H. J. V.

Tyrrell

The Origin and Present Status of

The University
Sheffield, England

Fick's Diffusion Law

serious experimental study of the

diffusion of one solution into another was first undertaken by Thomas Graham (I), who devised a number
of experimental techniques for studying the phenomenon and obtained a great deal of qualitative and
quantitative data. His achievements in this field
have perhaps been overrated and he is more justly
remembered for his distinction between crystalloids
and colloids in terms of the different rates a t which they
pass through membranes, and for his description of the
process of dialysis. A far more important name in the
early literature of the subject is that of Adolf Fick
(1829-1901). "Fick's Laws of diffusion" and the
"Fickian frame of reference" are terms which frequently appear in modern papers on transport properties. Virtually all experimental papers on diffusion
are concerned, in the first instance, with the determination of diffusion coefficients defined in a manner
similar to that proposed by Fick, namely.
J = -D dcldz
(1)

Ludwig, sixteen years older than he, was Professor of

Anatomy, and another brother, Heinrich, seven years
his senior, and later Professor of Commercial Law a t
Ziirich, was then privatdozent. Adolf's original intention was to take mathematics, but Heinrich, with
great perception, suggested that he should read for a
medical degree, studying mathematics as a subsidiary
subject, since his aptitude for this would be of inestimable value if he pursued a career in medical research. This judgement was later fully vindicated,
because Adolf Fick became an outstanding figure in
that small group of nineteenth century physiologist8

where J is the one-dimensional flow in moles or grams

per unit area per unit time across a reference plane, D
is the diffusion coefficient,and c is the concentration in
moles or grams per unit volume. An equation of this
Form is often termed Fick's first law, and the derived
equation.

Fick's second law. Furthermore, attempts to obtain

semiempirical equations to predict values of diffusion
coefficients from other physical properties have a long
history, while a priori calculations of transport coefficients, including diffusion coefficients, in liquids
from statistical-mechanical considerations have attracted considerable attention from modern theoreticians. Clearly, Fick's ideas have proved fruitful for
more than a century, almost as fruitful in fact as the
very similar ideas of Ohm in the field of electrical
conduction, and of Fourier in that of heat conduction.
Yet his paper on diffusion is usually misquoted;
his name, unlike those of Fourier and Ohm does not
appear in general works of reference in English, such as
"Encylopaedia Rrittanica" or "Chamber's Encyclopaedia," and the details of his career are unknown to
most physical scientists, though net to physiologists.
This paper is not intended as a biographical memoir
but rather as a review of the ideas which led Fick to
his conception of the basic diffusion law (Grundgesetz)
and an assessment of its present importance.
Fick, the youngest of nine children, showed a strong
mathematical bent a t school in Cassel, where he
graduated from the gymnasium in 1847. He proceeded
to the University of Marburg where his eldest brother

Ad0lf Fick

who applied the concepts and methods of physics to the

study of living organisnls, and thereby laid the foundations of modern physiology. He was graduated
as a doctor of medicine from Marburg in 1851, and
early in 1852 was invited by his friend and teacher
from earlier days in hlarburg, Carl Ludwig, then Professor of Anatomy and Physiology in Ziirich, to take
up a post as Assistant (Prosektor) in the Anatomy Dcpartment there. The contact with Ludwig was responsible for Fick's interest in diffusion, since Ludwig,
in common with several other physiologists of the
period, had been interested in diffusion through membranes for some years previously.
Fick's principal paper on diffusion entitled "Ueber
Diffusion" appeared in Poggendorff's Annalen in 1856
(94, p. 59). It was clearly intended t o be read by
physical scientists. Following a not uncommon practice, he comnlunicated the same results in a modiied
form to the Zeitschrift fur rationelle Medicin (6, 28%
(1955)) (2) in order to bring his ideas to the attention
of physiologists and medical men generally. The paper
in A n n a l a begins with the remark that hydrodiffusion
Volume 41, Number 7, July 1964

397

through membranes is an important physiological

problem which should also be of great interest to the
physicist. After quoting several references to earlier
work in this field, Fick stated that he had himself carried out experiments of this kind but with limited success. Smce it was not possible to carry the work further he proposed t o describe the results as far as they
went, and to draw attention to the analogy, which had
formerly been neglected, between diffusion through
membranes and the "simple spreading of a soluble substance in its solvent." The quantitative aspects of
Graham's work were strongly criticized and Fick went
on to announce his intention of discovering the basic
law governing the transfer of material from one layer
of solvent to the next. To do this he began by devising
a model to be used in describing the state of matter
taken up by a single substance and the process of the
diffusion of one substance into another. It was conceived in terns of attractive and repulsive forces between particles. A single substance was assumed to
be made up of two kinds of "atoms,"' "ponderable
atoms" attracting one another by gravitational forces,
and "aether atoms" which repelled one another with a
force proportional to the product of their masses and
to an inverse power of the distance r between them
which was greater than 2, indicated by f(r). Likewise,
the two types of "atoms" were assumed to attract one
another with a force, also proportional t o the product of
their masses, which fell off less rapidly with increasing
separation than f(r). This distance function was
termed ~ ( r ) . According to Fick, a model of this kind
leads to a system in which each "ponderable atom''
acts as the center of a sphere of "aether atoms," the
density of the latter falling off as the distance from the
central "atom" increases. Such aggregates he termed
"molecules." Diffusion is discussed in terms of two
assemblies, one of "molecules" of species A, and the
other of species B, placed side by side. Assuming that
the attraction between "n~olecules" of unlike species is
greater than that between those of like species, "molecules" of species A will be drawn into the region formerly
occupied by those of B and vice versa, until the "molecules" of each species are distributed uniformly over
the whole region. The model can be used to explain why
"molecules" of A and B do not coalesce entirely; the
gravitational attraction of the two species of "ponderable" atoms is balanced before this can occur by the increasing repulsive forces between the surrounding
spherical shells of "aether atoms." The most important aspects of this to modern eyes are the very
clear recognition of the part played by molecular motion
(using molecule in the modern sense) in diffusion processes, and the use of a model in which the macroscopic
properties of matter depend upon the balance of attractive and repulsive forces between microscopic
particles, an idea which is of central importance in all
modern theoretical treatments of similar problems. It
is of considerable historical interest to note that, a t
this date, Fick found it necessary to introduce a special hypothesis to explain the random distribution
of A and B "molecules" which was the final result of the
diffusion process. Graham was faced with the same

' Where the terms "atom" and "molecule" are used in a sense
other than a modern one, they areput hetween quotation marks.
398

Journol o f Chemical Education

problem and solved it in the same way, rather than by

applying the essentially correct explanation given by
Dalton of the interdiiusion of two gases (that the process was akin to the expansion of a gas into a vacuum)
to the case of liquid diffusion, though he mentions this
possibility in afootnote. The revolution brought about
by the kinetic theory of gases can be seen from the
standpoint taken up by Fick in a series of six popular
lectures "Die Naturkriifte in ihrer Wechselsbeziehung"
published in 1869, where his views on this subject are
essentially those held today (2).
Fick's reasoning can best be followed from this point
by direct quotation:
The first task would now he ta establish the basic law far this
mode of motion [i.e., of diffusion] from the general laws of motion, and this should, I hope, d m be possible without knowing
the functions f(r) and q(r). My attempts to do this have however had no success. On the other hand, when first considering
this hasic law, it occurred to me that a very similar supposition
has, in my opinion, been fully proved hy experiment. In fact, it
might be added, that from the outset, nothing is more probable
than this; the spreading of a. dissolved suhstance in a solvent,
provided that it is not affected by the exceptional influence of
molecular farce^,^ takes place according to the same law which
Fourier has suggested for the spreading of heat in a conductor,
and which Ohm has already applied to the spreading of electricity, where it is, of course, not quite correct.' It is only necessary to replace, in Fourier's law, the words "quantity of heat"
with the words "quantity of dissolved substance," and the word
"temperature" by "solution density." The conductivity, in our
case, corresponds ta a constant dependent on the affinity (verwmdsehaft) of the two substances.

Fick thus obtained the "Grundgesetz" for the diffusion process solely by analogy with the laws of Fourier
and Ohm, and could give no a priori reason for the
choice of the solution density gradient as the driving
force for diffusion. I n applying this law to real systems, he specified that the volume change on mixing
the diffusing solutions must be ignored; this proviso
enabled him to pass from the solution density gradient
to the concentration gradient as the driving force, concentration being expressed as weight units per unit
volume. He also pointed out that the density gradient had to be of such a kind that the heavier layers
of liquid lay below the lighter ones.
The mathematical development of the basic law proceeded as follows:
A quantity nf salt will he transferred into the elementary layer
between the planes z and z d z (where the concentration is y)
from that between x
d z and z
2 d z (wbere the concentration
d y l d z . d z ) equal to -Q k d y l d z dt, where Q is the crossis y
sectional area of the layer, and k s. constant dependent on the
nature of the substances. Naturally, an amount of water equal
in volume to this amount of salt enters from the upper into the
lower layer. Exactly by the method used for the development of
Fourier's law oneobtains from this basic law of diffusion flow, the
differential equation,

ay/st = - k(a2 y / a r z

+ 1/Q d ~ / d zs y / a z )

(3)

where Q is aasumed to be n function of the height of the plane

This proviso, which Fick did not discuss further, could be

stretched to cover all deviations from the hssic law. It is ao
general, however, that it can he regarded on Fick's part as seientifie caution rather than as exceptional prescience.
=Fick gives no indication of what he has in mind here. In
fact, Ohm's law is obeyed more exactly than Fourier's law in the
sense that the electrical conductivity of a substance is less dependent on the potentid gradient than is the thermal conductivity
on the temperature gradient.

above the base. If this is constant, the differential equation

becomes,

Fick's notation is here retained without any modification.

The equation now referred to as Fick's first law has
the form (1); the "basic law" referred to by Fick as a
modification of Fourier's law, is not exactly this, and
was never written down explicitly in the Annalen
paper. Equation (4) would now he called Fick's
second law for the restricted case where the diffusion
coefficient, written ask by Fick, is independent of concentration. Further points of particular interest are
that Fick was fully aware of the mutual nature of the
diffusion process, migration of one component in one
direction being balanced by the migration of an equal
volume of the other component in the opposite direction. The plane of reference, with respect to which the
(one-dimensional) flows are measured, while not
specifically discussed, is obviously one across which no
net volume transfer occurs. If there is no change in
volume on mixing as Fick assumed, this reference plane
will also he fixed with respect to the containing vessel.
This is probably the most useful way in which to define a diffusion coefficient, and Fick's contribution in
this respect can hardly he overrated. However, it
will be apparent that the detailed molecular model
which was to provide the basic law contributed nothing to its development, and the only real justification
for equations (3) and (4) would he a demonstration
that the quantity k (i.e., in modern terms, the diffusion
coefficient) is, in fact, independent of the concentration
gradient. This was the next task t o which Fick turned
his attention.
The first possibility considered was that of integrating equation (4) subject to appropriate boundary conditions in order to find the concentration y as a function
of position x and time, a procedure which is the basis
of ahnost all methods in common use a t the present
time. Fick rejected this possibility because of the
computational labor involved if the law were to be
tested adequately, even for those cases where an integral solution could be obtained in closed form;
this was the reason for not presenting any such solutions. A second method, found to be unsatisfactory,
was to establish a diffusion boundary in a tall cylindrical vessel by introducing concentrated sodium chloride
solution beneath a water layer, allowing the diiusion
process to continue for a definite period, then sampling
layers a t different levels in the liquid column, and analyzing. Ay/Ax, and A2x/Ay2 could in principle be
obtained from these measurements, but the experimental errors were such that their ratio (k) was variable. However, the qualitative change of concentration
with the coordinate x was that expected. A more successful technique was then devised. A column was
set up with sodium chloride crystals a t the bottom and
pure water, constantly renewed, a t the top. The concentration distribution along the column was allowed
to become that characteristic of a time-independent state
in which the flow of solute (and, in the reverse direction, of solvent) was the same across any horizontal
crosssection irrespective of the value of x. The sys-

tem is described more completely in the paper in the

Zeitschrift jfir ratimelle Medicin; and there Fick used
the terms, still in current use, "stationary state" and
"dynamic equilibrium" to describe conditions within
the column. The amount of solute diffusing out a t the
top of the column when the stationary state had been
attained was measured, and the concentration gradient
taken to he the ratio of the solubility of the salt to the
height of the column. This is only true if the diffusion
coefficient is independent of the concentration, and Fick's
"constants" are therefore averaged diffusion coefficients.
He does not seem to have considered the possibility
that the "constant" might vary in this way possibly
because this would not strictly he consistent with his
"basic law of Diffusion." Three columns of different
heights were used, and the coefficients derived from
experiments on each were considered t o be sufficiently
concordant for the diffusion law to be taken as experimentally established. Other experiments were done
with a cone-shapedcolumninorder toestablish thevalidity of an equation related to equation (2). The table
shows some of his results on columns of uniform cross
section, recalculated to modern units; some accurate
differential diffusion coefficients for sodium chloride in
water are also shown. It is worth mentioning that an
improved form of this method was used over fifty years
later by Clack (3) who measured the local concentration
gradient a t successive horizontal planes in diffusion
columns which had been allowed to reach the steady
state, and corrected for the solvent counterflow;
for many years they provided the best available diffusion data for electrolytes in water.
Fick's Average Diffusion Coefficients for Aqueous Sodium
Chlorides
colunk

Medium
column

Short
column

T ("c)

( x 109

( x lo5)

( x 105)

14.8-15.8
20
2W21

1.12
1.29
1.37

1.12
...

1 .07
1.27
1 .29

Tmw

...
Modern differential coefficients ( 2 5 W ( X 10-4'
0.5 M
2.0 M
4.0 M

1.474
1.514
1.58a

Values are recalculated to cm' seo-lunits. A value of 1.50.X

lo-' at

10-6 at 25' would correspond approximately to 1.34 X

20' and 1.20 X 10"at l 5 T .
1 STOKE%
R. H.,J . Am. Chem. Soc., 72, 2243 (1950).

Although Fick stated that his results had confirmed

the basic law, subsequent work has shown that, as in
the case of so many simple physical "hws," Fick's
"Grnndgesetz" is not correct in the general sense that its
author clearly hoped that it would he. His views did
not escape criticism a t the t i e . Beilstein (4) suggested that the effect of varying the concentration had
not been adequately investigated, and that there was
no more reason for choosing the first power of the concentration gradient as the driving force rather than some
power of it other than this. I n particular, Beilstein
proposed that the density gradient in Fick's formulation should be replaced by (density)"'/x, a curious suggestion later treated with some scorn by Fick (6).
He attacked the accuracy of Fick's experimental
Volume 41, Number 7, July 1964

399

data, and did many quantitative messurements designed to support Graham's conclusion (1) that the
amount of salt transferred into distilled water in unit
time was proportional to the weight of salt contained
in the solution from which the transfer took place.
Fick's reply to these criticisms forms the substance of
his third, and last, paper on diusion (6) and he claimed
that since, in his own experiments, the concentration
varied along the column from saturated solution a t the
base to pure solvent a t the top, the effect of concentration changes were automatically taken into account;
Fick does not seem to have realized that his assumption
about the constancy of the concentration gradient
in the colun~nwas the weak point in this argument.
As for the suggestion that the concentration gradient
was involved to some other power than the first, he
claimed that the observed constancy of his experimental
coefficients would not have been found if any other
assumption than this had been used. Later work has
shown that the diffusion coefficient usually varies appreciably with concentration, and Nernst (6) took the
gradient of osmotic pressure as the driving force for
diffusion. This is equivalent to taking this force as the
gradient of chemical potential, proposed independently
by Gibbs (7) (1899) and later by Schreiner (8) (1922)
and by Hartley (9) (1931) to whom is due the main
credit for the specific application of this conception t o
the diusion problem. This conclusion has been
confirmed by the application of steady-state thermodynamics to the problem (10). For dilute electrolyte
solutions the correction introduced into the original
diusion law4 suggests that D ( l
b In y/b in m ) ~ ;
should be independent of concentration and be equal to
the theoretical limiting Nernst value ( 6 ) which can be
calculated from the mobilities of the constituent ions.
Experiment has shown that this quotient is not quite
independent of concentration but does tend to the
limiting value a t low dilutions. The observed deviation
is due to neglect of certain interionic attraction terms,
as can be seen clearly from the thermodynamic treatment (11). These can be calculated successfully for
I : l electrolytes by a kinetic thecry due to Onsager and
Fuoss (10) which is, however, rather less successfnl for
other classes of electrolytes and breaks down for concentrated solutions. While detailed evidence is rather scanty as yet, for certain classes of nonelectrolyte
pairs insertion of the thermodynamic term does give
a quotient which is constant within experimental error.
For others. D (1
b in f/b 1nN) $; varies much
more with changing concentration than does D itself (18). Qualitatively, such variations are often
attributed to the formation of complex species in the
solution, each with its own characteristic diffusion coefficients (18), but ideas of this kind have not yet led to

7 is the activity coefficient on the molality scale, and f that

on the mole fraction scale.

400

Journal of Chemical Educaiion

quantitative developments. Indeed, a fully sahfactory explanation of the observed values of diusion
coefficient and of their variation with concentration
would not appear to be an immediate prospect.
More than a century of experience has shown that
the value of Fick's contribution to the study of diusion
in liquids, and also in its later application to gases, lies
preeminently in the stimulus i t gave, and is still giving,
t o accurate experimental work, and in the provision of
a concise and easily appreciated form for the expression
of experimental data. A glance a t Graham's extensive,
and almost unreadable, descriptions of quantitative
studies on diffusion, will show how great a contribution
this was. The original concept has been extended to
cover the phenomenon of self-diffusion, and, for a
binary mixture, a full description of the transport of
matter within the system requires three diffusion
coefficients, one for inter-diiusion of the two components, and two self-diffusion coefficients. Fick's hope
that diffusion coefficients could be calculated from considerations of the repulsive and attractive forces between molecules of the same, and of different, species
bore no fruit a t the time or for many years afterwards.
Hydrodynamic theories, like the Stokes-Einstein theory
held, and in many ways still hold, the field; but increasing attention is being paid to the statistical mechanical
description of transport processes in liquids (14) which,
it can reasonably he hoped, will eventually be successful
in realizing the ambitious program set out by Fick in
1855.
Literature Cited
(1) GRAHAM.
T.. Phil. Trans. Rou. Soc. L a d n . 140. 1. 805
(1850);. 141,483 (1851); is:, 183 (1861); A m . c&. 77,
58, 129 (1851); 80, 197 (1851).
FICK,A,, "Gesammelte Schriften," Stahelverlag, Wiirshurg,
1903-04, R. Fick, editor. (The papera referred to here
are reprinted in volume 1 of this four volume work. The
portrait here reproduced is also taken from this work.)
CLACK,
B., Proc. Phys. Soc., 21, 374 (1908); 24,40 (1912);
33. 259 11921).
27. 56 (1914): 29.. 49 (1916):
.
BEIL&&_
F R . : ' A ~ Chem..
~.
lbb. 165 (1856)
: h a . , 102; 97 ?1857j, (also reprinted in
(5j FICK,A., ~ n n C
reference (2)).
(6) NERNST,
W., Z. physik. Chernie, 2 , 613 (1888).
(7) "The Scientific Papers of J. Willard Gibb~," Longmars,
Green, snd Co. New York, 1906, p. 429. (Letter dated
1899, to Wilder D. Bancroft).
E., TidSskr. Kjemi, Bwgvesa Met., 2, 151
(8) SCHREINER,
(1922); ef. L m , O., S v m k Ka. Tidskr., 72, l(1960).
(9) HARTLEY,
GI S., Phil. Mag., (7) 12,473 (1931).
(10) ONSAGER,
L., AND FUOSS,R., J . Phys. Chem., 62,404 (1958).
H.J . V., "Diffusion and Heat Flow in Liquids,"
(11 ) TYRRELL,
Butterworth, London, 1961, chsp. 4.
D.K., HALL,J. R.,AND BABB,A. L., J . Phys.
(12) ANDERSON,
Chem., 62, 404 (1958).
A. W., AND IRANI,R., J. Phys. Chem., 64, 199
(13) ADAMSON,
11960).
(14) BEARMA,R. J., J. Phys. Chem., 65, 1961 (1961); RICE,
H. L., Ann. Rev. Phys. Chem., 11, 187
S. A,, AND FRISCH,
F. C., AND RAPFEL,H., Advan. Chem.
(1962); COLLINS,
Phys., 1,135 (1958).