VI.

OPERATORS & EXPECTATION VALUES

1. The Calculation of Average Values

The description of quantum states in terms of probability
amplitudes at once implies the possibility of using these probabi-
lity amplitudes to calculate the average values of various quanti-
ties for systems that are in a a specified state. Consider, for

example, a bound state of a particle in a one-dimensional poten-

tial. Such a state has a unique value of the energy,'but there
is a continuous distribution of the possible values of the position,
x. The probability amplitude associated with a particular value of
x is given by 4>(x). The probability that the particle will be
l to !j^(x)[I 2 dx, provided
found between x and x + dx is then equa

that tJKx) is normalized in the usual way:

X - I (6.1)

If we ask "What is the average position of a particle described

by iKx)?"- w^ can supply the answer by evaluating a weighted mean
value, just as in classical probability calculations. We multiply

each particular value of x by the probability of having that

value, and sum (integrate) over all possibilities:

* ,

Such averages are called expectation values in quantum mechanics.

This name vividly expresses the essentially statistical nature o:5

; -1-
 -2-

the quantum-mechanical description. One can imagine a huge

number of identical systems, each consisting of a particle in

the state described by ip(x). If one carried out a set of ex-
periments to determine the location of each of these particles ,
the results would be distributed in accordance with the relative
2
1
The result of an indivi-

dual observation is not predictable, but the mean of all the

measurements is well defined. It is customary to denote the ex-
pectation value of any quantity Q by enclosing it within angular
brackets ; thus we put
Expectation value of x

(6.2)

To take a specific example, consider any one bound state

of a particle in a "violin-string" state:

VI
The expectation value of x for this state is then given by
L
? C ' 2- / Iffi
t J X Sm
O
It is easy to verify that the result of this calculation is,that

(x^ is equal to L/2 for any n. Notice that the expectation

value of a quantity is in principle not at all the same thing as

the most probable value. For example, if we are concerned with
measurements of position, the most probable value of x is the value
 -3-

at which J«y(x) ( is greatesti If we consider the violin-string

states u/(x), then for n = 1 the expectation value and the
most probable value do coincide, but for n = 2 the value of
is zero; thus a particle in this state would never be found
at a value of precisely equal to C^ .
x
*N
Given that I MX) j dx is the probability associated with
a particular value of x , we can use it to calculate the expecta-
tion value of any function of x .. We simply have

<fM> -
Thus we can, for example, find the mean squared value of x2

2. Expectation Values involving Operators

Suppose that we again consider any individual energy-state
of a particle in a one-dimensional box, and this time ask what is
the expectation value of the linear momentum p j^ . A consideration

of the symmetry of the situation suggests, without the need for

any calculation, that the answer is zero; the particle is equally
likely to be found moving to right or left. We can spell this

out more formally by noting that the wave function can be written
as a superposition, with equal weights, of plane-wave amplitudes
belonging to equal and opposite values of the wave-number k :

m kM x
I
- fit •
yv rt
- ne/\

Thus k = + k n , and the momentum p*x (=ftk) is equal to + ^k n

^
This makes it clear that the probability associated with each
 4

value of the two values of p is 1/2, and so we have

<px> =
[Notice, again, the distinction between expectation value and
most probable value. Individual measurements on this state would
never yield a value of k equal to zero, but only + k .]
The above calculation is very simple and straightforward,
but this is mainly because the situation itself is very simple.
What we shall now do is to introduce a method of using the
Schrfidinger amplitude itself to calculate ^p > and related
quantities. With its help we can then handle problems that are
far less obvious than the one we are now using as an example. The
starting point is the following very important concept:
In quantum mechanics we can associate various
dynamical quantities with particular mathematical
operators .
This is a rather general statement, but we shall at once make it
explicit in connection with linear momentum. We have seen that
the probability amplitude for a free particle of given momentum
v is of the form /1
t(k*~
- ft e
where k = p /fi. Thus we can put

or

We then propose that, in general, the operator *~t*\C0fbty r operating

 -5-

on a Schrodinger amplitude ip , is equivalent to multiplying

by p . Thus we put
fi
(6 - 3)

We then proceed to evaluate the expectation value of p by means

of the following equation:

< (6.4)

If, as in any one of the bound states of a particle in a box,

the SchrOdinger amplitude $ embodies contributions from more than
one value of p , the mathematical operator corresponding to '
p ,
«V
*k

applied to J , yields all these component momenta and the integral

in eq. (6.4) is an automatic prescription for forming the appro-
priate weighted average. It is important to observe the order of
the various factors in this integral. Now that p j^> is represented
by a differential operator/ applying to anything that follows it,
we must place the factor $* to its left so as to make clear that
it is not subject to this operation.
Suppose once again that ip represents a particular energy-
state of a particle in a one-dimensional box:

L
Then 2.

o
 -6-

But the functions sin k x and cos k x are orthogonal (i.e.,

0 Jj sin k n x cos k n x dx = 0) , so this calculation gives

^ vp
N ^x / = 0
as we know it must. However, a calculation of p by the
same technique gives a non-zero result. The operator corre-
spending to p jf± is -# ( d /^x ) , and so for this same state
we have ^ Nr
2 w 2 'f (sin
' fv\
knx)[^/sin kRx) dx
2 fL 2
TL n Jf sin k n x dx
0
/ \. ty,
i.e., <px > = V kn 2

This is just what we expect, since the (kinetic) energy E of

the state will be given by

F <P 2 > n
n 2m 2m

With k = nTT/L, this gives us the familiar result

2 2
n h
n 8mL2

We could, of course, have obtained this expectation value of p wrv

much more readily by going back to the description of the state
as a. 50:50 mixture of the two momentum components Ik . On
this basis we have simply

<PX2 > = K 2 <k2 > = # 2 [ i(kn ) 2 + i(-k n ) 2] =

Thus, up to this point, our use of the momentum operator technique

has been like the proverbial sledge-hcmmer to kill a fly. But let
us now take an example where it is not. so trivial.
 -7-

Consider a particle in
one of the characteristic
energy states in a harmonic-
oscillator potential. Now
the value of p , as defined x-0
by + /2m[E - V(x)J , takes on an infinite number of different
values (including imaginary ones) . Given a knowledge of 1|/, we
can however evaluate with the help of eq. (6.4). Suppose
that the state in question is the lowest state of the oscillator
The spatial factor in its wave function is of the form
2/0 2
= A e -x /2a (6.5a)

where a 2 =s mC
[cf. Notes, Ch. Ill, pp. 19 ff.]
The normalization of \|/ leads to the condition

A2 = (6.5b)
a ATT
The value- of ) is given by

<PX> - -inK A*2 JC e -* //2a d_ dx

= %- j» "CO
-x 2~,/a 2 dx

Since the integrand is the product of an odd function (x itself)

with an even function, the integral between the limits + oo is
zero. This is another result that could have been expected.
 -8-

y 2 v
More interesting, perhaps, is the calculation of \p * > . For
this we have
00 2 2 2 22
v /Oa
— X//:d inl
U —v /Oa
,^A/Zcl x ^^
<px2 > = -*2A2 f
dx
A' y 2 ,
^2A2 f -x 2/a 2o e -x 2 /a
e ' dx - 4
x 7 dx
—(30

2a

Substituting the value of A2 from eq.(6.5b), this gives us

<PX2 > = - (6.6)

2a

The calculation of expectation values of the energy E can

likewise be based on the use of an equivalent operator. Referring
again in the first instance to a free-particle wave function, we have

= TV i (kx-Et/fi)
Ae N '
and hence

or E <^-> ih ^ (6.7)
^t
If, now, we take a wave-function that represents a superposition
of different energy-states, we have

e-^n*/*
ifl

f /f \/ _ •
Then E = C 7' ' E
 -9-

If the different (| f s are orthogonal and normalized, this

reduces to the equation

< E>= rA n
Thus/ again as one might expect/ the value of ^E^ is just a
weighted average of the energies of the component states , with
weighting factors equal to the integrated probabilities associated
with the individual components .
In some circumstances it may be of interest to calculate
the mean values of the kinetic energy and the potential energy
separately for a particle in a given state. The ways of evaluat-
ing these quantities are essentially contained in what we have
already done. The operator for the kinetic energy K is the
2
operator corresponding to p /2m ; thus for one-dimensional
problems we put

K ~~ (6-8a)
and for three dimensions we have

K <r~> -_ (6.8b)

In a one-dimensional system we therefore have

For the mean value of the potential energy/ we simply use the
procedure that applies to an arbitrary function of position, so that

V(x) I//" <^x (6.10)

 -10

If we consider the (one-dimensional) Schrodinger equation as

it applies to an individual energy-state, we have

dx
-*
Multiplying by \^ and integrating over all x then leads , with
the help of eqs. (6.9) and (6.10), to the reasonable result

<K> + <V> = E (6.11)

3. Variances and Uncertainty Relations

In statistical analysis , when one is dealing with a quantity
that has a certain distribution of values, an automatic measure of
the spread of the distribution around its mean value is provided
by the standard deviation, CT . The square of CT , known as the
variance, is the mean squared deviation of the individual values
from the mean. Thus for some arbitrary quantity Q , we have

Variance = -CT 2 = (Q - <Q> ) 2

>C ClV 4

But (Q - <Q> ) 2 = Q 2 - 2Q <Q> + <Q> 2 ,

and the variance is the expectation value of this expression.

This gives
*

<rQ2 = <Q2 > - 2 <Q><Q> 4- <Q> 2

tf

i.e. ,
<TQ2 = <Q2 > - <Q> 2 (6.12)
t

Only if Q
is limited to a single sharp value does the variance
drop to zero; in all other cases the difference between 2

and Q is positive.
 -11-

We can use this measure of the width of a probability

distribution to examine the uncertainty relations in more
specific terms. Consider, for example, the lowest state of the
harmonic oscillator, as described by eqs.(6.5). For this we have
= o

fir a3 ^
2,

Thus CT 2

Now in Section 2 we calculated the values of p J\. > and

the results were

<PX> =o
,<PV 2,> =
-h2
—2
2 a'
h2
Hence 2 a2

It follows, then, that the following relation holds:

x Px = ^ (6.13)

This uncertainty product is thus independent of the parameter a

that characteri2es the width of the position probability distribu-
tion. It happers that the value fc/2 , characteristic of these
 -12-

Gaussian error-function probability distributions, represents

the smallest achievable value of Ap « /\ x , where A p and Ax

are defined as the standard deviations of the probability distri-

butions in question.

[4. The Quantum-Mechanical Equivalent of Newton's Law

An interesting application of the calculation of expecta-
tion values is the formulation of an equation that parallels the
basic law of motion in classical dynamics. We begin with the

expectation value of p« £ as defined in eq. (6.4) , and then con-

sider its time derivative:

Now with the help of the time-dependent Schrfldinger equation we

can convert the integrands on the right into forms that do not
explicitly involve t: .

it
w
*
 -13-

Hence

at

-JL fi
2.w J^

For any reasonable wave- function, the values of \ and its

derivatives vanish at x = + cx> Thus we are left with the
result

Since the negative gradient of the potential energy is equal to

the force derived from that potential, eg. (6.14) corresponds to a
statement of F = dp/dt in terms of the expectation values of these
quantities.
If one wants to carry the analysis a stage further back,
one can show by similar methods that the following result also holds:

m ft < x> = <' px>

 -14-

5. Expectation Values in Superpositions of States

If we have a superposition of two different energy-states
of a quantum-mechanical system, there is a harmonic time-dependence
of the expectation values of position and related quantities.
Suppose that the state in question is given by .

+
Then

2-f./il/,^x) VMd*} cos cot

where ^0 = (E 2 - Ej)/K

Each of the three integrals in the above expression is a definite

integral representing a quantity of the dimension of length. Thus
the equation for ^x> can be written in the simplified form

<x> = A + B cos 60 1 (6.16)

If the wave-function in question describes an electrically charged

particle , one can see here the basis of a picture of an oscillat-
ing electric dipole, with the implied possibility of associated v
radiation characterized by the frequency 0) .