation of
production & loss rates requires computation of reaction rates. The following summarizes a few
key facts about reaction rates:

1. Reaction rates are volume integrals of reaction rate densities.

2. A reaction-rate density involves a microscopic cross section, which depends fundamentally


on:

(a) the nuclide (i.e., A and Z),


(b) the reaction (i.e., capture or inelastic scattering),
(c) the relative speed between the neutron and nucleus, vr |~v V ~ | (or equivalently, the
kinetic energy of the two bodies in the center-of-mass reference frame).

3. Reaction rate densities are fundamentally integrals over six variables: three that describe
neutron velocity (or energy and direction) and three that describe nucleus velocity (or energy
and direction). The reason is that the relative speed between the neutron and nucleus is a
function of all six variables.

4. It is often the case that nuclei have approximately isotropic velocity distributions, and further
that these distributions are approximately Maxwellians, which means they are characterized
by a single number: the material temperature. In such a case, the product of vr cold (vr ) can
be pre-averaged over nucleus velocities as a function of material temperature and lab-frame
neutron speed. This averaged vr cold can be divided by that lab-frame neutron speed, v,
to give an averaged cross section that depends only on neutron lab-frame speed (or kinetic
energy) and the material temperature.

5. For every microscopic cross section there is a corresponding macroscopic cross section, which
is a nucleus density times a microscopic cross section.

6. The macroscopic cross section for a mixture


P iisi the sum of the macroscopic cross sections for
each nuclide in the mixture: x (E) = i N x (E).

7. x (E) is expected reactions of type x per unit path length.

8. If microscopic cross sections are pre-averaged over nucleus velocities, a reaction-rate density
can be written as an energy integral of an energy-dependent scalar flux times a macroscopic
cross section that depends on the neutron lab-frame kinetic energy (or speed) and the material
temperature. That is, the six-dimensional integral collapses to a one-dimensional integral.

9. Neutron scalar flux, , is always the product of neutron speed neutron density. This is
true for energy-dependent , velocity-dependent , speed-dependent , etc.

96
NUEN 601, M.L. Adams notes, 2017 Chapter 4. Nuclear Reactions

10. (~r, E) is path-length rate per unit volume per unit energy at position
~r. (~r, v)
is path-
length rate per unit volume per unit speed at position ~r. (~r) = dE(~r, E) = dv(~r, v)
is path-length rate per unit volume.

11. If neutrons are moving in beams, then the scalar flux is the sum of the beam intensities. Note
that the intensity of a particle beam is the particle density times the particle speed.

12. The intensity of the uncollided neutrons in a beam is exponentially attenuated as the beam
penetrate matter. The exponent is the total cross section multiplied by the penetration
distance.

13. The mean free path for reactions of type x, which is the average total distance a neutron of
energy E travels before having a reaction of type x, is 1/x (E). This average distance is
independent of the neutrons historyit holds for any neutron of energy E in the material
whose cross section is x (E), even if that neutron has already traveled some distance in the
material! (The neutron has no memory of its previous travels.)

14. Note that the exponent in an attenuation factor is t d, where d is the attenuation distance,
if t is constant. Note that this is just the distance, d, expressed in units of mean-free paths:
d distance of attenuation
t d = = (4.134)
1/t distance of 1 mean-free path

15. The number of mean-free paths along a path through a material is in general the integral of
t along that path. If t is constant, then this simplifies to t times the distance. Otherwise
it can be viewed as the sum of (t distance) values over an infinite number of infinitesimal
distances.

16. Differential scattering cross sections are products of scattering cross sections probabil-
ity density functions that describe the distribution of post-scattering neutrons energies or
directions or both.

17. Given elastic scattering from stationary nuclei, the post-scatter lab-frame neutron energy is
a linear function of the cosine of the center-of-mass scattering angle.

18. If scattering is isotropic in the center-of-mass reference frame, then neutrons of a given pre-
scattering energy have a uniform distribution in post-scattering energy.

97
NUEN 601, M.L. Adams notes, 2017

Part III

Conservation of neutrons: transport


and diffusion equations

98
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Chapter 5

Flux, Current, Etc.

5.1 Introduction

A reactors behavior depends on the production rates and loss rates of neutrons in the reactor. We
also know that some production and loss is due to neutron-nucleus reactions, and that some is due
to neutron leakage.

Reaction rates usually can be calculated from the neutron scalar flux. Net leakage rates (net flow
rates across surfaces) can be calculated from the neutron net current density (often called current
for short). One-way leakage rates can be calculated from neutron partial current densities. All of
these can be calculated from the angular flux. In this chapter we study the scalar flux, angular
flux, net current density, and partial current densities, and we explore how to use them to calculate
quantities of interest.

5.2 Scalar flux

5.2.1 One-speed (mono-energetic) case

Consider some volume, V , in which at time t the density of neutrons is n(t) neutrons/cm3 , with
all neutrons moving at the same speed v but in many different directions (see Fig. 5.1).

What is the rate at which neutrons make tracks in this volume? To say it differently, what is the
total distance traveled by neutrons in this volume per unit time? In time t, each neutron travels:

distance = v t units: [cm/s] [s] = [cm] (5.1)

99
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Figure 5.1: One-speed neutrons in a volume V .

If the neutron density at time t is n(t), then there are n(t) V neutrons in the volume at time t.
Thus, total distance traveled by neutrons in V between t and t + t is

distance traveled (path-length) in V in (t, t + t) is n(t) V v t .

[n/cm3 ] [cm3 ] [cm/s] [s] = [n-cm] (5.2)


If we divide by the volume (to get a density) and by the time interval (to get a rate), we find:

(~r, t) = n(~r, t) v = path-length rate density


in units of: [n/cm3 ] [cm/s] = [n.cm/(cm3 .s)] = [n/(cm2 .s)] (5.3)

This is a general result:

Scalar flux is always speed times density.

Be careful: density is context-dependent. It is always some number of things per unit volume,
but volume may not be the physical x-y-z volume. There are many cases in which density
is number per something else. For example, population density is inhabitants per unit area, car
density on the highway is cars per unit length, etc. See Appendix B and recall the general definition
below.

Note:

Every density is a limit. Mathematically, the density of things at some point ~x in some phase
space is
 
expected number of things in V around ~x
d(~x) = lim (5.4)
V small V

where ~x is a point in a phase space and V is a phase-space volume that contains the point ~x.

100
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

5.2.2 Energy-Dependent Scalar Flux: distribution of speeds (energies)

Neutrons do not bounce around with a single energy in a nuclear reactor; they have a distribution
of energies.

Question: How many neutrons in a commercial reactor have exactly 1 MeV kinetic energy?

Answer: Exactly ZERO.

If we asked how many neutrons have energies between 0.999 and 1.001 MeV, instead of exactly
1.0000000000000000. . . MeV, the answer would be non-zero. This chapter will give you the tools
to find this answer.
We define the energy-dependent neutron density n(~r, E, t) such that:

n(~r, E, t)dxdydzdE = expected number of neutrons in volume d3 r at position ~r


whose energies are in the interval dE at energy E
at time t . (5.5)

What are the units of n(~r, E, t)? The phase space volume here is the product of x, y, z, and E
intervals, so the units are:

energy-dependent neutron density units = neutrons / (length3 energy).

(Note that n dx dy dz dE is number of neutrons, d3 r = dxdydz is spatial volume, dE is


energy.) Typically, the energy is expressed in MeV (or keV or eV). We will mainly use MeV but
be sure to check every time, because it does make a difference!

Likewise, we define the energy-dependent scalar flux:

(~r, E, t) n(~r, E, t) v(E) (5.6)

= path-length rate density


n cm n.cm n
Typical units: 3
= 3
= 2
cm .M eV s cm .M eV.s cm .M eV.s
Example
Suppose that in some reactor, the energy-dependent scalar flux in the moderator is given by:
0
(~r, E, t) = for E (10 eV, 10 keV)
E

101
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Question: What are the units of and 0 ?

Answer: has units n/(cm2 -s-MeV), so 0 must have units n/(cm2 -s).

Question: What is the density (n/cm3 ) of neutrons that have energies between 1 and 10 keV?
p
Answer: Since v(E) = 2E/m, where m is neutron mass, we have [see Eq. (5.3)]:

(~r, E) 0
n(~r, E, t) = = p
v(E) E 2E/m

Thus,
10 keV 10 keV
dE
answer = p 0 =p 0 E 3/2 dE
2/m 1 keV E E 2/m 1 keV

h i10 keV
= p 0 (2) E 1/2
2/m 1 keV
" #
1 1
= 20 p p
2 [1 keV] /m 2 [10 keV] /m
 
1 1
= 20
v(E)E=1 keV v(E)E=10 keV

5.3 Angular Flux

Often we require knowledge of not only the distance traveled by neutrons per unit time per unit
spatial volume per unit energy (which is (~r, E, t), but also of their distribution in direction. Let
us define the neutron angular density as follows.

n(~r, E, , t)d3 rdEd = number of neutrons in volume d3 r at position ~r


whose energies are in the interval dE at energy E
and whose directions are in the cone d at direction
at time t . (5.7)

Recall that

d3 rdEd = dx dy dz dE d d sin
= dx dy dz dE d d ,

where cos . That is, this is a differential volume in a space with six independent axes.
That is, the space of position-energy-direction is a six-dimensional phase-space volume. The
neutron angular density is a density in this 6-D phase space.

102
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

The neutron angular flux is defined to be speed times angular density:

(~r, E, , t) n(~r, E, , t) v(E) (5.8)

= path-length rate density


n cm n.cm n
Typical units: 3
= 3
= 2
cm .M eV.ster s cm .MeV.ster.s cm .M eV.ster.s
Note the relations among the three neutron densities we have encountered:


n(~r, E, t) = d n(~r, E, , t) (5.9)
4

2
= d d sin n(~r, E, , , t) (5.10)
0 0
2 1
= d d n(~r, E, , , t) (5.11)
0 1


n(~r, t) = dE n(~r, E, t) (5.12)
0

= dE d n(~r, E, , t) (5.13)
0 4

Also be aware of our notational sins: we use the same symbol, n, to represent three different
neutron-density functions. When you encounter a function called n, you must infer from the
context which density it representsthat is, you must infer whether it is a density in position &
energy & direction, in position & energy, or just in position.

103
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

The following are direct results of the relations above:


(~r, E, t) = d (~r, E, , t) (5.14)
4
show units on this line

2
= d d sin (~r, E, , , t) (5.15)
0 0
2 1
= d d (~r, E, , , t) (5.16)
0 1

(~r, t) = dE (~r, E, t) (5.17)
0

= dE d (~r, E, , t) (5.18)
0 4
show units on this line

We refer to (~r, t), the scalar flux due to neutrons of all energies, as the total scalar flux.

Again, be aware of our notational sins: we use the same symbol, , to represent two different
scalar-flux functions. When you encounter a function called , you must infer from the context
which function it representsthat is, you must infer whether it is a path-length-rate density in
position & energy or just in position.

5.4 Reaction Rates

We saw in Chapter 4 that the general expression for the rate per unit volume at which a reaction
of type r occurs at position ~r at time t can be written as
 
RRDr (~r, t) = dvx dvy dvz n(~r, ~v , t) ~ , t) ~v V
dVx dVy dVz N(~r, V ~ r,cold ~v V
~
neutron vels. nucl. vels.

We can rewrite this in terms of energy- and direction-dependent densities, noting that (for example)
~v = v(En )n :
n
RRDr (~r, t) = dEn dn n(~r, En , n , t)
0
4
dEnuc dnuc N(~r, Enuc , nuc , t) v(En )n V (Enuc )nuc

0 4 o

r,cold v(En )n V (Enuc )nuc

104
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

As we saw in Chapter 4, we can define an averaged x such that this expression becomes
n  o
RRDr (~r, t) = dEn dn n(~r, En , n , t)N (~r, t)v(En )r v(En )n
0 4 h ih  i
= dEn dn n(~r, En , n , t)v(En ) N (~r, t)r v(En )n
0 4  
= dEn dn (~r, En , n , t)r ~r, v(En )n , t (5.19)
0 4

We also saw in Chapter 4 that we can almost always take the nuclei to be vibrating isotropically,
in which case the averaged cross section no longer depends on the neutron direction, but only on
the neutron speed, or equivalently energy:

isotropic nuclei
RRDr (~r, t) dEn dn (~r, En , n , t) r (~r, En , t)
0 4
| {z }
(~
r,En ,t)


= dE (~r, E, t)r (~r, E, t) (5.20)
0

The averaged r , which is contained within r , depends on the distribution of nucleus speeds,
which we can often characterize by the material temperature. We often omit temperature from
the argument list, but the dependence is implied by the form of the equation. That is, we know
that if we write Eq. (5.20) for RRD, then the cross section must have been averaged over nucleus
velocities, and those velocities depend on temperature.

We can also define an energy-dependent reaction-rate density for reactions of type r:

RRDr (~r, E, t) = r (~r, E, t) (~r, E, t) . (5.21)

Units:      
reactions reactions cm of n travel
=
cm3 .MeV.s cm of n travel cm3 .MeV.s

105
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Important!

1. (~r, E, t) is a density in space and in energy. It is neutron path length rate per cm3 per MeV.

2. It does not make sense to ask what the path-length rate is at a point in phase space, such as
~r and/or E. It does make sense to ask what the path-length rate density is at a point.

3. (~r, E, t) is not a density


in space or in energy. It does make sense to ask what is at ~r and at E.

4. s (~r, Ei Ef , t) is not a density in space or in incident energy, Ei .


It is a density in outgoing energy, Ef .
How many neutrons have outgoing energy of exactly 100 keV after a scattering event?
Exactly zero.

5. s (~r, Ei Ef , i f , t) is not a density in space, in incident energy, Ei , or in incident


direction, i .
It is a density in outgoing energy, Ef , and outgoing direction, f .
How many neutrons have outgoing direction f ey after a scattering event?
Exactly zero.

5.5 Examples

Suppose you have been analyzing a reactor that is operating in steady state, and you have computed
the energy-dependent scalar flux, (~r, E), everywhere in the reactor and for all neutron energies.
You also know all the macroscopic cross sections everywhere in the reactor.
Question #1: In terms of the known function (~r, E), how many neutrons are in the reactor?

Answer: First, the number of neutrons is the volume integral of the neutron density:

ntot [neutrons] = d3 r n(~r) Note: n(~r) units are [ n/cm3 ]
volume

Second, the regular neutron density is the energy integral of the energy-dependent neutron den-
sity:

3
n(~r) [n/cm ] = dE n(~r, E) Note: n(~r, E) is in [ n/(cm3 .MeV)]
0

Third, the energy-dependent neutron density is the energy-dependent scalar flux divided by speed:

(~r, E) (~r, E)
n(~r, E) [ n/(cm3 .MeV) ] = =p
v(E) 2E/m
Note: It is OK to use the classical expression for kinetic energy of neutrons in reactors. Very few
neutrons have energies as high as 10 MeV. For a 10-MeV neutron, v/c 1/7 and the error in the
classical expression for kinetic energy is 1.5%. The error for lower-energy neutrons is smaller.

106
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Putting everything together, we obtain an expression for the total neutron population in the reactor,
in terms of the energy-dependent scalar flux function:

3 (~r, E)
nnot [neutrons] = d r dE p .
volume 0 2E/m

Question #2: Assuming each fission produces 200 MeV of energy and 190 MeV of this gets
deposited in the reactor (with the other 10 MeV carried away by leaking neutrinos, gammas, and
neutrons), what is the reactor power that is coming from fissions?

Answer:

energy energy deposited


power = = fission rate ,
s fission

energy deposited
= 190 MeV/fission ,
fission

fission rate = d3 r (fission rate density)
volume


fission rate density = dE f (~r, E)(~r, E)
0

Combine everything:

190 MeV 3
power [MeV/s] = d r dE f (~r, E)(~r, E)
fission volume 0

show units on this line show units on this line

Question #3: At what rate are neutrons being produced by fission in the reactor?

Answer: First, a rate is the volume integral of a rate density (rate per unit volume):
hni
rate = d3 r (production rate density)
s volume

Second, a rate density involving neutrons is the integral over energy of an energy-dependent rate
density. We use the symbol (~r, E) to represent the average number of neutrons that are emitted
(produced) by a fission that is caused by a neutron of energy E given the material at ~r:
h n i
production rate density = dE (~r, E)f (~r, E)(~r, E)
cm3 .s 0

107
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Combine everything:


3
rate [n/s] = d r dE (~r, E)f (~r, E)(~r, E)
volume 0

show units on this line show units on this line

Question #4: What is the rate at which neutrons scatter from energies above 1 MeV to energies
below 0.5 MeV in the reactor?

Answer: First, a rate is the volume integral of a rate density (rate per unit volume):

rate [n/s] = d3 r (downscattering rate density)
volume
show units on this line show units on this line

Second, the downscattering rate density that was requested requires integration over initial neutron
energies and integration over final neutron energies. The quantity that is integrated is the scalar
flux times a differential scattering cross section:

Ei = Ef =0.5 MeV
downscattering rate density = dEi dEf s (~r, Ei Ef )(~r, Ei )
Ei =1 MeV Ef =0

Combine everything:

Ei = Ef =0.5 MeV
3
rate [n/s] = d r dEi (~r, Ei ) dEf s (~r, Ei Ef )
volume Ei =1 MeV Ef =0

show units on this line show units on this line

Ei = Ef =0.5 MeV
= d3 r dEi (~r, Ei )s (~r, Ei ) dEf P (~r, Ei Ef )
volume Ei =1 MeV Ef =0

show units on this line show units on this line

5.6 Net Current Density and Leakage Rates

In this section we will find that the net leakage of neutrons across a surface can be written in terms
of a vector quantity called

neutron net current density.

108
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

We will find that the net leakage rate, across a flat area dA whose unit normal is en , is just

en J~ dA (5.22)

where J~ is the neutron net current density [n/(cm2 .s)]. Sometimes we call this simply the neutron
current density or net current density or net current or even just current. All of these terms
refer to the same thing, which most precisely is called the neutron net current density.

5.6.1 One-speed Case

Consider a beam of mono-energetic, mono-directional neutrons crossing a surface that is perpen-


dicular to the x axis (see Fig. 5.2). Here is defined to be a unit vector in the direction of the
beam. The number of neutrons crossing the surface in time t is:

number crossing area A in time t = n v cos A t , (5.23)


s show units on this line

where is the angle between and ex .

Figure 5.2: Beam crossing surface perpendicular to x axis. label x and y axes; label , ex , and ;
show time t and t + t; show vt and vt cos

Now, because
cos = ex (5.24)
and because
n v = I = beam intensity (5.25)
we have

number crossing area A in time t = I ex At (5.26)

109
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Thus, the crossing rate per unit area is:

crossing rate per unit area = I ex (5.27)

Question: What if > 90 ?

Answer: Then cos < 0 and the crossing rate is negative.

This means neutrons flow from the +ex side of the surface to the ex side.

Now consider the same beam again, but with a different surface (see Fig. 5.3). The only difference
between this case and the previous case is the orientation of the surface. In particular, the unit
vector normal to the surface is now ey instead of ex ; thus,

cos = ey (5.28)

Figure 5.3: Neutron beam crossing surface perpendicular to ey . label x and y axes; label , ey ,
and ; show time t and t + t; show vt and vt cos

Everything else is the same, so

crossing rate per unit area = I ey (5.29)

In fact, given a surface with any orientation (see Fig. 5.4), we have

crossing rate per unit area = I en (5.30)

110
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Figure 5.4: Neutron beam crossing surface with arbitrary orientation.

Notice that no matter what the orientation of the surface was, we needed the quantity

I (5.31)

in order to compute the crossing rate. In our simple beam example, this is the net current density:

J~ = net current density = I [only for a beam !!!!] (5.32)


This formula is valid only for a mono-directional beam!

Note:

1. The net current density has the same units as the scalar flux.

2. It does not have the same physical meaning. Scalar flux is fundamentally a volumetric
density (distance traveled by neutrons per unit VOLUME per unit time), whereas the
net current density is fundamentally an areal density (when dotted with en , net neutrons
per unit AREA per unit time).

3. This may seem confusing at first. Im sorry. Thats just the way it is.

5.6.2 Relation of J~ and to Beam Intensities

Consider some volume in which several different mono-directional beams are crossing. In region
where they cross, the neutron net current density is the vector sum of the net current density from
each beam:
# ofX
beams
J~ = Ik k [only for beams !!!!] (5.33)
k=1

111
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Similarly, in a region where multiple beams are crossing, the neutron scalar flux is the scalar sum
of the beam intensities:
# ofX
beams
= Ik [only for beams !!!!] (5.34)
k=1

In the many-beam case, the net current density from some beams may cancel the net current
density from others. For example, given two beams of the same intensity but opposite directions,
we would have:
I1 = I2 and 2 = 1 , (5.35)
which means
 
J~ = I1 1 + I2 2 = I1 1 + 2 = ~0 [opposing beams of same I] (5.36)

Thus, J~ is related to the net rate at which neutrons cross surfaces, which is why we call it the

net current density

5.6.3 Energy-Dependent Neutron Current Density

We know that neutrons dont travel about mono-energeticallyneutrons in reactors are distributed
in energy. As we did with the scalar flux, we define an energy-dependent net current density:
~ r, E)dE = net current density of neutrons at position ~r
J(~
whose energies are in the interval dE at energy E (5.37)

~ r, E)dE has units of net current density, which are:


We see that the product J(~

n/(cm2 -s).

~ r, E), must have units of


Thus, the energy-dependent net current density, J(~

n/(cm2 -s-MeV).

5.6.4 Net Current Density: General Case (Not Beams)

Neutrons do not normally move around in mono-directional beams, and they certainly do not do so
in nuclear reactors. In real-world scenarios neutrons have directional distributions. As we have
seen, these distributions are quantified by the direction-dependent neutron density, n(~r, E, , t) or,
equivalently, by the angular flux, (~r, E, , t).

112
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

We can view the product


(~r, E, , t) d dE
show units on this line
as a mini-beam intensity, involving only neutrons with directions and energies E. Recall
Eq. (5.33), which says that in the case of beams, J~ is the sum of the I for all the beams. In the
limit of infinitely many mini-beams, with one beam for each d, this sum becomes an integral
over all directions:


~ r, E, t) =
J(~ d (~r, E, , t) (5.38)
4
show units on this line

2
= d d sin (~r, E, , , t) (5.39)
0 0
2 1
= d d (~r, E, , , t) (5.40)
0 1

~ r, t) =
J(~ ~ r, E, t)
dE J(~ (5.41)
0

= dE d (~r, E, , t) (5.42)
0 4
show units on this line

~ r, t), which is the net current density due to neutrons of all energies, is called the total net
J(~
current density.

Again, be aware of our notational sins: we use the same symbol, J, ~ to represent two different
net-current-density functions. When you encounter a function called J,~ you must infer from
the context which function it representsthat is, you must infer whether it is a density in area &
energy or just in area.

113
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

More explicitly for each component:


2 1
Jx (~r, E, t) = d d (1 2 )1/2 cos (~r, E, , , t) (5.43)
0 1 | {z }
x
2 1
Jy (~r, E, t) = d d (1 2 )1/2 sin (~r, E, , , t) (5.44)
0 1 | {z }
y
2 1
Jz (~r, E, t) = d d (~r, E, , , t) (5.45)
0 1 |{z}
z

Note that we are using the same symbol, J, ~ for two different functions: the energy-dependent net
current density and the total net current density. This is not ideal, but it is standard practice.

5.6.5 Leakage Rates

We need often need to use the net current density to calculate

net leakage rates

across surfaces. Heres how it goes. Consider a point ~rs on some surface, with outward unit normal
en (~rs ) (see Fig. 5.5). Then, given what we learned previously for net leakage rates across surfaces,

the net leakage rate density


~ rs )
h n i
outward through surface at point ~rs = en (~rs ) J(~ (5.46)
cm2 .s
Note that this is an

areal rate density (rate per square centimeter)

To get the net leakage rate through some surface, we integrate the net leakage rate density over
the area. If we denote an infinitesimal surface area as d2 r, then

net leakage rate



outward through surface S = ~ rs )
d2 r en (~rs ) J(~ (5.47)
S

It really is that simple!

114
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Figure 5.5: A point on some surface, with outward unit normal en . show d2 r and en

Example
You have been analyzing a parallelepiped reactor, whose spatial domain is given by:

a a
x
2 2
b b
y
2 2
c c
z
2 2
The reactor is operating in steady state, and you have somehow figured out that the total net
current density (i.e., energy-dependent net current density integrated over all energies) is:

~ y, z) = J0 ex sin x + ey sin y + ez sin z ,


h i
~ r) = J(x,
J(~ (5.48)
a h i b c
n
units of J~ :
cm2 .s
Note: This means that the three components of the net current density are:

~ r) ex = Jx (x, y, z) = Jx (x) = J0 sin x ,


J(~
a
~ y
J(~r) ey = Jy (x, y, z) = Jy (y) = J0 sin ,
b
~ r) ez = Jz (x, y, z) = Jz (z) = J0 sin z .
J(~
c
Question 1: What are the units of the constant J0 ?

Answer: J0 must have units of n/(cm2 -s), because the sin functions are dimensionless.

115
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Question 2: Which of the following is a vector and which is a scalar?


~ r), Jx (~r), Jy (~r), Jz (~r), J0
J(~

Answer: vector, scalar, scalar, scalar, scalar

Question 3: At what net rate do neutrons leak outward through the top (z = +c/2) surface of
the reactor?

Answer:

net leakage rate


a/2 b/2  c
outward through top surface = dxdy ez J~ x, y,
a/2 b/2 2
a/2 b/2

= J0 dx dy sin = J0 [a][b][1]
a/2 b/2 2
hni
= J0 ab units:
s
Question 4: What is the total net rate at which neutrons leak out of the reactor?

Answer:
We will compute leakage out of the bottom (z = c/2) surface, then write the others by inspection.

net leakage rate outward


a/2 b/2  
~ c
through bottom surface = dx dy (ez ) J x, y,
a/2 b/2 2
a/2 b/2

= J0 dx dy (1) sin = J0 [a][b][1]
a/2 b/2 2
hni
= J0 ab units:
s
The other surfaces are similar; the final answer is

net leakage rate out of reactor = [top + bottom + back + front + right + left] leakage rates
= = 2J0 (ab + bc + ac)

116
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Question 5: At what net rate do neutrons leak from back to front across the surface at x=a/4?

Answer:
Since we said back to front, the unit normal we need is +ex . Thus,

net leakage rate


from back to front across surface
b/2 c/2 a 
at x = a/4 is: dz ex J~ , y, z
dy
b/2 c/2 4
b/2 c/2

= J0 dy dz sin = J0 [b][c][1/ 2]
b/2 c/2 4
J0 bc hni
= units:
2 s
Note that this leakage rate

is negative!

This is okay. It just means that more neutrons are flowing toward the back than toward the front
across this particular surface.
If the question had asked for the net rate of leakage from front to back across the surface x=a/4,
we would have gotten

J0 bc
rate = + .
2
because we would have used the unit normal ex instead of +ex . Make sure you understand this!

Example
You have been analyzing a parallelepiped reactor, whose spatial domain is given by:

a a
x
2 2
b b
y
2 2
c c
z
2 2
and you have found that the energy-dependent net current density is:
 
~ ~ E E h x y z i
J(~r, E) = J(x, y, z, E) =J0 2 exp ex sin + ey sin + ez sin ,
E0 E0 a b c

h n i
with units of
cm2 .s.MeV

117
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Question 1: What is the net neutron loss rate due to leakage from the reactor?

Answer: First lets compute net loss rate out of the back face (x=a/2) of the reactor:


net leakage back = d2 r ~ r, E)
dE en (~r) J(~
back surf 0
b/2 c/2  
~ a
= dy dz (ex ) J , y, z, E
b/2 c/2 2
b/2 c/2

 
E E
= J0 dy dz (1) sin dE 2 exp
b/2 c/2 2 0 E0 E0
 
J0 E
= [b][c](1)[1] dE E exp
E02 0 E0
J0 h 2 E/E0
iE=
= bc (E 0 E + E0 )e
E02 E=0

= J0 bc
hni
units:
s

Similarly,


net leakage front = d2 r ~ rs , E)
dE en (~rs ) J(~
front surf 0
b/2 c/2  
+a
= dy dz (+ex ) J~ , y, z, E
b/2 c/2 2
b/2 c/2
+
 
E E
= J0 dy dz (+1) sin dE 2 exp
b/2 c/2 2 0 E0 E0
hni
= = J0 bc , with units =
s
The other faces are handled the same way. The final answer is

net leakage rate out of reactor = [top + bottom + back + front + right + left] leakage rates
= 2J0 (ab + bc + ac)

118
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Question: What are the units of J0 ? How about E0 ?

Answers:

1. We know from our answer above that (area)(J0 ) must be neutrons/s; thus,
J0 units must be n/(cm2 .s)

2. We know that exponents are dimensionless; thus


E0 must have the same units as E, i.e., energy units

5.7 Partial Current Densities and One-Way Flow Rates

5.7.1 Connection to Angular Flux

It follows from what we have discussed here that

angular flux is a path-length rate density.

Thus, if we integrate the angular flux over some portion of a phase-space volume, we will get the
expected total distance traveled by neutrons in that portion of the phase-space volume, per unit
time.

It is also true that the angular flux defines an infinite set of beam intensities, with one beam for
every d-dE combination:

(~r, E, , t)dEd = beam intensity of the beam of neutrons


whose directions are in the infinitesimal cone d
and whose energies are in the infinitesimal interval dE
at position ~r at time t . (5.49)

Consider Fig. 5.4, which illustrates a beam of neutrons moving in direction through a small
patch of surface area, d2 r, that has unit normal en . Now consider a beam intensity I equal to
(~r, E, , t)ddE. Then we have
h i
neutron crossing rate per area = (~r, E, , t)ddE en (5.50)
| {z }
I= beam intensity

This is the crossing rate considering only neutrons whose energies are in dE at E and whose
directions are in the cone d around . A negative number would simply mean that the neutrons
are crossing from the +en side to the en side.

119
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Question: What is the net rate per unit area (n/cm2 -s) at which neutrons whose energies are in
dE at E are crossing the surface perpendicular to en , going from the en side to the +en side,
taking into account all possible directions?

Answer: The question asks for a net rate, which means neutrons that cross from the en side to
the +en side make a positive contribution whereas neutron that cross from the +en side to the
en side make a negative contribution. The sign of en automatically takes care of the positive
and negative for us:

n h i
net = d (~r , E, , t)dE en
cm2 -s 4

h i
~ r, E, t)dE en
= J(~ (5.51)

Note that if en > 0, then neutrons are crossing to the +en side of the surface, and if en < 0,
then neutrons are crossing to the en side of the surface. The integral therefore gives the net
crossing ratethe difference between two one-way crossing rates.

Question: What is the net rate per unit area (n/cm2 -s) at which neutrons whose energies are
between 1 eV and 1 keV are crossing the surface perpendicular to en , going from the en side to
the +en side, taking into account all possible directions?

Answer: The question asks us about neutrons in a finite energy range, which means we must
integrate over energy:
1keV
n
net = dE d (~r, E, , t) en
cm2 -s 1eV 4

 1keV 
= ~
dE J(~r, E, t) en (5.52)
1eV

Question: What is the net rate per unit area (n/cm2 -s) at which neutrons cross the surface
perpendicular to en , going from the en side to the +en side, taking into account all possible
directions and energies?

Answer: The question asks us to integrate over the complete range of possible neutron energies:

n
net = dE d (~r, E, , t) en
cm2 -s 0 4

~ r, t) en
= J(~ (5.53)

Note that J~ without the E argument is the total net current density, which is J(~
~ r, E, t) inte-
grated over all energies.

120
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Question: What if we want to count only the neutrons that cross from the en side to the +en
side of the surface? How would we calculate such a one-way crossing rate?

Answer: This question asks about a one-way flow rate, which means we consider only half of
the possible directions of neutron travel. That is, we integrate over a particular 2 steradians that
contain the directions of interest, instead of all 4 steradians. If we want the rate per unit area at
which neutrons cross and end up on the +en side of the surface, we integrate Eq. (5.50) over all of
the directions that satisfy en > 0:

n
one-way = d dE(~r, E, , t) en (5.54)
cm2 -s 2:en >0 0

Note that the direction integral has become a half-range integral.

Question: What if we want to count only the neutrons that cross from the +en side to the en
side of the surface? How would we calculate this particular one-way crossing rate?

Answer: Again, we integrate over all of the directions of interest. If we want the rate per unit area
at which neutrons cross and end up on the en side of the surface, we integrate Eq. (5.50) over all
of the directions that satisfy en < 0, or to put it another way, those that satisfy (en ) > 0:

n
one-way = d dE(~r, E, , t) (en )
cm2 -s 2:(en )>0 0
 
= d dE(~r, E, , t) en
2:(en )<0 0

= d dE(~r, E, , t) en (5.55)

2:(en )<0 0

Note that the direction integral is a half-range integral. Note that the answer must be non-
negativethis is easiest to see in the last line of the equation, where we have expressed the dot-
product term as an absolute value.

121
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

PHYSICAL DEFINITION OF PARTIAL CURRENT DENSITIES

n
Jn+ (r, E, t)dE d2 r = expected that cross surface area d2 r at ~r, whose normal is en ,
cm2 -s
crossing from the en side to the + en side, counting only
those neutrons whose energies E are in the interval dE at E. (5.56)

n
Jn (r, E, t)dE d2 r = expected that cross surface area d2 r at ~r, whose normal is en ,
cm2 -s
crossing from the + en side to the en side, counting only
those neutrons whose energies E are in the interval dE at E. (5.57)

The subscript n reminds us that these partial-current densities are for a surface perpendicular
to the unit vector en .

MATH DEFINITION OF PARTIAL CURRENT DENSITIES


Jn (~r, E, t) = d en (~r, E, , t) (5.58)

2:(en )0
show units on this line

Note that these are densities in both spatial area and energy. To get a number of neutrons that
cross an area, one must integrate over spatial area and integrate over energy.

122
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Common examples (pay attention to signs and integration limits):



Jz+ (~r, E, t) = d |z | (~r, E, , t)
2:z >0
2 1
= d d || (~r, E, , , t) (5.59)
0 0

2 1
= d d (~r, E, , , t) (5.60)
0 0


Jz (~r, E, t) = d |z | (~r, E, , t)
2:z <0
2 0
= d d || (~r, E, , , t) (5.61)
0 1

2 0
= d d (~r, E, , , t) (5.62)
0 1


Jy+ (~r, E, t) = d |y | (~r, E, , t)
2:y >0

1
1/2
= d d 1 2 sin (~r, E, , , t) (5.63)

0 1

1 1/2
= d d 1 2 sin (~r, E, , , t) (5.64)
0 1


Jy (~r, E, t) = d |y | (~r, E, , t)
2:y <0

2 1
1/2
= d d 1 2 sin (~r, E, , , t) (5.65)

1

2 1 1/2
= d d 1 2 sin (~r, E, , , t) (5.66)
1

123
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Note:

1. A partial current density is always a non-negative scalar. Note the absolute value in the
integral in Eq. (5.59).

2. In Staceys book there is no absolute value, and his partial currents can be negative. This is
non-standard terminology and we will not use it. Remember Note 1!

3. Partial currents are defined relative to some direction (such as +ex or ez ), so we often say
partial current in the +x direction or partial current in the z direction. But remember:
any partial current is due to neutrons moving in a wide range of directionshalf of all possible
directionsnot just those moving in a single direction.

Example: Suppose we have figured out that the angular flux is linearly isotropic and we express
it as follows:

(~r, E, , t) = a(~r, E, t) + bx (~r, E, t)x + by (~r, E, t)y + bz (~r, E, t)z

What is the partial current density in the z direction?



Answer: We must multiply the angular flux by ez and integrate over the appropriate half-

range of directions:


Jz (~r, E, t) = d |z | (~r, E, , t)
2:z <0
2 0
= d d || (~r, E, , , t)
0 1
2 0
= d d (~r, E, , , t)
0 1

Now we insert what we know about the angular flux and perform the integrations:
2 0
Jz (~r, E, t) = d d [a(~r, E, t) + bx (~r, E, t)x + by (~r, E, t)y + bz (~r, E, t)z ]
0 1

2 0
= a(~r, E, t) d d
0 1
2 0 h i
1/2
bx (~r, E, t) d d 1 2 cos
0 1
2 0 h i
1/2
by (~r, E, t) d d 1 2 sin
0 1
2 0
bz (~r, E, t) d d [ ]
0 1

124
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

or
  
1
Jz (~r, E, t) = a(~r, E, t) (2)
2
bx (~r, E, t) [0]
by (~r, E, t) [0]
  
1
bz (~r, E, t) (2)
3

 
2
= a(~r, E, t) bz (~r, E, t)
3

5.7.2 Connection Between Partial and Net Current Densities

Recall the math definition of partial current density, and observe the math associated with the
difference of the + and partial current densities relative to some surface:


Jn+ (~r, E, t) Jn (~r, E, t) = d en (~r, E, , t)

2:(en )>0

d en (~r, E, , t)

2:(en )<0

= d en (~r, E, , t)
2:(en )>0

+ d en (~r, E, , t)
2:(en )<0

= d en (~r, E, , t)
4

 
= d (~r, E, , t) en
4

~ r, E, t) en
= J(~ (5.67)

This is as it should be: the net flow rate density across a surface can be expressed either in terms
of the net current density or as the difference between two partial current densities.

125
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

5.8 Production and Loss Rates in Terms of Angular Flux

We have seen that reaction rates can be computed in terms of the scalar flux, (~r, E, t), and
macroscopic cross sections, x (~r, E, t) that have been averaged over nucleus motion. We have also
seen that the scalar flux can be expressed as a direction integral of the angular flux. Further, we
have seen that both one-way and net surface-cross rates can be expressed in terms of the angular
flux.

That is, we can calculate any reaction rate and any surface-crossing rate if we know the angular
flux, (~r, E, , t). Here are expressions for four high-level gain and loss rates of neutrons in a
reactor that is operating in steady state:


3
Prod. rate from fission = d r dE (~r, E)f (~r, E) d (~r, E, ) (5.68)
volume 0 4


3
Absorption rate = d r dE a (~r, E) d (~r, E, ) (5.69)
volume 0 4


Outleakage rate = d2 r dE d en (~r) (~r, E, ) (5.70)

surface 0 2:en (~
r)>0


Inleakage rate = d2 r dE d en (~r) (~r, E, ) (5.71)

surface 0 2:en (~
r)<0

We can also use the angular flux to calculate much more detailed reaction rates. For example,
suppose we want to know the rate at which neutrons in the spatial volume {x (x1 , x2 ), y (y1 , y2 ),
z (z1 , z2 )} scatter from energies between 20 keV and 30 keV to energies between 100 eV and 500
eV, counting only those whose pre-scatter directions were upward and post-scatter directions were
downward. We would have:
30keV 500eV  2 1
rate = d3 r dEi dEf di di
volume 20keV 100eV 0 0

2 0 
df df s (~r, Ei Ef , i f ) (~r, E, i ) (5.72)
0 1

126
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

5.9 Summary

A key goal in this course is to understand how reactors behave and why they behave that way.
This requires, at the very least, an understanding of neutron production and loss rates in reactors.

In this chapter we discussed neutron-nucleus reaction rates (RRs). We found that


reaction rate [units = n/s] = d3 r (reaction rate density at ~r)
volume

 
reactions
RR density at ~r units =
cm3 .s

= dE (energy-dependent RR density at ~r and E)
0

 
reactions
energy-dependent RR density at ~r, E units = = x (~r, E)(~r, E)
cm3 .s.MeV

where
(~r, E) = energy-dependent scalar flux [n/(cm2 -s-MeV)] = n(~r, E) v(E) ,
n(~r, E) = energy-dependent neutron density [n/(cm3 -MeV)] ,
x = macroscopic cross section for reaction of type x [cm1 ], and
v(E) = speed of neutron whose kinetic energy is E.

Note that

x is the sum of the macroscopic cross sections for all of the nuclides.

x for a given nuclide is N x for that nuclide.

x for a given nuclide and a given neutron lab-frame energy must be pre-averaged over nucleus
motion in order for the expressions above to give the correct rates.

127
NUEN 601, M.L. Adams notes, 2017 Chapter 5. Flux, Current, Etc.

Production of neutrons in reactors is mainly by fission. The production rate due to fission, by the
above equations, is as follows in a steady-state reactor:

hni
3
production rate = d r dE (~r, E)f (~r, E)(~r, E)
s volume 0

where (~r, E) is the average number of neutrons emitted from a fission caused by a neutron of
energy E given the material at position ~r. [If the reactor is not in steady state, then the production
rate from fission will include separate terms for prompt neutrons and delayed neutrons.]
Loss of neutrons in reactors is by absorption and leakage. By the above equations,

hni
3
absorption rate = d r dE a (~r, E)(~r, E)
s volume 0

The other neutron-loss mechanism in reactors is leakage, which we also discussed in this chapter.
We found that

hni
net leakage rate outward through surface S in is: ~ rs )
d2 ren (~rs ) J(~
s S

where en (~rs ) = outward unit normal at position ~rs on the surface,

~ r) =
J(~ total net current density

= ~ r, E)
dE J(~
0

~ r, E) = energy-dependent net current density [n/(cm2 -s-MeV)] .


J(~

The net loss rate due to leakage is just the net outleakage rate (or net leakage rate outward, or
outleakage rate minus inleakage rateall the same thing).

Thus, if we know the energy-dependent scalar flux, the energy-dependent net current density, and
all macroscopic cross sections (properly averaged over nucleus motion), we can compute neutron
production and loss rates in a reactor.

We also found in this chapter that if we know the energy-dependent angular flux, (~r, E, , t), we
can calculate the scalar flux, the net current density, and any desired partial current density.

128
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

Chapter 6

The Neutron Transport Equation

6.1 Introduction

We know that a reactors behavior depends on the gain and loss rates of neutrons in the reactor.
We also know that some gain and loss is due to neutron-nucleus reactions, and that some is due to
neutron leakage. In the previous chapter we developed mathematical expressions for reaction rates
and leakage rates. In this chapter we put these expressions into statements of conservation to see
what we can learn about how the neutron population changes with time in reactors and how the
neutrons distribute themselves in position and energy.

It turns out that if we write down a detailed mathematical statement of conservation of free neu-
trons, we will find that this statement is

an integro-differential equationthe neutron transport equation.

We will further find that this equation has

a unique solution.

Think about this! It says that there is only one possible neutron distribution that can satisfy
conservation (change rate = gain rate loss rate) everywhere in the phase-space volume!

The unique solution is the angular flux, (~r, E, , t), which tells us everything that the neutrons
are doing throughout the reactor. If we can solve the equation, thereby obtaining the angular flux,
we can use it to compute any gain rate or loss rate that we want.

In this chapter we derive the neutron transport equation from a detailed statement of conservation
of free neutrons. We then briefly explore properties of its solutions in problems of interest to nuclear
engineering, setting the stage for more detailed studies in forthcoming chapters.

129
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.2 Mechanisms for Gaining and Losing Free Neutrons

Our conservation statement will take the form

Change rate = Gain rate Loss Rate .

We will write each term as a mathematical expression. The change rate will simply be the time
derivative of the population in question. The gain rate will be the sum of several terms, each one
mathematically describing the rate due to a different gain mechanism. The loss rate will also be the
sum of several terms, each associated with a different loss mechanism. Our first step is identifying
the mechanisms by which neutrons can be gained and lost in any phase-space volume of interest.

6.2.1 Gain and Loss Mechanisms to be Included

Here we identify the most significant mechanisms by which neutrons can be gained and lost in any
phase-space subvolume of interest in a typical nuclear reactor. As we have seen previously, the
relevant phase space is six-dimensional, with the following independent variables typically used as
the coordinates in the phase space:

Spatial Position: 3 independent variables

Cartesian: x, y, z
Spherical: r, spatial , spatial
Cylindrical: r, spatial , z

Energy: 1 independent variable, E = lab-frame neutron kinetic energy

Direction: 2 independent variables, = cosine of polar angle and = azimuthal angle

A phase-space subvolume is a six-dimensional subvolume that can be described by

6 V = six-dimensional phase-space subvolume

= 3 rE , (6.1)

where

3 r the spatial subvolume. (6.2)

What are the significant mechanisms by which free neutrons (meaning those that are not bound in
nuclei) can be gained in some arbitrary subvolume, 6 V , of phase space in a reactor? Let us list
them.

130
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

SIGNIFICANT GAIN MECHANISMS

1. Emission from fixed sources (those that are not triggered by neutron interactions,
but spontaneously emit neutrons), such as Americium-Beryllium (AmBe), Plutonium-
Beryllium (PuBe), Polonium-Beryllium (PoBe), Californium-252 (spontaneous fission),
etc.

2. Prompt emission of neutrons from neutron-induced fission

3. Delayed emission of neutrons originating from neutron-induced fission

4. In-scattering,
in which a neutron in 3 r but not in E scatters off of some nucleus and emerges in
E

5. In-leakage,
in which a neutron in E enters 3 r through its surface

Similarly, let us list the significant mechanisms by which free neutrons (meaning those that are not
bound in nuclei) can be lost in some arbitrary subvolume, 6 V , of phase space in a reactor.

SIGNIFICANT LOSS MECHANISMS

1. Absorption

2. Out-scattering,
in which a neutron in 3 r and E scatters off of some nucleus and emerges outside
of E

3. Out-leakage,
in which a neutron in E and 3 r leaves 3 r through its surface

131
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.2.2 Gain and Loss Mechanisms to be Neglected

The gain and loss mechanisms listed above are incompletewe have ignored some mechanisms
because our experience and analyses show that they are not usually significant. Some mechanisms
we have ignored are:

Loss from neutron decay (T1/2 10.2 minutes )

Gain and loss from neutrons scattering off of other neutrons

Moving into or out of E because of gravitational force

In what follows we shall also ignore the following complicating factors in our mathematical models
of the gain and loss mechanisms that we are explicitly considering:

Anisotropic nucleus motion (which would cause our averaged cross sections to depend on
neutron direction)

Non-local interactions with matter (as with very slow neutrons that have long wavelengths).
We can include diffraction effects, but we approximate every scattering event as happening
at a single spatial point.

Multiple neutrons interacting simultaneously with a single nucleus.

Experience and detailed analyses show that these simplifications do not introduce substantial error
in typical reactor-related calculations. Nevertheless, it is wise to remember that the mathematical
models we employ, including the neutron transport equation that we derive in this chapter, contain
approximations and omissions compared to the full reality of nature.

6.2.3 Delayed-neutron physics and math

Let us review the processes that lead to emission of delayed neutrons. Consider the decay scheme
for Bromine-87, as shown in Fig. 6.1. The 87 Br atoms that eventually decay to metastable
87 Kr (and thus yield a neutron) are called

delayed-neutron precursors

(These are only 2.6% of the 87 Br atoms.) Notice that the precursor does not actually expel a
neutron. Instead, it decays to an excited state of another nuclide, which then emits the neutron.
Also note that the precursor does not have to be a direct fission product, but may simply be in the
decay chain of a direct fission product. One way to produce Bromine-87, for example, is by:

fission 87 As 87 Se 87 Br.

The delayed neutron in question is delayed by the sum of the decay times of the precursor and its
parents, in the whole chain all the way back to the fission event whose fission product is the start
of the chain.

132
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

Br-87 (55 sec)

Kr*-87
Kr-86 + neutron

Kr-87

Figure 6.1: Decay scheme for 87 Br, 2.6% of the atoms of which are delayed-neutron precursors.
(Show 2.6% and 97.4%.)

Approximately 270 delayed-neutron precursors have been identified to date. Of these, only a few
dozen are of much practical significance. These can be grouped into precursor types according
to the average delay between the originating fission event and the ultimate neutron emission. That
is, if two precursors emit neutrons with approximately the same delay, we can lump them together
as one type of precursor in our models.
We seek an equation or set of equations that will give us Sdn (~r, E, t), the number of delayed neutrons
emitted per unit spatial volume at position ~r and per unit energy at energy E at time t. Let us
define some helpful quantities:

Ci (~r, t) = expected density of type-i precursors at position ~r at time t.


i (t) = decay constant of type-i precursors.
i (E) = energy spectrum of neutrons emitted after decay of type-i precursors.

The definition of i means that i (E)dE is the fraction of type-i delayed neutrons that are emitted
with energies in dE at E.

From these definitions it follows that


# of d.n. precursor
X types
Sdn (~r, E, t) = i (E)i Ci (~r, t) (6.3)
i=1

This is fine, but it doesnt help much unless we know Ci (~r, t) for each precursor type, i. How do
we find these precursor densities?
As usual, we turn to our favorite equation:

Change rate = gain rate loss rate.

133
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

For what population of things should we write this equation?

population = type-i precursors in spatial volume 3 r

We already know the loss rate from decay of the precursors: it is just i times the precursor
population. (There is another loss mechanism: absorption of a neutron. But these precursors have
such short half-lives that they are far more likely to decay than to absorb a neutron, so we ignore
their destruction by neutron absorption.) But what is the gain rate?

Note that each precursor had its ultimate origin in fission. That is, each precursor is either a fission
product or originated from the decay of a fission product. So we know that the production rate
density of precursors will be tied to

the fission-rate density

Also note that

each precursor ultimately produces exactly one delayed neutron

This follows from our definition of precursor. It is useful and customary to define

di (~r, E, t) expected number of type-i precursors produced from


a fission that is caused by a neutron of energy E
at position ~r at time t . (6.4)

Because each precursor ultimately produces exactly one delayed neutron, di is also the expected
number of delayed neutrons that will ultimately be emitted from type-i precursors, per fission
caused by a neutron of energy E.

With these definitions we can now write our conservation equation for precursors of type i:


d
Ci (~r, t) = dE di (~r, E, t)f (~r, E, t)(~r, E, t) i Ci (~r, t) . (6.5)
dt 0

6.3 Derivation of the Neutron Transport Equation

Here we consider the expected number of neutrons in some arbitrary phase-space subvolume, 6 V
3 rE, in a nuclear reactor. For this number we write the conservations statement change rate
= gain rate loss rate. We then mathematically express each term in this conservation equation.
This produces an equationthe neutron transport equationwhose solution (the angular flux, )
tells us all we need to know about neutron behavior in reactors. The main catch is that the
equation is difficult to solve.

134
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.3.1 Preliminaries

Many of our mathematical expressions for gain and loss rates must be expressed in terms of the
neutron distribution in position, energy, and direction. We recall the definition of the direction-
and energy-dependent neutron density:

n(~r, E, , t)d3 r dE d expected number of neutrons in the volume d3 r at ~r,


with lab-frame kinetic energy in the interval dE at E,
and with direction in the solid-angle cone d around the direction .

This is a density in a six-dimension phase space: 3 spatial dimensions, 1 energy dimension, and 2
direction dimensions. (It takes two numberssay a polar angle and an azimuthal angleto define
a direction.) The units are neutrons per unit volume per unit energy per unit solid angle. The
usual unit of solid angle is the steradian. So in our usual units we have n/(cm3 -MeV-ster).
It follows that

3
d r dE d n(~r, E, , t) = expected number of neutrons
3 r E
in the phase-space subvolume 6 V at time t . (6.6)

This is the population for which we will write our conservation equation. We can now write a
mathematical expression for the change rate of this population:
 
d 3
change rate d r dE d n(~r, E, , t) (6.7)
dt 3 r E

Because our phase-space subvolume, 6 V , does not change with time, we can rewrite this term as
follows:

n(~r, E, , t)
change rate d3 r dE d (6.8)
3 r E t

We recall the definition of the energy-dependent angular flux:

(~r, E, , t) v(E)n(~r, E, , t) , (6.9)

Here v(E) the lab-frame speed of a neutron whose lab-frame kinetic energy is E.

We can rewrite our change-rate term as follows, if we so choose


3 1 (~r, E, , t)
change rate d r dE d (6.10)
3 r E v(E) t

135
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.3.2 Math descriptions of gain rates

Recall that we have identified the following significant gain mechanisms:

Emission from fixed sources, if any

Prompt emission from neutron-induced fission

Delayed emission from neutron-induced fission

In-scattering from energies outside the energy sub-interval E and/or directions outside
the directional cone to energies in the sub-interval and directions in the cone.

In-leakage through the spatial sub-volume surface


We consider each of these in turn and express it mathematically. We begin by defining the fixed
source rate density:

Sfixed (~r, E, , t)d3 r dE d expected rate at which neutrons are emitted


in the volume d3 r at ~r,
with lab-frame kinetic energy in the interval dE at E,
and with direction in the solid-angle cone
d around the direction .
Then the expected rate at which we gain neutrons in the phase-space subvolume 6 V must be the
integral of this source rate density over that subvolume:


3
gain rate from fixed sources d r dE d Sfixed (~r, E, , t) (6.11)
3 r E

We turn to the prompt emission rate of neutrons from fissions that are induced by neutron absorp-
tion. Recall our previous definition of :

(~r, E, t) expected number of neutrons emitted per fission that is


induced by absorption of a neutron of energy E
at position ~r at time t. (6.12)

We break into the sum of two terms, one for promptly emitted neutrons and one for those that
emerge (from precursor decay) some time after the fission event:

(~r, E, t) p (~r, E, t) + d (~r, E, t) (6.13)

# of dnp
X types
= p (~r, E, t) + di (~r, E, t) (6.14)
i=1

136
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

We can now express the gain rate density for neutrons emitted promptly from neutron-induced
fission:

1
prompt-neutron gain rate density = p (E) dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t) (6.15)
4 0
show units on this line

Note that for this to be correct, f must have been averaged over nucleus velocities and the nucleus
velocity distribution must be isotropic. Also note that prompt neutrons are born with a different
energy distribution than are delayed neutrons, and we specify the prompt-neutron spectrum as
p (E).

So the expected rate at which neutrons are emitted promptly from neutron-induced fission into the
phase-space subvolume 6 V must be the integral of this emission rate density over that subvolume:

prompt-n

3 p (E)
gain rate d r dE d dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t)
3 r E 4 0
point out the connection among the Ei0 s (6.16)

The expected rate at which we gain delayed neutrons in the phase-space subvolume 6 V must be
the integral of the delayed-neutron source rate density over that subvolume:


3 1
delayed-n gain rate d r dE d Sdn (~r, E, , t)
3 r E 4

# of dnp
X types
3 1
= d r dE d i (E)i Ci (~r, t) (6.17)
3 r E 4
i=1

We turn to the in-scattering term. The in-scattering rate density at the phase-space point
(~r, E, ), with no restrictions on initial energy or direction, is:


in-scat RD(~r, E, , t) = dEi di s (~r, Ei E, i , t)(~r, Ei , i , t) (6.18)
0 4
show units on this line; connect initial variables; connect final variables

137
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

The expected rate at which neutrons are scattered from outside the phase-space subvolume 6 V
to inside it is the total rate at which scattered neutrons appear in the subvolume minus the within-
subvolume scattering rate. The latter is the rate at which neutrons with initial energies and
directions in E and scatter and have final directions that are also in E and . So we
write the in-scattering rate as the difference of two terms:


3 f ake.to.make
in-scat rate d r dE d
3 r E f ence.high


dEi di s (~r, Ei E, i , t)(~r, Ei , i , t)
0 4


dEi di s (~r, Ei E, i , t)(~r, Ei , i , t) (6.19)
Ei E i

Our final gain term is from in-leakage. Unlike previous terms we have seen, this one involves an
integral over the incoming surface of the spatial sub-volume, 3 r:

 
in-leakage rate d2 r dE d en (~r) (~r, E, , t) (6.20)
3 r E : en (~
r)<0
show units on this line; show that minus = | |

Here we have used the common mathematical notation V to denote the surface of a volume V .

6.3.3 Math descriptions of loss rates

Our loss mechanisms are:

Absorption

Out-scattering from energies in the sub-interval and directions in the cone. to energies
outside the energy sub-interval E and/or directions outside the directional cone

Out-leakage through the sub-volume surface

138
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

We consider each of these in turn and express it mathematically. We begin with the absorption
rate, which is a phase-space volume integral of the absorption rate density:


absorption rate d3 r dE d a (~r, E, t)(~r, E, , t) (6.21)
3 r E

The expression for the expected rate at which neutrons are scattered from inside the phase-space
subvolume 6 V to outside it is developed similarly to the in-scattering rate:


3 f ake.to.make
out-scat rate d r dE d
3 r E f ence.high


dEf df s (~r, E Ef , f , t)(~r, E, , t)
0 4

)
dEf df s (~r, E Ef , f , t)(~r, E, , t) (6.22)
Ef E f

The expression for the expected rate at which neutrons leak out of the spatial part of the phase-
space sub-volume is similar to that for in-leakage:

 
out-leakage rate d2 r dE d en (~r) (~r, E, , t) (6.23)
3 r E : en (~
r)>0
show units on this line;

139
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.3.4 Manipulations

Before we assemble our conservation statement we arrange certain terms into more convenient
forms. Note that gain rate loss rate will include (in-leakage rate out-leakage rate):

in-leakage rate out-leakage rate


 
2
d r dE d en (~r) (~r, E, , t)
3 r E : en (~
r)<0
 
d2 r dE d en (~r) (~r, E, , t) (6.24)
3 r E : en (~
r)>0

Both terms have minus signs. In one term the integral is restricted to incoming directions in
and in the other it is outgoing. Their sum is the integral over all directions in :

in-leakage rate out-leakage rate


 
d2 r dE d en (~r) (~r, E, , t) (6.25)
3 r E

Recall Gausss divergence theorem:



d2 r en (~r) ~u(~r) = ~ ~u(~r),
d3 r (6.26)
V V

where V is the surface that encloses volume V .

We can apply this to the leakage term if we identify as the vector function:


2
d r en (~r) (~r, E, , t) = ~ (~r, E, , t)
d3 r
3 r 3 r


= ~ r, E, , t)
d3 r (~ (6.27)
3 r

It follows that


in-leakage rate out-leakage rate = 3
d r dE ~ r, E, , t)
d (~
3 r E
(6.28)

140
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

The gain rate loss rate will also include (in-scattering rate out-scattering rate). Subtract our
previous expressions for these, and the within-subvolume scattering terms will cancel:

in-scat rate out-scat rate



f ake.to.make
= d3 r dE d
3 r E f ence.high


dEi di s (~r, Ei E, i , t)(~r, Ei , i , t)
0 4


dEi di s (~r, Ei E, i , t)(~r, Ei , i , t)
Ei E i


dEf df s (~r, E Ef , f , t)(~r, E, , t)
0 4

)
+ dEf df s (~r, E Ef , f , t)(~r, E, , t)
Ef E f

After canceling the within-subvolume scattering terms we have:

in-scat rate out-scat rate



f ake.to.make
= d3 r dE d
3 r E f ence.high


dEi di s (~r, Ei E, i , t)(~r, Ei , i , t)
0 4


dEf df s (~r, E Ef , f , t)(~r, E, , t) (6.29)
0 4

141
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.3.5 Transport Equation

Now our conservation equation, [change rate = gain rate loss rate], for neutrons in 3 rE,
is

1 (~r, E, , t)
d3 r dE d
3 r E v(E) t


3
= d r dE d Sfixed (~r, E, , t)
3 r E


3 p (E)
+ d r dE d dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t)
3 r E 4 0

# of dnp
X types
3 1
+ d r dE d i (E)i Ci (~r, t)
3 r E 4
i=1


+ d3 r dE d dEi di s (~r, Ei E, i , t)(~r, Ei , i , t)
3 r E 0 4


3
d r dE d dEf df s (~r, E Ef , f , t)(~r, E, , t)
3 r E 0 4


3
d r dE d a (~r, E, t)(~r, E, , t)
3 r E


d3 r dE ~ r, E, , t) .
d (~ (6.30)
3 r E

We recognize that the integral over all final directions and energies of a double-differential scattering
cross section is just a scattering cross section:

dEf df s (~r, E Ef , f , t) s (~r, E, t) (6.31)
0 4

142
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

Then we add:

d3 r dE d a (~r, E, t)(~r, E, , t)
3 r E


3
+ d r dE d s (~r, E, t)(~r, E, , t)
3 r E


3
= d r dE d t (~r, E, t)(~r, E, , t) (6.32)
3 r E

We recognize that every term in Eq. (6.30 is an integral over 3 rE. We collect all the terms
into one integral:
(
3 1 (~r, E, , t) ~ r, E, , t)
d r dE d + (~
3 r E v(E) t

# of dnp
X types
1
+ t (~r, E, t)(~r, E, , t) Sfixed (~r, E, , t) i (E)i Ci (~r, t)
4
i=1


p (E)
dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t)
4 0


dEi di s (~r, Ei E, i , t)(~r, Ei , i , t) = 0 . (6.33)
0 4

This equation states that the integral of a certain collection of terms over some phase-space subvol-
ume, 3 rE, is zero. This is true no matter what subvolume we choosewe did not specify
anything special or unusual about the one we examined, and in fact let it be arbitrary. Now we
recognize:

If the integral of a function = 0 no matter what the range of integration, then

the function = 0 everywhere.

143
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

That is, the collection of terms inside the { } in Eq. (6.33) is identically zero. We rewrite this
statement and obtain the transport equation:

TIME-DEPENDENT NEUTRON TRANSPORT EQUATION

1 (~r, E, , t) ~ r, E, , t) + t (~r, E, t)(~r, E, , t)


+ (~
v(E) t

# of dnp
X types
1
= Sfixed (~r, E, , t) + i (E)i Ci (~r, t)
4
i=1


p (E)
+ dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t)
4 0


+ dEi di s (~r, Ei E, i , t)(~r, Ei , i , t) . (6.34)
0 4

The neutron transport equation is a single integro-differential equation for the single unknown
function . We can show mathematically that it has a solution and that this solution is unique,
given appropriate boundary and initial conditions. The transport equation is an essentially exact
description of the behavior of neutrons in reactors. It is a relatively complicated equation that is
relatively difficult to solve. Fortunately, we can gain a great deal of insight into reactor behavior
by studying equations (such as the neutron diffusion equation) that contain simplifying approxima-
tions. With modern computing and modern numerical techniques, we can also attack the transport
equation directly in many problems of interest and obtain sufficiently accurate solution to be of
high value.

6.4 Types of Transport Problems

There are three basic types of transport problems:

1. Time-dependent

2. Steady-state

3. Eigenvalue (with many sub-types)

144
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.4.1 Operator Notation

If we define a concise operator notation we will be able to write our various transport equations
easily and compactly:

p (E)
Fp dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t) (6.35)
4
0
(E)
F dEi (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t) (6.36)
4 0
~ r, E, , t) + t (~r, E, t)(~r, E, , t)
L (~

dEi di s (~r, Ei E, i , t)(~r, Ei , i , t) (6.37)
0 4

6.4.2 Time-dependent

The time-dependent neutron transport equation is given in Eq. (6.45). We can also write it in
operator notation:
1
+ L = Sf ixed + Sdn + Fp . (6.38)
v t
Here Sf ixed and Sdn are the fixed and delayed-neutron source-rate densities, respectively.

In a time-dependent problem the angular flux must be specified throughout the 6D phase space at
the start time. This is the initial condition:

(~r, E, , t0 ) = initial (~r, E, ) = known. (6.39)

Also, in a time-dependent problem the boundary conditions that define the incident angular flux on
the problem boundary (described in later subsections) must be specified as a function of time. If
fissioning nuclides are present, then the delayed-neutron precursor concentrations must be evolved in
time along with the angular flux, and their initial values must be specified (Ci (~r, t0 ) = Ci,initial (~r) =
known, i = 1, . . . number of delayed-neutron-precursor types).

If initial and boundary conditions are properly specified and physically realistic cross sections are
used, then the transport solution, , is unique.

We will discuss time dependence in more detail in a later chapter.

6.4.3 Steady state

In a steady-state neutron transport problem the time variable and time-derivative term are not
present. In addition, the neutron production term from neutron-induced fission is significantly
simplifieddelayed neutrons do not have to be explicitly treated in a steady-state problem. Note

145
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

that if the delayed-neutron emission rate density is steady (constant) in time, then it is the same
as if there were no delay. The steady-state equation follows.

STEADY-STATE NEUTRON TRANSPORT EQUATION

~ r, E, ) + t (~r, E)(~r, E, ) = Sfixed (~r, E, )


(~


(E)
+ dEi (~r, Ei )f (~r, Ei )(~r, Ei )
4 0


+ dEi di s (~r, Ei E, i )(~r, Ei , i ) . (6.40)
0 4

In operator notation this becomes

L = F + Sf ixed . (6.41)

It is possible, and in fact easy, to define a transport problem that has no steady-state solution.
A simple example is a supercritical reactor that contains a fixed source. However, the steady-
state neutron transport problem does have a unique solution for subcritical problems with properly
specified boundary conditions and physically realistic cross sections.

6.4.4 Eigenvalue

Eigenvalue problems can usually be written in the form:

Afn (~x) = n fn (~x) , (6.42)

where

A is an operator,

fn is the n-th eigenfunction or eigenvector of A,

n is the eigenvalue of A associated with fn ,

~x is a collection of coordinates that define a point in the space in which f is defined.

146
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.4.4.1 k-eigenvalue problems

In nuclear reactor analysis we are often concerned with the

k-eigenvalue problem,

which is written as follows.

k-EIGENVALUE NEUTRON TRANSPORT EQUATION

~ n (~r, E, ) + t (~r, E)n (~r, E, )



1 (E)
= dEi (~r, Ei )f (~r, Ei )n (~r, Ei )
kn 4 0


+ dEi di s (~r, Ei E, i )n (~r, Ei , i ) . (6.43)
0 4

In operator notation:
1
L = F L1 F = k (6.44)
k
A few remarks about eigenvalue problems:

Eigenvalue problems cannot contain driving terms (or forcing functions). They seek to
learn what the system being studied can do on its own. This means

there is never a fixed source in an eigenvalue problem

and never any neutrons introduced through the problem boundary.

A solution of an eigenvalue problem is an {eigenvalue, eigenfunction} pair. A solution of a


k-eigenvalue transport problem is a pair {k, }.

There are infinitely many solutions of a given neutron transport eigenvalue problem.

All of the k-eigenvalues are real and positive.

Only one k-eigenfunction is real and positive (or more precisely, does not change sign) through-
out the phase-space domain. We call this special eigenfunction

the fundamental mode

The eigenvalue associated with the fundamental-mode eigenfunction is the largest of the
k-eigenvalues. This largest k-eigenvalue is the multiplication factor of the reactor.

147
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.4.4.2 Alpha () eigenvalue problems

Another type of eigenvalue problem that is often of interest is the

-eigenvalue problem,

sometimes called by the more descriptive name time-absorption eigenvalue problem. This prob-
lem is written as follows.

-EIGENVALUE NEUTRON TRANSPORT EQUATION

 
~
n (~r, E, ) + t (~r, E) + n (~r, E, )
v(E)


(E)
= dEi (~r, Ei )f (~r, Ei )n (~r, Ei )
4 0


+ dEi di s (~r, Ei E, i )n (~r, Ei , i ) . (6.45)
0 4

In operator notation:

= v [F L] (6.46)

148
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

A few remarks about -eigenvalue problems:

As with all eigenvalue problems, -eigenvalue problems cannot contain driving terms (or
forcing functions). They seek to learn what the system being studied can do on its own.
There is never a fixed source in an eigenvalue problem and never any neutrons being intro-
duced through the problem boundary.

A solution of an -eigenvalue problem is a pair {, }.

There are infinitely many solution pairs that satisfy a given neutron transport eigenvalue
problem.

Only one -eigenfunction is real and positive (or more precisely, does not change sign)
throughout the phase-space domain. We call this special eigenfunction

the fundamental mode of the -eigenvalue problem

The -eigenvalue associated with the fundamental--mode eigenfunction is the algebraically


largest (not the largest in magnitude) of the -eigenvalues. We shall call this max .

If the system is supercritical, then at least one -eigenvalue, max , is real and positive. If the
system is just critical, the algebraically largest = max = 0. If the system is subcritical,
the algebraically largest = max is real and negative.

If neutrons are introduced into a system whose cross sections and geometry remain fixed
in time, then eventually the neutron population will take the shape of the fundamental
-mode and its amplitude will vary as et .
To see this, substitute (~r, E, , t) = et f (~r, E, ) into the time-dependent transport
equation, carry out the time derivative, and cancel the common et term in the resulting
equation. You should obtain the -eigenvalue equation.

The fundamental mode of the problem is, in general, different from the fundamental mode
of the k problem. The two are identical if the system is just critical. They are also identical
if the energy-dependent transport equation is replaced by a one-speed or one-group diffusion
model.

6.5 Initial and Boundary Conditions for the Transport Equation

A well-posed problem involving differential or integro-differential equations must include proper


boundary and initial conditions in addition to the equations themselves.

149
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.5.1 Transport Initial Conditions

The initial condition that makes a time-dependent transport problem well-posed is just what your
common sense would tell you: You must know the neutron distribution at the initial time if you
expect to know it at later times. The mathematical statement of this is:

(~r, E, , t0 ) = initial (~r, E, ) = known. (6.47)

If fissioning material is present, then you also must know the initial delayed-neutron-precursor
concentrations:

Ci (~r, t0 ) = Ci,initial (~r) = known. (6.48)

Neither steady-state nor eigenvalue problems need initial conditions, for they have no time variable.

6.5.2 Transport Boundary Conditions

There are several kinds of boundary conditions that make a transport problem well-posed. They all
do just what your common sense would tell you that they must do: they specify the distribution
of neutrons that come into the spatial domain through its boundary. There are several ways to do
this, many of which we outline below.

For time-dependent problems the boundary conditions must be specified as a function of time, so
in what follows we include t as an argument. For steady-state and eigenvalue problems there is no
time variable, so for such problems the t argument should be omitted.

6.5.2.1 Explicitly Specified Incident Flux

The most obvious way to specify what is coming into the spatial domain through its boundary is
to spell it out as an explicit, known function for all positions on the boundary. Let the spatial
domain be given by V and its boundary by V . Let the outward unit normal at position ~r on the
boundary be given by en (~r). Then the specified-incident boundary condition is:

(~r, E, , t) = inc (~r, E, , t) = known for all : en (~r) < 0, for all ~r Vinc , (6.49)

where Vinc is the portion of the boundary on which the specified-incident boundary condition is
imposed.

150
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.5.2.2 Specular Reflection (Mirror) Boundary Condition

If a problem has mirror-image symmetry along a plane, then we have the option of solving the
transport equation on only one side of the plane and imposing a mirror boundary condition at
the plane. This condition specifies that the angular flux that enters the problem on the plane
of symmetry is a specular reflection of the angular flux that exits the problem on that plane.
Mathematically:

(~r, E, , t) = (~r, E, refl , t) for all : en (~r) < 0, for all ~r Vmir , (6.50)

where

refl the direction that reflects off of Vmir onto , (6.51)

and where Vmir is the portion of the boundary on which the mirror boundary condition is imposed.

Note that there may be more than one plane of symmetry in a given problem. Note also that:

the net flow across a reflecting boundary is zero,

which means that the normal component of the net current density is zero on a reflecting boundary:
~ r, E, t) = 0
en (~r) J(~ for ~r on a reflecting boundary. (6.52)

6.5.2.3 Diffuse Reflection (White) Boundary Condition

Sometimes we wish to model an interface as having no net flow across it, as is the case with specular
reflection, but with the incident neutrons distributed isotropically across the incoming half of the
directions. This is called diffuse reflection, sometimes known as a white boundary condition.
Mathematically it can be stated as follows:
1 +
(~r, E, , t) = J (~r, E, t) for all : en (~r) < 0, for all ~r Vwhite , (6.53)
n
where

Jn+ (~r, E, t) d0 en (~r) 0 (~r, E, 0 , t) for all ~r Vwhite , (6.54)
0 :en (~
r)0 >0

where Vwhite is the portion of the boundary on which the white boundary condition is imposed.

Note that

the net flow across a white boundary is zero,

which means that the normal component of the net current density is zero on a reflecting boundary:
~ r, E, t) = 0
en (~r) J(~ for ~r on a white boundary. (6.55)

151
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.5.2.4 Periodic Boundary Condition

If a problem has an infinite repeating structure, we can solve it on the smallest repeating unit and
impose periodic boundary conditions. An example of a mathematical expression of a periodic
boundary condition is:
~ r), E, , t)
(~r, E, , t) = (~r + d(~ for all : en (~r) < 0, for all ~r Vper , (6.56)

where

d~ a translation vector whose magnitude = one dimension of the repeating unit, (6.57)

and Vper is the portion of the boundary on which the periodic boundary is imposed.

More complicated periodic boundary conditions are also possible, but the concept is the samethe
incident angular flux on one surface is defined to be the exiting angular flux from a different surface.

6.5.2.5 Albedo Boundary Condition

Sometimes we wish to solve a transport problem on a domain that is adjacent to a second domain.
The second domain affects the first, in general, because if neutrons leave the first domain they may
scatter in the second, thereby changing directions and energies and possibly re-entering the first
domain. An obvious example is a reflector next to a reactor core. The most accurate treatment
of this situation would be to solve the transport equation on the union of the two domains, but
sometimes this is more difficult than we would like. An alternative is to treat the second domain
as a kind of reflector, but one that in general reflects only a fraction of the neutrons that enter it.
The albedo boundary condition accomplishes this. An example:

(~r, E, , t) = d0 (~r, E, 0 , t)(~r, E, 0 , t) for all ~r Valbedo , (6.58)
0 :en (~
r)0 >0

where Valbedo is the portion of the boundary on which the albedo boundary condition is imposed.
Here

(~r, E, 0 , t) = the albedo function at ~r. (6.59)

In this example, has units of

inverse steradians.
Physically,

(~r, E, 0 , t)d = the fraction of neutrons exiting in direction 0


that re-enter the problem with directions
in the cone d around . (6.60)

152
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

More complicated albedo functions can be applied. In particular, note that physically realistic
albedos usually must specify the fraction of neutrons, exiting with energy E 0 and direction 0 , that
re-enter the problem in the interval dE around E and the cone d around . In many cases the
spatial points on the boundary are also coupledneutrons that leave at position r~0 may re-enter
at ~r.
Note that the mirror and white boundary conditions are special cases of the albedo condition:
mirror
(~r, E, 0 , t) ( 0 refl )(0 refl ) (6.61)
white en (~r) 0
(~r, E, 0 , t) (6.62)

Here is the Dirac delta-function, which is described in many textbooks and in our special-functions
appendix.

6.6 Other Neutron-Producing Reactions

So far we have included gain terms from only two kinds of neutron-nucleus reactions from which
neutrons emerge:

1. scattering

2. fission
There are many other reaction types from which neutrons emerge. All of these contribute to the
gain rate density in reality, so all that are significant in a given problem should be included on the
right-hand side of our transport equation. (We have already included the loss of the neutron
that initiates any type of reaction on the left-hand side, because all reaction types are included in
t . We dont need to change anything there.)

Consider the (n,2n) reaction. As we have seen with scattering and fission, when we write an
expression for the gain-rate density from (n,2n) reactions we will need an expression for how the
emerging neutrons are distributed in energy and direction. Let us express this with the following
probability density function:

Pn,2n (~r, Ei E, i , t)dEd probability that one of the emerging neutrons


emerges with energy in dE at E and direction in d at ,
given that the (n,2n) reaction was caused at position ~r
and time t by a neutron
with energy Ei moving in direction i . (6.63)

Remark: The Pn,2n function plays the same role that (E)/4 plays for the neutrons emerging
from fission. Note that if you integrate either function over all E and , the answer is 1.
Question: What plays the role in the (n,2n) gain term that plays in the fission term?

Answer: The number 2.

153
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

Given these observations and definitions it is not difficult to write down the gain rate density from
(n,2n) reactions:

gain rate density



from (n,2n) = dEi di 2n,2n (~r, Ei , t)Pn,2n (~r, Ei E, i , t)(~r, Ei , i , t)
0 4
(6.64)

We can similarly define gain-rate densities for other reactions, such as (n,3n), (n,4n), (n,p+n),
(n,d+n), (n,t+n), (n,+n), etc., where p, d, and t represent proton, deuteron, and
triton, respectively. We can then collect all of these, along with the differential scattering cross
section, into a single transfer cross section as follows:

X (~r, Ei E, i , t) s (~r, Ei E, i , t)
+ 2 n,2n (~r, Ei , t)Pn,2n (~r, Ei E, i , t)
+ 3 n,3n (~r, Ei , t)Pn,3n (~r, Ei E, i , t)
+ n,p+n (~r, Ei , t)Pn,p+n (~r, Ei E, i , t)
+ n,d+n (~r, Ei , t)Pn,d+n (~r, Ei E, i , t)
+ n,t+n (~r, Ei , t)Pn,t+n (~r, Ei E, i , t)
+ n,+n (~r, Ei , t)Pn,+n (~r, Ei E, i , t)
+ ... (6.65)

With this definition, we see that if we wish to account for neutron-emitting reactions other than
scattering and fission, we simply need to replace the differential scattering cross section in our
previous equations with X . In terms of operators, we simply re-define the net loss operator, L,
as follows:
~ r, E, , t) + t (~r, E, t)(~r, E, , t)
L (~

dEi di X (~r, Ei E, i , t)(~r, Ei , i , t) (6.66)
0 4

With this definition of L (which simply replaced s with X ), all of our previous equations will
properly account for all neutron-producing reactions.

154
NUEN 601, M.L. Adams notes, 2017 Chapter 6. The Neutron Transport Equation

6.7 Summary

In this chapter we derived the neutron transport equation, by simply writing a statement of con-
servation (change rate = gain rate - loss rate) for the neutrons in some an arbitrary spatial volume,
arbitrary energy interval, and arbitrary cone of directions.
We included all of the gain and loss mechanisms that are most significant in typical nuclear reactors,
and we listed some mechanisms and physical phenomena that we ignore.
In our conservation equation the unknown function is (~r, E, , t), the neutron angular flux. We
have seen in previous chapters that if we know , we can compute the scalar flux, , the net current
~ and any partial-current density, Jn , of interest.
density, J,
We discussed three different kinds of neutron transport problems:

1. Time-dependent
2. Steady-state
3. Eigenvalue. We showed two types of eigenvalue problems:
(a) k
(b)

In time-dependent problems we must keep up with both prompt and delayed neutrons. We do
not need to separate prompt and delayed neutrons in steady-state or eigenvalue problems. In this
chapter we described the process by which delayed neutrons are created and developed mathe-
matical equations that describe the rates at which delayed-neutron emission occurs. This requires
knowledge of the concentrations of delayed-neutron precursors as functions of time and position.
We developed equations whose solutions are these concentrations.
We noted, but did not prove, that the solution of the transport equation is unique: only one
distribution of neutrons can satisfy conservation everywhere in the phase-space domain and also
satisfy the boundary and initial conditions.
We described the initial condition that is required for a well-posed time-dependent transport
problem.
We described several types of boundary conditions that satisfy the requirements for well-posed
transport problems of any kind. These include:

Explicitly specified incident flux


Specular reflection (mirror)
Diffuse reflection (white)
General reflection (albedo)
Periodic

155
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Chapter 7

The Neutron Diffusion Equation

7.1 Introduction

In a previous chapter we derived the neutron transport equation, which is an accurate mathematical
model of what happens in reactors in the real world. The transport equations solution is the angular
flux, which in general is a function of seven independent variables: three spatial coordinates, two
direction coordinates, one energy variable, and in time-dependent problems one time variable.

The combination of the high-dimensional phase space and the hyperbolic character of the equation
makes the transport equation difficult to solve. Analytic solutions can be obtained only in the
simplest of problems. Numerical solutions are challenging because of the high dimensionality and
the lack of smoothness of the actual solution. Suppose we discretized the phase space by dividing
each of the x, y, z, E, , and ranges into 100 mesh intervals. This divides phase space into 1006 =
1012 mesh cells. If there is a single angular-flux unknown per mesh cell, then there are 1012 = one
trillion unknowns, and we have a 1012 1012 matrix system to solve. In a time-dependent problem
this would need to be solved for every time step. Further, in many problems of practical interest,
the 1006 phase-space mesh described here would be too coarse to provide sufficiently accurate
solutions. Even on the fastest supercomputers, accurately solving the 3D transport equation is
significant challenge.

In this chapter we do not explore a direct numerical attack on the transport equation. Instead we
make an analytic approximation to the equation itself, which will have the following effects:

1. It introduces an approximation that reduces the accuracy of the mathematical model. In


some problems the accuracy remains high; in others the accuracy becomes poor.

2. It removes (integrates out) the direction variables, thus removing two of the phase-space
dimensions.

3. It changes the character of the solution, making it smoother and easier to approximate.

156
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

The approximation we make is the

diffusion approximation,

also known as Ficks Law (although it is not really a law). The approximation expresses the
~ in terms of the scalar flux, . In this chapter we show one way to derive
net current density, J,
this approximation from a more fundamental approximation, namely that the angular flux () is
a linear function of the direction-vector components (x , y , and z ). We mention, but do not go
into detail about, other ways to derive the same diffusion approximation.

After we derive the diffusion equation we explore its solutions, beginning with simple problems and
working our way to more realistic reactor problems.

7.2 The P1 Equations

Recall the time-dependent neutron transport equation, written here under the assumption that
the material vibrates isotropically and presents no preferred directions to the neutrons, and the
microscopic cross sections contained in the macroscopic cross sections have been appropriately
averaged over the nucleus velocity distributions:

1 (~r, E, , t) ~ r, E, , t) + t (~r, E, t)(~r, E, , t)


+ (~
v(E) t

# of dnp
X types
1
= Sfixed (~r, E, , t) + i (E)i Ci (~r, t)
4
i=1


p (E)
+ dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t)
4 0


+ dEi di s (~r, Ei E, i , t)(~r, Ei , i , t) . (7.1)
0 4

157
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.2.1 0th Moment

Without approximation let us integrate


this equation over all directions. That is, let us operate on
every term of the equation with 4 d. As we do this, we recognize that:

d (~r, E, , t) = (~r, E, t), (7.2)
4

 
~ r, E, , t) =
d (~ ~ d (~r, E, , t)
4 4

~ J(~
= ~ r, E, t), (7.3)


1
d = 1, (7.4)
4 4


d dEi di s (~r, Ei E, i , t)(~r, Ei , i , t)
4 0 4
 
= dEi di ds (~r, Ei E, i , t) (~r, Ei , i , t)
0 4
| 4 {z }
r,Ei E,t)
s (~

 
= dEi s (~r, Ei E, t) di (~r, Ei , i , t)
0
| 4 {z }
(~
r,Ei ,t)


= dEi s (~r, Ei E, t)(~r, Ei , t), (7.5)
0

and we define

Sfixed,0 (~r, E, t) d Sfixed (~r, E, , t). (7.6)
4

158
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

With these recognitions and this definition, integration of the transport over all directions yields:

0TH ANGULAR MOMENT OF TRANSPORT EQUATION


(direction integral of equation)

1 (~r, E, t) ~ ~
+ J(~r, E, t) + t (~r, E, t)(~r, E, t)
v(E) t

# of dnp
X types
= Sfixed,0 (~r, E, t) + i (E)i Ci (~r, t)
i=1


+ p (E) dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t)
0


+ dEi s (~r, Ei E, t)(~r, Ei , t) . (7.7)
0

Remember: We made no approximations herewe just integrated the transport equation. This
has produced an equation with no direction variable. However, it is a single equation that contains
four unknown functions:

, Jx , Jy , Jz

159
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.2.2 1st Moment

As the caption above denotes, when we integrate an expression over all directions we are taking
the 0th angular moment of that expression. We now take the first angular moment of the
transport equation. To do this, we multiply the transport equation by and then integrate over
all directions. That is, we operate on each term in the equation with 4 d. This is really taking
three separate moments and generating three separate equations at onceone each for x , y , and
z and writing the three of them in vector notation. As we do this, we recognize that:


~ r, E, t),
d (~r, E, , t) = J(~ (7.8)
4

 
~ r, E, , t) =
d (~ ~ d (~r, E, , t)
4 4

~ ~~P(~r, E, t),
(7.9)


1
d = ~0, (7.10)
4 4


d dEi di s (~r, Ei E, i , t)(~r, Ei , i , t)
4 0 4

 
= dEi di d s (~r, Ei E, i , t) (~r, Ei , i , t)
0 4
| 4 {z }
r,Ei E,t)
i s1 (~

 
= dEi s1 (~r, Ei E, t) di i (~r, Ei , i , t)
0
| 4 {z }
~ r,Ei ,t)
J(~


= ~ r, Ei , t),
dEi s1 (~r, Ei E, t)J(~ (7.11)
0

(there will be more about s1 later) and we define



~fixed,1 (~r, E, t)
S d Sfixed (~r, E, , t). (7.12)
4

160
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

~
We have defined the pressure tensor, ~P

~
~P d (~r, E, , t). (7.13)
4

This is just a shorthand notation that allows us to compactly express the second angular moments
of :

Puw d u w (~r, E, , t), (7.14)
4

where u = x, y, or z and w = x, y, or z.

With these recognitions and this definition, integration of the transport over all directions yields
the following vector equation (three scalar equations written in vector notation):

1ST ANGULAR MOMENT OF TRANSPORT EQUATION


(direction integral of equation.)

~ r, E, t)
1 J(~ ~ ~~P(~r, E, t) + t (~r, E, t)J(~
+ ~ r, E, t)
v(E) t


~fixed,1 (~r, E, t) +
=S ~ r, Ei , t) .
dEi s1 (~r, Ei E, t)J(~ (7.15)
0

Remember: We made no approximations herewe just multiplied the transport equation by and
integrated.

In the first-moment equation appears a first-moment scattering cross section. We leave it as an


exercise to show that this cross section satisfies the following definition:
1
s1 (~r, Ei E, t) = d0 0 s (~r, Ei E, 0 , t), (7.16)
1

where 0 i = scattering-angle cosine (lab frame) and s (~r, Ei E, 0 , t) is a per-cosine


quantity, not per-steradian.

161
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Remark: If scattering is elastic, then for a given nuclide mass there is only one possible scattering
cosine for an energy change from Ei to E, if the nucleus is taken to be at rest in the lab frame.
This means ei (Ei E, 0 ) is a delta-function in 0 :
elastic, nuclide i at rest in lab
ei (~r, Ei E, 0 ) ei (Ei E)(0 0,Ei E ). (7.17)

Here 0,Ei E is defined to be the unique lab-frame scattering-angle cosine of a scattering event off
of nuclide i that changes the neutrons lab-frame energy from Ei to E.

7.2.3 P1 Approximation

The 0th and 1st angular-moment equations are really four scalar equations. In these equations are
the following unknown functions:

scalar flux, ,

net current density, which is three scalar functions: Jx , Jy , and Jz ,

pressure tensor, which is six scalar functions: Pxx , Pxy = Pyx , Pxz = Pzx , Pyy , Pyz = Pzy ,
and Pzz . (One of these is redundant with , because 2x +2y +2z = 1 Pxx +Pyy +Pzz = .)

~
If there were a way to express ~P in terms of and J,
~ then we could eliminate the pressure tensor
from the 0th- and 1st-moment equations and would then have four equations for the four remaining
unknown functions , Jx , Jy , and Jz . There are many ways to do this, but all of them involve
~
approximations. One of the most common approximations that allows us to eliminate ~P is the
so-called

P1 approximation.

The P1 approximation is that the angular flux is a Polynomial of order 1 in the components of the
direction vector:
P1 approx
(~r, E, , t) a(~r, E, t) + x bx (~r, E, t) + y by (~r, E, t) + z bz (~r, E, t)

= a(~r, E, t) + ~b(~r, E, t) (7.18)

162
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We can express the a and b coefficients in this approximation in terms of and J. ~ If we integrate
Eq. (7.18) over all directions we obtain:

d(~r, E, , t) (~r, E, t)
4
P1 approx
d [a(~r, E, t) + x bx (~r, E, t) + y by (~r, E, t) + z bz (~r, E, t)]
4

= . . . = 4a(~r, E, t)

1
a(~r, E, t) = (~r, E, t) (7.19)
4

If we multiply Eq. (7.18) by and then over all directions we obtain:

~ r, E, t)
d (~r, E, , t) J(~
4
P1 approx
d [a(~r, E, t) + x bx (~r, E, t) + y by (~r, E, t) + z bz (~r, E, t)]
4

4 ~
= ... = b(~r, E, t)
3

3 ~
~b(~r, E, t) = J(~r, E, t) (7.20)
4

Thus, we can write the P1 approximation for the angular flux in the following form:
P1 approx 1 h ~ r, E, t)
i
(~r, E, , t) (~r, E, t) + 3 J(~ (7.21)
4

163
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We can calculate the pressure tensor in terms of the P1 -approximate angular flux:

1 h ~ r, E, t)
i
Puw d u w (~r, E, t) + 3 J(~
4 4


1 3 ~
= (~r, E, t) d u w + J(~r, E, t) d u w . (7.22)
4 4 4 4

We leave it as an exercise to show the following identities:



4
d u w = uw , (7.23)
4 3


d u w = ~0, (7.24)
4

where uw , the Kronecker delta, equals 1 if u = w and 0 if u 6= w.

It follows that if the angular flux is linearly anisotropic, then the off-diagonal terms of the pressure
tensor are zero and the diagonal terms are /3. From this we find that

P1 approx 1
~ ~
~P(~r, E, t) (~~ r, E, t) (7.25)
3

This is somewhat unsatisfying, because we have no idea how close the angular flux is to linearly
anisotropic in problems of interest. We will need to explore this question before we trust calculations
that are based on this approximation.

164
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

If we make this substitution in the 1st-moment equation, then the (exact) 0th-moment equation and
the (approximate) 1st-moment equation constitute four equations for the four unknown functions
, Jx , Jy , and Jz :

THE P1 EQUATIONS (TIME-DEPENDENT)

1 (~r, E, t) ~ ~
+ J(~r, E, t) + t (~r, E, t)(~r, E, t)
v(E) t

# of dnp
X types
= Sfixed,0 (~r, E, t) + i (E)i Ci (~r, t)
i=1


+ p (E) dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t)
0


+ dEi s (~r, Ei E, t)(~r, Ei , t) . (7.26)
0

~ r, E, t) 1
1 J(~ ~ r, E, t) + t (~r, E, t)J(~
+ (~ ~ r, E, t)
v(E) t 3


~fixed,1 (~r, E, t) +
=S ~ r, Ei , t) .
dEi s1 (~r, Ei E, t)J(~ (7.27)
0

165
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

The steady-state version of the P1 equations is obvious (time derivative and time arguments vanish,
and delayed neutrons do not need to be considered separately from prompt neutrons):

THE P1 EQUATIONS (STEADY-STATE)

~ J(~
~ r, E, ) + t (~r, E)(~r, E)


= Sfixed,0 (~r, E) + (E) dEi (~r, Ei , t)f (~r, Ei )(~r, Ei )
0


+ dEi s (~r, Ei E)(~r, Ei ) . (7.28)
0

1~ ~ r, E)
(~r, E, t) + t (~r, E)J(~
3

~fixed,1 (~r, E) +
=S ~ r , Ei ) .
dEi s1 (~r, Ei E)J(~ (7.29)
0

The P1 equations for the k-eigenvalue problem are:

THE P1 EQUATIONS (k-eigenvalue)


~ r, E) + t (~r, E)(~r, E) = (E)
~ J(~ dEi (~r, Ei )f (~r, Ei )(~r, Ei )
k 0


+ dEi s (~r, Ei E)(~r, Ei ) , (7.30)
0


1~ ~ r, E)= ~ r , Ei ) .
(~r, E) + t (~r, E)J(~ dEi s1 (~r, Ei E)J(~ (7.31)
3 0

166
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.3 The Diffusion Equation

To obtain the P1 equations we approximated the angular flux as linearly anisotropic. To obtain
the diffusion equation from the P1 equations we need to make one more approximation. For time-
dependent problems the approximation is:
(
~ r, E, t) 
1 J(~ ~ r, E, t) ~ r, Ei , t)
+ t (~r, E, t)J(~ dEi s1 (~r, Ei E, t)J(~
v(E) t 0

diffusion ~ r, E, t) ,
tr (~r, E, t)J(~ (7.32)

where

tr (~r, E, t) the transport cross section. (7.33)

It is not easy in general to define tr such that this approximation is accurate, and in many cases
even the best value for tr does not make the approximation accurate for all three components of
~ Nevertheless, we proceed boldly onward. With this approximation, we can eliminate J~ in terms
J.
of :

DIFFUSION APPROXIMATION

diffusion
~ r, E, t) 1 ~ r, E, t)
J(~ (~ (7.34)
3tr (~r, E, t)

~ r, E, t),
D(~r, E, t)(~ (7.35)

where
1
D(~r, E, t) = the diffusion coefficient. (7.36)
3tr (~r, E, t)

167
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We can insert our approximate expression for J~ into the exact 0th-moment equation to obtain the
diffusion equation:

THE DIFFUSION EQUATION (time-dependent)

1 (~r, E, t) ~ h i
~ r, E, t) + t (~r, E, t)(~r, E, t)
D(~r, E, t)(~
v(E) t

# of dnp
X types
= Sfixed,0 (~r, E, t) + i (E)i Ci (~r, t)
i=1


+ p (E) dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t)
0


+ dEi s (~r, Ei E, t)(~r, Ei , t) . (7.37)
0

In a steady-state or k-eigenvalue problem, the approximation we must make also involves defining
a transport cross section:
n 
~
t (~r, E)J(~r, E) ~
dEi s1 (~r, Ei E)J(~r, Ei )
0

diffusion, steady-state or eigenvalue


~ r, E) .
tr (~r, E)J(~ (7.38)

We define the diffusion coefficient as before:


1
D(~r, E) . (7.39)
3tr (~r, E)

168
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We can insert our approximate expression for J~ into the exact 0th-moment equation to obtain the
diffusion equation for the k-eigenvalue problem:

THE DIFFUSION EQUATION (k-eigenvalue)

h i
~ D(~r, E)(~
~ r, E) + t (~r, E)(~r, E)


(E)
= dEi (~r, Ei )f (~r, Ei )(~r, Ei )
k 0


+ dEi s (~r, Ei E)(~r, Ei ) . (7.40)
0

7.4 Initial and Boundary Conditions

The diffusion equation is an integro-differential equation for the scalar flux, . A well-posed problem
involving differential or integro-differential equations must include proper boundary and initial
conditions in addition to the equations themselves.

7.4.1 Initial

The initial condition that makes a time-dependent diffusion problem well-posed is just what your
common sense would tell you: You must know the scalar flux at the initial time if you expect to
know it at later times. The mathematical statement of this is:

(~r, E, t0 ) = initial (~r, E) = known. (7.41)

7.4.2 Boundary

Recall that transport boundary conditions specify the angular flux of neutrons entering the problem
domain through its boundary. This can be specified explicitly or specified as a function of the exiting
angular flux (as is the case with reflecting, albedo, and periodic boundary conditions). Here we
examine the analogous boundary conditions for the diffusion equation.

169
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Mathematically, at every point on the spatial domain, the diffusion boundary condition must specify
exactly one of the following:

1. The scalar flux,


~
2. The normal component of the gradient of the scalar flux, en
~ =
3. An equation relating the two: + en

We shall see how to do this for several kinds of boundary conditions that are of interest in reactor
analysis.

For time-dependent problems the boundary conditions must be specifed as a function of time, so
in what follows we include t as an argument. For steady-state and eigenvalue problems there is no
time variable, so for such problems the t argument should be omitted.

7.4.2.1 Explicitly Specified Incident Flux

We have seen previously that if the angular flux entering the problem domain is known, then the
transport boundary condition is simpleit just says that the angular flux on the boundary equals
the specified incident flux for incoming directions:

(~r, E, , t) = inc (~r, E, , t) = known for all : en (~r) < 0, for all ~r Vinc . (7.42)

However, this is not a simple case for the P1 or diffusion equations, mainly because their unknowns
are the scalar flux and net current density, which are both full-range integrals of the angular flux.
The scalar flux knows nothing about the difference between incoming and outgoing directions, and
the net current density has only crude information about directionality.

As we saw above, the diffusion approximation is consistent with the P1 approximation for the
angular flux. That is, it is consistent with approximating the angular flux as linearly anisotropic.
What if we simply insert this approximation for into the transport boundary condition? Let us
try it:
1 h ~ r, E, t)
i
?
(~r, E, t) + 3 J(~ = inc (~r, E, , t), : en (~r) < 0, ~r Vinc . (7.43)
4
We see that this cannot work in general:

The left side is linear in x , y , and z ,

but the right side could be any function.

Given that the P1 approximation cannot satisfy the detailed transport boundary condition in the
case of an arbitrary incident flux, what shall we do?

170
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

The next best thing is to make the P1 or diffusion boundary condition capture the most important
information from the transport boundary condition. For example, it makes a lot of sense to require
that the

incoming partial current.

be correct. This would mean that the diffusion problem would at least respect the correct in-
coming particle flow rate, even though it would not know the difference between different angular
distributions that produce the same flow rate. Mathematically, we can enforce this by multiplying
Eq. (7.43) by en and integrating over incoming directions:
 1 h i
d en
~
(~r, E, t) + 3 J(~r, E, t)
:en <0 4

= d en inc (~r, E, , t) Jn (~r, E, t) ,

:en <0

or
   
1 2 ~~ ~
(~r, E, t) 3en I J(~r, E, t) = Jn (~r, E, t) ,
4 3
or

Marshak boundary condition for P1 equations

1 1 ~ r, E, t) = Jn (~r, E, t)
(~r, E, t) en J(~ = known , ~r Vinc . (7.44)
4 2

Eq. (7.57) is sometimes called a

Marshak

boundary condition, named after R.E. Marshak. Note that it results from forcing the P1 -
approximate flux to match the correct incoming flow rate at the boundary.

171
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

An alternative to the Marshak boundary condition is the Mark boundary condition, named after
J.C. Mark. The Mark boundary condition forces the P1 -approximate flux to match the incident
angular flux at a particular value (1/ 3) of the cosine of the incident angle. The Mark condition
is

1 1 3 ~ r, E, t)
(~r, E, t) en J(~
4 2 2
2 
1
= dn inc (~r, E, , t)
= known , ~r Vinc . (7.45)
2 0 |en |=1/ 3

Here n is the azimuthal angle about the axis perpendicular to the boundary. The cosine of the
polar angle relative to this axis is en .

As written, the Mark and Marshak boundary conditions can be applied to either the P1 or diffusion
equations. To apply them directly to the diffusion equation it is often convenient to replace J~ with
~
D. For example, the Marshak conditionwhich is the one we most often use in reactor
computationscan be written as follows, given the diffusion approximation:

Marshak boundary condition for the diffusion equation

 
1 1 ~
+ Den = Jn (~r, E, t) = known , ~r Vinc . (7.46)
4 2 ~
r,E,t

7.4.2.2 Extrapolated Boundary Condition

Suppose we multiply Eq. (7.46) by 4:


~ r, E, t)
(~r, E, t) + 2D(~r, E, t) en (~ = 4Jn (~r, E, t) = known , ~r Vinc . (7.47)

Notice that the left side is just

a linear extrapolation of the scalar flux to a distance = 2D outside the boundary.

That is, we can rewrite this expression as

linear.extrap (~r + dlin.extrap en , E, t) = 4Jn (~r, E, t) = known , ~r Vlin.extrap , (7.48)

where dlin.extrap = 2D. This is equivalent to Eq. (7.46)it is just a different way to write it. Fig. 7.1
provides a graphical depiction for the case of a vacuum boundary, which means Jn = 0.

172
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Figure 7.1: Illustration of extrapolated boundary condition, both linear and with curvature. Show
dlin.extrap and d.

Studies of neutron transport solutions in simplified problems suggest that a slightly longer extrapo-
lation distance produces a more accurate diffusion solutionone that better matches the transport
solution in the interior of a large diffusive problem:
transport-matching
dlin.extrap (3 0.71044 . . .) D (7.49)

This holds rigorously only for problems that are many, many mean-free paths thick, with approx-
imately planar boundaries, and with a material composition such that each collision on average
produces very nearly one exiting neutron either by scattering or fission.

It is a small step to modify the linear-extrapolation boundary condition to create one in which
the complete scalar flux function (not just a linear extrapolation, but with whatever curvature is
present inside the problem domain) is evaluated outside the problem boundary:

(~r + dextrap en , E, t) = 4Jn (~r, E, t) = known , ~r Vextrap . (7.50)

We shall refer to this as extrapolation with curvature to distinguish it from the straight-line
extrapolation of the previous equation. Notice that if the solution is concave down, as it is in a
material with k > 1, we need dextrap < dlin.extrap if we want the two conditions to produce the
same solution. This is the case illustrated in Fig. 7.1. If k = 1, we need dextrap = dlin.extrap , and if
k < 1 we need dextrap > dlin.extrap . We will typically use an extrapolation-with-curvature distance
of 2D, which is smaller than the transport-matching dlin.extrap , unless there we have reason to use
something different:

dextrap 2D. (7.51)

173
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.4.2.3 Vacuum Boundary Condtion

A vacuum boundary is a special case of the specified-incident boundary, with an incident partial
current of zero. The extrapolation-with-curvature boundary condition for the vacuum case is
simply:

VACUUM-BOUNDARY EXTRAPOLATED BOUNDARY CONDITION

(~r + d en , E, t) = 0, ~r Vvac (7.52)

where d is the extrapolation-with-curvature distance (usually taken to = 2D.)

7.4.2.4 Connection with statement about allowed boundary conditions

We can see both the Marshak and Mark boundary conditions specify the value of a particular linear
combination of the scalar flux and its normal derivative at the boundary. Previously we listed three
types of boundary conditions that produce a well-posed diffusion problem. The Marshak and Mark
conditions meet the definition of the third type. Thus, the linear-extrapolation boundary condition,
which can be made equivalent to either the Mark or Marshak condition through proper choice of
dlin.extrap , is also of the third allowed type.

It takes slightly more effort to match the extrapolation-with-curvature boundary condition to one
of the three allowed types. It is actually a combination of a physical approximation and the first
listed type of condition. The physical approximation is that that problem domain is extended to a
distance d outward from the true boundary. Then the boundary condition corresponds to specifying
the scalar flux on this extrapolated boundary.

7.4.2.5 Reflecting Boundary Condition

For the P1 and diffusion equations there is no difference among specular reflection (mirror bound-
ary), diffuse reflection (white boundary), and other kinds of conditions that send all exiting
neutrons back into the problem. All of these conditions have one thing in common:

incident partial current = exiting partial current


To say this another way:

the net flow rate is zero

across reflecting surfaces. The P1 statement of this condition is simple:


~ r, E, t) = 0
en J(~ for all ~r Vref (7.53)

174
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

The diffusion statement is obtained by inserting the diffusion approximation for the net current
density. Since the diffusion coefficient cannot be zero, it divides out, leaving:

DIFFUSION REFLECTING BOUNDARY CONDITION

~ r, E, t) = 0
en (~ for all ~r Vref (7.54)

7.4.2.6 Periodic

The diffusion and P1 versions of the periodic boundary condition follow directly from the transport
condition:
~ r), E, t),
(~r, E, t) = (~r + d(~ ~r Vper , (7.55)

where d~ is a translation vector whose magnitude is one dimension of the domain, so that for
~ r) is another point on the boundary.
~r Vper , ~r + d(~

7.4.2.7 Albedo

The transport albedo boundary condition specifies the incident angular flux in terms of the outgoing
angular flux. The P1 and diffusion versions of this specify the incident partial current in terms of
the exiting partial current. For P1 the expression is:
 
1 1 1 1
~ r, E, t) = (~r, E, t) (~r, E, t) + en J(~
~ r, E, t) , ~r Valbedo . (7.56)
(~r, E, t) en J(~
4 2 4 2

As usual we obtain the diffusion condition by inserting the diffusion approximation (J~ = D)
~
for the net current density:
1 1 ~ r, E, t)
(~r, E, t) + D(~r, E, t)en (~
4 2  
1 1 ~
= (~r, E, t) (~r, E, t) D(~r, E, t)en (~r, E, t) , ~r Valbedo . (7.57)
4 2

7.5 Interface and Other Conditions

7.5.1 Interface Conditions

A problem with multiple material regions can be viewed as multiple problems connected by in-
terface conditions. These conditions take the place of boundary conditions for the regions that
touch a given interface.

175
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

The interface conditions that are appropriate for the P1 and diffusion equations are simply state-
ments of continuity, which follow from the continuity of the angular flux in the transport equation.
They are:

1. The scalar flux is continuous.

2. The normal component of the net current density is continuous.

Let A and B denote two regions that meet at an interface. Let A and B denote the scalar flux in
regions A and B, respectively, with similar notation for the net current density. Then the continuity
statements for the P1 equations become:

A (~r, E, t) ~rA/B interface = B (~r, E, t) ~rA/B interface



(7.58)


en (~r) J~A (~r, E, t) = en (~r) J~B (~r, E, t) (7.59)

rA/B interface
~ rA/B interface
~

The scalar-flux continuity equation is the same for the diffusion equation as for P1 . The net-current
condition becomes the following for diffusion theory:
h i
A ~ A
D (~r, E, t)en (~r) (~r, E, t)
rA/B interface
~

h i
~ B (~r, E, t)
= DB (~r, E, t)en (~r) (7.60)
rA/B interface
~

Two important things to remember about these continuity conditions:

1. If the diffusion coefficient is different in the two regions, then the derivative of the scalar flux
is

discontinuous at the interface.

2. Only the normal component of the net current density is required to be continuous. Parallel
components may be discontinuous.

7.5.2 Other Helpful Conditions

In most diffusion problems the scalar flux must remain finite throughout the problem domain. (One
exception to this rule is a problem with a singular source, in which neutrons are emitted at a finite
rate from a source with zero volume or at an infinite rate from a source with finite volume.) In
infinite-medium problems, the boundary condition at infinity is often simply a statement that

the scalar flux remains finite.

176
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Diffusion problems that are described in spherical or cylindrical coordinates need boundary con-
ditions at the origin (spherical) or on the axis (cylindrical). The following boundary conditions,
which express the symmetry of the solutions, are valid for such problems:
 
dg (r)
=0, spherical (7.61)
dr r=0

 
g (r, z)
= 0 , all z , cylindrical (7.62)
r r=0

It works just as well to enforce finiteness at r = 0 in problems with spherical or cylindrical coordi-
nates. Sometimes this is easier than taking the derivative and setting it to zero.

7.6 Multigroup Transport, P1 , and Diffusion Equations

We have derived the neutron transport, P1 , and diffusion equations, and we have shown the form
of each for time-dependent, steady-state, and eigenvalue problems. We have noted that the P1
equation is far simpler than the transport equation, and the diffusion equation is even simpler.

Nevertheless, the energy-dependent diffusion equation is too complicated to solve for realistic reac-
tor problems. To make it tractable, reactor analysts usually employ the multigroup approximation,
which converts the integro-differential equations into coupled sets of partial differential equations.

In this section we describe the multigroup approximation and show how it reduces an energy-
dependent problem to a coupled set of energy-independent problems. We shall see that the multi-
group approximation is rather simple: we divide the energy range into intervals, define a group of
neutrons to be those that have energies in one interval, and then average the cross sections over each
interval. The averaging is where the approximation comes in. After we make the approximation
and develop the equations, we discuss the accuracy of the approximation. (In many problems of
interest it is quite good.)

The multigroup transport equations are much more tractable than the energy-dependent transport
equation, but they still are too difficult to solve routinely for a problem as large as a commercial-
scale nuclear reactor. The multigroup P1 and diffusion equations are much easier to solve. We will
focus mainly on multigroup diffusion.

With the multigroup diffusion equations in hand we consider first the simplest case: a single energy
group. We apply this one-group diffusion equation to homogeneous reactors and find that in this
simple setting we can obtain analytic solutions for the k-eigenvalues and eigenfunctions. Buoyed by
this success, we make the leap to the multigroup diffusion equations with two energy groups. We
find again that, if we employ simple enough boundary conditions, we can obtain analytic solutions
for homogenous reactors.

The multigroup equations require weighting functions (spectra) to generate the energy-averaged
cross sections that they employ. If these are not accurate approximations of the real-world energy

177
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

shape of the flux, then the multigroup equations will not produce accurate solutions. So how do we
generate accurate weighting spectra? We address this question by applying the energy-dependent
(not multigroup) diffusion equation to homogeneous reactors. We find that for some problems of
this type, given certain approximate boundary conditions, we can obtain analytic or semi-analytic
solutions for the energy shape of the solution. With this information in hand, we go on to discuss
how the spectra are calculated and applied to the analysis of real (heterogeneous) reactors.

7.6.1 Multigroup Transport

Let us begin by dividing the energy axis into G non-overlapping intervals that cover the range from
zero to some maximum energy (say 10 or 12 MeV) above which we assume there are no neutrons.
See Fig. 7.2 for an illustration.

Figure 7.2: Energy axis showing intervals used in the multigroup approximation. Bottom is EG ;
top is E0 ; interval shown is the g-th, ranging from Eg to Eg1 .

The neutrons in the g-th interval are called the g-th group of neutrons.

We illustrate the derivation of the multigroup approximation using the k-eigenvalue transport equa-
tion. The derivation is similar for other kinds of problems and for the P1 and diffusion equations.
Recall the k-eigenvalue transport equation:
~ r, E, ) + t (~r, E)(~r, E, )
(~


1 (E)
= dEi (~r, Ei )f (~r, Ei )(~r, Ei )
k 4 0


+ dEi di s (~r, Ei E, i )(~r, Ei , i ) . (7.63)
0 4

178
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We integrate the equation over the g-th energy interval, which ranges from Eg to Eg1 :
Eg1 Eg1
~
dE (~r, E, ) + dE t (~r, E)(~r, E, )
Eg Eg

" #
Eg1
1 (E)
= dE dEi (~r, Ei )f (~r, Ei )(~r, Ei )
k Eg 4 0

"
Eg1
#
+ dEi di dE s (~r, Ei E, i ) (~r, Ei , i ) . (7.64)
0 4 Eg

We define:
Eg1
g (~r, ) dE (~r, E, ) , (7.65)
Eg

Eg1
g (~r) dE (~r, E) , (7.66)
Eg
Eg1
g dE(E) , (7.67)
Eg

Eg1
s (~r, Ei g, i ) dE s (~r, Ei E, i ) . (7.68)
Eg

Manipulate:

"
Eg1 #
Eg1
Eg dE t (~r, E)(~r, E, ) Eg1
dE t (~r, E)(~r, E, ) = Eg1 dE (~r, E, )
Eg Eg dE (~r, E, ) Eg

D E
= t (~r, ) g (~r, ) , (7.69)
g

179
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

G
X Eg0 1
dEi (~r, Ei )f (~r, Ei )(~r, Ei ) = dEi (~r, Ei )f (~r, Ei )(~r, Ei )
0 g 0 =1 Eg0

E
Eg0 1
"
g 0 1
G dEi (~r, Ei )f (~r, Ei )(~r, Ei )
#
X Eg0
= Eg0 1 dEi (~r, Ei )
g 0 =1 E 0 dE i (~
r , E i ) Eg0
g

G D
X E
= (~r)f (~r) 0 g0 (~r) (7.70)
g
g 0 =1


dEi s (~r, Ei g, i )(~r, Ei , i )
0
E 0
Eg0 1
"
g 1
G dEi s (~r, Ei g, i )(~r, Ei , i )
#
X Eg0
= Eg0 1 dEi (~r, Ei , i )
0
g =1 E 0 dEi (~r , E i , i ) Eg 0
g

G D
X E
= s (~r, i ) g0 (~r, i ) (7.71)
g 0 g
g 0 =1

Important observation: at this point we have simply integrated the transport equation over an
energy interval and made some definitions. We have not yet introduced approximations!

We have defined several averaged cross sections, denoting them with angle brackets, hi. Each is a
weighted average, with either the exact angular flux or exact scalar flux as a weight function. With
these definitions, the transport equation integrated over the g-th energy interval can be written as:
G
~
D E 1 g X D E
g (~r, ) + t (~r, ) g (~r, ) = (~r)f (~r) 0 g0 (~r)
g k 4 0 g
g =1
G
X D E
+ di s (~r, i ) g0 (~r, i ) . (7.72)
4 g 0 g
g 0 =1

180
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

If we were given the averaged cross sections, we could solve these equations far more easily than
we could solve the energy-dependent equations (if we could solve the latter at all). However, we
have defined these averaged cross sections in terms of the

angular and scalar fluxes,

which of course are unknown. This is where the approximation comes in: We generate our energy-
averaged cross sections using approximate weighting functions. That is, we somehow obtain an
approximate flux spectrum, f (~r, E), and then define:
Eg1
Eg dE t (~r, E)f (~r, E)
t,g (~r) Eg1 (7.73)
Eg dE f (~r, E)

Eg0 1
Eg0 dEi (~r, Ei )f (~r, Ei )f (~r, Ei )
f,g0 (~r) Eg0 1 (7.74)
E 0 dEi f (~r, Ei )
g

Eg0 1
Eg0 dEi s (~r, Ei g, i )f (~r, Ei )
s,g0 g (~r, i ) Eg0 1 (7.75)
E 0 dEi f (~r, Ei )
g

Now we insert our approximate averaged cross sections into our exact energy-integrated transport
equation to obtain the (approximate) multigroup transport equations:
G
~ 1 g X
g (~r, ) + t,g (~r) g (~r, ) = f,g0 (~r) g0 (~r)
k 4 0
g =1

G
X
+ di s,g0 g (~r, i )g0 (~r, i ) . (7.76)
g 0 =1 4

181
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Remarks:

If the weighting spectrum (which we called f (E) in the equations above) accurately ap-
proximates the energy distribution of the actual transport solution, then the solution of the
multigroup equations will be an accurate approximation of the true energy-integrated fluxes.

The exact weighting function for t and s is the angular flux, which varies with the direc-
tion . Thus, the perfectly averaged t and s will vary with . The essentially universal
approach to multigroup approximations, however, is to use an approximate weighting spec-
trum (f ) that is isotropic. In general, no matter how good the weighting spectrum is, if it is
isotropic it cannot produce an averaged cross section that is perfect for all values of .

The solution of the multigroup equations is the set of group-integrated angular fluxes, {g , g =
1 . . . G}, each of which is a function of position and direction (not energy!).

The group-integrated fluxes are not densities in energywe have integrated over this density.
Thus, the units of the group-integrated angular and scalar fluxes are:

n/(cm2 -s-ster) and n/(cm2 -s)no per-unit-energy units.

The multigroup transport equations are a coupled set of differential equations. They are
coupled only through the scattering and fission terms. Each differential equation (one per
group) requires a set of boundary conditions that somehow define the incoming angular flux
for the group, at all points on the problem boundary.

The multigroup approximation has allowed us to trade one problem in which the solution depends
on six continuous variables (x, y, z, E, , ) for G coupled problems in which each solution depends
on only five of the original six continuous variables: (x, y, z, , ).

7.6.2 Multigroup P1

The multigroup P1 equations can be obtained either by taking 0th and 1st angular moments of
the multigroup transport equation and assuming that each g is linearly anisotropic, or by making
multigroup approximations to the continuous-energy P1 equations. Either way, for the k-eigenvalue
problem we obtain:
G G
~ J~g (~r) + t,g (~r) g (~r) = g
X X
f,g0 (~r) g0 (~r) + s,g0 g (~r)g0 (~r) (7.77)
k 0 0
g =1 g =1
G
1~
g (~r) + t,g (~r) J~g (~r) = s1,g0 g (~r)J~g0 (~r)
X
(7.78)
3
g 0 =1

182
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.6.3 Multigroup Diffusion

The multigroup diffusion equations are obtained from the multigroup 0th-moment equation plus
~ g . This yields:
the diffusion approximation for each groups net current density, J~g Dg

MULTIGROUP DIFFUSION k-EIGENVALUE EQUATIONS

h i
~ ~
Dg (~r)g (~r) + t,g (~r) g (~r)

G G
g X X
= f,g0 (~r) g0 (~r) + s,g0 g (~r)g0 (~r) . (7.79)
k 0 0
g =1 g =1

7.6.4 Weighting Spectra

The multigroup equationswhether transport, P1 , diffusion, or otherrequire weighting functions


(spectra) to generate the energy-averaged cross sections that they employ. These are denoted
f (~r, E) in the equations above. If these functions are not accurate approximations of the real-world
energy shape of the flux, then the multigroup equations may not produce accurate solutions.

So how do we generate accurate weighting spectra? Answering this question has been the sub-
ject of a substantial body of research and development over many decades. In a later chapter
we shall briefly address this question by solving the energy-dependent (not multigroup) diffusion
equation for homogeneous reactors. We will find that for some problems of this type, given certain
approximate conditions, we can obtain analytic or semi-analytic solutions for the energy shape of
the solution. This provides significant insight into how neutrons distribute themselves in energy
in nuclear reactors. After gaining this insight for simple reactors, we will outline how weighting
spectra are calculated and applied to the analysis of real (heterogeneous) reactors.

7.7 One-Group Diffusion Solutions

In our study of the multigroup diffusion equations and their solutions, we consider first the simplest
case: a single energy group. We begin by solving several fixed-source steady-state problems to
remind ourselves of solution techniques for second-order partial differential equations. We then
apply the one-group diffusion equation to the k problem for homogeneous reactors and find that
in this simple setting we can obtain analytic solutions for the k-eigenvalues and eigenfunctions.
Buoyed by this success, we make the leap to the multigroup diffusion equations with two or more

183
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

energy groups. We find again that, if we employ simple enough boundary conditions, we can again
obtain analytic solutions for homogenous reactors.

First we rewrite and manipulate the multigroup diffusion (MGD) equations for the one-group case
to obtain the 1-group diffusion (1GD) equation. We begin with the k-eigenvalue equation Since
there is only one group we do not need a group index:

~ r) + t (~r) (~r) = 1 f (~r) (~r) + s (~r)(~r) .


h i
~ D(~r)(~
(7.80)
k
We can subtract the scattering term from both sides to obtain

1-GROUP DIFFUSION k-EIGENVALUE EQUATION

~ r) + a (~r) (~r) = 1 f (~r) (~r) .


h i
~ D(~r)(~
(7.81)
k

It is easy to write down the steady-state fixed-source version of the 1-group diffusion equation:
h i
~ D(~r)(~
~ r) + a (~r) (~r) = f (~r) (~r) + S(~r) . (7.82)

or

1-GROUP DIFFUSION STEADY-STATE EQUATION

h i
~ D(~r)(~
~ r) + r (~r) (~r) = S(~r) , (7.83)

Here we have defined a one-group net removal cross section:

r (~r) a (~r) f (~r) .

= t (~r) s (~r) f (~r) . (7.84)

In a non-multiplying medium (one with no fission), r = a . But Eq. (7.83) is valid even if
fissioning material is present, if a steady-state solution exists. We will see later that a steady-state
solution does exist in a fixed-source problem with fissioning material, if the system is subcritical.

184
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.1 Simplification for Homogeneous Problems

We use homogeneous to describe a problem or region in which material properties are the same
at every position, so that the Ds and s are not functions of ~r. In such a problem or region,
D comes outside of the ~ operator, and we obtain ~
~ = 2 . The one-group steady-state
diffusion equation for a homogeneous problem is:

D2 (~r) + r (~r) = S(~r) , (7.85)

where 2 is the Laplacian operator. See Table 7.1 for a description of the grad, divergence, and

Laplacian operators in various coordinate systems.

Table 7.1: Gradient, Divergence, and Laplacian operators, various coordinate systems.

Coord. Sys. ~
f ~ ~u
~ f
~ = 2 f

f f f ux uy uz 2f 2f 2f
3D Cartesian ex + ey + ez + + + +
x y z x y z x2 y 2 z 2
df dux d2 f
1D Cartesian ex
dx dx dx2
1 d r2 ur
  
df 1 d 2 df
1D spherical er r
dr r2 dr r2 dr dr
 
df 1 d (rur ) 1 d df
1D cylindrical er r
dr r dr r dr dr
2f
 
f f 1 (rur ) uz 1 f
2D cylindrical er + ez + ez r + 2
r z r r z r r r z

7.7.2 Steady-State One-Group Diffusion Equation: Various Coordinate systems

Different transport and diffusion problems have different symmetries and other characteristics that
lend themselves to description in different coordinate systems. The main difference this makes in
~ term. The main difference in diffusion is the
the transport equation is in the form of the
~ ~
form of the D term. Here we catalog some interesting examples.

185
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.2.1 Cartesian

Given Cartesian coordinates with a solution that depends on all three variables, we have the
following for our steady-state one-group diffusion equation:

     
(x, y, z) (x, y, z) (x, y, z)
D(x, y, z) D(x, y, z) D(x, y, z)
x x y y z z

+ r (x, y, z) (x, y, z) = S(x, y, z) , 3D Cart., heterogeneous (7.86)

2 (x, y, z) 2 (x, y, z) 2 (x, y, z)


 
D + +
x2 y 2 z 2

+ r (x, y, z) = S(x, y, z) , 3D Cart., homogeneous (7.87)

Given Cartesian coordinates with a solution that depends on only x and y (infinite and uniform in
z), we have the following for our steady-state one-group diffusion equation:

   
(x, y) (x, y)
D(x, y) D(x, y)
x x y y

+ r (x, y) (x, y) = S(x, y) , 2D Cart., heterogeneous (7.88)

2 (x, y) 2 (x, y)
 
D +
x2 y 2

+ r (x, y) = S(x, y) , 2D Cart., homogeneous (7.89)

Given Cartesian coordinates with a solution that depends on only x (infinite and uniform in y and
z), we have the following for our steady-state one-group diffusion equation in slab geometry:

 
d d(x)
D(x) + r (x) (x) = S(x) , 1D slab, heterogeneous (7.90)
dx dx

d2 (x)
D + r (x) = S(x) , 1D slab, homogeneous (7.91)
dx2

186
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.2.2 Spherical

Given a symmetric problem in which everything depends only on distance from the origin, spherical
coordinates are most convenient, and only one spatial coordinate matters: r, the distance from the
origin. This is sometimes called 1D spherical geometry:

 
1 d 2 d(r)
2 r D(r) + r (r) (r) = S(r) , 1D spherical, heterogeneous (7.92)
r dr dr

 
1 d 2 d(r)
D 2 r + r (r) = S(r) , 1D spherical, homogeneous (7.93)
r dr dr

7.7.2.3 Cylindrical

Given a symmetric problem in which everything depends only on distance from an axis of symmetry
(uniform in azimuthal coordinate; infinite and uniform in z), cylindrical coordinates are most
convenient, and only one spatial coordinate matters: r, the perpendicular distance from the axis.
This is sometimes called 1D cylindrical geometry:

 
1 d d(r)
rD(r) + r (r) (r) = S(r) , 1D cylindrical, heterogeneous (7.94)
r dr dr

 
1 d d(r)
D r + r (r) = S(r) , 1D cylindrical, homogeneous (7.95)
r dr dr

Given a symmetric problem in which things depends on both distance from an axis of symmetry
and the position along the axis (z coordinate), but not on the azimuthal coordinate around the
axis, cylindrical coordinates are most convenient, and only two spatial coordinates matter: r, the
perpendicular distance from the axis, and z, the height parallel to the axis. This is sometimes
called 2D cylindrical or r-z geometry:

   
1 (r, z) (r, z)
rD(r, z) D(r, z) + r (r, z) (r, z) = S(r, z) ,
r r r z z
2D cylindrical, heterogeneous (7.96)

2 (r, z)
   
1 d d(r, z)
D r + + r (r) = S(r) ,
r dr r z 2
2D cylindrical, homogeneous (7.97)

187
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.3 Steady-State One-Group Diffusion Solutions

7.7.3.1 Constant Source, Infinite Homogeneous Medium

Consider an infinite homogeneous medium (which means uniform material properties) with a uni-
formly distributed source emitting S n/(cm3 -s). In such a universe nothing depends on position.
Our diffusion equation does not have a diffusion term; it is simply:

r = S . (7.98)

The differential equation has become an algebraic equation. The solution is obvious:
S
= . (7.99)
r

Under what conditions does this solution make sense?

Only if r > 0.

Recall the definition of r in our one-group equations:

r a f .

f
r > 0 IF AND ONLY IF < 1. (7.100)
a

Let us explore this a bit further. Consider the k-eigenvalue problem for an infinite homogeneous
medium in the one-group case:
1
a = f .
k

f
k= k for 1-group medium (7.101)
a

It turns out we have discovered a general truth. The following is true regardless of geometry, and
it holds for all reasonable mathematical models, including energy-dependent transport, multigroup
transport, energy-dependent diffusion, multigroup diffusion, etc.

A steady-state solution exists in a fixed-source problem if and only if k < 1.

188
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.3.2 Constant Source, Homogeneous Slab

Consider a problem that is infinite and uniform in y and z, with a homogeneous material between
x = a/2 and x = a/2, with vacuum on the outside. Suppose there is a uniformly distributed
source emitting S neutrons per unit time per unit spatial volume. Then we have the following for
our steady-state one-group diffusion equation in this slab-geometry problem

d2 (x)  a a
D + r (x) = S , x , . (7.102)
dx2 2 2

We need boundary conditions. We have vacuum boundaries, so let us use extrapolation with
curvature, which means the function that ultimately describes the scalar flux inside the problem
would extrapolate to 0 a distance d outside the problem boundary:
 a 
d =0 (7.103)
 a2 
+d =0. (7.104)
2
Recall that unless otherwise specified we will use d = 2D:
   
a + 4D a + 4D
= =0. (7.105)
2 2

This defines an extrapolated width of our slab:


d=2D
a a + 2d a + 4D, (7.106)

which allows us to write the boundary conditions compactly:


 
a
=0. (7.107)
2

189
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Quick review of solution procedure for 2nd-order ordinary differential equations (ODEs):

PROCEDURE FOR SOLVING 2ND-ORDER ODE

1. Find a particular solutionanything that satisfies the given differential equation. (Try
simple functions, such as constants or linear functions, until you find one that works.)

2. Find the general form of the solution of the equation with no source. Let us agree to
call this the no-source solution. Mathematicians call this the homogeneous solution,
but we would like to use that term for materials that have the same properties at all
positionsmaterials that have effectively been homogenized.

3. Add the particular + no-source solutions. The no-source solution has arbitrary constants.
Apply the boundary conditions to the total (particular + no-source) solution to determine
the constants.

Let us follow this procedure for our problem:

1. Find a particular solution.


Guess part (x) = constant. Then dpart /dx = 0, and we find that the following simple
solution satisfies the differential equation:
S
part =
r

2. Find the general form of the solution of the equation with no source.
The no-source equation is:
d2 ns (x) r ns
2
= (x)
dx D
That is, we seek functions whose derivative is a positive constant times itself. It is easy to
see that exponentials satisfy this:
ns (x) = Aex/L + Cex/L ,
where
r
D
L the diffusion length. (7.108)
r

Equivalently, we could express the linear combination of exponentials as a linear combination


of hyperbolic (or hypergeometric) sine and cosine functions:
x x
ns (x) = A cosh + C sinh .
L L
This is usually more convenient for finite problems.

190
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

3. Add the particular + no-source solutions. The no-source solution has arbitrary constants.
Apply the boundary conditions to the total (particular + no-source) solution to determine
the constants.
We have
S x x
(x) = + A cosh + C sinh solution (7.109)
r L L

   
S a a
0= + A cosh + C sinh left bdy (7.110)
r 2L 2L
   
S a a
0= + A cosh + C sinh right bdy (7.111)
r 2L 2L

Recall that cosh and sinh are even and odd functions, respectively, so that cosh() = cosh() and
sinh() = sinh(). Given this, if we subtract the boundary-condition equations we find:

C=0

If we add the boundary-condition equations we find:


S 1
A= a
 (7.112)
r cosh 2L

The solution is therefore


" #
x
S cosh L 
(x) = 1 a
(7.113)
r cosh 2L

See Fig. 7.3.

7.7.3.3 Incident partial current, Homogeneous Slab

Consider a homogeneous slab that ranges from x = 0 to x = . This is a homogeneous half-


space. Suppose there is no fixed source, but there is an incident partial current at x = 0. Then
our problem statement is:

d2 (x)
D + r (x) = 0 , x (0, ) equation (7.114)
dx2

(x)x=2D = 4Jinc , left bdy (7.115)

(x )= finite, right bdy (7.116)

191
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Figure 7.3: One-group diffusion solution given uniform fixed source in bare homogeneous slab.
Horiz axis is x. It goes a/2 to a/2. Also mark a/2 and 0 and a/2. Vertical axis is (x). Show
S/r as dotted line. Show solutions for L = 0.05a, L = 0.2a, and L = a.

Particular Solution. The particular solution in a source-free problem or region is 0.

No-source solution In this infinite problem it is most convenient to represent the no-source solu-
tion (which is the entire solution since there is no source) using exponentials instead of hyperbolics:

(x) = Aex/L + Cex/L ,

The boundary condition at x immediately tells us that

A=0

The left boundary condition is then


h i
Cex/L = 4Jinc C = 4Jinc e2D/L (7.117)
x=2D

See Fig. 7.4.

192
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Figure 7.4: One-group diffusion solution given incident partial current on homogeneous half-space.
Horiz axis is x. It goes x = 2D to x = 3L. Also mark x = 0 and x = L and x = 2L. Vertical
axis is (x). Solution is decaying exponential.

Consider the same problem except for a finite slab ranging from x = 0 to x = a. There is no fixed
source, but there is an incident partial current at x = 0, and now we specify a vacuum on the right
side, at x = a. Then our problem statement is:

d2 (x)
D + r (x) = 0 , x (0, a) equation (7.118)
dx2

(x)x=2D = 4Jinc , left bdy (7.119)

(a + 2D)= 0, right bdy (7.120)

Particular Solution. The particular solution in a source-free problem or region is 0.

No-source solution There are many ways to represent the no-source solution. We will use expo-
nentials since we used them for a similar half-space problem:

(x) = Aex/L + Cex/L ,

The right boundary condition tells us:


h i
Aex/L + Cex/L = 0 A = Ce2(a+2D)/L (7.121)
x=a+2D

193
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Then the left condition tells us


h i
C ex/L e2(a+2D)/L ex/L = 4Jinc (7.122)
x=2D

or
4Jinc
C= 2(a+2D)/L
(7.123)
e2D/L e e2D/L

4Jinc e2D/L
= , (7.124)
1 e2a/L

where a = extrapolated slab width = a + 4D. See Fig. 7.5. Notice the changes relative to the
half-space solution. The coefficient of the decaying exponential is slightly larger, but there is also
a growing exponential that has a negative coefficient, so that the solution drops fast enough that
it extrapolates to zero at the right extrapolated boundary.

Figure 7.5: One-group diffusion solution given incident partial current on finite homogeneous slab.
Horiz axis is x. It goes x = 2D to x = a + 2D. Also mark x = 0 and x = a. Vertical axis is
(x). Solution is dominated by a decaying exponential but contains a growing exponential with a
negative sign.

194
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.3.4 Constant Source, Homogeneous Sphere

Consider homogeneous sphere of radius R, with vacuum on the outside. Suppose there is a uniformly
distributed source emitting S neutrons per unit time per unit spatial volume. Then we have the
following for our steady-state one-group diffusion equation in this 1D spherical geometry problem

 
1 d 2 d(r)
D 2 r + r (r) = S , r (0, R) . (7.125)
r dr dr

We need boundary conditions. We have vacuum boundaries, so let us use extrapolation with
curvature, which means the function that ultimately describes the scalar flux inside the problem
would extrapolate to 0 a distance d outside the problem boundary:
 
R = 0 (7.126)
 
d(r)
=0 OR (r)r=0 < , (7.127)
dr r=0

where R is the extrapolated radius:


usually
R R + d R + 2D.
In spherical problems it is often helpful to make the following substitution:

(r) = w(r)/r .

If we do so here, we obtain the following equation for w:

d2 w(r)
D + r w(r) = rS , r (0, R) . (7.128)
dr2
or
d2 w(r) 1 rS
+ 2 w(r) = , r (0, R) . (7.129)
dr2 L D
We already know the general solution of this equation, for it is almost identical to the slab-geometry
equation:
rS r r
w(r) = + A cosh + C sinh (7.130)
r L L

S 1h r  r i
(r) = + A cosh + C sinh (7.131)
r r L L

195
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We can apply either of the boundary conditions at r = 0, and we quickly find that:

A = 0.

Then the outer boundary condition says:


!
R S 1 R
0= + C sinh
r R L

S R
C=   (7.132)
r sinh R
L

Our final solution is:



S R sinh Lr
(r) = 1   . (7.133)
r r sinh R
L

This is sketched in Fig. 7.6.

Figure 7.6: One-group diffusion solution given uniform fixed source in homogeneous sphere. hor-
izontal axis is r, from 0 to R + 2D. Show r = R tick mark. Show solutions for L=0.05R, 0.2R,
and 0.5R. The smaller-L solutions will approach a flat S/r for r close to 0.

196
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.3.5 Constant Source, Homogeneous Infinitely Tall Cylinder

Consider homogeneous cylinder of radius R, infinite and uniform in z, with vacuum on the outside.
Suppose there is a uniformly distributed source emitting S neutrons per unit time per unit spatial
volume. Then we have the following for our steady-state one-group diffusion equation in this 1D
cylindrical geometry problem

 
1 d d(r)
D r + r (r) = S , r (0, R) . (7.134)
r dr dr

We need boundary conditions. We have vacuum boundaries, so let us use extrapolation with
curvature, which means the function that ultimately describes the scalar flux inside the problem
would extrapolate to 0 a distance d outside the problem boundary:
 
R = 0 (7.135)
 
d(r)
=0 OR (r)r=0 < , (7.136)
dr r=0

where R is the extrapolated radius:


usually
R R + d R + 2D.
Unlike the spherical case, there is no simple substitution in cylindrical geometry that will transform
the equation into an equivalent Cartesian-coordinate equation. However, this is a well-studied and
well-understood equation with a known no-source solution. Let us carry out the derivatives to put
the equation in a different form:

d2 (r) 1 d(r) n2
 
1 S
2
+ 2
+ 2 (r) = , r (0, R) . (7.137)
dr r dr L r D

with n = 0. In this form you may recognize that the no-source equation is

a modified Bessels equation.

Solutions are modified Bessel functions, In (r/L) and Kn (r/L), with n = 0. The complete solution
including the usual particular solution is:
S r r
(r) = + A I0 + C K0 (7.138)
r L L

Bessel functions are perhaps less familiar than exponentials, trigonometric functions, and hyper-
trigonmetric functions, but they are just as well behaved and just as well understood. Figure 7.7
shows the behavior of the zero-order Bessel functions, both ordinary (J0 (r) and Y0 (r)) and modified
(I0 (r) and K0 (r)).

197
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Figure 7.7: Ordinary and modified Bessel functions of order zero.

Note that K0 (r/L) as r 0, and that dI0 /dr = 0 at r = 0. It follows from these observations
and the boundary condition at r = 0 that

C = 0.
The outer boundary condition then immediately gives
S S 1
0= + A I0 (R/L) A= , (7.139)
r r I0 (R/L)

and our complete solution is


 
S I0 (r/L)
(r) = 1 (7.140)
r I0 (R/L)

This solution is sketched in Fig. 7.8.

198
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Figure 7.8: One-group diffusion solution given uniform fixed source in homogeneous cylinder of
infinite height. horizontal axis is r, from 0 to R + 2D. Show r = R tick mark. Show solutions for
L=0.05R, 0.2R, and 0.5R. The smaller-L solutions will approach a flat S/r for r close to 0.

7.7.3.6 Constant Source, Homogeneous Rectangle, Infinite in z

Consider homogeneous rectangle given by x (a/2, a/2) and y (b/2, b/2), infinite and uniform
in z, with vacuum on the outside. Suppose there is a uniformly distributed source emitting S
neutrons per unit time per unit spatial volume. Then we have the following for our steady-state
one-group diffusion equation in this x-y geometry problem

2 (x, y) 2 (x, y)
 
D + + r (x, y) = S , x (a/2, a/2), y (b/2, b/2). (7.141)
x2 y 2

199
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We need boundary conditions. Notice that this problem has mirror symmetry at the central
planes x = 0 and y = 0. We shall therefore solve the problem in only 1/4 of the domain, with
reflecting boundary conditions to account for the rest of the domain. We have vacuum boundaries on
the outside, so let us use extrapolation with curvature, which means the function that ultimately
describes the scalar flux inside the problem would extrapolate to 0 a distance d 2D outside the
problem boundary. The conditions:
   
(x, y) a
= , y = 0, y (0, b/2), (7.142)
x x=0 2

  !
(x, y) b
= x, = 0, x (0, a/2). (7.143)
y y=0 2

This equation has an important and fundamental difference compared to those we have solved so
far: It is a

partial

differential equation instead of an

ordinary

differential equation. This is a big difference! It makes the solution much more complicated to find
and describe. In particular, the complete no-source solution now has

infinitely many

linearly independent functionseach with its own unknown constantinstead of only two. As
a result, the solution will be expressed as an infinite sum of functions. In most problems of
practical interest, truncating the sum at a finite number of terms produces an acceptably accurate
approximate solution.

One powerful technique for solving partial differential equations, which means generating the infinite
sums that represent the solutions, is

eigenfunction expansion.

This method proceeds as follows:

1. Find the eigenvalues and eigenfunctions of the differential operator.

2. Verify that the eigenfunctions for a complete set in the function space of interest, so that
sources and solutions can be represented as linear combinations of those eigenfunctions.

3. Expand sources and solutions in eigenfunctions.

4. Use the linear independence of the eigenfunctions to find the coefficients in the solution
expansion.

200
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We shall follow this procedure for our one-group diffusion problem on a symmetric homogeneous
rectangle. First we must solve the eigenvalue problem:
 2
(x, y) 2 (x, y)

+ = B 2 (x, y), (7.144)
x2 y 2

subject to
   
(x, y) a
= ,y = 0, y (0, b/2), (7.145)
x x=0 2

  !
(x, y) b
= x, = 0, x (0, a/2). (7.146)
y y=0 2

Here B 2 is the eigenvalue and is the eigenfunction.

This problem is itself a partial differential equation, so how do we solve it? We employ a helpful
and powerful technique known as

separation of variables,

which simply means that we guess that the unknown function ( in this case) is the product of
individual functions of the individual independent variables (x and y in this case):
?
= X(x)Y (y). (7.147)

We insert this guess into our eigenvalue equation and manipulate:

d2 X(x) d2 Y (y)
 
Y (y) + X(x) = B 2 X(x)Y (y)
dx2 dy 2
or
1 d2 X(x) 1 d2 Y (y)
+ = B2 (7.148)
X(x) dx2 Y (y) dy 2
| {z } | {z }
function of x only function of y only

It follows that each term in the equation is a constant. (If we vary x, the 2nd and 3rd terms do
not change, so the first term cannot change with x either, etc.)

201
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We name the two newly discovered constants and collect all of the equations that must be satisfied:

1 d2 X(x) d2 X(x)
Bx2 = Bx2 X(x) , x (0, a/2), (7.149)
X(x) dx2 dx2

 
dX(x)
= X(a/2) = 0, (7.150)
dx x=0

1 d2 Y (y) d2 Y (y)
By2 = By2 Y (y) , y (0, b/2) (7.151)
Y (y) dy 2 dy 2

 
dY (y)
= Y (b/2) = 0, (7.152)
dy y=0

Bx2 + By2 = B 2 . (7.153)

Let us examine the problem for X. We first address the question of the sign of the eigenvalue,
Bx2 . If Bx2 < 0, then X would have to be a combination of sinh(Bx x) and cosh(Bx x). The
reflecting condition at x = 0 eliminates the sinh functions, leaving only cosh(Bx x) as a candidate
eigenfunction. But cosh() 1 for all real , so this could not satisfy the extrapolated boundary
condition.

We conclude that we must have Bx2 > 0, which means

X(x) = A cos(Bx x) + C sin(Bx x) . (7.154)

The reflecting condition eliminates the sine function. The extrapolated condition then says:

A cos(Bx a/2) = 0 . (7.155)

We cannot have A = 0, for this would mean X(x) = 0, which is not a legitimate eigenfunction (by
definition). Instead we must find the values of Bx that make the boundary condition work when
A 6= 0. There are infinitely many:
i
Bx,i = , i = 1, 3, 5, . . . (7.156)
a
The complete solution of the X problem is therefore
 
ix
Xi (x) = Ai cos , i = 1, 3, 5, . . . (7.157)
a
 2
2 i  2
Bx,i = = i2 , i = 1, 3, 5, . . . (7.158)
a a

202
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

The Y solution is similar


 
jy
Yj (x) = Cj cos , j = 1, 3, 5, . . . (7.159)
b
 2  2
2 j 2
By,j = =j , j = 1, 3, 5, . . . (7.160)
b b
It follows that the solutions of our original eigenvalue problem are
   
ix jy
ij (x, y) = Aij cos cos , i = 1, 3, 5, . . . ; j = 1, 3, 5, . . . (7.161)
a b
 2  2
2 i j
Bij = + , i = 1, 3, 5, . . . ; j = 1, 3, 5, . . . (7.162)
a b
Note that:

1. Each eigenfunction satisfies all boundary conditions. Thus, any linear combination of the
eigenfunctions will also satisfy the boundary conditions.

2. The constants Aij are not determined by the eigenvalue problem. Solution amplitudes are
never determined by eigenvalue problems. The eigenvalue problem is satisfied no matter what
value of Aij is included in Eq. (7.161).

3. The eigenfunctions are orthogonal:


a/2 b/2
ab 2
dx dy ij (x, y)mn (x, y) = im jn A (7.163)
0 0 8 ij

4. The eigenfunctions form a complete set: essentially any function that satisfies the boundary
conditions can be represented as a linear combination of the {ij }.

We expand the solution and the source as linear combinations of the eigenfunctions:
X X
(x, y) = ij ij (x, y) (7.164)
i=1,3,5,... j=1,3,5,...
X X
S= Sij ij (x, y) (7.165)
i=1,3,5,... j=1,3,5,...

and we insert these expansions into the diffusion equation:


 2
ij (x, y) 2 ij (x, y)
X X    X X
ij D + + r ij (x, y) = Sij ij (x, y)
x2 y 2
i=1,3,5,... j=1,3,5,... i=1,3,5,... j=1,3,5,...

2:
or, recognizing that ij is an eigenfunction of the Laplacian operator with eigenvalue Bij
X X X X
2
 
ij DBij + r ij (x, y) = Sij ij (x, y) (7.166)
i=1,3,5,... j=1,3,5,... i=1,3,5,... j=1,3,5,...

203
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

The eigenfunctions ij are

linearly independent.

Thus, the only way that the linear combination on the left can equal the linear combination on the
right is for the coefficients of each function to be the same:
2
 
ij DBij + r = Sij , i = 1, 3, 5, . . . ; j = 1, 3, 5, . . . (7.167)

Sij
ij = 2 . (7.168)
r + DBij
Our solution will be complete once we find the Sij . Recall the expansion:
X X
S(x, y) = Sij ij (x, y) (7.169)
i=1,3,5,... j=1,3,5,...

Multiply both sides by mn (x, y) and integrate over the extrapolated domain, recognizing that
S(x, y) = S inside the physical domain and S(x, y) = 0 outside of it:
a/2 b/2  mx   
ny
dx dy S cos cos
0 0 a b
a/2 b/2  mx       
X X ny ix jy
= Sij dx dy cos cos cos cos
0 0 a b a b
i=1,3,5,... j=1,3,5,...

Evaluate each side:


a/2 b/2  
a  a m  b b n
dx dy S cos(mx/a) cos(ny/b) = S sin sin
0 0 m a 2 n b 2
 
ab  a m  b n
=S sin sin
mn 2 a 2 b 2

a/2 b/2  mx       
X X ny ix jy
Sij dx dy cos cos cos cos
0 0 a b a b
i=1,3,5,... j=1,3,5,...

Eq. (7.163) ab
Smn (7.170)
16
Thus,
 
16  a m  b n
Smn = S sin sin (7.171)
mn 2 a 2 b 2
This completes our solution:
 
sin bb j a i

X X (16) sin 2 a 2

ix
 
jy

(x, y) = S 2 2 cos cos (7.172)
(ij )(r + DBij ) a b
i=1,3,... j=1,3,...

204
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We illustrate in Fig. 7.9 how this expression converges to the exact solution as we keep more terms
in the infinite sum.

Figure 7.9: One-group diffusion solution given uniform fixed source in homogeneous rectangle,
infinite in z, with vacuum boundaries. In this example S = 1 n/cm3 -s, D = 0.5 cm, r = 0.1 cm1
d = 2D, a = 28 cm, and b = 48 cm. Approximate solutions are shown with the sums in Eq. (7.172)
truncated at i = j =: 3 (top left), 7 (top right), 15 (bottom left), and 25 (bottom right). The
bottom right solution is accurate enough that its plot is essentially indistinguishable from a similar
plot of the exact solution.

205
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.3.7 Constant Source, Homogeneous Slab with k = 1

Consider a problem that is infinite and uniform in y and z, with a homogeneous material between
x = a/2 and x = a/2, with vacuum on the outside. Suppose there is a uniformly distributed
source emitting S neutrons per unit time per unit spatial volume. Then we have the following for
our steady-state one-group diffusion problem in this slab-geometry problem

d2 (x)  a a
D + r (x) = S , x , , (7.173)
dx2   2 2
d
=0, (7.174)
dx x=0
 
a
=0. (7.175)
2

We have solved this problem before under the assumption that r = a f > 0, which means
k = f /a < 1. Let us now explore the solutionif it existsfor the case of k = 1, which
means r = 0. In this case our equation becomes

d2 (x)
D =S (7.176)
dx2
and our solution must therefore be a quadratic function of x (because its second derivative is
constant):

(x) = A0 + A1 x + A2 x2 . (7.177)
The mirror condition at x = 0 implies that

A1 = 0.

The equation itself implies that


S
D(2A2 ) = S A2 = . (7.178)
2D
The vacuum boundary condition then says
S
0 = A0 + A2 (a/2)2 A0 = A2 (a/2)2 = (a/2)2 . (7.179)
2D
and the solution is therefore the concave-down parabola
"  #
S a 2 2
(x) = x . peak grows with a; S/r = (7.180)
2D 2
We sketch this solution in Fig. 7.10. Recall that in problems with k < 1, in portions of the
problem domain that are more than a few diffusion lengths (a few L) from boundaries, the
solution asymptotes to an upper limit of max = S/r . This does not happen herethe peak flux
increases as the square of the slab width.

206
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Figure 7.10: One-group diffusion solution given uniform fixed source in bare homogeneous slab of
material with k = 1. Horiz axis is x. It goes 0 to a/2. Also mark a/4 and a/2. Vertical axis is
(x). Solution is concave-down parabola going from S/2D a2 /4 to 0.

7.7.3.8 Constant Source, Homogeneous Slab with k > 1 (Really!)

Consider once more a steady-state one-group diffusion problem in a symmetric homogeneous slab.
We have solved this problem before under the assumption that r = a f > 0, which means
k = f /a < 1. We have also solved it for k = 1, which means r = 0. We now ask under
what circumstancesif anya steady state solution might exist if k > 1 r < 0. In this case
we have
d2 (x)  a
D ( f a ) (x) = S , x 0, , (7.181)
dx2 | {z } 2
>0

   
d a
= =0. (7.182)
dx x=0 2

The particular solution is negative:


S
part (x, y) = <0. (7.183)
f a

207
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

The no-source solution is now sines and cosines:

ns (x) = A cos(Bm x) + C sin(Bm x) , (7.184)

2 f a
where Bm (7.185)
D
The mirror condition at x = 0 implies that

C = 0.

The vacuum boundary condition then says


S S 1
0= + A cos(Bm a/2) A= . (7.186)
f a f a cos(Bm a/2)

For this solution to be physically meaningful, the constant A must be > 0, because the particular
solution is negative. Moreover, the function cos(Bm x) must remain positive for x inside the problem
domain. That is, a physically meaningful steady-state solution exists only if the following condition
is met:

Bm a/2 < /2 Bm < . (7.187)
a
If this condition is metwhich later we will see means k < 1then the following steady-state
solution exists given a fixed source in a slab of material whose k > 1:
 
S cos(Bm x) f a  2
(x, y) = 1 , only if < . (7.188)
f a cos(Bm a/2) D a

We sketch this solution in Fig. 7.11. Recall that in problems with k < 1, in portions of the
problem domain that are more than a few diffusion lengths (a few L) from boundaries, the
solution asymptotes to an upper limit of max = S/r . In problems with k = 1 the solution is
a parabola. Now we see that in problems with k > 1 but k < 1, the solution is a constant times
a cosine function minus another constant. As the slab width approaches the critical width, the
constant multiplying the cosine function grows without bound, and the solution becomes a pure
cosine function except very near the boundary.

208
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Figure 7.11: One-group diffusion solution given uniform fixed source in bare homogeneous slab of
material with k > 1 but k < 1. Horiz axis is x. It goes 0 to a/2. Also mark a/2. Vertical axis is
(x). Solution is a cosine function minus a constant, and typically the constant will be negligibly
small except near the boundary.

7.7.3.9 Two-Region Slab

Consider a problem that is infinite and uniform in y and z, with a symmetry plane at x = 0, the
homogeneous material C between x = 0 and x = a, the homogeneous material R between
x = a and x = b, and vacuum on the outside. Suppose there is a uniformly distributed source
emitting S neutrons per unit time per unit spatial volume only in material C. Then we have the
following for our steady-state one-group diffusion equation in this slab-geometry problem

d2 C (x)
DC + C C
r (x) = S , x (0, a) , (7.189)
dx2
d2 R (x)
DR + R R
r (x) = 0 , x (a, b) , (7.190)
dx2
need to insert a figure!!!

209
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Boundary and interface conditions:

dC
 
=0, (7.191)
dx x=0

C (x)x=a = R (x)x=a , (7.192)

dC dR
   
DC = DR , (7.193)
dx x=a dx x=a

 
R b + 2DR R b = 0 .

(7.194)

Particular solutions:
S
part,C (x) = , (7.195)
C
r
part,R (x) = 0 . (7.196)

No-source solutions:
q
ns,C (x) = AC cosh(x/LC ) + CC sinh(x/LC ) , LC DC /C
r , (7.197)
q
ns,R (x) = AR cosh((b x)/LR ) + CR sinh((b x)/LR ) , LR DR /R
r . (7.198)

The reflecting condition immediately gives

CC = 0.

The extrapolated condition immediately gives

AR = 0.

Then the interface conditions are:


S
+ AC cosh(a/LC ) = CR sinh((b a)/LR ) , (7.199)
C
r

DC DR
AC C
sinh(a/LC ) = CR R cosh((b a)/LR ) . (7.200)
L L
Everything in these equations is known except for the constants AC and CR . It is easy to see that
CR > 0 and AC < 0. Also, note that the solution in region C will be concave down while the
solution in region R will be concave upward. A sketch is illustrated in Fig. 7.12.

210
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Figure 7.12: One-group diffusion solution in a two-region symmetric slab with source in the inner
region, no source in the outer region, and vacuum on the outside. Constant cosh function from
0 to a, then sinh(b x) function from a to b.

7.7.3.10 Point Source in Infinite Homogeneous Medium

Suppose at r = 0 in an infinite homogeneous medium there is an infinitesimally small source of


neutrons emitting

S0 [units = n/s].

This is a 1D problem in spherical coordinates. Away from r = 0, the problem is source-free, and
we have:
 
1 d 2 d(r) 1
2 r + 2 (r) = 0 for r 6= 0 . (7.201)
r dr dr L

The general solution of this homogeneous ODE is

A exp(r/L) + C exp(r/L)
(r) = (7.202)
r
The boundary conditions are as follows:

1. as r +, the flux must remain finite

2. at r = 0, we need a condition related to the point source

211
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

By our finiteness at + requirement we see that

C = 0.

Thus, all that remains is finding A. To do this, we take a close look at the neutron leakage rates
(i.e., the rate at which neutrons cross defined surfaces) near the source point. Consider the net
neutrons per s crossing a spherical shell of radius r, in the outward direction. We know that this
is just
h i
~ r) , or 4r2 Jr (~r) .
4r2 ~er J(~ (7.203)

We also know that this is just

crossing rate outward crossing rate inward .

As r approaches 0, the second term vanishes, because the surface area gets vanishingly small. But
all of the source neutrons must still cross outward. Thus,

lim 4r2 Jr (r) = S0 . (7.204)


r0

We can use this to determine A. In spherical coordinates we have

~ r) = D d(~r)
h i
Jr (r) = ~er D(~ (7.205)
dr

Thus,
  
2 d A exp(r/L)
2
S0 = lim 4r Jr (r) = lim D4r
r0 r0 dr r
  
1 1
= lim D4r2 A exp(r/L) 2 +
r0 r rL

= D4A (7.206)

Thus,

S0 exp(r/L)
(r) = , r 6= 0 . (7.207)
4D r

Diffusion solution, point source in infinite homogeneous medium

212
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Note that, as r 0, the flux behaves like 1/r and goes to infinity at r = 0!!!.

Question: How can this be? Did we make an error?

Answer: This physically impossible flux is caused by the physically impossible source, which emits
a finite neutrons from 0 volume. Note that the outgoing partial current density crossing a spherical
surface around the origin must as the surface gets close to the origin.

Even with the scalar flux at r = 0, we still obtain physically reasonable answers for observable
quantities, such as total number of neutrons in a finite volume or total path-length rate in a finite
volume. For example, how many neutrons are within a distance R of the source?
R
2 (r) S0 R
number = 4r dr = r exp(r/L)dr
0 v vD 0
S0
= [1 (1 + R/L) exp(R/L)] (7.208)
va

If we let R we find a total of


S0
va neutrons

in the entire infinite medium.

If we consider small R, so that R/L = d << 1, then

S0 2
R/L=<<1 S0 d
number = [1 (1 + R/L) exp(R/L)] (7.209)
va va 2

That is, the number of neutrons within a small distance d of the source is proportional to d2 , which
shrinks to zero for small d even though the scalar flux becomes infinite at the source point.

7.7.4 The Diffusion Length (L)

The diffusion length L, defined as

r
D
L = , [cm]
a

= diffusion length (7.210)

does have a physical interpretation. It is proportional to an average straight-line distance between


neutron birth point and absorption point, at least in an infinite medium.

213
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Recall that if Stot neutrons/second are born at a point in an infinite medium, the scalar flux a
distance r always is (in the diffusion approximation) given by:

S0 exp(r/L)
(r) = , (7.211)
4D r
a result we derived a few pages ago. The absorption-rate density a distance r from the source is
therefore

S0 a exp(r/L)
A.R.D. = (7.212)
4 D r
and the absorption rate between r and r + dr is the A.R.D. times the volume of the spherical shell
between r and r + dr:

S0 a exp(r/L) 1
4r2 dr = S0 2 exp(r/L)rdr (7.213)
4 D r L

Thus, the probability of absorption between r and r + dr, where r is the distance from a point
source in an infinite medium, is just the absorption rate divided by the source rate:
1
p(r)dr = exp(r/L)rdr . (7.214)
L2

Recall: General definition of an average:



w(x)f (x)dx
hf i = (7.215)
w(x)dx

In our case, the variable is r weight function is the probability density function for absorption, p(r):

p(r)f (r)dr
hf i = 0 (7.216)
0 p(r)dr

The denominator = 1: the probability that a neutron gets absorbed somewhere in the infinite
medium is unity:

1
exp(r/L)rdr = 1 (7.217)
0 L2

214
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

If we let f (r) = r2 , so we are computing the mean squared distance from birth point to
absorption point, we get


2 1
r = exp(r/L)r3 dr = = 6L2 . (7.218)
0 L2
This r2 is the

average squared straight-line distance (not path traveled)


between birth and absorption points in an infinite homogeneous medium.

We see from Eq. (7.218) that


1
2
L2 = r . (7.219)
6
We conclude that:

The diffusion length, L, is related to how far neutrons migrate from birth point to absorption
point, on average.

If the diffusion length is long, neutrons tend to travel a long way. If it is short, they dont. If you
have a material with a large diffusion length, you will need a lot of it to keep the neutrons from
escaping!

7.7.5 Greens functions

We can write the point-source solution in terms of Cartesian coordinates, with the point r~0 defined
as the location of the point source:

S0 |~rr~0 |/L
(~r) = e . (7.220)
4D

The flux due to a superposition of sources is the superposition of the individual fluxes due to each
~0
source. A distributed source can be viewed as a superposition of point sources. If S r is source
 
rate density at position r~0 , then S r~0 d3 r0 is effectively a point source. So the solution given a
distributed source in an infinite medium is:

  exp ~r r~0 /L
 

(~r) = S r~0 d3 r0 (7.221)


~0
4D ~r r

215
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

The
 function
 that relates a source at one point to the flux at another is called a Greens function,
~0
G ~r, r . For an infinite homogeneous medium, we see that the one-group-diffusion Greens
function is:  
  exp ~r r~0 /L
G ~r, r~0 = . (7.222)
4D ~r r~0

A line source or plane source can be viewed as a collection of infinitesimally small sources. As a
result, we could use Eq. (7.221) to generate the plane-source and line-source solutions we found
previously. It would be helpful for you to try this yourself!

7.7.6 One-Group k-Eigenvalue Diffusion Solutions

7.7.6.1 k-Eigenvalue Solution: Bare Homogeneous Slab

We consider first a bare, homogeneous slab reactor (infinite in y and z), of width a in the x
direction. Bare means it is

surrounded by nothing but vacuum.

Homogeneous means

all material properties are constant.

We want to know whether this reactor is critical, and also what the neutrons are doing in it. To
find out, we must solve the following eigenvalue problem:

d2 n (x) 1  a a
D + a n (x)= f n (x) , x ; , (7.223)
dx2 kn 2 2

n (e
a/2)= n (e
a/2) = 0 . (7.224)

where
d2D
a = extrapolated slab width a + 2d a + 4D
e

with d is the extrapolation distance. (Unless you are told otherwise, use d = 2D. Remember this!)
We have introduced the subscript n because we know there will be an infinite number of solutions
(each of which is a {k, (x)} pair), and we need a way to distinguish one solution from another.

Let us rewrite our equation for the nth {eigenvalue, eigenfunction} solution as follows:
f
d2 n (x) kn a  a a
+ n (x) = 0 , x ; (7.225)
dx2 D 2 2

216
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

and define the number Bn2 :


f 1 f k
kn a kn a 1 kn 1
Bn2 = = (7.226)
D L2 L2
This gives us
d2 n (x)  a a
+ Bn2 n (x) = 0 , x ; (7.227)
dx2 2 2

Note and remember: there is a simple one-to-one relationship between Bn2 and the eigenvalue
kn . In fact, we can think of Bn2 as an eigenvalue itself. Equation (7.227) shows, in fact,
that Bn2 is an eigenvalue of the negative of the Laplacian operator. That is, B 2 is an eigenvalue
of 2 . From the definition of Bn2 , Eq. (7.226), we find that kn expressed in terms of B 2 as:

f
kn = (7.228)
a + DBn2

Our equation and its boundary conditions cannot be satisfied by an arbitrary value of Bn2 ; only
certain values that are characteristic of the reactor size and shape will work. We must find those
values.
Let us solve the problem. It is clear that a sine function or a cosine function will satisfy our
equation, because B 2 > 0 (as we showed in a previous section, and as is obvious from Eq. (7.226)
if we recognize that k > kn ). It follows that the most general n that satisfies the equation is a
linear combination of sines and cosines:

n (x) = An sin(Bn x) + Cn cos(Bn x) (7.229)

Let us apply the boundary conditions:


   
a a
An sin Bn + Cn cos Bn =0 (7.230)
e e
2 2
   
a a
An sin Bn + Cn cos Bn =0 (7.231)
e e
2 2

If these equations had to be true for all constants B, the only solution would be

the trivial solution: An = Cn = 0.

(To see this, add the equations. The sine terms cancel. If the result had to be true for all Bn this
would mean Cn = 0. If you subtract the equations you find that if the result had to be true for all
Bn , then it would mean An = 0.)

217
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We are not interested in zero solutions. Instead, we seek those characteristic values of Bn that allow
nontrivial solutions. If we add the boundary-condition equations (7.230 and 7.231), we obtain
 
a
2Cn cos Bn =0 (7.232)
e
2

Either Cn or the cosine term must be zero. The cosine term = 0 for the following values of Bn :
n
Bn = , n = 1, 3, 5, (7.233)
a
e

These are some of the eigenvalues of our slab problem. The associated eigenfuntions are
 nx 
Cn cos(Bn x) = Cn cos (7.234)
a
e
where Cn is any constant (and n is odd).

If we subtract the boundary-condition equations (7.230 and 7.231) we obtain


 
a
2An sin Bn =0. (7.235)
e
2

Either An or the sine term must be zero. The sine term will be zero for the following values of Bn :
n
Bn = , n = 2, 4, 6,
a
e

These are the rest of the eigenvalues of our slab problem. The associated eigenfuntions are
 nx 
An sin(Bn x) = An sin ,
a
e
where An is any constant (and n is even).
Thats it. We have completely solved the eigenvalue problem specified by Eq. (7.225), and we have
obtained the following:
f f
kn = 2
= 2 , (7.236)
a + DBn a + D n
a
e

nx

Cn cos a
e , n = 1, 3, 5, . . . ,
n = (7.237)
nx

An sin , n = 2, 4, 6, . . . ,

a
e

where An and Cn are constants that are not determined by the eigenvalue problem (because eigen-
value problems do not determine eigenfunction amplitudesonly shapes).
Now we will discuss the solution at some length.

218
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.6.2 Modes

The nth eigenfunction is often called the

nth mode.

There is only one eigenfunction that can be positive throughout the problem domain. It is called
the

fundamental mode.

In our problem the fundamental mode is


 x 
fundamental mode = C1 cos(B1 x) = 0 cos
a
e
We have recognized that the constant C1 is the value of the flux at the center of the slab, so we
have renamed it 0 . Remember, this constant is not fixed by the eigenvalue problem. Sometimes
we will omit this constant when we talk about eigenfunctions.

Sometimes we expand the solution of a steady-state or time-dependent problem in terms of eigen-


functions. It is often useful, then, to think and communicate in terms of the modes of the
solution.
A very important special case is the source-free steady-state reactor. Remember this:

In a source-free steady state reactor, only the fundamental mode is present.

Note that eigenvalue problems tell us only the shapes of the modes. Even if we were told that
our reactor is source-free and steady-state, we would still need more information to determine the
amplitude or magnitude of the solutionthat is, to determine

the value of C1 .

The additional information might be the total reactor power, the value of the peak scalar flux,
the total net outleakage rate, etc.something that nails down the amplitude of the flux.

In Figure 7.13 we sketch the first few eigenfunctions (modes) of our bare homogeneous slab
reactor. As noted previously, only the fundamental mode is positive throughout the domain. Note
that the fundamental mode is less leaky than the others: it has a

smaller gradient at the boundaries.

219
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

-0.5 -0.3 -0.1 0.1 0.3 0.5

-1

Figure 7.13: Sketch of first 3 eigenmodes for a 1D slab centered at the origin. axes are (n)(x) and
x/e
a.

7.7.6.3 Important Concepts, Terms, and Definitions

In this subsection we define some terms that nuclear engineers find very useful in analyzing, de-
signing, and discussing reactors. You should understand them all.

7.7.6.3.1 Geometric buckling Recall the eigenvalue equation that we have been working on,
and write it for the nth {eigenvalue, eigenfunction} pair:

d2 n (x)
+ Bn2 n (x) = 0 (7.238)
dx2

Divide through by n (x):

1 d2 n (x)
Bn2 = (7.239)
n (x) dx2

Note:

1. Bn2 is a measure of solution curvature.

2. Every Bn2 > 0. Thus, where n is positive it is concave down, and where n is negative it is
concave up. Where n = 0 there is an inflection point. Check this by examining Fig. 7.13.

220
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We have a name for the Bn2 associated with the fundamental mode:

Fundamental-mode Bn2 Bg2 geometric buckling. (7.240)

The geometric buckling, Bg2 , is a measure of how much the fundamental mode is curved, or buck-
led. In a bare slab we have
 2
Bg2 = geometric buckling, bare homogeneous slab (7.241)
a
e
If you forget this, bad things are likely to happen (for example, to your grade). (Build your
intuition: how does curvature change as the reactor width increases? What is the curvature of the
fundamental mode in an infinite medium?)

7.7.6.3.2 Connection with k The problem we wrote down originally was

d2 n (x) f n (x)
D 2
+ a n (x) = (7.242)
dx kn

We have written it here explicitly denoting the nth {kn , n } pair. We then wrote it in a slightly
simpler-looking form by defining
1 f 1
kn a 1 kn f a
Bn2 = = (7.243)
L2 D
We also recognized that:
f
a k f
kn = = = (7.244)
1 + L2 Bn2 2 2
1 + L Bn a + DBn2

The kn associated with the fundamental mode has a special name:

Fundamental kn keff = k-effective = multiplication factor . (7.245)

It is obvious that we have:


f
a f
keff = = , (7.246)
1 + L2 Bg2 a + DBg2

because Bg2 is always the smallest of the Bn2 eigenvalues.

221
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.6.4 Materials Buckling

If our reactor is critical, then keff = 1 and

f a
Bg2 = (7.247)
D

The term on the left is geometric buckling, which depends only on reactor geometry (shape and
size). (In a bare slab, the only geometry information is the extrapolated slab width.) The term
on the right depends only on

material properties.

We assign this term a special name:


f a 2
= Bm = materials buckling, (7.248)
D

If the two bucklings are equalthat is if the material properties and the reactor geometry are a
perfect match for each otherthen:

the reactor is critical.

In general,
2
Bm < Bg2 subcritical
2
Bm = Bg2 critical
2
Bm > Bg2 supercritical

The critical relationship is easy to remember. To easily remember the others, imagine what must
happen when f gets larger or smaller (which makes the materials buckling larger or smaller).

The above relations apply to any bare homogeneous reactor, regardless of shape (according to
one-group diffusion theory). Only the definition of Bg2 , which of course depends on geometry, will
change.

222
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

IMPORTANT NOTE

We know intuitively that to design a critical reactor, we must perfectly match:

{material properties} with {reactor size & shape}.

We see now that according to one-group diffusion theory, for bare homogeneous reactors,

The only geometry property that matters is the geometric buckling, Bg2 , and
2 .
The only material property that matters is the materials buckling, Bm

If these two numbers are equal, the reactor is critical (according to one-group diffusion theory).

7.7.6.5 Bare Homogeneous Sphere

We consider next a bare homogeneous spherical reactor. Bare means surrounded by nothing but
vacuum (no neutrons entering from outside). Homogeneous means all material properties are
constant (the same everywhere in the sphere). We want to know whether or not this reactor is
critical, and also what the neutrons are doing in it. We must solve the following eigenvalue problem:

 
1 d 2 dn (r) 1
D 2 r + a n (r) = f n (r) , r [0; R) (7.249)
r dr dr kn

n (R)
e = 0 [thats 1 BC] (7.250)

dn
= 0 or n (r)|r=0 < [thats the 2nd BC] (7.251)
dr r=0

where
R
e = extrapolated radius = R + d ,

with d is the extrapolation distance. (Unless you are told otherwise, use d = 2D.) Note that you
have a choice of conditions you can impose at r = 0. In a symmetric sphere we know that the
derivative of the scalar flux must equal zero at the origin; this is one condition you could impose.
Alternatively, since there are no singular sources in this problem we know that the scalar flux
remains finite everywhere in the sphere, including at r = 0. Sometimes it is easier to impose
this condition, and it gives the same answer.

We have recognized, by using the subscript n, that our eigenvalue problem will have many {eigenvalue,
eigenfunction} pairs as solutions.

223
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

If we bring all terms to the left side of Eq. (7.249) and divide through by (D), we obtain
 
1 d 2 dn (r)
r + Bn2 n (r) = 0 for x [0; R]
r2 dr dr

where we have defined B 2 just as before:


f f
kn a a kn 1
Bn2 = = (7.252)
D L2
Note that there is a one-to-one relationship between Bn2 and the eigenvalue kn . In fact, we can
think of Bn2 as an eigenvalue itself. Eigenvalue means characteristic value. Our equation cant be
satisfied by just any old Bn2 ; only certain characteristic values will work.

Lets solve the problem. You can verify (by plugging it in) that the following satisfies our equation:

An sin(Bn r) + Cn cos(Bn r)
n (r) = (7.253)
r
(You can derive this solution by making the following substitution: n (r) = w(r)/r, which will
produce a slab-like equation for w(r). Try it!)
Now we apply boundary conditions. First consider r = 0. As r 0, sin(Bn r)/r Bn , so the
term multiplied by An remains finite. However, as r 0, cos(Bn r)/r . Thus, the solution
can remain finite at r = 0 only if
Cn = 0 . (7.254)

Now the vacuum boundary condition looks like:

An sin(Bn R)
e
=0 (7.255)
Re

If this equation had to be true for all constants Bn , the only solution would be the trivial solution,
An = 0. We are not interested in zero solutions. Instead, we seek those certain values of B
characteristic values for this geometrythat allow nontrivial solutions. The sine term will be zero
for the following values of Bn :
n
Bn = for n = 1, 2, 3, . . .
Re

The squares of these values, Bn2 , are the eigenvalues of our spherical Laplacian eigenvalue problem.
The associated eigenfunctions are:

An sin(Bn r)
n (r) = (7.256)
r
where An is any constant.

224
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Thats it. We have completely solved the diffusion k-eigenvalue problem for a bare homogeneous
sphere, and we have found:
f f
kn = =  2 , (7.257)
a + DBn2
a + D n
e R

 
1 nr
n (r) = An sin . (7.258)
r R
e

Now we will discuss the solution.

The one eigenfunction that does not change sign anywhere in the problem domain is called the
fundamental mode. In our problem the fundamental mode is
   
1 r R r
fundamental mode = fund (r) = A1 sin = 0 sin (7.259)
r Re r Re

We have replaced the constant A1 in terms of 0 , which is the value of the flux at the center of the
sphere. This value is not determined by the eigenvalue problemeigenvalue problems determine
only the

shape

of the eigenfunctions.

In a given problem, if the flux is expanded in terms of eigenfunctions, then the amplitudes of the
various modes must be determined by other conditions. For example, we may be told the reactor
is operating in steady state at a certain power level. If Eq. (7.259) describes the thermal flux,
then the reactor power is approximately the fast-fission factor () times the power produced by
thermal-neutron-induced fission. (Think about it!) This gives us an expression that we can solve
for the constant A1 , or equivalently the constant 0 :

P = known power [energy per unit time]

  R  
200 MeV Re r
0 f,th sin 4r2 dr . (7.260)
fission 0 r Re

Here we have recognized that the volume element in spherical coordinates= 4r2 dr.

We sketch the first few eigenfunctions of our spherical reactor in Figure 7.14. Note that the
fundamental mode is the only one that doesnt change sign in the reactor volume, and note that
the modes become increasingly oscillatory as n increases. That is, higher-n modes oscillate positive-
negative-positive over shorter distances.

225
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

0.5

0
0 0.2 0.4 0.6 0.8 1

-0.5

Figure 7.14: Sketch of the first 3 eigenmodes for a bare homogeneous spherical reactor. axes are
n (r) and r/R.
e

Recall that:

The Bn2 associated with the fundamental mode Bg2 = geometric buckling.

It is a measure of how much the fundamental mode is curved, or buckled. In a bare sphere we
have
 2
2
Bg = geometric buckling, bare sphere (7.261)
R
e

Recall the connection with the k-eigenvalues:


f
a kn 1 f
Bn2 = kn = (7.262)
L2 a + D Bn2

And of course, we still have:


f
f a
keff = klargest = = (7.263)
a + D Bg2 1 + L2 Bg2

226
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

These relations apply to any bare homogeneous reactor, regardless of shape, given the one-
group diffusion approximation. Only the definition of Bg2 , which of course depends on geometry,
will change.

7.7.6.6 Bare Shoebox (= Brick = Rectangular Parallelepiped)

In a bare homogeneous rectangular parallelepiped (brick- or shoebox-shaped) reactor, our diffusion


k-eigenvalue equations are:

2 (x, y, z) 2 (x, y, z) 2 (x, y, z)


 
D + +
x2 y 2 z 2

f
+ a (x, y, z) = (x, y, z) (7.264)
k

h a ai  b b  h c c i
for {x, y, z} , , , . (7.265)
2 2 2 2 2 2

along with the extrapolated boundary conditions

(e
a/2, y, z) = (e a/2, y, z) = 0 , (7.266)
(x, eb/2, z) = (x, eb/2, z) = 0 , (7.267)
(x, y, e
c/2) = (x, y, e
c/2) = 0 , (7.268)

where the e quantities are extrapolated widths & heights & lengths:

a = a + 2d ,
e eb = b + 2d , c = c + 2d ,
e (7.269)

and d is the extrapolation distance. (Unless you are told otherwise, use d = 2D.)

There is a fundamental difference between this and our previous k-eigenvalue problems: our equa-
tion is now a

partial

differential equation. How do we solve it? A standard approach to partial differential equations
one that works in many problems of interestis guess a separable solution. That is, we guess
that the unknown function, which in our case is (x, y, z), is the product of functions of single
variables:
guess
(x, y, z) X(x) Y (y) Z(z) . (7.270)

227
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

We insert this hypothesis into our equation (noting that the derivatives become full derivatives
because each function depends on a single variable):

d2 X(x) d2 Y (y) d2 Z(z)


 
D Y (y)Z(z) + X(x)Z(z) + X(x)Y (y)
dx2 dy 2 dz 2
f
+ a X(x)Y (y)Z(z) = X(x)Y (y)Z(z) (7.271)
k

and then divide through by X(x)Y (y)Z(z):

1 d2 X(x) 1 d2 Y (y) 1 d2 Z(z)


 
f
D 2
+ 2
+ 2
= a (7.272)
X(x) dx Y (y) dy Z(z) dz k

We also divide through by D, and as before we define


f
a k 1
B2 = (7.273)
L2

This produces
1 d2 X(x) 1 d2 Y (y) 1 d2 Z(z)
+ + = B 2 (7.274)
X(x) dx2 Y (y) dy 2 Z(z) dz 2

Note:

the first term depends only on x,

the second term depends only on y,

the third term depends only on z,

the right-hand side doesnt depend on x, y, or z.

But if the second, third, and fourth terms dont change when we vary x, the first term doesnt
change either (because they are in an equation together). This (and similar reasoning for y and z)
means that every term is a constant! We give these constants names:

1 d2 X(x)
= constant Bx2 , (7.275)
X(x) dx2
1 d2 Y (y)
= constant By2 , (7.276)
Y (y) dy 2
1 d2 Z(z)
= constant Bz2 . (7.277)
Z(z) dz 2

228
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

If we add these three equations we see that we must have the following relationship between these
constants and the B 2 defined above:

B 2 = Bx2 + By2 + Bz2 . (7.278)

Equations 7.275 through 7.277 are three separate eigenvalue problems of the type we have already
solved. For example, the X problem looks like:

d2 X(x)
+ Bx2 X(x) = 0
dx2
with  
a
X =0
e
2

The Y and Z problems are similar. We have already solved these problems! Each is just a slab-
geometry problem! The solutions are:

i
Bx,i = , i = 1, 2, 3, . . . (7.279)
a
e

Cx,i cos ix

a
e , i = 1, 3, 5, . . .
Xi (x) = (7.280)
ix

Ax,i sin , i = 2, 4, 6, . . .

a
e

j
By,j = , j = 1, 2, 3, . . . (7.281)
eb
 
jy
C y,j cos , j = 1, 3, 5, . . .
b

e
Yj (y) =   (7.282)
Ay,j sin jy ,

j = 2, 4, 6, . . .

e b

and

`
Bz,` = , ` = 1, 2, 3, . . . (7.283)
c
e

Cz,` cos `z

c
e , ` = 1, 3, 5, . . .
Z` (z) =
`z

Az,` sin , ` = 2, 4, 6, . . .

c
e

229
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Recall that our original equation was

2 2 2
+ 2 + 2 + B2 = 0
x2 y z

where
f
a k 1
B2 (7.284)
L2

2
We see now that ij` (x, y, z) and Bij` satisfy this equation if

(i, j, `)-th eigenfunction (mode) = ij` (x, y, z) = Xi (x)Yj (y)Z` (z) (7.285)

 2  2  2
2 2 i j `
(i, j, `)-th B eigenvalue = Bij` = + + (7.286)
a
e eb c
e

2 f
(i, j, `)-th k eigenvalue = kij` = 2 (7.287)
a + DBi,j,`

We have solved the problem! So: What is the fundamental mode, and what is the geometric
buckling?

 
x
 y z

fundamental mode: fund (x, y, z) = 111 (x, y, z) = [constant] cos a cos cos c
e eb e

 2
2 2
geometric buckling: Bg2 = B111
2 =
 
a + + c
e b
e e

Fundamental Mode and Geometric Buckling,


Bare Homogeneous Rectangular Parallelepiped

What is keff for a bare homogeneous brick-shaped reactor, according to diffusion theory?

f f
keff = = (7.288)
a + DBg2
  2 
2 2
 
a + D a + + c
e e b e

230
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.7.6.7 Bare Homogeneous Infinite Cylinder

We consider next a bare, homogeneous infinitely-long cylindrical reactor of radius R. We want


to know whether or not this reactor is critical, and also what the neutrons in it are doing. We
have to solve the same kind of eigenvalue problem we have already solved for slabs, spheres, and
bricks. The only difference is the form of the Laplacian operator. In an infinitely long homogeneous
cylinder we have the following eigenvalue problem:
 
1 d dn (r) f n (r)
D r + a n (r) = for r < R (7.289)
r dr dr kn

n (R)
e = 0 (7.290)

dn
=0 or n (r)|r=0 < (7.291)
dr r=0

where Re = extrapolated radius = R + d, with d is the extrapolation distance. (Use d = 2D unless


told otherwise.) Note that the Laplacian has 1/r and an r factors that the Cartesian Laplacian
does not have. Also note that there is a choice of conditions to impose at the origin. The finiteness
condition is fair game because there are no singular sources in this problem.

Let us rewrite our equation as we have done before:


 
1 d dn (r)
r + Bn2 n (r) = 0 for r < R (7.292)
r dr dr

where weve defined B 2 as usual


f
kn a
Bn2 = (7.293)
D
At this point in our slab-geometry analysis we recognized that the solutions of our differential
equations were sines and cosines. At this point in our spherical-geometry analysis we recognized
that we could perform a substitution, (r) = w(r)/r, to find that the solutions were 1/r sines
and cosines. But cylindrical geometry is slightly different. Its solution will not involve sines and
cosines.

To proceed, let us carry out the derivatives in the Laplacian term and thus rewrite our equation in
a form that you may recognize:

d2 n 1 dn (r)
+ Bn2 0 n (r) = 0 for r < R

2
+ (7.294)
dr r dr

We now recognize that this is an

ordinary Bessels Equation

231
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

Given that B 2 > 0, the solution is a linear combination of

ordinary Bessel functions

of the first and second kinds (see special-function appendix of this set of notes, or see any reference
about Bessel functions):
n (r) = An J0 (Bn r) + Cn Y0 (Bn r) (7.295)

If we were solving fixed-source steady-state subcritical problems, then 1/L2 would play the role
of B 2 in our equations, and our solution would be a combination of modified Bessel functions.
This is much like the slab and sphere cases in which the eigenvalue problem involves sines
and cosines but the fixed-source problem involves hyperbolic sines and cosines. That is, the
modified Bessel functions can be thought of as the hyperbolic versions of Bessel functions,
and in fact this terminology has been used in the past.

Back to our problem. The boundary condition at r = 0 requires that the derivative of the scalar
flux be 0 or that the scalar flux be finite. The function Y0 (z) is unbounded when z 0, whereas
J0 (z) remains bounded and has a zero derivative as r 0. It follows that we must have

Cn = 0 .

and therefore
n (r) = An J0 (Bn r) . (7.296)

Our extrapolated boundary condition tells us that either

An = 0 [which doesnt interest us]


or
 
J0 Bn Re =0

Thus, the eigenvalues Bn are the numbers such that

Bn R
e = n , where n = the nth zero of J0 .

This is analogous to Bn being the n-th zero of the sine function in the spherical problem. The
zeroes of the sine are simple: the sine function = 0 when its argument = n for n = 1, 2, 3... The
zeroes of J0 are not that easy to write down, but the concept is the same and the zeroes are known.

The smallest zero of J0 is


min = 2.405 . . . (7.297)
See Fig. 7.7 to see how the J0 function behaves and note that it crosses the axis when its argument
2.405.

232
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

The geometric buckling of a bare, infinitely long homogeneous cylinder, according to diffusion
theory, is therefore
 2
2.405 . . .
Bg2 = geometric buckling, infinitely long cylinder (7.298)
Re

The fundamental mode is always the eigenfunction associated with the smallest B 2 (which is
the geometric buckling):
 
2.405 . . .
fund (r) = A0 J0 r fund. mode, infinitely long cylinder (7.299)
Re

7.7.6.8 Bare Homogeneous Finite Cylinder

We consider next a bare, homogeneous cylindrical reactor of radius R and height H. We want
to know whether or not this reactor is critical, and also what the neutrons in it are doing. We
have to solve the same kind of eigenvalue problem we have already solved for slabs, spheres, and
bricks. The only difference is the form of the Laplacian operator. In an infinitely long homogeneous
cylinder we have the following eigenvalue problem:

2 n (r, z)
   
1 n (r, z) f n (r, z)
D r + 2
+ a n (r, z) = ,
r r r z kn
 
H H
r < R, z , , (7.300)
2 2

 
H H
n (R,
e z) = 0, z , (7.301)
2 2

!
dn H
e He
= 0 or n (r)|r=0 < , z , , (7.302)
dr r=0 2 2

!
H
e
n r, = 0, r < R
e, (7.303)
2

where Re = extrapolated radius = R + d and H e = extrapolated height = H + 2d, with d is the


extrapolation distance. (Use d = 2D unless told otherwise.) Note that the Laplacian has 1/r
and an r factors that the Cartesian Laplacian does not have. Also note that there is a choice

233
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

of conditions to impose at the origin. The finiteness condition is fair game because there are no
singular sources in this problem.

Let us rewrite our equation as we have done before:

2 n (r, z)
   
1 n (r, z)
r + + Bn2 n (r) = 0 (7.304)
r r r z 2

where weve defined B 2 as usual


f
kn a
Bn2 = (7.305)
D
As with the other bare homogeneous shapes, the solution of this problem boils down to finding
the eigenvalues and eigenfunctions of the Laplacian, with the associated boundary conditions. As
with the shoebox problem, we can attack this problem with separation of variables. We find, not
surprisingly, that each eigenfunction is the product of a function of z and a function of r. The z
functions are sines and cosines. The r functions are Bessel functions. The mechanics are left as an
exercise. The fundamental-mode eigenvalue and eigenfunction are listed in the table later in this
chapter.

7.8 When Can You Trust Diffusion Theory?

Remember: Diffusion theory is based on the diffusion approximation. It is not perfect. However,
we have argued that it is reasonably accurate in many practical problems. But how do you know
when to trust it and when not to?

You can trust diffusion theory to give an accurate solution if all of the following are true:

1. The region of interest is


many mean-free paths from boundaries or material interfaces.

2. The region of interest is many mean-free paths from


strong localized sources.

3. Scattering and/or fission are the dominant reactions in the region of interest.

In particular: in a large, homogeneous nuclear reactor, diffusion theory is pretty good!

234
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.9 Homogeneous-Reactor Eigenvalues and Eigenfunctions

Table 7.2: Geometric Bucklings & Fundamental Modes, Bare Homogeneous Reactors.

Geometry Extrapd dims. Bg2 f und


Slab  2  x 
cos
a = a + 4D a a

Sphere  2  
Re r
sin
R = R + 2D R
e r Re

cylinder 2  
min r

min J0
R = R + 2D R R

Finite Cyl.    
min r z
 2  2 J0 cos
R = R + 2D min R H
+
R H
H = H + 4D

Brick
 2    x   
a = a + 4D
 2 2 y  z 
+ + cos cos cos
a b c a b c
b = b + 4D
c = c + 4D

Here min 2.405 . . . = the smallest zero of the J0 Bessel Function.

235
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

7.10 Summary

Our basic goal in this course is to gain an understanding of how reactors behave and why they
behave that way. This requires, at the very least, an understanding of neutron production and loss
rates in reactors. In this chapter we went a long way toward that. We found that:

1. The neutron transport equation (either one-speed or energy-dependent) comes from simply
counting neutrons. It is essentially exactthe things it ignores are not important in reactors.

2. We usually make the multigroup approximation to convert the transport equation (or its
approximations) from energy-dependent to multigroup, where the latter is a finite set one-
group equations, coupled by the fission and scattering terms.

3. The neutron transport equation is difficult to solve. If we take its 0th angular moment, we
~ If we had another equation to relate the net
obtain a balance equation that contains and J.
~
current density (J) and the scalar flux (), we could potentially avoid solving the full-blown
transport equation.

4. Ficks Law, or the diffusion approximation, can be used to relate J~ and :


~ r, E, t) = D(~r, E, t)(~
J(~ ~ r, E, t) (7.306)

It is an approximation. It is a reasonably good approximation if all of the following conditions


are met:

(a) away from sources,


(b) away from material interfaces (including problem boundaries),
(c) away from strong localized sources,
(d) when << s + f ,
You must know when the diffusion approximation is reasonable!

5. If we approximate J~ in the balance equation using the diffusion approximation, we get the
diffusion equation. (Either one-speed or energy-dependent.)

6. Each term in the diffusion equation is a physically-meaningful rate density:

~ D(~r, t)(~
(a) ~ r, t) = net outleakage rate density [n/(cm3 -s)],
(b) a (~r, t)(~r, t) = absorption rate density [n/(cm3 -s)],
(c) f (~r, t)(~r, t) = fission-neutron production rate density [n/(cm3 -s)],
(d) Sext (~r, t) = extraneous or fixed source rate density [n/(cm3 -s)].

Similar interpretations hold for the energy-dependent diffusion equation, although in that
case each term is a density in energy as well as space, and thus has units n/(cm3 -s-MeV).

7. On the boundary of a diffusion problem we must specify:

236
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

(a) ,
~
(b) ~en
~
(c) or a combination of and ~en

in order to have a completely specified (well-posed) problem.

8. At an interface between different materials, we have continuity of

(a) scalar flux,


(b) normal component (~en J~ = D~en )
~ of net current density.

9. We applied the diffusion equation to many problems of interest. First we studied fixed-source
diffusion problems:

(a) for both homogeneous and multi-region problems,


(b) with localized or distributed sources

10. To solve a fixed-source diffusion problem, we

(a) write the diffusion equation for each homogeneous region,


(b) write all interface and boundary conditions,
(c) find a particular solution and the no-source solution; add them to get the general solution,
(d) use interface and boundary conditions to solve for the constants in the general solution.

11. We also solved k-eigenvalue diffusion problems. To write down a k-eigenvalue diffusion equa-
tion you can follow these steps:

(a) Write down the diffusion equation that would be true if the reactor were critical and
operating in steady state. This means there is no fixed source and no time variable. It
also means you dont have to distinguish between prompt and delayed neutrons.
(b) Multiply the fission term by 1/k.

12. To solve a k-eigenvalue diffusion problem for a bare homogeneous reactor, we

(a) write the diffusion k-eigenvalue equation, which contains the Laplacian for the particular
geometry of your reactor,
(b) write appropriate boundary conditions,
(c) define B 2 and obtain 2 (~r) + B 2 (~r) = 0,
(d) find all of the (B, ) pairs that satisfy both 2 (~r) + B 2 (~r) = 0 and the boundary
conditions,
(e) recognize that Bg2 = geometric buckling = the smallest of the B 2 values, that keff =
multiplication factor = the k value associated with that smallest B 2 value, and that
the fundamental-mode scalar flux, fund (~r), is the eigenfunction associated with that
smallest B 2 value and thus with the multiplication factor, keff .

237
NUEN 601, M.L. Adams notes, 2017 Chapter 7. The Neutron Diffusion Equation

13. Here is a (long-ish) review on boundary/interface conditions

(a) The diffusion is a second-order differential equation (it has second derivatives in spatial
variables).
(b) As a consequence, in one-dimensional problems (those that have only a single spatial
variable), its solution, for each homogeneous material zone, is expressed as a linear
combination of two independent solutions, plus the particular solution.
(c) Thus, we need two constraints per homogeneous material zone. Eg., for a 3-zone reactor,
we need 6 constraints.
(d) In heterogeneous media, we have at each interface, the continuity of the scalar flux
and continuity of the normal component of the current (2 interface conditions
per interface). That is, for a 3-zone slab or spherical reactor, there are 2 interfaces,
hence 4 constraints. We still need 2 more conditions to solve (next item below).
(e) No matter what configuration, you always need to apply 2 boundary conditions (per
dimension). Let us summarize them in detail:
i. Infinite medium. The flux must remain finite [unless there are singular sources, as
in the point-source problem where the flux goes to infinity at the origin].
ii. Vacuum : use the extrapolated boundary condition or use no incoming partial
current of neutrons.
iii. The presence of localized singular sources can also lead to the derivation of a
special type boundary condition know as the source condition, which signifies
that the net leakage rate [not the current density] of a small volume surrounding
the source is equal to the source intensity.

We end with the usual disclaimer: This was not a complete list of everything you need to know
from this chapter.

238
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Chapter 8

Time Dependence

8.1 Introduction

We know that a reactors behavior depends on the gain and loss rates of neutrons in the reactor.
We know that gain and loss rates depend on how the neutrons are distributed spatially in the
reactor, distributed in energy, and distributed in direction. These distributions can change with
time.
Other chapters address how neutrons distribute themselves in position, energy, and direction. In
this chapter we study how neutron populations evolve in time. We first recall the equations that
describe how delayed-neutron precursor populations evolve in time. We then consider the one-group
diffusion description of time-dependent neutronics for a homogeneous reactors. From the precursor
and one-group diffusion equations we shall derive an important set of equations:

the point-reactor kinetics equations (PRKEs).


We then turn to a much more accurate description of neutronics, namely the energy-dependent
transport equation for heterogeneous reactors. We show that (remarkably) we can use this nearly
perfect equation along with the delayed-neutron precursor equations to derive the PRKEs in exactly
the same form that we obtained from one-group diffusion theory for a homogeneous reactor.
The PRKEs tell us a great deal about how neutron populations evolve over time in nuclear reactors.
Some common terminology includes:

Reactor dynamics: Analysis of time-dependent behavior of the reactor, including but not
limited to the neutron population. This is a general, all-encompassing term.

Reactor kinetics: Analysis of time-dependent behavior of the neutron population, with at


most a simple treatment of other phenomena (such as heat transfer and fluid flow).

Point-Reactor Kinetics: Reactor kinetics analysis that assumes that only the amplitude of
the neutron distribution changes with time, with the shape of the distribution (in position,
energy, and direction) remaining fixed. The term point comes from the zero-dimensional
treatment of what is really a six-dimensional phase space.

239
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

In this chapter we will focus mostly on kinetics but will also touch on the coupling to other phe-
nomena (getting into the more general dynamics), including fission-product buildup and depletion
as well as feedback from temperature changes.

8.2 Phenomena and Time Scales

Time dependence of reactor state variables (including neutron distribution, temperature distribu-
tion, nuclide concentrations, etc.) can be induced by a variety of phenomena, and many of these
phenomena also play a role in the time evolution of the state variables after a transient has been
initiated. These phenomena act on vastly different time scales, and this introduces different time
scales into the time-dependent behavior of the reactor. Some examples are below.

8.2.1 Fuel burnup

Initially loaded fissionable material is consumed in an operating reactor at a rate proportional to


the power of the system. If fertile materials, such as 232 Th or 238 U, are present, fissile nuclei will be
produced (and also consumed) as the reactor operates. Thus, the fuel composition changes in time,
and this effect naturally has an impact on the multiplication factor and the neutron distribution.
If the reactor operates for long enough and fuel is not replaced, it must eventually deplete its fuel
and become unable to sustain a chain reaction.
Significant fuel burnup takes place on a time scale of

weeks or months.

in most reactors. There are exceptions:

The High-FLux Isotope Reactor (HFIR) must replace its core

every 3 full-power weeks!

The NSC core lasts for decades.

Question: What determines how long it takes to incur significant fuel burnup?

Answer: Power per unit mass of fuel

8.2.2 Fission product buildup and decay

Some fission products and/or their daughters have high cross sections for

absorption.

240
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

As they are produced and destroyed in the reactor they can significantly affect k and thus signif-
icantly affect the time behavior of the population. Significant changes can take place on a time
scale of

hours.

Recall the buildup and decay of 135 Xe and 135 I following reactor shutdown, for example, as illus-
trated in Fig. 8.1.

Figure 8.1: Illustration of time scale for fission products, using 135 I and 135 Xe. (Put hours on horiz.
axis

8.2.3 Control-rod motion and related changes

Control-rod motion is an example of an externally-initiated significant change that can take place
on a time scale of

seconds.
In a PWR that controls the multiplication factor partially through the use of

soluble boron
in the coolant, dilution or boration can have a significant effect on a time scale of

minutes.
Changes in coolant flow rate or coolant incoming temperature can occur on time scales of

seconds or minutes,

with significant impact on neutron multiplication.

241
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

8.2.4 Temperature changes

The temperature in some reactor components can change on

sub-second.

time scales. For example, fuel temperature can increase approximately as fast as power can increase.
In some transients this can be very fast.

Can fuel temperature decrease as fast as power can decrease? Think about it . . .

The temperature in other components, such as coolant and structural materials, changes more
slowly (why? think about it and understand it!), but can be significant on time scales of

seconds or minutes.

8.2.5 Accidents

Accidents can be viewed, for our purposes in this chapter, as the triggering events that cause
externally-initiated changes of significance to the neutron population. For example, a control-rod
ejection accident is a special case of control-rod motion. A coolant pipe break initiates a special
case of a change in coolant flow and inlet temperature.

8.2.6 Summary of time scales

Time scales of significant changes in neutron population have a tremendous range. The shortest
time scale that we usually encounter in reactor dynamics is the

prompt-neutron lifetime,

which is usually in the range of

107 to 104 s.
[consider exercise: 1MeV n, sea of U235, 19g/ml, sig: el 4.6, f 1.23, c 0.09; e-8 s].

with the shortest lifetimes corresponding to fast-spectrum systems and the long lifetimes corre-
sponding to thermal-spectrum systems.

The longest time scale is usually the scale of fuel burnup, which ranges from

weeks to years

Note that if we are interested in reactor behavior in a time range of a few seconds, we can ignore
changes in burnup, fission-product inventories, and other slowly-varying parameters. That is, we
need to model only the phenomena that are important on the time scale we study.

242
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

8.3 Delayed Neutron Precursors

Recall the definitions:

Ci (~r, t) = expected density [number/volume] of type-i precursors at position ~r at time t.


i (t) = decay constant of type-i precursors.
i (E) = energy spectrum of neutrons emitted after decay of type-i precursors.
di (~r, E, t) = expected number of type-i precursors produced from a fission
caused by a neutron of energy E at position ~r at time t . (8.1)

The definition of i means that i (E)dE is the fraction of type-i delayed neutrons that are emitted
with energies in dE at E.

Also note that

each precursor ultimately produces exactly one delayed neutron

This follows from our definition of precursor. Thus, di is also the expected number of delayed
neutrons that will ultimately be emitted from type-i precursors, per fission caused by a neutron of
energy E. If we use the definitions above we can obtain the conservation equation for precursors
of type i:


d
Ci (~r, t) = dE di (~r, E, t)f (~r, E, t)(~r, E, t) i Ci (~r, t) . (8.2)
dt 0

8.4 Time-Dependent One-Group Diffusion

The one-group diffusion approximation applied to a time-dependent neutronics problem is:

1 (~r, t) ~ h i
~ r, t) + a (~r) (~r, t) = p f (~r) (~r, t) + Sf ixed (~r, t)
D(~r)(~
v t

# of dnp
X types
+ i Ci (~r, t) . (8.3)
i=1

We see that

The scalar flux depends on the delayed-neutron precursor concentrations .

The delayed-neutron precursors concentrations depend on the scalar flux .

243
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Thus, we have a

coupled system

of equations to solve. While this may look like a complicated system that could be difficult to solve,
it is at least a well-posed set of equations that has a unique solution (given initial conditions for
the scalar flux and the precursor concentrations).

Our initial explorations will be for

homogeneous reactors

in which we take the cross sections and diffusion coefficient to be independent of position and time
during the time period of interest. In this case our coupled system becomes:

# of dnp
X types
1 (~r, t)
= D2 (~r, t) + [p f a ] (~r, t) + i Ci (~r, t) + Sf ixed (~r, t) . (8.4)
v t
i=1

d
Ci (~r, t) = di f (~r, t) i Ci (~r, t) . (8.5)
dt

8.5 Reactor Kinetics Equations

Let us see what we can learn about solutions of Eqs. (8.4) and (8.7). Reminder: these are for a
homogeneous reactor, we have made the diffusion approximation, and we are using only one energy
group.

Rewrite the equations in terms of neutron density:

# of dnp
X types
n(~r, t)
= Dv2 n(~r, t) + v [p f a ] n(~r, t) + i Ci (~r, t) + Sfixed (~r, t) . (8.6)
t
i=1

d
Ci (~r, t) = v di f n(~r, t) i Ci (~r, t) . (8.7)
dt

244
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Define total quantities, integrated over the reactor volume



ntot (t) d3 rn(~r, t) , (8.8)
V ol


Stot (t) d3 rSfixed (~r, t) , (8.9)
V ol


Ci,tot (t) d3 rCi (~r, t) . (8.10)
V ol
Define the

Delayed-neutron fraction, ,
as follows:

d p
=1 (8.11)

= fraction of fission neutrons born with delay. [1G diffusion, homog. reactor]

Here we have defined


# dnp
Xtypes
d di (8.12)
i=1

and noted that p + d = .


Similarly, we define
di
i (8.13)

= fraction of fission neutrons that are born delayed, from precursors of type i (8.14)

If we integrate the time-dependent diffusion equation over the reactor volume, we now have:


d
ntot (t) = Dv d3 r2 n(~r, t)
dt V ol

# of precursor
X types
+ v [p f a ] ntot (t) + Stot (t) + i Ci,tot (t) . (8.15)
i=1

245
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Let us multiply and divide the leakage term by v ntot :



Dv V ol d3 r2 n(~r, t)

Dv 3 2
d r n(~r, t) = v d3 r n(~r, t) . (8.16)
v d3 r n(~
r , t)
V ol V ol V ol

The term in brackets has an unambiguous physical interpretation. The numerator is the net
inleakage rate (the negative of the net outleakage rate). (To see this, recall that in the diffusion
approximation for a homogeneous reactor, Dv2 n = D2 = ~ J,
~ and the volume integral of
~ J~ equals the surface integral of ~en J.)
~ The denominator is the rate at which neutrons are
making tracks in the reactor. (It is the volume integral of the total scalar flux.) The ratio in the
brackets is therefore the negative of:

[the net number of neutrons that


leak out of reactor per unit path length of neutron travel]

Recall the physical interpretation of the absorption cross section:

number of neutrons absorbed per unit path length of neutron travel

We see that the term in brackets has much in common with a . The difference is that its mechanism
for removing neutrons is leakage, or neutron escape. In recognition of this, we define the following
short-hand notation for the bracketed ratio:


Dv V ol d3 r2 n(~r, t)
esc . [Escape Cross Section, 1G diffusion] (8.17)
v V ol d3 r n(~r, t)

Our volume-integrated system of equations can now be written as follows.


# of dnp
X types
dntot (t)
= v [p f a esc ] ntot (t) + i Ci,tot (t) + Stot (t) . (8.18)
dt
i=1

d
Ci,tot (t) = v di f ntot (t) i Ci,tot (t) . (8.19)
dt

We manipulate the first term on the right-hand side of the volume-integrated diffusion equation:
" p f # " #
a +esc 1 (1 ) a +fesc 1
v [p f a esc ] = v 1 = 1 (8.20)
a +esc v(a +esc )

246
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Two ratios in this expression have important physical meanings. One is the multiplication factor
of the reactor, provided that the escape cross section is defined using the fundamental-mode
neutron distribution. To see this, recall the equation satisfied by the fundamental-mode scalar flux
and the multiplication factor (which we shall just call k, omitting the eff subscript):

1
D2 fund (~r) + a fund (~r) = f fund (~r) (8.21)
k


d3 rf fund (~r)
k= 3
V ol
r) V ol d3 D2 fund (~r)
V ol d ra fund (~

f f
= =
D V ol d3 2 fund (~
r) Dv V ol d3 2 nfund (~r)
a + a +
3 v V ol d3 rnfund (~
V ol d rfund (~
r) r)

f
k= (8.22)
a + fesc
und

where fesc
und
is defined to be the escape cross section under the condition that the neutrons are in
the fundamental mode.

The second ratio is the average time from neutron birth to death. Note that neutrons ultimately
either leak out (with expected number per unit neutron travel distance given by esc ) or get
absorbed (with expected number per unit neutron travel distance given by a ). This means that
a + esc = expected number of neutrons lost per unit neutron travel distance, which means:
1
= the loss mean-free path. (8.23)
a + esc

It follows that the mean time from neutron birth to loss is this distance divided by v, or:
1
= average lifetime of a prompt neutron.
v(a + esc )

`p . (8.24)

247
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Using these defined quantities we can rewrite our volume-integrated diffusion equation as follows:
# of dnp
X types
" #
dntot (t) (1 ) a +fesc 1
= 1 ntot (t) + i Ci,tot (t) + Stot (t) .
dt v( + )
a esc i=1

  # of dnp
X types
(1 )k 1
= ntot (t) + i Ci,tot (t) + Stot (t)
`p
i=1

  # of dnp
X types
k 1 k
= ntot (t) + i Ci,tot (t) + Stot (t)
`p
i=1

# of dnp
X types
" #
k1
k
= ntot (t) + i Ci,tot (t) + Stot (t) . (8.25)
`p /k
i=1

We define

k1
reactivity = , (8.26)
k

`p
mean generation time = (8.27)
k

= expected time for N neutrons to have N descendants . (8.28)

We manipulate a term in the precursor equation:

v di f = vi f
f
a +esc k
= i 1 = i
v(a +esc )
`p

i
= (8.29)

248
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

With all of these definitions and manipulations, we can write our volume-integrated one-group
diffusion equation and our volume-integrated precursor equation in the following famous form
with a famous name:

POINT-REACTOR KINETICS EQUATIONS (PRKEs)

  # prec.
Xtypes
d
ntot (t) = ntot (t) + Stot (t) + i Ci (t) , (8.30)
dt
i=1

d i
Ci (t) = ntot (t) i Ci (t) . (8.31)
dt

Several remarks are in order:

The quantity is defined in terms of the defined parameter k. The quantity is defined in
terms of both k and the parameter `p .

Both k and `p are defined in terms of given cross sections, a and f , and the defined
parameter esc .

The parameter esc is defined in terms of the neutron density or scalar flux distribution,
n(~r, t) or (~r, t), which in general depends on time. Thus, in general, the exact value of esc
will depend on time, as will parameters (including and ) that depend on esc .

In many interesting problems, and are very nearly constant during interesting portions
(or all) of the transient, because only the amplitude (not spatial shape) of n(~r, t) is changing
during those portions of the transient. Thus, it is useful for us to study the solution these
coupled first-order ODEs in this interesting constant-parameter case. This case is simple
enough to permit analytic solutions that are enormously helpful in building an understanding
of nuclear-reactor behavior.

We began with the time-dependent one-group diffusion equations and the time-dependent
delayed-neutron precursor equations. If esc is defined in terms of the exact solution, n(~r, t),
then the PRKEs do not contain any approximations that were not already imbedded in the
diffusion and precursor equations.

As we shall see, we can derive exactly the same PRKEs beginning from our most accu-
rate description of neutron behaviorenergy-dependent-transporteven for the most com-
plicated, heterogeneous reactor. That is, these equations are not restricted to one-group
diffusion theory for homogeneous reactors. The only difference is in the definitions of the
coefficients i , , , and .

249
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

8.6 PRKE Solutions

We began with equations for functions that depended on both position and time [(~r, t) and Ci (~r, t)].
We have now written those equations in the form of

coupled first-order ODEs ,

with solution functions [ntot (t) and Ci (t)] that depend only on time (because we integrated over
all positions). The PRKEs are used extensively to study the time-dependent behavior of nuclear
reactors. We can learn a lot about the behavior of the neutron population by studying and solving
these equations! Let us see what they tell us.

8.6.1 Constant Coefficients

We first consider the case in which , , i , and are constant. (During many transients of interest,
this is a very good approximation for large segments of the transient, so this is an important case to
understand.) In this case of constant coefficients the Point-Reactor Kinetics Equations (PRKEs)
are a set of coupled first-order ODEs with constant coefficients. We know from previous
mathematical studies that:

1. The solution of each equation is the sum of

a particular solution plus a homogeneous (no-source) solution.

2. The no-source solution of each equation is a sum of exponentials.

Given our assumption that Stot and the coefficients are constants, it is easy to show that constant
particular solutions satisfy the PRKEs (with the exception of one tricky case). You should verify
that the following solutions do satisfy the equations:
i particular
Ciparticular = n , (8.32)
i tot

 

nparticular
tot = Stot . (8.33)

You can see that there is a problem with this particular solution if the reactivity, , is zero. This
is the tricky casethe case of a critical reactorwhich we will address later. Now let us address
the homogeneous solutionthe solution that satisfies the equations with Stot removed. If we
define
I = number of types of delayed-neutron precursors that we are tracking, (8.34)

250
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

then the PRKEs are a set of I + 1 coupled ODEs (with constant coefficients in the case we are
studying right now). We know from our previous mathematics studies that each homogeneous
solution (ntot (t), C1 (t), C2 (t), . . . , CI (t)) will be

a sum of I + 1 exponentials.

and that the time constants in the exponentials for a given function (such as C2 (t)) are

the same as the time constants in the other functions.

Thus, our PRKE solutions have the following form (excluding the tricky = 0 case and continuing
to assume that Stot is constant in time):
I+1
X
ntot (t) = Aj e s j t Stot . (8.35)

j=1

I+1
X i particular
Ci (t) = Ci,j esj t + n
i tot
j=1

I+1
X i
= Ci,j esj t Stot . (8.36)
i
j=1

But how do we find the constants sj , j = 1, , I+1? This is not terribly complicated: we insert
the solutions into the equations and find the s values that allow the solutions to be non-trivial
(i.e., the solutions that can have non-zero Aj and Ci,j coefficients in front of the exponentials in
the solutions). After a bit of algebra (okay, more than just a bit) we find that each of the s values
must satisfy the following equation:

I
`p s 1 X i s
= + . (8.37)
1 + `p s 1 + `p s s + i
i=1

INHOUR EQUATION

There are exactly I + 1 distinct values of s that satisfy this equation, which is called the inhour
equation. This equation gives us the s values that go into Eqs. (8.35-8.36). These values deter-
mine how quickly or slowly each part of the solution changes with time. Note that the s values
have units of

inverse time

In the early days of reactor analysis, people often used inverse hours as the unit for the s values.
This led to the name in-hour, which the equation still bears.

251
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

It is not difficult to show, for example by a graphical analysis outlined in Fig. 8.2, that the I+1
values of s have the following properties:

I of them are always negative, and in fact < min .

The other one (call it s1 ) has the same sign as the reactivity, .

If = 0, then s1 = 0.

If < 0, then s1 is negative but is closer to 0 than the other s values:

s1 (min , 0) .

Figure 8.2: Graphical solution of the inhour equation. RHS(s) goes from at s > 1 to 1 at
s ; from to + between the i lines; and drops from + at 1/`p to 1 as s .
Use dashed lines to indicate the s values, where = RHS(s).

That is, I values of s will always be real and negative; the other value, which is the largest
algebraically, is real and has the same sign as the reactivity, . If < 0, there is one s value that
can never be more negative than min , where min is the shortest of the precursor decay constants.
This corresponds to the precursor with the longest half-life, which is87 Br, with T1/2 = 55 s. Thus,

252
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

1
min .
80s

This simple information about the s values tells us a great deal about the time variation of the
neutron population in a reactor. We explore this for the six possible cases in the subsections that
follow.

8.6.2 Subcritical Reactor with Fixed Source


If the reactor is subcritical, then < 0 and, according to the discussion above, all exponential time
coefficients {sj , j = 1...I + 1} are negative. It follows that for large t, the exponentials become
vanishingly small, and we are left with only the particular part of the solution:

large t `p /k `p
ntot (t) Stot (t) = Stot = Stot . (8.38)
(k 1)/k 1k
That is, we have discovered that

In a source-driven subcritical system,


the neutron population eventually reaches a steady state .

The steady-state neutron density is proportional to:

the source strength,

the prompt-neutron lifetime, `p ,

1/(1 k).
The factor of (1k) in the denominator describes the phenomenon of

subcritical multiplication.
We see that because of subcritical multiplication, if k is arbitrarily close to 1, then the steady-state
neutron population can be arbitrarily large, even with a small source and a subcritical reactor.
Note further that for k close to unity, is close to zero and thus s1 will be close to zero. This
means that one exponential in the expression for ntot will be slow to decay away, which means it
will take a long time for the neutron population to attain its steady-state level.
8.6.3 Supercritical Reactor with Fixed Source
This case is relatively simple. The particular solution is a constant, I of the exponentials decay
with time, and one of the exponential terms has a positive s value. The increasing exponential
eventually dominates all other terms, and
large t
ntot (t) A1 es1 t . (8.39)

253
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

That is, we have discovered that

In a source-driven supercritical system,


the neutron population eventually increases exponentially with time .

8.6.4 Critical Reactor with Fixed Source


If the reactor is critical, then = 0 and, according to the discussion above, all exponential time
constants {sj } are negative except for one that is zero. However, we find that a constant particular
solution does not satisfy the equations when =0. (The 1/ factor in Eq. (8.33) is a hint that this
doesnt work!) We try the next-simplest function we can imagine, namely one that is linear in t.
We find then that the following particular solution does satisfy the equations with =0:

nparticular
tot (t) = tCp , (8.40)

and  
i 1
Ciparticular (t) = t Cp , (8.41)
i i
where

Cp PI i
Stot = a positive constant. (8.42)
+ i=1 i

Equation (8.40) says the particular portion of the solution for ntot is a linearly increasing function
of time. What about the homogeneous portion? It has I exponentials that decay away in time,
so they soon become unimportant. The other exponential is exp(0), because

s1 =0 when =0.

Of course, exp(0)=1. We therefore have


" #
large t `p
ntot (t) A1 + t PI i
Stot . (8.43)
`p + i=1 i

Question: Why were we able to replace with `p ?

Answer: Because = `p /k, and we are looking at =0 k=1.

Thus, we have now found that

In a source-driven critical system,


the neutron population eventually increases linearly with time .

254
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

8.6.5 Subcritical Reactor with No Fixed Source


In a source-free reactor the particular solution is zero for all cases. If the reactor is subcritical,
then < 0 and, as discussed above, all exponential time constants {sj } are negative. If we continue
to let s1 represent the algebraically largest s (the one closest to zero, in this case), then the other
exponential terms decay more rapidly than the s1 term, and we have
large t
ntot (t) A1 e|s1 |t . (8.44)
That is,

In a source-free subcritical system,


the neutron population eventually decreases exponentially with time .

Important: At the beginning of a transient (for example, immediately after rapid insertion of
control rods) the population may drop much faster than this single exponential, because of
the other rapidly-varying exponentials that are part of the solution. But the population eventually
approaches this single decaying exponential with time constant s1 .
Recall that in a subcritical system,
s1 (1 , 0) . (8.45)
Also note that 1 min is associated with the longest-living delayed-neutron precursors, which
have half-lives of around 55 seconds, so
ln 2 1
min = . (8.46)
55 s 80 s
This means it takes the neutron population at least 80 seconds to decrease by a factor of e. This
is not a fast rate of decrease! But remember:

the population may decrease


much faster than this in the early stages of a transient!

8.6.6 Supercritical Reactor with No Fixed Source


In this case one of the exponential terms has a positive s value while all other exponential terms
are decaying in time. The increasing exponential eventually dominates the others, and
large t
ntot (t) A1 es1 t . (8.47)
That is,

In a source-free supercritical system,


the neutron population eventually increases exponentially with time .

255
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

At the beginning of a transient the population may grow more slowly or more rapidly than this, or
even decrease for a little while (!), but it eventually approaches this single increasing exponential.

8.6.7 Critical Reactor with No Fixed Source


If the reactor is critical, then =0 and, as discussed above, all exponential time constants {sj } are
negative except for one that is zero. It follows that
large t
ntot (t) A1 . (8.48)
That is,

In a source-free critical system,


the neutron population eventually reaches a steady state .

This is just what we would expectthe definition of a critical reactor is that it is able to sustain
a steady chain reaction, without help from extra sources of neutrons. However, note that the
population could change significantly while it is on its way to the steady value. That is,

even in a source-free critical reactor, the neutron population


may change for a while before it settles into steady state.

8.6.8 Summary of PRKE Solutions with Constant Coefficients

We have studied PRKE solutions for situations in which Stot (fixed source), (reactivity),
(delayed-neutron fraction), i (type-i delayed-neutron fraction), and (mean generation time) are
all constants. This is a realistic case that does arise in nature! We studied six sub-cases, which are
the combinations of subcritical, supercritical, and critical reactors with and without fixed sources.
Interestingly, we find that while the presence of delayed neutrons can cause interesting and compli-
cated behavior early in a transient, eventually the neutron population behaves quite predictably:

1. If there is no fixed source, then the neutron population:

(a) eventually grows exponentially if > 0 ;


(b) eventually falls exponentially if < 0 ;
(c) eventually stays steady if = 0 ;

2. If there is a fixed source, then the neutron population:

(a) eventually grows exponentially if > 0 ;


(b) eventually reaches a steady value if < 0 ;
(c) eventually grows linearly in time if = 0 ;

256
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

If there were no delayed neutrons, then it is easy to see (by looking at the inhour equation with
=0) that
=0 k 1
s . (8.49)
`p

In a typical commercial reactor, `p is on the order of 0.0001 s. Now consider a slightly supercritical
reactor, with k=0.001. How much would the neutron population (and thus the power) change in 1
second if there were no delayed neutrons?
 
0.001
ntot (t)|t=1 s = ntot (0) exp (1 s) = ntot (0)e10 > 22, 000ntot . (8.50)
0.0001 s

Imagine how difficult it would be to control a reactor if its power could change this quickly!

Delayed neutrons slow this down dramatically. They make it relatively easy to control reactors
that are slightly supercritical or subcritical. There will be exercises in which you will quantify this
for yourself.

Included here are a series of important sketches that summarize what we found above. These are
Figs. 8.3-8.6. Study these! Be able to explain them to a high-school student. Practice explaining
them to each other. Work all of this into your intuition about the way neutrons behave in reactors.

Figure 8.3: Behavior of neutron population in a CRITICAL reactor. The with-fixed-source figure
assumes the source is inserted at t=0. Note the jump before the smooth increase that becomes
linear in time.

257
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Figure 8.4: Behavior of neutron population in a SUBCRITICAL reactor (linear and semi-log).
The reactor is at steady state until time zero, when there is an insertion of negative reactivity
(perhaps by insertion of control rods). Note the rapid drop in population prior to the slower
exponential decay to either zero (if no source) or a new steady value (if source is present).

258
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Figure 8.5: Behavior of neutron population in a SUPERCRITICAL reactor. The reactor is at


steady state until time zero, when there is an insertion of positive reactivity (perhaps by withdrawal
of control rods). Note the rapid jump in population prior to the slower exponential increase.

Figure 8.6: Behavior of neutron population in a SUPERCRITICAL reactor. The reactor is at


steady state until time zero, when there is an insertion of positive reactivity (perhaps by withdrawal
of control rods). This zoom-in view of early time more clearly shows the rapid jump in population
prior to the slower exponential increase.

259
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Figure 8.7: Behavior of neutron population in a CRITICAL reactor with source is inserted at
t=0. This zoom-in view of early time more clearly shows the rapid jump before the smooth increase
that becomes linear in time).

8.7 Feedback

A few words are in order about an important phenomenon that we have not yet described:

feedback

In a nuclear reactor, feedback happens when changes in the neutron population cause changes in
the reactor properties. These changes in properties then alter the way that the neutron population
behaves. Most of the feedback in a nuclear reactor is caused by

temperature changes

in the reactor material. Remember that temperature can be strongly influenced by the

heat-generation rate

which of course is determined by the

fission rate

which of course is determined by the

neutron population

Also recall that

cross sections

260
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

depend upon the

material temperature

and that

reactivity (or multiplication factor)

is a function of the cross sections. That is, we need to know the temperature to get the cross sections
that allow us to calculate the neutron population, but we need to know the neutron population to
calculate the fission source that drives the temperature.

Illustration:

1. Consider a reactor that is critical at a very low power level, such that fission is not causing
the fuel to be noticeably hotter than its surroundings.

2. Now suppose that control rods are pulled out so that the reactivity becomes positive (because
the absorption cross section has been reduced). As we have seen, the neutron population will
soon be on an exponential increase. Thus, the fission rate will increase exponentially with
time.

3. Eventually this will cause the fuel to heat up significantly. This will change the cross sections
in the fuel (rememberthe cross sections are averaged over nucleus motion, and temperature
determines the motion). In fact, in almost all reactors this will cause the absorption cross
section to increase more than the fission cross section, which causes the multiplication factor
to decrease.

4. As temperature increases further, the multiplication factor decreases further. This continues
until k reaches 1 and the reactor settles into steady state.

The above is an example of

negative feedback .

This is what we want. Negative feedback gives us

stability.

which means that fluctuations in neutron population are damped out in time. Positive feedback,
on the other hand, would amplify changes. Imagine a reactor in which increased temperature
caused an increase in k (and thus ). In such a reactor, a small increase in neutron population
would produce an increase in reactivity, which would cause the population to increase, which would
further increase reactivity, etc. The reactor power would spiral upward until something (such as
melting fuel) introduced negative feedback. Not good.

So we design our reactors with negative feedback. (Remark: The Chernobyl accident began largely
because the operators placed the reactor into a state in which there was positive feedback.)

261
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

8.8 Kinetics with Realistic Transport Model

Previously we examined the time-dependent neutronics problem assuming a simplistic model: one-
group diffusion theory for a homogeneous reactor. Here we show that even with the most realistic
model we ever considerenergy-dependent transport theory for a heterogenous reactorwe can
cast the time-dependent equations in exactly the same PRKE form as before.
If we integrate the time-dependent neutron transport equation over all directions we obtain the
0th-moment equation, expressing conservation in every element of the x-y-z-E phase-space:

1 (~r, E, t) ~ ~
+ J(~r, E, t) + t (~r, E, t)(~r, E, t)
v(E) t

I
X
= Sfixed,0 (~r, E, t) + i (E)i Ci (~r, t)
i=1


+ p (E) dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t)
0


+ dEi s (~r, Ei E, t)(~r, Ei , t) . (8.51)
0

(We continue to define I number of d.n.p. types. We have introduced Ci to represent precursor
concentration, reserving Ci for the total number of type-i precursors in the reactor.) There are
no approximations here beyond those in the neutron transport equation. Now we integrate this
equation over all energies and over the entire spatial volume of the reactor:

d 3 2 ~ r, E, t)
d r dE n(~r, E, t) + d r dE en (~r) J(~
dt V ol 0 Surf 0


3
+ d r dE t (~r, E, t)(~r, E, t)
V ol 0

I
X
= d3 r dE Sfixed,0 (~r, E, t) + i d3 r Ci (~r, t)
V ol 0 i=1 V ol


3
+ d r dEi p (~r, Ei , t)f (~r, Ei , t)(~r, Ei , t)
V ol 0


3
+ d r dE dEi s (~r, Ei E, t)(~r, Ei , t) . (8.52)
V ol 0 0

262
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

The energy integrals of the functions equal unity, so these functions have disappeared. The scat-
tering integral on the right-hand side becomes simply the total scattering rate in the reactor, which
can be subtracted from the total collision rate on the left-hand side, leaving the total absorption
rate. Now, without approximation, we define several terms:


3
ntot (t) d r dE n(~r, E, t) (8.53)
V ol 0


Ci (t) d3 r Ci (~r, t) (8.54)
V ol


Stot (t) d3 r dE Sfixed,0 (~r, E, t) (8.55)
V ol 0


3
tot (t) d r dE (~r, E, t) (8.56)
V ol 0


D1E ntot (t) d3 r 0 dE n(~r, E, t)
= V ol
3
(8.57)
v tot (t) V ol d r 0 dE v(E)n(~ r, E, t)

1
tot (t) = ntot D E (8.58)
1
v


d3 r 0 dE a (~r, E, t)(~r, E, t)
V ol
ha i (8.59)
3
V ol d r 0 dE (~r, E, t)


d3 r 0 dE p (~ r, E, t)f (~r, E, t)(~r, E, t)
hp f i V ol (8.60)
3
V ol d r 0 dE (~ r, E, t)


d3 r 0 dE di (~r, E, t)f (~r, E, t)(~r, E, t)
hdi f i V ol (8.61)
3
V ol d r 0 dE (~ r, E, t)

I
X
hf i hp f i + hdi f i (8.62)
i=1


d2 r ~
Surf 0 dE en (~
r) J(~r, E, t)
hesc i (8.63)
3
V ol d r 0 dE (~ r, E, t)

263
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence


hdi f i d3 r 0 dE di (~r, E, t)f (~r, E, t)(~r, E, t)
i = V ol
3
(8.64)
hf i V ol d r 0 dE (~ r, E, t)f (~r, E, t)(~r, E, t)

I PI
i=1 hdi f i hp f i
X
i = =1 (8.65)
hf i hf i
i=1

hp f i = (1 )hf i (8.66)

D1E 1
`p (8.67)
v ha i + hesc i

hf i
k (8.68)
ha i + hesc i
We continue to define reactivity and mean generation time exactly as before:
k1
(8.69)
k

`p
(8.70)
k
With these definitions (not approximations! ) we can rewrite our integrated transport equation as:
I
d X
ntot (t) = [hp f i ha i hesc i] tot (t) + i Ci (t) + Stot (t) (8.71)
dt
i=1

Now we manipulate and recognize more of our defined terms:


I
d 1 X
ntot (t) = [(1 )hf i (ha i + hesc i)] D E ntot (t) + i Ci (t) + Stot (t)
dt 1
v i=1

f h i I
(1 ) ha i+h esc i
1 X
= D E ntot (t) + i Ci (t) + Stot (t)
1 1
v ha i+hesc i i=1

I
(1 )k 1 X
= ntot (t) + i Ci (t) + Stot (t)
`p
i=1
k1 I
k X
= ntot (t) + i Ci (t) + Stot (t) (8.72)
`p /k
i=1

264
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

We perform similar operations and manipulations on the precursor equations. When the dust
settles we have:

REACTOR KINETICS EQUATIONS

I
d (t) (t) X
ntot (t) = ntot (t) + i Ci (t) + Stot (t) (8.73)
dt (t)
i=1
d i (t)
Ci (t) = ntot (t) i Ci (t) (8.74)
dt (t)

Obviously, these equations have

exactly the same form as those we derived from 1GD theory for homogeneous reactors!

The difference is only in the definitions of the coefficients, namely , {i } (and thus ), and .
Several observations are in order:

1. The coefficients , {i }, and are defined in terms of the exact transport quantities (~r, E, t)
~ r, E, t), which of course depend on time.
and J(~

2. In more detail, the coefficients depend on the spatial, energy, and directional shape of the
angular flux, which can evolve in time. Thus, in general, the coefficients depend on time.

3. The coefficients do not depend on the amplitude of the angular fluxthe amplitude cancels
from numerator and denominator in the relevant definitions.

4. If we assume that the flux shape does not change with time, but that only its amplitude
changes, then our kinetics equations become the point reactor kinetics equations (PRKEs).
In this case , {i }, and are all constants unless something changes the reactor geometry
or macroscopic cross sections. We can learn a lot from the solution of our coupled first-order
ODEs in this interesting constant-coefficients case.

5. We made no significant approximations here, and our setting is the real world. That is, the
RKEs are an essentially exact description of how the neutron population varies in time in
real reactors! The only question is how to calculate the coefficients in the equations.

8.9 Solutions of Reactor Kinetics Equations

In this section we discuss solutions of the RKEs. We develop tools for solving them and develop a
quantitative understanding of them.

265
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

8.9.1 Single delayed-neutron precursor type (I = 1)

If:

there is only one delayed-neutron precursor type (I=1);

the kinetics parameters are all constant in time; and

the source is constant in time;

then the PRKEs become:


d
n(t) = n(t) + C(t) + S (8.75)
dt

d
C(t) = n(t) C(t) (8.76)
dt
Here we have dropped the subscript tot on the neutron population and source rate to simplify
notation. It is understood that n is the total neutron population in the reactor and S is the total
source rate (neutrons per unit time).

We established previously that the solutions n and C are sums of particular and no-source
solutions, and we found the particular solutions. For the case of a single dnp type:
`p
S = S 1k ,
6= 0,
part
n (t) =   (8.77)
`p
`p +/ S t, = 0,


S ,
6= 0,
part
C (t) =   (8.78)

t 1 , = 0.

S


+`p

We know that the no-source part of the solutions are sums of two exponentials; thus, the complete
solutions are of the form:
2
X
n(t) = Aj esj t + npart (t) , (8.79)
j=1
2
X
C(t) = Fj esj t + C part (t) . (8.80)
j=1

266
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Here A1 , A2 , F1 , F2 , s1 , and s2 are constants that must be determined. When we insert these
expressions into the PRKEs, we find that the particular portions and the fixed source cancel out
of the equations, leaving the no-source solutions in the no-source equations:
2 2 2
d X X X
Aj esj t = Aj esj t + Fj esj t (8.81)
dt
j=1 j=1 j=1

2 2 2
d X X X
Fj esj t = Aj esj t Fj esj t (8.82)
dt
j=1 j=1 j=1

Important observation:

If s1 6= s2 , then es1 t and es2 t are linearly independent.

It follows that the no-source equations must hold independently for j = 1 and j = 2:
d
Aj e s j t = Aj esj t + Fj esj t , j = 1, 2 , (8.83)
dt

d
Fj esj t = Aj esj t Fj esj t , j = 1, 2 . (8.84)
dt
If we carry out the derivatives, the exponentials cancel out of each equation, and we have
 

sj Aj + [] Fj = 0 , j = 1, 2 , (8.85)

 

Aj + [sj + ] Fj = 0 , j = 1, 2 . (8.86)

For either value of j we can write this in matrix form. Let us do so and drop the j index for now:

s

A 0
= (8.87)
F 0
s+

We seek solutions with nonzero A and F coefficients, which means that the matrix must not be
invertible. This means its determinant must be zero:
 

s (s + ) = 0 (8.88)

This is a quadratic equation for s. It has two solutions, which we call s1 and s2 . It is equivalent to

the inhour equation for the case of 1 d.n.p. type

267
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

The two solutions are (if our algebra is correct):


 
1
q
2
s= ( ) + 4 (8.89)
2
" s #
4
= 1 1+ (8.90)
2 ( )2

Now that we have the s values, how do we find the A and F coefficients? There are two each, or
four total coefficients to find, so we need four equations.

Two of the equations are the initial conditions. If t = 0 is the initial time, then the initial conditions
are:

n(0) = A1 + A2 + npart (0) = n0 = given , (8.91)


part
C(0) = F1 + F2 + C (0) = C0 = given . (8.92)

We also have the four equations (8.85)-(8.86). Does this give us too many equationssix equations
for four unknowns? Not really:

Each Eq. (8.85) is just a multiple of the corresponding Eq. (8.86).

This is because the s values make Eqs. (8.85) and Eq. (8.86) linearly dependent.
So for each value of j, only one of those equations is independent. We need to pick two independent
ones (one for each j). An easy way to do this is to pick the precursor concentrations equation and
use it with each of the two s values (sj for j = 1 and 2). That is, our two linearly independent
equations can be:
 

A1 + [s1 + ] F1 = 0 , (8.93)

 

A2 + [s2 + ] F2 = 0 . (8.94)

With this choice, then Eqs. (8.91), (8.92), (8.93), and (8.94) are the four equations that determine
the four coefficients A1 , A2 , F1 , and F2 .

8.9.2 I delayed neutron precursor types

The solution procedure with I delayed-neutron precursor types is the same as for the special case
of 2 types. The difference is that there are more equations and more algebra. In the general case
of I types, we still have the particular solution for n, Eq. (8.77), and a particular solution for each
precursor types concentration, which for the i-th type is Eq. (8.78) with i subscripts on and .
We still have Eqs. (8.85) and we have one Eq. (8.86) for each precursor type, with both and
now taking on i subscripts and with Fi,j replacing Fj . So instead of 2 equations we now have I + 1
for each value of j, and now we expect these to hold for I + 1 different values of j.

268
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

When we put these equations into matrix form and require that the determinant equal zero, we
get an (I+1)-order polynomial equation for s, which is the generalization of the quadratic equation
(8.88). We can algebraically manipulate this polynomial equation into the following form:
I
`p s 1 X i s
= + , (8.95)
1 + `p s 1 + `p s s + i
i=1

which we call the inhour equation. This Eq. determines the time constants {sj , j = 1, . . . , I + 1}.

Finally, we must determine the (I + 1)2 coefficients Aj and Fi,j for j = 1, . . . , I + 1, which means we
need (I + 1)2 independent equations in which everything except the coefficients is known. At first
it may seem that there are too many equations: There are I + 1 initial conditions and also (I + 1)2
no-source equations (I + 1 no-source equations for each of the I + 1 values sj , j = 1, . . . , I + 1) that
must be satisfied. However, we recall that each of the sj values causes the no-source equations to
be linearly dependent, which means for each sj value we can remove one of the no-source equations
from our list (because it is just a combination of the other I no-source equations). When the dust
settles, we can impose the following (I + 1)2 equations to determine the coefficients in the analytic
solution:

The I + 1 initial conditions: Eq. (8.91) and the I Eqs. (8.92)

The I Eqs. (8.86), for each of the I + 1 values of sj

8.9.3 Reactor Kinetics Terminology: Period, Dollars, Cents, -effective

The stable reactor period is defined to be the time it takes for the reactor power to change by a
factor of e, after initial transients have died away and the neutron population (or reactor power) is
changing exponentially with time. It should be obvious that:
1
T stable reactor period = (8.96)
|s1 |

A long period means slowly changing power; a short period means rapidly changing power.

It is usually convenient to express the reactivity, , in terms of the delayed-neutron fraction, :



in dollars (8.97)

So, for example, if = /2 we say that the reactivity is

negative 50 cents, or -$0.5

269
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

The fraction of neutrons that emerge from fission with a time delay (through production and decay
of precursors) depends on the kind of nuclide that fissioned and on the energy of the neutron
that induced the fission. See Table 5.1 from Lewiss book, which provides data for different fissile
nuclides and for fissions caused by thermal neutrons.

Our Reactor Kinetics Equations (RKEs) contain delayed-neutron fractions that are averaged over
the different kinds of nuclides and averaged over the energies of the incident neutrons. For example,
when we derived the Reactor Kinetics Equations by integrating and manipulating the energy- P
dependent transport equation, Eqs. (8.64) and (8.65) naturally came about for i and i i .
These are flux-weighted averages over both the nuclides in the reactor and the energies of the
incident neutrons.
More sophisticated derivations of the Reactor Kinetics Equations are often employed, and these
derivations give rise to slightly different definitions of the kinetics parameters (t), i (t), (t), and
(t). These more sophisticated definitions also include flux-weighted averages, but in addition they
take into account the relative importance of each neutron at its birth, which depends on the
position and energy of the birth event. (The importance function is the adjoint flux. This is a
more advanced topic that you will encounter later in your studies. We will address this to some
extent in Chapter 10 of these notes.)

The properly defined i (t) and (t), complete with averaging and weighting according to impor-
tance, are called

effective delayed-neutron fractions

They are often denoted by the subscript eff :

i,eff effective delayed-neutron fraction for precursors of type i (8.98)


eff effective delayed-neutron fraction (8.99)

This distinguishes them from the physical delayed-neutron fractions, which depend only on the
nuclide (strong dependence) and the energy of the incident neutron (weak dependence).

We adopt the following convention: The i and in our RKEs are understood to be the proper
effective delayed-neutron fractions, even though we shall not employ the subscript eff.

8.9.4 PRKE Solution: Large Positive Reactivity

Consider the reactor kinetics equations (RKEs) and assume that the neutron flux shape in posi-
tion, energy, and direction are constant, with only the amplitude changing. Suppose the reactivity,
, is large:

assume > (8.100)

270
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

We assume that is so large that s1  i for all i. In this case we can make simplifying approxi-
mations in the inhour equation for s1 :
I I
`p s1 1 X i s1 s1 i `p s1 1 X
= + + i
1 + `p s1 1 + `p s1 s1 + i 1 + `p s1 1 + `p s1
i=1 i=1
s1 `p +
= (8.101)
1 + s1 `p

Solve for s1 :
1
t
s1 = n(t) A1 e `p (8.102)
1 `p `p

To develop an intuition about what this means, recall that in most reactors the prompt neutron
lifetime is a small fraction of a second: typically less than 104 s in a thermal-spectrum reactor
and 107 in a fast-spectrum reactor. As you can see in Staceys Table 5.1, is in the range
0.002 0.016. Consider the following values for example:

Suppose: `p = 104 s
= 0.007
= $1.5 = 1.5 = 0.0105

1.5 0.0035
Then: s1 = 4 = 35 s1 . (8.103)
`p 10 s

In this case, the neutron population would change by the rather dramatic factor e35 1015 in
one second! Of course, before this could really happen, feedback would kick in and reduce , thus
preventing such a large excursion.

If a reactor has > , which leads to extremely rapid changes in neutron population, it is called

prompt supercritical

This is because it is supercritical on prompt neutrons aloneit does not need delayed neutrons
in order to sustain a growing chain reaction. We normally try to avoid this situation in reactors,
although in some research and test reactors we occasionally explore this situation intentionally. An
example is pulsing a TRIGA reactor.

8.9.5 PRKE Solution: Small Positive Reactivity

Recall from our graphical solution of the inhour equation that if reactivity is small, then s1 will
be small. Let us quantify this. Suppose that is small enough that the following is true:

s1  i , all i

271
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

In this case we can make simplifying approximations in the inhour equation for s1 :
I I
`p s1 1 X i s1 s1 i X i
= + `p s1 + s1
1 + `p s1 1 + `p s1 s1 + i i
i=1 i=1
I
" #
X i
= s 1 `p +
i
i=1


I  
X 1
= s1 (1 ) `p + i `p +

i

| {z } |{z} |{z}
i=1 delayed frac |
prompt frac prompt life {z }
delayed life

= s1 h`i (8.104)

where we have defined the average neutron lifetime


I  
X 1
h`i|| (1 )`p + i `p + (8.105)
i
i=1

It follows that

t
s1 n(t) A1 e h`i (8.106)
h`i

This is a much slower rate of change than in the large-reactivity change, because the average
neutron lifetime, h`i, is much longer (orders of magnitude) than the prompt-neutron lifetime, `p .
(See the definition and recall that one value of 1/i equals 80 seconds.)

272
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

8.9.6 PRKE Solution: Large Negative Reactivity

From our graphical investigation we know that as reactivity becomes large and negative, s1 min .
It follows that all terms except one on the right-hand side of the inhour equation are insignificant,
and
I
`p s 1 1 X i s1  1 min s1
= +
1 + `p s1 1 + `p s1 s1 + i 1 + `p s1 s1 + min
i=1
min s1
(8.107)
s1 + min
It follows that
 min min || 1 |/|
s1 = (8.108)
min min + || 80 s (min /) + |/|

We already knew that s1 approaches min for large negative reactivity. This equation quantifies
how it approaches this limiting value as a function of reactivity.

8.9.7 PRKE Solution: Prompt-Jump Approximation

Suppose a reactor is operating in steady state (SS) with neutron population nss . The RKE for the
precursor concentration tells us that in steady state the following must be true:
i
0= nss i Ci,ss (8.109)

which implies that
i
Ci,ss = nss (8.110)
i
I
X
and thus i Ci,ss = nss (8.111)

i=1

Now suppose that at time t = 0 the reactivity instantly changes from ss to new , and suppose ,
, and Stot are not changing with time. Then the RKE for the neutron population is:
I
d new X
n(t) = n(t) + i Ci (t) + Stot (8.112)
dt
i=1

273
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

In the earliest stages of the transient, the delayed-neutron precursor concentrations have not had
time to change appreciablythey approximately retain their pre-transient steady-state values. Let
us solve for the neutron population under this approximation. If we use an integrating factor and
rearrange we find:
t " I #
new X new 0
t 0
n(t) = nss e + dt i Ci,ss + Stot e (tt )
0 i=1
I
" #
new
t
X  new

= nss e i Ci,ss + Stot 1e t (8.113)
new
i=1

Now consider the usual case of new < . In this case (new )/ is negative, and the expo-
nentials decay to zero. In many cases of interest, this happens very quickly, before the precursor
concentrations have time to change by very much.

Consider the following numerical example. Suppose new = 0.005 and = 104 seconds.
new
Then after only 0.1 seconds, e t e5 < 0.01, and after only 0.2 seconds it is less than 104 .
The precursor concentrations cannot change appreciably over such a short time. Note that if the
prompt neutron lifetime is shorter, then the exponential decay happens faster.
The important point is that there can be a time range in which the exponentials in Eq. (8.113) be-
come negligibly small but the precursor concentrations are still at approximately their pre-transient
levels. During this time window we have
" I #
X
n(t) = nss [0] i Ci,ss + Stot (1 0)
new
i=1

Thus,
 

Eq. (8.111)
n(t) nss + Stot , t in specified time window (8.114)
new

An example of how the neutron population quickly transitions to this new quasi-steady level is
shown in Fig. 8.6. In this example the population increases by more than 15% in the first 0.03
s, then increases at a much slower rate thereafter (as the precursor concentrations increase). For
an example with negative reactivity insertion see Fig. 8.7. In this example the neutron population
quickly decreases to approximately 1% of its initial steady-state value before transitioning to a
slowly decaying exponential.
In the limit of infinitesimally short prompt-neutron lifetime, the exponentials immediately go to
zero and the neutron population jumps instantly to the new level given by Eq. (8.114). If we
pretend that this jump happens instantly, we have made

THE PROMPT-JUMP APPROXIMATION

274
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Even when we dont pretend it happens instantly we recognize that rapid changes in reactivity
induce initial rapid changes in neutron population, followed by a transition to a slower rate of
change. The exception to this is when > , in which case the exponentials in Eq. (8.113 grow
instead of decaying, and the reactor (which is prompt supercritical in this case) does not transition
to a slower rate of change.
Consider the following prompt-jump results, which follow from the above development.

1. Critical steady state followed by reactivity insertion new < . There are no sources in this
case. We have:
jump 1
nnew nss = nss = nss , (8.115)
new new 1 $new
where $new new / = transient-inducing reactivity in dollars. That is, if we begin at
critical steady state and insert reactivity, the population jumps by a factor of 1/(1 $).
Important: This holds for all negative reactivity insertions and for positive insertions that
are sufficiently lower than (sufficiently far from prompt-supercritical).

2. Critical steady state followed by insertion of a fixed source, with reactivity remaining at 0.
In this case the neutron population jumps quickly to:
 
jump
nnew nss + Stot = nss + Stot , (8.116)

3. Subcritical steady state (which means constant source and negative pre-transient reactivity)
followed by reactivity insertion new < . We now have
 
jump
nnew nss + Stot (8.117)
new
In this case the SS RKEs tell us the relation between the source and the steady-state pre-
transient neutron population:
ss
Stot = nss

and we have
 
jump ss ss 1 $ss
nnew nss = nss = nss (8.118)
new new 1 $new

4. Subcritical steady state (which means constant source and negative pre-transient reactivity)
followed by removal of source but no change in reactivity. We now have
 
jump 1
nnew nss + 0 = nss = nss , (8.119)
ss ss 1 $ss

275
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

This is much like the result we obtained in the source-free case with reactivity insertion. Note
that since ss < 0, the denominator is > 1 and the population jumps to a lower value. Note
further that if the jump can be measured, then the pre-transient reactivity can be inferred.
This technique for measuring reactivity is called the source-jerk method. However, it can
be difficult to measure the jump (drop), because the population does not stay for very long
at all at the value to which it promptly dropsit continues to decrease from there.

Examples

1. A reactor is critical and operating in steady state at 10 kW. Control rods are quickly dropped
into the reactor, inserting negative reactivity of magnitude 9 dollars. What is the power
following the prompt drop?

Answer: Approximately (10 kW) 1/(1 (9)) = 1kW.

2. A reactor is critical and operating in steady state at 10 kW. Control rods are quickly pulled
from the reactor, inserting positive reactivity of magnitude 10 cents. What is the power
following the prompt jump?

Answer: Approximately (10 kW) 1/(1 (0.1)) 11.1kW.

3. A reactor is critical and operating in steady state at 10 kW. Control rods are quickly pulled
from the reactor, inserting positive reactivity of magnitude 40 cents. What is the power
following the prompt jump?

Answer: Approximately (10 kW) 1/(1 (0.4)) 16.7kW, but this may not be terribly
accurate.

4. A reactor is critical and operating in steady state at 10 kW. Control rods are quickly pulled
from the reactor, inserting positive reactivity of magnitude 1 dollar. What is the power
following the prompt jump?

Answer: The prompt-jump approximation does not hold here. There is no quasi-steady
state; rather, the population continues its rapid rise.

276
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

8.9.8 PRKE Solutions: Neutrons or Power?

So far we have studied RKEs that describe neutron population and delayed-neutron-precursor
concentration. It is often convenient to replace the neutron-population equation with an equation
for reactor power. Note that reactor power is in many cases well approximated by:

n(t)
P (t) = hf i
|{z} |{z} h1/vi
[ recoverable
s
energy
] [ recoverable
fission
energy |
] {z }
[ fissions
s ]
Note This is not always rigorously true, for the following reasons:

1. Some recoverable energy comes from the decay of fission products. Unless the reactor is
operating in steady state and has been for a long time, the energy deposition rate from this
decay at a given time and location is not necessarily proportional to the fission rate at that
time and location.
Note in particular that if we shut down a reactor with no fixed sources, we can drive the
neutron population to zero, but we cannot drive the decay heat to zero, and thus we cannot
drive the power to zero.

2. Exothermic reactions (such as radiative capture) other than fission contribute energy that is
deposited in the reactor. However, the rate at which these reactions occurs is usually roughly
proportional to the rate at which fission occurs, so this can be treated by adjusting .
It follows that if we multiply the RKEs by hf i/h1/vi, we obtain exactly the same equations with
the following replacements:

replaced by hf i
n(t) n(t) P (t) (8.120)
h1/vi
replaced by hf i
[Ci (t)]actual [Ci (t)]actual [Ci (t)]new definition (8.121)
h1/vi
replaced by hf i
[S(t)]tot,actual [S(t)]tot,actual [S(t)]new definition (8.122)
h1/vi
That is, in the RKEs we can simply replace the neutron population, n(t), with the reactor power,
P (t), provided that:

Power is dominated by fissions and other exothermic neutron-nucleus reactions (not by decay
heat), and

We interpret Ci and S as being hf i/h1/vi their actual physical values.

8.10 Impact of Feedback on Kinetics Solutions

The RKEs provide an accurate description of reactor behavior under transient conditions, as long
as the coefficients , i , and are calculated using the correct flux-weighted average quantities

277
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

contained in their definitions. This is true even if feedback is changing the underlying material
properties, as long as the correct properties at any given time are used in the definitions of the
coefficients.
The PRKEs assume that the flux shape (distribution in position, energy, and direction) contained
in the definition of the coefficients does not change during the time period of interest. The flux
amplitude is, of course, allowed to change. The PRKEs can be applied to problems with feedback
as long as the correct material properties at any given time are used in the definitions of the
coefficients, and they remain accurate in this case as long as the flux shape does not change much
during the time period of interest.

The PRKEs express a relationship between power (P (t)) and reactivity ((t)). We have primarily
looked upon them as a way to calculate P (t) given some known (t). In reality, however, (t) is
not completely given if the power is changing the material properties, but instead is influenced
by the power itself. From our previous definitions it is easy to see that

k1 hf i ha i hesc i
(t) = (8.123)
k hf i

Because the macroscopic cross sections in this expression depend on

number densities and thermal motion of atoms,

which depend on temperature, we see that depends on the temperature throughout the reactor.
We also know that in many cases of interest, temperature will depend on reactor power. Thus, in
general,

Changes in reactor power induce changes in

It is useful to split the reactivity into two additive components:

(t) = ext (t) + f (t) (8.124)

where

ext (t) externally imposed reactivity (8.125)


f (t) feedback reactivity ( induced by changes in P ) (8.126)

The relationships among power and these reactivity components are illustrated in a circuit type
diagram in Fig. 8.8.

Another representation of the same process hides the feedback loop inside a bigger box, as in
Fig. 8.9.

278
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Figure 8.8: Sketch of feedback reactivity circuit. PRKEs take (t) as input and give P (t) as
output. The feedback process takes P (t) as input (along with other variables, perhaps) and gives
f as output. Power history determines delayed-neutron precursor concentrations as well as decay
heat. Top: ext in, + in circle, = ext + f , PRKEs in box, P (t) outgoing and down, power
history in oval, f [P (t), . . .] in box, f coming up

Figure 8.9: Different sketch of feedback reactivity circuit. This simply draws a box around the
previous sketchs two boxes and sum operator. Top: ext in, + in circle, = ext + f , PRKEs in
box, P (t) outgoing and down, power history in oval, f [P (t), . . .] in box, f coming up

8.10.1 Quantifying Feedback: Reactivity Coefficients

We have seen that reactivity depends on material composition, geometry, temperature, neutron
distribution (flux shape), and neutron-population amplitude. We have seen that reactivity can be
separated into two components: reactivity imposed on the reactor from the outside, and reactivity
induced within the reactor by changes in the neutron population. In this section we develop
a framework for quantifying the changes in reactivity that are induced by changes in neutron
population (which can induce changes in composition, geometry, and temperature). To simplify
our task initially we will make some mild assumptions:

1. We assume that the reactor remains intact. No melting or explosions.

2. We assume that the balance of plant stays about the same


during the time period we study

279
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

The first assumption means no dramatic changes in reactor geometry. The second means no major
changes in inlet temperature and pressure of the coolant.

Under these assumptions, power changes will cause short-term reactivity changes only by changing
the temperatures of reactor components. (We take the view that density changes are the result of
temperature changes.) That is, except for the reactivity due to fission products,

f (t) = f [T (~r, t)] , (8.127)

T (~r, t) = functional of P (t) and power history (8.128)

It is useful to define various coefficients of reactivity, or reactivity coefficients. These coefficients


quantify the amount of feedback reactivity that is introduced per unit change in another quantity,
such as temperature of a component. For example

Tmod moderator temperature coefficient (8.129)
Tmod


Tfuel fuel temperature coefficient (8.130)
Tfuel

If temperature at time t is determined completely by power at time t, then it is possible to define


a power coefficient of reactivity that includes all of the temperature-induced reactivity changes:

P power coefficient (8.131)
P
Note that the power coefficient of reactivity is meaningful only if there is a one-to-one relation
between power and the temperature of any given component. This can be the case if the two
assumptions given above are met, but it is not always the case. For example, following a reactor
scram (rapid shutdown), temperature will be determined more by decay heat (which is a function
of power history) than by power at the current time.

Reactivity coefficients quantify the feedback reactivity introduced from changes in temperatures or
power. For example, the feedback reactivity introduced by changing fuel temperature from T1 to
T2 is:
T2 T2
fuel
f = dTf uel T = dTf uel = []Tfuel =T2 []Tfuel =T1 (8.132)
T1 T1 Tfuel

Suppose temperature at time t is determined by power at time t, and suppose power coefficient of
reactivity is approximately constant. Then how much feedback reactivity is introduced by changing
power from P1 to P2 ? Answer:
P2
constant
P
f = dP P P (P2 P1 ) (8.133)
P1

280
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Earlier we discussed dynamic phenomena that take place on vastly different time scales in reactors.
The following are among the more important phenomena that induce feedback reactivity on time
scales ranging from sub-second to hours.

Doppler broadening of resonances

Changes in moderator/coolant temperature and density

Changes in fission-product inventory

Of course, nuclide production and depletion also has a dramatic effect on reactivity, but this
takes place on longer time scales. We will address depletion and production later. There are
also other phenomena in some reactors that can cause significant feedback on short time scales,
such as geometric changes in fuel or structural components due to temperature changes. These
vary significantly from one reactor to another and must be carefully considered reactor by reactor.
Expansion of fuel as it heats up, for example, is an important negative-feedback mechanism in some
reactors.

In what follows we briefly discuss the phenomena listed above and how their impacts can be
approximated via reactivity coefficients.

8.10.1.1 Doppler Broadening

We have stressed repeatedly that the expressions for reaction rate densities that appear in our
transport and diffusion equations are correct only if the microscopic cross sections are properly
averaged over nucleus velocities. We have seen that the effect of this averaging is to smooth or
smear sharp features in the cross sections. This smoothing has the most dramatic effect on the
sharpest features, which are the resonances. Figure 4.7 illustrates this effect.

The important question is: how does this affect reactivity? To arrive at the answer we must first
understand the effect of Doppler broadening on reaction rates, and in particular on fission and
capture rates. Consider the expression for the absorption rate density due to a single resonance:

A.R.D. = dE a (E, T )(E, T ) (8.134)
Eres

It turns out that as the resonance broadens, the integral of a stays about the same. However,
as we shall see in our studies of neutron distributions in energy, the scalar flux is depressed at
energies for which is high. In a broadened resonance the scalar flux is not as depressed as in the
un-broadened case, and thus as temperature goes up the average scalar flux increases at resonance
energies. We conclude that

As temperature increases, resonance reaction rates increase.

281
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

This is an important result, but it does not answer our question about reactivity: does Doppler
broadening cause reactivity to increase, decrease, or stay the same? The answer depends on the
increase in the fission rate relative to the increase in the non-fission component of the absorption
rate (which is almost entirely radiative capture in the energy range where resonances dominate).
Both rates increase, but which one increases more?
Consider first an easy case: a fissionable (not fissile) material such as 238 U. In this case, absorption
at resonance energies is almost always a capture event, because the neutron energy is not high
enough to excite the compound nucleus beyond the fission threshold. It is therefore clear that:

increased temperature in fissionable (not fissile) nuclides induces negative feedback

What about fissile nuclides such as 235 U? Here both the resonance capture and resonance fission
rates will increase, and the effect on reactivity is not obvious. Does the neutron have more chance
of causing fission if it is absorbed in a resonance, or if it escapes the resonance and reaches thermal
energies? This depends on leakage probabilities as well as how the capture-to-fission ratio changes
with neutron energy. We can safely say the following, though, without knowing those details:

Doppler broadening in fissile nuclides has a relatively weak feedback effect

Here relatively means relative to fissionable (not fissile) nuclides.

We can make a strong fairly general statement:

Doppler broadening induces strong negative feedback


in thermal reactors fueled by low-enriched uranium.

In commercial LWRs the reactivity change caused by increased fuel temperature is substantial.
The fuel-temperature change associated with going from 1% to 100% power can cause reactivity to
change by -0.01 or more, which has magnitude > 1$.

A final remark about the feedback from Doppler broadening: As we noted early in this chapter,
this feedback is essentially instantaneous when power increases, because the energy deposition from
fission fragments to the fuel material (and thus the temperature increase caused by an increase in
power) takes place on sub-microsecond time scales. (The Doppler feedback from a decrease in
power will take longer, because it depends on transferring heat out of the fuel.)

8.10.1.2 Moderator Density

Changes in moderator density produce increases and/or decreases in reactivity. Consider the case
in which a change in moderator density causes a change in the moderator inventory in the core
and thus a change in the ratio of moderator atoms to fuel atoms. (This is usually the case if the
moderator is a fluid.) The competing effects lead to the basic picture crudely sketched in Fig. 8.10.

282
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Figure 8.10: Notional sketch of k as a function of moderator/fuel ratio, dividing the range into
under-moderated and over-moderated regions. Vertical axis = k. Horizontal = M/F. In-
tuitively you know M/F too low means low k; same for M/F too high. Must be a maximum
somewhere. Draw essentially a down-pointing parabola.

We see that there are basically two regimes:

1. Under-moderated: Mod. density reactivity

2. Over-moderated: Mod. density reactivity

Since increases in reactor power cause the moderator density to decrease, we see that

Under-moderated negative feedback [good]

Over-moderated positive feedback [not good]

Thus, it is a good idea to design reactors to be under-moderated, at least if the moderator is a


liquid or gas.

8.10.1.3 Other Temperature Effects

In most reactors the primary effect on reactivity of a change in moderator temperature is the effect
caused by the temperature-induced change in moderator density. This is discussed above. However,
there can also be other effects, mostly caused by the change in the neutron energy distribution in
the thermal (low) energy range.

283
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Recall that in the thermal range, the fission rate density and the absorption rate density:

are independent of the neutron and nucleus velocity distributions


IF
the absorption and fission cross-sections are proportional to 1/v.

It follows that if every nuclide had 1/v cross sections, changing temperatures would not cause
significant changes in thermal absorption and fission rates. Of course, many of the heavier nuclides
do not have 1/v cross sections in the thermal energy range. The most important deviations are
caused by

low-lying resonances

Given nuclides with low-lying resonances, the changes in neutron (and nucleus) velocity distribu-
tions that are induced by temperature changes can cause noticeable changes in reactivity. In fact,
in the TRIGA reactor this is a very important negative-feedback mechanism:

increased temperature shifts the thermal-neutron spectrum to higher energies, causing more
thermal neutrons to be absorbed by a nuclide such as erbium, which has a low-lying capture
resonance.

Another effect of increasing temperature is expansion of fuel. You may recall from homework
exercises that compressing a reactor causes its multiplication factor to increase, because the density
increases by a larger factor than the distance from any point to the surface, making leakage less
likely. Expansion has the opposite effect: it makes leakage more likely and can therefore cause
reactivity to decrease. This is an important negative feedback mechanism in some reactors
mainly fast-spectrum reactors and most noticeably those with metal (not oxide) fuel, which has a
higher thermal expansion coefficient.

8.10.1.4 Fission-Product Inventories

Many fission products or their daughters are neutron parasites and therefore cause a decrease in
reactivity as they build up (or an increase if they decay or get depleted). One nuclide in particular,
135 Xe, can have significant effects on reactor operations, because:

its concentration can change dramatically on a time scale of hours,

its thermal absorption cross section is 3 106 barns.


135 Xe production and loss mechanisms are sketched in Fig. 8.11.

284
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Figure 8.11: 135 Xe production and loss mechanisms. Mention that some references indicate slightly
different percentages, but this doesnt change anything at the level of our discussion here.

Observations:

Feedback from 135 Xe is delayed

Feedback from 135 Xe is substantial:

In a commercial LWR, equilibrium 135 Xe at 100% power can change reactivity by


5$ or more!

Following a power increase, the 135 Xe concentration will

first decrease, then increase to higher equilibrium

Following a power decrease, the 135 Xe concentration will

first increase, then decrease to a lower equilibrium

Interestingly, the 135 Xe concentration can vary significantly at different locations in a large core.
If the spatial distribution of power changes, for example because of control-rod motion, then the
135 Xe concentration responds accordingly in different spatial regions. This can lead to

xenon oscillations,

which in extreme cases can grow over time and significantly disrupt reactor operations.

285
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

There is another fission product that is worth mentioning: samarium-149 (149 Sm). Remarks:

149 Smhas a lower thermal absorption cross section ( 5 104 b) than 135 Xe, and is produced
by only 1.4% of fissions.

It is stableit does not decay. Thus, it builds up following shutdown as its parents decay, but
it then stays in the core until the reactor returns to power and provides neutrons to destroy
it via absorption reactions.

Its reactivity worth is significant: it can be greater than $1

8.10.2 Solutions Including Feedback

We begin with some assumptions:

1. We assume that the reactor remains intact. [No melting or explosions.]

2. We assume that the balance of plant stays about the same during the time period we study.

The first assumption means there are no major changes in the reactor geometry. The second means
there are no major changes in the inlet temperature and pressure of the coolant.

Under the above assumptions, power changes will cause short-term reactivity changes only by
changing the temperatures of reactor components. (We are taking the view that density changes
are the result of temperature changes.) That is, except for the reactivity due to fission products,

f (t) = f [T (~r, t)] , with T (r, t) = functional of P (t) (8.135)

This means the power coefficient of reactivity, P , will be quite useful in our calculations.

IMPORTANT EXAMPLE FOLLOWS

Example including feedback from temperature and 135 Xe

A critical reactor that had operated at high power for many months was shut down and has been
operating in steady state at hot zero power for a week. (Heat added by the coolant pumps keep
the temperature hot. Decay heat is contributing as well; see Fig. 8.12 to get an idea for how
much. Zero power means the fission rate is so low that it does not add noticeably to the reactor
temperature.) You are given the power coefficient of reactivity:

0.016% k/k
P = (8.136)
% power

286
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Figure 8.12: Decay heat rate as a function of time after operating for one year at a given power
level. Rate is expressed as a percentage of the previous steady power level. After 0.01 day ( 15
minutes) decay heat is down to 1.4% of previous power level. After 1 day it is 0.5%, but it takes10
days to reach 0.2% and about 3 years to reach 0.01%. It would take longer than 3 years to reach
0.01% if the reactor had operated for more than one year before shutdown. [Curve generated using
approximate function H(t) = P0 0.0065(days0.2 [(t)0.2 (t + t0 )0.2 ], where t = time since
shutdown and t0 = time of operation before shutdown. See Lewis chapter 1. Curve is roughly
straight line on log plot until t approaches t0 .

Be careful with units for power coefficients. The power coefficient given here says that if we
change from 0% to 100% power, the resulting temperature change will induce feedback reactivity
of (100%)(0.016%k/k) = 1.6% k/k = 0.016 = more than 2$ of negative reactivity.)

Question 1: Over the course of a few seconds, we insert positive reactivity = 0.01 by withdrawing
control rods. In the short term (several minutes), what power level will the reactor attain?

Answer 1: The reactor must attain an equilibrium power level such that

ext + f = 0

287
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

This is an important concept to understand and remember! The power must increase until the
feedback reactivity exactly balances out the inserted reactivity, thereby making the reactor critical
at the new power level. That is,

f = ext = 0.01 (8.137)

We know that the feedback reactivity associated with moving from power P1 to power P2 is:
if = constant
P
f P [P2 P1 ] (8.138)

These two equations tell us what P2 P1 must be:


P2 P1 0.01
0.01 = 0.00016 P2 P1 = % power = 62.5% power (8.139)
% power 0.00016

Recall that we started at hot zero power, P1 0. Thus we find that in the short term, following
an insertion of positive reactivity of magnitude 0.01, the reactor will stabilize at a power level of

P2 = 62.5%

Question 2: What will happen during the next day or two? [Think about 135 Xe!]

Answer 2: Because the reactor was at HZP for a week, there was essentially no 135 Xe in the reactor
when the control rods were moved. Now 135 Xe buildup will introduce negative reactivity over a
time period of several hours. This will cause power to decrease, which adds positive reactivity from
feedback to exactly balance the negative reactivity from 135 Xe buildup, until the reactor eventually
stabilizes with an equilibrium 135 Xe concentration at a new power level lower than 62.5%.

Question 3: Define
135
Xe (P ) reactivity induced by equilibrium Xe concentration at a power level of P.

Then if P = power coefficient of reactivity = constant, what is the equation for the new stable
power level? Call this P3 . Note that as defined, Xe (P ) is negative.

Answer 3: The power drop from 62.5% to the new power level must introduce just enough positive
reactivity to balance the negative reactivity introduced by equilibrium xenon at the new power level.
Thus the new P3 must satisfy the equation:

Xe (P3 ) + P [P3 62.5%] = 0 point out that P3 62.5% must be < 0 (8.140)

This equation has a unique solution for P3 . We could solve for P3 if we were given Xe as a function
of power.

288
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Question 4: After this new equilibrium power level is achieved, what will happen if we insert
positive reactivity of 0.005? What will happen in the next few minutes, the next few hours, and
the next day?

Answer 4a: In the next few minutes power will increase to a new stable level such that the
negative feedback will exactly match the positive insertion of 0.005. Let us call this P4 . We can
see from our previous calculation that this means:

power will increase by 31.25% to P4 = P3 + 31.25% .

(Note that Xe concentration cannot change much in a few minutes.)

Answer 4b: At this time the 135 Xe production rate will be almost the same as it was at the previous
power level, because it comes mainly from the decay of existing 135 I. However, its destruction rate
will be increased by the increased scalar flux associated with the increased power. Thus, in the
next few hours
135 Xe concentration will decrease and power will increase further.

Answer 4c: But the 135 I concentration grows during this time because the fission rate grows.
Eventually the production rate of 135 Xe from 135 I will exceed its destruction rate, and the 135 Xe
concentration will climb to a new, higher equilibrium level than it had at power P3 . Power will
decrease accordingly to a new steady level. Call it P5 . This new level must satisfy:

Xe (P5 ) + P [P5 0] + ext = 0 point out that f hs two parts (8.141)

Here ext and the feedback reactivities are written relative to the starting point at hot zero power.
[You can pick any reference point at which you know the power and Xe at that power.] Given our
reference point,

ext = 0.01 + 0.005 = 0.015 .


Again, if we were given the function Xe (P ), we could find the value P5 that satisfies Eq. (8.141).

289
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Question 5: At this point we have introduced positive reactivity totaling 0.015, and the reactor
has responded with feedback from increased temperature and feedback from buildup of 135 Xe. If
we now remove this reactivity of 0.015 that we originally introducedthat is, if we now introduce
(0.015) of reactivitywhat will happen in the next several minutes, the next few hours, and the
next day?

Answer 5a: There will be a prompt drop in power, then in the next several minutes the power
will continue to drop. Will the power reach approximately zero, where it began when we first
introduced the positive reactivity that we now remove? Or will feedback prevent this?

On a time scale of minutes, the 135 Xe concentration will begin to grow slowly (why?), but it will
take more than a few minutes to change appreciably. The temperature will drop in response to the
power drop, and this will introduce positive reactivity. Let the new power level be P6 . Then the
reactivity at this new power level must satisfy

= f [P6 ]few minutes + ext = Xe (P5 ) + P [P6 0] + 0 < 0 (8.142)

We see that no matter how low the power goes, since P6 cannot go below zero, the reactivity must
remain negative. (Recall that Xe and P are both negative. Also note that the Xe feedback
reactivity is correctly evaluated at P5 , not P6 , because it has not had time to change much.) Since
reactivity remains negative, the power (ignoring decay heat) must go to zero. That is,

P6 0,

and the reactor temperature returns to the level it had originally, before the first introduction
of positive reactivitythe temperature caused by heat from the pumps. (Here we ignore possible
temperature changes caused by changes in decay heat. We are trying to keep the example relatively
simple!)

Answer 5b: In the next few hours, what happens to reactivity and power? The change that
occurs on this time scale is in the xenon concentration, which now increases because its production
rate from 135 I decay is roughly the same as it was at P5 but now there is no loss from neutron
absorption. This means reactivity becomes even more negative. This has no appreciable effect on
power, because it is already approximately zero. The neutron population will continue to drop.

Answer 5c: In the next few days what happens? Now the 135 Xe will decay away, slowly removing
the negative reactivity that it had added, which adds positive reactivity. Let us define the new
steady-state power level as P7 . Then the new steady-state reactivity must satisfy

= 0 = Xe (P7 ) + P [P7 0] + 0 (8.143)

It is clear that
135 Xe = 0 and P7 = 0

are the unique steady-state solutions to this equation. (Recall that f and 135 Xe and are both
negative. If P7 were positive, then could not be zero.) So the reactor has returned to

its original critical state.

290
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

This makes sense, because it has the same materials, temperature, and geometry that it originally
had.
Note:

1. We have ignored samarium here. If the reactor had originally had zero 149 Sm, then after
our hypothetical power increase and decrease it will not have exactly the same materials
as it originally had, because now it will have some 149 Sm. This would keep the reactivity
negativethe reactor would not return to critical after its xenon decayed. (149 Sm is stable
and thus does not decay away.)

2. We have also ignored changes in decay heat. In reality, decay heat after our proposed power-
ups and power-downs would probably be slightly higher than it was at the beginning (after a
week at HZP). This would make the temperature slightly higher, at least in the fuel.

8.11 Nuclide Depletion and Production

So far in this chapter we have studied reactor phenomena that occur on the following time scales:

sub-second (fission-product energy deposition in fuel, or life span of a free neutron)

seconds (heat transfer from fuel to other components)

minutes (delayed-neutron emission)

hours-days (changes in fission-product inventories)

Now we examine longer-term changes, namely changes in nuclide inventories that occur as fuel
fissions. This depletion and production phenomenon is easy to understand. The main challenge is
keeping up with a host of production and loss mechanisms and rates for all of the nuclides that
affect the concentrations of the nuclides that we care about. For example, suppose we are interested
in how the Pu concentration changes at some position in a uranium-fueled reactor. We begin by
listing in Table 8.1 the production and loss mechanisms.

A key point to take away from Table 8.1 is that the production rate of a given nuclide depends on
the concentrations of many other nuclides. This means that tracking one nuclides concentration
requires that we track many other nuclides concentrations. For example, if we are interested in
calculating the 241 Pu concentration as a function of time, and we determine that alpha decay of
245 Cm is a significant source of 241 Pu, then we must also calculate the 245 Cm concentration as a

function of time, as well as any nuclides that provide sources of 245 Cm, and nuclides that provide
their sources, etc.

291
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Table 8.1: Production and loss mechanisms for Pu isotopes in U-fueled reactor.

Pu Isotope Production reactions Loss reactions


238 Pu 237 Pu(n,), 239 Pu(n,2n), 240 Pu(n,3n), decay, neutron absorption
238 Np(n,p), 238 Np
decay, 238 Am
elec-
tron capture, 242 Cm decay, 241 Cm(n,),
...
239 Pu 238 Pu(n,), 240 Pu(n,2n), 241 Pu(n,3n), decay, neutron absorption
239 Np(n,p), 239 Np
decay, 239 Am EC,
243 Cm decay, 242 Cm(n,), . . .

240 Pu 239 Pu(n,), 241 Pu(n,2n), 242 Pu(n,3n), decay, neutron absorption
240 Np(n,p), 240 Np
decay, 240 Am EC,
244 Cm decay, 243 Cm(n,), . . .

241 Pu 240 Pu(n,), 242 Pu(n,2n), 243 Pu(n,3n), decay, neutron absorption
241 Np(n,p), 241 Np
decay, 241 Am EC,
245 Cm decay, 244 Cm(n,), . . .

242 Pu 241 Pu(n,), 243 Pu(n,2n), 244 Pu(n,3n), decay, neutron absorption
242 Np(n,p), 242 Np
decay, 242 Am EC,
246 Cm decay, 245 Cm(n,), . . .

etc. etc. etc.

Since we do not want to track the entire chart of the nuclides, in practice we achieve a manageable
scope for depletion and production calculations by ignoring certain production mechanisms that
are judged to be relatively insignificant. It should be clear that:

After the sets of significant production and loss mechanisms are chosen for a chosen set of
nuclides, we must solve a
coupled set of ordinary differential equations
(each one stating change rate = gain rate loss rate)
to find all of the concentrations as functions of time.
Figures 8.13, 8.14, and 8.15 illustrate the effect that isotope production and depletion have on the
multiplication factors of two particular sodium-cooled fast reactors. (The plots were calculated by
Dr. Sunil Chirayath assuming that all control rods were out of the core, which is why they show
k > 1.) These reactors are designed to produce fissile materialthey have high conversion ratios,
where the conversion ratio is the ratio of fissile atoms produced to fissile atoms consumed. This is
why their multiplication factors do not change more than they do over the course of each cycle.

292
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Figure 8.13: k as a function of time in two example fast-spectrum reactors. Results from Dr. Sunil
Chirayath, who generated them using MCNP and Monteburns.

Figure 8.14: 239 Pu inventory as a function of time in two example fast-spectrum reactors. Results
from Dr. Sunil Chirayath, who generated them using MCNP and Monteburns.

293
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

Figure 8.15: Inventory of two U isotopes as a function of time in two example Th-fueled fast-
spectrum reactors. Results from Dr. Sunil Chirayath, who generated them using MCNP and
Monteburns.

A few notes of interest:

1. One core uses Pu for its fissile material and depleted U (almost all 238 U) as its fertile material.
It converts 238 U to 239 Pu.

2. The other core replaces part of the Pu-U fuel with a Th-U mixture. This core converts 232 Th
to 233 U in addition to converting 238 U to 239 Pu.

3. Both reactors operate for 180 full-power days before shutting down for 60 days.

4. Note that k increases somewhat during the 60-day shutdown, before new fuel is added (which
of course causes a large k increase).
Question: Why is this?
Answer: Because 239 Np and 233 Pa continue to decay, producing 239 Pu and 233 U

These figures illustrate the kinds of calculations that are now performed routinely in reactor anal-
ysis. This hopefully provides a sense of the importance of depletion and production in long-time
reactor behavior, as well as a sense of the methods that are now available to calculate these phe-
nomena.

294
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

8.12 Summary

In this chapter we have studied the time-dependent behavior of the neutron population in reactors,
a subject known as

reactor dynamics
We examined phenomena that take place on time scales ranging from picoseconds to years, noting
that phenomena on these widely varying time scales all have significant impacts on reactor behavior.
We focused considerable attention on

reactor kinetics,

which considers changes in neutron population while mostly ignoring the coupling to other phe-
nomena (such as heat transfer and fluid flow). We derived the

reactor kinetics equations

to describe this.
We noted that if the shape of the neutron populationthe neutron distribution in position,
energy, and directiondoes not change much over a given time period, so that only the amplitude
of the angular flux function changes with time, then many coefficients in the kinetics equations
become constant in time. The resulting model is called the

point reactor kinetics equations (PRKEs)


This is the model that we employed the most to describe reactor behavior on time scales from
sub-second to many minutes. We studied PRKE solutions in the six cases that are possible:

subcritical, critical, supercritical

each with

fixed source present or not present.

We sketched how the neutron population (or power) changes with time in each case.
We studied the inhour equation and its solutions, which are the time constants in the exponentials
that describe the PRKE version of the neutron population during a transient. We found that
the dominant exponential has a time constant, which we called s1 , that has the same sign as the
reactivity.
We derived simple expressions for s1 in several limiting cases of interest, including small reactivity,
large negative reactivity, and large positive reactivity.
We studied the prompt-jump approximation and noted that it can be quite useful for any negative
reactivity insertion and for small positive insertions.
We discussed feedback reactivity and introduced reactivity coefficients that quantify this feedback
reactivity. We discussed feedback mechanisms, most of which are driven by temperature changes.
We also discussed feedback due to changes in fission-product inventorieschanges that occur on
time scales much longer than temperature changes.

295
NUEN 601, M.L. Adams notes, 2017 Chapter 8. Time Dependence

We presented a detailed example of how reactor power responds to various combinations of imposed
reactivity, temperature-induced feedback reactivity, and Xe-induced feedback reactivity. We noted
that this example is important!

Finally, we discussed the effects of long-term nuclide depletion and production, how to calculate
nuclide inventories, and some examples of nuclide inventory calculations.

296
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

Chapter 9

Neutron Distribution in Energy

9.1 Introduction

In previous chapters we have studied how neutron populations in reactors change with time un-
der different circumstances and how neutrons distribute themselves spatially in reactors. We also
derived the multigroup transport (MGT) and multigroup diffusion (MGD) equations, which con-
tain energy-averaged cross sections. We did not discuss where one might obtain the approximate
spectrum functions that are needed in order to generate the averaged cross sections.

In this chapter we seek insight into the energy spectrumhow neutrons distribute themselves in
energy in reactors. We start with the energy-dependent diffusion equation for the k-eigenvalue
problem, and once again we begin with the simple case of homogeneous reactors

9.2 Energy-Dependent Diffusion Theory

The energy-dependent diffusion k-eigenvalue equation is:



~
h
~
i (~r, E)
D(~r, E)(~r, E) + t (~r, E)(~r, E) = dE 0 (~r, E 0 )f (~r, E 0 )(~r, E 0 )
k
0

+ dE 0 s (~r, E 0 E)(~r, E 0 ) (9.1)


0

297
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

For a bare homogeneous reactor, if we accept the same extrapolation distance in the bound-
ary condition for every energy (which is not our most accurate boundary condition, but which
makes the forthcoming math tractable), the k-eigenvalue problem becomes

(E)
2
D(E) (~r, E) + t (E)(~r, E) = dE 0 (E 0 )f (E 0 )(~r, E 0 )
k
0
+ dE 0 s (E 0 E)(~r, E 0 ) (9.2)
0

(~r, E) = 0, ~r V + , (9.3)

where

V + = spatial domain enclosed by the extrapolated boundary of reactor,


V + = extrapolated boundary

Our equation involves derivatives in spatial variables and integrals in the energy variable. Let us
see if separation of variable leads to a solution:
?
(~r, E) = f (E)X(~r) (9.4)

When we insert this guess and perform algebraic manipulations we obtain


(E)
1 t (E)f (E) dE 0 (E 0 )f (E 0 )f (E 0 ) dE 0 s (E 0 E)f (E 0 )
2 X(~r) = k 0 0
(9.5)
X(~r) D(E)f (E)

Observe that

the right-hand side does not depend on position


and

the left-hand side does not depend on energy

Thus, each side must equal a constant. In keeping with our established tradition, we shall call this
constant Bn2 for the n-th eigensolution. We obtain two separate problems.

Spatial problem:

2 Xn (~r) = Bn2 Xn (~r), Xn (~rs ) = 0, ~rs V + , (9.6)

Energy problem:

(E)
D(E)Bn2 dE 0 (E 0 )f (E 0 )fn (E 0 )
 
+ t (E) fn (E) =
kn
0

+ dE 0 s (E 0 E)fn (E 0 ) (9.7)
0

298
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

We have encountered the spatial problem many times now, and we have solved it for several
interesting reactor shapes. We therefore assume that the solutions {Bn2 , Xn (~r)} of the spatial
problem are known.

The energy problem, Eq. (9.7), looks like a transport problem for an infinite uniform medium, with
added extra absorption equal to D(E)Bn2 . If we can solve it, its solution gives the energy shape of
the n-th mode of the eigenvalue problem as well as the n-th k-eigenvalue.

Let us sketch the solution procedure, beginning with some observations.

1. The amplitude of the solution fn (E) is arbitrary. If f(E) is any solution of Eq. (9.7), then
so is any constant times f(E). That is, this equation says nothing about fs amplitude. It
gives information only about f(E)s shape (which describes the energy distribution of the
neutrons in the reactor). Suppose f and k satisfy the equation. Then define the constant
Cint as follows:

1
Cint dEi f (Ei )f(Ei ). (9.8)
k
0

This is just a numbera constant. Divide the equation by this constant:



 2
 f(E) f(Ei )
D(E)B + t (E) = (E) + dEi s (Ei E) (9.9)
Cint Cint
0

This shows that the function f (E) f(E)/Cint , which is just a constant times the original
solution function f(E) (and thus is also a solution), satisfies the following:

D(E)Bn2 + t (E) fn (E) = (E) +
 
dEi s (Ei E)fn (Ei ) (9.10)
0

2. The fixed source in this equation, (E), is 0 for all E. The solution, fn (E), is therefore
non-negative for all E, which means its integral is positive.

3. Once we have the solution, fn (E), we recover kn by integrating Eq. (??) over all energies and
rearranging:

dE (E)f (E)fn (E)
kn = 0 2
(9.11)
0 dE [D(E)Bn + a (E)] fn (E)

This is recognizable as a correct expression for a k-eigenvalueit looks like production rate divided
by loss rate.

299
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

9.2.1 Solution of Fixed-Source Energy Problem

The energy problem is an infinite-medium slowing-down transport problem with a fixed source.
For simplicity of notation we shall drop the subscript n and write the equation as:
E0
2
dE 0 s (E 0 E)f (E 0 ),
 
D(E)B + t (E) f (E) = (E) + 0 < E < E0 . (9.12)
0

Here we have defined E0 to be the energy above which there are essentially no neutrons. In a fission
reactor this is around 10-12 MeV.
The energy problem breaks naturally into three ranges in a thermal reactor:

1. High-energy rangedominated by fission source: E > Es ,

2. Slowing-down rangedominated by downscattering: E (Eth , Es ),

3. Low-energy (thermal) rangein which upscattering is important: E < Eth ,


where

Es = a few hundred keV


[Note: 99% of fission neutrons are born with energy > 100 keV.]
[See Figure 9.1.]
Eth = thermal cutoff energy = energy above which there is negligible upscattering.
usually between 0.5 eV and 4 eV

We seek accurate simplifying approximations that will make it easier to solve our slowing-down
problem and to understand the solution. The following observation will motivate our first approx-
imation:

Elastic scattering off of heavy nuclides

causes little change in the neutron energy


if the nuclide is initially at rest.

This motivates the approximation:

A1 : For E > Eth , we shall not include elastic scattering from heavy nuclides in the explicit
t (E) or s (E 0 E) terms, although we do include it in the cross sections that go into D(E).
That is, we allow direction changes via scattering from heavy nuclides but not energy changes.

With this approximation we are basically subtracting heavy-nuclide elastic scattering from both
sides of the infinite-medium equation. Note that we still include the contribution of elastic scattering
from heavy nuclides in tr and thus in the diffusion coefficient, D.

300
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

1.2 4.0E-07

3.5E-07
1

3.0E-07

0.8
2.5E-07

0.6 CDF 2.0E-07


Fission Spectrum, E_inc=1MeV

1.5E-07
0.4

1.0E-07

0.2
5.0E-08

0 0.0E+00
0.E+00 2.E+06 4.E+06 6.E+06 8.E+06 1.E+07

1.E+00
Fission Spectrum (inverse MeV)

1.E-01

1.E-02

1.E-03
1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01
Energy (MeV)

Figure 9.1: Fission spectrum, with both probability density function (E) and cumulative distri-
E
bution function, CDF(E) = 0 dE 0 (E 0 ). Top: linear scale, with energy in eV and in eV1 .
Bottom: (E) on log-log scale, with energy in MeV and in MeV1 . You can see from the CDF in
the top figure that only 5% of fission neutrons are born with energies below 250 keV. It is difficult
to see from this figure, but < 1% are born below 100 keV. Only 6% are born with energies higher
than 5 MeV.

301
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

Inelastic scattering is a different story, because it can cause significant change in the neutron energy.
We will take a closer look at this. The cross-section plots in Figure 9.2 allow us to compare the
importance of inelastic scattering from U-235 and U-238 against the importance of scattering from
H-1 and O-16.

Figure 9.2: Scattering cross sections for key LWR nuclides. (From highest to lowest at 105 eV: H-1
elastic, O-16 elastic; U-238 inelastic, and U-235 inelastic.).

The figure shows microscopic cross sections, but it is the macroscopic cross sections that matter in
our equation. In a typical LWR there are several times as many H-1 or O-16 atoms as there are
U atoms on the order of 5 times as many. This and the figure tell us that elastic scattering from
hydrogen and oxygen has a much bigger impact than inelastic scattering from uranium nuclides,
especially for energies below a few hundred keV. This brings us to our second approximation:

A2 : For E < Es . we shall not include inelastic scattering in t (E) or s (E 0 E).

Finally, we observe that in a water-moderated reactor there will usually be more hydrogen atoms
than oxygen atoms, and we see from the figure that hydrogens scattering cross section is signifi-
cantly larger than oxygens over almost all of the energy range. This says that there will be more
scatters off of hydrogen than off of oxygen. This is further accentuated by the fact that on average a
neutron loses (1H )/2 half of its energy in a scatter from hydrogen but only (1O )/2 11%
of its energy in a scatter from oxygen.

302
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

The bottom line is that scattering off of hydrogen is by far the dominant mechanism for
neutron slowing down in LWRs. As we have seen, there are three reasons for this:

1. The scattering cross section for hydrogen is much larger than that for oxygen or for fissionable
nuclides.

2. There is more hydrogen than anything else.

3. The average energy loss per scattering collision is much greater for scattering from hydrogen
than from heavier nuclei.
We summarize and state mathematically the approximations we have made:

heavy
t (E) elastic (E), E > Es ,





t (E) t (E) heavy
s (E), E (Eth , Es ),





(E),
t E < Eth .




s (E 0 E) heavy 0
elastic (E E), E 0 > Es ,



s (E 0 E) s (E 0 E) heavy
s (E 0 E), E 0 (Eth , Es ),




(E 0 E),

E 0 < Eth .
s

Note: If we analyze reactors other than LWRs we must re-evaluate the approximations we have
made in this section.

9.2.2 Eth < E < E0 : Fast + slowing-down energy ranges

Previously we recognized that in a LWR, most of the slowing down of neutrons is cause by scattering
off of H-1. Because of this and because of the special simplicity of the H-1 differential scattering
cross section, it makes sense to solve a problem with H-1 as the only scatterer. We do this next,
before considering the more general problem.

303
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

9.2.2.1 Slowing down: H-1 plus heavy absorbers

Here we consider our fixed-source problem that looks mathematically like an -medium problem:
E0
2
dE 0 s (E 0 E)f (E 0 ),
 
D(E)B + t (E) f (E) = (E) + 0 < E < E0 . (9.13)
0

with the following cross sections:

t (E) = H F
s (E) + a (E), (9.14)
1
E < E0,

E0 ,
0 0 0
s (E E) = H
s (E E) = s (E ) (9.15)
0, E> E0

where F refers to fuel + other non-hydrogen absorbers

We can solve this equation analytically. The solution is

SOLUTION, SLOWING DOWN IN HYDROGEN WITH MASSIVE


ABSORBER


1 E0
 
1 0 0 0 0
f (E) = (E) + dE (E )C(E )p surv (E E) (9.16)
t (E) + D(E)B 2 E E

where
s (E 0 )
C(E 0 ) (9.17)
D(E 0 )B 2 + t (E 0 )

( )
E0
1 D(E 00 )B 2 + a (E 00 )
psurv (E 0 E) exp dE 00 00 (9.18)
E E D(E 00 )B 2 + t (E 00 )

( )
E0
1
dE 00 1 C(E 00 )
00

= exp (9.19)
E E

The first term, C(E 0 ), is the probability that a neutron, born from fission with energy E 0 , scatters
before it leaks or gets absorbed. The second term, psurv (E 0 E), is the probability that a neutron
of initial energy E 0 survives to slow down to energy E, given that it scattered on its first event.

304
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

As we have seen, (E) is small for E < Es . As a result:


 
E<Es 1 1
f (E) psurv (Es E) Seff (9.20)
E t (E) + D(E)B 2

where
E0
Seff dE 0 (E 0 )C(E 0 )psurv (E 0 Es ) (9.21)
Es

[Here we have used the identity psurv (E1 E3 ) psurv (E1 E2 ) psurv (E2 E3 ).]
Let us examine this solution. Recall that for most of the slowing-down energy range:

1. a (E)  s (E) except at resonances

2. s (E) constant except at resonances

It follows that for most of the energy range of interest,



constant, E
/ a resonance,
2
t (E) + D(E)B = (9.22)
much larger, E a resonance.

Further, for E between resonances, psurv (Es E) changes very little, because a (E) 0 except
in resonances, and D(E)Bg2 is small in a large reactor. [See definition of psurv .]
If we put all of this together we observe that:

1. The basic spectrum of the neutron flux in the slowing-down region in a thermal reactor is:

constant times 1/E

2. This basic shape is perturbed as follows:

(a) Leakage causes


a gradual drop below constant/E
as E decreases.
(b) Resonances cause
down-spikes at resonance energies
and each resonance causes
an additional small drop
below constant/E (because it causes a small decrease in psurv (Es E)).
If we zoom in on an energy interval that contains a few resonances, we should therefore see
something like the sketch shown in Fig. 9.3.

305
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

Figure 9.3: Scalar flux spectrum in slowing-down region near resonances. Horizontal: log[E].
Bottom curve is log[t (E)]. Top straight dashed line = log[constant/E]. Solid line underneath =
log[(E)].

9.2.2.2 Slowing-Down Terminology

In this section we briefly define and discuss some standard variables, nomenclature, and terminology.

Lethargy (u)

The lethargy variable is often used in place of neutron kinetic energy. Definition:

 
E0
u = lethargy = ln (9.23)
E

where E0 is some reference energy, usually chosen to be the high-energy cutoff. Note that as E
decreases, the lethargy increases. (Slower neutrons are more lethargic!)

306
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

The lethargy variable is used extensively in slowing-down studies, chiefly because it simplifies the
equations. Notice:

As E decreases, u increases,

As E 0, u

If E < E0 , u > 0 .
h i h i h i
E0 E0 Ei
uf ui = ln E f
ln Ei = ln Ef

E = E0 exp(u)

We often perform a change of variable and talk about densities in lethargy instead of densities
in energy. For example, a lethargy-dependent density, Fleth (u), is related to the corresponding
energy-dependent collision density, Fen (E), as follows:

dE
Fleth (u)du = Fen (E)dE Fleth (u) = Fen (E) (9.24)
du

(The minus sign is because u increases when E decreases.) This is like a change of variable in an
integral.

From the definition of u,


dE d 
E0 eu = E0 eu = E .

= (9.25)
du du

Thus,
Fleth (u) = E Fen (E). (9.26)

The energy-dependent scalar flux is a density in energy. Thus, we can define a lethargy-
dependent scalar flux, (u), as follows:

(u) = E(E) (9.27)

We are abusing notation herewe shouldnt use the same symbol for the two different functions
(u) and (E). However, this is common practice in nuclear engineering.

Note that t (u) = t (E) (no factor of E), because is not a density in energy or lethargy!
Similarly, D(u) = D(E) (no factor of E).

Question: What are the units of (u) and (E)?

Answer: (E) has units (n-cm)/(cm3 -s-MeV); (u) has units (n-cm)/(cm3 -s)

307
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

Average Lethargy Gain Per Collision: (Worm)

A quantity of considerable interest is the

average lethargy gain per scattering collision.

Lethargy gain is the same thing as logarithmic energy decrement:

uf ui = ln(E0 /Ef ) ln(E0 /Ei ) = ln(Ei /Ef ) = ln(Ei ) ln(Ef ) (9.28)

We give the average lethargy gain per scattering collision the symbol , which we lovingly call
worm. (It is really the Greek letter xi.) Its definition is:


(Ei ) = dEf [u(Ef ) u(Ei )] P (Ei Ef )
0
 
Ei
= dEf ln P (Ei Ef ) (9.29)
Ef
0

We cannot evaluate this analytically, in general. However, given the special case of s-wave elastic
scattering, we can:

 
Ei
(Ei ) = dEf ln p(Ei Ef ) (9.30)
Ef
0
Ei  
s-wave elastic Ei 1 1
dEf ln (9.31)
Ef 1 Ei
Ei
1  
x=Ef /Ei 1 1 1
Ei dx ln (9.32)
1 Ei x

1
= [1 + ln ] (9.33)
1
Hence,

ln
=1+ (9.34)
1
Only if scattering is elastic and isotropic in the CM frame

308
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

If we have a mixture of different nuclides, we can define an average , denoted (worm-bar):


#nuclides
P
s (E)i (E)
i=1
(E) = #nuclides
(9.35)
P
s (E)
i=1

Note: Because in general the scattering cross sections depend on energy, so in general will (E)
worm-bar. However, for energies in the slowing-down range (1 eV to 100 keV), the scattering
cross sections of most light nuclides are nearly constant. Thus, in back-of-the-envelope calculations,
we often ignore the energy dependence of worm-bar and simply use a suitable average value over
the slowing-down energy range.

Values of worm-bar for several moderating materials of interest are given in Table 9.1.

Table 9.1: Average logarithmic energy decrement (= avg. lethargy gain) per scatter for
various common moderator materials. (Uranium is included for comic contrast.)

Material A
H 1 1.000
H2 O 1 and 16 0.93
D2 O 2 and 16 0.509
4 He 4 0.425
C (graphite) 12 0.158
238 U 238 0.008

309
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

9.2.2.3 Slowing down solution in terms of lethargy: 1 H plus massive absorber

In Eq. (9.16) we presented the energy-dependent flux in a homogeneous medium of hydrogen plus a
massive (A  1) absorber. We can rewrite this solution in terms of lethargy-dependent quantities
if we recognize the identities:

f (E) = f (u)/E (9.36)


(E) = (u)/E (9.37)
t (E) = t (u) (9.38)
D(E) = D(u) (9.39)
0 0
C(E ) = C(u ) (9.40)
0 0
psurv (E E) = psurv (u u) (9.41)
0 0
s (E E) = s (u u)/E (9.42)
0 0
dE = Edu (9.43)
0 0 0 0
E =Eu =u and E = E0 u = 0 (9.44)

These lead to:



1 u 0 0 (u0 )
 
f (u) 1 (u) 0 0
= + du E C(u )psurv (u u)
E t (u) + D(u)B 2 E E 0 E0
or
 u 
1 0 0 0 0
f (u) = (u) + du (u )C(u )psurv (u u) (9.45)
t (u) + D(u)B 2 0

Similarly,
 
u>us 1
f (u) psurv (us u) Seff (9.46)
t (u) + D(u)B 2

Note that while the energy-dependent f (E) is proportional to 1/E for energies between reso-
nances, the lethargy-dependent f (u) is constant between resonances.

9.2.2.4 Slowing down: including non-hydrogen scatterers

When we include scattering from nuclides other than 1 H, we cannot solve the slowing-down equation
so easily. However, we can show that for E  Es , the solution is well approximated by the following
function, which is not tremendously different from what we had with 1 H only:
 
E<Es 1 1 1
f (E) psurv (Es E)Seff (9.47)
E t (E) + D(E)B 2

310
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

We see from Eq. (9.47) that one change in the general-moderator case relative to the hydrogen-
only case is the presence of 1/worm-bar. Another change is that while psurv has the same physical
meaning as before, it is

no longer a simple exponential,

but is a more complicated function than when 1 H is the only scatterer.

9.2.2.5 How good is a given moderator?

Clearly, worm-bar is an important number in the evaluation of moderators. However, it is not the
entire story.

A good moderator should have

1. a high (or ) so that neutrons loose a lot of kinetic energy per collision;

2. a high scattering cross section, so that the moderating collisions will occur with high fre-
quency;

3. a low absorption cross section, so it doesnt cause neutron losses.

Table 9.2 provides these figures of merit for a few different moderators.

This reasoning leads to two figures of merit for moderators:

moderating power = s (9.48)

s
moderating ratio = (9.49)
a

Table 9.2 provides these figures of merit for a few different moderators.

You will notice that we have included average number of collisions from 2 MeV to 1 eV in our
table. What we have actually shown is the average number of collisions to give the collided neutrons
an average lethargy equal to that of a 1-eV neutron. This interesting quantity is easy to calculate
once we know :
h 6i
110
total lethargy gain ln 1 14.5
h# collisionsi = D E = =
lethargy gain
collision

311
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

Table 9.2: Various figures of merit for various moderators of interest.

avg. # of collisions
s
Moderator A from 2 MeV to 1 eV s
a
H2 0 1 and 16 0.920 16 1.35 71
D2 0 2 and 16 0.509 29 0.176 5670
He 4 0.425 43 n/a 83
C 12 0.158 91 0.06 192
U-238 238 0.008 1730 0.003 0.009

9.2.3 E > Es : Energy range in which fission source is prominent

If we ignore for the moment that the in-scattering source rate density, Sinscat (E), depends on the
solution, f (E), then we can solve Eq. (9.12), including the approximations we have made to
ignore certain kinds of scattering in certain energy ranges, for the neutron energy distribution:

(E) + Sinscat (E)


f (E) = (9.50)
D(E)B 2 + t (E)
where

t (E) t (E) heavy


elastic (E), (9.51)

1
D(E) = , with actual tr (E) (9.52)
3tr (E)

In a large reactor the denominator is dominated by e t (E). (It is easy to show that the leakage
terms effect on the denominator is proportional to [(mean-free path)/(reactor dimension)]2 ). Thus,
in a large reactor, we can understand the neutron energy distribution in the high-energy range if
we understand three functions:

1. The fission spectrum, (E),

2. The in-scattering source-rate density, Sinscat (E), and

3. The effective macroscopic total cross section, t (E).

9.2.3.1 Fission Spectrum

We have seen plots of the fission spectrum and understand its characteristics. The most likely birth
energy for a neutron emerging from a fission event is 700-800 keV. The average birth energy is
2 MeV. Less than 1% of neutrons emerge from fission with energies below 100 keV.

312
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

9.2.3.2 Inscattering

Recall the following about the scattering processes for high-energy neutrons:

Each elastic scattering off of a heavy nucleus causes the neutron to


lose very little energy
and thus has little effect on the energy distribution. From the point of view of the energy
distribution, it is almost as if the event never happened.

Each inelastic scattering event causes the neutron to


lose substantial energy.
This almost always removes the neutron from the energy range we are considering (which
is E > 100 keV). That is, inelastic scattering contributes relatively little to the scattering
source in this energy range.

Each elastic scattering off of a light nucleus causes the scattered neutron to emerge with
energy
between Ei and Ei ,
where Ei was the pre-scatter neutron energy, = [(A 1)/(A + 1)]2 for that nuclide, and
has a value well below unity.

The qualitative result of all of this is:

In the high-energy range (E > 100 keV),

Sinscat (E) is relatively low at the highest energies and increases as E decreases.

9.2.3.3 1/t (E) Term

We turn next to the 1/t (E) term in the solution. Consider reactors moderated by H2 O or D2 O,
which includes almost all of the commercial reactors built to date. Observe the following about the
total cross sections of the main materials in such reactors, for E ranging from 100 keV to 10 MeV:

The t (E)s for the uranium and plutonium isotopes decrease smoothly as E increases in this
range.

t (E) for 1 H decreases smoothly from > 10 b to < 1 b as E increases from 105 eV to 107 eV.

t (E) for 2 D decreases smoothly from 3 b to 1 b as E increases from 105 eV to 107 eV.

t (E) for 16 O has resonances in the energy range of interest here, superimposed on a smoothly
decreasing function that ranges from 4 b to 1.5 b as E increases from 105 eV to 107 eV.
The resonances produce a peak cross section of 17 b and also produce a sharp minimum
of a fraction of a barn and an energy between 2 and 3 MeV.

313
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

It follows that there are two main effects of the 1/t (E) multiplier:

1. The multiplier is higher at high energies and lower at lower energies, in the energy range
considered here.

2. The multiplier has sharp minima at energies at which oxygen has resonances, and sharp
maxima at energies at which oxygen has sharp minima between its resonances. (See plot of
oxygen cross section in Fig. 9.2.)

9.2.3.4 Combination

If we combine what we have learned about the uncollided and scattered neutrons, we conclude the
following about the distribution of neutron energies in a large homogeneous reactor:

For E > 100 keV, the neutron energy distribution can be viewed as a modified fission spec-
trum. The modifications include

1. The addition of the in-scattering source, which has little effect for the highest energies
but becomes stronger and stronger at lower and lower E.

2. The overall effect of the 1/t (E) multiplier, which is to raise the spectrum at the higher
energies in the range and reduce it at the lower energies in the range.

3. The detailed effects of resonances on the 1/t (E) multiplier, which cause sharp down-
spikes in f (E) at resonance energies and up-spikes at energies where resonance effects
cause t (E) to have sharp minima.

This will be illustrated when we plot f (E) in later sections.

9.2.4 E < Eth : Thermal energy range

In this range the fission source is zero, so the right-hand side has only the inscattering term. It is
useful to break this term into two parts: scattering from E 0 > Eth (downscattering) and scattering
from E 0 < Eth (thermal scattering, which includes downscattering and upscattering). We have:
Eth E0
0 0 0
2
dE 0 s (E 0 E)f (E 0 )
 
D(E)B + t (E) f (E) = dE s (E E)f (E ) + (9.53)
0 Eth
or
Eth
D(E)B 2 + t (E) f (E) = dE 0 s (E 0 E)f (E 0 ) + Sds (E)
 
(9.54)
0

where the downscattering source, Sds , plays the role of a fixed source for thermal neutrons:
E0
Sds (E) dE 0 s (E 0 E)f (E 0 ) (9.55)
Eth

314
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

9.2.4.1 Simplest thermal problem: no source, absorption, or leakage

Before we attempt the solution of our general thermal-range transport problem, let us discuss the
solution of a simpler equation, one with no sources or sinksonly scattering:
Eth
s (E)f (E) = dE 0 s (E 0 E)f (E 0 ) (9.56)
0

The solution of this problem is independent of the details of the cross sections, for physically
realizable cross sections. It is the Maxwellian distribution:
no source, absorption, or leakage
f (E) M (E, T ) (9.57)

"  1/2 #    
2 1 2kT E E
= nth exp (9.58)
kT m kT kT

Here M (E, T ) is the Maxwellian distribution for the scalar flux at a temperature of T , with:

nth thermal-neutron density [n/cm3 ] ,


m neutron mass ,
k Boltzmanns constant
eV 1 eV
8.617065 105
Kelvin 11605 K

T = temperature (absolute) .

The rather amazing answer, Eq. (9.58),

is true for ANY differential scattering cross section, s (E 0 E),

that is physically realizable. The proof of this is by statistical mechanics and is beyond the scope
of our study. In a nutshell: This is the distribution that has the highest entropy, which is what the
second law of thermodynamics would say the neutrons must attain. Do not forget the following
main result Eq. (9.58), which we now restate:

In the absence of absorption, leakage, and sources,

neutrons attain the Maxwellian distribution,

which is given by Eq. (9.58).

315
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

Notes on units and meanings:


E
1. kT has units of energy. kT is a dimensionless energy variable.

2. The energy dependence of the Maxwellian scalar flux is of the form  exp(), where  =
E/(kT ). The other terms do not contain the E variable. So if you want to see the shape of
the Maxwellian scalar flux, plot  exp().

3. The term (2kT /m)1/2 has units of speed, and 1/(kT ) has units of 1/energy. So the Maxwellian
flux has units of (neutron density) speed divided by energy, or (n-cm)/(cm3 -MeV-s), as it
should.
In Fig. 9.4 we plot the Maxwellian for two different temperatures. You can observe in the figure
something that is easy to prove: The peak of the Maxwellian scalar-flux distribution is at the
following energy:

Epeak = kT.

[You can prove this by setting dM /dE = 0 and solving for E.] That is, the most probable
energy for scalar flux is simply kT if the neutrons are in a Maxwellian at temperature T . Recall
that at room temperature, kT is approximately

k (293.15K) 0.0253eV.

A neutron with this energy has approximately the following speed:

2200 m/s = 220, 000 cm/s.

9.2.4.2 Realistic thermal problem with source, absorption, and leakage

For a thermal problem with leakage, absorption, and a downscattering source, the thermal flux
is no longer exactly a Maxwellian at the termperature of the background material. However the
thermal flux is often well approximated by a perturbed Maxwellian at a different temperature,
which is called the neutron temperature and given the symbol Tn .

316
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

10

0.1

kT = 0.0 1

0.01 kT = 0.0 2

0.001
0.0001 0.001 0.01 0.1

Figure 9.4: Energy-dependent scalar flux (not neutron density) for Maxwellian distributions at two
temperatures. Dashed line is for kTn = 0.01 eV (Tn 116 K) and solid is for kTn = 0.02 eV
(Tn 232 K).

Following are the main perturbations introduced by absorption, leakage, and the downscattering
source.

Absorption by 1/v absorbers: Preferentially depletes the lower-energy neutrons, shifting


the peak to the right and thus increasing Tn . Often called absorption heating.

Leakage: Preferentially depletes the higher-energy neutrons, shifting the peak to the left
and thus decreasing Tn . Often called diffusion cooling.

Downscattering source: Fills in the tail of the Maxwellian (increases the neutron population
at higher thermal energies) and blends it into the 1/E shape of f (E) in the slowing-down
range. (See section on slowing-down energy range.)

Absorption by low-lying resonances: Some heavy nuclides have resonances at the upper end
of the thermal-neutron-energy range. Neutrons are heavily absorbed in their vicinity. This
distorts the Maxwellian spectrumthe flux is depressed near the resonance energy. See Fig.
9.5 for an example of how the Pu in Mixed-Oxide (MOX) fuel perturbs the spectrum, relative
to Uranium-Oxide (UOX) fuel.

317
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

Figure 9.5: Comparison of the Maxwell distribution with the thermal spectrum in UOX fuel and
MOX fuel.

In analytic studies we usually assume that the thermal flux is well approximated by a Maxwellian
at a neutron temperature that may not equal the material temperature. Usually the effects of
absorption outweigh those of leakage, and the neutron temperature is higher than the material
temperature.

9.2.5 All Together: Thermal-Reactor Energy Distribution (Spectrum)

In the preceding sections we have figured out how neutrons should distribute themselves in energy
inside a large homogeneous thermal reactor (e.g., a reactor designed to slow down most of the
neutrons before they cause fission). Here we test our understanding by examining detailed solutions
of the neutron transport equation in a simple problem: An infinite square lattice in which each
square in the x-y plane contains one infinitely tall circular UO2 fuel rod (or pin) surrounded by
water. This is not a homogeneous reactor, so the solution will be different at different positions.

Figure 9.6 shows a high-fidelity calculation of the scalar flux at four different positions in the unit
cell, usually called a pin cell. The locations are: in the water between fuel pins, inside the fuel
but near the fuel/moderator interface, near the center of the fuel pin, and between the center and
outside of the fuel pin. The temperature of each material was taken to be 300 K in this example.

318
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

100
10-1 Inner UO2
10-2
Middle UO2
Outer UO2
10-3 Moderator
Flux (arb./eV)

10-4
10-5
10-6
10-7
10-8
10-9
10-1010-5 10-4
10-3 10-2 10-1 100 101 102 103 104 105 106 107 108
Energy (eV)
Figure 9.6: Energy-dependent scalar flux at various positions in an infinite square lattice of circular
UO2 pins in H2 O. Inner, Middle, and Outer UO2 refer to positions in near the center of the fuel
pin, about halfway between the center and surface, and inside the pin near the pin/water interface.
Moderator is in the water between fuel pins. All materials were at room temperature (near 300
K). (Plot courtesy of Andrew Till, who generated it using the CENTRM code in the SCALE code
system.)

319
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

There are many features worth noting in the spectrum plots of Fig. 9.6. The low-energy (ther-
mal) spectrum is

roughly Maxwellian

with a peak between 0.01 and 0.1 eV. In the moderator it is not far from kTmaterial 0.025 eV.
Detailed inspection reveals that the peak becomes lower and moves to the right as we move from
the moderator to the outer, middle, and inner portions of a fuel pin. This is an example of

absorption hardening or absorption heating.

(Absorption of thermal neutrons is strong in the fuel and weak in the water.) This is as predicted
in our discussion of the scalar flux in the thermal energy range.

Moving to slightly higher energies we see a steep drop in the spectrum just to the right of the
peak, as is characteristic of a Maxwellian. However, whereas a pure Maxwellian drops by a factor
of 1000 when energy goes from E = Epeak = kT to E = 10 Epeak , the figure shows a less
dramatic drop, because the exponential tail of the Maxwellian blends into the 1/E shape in
the slowing-down range. This was predicted in our discussion of the scalar flux in the thermal
range.

320
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

Figure 9.7: Energy-dependent scalar flux at various positions in an infinite square lattice of circular
UO2 pins in H2 O. Inner, Middle, and Outer UO2 refer to positions in near the center of the fuel
pin, about halfway between the center and surface, and inside the pin near the pin/water interface.
Moderator is in the water between fuel pins. This figure has some dashed lines superimposed to
help with quantitative interpretation. The vertical and horizontal lines are four orders of magnitude
apart. The function constant/E passes diagonally through two of the intersections, as shown by
the diagonal dashed line.

321
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

We turn next to the slowing-down energy range, which in this room-temperature reactor extends
from below 1 eV to 105 eV (100 keV). See figure 9.7, which is the same as the previous figure
except that some dashed lines have been added to help us quantify the behavior of the scalar flux.
Observe that in this range:

1. The overall shape of the scalar flux is


very nearly = constant/E,
as predicted by the slowing-down analysis in a preceding section.

2. The solution at each spatial point shows


sharp down-spikes
in the scalar flux at energies for which either 238 U or 235 U has a resolved resonance (which
means a resonance that can be well characterized because it is well enough known and its
extent is not too strongly overlapped by a neighboring resonance). The resolved resonance
range in this calculation ends at approximately 20 keV (2 104 eV); this is why the sharp
down-spikes in the flux plots do not appear above that energy. (There are plenty of resonances
above 20 keV, but they are
unresolved
and are not individually modeled in this calculation.)

3. The down-spikes are deepest for the inner and middle portions of the fuel pins (red and light-
blue plot lines, with red tips visible below some light-blue tips). They are not as deep at the
outer portion of the fuel pins, near the water (green curve, plotted on top of the light-blue).
The are not deep at all (but still visible) in the scalar flux in the water, between fuel pins
(dark-blue curve, plotted on top of the green), but they are still easily visible even on this
log-scale plot.

4. Close inspection shows that the scalar flux at all spatial points (in both fuel and water) is
on the dashed constant/E line at E 20 keV, but below this dashed line at E = 1 eV. If
there were no losses, the flux would remain on the constant/E line. The ratio of the actual
flux to the constant/E line at E = 1eV is psurv (20 keV 1 eV), which is the probability
that a neutron survives the downscattering journey from 20 keV to 1 eV. See Eq. (9.47).

5. The plot comes from an infinite-medium calculation. If there were leakage, then the scalar
flux would fall even farther below the constant/E curve, by an amount determined by the
probability of neutrons leaking during the slowing-down process. In a large reactor this
probability is low, and the effect of leakage is virtually undetectable in a plot of the energy-
dependent scalar flux.

6. Above the resolved-resonance range, the scalar flux is almost the same in the water as in each
part of the fuel: the red, light-blue, green, and dark-blue curves very nearly coincide. This
is because the fuel and water segments are not very many mean-free paths thick except for
neutron energies at which the fuel cross section is high, which means neutrons whose energies
are either low or at uranium resonances.

322
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

Finally we turn to the high-energy range, E & 105 eV. Here we see the effects of the fission
source, with energy distribution (E), superimposed on the constant/E line. We also see the
influence of the resonances in oxygen cross section in the range from 200 keV to a few MeV, with
down- and up-spikes in the scalar flux corresponding to up- and down-spikes in the oxygen cross
section. (See oxygen cross-section plot in Appendix C.) This is just as predicted in the discussion
of the high-energy range.

9.3 Complete Position-Energy Solutions

Now that we have found fn (E), which is the energy spectrum associated with the n-th spatial
mode, we can recover the n-th k-eigenvalue using Eq. (9.60):

dE (E)f (E)fn (E)
kn = 0 2
(9.59)
0 dE [D(E)Bn + a (E)] fn (E)

Of course, the multiplication factor is the eigenvalue associated with the geometric buckling:

dE (E)f (E)ff und (E)
k = 0  2
 (9.60)
0 dE D(E)Bg + a (E) ff und (E)

The fundamental-mode scalar flux as a function of position and energy is recovered from our
separation-of-variables equation:

f und (~r, E) = Xf und (~r)ff und (E) (9.61)

where ff und (E) is the solution of the energy equation with Bn2 = Bg2 = the geometric buckling, and
where Xf und (~r) is the eigenfunction of the Laplacian operator, on the given reactor spatial domain,
associated with the smallest B2 eigenvalue (which is the geometric buckling).

For example, according to our energy-dependent diffusion theory analysis, the energy-dependent
scalar flux of the fundamental mode in a bare homogeneous cylinder is

   
2.405r z
(r, z, E) 0 J0 cos ff und (E) (9.62)
R H


(E)+Sds (E)


D(E)Bg2 +t (E)
, E > Es ,

   
2.405r z 1 1
0 J0 cos E D(E)Bg2 +t (E) psurv (Es E)Seff , E (Eth , Es ), (9.63)
R H





M (E, Tn ), E < Eth .

The definitions of the terms are given in previous sections.

323
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

9.4 Multigroup Diffusion Solutions

In a previous section we addressed the solution of the energy-dependent diffusion equation for a bare
homogeneous reactor. We found that we could obtain a separation-of-variables solution if we used
a single extrapolation distance for all energies in the vacuum-boundary condition. Here we apply
the same basic approach to the multigroup diffusion k-eigenvalue problem for a bare homogeneous
reactor, which we write as:
G G
g X X
Dg 2 (~r, E) + t,g g (~r) = g0 f,g0 g0 (~r) + s,g0 g g0 (~r), g = 1, . . . , G, (9.64)
k 0 0
g =1 g =1

g (~r) = 0, ~r V + , (9.65)

Let us see if separation of variable again leads us to a solution:


?
g (~r) = fg X(~r) (9.66)

When we insert this guess and perform algebraic manipulations we obtain


g PG PG
1 t,g fg k g 0 =1 g 0 f,g 0 fg 0 g 0 =1 s,g0 g fg0
2 X(~r) = (9.67)
X(~r) Dg fg

Each side must equal a constant. In keeping with our established tradition, we shall call this
constant B 2 . We obtain two separate problems.

Spatial problem:

2 X(~r) = B 2 X(~r), X(~rs ) = 0, ~rs V + , (9.68)

Energy problem:
G G
g X X
Dg Bn2 + t,g fn,g =
 
g0 f,g0 fn,g0 + s,g0 g fn,g0 (9.69)
kn 0 0
g =1 g =1

We have encountered the spatial problem many times now, and we have solved it for several
interesting reactor shapes. We therefore assume that the solutions {Bn2 , Xn (~r)} of the spatial
problem are known.

The energy problem looks like a multigroup transport problem for an infinite uniform medium,
with added extra absorption equal to Dg Bn2 . If we can solve it, its solution gives the multigroup
energy shape of the n-th mode of the eigenvalue problem as well as the n-th k-eigenvalue.

324
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

The fission term is just the fission spectrum, g , times a constant, where the constant is the fission
source-rate density divided by kn . This constant is arbitrary, because the amplitude of the solution
{fg } is arbitrary. We choose an amplitude such that:
G
1 X
g0 f,g0 fn,g0 = 1 (9.70)
kn 0
g =1

Then we can write the energy problem in matrix form as

D1 Bn2 + t,1 s,11



s,21 ... s,G1 fn,1
D B 2 +
s,12 2 n t,2 s,22 . . . s,G2 fn,2

.. .. ..

..
. . . .
s,1G s,2G ... DG Bn2 + t,G s,GG fn,G

1
2
= . (9.71)

..


G
The two-group version, with the boundary between groups chosen such that all neutrons are born
in group 1 and up-scattering from group 2 to group 1 is negligible, is:

D1 Bn2 + t,1 s,11


     
0 fn,1 1
= (9.72)
D2 Bn2 + a,2

s,12 fn,2 0
The solution for the flux spectrum, which in a two-group problem is just the ratio of the two group
fluxes, is easy to obtain:
s,12
fn,2 = fn,1 (9.73)
D2 Bn2 + a,2
The solution for kn follows from summing the multigroup equations over all groups and rearranging
into a ratio of production rate to loss rate:
P2
g=1 g f,g fn,g
kn = P2 (9.74)
[D B 2 + ]f
g=1 g n a,g n,g

P2 fn,g
g=1 g f,g fn,1
= P2 f
2
g=1 [Dg Bn + a,g ] fn,g
n,1


1 f,1 + 2 f,2 D2 Bs,12
2 +
a,2
n
=
[D1 Bn2 + a,1 ] + [D2 Bn2 + a,2 ] D2 Bs,12
2 +
a,2
n

1 f,1 s,12 2 f,2


= + (9.75)
D1 Bn2 + a,1 + s,12 D1 Bn2 + a,1 + s,12 D2 Bn2 + a,2

325
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

The multiplication factor in a bare homogeneous reactor, according to two-group diffusion theory,
is this kn with the geometric buckling used for Bn2 :

1 f,1 s,12 2 f,2


k= + (9.76)
D1 Bg2 + a,1 + s,12 D1 Bg2 + a,1 + s,12 D2 Bg2 + a,2

a,1 + s,12 a,1 1 f,1 s,12 a,2 2 f,2


= + (9.77)

2
D1 Bg + a,1 + s,12 a,1 + s,12 a,1
2
a,1 + s,12 D2 Bg + a,2 a,2
| {z } | {z } | {z } | {z }| {z } | {z }
PFNL (1p) uF F p PTNL f T

We have multiplied and divided by various expressions to allow us to identify the terms that make
up the classic six-factor formula for k.

We can combine the fundamental-mode multigroup spectrum, {ff und,g , g = 1, . . . , G}, with the
fundamental-mode spatial eigenfunction for a bare homogeneous reactor of any shape to find the
multigroup diffusion approximation of the fundamental mode distribution in such a reactor. For
example, two-group diffusion theory says that the fundamental-mode flux in a bare homogeneous
shoebox-shaped reactor is:

 x    1, g = 1,
y  z

g (x, y, z) 0 cos cos cos (9.78)
a b c s,12 , g = 2.
D2 B 2 +a,2
g

Here 0 is a constant that equals the group-1 (fast) flux value at the reactor center.

326
NUEN 601, M.L. Adams notes, 2017 Chapter 9. Neutron Distribution in Energy

9.5 Summary

In this chapter we studied how neutrons distribute themselves in energy in reactors. We formally
considered only homogeneous reactors and used the energy-dependent diffusion equation as our
model. In addition, we accepted a single extrapolation distance for neutrons of all energies.

We found that the diffusion k-eigenvalue problem for a homogeneous reactor separates into two
problems:

1. a Helmholtz equation (Laplacian eigenvalue problem) for the spatial distribution functions;
and

2. a fixed-source steady-state infinite-medium problem, with an extra DB 2 term that functions


as an extra absorption cross section, for the energy distribution function.

When we studied the problem for the energy distribution we found that it divides naturally into
three different problems:

1. The problem for E > Es one or a few hundred keV, in which the fission source dominates.

2. The problem for Eth < E < Es , in which downscattering dominates and resonance absorption
is important.

3. The problem for E < Eth , in which neutrons attain approximate thermal equilibrium with
the material.
In the high-energy range, the flux spectrum is approximately the fission spectrum divided by the
total cross section, with a leakage correction term added to the total cross section.

In the downscattering (slowing-down) range, the flux spectrum is approximately a constant times
1/E, modified by down-spikes at resonances and also by some degradation from a 1/E shape due
to absorption and leakage losses.

In the thermal range, the flux spectrum is approximately a Maxwellian characterized by a neutron
temperature that is typically slightly higher than the material temperature.

Although we performed the mathematics only for a homogeneous reactor, we examined high-fidelity
computed solutions from a heterogeneous pin cell with UO2 fuel surrounded by water. We
found that these heterogeneous-reactor solutions agreed extremely well with the diffusion-theory
predictions we made using diffusion theory for homogeneous reactors.

327
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

Chapter 10

Perturbation Theory

10.1 Introduction

We can obtain exact results only for simple problems. However, we can use perturbation theory
to obtain approximate results for a more complicated problem if it is close to a problem with
a known solution. Perhaps more importantly, we can use this theory to gain accurate estimates
of the difference in some quantity of interest that is caused by a change in a reactor, even if our
calculations of the before and after quantities are less accurate.

The basic idea is to assume that the difference between the solutions of two similar problems is
small, then to derive an equation for the resulting change in some quantity of interest, making the
approximation that the product of two small term is negligible compared to a single small term.

In this chapter we assemble the mathematical tools that we will need for perturbation theory,
namely adjoints and the Fredholm Alternative Theorem, then apply these tools to perturbation
theory for one-group and multi-group diffusion problems.

10.2 Warmup

Suppose we know the fundamental-mode scalar flux, , and the multiplication factor, k for a reactor
in some configuration, according to a one-group diffusion (1GD) model. That is, we know and k
such that

~ r) + a (~r)(~r) = 1 f (~r)(~r),
h i
~ D(~r)(~ (10.1)
k

(~r) = 0 for r V + , (10.2)

where V + is the extrapolated boundary of the reactor.

328
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

Now suppose the reactor has water coolant with boric acid for reactivity control, and suppose we
make a small change to the boric acid concentration. The main change to the reactor properties
will be a change in the absorption cross section in the water. We define

0a (~r) perturbed cross section, (10.3)

a (~r) 0a (~r) a (~r) (10.4)

= the change, or perturbation, in a .

We denote the perturbed solution with primes (0 ). The perturbed eigensolution must satisfy:

~ 0 (~r) + 0 (~r)0 (~r) = 1 f (~r)0 (~r),


h i
~ D(~r)
(10.5)
a
k0

0 (~r) = 0 for r V + . (10.6)

Let us see what we can figure out about the change in k that is induced by the change in a . If
we multiply the original equation by the perturbed solution and multiply the perturbed equation
by the original solution, we obtain the following two equations:

~ r) + a (~r)0 (~r)(~r) = 1 f (~r)0 (~r)(~r),


h i
~ D(~r)(~
0 (~r) (10.7)
k

~ 0 (~r) + 0a (~r)0 (~r)(~r) = 1 f (~r)0 (~r)(~r) .


h i
~ D(~r)
(~r) (10.8)
k0
We integrate each equation over the extrapolated domain and then subtract:
 h i h i
~ D(~r)
d3 r (~r) ~ 0 (~r) 0 (~r)
~ D(~r)(~
~ r)
V+

 
0 1 1
+ 3
d r a (~r) (~r)(~r) = 0
d3 r f (~r)0 (~r)(~r) (10.9)
V + k k V +

Observe that by Greens theorem (a multi-dimensional form of integration by parts):


h i
~ D(~r)
d3 r(~r) ~ 0 (~r) = d2 r (~r)D(~r)en (~r) 0 (~r)
V+ V +


~ 0 (~r) (~
d3 rD(~r) ~ r) (10.10)
V+

329
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

Observe further that both and 0 satisfy extrapolated boundary conditions, which means they
= 0 in the surface integrals that we obtain after integration by parts. Further, the remaining
volume integrals are symmetric in and 0 , so they will subtract out of Eq. (10.9). The bottom
line is that the leakage terms disappear from Eq. (10.9), leaving us with
 
0 1 1
3
d r a (~r) (~r)(~r) = d3 r f (~r)0 (~r)(~r) (10.11)
V+ k0 k V+

Rearrange:

1 1 d3 r a (~r)0 (~r)(~r)
= V + (10.12)
k0 k 3
V + d r f (~ r)0 (~r)(~r)

Recall that
k1 1
=1 . (10.13)
k k
It follows that
   
0 1 1 1 1
= 1 0 1 = 0 (10.14)
k k k k

Let us define this change in reactivity to be . Then we now have, without approximation, the
following result:
3 r)0 (~r)(~r)
+ d r a (~
= V 3 (10.15)
r)0 (~r)(~r)
V + d r f (~

If the reactor is homogeneous, so that f and a are independent of position, then the result
becomes quite simple:

homogeneous a
(10.16)
f

Remark: we could have obtained this result directly from our knowledge of the 1GD k for bare
homogeneous reactors. Still, the manipulations we learned how to perform here will serve us well
in the more complicated problems to come.

10.3 Adjoint Operators and Adjoint Functions

Before we dive into perturbation theory we remind ourselves of some important mathematical
definitions and truths.

330
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

10.3.1 Adjoint definition

Consider a linear operator L, which operates on functions in some defined function space. For
example, if L is the negative of the Laplacian, L = 2 , then we might be interested in it operating
on the space of functions that satisfy extrapolated boundary conditions and whose second spatial
derivatives exist. In this case the equation

g = Lf

would mean that

g(x, y, z) = 2 f (x, y, z),

and would imply that

f (~r) = 0 for r V +

where

V + extrapolated boundary of the domain V.

We emphasize that the following is completely general, not restricted to Laplacians or to anything
else we encounter nuclear engineering: Given any operator L, the adjoint operator, L , is defined
as the operator that satisfies the following equation for all functions f and g in their defined function
spaces:

DEFINITION OF ADJOINT OPERATOR

 
f, L g = (Lf, g) (10.17)

where (f1 , f2 ) is the defined inner product of the functions f1 and f2 .

It is important to recognize that the definition of any adjoint operator depends fundamentally on

the definition of an inner product.

For some product of two functions or vectors to be a proper inner product, it must have several
important properties. Some of these are:

(f, g) = (g, f ) = a scalar (just a number) (10.18)


(f, g) = (f, g), any scalar (10.19)
(f, f ) 0 (10.20)
(f, f ) = 0 if and only only if f = 0 (10.21)

331
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

In most settings there are different inner products that could be chosen, and they produce different
adjoint operators. Many times, however, there are good reasons to choose a particular inner
product. In our applications this usually arises from physical considerations.

Example

Let us consider a simple example of a 22 matrix with real entries as our operator, and the usual
dot product as our inner product. To find the adjoint of the matrix A, then, we are seeking a
matrix A such that the following is true:

a11 a12

f1 g1 a11 a12 f1 g1
= (10.22)
f2 a21 a22 g2 a21 a22 f2 g2

We leave it as an exercise (a very simple one) to show that in this setting, the transpose of A is
the adjoint operator (matrix), A . That is,

a11 a12

a11 a21
A = = = transpose of A (10.23)
a21 a22 a12 a22

It follows that in the world of matrices and vectors, a symmetric matrix with real elements is

self-adjoint

In general, an operator L is called self-adjoint if L = L.

10.3.2 Fredholm Alternative Theorem

Given a linear operator L, suppose there is a nonzero function f that satisfies

Lf = 0 (10.24)

(This means L is not invertibleit is singular.) It follows that there exists a nonzero function f
that satisfies

L f = 0 (10.25)

332
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

Fredholm Alternative Theorem

If L is singular, then the equation

Lg = q (10.26)

has a solution g if and only if

(q, f ) = 0, where f : Lf = 0. (10.27)

That is, the solution of a singular equation with a driving term exists only if the driving term
is orthogonal to the adjoint eigenfunction that corresponds to a zero eigenvalue.

We will not prove the complete if and only if statement of the theorem, but we will show the
only if part. That is, we shall postulate that Lg = q and show that this can be true only if
(q, f ) = 0:

L f =0 if Lg=q
0 = (0, g) (L f , g) = (f , Lg) = (f , q). (10.28)

We shall simply take Fredholms word (and that of many others who have verified this) for the if
part of if and only if, namely that if (q, f ) = 0, then Lg = q does have a solution g.

10.4 Adjoints and Importance in Neutronics

10.4.1 Physical Discussion and Motivation

Suppose a neutron detector is in a fixed physical system. Suppose further that we wish to know
the detector response for any distribution of sources in the system. One option is to solve the
transport or diffusion equations once for each source distribution of interest, but this could be
time-consuming if we are trying to optimize a source distribution, for example, and there are lots
of trial-and-error tests. If we are clever enough, however, we might figure out a better way.

Suppose we could find an equation that transported neutrons backwards from the detector to
the rest of the problem domain, such that the probability of reaching any point in the domain
is the same as the forward probability of a source particle emitted at that point reaching the
detector. Then we could solve the backwards problem once for a given detector, and we could
use its solution to tell us the detector response to any postulated source in the problem.

The backwards-transport problem is the adjoint problem. If we solve the adjoint prob-
lem in a source/detector configuration, we obtain a function that tells us the importance of a
source at any point in the problem, where importance is proportional to the detector response
that the given source would generate.

333
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

10.4.2 Adjoint Operator for One-Group Diffusion

Let us begin by finding the adjoint operator for a one-group diffusion (1GD) fixed-source problem.
We shall assume for simplicity that our material has k < 1 and that it is surrounded by vacuum.
We define the non-negative net-removal cross section as
assumed
r (~r) a (~r) (~r)f (~r) 0. (10.29)

Then the 1GD operator is defined by


h i
~ D(~r)f
Lf ~ (~r) + r (~r)f (~r) (10.30)

We wish to find the adjoint operator, L . As we have learned, this requires that we define an inner
product. For our purposes a simple integral works well:

(f, g) d3 r f (~r)g(~r) (10.31)
V+

Given this inner product, let us find our adjoint operator. By definition, this operator must satisfy
the following:

(L g, f ) = (g, Lf ), (10.32)

where f and g are differentiable functions that satisfy the extrapolated boundary condition. In-
serting definitions, we have
h i
d3 r L g(~r) f (~r) = d3 r g(~r) [Lf (~r)] (10.33)
V+ V+

 h i 
= ~ D(~r)f
d3 r g(~r) ~ (~r) + r (~r)f (~r)
V+


= 2 ~ (~r) +
d r g(~r)D(~r)en (~r) f ~ r) f
d3 r D(~r)g(~ ~ (~r)
V + V +


+ d3 r r (~r)g(~r)f (~r)
V +


g=0 on V +
 
d3 r ~ r) f
D(~r)g(~ ~ (~r) + r (~r)g(~r)f (~r) (10.34)
V+

334
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

We observe that if we interchange f and g in this expression, it does not change. That is, the inner
product (g, Lf ) is the same as the inner product (f, Lg), which is the same as (Lg, f because order
does not matter in an inner product. It immediately follows from Eq. (10.33) that

THE ONE-GROUP DIFFUSION OPERATOR IS SELF-ADJOINT

10.4.3 Importance for One-Group Diffusion in Source/Detector Problem

The 1GD equation for the forward problem is


h i
~ D(~r)(~
~ r) + r (~r)(~r) = S(~r) (10.35)

Let us assume vacuum boundaries for simplicity in this illustration, and use the usual extrapolated
boundary condition to represent this:

(~r) = 0 for r V + . (10.36)

The quantity of interest (QoI) in our postulated source/detector problem is the detection rate,
which is the integral of a detection rate density over the volume of the detector. The detection rate
density is a detection cross section, d , multiplied by the scalar flux. That is, our desired detection
rate, Rd , is

Rd d3 r d (~r)(~r) (10.37)
Vd

Obviously, if we were given S(~r) and then solved the forward equation, Eq. (10.35), we could
compute Rd . But then if we changed the source location or distribution, we would have to solve
Eq. (10.35) again. If we want to answer the general question, What is the detection rate for a
given source position?, we would have to solve Eq. (10.35) for each different source position of
interest.
There is another way to obtain Rd for any S(~r)a way that requires only one solution of a
differential equation. Consider the the following adjoint problem:
h i
~ D(~r)
~ (~r) + r (~r) (~r) = d (~r) (10.38)

with boundary condition

(~r) = 0 for r V + . (10.39)

Note that the source term in our adjoint problem is actually the detector cross section! Intuitively
this makes sense because we want to transport (or diffuse) particles backwards from the detector
to all other locations in the problem.

335
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

Let us now perform the mathematical manipulations that show us how to get the desired Rd from
the adjoint solution. We multiply the forward equation by the adjoint solution and vice versa, then
integrate and subtract:
 h i 
~ D(~r)(~
d3 r (~r) ~ r) + r (~r)(~r) S(~r)
V+

 h i 
~ D(~r)
d3 r (~r) ~ (~r) + r (~r) (~r) d (~r)
V+

=0 (10.40)

The r terms subtract out. Each volume-integrated leakage term transforms into a surface integral
and a different volume integral. Each surface integral disappears because both f and g are zero on
the extrapolated boundary. The different volume integrals both have the integrand D ~ .
~
These volume integrals are identical and subtract out. We are left with

0 = d3 r (~r)S(~r) d3 r (~r)d (~r) (10.41)
V + V +
| {z }
Rd

or

Rd = d3 r (~r)S(~r) (10.42)
V+

That is, we can obtain the desired detection rate

for any source distribution, S(~r)

by performing a volume integration of (~r) S(~r). That is, the value of at a given position
tells you how important a source would be at that position. This gives rise to the terminology

(~r) = Importance function

10.4.4 Adjoint Operator for Multi-Group Diffusion

Let us now find the adjoint operator for multi-group diffusion (MGD). The forward operator is
defined as follows:
h i
~ Dg (~r)f
[Lf ]g ~ g (~r) + t,g (~r)fg (~r)

G
X G
X
s,g0 g (~r)fg0 (~r) g (~r) f,g0 (~r)fg0 (~r) (10.43)
g 0 =1 g 0 =1

336
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

We wish to find the adjoint operator, L . As we have learned, this requires that we define an inner
product. We continue to use a phase-space integral, which in the multigroup world includes a sum
over groups:
G
X
3
(f, u) d r fg (~r)ug (~r) (10.44)
V+ g=1

Given this inner product, let us find our adjoint operator. By definition, this operator must satisfy
the following:

(L u, f ) = (u, Lf ), (10.45)

where both fg and ug satisfy the boundary condition for all group indices g. Inserting definitions,
we have the following (where we omit position arguments but remember that all functions in this
expression will in general depend on position):

G h
X i G
X
d3 r L u fg = d3 r ug [Lf ]g,~r (10.46)
V+ g,~
r V+
g=1 g=1

G

h i G G
~ Dg f
~ g + t,g fg
X X X
= d3 r ug s,g0 g fg0 g f,g0 fg0
g=1 V+ 0 0 g =1

g =1

G G
~ g+ ~ g f
~ g
X X
2
= d r ug Dg en f d3 r Dg u
g=1 V + g=1 V+

G
X G X
X G
d3 r t,g ug fg d3 r

+ s,g0 g + g f,g0 ug fg0
g=1 V+ V+ g=1 g 0 =1

G
ug =0 on V +
~ g f
~ g
X
d3 r Dg u
g=1 V+

G
X G X
X G
3
d3 r

+ d r t,g ug fg s,g0 g + g f,g0 ug fg0 (10.47)
g=1 V+ V+ g=1 g 0 =1

337
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

We now change the g and g 0 indices in the double summations to make it easier to recognize what
this equation is telling us:
G h i G
~ g f
~ g
X X
3
d r fg Lu = d3 r Dg u
V+ g,~
r V+
g=1 g=1

G
X G X
X G
d3 r t,g ug fg d3 r

+ s,gg0 + g0 f,g ug0 fg (10.48)
g=1 V+ V+ g=1 g 0 =1

or
G
X h i
d3 r fg L u
V+ g,~
r
g=1










G
h i G G

~ ~
X X X
3
= d r fg Dg ug + t,g ug s,gg ug f,g
0 0 g0 ug0
V + 0 0

g=1

g =1 g =1


| {z }


L u

(10.49)

The term in braces is the adjoint MGD operator operating on the function ug (~r). We see that
the leakage and total-collision portions of the operator are normal, but the scattering and fission
processes are backwards. This is what we should expect! Let us remember, then:

THE MULTI-GROUP DIFFUSION OPERATOR IS NOT SELF-ADJOINT

338
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

10.5 Perturbation Theory for One-Group Diffusion

Suppose we know the solution, k and , of the following one-group diffusion problem:
 
~ ~ 1
L = D(~r) + a (~r) f (~r) (~r) = 0, ~r V + , (10.50)
k

(~r) = 0, ~r V + (10.51)

Now suppose we wished to know the change in reactivity induced by the following changes in
material properties:

D0 (~r) = D(~r) + D(~r), (10.52)


0a (~r) = a (~r) + a (~r), (10.53)
0f (~r) = f (~r) + f (~r) (10.54)

The new solution, k 0 and 0 , must satisfy


h i
~ D(~r) ~ + a (~r) 0 (~r)

 
~ ~ 1 1
= D(~r) a (~r) + 0 f (~r) + 0 f (~r) 0 (~r) (10.55)
k k

with the same boundary conditions. We define the flux and reactivity perturbations

(~r) 0 (~r) (~r) (10.56)

   
0 1 1 1 1
= 1 0 1 = 0 (10.57)
k k k k

We now rewrite Eq. (10.55) further:


 
~ ~ 1
D(~r) + a (~r) f (~r) ((~r) + (~r))
k

   
~ D(~r)
~ a (~r) + 1 1 1
= 0
f (~r) + 0 f (~r) 0 (~r)
k k k

   
~ D(~r)
~ + a (~r) f (~r) + 1
= f (~r) 0 (~r) (10.58)
k

339
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

The left-hand side is L( + ), and we know that L = 0 by Eq. (10.50). Thus,


   
~ ~ 1
L((~r)) = D(~r) a (~r) f (~r) + f (~r) 0 (~r) (10.59)
k

We have made no approximations yet; weve just done some algebra.

We now apply the Fredholm Alternative Theorem to Eq. (10.59). Recall that L = 0, which means
L is singular. Recall that L = L (the one-group diffusion operator is self-adjoint). The theorem
says that a solution exists if and only if the inner product of with the right-hand side of
Eq. (10.59) vanishes. Because the one-group diffusion operator is self-adjoint, is just , so we
conclude that the inner product of with the right-hand side of Eq. (10.59) must be zero:
   
~ ~ 1
3
d r (~r) D(~r) a (~r) f (~r) + f (~r) 0 (~r) (10.60)
V+ k

or, solving for the change in reactivity,


h i
d3 r (~r ) ~ D(~r)
~ a (~r) + 1 f (~r) 0 (~r)
V+ k
= (10.61)
3
V + d r (f (~ r) + f (~r)) (~r)0 (~r)

This equation is still exact. It is also practically useless, because the unknown new scalar flux,
0 , appears in it. Thats where the perturbation theory comes in. We shall now assume that the
perturbations in material properties were small, and that the perturbation in the solution is also
small. We then throw out terms that are small under these conditions:
h i
d3 r (~r ) ~ D(~r)
~ a (~r) + 1 f (~r) ((~r) + (~r))
V+ k
(10.62)
3
V + d r (f (~ r) + f (~r)) (~r) ((~r) + (~r))
strike term in num & terms in denom

leaving

PERTURBATION-THEORY RESULT FOR , ONE-GROUP DIFFUSION

 
3 ~ ~ 2 1 2
d r (~r) D(~r)(~r) a (~r) (~r) + f (~r) (~r)
k
V+ (10.63)
d3 r f (~r)2 (~r)
V+

Note in particular that every term has 2 or a product of times derivatives of .

340
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

If we have a bare homogeneous reactor, with homogeneous perturbations, our result can be
simplified:
3
1
k f a D + d r (~
V r)2 (~r)
+ 3 2 r)
[Bare Homogeneous 1GD] (10.64)
f f V + d r (~

If in addition the unperturbed solution in the bare homogeneous reactor was in the fundamental
mode, the result can be further simplified:
1
f a DBg2
k [Bare Homogeneous, 1GD, fund. mode] (10.65)
f

Example Problem
A bare homogeneous slab of width a is critical. A thin slice, from x0 to x0 + x, is replaced by
a nonfissioning material with the same D and a . Using one-group diffusion theory, estimate the
change in reactivity.

Solution
We are given that:

D(x) = 0
a (x) = 0

f , x0 x x0 + x,
f (x) =
0, otherwise

We also know that:

k = 1, and
 x 
(x) = 0 cos ,
a
where 0 is a constant.

Thus, our general 1GD perturbation-theory result, Eq. (10.63), tells us


x0 +x x0 +x
dx k1 f 2 (x) dx cos2 x
  
x0 x0 a
a/2 = a/2 x

2 (x)] 2
a/2 dx [f a/2 dx cos a

   
x a 2(x+x) 2x
2 + 4 sin a sin a
=
a/2

Ask them what would be if the perturbed range were the whole reactor.

341
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

10.6 Perturbation Theory for Multi-Group Diffusion

We have discussed perturbation theory in the context of one-group diffusion (1GD) theory. We
now do the same in the context of multigroup diffusion (MGD) theory.

Suppose we know the solution {k and g , g = 1 . . . G} of the problem


G G
~ Dg (~r)
~ g + t,g g g X X
[L]g = g0 f,g0 g0 s,g0 g g0 = 0, (10.66)
k 0 0
g =1 g =1

1 1 ~ g (~rs ) = 0,
g (~rs ) + Dg (~rs )en (~rs ) ~r Vvac , (10.67)
4 2

~ g (~rs ) = 0,
en ~rs Vref (10.68)

We have suppressed the position argument in the MGD equation, but every term does depend on
position (~r) in general.

Note that we are not using the extrapolated boundary condition, but rather have expressed that the
incoming partial current is zero on the vacuum portion of the boundary and the normal derivative
of the scalar flux is zero on the reflecting portion. (On the vacuum portion, this is equivalent to
a linear extrapolated boundary condition instead of our usual extrapolation with curvature.) We
choose these boundary conditions to emphasize that the adjoint and how it is used are not restricted
to the case of extrapolation with curvature, but actually hold for all the boundary conditions we
would normally encounter in physical applications.

Now suppose we would like to know the multiplication factor, k 0 , of a similar problem with altered
material properties:
G G
~ 0 + 0 g g
~ D0 (~r)
 0 0 X X
0 0 0
L g = g g t,g 0
g f,g g 0 0 0s,g0 g g0 = 0, (10.69)
k0 0 0
g =1 g =1

1 0 1 ~ g (~rs ) = 0,
g (~rs ) + Dg0 (~rs )en (~rs ) ~r Vvac , (10.70)
4 2

~ 0g (~rs ) = 0,
en ~rs Vref , (10.71)

where the altered material properties can be expressed in terms of changes relative to the originals:

Dg0 (~r) Dg (~r) + Dg (~r), (10.72)


0t,g (~r) t,g (~r) + t,g (~r), (10.73)
0s,g0 g (~r) s,g0 g (~r) + s,g0 g (~r) (10.74)
g0 0f,g (~r) g f,g (~r) + g f,g (~r) (10.75)

342
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

We define the changes in reactivity and scalar flux:

g (~r) 0g (~r) g (~r) (10.76)

   
0 1 1 1 1
= 1 0 1 = 0 (10.77)
k k k k

Now we manipulate the equation for the altered solution into a form that is more convenient for
our study. First we define the perturbation in the operator, L, such that the following is true for
any function {fg } in the function space:
 0 
L f g [Lf ]g + [Lf ]g (10.78)

Then we have

L0 0
 
g
= [(L + L)( + )]g

= [L]g + [L ]g + [(L)( + )]g (10.79)

The equation for the perturbed solution in the perturbed reactor is [L0 0 ]g = 0, which means we
must have

[L ]g = [L]g [(L)( + )]g (10.80)

Of course, the original solution satisfies the original equation:

[L]g = 0 (10.81)

so we must have

[L ]g = [(L)( + )]g = (L)(0 ) g


 
(10.82)

We have not made approximationswe have simply manipulated the original and altered equations.
Now we invoke the Fredholm alternative theorem. There is a singular operator on the left-hand side
of our equation, so Fredholm tells us that this equation has a solution if and only if the right-hand
side is orthogonal to the adjoint eigenfunction of that operator:
  
, (L)0 = 0, (10.83)

where is defined such that


h i
L =0 (10.84)
g

343
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

Let us spell this out, employing the inner product that we previously defined for multigroup diffusion
problems:
G
X
d3 r g (~r) (L)0 ~r,g
 
0= (10.85)
g=1 V

G

G
g0 0 0f,g0
!
X
~ ~ 0g
X g0 f,g0
= 3
d r g Dg + t,g 0g g 0g0
V k0 k
g=1 g 0 =1


G
X
s,g0 g 0g0 (10.86)
g 0 =1

Manipulate the fission term:


 
1 0 0 1 1 1
0
g f,g g f,g = g0 0f,g g f,g
k k k k

1
= g f,g g0 0f,g (10.87)
k
Use this in the previous equation:

G

G  
X
~ ~ 0 0
X 1
0= 3
d r g Dg g + t,g g g g0 f,g0 g0 f,g0 0g0
0 0
V 0
k
g=1 g =1


G
X
s,g0 g 0g0 (10.88)
g 0 =1

Solve for :
PG h i
g=1 V d3 r g ~ 0g t,g 0g + g PG0 g0 f,g0 0 0 + PG0 s,g0 g 0 0
~ Dg
k g =1 g g =1 g
= PG 3 PG 0 0 0
g=1 V d r g g g 0 =1 g 0 f,g 0 g 0
(10.89)
This equation is still exact. It is also practically useless, because the unknown new scalar flux,
0 , appears in it. Thats where the perturbation theory comes in. We shall now assume that the
perturbations in material properties were small, and that the perturbation in the solution is also
small. We then throw out terms that are small under these conditions:

1. In the numerator and denominator, neglect compared to .

2. In the denominator, neglect g f,g compared to g f,g .

344
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

This leaves us with

MULTIGROUP DIFFUSION PERTURBATION-THEORY RESULT

PG h i
g=1 V d3 r g ~ g t,g g + g PG0 g0 f,g0 g0 + PG0 s,g0 g g0
~ Dg
k g =1 g =1
PG 3 PG
g=1 V d r g g g 0 =1 g 0 f,g 0 g 0
(10.90)

Recall that in the one-group diffusion case, the adjoint flux and the regular (or forward) flux
were the same. Thus, with 1GD theory, we can use perturbation theory without solving a separate
adjoint problem. With MGD,

we must solve both the forward and adjoint unperturbed problems

in order to use our perturbation-theory result. This may still have great benefit, though, if we want
to study the effects of more than one perturbation.

We remark that once is obtained from Eq. (10.90), it is easy to compute the multiplication factor
of the perturbed reactor:
1 1 1
= k0 = (10.91)
k0 k 1
k

10.7 Applications of Perturbation Theory

There are a host of practical applications of perturbation theory in nuclear reactor studies. Two of
the more common applications:

1. Estimating the reactivity change that will be introduced during an experiment. For example,
how will the multiplication factor change if we insert some material into a water hole in a
research reactor (perhaps to test its response to radiation)? Similar question for inserting
a detector, such as a fission chamber or a boron-loaded ion chamber, into some reactor
location? We would like to be able to quickly answer these questions for a variety of locations
and inserted materials/objects.

2. Estimating control-rod worth: how much does reactivity change if we move a given control
rod from height h1 to height h2 ? We would like to know this for a variety of h1 and h2 , and
for each control rod.

Perturbation theory allows us to answer these questions with only two full-core transport or diffusion
calculations (one forward and one adjoint).

345
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

10.8 Examples

10.8.1 Example: 2GD k-eigenvalue adjoint

Recall our usual 2-group diffusion equations, in which we assume negligible upscattering from group
1 (fast group) to group 2 (thermal group) and negligible neutrons born from fission into group 2,
and apply them to a bare homogeneous reactor:
1
D1 2 1 (~r) + [t,1 s,11 ] 1 (~r) = [1 f,1 1 (~r) + 2 f,2 2 (~r)] , (10.92)
k

D2 2 2 (~r) + [t,2 s,22 ] 2 (~r) = s,12 1 (~r), (10.93)

g (~r + 2Dg en (~r)) = 0, ~r V, g = 1, 2. (10.94)

The corresponding adjoint equations are:


1
D1 2 1 (~r) + [t,1 s,11 ] 1 (~r) = 1 f,1 1 (~r) + s,12 2 (~r), (10.95)
k

1
D2 2 2 (~r) + [t,2 s,22 ] 2 (~r) = 2 f,2 1 (~r), (10.96)
k

g (~r + 2Dg en (~r)) = 0, ~r V, g = 1, 2. (10.97)

Pay close attention to the fission and scattering terms. They are backwards from what you are
used to seeing. This is exactly what we should see in any adjoint problemthe physics operates
backwards.
Adjoint diffusion problems can be solved using exactly the same techniques we have used for solving
ordinary diffusion problems. For homogeneous problems, it remains true that if you know the B 2
eigenvalues of the Laplacian and you know the corresponding eigenfunctions, it is easy to solve
essentially any problem, including adjoint problems.

10.8.2 Example: 2GD perturbation

Suppose the two-group diffusion multiplication factor and fundamental-mode solution are known
for a bare homogeneous cylindrical reactor, radius R and height H, that is operating in
source-free steady state. The two-group diffusion fundamental-mode adjoint solution is also known.
(The fundamental-mode eigenvalues are the same for the forward and adjoint problems.)
A thin-walled air-filled tube of radius Rt  R is inserted along the axis into the top half of the
reactor. The neutronic effect is the same as if the cross sections inside the tube were set to zero
and the diffusion coefficient were set to a large number, say Dg0 (r, z) = D0  Dg for (r, z) in the
tube.

346
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

What is the approximate change in reactivity due to the insertion of the tube?

Solution:
We begin with the general MGD perturbation-theory expression:
PG 3 h ~ ~ g PG PG i
g=1 V d r g Dg g
t,g g + k 0
g =1 0
g f,g g 0 0 + 0
g =1 0
s,g g g 0
PG 3 PG
g=1 V d r g g g 0 =1 g 0 f,g 0 g 0
(10.98)

We recognize that perturbations in all material properties are zero except in the range r (0, Rt ),
z (0, H/2). In that range we have:

Dg (r, z) = D0 Dg , g = 1, 2, r (0, Rt ), (10.99)

t,g (r, z) = t,g , g = 1, 2, r (0, Rt ), (10.100)

s,g0 g (r, z) = s,g0 g , g 0 = 1, 2, g = 1, 2, r (0, Rt ). (10.101)

We recognize that the original reactor was operating in the fundamental mode, and we make the
approximation that its fast and thermal fluxes had approximately the same extrapolation distance.
This means we know that:
   
0 r z
1 (r, z) f1 J0 cos , (10.102)
R H
   
0 r z
2 (r, z) f2 J0 cos , (10.103)
R H
where

f1 0 , (10.104)

s,12
f2 0 , (10.105)
D2 B 2 + a,2

0 2.405 . . . = smallest zero of the J0 Bessel function, (10.106)


 2  2
2 0
B + = geometric buckling, (10.107)
R H

1, g = 1,
g = (10.108)
0, g = 2
k = 1. (10.109)

347
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

We also know that:


   
0 r z
1 (r, z) f1 J0 cos , (10.110)
R H
   
0 r z
2 (r, z) f2 J0 cos , (10.111)
R H

but we leave it as an exercise to determine the relation between f1 and f2 .

If we take this into account, the general perturbation-theory result for our specific case becomes:

[D0 Dg ]B 2 + t,g fg

Rt H/2
h    i2
P2 0 r z
g=1 fg 2rdr dz J0 cos

0 0 R H
g PG0 g0 f,g0 fg0 PG0 s,g0 g fg0
g =1 g =1

R H/2 h    i2
P2 PG 0 r z
g=1 fg g g 0 =1 g0 f,g0 fg0 0 2rdr H/2 dz J0 R
cos H
(10.112)

Some simplification is possible. In the denominator we can easily take advantage of knowing that
1 = 1 and 2 = 0. Also note that the z integral in the numerator is half of that in the denominator:

[D0 Dg ]B 2 + t,g fg

P2 Rt h  i2
f
g=1 g

0 rdr J0 R0 r
PG PG
1 g g0 =1 g0 f,g0 fg0 g0 =1 s,g0 g fg0
R h  i2 (10.113)
2
f1 G 0 r
P
0 0
g =1 g f,g g 0
0 f 0 rdr J0 R

Also note that because we know the original reactor was critical and in the fundamental mode, the
following must be true:
G
X G
X
Dg B 2 fg + t,g fg g g0 f,g0 fg0 s,g0 g fg0 = 0, g = 1, 2. (10.114)
g 0 =1 g 0 =1

This allows tremendous simplification in the numerator:


Rt
  h  i2
P2
fg 2 rdr J0 R0 r
1 g=1 D0 B fg 0
R h  i2 (10.115)
2
f1 G J0 R0 r
P
0
g =1 g 0 f,g 0 fg 0
0 rdr

348
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

We can go even farther if we care to, because we know the relation between f2 and f1 :
Rth  i2
D0B2 f1 f1 + f2 f2 0 rdr J0 R0 r
(10.116)
f1 (1 f,1 f1 + 2 f, 2f2 ) R rdr J0 0 r
h  i2
2
0 R

f2 f2 Rt h  i2
D0B2 1+
f1 f1 0 rdr J0 R0 r
= R h  i2 (10.117)
2 1 f,1 + 2 f,2 ff12 rdr J0 R0 r
0

f2 s,12 Rt h  i2
D0B2 1+
f1 D2 B +a,2
2
0 rdr J0 R0 r
= R h  i2 (10.118)
2 1 f,1 + 2 f,2 D2 Bs,12
2 +
a,2 0 rdr J0 R0 r

If we solved the 2-group diffusion adjoint problem, which is not difficult, we would obtain an
expression for f2 /f1 in terms of the cross sections, diffusion coefficients, and geometric buckling.
This would complete the analysis, giving a numerical estimate for , provided we could perform
the integration of the Bessel function (which is not difficult).

349
NUEN 601, M.L. Adams notes, 2017 Chapter 10. Perturbation Theory

10.9 Summary

In this chapter we introduced adjoint operators and presented the general definition of such an
operator. We noted that the definition of an adjoint depends on the definition of the function or
vector space of interest and also on the definition of an inner product.

We gave a simple example of a matrix with real coefficients, using the usual dot product as the
inner product in our standard vector space of real numbers, and we found that the adjoint of a
matrix of real coefficients in this setting is simply the transpose of the matrix.

This provided an opportunity to point out that symmetric real matrices are self-adjoint, where
self-adjoint describes an operator whose adjoint is itself.

We introduced the Fredholm Alternative Theorem, which declares that if L is singular, then Lu = q
has a solution if and only if q is orthogonal to an adjoint eigenfunction:

Lu = q has solution if and only if (f , q) = 0 where L f = 0

We defined an inner product for our diffusion-theory studies. It is simply an integral over volume
and sum over energy groups (or integral over energy if we are working with energy-dependent
equations).

We derived the adjoint of the one-group diffusion operator and found it equal to the original
operator. That is, we showed that the one-group diffusion operator is self-adjoint.

We showed how the adjoint operator and its solution can be used to save a great deal of effort
in certain kinds of source/detector problems. We introduced the term importance function and
showed that an adjoint problem can be defined such that the adjoint solution has a clear physical
interpretation as such an importance function.

We turned to the multigroup diffusion (MGD) operator and derived its adjoint. We found that
the MGD operator is not self-adjoint, because all of the collisional group-to-group exchanges are
backwards.
With our adjoint operators defined for one-group and multi-group diffusion theory, we presented
perturbation theory as applied to estimating the change in reactivity, , that is introduced by a
small change in a reactor, such as small movement of a control rod or a small change in temperature.
We derived general results for the one-group and multi-group cases.

We discussed typical applications of perturbation theory and provided examples.

This is not an exhaustive list of everything that is covered in this chapter!

350
NUEN 601, M.L. Adams notes, 2017

Part IV

Heat Transfer in Nuclear Reactors

351
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

Chapter 11

Heat Transfer in Reactors

11.1 Introduction

In previous chapters we have studied how neutron populations in reactors change with time under
different circumstances and how neutrons distribute themselves in position and energy. The be-
havior in time, position, and energy depend on material properties and geometry, as we knew from
the beginning. The dependence on material properties enters via cross sections.

We have recognized from the beginning of the course that cross sections depend on material tem-
perature. Macroscopic cross sections (s) are products of microscopic cross sections (s) times
nuclide number densities (N s). Number densities are proportional to mass density, which is influ-
enced by temperature (T ). Microscopic cross
sections
depend fundamentally on the relative speed
~
between the neutron and nucleus (vr = ~v V ). A neutron of a given velocity (~v ) sees nuclei

with a range of relative speeds, because the nuclei at any position in any material have a range of
~ s). As we have repeatedly noted, this means that the cross sections that appear in
velocities (V
our transport and diffusion equations must be properly averaged over the distribution of nucleus
velocities.
The properly averaged microscopic cross section is

~ )t,cold (|~v V
Ni (V ~ |)|~v V
~ |dVx dVy dVz
i i nucleus vels.


t ~v , N = (11.1)
v Ni (V~ )dVx dVy dVz
nucleus vels.

~ ) is the nucleus velocity distribution function for the nuclide of type i. In most cases of
Here Ni (V
practical interest, a sufficiently accurate approximation is that the nucleus velocity distribution is
isotropic and depends only on the material temperature. In this case the properly averaged cross
section becomes a function of neutron lab-frame speed and material temperature:
 usually
ti ~v , Ni ti (v, T ) (11.2)

352
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

The bottom line:

We cannot know the properly averaged cross sections at a given position if we do not know the
temperature at that position.
This is one reason we must understand heat transfer in our reactors. An equally important reason
is obvious: most reactors are designed to heat some working fluid, which is then used to do work
(including turning turbines connected to electrical generators), and we cannot properly design this
system if we do not understand heat transfer in the reactor.
These observations motivate this chapter, which briefly outlines heat transfer in fluid-cooled reactors
and develops some equations that model the heat-transfer phenomena and whose solution provides
temperature as a function of position.

11.2 Heat-Transfer Basics

It should be no surprise that we begin with our favorite equation:

Change Rate = Gain Rate - Loss Rate


Throughout the course we have applied this conservation statement to various neutron populations.
Now we shall apply it to the energy contained in some arbitrary volume of matter. We call this
volume V and denote its enclosing surface as V . We define

(~r, t) thermal energy density (energy per unit volume)


stored in the material at position ~r at time t,
= analog of neutron density in neutronics, (11.3)
000
q (~r, t) energy deposition rate density
= energy added to material per unit volume per unit time at position ~r at time t,
= analog of total neutron source rate density in neutronics, (11.4)
q~00 (~r, t) energy-flux vector
= net energy flow rate per unit area across a surface eq q~00 /|q~00 |
at position ~r at time t,
= analog of net current density in neutronics. (11.5)

With these definitions, we can write conservation of energy for volume V as follows:

3
d r (~r, t) = 3 000
d r q (~r, t) d2 r en q~00 (~r, t) (11.6)
t V V V

As we have seen before, we can use Gausss divergence theorem to convert the surface integral into
a volume integral:

3 3 000 ~ q~00 (~r, t)
d r (~r, t) = d r q (~r, t) d3 r (11.7)
t V V V

353
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

As we have seen before, if this equation holds for any arbitrary volume V , then the following must
be true:
~ q~00 (~r, t) = q 000 (~r, t)
(~r, t) + (11.8)
t
Our task in this chapter is to develop mathematical approximations for the q terms in this equation
so that ultimately we obtain equations whose solutions will tell us the temperature distribution
in the reactor. In what follows we shall express the energy-flux vector as the sum of energy-flux
vectors from three kinds of heat transfer, namely conduction, convection, and radiative:

q~00 (~r, t) = q~00 cond (~r, t) + q~00 conv (~r, t) + q~00 rad (~r, t) (11.9)

11.2.1 Volumetric energy deposition rate (q 000 )

In nuclear reactors the dominant source of energy deposition into the reactor material is transfer of
kinetic energy of fission fragments as they quickly slow down through Coulombic interactions after
they emerge with high energy from fission events. However, there are other important sources of
energy deposition. We list some potentially important sources and their mathematical expressions
below:

1. Fission fragments slowing down,

2. Recoil of matter electrons / nuclei / atoms / molecules / lattices from scattering by neutrons,
photons, or other particles,

3. Absorption of neutrons or photons,

4. Annihilation of anti-particles, such as positrons,

5. Electrical heating (sometimes used in experimental apparatuses in reactors),

6. etc.

For purposes of this chapter we shall assume that q 000 (~r, t) is given, and sometimes we may assume
that it is dominated by slowing down of fission fragments, delivering energy essentially at the
location and time of the fission event.

354
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

11.2.2 Conduction

Our math model for heat conduction is much like the diffusion approximation in neutronics. It
states that energy flow rate due to conduction is proportional to the negative of the temperature
gradient, which is an analog of stating that the neutron flow rate is proportional to the negative
of the scalar-flux or neutron-density gradient. Whereas Fick gets credit in the context of particle
diffusion, Fourier gets credit for heat conduction:

q~00 cond (~r, t) = kc (~r, t)T


~ (~r, t), (11.10)

n o
~ r, t) = D(~r, t)(~
neutron analog: J(~ ~ r, t)

where

kc (~r, t) thermal conductivity at position ~r at time t. (11.11)

The units of thermal conductivity are:


energy
.
length-time-degree

Common units are W/(m-K) (watts per meter-Kelvin). Thermal conductivity is given data
for our purposes, much like macroscopic cross sections are given for neutronics calculations.
Extensive tabulations of kc exist for a wide variety of material. Thermal conductivity is a function
of the material composition, temperature, and density (and sometimes material microstructure).

11.2.3 Convection

The math model we shall employ for convective heat transfer across a solid/fluid interface is simple:

q~00 conv (~rs , t) h(~rs , t) [Ts (~rs , t) Tf l (~rs , t)] es (~rs ), ~rs solid/fluid interface, (11.12)

where
 
energy
h(~rs , t) convective heat-transfer coefficient , (11.13)
area-time-degree
Ts (~rs , t) temperature of solid at interface, (11.14)
Tf l (~rs , t) temperature of bulk fluid near interface (not in boundary layer) (11.15)

The convective heat-transfer coefficient is often obtained from correlations, which are fits to ex-
perimental data that often have some basis in theory. A correlation for a given solid / fluid material
combination is typically expressed as a function of dimensionless parameters that characterize the
state of the materials.

355
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

11.2.4 Radiative Heat Transfer

Matter emits energy in the form of thermal photons at a rate proportional to T 4 times an
emission opacity, where T is absolute temperature. These photons stream at the speed of light
until interacting, with the dominant interaction typically being photo-electric absorption, which
deposits energy into the matter at the site of the absorption. Thus, thermal radiation can transfer
energy from one place to anotherthe phenomenon of radiative transfer.

In most nuclear reactors, conduction and convection are the dominant heat-transfer processes, and
thermal radiation is not usually modeled explicitly. We shall include a mathematical description
here for completeness, but we shall also ignore it in our discussion of the temperature distribution
in reactors, below.

The transport of thermal photons is described relatively accurately by the same equation that
we use to describe neutron transport, namely the linearized Boltzmann equation. If we neglect
scattering, which is usually a good approximation unless temperature is quite high, we have:
1 r ~ r (~r, , , t) + a (~r, , t)r (~r, , , t) = e (~r, , t)B(T (~r, t), ),
+ (11.16)
c t
where
energy
r radiation intensity, units = (11.17)
frequency-area-time-steradian
photon frequency, (11.18)
hP photon energy, (11.19)
34
hP Plancks constant = 6.62606957 10 J-s , (11.20)
a absorption opacity, units = length1 , (11.21)
1
e emission opacity, units = length , (11.22)
c speed of light, (11.23)
2h 3 1
B(T, ) Planck function = , (11.24)
c2 exp hP 1
 
kB T
energy
units =
frequency-area-time-steradian

1 eV
kB Boltzmanns constant . (11.25)
11605K
We can show that
4 5
2kB
d dB(T, ) = 4 2 h3
T4 , (11.26)
4 0 15c
| {z }
SB

356
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

where SB is the Stefan-Boltzmann constant. That is, the integral of the Planck function over all
emission frequencies

grows like the fourth power of T .

This causes radiative transfer to quickly become important as temperature increases. It is the
dominant heat-transfer mechanism in high-temperature systems, such as stars and high-energy-
density laboratory experiments. But it is usually not a significant player in most nuclear reactors.

The opacities in the radiative transfer equation are strong functions of temperature, as is the
Planckian, B. If radiative heat transfer is important in a given problem, then the heat-transfer
equations must be solved along with the thermal-radiation transport equation (or approximation
thereof)they form a tightly coupled system, with temperature and radiation intensity depending
strongly on each other.

If we do need to model radiative transfer, it enters into our energy conservation equation as follows:
h i
~ q~00 rad (~r, t) = d d e (~r, , t)B(T (~r, t), ) a (~r, , t)r (~r, , , t) (11.27)
0 4

That is, the matter loses energy by Planckian emission and gains it by photon absorption.

11.3 Simple Model of Single-Channel Heat Transfer

In this section we apply the energy conservation equation to a single fuel pin in a typical water-
moderated reactor, along with its surrounding coolant channel. This collection of pin and coolant
is referred to as a

pin cell.

When we study heat-transfer analysis of such a portion of a reactor, especially when we ignore
cross flow of coolant from one channel to and from its neighboring channel, we are performing

single-channel

analysis. We shall do this here.

We shall use physical reasoning to simplify many terms in the equations and obtain analytic results
that help greatly in allowing us to understand the main features of the temperature distribution in
nuclear reactorsat least those with long cylindrical fuel rods with coolant flowing past them. We
consider the steady-state case, for which our energy conservation equation becomes:

~ q~00 (~r) = q 000 (~r)


[steady state] (11.28)

357
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

The physical interpretation of this equation is:

net energy outflow rate density = energy gain rate density

The same equation would apply to a steady-state neutron population if there were no collision
terms.

11.3.1 Radial Temperature Profile in Pin Cell

We begin in the interior of a fuel pin, which we approximate as a tall solid cylinder of fuel inside
a cladding tube with a very thin annular gap between the fuel and cladding. In a typical water-
moderated reactor the fuel meat is less than 1 cm in diameter and more than 300 cm tall, so it
is strikingly tall and thin. We let z be the height above the bottom of the fuel and r be the
radial distance from the fuel-pin centerline. See Fig. 11.1 We shall analyze the heat transfer and
temperature profile moving from the inside to the outside in our single pin cell (single channel),
making reasonable simplifying assumptions as we go.

Figure 11.1: Illustration of simplified fuel pin. Show cross section, with notation Rf , rgap ,
Rg Rf + rgap , Rc , and rc = Rc Rg . Show 3D view and emphasize that H/Rf > 600.

11.3.1.1 Fuel interior

Inside the fuel, the heat transfer is dominated by conduction:



~00
q (~r)

q~00 cond (~r) (11.29)
rfuel
~

~ (~r)
kc (~r)T (11.30)

358
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

The volumetric heat generation (energy deposition) rate density, q 000 , has approximately the same
spatial shape as the fission rate density, which has approximately the same spatial shape as the
scalar flux. We now consider the axial, azimuthal, and radial shapes of q 000 .
As we have seen, if the reactor has approximately constant material properties from bottom to
top, the axial shape is approximately a sine or cosine function that peaks at the midplane. Het-
erogeneities such as partially inserted control rods, axial variations in coolant density, and axial
variations in temperature perturb this cosine shape, but for most of the length of the fuel rod, q 000
does not vary in the axial direction by more than a factor of 2 or 3, and the variations are mostly
smoothno large gradients, except possibly near the tip of a control rod or blade.
Across much of a commercial power reactor, pin cells have neighbors that look a lot like themselves.
As we have seen (see Chapter 9) the scalar flux does not change much across a fuel pin except at
neutron energies where the fuel has high cross sections, namely at resonances and at very low
neutron energies. In particular, the azimuthal variation in a given annular ring inside the fuel is
much smaller than the radial variation. We therefore make our

First simplifying approximation:

q 000 (~r) q 000 (r, z), ~r single fuel pin, r distance from pin axis (11.31)

That is, we ignore azimuthal spatial variations in the energy deposition rate in a single fuel pin.
This introduces negligible error into our analysis.
We noted above that the scalar flux does not change much across a fuel pin except at neutron
energies where the fuel has high cross sections. At such neutron energies there is a significant
variation in the scalar flux as a function of distance from the fuel-pin axis, but at other energies
there is not. For simplicity in our first basic analysis, we shall ignore radial variations in the energy
deposition within a fuel pin, even though we recognize that this introduces some error in the details
of the heat-transfer process but does not change the overall picture.

Second simplifying approximation:

q 000 (r, z) q 000 (z), r single fuel pin (11.32)

Remark:

We are ignoring q 000 radial and azimuthal variations within a given fuel pin.

We are not ignoring pin-to-pin variations in q 000 it may have a different magnitude and/or
different axial shape in different fuel pins, depending on their global positions.

We insert the approximations made so far into our steady-state heat-transfer equation for the
interior of a single fuel pin:
h i
~ kc (r, z)T
~ (r, z) = q 000 (z), r (0, Rf ) (11.33)

359
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

Our purpose here is to gain a qualitative and semi-quantitative understanding of heat transfer and
temperature profile in a nuclear reactor. We are not seeking high quantitative accuracy. It suits
our purpose to make a

Third simplifying approximation:

kc (r, z) constant = kcf , (r, z) fuel. (11.34)

This introduces some error in the details of the heat-transfer process but does not change the
overall picture.

This allows us to pull kc outside the divergence operator, leaving us with a Laplacian operating on
the temperature T :

kcf 2 T (r, z) = q 000 (z), r (0, Rf ) (11.35)

(Have you ever seen anything like this before?) Spelling out the Laplacian for our cylindrical fuel
pin, we have:

2 T (r, z)
   
1 T (r, z)
kcf r + = q 000 (z), r (0, Rf ) (11.36)
r r r z 2

Recall that the axial variation of q 000 is smooth and not steep, and recall that the fuel-pin height is
more than 500 times its radius. It follows that
T T
 (11.37)
z r
This leads to our

Fourth simplifying approximation:

2 T (r, z)
   
1 T (r, z) 1 T (r, z)
r + r (11.38)
r r r z 2 r r r

This introduces negligible error into our analysis, except possibly near axial interfaces such as the
fuel bottom, fuel top, or the end of a nearby control rod/blade.

We finally arrive at our heat-transfer model for the interior of a fuel pin:
 
1 T (r, z)
kcf r = q 000 (z), r (0, Rf ) (11.39)
r r r

Note that the axial position, z, is just a parameter in this equationthere are no derivative or
integral or algebraic operations that involve z. That is, each axial layer is independent of each
other axial layer, given the approximations we have made.

360
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

We can solve our radial conduction equation by rearranging and then integrating it twice:
 
T (r, z) r 000
r = q (z), (11.40)
r r kcf

T (r, z) r2 000
r = q (z) + C1 , (11.41)
r 2kcf

T (r, z) r 000 C1
= q (z) + , (11.42)
r 2kcf r

r2 000
T (r, z) = q (z) + C1 ln(r) + C2 . (11.43)
4kcf

We immediately recognize that


C1 = 0,

because otherwise the temperature would reach at r = 0. We can express C2 in terms of the
(yet unknown) temperature at the fuel surface, which is the fuel/gap interface. We call this Tf g .
We have
Rf2 Rf2
Tf g (z) = C2 q 000 (z) C2 = Tf g (z) + q 000 (z) = T (0, z) T0 (z) (11.44)
4kcf 4kcf
Thus, we can write our within-fuel temperature profile as follows, and we see that it is a constant
minus a parabola:

Rf2 r2 Rf2
T (r, z) = Tf g + q 000 (z) and T0 (z) = Tf g (z) + q 000 (z) (11.45)
4kcf 4kcf

We sketch this solution in Fig. 11.2.

11.3.1.2 Gap

We model the heat transfer across the fuel/gap interface and the gap/cladding interface using
convective-type coefficients:
h i
er q~00 hf g [T (Rf ) Tg ] = hf g [Tf g Tg ] (11.46)
Rf
| {z }
qr00 (Rf ,z)

h i
er q~00 hgc [Tg T (Rc rc )] = hgc [Tg Tgc ] (11.47)
Rf +rgap
| {z }
qr00 (Rg ,z)

361
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

Rf2 000
Figure 11.2: Sketch of temperature profile inside fuel meat. T (r = 0) = Tf g + 4kcf q (z), and it
drops quadratically to Tf g at the outer radius.

Here Tgc temperature at the inner surface of the cladding, at the gap/cladding interface.

In steady state, if we continue to ignore axial heat transfer, continuity of energy flow means:
 
00 T
qr (Rf )2Rf dz = kcf 2Rf dz = hf g [Tf g Tg ] 2Rf dz (11.48)
r rRf

 
2r 000
kcf q (z) = hf g [Tf g Tg ]
4kcf rRf

Rf 000
Tf g (z) = Tg (z) + q (z) (11.49)
2hf g

Sanity check: Since we are ignoring axial heat transfer, steady state requires that the energy flowing
out of a tiny axial slice of the fuel, of height dz, must equal the energy deposited in in the fuel in
the same axial slice. Stated mathematically this means:
Rf 000
qr00 (Rf , z)2Rf dz = q 000 (z)Rf2 dz qr00 (Rf , z) = q (z) (11.50)
2
The sanity check is passed.

362
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

Equation (11.49) gives the temperature drop from the fuel surface to the bulk gap material. If
we continue to ignore axial transfer, then the energy that flows into the gap from the fuel must
flow from the gap into the cladding:

hgc [Tg Tgc ] 2Rg dz = qr00 (Rf , z)2Rf dz (11.51)

Rf2
Tg (z) = Tgc (z) + q 000 (z) (11.52)
2hgc Rg

This gives the temperature drop from the bulk gap material to the inner surface of the cladding.
Note that Rg Rf , so without much error we could simplify our expression somewhat.

11.3.1.3 Cladding

Inside the cladding the heat transfer is dominated by radial conduction, as it is in the fuel. Unlike
the fuel, the cladding has negligible internal energy deposition, so our simple model in the cladding
is:
 
1 T (r, z)
kcc r = 0, r (Rf + rgap , Rc ) (11.53)
r r r

The solution procedure is the same as in the fuel, and the result is the same except that there is
no q 000 term:

T (r, z) = C1 ln(r) + C2 . (11.54)

The energy flow rate from the gap into the cladding must satisfy

 
T
kcc 2Rg dz = hgc [Tg Tgc ] 2Rg dz (11.55)
r rRg

We use the solution Eq. (11.54) to evaluate the derivative on the left-hand side, and we use the
conservation statement of Eq. (11.51) to replace the right-hand side:

 
C1
kcc 2Rg dz = q 00 (Rf , z)2Rf dz
r r=Rg
Rf Rf2 000
C1 = q 00 (Rf , z) = q (z) (11.56)
kcc 2kcc

363
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

We can now evaluate the temperature drop across the cladding:


" # " #
Rf2 000 Rf2 000
T (Rc ) T (Rg ) = q (z) ln(Rc ) + C2 q (z) ln(Rg ) + C2 (11.57)
2kcc 2kcc

or
Rf2
Tg (z) = Tc (z) + q 000 (z)[ln(Rc ) ln(Rg )]
2kcc

Rf2
 
Rc
= Tc (z) + ln q 000 (z) (11.58)
2kcc Rg

11.3.1.4 Cladding/Coolant interface

We model the heat transfer from the cladding outer surface using a convective heat-transfer coef-
ficient:

q 00 (Rc ) hcw [Tc Tw ] (11.59)

In steady state, if we continue to ignore axial heat transfer, continuity of energy flow means:
 
T
kcc = hcw [Tc Tw ] (11.60)
r rRc

We can evaluate the left-hand side using the solution of the conduction equation in the cladding:
2 Rf2 000
kcc Rf 000
   
T C1
kcc = kcc = q (z) = q (z) (11.61)
r rRc Rc Rc 2kcc 2Rc

This immediately yields the temperature drop from the cladding surface to the bulk coolant:

Rf2
Tc (z) = Tw (z) + q 000 (z) (11.62)
2Rc hcw

364
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

11.3.1.5 Summary of radial profile

We now have expressions for all of the temperature drops from the fuel centerline to the bulk
coolant:
Rf2
T0 (z) = Tf g (z) + q 000 (z) (11.63)
4kcf
Rf 000
Tf g (z) = Tg (z) + q (z) (11.64)
2hf g
Rf2
Tg (z) = Tgc (z) + q 000 (z) (11.65)
2hgc Rg
Rf2
 
Rc
Tgc (z) = Tc (z) + ln q 000 (z) (11.66)
2kcc Rg
Rf2
Tc (z) = Tw (z) + q 000 (z) (11.67)
2Rc hcw

We sum these equations to obtain the drop from the centerline to the bulk coolant:
   
1 1 1 1 Rc 1
T0 (z) = Tw (z) + + + + ln + Rf2 q 000 (z) (11.68)
4kcf 2hf g Rf 2hgc Rg 2kcc Rg 2Rc hcw

It is convenient and instructive to recognize the linear heat-generation rate:



0 our assumptions energy
q (z) d2 rq 000 (r, z) = Rf2 q 000 (z) (11.69)
pincell area length-time

We define a shorthand notation for the sum of the heat-transfer parameters, divided by to make
it convenient to use q 0 :
 
1 1 1 1 Rc 1
(z) + + + ln + (11.70)
4kcf 2hf g Rf 2hgc Rg 2kcc Rg 2Rc hcw

(We allow to vary with axial position to account for the fact that the heat-transfer coefficients
depend on details that can change with axial position.) This allows us to write our temperature
drop compactly:

T0 (z) = Tw (z) + (z)q 0 (z) (11.71)

Figure 11.3, taken from the Lewis book, sketches the radial profile of the temperature in the pin
cell. We see the quadratic shape of the temperature in the fuel, a drop across the tiny gap, not
much drop across the thin metal cladding, and a smooth drop to the bulk coolant temperature.

365
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

Figure 11.3: Illustration of temperature profile across a single pin cell. Taken from Lewis. Show
T0 , Tf g , Tgc , Tc , Tw . Change to r.

Remarks:

1. The analysis we have performed is straightforward and not complicated. With the simplifying
assumptions we made, by far the most difficult part of this analysis is figuring out what to
use for the heat-transfer coefficients kcf , kcc , hf g , hgc , and hcw .

2. It is straightforward to remove most of our simplifying assumptions if we are willing to perform


a more complicated analysis, especially if we are willing and able to employ some computing
power. The trickiest part remains figuring out what to use for the heat-transfer coefficients.

11.3.2 Axial Temperature Profile in Pin Cell

We have addressed the radial heat transfer in an arbitrary pin cell (single channel) and characterized
the temperature profile at height z as a function of the linear heat-generation rate, q 0 (z), and the
bulk coolant temperature, Tw (z). Here we discuss the axial shape of q 0 (z) and develop an expression
for Tw (z) as a function of inlet temperature, flow rate, and q 0 (z).

366
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

Recall that in a bare homogeneous rectangular parallelepiped reactor, diffusion theory predicts that
the axial shape of the flux, and hence the power density, is a cosine function. The same is true for
a homogeneous cylindrical reactor, and in fact it is true regardless of the reactors shape in the x-y
plane. Moreover, it is not difficult to show that according to diffusion theory, even if the reactor
is heterogeneous in the radial (or x-y) plane, as long as material properties are uniform in z, the
power profile is a cosine function.

In a real reactor several factors cause the axial power profile to deviate from a simple cosine function:

1. The top and bottom of the core are

not exposed to vacuum,

but rather are adjacent to materials that can scatter neutrons back into the reactor. This
flattens the power profileit still has a cosine shape, but it is stretched out so that it would
not extrapolate to zero close to the core top or bottom.

2. The material properties in the core are

not uniform.

The largest non-uniformity arises from partially inserted control rods, which strongly depress
the power profile in the axial range where they are deployed.

3. An additional non-uniformity arises from

the temperature increase

across the core from bottom to top. The coolant is typically less dense at the top than at the
bottom (causing the top to be more under-moderated than the bottom), and the properly
averaged cross sections are different at the top than the bottom (because they depend on
temperature). These phenomena cause smooth changes to the power profile, unlike partially
inserted control rods/blades.

For our heat-transfer analysis here, we assume that our neutronics team can provide us with the
power profile, q 0 (z), for any pin cell in the reactor, provided we can provide the temperature and
density distribution. So we take q 0 (z) to be given.

As the coolant flows upward from z to z + dz it collects energy at the rate q 0 (z)dz. We define:
mass
mch coolant mass flow rate in the channel, (11.72)
time
energy
cp coolant specific heat, (11.73)
mass-degree
Then conservation of energy requires:

mcp [Tw (z + dz) Tw (z)] = q 0 (z)dz (11.74)

367
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

It follows that
z
1
Tw (z) = Ti + dz 0 q 0 (z 0 ), (11.75)
mcp H/2

where

Ti coolant inlet temperature for the channel (11.76)

11.4 Summary: Global Temperature Distribution in Reactors

The preceding sections thoroughly describe the temperature distribution in a single pin cell in a
nuclear reactor with cylindrical pins and coolant flowing from top to bottom. The needed inputs
for the analysis are:

q 0 (z), the linear heat-generation rate as a function of height, for the given pin cell. This comes
from the global power shape, which comes from solving a transport or diffusion equation for
the entire core volume.

Ti , the coolant inlet temperature. This comes from knowledge of the coolant flow system and
conditions outside of the reactor.

m, the coolant mass flow rate in the given channel. This comes from knowledge of the coolant
flow system and conditions outside of the reactor.

Various material properties and heat-transfer coefficients.

If these inputs can be found, then our single-channel analysis can provide any needed temperatures.

Of course, the heat-generation rate depends on the scalar flux, which depends on cross sections,
which depends on temperature. So the neutronics, heat transfer, and fluid flow are

coupled.

In practice, the solutions are obtained iterativelya previous iterate or guess for the temperatures
is used to generate cross sections for a neutronics calculation, which feeds power densities into the
thermal-hydraulic analysis, which returns the next temperature iterate, etc.

368
NUEN 601, M.L. Adams notes, 2017 Chapter 11. Heat Transfer in Reactors

Appendices

369
NUEN 601, M.L. Adams notes, 2017 Chapter A. Direction Vectors and Solid Angle

Appendix A

Direction Vectors and Solid Angle

A.1 Introduction

In studies of particle transport we must describe functions of direction, including distributions


(densities) in direction-space. Here we develop the required vocabulary and machinery.

A.2 Angle in 2D: Radians

Let us begin with something familiar: measures of ordinary angles and direction vectors that live
in a plane. Recall the definition of the radian, a measure of angle, as depicted in Fig. A.1:

= angle subtended by arc of length `


`
. (A.1)
R

Figure A.1: Radiansmeasures of angle in the plane

Recall that the circumference of a circle is 2R, which implies that

the angle subtended by a full circle is 2 radians (2R/R = 2).

1
NUEN 601, M.L. Adams notes, 2017 Chapter A. Direction Vectors and Solid Angle

A.3 Solid Angle in 3D: Steradians

We turn to 3D space instead of 2D planes. Instead of an arc of a circle, we now consider a portion
of the surface of a sphere of radius R. In particular, let us consider an infinitesimal surface area,
dA, on the sphere. Instead of an angle in radians, we shall need a solid angle in steradians to
describe what is subtended by such an area. See Fig. A.2.

Figure A.2: Steradiansmeasures of solid angle in three dimensions. axes are x, y, z; show
, d, Rd, R sin , , d, R sin , dA

DEFINITION

d = infinitesimal increment of solid angle


= solid angle subtended by dA
dA
2 . (A.2)
R

2
NUEN 601, M.L. Adams notes, 2017 Chapter A. Direction Vectors and Solid Angle

Develop an expression for dA:

dA = (differential length 1) (differential length 2)


= (Rd) (R sin d)
= R2 sin dd (A.3)

Combine with the definition of the solid angle subtended by dA:

dA
d = sin dd . (A.4)
R2

Solid angle is a two-dimensional quantity. One way to see this is to recognize its relation to

surface area

on a sphere. Another way to see this is to note that it contains the product of two ordinary angle
increments, d and d.

Let us exercise our solid-angle machinery by deriving the formula for the entire surface area of a
sphere:
2  
2
Area = dA = R sin d d
area 0 0
 2   
2
=R d sin d
0 0

= R2 [2] [cos(0) cos()]

= 4R2 . (A.5)

We see that:

the solid angle subtended by a full sphere is 4 steradians.

It is often convenient to change variable, replacing the polar angle () with its cosine:

Define cos d = sin d . (A.6)

Then

=0 = 1 , and
= = 1 . (A.7)

3
NUEN 601, M.L. Adams notes, 2017 Chapter A. Direction Vectors and Solid Angle

With this change of variables, integrals over solid angle can be rewritten:
2 2 1
d d sin f (, ) = d d f (, ) (A.8)
0 0 0 1

Question: What fraction of the Earths surface is North of latitude 45 N, meaning more than 45
degrees north of the equator?

Answer The fraction is the solid angle subtended by the defined area, divided by 4:
2 1
1
fraction = d d
4 0 cos 45
1
= (2)(1 cos 45 )
4
!
1 2
= 1 14.6% (A.9)
2 2

A.4 Direction Vectors

We now consider direction vectors. Returning to the 2D-planar world, let us examine how we might
describe a direction. First, we define a unit vector in the direction of some vector ~v (which in our
problems could be a velocity, for example):
~ planar ~v / |~v | = ~v /v .

The convention in studies of particle transport is that direction vectors have unit lengththey are
unit vectors.
In the 2D planar world, specifying a single angle is sufficient to uniquely define a unit vector in the
plane. Suppose is the angle between the planar direction vector and the x axis. Then we can
completely express the direction vector in terms of :
~ planar = ex cos + ey sin .
(A.10)

Now we turn to the 3D world and define our unit direction vector exactly the same way:
~ ~v / |~v | = ~v /v .

This time, however, we must specify two angles, or two direction cosines, or one angle and one
cosine, to uniquely define the unit vector. Consider Figure A.3, which illustrates how a direction
vector is uniquely determined by a polar angle () and an azimuthal angle (), or how a direction
vector uniquely determines those two angles.

Think about latitude and longitude, which uniquely specify a point on the surface of a sphere (such
as Earth, approximately). Note that there is a one-to-one correspondence between a direction vector

4
NUEN 601, M.L. Adams notes, 2017 Chapter A. Direction Vectors and Solid Angle

Figure A.3: Polar angles (latitudes), azimuthal angles (longitudes), and direction vectors. show
arcs for angles, show u = definitions.

and a point on the surface of the spherea direction uniquely specifies a point on the surface, and
vice versa. Since the two angles latitude and longitude uniquely specify a point on the spheres
surface, they also uniquely determine a direction.

The math associated with these remarks is:


~ = x ex + y ey + z ez ,
(A.11)

where
~ ex = sin cos =
p
x 1 2 cos , (A.12)
~ ey = sin sin = 1 2 sin ,
p
y (A.13)
~ ez = cos .
z (A.14)

(Here we have define cos , as is customary.)

~ is a unit vector, as it should be.


It is easy to show that ~
~ is the square of the length of :
~

~
~ = sin2 cos2 + sin2 sin2 + cos2
= sin2 cos2 + sin2 + cos2


= sin2 (1) + cos2 = 1 . (A.15)

~ has length = 1.
Thus,

5
NUEN 601, M.L. Adams notes, 2017 Chapter B. Densities and Expected Values

Appendix B

Densities and Expected Values

B.1 Introduction

Our expressions for various gain and loss rates are written in terms of various densities that must
be integrated over various generalized (phase-space) volumes to describe the rates that we want.
This appendix is intended to clarify a few points about densities.

We are familiar with mass density expressed as mass per unit volume, for example in units of
g/cm3 . All densities have units of

something per unit volume,

at least if we are liberal in our interpretation of the term volume. The volume in question is
usually called the

phase-space volume.

B.2 Simple Examples

Figure B.1 is a schematic illustration of a string with knots tied along its length. If you were
asked about the knot density in the string, you would probably quickly assess that it is about 3
knots/cm. Would it bother you that the density is number per unit length instead of number
per unit (length3 )? (Probably knot.) Because the string is essentially a one-dimensional object,
the volume in our knot density is a length.

1
NUEN 601, M.L. Adams notes, 2017 Chapter B. Densities and Expected Values

0 1 2 3

Figure B.1: Knots in a string. (What kind of knots?)

This illustrates an important concept:

The volume in a density may not be a 3D spatial volume and thus may not have units of
[length]3 . It could be simpler (just length) or more complicated (with an arbitrarily large
number of dimensions, not limited to 3).

Let us continue to explore our knots-in-string example. If we plotted knot density as a function of
position along our string, and we were faithful to the details shown in Fig. B.1, we would obtain
the plot shown in Fig. B.2.

10

knots

cm

0 1 2 3

Figure B.2: Detailed knot density as a function of position along the string of Fig. B.1

This is a highly detailed plot of knot density. It shows where each individual knot is. You can see
that each knot is 1/9 cm long, so where there is a knot, the density is 9 knots / cm, but the density
is zero between knots.

We would not normally need such a detailed picture of a density. (Imagine the analogous picture of
mass density in a material, knowing that most of the volume of an atom is essentially empty space!)

2
NUEN 601, M.L. Adams notes, 2017 Chapter B. Densities and Expected Values

We usually want densities that are really averages over small volumes. For example, Fig. B.3 might
be a more useful picture of knot density in our string.

5
knots

cm

0 1 2 3

Figure B.3: Knot density as a function of position along the string of Fig. B.1, using averages over
small portions of the phase-space volume.

Figure B.3 indicates that the knot density is uniform, with a value of 3 knots/cm. It is obtained
by averaging the detailed plot over small intervals.

If interpreted literally, our second picture of knot density implies that the knots are a continuum.
We know this is not true, just as we know that matter is not a continuum. However, if we never
need to know exactly how many knots are in some tiny interval, the simpler picture is adequate.
For example, if we only cared how many knots were in one-kilometer intervals of string, our simpler
picture would give the same answer as the detailed and complicated one.

As a second example of a density, let us consider population density:

population density = people/mi2

In this example, the volume is spatial area. Again, we rarely need to know a detailed, microscopic
picture of population density. Such a picture would show a zero density almost everywhere, with a
spike at each persons current location, and it would change constantly.

A third example is number density of atoms:

atom number density = atoms/cm3

Here the volume is the spatial volume that we all know and love. Number density is much like
mass densityit just counts atoms instead of grams.

3
NUEN 601, M.L. Adams notes, 2017 Chapter B. Densities and Expected Values

B.3 Definition

DEFINITION

Every density is a limit. We define the density of things at some point ~x in some phase space
as follows:
 
expected number of things in V, where V contains ~x
d(~x) = lim (B.1)
V small V

where ~x is a point in phase space and V is a portion of phase-space volume that contains ~x.

B.4 Expected Values

Our general definition of a density uses the term expected number. This deserves discussion.
Imagine a small volume in a nuclear reactorsay a few cubic angstromsthat is operating in
steady state. At a given instant it would be unlikely that a neutron would be in this tiny volume.
However, if we observed this volume over a long time we would find that the average number of
neutrons in the volume is not zero. This average number would be the expected number of
neutrons in the volume. It would be less than 1.0, but this should not bother us.

Notice that steady state does not apply microscopically. In any phase-space volume we can
expect fluctuations about the expected value of the neutron population. The fluctuations become
insignificant relative to the expected value if the expected value is large, but become large relative
to the expected value if the expected value is small. Steady state refers to the macroscopic
behavior of the system (e.g., reactor). It is an ideal state that is never realized perfectly in practice.

Now consider the same volume in the same reactor but remove the steady-state restriction, so that
even macroscopically we allow things to be changing significantly in the reactor. Now imagine a
large number of macroscopically identical reactors, all macroscopically doing the same thing. Then
take an ensemble averagean average over all of the reactorsof the number of neutrons in the
tiny volume. This average is the expected number of neutrons in the volume.

When we write down our conservation equations, they will be in terms of densities that describe
expected values. A solution is perhaps best thought of as representing an average over a large
number of systems. If all we care about is quantities for which the expected value involves a
large number of neutrons or events (such as the power generated by an entire fuel rod in a reactor
operating at full power), then we can expect the fluctuations (either temporal or system-to-system)
to be small compared to the expected value, and the quantity for a particular reactor will be about
the same as the average over a large number of identical reactors.

4
NUEN 601, M.L. Adams notes, 2017 Chapter B. Densities and Expected Values

Let us define

hQi expected value of quantity Q (B.2)


hQi expected value of magnitude of fluctuation of quantity Q (B.3)

Then as a general rule, hQi/hQi will be small if hQi involves a large number of particles or events,
and it will be large if hQi involves a small number of particles or events. For example, if Q =
the total neutron population in a reactor operating at high power, then fluctuations will be small
compared to Q itself.

At the other extreme, if Q = the number of neutrons in a particular cubic angstrom (which has
a volume of 1030 m3 ) in the shield wall of a reactor building, then hQi  1 and the fluctuations
will be large compared to this expected value.

B.5 More Examples

Consider gas molecules bouncing about in a room. Suppose we do not care where in the room they
are, but we want to know about their velocities. This is described by a density in three-dimensional
velocity space:

molecular velocity density = molecules / (cm3 /s3 )

See Figure B.4.


Define the molecular velocity density function as follows:

D(vx , vy , vz )dvx dvy dvz number of molecules whose velocities are in the intervals
dvx containing vx , dvy containing vy , and dvz containing vz . (B.4)

In more compact notation:

D(~v )d3 v number of molecules whose velocities are in


d3 v containing ~v . (B.5)

To find how many molecules are going upward, we could integrate this density over all x velocities,
all y velocities, and all positive z velocities:


upward-going molecules = dvx dvy dvz D(~v ) . (B.6)
0

Note how the units work out:

cm cm cm molecules
molecules = . (B.7)
s s s cm3 /s3

5
NUEN 601, M.L. Adams notes, 2017 Chapter B. Densities and Expected Values

cm3
d3 v = dvx dvy dvz units =
s3

~v
vz

vx
vy

Figure B.4: A velocity vector (~v ) defines a point (with coordinates (vx , vy , vz )) in a 3D velocity
space. A differential element of phase-space volume is a velocity cube, dvx dvy dvz , which has
units of length3 /time3 .

We encounter densities in other volumes as well. For example, the angular neutron density is a
density in position and velocity. We often replace velocity by the equivalent variables energy and
direction. In this case:

neutron density = neutrons / (cm3 .MeV.ster) .

We define this energy- and direction-dependent neutron density function as follows:

n(x, y, z, , , E)dxdydz sin dddE number of neutrons whose position coordinates


are in the intervals dx containing x,
dy containing y, and dz containing z, whose
direction coordinates are in the intervals
d containing and d containing , and whose
energies are in the interval dE containing E. (B.8)

6
NUEN 601, M.L. Adams notes, 2017 Chapter B. Densities and Expected Values

In shorthand,
~ E)d3 rddE number of neutrons with positions
n(~r, ,
in d3 r containing ~r,
~ and
directions in d containing ,
energies in dE containing E. (B.9)

The phase-space volume here is a six-dimensional quantity: three dimensions in position, two in
direction, one in energy. Integrating this density over all energies and directions gives an ordinary
neutron densitythe number of neutrons per cm3 :
2
ntot (~r) = dE d sin ~ E) .
d n(~r, , (B.10)
0 0 0

Note how the units work out:

n n
3
= M eV.ster. 3 . (B.11)
cm cm .M eV.ster

B.6 Changes of Variables

Densities are functions of independent variables. The independent variables are coordinates that
define points in phase space. We can use different coordinate systems to describe the same phase
space, and often a problem can be greatly simplified by choosing a coordinate system that is
well matched to it. We therefore must understand the relationships among density functions that
describe the same physical quantity in different coordinate systems.

Changing coordinates (independent variables) in a density function is closely related to changing


variables in an integral, because the integral of a density over a phase-space subvolume is the
number of things in that subvolume, which is a physical quantity that must be independent of
the coordinate system that is used.

B.6.1 Single independent variable (one-dimensional phase spaces)

Consider the knot density function in a previous example. This density must have units of knots
per unit length, but any units can be chosen for the length variable. Let f (x) be the density
expressed in knots per inch, with x expressed in inches, and s(`) be the density in knots per cm,
with ` expressed in cm. Then we must have:

f (x)dx = s(`)d` . (B.12)

7
NUEN 601, M.L. Adams notes, 2017 Chapter B. Densities and Expected Values

This illustrates a fundamental truth in changes of variables for density functions: the number of
things must be the same no matter what variables are used to describe the density. That is,
we must begin with a statement of equality in which the terms express a number of things. This
means we will always have differential elements of phase-space volume, expressed in the different
variables, in our statement that relates two densities in different variables.

What is the relation between d and s? It is easy to find:



d`
f (x) = s(`) 2.54s(`) , (B.13)
dx

because there are 2.54 cm/inch. Of course, this case is simple enough that you could have
written down the relation from the beginning, but it illustrates a process that can be used in more
complicated cases. Remark: the absolute value operation is included because densities are defined
to be non-negative, whereas it is possible to choose coordinate systems such that the derivative in
question is negative.

Consider now a density in a speed variable, and call the speed-dependent density f (v). Suppose we
wish to use kinetic energy instead of speed, and we define the energy-dependent density to be g(E),
where E = mv 2 /2. Remember, these densities are describing the same physical quantity (which
could be neutrons in this case). The relation between the two densities is as before:


dE
f (v)dv = g(E)dE f (v) = g(E) = mvg(E) = 2Em g(E) . (B.14)
dv

Remark: We often commit a notational sin and use the same letter for a density in two different
independent variables, even though the density functions are obviously different. That is, we often
write

n(v) = 2Em n(E) . (B.15)

We do this even though the n function on the left is different from the n function on the right.
This has been going on for decades in studies of neutron transport, so I have no hope of reversing
it. If we are going to use the same symbol, n, for these two different functions, it would be less
confusing if we at least added descriptive subscripts, such as:

nspeed (v) = 2Em nenergy (E) . (B.16)

But neither we nor other authors will consistently do this. Therefore, one must always be careful
to figure out from context exactly what density function (i.e., what coordinate system) is being
used in any given set of equations.

Remark: Densities expressed in different coordinate systems have different units. In the exam-
ple above, the speed-dependent neutron density has units n/(cm3 .(cm/s)), whereas the energy-
dependent neutron density has units n/(cm3 .MeV).

8
NUEN 601, M.L. Adams notes, 2017 Chapter B. Densities and Expected Values

B.6.2 Multiple independent variables (multi-dimensional phase spaces)

As is the case in multi-dimensional integrals with changes of variables, the relation between densities
in multi-dimensional phase spaces involves a Jacobian. Consider a density function f (~x) in an
N -dimensional phase space, with coordinates x1 , x2 , . . . , xN , and another density function g(~y )
describing the same physical quantity in terms of coordinates y1 , y2 , . . . , yN . Then we have

f (~x)dx1 dx2 . . . dxN = g(~y )dy1 dy2 . . . dyN , (B.17)

and

f (~x) = g(~y ) |J| , (B.18)

where
y1 yN

...
x1 x1

J .. .. ..
. . .


y1 yN
...
xN xN

= the Jacobian of the transformation (B.19)

and |J| is the absolute value of the determinant of the Jacobian matrix.

Consider for example a mass density expressed in terms of Cartesian coordinates, xyz (x, y, z), and
the same density expressed in terms of polar spatial coordinates, polar (r, , ). Then we have

polar (r, , ) = xyz (x, y, z) |J| , (B.20)

where
x x x




r


y y y
|J| . (B.21)

r



z z z
r

9
NUEN 601, M.L. Adams notes, 2017 Chapter B. Densities and Expected Values

Observe that

x = r sin cos , (B.22)


y = r sin sin , (B.23)
z = r cos . (B.24)

Here we have defined the polar angle to be and the azimuthal angle to be . Now we turn the
crank:

sin cos r cos cos r sin sin


|J| sin sin r cos sin r sin cos = . . . = r2 sin

(B.25)


cos r sin 0

It follows that

polar (r, , ) = xyz (x, y, z)r2 sin = xyz (x, y, z) x2 + y 2 + z 2 sin .



(B.26)

B.7 Observations

We conclude our discussion of densities (distributions) with some summary observations:

It makes no sense to ask how many objects exist at a point in phase space. It does make
sense to ask what the density is at a point in phase space.

To get a number, integrate a density over a phase-space volume.

The volume is not always in x-y-z space.

We treat densities much as we would if the things they describe were continuous. (See the
two different plots of knot density in a string. We use the second view.) This does not mean
that we ignore the fact that particles exist as discrete packets. We are simply using expected
values in our definition of densities.

A density has units of things per unit phase-space volume. The units will differ if we choose
different coordinates to represent points in phase space. The transformation of a density from
one coordinate system to another involves the determinant of a Jacobian matrix.

10
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Appendix C

Selected Cross-Section Plots

In this appendix we provide plots of some cross sections, highlighting behavior that has a significant
impact on neutron distributions in energy, for several nuclides of interest to nuclear engineers.

1
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.1: 11 H (hydrogen) cross sections. Note several features: A) Except for E . 0.1 eV,
the total and elastic-scattering cross sections are almost the same. That is, except for very slow
neutrons, 1 H is essentially a pure elastic scatterer. B) The 1 H scattering cross section is remarkably
large, 20 barns. Compare this to other low-Z nuclides, such as 4 He, 12 C, and 16 O. C) 1 H is a
1/v absorber, which means its absorption cross section is proportional to 1/(neutron speed),
which means proportional to 1/ E. (Verify that every decrease in energy of 100 results in an
increase in absorption cross section of 10.) This is a common feature of light (low-A) nuclides.
D) The elastic-scattering cross section is almost constant for E < 10 keV. (E) The cross section
shown here is for 1 H in a free gas, that is, under the assumption that each hydrogen atom is not
bound to any other atom. For low-energy neutrons the cross section is significantly different if the
hydrogen is bound, as it almost always is in real materials (such as water or polyethylene).

2
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.2: H2 O (regular water) scattering cross section, illustrating the substantial effect of molec-
ular bonds on the scattering of low-energy neutrons. The curve that looks flat [marked as 2s (H)
+s (O)] is the sum of the cross sections for two unbound hydrogen nuclei plus one unbound oxy-
gen nucleus. The curve that increases as neutron energy decreases is the sum of the cross sections
for two hydrogen nuclei plus one oxygen nuclei when the atoms are bound in an H2 O molecule. The
difference is mainly in the scattering cross section for the bound hydrogen, not the oxygen. That is,
the scattering cross section for bound hydrogen is much higher than the cross section for free
hydrogen, for low-energy neutrons. Note that the strength of the bond that holds one hydrogen
atom to the oxygen atom in a water molecule is 4.4 eV, which is about where the bound and
free cross sections become noticeably different.

3
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Deuterium, ENDF/B-VII
101
(n,total) xsec
(n,elastic) xsec
(n,2n) xsec
(n,gamma) xsec
0
10
Cross Section (barns)

10-1

10-2

10-3 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9
10 10 10 10 10 10 10 10 10 10 10 10 10 10
Energy (eV)

Figure C.3: 21 H (deuterium, or hydrogen-2) cross sections. Note several features: A) Except for
E > 3 MeV and E < 105 eV, the total and elastic-scattering cross sections are almost the
same. That is, except for very slow and very fast neutrons, deuterium is essentially a pure elastic
scatterer. B) 2 H is a 1/v absorber, which means its absorption cross section is proportional to
1/(neutron speed), which means proportional to 1/ E. C) Fast neutrons can convert deuterium
to ordinary hydrogen via the (n,2n) reaction, although its cross section is not large.

4
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

He-3 total, elastic, (n, proton+triton); ENDF pointwise


106
(n,total) xsec
5 (n,elastic) xsec
10
(n,p) xsec
Cross Section (barns)

104

103

102

101

100

10-1 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
10 10 10 10 10 10 10 10 10 10 10 10 10
Energy (eV)

Figure C.4: 32 He (helium-3) cross sections. Note several features: A) Helium-3 is a strong neutron
absorber, with a  s for E < 105 eV. Interestingly, the absorption reaction is not capture,
but is (n,p), with reaction products of a proton and a triton. The combination of a high cross
section and charged-particle products make 3 He an attractive material for neutron detectors. B)
The scattering cross section is almost constant for all E below 50 keV. C) 3 He is a 1/v
absorber, which means its absorption cross section is proportional to 1/(neutron speed), which
means proportional to 1/ E.

5
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.5: 42 He (helium-4) cross sections. Note several features: A) For E (105 , 2 107 eV),
the total and elastic-scattering cross sections are identical. That is, helium-4 prefers to simply
scatter neutrons elastically. This makes it an excellent choice for a coolant if one wishes to build
a gas-cooled reactor. B)The cross section is constant for E < 105 eV. C) The cross section grows
significantly (factor of 10!) for E > 105 eV, because of a resonance just above 1 MeV.

6
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Li-6 total, elastic, inelastic, (n,alpha T), capture (ENDF/B


104
(n,total) xsec
(n,t) xsec
103 (n,elastic) xsec
(n,inelastic) xsec
(n,gamma) xsec
Cross Section (barns)

102

101

100

10-1

10-2 -3 -2 -1 0 1 2 3 4 5 6 7 8
10 10 10 10 10 10 10 10 10 10 10 10
Energy (eV)

Figure C.6: 63 Li (lithium-6) cross sections. Note several features: A) Lithium-6 is a strong neutron
absorber, with a  s for E < 104 eV. Interestingly, the dominant absorption reaction is not
capture, but 10 n + 63 Li 31 T + 42 He, often called the n, T reaction. This is the reaction that it
most often used to create tritium, which is a key ingredient for D+T fusion reactions and is therefore
needed for experiments that study fusion and also for some nuclear weapons. B) The scattering
cross section is almost constant for all E below 100 keV. C) 6 Li is a 1/v absorber, which
means its absorption cross section is proportional to 1/(neutron speed), which means proportional
to 1/ E, for energies below the resonance that is just over 100 keV. This is also true individually
for the capture reaction and the (n, T) reaction, which together make up the absorption cross
section in that energy range.

7
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Li-7 total, elastic, inelastic (ENDF/B VII); Li-7 (n,n alpha


101
(n,total) xsec
(n,elastic) xsec
(n,inelastic) xsec
Li7(n,na)T xsec
Cross Section (barns)

100

10-1

10-2 5 6 7
10 10 10
Energy (eV)

Figure C.7: 73 Li (lithium-7) cross sections in the high-energy range. Note: For neutron energies
above 5 MeV there is a cross section of a few tenths of a barn for an interesting reaction: 10 n + 73 Li
3 T + 4 He + 1 n, or (n, T n). Note that this reaction produces a tritium nucleus and a low-energy
1 2 0
neutron. If 6 Li is also present (which it usually is if 7 Li is present), this low-energy neutron could
produce a second tritium nucleus! Thus, there is a way to produce two tritium atoms from a single
neutron if that neutron had energy > 5 MeV. Without some possibility of generating more than
one tritium atom from one high-energy neutron, large-scale power production from D+T fusion
would be impractical, because there would not be enough tritium to keep the reactors running.

8
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.8: 94 Be (beryllium-9) cross sections. Note several features: A) Beryllium-9 is a metal and
is not a strong neutron absorberits capture cross section at 0.025 eV is less than 0.01 barnso it
is often used as a neutron reflector or as a structural material that doesnt eat neutrons. B) Be-9
can act as a neutron multiplier, for it has a non-negligible (n,2n) cross section for neutron energies
> 3 MeV.

9
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Boron-10
106

ENDF-VI (n,total) xsec


ENDF-VI (n,elastic) xsec
105

104
Cross Section (barns)

103

102

101

100
10-5 10-4 10-3 10-2 10-1 100 101 102 103 104 105 106 107
Energy (eV)

Figure C.9: 10
5 B (boron-10) cross sections. Note several features: A) Boron-10 is a strong neutron
absorber, with a  s for E < 105 eV. The absorption interaction is not capture, but rather
(n,), in which n + 10 B + lithium-7. The high cross section and relative abundance makes
boron a common choice for control-rod or soluble-poison material in reactors. The charged-particle
products make B-10 an attractive choice for some kinds of neutron detectors. B) The scattering
cross section is almost constant for all E below the resonance that peaks at 300 keV. C) 10 B is
a 1/v absorber, which meansits absorption cross section is proportional to 1/(neutron speed),
which means proportional to 1/ E. D Boron-10 is 20% of natural boron, with the other 80%
being boron-11.

10
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

C-12 total and elastic from JENDL-3.2 pointwise


7

6 (n,total) xsec
(n,elastic) xsec
Cross Section (barns)

0 0 1 2 3 4 5 6 7
10 10 10 10 10 10 10 10
Energy (eV)

Figure C.10: 12 C (carbon-12) cross sections. Note several features: A) For E < 5 MeV, the total
and elastic-scattering cross sections are almost the same. That is, carbon-12 prefers to simply
scatter neutrons elastically. B) The scattering cross section is almost constant for all E below
20 keV. C) For E > a few MeV, inelastic scattering is the main reaction that comes into play in
addition to elastic scattering. For E = 14 MeV the inelastic cross section is 0.43 barns.

11
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Oxygen-16
18

ENDF-VI (n,total) xsec


16 ENDF-VI (n,elastic) xsec

14

12
Cross Section (barns)

10

0
10-5 10-4 10-3 10-2 10-1 100 101 102 103 104 105 106 107
Energy (eV)

Figure C.11: 16 O (oxygen-16) cross sections. Note several features: A) For E < 5 MeV, the total
and elastic-scattering cross sections are almost the same. That is, oxygen-16 prefers to simply
scatter neutrons elastically. (There is 1/v absorption, but it is quite small for E > 105 eV.) B)
The scattering cross section is almost constant for all E below 20 keV. C) For E > a few MeV,
inelastic scattering is the main reaction that comes into play in addition to elastic scattering. For
E = 14 MeV the inelastic cross section is 0.5 barns.

12
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

total Fe(n,)? [ndf/endl2009.0/neutron]

Fe(n,n)Fe [ndf/endl2009.0/neutron]

Fe(n,2n)Fe [ndf/endl2009.0/neutron]

Fe(n,p)Mn [ndf/endl2009.0/neutron]
cross section (barn)

Fe(n,He)Cr [ndf/endl2009.0/neutron]

Fe(n,n')Fe [ndf/endl2009.0/neutron]

Fe(n,n')Fe [ndf/endl2009.0/neutron]

Fe(n,g)Fe [ndf/endl2009.0/neutron]

Fe(n,g)Fe [ndf/endl2009.0/neutron]

E (MeV)

Figure C.12: 26 Fe (natural isotopic mix) cross sections. Note several features: A) For E (10 eV,
1 MeV), the total and elastic-scattering cross sections are almost the same and are mostly in the
range of 3 to 11 barns. That is, in this energy range, iron is almost a pure scatterer. B) There is
1/v capture, with cross section of roughly 1 barn at 0.1 eV (and thus 10 barns at 0.001 eV, which
is off the scale to the left). C) There are lots of threshold reactions that come into play for neutron
energies above 1 MeV.

13
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Natural Cadmium
105

(n,total) xsec
10
4 (n,gamma) xsec
Cross Section (barns)

103

102

101

100
10-5 10-4 10-3 10-2 10-1 100 101 102 103 104 105 106 107
Energy (eV)

Figure C.13: Natural cadmium (48 Cd) cross sections. Cadmiums cross section jumps by a factor
of 1000 from E >1eV to E <1eV. This huge low-lying resonance makes Cd a non-1/v absorber
(although you can see that even this becomes 1/v at very low energies). This huge jump makes Cd
a very useful material for screening out thermal neutrons. Cd covers are often placed on foils for
this purpose in neutron-absorption experiments.

14
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

In-115 total and capture, JENDL-3.2-pointwise


105
(n,total) xsec
(n,gamma) xsec

104
Cross Section (barns)

103

102

1
10

100 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
10 10 10 10 10 10 10 10 10 10 10 10 10
Energy (eV)

Figure C.14: 115


49 In (indium-115) cross sections. Indium-115, which is 95.7% of natural indium, has
a big fat resonance at 1.45 eV, which makes it good for absorbing neutrons that are almost thermal.
If you cover an In foil with Cd, the In will not see thermal neutrons and will absorb mostly neutrons
of energies close to 1.45 eV. The resulting product, In-116, has a metastable state that -decays
with half-life 54 minutes, which makes it relatively straightforward to count it in a detector and
infer how many absorptions took place.

15
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Xe-135, JENDL-3.2-pointwise
108

107 (n,total) xsec


(n,gamma) xsec
6
10
Cross Section (barns)

105

104

103

102

1
10

100 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
10 10 10 10 10 10 10 10 10 10 10 10 10
Energy (eV)

Figure C.15: 13554 Xe (xenon-135) cross sections. The most striking feature of xenon-135 is how
enormous its cross section is for slow neutronsmore than a million barns! Xe-135 is a fission
product, and it is also the decay product of another fission product (Iodine-135). In fact, > 6%
of fissions ultimately produce this neutron-hungry nuclide. We must design our reactors to have
enough excess reactivity to stay critical even after Xe-135 builds up.

16
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.16: 235


92 U (uranium-235) cross sections. Note several features: A) The fission cross section
is significantly larger than the capture cross section, and it is most of the total for thermal neutrons.
Quantitatively, the 2200-m/s fission cross section ( E=0.0253 eV) is a bit below 600b and the
capture cross section is only 100b, so fission is by far the most likely interaction for a thermal
neutron with U-235. B) There is a low-lying resonance (at about 0.3 eV) that perturbs the 1/v
functional form for thermal neutrons. C) We see the typical behavior of a fissile nuclide: a 1/v
region at very low E, a resolved-resonance region from < 1eV to 20 keV, a fall-off of the capture
cross section for E > 1 MeV, inelastic scattering being significant in the energy range where
neutrons are born from fission, and (n,2n) reactions being possible for neutrons of several MeV.

17
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.17: 235 U (uranium-235) cross sections below 10 eV, showing the details of low-lying
92
resonances.

18
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.18: 235


92 U (uranium-235) cross sections from 3 eV to 30 keV, showing details of the resolved-
resonance range.

19
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.19: 235


92 U (uranium-235) cross sections on a linear-linear scale. This view is useful for
determining what happens to neutrons if they interact with U-235 after they are first born from
fission.

20
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.20: 235


92 U (uranium-235) inelastic scattering cross section (black) and the portions of the
inelastic scattering cross section that are due to the nucleus being left in the first six excited states
(the six colored curves), on a linear-linear scale. You can see that the threshold for exciting each
level is higher than that for exciting the previous level. For example, the fourth level has a threshold
of about 52 keV and the fifth level has a threshold of about 82 keV. You can see that the overall
inelastic scattering cross section is the sum of the cross sections for these six levels, until the 7th
level kicks in somewhere around 135 keV.

21
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.21: 235


92 U (uranium-235) inelastic scattering cross section (black) and the portions of the
inelastic scattering cross section that are due to the nucleus being left in the first six excited states
(the six colored curves), on a log-log scale. The log-log scale makes it easier to see the threshold
nature of the each interaction.

22
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.22: 238 U (uranium-238) cross sections. Compare against U-235. Note several features:
A) The U-238 cross section is much lower than that of a fissile nuclide for thermal neutrons, and
fission does not become important until E  1 MeV. B) There is a low-lying resonance (at 6.6 eV)
that strongly perturbs the 1/v functional form for thermal neutrons. C) This is typical behavior
for a fissionable (not fissile) nuclide: a 1/v region at very low E, a resolved-resonance region at
intermediate energies, a fall-off of the capture cross section for higher E, inelastic scattering being
significant in the energy range where neutrons are born from fission, and (n,2n) reactions being
possible for neutrons of several MeV.

23
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.23: 238


92 U (uranium-238) cross sections on a linear-linear scale. This view is useful for
determining what happens to neutrons if they interact with U-238 after they are first born from
fission.

24
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Figure C.24: 239 Pu (plutonium-239) cross sections. Looks a lot like U-235, because both are fissile.
Main differences: its low-lying resonance is stronger and its cross section is higher for thermal
neutrons. Its capture/fission ratio is also higher.

25
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

Pu fission and capture cross sections, 238 to 242


105
Pu238 fission
104 Pu238 capture
Pu239 fission
Pu239 capture
3
10 Pu240 fission
Cross Section (barns)

Pu240 capture
Pu241
102 Pu241 capture
Pu242 fission
Pu242 capture
101

100

10-1

-2
10

10-3 -3 -2 -1 0 1 2 3 4 5 6 7
10 10 10 10 10 10 10 10 10 10 10
Energy (eV)

Figure C.25: Cross sections for several plutonium isotopes. Pu-239 and 241 are fissile, and exhibit
typical fissile behavior. Pu-240 and 242 are fissionable, and exhibit typical fissionable behavior.
Pu-238 is somewhere in between. Note that if the fissile nuclides are present in a reactor, then
capture reactions will produce the fissionable nuclides. Note that all fission cross sections become
comparable for E > 1 MeV, and all capture cross sections fall off in that range. This is why
fast-spectrum reactors can be designed to consume more actinides than they create.

26
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

ENDF-VI H-1a
(n,elastic) ang. dist.

0 .6

0 .4
Prob/Cos

10 6
0 .2

10 5
y
rg
ne
1.

E
0

10 4
0. Co
5 s
0. ne
0
-0
i

.5

3
-1

10
.0

Figure C.26: 1 H (hydrogen-1) probability density function for scattering angle in the center-of-mass
frame. The functions plotted are P (Ei , 0C ) = s (Ei , 0C )/s (Ei ), as a function of 0C for several
values of Ei . A flat plot with value 0.5 defines isotropic scattering in the center-of-mass
frame, which is called s-wave scattering. You can see that H-1 is a perfect s-wave scatterer.

27
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

ENDFB-VII-neutron H-2
(n,elastic)

2 .0

1 .5
Prob/Cos

10 7
1 .0

10 6
0 .5

10 5

y
rg
ne
1.

E
0
0. Co

10 4
5 si
0. ne
0
-0
.5

3
-1

10
.0

Figure C.27: 2 H (deuterium) probability density function for scattering angle in the center-of-mass
frame. The functions plotted are P (Ei , 0C ) = s (Ei , 0C )/s (Ei ), as a function of 0C for several
values of Ei . You can see that deuterium begins to deviate from s-wave scattering for neutron
energies 100 keV.

28
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

ENDF-neutron C-nata
(n,elastic)

3
Prob/Cos

10 7
2

10 6

y
rg
ne
1.

E
0
0. Co
5 si
0. ne
0
-0
.5

5
-1

10
.0

Figure C.28: 12 C (carbon-12) probability density function for scattering angle in the center-of-mass
frame. The functions plotted are P (Ei , 0C ) = s (Ei , 0C )/s (Ei ), as a function of 0C for several
values of Ei . You can see that carbon begins to deviate from s-wave scattering for neutron energies
> 2 MeV. You can also see that the pdf becomes complicated at high energies and that at some
high energies it is forward-peaked (probability of small-angle scattering is greater) while at other
energies it is backward-peaked (probability of backscattering is greater).

29
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

ENDF-VI O-16
(n,elastic) ang. dist.

1 .5

1 .0
Prob/Cos

10 6
0 .5

10 5
y
rg
ne
1.

E
0

10 4
0. Co
5 s
0. ne
0
-0
i

.5

3
-1

10
.0

Figure C.29: 16 O (oxygen-16) probability density function for scattering angle in the center-of-mass
frame. The functions plotted are P (Ei , 0C ) = s (Ei , 0C )/s (Ei ), as a function of 0C for several
values of Ei . You can see that O-16 begins to deviate from s-wave scattering for neutron energies
200 keV. You can also see that the pdf becomes complicated and that at some high energies
it is forward-peaked (probability of small-angle scattering is greater) while at other energies it is
backward-peaked (probability of backscattering is greater).

30
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

ENDF-VI U-238b
(n,elastic) ang. dist.

2
Prob/Cos

10 6
1

10 5
y
rg
ne
1.

E
0

10 4
0. Co
5 s
0. ne
0
-0
i

.5

3
-1

10
.0

Figure C.30: 238 U (uranium-238) probability density function for scattering angle in the center-of-
mass frame. The functions plotted are P (Ei , 0C ) = s (Ei , 0C )/s (Ei ), as a function of 0C for
several values of Ei . You can see that U-238 begins to deviate from s-wave scattering for neutron
energies 50 keV. You can also see that the pdf forward-peaked (probability of small-angle
scattering is greater) at high energies.

31
NUEN 601, M.L. Adams notes, 2017 Chapter C. Selected Cross-Section Plots

ENDF-VI U-238b
(n,elastic) ang. dist.

30

20
Prob/Cos

10 7
10

10 6
y
rg
ne
1.

E
0

10 5
0. Co
5 s
0. ne
0
-0
i

.5

4
-1

10
.0

Figure C.31: 238 U (uranium-238) probability density function for scattering angle in the center-
of-mass frame. The functions plotted are P (Ei , 0C ) = s (Ei , 0C )/s (Ei ), as a function of 0C
for several values of Ei . Note the vertical scalescattering becomes extremely forward-peaked in
U-238 for neutron energies above 1 MeV!

32
NUEN 601, M.L. Adams notes, 2017 Chapter D. Tensors

Appendix D

Tensors

In this appendix we describe and hopefully de-mystify tensors. We begin by noting the following:
it is possible to write all of our equations without using vectors or tensors. We choose to use vectors
and tensors because this dramatically simplifies our mathematical expressions. For example, we
write
~ ~u
(D.1)

instead of
ux uy uz
+ + . (D.2)
x y z
The two expressions are equivalent, but the first expression is more compact, and moreover it is
written the same way for Cartesian, cylindrical, or spherical coordinate systems.

Similarly, we choose to write entire equations with each term a vector instead of writing out a
scalar equation for each component. For example, we may write the vector equation
~ + ~g = ~q
f (D.3)

instead of the three equations


f
+ gx = qx , (D.4)
x
f
+ gy = qy , (D.5)
y
f
+ gz = qz . (D.6)
z
Again, the vector expression is equivalent to the set of three scalar expressions, but the vector
expression is more compact and is written the same way for all coordinate systems.

1
NUEN 601, M.L. Adams notes, 2017 Chapter D. Tensors

D.1 Introduction and Notation

You can think of the concept of a tensor as an extension of the concept of a vector. A vector is a
collection of components, with one component for each dimension in some space. A rank-2 tensor
is also a collection of components, with one component for each pair of dimensions in some space.
A rank-3 tensor has one component for each triplet of dimensions, and so on. A vector in our usual
position space has three components, one each for x, y, and z. We can write such a vector, ~u for
example, as the sum of the products of components times unit vectors:

~u = ux ex + uy ey + uz ez . (D.7)

Another way to write a vector is as a column that contains its components in a pre-agreed order,
such as x then y then z:

ux
~u = uy (D.8)
uz

A rank-2 tensor in our usual position space has nine components, one each for xx, xy, xz, yx,
yy, yz, zx, zy, and zz. We can write a rank-two tensor as a sum of the products of components
times pairs of unit vectors:

~
~t = txx ex ex + txy ex ey + txz ex ez
+ tyx ey ex + tyy ey ey + tyz ey ez
+ tzx ez ex + tzy ez ey + tzz ez ez (D.9)

Another way to write a tensor is as a matrix that contains its components in a pre-agreed order,
such as xx then xy then xz in the first row, etc.:

txx txy txz
~
~t = tyx tyy tyz (D.10)
tzx tzy tzz

2
NUEN 601, M.L. Adams notes, 2017 Chapter D. Tensors

D.2 Operations With Vectors and Tensors

D.2.1 Tensor-vector dot products

If we dot a vector with a tensor we obtain a vector. The order matters in general. That is, in
general ~~t ~u 6= ~u ~
~t. Let us first examine a 3-vector dotted into 3 3 tensor from the left, and let
us spell out all 27 terms (3 3 3 = 27) in gory detail:

~u ~~t = [ux ex + uy ey + uz ez ] [txx ex ex + txy ex ey + txz ex ez


+ tyx ey ex + tyy ey ey + tyz ey ez
+tzx ez ex + tzy ez ey + tzz ez ez ]

= ux txx ex ex ex + ux txy ex ex ey + ux txz ex ex ez


| {z } | {z } | {z }
1 1 1
+ ux tyx ex ey ex + ux tyy ex ey ey + ux tyz ex ey ez
| {z } | {z } | {z }
0 0 0
+ ux tzx ex ez ex + ux tzy ex ez ey + ux tzz ex ez ez
| {z } | {z } | {z }
0 0 0
+ uy txx ey ex ex + uy txy ey ex ey + uy txz ey ex ez
| {z } | {z } | {z }
0 0 0
+ uy tyx ey ey ex + uy tyy ey ey ey + uy tyz ey ey ez
| {z } | {z } | {z }
1 1 1
+ uy tzx ey ez ex + uy tzy ey ez ey + uy tzz ey ex ez
| {z } | {z } | {z }
0 0 0
+ uz txx ez ex ex + ux txy ez ex ey + ux txz ez ex ez
| {z } | {z } | {z }
0 0 0
+ uz tyx ez ey ex + ux tyy ez ey ey + ux tyz ez ey ez
| {z } | {z } | {z }
0 0 0
+ uz tzx ez ez ex + ux tzy ez ez ey + ux tzz ez ez ez
| {z } | {z } | {z }
1 1 1

= (ux txx + uy tyx + uz tzx ) ex + (ux txy + uy tyy + uz tzy ) ey + (ux txz + uy tyz + uz tzz ) ez (D.11)

3
NUEN 601, M.L. Adams notes, 2017 Chapter D. Tensors

We could express the same math in a more compact summation notation:


" #

~u ~
X X X XXX
~t = ui ei tjk ej ek = ui tjk ei ej ek
| {z }
i j k i j k
ij
!
X X
= ui tik ek (D.12)
k i

Here each index i, j, and k takes on the values x, y, and z.

If we use matrix-vector notation and follow the rules of matrix-vector multiplication we can obtain
the same result. Since we have the vector on the left, we must transpose it into a row vector, and
the result will also be a row vector:

txx txy txz
~u ~~t = ux uy uz tyx tyy tyz
 

tzx tzy tzz


 
= (ux txx + uy tyx + uz tzx ) (ux txy + uy tyy + uz tzy ) (ux txz + uy tyz + uz tzz ) (D.13)

We can transpose into the more standard column format if we wish:



(ux txx + uy tyx + uz tzx )
(ux txy + uy tyy + uz tzy )
(ux txz + uy tyz + uz tzz )

4
NUEN 601, M.L. Adams notes, 2017 Chapter D. Tensors

Now we examine a 3-vector dotted into 3 3 tensor from the right, and we spell out all 27 terms
(3 3 3 = 27) in gory detail:

~~t ~u = [t e e + t e e + t e e
xx x x xy x y xz x z
+ tyx ey ex + tyy ey ey + tyz ey ez
+tzx ez ex + tzy ez ey + tzz ez ez ] [ux ex + uy ey + uz ez ]

= txx ux ex ex ex +txy ux ex ey ex +txz ux ex ez ex


| {z } | {z } | {z }
1 0 0

+ tyx ux ey ex ex +tyy ux ey ey ex +tyz ux ey ez ex


| {z } | {z } | {z }
1 0 0

+ tzx ux ez ex ex +tzy ux ez ey ex +tzz ux ez ez ex


| {z } | {z } | {z }
1 0 0

= txx uy ex ex ey +txy uy ex ey ey +txz uy ex ez ey


| {z } | {z } | {z }
0 1 0
+ tyx uy ey ex ey +tyy uy ey ey ey +tyz uy ey ez ey
| {z } | {z } | {z }
0 1 0
+ tzx uy ez ex ey +tzy uy ez ey ey +tzz uy ez ez ey
| {z } | {z } | {z }
0 1 0
= txx uz ex ex ez +txy uy ex ey ez +txz uy ex ez ez
| {z } | {z } | {z }
0 0 1

+ tyx uz ey ex ez +tyy uy ey ey ez +tyz uy ey ez ez


| {z } | {z } | {z }
0 0 1

+ tzx uz ez ex ez +tzy uy ez ey ez +tzz uy ez ez ez


| {z } | {z } | {z }
0 0 1

= (txx ux + txy uy + txz uz ) ex + (tyx ux + tyy uy + tyz uz ) ey + (tzx ux + tzy uy + tzz uz ) ez (D.14)

Again, we could express the same math in a more compact summation notation:
" #
~
~t ~u =
X X
tjk ej ek
X
ui ei =
XXX
tjk ui ej ek ei
| {z }
j k i j k i ki
!
X X
= tji ui ej (D.15)
j i

And again, if we use matrix-vector notation and follow the rules of matrix-vector multiplication we
can obtain the same result:

txx txy txz ux (txx ux + txy uy + txz uz )
~
~t ~u = tyx tyy tyz uy = (tyx ux + tyy uy + tyz uz ) (D.16)
tzx tzy tzz uz (tzx ux + tzy uy + tzz uz )

5
NUEN 601, M.L. Adams notes, 2017 Chapter D. Tensors

Examination of the results for ~u ~


~t and ~~t ~u shows that in general they are different, but they are
equal if the tensor is symmetric. That is,

IF: txy = tyx , txz = tzx , and tyz = tzy ,

THEN: ~u ~~t = ~~t ~u (D.17)

D.2.2 Vector-tensor-vector dot products

It is common to encounter the following type of expression:

~u ~~t ~v

The associative property holds, so one can perform the dot operations in either order:

~t ~v = ~u ~~t ~v = ~u ~~t ~v
~u ~
   
(D.18)

Note that the result in either set of parentheses is a vector, which is then dotted into another vector.
The final quantity is a scalar.

D.3 Identity Tensor

The identity tensor is given by Iij = ij :

~
I~ = (1)ex ex + (0)ex ey + (0)ex ez
+ (0)ey ex + (1)ey ey + (0)ey ez
+ (0)ez ex + (0)ez ey + (1)ez ez

= ex ex + ey ey + ez ez


1 0 0
= 0 1 0 (D.19)
0 0 1
It is easy to verify that any vector dotted into the identity tensor from either side equals that
vector:
~ ~
~u I~ = I~ ~u = ~u (D.20)
It follows from this that
~
~u I~ ~v = ~u ~v . (D.21)

Note that this quantity is a scalar.

6
NUEN 601, M.L. Adams notes, 2017 Chapter D. Tensors

D.4 Higher Rank Tensors

In transport theory it is sometimes convenient to use tensors of rank higher than two. For example,
we may encounter the third-rank tensor:

d .
4

One way to write this, as we can deduce from expressions for rank-two tensors, is with summation
notation:
X X X  
d = d i j k ei ej ek (D.22)
4 i=x,y,z j=x,y,z k=x,y,z 4

The ijk-th component of this rank-three tensor is the direction integral in parentheses. It is easy
to show that in this example, every component of the tensor equals zero. We could write this in
the compact notation:
~
d = ~~0 .
4
This makes it obvious that no matter what this is dotted into, the result will be zero.

7
NUEN 601, 2017 HOMEWORK 1, DUE FRIDAY, FEB. 6

(1) Read and study:


(a) Chapter 1 of the Lewis book (Fundamentals of Nuclear Reactor Physics).
(b) Chapter 2 of the Lewis book.
(c) Appendix A of the course notes.
(d) Appendix B of the course notes.

(2) Read Dr. Charltons lecture notes, Charlton 01.pdf, found on the course web site under Study
Material. Then answer the following.
(a) [5 points] List five kinds of radiation.
(b) [10 points] The energy required to separate an electron from a neutral 11 H atom in its
ground state is between 13 and 14 eV. If you wanted to calculate the electron-proton binding
energy in this atom to an accuracy of 3 digits, by looking at electron mass + proton mass
11 H mass, how many digits of accuracy would you need in the mass of the proton and the
1 H atom? Explain your reasoning.
1

(c) [10 points] Show that for small v/c, the relativistic expression
for kinetic energy reduces
to the classical expression. [Hint: Note that for small x, 1 x 1 x/2 x2 /8 . . ..
Note that for small y, 1/(1 y) 1 + y + y 2 + . . .]

(3) Consider a 235 U nucleus in its ground state, and suppose it absorbs a neutron that has very little
kinetic energy. This forms a compound nucleus: 236 U in an excited state.
(a) [10 points] How excited is the 236 U compound nucleus? That is, how much binding energy
is added when the neutron gets stuck in the nucleus? Express your answer in MeV. [Hint:
take the difference between (neutron mass + 235 U mass) and (236 U mass).]
(b) [10 points] Suppose the compound nucleus fissions. Describe a reasonably likely result
(choose a number of free neutrons that come out and choose the A and Z of each of two
fission fragments). Dont choose the same thing that I chose in the notes or that any of
your classmates choose, as far as you know. For the scenario youve chosen, calculate how
much energy is released promptly by the absorption-plus-scission event.
(c) [5 points] What fraction of the original mass (the mass of one neutron plus one 235 U atom)
is converted to energy in your postulated absorption-plus-scission event?

(4) Consider a commercial nuclear reactor that generates 3.2 GW of recoverable thermal energy. (This
is approximately what it takes to generate 1 GW of electricity.) Suppose that each fission corre-
sponds to 185 MeV of this recoverable energy. How many fissions take place in a year of operation
at 3.2 GW thermal?

(5) Consider a nuclear weapon that produces 15 kilotons of yield, and assume that the yield comes from
the fission of 239 Pu. Assume that each fission produces approximately 200 MeV of yield. What
mass of Pu actually fissions when this device detonates? [1 kiloton 4.2 TJ.] How long must a
reactor operate at 3.2 GW-thermal in order to produce recoverable energy = this devices yield?
1
2 NUEN 601, 2017 HOMEWORK 1, DUE FRIDAY, FEB. 6

(6) Consider a velocity-dependent density function that describes a Maxwellian distribution:


" #
 m 3/2 m vx2 + vy2 + vz2
nveloc (~v ) = n0 exp
2kT 2kT

(a) [20 points] Derive the corresponding speed- and direction-dependent density function,
nspeed-dir (v, , ), where v is the speed, is the polar angle, and is the azimuthal an-
gle. That is, switch to polar coordinates for velocity. (This involves a Jacobian. Show all
steps, including taking the determinant. See Appendix B of your course notes.) Note that
vx = v sin cos ,
vy = v sin sin ,
vz = v cos .

(b) [10 points] Derive the corresponding energy- and direction-dependent density function,
nenergy-dir (E, , ), where E = kinetic energy = mv 2 /2, and and are as before.

(c) [10 points] Show, by explicitly evaluating all of the integrals, that
2
dvx dvy dvz nveloc (~v ) = dv d dnspeed-dir (v, , )
0 0 0

Hint: It may help to recognize that your integrals involve the normal or Gaussian
probability density function (pdf), and the integral of a pdf, over all possible values of the
independent variable, equals 1. Or you can look up the integrals if you like.

(7) [15 points] Recall the energy- and direction-dependent Maxwellian distribution you derived above
(nenergy-dir (E, , )). Integrate this function over all directions to obtain the energy-dependent
Maxwellian distribution, nenergy (E). (This is easy.) What is the corresponding energy-dependent
scalar flux, (E)? That is, what is (E) if the neutrons are in a Maxwellian distribution? At what
energy does its peak occur?

(8) [10 points] Consider a particle of mass m and velocity ~v , and a second particle of mass M and
velocity V~ . Find the relationship between the total kinetic energy of the two particles in their
center-of-mass reference frame and the relative speed between them. Use the classical expression
for kinetic energy, massspeed2 /2. You should find that Ec = total kinetic energy in c-o-m frame
~ |. What expression do you obtain for ?
= vr2 /2, where is the reduced mass and vr = |~v V
NUEN 601, 2017 HOMEWORK 1, DUE FRIDAY, FEB. 6 3

(9) Consider a sample that contains N0 radioactive atoms at time t = 0, with decay constant .
Suppose there are no mechanisms for producing these atoms and decay is the only loss mechanism.
(a) [10 points] What is the expected number of the original atoms remaining at time t?
(b) [5 points] What is the expected time it will take for half of the population to decay? (This
is the half-life, T1/2 .) Show the math that leads to your answerdont just write it down.
(c) [5 points] What is the probability that at time t, no atoms have decayed? [Think about
the probability of decay for each atom, and go from there.]
(d) [5 points] What is the probability that at time t, all N0 of the atoms have decayed?
(e) [5 points] What is the probability that at time t, exactly one of the atoms has decayed?
[Hint: Find the probability that the first atom decays and the others do not, then add the
probability that the second atom decays and all the others do not, . . . ]

(10) The radius of the Earth is approximately 4000 miles.


(a) [10 points] What is the area between the tropic of Cancer and the tropic of Capricorn?
Assume that these two lines of latitude are 23.5 from the Equator. Compare this
tropical area to the areas inside the Arctic and Antarctic circles, assuming each circle is
23.5 from its pole. [See Appendix A of notes.]
(b) [10 points] The span of latitudes, in degrees, is the same for the tropical and polar areas.
However, if we measure the span of latitudes in terms of the cosines of the polar angles
instead of the angles themselves, it looks different. Divide your Arctic+Antarctic area by
your tropics area. Then divide the Arctic+Antarctic cosine span, (cos 0 cos 23.5 ) +
(cos 156.5 cos 180 ), by the tropical cosine span. How do the ratios compare?
4 NUEN 601, 2017 HOMEWORK 1, DUE FRIDAY, FEB. 6

(11) 135 Xe
and 135 I play important roles in the steady-state and transient behavior of nuclear reactors.
The most important production and loss mechanisms are:
135 I production: beta decay of 135 Te (which has a 19-second half-life)
135 I loss: beta decay to 135 Xe (6.6 hr half-life)

135 Xe production: beta decay of 135 I

135 Xe loss: 1) beta decay (9.1-hr half-life) and 2) absorption of a neutron


Important: 135 Te is a fission fragment produced by 6.3% of fissions. It decays so quickly to 135 I

that for our purposes we can assume that 135 I is produced directly by 6.3% of fissions.
(a) [10 points] Write down conservation equations for the populations of 135 I and 135 Xe, taking
into account the production and loss mechanisms listed above. (Remember: you can skip
135 Te and pretend that 135 I is produced directly from fission.) Define:

135
NI (t) average I number density in the reactor at time t
135
NX (t) average Xe number density in the reactor at time t

Assume that at time 0 there are no 135 I or 135 Xe atoms in the reactor. Use the following
for the average fission rate density and the absorption rate density in 135 Xe:
fissions
= hihf i
cm3 .s
absorptions in 135 Xe
= hihaXe iNX (t)
cm3 .s
(b) [10 points] As t the 135 I and 135 Xe number densities approach asymptotic (equilib-
rium) values. What are the expressions for these values? [Hint: in this limit, the rate of
change is zero, so your differential equations become easily-solved algebraic equations.]
(c) [15 points] Solve your conservation equations for the average number densities of 135 I and
135 Xe as a function of time, assuming that neither nuclide is present at t = 0.

(d) [15 points] Plot your solutions from t = 0 to t = 80 hours. Use the following values:
n
hi = 1013
cm2 .s
hf i = 0.5 cm1
haXe i = 3 106 b
Do your solutions approach the asymptotic values given by your previous answer? Do you
think they should approach these values? Explain your reasoning.
(e) [10 points] Suppose that the 135 I and 135 Xe concentrations have attained their asymptotic
values, and then at time t0 the reactor is quickly shut down so that hi quickly goes to
zero. Solve for and plot the 135 I and 135 Xe concentrations (using the flux and cross-section
values given above) for the 40-hour period following the shutdown.
(f) [10 points] Discuss what happens to the 135 Xe concentration after shutdown. How does the
concentration compare to the equilibrium concentration that is attained during operation?
Give a physical explanation of why the concentration changes the way it does. Given that
the 135 Xe cross section for absorbing neutrons is very large, do you think that the post-
shutdown behavior of 135 Xe could cause operational problemsfor example if there were a
desire to restart a few hours after a shutdown?
NUEN 601, 2016 HOMEWORK 2, DUE FRIDAY, FEB. 10

(1) [15 points] Consider the expression we derived for a microscopic cross section that takes into
account the motion of nuclei. Equation (4.43) of the course notes gives this expression for the
total cross section. The expression for other reactions (fission, capture, scattering, etc.) looks the
same except for the subscript t on the symbols. Now consider the limit of very low neutron
speeds, such that essentially all of the nuclei are moving much faster than the neutrons. That is,
consider the limit in which |V ~ | >> |~v | |~v V
~ | |V
~ |. Show that in this limit, for any given
nucleus velocity distribution (N(V ~ )), the averaged r for reactions of type r is proportional to
1/v, where v = |~v | = the lab-frame neutron speed. Does this result depend on any particular form
of r,cold (vrel )?
[Goals: Understand that our averaged is really an average of [vrel cold (vrel )] that is then divided
by lab-frame speed v. Understand that this gives rise to 1/v behavior for sufficiently low neutron
energies even if the cold cross section is constant, given nuclei that are in motion.]

(2) [15 points] Discuss whether or not it is possible to have a nuclear fission reactor with multiplication
factor k > 3. The key to this is recognizing values of , the number of fission neutrons emitted per
absorption by a fuel nucleus, for fuel nuclides that exist. You have seen plots of this for key fuel
nuclides. In your answer I am looking for understanding, not precise quantification of the maximum
possible value of k.
[Goal: Build intuition about the connection between basic physical quantities and reactor multi-
plication factor. Build intuition about reasonable range of values of k.]

(3) [20 points] The gradient operator is defined by

~ (x, y, z) = ex f + ey f + ez f
f
x y z
Show that for any differentiable function g(x, y, z),

~ = 4 g , for u = x, y, and z .
du ( g)
4 3 u
The expression on the left involves several integrals over solid angle. Each such integral is a double
integral over polar and azimuthal angles. I want you to show the details of these integrations and
see if you obtain the right-hand side.
[Goal: Become more familiar with gradient operators, direction vectors, and solid-angle integrals.]
1
2 NUEN 601, 2016 HOMEWORK 2, DUE FRIDAY, FEB. 10

(4) The angular flux, , is in general a function of seven independent variables: 3 in position (such as
x, y, and z), one energy variable (E), one time variable (t), and two direction variables (such as
polar cosine, , and azimuthal angle, ). In some problems of practical interest the angular flux is
not a strong function of directionit is almost isotropicand its dependence on direction can be
well approximated as being linear in the direction components (also called the direction cosines).
The mathematical statement of this is:
linearly anisotropic
(~r, E, , t) a(~r, E, t) + x bx (~r, E, t) + y by (~r, E, t) + z bz (~r, E, t)
= a(~r, E, t) + ~b(~r, E, t)

(a) [10 points] Assume that is linearly anisotropic with the form shown above. Perform the
integral over all directions to obtain the scalar flux:

(~r, E, t) = d(~r, E, , t)

That is, express (~r, E, t) in terms of the a and b functions.

(b) [10 points] Assume that is linearly anisotropic with the form shown above. Multiply
by and itegrate over all directions to obtain the net current density, which is a vector
function:
~
J(~r, E, t) = d(~r, E, , t)
~ r, E, t) in terms of the a and b functions.
That is, express J(~

(c) [10 points] Assume that is linearly anisotropic with the form shown above. Multiply by
ez and itegrate over all positive values of ez to obtain the partial current density
in the +z direction, which is a scalar function:

+
Jz (~r, E, t) = d ez (~r, E, , t)

:ez >0

That is, express Jz+ (~r, E, t) in terms of the a and b functions.


[Goals: Gain familiarity with integrals over direction, especially with linear and quadratic functions
of the direction cosines. Gain familiarity with the connections among angular flux, net current
density, and partial current density.]
NUEN 601, 2016 HOMEWORK 3, DUE FRIDAY, FEB. 17

(1) Read and study:


(a) Chapter 2 of the Lewis book (especially the section on resonances).
(b) Appendix C of the course notes.

(2) Consider the Breit-Wigner formula for the cross sections for radiative capture and for elastic scat-
tering, (E) and e (E), near a resonance. See Eqs. (2.41) and (2.42) of the Lewis text, for example.
Note that the variable E in Lewiss equations is the total kinetic energy in the center of mass
frame. For the effective radius of the nucleus, use the expression in Lewiss problem 1.8 at the
end of his Chapter 1. Note that the reduced wavelength (0 in Lewiss equations) is the neutron
wavelength divided by 2, and the neutron wavelength is h/pCM , where h = Plancks constant and
pCM is the neutron momentum in the center of mass frame.
(a) [15 points] Assume: nucleus is at rest in the lab frame, nucleus mass is A neutron mass
(mn ), A = 115, and neutron speed is low enough that non-relativistic expressions can be
used for all energy and momentum values
1
kinetic energy = (mass)(speed)2 ,
2
momentum = (mass)(speed).
Under these assumptions, for a neutron whose lab-frame kinetic energy is Elab , derive
expressions for:
vCM , the lab-frame speed of the center of mass,
E = total kinetic energy (neutron plus nucleus) in the center-of-mass frame,
0 , the reduced wavelength of the neutron in the center-of-mass frame.
(b) [30 points] Assume that t (Elab ) = (Elab ) + s (Elab ). Plot t (Elab ) using the following
parameter values:
0 = 3 104 b
= 0.0799 eV E0 = 1.45 eV
n = 0.0001 eV A = 115

(c) [10 points] Provide your plot on a log-log scale, with lab-frame neutron kinetic energy
ranging from 105 eV to 106 eV.
(d) [10 points] Discuss similarities and differences between your t which is a theoretical
expression for a cross section due potential scattering and one single resonanceand the
actual t for 115 In, which is in Appendix C of your notes and which includes contributions
from all of its resonances plus potential scattering plus any other reactions that might
be possible. How closely does your cross section match the real one in the immediate
neighborhood of the big resonance at 1.45 eV? How closely do the 1/v tails match? How
do the cross sections compare for E 10 to 2000 eV? Above 10,000 eV?
[Goals: Learn the basics of the Breit-Wigner formula. Understand that it can provide an accurate
description of cross sections. Understand that it contains terms proportional to E 1/2 , or 1/v.]
1
2 NUEN 601, 2016 HOMEWORK 3, DUE FRIDAY, FEB. 17

(3) The angular flux, , is in general a function of seven independent variables: 3 in position (such as
x, y, and z), one energy variable (E), one time variable (t), and two direction variables (such as
polar cosine, , and azimuthal angle, ). In some problems of practical interest the angular flux is
not a strong function of directionit is almost isotropicand its dependence on direction can be
well approximated as being linear in the direction components (also called the direction cosines):
linearly anisotropic
(~r, E, , t) a(~r, E, t) + x bx (~r, E, t) + y by (~r, E, t) + z bz (~r, E, t)
= a(~r, E, t) + ~b(~r, E, t)

(a) [10 points] Assume that is linearly anisotropic with the form shown above. Use the
results you obtained in a previous homework to express in terms of and J~ instead of
a and ~b. (This is easydont make it difficultbut it is important: it shows you how to
construct the angular flux from the scalar flux and net current density, when the angular
flux is linearly anisotropic.)
(b) [10 points] Assume that is linearly anisotropic with the form shown above. Use the
results you obtained in a previous homework to express Jz+ in terms of and J~ instead of
a and ~b.
[Goals: Gain familiarity with integrals over direction, especially with linear and quadratic functions
of the direction cosines. Gain familiarity with the connections among angular flux, scalar flux, net
current density, and partial current densities.]

(4) In this problem you will use conservation arguments to construct a particle-transport model. Then
you will use your model to calculate buildup factors. The setting is simplified compared to the real
world: we will pretend that all particles have the same kinetic energy and that they can move only
right or left along the x axis. Given this setting, we can describe the solution in terms of two beam
intensities, one going to the right and one going to the left. We shall call them I + (x) and I (x).
Consider a slab of material that ranges from x = 0 to w, where w = the slab width, with a
and s known and constant throughout the material. Suppose a beam of intensity I0+ is incident
perpendicularly on the x = 0 face, and nothing enters the slab on the x = w face. Suppose there
are no particle-emitting sources inside the slab.
Suppose further that when a particle scatters, it has an equal chance of emerging as a right-going
or left-going particle. This corresponds to isotropic scattering in our simplified world of particles
going only right or left.

(a) [20 points] Derive the two first-order ordinary differential equations that describe conservation of
right-going particles (one equation) and left-going particles (the other equation).
The derivation for I + (x) should closely follow the derivation of uncollided beam intensity in Section
4.14.1 of your notes. The difference is that in your derivation, the term intensity added between
x1 and x2 is not zero, but rather is half of (x2 x1 ) the scattering rate density in that little
interval (because half of the scattered particles go to the right). The scattering rate density is the
scalar flux times the scattering cross section. The scalar flux is the sum of the two beam intensities:
(x) = I + (x) + I (x).
The derivation for I (x) is much like that for I + (x), but be careful with the sign of the derivative
term.
You should get an equation for I + that contains I as part of its scattering-source term and an
equation for I that contains I + as part of its scattering-source term.
NUEN 601, 2016 HOMEWORK 3, DUE FRIDAY, FEB. 17 3

(b) [20 points] Solve your system of equations, which is a set of coupled first-order ODEs with constant
coefficients. Hints:
Guess a solution of the form:
I + (x) = A+ eBx
I (x) = A eBx
where A are constants.
Insert the guess into your two equations. What values of B allow the equations to be
satisfied by nonzero solutions? You should be able to write your two equations as
  +   
a b A 0
=
c d A 0
Then a non-zero solution demands that the determinant of the 22 matrix be 0. The matrix
elements (a, b, c, d) contain B, t , and s . What values of B give a 0 determinant?
You should find two values of B (lets call them B1 and B2 ) that allow non-zero solutions.
(One is the negative of the other.) This means the general solution for each intensity is a
linear combination of two exponentials:
I + (x) = A+
1e
B1 x
+ A+
2e
B2 x

I (x) = A
1e
B1 x
+ A
2e
B2 x

For each value of B, the matrix equation forces a certain ratio of the A+ and A coefficients.
For example, let a1 and b1 be the matrix coefficients corresponding to the value B1 , and let A
1
and A+ be the associated A values. Then the matrix equation requires that a A+ = b A , or
1 1 1 1 1
A +
1 = R1 A1 , where R1 = a1 /b1 = known. A similar result holds for the other value of B (B2 ).
This leaves us with
I + (x) = A+
1e
B1 x
+ A+
2e
B2 x

I (x) = R1 A+
1e
B1 x
+ R2 A+
2e
B2 x

where R1 and R2 are known.


(c) [10 points] Apply the boundary conditions, I + (0) = I0 and I (w) = 0, to determine the remaining
unknown coefficients, A+ +
1 and A2 .
+
(d) [20 points] Define the buildup factor f (x) (x)/Iuncollided (x). Plot this as a function of x for
the following cases:
w = 1 mean free path = 1/t = 0.1 scattering mfp = 0.1/s
w = 1 mean free path = 1/t = 0.99 scattering mfp = 0.99/s
w = 10 mean free paths = 10/t = 0.1 scattering mfp = 0.1/s
w = 10 mean free path = 10/t = 1 scattering mfp = 1/s
w = 10 mean free path = 10/t = 9.9 scattering mfp = 9.9/s
NUEN 601, 2016 HOMEWORK 4, DUE FRIDAY, FEB. 24

(1) Consider elastic scattering of a neutron off of a stationary nucleus of mass M = A m, where m
is the neutron mass. Define 0C cos C , where C is the scattering angle in the center-of-mass
frame. Define 0L cos L , where L is the scattering angle in the lab frame. The following
graphics illustrate the relationships among the key variables.

Before and after pictures in lab and center-of-mass frames

Post-scatter picture in lab frame with center-of-mass information superimposed

The following equations come from conservation of momentum and energy and from the geometry
of the figures shown above, with C cos c and L cos L .
0 A
vC = vC = vL
A+1
1
vCM = vL
A+1
vL0 sin L = vC
0
sin C
vL0 L = vCM + vC
0
C
1
2 NUEN 601, 2016 HOMEWORK 4, DUE FRIDAY, FEB. 24

(a) [20 points] Derive an expression for PL (L ) as a function of PC (C ), where PL and PC are
probability density functions for scattering through the associated angles in the associated
reference frames. Recall the rule that PL (L )dL = PC (C )dC . Notes:
If you divide the last two equations cited above you obtain a relation involving sines
and cosines of L and C . (Use the first two equations to eliminate everything else
except A.) If you square the relation you can express everything in terms of cosines,
and in fact you should be able to solve for 2L in terms of A and c .
Do this, then take the square root to obtain L = a function of A and C .
Now you can find dL /dC , which you will need to relate PL to PC .
(b) [10 points] Plot PL (L ) for 1 H. Note that for 1 H scattering is isotropic in the center-of-
mass frame. This means you know PC (C ). For this exercise, pretend that A exactly equals
1. In this very special A = 1 case, L can never be > 90 degrees, which means L cannot
be < 0. [Hint: Pick a set of C values that range from 1 to +1. For each of these, figure
out L and PL (L ), and plot the point.]
(c) [10 points] Plot PL (L ) for 16 O, assuming an incident neutron energy low enough that
scattering is isotropic in the center-of-mass frame. You may pretend that A = 16. (Its real
value is 15.85.)
(d) [10 points] Plot PL (L ) for 238 U, assuming an incident neutron energy low enough that
scattering is isotropic in the center-of-mass frame. You may pretend that A = 238.
(e) [15 points] For each of the three cases you plotted, find the average lab-frame scattering
cosine, hL i. This is a weighted average of L with weight function PL . Use your three
calculated values to plot Ahl i > as a function of A for the case of s-wave elastic scattering
(elastic scattering that is isotropic in the center-of-mass frame). That is, plot your three
values of Ahl i >.
(f) [10 points] What do you conclude from your analyses and plots about how the nucleus
mass affects the relation between the lab-frame and center-of-mass scattering angles and
their associated probabilities? What is your hypothesize for a simple relation between hL i
and A for the case of s-wave elastic scattering?
[Goals: Become familiar with, and hopefully unafraid of, the mathematics and physics that relate
scattering angles in the lab and center-of-mass frames. Develop intuition about differential scatter-
ing cross sections in direction, in the lab frame (where we need them). Understand that isotropic
scattering in the COM frame is forward-peaked in the lab frame, to an extent that varies with A,
and that this can be described by a simple relation.]

(2) (a) [20 points] Given H2 O at 1 g/cm3 , what is H s (Ei Ef ), the differential scattering cross
section for the hydrogen, for Ei = 1 keV? Same question for O s (Ei Ef )? How about for
H O
the mixturewhat is s (Ei Ef )? The complete answer will contain the variables Ei
2

and Ef but everything else should be numbers with units, because you are given the water
density and you can get the microscopic cross sections you need from the plots you were
given.
(b) [10 points] Consider a large pool of water and imagine that 106 neutrons, each with energy
1 keV, are introduced in the center, far from boundaries. How many neutrons have energies
between 700 eV and 850 eV after one scattering event?
[Goals: Become familiar with differential scattering cross sections. Learn what real ones look like
for the most widely used moderating material in nuclear reactors (water). Review how to calculate
number densities. Solidify understanding of macroscopic cross sections for mixtures.]
NUEN 601, 2016 HOMEWORK 4, DUE FRIDAY, FEB. 24 3

(3) Consider elastic scattering of a 400-keV neutron off of 16 O. Let us approximate the differential
scattering cross section in the center-of-mass scattering-angle cosine as follows:
3
1 + 2C

PC (C )
8

(a) [5 points] Show that PC is properly normalizedthat when you integrate it over the
appropriate range you get the answer that you should.
(b) [10 points] Plot PC (C ).
(c) [15 points] Given Ei = 400 keV and elastic scattering off of 16 O with PC (C ) = 38 1 + 2C ,


plot P (Ei Ef ) as a function of Ef . (See section on differential scattering cross sections


in Chapter 4 of course notes.)
(d) [10 points] Discuss any similarities you see in your plots of PC (C ) and P (Ei Ef ). You
should see a striking similarity. Be quantitative in your discussion about this.
[Goals: Understand the relation between center-of-mass scattering cosine and lab-frame post-scatter
neutron energy, given elastic scattering.]
NUEN 601, 2016 HOMEWORK 5, DUE FRIDAY, MAR. 10

(1) One of your classmates has looked ahead and figured out an approximate solution for the angu-
lar flux in a critical steady-state rectangular parallelepiped reactor that has a uniform material
composition and that is surrounded by vacuum. The reactor size and shape are given by:
a a b b c c
x , y , z
2 2 2 2 2 2
Your classmate figures that the following is a reasonable approximation:
1 n  x   y   z 
(~r, E, , t) = 0 f (E) cos cos cos
4 a b c
1 h  x   y   z i
~
cos cos cos ,
t (E) a b c

where 0 is a known constant and f (E) is a known function. Also known are the functions
a (E), (E), f (E), t (E), and s (Ei Ef , i f ), all properly averaged over nucleus motion.
In what follows, perform all integrals for which there is enough information.
(a) [10 points] What is (~r, E, t)?
(b) [10 points] What is (~r, t)?
~ r, E, t)?
(c) [10 points] What is J(~
~ r, t)?
(d) [10 points] What is J(~
(e) [10 points] At what rate (n/s) are neutrons being absorbed in the reactor?
(f) [5 points] At what rate (n/s) are neutrons causing fission in the reactor?
(g) [5 points] At what rate (n/s) are neutrons being produced by fission in the reactor?
(h) [10 points] At what rate (n/s) are neutrons of energies > 100 keV scattering in the reactor?
(i) [10 points] Consider neutrons of energies > 100 keV that suffer a scattering collision.
What fraction emerge with energies < 40 keV?
(j) [10 points] What is Jx (~r, E, t)?
(k) [10 points] At what rate (n/s) are neutrons crossing from the top half of the reactor (z > 0)
to the bottom half (z < 0)? (One-way, not net rate.)
(l) [10 points] At what net rate (n/s) are neutrons crossing the plane at y = b/4 from
y < b/4 to y > b/4?
[Goals: Become familiar with, and hopefully unafraid of, expressing physical quantities in mathe-
matical terms. Become familiar with, and hopefully speedy at, the integrations that we frequently
encounter. Become more familiar with the relations among angular flux, scalar flux, net current
density, partial current density, reaction rates, one-way leakage rates, and net leakage rates.]
1
2 NUEN 601, 2016 HOMEWORK 5, DUE FRIDAY, MAR. 10

(2) Consider the following hypothetical reactor:


nothing (including ) depends on x or y (one-dimensional slab problem in z)
the angular flux does not vary with , the azimuthal angle around the z axis
the slab material is homogeneous (macroscopic cross sections do not depend on position),
and macroscopic cross sections are given that have been averaged over nucleus motion
slab goes from z = H/2 to z = +H/2
define z

(a) [15 points] Write down the neutron k-eigenvalue transport equation for this reactor. Spell
out all derivatives and integralsdo not use the or d symbols. In the inscattering term,
begin with the usual differential scattering cross section, s (Ei E, i ), and show how
to go from this to s (Ei E, i ) for this case in which the angular flux does not
depend on . That is, define s (Ei E, i ) in terms of s (Ei E, i ) and use
it in your k-eigenvalue equation.
(b) [10 points] Suppose the slab is surrounded by vacuum, which means that no neutrons can
enter the problem through its boundaries. Write down the mathematical expression for
this vacuum boundary condition. Your transport equation and your boundary conditions
should constitute a well-posed math problem.
(c) [5 points] A solution of an eigenvalue problem is an {eigenvalue, eigenfunction} pair. How
many solutions exist to the problem you have specified?
(d) [5 points] How many of these solutions have eigenfunctions that do not change sign, so that
a constant times the eigenfunction can be positive throughout the entire spatial, directional,
and energy domain?
(e) [10 points] Write down the k-eigenvalue P1 equations for this reactor. Spell out all deriva-
tives and integralsdo not use the or d symbols. Note: only one of the first-moment
equations is non-zero.
(f) [10 points] Write down the k-eigenvalue diffusion equation for this reactor. Spell out all
derivatives and integralsdo not use the or d symbols.
(g) [10 points] Write down appropriate boundary conditions for your diffusion equation.
[Goals: Solidify understanding of transport equations, P1 equations, diffusion equations, and bound-
ary conditions. Gain practice and understanding in the translation of physical descriptions into
mathematical equations. Reinforce understanding of eigenvalue problems.]

(3) [15 points] Show that any half-range integral of the tensor equals 2/3 times the identity
tensor. That is, show the following:

2 ~~
d = I
:en >0 3
There are nine terms in the tensor . You need to integrate each term over the indicated
directional half range. To make this easier you can choose any convenient coordinate system
for your integrations. For example, you might want to define the z axis of your coordinate system
to point in the direction of en . You may wish to consult the course handout on tensors.
[Goals: Become more familiar with tensors, especially as they are used in particle transport studies.
Gain experience with half-range integrals in the direction domain.]
NUEN 601, 2016 HOMEWORK 5, DUE FRIDAY, MAR. 10 3

(4) Consider a beam of neutrons, all moving with the same velocity in the lab frame. Suppose the
beam is incident on a slab of material occupying the following spatial volume: x (0, 8 cm),
y (, +), and z (, +). The beam is of infinite extent in y and z. The beam has the
following characteristics:
neutron density in beam = 3 104 n/cm3
neutron lab-frame speed = 5 105 cm/s
beam direction = = 53 ex + 45 ey

(a) [5 points] At what rate per unit y-z area (n/[cm2 -s]) are neutrons entering the slab? How
is this rate related to the partial current density Jx+ (x)x=0 ?
(b) [10 points] If the slab material only captures neutrons (no scattering or fission or other
neutron-releasing reactions), with cross section 0.5 cm1 , then at what rate per unit y-z
area (n/[cm2 -s]) are neutrons exiting the slab? [Hint: think about attenuation, and think
carefully about the path length through the material.] How is this related to the partial
current density Jx+ (x)x=8 cm ?
(c) [10 points] What is the capture rate density, n/[cm3 -s], as a function of x in the slab?
(d) [10 points] Integrate the capture rate density over the volume x (0, 8 cm), y y, and
z z to obtain the capture rate (n/s) in that volume. Then divide by yz to obtain
the capture rate per unit y-z area.
(e) [10 points] You have now computed entering, exiting, and capture rates per unit y-z area.
Do your answers form a conservation statement that makes sense? That is, does you gain
rate = loss rate in this steady-state problem?
[Goals: Solidify understanding of crossing rates, attenuation, reaction rate densities, reaction rates,
and conservation. Practice calculating all of these for a specific problem.]

(5) Consider a source-free homogeneous slab, infinite in y and z, and assume the one-group diffusion
approximation. Suppose Jinc = 103 n/(cm2 -s) enter the left face of the slab at x = 0, and suppose
there is a vacuum on the right at x = a = 12 cm. Define:
f
k ,
a
r a f = a (1 k ),
D = 0.2 cm,
a = 0.05 cm1 .
(a) [10 points] Write down the one-group diffusion equation for this problem.
(b) [10 points] Write down appropriate boundary conditions.
(c) [10 points] Solve the problem for the case k = 0.9.
(d) [10 points] Solve the problem for the case k = 1.
(e) [10 points] Solve the problem for the cases k = 1.05 and k = 1.1.
(f) [15 points] Plot your four solutions on the same plot. Discuss how the solution shape
changes as a function of k . Include concavity in your discussion and relate it to the sign
of the second derivative. Would you expect to see maxima or minima inside the problem
domain for any of these cases? What would you expect as k 1?
[Goals: Solidify understanding of diffusion solutions, and learn how the character of the solution of
steady-state problems changes depending on the sign of k 1.]
NUEN 601, 2017 HOMEWORK 6, DUE FRIDAY, APR. 7

(0) (a) Read Chapter 6 of the Lewis book.


(b) Read Sections 1-5 of Chapter 7 of the Lewis book.

(1) [15 points] Solve problem 6.1 from the Lewis book. (Use one-group diffusion theory for this.
Lewiss s00 is our Jincident . Assume that everything is infinite and uniform in y and z and that no
neutrons enter the slab through the right surface.)

(2) We have found that according to one-group diffusion (1GD) theory, the fundamental-mode scalar
flux in a bare homogeneous brick-shaped reactor has the following shape:
(x, y, z) = 0 cos(x/a) cos(y/b) cos(z/c)
where a, b, c are the extrapolated dimensions, 0 is a constant, and (0,0,0) is the reactors center.
(a) [15 points] What is the pointwise peaking factor in this reactor? [The pointwise peaking
factor is defined to be the maximum power density divided by the average power density.
Power density is proportional to scalar flux. You are permitted, and even encouraged, to
include the entire extrapolated reactor dimensions in the integrals that you will need to
perform. This adds very little error and makes the math simpler.]
(b) [15 points] If a = b = c = 1 meter, the one-group fission cross section is 0.8 cm1 , and the
bare homogeneous cubical reactor is steadily undergoing 1020 fissions/second without any
fixed sources present, then what is the scalar flux at the center of the reactor?
(c) [15 points] Consider again the bare, source-free, homogeneous cubical reactor from the
previous question, still with 1020 fissions/s taking place in steady state. If the one-group
total cross section is 3 cm1 , then at what rate (neutrons per second) are neutrons leaking
out of the bottom surface?
[Goals: Develop understanding of the fundamental mode in a 3D reactor, peaking factors, the
connection between scalar flux and net current density in diffusion theory, and net leakage rates.]

(3) Consider a commercial LWR. Quantify the time scales on which the following phenomena are
capable of causing a significant change in reactivity:
(a) [5 points] production and depletion of fuel nuclides
(b) [5 points] production and depletion of fission-product nuclides
(c) [5 points] fuel-temperature change in response to a rapid power increase
(d) [5 points] fuel-temperature change in response to a power decrease (Hint: heat has to
escape for temperature to drop)
(e) [5 points] moderator-temperature change in response to a power change
(f) [5 points] control-rod motion
[Goals: Reinforce understanding of time scales relevant to reactor transients.]
1
2 NUEN 601, 2017 HOMEWORK 6, DUE FRIDAY, APR. 7

(4) Consider a simplified version of enriched uranium consisting of only 235 U and 238 U, with W
N235 /(N235 + N238 ). Assume uranium metal of density 19 g/cm3 .
In this problem you will use one-group diffusion theory to address several interesting questions
about configurations of this uranium metal surrounded by vacuum. You will need cross sections that
are weighted averages over energy, with a weighting function that should be a good approximation
to the energy-dependent scalar flux in your problems. As a crude but reasonable approximation, we
shall assume that the energy-dependent scalar flux is proportional to (E), the fission spectrum.
This is somewhat reasonable because neutrons are born with a fission spectrum and most neutrons
in a pure-uranium system will not lose very much energy before they leak or get absorbed.
The following microscopic cross sections have been averaged over energy using (E) as the weight
function:
235 U 238 U

t 7.70 b 7.78 b
f 1.24 b 0.31 b
0.09 b 0.07 b
e 4.57 b 4.80 b
in 1.80 b 2.60 b
2.50 2.50

(a) [20 points] Use your one-group cross sections to compute k for the cases of W = 0.9,
0.6, 0.4, and 0.2.
(b) [30 points] Estimate (i.e., compute using one-group diffusion theory) the critical mass of
the uranium mixture with W = 0.9. Critical mass is the uranium mass needed to make a
bare sphere critical.
(c) [20 points] Repeat this for W = 0.6, 0.4, and 0.2.
(d) [20 points] Plot: (a) critical mass, and (b) mass of 235 U that makes a bare sphere critical,
as functions of W , on a single graph. Start with the four W values you were given, and
then calculate points at additional values of W as needed to get the shapes of the curves
about right. Comment on how quickly the required amount of 235 U grows as the enrichment
decreases.
(e) [20 points] Return to the case of W = 0.9. If the critical mass were formed into a cube,
what would be the value of k? [Hint: it is less than 1.] If you had two such subcritical
cubes, each containing the mass that would be critical if in a spherical shape, and put them
together face to face, what would k be for the resulting rectangular parallelepiped?
(f) [10 points] Consider W = 0.2, and again consider two cubes, with each formed from one
critical mass. What is k for one such cube, and what is it when you put two such subcritical
cubes together face to face? When you compare your answer for the W = 0.9 and 0.2 cases,
what do you conclude about the roles of enrichment and size in forming systems with high
multiplication factors? (You must answer this question if you want full credit for this
problem.)
(g) [20 points] By what factor does the critical radius of a bare sphere change if the density
of the material is doubled? By what factor does the critical mass change in this case?
[Goals: Learn to estimate physical quantities using one-group diffusion theory. Explore how critical
mass and multiplication factor vary with size, shape, enrichment, and density.]
NUEN 601, 2017 HOMEWORK 7, DUE MONDAY, APR. 17

(1) Consider a simplified version of enriched uranium consisting of only 235 U and 238 U, with W
N235 /(N235 + N238 ). Assume uranium metal of density 19 g/cm3 .
In this problem you will use two-group diffusion theory to address several interesting questions
about configurations of this uranium metal surrounded by vacuum.
Think about the physics of such a system. Neutrons are born from fission, with energy distribution
(E), which means > 99% are born with energies > 100 keV. Here is what (E) looks like:
0.40

0.35

0.30
Fission Spectrum (inverse MeV)

0.25

0.20

0.15

0.10

0.05

0.00
0 1 2 3 4 5 6 7 8 9 10
Energy (MeV)

Almost every neutron will have one of the following life histories:
1. leak or get absorbed before scattering, or
2. scatter elastically, perhaps many times, before leaking or getting absorbed (still with ap-
proximately the energy it had at birth, given that uranium nuclei are so massive), or
3. scatter inelastically at some point in its history before ultimately leaking or getting ab-
sorbed.
Neutrons in the first two categories maintain approximately a fission-spectrum energy distribution,
so their interactions can be well characterized by fission-spectrum-averaged cross sections. However,
if a neutron scatters inelastically it emerges in a different (significantly lower) energy range, for
which the cross sections are different and in particular the 238 U fission cross section is almost zero.
You will account for this by using two energy groups in your diffusion model. The fast group will
include neutrons that have never scattered inelastically. It will cover the range E & 100 keV. The
slow group will have neutrons that have inelastically scattering. The cross section for scattering
from fast the slow is just the fast-groups inelastic scattering cross section.
1
2 NUEN 601, 2017 HOMEWORK 7, DUE MONDAY, APR. 17

You will need cross sections that are weighted averages over the energy range corresponding to
each of your groups, with a weighting function that should be a good approximation to the energy-
dependent scalar flux in the given energy range. As a crude but reasonable approximation, we shall
assume that the energy-dependent scalar flux in the fast group (E > 100 keV) is proportional
to (E), the fission spectrum. This is somewhat reasonable because neutrons are born with a
fission spectrum and most neutrons in a pure-uranium system will not lose very much energy before
they leak or get absorbed.
The following microscopic cross sections have been averaged over energy using (E) as the weight
function:
Fast-group cross sections
235 U 238 U

t 7.70 b 7.78 b
f 1.24 b 0.31 b
0.09 b 0.07 b
e 4.57 b 4.80 b
in 1.80 b 2.60 b
2.50 2.50
The following cross sections are crude but reasonable averages for the (slower) neutrons that have
inelastically scattered:
Slow-group cross sections
235 U 238 U

t 16 b 11 b
f 3 b 0 b
1 b 1 b
e 12 b 10 b
in 0 b 0 b
2.50 2.50

Your two-group diffusion equations are:


1
D1 2 1 (~r) + [a1 + s,12 ] 1 (~r) = [1 f 1 1 (~r) + 2 f 2 2 (~r)] ,
k
D2 2 2 (~r) + a2 2 (~r) = s,12 1 (~r) ,
where
s,12 in,1 in our model for this problem, and
1 (~r) = 2 (~r) = 0 for ~r on the extrapolated boundary.

For this problem, use the same extrapolated boundary for both energy groups. I suggest that you
use 2D1 for extrapolation distance. Given the same extrapolated boundary for each group, then
each group flux will be proportional to the same spatial function (eigenfunction of the Laplacian):
1 (~r) = f1 (~r),
2 (~r) = f2 (~r), where
(~r) = Bg2 (~r) and
2

(~r) = 0 for ~r on the extrapolated boundary

You know what the and Bg2 are for reactors with commonly considered shapes, such as the spheres
and bricks you will consider here.
NUEN 601, 2017 HOMEWORK 7, DUE MONDAY, APR. 17 3

From these equations it follows that your f1 , f2 , and k satisfy:

1
D1 Bg2 + a1 + s,12 f1 = [1 f 1 f1 + 2 f 2 f2 ] ,
 
k
D2 Bg2 + a2 f2 = s,12 f1 ,
 

These two equations yield an expression for k in terms of cross sections and Bg2 . (Solve the second
equation for f2 in terms of f1 , substitute into the first equation, and solve for k.)

Using your two-group diffusion model, perform the following calculations.


(a) [20 points] Compute k for the cases of W = 0.9, 0.6, 0.4, and 0.2.
(b) [20 points] Compute the critical mass of the uranium mixture with W = 0.9. Critical
mass is the uranium mass needed to make a bare sphere critical. How does this compare
to your previous one-group answer?
(c) [20 points] Repeat this for W = 0.6, 0.4, and 0.2. How do your answers compare to your
previous one-group answers?
(d) [20 points] Plot:
(1) critical mass of the uranium mixture, and
(2) mass of 235 U contained in that mixture,
as functions of W , on a single graph. Start with the four W values you were given, and then
calculate points at additional values of W as needed to get the shapes of the curves about
right. Comment on how quickly the required amount of 235 U changes as the enrichment
decreases.
(e) [20 points] Return to the case of W = 0.9. If the critical mass were formed into a cube,
what would be the value of k? [Hint: it is less than 1.] If you had two such subcritical
cubes, each containing the mass that would be critical if in a spherical shape, and put them
together face to face, what would k be for the resulting rectangular parallelepiped?
(f) [10 points] Consider W = 0.2, and again consider two cubes, with each formed from one
critical mass. What is k for one such cube, and what is it when you put two such subcritical
cubes together face to face? When you compare your answer for the W = 0.9 and 0.2 cases,
what do you conclude about the roles of enrichment and size in forming systems with high
multiplication factors? (You must answer this question if you want full credit for this
problem.)
[Goals: Learn to estimate physical quantities using two-group diffusion theory. Learn how inelastic
scattering dramatically changes the behavior of fast systems that contain 238 U. Explore how
critical mass and multiplication factor vary with size, shape, enrichment, and density, using a
model that is more realistic than one-group diffusion theory.]
NUEN 601 STUDY PROBLEMS, EXAM 1, 2017

Material: Notes through Section 7.7.3 + Homeworks 1-5

(1) You are given that the angular flux, (~r, E, , t), is linearly anisotropic. Express in terms
~ r, E, t).
of the scalar flux, (~r, E, t), and the net current density, J(~

(2) What is the physical meaning of:


(a) (~r, E, t) integrated over all energy and over a given spatial volume?
~ r, E, t) integrated over all energy and over a given spatial volume?
(b) J(~
(c) (~r, E, t) integrated over all energy and over a given surface area?
~ r, E, t) integrated over all energy and over a given surface area, where en (~r)
(d) en (~r) J(~
is the outward unit normal at position ~r on the surface?
(e) The macroscopic absorption cross section for a given material?

(3) Explain why there are resonances in cross sectionswhy the cross section for nuclide A ZX
can be hundreds of times as large at certain neutron energies than it is at slightly larger or
smaller energies.

(4) Consider the multiplication factor, k, of a reactor.


(a) Provide two common definitions of this factor.
(b) Explain why these definitions are not always correct, and state the situation under
which they are correct.

(5) Consider the following sequence of reactions:


1 238 239
0n +92 U 92 U +
239 239
92 U 93 Np + e + +
239 239
93 Np 94 Pu + e + +

(a) How much energy is released by this sequence? (That is, what is the total Q value for
the three consecutive reactions?) Potentially helpful numbers are below. Note that
the masses for the U, Np, and Pu are for neutral atoms, not just nuclei. Note that the
mass of a neutrino is negligibly small relative to the mass of an electron. State your
assumptions and approximations.
neutron: 1.008664915 amu
proton: 1.007276467 amu
electron: 0.0005485799 amu
238 U: 238.0507826 amu
239 U: 239.0542878 amu
239 Np: 239.0529314 amu
239 Pu: 239.0521565 amu
c2 1 amu = 931.49 MeV.
1
2 NUEN 601 STUDY PROBLEMS, EXAM 1, 2017

(b) Write down differential equations for NU9 (~r, t), the number density of 239 U in a reactor
at time t; NNp (~r, t), the number density of 239 Np in a reactor at time t; and NPu (~r, t),
the number density of 239 Pu in a reactor at time t; under the following assumptions:
(~r, E) = given and independent of time
all microscopic cross sections are given (including capture and absorption cross
sections for each nuclide), pre-averaged over nucleus velocities, and independent
of time
238 U number density, NU8 (~r), is given and independent of time
239 U is produced only by neutron capture in 238 U, 239 Np is produced only by
decay of 239 U, and 239 Pu is produced only by decay of 239 Np (assume that other
production rates are negligible)
239 U, 239 Np, and 239 Pu can be lost either by decay or by absorbing a neutron.

(6) Each car in a coal train holds approximately 100 metric tons of coal. A typical coal train
contains 100 of these cars, is one mile long, and supplies a large coal-fired power plant for
one day at full power, which is roughly 3GW thermal. A 3-GWth nuclear power plants
reactor contains roughly 100 metric tons of uranium. If 5% of this uranium fissions over
the lifetime of the fuel, and each fission produces 190 MeV of recoverable energy, how many
total Joules come from the fuel in one 100-ton reactor core? Now consider how much coal
must be burned to produce the same number of Joules. How many miles of train cars would
be required to hold this much coal?

(7) The half-life of Au-198 is 2.695 days. What is the average lifetime of an Au-198 nucleus?

(8) Describe the production and loss mechanisms for Xe-135 in a reactor. Discuss how the
Xe-135 concentration changes when reactor power changes does it change immediately or
is there a time delay? Does it go monotonically to a new equilibrium value or does it first
go through a maximum or minimum? How might this affect reactor operations, including
startup following a shutdown?

(9) If a shield of thickness w reduces an uncollided beam intensity by a factor of 10, how thick
should it be to reduce the uncollided intensity by a factor of 1000?

(10) Show that for any differentiable function g(x, y, z),



~ = 4 g(x,
d( g) ~ y, z)
4 3

(11) Consider 10-keV neutrons injected uniformly into an infinite medium of 4 He. Suppose the
4 He density is and the scalar flux of 10-keV neutrons is . Give an equation for S (E),
0 1
the source-rate density of once-scattered neutrons. Be explicit about energy ranges over
which different expressions apply.

(12) Some research reactors have water holes in which various materials can be placed for
irradiation. Consider a water hole in such a reactor, and suppose it is filled with H2 O,
with all hydrogen being 1 H (no deuterium or tritium) and all oxygen being 16 O. Suppose
the density of the water is 0.95 g/cm3 . Suppose the density of thermal neutrons (neutrons
of energy below approximately 1 eV) in some portion of the water hole is 5] 106 n/cm3 .
Suppose the water molecules are vibrating with a Maxwellian distribution associated with
a temperature of 370 K.
NUEN 601 STUDY PROBLEMS, EXAM 1, 2017 3

At what rate per cm3 is hydrogen absorbing thermal neutrons in this portion of the water
hole? [Hint: 1 H is a 1/v absorber.] Assume that you are given the 2200-m/s absorption
cross section for 1 H. [Did you really need to know the temperature or that the water
molecules are in a Maxwellian?]

(13) Consider a beam of neutrons, all moving with the same velocity in the lab frame. Suppose
the beam is incident on a slab of material whose faces are perpendicular to the x axis and
whose thickness in the x direction is 5 cm. (The slabs dimensions in y and z are very
large.) The numbers:
neutron density in beam = 2E+03 n/cm3
neutron lab-frame speed = 3E+07 cm/s
beam direction = cos(/3)ex + sin(/3)ey
beam illuminates an area of 5 cm2 on the face of the slab
(a) At what rate (n/s) are neutrons entering the slab?
(b) If the slab material only captures neutrons (no scattering or fission or other neutron-
releasing reactions), with cross section 0.5 cm1 , then at what rate (n/s) are neutrons
exiting the slab?
(c) At what rate (n/s) are neutrons in the slab being captured?
(d) In the portion of the slab that is illuminated by the beam, what is (x), the scalar flux
as a function of x?
~
(e) In the portion of the slab that is illuminated by the beam, what is J(x), the net current
density as a function of x? Note that this is a vector quantityyou must specify all
components.

(14) (a) What is the elastic scattering cross section for the following nuclides given a neutron
energy in the range E (5 eV, 5 keV)? H-1, C-12, He-4, O-16.
(b) Do larger nuclei always have larger elastic scattering cross sections? If not, why not?

(15) Consider elastic scattering of 10-keV neutrons off of a nucleus whose mass is A times the
neutron mass. Suppose that at an incident neutron energy of 10 keV, scattering is s-wave.
What is the energy distribution of the once-scattered neutrons?

(16) Show by performing the integrals on the left side the following equation that the expression
on the right-hand side is the P1 -approximations expression for an incoming partial current:
(a) Show that Ficks Lawthe diffusion approximationfollows directly from the P1 equa-
tions if we have one energy group and either:
a k-eigenvalue problem, or
a steady-state problem with a fixed isotropic source
(b) Show that Ficks Law does not follow directly from the multigroup P1 equations with
more than one energy group or from the one-group P1 equations if there is time de-
pendence.

(17) Write down the diffusion equations and all necessary conditions for a well-posed problem, for
the following problems. Assume that all macroscopic cross sections and diffusion coefficients
are given.
4 NUEN 601 STUDY PROBLEMS, EXAM 1, 2017

(a) One energy group, symmetric slab, infinite in y and z, plane of symmetry at x = 0,
material in x (0, a), material in x (a, b), vacuum on the outside, fixed source
of strength S only in material , with material having non-zero fission cross section
but material having no fissioning material
(b) Same problem except three energy groups, with source emitting neutrons only in the
highest-energy group
(c) Same problem, one group again, except k-eigenvalue problem (of course with no fixed
source)
(d) One energy group, symmetric sphere of radius R, with fixed source emitting S n/(cm3 -
s) only in the region r (0, R/3), with vacuum on the outside
(e) One energy group, cube of width w with planes of symmetry at x = 0, y = 0, and z = 0,
with uniform fixed source of strength S, with vacuum at x = w/2 and at y = w/2, but
with incident partial current Jinc entering through the top z = w/2 surface. (Same on
the bottom, but symmetry makes this clear.)
(f) One energy group, finite cylinder of radius R and height H, so that the axial variable
z ranges from H/2 to H/2, with fixed source of strength S only in the volume given
by r (0, R/2) z (H/2, 0), with vacuum outside the cylinder.
(g) One energy group, k-eigenvalue problem, finite cylinder of radius R and height H with
uniform material properties, with vacuum outside the cylinder.

(18) Write down the transport equation and all necessary conditions for a well-posed problem,
for the following problems. Assume that all macroscopic cross sections are given.
(a) One energy group, symmetric slab, infinite in y and z, plane of symmetry at x = 0,
material in x (0, a), material in x (a, b), vacuum on the outside, fixed source
of strength S only in material , with material having non-zero fission cross section
but material having no fissioning material
(b) Same problem except three energy groups, with source emitting neutrons only in the
highest-energy group
(c) Same problem, one group again, except k-eigenvalue problem (of course with no fixed
source)

(19) Explain why fusion of light nuclides and fission of heavy nuclides both release energy.

(20) Describe the neutron-induced fission process, beginning with absorption of a neutron. De-
scribe what emerges on what time scales with what energies.

(21) Given the velocity-dependent neutron distribution and the velocity-dependent nucleus dis-
tribution, and given that neither is isotropic, write down the general form of the absorption
rate density in a material consisting of a single nuclide.

(22) All homework problemsespecially those that I can work in 10 minutes or less.

(23) Notes Chapters 1-7. You should read them. They will help solidify your understanding.
NUEN 601 STUDY PROBLEMS, EXAM 1, 2017

Material: Notes through Section 7.7.3 + Homeworks 1-5

(1) You are given that the angular flux, (~r, E, , t), is linearly anisotropic. Express in terms
~ r, E, t).
of the scalar flux, (~r, E, t), and the net current density, J(~

(2) What is the physical meaning of:


(a) (~r, E, t) integrated over all energy and over a given spatial volume?
~ r, E, t) integrated over all energy and over a given spatial volume?
(b) J(~
(c) (~r, E, t) integrated over all energy and over a given surface area?
~ r, E, t) integrated over all energy and over a given surface area, where en (~r)
(d) en (~r) J(~
is the outward unit normal at position ~r on the surface?
(e) The macroscopic absorption cross section for a given material?

(3) Explain why there are resonances in cross sectionswhy the cross section for nuclide A ZX
can be hundreds of times as large at certain neutron energies than it is at slightly larger or
smaller energies.

(4) Consider the multiplication factor, k, of a reactor.


(a) Provide two common definitions of this factor.
(b) Explain why these definitions are not always correct, and state the situation under
which they are correct.

(5) Consider the following sequence of reactions:


1 238 239
0n +92 U 92 U +
239 239
92 U 93 Np + e + +
239 239
93 Np 94 Pu + e + +

(a) How much energy is released by this sequence? (That is, what is the total Q value for
the three consecutive reactions?) Potentially helpful numbers are below. Note that
the masses for the U, Np, and Pu are for neutral atoms, not just nuclei. Note that the
mass of a neutrino is negligibly small relative to the mass of an electron. State your
assumptions and approximations.
neutron: 1.008664915 amu
proton: 1.007276467 amu
electron: 0.0005485799 amu
238 U: 238.0507826 amu
239 U: 239.0542878 amu
239 Np: 239.0529314 amu
239 Pu: 239.0521565 amu
c2 1 amu = 931.49 MeV.
1
2 NUEN 601 STUDY PROBLEMS, EXAM 1, 2017

(b) Write down differential equations for NU9 (~r, t), the number density of 239 U in a reactor
at time t; NNp (~r, t), the number density of 239 Np in a reactor at time t; and NPu (~r, t),
the number density of 239 Pu in a reactor at time t; under the following assumptions:
(~r, E) = given and independent of time
all microscopic cross sections are given (including capture and absorption cross
sections for each nuclide), pre-averaged over nucleus velocities, and independent
of time
238 U number density, NU8 (~r), is given and independent of time
239 U is produced only by neutron capture in 238 U, 239 Np is produced only by
decay of 239 U, and 239 Pu is produced only by decay of 239 Np (assume that other
production rates are negligible)
239 U, 239 Np, and 239 Pu can be lost either by decay or by absorbing a neutron.

(6) Each car in a coal train holds approximately 100 metric tons of coal. A typical coal train
contains 100 of these cars, is one mile long, and supplies a large coal-fired power plant for
one day at full power, which is roughly 3GW thermal. A 3-GWth nuclear power plants
reactor contains roughly 100 metric tons of uranium. If 5% of this uranium fissions over
the lifetime of the fuel, and each fission produces 190 MeV of recoverable energy, how many
total Joules come from the fuel in one 100-ton reactor core? Now consider how much coal
must be burned to produce the same number of Joules. How many miles of train cars would
be required to hold this much coal?

(7) The half-life of Au-198 is 2.695 days. What is the average lifetime of an Au-198 nucleus?

(8) Describe the production and loss mechanisms for Xe-135 in a reactor. Discuss how the
Xe-135 concentration changes when reactor power changes does it change immediately or
is there a time delay? Does it go monotonically to a new equilibrium value or does it first
go through a maximum or minimum? How might this affect reactor operations, including
startup following a shutdown?

(9) If a shield of thickness w reduces an uncollided beam intensity by a factor of 10, how thick
should it be to reduce the uncollided intensity by a factor of 1000?

(10) Show that for any differentiable function g(x, y, z),



~ = 4 g(x,
d( g) ~ y, z)
4 3

(11) Consider 10-keV neutrons injected uniformly into an infinite medium of 4 He. Suppose the
4 He density is and the scalar flux of 10-keV neutrons is . Give an equation for S (E),
0 1
the source-rate density of once-scattered neutrons. Be explicit about energy ranges over
which different expressions apply.

(12) Some research reactors have water holes in which various materials can be placed for
irradiation. Consider a water hole in such a reactor, and suppose it is filled with H2 O,
with all hydrogen being 1 H (no deuterium or tritium) and all oxygen being 16 O. Suppose
the density of the water is 0.95 g/cm3 . Suppose the density of thermal neutrons (neutrons
of energy below approximately 1 eV) in some portion of the water hole is 5] 106 n/cm3 .
Suppose the water molecules are vibrating with a Maxwellian distribution associated with
a temperature of 370 K.
NUEN 601 STUDY PROBLEMS, EXAM 1, 2017 3

At what rate per cm3 is hydrogen absorbing thermal neutrons in this portion of the water
hole? [Hint: 1 H is a 1/v absorber.] Assume that you are given the 2200-m/s absorption
cross section for 1 H. [Did you really need to know the temperature or that the water
molecules are in a Maxwellian?]

(13) Consider a beam of neutrons, all moving with the same velocity in the lab frame. Suppose
the beam is incident on a slab of material whose faces are perpendicular to the x axis and
whose thickness in the x direction is 5 cm. (The slabs dimensions in y and z are very
large.) The numbers:
neutron density in beam = 2E+03 n/cm3
neutron lab-frame speed = 3E+07 cm/s
beam direction = cos(/3)ex + sin(/3)ey
beam illuminates an area of 5 cm2 on the face of the slab
(a) At what rate (n/s) are neutrons entering the slab?
(b) If the slab material only captures neutrons (no scattering or fission or other neutron-
releasing reactions), with cross section 0.5 cm1 , then at what rate (n/s) are neutrons
exiting the slab?
(c) At what rate (n/s) are neutrons in the slab being captured?
(d) In the portion of the slab that is illuminated by the beam, what is (x), the scalar flux
as a function of x?
~
(e) In the portion of the slab that is illuminated by the beam, what is J(x), the net current
density as a function of x? Note that this is a vector quantityyou must specify all
components.

(14) (a) What is the elastic scattering cross section for the following nuclides given a neutron
energy in the range E (5 eV, 5 keV)? H-1, C-12, He-4, O-16.
(b) Do larger nuclei always have larger elastic scattering cross sections? If not, why not?

(15) Consider elastic scattering of 10-keV neutrons off of a nucleus whose mass is A times the
neutron mass. Suppose that at an incident neutron energy of 10 keV, scattering is s-wave.
What is the energy distribution of the once-scattered neutrons?

(16) (a) Show that Ficks Lawthe diffusion approximationfollows directly from the P1 equa-
tions if we have one energy group and either:
a k-eigenvalue problem, or
a steady-state problem with a fixed isotropic source
(b) Show that Ficks Law does not follow directly from the multigroup P1 equations with
more than one energy group or from the one-group P1 equations if there is time de-
pendence.

(17) Show by performing the integrals on the left side the following equation that the expression
on the right-hand side is the P1 -approximations expression for an incoming partial current:

1 h ~ r) = 1 (~r) 1 en J(~
i
~ r)
d (~r) + 3 J(~
:en <0 4 4 2
4 NUEN 601 STUDY PROBLEMS, EXAM 1, 2017

(18) Write down the diffusion equations and all necessary conditions for a well-posed problem, for
the following problems. Assume that all macroscopic cross sections and diffusion coefficients
are given.
(a) One energy group, symmetric slab, infinite in y and z, plane of symmetry at x = 0,
material in x (0, a), material in x (a, b), vacuum on the outside, fixed source
of strength S only in material , with material having non-zero fission cross section
but material having no fissioning material
(b) Same problem except three energy groups, with source emitting neutrons only in the
highest-energy group
(c) Same problem, one group again, except k-eigenvalue problem (of course with no fixed
source)
(d) One energy group, symmetric sphere of radius R, with fixed source emitting S n/(cm3 -
s) only in the region r (0, R/3), with vacuum on the outside
(e) One energy group, cube of width w with planes of symmetry at x = 0, y = 0, and z = 0,
with uniform fixed source of strength S, with vacuum at x = w/2 and at y = w/2, but
with incident partial current Jinc entering through the top z = w/2 surface. (Same on
the bottom, but symmetry makes this clear.)
(f) One energy group, finite cylinder of radius R and height H, so that the axial variable
z ranges from H/2 to H/2, with fixed source of strength S only in the volume given
by r (0, R/2) z (H/2, 0), with vacuum outside the cylinder.
(g) One energy group, k-eigenvalue problem, finite cylinder of radius R and height H with
uniform material properties, with vacuum outside the cylinder.

(19) Write down the transport equation and all necessary conditions for a well-posed problem,
for the following problems. Assume that all macroscopic cross sections are given.
(a) One energy group, symmetric slab, infinite in y and z, plane of symmetry at x = 0,
material in x (0, a), material in x (a, b), vacuum on the outside, fixed source
of strength S only in material , with material having non-zero fission cross section
but material having no fissioning material
(b) Same problem except three energy groups, with source emitting neutrons only in the
highest-energy group
(c) Same problem, one group again, except k-eigenvalue problem (of course with no fixed
source)

(20) Explain why fusion of light nuclides and fission of heavy nuclides both release energy.

(21) Describe the neutron-induced fission process, beginning with absorption of a neutron. De-
scribe what emerges on what time scales with what energies.

(22) Given the velocity-dependent neutron distribution and the velocity-dependent nucleus dis-
tribution, and given that neither is isotropic, write down the general form of the absorption
rate density in a material consisting of a single nuclide.

(23) All homework problemsespecially those that I can work in 10 minutes or less.

(24) Notes Chapters 1-7. You should read them. They will help solidify your understanding.
NUEN 601 STUDY PROBLEMS, EXAM 2, 2017

Material:

Notes, with emphasis on chapters 7-9


Homeworks, with emphasis on #s 5-7
Everything this study guides problems ask about
Texts, all assigned readings

(1) For each question below, assume that a critical reactor has been operating in steady state for several
days at such a low power level that the energy from fission is not enough to change the temperature
significantly. This continues until time t0 . Each of the following questions (except the first) begins
at t0 . That is, the changes are not cumulative from one question to the next. For some questions
you must take feedback into account.
(a) Is there a fixed source of neutrons in or near the reactor, or is there insufficient information
to tell?
(b) Assume that at t0 positive reactivity of $0.10 is added quickly, and sketch P (t) vs. t for t
ranging from t0 to t0 + 100 s. Be as quantitative as you can, label key features and values
of your graph, and state your assumptions.
(c) Repeat for a positive reactivity of $0.30.
(d) Repeat for a positive reactivity of $1.50.
(e) Repeat for a negative reactivity insertion of $5.00.
(f) Assume that at t0 a neutron source is placed next to the reactor, so that neutrons stream
into the reactor from outside. Sketch P (t) vs. t for t ranging from t0 to t0 + 100 s. Label
key features and values of your graph, and state your assumptions.
(g) Assume that at t0 a neutron source is placed next to the reactor while at the same time
there is and addition of $2.00. Sketch P (t) vs. t for t ranging from t0 to t0 + 500 s. Label
key features and values of your graph, and state your assumptions.

(2) Sketch a log-log plot of (~r, E) as a function of E in a large thermal reactor, for some position that
is far from the reactor boundary. Explain key features of your plot. You should have interesting
things to show and describe in the thermal range (sub-eV), the slowing-down range (eV to 100
keV), and the high-energy range (above 100 keV).

(3) Consider the 2-group diffusion equations for a bare homogeneous reactor. Assume that the bound-
ary between the fast (group 1) and thermal (group 2) neutrons is 4 eV.
(a) Write down the 2-group diffusion equations for this situation. Take advantage of the 4-eV
boundary between groups to make simplifications. What are those simplifications?
(b) Assume the same extrapolation distance in the boundary conditions for both groups. As-
sume that the eigenvalues and eigenfunctions of the Laplacian are known for the reactor
geometry, with Bg2 being the smallest of those eigenvalues. Derive an expression for the
multiplication factor, k.
1
2 NUEN 601 STUDY PROBLEMS, EXAM 2, 2017

(4) Consider 1GD theory for a bare homogeneous sphere, and consider the fundamental-mode distri-
bution. Assume for this exercise that the extrapolation distance is small compared to the sphere
radius.
(a) Use the diffusion approximation to obtain an expression for the net current density (whose
only nonzero component will be radial). Use your expression for the net current density to
evaluate the net leakage rate out of the outer surface.
(b) We have noted that in diffusion theory for homogeneous reactors, the quantity DBg2 behaves
like an escape cross section. Multiply the fundamental-mode scalar flux by this quantity
and integrate it over the volume of the sphere. Do you obtain the same answer as in part
(a)?

(5) 2-keV neutrons are introduced uniformly into an infinite medium of C-12. (A plot of C-12 cross
sections is given below.) The C-12 density is [mass/vol] and the scalar flux of 4-keV neutrons
is 0 [(n/(area-time)]. In all of the following, be explicit about energy ranges over which different
expressions apply, and define symbols you use other than and 0 . Take advantage of what
you know about neutron interactions with C-12 in the energy ranges that matter here. (Is there
absorption? Inelastic scattering? S-wave or other?)
(a) What is S1 (E1 ), the source-rate density of once-scattered neutrons, where E1 represents
the energy of the once-scattered neutrons? [n/(cm3 -s-keV)] Spell this out for all values of
E1 from 0 to .
(b) What is 1 (E1 ), the scalar flux of once-scattered neutrons? Include units. Spell this out
for all values of E1 from 0 to .
(c) What is S2 (E), the source-rate density of twice-scattered neutrons? [n/(cm3 -s-keV)] Spell
this out for all values of E2 from 0 to .

C-12 total and elastic-scattering cross sections, ENDF/B-VII


10

9 (n,elastic) xsec
(n,total) xsec

8
Cross Section (barns)

0
100 101 102 103 104 105 106
Energy (eV)
NUEN 601 STUDY PROBLEMS, EXAM 2, 2017 3

(6) Two bare homogeneous reactors are critical and operating in steady state. One reactor is twice as
big as the other in every dimension (and thus has 8 times the volume and 4 times the surface area).
The reactors have the same diffusion coefficients, same , and same f . The diffusion coefficient is
small relative to the reactor dimensions (but not zero). If the two reactors have the same peak flux
value, then according to 1GD theory approximately what is the ratio of the total neutron leakage
from the large reactor to that from the small reactor? [Think about a simple way to get net leakage
rate out of the entire reactor. It may help to think about the escape cross section.]]

(7) The positive-period method can be used to determine the approximate reactivity worth of a
change that induces positive reactivity, such as motion of a control rod. (Reactivity worth is
the amount by which the reactivity changes when the change occurs.) Consider the following
experimental procedures and measurements for determining the worth of a particular control rod
when it is moved from height h1 to height h2 :
(a) With the control rod at height h1 , other rods are adjusted until the reactor is critical and
operating in steady state at a low power level.
(b) The control rod is moved from h1 to h2 .
(c) Power is recorded several times per second as it rises, until it reaches a higher power level
at which the reactor temperature begins to increase noticeably.
Assume that you know all of the kinetics parameters (`p , the {i }, and the {i }) and you are
given the set of (t, P (t)) data. How would you determine the reactivity that was inserted by the
control-rod motion?
(8) For the questions below use the following data and assume that reactivity feedback from nuclides
other than 135 Xe is negligible.
P = power coefficient of reactivity = $0.001/MW [This does not include xenon effects.]
2P/1GW
Equilibrium 135 Xe worth = 5$ 1+2P/1GW .
(Note that P/1GW is just power expressed in GW units.)
In each part below, you must explain your reasoning and any equations that you usedont just
give an answer without an explanation, or you will not get credit. I highly recommend sketching
power and xenon concentration as a function of time to help answer the questions.
A source-free reactor operates at 2 GW for one week before control-rod insertion inserts reactivity
of $1.0.
(a) What is the power level a few minutes after the control-rod motion? Call this Pa and
compute the number.
(b) If the operators take no action, a few hours later will the power (call it Pb ) be greater than,
less than, or equal to Pa ?
(c) If the operators take no further action, a few days later what will be the power? Call this
Pc and compute the number.
(d) If at this time the operators insert +$1.5 of reactivity, what will the reactivity be after
several minutes? What will the power do as a result?
(e) If the operators take no further action, describe what will happen to the power level over
the course of the next few days.
4 NUEN 601 STUDY PROBLEMS, EXAM 2, 2017

(9) In this problem you will explore what happens when a neutron back-scatters off of a heavy nucleus
that is not at rest in the lab frame. We define the x axis to be in the direction of neutron motion
before the scattering event, with the heavy nucleus moving toward the neutron in the ex direction.
Define:
m neutron mass
M
A , where M nucleus mass
m
~v vex = lab-frame velocity of neutron before scatter
v lab-frame neutron speed before scatter
v~0 v 0 ex = lab-frame neutron velocity after scatter
v 0 lab-frame neutron speed after scatter
~ V ex = lab-frame nucleus velocity before scatter
V
V lab-frame speed of nucleus before scatter
~vCM lab-frame velocity of the center of mass
(a) What is the ratio of nucleus pre-scatter speed to neutron pre-scatter speed above which
the neutron will gain energy in the backscattering collision? [Hint: recall that in the CM
frame, the pre- and post-scattering speeds are the same. If the CM is at rest in the lab,
then the pre- and post-scattering lab-frame speeds must also be the same, and thus the
neutron would neither gain nor lose energy in the collision, right? What nucleus pre-scatter
speed causes ~vCM = 0?]
(b) Consider a neutron with lab-frame kinetic energy = 6.6 eV. Suppose the neutron elastically
backscatters off of a 238 U nucleus that is moving directly toward the neutron with lab-frame
kinetic energy = 0.1 eV. What is the lab-frame kinetic energy of the scattered neutron?
[Hint: first figure out the center-of-mass neutron speed, then transfer back to the lab frame.]
[Goals: Discover the surprising truth that neutrons well above the thermal range can gain energy
in collisions with heavy nuclei even if the nuclei are not much faster than average. Reinforce basic
2-body kinematics and recognize that conservation of momentum and energy allow you to figure
out anything you need about elastic scattering events.]

(10) Consider a bare cubical reactor of width 3 meters, composed of a homogeneous mixture of moder-
ator and fuel with known energy-dependent cross sections (including the transport cross section).
According to continuous-energy diffusion theory, what is a good approximation to the fundamental-
mode energy-dependent scalar flux as a function of x, y, z, and E in this reactor? Assume the reactor
is at low power and room temperature. Express your answer in terms of the given dimensions and
the energy-dependent cross sections to the extent that you can. You may need to divide the energy
range into segments to express your answer. If there are unknown constants in your expressions,
define and/or describe them.

(11) Write down the differential equations, interface conditions, and boundary conditions for a 1GD
description of a reactor with a spherical-shell homogeneous moderator of thickness b wrapped
around a spherical homogenous core of radius R. (This is a two-region problem. Write the 1GD
equation for each region, giving the flux function a different name in each region (maybe using
subscripts c for core and r for the reflector), then write down the two boundary conditions and two
interface conditions.)
NUEN 601 STUDY PROBLEMS, EXAM 2, 2017 5

(12) Consider an infinite uniform medium and assume the one-energy-group approximation. Assume
that all macroscopic cross sections are given, the medium is subcritical, and the neutrons are
distributed uniformly in direction and position throughout the medium.
(a) Write down the time-dependent one-group transport equation (along with other equations
(precursors) that need to go along with it) for this problem.
(b) Suppose there is a uniformly distributed isotropic fixed source of strength Sfixed n/(cm3 -s).
What is the steady-state scalar flux that will be achieved in the medium after transients
have died out? [Set the d/dt term to zero in each equation. If any other derivative terms
are zero in this problem, set them to zero as well. Integrate your transport equation over
all directions. Then do the algebra to find the stead-state scalar flux.]
(c) In the notes is a PRKE solution for a subcritical system with a source. Show whether or
not your steady-state scalar-flux solution agrees with the steady-state limit of this PRKE
solution. [You will need to recall the relation between scalar flux and neutron density. You
will need to recall or figure out the relation between prompt neutron lifetime and other
relevant quantities. You will need to recall or figure out the relation between k and the
macroscopic cross sections.]

(13) Use the appropriate figure from the notes to estimate the fraction of neutrons that emerge from
fission events with energies:.
(a) below 0.5 MeV
(b) between 0.5 MeV and 2 MeV
(c) between 2 MeV and 5 MeV
(d) between 5 MeV and 8 MeV
(e) greater than 8 MeV

(14) Consider an infinite homogeneous medium of material with neutrons introduced at high energies,
Esource > Es and left to slow down via collisions or be absorbed along the way. Recall the equation
developed in the notes for the energy distribution of neutrons in such a problem (Eq. (9.55) with
B 2 0) for the energy range E (Eth , Es ). With B 2 = 0 the equation is:
E<Es 1
f (E) psurv (Es , E) Seff
E t (E)

(a) Consider an infinite medium of oxygen and consider the energy range E (1 eV, 10 keV),
in which sO (E) constant and aO (E) 0. Insert the proposed solution, shown above,
into the infinite-medium transport equation for this problem and show whether or not it
is satisfied. That is, insert the proposed solution along with the correct expression for the
differential scattering cross section (under the assumption sO (E) constant) and see if it
works.
(b) Consider an infinite medium of CO2 with the same assumptions of constant scattering cross
sections and no absorption in the range of interest. Again insert the proposed solution
along with the correct expression for the differential scattering cross sections (under the
assumption s (E) constant) and see if it works.
6 NUEN 601 STUDY PROBLEMS, EXAM 2, 2017

(15) Suppose a reactor with prompt-neutron lifetime `P 104 s has multiplication factor k = 0.96.
Suppose the reactor is operating in steady state at room temperature with a total neutron popu-
lation of 108 neutrons.
(a) Is there a fixed source adding neutrons to the reactor? If so, estimate how many neutrons
per second it is putting in the reactor.
(b) Control rods are withdrawn to new positions, and measurements indicate that the total
neutron population settles at a new value of 4 108 neutrons. What is the new value of the
multiplication factor? What is the change in reactivity that is induced by the control-rod
withdrawal?
(16) Know the expressions for the geometric bucklings and fundamental modes for the reactor shapes
that we studied.
(17) Know how to take angular moments and energy integrals of the transport equation. Especially
look at the scattering termit is the trickiest one for these.

(18) Know how to find a critical size and shape for a reactor made of some given material whose cross
sections are known, under the assumption of either one-group or two-group diffusion theory.

(19) Read the notes!


NUEN 601 EXAM 1, MARCH 13, 2015

INSTRUCTIONS:

(1) Dont write where the staple will go (top left corner of each page).
(2) Write your name on the first page of your solutions.
(3) Turn in solutions, pages stapled together, but not the printed exam.
(4) Do not help others or receive help from people, notes, books, or electronics.
(5) Include units for each answer unless it is dimensionless.
(6) Calculate means write an expression that I could enter into a calculator.

(1) What is the physical meaning of:


(a) [10 points] (~r, E, t) integrated over all energy, over the spatial volume V , and over the
time period t?
(b) [10 points] (~r, E, t) integrated over all energy, over the spatial surface S, and over the
time period t?

(2) [15 points] Consider the following reaction:


239
92 U 239
93 Np + e +
How much energy is released in this reaction? Potentially helpful numbers are below. Masses given
for the U and Np are for neutral atoms, not just nuclei. State assumptions and approximations.
neutron: 1.008664915 amu
proton: 1.007276467 amu
electron: 0.0005485799 amu
238 U: 239.0542878 amu
238 Np: 239.0529314 amu
c2 1 amu = 931.49 MeV.

(3) Consider a mono-velocity beam of neutrons incident on a slab of material whose faces are perpen-
dicular to the x axis. (The slabs dimensions in y and z are very large.) The numbers:
neutron density in beam = 2E+04 n/cm3
neutron lab-frame speed = 4E+06 cm/s
beam direction = cos(/6)ex + sin(/6)ez
beam illuminates an area of 5 cm2 on the face of the slab
t = = 2cm1 in the slab

(a) [10 points] At what rate (n/s) are neutrons entering the slab?
(b) [10 points] Calculate (see instructions) the width of the slab that would cause the slab
to absorb 90% of the incident neutrons.
1
2 NUEN 601 EXAM 1, MARCH 13, 2015

(4) [10 points] At a particular location in a pool of H2 O, the water density is 0.9 g/cm3 and the
density of thermal neutrons is 4 105 n/cm3 . The water molecules are in a Maxwellian distribution
at a temperature of 350 K. The 2200-m/s absorption cross section for hydrogen is 0.33 barns. For
oxygen it is 0.0002 barns.
Calculate (see instructions) the number of thermal neutrons per cm3 per second being absorbed
by the H2 O at the given location in the pool.

(5) Write down the diffusion equations and all necessary conditions for a well-posed problem, for the
following problems. Assume that all macroscopic cross sections and diffusion coefficients are given.
(a) [10 points] One energy group, k-eigenvalue problem, finite cylinder of radius R and height
H with uniform material properties (i.e., a homogeneous mixture), with vacuum outside
the cylinder.
(b) [10 points] Same problem except two energy groups.

(6) Suppose someone solves the two-group diffusion k-eigenvalue problem for a bare finite homogeneous
cylinder of radius R and height H. Let z = 0 be in the center of the reactor. In terms of the
diffusion solution, the two group diffusion coefficients, and the two group cross sections, write down
the following rates.
(a) [10 points] The fission rate (fissions/second) in the reactor.
(b) [5 points] The rate (neutrons/second) at which neutrons leak out of the top surface (at
z = H/2).

Extra Credit [5 points] 106 neutrons of energy almost exactly 1 keV suffer scattering collisions in
H2 O at room temperature. At this energy, sH 20 barns and sO 4 barns. How many neutrons
emerge from their first scattering events with energies between 700 and 900 eV?
NUEN 601 EXAM 2 MAKEUP C, MAY 7, 2015

INSTRUCTIONS:

(1) Dont write where the staple will go (top left corner of each page).
(2) Write your name on the first page of your solutions.
(3) Turn in SOLUTIONS, stapled, but keep the exam questions if you like.
(4) Do not help others or receive help from people, notes, books, or electronics.
(5) Include units for each answer unless it is dimensionless.
(6) Calculate means write an expression that I could enter into a calculator.

(1) [15 points] Sketch a log-log plot of (~r, E) as a function of E in a large thermal reactor, for some
position (~r) that is far from the reactor boundary. Describe and/or explain key features of your
plot. This should include approximate functional forms that attains in various energy ranges,
along with explanations of the origins of other interesting features..

(2) Consider the following experimental procedure and measurements:


A reactor is critical and in steady state at a low power level with a control rod at height
h1 . It remains in this steady state for a long time.
The control rod is moved (withdrawn) from h1 to h2 at time 0.
A neutron detector near the reactor records count rates as shown in the table:

time count rate


t<0 600 cpm
t=0 600 cpm
t = 1s 840 cpm
t = 2s 960 cpm
t = 10s 3000 cpm
t = 18s 9000 cpm
t = 26s 27000 cpm

(a) [10 points] If you had a calculator or spreadsheet, how would you find the reactor period
from this data? Your answer should clearly tell a person with a calculator or spreadsheet
how to obtain the correct number, with units.
(b) [10 points] Let T be the correct reactor period from part (a). Recall the inhour equation:
I
`p s 1 X i s
= + ,
1 + `p s 1 + `p s s + i
i=1
Assume that T and the {i }, {i }, and `p are all known. How would you estimate the
reactivity that was inserted by the control-rod motion? Provide an expression that contains
only these given quantities and that I could punch into a calculator. (Dont make this hard.)
1
2 NUEN 601 EXAM 2 MAKEUP C, MAY 7, 2015

(3) For the questions below use the following data and assume that reactivity feedback from nuclides
other than 135 Xe is negligible.
$1.5
P = power coefficient of reactivity = GW [This does not include xenon effects.]
135 Xe 2P/1GW
Equilibrium worth at power level P = 5$ 1+2P/1GW .

In each part below, you must explain your reasoning and any equations that you usedont just
give an answer without an explanation, or you will not get credit. I highly recommend sketching
power and xenon concentration as a function of time to help answer the questions.
A source-free reactor operates at 3 GW for one week before control-rod motion inserts reactivity
of $1.5. Then:
(a) [10 points] What is the power level several minutes after the control-rod motion? Call
this Pa and compute the number.
(b) [10 points] If the operators take no action, a few hours later will the power (call it Pb ) be
greater than, less than, or equal to Pa ?
(c) [10 points] If the operators take no further action, a few days later what will be the power?
Call this Pc and provide an equation in which Pc is the only unknown.

(4) [20 points] Write down the differential equations, interface conditions, and boundary conditions
for the two-group diffusion k-eigenvalue problem for a two-region cylindrical reactor:
The core is homogeneous and contains fissile material. Its axial range is z (H/2, H/2).
Its radial range is r (0, RC ).
The reflector is homogeneous with material that does not fission. It is an annular region
of thickness b that surrounds the core along the entire height z (H/2, H/2).
There is vacuum above z = H/2, below z = H/2, and outside of r = RC + b.

2-region reactor geometry

(5) [15 points] Suppose you and a classmate are given identical amounts of a homogeneous mixture of
fissile material and moderator. Your friend shapes her material into a cube and finds that it is just
critical. You shape yours into a cylinder with height = radius (H = R). Is your reactor subcritical,
supercritical, or critical? Justify your answer with convincing analysis. There will be no credit
for stating an answer that is not explained. You may assume that the dimensions are much greater
than the size of the diffusion coefficient in the material.

Extra Credit [5 points] There is a steady-state neutron population of n0 in a room-temperature


reactor whose multiplication factor is k = 0.98. At t = t0 control rods are moved slightly, causing
k to change to 0.99. Sketch what you believe will be the neutron population as a function of time
from just before t0 to t0 + 200 seconds. Be as quantitative as you can for the late-time portion of
your sketch.
Introduction to Nuclear Systems
and the
Basic Elements of Reactor Physics

Dr. William S. Charlton


NUEN 601
Nuclear Reactor Theory
Texas A&M University
Nuclear Systems
Nuclear systems are simply any system that uses
nuclear process to achieve some goal
These could include
nuclear reactors for power production
accelerator-driven systems for the transmutation of
waste
large radioactive sources for cancer therapy
fusion reactors
Our primary interest is nuclear reactors
but the physics learned here will be applicable to other
systems as well
Nuclear Reactors
Nuclear reactors rely on a
controlled nuclear chain
reaction to supply energy
This energy could be:
heat
electrical power
radiation
1 of 4 reactor categories:
power
research
production
propulsion
Nuclear Reactors
(continued)
Power reactors are used Production reactors are
to supply ~20% of U.S. used to produce nuclear
electrical needs materials
primarily for military
applications
Research reactors provide
numerous services:
medical isotope production
Propulsion reactors have
basic physics research
been used for
naval vessels
nuclear fuels tests
spacecraft
nuclear analytical services
aircraft
Pressurized Water Reactors
Originally designed by
Westinghouse (Bettis) for
military ship applications

Westinghouse Nuclear Power


Division designed first
commercial reactor
the first commercial PWR plant in
the United States was
Shippingport
located near Pittsburgh,
Pennsylvania
Pressurized Water Reactors
(continued)
Characteristics:
Pressurizer pressurized water
Primary moderated and cooled
Loop
LEU oxide fuel
Rx SG
Coolant is pressurized to
inhibit boiling
Pump Primary and secondary
T loop
decrease risk of turbine
Secondary contamination
Loop C
aid in reactor control
Pump Refuelings must be done
with the plant shutdown
Boiling Water Reactors
Originally designed by Allis-
Chambers and General Electric
(GE)
General Electric design has
survived
all Allis-Chambers units are now
shutdown

The first GE US commercial


plant was at Humboldt Bay
(near Eureka) in California
Boiling Water Reactors
(continued)
Characteristics:
low-pressure, light-water
moderated and cooled
LEU oxide fuel
T

BWR uses a single loop:


steam is generated directly
Rx in the reactor core through
C bulk boiling of the coolant
steam directly into the
Pump
turbine

Refuelings must be done


with the plant shutdown
Liquid Metal Fast Breeder
Reactors
Liquid metal cooled with no
moderator

May use highly enriched


uranium fuel drivers and
natural uranium targets to
breed Pu

Coolant is usually sodium,


or a combination of sodium
and potassium
liquid metal has excellent
heat transfer properties
LMFBR
(cont.)
Gas Reactors
Core is moderated with
graphite

Carbon dioxide or helium


gas is used to cool the fuel
and the graphite
moderator

Due to the graphite, the


cores can be designed
using natural uranium fuel
Canadian Deuterium-Uranium
Reactors
Heavy water moderated and
cooled

Uses natural uranium metal


fuel

Does not require a pressure


vessel

Can be refueled online

Fuel generally achieves a low


burnup
CANDU
(continued)
CANDU
(continued)
Russian Reactor Systems
The VVER is a Russian
version of a PWR
consists of a hexagonal
lattice
uses Zr-1%Nb clad

RBMK is a graphite-
moderated, water-cooled
reactor that uses natural
uranium fuel

BNs are Russian


LMFBRs
Accelerator Driven Systems
Sub-critical system Neutrons Per Proton Versus Incident
Proton Energy for Various Target
Materials

Uses an accelerator to 60
produce neutrons through
50
a process called spallation U

Neutrons Per Proton


40

Increased safety
30
Pb
Th
Accelerator is expensive 20
Sn
10
Sub-critical systems have Be
some advantages in 0
0.4 0.8 1.2 1.6
transmutation Proton Energy (GeV)
Fusion Reactor Systems
Relies on nuclear fusion
to produce energy instead
of fission

Chain reaction is
maintained through the
heating and confinement
of plasma fuel

Fusion is
clean because it produces
no fission products
safe because it will always
cool itself down
Fusion Reactor Systems
(continued)
Critical and Subcritical
Assemblies
Small assemblies that are
used to
test criticality and nuclear data
provide a neutron/gamma-ray
source
benchmark reactor physics
codes

Come in all shapes and


sizes and all materials

LANL and ORNL lead the


world in critical assembly
testing
Research Reactors
Generally smaller cores
than found for power or
production reactors
Common types include
TRIGA, MTR, and IRT
reactors
Research can include:
isotope production
neutron activation analysis
neutron and gamma
radiography
neutron physics
Large Radioactive Sources
These facilities are often
used in industry and
research
These could include
Cf-252 sources for neutron
radiography
PGAA for realtime process
feedback
Co-60 sources for food
irradiation
thermal generators
Simpler to operate than a
reactor
Other Nuclear Systems
Fuel fabrication and enrichment facilities

Spent fuel reprocessing facilities

Spent fuel storage facilities

Space reactors and nuclear rockets

Pulse power systems

Radiation detection systems


Reactor Physics
Reactor physics is the study of the behavior and
characteristics of neutron fission chain reacting
systems
In most cases, our primary interest is in
determining
the rate at which different reactions are occurring in the
system
These reactions allow us to determine:
reactor criticality and reactor power
doses to humans and radiation damage to materials
nuclear material production
Nuclear Reaction Rates
Nuclear reactions are generally induced by either
radioactive decay or radiation bombardment
To know the rate at which reactions are
occurring, we need to know:
rate at which radiation travels through a material
probability that radiation will cause a reaction
probability that any radioactive products in the material
will decay as a function of time
Thus, we need to have nuclear data and have the
ability to predict the transport of radiation
Reactor Physics Elements

Nuclear
Data

Radiation Buildup
Transport and Decay
The Importance of Basic Nuclear
Physics
To be able to understand any of the three basic
building blocks of reactor physics
we need to have some basic nuclear physics
knowledge

We need to understand the constituents of a


nucleus, mass and energy, radiation, radioactive
decay, nuclear reaction physics, etc

We will study these basic processes in some


detail today and over the next few weeks
Fundamental Particles
The world is composed of various subatomic
particles
also referred to as fundamental particles

Many fundamental particles are composed of


quarks
and are bound together by an intermediary referred to
as gluons

The fundamental particles are divided into:


leptons (electrons, positrons, and neutrinos)
hadrons (a subset of which includes protons and
neutrons, also called baryons)
Fundamental Particles
(continued)
The leptons are subject to the weak nuclear force

The hadrons are subject to both the weak and


strong nuclear forces

The hadrons are composed of quarks


the exchange of gluons between the quarks is
responsible for the strong nuclear force

To understand nuclear reactions


we need to have a basic understanding of some of
these fundamental particles
Fundamental Particles
(continued)
For the purposes of reactor physics, we will
concern ourselves with only the following
fundamental particles:
Type Particle Symbol Charge Rest Mass Half-life
baryons proton p +e 1.007 amu stable?
neutron n 0 1.009 amu 10.4 min
leptons electron -, e- -e 5.486E-4 amu stable
positron + +e 5.486E-4 amu stable
neutrino 0 ? amu stable
gluons photon 0 0 stable
Atomic and Nuclear Structure
Atoms are the building blocks of all matter
atoms consist of a massive nuclei surrounded by a
cloud of electrons

The total number of protons in a nucleus is called


the atomic number (or Z)

The total number of neutrons in a nucleus is


called the neutron number (or N)

The total number of nucleons in a nucleus is


called the atomic mass number (or A)
Radiation
Radiation is simply particles
in motion
Radiation includes:
gamma-rays
neutrons
-
beta-rays
alpha particles
neutrinos
cosmic rays
Radiation is produced
n via decay
as a biproduct of nuclear and
atomic reactions
Mass and Energy
The energy produced in nuclear systems is
derived from Einsteins famous formula:
E0 = m0c 2

where c is the speed of light


E0 is the rest energy
m0 is the rest mass
Mass and Energy
(continued)
It is important to note that the mass of a particle
in motion is different than its rest mass due to
Einsteins general theory of relativity:
m0
m=
2
1 v
c 2

thus as v increases the particle mass increases


What is the particle mass when the particle speed is
v=c?
Mass and Energy
(continued)
The total energy of a particle (i.e., its rest mass
energy plus its kinetic energy) is given by
E = mc 2

Thus, the kinetic energy is given by

KE = mc 2 m0 c 2

1
2
KE = m0 c 1
v 2
1
c2
Mass and Energy
(continued)
Is this kinetic energy expression the same as the
classical expression?
1
KE = mv 2
2
Mass and Energy
(continued)
We can expand the radical term in the relativistic
kinetic energy expression as a binomial
expansion
Mass and Energy
(continued)
If v/c is small,
then we can discard all of the terms past the second
term in the expansion, and we get
Radioactive Decay
and
Number Densities

Dr. William S. Charlton


NUEN 601
Nuclear Reactor Theory
Texas A&M University
Radioactive Decay
The figure shows all of the
stable nuclides as a Atomic Number (Z) Versus Neutron
Number (N) for All Known Stable
function of Z and N Isotopes
note that there are more 140
neutrons than protons in Z=N Line
120 Neutron
most nuclei Poor
100

the extra neutrons provide +
80
nuclear stability

Z
60
-
40
There are only certain Neutron
Rich
20
combinations of N and Z
0
that produce stable nuclei 0 20 40 60 80 100 120 140
N
the rest are radioactive
Radioactive Decay
(continued)
The mode of decay for
any isotope is determined Atomic Number (Z) Versus Neutron
Number (N) for All Known Stable
by its position on the chart Isotopes

140

Heavy elements are 120 Neutron


Poor
Z=N Line

typically 100

80 +

A A 4 4
Z
X Z 2 2
Y + 60
Z -
40
Neutron
20 Rich
Heavy elements might
0
also 0 20 40 60 80 100 120 140
N
Radioactive Decay
(continued)
Neutron rich isotopes are
typically - emitters Atomic Number (Z) Versus Neutron
Number (N) for All Known Stable
Isotopes
A A 0 0
Z X Z +1 1 0
Y + + 140

Neutron Z=N Line


Neutron poor isotopes are 120
Poor
generally + emitters (or 100
+

80
E.C.)
Z
60
A A 0 0
Z 1 1 0
+ + -
Z X Y 40
Neutron
20 Rich
A 0 A 0
Z X + 1 e Z 1 0
Y + 0
0 20 40 60 80 100 120 140
Fission products are N

typically neutron rich


Radioactive Decay
(continued)
Consider the excerpt from
the Chart of the Nuclides

The element sodium


(Z=11) has various
isotopes
(N=11, N=12, N=13, )
only Na-23 (Z=11, N=12) is
stable
Na-22 decays by positron
(+) emission
Na-24 decays by -
emission
Radioactive Decay
(continued)
Nuclei (such as O-15) which are lacking in
neutrons undergo + decay:
15 15 0 0
8 O 7 N + +1 0
+

On the other hand, nuclei (such as O-19) that are


excessively neutron rich undergo - decay:

19 19 0 0
8 O 9 F + 1 0
+
Radioactive Decay
(continued)
Often, the daughter
nucleus is also unstable
and will undergo an
additional - decay

This leads to a decay


chain:
20 20 0 0
8 O 9 F + 1 0
+
20 20 0 0
9 F 10 Ne + 1 0
+

where Ne-20 is stable


Radioactive Decay
(continued)
The nucleus formed as the result of a decay,
electron capture, or -decay is often left in an
excited state

The excited nucleus can relieve this excitation


energy by emitting one or more

It should be noted that


these -rays are frequently attributed to the parent
nucleus decay not the de-excitation of the daughter
nucleus since this de-excitation occurs very quickly
Radioactive Decay Example
60Co decays as follows:
60 60 0 0
+ 1 0 +
+
60Co
27 Co 28 Ni

2.506 MeV
Thus, most of the time 99+%
60Co emits two -rays of 0.013%
2.158 MeV
0.12%
1173 and 1332 keV
1.332 MeV

60Co has a half-life of 5.27


years or
t1 / 2 = 5.27 yrs
60Ni
Modes of Decay
Decay Product Principle Other Radiations
Mode Nucleus Radiations
A-4, Z-2 -lines -lines (often)

- Z+1, isobaric continuous -lines (often)

+ Z-1, isobaric + continuous -lines (often), annihilation


(511 keV)
Z-1, isobaric x-rays, auger electrons,
lines of daughter
I.C. Z, A e- -lines x-rays, auger electrons,
lines of parent
I.T. Z, A -lines n/a

S.F. distribution fission fragments n, -, , etc...


Activity
Source activity (A) Decay constant ()
the rate of radioactive decay probability of decay per unit
(decays/sec) time per radioactive atom
common units is constant for all time
Becquerel (Bq) one Bq is and for all atoms of the
one decay per second same species
Curie (Ci) one curie is is equal to
equal to 3.7 x 1010 decays
per second
calculating activity:

t1/2 is the half-life of the


N is the number of atoms radioactive atoms in sample
in the sample
Decay Calculations
We can derive most dN
= 0 N ( t ) = N ( t )
activity and decay dt
relationships from the 1
simple balance equation: dN = dt
N (t )
dN N (t ) t
= [rate of gain ] [rate of loss ] 1
dN ' = dt '
dt N'
N0 0
Assume we ln[N (t )] ln[N 0 ] = t 0
have a sample which N (t )
contains only N0 radioactive ln = t
atoms at time t=0 N0
N (t )
= e t
interested in the number of
atoms [N(t)] at new time t N0
Decay Calculations
(continued)
Thus, our number of If the N(t1/2) is half of the
atoms is number of atoms we
began with:
N (t1 / 2 ) 1
=
This also means the N0 2
activity of our sample is 1
= e t1 / 2
A = N 2
ln(1) ln( 2) = t1 / 2
A(t ) = N 0e t
0 ln( 2) = t1 / 2
A(t ) = A0e t
where A0 is the initial activity
of the sample
Mean Life
Probability of an atom surviving from 0 to time t is e-t
Probability that an atom will decay between t and dt is:
probability = ( prob of surviving from 0 to t )( prob of decay in dt )
( )
probability = e t (dt ) = e t dt

Therefore, the mean life is given by



t
t e dt

= t = 0 = te t dt
t 0
e dt
0
Production of Radioactive
Species
If we had radioactive Solution to this system is
atoms being produced at
some constant rate R
then our balance equation
would become: where t is the irradiation
time
dN
= R N N(t)=R/
dt Saturation

N(t)
lets assume that at t=0
there are no radioactive
atoms in the sample: t

N (t = 0) = 0 we will assume we reach


saturation after 4-5 t1/2
Production of Radioactive
Species Example
Suppose we begin with a sample of 27Al
27
Al + n 28Al +
28Al has a half-life of 2.25 minutes
thus, if we allow the sample to irradiate for ~10 minutes
we will have saturated the sample with 28Al
For an activated product
(
A(t ) = R 1 e t )
as we saturate the sample, the activity will approach

saturation activity
Number Densities
In previous calculations,
we have been essentially
solving for the atom number
density
in atoms/cc, atoms/sample,
etc.

Number densities are a


pivotal though simple
concept in reactor physics

One should note the


macroscopic nature of
number densities
Number Densities
(continued)
It is important to be able to calculate the atom
densities of a material when given the mass
density of that material

where
is the mass density (in g/cc)
Na is Avogadro's Number (in atoms/mole), and
M is the molecular mass (g/mole)
Number Density Example 1
Suppose we have a sample of aluminum (Al)
We know that Al is composed of 100% 27Al
We also know
=2.70 g/cc
M =26.98154 g/mole
Na =6.022E23 atoms/mole
Thus:
N a (2.70 g / cc)(6.022 E 23 atoms / mole)
N ( Al ) = =
M (26.98154 g / mole)
N ( 27 Al ) = N ( Al ) = 6.026 E 22 atoms / cc
Mixtures of Nuclides
The previous formulation works for a single
nuclide
with the density and atomic mass given for that nuclide
Often samples consist of a mixture of nuclides
mixture can be specified in terms of weight percents
(wi)
wi N A
Ni =
100 M i
or in terms of atom percents (i)
i NA
Ni =
100 M
Mixtures of Nuclides
(continued)
The atomic mass of a mixture of isotopes can be
calculated as follows:

where the mixture is made of I isotopes


dont try to use these for a molecule though
Number Density Example 2
Suppose we have a sample of natural uranium

NatU is composed of
99.2745 a/o 238U
0.7200 a/o 235U
0.0055 a/o 234U

We also know the following about NatU:


=18.90 g/cc
M =238.0289 g/mole
Na =6.022E23 atoms/mole
Number Density Example 2
(continued)
Thus we can find:
N a
N (U ) =
M

(18.90 g / cc )(6.022 E 23 atoms / mole)


N (U ) =
( 238.0289 g / mole)

N (U ) = 4.782 E 22 atoms of U / cc
Number Density Example 2
(continued)
[ ]
N ( 234U ) = atoms 234U atom U N (U )
N ( 234U ) = (5.5E 5)( 4.782 E 22) = 2.630 E18 atoms 234U / cc

[ ]
N ( 235U ) = atoms 235U atom U N (U )
N ( 235U ) = (0.0072)( 4.782 E 22) = 3.443E 20 atoms 235U / cc

[ ]
N ( 238U ) = atoms 238U atom U N (U )
N ( 238U ) = (0.992745)( 4.782 E 22) = 4.747 E 22 atoms 238U / cc
Number Density Example 3
(continued)
Suppose we have a sample of low enriched
uranium (LEU) metal

Our LEU is composed of


97 w/o 238U
3 w/o 235U

We also know the following about our LEU


=18.90 g/cc
M(238U) =238.050785 g/mole
M(235U) =235.043924 g/mole
Na =6.022E23 atoms/mole
Number Density Example 3
(continued)
We can determine the mass concentrations of
235U and 238U as follows:

(U ) = 18.90 g / cc

235 g 235U
( U ) = (U ) gU
= (18.90)(0.03)

( 235U ) = 0.567 g 235U / cc

238 g 238U
( U ) = (U ) g U
= (18.90)(0.97)

( 238U ) = 18.333 g 238U / cc
Number Density Example 3
(continued)
Thus, the 235U atom densities are

( 235U ) N a
N ( 235U ) =
M ( 235U )
235 (0.567 g / cc )(6.022 E 23 atoms / mole)
N( U)=
( 235.043924 g / mole)

N ( 235U ) = 1.459 E 21 atoms of 235U / cc


Number Density Example 3
(continued)
Thus, the 238U atom densities are

238 ( 238U ) N a
N( U)=
M ( 238U )
(18.333 g / cc )(6.022 E 23 atoms / mole)
N ( 238U ) =
( 238.050785 g / mole)

N ( 238U ) = 4.638 E 22 atoms of 238U / cc


Number Density Example 4
UO2 nuclear fuel:
= 10.6 g/cc
uranium is 88.147 weight percent of the molecule
U is enriched to 3 a/o 235U

M (O ) = 15.9994 g / mole

( )
M 235U = 235.043924 g / mole

( )
M 238U = 238.050785 g / mole
Number Density Example 4
(continued)
Atomic mass of U and molecular mass of UO2:

M (U ) =
3
100
(
M 235U +) 97
100
( )
M 238U = 237.9606 g / mole

M (UO2 ) = M (U ) + 2 M (O ) = 269.9594 g / mole

Number density of UO2 molecules and U atoms:


(UO2 ) N A
N (UO2 ) = = 2.3646 E 22 molecules / cc
M (UO2 )

N (U ) = N (UO2 ) = 2.3646 E 22 atoms / cc


Number Density Example 4
(continued)
Thus, we can calculate the 235U atom density

N ( U )=
235 ( )
235U
N (U )
100

( )
N 235U = 7.0937 E 20 atoms / cc

I actually gave you enough information to


calculate this value another way
I will leave it as an exercise to the student to see if the
other method gives you the same result
Radioactive Decay Chain
Calculations

Dr. William S. Charlton


NUEN 601
Nuclear Reactor Theory
Texas A&M University
Three Member Decay Chain
Many nuclides decay
through complicated
A -Parent parent-daughter chains

a The concentrations of
B -1st Daughter these nuclides can be
determined analytically
b
2nd Daughter- C We will assume the
following initial conditions:
NA(t=0)=N0
NB(t=0)=0
NC(t=0)=0
Decay Chain Solution for A
The governing equations The solution for A is
for this system are 1
dN A = a dt
N A (t )
dN A
= a N A (t ) N A (t )
1 t
dt dN ' = a dt '
N A'
N0 0
dN B N A (t )
= a N A (t ) b N B (t )
dt ln = a t
N0

dN C
= b N B (t )
dt
weve done this one before
right?
Decay Chain Solution for B
Thus, the equation for B becomes
dN B
= a N 0 e a t b N B
dt

The solution for B is


dN B
+ b N B = a N 0e a t
dt
dN B bt
e + b N B ebt = a N 0e a t ebt
dt
d
dt
[ ]
N B ebt = a N 0e(b a )t
Decay Chain Solution for B
(continued)
Solving the ODE yields:
N B ebt = a N 0e(b a )t dt
a N 0 (b a )t
N B ebt = e +D
b a
a N 0 a t
N B (t ) = e + De bt
b a

Our initial condition is:

N B ( t = 0) = 0
Decay Chain Solution for B
(continued)
Substituting t=0 into our equation above:
a N 0
N B ( t = 0) = +D=0
b a
N
D= a 0
b a
N N
N B (t ) = a 0 e a t a 0 e bt
b a b a
Thus, the solution for B is
Decay Chain Solution for C
The equation for C then becomes

dt
=
b a
e (
dN C b a N 0 a t
e b t )
The solution for C is as follows:

dN C = e (
ba N 0 a t bt
e dt )
b a

N 1 1
N C (t ) = b a 0 e a t + e bt + F
b a a b
Decay Chain Solution for C
(continued)
We use our initial condition to get F:
N C ( t = 0) = 0
b N 0 a N 0
N C ( t = 0) = + +F
b a b a
b a
F= N0 = N0
b a

The solution for C is then:


Decay Chain Results
If we know the half-lives of
Normalized Atom Density
the parents and daughters Versus Time for Three-Member
Decay Chain
t1a/ 2 = 15 sec
1
ln( 2) A
= 0.0126 sec1
0.9
a =

Normalized Atom Density


B
0.8
15 0.7
C

t1b/ 2 = 55 sec 0.6


0.5

b = 0.0462 sec1
0.4
0.3
0.2
t1c/ 2 = 0.1
0
c = 0.00 sec1 0 20 40 60 80

Time (sec)
Transient Equilibrium
If b > a, then when t gets Normalized Atom Density
large Versus Time for Decay Chain
with Transient Equilibrium

N B (t ) =
b a
(
a N 0 a t
e e b t ) 1

Normalized Atom Density


a N 0 a t A
N B (t ) = e B
b a 0.1 Equilibrium

0.01
This is called transient 0 100 200 300 400 500
equilibrium Time (sec)
Secular Equilibrium
If parent has extremely long Normalized Atom Density
half-life (i.e., a is very small): Versus Time for Decay Chain

( )
with Secular Equilibrium
a N 0
N B (t ) 1 e b t 1
b a

Normalized Atom Density


This implies b>>a

( )
0.1 A
a N 0
N B (t ) 1 e b t B
b Equilibrium

(
b N B (t ) a N 0 1 e b t ) 0.01

After 4-5 half-lives, we reach


secular equilibrium 0.001
0 100 200 300 400 500
Time (sec)

You might also like