Visualizing the phonon wave function

Scott C. Johnsona) and Thomas D. Gutierrezb)

Department of Physics, University of California, Davis, California 95616
共Received 5 August 2001; accepted 6 December 2001兲
A phonon often is described as ‘‘a quantum of lattice vibration,’’ but this description can be difficult
to reconcile with the wave functions explored in a typical undergraduate quantum mechanics class.
A phonon wave function is similar to the harmonic oscillator wave functions studied in introductory
quantum mechanics, except that it is many-dimensional. We suggest a way to visualize the
probability density for this very high-dimensional wave function. The resulting pictures are
especially clear and intuitive for a coherent state, which is both a good approximation to a sound
wave and a discrete analog to laser light. These pictures can also provide a qualitative introduction
to quantum field theory. © 2002 American Association of Physics Teachers.
关DOI: 10.1119/1.1446858兴

I. INTRODUCTION We first do a classical analysis. The system is shown in

Fig. 1. Each mass is constrained to move vertically, and we
Phonons and quantum field theory are usually regarded as consider only small oscillations so that this system is a
graduate topics but undergraduate students often are curious coupled pair of one-dimensional 共1D兲 harmonic oscillators.
about them. The goal of this article is to give a semi- Vertical oscillations are chosen because they correspond di-
quantitative introduction to these topics at a level appropriate rectly to the axes of our plots. The position coordinates of the
for the final weeks of an introductory quantum mechanics masses are q 1 and q 2 . The momenta, which are directed
class. At this level, students are comfortable with one-
vertically along each coordinate, are denoted as p 1 and p 2 ,
dimensional wave functions and may have been introduced
respectively. The masses of the two particles are assumed to
to higher dimensional wave functions. This article takes this
background into account and presents a phonon or a quantum be identical, m, and the spring constants are also identical, ␬.
field as simply another wave function, albeit a wave function The key to analyzing this system and to understanding its
of many coordinates instead of the usual one, two, or three. It behavior is to use normal mode coordinates, denoted by the
is challenging to visualize a function in more than three di- upper-case letters Q 1 and Q 2 . For this system, they are
mensions, but we claim that some intuition can be gained by 1
looking at groups of one- and two-dimensional projections of Q 1⫽ 共 q 2 ⫹q 1 兲 , 共1a兲
such functions. A related series of pictures can be generated &
by the software in Ref. 1.
Understanding Fig. 18 is the key to our visualization tech- 1
Q 2⫽ 共 q 2 ⫺q 1 兲 . 共1b兲
nique. It shows a propagating coherent state of the phonon &
field. Most of this article up to this point describes the vari-
ous concepts and techniques used to generate this figure. The momentum coordinates conjugate to Q 1 and Q 2 are P 1
First, the analysis of a two-particle coupled harmonic oscil- and P 2 , and are related to the momenta of each mass by a
lator system is given using our visualization scheme, both similar set of equations,
classically and quantum mechanically. This system is consid-
ered to establish our notation and to illustrate the idea of 1
P 1⫽ 共 p 2⫹ p 1 兲, 共2a兲
using multiple projections of a function to visualize it. Then, &
an eight-particle lattice is analyzed in detail, emphasizing
some of the very interesting states of this system and culmi- 1
nating in Fig. 18. Some other interesting states of the eight- P 2⫽ 共 p 2⫺ p 1 兲. 共2b兲
oscillator system are then discussed and a comparison is &
made to a quantum field. In position coordinates the equations of motion are coupled
as indicated by a nondiagonal force matrix in the equation of
II. A TWO-PARTICLE SYSTEM motion,

A system of two coupled harmonic oscillators makes an

excellent system for demonstrating many phonon concepts
冋 2 ␬ /m
⫺ ␬ /m
⫺ ␬ /m q 1
2 ␬ /m q 2
册冋 册 冋 册
⫹m
q̈ 1
q̈ 2
⫽0. 共3兲
because this system can be thought of as a very small 共two- In normal mode coordinates, however, the equations of mo-
particle兲 lattice. In particular, it demonstrates normal mode tion are not coupled as indicated by the diagonal force ma-
decomposition and how the wave functions are separable, trix,

冋 册冋 册 冋 册
and hence much simpler, when expressed in normal mode
coordinates. Best of all, the probability density of the entire ␬ /m 0 Q1 Q̈ 1
wave function in position space can be fully visualized, so ⫹m ⫽0. 共4兲
0 3 ␬ /m Q 2 Q̈ 2
that we can develop some intuition about projections onto
the one-dimensional position or normal mode axes. An ex- This is the reason for introducing and using normal mode
cellent treatment of this system is given in Shankar.2 coordinates—each of the normal modes can be treated as if it

227 Am. J. Phys. 70 共3兲, March 2002 http://ojps.aip.org/ajp/ © 2002 American Association of Physics Teachers 227
Fig. 1. A two-particle system, showing the position coordinates q 1 and q 2 of
the two masses 共dots兲, which are constrained to move vertically, connected
by springs.

were an independent, one-dimensional harmonic oscillator.

Figure 2 shows three ways of plotting the configuration of
this system at any time. The pair of one-dimensional q 1 and
q 2 axes on the left simply gives the location of each of the
masses. 共The horizontal line crossing both axes indicates the
origin or equilibrium position.兲 The two-dimensional plot in
the center specifies the location of both masses with a single
point. The pair of plots on the right gives the projection onto
each of the normal mode axes. These three views will be
used consistently to visualize the two-oscillator system in Fig. 3. Plot of the system in each of its two eigenmodes 共a兲 and 共b兲. The
this section. amplitudes of the displacements are arbitrary. 共c兲 The system with one mass
It is helpful to look at a few examples of configurations displaced but the other at its equilibrium position.
plotted on these axes, as given in Fig. 3. In Fig. 3共a兲, for
example, the dot lies along the Q 1 axis and the system is in
one of its two normal modes. The projections onto q 1 and q 2
where ␻ 2 ⫽ ␬ /m. To make the transition from the classical to
show that in this normal mode, both masses are displaced
the quantum mechanical analysis of the same system, the
identically. The projections onto Q 1 and Q 2 show that the
position and momentum coordinates are changed to opera-
system is only in normal mode 1. In this mode, the dot on the
tors as indicated by the caret. For example, q̂ 1 is the operator
two-dimensional 共2D兲 plot will oscillate only along the Q 1
corresponding to the coordinate q 1 .
axis about the origin and the masses will oscillate in unison
The Hamiltonian, as written in position coordinates in Eq.
about the equilibrium positions. Figure 3共b兲 shows the other
共6兲, does not give a separable Schrödinger wave equation
normal mode, where the masses oscillate against each other.
Figure 3共c兲 shows only mass 2 being displaced. The time because of the (q̂ 1 ⫺q̂ 2 ) 2 term. However, in normal mode
evolution from this initial displacement will not be a simple coordinates,
oscillation like a normal mode, but a more complex motion P̂ 21 P̂ 22
1 1
that is a superposition of the two normal modes. Ĥ⫽ ⫹ ⫹ m ␻ 21 Q̂ 21 ⫹ m ␻ 22 Q̂ 22 , 共7兲
We next analyze this system quantum mechanically. We 2m 2m 2 2
seek state vectors 兩 ␺ (t) 典 that satisfy the Schrödinger equa- it does give a separable wave equation. Here, ␻ 21 ⫽ ␬ /m and
tion ␻ 22 ⫽3 ␬ /m are the eigenvalues for normal mode coordinates
d Q 1 and Q 2 , which are the eigenmodes of the system.
iប 兩 ␺ 共 t 兲 典 ⫽Ĥ 兩 ␺ 共 t 兲 典 , 共5兲 Because the wave equation separates in normal mode co-
dt
ordinates, these coordinates are the easiest to use. Solutions
where Ĥ is the Hamiltonian operator. In position coordinates, to the Schrödinger equation for this system can be written as
the Hamiltonian is simple products of the familiar one-dimensional harmonic
oscillator states. For example, the ground state of the system
p̂ 21 p̂ 22
1 is the product of ground state wave functions for Q 1 and for
Ĥ⫽ ⫹ ⫹ m ␻ 2 关 q̂ 21 ⫹q̂ 22 ⫹ 共 q̂ 1 ⫺q̂ 2 兲 2 兴 , 共6兲
2m 2m 2 Q2 ,
␺ 共 Q 1 ,Q 2 兲 ⫽ ␺ 0 共 Q 1 兲 ␺ 0 共 Q 2 兲

⫽ 冉 冊 冉
m␻1
␲ប
1/4
exp ⫺
m ␻ 1 Q 21
2ប
冊
⫻冉 冊 冉
m␻2
␲ប
1/4
exp ⫺
m ␻ 2 Q 22
2ប
冊
⫽ 具 Q 兩 0,0 典 . 共8兲
Fig. 2. Three views of the two-dimensional space formed by q 1 and q 2 . The The last line uses Dirac notation. The state vector, or ket, is
center view shows the single point in this space which gives the locations of written as 兩 0,0典 , which is of the form 兩 n 1 ,n 2 典 , where n 1 and
both masses. The position axes q 1 and q 2 are shown as solid lines and the
normal mode axes Q 1 and Q 2 are shown as dashed lines. The left view
n 2 indicate the energy eigenstate of the wave function along
shows its projection onto the q 1 and q 2 axes, and the right view onto the Q 1 and Q 2 respectively. Hence, 兩 0,0典 means that n 1 ⫽0 and
normal mode axes Q 1 and Q 2 . n 2 ⫽0 and the system is in its ground state along both normal

228 Am. J. Phys., Vol. 70, No. 3, March 2002 S. C. Johnson and T. D. Gutierrez 228
axes. This state vector exists in an infinite-dimensional Hil-
bert space and can be projected onto any set of basis vectors
that span the space, such as an energy basis, a coordinate
basis, or a momentum basis. One such set of coordinate basis
vectors corresponds to the infinite set of points on the
Q 1 – Q 2 plane and is represented by the shorthand notation
具 Q 兩 . The Hilbert space projection 具 Q 兩 0,0典 gives the two-
dimensional wave function ␺ (Q 1 ,Q 2 ). Note that all the
other projections in this section are done in this 2D coordi-
nate space, not in the Hilbert space.
Note that the one-dimensional harmonic oscillator states
do not lie along the position coordinates q 1 and q 2 , but
along the normal mode coordinates Q 1 and Q 2 . Thus, it is
easy to project the two-dimensional total probability density
onto the one-dimensional Q 1 and Q 2 axes,
P 共 Q 1 兲 ⫽ 兩 ␺ 0共 Q 1 兲兩 2. 共9兲
It is not so easy, however, to project it onto the q axes,

P共 q1兲⫽ 冕⫺⬁
⬁
兩 ␺ 共 Q 1 ,Q 2 兲 兩 2 dq 2 . 共10兲

Typically, this integration must be done numerically. For the

plots in this section, the numerical integration is done using
the standard trapezoid method. For the plots in Sec. III, it is
done using a Monte Carlo method.
As for any one-dimensional harmonic oscillator, we can
define raising and lowering operators 共also called creation Fig. 4. Probability densities of four states of the two-particle system, 共a兲
and annihilation operators兲. Each normal mode coordinate 兩 0,0典 , 共b兲 兩 0,1典 , 共c兲 兩 1,0典 , and 共d兲 兩 1,1典 .
has one raising and one lowering operator,

 †1 ⫽ 冑 m␻1
2ប 1
Q̂ ⫺i 冑 1
P̂ ,
2m ␻ 1 ប 1
共11a兲

冑 冑
each state, the center, two-dimensional plot is the most com-
m␻1 1 plete representation of the probability density function. It has
 1 ⫽ Q̂ ⫹i P̂ , 共11b兲
2ប 1 2m ␻ 1 ប 1 the disadvantage, however, that such a plot cannot be made
for functions of higher than three dimensions. The other sets
 †2 ⫽ 冑 m␻2
2ប 2
Q̂ ⫺i 冑 1
P̂ ,
2m ␻ 2 ប 2
共11c兲
of one-dimensional plots do not give as much information
and are probably not as easy to interpret, but they can be
easily adapted to visualizing higher-dimensional functions.
 2 ⫽ 冑 m␻2
2ប 2
Q̂ ⫹i 冑 1
P̂ .
2m ␻ 2 ប 2
共11d兲
This feature of the one-dimensional projections is the reason
we use them to visualize the many-dimensional phonon
wave function considered in Sec. III.
The action of these operators on a ket is to raise or lower one The projections onto the position axes 共the left-hand plots兲
of the n values, the energy eigenstate of the corresponding give the probabilities of finding masses 1 and 2 at various
mode, locations along the position coordinates q 1 and q 2 . The hori-
 †1 兩 n 1 ,n 2 典 ⫽ 冑n 1 ⫹1 兩 n 1 ⫹1,n 2 典 ,
zontal width of these plots has no physical meaning and is
共12a兲
chosen to make the probability densities easy to see. The
 1 兩 n 1 ,n 2 典 ⫽ 冑n 1 兩 n 1 ⫺1,n 2 典 , 共12b兲 projections onto the normal mode axes Q 1 and Q 2 共the right-
hand plots兲 are similar except they usually lack the intuitive
 †2 兩 n 1 ,n 2 典 ⫽ 冑n 2 ⫹1 兩 n 1 ,n 2 ⫹1 典 , 共12c兲 explanation of the q 1 and q 2 projections. For this system, the
Q 1 projection gives the probability density for the center of
 2 兩 n 1 ,n 2 典 ⫽ 冑n 2 兩 n 1 ,n 2 ⫺1 典 . 共12d兲 mass of the system because Q 1 ⫽1/&(q 1 ⫹q 2 ); the Q 2 pro-
jection gives the probability density for the relative coordi-
A few examples are  †1 兩 0,0典 ⫽ 兩 1,0典 ,  †1  †1 兩 0,0典 ⫽& 兩 2,0典 ,
nate 1/&(q 1 ⫺q 2 ). In these plots, it is clear that the com-
(Â †1 ) 3 兩 0,0典 ⫽ 冑6 兩 3,0典 , Â 1 兩 4,1典 ⫽2 兩 3,1典 , Â †2 兩 2,8典 ⫽3 兩 2,9典 , plete wave function is a product of functions along the
and  2 兩 3,0典 ⫽0. dashed Q 1 and Q 2 axes, not the solid q 1 and q 2 axes.
The quantum mechanical version of this system can be All the states shown in Fig. 4 are eigenfunctions of this
visualized using the same sets of coordinates as the classical system, so they do not vary with time. The superpositions of
version. Instead of plotting a single point on each plot, the these states, however, will change with time.
probability density is plotted at each point using a gray scale. Figures 5 and 6 show the time evolution of a particularly
Figure 4 shows four examples. Figure 4共a兲 shows the ground interesting type of state for harmonic oscillator systems, a
state for the two-particle system as a whole, given explicitly coherent state. This state is a superposition of an infinite
by Eq. 共8兲. Each of the other plots is an excited state. For number of two-particle eigenfunctions

229 Am. J. Phys., Vol. 70, No. 3, March 2002 S. C. Johnson and T. D. Gutierrez 229
 Fig. 7. Probablilty density for q̂ †1 兩 0,0典 .

冉 冊兺
⬁
兩␣兩2 ␣n † n
␺ 共 Q 1 ,Q 2 兲 ⫽ 具 Q 兩 exp ⫺ 共 Â 兲 兩 0,0典
2 n⫽0 n! 1

冉 冊兺 冑
⬁
兩␣兩2 ␣n
⫽exp ⫺ ␺ n共 Q 1 兲 ␺ 0共 Q 2 兲 ,
2 n⫽0 n!
共13兲
where ␺ n (Q 1 ) is the nth excited 1D harmonic oscillator
wave function, evaluated along coordinate Q 1 , and ␣ is a
constant specifying the amplitude of oscillation. 共See Ref. 3
for other excellent illustrations of this type of state.兲 These
states feature a single peak that is a Gaussian along any
eigenmode axis and that stays Gaussian as it moves in an
elliptical orbit around the origin. They are often called clas-
sical states because the center of the peak follows the trajec-
tory of a classical particle.
This particularly strong correspondence between the time
Fig. 5. Time evolution of a classical state of the two-particle system. In this evolution of a quantum mechanical wave function and the
state, the two masses are oscillating in synchronization in one of their eigen- motion of a classical particle makes coherent states particu-
modes. larly useful for demonstrating the transition from classical to
quantum mechanical models of systems. For example, Fig. 5
shows a state that corresponds to the classical oscillators os-
cillating in synchronization, like Fig. 3共a兲. The q 1 and q 2
plots show the particles oscillating in synch with each other
and the Q 1 and Q 2 plots show that only one eigenmode is
excited. Figure 6 shows an oscillation corresponding to the
other classical eigenmode where the two masses oscillate
against each other, like Fig. 3共b兲. The q 1 and q 2 plots again
show this behavior in an intuitive way.
Another interesting wave function is generated by the op-
erator q̂ 1 , which can be calculated from the coordinate trans-
forms of Eq. 共1兲. The resulting state q̂ 1 兩 0,0典 is shown in Fig.
7. This plot is similar to Fig. 4共c兲, which shows the state
Q̂ 1 兩 0,0典 . In particular, the q 1 – q 2 plots for q̂ 1 兩 0,0典 look like
the Q 1 – Q 2 plots for Q̂ 1 兩 0,0典 . There is a significant differ-
ence between these states, however. Because 具 Q 兩 Q̂ 1 兩 0,0典 is
an eigenfunction of the system, it is constant in time, but
具 Q 兩 q̂ 1 兩 0,0典 is not an eigenfunction so it will change with
time. Most notably, the two-peak pattern in the q 1 – q 2 plots
is not constant.
The main point of this section is to demonstrate in a visual
way the requirements for plotting a system of two particles
moving in one dimension. It requires a two-dimensional
space to show the complete probability density, but some
insight can be gained from groups of 1D projections. The
two sets of axes that are most useful for projections are the
position coordinate axes q 1 and q 2 and the normal mode
coordinate axes Q 1 and Q 2 .

III. AN EIGHT-PARTICLE LATTICE

Fig. 6. Time evolution of a classical state of the two-particle system, shown
at a different time scale than Fig. 5. In this eigenmode, the two masses The visualization techniques introduced in Sec. II can be
oscillate against each other. extended to higher dimensions to show the lattice vibrations

230 Am. J. Phys., Vol. 70, No. 3, March 2002 S. C. Johnson and T. D. Gutierrez 230
Fig. 8. The eight-particle system of harmonic oscillators connected by
springs of spring constant ␬. All masses are constrained to move vertically Fig. 9. The coordinate axes used to plot the eight-particle lattice.
along the coordinates q 1 through q 8 . Mass 8 is connected by a spring to
mass 1, as if the masses were arranged in a ring, giving periodic boundary
conditions.
and are collectively called P k . The oscillation frequency for
each normal mode is given by

of a crystal 共that is, phonons兲. As with the two-particle os-

cillator, we first consider the classical case. A system with
␻ k ⫽2 冑␬ m
sin
␲ 兩k兩
2 4
, 共16兲
eight oscillators is shown in Fig. 8. This system has periodic
which is the dispersion relation for this system. As for the
boundary conditions, which is a popular choice4 – 6 for dem-
two-particle case, the reason for using the normal mode co-
onstrating propagating waves. The masses and the spring
ordinates is that the equations of motion are decoupled using
constants are all identical. The position space coordinates are
these coordinates.
q 1 through q 8 and are collectively referred to as q x where the
The locations of all the masses, q x , can be specified by a
index x ranges from 1 through 8. The momenta of the
single point in an eight-dimensional space. Because such a
masses, which are directed vertically along each axis, are p 1 space cannot be drawn clearly on a two-dimensional page,
through p 8 or collectively p x with x ranging from 1 through we draw only the projections onto various axes in this space.
8. The axes that will be used in this section are shown in Fig. 9.
As with the two-particle problem, the key is to find the The position axes q x are on the left of each set of plots.
normal modes. The normal mode coordinates we will use are The eight dots give the positions of the eight masses. Rather
given by discrete cosine or sine transforms, than draw and label each of the eight position axes, q 1
8 through q 8 , we draw a single vertical axis labeled q x and
1 ␲
Q k⫽ 兺 q x cos 4 kx
x⫽1 冑8
共 k⫽0,1,2,3,4 兲 , 共14a兲 label the horizontal axis the x axis. The variable x is a di-
mensionless index; the distance from the leftmost mass to
8
mass number x is x ␴ , where ␴ is the distance between each
1 ␲
兺 q x sin 4 kx
mass.
Q k ⫽⫺ 共 k⫽⫺1,⫺2,⫺3 兲 , 共14b兲
x⫽1 冑8 The normal mode axes Q k are used in the center plot. As
with the q x axes, a single vertical axis labeled Q k is drawn at
and are collectively referred to as Q k . A positive k value k⫽0 and the horizontal axis is labeled k. The use of the
indicates a ‘‘cosine mode’’ and a negative k value indicates a letter k intentionally suggests a momentum, however, k is
‘‘sine mode.’’ Modes with identical 兩 k 兩 are degenerate; they not the conjugate momentum. Like x, k is a dimensionless
have identical energies and frequencies. For example, the index. The value of k is the wave number of the mode, which
cosine mode Q 1 is degenerate with the sine mode Q ⫺1 . for this system is the number of complete wave cycles in the
These normal mode coordinates are similar to, but distinct lattice. 共See the momentum discussion below.兲
from, the more commonly used coordinates based on a dis- The Q 1,⫺1 plot shows the projection onto both 兩 k 兩 ⫽1 axes
crete Fourier transform Q F,k ⫽ 兺 x⫽1
8
(1/冑8) q x e i ␲ /4kx . The simultaneously as a single point on a plane. The plot contains
Fourier modes are linear combinations of the modes we use, no new information; it is simply a different view of the two
Q F,1⫽(1/&)(Q 1 ⫹iQ ⫺1 ) and Q F,⫺1 ⫽(1/&)(Q 1 ⫺iQ ⫺1 ). axes Q 1 and Q ⫺1 which are already shown in the Q k plot.
Conversely, our modes are linear combinations of the Fourier However, it is useful because it shows more clearly the time
modes, Q 1 ⫽1/&(Q F,1⫹Q F,⫺1 ) and Q ⫺1 ⫽i/&(Q F,1 behavior of these two degenerate modes.
⫺Q F,⫺1 ). Fourier transform modes are complex, which Figure 10 shows all eight normal modes for this system.
makes them easier to manipulate than the pair of sine and Each eigenmode has a particularly simple projection on the
cosine modes, but also makes them more difficult to plot. Q k axes. Most of the normal modes 共except for k⫽0 and k
Because the goal of this article is to demonstrate phonon ⫽4兲 are degenerate pairs, with identical resonant frequencies
modes graphically, we have chosen the pure real sine and and identical wavelengths on the position axes. One mode of
cosine modes. each of these pairs corresponds to a sine function and the
The conjugate momenta 共see the discussion on momenta other to a cosine function. Linear combinations of these
below兲 for these normal mode coordinates have nearly iden- mode pairs can produce sinusoidal waves with varying phase
tical transforms, angles. Figure 11 shows such a wave propagating in time.
8
Note the time behavior of the projection on the Q 1,⫺1
1 ␲
兺 p x cos 4 kx
plane—the projected point moves in a circle as the wave
P k⫽ 共 k⫽0,1,2,3,4 兲 , 共15a兲
x⫽1 冑8 propagates along the position axes.
This system has two different momentum concepts, con-
8 jugate momentum and wave number, which can be difficult
1 ␲
P k⫽ 兺 p x sin 4 kx
x⫽1 冑8
共 k⫽⫺1,⫺2,⫺3 兲 , 共15b兲 to grasp.7 The conjugate momentum is proportional to the
time derivative of the coordinates. It is the familiar Newton-

231 Am. J. Phys., Vol. 70, No. 3, March 2002 S. C. Johnson and T. D. Gutierrez 231
 Fig. 11. Time evolution of the system showing a propagating 兩 k 兩 ⫽1 wave.

it is a quantity that is used in conservation laws for interac-

tions. The phonon momentum for this system is directed
horizontally along the x axis.
The following example qualitatively illustrates the nature
of phonon momentum. Consider a model for the absorption
of light by an ionic crystal. We neglect absorption by the
electrons and assume all the light is absorbed as the ions
vibrate in the electromagnetic field of the light 共which is a
good model for an ionic crystal absorbing infrared light兲.
Absorption is proportional to the amplitude of vibration of
the ions, so light will be appreciably absorbed only when the
ions are vibrating near one of their resonant frequencies. For
a lattice, the resonant frequencies are the normal mode fre-
quencies of Eq. 共16兲 and each corresponds to a particular
normal mode with a particular wave number k. Light with
both the right frequency ␻ k and wave number k will push all
the ions of the crystal in the right direction at the right time
Fig. 10. All eight normal modes of the eight-particle lattice, from k⫽⫺3 in to increase the amplitude of a particular normal mode. Light
共a兲 to k⫽4 in 共h兲, displayed using the coordinates used in Sec. III. at a frequency ␻ k ⬘ but with a wave number k ⬙ that does not
match the wave number of the normal mode will push dif-
ferent parts of the crystal out of phase with each other and
will not increase the amplitude of the normal mode. Thus,
ian momentum of each of the masses in the chain. For this the light must have the same frequency and wave number as
system, it is only directed vertically. Each of the eight indi- a particular phonon mode to interact with it. So, how is the
vidual masses has a time-varying value for this momentum. phonon momentum a conserved quantity? As the light is ab-
The other momentum concept, the wave number k, is a bit sorbed, its amplitude at a particular k decreases and the am-
more subtle. It is also called the phonon momentum or the plitude of the oscillation of the crystal at the same k in-
crystal momentum and is not a physical momentum.4 Instead, creases. In this sense, the quantity ‘‘amplitude at k’’ is

232 Am. J. Phys., Vol. 70, No. 3, March 2002 S. C. Johnson and T. D. Gutierrez 232
conserved. Quantum mechanically, the amplitudes are quan-
tized and an interaction term in the Hamiltonian will be of
the form B̂ k  †k , where B̂ k removes a photon of wave vector
k from the electromagnetic field and  †k adds a phonon of
wave vector k to the crystal. This is part of the fascinating
topic of interacting quantum fields, which is beyond the
scope of this article. Fig. 12. Ground state of the quantum mechanical lattice. The apparent ab-
We next analyze this eight-particle system quantum me- sence of a Q 0 projection is discussed in the text.
chanically. As with the two-particle system, we seek state
vectors 兩 ␺ (t) 典 that satisfy the Schrödinger equation
the Q k axes are also Gaussian with varying widths. Higher
d
iប 兩 ␺ 共 t 兲 典 ⫽Ĥ 兩 ␺ 共 t 兲 典 . 共17兲 兩 k 兩 values have narrower Gaussian profiles, corresponding to
dt the higher energies and higher frequencies of these modes, as
In position coordinates, the Hamiltonian is specified in the dispersion relation of Eq. 共16兲.
8 8
The Q 0 mode in Fig. 12 appears to be missing because we
p̂ 2x 1 have chosen a delta function ␺ (Q 0 )⫽ ␦ (Q 0 ) for the Q 0 nor-
Ĥ⫽ 兺 ⫹ x⫽1
x⫽1 2m
兺 2 m ␻ 2共 q̂ x⫹1 ⫺q̂ x 兲 2 , 共18兲
mal mode. Thus, the ground state wave function is

具 Q 兩 0 典 ⫽ ␦ 共 Q 0 兲 兿 ␺ 0共 Q k 兲 ,
which again does not give a separable wave equation be-
共23兲
cause of the (q̂ x⫹1 ⫺q̂ x ) 2 term. In normal mode coordinates k⫽0
4
P̂ 2k
4
1 where ␺ 0 (Q k ) is the 1D ground state harmonic oscillator
Ĥ⫽ 兺 ⫹ k⫽⫺3
k⫽⫺3 2m
兺 2 m ␻ 2k Q̂ 2k , 共19兲 wave function for normal mode k. Normal mode Q 0 repre-
sents the motion of the center of mass of the entire system.
which does give a separable wave function. The eight- Our eight-particle system is not anchored to any fixed refer-
dimensional wave function can be written as a product of 1D ence system because of periodic boundary conditions, so the
harmonic oscillator wave functions, motion of its center of mass coordinate Q 0 is that of a free
4 particle in space. Because the probability density for a free
␺ 共 Q ⫺3 ,Q ⫺2 ,...,Q 4 兲 ⫽ 兿
k⫽⫺3
␺ k共 Q k 兲 particle is uniform over space, using the free particle wave
function for ␺ 0 (Q 0 ) would produce a uniform probability
density for all q x coordinates, which would not be useful for
⫽ 具 Q 兩 n ⫺3 ,n ⫺2 ,n ⫺1 ;n 0 ,n 1 ,n 2 ,n 3 ,n 4 典 ,
visualization. So, instead, we use the center of mass of the
共20兲 system as the origin for our coordinate system. This choice is
where each ␺ k can be any function of one variable. The equivalent to transforming into the center of mass reference
Dirac-style notation on the last line is defined similarly to frame. The result is a delta function for the probability den-
that of the two-particle system and is a particularly useful sity of Q 0 .
notation for the state of the system. The quantum mechanical As for the two-particle system normal modes, or indeed
state of such a many-particle system is often called a Fock any 1D harmonic oscillator, we can define raising and low-
state. ering operators for this system. Each normal mode coordi-
Projecting the eight-dimensional probability density func- nate has one raising and one lowering operator,
tion onto one of the normal modes is easy,
P 共 Q 1 兲 ⫽ 兩 ␺ 1共 Q 1 兲兩 2, 共21兲
 †k ⫽ 冑 m␻k
2ប k
Q̂ ⫺i 冑 1
P̂ ,
2m ␻ k ប k
共24a兲

but it is not so easy to project it onto one of the q x axes,

 k ⫽ 冑
m␻k
Q̂ ⫹i 冑 1
共24b兲
冕
P̂ .
⬁ 2ប k 2m ␻ k ប k
P共 q1兲⫽ dq 2 dq 3 dq 4 dq 5 dq 6 dq 7 dq 8 兩 ␺ 兩 2 . 共22兲
⫺⬁ The action of these operators on a Fock space ket is to raise
The integration over a seven-dimensional subspace of the or lower one of the n k values, the energy eigenstate of the
eight-dimensional function must be done numerically. We corresponding mode. For example, the action of the k⫽1
use a Monte Carlo method as follows.8 We generate a large raising operator is  †1 兩 n ⫺3 ,n ⫺2 ,n ⫺1 ;n 0 ,n 1 ,n 2 ,n 3 ,n 4 典
number 共typically 105 兲 of random points in the eight- ⫽ 冑n 1 ⫹1 兩 n ⫺3 ,n ⫺2 ,n ⫺1 ;n 0 ,n 1 ⫹1,n 2 ,n 3 ,n 4 典 and of the k
dimensional Q k space with a distribution that matches the ⫽1 lowering operator is  1 兩 n ⫺3 ,n ⫺2 ,n ⫺1 ;n 0 ,n 1 ,n 2 ,n 3 ,
probability density in that space. This is done using the Me-
tropolis algorithm.9 Each of those points is transformed to q x
coordinates using the inverse transforms of Eq. 共14兲. A his-
togram is constructed for each q x and these histograms are
plotted on the q x vs x plots, with greater numbers of points
corresponding to darker shades of gray.
The ground state of the eight-particle quantum mechanical
system is shown in Fig. 12 and is written as
兩 0,0,0; 0,0,0,0,0 典 or simply 兩 0 典 . The projections onto the q x
axes are Gaussian and are all identical. The projections onto Fig. 13. The quantum mechanical lattice with one k⫽1 phonon, Â †1 兩 0 典 .

233 Am. J. Phys., Vol. 70, No. 3, March 2002 S. C. Johnson and T. D. Gutierrez 233
Fig. 14. The quantum mechanical lattice with one k⫽2 phonon, Â †2 兩 0 典 . Fig. 16. The quantum mechanical lattice with four k⫽1 phonons, three in
cosine modes and one in a sine mode, (Â †1 ) 3 (Â ⫺1
†
) 1兩 0 典 .

n4典⫽冑n 1 兩 n ⫺3 ,n ⫺2 ,n ⫺1 ;n 0 ,n 1 ⫺1,n 2 ,n 3 ,n 4 典 . Note that the

action of the k⫽1 lowering operator is not the same as that phonons as the one in Fig. 15, and because they are all 兩 k 兩
of the k⫽⫺1 raising operator, which is  ⫺1 †
兩 n ⫺3 ,n ⫺2 , ⫽1 phonons, the total energy of both systems is identical.
n ⫺1 ;n 0 ,n 1 ,n 2 ,n 3 ,n 4 典 ⫽ 冑n ⫺1 ⫹1 兩 n ⫺3 ,n ⫺2 ,n ⫺1 ⫹1;n 0 ,n 1 , The q x projection still shows an expectation value of zero for
n 2 ,n 3 ,n 4 典 . each mass.
The word phonon refers to an increase in the excitation To get a nonzero position expectation value for any of the
number of the system. The  †k operators add one phonon masses, a superposition of states is required. Figure 17 shows
a superposition of the ground state and a one-phonon state.
each and the  k operators remove one phonon each from a The probability densities of many of the masses are now
system. The ground state has zero phonons. Using an  †k clearly centered above or below the equilibrium position,
operator on the ground state creates a one-phonon state. For indicating a nonzero expectation value. Figure 17 shows the
example, Â †1 兩 0,0,0; 0,0,0,0,0 典 ⫽ 兩 0,0,0; 0,1,0,0,0 典 is a sys- time evolution of this state by showing it at three separate
tem with one k⫽1 phonon. A more compact notation is times. The peak of q x probabilities follows a cosine shape in
space and oscillates in time like a vibrating string.
 †1 兩 0 典 . This one-phonon state is shown in Fig. 13. In the The time behavior of these states is calculated from the
normal mode space, we see that the probability along Q 1 is Schrödinger equation. Each 1D harmonic oscillator wave
that of a first excited harmonic oscillator and all the other function is multiplied by a phase factor e i ␻ k t , where ␻ k is
normal modes are in their ground states. The Q 1,⫺1 axes determined by the dispersion relation of Eq. 共16兲. The norm
emphasize that the first excited state is along a cosine mode. of the resulting wave function gives the probability density
On the q x position axes, we do not see a cosine wave as we that is plotted.
might expect. The expectation value of the position for any At this point, we have introduced all the background
mass is still zero. However, the width of the probability dis- needed to understand the most important plot in this paper,
tribution now varies with x and this variation follows a co- Fig. 18. It is a coherent state similar to those introduced in
sine function. the discussion of the two-particle system. To get a coherent
Similar observations apply to Fig. 14, which shows the state in any one mode, we would use an expression similar to
result of  †2 兩 0,0,0; 0,0,0,0,0 典 ⫽ 兩 0,0,0; 0,0,1,0,0 典 or  †2 兩 0 典 . Eq. 共13兲,

冉 冊兺
⬁
This system has one k⫽2 phonon. Its Q 2 probability density 兩␣兩2 ␣ n i␾ † n
is that of a first excited harmonic oscillator state. The width exp ⫺ e n 共 Â k 兲 兩 0 典 , 共25兲
2 n⫽0 n!
of its q x probability densities vary with x following a cosine
function, except that the wavelength of this cosine function
is shorter than for the k⫽1 phonon.
We can put several phonons into a system by using the
raising operator several times. For example, (Â †1 ) 4 兩 0 典
⫽ 兩 0,0,0; 0,4,0,0,0 典 is a system with four k⫽1 phonons and
is shown in Fig. 15. The Q k plot shows that the k⫽1 normal
mode is in its fourth excited state and the q x plot clearly
shows the cosine variation in the width of the probability
density of the masses. The amplitude of the width variation
is greater with more phonons in the system.
A system can also contain several phonons of arbitrary k
values. Figure 16 shows the system with three k⫽1 phonons
and one k⫽⫺1 phonon. This system has the same number of

Fig. 17. Time sequence for the quantum mechanical lattice in a superposi-
Fig. 15. The quantum mechanical lattice with four k⫽1 phonons, (Â †1 ) 4 兩 0 典 . tion of states, the ground state and a one-phonon state, 1/& 关 Â †1 兩 0 典 ⫹ 兩 0 典 ].

234 Am. J. Phys., Vol. 70, No. 3, March 2002 S. C. Johnson and T. D. Gutierrez 234
 Fig. 19. A fifty-particle phonon system in a coherent k⫽⫺1 共sine-type
mode兲 state.

probability distribution is true for all states of this system,

and the coherent state simply makes the connection to the
classical system most vivid.
In fact, the basic idea of replacing each classical value
with a probability distribution applies to any quantum field,
and we believe that this idea is a good way to introduce
quantum field theory. Figure 19 shows a coherent state for a
fifty-particle quantum mechanical lattice. Using fifty par-
ticles makes this system a good approximation to a continu-
ous, scalar field in one dimension.
The coherent propagating waves of Figs. 18 and 19 are
good approximations to several real-world quantum fields.
For a lattice, these represent sound waves traveling through
the lattice. In contrast, noncoherent sums of phonon states 共in
the proper proportions兲 can represent thermal vibrations of
the lattice. Using the scalar field to represent one component
of an electromagnetic field, Figs. 18 and 19 are qualitatively
correct representations of the coherent light of a laser. They
could also represent the electric waves broadcast by a radio
antenna. In contrast, noncoherent sums of these electric field
states can represent thermal radiation.
It is interesting to see a few other states of the eight-
particle system. Figure 20 shows a noncoherent sum of sev-
Fig. 18. Time evolution of the quantum mechanical lattice in a coherent,
propagating 兩 k 兩 ⫽1 state. This is a good visualization of a sound wave in a
eral 兩 k 兩 ⫽1 states, each with the same energy. Without coher-
solid. ence, though, the q x probability distributions are all centered
on zero. This sum of states is a good analogy for monochro-
matic but noncoherent light.
Figure 21 shows a squeezed state.10 Squeezed states of
where ␾ n ⫽(n⫹ 21 ) ␻ k t gives the time dependence. This state light are similar to laser light and can be produced by various
would produce a coherent standing wave corresponding to nonlinear optical techniques. A squeezed state is similar to a
mode k. However, Fig. 18 is a propagating wave which re- coherent state, but the width of the Gaussian function plotted
quires a superposition of a cosine mode and a sine mode with in the Q 1,⫺1 plane varies with time. The squeezing is greatest
the right relative phase. Such a state is constructed by in Fig. 21共c兲, where the width of the Gaussian is a minimum.
⬁ ⬁ A measurement of the 兩 k 兩 ⫽1 modes of the system at that
␣ n i␾ † n ␣ m i␾
exp共 ⫺ 兩 ␣ 兩 兲
2
兺
n⫽0 n!
e n 共 Â 1 兲 兺
m⫽0 m!
†
e m 共 Â ⫺1 兲 m兩 0 典 , time will have less uncertainty than a measurement on an
otherwise identical coherent state. This squeezing is balanced
共26兲 by the width of the Gaussian in Figs. 21共a兲 and 21共e兲, which
with ␾ n ⫽(n⫹ 21 ) ␻ 1 t and ␾ m ⫽(m⫹ 21 )( ␻ ⫺1 t⫺ ␲ /2), show- is larger than for an otherwise identical coherent state. Thus,
a squeezed state allows the system to have a lower uncer-
ing that the sine modes are shifted by ⫺ ␲ /2 relative to the
tainty in its position at some times, but compensates by hav-
cosine modes. 共With no shift, the result would be another
ing a higher uncertainty at other times. This reduction in
standing wave.兲 Figure 18 shows the time evolution of such
uncertainty can be used to improve the sensitivity of some
a 兩 k 兩 ⫽1 propagating coherent state. types of measurements, for example, a gravitational wave
The q x projection is the most pedagogically useful. If we detector.11
compare Fig. 18 with the classical propagating wave in Fig.
11, we see that the pictures look qualitatively similar with a
sinusoidal wave propagating to the right. In the quantum
case, however, the locations of each mass are not certain, but
are given by a probability distribution that is centered on the
classical location. We claim that Fig. 18 is an intuitive and
correct picture of a propagating phonon. It is important to
note that this picture is of a rather special, coherent state
instead of a single phonon. Still, the idea of replacing the
definite classical values of the positions of each mass with a Fig. 20. A noncoherent superposition of several 兩 k 兩 ⫽1 four-phonon states.

235 Am. J. Phys., Vol. 70, No. 3, March 2002 S. C. Johnson and T. D. Gutierrez 235
 Fig. 23. A local excitation at x⫽3 from the q̂ †3 field operator, q̂ †3 兩 0 典 .

particle system, this is done by using the q̂ x operator, which

is often called a field operator. Like q̂ 1 for the two-particle
system, this operator can be expressed in terms of the normal
mode operators using the inverse of the discrete sine and
cosine transforms of Eq. 共14兲,
4 ⫺1
1 ␲ 1 ␲
q̂ x ⫽ 兺
k⫽0 冑8
Q̂ k cos
4
kx⫹ 兺
k⫽⫺3 冑8
Q̂ k sin kx.
4
共27兲

Figure 23 shows the state q̂ †3 兩 0 典 which puts such a localized

excitation at location x⫽3. The q x plot of this system is quite
similar to the Q k plot of Fig. 13. In both plots, most of the
coordinates have ground state probability densities and just
one coordinate has a first excited state probability density.
However, the Q k plot shows an eigenfunction so it will be
stationary in time, whereas the q x plot shows a superposition
so it will vary in time. In particular, the q x plot’s appearance
is short-lived and quickly decays to be barely distinguishable
from the ground state.
The word ‘‘particle’’ has a different meaning in quantum
field theory than it does in many other contexts. Usually,
particle refers to an object localized in space. However, a
phonon is usually called a particle, and it is distributed over
Fig. 21. Time development of a squeezed state.
the entire lattice. The same observation can be made, for
example, for a photon 共an excitation of the electromagnetic
field兲 or for an electron 共an excitation of its associated field,
Figure 22 shows a one-phonon state in one of the alternate often called the Dirac field兲. The local excitation q̂ †x 兩 0 典 de-
discrete Fourier transform normal mode coordinates, dis-
scribed above is not a phonon; it is in fact a linear combina-
cussed after Eq. 共14兲. This state is a complex linear combi-
tion of many phonon states. In some contexts, this local ex-
nation of the k⫽1 and k⫽⫺1 normal mode coordinates. It citation is called a particle.
has the advantage that this one-phonon state represents a
traveling wave. States with positive k travel in one direction
and negative k travel in the other direction. However, a plot IV. CONCLUSION
reveals why we chose not to use these coordinates—they are A one-dimensional lattice of N coupled harmonic oscilla-
featureless on a probability density plot. If phase were plot-
tors is a good demonstration system for phonons and, as N
ted on these plots, such as by using a color code, a point of
constant phase on the Q ⫾1 plot would travel in a circle at →⬁, for quantum fields. Classically, this system demon-
strates the normal mode decomposition needed to solve such
frequency ␻ k . 共For beautiful pictures of one- and two-
many-particle or field theory problems and also shows the
dimensional wave functions showing phase information, see
two types of momentum, p x and k, that such systems have.
Ref. 12.兲
It is possible, and often useful, to define an operator that Quantum mechanically, the wave function is a complex func-
creates a localized excitation of the lattice 共for phonons兲 or tion of N dimensions which cannot be plotted for N⬎3.
the field 共for photons and electrons and such兲. For our eight- However, the probability density is a positive, real number at
each point in this N-dimensional space and can be projected
onto groups of coordinate axes such as the q x position axes
and Q k normal mode axes. The resulting probability density
plots present a somewhat intuitive picture of the phonon
wave function, or of the wave function of any quantum field.
What they show for a particular state of the system, the co-
herent state, is that displacement of each classical mass is
replaced by a probability density centered on the classical
location. Thus, just like the location of a single particle is
Fig. 22. A one-phonon state using one alternate normal mode coordinate, blurred in quantum mechanics, the location of every particle
关  †1 ⫹i ⫺1
†
兴兩0典. in a lattice, or every value of a field, is blurred in quantum

236 Am. J. Phys., Vol. 70, No. 3, March 2002 S. C. Johnson and T. D. Gutierrez 236
 a兲
field theory. We hope that these pictures can serve as an Present address: Intel Corporation, 5200 NE Elam Young Parkway, Mail
enticing introduction to the fascinating but difficult subject Stop RA1-329, Hillsboro, Oregon 97124; electronic mail:
of quantum field theory. [email protected]
b兲
Electronic mail: [email protected]
1
The software, BlochApp, FieldApp, JCApp, and QEDApp, by G. Andrew
V. PROBLEMS Antonelli and Wolfgang Christian, is described in Denis Donnelly, ‘‘CIP’s
sixth annual software contest,’’ Comput. Phys. 9, 594 – 601 共1995兲.
共1兲 Find the time dependence of the positions of the two 2
Ramamurti Shankar, Principles of Quantum Mechanics 共Plenum, New
classical masses for each configuration in Fig. 3. Use Q 1i York, 1994兲, 2nd ed. Example 1.8.6 is the classical coupled harmonic
and Q 2i or q 1i and q 2i as the coordinates at time t⫽0. First oscillator, examples 10.1.3 and 10.2.2 treat the quantum version, and
use the Q coordinates and then transform to the q coordi- Chapter 7 is on the one-dimensional oscillator.
3
nates. Siegmund Brandt and Hans Dieter Dahmen, The Picture Book of Quantum
共2兲 Write explicit expressions for the four wave functions Mechanics 共Wiley, New York, 1985兲, Secs. 7.1, 7.2, 9.1, 12.3.
shown in Fig. 4. 4
Charles Kittel, Introduction to Solid State Physics 共Wiley, New York,
1976兲, 6th ed., Chap. 4.
共3兲 Show that Q̂ 1 兩 0 典 is an eigenstate of the two-particle 5
Neil Ashcroft and N. David Mermin, Solid State Physics 共Saunders Col-
system, but that q̂ 1 兩 0 典 is not. Hint: find Q̂ 1 and q̂ 1 in terms lege Publishing, 1976兲, p. 430.
of  1 and  †1 using Eqs. 共11兲 and 共1兲 and show that operating
6
Walter Greiner and Joachim Reinhardt, Field Quantization 共Springer-
Verlag, Berlin, 1996兲, Chap. 1.
on 兩 0 典 with Q̂ 1 gives a single particle state, but that q̂ 1 gives 7
John Robinson Pierce, Almost All About Waves 共MIT, Cambridge, 1974兲,
a superposition of states. Chaps. 10 and 11.
共4兲 Find the time dependence of the positions of the eight 8
Verissimo M. de Aquino, Valdir C. Aguilera-Navarro, Mario Goto, and
classical masses in Fig. 11. Use Q xi or q xi as the coordinates Hiromi Iwamoto, ‘‘Monte Carlo image representation,’’ Am. J. Phys. 69,
at time t⫽0. First use the Q coordinates, and then transform 788 –792 共2001兲.
9
into the q coordinates. See, for example, B. L. Hammond, W. A. Lester, Jr., and P. J. Reynolds,
Monte Carlo Methods in Ab initio Quantum Chemistry 共World Scientific,
共5兲 共Advanced兲 Evaluate the expectation value for Q̂ 1 and Singapore, 1994兲.
Q̂ 2 for the states in Figs. 13 and 17 共at time t⫽0 only兲. Hint: 10
Marlan O. Scully and M. Suhail Zubairy, Quantum Optics 共Cambridge
the expectation value of Q̂ k for a state 兩 ␺ 典 is 具 ␺ 兩 Q̂ k 兩 ␺ 典 , 11
U.P., Cambridge, 1997兲.
Peter Knight, ‘‘Quantum optics,’’ in The New Physics, edited by Paul C.
where it is usually helpful to express Q̂ k and 兩 ␺ 典 in terms of W. Davies 共Cambridge U.P., Cambridge, 1992兲, Secs. 10.6 and 10.9.
 k and  †k . For example, 兩 ␺ 典 ⫽ †1 兩 0 典 and 具 ␺ 兩 ⫽ 具 0 兩  1 . Use 12
Bernd Thaller, Visual Quantum Mechanics: Selected Topics with
the result that 具 n ⫺3 ...n 4 兩 m ⫺3 ...m 4 典 ⫽ 兿 k⫽⫺3
4
␦ n k m k . That is, Computer-Generated Animations of Quantum-Mechanical Phenomena
共Springer/TELOS, Berlin, 2000兲.
具 n ⫺3 . . . n 4 兩 m ⫺3 . . . m 4 典 is 0 unless both sides have the 13
Rubin H. Landau, Quantum Mechanics II: A Second Course in Quantum
same number in each mode. For more of this sort of problem, Theory 共Wiley, New York, 1996兲, Sec. 21.2, Coherent States of the Radia-
see Ref. 13. tion Field: Tutorial.

237 Am. J. Phys., Vol. 70, No. 3, March 2002 S. C. Johnson and T. D. Gutierrez 237