LECTURE NOTES ON QUANTUM MECHANICS

PHY 851/852 – 2020/2021/2022

Scott Pratt
Department of Physics and Astronomy
Michigan State University
 FORWARD
These notes are for the two-semester graduate level quantum mechanics class taught at Michi-
gan State University. Although they are more terse than a typical text book, they do cover all
the material used in PHY 851/852. The notes presume a familiarity with basic undergraduate
concepts in quantum mechanics.
Anybody is welcome to use the notes to their heart’s content, though the text should be treated
with the usual academic respect when it comes to copying material. If anyone is interested in
the LATEX source files, they should contact me (prattsc@msu.edu). Solutions to the end-of-chapter
problems will also be provided on the course web site, https://people.nscl.msu.edu/~pratt
/phy851. Please beware that this is a web manuscript, and is thus a living document and subject
to change at any time.
Contents
1 States and Operators 2
1.1 States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Hermitian Conjugate: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Unitary Operators and Transformations . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Density Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Rotations of Two-Component Spin-Half Systems . . . . . . . . . . . . . . . . . . 10
1.7 Polarization of Photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.8 Evolution in Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.9 Evolution of Two-component Systems . . . . . . . . . . . . . . . . . . . . . . . 16
1.10 Heisenberg and Schrödinger representations . . . . . . . . . . . . . . . . . . . . 18
1.11 Uncertainty Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.12 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2 Coordinate and Momentum Space 27

2.1 Schrödinger’s Wave Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 Momentum States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3 Local Charge Conservation and the Equation of Continuity . . . . . . . . . . . . 33
2.4 Potential Problems in One Dimension . . . . . . . . . . . . . . . . . . . . . . . . 34
2.5 Wave Packets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.6 The Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.7 Propagators, Green’s Functions and Integral Equations . . . . . . . . . . . . . . . 46
2.8 Phase Space Density and Wigner Functions . . . . . . . . . . . . . . . . . . . . . 47
2.9 Numerically Solving Schrödinger’s Equation . . . . . . . . . . . . . . . . . . . . 49
2.10 Combining Coordinate Space and Spin . . . . . . . . . . . . . . . . . . . . . . . 50
2.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3 Charged Particles in Electromagnetic Fields 54

3.1 Vector and “Scalar” Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Gauge Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.3 Interaction with Uniform Magnetic Fields . . . . . . . . . . . . . . . . . . . . . . 57
3.4 Motion with Uniform Electric and Magnetic Fields . . . . . . . . . . . . . . . . . 58
3.5 Path Integrals and the Aharanov-Bohm Effect . . . . . . . . . . . . . . . . . . . 59
3.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4 Angular Momentum and Central Potentials 64
4.1 The Baker-Campbell-Hausdorff Relation . . . . . . . . . . . . . . . . . . . . . . 64
4.2 Three-Dimensional Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3 A Greatly Abbreviated Guide to Group Theory . . . . . . . . . . . . . . . . . . . 68
4.4 ⃗ 2 Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The |L| 69
4.5 Raising and Lowering Operators for Angular Momentum . . . . . . . . . . . . . 70
4.6 Spherical Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.7 Wigner D Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.8 Separating Relative and Center-of-Mass Coordinates . . . . . . . . . . . . . . . . 75
4.9 Coupled Oscillators in Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . 77
4.10 Solving the Radial Wave Equation for Spherically Symmetric Cases . . . . . . . . 79
4.11 Spherical Harmonic Oscillator – Spherical Basis . . . . . . . . . . . . . . . . . . 81
4.12 The Hydrogen Atom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.13 Adding Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.14 Clebsch-Gordan Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.15 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

5 Symmetries and Conservation Laws 95

5.1 Continuous and Discrete Symmetries . . . . . . . . . . . . . . . . . . . . . . . . 95
5.2 Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.3 Time Reversal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.4 Time Reversal and Angular Momentum . . . . . . . . . . . . . . . . . . . . . . 100
5.5 A Particle in a Periodic Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.6 Object with Discrete Rotational Symmetry . . . . . . . . . . . . . . . . . . . . . 103
5.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

6 Approximation Methods 107

6.1 The WKB Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.2 Variational Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.3 The Sudden Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.4 Stationary-State Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . 113
6.5 Degenerate-State Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . 116
6.6 The Interaction Representation (Picture) . . . . . . . . . . . . . . . . . . . . . . 117
6.7 Time-Dependent Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . 118
6.8 Fermi’s Golden Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.9 Harmonic Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
 6.10 An Exactly Solvable Model of a Time-Dependent Interaction . . . . . . . . . . . . 123
6.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

7 Perturbative Approaches to Scattering 129

7.1 Cross Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.2 The Born approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.3 Structure Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.4 Form Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.5 Higher Order Expansions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7.6 Propagators (Time-Ordered Evolution Operators) . . . . . . . . . . . . . . . . . 138
7.7 The Fourier Transform of the Propagator, G̃(ω) . . . . . . . . . . . . . . . . . . . 139
7.8 Propagating to the Same State . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.9 Self-Energy of a Particle in Medium . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.10 Resonant Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.11 Resonant Scattering, Simplified . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.12 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

8 Scattering at Lower Energies 150

8.1 Partial Waves and Phase Shifts . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.2 The Optical Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
8.3 Calculating Phase Shifts Numerically . . . . . . . . . . . . . . . . . . . . . . . . 157
8.4 The Low Energy Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
8.5 Levinson’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
8.6 Effective Range Theory for ℓ = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . 163
8.7 Coulomb Waves and Gamow Factors . . . . . . . . . . . . . . . . . . . . . . . . 164
8.8 Classical Interpretation of the Scattering Wave Function . . . . . . . . . . . . . . 166
8.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

9 Second Quantization and Radiation 172

9.1 Quantum Fields and Second Quantization . . . . . . . . . . . . . . . . . . . . . 172
9.2 Field Operators: Creation and Destruction Operators in Coordinate Space . . . . 174
9.3 Energy and the Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
9.4 Interaction with an External Potential . . . . . . . . . . . . . . . . . . . . . . . . 177
9.5 Electromagnetic Decays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
9.6 The Dipole Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
9.7 Magnetic Dipole and Electric Quadrupole Radiation . . . . . . . . . . . . . . . . 186
9.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
10 Advanced Topics in Angular Momentum 190
10.1 Isospin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
10.2 Combining Three Angular Momenta . . . . . . . . . . . . . . . . . . . . . . . . 192
10.3 Notation: Conventions for Clebsch-Gordan Coefficients and Wigner 3j or 6j Symbols193
10.4 Irreducible Tensor Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
10.5 Wigner D matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
10.6 The Two-Dimensional Analogy of the Wigner-Eckart Theorem . . . . . . . . . . . 200
10.7 The Wigner-Eckart Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
10.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

11 Fermions 210
11.1 The Spin Statistics Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
11.2 The Pauli-Exclusion Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
11.3 Fermi Gases and Neutron Stars . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
11.4 Multi-Particle Symmetrization and Slater Determinants . . . . . . . . . . . . . . 218
11.5 Fermi Creation and Destruction Operators . . . . . . . . . . . . . . . . . . . . . 220
11.6 Fermionic Field Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
11.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226

12 Non-Perturbative Approaches for Many-Fermion Systems 228

12.1 Interacting Fermi Systems – Two and Four-Point Interactions . . . . . . . . . . . 228
12.2 The Thomas-Fermi Approximation . . . . . . . . . . . . . . . . . . . . . . . . . 230
12.3 The Hartree-Fock Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . 231
12.4 Hartree-Fock for Atoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
12.5 The Periodic Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
12.6 Configuration Splitting and Hund’s Rules . . . . . . . . . . . . . . . . . . . . . 234
⃗ and Total S
12.7 Constructing States According to Total L ⃗ . . . . . . . . . . . . . . . . 235
12.8 Permutation Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
12.9 Zeeman Effect and the Landé g Factor . . . . . . . . . . . . . . . . . . . . . . . . 237
12.10Molecules and the Adiabatic Approximation . . . . . . . . . . . . . . . . . . . . 239
12.11The Hydrogen Molecule and Pairing . . . . . . . . . . . . . . . . . . . . . . . . 242
12.12Superconductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
12.13Cooper Pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
12.14The Gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
12.15Landau Levels and the Integer Quantum Hall Effect . . . . . . . . . . . . . . . . 247
12.16Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
PHY 851/852

13 Relativistic Quantum Mechanics 253

13.1 The Klein-Gordon Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
13.2 Current Conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
13.3 Coupling to the Electromagnetic Field . . . . . . . . . . . . . . . . . . . . . . . . 256
13.4 The Dirac Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
13.5 Conserved Quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
13.6 Solutions for Free Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
13.7 Non-Relativistic Limit and the g Factor of the Electron . . . . . . . . . . . . . . . 264
13.8 The Spin-Orbit Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
13.9 More Notation: γ Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
13.10Dirac Field Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
13.11The Dirac Equation and the Mass Term . . . . . . . . . . . . . . . . . . . . . . . 270
13.12Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271

14 States Without Conserved Particle Number 274

14.1 Bogoliubov Operators and Cooper Pairs . . . . . . . . . . . . . . . . . . . . . . 274
14.2 Coherent States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
14.3 Emission Via Coupling to a Classical Current . . . . . . . . . . . . . . . . . . . . 277
14.4 Completeness Relations for Coherent States . . . . . . . . . . . . . . . . . . . . . 278
14.5 Path Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
14.6 Coherent Fields in Coordinate Space – Ginzburg Landau Theory of Superconduc-
tivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
14.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282

1
PHY 851/852 1 STATES AND OPERATORS

1 States and Operators

1.1 States
Quantum mechanics consists of states and operators. For any finite system, states are discrete
and can be assigned labels. The discrete nature of states is what is what is meant by the word
“quantized”. For some systems the number of such discrete states is finite, e.g. a particle in
either spin-up or spin-down state, whereas in other systems the number of states are infinite,
e.g. levels of a harmonic oscillator. In this chapter we concentrate on those systems where the
number of states are finite, and proceed to the latter case in the next chapter.
Any physical state can be assigned a label, in this case ψ. This state can always be expressed as
a linear combination of basis states, |i⟩, in an orthonormal basis. These states can be expressed
as vectors in a vector space of size n,
     
1 0 0
 0   1   0 
     
|1⟩ = 1̂ =  0  , |2⟩ = 2̂ =  0  , · · · |n⟩ = n̂ =  0  . (1.1)
     
 ..   ..   .. 
 .   .   . 
0 0 1

By inspection, the basis is orthonormal,

⟨i|j⟩ ≡ î∗ · ĵ = δij . (1.2)

In quantum mechanics the vector algebra is complex, and the adjoint vector ⟨ψ| is represented
by the complex transpose. Any orthonormal set of vectors can be expressed in this basis. The
notational choice of using ⟨ψ| to refer to the adjoint vector and |ψ⟩ to denote the vector is known
as bras and kets respectively (a take on the word “bracket”), and is known as Dirac notation,
though it derives from Hermann Grassmann’s work, https://en.wikipedia.org/wiki/Herman
n_Grassmann, one hundred years earlier.
Whenever the notation has a bra followed by a ket, it implies the dot product between the two
vectors, i.e.
X
⟨v|u⟩ = vi∗ ui , (1.3)
i

and is sometimes referred to as an inner product. A sum over vector indices is not implied if the
ket precedes the bra, and
|u⟩⟨v| = ui vj∗ (1.4)
is known as an outer product. However, if one sees the notation,
|α⟩⟨α|, (1.5)
P
a sum over the vectors α is implied (unless otherwise stated), as the summation α is implicit.
Ultimately, for any physical observable, all bras and kets will be closed with a bra being the first

2
PHY 851/852 1 STATES AND OPERATORS

to the left and a ket being the last to the right, with all vector indices being summed over within
each bra-ket factor.
One can form new states by taking linear combinations of basis states |i⟩. For example,
X
|ψ⟩ = aψ,i |i⟩
i
X
⟨ψ| = a∗ψ,i ⟨i|,
i

where ai,ψ is a complex number. By taking the overlap of |ψ⟩ with ⟨i|, one sees that
ai,ψ = ⟨i|Ψ⟩. (1.6)
If the basis states |i⟩ are orthonormal and complete, one can state that
X
|i⟩⟨i| = I. (1.7)
i

This is known as the completeness relation. It is easy to see that completeness relation holds for
the simple basis in Eq. (1.1). Any other orthonormal basis can be related to this simple basis by
a basis transformation, which is an example of a unitary transformation. A basis transformation
relates a set of basis states |i⟩ to a set of basis states |j⟩. One can express the state |i⟩ in terms of
the basis j,
X
|i⟩ = |j⟩⟨j|i⟩. (1.8)
j

If this is to work for any state i⟩, the completeness relation in Eq. (1.7) must hold.
Furthermore, by taking the overlap of ψ with itself (the norm),
X
⟨ψ|ψ⟩ = ⟨ψ|i⟩⟨i|ψ⟩ (1.9)
i
X X
= |⟨ψ|i⟩|2 = |aψ,i |2 .
i i

If the state |ψ⟩ is normalized, i.e. ⟨ψ|ψ⟩ = 1 = i |⟨i|ψ⟩|2 = i |aψ,i |2 , the squared element,
P P
|⟨i|ψ⟩|2 , can be interpreted as a probability because each term is positive and because the sum
yields unity. Thus, if one prepares a state |ψ⟩, the chance one finds that one is in a state |i⟩ is
|⟨i|ψ⟩|2 . Up to this last sentence, where the connection to probability is stated, the discussion
has been one of standard linear algebra and notation. Of course, once one is in the state i, you
know that the system is in the state i, and any knowledge of being in the original state ψ is lost.
Associating the squared overlap as a probability is the profound intellectual jump that makes
quantum mechanics physics. Most of the expressions applied in this course are derived, almost
inexorably, from this conceptual leap combined with arguments about symmetry and the need
to reproduce classical mechanics in some limit.
Equation (1.9) also suggests the completeness relation,
X
|i⟩⟨i| = I. (1.10)
i

3
PHY 851/852 1 STATES AND OPERATORS

This is easy to see with the simple basis states in Eq. (1.1). In that case
   
1 0 ··· 0 0 0 ··· 0
 0 0 ··· 0   0 1 ··· 0 
|1⟩⟨1| =   , |2⟩⟨2| =  , ··· (1.11)
   
.. ..
 .   . 
0 ··· , 0 0 ··· , 0

The completeness relation will also work for anyP set of basis states, because the basis can always
be transformed to the simple basis, and because i |i⟩⟨i| = I and the unit matrix is unchanged
by a transformation of basis, a.k.a. a unitary transformation. This will be demonstrated ahead.
As mentioned previously, all states can be expressed as vectors. For instance, if a basis has two
states, the two states might be defined in terms of vectors as,
   
1 0
| ↑⟩ = , | ↓⟩ = (1.12)
0 1

The choice of these two vectors is arbitrary. As long as the two vectors are orthonormal, any
state can be represented as a linear sum of the two states. Any set of orthonormal states that
spans the space is known as a “basis”, with each orthonormal state being a basis state.
Example 1.1: Probabilities and Overlaps
Consider the state
1
 
1
|ψ⟩ = .
Z 1/2 i

a) Find Z so that the state is normalized.

b) What is the probability that |ψ⟩ would be measured in the | ↓⟩ state?

Solution:
a) Squaring |ψ⟩,

1
 

1
⟨ψ|ψ⟩ = 1 −i
Z i
2
= ,
Z
Z = 2.

b) The probability is

P (↓) = |⟨↓ |ψ⟩|2

1 
1 2
  
=  0 1 
2 i 
1 1
= (−i)(i) = .
2 2

4
PHY 851/852 1 STATES AND OPERATORS

1.2 Operators
Operators operate on vectors and return another vector. All operators may be described in terms
of bras and kets as a linear combination of outer products,
X
A= aij |i⟩⟨j|. (1.13)
ij

Just as any state can be equivalently expressed as a vector, any operator can be represented by a
matrix, in this case aij . Just as a state’s expression in terms of a vector depends on the basis, so
does the expression of an operator in terms of a matrix. Knowing the coeficients aij is sufficient
to define the matrix. If the basis is defined by
   
1 0
 0   1 
|1⟩ =   , |2⟩ =   , · · · , (1.14)
.. ..
. .

the matix and the coefficients are synonymous Aij = aij . Otherwise, one can express the states
|i⟩ and |j⟩ as vectors, v(i)m and v(j)n . In that case
X
Amn = aij v(i)m v ∗ (j)n , (1.15)
ij

where v(i)m is the mth component of the basis vector v(i).

1.3 Hermitian Conjugate:

Consider a matrix element, ⟨ϕ|A|ψ⟩. Expressing the operators as matrices and the states as
vectors, then taking the complex conjugate one sees that,

⟨ϕ|A|ψ⟩ = ϕ∗i aij ψj (1.16)

∗
⟨ϕ|A|ψ⟩∗ = ϕ∗i aij ψj
= ψj∗ a∗ij ϕj .

Thus, if one wishes to define an operator A† such that

⟨ψ|A† |ϕ⟩ = ⟨ϕ|A|ψ⟩∗ , (1.17)

for any states |ϕ⟩ and |ψ⟩, the operator A† must be represented by the complex-conjugate of the
matrix that represents A, then transposed.
X
A† = (a† )ij |i⟩⟨j|, (1.18)
ij

(a )ij = a∗ji
†

The operator A† is known as the Hermitian conjugate of A. A Hermitian operator is one that
obeys the relation,
A = A† . (1.19)

5
PHY 851/852 1 STATES AND OPERATORS

The eigenvalues of a Hermitian operator are real, i.e. one can always find a change of basis
that diagonalizes A, with the elements all being real. To see this, assume the operator A has
eigenvalues ai and eigenvectors |i⟩.

A|i⟩ = ai |i⟩, (1.20)

⟨i|A† = a∗i ⟨i|,
⟨i|(A − A† )|i⟩ = ai − a∗i .

If A is Hermitian, A = A† , and ai must equal a∗i , so the eigenvalues are real.

Hermitian operators are important for two reasons,

1. They yield real numbers as expectations, ⟨ψ|K|ψ⟩∗ = ⟨ψ|K|ψ⟩. Because all physical
observables are real, they need to be represented by Hermitian operators. An example is
the Hamiltonian H which represents the energy of a system.
2. They can be used to generate unitary transformations, e.g. eiKθ . An example is the time
evolution operator e−iHt . This will be discussed briefly in the next sub-section, but plays
a much larger role when we have a general discussion of symmetries.

1.4 Unitary Operators and Transformations

A class of operators that often play an important physical role is defined by those that corre-
spond to a change of basis, while keeping the new basis orthonormal. Examples are rotations,
translations and reflections. A common unitary operator is the evolution operator, U = e−iHt/ℏ ,
which enacts translations in time. Consider two sets of basis states |i⟩ and i′ ⟩, related by a trans-
formation operator R,
X
|i′ ⟩ = Ri′ i |i⟩, (1.21)
i
X
′
⟨i | = Ri∗′ i ⟨i|
i
X †
= ⟨i|Rii ′.

Here, the set of states |i′ ⟩ will refer to a new basis. The coefficients Ri′ i must be constructed to
retain the orthormality properties,

⟨i′ |j ′ ⟩ = δi′ j ′ (1.22)

X †
= Rii′ Rj ′ j ⟨i|j⟩
ij
X †
= Rj ′ i Rii ′

i
= (RR† )j ′ i′ .

Here, the Hermitian conjugate R† was defined in Eq. (1.18). In simple words, the Hermitian
conjugate of a unitary matrix is its inverse, i.e., RR† = I, or equivalently, R−1 = R† . An

6
PHY 851/852 1 STATES AND OPERATORS

operator that satisfies this condition is called unitary. The word unitary follows from the fact
that a state U |ψ⟩ has the same norm as |ψ⟩, implying that the net probability is unchanged by
the unitary transformation U .
Hermitian operators are often used to generate unitary transformations,

U = e−iKθ . (1.23)

It is easy to see that such an operator is unitary if K = K† ,

†θ
U U † = e−iKθ eiK (1.24)
= e−iKθ eiKθ = e−i(K−K)θ
= I.

If an operator U is unitary, it represents unitary transformations, under which a vector |ψ⟩

transforms as

|ψ ′ ⟩ = U |ψ⟩, (1.25)

and an operator A transform as

A′ = U AU −1 . (1.26)

The latter definition allows the vector |ϕ⟩ = A|ψ⟩ to transform as

|ϕ′ ⟩ = (U AU −1 )U |ψ⟩ (1.27)

= U (A|ψ⟩) = U |ϕ⟩.

Summarizing, for a unitary transformation U ,

U † = U −1 , (1.28)
|ψ ′ ⟩ = U |ψ⟩,
⟨ψ ′ | = ⟨ψ|U † ,
A′ = U AU † ,
< ψ ′ |A′ |ψ ′ ⟩ = ⟨ψ|A|ψ⟩.

This last line emphasizes the fact that if you transform BOTH the operators and states, the matrix
element is unchanged. However, in some cases you transform the states, leaving the operator
unchanged, or you might transformed the operator and leave the state unchanged. This leads
to new matrix elements. It can be painful to keep track of which objects are being transformed.
This is akin to performing a rotation, where rotating an object by ϕ or rotating the coordinate
system by −ϕ has the same effect. Also, if you rotate both the coordinate system and the object
by the same angle, nothing changes.

1.5 Density Matrices

First, we stop and consider what forms of matrix elements might be considered as an observable.
Observables must be real and independent of the basis, i.e. all unitary transformations that act

7
PHY 851/852 1 STATES AND OPERATORS

both on the bras and kets and on the operators should leave the observable unchanged. All
observables can be expressed as either the expectation of a Hermitian operator,

⟨ψ|K|ψ⟩, (1.29)

or as the squared overlap of two states,

|⟨ϕ|U |ψ⟩|2 . (1.30)

Instead of describing a state ψ by a vector, one could describe it by a density matrix,

ρψ = |ψ⟩⟨ψ|, (1.31)

or as a matrix

(ρψ )ij = ψi ψj∗ . (1.32)

By inspection, one can see that the density matrix is Hermitian. From the definition of the density
matrix, one can write the state |ψ⟩ as a vector and the operator A as a matrix, and that
X
⟨ψ|A|ψ⟩ = ψi∗ Aij ψj (1.33)
i,j

= Tr ρψ A,

and that

|⟨ϕ|A|ψ⟩|2 =< ψ|A† |ϕ⟩⟨ϕ|A|ψ⟩ (1.34)

= Tr ρψ A† ρϕ A.

Thus, density matrices are sufficient to generate all observables. The trace of any product of op-
erators, or matrices, is invariant to unitary transformations so the answer should be independent
of basis, as long as the density matrices and the operator A are all transformed.
Example 1.2: Density Matrix for a Two-Component System
Consider the state
 
cos θ
|ψ⟩ = .
eiϕ sin θ

• Calculate the density matrix and find the trace. Then find the eigenvalues of ρψ .
Solution:
Write down the outer product of the state,
 
cos θ
cos θ e−iϕ sin θ


ρψ = iϕ
e sin θ
e−iϕ sin θ cos θ
 
cos2 θ
= .
eiϕ sin θ cos θ sin2 θ

8
PHY 851/852 1 STATES AND OPERATORS

The trace is cos2 θ + sin2 θ = 1. The eigenvalues, λ, are found by

|ρψ − λI| = 0
 
 cos2 θ − λ e−iϕ sin θ cos θ 
=  iϕ ,
e sin θ cos θ sin2 −λ 
0 = λ2 + cos2 θ sin2 θ − λ(cos2 + sin2 θ) − sin2 θ cos2 θ,
λ2 − λ = 0.

The eigenvalues are zero and 1, as expected for a pure state.

• What is ρψ if one averages over all phases ϕ? Did this change the trace? What are the
eigenvalues?
Solution:
The off-diagonal terms average to zero and
 
cos2 θ 0
ρψ =
0 sin2 θ

The trace is still unity. The eigenvalues are cos2 θ and sin2 θ.

For a state |ψ⟩, one can always choose a basis where

 
1
 0 
|ψ⟩ =  ..  . (1.35)
 
 . 
0

In this basis ρψ is diagonalized and has the form,

 
1 0 ··· 0
 0 0 ··· 0 
ρψ =  . (1.36)
 
..
 . 
0 0 ··· 0

A density matrix that can be diagonalized in such a way is known as a pure state. A projection
operator is a density matrix corresponding to a pure state. The reason that it is often called a
projection operator is that

Pψ2 = Pψ . (1.37)

Further, for a pure state,

Tr Pψ = 1. (1.38)

Projection operators can also play the role of a filter, e.g. a polarization filter. If one has a state
that is a linear combination of various states, a|ψ⟩ + b|ϕ⟩, the projection operator Pψ acting on
such a state returns a|ψ⟩.

9
PHY 851/852 1 STATES AND OPERATORS

However, density matrices can be more general. If one has a density matrix which is an incoher-
ent sum over several states, the resulting diagonalized density matrix could have more than one
non-zero element, though the trace would have to remain equal to unity. Thus, density matrices
can be used to express non-pure states such as unpolarized beams. For instance, one can define
the following density matrix,
1 1
ρ = |1⟩⟨1| + |2⟩⟨2|. (1.39)
2 2
As with the pure state, the trace remains at unity, but after being diagonalized the strength is
spread along the diagonal. This density matrix describes being in state |1⟩ 50% of the time and
in state |2⟩ 50% of the time. A way to write a state that is described by this density matrix, would
be
1 1
√ |1⟩ + eiϕ √ |2⟩, (1.40)
2 2
with ϕ being treated as a random phase to justify ignoring the off-diagonal terms in the density
matrix. When performing calculations with such a state, one would ignore all terms in the
density matrix with a leftover phase, eiϕ , to account for the randomness of the phase. Note that
a density matrix ρ for an impure state is not a projection operator, i.e. ρ2 ̸= ρ.
Density matrices play an essential part in thermodynamics. In that case, one considers incoher-
ent sums over many states weighted by the energy. If one is in a basis where the Hamiltonian is
diagonalized, the density matrix takes the form
 
e−βE1 0 ··· 0
 0
1  e−βE2 · · · 0 
ρthermal = , (1.41)

 .. .. .. ..
Z  . . . . 
0 ··· 0 e−βEN
X
Z= e−βEi ,
i

where β is the inverse temperature.

One might ask the question whether it makes sense to do quantum mechanics in terms of den-
sity matrices rather than with wave functions. Density matrices are Hermitian, and given the
constraint that the trace is unity, and N × N density matrix can be represented by N 2 − 1
real numbers. A wave function has N complex components, and given the normalization, have
2N − 1 components. Of course, there is also a meaningless phase, so a wave function can be
physically represented by 2N − 2 numbers. Clearly, the wave function is more efficient. How-
ever, the density matrix can also account for incoherent mixtures of states. This explains the
need for additional information.

1.6 Rotations of Two-Component Spin-Half Systems

As an example of a unitary transformation we consider rotations in a two-component system.
For spin-1/2 systems the spin operator is
ℏ
⃗=
S ⃗
σ. (1.42)
2

10
PHY 851/852 1 STATES AND OPERATORS

For two-component systems, all operators can be written as a linear combination of the Pauli σ
matrices and the unit matrix.
       
1 0 0 1 0 −i 1 0
σz = , σx = , σy = , I= . (1.43)
0 −1 1 0 i 0 0 1

As we will see later, under rotations the three Pauli matrices transform like components of a
vector. For now, we only notice that each matrix is Hermitian and that when each sigma matrix
is squared it gives the unit matrix. Note that the Pauli matrices have the properties,

σi2 = 1, (1.44)
{σi , σj } = 2δij ,
[σi , σj ] = 2iϵijk σk ,
σi σj = δij + iϵijk σk .

where the anti-commutator is noted by {A, B} ≡ AB + BA for operators A and B, and the
regular commutator is noted by [A, B] ≡ AB − BA. From these expressions one can readily
show that for a unit vector n̂,

σ · n̂)2 = I.
(⃗ (1.45)

For i ̸= j one can see

σi σj̸=i = iϵijk σk , (1.46)

σx σy = iσz , σy σz = iσx , σz σx = iσy .

For a state with spin-up or spin-down along the z axis, we choose the basis
   
1 0
| ↑⟩ = , | ↓⟩ = . (1.47)
0 1

Rotations by an angle θ are given by:

⃗ = e−iθ·⃗σ/2 , ⃗
R(θ) (1.48)
⃗ is along the axis of rotation n̂, or equivalently, θ
where the direction of θ ⃗ = θ n̂.
From performing a Taylor expansion, and using Eq. (1.44),

e−iθ⃗σ·n̂/2 = cos(θ/2) − i sin(θ/2)⃗

σ · n̂. (1.49)

This trick comes in handy for a large number of physics examples, not just rotations.
Example 1.3: Rotating Spin-1/2 Systems
Consider a fermion’s spin to originally be in the spin-up state, where “up” is defined by the z
axis.
 
1
|ψ⟩ = | ↑⟩ = .
0

11
PHY 851/852 1 STATES AND OPERATORS

a) Find |ψ⟩ after rotating by angles of θ = 90,◦ , 180◦ and 360◦ about the x axis.
b) What are the expectations of ⟨ψ|σi |ψ⟩ for each rotation.
c) What is |ψ⟩ if it is rotated by an angle ϕ about the z axis?

Solution:
a) Inserting σx into Eq. (1.49),

R(θ) = cos(θ/2) − i sin(θ/2)σx ,

and the answers are:

1
 
◦ ◦ 1
R(θ = 90 )| ↑⟩ = |ψ(90 )⟩ = √ ,
2 −i
 
◦ 0
|ψ(180 )⟩ = ,
−i
 
◦ −1
|ψ(360 )⟩ = .
0

b) Just stick in the Pauli matrices,

⟨ψ(0◦ )|σx |ψ(0◦ )⟩ = 0, ⟨ψ(0◦ )|σy |ψ(0◦ )⟩ = 0, ⟨ψ(0◦ )|σz |ψ(0◦ )⟩ = 1,

1 1
⟨ψ(90◦ )|σx |ψ(90◦ )⟩ = 0, ⟨ψ(90◦ )|σy |ψ(90◦ )⟩ = , ⟨ψ(90◦ )|σz |ψ(90◦ )⟩ = ,
2 2
⟨ψ(180◦ )|σx |ψ(180◦ )⟩ = 0, ⟨ψ(180◦ )|σy |ψ(180◦ )⟩ = 0, ⟨ψ(180◦ )|σz |ψ(180◦ )⟩ = −1,
⟨ψ(360◦ )|σx |ψ(360◦ )⟩ = 0, ⟨ψ(360◦ )|σy |ψ(360◦ )⟩ = 0, ⟨ψ(360◦ )|σz |ψ(360◦ )⟩ = 1.

c) Just reading it off from Eq. (1.49),

e−iϕ/2
 
◦
|ψ(ϕ = 90 )⟩ = ,
0

which is simply the original state multiplied by a phase factor. One odd feature here is that
the rotating by 360◦ returns the same state multiplied by −1.

It is important to contrast the behavior for rotating spin 1/2 systems, which are two-component
systems in a three dimensional world, to rotating normal vectors in a two-dimensional world.
For normal vectors, the rotation matrix would be
 
cos ϕ − sin ϕ
R(ϕ) = (1.50)
sin ϕ cos ϕ
= e−iϕσy
= cos ϕ − iσy sin ϕ.

Note the lack of the factor 1/2 in the argument, and that σy is being used even though this is a
rotation about the z axis. Also, for a rotation of 360◦ , one returns to the same state, without the

12
PHY 851/852 1 STATES AND OPERATORS

additional factor of −1. The fact that both kinds of transformations represent the same rotations
is something that will be discussed during the brief discussion of group theory in 4.2.

1.7 Polarization of Photons

Electromagnetic waves are comprised of quantized photons. For an electromagnetic wave trav-
eling in some direction, the light can be polarized, with the electric field oscillating being in some
direction perpendicular to the direction of propagation. If the light is moving in the z direction,
the field might be oscillating in the x direction. For a photon the spin is quantized into two pos-
sible polarization states. For a photon propagating in the z direction, those states might be |x⟩
or |y⟩, which would correspond to light polarized in the x or the y directions. The two states
might be represented by the vectors
   
1 0
|x⟩ = , |y⟩ = . (1.51)
0 1

Note that one switches from the |x⟩ to the y⟩ if rotated by 90◦ . This is in contrast to spin-
half particles, where one rotates by 180◦ to switch to the orthonormal state. Of course, the
other difference is that for photons we are considering only rotations about the direction of
propagation, whereas for spin-half particles, one must consider three dimensions of rotations.
The wave function for photons moving in the z direction has the form
 
−iωt+ikz a
ϕ(x, t) = e , (1.52)
b

where the vector is known as the polarization vector, which we consider to be normalized a2 +
b2 = 1. The following polarization vectors describe linearly polarized photons,
 
1
, linear polarization along the x axis
 0 
0
, linear polarization along the y axis (1.53)
 1 
1 1
√ , linear polarization along a direction 45◦ from x axis.
2 1

Another linear combination of |x⟩ and y⟩ states are right- and left-circularly polarized photons.
Those states are
1 1
 
1
|R⟩ = √ (|x⟩ + i|y⟩) = √ , (1.54)
2 2 i
1 1
 
1
|L⟩ = √ (|x⟩ − |y⟩) = √ ,
2 2 −i

To see why these are called linear polarized photons, one uses the fact that the electric fields are
proportional to the corresponding components of the wave function. Looking at those points

13
PHY 851/852 1 STATES AND OPERATORS

with z = 0, for the polarization with upper and lower components a and b, the electric fields
behave as

Ex = ReE0 ae−iωt , (1.55)

Ey = ReE0 be−iωt .
√ √
For a = 1/ 2 and b = i 2, this becomes

E0 E0
Ex = √ cos ωt, Ey = √ sin ωt. (1.56)
2 2
At t = 0 the electric field is pointed along the positive z axis, but after one fourth of a pe-
riod it is pointed along the positive y axis. At this fixed value of z, the direction of the field
is then rotating about the z axis√and is called right
√ circularly polarized. If one had considered
the polarization vector a = 1/ 2, b = −i/ 2, the polarization vector would have rotated
in the opposite direction, and be called left-circularly polarized. As long as one fixes the mo-
mentum of the photon, there are two polarizations and the polarization states can be described
in a two-component basis. If one wants the polarization vector to describe the direction of the
polarization even when the photon is propagating away from the z axis, polarization vectors
must then become three-component vectors. But because polarizations are perpendicular to the
direction of propagation, p̂, one must project away the contribution of those vectors in the p̂
direction.
Example 1.4: Photon Polarization
Photons are traveling along the z axis and are polarized along the x axis. They pass through
a polarization filter which only permits passage of photons whose polarization is at an angle
30◦ to the x axis. What fraction of photons pass through the filter.
Solution:
Consider a single photon. The original polarization vector is
 
1
|ϕ = 0⟩ = ,
0

and the state that passes through the filter is

   √ 
3
◦ cos ϕ
|ϕ = 30 ⟩ = = 2
1 .
sin ϕ 2

The probability of being in the new state is the squared overlap,

3
|⟨ϕ = 30◦ |ϕ = 0⟩|2 = .
4

14
PHY 851/852 1 STATES AND OPERATORS

1.8 Evolution in Time

Any function of time can be expressed as a Taylor series. For a state,
d τ 2 d2 τ n dn
|ψ(t0 + τ )⟩ = |ψ(t0 )⟩ + τ |ψ(t)⟩t=t0 + |ψ(t)⟩t=0 + · · · + |ψ(t)⟩t=t0 + · · · .
dt 2 dt2 n! dtn
(1.57)
Using the fact that the Taylor expansion for ex = 1 + x + · · · + xn /n! + · · · , one can see that
|ψ(t0 + τ )⟩ = U (t0 , t0 + τ )|ψ(t)⟩t=t0 (1.58)
 
d
U (t0 , t0 + τ ) = exp τ .
dt
The operation U (t0 , t0 +τ ) is referred to as the evolution operator as it evolves the state forward
(or backward) in time by an amount τ . The norm of a state, ⟨ψ|ψ⟩, represents the probability so
it which must remain equal to unity. This puts a condition on U ,
⟨ψ(t)|ψ(t)⟩ = ⟨ψ(0)|ψ(0)⟩ (1.59)
†
= ⟨ψ(0)|U (0, t)U (0, t)|ψ(0)⟩,
U † (0, t)U (0, t) = I.
Thus, the evolution operator must be a unitary operator.
For small times,
d
U (t, t + δt) = 1 + δt , (1.60)
dt
where d/dt returns the derivative w.r.t. at time t. We can define an operator H as
d
H(t) = iℏ , (1.61)
dt
d
H(t)|ψ(t)⟩ = iℏ |ψ(t)⟩,
dt
which is Schrod̈inger’s equation in general form and is nothing more than a definition of H. The
evolution operator for small times is
H(t)
U (t, t + δt) ≈ 1 − i δt, (1.62)
ℏ
H(t) − H † (t)
U † (t, t + δt)U (t + δt) ≈ 1 − i δt.
ℏ
Because U is unitary, one can state that H(t) = H † (t), or that H(t) must be Hermitian.
If H is independent of time, one can take inspect the Taylor expansion in Eq. (1.57) and see that
U (t, t + τ ) = e−iHτ /ℏ . (1.63)
For many considerations, e.g. in time-dependent perturbation theory, H depends explicitly on
time, which then invalidates the simple expression above.

15
PHY 851/852 1 STATES AND OPERATORS

1.9 Evolution of Two-component Systems

Many problems in quantum mechanics can be reduced to two-state problems. Aside from the
“spin-up spin-down” problem, the two-kaon problem (see Baym), the solar neutrino problem
along with many other examples are really simple variations of the two-state problem.
The two-state problem is especially nice because all two-by-two matrices can be written as a
linear combination of the unit matrix and the three sigma matrices. For convenience, we repeat
the properties of the Pauli matrices,
       
1 0 0 1 0 −i 1 0
σz = , σx = , σy = , I= . (1.64)
0 −1 1 0 i 0 0 1

As presented earlier, the Pauli matrices are Hermitian, traceless, and obey simple commutation
relations,
[σi , σj ] = 2iϵijk σk , {σi , σj } = 2δij . (1.65)
In fact, as we will see later, the matrices ℏ⃗
σ /2 obey the same commutation rules as angular
momentum. From the anti-commutation relations, the square of any σ matrix is unity,

σ · n̂)2 = I
(⃗ (1.66)

The evolution of states under a Hamiltonian,

H = β⃗
σ · n̂, (1.67)

is especially simple. In this case the evolution operator is

U (t) = e−iHt/ℏ (1.68)

= 1 − iβt⃗
σ · n̂ + (−iβt⃗ σ · n̂)2 /2! + · · · + (iβt⃗
σ · n̂)N /N ! + · · ·
= cos(βt) − i⃗σ · n̂ sin(βt),

because
I,

N N = even
(⃗
σ · n̂) = (1.69)
σ · n̂,
⃗ N = odd

Example 1.5: Evolving Spin 1/2 Particles

A spin-up(along the z-axis) particle is placed in an environment at t = 0 where it interacts
with a magnetic field pointed along the x axis,

H = βσx .

Find the probability of being in the “up” state as a function of time.

Solution:

16
PHY 851/852 1 STATES AND OPERATORS

|ψ(t)⟩ = e−iHt/ℏ |ψ(0)⟩

 
1
= (cos(βt) − i sin(βt)σx )
0
 
cos(βt)
= .
−i sin(βt)

Thus, the probability of being in the “up” state is cos2 (βt).

Example 1.6: Neutrino Mixing

Two species of neutrinos, the νµ and ντ , have masses mµ and mτ . The Hamiltonian that
describes these masses could be written:
 
mµ 0
H0 = .
0 mτ
Now, we consider an extra term added to the Hamiltonian that mixes the two flavors of neu-
trinos,  
0 α
Hmix = .
α 0
First, let us find the energies of two new states, which requires finding the eigenvalues of the
Hamiltonian. We do this by writing the Hamiltonian in terms of the sigma matrices.
1 1
H = (mµ + mτ )I + (mµ − mτ )σz + ασx
2 2
For a three dimensional vector, one can rewrite a vector,
A⃗ = Ax x̂ + Ay ŷ + Az ẑ
⃗
= |A|n̂,
q
⃗ = A2 + A2 + A2 ,
|A| x y z

⃗
n̂ = (Ax x̂ + Ay ŷ + Az ẑ)/|A|.
Because the σ matrices rotate amongst one another in the same way as x̂, ŷ and ẑ, we are
motivated to rewrite the Hamiltonian as
1
H = (mµ + mτ )I + β⃗ σ · n̂.
2
Here, β is the magnitude of the two terms that multiply sigma matrices,
q
β = α2 + (mτ − mµ )2 /4,

and n̂ is a unit vector pointing in the direction,

n̂ = ẑ cos θ + x̂ sin θ
α
sin θ =
β

17
PHY 851/852 1 STATES AND OPERATORS

Finding the eigenvalues is simply a matter of rotating the σ matrices so that n̂ is in the z
σ ·⃗ˆ= σz and the Hamiltonian becomes diagonal. The two energies
direction, at which point ⃗
are then,
1
E± = (mµ + mτ )c2 I ± βc2 .
2
One could also just diagonalize the two-by-two matrix, which would lead to the same answer.

1.10 Heisenberg and Schrödinger representations

Usually, one wishes to calculate expectations of operators, e.g. ⟨ϕ|AB · · · C|ψ⟩, where the states
evolve as a function of time, but the states are considered independent of time. Considering an
evolution operator, U = e−iHt/ℏ , one can express the time development of the expectation of
AB · · · C in either of two equivalent representations,

⟨ϕ(t)|AB · · · C|ψ(t)⟩ = ⟨ϕ(0)|U † AB · · · CU |ψ(0)⟩ (1.70)

= ⟨ϕ(0)| U † AU U † BU · · · U † CU |ψ(0)⟩.

The upper line of Eq. (1.70) is known as the Schrödinger representation, with the states evolving
while the operators are fixed. The operators above will be labeled AS (t) = A, with the subscript
referring to the fact that one is in the Schrödinger representation and distinguish it from the
operators in Heisenberg representation. The time dependence of AS (t) only exists if there is an
explicit time dependence added to the problem, e.g. a time-dependent potential. In the lower
line of Eq. (1.70), one can consider the states as being fixed, then consider the evolution of the
operators, AH (t) = U † (t)AS U (t). Even if AS has no time dependence, the operators AH (t)
will depend on time unless AS commutes with the Hamiltonian.
In the Heisenberg representation, the time development of the operators can be expressed as a
differential equation, where the rate of change of the operator is given by the commutation of
the Hamiltonian with AS .

AH (t) ≡ U † (t)AS U (t) (1.71)

d ∂  † 
AH (t) = U (t)AS (t)U (t)
dt ∂t  
i † †
∂
= U (t)[H(t), AS (t)]U (t) + U (t) AS (t) U (t).
ℏ ∂t
Here, the subscripts S and H refer to Schrödinger and Heisenberg representations respectfully.
If there is no explicit time dependence in AS , then any operator that commutes with the Hamil-
tonian represents a constant of the motion. The most obvious such operator is H itself. Thus,
if the Hamiltonian has no explicit time dependence, the expectation of H, a.k.a. the energy, is a
constant of the motion.
Example 1.7: Spin Precession in Heisenberg Representation
The spin-precession example from before can also be described by considering how the opera-

18
PHY 851/852 1 STATES AND OPERATORS

tors, in the Heisenberg representation, change with time, as opposed to how the states change
with time in the Schrödinger representation. Again, we consider the Hamiltonian, H = βσx .
The time development of σz (t) and σy (t) in the Heisenberg representation could be written
as,
d i
σz (t) = U † (t)[H, σz ]U (t) (1.72)
dt ℏ
i 2β
= βU † (t)[σx , σz ]U (t) = σy (t)
ℏ ℏ
d i 2β
σy (t) = βU † (t)[σx , σy ]U (t) = − σz (t),
dt ℏ ℏ
or equivalently,
d2 4β 2
σz (t) = − σz (t). (1.73)
dt2 ℏ2
Thus, the expression for ⟨σz (t)⟩ must have the form

⟨σz (t)⟩ = A cos (2βt/ℏ) + B sin (2βt/ℏ), (1.74)

where A and B are chosen to satisfy the initial conditions. If the initial spin is along the z axis
A = 1 and B = 0.

Finally, we should point out that for any two Hermitian operators A and B that commute, a set
of states can be found that are eigenstates of both A and B. To see this, consider eigenstates of
B,
B|b⟩ = b|b⟩. (1.75)
If A and B commute,
⟨b′ | [A, B] |b⟩ = 0 = (b − b′ )⟨b′ |A|b⟩. (1.76)
Thus, the operator A does not mix states with different eigenvalues of B. Or in other words, one
can take the subset of states which have a given eigenvalue b, and this set should provide a basis
for simultaneously diagonalizing A.
The implications of this simple statement are profound. For example if the angular momentum
operator Lz commutes with the Hamiltonian, the subset of states with the same eigenvalue of
Lz , usually denoted by m, can be considered by themselves while diagonalizing the Hamilto-
nian.
This also implies that for any operator A and a Hamiltonian H, where there is no explicit
time dependence in either operator, or equivalently the operator has no time dependence in
the Schrödinger representation, the time rate of change of AH (t) is,

d d
eiHt/ℏ AS e−iHt/ℏ


AH (t) = (1.77)
dt dt
i
= eiHt ([H, As ])e−iHt .
ℏ
Thus, any operator that commutes with the Hamiltonian is a constant of the motion.

19
PHY 851/852 1 STATES AND OPERATORS

1.11 Uncertainty Relations

In many examples one is interested in the expectations of the product of two or more Hermi-
tian operators. For instance, one might be interested in both the Hamiltonian and the number
operator. Here, we consider two Hermitian operators, A and B. One can choose states to be
eigenstates of either operator, but not necessarily both. This depends on whether the two oper-
ators commute. If there existed a basis where all the basis states were eigenstates of both A and
B, one could state that

AB|ab⟩ = ab|ab⟩, (1.78)

BA|ab⟩ = ab|ab⟩.

In this case the products AB and BA would return the same result acting on any basis state and
one could say

AB = BA, (1.79)
[A, B] = 0.

Thus, the commutation of operators is essential if one is to be able to find a basis that serves as
eigenstates for both operators.
For the case where the operators no longer commute, one can find an uncertainty relation. The
non-commutation results in some non-zero operator C,

[A, B] = iC. (1.80)

The operator C must be Hermitian because the commutation of two Hermitian operators must
be anti-Hermitian,

(AB − BA)† = (BA − AB) (1.81)

= − (AB − BA) .

Equivalently, multiplying the commutation by i results in a Hermitian operator.

To derive the uncertainty relation, one begins with the Cauchy-Schwarz inequality, https://en
.wikipedia.org/wiki/Cauchy\T1\textendashSchwarz_inequality. For any two operators δA
and δB sandwiched between any state |ψ⟩,

|⟨ψ|δAδB|ψ⟩|2 ≤ ⟨ψ|δA2 |ψ⟩⟨ψ|δB2 |ψ⟩. (1.82)

⃗ and Q,
This is related to the fact that for vectors P ⃗ |P
⃗ |2 |Q|
⃗ 2 ≥ (P
⃗ · Q)
⃗ 2 . For our purposes, we
define
δA ≡ A − ⟨ψ|A|ψ⟩, δB ≡ A − ⟨ψ|B|ψ⟩. (1.83)
Next, one can see that
1 1
⟨ψ|δAδB|ψ⟩ = ⟨ψ|[A, B]|ψ⟩ + ⟨ψ|{δA, δB}|ψ⟩, (1.84)
2 2
i 1
= ⟨ψ|C|ψ⟩ + ⟨ψ|{δA, δB}|ψ⟩.
2 2

20
PHY 851/852 1 STATES AND OPERATORS

The first term is imaginary and the second term is real because the anti-commutator of two
Hermitian operators is Hermitian. Thus,
1 1
|⟨ψ|δAδB|ψ⟩|2 = ⟨ψ|C|ψ⟩2 + ⟨ψ|{δA, δB}|ψ⟩2 . (1.85)
4 4
Plugging this into the Cauchy-Schwarz inequality, one obtains the Schrödinger-Robertson un-
certainty relation,
1 1
⟨ψ|δA2 |ψ⟩⟨ψ|δB2 |ψ⟩ ≥ ⟨ψ|C|ψ⟩2 + ⟨ψ|{δA, δB}|ψ⟩2 . (1.86)
4 4
Both terms on the r.h.s. are manifestly positive. The expectations ⟨ψ|δA2 |ψ⟩ and ⟨ψ|δB2 |ψ⟩
represent the degree to which either operator is well determined by |ψ⟩. If the state |ψ⟩ were an
eigenstate of A, δA would be zero and the l.h.s. would vanish. Because the r.h.s. is positive, this
would require ⟨ψ|δB2 |ψ⟩ to be infinite. Thus, one cannot simultaneously find any state where
both A and B are arbitrarily well determined.
The usual uncertainty relation involves ignoring the latter term, but including the latter term
strengthens the relation. For the famous Heisenberg uncertainty relation, the operators are the
position and momentum operators, X and P = −iℏ∂x , so C = ℏ, and ignoring the last term in
the Schrödinger-Robertson uncertainty relation,
ℏ2
⟨ψ|δX 2 |ψ⟩⟨ψ|δP 2 |ψ⟩ ≥ . (1.87)
4

Example 1.8: Uncertainty Relations for Angular Momentum

Find the uncertainty relation for the angular momenta Sx and Sy .
Solution: Use the fact that [Sx , Sy ] = iℏSz (which is true for any spin, not just spin 1/2).
Here,
C = −i[Sx , Sy ] = ℏSz .
Further, {Sx , Sy } = 0. The uncertainty relation is thus

ℏ2
⟨ψ|δSx2 |ψ⟩⟨ψ|δSy2 |ψ⟩ ≥ ⟨ψ|δSz2 |ψ⟩.
4

1.12 Exercises
1. Photons, traveling along the z axis can be polarized either linearly along the x or y axis, or
a linear combination of the two states.
(a) Write the operator that rotate states by 45◦ about the z axis in terms of |x⟩, |y⟩ and
the corresponding bras.
(b) Continuing with the photon moving in the z direction, and choosing the basis,
   
1 0
|x⟩ → , |y⟩ → ,
0 1
write the matrix that rotates the states by ϕ about the z axis.

21
PHY 851/852 1 STATES AND OPERATORS

(c) Right-hand circularly polarized (RCP) light is made of a linear combination of x and
y polarized light.
1
|R⟩ = √ (|x⟩ + i|y⟩) .
2
Light traveling along the z axis passes through a thin slab of thickness Z whose index
of refraction, k = nω/c, is different for light polarized in the x and y directions. In
terms of nx , ny and Z find the polarization vector for light which enters the slab as
right-circularly polarized.
HINT: The wave has a form e−iωt+ikz . The two components have the same ω but
different k while in the medium.
(d) Find the density matrix for right-circularly polarized light in the basis defined above.
(e) Using the basis described above, write the density matrix for light that is an incoherent
mixture, 50% polarized along the x direction and 50% along the y direction.

2. Considering a photon’s polarization for a photon moving in the z direction, calculate

⟨x|R(ϕ)|x⟩ for ϕ = π/2, π, 2π, where the rotation is about the z axis.
3. For a spin 1/2 particle, calculate ⟨z, +|R(θ)|z, +⟩, for the same angles, θ = π/2, π, 2π,
when the rotation is about the y axis.
4. Show that the unit matrix I, which can be considered as an operator, is unchanged by a
unitary transformation. Begin with the fact that for any matrix M, MI = IM = M.
5. Consider the rotation matrix for rotating Pauli spinors by an angle 90◦ about the z axis.
Using Eq. (1.49),
1
U = e−iσz π/4 = √ (1 − iσz ).
2
(a) Using the commutator and anti-commutator relations for the σ matrices, show that
the transformation of σx is
U σx U † = σy .
(b) Show that rotating the state, |+, x⟩, which refers to an eigenstate of σx with eigen-
value of +1, gives
U |+, x⟩ = |+, y⟩,
which is the eigenstate of σy with eigenvalue +1.

6. Consider some Hermitian N ×N matrix Kij , with eigenvalues λ(n) and the corresponding
normalized eigenvectors v (n) ,
Kv (n) = λ(n) v (n) .
(n)
The N eigenvectors each have N components, vi . Create an N × N matrix
∗(i)
Uij = vj .

Thus, one is making a matrix by having each row be one of the eigenvectors.

(a) Show that U is unitary.

22
PHY 851/852 1 STATES AND OPERATORS

(b) Show that the j th component of the vector U v (n) is

(U v (n) )j = δnj ,
Thus, the vectors U v (n) are
   
1 0

 0 


 1 

(1)
(U v ) = 
 0 (2)
 , (U v ) = 
  0  , ···

 ..   .. 
 .   . 
0 0
Now, consider the matrix
K ′ = (U KU † ),
and have it act on the vectors above. Show that
K ′ (U v (n) ) = λn (U v (n) ).

This shows that the vectors (U v (n) ) are eigenvectors of the matrix K ′ with eigenval-
ues λn . Given that the eigenvectors are of the simple form above, the matrix (U KU † )
must be diagonal. Thus, the matrix U defined above provides the unitary matrix for
transforming the matrix K into its diagonal form.
7. Consider the matrix: 

1 0 0
M= 0 0 1 
0 1 0
(a) What are the eigenvalues of M?

(b) Find eigenvectors of M?

8. Consider the 2×2 matrix
A C∗
 
K=
C B
(a) What are the eigenvalues of K?
(b) What are the eigenvectors of K?
9. A beam of light with wavelength 660 nm is sent along the z axis through a polaroid fil-
ter that passes only x polarized light. The beam is initially polarized at 30◦ to the x axis,
and the total energy of the pulse is exactly 10 Joules. Estimate the fluctuations of the en-
ergy of the transmitted beam, ⟨(E − Ē)2 ⟩1/2 . Express the fluctuations as a fraction of the
average transmitted energy. (Hint: Consider the binomial distribution, with N tries with
probability p of success of each try.)
10. Consider light moving along the z axis and using the following definitions for |R⟩ and |L⟩
in terms of x and y polarized light,
1 1
|R⟩ ≡ √ (|x⟩ + i|y⟩), |L⟩ ≡ √ (|x⟩ − i|y⟩),
2 2

23
PHY 851/852 1 STATES AND OPERATORS

(a) In terms of |R⟩(RCP) and |L⟩(LCP) write the states |45⟩ and |135⟩ which are linearly
polarized at 45◦ and 135◦ relative to the x axis.
(b) Calculate the 2 × 2 transformation matrix from the 45, 135 basis, where
   
1 0
|45⟩ = , |135⟩ = ,
0 1
to the R, L basis.
(c) Show that this transformation is unitary.
11. The probability that a photon in state |Ψ⟩ passes through an x-polaroid is the average value
of a physical observable which might be called the x-polarizedness..

(a) Write down the operator Px , as a matrix in the XY basis where

   
1 0
|X⟩ = , |Y ⟩ = .
0 1
The projection ⟨Ψ|Px |Ψ⟩ is the probability that |Ψ⟩ makes it through the filter.
(b) What are its eigenvalues and eigenstates?
(c) Write the matrix Px in the RL basis, where RCP and LCP states are
   
1 0
|R⟩ = , |L⟩ = ,
0 1
and show that the eigenvalues are the same as in the XY basis. Also, show that the
this matrix is a projection operator by explicitly multiplying Px by itself.

12. The trace of a matrix A is defined as:

X
TrA ≡ Aii
i

(a) Show that the trace of A is invariant under a transformation of basis,

A → U † AU

(b) Show that TrAB = TrBA.

13. A plane polarized photon at θ = 45◦ enters a special crystal with indices of refraction:
nx =1.50 for photons polarized along the x axis
ny =1.52 for photons polarized along the y axis.
Assuming the wavelength of the light is 660 nm before it enters the crystal, choose the
thickness of the crystal such that the outgoing light is right circularly polarized. Assume
the dispersion is linear, k = nω/c.
14. Consider the matrix for rotation about the z axis,
R(ϕ) = e−iσz ϕ/2 . (1.88)
Show that after rotation about the z axis,
R(ϕ)σx R−1 (ϕ) = σx cos(ϕ) + σy sin(ϕ) (1.89)

24
PHY 851/852 1 STATES AND OPERATORS

15. Consider a basis for spin-up and spin-down electrons (along the z axis),
   
1 0
| ↑⟩ = , | ↓⟩ =
0 1

(a) Write down the 4 vectors describing an electron with spin pointed along the posi-
tive/negative directions of x and y axes.
(b) Write the six density matrices describing electrons polarized along the positive/neg-
ative directions of each of the three axes.
(c) Write the density matrix describing an incoherent mixture of 60% spin-up and 40%
spin down.
(d) Using the density matrix, calculate < y, +|Sz |y, +⟩.

16. Neutral Kaon Oscillations: There are two kinds of neutral kaons one can make using down
and strange quarks,
¯
|K 0 ⟩ = |ds̄⟩, |K̄ 0 ⟩ = |sd⟩.

If it weren’t for the weak interaction, the two species would have equal masses, and the
Hamiltonian (for a kaon with zero momentum) would be
 
M 0
H0 = .
0 M

However, there is an additional term from the weak interaction that mixes the states,
 
0 ϵ
Hm = .
ϵ 0

The masses of a neutral kaon are 497.6 MeV, without mixing, but after adding the mixing
term the masses differ by 3.56µeV. The two eigenstates are known as KS (K-short) and KL
(K-long), because they decay with quite different lifetimes.

(a) What is ϵ?
(b) If one creates a kaon in the K0 state at time t = 0, find the probability it would be
measured as a K̄ 0 as a function of time.
(c) A beam kaons is created in the K0 channel and has a kinetic energy of 600 MeV per
kaon. Plot the probability that the kaon is in the K0 state as a function of the distance
traveled, x. Ignore the fact that the kaons decay.
(d) Repeat (c), but take into account the decays. The states

1
|KS ⟩ = √ (|K 0 ⟩ + |K̄ 0 ⟩,
2
1
|KL ⟩ = √ (|K 0 ⟩ − |K̄ 0 ⟩,
2

25
PHY 851/852 1 STATES AND OPERATORS

known at K-short and K-long, represent the eigenstates of the Hamiltonian. The
lifetime of a KL is 51.2 ns, and the lifetime of the KS is 0.0896 ns. Note that the wave
function should be modified by the factor e−t/(2τ ) to take decays into account decays
of lifetime τ . It is often convenient to remember that ℏc=197.327 MeV fm = 197.327
eV nm.

FYI: If the above were exactly true, the KS state would be even under CP while the KL
would be odd under CP . Here, CP is an operator that changes particles to anti-particles.
If the particle-antiparticle symmetry were exact, the CP operator would commute with
the Hamiltonian and the eigenstates of the Hamiltonian, KS and KL , would have to be
eigenstates of CP . The KS would then decay to two pions and the KL could decay to
three pions. However, there is an additional small CP violating term in the Hamiltonian
which allows KL to have a small probability of decaying to two pions. This was the first
experimental laboratory observation of CP violation. CP violation is required to explain
the preponderance of matter vs. anti-matter in the universe.
17. Neutrino Oscillations: There are three kinds of neutrinos corresponding to the three lepton
families, and recent evidence has suggested that they may oscillate between generations.
Here we consider two flavors, the µ neutrino and the τ neutrino. Suppose that the Hamil-
tonian can be written as a free term plus a term that mixes the µ and τ neutrinos, which is
proportional to α.
q !  
2
k + mµ 2 0 0 1
H= +α
0
p
k2 + m2τ 1 0

(a) Supposing you are in the rest frame of the neutrino and that the momentum k is zero,
show that the evolution operator e−iHt/ℏ can be written as
e−i(mµ +mτ )t/2ℏ {cos ωt − iσn sin ωt} ,
where r
mτ − mµ
ℏω ≡ α2 + ( )2
2
mµ −mτ
2
σz + ασx
σn ≡
ℏω
(b) If a neutrino starts as a µ neutrino, what is the probability, as a function of time, of
being a τ neutrino?

(c) As a function of the masses and α, what is the oscillation time? I.e. the time to return
to its original flavor.
(d) If the neutrinos are extremely relativistic, k >> m, describe how the oscillation time
translates into an oscillation as a function of the distance from the creation.

Note: Here the “masses” are the rest energies (mc2 ).

18. Show that
TrAS BS CS = TrAH (t)BH (t)CH (t),
where the subscripts refer to Schrödinger and Heisenberg representations.

26
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

2 Coordinate and Momentum Space

In the previous section we considered problems with a finite number of discrete states. If there
were N discrete states, the problem was reduced to a linear algebra exercise involving N -
dimensional vectors to represent the states, and N × N dimensional matrices to represent the
operators. For particles confined to a finite volume, such as a harmonic oscillator or a infinite
potential well, the system still has discrete levels, but the number of such levels is unbounded as
the states can have arbitrarily high energy. A vector space, where the dimensionality approaches
infinity is known as a Hilbert space. Thus, linear algebra is not applicable without some sort of
truncation. In its place, differential equations often provide the dominant mathematical tool to
represent the physics.
Having the number of states approach infinity is typically the result of either (a) a continuum of
points in coordinate space or (b) an infinite extent to coordinate space, or (c) both. An example
of (a) is an infinite square well or a harmonic oscillator. The solutions to these problems lead to
an infinite number of discrete states. An example of (b) would be an infinitely long chain with
sites separated by some distance ℓ. In this case there is a continuum of momentum solutions,
e.g. phonons, but with a maximum momentum of the order ℏ/ℓ. Finally, for (c), one has a
continuum of solutions, and no upper bound on the energy or momentum. For problems with
a continuum in coordinate space, the physics is often described by some sort of wave equation,
most famously Schrödinger’s wave equation. The qualifier wave is invoked to distinguish it
from Hψ = Eψ, which is not necessarily a wave equation, and might also be referred to as
Schrödinger’s equation.

2.1 Schrödinger’s Wave Equation

One of the most famous equations in physics is Schrödinger’s wave equation,

∂ ℏ2 ∂ 2
iℏ ψ(x) = − ψ(x) + V (x)ψ(x), (2.1)
∂t 2m ∂x2
where ψ is a complex function of the continuous variable x. Rather than thinking of ψ(x) as a
wave function, it is more revealing to recognize it as overlap of the state |ψ⟩ with the state |x⟩.

ψ(x) = ⟨x|ψ⟩. (2.2)

The only difference between a label in coordinate space and a label that denotes a discrete vari-
able such as spin, is that because the x label is continuous, the normalization has to be changed.

⟨x′ |x⟩ = δ(x − x′ ), (2.3)

where δ(x − x′ ) is a Dirac delta function, as opposed to the usual Kronecker delta function, δmn ,
used for discrete states. Dirac delta functions are zero when x ̸= x′ and infinty for x = x′ such
that
Z
dx′ δ(x − x′ ) = 1, (2.4)

27
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

whereas for Kronecker deltas

X
δmn = 1. (2.5)
n

Equation (2.3) implies that |x⟩ has dimensions of inverse length to the one-half power. The
completeness relation becomes
Z
dx |x⟩⟨x| = I, (2.6)
Z
dx ⟨ϕ|x⟩⟨x|ψ⟩ = ⟨ϕ|ψ⟩.

Next, we wish to show that the Schrödinger equation is merely the continuum limit of a matrix
equation, with ψ0 , ψ1 · · · ψN representing ψ(x) at x = 0, δx, 2δx · · · . To do this we write the
second derivative as
∂2 ∂ ψ(x + δx/2) − ψ(x − δx/2)
ψ(x) = lim (2.7)
∂x2 δx→0 ∂x δx
ψ(x + δx) − 2ψ(x) + ψ(x − δx)
= lim .
δx→0 δx2
Thus, by making the substitution,
1
ψ(x) → ψi √ , (2.8)
δx
the normalization becomes:
X
|ψi |2 = 1. (2.9)
i

One may now rewrite Schrödinger’s equation in terms of discrete vectors,

∂ ℏ2
i ψi (t) = − (ψi+1 (t) − 2ψi (t) + ψi−1 (t)) + V (x)ψi (t). (2.10)
∂t 2mδx2
If one takes the limit of δx → 0, the number of states |x⟩ within a finite length approaches
infinity.
For the case where δx is finite one may also write the Hamiltonian as a matrix.
   
··· ··· ··· ··· ··· ··· ··· ··· ··· ···
ℏ2 
 · · · −2 1 0 · · · 
 
 · · · V (x i−1 ) 0 0 ··· 

H =−  ··· 1 −2 1 · · ·  +  · · ·
  0 V (xi ) 0 ··· 
.
2mδx2  ··· 0 1 −2 · · ·   ··· 0 0 V (xi+1 ) ··· 
··· ··· ··· ··· ··· ··· ··· ··· ··· ···
(2.11)
Here, the potential term is diagonal while the kinetic term is band diagonal.

28
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

2.2 Momentum States

A momentum state is no more than a linear combination of coordinate-space states,
Z
|p⟩ = dx eipx/ℏ |x⟩, (2.12)
Z
′
⟨x|p⟩ = dx′ eipx /ℏ ⟨x|x′ ⟩
Z
′
= dx′ eipx /ℏ δ(x − x′ )

= eipx/ℏ .

With this definition of the state |p⟩, the normalization becomes

Z
′ ′ ′
⟨p |p⟩ = dxdx′ ei(px−p x )/ℏ ⟨x′ |x⟩ (2.13)
Z
′
= dx ei(p−p )x/ℏ

= 2πℏδ(p − p′ ),

and the dimensionality of |p⟩ is length to the one-half power.

Aside: Fourier Transforms and Delta Functions

The relation above is often used to manipulate Fourier transforms. To derive the fact that inte-
grating the phase eikx gives a delta function, one can add a factor e−ϵ|x| to the integrand and
take the limit ϵ → 0. The proof then proceeds,
Z ∞ Z 0 Z ∞ 
−ϵx+ikx 
ikx ϵx+ikx

dx e = dx e + dx e  (2.14)
−∞ −∞ 0 ϵ→0
1 1
= +
ϵ + ik  ϵ − ik
2ϵ 
= 2 .
ϵ + k2  ϵ→0

The r.h.s. is clearly zero when k ̸= 0, but is infinity for k = 0. To see that this is proportional to
δ(k), one need only integrate the function and see,
Z ∞
2ϵ
dk 2 = 2πδ(k). (2.15)
−∞ ϵ + k2

To perform this last integral, one can do a trigonometric substitution, k = ϵ tan θ.

The inverse transformation is

1
Z
|x⟩ = dpe−ipx/ℏ |p⟩. (2.16)
2πℏ

29
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

Expressed as a completeness relation,

dp dp ip(x′ −x)/ℏ
Z Z
′
⟨x|p⟩⟨p|x ⟩ = e (2.17)
(2πℏ) (2πℏ)
= δ(x′ − x) = ⟨x|x′ ⟩.

Thus,
1
Z
dp |p⟩⟨p| = I. (2.18)
2πℏ
Sandwiching this between some normalized state ψ,

1
Z
dp ⟨ψ|p⟩⟨p|ψ⟩ = ⟨ψ|ψ⟩ (2.19)
2πℏ Z
1
dp ψ ∗ (p)ψ(p) = 1,
2πℏ
ψ(p) = ⟨p|ψ⟩.

Thus, the probability per differential momentum for observing a particle with momentum p in
some state ψ is
dPψ 1
= ψ ∗ (p)ψ(p). (2.20)
dp 2πℏ

Changing the problem to n dimensions only affects the expressions here by changing dp to dn p
and replacing (2πℏ) with (2πℏ)n .
The label p here refers to a continuum of momentum states. If a system is confined, then no
eigenstate of the momentum operator really exists, but one still uses the label p to point to
eigenstates of the momentum operator within the volume, but with discrete values of p rather
than a continuum. In that case the normalizations are different. It is the duty of the watchful
reader to accurately interpret the notation.
Momentum and position are also operators, and as operators can be expressed in terms of bras
and kets.
Z
X = dx x|x⟩⟨x| (2.21)
dp
Z
P = p|p⟩⟨p|.
2πℏ
From the definition of X , one can see that
Z
⟨ϕ|X |ψ⟩ = dx ⟨ϕ|x⟩x⟨x|ψ(x)⟩ (2.22)
Z
= dx ϕ∗ (x)xψ(x).

30
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

One can also perform a similar operation with the momentum operator.
dp ∗
Z
⟨ϕ|P|ψ⟩ = ϕ (p)pψ(p), (2.23)
2πℏ
ψ(p) = ⟨p|ψ⟩.

However, the customary way to view the momentum operator is not in momentum space, but
as a derivative in coordinate space. Expanding |p⟩ in terms of coordinate space states, one can
use completeness to express the ⟨ϕ|P|ψ⟩ in terms of wave functions,
dp
Z
⟨ϕ|P|ψ⟩ = dxdy ⟨ϕ|x⟩⟨x|p⟩p⟨p|y⟩⟨y|ψ⟩ (2.24)
2πℏ
dp
Z
= dxdy ⟨ϕ|x⟩eip(x−y)/ℏ p⟨y|ψ⟩
2πℏ
The factor p can be changed into a derivative of the phase e−ipy/ℏ with respect to y,
dp
Z
iℏ∂y eip(x−y)/ℏ ⟨ϕ|x⟩⟨y|ψ⟩


⟨ϕ|P|ψ⟩ = dxdy (2.25)
2πℏ
dp ip(x−y)/ℏ
Z
= dxdy e ⟨ϕ|x⟩ (−iℏ∂y ⟨y|ψ⟩) .
2πℏ
One can now perform the integral over p to get a factor of δ(x − y), which then allows one to
integrate over y,
Z
⟨ϕ|P|ψ⟩ = dx ⟨ϕ|x⟩ (−iℏ∂x ) ⟨x|ψ⟩ (2.26)
Z
= dx ϕ∗ (x) (−iℏ∂x ) ψ(x),
Z
P = dx |x⟩ (−iℏ∂x ) ⟨x|.

This algebra also shows that the one can express the momentum operator as a left-acting deriva-
tive, but with a change of sign.
Thus, the momentum operator can be thought of a −iℏ∂/∂x. With some algebra, one can see
that commuting P with X yields,
Z Z
[P, X ] = dx|x⟩(−iℏ∂x )(x⟨x|) − dx |x⟩(x(−iℏ∂x ))⟨x| (2.27)
Z
= −iℏ dx |x⟩⟨x|

= −iℏI.

One should keep in mind that in many applications the definition of a momentum state in Eq.
(2.12) is replaced with
 ipx √
e / L, 0 < x < L
⟨x|p⟩ ≡ , (2.28)
0, otherwise

31
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

where L → ∞. For this definition, ⟨x|p⟩ has dimensions of length−1/2 , and the values of p are
discrete. The normalizations of the momentum states are then
⟨p|p′ ⟩ = δpp′ , (2.29)
which are Kronecker delta functions instead of Dirac delta functions. These states are not ac-
tually eigenstates of the momentum operator because −iℏ∂x ⟨x|p⟩ ̸= p⟨x|p⟩ when x < 0 or
x > L. Typically, this choice is useful when one is considering emission into a continuum of
momentum states, and the density of such states factors into the answer. Because the density of
states is proportional to L, the arbitrarily large lengths typically cancel from the answer. In prac-
tice, when reading texts or literature, or the later chapters of these notes, the choice is usually not
explicitly stated, and the reader must discern the choice of convention based on whether factors
of the length or volume appear.

Aside: Schrödinger’s Wave Equation as an Operator

Before we find solutions, we regress to consider the notation and vocabulary associated with
Schrödinger’s equation. One often writes the Hamiltonian in this way,
ℏ2 ∂ 2
Hψ(x) = Eψ(x), H = − + V (x). (2.30)
2m ∂x2
Referring to the Hamiltonian H as an operator is actually incorrect, as it does not represent
the fact that the Hamiltonian is an operator and should be sandwiched between a bra and ket.
Operators are defined by matrix elements, and should referenced by two labels. In this case
⟨x′ |H|x⟩, which should be considered a function of both x and x′ . More correctly, beginning
with H defined as an operator, then using completeness,
H|ψ⟩ = E|ψ⟩ (2.31)
Z
dx′ ⟨x|H|x′ ⟩⟨x′ |ψ⟩ = E⟨x|ψ⟩.

Now, one assumes that H does not mix |x⟩ with |x′ ⟩ unless x and x′ are a the same position
(or neighboring positions in the limit that one discretizes coordinate space). In that case one can
define a function h(x) as,
⟨x|H|x′ ⟩ = h(x)δ(x − x′ ), (2.32)
and Schrödinger’s equation becomes
h(x)⟨x|ψ⟩ = E⟨x|ψ⟩, (2.33)
h(x)ψ(x) = Eψ(x),
ℏ2 ∂ 2
h(x) = − + V (x).
2m ∂x2
This is not simply an exercise in notation. It emphasizes that the Hamiltonian operator does not
mix states with x far from x′ . Without such a constraint, one would not have local current con-
servation. Local conservation of current requires that charge cannot instantaneously transport
across non-zero distances.

32
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

2.3 Local Charge Conservation and the Equation of Continuity

The equation of continuity is omnipresent throughout physics,

r , t) = −∇ · ⃗j(⃗
∂t ρ(⃗ r , t). (2.34)

To see that this infers that charge conservation is local, one applies the divergence theorem. For
any volume V ,
Z Z
∂t 3
r , t) = −
d r ρ(⃗ d3 r ∇ · ⃗j(⃗ r , t) (2.35)
V V
d
I
Q = − dA ⃗ · ⃗j(⃗
r , t),
dt
where Q is the charge within the volume V . By considering arbitrarily small volumes, this
shows that the change in charge within that volume can be identified with the current density
⃗j exiting the surface of the volume. Thus, charge cannot leave a small volume in East Lansing,
only to instantaneously appear in Kalamazoo. Instead, it appears in the neighboring volume,
and might then make its way to Kalamazoo over time.
Schrödinger’s equation allows one to identify both the charge and current densities. First, con-
sider Schrödinger’s wave equation and its complex conjugate,
−ℏ2 ∇22 ∂
ψ ∗ (⃗
r1 , t1 )ψ(⃗ r2 , t2 )ψ ∗ (⃗
r2 , t2 ) + V (⃗ r1 , t2 )ψ(⃗
r2 , t2 ) = −iℏ ψ ∗ (⃗
r1 , t1 )ψ(⃗
r2 , t2 ),
2m ∂t2
(2.36)
2
−ℏ ∇21 ∂
ψ ∗ (⃗
r1 , t1 )ψ(⃗ r1 , t1 )ψ ∗ (⃗
r2 , t2 ) + V (⃗ r1 , t2 )ψ(⃗
r2 , t2 ) = iℏ ψ ∗ (⃗
r1 , t1 )ψ(⃗
r2 , t2 ).
2m ∂t1

The second equation is found by taking the complex conjugate of the first equation, then switch-
ing (1 ↔ 2). Subtracting the two equations, and using the fact that ∇21 − ∇22 = (∇1 − ∇2 ) ·
(∇1 + ∇2 ),
ℏ2
− (∇1 + ∇2 ) · (∇1 − ∇2 ) [ψ ∗ (⃗ r1 , t1 )ψ(⃗ r2 , t2 )] (2.37)
2m
+ [V (⃗ r2 , t)] [ψ ∗ (⃗
r1 , t) − V (⃗ r1 , t1 )ψ(⃗r2 , t2 )]
 
∂ ∂
= −iℏ + [ψ ∗ (⃗
r1 , t1 )ψ(⃗r2 , t2 )] .
∂t1 ∂t2
Then dividing by iℏ,
−iℏ
(∇1 + ∇2 ) · (∇2 − ∇1 )ψ ∗ (⃗
r1 , t1 )ψ(⃗
r2 , t2 ) (2.38)
2m
−i
+ [V (⃗ r2 , t)] [ψ ∗ (⃗
r1 , t) − V (⃗ r1 , t1 )ψ(⃗ r2 , t2 )]
iℏ  
∂ ∂
= + ψ ∗ (⃗r1 , t1 )ψ(⃗ r2 , t2 ).
∂t1 ∂t2

33
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

r1 = ⃗
Setting ⃗ r2 and t1 = t2 , one finds

∂t ρ(⃗r , t) = −∇ · ⃗j(⃗ r , t), (2.39)

∗
ρ(⃗r , t) = ψ (⃗r1 , t1 )ψ(⃗ r2 , t2 )
⃗j(⃗ iℏ ∗ iℏ
r , t) = − ψ (⃗ r , t)∇ψ(⃗ r , t) + [∇ψ ∗ (⃗
r , t)]ψ(⃗
r , t).
2m 2m
⃗
If one associates iℏ∇/m with the velocity, then the current looks like the velocity density. For
Schrödinger’s equation, the net number of particles is a conserved quantity. If the described
particle has some charge q, then the charge density is qψ ∗ ψ.

2.4 Potential Problems in One Dimension

Although few physics applications are in one dimension, many problems reduce to one dimen-
sion due to symmetry. For example, once there is rotational invariance, a three-dimensional
solution to the Schrödinger wave function can be factored into products of a radial and angular
pieces. The radial equation can then be mapped into a simple one-dimensional equation with an
effective centrifugal potential. One-dimensional solutions form the foundation for understand-
ing a wide range of physics beyond one-dimension.
Consider Schrödinger’s wave equation in one dimension,

ℏ2 ∂ 2 ∂
− Ψ(x, t) + V (x)Ψ(x, t) = iℏ Ψ(x, t). (2.40)
2m ∂x2 ∂t
If Ψ(x, t) is an eigenstate of the Hamiltonian with eigenvalue E, the solution becomes

Ψ(x, t) = e−iEt/ℏ ψ(x), (2.41)

and the time derivative iℏ∂/∂t can be replaced by E in Schrödinger’s wave equation. Because
the time dependence is then simply a phase factor, all observables become fixed in time for eigen-
states. For many applications, we will consider the steady-state case by looking for eigenstates.
However, the fact that these are eigenstates does not necessarily imply that the currents are zero.
An example is a momentum eigenstate. One can set boundary conditions such that current en-
ters and exits the boundary, thus resulting in a steady-state solution, but one where the charge
moves.
To find solutions for problems where the potential is continuous, it is sufficient to find solutions
of the wave equation that have the correct behavior at x → ±∞. If the potential is discontinu-
ous at certain points, boundary conditions must be enforced at every point where a discontinuity
is formed.
If the discontinuity is finite (not a delta function or an infinite potential wall), the boundary
conditions at the discontinuity are:

1. ψ(x) is continuous.
2. ∂x ψ(x) is continuous.

34
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

If ψ were discontinuous, the derivative would be infinite at that point and the current would be
infinite at the discontinuity, which would make it impossible to enforce the equation of continu-
ity. If the first derivative were discontinuous at some point y, the second derivative would be
infinite at that point. To understand the second boundary conditions we integrate Schrödinger’s
equation from x = y − ϵ to x = y + ϵ, where ϵ → 0. This yields

ℏ2
  Z y+ϵ
∂ ∂
− ψ(x)|y+ϵ − ψ(x)|y−ϵ = dx(E − V (x))ψ(x). (2.42)
2m ∂x ∂x y−ϵ

For finite potentials, the integrand in the integral of the r.h.s. of this equation is manifestly finite,
thus as ϵ goes to zero, the r.h.s. must vanish, which demonstrates the boundary condition. One
should note that for an infinite potential step the boundary conditions can be met by simply
having the wave function vanish. In this case it would appear the slope changes suddenly, but
because the wave function is zero there is no energy in the infinitesimal region.
Example 2.1: Sudden Disintegration of Square Well
Consider the simple potential,

 ∞, x < −a
V (x) = 0, −a < x < a
∞, x>a


The ground state wave function is

ℏ2
r 2
1

π
ψ(x) = cos(πx/2a), E = .
a 2m 2a

If the well suddenly dissolves, what is the probability the particle will be observed with mo-
mentum p?
Solution: From Eq. (2.20), the probability of observing p is

dPψ 1
= |⟨p|ψ⟩|2 ,
dp 2πℏ
Z
⟨p|ψ⟩ = dx ⟨p|x⟩⟨x|ψ⟩
Z
= dx e−ipx/ℏ ψ(x)
Z a
1
= √ dx [cos(k) − i sin(kx)] cos(k0 x)]
a −a
Z a
1
= √ dx cos(kx) cos(k0 x),
a −a
π p
k0 ≡ , k≡ .
2a ℏ

35
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

0.3

0.2

( /a)dPψ /dp 0.1

0.0−3π −2π −π 0 π 2π 3π
pa/
Figure 2.1: Differential probability of observing particles of momentum p from a suddenly disintegrating
square well of width 2a, solved in Example 2.1.

The term with sin(kx) was discarded by symmetry. Integrating,

Z a
1
⟨p|ψ⟩ = √ dx {cos[(k + k0 )x] + cos[(k − k0 )x]}
2 a −a
1 sin[(k + k0 )a] sin[(k − k0 )a]
 
= √ +
a k + k0 k − k0
√ 1 1
 
= a cos(ka) −
ka + π/2 ka − π/2
√ π
= a cos(ka) .
(ka)2 − π 2 /4

The dimension of |⟨p|ψ⟩|2 is that of length, and given that the units of ℏ are
momentum×length (or energy×time), the probability dPψ /dp indeed has units of inverse
momentum. The probability is displayed in Fig. 2.1.

For many exercises a potential is chosen that is constant within some range of x, e.g. 0 < x < L.
In that region the form of the wave function is either sinusoidal or exponential, depending or
whether the energy is above or below the potential If the energy is above the potential, i.e. the
kinetic energy would appear positive, the form of the wave function is sinusoidal,
ψ(x) = Aeikx + Be−ikx , (2.43)
or equivalently
= C cos(kx) + D sin(kx).
The wave number k is chosen to match the kinetic energy,
ℏ2 k2
= KE = E − V (x). (2.44)
2m

36
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

If the energy is below the potentia., i.e. the kinetic energy would be negative, the form is expo-
nential, or equivalently in terms of hyperbolic sines and cosines,
ψ(x) = Ae−κx + Beκx , (2.45)
or equivalently
= C cosh(κx) + D sinh(κx).
The exponential scale κ is again chosen according to the energy
ℏ2 κ2
= V (x) − E. (2.46)
2m
For the latter case the particles are energetically disallowed, but can penetrate into the disal-
lowed region with an exponential form. In both cases, the constants A, B, C or C can be com-
plex coefficients.
Example 2.2: Finite Potential Well
Consider the potential,

 ∞, x<0
V (x) = −V0 , 0 < x < a
0, x>a


Solve for the binding energy and wave function of the lowest energy state:
Solution: Assume the energy is negative. In the region of the
p well solutions of2Schrödinger’s
equation are sines and cosines with wave number, k = (2m(V0 − B)/ℏ ), where B is
the binding energy (a positive number between zero and V0 ). The boundary condition at the
origin is that the wave function must go to zero due to the infinite potential. Thus, only the
sine piece remains in region I (0 < x < a),
ψI (x) = sin(kx).
We have chosen an arbitrary normalization constant of unity.
In the second region, (x > a), the potential is zero and E − V < 0, so the solutions
are exponentially
p growing or decaying. They are characterized by a decay constant of
2
q = (2mB/ℏ ). The exponentially growing piece can be thrown out as we wish to find a
solution where some probability is near the origin.
ψII (x) = Ae−qx
Thus far, we have neglected the two boundary conditions at x = a, which are necessary to
determine the two unknowns, the binding energy B and the normalization factor A. Writing
the two boundary conditions,
sin(ka) = Ae−qa (2.47)
−qa
k cos(ka) = −qAe
Dividing the two boundary conditions eliminates A and gives the relation,
−k
tan(ka) = ,
q
p p
k = 2m(V0 − B)/ℏ2 , q = 2mB/ℏ2 .

37
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

This is a transcendental equation for B (or equivalently for k or q) which can be solved
graphically,
p or on a computer, and keeping in mind that k must be between zero and
kmax = 2mV0 /ℏ2 . In order for the two functions above to intersect and have p the same
slope, the wave function must have a negative slope, and because k < kmax = 2mV0 /ℏ2 ,
this can only happen when kmax a > π/2. Thus, the solution disappears if the depth, V0 ,
is too small or if the the width a is too narrow. At the point where kmax a = π/2, the solu-
tion has binding energy zero and q → 0. Using the expression for tan(ka) one can see that
tan(ka) = ∞, or ka = π/2 as just stated.
In one dimension, if a potential remains zero or below zero for all x, there is always at least
one bound state even if the potential depth or range become infinitesimally small. However,
in this case the potential is positive, and in fact infinite, for x < 0, and the potential needs to
have sufficient strength and extent to result in a bound state.

Example 2.3: Plane Wave Incident on Finite Barrier

Consider a plane wave of wave number k incident on a positive barrier,


0, x < 0
V (x) = .
V0 , x > 0

Find the reflection and transmission probabilities for a particle with energy E which is greater
than V0 .
Solution: In region I (x < 0) the solution consists of both an incoming, ∝ eikx , and a reflected,
∝ e−ikx , component. Whereas in region II (x > 0) one only has a component moving in the
positive x direction.
p
ψI (x) = eikx + Ae−ikx , k = 2mE/ℏ2 ,
p
ψII (x) = Beiqx , q = 2m(E − V0 )/ℏ2 .

The boundary conditions yield

1 + A = B,
k(1 − A) = qB.

Here, k and q are known, and the unknowns are A and B. Solving for A and B,
k−q 2k
A= , B=
k+q k+q
Thus, one quickly obtains the reflection and transmission probability.
In order to check the answer, one can see that the outgoing fluxes sum to the incoming flux.
ℏk ℏq ℏk
|A|2 + |B|2 = .
m m m
Fluxes are given by the square of the amplitude, which represents the probability density,
multiplied by the velocity, or in this case the momentum divided by the mass.

38
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

We now consider a discontinuity at y, and perform the integral over an infinitesimal range cen-
tered about y. From Eq. (2.42) one could see that if the potential is finite in the neighborhood of
y one obtains
ℏ2
 
∂ ∂
− ψ(x)|y+ϵ − ψ(x)|y−ϵ = 0. (2.48)
2m ∂x ∂x
However, if the potential is a delta function, integrating the right-hand side of Eq. (2.42) over
the infinitesimal range can yield a finite value. Thus, if

V (x) = βδ(x − y), (2.49)

the boundary condition becomes

ℏ2
 
∂ ∂
− ψ(x)|y+ϵ − ψ(x)|y−ϵ = −βψ(y). (2.50)
2m ∂x ∂x
The slope of the wave function is then not continuous at the delta function discontinuity.
Example 2.4: Delta Function Potential
Consider the potential,

V (x) = −βδ(x), β > 0.

Find the binding energy of the bound state.

Solution: Assume the existence of a bound
p state of binding energy B. The solutions are
exponentials with decay constant k = 2mB/ℏ2 , and the requirements that they go to zero
at x = ±∞, and that they are equal at x = 0 demands

exp(−kx) x > 0
ψ(x) = .
exp(kx) x < 0

Plugging them into the boundary condition at x = 0, Eq. (2.50), gives

ℏ2 k
= β. (2.51)
m
The binding energy is
ℏ2 k2 β2m
B= = ,
2m 2ℏ2
and a bound state exists for any value of β.

2.5 Wave Packets

A plane wave confined to a finite region can be written as

 0, x<0
1 ipx/ℏ
ψp,L (x) = ⟨x|p, L⟩ = √L e , 0<x<L (2.52)
0, x > L


39
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

which is normalized in the length L, but is not really an eigenstate of the momentum operator
due to the sharp cutoff at the boundaries. A more physical description of a wave that is confined
to a finite region is a wave packet, which confines the probability to a finite region, and because
of the uncertainty principle is accompanied by a spread in the momentum. This state |Q⟩ will
NOT be an eigenstate of the momentum operator, but will instead be a linear combination of
states |p⟩, which are eigenstates of the momentum operator P = −iℏ∇.
dp
Z
|Q⟩ = g(p − Q)|p⟩, (2.53)
2πℏ
g(p − Q) = ⟨p|Q⟩.

For our purposes we will assume g to be of a Gaussian form,

−(p − Q)2
 
g(p − Q) = α exp . (2.54)
4∆2
As an exercise, one should check that this resulting wave packet is properly normalized by

ℏ
r
1/4
α = (2π) . (2.55)
∆
With this definition, the width of the wave packet in momentum space is ∆, i.e.
1
Z
2
⟨Q|(P − P̄ ) |Q⟩ = dp (p − P̄ )2 [g(p − Q)]2 (2.56)
2πℏ
= ∆2 ,
P̄ ≡ ⟨Q|P|Q⟩.

To understand the spatial shape of the packet, one can see that
dp
Z
ψQ (x) = ⟨x|Q⟩ = ⟨x|p⟩⟨p|Q⟩ (2.57)
2πℏ
dp ipx/ℏ
Z
= e g(p − Q)
2πℏ
r
α∆ 2 iQx/ℏ −x2 ∆2 /ℏ2
= e e ,
ℏ rπ
2∆ 1 (2∆)2 x2
 
2
|ψQ (x)| = exp − ,
ℏ 2π ℏ2 2
which is also of Gaussian form, with the spatial spread being ℏ/(2∆).
Thus, the product of the spread in momentum space multiplied by the spread in coordinate
space is ℏ/2, exactly the minimum allowed by the uncertainty principle. If the packet had been
described with eipxℏ → eip(x−x0 )/ℏ , the packet would be centered at x0 instead of the origin.
For the wave packet at arbitrary times,
dp
Z
ψQ (x, t) = e−iEp t/ℏ+ipx/ℏ g(p − Q), (2.58)
2πℏ

40
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

we expect the packet to move in time. The packet has contributions from all momenta which
represent differential additions with a wide variety of phases for each p. At most values of x
these phases vary to the degree that the resulting wave function is near zero. However, at any
given time t there is a point x for which the phases are constant as p is varied near Q. At this
point, the differential contributions add in-phase and the wave function in coordinate space is
a maximum. To find that point we take a derivative of the phase with respect to p at Q, and
require it to be zero.
d
(−iEp t + ipx)|p=Q = 0, (2.59)
dp
dEp
x= t.
dp
Given that dE/dp|p=Q = v, even for relativistic particles, one sees that

x = vt (2.60)

which is not surprising. Here, g(p) was assumed to be a real function, but if it were given a
momentum-dependent phases, eiϕ(p) , then the same condition would lead to

x = vt − ℏdϕ/dp|p=Q , (2.61)

and the additional term would set the position of the wave packet away from the origin, even at
t = 0.
Example 2.5: Wave Packet Reflecting off Finite Barrier
As an example where wave packets are involved, we will consider a packet as described
above, incident on a potential barrier. The goal is to find the time delay of scattering. We
consider a simple square well potential of the form,

0, x > 0
V (x) =
V0 , x < 0
We also assume that the packet is narrow and that V0 is larger than Ek , so the wave packet
should completely reflect (assuming that g(p − k) is narrow).
The incoming wave packet can have the form,
1
Z
ψin (x, t) = dp e−iEp t/ℏ−ipx/ℏ g(p − k).
2πℏ
Such a packet should have a reflection, with the reflected wave packet,
1
Z
ψout (x, t) = − dp e−iEp t/ℏ+ipx/ℏ e+2iδ(p) g(p − k).
2πℏ
The negative sign is chosen so that in the limit of an infinite potential δ(p) = 0. The reflected
packet has the same amplitude, because the flux must be the same both ways given that the
reflection is total. However, there could be a phase factor that might be momentum depen-
dent. The factor of two in the phase is a convention which we will encounter again when we
discuss scattering theory.

41
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

Solving for the phase shift, we assume the solutions in regions I (x > 0) and II (x < 0) are

ψI (x) = e−ipx/ℏ − eipx/ℏ+2iδ(p) ,

p
ψII (x) = Beqx/ℏ , q = 2m(V0 − E).

The boundary conditions give

1 − e2iδ = B
−ip(1 + e2iδ ) = qB.

The solution for the phase shift is

p
tan δ = .
q

Looking for the point x where the phase is stationary, one finds the expression

d
x = vt − 2 δ(p).
dp

Using p2 + q 2 = 2mV0 , one finds

dδ 2mV0
= .
dp q3
From the equation for ψout , one can see that the condition for a stationary phase in the outgo-
ing wave packet becomes

d
{−iEp t/ℏ + ipx/ℏ + 2iδ(p)} = 0, (2.62)
dp
d
∆x = −2ℏ δ(p),
dp
ℏ d d
∆t = 2 δ(E) = 2ℏ δ(E).
vp dp dE

This time delay is relative to the case where V0 = ∞. The same ideas will be applied to spher-
ical waves when we cover scattering theory, where the phase shifts, δ, provide a standard, and
insightful, means for understanding the effects of a potential in scattering.

2.6 The Harmonic Oscillator

The harmonic oscillator provides the basis for much of physics. In fact, quantum field theory
considers every point in space to have it’s own oscillators for every type of quantum field. We
begin with a single oscillator with the Hamiltonian,

P2 mω 2
H = + X 2, (2.63)
2m 2

42
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

where the spring constant is expressed in terms of a frequency ω, k = mω 2 . Here, X and P are
operators.
Dirac’s solution to the problem involves first defining two new operators,
r r r r
mω 1 mω 1
a≡ X +i P, a† = X −i P. (2.64)
2ℏ 2ℏmω 2ℏ 2ℏmω
The operators are known as the annihilation (or destruction) and creation operators respectively
for reasons to be seen below. In some texts, these operators are referred to as ladder operators.
The operators satisfy simple commutation relations,
r r
mω 1
[a, a† ] = 2i [P, X ] = 1. (2.65)
2ℏ 2ℏmω
Furthermore, by substituting the expressions for X and P in terms of a and a† ,

ℏ
r
X = (a + a† ), (2.66)
2mω
ℏmω
r
P = −i (a − a† ),
2
into the Hamiltonian, Eq. (2.63), the Hamiltonian may be expressed as
" r #2 "r #2
1 ℏmω mω 2
ℏ
H = −i (a − a† ) + (a + a† ) (2.67)
2m 2 2 2mω
ℏω 
−a2 + aa† + a† a − (a† )2 + a2 + aa† + a† a + (a† )2

=
4
ℏω
= (aa† + a† a)
2 
1

= ℏω a a +†
.
2

To see that the creation operator does exactly what it sounds like it should do, consider an eigen-
state of the Hamiltonian such that
a† a|n⟩ = n|n⟩. (2.68)
At this point we do not know that the eigenvalues, n, of a† a are integers. By using the commu-
tation relations, aa† = a† a + 1, one can see that

a† a(a† |n⟩) = a† (a† a + 1)|n⟩ (2.69)

†
= (n + 1)a |n⟩

by commuting the a† to the far left. Doing the same with the state a|n⟩ one finds

a† a(a|n⟩) = (aa† − 1)a|n⟩ (2.70)

= (n − 1)(a|n⟩).

43
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

Thus, to within a normalization constant,

a† |n⟩ ∝ |n + 1⟩, a|n⟩ ∝ |n − 1⟩. (2.71)
To calculate the normalization constant Z,
1
a† |n⟩ = √ |n + 1⟩, (2.72)
Z
1
⟨n|aa† |n⟩ =
Z
= ⟨n|(a† a + 1)|n⟩ = n + 1,
s
1
Z= ,
n+1
√
a† |n⟩ = n + 1|n + 1⟩.
Similarly for the destruction operator,
1
a|n⟩ = √ |n − 1⟩, (2.73)
Z
†
⟨n|a a|n⟩ = n,
1
Z= √ ,
n
√
a|n⟩ = n|n − 1⟩.
The operator a† a is referred to as the number operator, and because the energy is expressed in
terms of the number operator, and because that energy must not have arbitrarily small negative
values, the values of n cannot be arbitrarily negative. Thus, the sequence of repeatedly applying
the operator a to some state n must terminate at some point. This only happens if there is a state
with n = 0,
a|n = 0⟩ = 0, (2.74)
which constrains all the n to be integers. With the constraint that n is an integer, the normaliza-
tions can be expressed as
√
(a† )n |0⟩ = n!|n⟩. (2.75)
and the eigenenergies are
En = (n + 1/2)ℏω, (2.76)
n = 0, 1, 2, · · ·
We will see later on that similar tricks are used with angular momentum raising and lowering
operators.
Finally, it is straight-forward to find the ground state wave function if one is sufficiently prescient
to guess that the form of the solution is a Gaussian with some as-yet-undetermined width b.
2 /(2b2 )
ψ0 (x) = e−x . (2.77)

44
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

For the moment, we neglect the normalization. To show that this is a solution we first take
derivatives with respect to x.

d x d2 x2 1
ψ0 (x) = − ψ (x),
2 0
ψ (x) =
2 0
ψ (x) −
4 0
ψ0 (x). (2.78)
dx b dx b b2
Plugging this into the Schrödinger equation,

ℏ2 x2 1 mω 2 x2
 
− − 2 ψ0 (x) + ψ0 (x) = Eψ0 (x). (2.79)
2m b4 b 2

allows one to determine b and E by inspection.

ℏ
r
1
b= , E = ℏω. (2.80)
mω 2
Indeed, the energy is what was expected from the earlier arguments with creation and destruc-
tion operators. One may also calculate the normalization Z by enforcing the constraint
Z ∞
−1/2 −x2 /(2b2 ) 2 2
ψ0 (x) = Z e , Z= dxe−x /b . (2.81)
−∞

This gives Z = (π 1/2 b).

If one were to consider an n dimensional problem,

ℏ2 d2 ℏ2 d2 ℏ2 d2 mω 2 mω 2 mω 2
HΨ = − Ψ− Ψ··· − Ψ+ x21 Ψ + x22 Ψ · · · + x2n Ψ,
2m dx21 2m dx22 2m dx2n 2 2 2
(2.82)
one would write the solution as

Ψ(x1 , x2 , · · · xn ) = ψ1 (x1 )ψ2 (x2 ) · · · ψn (xn ), (2.83)

where ψi are the solutions to the 1-d Schrödinger equation with eigenenergies ℏω(ni + 1/2),
and the total energy is X
E = ℏω (ni + 1/2), (2.84)
i

with the ground state energy being for n = 0, E0 = ℏω/2. For a three-dimensional oscillator,
one can consider the x, y and z motion as three separate oscillators, and the ground state energy
would be 3ℏω/2.
Example 2.6: Practice with Creation and Destruction Operators
Evaluate the following matrix element:

⟨m|aa† a† aa|n⟩.

Solution:

45
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

√ √
Using the facts that a|n⟩ = n|n − 1⟩ and a† |n⟩ = n + 1|n + 1⟩,
√
⟨m|aa† a† aa|n⟩ = n⟨m|aa† a† a|n − 1⟩
√
= (n − 1) n⟨m|aa† |n − 1⟩
= (n − 1)n⟨m|a|n⟩
= (n − 1)n3/2 ⟨m|n − 1⟩
= (n − 1)n3/2 δm,n−1 .

2.7 Propagators, Green’s Functions and Integral Equations

A propagator is merely an incarnation of an evolution operator and is often confused with being
the same as a Green’s function, which it sometimes is. A propagator in coordinate space is
defined as
−iH(t − t′ )
 
′ ′
K(x, t; x , t ) ≡ ⟨x| exp |x′ ⟩, (2.85)
ℏ
and represents the amplitude from evolving from a position |x′ ⟩ at time t′ to a position |x⟩
at time t. Rather than propagating from x′ to x, one can also define propagators from some
momentum state p′ to some final momentum state p. Propagators form the basis of diagram-
matic perturbation theory, and in that application provide a physical understanding of specific
processes.
If the eigenstates, |a⟩, of H are known the propagator may be written as
X
K(x, t; x′ , t′ ) = e−iϵa t/ℏ ⟨x|a⟩⟨a|x′ ⟩. (2.86)
a

Because K is the evolution operator, one sees that determining the propagator is equivalent to
solving the Schrödinger equation,
Z
ψ(x, t) = d3 x′ K(x, t; x′ , t′ )ψ(x′ , t′ ). (2.87)

In fact, the propagator is a solution of Schrödinger’s equation for t > t′ with the boundary
condition that at t′ the wave function is at x′ ,
ℏ2 ∇2
 
∂
− + V (x) K(x, t; x′ , t′ ) = iℏ K(x, t; x′ , t′ ), (2.88)
2m ∂t
while being zero for t < t′ and equal to δ 3 (x − x′ ) when t = t0+ .
We now consider the simple case of a free particle in one dimension. In that case the eigenstates
of the Hamiltonian are momentum states and
1 ip · (x − x′ ) ip2 (t − t′ )
 Z  
′ ′
K0 (x, t; x , t ) = dp exp − (2.89)
2πℏ ℏ 2mℏ
im(x − x′ )2
 
m
r
= exp .
2πiℏ(t − t′ ) 2ℏ(t − t′ )

46
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

The integral in the last step was performed by completing the square. Note that the phase in the
exponential looks (1/2)mv 2 t/ℏ, where the velocity is given by ∆x/∆t. A similar expression
can easily be found in higher dimension. To see that this form approaches δ(x − x′ ) as t → t′ ,
one can check:

1. that dxK(x, t; x′ , t′ ) = 1 for any time t > t′ .

R

2. that at small time differences the phase oscillates infinitely quickly with varying x except
when x = x′ . This rapid variation of phases would cause any integral of the propagator
over a finite interval to vanish unless that interval includes x = x′ .

Adding a potential makes finding the propagator much more difficult. One can see that if the
propagator is expressed as
1
Z
′ ′ ′ ′
K(x, t; x , t ) = K0 (x, t; x , t ) + dt′′ dx′′ K0 (x, t; x′′ , t′′ )V (x′′ , t′′ )K(x′′ , t′′ ; x′ , t′ ),
iℏ
(2.90)
the Schrödinger equation will be satisfied. This expression for the propagator can be viewed
as the integral representation of Schrödinger’s wave equation. Note that the second propagator
in the right side is the “full” propagator which means that solving the equation is not simply a
matter of performing an integral. Instead, one must find a self-consistent solution.
By replacing K with K0 in the right-hand side of Eq. (2.90), one has found the “first-order
perturbative corrections to the propagator”. We will see this again later.

2.8 Phase Space Density and Wigner Functions

We have seen that |ψ(x)|2 = |⟨x|ψ⟩|2 can be identified as the density dN/dx, and |ψ(p)|2 =
|⟨p|ψ⟩|2 can be identified as the number per dp/(2πℏ). One can also get a measure of the
number per dpdx, which simultaneously expresses a probability of observing the particle with
both a specific momentum and density by performing what is called a Wigner transform of the
wave function,
Z
f (p, x) = dδx ψ ∗ (x + δx/2)ψ(x − δx/2)eipδx/ℏ . (2.91)

With this definition,

dp dp
Z Z Z
∗
f (p, x) = dδx ψ (x + δx/2)ψ(x − δx/2) eipδx/ℏ (2.92)
2πℏ 2πℏ
= ψ ∗ (x)ψ(x).
Further, one can express f (p, x) in terms of ψ(p),
dp1 −ip1 (x+δx/2)/ℏ ∗ dp2 ip2 (x−δx/2)/ℏ
Z Z
f (p, x) = dδx e ψ (p1 ) e ψ(p2 )eipδx/ℏ . (2.93)
2πℏ 2πℏ
Integrating over δx,
dδp
Z
f (p, x) = ψ ∗ (p + δp/2)ψ(p − δp/2)e−iδpx/ℏ , (2.94)
2πℏ

47
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

and integrating over x,

Z
dx f (p, x) = ψ ∗ (p)ψ(p). (2.95)

Thus, when one integrates over either x, as seen in Eq. (2.92), or if one integrates over p, as in
Eq. (2.95), it appears that f (p, x) behaves like the phase space density,

dN f (p, x)
= . (2.96)
dpdx 2πℏ

This is called the phase space density, or occupancy, because it can be related to the number of
particles per single-particle level. To demonstrate this, we consider the limit of a large box of
length with an infinite square well potential confining the particle to within 0 < x < L, the
boundary conditions give solutions of

ψn (x) = sin(pn x/ℏ) = sin(nπx/L), n = 1, 2, 3, · · · , (2.97)

pn = nπℏ/L.

The number of states with momentum p within some range ∆p is

L
∆N = ∆p. (2.98)
πℏ
If one uses plane waves, which allows the states p to be both positive or negative, one alters the
expression to

L
∆N = ∆p. (2.99)
2πℏ
Because f (p, x) gives the number of particles per unit ∆x∆p/(2πℏ), and the number of states
is ∆x∆p/(2πℏ), f (p, x) can also be viewed as the number of particles per single particle level.
Of course, this interpretation is based on the idea that ∆x = L is large enough so that ∆n can
be treated as a continuous function.
The interpretation of f (p, x) might seem to violate the uncertainty principle, which forbids the
simultaneous measurement of p and x. However, simply having an expression for f (p, x) does
not mean that you can measure both p and x simultaneously. In fact f (p, x) can be negative, or
greater than unity, for some values of p and x, and for some wave functions. This disqualifies
it as a probabilistic measure because probabilities must lie between zero and unity. It can only
be identified probabilistically in the classical limit, where the averaging is over a sufficiently
large range of p and x to include many quantum state. Nonetheless, even away from the clas-
sical limit, f (p, x) has the property that if it is integrated over p one obtains the probabilistic
density in coordinate space, and if one integrates over x, one finds the probabilistic density in
momentum space.
Example 2.7: Wigner Transform of Harmonic Oscillator Wave Functions
Here, we find the phase space density, a.k.a. the Wigner transform, of the ground state of the

48
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

harmonic oscillator.
1 2 2
ψ0 (x) = e−x /2b ,
π 1/4 b1/2
1
Z
2 2 2 2
f (p, x) = 1/2 dδx eipδx/ℏ e−(x+δx/2) /2b e−(x−δx/2) /2b
π b
1
Z
−x2 /b2 2 2
= 1/2 e dδx eipδx/ℏ e−δx /4b
π b
2 2 2 2 2
= 2e−x /b −p b /ℏ .

One can check the normalization by showing that

dpdx
Z
f (p, x) = 1.
2πℏ

2.9 Numerically Solving Schrödinger’s Equation

Here, we provide algorithms for solving one-dimensional Schrödinger’s equations for two types
of problems. For both types we consider potentials that are infinite for x < 0. This bound-
ary condition also applies for spherical wave decompositions in three dimensions, where for a
particular spherical wave the problem reduces to a one-dimensional problem with an effective
potential for angular momentum, but because there are no negative radii, one sets a bound-
ary condition that ψ(r = 0) = 0. The first type of problems to be considered is solving for
a bound state energy for a finite (for x > 0) potential that vanishes beyond some position,
V (x > a) = 0, and the second type of problem involves finding the phase shift for the same
kind of potential.
To find a bound state energy numerically, one must first guess at a binding energy B. For the
sake of convenience, instead of B we will refer to q, where ℏ2 q 2 /2m = B. For x > a, the
bound state wave function behaves as Ae−qx , where A is an unknown constant. For x < a, we
solve the problem numerically on a grid of resolution ∆x,

ψ(x = 0) = 0, ψ(x = ∆x) = ψ1 . (2.100)

Here, ψ1 can be any value. Normalization will be chosen later. For the sake of convenience, one
might choose a real value such as dx. Next, one uses the discrete representation of Schrödinger’s
equation,

−1 2mV (xn )
 
2
(ψn+1 − 2ψn + ψn−1 ) = −q − ψn , (2.101)
(∆x)2 ℏ2

to solve for ψn+1 ,

2mV (xn )
 
n+1 2 2
ψ = 2ψn − ψn−1 − (∆x) −q − ψn . (2.102)
ℏ2

49
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

One can iterate this forward, first solving for ψ2 , then ultimately for ψN , where N = a/dx.
At that point one can consider the boundary conditions. By comparing logarithmic derivatives,
(dψ/∆x)/ψ, one can ignore the constant A, and see that
ψN +1 − ψN −1
= −q. (2.103)
2(∆x)ψN
The difficulty is finding the value of q that satisfies this boundary condition at x = a. This
involves guessing a value, then comparing the equivalence. One then adjusts q until one finds
a satisfactory fit. This can be done with Newton’s method, or with a root-finding package. The
bound state energy is then E = −B = −ℏ2 q 2 /(2m).
The second class of problem to be considered is that of finding the phase shift for a given in-
coming wave number k. This is easier because, the energy is given. Again, discretize the wave
function, and assume that the wave function behaves like the incoming wave, e−ikx ,

ϕN +1 = e−ik(N +1)∆x , ϕN = e−ikN ∆x . (2.104)

One then iterates forward,

2mV (xn )
 
2 2
ϕn−1 = 2ϕn − ϕn+1 − (∆x) q − ϕn . (2.105)
ℏ2
Ultimately, one finds ϕ0 , which of course does not satisfy the boundary condition. Because
Schrödinger’s equation is real, one also knows that ϕ∗ is also a solution, also not satisfying the
boundary condition. However, one can take the combination,

ψn = ϕn − e2iδ ϕ∗n , (2.106)

choosing the phase shift to fit the boundary condition,

ϕ0
e2iδ = . (2.107)
ϕ∗0
With this choice of δ, the net wave function at x = 0 vanishes. This method even works for cases
where the potential blows up at the origin, although one may have to extrapolate the phase to
x = 0 from x = ∆x and x = 2∆x.

2.10 Combining Coordinate Space and Spin

The overlap ψ(x) = ⟨x|Ψ⟩ describes the overlap of the a state |Ψ⟩ with a state located at x.
The wave function ψ(x) can be thought of as having values at each point x. For spin-half, one
represents the overlap with the | ↑⟩ and | ↓⟩ states with a two-component vector. Of course,
states can vary in both position and spin, and the spin might be correlated with position. In that
case, one must consider the overlap, ⟨s, x|Ψ⟩, where s refers to the spin. In general, it can be
represented by a two-dimensional vector, with each component depending on x,
 
ψ↑ (x)
⟨s, x|Ψ⟩ = . (2.108)
ψ↓ (x)

50
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

The functions ψ↑ and ψ↓ can be any functions as long as the overall wave function is normal-
ized,
Z  
⟨Ψ|Ψ⟩ = dx ψ↑∗ (x)ψ↑ (x) + ψ↓∗ (x)ψ↓ (x) = 1. (2.109)

If the position and spin information are independent, i.e. the upper and lower components have
the same spatial wave function, one might represent the overall wave functions as

⟨s, x|Ψ⟩ = ψ(x)χ, (2.110)

where χ is some two component wave function. But, this last form is not as general at that in
Eq. (2.108), which accommodates the possibility that the spin features varies with position. For
particles of higher spin, one needs to increase the dimensionality of the vector. For example,
for massive spin-one particles, there are three spin states, m = −1, 0, 1 and one needs to use
three-dimensional vectors.

2.11 Exercises
1. Proof that ℏ = 0: Consider a normalized momentum eigenstate of the momentum operator
|q⟩, i.e. P|q⟩ = q|q⟩ and ⟨q|P = ⟨q|q. Consider the expectation,

⟨q|(PX − X P)|q⟩ = ⟨q|(qX − X q)|q⟩

= q⟨q|(X − X )|q⟩ = 0.

However the commutation relation, PX − X P = −iℏ, so we also have

⟨q|(PX − X P)|q⟩ = −iℏ.

Comparing the two equations, ℏ = 0.

What went wrong?
2. Prove that the average kinetic energy is always positive, i.e.
ℏ2 ∂x2 ℏ2
Z
⟨− ⟩=− dx ψ ∗ (x)∂x2 ψ(x) > 0.
2m 2m

3. Consider the one-dimensional potential,


 0, x < −a
V (x) = −V0 , −a < x < a
0, x>a


V0

a
For fixed a, find the minimum V0 for the number of bound states to equal or exceed 1,2,3....

51
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

4. Consider a particle of mass m under the influence of the potential,

ℏ2
V (x) = V0 θ(−x) − βδ(x − a), V0 → ∞, β > 0.
2m
(a) Find the transcendental equation for the energy of a bound state?
(b) What is the minimum value of β for a ground state?
(c) For increasing β can one find more than one bound state?

5. Consider a plane wave moving in the −x̂ direction to be reflected off the delta function
potential, For(x > a) the plane wave will have the form

e−ikx − e2iδ eikx .

(a) Find the phase shift δ as a function of ka, and plot for βa = 0.5 and for 0 < ka < 10.
Because addition of nπ to the phase shift is arbitrary, translate all phases to angles
between zero and π.
(b) Repeat for βa = 0.99, 1.01, 1.5.

6. Consider a particle of mass m interacting with a repulsive δ function potential,

ℏ2
V (x) = βδ(x).
2m
Consider particles of energy E incident on the potential.

(a) What fraction of particles are reflected by the potential?

(b) Show that the currents forx > and for x < 0 are the same.

7. Consider a three-dimensional harmonic oscillator with quantum numbers nx , ny and nz .

How many states are there with a given N = nx + ny + nz ? Find a closed expression (no
sum). Test it for all n ≤ 3.
8. Calculate ⟨0|aaa† aa† a† |0⟩ and ⟨n|a† a† a† a|m⟩.
9. Find ψ1 (x), the wave function of the first excited state by applying a† , defined in Eq. (2.64),
to the ground state.
10. Consider a particle of mass m in a harmonic oscillator with spring constant k = mω 2 .

(a) Write the momentum and position operators for a particle of mass m in a harmonic
oscillator characterized by frequency ω in terms of the creation and destruction oper-
ators.
(b) Calculate ⟨n|X 2 |n⟩ and ⟨n|P 2 |n⟩. Compare the product of these two matrix ele-
ments to the constraint of the uncertainty relation as a function of n.
(c) Show that the expectation value of the potential energy in an energy eigenstate of the
harmonic oscillator equals the expectation value of the kinetic energy in that state.

11. (a) What is the representation of the position operator in the momentum basis – how is
⟨p|X |Ψ⟩ related to ⟨p|Ψ⟩? Use the completeness relation, dx|x⟩⟨x| = I and the
R

fact that ⟨p|x⟩ = e−ipx/ℏ .

52
PHY 851/852 2 COORDINATE AND MOMENTUM SPACE

(b) Suppose that the potential is v(x) = (k/2)x2 . What is the Schrödinger equation
written in momentum space; i.e. what is the equation of motion of the amplitude
⟨p|Ψ(t)⟩?
12. Consider a potential
0, x < −a
V (x) = u(x), −a < x < a
0, x>a
where u(x) is an arbitrary real function. Consider a wave incident from the left. Suppose
that the transmission amplitude, defined as the ratio of the transmitted wave at x = a
to the incident wave at x = −a, is S(E). Now consider a wave incident from the right.
Show that the transmission amplitudes, |S(E)|, are the same for both directions. (Hint: the
Schrödinger equation in this case is a real equation, so the complex conjugate of a solution is also a
solution.)
13. Consider the creation and annihilation operators for the harmonic oscillator in the Heisen-
berg representation, a† (t) and a(t).

(a) Derive and solve the equations of motion for a(t) and a† (t) for the harmonic os-
cillator. I.e., first find (d/dt)a(t) and (d/dt)a† (t). Then write the solution of the
differential equation given a(t0 = 0) = a and a† (t = 0) = a† .
(b) Calculate [a(t), a† (t′ )].

14. Calculate the correlation function ⟨0|x(t)x(t′ )|0⟩ for the harmonic oscillator where |0⟩ is
the harmonic oscillator ground state, and x(t) is the position operator in the Heisenberg
representation. Hint: use the expressions for a(t) and a† (t) from the previous problem.
Then solve for the equations of motion for both x(t) and p(t).
15. What are the matrix elements of the operator 1/|⃗
p| in the position representation? That is,
find
1
⟨r| |r′ ⟩.
|p|
Work the problem in three dimensions.
16. Calculate the Wigner transform f (p, x) for a particle in the ground state of an infinite
square well potential,

 ∞, x<0
V (x) = 0, −a/2 < x < a/2 .
∞, x>a


Are there any regions with phase space densities either greater than unity or less than
zero?

53
PHY 851/852 3 CHARGED PARTICLES IN ELECTROMAGNETIC FIELDS

3 Charged Particles in Electromagnetic Fields

3.1 Vector and “Scalar” Potentials

⃗ and the scalar potential Φ,
Electric and magnetic fields are determined by the vector potential A

⃗
1 ∂A
⃗ = −∇Φ −
E ⃗ = ∇ × A.
, B ⃗ (3.1)
c ∂t

The units of eΦ and eA ⃗ are both energy. This is consistent with cgs units, where the magnetic
and electric fields have the same units. The coupling e must change sign if the charge of the
particle is reversed, i.e. this sign would be reversed for an electron.
To add electromagnetic interactions to the Schrödinger equation, one begins with a Hamiltonian
where such interactions are missing,

(−iℏ∇)2
iℏ∂t = + V (⃗
r ), (3.2)
2m
where V (r) refers to potentials other than those from electromagnetism. Next, one performs
“minimal substitution”, which is to make the replacements

⃗
−iℏ∇ → −iℏ∇ − eA/c, (3.3)
iℏ∂t → iℏ∂t − eΦ.

Relativistically, the “scalar” potential is the zeroth component of the four-vector, A0 = Φ, so the
principal of minimal substitution becomes iℏ∂µ → iℏ∂µ − eAµ . Thus,there is nothing “scalar”
about the scalar potential, as it is not invariant under Lorentz transformations. Including the
electromagnetic interaction, Schrödinger’s wave equation becomes:
h i2
⃗ r , t)/c
−iℏ∇ − eA(⃗
r , t) = [iℏ∂/∂t − eΦ(⃗
ψ(⃗ r , t)] ψ(⃗
r , t), (3.4)
2m

r , t) = iℏ∂t ψ(⃗
Hψ(⃗ r , t),
!2
1 ⃗
eA
H = ⃗−
P + eΦ(⃗
r , t)
2m c
"  2 #
1 e e
= P2 − ⃗ ·A
(P ⃗+A
⃗ · P)
⃗ + A2 + eΦ(⃗
r , t).
2m c c

Solving for the equations of motion of the displacement operator allows one to identify an oper-

54
PHY 851/852 3 CHARGED PARTICLES IN ELECTROMAGNETIC FIELDS

ator that represents the velocity,

d⃗
r [⃗
r , H]
= (3.5)
dt iℏ
d 1
ri = {ri (Pj − eAj /c)(Pj − eAj /c) − (Pj − eAj /c)(Pj − eAj /c)ri }
dt 2miℏ
1
= {ri (Pj − eAj /c)(Pj − eAj /c)
2miℏ
− (Pj − eAj /c)ri (Pj − eAj /c) + (Pj − eAj /c)(iℏδij )}
(Pi − eAi /c) Πi
= = ,
m m
⃗ ≡P
Π ⃗ − eA/c.
⃗

Here, Π⃗ is the canonical momentum, i.e. the mass multiplied by the velocity. It is the same
combination of P ⃗ and A ⃗ that appeared in kinetic energy term of the Hamiltonian in Eq. (3.4).
The momentum P ⃗ = −iℏ∇ is NOT the operator associated with the mass multiplied by the
velocity in the classical limit. Instead, the canonical momentum Π ⃗ plays that role. In fact, the
current density from the previous chapter is also redefined,

−ieℏ ∗ ⃗
e2 A
⃗j(⃗
r , t) = [ψ (⃗ r , t) − (∇ψ ∗ (⃗
r , t)∇ψ(⃗ r , t)] −
r , t))ψ(⃗ r , t)|2 .
|ψ(⃗ (3.6)
2m mc
Showing that this definition is consistent with the equation of continuity is left as a homework
problem.
⃗
One can also solve for dΠ/dt, but first we calculate the commutator,

⃗ − eA)
[Πi , Πj ] = [(P ⃗ i , (P
⃗ − eA/c)
⃗ j] (3.7)
= (iℏ) (e∂i Aj /c − e∂j Ai /c) = iℏeϵijk Bk /c.

Moving forward,

[Π2i , Πj ] = Πi [Πi , Πj ] + [Πi , Πj ]Πi (3.8)

X
[Π2i , Πj ] = (iℏe)ϵijk (Bk Πi + Πi Bk )/c
i
d i
⃗ =
Π ⃗
[H, Π]
dt ℏ
e  
= ⃗ ⃗ ⃗ ⃗
Π × B − B × Π − ∇Φ.
2mc
⃗
This becomes the usual, dP/dt ⃗ + e⃗
= eE ⃗
v × B/c, ⃗ with Π
if one replaces P ⃗ and if one ignores
⃗ and Π
the fact that B ⃗ = m⃗ v might not commute. They do not commute if B ⃗ depends on ⃗ r
⃗
because P has a gradient.

55
PHY 851/852 3 CHARGED PARTICLES IN ELECTROMAGNETIC FIELDS

3.2 Gauge Invariance

Gauge invariance in electromagnetism shows that if one alters the vector and scalar potentials
in the following manner,
1 ∂Λ(⃗
r , t)
⃗ r , t) → A(⃗
A(⃗ ⃗ r , t) + ∇Λ(⃗
r , t), Φ(⃗
r , t) → Φ(⃗
r , t) − , (3.9)
c ∂t
⃗ and B
that E ⃗ are unchanged. However, the Hamiltonian is a function of A ⃗ and Φ, not E ⃗ and B, ⃗
which makes one question whether physics is invariant under a gauge transformation. Here, we
show that even though the changes from the gauge transformation to A ⃗ and Φ indeed yield a
new Hamiltonian, the solutions are identical to those of the original Hamiltonian after applying
a simple phase factor,  
ieΛ(⃗ r , t)
ψ(⃗ r , t) → exp ψ(⃗ r , t). (3.10)
ℏc
If one now considers the canonical momentum operator, altered by a gauge transformation,
acting on a wave function, also altered by gauge invariance, one can see that the result is the
same,
 
ieΛ(⃗r , t)
(−iℏ∇ − eAc ⃗ − e∇Λ(⃗ r , t)/c) exp ψ(⃗ r , t) (3.11)
ℏc
  
ieΛ(⃗
r , t) ieΛ(⃗r , t)
= exp ⃗
−iℏ∇ − eAc − e∇Λ(⃗ r , t)/c − iℏ∇ ψ(⃗
r , t)
ℏc ℏc
 
ieΛ(⃗
r , t) h i
= exp −iℏ∇ − eAc ⃗ ψ(⃗ r , t).
ℏc
Following through with more factors of Π one can see that ⟨ϕ|Πn |ψ⟩ will be unchanged by the
gauge transformation as long a both the operator Π and the states |ψ⟩ and ϕ⟩ are transformed
accordingly.
Similarly the term iℏ∂t − eΦ acting on the altered wave function also shows an invariance.
1
   
ieΛ(⃗
r , t)
iℏ∂t − eΦ + ∂t Λ(⃗ r , t) exp ψ(⃗
r , t) (3.12)
c ℏc
 
ieΛ(⃗
r , t)
= exp [iℏ∂t − eΦ(⃗r , t)] ψ(⃗
r , t).
ℏc

Showing that the charge and current densities are unaffected by the transformation is left as a
homework problem.

Aside: Checking Dimensions

Checking dimensions for relations involving electromagnetism can be painful. Here, the nota-
⃗ and Φ have the same dimension. Both eA
tion is consistent with cgs units in that A ⃗ and eΦ have
dimensions of energy. Both B ⃗ and E ⃗ also have the same dimensions, as the dimensions of eE⃗
and eB ⃗ are both energy per length. In these units, magnetic forces are given as e⃗ ⃗
v × B/c.

56
PHY 851/852 3 CHARGED PARTICLES IN ELECTROMAGNETIC FIELDS

3.3 Interaction with Uniform Magnetic Fields

A constant magnetic field in the z direction can be described with the vector potentials,

Az = 0, Aρ = 0, Aϕ = ρB/2, (3.13)

which has azimuthal symmetry about the z axis as the vector potential winds around the z axis.
Through a gauge transformation, one can also rewrite the vector potential as

Ay = Bx, Ax = 0, Az = 0, (3.14)

which gives the identical magnetic field. This appears to violate the azimuthal symmetry but
has the advantage is that it is easy to solve. The wave function can be written in the form

ψ(x, y, z) = eikz z+iky y ϕ(x) (3.15)

because both Py and Pz commute with the Hamiltonian.

The differential equation for ϕ then turns out to be

ℏ2 1
Exy ϕ(x) = − ∂x2 ϕ(x) + (ℏky − eBx/c)2 ϕ(x) (3.16)
2m 2m
ℏ2 e2 B 2
=− ∂x2 ϕ(x) + (x − x0 )2 ϕ(x),
2m 2mc2
x0 = ℏcky /(eB),
E = Exy + ℏ2 kz2 /2m.

This is the harmonic oscillator Hamiltonian with a spring constant mω 2 where

eB
ω= . (3.17)
mc
The harmonic oscillator frequency is the same as the orbital frequency for a classical particle in
a magnetic field of strength B, referred to as the cyclotron frequency.
The solutions thus look like a particle whose x position is centered about x0 , which is deter-
mined by ky and whose solutions in the y and z directions are eigenstates of the momentum
operators. It seems odd that a particle with circular motion would be an eigenstate of Py . The
solution to this paradox is that if Py is constant, it does not imply that vy = (ℏky − eAy /c)/m
is constant. Instead,
mvy + eBx/c = ℏky , (3.18)
where py is the eigenvalue of Py . In fact this does describe circular motion centered about a
point (xc , yc ) with frequency ω = eB/mc. To see this consider the general form for circular
trajectory of frequency ω,

x = x0 + R cos(ωt + ϕ), y = y0 + R sin(ωt + ϕ), (3.19)

whose velocities are

vx = −ωR sin(ωt + ϕ), vy = ωR cos(ωt + ϕ). (3.20)

57
PHY 851/852 3 CHARGED PARTICLES IN ELECTROMAGNETIC FIELDS

Inserting the expressions for x0 and ω in Eq.s (3.16) and (3.17) into Eq.s (3.19) and (3.20), one
sees that that Eq. (3.19) is simply a statement of circular motion with the x position centered
around the minimum of the harmonic oscillator, which is determined by ky . The center of the y
motion is arbitrary, as this solution is a linear combination of orbits with different y0 values.
The solutions are thus that of a harmonic oscillator for the motion in the x − y plane multiplied
by a plane-wave form for the z motion. The eigenenergies are

ℏ2 kz2
En (kz ) = + (n + 1/2)ℏω, (3.21)
2m
where ω is the cyclotron frequency described above. Energies are independent of ky . The energy
levels are known as Landau levels. The degeneracy of the levels with respect to ky will play the
central role in describing the integral quantum Hall effect which we will discuss next semester.

3.4 Motion with Uniform Electric and Magnetic Fields

It is interesting to study the consequences of adding an electric field directed perpendicular to
the magnetic field. Consider the case where the field is in the x direction. The Hamiltonian picks
up an extra term, −eEx. In this case the differential equation for ϕ becomes

ℏ2 1
Exy ϕ(x) = − ∂x2 ϕ(x) +(ℏky − eBx/c)2 ϕ(x) − eExϕ(x) (3.22)
2m 2m
ℏ2 2 e2 B 2 mc2 E 2 ℏky E
 
2
=− ∂x ϕ(x) + (x − x 0 ) ϕ(x) − ϕ(x) − ϕ(x),
2m 2mc2 2 B B
ℏcky mc2 E
x0 = + .
eB eB 2
The x position averaged over time would simply be x0 , the center of the harmonic oscillator.
One can also calculate the average vy ,
Πy ℏky eBx
vy = = − (3.23)
m m mc
ℏky eBx0
v̄y = −
m mc
Ec
=− .
B
Here, the fact that the average x was x0 , as given in Eq. (3.22), was used to find the average
velocity in the y direction. In the absence of an electric field, this average would be zero as
expected given that classically, the motion is circular. Once the electric field is added, the particle
moves, on average, in the y−direction with velocity −(E/B)c.
It might seem odd given that the average motion is in the −ŷ direction while the electric field
was applied in the x−direction. However, this is exactly the same result one finds for a classical
trajectory. To gain an understanding of this, one can consider the motion in the absence of an
electric field. This is clockwise circular motion of some radius R. Adding a small electric field in
the x direction accelerates the particle on the upper part (y > 0) half of the trajectory, leading to

58
PHY 851/852 3 CHARGED PARTICLES IN ELECTROMAGNETIC FIELDS

the speed being greatest when x = +R, and the speed being smallest when x = −R. Because
the speed is higher when the particle is moving downward, and lower when the particle is
moving upward, it is not surprising that the average vy is negative. To understand why the
particle does not drift in the direction of the electric field, one has to consider that the magnetic
force acting on the particle when x = +R is stronger than for x = −R because the magnitude
of the magnetic force is proportional to the speed. Thus the average magnetic force acting in the
x direction is negative, which then allows it to cancel the added electric force.
One can also understand the result by considering the relativistic transformation properties of
electromagnetic fields. If one begins with a magnetic field Bz in the z direction, then boosts
in the y direction by a small velocity vy , one generates an electric field in the x direction of
strength Ex = −vy Bz . Hence, if one has an electric field in the x direction, but views it in a
frame moving with velocity vy = −Ex /Bz , one sees no electric field and the motion is purely
circular.

3.5 Path Integrals and the Aharanov-Bohm Effect

One way to perform Quantum Mechanics is through path integrals. Path integrals are usually a
rather inconvenient way to go, but sometimes come in handy. The name refers to the fact that a
sum over all intermediate states can be thought of as a path. To get a better idea we consider the
matrix element
X
⟨ψf |e−iHt |ψi ⟩ = ⟨ψf |e−iHδt |xN −1 ⟩⟨xN −1 | · · · |x2 ⟩⟨x2 |e−iHδt |x1 ⟩⟨x1 |e−iHδt |ψi ⟩
x1 ···xN −1
(3.24)
X
≈ ⟨ψf |(1 − iHδt)|xN −1 ⟩⟨xN −1 | · · · |x2 ⟩⟨x2 |(1 − iHδt)|x1 ⟩
x1 ···xN −1

⟨x1 |(1 − iHδt)|ψi ⟩.

Here, δt = t/N and the approximation becomes exact in the limit of large N .
If the set of states x1 → xn correspond to positions in coordinate space arranged in a mesh of
size δx, the succesive points in the path, xi and xi+1 , are constrained to be neighbors because H
is local. One can appreciate the locality by returning to the picture from Chapter 2, where space
was discretized in steps of δx and the kinetic term of the Hamiltonian only mixed neighboring
sites,
ℏ2
⟨i|H|i + 1⟩ = − , (3.25)
2mδx2
and the potential terms were diagonal. This means that each term in the sum can be thought of
as a continuous trajectory where at each step in time the trajectory either remains at the same
position or moves by ±δx. This motivates the name path integral, though one might more ac-
curately state that one sums over all trajectories rather than over all paths. The classical limit
of quantum mechanics comes from the constraint of choosing the trajectory for which the phase
becomes fixed with respect to small variations of the trajectory. This line of approximation is
what motivates the simple calculations of interference phenomena, such as 2-slit interference, in
elementary physics courses.

59
PHY 851/852 3 CHARGED PARTICLES IN ELECTROMAGNETIC FIELDS

Figure 3.1: Illustration of the Aharanov-Bohm effect: The magnetic field B ⃗ in the solenoid changes the
relative phase of two paths of a charged particle which produce the interference pattern on the screen to
the right. This modification derives from the changing the vector potential A ⃗ in the region of the paths.
The interference pattern shifts despite the fact that the paths never sample a region with non-zero B.⃗

In principle, all matrix elements can be considered in this fashion. Lattice gauge theory, which
provides a powerful tool for numerically calculating the structure of the non-perturbative QCD
vacuum where the trajectory accounts for all possible field configurations at each point in space-
time, is built upon exactly such concepts. Path integral techniques are also often used in statisti-
cal mechanics by making the analogous decomposition,
X
e−βH = e−δβH |αN −1 ⟩⟨αN −1 | · · · |α2 ⟩⟨α2 |e−δβH |α1 ⟩⟨α1 |e−δβH (3.26)
α1 ···αN −1

⃗ also contributes an off-diagonal piece due to the

The interaction with an electromagnetic field A
presence of the term (−ieℏ/mc)A ⃗ · ∇. The off-diagonal term to the Hamiltonian is then
ℏ2 ℏ2
 
ieℏ ieδx
⟨i|H|i + 1⟩ = − − Ai = − 1+ Ai . (3.27)
2mδx2 2mcδx 2mδx2 ℏc

Now, if we consider the evaluation of element ⟨⃗ ri |e−iHt |⃗

rf ⟩ as a sum over trajectories as de-
scribed above. Each must include a series of links between neighboring sites. The effect of the
vector potential is to add a product of the terms
Y ieδx

1+ Ai (3.28)
links
ℏc
to each trajectory.
Thus, the effect of the vector potential for propagation along a specific path from ⃗
r1 to ⃗
r2 is to
modify the matrix element by a phase factor,
 Z 
ie
⟨⃗
r1 |e−iHt
|⃗
r2 ⟩ → ⟨⃗
r1 |e−iHt
|⃗
r2 ⟩ exp ⃗
r·A ,
d⃗ (3.29)
ℏc path

60
PHY 851/852 3 CHARGED PARTICLES IN ELECTROMAGNETIC FIELDS

where the path extends from ⃗ r1 to ⃗

r2 . This added phase depends only on the path taken, but
not the time-dependence of the trajectory because we have no time dependence and because the
⃗ is contracted only with d⃗
vector potential A r . Note that we have taken the liberty to jump to
three dimensions at this point.
Now, we consider a two-slit interference experiment, where between the slits lies a small mag-
netic solenoid. In the upper path, the particle goes above the solenoid, while in the lower path
the particle travels below the solenoid. Neither path samples the region inside the solenoid,
where there is a magnetic field B and a magnetic flux, ΦB = Bσ. Here, σ is the cross sectional
area of the solenoid.
However, there is a vector potential outside the solenoid which must satisfy Stoke’s theorem,
I
A⃗ · dℓ = ΦB = |B|σ.
⃗ (3.30)

The two path’s contributions to the amplitude have phases which differ by this factor, multiplied
by the charge,
eΦB
∆ϕ = ϕupper path − ϕlower path = . (3.31)
ℏc
Here, the difference of the two paths is the same as adding the contribution to the phase around
a closed loop, because subtracting the contribution from the lower path is the same as adding a
contribution where one returns along the lower path. Because the expression for ∆ϕ does not
depend on the precise paths taken from the source to the screen, as long as one path is above
the solenoid and one is below, it exactly sums up how any interference is adjusted between the
upper and lower paths due to the presence of the vector potential A.⃗ When the phase difference
equals π, the maxima and minima of the interference pattern will have traded positions, and
the interference pattern will have shifted by half a fringe. This phenomena has been observed
experimentally with a thin magnetized iron filament called a whisker, as illustrated in Fig. 3.1.
Thus, altering the current through the solenoid shifts the interference pattern even though there
are no magnetic or electric fields along the sampled trajectories. This makes it clear that A ⃗ and
Φ that are the fundamental fields in nature, not E ⃗ and B. ⃗ Even though this means that A ⃗ is
fundamental, gauge transformations do not change any physical observables, even though a
⃗ This follows because the integral A ⃗ · d⃗
H
gauge transformation changes A. ℓ, is also independent
of a gauge transformation, i.e. A⃗→A ⃗ + ∇Λ.
If one considers an eigenstate of a charged-particle’s wave function, one can also consider the
relative phases described by Eq. (3.29). The phase must return to zero, or a multiple of 2π if the
phase is to have a single value return. For a charge q,
iq
Z
r·A
d⃗ ⃗ = 2nπ. (3.32)
ℏc path
By Stoke’s theorem this leads to a quantization of magnetic flux,
2nπℏc
ΦB = . (3.33)
q
Thus, the magnetic field in some fixed area cannot be assigned arbitrary values, but instead
comes in integral units, if one is to have an eigenstate. Note that the two-slit experiment is not

61
PHY 851/852 3 CHARGED PARTICLES IN ELECTROMAGNETIC FIELDS

an eigenstate. This seems peculiar because the integral units of flux depend on the charge. The
Aharanov Bohm effect plays a critical role in Josephson’s junctions. In that case the relevant
charge is 2e, the charge of a Cooper pair. This is not to state that one cannot have a non-integer
amount of flux in a loop, only that the if the charged particles are in an eigenstate that they will
affect the magnetic field in such a way that the flux in the loop settles into quantized levels. The
quantized unit of flux is
2πℏc
Φ0 = = 2.068 × 10−15 Tm2 .
2e

3.6 Exercises
1. Using the equations of motion for the wave function, show that the density and current
defined by

r , t)|2 ,
r , t) = |ψ(⃗
ρ(⃗
−iℏ ∗ ⃗
eA
⃗j(⃗
r , t) = (ψ (⃗ r , t) − (∇ψ ∗ (⃗
r , t)∇ψ(⃗ r , t)) −
r , t))ψ(⃗ r , t)|2 ,
|ψ(⃗
2m mc
satisfies the continuity equation,

∂t ρ + ∇ · ⃗j = 0.

2. Consider a particle of charge e traveling in the electromagnetic potentials

1 ∂Λ(r, t)
A(r, t) = −∇Λ(r, t), Φ(r, t) =
c ∂t
where Λ(r, t) is an arbitrary scalar function.

(a) What are the electromagnetic fields described by these potentials?

(b) Show that the wave function of the particle is given by
 
ie
ψ(r, t) = exp − Λ(r, t) ψ0 (r, t),
ℏc
where ψ 0 solves the Schrödinger equation with Λ = 0.
(c) Let V (r, t) = eΦ(t) be a spatially uniform time varying potential. Show that

ie t
 Z 
′ ′
ψ(r, t) = exp − Φ(t )dt ψ0 (r, t)
ℏ −∞
is a solution if ψ0 is a solution with Φ = 0.

3. For a gauge transformation, described in Eq. (3.9), including the associated the phase
change to the wave function ψ, described in Eq. (3.10),

(a) Show that the charge density eψ ∗ ψ is unchanged by the gauge transformation

62
PHY 851/852 3 CHARGED PARTICLES IN ELECTROMAGNETIC FIELDS

(b) Show that the current

⃗j = 1 [ψ ∗ (−iℏ∇ψ) + (iℏ∇ψ ∗ )ψ] − e Aψ

⃗ ∗ ψ.
2m mc
is unchanged.
(c) Show that ⟨χ|H|ψ⟩ is unchanged in a gauge transformation where Λ is independent
of time.

4. Find (or guess) the function Λ(⃗ r , t) that corresponds to the gauge transformation in Eq.
(3.9) responsible for re-expressing the vector potential in Eq. (3.13) to the form of Eq. (3.14),
and show that both forms give the same magnetic field.
5. The expression for the v̄y in Eq. (3.23) is only valid for non-relativistic velocities, where
|E| << |B|. For a uniform magnetic field B ẑ, with no electric field, consider the form for
the vector potential in Eq. (3.14). Performing a relativistic boost (Lorentz transformation),
but for non-relativistic velocities, in the y direction by a velocity vy , what is the resulting
zeroth component of the vector potential A0 ? Equating this with the electric scalar poten-
tial, express the strength of the resulting electric field in terms of vy and B.
6. In this problem, we reconsider the problem of a charged particle in the presence of both an
electric and magnetic field, but do so in a different gauge. The electron is placed in a region
of constant external magnetic field B directed along the z axis and of constant electric field
E in the y direction.

(a) Choosing the vector potential to lie along the y axis and describe both the electric and
magnetic fields, show that the Hamiltonian may be written in the form,

Pz2 Px2 1
H = + + mω 2 (x − x0 − v0 t)2 ,
2m 2m 2
and find ω, and v0 in terms of E, B, e, m, ky and c, where ℏky is the eigenvalue of
⃗ = −(1/c)∂t A.
Py . Hint: Choose a gauge such that E ⃗
(b) Show that Schrödinger’s equation, i(∂/∂t)Ψ = HΨ is satisfied by the form

Ψ(x, y, z, t) = e−iϵn t/ℏ+imv0 x/ℏ+ikz z+iky y ϕn (x − x0 − v0 t) ,

where ϕn refers to a harmonic-oscillator wave function characterized by the frequency

ω and ϵn = (n + 1/2)ℏω + mv02 /2.

63
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

4 Angular Momentum and Central Potentials

Understanding the role of angular momentum in any quantum problem first requires taking
stock of what rotational symmetries are either satisfied, or approximately satisfied. Because the
understanding of symmetry is based on group theory, this chapter includes an extremely brief
discussion of group theory before we launch into the various technical skills required to classify
states based on angular momenta or to understand conservations and constraints of various
transitions involving such symmetries.

4.1 The Baker-Campbell-Hausdorff Relation

Before our brief foray into group theory, we present the Baker-Camblel-Hausdorff relation, which
is central to demonstrating the group nature of rotations. This relation expresses how two trans-
formations might be expressed as a single transformation.
The Baker-Campbell-Hausdorff relation expresses how the product of two unitary transforma-
tions, eA and eB , are not necessarily the same as the transformation eA+B . Instead, the relation
is
1 1 1
 
A B
e e = exp A + B + [A, B] + [A, [A, B]] − [B, [A, B]] + · · · . (4.1)
2 12 12
If the operators commuted, one would get the simple result eA eB = eA+B . If when commut-
ing A and B one gets an operator C = [A, B] that commutes with both A and B, e.g. C is
proportional to the unit matrix, the result is simpler

eA eB = eA+B eC/2 . (4.2)

For angular momentum [Li , Lj ] = iℏϵijk Lz and, because Lz does not commute with Lx and
Ly , this simpler relation doesn’t apply. Nonetheless, it helps clarify the role of commutators and
is applicable in some other cases.
To demonstrate the relation where C is a constant, we consider the expansion
X (A + B)n
eA+B = . (4.3)
N
n!

Expanding (A + B)n gives all terms with all orderings of the n operators. For example, for
n = 5 one of the terms is ABBAB. We wish to move all the A operators to the left, which
requires commuting them past the B operators. Every time an A operator moves past a B
operator one must add a term where the BA pair is replaced by −C = [B, A]. Using the
binomial theorem, one can then write

(A + B)n X Ai B j n! Ai−1 B j−1 n! 2

Ai−2 B j−2 n!
= + (−C) N̄1 (i, j) + (−C) N̄2 (i, j)
n! i+j=n
n! i!j! n! i!j! n! i!j!
(4.4)
i−ℓ j−ℓ
A B n!
+ · · · + (−C)ℓ N̄ℓ (i, j) + · · ·
n! i!j!

64
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

where N̄ℓ is the average number of ways to pick ℓ AB pairs from an order n term, under the
constraint that the B operators in the pair were initially to the right of the A terms. This number
is simply the number of such independent pairs times (1/2)ℓ to account for the fact that only
half the time does a given pair start off with the BA ordering,
 ℓ
1 i(i − 1) · · · (i − ℓ + 1)j(j − 1) · · · (j − ℓ + 1)
N̄ℓ = . (4.5)
2 ℓ!
One can then factor the exponentials in the expression above to get Eq. (4.1),
X Ai B j (−C/2)ℓ
e(A+B) = (4.6)
ijℓ
i!j! ℓ!
= eA eB e−C/2 .

Example 4.1: Rotations and Translations of Operators

One can rotate around the z axis by an angle α with the operator,
∞
α∂ϕ
X 1
R(α) = e = (α∂ϕ )n .
n=0
n!
This demonstrates how the exponential form for rotations is related to a Taylor expansion.
This can also be written as eiLz ϕ/ℏ . Similarly, one can show how translations by a distance α
are related to eα∂x = eiPx α/ℏ . For an operator B(ϕ), show that with R(α) = eα∂ϕ that
R(α)B(ϕ)R(−α) = B(ϕ + α).
Solution:
First consider
X 1
eα∂ϕ B(ϕ) = (αn ∂ϕ )n B(ϕ) · · ·
n
n!
The additional · · · is to emphasize that there may be quantities further to the right that of
B(ϕ) on which the derivatives act, i.e. there are functions of ϕ to the right of B. Next, one
moves the derivatives from the left to the right. For each term ∂ϕn , one must account for m
occurrences where the derivative acts on B(ϕ) rather than on the quantities to the left.
∞
n X
X αn n!
eα∂ϕ B(ϕ) = (∂ϕn B(ϕ))∂ϕn−m .
m=0 n=0
n! m!(n − m)!
Here, the derivatives inside the parenthesis act only on B(ϕ), and given the substitution
ℓ = m − n,
!
X αm ∂ϕm X αℓ ∂ϕℓ
eα∂ϕ B(ϕ) = B(ϕ)
m
m! ℓ!
!ℓ
X (α∂ϕ )m
= B(ϕ) eα∂ϕ
m
m!
= B(ϕ + α)eα∂ϕ .

65
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

Thus,
eα∂ϕ B(ϕ)e−α∂ϕ = B(ϕ + α)eα∂ϕ e−α∂ϕ = B(ϕ + α).
To derive a more general relation, one could follow the same steps and show that for any
operators A and B,

1 1
eA Be−A = B + [A, B] + [A, [A, B]] + [A, [A, [A, B]]] + · · ·
2 3!

4.2 Three-Dimensional Rotations

In quantum mechanics the angular momentum of a state, or an operator, tells us how the object
behaves under rotation. Rotations are unitary transformations which can be classified as a group,
⃗ ⃗
and the operators that represent such transformations can be written in the form eiL·θ/ℏ , where
⃗ is the angle of the rotation about an axis the θ̂ direction. If an object does not depend
the |θ|
on angle, then a rotation should have no effect on it. These are scalars, and examples are time,
mass, or the dot product of two vectors, u⃗ ·⃗v . Under rotations, both u ⃗ and ⃗v change but the
angle between the two vectors remains the same and u ⃗ ·⃗
v is unchanged.
The rotation group consists of unitary operators,
⃗
α) = eiL·⃗α/ℏ .
R(⃗ (4.7)

To understand why this a rotation, consider the case where α ⃗ is along the z axis. One can then
see that  
∂ ∂ ∂
Lz = xPy − yPx = −iℏ x −y = −iℏ , (4.8)
∂y ∂x ∂ϕ
which means that for a rotation of angle ϕ about the z-axis,

eiLz α/ℏ f (ϕ) = eα∂ϕ (4.9)

α2 ∂ 2
 
∂
= 1+α + · · · f (ϕ)
∂ϕ 2! ∂ϕ2
= f (ϕ + α).

In order to be a true group, two consecutive rotations of α ⃗ must be identical to a single

⃗ and β
rotation ⃗
γ.
⃗ ⃗ ⃗ ⃗
eiL·β/ℏ eiL·⃗α/ℏ = eiL·⃗γ /ℏ . (4.10)
Because the different components of L ⃗ do not commute, it can be non-trivial to find the equiva-
lent single rotation ⃗
γ given α ⃗
⃗ and β.
⃗ =⃗
Given the definition L r×p
⃗, it is straightforward to find the commutation relations,

[Li , Lj ] = iℏϵijk Lk . (4.11)

Our goal in this section is to discuss the requirement of using different operators, Sx , Sy and Sz
to generate rotations, not in coordinate space but in a discrete vector space, meaning that S ⃗ can

66
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

be expressed as matrices. The important requirement for S to be considered a rotation is that

⃗ ⃗ ⃗ ⃗
eiS·β/ℏ eiS·⃗α/ℏ = eiS·⃗γ /ℏ , (4.12)

where the same ⃗ γ results from a given α⃗ and β ⃗ as would have resulted if one had inserted
⃗ = −iℏ⃗
L r × ∇ instead of S ⃗ everywhere above.
We wish to demonstrate that if the components of S ⃗ obey the same commutation laws of the
components of L,⃗ the rotations will be identical. To see this we divide the two rotations into N
smaller rotations with N → ∞.
⃗ ⃗ ⃗ ⃗ ⃗ ⃗ ⃗ ⃗ ⃗ ⃗ ⃗ ⃗
eiS·β/ℏ eiS·⃗α/ℏ = eiS·β/(ℏN ) eiS·β/(ℏN ) · · · eiS·β/(ℏN ) eiS·⃗α/(ℏN ) eiS·⃗α/(ℏN ) · · · eiS·⃗α/(ℏN ) . (4.13)

To find the equivalent single rotation, one must expand each exponential then commute them in
a manner as was done for the Baker-Campbell-Hausdorff lemma earlier. There are of order N 2
such commutations.
For our purposes we wish to consider the inner two exponentials,
⃗ ⃗ ⃗ ⃗ · (⃗ ⃗
eiS·β/(ℏN ) eiS·⃗α/(ℏN ) = 1 + iS α + β)/(ℏN ) (4.14)
1 
⃗
 2 1 ⃗ +O 1
− S⃗ · (⃗
α + β) + ⃗·α
[S ⃗, S ⃗ · β]
2(ℏN ) 2 2(ℏN ) 2 (ℏN )3
1 1
  
= exp i S ⃗ · (⃗ ⃗
α + β)/(ℏN )+i [ ⃗·α
S ⃗ , ⃗ · β]
S ⃗ + O
2(ℏN )2 (ℏN )3
1 1
  
= exp i S ⃗ · (⃗ ⃗
α + β)/(ℏN )− S j ϵ jkℓ α k β ℓ + O
2(ℏN )2 (ℏN )3
αj + βj 1 1
  
= exp iSj − 2
ϵjkℓ αk βℓ +O
(ℏN ) 2(ℏN ) (ℏN )3
1 1
 
= exp i S ⃗ · δ⃗γ +O
ℏ (ℏN )3
⃗ +β
α ⃗ 1
γ=
δ⃗ − ⃗ × β.
α ⃗
N 2(ℏN )2
Because there are of order N 2 such commutations we must perform to find ⃗ γ , we may throw
away all terms of order 1/N 3 or higher. The angle δγ was determined purely from the com-
mutation relations. Initially, there were 2N elements, but after this reduction there are 2N − 1
⃗
elements. One can then repeat the procedure and combine the element eiS·δγ/ℏ with one of the
neighboring elements to reduce it further. Eventually, one can reduce everything to a single ele-
ment described by a single angle ⃗ γ . Expressing ⃗
γ in terms of α
⃗ and β⃗ is difficult, but that is not
the point here. We only need to show that determining the angle depends purely on the com-
mutation relations. Thus, if the commutation relations for the components for S ⃗ are identical to
the commutation relations for L, ⃗ then the equivalent angle ⃗γ will be the same in both cases.
The simplest example of three operators, Sx , Sy and Sz , which generate such rotations is the
2 × 2 representation,
ℏ ℏ ℏ
Sx ≡ σx , Sy ≡ σy , Sz ≡ σz . (4.15)
2 2 2
67
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

As a homework problem, you will able to find the equivalent angle ⃗γ , describing two successive
rotations α ⃗
⃗ and β, by manipulating Pauli matrices. The arguments above also hold for any set of
matrices where the commutation of any two such matrices results in something linearly propor-
tional to the matrices. Although in such cases the mapping of α⃗ and β to ⃗ γ would differ from
this case for three-dimensional rotations, the mapping would also be determined completely by
the commutation relations.
Many physical systems exhibit rotational symmetry, which leads to the fact that angular mo-
mentum is conserved. This can be seen by commuting the rotation operator for an infinitesimal
rotation with the Hamiltonian. If Lz commutes with the Hamiltonian, then the eigenstates of H
can be simultaneously chosen as eigenstates of Lz . However, if the Hamiltonian has terms such
⃗ · S,
as L ⃗ neither Lz nor Sz commutes with the Hamiltonian. However, in that case the operator
Jz ≡ Lz + Sz does commute. We will perform such an example in class later in this chapter.

4.3 A Greatly Abbreviated Guide to Group Theory

Groups are made of elements that represent transformations. The elements of a group are NOT
the object being transformed, they are the transformations themselves. To be classified as a
group, the group of elements Ri must satisfy the following conditions:
1. The combination of any two elements returns an element of the group, Ri Rj = Rk . This
is known as closure. Note this need not be commutative, Ri Rj ̸= Rj Ri .
2. Associativity, (Ri Rj )Rk = Ri (Rj Rk )
3. There must exist an identity element I, such that IRi = Ri and Ri I = Ri .
4. Every element Ri must have an inverse element R−1 −1
i , such that Ri Ri = I.
Often the elements of a group represent symmetry transformations. For example, consider the
symmetries of an equilateral triangle, where the base lies along the x axis. The six group ele-
ments are:

1. R1 , The identity
2. R2 , Rotating by 120◦
3. R3 , Rotating by 240◦
4. R4 , Reflecting about an axis through the
center of the triangle in the 30◦ direction
5. R5 , Reflecting about an axis through the
center of the triangle in the 90◦ direction
The elements of the symmetry group
6. R6 , Reflecting about an axis through the
of an equilateral triangle are the
center of the triangle in the 150◦ direction
identity, two rotations, and three re-
flections about the dashed lines.
The group is defined by the integer coefficients a(i, j) = 1 − 6, where Ri Rj = Ra(i,j) . The
elements a(i, j) can be expressed as an N × N matrix, known as a Cayley Table, https://en.w
ikipedia.org/wiki/Cayley_table.

68
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

Rotations are an example of a Matrix Lie Group, where in this context it simply means that there
are a continuum of elements. For example, rotations can be taken about an axis in any direction,
where the direction of the axis is described by the angles θ and ϕ in spherical coordinates, and
the size of the rotation angle can be varied. Using α̂ to describe the rotation axis and |⃗α| to
describe the rotation angle, the vector α⃗ represents an element of the rotation group. The Lie
groups used in physics are generally of the form
P
iGn θn
R(θ) = e n=1,N . (4.16)
The group elements are denoted by the N −dimensional vector θ, and the group elements are
infinite in number. The matrices Kn are the generators of the group, and are often (but not al-
ways, e.g. the Lorentz transformation) Hermitian matrices, which makes each transformation
unitary. As long as the commutation of any two generators gives a linear combination of gen-
erators, the multiplication of two group elements will be equivalent to a single element. These
commutation rules are encapsulated by what are referred to as structure constants fijk
[Gi , Gj ] = ifijk Gk . (4.17)
The commutation rules for the N generators effectively define the entire group. Rotations are
an example of a group, and for three dimensional rotations one needs three generators, corre-
sponding to the rotations described by the three Euler angles.
To understand the role of the group conditions for the rotations of spin, one can imagine some
quantum mechanical state with spins, and also perhaps with orbital angular momentum. If one
rotates the entire system twice, by angles θ⃗1 and θ
⃗2 , where the direction is that of the rotational
⃗ That single angle Θ
axis, it should be equivalent to a single rotation Θ. ⃗ should not depend on
whether one is rotating a spin wave function, or the orbital wave function. Even though rotating
a spin-1/2 particle uses two-by-two matrices, and rotating a spin-one particle involve three-by-
three matrices, the mapping of θ ⃗1 and θ⃗2 to Θ
⃗ must be the same for both matrices. This will
follow if the generators for the two cases have the same structure constants, even if the genera-
tors have different dimensionality, or in group theory parlance they are different representations
of the same group but are isomorphic.

4.4 ⃗ 2 Operator
The |L|
The angular momentum operator Lz = −iℏ∂/∂ϕ commutes with the Hamiltonian if the
Hamiltonian is invariant to rotations about the z axis, i.e. the Hamiltonian does not depend
on ϕ. Furthermore, if the Hamiltonian is invariant to rotations about any axis, all three compo-
⃗ commute with H. One may then define states which are simultaneously eigenstates
nents of L
of H and Lz , or Lx or Lz . But, one may not necessarily find states which are simultaneously
eigenstates of Lx , Ly and Lz because these operators do not commute with one another.
However, the operator L2 = L2x + L2y + L2z is spherically symmetric and commutes with any
of the three components of L.⃗ This is because L2 is a scalar, and scalars are not changed by
rotation. One may therefore define eigenstates of a spherically symmetric Hamiltonian that are
also eigenstates of both L2 and Lz . We define the eigenvalues in terms of m and ℓ.
Lz |ℓ, m⟩ = mℏ|ℓ, m⟩ , L2 |ℓ, m⟩ = ℓ(ℓ + 1)ℏ2 |ℓ, m⟩. (4.18)

69
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

The curious choice of ℓ(ℓ + 1) will become apparent in the next subsection, where it will be
shown that this choice is consistent with ℓ being an integer.

Example 4.2: The |L| ⃗ 2 operator in Cartesian coordinates

Express |L|⃗ 2 in terms of x, y, z and Px , Py , Pz , with all the coordinate space operators to the
left of the momentum operators.
Solution:
One must be careful of the ordering of the spatial components ri and the momentum operators
Pj , because they do not commute. Given that Li = ϵijk rj Pk ,

⃗ 2 = ϵijk ϵimn rj Pk rm Pn
|L|
= (δjm δkn − δjn δmk )rj Pk rm Pn
= ri Pj ri Pj − ri Pj rj Pi
= r 2 P 2 − iℏ⃗r·P ⃗ − ri rj Pi Pj + 3iℏ⃗ ⃗
r·P
= x2 (Py2 + Pz2 ) + y 2 (Pz2 + Px2 ) + z 2 (Px2 + Py2 )
− 2xyPx Py − 2yzPy Pz − 2zxPz Px + 2iℏxPx + 2iℏyPy + 2iℏzPz .

In classical physics the last three terms could be neglected because they are proportional to ℏ.

4.5 Raising and Lowering Operators for Angular Momentum

We define operators,
L± ≡ Lx ± iLy . (4.19)
These operators have the fortuitous property that

[Lz , L± ] = [Lz , Lx ] ± i[Lz , Ly ] (4.20)

= iℏLy ± Lx
= ±ℏL± ,

which means that

Lz (L± |m⟩) = L± Lz |m⟩ ± ℏL± |m⟩ (4.21)

= (m ± 1)ℏ (L± |m⟩) .

This means that L± effectively change an eigenstate of Lz to a new eigenstate with the eigen-
value either raised or lowered by ℏ.
One can also find the normalization of the new states by noting that

⟨m|L∓ L± |m⟩ = ⟨m|L2x + L2y ± i[Lx , Ly ]|m⟩ (4.22)

= ⟨m|L2x + L2y ∓ ℏLz |m⟩
= ⟨m|L2 − L2z ∓ ℏLz |m⟩
= ℓ(ℓ + 1) − m2 ∓ m ℏ2 .

70
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

⃗ 2 is assigned the value of ℓ(ℓ + 1), which defines ℓ, but we have not
Here, the eigenvalue of |L|
yet shown that ℓ must be a integer or half integer.
In terms of normalized states, the raising and lowering operators acting on a state |m⟩ become

L± |m⟩ = ℏ[ℓ(ℓ + 1) − m2 ∓ m]1/2 |m ± 1⟩. (4.23)

By inspection, one sees that if the sequence of ms is to be finite that they must begin at −ℓ and
end at ℓ, thus giving 2ℓ + 1 values of m for a given ℓ. Because the ms must be separated by unit
steps, ℓ and m must therefore be either integer or half integer.

4.6 Spherical Harmonics

After extensive application of the chain rule, the kinetic energy term in Schrödinger’s wave
equation may be written in spherical coordinates,
∇2 ℏ2 ℏ2
 2
2 ∂ 1 ∂2 1 ∂
  
2
∂ ∂
HK = −ℏ =− + − + sin θ . (4.24)
2m 2m ∂r 2 r ∂r 2mr 2 sin2 θ ∂ϕ2 sin θ ∂θ ∂θ
Furthermore, the components of angular momentum may be written in terms of angular deriva-
tives,
∂
Lz = −ixPy + iyPx = −iℏ , (4.25)
∂ϕ
 
∂ ∂
Lx = −iyPz + izPy = −iℏ − sin ϕ − cot θ cos ϕ ,
∂θ ∂ϕ
 
∂ ∂
Ly = −izPx + ixPz = −iℏ cos ϕ − cot θ sin ϕ ,
∂θ ∂ϕ
 
±iϕ
∂ ∂
L± = −iℏe ±i − cot θ .
∂θ ∂ϕ

Because Lz = −iℏ∂ϕ , one can see that a Hamiltonian with rotational invariance about the z
axis, i.e. no ϕ dependence, will commute with Lz . Thus, the conservation of Lz is tied to az-
imuthal symmetry. Further, the conservation of Lx or Ly depends on whether the Hamiltonian
is invariant under rotations about the x or y axes.
Using the relation, L2 = L2z + L+ L− + iℏLz , one can see that the kinetic energy may be written
as
ℏ2 2 ℏ2
 2
2 ∂ L2

∂
− ∇ =− + + , (4.26)
2m 2m ∂r 2 r ∂r 2mr 2
( 2 )
1 1

∂ ∂ ∂
⃗ 2 = −ℏ2 −
|L| − sin θ .
sin2 θ ∂ϕ sin θ ∂θ ∂θ

The last term reproduces the usual relation from classical physics for the effective centrifugal
potential. If the potential is spherically symmetric, the Hamiltonian commutes with each com-
ponent of L because they only involve angular derivatives, and each component of L commutes

71
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

with L2 . Thus, we may write the solution as an eigenstate of a single component of L, ⃗ such
2
as Lz , and also as an eigenstate of L . However, one may not find a set of basis states that
are simultaneously eigenstates of two different projections, e.g. Lx and Lz , because the three
projections do not commute amongst themselves.
If the potential V (⃗
r ) = V (r) is a function only of r, and is independent of both θ and ϕ, Lz
⃗ commute with the Hamiltonian and the eigenstates of H are also eigenstates of Lz and
and |L| 2
2
L . These eigenstates will be labeled by m and ℓ,

Ψ(⃗
r ) = ϕℓ (r)Yℓ,m (θ, ϕ). (4.27)

Being an eigenstate of L2 , the operator L2 in Eq. (4.26) can be replaced with its eigenvalue
ℏ2 ℓ(ℓ + 1). The spherical harmonic then factors out, and there is no mention of m in Schrödin-
ger’s equation, which justifies having ϕℓ (r) being independent of m. Schrödinger’s equation
for an energy E is then

ℏ2 ℏ2 ℓ(ℓ + 1)
 2
2 ∂

∂
Eϕℓ (r) = − + ϕ ℓ (r) + ϕℓ (r) + V (r)ϕℓ (r). (4.28)
2m ∂r 2 r ∂r 2mr 2
The radial function has a label ℓ because the radial wave function, Eq. (4.28), depends on ℓ, but
not on m.
Our immediate goal is to understand the angular functions Yℓ,m (θ, ϕ) which are eigenstates of
Lz and L2 and are referred to as spherical harmonics. In terms of bras and kets,

Yℓ,m (θ, ϕ) = ⟨n̂|ℓ, m⟩, (4.29)

which implies the normalization,

Z
dϕd cos θ |Yℓ,m (θ, ϕ)|2 = 1. (4.30)

The eigenstates of Lz must go as eimϕ . Combined with the requirement that the raising operator
in Eq. (4.25) working on Yℓ,ℓ gives zero, one can derive the expression for Yℓ,ℓ (θ, ϕ),

Yℓ,ℓ = fℓ (θ)eiℓϕ , (4.31)

 
∂ ∂
±i − cot θ fℓ (θ)eiℓϕ = 0,
∂θ ∂ϕ
(∂θ − ℓ cot θ) fℓ (θ) = 0,
fℓ (θ) = cℓ sinℓ (θ),
Yℓ,ℓ (θ) = cℓ eiℓϕ sinℓ (θ),

where the normalization is given by

s
ℓ
(−1) (2ℓ + 1)(2ℓ)!

cℓ = . (4.32)
2ℓ ℓ! 4π

The factor (−1)ℓ above is a matter of convention.

72
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

By operating on the known Yℓ,m s with L− one may generate Yℓ,m s for successively lower values
of m. Because the Yℓ,m s are eigenstates of Lz the m dependence is always trivial as it goes
proportional to eimϕ , but the θ dependence can be messy. Examples of a few spherical harmonics
are
1
Y0,0 = √ (4.33)
4π
r
3
Y1,0 = cos θ
4π
r
3
Y1,±1 = ∓ sin θe±iϕ
r 8π
5
Y2,0 = (3 cos2 θ − 1)
16π
r
15
Y2,±1 = ∓ sin θ cos θe±iϕ
r 8π
15
Y2,±2 = sin2 θe±2iϕ
32π
One might ask why half integer values of ℓ are never mentioned for spherical harmonics. The
problem is that the wave functions then become discontinuous as ϕ goes past 2π. Thus, half-
integral angular momenta can only be used for intrinsic spins and not as labels for a spatial wave
function because the spatial part of the Hamiltonian involves spatial gradients.
Finally, Legendre polynomials are defined as
s
4π
Pℓ (cos θ) ≡ Yℓ,m=0 (θ). (4.34)
2ℓ + 1

The Yℓ,m s with odd ℓ have odd parity, i.e. under the transformation (θ → π − θ, ϕ → ϕ + π)
the Yℓ,m s with odd ℓ switch sign. This will play an important role in determining that many
matrix elements will be zero. Note that the radial wave function, ϕℓ (r), is always even under
parity.

4.7 Wigner D Matrices

This is for the most point an exercise in notation, but it does emphasize that rotations mix states
within a given multiplet of angular momentum ℓ, where the degeneracy is 2ℓ + 1. Rather than
expressing rotations as a function of αx , αy and αz , one can express a rotation as a function of
the three Euler angles, which represent consecutive rotations about the z, y and the new z axes.
−iSz α −iSy β −iSz γ
     
ℓ
Dmm′ (α, β, γ) = exp exp exp . (4.35)
ℏ ℏ ℏ
The label ℓ refers to the dimension of the matrices, 2ℓ + 1, used to represent the rotations, e.g.
for two component matrices ℓ = 1/2, while m and m′ refer to the components of the matrices,
−ℓ ≤ m ≤ ℓ.

73
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

For instance, for ℓ = 1/2,

−iσz α −iσy β −iσz γ

     
1/2
D = exp exp exp (4.36)
2 2 2
= [cos(α/2) − iσz sin(α/2)][cos(β/2) − iσy sin(β/2)][cos(γ/2) − iσz sin(γ/2)]
 −i(α+γ)/2
cos(β/2) −e−i(α−γ)/2 sin(β/2)

e
= .
ei(α−γ)/2 sin(β/2) ei(α+γ)/2 cos(β/2)

Example 4.3: Rotations within a Spin−1/2 Multiplet Using Wigner D matrices

Using Wigner D matrices, rotate the state,

 
1
| ↑⟩ = ,
0

about the x axis by an angle ψ.

Solution:
For Euler angles, the first rotation is about the z axis, followed by a rotation about the y axis,
and the last rotation is about the new z axis. In order to rotate about the x axis, one must
first rotate the system with about the z axis by angle α = −π/2, then rotate about the y axis
by ψ. Finally, to restore the coordinate system, you need to rotate again about the z axis by
γ = π/2. Using Eq. (4.36), the rotation matrix is

−e−iπ/2 sin(ψ/2)
 
1/2 cos(ψ/2)
D (π/2, ψ, −π/2) = .
eiπ/2 sin(ψ/2) cos(ψ/2)
⃗
One can check that this is the rotation one would expect, e−i⃗σ·ψ/2 ,
⃗
e−i⃗σ·ψ/2 = cos ψI − iσx sin ψ
 
cos(ψ/2) −i sin(ψ/2)
= .
−i sin(ψ/2) cos(ψ/2)

Then, multiplying | ↑⟩,

 
1/2 cos(ψ/2)
D (π/2, ψ, 0)| ↑⟩ = .
−i sin(ψ/2)

By inspection, one can see that if one rotates by π the state will become | ↓⟩ as expected.
Further, if the rotation is by ψ = π/2 the resulting state will be an eigenstate of σy with
eigenvalue −1. Again, this would be expected for a rotation about the x axis.

The D matrices can also rotate operators, if the operators carry the labels of angular momentum,
e.g J and M . For example, if one considers the operators X, Y and Z, one can express three

74
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

similar operators

R0 = Z, (4.37)
√
R+ = (X + iY )/ 2,
√
R− = −(X − iY )/ 2.

By inspecting the form for spherical harmonics, one can see that these operators rotate just like
the spherical harmonics, Yℓ=1,m . Thus, even though they are operators, not states, their rotations
can be expressed with the same D ℓ=1 matrices used to rotate states with labels ℓ = 1 and m.
This may seem uneasy, as we are more accustomed to seeing operators, A, transforming as
U AU † . However, in this case the operators do not have two indices m, m′ , i.e. Amm′ , and
instead have a single index m. This realization plays a critical role in understanding the Wigner
Eckart theorem later in the course.

4.8 Separating Relative and Center-of-Mass Coordinates

Consider the Schrödinger equation for two particles interacting through a potential V (⃗
r1 − ⃗
r2 ).
The kinetic energy term,
ℏ2 ℏ2
HK = − ∇21 − ∇2 (4.38)
2m1 2m2 2
needs to be rewritten such that the derivates are with respect to center-of-mass and relative
coordinates,
m ⃗r + m2 ⃗ r2
⃗ ≡ 1 1
R r≡⃗
, ⃗ r1 − ⃗
r2 . (4.39)
m1 + m2
Using these definitions, one can show that that the kinetic energy becomes

ℏ2 ℏ2
HK = − ∇2r − ∇2r , (4.40)
2M 2µ

where M ≡ m1 + m2 and µ ≡ m1 m2 /(m1 + m2 ).

If the potential is only a function of r, with no mention of R, the total momentum operator,
−iℏ∇R , commutes with the Hamiltonian. The wave function may then be written as a product
of center-of-mass and relative coordinates,

Ψ(⃗ ⃗ = eiK·R⃗ ϕrel. (⃗

r , R) r ), (4.41)

with the center-of-mass dependence being characterized by a plane wave. The overall energy is
then a sum of the eigenenergy of the relative wave function plus ℏ2 K 2 /2M .
Note that if one of the masses is much larger than the other that the reduced mass µ approaches
the smaller of the two masses. If both masses are equal, the reduced mass is half the mass
of either of the two individual masses. For our purposes, we will solve problems such as the
hydrogen atom assuming the potential is fixed. For the real case, one need only replace the mass
with the reduced mass to include the effect that the source of the potential is itself mobile.

75
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

Separating the center-of-mass coordinates is convenient whenever one has a potential that is a
r1 − ⃗
function of ⃗ r2 only. Writing Schrödinger’s equation,
ℏ2 ℏ2 2
 
2 ⃗ ⃗ ⃗
− ∇R − ∇r eiK·R ϕ(⃗r ) + V (r)eiK·R ϕ(⃗
r ) = EeiK·R ϕ(⃗
r ), (4.42)
2M 2µ
⃗
One can factor out eiK·R after operating with ∇2R and get a simple equation for the relative wave
r ).
function ϕ(⃗
ℏ2
− ∇2r ϕ(⃗
r ) + V (r)ϕ(⃗
r ) = Erel ϕ(⃗
r) (4.43)
2µ
ℏ2 K 2
E= + Erel
2M
The factorization above works for any non-relativistic problem whenever the potential is a func-
tion of ⃗ ⃗
r and not R.
Example 4.4: Two Masses Coupled by a Spring
One especially easy example of factorization is the three-dimensional harmonic oscillator in
⃗ =
Cartesian coordinates. In this case the problem is even more factorizable. Again, using R
(⃗
r1 + ⃗
r2 )/2 and ⃗
r=⃗ r1 − ⃗r2 ,

ℏ2 ∇2R ℏ2 ∇2r
H =− +
+ V (⃗r ), (4.44)
2M 2µ
1 1 
r |2 = k x2 + y 2 + z 2 .

V (⃗
r ) = k|⃗
2 2
This potential is spherically symmetric, and is thus a candidate to solve using wave functions
that are products of radial wave functions and Yℓm s. However, because this potential can be
written as the sum of an x-dependent, a y-dependent and a z-dependent piece, it also works
to factorize the wave function as
⃗ ⃗ ⃗ ⃗
ϕ(R, r ) = ϕx (x)ϕy (y)ϕz (z)eiK·R , (4.45)

where x, y and z are the components of ⃗

r . One can separate the Schrödinger equation for the
three-dimensional relative wave function into three one-dimensional Schrödinger equations.

ℏ2 1
− ∂x2 ϕx + kx2 ϕx (x) = Ex ϕx (x) (4.46)
2µ 2
ℏ 2
1
− ∂y2 ϕy + ky 2 ϕy (y) = Ey ϕy (y)
2µ 2
ℏ2
1
− ∂z2 ϕz + kz 2 ϕz (z) = Ez ϕz (z)
2µ 2

Here, E = Ex + Ey + Ez + ℏ2 K 2 /2M . By multiplying the first equation by ϕy ϕz , the second

equation by ϕx ϕz and the third by ϕx ϕy , then adding the three equations, one finds that they
provide a solution to the three-dimensional equation.

76
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

Thus, one is able to take a six-dimensional equation, factor out the center-of-mass motion, and
recognize the factorizability, and reduce the problem to three trivial one-dimensional equa-
tions of motion. The energies are

Ex = (nx + 1/2)ℏω, Ey = (ny + 1/2)ℏω, Ez = (nz + 1/2)ℏω, (4.47)

p
where ω = k/µ. The total energy is then the sum

ℏ2 K 2
E= + (nx + ny + nz + 3/2)ℏω. (4.48)
2m
We refer to these solutions as the Cartesian-basis solutions to the harmonic oscillator. Because
the harmonic oscillator presented above has spherical symmetry, one can also write solutions
in spherical coordinates where each solution has labels ℓ and m and has a factor Yℓ,m (θ, ϕ).
The spherical basis will be discussed in Sec. (4.10).

4.9 Coupled Oscillators in Cartesian Coordinates

One convenient aspect of harmonic oscillator potentials is that they can be solved for the N -body
case even when N > 2. Here, we consider N particles of mass m interacting through mutual
harmonic oscillator potentials, with every mass being coupled to every other mass. Consider the
potential, !
1 X
V (⃗r1 , · · · ⃗
rN ) = k |⃗ rj |2 .
ri − ⃗ (4.49)
2 i<j

This potential is independent of the center-of-mass coordinates. That is, a translation of all co-
ordinates does not affect the potential. One can therefore write the solution as a product of a
center-of-mass wave function and a wave function that depends on N − 1 relative coordinates.
The above problem would be intractable if not for a trick which is unique to the harmonic oscil-
lator. One adds a fictitious potential that depends only on the center-of-mass coordinate.

1 1
Vf = r1 + ⃗
k|⃗ rN |2 =
r2 + · · · ⃗ N 2 k|Rcm |2 . (4.50)
2 2
With this choice, the sum of the real and fictitious potentials cancels all the cross terms and
becomes,
1
Vtot = Vf + V = N k r12 + r22 + · · · rN 2


. (4.51)
2
The total energy can be written as the sum of the center-of-mass energy plus the relative energy,
p
Etot = (nf + 3/2)ℏ N k/m + Erel , (4.52)

Erel is the energy of relative motion which is our ultimate goal. Here the term (nf +
wherep
3/2)ℏ N k/m replaces the usual kinetic energy of the center-of-mass. The single factor of

77
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

N inside the square root comes from the two factors of N in the expression for the fictitious
potential canceled by the one factor of N in the total mass,
s
N 2k
ωf = , M = N m. (4.53)
M
The total energies are those of N harmonic oscillators with mass m and spring constant N k,
and have eigenenergies
p
Etot = (n1 + n2 + · · · nN + 3N/2)ℏ N k/m. (4.54)

The ground statepenergy we are interested in is thus Etot (ni = 0) minus the energy of the center
of mass, (3/2)ℏ N k/m.
3N p 3 p
E0,rel = ℏ N k/m − ℏ N k/m (4.55)
2 2
3(N − 1) p
= N k/m.
2
For N = 2, the ground state energy is that of a single oscillator with reduced mass of µ = m/2.
The wave function of the ground state can be written as the product of all the ground state
wave functions, divided by the wave function of the center-of-mass. If one wanted to include
the center-of-mass motion, one could assign a momentum to the center of mass, P ⃗ , then add
⃗ ⃗
⃗ |2 /2m to the energies, and add a factor of eiP ·Rcm /ℏ to the wave function.
|P
Example 4.5: Three Coupled Harmonic Oscillators
rj |2 /2.
ri − ⃗
Consider three identical masses m, each coupled to the other by a potential k|⃗
1. What are the energies of the ground state and the first excited state?
2. What is the degeneracy of the first excited state?
Solution:
The ground state energy is found by applying Eq. (4.55) with N = 3.
3(3 − 1) p 3p
E0 = 3k/m = 6k/m.
2 2
The first excited state energy is found by taking the first excited state energy from the total
energy in Eq. (4.54), p
Etot = (3/2 + ntot )ℏ N k/m,
where ntot is some non-negative integer. Then, one must subtract the energy of the center of
mass, noting that the center-of-mass might also be excited, i.e.,
p
Ec.m. = (3/2 + nf )ℏ N k/m.

The difference gives the energy

p
Erel = (3/2 + ntot − nf )ℏ N k/m.

78
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

We are only looking for the excitation energy for given state of the center of mass, so we put
p nf = 0, and see that the first excited state of the relative
the center of mass in its ground stat,
energy is higher by an amount ℏ N k/m, with N = 3.
Finally, finding the degeneracy is a bit tricky. There exist 3N possibilities of n1 + n2 + n3 +
nN = 1. The factor of three accounts for the fact that these are three-dimensional harmonic
oscillators. However, this includes the possibility that the center of mass was excited. Thus,
the answer is 3(N − 1). So, for three coupled oscillators, the degeneracy of the first excited
state is 6.

4.10 Solving the Radial Wave Equation for Spherically Symmetric Cases
Similar factorization ideas are applied any time one deals with a spherically symmetric potential.
In that case one can write the wave function as a product of ϕℓ (r) and Yℓ,m (θ, ϕ) and reduce the
problem to a one-dimensional problem of the radial coordinate. The one-dimensional Schrödin-
ger equation for ϕℓ (r) then becomes

ℏ2 ℏ ℓ(ℓ + 1)
 2
2
  
2
− ∂r + ∂r ϕℓ (r) + + V (r) ϕℓ (r) = Eϕℓ (r). (4.56)
2µ r 2µr 2

Although this is a one-dimensional differential equation, it is not a one-dimensional Schrödinger

equation due to the extra derivative term (2/r)∂r . The one-dimensional Schrödinger form can
be regained by defining
uℓ (r) ≡ rϕℓ (r). (4.57)
The wave equation for u looks like a 1-d Schrödinger equation with a centrifugal potential,

ℏ2 ∂ 2 ℏ ℓ(ℓ + 1)
 2 
− u(r) + + V (r) u(r) = Eu(r). (4.58)
2µ ∂r 2 2µr 2

The boundary condition for u is that it must go to zero at the origin so that ϕ is finite, and
analytic, at the origin. The basic strategy is to solve the equation for u(r), then divide by r to
get ϕℓ (r).
The wave equation is particularly simple for s waves (ℓ = 0) because such solutions reduce to
simple one-dimensional problems with an infinite potential for r < 0. Considering non-zero
ℓ introduces a divergent centrifugal potential at the origin, proportional to ℓ(ℓ + 1)/r 2 . In the
neighborhood of r = 0 the solution looks like either

uℓ (r) ∼ r ℓ+1 1 + Or + Or 2 · · · , the regular solution,

(4.59)

or
uℓ (r) ∼ r −ℓ 1 + Or + Or 2 · · · , the irregular solution.


(4.60)
Equivalently, one could state that the solutions for ϕℓ = uℓ /r behave as r ℓ and r −ℓ−1 for the reg-
ular and irregular solutions respectively. Clearly, only the regular solution satisfies the boundary
conditions. Even when a Coulomb potential is added, which is also divergent at the origin only
less so, the behavior at the origin can be expanded as shown above. If the divergence of the

79
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

potential at the origin is as strong or stronger than 1/r 2 , one needs to rethink the expansion
above.
When the potential V (r) is zero, i.e. free particles, the regular solutions are known as spherical
bessel functions, jℓ (kr), while the irregular solution are referred to as spherical Neumann func-
tions, nℓ (kr). These correspond to solutions to the wave equations for ϕℓ above, rather than
for uℓ . Before writing
p these down, we rewrite Schrödinger’s equation with V = 0 in terms of
2 2 2
x = kr with k = 2mE/ℏ . Multiplying Schrödinger’s equation by 2µ/(ℏ k ),

2 ℓ(ℓ + 1)
 
2
− ∂x + ∂x ϕℓ (x) + ϕℓ (x) = ϕℓ (x). (4.61)
x x2

For the first few ℓs, the solutions are

sin x cos(x)
j0 (x) = , n0 (x) = − , (4.62)
x x
sin x cos x cos x sin x
j1 (x) = 2
− , n1 (x) = − 2
− ,
x x x x
3 1 3 3 1 3
j2 (x) = 3
− sin x − 2 cos x, n2 (x) = − 3
− cos x − 2 sin x,
x x x x x x
.. ..
. .

By taking a linear combination of jℓ and nℓ , one can find a solution which behaves like an out-
going wave at large r,

hℓ (kr) = jℓ (kr) + inℓ (kr) (4.63)

eikr
≈ (−i)ℓ+1 , as r → ∞.
kr
These are known as Hankel functions. As an example, for ℓ = 1,

eix
 
i
h1 (x) = −1 − . (4.64)
x x

Example 4.6: Spherical Wells

1. Solve for the lowest energies of the ℓ = 0 and ℓ = 1 states of an infinite spherical well
of radius R. Note that for the ℓ = 1 solutions, a transcendental expression will remain.
Solution: The BC and solutions for ℓ = 0 are

j0 (kR) = 0,
sin(kR) = 0,
π
k= ,
R
ℏ2 π 2
E0 = .
2mR2

80
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

For ℓ = 1,

j1 (kR) = 0,
sin(kR)
− cos(kR) = 0,
kR
tan(kR) = kR, (transcendental eq.for k)
ℏ2 k2
E1 = .
2m
The values of kR for which jℓ (kR) = 0 are known as the zeros of the spherical Bessel
function. The values are:

ℓ n=1 n=2 n = 3 ···

0 π 2π 3π ···
1 7.72525 10.9041 · · ·
4.4934
2 9.0950 12.3299 · · ·
5.7635
3 10.4171 13.6980 · · ·
6.9879
.. .. .. .. ..
. . . . .
th
Values of the n zeros of spherical Bessel function of order ℓ

From the table, one can see that the first excited state has ℓ = 1, and that the first ℓ = 2
state is also of lower energy than the second ℓ = 0 state.
2. Outline how one would solve for the bound state of a well of finite depth −V0 and width
R for ℓ = 1.
Solution: One would choose j1 (kr) for the interior solution (region I), with
p
k = 2m(V0 − B)/ℏ2 ,

then match to a solution for negative energy (imaginary k) outside. The outer solution
would be chosen as
e−qr 1
 
Ah1 (iqr) = −A 1+ ,
iqr qr
p
where q = 2mB/ℏ2 and A is an unknown constant. One then needs to satisfy the two BC
(continuous value and slope) at r = R by varying the binding energy B and the constant A.
This would lead to a transcendental equation for B.

4.11 Spherical Harmonic Oscillator – Spherical Basis

The spherically symmetric harmonic oscillator can be approached either through Cartesian co-
ordinates as performed earlier or in a spherical basis. First, we review the Cartesian solutions.

81
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

Because the potential can be written as

1 1
kr 2 = k x2 + y 2 + z 2 ,


V (⃗
r) = (4.65)
2 2
the solutions factorize into x, y and z-dependent functions,

Ψ(⃗
r ) = ψnx (x)ψny (y)ψnz (z). (4.66)

Each piece is a solution of the 1-dimensional Schrödinger’s equation and the total energy is
p
E = (N + 3/2)ℏω , N ≡ nx + ny + nz , ω k/µ. (4.67)

The N = 0 and N = 1 eigenstates have the form

2 /(2a2 )
r ) ∼ e−r
ϕnx =ny =nz =0 (⃗ (4.68)
−r 2 /(2a2 )
ϕnx =1,ny =nz =0 (⃗
r ) ∼ xe
2 /(2a2 )
r ) ∼ ye−r
ϕnx =0,ny =1,nz =0 (⃗
2 /(2a2 )
r ) ∼ ze−r
ϕnx =ny =0,nz =1 (⃗

By looking at the form of the Yℓ,m s, one can see that the N = 0 state has a ℓ = 0, m = 0 an-
2 2
gular dependence, and can therefore be written as a spherically symmetric function, e−r /(2a0 ) ,
multiplied by Y0,0 . The solution for the nx = ny = 0, nz = 1 state can be written as a prod-
2 2
uct of Y1,0 and the radial function re−r /(2a0 ) . By taking linear combinations, ϕnx =1,ny =nz =0 ±
ϕnx =0,ny 1,nz =0 , one finds solutions which can be written as the same radial wave function mul-
tiplied by Y1,±1 .
In spherical coordinates the labels nx , ny and nz are replaced by N , ℓ and m. Mapping the
solutions for higher N is a bit tricky. For N ≥ 2, one can count the states from a Cartesian
perspective. One needs to know the number of ways to get three integers to add to N . First, the
number of ways, d⊥ , to get two positive integers to add to N⊥ is
X
d⊥ (N⊥ ) = = N⊥ + 1. (4.69)
nx =0,N⊥

Requiring that a third integer adds to N gives the total degeneracy d(N ),
X (N + 1)(N + 2)
d(N ) = d⊥ (N⊥ ) = . (4.70)
N⊥ =0,N
2

Thus, there is one way to get N = 0, 3 ways to get N = 1, six ways to get N = 2, etc.
To determine which ℓ multiplets combine to create the dN Cartesian solutions with a given en-
ergy, (N + 3/2)ℏω, consider two pieces of evidence: First, creating the Cartesian state with
nz = 0, nx = ny = 0 is represented by the spherical state Yℓ=N,m=N , therefore there must

82
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

be at least one ℓ = N multiplet. Secondly, all the multiplets for N − 2 must have correspond-
ing multiplets which are generated by operating on those states with the spherically symmetric
operator a2x + a2y + a2z . We can now determine the multiplets by:

(N + 1)(N + 2)
d(N ) = = (2N + 1) + d(N − 2) + any others. (4.71)
2
Solving for the number of others, one finds there are no others. Hence the excitation N + 2
states have the same states, but with one more multiplet of ℓ = N + 2. Because one knows that
for N = 0 there is one ℓ = 0 state and for N = 1 there is one ℓ = 1 multiplet, one can quickly
find all the multiplets for any N .
As an example the N = 5 states are covered by one ℓ = 5 multiplet, one ℓ = 3 multiplet and
one ℓ = 1 multiplet. Note that all states with even N have even parity and all states with odd
N have odd parity.

4.12 The Hydrogen Atom

There are three standard problems of spherically symmetric potentials where the solutions are
analytic, the inifinite well, the harmonic oscillator and the Coulomb potential. Here, we consider
the case where the potential is attractive,

e2
V (r) = − . (4.72)
r
One may rewrite the Schrödinger equation,

∂2 ℓ(ℓ + 1)
 
Z1 Z2
− 2+ 2
− uℓ (r) = −k2 uℓ (r), (4.73)
∂r r a0 r

where the Bohr radius is defined a0 ≡ ℏ2 /(µe2 ) and k2 = −2µE/ℏ2 . For large r the potential
and centrifugal terms are negligible and the wave function must behave as e±ikr multiplied by
terms that vary more slowly in r.
The solutions to the Schrödinger equation can be written in terms of associated Laguerre poly-
nomials.
( )1/2
2 3 (n − ℓ − 1)! 2r ℓ 2ℓ+1 2r
    
un,ℓ −r/(na0 )
Rn,ℓ (r) = = e Ln+ℓ . (4.74)
r na0 2n[(n + ℓ)!]3 na0 na0

For a given ℓ there are many solutions labeled by the integer n > ℓ. The eigenenergies can be
written simply as
e2 1
En = − . (4.75)
2a0 n2
If larger charges are used, the above expressions are modified by scaling a0 by 1/(Z1 Z2 ).

83
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

Writing a few solutions,

2
R1,0 = 3/2
e−r/a0 (4.76)
a0
1
 
r
R2,0 = 2− e−r/(2a0 )
(2a0 )3/2 a0
1 r
R2,1 = √ e−r/(2a0 )
(2a0 )3/2 a0 3

The degeneracy, where energies with different ℓ have the same energies seems accidental, but is
related to the similar degeneracy in the harmonic oscillator. Both degeneracies can be explained
by considering the Lenz vector, which commutes with the Hamiltonian for the special case of
the Coulomb potential.
1 e2
Q⃗ ≡ (⃗
p×L−L×p ⃗) − ⃗ r. (4.77)
2µ r
This operator is Hermitian and, if the commutation between p ⃗ is ignored, is identical to
⃗ and L
the classical expression for the Lenz vector.
One can also define a scaled operator,
r
−m
⃗ ≡
K ⃗
Q, (4.78)
2H
which is a little odd because one is defining the square root of an operator. However, because we
are considering only eigenstates of the Hamiltonian with negative energies, this is not too sick.
One can show that the components of K ⃗ and the components of the angular momentum L ⃗ obey
simple commutation relations,
[Ki , Kj ] = iℏϵijk Lk , [Ki , Lj ] = iℏϵijk Kk (4.79)
These commutation relations are reminiscent of angular momentum commutation relations, and
in fact, if one defines two new operators,
⃗ +K
L ⃗
⃗ ≡
M (4.80)
2
⃗ −K
L ⃗
⃗ ≡
N ,
2
one can see that M and N obey the same commutation relations as L,

[Mi , Mj ] = iℏϵijk Mk , [Ni , Nj ] = iℏϵijk Nk , [Mi , Nj ] = 0 (4.81)

⃗ and N
Because M ⃗ are linear combination of the L
⃗ and K,
⃗ they commute with the Hamiltonian,
and because they commute with one another they may simultaneously be defined.
Furthermore, some algebra reveals that the Hamiltonian may be written as
me4 me4
H =− = (4.82)
⃗2+L
2(K ⃗ 2 + ℏ2 ) ⃗ 2 + 2N
2(2M ⃗ 2 + ℏ2 )

84
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

Finally, note that the Lenz vector is always perpendicular to the angular momentum, which
means that ⃗r·L ⃗ =K ⃗ ·L⃗ = 0. From the definitions of M
⃗ and N⃗ , one then sees the constraint
that
M =N (4.83)
⃗ and N
where M and N are the quantum numbers denoting the magnitude of the vectors M ⃗ in
⃗ The eigenenergies are thus,
the same way that ℓ denotes the magnitude of L.

me4 me4
EM = − =− (4.84)
2ℏ2 (4M (M + 1) + 1) 2ℏ2 (2M + 1)2

Because M could be either integral over half integral, the numbers

n ≡ 2M + 1 (4.85)

are integral values.

These states are not eigenstates of L, but given the fact that there must exist eigenstates of L, one
can conclude that several states of the same ℓ might be degenerate.

4.13 Adding Angular Momentum

In a spherically symmetric potential, the orbital angular momentum, L, commutes with the
Hamiltonian. Thus, one can choose eigenstates of Lz and |L| ⃗ 2 as a subset of states, knowing
that such states can also be eigenstates of the Hamiltonian. The eigenvalues of are ℏm and
ℏ2 ℓ(ℓ + 1), the labels ℓ and m are often called “good” quantum numbers, because they form
a basis that is convenient for diagonalizing the Hamiltonian. Many particles also have intrinsic
spin, even those particles which are currently considered as point particles such as electrons and
photons. Thus, in addition to the orbital quantum numbers two more quantum numbers may
be used to describe the eigenstates of a single particle in the potential, s and ms , which describe
the magnitude and projection of the spin angular momentum.
Often a term exists in the Hamiltonian which couples the two types of spin. One example is the
spin-orbit interaction,
⃗ · S.
Hs.o. = αL ⃗ (4.86)
This term originates from relativistic considerations which we will see later in the course. Be-
cause the term is written as a rotational scalar and does not involve an external field, which
would explicitly break the rotational symmetry, we expect that the overall angular momentum
remains conserved. Indeed, one can see that each component of the total angular momentum

J⃗ ≡ L
⃗ +S
⃗ (4.87)

commutes with the spin-orbit term, even though none of the components of L ⃗ or S
⃗ commute
with Hs.o. individually. Furthermore, the total squared orbital and total spin angular momen-
tum, L2 and S 2 , also commute with Hs.o. . Thus, there are two new quantum numbers j and
mj which replace mℓ and ms as good quantum numbers, while mℓ and ms are no longer good
quantum numbers.

85
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

A clearer insight into the spin-orbit term can be attained by rewriting it explicitly in terms of j,
ℓ and s.
 2
⃗ +S
L ⃗ =L ⃗2 + S⃗ 2 + 2L
⃗ ·S⃗ (4.88)
1 2 
L⃗ ·S⃗= J⃗ − L ⃗2 − S
⃗2 ,
2
which means that the spin-orbit term for a state in the basis labeled by j, mj , ℓ and s may be
expressed as,
αℏ2
Hs.o. = [j(j + 1) − ℓ(ℓ + 1) − s(s + 1)] . (4.89)
2
The coupling of spins is a common occurrence in all branches of physics. In nuclear physics,
the spin-orbit term is surprisingly large, and is responsible for the basic scheme for nuclear shell
structure. In describing hadron spectroscopy, a spin-spin interaction is largely responsible for
the difference of the spin 3/2 delta baryon and the spin 1/2 proton which are comprised of
quarks of the same flavor. The coupling of angular momentum in physics thus often involves
changing from the ℓ, s, mℓ , ms basis to the ℓ, s, j, mj basis.
When undergoing a change of basis, the number of states involved is (2s + 1)(2ℓ + 1) as can be
determined by considering the number of combinations of mℓ and ms . However, changing to
the j, mj basis only mixes states with identical mj = mℓ + ms . Because the states in a multiplet
described by j must be complete, running from −j to j, we see that the maximum value of j is
j ≤ ℓ + s. (4.90)
Because there is only one state with mℓ = ℓ and ms = s, there is only one state with mj = ℓ + s
and thus only one j multiplet with j = ℓ + s. Counting the number of pairs of mℓ and ms that
add up to a specific value of mj , and realizing that every j multiplet must be complete lets one
see that the values of j involved are
jmax = ℓ + s (4.91)
jmin = |ℓ − s|.
These are known as the triangle relations, as they can be considered as constraints involved in
adding vectors. One cannot add two vectors of lengths ℓ and s and obtain a vector of length j
outside this range. A common error made by students is to only consider the j = ℓ + s and
j = ℓ − s multiplets, and forget the multiplets with intermediate values of j.
One can check that the net number of states is unchanged. Each j multiplet has 2j + 1 states.
Assuming s ≤ ℓ the number of j multiplets is (2s + 1) while the average j of the multiplets
is ℓ, which means that the number of states is ⟨2j + 1⟩(2s + 1) = (2ℓ + 1)(2s + 1). This is
consistent with the total number of states calculated in the mℓ , ms basis.

4.14 Clebsch-Gordan Coefficients

Changing from the mℓ , ms basis to the j, mj basis is described by the overlap of matrix ele-
ments,
⟨ℓ, s, j, mj |ℓ, s, mℓ , ms ⟩. (4.92)

86
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

Such matrix elements are known as Clebsch-Gordan coefficients and are referred to through
a variety of confusing notations, such as C(ℓ, s, mℓ , ms ; j, m) and nearly every other possi-
ble permutation of the arguments. Sometimes the coefficients are labeled by superscripts and
subscripts and sometimes they look like matrix elements ⟨j, mj |ℓ, s, mℓ , ms ⟩, where ℓ and s
are implied in the bra. The notations are remarkably confusing given that the only purpose of
not writing down eight labels, with two repeated as in Eq. (4.92), is that the labels ℓ and s,
which are always the same in the bra and ket, are not written down twice. To emphasize that
the coefficients behave as matrix elements we will refer to the coefficients either in the form
⟨j, mj |ℓ, s, mℓ , ms ⟩ or with ℓ and s repeated in the bra as in Eq. (4.92).
The matrix elements are usually used as part of a basis transformation, e.g. expressing states
labeled by |ℓ, s, j, mj ⟩ as a linear combination of |ℓ, s, mℓ , ms ⟩. Using completeness,
X
|ℓ, s, mℓ , ms ⟩⟨ℓ, s, mℓ , ms | = I, (4.93)
mℓ ,ms
X
|ℓ, s, j, mj ⟩ = |ℓ, s, mℓ , ms ⟩⟨ℓ, s, mℓ , ms |ℓ, s, j, mj ⟩
mℓ ,ms
X
= ⟩⟨ℓ, s, mℓ , ms |ℓ, s, j, mj ⟩|ℓ, s, mℓ , ms ⟩
mℓ ,ms
X
= ⟨j, mj |ℓ, s, mℓ , ms ⟩|ℓ, s, mℓ , ms ⟩.
mℓ ,ms

The last line was justified by the fact that the Clebsch-Gordan coefficients are real (because the
raising and lowering operators don’t introduce any complex phases).
Finding the matrix elements is straight-forward, though tedious. They can be found using the
algebra for raising and lowering angular momentum. First, remember that the matrix elements
are all proportional to δmj ,mℓ +ms . Because there is only one multiplet with mj = ℓ + s that
matrix element is simple to express.

⟨ℓ, s, j = ℓ + s, mj = ℓ + s|ℓ, s, mℓ = ℓ, ms = s⟩ = 1. (4.94)

To generate the coefficients involving the same j = ℓ + s but reduced mj , one can use the
lowering operators,

1
|ℓ, s, j, mj − 1⟩ = p J− |ℓ, s, j, mj ⟩ (4.95)
j(j + 1) − mj (mj − 1)
1
= p (L− + S− ) |ℓ, s, j, mj ⟩.
j(j + 1) − mj (mj − 1)

Applying this to the case where mj = j = ℓ + s one generates an expression for the matrix

87
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

elements with mj = ℓ + s − 1,

1
|ℓ, s, j = ℓ + s, mj = ℓ + s − 1⟩ = p (4.96)
j(j + 1) − mj (mj − 1)
· (L− + S− ) |ℓ, s, mℓ = ℓ, ms = s⟩
1
= p
j(j + 1) − mj (mj − 1)
p
· ℓ(ℓ + 1) − mℓ (mℓ − 1)|ℓ, s, mℓ = ℓ − 1, ms = s⟩
p 
+ s(s + 1) − ms (ms − 1)|ℓ, s, mℓ = ℓ, ms = s − 1⟩ .

One can now read off the Clebsch-Gordan coefficients. For instance,
p
ℓ(ℓ + 1) − mℓ (mℓ − 1)
⟨ℓ, s, j = ℓ + s, mj = ℓ + s − 1|ℓ, s, mℓ = ℓ − 1, ms = s⟩ = p .
j(j + 1) − mj (mj − 1)
(4.97)
Finding the Clebsch-Gordan coefficient for j ̸= ℓ + s is a bit trickier. By knowing that the
j = ℓ+s−1 states are orthogonal to the j = ℓ+s states, allows one to express the j = ℓ+s−1
states down by inspection. For instance,

|ℓ, s, j = ℓ + s − 1, mj = ℓ + s − 1⟩ = (4.98)
1  p
p · ℓ(ℓ + 1) − mℓ (mℓ − 1)|ℓ, s, mℓ = ℓ, ms = s − 1⟩
j(j + 1) − mj (mj − 1)
p 
− s(s + 1) − ms (ms − 1)|ℓ, s, mℓ = ℓ − 1, ms = s⟩

Of course, one could multiply the states by any arbitrary phase and the coefficients would work
as well. The convention is that the coefficient ⟨j1 , j2 , j, j|j1 , j2 , m1 = j1 , m2 = j − j1 ⟩ is real
and positive.
Example 4.7: Calculating Clebsch-Gordan Coefficients
Find ⟨j = 3/2, m = 3/2|j1 = 3/2, j2 = 1, m1 = 3/2, m2 = 0⟩.
First, using J− = J1,− + J2,−

|j = 5/2, m = 5/2⟩ = |m1 = 3/2, m2 = 1⟩

p
(5/2)(7/2) − (5/2)(3/2)|j = 5/2, m = 3/2⟩ =
p
(3/2)(5/2) − (3/2)(1/2)|m1 = 1/2, m2 = 1⟩
p
+ (1)(2) − (1)(0)|m1 = 3/2, m2 = 0⟩,
|j = 5/2, m = 3/2⟩ =
r r
3 2
|m1 = 1/2, m2 = 1⟩ + |m1 = 3/2, m2 = 0⟩.
5 5

88
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

Now, because the state |j = 3/2, m = 3/2⟩ is orthogonal to the state above,
r r
3 2
|j = 3/2, m = 3/2⟩ = |m1 = 3/2, m2 = 0⟩ − |m1 = 1/2, m2 = 1⟩. (4.99)
5 5
The Clebsch-Gordan coefficient is then
r
3
⟨j = 3/2, m = 3/2|j1 = 3/2, j2 = 1, m1 = 3/2, m2 = 0⟩ = . (4.100)
5

Example 4.8: Neutron and Proton with Spin-Spin and Magnetic Interaction
A neutron and proton are each in an s wave of a nuclear potential. The two particles feel a
spin-spin interaction,
Vss = −αS ⃗p · S
⃗n .
The nucleons also are in region with external magnetic field of strength B,

⃗ ·S
Vb = −µp B ⃗p − µn B
⃗ ·S
⃗n .

In terms of α, µp , µn and B, find the four energy eigenvalues.

Solution:
This problem is made difficult by the fact that Vb is diagonal in the mp , mn basis, while Vss is
diagonal in the j, m basis. (We will omit the sp = 1/2, sn = 1/2 labels in the bras and kets
to save space.)

′ ′
αℏ2
⟨j, m|Vss |j , m ⟩ = δ j,j ′ δ
m,m′ (j(j + 1) − sp (sp + 1) − sn (sn + 1))
2
⟨mp , mn |Vb |m′p , m′n ⟩ = −δmp ,m′p δmn ,m′n Bℏ (µp mp + µn mn )

To proceed further, one must choose a basis. We choose the j, m basis with the following
eigenvalues,
   
1 0
 0   1 
|j = 1, m = 1⟩ = 
 0  , |j = 1, m = −1⟩ =  0  ,
  

0 0
   
0 0
 0   0 
|j = 1, m = 0⟩ = 
 1  , |j = 0, m = 0⟩ =  0  .
  

0 1

In this basis Vss is diagonal,

 
−αℏ2 /4 0 0 0
 0 −αℏ2 /4 0 0 
Vss = 2

 0 0 −αℏ /4 0 
2
0 0 0 3αℏ /4

89
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

while writing Vb requires first rewriting each of the states in the mn , mp basis.
|j = 1, m = 1⟩ = |mp = 1/2, mn = 1/2⟩,
|j = 1, m = −1⟩ = |mp = −1/2, mn = −1/2⟩,
1
|j = 1, m = 0⟩ = √ (|mp = 1/2, mn = −1/2 + |mp = −1/2, mn = 1/2⟩) ,
2
1
|j = 0, m = 0⟩ = √ (|mp = 1/2, mn = −1/2⟩ − |mp = −1/2, mn = 1/2⟩) .
2
If the problem involved higher angular momentum, one would have to go through the proce-
dure of the previous exercise, utilizing raising and lowering operators, to write the j, m states
in terms of the m1 , m2 basis.
From the above form, we can see that Vb will mix the two states with with m = 0 as they are
not eigenstates in the mp , mn basis.
 
(µp + µn ) 0 0 0
ℏB  0 −(µp + µn ) 0 0 
Vb = −  .
2  0 0 0 (µp − µn ) 
0 0 (µp − µn ) 0
Thus, the first two eigenvalues are simple to find
ϵ1 = −αℏ2 /4 + (µp + µn )ℏB/2
ϵ2 = −αℏ2 /4 − (µp + µn )ℏB/2,
while the last two eigenvalues are most easily found by writing the lower-right 2 × 2 sub-
matrix in terms of σ matrices,
V2×2 = αℏ2 /4 − (αℏ2 /2)σz + [(µp − µn )ℏB/2] σx
The eigenvalues of the submatrix are
q
ϵ± = αℏ /4 ± α2 ℏ4 /4 + (µp − µn )2 ℏ2 B 2 /4
2

4
Energy levels of a proton and neutron ex-
2 periencing a spin-spin interaction while in
2/4)

a magnetic field. The lower levels cor-

0 respond to those of the deuteron, which
E/(

has S = 1, whereas the upper line corre-

2 sponds to an excited (unbound resonance)
state.
4
0.0 0.2 0.4 0.6 0.8 1.0
NB/( 2)

90
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

4.15 Exercises
r 2 = x2 + y 2 + z 2 commutes with Lz .
1. (a) Show that ⃗

r·p
(b) Show that ⃗ ⃗ commutes with Lz .
2. Any two rotations, α ⃗ can be written as a single rotation by ⃗
⃗ and β, γ , which in the spin 1/2
basis means
⃗
eiβ·⃗σ/2 ei⃗α·⃗σ/2 = ei⃗γ ·⃗σ/2 .
Show that the equivalent angle ⃗
γ may be written in terms of α ⃗ as
⃗ and β

cos(γ/2) = cos(β/2) cos(α/2) − β̂ · α̂ sin(α/2) sin(β/2)

γ̂ sin(γ/2) = cos(β/2) sin(α/2)α̂ + cos(α/2) sin(β/2)β̂ + sin(β/2) sin(α/2)α̂ × β̂,

where α̂, β̂ and γ̂ are the corresponding unit vectors. Note that these relations would hold
for any rotation, not just the spin 1/2 system. Thus, they describe the rotation group.
Hints: Use the fact that ei⃗a·σ = cos(a) + i ⃗σ|⃗a·⃗a| sin(a). Also use the identity σi σj = δij +
iϵijk σk .
3. Consider the matrices,
     
ℏ 0 1 0 ℏ 0 −i 0 1 0 0
Sx = √  1 0 1  , Sy = √  i 0 −i  , Sz = ℏ  0 0 0  .
2 0 1 0 2 0 i 0 0 0 −1
These represent the rotation matrices for angular momentum S = 1, S(S + 1) = 2. Note
that the eigenvalues of Sz are -1,0,1 as expected.
(a) Explicitly multiply the matrices to show that
[Si , Sj ] = iℏϵijk Sk .
For efficiency, just pick one of the three combinations to check.
(b) Explicitly multiply the matrices to show that
X
Si2 = 2ℏ2 I = ℏ2 S(S + 1)I.
i

4. Using the definition, Lz = −iℏ(x∂y − y∂x ) = −iℏ∂ϕ , express X(α) = eiLz α/ℏ xe−iLz α/ℏ
in terms of x, y and z. Hint: Note that
eα∂ϕ f (ϕ)e−α∂ϕ = f (ϕ + α).
because eα∂ϕ generates a Taylor expansion.
5. Consider the six group elements for the symmetry of the equilateral triangle listed in Sec.
4.3. As a six-by-six matrix, find the coefficients aij .
6. Using the commutation relations for angular momentum, [Lx , Ly ] = iℏLz , and the defi-
nition L± = Lx ± iLy , show that
⃗ 2 = L2 + L+ L− − ℏLz .
|L| z

91
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

7. In terms of ℓ, m1 and m2 find expressions for:

(a) ⟨ℓm1 |L2x |ℓm2 ⟩ (Warning: this is messy)

(b) ⟨ℓm1 |L2x + L2y |ℓm2 ⟩

8. Consider a particle of mass m in a spherical well of radius R, where the potential is +∞

for r > R and zero for r < R.

(a) Find the ground state energy.

(b) Describe how one would find the energy of the first excited state of the same well.
(c) If the particle is an electron and the radius of the well is 0.15 nm, give a numerical
value for the energy of the ground state in eV.

9. (a) Estimate the ground state binding energies of the following atoms. You can use the
fact that the binding energy for hydrogen is 13.6 eV, the mass of an electron is 0.511
MeV the mass of a muon is 105.7 MeV, the mass of a proton is 938.3 MeV and the
charge of Pb is 82. Then scale the hydrogen values to get the desired results. The Bohr
radius of H is 0.053 nm.
i. e, P b
ii. µ− , p
iii. e+ e−
iv. p̄, P b
The mass of a muon is 205 times larger than that of an electron.
(b) For the same cases above, find the associated Bohr radii. (Treat the Pb nucleus as a
point particle)
10. For the Hydrogen atom, calculate the expectation of the operator X between the ground
state and each of the four n = 2 states. You can express answer in terms of a0 .
11. Prove the following recurrence relation for spherical Bessel functions:

ℓ
jℓ+1 (z) = −jℓ′ (z) + jℓ (z).
z
To accomplish this, assume the equation is true and that jℓ (z) is a solution to:

2 ℓ(ℓ + 1)
−jℓ′′ (z) − jℓ′ (z) + jℓ (z) = jℓ (z).
z z2
Then show that using the assumed expression for jℓ+1 (z) will be a solution to:

′′
2 ′
(ℓ + 1)(ℓ + 2)
−jℓ+1 (z) − jℓ+1 (z) + jℓ+1 (z) = jℓ+1 (z).
z z2
This last expression is the same differential equation as the one just above, but with ℓ →
ℓ + 1.
12. Find the Clebsch-Gordan coefficient

⟨ℓ = 1, s = 1, j = 0, m = 0|ℓ = 1, s = 1, mℓ = 1, ms = −1⟩

92
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

13. Calculate the Clebsch-Gordan Coefficients ⟨ℓ = 12, s = 1, j = 12, mj = 12|ℓ = 12, s =

1, mℓ , ms ⟩ for all mℓ and ms .
14. An electron is in an ℓ = 1 state of a hydrogen atom. It experiences a spin orbit interaction,
⃗ ·S
Vs.o. = αL ⃗

and also feels an external magnetic field

 
⃗ ⃗ ⃗
VB = −µB · L + 2S .

a) Using the basis

  
1 0
 0   1 
   
 0 
 , |J = 3/2, M = −3/2⟩ =  0 
 
|J = 3/2, M = 3/2⟩ = 
 0   0 
   
 0   0 
0 0
   
0 0
 0   0 
   
 1 
 , |J = 1/2, M = 1/2⟩ =  0 
 
|J = 3/2, M = 1/2⟩ = 
 0   1 
   
 0   0 
0 0
   
0 0
 0   0 
   
 0   0 
|J = 3/2, M = −1/2⟩ =   , |J = 1/2, M = −1/2⟩ = 
  
0
 
 0 
 
 1   0 
0 1
write the Hamiltonian components Vs.o. and VB as 6×6 matrices. To assist you, the |J, M ⟩
states can be written in the |mℓ , ms ⟩ basis as
|J = 3/2, mj = 3/2⟩ = |mℓ = 1, ms = 1/2⟩, (4.101)
|J = 3/2, mj = −3/2⟩ = |mℓ = −1, ms = −1/2⟩,
p √ p
15/4 − 9/4 + 3/2|J = 3/2, mj = 1/2⟩ = 2|0, 1/2⟩ + 3/4 − 1/4 + 1/2| − 1, −1/2⟩,
r
2 1
|J = 3/2, mj = 1/2⟩ = |0, 1/2⟩ + √ |1, −1/2⟩,
3 3
r
1 2
|J = 1/2, mj = 1/2⟩ = √ |0, 1/2⟩ − |1, −1/2⟩,
3 3
r
2 1
|J = 3/2, mj = −1/2⟩ = |0, −1/2⟩ + √ | − 1, 1/2⟩,
3 3
r
1 2
|J = 1/2, mj = −1/2⟩ = √ |0, −1/2⟩ − | − 1, 1/2⟩.
3 3

93
PHY 851/852 4 ANGULAR MOMENTUM AND CENTRAL POTENTIALS

b) What are the six eigenvalues of H?

15. A spin 1/2 particle is bound to a fixed center by a spherically symmetric potential. The
particle is in an ℓ = 0 state with spin-up, i.e.
 
1
Ψ(⃗r , m) = ψ(r) .
0

In terms of ψ(r) and ⃗

r , write the matrix element for

⟨⃗
r , ms |⃗
σ·⃗
r |Ψ⟩

(a) for ms = 1/2

(b) for ms = −1/2

94
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

5 Symmetries and Conservation Laws

5.1 Continuous and Discrete Symmetries

Symmetries can be classified as either continuous or as discrete. Examples of continuous sym-
metries are rotational and translation symmetries while parity and time reversal are examples of
discrete symmetries. Rotational and translational symmetries can be continuous because of the
continuum of rotations under which the system can be rotated. However, sometimes symme-
tries involve rotations or translations by a finite angle. For instance, a lattice is invariant under
translations of size na where n is an integer and a is the lattice spacing. A circular chain of n
identically spaced identical objects is unchanged after rotation by an angle 2π/n.
In classical mechanics and in field theory, one studies Noether’s theorem which associates a
conserved charge with any continuous symmetry. There exists an analogous relation in quantum
mechanics which is perhaps easier to see. A continuous symmetry can be expressed
U † (α)HU (α) = H, (5.1)
where H is the Hamiltonian and U represents the unitary transformation over some coordinate
θ by a continuous variable α. One can write such a transformation as
U = e−iGα/ℏ , (5.2)
where G is an operator which generates the transformation. If the operator G behaves as
G = iℏ∂θ , (5.3)
then one can quickly see that
1 αn
U = 1 + α∂θ + α2 ∂θ2 + · · · + ∂θ2 + · · · , (5.4)
2 n!
is simply the Taylor expansion, i.e. when it operates on f (θ) it gives f (θ + α).
For small α,
U = 1 − iGα/ℏ, (5.5)
and
U † (α)HU (α) = H − i[H, G]α/ℏ, (5.6)
which shows that H must commute with G if the Hamiltonian is to be left invariant under the
transformation. Also, such a commutation implies that ⟨G⟩ is a constant of the motion. In the
Heisenberg representation
d
⟨G⟩ = i⟨[H, G]⟩/ℏ = 0 (5.7)
dt
Thus, the invariance of the Hamiltonian with respect to a transformation by θ is equivalent to
stating that the generator of the transformation, G, is constant in time.
The degeneracies of levels is also intimately related to symmetry. For instance, if one considers
an electron orbiting a proton, even with spin-orbit splitting, the Hamiltonian commutes with
J 2 , and because one may rotate such states into one another by changing the coordinate system,
the (2j + 1) states must be degenerate. If one explicitly breaks the symmetry by adding an
interaction with an external magnetic field, the degeneracy is no longer expected to exist.

95
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

5.2 Parity
The parity operator Π reflects the system about some point, x → −x, y → −y, z → −z, or
equivalently three planar reflections. Parity is an example of a discrete symmetry. One can also
have a system invariant under only a subset of the parity operator, e.g. a reflection about the
x = 0 plane, Πx . Both the coordinate and momentum operators should flip sign under parity.
Because Π2 should return one to the original state, eigenstates of Π should have eigenvalues of
±1.
The angular momentum operator,
L ≡ r × p, (5.8)
is invariant under parity because both ⃗ r and p⃗ change signs. Pseudo-vectors, which are defined
by taking the cross products of two vectors, rotate like vectors but have the opposite behavior
under parity. Examples of pseudo-vectors are angular momenta and the magnetic field B. ⃗ In
fact, defining the direction of pseudo-vectors requires arbitrarily choosing a right-hand vs left-
hand rule, and if parity is conserved, there should not be any behavior that differs by this choice.
For instance, if some atom had its angular momentum (a pseudo-vector) oriented in the z direc-
tion by a magnetic field (also a pseudo-vector) in the z direction, the emission of a charged
particle from radioactive decays should be the same parallel or anti-parallel to the applied field.
After all, whether a direction is anti-parallel or parallel to the field depends on one’s choice of
right- vs left-handed convention, whereas the direction of a charged particle certainly defines
a direction in a non-arbitrary way. In fact the electric current density is a vector. Shockingly
however, experiments proposed by T.D. Lee and performed by Madame Chien-Shiung Wu Wu
in 1956 showed that the weak interaction maximally violates parity by observing the decay of
polarized 60 Co, https://en.wikipedia.org/wiki/Wu_experiment. The angular distribution of
decaying electrons was strongest along the direction of the applied field, with zero probability
density for a decaying electron being observed in the perfectly anti-parallel direction.
Most Hamiltonians are even under parity. In that case, the eigenstates can be either even or
odd because they must be eigenstates of the parity operator. The eigenvalues are ±1 because
Π2 must return one to the original state. If the Hamiltonian has any odd-parity terms, then
the eigenstates can have mixtures of even- and odd-parity components. Further, if some origi-
nal state has parity +1, it would then develop some admixture of parity −1 over time, which
would then not be eigenstates of Π. For instance, the weak interaction of an electron with a
nucleus mixes the 2s and 2p states in the Hydrogen atom. Parity conservation refers to the state
maintaining its parity, and if the Hamiltonian is not perfectly even in parity, parity is not con-
served. In practice, the weak interaction violates parity, and all states have some small, often
immeasurably small, admixtures of opposite parity components.
Finally, we point out that parity allows one to recognize many matrix elements as being zero.
For instance, if one considers the matrix element,

⟨ϕ|A|ψ⟩, (5.9)

where the parities of the two states and the operator are (−1)pϕ , (−1)pA and (−1)pψ , the matrix
element will be zero if pψ + pϕ + pA is an odd number. This can be understood by considering
what should happen when integrating over even or odd functions of the coordinate. One can

96
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

consider a matrix element for an operator O as being written in the form

Z
∗
d3 r ψα,ℓ,m (⃗
r )O(⃗
r )ψβ,ℓ′ ,m′ . (5.10)

∗
For every point ⃗r there is point −⃗
r . If the product ψα Oψβ is odd overall, the integral must
vanish. Even if the operator involves derivative, the argument still holds. The operator Px =
−i∂x changes the parity by a factor of −1, just the same as the operator x.
The parity of wave functions in spherical coordinates depends on the orbital quantum number,

Π|ℓ, mℓ , s, ms ⟩ = (−1)ℓ |ℓ, mℓ , s, ms ⟩. (5.11)

One can see this by considering the form for the Yℓm s. Under parity, cos θ → − cos θ whereas
sin θ does not change. The azimuthal angle ϕ → ϕ + π, or equivalently eimϕ → (−1)m eimϕ .
The harmonics Yℓ,m=ℓ have no ϕ dependence and are even or odd functions of cos θ. The low-
ering operator, defined in Eq. (4.25),
 
±iϕ
∂ ∂
L± = −iℏe ∓i sin θ − cot θ , (5.12)
∂ cos θ ∂ϕ

is even under parity, as the term e±iϕ is odd as is the cos θ dependence. Thus, all the states
within a multiplet have the same parity. Further in the course, multi-particle wave functions
will be considered. In that case, there is a quantum number for the total angular momentum, L.
In that case the parity of the wave function will behave as (−1)L . Because parity only acts on
the spatial coordinates, x, y, z, the intrinsic spin part of the wave functions is unchanged under
parity.
Example 5.1: Identifying Vanishing Matrix Elements by Parity
Which of the following matrix elements are zero due to parity? Here α and β refer to all other
quantum numbers.

a) ⟨α, ℓ = 2, m = 1|x|β, ℓ = 1, m = 0⟩

b) ⟨α, ℓ = 1, m = 1|(x2 + y 2 − z 2 )|β, ℓ = 3, m = 1⟩

c) ⟨α, ℓ = 3, m = 0|xyz|βℓ = 1, m = 0⟩

d) ⟨α, ℓ = 3, m = 0|xyPz |βℓ = 1, m = 0⟩

e) ⟨α, ℓ = 3, m = 0|xyPz |βℓ = 1, m = 0⟩

Solution:

a) Can be non-zero. The operator is odd under parity, and the bra and the ket have different
parities.

b) Can be non-zero. In this case the operator is even under parity and the bra and ket have
the same parities.

97
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

c) Must be zero. The bra and kets have the same parity, but the operator is odd under
parity.

d) Must be zero. The operator has the same parity as (c).

e) Can be non-zero. The operator Lz has even parity, so the operator has even parity and
can link states of the same parity.

5.3 Time Reversal

Time reversal is an odd kind of symmetry. It suggests that a motion picture of a physical event
could be run in reverse without the viewer being able to tell something is wrong. Of course, this
does not apply whenever dissipation (friction) is included. In classical mechanics the motion of a
particle in a potential V (r) should be equally valid when watched in reverse. In electrodynam-
ics, the reverse motion should be fine as long as the direction of the magnetic field is reversed.
Summarizing time reversal behavior of particles in classical electrodynamics,
r→⃗
⃗ v → −⃗
r, ⃗ v , E → E, B → −B, j → −j, ρ → ρ. (5.13)
should lead to an equally valid evolution. To understand the signs above, imagine a charged
particle moving in an electric field from times t0 to tf . If at tf you reversed the velocity, the
particle would trace out the identical trajectory. However, to get a particle to retrace its trajectory
in a magnetic field, one would have to flip the sign of the magnetic field. The charge density is
not a function of time reversal, and because the velocity clearly changes sign, and because the
current density is ρ⃗
v , clearly the current density is odd under time reversal.
The electric potential Φ and the vector potential A ⃗ appear in the combinations, H0 − eΦ and
P⃗ − eA/c,
⃗ so one expects Φ and A ⃗ to have the same time-reversal properties as the Hamiltonian
and as the momentum, respectively, i.e. Φ is even and A ⃗ should be odd. In fact, for all Lorentz
four-vectors, the three spatial components must have the opposite behavior under time reversal
as the “zeroth “ component.
In quantum mechanics, we expect the following behavior under time reversal,
t → −t, r → r, p → −p. (5.14)
The gradient operator and ∂t should be even and odd, respectively, under time reversal. How-
⃗ = −iℏ∇ and H = iℏ∂t are odd and even respectively if they are to represent the
ever P
momentum and energy. Thus, for quantum mechanics, the time reversal operator must also
involve taking the complex conjugate. Under time reversal,
iℏ∂t → iℏ∂t , (5.15)
−iℏ∇ → iℏ∇.
One may also see that time reversal entails taking the complex conjugate by considering Schrödin-
ger’s wave equation,
ℏ2 ∂ 2 ∂
− Ψ(x, t) + V (x)Ψ(x, t) = iℏ Ψ(x, t). (5.16)
2m ∂x2 ∂t
98
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

If Ψ(x, t) = ψ(x)e−iEt is a solution, taking the complex conjugate and reversing the time also
gives a solution to the Hamiltonian. In fact, because H is real one may write ψ(x) as a purely
real function or purely imaginary function. A purely real function ψ(x) would be even under
time reversal symmetry while a purely imaginary function would be odd. For instance a plane
wave solution, ψ(x) is neither even nor odd as it has both real and imaginary parts. In that
case taking the complex conjugate changes the direction of the momentum, thus signifying the
motion reversal mentioned before.
Considering the example of plane waves,
Ψ(x, t) = e−iEt+ikx . (5.17)
By both taking the complex conjugate and reversing the sign of t, one obtains a new solution
where the momentum is reversed but the energy remains the same. These are eigenstates of
the Hamiltonian, but are not eigenstates of the time-reversal operator, even though the time-
reversal operator commutes with the Hamiltonian. This puzzle resolves itself by realizing that
the time-reversed state has the same energy, so one can make combinations,
1
Ψ± (x, t) = √ e−iEt+ikx ± e−iEt+ikx ,


2
which are eigenstates of both the Hamiltonian, and of the time-reversal operator, with eigenval-
ues ±1.
Of course, when one gets away from the Schrödinger equation, there are many Hamiltonians
which are not real but are Hermitian, e.g. those that involve σy . Nonetheless, there usually exists
a time-reversal symmetry, but expressing the symmetry may be more complicated than merely
taking the complex conjugate. In particle physics, the symmetry of switching times is linked to
the existence of solutions with opposite energy (anti-particles). In that case the operation which
includes taking the complex conjugate is associated with finding negative-energy or antiparticle
solutions, but that is material for another course.
The time reversal operator acts only to the right because it entails taking the complex conjugate.
In terms of bras and kets the time-reversed matrix elements satisfy the relations below, using the
notation, |α̃⟩ = Θ|α⟩, α̃i = α∗i .
⟨β̃|α̃⟩ = (βi∗ )∗ α∗i = α∗i βi = ⟨α|β⟩ = ⟨β|α⟩∗ . (5.18)
Similarly, the expectation of operators satisfies the relations,
⟨β̃|ΘAΘ−1 |α̃⟩ = ⟨β̃|ΘA|α⟩ = (βi∗ )∗ A∗ij α∗j = ⟨α|A† |β⟩ = ⟨β|A|α⟩∗ , (5.19)
or is equivalent to saying that the time reversed operator ΘAΘ−1 sandwiched between time-
reversed states gives the complex conjugate of the same matrix element without the time rever-
sals.
Most operators of interest are either even or odd under time reversal.
ΘBΘ−1 = ±B (5.20)
Clearly an operator that is a linear combination of an odd and an even operator would be neither.
Examples of odd and even operators are:
⃗ Θ−1 = −P
ΘP ⃗ , ΘAΘ
⃗ −1 = −A,
⃗ Θ⃗
r Θ−1 = ⃗ ⃗ −1 = −L,
r , ΘLΘ ⃗ (5.21)

99
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

⃗ is the electromagnetic vector potential.

where A
Note that the commutation relations,

[x, p] = iℏ , [Li , Lj ] = iℏϵijk Lk , (5.22)

have apparently different behaviors under reversal on the left and right-hand sides of the equa-
tions, until one remembers to consider the iℏ, which flips sign under the time-reversal operator.
Notice that the raising and lowering operator for a harmonic oscillator, x + ip, is even under
time reversal.
Hamiltonians are usually invariant under time reversal. Here we list a few terms which might
appear in a Hamiltonian and discuss whether they violate time reversal or parity.

1. p2 /2m is invariant under both.

2. p · r is invariant under parity but not time reversal.
3. L · p is invariant under time reversal but not parity.
4. S · B and p · A are invariant under both.

5.4 Time Reversal and Angular Momentum

The time reversal operator acts in a surprisingly complex manner when operating on eigenstates
of angular momentum. For integer-spin particles, one can understand the behavior by consider-
ing properties of the Yℓ,m s. By taking the complex conjugates, one sees

Θ|j, m⟩ = (−1)m |j, −m⟩. (5.23)

However, we should realize that the (−1)m is basically the result of a phase convention and
should not be taken too seriously.
The case of spin 1/2 particles is more surprising. In this case the spins cannot be represented
by Yℓ,m s and one must instead consider the two-component system where |+⟩ and |−⟩ refer
to spin-up and spin-down with the direction being chosen along the z axis. In this basis an
arbitrary normalized state, with spin in some arbitrary direction, may be written,
 
iδ1 cos θ/2
|α⟩ = e . (5.24)
eiγ1 sin θ/2

Because the time-reversed state must be a state with opposite spin, it must be orthogonal. The
time reversal operator must create the orthogonal state,

−e−iγ1 sin θ/2

 
iδ2
Θ|α⟩ = e (5.25)
cos θ/2

The phases δ1 and δ2 are arbitrary, and labeling their difference as η ≡ δ2 − δ1 we see that Θ
must equal  
iη 0 1
Θ=e K, (5.26)
−1 0

100
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

where K is an operator which takes the complex conjugate of any quantities to the right. The
surprising property of Θ is viewed by squaring the operator,
 2
2 0 1
Θ = = −1, (5.27)
−1 0

independent of the choice of η.

Making an arbitrary choice of eiη = −i, one sees that

Θ|+⟩ = i|−⟩ , Θ|−⟩ = −i|+⟩. (5.28)

This allows one to summarize both the integral and half-integral results with

Θ|j, m⟩ = i2m |j, −m⟩ (5.29)

Thus, performing two time-reversal operations does not return the same state in the case of
half-integral spin, but it does for the case of integral spin.
To make the time-reversal operator in spin 1/2 systems seem a bit more peculiar it is often
written as
Θ = σy K, (5.30)
which makes the operator look odd because the y direction appears to be preferentially singled
out. However, this has nothing to do with the y direction, but only with the fact that the only the
operator which flips spin must change the place of the two components and change one sign to
make an orthogonal state. One could have associated any of the three sigma matrices with the
x, y and z directions, but the time reversal operator must be proportional to the anti-symmetric
off-diagonal matrix.
Note that the eigenstates of L2 and Lz are not eigenstates of the time-reversal operator, unless the
projection m equals zero, because they have the form eimϕ . However, one can make eigenstates √
of the time reversal operator by combining the states |m⟩ and | − m⟩, e.g. (|m⟩ ± | − m⟩)/ 2.

5.5 A Particle in a Periodic Potential

A second class of discrete symmetries involves particles moving in a periodic lattice, which for
our purposes will be considered to be a one-dimensional chain. Two- and three-dimensional
lattices are more difficult to describe, but many basic features, such as the existence of band
structure, occur in all dimensionalities. The translation operator, τa , translates the wave func-
tion by a distance a, where the lattice separation is a. Translational invariance implies that τa
commutes with the Hamiltonian,

τa Hτa−1 = H or V (x + a) = V (x). (5.31)

One thus expects the solution to be an eigenstate of the translation operator,

τa ψk (x) = ψk (x + a) = eika ψk (x), (5.32)

where the eigenvalue is eika . We label the eigenvalue by k, but emphasize that k is not the
eigenvalue of the momentum operator and |k⟩ is not an eigenstate of the momentum operator.

101
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

Instead, k is a label that denotes that the eigenvalue of τa is eika . The translation operator τa
commutes with the Hamiltonian, but it is not Hermitian. Thus, the eigenvalue can be complex.
If the physics is to be unchanged by translation, the state must change only by a phase factor.
Hence, the form of the eigenvalue can be chosen as eika with ka being real. The value ka can
take on any value, but must be in a range of ∆ka = 2π, e.g. −π < ka < π. Any value
of ka outside this range gives the same eigenvalue as one inside the range where you pick a
value within the range that differs by integral numbers of 2π. For a given ka, one can have a
multitude of eigenstates, as will be illustrated below.
The Kronig-Penney model is enormously useful in illustrating how band structure emerges from
the periodicity of a lattice. It describes a one-dimensional geometry with delta-function poten-
tials separated by a,
X∞
V (x) = βδ(x − na). (5.33)
n=−∞

For positions 0 < x < a there is no potential energy, so if the eigenenergy is ℏ2 q 2 /2m, the form
of the wave function is
ψ(x) = eiqx + Be−iqx , (5.34)
with boundary conditions,
d d 2mβ
ψ(a) = eika ψ(0) , ψ(x)|x=a−ϵ − eika ψ(x)|x=ϵ + ψ(a) = 0. (5.35)
dx dx ℏ2
The boundary conditions become
eiqa + Be−iqa = eika (1 + B) (5.36)
2mβ
iq eiqa − Be−iqa − eika + eika B = − 2 eika (1 + B)


ℏ
Eliminating B, one can find a transcendental expression for q.
p sin(qa) + 2q cos(qa) − 2q cos(ka) = 0, (5.37)
where p ≡ 2mβ/ℏ2 . Because the solution only depends on cos(ka), there are solutions for
−π < ka < π with the solutions symmetric about k = 0. However, for any k there are a
variety of solutions q to the transcendental equation. Plotting the solutions as a function of k
yields bands. In the limit p → 0, the solutions are given by q = k, otherwise they are shifted,
especially near the points where cos(ka) = ±1 as seen in the example in Fig. 5.1.
The bands refer to the energies that give solutions. Those energies, ℏ2 q 2 /2m, can be found for
the values of qa, which are plotted along the vertical axis in Fig. 5.1. There are values of qa, or
equivalently of the energy, that are not solutions for some value of ka. These are called gaps,
and the gap typically refers to the jump in energy between bands. If the strength of the potential,
β were set to zero, then the term proportional to p in Eq. (5.37) would vanish, and one would
find
2nπ
q(β = 0) = k ± , (5.38)
a
ℏ2 q 2 ℏ2 (k ± 2nπ/a)2
E(k) = = ,
2m 2m

102
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

4π

3π
Figure 5.1: First four bands of the Kronig Penny
qa

2π model from Eq. (5.37) where the lattice spacing is a,

and for a coupling where pa = 5. Energies would
be given by ℏ2 q 2 /2m.
π

0−π π
−π/2 0 π/2
ka
where n = 0 refers to the first band. So, by mapping k → k ± 2nπ/a, the energies would be
the same as for free particles.
The structure of bands plays a critical role in understanding the electronic properties of solids.
Depending on the density of electrons, the bands are filled up to some level. If that level occurs
in the middle of a band, it is easy for electrons to become excited within the band and arrange
themselves with more positive momenta than negative momenta and yield a current. Such ma-
terials can be good conductors. If the density of electrons is such that the cutoff for filling levels
is between bands, then it requires energy to overcome the band gap to create currents. These
materials are insulators. If the cutoff is between two bands with a small gap (compared to the
temperature), the material is a semi-conductor. This concept extends to multiple dimensions,
and the band structure will depend on the geometrical symmetries of the underlying ionic lat-
tice.

5.6 Object with Discrete Rotational Symmetry

Another example of a discrete symmetry is that of the rotations of a symmetrical object. Unlike
the Kronig-Penney model, there is no particle moving in a symmetric potential. Rather, the
symmetry is the object itself. Even an object with no noticeable symmetry will be invariant
under rotations of 2π. In two-dimensions, an oval is invariant under rotations by π, and a
square is invariant under rotations of π/2. The wave function of such an object would use and
angle ϕ to describe the position, but the angle would only have values between zero and π for
the oval or between zero and π/4 for the square. Rotations for larger angles would be equivalent
to rotations within the smaller range. The wave functions would still behave as eimϕ , but with
the constraint that

eimΦ = 1 (5.39)

when Φ corresponds to a rotation that brings one back to the original state. For the oval,
where Φ = π, this then constrains m to the values 0, 2, 4, 6 · · · , whereas for the square m =

103
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

0, 4, 8, 12, 16 · · · . This requires a perfect symmetry. For example, if the square had distinguish-
able particles located on each corner one could distinguish between a rotation of π/2 and rota-
tion of zero. Thus, the values of m would return to 0, 1, 2, 3 · · · . Even if the particles on each
had the exact same mass and charge, the values would be 0, 1, 2, 3. Thus, the first rotational
state is at a higher energy if the symmetry is higher. In fact if the object was perfectly circular,
n → ∞, one could not rotate it.
This has implications for nuclear and atomic physics. Many nuclei are elliptical and have a
reflection symmetry, ψ(n̂) = ψ(−n̂). These are therefore invariant under rotations by π, and
only even values of n are allowed. For three-dimensional objects only even ℓ are allowed for
elliptic objects, and the sequence of energy levels is constructed with ℓ = 0, 2, 4 · · · . These are
referred to as rotational bands, with the term “band” having a different meaning than in the
Kronig-Penney model. To emphasize the importance of symmetry, one could consider an O2
molecule. If one of the atoms were a different isotope, then both even and odd values of ℓ would
be permitted, but if both were the same isotope only even values of ℓ would be permitted.
Example 5.2: Rotating Chain
Here, we consider an example of a circular chain of mass M and radius R with n equally
spaced identical particles fixed to the chain that can rotate together. The chain exists in two
dimensions. Find the rotational energies.
Solution:
The chain is invariant under rotations of 2π/n. Because this is rotation in a plane, the position
of the chain is specified by a single angle

0 < ϕ < 2π/n.

Any rotation outside this range would be equivalent to a rotation inside the range. The part
of the Hamiltonian representing the rotational kinetic energy would be of the form

L2 ℏ2
HK = =− ∂ϕ2 .
2M R2 2M R2
Because the object rotated by 2π/n is indistinguishable, the boundary condition must be that
the wave function is returned to the same state, within a phase, by a rotation by 2π/n,

ψ(ϕ + 2π/n) = ψ(ϕ).

This differs from the Kronig-Penny model in that the object carries the symmetry, where in the
Kronig-Penny model the object was a point particle moving in a periodic potential. When this
chain is rotated by 2π/n one has the same state, but when an electron is translated a distance
a in a periodic lattice, it is indeed a different state.

ψm (ϕ) = eimϕ , m = · · · , −2n, −n, 0, n, 2n, · · · ,

ℏ2 m2
Em = .
2M R2
If one link in the chain were different, none of these arguments would apply as only rotations
by 2π would return one to the original state and m could be any integer.

104
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

5.7 Exercises
⃗ D(n̂, ϕ), the rotation oper-
1. Let Td⃗ denote the translation operator (displacement vector d);
ator; and π the parity operator. Which, if any, of the following pairs commute? Why?

(a) Td⃗ and Td⃗′ (d⃗ and d⃗′ are in different directions.)
(b) D(n̂, ϕ) and D(n̂′ , ϕ′ ) (n̂ and n̂′ are in different directions.)
(c) Td⃗ and Π.
(d) D(n̂, ϕ) and Π.
2. Because of weak (neutral-current) interactions there is a parity-violating potential between
the atomic electron and the nucleus as follows:
V = λ δ 3 (⃗r )S · p + S · pδ 3 (⃗
 
r)
where S and p are the spin and momentum operators of the electron, and the nucleus is
assumed to be situated at the origin. As a result, the ground state of an alkali atom, usually
characterized by |n, ℓ, j, m⟩ actually contains tiny contributions from other eigenstates as
follows X
|n, ℓ, j, m⟩ → |n, ℓ, j, m⟩ + Cn′ ,ℓ′ ,j ′ ,m′ |n′ , ℓ′ , j ′ , m′ ⟩
n′ ,ℓ′ ,j ′ ,m′

On the basis of symmetry considerations alone, what can you say about (n′ , ℓ′ , j ′ , m′ )
which give rise to non-vanishing contributions?
3. Suppose a spinless particle is bound to a fixed center by a potential V (⃗
r ) so asymmetrical
that no two energy levels are degenerate. Using time-reversal invariance prove
⟨L⟩ = 0.
for any energy eigenstate. Use the fact each eigenwave function ψ(⃗ r ) must be an eigenstate
of the time reversal operator with eigenvalue eiγ , thus ψ ∗ (⃗
r ) = eiγ ψ(⃗r ). Also use the fact
that ⟨α|Li |α⟩ is real because Li is a Hermitian operator. (This is known as quenching of
orbital angular momentum.)
4. Consider the time-reversal operator for spin-1/2 particles, Θ = σy K, where K takes the
complex conjugate of all quantities to its right. Show that Θ commutes with the rotation
operator,
⃗ = cos(θ) + i⃗
R(θ) σ · θ̂ sin(θ).

5. Consider a particle of mass M confined to a two-dimensional circle of radius R.

(a) Write down the Schrödinger equation for the wave function ψ(ϕ), where the potential
depends only on ϕ, and radial motion is ignored.
(b) Assuming the potential is periodic,
V (ϕ + 2π/N ) = V (ϕ),
where N is an integer. Write the boundary condition relating ψ(ϕ) and ψ(ϕ+2π/N ),
where the eigenvalue of the rotation operator, R(2π/N ), is eiγ . What values of γ are
allowed?

105
PHY 851/852 5 SYMMETRIES AND CONSERVATION LAWS

(c) Assume the potential ,

X
V (ϕ) = β δ(ϕ − 2πk/N ),
k=1,N

Assume the wave function has the form,

ψ(ϕ) = eimϕ + Be−imϕ , 0 < ϕ < 2π/N,

where m is not necessarily an integer. Find a transcendental expression for m in terms

of β, M , γ and n. Hint: Note the similarity to the Kronig-Penny model, where the
solution in Eq. (5.37) translates to this problem with qa → mα, ka → γ, and a → α,
with α = 2π/N .

106
PHY 851/852 6 APPROXIMATION METHODS

6 Approximation Methods
Physics is rarely solved exactly, and the art of approximating the correct answer is the essence
of any physics problem. Even if a problem cannot be solved exactly, one may be able to sat-
isfactorily approximate the true solution. A wide range of approximation methods exist, that
vary widely in the circumstances under which they become accurate. For instance, the WKB ap-
proximation provides a crude estimate of tunneling probabilities that is valid when the potential
changes slowly with position. Variational methods encompass a wide variety of approaches and
becomes exact when the true wave function can be well described with some form involving a
few adjustable parameters. For transitions, it is good to understand whether the Hamiltonian is
changing quickly or slowly, as if the time scale describing the change is either much shorter or
much longer than any characteristic frequencies of the problem, one can apply either the sudden
or adiabatic approximation. Finally, if a Hamiltonian differs by a small amount from one which
is exactly solvable, the exact solution can be expressed as a converging sum of perturbative cor-
rections. This chapter covers five different classes of approximation:

1. The WKB Approximation

2. Variational Theory
3. The Sudden Approximation
4. Stationary-State Perturbation Theory
5. Time-Dependent Perturbation Theory

The choice of approximation method depends both on the justifiability of the approximation
and on what is being asked. For instance, time-dependent perturbation theory is applied for
problems where a transition is being studied, such as a decay or a scattering. For addressing
questions about a stationary states, e.g. how is the binding energy of the ground state affected
by a small additional potential, one might apply stationary-state perturbation theory or a vari-
ational method. The methods presented here by no means constitute an exhaustive list. Lattice
gauge theory the method becomes exact in the limit that the space time mesh becomes infinitely
fine.

6.1 The WKB Approximation

The WKB (Wentzel-Kramer-Brillouin) approximation is a useful method for estimating wave
functions and tunneling probabilities for smooth potentials or for potentials with only a few
discontinuities. The WKB approximation for a wave function can be written as:

ψ(x) = A+ (x)eiϕ(x) + A− (x)e−iϕ(x) (6.1)

Z x
ϕ(x) = dx′ p(x′ )/ℏ,

where the lower limit of the integral is absorbed by the arbitrary phases in A+ and A− . The
function p(x) is defined by p
p(x) ≡ 2m(E − V (x)), (6.2)

107
PHY 851/852 6 APPROXIMATION METHODS

and can be thought of the momentum of a classical particle with energy E at position x.
To assess the accuracy of the approximation one can apply the Schrödinger equation to the as-
sumed form for ψ. Here we use only the A+ term.
∂2
(H − E)ψ(x) = −ℏ2 2 ψ(x) − (E − V )ψ(x) (6.3)
 ∂x2 
p(x)
= − E − V (x) A+ (x)eiϕ(x)
2m
 
iℏ ∂ ∂
+ 2p(x) A+ (x) + A+ (x) p(x) eiϕ(x)
2m ∂x ∂x
ℏ2
 2 
∂
− 2
A+ (x) eiϕ(x) .
2m ∂x
The first term disappears from our choice of p(x) while the second term will vanish if we choose
A+ (x) ∝ p(x)−1/2 . (6.4)
The last term does not vanish, but is neglected in the limit that ℏ is small as it is proportional to
ℏ2 . One can understand the accuracy of the approximation by taking the second derivative of
A+ (x) in the last term and comparing it to the other terms. One then sees that the approxima-
tion becomes exact when characteristic length scales of the the potential are much longer than
ℏ/p(x).
Physically, one may understand the p−1/2 dependence of A+ by realizing that the WKB ap-
proximation has no reflection associated with it, at least when E > V and p(x) is real. Thus,
conservation of flux requires p|A|2 to be constant.
To estimate the binding energy, one sets the phase ϕ(x) to be zero at one turning point (a point
where p(x) = 0), then solves for ϕ(x) at the other turning point. By finding an energy for which
the phase changes by π one then has a solution. For one may then combine such a solution with
the A− solution to return to the original turning point incurring a net phase change of 2π.
Example 6.1: Ground State of Harmonic Oscillator in WKB approximation
Consider a particle of mass m in the one-dimensional harmonic oscillator potential,
1
V (x) = mω 2 x2 . (6.5)
2
estimate the energies of the eigenstates.
Solution:
For bound states, one typically applies the WKB approximations between the classical turning
points. This involves assuming that the WKB phase is a multiple of π between the R points
defined by V (x) = E. In this case one fourth of 2nπ must result from integrating k(x)dx
from zero to the turning point.
√ r
nπ 2m a 1 1
Z
= dx mω 2 a2 − mω 2 x2 , (6.6)
2 ℏ 0 2 2
where the turning point a depends on the unknown ground state energy, E = mω 2 a2 /2.

108
PHY 851/852 6 APPROXIMATION METHODS

Solving the equation above for a gives

2nℏ
a2 = (6.7)
mω
E = nℏω,

which is correct aside from an extra ℏω/2. Because it was assumed that the wave function
vanished at the turning point (where p(x) = 0), rather than allowing the wave function to
gradually decay in the region where p(x) is imaginary, one expects it to consistently give an
over-prediction of the energies.

Because other methods tend to be much more accurate, the WKB approach is rarely used to
estimate ground state energies. The most common use of the WKB approximation is to esti-
mate tunneling probabilities. In this case the wave functions have exponentially growing and
decaying amplitudes.

ψ(x) = A+ (x)eϕ(x) + A− (x)e−ϕ(x) (6.8)

Z x
ϕ(x) = dx′ q(x′ )/ℏ,

where the approximation is applied to tunneling regions, V (x) > E. The function q(x) is
defined as p
q(x) ≡ 2m(V (x) − E), (6.9)

In fact, one usually ignores the x dependence of the amplitude and merely states that the tun-
neling probability form going from turning point a to turning point b is
b
2
 Z 
p
Pa→b ≈ exp − dx 2m(V (x) − E) , (6.10)
ℏ a

with the factor of two coming from squaring the amplitude.

Example 6.2: Tunneling Through Barrier
Consider a particle in a potential,
 

 ∞, x<0 

0, 0<x<L
 
V (x) =

 V0, L < x < 2L 

0, x > 2L
 

where the barrier, V0 is large. Estimate the tunneling probability of a particle stuck in the well.
Solution:

109
PHY 851/852 6 APPROXIMATION METHODS

First, estimate the energy of the ground state ignoring any penetration,
Z L √
π= dx 2mE/ℏ
0
L√
= 2mE,
ℏ
ℏ2 π 2
E= .
2mL2
Next, estimate the rate at which the particle collides with the right p
side to try and escape. The
distance back and forth across the well is 2L, and if the velocity is 2E/m, the rate is

v ℏπ
Γtry = = .
2L 2mL2
Finally, one can apply the WKB approximation to estimate the tunneling probability,

2 2L
 Z 
Ptunnel = exp − dx p(x) ,
ℏ L
p p
p(x) = 2m(V (x) − E) = 2m(V0 − E),
2L p
 
Ptunnel = exp − 2m(V0 − E) ,
ℏ
 s 
 2L ℏ π 
2 2
= exp − 2m(V0 − ) .
 ℏ 2mL2 

The factor of two in the exponential is from squaring the wave function, which is assumed to
be unity at x = L. The estimate for the decay rate is then

Γ = Γtry Ptunnel .

Every step of the WKB approximation is crude, but having an estimate of the decay rate through
a barrier can be useful even if the accuracy is only within an order of magnitude given that
decay rates for processes like fission can range from 10−20 seconds to many times the age of the
universe.

6.2 Variational Theory

Variational calculations are used to calculate ground state wave functions by using the simple
fact that all states must have energies greater or equal to that of the ground state. Thus, by
writing a state in terms of some parameters αi , and minimizing E = ⟨⃗ α|H|⃗α⟩ with respect
to the parameters αi , one knows that the energy can never fall below the true ground state
energy. The minimization procedure therefore can be used to estimate the ground state energy
and ground state wave functions.

110
PHY 851/852 6 APPROXIMATION METHODS

Example 6.3: Variational Estimate of Ground State Energy of Hydrogen Atom

Consider the Coulomb potential,
e2
V (r) = − .
r
We suppose that we were lucky and guessed an exponential form for ground-state hydrogen
atom wave function by assuming a trial form,

2
ψ(r) = √ e−r/a ,
4πa3
where a is the variational parameter and the prefactor was chosen to normalize the wave
function. One would then minimize the expectation of ⟨H⟩ with respect to a.

ℏ2 e2
Z  
2 ∗ 2
⟨H⟩ = 4π r dr ψ (r) − (∂ + (2/r)∂r ) − ψ(r),
2m r r

and after doing the integrals,

ℏ2 e2
⟨H⟩ = − .
2ma2 a
The minimization, (∂/∂a)⟨H⟩ = 0, yields

ℏ2
a= ,
me2
e2
⟨H⟩ = − .
2a
In this case this gives the exact wave function and ground state energy, but only due to the for-
tunate choice for the form of the wave function. In general, one would obtain an approximate
wave function with on overestimate of the ground state energy.

Variational calculations are popular in a variety of many-body applications where the interac-
tions and many-body wave functions can be extremely complicated.

6.3 The Sudden Approximation

The sudden approximation is typically applied when calculating transition probabilities for
cases where the Hamiltonian changes rapidly between two times t1 and t2 . The approxima-
tion becomes exact in the limit that the Hamiltonian changes instantaneously. In that limit one
can approximate the transition amplitude by a simple overlap,
P (β → α) = |⟨α|U (t2 , t1 )|β⟩|2 , (6.11)
⟨α|U (t2 , t1 )|β⟩ ≈ ⟨α|β⟩,
where |β⟩ is usually an eigenstate of the Hamiltonian before t1 , and |α⟩ is usually an eigen-
state of the Hamiltonian after t2 . A previous example of a sudden transition was presented in

111
PHY 851/852 6 APPROXIMATION METHODS

Example 2.1.
The sudden approximation can be used whenever the times t1 and t2 are so close that the rel-
evant energy changes multiplied by (t2 − t1 )/ℏ are small. The sudden approximation is com-
monly used in nuclear reaction theory. If a high-energy particle knocks a proton or neutron out
of the nucleus with high energy, the remainder of the nucleus finds itself suddenly in the wrong
wave function, and calculating the probability it will be found in a particular state is approxi-
mated by simply taking the overlap of wave functions.
When a potential is changed slowly, the probability remains assigned to the same state. This
is because for each differential change in the potential, a differential change ϵ is induced in the
wave function. However, if the changes occur at much different times, as would be the case
for a slow change, the differentialPamplitudes contribute with uncorrelated phases and the net
change in the probability goes as |ϵ|2 = 0. Thus, if a particle is in the ground state, and the
well changes slowly, it remains in the ground state afterward. This also implies that entropy is
not generated, hence the term adiabatic.
Example 6.4: Expanding Well
The most common example used to illustrate the sudden approximation is the case of a par-
ticle of mass m in an expanding well. Here, we consider an infinite square well confining
particles to the region 0 < x < a which suddenly expands at time t = 0 to allow particles to
occupy the region 0 < x < 2a.
Assuming a particle was in the ground state of the old well,

a) What is the probability of being in the state n of the new well? The wave functions of
the new eigenstates are
r
1
ψn (x) = sin [nπx/(2a)] , 0 < x < 2a,
a
where the new ground state has n = 1.

b) What is the expectation of the energy ⟨H⟩ after the expansion of the well?

Solution:

112
PHY 851/852 6 APPROXIMATION METHODS

a) Take the overlap of the two ground state wave functions then square it,

P (0 → n′ ) = |M0→n |2 ,
r r Z a
2 1
M0→n = dx sin(πx/a) sin(nπx/2a)
a a 0
√ Z a
2
dx −ei(n+2)πx/2a − ei(−n−2)πx/2a + ei(n−2)πx/2a + ei(−n+2)πx/2a
 
=
4a
√ Z0 a
2
= dx {cos[(n − 2)πx/2a] − cos[(n + 2)πx/2a]}
2a 0
√ 
2 sin[(n + 2)π/2] sin[(n − 2)π/2]

= − +
π (n + 2) (n − 2)
√
2 1 1
 
= sin(nπ/2) −
π n+2 n−2
 √
 4 π 2 (−1)(n+1)/2 /(n2 − 4), n = odd

1
= √
2
, n=2

 0, n > 2 and even
 2 2 2
 32/[π (n − 4) ], n = odd
1
P (0 → n) = 2
, n=2
0, n > 2 and even


The ground state is n = 1 and that probability is 32/9π 2 .

b) The energy of the states n are En = ℏ2 n2 π 2 /8a2 , summing over n multiplied by

P (0 → n) should give the original energy because the wave function does not change
in the instant the well is moved, and the kinetic energy depends only on the form of the
wave function, and not on the potential. Thus, the kinetic energy will not change if the
potential suddenly changes. However, if the potential changes in the region where the
wave function is non-zero, the total energy will change. Thus, one could put in great
effort to perform the sum but the result would be
X ℏ2 π 2
En |M0→n |2 = .
n
2ma2

6.4 Stationary-State Perturbation Theory

Here, we present stationary-state perturbation theory which is also known as Rayleigh-Schrödinger
perturbation theory. Here, one solves for states and their energies in an expansion of powers of
λ where the Hamiltonian is
H = H0 + λV. (6.12)
The potential V is known as the perturbation and is assumed to be small while λ is assumed
to be unity and is only used to keep tally of the expansion of V , i.e. the expansion in λ is an

113
PHY 851/852 6 APPROXIMATION METHODS

expansion in V .
Before one embarks on perturbation theory, one assumes that one has already solved for the
states |n⟩, which are eigenstates of H0 , with eigenenergies ϵn . The goal is to express solutions
for the new eigenenergies and eigenfunctions of the full Hamiltonian, H, as an expansion in
powers of λ, where each term in the expansion is expressed in terms of |n⟩, ⟨m|V |n⟩ and ϵn .
We assume that both the eigenstates and eigenenergies of the new Hamiltonian can be written
as an expansion if powers of λ,

|N ⟩ = |n⟩ + λ|N (1) ⟩ + λ2 |N (2) ⟩ + · · · (6.13)

En = ϵ n + λEn(1) +λ 2
En(2) + ···

Here, the terms |N (j) ⟩ and En(j) denote the corrections to the specific eigenstate |n⟩ and energies
of order λj . We are also free to make an assumption about normalization of the state |N ⟩.

⟨n|N ⟩ = 1, (6.14)

which is equivalent to saying that the additional parts of the wave function have no |n⟩ compo-
nent,
⟨n|N (j) ⟩ = 0, for j > 0. (6.15)

The Schrödinger equation,

(H0 + λV )|N ⟩ = En |N ⟩, (6.16)
must be satisfied to every power of λ individually. By inspecting the j th power of λ, this gives
X
H0 |N (j) ⟩ + V |N (j−1) ⟩ = En(k) |N (j−k) ⟩. (6.17)
k=0,j

Here, the sum over k goes from zero to j with the understanding that |N (0) ⟩ = |n⟩ and E (0) =
ϵn . We solve for the expressions iteratively. That is, one first finds E (k) then finds |N (k) ⟩, then
given those states move onto k + 1. To find En(k)) , one takes the overlap of Eq. (6.17) with ⟨n|
and using the normalization definitions one obtains

⟨n|V |N (j−1) ⟩ = En(j) . (6.18)

For the case where j = 1, one obtains the lowest-order perturbation theory answer for the
energy.
En(1) = ⟨n|V |n⟩. (6.19)

The state |N (j) ⟩ is defined by its overlap with the states ⟨m| where m ̸= n. Given that En(j) is
now known, one can find the ⟨m|N (j) ⟩ by taking the overlap of ⟨m| with Eq. (6.17),
X
ϵm ⟨m|N (j) ⟩ + ⟨m|V |N (j−1) ⟩ = En(k) ⟨m|N (j−k) ⟩. (6.20)
k=0,j

Solving for the j th order, which is unknown,

!
1 X
⟨m|N (j) ⟩ = −⟨m|V |N (j−1) ⟩ + En(k) ⟨m|N (j−k) ⟩ . (6.21)
ϵm − ϵ n k=1,j

114
PHY 851/852 6 APPROXIMATION METHODS

The first-order correction to the wave function is then

X 1
|N (1) ⟩ = − |m⟩ ⟨m|V |n⟩. (6.22)
m̸=n
ϵ m − ϵ n

Using the state |N (1) ⟩, one can then find the expression for En(2) ,
X |⟨m|V |n⟩|2
En(2) = − . (6.23)
m̸=n
ϵm − ϵ n

Several important principles can be realized by observing the form of En(2) . First, two state’s
energies are pushed apart in 2nd-order perturbation theory, a phenomena known as level repul-
sion. Secondly, if the levels are initially close, the energies are more affected. In fact, if they are
degenerate, perturbation theory breaks down, and one must apply degenerate perturbation the-
ory which is the topic of the next sub-section. Of special significance is noticing that the ground
state is always lowered in second-order perturbation theory.
Example 6.5: Harmonic Oscillator with Linear Perturbation
We will find the second-order correction to the ground state energy of the Harmonic oscillator
with a linear perturbation,

ℏ2 1
H0 = − ∂x2 + mω 2 x2 ,
2m 2
V = βx,

Perhaps surprisingly, this gives the exact answer for this case.
Solution:
To proceed one can write the perturbation in terms of creation and destruction operators,

ℏ
r
V =β (a + a† ).
2mω
The ground state then overlaps with only the first excited state,

ℏ
r
⟨1|V |0⟩ = β ,
2mω
and the correction to the energy is

|⟨1|V |0⟩|2 β2
E (2) = − =− .
ℏω 2mω 2
One could have also solved this exactly by completing the square of the potential,
1
V (x) = mω 2 x2 + βx
2
2
1 β2

2
β
= mω x + − ,
2 mω 2 2mω 2

115
PHY 851/852 6 APPROXIMATION METHODS

and the last term is a constant correction to the energy, while the rest of the change is simply
a translation of the potential. Given that the exact solution is proportional to β 2 , it must also
equal the perturbative correction to second order because any other order of correction would
involve β to a higher power.

Finally, we point out that there is a competing scheme for perturbation theory, know as Brillouin-
Wigner theory. It differs in that E (n) appears in the expressions for higher-order corrections to
the wave functions. Operationally, results are the similar as the Rayleigh-Schrödinger form.
The Brillouin-Wigner scheme will not be pursued here, but the interested reader can easily find
descriptions of it in standard texts. Baym’s Lectures on Quantum Mechanics contains a nice
description.

6.5 Degenerate-State Perturbation Theory

Due to terms of the form
⟨m|V |n⟩
,
ϵm − ϵ n
perturbation theory falls apart when the perturbation mixes degenerate states, ϵm = ϵn . This
can be corrected by first separating the part of the potential that mixes the degenerate states, Vd ,
from the remainder of the potential, V ′ , then diagonalizing Vd and using perturbation theory
for V ′ .
Formally, this means dividing the potential into parts using projection operators,
V = Vd + V ′ (6.24)
Vd = Pd V Pd
V ′ = (I − Pd )V (I − Pd ) + Pd V (I − Pd ) + (I − Pd )V Pd ,
where Pd is the projection operator that projects the subset of states that are degenerate
X
Pd = |m⟩⟨m|. (6.25)
m∈d

Because ⟨m1 |V ′ |m2 ⟩ = 0 when m1 and m2 are in the degenerate set, there is no longer any
problem. Note that by diagonalizing Vd , Vd essentially becomes part of H0 .
Example 6.6: Stark Effect
Here, we consider the diagonalization of Vd and forget about the perturbative part entirely.
The Stark effect refers to the placement of a hydrogen atom in an electric field,

V = eEz.

and are interested in calculating the splitting of the 2s and 2p levels. In this case the matrix
elements of V are zero along the diagonal of the 4×4 matrix that describes V . This arises
because the three p states and the s state have specific parities while the interaction is has
odd parity, and an odd-parity interaction can only result in non-zero matrix elements when
sandwiched between states of opposite parities. The s state can be mixed with the p states by

116
PHY 851/852 6 APPROXIMATION METHODS

the interaction. Thus, there is an off-diagonal matrix element

⟨n = 2, ℓ = 1, mℓ = 0|V |n = 2, ℓ = 0⟩ ̸= 0.

All other matrix elements are zero. Using the forms for the wave functions given in a previous
lecture, one can solve for the matrix element,
Z
⟨2s|V |2p, m = 0⟩ = dΩr 2 drR2,0 (r)R2,1 (r)r cos θY0,0 (θ, ϕ)Y1,0 (θ, ϕ) = 3ea0 |E|.

Because Vd , in the relevant part of the Hamiltonian matrix is a 2 × 2 matrix proportional

to σx , the eigenstates of Vd are ±3ea0 |E|. One can then apply perturbation theory to find
corrections due to mixing with other shells. When going forward, one must use eigenstates of
Vd as the basis.

6.6 The Interaction Representation (Picture)

Before one can begin time-dependent perturbation theory, we need to present the interaction
representation, which is an alternative to the Schrödinger and Heisenberg representations. Sum-
marizing the representations,

|ψ(t)⟩S = e−iHt/ℏ |ψ(t = 0)⟩S , (6.26)

|ψ⟩H = |ψ(t = 0)⟩S (Heisenberg states are fixed in time),
|ψ(t)⟩I = eiH0 t/ℏ |ψ(t)⟩S = eiH0 t/ℏ e−iHt/ℏ |ψ(t = 0)⟩S ,

where the Hamiltonian is divided into an understood part, H0 , and a perturbation V . In almost
all instances, the basis is chosen to correspond to eigenstates of H0 . Note that if V = 0 that the
interaction states become equal to the Heisenberg states.
Because matrix elements ⟨ϕ|ABC|ψ⟩ must not depend on the representation, operators in the
Heisenberg and Interaction representations must be re-defined. In terms of AS , the operator in
the Schrödinger representation, then become

AH (t) = eiHt/ℏ AS e−iHt/ℏ (6.27)

AI (t) = eiH0 t/ℏ AS e−iH0 t/ℏ ,

where we have assumed that H is not time dependent. Otherwise, e−iHt/ℏ would be replaced
with an evolution operator.
Solving for the evolution of a state in the interaction representation,

d d
iℏ |ψ(t)⟩I = iℏ eiH0 t/ℏ |ψ(t)⟩S (6.28)
dt dt
= −eiH0 t/ℏ H0 |ψ(t)⟩S + eiH0 t/ℏ (H0 + V )|ψ(t)⟩S
= eiH0 t/ℏ V e−iH0 t/ℏ eiH0 t/ℏ |ψ(t)⟩S
= VI (t)|ψ(t)⟩I .

117
PHY 851/852 6 APPROXIMATION METHODS

The subscript I is omitted in most literature and the knowledge of which representation is being
used is almost always left to the astute reader, who is assumed to recognize the representation
by context. If the potential has an explicit time dependence, the explicit time dependence must
be absorbed into VI (t). Working in a basis of eigenstates of H0 ,
X
VS (t) = βmn (t)|m⟩⟨n|, (6.29)
mn
X
VI (t) = βmn (t)ei(ωm −ωn )t |m⟩⟨n|,
mn

where the eigenstates of H0 are ℏωn . In many applications the potential does not explicitly
depend on time and βmn would not depend on time.

6.7 Time-Dependent Perturbation Theory

Many problems in quantum mechanics involve time-dependent interactions. Obvious examples
include spin magnetic resonance problems where the interaction explicitly varies in time as the
magnetic field is oscillating. A less obvious example is a scattering problem where the incom-
ing wave packet slowly enters the region where it feels the potential then leaves. This problem
is treated by considering asymptotic momentum states with a potential that slowly turns off
and on with time, rather than with a fixed potential with wave packets. In fact, nearly all per-
turbative scattering (e.g. Feynmann diagrams) treatments are applications of time-dependent
perturbation theory even though the potential is not actually varying with time.
Using Eq. (6.28), the evolution operator in the interaction picture is defined by

|ψ(t)⟩I = UI (t, t0 )|ψ(t0 )⟩I , (6.30)

iH0 (t−t0 )/ℏ −iH(t−t0 )/ℏ
=e e |ψ(t0 )⟩I ,
iH0 (t−t0 )/ℏ −iH(t−t0 )/ℏ
UI (t, t0 ) = e e .

The equations of motion for the evolution operator are then,

d d
eiH0 t/ℏ e−iHt/ℏ


iℏ U (t, t0 ) = iℏ (6.31)
dt dt
= eiH0 t/ℏ V e−iHt/ℏ
= VI (t)U (t, t0 ).

The boundary condition is

U (t = t0 , t0 ) = 1. (6.32)
One can rewrite the differential equation as an integral equation,
t
i
Z
UI (t, t0 ) = 1 − dt′ VI (t′ )UI (t′ , t0 ). (6.33)
ℏ t0

One can see that this is a solution to the differential equation by taking derivatives of it w.r.t.
time. One can take the expression for UI on the left-hand side and insert into the place of

118
PHY 851/852 6 APPROXIMATION METHODS

UI (t′ , t0 ) on the right-hand side. If one continues this substitution ad nauseum,

 2 Z t Z t′
i t ′ −i
Z
′ ′
UI (t, t0 ) = 1 − dt VI (t ) + dt dt′′ VI (t′ )VI (t′′ ) (6.34)
ℏ t0 ℏ t0 t0
 n Z t Z t′ Z t(n−1)
−i ′ ′′
··· + dt dt · · · dt(n) VI (t′ )VI (t′′ ) · · · VI (t(n) ) + · · · .
ℏ t0 t0 t0

The integral equation is known as the Dyson series, named after Freeman J. Dyson, https:
//en.wikipedia.org/wiki/Freeman_Dyson. Keeping only the first term in the expansion in V
amounts to first-order perturbation theory.
Also note the the operator UI is related to the Schrödinger evolution operator by
US = e−iH0 t/ℏ UI eiH0 t/ℏ . (6.35)
Transition probabilities, which are the square of matrix elements |⟨n|US |i⟩|2 with n and i refer-
ring to eigenstates of H0 , are similarly |⟨n|UI |i⟩|2 in the interaction representation.
To second order perturbation theory, the matrix element becomes
i t ′
Z
⟨n|UI (t, t0 )|i⟩ = − dt ⟨n|VS (t′ )|i⟩ei(ϵn −ϵi )t/ℏ (6.36)
ℏ t0
 2 X Z t Z t′
−i ′ ′ ′′
+ dt dt′′ ⟨n|VS (t′ )|m⟩⟨m|VS (t′′ )|n⟩ei(ϵn −ϵm )t /ℏ ei(ϵm −ϵi )t /ℏ .
ℏ m t0 t0

It is straight-forward to find expressions for the higher-order terms, but the expressions become
lengthy. Note that the this expression was accomplished by invoking
⟨n|VI (t)|m⟩ = ⟨n|VS (t)|m⟩ei(ϵn −ϵm )t/ℏ , (6.37)
where the Schrödinger operator VS (t) usually has no time dependence unless the interaction
explicitly depends on time.

6.8 Fermi’s Golden Rule

We will now derive one of the most important expressions derived in this class, Fermi’s Golden
Rule for the transion rate from the state |i⟩ to the state |n⟩, where |i⟩ and |n⟩ are eigenstates of
H0 . The transition probability will be calculated to first order in perturbation theory. We will
derive the same expression twice to demonstrate how robust the derivation is.
First we consider the case where we turn on the perturbation slowly according to an exponential,
⟨n|VS (t)|m⟩ = eηt Vnm , η → 0+ (6.38)
where the times being considered are less than or near zero, meaning the interaction was turned
on slowly from t = −∞. To first order,
Z t
i ′ ′
⟨n|UI (t, −∞)|i⟩ = − Vni dt′ ei(ϵn −ϵi )t /ℏ+ηt (6.39)
ℏ −∞
Vni
= ei(ϵn −ϵi )t/ℏ+ηt
ϵn − ϵi − iℏη

119
PHY 851/852 6 APPROXIMATION METHODS

Thus, the probability goes as,

Pi→n (t) = |⟨n|UI (t, −∞)|i⟩|2 (6.40)

|Vni |2
= e2ηt ,
(ϵn − ϵi )2 + ℏ2 η 2

In obtain Fermi’s Golden Rule, the goal is to find the transition rate,

d
Ri→n = Pi→n (t) (6.41)
dt
2η|Vni |2
= ,
(ϵn − ϵi )2 + ℏ2 η 2

where we have made the approximation that we are considering the rate at a time where ηt is
small. From an end-of-chapter problem, you can see that the η-dependent terms can be replaced
with a δ function as η → 0+ ,
2π
Ri→n (t) = |Vni |2 δ(ϵn − ϵi ) (6.42)
ℏ

Thus, there is only a non-zero transition rate if the final state has the same energy as the initial
state. This works for scattering or decays, where indeed there are many final states with a given
energy. The δ function looks a bit peculiar, but makes sense when the states n are in a continuum.
For instance, one does not calculate the electromagnetic decay rate of a radioactive nucleus to
a specific state where a photon has a momentum k, but instead to any state within a specific
angle. By summing over all such states,
R and using the density of states to transform the sum to
an integral over the photon energy, dϵγ · · · , the delta function is effectively replaced by the
density of photon states.

Aside: Changing Sum over Momentum States to Integral

Fermi’s golden rule features a delta function in energy, δ(Ef − Ei ). This suggests you can only
transform to states with exactly the same energy, and if the state does have exactly the same
energy, the transition rate would explode. In reality, the delta function has a finite width, ℏη,
and only as η → 0, δ(Ef − Ei ) becomes a true delta function. Another limit one considers
is that momentum states are defined in some length, or some volume in three dimensions, that
approaches infinity. Summing over final states means summing over final momenta, which in
the limit of L → ∞ means that the number of states within some small energy range will
become infinity. Thus, we will assume that η → 0, but that the length approaches ∞ such that
ηL → ∞.

120
PHY 851/852 6 APPROXIMATION METHODS

As shown in Chapter 2, if one has a one-dimensional infinite square well of length L, the bound-
ary conditions give
sin(kL) = 0, (6.43)
kL = nπ, (k > 0)
dn L
= ,
dk π
If one considers momentum states, rather than the sin(kx) forms which describe a linear com-
bination of both forward and backward waves, the number of states per dk is
dn L
= , − ∞ < k < ∞. (6.44)
dk (2π)
One can extend this into three dimensions where the infinite volume is Ω,
d3 n Ω
= . (6.45)
d3 k (2π)3
If the particle also had spin, a factor of (2s + 1) would be added. The sum over states in three
dimensions thus becomes
Z ∞
X L
→ dk, one dimension, (6.46)
k
2π −∞
Z ∞
X A
→ 2
d2 k, two dimensions,
k
(2π) −∞

Ω
X Z
→ 3
d3 k, three dimensions.
k
(2π)

For Fermi’s golden rule, the delta function is expressed in energy, so in order to integrate the
differential dk must be changed to dE,
dE dE
dk = , = (6.47)
dE/dk ℏvk
m
= dE 2 , non − relativistic,
ℏ k
E
= dE 2 , relativistic,
ℏ k
dE
= , massless.
ℏc
Here, vk = dE/dp is the velocity of a particle of momentum p. Integrals of the type one
encounters with Fermi’s Golden rule can then be easily performed,
f (k) 
Z 
dk f (k)δ(Ef − Ei ) = . (6.48)
ℏvk with k such that Ef =Ei

121
PHY 851/852 6 APPROXIMATION METHODS

6.9 Harmonic Perturbations

Rather than having a potential that turned on slowly, one could envisage a potential with an
explicit time dependence characterized by some frequency ω,

1
⟨n|VS (t)|m⟩ = Vnm eηt cos(ωt) = Vnm eηt eiωt + e−iωt .


(6.49)
2
One can then follow the same derivation by noting that the only two differences are the factor
of 1/2, which get squared to obtain a probability, and the extra phases in the two terms. If we
consider each term separately eiωt and e−iωt , one can write the answer without much thought.

d 2π |Vni |2
Pi→n (t) = [δ(ϵn − ϵi + ℏω) + δ(ϵn − ϵi − ℏω)] . (6.50)
dt ℏ 4
Additionally there would be cross term from squaring the matrix elements, but that term be-
haves as cos(2ωt) and is disregarded when averaging over time.
Example 6.7: Exciting a Particle with a Radiative-Like Interaction
Consider a particle of mass m in the ground state of a δ function potential,

V0 (x) = −βδ(x)

The particle feels a harmonic potential

V (t) = eEx cos(ωt), ℏω > |G.S. energy|

Estimate the ionization rate using first-order perturbation theory. To simplify the problem,
assume the outgoing momentum is high enough that the outgoing wave can be treated as a
plane wave, i.e. the corrections due to the delta function potential are small. This is a one-
dimensional example that has much in common with radiative excitation.
To solve this problem,

1. Find the G.S. wave function and G.S energy of the delta function potential.
√
2. Calculate the matrix element ⟨k|V |0⟩ using a normalized state ⟨x|k⟩ = eikx / L where
L is an arbitrarily large size to the box.
3. Sum the probability over states by writing it as a density of states. Notice how the size
of the box falls out of the problem.

(1) The solution to the delta function potential from Example (2.4), and the plane wave func-
tions are,
p mβ
ψ0 (x) = q/2e−q|x| , q = ,
ℏ2
eikx p
ψk (x) = √ , k = 2m(ℏω − B)/ℏ2 ,
L

122
PHY 851/852 6 APPROXIMATION METHODS

where B = ℏ2 q 2 /2m is the binding energy.

(2) The matrix element is
r
q
Z
Vk0 = dx eExe−ikx e−q|x|
2L
r
q
Z
= −ieE dx x sin(kx)e−q|x|
2L
r
q d
Z
= ieE dx cos(kx)e−q|x|
2L dk
r
1 1
 
q d
= ieE +
2L dk q + ik q − ik
r 3/2
2 2kq
= ieE .
L (q 2 + k2 )2

(3) Using Fermi’s golden rule, the ionization rate is

π Ldk
Z
R= |Vk0 |2 δ(ϵk − ℏω − B)
2ℏ 2π
1 e2 E 2 k2 q 3 4
=
ℏ (q 2 + k2 )4 |dϵk /dk|
4m e2 E 2 kq 3
= 3 .
ℏ (q 2 + k2 )4

A factor of two entered the expression because the delta function picked up contributions
from ±k.
As stated in the problem,√an approximation was made in that the outgoing wave function was
assigned the form eikx / L. More realistically, the outgoing wave should have been a plane
wave modified by the delta function potential. In that case, the form would be

1 e + Ae−ikx , x > 0
 ikx
ψk (x) = √ .
L Beikx , x < 0

A similar form would have been used for the solution with outgoing wave e−ikx . One could
have solved for A and B, then calculated the integrals as done before. This improvement
requires some work, but is certainly tractable. In nuclear physics fixing the outgoing wave to
include the potential is known as the distorted wave Born approximation (DWBA).

6.10 An Exactly Solvable Model of a Time-Dependent Interaction

The problem described here is quite specific, as it only applies to two-component problems.
However, it does illustrate how time dependence manifests itself in the evolution. Consider two

123
PHY 851/852 6 APPROXIMATION METHODS

states |1⟩ and |2⟩ interacting through the time-dependent potential,

ℏ ℏ
H0 = ℏω1 |1⟩⟨1| + ℏω2 |2⟩⟨2| = (ω1 + ω2 )I + (ω1 − ω2 )σz
2 2
Vt = γ cos(ωt)σx + γ sin(ωt)σy

or written as a matrix,
E1 γe−iωt
 
H = . (6.51)
γeiωt E2
This is the form of an interaction with a time-dependent magnetic field,

H = H0 + µ ⃗
⃗ · B(t), B(t) = B0 ẑ + B⊥ [x̂ cos(ωt) + ŷ sin(ωt)] . (6.52)

One can make the substitution,

E1 + E2
|ψ⟩ → e−iω̄t |ψ⟩, ω̄ ≡ (6.53)
2ℏ
to write the new Hamiltonian as
ω12 /2 γe−iωt
 
H = , (6.54)
γeiωt −ω12 /2

where ω12 ≡ (E1 − E2 )/2.

Writing the evolution in the interaction picture for the components ψ1 and ψ2 ,

d ω12 γ
ψ1 (t) = i ψ1 (t) + i e−iωt ψ2 (t)
dt 2 ℏ
d ω12 γ
ψ2 (t) = −i ψ2 (t) + i eiωt ψ1 (t)
dt 2 ℏ

Now, by making the substitution,

ψ1′ ≡ eiωt/2 ψ1
ψ2′ ≡ e−iωt/2 ψ2 ,

one can derive the evolutions,

d (ω12 − ω) γ
ψ1′ (t) = −i ψ1′ (t) + ψ2′ (t)
dt 2 ℏ
d (ω12 − ω) γ
ψ2′ (t) = i ψ2′ (t) + ψ1′ (t).
dt 2 ℏ
The problem now looks like a Hamiltonian without a time dependent interaction,

ℏ(ω12 − ω)
H′ = σz + γσy . (6.55)
2
124
PHY 851/852 6 APPROXIMATION METHODS

This is the same problem as we worked out for neutrino oscillations. The evolution operator
becomes
′ t/ℏ
e−iH = cos(Ωt) + σn sin(Ωt),
s
(ω12 − ω)2 γ2
Ω= + 2
4 ℏ
σn = cos(θ)σz + sin(θ)σy
2γ
tan(θ) =
ℏ(ω12 − ω)

If the state begins polarized in the z direction, the maximum probability of becoming a spin-
down state is
−iH ′ t/ℏ 2
γ 2 /ℏ2
max |⟨↓ |e | ↑⟩| = 2 2 (6.56)
γ /ℏ + (ω − ω12 )2 /4
The resonant form in this example clearly displays that driving the system at the resonant fre-
quency, ω = ω12 , results in the greatest chance for flipping the spin. NMR works on very similar
principles, only in this case the time-dependent field usually oscillates only in one plane, i.e.
⃗
B(t) = B0 ẑ + B⊥ x̂ cos(ωt).
This is a bit harder to work out as compared to our example but the resonant conditions remain
the same.
The functional form above,
1 Γ/2
f (ω) = , (6.57)
π 1 + (ω − ωr )2 /(Γ/2)2
is known as a Lorentzian. This function has a maximum at ω = ωR , and integrates to unity. The
parameter Γ is known as the width, or the half-width, meaning that the width of the function
measured at half its maximum height is Γ. Lorentzian forms show up in resonant forms, e.g. for
driven harmonic oscillators. When the driving frequency, ω, matches the fundamental, or reso-
nant, frequency, ωR , the resulting amplitude is a maximum. In quantum mechanics, adjusting
the energy can be thought of as adjusting the frequency, and when an energy from some channel
matches the energy of a different channel, the channels couple most strongly. For instance, if the
beam energy of scattering matches the energy of an unstable state, the scattering cross sections
often maximize at that energy. For this reason, excited states encountered in nuclear or particle
physics are often called resonances.

6.11 Exercises
1. Using the WKB approximation, you can estimate the lifetime of a particle of mass m ini-
tially trapped in the “ground state” of a one-dimensional rectangular well using Eq. (6.10),

 ∞, x < 0
V (x) = 0, 0 < x < a
 α
x
, a<x

125
PHY 851/852 6 APPROXIMATION METHODS

Assume the barrier is sufficiently high that the wave function in the well can be approx-
imated as that of an infinite well and that the frequency of tunneling attempts can be
thought of as the rate at which a classical particle would impact the barrier at that energy.

(a) Estimate the energy E, by treating the potential as if its infinity at x = a.

(b) To what point, xf , does the particle need to tunnel to escape the well.
(c) Estimate the tunneling probability, e−2ϕ , where ϕ is calculated using the WKB ap-
proximation. I.e. find an expression for ϕ. Warning: The integral is not trivial.
(d) In terms of ϕ what is the lifetime?

2. A particle of mass m is initially in the ground state of a one-dimensional harmonic oscil-

lator of frequency ω centered at x = 0. Suddenly, at time t = 0, the center of the well is
moved to x = ℓ. Here, you will calculate the probability that the particle will be found in
the state |n⟩ of the new well, where |0⟩ is the new ground state.

(a) The ground state of the old well can be written as

|ϕ0 ⟩ = eℓ(d/dx) |0⟩,

where ϕ0 is the ground state of the old well and |n⟩ refers to eigenstates of the newly
positioned well. Writing the operator d/dx in terms of creation and destruction op-
erators of the new well, find an expression for |ϕ0 ⟩ as a linear combination of |n⟩.
HINT: You will probably wish to use the Baker-Campbell-Hausdorff relation.
(b) What is the probability of finding the particle in the state |n⟩.
(c) What is the expectation of the energy ⟨H⟩ after the well is shifted?
(d) If the well were shifted slowly instead of suddenly to its new position, what would be
the probability of finding the particle in the ground state of the new well?

3. Estimate the ground state energy of the hydrogen atom using a three-dimensional har-
monic oscillator ground state wave function as a trial function. Give your answer in terms
of the electron mass m and the coupling e2 .
4. Estimate the ground state energy of the three-dimensional harmonic oscillator using the
hydrogen atom wave function as the trial wave function. Give your answer in terms of the
mass m and the characteristic frequency of the harmonic oscillator ω.
5. Consider a particle in an infinitely deep square well of width a.

 ∞, x < −a/2
V0 (x) = 0, −a/2 < x < a/2
∞, x > a/2


A particle feels a perturbative potential,

V1 (x) = β sin(πx/a)

(a) What is the change in the ground state energy in lowest (non-zero) order perturbation
theory?
(b) What is the correction to the energy of the first excited state to the same order?

126
PHY 851/852 6 APPROXIMATION METHODS

(c) What is the correction to the wave function of the ground state to lowest non-zero
order?

6. Consider the Hamiltonian:

H0 = ασz
and the perturbation
V = βσx

(a) What is the correction to the ground state energy to second order in perturbation
theory?
(b) What is the correction to the excited state’s energy to the same order?
(c) Find the exact expression for the energy of the first state, and show that it gives the
same answer as part a when expanded in powers of β.

7. An electron of mass m initially in the ground state of a three-dimensional harmonic os-

cillator potential characterized by frequency ω, i.e. the ground state energy is 3ℏω/2, is
placed in a region with uniform electric field E.

(a) By finding corrections to the ground state wave function in first order perturbation
theory, find an expression for the electric dipole moment induced in the atom.
(b) An alternative method for calculating the dipole moment is to differentiate the energy
with respect to the electric field. Show that this method yields the same expression
found in (a) when one uses second order perturbation theory to find the correction to
the energy.

8. Two electrons whose positions are defined by r1 and r2 relative to the centers of their
⃗
confining potentials. The confining potentials are then separated by a distance R.

1
V0 (r1 , r2 ) = mω 2 (r12 + r22 ).
2
Positive charges +e are fixed at the centers of the potentials. The electromagnetic energy
between the two wells is:
e2 e2 e2 e2
V = + − − .
R ⃗ +⃗
|R r1 − ⃗
r2 | ⃗ +⃗
|R r1 | ⃗ −⃗
|R r2 |

Here, the repulsive interaction between the two positive ions is described by the first term,
and the repulsive interaction between the electrons is described by the second term. The
last two terms describe the attractive interaction between the electron and the ion in the
other well. The electromagnetic energy between each electron and its confining ion is as-
sumed to be part of the confining potential, and not part of the perturbation.
Assume that the separation R is much larger than either r1 or r2 . In terms of the separation
between the wells, R, the mass of the electrons m, the charge e and ω,

127
PHY 851/852 6 APPROXIMATION METHODS

(a) Show that for large R, the interaction may be approximated as a dipole-dipole inter-
action,
e2
V = 3 (x1 x2 + y1 y2 − 2z1 z2 ) ,
R
⃗
where the z axis is along the direction of R.
(b) Use second-order perturbation theory to find the electromagnetic attraction of the
two wells, V (R). This motivates the form for the London dispersion force, https:
//en.wikipedia.org/wiki/London_dispersion_force, which is the long-range
attractive force between neutral molecules.

9. Consider the function

1 η
g(ω) ≡ Im =
ω − iη ω2 + η2
where η is a positive real constant that approaches zero.

(a) What is g(ω = 0)?

(b) What is g(ω ̸= 0)?
(c) Using trigonometric substitutions, evaluate
Z ∞
dω g(ω).
−∞

(d) Write an expression for a delta function in terms of g(ω).

10. A bob particle is in the ground state of a 3-dimensional harmonic oscillator characterized
by a frequency ω,
1
V0 = mω 2 r 2
2
A perturbation is added that allows a bob particle to undergo a transformation into a mary
particle. The mary particle does not feel the effects of the oscillator potential. The bob and
mary particles have the same mass m. The perturbation is of the form,
Z
∗
Vbm = g⃗ϵs · d3 rψbob (r)∇ψmary (r),

where ϵs is the unit polarization vector of the mary particle with polarization s, which
refers to any of three directions.

(a) Find the total decay rate Γ (to any polarization). Hint: The answer will of the form

mg 2 b3 k3
Γ=A , (6.58)
ℏ3
where A is a dimensionless constant and b is the size of the harmonic oscillator wave
function. You need to find A.
(b) Find the differential decay rate, dΓs /d cos θ, for a polarization ⃗
ϵs = ẑ.

128
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

7 Perturbative Approaches to Scattering

7.1 Cross Sections

Scattering experiments provide much of the basis from which knowledge of atomic and sub-
atomic physics is extracted. This ranges from investigating the forces between the projectile and
target to understanding the elementary constituents and substructure of matter. Experiments
often involve aiming a beam of projectiles at a target with a given energy, then observing the
probability with which they scatter at given angles. Scattering cross sections, σif , can be related
to the rate for scattering of a single target particle. This rate can then be expressed in terms of
Fermi’s golden rule. For an incoming wave packet with an area of extent A and a length L, the
probability of colliding is

(N ) Nσ
Pscatt = , (7.1)
A
where there are N target particles. If a target has thickness Z, area A and a number density of
scatterers, nt ,

N = nt AZ, (7.2)
(N )
Pscatt = (nt Z)σ,
Pscatt
σ= .
nt Z
Thus, armed with knowledge of the target density and thickness, an experimentalist can deter-
mine the cross section for a single scatterer.
The choice of using the cross section to describe the result of the scattering is to use an observ-
able that is independent of the size and density of the target, and encapsulates the physics of
scattering off a single target particle.
The cross section can be related to a rate for the scattering by a single particle by considering
an incoming wave packet of longitudinal extent (along the direction of the incoming wave) L
and transverse size A. Relating σ to the rate is necessary because perturbative approaches, like
Fermi’s golden rule, provide expressions for the rate.The rate for scattering off a single scatterer
is
(1)
Pscatt
R= , (7.3)
L/vp
where the velocity of the projectile is vp and the time for the packet to pass is L/vp . Given that
the probability for scattering off a single scatterer is σ/A,
σ
R = vp , (7.4)
W
W = AL.

Here W is the volume of the wave packet. To calculate the cross section in perturbation theory
one can find an expression for the rate in terms of Fermi’s golden rule, then obtain the cross

129
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

section,
W
σ= R. (7.5)
vp

The dependence on the wave-packet volume, W , will disappear for the theoretical expressions
used to calculate the cross section, as the cross section should depend solely on the projectile
mass, m, the projectile velocity vp and the potential, V (r), experienced by the particle.
Theoretical descriptions of scattering mainly fall into two classes. The first is perturbative, the
simplest being the Born approximation. These are typically performed in a plane-wave basis,
with the matrix elements being expressed in the form V⃗k,⃗k′ . The second class of treatments are
built around a partial wave basis and are non-perturbative. Here, one considers incoming and
outgoing spherical waves, known as partial waves. Using angular momentum conservation,
one can divide the problem into a few one-dimensional problems, which can be solved non-
perturbatively, often by solving Schrödinger’s equation numerically. For lower energies, the
latter becomes more tractable because only a few partial waves contribute. The momentum
multiplied by the range of the potential sets the scale for what values of angular momentum are
relevant. Furthermore, at low energies, the potential energy is not much smaller than the kinetic
energy, which would invalidate the perturbative class of treatments, which are perturbative in
V . For the next several sections these notes focus on the perturbative picture, then switch to
partial wave treatments in the next chapter.
If one is interested in only those scatterings from the initial momentum state ⃗
ki to a specific final
⃗
state kf , the cross section becomes

σif = W
vp
R(⃗
ki →⃗
kf ). (7.6)

If one considers the probability per solid angle that scattered particles are emitted in a given
direction, int into a solid angle element of size dΩ, the expression becomes

dσ 1 dPscatt
= . (7.7)
dΩ nt Z dΩ
The differential solid angle is dΩ = sin θdθdϕ = d cos θdϕ, with θ and ϕ referencing the
direction of the final momentum. The solid angle is a measure of the angular extent in spherical
R
coordinates. For example, the area on a sphere is r 2 dΩ. The total angular coverage is dΩ =
4π, which is why many hermetic detectors are called 4π detectors. The quantity dσ/dΩ is
referred to as the differential cross section, and integrates to the total cross section,

dσ
Z
dΩ = σtot . (7.8)
dΩ
In essence, experiments measure dPscatt /dΩ, and given their knowledge of the thickness of the
target and the density of scatterers, infer the cross section. Rather than comparing scattering
rates, which depend on the size of the projectile’s wave packet, the target thickness and the
target’s number density, experiments report on differential or total cross sections.

130
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

7.2 The Born approximation

From Eq. (7.6) one can write the cross section for scattering into any state in terms of a rate.
Then, using Fermi’s golden rule to approximate the rate,
W X
σ= R(⃗
ki → ⃗
kf ) (7.9)
v ⃗
kf
2πW X ⃗
= |⟨kf |V|⃗
ki ⟩|2 δ(ϵf − ϵi ).
vℏ ⃗
kf

One may express the sum over states as an integral,

W W k2
X Z Z
3
= d k f = dΩ dE, (7.10)
⃗
(2π) 3 (2π) 3 ℏv
kf

and write the matrix element as,

⃗ ⃗
e−ikf ·⃗r eiki ·⃗r
Z
⟨⃗
kf |V|⃗
ki ⟩ = d3 r √ V(⃗
r) √ (7.11)
W W
to obtain an expression for the cross section where the volume has canceled,
2
1
Z Z
3
 3 i(⃗k − ⃗
k )·⃗
r

σ= d k f δ(ϵ f − ϵ i )  d r V(⃗ r )e f i  (7.12)
4π 2 ℏv  
Z 2
m
Z
i(⃗kf −⃗
3
 3 ki )·⃗r

= d p f δ(ϵ f − ϵ i )  d r V(⃗ r )e
4π 2 ℏ4 p  
 2
m
Z Z

 d3 r V(⃗ ⃗ ⃗
i(kf −ki )·⃗
r

= pmdϵ f dΩ δ(ϵ f − ϵ i ) r )e
4π 2 ℏ4 p  
m2 2
Z  Z 
i(⃗
kf −⃗

= dΩ  d3 r V(⃗ r )e ki )·⃗
r
4π 2 ℏ4   .

As promised, there is no dependence on the wave-packet volume W . This is known as the Born
approximation for the cross section, and is an example of first order perturbation theory, where
in this case one is calculating the rate from an initial to some final states ⃗
kf using Fermi’s golden
rule. To change this to the differential cross section, one can simply divide both sides by dΩ to
find the differential form of the Born approximation,
dσ m2
= 2 ℏ4
q )|2 ,
|Ṽ(⃗ (7.13)
dΩ 4π
Z
Ṽ(⃗
q) = d3 r V(⃗ r )ei⃗q·⃗r ,

⃗=⃗
q ki − ⃗
kf .

The differential cross section for scattering into the direction ⃗

kf is determined by the Fourier
transform of the potential, where the momentum transferred to the target, ⃗ ki − ⃗
kf , enters the

131
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

Fourier transform. For elastic scattering, where no energy is lost by the beam and |ki | = |kf |,
the momentum transfer is related to the scattering direction noted by θ and ϕ,

⃗
ki − ⃗
kf = |k| [(1 − cos θ)ẑ − sin θ cos ϕx̂ − sin θ sin ϕŷ] , (7.14)

where the initial beam was traveling in the +ẑ direction.

Example 7.1: Born Approximation for a Spherically Symmetric Gaussian Potential
Consider the potential
2 /(2a2 )
V(r) = V0 e−r .

a. As a function of the scattering angle θ find the differential cross section.

Solution:
First calculate the Fourier transform of the potential which depends only on the magni-
tude of ⃗ki − ⃗
kf .
Z
2 2
q ) = V0 d3 r e−r /(2a ) ei⃗q·⃗r
Ṽ(⃗
Z
2 2 2 2 2
= V0 d3 r e−(⃗r−i⃗qa ) /(2a ) e−q a /2
2 a2 /2
= V0 a3 (2π)3/2 e−q ,

where

q = |⃗
ki − ⃗
kf | (7.15)
q
= k (1 − cos θ)2 + sin2 θ
p
= k 2(1 − cos θ)
= 2k sin(θ/2).

Thus, the differential cross section is

dσ 2πm2 a6 V02 2 a2 sin2 (θ/2)

= e−4k .
dΩ ℏ4

b. Find the total cross section in this approximation.

Solution:
This involves integrating dσ/dΩ over dΩ = 2πd cos θ.

4π 2 m2 a6 V02
Z Z
2 a2 sin2 (θ/2)
σ= dσ = e−4k d cos θ,
ℏ4
8π 2 m2 a6 V02 1
Z
2 a2 sin2 (θ/2)
= d sin(θ/2) e4k sin(θ/2),
ℏ4 0

132
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

where a double angle formula, cos θ = 1 − 2 sin2 (θ/2), was applied. Integrating over
sin θ/2,

π 2 m2 a4 V02
σ= .
ℏ4 k2
The first surprise might be that the total cross section is not infinite even though the
potential does extend to infinity despite falling rapidly with r. This is because the tails
of the potential are not captured by the fluctuating phase, ei⃗q·⃗r , in the Fourier transform.
For higher k, the phase oscillates more throughout the integral, hence as k → ∞ the
total cross section tends to zero. Another surprise is that the cross section approaches
zero as k → 0. But, one must remember that the Born approximation is perturbative
and become increasingly invalid as the strength of the potential, V0 , becomes larger than
the kinetic energy.

Example 7.2: Born Approximation for the Coulomb Potential

The potential for a charge e scattering off a charge Ze is
Ze2
V(⃗
r) = ,
r
and the cross section will be the same whether the potential is attractive of repulsive because
the matrix element is squared.
Performing the Fourier transform,
Ze2
Z Z 1
2
Ṽ (⃗
q ) = 2π r dr dxeiqrx
−1 r
2 Z
4πZe
= dr sin qr
q
4πZe2
=− 2
cos(qr)|∞0
q
4πZe2
=
q2
πZe2
= 2 ,
k sin2 (θ/2)
The limit at r → ∞ can be realized by adding an exponential damping term to the potential
e−ηr , η → 0+ , and repeating the integral to see that one gets the same answer with the
evaluation at ∞ going to zero.
The expression for the differential cross section, is then,
dσ m2 Z 2 e 4
= .
dΩ 4(ℏk)4 sin4 (θ/2)
This answer is identical to the Rutherford cross section in classical mechanics. Note that this
cross section is ill-behaved at θ → 0. It is a non-integrable singulatrity and the total cross
section is infinite.

133
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

7.3 Structure Functions

Often scattering is performed within a material for the purpose of learning about the structure
of the material rather than understanding the potential due to a single scatterer. In that case the
Fourier transform of the net potential, i.e. the potential from a large number of scatterers, can be
written as,
XZ
Ṽ(⃗
q) = d3 rei⃗q·⃗r V(⃗
r −⃗
a) (7.16)
a
X Z
q ·⃗
= e i⃗ a
d3 rei⃗q·(⃗r−⃗a) V(⃗
r −⃗
a)
a
= ṽ(⃗
q )s(⃗
q ),

where
X
q) ≡
s(⃗ ei⃗q·⃗a , (7.17)
a

and ṽ(⃗q ) is the Fourier transform of the potential from a single scattering center placed at the
origin.
This allows us to write the differential cross section,

dσ 1 m2
= q )|2 |s(⃗
|ṽ(⃗ q )|2 , (7.18)
dΩ N 4π 2 ℏ4
where N is the number of scatterers. The factor 1/N enters because the cross sections are related
to the rate per scatterer. If the potential for a single scatterer is understood, one may therefore
determine s(⃗ q ).
The differential cross section is related to the probability that two scattering centers are separated
by a distance ⃗
a,
′
X
q )|2 =
|s(⃗ ei⃗q·(⃗a−⃗a ) (7.19)
a,a′
X
=N ei⃗q·δ⃗a
δ⃗
a

= N S̃(⃗q ),
2
dσ m
= q )|2 S̃(⃗
|ṽ(⃗ q ).
dΩ 4π ℏ
2 4

q ) is the Fourier transform of the structure function,

Here, S̃(⃗

dσ/dΩ
q) =
S̃(⃗ (7.20)
dσ/dΩ|isolated
X
= ei⃗q·δ⃗a .
δ⃗
a

134
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

This is the ratio of the measured cross section of target particles in matter to the cross section one
would measure if the target particles were measured in isolation, i.e. if there was no interference
contributions from different target particles. The sum over δ⃗ a include δ⃗ a = 0, i.e. the particle
with itself. For large momentum transfers the arguments of the phase factors become large and
the terms, ei⃗q·δ⃗a , become random. Only the δ⃗
a term survives and the S̃(⃗ q ) can be set to unity.
One can change the sum over relative positions to an integral where the probability of finding a
second scattering center at relative position δ⃗ a)d3 a,
a is S(δ⃗
X Z
q ·δ⃗
e i⃗ a
→ S(δ⃗a)d3 a ei⃗q·δ⃗a , (7.21)
δ⃗
a
Z
q) =
S̃(⃗ d3 r ei⃗q·⃗r S(⃗
r ).

The function S(⃗ r ) is the structure function in coordinate space. Even in a perfect crystal particles
a), is more realistic than
are oscillating around their lattice sites. Thus, a differential density, S(δ⃗
considering a sum over discrete positions, δ⃗ a.
For a cubic crystal, S(⃗ r ) has spikes when ⃗ r = nx ax̂ + ny aŷ + nz aẑ where a is the separation
between lattice sites. Then, S̃(⃗ q ) has spikes for values of q
⃗ which correspond to in-phase con-
tributions from different sites. For a perfect crystal, these spikes can be infinitely sharp because
for certain values of q ⃗ because the contributions can stay in phase as ⃗ r → ∞. This coherence
is known as Bragg’s law, https://en.wikipedia.org/wiki/Bragg%27s_law. Whereas, if the
material is a liquid, the spikes are greatly muted because there is no long range order. For a
liquid, S(⃗ r ) typically has a hole near ⃗ r = 0 and perhaps a few wiggles before being flat for
large ⃗r . The momentum transfer in nuclear or particle experiments is typically high compared
to the inverse inter-atomic distance between scatterers. Thus, one often neglects the structure of
the target in those experiments. Structure comes into play for X-ray scattering or for low-energy
neutron scattering because the characteristic momentum transfers are smaller. However, even
with a high energy beam, structure can come into play in the limit that the momentum transfer
is sufficiently small, i.e. the scattering is nearly forward.
The considerations used to derive Eq.s (7.19) included an implicit assumption, that the scattering
off one target particle at ⃗ a cannot be distinguished from the scattering off a different particle at
a′ . If the two target particles are in a lattice, and if the scattering is elastic, the lattice is unchanged
⃗
by the scattering and the assumption is justified. For higher momentum transfers most of the
scattering becomes inelastic, and measuring the structure function becomes more difficult.
Example 7.3: Two Scattering Centers
Consider two scattering centers, one placed at ⃗r = 0 and the second at ⃗ r = aẑ, where the
beam is in the z−direction. Assuming the two scatterers contribute identically, and that the
scatterings can be treated perturbatively, at what scattering angles does the differential cross
section disappear. Assume the incoming wave number is kẑ.

135
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

Solution: The overall scattering matrix element has the form

α(⃗q ) = ṽ(⃗
q )s̃(⃗q ),
X
q) =
s̃(⃗ ei⃗q·⃗a ,
a
⃗ = |k|((1 − cos θ)ẑ − sin θ cos ϕx̂ − sin θ sin ϕŷ).
q

Here, ṽ is the matrix element for a single scatterer. The cross sections will disappear for direc-
tions where s̃(⃗q ) = 0. This requires

1 + ei⃗q·aẑ = 0,
ka(1 − cos θ) = (2n + 1)π,

where n is a positive integer. This gives an angle for each n,

(2n + 1)π
cos θn = 1 − ,
ka
(2n + 1)π
 
−1
θn = cos 1− .
ka

For small ka there are no angles at which the cross section vanishes because the arguments
of the inverse cosine function must be between ±1. As ka increases, the number of solutions
increases.

7.4 Form Factors

Even if there is only one scattering center, it can be considered as having contributions from
many individual points within the single target. If the positions of the points are described by a
continuous distribution ρ(δ⃗a), one can set N = 1 above and replace the sums in Eq. (7.19) with
integrals,
Z
2 ′
|F (⃗
q )| = d⃗ad⃗a′ ρ(⃗ a′ )ei⃗q·(⃗a−⃗a ) .
a)ρ(⃗ (7.22)
R
Here, ρ is the density of scattering centers such that d⃗ a ρ(⃗ a) = 1. With this definition, |F (⃗
q=
2
0)| = 1. The differential cross section is again a product of the cross section one would have if
the scattering centers were point-like and the factor |F (⃗ q = 0)|2 = 1, which in the context of
scattering off individual object is known as a form factor, F (⃗ q = 0),

dσ dσ 
= q )|2 .
|f (⃗ (7.23)
dΩ dΩ point like


For Coulomb scattering of electrons, the point-like differential cross section is simply the Ruther-
ford form. Then, by measuring the true cross section one determines the form factor, s(⃗ q ). By
Fourier transforming, on then extracts information about ρ, which describes how charge is dis-
tributed about the scattering object.

136
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

Example 7.4: Form Factor for a Gaussian Distribution of Charge

Assume the charge density of a proton is distributed as a Gaussian,
e 2 /2a2
ρq (⃗
r) = e−r .
(2πa2 )3/2

what is the squared form factor?

Solution: For the form factor define ρ(⃗ r ) = ρq /e, because it represents the probability the
charge is found at ⃗ r , i.e., the point-like cross section already has the total charge accounted
for, just not its distribution. This gives

1
Z
2 2
F̃ (⃗
q) = d3 r 2 3/2
e−r /2a ei⃗q·⃗r
(2πa )
2 a2 /2
= e−q ,
−q 2 a2
q )|2 = e
|F (⃗ .

7.5 Higher Order Expansions

Calculating decays and scattering cross sections are the main instances where one invokes Fermi’s
golden rule. Fermi’s golden rule is based on a first-order perturbative correction to the evolution
matrix. Higher-order perturbative corrections can be included by replacing the potential V with
a more sophisticated operator.
Writing the transition matrix element to second order in time-dependent perturbation theory,
t
i
Z
′ ′
⟨n|UI (t, −∞)|i⟩|n̸=i = − Vni dt′ ei(ϵn −ϵi )t /ℏ+ηt (7.24)
ℏ −∞
 2 Z t Z t′
−i ′ ′ ′ ′′ ′′
+ Vnm Vmi dt dt′′ ei(ϵn −ϵm )t +ηt ei(ϵm −ϵi )t /ℏ+ηt
ℏ −∞ −∞
Z t
i ′
=− dt′ ei(ϵn −ϵi )/ℏ+ηt
ℏ −∞
Z t′ !
−i ηt′ ′ ′′ ′′ ′
· Vni + e Vnm Vmi dt′′ ei(ϵi −ϵm )(t −t )/ℏ+η(t −t ) .
ℏ −∞

By inspection, one can see that by making the substitution,

t′
i
Z
′′ −t′ )/ℏ−η(t′ −t′′ )
Vni → Vni − Vnm Vmi dt′′ ei(ϵi −ϵm )(t (7.25)
ℏ −∞

Vnm Vmi 
= Vni −  ,
ϵi − ϵm − iℏη η→0

one sees that the second order solution looks like the first order solution, but with replaced with
the more complicated form, which is known as the T-matrix, T . Sometimes, T refers to some

137
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

given order in perturbation theory, but in some contexts might refer to all orders. To all orders
the T becomes, extending the derivation for second order in Eq.s (7.24) and (7.25),
X Vnm Vni
Tni = Vni + (7.26)
m
(Em − Ei + iℏη)
X Vnm Vmm′ Vm′ i
+ + ···
mm′
(Em − Ei + iℏη)(Em′ − Ei + iℏη)

7.6 Propagators (Time-Ordered Evolution Operators)

Propagators, also known as a Green’s functions, are the building blocks of perturbative scatter-
ing approaches. A propagator is nothing more than the evolution operator with a step function
tacked on,
G(t) ≡ e−iHt/ℏ Θ(t), (7.27)
g(t) ≡ e−iH0 t/ℏ Θ(t).
The step function is added to account for the time-orderings involved in the expression for the
evolution operator in the interaction representation. Remembering that the evolution operator
in the interaction representation is
UI (t − t0 ) = eiH0 (t−t0 )/ℏ e−iH(t−t0 )/ℏ (7.28)
 Z t
−i
=1+ dt1 eiH0 t1 /ℏ V e−iH0 t1 /ℏ
ℏ t0
 2 Z t Z t1
−i −iH0 t1
+ dt1 eiH0 t1
Ve dt2 eiH0 t2 /ℏ V e−iH0 t2 /ℏ + · · · ,
ℏ 0 0

one can write the propagator as

G(t − t0 ) = e−iH0 (t−t0 )/ℏ UI (t)Θ(t − t0 ) (7.29)
= e−iH0 (t−t0 )/ℏ Θ(t − t0 )
 Z ∞
−i
+ dt1 e−iH0 (t−t1 )/ℏ Θ(t − t1 )V eiH0 (t1 −t0 )/ℏ Θ(t1 − t0 )
ℏ −∞
 2 Z ∞
−i
+ dt1 dt2 e−iH0 (t−t1 )/ℏ Θ(t − t1 )V eiH0 (t1 −t2 )/ℏ Θ(t1 − t2 )
ℏ −∞

· V eiH0 (t2 −t0 )/ℏ Θ(t2 − t0 )

+ ···
 Z ∞
−i
= g(t − t0 ) + dt1 g(t − t1 )V g(t − t0 )
ℏ −∞
 2 Z ∞
−i
+ dt1 dt2 g(t − t1 )V g(t1 − t2 )V g(t2 − t0 )
ℏ −∞
 3 Z ∞
−i
+ dt1 dt2 dt3 g(t − t1 )V g(t1 − t2 )V g(t2 − t3 )V g(t3 − t0 ) + · · ·
ℏ −∞

138
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

This expression has a nearly identical form to that for the evolution operator, the only benefit
being that the incorporation of the Θ functions allows one to eliminate the limits on the integrals
over time by absorbing the limits into the definitions of g. The real benefits to this formalism
comes when one considers the Fourier transform of G in frequency.

7.7 The Fourier Transform of the Propagator, G̃(ω)

Let us first consider the Fourier transform of g,
Z
g̃(ω) ≡ dt eiωt g(t), (7.30)
dω −iωt
Z
g(t) = e g̃(ω).
2π
Assuming the states |n⟩ and |m⟩ are eigenstates of H0 ,
Z
g̃mn (ω) = δnm dt ei(ω−iϵn /ℏ)t−ηt Θ(t)
iδnm
= , η → 0.
ω − ϵn /ℏ + iη
The infinitesimal η → 0+ represents an extremely slow exponential decay at large time differ-
ences. By using Cauchy’s theorem, one can integrate g̃(ω)e−iωt /(2π) over all ω to see that the
propagator g(t) is recovered. In fact, if one flipped the sign of η to 0− , the result would look the
same except with Θ(−t).
The pole in g̃(ω) tells us at what frequency the particle propagates. In this case one sees that the
pole (where the propagator blows up) is at ω = ϵn /ℏ − iη. If η were replaced with a finite value
Γ/2, the square of the propagator would behave as e−Γt/ℏ , and Γ/ℏ would be associated with
the exponential decay constant.
The advantage to using the operators frequency, rather than as integrals over time, is that the
perturbative expansion becomes simpler,
−i
Z  Z Z
G̃(ω) = dteiω(t−t0 )
g(t − t0 ) + dte iω(t−t0 )
dt′ g(t − t′ )V G(t′ − t0 ) (7.31)
ℏ
−i dtdt dω dω ′′
′ ′
 Z
′ ′ ′′ ′
= g̃(ω) + g̃(ω ′ )V G̃(ω ′′ )eiω(t−t0 )−iω (t−t )−iω (t −t0 )
ℏ (2π) 2

−i
 
= g̃(ω) + g̃(ω)V G̃(ω)
ℏ
X  −i n
= g̃(ω) + g̃(ω) (V g̃(ω))n .
n>0
ℏ

This is the Dyson series in frequency space, which unlike the one for time, does not involve
integrals. This simplification was made possible by incorporating the term Θ(t − t′ ) into the
definition of the Green’s function so that the limits of all the integrations over time in Eq. (7.29)
could run over all times. Although one typically chooses a basis so that g(t) and g̃(ω) are
diagonal, V is not usually diagonalized in that basis.

139
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

7.8 Propagating to the Same State

The propagator, Gmn (ω), carries information of how one evolves into the state n from the state
m. The diagonal part of that matrix, Gmm (ω), describes how one remains in, or returns to, the
same state, even if the path it involves leaving the state m during intermediate times. By itself,
Gmm (ω) provides information on how a particle’s energy is modified by the interaction, and
how it decays. These considerations form the basis for numerous phenomenological applications
of scattering and for numerous descriptions of how particle are modified in medium.
If V is diagonal in the same basis as H0 , one can forego the matrix notation and consider the
states one-at-a-time. One can then write the expression above for G̃ in a recursive form,

−i
G̃mm (ω) = g̃mm (ω) + g̃mm (ω)Vmm G̃mm (ω) (7.32)
ℏ
g̃mm (ω)
=
1 + (i/ℏ)g̃mm (ω)Vmm
i
= .
ω − ϵm /ℏ − Vmm /ℏ + iη

Thus, the only difference between the full propagator G̃ and the original propagator g̃ is that the
energy of the pole is shifted by V , exactly as one would expect from knowing the eigenvalues of
H = H0 + V .
One can not calculate G so easily when V has off-diagonal elements. However, by inspection of
the expansion of G̃ in the Dyson expansion, Eq. (7.31), one can separate out the contributions
from off-diagonal elements in expansion by defining the matrix T .

−i
X  
Tkm (ω) ≡ Vkm + Vki′ g̃i′ i′ (ω)Vi′ m (7.33)
′
i ̸=m
ℏ
 2
X −i
+ Vki′ g̃i′ i′ (ω)Vi′ j ′ g̃j ′ j ′ (ω)Vj ′ m + · · · .
′ ′
i ̸=m,j ̸=m
ℏ

Thus, T absorbs all the off-diagonal terms in the expansion of G,

−i
 
G̃mm (ω) = g̃mm (ω) + g̃mm (ω)Tmm (ω)g̃mm (ω) (7.34)
ℏ
 2
−i
+ g̃mm (ω)Tmm (ω)g̃mm (ω)Tmm (ω)g̃mm + · · ·
ℏ
i
= .
ω − ϵm /ℏ − Tmm /ℏ + iη

This propagator sums over all paths from some initial state m back into the same final state. All
the forays into states other than m are absorbed into T .
The first order correction to T is simply Tmm = Vmm . The second order correction in the

140
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

potential, separated into real and imaginary parts, is

X 1
Tmm (ω) = Vmi′ Vi′ m (7.35)
i′ ̸=m
ℏω − ϵi′ + iℏη
X  Vmi′ Vi′ m 
= + πiδ(ℏω − ϵi′ )Vmi′ Vi′ m .
i′ ̸=m
ℏω − ϵi′

Here we have used

1 P
= + πiδ(ω − E/ℏ) as η → 0, (7.36)
ω − E/ℏ + iη ω − E/ℏ

The symbol P simply points to the fact that the denominator is missing the iη, and that when
integrating across the pole one confines the integral to being away from the singularty by some
infinitesimal amount ±ϵ. Because the part of the integrand that explodes as 1/(ω −E/ℏ), which
switches sign as it passes the singularity, the singular part of the integral will not contribute as
ϵ → 0. One can now view the expression for G̃mm (ω) for ω = ϵm to see how the propagator is
affected to second order,
iℏ
G̃mm (ω) = , (7.37)
ℏω − ϵm − ∆ϵm + iℏΓm /2
X Vmi′ Vi′ m
∆ϵm = Vmm − ,
ϵ i ′ − ϵm
′
i ̸=m
2π X
Γm = Vmi′ Vi′ m δ(ϵm − ϵi′ ).
ℏ i′ ̸=m

Thus, a pole of the propagator is adjusted by the interaction in such a way that the real part
of the pole moves by an amount consistent with stationary state perturbation theory, while the
imaginary part is consistent with Fermi’s golden rule for calculating the rate at which one of the
states decays into the other channel.

7.9 Self-Energy of a Particle in Medium

If the propagator in Eq. (7.37) refers to momentum states, the propagation into the same mo-
mentum state k is
iℏ
G̃kk (ω) = , (7.38)
ℏω − ϵk − ∆ϵk + iℏΓm /2
X Vkk′ Vk′ k
∆ϵk = Vkk − ,
ϵ k ′ − ϵk
′
k ̸=k
2π X
Γk = Vkk′ Vk′ k δ(ϵk − ϵk′ ).
ℏ k′ ̸=k

141
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

The matrix element Vkk has the form,

1
Z
⃗ ⃗
Vkk = ⟨k|V |k⟩ = r )e−ik·⃗r
d3 r eik·⃗r V(⃗ (7.39)
W
1
Z
= d3 r V(⃗
r ).
W
Here, the volume of the plane wave, W , usually goes to infinity. This kills the contributions
to both ∆ϵk and Γk , and the position of the propagators pole does not change in a scattering
experiment. In fact, if one calculated higher-order corrections to ∆ϵ, they would all scale as
1/W .
However, if the particle is traversing a medium, where the density of such scatters in ns , the vol-
ume per scatterer becomes 1/ns , and the factor 1/W becomes ns . In this case the propagator’s
pole, essentially the energy of the particle, is altered a finite amount,
Z
∆ϵk = ns d3 rV(⃗ r ), (7.40)

to first order in perturbation theory. After changing the sum over intermediate states k′ to an
integral,
W
X Z
= 3
k′2 dkdΩ′ , (7.41)
k ′ (2π)

the decay rate in Eq. (7.38), Γk , becomes

2
2π
Z Z
′2 ′ ′
 3 i(⃗k−⃗k ′ )·⃗
r

Γk = ns k dk dΩ δ(ϵ k − ϵ k ′)  d r V(⃗
r )e  (7.42)
ℏ(2π)3  
Z 2
mk
Z
⃗ ⃗′
dΩ′  d3 r V(⃗
 
= ns 2 3 r )ei(k−k )·⃗r  .
4π ℏ
Using Eq. (7.13), and replacing v = ℏk/m, this becomes

Γk = ns σv, (7.43)

which is the scattering rate in a medium of density ns .

The correction to the particle’s energy, ∆ϵk , is known as the real part of the particles’ self energy,
and ℏΓk /2 is the imaginary part. The change in the pole of the propagator due to the interaction
represents the change in the wave function of a particle of momentum p ⃗,
(0) (p)t/ℏ+i⃗ (0) (p)+∆ϵ(p))t/ℏ+i⃗
e−iϵ p·⃗
r /ℏ
→ e−i(ϵ p·⃗
r −Γ(p)t/2
. (7.44)

Squaring the wave function results in e−Γ(p)t , illustrating the decaying nature of the wave in
the medium. In addition to the energy changing the velocity also changes in the medium. The
velocity of a wave is given by the expression,

d(ϵ(0) (p) + ∆ϵ(p)

vi = . (7.45)
dpi

142
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

For example, if photons of energy ϵ are traversing a medium where they might excite molecules
into some excited state E ∗ , the correction ∆ϵ might be positive or negative depending on whether
the excited state is above or below ϵ. This switch of sign results from the denominator of the ex-
pression for ∆ϵ changing sign if the energy of the intermediate state passes ϵ. The group velocity
follows suit, and can be significantly altered for ϵ ≈ E ∗ . Of course, when ϵ ≈ E ∗ , the cross
section, and thus the decay rate, is also large. This won’t be pursued further in this course, but it
is clear that the self energy and its momentum dependence play a critical role in understanding
how a particle’s properties are altered in medium.

7.10 Resonant Scattering

Here we consider the problem where an initial state k scatters to a final state k′ through a reso-
nant channel R. For simplicity, we consider the matrix element of the momentum state k with
the resonant state R to be
α(k)
⟨R|V |k⟩ = √ , (7.46)
W
We will assume that one has performed√ all the necessary integrations to find the matrix element,
and that we call it α aside from the 1/ W due to the normalization of the plane wave in the
volume W . To simplify our discussion we will assume that α has no dependence with respect
to the direction of k.
One could calculate the rate of decay of the resonance using Fermi’s golden rule and obtain

2π |α|2
ΓR = ρk (ϵR ), (7.47)
ℏ W
k2
= |α|2
πvk ℏ2
where ρk (ϵR ) is the density of states of the outgoing particle.
An example of resonant scattering could be a photon scattering off an atom. Here k labels the
momentum of the photon, while R would refer to a specific excited state of the atom that could
be attained due to the interaction with the photon. To simplify our derivation, we will assume
that α is independent of k.
It was shown in the previous lecture that the cross section for scattering could be written to
second order as
2πV X
σ= |Tk′ k |2 δ(ϵk′ − ϵk ) (7.48)
vℏ k′

−i X i 
Tk′ k = Vk′ k + Vk′ m Vmk  .

ℏ ω − ϵm /ℏ + iη 
m ℏω=ϵk

Using the derivations of the previous section, one could have inserted the full propagator into

143
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

the intermediate state, which would then yield


−i X 
Tk′ k = Vk′ k + Vk′ m G̃mm′ (ω)Vm′ k  . (7.49)

ℏ ′

mm ℏω=ϵk

By resonant scattering we mean that the only interaction is between the resonant state R and the
momentum states. The matrix element then becomes
−i 
Tk ′ k = Vk′ R G̃RR (ω)VRk  , (7.50)

ℏ ℏω=ϵk
i
G̃RR (ω) = ,
ω − ϵR /ℏ + iΓR /2

where Γ is the decay rate of the resonance.

The cross section is no longer infinite when ϵk = ϵR due to the factor ΓR appearing in the
denominator. Plugging the expression for T into the expression for the cross section one obtains

2πW X |α|4 1
σ= δ(ϵk′ − ϵk ) (7.51)
vℏ k′
W 2 (ϵk − ϵR )2 + ℏ2 Γ2R /4
2πW |α|4 1
= ρk (E)
vℏ W 2 (ϵk − ϵR )2 + ℏ2 Γ2R /4
k2 4
1
= |α|
πv 2 ℏ2 (ϵk − ϵR )2 + ℏ2 Γ2R /4
4π (ℏΓR /2)2
= 2 .
k (ϵk − ϵR )2 + (ℏΓR /2)2

The last line is known as the Breit-Wigner form for scattering through a resonance. Note that the
cross section is determined by two numbers, the width of the resonance and the Energy of the
resonance.
If the resonance has spin SR , then the effect is multiplied by the number of degenerate states
through which one might scatter. If the incoming particles have spins S1 and S2 , the effect is
correspondingly reduced due to the fact that many of the states of the resonance would not be
reached with particular combinations of S1 and S2 .
The Breit-Wigner form for resonant scattering is then

(2SR + 1) 4π (ℏΓR /2)2

σ= . (7.52)
(2S1 + 1)(2S2 + 1) k2 (ϵk − ϵr )2 + (ℏΓR /2)2

Example 7.5: Resonant Scattering through the ρ Meson

Here, we consider π − π resonant scattering through the ρ meson. Consider a π+ and a π0
which are scalar mesons with masses 140 MeV/c2 and 135 MeV/c2 . They scatter through the
ρ+ resonance which has spin 1, a mass of 770 MeV/c2 and a width ℏΓ = 151 MeV. Find the
cross section at resonance.

144
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

Figure 7.1: An example of resonant scattering of p+13 C. Each peak corresponds to a state in 14 O. Figure
from Y. Wang et al., “Study of Elastic Resonance Scattering at Ciae” (2011).

Solution: First, one must find the relative momentum p

⃗ in the center of mass frame. In this
⃗ and −⃗
frame the particles move with p p.
q q
2
mρ = m+ + p + m20 + p2 ,
2

2
m4ρ + m4+ + m40 − 2m2ρ m2+ − 2m2ρ m20 − 2m2− m2+
p = .
4m2ρ

Here, we have neglected the factors of c. The value of p at resonance is then 360 MeV/c.
The cross section at resonance is then
4π 12π
σR = (2S + 1) 2
= = 11.3 fm2 = 113 mb.
kR (360/197.326)2

Here, ℏc = 197.326MeV·fm and one fm2 equals 10 millibarns. Note that the maximum cross
section depended only on the relative momentum at resonance and not on the width.

7.11 Resonant Scattering, Simplified

Repeating Eq. (7.25), one can extend Fermi’s golden rule to second order via the substitution,
Z t′
i ′′ ′ ′ ′′
Vf i → Vf i − Vf R VRi dt′′ ei(ϵi −ϵm )(t −t )/ℏ−η(t −t ) (7.53)
ℏ −∞

Vf R VRi 
= Vf i −  .
ϵi − ϵR − iℏη η→0

145
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

The second term represents the part of the evolution that began in state i at t = −∞, then
switched to state R at time t′′ , then had the state R decay back to a state f at time t′ , which
due to the delta function in Fermi’s golden rule would have the same energy as the state i. One
might modify the exponential term in the integral to account for the decay,
′′ −t′ )/ℏ−η(t′ −t′′ ) ′′ −t′ )/ℏ−Γ(t′ −t′′ )/2
ei(ϵi −ϵm )(t → ei(ϵi −ϵm )(t , (7.54)
with the decay rate being Γ. The fact that Γ/2 is being used is motivated by the fact that squaring
the amplitude would give the factor e−Γt . This ansatz is intuitive, but nonetheless is an assump-
tion and is therefore less rigorous than then derivation of the previous section. Eq. (7.53) now
becomes,
Vf R VRi
Vf i → Vf i − . (7.55)
ϵi − ϵR − iℏΓ/2
If the magnitude of Vf R is independent of the direction of the momentum for all states with the
same energy, one can replace Vf R and VRi with |V |. One can also express the decay rate in terms
of V using Fermi’s golden rule for the decay„
2π
|V |2 ρ(ϵi ),
Γ= (7.56)
ℏ
where the delta function in Fermi’s golden rule has been replaced by the density of states,
W
Z
ρ(ϵi ) = 4πk2 dkδ(ϵk − ϵi ) (7.57)
(2π)3
mk
=W ,
2π 2 ℏ2
where W is the large volume and k is chosen to satisfy the delta function.
Substituting for V and neglecting the first term for Vf i ,
ℏΓ 1
Vf i → (7.58)
2πρ(ϵf ) (ϵi − ϵR − iℏΓ/2)
1 ℏΓ/2
= .
πρ(ϵf ) (ϵi − ϵR − iℏΓ/2)
Repeating Eq. (7.9), the cross section expressed in Fermi’s golden rule,
2πW
σ= |V |2 ρ(ϵf ). (7.59)
vf ℏ
Here, |V | will be replaced with the expression in Eq. (7.55). One can now express the cross
section in terms of Γ,
2W (ℏΓ/2)2
σ= (7.60)
vf πℏρ(ϵf ) (ϵi − ϵR )2 + (ℏΓ/2)2
4π (ℏΓ/2)2
= 2 .
k (ϵi − ϵR )2 + (ℏΓ/2)2
The velocity of the outgoing particle was replaced with mvf = ℏk. This is the same result for
the Breit-Wigner cross section in Eq. (7.52) derived more rigorously before.

146
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

Example 7.6: Narrow Resonance

Imagine that a proton could decay to a positron and a photon with a long lifetime τ = 1035
years. Rather than waiting for a decay, you build an accelerator that collides photons with
electrons with the required center-of-mass energy. To simplify the calculation, you can make
the approximation that the electron is massless its energy in the center-of-mass frame, ∼ 470
MeV, is much larger than the electron mass.
a) What is the cross section at resonance?
b) If one can tune the beam so that the center-of-mass energy is mp ± 1 eV, what is the
averaged cross section over the energy range?
Solution:
a) The Breit-Wigner cross section, using the degeneracy factors for the proton, photon and
electron, is
1 4π
σR = 2
2 kR
ℏkR c = 470 MeV,
470
kR = = 2.38 fm−1 ,
197.326MeV fm
σR = 1.11 fm = 11.1 mb.
Collider experiments readily measure cross sections many thousands of times smaller than
this.
b) Now, consider the energy-averaged cross section
Z mp +∆E
1
σ̄ = dE σ(E).
2∆E mp −∆E
The width, ∆E = 1 eV, is much larger than the width, ℏΓ, of the resonance, so the limits of
the integral can be taken to infinity. Further, ∆E is small enough that the prefactor 4π/k2 can
2
be set to 4πkR . Thus,
Z ∞
σR (ℏΓ/2)2
σ̄ = dE
2∆E −∞ (E − mp )2 + (ℏΓ/2)2
σR ℏΓ
Z ∞
1
= dx 2
4∆E −∞ x +1
πℏΓ
= σR .
4∆E
Using ℏ = 6.58 × 10−16 eVs, and Γ = 1035 years =3.1542 s,
ℏ
ℏΓ = = 3.17 × 10−59 eV.
τ
This gives
σ̄ = 2.8 × 10−58 mb.
This is FAR too small to be measured.

147
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

7.12 Exercises
1. Using the Born approximation estimate the differential scattering cross section, dσ/dΩ for
particles of mass m scattering off the following potentials.
(a) V(⃗r ) = V0 Θ(a − r).
(b) V(⃗r ) = a3 V0 δ 3 (⃗
r ).
3 3
(c) V(⃗r ) = a V0 [δ (⃗ r − aẑ) + δ 3 (⃗
r + aẑ)].
3 3 3
(d) V(⃗r ) = a V0 [δ (⃗ r − aẑ) − δ (⃗ r + aẑ)].
3 3 3
(e) V(⃗r ) = a V0 [δ (⃗ r − ax̂) − δ (⃗ r + ax̂)].
−r/a
(f) V0 e /r.
You can express your answer either in terms of the momentum transfer q
⃗, or in terms of
the beam momentum k and the scattering angle θ, ϕ.
2. By taking two derivatives of the form factor at q = 0,

∂ ∂ 
F (⃗q ) ,
∂qi ∂qj q=0

one can generate moments of the charge distribution,

Z
⟨(ri − r̄i )(rj − r̄j )⟩ ≡ d3 r ρ(⃗
r )(ri − r̄i )(rj − r̄j ).

(a) Working in a coordinate system where r̄i = 0, prove the relation above.
(b) Test your answer by comparing to the result of Example 7.4. First calculate ⟨ri rj ⟩
by integrating to get a weighted average of ri rj using ρq (⃗
r ), then compare to the
expression above using derivatives of the form factor.
3. The cross section for scattering a particle with momentum ℏk off a single target is
dσ
= α, (7.61)
dΩ
which is independent of the scattering angle. Here, we assume that the cross section is
small. Now, two targets are placed a distance a apart, separated along the z axis (the
same axis along which the incident beam is directed). At what scattering angles does the
differential cross section, dσ/dΩ, equal zero? (This is worked out as an example in the
notes)
4. The cross section for scattering a particle with momentum ℏk off a single target is
dσ
= α, (7.62)
dΩ
which is independent of the scattering angle. Here, we assume that the cross section is
small. Now, two targets are placed a distance a apart, separated along the x axis (perpen-
dicular to the axis along which the incident beam is directed). At what scattering angles
does the differential cross section, dσ/dΩ, equal zero? Specify both the polar angle θ and
the azimuthal angle ϕ.

148
PHY 851/852 7 PERTURBATIVE APPROACHES TO SCATTERING

5. Consider a charge Z that is uniformly distributed within a sphere of radius R. A

(a) Find the squared form factor F (⃗

q ).
(b) For an incoming wave with wave number kẑ what is F as a function of the scattering
angle θ? Express F in terms of kR and θ.
(c) Plot |F |2 from (b) for the cases where kR = 1, 2 and 5 as a function of θ.

6. A π + , which is a spin-zero meson, scatters off a proton through a ∆++ resonance(which

is comprised of three up quarks). The ∆++ is spin 3/2 baryon. The masses of the pion,
proton and delta are 139.58 MeV/c2 , 938.28 MeV/c2 and 1232 MeV/c2 respectively. The
width of the ∆ is 120 MeV.
p
(a) Using relativistic dispersion relations, E = p2 c2 + m2 , what is the relative mo-
mentum, q, of the pion and proton at resonance? This is one half the difference of the
two momenta in the center-of-mass frame. I.e. ϵπ (pR ) + ϵp (pR ) = M∆ .
(b) Estimate the cross section at resonance? Give your answer in millibars. One mb equals
10−24 cm and 10 mb= 1 fm2 .

7. Consider a particle of mass m that could be confined to a spherical well,


 0, r<a
V(r) = V0 , a < r < 2a
0, r > 2a


(a) Use the WKB method to estimate the decay rate of a particle of mass m escaping from
a spherical trap defined by the potential. Assume the barrier is sufficiently high to
approximate the energy of the trapped particle with an infinite well.
(b) Find an expression to estimate the cross section for a particle scattering off the poten-
tial well with an energy near the ground state energy described above. You can give
your answer as a function of the incident energy, E, m, V0 , a, and the width Γ.

8. Consider the function

i
f˜(ω) = .
ω − E/ℏ + iΓ/2

Show that the Fourier transform is

1
Z
f (t) = dω e−iωt f˜(ω)
2π
= e−iEt/ℏ−Γt/2 Θ(t).

HINT: Use contour integration.

149
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

8 Scattering at Lower Energies

Perturbative approaches do not work when the potential, which is treated as a perturbation,
is larger than the kinetic term. This then invalidates the approaches of the previous chapter.
Fortunately, when the kinetic energy is small, the relevant angular momenta are also small. If
the range of the potential is a, the angular momenta that contribute to the scattering are in the
range, L ≲ pa, where p is the momentum of the beam particles. Many experiments work in
the regime where only a few units of angular momenta need to be considered, and in fact, a
large amount of information about potentials is extracted from experiments where one need
only consider ℓ = 0.
Of course, the incoming and outgoing measurements of scattering experiments are plane waves,
eigenstates of momentum, not angular momentum. However, plane waves can be decomposed
into spherical waves (known as partial waves), i.e. waves that are eigenstates of angular mo-
mentum, with quantum numbers ℓ = 0, 1, 2 · · · . In this chapter, we will treat scattering by
considering a few partial waves. The scattering wave is then considered as the original plane
wave, plus the alterations of those partial waves with low ℓ. The alterations to the partial waves
due to the potential then provide one with expressions for the differential cross section. If the
potential is spherically symmetric, one can consider partial waves of a given ℓ, individually. This
effectively reduces the three-dimensional scattering problem to a few one-dimensional problems
with fixed ℓ.

8.1 Partial Waves and Phase Shifts

If potentials have spherical symmetry, angular momentum is conserved and one can solve the
scattering problem through the consideration of spherical waves rather than plane waves. This
effectively reduces the problem to the solution of one-dimensional Schrödinger equations, where
each partial wave is characterized by a specific angular momentum ℓ and projection m. Each
partial requires a different centrifugal potential as the angular momentum barrier depends on ℓ.
Because the energies are low, the relative momenta and the angular momenta are small. For low
energy scattering only the lowest few ℓ need to be considered, and when the scattering energy
approaches zero, scattering is dominated by the ℓ = 0 contributions.
First, we consider the kinetic part of the Hamiltonian. The Schrödinger equation can be written
in either Cartesian coordinates or spherical coordinates, and the kinetic term is

ℏ2 ℏ2 1 ∂ 2
 2
∂2 ∂2 1 ∂2 1
  
∂ ∂ ∂
− + + =− r+ 2 + 2 sin θ
2m ∂x2 ∂y 2 ∂z 2 2m r ∂r 2 r sin2 θ ∂ϕ2 r sin θ ∂θ ∂θ
(8.1)
ℏ 1 ∂
2 2
L 2
=− r + .
2m r ∂r 2 2mr 2
If angular momentum is conserved, L2 is a constant and can then be replaced with its eigenvalue
ℏ2 ℓ(ℓ + 1). Then ℏ2 ℓ(ℓ + 1)/2mr 2 appears like a repulsive potential energy in Schrödinger’s
equation, and is thus referred to as a centrifugal potential. If the potential is spherically symmet-
ric, angular momentum is a good quantum number and one can write solutions corresponding

150
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

to a specific ℓ and m.
Hψℓ,m (⃗
r ) = Eψℓ,m (⃗
r ), (8.2)
with the wave function being written as a product of a radial part and an angular part,
ψℓ,m (⃗
r ) = Yℓ,m (θ, ϕ)Rℓ (r). (8.3)
The radial wave function is a solution of the equation
ℏ2 1 ∂ 2 ℏ2 ℓ(ℓ + 1)
− 2
rRℓ (r) + Rℓ (r) + V (r)Rℓ (r) = ERℓ (r). (8.4)
2m r ∂r 2mr 2
Remember that the radial wave function depends only on ℓ and not m because the centrifugal
potential is determined by ℓ only.
If the potential is zero, the solutions Rℓ (r) are spherical Bessel functions jℓ (kr), first presented
in Chapter 4. Plane wave solutions can be expanded in terms of the spherical solutions through,
⃗
X
eik·⃗r = (2ℓ + 1)iℓ jℓ (kr)Pℓ (cos θ), (8.5)
ℓ

where cos θ ≡ k̂ · r̂. This is known as the partial wave expansion. Note that the expansion is in
terms of the Legendre polynomials,
s
4π
Pℓ (cos θ) = Yℓ,m=0 (θ, ϕ). (8.6)
2ℓ + 1
⃗
All angular functions can be expressed in terms of Yℓ,m s, and because eik·⃗r = eikr cos θ does not
depend on ϕ, it is not surprising that the expansion contains only Yℓ,m=0 terms. The partial wave
expansion can be derived by combining the orthogonality relations of the Legendre polynomials
and spherical Bessel functions with the Rodriquez formula,
1 dℓ (x2 − 1)ℓ
Pℓ (x) = . (8.7)
2ℓ ℓ! dxℓ
The solutions jℓ are normalized so that for large x = kr they behave as,
(−i)ℓ eix − (iℓ )e−ix
jℓ (x)|x→∞ = . (8.8)
2ix
They are a linear combination of outgoing and incoming waves, each of which is a solution to the
Schrödinger equation. The relative phase between the incoming and outging waves is chosen so
that the solution goes to zero at x = 0. In fact,
xℓ


jℓ (x) ≈  . (8.9)
(2ℓ + 1)!! x≈0
One can verify the small x expansion by applying Schrödinger’s equation with the centrifugal
potential. Using the definition, uℓ (x) ≡ xRℓ (x), we assume that uℓ is an expansion in powers
of x, with the lowest power being n. Schródingers’s equation becomes
ℓ(ℓ + 1)
−∂x2 uℓ (x) + uℓ (x) = uℓ (x), (8.10)
x2
uℓ (x) = xn + axn+1 + bxn+2 + · · ·

151
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

In order for the expansion to not include lower powers in m one must have

n(n − 1) ℓ(ℓ + 1)
− + = 0, (8.11)
x2 x2
where the factor n(n − 1) comes from taking two derivatives of xn . This give n = ℓ + 1 or
n = −ℓ. Because Rℓ ∼ uℓ /r, the radial wave functions behave as r ℓ or r −ℓ−1 .
The second set of solutions, those where uℓ behaves as x−ℓ , do not satisfy the boundary condi-
tions at zero as they begin with x−ℓ−1 . These are known as spherical Neumann functions and
have the opposite relative phase between the incoming and outgoing parts, while being quite
divergent at the origin.
(2ℓ − 1)!!
nℓ (x) ≈ . (8.12)
xℓ+1
Examples of a few spherical functions for low ℓ, repeated from Chapter 4, are

sin x cos x
j0 (x) = , n0 (x) = − (8.13)
x x
sin x cos x sin x cos x
j1 (x) = 2
− , n1 (x) = − 2 −
x x x x
3 1 3 3 1 3
 
j2 (x) = − sin x − cos x, n2 (x) = − − cos x − sin x
x3 x x2 x3 x x2

Both jℓ and nℓ are real.

By making combinations, jℓ ± inℓ , one finds solutions that correspond to incoming or outgoing
waves. The spherical Hankel functions are defined in terms of jℓ and nℓ ,

(−i)ℓ+1
hℓ (x) ≡ jℓ (x) + inℓ (x) ≈ eix |x→∞ (8.14)
x
iℓ+1
h∗ℓ (x) ≡ jℓ (x) − inℓ (x) ≈ e−ix |x→∞
x
Here hℓ and h∗ℓ behave as outgoing and incoming waves respectively.
When adding a potential of finite range, there still exists solutions which look like hℓ (kr) or
h∗ℓ (kr) for r beyond the range of the potential, but have modified forms at small r. Just as in the
case with no potential, one can find a linear combination of the incoming and outgoing solutions
which goes to zero at r = 0. However, the relative phase between the incoming and outgoing
phase will be adjusted by a phase e2iδℓ due to the existence of the potential. The modification
for r outside the range of the potential can only be a phase factor because the incoming flux and
outgoing flux must be the same strength. The large x = kr behavior is then

1
e2iδℓ hℓ (x) + h∗ℓ (x) ,


Rℓ (x)|x>ka = (8.15)
2
where δ is known as the phase shift. Here a is any distance large enough such that the potential
is zero (keep in mind that this is never true for the Coulomb potential). We define the overall

152
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

phase of Rℓ so that the incoming phase has the same phase as does the incoming part of jℓ . For
the s−wave, this becomes
eiδ sin(kr + δ)
Rℓ=0 (k, r > a) = . (8.16)
kr
If one scatters a plane wave off a potential, one can consider the solution to be the original plane
wave, expanded in terms of partial waves, plus the correction due to the interaction,
X
ψ⃗k (⃗
r) = (2ℓ + 1)iℓ Rℓ (k, r)Pℓ (cos θ) (8.17)
ℓ
i⃗
X
k·⃗
r
=e + (2ℓ + 1)iℓ (Rℓ (k, r) − jℓ (kr)) Pℓ (cos θ),
ℓ

where the choice of phases in the definition of Rℓ allows the incoming waves to be identical to
those of the solution with no potential. Expanding the answer at large r, one obtains

⃗
X
eikr
r )|r→∞ = eik·⃗r +
ψ⃗k (⃗ (2ℓ + 1) e2iδℓ − 1 Pℓ (cos θ) (8.18)
ℓ
2ikr
⃗
X eikr
= eik·⃗r + (2ℓ + 1)eiδℓ sin δℓ Pℓ (cos θ)
ℓ
kr

Only the latter term contributes to scattering as the plane wave continues to travel forward after
the wave packet leaves the region of the scatterer. One defines a quantity f (Ω) as the scattering
amplitude,
X 1
f (Ω) ≡ (2ℓ + 1)eiδℓ sin δℓ Pℓ (cos θ) (8.19)
ℓ
k

i⃗
k·⃗
r
eikr
ψ⃗k (⃗
r )|R→∞ = e + f (Ω).
r
The scattering amplitude, f (Ω) is a function of the scattering angle θ and has dimensions of
length. If there were no spherical symmetry, f (Ω) could also depend on the azimuthal angle ϕ.
Note that f no longer depends on r. It can also be related to the differential cross section. To see
this, we first relate the differential cross section to the flux of particles per solid angle,
v dσ dN
= . (8.20)
V dΩ dΩdt
The flux per unit area can be found by multiplying the square of the wave function in Eq. (8.19)
by the velocity and dividing by the volume,
dN v |f (Ω)|2
= , (8.21)
r 2 dΩdt V r2
Comparing the two equations above allows one to see that f (Ω) is directly related to the differ-
ential cross section.
dσ
= |f (Ω)|2 . (8.22)
dΩ
153
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

Combined with Eq. (8.19), one can see that the differential cross section at a given energy is
determined solely by the phase shifts δℓ (k).
Squaring the scattering amplitude in Eq. (8.19) includes cross terms with different values of ℓ,
but the contribution from the cross terms disappears after integrating over dΩ to obtain the total
cross section. The cancellation is due to the orthogonality of the Legendre Polynomials. The
total cross section is then
4π X
σ= 2 (2ℓ + 1) sin2 δℓ . (8.23)
k ℓ
For resonances of a given ℓ the peak of the resonance is for where sin2 (δℓ (k)) is near unity. The
value of σ at that point, 4π/k2 , then matches the same value for the Breit-Wigner resonance in
Eq. (7.52).
Example 8.1: Hard Sphere Scattering
Consider a hard sphere of radius a. Find the contribution to the total cross section from s and
p wave scattering as a function of the momentum.
Solution:
The ℓ = 0 case is simple as the solutions for ingoing and outgoing waves in the region r > a
are the Hankel functions which must go to zero at r = a.
1
e2iδ0 h0 (ka) + h∗0 (ka) = 0


Rℓ=0 (a) =
2
Plugging in the expressions for h0 , one obtains,
0 = e2iδ0 −ieika + ie−ika

= eiδ0 −iei(ka+δ0 ) + iei(ka−δ0 )

= 2eiδ0 sin(ka + δ0 ).
which gives the ℓ = 0 phase shifts,
δ0 = −ka.
The contribution to the cross section from the ℓ = 0 partial waves is then,
4π
σ0 = sin2 (ka).
k2
As k → 0, the cross section approaches 4πa2 , four times the expected geometric cross section.
Calculating the contribution for the ℓ = 1 partial waves is a bit more difficult. In this case the
incoming and outgoing waves are
1
e2iδ1 h1 (ka) + h∗1 (ka) = 0.


Rℓ=1 (a) =
2
Using Eq.s(8.13) and (8.14),
eix eix
h1 (x) = − −i 2 (8.24)
 x x
cos(x) sin(x) sin(x) cos(x)
  
= − + +i − − ,
x x2 x x2

154
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

which yields the following expression for δ1 when requiring that Rℓ=1 (a) = 0,

e2iδ1 h1 (ka) + h∗1 (ka) = 0,

h∗1 (ka)
e2iδ1 = −
h1 (ka)
i [cos(ka)/ka − sin(ka)/(ka)2 ] + [sin(ka)/ka + cos(ka)/(ka)2 ]
= .
−i [cos(ka)/ka − sin(ka)/(ka)2 ] + [sin(ka)/ka + cos(ka)/(ka)2 ]

This can be solved for δ1 ,

cos(ka) − sin(ka)/(ka)
tan δ1 = .
sin(ka) + cos(ka)/(ka)

The contribution to the total cross section is

12π
σ1 = sin2 δ1 .
k2
At low momentum, this becomes

[−(ka)2 /2 + (ka)2 /6]

δ1 ≈ (8.25)
1/ka
(ka)3
=− ,
3
4π
σ1 ≈ k4 a6 .
3
In general, for potential scattering, phase shifts tend to behave as k2ℓ+1 as k → 0, which
means that cross sections behave as k4ℓ . Thus, whereas the cross section for s-wave scattering
approaches a constant as k → 0, the p-wave contribution rises very slowly.

8.2 The Optical Theorem

Looking at the expansion of the wave function for large r,

⃗ eikr
r )|R→∞ = eik·⃗r +
ψ⃗k (⃗ f (Ω), (8.26)
r
one might ask how the scattered flux, v|f (Ω)|2 , is balanced by a loss of flux in the plane wave.
When one squares ψ⃗k , three terms arise. The first is the squared plane wave, which is unity,
regardless of whether scattering occurred. The second part is the flux of the scattered wave,
which behaves as |f (Ω)|2 . The third possibility term is the cross term between the plane wave
and the scattered wave, which will be linear in f (Ω). It is this term which must somehow
represent the loss of forward going flux.
The scattering amplitude, f (Ω), defined in Eq. (8.19), is complex. Further ahead, it will be seen

155
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

that only the imaginary part of f (Ω) contributes to the flux. Taking the imaginary part,
X 1
ℑ f (Ω) ≡ ℑ (2ℓ + 1)eiδℓ sin δℓ Pℓ (cos θ) (8.27)
ℓ
k
X 1
= (2ℓ + 1) sin2 δℓ Pℓ (cos θ).
ℓ
k

In the forward direction, θ = 0, Pℓ (θ = 0) = 1 for all ℓ, and

1X
ℑ f (Ω) = (2ℓ + 1) sin2 δ, (8.28)
k ℓ

and after comparing to expression for the total cross section in Eq. (8.23) becomes

4π
σ= ℑf (Ω = 0). (8.29)
k
This is known as the optical theorem.
The forward scattering amplitude is thus related to the dissipation of the plane wave. This was
not surprising given the considerations of Sec. 7.9 which showed how the propagator in the
forward direction was related to the cross section. To see that this accounts for the missing flux,
one can write a term for the flux,
1 −iℏ  ∗
⃗ = r ) − (∇ψ⃗k∗ (⃗

F ψ⃗k (⃗
r )∇ψ⃗k (⃗ r ))ψ⃗k (⃗
r) . (8.30)
W 2m
where W is some arbitrarily large volume. The net rate at which particles enter/leave a spherical
surface of radius r enclosing the scattering point is

dN
Z
=r 2 ⃗ ·F
dΩ ⃗ (⃗
r) (8.31)
dt
ℏ 2
Z
= r ⃗ ·⃗
dΩ k.
2m
Taking the gradient, then keeping only those terms which fall slowest in 1/r,

i k k |f (Ω)|2
⃗ =
∆F (r̂ + ẑ)ℑf (Ω)eikr(1−cos θ) + r̂ . (8.32)
W mr W mr 2

Here, the flux of the original plane wave, ℏ⃗

k/mW , is subtracted. As kr → ∞,

 δ(1 − cos θ)
eikr(1−cos θ) kr→∞ = i . (8.33)
kr
To see that the delta function ensues, one can first see that as long as θ ̸= 0 that for large r that
the function oscillates infinitely rapidly as a function of k. It can thus be considered to be zero
for any wave packet constructed over some small, but non-zero, range of k. But, this oscillation

156
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

disappears for θ = 0 and because integrating over cos θ gives a constant it must be proportional
to a delta function. This means that
1
⃗ = 2

∆F −k(r̂ + ẑ)δ(1 − cos θ)ℑf (Ω) + |f (Ω)| . (8.34)
mW r 2
The net rate of particles through the enclosing surface is then
1
Z  Z 
r 2 ⃗ ⃗
dΩ · F = 4πℑf (Ω = 0) + k dΩ|f (Ω)| 2
(8.35)
mW
1
 
dσ
Z
= 4πℑf (Ω = 0) + k dΩ
mW dΩ
1
= {4πℑf (Ω = 0) + kσ} .
mW
Thus, if the net flux is to be zero, one must satisfy the optical theorem, Eq. (8.29). This shows
that the optical theorem is equivalent to stating that the forward scattered wave must interfere
with the original plane wave in such a way that the interference term between the plane wave
and the scattered wave results in a contribution to the flux that is equal but opposite to the net
flux from the scattered wave.

8.3 Calculating Phase Shifts Numerically

Consider a potential that goes to zero for r > b. Rather than considering a solution for the radial
wave function Rℓ (r), it is easier to consider a solution for uℓ (k, r) = krRℓ (r).
uℓ (r) ≡ krRℓ (r), (8.36)
ℏ k
2 2
ℏ ∂
2 2
ℏ ℓ(ℓ + 1)
2
uℓ (r) = − u ℓ (r) + uℓ (r) + V (r)uℓ (r).
2m 2m ∂r 2 2mr 2
Thus, uℓ appears to be the solution to a one-dimensional Schrödinger equation with the effective
potential,
ℏ2 ℓ(ℓ + 1)
Veff (r) = V (r) + , (8.37)
2mr 2
and the extra condition that uℓ (r → 0) = 0.
The solution uℓ can be considered as the linear combination of an incoming and phase-shifted
outgoing wave.
uℓ (r) = uinℓ (r) + e
2iδℓ out
uℓ , (8.38)
where the incoming and outgoing waves are known solutions when r > b,
∗
uin
ℓ (r > b) = krhℓ (r), (8.39)
in,∗
uout
ℓ (r > b) = uℓ (r).

To find the phase shift numerically, one can discretize space into steps of ∆, then choose two
points rn+1 and rn which are chosen beyond b. One can calculate uin in
ℓ (rn+1 ) and uℓ (rn ). One
can then numerically solve the discretized Schrödinger equation,
ℏ2 ℏ k
 2 2 
in in in
− Veff (rn ) uin


− u (rn+1 ) − 2uℓ (rn ) + uℓ (rn−1 ) = ℓ (rn ), (8.40)
2m∆2 ℓ 2m

157
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

to find uin in
ℓ (rn−1 ). One can continue iteratively until one finds uℓ (r → 0). The phase shift δℓ is
then chosen to make uℓ (r → 0) = 0. Thus, δℓ is determined by the phase of uin ℓ as r → 0.

in,∗ 
u
e2iδℓ = − ℓin  . (8.41)

uℓ 
r→0

An alternative method is to integrate from r = 0, then match logarithmic derivatives at r = b

to find the phase shifts. However, this method can be troublesome when it comes to larger ℓ
because the wave functions grow as r ℓ+1 near r = 0.
In summary, phase shifts offer a convenient means to express all the information required to de-
scribe scattering. In the presence of a spherically symmetric potential, each partial wave becomes
a linear combination of the original incoming wave and a phase shifted outgoing wave. Because
measurements are made outside the range of the potential the phase shifts, δℓ (E), completely
describe all behavior.
One does not usually need to include higher values of ℓ in a phase shift analysis. This is because
the partial waves tend to sample distances of order r > ℓ/k. As k → 0, s-wave scattering
dominates the picture. Rarely does one see analyses for ℓ > 3.
Phase shift analyses are common in low-energy nuclear scattering and in atomic physics. They
become rather irrelevant at high energy due to the fact that inelastic channels (A + B → C +
D + E) tend to open at high energy, at which point the phase shift vocabulary is insufficient.
Even when ℓ is not a good quantum number, e.g. spin-orbit scattering, but where j is a good
quantum number, one can still extract phase shifts using polarized beams, although in these
cases the vocabulary includes mixing terms.
Finally, as an example we present a plot of proton-proton phase shifts to give an impression of
the degree to which these quantities are analyzed. In this figure, the data (circles) are indistin-

158
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

guishable from two well-known models (Nijmegan and Bonn).

60
1 3
S0 10 P0

Phase Shift (deg)

Phase Shift (deg)

40

0
20

-10
0

-20 -20
0 100 200 300 0 100 200 300
Lab. Energy (MeV) Lab. Energy (MeV)

3 1
0 P1 12 D2
Phase Shift (deg)

Phase Shift (deg)

-10
8

-20
4

-30
0
0 100 200 300 0 100 200 300
Lab. Energy (MeV) Lab. Energy (MeV)

20
3 3
P2 1.5 F2
Phase Shift (deg)

Phase Shift (deg)

15

1
10

0.5
5

0 0
0 100 200 300 0 100 200 300
Lab. Energy (MeV) Lab. Energy (MeV)

8.4 The Low Energy Limit

At low energy, one can solve the Schrödinger equation for a given partial wave between r = 0
and some point r = b where b is sufficiently large such that the potential is zero. Although the
norm of Rℓ is arbitrary, the logarithmic derivative is completely determined by the potential, the
energy, and the point b.
dRℓ (k, r)/dr 

αℓ (k, b) ≡ (8.42)
R (k, r)  ℓ r=b

We will consider k ≈ 0, one will ignore the energy dependence of αℓ by considering a small
range of energy.

159
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

For r > b, Rℓ (k, r) must have the form

Rℓ (k, r) ∝ h∗ℓ (kr) + e2iδ hℓ (kr) ,

(8.43)

which means that the logarithmic derivative at r = b becomes

∂r h∗ℓ (kr) + e2iδℓ hℓ (kr)) |r=b

α(k, b) = . (8.44)
h∗ℓ (kb) + e2iδℓ hℓ (kb)

Using the definition of hℓ = jℓ + inℓ allows one (with some algebra) to write the phase shift in
terms of α, jℓ and nℓ ,
∂nℓ /∂r|r=b − αnℓ (kb)
cot(δℓ ) = . (8.45)
∂jℓ (kr)/∂r|r=b − αjℓ (kb)
Thus, by finding the logarithmic derivative at r = b, one determines the phase shift. Note that if
the potential is zero, Rℓ would be proportional to jℓ and the denominator would diverge forcing
the phase shift to zero.
We are now in a position to consider the behavior at low k where jℓ and nℓ have the following
behavior,

(kr)ℓ
jℓ (kr) → (8.46)
(2ℓ + 1)!!
(ℓ − 1)!!
nℓ (kr) → ,
(kr)ℓ+1

where (2ℓ + 1)!! ≡ 1 · 3 · 5 · · · (2ℓ + 1). Inserting these into the expression for the phase shift
above.
ℓ + 1 + bαℓ (k, b)
cot δℓ (k) ≈ (kb)−(2ℓ+1) (2ℓ − 1)!!(2ℓ + 1)!! . (8.47)
ℓ − bαℓ (k, b)
For low k, the kinetic term in Schrödinger’s equation is negligible compared to the potential and
αℓ approaches a constant. Thus, the momentum dependence of the phase shifts at low relative
momentum is
sin δℓ (k) ∝ k2ℓ+1 . (8.48)
One can see that all phase shifts tend to an integral multiple of π at k = 0, and that the cross
section is dominated by the s-wave contribution at low energy. In fact, the scattering length, a, is
defined as the derivative of the ℓ = 0 phase shift at k = 0,

∂
a≡− δ0 (k)|k=0 . (8.49)
∂k
The cross section at very low energy is then,

4π
σ≈ sin2 (ka) = 4πa2 . (8.50)
k2

160
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

Example 8.2: Scattering off Spherical “Square-Well” Potential

A simple example that is common on exams is that of s-wave scattering off a square well
potential. Consider the repulsive potential

V0 , r < b
V (r) = (8.51)
0, r > b

Find the scattering length and the cross section at k ≈ 0 for a particle of mass m.
We need only consider the s wave in this case. Using the definition u(k, r) ≡ rRℓ=0 (k, r),
one knows that the Schrödinger equation for u(k, r) looks exactly like a one-dimensional
Schrödinger equation. Furthermore, the solution has the following form in the two regions,
s
2m(V0 − E)
uI (k, r) = A sinh κr, κ ≡ , (8.52)
ℏ2
s
2m(E)
uII (k, r) = sin(kr + δ), k ≡
ℏ2

Matching logarithmic derivatives at the boundary gives

1 1
tanh κb = tan(kb + δ). (8.53)
κ k
Solving for δ for small k,


−1
k
δ = −kb + tan tanh κb (8.54)
κ
1
 
≈ k −b + tanh κb
κ

The scattering length and cross section are thus

1
a=b− tanh κb (8.55)
κ
2
σ(k = 0) = 4πa .

Note that as V0 → ∞ the phase shift becomes δ → −kb, which agrees with the result for a
hard sphere in Example 8.1. Also, if the potential is small, κ → 0, which leads to a → 0 and
σ(k = 0) → 0

8.5 Levinson’s Theorem

As shown in the previous section, all phase shifts begin life at multiples of π. If the phase shift
at k = 0 were anything else an infinite cross section would result at small k because the cross
section behaves as 4π sin2 δ/k2 . As k → ∞, phase shifts tend to zero because the kinetic
energy should overwhelm the potential energy. This asymptotic behavior at large k can be also

161
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

understood by realizing that the in the high-energy limit the phase shift is near V ∆t/ℏ, where
∆t is the time spent in the potential, which goes to zero as the particle moves very quickly.
Levinson’s theorem relates the phase shift at zero energy, which is a multiple of π, to the number
of bound states.
(2ℓ + 1)δℓ (k = 0) = NB π, (8.56)
where NB is the number of bound states of angular momentum ℓ.
In order to explain the physical motivation of Levinson’s theorem we digress to consider the
density of states of particles in a large sphere of radius R, which feels a short range potential
V (r), with the origin being at the center of the sphere. The wave function at large r is
ψ(r → ∞) ∝ sin(kr + δℓ ), (8.57)
so that the boundary conditions restrict the possible values of k to
kR + δℓ (k) = nπ. (8.58)
Thus, the density of states in momentum is
dn R 1 dδℓ
= + . (8.59)
dk π π dk
If one considered ℓ ̸= 0, an additional factor of (2ℓ + 1) would be added. The change of the
density of states due to the non-zero potential is (2ℓ + 1)(1/π)dδℓ /dk. The integrated number
of extra states inserted between k = 0 and k = ∞ due to the potential is
δℓ (k → ∞) − δℓ (k = 0)
∆Ncont. = (2ℓ + 1) . (8.60)
π
However, the net number of states under consideration is not affected by the potential. If
∆Ncont. states were pushed out of the continuum, then they must have become bound states.
NB + ∆Ncont. = 0, (8.61)
where NB is the number of bound states. Combining this constraint with the Eq. (8.60) and with
the fact that δℓ (k → ∞) = 0 gives Levinson’s theorem, Eq. (8.56).
Levinson’s theorem is important as it gives one an idea of the general behavior to expect from
phase shifts. Attractive potentials tend to have positive phase shifts. If no bound state exists,
the phase shifts rise near k = 0 indicating that the states in the continuum were pulled down
toward k = 0. Then at higher k, the phase shifts fall, indicating that the density of continuum
states was depleted at higher k. If a bound state exists, the phase shift would generally start at
π and usually fall as a function of k. The falling phase shift denotes a negative correction to the
density of states. These are the states from which the bound state was formed.
Phase shifts behave rather peculiarly when the potentials are at the threshold of creating a bound
state. A small change in the potential causes the δ(k = 0) to jump from zero to π. In these
instances scattering lengths can be anomalously long. Such an example is neutron-neutron scat-
tering. The scattering length is nearly -20 fm, which gives a cross section over 1000 fm2 even
though the range of the potential is only of the order of one fm. Thus, the neutron-neutron cross
section at low momentum is ∼ 100 times larger than πR2 , where R is the range of the strong
interaction.

162
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

8.6 Effective Range Theory for ℓ = 0

The mathematics of low-energy scattering theory is often referred to as effective range theory,
and goes back to the work of Schwinger and of Hans Bethe in 1949, https://journals.aps.org
/pr/abstract/10.1103/PhysRev.76.38, and was mainly applied to nucleon-nucleon scattering.
As k → 0, scattering is dominated by the ℓ = 0 partial wave. Phase shifts (modulo integer
numbers of π) at small momentum rise with powers of k2ℓ+1 , so cross sections, which behave as
(4π sin2 δ)/k2 , behave as k4ℓ at low momentum, and for many experiments only the s−waves
contribute. Effective range theory parameterizes how the ℓ = 0 cross sections are principally
decided by two properties of the potential, the scattering length and the effective range. Given
that we are interested in the low-momentum limit, and are thus interested in viewing something
akin to a Taylor expansion of the phase shift with respect to the momentum, it is not surprising
that two properties of the potential, e.g. the width and depth, would determine the first two
terms in the Taylor expansion. However, the effective range expansion, which is an expansion
of cot δ, tends to do remarkably well as the higher-order terms play little role until the momenta
exceed the inverse scale of the potential.
The principal relation of effective range theory is derived as follows. First consider the radial
solution for the ℓ = 0 partial wave, u(k, r) = rRℓ=0 (k, r). If the potential vanishes for r > R,
the solution for r > R is of the form,
u(k, r > R) = sin(kr + δ(k))/Z(k). (8.62)
We will pick real solutions (which you can do if the potential is real) and leave off the factor eiδ
from Eq. (8.16). We will also consider a solution to the Schroödinger equation, w(k, r), without
the potentil, that matches u(k, r) for r > R, i.e., w(k, r > R) = u(k, r). This solution will not
satisfy the boundary condition at the origin, i.e.,
w(k, r > 0) = sin(kr + δ(k))/Z(k). (8.63)
Further, we will define the arbitary normalization constant Z(k) = sin δ(k). With this def-
inition, w(k = 0) = 1. Now, considering solutions u1 and u2 at two momenta, k1 and k2
respectively, one can consider the Schrödinger equations,
k12 u1 = −∂r2 u1 + 2mℏ2 V (r)u1 , (8.64)
k22 u2 = −∂r2 u2 + 2mℏ2 V (r)u2 ,
k12 w1 = −∂r2 w1 ,
k22 w2 = −∂r2 w2 .
Then, taking the products, u1 u2 and w1 w2 ,
Z R
2 2
(k1 − k2 ) dr (u1 u2 − w1 w2 ) (8.65)
0
Z R
dr −(∂r2 u1 )u2 + (∂r2 u2 )u1 − (∂r2 w1 )w2 + (∂r2 w2 )w1 .
 
=
0

The potential V (r) has fallen out of the expression, and be recognizing that (∂r2 u1 )u2 −u1 ∂r2 u2 =
∂r [(∂r u1 )u2 − u1 ∂r u2 ] is a total derivative, one can perform the integration,
Z R
2 2
(u1 ∂r u2 − u2 ∂r u1 + w1 ∂r w2 − w2 ∂r w1 )r=0 = (k1 − k2 ) dr (u1 u2 − w1 w2 ). (8.66)
0

163
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

This expression is exact and applies for any momenta k1 and k2 . The upper limit of the integra-
tion at R vanished because wi = ui for r > R. Now, because u1 (r = 0) = u2 (r = 0) and
wi (r = 0) = sin(δi ), the expression simplifies further,
Z R
2 2
(w1 ∂r w2 − w2 ∂r w1 )r=0 = (k1 − k2 ) dr (u1 u2 − w1 w2 ),
0
Z R
k2 sin(δ1 ) cos(δ2 ) − k1 cos(δ1 ) sin(δ2 ) 2 2
= (k1 − k2 ) dr (u1 u2 − w1 w2 ). (8.67)
sin(δ1 ) sin(δ2 ) 0

This last expression forms the basis for understanding the low-momentum expansion. For ex-
ample, taking the limit k1 → 0, and setting k2 = k,
1 1
k cot(δ) = − + ρk2 (8.68)
a 2
Z R
ρ≡2 dr [w(k = 0, r)w(k, r) − u(k = 0, r)u(k, r)] ,
0 
dδ 
a≡− .
dk k=0
This is the effective range formula. The quantity ρ is known as the effective range. Because
u(r = 0) = 0 and w(r = 0) = 1, it provides a characteristic distance over which the u returns
to the asymptotic form, and should be of the length scale of the potential. The scattering length
a was defined earlier. For small momenta, the cross section is
4π
σ= sin2 δ ≈ 4πa2 . (8.69)
k2
The scattering length need not be of the range of the potential. In fact, it can be arbitrarily large
in the limit that the potential is adjusted to very nearly providing a bound state. For higher
momenta terms of higher power of k contribute to the effective range expansion. The scale at
which they become important is given by the effective range ρ, i.e., Eq. (8.68) above should be
accurate for kρ < 1.

8.7 Coulomb Waves and Gamow Factors

The Coulomb potential is not short-range, and even the incoming wave is altered by the potential
at large distance. Hence, the incoming and outgoing spherical waves are not Hankel functions
but are instead Coulomb waves. At large r the spherical Coulomb waves behave as
1
R(r) ∼ e±i(kr−γ log r) , (8.70)
r
where
µZ1 Z2 e2 1
γ≡ =−. (8.71)
ℏ2 k a0 k
When the potential is attractive, e.g. the Hydrogen atom, γ is negative. The factor γ is known
as the Sommerfeld parameter.

164
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

The entire solutions for Coulomb waves are known as confluent hypergeometric functions which
can be found in books of special functions such as https://en.wikipedia.org/wiki/Abramowi
tz_and_Stegun. They are also often simply called Coulomb wave functions.
When an extra potential is added to the problem, one can still classify the behavior in terms of
phase shifts, but the problem becomes one of phase-shifted Coulomb waves rather than phase-
shifted Hankel functions. Of course, all the relations for total cross sections are modified because
the Coulomb force produces a scattering all it’s own.
The expression for a plane wave plus it’s scattered portion is written in terms of hypergeometric
functions as
ψk (r) = Γ(1 + iγ)e−πγ/2 eikz1 F1 (−iγ; 1; ik(r − z)). (8.72)
For small r (r << a0 ) the hypergeometric function 1 F1 goes to unity and the solution is ap-
proximately,

ψk (r << a0 ) = Γ(1 + iγ)e−πγ/2 eikz (8.73)

2 2 −πγ
|ψk (r << a0 )| = |Γ(1 + iγ)| e
2πγ
= 2πγ .
e −1
This last factor is referred to as the Gamow factor or penetrability.
Note that the Gamow factor is independent of r. For repulsive potentials, it can be thought of
as the penetration probability of reaching the origin, relative to what it would have been if there
were no Coulomb. Of course, in classical physics a charged particle never reaches the origin
when there is a repulsive Coulomb barrier. If the potential is repulsive, gamma is positive. As
k → 0, γ → ∞ and the Gamow factor goes to zero, meaning that low energy waves have great
difficulty penetrating the barrier. As k is increased and approaches (2πa0 )−1 the penetrability
rises. For attractive potentials, the Gamow factor is greater than unity. In both the attractive and
repulsive cases, the Gamow factor approaches unity as k → ∞.
The Gamow penetrability plays an important role in astrophysical rates. Fusion reactions would
happen much more quickly if it were not for the Coulomb barrier. Rather than listing cross
sections, reactions are often described by what is known as the astrophysical S factor,

S(E) = Ee2πγ σ(E). (8.74)

For large barriers this cancels the most divergent part of the Gamow factor, and the resulting
quantity, S(E) is well behaved as E → 0.
Example 8.3: When Does a Potential Become Long Range
If a potential falls as α/r m , for what values of m can one define the scattering with phase
shifted partial waves?
Solution:
Many forms for the potential never really go to zero at large r, e.g. the Yukawa form e−αr /r,
but they do approach zero well enough. For the power-law form considered here, one can
realize that for large r the potential falls sufficiently slowly to consider the estimate the con-
tribution to the phase shift from the potential outside some large distance R with the WKB

165
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

approximation,
Z
∆δ(p) = ∞dr ∆p(r)/ℏ.
R

Here, ∆ denotes that this is the contribution to the phase shift solely p
from the potential out-
side R. The momentum p(r) is a function of the potential, p(r) = 2m(E − V (r)), and
expanding for small V ,
Z √ V (r)
∆δ(p) ∞dr 2mE .
R 2E
If the potential falls of more quickly than 1/r, the integral vanishes for sufficiently large R.
However, for the Coulomb potential, the integral is undefined for any R. Thus, the answer is
that phase shifted Hankel waves can be considered as long as m > 1. One can also see that
for any exponentially falling potential, such as the Yukawa form, one can also apply the usual
scattering theory.

8.8 Classical Interpretation of the Scattering Wave Function

⃗
The scattered state wave function, ψ(⃗ r ), normalized so that the incoming wave is eik·⃗r , can be
k, ⃗
thought of as a measure of the relative probability density (after being squared) for a particle of
incoming momentum ⃗ k to be found at ⃗
r . The solution includes outgoing momenta distributed
over all directions. If one takes the complex conjugate of the scattered wave function, one finds a
solution where outgoing parts of the solution match a plane wave of momentum ⃗ k. The incom-
ing components then include the complex conjugate of the outgoing wave in scattering theory
and describe the incoming flux from all directions, that are joined together precisely to form an
outgoing wave that matches a plane wave.
One insightful relation is that of the wave function normalization compared to the change in the
density of states in terms of (1/π)dδ/dk. This can be derived in a similar approach to find the
effective range formula above. For a given partial wave,
1 dδℓ
Z  
2 (0) 2
dr |uℓ (r)| − |uℓ | = , (8.75)
2 dk
(0)
where uℓ (k, r) = rRℓ (k, r) and uℓ (k, r) = krjℓ (kr) are the solutions for partial waves with
and without potentials. This provides insight into the meaning of the squared scattered wave
function. It describes the strength with which one can emit a particle into a given outgoing mo-
mentum state. One can equivalently think of the fact that a particle with asymptotic momentum
⃗
k is more or less likely to be in the region of the potential due to the presence of the potential,
or one can think of being more likely to be in the region of the potential because the density of
states changed.
The classical analog to the squared scattered wave function is in terms of the ratio of the phase
space at the point ⃗
r compared to the the size of the same phase space element in its asymptotic
momentum state. If one emits a particle at position ⃗
r0 into some differential momentum volume

166
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

d3 p0 centered around p⃗0 , the potential will alter the momentum into some differential volume
3
d pf centered about p⃗f . In the limit that the momenta are larger than characteristic inverse sizes
of the potential, p >> ℏ/R, the wave functions become

2 d3 p0
|ϕ(⃗ r )| →
p, ⃗ . (8.76)
d 3 pf
This latter ratio is a function of ⃗ r0 and p
⃗f , and can be found by solving for the classical trajec-
⃗0 (⃗
tories, p r0 ). For the case of the Coulomb interaction, this ratio can be found analytically,
pf , ⃗
https://journals.aps.org/prc/pdf/10.1103/PhysRevC.45.387, using energy conservation,
angular momentum conservation, and conservation of the Lenz vector.

8.9 Exercises
1. Show that if the function uℓ (kr) is defined in terms of Rℓ (r)
uℓ (kr) ≡ rRℓ (r),
where Rℓ is a solution to the radial Schrödinger equation
ℏ2 1 ∂ 2 ℏ2 ℓ(ℓ + 1) ℏ2 k2
 
− r + + V (r) R ℓ (r) = Rℓ (r),
2m r ∂r 2 2m r2 2m
that uℓ satisfies the differential equation,
 2
ℓ(ℓ + 1)

d
+ 1 u ℓ (x) = uℓ (x) + β(x)uℓ (x),
dx2 x2
where β is proportional to the potential,
2m
β(x) = V (x/k).
ℏ2 k2
2. Recurrence relations for Bessel functions provide you the ability to find forms for solutions
at higher ℓ given you know the form for ℓ = 0 and ℓ = 2

(a) Show that in the case of zero potential that the solutions uℓ satisfy the recurrence
relation.
(ℓ + 1) d
uℓ+1 (x) = uℓ (x) − uℓ (x).
x dx
Use the expressions from the previous problem,
 2
ℓ(ℓ + 1)

d
+ 1 u ℓ (x) = uℓ (x) + β(x)uℓ (x). (1)
dx2 x2
(b) Show that this recurrence relation can be equivalently expressed as
ℓ d
fℓ+1 (x) = fℓ (x) −
fℓ (x),
x dx
where fℓ is a solution to the radial Schrödinger equation, fℓ (kr) ≡ uℓ (kr)/(kr),
which means that fℓ might be any linear combination of jℓ and nℓ .

167
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

(c) One can also show that a second recurrence relation is satisfied,

(ℓ + 1) d
fℓ−1 (x) = fℓ (x) + fℓ (x).
x dx
Given this recurrence relation, plus the one from the previous problem, show that

(2ℓ + 1)
fℓ−1 (x) + fℓ+1 (x) = fℓ (x)
x

(d) Using expressions for j0 , j1 , n0 and n1 , use recurrence relations to find expressions
for j2 and n2 .
(e) Using the recurrence relations, show that jℓ (z) and nℓ (z) behave as z ℓ and z −(ℓ+1)
respectively for z → 0. Begin with the facts that j0 (z) and n0 (z) behave as z 0 and
z −1 respectively, and that they are even and odd functions in z.

3. Consider a particle of mass m that interacts with a spherically symmetric attractive poten-
tial. 
−V0 , r < b
V (r) =
0, r > b

(a) What is the minimum depth Vmin that allows a bound state?
(b) Find an expression for the phase shift in terms of a particle whose momentum is p.
(c) Assuming the depth is V0 = 0.99 · Vmin , plot the s-wave phase shift for momenta in
the range 0 < p < 5ℏ/b. Use units of ℏ/b for the momenta.
(d) Repeat the above problem for V0 = 1.01 · Vmin .
(e) What are the scattering lengths for the two potentials?

4. Consider a proton scattering off of a an attractive one-dimensional potential,



  ∞,
 x<0
r2
V (x) = −V0 1 − R2 , 0 < x < R

0, r>R


For this example, we will consider R = 2.5 fm, and V0 =16 MeV. If you wish, to make the
units more natural, you may consider ℏc=197.327 MeV·fm, and mp = 938.27 MeV/c2 .
Consider a particle incident on the well with energy E that enters and leaves the well with
energy E. Far away, the solutions are of the form,

ψ(x) = e−ipx/ℏ − e2iδ+ipx/ℏ , x >> R

(a) Programming in either PYTHON or C++, construct a program that runs and returns
a listing of δ vs. p for 0 < p < 600 MeV/c, in steps of 2.0 MeV/c.

168
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

A graph of the results:

(b) EXTRA CREDIT Make a graph like the one above, except for the region between
p=0 and p=1.0 MeV, and consider two strengths of the potential, V0 = 17.0 MeV
and V0 = 17.025 MeV. Be sure to calculate values for very small values of p, in
steps of .001 MeV. For this problem, turn in a paper copy of the graph.

5. Consider a potential which gives non-zero phase shifts for 0 ≤ ℓ ≤ ℓmax , where ℓmax is
a large number. Assume these phase shifts can be considered as random numbers, evenly
distributed between zero and 2π. Using the expression for the cross section,

4πℏ2 X
σ= 2
(2ℓ + 1) sin2 δℓ ,
p ℓ

(a) Find the overall cross section by averaging over the expectation of the random phases.
Give your answer in terms of ℓmax and the incoming momentum p.
(b) Consider a problem classically where one scatters off a strong central potential whose
maximum range is Rmax . From classical arguments, what is the maximum angular
momentum of a particle that scatters? Give your answer in terms of Rmax and the
incoming momentum p. What is the total cross section in terms of Rmax in the limit
that ℓmax is large.

6. A particle of mass m experiences an attractive spherically symmetric potential,

V (r) = −βδ(r − a),

where β > 0.

(a) In terms of a, and the electron mass m, what is the minimum value of β that results
in a bound state?
(b) What is the scattering length and the cross section in the limit that the incident beam
energy is zero.

169
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

(c) If a scattered wave in a large volume behaves as

⃗
ψ(⃗ r , t) ∼ eik·⃗r−iωt , t → ∞
k, ⃗
in the outgoing limit (large time after interacting with potential), what is the relative
probability,
ρ(⃗r = 0)
α(k) = ,
ρ0 (⃗
r = 0)
that it will appear at the origin while interacting with the potential? Here ρ0 is the
probability density (per unit volume) in the absence of the potential, and ρ is the
probability density with the potential in place. FYI: The ratio α would be the same if
the boundary conditions specified an incoming plane wave, instead of matching to an
outgoing plane wave.
(d) Assume β is sufficiently large to bind a particle, and that the ground state energy
is −B. For the ground state what is the probability density of finding the particle at
r = 0? Refer to this as ρb (⃗
⃗ r = 0)? Given answer in
pterms of a and the binding energy
B (or equivalently the decay wave number, q ≡ 2mB/ℏ2 ). HINT: You don’t need
to solve for the binding energy!
7. Near a resonance of energy ϵR , a phase shift behaves as:
Γ/2
tan δℓ = ,
ϵR − E
where E is the c.m. kinetic energy. For the following problems, assume that Γ << ϵR ,
so that the 4π/k2 prefactor in the expression for the cross section can be considered as a
constant.
(a) Write down the cross section σℓ (E).
(b) What is the maximum cross section for a narrow cross section (as E is varied) for
scattering through that partial wave? (How does it depend on ϵR , Γ, the reduced
mass µ, and ℓ)?
R
(c) What is the energy integrated cross section ( σℓ (E)dE)?
8. The temperature at the center of the sun is 15 million degrees Kelvin. Consider two protons
with a relative kinetic energy characteristic of the temperature,
ℏ2 k2 3
= kT.
2µ 2
(a) What is the Gamow penetrability factor? Give a numeric value.
(b) If the two particles were a proton and a 12 C nucleus, what would the penetrability
factor become?
9. Consider a particle of mass m undergoing a repulsive spherically symmetric Coulomb
potential, V = Ze2 /r. The classical analogue of the squared wave function is
d3 pi
|ϕ(⃗ r )|2 →
pf , ⃗ .
d 3 pf

170
PHY 851/852 8 SCATTERING AT LOWER ENERGIES

Here, p
⃗i is the momentum when the particle is at position ⃗
r , and p
⃗f is the asymptotic
momentum at large times.

(a) If one averages over all directions of the final momentum, what is ⟨|ϕ(⃗ r )|2 ⟩? Give
pf , ⃗
a sketch of the classical approximationn to ⟨|ϕ(⃗ r )|⟩ as a function of r for fixed p.
pf , ⃗
(b) Repeats (a) but working in two dimensions, i.e. find d2 pi /d2 pf .
(c) Repeats (a) and (b) but working in one dimension, i.e. find dpi /dpf .

171
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

9 Second Quantization and Radiation

9.1 Quantum Fields and Second Quantization

Thus far, our studies of wave functions, ψ(⃗ r ) in coordinate space or ψk in momentum space,
have considered the quantum mechanics of a single particle. The particle exists in some kind
of state or at some position, and exists in perpetuity – at any time there is a single particle. In
reality, particles are created and destroyed and the net particle number is in constant flux. The
most obvious example is the emission or absorption of a photon. But even charged particles
are constantly being created in particle-antiparticle pairs. For example, the eigenstate referred
to a as photon includes contributions from electron-antielectron pairs. Even the vacuum in-
cludes the creation of such pairs, though the pairs exist only for a short time determined by the
∆E∆t ≳ ℏ/2 uncertainty principle. In addition to electromagnetic decays, most other decays in
particle physics involve the creation of new particles such as a Higgs boson decaying to a bottom
and anti-bottom quark. In relativistic quantum field theory, all interactions, e.g. the Coulomb
force, are realized through the creation and absorption of particles. The potential terms in Hamil-
tonians involve operators that create or destroy particles, with such operators sometimes being
referred to a second quantization. For this class, we will focus our study of second quantization
on decays.
Creation and destruction operators were introduced to describe the creation of energy quanta in
a harmonic oscillator. The operators obeyed the relations,

[ai , a†j ] = δij , [ai , aj ] = 0, [a†i , a†j ] = 0, (9.1)

where i specifies which oscillator is being affected. If one had N harmonic oscillators, one
would have N independent creation operators and N independent destruction operators. Note
that the operators corresponding to different oscillators commute with one another as they are
unrelated.
The essential feature of creation and destruction operators is that they increase/decrease the
number of quanta, where the number of quanta of the oscillator i is found by using the operator
Ni = a†i ai . However, instead of counting only energy quanta, as in the case of the harmonic
oscillator, the number operator could also refer to a number of particles in some single-particle
level i. Let us then consider a creation operator, a†k , for each mode, k, where a mode corresponds
to a single-particle eigenstate. The one-particle state, |k⟩, of momentum k, is created by a†k
operating on the vacuum.
|k⟩ = a†k |0⟩. (9.2)
The state is normalized to unity, just as one would expect for creation operators.

⟨k′ |k⟩ = ⟨0|ak′ a†k |0⟩ = δk′ k . (9.3)

By operating twice with a†k , one creates a state with two particles of momentum k. Because the
mode k might have any number of such √ particles the complete set of states for a single mode
† † n
would include |0⟩, ak |0⟩, · · · , [(ak ) / n!]|0⟩, · · · . This implies that the particles are bosons,
because the levels can be multiply occupied. Field operators for fermions is a topic for another
chapter.

172
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

Example 9.1: Coupled Harmonic Oscillators

Here, we consider two coupled harmonic oscillators. The oscillators are described by creation
operators a†1 and a†2 . Let the Hamiltonian be

H = H0 + V, (9.4)
H0 = ϵ1 a†1 a1 + ϵ2 a†2 a2 ,
β a†2 a1 + a†1 a2 .


V =

Consider the operators b†1 and b†2 defined by

b†1 ≡ a†1 cos θ + a†2 sin θ, (9.5)

b†2 ≡ a†2 cos θ − a†1 sin θ.

a) Show that b1 , b2 , b†1 and b†2 obey the commutation rules for destruction operators.
b) Find E1 , E2 and θ so that

H = E1 b†1 b1 + E2 b†2 b2 .

Solution:
(a) To test that bi and b†i act as creation operators,

[b1 , b†1 ] = [(a1 cos θ + a2 sin θ), (a†1 cos θ + a†2 sin θ)]
= cos2 θ + sin2 θ = 1,
[b2 , b†2 ] = [(a2 cos θ − a1 sin θ), (a†2 cos θ − a†1 sin θ)]
= cos2 θ + sin2 θ = 1,
[b1 , b†2 ] = [(a1 cos θ + a2 sin θ), (a†2 cos θ − a†1 sin θ)]
= − sin θ cos θ + sin θ cos θ = 0,
[b1 , b1 ] = [b2 , b2 ] = [b1 , b2 ] = 0.

By taking the complex conjugates, one can see that [b†1 , b†1 ] = [b†2 , b†2 ] = [b†1 , b†2 ] = 0.
(b) First, expand the Hamiltonian in the desired form in terms of ai and a†i using Eq.s (9.5),

H = E1 b†1 b1 + E2 b†2 b2
E1 + E2 †
E1 − E2 †
a1 a1 + a†2 a2 + a1 a1 − a†2 a2 cos 2θ


=
2 2
E1 − E2 †
a1 a2 + a†2 a1 sin 2θ,


+
2
then compare it to the original Hamiltonian in Eq. (9.4),
ϵ1 + ϵ2
ϵ1 − ϵ2 †
a†1 a1 + a†2 a2 + a1 a1 − a†2 a2 + β a†1 a2 + a†2 a1 .



H =
2 2

173
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

To determine E1 , E2 and β set the two Hamiltonians to be the same,

ϵ1 + ϵ2 = E1 + E2 ,
ϵ1 − ϵ2 = (E1 − E2 ) cos 2θ,
1
β = (E1 − E2 ) sin 2θ.
2
The Hamiltonians are equivalent when

2β
tan 2θ = ,
ϵ1 − ϵ2
E1 + E2 = (ϵ1 + ϵ2 ),
E1 − E2 2 ϵ1 − ϵ 2 2
   
2
=β + .
2 2

One can note the algebraic equivalence of this problem to the two-component problem with a
Hamiltonian
ϵ1 + ϵ 2 ϵ1 − ϵ2
H = I+ σz + βσx ,
2 2
where one is asked to find the eigenvalues E1 and E2 .

9.2 Field Operators: Creation and Destruction Operators in Coordinate Space

The operator a⃗†k will create a particle with momentum ⃗ k, where the momentum states are as-
sumed to be discrete. The operator a⃗k will remove a particle from such a state. The particle
might be even be massive or charged. If the state |0⟩ is the vacuum, a⃗†k |0⟩ is a state with one
particle in momentum state ⃗ k. Field operators are the coordinate-space analogs to a⃗k and a⃗†k .
They are defined as
⃗
X eik·⃗x
Ψ(⃗
x) = a⃗k √ , (9.6)
⃗
V
k
−i⃗
k·⃗
x
†e
X
†
Ψ (⃗
x) = a⃗k √ .
⃗
V
k

174
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

They obey the commutation rules,

1 X ⃗ ⃗′
x), Ψ† (⃗
[Ψ(⃗ y )] = ei(k·⃗x−k ·⃗y) [a⃗k , a⃗†k′ ] (9.7)
V ⃗
k,⃗
k′
1 X ⃗
= eik·(⃗x−⃗y)
V ⃗
k
1
Z
⃗
= d3 k eik·(⃗x−⃗y)
(2π)3
= δ 3 (⃗
x−y
⃗ ).

Further, they return zero when acting on the vacuum,

x)|0⟩ = 0, ⟨0|Ψ† (⃗
Ψ(⃗ x) = 0. (9.8)

These operators create the state |⃗

x⟩,

x⟩ = Ψ† (⃗
|⃗ x)|0⟩, ⟨⃗
x| = ⟨0|Ψ(⃗
x). (9.9)

Here, |⃗
x⟩ is a state with one particle at the position ⃗
x, and is normalized as

⟨⃗ y ⟩ = ⟨0|Ψ(x)Ψ† (⃗
x|⃗ y )|0⟩ = δ 3 (⃗
x−y
⃗ ). (9.10)

One can easily check that

⃗
⃗ † eik·⃗x
⟨⃗
x|k⟩ = ⟨0|Ψ(⃗
x)a⃗k |0⟩ = √ . (9.11)
V
One should keep in mind that Ψ(⃗ x) is an operator, not a wave function. To remind the reader of
the difference, a wave function, ⟨⃗ x|ψ⟩ = ψ(⃗ x), is an expectation value and is a complex number.
It includes a bra and ket. For the case of a finite number of states, a bra or ket is a vector and an
expectation has the form ⟨χ|A|ϕ⟩ = χ∗i Aij ϕj and there are no remaining indices. In contrast
an operator is a matrix, e.g., Aij . Each ij element of the operator projects out the j th element of
the vector described by the ket, then transforms it to |i⟩. The creation operator Ψ† (⃗ x) operating
on some state ϕ replaces ϕ with the same state, but with a particle added at ⃗ x. Unlike the
cases from the previous chapters the position labels ⃗ x or the momentum labels p ⃗ are insufficient
to describe a state. A complete basis must include configurations with different numbers of
particles. Thus, one needs operators that describe the change from states with different particle
number. Creation and destruction operators provide a means to describe elements involving n
and n ± 1 particles. A product of creation or destruction operators is required to describe a
transition between states which differ by more than 2 in particle number. The fact that much
of physics, including the standard model of particle physics, is modeled by a Hamiltonian with
powers of creation operators no higher than the quartic level is indeed a profound property of
nature.

175
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

If |ϕ⟩ refers to a one-particle state, the state is related to the wave function ϕ(x) by
Z
|ϕ⟩ = d3 x ϕ(x)Ψ† (⃗ x)|0⟩, (9.12)
Z
Ψ(⃗x)|ϕ⟩ = d3 x′ ϕ(⃗x′ )Ψ(⃗x)Ψ† (⃗
x′ )|0⟩

= ϕ(⃗
x)|0⟩,
⟨⃗
x|ϕ⟩ = ⟨0|Ψ(⃗x)|ϕ⟩
= ϕ(⃗
x).

Charge densities and currents can also be considered as operators. For particles of charge e,

ρ(⃗x) = eΨ† (⃗
x)Ψ(⃗x), (9.13)
⃗j(⃗ eℏ
(−∇Ψ† (⃗ x) + Ψ† (⃗


x) = x))Ψ(⃗ x)∇Ψ(⃗
x) .
2mi
If the state |ϕ⟩ and |χ⟩ are one-particle states, one can use the relations above to verify the
previous relations,

⟨χ|ρ(⃗x)|ϕ⟩ = eχ∗ (⃗
x)ϕ(⃗x), (9.14)
eℏ
⟨χ|⃗j(⃗
x)|ϕ⟩ = [χ∗ (⃗ x) − (∇χ∗ (⃗
x)∇ϕ(⃗ x)) ϕ(⃗
x)] .
2mi

9.3 Energy and the Hamiltonian

One can write the Hamiltonian for freely moving particles:

ℏ2
Z
H0 = − d3 xΨ† (⃗
x)∇2 Ψ(⃗
x). (9.15)
2m
ℏ2
Z
⃗ ⃗′
X
= d3 x k2 ei(k−k )·⃗x a⃗†k a⃗k′ .
2mV ⃗
k,⃗
k′

The integral over ⃗x gives zero due to the varying phase unless ⃗
k=⃗k′ , at which point the phase
is unity and the integral over ⃗
x cancels the volume in the denominator,
X ℏ2 k2
H0 = a⃗†k a⃗k . (9.16)
⃗
2m
k

Thus, even though the Hamiltonian, expressed in terms of the field operators Ψ and Ψ† , looks
like the familiar expression for a wave function, it is far more powerful as it correctly expresses
the energy even when many particles are present in the system, and even in those cases where
multiple particles occupy the same single-particle level.
As will become apparent throughout this chapter, the profound strength of field operators is
that the Hamiltonian can be expressed in a remarkably compact form between any two states,

176
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

where the bra and the ket might have any number of quanta. For the case of a single particle,
the Hamiltonian could in general be expressed as a matrix, Hkk′ , where k and k′ represent any
single-particle states. Once particle number is no longer fixed, the labels k or k′ might refer to
any configuration of labels for any number of particles. Thus, if expressed as a matrix, the most
general Hamiltonian would have require knowledge of how 16 particles in any given configura-
tion would be linked to every possible 17-particle configuration. However, if the Hamiltonian is
expressed in terms of field operators, the Hamiltonian can be expressed in terms of a relatively
small number of fields, each with a single-particle label. The matrix elements ⟨i|H|j⟩ can then
be easily found even if the states |i⟩ and |j⟩ have numerous quanta, each in some unique state.

9.4 Interaction with an External Potential

An interaction with an external potential can be written as
Z
Hint = r )Ψ† (⃗
d3 rV(⃗ r )Ψ(⃗
r ). (9.17)

This can be written in terms of momentum-space creation and destruction operators by substi-
tuting the expressions for Ψ† from Eq. (9.6),
1
Z
⃗ ⃗′
X †
Hint = d3 r a⃗k a⃗k′ e−i(k−k )·⃗r V(⃗
r) (9.18)
V ⃗ ⃗′ k,k
1 X
= Ṽ(⃗ k′ )a⃗†k a⃗k′ ,
k −⃗
V ⃗
k,⃗
k′

where Ṽ is the Fourier transform of V. Thus, the form of the interaction is similar to that of the
Ex. 9.1 in that this features terms where one state is created and another destroyed.
In fact, one can reduce the solution of the problem (finding the energy eigenvalues) to that of
diagonalizing a matrix. X
H = Aji a†j ai . (9.19)
i,j

The simplified Hamiltonian results by linearly transforming the states to a new basis where
A → B, and B is diagonalized. In this new basis, one transforms the operators with a unitary
transformation U ,
ai = Uii′ bi′ , (9.20)
a†j = Ujj
† †
′ bj ′ ,

H = b†j ′ Uj†′ j Aji Uii′ bi′ .

If the transformation diagonalizes the Hamiltonian, where the diagonalized Hamiltonian is
Bℓℓ′ = Bℓℓ δℓℓ′ ,
†
Bℓℓ′ = Uℓj Aji Uiiℓ′ , (9.21)
X
H = Bℓℓ b†ℓ bℓ .
ℓ

177
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

Thus, the same transformation used to diagonalize the matrix Aij can be used to transform the
operators to a basis where the Hamiltonian looks like a series of independent modes ℓ, where
each mode can have any number of particles, nℓ , with energies Eℓ (nℓ ) = nℓ Bℓℓ .
One should keep in mind that even if there are only a few single-particle levels, one still has
an infinite number of states because each mode ℓ can hold an arbitrary number of particles.
The ability of this formalism to include an arbitrary number of particles makes it the starting
point for any study of many-body physics. It also represents the starting point for the study of
relativistic physics due to the fact that even the vacuum might contain an arbitrary number of
virtual particle-antiparticle pairs.
The true challenges for many-body physics arise when the interaction is not that with an ex-
ternal potential but when the potential has explicit interactions between constituents.
R 3 3 ′ The in-
teraction term then has terms to higher powers in the field operators, e.g d rd r V (⃗ r −
r ′ )ψ † (⃗
⃗ r )ψ † (⃗
r ′ )ψ(⃗
r ′ )ψ(⃗
r ). The problem cannot then be solved by rearranging the single-particle
states into a basis where the Hamiltonian is diagonalized.
The most common example where particles are created or destroyed is in the radiative decay of
atomic or nuclear states. For example, a hydrogen atom in an excited state might decay to the
ground state via the emission of a photon. The photon must be created in the process. Creation of
a photon is surprisingly complicated due to the fact that the photon is a massless spin-1 particle
coupled to the current via the polarization vector. Because the algebraic gymnastics of polariza-
tion and spins adds additional complication, we will first consider the following example of a
decay through the emission of a scalar (spinless) particle.
Example 9.2: Decays Via Emission of a Created Particle, Radiation

Consider a hypothetical massive particle, of mass m, created by the field operator Ψ† and de-
stroyed by Ψ. It is coupled to operators for a massless particle which is created by the field op-
erator Φ† and destroyed by Φ. We assume both particles are spinless. The massive particle is
in the first excited state of a three-dimensional Harmonic oscillator characterized by frequency
ω0 . The first-excited state is three-fold degenerate, and in the Cartesian basis these states can
be written as a product of three one-dimensional wave functions, ψa (x)ψb (y)ψc (z). For this
problem, we assume the excited state applies to the z−dependence, i.e. a and b reference the
ground state and c refers to the first excited state. We wish to calculate the rate for decaying
to the ground state via the emission of the massless particle. The interaction term is expressed
as
Z
Hint = g d3 rΨ† (⃗ r ) Φ† (⃗


r ) + Φ(⃗
r ) Ψ(⃗r ),

which is manifestly Hermitian by noticing that Φ† (⃗ r ) + Φ(⃗

r ) is Hermitian. The factor g is
typically referred to as a coupling constant.
Using Fermi’s golden rule the decay rate is
2π X
Γ= |⟨nx , ny , nz = 0; kβ |Hint |nx , ny = 0, nz = 1⟩|2 δ(E1 − E0 − Ekβ ),
ℏ k
β

where E0 and E1 are the energies of the ground state energy and of the first excited state, and
Ekβ is the energy of the emitted particle. Because we are interested in the decay of the atom,

178
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

which creates a particle through Φ† , the destruction term, Φ, can be neglected. Also, note that
the initial state was assumed to be polarized along the z axis. This choice is arbitrary because
we are summing equally over all directions of k. If we were calculating a differential decay
rate, dΓ/dΩk , the angular dependence would depend on the initial polarization.
One can now calculate the matrix element,

⟨nx , ny , nz = 0; k|Hint |nx , ny = 0, nz = 1⟩

Z
= g d3 r⟨nx , ny , nz = 0; 0|ak Ψ† (⃗ r )(Φ† (⃗
r ) + Φ(⃗
r ))Ψ(⃗
r )|nx , ny = 0, nz = 1⟩,

where the state ⟨· · · ; k| was re-expressed as ⟨· · · ; 0|ak . Expanding the operator Φ† according
to Eq. (9.6), and using Eq. (9.12) to bring out the wave functions, Ψ(⃗ r )|ψ⟩ = ψ(⃗r )|0⟩,

⟨nx , ny , nz = 0; k|Hint |nx , ny = 0, nz = 1⟩

1
Z
⃗
= g d3 rψ0∗ (⃗r ) √ e−ik·⃗r ψnz =1 (⃗
r ).
V
Here, the harmonic oscillator wave functions can be written as a product of wave functions of
x, y and z respectively. If the normalized 1-d harmonic oscillator wave functions are denoted
ψn (x),

⟨nx , ny , nz = 0; k|Hint |nx , ny = 0, nz = 1⟩

r
g 2mω0 2
Z Z Z
−ikx x 2 −iky y 2 −ikz z
= √ dx e ψ0 (x) dy e ψ0 (y) dz e z ψ0 (z)
V ℏ
igkz α −k2 α2 /4
= √ e ,
2V
where α is the characteristic size of the ground state,
ℏ
α2 ≡ .
mω0
The wave function ψnz =1 could have been calculated by writing the raising operator for the
harmonic oscillator in terms of z and pz , then applying it to ψnz =0 .
Putting all this together,
2π X
Γ= |⟨nx , ny , nz = 0; k|Hint |nx , ny = 0, nz = 1⟩|2 δ(ℏkc − ℏω0 )
ℏ k
2
πg X 2 α2 /2
= kz2 e−k δ(ℏkc − ℏω0 )
mωV k
g2 1
Z Z
2 α2 /2
= 4
k dk d cos θ cos2 θ e−k δ(ℏkc − ℏω)
4mωπ −1
g 2 k3 2 α2 /2
= e−k ,
6πmℏc2
where k = ω/c, and ω is determined by energy conservation, ℏω = E1 − E0 .

179
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

9.5 Electromagnetic Decays

Electromagnetic decays are further complicated by the ⃗j · A ⃗ nature of the coupling. Remember
⃗ → (⃗
that minimal substitution, p ⃗
p − eA/c), results in a term in the Hamiltonian which looks
like  2 
e
Z
Hint = − d r ⃗j(⃗
3 ⃗ r )/c +
r ) · A(⃗ ⃗ r ) ρ(⃗
A(⃗ 2
r) , (9.22)
2mc2
representing the interaction with the vector potential. The last term will be neglected for now,
but plays a role in the quantum Hall effect. The current is given by the operator

⃗j(⃗ −ieℏ  †
r ) − [∇Ψ† (⃗

r) = Ψ (⃗
r )∇Ψ(⃗ r )]Ψ(⃗
r) , (9.23)
2m
where Ψ† and Ψ are field operators. Note that if sandwiched between the bra ⟨ψ| and the ket
|ψ⟩ one gets the usual expression for the current in terms of wave functions as Ψ(⃗
r ) → ψ(⃗
r)
†
and Ψ (⃗r ) → ψ(⃗r ).
First we must define the electromagnetic field operator in terms of creation and destruction
operators that make real photons,
s
2πℏ2 c2 X 1  ⃗ i⃗k·⃗r−iEk t/ℏ ⃗

⃗ r , t) =
A(⃗ √ ϵs (k)e
⃗ ak,s + ⃗ k)e−ik·⃗r+iEk t/ℏ a†k,s .
ϵ∗s (⃗ (9.24)
V k,s
Ek

Here s refers to the polarization (or spin) of the photon. Polarization vectors can be complex. For
example, for photons moving √ in the ẑ direction, the polarization vector for right-circularly po-
larized light is (x̂ + iŷ)/ 2. For each k there are two polarizations. Each must be perpendicular
to the direction of ⃗k, and normalized,

ϵ∗s (⃗
⃗ ϵs′ (⃗
k) · ⃗ k) = δss′ . (9.25)

⃗ looks peculiar for the complicated prefactors

Aside from√the polarization, the expression for A
and the 1/ Ek term inside the sum. However, all these factors are necessary to ensure that
Z ⃗2 + B
E ⃗2 X 
1

3 †
d r = Ek a⃗k,s a⃗k,s + , (9.26)
8π ⃗
2
k,s

⃗ are
where the electromagnetic fields resulting from A

⃗
1 ∂A
⃗ =−
E ,
c ∂t
⃗ = ∇ × A.
B ⃗

We are now in position to consider the general problem of electromagnetic decays, where a
particle of charge e changes from state i to state f while emitting a photon of momentum k and
polarization s. Outlining the steps to solving the problem:

180
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

1. Write down Fermi’s golden rule,

2π X 1
Z
Γ= ⃗
| ⟨f ; k, s| d3 r ⃗j(⃗ ⃗ r )|i⟩|2 δ(Ei − Ef − ℏkc).
r ) · A(⃗ (9.27)
ℏ ⃗ c
k,s

2. The operator A ⃗ operates on the photon degrees of freedom, while the current operator
acts only on the charged particle’s degrees of freedom. This allows the matrix element to
be written as a product of matrix elements, where one involves only the charged particle,
while the other captures the degrees of freedom of the photon,
1
Z Z
⃗
⟨f ; k, s| 3 ⃗ ⃗
r ) · A(⃗
d r j(⃗ r )|i⟩ = r )|i⟩ · ⟨⃗
d3 r ⟨f |J⃗(⃗ ⃗ r )/c|0⟩.
k|A(⃗ (9.28)
c
Using the definition of the electromagnetic field operator in Eq. (9.24) one can easily write
⟨⃗ ⃗ r )/c|0⟩ in terms of the plane wave function for the outgoing photon, and similarly,
k, s|A(⃗
one can write the current operator in terms of field operators to express ⟨f |J⃗(⃗
r )|i⟩ in terms
of the initial and final wave functions for the massive particle,
−ieℏ n ∗ o
⟨f |J⃗(⃗
r )|i⟩ = ψf (⃗
r )∇ψi (⃗ r ) − [∇ψf∗ (⃗
r )]ψi (⃗
r) , (9.29)
2m
s
2πℏ2 ∗ ⃗ −i⃗k·⃗r
⟨⃗ ⃗ r )/c|0⟩ =
k, s|A(⃗ ϵ (k)e
⃗ .
Ek V s

The overall matrix element then becomes

s
1 eℏ 2π ∗ ⃗
Z
⟨f ; ⃗
k, s| d3 r ⃗j(⃗ ⃗ r )|i⟩ =
r ) · A(⃗ ϵ (k) · M.
⃗ ⃗ (9.30)
c m Ek V s
ℏ
Z
⃗ ⃗
⃗
M(k, i, f ) ≡ d3 r e−ik·⃗r ψf∗ (⃗
r )∇ψi (⃗r)
2i
ℏ
Z
⃗
− d3 r e−ik·⃗r [∇ψf∗ (⃗r )]ψi (⃗
r ).
2i
Note M does not depend on the polarization, but does depend on the photon’s momen-
⃗
tum due to the factor e−ik·⃗r .
3. Change the sum over ⃗ k to an integral.
V V
X Z Z
3
→ 3
d k= 3
k2 dkdΩk (9.31)
⃗
(2π) (2π)
k

Then eliminate the delta function in Fermi’s golden rule by integrating over k,

2π 2πe2 ℏ2 V 1
Z X
Γ= k 2
dk dΩ k ϵ∗s (⃗
|⃗ ⃗ 2 δ(Ei − Ef − ℏck)
k) · M| (9.32)
ℏ V (2π) 3 Ek s
2
e k
Z X
= dΩk ϵ∗s (⃗
|⃗ ⃗ 2.
k) · M|
2πℏm2 c2 s

181
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

One can check the units of the above expression by noting that e2 k has dimensions of
energy (as it has the same dimensions as e2 /r) and that M has units of momentum which
are the same as the units of mc. Thus, the overall expression has units of energy over
ℏ which is an inverse time. Remember that in many books and tables Γ often refers to
energies rather than rates, in which case one erases the ℏ in the denominator.
4. To find the polarization for a specific direction, Ω, and for a specific polarization s, one can
rewrite the expression without integrating over the solid angle or summing over polariza-
tion,
dΓs e2 k 
 ∗ ⃗
2
= ⃗
ϵ
 s ( k) · ⃗
M  .

(9.33)
dΩk 2πℏm2 c2

⃗ with
If one wishes to sum over the polarization directions, one can replace M
⃗ =M
P ⃗ − k̂(k̂ · M).
⃗ (9.34)

This is the momentum operator with the component along k̂ projected away. The norm is
⃗ 2 = |M
|P| ⃗ |2 − |k̂ · M|
⃗ 2. (9.35)

The sum over s can now include the third polarization vector k̂, and

dΓ e2 k X
= |⃗ ⃗ 2
ϵ∗s · P| (9.36)
dΩk 2πℏm2 c2 s=1,2,3
2
e k
= ⃗ 2
|P|
2πℏm2 c2
e2 k n o
= ⃗ 2 ⃗ 2
|M| − |k̂ · M| .
2πℏm2 c2
Both polarization vectors, ⃗ϵ1 (⃗ ϵ2 (⃗
k) and ⃗ k) are orthogonal to the direction of the photon,
∗ ⃗
ϵs · P = ⃗
k̂, and ⃗ ∗ ⃗
ϵs · M for s = 1, 2. The third component is ⃗ ϵ3 = k̂ and is projected
away in the dot product ⃗ ∗ ⃗
ϵ · P. Thus, adding the third component does not change the
answer. Further, because P ⃗ is orthogonal to k̂, the polarization vector of the photon is
indeed perpendicular to ⃗ k. Thus, knowing the directions of M ⃗ and k̂ sets the polarization
of the outgoing light to be
⃗ P|.
ϵ(k̂) = P/|
⃗ ⃗ (9.37)

⃗ then P
Note that if k̂ is parallel to M, ⃗ = 0 and the intensity vanishes in that direction.
5. Finally, to obtain the total rate Γ one must integrate over all directions of the photon. In
⃗
general, this can be difficult, but it simplifies if can make the approximation, e−ik·⃗r ≈ 1
inside the integral for calculating M.⃗ This approximation is known as the dipole approx-
imation and is discussed ahead. In that case M ⃗ does not depend on k̂, which greatly
simplifies the integral. To integrate the first term, one can simply multiply dΓ/dΩk by 4π.

182
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

However, the second term depends on k̂. For any fixed vector V ⃗ , the integral of |k̂ · V
⃗ |2
⃗ |2 . Thus the integral over dΩk is
over all directions of k̂ is one third of the integral of |V
2/3 of the integral if the second term were neglected, and the integrated decay rate in the
dipole approximation is

4e2 k
Γ= ⃗ 2.
|M| (9.38)
3ℏm2 c2

9.6 The Dipole Approximation

For nuclear examples, typical γ energies are on the order of one MeV, which corresponds to
wavelengths of a few hundred fm, approximately 100 times the size of a nucleus. For atomic
examples, emitted photons have energies on the order of eV, or equivalently wavelengths of
hundreds of nm, which is thousands of times the size of a typical atom. Thus, one might consider
approximating the phase factor,
⃗
e−ik·⃗r ≈ 1. (9.39)
This allows us to approximate the matrix element M,⃗ defined in Eq. (9.30) as

⃗ ≈ ⟨f |P
M ⃗ |i⟩ (9.40)
im
= ⟨f |[H0 , ⃗
r ]|i⟩
ℏ
im(Ef − Ei )
= ⟨f |⃗
r |i⟩.
ℏ
This is known as the dipole approximation because the matrix element is proportional to the
dipole operator, e⃗r , sandwiched between the initial and final states. The second line in Eq.
(9.40) required commuting H0 = P ⃗ 2 /2m with ⃗ r , using the fact that
X
⃗ |2 , ri ] =
[|P [Pj2 , ri ] (9.41)
j
X
= (Pj2 ri − ri Pj2 )
j
X
= (Pj (ri Pj + [Pj , ri ]) − ri Pj2 )
j
X
= (Pj (ri Pj − iℏδij ) − ri Pj2 )
j
X
= −iℏPi + (ri Pj2 + [Pj , ri ]Pj − ri Pj2 )
j

= −2iℏPi ,

⃗ 2, ⃗
or [P ⃗.
r ] = −2iℏP
The name “”dipole approximation” is somewhat misleading because the only approximation is
that kr is small, or equivalently that the wavelengths are long compared to the size of the emit-
ting region. A more reasonable name might be the “long wavelength approximation”. However,

183
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

it is the approximation, eikr → 1, that enabled rewriting the gradient operator as the coordinate
operator multiplied by the energy difference of the initial and final states.
One consequence of dipole radiation is the dipole sum rule. Using the expression for the matrix
element in Eq. (9.40),
⃗ 2 ℏ2
|M|
2
|⟨f |⃗
r |i⟩| = . (9.42)
m2 (Ef − Ei )2
Rearranging Eq. (9.32),

2πℏm2 c2
Z X
dΩk ϵ∗s
|⃗ ⃗ 2 = Γi→f
· M| , (9.43)
s
e2 k
8π
= ⃗ 2.
|M|
3
The last step introduced a factor of 2/3 from averaging over the directions of ⃗ k, using the fact
∗ ⃗ ⃗
ϵs (k) is orthogonal to k. Combining these last two expressions, and replacing k with (Ei −
that ⃗
Ef )/ℏc,

3ℏ4 c3
r |i⟩|2 =
|⟨f |⃗ Γi→f . (9.44)
4e2 (Ei − Ef )3
Finally, if one sums over all states i, one can use completeness to reduce the l.h.s. to

2
X 3ℏ4 c3
⟨f |r |f ⟩ = Γi→f . (9.45)
i
4e(Ei − Ef )3

This is the dipole sum rule, and it relates the charge radius squared to a weighted sum over rates
from all initial states i to some final state f . Although the derivation suggests that emission is
through a single charge e, it can be derived more generally so that the l.h.s. is the charge radius
squared even for systems with charge spread over multiple particles with complicated wave
functions. For objects composed of multiple charges, collective states can dominate the sum on
the right hand side. When there are many charges that add to Z, and if those charges move
together, i.e. collectively, they emit proportional to Z 2 e2 , as opposed to single-particle motion
which emits proportional to e2 . For example, a giant dipole resonance in 208 Pb nuclei involves
the 82 protons oscillating together, while the 126 neutrons oscillate in the opposite direction.
Such a state absorbs and emits photons readily.
Example 9.3: Electromagnetic Decay of the Harmonic Oscillator
Consider a harmonic oscillator in the first excited state, where in Cartesian coordinates the
wave function is excited in the ẑ direction. The wave function is then
ψ1 (⃗
r ) = ϕ0 (x)ϕ0 (y)ϕ1 (z),
where ϕ0 and ϕ1 represent solutions to the one-dimensional harmonic oscillator.
1. Within the dipole approximation, calculate the rate per solid angle for decays of unpo-
larized photons.

184
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

Solution:
⃗
Using Eq.s (9.30) and (9.32) with e−ik·⃗r = 1„
e2 k
Z X
Γ= dΩk |⃗ ⃗ 2,
ϵ∗s · M|
2πℏm2 c2 s
ℏ
Z
⃗ ⃗
M(k, i, f ) = d3 r ψf∗ (⃗
r ) ∇ψi (⃗
r ).
i
Next, we apply the dipole approximation, Eq. (9.40), using the fact that Ei − Ef = ℏω.
Z
⃗ ⃗
M( k, i, f ) = imω d3 r ψf∗ (⃗r )⃗
r ψi (⃗
r ).

Using the fact that the original state is excited along the ẑ direction, one can write the
matrix element as a product of three integrals,
Z Z
∗
3
d r ψf (⃗r )⃗
r (⃗
r) = dxdydz |ϕ0 (x)|2 |ϕ0 (y)|2 |ϕ∗0 (z)ϕ1 (z)(xx̂ + y ŷ + z ẑ).
(9.46)
The terms with x̂ and ŷ vanish by symmetry, and for the integral over z, the integrals
over x and y become unity by the normalization of ϕ0 . This leaves a one-dimensional
integral,
Z
⃗ ⃗
M(k, i, f ) = imω ẑ dz ϕ∗0 (z)zϕi (z).

Using solutions for the harmonic oscillator wave functions in Sec. (2.6) for the ground
state and the excited states of the harmonic oscillator,
⃗ ⃗
M(k, i, f ) = i(ℏmω/2)1/2 ẑ.
Inserting into the expression for Γ in Eq. (9.38),
4e2 k ℏmω
Γ=
3ℏm2 c2 2
2e2 ω 2
= .
3mc3
2. What is the angular distribution of photons? – summed over polarizations
Solution:
One can apply Eq. (9.36),
dΓ e2 k n o
= ⃗ 2 − |k̂ · M|
|M| ⃗ 2 ,
dΩ 2πℏm2 c2
e2 ω 2 n 2
o
= 1 − ( k̂ · ẑ)
4πmc3
e2 ω 2
= 3
sin2 θk .
4πmc

185
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

There are no photons emitted in the ẑ direction because the matrix element M ⃗ is in
the ẑ direction and because the polarizations for a photon in the ẑ direction must be
perpendicular to ẑ when k̂||ẑ.

3. What is the polarization of a photon emitted in the x̂ direction?

Solution:
By inspection, one can see that ϵ̂ must be parallel to ẑ. Thus, the photon would be plane
polarized along the z direction.

9.7 Magnetic Dipole and Electric Quadrupole Radiation

The matrix element may be zero in the dipole approximation due to symmetry considerations,
usually parity constraints. In that case, one should keep the next order term in the expansion of
⃗
the factor e−ik·⃗r ≈ 1 − i⃗ k·⃗ r + · · · in Eq. (9.30). One then obtains terms in the matrix element
that look like
1n ⃗ o 1n o
(⃗
k·⃗ ⃗ ·⃗
r )(P ϵ∗ ) = (k · ⃗ ⃗ ·⃗
r )(P ϵ∗ ) − (⃗ϵ∗ · ⃗r )(⃗
k·P⃗) + (⃗
k·⃗r )(P⃗ ·⃗
ϵ∗ ) + (⃗
ϵ∗ · ⃗r )(⃗
k·P ⃗)
2 2
(9.47)
1 ⃗
n o 1 n o
= (k · ⃗
r )(P⃗ ·⃗ϵ∗ ) − (⃗ r )(⃗
ϵ∗ · ⃗ k·P ⃗) + ⃗ ·⃗
(P ϵ∗ )(⃗
k·⃗ r )(⃗
ϵ∗ · ⃗
r ) + (⃗ k·P ⃗) .
2 2

Here, the momentum operator P ⃗ = −iℏ∇ acts on the operators of the charged particle, and
should not be confused with ℏ⃗ k, the momentum of the photon. The last step, which involved
⃗
flipping the (k · ⃗ ⃗
r ) and (P · ⃗ ⃗
ϵ∗ ) factors in the second bracket was justified by the fact that P
commuting with ⃗ k·⃗ r yields a result proportional to ⃗
k which is in turn orthogonal to ⃗
ϵ.
Using a vector identity the first term in Eq. (9.47) can be written

1n ⃗ o
(k · ⃗ ⃗ ·⃗
r )(P ϵ∗ ) − (⃗ r )(⃗
ϵ∗ · ⃗ ⃗ ) = (⃗
k·P ⃗ ) · (⃗
r×P ϵ∗ ).
k ×⃗ (9.48)
2

This operator looks like L ϵ∗ × ⃗

⃗ · (⃗ k)/2, thus it looks like dot product of the magnetic dipole
⃗ ⃗
operator, eL/2m, and k × ⃗ ∗
ϵ . Remembering that A ⃗ is defined by the direction of ϵ, and that
⃗
k×A ⃗ is generated by ∇ × A, ⃗ one can think of this term as L⃗ × B,
⃗ which motivates the term
magnetic dipole.
Using the identity
m
⃗ =
P [⃗
r , H0 ], (9.49)
iℏ

186
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

the second term in Eq. (9.47) can be written as

1 n o
⟨f | (⃗
k·⃗ ⃗ ·⃗
r )(P ϵ∗ ) + (⃗ r )(⃗
ϵ∗ · ⃗ ⃗ ) |i⟩
k·P (9.50)
2
m
= (Ei − Ef )⟨f |(⃗ r ) · (⃗
ϵ∗ · ⃗ k·⃗
r )|i⟩
iℏ
m
= (Ei − Ef )ϵ∗i kj Qij ,
iℏ
1
Qij ≡ ⟨f |ri rj |i⟩ − δij ⟨f |r 2 |i⟩.
3
The final term in the quadrupole operator Qij which is proportional to the delta function does
ϵ and ⃗
not contribute because ⃗ k are orthogonal. The motivation for labeling this term “quadrupole”
comes from the fact that all such operators can be written in terms of Yℓ=2,m s.
Both the magnetic dipole term and the electric quadrupole term are linear when expanding in
⃗
k. They are smaller than the electric dipole piece only because kr is small. It is important to
keep in mind that radiation calculations are simultaneously built on two expansions. The first is
the perturbative assumption that e2 is small. Higher order perturbation theory would involve
multiple photons. Even if the final state has only a single photon, intermediate states appear in
higher-order perturbation theory (higher powers of e2 ) and such states might include a photon.
Such considerations are typically emphasized in a course on perturbative field theory. For ex-
ample, the vertex correction for the electron’s spin gyromagnetic factor, g − 2, or the running
coupling constant in QCD are examples of where higher-order perturbation theory comes into
play. In addition to assuming higher orders in e2 are negligible, radiative calculations also ex-
ploit the fact that kr is usually quite small. This second assumption is the foundation of the
dipole-quadrupole· · · paradigm.
The symmetry constraints of the radiation pattern will be a major point of discussion later in the
course when we discuss the Wigner-Eckart theorem.

9.8 Exercises
1. As part of an elaborate calculation using Fermi’s Golden rule, you find yourself needing to
calculate the following matrix element squared

|Mf i |2 = |⟨f |Ψ†A (⃗ r )|i⟩|2 .

r )ΨB (⃗

The initial state i is composed of NB particles of type B, which are all in the same single-
particle state of momentum ⃗ kB . The final state |f ⟩ is composed of NB − 1 particles of
type B, in the same level ⃗kB , along with one particle of type A with momentum ⃗ kA . The
momentum states are defined within some large volume V .

(a) Find |Mf i |2 . The momentum states are defined within some large volume V .
(b) Repeat but in this case assume the NB particles are all in different momentum states,
⃗
kn , n = 1 · · · NB , with the same values in the final state, with the exception of ⃗
k1 ,
which is missing.

187
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

2. Consider b-particles of mass m confined by a one-dimensional harmonic oscillator poten-

tial characterized by a frequency ω. The b particles interact with massless and spinless
a-particles through their respective field operators,
Z
Hint = g dxΨ† (x)Φ(x)Ψ(x),

where Φ and Ψ are the field operators for the a-particles and b-particles respectively. As-
sume the b particles are sufficiently heavy to ignore their recoil energy.
1 X 1  −ikx † 
Φ(x) = √ √ e ak + eikx ak
L k Ek
1 X −ikx †
Ψ† (x) = √ e bk ,
L k

(a) What is the dimension of g?

(b) What is the decay rate of a b particle in the first excited state.

3. Show that Eq. (9.26) is satisfied by using the electric and magnetic fields defined in Eq.
(9.24). Note: After squaring E ⃗ and B,
⃗ ignore any cross terms when you involving rapid
oscillations in time, i.e. those that behave as e±2iEk t/ℏ
From notes, the 2 equations are
Z ⃗2 + B
E ⃗2 X 
1

3 †
d r = Ek a⃗k,s a⃗k,s + , (9.51)
8π ⃗
2
k,s

and
s
2πℏ2 c2 X 1  ⃗ i⃗k·⃗r−iEk t/ℏ ⃗

⃗ r , t) =
A(⃗ √ ϵs (k)e
⃗ ak,s + ⃗ k)e−ik·⃗r+iEk t/ℏ a†k,s . (9.52)
ϵ∗s (⃗
V k,s
Ek

4. A proton in a nucleus decays from an excited state to its ground state by emitting a photon
of momentum ℏ⃗ k and polarization ⃗
ϵs . The matrix element describing the decay is
⃗
e−ik·⃗r
Z
3
ϕ∗0 (⃗ r ) − [∇ϕ∗0 (⃗


⟨0, k, s|V |1⟩ = β⃗
ϵs · d r √ r )∇ϕ1 (⃗ r )]ϕ1 (⃗
r) .
V
The factor β absorbed all the various factors involved in defining the vector field in Eq.
(9.24). Assume the ground and excited states are modeled with a three-dimensional har-
monic oscillator of frequency ω. If the excited state is in the first level of a harmonic oscil-
lator and has an angular momentum projection m, what is the shape (θϕ dependence) of
the angular distribution of the photons, dΓ/dΩ, for each m. Assume that the wavelength
⃗
of the photon is sufficiently long that the phase eik·⃗r ≈ 1. Remember that the two polar-
izations of the photon must be perpendicular to ⃗ k. You need only calculate the angular
shape of the distribution – ignore the prefactors.

188
PHY 851/852 9 SECOND QUANTIZATION AND RADIATION

Some help: The first excited state of the harmonic oscillator is three-fold degenerate. In
the Cartesian basis these have the form ∼ xϕ0 , yϕ0 and zϕ0 , where ϕ0 is the ground-state
wave function. These can be mapped to three states that are eigenstates of angular mo-
mentum, ℓ = 1; m = 1, 0, −1 as discussed in Chapter 4. The wave functions of states
with m = ±1 have to have an angular dependence given by Y1±1 ∼ sin θe±iϕ , whereas
the wavefunction for m = 0 has to be proportional to Y10 ∼ cos θ. Using the fact that
r cos θ = z and r sin θe±iϕ = x ± iy, the m = ±1 wave functions are proportional to
x ± iy, whereas the m = 0 wave is proportional to z.
5. A spinless particle of mass M and charge e is in the first excited state of a three-dimensional
harmonic oscillator characterized by a frequency ω. Assume the particle is in the Cartesian
state of a harmonic oscillator with nz = 1, i.e. m = 0. Using the interaction

Hint = ⃗j · A/c,
⃗

(a) Calculate the decay rate of the charged particle into the ground state of the oscillator
in the dipole approximation.
(b) Calculate dΓ/dΩ as a function of the emission angles of the photon, θ and ϕ.
(c) In terms of the unit vectors k̂, θ̂ and ϕ̂, the two polarization vectors which are allowed
for emission of a photon at an angle θ, ϕ are θ̂ and ϕ̂. For each of these two polariza-
tion vectors, calculate dΓs /dΩ, the probability of decaying via emission of a photon
emitted in the θ, ϕ direction with polarization s.

6. Again consider a spinless particle of mass M and charge e n the first excited state of a
three-dimensional harmonic oscillator characterized by a frequency ω. However, this time
assume the charged particle is originally in a state with angular momentum projection
m = +1 along the z axis. Using the interaction

Hint = ⃗j · A/c,
⃗

and applying the dipole approximation,

(a) Find the decay rate Γ of the first excited state.

(b) Find the differential decay rate dΓ/dΩ.
(c) Describe the polarization of a photon emitted in the x̂ direction.
(d) Describe the polarization vector of a particle emitted in the ẑ direction.

189
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

10 Advanced Topics in Angular Momentum

10.1 Isospin
The angular momentum algebra developed last semester has a broad range of applications,
many of which have nothing to do with rotations in coordinate space but instead deal with
symmetries in some other context. One such example is isospin. In nuclear and particle physics
isospin mimics the spin mechanics used to describe spin 1/2 particles. In quark language, the
up quark is an I = 1/2, I3 = 1/2 particle while the down quark is an I = 1/2, I3 = −1/2
particle. Here, the total isospin is I, which is analogous to the total angular momentum J , and
the isospin projection is I3 , which is analogous to the spin projection M . In nuclear physics the
proton and neutron form an iso-doublet with the proton being the I3 = 1/2 particle. The strong
interaction is invariant to “rotations” in isospin space. In Chapter 4 we showed that rotations of
⃗˙
angle θ about axis in the n̂ direction were generated with the operator e−iLˆnθ/ℏ . For spin-1/2
particles, the two components states were rotated by the operator e−i⃗σ·n̂θ/2 . If one expresses the
proton and neutron as two-component states, the transformation e−i⃗σ·n̂θ/2 can be thought of as
a “rotation” in isospin space. One can combine particles with a given isospin to form states with
a specific net isospin in the same way that angular momentum are coupled. This means that the
symmetry is analogous to the rotational symmetries discussed with angular momentum, and if
the rotational symmetry in isospin is conserved (sometimes referred to as being a “good” sym-
metry) then the total isospin of the system is conserved. For instance, composite particles (those
made of several quarks) have good isospin, which is determined by adding the isospin of sev-
eral particles. Three up or down quarks can be coupled to either I = 1/2 states, the proton or
neutron, or to an I = 3/2 state, the ∆ baryon, in the same way three spin-half particles can be
coupled to either J = 3/2 or J = 1/2 multiplets. The ∆ baryon multiplet, which has I = 3/2,
consists of four states with isospin projection I3 = −3/2, −1/2, 1/2 and 3/2. Because the
down quark has electric charge −1/3 and the up quark has electric charge +2/3, the four ∆s
are ∆− ∆0 , ∆+ and ∆++ .
If one considers two nucleons (a nucleon is a proton or a neutron), and assume perfect isospin
symmetry, one would assign the pair a total isospin of I = 0 or I = 1, with projections
I3 = −1, 0, 1 for I = 1. This is the same as listing all the ways to combine two spin 1/2 parti-
cles, which combine into a triplet with S = 1 or a singlet with S = 0. The state with I = 1 and
I3 = 1 is made of two protons, and the state with I = 1, I3 = −1 is made of two neutrons. The
state with I = 1, I3 = 0 is comprised of one proton and one neutron in a symmetric fashion. By
symmetric, the two particle wave function would look like Ψ(x1 , p; x2 , n) = Ψ(x1 , n; x2 , p).
One would find this state by taking the two-proton state, then applying the isospin lowering
operator, which would result in a symmetric state. The I = 0, I3 = 0 state would be anti-
symmetric. Because of the symmetry, one would expect the three I = 1 states to behave simi-
larly, while the I = 0 state might be different. Thus, the deuteron, which is made of a proton
and a neutron has an I = 0 bound state, while the three I = 1 combinations have no bound
states. Because the deutreron’s isospin wave function is anti-symmetric, and because nucleons
are spin-half fermions, the overall wave function must be anti-symmetric. This means that the
deuteron has a symmetric spin wave function, and has S = 1. The three I = 1 states then have
an anti-symmetric spin wave function, with S = 0. The pn combination thus has a bound state
with I = 0, S = 1, and resonant behavior just above threshold in the S = 0 channel. This

190
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

resonant state is part of an I = 1 iso-triplet, i.e. the same S = 0 resonance appears in the pp
and nn systems.
Isospin is an approximate symmetry. The symmetry is broken because the up and down quarks
have slightly different masses, and because of the Coulomb interaction, which is obviously dif-
ferent for up quarks (charge 2/3) and down quarks (charge -1/3). Thus, the proton and neutron
masses, 938.27 and 939.57 MeV/c2 respectively, are not identical.
Example 10.1: The ρ meson
Consider the ρ+,0,− mesons (they have charge +1, 0, −1), which form an isotriplet (I = 1)
and can decay into two pions. Pions, π +,0,− , also form an iso-triplet. Find the decay branches
for each of the three ρ → 2π decays.
Solution: Because isospin is conserved in the strong interaction, and because the ρ0 has total
isospin I = 1 and projection I3 = 0, the two-pion state must also have the same total isospin
and projection. First, look at the ρ+ or ρ− . The only branchings for each decay are

ρ+ → π + π 0 , ρ− → π − π 0 .

However, the ρ0 might go into one of two branches

ρ0 → π + π − or π 0 π 0 .

To solve for the relative strengths of the two branches, find how to add two pion isospins to
form an I = 1, I3 = 0 state. Start by writing the I = 2, I3 = 2 state

|I = 2, I3 = 2⟩ = |π + , π + ⟩,

then use the lowering operator to find the I = 2, I3 = 1 state. The lowering operator behaves
the same as for angular momentum,
p
I− |I, I3 ⟩ = I(I + 1) − I3 (I3 − 1)|I, I3 − 1⟩.

Thus,
1 1
|I = 2, I3 = 1⟩ = √ |π 0 , π + ⟩ + √ |π + , π 0 ⟩.
2 2
By orthogonality,
1 1
|I = 1, I3 = 1⟩ = √ |π + , π 0 ⟩ − √ |π 0 , π + ⟩
2 2
This combination corresponds to the decay of the ρ+ meson. By applying the lowering oper-
ator to the |I = 1, I3 = 1⟩ state,
p
p 1 · (1 + 1) + 0 · (0 − 1) − +
1 · (1 + 1) − 1 · (1 − 1)|I = 1, I3 = 0⟩ = |π π ⟩
p 2
1 · (1 + 1) + 0 · (0 − 1) + −
− |π π ⟩,
2
1 1
|I = 1, I3 = 0⟩ = √ |π − π + ⟩ − √ |π + π − ⟩
2 2

191
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

The I = 1, I3 = 0 combination refers to the ρ0 and has no π 0 π 0 contribution. Thus, the two
pions must both be charged if they couple to the ρ0 .
Also, the fact that the two terms appear identical, except for the switching of the π + and π −
tells us that the two final spatial (momentum) states of the pions must be anti-symmetric to
cancel the anti-symmetry of the isospin wave function. This is accomplished by requiring
the ρ to decay into a p wave. Note that the ρ+ and ρ− isospin wave functions were also
anti-symmetric.

Angular momentum algebra applies to any system where a “rotation”, e.g. eiσi ϕ/2 , leaves the
Lagrangian unchanged. Isospin is not an “exact” symmetry. Although the strong interaction
conserves isospin, the electromagnetic interaction manifestly violates isospin. Thus, the ρ0 does
not decay with 100% probability into two charged pions, but has a very small probability to
decay into π0 π0 that is not even listed in the particle data book, http://pdg.lbl.gov).
Conservation of isospin plays a large role in nuclear physics, where the proton and neutron form
the basis. For example, 12 C has six protons and six neutrons, thus any state of 12 C has I3 = 0.
However, the states of 12 C might have a total isospin, I = 0, 1, 2, · · · 6. The nuclei 12 N and
12
B each have I3 = 1 so they cannot have I = 0. The ground state of 12 C has I = 0 and has
no analog in Boron or Nitrogen, but there does exist an excited state of 12 C with I = 1 which
has very similar properties to the ground states of 12 N and 12 B. Sometimes this is referred to as
“mirror” symmetry.
Finally, we should mention that sometimes symmetries involve more than just the rotation of
two states into one another. For instance, although the strange quark is much more massive
than the up and down quarks, one can consider the three quarks to form a basis. If one ignored
the quark masses, this basis has a greater symmetry as it involves rotations among three con-
stituents. The three-components symmetry is SU(3), rather than the SU(2) symmetry discussed
before. In SU(3) the mechanics are more complicated compared to the SU(2) mechanics used to
study angular momentum or isospin.

10.2 Combining Three Angular Momenta

When we studied two angular momenta, we described the “addition” as a change of basis where
the state with labels |jj , j1 , m1 , m2 ⟩ was written as a linear combination of states with labels
|j1 , j2 , J, M ⟩. When adding three angular momenta the change of basis is

|j1 , j2 , j3 , m1 , m2 , m3 ⟩ ↔ |(j1 , j2 ), j3 , J12 , J, M ⟩. (10.1)

Again, M is the projection of the total angular momentum J with values from −J to J .
In order to couple the three angular momenta j1 and j2 were first coupled to J12 before j3 and
J12 were coupled to J . Thus, J12 survives as a quantum number, which is necessary as the
original state had six labels, which requires six labels for the final state.
The change of basis can be described in terms of Clebsch-Gordan coefficients. First we describe

192
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

an intermediate state with j1 and j2 coupled to J12 .

X
|j1 , j2 , J12 , j3 , M12 , m3 ⟩ = |j1 , j2 , j3 , m1 , m2 , m3 ⟩⟨j1 , j2 , j3 , m1 , m2 , m3 | (10.2)
m1 ,m2

|j1 , j2 , J12 , j3 , M12 , m3 ⟩

X
= ⟨j1 , m1 , j2 , m2 |J12 M12 ⟩|j1 , j2 , j3 , m1 , m2 , m3 ⟩.
m1 ,m2

The first line above is simply a statement of completeness. To write the second line one uses the
fact that the labels j3 , m3 in the matrix element ⟨j1 , j2 , j3 , m1 , m2 , m3 |j1 , j2 , J12 , j3 , M12 , m3 ⟩
are unaffected by this change of basis, and can be neglected. The matrix element can then be
replaced by a Clebsch-Gordan coefficient. One can now consider the states with total angular
momentum J created by coupling J12 to j3 .
X
|(j1 , j2 ), j3 , J12 , J, M ⟩ = |j1 , j2 , j3 , J12 , M12 , m3 ⟩⟨j1 , j2 , j3 , J12 , M12 , m3 | (10.3)
M12 ,m3

|(j1 , j2 ), j3 , J12 , J, M ⟩
X
= ⟨J12 , M12 , j3 , m3 |J M ⟩⟨j1 , m1 , j2 , m2 |J12 M12 ⟩
m1 ,m2 ,m3 ,M12

|j1 , j2 , j3 , m1 , m2 , m3 ⟩.

The choice of coupling j1 with j2 to J12 was arbitrary, as one might have instead chosen to
couple j1 to j3 or j2 to j3 . States in the |j1 , j2 , j3 , J, J12 , M ⟩ basis can be expressed in terms of
states in the |j1 , j2 , j3 , J, J23 , M ⟩ basis This change of basis which can be represented by Racah
coefficients W .

⟨(j1 , j2 ), j3 , J12 , J, M |j1 , (j2 , j3 ), J23 , J, M ⟩ (10.4)

p
= δM,M ′ (2J12 + 1)(2J23 + 1)W (j1 , j2 , j3 , J ; J12 , J23 )
 
j1 +j2 +j3 +J
p j1 j2 J12
= δM,M ′ (−1) (2J12 + 1)(2J23 + 1) ,
j3 J J23

where the quantity in brackets is known as the Wigner 6 − j symbol, which is simply another
incarnation of the Racah coefficient. As a homework problem you will be asked to express the
Racah coefficient W in terms of Clebsch-Gordan coefficients.

10.3 Notation: Conventions for Clebsch-Gordan Coefficients and Wigner 3j

or 6j Symbols
Unfortunately, there is no overwhelmingly accepted convention for writing Clebsh-Gordan co-
efficients. In order to stress that they are simply matrix elements, we have chosen to use the
notation,

⟨j1 , m1 , j2 , m2 |J M ⟩ = ⟨j1 , m1 , j2 , m2 |j1 , j2 , J, M ⟩. (10.5)

193
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

The indices j1 and j2 are not affected by the change of basis, thus they are typically not repeated
in the ket above. By convention, the states |J M ⟩ are defined in such a way that the Clebsch-
Gordan coefficients are real. Thus, the labels in the bra and the ket can be reversed. Sometimes
semicolons are inserted between labels, and sometimes the commas above are left out. It is good
practice to choose labels carefully so that is clear which projections (ms) are associated with
which total angular momenta (js).

⟨j1 , m1 , j2 , m2 |j3 m3 ⟩ = Cjj13,m

m3
1 ,m2 ,m2
(10.6)
 
p j1 j2 j3
= (−1)−j1 +j2 −m3 2j3 + 1 .
m1 m2 −m3
These latter notations mainly appeal to those with masochistic tendencies, and are often seen in
older literature or textbooks.
Aside from the convention that Clebsch-Gordan coefficients are real, there is also a choice in
sign. For example, when generating Clebsch-Gordan coefficients by using ladder operators, one
first considered the highest possible total angular momentum J and highest projection M , then
used lowering operators to generate those with lower M but the same J . Then to find those
with J lowered by one, the matrix element was found by the orthogonality constraint. But this
next element could have arbitrary sign. Thus, an additional convention is applied
⟨j1 , m1 = j1 , j2 , m2 = J − j1 |J J ⟩ > 0. (10.7)
The coefficients have a number of symmetry properties:
⟨j1 m1 j2 m2 |J M ⟩ = (−1)j1 +j2 −J ⟨j1 , −m1 , j2 , −m2 |J, −M ⟩ (10.8)
= (−1)j1 +j2 −J ⟨j2 , m2 , j1 , m1 |J M ⟩
s
2J + 1
= (−1)j1 −m1 ⟨j1 , m1 , J, −M |j2 , −m2 ⟩
2j2 + 1
s
2J + 1
= (−1)j2 +m2 ⟨J, −M, j2 , m2 |j1 , −m1 ⟩
2j1 + 1
s
2J + 1
= (−1)j1 −m1 ⟨J, M, j1 , −m1 |j2 , m2 ⟩
2j2 + 1
s
2J + 1
= (−1)j2 +m2 ⟨j2 , −m2 , J, M |j1 , m1 ⟩.
2j1 + 1

10.4 Irreducible Tensor Operators

Often physics involves calculating transition elements of the form
⟨J ′ , M ′ |Tqk |J, M ⟩, (10.9)
where the labels in the kets and bras denote the angular momentum of the initial and final
states and Tqk is an operator that is part of a set of (2k + 1) operators, T−k
k k
, T−k+1 · · · Tkk that

194
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

transform like spherical harmonics Yk,q under rotation. To explain what that entails, we first
write the spherical harmonics,
1
Y0,0 = √ , (10.10)
4π
r
3
Y1,0 = cos θ,
4π
r
3
Y1,±1 = ∓ sin θe±iϕ ,
r 8π
5
Y2,0 = (3 cos2 θ − 1),
16π
r
15
Y2,±1 = ∓ sin θ cos θe±iϕ ,
r 8π
15
Y2,±2 = sin2 θe±2iϕ ,
32π
∗
Yℓ−m (θ, ϕ) = (−1)m Yℓm (θ, ϕ).
For example, one set of operators that transforms like the three Yℓ=1,m spherical harmonics is
−x − iy, (10.11)
x − iy,
z.
p
By inspection, one can see that if these three functions were all multiplied by (1/r) 3/(4π),
which is invariant under rotations, the three Yℓ=1 spherical harmonics would result. In fact,
one could multiply the three functions above by any scalar, and it will still result in a perfectly
well defined set of irreducible tensor operators. Examples of scalars are constants, such as π, or
⃗ ·p
any operator that is invariant under rotation, e.g. L ⃗. The important criteria is that the three
operators transform amongst each other under rotation the same as the three Yℓ=1,m s. Another
perfectly fine set is
√
−(Lx + iLy )/ 2, (10.12)
√
(Lx − iLy )/ 2,
Lz .
Any operator with a vector index, e.g. Pz , infers the existence of two other operators Px and Py ,
where rotations transform the 3 operators amongst one another. Similarly, any operator with
indices with angular momentum indices, e.g. k and q, infers the existence of the remainder of
the set of 2k + 1 operators where −k ≤ m ≤ k. This set of 2k + 1 operators is the irreducible
set.
Such operators are known as irreducible tensor operators. Examples are

1. Lz , which transforms like Y1,0 . Three operators are

1
√
T±1 = − ∓ (Lx ± iLy )/ 2,
T01 = Lz .

195
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

2. x, which transforms as Y1,−1 − Y1,1 . The operators are

1
√
T∓1 = ∓(x ± iy)/ 2,
T01 = z,
1
√
x = (T−1 − T12 )/ 2

3. r 2 , which transforms as Y0,0 , i.e you can define some an operator T00 = r 2 .
4. Px Py which transform as Y2,2 − Y2,−2 . The set of operators is
2
T±2 = (Px + iPy )2 ,
2
T±1 = ±2Pz (Px ± iPy ),
√
T02 = (2Pz2 − Px2 − Py2 ) 2,
Px Py = −i(T22 − T−2
2
)/4.

One can express any analytic function of x, y and z as a sum of irreducible tensor operators. But
not all functions can be expressed in terms of operators with a single value of k. For example,
writing x2 + y 2 = (2/3)(x2 + y 2 + z 2 ) − (2z 2 − x2 − y 2 )/3 requires both T k=0 and T k=2
pieces. For the examples above, one can scale any of the sets of 2k + 1 operators by any scalar
function, e.g. multiply by a constant or by r 2 , and it is still a set of irreducible tensor operators.
However, all the terms in the set must be scaled similarly to preserve the way in which the rotate
into one another. The term “irreducible tensor operator” refers to the fact that under rotations
the operators mix only amongst irreducible subsets,

α)Tqk R−1 (⃗
R(⃗ k
α) = Dq,q α)Tqk′ .
′ (⃗ (10.13)
k k
Here, the matrices Dq,q ′ describe the linear transformation of the 2k + 1 elements of T . The

rotations mix only the 2k+1 operators with the same k but different q. Specifically, “irreducible”
refers to the fact that rotations can mix in any of the different q components. For example, if one
considered the set of six operators pi pj (six and not nine because the matrix is symmetric) one
could express any of the six operators in terms of the five ℓ = 2 operators, plus one ℓ = 0
operator, p2 = p ⃗·p⃗. The full set of six operators would be reducible, because the single ℓ = 0
operator does not mix with the other five during rotations. But, the set of five that transform as
the Ymℓ=2 s would indeed be an irreducible set.
As stated above the form of the spherical harmonics can be used to express polynomials of x, y
and z in terms of irreducible tensor operators. To that end, a list of the spherical harmonics and
associated polynomials to second order is presented here.
For k = 0, 1, 2 one can define a set of irreducible tensor operators using powers of x, y and z.
Defining
p
Tqk ≡ r k Yk,q 4π/(2k + 1), (10.14)

allows one to define the following polynomials in terms of irreducible tensor operators, using

196
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

z = r cos θ and (x + iy) = r sin θeiϕ .

T00 = 1, (10.15)
1
T11 = − √ (x + iy),
2
1
T0 = z,
1
1
T−1 = √ (x − iy),
2
r
3 2
T22 = (x + 2ixy − y 2 ),
8
r
3
T12 = − z(x + iy),
2
1
T02 = (3z 2 − r 2 ),
2
r
2
3
T−1 = z(x − iy),
r2
2
3 2
T−2 = (x − 2ixy − y 2 ).
8
p
The prefactor r 4π(2k + 1) was arbitrary. Any function of k is fine. It is only important that
the ratio of elements within a given k multiplet are unchanged. These expressions can then be
inverted to express the polynomials in terms of irreducible tensor operators,
1 = T00 , (10.16)
1 1
x = √ (T−1 − T11 ),
2
i 1
y = √ (T−1 + T11 ),
2
1
z = T0 ,
r
2
1 2 2 2
1 1
x = (T2 + T−2 ) − T02 + T00 r 2 ,
2 r 3 3 3
1 2 2 1 1
y2 = − (T2 + T−2 2
) − T02 + T00 r 2 ,
2 3 3 3
2 1
z 2 = T02 + T00 r 2 ,
3 3
1 2
xy = i √ (T−2 − T22 ),
6
1 2
xz = √ (T−1 − T12 ),
6
i 2
yz = √ (T−1 + T12 ).
6
This procedure could have been used to express other polynomials, e.g. x, y, z → Lx , Ly , Lz ,
because they behave identically under rotation.

197
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

For operators created by taking products of two different operators, e.g. Px Lz , one needs be
more careful, because in this case Px Lz does not behave like xz under rotation. If one considers
Pi Lj as a 3 × 3 matrix, there are 9 independent terms. In contrast the combination Pi Pj would
have only 6 terms because it is manifestly symmetric. For rank-2 tensors combined from two
⃗ and B,
different vector operators A ⃗ one can consider the following combinations, A ⃗ · B,
⃗ A i Bj −
Bi Aj and Ai Bj + Bi Aj − 2A ⃗ · B/3.
⃗ ⃗ ·B
The first combination, A ⃗ is clearly a scalar, and behaves
k=0
like an irreducible tensor operator, Tq=0 . The second is antisymmetric and behaves as part of
the three pseudo-vector components, ϵijk Aj Bk , or equivalently A ⃗ × B.⃗ These can be expressed
k=1
as a linear sum of the three T q operators as was done above for x, y and z. This leaves the
traceless symmetric combination at the end. These five independent operators can be written in
terms of Tqk=2 operators. For example, the operator Px Ly would be needed to be expressed as a
sum over some pieces with some operators Tqk=2 and combined with contributions from a set of
irreducible operators with Tq′k=1 .

Example 10.2: Combining operators to make irreducible tensor operators

Express the operator Px Ly as a linear sum over irreducible tensor operators.
Solution:
Following the discussion above, we first write Pi Lj in general.

1 Pi Lj + Pj Li 1 Pi Lj − Pj Li
   
Pi Lj = P ⃗ · Lδ
⃗ ij + − P ⃗ · Lδ
⃗ ij + .
3 2 3 2

The first term is a scalar and is invariant under rotation, i.e. it does not mix with any other
operators. The product Px Ly has a contribution from this element. The second term will
transform amongst a set of k = 2 irreducible tensor operators, while the third transforms like
the components of a vector A ⃗ =P ⃗ × L/2,
⃗ and can be expressed as a set of k = 1 operators.
Defining the k = 2 operators using Eq. (10.16) with both P ⃗ and L ⃗ being replaced by the
components x, y and z,

1 2
(Px Ly + Py Lx )/2 = i √ (S−2 − S22 ).
6

Here, Sqk is defined as irreducible tensor operators using Eq. (10.15) where each term ri rj is
replaced by (Pi Lj + Pj Li )/2. The third term is also expressed using Eq. (10.15),

⃗≡P
A ⃗ × L/2,
⃗
Az = Qk=1
q=0 ,
√
±(Ax ± iAy )/ 2 = Qk=1
±1 ,

and

(Px Ly − Py Lx )/2 = Az = Qk=1

0 .

This illustrates the meaning of the term “irreducible”. Under rotation, the nine operators Pi Lj
rotate amongst each other, but the nine operators can be divided, or reduced, into the three

198
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

sets, k = 0, 1, 2, where an element of each set does not mix with an element of the other two
sets under rotation.
This then provides the solution to the proposed question. Using the expressions for (Px Ly +
Py Lx )/2 and for (Px Ly − Py Lx )/2 above, one can see that

i 2
Px Ly = √ (S−2 − S22 ) + Q10 .
6
If the question had asked you to express Px Lx in terms of irreducible tensor operators, that
expression would have involved a scalar operator, with k = q = 0, from the P ⃗ ·L
⃗ term.

10.5 Wigner D matrices

The Wigner D matrices describe how objects with labels J and M transform amongst one an-
other under rotation. As stated in the previous chapter, rotations do not mix objects with differ-
ent J , thus the matrix has indices m and m′ ,
(J)
A′ (J, m) = Dmm′ (⃗
α)A(J, m′ ). (10.17)

where α⃗ are the three Euler angles. Any object, whether it be a state, or an operator, that has in-
dices J and M , denoting total angular momentum rotate this way. This statement is tautological
because angular labels are defined by how objects behave under rotation. Thus, for states,
(J)
R(⃗ α)|J, m′ ⟩,
α)|J, m⟩ = Dmm′ (⃗ (10.18)

or for operators
(k)
α)Tqk R−1 (⃗
R(⃗ α)Tqk′ .
α) = Dqq′ (⃗ (10.19)

This latter statement can be a bit confusing because one normally thinks of operators as trans-
forming under unitary transformation which multiplies the operators by U and U † on each side
of the operator. But, the operators have the dimensionality of the entire vector space (number of
all states × number of all states). Rotations indeed can be written as U OU † , but such rotations
rotate a set of k operators amongst one another if the operators are irreducible tensor operators.
k
The matrices Dq,q ′ (⃗
α) are defined in terms of Euler angles,

R(ϕ, θ, ψ) = e−iϕJz /ℏ e−iθJy /ℏ e−iψJz /ℏ , (10.20)

ℓ ′
Dm,m ′ (⃗
α) = ⟨ℓm |R(⃗
α)|ℓ, m⟩.

The order of the rotations might seem odd above. The first Euler rotation is described by ϕ, yet
it appears to operate last on the ket. To understand why the Euler angles are expressed in the
particular order, one can write

R(ϕ, θ, ψ) = e−iJz′ ψ/ℏ e−iJy′ θ/ℏ e−iJz ϕ/ℏ , (10.21)

199
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

where the y ′ axis is the y axis after an initial rotation around the original z axis by an angle ϕ
and z ′ is the new z axis formed after the rotation of θ about the y ′ axis. Thus,

e−iJy′ θ/ℏ = e−iJz ϕ/ℏ e−iJy θ/ℏ eiJz ϕ/ℏ (10.22)

e−iJz′ ψ/ℏ = e−iJy′ θ/ℏ e−iJz ψ/ℏ eiJy′ θ/ℏ .

Substituting these into the expression for R allows one to write the rotation matrices with-
out mentioning primed axes. Aside from the primes disappearing, the angles ϕ and ψ have
swapped their order,
R(ϕ, θ, ψ) = e−iJz ϕ/ℏ e−iJy θ/ℏ e−iJz ψ/ℏ . (10.23)

The D matrices, or Wigner D matrices, can be written in closed form,

j ′
−im ϕ j
Dm ′ m (ϕ, θ, ψ) = e dm′ m (θ)e−imψ , (10.24)
 1/2  −1/2
j λ 2j − k k+b
dm′ m (θ) = (−1)
k+a b
(a,b)
· (sin(θ/2))a (cos(θ/2))b Pk (cos θ),

where

k = min(j + m, j − m, j + m′ , j − m′ ),


 j + m : a = m′ − m; λ = m′ − m
j − m : a = m − m′ ; λ = 0

k= ,

 j + m′ : a = m − m′ ; λ = 0
j − m′ : a = m′ − m; λ = m′ − m


b = 2j − 2k − a.

Here, Pka,b (cos θ) are Jacobi polynomials. A description of their properties can be found in
https://en.wikipedia.org/wiki/Jacobi_polynomials. Fortunately, routines for calculating D
matrices are readily available. For more properties of D matrices one can visit https://en.wik
ipedia.org/wiki/Wigner_D-matrix.
The D matrices have a variety of properties, many of which are related to those of the spherical
harmonics. For example, the orthogonality property is

δj1 ,j2 δm1 ,m2 δm′1 ,m′2

Z
(j )∗ (j )
dω Dm1′ ,m1 (⃗
ω )Dm2′ ,m2 (⃗
ω) = . (10.25)
1 2 2j1 + 1
Here the integral over dω is shorthand for averaging all three Euler angles, i.e. the integration
2
Rcovers all the 0 < α1 , ψ < 2π, and −1 < cos θ < 1, then divides by the net angle, 8π so that
dω = 1. Note that for half integer spins, the integrals must subtend twice the angular range.

10.6 The Two-Dimensional Analogy of the Wigner-Eckart Theorem

Before launching into the Wigner-Eckart theorem, it is much easier to consider the two-dimen-
sional example. Consider bras and kets, which transform as eimϕ , and additionally, consider

200
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

operators, Tq , which also transform as eiqϕ . For example the operator x can be written as a sum
of two pieces, each of which transforms in this manner.
r
x= (eiϕ + e−iϕ ). (10.26)
2
The two terms will be considered separately, the one that transforms as eiϕ and the one that
transforms as e−iϕ . In the three-dimensional analog, each term will be called an irreducible
tensor operator. Under rotations, the states transform as

R(ϕ)|m⟩ = eimϕ |m⟩, (10.27)

−1 −imϕ
⟨m|R (ϕ) = ⟨m|e ,

and the operators transform as

R(ϕ)Tq R−1 (ϕ) = eiqϕ Tq . (10.28)

Thus, under rotation,

⟨m′ |Tq |m⟩ → ⟨m′ |R−1 (ϕ)R(ϕ)Tq R−1 (ϕ)R(ϕ)|m⟩ (10.29)

i(m+q−m′ )ϕ
=e ⟨m′ |Tq |m⟩.

The expectation must not depend on the rotation angle ϕ. This was certainly expected, because
if you rotate both the states and the operator by the same angle, nothing should change. Thus,
the matrix element must be zero unless m + q − m′ = 0. One could state this fact by

⟨β ′ , m′ |Tq |β, m⟩δm′ ,m+q F (m, q), (10.30)

where finding F (m, q) would typically involve solving for the matrix element for each m and
each q. The labels β and β ′ simply encapsulate all other labels needed to describe the states, but
which are not affected by rotation, e.g. the radial wave function. It should be emphasized that
the operator Tq could be any operator that transforms by a factor eiqϕ under rotation, such as
X + iY or Px + iPy .
The proofs of the two-dimensional and three-dimensional versions have much in common, but
the three-dimensional version is more powerful. Jumping ahead, the three-dimensional version
is
⟨β̃, J, M |Tqk |β, ℓ, mℓ ⟩ = ⟨k, q, ℓ, m|J M ⟩F (β̃, J˜, k, β, J ), (10.31)
where F might need to be calculated for each of the five arguments. Like the two-dimensional
version, there is a constraint on summing the projections, m + q = M , which is enforced by
the Clebsch-Gordan coefficient. Unlike the three-dimensional version, the function F does not
depend on the projections, m, M and q. This has to do with the fact that in three dimensions one
can rotate the states and operator about the x or y axis, which results in mixing various values
of m, but does not mix different values of J . Thus, one can calculate F for one combination
of m, q and M , as long as the Clebsch-Gordan coefficient is non-zero, then generate all other
combinations by taking ratios of Clebsch-Gordan coefficients. This proof is detailed below.

201
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

Example 10.3: Expressing a Two-Dimensional Function in Terms of Irreducible Tensor Op-

erators
Express the function, f (x, y) = xy, in terms of irreducible tensor operators. I.e., in polar
coordinates where the angular dependences are of the form eiϕ .
Solution: First, write
r
x= (eiϕ + e−iϕ ),
2
r
y= (eiϕ − e−iϕ ),
2i
r2
e2iϕ − e−2iϕ .


xy =
4i
Thus, xy can be considered as two terms, one which rotates as e2iϕ and one which rotates
as e−2iϕ . The two terms would be considered separately. If one were to calculate the matrix
element ⟨β ′ , m′ |xy|β, m⟩, the result would be zero unless m and m′ differed by ±2. If the
operator in question had been Px Py , the conclusion would have been the same because it can
also be written in terms of irreducible tensor operators with q = ±2.

We should qualify the statement above that the expectation of the matrix element ⟨α|O|β⟩
should not change under rotations if you rotate both the states and the operator, i.e.,

⟨α|R−1 (ϕ) R(ϕ)OR−1 (ϕ) R(ϕ)|β⟩ = ⟨α|O|β⟩.

 
(10.32)

Sometimes there is an external potential or field. The operator might be defined in terms of the
field, e.g. the interaction with a magnetic field, αB ⃗ · S.
⃗ For the statement above to be true,
one must also rotate the external field. If one factors out the magnetic field and only considers
the operators that operate on the state information represented by the bra and ket, then one
should only apply the Wigner-Eckart theorem to that portion of the matrix element. In this case
⃗
⟨α|S|β⟩. One would not treat B⃗ ·S
⃗ as being a scalar operator unless the source of the magnetic
field was also being described by the bra and ket. Thus, one separates the “external” operators
from the “internal” ones before applying either the two- or three-dimensional Wigner-Eckart
theorems.

10.7 The Wigner-Eckart Theorem

The Wigner-Eckart theorem states that the matrix element,

⟨β̃, J, M |Tqk |β, ℓ, mℓ ⟩ = ⟨k, q, ℓ, mℓ |J M ⟩f (β, β̃, J, k, ℓ). (10.33)

This is profoundly useful for two reasons, both related to the fact that the function f does not
depend on any of the projections q, m, or M . First, many matrix elements turn out to be zero
because of the vanishing Clebsch-Gordan coefficient. Secondly, if one needs multiple matrix
elements differing only by changes in the projections M, mℓ of q, one need only calculate for one
set of projections, then obtain the remaining ones by taking ratios of Clebsch-Gordan coefficients.

202
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

To prove the Wigner-Eckart theorm, one first defines the state,

X
|β ′ , k, ℓ, j, m⟩ ≡ ⟨k, q, ℓ, mℓ |jm⟩Tqk |β, ℓ, mℓ ⟩. (10.34)
q,mℓ

This must rotate as an object with angular momentum J ′ and projection M ′ . Using the com-
pleteness of Clebsch-Gordan coefficients, one can also state that
X
Tqk |β, ℓ, mℓ ⟩ = ⟨k, q, ℓ, mℓ |jm⟩|β ′ , k, ℓ, j, m⟩. (10.35)
jm

To prove the Wigner-Eckart theorem, one must prove that the matrix element
⟨β̃, J, M |β ′ , k, ℓ, j, m⟩ is
a) proportional to δM m .
b) proportional to δJj .
c) independent of M (aside from the delta function).

Proving (a) and (b) rely on noting that the overlaps are independent of rotation if both the bra
and ket rotate together,

⟨β̃, J, M |β ′ , k, ℓ, j, m⟩ = ⟨β̃, J, M |R−1 (⃗ α)|β ′ , k, ℓ, j, m⟩.

α)R(⃗ (10.36)

Under rotations about the z axis the bra and ket simply pick up a phase under rotation, e.g.
|β̃, J, M ⟩ → |β̃, J, M ⟩eiM ϕ . Thus, if one averages the overlap over all azimuthal angles, the
result will be proportional to δM m . This proves (a). To prove (b), one can consider the average
over all possible rotations of the three Euler angles. If one expresses the rotated states in terms
of the D matrices, the bra and ket become
X j
R(⃗ω )|β ′ , j, m⟩ = Dmm′ (⃗ ω ))∗ |β ′ , j, m′ ⟩, (10.37)
m′
X
⟨β̃, J, M |R−1 (⃗
ω) = ⟨β̃, J, M ′ |DM
J
′ M (⃗
ω ).
M′

Taking the overlaps,

X j
⟨β̃, J, M |R−1 (⃗ ω )|β ′ , j, m⟩ =
ω )R(⃗ ⟨β̃, J, M ′ |DM
J
′ M (⃗
ω )Dmm ω ))∗ |β ′ , j, m′ ⟩.
′ (⃗

M ′ m′

Next, we average the expression over all rotations, ω

⃗ , by integrating over all the angles. We then
invoke the orthogonality relations of Wigner D functions, Eq. (10.25). This gives
1 X
⟨β̃, J, M |β ′ , j, m⟩ = ⟨β̃, J, M ′ |β ′ , j, m′ ⟩δJ,j δM ′ m′ δM m (10.38)
2J + 1 M ′ m′
1 X
= δJj δM m ⟨β̃, J, M |β ′ , Jm⟩ δm′ M ′
2J + 1 m′ M ′

= δJj δM m ⟨β̃, J, M |β ′ , JM ⟩.

203
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

To prove (c), one needs to show that, aside from the δM m factor, there is no other M dependence,
or equivalently that ⟨β̃, J, M |β ′ , k, ℓ, j, M ⟩ is independent of M . To demonstrate this consider
the raising and lowering operators. One can see that
1
|β ′ , k, ℓ, J, M − 1⟩ = p J− |β ′ , k, ℓ, J, M ⟩, (10.39)
J (J + 1) − M (M − 1)
1
|β̃, J, M − 1⟩ = p J− |β̃, J, M ⟩.
J (J + 1) − M (M − 1)
Thus,
⟨β̃, J, M − 1|β ′ , k, ℓ, J, M − 1⟩ (10.40)
1
= ⟨β̃, J, M |J+ J− |β ′ , k, ℓ, J, M ⟩
J (J + 1) − M (M − 1)
1
= ⟨β̃, J, M |Jx2 + Jy2 + Jz |β ′ , k, ℓ, J, M ⟩/ℏ2
J (J + 1) − M (M − 1)
1
= ⟨β̃, J, M |J (J + 1) − M (M − 1)|β ′ , k, ℓ, J, M ⟩
J (J + 1) − M (M − 1)
= ⟨β̃, J, M |β ′ , k, ℓ, J, M ⟩.
Putting (a), (b) and (c) together, Eq. (10.35) becomes
X
⟨β̃, J, M |Tqk |β, ℓ, mℓ ⟩ = ⟨k, q, ℓ, mℓ |jm⟩⟨β̃, J, M |β ′ , k, ℓ, j, m⟩ (10.41)
jm

= ⟨k, q, ℓ, mℓ |J M ⟩⟨β̃, J, M |β ′ , k, ℓ, J, M ⟩,
and the final matrix element is independent of M . To make it clear that the matrix element is
independent of M , the Wigner-Eckart theorem is expressed as
⟨β̃, J, ||T (k) ||β, ℓ⟩
⟨β̃, J, M |Tqk |β, ℓ, mℓ ⟩ = ⟨k, q, ℓ, mℓ |J M ⟩ √ . (10.42)
2J + 1
The odd form with the two vertical bars carries no significance, aside from stating that the matrix
element√ does not depend on the projection M , but can depend on J , β, β̃, k and ℓ. The choice
of the 2J + 1 in the denominator is also arbitrary, but conventional. The matrix element with
the double bars is known as the reduced matrix element. It is confusing as it implies that it is an
object that one may calculate directly. This is not true. One must first calculate one of the matrix
elements for a specific combination of M, mℓ and q, then deduce the reduced matrix element
using Eq. (10.42). Finally, from the reduced matrix element one can express all other matrix
elements
√ as the reduced matrix element multiplied by a Clebsch-Gordan coefficient divided by
2J + 1.
The Wigner-Eckart theorem is profound for two reasons.
1. An irreducible tensor operator Tqk will not link two states of good angular momentum
unless the Clebsch-Gordan coefficient coupling the angular momentum of the ket with the
angular momentum of T to the angular momentum of the bra is not vanishing. Thus, a
vector operator, k = 1, cannot connect a j = 2 state to a j = 0 state. Also, the projections
must sum to zero, q + m = m′ .

204
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

2. If one needs to calculate ⟨β ′ , k, j, j ′ , m′ |Tqk |β, j, m⟩ for some given set of m, m′ and q,
one can choose any values of m, m′ and q that might make calculation of the matrix ele-
ment most simple then use the fact that the desired matrix element is the calculated matrix
element multiplied by the ratio of the Clebsch-Gordan coefficients. For example if one
needs to calculate a matrix element with m′ = 1, m = −1 and q = 2, one could simple
calculate the matrix element with all three projections equal to zero, then multiply by the
ratio of Clebsch-Gordan coefficients.
⟨β ′ , k, j, j ′ , m′ = 1|Tq=2
k
|β, j, m = −1⟩ (10.43)
′
⟨k, j, j , 1|k, j, 2, −1⟩
= ⟨β ′ , k, j, j ′ , m′ = 0|Tq=0
k
|β, j, m = 0⟩ .
⟨k, j, j ′ , 0|k, j, 0, 0⟩
This is handy because the integrals required to find the matrix element may be much easier
to perform with the projections all set to zero, and if one needs to find the elements for
many values of the projections, the matrix element must be calculated only once for each
J , as ratios of Clebsch-Gordan coefficients may be used to find all other matrix elements.

Example 10.4: Dipole Operators and the Wigner-Eckart Theorem

Suppose one needs to calculate the matrix elements,
⟨β̃, J ′ = 2, M ′ |x|β, J = 1, M ⟩,
for all M and M ′ .
1. For which values of M and M ′ are the elements zero?
2. If one had calculated the specific matrix element
γ ≡ ⟨β̃, J ′ = 2, M ′ = 0|z|β, J = 1, M = 0⟩,
express all the non-zero matrix elements above in terms of γ and Clebsch-Gordan coef-
ficients.
Solution:
1.) Using Eq. (10.16) one can write x, y and z in terms of irreducible tensor operators,
1
x = √ (−T11 + T−11
),
2
i
y = √ (T11 + T−1
1
),
2
z = T01 .
From this, one can see that from the fact that x either raises or lowers the projection,
⟨β̃, J ′ = 2, M ′ = 2|x|β, J = 1, M = 0⟩ = ⟨β̃, J = 2, M = 2|x|β, J = 1, M = −1⟩
= ⟨β̃, J ′ = 2, M ′ = 1|x|β, J = 1, M = 1⟩ = ⟨β̃, M = 1|x|β, J = 1, M = −1⟩
= ⟨β̃, J ′ = 2, M ′ = 0|x|β, J = 1, M = 0⟩ = ⟨β̃, J ′ = 2, M ′ = −1|x|β, J = 1, M = 1⟩
= ⟨β̃, M = −1|x|β, J = 1, M = −1⟩ = 0 = ⟨β̃, J ′ = 2, M ′ = −2|x|β, J = 1, M = 1⟩
= ⟨β̃, J = 2, M = −2|x|β, J = 1, M = 0⟩ = 0.

205
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

2.) The non-zero elements are then given by the one element you know, γ, and ratios of
Clebsch-Gordan coefficients. First, write the one element you know in terms of γ.

⟨β̃, J ′ = 2, M ′ = 2|T01 |β, J = 1, M = 1⟩ = γ.

Next, express the elements you want in terms of ⟨Tqk ⟩, then use the Wigner-Eckart theorem,

⟨β̃, J ′ = 2, M ′ = 2|x|β, J = 1, M = 1⟩
−1
= √ ⟨β̃, J ′ = 2, M ′ = 2|T11 |β, J = 1, M = 1⟩
2
γ ⟨k = 1, q = 1, J = 1, M = 1|J ′ = 2, M ′ = 2⟩
= −√ .
2 ⟨k = 1, q = 0, J = 1, M = 0|J ′ = 2, M ′ = 0⟩
The other elements are calculated in the same way.

γ ⟨1, 1, 1, 0|2, 1⟩
⟨β̃, J ′ = 2, M ′ = 1|x|β, J = 1, M = 0⟩ = − √ ,
2 ⟨1, 0, 1, 0|2, 0⟩
γ ⟨1, 1, 1, −1|2, 0⟩
⟨β̃, J ′ = 2, M ′ = 0|x|β, J = 1, M = −1⟩ = − √ ,
2 ⟨1, 0, 1, 0|2, 0⟩
γ ⟨1, −1, 1, 1|2, 0⟩
⟨β̃, J ′ = 2, M ′ = 0|x|β, J = 1, M = 1⟩ = √ ,
2 ⟨1, 0, 1, 0|2, 0⟩
γ ⟨1, −1, 1, 0|2, −1⟩
⟨β̃, J ′ = 2, M ′ = −1|x|β, J = 1, M = 0⟩ = √ ,
2 ⟨1, 0, 1, 0|2, 0⟩
γ ⟨1, −1, 1, −1|2, −2⟩
⟨β̃, J ′ = 2, M ′ = −2|x|β, J = 1, M = −1⟩ = √ .
2 ⟨1, 0, 1, 0|2, 0⟩

The Wigner-Eckart theorem explains many of the selection rules associated with atomic or nu-
clear transitions. In the dipole approximation the transition operator behaves as if it has J = 1.
Thus, there are no transitions with |∆J | > 1 because the Clebsch-Gordan coefficients would
then vanish.
Example 10.5: Wigner-Eckart Practice
A group of particles is bound into a state |α, J, M ⟩, where J and M reference the total angu-
lar momentum and α is the set of all other labels required to define the state.
1. Which of the matrix elements below might be non-zero?
(a) ⟨α′ , J ′ = 2, M ′ = 1|Px2 + Py2 |α, J = 4, M = 3⟩
(b) ⟨α′ , J ′ = 2, M ′ = 1|Px Py |α, J = 4, M = 1⟩
(c) ⟨α′ , J ′ = 2, M ′ = 1| ijk ϵijk Ji Rj Pk |α, J = 2, M = 1⟩
P

(d) ⟨α′ , J ′ = 2, M ′ = 1|Px |α, J = 3, M = 1⟩

2. With great effort you calculated the matrix element
M = ⟨α′ , J ′ = 2, M ′ = 0|Px2 + Py2 − 2Pz2 |α, J = 4, M = 0⟩.

206
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

In terms of M and Clebsch-Gordan coefficients, express the following matrix elements

(a) ⟨α′ , J ′ = 2, M ′ = 1|Px2 + Py2 − 2Pz2 |α, J = 4, M = 1⟩

(b) ⟨α′ , J ′ = 2, M ′ = 2|Px2 + Py2 |α, J = 4, M = 2⟩
(c) ⟨α′ , J ′ = 2, M ′ = 0|Px Py |α, J = 4, M = 0⟩
(d) ⟨α′ , J ′ = 2, M ′ = 0|Px2 + Py2 − 2Pz2 |α, J = 2, M = 0⟩

Solution:
1a) The operator can be written as a sum of T02 and T00 operators. It is zero becaue M ̸= M ′ .
1b) The operator can be written in terms of T12 and T−12
operators. It is zero because the bra
and ket have the same M .
1c) The operator is a scalar, and the bra and ket have the same J and M , so this can be non-
zero.
2
1d) The operator can be written in terms of T12 and T−1 operators, so it must be zero because
′
M =M.
Next, we address the question 2a-2d. Using Eq.s (10.15) and (10.16) for guidance one can write

Px2 + Py2 − 2Pz2 = −2T02 ,

2 2
1h 2 2 2 2 2 2
i
Px + Py = (Px + Py − 2Pz ) + 2(Px + Py + Pz )
3
2 2
= − T02 + T00 ,
3 3
i 2
Px Py = √ (T−1 − T12 ).
6
Using the Wigner-Eckart theorem, first find one of the matrix elements of the irreducible ten-
sor operator,

M
⟨α′ , J ′ = 2, M ′ = 0|Tq=0
k=2
|α, J = 4, M = 0⟩ = − .
2
2a)
−M ⟨2, 0, 4, 1|2, 1⟩
(−2)
2 ⟨2, 0, 4, 0|2, 0⟩
2b) Note that Px2 + Py2 will be a combination of T02 and a different irreducible operator T00 ,
but this last part gives zero because it a k = 0 operator can’t couple with a J = 4 state to
create a J = 2 state.
−M ⟨2, 0, 4, 2|2, 2⟩ −2
 

2 ⟨2, 0, 4, 0|2, 0⟩ 3
2c) This is zero because q + M ̸= M ′
2d) The matrix element might be non-zero but it cannot be written in terms of M because
J = 2 in the ket whereas J = 4 for the ket used to calculate M.

207
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

10.8 Exercises
1. The ∆++,+,0,− baryons have isospin 3/2 while the π +,0,− mesons form an isotriplet. Cal-
culate the branching ratios of all four ∆ decays into the corresponding pπ or nπ channels.
(For instance, what fraction of the ∆+ s decay into pπ 0 vs the nπ + channels.)
2. The S(975) meson is an isoscalar (I = 0), and decays into two pions. What fraction of the
two-pion decays are expected to go into the neutral pion channel?
3. Write the Racah coefficient, W (j1 , j2 , j3 , J ; J12 , J23 ) which is defined by
⟨(j1 , j2 ), J12 , j3 , J, M ′ |j1 , (j2 , j3 ), J23 , J, M ⟩
p
= δM,M ′ (2J12 + 1)(2J23 + 1)W (j1 , j2 , j3 , J ; J12 , J23 ),
in terms of Clebsch-Gordan coefficients.
4. For each operator, define a set (or sets) of irreducible tensor operators Tqk , from which one
can then define the given operator as a linear sum of the irreducible operators. (When
defining a set, write down Tqk for all possible q.)
(a) z
(b) px
(c) x2
(d) Lx Ly
(j)
5. In terms of the Pauli matrices, find the rotation matrix Dmm′ (ϕ, θ, ψ) for the case where
j = 1/2, and ϕ, θ and ψ are Euler angles.
6. Using Eq.s (10.21) and (10.22) derive Eq. (10.23).
7. Circle the matrix elements that might be non-zero.
• ⟨α′ , J ′ = 2, M ′ = 1|Px2 + Py2 |α, J = 4, M = 3⟩
• ⟨α′ , J ′ = 2, M ′ = 1|Px Py |α, J = 4, M = 1⟩
• ⟨α′ , J ′ = 2, M ′ = 1|ϵijk Ji Rj Pk |α, J = 2, M = 1⟩
• ⟨α′ , J ′ = 3, M ′ = 1|ϵijk Ji Rj Pk |α, J = 2, M = 1⟩
• ⟨α′ , J ′ = 2, M ′ = 1|Px |α, J = 3, M = 1⟩
• ⟨α′ , J ′ = 2, M ′ = 1|Px |α, J = 3, M = 0⟩
• ⟨α′ , J ′ = 2, M ′ = 1|Px |α, J = 2, M = 0⟩
8. Assume one is calculating matrix elements for transitions from a d state to a s state via a
quadrupole type coupling, and that one has performed an integral and found
I ≡ ⟨ℓ′ = 0, m′ = 0|(z 2 − r 2 /3)|ℓ = 2, m = 0⟩.
Given that one knows I, find
⟨ℓ′ = 0, m′ = 0|(x2 − r 2 /3)|ℓ = 2, m⟩
for all five values of m, in terms of I. You can leave your answer in terms of Clebsch-
Gordan coefficients.

208
PHY 851/852 10 ADVANCED TOPICS IN ANGULAR MOMENTUM

9. The matrix element for the electromagnetic decay of an atomic d state with mi to a p state
with mf is given by the matrix element,

ϵ∗ · ⟨ℓ = 1, mf |⃗
M ≡ α⃗ r |ℓ = 2, mi ⟩

where α⃗ ϵ∗ · ⃗
r is the interaction responsible for the decay, and ⃗
ϵ is the polarization vector
of the outgoing photon.
Consider the intensity of RCP light that
√ is emitted along the z axis. The polarization vector
of such light can be written as (1/ 2)(x̂ + iŷ). Find the RELATIVE intensities of such
light for all 15 combinations of mi and mf . You can get Clebsch-Gordan coefficients from
a table, e.g. https://en.wikipedia.org/wiki/Table_of_Clebsch-Gordan_coefficients,
or use an on-line calculator, e.g. https://www.wolframalpha.com/input/?i=Clebsch-Gor
dan+calculator.

209
PHY 851/852 11 FERMIONS

11 Fermions
When a system is comprised of a single particle, it does not matter whether it is a fermion or a bo-
son. However, when multiple particles are present, behaviors can differ radically. The difference
between fermions and bosons can be explained by a variety of perspectives.

1. When considering populations of single-particle levels, no more than one fermion of a

given type and spin can be assigned to a particular level, while an arbitrary number of
bosons can be placed in a given level. This constraint on fermions is known as the Pauli
exclusion principle.
2. For identical fermions of the same spin in three single-particle states a, b, c, the multi-
particle wave function ψabc (x1 , x2 , x3 ) must be completely anti-symmetric with respect
any interchange of the three spatial coordinates or under interchange of any of the labels
a, b, c. In contrast, the wave function for bosons must be symmetric under such inter-
changes.
3. Creation and destruction operators for fermions obey anti-commutation relations while
those for bosons obey commutation relations.

Examples of fermions are electrons, quarks and neutrinos. Examples of bosons are photons and
gluons. Composite particles made of an odd number of fermions, e.g. a proton which is made
of three quarks, are also fermions. Composite particles made of an even number of fermions are
bosons, e.g. a 12 C atom which is made of 6 protons, 6 neutrons and 6 electrons. All bosons have
integral spin while all fermions have half-integral spin.
First, we will consider consequences of the Pauli exclusion principle. Later sections will presents
discussions from the perspectives of anti-symmetrized wave functions or anti-commuting field
operators.

11.1 The Spin Statistics Theorem

Consider the overlap of a state with N particles, where we refer to this N -particle state as |Φ⟩.
One can take the overlap of this state with a state defined in coordinate space,

r1 , s1 ; ⃗
ϕ(⃗ r2 , s2 ; · · · ; ⃗
rN , sN ) = ⟨⃗
r1 , s1 ; ⃗
r2 , s2 ; · · · ; ⃗
rN , sN |Φ⟩, (11.1)

where si is the spin of the ith particle, which is at ⃗ ri . If one writes the same overlap, but with a
set of two indices permuted, e.g. ⃗ r4 , s4 ↔ ⃗ s7 , the new overlap will be the same but with a
r7 , ⃗
relative sign of ±1. The permutation operator P, which in this example permutes the i = 4 and
i = 7 particles, cannot change the physics if the particles are identical. Thus, it must commute
with the Hamiltonian, and because P 2 must return to the original state, its eigenvalues must be
±1. The spin-statistics theorem states that for half-integral-spin (s = 1/2, 3/2, 5/2 · · · ) parti-
cles the eigenvalue is -1, and for integral spins the eigenvalue is unity. Thus, wave functions for
integral-spin particles (bosons) are symmetric and the wave functions for half-integral particles
(fermions) are anti-symmetric under permutation.
To motivate the spin-statistic theorem, one can consider two spin half particles, each with spin
⟨Sz ⟩ = 1/2. The two particles are in the x − y plane at positions at opposite sides of a circle

210
PHY 851/852 11 FERMIONS

centered at the origin. If one were to rotate about the z−axis by a phase ϕ =180 degrees, the
spin wave function for each particle would change by a phase,

eiσz ϕ/2 = cos(ϕ/2) + iσz sin(ϕ/2). (11.2)

For spin-up particles, this introduces a phase factor of i when the angle of rotation is 180 degrees.
A second phase factor of i comes from rotating the second particle, resulting in an overall phase
factor of −1. The two particles are now in the same state but permuted. Thus, the permutation
is equivalent to this 180 degree rotation and has a phase factor of −1 relative to the unrotated, or
un-permuted, state. For spin-down fermions one would have an overall phase factor of −1. For
bosons, rotating by 180 degrees yields a factor of ±1 for each particle, which gives an overall
factor of +1. Thus, half-integral-spin particles should be anti-symmetric under permutation,
while permutation of integral-spin particles should yield the same overall state, without a sign
change. For the two-particle wave functions of fermions,

ψab (⃗ r2 ) =
r1 , ⃗ ⟨⃗ r2 |a, b⟩
r1 , ⃗ (11.3)
= −ψab (⃗ r1 ) = −⟨⃗
r2 , ⃗ r1 |a, b⟩
r2 , ⃗
= −ψba (⃗ r2 ) = −⟨⃗
r1 , ⃗ r2 |b, a⟩
r1 , ⃗
= ψba (⃗ r1 ) = ⟨⃗
r2 , ⃗ r1 |b, a⟩.
r2 , ⃗

The anti-symmetry property applies to either the ⃗

ri labels or to the states a, b.
If the two-particle wave function is anti-symmetric, the two particles cannot occupy the same
single-particle state. If one uses field operators to create or destroy particles, the operators must
obey anti-commutation relations, rather than the commutation relations used in Sec. 9.2, if the
particles are fermions. Fermi field operators are discussed in Sec. 11.5.

11.2 The Pauli-Exclusion Principle

The wave-function ψa,sa ;b,sb (⃗
r1 , s1 ; ⃗
r2 , s2 ) must be anti-symmetric under interchange of a and
b. So if a = b the wave function must be zero. If one considers the set of eigenstates of an
external potential V (⃗
r ), then each single-particle eigenstate can only host a single particle if
the spins a = b and sa = sb . If the spins, sa and sb are different, or if a ̸= b, then one can
write an anti-symmetrized version of the wave function. For that reason for each single-particle
eigenstate of an external potential, only one fermion can occupy a given eigenstate, or orbital.
If there is spin-degeneracy, then the same energy level might accommodate several different
fermions if they have different angular momentum or spin. This constraint is prominent in
understanding electronic structures of multi-electron atoms of in understanding how neutrons
and protons can arrange themselves to form nuclei.
Example 11.1: Filling the Harmonic Oscillator with Fermions
Here, we consider the filling of harmonic oscillator energy levels with fermions. This simple
picture provides a surprising amount of insight into the structure of light nuclei.
First, we must understand the degeneracy of single-particle levels in a 3-d harmonic oscillator.
Each level has an energy of (N + 3/2)ℏω. Considering the problem in a Cartesian basis, there
are N⊥ + 1 ways to arrange nx + ny = N⊥ (Here, we consider only spin-up fermions). The

211
PHY 851/852 11 FERMIONS

number of ways to arrange nx , ny and nz to add up to N is

N
X (N + 1)(N + 2)
d(N ) = N⊥ + 1 = .
N⊥ =0
2

Thus, there is one combination of nx , ny and nz to get N = 0, three combinations to get

N = 1, six to get N = 2, etc.
Next, we wish to calculate the number of states of a specific N that have a given orbital
angular momentum ℓ. This will come in handy when we consider the spin-orbit coupling on
top of the harmonic oscillator structure. The solution will be that the states with excitation
N are accounted for with one ℓ = N multiplet, one ℓ = N − 2 multiplet, one ℓ = N − 4
multiplet, etc. To prove that this is the case, we make an inductive proof assuming it is true for
N −2. For every state with excitation N −2 there is a state with the same angular momentum
with excitation N which is reached by applying the additional operator (a†x a†x +a†y a†y +a†z a†z )
which is a rotational scalar. There are (N − 1)N/2 such states. Furthermore, there must be at
least one ℓ = N multiplet because the state (a†x + ia†y )N |0⟩ transforms like part of a ℓ = N
multiplet. But one ℓ = N multiplet with degeneracy (2N + 1), and the (N − 1)N/2 states
accounted for by counting the states with excitation N − 2 completely account for all the
states with excitation N ,
(N + 1)(N + 2) (N − 1)N
= + (2N + 1).
2 2
Thus, increasing the excitation by 2ℏω adds one more multiplet with ℓ = N . As an example,
for N = 4, there are multiplets with ℓ = 4, 2, 0 with degeneracy 9 + 5 + 1 = 15 = 5 · 6/2.
If no spin-orbit terms were present, adding neutrons to a harmonic oscillator would lead to
shell closures with neutron numbers equal to 2, 8, 20, 40 and 70. We now consider filling all
the N ≤ 3 shells, which requires 40 neutrons. Note that these numbers accounted for both
spin-up and spin-down levels. To calculate this number for N ≤ 3 one would consider that
there is 1 · 2/2 = 1 way to arrange nx , ny , nz for N = 0 or E = 3ℏω/2, 2 · 3/2 = 3 ways
to make arrangements for E = 5ℏω/2 single-particle levels, 3 · 4/2 = 6 single-particle levels
with E = 7ℏω/2 and 4 · 5/2 = 10 single-particle levels with E = 9ℏω/2. This sums to
1 + 3 + 6 + 10 = 20 single-particle levels with E ≤ 9ℏω/2. Because each single-particle
level can have either spin-up or spin-down, this gives 40 single-particle levels for N ≤ 3, or
equivalently E ≤ 9ℏω/2, for spin 1/2 particles. Thus, one could place up to 40 neutrons into
the 40 single-particle levels without having any particle’s energy exceed 9ℏω/2.
Adding the spin-orbit term adjusts the single-particle energies by an amount

βℏ2
Es.o. = −β ⃗
ℓ·⃗
s=− [j(j + 1) − ℓ(ℓ + 1) − s(s + 1)] .
2
For every value of ℓ there are two values of j, j = ℓ ± 1/2. Thus, one finds the energy levels
by first labeling the states by N and ℓ. Then splitting each level into it’s two values of j and
finding its energy,

βℏ2
E = (N + 3/2)ℏω − [j(j + 1) − ℓ(ℓ + 1) − s(s + 1)] − αℓ(ℓ + 1).
2

212
PHY 851/852 11 FERMIONS

An additional term, proportional to L2 , was also added to phenomenologically better repro-

duce levels. This is known as the Nilsson model, https://en.wikipedia.org/wiki/Nils
son_model. The shell structure changes due to the spin-orbit interaction, and the numbers of
neutrons required to reach a condition where there is a large gap are called magic numbers.
In nuclei, the magic numbers are 2, 8, 20, 28, 50, 82 and 126. Nuclei are considered doubly
magic if both the neutron and proton numbers are magic numbers. Examples are 4 He, 16 0,
40 48
20 Ca, Ca, 56
28 Ni,
48
Ni, 78 Ni, 100
50 Sn,
132
Sn and 208
82 Pb. The magic numbers of nuclei were only
explained by assuming an anomalously large spin-orbit coupling. The reason for this sur-
prisingly large coupling will be explained later in the semester when the Dirac equation is
presented.

11.3 Fermi Gases and Neutron Stars

Consider a large number N of fermions of spin s and mass m in a large box of volume V . Here,
the box is defined by a confining potential that is zero inside, and infinite outside, the box. If
the fermions are placed in the lowest levels consistent with the Pauli exclusion principle, the
highest single-particle energy is known as the Fermi energy ϵf and the momentum of that state
is known as the Fermi momentum pf . The density is a function of the Fermi momentum,
Z p<pf
V
N = (2s + 1) d3 p, (11.4)
(2πℏ)3
N (2s + 1) 3
n= = p ,
V 6π 2 ℏ3 f
where s is the intrinsic spin of the fermion, e.g. 1/2, 3/2 · · · . Note that the Fermi momentum
is determined by the density and does not depend on the particle’s mass, whereas the Fermi
energy does depend on the mass, ϵf = p2f /(2m). For this course, we use the term “Fermi
energy” to refer to the kinetic energy of a particle at the top of the Fermi surface. The term Fermi
surface refers to the surface in momentum space of the sphere defining the occupied region of
momentum space. Contributions to the energy from the mass or potential also play a role, but
the term “Fermi energy” will only refer to the kinetic part of the energy. When reading the
literature, one needs to be careful in understanding how the author is using the term.
One example where the Fermi energy plays a pivotal role is in neutron stars. Due to beta decays
neutrons can change into protons through the emission of an electron and a neutrino,

n → p + e + ν̄, p + e → n + ν. (11.5)

Because neutrinos can exit the star due to their negligible masses and small cross-sections, they
need not be considered with respect to conservation laws. However, baryon number (the net
number of neutrons and protons) and electric charge must be conserved, and in fact, the net
electric charge density must be zero. Thus, beta decays, and inverse beta decays, can proceed as
long as the following constraints are met,

nn + np = nB , (11.6)
ne = np ,

213
PHY 851/852 11 FERMIONS

where nn , np and ne are the neutron, proton and electron densities respectively and nB is the
baryon density. Beta decays will proceed until the energy is minimized for the given value of
the baryon density nB . Neutrons are more massive than protons by an amount (mn − mp )c2 =
1.30 MeV which more than accounts for the electron’s mass of 0.511 MeV/c2 . When the energy
is minimized, changing a neutron to a proton plus electron must leave the energy unchanged if
the particles are added and removed from the top of the Fermi surfaces. If net energy changes
when undergoing such a reaction, reactions would proceed whichever way lowers the energy.
These reactions would emit either a neutrino or anti-neutrino of vanishing energy to bring the
(p) (e) (n)
energies into balance. For example if ϵf + ϵf + mp + me > ϵf + mn , one could lower the
energy by the reaction p + e → n + ν. This reaction would proceed until enough protons and
electrons were taken from their Fermi surfaces, and enough neutrons were added to the top of
the neutron Fermi sea until the following condition was met,
p
ϵn e
f + 1.30 MeV = ϵf + ϵf + 0.511 MeV. (11.7)

Combining this equality with the two constraints in Eq. (11.6) allows one to find all three Fermi
momenta in terms of nB .
At very low density, the fact that neutrons are heavier than protons by 1.30 MeV/c2 leads to
protons being more populous than neutrons even though each proton must be accompanied
by an electron with a mass of 0.511 MeV/c2 . Therefore, in stars like our sun protons are much
more numerous than neutrons. At high density the electron’s Fermi energy plays the pivotal role.
Because electrons and protons have the same density, they have the same Fermi momentum, but
the electron’s Fermi energy is much higher due to its lighter mass. For a proton,
q
(p)
ϵf = m2p + p2f − mp ≈ p2f /(2mp ), (11.8)

which would be smaller if the protons mass were even larger. Electrons are lighter and thus
have a much higher Fermi energy than protons, and in fact the densities tend to be so high
that the electron’s Fermi momenta well exceed the mass. The electron energies are then highly
relativistic,
q
(e)
ϵf = m2e + p2f − me ≈ pf . (11.9)

Thus, at high density the system must try to reduce the electron density which in turn requires
a reduction in the proton density in order to maintain charge neutrality. This results in a large
excess of neutrons in order to achieve the required baryon density. This is certainly the case for
the interior of neutron stars where densities are of the order 0.1 baryons per cubic fm. The re-
sulting electron Fermi energy is on the order of 10 MeV as is the resulting neutron Fermi energy.
At these densities the electrons are highly relativistic, because the electron Fermi energy is much
higher than the electron mass, while the protons and neutrons, whose masses are approximately
939 MeV/c2 , remain largely non-relativistic.
Example 11.2: Zero-Point Surface Energy for Fermions
It costs energy to divide a piece of metal (which is approximated here as a non-interacting
electron gas) into two pieces. The associated surface energy, energy per surface area, has
a component deriving from the penalty associated with the kinetic energies of the particles

214
PHY 851/852 11 FERMIONS

L L/2 L/2
Figure 11.1: For Example 11.2: Energy levels of a one-dimensional infinite square well of length L are
displayed in the left figure, while the levels of two wells of length L/2 are displayed in the two right
panels. Each level can hold one fermion of a given spin. By splitting the well in two, the energy penalty
is equivalent to moving fermions from the 1, 3, 5, 7 · · · levels to the 2, 4, 6, 8 · · · levels. Because half the
fermions were moved up by one step in the ladder, the net energy penalty is ϵf /2 for each spin in the
limit of a large system where there are many levels.

known as the zero-point energy. To understand the source of this energy, we first consider the
one-dimensional case. Consider a box of length L, which is divided into two boxes, each of
length L/2. The initial energy levels of the larger box and the energy levels of the two smaller
boxes are illustrated in Fig. 11.1. The levels are described by,
ℏ2 kn2
En = , kn L = nπ, n = 1, 2, 3, · · ·
2m
But after splitting the box the energy levels are given by
ℏ2 kn2
En = , kn L = nπ, n = 2, 4, 6, · · · ,
2m
with the new levels being able to hold twice as many particles because there are now two
boxes.
Thus, in one dimension the energy penalty can be thought of as arising from forcing half the
particles to move up by one rung on the energy diagram ladder. Neglecting spin, if all the
particles had moved up one rung, the net penalty would be the Fermi energy, but because
half the particles moved up, the penalty is
∆E = ϵf /2.

215
PHY 851/852 11 FERMIONS

The energy penalty would then be doubled to account for the two spins.
In three dimensions, there are three quantum numbers, nx , ny and nz , corresponding to kx ,
ky and kz .
kx Lx = nx π, ky Ly = ny π, kz Lz = nz π.
If the box is divided in two along the x axis, Lx → Lx /2, the penalty for all levels with a
fixed ny and nz is,
1 ℏ2 (kf2 − ky2 − kz2 )
δE(ny , nz ) = .
2 2m
This result follows from the fact that half the particles moved up a notch on the nx ladder,
with the top energy being such that the overall energy (including the y and z components)
adds up to the Fermi energy.
One can now find the entire energy cost by summing over ny and nz .

Ly Lz
Z
∆E = dky dkz δE(ny , nz )
(2π)2 |k|<kf
ℏ2 Ly Lz kf
Z
= kdk (kf2 − k2 )
8mπ 0
ℏ2 kf4
= Ly Lz
32mπ
kf2
= Ly Lz ϵf .
16π
In the second step, the variables kx and ky were transformed to polar coordinates, k =
q
kx2 + ky2 . Dividing by 2Ly Lz (as extra area is added to both sides of the split) gives the
surface energy. One would also multiply the result by the spin degeneracy.

∆E kf2
= (2s + 1) ϵf .
A 32π
In real systems the surface energy is of this order. It is lessened by the fact that the electron
density is not sharply cut off at the surface, but dies out smoothly. An additional positive con-
tribution can result from finite-range attractive interactions between the particles. In nuclear
physics, the above expression over-predicts the surface energy of nuclear matter by approxi-
mately 25%.

Example 11.3: Magnetization of a Fermi Gas

A magnetic field lowers the energies of those particles whose spin aligns with the magnetic
field, while raising the energies of those that are anti-parallel. For a magnetic field pointing
in the z direction, this results in a surplus of spin-up particles vs spin-down particles. The
magnetic energy is

⃗
S
HB = −gµB ⃗
· B,
ℏ

216
PHY 851/852 11 FERMIONS

where µB = eℏ/2mc is the Bohr magneton, and for electrons g = 2. For a gas of non-
interacting electrons of number density ne , find the net spin, n↑ − n↓ for a small magnetic
field. Also, find the magnetic susceptibility, χ ≡ µB (n↑ − n↓ )/B.
Solution: The difference between the spin-up and spin-down energies is
eℏB
∆E = .
mc
In terms of the two Fermi momenta, the densities are
1
n↑ = k↑3 ,
6π 2
1
n↓ = k3 ,
6π 2 ↓
ne = n↑ + n↓ .

The Fermi energies differ by ∆E,

ℏ2 k↓2 ℏ2 k↑2
− = ∆E.
2m 2m
For small fields, we can write the expressions above in terms of ∆k = k↑ − k↓ , and if the
original Fermi momentum was kf , then k↑ = kf + ∆k/2, k↓ = kf − ∆k/2.

ℏ2 kf
∆E = ∆k,
m
3kf2
n↑ − n↓ = ∆k,
6π 2
kf2 m
= ∆E
2π 2 ℏ2 kf
mkf
= ∆E
2π 2 ℏ2
ekf
= B
2π 2 cℏ
e
= (3π 2 ne )1/3 B.
2π 2 cℏ
The magnetic moment of the electron is µB , so the magnetic moment density is

M = µB (n↑ − n↓ )
eµB
= 2
(3π 2 ne )1/3 B.
2π cℏ
The magnetic susceptibility is the ratio of the induced magnetic moment density to a small
applied field,
eµB
χ= (3π 2 ne )1/3 .
2π 2 cℏ

217
PHY 851/852 11 FERMIONS

11.4 Multi-Particle Symmetrization and Slater Determinants

Consider a multi-particle state |α1 , α2 · · · ⟩ of distinguishable particles, where the ordering
α1 , α2 , · · · signifies that the first particle is in the state α1 , the second in α2 , etc. The per-
mutation operator Pij is defined by

P(12)|α1 , α2 , α3 , · · · ⟩ = |α2 , α1 , α3 , · · · ⟩. (11.10)

If any two particles are identical, and indistinguishable, the multi-particle state |ϕ⟩ should be an
eigenstate of the permutation operator,

P(12)|α1 , α2 , α3 , · · · ⟩ = |α2 , α1 , α3 , · · · ⟩ = p|α1 , α2 , α3 , · · · ⟩. (11.11)

If the permuted state differed from the un-permuted state, aside from a phase factor, the observ-
ables related to that state might change. This should not be possible if the particles are identical.
All states of the system should be eigenstates of the permutation operator P(i, j),

P(i, j)|ϕ⟩ = (±1)|ϕ⟩. (11.12)

Only ±1 are possible eigenvalues of the permutation operator because applying the same per-
mutation operator twice yields unity, P 2 (i, j) = 1. In order for the state to be an eigenstate of
all permutations, an N -particle state of all identical particles must be either totally symmetric
(bosons) or totally anti-symmetric (fermions).
1 X
|ϕ⟩ = √ (±1)np Pp |α, β, γ · · · ⟩, (11.13)
N ! perm.s p

where np is the number of pairwise permutations required to make the desired permutation.
For fermions, no two states, e.g. α and β can be identical. As an example, the three particle
antisymmetric state is
1
|ϕ⟩ = √ {|α, β, γ⟩ − |β, α, γ⟩ − |γ, β, α⟩ − |α, γ, β⟩ + |β, γ, α⟩ + |γ, α, β⟩} . (11.14)
6
Note that the fifth and sixth orderings required two pairwise permutations, whereas the second,
third and fourth orderings could be generated from a single pair-wise switch of labels. Similarly,
one can do the same for a state with coordinate-space labels,
1
|ϕ′ ⟩ = √ {|⃗
ri , ⃗ rk ⟩ − |⃗
rj , ⃗ rj , ⃗ rk ⟩ − |⃗
ri , ⃗ rk , ⃗ ri ⟩ − |⃗
rj , ⃗ ri , ⃗ rj ⟩ + |⃗
rk , ⃗ rj , ⃗ ri ⟩ + |⃗
rk , ⃗ rk , ⃗ rj ⟩} .
ri , ⃗
6
(11.15)

Taking the overlap of ⟨ϕ′ |ϕ⟩ gives the 3-particle wave function. The 36 terms reduce to six terms
because so many are redundant,

⟨ϕ′ |ϕ⟩ = ϕα,β,γ (⃗

ri , ⃗ rk )
rj , ⃗ (11.16)
=ϕα (⃗
ri )ϕβ (⃗
rj )ϕγ (⃗ rk ) − ϕβ (⃗
ri )ϕα (⃗rj )ϕγ (⃗
rk )
− ϕγ (⃗
ri )ϕβ (⃗rj )ϕα (⃗ rk ) − ϕα (⃗
ri )ϕγ (⃗rj )ϕβ (⃗
rk )
+ ϕβ (⃗
ri )ϕγ (⃗rj )ϕα (⃗ rk ) + ϕγ (⃗
ri )ϕα (⃗rj )ϕβ (⃗
rk ).

218
PHY 851/852 11 FERMIONS

This wave function is anti-symmetric with respect to the pairwise interchange of any of the
three indices α, β, γ, and is also anti-symmetric with respect to the pairwise interchange of any
of three spatial indices ⃗
ri , ⃗ rk . Ignoring spin, the overlap of such a state in coordinate space,
rj , ⃗
i.e. for any number of particles, n, the multi-particle wave function, ϕα1 ···αn (⃗ r1 · · · ⃗
rn ) must also
be an eigenstate of the permutation operator, and this must be true whether one is permuting
the αi indices or the ⃗
r indices.
Notation in terms of bras and kets can be confusing. Above, the state |α, β, γ⟩ denoted a state of
three distinguishable particles, where the first particle occupied the level α, the second occupied
β and the third γ. Given that the particles are actually indistinguishable, from here on out, as
is nearly always the case in the literature, the state |α, β, γ⟩ will simply refer to the fact that
α, β and γ are occupied and will not imply a specific ordering. After all, it makes no sense
to state“first particle in in level α, second is in β, · · · ”, because there is no way to distinguish
which particle is first, second or third. Using fermionic field operators to define states allows
one to to avoid this confusion, and is the subject of the next section.
If the single particle wave functions ϕα , ϕβ , · · · are orthonormal, the multi-particle wave func-
tion will also be normalized,
1
Z
d3 r1 · · · d3 rn |ϕα1 ···αn (⃗ rn )|2 = 1.
r1 · · · ⃗ (11.17)
N!
The 1/N ! factor accounts for the fact that integrating over all positions redundantly counts for
any given configuration of ⃗ ri . One can express the multi-particle wave function in terms of
normalized single-particle wave functions, ϕα (⃗ r ), in the form of a determinant
 
r1 ) ϕα1 (⃗
 ϕα1 (⃗ r2 ) · · · ϕα1 (⃗
rn ) 

 ϕ (⃗ r ) ϕα2 (⃗ r2 ) · · · ϕα2 (⃗
rn ) 
ϕα1 ···αn (⃗ rn ) =  α2 1
r1 · · · ⃗ . (11.18)
 ··· ··· ··· ··· 
 ϕαn (⃗
r1 ) ϕαn (⃗ r2 ) · · · ϕαn (⃗
rn ) 

The expression of the anti-symmetric N -particle wave

√ function as a Slater Determinant in Eq.
(11.18) is often expressed with a pre-factor of 1/ N !. When this convention is chosen, the
completeness/normalization condition in Eq. (11.17) is simultaneously altered by removing the
prefactor 1/N !.
This becomes more complicated when one includes spin, or isospin indices. In fact, one might
have a wave function that is symmetric in coordinate space, but anti-symmetric in spin, or vice
versa. Such cases will be discussed in the next chapter.
The Slater determinant enforces the Pauli exclusion principle. If any of the indices, α1 · · · αn , are
identical, the multi-particle wave function vanishes. The determinant also vanishes in the limit
that any two positions become equal. It should be emphasized that not all multi-particle wave
functions can be written in terms of a Slater determinant. If one were considering distinguishable
particles, the multi-particle wave function would be generally of the form ϕα1 ···αn (⃗ r1 · · · ⃗
rn ).
Assuming that ϕ factorizes into a product of single-partcle wave functions represents a strong
assumption. Similarly, although the Slater determinant is not a simple product wave function,
the assumption that it can be written as a sum of simply permuted product wave functions rep-
resents an equally strong assumption. In fact, this assumption is the foundation of the Hartree-
Fock approximation, which is discussed in the next chapter.

219
PHY 851/852 11 FERMIONS

Example 11.4: Writing Three-Particle Wave Function with Slater Determinants

Imagine one has 3 electrons in a Coulomb potential. Two of the electrons are in the 1s state.
The radial wave function for this state is noted by ϕ1s (r). The third electron is in the 2p level
and has spin-up, where the radial wave function is labeled ϕ2p (r). Write the three-particle
wave function, where particles are at positions ⃗ r3 with spin projections s1 , s2 =↓
r2 and ⃗
r1 , ⃗
and s3 respectively.
Solution: Use the Slater Determinant, with the labels being (s, ↑), (s, ↓) and (p, ↑).

Ψ((⃗
r1 , s1 ), (⃗
r2 , s2 ), (⃗
r3 , s3 ))
 
 ϕ1s,↑ (⃗ r1 , s1 ) ϕ1s,↑ (⃗r2 , s2 ) ϕ1s,↑ (⃗
r3 , s3 ) 
 
=  ϕ1s,↓ (⃗ r1 , s1 ) ϕ1s,↓ (⃗r2 , s2 ) ϕ1s,↓ (⃗
r3 , s3 ) 

 ϕ2p,↑ (⃗ r1 , s1 ) ϕ2p,↑ (⃗
r2 , s2 ) ϕ2p,↑ (⃗
r3 , s3 ) 
= ϕ1s,↑ (⃗
r1 , s1 )ϕ1s,↓ (⃗
r2 , s2 )ϕ2p,↑ (⃗r3 , s3 ) − ϕ1s,↑ (⃗r1 , s1 )ϕ1s,↓ (⃗
r3 , s3 )ϕ2p,↑ (⃗r2 , s2 )
− ϕ1s,↑ (⃗ r2 , s2 )ϕ1s,↓ (⃗
r1 , s1 )ϕ2p,↑ (⃗r3 , s3 ) + ϕ1s,↑ (⃗r2 , s2 )ϕ1s,↓ (⃗
r3 , s3 )ϕ2p,↑ (⃗r1 , s1 )
+ ϕ1s,↑ (⃗ r3 , s3 )ϕ1s,↓ (⃗
r1 , s1 )ϕ2p,↑ (⃗r2 , s2 ) − ϕ1s,↑ (⃗r3 , s3 )ϕ1s,↓ (⃗
r2 , s2 )ϕ2p,↑ (⃗r1 , s1 ).

The three states, (1s, ↑), (1s, ↓) and (2p, ↑) were chosen to be different. If any two of the
states were the same, e.g. replace (1s, ↓) with (1s, ↑) one can see that the resulting three-
particle wave function would vanish. Similarly, if one chooses to evaluate the wave functions
with two of the particles at the same position and the same spin, e.g. (⃗
r1 , s1 ) = (⃗
r2 , s2 ), the
three-particle wave function would also vanish. But, it does not vanish if two positions are
r1 = ⃗
the same, but with different spins, e.g. ⃗ r2 , s1 ̸= s2 .

11.5 Fermi Creation and Destruction Operators

Writing states as a sum over various permutations is a rather clumsy way to consider the Fermi
nature of the particles as it requires assuming there is a “first” particle, “second” particle, and so
on, then include N ! terms in the various expressions. The algebra of anti-commuting creation
and destruction operators offers a more natural means to incorporate anti-symmetrization. With
this formalism, a state can be noted simply by their labels with no mention of permutations.
Matrix elements are then calculated according to the algebra of the creation and destruction
operators, which account for the symmetrization or anti-symmetrization. This approach allows
one to address systems with many fermions, even in the macroscopic limit.
Fermi creation and destruction operators obey the algebra

{aα , a†β } = δα,β , (11.19)

{aα , aβ } = {a†α , a†β } = 0.
The curly brackets denote “anti”-commutation, i.e. {A, B} = AB + BA. One consequence of
anti-commutation relations is that any time two creation operators, or two destruction operators,
with the same index are next to one another the result is zero, aα aα = a†α a†α = 0. This enforces
the Pauli exclusion principle. Similar to the Bose example, the vacuum is annihilated by the
destruction operator.
aα |0⟩ = 0. (11.20)

220
PHY 851/852 11 FERMIONS

The commutation rules written above assumed the states α, β, · · · were orthogonal. It is more
complicated if the indices refer to states from different bases. The first basis will be labeled with
roman letters i, j, k · · · and the second will be with greek, α, β, γ · · · . First, as seen in Chapter
9, one can see that under a basis transformation creation and destruction operators transform
like single-partcle states,
X
⟨α| = ⟨α|k⟩⟨k|, (11.21)
k
X
⟨0|aα = ⟨0| ak ⟨α|k⟩,
k
X
aα = ⟨α|k⟩ak .
k

The anti-commutation algebra becomes

X
{aα , a†j } = ⟨α|k⟩{ak , a†j } = ⟨α|j⟩. (11.22)
k

We are now able to calculate arbitrary matrix elements for N particle states,

⟨i, j, k, · · · |α, β, γ, · · · ⟩ = ⟨0|(ai aj ak · · · )(a†γ a†β a†α · · · )|0⟩ (11.23)

X
= (−1)np ⟨i|α⟩⟨j|β⟩⟨k|γ⟩ · · ·
perm.s of i,j,···

As an example we consider two-particle matrix elements,

⟨i, j|α, β⟩ = ⟨i|α⟩⟨j|β⟩ − ⟨j|α⟩⟨i|β⟩. (11.24)

If i and j refered to positions x and y, the result would read

⟨x, y|α, β⟩ = ϕα (x)ϕβ (y) − ϕα (y)ϕβ (x), (11.25)

which signifies that the two particles can not be at the same position. Note that the overlap of
an N -particle matrix element would yield N ! terms. Thus, it seems little has been gained using
field operators rather than writing symmetrized/anti-symmetrized wave functions which also
have N ! terms. However, when writing the bra and ket with all permutations,
√ both the bra and
ket have N ! terms, with the extra N ! being cancelled by the 1/ N !s in the normalization of
the wave functions. This rapidly becomes onerous.
It should also be noted that all the same results are valid for bosonic operators, except that the
(−1)np factors disappear.

11.6 Fermionic Field Operators

The field operators Ψ†s (x) and Ψs (x) create and destroy a particle at position x with spin s.
They obey the anti-commutation relations,

x), Ψ†s′ (⃗
{Ψs (⃗ y )} = δ 3 (⃗
x−y
⃗ )δss′ . (11.26)

221
PHY 851/852 11 FERMIONS

Just like the non-fermionic operators defined in Chapter 9, they can be defined in terms of cre-
ation and destruction operators for momentum states,
1 X −i⃗k·⃗r † ⃗
Ψ†s (⃗
r) = √ e as (k). (11.27)
V ⃗
k

The field operators commute with other operators as

x), a†α } = ϕα (s, ⃗

{Ψs (⃗ x), (11.28)

where ϕα (s, ⃗x) = ⟨⃗

x, s|α⟩ is the single particle wave function of a particle of spin s in the state
α. Here, the label α encapsulates all the information of the state, including the spin. Analo-
gously to the Bose field operators discussed in Chapter 9, Fermi field operators are no different
than other Fermi creation and destruction operators, except that their dimension is length−3/2
and their anti-commutation relations are expressed in terms of Dirac deltas instead of Kronecker
deltas.
Considering a state where the orthonormal single-particle states are α, β · · · , the expectation of
the density operator is X
x) =
ρ(⃗ Ψ†s (⃗
x)Ψs (⃗x). (11.29)
s

The expectation of the density operator in the state |α, β, · · · ...⟩ is

⟨α, β, γ, · · · |Ψ†s (⃗ x)|α, β, γ, · · · ⟩ = ⟨0|(aα aβ aγ · · · )Ψ†s (⃗

x)Ψs (⃗ x)(· · · a†γ a†β a†α )|0⟩
x)Ψs (⃗

(11.30)
X
= ϕ∗κ (s, ⃗
x)ϕκ (s, ⃗
x).
κ∈α,β···

Then density operator can also appear in a transition element. Consider the bra and ket to differ
in one label, γ → γ ′ ,
X
⟨α, β, γ ′ , · · · |Ψ†s (⃗ x)|α, β, γ, · · · ⟩ = ⟨0|(aα aβ aγ ′ · · · )Ψ† (⃗
x)Ψs (⃗ x)(· · · a†γ a†β a†α )|0⟩
x)Ψ(⃗
s
(11.31)
X
= ϕ∗γ ′ (s, ⃗
x)ϕγ (s, ⃗
x).
s

This simple result followed because it was assumed that the final single-particle states in the
bra, were used to describe the ket. Thus, the N -particle state was specified by listing N single-
particle labels. In more realistic calculations, one might study a transition between states where
many of the wave functions were continually altered as particles were added. For example,
in a many-electron atom, if one adds the N th particle it affects the structure of the previous
N − 1 particles. This assumption allows one to write the state as a single product of creation
operators acting on the vacuum. A product state is produced by operating with a single string of
orthogonal creation operators on the vacuum. More generally, a state might be comprised from a
linear combination of product states. Going beyond simple product states requires sophisticated
many-body treatments, which is beyond the scope of this class.

222
PHY 851/852 11 FERMIONS

Example 11.5: Correlations in a Fermi Gas

Here, we calculate correlations in a Fermi gas due to the Fermi nature of the particles. Con-
sider a one-dimensional gas of fermions moving in a large region of length L where the den-
sity corresponds to a Fermi momentum kf . Here, we suppress the spin degrees of freedom,
or equivalently we are considering only electrons of a given spin projection.

(a) Find the correlation function, relating the ratio of the probability of finding two particles
separated by r to the probability of finding the two particles at locations far removed
from each other.
Solution: To do this problem, we first write the two-particle probability,

ρ2 (x1 , x2 ) = ⟨ϕ|Ψ† (x1 )Ψ† (x2 )Ψ(x2 )Ψ(x1 )|ϕ⟩.

Y †
|ϕ⟩ = ak |0⟩.
k<kf

Commuting the field operators Ψ(x2 ) and Ψ(x1 ) toward the ket yields
   
Y † X eik2 x2 eik1 x1 Y †
Ψ(x2 )Ψ(x1 )  ak  |0⟩ = √ √ ak2 ak1  ak  |0⟩
k<kf k1 ,k2 <kf L L k<kf
 
ik2 x2 ik1 x1
X e e Y
= (±)(−)n1 +n2 √ √  a†k  |0⟩.
k1 ,k2 <kf L L k<kf ,k̸=k1 ,k̸=k2

Here, the sign (±) depends on whether a(k1 )a(k2 ) appear in the same/opposite order
as a† (k1 )a† (k2 ) in the product of creation operators. One can find a similar expression
for ⟨ϕ|Ψ† (x1 )Ψ† (x2 )|ϕ⟩,
 
Y
⟨0|  ak′  Ψ† (x1 )Ψ† (x2 )
k′ <kf
 
′
−ik2 x2 −ik1 x1 ′
′ ′ e e
X Y
= ⟨0|  ak′  (±)(−)n1 +n2 √ √ .
′ ,k′ <k
k1 f
′ ,k̸=k′
k′ <kf ,k′ ̸=k1
L L
2 2

The overlap of the two states is zero unless k1 and k2 from the ket are matched with
k1′ and k2′ in the bra. However, this can happen with either k1 = k1′ , k2 = k2′ , or with
k1 = k2′ , k2 = k1′ . For the first ordering, the two phases exactly cancel. The signs of
(−1)n··· are also identical and cancel. For the second ordering, the product of phases
yields ei(k1 −k2 )(x1 −x2 ) results, plus there is an extra sign due to the permutation of the
creation operators, ak1′ ↔ ak2′ . This then yields
X 1
⟨ϕ|Ψ† (x1 )Ψ† (x2 )Ψ(x2 )Ψ(x1 )|ϕ⟩ = 1 − ei(k1 −k2 )(x1 −x2 ) .


k1 ,k2 <kf
L2

223
PHY 851/852 11 FERMIONS

Because for every pair k1 , k2 , the same pair appears again as k2 , k1 . Thus, only the
real part of the phase factor ei(k1 −k2 )(x1 −x2 ) enters, and the matrix element is indeed
Hermitian.
Taking the ratio of the two-particle density to the square of the one particle densities
gives the correlation function.

P2 (x1 , x2 )
g(r ≡ x2 − x1 ) ≡ 2
P n
1 − ei(k1 −k2 )(x1 −x2 )


k1 ,k2 <kf
= P .
k1 ,k2 <kf

The sums can be changed into integrals,

g(r) = 1 − |I|2 ,
R
−kf <k<kf
dkeikr
I= R
−kf <k<kf
dk
sin(kf r)
= .
kf r

The two-particle density is thus

( 2 )
sin(kf r)

ρ2 (x1 , x2 ) = n2 1− , r ≡ |x1 − x2 |.
kf r

The correlation is displayed in Fig. 11.2.

(b) The particles now feel a mutual interaction,

V (r) = αΘ(a − r),

where r is again the relative distance between two particles. Find the exchange con-
tribution to the energy per unit length due to the interaction in first-order perturbation
theory.
Solution: In first-order perturbation theory the correction to the energy is

1
Z
⟨H⟩ = dx1 dx2 ⟨ϕ|Ψ† (x1 )Ψ† (x2 )Ψ(x2 )Ψ(x1 )|ϕ⟩V (x1 − x2 )
2 Z
1
=L dr⟨ϕ|Ψ† (0)Ψ† (r)Ψ(r)Ψ(0)|ϕ⟩V (r),
2
where we have used the fact that the two-particle density depends only on r = x1 − x2 .

224
PHY 851/852 11 FERMIONS

1.0

g(r) 0.5

0.0
0.0 2.0 4.0 6.0 8.0
kf r
Figure 11.2: The electron-electron correlation (for electrons of one sign) due to anti-symmetrization in a
one-dimensional system, as derived in Example 11.5.

By using the results of (a) one can find the energy per length,
∞
⟨H⟩ 1
Z
2
= n dr g(r)V (r)
L 2 −∞
(  )
αn2 a sin(kf r) 2
Z 
= dr 1 −
2 −a kf r
( a Z )
a
1  2 sin(kf r) cos(kf r)
= αn2 a + 2 sin2 (kf r) − dr

kf r  0 kf r
0
( )
2
sin2 (kf a) 1
= αn a + 2
− Si(2kf a)
kf a kf
( )
2
sin (πna) 1
= αn2 a + − Si(2πna) ,
π 2 n2 a πn

where Si is the sine-integral function, and n is the number of particles per unit length.
Note that the first term describes the potential energy one would expect from particles
with a uniform density interacting through the potential V (x1 − x2 ). This is called the
direct term, while the remainder is referred to as the exchange term.

225
PHY 851/852 11 FERMIONS

11.7 Exercises
1. Consider a two-dimensional zero-temperature non-relativistic gas of identical spin-up fermions
whose mass is m. They are confined in a two-dimensional box of dimensions, Lx and Ly .
The quantum numbers characterizing the single-particle eigenstates are nx and ny . The
box is divided in half along the x axis. The eigenstates with odd nx now disappear, while
there are now two solutions (one for each half of the box) for each even value of nx . As-
suming the size of the box is large compared to the inverse Fermi momentum, find the
penalty, expressed as an energy per unit length (∆E/2Ly ), for dividing the box. Express
your answer in terms of the Fermi momentum and the mass.
2. In the interior of a neutron star, the neutron-to-proton ratio is very high. This results de-
spite the fact that the proton mass is 1.3 MeV higher than the neutron mass. This occurs
because protons must balanced by an equal number of electrons. Furthermore, protons
and neutrons may be interchanged via the reaction,
p+e↔n
(We have neglected the neutrinos in this reaction because they are free to leave the star due
to their massless nature.)
The masses of the particles are:
mp c2 = 938.27 MeV, mn c2 = 939.57 MeV, me c2 = 0.511 MeV
For the problems below, you may assume the protons and neutrons are non-relativistic,
E = mc2 + (ℏck)2 /(2mc2 ), but the electrons must be treated as relativistic particles,
E = [(ℏck)2 + (mc2 )2 ]1/2 . It is useful to remember that ℏc = 197.326 MeV·fm.

(a) If the baryon(neutrons or protons) density equals nB , express the corresponding con-
straint involving the Fermi momenta of the protons and neutrons.
(b) From the constraint that the system is electrically neutral, describe how the proton’s
Fermi momentum is related to the electron’s Fermi momentum.
(c) Describe how minimizing the overall energy results in a constraint involving the three
Fermi momenta.
(d) For 4 cases, nB = 0.0001, 0.001, 0.1, 1.0 baryons per cubic Fermi, solve the expres-
sions above and plot the neutron/proton ratio as a function of nB . Solving the three
equations may involve finding roots numerically.

3. Calculate the magnetic susceptibility of a two-dimensional electron gas with Fermi wave
number kf .
4. Consider the wave function of three identical particles in Eq. (11.16) where the three single-
particle wave functions are orthonormalized.
(a) Show that
1
Z
d3 r1 d3 r2 d3 r3 |ϕα,β,γ (⃗
r1 , ⃗ r3 )|2 = 1
r2 , ⃗ (11.32)
3!
if the three indices are different, α ̸= β, α ̸= γ, β ̸= γ, and that the overlap is zero if
any two of the indices are the same.

226
PHY 851/852 11 FERMIONS

(b) Show that ϕα,β,γ (⃗

r1 , ⃗ r3 ) = 0 if any two positions are the same.
r2 , ⃗

5. Correlation/anti-correlation in a quantum gas: Consider a uniform gas of non-interacting

spin-half particles in the ground state. The wave function may be written
Y
|ϕ⟩ = a†α |0⟩,
α,|kα |<kf

where the product includes all states α with momentum kα < kf and spin sα . The
density-density correlation function is defined as

⟨ϕ|Ψ†s1 (⃗
x1 )Ψ†s2 (⃗
x2 )Ψs2 (⃗
x2 )Ψs1 (⃗
x1 )|ϕ⟩
gs1 ,s2 (x⃗2 − ⃗
x1 ) ≡
⟨ϕ|Ψ†s1 (⃗ x1 )|ϕ⟩⟨ϕ|Ψ†s2 (⃗
x1 )Ψs1 (⃗ x2 )Ψs2 (⃗
x2 )|ϕ⟩

This function expresses the correlation of two particles with spin s1 and s2 being separated
x2 −x⃗1 . It is defined in such a way that it is unity if the probability of seeing two particles
by ⃗
at ⃗
x1 and ⃗
x2 is the product of the probabilities of observing each particle independently.

(a) Show that the density-density correlation function can be written as

⃗ ⃗
ei(kα −kβ )·(⃗x2 −⃗x1 )
P
kα ,kβ
gs1 ,s2 (x⃗2 − ⃗
x1 ) = 1 − δs1 s2 P ,
kα ,kβ

where the sums are over all momentum states with kα < kf and kβ < kf .
(b) By changing the sum over states to a three-dimensional integral over ⃗ k, find an an-
alytic expression for the density-density correlation function in terms of the Fermi
momentum kf .

6. Again, consider a non-interacting quantum gas of particles of mass m, with the ground
state being expressed as Y
|ϕ⟩ = a†α |0⟩,
α,|kα |<kf

where the product includes all states α with momentum kα < kf . For this problem,
ignore the spin indices (as if only spin ↑ particles existed. Consider an interaction between
the particles of the form,
1
Z
Hint = d3 r1 d3 r2 V (⃗ r2 )Ψ† (⃗
r1 − ⃗ r1 )Ψ† (⃗
r2 )Ψ(⃗
r2 )Ψ(⃗
r1 )
2
V (⃗
r1 − ⃗ r2 ) = βδ(⃗r2 − ⃗r1 )

Find the first-order perturbative correction for the energy, ⟨ϕ|Hint |ϕ⟩, for particles in the
gas in terms of kf , β and m. Use the result from the previous problem assuming only one
spin exists,
" #2
sin(kf ∆x)/∆x3 − kf cos(kf ∆x)/∆x2
C(∆x) = 1 − .
kf3 /3

227
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

12 Non-Perturbative Approaches for Many-Fermion Systems

In the previous chapter, interactions between fermions were mainly ignored. A one-body inter-
action is typically introduced via a spatial potential V (⃗
r ). Even if this is a potential that depends
on the relative position of two particles, V (⃗
r2 − ⃗
r2 ), one can work in center of mass coordinates
and effectively reduce the problem to a two-body problem to a one-body problem (at least in
non-relativistic physics). Bound state and scattering observables can then all be calculated with
theories that involve only the relative coordinate, ⃗ r = ⃗ r1 − ⃗r2 . As soon as more than two
particles are present, the problem no longer reduces to a one-body problem. In this chapter
we focus mainly on such problems where there is an external potential and ignore interaction
terms between constituents. This allows one to assume that the many-particle wave functions
are products, or anti-symmetrized products, of single particle wave functions. More generally,
interaction terms simultaneously modify the behavior of more than one particle at a time and
the true solutions cannot be expressed as products. Such treatments of more realistic interactions
are typically under the moniker of “many-body theory”, which is largely beyond the scope of
this course. Nonetheless, the one-body treatments, although somewhat crude, provide a great
deal of insight and in some cases can be remarkably accurate. One example is the Hartree-Fock
approximation, which is the topic of the Sec. 12.3.

12.1 Interacting Fermi Systems – Two and Four-Point Interactions

The simplest sort of interaction is a “two-point” interaction.
X
H = Vαβ a†α aβ . (12.1)
α,β

A two-point interaction has one creation and one destruction operator in each term. The inter-
action alters the momentum of a single particle, which makes it a one-body interaction. The
particles behave independently. To demonstrate what is meant by “independent”, consider the
product state

|ϕ(t = 0)⟩ = a†δ · · · a†β a†α |0⟩. (12.2)

Imagine one were to make a linear transformation of the creation operators using a unitary
matrix U ,

b†α = Uαβ a†β . (12.3)

These new operators would also satisfy the anti-commutation laws,

{bα , b†β } = Uγα

†
Uβδ {aγ , a†δ } (12.4)
= {aα , a†β } = δαβ .

Thus, a unitary matrix acting on the space of creation operators results in a new set of operators
that still obey the anti-commutation laws. Further, if that unitary matrix is chosen to be the same

228
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

matrix that diagonalizes Vαβ ,

X
†
H = Vαβ Uγα Uβδ b†γ bδ (12.5)
αβ

= (U † V U )αβ b†α bβ .

Thus, if there is a two-body Hamiltonian, one can transform to a basis where the Hamiltonian is
diagonalized, i.e.
X
H = Eα b†α bα . (12.6)
α

Eigenstates of the n − body problem can then be expressed as a product state,

Y
|αβ...γ⟩ = b†α b†β · · · b†γ |0⟩, (12.7)
αβ···γ

and the eigenenergies are E = Eα + Eβ + · · · + Eγ .

This is not the case if the Hamiltonian has terms such as

Vαβγδ a†α a†β aγ aδ . (12.8)

These terms take a pair of particles in one-body states γ and δ and put them into states α and β.
Scattering does precisely this. If one writes a two-body potential with field operators, e.g.,

1
Z
d3 r1 d3 r2 Ψ† (⃗
r1 )Ψ† (⃗
r2 )V (⃗
r1 − ⃗
r2 )Ψ(⃗
r2 )Ψ(⃗
r1 ), (12.9)
2
four-body terms are inevitable.
However, if the the potential is from an external one-body potential,
Z
V = d3 r V (r)Ψ† (⃗
r )Ψ(⃗
r ), (12.10)

only two-body terms appear, and one can find eigenstates that are products of creation and
destruction operators. Unfortunately, physics is not usually so simple. If all interactions were of
the two-point type, all many-body problems could be solved by diagonalizing the matrix of the
one-body problem.
An interaction between particles separated by ⃗ r=⃗ r1 − ⃗r2 is written classically as

1
Z
Hint = d3 r1 d3 r2 ρ(⃗
r1 )ρ(⃗
r2 )V (⃗
r1 − ⃗r2 ). (12.11)
2
The factor of 1/2 corrects for double counting because the integral is without a r1 < r2 qualifier.
Written in terms of field operators, this interaction becomes

1
Z
Hint = d3 r1 d3 r2 Ψ† (⃗
r1 )Ψ† (⃗
r2 )V (⃗
r1 − ⃗
r2 )Ψ(⃗
r2 )Ψ(⃗
r1 ). (12.12)
2

229
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

For the moment, we will omit spin indices and assume only particles of a specific spin are in-
volved. One peculiar aspect of the Hamiltonian is that the two Ψ† s are on the left while the two
Ψs are on the right, whereas the product of densities would suggest a product of Ψ† Ψs. How-
ever, such a product would result in an interaction energy for a one-particle state, as the particle
would interact with itself.
Four-point interactions (or three-point interactions, e.g. the electromagnetic interaction J⃗ · A)⃗
are infinitely more complicated than two-point interactions. For one, if one wishes to calculate
the evolution of a†α (t), one needs to commute the Hamiltonian with a†α . However, if Hint is a
product of four creation and destruction operators, commuting Hint with aα or a†α results in a
product of three operators. This makes it impossible to consider the problem as a simple mixing
of the creation and destruction operators. The eigenstates of a Hamiltonian with three-point
or four-point terms is not a product state, but is instead a complicated linear combination of
product states. Such problems are rarely solvable, forcing one to resort to approximations. This
is what is meant by the many-body problem, a moniker which encompasses most of the effort
in theoretical physics. Examples of such approximation methods are perturbation theory or the
Hartree-Fock approximation.

12.2 The Thomas-Fermi Approximation

A very crude way to find the density of fermions in an external well is the Thomas-Fermi ap-
proximation. This approximation is semiclassical and assumes that the local density is a function
of the local potential only,

2s + 1 3
ℏ2 kf (⃗
r )2
r) =
n(⃗ kf (⃗
r) , + V (⃗
r ) = µf . (12.13)
6π 2 2m
The density would vanish outide the region where V (⃗ r ) < µf . Given the potential and the
chemical potential µf , it is straight forward to find the density. For a finite system with a fixed
particle number, one would have to adjust µf to get the correct total number of particles. The
Thomas-Fermi approximation is crude, and gives rather non-sensical results for atoms due to the
small number of electrons in atoms. It is less unreasonable for estimating the electron density
near surfaces. As a general rule it is valid if the systems are large and the potential changes
slowly.
Example 12.1: The Thomas Fermi Approximation for the Harmonic Oscillator
Calculate the Fermi energy (as measured from the bottom of the harmonic oscillator) for N
electrons in a one-dimensional harmonic oscillator characterized by angular frequency ω.
Solution: The density of electrons for a potential V (x) in one dimension is:
kf
1
Z
n(x) = (2s + 1) dk
2π −kf
2
= kf (x),
π
2 q
= 2m(µf − V (x)).
ℏπ

230
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

For the harmonic oscillator, this gives

Z xmax
N = dx n(x)
−xmax
p
4 2mµf xmax
Z p
= dx 1 − α2 x2 ,
ℏπ 0
s r
mω 2 2µf
α= , xmax = .
2µf mω 2

Substituting u = αx,
1
8µf
Z p
N = du 1 − u2
ℏωπ 0
2µf
= .
ℏω
If one were to set µf = (n + 1/2)ℏω, the true ground state energy of the harmonic oscillator,
one would find N = 2n + 1. Whereas if one filled all the levels, each with a spin-up and a
spin-down fermion, the number of particles would be 2n. This is a good approximation for
large N , but is poor for a small number of particles. The conditions under which this method
becomes accurate is similar to the conditions discussed for the WKB approximation in Sec.
6.1.

12.3 The Hartree-Fock Approximation

The Hartree-Fock approximation yields an expression, that when solved gives an approxima-
tion to the ground state of a many-fermion system. The approximation is variational in nature.
But rather than varying a few parameters to modify the wave function, the value of the wave-
function at each point is treated as a variational parameter. In a variational approximation one
assumes some form for the variational wave function. In this case, one assumes that the solution
is in some sort of product state,
|ϕ⟩ = a†δ · · · a†β a†α |0⟩. (12.14)
The many-particle wave function, ⟨⃗ rn |a†n · · · a†1 |0⟩, will not be expressed as a product of
r1 · · · ⃗
single-particle wave functions. But that is only a matter of anti-symmetrization. Given one of
the simple product terms, the other terms of the many-particle wave-function can be generated
by permuting the labels of the first term.
To derive the Hartree Fock equations, one writes down the expectation of the Hamiltonian,
ℏ2 2
Z   Z
†
⟨ϕ|H|ϕ⟩ = ⟨ϕ| d rΨ (⃗3
r) − ∇ Ψ(⃗ r ) + d3 rU (r)Ψ† (⃗ r )Ψ(⃗
r )|ϕ⟩ (12.15)
2m
1
Z
+ ⟨ϕ| r1 )Ψ† (⃗
d3 r1 d3 r2 V (r⃗2 − ⃗ r1 )Ψ† (⃗
r2 )Ψ(⃗
r2 )Ψ(⃗
r1 )|ϕ⟩,
2
where the spin indices have been omitted temporarily.

231
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

One can now express the energy in terms of the wave functions of the occupied states α.

ℏ2 2
XZ  
3 ∗
⟨ϕ|H|ϕ⟩ = d r ϕα (⃗r) − ∇ + U (⃗ r ) ϕα (⃗ r) (12.16)
α
2m
1X
Z
d3 rd3 r ′ V (⃗ r ′ ) ϕ∗α (⃗
r )ϕ∗α′ (⃗
r ′ )ϕα′ (⃗
r ′ )ϕα (⃗
r ) ± ϕ∗α (⃗
r )ϕ∗α′ (⃗
r ′ )ϕα (⃗
r ′ )ϕα′ (⃗

+ r−⃗ r) .
2 α,α′

The second term in the two-body interaction, where the ± refers to bosons/fermions, is known
as the exchange term, with the name coming from the exchange of the α and α′ indices.
The next step in the variational procedure is to minimize ⟨ϕ|H|ϕ⟩ with respect to changes in
the wave functions subject to the constraint that each wave function is properly normalized.
One accounts for the constraint by multiplying the constraint by a Lagrange multiplier λγ then
adding it to the function one wishes to minimize,
 
δ
Z
3 ∗
⟨ϕ|H|ϕ⟩ − λγ d rϕγ (⃗ r )ϕγ (⃗
r ) = 0. (12.17)
δϕγ (⃗
r)

More correctly, one would vary both the real and imaginary parts of ϕγ which is equivalent to
varying either ϕγ or ϕ∗γ . The resulting expression is

ℏ2 2
 
λγ ϕγ (⃗
r) = − ∇ + U (⃗ r ) ϕγ (⃗ r) (12.18)
2m
XZ
d3 r ′ V (⃗ r ′ ) ϕ∗α′ (⃗
r ′ )ϕα′ (⃗
r ′ )ϕγ (⃗
r ) ± ϕ∗α′ (⃗
r ′ )ϕγ (⃗
r ′ )ϕα′ (⃗

+ r−⃗ r) .
α′

The sum over all α′ can be rewritten as a sum over all α′ ̸= γ, because the α′ = γ term vanishes.
The two terms in Eq. (12.18) involving V (⃗ r−⃗ r ′ ) are known as the Hartree and Fock terms
respectively. The Hartree term looks like a potential felt by the particles due to the presence
of the other particles. Thus, the Hartree equation can be solved by self-consistently finding the
solutions to the effective potential
XZ
VHartree (⃗
r) = d3 r ′ V (⃗ r−⃗ r ′ )ϕ∗α′ (⃗
r ′ )ϕα′ (⃗
r′ ) (12.19)
α′

The Hartree equations are often solved iteratively. One guesses at the wave functions, finds the
Hartree-potential, solves the Schrödinger equation for the single-particle wave functions, then
iterates the procedure until the wave functions converge. One issue with the Hartree approxi-
mation is that a particle also feels the potential due to itself unless one restricts the sum over α′
to ignore the term α = γ in Eq. (12.18). If one excludes this term, then the potential is different
for each state, which then leads to the solutions ϕγ not being orthogonal.
The Fock term presents a different challenge because ϕγ (⃗ r ) cannot be factored out of the differ-
r ′ ) appears in its place.
ential equation. The Fock term is non-local in that ϕγ (⃗
The Lagrange multiplier λγ plays the role of a single-particle energy. However, one should re-
member that because the interaction energy between two states is contained in the single-particle

232
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

energy for both states, the total energy is not the sum of the λγ s. The Hartree-Fock approxima-
tion is especially useful for estimating density distributions or even changes in energies (e.g.
separation energies), but is not particularly accurate at calculating total binding energies.
Finally, we return to the problem of including the spin indices. When spin indices are included
the Fock term disappears unless the two states are the same species and the same spin. Hence a
factor δs,s′ accompanies the Fock term.

12.4 Hartree-Fock for Atoms

The principal difficulty in handling multi-electron atoms comes from the fact that the electrons
interact with one another. If it were not for the mutual interactions between electrons, one could
treat the electrons independently. Of course, this would yield horribly unphysical results. For
instance, one could place an infinite number of electrons into the the hydrogen atom, where in
reality no more than two electrons can be bound to hydrogen.
The Hartree approximation is the first method one might consider for treating such atoms. One
then solves the Schrödinger equation with the potential,

Ze2 XZ e2
Vi (⃗
r) = − + d3 r ′ |ϕj (⃗
r ′ )|2 (12.20)
r j̸=i
|⃗ r′ |
r−⃗

If there are n electrons, one must solve n coupled equations.

One can go one step further and solve the Hartree-Fock equations,

ℏ2 ∇2 Ze2 e2
  XZ
− − ϕi (⃗
r )+ d3 r ′ ϕ∗ (⃗
′| j
r ′ ) [ϕj (⃗
r ′ )ϕi (⃗
r ) − δs,s′ ϕj (⃗ r ′ )] = λi ϕi (⃗
r )ϕi (⃗ r)
2m r j
|⃗
r − ⃗
r
(12.21)
to find the wave functions. One can then find the energy by calculating ⟨ϕ|H|ϕ⟩. Again, this
requires solving coupled differential equations, but the Fock term also makes the equations non-
local which brings along an added computational difficulty.
The difference of the energy found by solving the Hartree-Fock equations and the Hartree equa-
tions is the exchange energy. Because the Hartree-Fock wave functions force the electrons to stay
away from one another due to anti-symmetrization, the repulsive Coulomb interaction between
electrons is weakened which means that the exchange energy is negative, affecting the solution
like an attractive contribution to the potential.
The Hartree-Fock equations were derived with the assumption that the solution is a product
state. As that is a variational assumption, the true ground state energies are about 1 eV lower
than Hartree-Fock solutions. This difference is referred to as “correlation” energy. It can be
calculated perturbatively by considering the mixing of one-particle-one-hole and two-particle-
two-hole states into the wave functions. This is done by considering the n-electron Hartree-Fock
ground state as the ground state. Then one considers excited states as being those n-particle n-
hole states formed by considering n of the single-particle solutions of the Hartree-Fock ground
state to be replaced by n single-particle solutions from the set of single-particle states left unfilled
in the Hartree-Fock ground state.

233
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

12.5 The Periodic Table

Single-electron wave functions are labeled by n, ℓ, s, j and mj . In an atom with only one elec-
tron, the energy depends principally on n, as states with same n but different ℓ = 0, 1, · · · n − 1
are degenerate aside from the spin-orbit interaction. Of course, s always equals 1/2 and j can
be either ℓ + 1/2 or ℓ − 1/2.
The existence of the other electrons destroys the “accidental” degeneracy of the hydrogen atom
and allows states with different ℓ and the same n to have significantly different energy. This
is due to the screening of the positive charge. The “accidental” degeneracy allowed states with
fewer nodes in the radial wave function but larger angular momentum to have the same en-
ergy as states with smaller angular momentum but more nodes in the radial wave function. By
screening the charge, an advantage is created for states that have a relatively greater probability
of being near the origin. Because radial wave functions behave as r ℓ near the origin, a state with
a lower ℓ but the same n will have lower energy due to screening. This difference can be large
enough at times to allow a state to move lower than states of higher n but higher angular mo-
mentum. For instance, in some cases the 4s states can move below the 3d states, and similarly
the 4d and 5s shells compete as well. The 5s shell is always well below the 4f shell. When shells
compete, e.g. the 4s and 3d electrons, the choice of orbitals is non trivial. In these cases the con-
figuration can vary from one element to the next, and in fact, solutions might contain a mixture
of configurations. Chemical properties are determined largely by the outermost electrons. When
shells are filled, the elements are less reactive. The rare gases (also know as inert or noble gases)
all have filled p shells with the exception of Helium.
The set of orbitals with a specific n and ℓ is known as a “shell”. The degeneracy of a shell is
4ℓ + 2. Electronic “configurations” are labeled by the shells and the filling, e.g.,

(1s)2 (2s)2 (2p)3 ,

with the superscript labeling the number of electrons in the shell.

12.6 Configuration Splitting and Hund’s Rules

Different configurations are generally split by a few electron volts. There are 4ℓ + 2 single-
particle orbitals in a shell, and the number of ways to arrange several electrons among these
levels can be rather large. For multi-electron atoms one first considers the single-particle orbitals
with a given radial quantum number n and orbital quantum number ℓ (of course s = 1/2). If
all the levels of a given shell are filled, the net angular momentum of that shell is zero. But if
the shell is not either completely filled or completely empty, one might be able to configure the
electrons into numerous states. The energies of these configurations depends on the total (of the
particles in the semi-filled shell) spin angular momentum S, the total orbital angular momentum
L, and the total momentum J . Ignoring the spin-orbit interaction, the total angular momentum
⃗ and total spin S
L ⃗ of the electrons commute with the Hamiltonian. The (2L + 1)(2S + 1)
states of an LS multiplet are then further split by the spin-orbit interaction. Thus, the spin-orbit
interaction invalidates ML and MS as good quantum numbers, and replaces them with J and
MJ .
Hund’s rules determine which LSJ combination has the lowest energy:

234
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

1. The largest S is most favorable as it has the most symmetric spin wave functions, there-
fore totally anti-symmetric spatial wave functions. This minimizes the probability that
electrons are close to one another, and therefore minimizes their repulsive interaction.
2. For a given S, the LS multiplet with the largest L has the lowest energy. This can be under-
stood by considering two electrons rotating about a nucleus. If L is large, both electrons
are moving about with essentially the same orbit as m1 and m2 can be large and of the
same sign. As such electrons spend less time crossing one another’s path they miminize
their repulsive interaction.
3. Different LS multiplets are typically split by tenths of electron volts. A given LS mul-
tiplet is again split by the spin-orbit interaction but only by a few hundredths or a few
thousandths of an electron volt. The spin-orbit interaction gives preference to states with
the minimum J when the shells are less than half filled and behaves in the opposite way
when the shells are more than half filled.

Configurations are labeled by S, L and J , which should not be confused with the ℓ of the single-
particle orbitals. The notation is
2S+1
LJ .
For instance, the state 3 D3 would have the spins coupled to S = 1, L = 2 and J = 3. Note the
upper-case angular momentum labels S, P, D, F · · · .
It should be remembered that considering only ways to arrange particles in a given shell is an
approximation. There is no reason that the true ground state shouldn’t have some mixture of
states from different shells.

12.7 ⃗ and Total S

Constructing States According to Total L ⃗

The overall wave function for a many-electron state described by L, S and J can be written as
the sum over products of angular and spin wave functions.
X
|L, S, J, MJ ⟩ = ⟨J, MJ , |L, ML , S, MS ⟩|L, ML ⟩|S, MS ⟩ (12.22)
ML ,MS

The angular/spin wave functions |L, ML ⟩|S, MS ⟩ must first be written in terms of products of
singular particle wave functions summed over with the help of Clebsch-Gordan technology.
As an example we consider a two-electron state with orbital/spin wave functions in a shell of

235
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

angular momentum ℓ.

ΨL,S,J,MJ (Ω, Ω′ ; ms , m′s ) = ⟨Ω, Ω′ ; ms , m′s |L, S, J, MJ ⟩ (12.23)

X
= ⟨J, MJ |L, ML , S, MS ⟩⟨Ω, Ω′ |L, ML ⟩⟨ms , m′s |S, MS ⟩
ML ,MS
X
= ⟨J, MJ |L, ML , S, MS ⟩
ML ,MS
X
· ⟨L, ML |ℓ, mℓ , ℓ, m′ℓ ⟩Yℓ,mℓ (Ω)Yℓ,m′ℓ (Ω′ )
mℓ ,m′ℓ

· ⟨S, MS |1/2, ms , 1/2, m′s ⟩

X
= ⟨J, MJ |L, ML , S, MS ⟩⟨L, ML |ℓ, mℓ , ℓ, m′ℓ ⟩
ML ,Ms ,mℓ ,m′ℓ

⟨S, MS |1/2, ms , 1/2, m′s ⟩Yℓ,mℓ (Ω)Yℓ,m′ℓ (Ω′ ).

The expression would be even more complicated if there were three electrons in the shell.
Example 12.2: Electrons in Carbon
Here we construct the 1 S0 state in Carbon. The two electrons in the p shell of Carbon can be
in any number of configurations. Using the expression above,

ΨL=0,S=0,J=0,MJ =0 (Ω, Ω′ , ms , m′s ) (12.24)

X
= ⟨J, MJ |L, ML , S, MS ⟩⟨L = 0, ML = 0|ℓ = 1, mℓ , ℓ = 1, m′ℓ ⟩
mℓ ,m′ℓ

· ⟨1/2, 1/2, S = 0, MS = 0|ms , m′s ⟩Yℓ=1,mℓ (Ω)Yℓ=1,m′ℓ (Ω′ )

X
= ⟨L = 0, ML = 0|ℓ = 1, mℓ , ℓ = 1, m′ℓ ⟩Yℓ=1,mℓ (Ω)Yℓ=1,m′ℓ (Ω′ )
mℓ ,m′ℓ
1

· √ δms ,1/2 δm′s ,−1/2 − δms ,−1/2 δm′s ,1/2
2
1
= √ {Y1,1 (Ω)Y1,−1 (Ω′ ) + Y1,−1 (Ω)Y1,1 (Ω′ ) − Y1,0 (Ω)Y1,0 (Ω′ )}
6

· δms ,1/2 δm′s ,−1/2 − δms ,−1/2 δm′s ,1/2

Note that the wave function is symmetric with respect to interchange of Ω with Ω′ and anti-
symmetric with respect to interchange of ms and m′s , making the overall wave function anti-
symmetric.

12.8 Permutation Symmetry

When coupling two particles from the same shell (Thus ℓ1 = ℓ2 ) together the permutation
symmetry goes as (−1)L for the spatial part, while the interchange of spins is symmetric/anti-
symmetric for S = 1/0. For this reason one can notice that for two electrons in the (2p) shell

236
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

(carbon) the orbital states must be L = 0 or L = 2 if the the spin state is S = 0 and must be
L = 1 if S = 1. Thus, the possible states in carbon are 1 S0 , 1 D2 , 3 P0 , 3 P1 and 3 P2 .
Coupling a higher number of particles together can lead to “mixed” symmetries. In this case the
angular wave functions may be neither symmetric or anti-symmetric, while the spin wave func-
tions might be mixed as well. However, the overall wave function needs to be anti-symmetric.
For example, the flavor and spin wave functions of the three quarks that constitute a proton or
neutron are in states of mixed symmetry.

12.9 Zeeman Effect and the Landé g Factor

A particle in a magnetic field feels the interaction,

eB
Hmag = − (Lz + 2Sz ), (12.25)
2mc
where the magnetic field is assumed to point in the z direction. The factor g = 2 before Sz is
natural in the Dirac equation, though higher-order terms in perturbation theory provide correc-
tions of order e2 /ℏc = 1/137.036. We wish to calculate the change in energy for an atom in a
state of good J, MJ due to the interaction

eB
∆ELSJMJ = − ⟨L, S, J, MJ |Lz + 2Sz |L, S, J, MJ ⟩ (12.26)
2mc
eB
=− (ℏMJ + ⟨L, S, J, MJ |Sz |L, S, J, MJ ⟩) .
2mc

The challenge in calculating the splitting comes from finding ⟨Sz ⟩. We expect this to be propor-
tional to MJ as it is the only label available. The Wigner Eckart theorem and Clebsch-Gordan
technology come to the rescue.

⟨L, S, J, MJ |Sz |L, S, J, MJ ⟩ ⟨L, S, J ||S||L, S, J ⟩

= . (12.27)
⟨L, S, J, MJ |Jz |L, S, J, MJ ⟩ ⟨L, S, J ||J ||L, S, J ⟩

This follows by applying the Wigner Eckart theorem to both ⟨Sz ⟩ and ⟨Jz ⟩ and noticing that the
same Clebsch-Gordan coefficients appear in both terms.
Thus, if we can find the ratio of the reduced matrix elements we will have fulfilled our mission.
To do this, we first step aside to perform a proof. Consider a vector operator A,⃗ which will be
⃗ further below. Using the identity,
chosen to be S
X
⟨J M |A⃗ · J⃗|J M ⟩ = ⟨J M |Am′ J−m′ |J M ⟩, (12.28)
−1≤m′ ≤1

then applying completeness,

X
⃗ · J⃗|J M ⟩ =
⟨J M |A ⟨J M |Am′ |J M ′′ ⟩⟨J M ′′ |J−m′ |J M ⟩, (12.29)
−1≤m′ ≤1,−J≤M ′′ ≤J

237
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

Note that the inserted states were only those within the same J M multiplet. This is valid be-
cause Jm does not mix different multplets. Now, by applying the Wigner Eckart theorem, one
can write the matrix element ⟨A⃗ · J⃗⟩ as

⃗ · J⃗|J M ⟩ = f (J, M )⟨J ||A||J ⟩⟨J ||J ||J ⟩

⟨J M |A (12.30)
X
f (J, M ) =
−1≤m′ ≤1,−J≤M ′′ ≤J
1
· ⟨1, J, J, M |m′ , M ′′ ⟩⟨1, J, J, M | − m′ , M ′′ ⟩
2J + 1
Because f is determined solely by Clebsch-Gordan coefficients, one can see that

⟨J M |A⃗ · J⃗|J M ⟩ ⟨J ||A||J ⟩

= (12.31)
⟨J M |J⃗ · J⃗|J M ⟩ ⟨J ||J ||J ⟩

Now that our proof is finished, as we can see that

⟨L, S, J ||S||L, S, J ⟩ ⟨L, S, J, MJ |J⃗ · S|L,

⃗ S, J, MJ ⟩
= (12.32)
⟨L, S, J ||J ||L, S, J ⟩ ⟨L, S, J, MJ |J⃗ · J⃗|L, S, J, MJ ⟩
J (J + 1) + S(S + 1) − L(L + 1)
= ,
2J (J + 1)
where we have taken advantage of the fact that
1h i
J⃗ · S
⃗= −(J⃗ − S)
⃗ 2 + J⃗2 + S
⃗2 (12.33)
2
ℏ2
= [J (J + 1) + S(S + 1) − L(L + 1)].
2
Finally, we are able to insert our result for the ratio of the reduced matrix elements into our
expression for the splitting to get
eℏB
∆E = −g MJ (12.34)
2mc
J (J + 1) + S(S + 1) − L(L + 1)
g =1+ .
2J (J + 1)
The factor g is known as the Landé g factor, with the g referring to gyration.
In some undergraduate modern physics books the Landé g factor is derived from simple ge-
ometric arguments. If one considers the vector J precessing about the z axis and the vector
S precessing about J⃗, one would expect that the expectation of S
⃗ for a given orientation of J⃗
would be
⃗ · J⃗
S
⃗ = J⃗
⟨S⟩ , (12.35)
|J |2
⃗ precesses about J⃗, the component of S
because as S ⃗ perpendicular to J⃗ averages to zero. Finally,
averaging over the direction of J⃗ gives the result we expected.

238
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

12.10 Molecules and the Adiabatic Approximation

The force felt between atoms largely derives from the distortion of the electronic wave functions
due to the proximity of a second potential. Because atoms move slowly with respect to the
electronic motion, electronic wave functions largely adjust in such a way as to remain in the
ground state for a given atomic separation. If N atoms are centered at positions R ⃗1 · · · R
⃗N,
one can calculate the net electronic binding energy E(R⃗1 · · · R
⃗ N ) for the given positions of the
atoms. One then solves a Schrödinger equation for the atomic positions, Ψ(R ⃗1 · · · R
⃗ N ), using
the net electronic binding energy as the potential. This is referred to as the adiabatic or Born-
Oppenheimer approximation.
Examples where the adiabatic approximation were applied included a particle in a slowly ex-
panding box. If a box slowly expandd or contracts, a particle remains in the ground state of the
box. Of course, this implies that the energy of the particle is not kept constant. For the example
of an expanding box, the ground state energy falls as the volume expands, and the lost energy
appears in the kinetic energy of the piston. In fact, that energy is equal to the work, P dV , done
by the expansion. In the case of atoms, the gain or loss of the electronic energy appears as a loss
or gain of the ion’s kinetic energy. The electronic energy, calculated as a function of the ion’s
separation R, serves as a potential for the atoms. In a previous exercise, 6.8, one was asked to
find the energy felt by two electrons in harmonic oscillator potentials situated far apart, such
that they felt a dipole-dipole interaction from the electromagnetic interaction. Calculating the
correction to the energy, it was found that the energy went as 1/R6 . This energy would then
serve as a potential energy for understanding the behavior of two atoms if each well was given
a mass and one were to solve a Schroödinger equation using this energy as a potential. How-
ever, in this case this only gives the potential at larger distances. Different approaches would be
necessary to estimate the electronic energy for two harmonic oscillators, or two atoms, brought
into closer contact.
Example 12.3: Interacting with Distant Ion
For this example we consider an electron in a well interacting with a distant positive ion,
where we assume the states available to the electron are described as those of a harmonic
oscillator with frequency ω. This differs from the exercise 6.8 in that the distant particle is not
another charged particle in a well, which would be overall neutral, but carries a charge Ze.
The ion is placed a distance R from the atom. The perturbative potential between the ion and
the atom is
( )
1 1
Z
Hint = Ze2 − d3 rρ(⃗ r)
R ⃗ −⃗
|R r|
z
Z
≈ −Ze2 d3 rρ(⃗ r) 2
R
2
Ze
= − 2 zop .
R

239
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

One can then calculate the energy to second order perturbation theory,

Z 2 e4 |⟨nz = 1|zop |0⟩|2

∆E (2) = −
R4 ℏω
Z 2 e4
=− .
2R4 mω 2
This energy serves as the potential between the atom and the ion. It shows that the well is
attracted toward the distant ion. Note that if one were to consider an induced dipole moment
proportion to the electric field, multiplied by the electric field which falls off as 1/R2 , one
would have expected the 1/R4 behavior. This can be considered as an interaction of a charge
with an induced dipole moment, as opposed to the exercise 6.8, which was a dipole-dipole
interaction, or more accurately an induced-dipole-induced-dipole interaction.

Example 12.4: 2 H positive ion

Here, we consider the + H2 ion, where two protons are separated by a distance R, with the ad-
dition of a single electron. In the adiabatic approximation the binding energy of the electron,
ϵ(R), serves as the potential between the protons when combined with the proton-proton in-
teraction. At large distances the binding energy is one Rydberg, while one would expect an
increase in the magnitude of the binding energy as R is shortened. At very short distances,
one expects the interaction to become repulsive when one penetrates the electronic cloud and
the interaction is dominated by the Coulomb interaction between the protons.
Calculating the binding energy as a function of R can be crudely accomplished with a varia-
tional calculation assuming a trivial form of the wave function,

r ) = C± [ψA (⃗
ψ(⃗ r ) ± ψB (⃗
r )] ,

where ψA and ψB are the bound state wave functions of an electron to each of the protons,
⃗
r ) = (πa30 )−1/2 e−|⃗r−RA |/a0 ,
ψA (⃗
⃗
r ) = (πa30 )−1/2 e−|⃗r−RB |/a0 .
ψB (⃗

The factor C± is merely a normalization constant,

1
C± = p
2 ± 2S(R)
Z
S(R) = d3 rψA (⃗
r )ψB ⃗r)
R2
 
R
= 1+ + 2
e−R/a0 .
a0 3a0

This can be considered as a variational calculation with zero variational parameters.

This integral S(R) is most easily calculated in elliptic coordinates, where the three compo-

240
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

nents of ⃗
r are replaced by

|⃗ ⃗ A | + |⃗
r−R ⃗ B|
r−R
u≡ ,
R
|⃗ ⃗ A | − |⃗
r−R ⃗ B|
r−R
v≡ ,
R
ϕ ≡ arctan y/x.
Thus, u is the scaled sum of the distances to the two protons, v is the scaled difference of the
two distances and ϕ is the usual azimuthal angle with respect to the axis defined by the two
protons. Some Jacobian manipulations would reveal,
Z ∞
R3 2
Z Z 1 Z 2π
3
d rf (⃗
r) = du dv dϕ (u − v 2 )f (⃗r ).
1 −1 0 8
Once one has made a transformation into these coordinates the integral becomes rather trivial
because
ψA (⃗ r ) = (πa30 )−1 e−uR/a0
r )ψB (⃗
If you have not worked with elliptic coordinates before, you can understand the name “ellip-
tical” because the set of points with a fixed sum of distances, |⃗ ⃗ A | + |⃗
r−R ⃗ B |, defines an
r−R
ellipse.
Now that the normalized wave function is known, one can calculate the expectation of the
energy,
⟨A|H|A⟩ + ⟨B|H|B⟩ ± 2⟨A|H|B⟩
⟨H⟩± = ϵ± (R) =
2 ± 2S
⟨A|H|A⟩ ± ⟨A|H|B⟩
= ,
1±S
where
e2 e2
Z
2
⟨A|H|A⟩ = ϵ1 + − ψA (⃗
r) d3 r
R |⃗ ⃗
r − RB |
e2
 
R
= ϵ1 + 1+ e−2R/a0 ,
R a0
and
e2 e2
  Z
⟨A|H|B⟩ = ϵ1 + S− d3 rψA (⃗
r )ψB (⃗
r)
R |⃗
r−R ⃗ B|
e2 e2
   
R
= ϵ1 + S− 1+ e−R/a0 .
R a0 a0
The term ϵ1 is merely the electronic binding energy of one electron with a proton, -1.0 Ryd-
bergs. The integrals were calculated with the help of the tranformation into elliptic coordi-
nates.

241
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

Figure 12.1: Potential for + H2 ion from Example 12.4.

The potential between the atoms is effectively

V± (R) = ⟨H⟩± − ϵ1 .

When plotted against R the resulting curves are shown in Fig. 12.1.
In the adiabatic approximation, where the mass of the proton is considered to be very large,
the binding energy should simply be the minimum of the potential. In reality the binding
energy is deeper by an entire eV. However, one could improve the calculation by considering
a variational calculation.
The difference between V+ and V− comes from the fact that the parities of the two wave func-
tions were positive and negative respectfully. The positive-parity solution has lower energy
because the probability density is larger in the region between the protons where the inter-
action with the protons is maximized, and because the negative-parity solution has an extra
node, which increases the kinetic energy.
This result for the potential should be contrasted with the previous example. In that case
the potential at large distances was found to fall as 1/R4 using perturbation theory. Even
though that calculation was based on the harmonic oscillator, one would still expect that for a
hydrogen atom the perturbative result to scale as the square of the electric field and behave as
1/R4 . In this calculation the potential fell exponentially at large R, so it rather poorly predicts
the behavior for large distances. This failure has nothing to do with the adiabatic nature
of the approximation, but rather with the poor choice of wave functions in the variational
approximation. In order to improve the large-R behavior the variational wave function would
need to incorporate excited states of the hydrogen atom.

12.11 The Hydrogen Molecule and Pairing

The neutral hydrogen molecule has two electrons. One can pursue a variational calculation
as performed in the previous problem; only in this instance the variational wave function is a

242
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

two-electron wave function. When we discussed multi-electron atoms, Hund’s first rule stated
that the prime criteria for multi-electon states to minimize the energy is to form states of largest
S so that the spatial wave function will be maximally anti-symmetrized thus minimizing the
electron-electron repulsive interaction. However, in the case of the H2 molecule, the driving
determination is that both electrons should be in the even-parity ψ+ state, which through the
Pauli exclusion principle requires that the electrons be in a spin singlet with S = 0.
When two electrons form a spin-singlet they are called a pair. Much of the systematics of molecu-
lar binding can be understood by considering pairing. Pairing that involves p states is somewhat
more complicated. Of the three p states, the mℓ = 0 state extends along the z axis and thus more
strongly samples the attractive interaction with the other proton. Pairing between the mℓ = ±1
states can also take place but is usually weaker. The two pairings are refered to as σ and π
bondings. The π label refers to the fact that the transverse pairing allows an angular momentum
about the molecular axis.

12.12 Superconductivity
Pairing of electrons is responsible for the phenomena known as superconductivity. The con-
nection between forming pairs and reducing the conductivity to zero is actually rather subtle.
The formed pairs, which can have zero net momentum, can coalesce and move as a coherent
unit through the material. Here, we concentrate on the pairing mechanism and offer only the
following rather heuristic discussion of how this leads to superconductivity.
One could imagine resistance, or dissipation, coming from three sources:

1. As we will see, a gap energy will be associated with the pairs. In order for the moving
pairs to de-excite one could excite the pairs by at least the gap energy, which would break
a pair. If one imagines a current moving through a circular loop at a temperature near
zero, there is not enough thermal energy available to break a pair. At small temperature, it
is extremely difficult to de-excite the condensate by breaking pairs or removing pairs from
the condensate.
2. One could imagine a drag force acting on the condensate due to moving the condensate
through the background of non-condensate particles. If the condensate is treated as a
macroscopic object, the drag force would then behave as

Fdrag ∝ u2 , (12.36)
P = F v ∝ u3 ,

where u is the condensate velocity. Because the power scales as u3 and because the resis-
tance is defined as

P = I 2 R, (12.37)
R = P/I 2 ,
I ∝ u, P ∝ u3
P
R = 2 ∝ u → 0.
I

243
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

Thus, drag can be neglected with respect to its contribution to the resistance. Because the
coalesced pairs move as a unit, they can carry a large current with a very small velocity
and thus vanishingly small drag.
3. The pairs condense into a coherent unit, and one could imagine exciting a pair, while leav-
ing it bound. However, because of attractive inter-pair interactions, the pairs in the con-
densate also require energy to be dislodged. This energy tends to exceed that required to
break a pair.

The best way to dissipate movement of the condensate is to break a pair. This requires an amount
of energy 2∆, where ∆ is known as the gap energy. Only those electrons with energies greater
than the gap can contribute to the dissipation. The conductivity then behaves as e−2∆/T
The phenomena of superconductivity has much in common with superfluidity. In both cases, it is
rather easy to motivate why one forms a condensate, but explaining why the condensate moves
without dissipation is subtle. Both an electron pair and a 4 He atom are bosonic, but formed of an
even number of fermions. Superconductivity might occur in any Fermi system, and is thought
to occur due to the pairing of protons or neutrons in the crusts of neutron stars. When nuclear
matter is compressed to extremely high density, several times higher than the density of matter
in nucleus, and becomes a Fermi gas of quarks, one also expects pairing of the quarks.

12.13 Cooper Pairs

Our goal in this section will not be to understand the correlated structure of the pairs, but only
to show that the existence of pairs lowers the energy compared to a Fermi gas. In order for
electrons to pair there must exist an attractive interaction between two electrons with momenta
near the Fermi surface. The idea of two electrons interacting attractively is most peculiar be-
cause electrons have the same charge. The excuse for assuming an attractive interaction is that
the electrons interact with one another via the lattice. An electron might exchange momentum
with an ion in the lattice, which then translates that momentum into a sound wave, or phonon,
which travels through the lattice. The phonon then interacts with the other electron through its
interaction with some other ion. Intuitively, one might expect such an interaction to provide an
effective polarizability that reduces the electric repulsion, but does not reverse it. However, elec-
trons are free to move and screen the repulsive electric interaction, as in Debye screening, while
not screening the interaction through a phonon. This allows the attractive interaction through
phonons to overwhelm the electromagnetic interaction.
For our purposes, we follow the work of Bardeen, Cooper and Schraefer (BCS) and assume a
simplified interaction between electrons,
 v0
⃗ ⃗ ⃗ ′ ⃗′ − V δk1 +k2 ,k1′ +k2′ , kf < k1 , k2 , k1′ , k2′ < ka
⟨k1 k2 |V|k1 k2 ⟩ = , (12.38)
0, otherwise

where v0 sets the scale of the interaction, V is the volume, and ka is close to kf . Thus, the model
assumes the attractive interaction is confined to particles within the neighborhood of the Fermi
surface.

244
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

Because total momentum is a good quantum number, eigenstates can have the form,
X′
⃗ paired⟩ =
|K ⃗ ⃗
a⃗k′ (K)| ⃗
k1′ = K/2 +⃗
k′ , ⃗ ⃗
k2′ = K/2 −⃗
k′ ⟩, (12.39)
⃗
k′

where the primed sum is over all relative momenta ⃗

k′ ≡ (⃗
k1′ − ⃗
k2′ )/2 such that both ⃗
k1′ and ⃗
k2′
are inside the region between kf and ka .
The overall Hamiltonian between states of total momentum K ⃗ can be written as,
′
X
H = (ϵK/2+ ⃗ ⃗ ⃗ ⃗ ⃗ ⃗ ⃗ ⃗
⃗ ⃗k + ϵK/2−
⃗ ⃗k )|K + k, K − k⟩⟨K + k, K − k| (12.40)
⃗
k
′ X
′
X −v0
+ ⃗
|K/2 +⃗ ⃗
k, K/2 −⃗ ⃗
k⟩⟨K/2 +⃗ ⃗
k′ , K/2 −⃗
k′ |
⃗ ⃗
V
k k′

Solving for the eigenstates

H|K,⃗ paired⟩ = E|K, ⃗ paired⟩, (12.41)

v0 X ′
− ⃗ = (E − ϵ1 − ϵ1 )a⃗ (K),
a⃗ ′ (K) ⃗
V ⃗′ k i
k
ϵ1 = ϵK/2+
⃗ k , ϵ2 = ϵK/2−
⃗ ⃗ ⃗k.

Dividing both sides by (E − ϵ1 − ϵ2 ), and summing both sides over ⃗

k,
′ ′ ′
X v0 X 1 X
⃗ −
a⃗k (K) ⃗ = 0.
a⃗k′ (K) (12.42)
⃗
V ⃗
E − ϵ1 − ϵ2 ⃗
k k k′

Now, one can see how the simplifying assumption that the matrix element vk1 ,k2 ,k1′ k2′ was inde-
pendent of k and k′ in the subspace simplifies the problem. It allows the sum over amplitudes
to be canceled from both sides of the equation above and result in,
1 1 X 1
− = (12.43)
v0 V ⃗
E − ϵ 1 − ϵ2
k′
′
1 X 1
=
V ⃗
E − ϵK/2+
⃗ k − ϵK/2−
⃗ ⃗ ⃗k
k
1 X 1
= ,
V ⃗
E − ℏ2 K 2 /(4m) − ℏ2 k2 /m
k

where the last step involved writing the energy as a center-of-mass energy plus an energy-of-
relative-motion.
One can solve for E in the equation above graphically. To illustrate the solutions, we consider
the function
′
1 X 1
ΦK (E) ≡ , (12.44)
V ⃗ E − ℏ2 K 2 /(4m) − ℏ2 k2 /m
k

245
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

Φ (E)

-1/v 0

Figure 12.2: Graphical solutions to Eq. (12.44).

and graph the function to see for what energies, E, the function Φ(E) equals −1/v0 . This is
illustrated in Fig. (12.2).
Every time E passes an energy, ℏ2 K 2 /(4m) + ℏ2 k2 /m, Φ changes from −∞ to +∞. Thus, for
every value of ⃗
k there exists a solution with E < ℏ2 K 2 /(4m) + ℏ2 k2 /m where Φ = −1/v0 .
The vertical dashed lines in the figure represent the energies ℏ2 K 2 /(4m) + ℏ2 k2 /m, which
would be the eigenenergies if v0 were to be zero. The intersections of the blue curves with the
horizontal dashed line represent the solutions.
⃗ = 0, the first value of ⃗
If K k that enters the primed sum is |⃗k| = kf . This state is composed
of momenta on opposite sides of the Fermi surface. From Fig. 12.2, one can see that the first
solution has an energy below that of the first vertical dashed line, with the first vertical dashed
line corresponding to |⃗
k| = kf . Thus, there exists a solution with energy E < 2ϵf , even though
the solution was formed from momentum states above kf . This is essentially a bound state,
because the paired state is energetically favorable compared to being a momentum state at the
Fermi surface. It cannot decay into lower energy particles because the states with k < kf are all
occupied. Note that as v0 is increased the solution for the energy becomes lower.

12.14 The Gap

The expression for Φ in Eq. (12.44) can be integrated analytically for the case with K = 0 if one
approximates the sum over states as
′
X Z ϵa
→ V ρ(ϵf ) dϵk . (12.45)
⃗ ϵf
k

246
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

This amounts to assuming the density of states per volume, ρ(ϵk ), is constant in the region of
integration. The expression for Φ then becomes
Z ϵa
1
Φ(E) = ρ(ϵf ) dEk (12.46)
ϵf E − 2ϵk
ρ(2ϵf )  2ϵa − E 
 
=− log  ,
2 2ϵf − E 

where
mkf
ρ(ϵf ) = . (12.47)
2π 2 ℏ3
Solving for E,
−2ϵa + 2ϵf e2/(ρv0 )
E= . (12.48)
e2/(ρv0 ) − 1
The difference between this energy and 2ϵf is known as the gap energy, ∆.

2ϵf − E
∆≡ (12.49)
2
ϵa − ϵf
= .
e2/[ρ(ϵf )v0 ] − 1
One can do a similar calculation for K ̸= 0, but the binding energy would be smaller because
the interaction links fewer pairs as both electrons in the pair, with momenta k± = K/2 ± k,
must lie within the shell in momentum space.
One can see that the energy required to break the pair is 2∆, because in order to break the pair
one must move the electrons outside the Fermi sea, that is, increase their energy beyond 2ϵf .

12.15 Landau Levels and the Integer Quantum Hall Effect

Consider particles of mass m and charge e moving in a two-dimensional world, the x-y plane.
⃗ = B ẑ. The Hamiltonian for such particles is
A magnetic field is present, B
 2
⃗ − eA/c
H = P ⃗ /(2m). (12.50)

If one chooses as a vector potential,

⃗ = Bxŷ,
A (12.51)
one obtains the desired magnetic field. Because there is no y dependence in the Hamiltonian,
solutions can be chosen as eigenstates of the operator Py with eigenvalue py . One can then
rewrite the Hamiltonian,
Px2 (py − eBx/c)2
H = + , (12.52)
2m 2m

247
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

where py is simply a number and Px is an operator. Because py is a number, one can consider
the second term as an offset harmonic oscillator potential,

Px2 1 py c 2
 
2
H = + mω x − , (12.53)
2m 2 eB
eB
ω≡ .
mc
Thus, the Hamiltonian looks like a one-dimensional Harmonic oscillator with a frequency equal
to the Larmor frequency. The harmonic oscillator is centered at x = py c/(eB).
Note that the eigenenergies, (n + 1/2)ℏω, do not depend on py . Thus, there are many solutions
with different py that have identical energies. The energy levels are referred to as Landau levels.
Each level has a degeneracy equal to the number of values of py for which there exist solutions.
Using the fact that the y−dependence behaves as eipy y/ℏ and applying the boundary conditions
that the wave function vanishes at y = 0 and y = Ly , one finds the same values of py as one
would have for particles in a box of size Ly . The density of such states is then
dN Ly
= , (12.54)
dpy 2πℏ

where Ly is the size of the sample in the y direction. The limits on py are determined by the
x dimension. Because py determines the center of the center of the Harmonic oscillator, x =
py c/(eB), and because the centers must be between x = 0 and x = Lx ,
py c
0< < Lx , (12.55)
eB
eBLx
0 < py < py,max = .
c
The number of such states is then
dN eBLx Ly
N = py,max = , (12.56)
dpy 2πℏc

Assume the sample has a number of conduction electrons per unit area n. If the number of
conduction electrons exactly fits an integer number of Landau levels, the levels will be exactly
filled and the conductance changes. Thus, varying the magnetic field one sees changes in the
resistivity when
eBLx Ly
nLx Ly = m , m is an integer, (12.57)
2πℏc
2πℏcn
m= .
eB
For fixed density n, the levels of B that cause jumps in the conductance are
2πℏcn
B= . (12.58)
em

248
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

— — — — — — — — — —
y
B Lorentz Force e velocity

Electric Force X
+ + + + + + + + +
Figure 12.3: The classical Hall effect. The electrons moving to the right feel a Lorentz force in the y
direction. This causes a build up of electrons on one side of the wire, and a deficit of electrons on the
opposite side, until an electric field is large enough to cancel the Lorentz force.

This is known as the integer quantum Hall effect. It corresponds to an integral number of levels
being exactly filled. From the constraint on magnetic flux given in the discussion of the Ahara-
nov Bohm effect in Eq. (3.33) one can see that these dips in the conductivity happen when the
area per electron multiplied by the magnetic field give a magnetic flux corresponding to some
integer factor of the relevant quantized units of flux.
The behavior of the resistivity as a function of the magnetic field is remarkable. The conductance
and resistivity are a property of the material described by a tensor, σij ,
Ji = σij Ej , (12.59)
Ei = ρij Jj ,
ρ = σ −1 .
The off-diagonal elements appear due to the presence of a magnetic field. Here, we assume that
B⃗ is applied in the z direction, and that the current is moving in the x direction along a surface
in the xy-plane. For free particles in a constant magnetic field B ẑ, the addition of an electric
field E x̂ induced a charge to move with an average velocity Ec/B in the y direction as derived
in Eq. (3.23). Thus, we expect an off-diagonal component in the conductivity tensor, σxy ̸= 0, if
a magnetic field is present. Additionally, one expects diagonal elements to the resistivity given
the presence of scattering in the medium.
To understand the classical Hall effect, one considers a two-dimensional “wire” carrying a steady
current in the x direction. There is no current in the y direction despite the Lorentz force acting
on the moving charge. As illustrated in Fig. 12.3 the electric ensues because the Lorentz force
moves charge density to the boundaries of the wire until an electric field builds up that cancels
the Lorentz force.
It is not too surprising that there are peaks in the diagonal components of the resistivity, ρxx ,
when the applied magnetic field corresponds to an exactly filled shell as are displayed in Fig.
12.4. Indeed, the sharp peaks are located at points where the magnetic field is adjusted to exactly
fill a shell.
The off-diagonal resistivity is more difficult to understand. One begins by considering a current
Jx moving down a wire. The net Lorentz force in the y direction is Fy = eN vy B, where N
is the number of electrons in some segment of the wire. This drives electrons to one side of the
wire. The electrons move to one side of the wire until an electric field in the x direction balances
out the force,
eN Ey = eN vx B. (12.60)

249
the classical model (figure 1), one quickly sees that classical physics does not give us the
PHY answer.
right 851/852 We12areNON-PERTURBATIVE APPROACHES
therefore going to use FOR MANY-FERMION
a quantum mechanical approach toSYSTEMS
describe
the system.

Figure 2: Resistivities of the integer quantum Hall system, as functions of the magnetic
field.
FigureThe
12.4: red line shows
Experimental the longitudinal
measurements resistivityand
of the diagonal(red) ⇢xx , it is zero
off-diagonal as long (green)
resistivities as ⇢xyinsits
the
quantum Hall effect. The off-diagonal resistivity appears at quantized values of 2πℏ/e 2 . Figure from
on a plateau level and spikes whenever ⇢xx changes from one plateau to the next. The
http://oer.physics.manchester.ac.uk/AQM2/Notes/Notes-4.4.html.
green line denotes the Hall resistivity ⇢xy , it takes on a plateau form, i.e. it is constant
over a range of magnetic fields. Image reproduced from [6].
The current per area is then

1.5 Quantum Treatment Jx = eN vx /A = eN Ex , (12.61)

A Bz
1.5.1 Eigenstates and Eigenvalues
so the transverse conductivity and resistivities are
The Hamiltonian for one electron in ae 2D N
system subjectB
to a perpendicular magnetic
field Bẑ is σxy = , ρxy = . (12.62)
BA 1 eN/A
2
H= (p + eA) (10)
2m The Hall voltage is Ey ly , where ly is the width
This result is known as the classical Hall effect.
where A is the vector potential describing the magnetic field: r ⇥ A = Bẑ. Note that
of the wire.
the momentum
However, operatorcontribute
not all electrons p is the to
canonical momentum
the resistivity. operator
If one assumes that and di↵ers from
the electrons not in the
the
filled Landau levels are pinned by impurities, then only those in the Landau levels contribute.
This assumption then leads to the quantum Hall5 effect. The number of electrons in Landau
levels, from Eq. (12.56), is
eBA
NL = ν , (12.63)
2πℏc
where ν is the number of filled Landau levels. Substituting NL into Eq. (12.62) for N , the
magnetic field cancels and
2πℏc
ρxy = . (12.64)
νe2

250
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

Aside from the factor ν, the transverse conductivity and resistivity are set by fundamental con-
stants. This explains the plateaus in the transverse resistivities in Fig. 12.4. The 1985 Nobel Prize
was awarded to Klaus von Klitzing for discovery of the quantum Hall effect. This provides a
means for measuring the fine structure constant, α = e2 /ℏc.
The longitudinal resistivity has maxima for the magnetic fields for which Landau levels become
full. This is normal behavior for conductors if one thinks of the levels as conduction bands.
Electrons in partially filled bands readily scatter. Thus, they do not undergo the circular motion
of a particle interacting with only the electromagnetic field and tend to diffuse parallel to the
applied electric field. When the Landau levels are perfectly full, there are few such electrons
and the longitudinal resistivity has peaks. The fact that electrons from the unfilled bands do
not contribute at all, or extremely little, to the transverse conductivity is beyond this author’s
expertise.
The fractional quantum Hall effect refers to conductance structure for fractional, e.g. one third,
values of m. This is much more difficult to explain, but is related to the sharing of electrons
between adjacent orbitals, similar to the sharing of electrons in the covalent bonds in solids.
The 1988 Nobel prize was awarded to Robert Laughlin, Horst Störmer and Daniel Tsui for its
discovery.

12.16 Exercises
1. Consider the two electron holes in the p-shell of a neutral oxygen atom.
(a) What is the L − S − J of the ground state.
(b) If the atom is in a magnetic field of 0.01 Tesla, find the magnetic energies of the origi-
nally degenerate 2J + 1 states.
2. One electron moves in a one-dimensional system and feels the interaction of two atoms.
Approximate the interaction between the electrons and the atoms with the potential

V (x − R) = −βδ(x − R),

where R is the position of an atom. Use the adiabatic approximation to answer the follow-
ing questions.
(a) Given the two atoms are separated by a distance r, find a transcendental equation
relating k and r where the electronic binding energy is ℏ2 k2 /(2m).
(b) Find the potential between the two atoms at small r,

V (r → 0) ∼ V (r = 0) − αr,

that is, find V (r = 0) and α. Do this by expanding the transcendental equation in

terms of r. Hint: First, find V (r = 0) by solving the transcendental equation with
r = 0. Take derivatives of the transcendental equation with respect to r, then solve
for dk/dr at r = 0, and finally find dE/dr to obtain α.
(c) Find the potential between the two atoms at large r,

V (r → ∞) = −γ exp(−2k∞ r),

251
PHY 851/852 12 NON-PERTURBATIVE APPROACHES FOR MANY-FERMION SYSTEMS

that is, find γ. Hint: Use first order perturbation theory, assuming the unperturbed
wave function is the bound state of one well, and the perturbation is the interaction
with the second well.

3. Consider a particle of mass M and charge e moving in the x − y plane under the influence
of a magnetic field in the z direction. Ignore motion in the z direction.

(a) Show that the vector potential,

B
⃗=
A (xŷ − y x̂) ,
2
describes a magnetic field in the z direction.
(b) Write the Schrödinger equation,
 2
⃗ − eA/c
P ⃗
ψ(r, ϕ) = Eψ(r, ϕ),
2M
in cylindrical coordinates.
(c) Show that Lz commutes with the Hamiltonian.
(d) Assuming the solution is an eigenstate of Lz with eigenvalue mℏ,

ψ(r, ϕ) = eimϕ ξm (r),

rewrite the Schrödinger equation for ξm (r).

(e) Extra Credit: Solve for ξm (r) and the eigenenergies for the case where m = 0.

4. Consider a surface with 10 electrons per µm2 . Lowering the magnetic field, at what mag-
netic field (in Tesla) do you find the first dip in conductivity due to the quantum Hall
effect?

252
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

13 Relativistic Quantum Mechanics

13.1 The Klein-Gordon Equation

In relativistic kinematics energy and momentum appear on equal footing, as do position and
time. For instance,

E 2 − p2 c2 = m2 c4 . (13.1)

Rewriting energy and momentum as operators, one obtains the Klein-Gordon equation,

∂2
 
−ℏ 2
+ ℏ c ∇ − m c ϕ(⃗
2 2 2 2 4
r , t) = 0. (13.2)
∂t2

In this version, the wave function ϕ can be a complex number. There are instances where one
uses a real field, but here we use a complex field so that we can find solutions that are plane
waves. It differs from the Schrödinger equation in that time appears in a second derivative.
The Klein-Gordon equation for a complex field has plane wave solutions,

r ) ∼ ei⃗p·⃗r/ℏ−iE± (p)t/ℏ ,
ϕp⃗ (⃗ (13.3)
p
E± (p) = ± p2 c2 + m2 c4 .

If one expands the energy for small p,

E+ (p) ≈ mc2 + p2 /2m. (13.4)

Thus, in the non-relativistic limit, the energy appears identical to the that of the Schrödinger
equation aside from the inclusion of the rest mass energy. If mass were conserved this would
2
merely add a constant to the energy along with an unobservable phase, e−iM c t/ℏ . However, for
cases where particles are generated, the phase plays an important role.
Perhaps the most unusual feature in the solutions is the appearance of negative-energy solutions.
The negative-energy solutions can only be understood when one writes field operators in second
quantization. The field operators (remember the case for photons) will then be written as

ℏc X 1  ip·r/ℏ †
r 
Φ(⃗ r , t) = p e bp⃗ + e−ip·r/ℏ ap⃗ (13.5)
V p Ep
ℏc X 1
r  
= p e−i⃗p·⃗r/ℏ eiEt/ℏ b†p⃗ + e−iEt/ℏ a−⃗p .
V p Ep

where p·r ≡ Ep t− p ⃗ ·⃗
r . Thus, in the language of field operators, the negative-energy “solution”
is associated with the destruction of antiparticles moving with opposite momentum. We note
that if the field were a real field, rather than a complex field, b†p⃗ would be replaced by a†p⃗ and bp⃗
by ap⃗ . It only seems natural that a relativistic theory should require the consideration of field
operators, because the creation and destruction of particles is an inherent feature of relativistic
treatments.

253
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

For non-relativistic field operators we had defined field operators as,

1 X −iEk t+i⃗k·⃗r
Ψ(⃗ r , t) = √ e a⃗k , (13.6)
V k
1 X iEk t−i⃗k·⃗r †
Ψ† (⃗
r , t) = √ e a⃗k .
V k

If the field operators refer to electrons, one can think of Ψ† (⃗

r ) as creating a negative particle at
r , while Ψ(⃗
⃗ r ) would destroy a negative particle. With relativistic fields, the operator Φ† creates
a positive particle, but also destroys the negative antiparticle. This can be considered a particle-
antiparticle symmetry that derives from the existence of negative-energy solutions. The same
symmetry will result from the Dirac equation, which is the topic of the next section.

Aside: Covariant-Contravariant Notation

Four vectors, e.g. time and position or energy and momentum, are usually referenced by Greek
indices numbering from zero to three. The zeroth component refers to the component which
rotate and boost like time, e.g. time or energy, and the other three refer to the components that
behave like spatial vectors, e.g. position or momentum. Contravariant vectors are labeled with
superscripts, e.g. Aµ , and the values are the actual values of the vectors. Covariant vectors are
labeled by subscripts and have their spatial components reversed. The components of pµ are
E, −px c, −py c, −pz c. Derivatives behave oppositely,

∂ µ
∂
∂µ = , ∂ = . (13.7)
∂xµ ∂xµ

This definition ensures that ∂µ (p · x) = pµ . One must worry about whether the position xµ
refers to (ct, x, y, z) or t, x/c, y/c, z/c. Here, we assume the former, with xµ having units of
distance. With this choice
1
∂0 = ∂t , ∂ 1 = ∂x , · · · (13.8)
c

It only makes sense to define four vectors so that all the components have the same units, because
boosts behave like rotations that mix time and position. Maintaining a different set of units for
time and position (or for energy and momentum) makes as much sense as using one set of units
for the x-component of the position and another for the y-component. For the remainder of
the chapter factors of c may appear or disappear from expressions. In practice, the best way to
proceed is to ignore factors of c and treat time and position as if they are measured in the same
units. Then, after the final expressions are derived, one reinserts the factors of c as needed. For
example, in high-energy nuclear physics distance is often measured in fm, and time is measured
in fm/c. In this set of units c = 1, so all the factors can be ignored. If one ultimately requires
knowing time in seconds, one can first calculate a numerical value for the time in fm/c, then
convert to seconds.

254
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

The use of Greek indices for four-vectors, and Roman indices for three-vectors, helps alleviate
some of the confusion with covariant and contravariant notation. For instance,

Ai=2 = Aµ=2 = −Aµ=2 . (13.9)

Thus, the term A2 is ambiguous, unless one one knows whether one is referring to a four-vector
or a three-vector. Dot-products of four vectors,

A · B = A µ Bµ = A 0 B0 − A x Bx − A y By − A z Bz , (13.10)

are invariant to both boosts and rotations. To distinguish these from the dot products of three-
vectors, A⃗ ·B⃗ = Ax Bx + Ay By + Az Bz , the vector (or in other text bold face) are not invoked,
⃗ · B.
i.e. the dot-product of a four-vector is A · B, not A ⃗

13.2 Current Conservation

The current density should have the form of a relativistic four-vector. Furthermore, it should be
conserved,

r , t) + ∇ · ⃗j(⃗
∂t ρ(⃗ r , t) = 0, (13.11)

or in four-vector notation,

∂µ J µ (⃗
r , t) = 0. (13.12)

If J 0 has dimensions of charge density, J i must refer to the normal current divided by c. To
accomplish this equivalence the four-current operator is defined in terms of the field operators
as
−ie  †
J µ (x) = Φ (x)∂ µ Φ(x) − (∂ µ Φ† (x))Φ(x) .

(13.13)
2

Using the definition of Φ(x) in Eq. (13.5),

ie ℏc X 1
J µ (x) =

p (13.14)
2 V p
⃗p⃗′
Ep Ep′
−ip′µ ′ ip′µ ′
(a†p⃗ eip·x/ℏc + bp⃗ e−ip·x/ℏc )( ap⃗′ e−ip ·x/ℏc + b†p⃗′ eip ·x/ℏc )
ℏc ℏc
ipµ ipµ ′ ′
a†p⃗ eip·x/ℏc − bp⃗ e−ip·x/ℏc )(ap⃗′ e−ip ·x/ℏc + b†p⃗′ eip ·x/ℏc ) .

−(
ℏc ℏc
Integrating over coordinate space, and setting x0 = 0, only the p
⃗ = p ⃗′ terms survive and the
factor of volume disappears.
Z X pµ 
d3 x J µ (x) = e a† (⃗ p)b† (⃗

p)a(⃗
p) − b(⃗ p) . (13.15)
p
⃗
E p

255
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

These look like the usual non-relativistic expressions, but particles and antiparticles contributing
with opposite signs.
A peculiar aspect of the expression is that the contribution from antiparticles has the creation
and destruction operators reversed, i.e. bb† rather than the number operator b† b. This implies
that an extra charge of -1 is associated with each momentum mode after making the substitution,

a† a − bb† = a† a − b† b − 1. (13.16)

This suggests that the vacuum has a charge of -1 associated with each mode, which translates
into an infinite negative charge because there are an infinite number of modes.
If one wrote down an expression for the Hamiltonian in terms of field operators (which could
be derived from Lagrangians) one would find,
X  
H = Ep a†p⃗ ap⃗ + bp⃗ b†p⃗ (13.17)
p
X  
= Ep a†p⃗ ap⃗ + b†p⃗ bp⃗ + 1 .
p

Thus, the vacuum has an infinite positve energy associated with it. When we consider the Dirac
equation, which is applicable for fermions, the same problem will appear, but with opposite
signs. That is, the energy of the vaccum will be −Ep for each mode. This result serves as one of
the motivations for super-symmetry, where every bosonic mode has a corresponding fermionic
mode.

13.3 Coupling to the Electromagnetic Field

In the Schrödinger equation, incorporating the electromagnetic field could be accomplished by
minimal substitution,

−iℏ∂⃗ → −iℏ∂ ⃗ − eA/c

⃗ (13.18)
∂ ∂
iℏ → iℏ − eΦ,
∂t ∂t
In undergraduate textbooks, the electric potential, Φ, is often referred to as the “scalar” potential.
However, it is actually the zeroth component of the four-vector field Aµ .
In relativistic treatments, coupling to the electromagnetic field is also incorporated via minimal
substitution. One result of coupling to an electromagnetic field is that the charge and current
densities are modified,
e2 A0 (⃗
 
0
ie ∗
∂ ∂ ∗ r , t) ∗
j (⃗r , t) = ϕ (⃗
r , t) ϕ(⃗ r , t) − ϕ (⃗r , t)ϕ(⃗r , t) − ϕ (⃗ r , t)ϕ(⃗ r , t) (13.19)
2c ∂t ∂t ℏc
−ie  ∗  ⃗ r , t)
e2 A(⃗
⃗j(⃗
r , t) = ϕ (⃗ ⃗ ⃗ ∗
r , t) − ∂ϕ (⃗
r , t)∂ϕ(⃗ r , t)ϕ(⃗r , t) − ϕ∗ (⃗
r , t)ϕ(⃗ r , t).
2 ℏc
The charge density can change signs if the energy, ⟨iℏ∂t ⟩, changes sign, or if the electric potential
A0 is altered. Negative charge density can be interpreted as the presence of antiparticles.

256
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

The electromagnetic field also affects the energy through minimal substitution,

⃗ c − eA)
(E − eA0 )2 = (P ⃗ 2 + m2 c4 , (13.20)
q
E = eA0 ± (P ⃗ c − eA)
⃗ 2 + m 2 c4 .

For the case where A ⃗ = 0 and I have a simple electromagnetic field, states with a given mo-
mentum P ⃗ might change from positive to negative energy depending on the strength and sign
of A0 . If a state has negative energy, and if you are searching for the minimum energy state, you
might expect the ground state to be populated with infinite numbers of quanta. Interpreting
such states can be challenging.

Aside: Units
The units for relativistic field operators are confusing and vary throughout the literature. Here,
the field Φ was given units of inverse length. With this definition Φ† ∂µ Φ had units of inverse
volume, i.e. a number density. In contrast, the vector potential operator was chosen √ such that
eA/c would have units of momentum by incorporating an additional √ factor of ℏc into its
definition. Finally, the vector potential also had an extra factor of 2π. With this definition
the Lagrangian density for a real scalar field (which we will not pursue further here) looks like
(1/2)Φ∂ 2 Φ, whereas for the electromagnetic field the it looks like (1/4π)Aµ ∂ 2 Aµ . This choice
is by no means consistent throughout the literature or throughout textbooks. Thus, this choice is
related to the choice of units for electromagnetism. For example, the fine structure constant,

e2
0 ℏc
, S.I. units
1

 4πϵ
e2
α= = ℏc
, c.g.s. units (13.21)
137.036 · · ·  e2
, natural units

4π

Fortunately, these choices will not cause much trouble in the limited discussions of this chapter.

Example 13.1: Reflecting off a Potential Step

A seemingly simple example that becomes surprisingly difficult is the reflection off a potential
step, where the potential step is due to jump in A0 . Consider the one-dimensional problem,

2 I. II. eΦ
A 0
E-mc


0, x < 0
A0 (x) =
A0 , x > 0

257
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

Consider a wave incident from the left, with a reflected and transmitted wave,

ϕI (x) = eikx − Be−ikx

′
ϕII (x) = Ceik x ,

where the energy of the particle is

p
E= ℏ2 k2 c2 + m2 c4 .

Before solving for B and C, one must find k′ . Applying the Klein-Gordon equation,

(E − eA0 )2 = ℏ2 k′2 c2 + m2 c4 ,
′2
(E − eA0 )2 − m2 c4
k = .
ℏ2 c2
One may then find the ranges for which k′2 is positive or negative,

1. k′2 > 0 when 0 < eA0 < E − mc2 ,

2. k′2 < 0 when E − mc2 < eA0 < E + mc2 ,
3. k′2 > 0 when eA0 > E + mc2 .

The first two ranges for A0 correspond to the usual case for non-relativistic physics. That is,
when the barrier is small the particle propagates as a plane wave with a reduced velocity. For
larger barriers, the particle is confined to region I and the wave function exponentially dies in
region II.
However, the surprise here is that for very large A0 , there again appears to be a solution with
the particle moving through the barrier. This surprising result can be understood by calculat-
ing the charge and current densities in region II. One finds that the charge density in region
II is opposite to that of region I. The solution describes an incoming wave that splits into two
outgoing waves, one of particles moving to the left and one of antiparticles moving to the
right. This behavior owes itself to the large field. When the voltage difference exceeds twice
the mass, it becomes possible to create pairs of particles. Because the antiparticle feels a po-
tential of −eA0 , it is perfectly willing to go to into region II, and many do so spontaneously if
the step has enough energy to overcome the mass penalty, 2mc2 . This behavior does not exist
should one consider a problem where the regions have different masses rather than different
values of eA0 . This example is assigned as an exercise at the end of the chapter.

13.4 The Dirac Equation

The Dirac equation is another example of a relativistic wave equation. However, the Dirac equa-
tion differs in that it describes spin 1/2 particles whereas the Klein-Gordon equation applies to
spinless particles.
Dirac was motivated by finding a linear equation that was consistent with the relativistic con-
straint H 2 = P 2 c2 + m2 c4 . By linear, Dirac was looking for an equation that was linear in the

258
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

derivatives. In order to accomplish this Dirac suggested a matrix equation,

H = αx cpx + αy cpy + αz cpz + βmc2 (13.22)
= c⃗
α·p ⃗ + βmc2 ,
where αi and β are Hermitian matrices. Given that one needs to satisfy our knowledge of
relativity,
H 2 = p2 c2 + m2 c4 , (13.23)
there are constraints on the matrices,
{αi , αj } = 2δij , β 2 = 1, {αi , β} = 0. (13.24)
Two-by-two matrices are insufficient, as one can only construct three anti-commuting matrices
(the Pauli matrices), but four are needed. Three-by-three matrices are also insufficient, but four-
by-four matrices are sufficient. The following four-by-four matrices satisfy the conditions,
 
  1 0 0 0
0 ⃗σ  0 1 0 0 
⃗ =
α , β=  0 0 −1 0  .
 (13.25)
σ 0
⃗
0 0 0 −1
Actually, there exist an infinite number of choices as one can transform the matrices by a unitary
transformation,
⃗ U †,
Uα (13.26)
and find a new choice of matrices that will still satisfy the constraints. The choice above is known
as the Dirac representation, which is convenient for massive particles. For highly relativistic
particles, it is sometimes convenient to employ the chiral representation,
 
  0 0 1 0
σ 0
⃗  0 0 0 1 
⃗ =
α , β = −  1 0 0 0 .
 (13.27)
0 −⃗ σ
0 1 0 0
We will mainly employ the Dirac representation.
It should be stressed that the four components of the wave function do not make a relativistic
four vector. They correspond to spin-up, spin-down, and the two corresponding anti-particle
solutions. Like any wave function, the components of the wave functions mean nothing in
isolation. Only expectations have a physical meaning, and those involve overlaps of vectors.
By changing the arbitrarily chosen representation, the physical meaning of the solutions is un-
changed.

13.5 Conserved Quantities

The particle current,

r , t) ≡ ψ † (⃗
j0 (⃗ r , t), ⃗j(⃗
r , t)ψ(⃗ r , t) ≡ cψ † (⃗
r , t)⃗
αψ(⃗
r , t), (13.28)

259
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

is conserved. One can demonstrate the conservation of the current by taking the four divergence,
∂µ j µ , and by applying the Dirac equation,

∂
Hψ = iℏ ψ, (13.29)
∂t
∂
iℏ ψ + iℏc⃗ ⃗ = mc2 βψ,
α · ∇ψ
∂t
∂ †
−iℏ ψ − iℏc⃗ ⃗ † = mc2 βψ † ,
α · ∇ψ
∂t
to see that
∂
j0 + ∇ · j = [(∂t + c∇ · α r , t)]† ψ(⃗
⃗ )ψ(⃗ r , t) + ψ † (⃗
r , t)(∂t + c∇ · α
⃗ )ψ(⃗
r , t) (13.30)
∂t
imc2 
r , t)† ψ(⃗
r , t) + ψ † (⃗

= −(βψ(⃗ r , t)βψ(⃗
r , t)
ℏ
= 0.

The fact that α

⃗ and β are Hermitian was exploited.
This contrasts with the Klein-Gordon equation in that no derivatives are used to construct the
current-density or the charge-density operators. In the Klein-Gordon equation both quantities
had derivatives, and in the Schrödinger equation the current density involves derivatives, but
the charge density does not.
Now, we consider conservation of angular momentum. Unlike the Schrödinger equation, the
Dirac Hamiltonian does not commute with the orbital angular momentum.
X X X
[H, ⃗
r×p ⃗]i = c[ α ℓ pℓ , ϵijk rj pk ] + mc2 [β, ϵijk rj pk ] (13.31)
ℓ jk jk
X
= −iℏc ϵijk αj pk ,
jk

[H, ⃗
r×p
⃗] = −iℏc⃗
α×p
⃗.

⃗
One can also define a spin operator, S = ℏΣ/2, where

i
Σi = − ϵijk αi αj (13.32)
2 
⃗ σ 0
⃗
Σ= in the Dirac representation.
0 ⃗ σ

260
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

The spin is also not conserved.

X ℏcpj
[H, Si ] = [αj , Σi ] (13.33)
j
2
X −iℏcpj
= ϵikl [αi , αk αl ]
jkl
4
X −iℏcpj
= (ϵijk αk − ϵikj αk )
jk
2
X
= iℏc ϵijk αj pk ,
jk
⃗ = iℏc⃗
[H, S] α×p
⃗.

However, the combination J⃗ = ⃗

r×p ⃗ is conserved,
⃗+S

[H, J⃗] = 0. (13.34)

It is often the case in physics that the spin and orbital angular momentum are not conserved
individually, but only in combination. The surprising result here is that they are not conserved
individually even in the absence of interaction.

13.6 Solutions for Free Particles

In analogy to the Klein-Gordon equation, the Dirac equation will have both positive-energy
and negative-energy solutions. Here, we consider those solution which behave as ei⃗p·r/ℏ . The
positive-energy solutions with a given momentum, p ⃗, are referred to as us (⃗
p) while the negative-
energy solutions are referred to as vs (−⃗
p). The label s refers to the spin.

p) = α
Ep u(⃗ ⃗ ·p p) + βmc2 u(⃗
⃗cu(⃗ p) (13.35)
2
−Ep v(−⃗
p) = α
⃗ ·p
⃗cv(−⃗p) + βmc v(−⃗ p),

where Ep is defined to be positive. The momentum labels on the negative-energy solutions

were labeled with the opposite momenta because once one switches to the language of fieled
operators they would correspond to the destruction of anti-particles with momentum −⃗
p rather
than the creation of particles of momentum p
⃗.
To find the solutions, first find the solutions for p = 0. The solutions are then
   
1 0
 0   1 
u↑ (p = 0) =   0  , u↓ (p = 0) =  0  ,
   (13.36)
0 0
   
0 0
 0   0 
v↓ (p = 0) =   1  , v↑ (p = 0) =  0  .
  

0 1

261
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

Note that the solutions v are labeled with a spin index opposite to the eigenvalue of Σz . Again,
this is because the solution will correspond to the destruction of an antiparticle. It is easy to
check that these solutions are eigenstates of the Dirac equation with eigenvalues ±mc2 .
Finding solutions for non-zero momentum can be accomplished by multiplying the zero-momentum
solutions by the operator ±Ep + α ⃗c + mc2 β. This results in a solution to the Dirac equation
⃗ ·p
because

(Ep − α ⃗c − βmc2 )(Ep + α

⃗ ·p ⃗c + βmc2 ) = 0.
⃗ ·p (13.37)

We define the solution us (⃗

p) to be

Ep + α ⃗c + βmc2
⃗ ·p
us (⃗
p) = p us (p = 0), (13.38)
2mc2 (mc2 + Ep )

which will also be a solution to the Dirac equation.

(E − α ⃗c − βmc2 )u(⃗
⃗ ·p p) = 0. (13.39)

The square root in the denominator was chosen to result in the normalization

Ep
u†s (p)us (p) = . (13.40)
mc2
The choice of normalization is motivated by the fact that u† u turns out to be related to the zeroth
component of the current, which must transform like the zeroth part of a four vector (E, p ⃗c)/m.
As an example, we find u↑ (⃗
p) where p
⃗ is along the z axis. For instance,
  
Ep + mc2 0 pz c 0 1
1  0 Ep + mc 2
0 −pz c  0 
u↑ (pz ) = p  2
 
2
2mc (Ep + mc ) 2  p z c 0 E p − mc 0  0 
0 −pz c 0 Ep − mc2 . 0
(13.41)
 
Ep + mc2
1  0 
= p   .
2mc2 (Ep + mc2 )  pz c 
0

At non-zero momentum the positive-energy solutions have a mixture of upper (top two) and
lower (bottom two) components, with the upper components dwarfing the lower components
when the motion is non-relativistic.
Example 13.2: The Infinite Square Well with a Scalar Potential
One can consider a binding potential as having a position dependent mass, i.e. a scalar field
Φ(x). Consider a spin-1/2 particle moving in one dimension with zero mass inside the well,

262
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

but infinite mass outside the well. The Dirac equation and well are
Eψ(x) = iℏcαx ∂x ψ(x) + Φ(x)βψ(x),

 M c2 , x<0
Φ(x) = 0, 0<x<L
M c2 , x>L


M → ∞.
1. For x < 0 and x > L show that the solution behaves as e−M c|x|/ℏ and e−M c|x−L|/ℏ
respectively.
2. Show that the boundary condition at the surface is
α · n̂ψ = ψ.
iβ⃗
Note this means that the wave functions are not vanishing at the boundary.
3. Using these boundary conditions find the ground state energy.
Solution:
(1.) Outside the well one can ignore the term in the Dirac equation proportional to E as that is
finite, and one must cancel the infinite terms. Ignoring the finite terms the equation becomes
(−iℏcαx ∂x − M c2 β)2 ψ(x) = 0,
−c2 ℏ2 ∂x2 ψ(x) + M 2 c4 ψ(x) = 0.
The term with ∂x can be arbitrarily large if the wave function changes a finite amount over
an infinitesimal distance, so it cannot be discarded. Here, the facts that α2x = β 2 = 1 and
{αx , β} = 0 were invoked. The solution must then behave as e−M c|x|/ℏ .
(2.) One can integrate the Dirac equation above,
Z 0 Z 0
2
dx[−iℏcαx ∂x ψ(x)] = −M c β dx ψ(x)
−∞ −∞
Z 0
−iℏcαx ψ(x = 0) = −M c2 β dx ψ(x = 0)e−M c|x|/ℏ ,
−∞
iαx ψ(x = 0) = βψ(x = 0).
Or more generally, iβ⃗ α · n̂ψ = ψ where n̂ is the unit surface vector point outside the well.
Thus, on the l.h.s. of the well iβαx ψ(x = 0) = ψ(x = 0) and on the r.h.s. iβαx ψ(x =
L) = −ψ(x = L).
(3.) Inside the well the particles are massless and it is easiest to consider the problem in the
chiral representation. In that case
 
σx 0
αx =
0 −σx
   
1 0
1  1  ikx
 1  0  e−ikx .
 
ψ+,k (x) = √  e , ψ+,−k (x) = √
2 0  2 1 
0 1

263
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

These are two positive-energy solutions for particles with spins in the x direction moving to
the left and right respectively. These solutions are also eigenstates of the spin operator Σx ,
with eigenvalue +1. There are also another two solutions that are eigenstates of Σx with
eigenvalues −1. Because Σx commutes with the one-dimensional Hamiltonian (both inside
and outside the well), we can choose one set of solutions.
To solve the B.C. one consider a linear combination,

ψ(x) = Aψ+,k (x) + Bψ+,−k (x).

To consider the B.C. one can write the matrix

 
0 σx
−iβαx = i ,
−σx 0

using the definitions of the Dirac matrices from the chiral representation in Eq. (13.27). At
x = 0, the B.C. can be applied,

iβαx {Aψ+,k (x = 0) + Bψ+,−k (x = 0)} = Aψ+,k (x = 0) + Bψ+,−k (x = 0),

iAψ+,−k (x = 0) − iBψ+,k (x = 0) = Aψ+,k (x = 0) + Bψ+,−k (x = 0).

This will be a solution of the B.C. if

B/A = i.

Repeating the considerations at x = L gives

B/A = −ie2ikL .

The negative sign appeared due to the fact that the unit surface vector switched directions at
the other end of the well. The conditions on k are thus

e2ikL = −1
kL = (n + 1/2)π, n = 0, 1, 2 · · ·

This differs from the Schrödinger equation result, kL = nπ. The ground-state momentum is
half the value of the Schrödinger case, but the separation of the values of k are the same. One
consequence of this is that if one considers a macroscopic Fermi gas of such particles, there is
no penalty is splitting the well into two wells, and thus there is no surface energy.

13.7 Non-Relativistic Limit and the g Factor of the Electron

Interaction with the electromagnetic field can be incorporated by replacing p
⃗ with p ⃗ and
⃗ − eA/c
iℏ∂/∂t with iℏ∂/∂t − eA0 , where A0 is the electric potential.
Using the Dirac representation, one can write the four components in terms of two two-component

264
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

vectors ϕ and χ, where ϕ and χ refer to the upper/lower components.

 
ϕ
ψ= . (13.42)
χ

One can now write the Dirac equation as two equations for ϕ and χ.

ℏ
 
∂ϕ e
iℏ =c ∇− A ⃗ ·⃗σ χ + (A0 + mc2 )ϕ, (13.43)
∂t i c
ℏ
 
∂χ e
iℏ =c ∇− A ·⃗⃗ σ ϕ + (eA0 − mc2 )χ. (13.44)
∂t i c

In the non-relativistic limit the lower components will be small. Given that the energy is domi-
nated by the mass, can make the approximation

∂
iℏ χ ≈ mc2 χ, (13.45)
∂t
which results in a simple substitution for χ using Eq. (13.44).

ℏ
 
e
2
2mc χ ≈ c ∇− A ⃗ ·⃗ σ ϕ. (13.46)
i c
Substituting this expression into Eq. (13.43),

ℏ
2
1
 
∂ e
iℏ ϕ= ∇− ⃗ ·⃗
A σ ϕ + (eA0 + mc2 )ϕ. (13.47)
∂t 2m i c
If the gradient and vector potential operators commuted with one another, one would use the
anti-commutation relations to obtain the usual kinetic energy piece in the Schrödinger equation
from squaring the p ⃗
⃗ − eA/c term and employing the anti-commutation relation. However,
taking account of such terms yields an extra term.

ℏ
2
1

∂ e eℏ X
iℏ ϕ= ∇− ⃗
A ϕ + (eA0 + mc2 )ϕ + σi σj (∂i Aj + Ai ∂j )ϕ.
∂t 2m i c 2mci i̸=j
(13.48)

Using the fact that σi σj = −σj σi for i ̸= j,

ℏ
2
1

∂ e eℏ X
iℏ ϕ= ∇− ⃗
A ϕ + (eA0 + mc2 )ϕ + σi σj (∂i Aj − Aj ∂i )ϕ
∂t 2m i c 2mci i̸=j
(13.49)
ℏ
2
1

e
= ∇− ⃗
A ϕ + (eA0 + mc2 )ϕ
2m i c
eℏ X
+ [σi , σj ](∂i Aj − ∂j Ai − Aj ∂i + Ai ∂j )ϕ.
4mci i<j

265
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

Finally, making use of the fact that for any function F (⃗

r ),

(∂i Aj − ∂j Ai − Aj ∂i + Ai ∂j )F (⃗
r ) = ϵijk Bk F (⃗
r ), (13.50)

and using the definition of the spin operator in Eq. (13.32, one finds the desired result,

ℏ
2
1

e eℏ
= ∇− ⃗
A ϕ + (eA0 + mc2 )ϕ − ⃗ · Bϕ.
Σ ⃗ (13.51)
2m i c 2mc

Because ℏΣ⃗ = 2S,⃗ the last term explains why the g factor of the electron is 2. Previously, this
had been inserted by hand. However, the Dirac equation requires that the g factor be exactly two
(though there are small perturbative corrections). Note that the g factor of the proton and neu-
tron are not equal to two, because they are composite particles. Being a composite particle, the
spin of the proton includes contributions from internal orbital angular momenta of the quarks
and gluons and from the intrinsic spins of the quarks and gluons.
The only difference between the expression above and the usual Schrödinger equation is in the
additional term mc2 . However, this merely adds a constant to the energy as long as mass is
conserved, and does not affect any observable in the non-relativistic limit. This rapid oscillation
was referred to as “zitterbewegung” (trembling motion) by Schrödinger, https://en.wikiped
ia.org/wiki/Zitterbewegung.

13.8 The Spin-Orbit Interaction

In the previous section an approximation was made that the iℏ∂t χ ≈ mχ in order to find
an expression for ϕ without χ. The terms in that expression that are not in the Schrödinger
equation, e.g. the coupling of the spin to the magnetic field, can be considered as an expansion
in powers of 1/m. One can find that expansion rigorously by performing the Foldy-Wouthuysen
transformation, . This refers to a unitary transformation of the Hamiltonian that eliminates terms
that mix upper and lower components. It generates an infinite sequence of additional terms in
H that can be arranged in powers of 1/m. As this transformation is a tad onerous, it will not be
described in detail here.
To order 1/m2 , the additional terms are,

p4 eℏ eℏ2
δH = − +i [(⃗ ⃗ σ·p
σ · E)(⃗ ⃗) − (⃗
σ·p
⃗)(⃗ ⃗ +
σ · E)] ∇2 A 0 , (13.52)
8m3 c2 8m2 c 2 8m2 c2
p4 eℏ eℏ2
=− − [⃗ ⃗ ×p
σ · (E ⃗]) + ∇ 2 A0 ,
8m3 c2 4m2 c 2 8m2 c2

where E ⃗ is the electric field. Here, we have kept terms that are of order 1/m2 of the usual non-
relativistic terms, noting that p2 /2m is one
p of the non-relativistic terms. The first term above is
merely the next-order expansion of E = p2 c2 + m2 c4 , while the second term is the spin-orbit
term. To show that the second term is the spin-orbit term, we note that

E(r)
⃗ ×p
E ⃗= r×p
⃗ ⃗, (13.53)
r

266
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

⃗ The final term in Eq. (13.52) is known as the Darwin term (derived
for a radial electric field E.
by Charles Galton Darwin, the grandson of the more famous Charles Darwin). For the Coulomb
potential, A0 = −Ze/r, it can be re-expressed as

Zℏ2 e2
HDarwin = δ 3 (⃗
r ), (13.54)
8m2 c2
and thus appears mainly for s− wave terms.
Finally, it is noted that one may go through the same exercise without any electromagnetic field,
but instead with a position dependent mass m(r). One then finds that the spin-orbit term looks
identical except that

∂m(r)c2
⃗ →
eE . (13.55)
∂r
This has the opposite contribution for an attractive interaction. This is important for understand-
ing nuclear physics where the spin-orbit interaction is surprisingly large. It can be explained by
an attractive scalar interaction (like a position dependent mass) and a repulsive vector interac-
tion (similar to a Hydrogen atom, but with opposite sign because it is repulsive). The interactions
cancel each other out to a large degree as far as the binding energies are concerned but the con-
tributions from the spin-orbit terms add together. Thus, the spin-orbit interaction in nuclear
physics is surprisingly large, and of the opposite sign as what one has in atomic physics.
The Dirac equation is one of the great triumphs of twentieth century physics. Motivated by
aesthetic considerations, several previously ad-hoc assumptions fall out naturally: the g factor,
particle-antiparticle symmetry, spin-orbit coupling, etc. When combined with the relativistic
coupling to the quantum electromagnetic field, phenomenally accurate calculations can be made
of g − 2 using perturbation theory. But, this is material for another course.

13.9 More Notation: γ Matrices

To more clearly demonstrate the covariant nature of the equations, it is common to define the
Dirac γ matrices.

γ 0 ≡ β, ⃗
γ ≡ β⃗
α. (13.56)

The four matrices γ µ then transform into one another as a four-vector. The three space-like γ
matrices are anti-Hermitian, while γ 0 is Hermitian. The convenient covariant behavior of the γ
matrices comes from the property,

{γ µ , γ ν } = 2g µν , (13.57)

where g µν is the metric tensor

 
1 0 0 0
 0 −1 0 0 
g µν =
 0 0 −1 0  .
 (13.58)
0 0 0 −1

267
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

This form confirms that γ µ was a four-vector.

A very common notation is
/p ≡ pµ γµ , (13.59)
which means that the Dirac equation (multiplied by γ0 ) can be expressed as
(/p − m)ψ = 0. (13.60)
This has the attraction that /p is a scalar.
One can write the current as
j µ = ψ̄γ µ ψ, (13.61)
where
ψ̄ ≡ ψ † γ0 . (13.62)
This seems a bit peculiar, but ψ̄ψ is a scalar rather than ψ † ψ which transforms like the zeroth
component of the four current. To see that ψ̄ψ is a scalar under boosts, we recall Eq. (13.38)
where the operator used to boost the solution from p ⃗ = 0 to p⃗ was
E+α⃗ ·p
⃗ + βm
B= p , (13.63)
2m(m + Ep )
which is Hermitian. Next, we consider the boosted value of ψ̄ψ where the boosted wave func-
tion is ψ and the p = 0 solution is ψ0 ,
ψ̄ψ = ψ0† B † βBψ0 (13.64)
E+α ⃗ ·p
⃗ + βm E α ⃗ ·p
⃗ + βm
= ψ0† p βp ψ0
2m(m + Ep ) 2m(m + Ep )
1
= ψ † β(E − α⃗ ·p⃗ + βm)(E α
⃗ ·p
⃗ + βm)ψ0
2m(m + Ep ) 0
1
= ψ † β(E 2 − p2 + m2 + 2mEβ − 2mβ⃗ α·p⃗)ψ0 .
2m(m + Ep ) 0
For the at-rest solution ψ0† βαi ψ0 = 0 because α mixes upper and lower components, and
ψ̄0 ψ = ψ0† ψ. Substituting E 2 = p2 + m2 then gives
ψ̄ψ = ψ̄0 ψ0 . (13.65)
Though this is not a proof, it certainly shows that ψ̄ψ a Lorentz invariant. Similarly from our
study of the currents, ψ̄γ µ ψ behaves as a four vector.
To more rigorously see the Lorentz structure of all the terms and more consistently derive the
expressions for the currents, it is best to start with the Lagrangian density,
L = ψ̄(/p + m)ψ, (13.66)
then use Noether’s theorem to derive the equations of motion (Dirac’s equation) and conserved
currents. The Lagrangian density above is manifestly boost-invariant, as it should be. To build
a scalar, one need only require that ψ̄ and ψ surround operators constructed from four-vectors
where all the Lorentz indices are contracted. The four matrices γ µ indeed behave like four
vectors under boosts.

268
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

13.10 Dirac Field Operators

For the non-relativistic case field operators were defined as
1 XX
Ψi (⃗
r , t) = √ p)e−iEp t/ℏ+i⃗p·⃗r/ℏ bs,⃗p .
us,i (⃗ (13.67)
V p⃗ s

Here, the two-component vector ui (s) describes how the spin s is projected onto the chosen
basis. For example i = 1 might refer to spin-up along the z axis and i = 2 to spin-down. The
spin label s might refer to some other basis, such as along the direction of p ⃗, or perhaps along
another axis. The field operator Ψi (⃗ r ) lowers the charge at position ⃗ r by destroying a particle
of spin projection i, whereas Ψ† (⃗r ) creates a particle at ⃗
r . The operators bs,⃗p destroy particles of
momentum p ⃗ with spin s. Because the basis for the spins when referencing momentum states,
labeled s, might differ from the basis used for coordinate space, labeled i, the two-by-two ma-
trix us,i (⃗
p) is simply the representation of a basis transformation which might differ for each
momentum value p ⃗.
For the Dirac case, the field operator will have four components, with i = 1 · · · 4. The extra
components account for the existence of anti-particles.
s
1 X m X 
Ψi (⃗
r , t) = √ us,i (⃗ p)eiEp t/ℏ−i⃗p·⃗r/ℏ d†s,⃗p . (13.68)
p)e−iEp t/ℏ+i⃗p·⃗r/ℏ bs,⃗p + vs,i (⃗
V p⃗ Ep s

The vectors us,i (⃗

p) are normalized as
u†s (⃗ p) = vs† (⃗
p)us′ (⃗ p)vs′ (⃗
p) = (Ep /m)δss′ , (13.69)
u†s (⃗
p)vs (−⃗
p) = vs† (⃗
p)us (−⃗
p) = 0.

u†s,i (⃗ p) = (Ep /m)δss′ . Here the operator d†s (⃗

P
i p)us′ ,i (⃗ p) creates an anti-particle, which is
the same as destroying the negative-energy state of momentum −⃗ p. Note that the overlaps
of us with vs are evaluated at opposite momenta. This is to ensure that any two operators
that are solutions of a given momentum p ⃗, i.e. the coordinate-space dependence is e+i⃗p·⃗r are
orthonormal.
The charge density operator Ψ†i (⃗ r , t)Ψi (⃗ r , t) can be integrated to yield the total charge,
m
Z Z X
d3 r Ψ†i (⃗
r , t)Ψi (⃗
r , t) = d3 r (13.70)
p
⃗,s,⃗ ′
p ,s ′
E p V
n
† p′ )·⃗
r /ℏ † † ′
us,i (⃗p)us′ ,i (⃗ ′
p )e p−⃗
i(⃗
bs,⃗p bs′ ,⃗p′ + vs,i (⃗ p′ )ei(⃗p −⃗p)·⃗r/ℏ ds,⃗p d†s′ ,⃗p′
p)vs′ ,i (⃗
′ ′
o
+u†s,i (⃗ p′ )ei(⃗p+⃗p )·⃗r/ℏ b†s,⃗p d†s′ ,⃗p′ + vs,i
p)vs′ ,i (⃗ †
(⃗ p′ )e−i(⃗p+⃗p )·⃗r/ℏ ds,⃗p bs′ ,⃗p′
p)us′ ,i (⃗
Xn † o
= bs,⃗p bs,⃗p + ds,⃗p d†s,⃗p
p
⃗,s
Xn † o
= bs,⃗p bs,⃗p − d†s,⃗p ds,⃗p + 1 .
p
⃗,s

The vacuum thus has a contribution to the charge density from each spin and for each value of
the momentum.

269
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

13.11 The Dirac Equation and the Mass Term

The Hamiltonian for the Dirac equation for massless particles can be written in terms of creation
and destruction operators,
X
H0 = ℏkc(b⃗†k b⃗k + d†−⃗k d−⃗k ), (13.71)
⃗
k

where b† and d† correspond to creation and destruction operators for particles and antiparticles
respectively. For this discussion we ignore the sums over spin or the vacuum energy.
If the mass term is treated like a perturbation, it can also be written in terms of the same cre-
ation/destruction operators,
Z
2
V = mc d3 r Ψ̄(⃗ r )Ψ(⃗
r) (13.72)
X
= mc2 (b⃗†k d†−⃗k + d−⃗k b⃗k ).
⃗
k

For each momentum mode we can define two new operators,

α†k ≡ cos θk b⃗†k + sin θk d−k , (13.73)

βk† ≡ cos θk d†−⃗k − sin θk b⃗k .

These operators satisfy the commutation relations,

{αk , α†q } = {βk , βq† } = δk,q , (13.74)

{αk , βq } = {βk , α†q } = {αk , βq† } = {α†k , βq† } = 0.

If one considers the Hamiltonian,

X
H = Ek (α†k αk + βk† βk ) (13.75)
k
X n h i h io
† † 2 † †
= Ek cos 2θk b⃗k b⃗k + d−⃗k d−⃗k + 2 sin θk + sin 2θk b⃗k d−⃗k + d−⃗k b⃗k ,
k

one may see that the H will equal H0 + V to within a constant if

Ek cos 2θk = ℏck, (13.76)

Ek sin 2θk = mc2 ,

or equivalently,
p
Ek = ℏ2 k2 c2 + m2 c4 , (13.77)
mc
tan 2θk = .
ℏk

270
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

Thus,
X
H = H0 + V + 2 Ek sin2 θk (13.78)
k
X
= H0 + V + Ek (1 − cos 2θk )
k
s !
X 1
= H0 + V + Ek 1−
k
1 + tan2 2θk
X p 
= H0 + V + (ℏck)2 + m2 c4 − ℏck .
k

The last term is a correction to the vacuum energy which is simply the difference between the
vacuum energies with and without the mass. This can be thought of as a correction to the energy
of the Dirac sea.

13.12 Exercises
1. To show why derivatives are defined as shown in Eq. (13.7), show that

∂µ x2 = 2xµ , and ∂ µ x2 = 2xµ ,

where x2 = x20 − x21 − x22 − x23 .

2. Consider a charged relativistic particle interacting with the electromagnetic field, and de-
scribed by the Klein-Gordon equation.

(iℏ∂t − eA0 )2 + c2 ℏ2 ∂x2 − m2 c4 ψ(x, t) = 0

 

The electrostatic potential A0 is illustrated in the diagram below.

I. II. eΦ
A 0
E-mc 2

Consider a solution for a particle incident from the left,

ψI (x, t) = e(−iEt+ikx)/ℏ + Be(−iEt−ikx)/ℏ

′
ψII (x, t) = Ce(−iEt+ik x)/ℏ ,
√
where E = m 2 c4 + k 2 .
Calculate the charge and current densities (include direction) in regions I and II for each of
the following three cases.

(a) eA0 < E − mc2 .

271
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

(b) E − mc2 < eA0 < E + mc2 .

(c) eA0 > E + mc2 .

3. Consider the same case as above, except with no electrostatic potential. Instead, consider
a different mass in region I and region II, with mII > mI . For each of the following two
cases, calculate the charge and current densities in regions I and II.

(a) E > mII c2

(b) E < mII c2

4. Consider the Dirac representation,

I 0
   
0 ⃗
σ
β= ⃗ =
α
0 −I σ 0
⃗

and the chiral representation,

   
0 −I σ 0
⃗
β= ⃗ =
α
−I 0 0 −⃗
σ

The spinors, u↑ and u↓ , represent positive-energy eigenvalues of the Dirac equation as-
suming the momentum is along the z axis.

(mβ + pz αz ) u(pz ) = Eu(pz ) ,

The spin labels, ↑ and ↓ refer to the positive and negative values of the spin operator, which
in both representations is  
σz 0
Σz =
0 σz
Write the four-component spinors u↑ and u↓ in terms of p, E and m :

(a) in the Dirac representation.

(b) in the chiral representation.
(c) in the limit pz → 0 for both representations.
(d) in the limit pz → ∞ for both representations.

5. Consider a solution to the Dirac equation for massless particles, u+ (⃗ p), where the + de-
notes the fact that the solution is an eigenstate of the spin operator in the p̂ directions,
⃗ · p̂)u+ (⃗
(Σ p, x) = u+ (⃗
p, x).

Show that the operator β operating on u+ (p) gives a negative energy solution but is still
⃗ · p̂ with eigenvalue +1.
an eigenstate of Σ
6. Show that the operator
1
P = (Ep + α
⃗ ·p
⃗ + βm)
2Ep
(a) is a projector, i.e. P 2 = P .

272
PHY 851/852 13 RELATIVISTIC QUANTUM MECHANICS

(b) and that P |ψ⟩ gives a positive-energy solution to the Dirac equation when operating
on any state |ψ⟩,
(Ep − α⃗ ·p⃗ − βm)P |ψ⟩ = 0.
7. Consider a massless spin half particle of charge e in a magnetic field in the ẑ direction
described by the vector potential
⃗ = Bxŷ.
A
The Hamiltonian is then

H = αx (−iℏ∂x ) + αy (−iℏ∂y − eBx).

(a) Show that the Hamiltonian commutes with −iℏ∂y and −iℏ∂z .
(b) The wave function can then be written as

ψky ,kz (x, y, z) = eiky y+ikz z ϕky ,kz (x),

After setting ky = kz = 0, show that the lowest energy can be found by solving the
equation
E 2 ϕ± (x) = (−ℏ2 ∂x2 + e2 B 2 x2 − eℏBΣz )ϕ± (x).
(c) Show that the eigenvalues of the operator H 2 are
2
E± = (2n + 1 ∓ 1)eℏB, n = 0, 1, 2·,

where the ± refers to eigenvalues of Σz . You can do this mapping to the harmonic
oscillator and then using the solutions to the harmonic oscillator from Chapter 3. Note
that when the the eigenvalue of Σz is +1, there exists a solution with E = 0.
8. Using the definition of field operators in Eq. (13.68), show that the Hamiltonian
Z
H = d3 r Ψ† (⃗ r , t)(−iℏ⃗ α · ∇ + βm)Ψ(⃗ r , t) (13.79)
X
= Ep (b†s,⃗p bs,⃗p + d†s,⃗p ds,⃗p − 1).
s,⃗
p

I.e. the vacuum energy for each mode is negative.

9. Using the definitions for αk and βk in Eq. (13.73),
(a) Show that

b⃗†k b⃗k − d†−⃗k d−⃗k = α† αk − β † βk .

This demonstrates that the eigenstates of the new Hamiltonian are still eigenstates of
the charge operator written in the old basis.
(b) Show that the state
|0̃⟩ ≡ cos θk |0⟩ + sin θk d†−⃗k b⃗†k |0⟩
is destroyed by both αk and βk , where |0⟩ is the vacuum in the old basis. Effectively,
this shows that |0̃⟩ is the vacuum in the new basis.

273
PHY 851/852 14 STATES WITHOUT CONSERVED PARTICLE NUMBER

14 States Without Conserved Particle Number

Thus far, the initial and final states that we have considered all have a fixed number of particles.
That number might have changed between the initial and final state, e.g. an excited state emit-
ting a photon, but the basis being used was purely one where the n particles were well defined,
typically by their momentum. However, there are numerous applications where the basis being
used is not one of fixed particle number. This is even the case when the particles carry charge. In
the next section the Bogoliubov treatment of pairing is presented. In this treatment a variational
wave function involving a mixture of states with charge zero (zero electrons) and charge −2e (2
electrons) is proposed. The subsequent section introduces the idea of coherent state for bosons.
This basis was applied by Roy Glauber to describe the photon pulse from a laser. It is also serves
as a basis for lattice gauge theory. These topics will not be pursued in great detail, but a student
should get a decent knowledge of such bases, and perhaps gain some level of comfort with this
peculiar choice of states.

14.1 Bogoliubov Operators and Cooper Pairs

The pairing interaction in BCS theory can be written in terms of creation and destruction opera-
tors as
v0 X′ † †
V =− b d d ⃗ ′ b⃗ ′ . (14.1)
Ω k,k′ ⃗k −⃗k −k k

Here, we have only considered pairs that sum to total momentum zero. A more general expres-
sion would involve all combinations of the four momenta that sum to zero. In this expression
there are constraints for both the b† d† and bd terms. The b† and d† operators refer to electron
spin-up and spin-down creation operators respectively. We do this rather than fouling the for-
malism with more subscripts. The interaction represents the scattering of particles of opposite
momenta ⃗ k′ and −⃗
k′ into states ⃗
k and −⃗ k. The primed sum limits the sum to a region near the
Fermi surface.
Now, we consider as a variational wave function a state where each momentum mode is defined
as
Y′  
|Ψ⟩ = cos θk + sin θk b†−⃗k d⃗†k |0⟩ (14.2)
k

When θk becomes non-zero this state is no longer an eigenstate of the number operator or electric
charge, because the state mixes components with different numbers of electrons. The expecta-
tions of the relevant creation/destruction operator combinations are

⟨Ψ|b⃗†k b⃗k + d†−⃗k d−⃗k |Ψ⟩ = 2 sin2 θk , (14.3)

1
⟨Ψ|b⃗k d−⃗k |Ψ⟩ = sin 2θk .
2

274
PHY 851/852 14 STATES WITHOUT CONSERVED PARTICLE NUMBER

Next, we consider the energy of particles in the primed region,

X′ v0 X′ X′
⟨Ψ|H0 + V |Ψ⟩ = ϵk ⟨Ψ|b⃗†k b⃗k + d†−⃗k d−⃗k |Ψ⟩ − ⟨Ψ|b⃗†k d†−⃗k |Ψ⟩ ⟨Ψ|b⃗k′ d−⃗k′ |Ψ⟩
k
Ω k k′
(14.4)
X′  ∆

2
= 2ϵk sin θk − sin 2θk ,
k
2

where
v0 X′
∆≡ sin 2θk′ (14.5)
2Ω k′

The important step here is in writing ⟨Ψ|b⃗†k d†−⃗k d−⃗k′ b⃗k′ |Ψ⟩ = ⟨Ψ|b⃗†k d†−⃗k |Ψ⟩⟨Ψ|b⃗k′ d−⃗k′ |Ψ⟩. This
follow from the ansatz for the product form of the state, Eq. (14.2), and from ignoring terms
where ⃗k=⃗ k′ . These terms are ignored because they are an infinitesimal fraction of the terms in
the continuum limit where many states are included in the sum.
Now, one must choose the angles θk so that the energy is minimized compared to the extra
particles coming from a reservoir with chemical potential µ. (Here, µ is the Fermi energy).

∆
 
∂ X′ 2
2(ϵk − µ) sin θk − sin 2θk = 0. (14.6)
∂θk k 2

After taking the derivatives, and remembering that ∆ depends on θ, this leads to the solution

∆
tan 2θk = . (14.7)
ϵk − µ

Because ∆ is a function of the angles, this is a transcendental equation. One can guess a a value
of ∆, then find the angles θk . One can then find the new ∆ using Eq. (14.5), and iterate until a
consistent solution is found.
The probability that a state ⃗
k is occupied by a particle,

1 1 (ϵk − µ)
⟨b⃗†k b⃗k ⟩ = sin2 θ⃗k = − p .
2 2 (ϵk − µ)2 + ∆2

The occupation is 1/2 when ϵk = µ, is zero for ϵk >> µ and is unity for ϵk << µ. Pairing
smooths the step function one expects for non-interacting particles with a scale ∆.
We also note the difference of this solution with the relativistic Dirac problem above. In that case
the ground state was a mixture of the old vacuum and a particle/anti-particle pair, i.e. |0⟩ and
b†k d†−k |0⟩. However, in this approach the solution is a mixture of zero-electron and two-electron
states, thus not even electric charge is fixed. The analogy between superconductivity and par-
ticle/nuclear physics is reflected in the jargon, as the generation of masses due to spontaneous
symmetry breaking is referred to as the creation of a gap.

275
PHY 851/852 14 STATES WITHOUT CONSERVED PARTICLE NUMBER

Example 14.1: Show that same-spin particles don’t pair

If we had considered particles of the same spin,
Y′  
|Ψ⟩ = cos θk + sin θk b⃗†k b†−⃗k |0k ⟩,
k

the resulting expectation for the pairing would be zero. This follows because
 
† †
(b⃗k b−⃗k + b−⃗k b⃗k ) cos θk + sin θk b⃗k′ b−⃗k′ |0⟩ = 0.

The cancellation results from the fact that ⃗k could equal either ⃗ k′ or −⃗
k′ . There is only one
possible connection for the case where the product of b⃗k b−⃗k is replaced by b⃗k d−⃗k . One way
of stating this result is that an s-wave coupling must combine with an anti-symmetric spin
combination because the overall pair wave function must be anti-symmetric. If the original
potential had odd-parity, e.g. it was proportional to k, then one could form a p wave like state
with symmetric spin. The wave function might have a form
Y′  
|Ψ⟩ = k) sin θk b⃗†k b†−⃗k |0k ⟩,
cos θk + f (⃗
k

where f is an odd-parity function of ⃗

k.

14.2 Coherent States

Our last example of a state which is not an eigenstate of particle number is referred to as a
“coherent” state.
∗ η/2 †
|η⟩ = e−η eηa |0⟩ (14.8)
∗ η/2
X (ηa† )n
= e−η |0⟩,
n
n!

where η is a complex number. Coherent states are combinations of states with different particle
number, i.e. when you expand the exponential you find a linear combination of a† |0⟩, (a† )2 )|0⟩, (a† )3 ∥0⟩.
The name “coherent” refers to the fact that the coefficients of the various terms, |n⟩, do not have
random phases. Of course, we are assuming that the excitations are bosonic, i.e. a† a† ̸= 0.
One can check to see that this state is properly normalized by calculating the overlap,

∗ †
X (η ∗ a)m X (ηa† )n
⟨0|eη a eηa |0⟩ = ⟨0| |0⟩ (14.9)
m
m! n
n!
X (η ∗ η)n
= ⟨0| 2
an (a† )n |0⟩
n
(n!)
∗
X (η η)n ∗
= = eη η .
n
n!

276
PHY 851/852 14 STATES WITHOUT CONSERVED PARTICLE NUMBER

A coherent state is an eigenstate of the destruction operator.

∗ η/2
X (ηa† )n
a|η⟩ = e−η a |0⟩ (14.10)
n
n!
X nη n (a† )n−1
= |0⟩
n
n!
X η n−1 (a† )n−1
=η |0⟩
n
(n − 1)!
= η|η⟩.

Thus, it is simple to calculate matrix elements of coherent states. For example,

⟨γ|(a† )3 a2 |η⟩ = (γ ∗ )3 η 2 ⟨γ|η⟩. (14.11)

Finally, the same coherent state in Eq. (14.8) can be expressed as

∗ η/2 † † −η ∗ a
|η⟩ = e−η eηa |0⟩ = eηa |0⟩. (14.12)

Because the argument of the exponential in the latter form is manifestly anti-Hermitian, it is easy
to see that ⟨η|η⟩ = 1. It is left as part of a homework problem to prove the equivalence of the
two forms (use the Baker-Campbell-Hausdorff theorem).

14.3 Emission Via Coupling to a Classical Current

The physical importance of coherent states comes from the fact that they are the solution to the
following Hamiltonian.

H(t) = H0 + V (t). (14.13)

H0 = ϵa† a,
V (t) = j(t)[a† + a].

Here j(t) plays the role of an external classical current that couples to a quantum field, e.g. j · A.
As a solution of the Hamiltonian, one can propose a state |η(t)⟩I in the interaction representa-
tion,
∗ (t)η(t)/2 † † ∗
|η(t)⟩I = e−η eη(t)a |0⟩ = eη(t)a −η (t)a |0⟩, (14.14)
−i t
Z
′
where η(t) ≡ dt′ eiϵt /ℏ j(t′ ).
ℏ −∞
This state is manifestly normalized because the argument of the exponential is anti-Hermitian.
Viewing the evolution of |η⟩I with time,
∂
|η(t)⟩I = j(t)eiϵt/ℏ a† + j(t)e−iϵt/ℏ a |η(t)⟩I


iℏ (14.15)
∂t
j(t) a† (t) + a(t)) |η(t)⟩I .

277
PHY 851/852 14 STATES WITHOUT CONSERVED PARTICLE NUMBER

Here, a(t) is the destruction operator in the interaction representation,

a(t) = e−iH0 t/ℏ aeiH0 t/ℏ = e−iϵt/ℏ a, (14.16)

a† (t) = e−iH0 t/ℏ a† eiH0 t/ℏ = eiϵt/ℏ a† .

This then gives

∂
iℏ |η(t)⟩I = VI (t)|η(t)⟩I , (14.17)
∂t
which demonstrates that our guess at a form for |η(t)⟩I was successful.
Coupling to an external current is a common consideration in a wide variety of physical prob-
lems. Even when spatial degrees of freedom are considered, e.g. j · A, one finds simple analytic
solutions. They then often have the form
 Z  Z
3 †
|η⟩ ∼ exp −i d p j(⃗ p)ap⃗ |0⟩, j(⃗p) ∼ d4 x eip·x j(x). (14.18)

This form assumes that the solution addresses at which the external current j(x, t) has vanished.
Any classical current that couples linearly to the field will radiate particles in a coherent state.
Some lasers are described with coherent states, when one can justify that the current behaves
classically. A classical current is one that is unaffected by the emission. When an atom emits a
photon the atom then changes to a new state. For a classical current, no such change occurs. To
justify treating a current as classical, one typically considers macroscopic emitters. For a laser,
one can think of a crystal which is coupled to the field as providing the classical current.

14.4 Completeness Relations for Coherent States

A useful property of any basis is completeness,
X
⟨m|α⟩⟨α|n⟩ = δmn . (14.19)
α

Coherent states provide a complete basis when all values of η in the complex plane are consid-
ered.
1
Z
dηR dηI ⟨m|η⟩⟨η|n⟩ = δmn , (14.20)
π
where ηR and ηI are the real and imaginary parts of η. To prove this, we expand the matrix
elements.
2 /2 (ηa† )m
⟨m|η⟩ = e−|η| ⟨m| |0⟩, (14.21)
m!
ηm 2 /2
= e−|η|
√ ,
m!
m ∗ n
−|η|2
η (η )
⟨m|η⟩⟨η|n⟩ = e √ .
n!m!

278
PHY 851/852 14 STATES WITHOUT CONSERVED PARTICLE NUMBER

Writing the integral over the real and imaginary parts of η as

Z Z
dηR dηI → dϕ|η|d|η|, (14.22)

where ϕ is the complex phase of η, allows one to see that the integral over ϕ will eliminate all
terms with m ̸= n because
Z
dϕei(m−n)ϕ = 2πδmn , (14.23)

2 |η|m+n
⟨m|η⟩⟨η|n⟩ = e−|η| ei(m−n)ϕ √ .
n!m!
One may now rewrite the expression using the fact that integrating over ϕ constrains the result
to be proportional to δmn .
1 |η|2n
Z Z
−|η|2
dηR dηI ⟨m|η⟩⟨η|n⟩ = 2δmn |η|d|η| e . (14.24)
π n!
By making the substitution u ≡ |η|2 , one can perform the integral and see that
1
Z
dηR dηI ⟨m|η⟩⟨η|n⟩ = δmn . (14.25)
π

14.5 Path Integrals

Path integrals provide an alternative means to express quantum mechanics. Rather than discrete
states, the basis for each mode becomes coherent states described by some complex number
η. Path integrals form the basis for some non-perturbative methods, principally lattice gauge
theory.
Completeness as displayed in the previous section, combined with the fact that the coherent
states are eigenstates of the destruction operator, allows one to calculate all matrix elements
X
⟨αf |e−iHt/ℏ |αi ⟩ = ⟨αf |e−iH(t−t1 )/ℏ |α1 ⟩⟨α1 |e−iH(t1 −t2 )/ℏ |α2 ⟩ (14.26)
α1 ,α2 ···

⟨α2 |e−iH(t2 −t3 )/ℏ |α3 ⟩ · · · ⟨αn |e−iH(tn −ti )/ℏ |αi ⟩
as an integral over complex fields ηn rather than a sum over discrete states αn . Furthermore,
if the Hamiltonian is normal ordered, that is each term has all creation operators pushed to the
left,
†
⟨ηn |e−iH(a ,a)δt/ℏ |ηn+1 ⟩ = 1 − iH(ηn∗ , ηn+1 )δt/ℏ ⟨ηn |ηn+1 ⟩.

(14.27)
Going further,
∗ η/2
⟨ηn |ηn+1 ⟩ = e−η ⟨0|eηa |η + δη⟩ (14.28)
−η ∗ η/2 η ∗ (η+δη)
=e e ⟨0|η + δη⟩
−η ∗ η/2 η ∗ (η+δη) ∗ (η+δη)/2
=e e e(η+δη)
∗ δη−δη ∗ η)/2
= e(η
∗ η̇−η̇ ∗ η)δt/2
= e(η .

279
PHY 851/852 14 STATES WITHOUT CONSERVED PARTICLE NUMBER

This gives
† ,a)δt/ℏ
⟨ηn |e−iH(a |ηn+1 ⟩ = 1 + −iH(ηn∗ , ηn+1 )/ℏ + (η ∗ η̇ − η̇ ∗ η)/2 δt.
 
(14.29)

Thus, the matrix element is transformed into a purely numerical function of the complex number
η. The process of breaking up the evolution operator into the product of many individual pieces
e−iHδt/ℏ , inserting a complete set of states between each piece, expressing them in terms of some
variables at each point (η) and integrating over all η is known as a path integral.
Finally, we point out that rather than thinking of the variable
√ η, one can also write the integral
in terms of the real variables p and q, where η = (p + iq)/ 2ℏ. In that case
1 1
dηR dηI → dpdq , (14.30)
π 2πℏ
and
† ,a)δt/ℏ
⟨ηn |e−iH(a |ηn+1 ⟩ = {1 + i[pq̇/2 − q ṗ/2 − H(p, q)]δt/ℏ} . (14.31)

The final matrix element is thus

dp(ti )dq(ti )
 Z 
⟨η(tf )|η(t0 )⟩ = Πi exp i dt[(pq̇ − q ṗ)/2 − H(p, q)]/ℏ . (14.32)
2πℏ
R
Because one integrates over all t, the contribution from a term ei dt(pq̇+qṗ)/ℏ is independent of
the intermediate values of p and q, as long as one is looking at specific initial and final states. I.e,
Z tf Z tf
dt[ṗq + pq̇] = dt(d/dt)(pq) (14.33)
ti ti
= p(tf )q(tf ) − p(ti )q(ti ).

Thus, it leads to a path-independent phase and can be neglected. This allows one to make the
replacement in Eq. (14.31),

i(pq̇ − q ṗ)/2ℏ → ipq̇/ℏ. (14.34)

The final matrix element becomes

dp(ti )dq(ti )
 Z 
⟨η(tf )|η(t0 )⟩ = Πi exp i dt[pq̇ − H(p, q)]/ℏ . (14.35)
2πℏ
Lo and behold, this phase looks like the Lagrangian integrated over time, i.e. the action. Thus,
quantum physics can be re-expressed as a path integral over coordinates p and q. In quantum
field theory, every point in space is reduced to a state, using such coordinates. This is how
Lagrangian field theory, which is the theoretical backbone of the standard model, is described in
terms of fields at each point in space encapsulated by numbers, can be translated into quantum
mechanics.
In the limit that ℏ → 0, the only path that should contribute is one where the phase,
Z Z
S≡ dt [pq̇ − H(p, q)] = dt L(p, q), (14.36)

280
PHY 851/852 14 STATES WITHOUT CONSERVED PARTICLE NUMBER

is stationary with respect to small variations of the path. The quantity S is known as the action,
and using the calculus of variations leads to Lagrange’s classical equations of motion.
In addition to being occasionally useful for calculating evolution operators, path integrals can be
applied to thermodynamic partition functions which require calculating the trace of e−βH . Real-
time path integrals tend to be problematic because of the complex phase factors. Because the
integrals are typically performed by Monte Carlo procedures there is a huge amount of cancella-
tion from different paths, which makes the calculations extremely noisy. Because the thermody-
namic trace considers a path in imaginary time, the calculations are far less noisy. Lattice gauge
theory is the numeric enterprise of calculating such path integrals, and is most often employed
for QCD.

14.6 Coherent Fields in Coordinate Space – Ginzburg Landau Theory of Su-

perconductivity
One can define a coherent state in a coordinate-space basis. Using bosonic field operators obey-
ing the commutation relations,
r ), Φ† (⃗
[Φ(⃗ r ′ )] = δ(⃗ r ′ ),
r−⃗ (14.37)
one constructs a normalized coherent state defined by a complex function η(x),
Z
†
R
|η⟩ = exp − dx |η(x)|2 /2e d⃗r η(⃗r)Φ (⃗r) |0⟩. (14.38)

With this definition,

Φ(⃗ r )|η⟩, ⟨η|Φ† (⃗
r )|η⟩ = η(⃗ r ) = η ∗ (⃗
r )⟨η|. (14.39)
Next, one can imagine a Hamiltonian,
1
Z
H = r Φ† (⃗
d⃗ r )∇2 Φ(⃗
r ) + V (⃗r ), (14.40)
2m
a2 b2
r ) = − Φ† (⃗
V (⃗ r )Φ(⃗r ) + Φ† (⃗ r )Φ† (⃗
r )Φ(⃗
r )Φ(⃗
r ).
2 4!
One can then choose η(x) to minimize the energy. This can be considered a variational calcu-
lation, where η(x) represents a set of variational parameters for each point in coordinate space.
To minimize the kinetic energy term it makes sense to pick a constant function. The value of η
is such that
a b
Minimum = − η ∗ η − (η ∗ η)2 . (14.41)
2 4!
There is a symmetry about the complex plane, so one can choose η to be real. Finding the value
of η to minimize the energy,
b
aη = η3, (14.42)
6
r
6a
η= .
b

281
PHY 851/852 14 STATES WITHOUT CONSERVED PARTICLE NUMBER

One can think of the field as being defined by a magnitude and a phase, η = |η|eiϕ . One can
r ) being a degree of freedom. With |η|
imagine fixing η to minimize the energy, then having ϕ(⃗
fixed the Hamiltonian becomes
ℏ2 |η|2
⟨η|H|η⟩ = Vmin + r ))2 .
(∇ϕ(⃗ (14.43)
2m
Thus, the phase of η becomes a degree of freedom. The Hamiltonian above is the basis of the
Ginzburg-Landau phenomenological theory of superconductivity. In that case, Cooper pairs
are the bosons. One can add electromagnetic interactions by minimal substitution, ℏ/i∇ →
⃗
ℏ/i∇ − q A/c, with q = −2e because a Cooper pair is composed of two electrons. This looks
somewhat like a Shrödinger equation for ϕ, but one must keep in mind that ϕ is real and that
ϕ2 should not be associated with a charge density.

14.7 Exercises
1. Consider bosonic creation and destruction operators, a† and a. Consider a linear combi-
nation,
b = αa + βa†
What is the constraint on the complex numbers α and β if one is to demand that [b, b† ] =
1?
2. Consider two oscillator levels described by the creation operators, a†1 and a†2 , where the
Hamiltonian is
H = ϵ1 a†1 a1 + ϵ2 a†2 a2 + β(a†1 a†2 + a1 a2 ).
Consider the operators

b†1 ≡ cosh η a†1 + sinh η a2 ,

b†2 ≡ cosh η a†2 + sinh η a1 .

(a) Show that bi and b†i behave like creation/destruction operators.

(b) Find the values of η, E0 , E1 and E2 that allow H to be written as

H = E0 + E1 b†1 b1 + E2 b†2 b2 .

This is known as a Bogoliubov transformation.

3. Consider the coherent state |η⟩ defined by,
∗ η/2
|η⟩ = e−η exp (ηa† )|0⟩.

(a) Show that |η⟩ can also be written as

∗ a+ηa†
|η⟩ = e−η |0⟩.

Hint: You may wish to use the Baker-Campbell-Hausdorff lemma.

282
PHY 851/852 14 STATES WITHOUT CONSERVED PARTICLE NUMBER

(b) Show that the overlap of two states is given by,

′ 2 /2−|η|2 /2+η ′∗ η
⟨η ′ |η⟩ = e−|η | .

4. Consider a coherent state

∗ η/2 †
|η⟩ = e−η eηa |0⟩.

(a) Show that N̄ = ⟨η|Nop |η⟩ = η ∗ η, where Nop = a† a is the number operator.
(b) Show that the variance equals the mean, i.e.,

⟨η|(Nop − N̄ )2 |η⟩ = N̄ .

This is characteristic of a Poissonian distribution.

283