2

The Motion of Wave Packets: Fourier

Analysis

Because we will need to work with wave packets of finite extent, it will be very
useful to first give a brief review of Fourier analysis.

A Fourier Series
We shall start by studying periodic functions of infinite extent in space. First con-
sider periodic functions f (x) with a periodicity interval 2rr, such that f (x + 2rr) =
f (x). For real functions f (x), we usually use Fourier expansions in cosine and sine
functions. For the complex functions of quantum theory, it will be advantageous
to use a Fourier expansion in exponential functions.
1. Fourier Expansion:

L
00

f(x) = ane inx , (1)

n=-oo

where we will exploit the orthogonality of the exponential functions.

2. Orthogonality:

1 7f

-7f
dx'ei(n-m)x' = 2rrlinm , (2)

which is expressed in terms of the usual Kronecker delta. With this orthogonality
relation, the expansion coefficients, all, can be determined via the Fourier inversion
theorem. If we multiply f(x) by the complex conjugate of a specific exponential,
say, e- imx , with some specific, fixed m, and integrate both sides of the resultant

K. T. Hecht, Quantum Mechanics

© Springer-Verlag New York, Inc. 2000
 A Fourier Series 9

equation over the periodicity interval, say, from -T( to +T(, the orthogonality
property will pick out one specific am, with value given by the Fourier coefficients.

1" .,
3. Fourier coefficients:

am = - 1 dx' f(x')e- 1mx . (3)

2T( _IT

Substituting this coefficient back into the Fourier expansion, we get the
4. Fourier expression for f(x):

f(x) = L -1 1" dx' f(x')ein(x-x'l.

DO (4)
n=-DO 2T( -IT

It will be convenient to introduce orthonormal functions, tPn (x),

1 .
tPn(x) = ~e'nx. (5)
y2T(
The four basic Fourier equations can then be rewritten as

L
DO

f(x) = bnrPn(x), (6)

n=-oo

(7)

bn = i: dx' f(x')rP:(x'), (8)

f(x) = nf;DO i: dx' f(x')rPn(x)tP:(x'). (9)

Finally, it will be convenient to use a periodicity interval of length (2/), where I

has the dimension of a length, where now f(x + 2/) = f(x) and the orthonormal
functions can be expressed as
1
- e-r .
,.fj]
The four basic Fourier equations can then be rewritten as

f() ~ 1 "!'U
(10)
X = ~ Cn Mje I ,
n=-DO y 21

-
11+/ , i(n-m)""
dx e I
,= 8 nm , (11)
21 _/

Cn =
1
Mj
1+/ , , .", ,'
dx f(x )e-'-r , (12)
y21 -/
10 2. The Motion of Wave Packets: Fourier Analysis

f(x) = ;/ ntool:l dx' f(x')e i T(x-x'). (13)

It will now be useful to introduce the wavenumber, k n

nrr 2rr II
kn = - = -; with An = -, (14)
I An n
so ¢n(x) = Jueik"X. This relation will be particularly useful in making the tran-
sition from the Fourier series to the Fourier integral for a wave packet of finite
extent.

B F01.l!-ier Integrals
Now suppose the repeating function, with periodicity interval (21), has the form of
a wave packetofextent~ a, with a < [, which repeats from -00 to +00, as shown
in Fig. 2.1. Now, suppose we let I ~ 00, keeping the wave packet unchanged,
with a fixed. Then, by taking the limit I ~ 00, provided f(x) ~ 0 sufficiently
rapidly as x ~ ±oo, we can make the transition from a periodic function to a
nonperiodic one, i.e., a transition from an infinite wave train to a wave packet of
finite extent in space. As I ~ 00, the spectrum of possible kn goes from a discrete
spectrum to a continuous one, because
rr
kn+! - kn = T ~ 0 as I ~ 00. (15)

Because the number of spectral terms in a k-space interval dk is (see Fig. 2.2)
dk dk
(interval between successive kn ) rr / [ ,
the discrete sum over n in the Fourier series goes over to a continuous integral
00
L~
foo -.
dk
n=-oo -00 rr / I

1\ {\ 1\ {\ 1\ f\ 1\ {\ 1\ {\
-Sl V -3/-V -I \T +1 V +31 \T +SI

----~--V'~A~-----
-31 -1 +1 V +31

~---------l----~-----+l-------~
FIGURE 2.1. Periodic wave fonn, 1---* 00, a fixed.
 C The Dirac Delta Function 11

d dk I'>

11111111"111111111"111111111111"1111 I'> k

mt
/\ (n+l)1t
L L
FIGURE 2.2. The spectrum of k values, kn = mr / L. The number of spectral terms in the
dk interval = [dk/~J

Thus, the Fourier expression for f(x) becomes

f(x) = -1 -
21rr
11 1 00

-00
dk 00

-00
dx' f(x')eik(x-X'). (16)

We can then think of the Fourier development in terms of a Fourier amplitude

function, g(k), as

f(x) = ~1°O dkg(k)e ikx , (17)

v2rr -00

with amplitude function g(k), the so-called Fourier transform of f(x), given by

g(k) = -1-
../2i
1 00

-00
dx' f(x')e-''k x.' (18)

Note, however, the orthonormality integral becomes divergent when k = k',

_1 JOO dx' eix'(k-k') = 8(k _ k'). (19)

2rr -00
The Kronecker delta becomes a Dirac delta function.

C The Dirac Delta Function

If we rewrite the Fourier series in terms of a limit of a sum over a finite number of
terms,

f(x) = lim j 7r

dx' f(x')
+N
L ¢n(x)¢;(x'); (20)
N-"oo -7r n=-N

or, similarly, if we rewrite the Fourier integral as

f(x) = lim J OO dx' f(x')-

1 jk O
dkeik(x-x'), (21)
ko-"oo -00 2n: -ko
12 2. The Motion of Wave Packets: Fourier Analysis

the function

K(x, x') =
+N
L
n=-N
¢n(X)¢Z(x') or K(x, x') = - 1
2rr
l+
-k o
ko
dkeik(x-x') (22)

becomes, in the limit of large N or large k o, a function strongly peaked at x = x'

with oscillations of very small amplitude for x f: x'. Keeping in mind that the
real limiting processes should be those expressed by eqs. (20) and (21), physicists
blithely interchange the infinite sum or the infinte k-integral with the x' -integral,
through the definition of the Dirac delta "function"

L
OX)

¢n(X)¢Z(x') = 8(x - x')

_1_/
11=-00

dkeik(x-x') = 8(x - x'),

00
(23)
2rr -00

where the Dirac delta "function" is not at all a function in the mathematician's
sense. It is what mathematicians call a "distribution" (see, e.g., an appendix in Vol.
I of the books by Messiah). The Dirac delta function "picks out" the value x' = x
for the function being integrated. It has meaning only through the integrals. By
itself, it diverges at the value x' = x. The Dirac delta function is defined through
the following properties:

8(x - x') =0 for x' f: x. (24)

For x' = x, the Dirac delta function becomes 00 in such a way that

I ab dx'8(x - x') = 1, if x' = x is in the interval (a, b), (25)

f:
and

dx' f(x')8(x - x') = f(x). (26)

Our limiting process, given through eq. (21), e.g., would give

~(
o X - X
') l'1m - 1
= ko-+oo 2rr
/k -ko
O
dk ik(x-x')
e I'1m sin ko(x - x')
= ko---+oc rr(x - x')
(27)

See Fig. 2.3 for a plot of this diffraction-like peaked function for finite ko. This
representation of the Dirac delta function is not, however, unique. Another example
(of the infinite number of possibilities) would be

,IE
8(x - x) = -lim----::--::__ (28)
rr E--->O [(x - x'? + E2]
 D Properties of the Dirac Delta Function 13

x-x'

x'=x
FIGURE 2.3. The function Sin[~o(x-;;')l.
Ir x-.(

D Properties of the Dirac Delta Function

The Dirac delta function is an even function of its argument
8( -x) = 8(x). (29)
Other properties, such as
d
x-8(x) = -8(x), (30)
dx
follow by integration by parts, because delta function relations have meaning only
through their applications within integrals

ib dxx8'(x) =
[]b
x8(x) a - ib dx8(x) =- ib dx8(x). (31)
14 2. The Motion of Wave Packets: Fourier Analysis

If a is a real number,
I
a(ax) = ~a(x). (32)
lal
Note the absolute value sign follows from

f oo
-00 dxa(ax) = ~
1100 -00 d(ax)a(ax) = ±~
1100 -00 dx'a(x'), (33)

°
where the upper sign applies for a > and the lower sign applies for a < 0,
because the change of variable ax = x' interchanges the limits in this latter case.
If the variable in the delta function is itself a function of x,
1
8(¢(x» = L-d-tP-8(x - x n ), (34)
n l(dx)x,,1
where the Xn are the zeros of the function, ¢(x). and the sum is a sum over all such
zeros. As a very specific example.
1
8(x 2 - a2) = -[/S(x - a) + a(x + a)]. (35)
21al

E Fourier Integrals in Three Dimensions

It is straightforward to generalize the Fourier series and Fourier integrals to
functions in ourthree-dimensional (3-D) space, f(x, y, z). For a 3-D wave packet,

f(x, y, z) = 13
-- foo dkx foo dky foo dk foo dx' foo dy'z
9rr) -00 -00 -00 -00 -00

x foo dz' f(x', y', z')ei[kx(x-X'Hk,(y-y')+k,(z-z')]. (36)

-00

It will be useful to introduce the following shorthand notation for this Fourier
integral expression

fer) = _1-/
(2rr)3
ik/ dr' fC;')eik.(r-r')
'
(37)

where

fer) = _1_3 / ikg(k)ei(k.r), (38)

(2rr)2

g(k) = _1_3 / dr' fer ')e-i(k-r'J. (39)

(2rr)2
(Note, in particular, the symbol, dr, when it follows an integral sign, is merely a
shorthand notation for dr := dxdydz and the single integral sign preceding dr is
shorthand for a triple integral over all of 3-D space.)
 G Wave Packets 15

F The Operation aax t

We note
1 a
-;--
l aX
fer) =
1
--3
(21r)i
f ~ ~
dkg(k)kxe'(
°k~~
or). (40)

Thus, we see, if g(k) is the Fourier transform of fCr). kg(k) is the Fourier transform
t
of ~ fCT). similarly, -(f. k)g(k) is the Fourier transform of V2 f(x, y, Z), and
soon.

G Wave Packets
A plane scalar wave propagating in the direction of the k vector can be given by
the scalar function
(41)

with constant amplitude, A, where the circular frequency, w, is in general related

to k through the dispersion law
w = f(k), or w = f(k), with k = Iki. (42)

where the latter is valid for an isotropic medium. Moreover, in a nondispersive

medium, in vacuum, e.g., w = ck.
To go from this infinite wave train to a wave packet of finite extent in space, we
need to form the wave packet from a superposition of amplitudes with different
k-values. For a 3-D wave packet,

(43)

To simplify the discussion, assume the wave packet proceeds in one dimension
only, say, the x-direction. Then,

1/I(x, t) = -1-
5
1 00

-00
dkA(k)ei(kx-wt). (44)

To use a very simple example, assume A(k) is different from zero only in an
4
interval, ko - 4!:1k :s k :s ko + !:1k, and moreover, assume A(k) has the constant
value, A, in this k-space interval. If the interval !:1k is not too large, we can expand
w(k) about ko, and retain only the dominant terms,

w(k) = w(ko) + (k - ko)(~~\ + . .. , (45)

and the wave function can be written as

1/1 (x , t) = __
5
A ei[kox-w(ko)t j lko + iLlk dkei(k-ko)[x-(~')otj
ko-tll.k
16 2. The Motion of Wave Packets: Fourier Analysis

~
! \.. Vpoup =(dw I dk)o
/
!
./
\

./
/

FIGURE 2.4. The wave packet of eq. (46).

= ;g .
_ Ae' [kox-w(ko)tl
n
sin(~k[x - (~~)ot])
_ _~_----; _ _ _,--
d •
[x - (d~)ot]
(46)

This wave packet is shown in Fig. 2.4. We note, in particular, the individual wavelets
travel with the phase velocity
w(ko)
Vphase =-,q;-' (47)

The wave train itself, the envelope of the packet, however, travels with the group
velocity

(48)

If we assume most of the energy of the wave train lies in the large central peak of the
wave envelope, we can take the extent of the wave packet to be ~x ~ 2 Eveni: .
for more sophisticated functions, A(k), we will find the Fourier integral analysis
always gives
~x~k ~ 2n, (49)

neglecting factors of order 2 in this approximation. This is the uncertainty relation

for a wave packet. Note, in particular, it follows for all wave packets, merely from
the Fourier analysis.
 H Propagation of Wave Packets: The Wave Equation 17

H Propagation of Wave Packets: The Wave Equation

The wave equation, the propagation law for the wave, is intimately related to the
dispersion law
w = f(k). (50)
In one dimension, with

l/F(x, t) = ~/ dkA(k)ei(kx-wr), (51)

- ~~l/F = _1_/dkWA(k)ei(kx-wt), (52)

I at ...tiii
( ~)n
I
~l/F =
ax n ...tiii
_1-/ dkkn A(k)ei(kx-wt). (53)

For functions f(k) that can be given by Taylor expansions,

f(k) = l:ankn,
n=O
we then have

f(~~)l/F
I ax
= _1-/
...tiii
dkf(k)A(k)ei(kx-wt). (54)

Eqs. (52) and (54) then lead to

[ -~~
at
I
- f(~~)Jl/F = _1_/ dk [w -
ax I ...tiii
f(k)]A(k)ei(kX--wr) = 0, (55)

so the dispersion law, W = f(k), leads to the wave equation

[ -~~at - f(~~)Jl/F
I ax = o. I
(56)

For the special case of a nondispersive medium, with W = ck, we would have
1 al/F - -
- -:-- c;al/F
- = / d k[w - ck]A(k)e i(kx-wt) =: O. (57)
I at I ax
So that, seemingly, the wave equation in this simple case of a nondispersive medium
becomes

(58)

This equation looks like a strange wave equation. however. Its solutions would be
l/F(x, t) = F(x -ct), where F is any arbitrary function. That is, this wave equation
would permit wave propagation only in the positive x-direction and, hence, would
correspond to a nonisotropic medium. The difficulty here is not with our method
of arriving at the wave equation, but that we have written the dispersion law in a
18 2. The Motion of Wave Packets: Fourier Analysis

way that builds in this anisotropy. For a nondispersive, isotropic medium, we have
to express the dispersion law in the form
(59)

or in three dimensions
(60)

The technique we have used to arrive at the wave equation would then give us
1 a2 1fr a2 1fr
----=0 (61)
c 2 at 2 ax 2
in one dimension, and

(62)

in three dimensions.
Note, finally, our method of arriving at the wave equation from the dispersion
law is not a derivation of the wave equation. Our method may also not give a unique
expression for the wave equation.