Sie sind auf Seite 1von 136

The Delights of

the Appendices to
The Joy of Quantum Physics
Michael A. Morrison
Homer L. Dodge Department of Physics & Astronomy
University of Oklahoma
Version 8.37: August 3, 2010
Date Printed: August 3, 2010
c _2010 by Michael A. Morrison.
Not to be distributed or copied without permission of the author.
Contents
Version 8.37: August 3, 2010
A. Quantum mechanics: Greatest Hits 1
A.1. Of states, ensembles, and measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
A.2. Three great ideas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
A.3. Four great postulatesand their consequences . . . . . . . . . . . . . . . . . . . . . . . . . . 4
A.4. Stationary states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
A.5. Non-stationary states and the method of eigenfunction expansion . . . . . . . . . . . . . . . 20
A.6. Generic problem-solving strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
A.7. Measurements in the microverse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
B. Dimensional extermination 26
B.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
B.2. Transformation procedure and tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
B.3. Transforming the Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
B.4. Conversion back to dimensional quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
B.5. Solving the dimensionless TISE for the Morse potential . . . . . . . . . . . . . . . . . . . . . 35
B.6. Scale transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
B.7. Final exhortations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
B.8. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
C. Constants and estimates 42
D. The Conversion Factory 47
E. SI units 50
F. Atomic units 53
F.1. Introducing atomic units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
F.2. How to make atomic units work for you . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
F.3. Atomic unit replacement rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
G. Dirac notation 64
G.1. Dirac shorthand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
G.2. Dirac notation and Hermiticity of operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
G.3. Eigenfunction expansions in Dirac notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
G.4. Useful properties in Dirac notation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
G.5. Projecting a state out of the time-independent Schrodinger equation . . . . . . . . . . . . . . 68
H. The mathematics of operators 70
H.1. Operators in quantum mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
i
CONTENTS ii
I. Matrices and determinants 79
I.1. Useful properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
I.2. Eigenvalues, eigenvectors, and matrix diagonalization . . . . . . . . . . . . . . . . . . . . . . . 81
I.3. Determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
J. Making sense of spectral data 88
K. Angular-momentum coupling 94
K.1. The total orbital angular momentum operator . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
K.2. Commutation relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
K.3. The fundamental expansions of angular-momentum coupling . . . . . . . . . . . . . . . . . . 96
K.4. Eigenvalue equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
K.5. Clebsch-Gordan coecients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
L. The prex dictionary 102
M.The Greek alphabet 103
N. The Dirac delta function 104
N.1. A Dirac-delta primer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
N.2. A Dirac-delta users guide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
O. Waves and the de Broglie-Einstein relations 109
P. The one-dimensional simple harmonic oscillator 113
P.1. Hamiltonian eigenfunctions and stationary-state energies . . . . . . . . . . . . . . . . . . . . . 113
P.2. Creation and annihilation operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
P.3. Momentum-space wave functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Q. The Seven Habits of Highly Eective Problem Solvers 120
Q.1. Why you need to develop new habits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Q.2. Implementation Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Q.3. The Principle of Brainstorming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
Q.4. The Principle of Symmetry Seeking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
Q.5. The Principle of Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Q.6. The Principle of Pattern Seeking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Q.7. The Principle of Back-to-Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Q.8. The Principle of Least Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Q.9. The Principle of Alert Awareness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Index 127
PrintAppendices Version: 8.37 Printed: August 3, 2010
Appendix A
Quantum mechanics: Greatest Hits
Version 8.42: August 3, 2010
Before laying it before the public, it would be as well, perhaps,
that I should refresh their memories as to the simpler facts
upon which this commentary is founded.
The Man with the Watches, by Sir Arthur Conan Doyle
Quantum mechanics is a collection of postulates and consequences thereof, formulated in the language of
mathematics, which provides tools for the analysis, prediction, and understanding of observed phenomena
in the microscopic domain. This appendix reviews the key ideas and postulates of quantum physics. It then
recaps the high points of the two types of quantum states found in nature, stationary and non-stationary
states. These introductory sections set the stage for a birds eye tour of the major mathematical machinery
of quantum mechanics and of strategies and tactics for conquering quantum problems. The last section
briey discusses the eect of measurement on a quantum state. You can nd additional information about
Dirac notation in Appendix G and about the mathematics of operators in Appendix H.
A.1 Of states, ensembles, and measurement
The most important notion in quantum physics is that of a state of a physical system. We associate a
particular quantum state with probabilistic information about the observables of the system. In classical
and in quantum physics, an observable is a physically measurable quantity. The mathematical tool we use
to represent an observable in quantum mechanics is an operator.
1
1
Notation: Some authors use the term dynamical variable for what I call an observable. Others use observable and
operator interchangeably. To distinguish the two, Ill put a little hat on operators but not on observables. Thus p is the
observable linear momentum, while p is the corresponding operator. The sole exception to this rule is the energy E, for which
the operator is the Hamiltonian

H.
Appendix 2
A.1.1 States and their wave functions
In introductory quantum mechanics, we represent a quantum state mathematically by a wave function. In
the simplest representation, the wave function depends on the position variables appropriate to the system
and on time. The most general wave function represents a state in which the various observables of the
system do not have specic values; rather, each observable can assume one of a number of possible allowed
values, each with a non-zero probability. The process of measuring a particular observable actualizes one
of these possibilities, leaving the system in a state where the measured observable does have a denite value
(see A.7).
Ensembles. Since the information contained in the wave function is inherently probabilistic, this infor-
mation does not describe individual systems. Rather it describes a quantum-mechanical ensemble (or,
simply, an ensemble): a huge number of identical systems all of which are in the same quantum state
(see A.1.2). Each system is called a member of the ensemble. According to the Born interpretation,
the wave function contains information about the probability that a member of the ensemble will exhibit a
particular value of an observable upon measurement of that observable. So when we talk about a quantum
state, were talking about an ensemble, not an individual system.
Bound versus continuum states. States come in two varieties: bound and continuum. Physically, the
distinction is simple: a particle in a bound state is spatially localized for all time in one region of spacenear
the potential energy that binds it. By contrast, a particle in a continuum state suers no such restriction.
If the continuum state is non-stationary, then the region of localization (that is, the region where the particle
is most likely and be found in a position measurement) can move throughout space as time passes. To clarify
the distinction, consider the hydrogen atom: one electron in the attractive Coulomb eld of a proton. In a
bound state, the probability of nding the electron far from the nucleus diminishes (exponentially) to zero
as distance from the nucleus increases to innity (Chap. 5). By contrast, in a continuum state, the electron
can escape the vicinity of the proton to be found at arbitrarily large distances from it. (In this case,
the electron is ionized.) Continuum states are especially important because of their role in the dynamics of
collisions. For this reason, a continuum state is often called a scattering state.
A.1.2 Measurement and the Born interpretation
In general, quantum mechanics cannot predict the value that a single member of an ensemble will exhibit
in a measurement of an observable. So we must be very careful when we interpret experiments. In quantum
physics, measurement has a precise meaning that follows from the intrinsically probabilistic character of
information. By a measurement on a microscopic system, we mean an ensemble measurement: identical
measurements performed on a huge number of identical systems that have been prepared so that prior to
measurement each is in the same quantum state.
2
The notion of an ensemble measurement helps us understand the nature of information in a wave func-
tion. From a wave function we can determine the fraction of members of the ensemble that exhibit each
allowed value of whatever observable were measuring. A familiar instance is position: according to the Born
interpretation, we calculate information about a particles position by squaring the modulus of its wave
function. Thus the wave function is a probability amplitude for various possible outcomes of a position
measurement. The generalized Born interpretation extends this approach to all observables. In a par-
ticular quantum state every observable is characterized by a probability distribution of allowed values; we
determine the probability for each allowed value from quantities we calculate from the wave function. The
mathematics we use to perform this minor miracle is the method of eigenfunction expansion (A.5).
2
Details: To ensure that all members of the ensemble are in the same state, they must be prepared identically. Of course,
identical preparation does not imply that these members will behave the same in a subsequent ensemble measurement. This
special outcomeall members exhibiting the same value of an observableoccurs only if the state prior to measurement was
an eigenstate of that observablethat is, in that state, the observable is sharp.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 3
A.2 Three great ideas
The intrinsic weirdness of quantum physics is prominently on display in the three great ideas which, early
in the 20
th
century, marked the transition from classical to quantum physics. These ideas are the heart
and soul of the physics of matter. All three violate both our intuition as human beings and the insight we
gained from studying classical physics. The break with classical physics is especially dramatic: To describe a
microscopic particle, we must abandon completely the notion that a particle follows a well-dened trajectory.
More radically, we must abandon the idea that we can study a physical system without interacting with it.
These principles force us to jettison the classical precepts of causality, objectivity, and determinism.
A.2.1 Quantization
Observables that in classical physics are continuous may be quantized in the microscopic realm. For example,
in classical physics the magnitude L of the orbital angular momentum of a negatively charged particle in
orbit around a positively charged particle at the origin can assume any valueand it does, as the particle
loses energy continuously and spirals into the origin. But the orbital angular momentum of an electron
bound to a proton in a hydrogen atom can assume only discrete values (Chaps. 4-5). Similarly, the energy
of the electron is quantized. The electron is forbidden to assume certain values of energy. Fortunately,
quantization keeps the electrons in an atom from plummeting into the nucleus.
We refer to the collection of allowed values of any observable as the spectrum of that observable. Not
all allowed values of an observable are necessarily quantized. For example, the energy of a free particle
is continuous; this observable can assume any value E 0. Moreover, some observables are allowed both
discrete and continuous values. In this case, we refer to the discrete part and the continuous part of the
spectrum. In quantum mechanics, the allowed values of an observable are the eigenvalues of the operator
that represents the observable (see Appendix H). We determine the spectrum of an observable by solving
the eigenvalue equation of the corresponding operator (Appendix I).
A.2.2 Uncertainty
The inherent fuzziness in nature is quantied by the quantum mechanical uncertainty, a quantity we
calculate from the wave function. The signature of fuzziness of an observable in a particular quantum state
is the following: in an ensemble measurement of the observable, various members of the ensemble return
dierent valuesall drawn, of course, from the list of eigenvalues of the corresponding operator. In such a
state the uncertainty in the observable is positive. Only in the special case that the quantum state is an
eigenstate of the observable will all members return the save value. In this case, the uncertainty is zero,
the observable has a single, well-dened value, and we say that in this state the observable is sharp.
3
A more general expression of this extraordinary property is the Heisenberg uncertainty principle
(HUP) (A.3.12). The HUP is one instance of the remarkable general principle that nature prohibits simul-
taneous precise measurement of certain pairs of observables. (The Heisenberg principle refers to position
and momentum.) In general, an uncertainty relation is a mathematical statement about the product of
the uncertainties of two observables. By the denition of the uncertainty, this product must be non-negative
(see A.6). If the state is an eigenstate of an observable, the uncertainty for that observable is zero.
But an uncertainty relation is not restricted to any particular state: rather, its a general statement
about the uncertainties in two observables in any quantum state. If the uncertainty product is positive
(rather than zero), then the uncertainty relation articulates a fundamental limitation on our knowledge of the
observable. Quantum mechanics provides a Generalized Uncertainty Principle (GUP)a prescription
3
A cautionary note: Its vital that you distinguish quantum mechanical uncertainty, which is a well-dened quantity we
calculate from wave functions, from experimental uncertainty. The latter arises from limitations inherent in the apparatus
we use to perform measurements and has nothing to do with the quantum mechanical uncertainty. Many physicists prefer the
word indeterminacy to uncertainty.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 4
for deriving the uncertainty relations for any two observables for any quantum state from the commutator
of the corresponding two operators (see A.3 and Appendix H).
Not all pairs of observables are constrained by uncertainty relations. We can determine whether such a
constraint applies to a pair of observables by evaluating their commutator. Operators whose commutator
is non-zero (operators that dont commute) are constrained by an uncertainty relation. Operators whose
commutator is zero (commuting operators) are not. We call two observables that are subject to an
uncertainty relation incompatible observables. We call all other pairs compatible observables:
Rule: If two observables are incompatible, then in any quantum state its impossible in principle to measure
both observables at the same time to innite precision.
A.2.3 Duality
The physical description of the macroscopic world elegantly and neatly divides phenomena into two irre-
deemably distinct categories: waves and particles. But in the microverse, matter and its constituents do
not behave according to our classical notions of either waves or particles. Microscopic thing-a-ma-jigs are
essentially other. But to avoid silly locutions, we go ahead and call them particles.
Quantum particles behave like waves in some experiments and like particles in others. According to
the Principle of Complementarity, microscopic particles never manifest both wave-like and particle-like
behavior in the same measurement.
The mathematical embodiments of this wave-particle duality are the Einstein-de Broglie relations
(Appendix O),
= h/p, and = E/h. (A.1)
Each equation relates a wave-like property (on the left-hand-side) to a particle-like property (on the right-
hand-side).
In replacing classical notions like determinism by quantization, uncertainty, and duality, quantum me-
chanics wreaks epistemological havoc. It transform the very nature of knowledge. In quantum physics,
knowledge is inherently, irredeemably probabilistic, statistical, and limited.
4
A.3 Four great postulatesand their consequences
In the absence of data we must abandon the analytic or scientic method of investigation,
and must approach it in the synthetic fashion. In a work, instead of taking known events
and deducing from them what has occurred, we must build up a fanciful explanation if it
will only be consistent with known events. We can then test this explanation by any fresh
facts which may arise. If they all t into their places, the probability is that we are upon
the right track, and with each fresh fact this probability increases in a geometrical
progression until the evidence becomes nal and convincing.
The Man with the Watches, by Sir Arthur Conan Doyle
The ideas in A.1 and A.2, along with spin (Chap. 7) and indistinguishability (Chap. 8), constitute the
conceptual core of quantum physics. To implement these ideas in the study of physical systems and the
design of technological devices we must express them mathematically. The primary mathematical devices
4
Read on: The question of the meaning and philosophical implications of quantum physics is one each individual must decide.
A ne, balanced non-technical introduction to these murky matters is David Alberts Quantum Mechanics and Experience
(Cambridge, MA: Harvard University Press). A more penetrating and provocative inquiry is James T. Cushings Quantum
Mechanics: Historical Contingency and the Copenhagen Hegemony (Chicago: University of Chicago Press, 1994). To whet
your appetite, heres a quote from Cushings introduction: It is astounding that there is a formulation of quantum mechanics
that has no measurement problem and no diculty with a classical limit, yet is so little known.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 5
of quantum mechanics are the operators and wave functions which represent a microscopic system, its
observables, and its states. To associate these mathematical devices with the core ideas of A.2 is the job of
the postulates reviewed in this section.
5
A.3.1 The wave function
Postulate 1 (The wave function)
Every physically realizable state of a system is represented in quantum mechanics by a wave function .
This function contains all accessible physical information about the system in that state.
An essential attribute of wave functions is that they satisfy the Principle of Superposition: If
1
and

2
represent two physically realizable states of a system, then any linear combination,
= c
1

1
+ c
2

2
, a superposition of two wave functions. (A.1)
with arbitrary complex constants c
1
and c
2
, represents a third physically realizable state. This acutely non-
classical principle follows from the linearity of the time-dependent Schrodinger equation (TDSE), which all
wave functions satisfy (see Postulate IV).
6
Requirements for physical admissibility. Not any old function of position and time can represent a
quantum state. In order that a wave function be subject to interpretation as a probability amplitude (see
Postulate II), it must satisfy four requirements of physical admissibility:
(1) The wave function must be a continuous function of position. This condition ensures, among
other things, that the probability density [[
2
is everywhere continuous.
(2) The wave function must be a smoothly varying function of position. That is, rst and
second derivatives with respect to all position variables must be continuous. Together with the rst
requirement, this ensures continuity of the probability current density (see Complement ?? to Chap. 5).
(3) The wave function must be single-valued. If this condition is met, then the probability density
the current density will be single-valued. Were this not the case, then we could associate dierent
position probabilities with a single point in spaceand even in quantum mechanics you cant do that.
7
(4) The wave function must be normalizable. For example, every wave function (x, t) of a particle
in one dimension (1D) must be such that the normalization integral is nite:
8
N
_

(x, t)(x, t) dx. normalization integral (1D) (A.2a)


5
Commentary: The basic assumptions (postulates) are not to be understood as mathematical axioms from which every-
thing can be derived without using further judgement and creativity. An axiomatic approach of this kind does not appear to
be possible in physics. The basic assumptions are to be considered as a concise way of formulating the quintessence of many
experimental facts. from Quantum Mechanics: Foundations and Applications (Third edition) by Arno Bohm, (New York:
Springer-Verlag, 1993).
6
Details: This statement of Postulate I pertains to the class of so-called pure states. An ensemble is in a pure state if
its members are in a maximally sharp quantum state, that is, a state in which all values of a complete set of commuting
operators are sharp (Chaps. 25). An ensemble whose preparation has not actualized a maximally sharp state is said to be in
a mixed state. We cannot represent a mixed state with a wave function, because in essence we dont know enough about the
state to write one down. The extension of Postulate I to mixed states avers the existence of a density matrix that contains all
the accessible information about the state. The physics of mixed states is governed by quantum statistical mechanics. Youll
nd a clear introduction to these ideas in Chap. 1 of Density Matrix Theory and Applications, Second Edition, by Karl Blum
(New York: Plenum, 1996).
7
Commentary: We impose the requirement of single-valuedness on the wave function itself rather than on the probability
density ||
2
to ensure that the probability density for superposition states will be single-valued: if
1
and
2
are single valued,
then so must be the probability density for the linear combination c
1

1
+c
2

2
.
8
A cautionary note: The requirement of normalizability does not imply that itself must be nite everywhere. Singularities
are allowed so long as the normalization integral remains nite.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 6
Only if N is nite can we normalize the wave function, which we do by multiplying the function
by 1/

N. The normalization integral for the resulting normalized wave function equals 1: this is the
famous normalization condition, which must be obeyed by all quantum mechanical wave functions:
_

(x, t)(x, t) dx = 1 normalization condition (1D). (A.2b)


Imposing this condition ensures that the total probability for nding the particle somewhere in the
domain < x < + will be unity, as it must be if the particle exists.
9
A.3.2 The interpretation of the wave function
Postulate 2 (The Born interpretation.)
We interpret the wave function as a position probability amplitude. In a one-dimensional state represented
by a wave function (x, t),
T(x, t) dx = [(x, t)[
2
dx =

(x, t) (x, t) dx, position probability density, (A.3)


is the probability that in an ensemble measurement of position at time t the particle will be detected in the
innitesimal volume element dx at x. Thus the position probability density [(x, t)[
2
is the probability per
unit length for detection of the particle will be detected in dx at x in a measurement at time t.
The generalization of this postulate to single-particle systems in two (2D) and three dimensions (3D)
is straightforward. In 3D, for example, the wave function depends on three spatial coordinates (x, y, z),
which we can collectively denote by r. According to the Born interpretation, the probability density
T(x, y, z; t) = [(x, y, z, t[
2
times the volume element d
3
v = dxdy dz is the probability per unit volume
for detecting the particle in volume element d
3
v at r at time t. Its generalization to many particles is
addressed in Chap. 9.
A.3.3 The integrated probability density
Actual position measurements concern nite regions of space. Still, however good our apparatus, it cant
detect particles in an innitesimal region. So when we apply Postulate II to actual experiments, we always
construct an integrated probability density. If, for example, the resolution of our apparatus can detect
the particle in a nite region [a, b], then the relevant quantity is
T([a, b], t) =
_
b
a
T(x, t) dx =
_
b
a
[(x, t)[
2
dx, integrated probability density. (A.4)
9
Details: We relax this requirement a bit for a continuum stationary state. The wave function of such a state cannot
be normalized because it pervades all space. The normalizability requirement for such a state is that the modulus of the
wave function must be nite for all values of its variables. To illustrate, consider a continuum stationary state of energy
E =
2
k
2
/(2m) and spatial function
k
(x). Instead of the usual normalization condition (A.2b), the condition for a continuum
stationary state must be written in terms of the Dirac delta function (Appendix N), as
_

k
(x)
k
(x) dx (k

k),
where the constant of proportionality is arbitrary. This condition is called Dirac delta function normalization. To learn
more about continuum states and their interpretation see A.4 and 8.5 in Understanding Quantum Physics (UQP).
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 7
A.3.4 Expectation values and uncertainties
The probabilistic interpretation of the wave function provides the foundation for calculating quantities that
characterize an ensemble as a whole. Although quantum mechanics gives only probabilistic information about
the results of measurement, it gives precise information about statistical properties of the ensemble. The
two most important such properties are the expectation value and the uncertainty of an observable.
10
In quantum mechanics, the average value of an observable (in a state) is its expectation value and the
standard deviation is its uncertainty. For position, say, the expectation value and the uncertainty from
the interpretation of (x, t) as a position probability amplitude, as well now see.
The expectation value. The ensemble average or expectation value of position is given by the
expectation value of the position operator x with respect to the wave function:
x(t)
_

xT(x, t) dx =
_

(x, t) x (x, t) dx. average position at t (1D) (A.5)


As its name implies, x(t) is the average position we would expect to get in an ensemble measurement of
this observable at time t.
The uncertainty. In general, the results of an ensemble measurement will be dispersed about the expecta-
tion value in a pattern characterized by the uncertainty (x)(t). Like the expectation value,l we calculate
the uncertainty from the wave function. For position in 1D, the uncertainty is
11
(x)(t)
_
(x x)
2
=
_
x
2
x
2
. position uncertainty at t (1D) (A.6)
In writing expressions that contain expectation values and other quantum mechanical devices, we often
resort to an elegant shorthand devised by the brilliant theoretician P. A. M. Dirac during the early years of
quantum theory (see Appendix G).
A.3.5 Probabilistic & statistical information about momentum
Armed with the Born interpretation and the denitions of the mean position and uncertainty, we can gure
out all that nature allows about position. But buried in the wave function is all information about the
state: contains information about all observables of the system. Happily, quantum mechanics provides
a powerful, general procedurethe method of eigenfunction expansionfor determining information
about any observable. One way to view this method is as a translation procedure: we translate the wave
function into the language of whatever observable were interested in, then apply an extension of the Born
interpretation to the result (Postulate II).
Before we review this procedure (see A.5), lets recall the familiar example of the linear momentum.
We can translate the position probability amplitude (x, t) into a momentum probability amplitude
(p) using the Fourier transform. This transform relates the initial wave function to an momentum
amplitude function A(k) by
(x, 0) =
1

2
_

A(k) e
i kx
dk. (A.7a)
We dene a momentum probability amplitude (p) by interpreting k as the wavenumber k = p/ of
the particle and writing
10
Jargon: Somewhat imprecisely, at least from the viewpoint of a statistician, the uncertainty is also often called the standard
deviation or the dispersion.
11
Notation: The argument (t) denotes the time dependence of (x)(t); it does not imply that x is a function of time. Ill use
this argument only when I want to emphasize this time dependence.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 8
(p)
1

A
_
p

_
, momentum probability amplitude. (A.7b)
The Fourier transform now directly relates the position and momentum amplitudes:
(x, 0) =
1

2
_

(p) e
i px/
dp. (A.7c)
If we want to calculate the momentum amplitude from the initial wave function, we just use the inverse
relation
12
(p) =
1

2
_

(x, 0) e
i px/
dx. momentum amplitude (A.7d)
At subsequent times t > 0, the relationship between the position and momentum amplitudes follows from
the time-dependent Schrodinger equation (TDSE) (see Postulate IV):
(x, t) =
1

2
_

(p) e
i [px(p)t]
dp. (A.8)
The function (p) is the dispersion relation. We determine this function from the appropriate expression
for the total energy. Thus for a free particle, the dispersion relation is
(p) =
p
2
2m
=
k
2
2m
. (free particle dispersion relation) (A.9)
Using F as shorthand for a Fourier transform, we can encapsulate Eqs. (A.7) in the easily remembered form
(x, 0) = F
1
[(p)]
(p) = F[(x, 0)]
(A.10)
When we refer to (p) as a momentum probability amplitude, we mean that from it we can
construct the (1D) momentum probability density at time t. The mathematical structure of this
momentum density is identical to that of the position density in Postulate II:
T(p, t) dp = [(p)[
2
dp =

(p)(p) dp, momentum probability density (A.11)


Like all probability amplitudes, this function must be normalized,
_

(p)(p) dp = 1. (A.12)
Thus we have translated the wave function into a function of momentum and generalized the Born inter-
pretation to this observable. We can similarly treat any other observable, be it discrete or continuous.
Armed with the momentum probability amplitude (p), we can calculate the expectation value and
uncertainty of the linear momentum using relationships analogous to those for position. The momentum
expectation value for a system with momentum amplitude (p) is
p
_

(p) p (p) dp, (A.13a)


where the action of the momentum operator p on a function of momentum is simply to multiply the function
by the variable p. (The action of p on a function of position is quite dierent.) The momentum uncertainty
12
Jargon: The new function (p) is sometimes called the wave function in momentum space or the momentum repre-
sentation of the state. Note that p is the variable conjugate to x; as such, the two are related by the commutator relation
_
x, p
_
= i . Such relationships are characteristic of conjugate variables in quantum mechanics.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 9
is dened by
p
_

_
p p
_
2
=
_
p
2
p
2
. (A.13b)
In practice, its usually easier to evaluate p and p directly from the wave function (x, t) using the
momentum operator (see Postulate III)
p = i
d
dx
. (A.14)
For example, an equivalent expression for the average momentum is
p = i
_

(x, t)
_
d
dx
(x, t)
_
dx, momentum operator (1D). (A.15)
A.3.6 Statistical quantities for arbitrary observables
By extension of Eqs. (A.5) and (A.6) for position and Eqs. (A.15) and (A.13b) for momentum, we can
evaluate the expectation value and uncertainty of an arbitrary observable Q:
Q =
_

(x, t)

Q(x, t) dx, (A.16a)


Q =
_
Q
2
Q
2
. (A.16b)
As noted above, the quantum mechanical uncertainty is analogous to the statistical standard deviation of an
observable of a state of a macroscopic system. In quantum mechanics, however, the uncertainty implicitly
refers to the standard deviation of individual results obtained in an ensemble measurement about their
average value. This average value is given by the expectation value Q.
A.3.7 Observables and operators
Postulate 3 (Operators)
Every observable of a system is represented by an operator. The operator is the mathematical tool we use
to extract information about the observable from wave functions. For an observable that is represented in
classical physics by the function Q(x, p), the corresponding operator is

Q(x, p), where the operators x and p
appear in Tbl. A.1.
Table A.1. A dictionary of operators for a
one-dimensional system. Each operator acts,
in general, on a wave function (x, t).
Observable Operator Instructions
position x multiply by x
momentum p i

x
total energy

1

T +

V (x)
kinetic energy

T

2
2m

2
x
2
potential energy

V multiply by V (x, t)
parity

invert x through the origin
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 10
Restrictions on operators. Just as we must restrict the class of functions that can serve as probability
amplitudes (to those that are continuous, normalizable, and single-valued), we must restrict the class of
operators that can represent observables:
(1) All quantum mechanical operators must be Hermitian.
(2) All quantum mechanical operators must be linear (or anti-linear).
13
Dos and donts of operator manipulation. Operators are so important that we must keep before us
the rules for manipulating them. Here are the key ones (for more, see Appendix H):
(1) Derive and simplify operator expressions and equations by letting operators act on an
arbitrary physically admissible function of all the relevant variables. For instance, a one-
dimensional operator expressed in terms of x and its derivatives should act on an arbitrary func-
tion f(x).
(2) Operators act on everything to their right unless their action is constrained by parentheses
or brackets.
(3) The product of two operators is a third operator.
(4) An operator product implies successive operation.
(5) The order in which the operators act is vital; never assume that two operators commute.
A.3.8 Hermiticity
In 1D an operator

Q is Hermitian if, for arbitrary well-behaved wave functions
1
(x, t) and
2
(x, t),
_

1
(x, t)
_

Q
2
(x, t)
_
dx =
_

Q
1
(x, t)

2
(x, t) dx. (A.17a)
In Dirac notation, we write this denition as

1
(t) [

Q
2
(t) =

Q
1
(t) [
2
(t)
1
(t) [

Q [
2
(t). (A.17b)
The double bar notation in the matrix element
1
(t) [

Q [
2
(t) signies that

Q can act on either

1
(x, t) or on
2
(x, t). The Hermiticity of an operator

Qa property of an operator, not of states of a
systemcan be expressed even more abstractly using the adjoint notation (Appendix H) as

Q

=

Q.
14
Consequences of Hermiticity. Hermiticity is an exceptionally powerful property. From it follow some of
the most important results of quantum mechanics. These results concern solutions of the eigenvalue equation
of an Hermitian operator.
(1) The eigenvalues of a Hermitian operator are real. The eigenvalues of an operator are the only
values that can be obtained in a measurement of the corresponding observable. Since all measurable
values are real numbers, we must use only Hermitian operators.
(2) The eigenfunctions of a Hermitian operator are orthogonal.
15
Two functions are orthogonal
if their overlap integral is zero. Two bound-state eigenfunctions of a one-dimensional Hamiltonian, for
example, satisfy this property:
13
Details: All operators that can be expressed in terms of the position and momentum operators are linear. Moreover, all
operators that represent continuous transformations of position coordinates are linear. The most important anti-linear operator
is the time reversal operator. This operator, which switches the signs of momentum (and spin) variables, is extensively used
in the quantum theory of scattering (see Chap. 1).
14
Jargon: Mathematicians often call a Hermitian operator a self-adjoint operator.
15
Details: This is a slight overstatement. If a particular eigenvalue is degenerate, which means that two or more eigenfunctions
share this eigenvalue, then these degenerate eigenfunctions may not be orthogonal. But if they arent we can always construct
an equal number of orthogonal eigenfunctions, each of which corresponds to the same eigenvalue, as linear combinations of the
functions in the original set. A common procedure for doing so is the Gram-Schmidt orthogonalization procedure.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 11

n
[
n
=
_

n
(x)
n
(x) dx = 0, (if n

,= n). (A.18a)
As always, these eigenfunctions must be normalized. A combined statement that the eigenfunctions of

1 are orthogonal and normalized is the orthonormality condition


16

n
[
n
=
_

n
(x)
n
(x) dx =
n

,n
(A.18b)
where the Kronecker delta function is

,n
=
_
0 if n

,= n
1 if n

= n
Kronecker delta function (A.18c)
Orthonormality plays a key role in the method of eigenfunction expansion mentioned above. We
use orthogonality to determine the coecients of an expansion in the complete set of eigenfunctions of
a Hermitian operator.
(3) The eigenfunctions of a Hermitian operator constitute a complete set. Completeness
is the heart of the method of eigenfunction expansion, since it justies our writing down the
expansion in the rst place! Completeness of a set of functions means that we can expand an (almost)
arbitrary function in the elements of the set. The function we want to expand must be mathematically
well-behaved and must obey the same boundary conditions as the functions in which were expanding
it. In quantum mechanics, this means that the function must satisfy the usual conditions of physical
admissibility (continuous, smoothly varying, and so on), and must obey the same boundary conditions
as the elements of the set in which were going to expand it.
17
This set is called the basis (or basis
set) for the expansion. Well see an important example of this principle in A.5.
(4) The eigenfunctions of a Hermitian operator satisfy closure. Although Ive singled out this
property, its just another way to express completeness; closure is a property of eigenfunctions at two
dierent positions. For a complete set of eigenfunctions
E
(x) , this property reads

E
(x

)
E
(x) = (x

x). closure (A.19)


16
Details: In the language of linear algebra, the orthogonality integral for two functions is called their inner product or
scalar product. Another way to state the orthogonality condition is to say that the scalar product of two eigenfunctions with
dierent eigenvalues is zero. Another way to state the normalization condition is to say that the scalar product of a normalized
eigenfunction with itself is 1. The latter quantity is usually called the norm of the function, so the normalization condition
amounts to the easily remembered statement that the norm of a normalized eigenfunction is 1. Well use such language in
studying electron spin in Chap. 6.
17
Details: For example, we couldnt expand a function that went to innity as x in a basis of eigenfunctions of the simple
harmonic oscillator Hamiltonian. In practice, this isnt an issue, because such a function wouldnt be physically admissible.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 12
A.3.9 Linearity and the principle of superposition
An operator

Q is linear if, for any functions
1
and
2
and complex number a and b (Appendix H),

Q(a
1
+ b
2
) = a

Q
1
+ b

Q
2
. a linear operator. (A.20)
Physically, this property of quantum mechanical operators is related to the Principle of Superposition,
Eq. (A.1), p. 5. In fact, this principle holds only because the Hamiltonian, which determines wave functions
via the TDSE, is a linear operator.
A.3.10 The commutator
We can determine whether two operators commute by evaluating their commutator. For operators

Q
1
and

Q
2
, the commutator is a third operator, dened as
_

Q
1
,

Q
2


Q
1

Q
2


Q
2

Q
1
. commutator (A.21)
If

Q
1
and

Q
2
commute, their commutator
_

Q
1
,

Q
2

is zero; if they dont, its non-zero:


_

Q
1
,

Q
2

= 0, commuting operators, (A.22a)


_

Q
1
,

Q
2

,= 0, non-commuting operators. (A.22b)


A key property of commuting operators is:
Rule: Operators that commute dene a complete set of simultaneous eigenfunctions.
Warning: Unlike numbers and functions, operators do not, in general, commute.
A.3.11 Eigenstates and simultaneous eigenfunctions
To understand the above rule, we need to recall the denition of a simultaneous eigenfunction. Our
starting point is the notion of an eigenstate. A system in state (x, t) is an eigenstate of an observable
Q if (x, t) is an eigenfunction of

Q:

Q
q
(x, t) = q
q
(x, t), system is in an eigenstate of Q. (A.23)
In this case, we often use the eigenvalue q as a subscript to label the wave function, as above.
In an eigenstate of Q, this observable is said to be sharp. This term has a very specic meaning: in an
ensemble measurement of Q, all members of the ensemble will give the same valuethe eigenvalue q. In an
eigenstate of the observable Q, the uncertainty of Q is therefore zero: Q = 0.
The most familiar eigenstate is the energy eigenstate. A system is in an energy eigenstate if its wave
function is an eigenfunction of the Hamiltonian. Another name for an energy eigenstate is a stationary
state (A.4).
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 13
A function
q,r
is said to be a simultaneous eigenfunction of two operators

Q and

R if it satises the
eigenvalue equations of both operators,
18

Q
q,r
= q
q,r

R
q,r
= r
q,r
.
a simultaneous eigenfunction (A.24a)
The converse also holds: any two operators that share a complete set of simultaneous eigenfunctions neces-
sarily commute. Well return to the remarkable consequences of this rule in A.7.
A.3.12 Commutators and uncertainty relations
As noted in A.2, if two operators commute, then we say that the corresponding observables are compatible.
If they dont commute, the observables are incompatible. These colorful names allude to the possibility
of measuring both observables simultaneously to innite precision (in principle). Incompatible observables
are subject to the constraints of an uncertainty relation. The most famous example of incompatible observ-
ables are position and momentum. The commutator of the corresponding operators is
_
x, p

= i . The
consequence of the incompatibility of x and p are that any quantum state of any system is subject to the
Heisenberg uncertainty principle
(x)(t) (p
x
)(t)
1
2
. Heisenberg Uncertainty Principle (A.25)
Rule: Uncertainty relations like Eq. (A.25) are general statements about systems; they are not limited to
particular states. In no state of any system is the uncertainty product (x)(t) (p
x
)(t) at any time less
than /2.
The Heisenberg uncertainty principle illustrates a more general result that pertains to any two quantum
mechanical operators, whether or not they commute. According to this Generalized Uncertainty Prin-
ciple (GUP), we calculate the uncertainty product for operators

Q
1
and

Q
2
from their commutator, as
19
(Q
1
) (Q
2
)
1
2

i
_

Q
1
,

Q
2

. Generalized Uncertainty Principle (A.26)


A.3.13 Constants of the motion
In quantum mechanics as in classical physics, certain observables have a special status. These are the con-
stants of the motion: physical quantities that remain unchanged as the system evolves with time. Knowledge
of the constants of the motion provides profound and powerful insights into the systemespecially into its
symmetry properties (see Chap. 1).
The meaning of constant of the motion in quantum physics is inherently statistical. An observable is a
constant of the motion if the expectation value of the observable for any state doesnt depend on time:
d
dt
Q =
d
dt
(t) [

Q [ (t) = 0 for any . constant of the motion (A.27)
18
Details: A familiar example is a stationary state of a particle in a one-dimensional symmetric potential. Since the spatial
function of such a state has denite parity, the stationary-state wave function is an eigenfunction of the Hamiltonian and the
parity operator.
19
Read on: You can nd a discussion of this important result in 11.4 in Morrison (1990) and remarks on its consequences
for measurement in 13.1 of that book.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 14
To determine whether an observable is a constant of the motion we call on the commutator of the
corresponding operator with the systems Hamiltonian. The connection between these quantities is the
general relationship
d
dt
Q =
i

_

1,

Q

+
_

Q
t
_
(A.28)
Unless the operator

Q explicitly depends on t (like the interaction Hamiltonians for time-dependent elds
in Chap. 18), the last term on the right-hand side of Eq. (A.28) is zero and, if the commutator
_

1,

Q

is
also zero, Eq. (A.27) immediately follows. But Q may also be a constant of the motion if the commutator
is non-zero, provided its expectation value with respect to every state of the system is zero.
A.3.14 The Correspondence principle
Since its initial development early in the 20th century, physicists have demanded that the equations of
quantum mechanics must go over continuously into their classical counterparts in the classical limit. This
is the essence of the Correspondence Principle:
Rule: In the classical limit, the laws of quantum mechanics must reduce to those of Newtonian Mechanics.
Here the phrase the classical limit is a bit vague, although we can make it specic when we discuss
particular systems. For example, in exploring the stationary states of a one-dimension simple harmonic
oscillator potential, we nd that as the quantum number n increases, the quantum mechanical probability
density [
n
(x)[
2
increasingly resembles the classical probability. Here the classical limit is the limit of large
quantum numbers. In other systems, the classical limit refers to large spatial dimensions or to the limit in
which quantities involving are negligible.
20
What does the Correspondence Principle imply for the expectation values and uncertainties we use
to characterize ensemble measurements? These quantities express the essentially non-classical nature of
quantum physics, its probabilistic character. In the classical limit, all evidence of this character must
disappear:
Q(t) Q(t)
Q(t) 0.
(in the classical limit) (A.29)
Ehrenfests Theorem. The Correspondence Principle has a subtle, fascinating implication for the classi-
cal idea of a particle trajectory. In some sense, the position probability density, the squared modulus of the
wave function, must reduce in the classical limit to the trajectory the particle would have were it described
by classical mechanics.
21
The quantum mechanical articulation of this idea are the equations of Ehrenfests
Theorem. These equations are written in terms of the expectation values of position and momentum for
a given quantum state. To clarify the connection to classical equations youve seen before, Ill write these
equations for a particle in 3D:
22
20
Jargon: Most commonly, youll see the classical limit indicated in sentences like, Quantum equations of motion must
reduce to their classical counterparts in the limit 0. Such locutions are not to be taken literally! Plancks constant is a
constant, so strictly speaking, writing 0 is meaningless. Rather, this limit is a shorthand to indicate that terms involving
are being neglected compared to other terms, which must be signicantly larger for this limit to apply.
21
Read on: Strictly speaking, this statement should be couched in terms of the ensemble of microscopic particles whose
common state is described by the wave function. That is, in the classical limit the quantum mechanical position probability
distribution must reduce to the corresponding statistical probability distribution for an ensemble of classical particles, each of
which has a trajectory. For more on this topic, see Sec. 15.1 of Quantum Mechanics, by Leslie E. Ballentine (Englewood Clis,
NJ: Prentice-Hall, Inc. 1990).
22
Details: The second of these equations holds rigorously only if the corresponding classical expression V (r) is linear
with respect to position, as in the case of a simple harmonic oscillator. More generally and rigorously, the right-hand-side
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 15
d
dt
r(t) =
1
m
p(t) (A.30a)
d
dt
p(t) = V (r). (A.30b)
A.3.15 The time-development of a quantum state
Postulate 4 (The Schrodinger equation.)
In the absence of measurement, the time development of the wave functions of a system is governed by the
time-dependent Schr odinger equation (TDSE),

= i

t
. (A.31)
The Hamiltonian

=

T +

V fully represents the system. To specify a particular state of the system, we must
give the initial condition: the value of the wave function at one time (usually t = 0).
The Hamiltonian, the operator that corresponds to the total energy of the system, is the sum of the kinetic
and potential energy operators. Provided no external magnetic elds act on the system, the kinetic energy
operator is just T = p
2
/(2m), and the Hamiltonian has the form
23

1 =

T +

V =
p
2
2m
. Hamiltonian (A.32)
A.3.16 The probability current density
Discussions of the time development of a state often refer to a quantity which contains information comple-
mentary to that of the probability density: the probability current density. This quantity plays a key
role in understanding the magnetic moment of a microscopic system (see Complement ?? of Chap. 5). In
1D, the probability current density is dened in terms of the wave function by
j(x, t)
i
2m
_

(x, t)

x
(x, t) (x, t)

x

(x, t)
_
. (A.33)
Conservation of probability. To illustrate the importance of the probability density, lets consider the
reasonable requirement that the probability associated with a particle must be conserved. We can articulate
this important law in two ways: conservation at a point and conservation in a nite region. In 1D, we write
the law of pointwise conservation of position probability as

t
T(x, t) +

x
j(x, t) = 0. (A.34a)
Integrating this equation in a nite region a x b yields
d
dt
T
a,b
([a, b], t) = j(a, t) j(b, t). (A.34b)
By extending the region [a, b] to encompass the entire domain of x, we regain the reassuring result that the
total probability of nding a particle somewhere in space doesnt change with time:
of Eq. (A.30b) is V (r). In most applications, the position uncertainty is small enough (compared to the range of the
potential energy) that the approximation in Eq. (A.30b) is excellent.
23
Details: In the presence of an external magnetic eld B with vector potential A, the momentum operator becomes [p +eA]
2
.
(This alteration is required by classical electrodynamics and is not peculiar to quantum mechanics.) This changes the kinetic
energy operator and hence the Hamiltonian, as described in Chaps. 13 and 18.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 16
d
dt
T([a, b], t) =
d
dt
_

T(x, t) dx = 0. (A.34c)
A.4 Stationary states
Stationary states are the simplest states in quantum mechanics. They are also the starting point for most
strategies for studying non-stationary states (see A.5). The dening property of a stationary state is the
mathematical structure of its wave function. Its signature is a wave function that is a separable function
of space and time with an exponential time dependence:

E
(x, t) =
E
(x) e
i Et/
. stationary state wave function (A.1)
Solutions of the TDSE of this form exist only if the potential energy term in the Hamiltonian

1 =

T +

V
does not depend on time. If so, then Eq. (A.1) follows from the TDSE by application of separation of
variablesa technique well use repeatedly in this book. If not, see Chaps. 17 and 18.
A.4.1 The physics of stationary states
The physical properties of a stationary state follows from Eq. (A.1). The time dependence of
E
(x, t) appears
entirely in the complex phase factor e
i Et/
. Therefore the product of
E
(x, t) and its complex conjugate
does not depend on time. Physically, therefore, the probability density of a stationary state is independent
of time. (Thats why physicists call the state stationary.) Consequently the statistical properties of a
stateits expectation value and uncertaintyare also independent of time. More generally, we have the
following important rule:
24
Rule: In a stationary state, all physical properties of the system are independent of time.
This rule pertains to all observablesnot just the energy. For any observable Q, Q and Q are independent
of time in a stationary state.
The energy of stationary state has a very special property: it is sharpit has a well-dened value, and
that value is the particular eigenvalue of

1 that appears in the phase factor e
i Et/
:
E = E, and E = 0 (in a stationary state) (A.2)
The spatial function. The spatial function
E
(x) in the wave function
E
(x, t) fully describes the
spatial properties of the state, such as position and momentum. Mathematically,
E
(x) is the eigenfunction of
the Hamiltonian with eigenvalue E. We determine spatial functions by solving the Hamiltonian eigenvalue
equation, the time-independent Schrodinger equation (TISE)

1
E
(x) = E
E
(x), time-independent Schrodinger equation . (A.3)
Rule: The TISE is is a linear, homogeneous, second-order ordinary dierential equation.
(In more than 1D and/or for more than one particle, its a partial dierential equation.) To specify its
solutions, therefore, we must specify two boundary conditions, two values of
E
(x). For a 1D system,
these conditions are specied in the asymptotic limits x . That is, rather than specify the value of
the function at two nite points, we specify its values in the limits of the domain < x < +.
24
Commentary: Even the probability current density j(x, t) of a stationary state is independent of time. But its not necessarily
zero. Indeed, in Chap. 5 we explore stationary states of atomic hydrogen in which a non-zero (but time independent) probability
current density leads to the orbital magnetic moment of the atom.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 17
An integral form of the TISE. In some methods for solving the TISEnotably the variational
method of Chap. 16its useful to transform this dierential equation into an integral for the energy.
This transformation is not restricted to stationary states. For an arbitrary quantum state with (normalized)
wave function (x, t), the expectation value of the energy at t is, by denition,
E(t) =
_

(x, t)

1(x, t) dx (A.4)
For a stationary state, the energy is sharp with value E. So the expectation valuethe average value of the
energy in this stateis time independent and evaluates to E:
E =
_

n
(x)

1
n
(x) dx = E. (for a stationary state) (A.5a)
Using Dirac notation as a shorthand for integration we write
25
E = [

1 [ = E. (for a stationary state) (A.5b)
A.4.2 Boundary conditions: bound versus continuum states
As noted in A.1, quantum states come in two varieties, bound and continuum. Amongst stationary states,
the two varieties are distinguished by the asymptotic boundary conditions on their spatial functions,
that is, the behavior of
E
(x) in the limits x . For a 1D system, boundary conditions are

n
(x)
x
0, bound stationary state, (A.6a)
lim
x
[
E
(x)[ < , continuum stationary state. (A.6b)
The continuum state boundary condition states that the magnitude of the spatial function must remain
nite for all values of r. Mathematicians say that such a function is bounded at innity.
26
Energy quantization. For a bound stationary state, the boundary conditions Eq. (A.6a) have profound
consequences. Most important, the energy of a bound stationary state is quantized. Only for certain, discrete
values of E do there exist solutions of the TISE that obey this boundary condition. Only these solutions are
normalizable, and hence only these solutions correspond to physically realizable states. Since each bound-
state energy is quantized, we can index the energy and the spatial function by an integer. This integer is
called the principal quantum number n. For the minimum value of n, the index for the ground state,
we choose n = 1, except when its not.
27
In two and 3D, to construct a unique name for a Hamiltonian
eigenfunction requires more than one quantum number.
25
A cautionary note: Notice that the normalized spatial function
E
(x) appears in the integral form of the TISE for a
stationary state.
26
Read on: We cant normalize the spatial function for a continuum stationary state. Nevertheless, we can extend our
interpretation of the wave function as a position probability amplitude to interpret continuum spatial functions as amplitudes
for relative probabilities. In 1D, the transmission coecient T(E) for scattering of a particle with incident kinetic energy E
by a potential well or barrier is dened as a ratio of continuum spatial functions (pure momentum state functions e
i kx
).
Hence the arbitrary normalization of the spatial function divides out of T(E), which is the physically measurable quantity. For
further discussion, see 8.5 in Morrison (1990).
27
A cautionary note: A prominent exception to this convention is the simple harmonic oscillator, where for reasons of
convenience we start indexing at n = 0. See Appendix P.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 18
A.4.3 Bound states in one dimension
For a single particle in 1D, the bound-state solutions of the TISEthe spatial functions
n
(x) which satisfy
the bound-state boundary conditions (A.6a)have properties that aid us in problem solving and physical
thinking.
28
(1) All bound-state energies are non-degenerate. To a non-degenerate eigenvalue E
n
, there corre-
sponds one and only one mathematically distinct spatial function
n
(x). By contrast, each continuum
state energy E > 0 is two-fold degenerate. For each such energy, there are two linearly independent
spatial functions that satisfy the TISE and the continuum boundary conditions (A.6b).
29
(2) There exists at least one bound state in any attractive potential. An attractive potential
V (x) has the property that, if we choose the zero of energy at the maximum value of the potential,
then the potential will be negative for part of the domain < x < +. For any such potentialno
matter how weakwe can nd at least one solution of the TISE that obeys the bound-state boundary
conditions. We say that such a potential supports at least one bound state.
30
(3) The bound states of a symmetric potential have denite parity. If a potential is symmetric
under inversion, then V (x) = V (x), and each of its bound states has denite parity. This prop-
erty describes the behavior of the Hamiltonian eigenfunction
n
(x) under inversion (Chap. 1):
: x x

= x. Inverting the coordinate x transforms functions of x, in general into dierent func-


tions. An arbitrary function f(x) will thus be transformed into a new function g(x):
f(x) g(x) = f(x). transformation under inversion (A.7a)
The equality in this equation is vital: it denes the transformed function g(x) as the function whose
value at x is equal to that of the untransformed function f at the point x

that results from the inversion.


The operator that eects this transformation of functions is the parity operator

:

f(x) = g(x) = f(x). parity operator (A.7b)


If the function f(x) has denite parity, the Eq. (A.7b) becomes an eigenvalue equation, with eigenvalues
1. A function with eigenvalue +1 is an even function and one with 1 is an odd function:

f(x) = f(x)
_
+ even
odd
(A.7c)
Rule: An even function is unaected by inversion; an odd function changes sign under inversion. A
function f(x) whose transform under inversion, g(x), is neither even nor odd does not have denite parity.
Rule: In 1D, the bound-state spatial functions of a symmetric potential are all even or odd.
31
States whose spatial functions are even are said to have even parity; those whose spatial functions are
odd have odd parity. The classication of the stationary states of a symmetric potential into even-
and odd-parity states is a powerful tool for understanding and simplifying the analysis of all states,
stationary and non-stationary.
28
A cautionary note: These properties pertain to single-particle systems in 1D only. For generalization to other systems,
see Chaps. 25.
29
Jargon: A set consisting of two or more functions, say { f
1
, f
2
, . . . , f
N
} is linearly independent if the only set of constants
{ c
1
, c
2
, . . . , c
N
} such that

N
j
c
j
f
j
= 0 is the set of zeroes, c
1
= c
2
= = 0. (This condition must hold for all values of
the independent variable(s) of f.) If even one coecient is non-zero, then the set is said to be linearly dependent. For
instance, { sin, cos } is linearly independent, because neither function in the set is proportional to the other function. But
{ sin , cos , e
i
} is linearly dependent, because e
i
= cos +i sin .
30
A cautionary note: This characterization doesnt apply to innite potentials like the innite square well or the simple
harmonic oscillator. For innite potentials, which are useful models but cannot be realized in nature, we choose the zero of
energy at the bottom of the potential. Innite potentials support an innite number of bound states and have no continuum.
31
Details: The spatial functions for a non-symmetric potential dont have denite parity.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 19
A.4.4 The stationary states of a free particle
The simplest quantum mechanical system is the free particle: a particle in the absence of any forces
whatsoever. Since a potential energy V = 0 cannot bind the particle, all its quantum states are continuum
states. In 1D, these states are called pure momentum states, because they are eigenstates of linear
momentum.
32
The free particle in one dimension. The Hamiltonian for a free particle is just the kinetic energy
operator,

1 =

T. We can write its eigenfunctions either in terms of the wavenumber k or the linear momen-
tum p = k:

k
(x) =
1

2
e
i kx
, or
k
(x) =
1

2
e
i px/
. (A.8)
The factors 1/

2 in
k
(x) and 1/

2 in
p
(x) ensure that these functions satisfy the Dirac delta
function normalization condition for continuum states,
_

k
(x)
k
(x) dx = (k

k). (A.9)
The wave function of a pure momentum state is the product of
k
(x) and the usual stationary-state
time factor e
i Et/
. We can write this wave function as a travelling wave using the de BroglieEinstein
relations (A.1). For a free particle of mass m, we have
E = =

2
k
2
2m
.
=
k
(x, t) =
k
(x) e
i Et/
=
1

2
e
(i kxt)
.
(for a free particle) (A.10)
It is sometimes convenient to write
k
(x, t) in terms of p and E rather than k and . We introduce an
additional factor of 1/

to ensure that the normalization condition (A.9) still holds, to wit:

p
(x, t) =
1

2
e
i (pxEt)/
.
pure momentum state wave
function for a free particle.
(A.11)
Each pure momentum state is two-fold degenerate, because both momenta p correspond to the same free
particle energy, E = p
2
/(2m).
The free particle in 3D. For a free particle in 3D with linear momentum
33
p = k = p
x
e
x
+ p
y
e
y
+ p
z
e
z
, (A.12)
we can separate variables x, y, and z in the TISE,

1
k
(x) =

T
k
(x) =

2
2m
_

2
x
2
+

2
y
2
+

2
z
2
_

k
(x). = E
k
(x). (A.13)
The resulting spatial function
k
(x) is the product of three single-variable functions, one for each coordinate,
each of the form of (A.8). Combining the exponential functions in this product, we get
32
Commentary: The Heisenberg Uncertainty Principle implies that in a pure momentum state we know nothing about position,
x = . Although an idealization, such states are useful in, for example, the description of collisions.
33
Commentary: In 3D the momentum p and the wave vector k are vectors. So when we say p is sharp, we mean that
all three Cartesian components of the linear momentum are sharp, px = py = pz = 0. In 1D, of course, the momentum
and wavenumber in these equations are scalars, p p and k k. The symbols ex, ey, and ez denote unit vectors along the
Cartesian axes.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 20

k
(x) =
1
(2)
3/2
e
i kr
, (A.14a)
The Dirac delta function normalization condition (A.9) becomes the three-dimensional integral
34
_

k
(x)
k
(x) d
3
v =
3
(k

k). (A.14c)
The corresponding wave function for a free particle in three dimensions is

k
(r, t) =
1
(2)
3/2
e
i (krt)
, (A.14d)
where the frequency is related to the magnitude k of the wave vector k by = [k[
2
/2m.
Physical properties of a pure momentum state.
(1) The linear momentum of a pure momentum state is sharp, p = 0, with value p = k.
(2) The energy of a pure momentum state is sharp (E = 0) with value E =
2
k
2
/2m, where k = [k[ is
the magnitude of the wave vector.
(3) The spatial function
k
(r) is an eigenfunction of

1 =

T and p =

k.
A.5 Non-stationary states and the method of eigenfunction expansion
Not all systems have stationary states. If the Hamiltonian of a system depends on time, then all of its
quantum states are non-stationary (see Chap. 17). Even for a system that has stationary states, many of
the most interesting states are non-stationary. So every quantum mechanician needs to know how to use
stationary states to analyze non-stationary states.
A non-stationary state is everything a stationary state is not. In a non-stationary state, the energy
isnt sharp [E > 0]. Both the position probability density [(r, t)[
2
and the current density j(r, t) vary
with time, as do the probabilities, expectation values, and uncertainties for most observables of the system.
Working with non-stationary states is especially challenging because their wave functions cannot be sepa-
rated in space and time; we must therefore solve the TDSE. To do so we invoke the powerful method of
eigenfunction expansion.
A.5.1 The method of eigenfunction expansion
Because the eigenfunctions of the Hamiltonian constitute a complete set, we can expand any function that
satises appropriate boundary conditions in this set. To implement the method of eigenfunction expan-
sion for a non-stationary state wave function (x, t), we proceed as follows:
(1) Expand the initial wave function (x, 0) in the complete set of Hamiltonian eigenfunc-
tions
E
(x) :
(x, 0) =

n
c
n
(0)
n
(x). (A.1a)
34
Notation: The symbol
3
(k

k) denotes the Dirac delta function in 3D. In Cartesian coordinates, this function is the
product of three 1D Dirac delta functions:

3
(k

k) = (k

x
kx) (k

y
ky) (k

z
kz). (A.14b)
Appendix N contains lots of useful stu about the Dirac delta function.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 21
This set is called the basis of the expansion. The expansion coecients c
n
(0) are, in general, complex
numbers. They carry an argument 0 to remind us that they correspond to t = 0; were we expanding
the wave function at time t, wed write c
n
(t).
(2) Evaluate all non-zero coecients c
n
(0) from the initial wave function
35
c
n
(0) =
n
[ (0) =
_

n
(x)(x, 0) dx, for all n. (A.1b)
(3) Use the resulting coecients and the eigenvalues E
n
to write the wave function for t > 0:
(x, t) =

n
c
n
(0)
n
(x) e
i E
n
t/
. (A.1c)
Thats all there is to it.
36
Warning: As in any eigenfunction expansion, we must include all of the eigenfunctions except those
whose coecients are zero.
Once we have completed Steps 1 and 2, however, we may approximate the expansion (A.1c) by neglecting
terms in this summation except those with the largest coecients. This approximation is called truncation.
Example A.1 (Derivation of the expansion coecients.)
The key to this derivation is the orthogonality of eigenfunctions of any Hermitian operatorin this case, of
the Hamiltonian, Eq. (H.6a), p. 72.
The most important step in deriving an equation for the n
th
coecient c
n
(0) in the eigenfunction expansion
Eq. (A.1a) is to isolate this coecient. That is, I must mathematically transform the sum on the right-hand
side into a single term. To do so, I multiply both sides by

n
(x), then integrate the equation with respect to
the independent variable, x:

n
(x)(x, 0) =

n
cn(0)

n
(x)n(x) (A.2a)
_

n
(x)(x, 0) dx =

n
c
n
(0)
_

n
(x)
n
(x) dx (A.2b)
Invoking orthogonality introduces a Kronecker delta function on the right-hand side, to wit:
_

n
(x)(x, 0) dx =

n
c
n
(0)
n

,n
. (A.2c)
When I now execute the sum, the Kronecker delta neatly isolates a single coecient:
_

n
(x)(x, 0) dx = c
n
(0). (A.2d)
Now I can drop the superuous prime from the principle quantum number and regain Eq. (A.1b).
35
A cautionary note: Before evaluating any of these coecients, think carefully about the symmetry properties of the Hamilto-
nian eigenfunctions and of the initial wave function; from such symmetry properties we can reason out that certain coecients
are zero and thus avoid evaluating them altogether! In the 1D quantum mechanics of a symmetric potential, parity may render
many coecients equal to zero.
36
Details: Except for articial models like the innite square well and the simple harmonic oscillator, potential energies V (x)
support both bound and continuum states. So in general we must include bound and continuum spatial functions in the initial
expansion (A.1a). Since continuum solutions of the TISE exist for any E > 0, these solutions contribute an integral, and the
generalized expansion of the initial wave function looks like
(x, 0) =

n
c
n
(0)
n
(x) +
_

0

E
(x)(x, 0) dE.
A similar form generalizes Eq. (A.1c) for t 0. See Chap. 12 in Morrison (1990) for details.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 22
A.5.2 Quantum states in the language of energy
Using Eqs. (A.1b) and (A.1c) we can mathematically represent any quantum state either by a function of
position, the wave function (x, t), or by a list of numbers, the energy expansion coecients c
n
(0) .
Either representation has the same information content; they just represent this information dierently.
Applying the translation metaphor of A.3, we can view the expansion of (x, t) in the basis
n
(x) as a
translation of the information about the state from the language of position to the language of energy.
To explain what this means we call upon the Born interpretation. According to its generalized form,
each number c
n
(t) is an energy probability amplitude. That is, the real number [c
n
(t)[
2
is the probability
that in an energy measurement at t = 0 we will detect the value E
n
: T(E
n
, t) = [c
n
(t)[
2
. The set of energy
probability amplitudes is ideal for calculation and interpretation of energy-related properties of the state.
Normalization in the language of energy. To use this interpretation properly we must ensure normal-
ization. This requirement holds regardless of how we choose to represent the stateby the wave function
(x, t), the momentum amplitude (p), or the energy amplitudes c
n
(0) . For the wave function, the
normalization condition is the integral (A.2b), p. 6. For the momentum amplitude, its the integral (A.12),
p. 8. But for the energy representation c
n
(0) its a simple sum:

n
[c
n
(0)[
2
= 1. normalization condition (A.3)
Physically, this equation states that the sum of probabilities for all possible energies of the system must be 1.
This makes sense: upon measurement of its energy, each particle must exhibit one of the values the list of
allowed energies, the eigenvalues of the Hamiltonian.
The expectation value and uncertainty in the energy. The energy amplitudes greatly facilitate
calculation of statistical energy-related properties of a non-stationary state. For instance, the mean value of
the energy at time t is the expectation value
E(t) = (t) [

1 [ (t). (A.4a)
We could, of course, evaluate this quantity directly from the wave function,
E(t) =
_

(x, t)
_


2
2m
d
2
dx
2
+ V (x)
_
(x, t) dx. (A.4b)
But this is arguably the least ecient way to perform this chore.
The easiest way to evaluate E(t) is to rst translate the wave function into the language of energy [that
is, calculate the expansion coecients a la Eq. (A.1b)]. Then we need only evaluate the sum
E(t) =

n
[c
n
(t)[
2
E
n
. average energy (A.5)
Evaluation of the energy uncertainty is equally easy, since to calculate
E =
_
E
2
E
2
, (A.6a)
we require only two expectation values, (A.5) and
E
2
(t) =

n
[c
n
(t)[
2
E
2
n
. (A.6b)
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 23
A.6 Generic problem-solving strategies
A great deal of your rst exposure to quantum physics was probably devoted to solving 1D model problems
like square wells and harmonic oscillators. Some of the techniques for solving such problems are specic to
those problems. Others, like power-series expansion of dierential equations, involve dense mathematical
detail; these we invoke as needed. But a few qualitative problem-solving gambits combine simplicity with
wide applicability. They are generic, in the sense that they apply to an enormous variety of problems. This
short section reviews such strategies.
Continuity. One consequence of the requirements that a physically admissible wave function must be
continuous and smoothly varying is that the spatial function of a stationary state and its rst spatial derivative
must be continuous at every point. We use this continuity requirement to connect pieces of a spatial
function we obtain by solving the TISE in separate regions of space.
Simplications due to symmetry. If the potential V (x) is symmetric under inversion, so that
V (x) = V (x), then we can use the parity of the Hamiltonian eigenfunctions to simplify or set to zero
many of the integrals we encounter in solving the TISE. The principles we use are
The integral over x from to + of an odd function is zero.
The integral over x from to + of an even function is twice the integral from 0 to
+ of the function.
Curvature and energy. The curvature of a function f(x) is its second derivative d
2
f/ dx
2
. Curvature
is therefore a measure of how rapidly the slope df/ dx changes with changing x. A function with larger
curvature turns over faster (per unit increment in x) than a function with smaller curvature. The sign of
the second derivative determines whether the slope of f(x) increases or decreases with changes in x.
37
According to the TISE, Eq. (A.3), p. 16, the curvature of a spatial function
E
(x) is given by
d
2

E
dx
2
=
2m

2
_
E V (x)

E
(x). (A.1)
The curvature of
E
(x) is proportional to
E
(x). Hence the sign of the curvature depends on the sign of
the function, as well as on the sign of E V (x). In a classically allowed region (CA region), E > V (x),
so d
2

E
(x)/ dx
2
has the opposite sign of
E
(x). A spatial function in a CA region is therefore oscillatory.
But in a classically forbidden region (CF region), E < V (x). Here the curvature of
E
(x) has the same
sign as the function, which is therefore an evanescent (or decaying) function. The boundary between a
CA and an adjacent CF region, the value of x where V (x) = E is called the classical turning point. At
this point, the curvature of
E
(x) is zero, and the function changes character from oscillatory to evanescent
(or vice-versa).
Equation (A.1) also shows that the curvature of the spatial function is related to its energy. Since E and
the particles de Broglie wavelength are inversely related,
E =

2
k
2
2m
=

2
(2)
2
2m
2
, (A.2)
the curvature inuences the wavelength. This connection makes sense: the faster a function turns over with
changing x (in a CA region), the smaller its wavelength.
38
Several important consequences follow from these
relations.
37
Details: In dierential geometry, the curvature at a point on a curve is dened to be the norm of the rst derivative of the
unit tangent vector at the point. That is, the curvature is the ratio of the rst derivative vector to the unit normal (the principal
normal) at the same point. The curvature is thus a quantitative, point-by-point measure of how much the curve bends. (A line
has zero curvature.) The reciprocal of the curvature is called the radius of curvature of the curve.
38
Details: Considerations of wavelength must be modied in a region where the potential energy is changing with x, since
strictly speaking a spatial function in such a region has no wavelength. To adapt to this situation we introduce the notion of a
local wavelength [see 9.2 in Morrison (1990) and Chap. 5 in this book]. The generic principles of this section then apply.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 24
The ground state has the largest allowed wavelength and no nodes. The ground state is
the stationary state whose spatial function has the largest curvature consistent with the boundary
conditions
n
(x) 0 as x . Consequently the ground state function never crosses the x axis:
it has no nodes.
39
This is also the state with the smallest allowed energy, if we measure energy from
the bottom of the potential well.
40
With increasing energy, the bound states gain additional nodes. As we go up the energy
ladder of bound states, from the ground state (with energy E
1
) to the rst excited state (E
2
) to the
second excited state (E
3
) and so, the curvature of
n
(x) increases. Since all bound-state functions
must satisfy the usual boundary conditions, each successive bound state has one additional node. If
we label the ground state with n = 1, the number of nodes for the stationary state with energy E
n
is n 1.
With increasing energy, the wavelengths of the bound states decrease. As the curvature of
an oscillatory function increases, its (local) wavelength decreases. So as we climb the energy ladder,
each successive bound state has a smaller wavelength.
A.7 Measurements in the microverse
A.7.1 Measurement and the collapse of the wave function
The TDSE fully describes the time evolution of a system with Hamiltonian

1 so long as no one performs a
measurement on the system. But neither this nor any other postulate describes what happens to the state
when a measurement is performed. Using the Born interpretation (and eigenfunction expansion) we can
predict the probabilities for various possible outcomes of an ensemble measurement but not the eect of the
measurement on the state. In this sense, the study of state evolution in quantum mechanics bifurcates into
two topics:
If no one is performing a measurement, the state evolves according to the TDSE.
At the instant a measurement is performed, the state changes (instantaneously) into the eigenstate
that corresponds to the value of the observable obtained in the measurement.
The question of what happens to a quantum state upon measurement of an observable teeters on the
precipice brink of metaphysics. According to the interpretation of quantum mechanics originally devised by
Niels Bohr and adapted to measurement by John von Neumann and others, the measurement itself transforms
the state into an eigenstate of the measured observable. Precisely how this miracle occurs, no one knows. This
instantaneous change of state as a result of measurement is called the collapse of the wave function. This
way of understanding the eect of measurement is part of the Copenhagen interpretation of quantum
mechanics.
We can begin to come to grips with this remarkable notion by thinking explicitly in terms of an ensemble.
Suppose we measure the energy of a system thats initially in a non-stationary state made up of contributions
from the ground and rst excited stationary states:
(x, 0) = c
1
(0)
1
(x) + c
2
(0)
2
(x). (A.1a)
39
Memory Jog: Recall that a node in a function f(x) is a nite value of x at which f(x) = 0. Since all bound states go to
zero in the asymptotic limits x , we dont count these as nodes.
40
A cautionary note: Be careful when you relate curvature to energy that youre aware of the zero of energy. For example,
consider a 1D symmetric nite square well that supports three bound states. If we measure energy from the bottom of the
well, then the bound-state energies are positive, with E
3
> E
2
> E
1
. But if, as is often the case, we put the zero of energy
at the top of the well, then the energies are negative and, although this string of inequalities still holds, the magnitudes of the
bound-state energies are related by |E
3
| < |E
2
| < |E
1
|. In either case, the wavelengths of the states inside the well (the CA
region) are related by
1
>
2
>
3
, with the ground state
1
(x) having the smallest curvature and the largest wavelength.
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix 25
Then at time t > 0 (absent an intervening measurement) the wave function is
(x, t) = c
1
(0)
1
(x) e
i E
1
t/
+ c
2
(0)
2
(x) e
i E
2
t/
. (A.1b)
Initially, the energy is uncertain: E(0) > 0, as it is at all subsequent timesuntil we measure the energy.
Heres what happens then.
Immediately before the measurement. From t = 0 until the instant of the measurement, all systems in
the ensemble are in the same state, with wave function (A.1b). This state has no denite, xed energy;
rather, it embodies two possible (allowed) energies, because the two energy probability amplitudes
c
1
(0) and c
2
(0) are nonzero.
At the instant of the measurement. The measurement transforms this single ensemble into two
subensembles. One subensemble consists of all particles that manifested energy E
1
in the mea-
surement; the other consists of particles that manifested E
2
. From the energy amplitudes we can
calculate the fraction of systems in the original ensemble in each subensemble: [c
1
(0)[
2
are in the
one with energy E
1
, and [c
2
(0)[
2
have energy E
2
.
Immediately after the measurement. Two distinct subensembles exist. In each subensemble, the en-
ergy is sharp (E = 0), in one with value E
1
, in the other with E
2
. Each subensemble is therefore in
an energy eigenstate. The measurement has somehow caused the wave function to collapse into
one of the two stationary-state wave functions in the initial state (A.1a).
PrintAppendices Version: 8.42 Printed: August 3, 2010
Appendix B
Dimensional extermination in quantum mechanics
Version 8.23: August 3, 2010
Errors disappear like magic.
Carton
B.1 Introduction
In working with dimensions, a little thought and common sense go a very long way. The primary goal of
dimensional extermination is to rewrite relationships among physical quantities (equations and inequalities)
in terms of dimensionless quantities.
1
Because physical quantitiesmass, energy, length, time, and so
forthhave dimensions, theyre called dimensional quantities. For instance, the left- and right-hand
sides of the de Broglie relationship, p = /, must both have dimension of momentum, M L T
1
. On the
right-hand side, the wavelength has dimensions of length, L , and has dimensions E T = M L
2
T
1
,
energy times time. So the dimensions of the left- and right-hand sides do indeed agree.
2
Physical Quantity Dimension Physical Quantity Dimension
Length L Time T
Mass M Temperature
Current I
Table B.1. Notation for dimensions of primary physical quantities.
For additional quantities, see the the tables in B.8.
Similarly, the time-independent Schrodinger equation (TISE),

1
E
= E
E
, equates quantities with di-
mensions of energy. The 1D TISE, for example, is
_

p
2
2m
+ V (x) E
_

E
(x) = 0. (B.1a)
Written explicitly in terms of x and its second derivative, this equation reads
1
Jargon: Some writers use the term dimensional analysis for this process. Usually, though, dimensional analysis refers
exclusively to using the required dimensions of a physical quantity to deduce, up to a constant, an expression for the quantity.
Thats why I use the term dimensional extermination. Other authors call this process dedimensionalizing. (I refuse to.)
2
Notation: I indicate dimensions by putting the physical quantity under discussion in square brackets, as for example
[m
e
] = M or [] = E T = M L
2
T
1
. Except for the Greek letter (capital theta), which by convention denotes the
dimension of temperature, I use roman capital letters for dimensions (see Tbl. B.1). For contrast, I math-italic symbols to
represent physical quantities. For example, M denotes the dimension of mass, while M (or m) is the symbol for the mass. A
quantity, such as the ne-structure constant e
2
0
/c, whose dimensions are 1 ([] = 1) is a dimensionless quantity.
Appendix 27
_


2
2m
d
2
dx
2
+ V (x) E
_

E
(x) = 0. (B.1b)
Equation (B.1a) contains quantities with dimensions of momentum, mass, length, and energy. Equa-
tion (B.1b) uses the denition of the momentum operator, p i d/dx, to eliminate momentum in favor
of d
2
/ dx
2
, with dimensions L
2
, and the fundamental constant , with [] = E T . Crucially, every term
in Eqs. (B.1) must have the same dimensionsthose of energy.
Dimensional extermination is useful in solving any problem in any eld of physics. In quantum physics,
this technique is valuable for, among other things,
(1) checking derived physical expressions and equations;
3
(2) organizing the physical mechanisms involved in a process in order of decreasing importance to gain
insight into the balance of those mechanisms;
(3) identifying likely approximations, such as omitting less important physical mechanisms in a model;
(4) minimizing the number of physical quantities you need to characterize a type of physical process (re-
ducing the size of the parameter space);
(5) minimizing the work required to solve equations via mathematical analysis and/or computation;
(6) To develop solely by dimensional analysis the mathematical structure of a relationship between
physical quantities, and to identify the minimum number of physical properties on which the relationship
depends.
B.2 Transformation procedure and tools
We need education in the obvious more than
investigation in the obscure.
Oliver Wendel Holmes
B.2.1 Dimensionless quantities and scale factors
To determine the dimensionless form of an expression or equation, we write it in terms of dimensionless
quantities. Each dimensionless quantity is related to the corresponding dimensional quantity by a
multiplicative scale factor:
dimensional quantity = scale factor dimensionless quantity (B.1)
I use a tilde to signify the dimensionless counterpart of a dimensional quantity. For instance, we can relate
an energy E to a dimensionless energy

E by
4
E = s
E

E

E =
E
s
E
. (B.2)
3
A physical equation is a mathematical statement of a relationship between physical quantities that describe a system.
Examples of physical quantities include independent variables (such as x and t), functions (such as the wave function),
material properties (such as m and q), and constants (such as c and ). In general, a physical quantity has both a dimension
and a value in a particular system of units.
4
Notation: Before long youll probably be sick of writing and seeing little tildes on everything in your transformed equations.
These little guys make the already complicated notation of quantum mechanics even more clutterednot a good thing. Alter-
natively, you could use a dierent symbol for the dimensionless quantity than for the dimensional quantity. For instance, you
might use for the dimensionless time, writing t = st. Finally, you could use the same symbols but drop the tildes as soon as
you can without confusing yourself. (This is the approach Ill take.) Its not a good idea to drop the tildes at the outset, since
that leads to ridiculous looking equations like t = s
t
t.
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 28
Since E and s
E
have dimensions of energy, their ratio

E is dimensionless.
Rule: A scale factor must carry the dimension of the physical quantity to which it is related.
This rule allows a lot of exibility. There is no unique or right scale factor for a dimensional quantity. For
instance, the scale factor s
E
can be any (real) quantity with dimensions of energy. Similarly we can dene
dimensionless counterparts of the other familiar quantities:
r = s
r
r r =
r
s
r
(B.3a)
x = s
x
x x =
x
s
x
(B.3b)
p = s
p
p p =
p
s
p
(B.3c)
t = s
t

t

t =
t
s
t
(B.3d)
= s




=

s

(B.3e)
Example B.1 (The dimensions of a wave function.)
The presence of the wave function in the list Eq. (B.3) may surprise you. But a wave function must be a
dimensional quantity, because its normalization integral must be dimensionless. For a bound state of a particle
in 1D, the normalization requirement is
_

(x, t) (x, t)
. .
L
1
dx
..
L
= 1. (B.4a)
As indicated by the braces, to cancel the dimension of the dierential element of length, [ dx] = L , the dimension
of the wave function must be [(x, t)] = L
1/2
. Now, to write the integral (B.4a) in dimensionless quantities,
we rst scale the dierential element of length d x = sx dx. To ensure that the resulting integral itself is a
dimensionless number we must also scale the wave function by s
1/2
x
, as
5

( x, t) =
1

sx
( x, t) =
_

( x, t)

( x, t) d x = 1. (B.4b)
The same reasoning holds for Hamiltonian eigenfunctions in any number of dimensions. In three dimensions,
for instance, the normalization requirement is
_

E
(r)
E
(r) d
3
r = 1. (B.4c)
The volume element d
3
v = d
3
r has dimensions [ d
3
v] = L
3
, so the dimensions of the wave function in Eq. (B.4c)
must be [E(r)] = L
3/2
, and when we introduce dimensionless quantities we must scale the eigenfunction
accordingly.
B.2.2 Getting started
To develop a dimensionless equation, we rst must answer two basic questions:
(1) What physical quantity is the equation about?
(2) What independent variables appear in the equation?
The answer to the rst question is right in front of us. For instance, de Broglies relation p = / is about
linear momentum: it gives an expression for linear momentum in terms of other quantities. The time-
independent Schrodinger equation (B.1a) is about energy: it equates two quantities with the dimensions of
5
Details: The state vector |(t) is dimensionless. Thats why the wave function (x, t) = x | (t) gets its dimensions
from the normalization integral Eq. (B.4a).
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 29
energy. The answer to this rst question identies one of the scale factors well need: for the TISE, we need
an energy scale factor s
E
.
The answer to the second question is equally evident. In the TISE for a one-dimensional system, for
example, the independent variable is x, a length; for the time-dependentSchrodinger equation (TDSE) of the
same system, there are two independent variables, length x and time t. The answer to this second question
identies additional scale factors well need: for the TISE, we need a length scale factor s
x
as well as s
E
;
for the TDSE we also need a time scale factor s
t
. The answers to these two questions determine which scale
factors youll need.
6
B.2.3 Characteristic values
In choosing scale factors you have almost limitless freedom. You need only follow the rule in B.2.1: Each
scale factor must have the same dimensions as the dimensional variable youre scaling. You
can, if you like, pick your mass scale factor to be the mass of the sun, your energy scale factor to be the
rest-mass of an electron, and your length scale factor to be the length of your best friends cat. To reap the
greatest benet from dimensional extermination, however, these choices are not optimal.
At this point thought and common sense enter the picture. There is no right choice of a scale factor.
But some choices provide greater simplications and insight than other choices. In almost every situation,
either the equation or the system will guide your choice of scale factors. All you have to do is pay attention.
Often you can deduce scale factors from basic features of the system. One way to do this is to notice the
characteristics value of each relevant physical quantity.
Denition (characteristic value) The characteristic value of a physical quantity Q answers the ques-
tion: roughlythat is, to an order of magnitudehow large is Q for systems of the type under consideration.
Here are a couple of examples. For atoms you already know a characteristic length: the size of most
atoms (in low-lying stationary states) is on the order of the rst Bohr radius a
0
. So a
0
is a sensiblethough
not uniquechoice for a length scale factor s
r
= a
0
.
For a one-dimensional simple harmonic oscillator (1D SHO), the potential energy m
2
0
x
2
/2 contains two
physical quantities: the mass m and the harmonic frequency
0
. This potential suggests the characteristic
time 1/
0
, and the characteristic energy
0
. So a sensible length scale factor is s
x
=
_
/m
0
. This
factor, as it happens, is 1/, where
_
m
0
/ appears in SHO eigenfunctions (Appendix P).
B.2.4 Tools
To eliminate dimensional quantities from operators, functions, expressions, and equations, you need a few
tools. One is Tbl. B.2, the dimensions of common physical quantities.
The others tools you need are the following rules for mathematical manipulation, including the
transformation of integrals and derivatives:
(1) Rule: in any physical equation, two or more physical quantities that are equal or that
are combined via addition or subtraction must have the same dimensions. Most important,
quantities on the left- and right-hand sides of physical equations must have the same dimensions.
6
Commentary: In some situations, a third question needs attention: What other physical parameters appear in the equation?
For example, the TISE for a hydrogen atom in an external electric eld contains, in addition to the usual kinetic energy operator

2
/2m
e

2
, and Coulomb potential energy Ze
2
0
/r, a eld-dependent term e E r (Chap. 13), where e > 0 is the magnitude
of the electronic charge. To retain the exibility to apply our dimensionless equations to dierent elds, we want to introduce
a scale factor for the electric eld strength, s
E
. Another way to treat systems with several parameters we want to leave exible
is illustrated for the Morse potential (see Example B.5).
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 30
Table B.2. Dimensions of frequently oc-
curring physical quantities in quantum
mechanics. The notation [Q] indicates the
dimension of the physical quantity Q. Note
that in the dimensions of wave function, d is
the product of the number of particles times
the dimensionality of the system. Thus for a
single particle in one dimension, d = 1; for a
single particle in three dimension, d = 3, for
two particles in three dimension d = 6, and so
forth. Note that L = [L[.
energy [E] = M L
2
T
2
Plancks constant [] = M L
2
T
1
= [E]T
linear momentum [p] = M L T
1
= []L
1
angular momentum [L] = M L
2
T
1
= []
wave function [] = L
d/2
frequency [] = T
1
(2) Rule: Except for simple power-law monomial functions, such as f(x) = x
2
, the arguments
of all mathematical functions must be dimensionless. For example, the arguments of sin, exp,
and special functions such as Bessel functions and Hermite polynomials must be dimensionless.
7
(3) Rule for transforming the n
th
derivative of an expression with respect to a variable: The
dimension of the n
th
derivative must equal the dimension of the original quantity times the dimension of
the variable to the power n. For example, taking the rst time derivative of position (dimensions L )
yields a speed v (dimensions L T
1
). Let v denote the dimensionless speed, and s
v
the speed scale
factor. To eliminate dimension from the n
th
derivative with respect to v, we must divide by s
n
v
:

n
v
n
=
1
s
n
v

n
v
n
=

n
v
n
= s
n
v

n
v
n
(B.5)
(4) Rule for transforming the integral of an expression with respect to a variable: the dimension
of the resulting expression must equal that of the original quantity times the dimension of the variable
of integration. To transform the dierential d
q
v in an integral over speed v to the dierential d
q
v,
where v = s
v
v, multiply the integrand by s
q
v
:
d
q
v = s
q
v
d
q
v (B.6)
Try This! 2.1. Dimensional transformation from the denition of a derivative.
By denition, the derivative of f(x) with respect to x is
df
dx
lim
x0
f(x + x) = f(x)
x
.
Explain why Eq. (B.5) is consistent with this denition.
Example B.2 (Transforming the momentum operator.)
The rules above apply immediately to the momentum operator. Each Cartesian component of the (congu-
ration space) momentum operator transforms as
px = i

x
= i
1
sx

x
(B.7)
Aside. The role of the Jacobian in transforming derivatives
To transform a derivative with respect to a variable x to a derivative with respect to a new variable x,
you divide by the Jacobian raised to the order of the derivative. By denition, the ij
th
element of the
Jacobian is
7
Example: Each Hamiltonian eigenfunction of a 1D SHO is proportional to a Hermite polynomial H
n
(x) whose argument
is x, where
_
m
0
/ has dimensions of length.
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 31
J
i,j

x
i
x
j
, Jacobian of the transformation | x
i
| x
i
. (B.8a)
To transform the rst derivative, one uses

x
i
=

x
j
_
J
i,j
_
1
. Jacobian of the transformation | x
i
| x
i
. (B.8b)
For a transformation from x to x, where
x = sx x x =
1
sx
x, (B.9a)
the corresponding rst partial derivative transforms according to

x
=

x
x
x
=
1
sx

x
. (B.9b)
Similarly, the q
th
partial derivative transforms according to Eq. (B.5). A similar argument shows how to
transform the rst time derivative in the TDSE:
t = st

t

t =
1
st
t, =

t
=
1
st

t
(B.10)
B.3 Transforming the Hamiltonian
I dont like nonrepeating decimals.
Pi makes me furious.
Ratners Star by Don DeLillo
Most equations in quantum mechanics involve operators. So you must know how to transform operators as
well as the functions they act upon. You need to be especially aware of whatever ancillary conditions
boundary conditions, orthogonality and normalization conditionsequations these functions must satisfy in
addition to the equation itself. Ill illustrate these points using the single-particle TISE, with an occasional
nod to the TDSE, to show you how to cope with more than one independent variable. You can apply these
ideas and tactics to any expression or equation in any eld of physics.
B.3.1 Transforming the kinetic energy operator
The fundamental operator in the TISE is the Hamiltonian,

1 =

T +

V. The kinetic energy operator,

T = p p /2m, depends only on the mass of the particle only. The potential energy, by contrast, depends on
the particular forces that act on the particle; these dier from system to system. It makes sense, therefore, to
begin with the kinetic-energy operator, which we can transform once and for all, and then turn to potential
energies, which we must treat on a case-by-case basis.
The kinetic energy operator for a single particle in one-dimension is
8

T(x) =

2
2m
d
2
dx
2
. (B.1)
8
Details: This operator, like others in this section, is written in position space. One can, of course, formulate quantum
mechanics in other spaces, such as momentum space. In that case, one can translate the procedures and equations in this
section into the appropriate space. The presence of x as an argument of

T is a reminder that this operator contains operators
with respect to x. Pay attention to whether or not a quantity carries a tildethose that dont are dimensional; those that do
are dimensionless.
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 32
The dimensions of

T are energy. Letting s
E
denote the energy scale factor and using a tilde to signify the
corresponding dimensionless operator, we can write
9

T(x) = s
E

T( x). (B.2a)
The dimensionless operator

T( x) involves operators with respect to the dimensionless length variable x,
which is dened via a length scale factor: x = s
x
x. Applying Eq. (B.9b), p. 31 twice, we nd that in terms
of x, Eq. (B.1) becomes

T( x) =

2
2m
1
s
2
x
d
2
d x
2
. (B.2b)
Now, we seek a relationship between the length and energy scale factors. To nd one, we equate
the right-hand side of Eq. (B.2b) to s
E

T( x) and get

2
ms
2
x
= s
E
=

T( x) =
1
2
d
2
d x
2
(B.3a)
The relationship is a bit arbitrary. Alternatively, you can include exclude the factor of 1/2 and instead use
or

2
2ms
2
x
= s
E
=

T( x) =
d
2
d x
2
. (B.3b)
It makes no dierence which relationship you choose. But having chosen a relationship, you must stick with
it. In this book, Ill use Eq. (B.3a).
10
B.3.2 Transforming the potential energy
Our work with the kinetic energy operator has established a relationship between the length and scale factors.
For either scale factor, we could choose any quantity with the appropriate dimensions. The idea, though, is
to choose a quantity that is appropriate to the system were studying. For guidance, we look to the potential
energy, which incarnates the forces that act on the particle.
Rule: Any potential energy must have dimensions of energy.
Except for the constant potential energy of an innite square well, all potential energies depend on position.
We can always write a potential energy in terms of dimensionless length variables: in 1D, for example, we
just replace every occurrence of x by s
x
x by writing V ( x) = V (x = s
x
x).
The strategy is to use the potential energy to determine either the length or energy scale factors,
whichever is easiest to get algebraically. We then use relationship (B.3) between these scale factors to
get the other factor. To see how this works, lets write

1(x) in terms of x:

1( x) =

2
2m
1
s
2
x
d
2
d x
2
+ V (s
x
x) (B.4a)
=

2
ms
2
x
_

1
2
d
2
d x
2
+
ms
2
x

2
V (s
x
x)
_
. .
dimensionless Hamiltonian

H( x)
. (B.4b)
The thing in square brackets is a dimensionless Hamiltonian. In general a dimensionless Hamiltonian
contains operators and function of the dimensionless length variables. Denoting this operator by

1( x), we
9
Commentary: The only reason for the minus sign is so the scale factor s
E
will be positive. Scale factors can be negative;
but by always choosing them to be positive, you free yourself from having to remember their sign.
10
Commentary: My research is in atomic and molecular physics. This choice, when applied to expressions for atoms and
molecules, leads automatically to atomic units (Appendix F), which is why I use it.
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 33
have

1(x) =

2
ms
2
x

1( x) =

1( x)
1
2
d
2
d x
2
+
ms
2
x

2
V (s
x
x) =
1
2
d
2
d x
2
+ s
E

V ( x) (B.4c)
We can now write the TISE (or the TDSE) in terms of the dimensionless Hamiltonian. The TISE is

1
E
E
E
= 0. Because the right-hand side of this equation is zero, we can divide out the prefactor

2
/2ms
2
x
in the Hamiltonian (B.4c) to get
_

1( x)
_
ms
2
x

2
s
E
_

E
_

E
( x) = 0. (B.5a)
At this point we invoke relationship (B.3) between s
E
and s
x
. Choosing Eq. (B.3a) causes the expression
in parentheses in Eq. (B.5a) to equal 1 and, unsurprisingly, simplies the dimensionless TISE to
_

1( x)

E
_

E
( x) = 0. dimensionless TISE for one particle in 1D. (B.5b)
Now is a good time to drop all the tildes, thereby reducing clutter at the (modest) cost of having to remember
that youre using dimensionless quantities.
Example B.3 (The dimensionless 1D SHO.)
In the TISE of a 1D SHO with harmonic frequency
0
,
_


2
2m
d
2
dx
2
+
1
2
m
2

2
0
x
2
E
_
E(x) = 0, (B.6a)
the potential energy function is
V (x) =
1
2
m
2
0
x
2
. (B.6b)
To determine s
x
and s
E
for this system, Ill rst write V (x) as a function of x:
V ( x) = V (x = s
x
x) =
1
2
_
m
2
0
s
2
x
_
x
2
. (B.7)
Any potential energy must have dimensions of energy; x in Eq. (B.7) is dimensionless, so the stu that
multiplies it must carry these dimensions:
_
m
2
0
s
2
x
_
= E . Choosing this as my energy scale factor gives me
one relationship between sx and sE :
s
E
= m
2
0
s
2
x
. (B.8)
But I already know a second relationship between these scaling factors: its s
2
x
=
2
/ms
E
, which I got by
exterminating dimensions in the kinetic energy operator [Eq. (B.3a), p. 32]. Combining this with relationship
with Eq. (B.8), I get the energy scale factor for the 1D SHO:
m
2
0


2
ms
E
= s
E
= s
E
=
0
(B.9a)
I can now plug sE into my relationship for s
2
x
in terms of sE to get the length scale factor for the 1D SHO:
s
2
x
=

m0
= sx =
_

m0
(B.9b)
Since the expressions for the eigenfunctions and eigenvalues of this system are extremely well known, I can
examine them to see if my scale factors make sense and are going to be helpful. In Appendix P, I nd that my
choice of s
x
is the inverse of the factor that multiples the (dimensional) length variable x in the eigenfunctions,
while my choice of s
E
shows up the the eigenvalues:
n(x) =
1

2
n
n!
_
m

_
1/4
Hn(x) e

2
x
2
/2
En =
_
n +
1
2
_
,
_

_
n = 0, 1, . . . , (B.10)
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 34
where by denition =
_
m
0
/.
Noteworthy features of Example B.3: The discovery that the logic of dimensional extermination led
to scale factors that appear in the exact solutions of the TISE has profound implications. Equations (B.10)
tell me that the quantities 1/ and
0
are the characteristic length and characteristic energy of this
system. But in Example B.3 I found those characteristic quantities without solving the Schrodinger equation
; I just used dimensional extermination and a little algebra.
Knowing characteristic values of relevant physical quantities is useful for any system. The SHO is one of
a (tiny) handful of quantum systems whose TISE can be solved exactly. But you can apply dimensional ex-
termination to any system. And, as you might expect, doing so will usually lead to appropriate characteristic
values for the physical quantities in the problem.
Example B.4 (The dimensionless TISE for the 1D SHO)
With the choices of scale factors from Example B.3, the TISE for the 1D SHO becomes simply
_

1
2
d
2
dx
2
+
1
2
x
2
E
_

E
(x) = 0, dimensionless TISE for a 1D SHO, (B.11a)
where for readability Ive dropped the profusion of tildes. I must solve this equation subject to the usual
normalization and boundary conditions for 1D bound states:
_

E(x)

E(x) dx = 1 and lim


x
E(x) = 0. (B.11b)
One can solve Eqs. (B.11) on a computer or by standard algebraic methods (either power-series expansion or
using raising and lowering operators) or, in our case, just look up the solution in Appendix P:

n
(x) =
1

1/4
1

2
n
n!
H
n
(x) e
x
2
/2
E
n
=
_
n +
1
2
_
_

_
n = 0, 1, . . . , (B.12a)
where H
n
(x) is the Hermite polynomial of order n. All quantities in Eqs. (B.12) are dimensionless. To generate
dimensional solutions, I use the scale factors for x, E, and E(x); doing so returns me to Eqs. (B.10).
Noteworthy features of Example B.4:
Look closely at the dimensionless TISE (B.11a). This equation contains no system-dependent quantities.
On the other hand, the dimensional TISE, Eq. (B.6a), contains the systems mass m and its harmonic
frequency
0
. When we eliminated dimensions from the TISE we also buried all quantities that depend
on a particular 1D SHO into the length scale factor, s
x
=
_
/m
0
. Those quantities reappear only when
we transform the results of calculations to dimensional quantities (B.4).
The signicance of the disappearance of system-dependent quantities from equations we have to solve
and expressions we have to evaluate is that we therefore must perform those chores only once. Were we to
work with dimensional equations and expressions, wed have to (in principle) do these chores anew for each
new system.
Furthermore, our ability to eliminate system-dependent quantities by writing (without approximation)
the TISE for the 1D SHO in terms of s
E
=
0
and s
x
=
_
/m
0
reveals that the wave functions of
this system depend on m and
0
only in the combination m
0
. Similar ndings emerge from standard
dimensional analysis and always bear fruitful insights the physics of a system.
Finally, Examples B.3 and B.4 illustrate ways to relate scale factors to one another. Using such rela-
tionships eliminates the need to determine independently one (or more) scale factors. Here is a collection of
useful relationships among scale factors:
11
s
E
=

2
ms
2
x
(energy & position) (B.13a)
11
Read on: For an additional, more elaborate example, see dimensional extermination in the pure-Coulomb Hamiltonian of
a hydrogenic atom, Example ?? in Complement ??.
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 35
s
p
=

s
x
(momentum & position) (B.13b)
s
t
=

s
E
(time & energy) (B.13c)
s
t
=
m

s
2
x
(time & position) (B.13d)
Aside. Dimensional deductions.
For the 1D SHO we can deduce scale factors and thereby develop, from dimensional analysis, an approx-
imate expression for the energies. To generate a scale factor s
x
with dimensions of length we must combine
physical quantities in such a way that [sx] = L . Since [] = E T = M L
2
T
1
, we can write the length
dimension as L =
_
(E T ) M
1
T . We now replace the dimensions in this equation with the symbols for
the SHO; doing so leads back to Eqs. (B.9).
Notice the procedure: First, combine the dimensions of the physical properties and constants in the
problem (here , m, and ) to generate a quantity with the desired dimensions (here, L ). Then replace the
dimensions in the resulting expression with the symbols that represent the physical properties and constants.
In this analysis I kept E T together as a unit rather than replacing the secondary dimension E with M L
2
T
2
,
because E T are the dimensions of , which we expect to nd in the scale factor sx.
12
B.4 Conversion back to dimensional quantities
Dont be so quick to answer correctly.
Tragic mistakes can result.
Ratners Star by Don DeLillo
Once youve chosen scale factors for length and energy, you solve the dimensionless TISE. Once youve
completed that chore, you must remember to transform the dimensionless eigenfunctions back to dimensional
eigenfunctions. In 1D, for example, we do so via

n
(x) =
1

s
x

n
_
1
s
x
x
_
(B.1)
Only after this dimensionalization step will our solutions obey the normalization condition
_

n
(x)
n
(x) dx = 1. (B.2)
Eigenfunctions (and wave functions) must obey this condition; otherwise they cant be interpreted as position
probability amplitudes and used in calculation of physical quantities.
B.5 Solving the dimensionless TISE for the Morse potential
The 1D SHO potential energy V (x) = m
2
0
x
2
/2 contains, in addition to the particles mass, a single system-
dependent parameter: the harmonic frequency
0
. We can choose the value of
0
to suit the system
were trying to model.
12
Commentary: Provocatively, Menzel (1961) notes that, for reasons no one understands, dimensionless factors that dimen-
sional analysis misses are almost always on the order of 1: The numerical multiplying factor[s], indeterminate by dimensional
methods, [are] surprisingly close to unity, and usually quite simple in form. 2 occurs very frequently as a factor. . . . Occasionally
the result is exact. . . . This fortunate circumstance means that the dimensionally derived equation, if unique, is very nearly ex-
act.
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 36
But many model potentials have more than one such parameter. To illustrate, this section applies
dimensional extermination to a widely used model of molecular vibration, an essential aspect of molecular
spectroscopy. If the amplitude of molecular vibrations is small, then we may be able to use the SHO
model, letting the variable x be x R R
eq
, the deviation of the internuclear separation R from its
equilibrium value R
eq
, a quantity that can be determined from experimental data.
13
But if the amplitude
of the vibrations is not small, then we must modify the SHO model to allow for anharmonicity.
One very useful model of an anharmonic molecular potential energy is the Morse potential
V (x) D
e
_
1 e
ax
_
2
. .
dimensionless
, dimensional Morse potential. (B.1)
This function contains two system-dependent parameters: the range parameter a > 0 and the well
depth D
e
; both are illustrated in Fig. B.1. The parameter D
e
is the depth of the well dened by the
Morse potential, since the value of the function (B.1) at equilibrium (x = 0) is V (0) = D
e
. Similarly, the
parameter 1/a characterizes the range of the potential. As indicated by the braces in Eq. (B.1), D
e
must
carry the dimensions of V (x), so [D
e
] = E . Since the argument of the exponential in this function must be
dimensionless, [a] = L
1
.
Figure B.1. The Morse potential of Eq. (B.1)
for two choices of the range parameter a. The
value 1/a is a rough estimate of the range of
this potentialthat is, of the size of the poten-
tial well: as 1/a increases, the Morse potential
broadens. The second Morse parameter is the
equilibrium dissociation energy De.
1 0 1 2 3 4
D
e
xRR
eq
M
o
r
s
e
p
o
t
e
n
t
i
a
l
e
n
e
r
g
y
1a 1
1a 12
The dimensional Hamiltonian that describes vibrations of a diatomic molecule via this model therefore
contains, in addition to these two parameters, the (system-dependent) reduced mass of the nuclei:

1
M
(x) =

2
2
d
2
dx
2
+ D
e
_
1 e
ax
_
2
, with
m
A
m
B
m
A
+ m
B
, (B.2)
where m
A
and m
B
are the masses of the two nuclei. The Hamiltonian Eq. (B.2) represents the relative
motion of the two nuclei. Its eigenvalues represent, in the Morse model, the bound-state vibrational energies
of the molecule, and the eigenfunctions are probability amplitudes for nding the nuclei separated by a
distance R = x + R
eq
. While the reduced mass is xed by the atoms that make up the molecule, were free
to adjust a and D
e
to any values we like. In practice, we choose values that reproduce, as accurately as
possible, measured properties of the molecule.
Example B.5 (The dimensionless TISE for the Morse potential.)
The mathematical structure of the Morse potential (B.1) strongly suggests choices for the length and energy
scale factors. Since the exponential factor implies [a] = L
1
, we could to choose s
x
= 1/a. Similarly, since
[De] = E , we could choose sE = De. With these choices, the dimensionless the Morse function is
sx =
1
a
and sE = De =

V ( x) =
_
1 e
x
_
2
dimensionless Morse potential. (B.3)
13
For example, the nuclei of the nitrogen molecule N
2
vibrate about an equilibrium internuclear separation Req = 1.097 68

A
with a harmonic frequency
0
= 2358.57 cm
1
, while the nuclei in H
2
vibrate about a much smaller equilibrium separation,
R
eq
= 0.7414

A with
0
= 4401.21 cm
1
(see Appendix J regarding the units of
0
). You can nd values for these and other
constants of molecules in ?.
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 37
Both choices are sensible and physically relevant: 1/a and D
e
dene characteristic length and energy scales
for this system. But these choices have one disadvantage: they leave system-dependent quantities in the
dimensionless kinetic-energy operator, which becomes [see Eq. (B.2b), p. 32]

T( x) =
1
2
_

2
a
2
D
e
_
d
2
d x
2
. (B.4)
The presence of these factors in

T( x) doesnt invalidate the choices in Eq. (B.3), but it doesnt lead to as clean
a Schr odinger equation as we might like. So lets consider an alternative approach.
If we want to eliminate a from the exponential in the Morse function, we must choose s
x
= 1/a. But
suppose we now apply our old standby, the relationship s
2
x
=
2
/msE [Eq. (B.13a), p. 34]. I get
sx =
1
a
and sE =

2
a
2

. (B.5a)
Choosing this scale factor introduces a dimensionless well depth

D
e

1
s
E
D
e
=
D
e

2
a
2
=

V ( x) =

D
e
(1 e
x
)
2
(B.5b)
is our dimensionless Morse function. Although

De is a perfectly reasonably symbol for this quantity, its
conventional (and convenient) to use
2
:

D
e

2
a
2
=

V ( x) =
2
(1 e
x
)
2
. (B.5c)
With the choices Eqs. (B.5), the TISE for the dimensionless Morse potential energy is therefore
_

1
2
d
2
dx
2
+
2
_
_
1 e
x
_
2
E
_
_
E(x) = 0 dimensionless TISE for a Morse potential, (B.6)
where I have dropped the tildes on all dimensionless quantities. The number will appear in the Morse
eigenvalues and eigenfunctions. For any molecule, we just choose a and D
e
, calculate , and o we go.
14

Noteworthy features of Example B.5: When we eliminated dimensions from the 1D SHO potential
energy (Example B.3), the system-dependent quantities m and
0
wound up in the length scale factor
s
x
=
_
/m
0
, and no system-dependent quantities remained in the dimensionless TISE [Eq. (B.11a), p. 34].
The Morse case is a bit dierent. In Example B.5 we explored two choices of s
x
and s
E
both of which left
system-dependent quantities in the dimensionless TISE. The second choice, in Eqs. (B.5), collected all such
quantities (, a, and D
e
) in a single dimensionless number , which appears in the dimensionless Hamiltonian
and so will appear in its eigenfunctions and eigenvalues.
Aside. Bound-state vibrational energies of the Morse potential.
One can (with a lot of work) derive exact algebraic expressions for the bound-state eigenvalues and eigen-
functions of the Morse Hamiltonian (see Morrison and Sun (1995) and references therein). The results
for the bound-state energies are a little complicated, but I want to show them to you so youll see how
the system-dependent parameters appear in physically signicant quantities that can be determined from
spectroscopic data. The essential result, in a form immediately comparable to spectroscopic constants, is
15
E
v
=
e
_
v +
1
2
_

e
x
e
_
v +
1
2
_
, energies of the Morse potential, (B.7a)
14
Commentary: The choice of a for the length scale factor is sensible but by no means unique or even necessarily optimal.
Another equally sensible choice would be s
x
= 1/ =
_
/
e
, our choice for the 1D simple harmonic oscillator of mass . We
could then relate the angular frequency e to the Morse parameters by e = a
_
2De/. With this length scale factor, we get
the energy scale factor s
E
=
2
/s
2
x
= e, which equals the spacing between adjacent energy levels of the corresponding SHO.
These alternative choices would be useful if the anharmonicity is weakin which case we might want to use time-independent
perturbation theory (Chap. 14) to approximate the bound-state energies.
15
Details: You may be wondering how to determine a and D
e
for an actual molecule. For most molecules, experimentalists (or
theoreticians) have determined accurate values for De, which is called the equilibrium dissociation energy of the molecule.
For example, the measured dissociation energy of the ground state of H
2
is De = 4.478 eV. For the ground state of N
2
,
its D
e
= 9.759 eV. Knowing D
e
, we can evaluate a by requiring the rst term in a Taylor-series expansion of the Morse
potential about x = 0 to equal the rst term in the empirical equation for the vibrational energies E
v
=
e
(v + 1/2), where
e
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 38
where the range of the vibrational quantum number is v = 0, 1, . . . , v
max
. (Unlike the SHO, the Morse
potential supports only a nite number of bound states.) Ideally, one would calculate the Morse parameters
from experimentally determined properties of the molecule, the equilibrium dissocation energy D
e
and
the equilibrium harmonic frequency e. From these numbers, the Morse range paramter is
16
a =
e
_

2De
, range parameter for a Morse parameters. (B.7b)
With e in hand, we can calculate the anharmonicity constant in Eq. (B.7a):
xe =
e
4De
, anharmonicity constant from Morse parameters (B.7c)
The expressions for the Morse eigenfunctions, although complicated, involve familiar functions: exponentials
and Laguerre polynomials.
17
B.6 Scale transformations
In some problems a type of change-of-variables dierent from dimensional extermination proves useful. Called
a scale transformation, this change of variables does not alter the dimensions of physical quantities.
Rather, it incorporates one or more system constants into a variable. Its best explained by example.
Example B.6 (A scale transformation for a screened Coulomb potential.)
Many problems in atomic, molecular, and even plasma physics can be modeled by a screened Coulomb
potential (see examples in Part III). The most general form of such a potential energy is
V (r) =
A
r
f(r), screened Coulomb potential energy. (B.1)
The function f(r) carries the dependence of this potential energy on the single system-dependent parame-
ter , which multiplies the radial variable r. The parameter A measures the overall strength of the potential
energy. The other parameter, the screening constant , measures the deviation of the potential energy from
the pure-Coulomb potential energy of an electron in the eld of a nucleus of atomic number Z, which in atomic
units is Z/r. Here are dimensionless forms of a couple of widely used screened Coulomb potentials, the
Yukawa potential and the Hulthen potential :
V
Y
(r) =
A
r
e
r
, Yukawa potential, (B.2a)
V
H
(r) =
A
r
_
r e
r
1 e
r
_
, Hulthen potential. (B.2b)
A scale transformation of a screened Coulomb potential incorporates the constant A into the radial vari-
able. (For an atom, Z is usually taken as the atomic number or, for a many-electron atom in the independent-
particle central-eld model of Chap. 11, as the eective atomic number Z
e
n,
.) We can implement this
transformation via the replacement r r/A in the radial Schr odinger equation. Modest algebra shows that
the energy E and radial wave function R(r) then depended on A and in an elegant, simple way:
18
is determined from measured vibrational spectra. Doing so yields a = e
_
/2De. Plugging this expression into our denition
of , Eq. (B.5c), gives =

2 D
e
/
e
.
16
Units: This equation assumes that De has dimensions of energy. Since the dimensions of a and are length and mass, this
equations gives the angular frequency in dimensions of T
1
. When multiplied by , this e has the correct dimensions, E . Note,
however, that tables of molecular constants usually give
e
in units of cm
1
, so you may need the conversions of Appendix J.
17
Details: One cant directly measure the equilibrium dissociation energy, because the lowest available vibrational energy
is the zero-point energy
e
/2, which corresponds to v = 0. One can, however, measure the amount of energy required to
dissociate the molecule from the v = 0 statea quantity denoted D
0
. From this number and
e
, the equilibrium dissociation
energy follows as De = D
0
+e/2.
18
Commentary: Its easy to combine a scale transformation with dimensionless extermination. For instance, if were applying
a screened-Coulomb potential to an atom and choose as our length scale the rst Bohr radius a
0
, then we can use a length scale
factor s
r
= a
0
/A. For the corresponding energy scale factor we can use s
E
=
2
/ms
2
r
= A
2
E
h
.
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 39
E(A, ) = A
2
E(1, /A)
R(r; A, ) = A
3/2
R(Ar; 1, /A)
(B.3)
where the arguments of the radial function after the semicolon are the parameters in the model potential
energy Eq. (B.2). Two consequences follow from these results. The rst is formal:
Rule: In general, the eigenvalues of a Hamiltonian whose potential energy depends on several parameters can be
treated mathematically as a function of those parameters.
For a screened-Coulomb potential, the energy depends not on A and separately but rather on the single
variable /A. The second consequence is practical:
Rule: We need not solve the Schrodinger equation or do any analysis with expressions that contain A. In our
analysis we set A = 1. When were done, we use a scale transformation to generate the nal Adependent quantities.
B.7 Final exhortations
Physics depends on a universe innitely centered on an equal sign.
House of Leaves,
by Mark Z. Danielewski
Dimensional extermination oers great benets with comparably little work. There is no cookbook for
dimensional analysis: you must consider each case on its own. The examples above and suggestions below
will serve you well in your own analysis.
(1) Think about the system. Try to identify characteristic values of physical quantities. Such values
are often ideal scale factors.
(2) Look at the equation. Identify each physical quantity (mass, length, natural frequency, and so forth).
Consider using one or more of these quantities, perhaps in combination, as a scale factor. If, as in
the Morse potential of Example B.5, the equation strongly suggests particular scale factors, consider
following its guidance. (But dont be a slave to the equationyou may prefer choices other than the
ones the potential energy recommends.)
(3) Seek mathematical functions in the equation whose arguments must be dimensionless:
exponentials, logarithms, special functions such as Hermite polynomials, Laguerre polynomials, and so
forth. The arguments of these functions often suggest scale factors.
(4) Use relationships between scale factors to derive other scale factors [Eqs. (B.13), p. 34].
(5) Dont forget that every term in an equation must have the same dimensions.
(6) Dont try to be precise. All you need is a scale factor that is of a sensible order of magnitude; its
actual value matters little. (Thats why characteristic values are so useful: they characterize the
value of the quantityroughly, to an order of magnitude.
(7) Explore ways to combine the physical quantities of the system and relevant fundamental
constants to produce whatever dimensionless quantities you need.
(8) Be exible. If, as you work with your dimensionless equations, you decide that a dierent choice of
scale factor is preferable to the one you made, change it.
Choosing appropriate, advantageous scale factors is not an art; its a skill. You dont have to be Einstein
or Feynman to master it. But you do have to think a bit about the system and the relevant quantities and
equations youre working with. And you do have to use common sense. In other words: keep your brain
engaged, not just the optic nerve!
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 40
B.8 Summary
Equation (B.1) shows how dimensional extermination works:
dimensional quantity = scale factor dimensionless quantity (B.1)
Summary of the procedure for eliminating dimensional quantities:
(1) Use the kinetic energy operator to develop a relationship between the energy and length scale factors.
(2) Use the potential energy to determine one of the factors, s
E
or s
r
, whichever is more convenient.
(3) Use the relationship from step (1) to determine the scale factor you didnt determine in step (2).
You can now write down the dimensionless Hamiltonian and TISE, which you can proceed to solve.
(4) If youre interested in the time-dependent Schrodinger equation, you need a second fundamental
relationship, one between the time and energy scale factors:
s
t
s
E
= =
_

T +

V ( r)
_
(r, t) = i

t
(r, t) (B.2)
Rule: When you transform a dierential equation to dimensionless quantities, dont forget to transform all initial
or boundary conditions that pertain to the equation. Also dont forget to transform any auxiliary equations such
as the normalization condition.
Table B.2 contains dimensions of the most-often-encountered quantities in quantum mechanics and of
additional widely used primary and some secondary physical quantities. Table B.4 focuses on energy-related
quantities and Tbl. B.5 on quantities related to electricity and magnetism; the latter are used in the study of
the interaction of radiation with matter (see Chap. 18). Finally, Tbl. B.6 lists the dimensions of important
physical constants.
Physical Quantity Dimension Physical Quantity Dimension
Length L Time T
Mass M Temperature
Current I
Area L
2
Specic heat capacity L
2
T
2

Velocity L T
1
Volume L
3
Charge Q = I T Density M L
3
Acceleration L T
2
Force M L T
2
Linear momentum M L T
1
Moment of inertia M L
2
Angular momentum M L
2
T
1
Torque L
2
M T
2
Energy M L
2
T
2
Power M L
2
T
3
Angle Frequency T
1
Pressure M L
1
T
2
Entropy M L
2
T
2

1
Table B.3. Dimensions of primary and some secondary physical quantities. Although I
use Q for the dimension of charge, charge is not a primary dimension. The Coulomb equals one
ampere-second: that is, Q = I T .
Dimensions for length, time, mass, current, and temperature are primary dimensions. (Additional
primary units are luminous intensity, which is not used in this book, and unit of substance, which in SI units
is the mole.) All other dimensions are secondary dimensions. The dimensions of any physical quantity
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix 41
Physical quantity with energy with linear momentum fundamental units
Linear momentum E
1/2
M
1/2
M L T
1
Energy P
2
M
1
M L
2
T
2
h, E T P L M L
2
T
1

2
/m E L
2
P
2
L
2
M
1
M L
4
T
2
e
2
0
e
2
/4
0
E L M L
3
T
2

2
/me
2
0
L
Table B.4. Dimensions of energy-related quantities.
Physical Quantity Dimension in terms of charge Q
Charge L
1/2
M
1/2
Q
Current L
1/2
M
1/2
T
1
T
1
Q
Potential Dierence M
1/2
L
3/2
T
2
M L
2
T
2
Q
1
Permeability () 0 M L Q
2
Permittivity () L
2
T
2
M
1
L
3
T
2
Q
2
Magnetic dipole moment L
5/2
L
1/2
T
1
L
2
T
1
Q
Magnetic eld M
1/2
L
1/2
T
1
M T
1
Q
1
Electric eld M
1/2
L
1/2
T
2
M L T
2
Q
1
Current density M
1/2
L
3/2
T
1
L
2
T
1
Q
Conductivity L
2
T M
1
L
3
T Q
2
Table B.5. Dimensions of physical quantities related to electricity and mag-
netism.
Table B.6. Dimen-
sions of physical
constants and useful
combinations.
Fundamental constant Symbol Dimension
Rationalized Plancks constant M L
2
T
1

2
/m M L
4
T
2
Fine-structure constant 1
c M L
3
T
2
Boltzmann constant k
B
M L
2
T
2

1
Stefan-Boltzmann constant M T
3

4
Bohr magneton
B
M
3/2
L
3/2
T
3
g factor g 1
gyromagnetic ratio M
1/2
L
1/2
permittivity of free space
0
M
1
L
3
T
4
I
2
can be expressed as a simple power-law monomial function of primary dimension such as L , M , and T .
(See also Appendix E.)
Aside. Would you like to know more?
Various uses of dimensional analysis are illustrated in Chaps. 2 and 3 of Howison (2005). The rst chapter
of Sneider (2004) also contains several useful examples. Part I of Menzel (1961) addresses units for elec-
tromagnetic quantities, lists denitions and dimensions of quantities that are important to elds other than
quantum physics, and itemizes a procedure for generating systems of simultaneous algebraic equations the
solutions of which yield dimensional relationships (see 14 and 15). For more detailed treatments, check
out the short books by Barenblatt published in 1987 and 2003.
PrintAppendices Version: 8.23 Printed: August 3, 2010
Appendix C
Fundamental constants and order of magnitude estimates
Version 8.13: August 3, 2010
Constants. This book uses two systems of units: primarily, the SI (Syst`eme International) system and
atomic units (see Appendix F). Although I dont use CGS units, much data in atomic physics and quantum
mechanics is reported in these units, so for some constants I give the appropriate exponent. Most fundamental
constants used in quantum mechanics are in Tbls. C.2 and C.3. Those associated with magnetic moments
are in Tbl. C.4, while those associated with Plancks constants are in Tbl. C.1. Note especially Tbl. C.5,
which contains approximate values of key fundamental constants; these are more easily remembered and
used that the full values given elsewhere. Except for Tbl. C.5, all values tables come from the NIST web site
http://physics.nist.gov/ and are reported to all accurate decimal digits as of January 2009.
Unit h
eVs 4.135 667 33 10
15
6.582 118 99 10
16
MeVs 4.135 667 33 10
21
6.582 118 99 10
22
J s 6.626 068 96 10
34
1.054 571 628 10
34
erg s 6.626 075 5 10
27
1.054 571 628 10
27
1
To convert to CGS units, multiply the value in J s
1
by 10
7
.
2
In atomic units, 1, so h 2 (Appendix F).
3
The following conversion factor is useful for working with
data from spectroscopic measurements (see Appendix J):
hc = 1.986 445 5 J m = 1.239 841 22 eV m.
Table C.1. Plancks constant in various units.
Quantity Symbol Mantissa SI exponent CGS exponent
speed of light in vacuum c 2.997 924 58 10
8
ms
1
10
10
cms
1
elementary charge
1
e 1.602 177 653 10
19
C 10
20
e.m.u.
e/h 2.417 989 40 10
14
10
14
A J
1
electron charge to mass ratio e/m 1.758 820 12 10
11
C kg
1
Rydberg constant
2
R

1.097 373 156 852 5 10


7
m
1
10
5
cm
1
ne-structure constant 7.297 353 08 10
3
10
3
1/ 137.035 999 11
Bohr radius
3
a
0
0.529 177 2108 10
10
m 10
8
cm
Compton wavelength (electron)
4

C
2.426 310 238 10
12
m 10
10
cm
classical electron radius
5
r
e
2.817 940 325 10
15
m 10
13
cm
Avogadros constant N
A
6.022 1415 10
23
mol
1
10
23
mol
1
unied atomic mass unit
6
1u 1.660 538 86 10
27
kg 10
24
g
energy equivalent of 1 m
u
E(u) 1.492 417 90 10
10
J 10
3
erg
931.494 043 MeV
1
In order to reduce clutter I use throughout this book the auxiliary quantity
e
2
0
e
2
/4
0
= 2.30708 10
28
kg m
3
s
2
. In Gaussian units,
0
= 1/4 so e
2
0
= e.
2
The subscript on the Rydberg constant refers to the approximation of innite nu-
clear mass. Note R

= m
e
c
2
/hc, where is the ne-structure constant e
2
0
/c.
Note also R

c = 3.289 841 960 360 10


15
Hz. The Rydberg unit of energy is
1 Ry = hcR

= 13.605 6923 eV.


3
For useful relationships involving the rst Bohr radius a
0
see Appendix F.
4
The value of the reduced Component wavelength,
C

C
/2 = a
0
, in SI units is

C
= 3.861 592 678 10
13
m.
5
The classical electron radius is related to the reduced Compton wavelength by r
e
=
2
a
0
=
C
.
6
This quantity is also called the atomic mass constant. Its equal to 1/12 the mass of a
12
C atom
and is related to the mass of one mole of a substance by 1u = 10
3
kg mol
1
/N
A
.
Table C.2. Fundamental constants. See also Tbl. C.3. For Plancks constant, see Tbl. C.1.
Quantity Symbol Mantissa SI exponent CSG exponent
electron rest mass
1
m
e
9.109 3826 10
31
kg 10
28
g
5.485 799 0945 10
4
u 10
4
u
proton rest mass
2
m
p
1.672 623 1 10
27
kg 10
24
g
1.007 276 466 88 u u
m
p
/m
e
1836.152 672 61
neutron rest mass
3
m
n
1.674 927 28 10
27
kg 10
24
gm
1.008 664 915 60 u u
m
p
/m
n
1.001 378 418 70
deuteron rest mass m
d
3.343 583 35 10
27
kg 10
24
gm
2.013 553 212 70 u u
m
d
/m
p
1.999 007 500 82
muon rest mass m

1.883 531 40 10
28
kg
0.113 428 9264 u u
m

/m
e
206.768 2838
permittivity of vacuum
4

0
8.854 187 817 . . . 10
12
F m
1
permeability of vacuum
5

0
12.566 370 614 10
7
N A
2
gravitational constant G 6.6742 10
11
m
3
kg
1
s
2
10
5
dyn cm
2
g
1
Boltzmann constant
6
k
B
1.380 6505 10
23
J K
1
10
16
erg K
1
8.617 343 10
5
eV K
1
k
B
/h 2.083 6644 Hz K
1
k
B
/hc 69.503 56 m
1
K
1
1
From the energy-mass relation E = m
e
c
2
, we getm
e
= 0.510 998 918 MeV/c
2
.
2
For the proton m
p
= 938.72 029 MeV/c
2
.
3
For the neutron m
n
= 939.565 360 MeV/c
2
.
4
Note that c = 1/

0
and
0
= 10
7
/(4c
2
) 1/36
2
10
9
C
2
N
1
m
2
. Also useful is the
value 1/4
0
= 8.98755 10
9
kg m
3
s
2
C
2
= 8.98755 10
9
N m
2
C
2
.
5
The value of the vacuum permeability is exact:
0
= 4 10
7
.
6
The Boltzmann constant k
B
= R/N
A
, where R is the molar gas constant.
Table C.3. More fundamental constants. See also Tbl. C.2. For Plancks constant, see Tbl. C.1.
Quantity Symbol value in SI and other useful units
Bohr magneton
1

B
9.274 009 49 10
24
J T
1
5.788 381 804 10
5
eV T
1
nuclear magneton
2

N
5.050 786 6 10
27
J T
1
3.152 451 66 10
8
eV T
1
electron magnetic moment
e
928.476 412 10
26
J T
1

e
/
B
1.001 159 652 1859

e
/
N
1838.281 971 07
proton magnetic moment
p
1.410 606 71 10
26
J T
1

p
/
B
1.521 032 206

p
/
N
2.792 847 351
neutron magnetic moment
n
0.966 236 45 10
26
J T
1

n
/
B
1.041 875 63

n
/
N
1.913 042 73
deuteron magnetic moment
d
0.433 073 75 10
26
J T
1

d
/
B
0.466 975 447 10
3

d
/
N
0.857 438 2329
electron g factor
3
g
s
2.002 319 304 3718
proton g factor 2
p
/
N
g
p
5.585 694 701
neutron g factor 2
n
/
N
g
n
3.826 085 46
electron gyromagnetic ratio 2[
e
[/
e
1.760 859 74 10
11
s
1
T
1
proton gyromagnetic ratio 2
p
/
p
2.675 222 05 10
8
s
1
T
1
neutron gyromagnetic ratio 2[
n
[/
n
1.832 471 83 10
8
s
1
T
1
1
In SI units the Bohr magneton is dened as
B
= e/2m
e
. A useful form for
current experimental physics is
B
= 1.4 MHz Gauss
1
.
2
The nuclear magneton is dened as
N
= e/2m
p
.
3
The electron g factor is dened as g
s
2(1 + a
e
), where a
e
is the electron
magentic moment anomaly a
e
[
e
[/
B
1 = 1.159 652 1859 10
3
.
Table C.4. Fundamental constants related to magnetic moments.
Observable Symbol Estimate
1 radian 60

speed of light in vacuum c 3 10


8
m s
1
Planck constant h 4 10
15
eV s
6.6 10
34
J s
rationalized Planck constant 6.6 10
16
eVs
1.0 10
34
J s
c 1240 eV

A
hc 1/8000 eV cm
elementary charge e 1.6 10
19
C
e
2
0
1.4 eV nm = 2.3 kg m
3
s
2
Rydberg constant R

13.6 eV
109 737 cm
1
Avogadros constant N
A
6 10
23
mol
1
ne-structure constant 1/137
Bohr radius a
0
0.53

A = 0.53 10
10
m
a
2
0
0.28

A
2
atomic radius (ground state) 1

A
nuclear radius (A=nuclidic mass number) R
n
1.2A
1/3
10
15
m
electron rest mass m
e
9.1 10
31
kg
m
e
c
2
0.5 MeV = 0.5 10
6
eV
proton rest mass m
p
1.7 10
27
kg
m
p
c
2
940 MeV 1 GeV
ratio m
p
/m
e
2000
neutron rest mass m
n
1.7 10
27
kg
permeability of vacuum
0
4 10
6
H m
1
permittivity of vacuum
0
8.8 10
12
F m
1
Boltzmann constant k
B
8.6 10
5
eV K
1
k
B
T for T = 300 K 1/40 eV = 0.025 eV
k
B
T for T = 10
4
K 1 eV
Table C.5. Handy estimates of key fundamental constants.
Appendix D
The Conversion Factory
Version 8.11: August 3, 2010
Introduction to unit conversions. To convert a physical quantity from one system of units to another,
multiply the value in the original units by the appropriate conversion factor. Ill use f to denote a generic
conversion factor. For instance, to convert energy from electron volts to joules, we use the denition of the
electron volt in SI units, which to four-digit precision is 1 eV = 1.602 10
19
J, to dene the conversion factor
f =
1.602 10
19
J
1 eV
. (D.1a)
This is the factor we need to perform the conversion:
E( J) = E( eV) f = E( eV) (1.602 10
19
J/eV). (D.1b)
For instance 2 eV = 3.204 10
19
J.
Warning: The numerical value of the conversion factor f is f = 1. It is dimensionless. Multiplying
E( eV) by f doesnt change the amount of energy. Rather, it changes the value in a particular unit: the
original unit is the electron volt, the new unit is the joule.
How to use conversion tables. Tables D.1 and D.2 contain conversion factors for mass and length,
respectively. Conversion factors for energy appear in Tbls. D.3 and D.4. To convert from the units in the
rst column (the original system of units) to the units in any other row (the new system), multiply by the
conversion factor in the corresponding row-column position.
1
How to dene a conversion factor not given in the tables. All you need is the (dimensional) equation
that denes the quantity you want to convert (see Appendix B) To obtain the desired conversion factor, just
replace each symbol in this equation by the conversion factor for that symbol. You can also use this gambit
to convert to so-called natural systems such as atomic units (see Appendix F).
2
1
Data in these tables come from the NIST web site http://physics.nist.gov/ and are reported to all accurate decimal
digits as of January 2009.
2
A cautionary note: As you use tables like those in this appendix, stay alert. For some quantities, the same symbol represents
a quantity that in dierent systems of units has dierent dimensions. This situation often arises in electromagnetism. For
instance, the SI unit of electric eld, V m
1
, is equivalent to the Gaussian unit 1/(3 10
4
) statvolt cm
1
. But since the
dimensions of electric eld in these two units are dierent, this quantity has dierent dimensions.
Appendix 48
MASS
Unit kg g Unied atomic mass unit (u)
1
1 kg 1 10
3
6.022 1367 10
26
1 g 10
3
1 6.022 1367 10
23
1 u 1.660 540 2 10
27
1.660 540 2 10
24
1
1
Convert from unied atomic mass units to atomic units (a.u.) of mass via
1 u = 1822.888 (au)
1au = 5.485 799 03 10
4
u.
Note that 1 u c
2
= 1.492 419 104 10
10
J = 931.494 3336 MeV.
Table D.1. Conversion factors for mass.
LENGTH
Unit m
1
cm mm

A fm
1 m (meter) 1 10
2
10
3
10
6
10
10
10
15
1 cm (centimeter) 10
2
1 10 10
4
10
8
10
13
1 mm (millimeter) 10
3
10
1
1 10
3
10
7
10
12
1 (micron) 10
6
10
4
10
3
1 10
4
10
9
1

A (angstrom) 10
10
10
8
10
7
10
4
1 10
5
1 fm (fermi) 10
15
10
13
10
12
10
9
10
5
1
1
1 inch = 0.254 m.
Table D.2. Conversion factors for length.
ENERGY: TABLE A
Unit electron volt (eV)
2
joule (J)
3
cm
1
1 eV 1 1.602 177 33 10
19
8065.541 0
1 J 6.241 5064 10
18
1 5.034 1125 10
22
1 cm
1
1.239 842 44 10
4
1.986 4475 10
23
1
1 K 8.617 385 10
5
1.380 658 10
23
0.695 038 7
1 E
h
27.211 396 4.359 748 2 10
18
2.194 746 306 7 10
5
1 Ry 13.605 6923 2.179 8741 10
18
1.0974 10
5
1 MHz 4.135 667 43 10
9
6.626 0755 10
28
3.335 640 952 10
5
1
1 keV = 10
3
eV, and 1 MeV = 10
6
eV.
2
The CGS unit of energy is the erg: 1 J = 10
7
erg = 10
7
dyn cm, where
1 dyn = 1 g cm s
1
.
3
In chemistry, energies are often reported in kilocalorie or kilocalorie per mole,
where 1 kcal = 2.613 194 10
22
eV and 1 kcal mol
1
= 0.0433 931 3 eV. (One
mole of a substance contains Avogadros number of particles of the substance.)
The conversion factors for atomic units are 1 kcal = 9.603 193 10
20
E
h
and
1 kcal mol
1
= 1.594 649 10
3
E
h
.
Table D.3. Conversion factors for energy.
PrintAppendices Version: 8.11 Printed: August 3, 2010
ENERGY: TABLE B
Unit K hartree ( E
h
) MHz
1
1 eV 1.160 445 10
4
0.036 749 309 2.417 988 36 10
8
1 J 7.242 924 10
22
2.293 7104 10
17
1.509 188 97 10
27
1 kcal/mole 5.0356 10
2
1.5946 10
3
1.0492 10
7
1 cm
1
1.438 769 4.556 335 267 2 10
6
c = 299 792 458
1 K 1 3.166 829 10
6
2.083 674 10
4
1 E
h
3.157 733 10
5
1 6.579 683 899 9 10
9
1 MHz 4.799 216 10
5
1.519 829 850 8 10
10
1
1
The most common unit in spectroscopy is the Hertz, 1 Hz = 1 s
1
. Various prexes
are attached to this unit, such as 1 MHz = 10
6
Hz. For translation of prexes, see
Appendix L. For more help with the arcana of spectroscopic data, see Appendix J.
2
Sometimes physicists report atomic and molecular energies in units of Rydberg. The
Rydberg is not the atomic unit of energy; that is the Hartree ( E
h
). For conversion
between energies in Rydberg and in Hartree and for more about the Rydberg unit,
see Appendix F.
Table D.4. More conversion factors for energy.
Appendix E
SI units (the International System of Units)
Version 8.9: August 3, 2010
Table E.1 contains the seven independent base SI (Syst`eme International) units. Derived units related to
angle are in Tbl. E.2, related to electromagnetic elds in Tbl. E.3, and other derived units in Tbl. E.4. For
values of constants in SI units, see Appendix C; for conversion factors, see Appendix D. For conversions
to and from atomic units, see Appendix F. Inconsistencies concerning jargon related to magnetic elds are
addressed in the Aside on p. 52; see also Chap. 13.
Physical quantity SI unit Dimension
Name Symbol
length meter m L
mass kilogram kg M
time second s T
electric current ampere A I
thermodynamic temperature kelvin K
amount of substance mole mol
luminous intensity candela cd
Additional notes
1 atomic mass unit = 1 u = 1.660 54 10
27
kg
Table E.1. The seven independent base quantities of SI units.
Physical quantity SI unit
Name Symbol
plane angle radian rad
solid angle steradian sr
Additional notes
The angular units radial and steradian are dimensionless.
Table E.2. SI derived units for angles.
Appendix 51
Physical quantity Symbol for SI unit dimension
electric current density j A m
2
L
2
I
electric dipole moment d C m L T I
magnetic dipole moment J T
1
L
2
I
electric eld strength c V m
1
= kg m s
3
A
1
M L T
3
I
1
electric displacement D C m
2
L
2
T I
electric charge density C m
3
L
3
T I
magnetic intensity H A m
1
L
1
I
vector potential A T m M L T
2
I
1
permeability H m
1
M L T
2
I
1
Additional notes
1 electron volt = 1 eV = 1.602 18 10
19
J
The unit tesla is denoted by T. The dimension time is denoted T .
Table E.3. SI derived units related to the electromagnetic eld.
Physical quantity SI unit Dimension
Name Symbol Denition
electric charge coulomb C A s Q = I T
energy joule J N m = kg m
2
s
2
M L
2
T
2
force newton N m kg s
2
M L T
2
frequency hertz Hz s
1
T
1
pressure pascal Pa N m
2
M L
1
T
2
power watt W J s
1
M L
2
T
3
electric potential volt V J C
1
M L
2
T
3
I
1
magnetic ux weber Wb kg m
2
s
2
A
1
M L
2
T
2
I
1
magnetic eld tesla T V s m
2
M T
2
I
1
angular velocity s
1
, rad s
1
T
1
wave number, decay constant m
1
L
1
density kg m
3
M L
3
energy density J m
3
= kg m
1
s
2
E L
3
charge density C m
3
= A m
3
s Q L
3
current density A m
2
Q L
3
T
1
permittivity
0
F m
1
= kg
1
m
3
s
4
A
2
Q
2
W
1
L
1
action J s M L
2
T
1
permeability
0
H m
1
= kg m s
1
A
2
electric capacitance farad F C V
1
Additional notes
The word permittivity (of free space) often refers to
0
4
0
.
Table E.4. SI derived units.
PrintAppendices Version: 8.9 Printed: August 3, 2010
Appendix 52
Aside. Jargon: magnetic induction and magnetic eld
In physics handbooks and many advanced textbooks the symbol B in SI units, which I call the magnetic
eld, is called the magnetic ux density or the magnetic induction. Its units are tesla (T), and its
the quantity that appears in Maxwells and related equations, such as B(r) = 0 and B(r) = A(r).
The quantity I call the magnetic intensity is, in other texts, called the magnetic eld (or, confusingly, the
magnetic-eld strength). This quantity is dened by H B/
0
M, where M is the magnetization (the
total magnetic moment per unit volume) and 0 is the permeability of the vacuum, 0 4 10
7
N/ A
2
.
If the magnetization is zero, then H B/
0
. It is H that appears in, for example, the Poynting vector
S E H, but B that appears in the Lorentz force law F = q(E +v B).
PrintAppendices Version: 8.9 Printed: August 3, 2010
Appendix F
Everything you always wanted to know about atomic
units but were afraid to ask
Version 8.22: August 3, 2010
How submerged does a reference have to be before it drowns?
Flauberts Parrot, by Julian Barnes.
F.1 Introducing atomic units
F.1.1 How units work
To explain atomic units, I have to rst remind you of something very basic about units in general. Ill do
so by example. Suppose you want to evaluate the rst Bohr radius a
0
in, say, SI units. You start with an
equation for a
0
, its denition
a
0


2
m
e
e
2
0
=

2

0
m
e
e
2
(F.1a)
As discussed in the Aside on p. 59, for this appendix Ive introduced a new symbol,

0
4
0
=
1
c
2
10
7
A
2
N
1
= 1.1126501 10
10
s
2
m
2
, (F.1b)
where the last equality gives its value in SI units.
1
To evaluate a
0
in SI units, I just look up the values in SI
units of the (so-called) rationalized Plancks constant , the electron mass m
e
, and the electronic charge e
in Appendix C:
a
0
=
(1.05457 10
34
J s)
2
(1.11265 10
10
C
2
J
1
m
1
)
(9.10938 10
31
kg) (1.60218 10
19
C)
2
= 5.29177 10
11
m. (F.1c)
Now suppose I want to calculate the same quantity in a dierent set of unitssay, electrostatic CGS
units. I proceed as above but plug into Eq. (F.1a) values of , m
e
, and e in this other system of units. (In
CSG units
0
= 1):
1
Notation: The dots, , indicate that this number is exact; I show only eight decimal digits. The quantity

0
1/
0
c
2
= 8.854 19 10
12
Fm
1
is the permittivity of the vacuum, with
0
= 4 10
7
N A
2
being the permeability
of the vacuum. Since the value of c is exact, c 299 792 458 m s
1
, so is the value of
0
. No other base unit in Tbl. F.1 is
exact.
Appendix 54
= 1.05457 10
27
erg s
m
e
= 9.10938 10
28
g
e = 4.80321 10
10
esu
_

_
= a
0
= 5.29177 10
9
cm, (F.2)
where the CGS unit of mass is the gram (g), of energy is the erg, and of charge is the statcoulomb (also
called the electrostatic unit esu). The result of my little calculation is the rst Bohr radius in the CGS
unit of length, the centimeter.
To isolate the numerical value of a
0
in these systems of units, I just divide each result by its unit:
a
0
m
= 5.29177 10
11
value of a
0
in SI units
a
0
cm
= 5.29177 10
9
value of a
0
in CGS units
(F.3)
F.1.2 How atomic units work
All this probably seems excruciatingly tedious. But if you understood what I just did, atomic units should
make sense. Atomic units is a coherent system of units, just like SI units and CGS units. Atomic units diers
from these systems in that the atomic unit of a physical quantity is the value of the property appropriate
to systems on the scale of atoms and molecules. For instance, the atomic unit of mass is the mass of the
electron, m
e
. Lets look at the mass of a proton:
m
p
= 1.67262 10
27
kg the proton mass in SI units (F.4a)
=
m
p
kg
= 1.67262 10
27
the value of the proton mass in SI units (F.4b)
or
m
p
m
e
= 1836.15 the value of the proton mass in atomic units, (F.4c)
where in the last step I used m
e
= 9.1094 10
31
kg. The logic of this little calculation parallels the logic
of Eqs. (F.1). Now, by convention, users of atomic units would write the value of the proton mass in atomic
units as m
p
= 1836.15 au.
Key point! The atomic unit of a physical quantity is a characteristic value (with units) of the quantity for
systems on an atomic and molecular scale. Only four physical quantities are independent: m
e
, e, , and
0
; these
are given in Tbl. F.1.
Table F.1. The independent
atomic units. Note that the
atomic unit of action is the ratio-
nalized Plancks constant, h/2,
not Plancks constant h. Also note
that the atomic unit of permittiv-
ity is
0
4
0
of Eq. (F.1b), not
0
.
The unit W is the Watt, the SI unit
of power.
quantity symbol dimension value in SI units
mass m
e
[m
e
] = M 9.1094 10
31
kg
charge e [e] = Q 1.602 10
19
C
action [] = E T = M L
2
T
1
1.0546 10
34
J s
permittivity
0
[
0
] = Q
2
E
1
L
1
10
7
c
2
C
2
J
1
m
1
PrintAppendices Version: 8.22 Printed: August 3, 2010
Appendix 55
F.1.3 Do-it-yourself atomic units.
In atomic units, only the four quantities in Tbl. F.1 are independent. All other physical quantities are
constructed as combinations of these four. I nd it easiest to construct the atomic unit of a quantity from its
dimension (see Appendix B), which is why I included this information in Tbl. F.1. I start with the question,
how can I combine m
e
, e, , and
0
to give me a quantity with the dimensions I need? The answer is the
atomic unit of that quantity. Here are some examples:
2
length:
_

2
m
e
e
2
0
_
= L = atomic unit of length, the Bohr a
0
, (F.5a)
time:
_

3
m
e
e
4
0
_
= T = atomic unit of time, (F.5b)
energy:
_
m
e
e
4
0

2
_
= E = atomic unit of energy, the hartree E
h
. (F.5c)
You can use this method to construct the atomic unit of any physical quantity, be it mechanical, electric,
or magnetic. Such machinations often yield physical insight into the atomic unit youve constructed. For
instance, in Eq. (F.5a), we discovered that the atomic unit of length is the rst Bohr radius. And in Eq. (F.5c),
we discovered that the atomic unit of energy is the expression weve known since Chap. 5 as the hartree.
3
Key point! To determine the atomic unit of any quantity, combine the base atomic units in Tbl. F.1 into an
expression with the dimensions of the quantity. From the resulting expression, calculate the value of the unit in
any system of units by plugging in values of m
e
, e, , and
0
all in that system of units.
Example F.1 (The atomic unit of speed)
Atomic units are based on the Bohr classical model of an atom. SO you can also generate atomic units using
classical expressions, if youre careful about interpretation. For instance, suppose you need the atomic unit of
speed and have lost Tbl. F.4. In classical mechanics, the speed of a particle is related to its kinetic energy by
E = mv
2
/2, which implies v =
_
2E/m. For an electron in the rst Bohr orbit, the magnitude of the kinetic
energy is E1 = E
h
/2. (We use the magnitude to ensure that the atomic unit is positive.) Using the expression
for the hartree E
h
in terms of the base units, E
h
m
e
e
4
0
/
2
, we get v
B
=
_
E
h
/m
e
= e
2
0
/, which is indeed
the expression the atomic unit of speed (see Tbl. F.4).
F.2 How to make atomic units work for you
F.2.1 Converting physical quantities
How do we determine the value in atomic units of a physical quantity that is not one of the big four
in Tbl. F.1. How, for instance, do we gure out the value of the speed of light in atomic units? Again, the
key is dimensions: [c] = L T
1
. The only combination of the big four with these dimensions is e
2
0
/, which,
as we learned in Example F.1, is the atomic unit of speed. To get the numerical value of the speed of light
in atomic units, we just divide by this unit:
c
e
2
0
/
=
c
e
2
0
=
(1.054 571 628 10
34
J s)(299 792 458 m s
1
)
2.307 077 13 10
18
J m
= 137.036. (F.1a)
Whoa! The number we got is the value of one over the (dimensionless) ne-structure constant, e
2
0
/c.
By blindly plugging numbers into the expression in Eq. (F.1) (something you should never do!) we missed
2
Commentary: That the atomic unit of length is the Bohr corresponds, in terms of the dimensional procedures of Appendix B,
to choosing a length scale factor that is proportional to a
0
. Similarly, the hartree corresponds to choosing an energy scale
factor proportional to E
h
.
3
Details: Equation (F.5b) hasnt got an ocial name (although many have proposed the atomic second), but its easy to
interpret: other equations in Eqs. (F.5) tell us that the atomic unit of time is the period of an electron in the rst Bohr orbit.
PrintAppendices Version: 8.22 Printed: August 3, 2010
Appendix 56
quantity symbol and equations
rst Bohr radius a
0


2
m
e
e
2
0
=
c
(m
e
c
2
)
=
e
2
0
4cR

Rydberg constant R

=
m
e
e
4
0
4
3
c
=

2
m
e
c
4
hartree E
h

m
e
e
4
0

2
=

2
m
e
a
2
0
=
e
2
0
a
0
=
2
(m
e
c
2
) = 2R

hc
ne-structure constant
e
2
0
c
=

m
e
c a
0
e
2
0

e
2
4
0
= E
h
a
0
Table F.2. Denitions and useful equations for key quantities of atomic units. The rst bohr
radius a0 is the atomic unit of length. The hartree E
h
is the atomic unit of energy. The quantity 0 is
the permittivity of the vacuum. The ne-structure constant is dimensionless. Plugging into the above
expressions the values for fundamental constants in SI units will generate the rst Bohr radius in meters
and the Hartree in Joule. To convert the former to angstrom, multiply to 10
10
; to convert the latter to
electron volts, divide by the elementary charge e. February 26, 2010
something important: the value of c in atomic units is one over :
4
c =
1

=
1
137.036
au speed of light in atomic units. (F.1b)
Key point! To evaluate a constant in atomic units, divide its symbol by the atomic unit that corresponds to
its dimensions. Then either plug in values from Tbl. F.1 (or, if you need the most accurate current values, from
http://physics.nist.gov/cuu/Constants), or use an equation that relates the appropriate combination of
m
e
, e, , and/or
0
to a constant whose value you know. You should use the latter approach whenever possible.
F.2.2 Converting expressions
A common chore in physics is calculating a quantity in a common system of units, such as SI units, from an
equation in atomic units. The following examples illustrate how to do this without making mistakes.
Example F.2 (Lengths in atomic units: the mean radius of a hydrogenic atom.)
The symbol a
0
is just shorthand for the combination of fundamental constants in Tbl. F.2. It has dimensions
of length, [a0] = L , but it has no value until we choose a system of units. To three decimal digits, its values are
a0 = 0.529 10
10
m = 0.529

A. (F.2)
When we write, say, the expression for the mean radius of a hydrogenic atom in a stationary angular momentum
eigenstate [
nm

) as (see Tbl. ??, p. ??)


r)
n
=
1
2Z
_
3n
2
( + 1)
_
a
0
, (F.3a)
4
A cautionary note: This development in Eqs. (F.1) highlights something important about units. The fundamental con-
stants e
2
0
and appear in Eq. (F.1a) only in the combination e
2
0
/. Since this combination is the inverse of the ne-structure
constant, which is one of the most accurately known constants in physics, we should calculate the value of c in atomic units
via Eq. (F.1b) rather than Eq. (F.1a). Doing so ensures the most accurate possible value of c at the time of calculation.
Moreover, realizing that in atomic units c is 1/ sheds insight into the underling nature of atomic processes.
PrintAppendices Version: 8.22 Printed: August 3, 2010
Appendix 57
we are not using atomic units; were just using the symbol a
0
. To transform Eq. (F.3a) into atomic units, we
invoke the replacement rule a0 1 (Tbl. F.3):
r)
n
=
1
2Z
_
3n
2
( + 1)
_
au, (F.3b)
Now, suppose we want to convert the mean radius of a hydrogen atom (Z = 1) in a 2p state from atomic
units to, say, angstrom. According to Eq. (F.3a),
r)
2p
= 5 a
0
= 5 bohr. (F.4a)
Equation (F.2) gives the conversion factor from the atomic unit of length, the bohr a0, to angstrom: 1 bohr is
about half an angstrom. More precisely (to three decimal digits) r)
2p
is
5
r)
2p
= 5 bohr
0.529

A
1 bohr
= 2.646

A. (F.4b)
This calculation illustrates how to determine the conversion factor from an atomic unit to some other system
of unit: the length conversion factor from atomic units (bohr) to angstrom (to three decimal digits) is
6
f( bohr

A) =
0.529 177

A
1 bohr
. (F.5a)
We can use f( bohr

A) to convert any algebraic expression for or value of a length from atomic units to
angstrom. Equation (F.5a) also denes the conversion factor for going the other way:
7
f(

A bohr) =
1 bohr
0.529 177

A
=
1.889 73 bohr
1

A
. (F.5b)
This example reveals another advantage to atomic units: results given in atomic units are independent of
the values of the fundamental constants at the time of the calculations.
8
Example F.3 (The atomic unit of energy: the Hartree versus the Rydberg.)
Sometimes, physicists report energies in the Rydberg unit (see Appendix J). The rst thing you must remember
is that the Rydberg is not the atomic unit of energy:
9
The Rydberg is dened as the value of the Rydberg
constant R, the SI units of which are m
1
, converted to energy units:
1 Ry hc R, the Ryberg unit of energy (F.6a)
5
Commentary: It would be confusing to introduce the symbol a
0
into the conversion factor: that is, to write f = a
0
/1 bohr.
Doing so is a bad idea because it leaves ambiguous the units of a
0
. For example, in

A, a unit dened by 1

A = 10
10
m, the
rst Bohr radius is a
0
= 0.529

A. How would we know whether f converted lengths in bohr to lengths in



A, in meters, or in
some other length unit?
6
A cautionary note: Whenever you express a physical quantity in a dimensional system of units, you must pay attention
to the accuracy of the conversion factor you use, which must be greater than the accuracy of the number youre converting. If
you convert, say, a length in angstrom to bohr, be sure your conversion factors carries more signicant gures than the number
youre converting. These issues dont arise if youre solving equationssay, solving the Schrodinger equation by numerical
integrationin atomic units or some other dimensionless system. For more on this point, see Appendix B.
7
A cautionary note: Equations like (F.5) obscure two important points. First, the value of the conversion factor is 1. And
second, the conversion factor is dimensionless. [Both points emerge if we plug Eq. (F.2) into Eq. (F.5a).] The value of any
conversion factor must be unity, and any conversion factor must be dimensionlessotherwise converting a physical quantity
from one system of units to another would change the actual value of the physical quantity rather than just its value in a
particular system of units.
8
Commentary: This remark may come as a shock. Surely fundamental constants are constant. Nope. These values are
based on measurements, so they change with improved measurement devices and techniques. For example, the recommended
value of Plancks constant in 1929 [cited in R. T. Birge, Rev. Mod. Phys., 1, 1 (1929)] was 6.457 10
34
Js, but the current
(2010) value is 6.6261 10
34
J s. This dierence wont aect an angular momentum quantity calculated in 1929 if it was
reported in atomic units; all that has changed since then is the conversion factor. Even the mass of the electron has changed:
its value in 1929 was 9.035 10
31
kg, but the current value is 9.1094 10
31
kg.
9
Jargon: Considerable confusion is sown by authors who refer to energy in atomic units when they mean energy in
Rydberg. This makes the phrase energy in atomic units ambiguous. What units does it refer to? Hartree? Rydberg? Only
the author knows for sure.
PrintAppendices Version: 8.22 Printed: August 3, 2010
Appendix 58
Using the Rydberg constant R

in terms of fundamental constants (see Tbl. F.2), we can relate the Rydberg
unit of energy to the atomic unit of energy, the Hartree as
10
1 Ry =
1
2
E
h
, = E
h
= 2hcR (F.6b)
This relationship denes factors for conversion of a quantity f between these two energy units:
f( Hartree Ry) =
2 Ry
1 Hartree
f( Ry Hartree) =
1 Hartree
2 Ry
. (F.6c)
Rule: One Rydberg of energy equals half a Hartree. So to covert a value or an expression for energy from atomic
units (Hartree) to Rydberg multiply by 2.
For example, in the pure-Coulomb model with the approximation of innite nuclear mass the bound-state
energies of a hydrogenic atom (Chap. 5) are
11
E
n
=
Z
2
2n
2
E
h
=
Z
2
n
2
Ry. (F.7a)
According to the replacement rule E
h
1, this equation in atomic units reads
E
n
=
Z
2
2n
2
au
bound-state energy of a hydrogenic
atom (in atomic units).
(F.7b)
To evaluate the ground-state energy in eV, we use the value of E
h
(or, if we prefer, of Ry) from Tbl. F.4:
E
1
=
1
2
E
h
= 2
_

1
2
Ry
_
= Ry = 13.6057 eV. (F.8)
Warning: The atomic unit of energy is the Hartree, not the Rydberg.
12
F.2.3 Converting functions and operators
You now know how to convert constants and expressions into atomic units. But you also need to be able
to convert functions and, because this is quantum mechanics, operators. The approach is the same, but we
must give due attention to the variables involved. To illustrate, Ill convert the Hamiltonian of a hydrogenic
atom (Chap. 5).
Example F.4 (Converting a Hamiltonian to atomic units.)
To be specic, Ill convert the operator that appears in the reduced radial equation (see Chap. 5) for an s state
(one with = 0), so there is no centrifugal barrier term:

h (r)

2
2m
d
2
dr
2

Ze
2
0
r
,
radial Hamiltonian for a hydrogenic
atom of atomic number Z for = 0.
(F.9)
Both terms in

h (r) have dimensions of energy, so my rst thought might be to divide each term by the atomic
unit of energy, E
h
m
e
e
4
0
/
2
:
V (r)
E
h
=
V (r)
m
e
e
4
0
/
2
=
Z
2
m
e
e
2
0
1
r
. (F.10a)
This hardly represents progress: my result is more complicated than the one I started with. Where I went
astray was in not thinking before I started doing. I should brainstorm. The function I want to convert, the
Coulomb potential energy, is inversely proportional to r, which has dimensions [r] = L . So I want to divide
10
Details: Plugging the values of fundamental constants in SI units into this expression gives the Hartree in Joule. To convert
to electron volts, divide by the elementary charge e.
11
Details: In the theory of the hydrogenic atom, the zero of energy is chosen at the ionization continuum, so all bound states
have negative energies. However, units are never negative, so E
h
> 0 by denition.
12
Details: In reporting energies determined via spectroscopy, scientists often actually report quantities that are not energies
at all. Appendix J oers a rst-aid kit for users of spectroscopic data.
PrintAppendices Version: 8.22 Printed: August 3, 2010
Appendix 59
V (r) by a form of the atomic unit of energy, E
h
, that contains one power of the atomic unit of length, a
0
.
After a little manipulation (or looking at Tbl. F.2), I nd what I need:
V (r)
E
h
=
V (r)
e
2
0
/a
0
=
Z
r/ a
0
= V (r) =
Z
r
au (F.10b)
where in the last step I wrote the potential energy in atomic units, as indicated by the qualier au, and,
therefore dropped the factor a
0
from the radial variable r (see F.3).
Warning: The qualier au or the statement that an expression or equation is in atomic units implies
that all quantities and variables in the expression or equation are in atomic units.
Now Ill apply the same reasoning to the kinetic-energy operator in (F.9). This operator,

t(r), is proportional
to the second derivative d
2
/dr
2
. So I want to divide

t(r) by a form of E
h
that contains the square of the
atomic unit of length. Such a form is E
h
=
2
/m
e
a
2
0
, whence

2
/2me

2
/m
e
a
2
0
d
2
dr
2
=
1
2
d
2
d(r/ a0)
2
=

t(r) =
1
2
d
2
dr
2
au (F.11)
Putting together the pieces, I get the Hamiltonian (F.9) in atomic units,

h (r) =
1
2
d
2
dr
2

Z
r
au,
radial Hamiltonian for a hydrogenic
atom for = 0 in atomic units.
(F.12)
Warning: Once youve transformed expressions to atomic units, be extremely careful with subsequent alge-
bra. This transformation eliminates a powerful tool for checking algebra: dimensional analysis. While you can
easily use the dimensions of the stu on the right-hand side of Eq. (F.9) to ensure that this expression has the
required dimensions (energy), you cant do this with its counterpart in atomic units, Eq. (F.12).
13
Aside. Absorbing the permittivity.
Elsewhere in this book I consistently use the shorthand
e
2
0

e
2
4
0
= 2.307 077 13 10
18
J m. (F.13)
to avoid having to clutter up my equations with 4
0
, which appears in SI units. In this appendix, Ive
changed this notational convention: as you can see in Eq. (F.1a), Ive explicitly used the electronic charge
e > 0 and have absorbed the permittivity into yet another symbol,
0
of Eq. (F.1b).
F.3 Atomic unit replacement rules
Now that you know where atomic units come from, how they work, how to generate your own atomic units,
and how to evaluate other constants in these units, youre ready to learn an extremely easy way to bypass
the steps we just went through. Instead of laboriously guring out appropriate quantities to eliminate the
four base atomic units, m
e
, e, , and/or
0
, we just replace these quantities by unity, as m
e
1, 1,
and so on. We can do the same thing with certain combinations of these atomic units, such as a
0
1 and
E
h
1. Ive tabulated a bunch of these replacement rules in Tbl. F.3, and will conclude this appendix with
some examples of their implementation, and two collections of conversions, Tbls. F.4 and F.5.
14
Warning: Transformation to atomic units does not result in replacement of all fundamental constants
by 1.
15
13
Details: From a dimensional point of view, Eq. (F.12) is ridiculous: it seems to equate a quantity with dimensions E to
the sum of an expression with L
2
plus an expression with L
1
. The reason, of course, is that transformation to atomic
units eliminates many fundamental constants and combinations thereof. This features is useful in formulating easily applied
replacement rules (F.3), but exacts a price: it vitiates dimensional analysis.
14
Commentary: To convert a quantity from the unit indicated in the third column of each table into atomic units, just divide
by the number in the fourth column.
15
A cautionary note: The classic mistake made by physicists who forget this warning is to replace the speed of light by 1.
The correct replacement rule for the speed of light is c 1/, where is the ne structure constant.
PrintAppendices Version: 8.22 Printed: August 3, 2010
Appendix 60
Table F.3. Atomic unit replacement
rules. Note that these are replacement
rules, not equalities. While its true that
the value of, say, the electron mass in
atomic units is 1, to avoid confusion, you
should transform equations and expres-
sions by making replacements rather than
by invoking equalities. (Dont confuse
the Rydberg energy unit Ry, with the
atomic unit of energy, the Hartree E
h
, or
with the Rydberg constant R.) In the
row for the permittivity of the vacuum,

0
4
0
[Eq. (F.1b), p. 53]. The quan-
tity e
2
0
/c is the ne-structure con-
stant.
quantity transformation rule
Plancks constant 1
Electron mass m
e
1
Speed of light c 1/
Bohr radius a
0
1
Hartree energy unit E
h
1
Rydberg energy unit Ry 1/2
Permittivity of vacuum
0
1/4 =
0
1
Elementary charge e 1 = e
2
0
1
Permeability of vacuum
0
4
2
Proton mass mp 1836.152 671
Neutron mass m
n
1838.683 66
Rydberg constant R /4
Example F.5 (Converting an operator to atomic units: the kinetic-energy operator.)
In position space, the kinetic energy operator for an electron in three dimensions is

T =

2
2me

2
. (F.1a)
To convert this expression to atomic units, we invoke the replacement rules me 1 and 1 to obtain
16

T =
1
2

2
,
kinetic-energy operator for
an electron (in atomic units)
(F.1b)
If the particle in question isnt an electron, then Eq. (F.1b) isnt right. If its mass is m ,= m
e
, then
17

T =
1
2m

2
, (where m is in atomic units). (F.1c)
Example F.6 (The Schrodinger equation in atomic units.)
Given a quantum-mechanical equation, you can generate its counterpart in atomic units in two ways. You can
choose choose appropriate length and energy scale factors and use the dimensional-manipulation procedures
of Appendix B. Or you can use the atomic unit replacement rules. For reasons explained in Appendix B, I
(strongly) recommend dimensional manipulation. Here is an example of how to use replacement rules.
Suppose I want to covert the time-independent Schr odinger equation (TISE) for a hydrogenic atom,
_


2
2m
e

Ze
2
0
r
+En
_

nm

(r) = 0. (F.2a)
The atomic unit replacement rules of Tbl. F.3 tell me to set 1, m
e
1, and e
2
0
1; doing so gives
18
_

1
2

Z
r
+E
n
_

nm

(r) = 0. (F.2b)
16
A cautionary note: This little derivation illustrates an important point: when you use Eq. (F.1b), you must remember
that the derivatives in the Laplacian are to be taken with respect to spatial variables in atomic units (bohr). You also must
remember, of course, that any energies you calculate using this form of

T will be in Hartree.
17
A cautionary note: Dont forget that m must be in atomic units. For, say, a helium atom, the atomic weight is
m
He
= 4.0026 u = 6.64648 10
27
kg. What you plug into Eq. (F.1c) for m, however, must be m = m
He
/me = 7296.3 au.
18
A cautionary note: Using atomic units requires you to remember a lot of stu. Implicit in Eq. (F.2b) is that the particle is
an electron, that Im treating the atom in the approximation of innite nuclear mass, that Z is the nuclear charge in atomic
units, that the radial variable and derivatives in
2
are dened in atomic units (bohr), and that the Hamiltonian eigenfunction
must be multiplied by a
3/2
0
to ensure proper normalization.
PrintAppendices Version: 8.22 Printed: August 3, 2010
Appendix 61
Aside. Atomic units: an historical expose.
Douglass R. Hartree introduced atomic units in 1928 to facilitate calculations in the atomic and subatomic
realms.
19
For microscopic particles, conventional SI (or CGS) units are inconvenient, because in these units
the values of most physical quantities are so small. This problem predates modern quantum mechanics: in
the Bohr model of atomic hydrogen, electrons were thought to move in circular orbits of radii r
n
= n
2
a
0
for n = 1, 2, 3, . . ., where a
0
= 0.529 10
10
m.
20
How much nicer if the value of a
0
were unity. Then
the magnitude of quantities like r) and r for atoms and molecules would be numbers on the order of a
few hundredths to a few tenshandy for calculating, graphing, and remembering. Hartree achieved this
ideal by choosing as the unit of length not the meter (the SI unit) but rather the particular combination
of fundamental constants that dene a
0
. The result is a new length unit he called the bohr. Since one
atomic unit (abbreviated a.u.) of length is one bohr, the radius of the rst Bohr orbit, 1 a0, in atomic units
is 1 au = 1 bohr; the radius of the second Bohr orbit is 4 au = 4 bohr. And so on.
Aside. The philosophical signicance of atomic units.
The central dening replacement rule for atomic units is 1. Because this rule puts observables like
momentum p and wavenumber k on a common footing, it leads to a profound interpretation of the de Broglie-
Einstein relations (Appendix O)pairs of physical quantities previously thought to be distinct are in fact
the same. From this perspective, the role of in p = k is to connect two quantities, momentum and
wavenumber, that carry the same information about the particle. Similarly, energy and frequency in E =
are, in this sense, the same physical quantity. Thus is a universal constant of nature, unrelated to the
particulars of the system in whose equations it appears. In SI units, the identity of p and k (and of E and )
is obscured by the factor . But in atomic units, does not appear, and the underlying truth that p = k and
E = emerges. This revealed truth is wave-particle duality; as Arno Bohm writes about the electron in
Sec. II.9 of Quantum Mechanics: Foundations and Applications, Third Edition (New York: Springer, 1993):
particles and waves are just two dierent appearances (of many possible appearances) of the quantum
physical system electron.
19
Read on: D. R. Hartree, Proc. Camp. Phil. Soc. 24, 89 (1928).
20
Read on: Except in tables, Ive quoted values of physical constants and conversion factors to a precision of three signicant
gures. For more accurate values refer to Appendix C and Appendix D. For additional information, see the web-based reference
on Constants, Units, and Uncertainty maintained by the National Institute of Standards and Technology at physics.nist.
gov/cuu/Constants/index.html.
PrintAppendices Version: 8.22 Printed: August 3, 2010
atomic unit of symbol (if any) and equation to covert to multiply by
charge e C 1.602 176 487 10
19
length a0 =
2
/me e
2
0

A 0.529 177 208 59
cross section a
2
0

A
2
0.280 028 518
a
2
0

A
2
0.879 735 534
mass m
e
kg 9.109 382 15 10
31
time
1

0
= / E
h
=
3
/m
e
e
4
0
s 2.418 884 327 10
17
energy E
h
= m
e
e
4
0
/
2
eV 27.211 383 86
J 4.3597 439 10
18
Ry = E
h
/2 eV 13.605 691 93
velocity v
B
= e
2
0
/ = c m s
1
2.187 691 2633 10
6
action (angular momentum) J s 1.054 571 628 10
34
frequency
2
E
h
/ Hz 4.134 137 333 657 10
16
wavenumber
3
1/a0 cm
1
1.889 726 132 886 10
8
linear momentum / a
0
= m
e
c kg m s
1
1.992 851 66 10
24
force e/
0
a
2
0
N 8.238 7221 10
8
1
The atomic unit of time is the time required for an electron to make one transit of the rst Bohr
orbit in atomic hydrogen. This unit is sometimes called the atomic second.
2
Note that cR

= 3.289 841 960 361 10


15
Hz.
3
Youre most likely to need to convert wave numbers from atomic units when youre working with
spectroscopic data. In spectroscopy, by long-standing convention, energies are given in inverse
centimeter, that is, what you actually nd labeled energy in tables are actually wave numbers
1/ (see Appendix J). Note that this conversion factor is not in SI units. (The SI unit of wave
number is called the Kayser.)
Table F.4. Atomic-unit conversions. To convert a quantity from the unit indicated in the third
column into atomic units, divide by the number in the fourth column. [MM: August 3, 2010]
atomic unit of symbol (if any) and equation to covert to multiply by
current e E
h
/ A 6.623 6176 10
3
charge density e/ a
3
0
C m
3
1.081 2023 10
12
electric potential
1
e/
0
a
0
V 27.211 383 87
electric eld strength c
0
= e/
0
a
2
0
V m
1
5.142 206 323 10
11
electric ux density
2
E
h
/0a
2
0
W m
2
6.436 408 811 10
11
power E
h
/
0
W 0.1802 378
intensity
3
c 0(E
h
/ea0)
2
/8 W m
2
3.509 4449 10
20
electric dipole moment
3
e a
0
C m 8.478 352 807 10
30
electric quadrupole moment e a
2
0
C m
2
4.486 551 07 10
40
electric polarizability e
2
a
2
0
/ E
h
C
2
m
2
J
1
1.648 777 251 10
41
magnetic dipole moment
4
e/m
e
J T
1
1.854 801 828 10
23
magnetic eld strength
5
/ea
2
0
T 2.350 517 384 10
5
1
The atomic unit of electric potential is the potential at a distance of 1 a
0
from a unit test
charge. In CGS units, use 1/(4
0
) = 1 and e = 4.803 207 esu cm.
2
By denition, the intensity is the time-averaged ux of the energy carried by the electric eld:
that is, the product of the energy density times the speed at which the energy is propagating
through space. So some authors refer to the electric ux density as the intensity. More
recently, some have used
0
c c
2
0
/2 for the atomic unit of intensity, where c
0
is the atomic
unit of electric eld strength. This quantity equals the atomic unit of electric ux density divided
by 8, where is the ne-structure constant.
3
The most common unit of dipole moment is the Debye, where 1 D = 3.335 641 10
30
C m.
The factor for conversion to atomic units is 1 au = 2.54175 D. Theoretical dipole moment are
often reported in units of e a
0
, where 1 e a
0
= 2.54 Debye = 8.478 3528 10
30
C m.
4
The atomic unit of magnetic dipole moment is twice the Bohr magneton, B = e/2me, whose
value in SI units is 927.400 915 10
26
J T
1
.
5
The unit of Gauss for a magnetic eld is related to Tesla by 1 T = 10
4
Gauss. In SI units, the
magnetic eld strength is usually called the magnetic ux density.
Table F.5. Atomic-unit conversions for quantities in electricity & magnetism. To convert a quantity
from the unit indicated in the third column into atomic units, divide by the number in the fourth column.
Appendix G
Dirac notation
Version 8.19: August 3, 2010
Throughout this book I use Dirac notation, mostly as shorthand for integrals (see G.1). But there is much
more to the story than Ive let on. A proper development of Diracs formulation of quantum mechanics
belongs in a more advanced book. But this appendix (together with Complement ?? of Chap. 7) pro-
vides enough information for you to exploit the power of this formulation without going into the detailed
mathematics behind it.
As discussed briey in Chap. 7, Dirac introduced symbols that represent a quantum state in a more
abstract way than do wave functions: these symbols represent elements of a linear vector space called a
Hilbert space. Diracs key idea was to represent a state of a system at a xed time t by an abstract
mathematical object he called a ket, or state vector, and denoted by [(t). According to one of the
fundamental postulates of quantum mechanics, [(t) contains all accessible physical information about the
system in that state at time t. The same state at a dierent time, t

, would be denoted by a dierent


ket, [(t

). Operators, according to Dirac, act in the Hilbert space to map one ket into another. For
instance, the kets that represent the same state at two dierent times are related by an operator called the
time evolution operator: [(t

) =

U(t

t) [(t). Unsurprisingly, the time-evolution operator can be


written in terms of the Hamiltonian operator

1.
To extract from [(t) physical information about possible outcomes of measurement of a particular
observable, we translate this abstract object into the language of the observable were going to measure.
We do this by constructing scalar products, such as
(x, t) = r [ (t) position: position-space wave function (G.1a)
(p, t) = p [ (t) momentum: momentum-space wave function (G.1b)
c
n
(t) =
n
[ (t) energy: energy probability amplitudes, (G.1c)
where [
n
represents the n
th
stationary state of the system whose state is [(t).
Having made the appropriate translation, we interpret the result according to the Born interpretation.
The quantity [(x, t)[
2
d
3
v, for example, is the probability of nding the particle in volume element d
3
v
at r. The quantity [c
n
(t)[
2
is the probability of obtaining the value E
n
in a measurement of the energy. And
so forth.
A Dirac ket need not correspond (via a scalar product) to a function. In Part II, while developing the
quantum mechanics of a spin-
1
/
2
particle, we often expanded an arbitrary spin state in a basis of eigenvectors
of

S
z
:
[(t) = c
+
(t) [
+
+ c

(t) [

, (G.2a)
In the Pauli representation, this expansion becomes a matrix equation involving 2 1 column matrices,
Appendix 65
(t) =
_
c
+
(t)
c

(t)
_
= c
+
(t)
+1/2
+ c

(t)
1/2
, (G.2b)
where
_

_
[
1/2

1/2

_
1
0
_
,
[
1/2

1/2

_
0
1
_
.
(G.2c)
The general spinor wave function is a 2 1 column matrix whose elements are ordinary wave functions.
G.1 Dirac shorthand
To use Dirac notation solely as shorthand, we proceed as follows:
we signify integration over spatial variables (and/or summation over spin variables, as in Chaps. 6
and 7) by putting the integrand (or summand) between brackets, and ;
we dont explicitly show the coordinate dependence of functions and operators;
1
we imply complex conjugation of a function by positioning this function between a left bracket and
a vertical line [.
Table G.1 summarizes this use of Dirac notation.
For example, we write the overlap integral of two one-dimensional (1D) wave functions,
1
(x, t)
and
2
(x, t), in Dirac and conventional notation by

1
(t) [
2
(t) =
_

1
(x, t)
2
(x, t) dx, overlap integral. (G.1a)
Similarly, we write the matrix element of an operator

Q with respect to these functions as

1
(t) [

Q [
2
(t) =
_

1
(x, t)

Q
2
(x, t) dx. matrix element (G.1b)
Mathematically, overlap integrals and matrix elements are scalar products (see ?? of Chap. 6).
Dirac shorthand further facilitates and generalizes writing basic quantum-mechanical equation such as
the omnipresent properties
(t) [ (t) = 0, orthogonal,
(t) [ (t) = 1, normalized,
_
orthonormal. (G.2)
Aside. Why you should learn Dirac notation.
One advantage of Dirac notation is its lack of specication of the spatial (or momentum) variables or even the
number of particles. Because Dirac notation suppresses this information, it allows us to write equations of
quantum mechanics in forms so general that they apply to a huge array of situations. The time-independent
Schr odinger equation written as

[E) = E [E), for instance, applies equally well to a single particle in
one-dimension, an electron in a hydrogen atom, all the electrons in a uranium atom, or the motion of two
nuclei in a H
2
molecule. This form of the TISE also pertains equally well to position spacein which,
for one particle in one dimension, the wave function is
E
(x)as to momentum space, in which the wave
function is a function of the linear momentum p rather than the spatial variable x. The key point is that
1
Notation: Sometimes, to clarify which of several variables are to be integrated (summed) over, I tack this information onto
the Dirac symbol as a subscript: for example, (t) | (t)
r
is the normalization integral for a state of a spinless particle (s = 0)
in three dimensions with wave function (x, t).
PrintAppendices Version: 8.19 Printed: August 3, 2010
Appendix 66
the symbol [
E
) doe not imply any particular choice of representation (position? momentum? other?), a
choice that is explicit when we symbolize a state by a wave function.
Dirac notation wave-function notation (1D)
quantum state [(t) (x, t)
overlap integral

(x)

(x) dx
matrix element

[

Q [

(x)

(x) dx
k
th
partial derivative

k
x
k
[


k
x
k

(x)
Table G.1. Key elements of Dirac notation. The last line shows the meaning of a partial
derivative of a ket; this symbol appears in the derivations of perturbation theory (Chap. 14). The
examples in the last column are for a spinless particle in one dimension (1D) but generalize to
particles with or without spin in two or three dimensions.
G.2 Dirac notation and Hermiticity of operators
One of the most important properties an operator can have is Hermiticity. In one dimension, for instance, an
operator

Q is Hermitian if, for arbitrary well-behaved wave functions
1
(x, t) and
2
(x, t), the following
equality holds:
2
_

1
(x, t)
_

Q
2
(x, t)
_
dx =
_

Q
1
(x, t)
_

2
(x, t) dx, (G.1a)
where the square brackets emphasize the wave function on which the operator acts. In Dirac notation, we
write Eq. (G.1a) as

1
(t) [

Q
2
(t) =

Q
1
(t) [
2
(t)
1
(t) [

Q [
2
(t). (G.1b)
The rst equality gives meaning to the matrix elements
1
(t) [

Q [
2
(t). The double bar notation
in the matrix element
1
(t) [

Q [
2
(t) signies that

Q can act on either
1
(x, t) (to the left) or on

2
(x, t) (to the right).
Example G.1 (How to express a physical property in Dirac notation: a stationary-state energy.)
If the state [(t)) of a particle conned to one dimension is stationary, then its energy is sharp with some value
E. So in this case, the expectation value of the Hamiltonian, the average energy in this state, doesnt depend
on time and equals E. In wave-function notation, we represent the state by E(x, t) and write the fact as
E) =
_

E
(x)

E
(x) dx = E, (for a stationary state). (G.2a)
In Dirac notation, we can write Eq. (G.2a) more simply and more generally:
E) = (t) [

[ (t)) = [

[ ) = E. (for a stationary state) (G.2b)
Equation (G.2b) pertains to any stationary state of any system.
2
Jargon: We can also express the Hermiticity of an operator

Q, which claries that Hermiticity is a property of an operator,
rather than of states, via the operator equation

Q

=

Q. An operator that is Hermitian is often said to be self-adjoint,
although strictly speaking the two terms are not synonymous. Youll nd a review of the properties of adjoints and a host of
other important aspects of operator algebra in Appendix H.
PrintAppendices Version: 8.19 Printed: August 3, 2010
Appendix 67
G.3 Eigenfunction expansions in Dirac notation
Arguably the most powerful approach to problem-solving in quantum mechanics is the method of eigen-
function expansion. In Dirac notation, the expansion of a state vector [(t) in some complete set [
n

assumes the elegantly simple form
3
[(t) =

n
[
n
c
n
(t), where c
n
(t) =
n
[ (t), (G.1a)
The sum implicitly runs over all eigenvectors in the set.
We can write this expansion even more elegantly as
[(t) =

n
[
n

n
[ (t) eigenfunction expansion of [(t)) in | [
n
) (G.1b)
The latter form reveals the presence in an eigenfunction expansion of the unit operator:

1 =

n
[
n

n
[ completeness of |
n
(G.2)
As the label indicates, Eq. (G.2) is shorthand for mathematical statement that [
n
is complete.
4
Example G.2 (Using completeness in Dirac notation to generate quantum-mechanical equations.)
Equation (G.2) can be used in a nifty, straightforward, and very powerful tactic for expressing equations in
forms we can evaluate from the coecients c
n
(t) =
n
[ (t)) of an eigenfunction expansion. For example, the
normalization condition reads (t) [ (t)) = 1. I can consider the scalar product (t) [ (t)) as the matrix
element of the unit operator

1 with respect to [(t)),
(t) [ (t)) = (t) [

1 [ (t)). (G.3)
Why would I do something so self-evident and apparently useless? Ah, because I can use the completeness
relation (G.2) to replace

1 by a sum of objects [
n
)
n
[. Doing so transforms the normalization requirement
(t) [ (t)) = 1 into a condition I must impose on the expansion coecients:
1 =
_
(t)

n
[n) n[

(t)
_
(G.4a)
=

n
(t) [
n
)
n
[ (t)) (G.4b)
=

n
[ (t))

n
[ (t)) (G.4c)
=

n
[n [ (t))[
2
(G.4d)
=

n
[cn(t)[
2
. (G.4e)
The nal condition,

n
[cn(t)[
2
= 1, is used throughout this book.
So useful is Eq. (G.2) that Dirac dened each object that appears in the summand in Eq. (G.2) as a
projection operator:
3
Details: To keep the notation clear, Im assuming that the set{ | n } includes only normalizable state vectors
(n | n = 1). This assumption precludes the set being eigenfunctions of an operator whose eigenvalues include a contin-
uum.
4
Jargon: Mathematicians call an equation like (G.2) a resolution of the identity.
PrintAppendices Version: 8.19 Printed: August 3, 2010
Appendix 68

P
n
[
n

n
[ , projection operator for [
n
) . (G.5a)
Why projection operator? Well, careful scrutiny of Eq. (G.1b) reveals that

P
n
[(t) = [
n
. (G.5b)
In words, the operator

P
n
projects the n
th
element of the set [
n
out of the state vector [(t). In
terms of the projection operators, the completeness of [
n
, Eq. (G.2), assumes the deeply meaningful,
beautiful mathematical form

1 =

P
n
completeness of | n . (G.5c)
For an alternative use projection, see G.5.
G.4 Useful properties in Dirac notation.
The following properties of scalar products (?? of Chap. 6) are useful in working with Dirac notation:

1
(t) [ c
2
(t) = c
1
(t) [
2
(t) (G.1a)
c
1
(t) [
2
(t) = c

1
(t) [
2
(t). (G.1b)
1(t) + 2(t) [ 3(t) + 4(t)) = 1(t) [ 3(t)) +1(t) [ 4(t)) +2(t) [ 3(t)) +2(t) [ 4(t)). (G.1c)
Also extremely useful are these properties of kets and bras:
([(t))

(t)[ (G.2a)
[(t) = [(t) (G.2b)
(t)[ =

(t)[ (G.2c)
(t) [ (t)

= (t) [ (t) (G.2d)


(t) [ (t)
2
(t) [ (t) (t) [ (t) (G.2e)
_
(t) + (t) [ (t) + (t)
_
(t) [ (t) +
_
(t) [ (t) (G.2f)
G.5 Projecting a state out of the time-independent Schrodinger equation
The chapters of Part IV, which concern approximation methods, use Dirac notation extensively. The major
extension introduced in those chapters is a procedure called projecting a state out of an equation. Consider
the time-independent Schrodinger equation (TISE): in Dirac and (1D) wave function notation, its
_

1 E
_
[
E
= 0
_

1 E
_

E
(x) = 0. (G.1)
Suppose we want to multiply the TISE by the complex conjugate of some other functionsay, (x)and
integrate the result over all space. In wave-function and Dirac notation, wed write this procedure as
_

(x)
_

1 E
_

E
(x) dx = 0. [

1 E [
E
= 0. (G.2a)
For a particle in 3D, the equation in Dirac notation looks the same, while in wave function notation it
changes to indicate integration over three spatial coordinates:
PrintAppendices Version: 8.19 Printed: August 3, 2010
Appendix 69
_

0
_

0
_
2
0

(r)
_

1 E
_

E
(r) r
2
dr sin d d. [

1 E [
E
= 0. (G.2b)
The process of constructing Eqs. (G.2) from the TISE (G.1) is called projecting [ out of the
TISE Eq. (G.1).
5
5
Read on: The word projection is not as arbitrary is may seem. The mathematical meaning of this word is analogous
to its meaning in vector algebra, where we talk about projecting a vector r onto some axis (for instance, x = e
x
r is the
projection of r onto the x axis). For a clear introduction to Diracs formulation of quantum mechanics, see Thaller (2000).
For an authoritative treatment, see Dirac (1958).
PrintAppendices Version: 8.19 Printed: August 3, 2010
Appendix H
The mathematics of operators
Version 8.14: August 3, 2010
H.1 Operators in quantum mechanics
Mathematically speaking, an operator is a rule that transforms one function into another function. Oper-
ators play a crucial role in quantum mechanics because of the following postulate:
Postulate 5 (Operators)
Every observable of a system is represented by a Hermitian operator. The operator is the mathematical tool
we use to extract information about the observable from wave functions.
For a single particle in one dimension, an observable (a physically measurable quantity) hat is represented in
classical physics by the function Q(x, p) is represented in quantum physics by the corresponding operator

Q(x, p), where x and p are the position and momentum operators (see Tbl. H.1).
Three types of operators are especially important in quantum mechanics:
1

=

Q
1

Q is unitary (H.1a)

=

Q

Q is Hermitian (H.1b)

Q (c
1
[
1
+ c
2
[
2
) = c
1

Q[
1
+ c
2

Q[
2


Q is linear (H.1c)
Linearity is of special importance in quantum mechanics because of its connection to the Principle of Super-
position. The inverse of a linear operator

Q is the operator

Q
1
such that

Q
1
=

Q
1

Q =

1, (H.2)
where

1 is the unit operator, which does nothing to whatever it acts upon.
1
Commentary: One operator in quantum mechanics is anti-unitary: the time-reversal operator. While a unitary operator
is linear, an anti-unitary operator is anti-linear: that is,

Q (c
1
|
1
+c
2
|
2
) = c

Q |
1
+c

Q |
2
. Both unitary and
anti-unitary operators preserve scalar products:

Q |

Q = | .
Appendix 71
Unitary operators. This special class of operators is introduced in Chap. 1. By denition, an operator
is unitary if its inverse equals its adjoint (see H.1.1):

Q
1
=

Q


Q =

Q

Q

=

1 a unitary operator (H.3)
The product of two or more unitary operators is itself a unitary operator. The eigenvalues of any unitary
operator are complex numbers of the form e
i
, where is a real number. The absolute value of each such
eigenvalue is 1.
Table H.1. A dictionary of operators for
a one-dimensional system. Each operator
acts, in general, on a wave function (x, t).
Observable operators instructions
position x multiply by x
momentum p i

x
total energy

1

T +

V(x)
kinetic energy

T

2
2m

2
x
2
potential energy

V multiply by V (x, t)
parity

invert x through the origin
H.1.1 The adjoint of an operator and Hermiticity
By denition, the adjoint of an operator

Q is the operator

Q

such that, for any two state functions


1
and
2
,

1
[
2
=
1
[

Q
2
. (H.4)
Table H.2 collects useful properties of operator adjoints.
Table H.2. Basic properties
of operators and their ad-
joints. These operators are
not assumed to be Hermitian.
Commutators and their mathe-
matical properties are discussed
in H.1.2. (In general, the ad-
joint of an operator is not equal
to the operator itself or to its
complex conjugate.)
the operator its adjoint
c
1

Q + c
2

R c

+ c

R

R

Q

_

Q,

R
_

Q

Q

Q

Q

Q

Q
_

Q
n
_

_
n
Not all mathematically well-dened operators can represent physical observables. These restrictions
apply:
(1) All quantum mechanical operators must be Hermitian.
(2) All quantum mechanical operators must be linear (or anti-linear).
A Hermitian operator is self-adjointthat is, the adjoint of a Hermitian operator

Q is the operator
itself:

Q

=

Q. For a Hermitian operator

Q, Eq. (H.4) becomes
2
2
Commentary: Note that Eq. (H.5) is the denition of the symbol
1
|

Q |
2
. Some authors, however, generalize this
notation to non-Hermitian operators in the following way. Using properties of adjoints in Tbl. H.2, one can generate the
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 72

Q
1
[
2
=
1
[

Q
2

1
[

Q [
2
(H.5)
Table H.3 lists combinations of Hermitian operators that are or are not Hermitian.
3
The operator is Hermitian if is not Hermitian if
a constant operator c

1 c is real if c is complex
c
1

Q
1
+ c
2

Q
2
c
1
and c
2
are real c
1
and c
2
are complex

Q
1

Q
2
_
Q
1
, Q
2

= 0
_
Q
1
, Q
2

,= 0

Q
1
n
for integer n is Hermitian
commutator
_

Q
1
,

Q
2

i
_

Q
1
,

Q
2
_

Q
1
,

Q
2

anti-commutator is Hermitian
a function f(x) f(x) is real f(x) is complex
Table H.3. Hermiticity of combinations of operators and functions. In this
table

Q1 and

Q2 are Hermitian operators. Note that a constant operator c

1 is often
written without the unit operator

1, as simply c.
Consequences of Hermiticity. The most important consequences of Hermiticity of an operator concern
solutions of the eigenvalue equation of Hermitian operators (see H.1.5):
(1) The eigenvalues of a Hermitian operator are real.
(2) The eigenfunctions of a Hermitian operator that correspond to dierent eigenvalues are
orthogonal.
4
Two functions are orthogonal if their scalar product is zero. Two bound-state eigen-
functions of a one-dimensional Hamiltonian, for example, satisfy this property:
5

n
[
n
=
_

n
(x)
n
(x) dx = 0, (if n

,= n) (H.6a)
As always, bound-state eigenfunctions must be normalizedthat is, the scalar product of an eigenfunc-
tion with itself must equal unity. A combined statement that the eigenfunctions of

1 are orthogonal
and normalized is the orthonormality condition

n
[
n
=
_

n
(x)
n
(x) dx =
n

,n
, orthonormality (H.6b)
following string of equalities:

1
|

Q

|
2
=

Q
1
|
2
=
2
|

Q
1

=
2
|

Q |
2

.
This equation, which generalizes Eq. (H.4), pertains to any operator, Hermitian or otherwise.
3
Details: Note that, while x is Hermitian, the rst derivative operator d/ dx is not Hermitian. Multiplying this operator by i
renders it Hermitian. Thats why the one-dimensional linear momentum operator is complex, p
x
= i d/ dx. Correspondingly,
the operator e
x
is Hermitian, as is e
i d/ dx
.
4
Details: If a particular eigenvalue is degenerate, which means that two or more (mathematically distinct) eigenfunctions
share that eigenvalue, then these degenerate eigenfunctions may not be orthogonal. But in that case we can always construct
a set of eigenfunctions that are orthogonal. Each eigenfunction in the set corresponds to the same eigenvalue, and the number
of eigenfunctions in the set equals the number of degenerate eigenfunctions (the degree of degeneracy of the eigenvalue). Each
orthogonal eigenfunction is a linear combinations of the functions in the original set. A common procedure for doing so is the
Gram-Schmidt orthogonalization procedure.
5
Jargon: The term scalar product comes from linear algebra. It is also sometimes called the inner product. In physics,
its sometimes called the overlap integral . The scalar product of an eigenfunction with itself is usually called the norm
of the function. The normalization condition amounts to the easily remembered statement that the norm of a normalized
eigenfunction is 1.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 73
where the Kronecker delta function is

,n

_
0 if n

,= n
1 if n

= n
Kronecker delta function (H.6c)
(3) The eigenfunctions of a Hermitian operator constitute a complete set, and all functions
in this set are orthogonal to one another. A set of functions is complete if an arbitrary function
can be expanded in the elements of the set. The function to be expanded must be mathematically
well-behaved. In quantum mechanics, this means that the function must satisfy the usual conditions
of physical admissibility (continuous, smoothly varying, and so forth) and that it must obey the same
boundary conditions as the elements of the set in which it is to be expanded.
6
A complete set of
functions is often called basis (or basis set or eigenbasis) for the expansion. Provided no eigenvalues
of the operator are degenerate, the basis is unique. Degeneracy results in an innite number of mutually
orthogonal bases.
(4) The eigenfunctions of a Hermitian operator satisfy closure. In position space closure is a
property of eigenfunctions at two dierent positions. For the eigenfunctions
n
(x) , this property
reads
7

n
(x

)
n
(x) = (x

x). closure (H.7)


H.1.2 Commutators, compatibility, and simultaneous eigenfunctions
The commutator of operators

Q
1
and

Q
2
is a third operator
_

Q
1
,

Q
2

dened as
_

Q
1
,

Q
2


Q
1

Q
2


Q
2

Q
1
. commutator (H.8)
If

Q
1
and

Q
2
commute, their commutator
_

Q
1
,

Q
2

is zero; if they dont, its non-zero:


_

Q
1
,

Q
2

= 0 commuting operators (H.9a)


_

Q
1
,

Q
2

,= 0. non-commuting operators (H.9b)


For any two operators

Q and

R,
_

Q,

Q commutator of

Q and

R (H.10a)
[

Q,

R]
+


Q

R +

Q

R anti-commutator of

Q and

R. (H.10b)
Rules for manipulating commutators are given in Eq. (H.13). Commutators of particular operators and
functions of operators appear in H.1.4.
Rule: Operators that commute dene a complete set of simultaneous eigenfunctions.
6
Commentary: For example, we couldnt expand a function that went to innity as x in a basis of eigenfunctions of
the simple harmonic oscillator Hamiltonian. In practice, this isnt an issue, because such a function wouldnt be physically
admissible.
7
Commentary: Although Ive singled out this property, its just another way to express completeness.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 74
Compatible and incompatible observables. Two observables Q and R are said to be
compatible if
_

Q,

= 0 (H.11a)
incompatible if
_

Q,

,= 0. (H.11b)
(1) If Q and R are compatible observables, then there exists a complete set of simultaneous eigenfunctions
of

Q and

R. In quantum states that are represented by eigenfunctions in this set, both observables Q
and R are sharpthat is, both have well-dened values with zero uncertainty. Only for such a state is
it possible in principle to measure Q and R precisely.
(2) If Q and R are incompatible observables, then there may exist simultaneous eigenfunctions, if
_

Q,

=

S and if

S is not proportional to the unit operator). But there does not exist a complete
set of simultaneous eigenfunctions. If
_

Q,

is a non-zero constant, there do not exist any simultane-


ous eigenfunctions of

Q and

R.
H.1.3 Operator algebra
Rule: Unlike numbers and functions, operators do not, in general, commute.
8
Keeping this rule in mind, one can perform operator algebra using the following relationships:

Q
1
_

Q
2

Q
3
_
=
_

Q
1

Q
2
_

Q
3
(H.12a)

Q
n

Q
m
=

Q
n+m
(H.12b)

Q
1

Q
2
=

Q
1
_

Q
2

_
. (H.12c)
The last of these equations indicates that the product of operators implies successive operation by one
operator after the other, starting with the right-most operator and proceeding to the left.
Dos and donts of operator manipulation. Here are they key rules for manipulating operators:
(1) Derive and simplify operator expressions and equations by letting all operators act on an arbitrary
physically admissible function of all the relevant variables. For instance, a one-dimensional operator
expressed in terms of x and its derivatives should act on an arbitrary function f(x).
(2) Operators act on everything to their right unless their action is constrained by parentheses or brackets.
(3) The product of two operators is a third operator.
(4) An operator product implies successive operation.
(5) The order in which the operators act is vital; never assume that two operators commute.
8
Commentary: The commutator and commutation rules are discussed in H.1.2. If

Q
1
and

Q
2
are both Hermitian and if
their product

Q
1

Q
2
is Hermitian, then these operators commute. Note that if

Q
1
and

Q
2
are Hermitian, then their commutator
is not Hermitian, because
_

Q,

Q
2
_
=
_

Q,

R
_
.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 75
Coping with commutators. For manipulating commutators, useful equations are
_

Q, c

= 0 for any constant c (H.13a)


_

Q,

Q
n

= 0 (H.13b)
_

Q,

=
_

R,

Q

(H.13c)
_

Q, c

= c
_

Q,

(H.13d)
_

Q +

R,

=
_

Q,

+
_

R,

(H.13e)
_

Q,

R +

=
_

Q,

+
_

Q,

(H.13f)
_

Q

R,

=

Q
_

R,

+
_

Q,

R (H.13g)
_

Q,

=
_

Q,

S +

R
_

Q,

(H.13h)
_

Q

R,

=

Q
_

R,

T +

Q

S
_

R,

+
_

Q,

R +

S
_

Q,

R (H.13i)
_

Q,

f(

Q)

= 0 (H.13j)
_

R,

f(

Q)

= 0 if
_

Q,

=0. (H.13k)
Operators that are functions of operators. The last two of these equations involves a function of
an operator,

f(

Q). Functions of operators are dened in terms of power-series expansions. For a linear
operator

Q,

f(

Q)

n=0
d
n

Q
n
, function of an operator

Q (H.14)
where the expansion coecients d
n
may be complex.
9
The most important example of such a function is
the exponential operator,
e
c

Q
=

n=0
_
c
n
n!
_

Q
n
=

1 + c

Q +
_
c
2
2!
_

Q
2
+ , the exponential operator, (H.15)
for any constant c. From the denition Eq. (H.14), it follows that

f(

Q) commutes with any function of



Q.
One can subdue commutators that contain powers of operators via
10
_

Q,

R
n

=
n1

m=0

R
m
_

Q,


Q
nm1
(H.16a)
_

Q
n
,

=
n1

m=0

Q
nm1
_

Q,

R
m
(H.16b)
Aside. Exponential operators involving non-commuting operators.
If

Q and

R do not commute, then

R does not commute with an arbitrary function of

Q. The most important
instance of this is the product of exponential operators of two non-commuting operators:
e

Q
e

R
,= e

Q+

R
(if
_

Q,

R
_
,= 0).
To cope with such a product of exponential operators, one must use
9
Details: The function

f(

Q) is Hermitian only if

Q is Hermitian and f(Q) is a real function of the observable Q. If

Q is
Hermitian and

f(

Q) is Hermitian, then the coecients d


n
in Eq. (H.14) will be real.
10
Details: In general, we evaluate the adjoint of a function of an operator via

f(

Q)

=

f

).
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 76
e

Q
e

R
= e

Q+

R
e
_

Q,

R
_
/2
.
A useful, if rather scary, extension of this relationship is
e

R e

Q
=

R +
_

Q,

R
_
+
1
2!
_

Q,
_

Q,

R
_ _
+
1
3!
_

Q,
_

Q,
_

Q,

R
_ _ _
+
Finally, a valuable if rather arcane operator relationship is
e


Q
e

R
e


Q
e

R
= exp
_

2
_

Q,

R
_
+O
_

3
_
_
, (H.17)
where the symbol O
_

3
_
indicates terms of order
3
or higher.
H.1.4 Commutation relations involving position and momentum operators
The most important commutation relations in quantum mechanics are those that involve the operators for
the Cartesian components of position and linear momentum. Of all possible combinations of x, y, z and p
x
,
p
y
, and p
z
, the only non-zero commutators are
_
x, p
x

= i p
z
(H.18a)
_
y, p
z

= i p
x
(H.18b)
_
z, p
x

= i p
y
. (H.18c)
From these you can derive any commutator that involves operators whose corresponding classical functions
depend on position, linear momentum, and (perhaps) time. (This class excludes, for example, the parity
operator.)
Commutation relations involving functions of position and momentum operators. Its conve-
nient to denote the Cartesian components of r and p by indices j = 1, 2, 3, as
r = (x, y, z) (r
1
, r
2
, r
3
) (H.19a)
p = (p
x
, p
y
, p
z
) (p
1
, p
2
, p
3
). (H.19b)
With this notation, the following useful commutation relations follow from Eqs. (H.18):
_
r
j
, p
n
j

= i p
n1
j
(H.20a)
_
p
j
, q
n
j

= i q
n1
j
(H.20b)
_
r
j
,

Q

= i

Q
p
j
(H.20c)
_
p
j
,

Q

= i

Q
r
j
, (H.20d)
where

Q =

Q(r, p , t). The partial derivative operators in Eqs. (H.20) are generated by taking the partial
derivative of the corresponding expression for the classical observable, expressed as a function of r and/or p,
then making the usual operator replacements:
x x, and p
x
p
x
= i

x
(H.21)
The following commutators involving position and momentum are especially useful:
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 77
_
x
2
, p
x

= 2i x (H.22a)
_
x, p
2
x

= 2i p (H.22b)
_
x
2
, p
2
x

= 2i (2xp i

1) (H.22c)
_
x
n
, p
x

= i nx
n1
, for m > 1 (H.22d)
For arbitrary operator functions

f(p) and

f(x),
_
x,

f(p)

= i
d

f(p)
dp
(H.22e)
_

f(x), p

= i
d

f(x)
dx
. (H.22f)
Commutation relations involving other operators. For a Hamiltonian

1 =

T +

V,
_

1, x

= i

m
p (H.22g)
_

1, p

= i
d

V
dx

1. (H.22h)
For the Cartesian components of the orbital angular momentum operator,

L
j
where j = 1, 2, and 3 cor-
respond to x, y, and z, respectively, with corresponding notation for the position and linear-momentum
operators, we have
_
r
i
,

L
j
,

= i
i,j,k
r
k
(H.23a)
_
p
i
,

L
j
,

= i
i,j,k
p
k
(H.23b)
Here
i,j,k
is the Levi-Civita symbol dened in Chap. 2. For example,
_
p
x
,

L
z

= i p
y
, and
_
p
y
,

L
z

= i p
x
. (H.24)
H.1.5 Eigenvalues, eigenvectors, and all that
The eigenvalue equation for an operator

Q has the form

Q
q
= q
q
, (H.25)
where
q
is the eigenfunction corresponding to eigenvalue q, which may be complex unless

Q is Hermitian.
In general, an operator

Q may have several mathematically distinct eigenfunctions all of which share a single
eigenvalue. In this case, the eigenvalue q is said to be degenerate with degree of degeneracy equal to
the number of eigenfunctions that share is. An eigenvalue q that is shared by g(q) eigenfunctions is said to
be g(q)fold degenerate. The adjective degenerate is widely applied to the eigenfunctions as well as to
the eigenvalue they share.
11
From the set of eigenvalues of

Q one can easily generate eigenvalues of any function

f(

Q), since

f(

Q)
q
= f(q)
q
, (H.26)
11
Details: Degenerate eigenfunctions must be mathematically distinct. That is, they must be linearly independent. To
uniquely label degenerate eigenfunctions, we must attach labels in addition to the shared eigenvalue q. In the simplest case,
this additional label is simply an index, 1, 2, . . . , g(q), and we write
q,i
. More generally, the labels include quantum numbers
that are not required to uniquely label the eigenvalue q.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 78
where f(Q) is the function whose operator counterpart is

f(

Q). An important instance of this manipulation


is

f(

Q) =

Q
1
, to which we can apply Eq. (H.26) provided

Q
1
exists.
H.1.6 Expectation values, uncertainties, and uncertainty principles
For any quantum mechanical operator

Q that represents an observable Q we dene
12
Q = (t) [

Q [ (t), expectation value of Q in state [(t) (H.27a)
Q =
_
Q
2
Q
2
uncertainty of Q in state [(t), (H.27b)
where, of course,

Q
2
= (t) [

Q

Q [ (t).
The uncertainties for any two observables Q and R are related by the generalized uncertainty prin-
ciple (GUP)
13
QR
1
2

_
_

Q,

, for any state [. (H.28)


This is a generalization of the Heisenberg uncertainty principle, which for a single particle in 3D reads
xp
x

1
2
, and y p
y

1
2
, and y p
z

1
2
. Heisenberg uncertainty principle (H.29)
12
Details: In statistical terms, Q is the average value of the observable Q while Q is its root-mean-square deviation, the
square root of its variance. Sometimes, to emphasize dependence on time t and/or quantum numbers that label eigenvalues
of

Q, I denote these quantities by Q(t) or Q and Q(t) or Q.
13
Notation: This principle is often written with a factor of i multiplying the commutator
_

Q,

R
_
on the rhs. This factor makes
no dierence to the value of the rhs, because we must take the absolute value. Its advantage is that i
_

Q,

R
_
is Hermitiana
technical nicety you need not worry about.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix I
Matrices and determinants
Version 8.14: August 3, 2010
Matrices and determinants are widely used throughout physics in general and quantum physics in particular.
In this book, matrices are prominent in Chaps. 7 and 16, while determinant show up primarily in Chaps. 14
and 16. This appendix collects a small subset of the vast array of tools mathematicians have developed to
manipulate matrices but, apart from a couple of examples, doesnt illustrate, let alone prove, anything. For
such details, see a good book on mathematical methods, linear algebra, or matrices; I recommend several
in the Aside on p. 85.
Fundamental denitions. Table I.2, p. 87 collects some of the technical terms used in matrix algebra.
Denition (dimensions of a matrix) The dimensions of a matrix are the numbers of rows n
r
and
the number of columns n
c
. This information is conventionally denoted n
r
n
c
, which reads a n
r
by n
c
matrix.
Warning: The rst element in n
r
n
c
is the number of rows, and the second elements is the number of
columns. Correspondingly, the element M
ij
belongs to row i and column j.
Denition (adjoint) Formation of the adjoint of a matrix Q requires taking the complex conjugate of
each element Q
ij
, then transposing the resulting matrix.
Denition (transpose) The transpose of a matrix M, which is denoted by M
T
, is constructed by
reecting each element through the diagonal (that is, interchange the rows and column.)
Matrix arithmetic.
( M + N)
ij
= M
ij
+ N
ij
, matrix addition (I.1a)
(M)
ij
= M
ij
multiplication by a complex number (I.1b)
N

i=1
( M N)
ij
=
N

k=1
M
ik
N
kj
, matrix multiplication (I.1c)
Noteworthy features of Eqs. (I.1):
(1) Although matrix multiplication is associativethat is, M ( N P) = ( M N) Pin general, matrix
multiplication is not commutative:
M N ,= N M. (I.2)
But any N N square matrix commutes with the N N unit matrix. And any two N N diagonal
matrices commute.
Appendix 80
(2) The matrix product M N is a meaningful operation only if the number of columns of M equals the
number of rows of N:
M
..
n
r
n
k
N
..
n
k
n
c
= M N
. .
n
r
n
c
, (I.3)
where the dot indicates matrix multiplication.
(3) The n
th
power of a matrix M is the matrix product of M times itself n times:
M
n
= M M M
. .
n factors
(I.4)
Special matrices. Table I.1, p. 86 tabulates some of the most useful matrices one can construct from a
given matrix M.
(1) The zero matrix 0 is the matrix all of whose elements are zero.
(2) The unit matrix 1 is a matrix all of whose o-diagonal elements are zero and all of whose diagonal
elements are 1. Thus the elements of the unit matrix are given by the Kronecker delta: ( 1)
ij
=
i,j
.
For instance, the 2 2 unit matrix is
1 =
_
1 0
0 1
_
, unit matrix (I.5)
In quantum mechanics, the unit matrix appears in two common contexts. The product of an invertible
(square) matrix times its inverse equals the unit matrix:
M
1
M = M M
1
= 1. (I.6a)
(The inverse of a matrix exists only if the matrix is square (of dimensions N N) and has a non-zero
determinant.) Similarly, the product of any Hermitian matrix times its adjoint equals the unit matrix:
Q

Q = Q Q

= 1. (I.6b)
Example I.1 (The inverse of a matrix.)
If the determinant of the matrix
M =
_
M
11
M
12
M21 M22
_
(I.7)
is non-zero, then M can be inverted. The elements of its inverse are
M
1
=
1
det M
_
M
22
M
12
M
21
M
11
_
, (I.8a)
where det M = M12 M21 M11 M22. (I.8b)
Rule: A matrix M has an inverse only if its determinant det M is non-zero. Such a matrix is said to be
non-singular.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 81
I.1 Useful properties
(1) The inverse of an orthogonal matrix is the transpose of the matrix.
(2) The inverse of a unitary matrix is the adjoint of the matrix.
(3) If the matrix is real, its transpose equals its adjoint, so M
T
= M

= M
1
.
(4) Any positive-denite matrix can be written in the form M = N N

for some N.
Useful properties of square matrices.
(1) Any N N square matrix can be written as the sum of a symmetric N N matrix and an anti-
symmetric N N matrix.
(2) The diagonal elements of any antisymmetric square matrix are all zero.
Useful properties of matrix multiplication.
_
M N P)
1
= P
1
N
1
M
1
(I.1a)
_
M N P
_
T
= M
T
N
T
P
T
(I.1b)
Tr ( M N) = Tr ( N M) (I.1c)
det( M N) = (det M)(det N) = det( N M). (I.1d)
Since determinants are numbers, the product (det M)(det N) refers to regular multiplication.
Useful properties of the trace of a matrix.
(1) The trace of a Hermitian matrix is the sum of its eigenvalues.
(2) The trace of the commutator of two matrices equals zero.
(3) The trace of the product M N equals the trace of N M whether or not M and N commute.
(4) The trace of M equals times the trace of M.
(5) The operation of calculating the trace is liner: Tr( M + N) = Tr M + Tr N.
(6) The trace and determinant of M are unchanged by a similarity transformation of M.
(7) The trace and determinant of any positive-denite matrix are non-negative.
I.2 Eigenvalues, eigenvectors, and matrix diagonalization
Matrix eigensystems. The eigenvalues q
i
of a square matrix Q are solutions of the matrix eigenvalue
equation
1
QQ
i
= q
i
Q
i
, for i = 1, 2, . . . , N, (I.1)
where the N 1 column matrix Q
i
is the i
th
eigenvector of Q. Each eigenvalue q
i
is a number. If Q is
Hermitian, all its eigenvalues are real; otherwise, one or more eigenvalues may be complex. The collection
of eigenvalues and eigenvectors of a matrix is its eigensystem.
One can determine the eigenvalues of a matrix using computer-algebra software such as Mathematica
or by solving the secular equation (or characteristic equation) of the matrix. The secular equation of
an N N matrix is an n
th
-order polynomial equation obtained by expanding the determinant of Q 1.
1
Jargon: The collection of eigenvalues of a matrix is sometimes called its spectrum. This term is more often applied to the
collection of eigenvalues of an operator. But since the eigenvalues of an operator are often found by diagonalizing its matrix
representation, this is a dierence that makes no dierence. Note, however, that this use of spectrum is purely mathematical
and is unrelated to transitions between stationary states, which generate a spectrum in physics.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 82
For a 2 2 matrix, for example, the characteristic equation is
2

Q
11
q Q
12
Q
21
Q
22
1

= 0 = q
2
+ c
1
q + c
0
= 0 (I.2)
This equation has two roots, the eigenvalues q
1
and q
2
. Plugging each root back into the equation gives
a value for the ratio of the coecients c
1
/c
0
for that root. To ascribe values of these coecients, one
must impose another condition; in quantum mechanics, the normalization condition [c
1
[
2
+[c
0
[
2
= 1. For
an N N matrix, solution of the secular equation yields ratios of N 1 coecients to c
0
. Normalization
then determines the values of all N coecients.
3
Denition (matrix diagonalization) The process of matrix diagonalization of M entails determing
a matrix N that transforms M (via a similarity transformation) into a diagonal matrix. The result of matrix
diagonalization of M is a diagonal matrix whose diagonal elements are the eigenvalues of M.
4
The N N matrix whose i
th
column is the i
th
eigenvector Q
i
of Q is the matrix N that diagonalize Q.
If the eigenvectors Q
i
are orthonormal, then N is unitary. Since the eigenvectors of Q are linearly inde-
pendent, N has an inverse. We can therefore generate a diagonal matrix whose elements are the eigenvalues
of Q as the product of the inverse of N times Q times N. For example, for a 2 2 matrix, we have
_
q
1
0
0 q
2
_
= N
1
Q N. (I.3)
A product of this form is called a similarity transformation on Q, and Q and M are said to be
similaryet another word whose meaning here is much more precise than in colloquial chat. Note that
5
P = N
1
Q N = Q = N P N
1
matrices Q and
N are similar
(I.4)
2
Details: In this simple case, the secular polynomial is a quadratic. In general, this polynomial has the form
q
n
+c
n1
q
n2
+ +c
1
q +c
0
. Any n
th
-order polynomial may have complex roots, and any such polynomial has at least
one root.
3
A cautionary note: If you use a program like Mathematica to determine eigensystems of matrices, be very careful about two
things. First, some software automatically normalizes its eigenvectors; other software doesnt. Some in the latter category oer
an option to make the command give normalized eigenvectors; other require you to program normalization yourself. Second,
some software, given a matrix to diagonalize, produces a list of eigenvectors in reverse order to the list of eigenvalues. Your
safest bet (by far) is to check that each eigenvector-eigenvalue pair satises the original matrix equation. Its easy to make a
computer do this chore, which would be odious do perform by hand.
4
Details: Numerical analysts have developed a variety of methods to diagonalize a matrix. Many of these methods are
extremely fast but pertain only to special types of matrices. The most widely used methods for simple quantum-mechanical
calculations in which the matrix to be diagonalized is Hermitian are the Jacobi method and the (much faster) Givens-
Householder-Wilkinson method. For details, see any good text on numerical analysis.
5
Jargon: If N is an orthogonal matrix, then this similarity transformation is called an orthogonal transformation. Orthog-
onal transformations play an important role in the mechanics of rotation. Similarly, if N is a unitary matrix, then this similarly
transformation is called a unitary transformation. All symmetry transformations (Chap. 1) are unitary transformation, as
are many other transformations in quantum mechanics.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 83
Degenerate eigenvalues. One or more eigenvalue of a matrix Q may be degenerate: if, for example,
q
i
= q
j
for i ,= j, then the number q
i
is said to be two-fold degenerate, indicating that there are two
eigenvectors, Q
i
and Q
j
, that share this eigenvalue.
Useful properties of eigensystems.
(1) The eigenvalues of a square Hermitian matrix are real. If all its eigenvalues are non-degenerate, then its
eigenvectors are linearly independent and orthogonal. If one or more eigenvalues are degenerate, then
one can construct from these degenerate eigenvectors an orthonormal set by taking linear combination
of them.
(2) The eigenvectors of a Hermitian matrix that correspond to dierent eigenvalues are orthogonal. Eigen-
vectors that share a (degenerate) eigenvalue are not necessarily orthogonal, but one can construct from
them an orthogonal set using methods such as the Gram-Schmidt procedure (see sources in the Aside
on p. 85.)
(3) The absolute value of each eigenvalue of a unitary matrix equals 1.
(4) The trace of a matrix equals the sum of its eigenvalues.
Tr M =
N

i=1
M
ii
(I.5)
(5) The determinant of a matrix equals the product of its eigenvalues.
det M =
N

i=1
M
ii
(I.6)
(6) If the eigenvalue q
i
is k-fold degenerate, then any linear combination of the k eigenvectors Q
i
that
share this eigenvalue is another eigenvector with this eigenvalue. If Q is Hermitian, then one can
construct a set of k linearly independent eigenvectors with eigenvalue q
i
.
6
(7) One can simultaneously diagonalize two matrices, M and N, only if the matrices commute:
M N = N M.
Rule:
(1) All eigenvalues of a Hermitian matrix are real.
(2) The magnitude of each eigenvalue of a unitary matrix equals 1.
6
A cautionary note: Such a construction is not possible for an arbitrary matrix M. All one can say in this case is that any
linear combination of degenerate eigenvectors is another eigenvector with the same eigenvalue.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 84
Aside. Linear dependence and the grammar of the Gramian
Just because your theory predicts that a set of eigenvectors should be linearly independent doesnt mean
that your computer will produce a set with this property. All computer calculations are limited by myriad
numerical errors many of which result from the nite word-length of a computers representation of a real
(or complex) number. Fortunately, you can make your computer calculate a determinant that will tell you
the extent to which a set of eigenvectors if linearly dependent and hence awed. This useful little guy is
called the Gram determinant or, more quaintly, the Gramian.
Denition The Gramian of a set of N vectors | Q
i
is the N
th
-order determinant whose elements are all
the scalar products one can construct from this set.
7
Each scalar product has the form Qi [ Qj) = Q
i

Q
j
with elements
Q
i
[ Q
j
)
lm
=
N

k=1
( Q
i
)

lk
( Q
j
)
km
, (for l, m = 1, 2, . . . , N) (I.7)
The Gramian of the set of eigenvectors | Q
1
, Q
2
, Q
3
of a 3 3 matrix Q is the determinant
( =

Q1 [ Q1) Q1 [ Q2) Q1 [ Q3)


Q
2
[ Q
1
) Q
2
[ Q
2
) Q
2
[ Q
3
)
Q
3
[ Q
1
) Q
3
[ Q
2
) Q
3
[ Q
3
)

, Gramian of a 3 3 matrix. (I.8)


Like any determinant, the Gramian is a number. Because the matrix of elements that denes the Gramian
is positive denite, the Gramian itself is a non-negative number. Heres what it tells you:
Rule: If the Gramian of a set of eigenvectors is non-zero, then the set is linearly independent. If the Gramian
is zero, then the set is linearly dependent.
8
I.3 Determinants
Denition (determinant) The determinant of an N N matrix M is a number evaluated from the
array of its elements. The rank N of M is the order of its determinant.
For notational reasons only, this array is surrounded by vertical lines. The determinant of a 3 3 matrix,
for example, is written
det M =

M
11
M
12
M
13
M
21
M
22
M
23
M
31
M
32
M
33

, a third-order determinant. (I.1)


We calculate the value of a second-order determinant as
det M =

M
11
M
12
M
21
M
22

= M
11
M
22
M
12
M
21
. (I.2)
Evaluating higher-order determinants is best done by using computer software. Failing that, one can expand
the determinant in minors.
Denition (cofactor and minor) The cofactor of the element M
ij
of an N N matrix M is the prod-
uct of a phase factor (1)
i+j
times the (N 1) (N 1) determinant obtained by deleting row i and col-
umn j from M. The resulting (N 1) (N 1) matrix is the minor of M
ij
. Each cofactor, therefore, is
a determinant of order N.
7
A cautionary note: You must multiply each eigenvector by its adjoint, not the other way around. Each eigenvector is an
N 1 column matrix and so must be multiplied on the left by an 1 N row vector for the matrix product to have any meaning.
8
Commentary: This rule promises an elegant, clean division into linearly dependent and linearly independent sets of vectors
your computer cant deliver. Almost invariably, a computer told to evaluate the Gramian of a set of properly calculated vectors
that should be linearly independent returns a number that isnt zero but is very small. This result means that strictly speaking,
your eigenvectors are numerically linearly dependent. Whether this purely numerical defect has serious consequences and must
be rectied by a calculation with higher precision depends on you needs in a particular calculation, the power of your particular
computer, and your patience.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 85
Example I.1. The determinant of a 3 3 matrix.
The determinant (I.1) of the 3 3 matrix i s
det M = M
11

M
22
M
23
M32 M33

M
12

M
21
M
23
M31 M33

+M
13

M
21
M
22
M31 M32

(I.3)
Each minor can now be expanded using Eq. (I.2). For examples of the use of minors to expand determinants of
order higher than 3, see a good book on matrix mathematics (see the Aside on p. 85).
Warning: Dont forget the phase factor (1)
i+j
that must multiply each minor. Thats why there is a
minus sign in the second term of Eq. (I.3).
Determinants satisfy several properties that play a vital role in the quantum-mechanics of many-electron
systems (Part III). You should memorize these:
Rule:
(1) If any two rows of a determinant are identical, then the determinant is zero. The same holds for a determinant
with any two identical columns:
identical rows: M
ij
= M
ik
for all i = 1, 2,. . . ,N and for j ,= k = det M = 0,
identical columns: M
ij
= M
kj
for all j = 1, 2,. . . ,N and for i ,= k = det M = 0.
(I.4)
(2) If every element of one (or more) rows or one (or more) columns of a determinant is zero, then the
determinant is zero:
one zero row: M
ij
= 0 for all j = 1, 2, . . . , N for any i = det M = 0,
one zero column: M
ij
= 0 for all i = 1, 2, . . . , N for any j = det M = 0.
(I.5)
(3) The determinant of the matrix M is det M.
(4) Interchanging any two rows (or any two columns) of a determinant det M results in det M.
(5) If we add to each element of one row the product of a constant times the corresponding element of another
row, the value of the determinant is unchanged:

M
11
+ M
21
M
12
+ M
22
M
21
M
22

M
11
M
12
M
21
M
22

(I.6)
The same result holds if we add the multiple of one column to another column.
Summary: useful properties of determinants.
det( M

) = (det M)

, determinant of the complex conjugate of M (I.7a)


det( M
T
) = det M, determinant of the transpose of M (I.7b)
det( M

) = (det M)

, determinant of the adjoint of M (I.7c)


det M = e
Tr ln M
(I.7d)
Aside. Would you like to know more?
The most comprehensive source for information about matrix mathematics is the encyclopedic reference
by Bernstein (2005). For a pedagogical treatment, see Chap. 1 of Hildebrand (1965). From there, I suggest
you proceed to Secs. 2.1 and 2.2 of Cushing (1975). Short book-length studies that concentrate on numerical
aspects of matrix mathematics are those by Jennings (1977), Forsythe and Moler (1967), and Sewell (2005).
Exceptionally useful treatments of numerical computation involving matrices appear in Chap. 5 of Wong
(1997), Sec. 5.6 of Kincaid and Cheney (2002), Chap. 5 of Allen and Isaacson (1998) (which is especially good
on eigenvalue problems), and Chap. 10 of Antia (2002). Chapter 2 of this book provides useful information
about numerical error, which should always be a concern in matrix (and other) calculations. At a more
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 86
equal M = N M
ij
= N
ij
for all i and j
scalar diagonal matrix with equal elements only
Hermitian Q = Q

Q
ij
= Q

ji
anti-Hermitian M = M

M
ij
= M

ji
real M = M

M
ij
= M

ij
imaginary M = M

M
ij
= i N
ij
, where N
ij
is real
symmetric M = M
T
M
ij
= M
ji
antisymmetric M = M
T
M
ij
= M
ji
, and M
ii
= 0
unitary M =
_
M

_
1
( M

M)
ij
=
i,j
orthogonal M = M
1
( M M)
ij
=
i,j
idempotent M
2
= M
nilpotent M
n
= 0 for some positive integer n
upper triangular M
ij
= 0 for all i > j
lower triangular M
ij
= 0 for all i < j
singular det M = 0 a singular matrix has no inverse
non-singular det M ,= 0 a non-singular matrix can be inverted
diagonal M
ij
= 0, for all i ,= j
tridiagonal M
ij
= 0 for all [i j[ > 1
banded M
ij
= 0 for all [i j[ > n for positive n
o-diagonal at least one M
ij
is not-zero
block diagonal all non-zero elements occur in blocks along the diagonal
positive denite all eigenvalues are real and non-negative
Table I.1. Types of matrices. The symbol M denotes an arbitrary matrix; Q denotes a Hermitian matrixa
matrix that equals its adjoint. The symbol det M denotes the determinant of M. Each block in a block-diagonal
matrix is itself a square matrix of rank less than that of the matrix itself. An antisymmetric matrix is sometimes
called a skew-symmetric matrix.
advanced level, Chap. 11 of Press et al. (1992) and its update Press et al. (1996) oer good advice. One
of the best books for learning how to apply computers to physics is ?, Chap. 8 of which focuses on matrix
computing. Finally, no one should tackle problems in physics that require extensive computation without
reading carefully two books by Acton (1970) and Acton (1996). Packed with advice born of extensive
experience, these books also boast more jokes per page than any other numerical analysis text Ive ever seen.
PrintAppendices Version: 8.14 Printed: August 3, 2010
notation denition
matrix M M
ij
transpose M
T
_
M
T_
ij
= M
ji
complex conjugate M

_
M

_
ij
= (M
ij
)

adjoint M

_
M
_
ij
= (M
ji
)

inverse M
1
_
M
1_
ij
=
cof Mij
det M
trace Tr M

N
i=1
M
ii
Table I.2. Terminology for matrix algebra. In the denition of the inverse of a
matrix, cof M
ij
is the cofactor of the element M
ij
. In the denition of the trace, N is the
dimension of the matrix. Note that some sources denote the transpose of a matrix M
by

M. The adjoint M

of a matrix M is sometimes called its Hermitian conjugate.


The trace of M is sometimes called its spur.
Appendix J
Making sense of spectral data
Version 8.16: August 3, 2010
For those who want some proof that physicists are human,
the proof is in the idiocy of all the dierent units
which they use for measuring energy.
Richard P. Feynamn, in The Character of Physical Law
In principle, dealing with data from spectroscopy should be among the easiest of chores. You need only
a few constants (Tbl. J.1) and a handful of equations from quantum mechanics, thermodynamics, and the
theory of waves:
1
E = h = , de-Broglie/Einstein relations, (J.1a)
= 2 = c k, frequency ; wave number k, (J.1b)
=
1
T
=
c

period T ; wavelength , (J.1c)


E = k
B
T temperature T. (J.1d)
Note that T denotes period and T denotes temperature. The fundamental constants in these equations
are the speed of light c, Plancks constant h, the reduced Plancks constant h/2, and Boltzmanns
constant k
B
.
What makes coping with atomic spectra hazardous is the bewildering variety of units in which this data
is reported and a few rarely stated conventions one must somehow know. You can nd tables of conversion
factors in Appendix D and constants galore in Appendix C. This appendix brings together the small subset
of this information you need to cope with spectral data and, more important, shows you how to handle those
units and conventions. Heres the bottom line:
Warning: When you see a quantity from spectroscopy called an energy, dont assume that its units are
actually those of energy.
Conversion of an energy into an energy follows from a rather eccentric application of Eqs. (J.1):
E = h =
hc

=
1

=
E
hc
energy in units of inverse length ( cm
1
) (J.2a)
E = h =
h
T
= =
1
T
=
E
h
energy in units of inverse time ( s
1
=Hz) (J.2b)
E = k
B
T = T =
E
k
B
energy in units of temperature (kelvin). (J.2c)
1
Read on: You can nd information about these equations and the quantities they contain in Appendix O.
Appendix 89
quantity value dimensions
h 6.626 068 96 10
34
J s [h] = E T
h
e
4.135 667 33 10
15
eVs
_
h
e

= E T
=
h
2
1.054 571 628 10
34
J s [] = E T

e
6.582 118 99 10
16
eVs
_

= E T
hc 1.986 4455 10
23
cm J [hc] = E L
hc
e
1.2398 419 10
4
cm eV
_
hc
e

= E L
1
hc
8065.5489 eV
1
cm
1
_
1
hc

= E
1
L
1
R

10 973 731.568 525 m


1
[R

] = L
1
R

hc 2.179 872 10
18
J [R

hc] = E
R

hc 13.605 691 93 eV [R

hc] = E
R

c 3.289 842 10
15
Hz [R

c] = T
E
h
4.3597 439 10
18
J [ E
h
] = E
E
h
e
27.211 383 86 eV
_
E
h
e

= E
Table J.1. Constants you need to cope with spectra. By convention, the unit Hz s
1
(Hertz) is used for frequency , not angular frequency = 2. Also by convention,
spectroscopists report the wave number 1/ (which they often call the energy) in
cm
1
, not in SI units ( m
1
). So you need to use 1 m = 100 cm for conversions. August 3,
2010
What a spectroscopist means by a wave number. To understand energies in spectroscopy, you
rst need to know that spectroscopists do not use the term wave number to mean what it usually means.
Strictly, the wave number k is 2 times the number of cycles per unit distance.
2
So the wave number is
related to the quantities in Eq. (J.1) by
3
k
2

=

c
=
2
c
, wave number, (J.3)
The dimensions of wave number are [k] = L
1
. But when spectroscopists talk about a wave number, they
mean k/2 = 1/, which they usually denote by :

1

=
k
2
=

c
, wave number (in spectroscopy). (J.4)
The overbar on is intended to remind you of two things: (1) is the actual wave number divided by 2,
and (2) is the frequency divided by the speed of light c.
Warning: Wave numbers are not reported in the SI unit of inverse length, m
1
, but rather in the
CGS unit, cm
1
= 100 m
1
.
2
Details: The propagation speed of an EM wave in a vacuum is the speed of light c. The wave number k is the magnitude
of the wave vector k = k

k, which points along the propagation direction



k of the wave.
3
Details: Widely used units of wavelength include 1

A = 10
10
m = 10
8
cm = 10
4
, where is the micron, short for
micrometer: 1 = 10
6
m.
PrintAppendices Version: 8.16 Printed: August 3, 2010
Appendix 90
When an energy isnt really an energy: units of inverse length. Many spectroscopists refer
to as an energy (although they omit the quotation marks). Of course, isnt an energy; it doesnt even
have the right dimensions. But you can easily transform into an energy using Eqs. (J.2). For instance,
heres how to generate (the wave number) from the energy in SI units (J) or in electron volts (eV):
E( eV) =
hc
e
( cm
1
) =
_
1.2398 419 10
4
cm eV
_
( cm
1
)
( cm
1
) =
e
hc
E( eV) =
_
8065.5489 eV
1
cm
1
_
E(eV )
(J.5)
Thus the statement the energy of the n = 2 state of atomic hydrogen is 2.743 10
4
cm
1
really means
that the energy is what comes out of Eq. (J.5), 3.401 eV. The crucial conversion factor is the product of
Plancks constant times the speed of light; you can nd values in Tbl. J.1; dont forget the factor of 100 to
take care of that pesky cm
1
convention.
When an energy isnt really an energy: units of inverse time. For many years, spectroscopists
have also reported energies in Hertz, which is a unit of inverse time: Hz s
1
, a unit of frequency.
4
Because
frequencies for atomic and molecular transitions are typically quite small, spectroscopists usually attach a
prex to the symbol Hz, as k for kilo (10
3
), M for mega (10
6
), and G for giga (10
12
) (see Appendix L). Using
the information in Tbl. J.1 and Eqs. (J.1), you can derive a host of conversion equations. For instance,
Especially in reporting very small energies for hyperne transitions and for transitions between the same
states in dierent isotopes, physicists usually report energies in MHz:
5
1 MHz 10
9
Hz 3.335 641 10
5
cm
1
(J.6a)
1 MHz 4.1356673 10
9
eV. (J.6b)
When an energy isnt really an energy: units of temperature. During the last decades of the
20
th
century, as technology enabled physicists to explore atomic and molecular processes at lower and lower
temperatures, many began to report energies in units of (get ready for it) Kelvin. In this case the conversion
factor is Boltzmanns constant:
k
B
RN
A
= 1.380 6504 10
23
J/ K,
=
k
B
e
= 8.617 342 8 10
5
eV/ K.
Boltzman constant. (J.7)
where R = 8.314 472 J mol
1
K
1
is the molar gas constant, e = 1.602 176 487 10
19
C is the elementary
charge and N
A
= 6.022 141 79 mol
1
is Avogadros constant.
4
Details: Thats frequency , not angular frequency = 2. The angular frequency characterizes the number of radians
per unit time, as opposed to the frequency , which characterizes the number of cycles per unit time.
5
Commentary: While the use of Hertz for frequency rather than angular frequency is conventional, this convention
is not uniformly followed. (By now this probably comes as no surprise to you.) Some authors use the word frequency and
report numbers in Hertz but actually mean , not . I wish I could tell you a foolproof way to tell what an authors means if
he or she doesnt explicitly say, but I dont know one. (spectroscopic data is rarely reported as wavelengths, because a measured
wavelength depends on the medium through which the wave propagates, while a measured frequency does not.)
PrintAppendices Version: 8.16 Printed: August 3, 2010
Appendix 91
Conversions are really correspondences. So varied a menu of energy units often results in weird-
looking conversion factors.
6
To indicate that an conversion is not a true equality, throughout this book I use
the symbol , which should be read corresponds to. For example, here are three correspondences I
use in my research:
1 cm
1
1 10
2
m
1
29.979 2458 GHz 30 10
9
Hz,
1 eV 8065.5489 cm
1
8065 cm
1
1 mK 10
3
K 20.836644 MHz 21 10
6
Hz
(J.8)
Transitions, the Rydberg constant, and the Rydberg. The primary application of all this stu is
to transitions. The fundamental equation for the wavelength of a transition (Chap. 18) is the Balmer
equation (see Chap. 5). Expressed in terms of the wave number and written for an emission from a
hydrogenic energy level E
n
to a lower-lying level E
n
, this equation reads

,n
= R

_
1
n
2

1
n
2
_
,
Balmer equation for a
transition wave number.
(J.9a)
Such equations introduce the Rydberg constant
R


m
e
e
4
0
4
3
c
Rydberg constant: approximation
of innite nuclear mass: [R

] = L
1
(J.9b)
Crucially, the dimensions of the Rydberg constant are inverse length, not energy. The actual energy of the
transition in Eq. (J.9a) is
E
n

,n
E
n
E
n
= hc R

_
1
n
2

1
n
2
_
(J.9c)
There it is again: the factor hc that converts from dimensions of inverse length to dimensions of energy. The
resulting prefactor, hcR

acquired a namethe Rydbergand a symbol, Ry:


Ry hc R

=
1
2
E
h
, the Rydberg unit of energy. (J.10a)
The second equality in Eq. (J.10a) is quite important, because it relates the Rydberg to the atomic unit
of energy, the Hartree (see Appendix F):
E
h

m
e
e
4
0

2
= 2hcR

the Hartree, the atomic unit of energy. (J.10b)


Warning: The Rydberg, Eq. (J.10a) is not the atomic unit of energy; the value of Ry is half the value
of the atomic unit of energy. Thus Ry( eV) = 13.605 691 93 eV, while E
h
( eV) = 27.211 383 86 eV.
7
6
Commentary: The confusion doesnt stop with energy: I recently read a book that gave, for the frequency of an electron
in the rst Bohr orbit, a value in dimensions of L
1
: 2.19 10
5
cm
1
.
7
A cautionary note: Boundless confusion pervades some of the literature of physics because authors state that their reported
energies are in atomic units but dont state that they mean Rydberg, not Hartree. Unless an author explicitly says Rydberg,
you should assume that the author means Hartreebut be wary and stay alert: watch for errant factors of 2 between what
you get and what you expect; such factors may indicate that the author really means Rydberg.
PrintAppendices Version: 8.16 Printed: August 3, 2010
Appendix 92
The Rydberg constant for atoms. As my label on Eq. (J.9b) indicates, R

implicitly invokes the


approximation of innite nuclear mass, as you can tell from the presence of m
e
rather than the reduced mass
(see Chap. 5). Fortunately, its easy to calculate the Rydberg constant for any atom you want:
R
Z
= R

Z
m
e
_
,
Rydberg constant for an
atom of atomic number Z.
(J.11)
The multiplicative factor
Z
/m
e
just replaces m
e
in Eq. (J.9b) with the reduced mass of your atom. If the
mass of the nucleus is m
Z
, then its reduced mass is

Z

m
e
m
Z
m
e
+ m
Z
. (J.12)
Atomic units and atomic mass units. Typically, atomic and molecular weights are tabulated in
the imposingly named unied atomic mass unit, which is more modestly denoted by u (but used to
be denoted by amu). The value of this quantity, which equals 1/12 the mass of a
12
C atom, provides the
conversion factor you need to get your masses into SI units:
1 u
1
12
m(
12
C) = 1.6605 388 10
27
kg an atomic mass unit (J.13a)
Since we work so often in atomic units (Appendix F), its important to have the factor that converts mass
in atomic mass units to mass in atomic units (the atomic unit of mass is the mass of the electron):
m( atomic units) = 1822.8885 m( u) (J.13b)
Final warnings.
(1) The factor that converts an energy given in cm
1
(that is, a wave number ) into an actual energy
is the product of the speed of light c times Plancks constant h = 6.6261 10
34
J s not the so-called
rationalized Plancks constant h/2. The correct factor to convert from cm
1
to eV is there-
fore hc = 1/8065.5489 cm eV = 1.2398 412 10
4
cm eV. Use this factor to convert all spectroscopic
constants to energy units.
(2) Dont forget to convert the speed of light c from SI units (ms
1
) to (cms
1
). That is
c 2.998 10
10
cm s
1
. (Note the exponent.)
PrintAppendices Version: 8.16 Printed: August 3, 2010
Useful relationships
= c (in vacuum)

k
= c (in vacuum)
E = h =
hc

= hv E = = ck
Useful conversion factors
hc =
1
8065.5489
eV cm = 1.2398 412 10
4
eV cm c = 1973.2686 eV

A
E(eV ) =
12 398.419
(

A)
E( E
h
) =
455.633 53
(

A)
Table J.2. Useful information about waves for analysis of atomic spectra. The equations relating
frequency and wavelength, and angular frequency and wave number refer to the vacuum. The conventional unit
of frequency is the hertz, Hz. Other frequently used frequency units are 1 MHz = 10
6
s
1
and 1 GHz = 10
9
s
1
.
The atomic unit of frequency is 4.1341410
16
Hz, the inverse of the atomic unit of time, 2.4189 10
17
s.
To convert a wave number in cm
1
to an energy in eV, multiply by the value of c in eV cm in this
table: 1 cm
1
= 1.2398 419 10
4
eV. In the conversion equations in the last row, is the vacuum wavelength
in Angstrom. For additional conversion factors, see Appendix C.
Appendix K
Angular-momentum coupling and Clebsch-Gordan
coecients
Version 8.14: August 3, 2010
This appendix concerns a construction project. Your have two quantum-mechanical angular momenta. You
want to construct a third angular momentum: their sum. This chore arises in lots of applications of quantum
mechanics. A prominent example in this book is the spin-orbit interaction discussed in Chaps. 5 and 14: in
that case, the angular momenta are the orbital angular momentum

L and spin angular momenta

S of one
electron. Another example is the treatment of any many-electron system, from the helium atom to the most
complicated molecule you can imagine. In setting up the quantum mechanics of a two-electron atom such
as helium (He: Z = 2) [Chap. 9], one must cope with four angular momenta: the orbital and spin angular
momenta of each electron. In a rigorous treatment of a two-electron atom, all four of these angular momenta
experience pairwise interactions, so one must sum all four. Because, in the quantum-mechanics of angular
momenta, the word sum has implications beyond its literal meaning in mathematics, the word coupling,
which suggests physical interactions, is preferredand this topic is usually called angular-momentum
coupling.
Addition of more than two angular momenta is a topic for a more advanced book. Indeed, the importance
of this topic in quantum mechanics has spawned a small library on the quantum theory of angular momentum
(the Aside on p. 100). My more modest goal is to summarize (not derive) key results you need to cope with
two angular momenta only.
First I must explain what I mean by the sum of two angular momenta. In classical mechanics, this
matter never arises. The sum of the orbital angular momenta of, say, two particles is the sum of two vectors
in ordinary space: J = J
1
+J
2
. In quantum mechanics, however, we must add vector operators

J
1
and

J
2
.
At rst glance, this construction seems straightforward: we just add their Cartesian components:

J
x
=

J
x1
+

J
x2
,

J
y
=

J
y1
+

J
y2
, and

J
z
=

J
z1
+

J
z2
. (K.0.1)
These equations are correct, and the third one,

J
z
=

J
z1
+

J
z2
, plays a key role in what is to come.
One reason addition of angular momenta in quantum mechanics diers so much from its classical counter-
part is that the sum

J

J
1
+

J
2
purports to represent an observable . . . not just any observable, a quantum
mechanical angular momentum. So we must not only verify Hermiticity of

J, we must also verify that
its Cartesian components obey the commutation relations required of any angular momentum. Moreover,
if we are to use

J for anything, we need simultaneous eigenvectors of

J
2
and

J
z
. How do we develop these
eigenvectors? The corresponding eigenvalues, in turn, introduce new quantum numbers, the total angular
momentum quantum number J and its projection quantum number, which Ill denote by M. What
are the allowed values of these quantum numbers? How are they related to the quantum numbers of the
constituent angular momenta

J
1
and

J
2
? No such questions never arise in classical mechanics.
Appendix 95
K.1 The total orbital angular momentum operator
We start with two quantum-mechanical angular momenta,

J
1
and

J
2
. Taking

J
1
as an example, each
of these two is Hermitian,

J
1

J
1
, and each and has Cartesian components

J
1
= e
x

J
x1
+e
y

J
y1
+e
z

J
z1
that satisfy the requisite commutation relations
_

J
1x
,

J
1y

= i

J
1z
,
_

J
1y
,

J
1z

= i

J
1x
,
_

J
1z
,

J
1x

= i

J
1y
, (K.1.1a)
and
_

J
2
1
,

J
1x
= 0

,
_

J
2
1
,

J
1y
= 0

,
_

J
2
1
,

J
1z
= 0

, (K.1.1b)
where

J
2
1
=

J
2
x1
+

J
2
y1
+

J
2
z1
. (K.1.1c)
Each individual angular momentum denes the usual simultaneous eigenvalue equations

J
2
1
[j
1
, m
1
= j
1
(j
1
+ 1)
2
[j
1
, m
1
(K.1.2a)

J
z1
[j
1
, m
1
= m
1
[j
1
, m
1
(K.1.2b)
for values of m
1
in integral steps in the range j
1
m
1
j
1
. Hermiticity of

J
1
guarantees the orthonor-
mality relations
j
1
, m
1
[ j
2
, m
2
=
j
1
,j
2

m
1
,m
2
, (K.1.3a)
j
1

, m
1

[ j
1
, m
1
=
j
1

,j
1

m
1

,m
1
, and j
2

, m
2

[ j
2
, m
2
=
j
2

,j
2

m
2

,m
2
. (K.1.3b)
The total angular momentum operator is dened as

J =

J
1
+

J
2
. (K.1.4)
The operator for its square is
1

J
2
=

J
2
1
+

J
2
2
+ 2

J
1

J
2
, (K.1.5a)
where

J
1

J
2
=

J
x1

J
x2
+

J
y1

J
y2
+

J
z1

J
z2
. (K.1.5b)
The term (K.1.5b), the last term in Eq. (K.1.5a), is the troublemaker: it forces us to form linear combinations
to get eigenfunctions of

J
2
, as youll see in the next section.
K.2 Commutation relations
Unlike the statement this is a fruitcake, which could mean all sorts of more-or-less scary things, the
statement this is a quantum-mechanical angular momentum means something very precise. As detailed
in Chaps. 3, 6, and 7, in quantum mechanics, an angular momentum

J by denition satises commutation
relations involving its Cartesian components.
From these dening relations there follows the all-important commutativity of the square and any com-
ponent of

J. This, in turn, implies the existence of a complete set of simultaneous eigenvectors of

J
2
and,
say,

J
z
.
Suitable (tedious) application of operator algebra shows that

J is Hermitian (hooray) and that its Carte-
sian components satisfy the commutation relations
_

J
x
,

J
y

= i

J
z
,
_

J
y
,

J
z

= i

J
x
,
_

J
z
,

J
x

= i

J
y
, (K.2.1a)
1
A cautionary note: While the sum of

J
1
and

J
2
satises the commutation relations that admit it to the brotherhood
of quantum-mechanical angular momentum operators, these commutation relations are not necessarily satised by arbitrary
combinations of

J
1
and

J
2
. Each such combination must be checked.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 96
and
_

J
2
,

J
x
= 0

,
_

J
2
,

J
y
= 0

,
_

J
2
,

J
z
= 0

. (K.2.1b)
Equations (K.2.1a) guarantee that

J is indeed a quantum-mechanical angular momentum. Equa-
tions (K.2.1b) guarantee the existence of a complete set of simultaneous eigenvectors of

J
2
and

J
z
.
Because (by assumption) the degrees of freedom of

J
1
are independent of those of

J
2
, we can write down
a whole host of other commutation relations such as
_

J
x1
,

J
y2

= 0 and so on:
2
Rule: Any component or function of

J
1
commutes with any component or function of

J
2
.
At this point, the questions we must answer are
(1) What are the eigenvalues of

J
2
and

J
z
?
(2) For a given set of quantum numbers (j
1
, m
1
) and (j
2
, m
2
), what are the values of the quantum num-
bers (J, M)?
(3) How can we write each eigenvector of

J
2
and

J
z
in terms of products of the known eigenvectors of the
individual operators

J
2
i
and

J
zi
?
The answers emerge, as answers so often do in quantum mechanics, from application of the method of
eigenfunction expansion.
K.3 The fundamental expansions of angular-momentum coupling
We begin by constructing simultaneous eigenvectors of

J
2
1
,

J
z1
,

J
2
2
, and

J
z2
. Because

J
1
and

J
2
act indepen-
dently, each of these eigenvectors is a product of single-angular momentum eigenstates:
3
[j
1
, j
2
; m
1
, m
2
= [j
1
, m
1
[j
2
, m
2
, UCAM eigenvectors. (K.3.1)
Ill call each [j
1
, j
2
; m
1
, m
2
an uncoupled-angular-momentum eigenstate (UCAM).
To determine each simultaneous eigenvector of

J
2
and

J
z
we expand in the set [ j
1
, j
2
; m
1
, m
2
. Ill
denote the eigenvectors we seek by [j
1
, j
2
; J, M and call each one a coupled-angular-momentum eigen-
state (CAM). Completeness of [ j
1
, j
2
; m
1
, m
2
allows us to expand each CAM eigenstate [j
1
, j
2
; J, M
in this set:
4
[j
1
, j
2
; J, M
. .
CAM
=
j
1

m
1
=j
1
j
2

m
2
=j
2
C(j
1
, j
2
, J; m
1
, m
2
, M) [j
1
, j
2
; m
1
, m
2

. .
UCAM
(K.3.2a)
The elaborate looking coecients C(j
1
, j
2
, J; m
1
, m
2
, M) in such expansions are called Clebsch-Gordan
coecients.
5
Although festooned with indices and subscripts, theyre just numbersin the phase conven-
tion used in this book, real numbers.
6
2
Commentary: By independent I mean that the operator

J
1
has no eect on functions of variables relevant to

J
2
, and vice
versa. For instance,

J
1
=

L acts on functions of the spatial angles and but has no eect on spinors, while

J
2
=

S acts on
functions of the spin variable but has no eect on functions of spatial variables.
3
Details: Mathematicians call these constructs direct products and write them as |j
1
, m
1
|j
2
, m
2
.
4
Details: We need not sum over j
1
and j
2
because |j
1
, j
2
; J, M and |j
1
, j
2
; m
1
, m
2
are both eigenvectors of

J
1

J
2
. So they
share the quantum numbers j
1
and j
2
5
Jargon: Many sources use so-called Wigner 3j symbols instead of Clebsch-Gordan coecients; the former are dened in
terms of the latter by
_
j
1
j
2
J
m
1
m
2
M
_
=
(1)
j
1
j
2
m
3
_
(2j
3
+ 1)
C(j
1
, j
2
, J; m
1
, m
2
, M).
Both Wigner 3j symbols and Clebsh-Gordan coecients oer advantages and disadvantages, but the distinction is merely
technical. Their physical signicance is the same.
6
Details: Clebsh-Gordan coecients are arbitrary to within an overall phase. Requiring the single coecient
C(j
1
, j
2
, J; j
1
, j
2
, J j
1
) to be a positive real number xes the phases of all coecients so they are (positive or negative)
real numbers. Thats why none of the relationships in this appendix that contain more than one Clebsh-Gordan coecient
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 97
Since the CAM set [ j
1
, j
2
; J, M is also complete, we can expand any UCAM state [j
1
, j
2
; m
1
, m
2
in
this set, obtaining the inverse of Eq. (K.3.2a):
7
[j
1
, j
2
; m
1
, m
2

. .
UCAM
=
j
1
+j
2

J=|j1j2|
J

M=J
C(j
1
, j
2
, J; m
1
, m
2
, M) [j
1
, j
2
; J, M
. .
CAM
(K.3.2b)
The number of terms in these summations is substantially constrained by the properties Eqs. (K.5.1) of
Clebsh-Gordan coecients:
Rule: Each eigenvector [j
1
, j
2
; J, M of

J
z
for a xed J and M is a linear combination of eigenvec-
tors [j
1
, j
2
; m
1
, m
2
with m
1
+ m
2
= M. No other [j
1
, j
2
; m
1
, m
2
contribute to [j
1
, j
2
; J, M.
In general the expansions (K.3.2) contain more than one term; but in special cases, all Clebsh-Gordan
coecients but one are zero; for instance,
Rule: For the particular coupled state [j
1
, j
2
; J, M with M = J, the linear combination Eq. (K.3.2a) reduces to
a single term, the uncoupled state [j
1
, j
2
; m
1
, m
2
with m
1
= j
1
and m
2
= j
2
, with a similar result for M = J:
[j
1
, j
2
; J, J
. .
CAM
= [j
1
, j
2
; j
1
, j
2

. .
UCAM
, and [j
1
, j
2
; J, J
. .
CAM
= [j
1
, j
2
; j
1
, j
2

. .
UCAM
. (K.3.3)
To keep straight the dierence between the UCAM and CAM sets, refer regularly to Tbl. K.1.
[UCAM [CAM
Dirac notation [j
1
, j
2
; m
1
, m
2
[j
1
, j
2
; J, M
dening operators

J
2
1
,

J
2
2

J
2
1
,

J
2
2

J
z1
,

J
z2

J
2
,

J
z

number of eigenstates (2j


1
+ 1)(2j
2
+ 1) (2j
1
+ 1)(2j
2
+ 1)
projection quantum numbers j
1
m
1
j
1
J M J
j
2
m
2
j
2
M = m
1
+ m
2
Table K.1. Properties of uncoupled and coupled angular momentum eigen-
states. For given values of j1 and j2, the number of eigenstates of the dening opera-
tors is the same. The allowed values of J are determined by the triangle rule .
_
j
1
j
2
J
_
of Eqs. (K.5.1a) and (K.5.2).
indicate complex conjugation of these numbers. Because of this convention, Clebsh-Gordan coecients satisfy symmetry rela-
tions like C(j
1
, j
2
, J; m
1
, m
2
, M) = (1)
Jj
1
j
2
C(j
2
, j
1
, J; m
2
, m
1
, M).You can nd more thrilling symmetry relations in the
readings in the Aside on p. 100.
7
Commentary: The sets {

L
2
,

S
2
,

L
z
,

S
z
} and {

L
2
,

S
2
,

J
2
,

J
z
} are complete with respect to the spatial variables and and
the spin variable (Chap. 6). Moreover, within each set, each pair of operators commutes. But, for a particle with spin s > 0
thats free to move in three dimension, these are not complete sets of commuting operators, because neither set includes the
Hamiltonian. Inclusion of the pure-Coulomb Hamiltonian

H =

T Ze
2
0
/r in either set leads to a CSCO. But inclusion of the
spin-orbit operator in this Hamiltonian gives

H =

T Ze
2
0
/r +

V
SO
. Because

V
SO

S, this Hamiltonian does not commute


with

L
z
or with

S
z
, so its inclusion leads to only one CSCO,
_

H,

L
2
,

J
2
,

Jz
_
.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 98
K.4 Eigenvalue equations
The commutation relations (K.2.1) guarantee the existence of a complete set of simultaneous eigenvectors
of

J
2
and

J
z
. This set is not, however, the set of product states [ j
1
, j
2
; m
1
, m
2
, because [j
1
, j
2
; m
1
, m
2

is not an eigenvector of

J
2
. (It is, however, an eigenvector of

J
z
.) The eigenvectors of

J
2
and

J
z
, the coupled
angular-momentum eigenstates, satisfy a whopping four eigenvalue equations:

J
2
1
[j
1
, j
2
; J, M = j
1
(j
1
+ 1)
2
[j
1
, j
2
; J, M , (K.4.1a)

J
2
2
[j
1
, j
2
; J, M = j
2
(j
2
+ 1)
2
[j
1
, j
2
; J, M , (K.4.1b)

J
2
[j
1
, j
2
; J, M = J(J + 1)
2
[j
1
, j
2
; J, M , (K.4.1c)

J
z
[j
1
, j
2
; J, M = M [j
1
, j
2
; J, M . (K.4.1d)
Prominently missing from this compendium are eigenvalue equations for

J
z1
and

J
z2
. The reason for their
absence is that [j
1
, j
2
; J, M are not eigenvectors of these operators!
K.5 Clebsch-Gordan coecients
Given values of j
1
, j
2
, m
1
, and m
2
, the Clebsh-Gordan coecient C(j
1
, j
2
, J; m
1
, m
2
, M) is zero unless J
and M satisfy the following requirements:
8

_
j
1
j
2
J
_
= [j
1
j
2
[ J j
1
+ j
2
(K.5.1a)
= J = [j
1
j
2
[, [j
1
j
2
[ + 1, . . . , j
1
+ j
2
1, j
1
+ j
2
, (K.5.1b)
and j
1
+ j
2
+ J is an integer, (K.5.1c)
and M = m
1
+ m
2
. (K.5.1d)
The symbol
_
j
1
j
2
J
_
is shorthand for Eq. (K.5.1a), which is called the triangle rule. This rule implies
that a particular C(j
1
, j
2
, J; m
1
, m
2
, M) equals zero unless the following inequality is satised:

_
j
1
j
2
J
_
= [j
1
j
2
[ J j
1
+ j
2
(K.5.2)
The constraint Eq. (K.5.1d) follows from the denition

J
z
=

J
z1
+

J
z2
:

J
z
[j
1
, j
2
; m
1
, m
2
=
_

J
z1
+

J
z2
_
[j
1
, j
2
; m
1
, m
2
(K.5.3a)
=
_

J
z1
+

J
z2
_
[j
1
, m
1
[j
2
, m
2
(K.5.3b)
= [j
2
, m
2


J
z1
[j
1
, m
1
+[j
1
, m
1


J
z2
[j
2
, m
2
(K.5.3c)
= (m
1
+ m
2
) [j
1
, j
2
; m
1
, m
2
. (K.5.3d)
So each [j
1
, j
2
; m
1
, m
2
is an eigenvector of

J
z
with eigenvalue m
1
+ m
2
. The only non-zero terms in the
fundamental expansions (K.3.2) are those with m
1
+ m
2
= M. For a given M, the number of terms in the
expansion (K.3.2a) equals the number of pairs (m
1
, m
2
) that satisfy this constraint. The total number of
states in [ j
1
, j
2
; J, M is
8
Details: The maximum values of m
1
and m
2
are j
1
and j
2
, respectively. Since M = m
1
+m
2
, as shown in Eqs. (K.5.3),
the maximum value of J is j
1
+ j
2
. But since M is a projection quantum number, its constrained by |M| J. This implies
that the maximum possible value of J is j
1
+ j
2
. A similar argument shows that the minimum value of J is |j
1
j
2
|. (The
absolute value signs are mandatory, since J 0).
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 99
j
1
+j
2

J=|j
1
j
2
|
(2J + 1) = (2j
1
+ 1)(2j
2
+ 1)
number of states in the
coupled and uncoupled sets
(K.5.4)
Crucially, the number of elements of [ j
1
, j
2
; J, M equals the number of elements of [ j
1
, j
2
; m
1
, m
2
.
This equality is useful when actually constructing elements of one set from elements of the other.
Useful properties of Clebsh-Gordan coecients. Hermiticity of the various angular-momenta weve
been discussing yield an abundance of useful properties. For example, the UCAM states satisfy the baroque-
looking orthogonality relations
j

1
, j

2
; m

1
, m

2
[ j
1
, j
2
; m
1
, m
2
=
j
1

,j
1

j
2

,j
2

m
1

,m
1

m
2

,m
2
. (K.5.5a)
A similar set of relations pertain to the CAM states,
j

1
, j

2
; J

, M

[ j
1
, j
2
; J, M =
j
1

,j
1

j
2

,j
2

J

,J

M

,M
. (K.5.5b)
From these relations there ows a veritable torrent of properties of the Clebsh-Gordan coecients. For
example, these coecients have their own orthonormality relations:
j1+j2

J=|j
1
j
2
|
J

M=J
C(j
1
, j
2
, J; m
1
, m
2
, M)C(j
1
, j
2
, J; m

1
, m

2
, M) =
m
1

,m
1

m
2

,m
2
, (K.5.6a)
j
1

m
1
=j
1
j
2

m
2
=j
2
C(j
1
, j
2
, J; m
1
, m
2
, M)C(j
1
, j
2
, J

; m
1
, m
2
, M

) =
J

,J

M

,M
(K.5.6b)
for values of J and M that satisfy the constraints Eq. (K.5.1). Among the host of other such properties,
derivations and examples of which you can nd sin the Aside on p. 100, here are two I nd particular handy
in practice:
j
1

m1=j1
j
2

m2=j2
C(j
1
, j
2
, J; m
1
, m
2
, M)
2
= 1, (K.5.7a)
j1+j2

J=|j1j2|
J

M=J
C(j
1
, j
2
, J; m
1
, m
2
, M)
2
= 1. (K.5.7b)
Special cases. For j
2
=
1
/
2
, one can generate any Clebsh-Gordan coecient via
[j
1
, j
2
; j
1
+
1
2
, M = cos
M
[j
1
, j
2
; M +
1
2
,
1
2
sin
M
[j
1
, j
2
; M
1
2
,
1
2
, (K.5.8a)
[j
1
, j
2
; j
1

1
2
, M = sin
M
[j
1
, j
2
; M +
1
2
,
1
2
+ cos
M
[j
1
, j
2
; M
1
2
,
1
2
, (K.5.8b)
where cos
M
=

j
1
M +
1
2
(2j
1
+ 1)
and sin
M
= (1)
2(j
1
+M)

j
1
+ M +
1
2
(2j
1
+ 1)
(K.5.8c)
Tables K.2 and K.3 compile values for especially important cases used in this book.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 100
Applications of Clebsch-Gordan coecients. At various places in this book, I make use of two results
that contain Clebsh-Gordan coecients. The rst is the integral of three spherical harmonics (of angles
and ) over the unit sphere:
_
2
0
_

0
Y
1,m1
Y
2,m2
, Y

3
,m
3
sin d d =

(2
1
+ 1)(2
2
+ 1)
4(2
3
+ 1)
C(
1
,
2
,
3
; 0, 0, 0)C(
1
,
2
,
3
; m
1
, m
2
, m
3
). (K.5.9)
This integral is zero unless the following conditions are met:
(1) The orbital angular momentum quantum numbers satisfy the triangle condition
[
1

3
[
2

1
+
3
;
(2) The projection quantum numbers satisfy m
1
+ m
2
= m
3
;
(3) The sum
1
+
2
+
3
is even.
The second result I use is the product of two spherical harmonics:
Y
1,m1
(, )Y
2,m2
(, ) =
1+2

L=|
1

2
|
L

M=L

(2
1
+ 1)(2
2
+ 1)
4(2 + 1)
C(
1
,
2
,
3
; m
1
, m
2
, m
3
)C(
1
,
2
,
3
; 0, 0, 0)Y
L,M
(, ). (K.5.10)
Key Points
For particular values of j
1
and j
2
, the allowed values of J are
J = [j
1
j
2
[, [j
1
j
2
[ + 1, . . . , j
1
+ j
2
1, j
1
+ j
2
. (K.5.11)
The number of values of J is 2 min(j
1
, j
2
) + 1, where min(j
1
, j
2
) is the smaller of j
1
and j
2
. For each J
there are 2J + 1 values of M, in integer steps in the range J M J.
For each pair (J, M) we construct the coupled angular momentum state [j
1
, j
2
; J, M as a linear combi-
nation of uncoupled states [ j
1
, j
2
; m
1
, m
2
. The only states [j
1
, j
2
; m
1
, m
2
that contribute to the
sum are those whose quantum numbers satisfy the requirements in Eqs. (K.5.1).
For M = j
1
+ j
2
, only one term contributes to the coupled state [j
1
, j
2
; J, J = [j
1
, j
2
; j
1
+ j
2
, j
1
+ j
2
:
the uncoupled state[j
1
, j
2
; m
1
, m
2
= [j
1
, j
2
; j
1
, j
2
.
Aside. Would you like to know more?
Most graduate-level textbooks on quantum mechanics contain good treatments of the addition of two an-
gular momenta: see, for example, Cohen-Tannoudji et al. (1977), Merzbacher (1998), and, at a rather more
formal level Messiah (1966). Some undergraduate books delve into this topic in more detail than I do:
see, for example, Bransden and Joachain (2000). But if you want the real story, you need to seek out
tomes devoted entirely to this topic. I mentioned many of these in the Selected Readings for Chap. 3
(??). Theres where youll nd the heart of the matter. Most symbolic math programs have the built-in
capability to generate Clebsh-Gordan coecients, but when using them you must be extremely careful.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 101
Not all sources adopt the phase convention I use, and there are about as many notations for these coe-
cients as there are books about quantum mechanics. Mathematica, for example, has the built-in function
ClebschGordan, which generates C(j
1
, j
2
, J; m
1
, m
2
, M) but expects the arguments in the following sequence:
[|j1, m1, |j2, m2, |J, M]. This dierence isnt a problem. . . unless you dont notice it, in which case all
your answers will be wrong. Apart from avoiding the excruciating tedium of calculating your own Clebsh-
Gordan coecients or typing them in from tables, a (major) advantage of such software is its ability to
generate and perform algebra with symbolic coecients, values of C(j1, j2, J; m1, m2, M) whose arguments
are symbols rather than, as in Tbls. K.2 and K.3, numbers.
j
1
=
1
2
j
2
=
1
2
J = 1 J = 0
m
1
m
2
M = 1 M = 0 M = 1 M = 0
1
2
1
2
1
1
2

1
2
_
1
2
_
1
2

1
2
1
2
_
1
2

_
1
2

1
2

1
2
1
Table K.2. Clebsh-Gordan coecients for j
1
= 1/2 and j
2
= 1/2. These co-
ecients are useful when coupling two electrons each of which is in an orbital with
= 0, as in the ground-state of helium. Empty entries signify that the corresponding
Clebsh-Gordan coecient equals zero.
j
1
= 1 j
2
=
1
2
J =
3
2
J =
1
2
m
1
m
2
M =
3
2
M =
1
2
M =
1
2
M =
3
2
M =
1
2
1
1
2
1
1
1
2
_
1
3
_
2
3
0
1
2
_
2
3

_
1
3
0
1
2
_
2
3
1
1
2
_
1
3
_
1
3
1
1
2
1
Table K.3. Clebsh-Gordan coecients for j1 = 1 and j2 = 1/2. These coecients are useful
when coupling the spin and orbital angular momentum of an electron, as in the hydrogen atom.
Empty entries signify that the corresponding Clebsh-Gordan coecient equals zero.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix L
The prex dictionary
Version 8.3: August 3, 2010
The names of multiples and submultiples of the units are formed with the prexes given in the table
below. Multiply the unit modied by the indicated prex or symbol by the power of ten in the column
labeled Factor. For example, 1m = 1 10
6
m. This table contains prexes commonly used in
spectroscopy. If you need to denote an order of magnitude not shown here, no one can stop you. For
starters, I recommend groucho, harpo, chico, and zeppo.
Prex Symbol Factor Prex Symbol Factor
exa E 10
18
deci d 10
1
peta P 10
15
centi c 10
2
tera T 10
12
milli m 10
3
giga G 10
9
micro 10
6
mega M 10
6
nano n 10
9
kilo k 10
3
pico p 10
12
hecto h 10
2
femto f 10
15
deka da 10 atto a 10
18
Note the following colloquial usages:
Prex Symbol Factor Usage
mega M 10
6
one million
giga G 10
9
one billion
tera T 10
12
one trillion
Appendix M
The Greek alphabet
Version 8.3: August 3, 2010
A, Alpha N, Nu
B, Beta , Xi
, Gamma O, o Omicron
, Delta , Pi
E, Epsilon P, Rho
Z, Zeta , Sigma
H, , Eta T, Tau
, Theta , Upsilon
I, Iota , Phi
K, Kappa X, Chi
, Lambda , Psi
M, Mu , Omega
Physicists always seem to be running out of letters to denote things. So we often use variant forms of Greek
letters, the most common of which are shown below.
variant usual name
, Gamma
Epsilon
Rho
Phi
theta
Theta
Kappa
Sigma
Appendix N
The Dirac delta function
Version 8.9: August 3, 2010
Shall I refuse my dinner because I do not fully understand
the process of digestion?
Oliver Heaviside
The Dirac delta function plays a huge role in many areas of physics. In this book, however, its primary use
is in the treatment of time-dependent Hamiltonians in Chap. 17. In fact, to understand the Golden Rule,
you have to understand a bit about the Dirac delta function. N.1 provides that bit. The basic denitions
and a bunch of useful properties are given in N.2; note especially Tbl. N.1, p. 107. A second use of the
Dirac delta function in this book is in comments on continuum functions; youll nd a word about this in the
Aside on p. 107.
N.1 A Dirac-delta primer
The most important thing to know about the Dirac delta function () is that its not a function.
1
Physically
measurable quantities must be represented by functions, which would seem to make the Dirac delta function
useless to physicists.
Not so! The Dirac delta function () is physically relevant when it appears in the integrand of an integral
over . We can write down the primary property of () in terms of such an integral whose integrand also
contains an arbitrary analytic function g():
_

b
a
g()( ) d =
_
g() if
a
< <
b
,
0 if <
a
or >
b
.
(N.1.1a)
This property makes an integral that contains a Dirac delta function incredibly easy to evaluate: the integral
equals the value of whatever multiplies the delta function [here, g()] evaluated for zero argument of the
delta function (here, for = ). (The integral equals zero, however, if the argument lies outside the range
of integration.)
The simplest application of Eq. (N.1.1a), in which we set g() = 1 and = 0 and integrate over all ,
leads to the useful result that the area under () equals 1:
1
Details: The Dirac delta function is not a function because its not analytic. Rather, its a thing mathematicians call a
generalized function or, more commonly, a distribution.
Appendix 105
_

() d = 1 (N.1.1b)
Equations (N.1.1) are the primary properties of the Dirac delta function. To relate () to transition
probabilities via the Golden rule (Chap. 17) we need to understand that () can be written in several ways
as limits of analytic functions of . Each involves a limit in which a real number goes to zero through
positive values. (In math-speak, such a limit is indicated by 0
+
). Indeed, the primary denition of the
delta function is
() lim
0
+
1

2
+
2
_
, (for real ). (N.1.2a)
For our purposes, three limiting forms of the delta function are useful:
() = lim
0
+
1

sin(/)

(N.1.2b)
() = lim
0
+

_
sin
2
(/)

2
_
(N.1.2c)
() = lim
0
+
1

2
/
2
. (N.1.2d)
Try This! 14.1. A highly relevant question.
To what physically relevant functions are Eq. (N.1.2b) and Eq. (N.1.2c) germane? For each function,
what physically relevant quantity plays the role of ?
The last of these limits, Eq. (N.1.2d), helps us understand how the delta function behaves in its dening
limit 0
+
. The function on the right side of Eq. (N.1.2d) is a normalized Gaussian in the variable
2
((; )
1

2
/
2
, where
_

((; ) d = 1 for any (N.1.3)


The function ((; ) attains its peak value at = 0, where its equal to 1/

. Its width is proportional


to 1/. Figure N.1 shows that as decreases towards zero, its peak height rapidly increases and its width
rapidly decreases; all the while, the area under ((; ) remains equal to 1. This gure thus illustrates how
the Gaussian function approaches () as goes to zero.
For applications of the Dirac delta function to switched-on perturbations, in which the switching function
involves step functions, its useful to know that the Dirac delta function is the rst derivative of the the
Heaviside step function (unit-step function):
(x) =
d
dx
(x), (N.1.4a)
where (t)
_
0 x < 0
1 x > 0
Heaviside step function (N.1.4b)
Finally, () is the limit of a pulse function of width 2:
( ) = lim
0
+
1
2
[( ) + ] + [( ) ] . (N.1.5)
2
Details: This Gaussian is centered at = 0. Its standard deviation is , and its full-width at half maximum is 2 ln 2. We
used a Gaussian as a switching function in Eq. (??); see Fig. ??, p. ??.
PrintAppendices Version: 8.9 Printed: August 3, 2010
Appendix 106
Figure N.1. A normalized
Gaussian function of decreas-
ing widths. The limit 0
+
of this function is one representa-
tion of the Dirac delta function. In
time-dependent perturbation the-
ory (Chap. 17) the parameter is
the inverse of the duration of the
perturbation.
0.4 0.2 0.0 0.2 0.4
0
5
10
15
20
0.03
0.06
0.13
0.25
N.2 A Dirac-delta users guide
The discussion in N.1 is geared for users of the Dirac delta function in perturbation theory. So there I
used as the independent variable. The results in this section are of wide generality, so Ill use x. Basic
dening properties appear in Tbl. N.1, p. 107.
Additional Properties. Notation:

(x) denotes the rst derivative d(x)/ dx.


(ax) =
1
a
(x) for a > 0

(x) =

(x)
x(x) = 0
(x
2
a
2
) =
1
2a
_
(x a) + (x + a)
_
for a > 0
f(x)(x a) = f(a)(x a)
[f(x)] =
1
[df/dx[
x=x0
(x x
0
), for x
0
a root of f(x)
_

0
f(p)(E
p
E
p
) dp =
m
p

f(p

), where E
p
= p
2
/2m
Integrals.
_

(x)f(x) dx = f

(0)
_
(x b)(a x) dx = (a b)
_

d
m
(x)
dx
m
f(x) dx = (1)
_
d
m
f
dx
m
_
x=0
PrintAppendices Version: 8.9 Printed: August 3, 2010
Appendix 107
Representations [see also Eqs. (N.1.2)]
(E
1
E
2
) =
i
2
lim
0
_
1
E
1
E
2
+ i

1
E
1
E
2
i
_
(x) =
1
2
_

e
i kx
dk [Fourier transform of 1/

2]
(x x

) =
1
2
_

e
i k(xx

)
dk
(r r

) =
1
(2)
3
_
e
i k(rr

)
d
3
k =
1
(2)
3
_
e
i p(rr

)/
d
3
p
(x x

) =
1

_

0
cos k(x x

) dk
(x x

) =
1
2

e
i n(xx

)
(x x

) =
1
2
_
1 +

1
2 cos n(x x

)
_
(x x

) = lim
0
e
(xx

)
2
/
2

(x x

) =
1

lim
0

(x x

)
2
+
2
(x) =
1
2L
+
1
L

n=1
cos
_
n
x
L
_
[Fourier series for region +L to L]
(x x

) =

n=0
1

2
n
n!
exp
_

_
x
2
+ x

2
2
__
H
n
(x)H
n
(x

).
In the last equation H
n
(x) is the n
th
-order Hermite Polynomial.
formal denition Eq. (N.1.2)
dening properties Eqs. (N.1.1)
normalization Eq. (N.1.1b)
nature (x x
0
) =
_
_
_
0 x ,= x
0
x = x
0
dimensions (1D) (x x
0
) has dimensions of L
1
dimensions (3D) (r r
0
) has dimensions of L
3
(x) is real

(x) = (x)
(x) is even (x) = (x)
Table N.1. Denitions and basic properties of the Dirac delta function.
Aside. The Dirac delta function and normalization of continuum eigenfunctions.
The wave function for a continuum stationary state cannot be normalized because it never decays to
zero. The normalization requirement for such a state is that the modulus of the wave function must be
PrintAppendices Version: 8.9 Printed: August 3, 2010
Appendix 108
nite for all values of its variables. Consider a continuum stationary state of energy E =
2
k
2
/2m with
spatial wave function
k
(x). The normalization condition for this continuum stationary state must be
written in terms of the Dirac delta function:
_

k
(x)
k
(x) dx (k

k), (N.2.1)
where the constant of proportionality is arbitrary. This condition called Dirac delta function normal-
ization. To learn more about these continuum states and their interpretation see 8.5 in Understanding
Quantum Physics.
PrintAppendices Version: 8.9 Printed: August 3, 2010
Appendix O
Waves and the de Broglie-Einstein relations
Version 8.8: August 3, 2010
This appendix contains basic equations and constants that relate wave-like and particle-like properties of
denizens of the microverse. The basic properties of waves are summarized in Tbl. O.1 and discussed below.
Particle-like properties include total energy E and linear momentum p; wave-like properties (for a vacuum)
include wavelength , frequency , angular frequency , and wave number k. Table O.2 contains a variety
of relationships between these quantities for photons and for massive particles. Finally, Tbl. O.3 lists values
of constants widely used in the study of waves.
quantity symbol key relationships SI units
period T T =
1

=

c
s
frequency = 2 Hz = s
1
angular frequency
1
=

2
rad s
1
wavelength
2
=
c

m
wave number
3
k k =
2

=

c
m
1
1
The angular frequency equals the number of radians per unit
time, while the frequency, equals the number of cycles per unit
time. For an EM wave propagating along the x axis, the angle
referred to in the name angular frequency is the polar angle of
the electric eld in the xy plane.
2
The wavelength is the spatial interval along the direction of prop-
agation specied by the unit vector e
k
k/k within which the
wave replicates itself.
3
The wave number k is 2 times the number of cycles per
unit distance.
Table O.1. Basic quantities for the description of waves in physics. The propagation
speed of the wave in a vacuum is the speed of light c. The wave number k is the magnitude
of the wave vector k = k

k, which points along the propagation direction



k of the wave. The
conventional unit of frequency is the hertz, Hz. [The angular frequency is sometimes called
the angular speed or (incorrectly) the angular velocity.] The quantity 1/ is a reciprocal
wavelength, not a wave number, a frequency, or an energy (see Appendix J).
Appendix 110
property p k
for a photon (rest mass equals zero)
energy E = pc h
hc

c k
linear momentum p =
E
c
h
c

k
for a free massive particle (rest mass greater than zero)
energy E =
p
2
2m
h
_
h
2
2m
_
1

2
_

2
2m
_
k
2
linear momentum p = mv
_
2m
h
_
2m

k
Table O.2. The de BroglieEinstein relations for photons and for massive particles (that is,
particles with rest mass m > 0). A photon has rest mass m = 0, travels at speed c; its energy is
purely kinetic, E = pc. A free massive particle has rest mass m > 0 and travels at speed v c.
Being free, its energy is also purely kinetic, E = p
2
/2m.
Basic denitions. Consider the following sinusoidal function of space and time (a standing wave),
g(x, t) = sin(kx t) = sin
_
2

x
2
T
t
_
= sin
_
2
_
x


t
T
__
= sin [2 (x t)] . (O.0.1)
This function repeats itself during a time interval T, which is called the period, and over a spatial interval ,
which is called the wavelength. The wavelength is usually dened as the distance between successive wave
crests. A wave moves forward a distance of one wavelength during a time interval of one period.
1
Outside
an interval of width , the function repeats itself.
2
The factor of 2 appears in Eq. (O.0.1) because the
wavelength of sin x is 2:
3
sin(x + 2) = sin x. (O.0.2)
In Eq. (O.0.1) the frequency of a sinusoidal function with period T is dened as

1
T
(O.0.3)
The corresponding angular frequency is dened as
2
2
T
(O.0.4)
The quantity is the number of times the function repeats itself during a time interval t = 1, while is
the number of repetitions during t = 2. The units of frequency are s
1
= Hz (Hertz), while the units
of angular frequency rad s
1
, include the (dimensionless) SI unit of angle, the radian (see Appendix E).
1
Details: For example, in the classical physics of a particle executing simple harmonic motion, the period is the time required
for one cycle of the motion, and the frequency is the number of cycles per unit time interval.
2
Jargon: The function is thus said to be periodic, although in this sense the allusion to the waves period is indirect.
3
Details: The function sin
2
x is not normalized on its periodicity interval, 2. To normalize this function we must multiply
it by 1/

.
PrintAppendices Version: 8.8 Printed: August 3, 2010
Appendix 111
The wave number k is dened as
4
k
2

= 2 (O.0.5)
The SI units of k are radians per meter; those of are inverse meter. Mathematically, equals the number
of repetitions of the basic pattern of the wave in a spatial interval x = 1that is, is a reciprocal
wavelengthwhile k is the number of repetitions in an interval of width x = 2. Last but not least, the
amplitude of the function g
k
(x, t) is A = 1, and its phase (argument) is kx t.
Relevance to quantum mechanics. In quantum physics, the wave number of a particle is related to its
linear momentum by the de Broglie relation, which in one dimension reads p = k with /2 being the
reduced Plancks constant. For more equations, see Tbl. O.2. In atomic units we set 1, and k equals
the linear momentum k = p. Note that since = 2k, in atomic units p = h /2.
A normalized sine function with arbitrary wave length. The function
h(x) =
_
2
L
sin
_
2
x
L
_
(O.0.6)
is normalized on an interval of width L. Its wavelength is = L, its wave number is k = 2/L, and its
amplitude is A =
_
2/L. This function is said to be periodic with a minimum periodicity interval of
width L:
5
h(x + L) = h(x). (O.0.7)
This function satises the periodic boundary conditions
h(0) = h(L) = 0. (O.0.8)
[Since I have not introduced a shift in the phase of the sine function in Eq. (O.0.6), the value of h(x) at
x = 0 and x = L are zero.] Correspondingly, the cosine function
h(x) =
_
2
L
cos
_
2
x
L
_
(O.0.9a)
is normalized on the same interval as the sine function, has the same wavelength and fundamental period,
but satises the periodic boundary conditions
h(0) = h(L) = 1. (O.0.9b)
Finally, the periodic exponential functions with fundamental period L are
e
2i x/L
=
_
cos
_
2
x
L
_
+ i sin
_
2
x
L
__
. (O.0.10)
Aside. Useful wavelength conversions.
In converting between wavelength and energy the following relationships are useful (units are given as
arguments):
(m) =
1.24
E( eV)
photon (O.0.11a)
(m) =
1.23
_
E( eV)
electron, (O.0.11b)
where 1m = 10
6
m = 10
4

A.
4
Jargon: In the literature of atomic and molecular spectroscopy, the quantity = 1/ is, misleadingly, called the wave
number (see Appendix J).
5
Jargon: This interval is often called the fundamental period of the function. Thus this sine function has fundamental
period L.
PrintAppendices Version: 8.8 Printed: August 3, 2010
h = 4.1357 10
16
eV s
= h/2 = 6.5822 10
16
eV s
c = 2.998 10
8
m s
1
hc = 12398.42 eV

A
1

A = 10
10
m
1 eV = 1.6022 10
19
J
Table O.3. Constants used in the quantum mechanics of wave-particle duality. You
can nd more precise values of these constants in Appendix C.
Appendix P
The one-dimensional simple harmonic oscillator
Version 8.4: August 3, 2010
P.1 Hamiltonian eigenfunctions and stationary-state energies
The potential energy of a particle of mass m conned to one dimension (x) and subject to a restoring force
that corresponds to a natural (angular) frequency
0
is
1
V (x) =
1
2
m
2
0
x
2
. (P.1.1)
The normalized Hamiltonian eigenfunctions of the one-dimensional (1D) simple harmonic oscillator
(SHO) are

n
() =

n! 2
n
H
n
() e

2
/2
, n = 0, 1, . . . (P.1.2a)
where H
n
() is the n
th
Hermite polynomial . In Eqs. (P.1.2a), the independent variable is
x, (P.1.2b)
where
_
m
0

. (P.1.2c)
Since the parameter has dimensions [] = L
1
, the variable is dimensionless.
From one can dene the characteristic length of a 1D SHO:
x
c

1

2
=
_

2m
0
characteristic length of a 1D SHO (P.1.3)
The factor 1/

2 is included by convention so the characteristic length will equal the standard deviation of x
2
in the ground state. Since x
n
= 0, this quantity is the position uncertainty in the ground state:
2
1
Details: In terms of the force constant by k
f
, the natural frequency (in units of radians per unit time) is =
_
k
f
/m.
The force constant equals the curvature of the potential energy evaluated at the value xm where it attains its minimum value:
k
f
= d
2
V/ dx|x=x
m
.
2
Details: The ground-state eigenfunction has the mathematical form of a Gaussian function. In terms of the characteristic
length, its full-width-at-half-maximum (FWHM) is x
fwhm
= 4

ln2 x
c
. Note that the FWHM does not equal the position
uncertainty in the ground state; rather, x
fwhm
= 4

ln2 x
0
.
Appendix 114
(x)
0
=
_
x
2

0
=
_

0
[ x
2
[
0
= (x)
0
= x
c
. (P.1.4)
The Hermite polynomials are solutions of the Hermite dierential equation
_
d
2
d
2
2
d
d
+ 2n
_
H
n
() = 0, Hermite dierential equation (P.1.5)
Table P.1 gives the rst six Hermite polynomials, and Tbl. P.2 the corresponding Hamiltonian eigenfunctions.
To preserve normalization, the eigenfunctions Eq. (P.1.2a), when expressed as a function of the displace-
ment from equilibrium x, must be written

n
(x) =
_

n
() (P.1.6a)
Explicitly, then, these eigenfunctions are

n
(x) =
_
m
0

_
1/4
1

2
n
n!
H
n
(x
_
m
0
/) e
m
0
x
2
/2
(P.1.6b)
All states of the SHO are bound. The stationary state energies are
E
n
=
_
n +
1
2
_

0
, n = 0, 1, . . . (P.1.7a)
In terms of the oft-used dimensionless energy
n
, these eigenvalues are

n

2E
n

0
= 2n + 1, n = 0, 1, . . . (P.1.7b)
The following matrix elements are useful when working with the SHO:
3

n
[ x [
m
=
_

_
1

_
n + 1
2
m = n + 1
1

_
n
2
m = n 1
0 otherwise.
(P.1.8a)

n
[ x
2
[
m
=
_

_
2n + 1
2
2
m = n
_
(n + 1)(n + 2)
2
2
m = n + 2
0 otherwise.
(P.1.8b)
3
Notation: Throughout, Dirac brackets signify integration over all space (< x < ), and primes denote derivatives:
n | x | m
_

n
(x) xm(x) dx

n
|

m

_

n
(x)
d
m
(x)
dx
dx
PrintAppendices Version: 8.4 Printed: August 3, 2010
Appendix 115

n
[

m
=
_

_
+
_
n + 1
2
m = n + 1

_
n
2
m = n 1
0 otherwise.
(P.1.8c)
_

n
(x)
2n+2
(x)

0
(x)
dx = 0. (P.1.8d)
Table P.3 lists statistical properties of the SHO which follow from these integrals.
Hermite polynomials H
n
()
H0() 1
H
1
() 2
H2() 4
2
2
H
3
() 8
3
12
H4() 16
4
48
2
+ 12
H
5
() 32
5
160
3
+ 120
Table P.1. The rst six Hermite polynomials.
n Hamiltonian eigenfunction n(x)
0
_

_
1/4
e

2
x
2
/2
1
_

_
1/4
xe

2
x
2
/2
2
_

_
1/4
1
8
_
4
2
x
2
2
_
e

2
x
2
/2
3
_

_
1/4
1
48
_
8
3
x
3
12x
_
e

2
x
2
/2
4
_

_
1/4
1
384
_
16
4
x
4
48
2
x
2
+ 12
_
e

2
x
2
/2
5
_

_
1/4
1
3840
_
32
5
x
5
160
3
x
3
+ 120x
_
e

2
x
2
/2
Table P.2. The rst six Hamiltonian eigenfunctions of the 1D SHO.
PrintAppendices Version: 8.4 Printed: August 3, 2010
Appendix 116
Statistical properties
x)
n
= 0 (x)
n
=
1

_
n +
1
2
=
_
E
n
m
2
0
p)n = 0 (p)n =
_
n +
1
2
=

mEn
(x)n(p)n =
_
n +
1
2
_

Table P.3. Statistical properties of stationary states of the 1D SHO.


PrintAppendices Version: 8.4 Printed: August 3, 2010
Appendix 117
P.2 Creation and annihilation operators
The dimensionless position and momentum operators for the 1D SHO are dened as

Q
_
m
0

x = x (P.2.1a)

P
1

m
0

p =
p

. (P.2.1b)
These quantities are related to the creation operator a

, the annihilation operator a, and the number


operator

N, whose denitions are
4
a

2
(

Q i

P) creation operator (P.2.2a)
a
1

2
(

Q + i

P) annihilation operator (P.2.2b)

N a

a. number operator (P.2.2c)


In terms of these operators, the dimensionless position and momentum operators become

Q =
1

2
( a

+ a) dimensionless position operator (P.2.2d)

P =
i

2
( a

a) dimensionless momentum operator (P.2.2e)


The number operator is related to the dimensionless length variable by

N =
1
2
_
d
2
d
2
+
2
+ 1
_
. (P.2.3)
The operator

Q is the counterpart of the dimensionless variable dened in Eq. (P.1.2b). From the
commutation relation
_
x, p

= i , one can show these new operators satisfy


5
_

Q,

P

= i (P.2.4a)
_
a, a

= 1 (P.2.4b)
_

N, a

= a

(P.2.4c)
_

N, a

= a (P.2.4d)
The Hamiltonian for the potential (P.1.1) can be written in the following more-or-less useful forms:
4
Jargon: These operators are also called ladder operators, with a

the raising operator and a the lowering operator.


The creation and annihilation in the 1D SHO refers to creation or annihilation of a quantum of energy. This delightfully
apocalyptic jargon is perhaps most appropriate in the theory of second quantization, where creation and annihilation operators
are used to create and destroy the number of identical particles in a system. See Chap. ??.
5
Read on: You can nd useful discussions of the application of these operators to the 1D SHO in Chap. 7 of Introductory
Quantum Mechanics, Third Edition, by Richard L. Libo (Reading, Mass: Addison-Wesley, 1997) and Sec. 2.3 of Introduction
to Quantum Mechanics by David J. Griths (Englewood Clis, NJ: Prentice-Hall, 1995). The more mathematically inclined
should check out Chap. 3 of An Introduction to Quantum Theory by Keith Hannabuss (New York: Oxford, 1997) for details of
both analytic and operator solutions of the 1D SHO. For an extensive investigation of the physical and mathematical features
of the 1D SHO see Chap. 9 of my book Understanding Quantum Physics (Englewood Clis, NJ: Prentice-Hall, 1990) or,
at a more advanced level, Chap. V of Quantum Mechanics by Claude Cohen-Tannoudji, Bernard Diu, and Franck Laloe (New
York: Wiley-Interscience, 1997).
PrintAppendices Version: 8.4 Printed: August 3, 2010
Appendix 118

1 =
p
2
2m
+
1
2
m
2
0
x
2
(P.2.5a)
=
1
2

0
(

P
2
+

Q
2
) (P.2.5b)
=
1
2

0
( a a

+ a

a) (P.2.5c)
=
0
( a

a +
1
2
) (P.2.5d)
=
0
( a a

1
2
) (P.2.5e)
=
0
(

N +
1
2
) (P.2.5f)
From these expressions follow the useful commutation relations
_
a,

1

=
0
a (P.2.6a)
_
a

,

1

=
0
a

. (P.2.6b)
Denoting the orthonormal eigenfunctions of

N by [n, we have

N [n = n[n number operator (P.2.7a)


a

[n =

n + 1 [n + 1 creation operator (P.2.7b)


a [n =

n [n 1 annihilation operator (P.2.7c)


a [0 = 0. ground state (P.2.7d)
When acting in succession on a state [n, the creation and annihilation operators return the original state,
a

a [n = n[n (P.2.8a)
a a

[n = (n + 1) [n . (P.2.8b)
The eect of the position and momentum operators of Eqs. (P.2.1) on a state [n is therefore
x [n =
_

2m
0
( a + a

) [n (P.2.9a)
=
_

2m
0
(

n [n 1 +

n + 1 [n + 1 (P.2.9b)
p [n = i
_
m
0
2
( a a

) [n (P.2.9c)
= i
_
m
0
2
(

n [n 1

n + 1 [n + 1 (P.2.9d)
In terms of these operators, the TISE for the oscillator assumes the forms

1[ a [n ] = (E
n

0
) [ a [n ] (P.2.10a)

1[ a

[n ] = (E
n
+
0
) [ a

[n ] (P.2.10b)
One can generate the bound stationary state [n from the ground state [0 via
[n =
1

n!
( a

)
n
[0 . (P.2.11)
PrintAppendices Version: 8.4 Printed: August 3, 2010
Appendix 119
The mean values of position and momentum for both stationary and non-stationary states evolve according to
d
dt
x =
1
i

_
x,

1

=
p
m
(P.2.12a)
d
dt
p =
1
i

_
p,

1

= m
2
0
x. (P.2.12b)
Finally, the following matrix elements are useful in working with the SHO:
n

[ a

[ n =

n + 1
n

,n+1
(P.2.13a)
n

[ a [ n =

n
n

,n1
(P.2.13b)
n [ x
2
[ n =
_
n +
1
2
_

m
0
(P.2.13c)
n [ p
2
[ n =
_
n +
1
2
_
m
0
. (P.2.13d)
P.3 Momentum-space wave functions
Although the Fourier transform of the SHO potential Eq. (P.1.1) does not exist, the time-independent
Schrodinger equation in momentum space can still be solved. The dimensionless momentum variable corre-
sponding to = x (not to be confused with the radial variable of plane polar coordinates in Chap. 2) is
6

1

p =
1

m
0

p =
1

2 p
c
p, dimensionless momentum variable of the 1D SHO (P.3.1)
In the last equality of Eq. (P.3.1) I introduced the characteristic momentum of a 1D SHO:
p
c

_
m
0

2
= p
2

0
= p
0 characteristic momentum of a 1D SHO (P.3.2)
Each momentum-space Hamiltonian eigenfunction
n
(p) is the Fourier transform of the corresponding
coordinate-space function:

n
(p) =
1

2
_

e
i p x

n
(x) dx, (P.3.3)
The mathematical nature of the SHO potential is such that the momentum-space eigenfunctions have the
same dependence on p as the conguration-space functions
n
(x) have on x [compare to Eq. (P.1.2a)]

n
() = i
n
_
1
m
0

_
1/4
1

2
n
n!
H
n
() e

2
/2
(P.3.4)
6
Details: In terms of the dimensionless length variable, the linear momentum operator p
x
= i /x becomes
p

= i / .
PrintAppendices Version: 8.4 Printed: August 3, 2010
Appendix Q
The Seven Habits of Highly Eective Problem Solvers
Version 8.14: August 3, 2010
No problem can stand the assault of sustained thinking.
Francois Marie Arouet de Voltaire
Q.1 Why you need to develop new habits
Most students bring to the study of college-level physics a bunch of problem-solving techniques they pick up
in high school. These techniques may work for simple plug-and-grind problems but are useless for the more
interesting problems one encounters in quantum mechanics. By now, many of these old, no-longer eective
habits are almost second nature. Yet, many of these habits are counterproductive, so youll have a lot more
success (and fun) if you change them. This appendix describes one approach to more thoughtful, enjoyable
problem solving.
Three premises underlie the approach I want you to learn:
(1) You already know much more than you realize; your diculty is accessing and using that knowledge
eectively.
(2) Knowledge is hierarchical and cumulative, so the most eective ways of accessing and using knowledge
take these features into account.
(3) Problem solving is a lot of work. So it should be as fruitful, painless and even fun as possible.
These principles constitute an approach that maximizes fun and insight and minimizes algebra, error, and
tedium.
Throughout this book, I develop, explain, and extensively illustrat this way of approaching physics
problems. This appendix codies the essential strategies and tactics of this way of thinking into seven easily
remembered habits. While there are many eective ways to approach problem solving in physics, the
many students who have learned these habits, both in quantum-mechanics and in other physics courses,
report that applying them makes problem-solving easier, less prone to error, much more ecient, and more
fun. To use these strategies eectively, you must apply them over and overto enough problems that they
become truly habitual, so much a part of your thinking that your mind will automatically apply them. So for
a while youll probably need a list like the one below prominently displayed on your deskto remind you,
for example, to begin work on every problem by brainstorming; to seek analogies and patterns between the
problem youre trying to solve and others youve already solved; to avoid algebraic complexity at each stage;
Appendix 121
and so forth. I have found that it usually takes students about two months of regular, frequent, conscientious
application to internalize these habits.
THE SEVEN PRINCIPLES
(1) Begin by brainstorming
(2) Seek symmetry simplications
(3) Make conceptual models
(4) Pursue familiar patterns
(5) Back to basics
(6) Eschew explicit expressions (The Principle of Least Algebra)
(7) Keep your brain engaged
The rest of this appendix briey summarizes the essence of each habit. These summaries are necessarily a
bit general; to really understand and learn a habit you need to see it used, which means you need to return
to examples in this book.
Q.2 Implementation Strategies
Most students Ive worked with get lost while problem solving for one of three reasons:
(1) They dont know where they want to go or how to get there. (This wont happen to you if you follow
The Principle of Brainstorming; see Q.3.)
(2) Theyre using equations that are more complicated than they need to be. (This wont happen to you
if you follow The Principles of Brainstorming (Q.3), of Least Algebra Q.8, and of Alert Awareness
(Q.9.)
(3) Theyre trying to do too many things at once (juggle equations, solve equations, plug numbers into
equations, etc.) (This wont happen to you if you follow The Principle of Alert Awareness; see
Q.9).
The core strategy I advocate is
Rule: Approach problems systematically, strategically, and, initially at least, in a spirit of play. Tackle problems
in three stages:
(1) Brainstorm the problem.
(2) Work the problem (do the math, calculations, graphs. . . whatever).
(3) Determine whether your results make sense.
Tips for getting started.
(1) When you decide to (or are assigned to) work a problem, read it immediately. Just read it: dont try
to solve it; dont fret about it. But read it with your mind, not just your eyes.
(2) If time permits, put the problem aside for a day or so and go about your life while your mind digests
and works on the problem. Then take out the problem and brainstorm (Q.3). Then and only then
take out your calculator, tables, text, scratch paper, and start work.
(3) Dont put yourself under unnecessary stress by letting the problem sit around too long. (Procrastination
isnt immoral, but neither is it cost eective.) Start work early. Give yourself a break: Work on
problems during frequent brief sessions, rather than one marathon slugfest.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 122
(4) Always be willing to put a problem aside for a while, then return to it later. Youll be surprised what
your mind gured out while you were doing other things!
With this background, lets talk tactics.
Q.3 The Principle of Brainstorming
The most powerful way to enhance your mastery of problem solving is brainstorming.
Rule: Get into the habit of brainstorming as the rst stage in tackling any problem.
When you brainstorm, youre not trying to work the problem; youre not trying to get the right answer; and
youre not doing calculations. What you are doing is playing with the ideas raised by the problem.
Guidelines to successful brainstorming.
(1) Jot down as many ways to approach the problem (or parts of it) as you can think of.
(2) Dont worry about practical issues (can I do this? how hard would it be?).
(3) Use few (or no) equations and few (or no) calculations. Remember: your goal is to wrap your mind
around the problem. Once youve done that, the equations and calculations will follow.
Key questions for a brainstorming session.
(1) Whatpreciselyis the problem asking for? For that matter, do I really understand the problem? To
check, try to restate the problem in your own words without referring to the original statement.
(2) What physical principles are involved in the problem? (This does not mean what equations do I
use?)
(3) Might I be able to simplify the problem via modeling or by invoking approximations? Can I think of
more than one relevant model and/or approximation?
(4) Can I express the problem in sketches? Draw lots of simple, rough, qualitative sketchesas many as
you can think ofto illustrate or clarify the situation posed in the problem.
Tip. Keep your brainstorming notes informal but neat enough that you can read them. Prepare a clear
written listof about half a pageof the key elements of the problem (quantities, physical principles, data supplied,
and so forth). When you start work on the problem, this list is your guide, your map out of any swamps of dreary
algebra into which you might stray.
Tip. Try to devise several approaches to the problem without fretting about how hard they are
or even whether theyll work. That is: once youve come up with one way to approach the problem, stop.
Reconsider. Ask yourself, What is another way to solve this problem? Can I think of several ways? If so, jot them
down. Then ask, Which way seems easiest and clearest and most direct?
Q.4 The Principle of Symmetry Seeking
From Chap. 1 Ive been urging you to begin considering a new physical system or problem in terms of the
systems symmetry properties. In its power to see through surface complexities to the essence of a system
or problem, symmetry is the hydrogen bomb of tactics.
Rule: Start by identifying symmetry properties of a system. Then think about their consequences for the
observables, operators, and wave functions of the system. To see a systems symmetries, consider rst its
geometric structure and its potential energy, which arises from forces acting on or within the system.
Its particularly crucial to seek symmetries in generic properties of the systems potential energy, not in
its explicit mathematical form. For instance, the essential symmetries of of the screened Coulomb model
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 123
(the Yukawa potential we used in Part III to model atoms) is that this potential energy is spherically
symmetric (that is, the system is rotationally invariant)not that the potential energy has the particular
form V (r) = e
r
/r. In fact, looking at the particular form of many potential energies obscures their
essential symmetry.
Tips for symmetry seekers.
(1) Try to nd a coordinate system that closely mimics the essential symmetry of the system. (For instance,
dont try to solve the Schrodinger equation for atomic hydrogen in Cartesian coordinates unless you
have a really good reason do so.)
(2) Follow through insights into symmetry to identify generic properties of wave functions and other quan-
tum critters. For instance, Chap. 4 is chock full of details about stationary-state wave functions of
all central-force problems. Once you identify a (single-particle) system as rotationally invariant, you
can just write down the form of its stationary angular momentum eigenstates and the structure of its
radial functions, including their limiting behaviors at the origin and in the limit r all without
solving a single equation!
(3) Continue seeking symmetry as you move from brainstorming to problem solving as such. In particular,
after you recast your equations in dimensionless form (see Appendix B), reconsider what they can tell
you about the symmetry of the system they describe. Sometimes the simplications that result from
exterminating dimensional quantities in equations reveal powerful symmetries previously obscured by
the clutter of constants and other quantities.
(4) Learn group theory. (I know, I know: Youre busy. Time is limited. And group theory is math.) But
group theory is really powerful math. Armed with The Principle of Symmetry Seeking and the
power of group theory, you can wreak magnicent, illuminating simplications in an enormous number
of otherwise fearsome problems (not that there are any like that in this book).
Q.5 The Principle of Modeling
A key part of the brainstorming stage of problem solving is devising a model of the physical system youre
studying. From the beginning of this book, weve seen that real physical systems are far too complicated
to be investigated except in the context of models. A model is a story; our rst model of a hydrogen
atom (Chap. 5) is, a hydrogen atom is an electron bound to an innitely massive point proton with
which it interacts only via the electrostatic Coulomb interaction. This story jettisons a lot of the physics
of a real hydrogen atom: the nite nuclear size, electron and nuclear spin, relativistic eects, and so forth.
Nevertheless, this model resembles an actual hydrogen atom closely enough to provide a context for answering
a lot of questions correctly, albeit approximately. Ideas for the model emerge naturally as you brainstorm
the problem.
1
Key questions for model building.
(1) Precisely what physical properties (or related quantities) do I need to evaluate?
(2) How accurately do I need to know these quantities?
(3) What physical features of the system might aect these quantities? (Often you can get insight into
this question by pondering the physical meaning of the terms in the potential energy.)
(4) Which of these features will least aect the quantities I want to the accuracy I need?
(5) Based on this analysis, which features of the physics must I retain to get the accuracy I desire in the
quantities of interest?
Tip. In your rst stab at a model, jettison every feature of the physics you dont think you absolutely must retain.
Retain only the most important interactions and features of the system. The resulting problem will be as simple as
1
Commentary: Model building is not the same as deciding which mathematical approximations you need to invoke to solve
the problem. For instance, deciding that your model of a hydrogen atom needs to include, say, the spin-orbit interaction is
dierent from deciding that youll include the spin-orbit interaction potential operator via perturbation theory (Chap. 14).
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 124
you can make it (without making it so simple that it no longer resembles the problem). Thats your rst model. If,
after calculating the necessary quantities in your rst model you nd that its not accurate enough, your analysis will
facilitate deciding what to put back in to create your next model.
Tip. Dont start looking for equations too soon! Only when you can write down a verbal description of your model,
accompanied by rough sketches, and identify all aspects of the physics that youre neglecting, . . . only then are you
ready to translate your ideas into mathematics (equations and data).
Q.6 The Principle of Pattern Seeking
Students with whom I have worked consistently underestimate how much they know. Facing a new problem,
they tend to approach it in a knowledge vacuum, as though they have never solved a similar problem before.
Dont do this! Dont sell yourself short! You know a lot of stu. Use what you know. Familiarity
breeds success. Both during brainstorming and during the subsequent machinations of problem solving,
seek patterns and analogies between the problem, equation, calculations, or whatever youre working on and
prior work. In almost every chapter of this book, I used this tactic to bypass lots of tedious work and/or
to nd roads to solutions of problems. For instance, by seeking patterns in the mathematical structure
of the reduced radial equation for atomic hydrogen, we realized that we already knew the hard part of the
solutions, the generalized Laguerre functions (Chap. 5). Similarly, when we used the Yukawa function to
model complicated interactions of an electron with other particles in atoms, we realized that the problem
we faced was just another central-force problem (Chap. 4) and therefore we could simply write down a lot of
the answer. When we set up the quantum mechanics of spin (Chap. 6), realizing that spin was just another
angular momentum let us write down, with no derivations whatsoever, almost all the equations we needed.
Rule: Before you start work, try to relate what you know to what you want to know. Seek patterns over and
over as you work. Pay special attention to symmetry properties of the system (see Q.4) and to the mathematical
structure of equations.
Key questions for pattern seeking.
(1) What is the main thing this problem is asking? What do I already know about this thing?
(2) What patterns and relationships do I see in this problem?
(3) How does this problem resemble problems Ive already solved? Once you have found an analogy between
the problem you want to solve and one or more that youve already solved, list the similarities and
dierences between the two problems. Often this step alone will trigger ideas for an ecient way to
solve the problem.
(4) How can I adapt results from other problems to the context of this problem?
Tip. If youre having trouble formulating the problem youve been asked to solve, make up and solve another
problem! That is, devise a simpler problem that involves the same physical principles and quantities. The simpler
the better! Then seek patterns between the simpler problem(s) and the one you want to solve.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Appendix 125
Q.7 The Principle of Back-to-Basics
As you move from strategic brainstorming mode, where youre playing with ways to solve the problem and
to relate it to stu you already know, into tactical problem-solving mode, where youre using practical tools
of mathematics and numerical analysis to implement your ideas, youre likely to nd yourself stuck. (I often
do, and Ive been at this for a long time!) You feel like youre on the right track, but youve hit a road
block. Or your answers look weird or feel wrong. Or youre not sure how to proceed, and the morass of
symbols, explicit equations, numbers, units, and algebra in front of you is too complicated to help you. At
this point, many students give up. They shouldnt. Often, they can gure a way forward by backing up to
basic elements of the problem they do understand.
Your brainstorming notes set you up to eectively use this principle. Heres how: Whenever you feel
yourself getting lost or swamped, stop, look at your list of key elements. Try to navigate back to the
basic physical concepts, quantities, and questions of the problem. This may mean temporarily abandoning
whatever tributary youve gone down, but thats okay; you can always return later.
When you feel yourself losing your way, the rst thing to do is
STOP!
Take a deep breath. Go get a cup of coee (or whatever). When you get back, ask yourself
(1) Am I following one of the approaches I devised while brainstorming, or have I wandered o on a detour?
If so, is it a productive detour, or am I just goong around?
(2) Why did I do the step just before I found myself blocked? Did I have to do this step? How about the
one before this? Can I nd a recent step where I felt more condent about what I was doing? (If so,
return to that step, review your brainstorming notes, and start from that step on a new sheet of paper,
without looking at the pages where you got lost.)
(3) Do I really have to use these complicated equations? Are there more variables in these equations
than appear in the lists I prepared while brainstorming? Must all the functions in these equations be
explicitly written out? Could I see more clearly the structure of the equation if I introduced auxiliary
variables or functions (as in many examples throughout this book)? Can I nd a simpler equation that
mathematically articulates the physical principles Ive identied?
(4) Am I trying to do several things at once?
Tip. The last point is worth emphasizing. Particularly when facing a new type of problem or one that is otherwise
unfamiliar, approach the problem in stages (see above). Choose relevant equations. Relate these equations to one
another. Try to simplify them. Above all, resist plugging explicit expressions into them (see Q.8). Interpret the
equations physically and/or graphically. But do only one of these things at a time; never never do all these things at
the same time.
Tip. As you work through the problem, refer frequently to your brainstorming notes. Update and annotate new
quantities you introduce, equations involving these quantities, and intermediate or auxiliary results. Keep it neat.
You need this map.
Q.8 The Principle of Least Algebra
We now come to the most important principle of eective problem solving. The basic premise is simple:
There is no problem which, however dicult, cannot be made more dicult by using explicit expressions.
The tendency to start problem solving with grossly over-complicated equations is the single most common
reason students get lost, make (lots of) algebraic errors, and, in some cases, ultimately abandon ship. It
follows that you can maximize the eciency and ease of problem solving if you regularly clean up your act.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Index 126
How to avoid algebraic quicksand.
(1) Before you start manipulating equations, make them look as simple as possible. (Sometimes just writing
equations without arguments on the functions in them reveals structures you previously didnt see.)
(2) Once youve formulated the problem, express everything in terms of dimensionless quantities. Doing
so will reduce clutter and expose underlying structures that can show you the way forward (see Ap-
pendix B).
(3) As you work, stay focused on the mathematical structure of the expressions and equations youre using.
(4) As you work, repeatedly try to simplify the mathematical structure of your equations.
(5) At frequent intervals, clear away the algebraic underbrush. Try to eliminate algebraic clutterfor in-
stance, by introducing auxiliary constants, variables, and functionsso you can see clearly the structure
of expressions.
(6) Above all, at each stage of your work on a problem, use the least explicit forms of physical expressions
and equations that you possibly can.
2
Q.9 The Principle of Alert Awareness
If The Principle of Least Algebra is the most valuable problem-solving habit, The Principle of Alert Awareness
runs a close second. Even very experienced physicists (like, say, your author) often nd that they have drifted
unawares into autopilot mode. You know what thats like: activity is taking place, youre getting more and
more tired, but youre no nearer an answer. Worse, nobodys home. Your pencil moves across the page.
Equations follow from other equations. Your ngers click keys that make your computer spew forth numbers.
(The propensity to enter autopilot mode is especially strong when using a computer.) Yet, you dont know
what youre doing, and you arent really aware that you dont know what youre doing.
Rule: Stay alert. Keep your brain engaged, not just your optic nerve and ngers. Dont let the problem lull you
into a mindless stupor.
Tactics to avoid autopilot model. First and foremost, stop. Regularly. Often. Get up. Move around.
Then ask yourself questions like the following,
(1) What am I trying to do (derive, calculate, or whatever)? Why am I trying to do it?
(2) What do these quantities Im working with mean physically?
(3) Is this problem becoming more complicated than I think it should?
(4) Do my answers or intermediate results make sense? Are they of sensible order of magnitude? Are they
dimensionally correct? Do they feel wrong somehow?
(5) Where does what Im trying to do t into my brainstorming plan?
Tip. Your work on a problem isnt through until youve checked dimensions, orders of magnitude, and simple
limitsand anything else you can think of!
2
Details: For instance, if you need to evaluate an expectation value of some function of r for a stationary state of a system with
a central force, dont plug in the explicit form of the spherical harmonic for the state; instead use the symbolic form Y
,m

(, ).
If you plug in the explicit form, all you see is a triple integral that contains complicated algebraic expressions of and . Youre
not likely to immediately recognize that the double angle integral equals 1 because spherical harmonics are normalized.
PrintAppendices Version: 8.14 Printed: August 3, 2010
Index
Version 8.14: August 3, 2010
Levi-Civita symbol, 77
time-independent Schrodinger equation, see TISE
adjoint, 10, 71
adjoint of a matrix, 79
ancillary conditions, 31
angular frequency, 90, 109, 110
angular-momentum coupling, 94
anharmonic molecular potential, 36
anharmonicity, 36
anharmonicity constant, 38
annihilation operator, 117
anti-linear, 70
anti-unitary operator, 70
associative, 79
asymptotic boundary conditions, see boundary conditions, 17
asymptotic limit, 17
asymptotic limits, 16
atomic mass constant, 43
atomic second, 62
atomic unit of action, 54
atomic unit of electric potential, 63
atomic unit of frequency, 93
atomic unit of intensity, 63
atomic unit of permittivity, 54
atomic unit of time, 62, 93
atomic units
signicance of, 61A
attractive potential, 18
average value, see expectation value
Avogadros constant, 90
Balmer equation, 91
basis, 11, 11, 20, 21, 73
basis set, 11, 73
bohr, 61
Bohr magneton, 45
Boltzmann constant, 44
Boltzmanns constant, 90
Born interpretation, 2, 6, 64
generalized, 2
bound state, 2
asymptotic boundary condition, 17
properties of (1D), 18
boundary conditions, 16
asymptotic (1D), 17
bounded at innity, 17
brainstorming, 122
. . .
CAM, 96
causality, 3
characteristic energy, 29, 34
characteristic equation, 81
characteristic length, 29, 34
characteristic length of a 1D SHO, 113
characteristic momentum of a 1D SHO, 119
characteristic time, 29
characteristic value, 29
characteristic values, 39
characteristics value, 29
classical electron radius, 43
classical limit, 14
classical turning point, 23, 23
classically allowed region, 23, 23
classically forbidden region, 23, 23
Clebsch-Gordan coecients, 96
closure, 11, 73
cofactor, 84
collapse of the wave function, 24
commutator, 4, 12, 73
commuting operators
and simultaneous eigenfunctions, 12, 73
and uncertainty relations, 4
compatible, 13
compatible observables, 4, 13, 74
and measurement, 4
and the commutator, 13
complete, 73
complete set
of eigenfunctions, 11, 73
Completeness, 11
conservation law
and constant of the motion, 13
conservation of probability, 15
constant of the motion, 13
continuity requirement, 5, 23
continuum state, 2, 6n
asymptotic boundary condition, 17
continuum stationary state, 6, 107
conversion factor, 47
Copenhagen interpretation, 4n, 24
Correspondence Principle, 14
corresponds to, 91
coupled angular-momentum eigenstates, 98
coupled-angular-momentum eigenstate, 96
creation operator, 117
127
Index 128
. . .
current density, see probability current density
curvature, 23
and the energy, 23
and the number of nodes, 23
of a 1D spatial function, 23
de Broglie wavelength, 23
Debye, 63
decaying, 23
decaying function, see evanescent function
degeneracy
lack of in 1D, 18
degenerate, 10, 72, 77, 83
degree of degeneracy, 77
density matrix, 5
determinant, 84
determinism, 3
dimensional analysis, 26, 35
dimensional quantities, 26
dimensional quantity, 27
dimensionless Hamiltonian, 32
dimensionless Morse potential energy, 37
dimensionless operator, 32
dimensionless quantities, 26
dimensionless quantity, 26, 27
dimensions of a matrix, 79
Dirac delta function
in 3D, 20n
Dirac delta function normalization, 6n, 6, 19, 20, 108
Dirac delta function normalization condition, 19
Dirac notation, 65
direct products, 96
dispersion, see uncertainty, 7
dispersion relation, 8
distribution, 104
duality, 4
dynamical variable, 1
eective atomic number, 38
Ehrenfests Theorem, 14
eigenbasis, 73
eigenfunction, 77
eigenfunction expansion
and continuum states, 21n
derivation of expansion coecients (1D), 21
expansion coecients, 20
eigenstate, 3, 12
eigensystem, 81
eigenvalue, 3, 77
eigenvalue equation, 77
eigenvalues, 81
eigenvector, 81
Einstein-de Broglie relations, 4
electric ux density, 63
electron magentic moment anomaly, 45
electrostatic CGS units, 53
electrostatic unit, 54
energy eigenstate, 12
energy expansion coecients, 22
energy probability amplitude, see probability amplitude, en-
ergy
energy scale factor, 55
energy scale factor for the 1D SHO, 33
ensemble, 2, 24
ensemble average, see expectation value, 7
ensemble measurement, 2
. . .
equilibrium dissocation energy, 38
equilibrium dissociation energy, 37
equilibrium harmonic frequency, 38
erg, 54
esu, 54
evanescent, 23
evanescent function, 23
even function, 18
even parity, 18
expansion coecients, see eigenfunction expansion
expectation value, 7, 9
evaluation in terms of expansion coecients, 22
of momentum, 8
of position, 7
of the energy, 22
experimental uncertainty, 3
rst Bohr radius, 43
Fourier transform, 7, 119
free particle, 19, 19
frequency, 90, 109, 110
full-width-at-half-maximum, 113
function of an operator, 75
fundamental period, 111
Gaussian function, 113
generalized Born interpretation, see Born interpretation, 2
generalized function, 104
Generalized Uncertainty Principle, 3, 13
generalized uncertainty principle, see uncertainty principle, 78
Givens-Householder-Wilkinson, 82
global phase factor, 16
gram, 54
Gram determinant, 84
Gram-Schmidt, 83
Gram-Schmidt orthogonalization, 10, 72
Gramian, 84
ground state, 17, 24
curvature of spatial function, 23
GUP, 13
Hamiltonian, 15
Hamiltonian eigenfunction (1D), 16
Hamiltonian eigenvalue equation, 16
harmonic frequency, 35
Hartree, 91
hartree, 55
Heaviside step function, 105
Heisenberg uncertainty principle, see uncertainty principle, 3,
78
Hermite dierential equation, 114
Hermite polynomial, 113
Hermitian, 66
Hermitian conjugate, 87
Hermitian operator, 71
Hermiticity
and the adjoint, 10
consequences of, 10
in 1D, 10
Hertz, 90
Hilbert space, 64
Hulthen potential, 38
HUP, see uncertainty principle, Heisenberg
incompatible, 13
incompatible observables, 4, 13, 74
PrintAppendices Version: 8.14 Printed: August 3, 2010
Index 129
. . .
and measurement, 4
and uncertainty relations, 13
inner product, 11n, 11, 72n
integral of three spherical harmonics, 100
integrated probability density, 6, 6
intensity, 63
internuclear separation, 36
inversion
in 1D, 18
Jacobi method, 82
Jacobian, 30
Kayser, 62
ket, 64
kilocalorie, 48
kilocalorie per mole, 48
Kronecker delta function, 11, 21, 73
ladder operators, 117
length scale factor, 55
length scale factor for the 1D SHO, 33
linear, 12
linear algebra, 11n, 72n
linear independence, 18n
linear momentum, 7
linear operators, see operators, linear
linearity
of operators, see operators, linear
of the TDSE, 5
linearly dependent, 18
linearly independent, 18
local wavelength, 23
lowering operator, 117
magnetic eld, 52
magnetic ux density, 52, 63
magnetic induction, 52
magnetic intensity, 52
magnetization, 52
mathematical structure, 8, 27, 124
matrix diagonalization, 82
matrix eigenvalue equation, 81
matrix element, 10, 65, 66
matrix product, 80
measurement, 2
method of eigenfunction expansion, 2, 11, see eigenfunction
expansion, 20, 67
method of eigenfunction expansion, 96
micrometer, 89
micron, 89
minor, 84
mixed state, 5
model, 123
molar gas constant, 90
mole, 43
molecular vibration, 36
momentum
expectation value, see expectation value
linear, see linear momentum
uncertainty, see uncertainty
momentum amplitude function, 7
momentum expectation value, 8
momentum operator, 30
momentum probability amplitude, 7, 8
momentum probability density, 8, 8
. . .
momentum representation, 8
momentum space, 8
momentum uncertainty, 8
Morse potential, 36
Morse range paramter, 38
node, 24
nodes, 24n
and bound states, 23
and curvature of the spatial function, 23
non-degenerate, 18
non-degenerate eigenvalue, 18
non-singular, 80
non-stationary state, 20
norm, 11n, 11, 72n
normalization
in terms of expansion coecients, 22
of a 1D wave function, 6
of a non-stationary state wave function (1D), 22
normalization condition, 6
normalize the wave function, 6
normalized Gaussian, 105
nuclear magneton, 45
number operator, 117
objectivity, 3
observable, 1, 9, 70
observables,compatible, see compatible observables
observables,incompatible, see incompatible observables
odd parity, 18
odd function, 18
operator, 1, 70
operators
anti-linear, 10n
Hermitian, 10
linear, 12, 70
manipulation rules, 10, 74
postulate, 9
restrictions on, 9
self-adjoint, 10n, 66n
orbital magnetic moment, 16
order, 84
orthogonal, 10, 72
orthogonal transformation, 82
orthogonality, 10, 21, 72
orthonormality
in 1D, 11, 72
orthonormality condition, 11, 72
oscillatory, 23
oscillatory function, 23
overlap integral, 65, 72
parity
denite, 18
parity operator, 18
in 1D, 18
period, 110
periodic, 111
periodic boundary conditions, 111
permeability of the vacuum, 53
permittivity of the vacuum, 53
phase factor, see global phase factor, 16
physical admissibility, see wave function, physical admissibil-
ity, 5
physical equation, 27
physical quantities, 27
PrintAppendices Version: 8.14 Printed: August 3, 2010
Index 130
. . .
pointwise conservation of position probability, 15
position probability amplitude, see wave function, 6
position probability density, 6
power of a matrix, 80
primary dimensions, 40
principal quantum number, 17
Principle of Complementarity, 4
Principle of Superposition, 5, 12, 70
probability amplitude, 2, 5
energy, 22
for position (1D), 6
probability current density, 15
probability density, 6
probability distribution, 2
probability, conservation of, see conservation of probability
problem solving strategies, 2324
product of two spherical harmonics, 100
projection operator, 67
projection quantum number, 94
properties of kets and bras, 68
pulse function, 105
pure momentum state, 19
physical properties of, 20
pure momentum states, 19
pure state, 5
quantization, 3
of energy, 17
quantum mechanical angular momentum, 94
quantum state, 1
quantum statistical mechanics, 5
quantum-mechanical angular momenta, 95
quantum-mechanical ensemble, 2
radius of curvature, 23
raising operator, 117
range parameter, 36
rationalized Plancks constant, 92
reciprocal wavelength, 111
reduced Component wavelength, 43
reduced mass, 36
relationship between the length and energy scale factors, 32
resolution of the identity, 67
Rydberg, 91
Rydberg constant, 43, 57, 91
Rydberg unit, 57
Rydberg unit of energy, 43, 58
scalar product, 11n, 11, 72n
scalar products, 68
scale factor, 27
scale transformation, 38
scattering state, see continuum state, 2
screened Coulomb potential, 38
screening constant, 38
second quantization, 117
secondary dimensions, 40
secular equation, 81
self-adjoint, 66
self-adjoint operator, 10
separable function, 16
separable wave function, see wave function
separation of variables, 16
in the 1D TDSE, 16
sharp, 2, 3, 12, 16, 74
similar, 82
. . .
similarity transformation, 82
simultaneous eigenfunction, 12, 13
simultaneous eigenfunctions, 12
and commuting operators, 12, 73
simultaneous eigenstate, see simultaneous eigenfunctions
single-valued requirement, 5
skew-symmetric matrix, 86
smoothly varying requirement, 5
spatial function, see Hamiltonian eigenfunction, 16
spectrum, 3, 81
spur, 87
square matrix, 81
standard deviation, see uncertainty, 7
standing wave, 110
statcoulomb, 54
state vector, 64
stationarity, 16
stationary state, 12, 16
statistical properties, 7
subensemble, 25
subensembles, 25
symmetric potential
and parity in 1D, 18
symmetry
using to simplify 1D integrals, 23
system-dependent parameter, 38
TDSE, see TDSE
postulate, 15
the method of eigenfunction expansion, 7
time evolution operator, 64
time reversal operator, 10
TISE
in 1D, 16
integral form, 16
total angular momentum quantum number, 94
transmission coecient, 17
transpose of a matrix, 79
travelling wave, 19
triangle condition, 100
triangle rule, 98
truncation, 21
two-fold degenerate, 18, 83
UCAM, 96
uncertainty, 3, 7, 9
evaluation in terms of expansion coecients, 22
experimental, 3
of momentum, 8
of position, 7
of the energy, 22
uncertainty principle
generalized, 3, 13
Heisenberg, 3, 13
uncertainty product, 3, 13
uncertainty relation, 3
and commuting operators, 4
and incompatible observables, 13
uncoupled-angular-momentum eigenstate, 96
unied atomic mass unit, 92
unit matrix, 80
unit-step function, 105
unitary transformation, 82
units of wavelength, 89
variational method, 17
PrintAppendices Version: 8.14 Printed: August 3, 2010
Index 131
. . .
volume element
in 3D, 6
wave function, 2, 5
as a probability amplitude, 6
physical admissibility, 5
postulate, 5
separable (1D), 16
wave number, 89, 109, 111
wave packet
of a free particle, 8, 19, 20
wave vector, 19, 20
wave-particle duality, see duality, 61
wavelength, 110
wavelength of a transition, 91
wavenumber, 7, 7
well depth, 36
Wigner 3j symbols, 96
Yukawa potential, 38
zero matrix, 80
PrintAppendices Version: 8.14 Printed: August 3, 2010
Bibliography
Version 8.14: August 3, 2010
F. S. Acton. Numerical Methods that (Usually) Work. Harper and Row, New York, 1970.
F. S. Acton. Real Computing made Real. Princeton University Press, P)rinceton, New Jersey, 1996.
M. B. Allen and E. I. Isaacson. Numerical Analysis for Applied Science. Wiley, New York, 1998.
H. M. Antia. Numerical Methods for Scientists and Engineers. Birkhauser, Berlin, second edition, 2002.
G. I. Barenblatt. Dimensional Analysis. Gordon and Breach Science Publishers, New York, 1987.
G. I. Barenblatt. Scaling. Cambridge University Press, New York, 2003.
D. S. Bernstein. Matrix mathematics: theory, facts, and formulas with applications to linear systems theory. Princeton, NJ,
New York, 2005.
B. H. Bransden and C. J. Joachain. Quantum mechanics. Prentice-Hall, New York, second edition, 2000.
C. Cohen-Tannoudji, B. Diu, and F. Laloe. Quantum Mechanics. Wiley-Interscience, New York, 1977.
J. T. Cushing. Applied Analytical Mathematics for Physical Scientists. John Wiley & Sons, New York, 1975.
P. A. M. Dirac. The Principles of Quantum Mechanics. Oxford, Clarendon Press, 4th edition, 1958.
G. E. Forsythe and C. B. Moler. Computer Solution of Linear Algebraic Systems. Prentice-Hall, Englewood Clis, New Jersey,
1967.
F. B. Hildebrand. Methods of Applied Mathematics. Prentice-Hall, Englewood Clis, NJ, second edition, 1965.
S. Howison. Practical Applied Mathematics: Modelling, Analysis, Approximation. Cambridge University Press, New York,
2005.
A. Jennings. Matrix Computation for Engineers and Scientists. Wiley, New York, 1977.
D. Kincaid and W. Cheney. Numerical Analysis: Mathematics of Scientic Computing. Brooke/Cole, Pacic Grove, CA, third
edition edition, 2002.
D. H. Menzel. Mathematical Physics. Dover, New York, 1961.
E. Merzbacher. Quantum mechanics. John Wiley & Sons, New York, third edition, 1998.
A. Messiah. Quantum mechanics. North-Holland, Amsterdam, 1966.
M. A. Morrison. Understanding Quantum Physics: A Users Manual, volume I. Prentice-Hall, Englewood Clis, NJ, 1990.
M. A. Morrison and W. Sun. How to calculate rotational and vibrational cross sections for low-energy electron scattering
from diatomic molecules using close-coupling techniques. In W. Huo and F. Gianturco, editors, Computational Methods for
Electron-Molecule Collisions, chapter 6, pages 131190. Plenum, New York, 1995.
W. H. Press, B. P. Flannery, S. A. Teukolsky, and W. T. Vetterling. Numerical Recipes in FORTRAN 77: The Art of Scientic
Computing, volume 1 of Numerical Recipies. Cambridge University Press, Cambridge, second edition, 1992.
132
Bibliography 133
W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery. Numerical Recipes in FORTRAN 90: The Art of Parallel
Scientic Computing, volume 2 of Numerical Recipies. Cambridge University Press, Cambridge, 1996.
G. Sewell. Compuational Methods of Linear Algebra. Pure and Applied Mathematics. Wiley-Interscience, New York, second
edition, 2005.
R. Sneider. A Guided Tour of Mathematical Methods for the Physical Sciences. Cambridge University Press, New York, second
edition, 2004.
B. Thaller. Visual Quantum Mechanics. Springer, New York, 2000.
S. S. M. Wong. Computational Methods in Physics and Engineering. World Scientic, New Jersey, second edition, 1997.
PrintAppendices Version: 8.14 Printed: August 3, 2010

Das könnte Ihnen auch gefallen