Anda di halaman 1dari 230

Lecture notes on Quantum Physics I, II (Phys.

591, 592)

Version 2.19

Kirill Tuchin1
1
Department of Physics and Astronomy, Iowa State University, Ames, IA 50011
(Dated: April 18, 2016)

Contents

I. Basic principles of Quantum Mechanics 1


§1. The uncertainty principle 1
§2. Wave function 2
§3. Wave function of a free particle 2
A. Wave packet 2
B. Time evolution of wave packet 4
C. Particle in a box 4
§4. Operators 5
A. Expectation values of coordinate and momentum 5
B. Hermitian operators 6
§5. Eigenfunctions and eigenvalues of operators 8
A. Properties of discrete spectrum 9
B. Properties of continuous spectrum 11
§6. Examples of eigenvalue problems 12
A. Operator of momentum p̂. 12
B. Operator of position r̂. 12
C. Operator of orbital angular momentum L̂. 13
§7. Description of quantum states 14
§8. Heisenberg relations 15
§9. Classical limit of quantum mechanics 16

II. Schrödinger equation 18


§1. Hamiltonian 18
§2. Stationary states 19
§3. Equations of motion in operator form 21
A. Time derivative of operators 21
B. Ehrenfest theorem 21
C. Virial theorem 22
§4. Schrödinger and Heisenberg pictures of time-evolution 23
A. Schrödinger picture 23
B. Heisenberg picture 24
C. Interaction picture 25
§5. Symmetries and conserved quantities 26
A. Translations 26
B. Rotations 26
C. Time evolution 28
D. Space inversion 28

III. Motion in one-dimension 29


§1. General properties of one-dimensional motion 29
§2. Delta-potential 31
A. Continuity of the wave function 31
B. Discrete spectrum 32
C. Continuous spectrum 32
§3. Rectangular potential well 33
A. Discrete spectrum 33
B. Continuous spectrum 35
§4. Infinite one-dimensional crystal 36
§5. Harmonic oscillator 39
A. Energy spectrum 39
B. Ladder operators 41
§6. Integral form of Schrödinger equation 42
A. Discrete spectrum 43
B. Continuous spectrum 44

IV. Representation theory 45


§1. Representations of states 45
A. Hilbert space 45
B. Transformation between a continuous and a discrete representations 46
C. Transformation between two continuous representations 46
§2. Representations of operators 47
A. Discrete representation 47
B. Continuous representation 48
§3. Eigenvalue problem in discrete represenation 50
§4. Unitary operators 51
§5. Schrödinger equation in momentum representation 53
§6. Occupation number representation of harmonic oscillator 55

V. Motion in central potential 59


§1. General properties of motion in central potential 59
§2. Spherical waves 61
§3. Spherical potential well 62
§4. Spherical harmonic oscillator 63
§5. Coulomb potential 64
A. Discrete spectrum 64
B. Continuous spectrum 66
§6. Effective electric potential of the hydrogen atom 67
§7. Conservation laws and degeneracy of energy spectrum 68

VI. Angular momentum 70


§1. Angular momentum operator 70
§2. Spin 74
§3. Addition of angular momenta 76

ii
§4. Matrix elements of vector operators 79

VII. Stationary perturbation theory 81


§1. Approximations for a discrete non-degenerate spectrum 81
§2. Approximations for a discrete degenerate spectrum 84
§3. Spectrum with two close levels 87
§4. Variational method 89

VIII. Motion in magnetic field 92


§1. Schrödinger equation in magnetic field 92
§2. Motion in constant magnetic field 93
A. Landau levels 93
B. Equations of motion 94
C. Degeneracy 95
§3. Time dependence of states 96

IX. Time-dependent perturbation theory 99


§1. Transition probability 99
§2. Perturbations finite at t → ∞ 101
§3. Periodic perturbations 101
§4. Quasi-stationary states 103
§5. Sudden interactions 104
A. Weak perturbations 104
B. Strong interactions 104
§6. Adiabatic perturbations 105
§7. Born-Oppenheimer approximation 109
§8. Berry phase 110
§9. Aharonov-Bohm effect 114

X. WKB method 117


§1. The classical limit of quantum mechanics 117
§2. The wave function in the quasi-classical approximation 118
§3. Bohr-Sommerfeld quantization rules 119
§4. Tunneling through potential barriers 123
§5. Double-well potential 127
§6. Quasiclassical approximation in central field 129
§7. Quasiclassical approximation for Coulomb potential 130

XI. Identical particles 133


§1. Spin-statistics theorem 133
§2. Second quantization of bosons 134
§3. Second quantization of fermions 136
§4. Supersymmetric harmonic oscillator 139
§5. Field operators ψ̂ 141

XII. Atoms 143


§1. Ground state of helium-like atoms 143
§2. Lowest excited states of helium-like atoms 145

iii
§3. Hartree-Fock method 146
§4. Thomas-Fermi model 148
§5. Atom in magnetic field 150
A. Weak field. Zeeman effect. 151
B. Strong field. Paschen-Back effect. 152
§6. Atom in electric field 152
A. Atoms other than hydrogen-like 152
B. Hydrogen-like atoms 153
C. Strong fields 154

XIII. Elastic scattering 155


§1. Scattering amplitude 155
§2. Free-particle Green’s function 156
§3. Born approximation 158
A. Conditions of applicability of the Born approximation 159
§4. Partial waves 162
§5. Properties of phase shifts 164
§6. Scattering at low energies 167
§7. Scattering at high energies 170
A. Wave function 170
B. Scattering amplitude 171
§8. Elastic scattering in Coulomb potential 174
§9. Exchange effects for identical spin-0 particles 176
§10. Exchange effects for identical particles of arbitrary spin 177
A. s = 1/2 177
B. Any s 178
§11. Quasi-elastic scattering of electron by atom 179
A. No exchange effects 179
B. Exchange effects 183
§12. Scattering matrix 184

XIV. Relativistic scalar particles 187


§1. Uncertainty principle in relativistic theory 187
§2. Klein-Gordon equation 187
§3. Non-relativistic limit of Klein-Gordon equation 189
§4. Scalar particle states with definite momentum 190
§5. Second quantization of scalar particles 191
§6. Scalar particle in electromagnetic field 192
§7. Scalar particle in Coulomb field 194
§8. Tunneling through potential barrier 195

XV. Relativistic spin-1/2 particles 198


§1. Dirac equation 198
§2. Spin and the Dirac equation 200
§3. Covariant form of the Dirac equation 200
§4. States of spin-1/2 particle with definite momentum 201
§5. Second quantization of spin-1/2 particles 204
§6. Spin and helicity states 205

iv
§7. Charged spin-1/2 particle in electromagnetic field 206
§8. Non-relativistic limit of Dirac equation 207
A. Free particle 207
B. Particle in external field 208
§9. Relativistic corrections to motion in spherically symmetric potential 211
§10. Relativistic corrections in hydrogen-like atom 213

XVI. Quantization of electromagnetic field 215


§1. Second quantization of E&M field 215
§2. Photon radiation by an atom 216

A. Fourier analysis 219

B. Dirac delta function 219

C. Levi-Civita symbol 220

D. Legendre polynomials 221

E. Bessel functions 223

F. List of books used in preparation of these notes 224

v
§1 The uncertainty principle I BASIC PRINCIPLES OF QUANTUM MECHANICS

I. BASIC PRINCIPLES OF QUANTUM MECHANICS

§1. The uncertainty principle

Quantum mechanics considers microscopic systems that are smaller than 10−6 cm. Physical laws that govern such
systems are very different form those that govern macroscopic systems, which are described by the classical mechanics.
Moreover, the microscopic phenomena can be observed only by means of a macroscopic apparatus that translates the
action of microscopic objects into the macroscopic language (examples: Geiger counter, bubble chamber). Quantum
mechanics is a coherent mathematical framework that describes the laws of microscopic systems and translates them
into the classical language. The ultimate success of quantum mechanics is evident in a great number of experiments.
However, its interpretation is still being debated. The goal of this course is to develop the mathematical framework
of quantum mechanics and to illustrate how it can be applied to number of important problems.
Historically, the first indications that the classical theory is not adequate for description of microscopic systems
came in the early 20th century. It was experimentally established that electromagnetic radiation possesses both wave
and corpuscular character. In particular, it is absorbed and emitted in separate portions, quanta, which we now call
photons. Photon energy E turned out to be proportional to its frequency ω as
E = ~ω , (1.1)
−27
where ~ = 1.054 · 10 erg·sec is Planck’s constant. In free space photons move with the velocity of light c and have
the momentum
p = ~k , (1.2)
where k is the wave-vector with the length k = |k| = 2π/λ = 1/λ. We know from the electromagnetic theory that
ω = kc. It follows that p = ~ω/c = E/c, which is the relation between the energy and momentum of a relativistic
particle of zero mass.
Similarly to the electromagnetic radiation, matter particles also posses both wave and corpuscular character. This
was most clearly seen in the electron diffraction experiment where a homogeneous beam of electrons passes through a
crystal. The emergent beam exhibits a pattern of alternate maxima and minima of intensity similar to the diffraction
of electromagnetic waves. In the two-slit experiment one considers two screens: one impermeable to electrons in which
two slits are cut and behind it another continuous screen. When electrons pass through the first slit, while the second
one is covered, one observes a certain distribution of intensity on the screen. A different distribution is obtained
when the first slit is opened and the second is covered. However, the intensity distribution that is obtained when
both slits are open is different from a simple sum of the two intensity distribution in contrast to expectation from the
classical physics. Using the diffractive experiments one can assign to a free particle a wave length λ that is uniquely
determined by its momentum
p 2π
k= , k= , (1.3)
~ λ
in analogy with the photon. In this context λ is called the de Broglie wavelength.
An important consequence of the electron diffraction experiment is that the classical notion of trajectory does not
apply to microscopic objects. Had electron beams moved along a certain trajectory they would have not interfered.
This absence of a trajectory is the essence of the uncertainty principle. Later in this course we will give a mathematical
formulation of this fundamental principle.
Suppose now that we measure the electron position with an apparatus (a classical devise). The more accurate is the
measurement, the stronger apparatus affects the electron, so that consecutive measurements would give a discontinuous
and disorderly results indicating absence of trajectory. Only at very low accuracy (e.g. in bubble chamber), the results
can be approximated by a smooth curve – classical electron trajectory. As a result, the classical definition of velocity
as a time derivative of the particle position is meaningless in quantum theory. Later we will construct a reasonable
definition of velocity in quantum mechanics that has correct classical limit.
In classical mechanics a full description of a system is achieved by providing the coordinates and velocities of all
particles at a given time. The uncertainty principle implies that such description is impossible in quantum mechanics.
Quantum description is less detailed than the classical one, and its predictions are less certain. For instance, given
an electron in a certain initial state, a subsequent measurement can yield various results. The problem in quantum
mechanics consists in determining the probability of obtaining of these different results in a measurement.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §1; Messiah, “Quantum
Mechanics”, Ch. I; Merzbacher, “Quantum Mechanics” 3rd edition, Ch. 1.

1
§3 Wave function of a free particle I BASIC PRINCIPLES OF QUANTUM MECHANICS

§2. Wave function

The properties of microscopic systems are described in quantum mechanics by means of the wave function ψ. It is an
auxiliary quantity which is not directly observed in experiment. The wave function is a complex continuous function
of coordinates and time. According to the statistical or Copenhagen interpretation the wave function determines the
probabilities of a different outcomes of a measurement. Let q denote the set of coordinates of a quantum system
(often referred to as the configuration space) and dq the product of differentials of these coordinates (i.e. element of
volume of the configuration space). For example, for one particle q = (x, y, z) and dq = dx dy dz, for two particles
q = (x1 , y1 , z1 , x2 , y2 , z2 ) and dq = dx1 dy1 dz1 dx2 dy2 dz2 , etc. According to the statistical interpretation, quantity
|ψ(q)|2 dq = ψ(q)ψ ∗ (q)dq (1.4)
is proportional to the probability that we find as the result of our measurement that values of coordinates lie within
the interval [q, q + dq]. The sum of the probabilities of all possible value of the coordinates of the system must be
equal to unity implying that
Z
|ψ(q)|2 dq = 1 . (1.5)

This equation is called the normalization condition. If a wave function is normalized according to (1.5), then ρ = |ψ(q)|2
is known in mathematics as the probability density. In some idealized systems integral over the configuration space
does not converge, viz.
Z
|ψ(q)|2 dq = ∞ . (1.6)

In this case |ψ(q)|2 cannot be interpreted as the probability density. Instead, it is meaningful to talk about the relative
probabilities at different q’s.
Wave function satisfies the following fundamental superposition principle: if a system can be in states described by
wave functions ψ1 and ψ2 , it can also be in all states described by the linear combination a1 ψ1 + a2 ψ2 , for any time-
independent complex numbers a1 and a2 . In particular, in view of the normalization condition, wave functions ψ and
aψ, with non-zero a, describe the same state. In view of (1.5) we can write a = eiα with any real α. This ambiguity
does not affect any of the physical results, because |ψ(q)|2 does not depend on it. Note that the superposition
principle in quantum mechanics is essentially different from the superposition of vibrations in classical physics, where
a superposition of a vibration onto itself gives a new vibration with a larger or smaller amplitude.
Apart of its dependence on coordinates, the wave function also depends on time. Its time-dependence, or evolution,
is determined by a differential equation that we will discuss in the next chapter. That equation must be linear so that
its solutions satisfy the superposition principle.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §2, Merzbacher, “Quan-
tum Mechanics” 3rd edition, Ch. 1; Messiah, “Quantum Mechanics”, Ch. IV, §15-19, Ch. V, §1,2.

§3. Wave function of a free particle

A. Wave packet

It is instructive to consider a simple example before we start developing mathematical framework of quantum
mechanics. Consider a free non-relativistic particle of mass m. Its energy E = p2 /2m and momentum p are conserved,
which reflects uniformity of space and time (will discuss this in more detail later). Therefore, its wave function must
depend on E and p. According to our discussion in I §1 this particle behaves in diffractive experiments as a wave with
wave-vector k = p/~ and energy ω = E/~. In optics, a wave with given k and ω is a plane wave
ψ(r, t) = N ei(k·r−ωt) , (1.7)
where N is a constant. Therefore, it is reasonable to assume that (1.7) describes motion of a free particle. This
argument is apparently an educated guess, which cuts many corners, but leads to the right answer as will be shown
later. A state of a free particle of mass m, momentum p0 and energy E 0 is described by
0 0
ψ 0 (r, t) = N ei(k ·r−ω t) . (1.8)

2
§3 Wave function of a free particle I BASIC PRINCIPLES OF QUANTUM MECHANICS

According to the superposition principle Ψ = aψ + a0 ψ 0 also describes a possible state of a particle. In this state
however, the particle does not have a definite momentum.
Generally, one can consider a linear combination of any number of states of form (1.7) with different momenta.
Suppose for simplicity that particle moves along the z-axis. Then its state is described by the following wave function
Z +∞
ψ(z, t) = ak ei(kz−ωt) dk (1.9)
−∞

with an arbitrary ak (This notation means that a is a function of k). In practice, particle momentum, say k0 , is
known with a certain accuracy ∆k, where ∆k  k0 (otherwise we cannot say that particle momentum is known). In
this case, (1.9) is called the wave packet and can be written down as
Z k0 +∆k
ψ(z, t) = ak ei(kz−ωt) dk . (1.10)
k0 −∆k

To estimate this integral, introduce a new variable l = k − k0 , and expand ω as a function of k near k = k0 :
 

ω(k) ≈ ω0 + l, (1.11)
dk 0

where ω0 = ω(k0 ). In this approximation integration in (1.10) produces


Z +∆k
ak0 ei(l+k0 )z−iω0 t−i( dk )0 lt dl

ψ(z, t) =
−∆k
Z +∆k
ei[z−( dk )0 t]l dl

i(k0 z−ω0 t)
= ak0 e
−∆k
  
sin z − dω dk 0 t ∆k i(k0 z−ω0 t)
= 2ak0  |e {z } (1.12)
z − dω t
| {z dk 0 } fast-oscillating
“amplitude”

The wave packet is a product of two factors: slowly changing “amplitude” and rapidly oscillating plane wave. At
t = 0 the wave packet has global maximum at z = 0 and then decreases as 1/z while oscillating with period 2π/∆k.
Zeros of the amplitude are located at zn = πn/∆k, where n = ±1, ±2, . . .. The spatial extent of the packet can be
estimated as a distance between the two zeros around the maximum at z = 0:

∆z = . (1.13)
∆k
In view of (1.3) we conclude that

∆z∆p = 2π~ . (1.14)

In other words, uncertainty in the particle position is inversely proportional to the uncertainty of its momentum. This
is consistent with the uncertainty principle that a microscopic particle does not have a trajectory.
According to (1.12) the maximum of the wave packet moves along the z-axis with group velocity
 

vg = . (1.15)
dk 0

Since ω = p2 /2m~ = ~k 2 /2m we find


p0
vg = , (1.16)
m
as it should be for a free particle. Note, that the phase velocity vp = ω/k = E/p = p/2m is not velocity of particle.

3
§3 Wave function of a free particle I BASIC PRINCIPLES OF QUANTUM MECHANICS

B. Time evolution of wave packet

Consider a wave packet that at t = 0 is given by


z 2
− 2ρ
ψ(z, 0) = N e 2
. (1.17)
Here ρ is a parameter that specifies the width of the Gaussian (1.17). Using (1.5) we find the normalization constant
1 z
− 2ρ
2

ψ(z, 0) = √ e 2
. (1.18)
π 1/4 ρ
Using (1.9) we can write
Z +∞
1 z
− 2ρ
2

ak eikz dk = √ e 2
. (1.19)
−∞ π 1/4 ρ
Inverting the Fourier transform (see App. A) we get
Z +∞ √
dz −ikz 1 z2
− 2ρ ρ k 2 ρ2
ak = e 1/4 √ e 2
= √ e− 2 . (1.20)
−∞ 2π π ρ 2π 3/4
Employing (1.9) again and recalling that ω = ~k 2 /2m we have
√ Z +∞
ρ k 2 ρ2 1 − z2
ψ(z, t) = √ ei(kz−ωt) e− 2 dk = p e 2(ρ2 +i~t/m) . (1.21)
2π 3/4 −∞ π 1/4 ρ + i~t/mρ
The probability density reads
mρ − z2
|ψ(z, t)|2 = √ p e ρ2 +~2 t2 /m2 ρ2 . (1.22)
π m2 ρ4 + ~2 t2
This shows that the width of the wave packet grows as a function of time, so that the particle position becomes less
certain at later time. This can be easily understood as follows. At t = 0 the width of the packet is ∆z ∼ ρ and
according to (1.14) ∆p ∼ ~/ρ. Components of the wave packet with different k move with different velocities, the
uncertainty in velocity being ∆v = ∆p/m ∼ ~/mρ. Thus, the packet width grows as ∆z(t) ∼ ~t/mρ in agreement
with (1.22). This effect is important at t > t0 = mρ2 /~, whereas at t  t0 the diffusion of the wave packet can
be neglected. To put this into a perspective, take ρ = 10−8 cm as a typical atomic scale. Then for electron we get
t0 = 10−16 sec. In contrast, for m = 1 gr, t0 = 104 years.

C. Particle in a box

Wave function of a free particle (1.7) is an example of a wave function that cannot be normalized with condition
(1.5), but rather its normalization integral is divergent as in (1.6). This divergence appears because the notion of free
particle is an idealization. In reality, we can only talk about a particle that can be considered approximately free in
a sufficiently large box, beyond which its interactions cannot be neglected. So, consider a free particle confined in a
cubic box of edge L  10−6 cm. On the surface of the box ψ must satisfy certain boundary conditions. Since L is
much larger than the particle de Broglie wavelength, these boundary conditions have very little effect on the particle
motion. Therefore, let’s set the periodic boundary conditions because they are the simplest to deal with:
ψ(x, y, z) = ψ(x + L, y, z) = ψ(x, y + L, z) = ψ(x, y, z + L) . (1.23)
A wave function that has the same form as (1.7) and satisfies the boundary condition (1.23) is (at t = 0)
1 2πnx 2πny 2πnz
ψk (r) = √ eik·r , kx = , ky = , kz = , (1.24)
L3 L L L
where nx ,ny ,nz are non-zero integers. This wave function satisfies the following relation
Z
ψk∗ 0 (r)ψk (r)d3 r = δk0 k , (1.25)

4
§4 Operators I BASIC PRINCIPLES OF QUANTUM MECHANICS

where the integration extends to the entire volume(1 . Indeed, if k 6= k0 the integral vanishes due to the periodic
boundary condition. Since momentum of a free particle is conserved, each function (1.24) corresponds to motion with
definite momentum p = ~k.
Consider now a particle in a box in an arbitrary state (i.e. not necessarily free) described by a wave function ψ(r).
We know from the Fourier analysis (see Appendix A) that the set of wave functions (1.24) is a complete set. Namely,
ψ(r) can be expanded as follows
X
ψ(r) = ak ψk (r) . (1.26)
k

This formula tells us that if a particle is in a state ψ(r) and we measure its momentum, the result can be any number
p = ~k from a discrete infinite set specified in (1.24). Multiplying this formula by ψk∗ 0 (r) on both sides, integrating
over the volume of the box and using (1.25) yields the Fourier coefficients
Z
ak = ψ(r)ψk∗ (r)d3 r . (1.27)

Sometimes coefficients ak ’s are called amplitudes, a term borrowed from optics. Notice, that it follows from (1.5),(1.26)
and (1.25) that
Z Z X X
1 = ψ ∗ (r)ψ(r)d3 r = ak a∗k0 ψk (r)ψk∗ 0 (r)d3 r = |ak |2 . (1.28)
k0 ,k k

It seems reasonable to interpret |ak |2 as the probability that measurement of momentum of a particle in state ψ(r)
gives p = ~k. Taking L → ∞ corresponds to the free space limit.

A simple example developed in this section illustrates many important features of the general theory. I will use it
as a reference in the forthcoming sections.

• Additional reading: Messiah, “Quantum Mechanics”, Ch. II, §1-3; Merzbacher, “Quantum Mechanics” 3rd edition,
Ch. 2.

§4. Operators

A. Expectation values of coordinate and momentum

Suppose now that a particle is in a state described by ψ(r, t). Average (or expectation) value of the position vector
is
Z
hri = ψ ∗ (r, t)rψ(r, t)d3 r . (1.29)

If F (r) is a function of r, than


Z
hF (r)i = ψ ∗ (r, t)F (r)ψ(r, t)d3 r . (1.30)

To determine the expectation value of momentum, we place the particle into a big box. We can always expand the
wave function into a complete set of states with definite momentum, see (1.26). The probability to find the value
of particle momentum p = ~k is given by |ak |2 , where ak is given by (1.27). Therefore, according to the statistical
interpretation,
X XZ Z

hpi = ~ ak ak k = ~ ψ (r)ψk (r)d rk ψ(r 0 )ψk∗ (r 0 )d3 r0
∗ 3
(1.31)
k k

(1 Kronecker delta δab is defined to be equal unity when a = b and zero otherwise.

5
§4 Operators I BASIC PRINCIPLES OF QUANTUM MECHANICS

It follows from (1.24) that i∇ψk∗ (r) = kψk∗ (r), so that (1.31) can be written as
XZ Z
hpi = i~ ψ ∗ (r)ψk (r)d3 r ψ(r 0 )∇ψk∗ (r 0 )d3 r0 (1.32)
k

Integrating by parts and using the periodic boundary conditions (1.23) implies
Z Z Z
ψ(r 0 )∇ψk∗ (r 0 )d3 r0 = ∇[ψ(r 0 )ψk∗ (r 0 )]d3 r0 − ∇ψ(r 0 )ψk∗ (r 0 )d3 r0 . (1.33)
| {z }
=0

Substituting (1.33) into (1.32) and employing the completeness relation (see home assignments)
X
ψk∗ (r)ψk (r 0 ) = δ(r − r 0 ) (1.34)
k

we obtain
Z
hpi = ψ ∗ (r)(−i~∇)ψ(r)d3 r . (1.35)

Taking limit L → ∞ we derive that (1.35) is valid also for a particle in entire space. Generalization of (1.35) for any
rational function F (p) is straightforward:
Z
hF (p)i = ψ ∗ (r)F (−i~∇)ψ(r)d3 r . (1.36)

For example, expectation value of kinetic energy is


 2 Z  2 2
p ~ ∇
= ψ ∗ (r) − ψ(r)d3 r . (1.37)
2m 2m

We can now generalize (1.30) and (1.37) as follows. Let F be a function of r and p such that

F (r, p) = F1 (r) + F2 (p) . (1.38)

Expectation value of F can be written in a symbolic form


Z
hF i = ψ ∗ (r)F̂ ψ(r)d3 r , (1.39)

where

F̂ = F1 (r) + F2 (−i~∇) (1.40)

is a differential operator.

B. Hermitian operators

Eq. (1.40) is an example of an operator corresponding physical quantity F . We can can further generalize this
definition to include any physical quantity, not necessarily depending on r and/or p. (Quantities that have no
classical analogue cannot be expressed as a function of r and p, e.g. spin). Namely, an operator F̂ can be put into
correspondence with any physical quantity F .(2

(2 In this context F is sometimes referred to as a c-number (c for “classical”) to distinguish it from the corresponding operator F̂ .

6
§4 Operators I BASIC PRINCIPLES OF QUANTUM MECHANICS

By convention, F̂ ψ means that operator F̂ acts on function ψ. Notation F̂ ψϕ, or equivalently F̂ (ψϕ), means that
F̂ acts on the product of ψ and ϕ, whereas (F̂ ψ)ϕ means that F̂ acts on ψ and then the result is multiplied by ϕ.
Most operators in quantum mechanics are linear, which means that

F̂ (ψ1 + ψ2 ) = F̂ ψ1 + F̂ ψ2 , F̂ (aψ) = aF̂ ψ , (1.41)

for any functions ψ1 , ψ2 and for any constant a.



Physical quantities are real, hence hF i = hF i . In view of (1.39) this means that
Z Z
ψ (q)F̂ ψ(q)dq = ψ(q)F̂ ∗ ψ ∗ (q)dq .

(1.42)

In linear algebra operator F̂ † is called adjoint of Hermitian conjugate of F̂ if


Z Z
ψ (q)F̂ ϕ(q)dq = ϕ(q)(F̂ † ψ(q))∗ dq .

(1.43)

Linear operator is self-adjoint or Hermitian if


Z Z
ψ (q)F̂ ϕ(q)dq = ϕ(q)(F̂ ψ(q))∗ dq .

(1.44)

A short-hand operator notation of (1.44) is

F̂ = F̂ † . (1.45)

Clearly (1.42) is a particular case of (1.44) with ϕ = ψ. Thus we can state that to any physical quantity there
corresponds a linear Hermitian operator.
Define the product F̂ K̂ of two operators F̂ and K̂ as a successive action of operator K̂ and then of operator F̂ .
Note, that generally, F̂ K̂ψ 6= K̂ F̂ ψ; in operator notation: F̂ K̂ 6= K̂ F̂ . Even though F̂ and K̂ represent physical
quantities, their product does not necessarily do so. We are interested to find a condition under which F̂ K̂ may also
represent a physical quantity. In mathematical language, given Hermitian operators F̂ and K̂, when is the operator
L̂ = F̂ K̂ also Hermitian? To answer this question use (1.44) twice in the following integral
Z Z Z Z Z
ψ ∗ L̂ϕdq = ψ ∗ F̂ (K̂ϕ)dq = (K̂ϕ)F̂ ∗ ψ ∗ dq = (F̂ ∗ ψ ∗ )K̂ϕdq = ϕ(K̂ F̂ ψ)∗ dq . (1.46)

On the other hand, according to (1.43)


Z Z

ψ L̂ϕdq = ϕ(L̂† ψ)∗ dq . (1.47)

Comparing (1.47) with (1.46) we can write down an operator equation

L̂† = (F̂ K̂)† = K̂ F̂ . (1.48)

We see that operator F̂ K̂ is Hermitian iff operators F̂ and K̂ commute with one another, viz.

F̂ K̂ = K̂ F̂ . (1.49)

Commutator of two operators is defined as

[F̂ , K̂] = F̂ K̂ − K̂ F̂ . (1.50)

With this notation (1.49) reads [F̂ , K̂] = 0.


Out of any two Hermitian operators F̂ and K̂ we can always construct Hermitian operators Ŝ = K̂ F̂ + F̂ K̂ and
Ĝ = i(K̂ F̂ − F̂ K̂) (see homework assignments). In particular, [F̂ , K̂] = iĜ.

∗ Examples.

7
§5 Eigenfunctions and eigenvalues of operators I BASIC PRINCIPLES OF QUANTUM MECHANICS

1. It is convenient to denote components of vector A by Ai , where i = 1, 2, 3. For example, components of r are


ri such that r1 = x, r2 = y, r3 = z; components of p are pi such that p1 = px , p2 = py , p3 = pz . We have the
following commutators

[r̂i , r̂k ] = 0 , [p̂i , p̂k ] = 0 , (1.51)

where p̂i = −i~ ∂r∂ i ≡ −i~∂i . From the following identities


 
∂ψ
x̂p̂x ψ(x) = x̂ −i~ (1.52)
∂x
∂(xψ) ∂ψ
p̂x x̂ψ(x) = −i~ = −i~ − i~ψ (1.53)
∂x ∂x
we derive

[x̂, p̂x ]ψ = i~ψ ⇒ [x̂, p̂x ] = i~ . (1.54)

In general,

[r̂i , p̂k ] = i~δik . (1.55)

2. Orbital angular momentum of a particle is L = r × p. The corresponding operator:

L̂ = −i~(r × ∇) . (1.56)

It is Hermitian as can be seen from (1.55) employing a theorem that we proved in the previous section. Cross
products can be written using the Levi-Civita symbol ijk defined in Appendix C: Li = ijk rj pk , where summa-
tion over the repeated indices is implied. One can prove that (homework assignment)

[L̂i , L̂k ] = i~ikl L̂l , (1.57)


2
[L̂ , L̂i ] = 0 , (1.58)
[L̂i , r̂k ] = i~ikl r̂l , (1.59)
[L̂i , p̂k ] = i~ikl p̂l . (1.60)

• Additional reading: Merzbacher, “Quantum Mechanics” 3rd edition, 4.1–1.4, 3.4, 11.3–11.5, Landau and Lifshitz,
“Quantum Mechanics: Non-relativistic theory”, §3,4.

§5. Eigenfunctions and eigenvalues of operators

As we have seen in I §3 that a particle can be in a state with definite value of momentum p = ~k. In general,
a system can be in a state with definite value of a certain physical quantity F . In such a state variation from the
average hF i must vanish. A measure of such variation is variance, defined as the expectation value of the square of
d = F̂ − hF i. Namely,
operator ∆F
Z Z  ∗ Z  ∗ Z 2


d ψdq = ∆F d ψ dq ,

d ∆F
(∆F )2 = ψ ∗ ∆F d ψdq = d ψ ∆F
∆F d ψdq = d ψ ∆F
∆F (1.61)

d ψ = 0. Since in such
where I used that fact the operator F̂ is Hermitian. Variance vanishes for states satisfying ∆F
states hF i of course coincides with F we can write this condition as

F̂ ψ = F ψ . (1.62)

Eq. (1.62) is usually supplemented by a boundary condition on ψ and is called the eigenvalue problem for operator F
with a particular boundary condition. In mathematics it is a subject of the Sturm-Liouville theory. One of the results
of this theory is that the eigenvalue problem has solutions only at certain discrete values of parameter F . These

8
§5 Eigenfunctions and eigenvalues of operators I BASIC PRINCIPLES OF QUANTUM MECHANICS

particular values are called the eigenvalues of the operator F̂ and the corresponding solutions the eigenfunctions. A
set of all eigenvalues is called its spectrum. I will denote eigenfunctions corresponding to the eigenvalue F as ψF .
For example, functions (1.24) are eigenfunctions of operator of momentum corresponding to momentum p = ~k. As
we have already seen before, it is often convenient to consider unbounded systems, in which case the spectrum is
continuous.
The physical meaning of the spectrum is that measurement of physical quantity F can yield only it eigenvalues.
If a system is in a state described by one of the eigenfunctions ψF , than the result of the measurement definitely
produces the corresponding eigenvalue F . Sometimes, several linearly independent eigenvalues will correspond to one
eigenvalue of operator. The number of eigenvalues is called the degeneracy of the eigenvalue. An eigenvalue having
degeneracy one is called non-degenerate.
Since F̂ corresponds to a physical quantity, its eigenvalues must be real numbers. To prove that this is indeed the
case, write F̂ ψ = F ψ and its complex conjugate F̂ ∗ ψ ∗ = F ∗ ψ ∗ and consider the following integrals
Z Z
(ψ F ψ − ψF ψ )dq = (ψ ∗ F̂ ψ − ψ F̂ ∗ ψ ∗ )dq .
∗ ∗ ∗
(1.63)

Since F̂ is Hermitian we can write this as


Z Z
(F − F ) |ψ| = (ψ F̂ ∗ ψ ∗ − ψ F̂ ∗ ψ ∗ )dq = 0
∗ 2
⇒ F − F∗ = 0. (1.64)

Now let us discuss properties of discrete and continuous spectra.

• Additional reading: Arfken et. al. “Mathematical methods for physicists”, 7th edition, Ch.8.

A. Properties of discrete spectrum

1. Consider a Hermitian operator F̂ with a non-degenerate discrete spectrum Fn , n = 1, 2, . . . and the corresponding
set of eigenfunctions ψn . According to the definition of a Hermitian operator (1.44)
Z Z

ψm F̂ ψn dq − ψn F̂ ∗ ψm

dq = 0 . (1.65)

On the other hand, F̂ ψn = Fn ψn and F̂ ∗ ψm∗


= Fm ψm∗
because Fm is real. It follows that
Z

(Fn − Fm ) ψm ψn dq = 0 . (1.66)

If m 6= n we arrive at the orthogonality property of the wave functions


Z

ψm ψn dq = 0 , m 6= n . (1.67)

It’s physical meaning is that measurement of F gives definitely either Fn in state ψn or Fm in state ψm for any
m 6= n. In other words, measurement of F cannot result in a system being in a superposition of eigenfunctions.
2. As I mentioned above (see discussion after (1.62)), an eigenvalue problem has a discrete spectrum
R if it is restricted
to a finite domain. In this case wave functions ψn rapidly decrease at q → ∞ implying that |ψn |2 dq is finite.
Thus, we can normalize these wave functions to unity. Together with (1.67) we obtain the orthonormality
property
Z

ψm ψn dq = δmn . (1.68)

3. An important result of the Sturm-Liouville


R theory is that eigenfunctions ψn , n = 1, 2, . . . , form a complete set.
Namely, any function ψ(q) such that |ψ(q)|2 dq is finite can be expanded as follows
X
ψ(q) = an ψn (q) . (1.69)
n

9
§5 Eigenfunctions and eigenvalues of operators I BASIC PRINCIPLES OF QUANTUM MECHANICS

Using the orthonormality property we can calculate the coefficients an :


Z
an = ψ(q)ψn∗ (q)dq . (1.70)

These equations generalize (1.25),(1.26),(1.27) for a particle in a box.


4. Eigenfunctions satisfy the completeness relation that reads (cp. (1.34))
X
ψn∗ (q 0 )ψn (q) = δ(q − q 0 ) . (1.71)
n

Indeed, substituting (1.70) into (1.69) we can write


Z X
ψ(q) = dq 0 ψ(q 0 ) ψn (q)ψn∗ (q 0 ) , (1.72)
n

which is self-consistent only if (1.71) holds.


5. Suppose now that Fn is N -fold degenerate, i.e. functions ψnl with l = 1, . . . , N are eigenfunctions corresponding
to the same eigenvalue Fn . Eigenfunctions with different n are still orthogonal, which can be proved as in (1.65)–
(1.67). However, eigenfunctions with the same n, but different l are not necessarily so. Since it is convenient to
deal with a set of orthogonal functions we can replace a set of N independent functions ψnl , l = 1, . . . , N , by
another set of functions which are eigenfunctions of F̂ and orthogonal. Let me show how this can be done in
the simplest case N = 2.
We start with two normalized eigenfunctions ψn1 and ψn2 . Define two new functions ϕn1 = ψn1 and ϕn2 =
a(ψn1 + λψn2 ). These two are also eigenfunctions of F̂ . The idea is two choose λ such that
Z Z Z
∗ ∗ −1 ∗
ϕn1 ϕn2 dq = 0 ⇒ a ψn1 (ψn1 + λψn2 )dq = 0 ⇒ λ = − ψn1 ψn2 dq , (1.73)

and choose a such that


Z
ϕ∗n2 ϕn2 dq = 1 . (1.74)

This procedure is called orthogonalization or Gram-Schmidt process. It can be easily generalized to arbitrary
N . The resulting spectrum satisfies the following relations:
Z

ψml (q)ψnk (q)dq = δmn δkl , (1.75)
X Z

ψ(q) = anl ψnl (q) , where anl = ψ(q)ψnl (q)dq , (1.76)
n,l
X

ψnl (q 0 )ψnl (q) = δ(q − q 0 ) . (1.77)
n,l

6. Expectation value of F in state ψ can be expressed through an as follows


Z Z X !∗ Z
X XX

hF i = ψ F̂ ψdq = an ψn F̂ am ψm dq = an am ψn∗ Fm ψm dq

n m m n
XX X
= Fm δmn a∗n am = |an |2 Fn . (1.78)
m n n

In particular, for F = 1 we have


X
|an |2 = 1 . (1.79)
n

Compare this with (1.28). As in I §3 we can interpret |an |2 as probability to find a system in state n.

10
§5 Eigenfunctions and eigenvalues of operators I BASIC PRINCIPLES OF QUANTUM MECHANICS

It is important to note that knowledge of probabilities |an |2 does not completely specify the wave function of a
quantum state. The wave function is not known unless all relative phases between different states ψn are also known.
States for which only probabilities are known are called the mixed states, whereas states with a wave function are
called the pure states. Instead of wave function, the mixed states are described by the density matrix. Such description
is not full even in quantum mechanical sense. Mixed states play in quantum statistical physics the same role as the
wave function in quantum mechanics.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §3; Messiah, “Quantum
Mechanics”, Ch. V, §5-7.

B. Properties of continuous spectrum

1. Consider
R operator F̂ that has continuous spectrum: F̂ ψF = F ψF , where F is continuous. In this case
|ψF |2 dq = ∞ because the system (or its part) can go to infinity. We can derive properties of continuous
spectrum by taking the limit L → ∞ as we did in the case of a particle in box. As a result we obtain that
functions ψF form a complete set, viz. any ψ(q) can be expanded as
Z
ψ(q) = aF ψF (q)dF . (1.80)

If ψ(q) is normalized, then aF must be normalized as well due to (1.79). Namely


Z Z
ψ ∗ (q)ψ(q)dq = a∗F aF dF = 1 . (1.81)

Let us show that this allows us to invert (1.80) similarly to (1.70). Write
Z Z Z Z Z 
∗ ∗ ∗ ∗ ∗
ψ (q)ψ(q)dq = dqψ(q) aF ψF (q)dF = dF aF dqψF ψ(q) (1.82)

and subtract from it (1.81). This gives


Z Z 
0= dF a∗F dqψF∗ ψ(q) − aF , (1.83)

which is true if
Z
aF = ψF∗ ψ(q)dq . (1.84)

2. Substitute (1.80) into (1.84) to find


Z Z
aF = dqψF∗ (q) aF 0 ψF 0 (q)dF 0 . (1.85)

This equation is self-consistent if the following delta-function normalization holds


Z
ψF∗ (q)ψF 0 (q)dq = δ(F − F 0 ) . (1.86)

Similarly, we can start with (1.80), substitute into it (1.84) and obtain that
Z
ψF∗ (q 0 )ψF (q)dF = δ(q 0 − q) . (1.87)

Thus, we learned how to normalize the wave functions of continuous spectrum.

11
§6 Examples of eigenvalue problems I BASIC PRINCIPLES OF QUANTUM MECHANICS

Finally, some operators have both discrete and continuous spectra. In this case the complete set of states includes
eigenfunctions from both parts of the spectrum. Any function ψ(q) can be expanded as
X Z
ψ(q) = an ψn (q)dq + aF ψF (q)dq . (1.88)
n
R 2
Normalization |ψ(q)| dq = 1 implies that
X Z
|an |2 + |aF |2 dF = 1 . (1.89)
n

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §5; Messiah, “Quantum
Mechanics”, Ch. V, §8-11.

§6. Examples of eigenvalue problems

A. Operator of momentum p̂.

1. Eigenvalue problem for the operator p̂x = −i~∂x reads


∂ψ(x)
−i~ = px ψ(x) . (1.90)
∂x
Solution to this equation is
ψpx (x) = N eipx x/~ . (1.91)
This is the eigenfunction of p̂x corresponding to eigenvalue px . Spectrum of this operator is continuous −∞ <
px < ∞. ψpx (x) describes motion of a particle along x-axis with definite momentum px .
The normalization constant N can be fixed using (1.86) and (B11) to be N = 1/(2π~)1/2 . Indeed,
Z Z ∞
1 0
ψp∗x (x)ψp0x (x)dx = ei(px −px )x/~ dx = δ(px − p0x ) . (1.92)
2π~ −∞
2. In three dimensions p̂ = −i~∇. The corresponding eigenvalue problem is solved by eigenfunctions
1
ψp (r) = eip·r/~ (1.93)
(2π~)3/2
corresponding to eigenvalues p. Normalization is similar to the one-dimensional case if one uses (B16).

B. Operator of position r̂.

This is another example of continuous spectrum. The eigenvalue problem reads


r̂ψr0 (r) = r0 ψr0 (r) . (1.94)
Action of r̂ on a function of r is a simple product. If r = r0 , then (1.94) is identity, if r 6= r0 , then (1.94) is satisfied
only if ψr0 = 0. Therefore, the properly normalized solution to (1.94) is
ψr0 (r) = δ(r − r0 ) . (1.95)
We can expand any wave function in complete set of eigenfunctions ψr0 (r) as follows
Z
ψ(r) = ar0 ψr0 (r)d3 r0 , (1.96)

where
Z
ar0 = ψ(r)δ(r − r0 )d3 r = ψ(r0 ) . (1.97)

We observe that |ar |2 = |ψ(r)|2 is the probability density to find a particle at position r. This agrees with the
interpretation we gave in the previous section.

12
§6 Examples of eigenvalue problems I BASIC PRINCIPLES OF QUANTUM MECHANICS

C. Operator of orbital angular momentum L̂.

1. Consider operator of projection of the orbital angular momentum on z-axis L̂z . In Cartesian coordinates it has
form (see (1.56))

L̂z = −i~ (x∂y − y∂x ) . (1.98)

In spherical coordinates

L̂z = −i~∂φ , (1.99)

where 0 ≤ φ ≤ 2π. The eigenvalue problem reads

−i~∂φ ψLz (φ) = Lz ψLz (φ) , (1.100)

with the periodic boundary condition

ψLz (φ) = ψLz (φ + 2π) . (1.101)

Since this eigenvalue problem is restricted to unit circle, its eigenvalues Lz are discrete. Solution to the differential
equation (1.100) is

ψLz (φ) = N eiLz φ/~ . (1.102)

Boundary condition implies that Lz = m~, with any integer m. We can thus label eigenvalues and eigenfunctions
of L̂z by m. Imposing the normalization condition
Z 2π

ψm ψm dφ = 1 (1.103)
0

we finally obtain the orthonormal set of eigenfunctions


1
ψm (φ) = √ eimφ . (1.104)

P3
2. Another operator that we will use extensively in this course is L̂2 = k=1 L̂2k . In spherical coordinates it reads
   
2 2 1 ∂ ∂ 1 ∂2
L̂ = −~ sin θ + . (1.105)
sin θ ∂θ ∂θ sin2 θ ∂φ2

The eigenvalue equation L̂2 ψlm = L2 ψlm becomes


   
2 1 ∂ ∂ 1 ∂2 L2
−~ sin θ + + 2 ψlm (θ, φ) = 0 . (1.106)
sin θ ∂θ ∂θ sin2 θ ∂φ2 ~
This eigenvalue problem is defined on a unit sphere. One of the boundary conditions is periodicity in φ, as in
(1.101). Another requires that the eigenfunction be finite at θ = 0 and θ = π. The resulting eigenfunctions are
spherical harmonics Ylm (θ, φ) and the corresponding eigenvalues are L2 = ~2 l(l + 1), where

l = 0, 1, 2, . . . and m = 0, ±1, ±2, . . . , ±l . (1.107)

Explicit form of the spherical harmonics is


s
(2l + 1)(l − m)!
Ylm (θ, φ) = Plm (cos θ)eimφ , (1.108)
4π(l + m)!

where Plm are associated Legendre polynomials. A list of spherical harmonics can be found in all books on
quantum mechanics. Normalization condition
Z

Ylm (θ, φ)Yl0 m0 (θ, φ)dΩ = δll0 δmm0 . (1.109)

13
§7 Description of quantum states I BASIC PRINCIPLES OF QUANTUM MECHANICS

Here dΩ = sin θdθdφ is an element of solid angle.


In addition to being eigenfunctions of L̂2 , spherical harmonics are also eigenfunctions of L̂z :

L̂z Ylm = −i~∂φ Ylm = ~mYlm . (1.110)

We make two important observations: (i) Eigenvalue l of operator L̂2 is 2l + 1-fold degenerate because 2m + 1
different values of Lz correspond to the same l, see (1.107). (ii) Two physical quantities can have definite values
in the same quantum state, i.e. two operators can share the same eigenfunction.

• Merzbacher, “Quantum Mechanics” 3rd edition, 11.3,11.4.

§7. Description of quantum states

Since, as we have seen in the previous section, two different physical quantities can simultaneously have definite
values, let us find a relationship between the corresponding operators. Consider operators F̂ and Ĝ that have definite
values in state ψn . Formally, this means that

F̂ ψn = Fn ψn , Ĝψn = Gn ψn . (1.111)

Acting on this equations with F̂ and Ĝ we get

ĜF̂ ψn = Fn Ĝψn = Fn GN ψn , (1.112)


F̂ Ĝψn = Gn F̂ ψn = Gn Fn ψn . (1.113)

Subtracting (1.112) from (1.113) gives

(ĜF̂ − F̂ Ĝ)ψn = 0 . (1.114)

Now consider any function ψ:


X X
(ĜF̂ − F̂ Ĝ)ψ = (ĜF̂ − F̂ Ĝ) an ψn = an (ĜF̂ − F̂ Ĝ)ψn = 0 . (1.115)
n n

The conclusion is

[F̂ , Ĝ] = 0 . (1.116)

Converse statement is also true, viz. if two operators commute, they have the same eigenfunctions. Indeed, let
[F̂ , Ĝ] = 0 and Ĝψn = Gn ψn . Then

Ĝ(F̂ ψn ) = F̂ (Ĝψn ) = Gn (F̂ ψn ) . (1.117)

This implies that function F̂ ψn is also an eigenfunction of Ĝn corresponding to eigenvalue Gn . If n is not degenerate,
than F̂ ψn ∝ ψn . Denoting the proportionality
P factor by Fn we obtain F̂ ψn = Fn ψn . If n is degenerate, we can
construct linear combinations ϕnl = k alk ψnk that are eigenfunctions of F̂ .
In summary, two physical quantities can have simultaneously definite values iff the corresponding operators com-
mute. Based on this result, one can decide which set of physical quantities can be chosen for a description of a
quantum system. If we know the values of all independent quantities with definite values in a given state, then the
wave function in that state must be an eigenfunction of all corresponding operators. We have already discussed two
examples: states with definite value of momentum p (its three projections commute with each other), see (1.93),
(1.51) and states with definite values of square of orbital angular momentum and its projection along some direction,
see (1.108), (1.58). More examples will be considered in forthcoming chapters.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §4; Messiah, “Quantum
Mechanics”, Ch. V, §13-18.

14
§8 Heisenberg relations I BASIC PRINCIPLES OF QUANTUM MECHANICS

§8. Heisenberg relations

Let K̂ and F̂ be Hermitian operators. Their commutator can be written as [K̂, F̂ ] = iĜ, where Ĝ is also a Hermitian
operator (see below (1.50)). Define

d = F̂ − hF i ,
∆F d = K̂ − hKi .
∆K (1.118)

It is easy to see using this definition that


d ∆F
K̂ F̂ − F̂ K̂ = ∆K d − ∆F
d ∆K
d = iĜ . (1.119)

Consider now the following auxiliary function I(α) of a real parameter α:


Z 2
d d )ψ dq ≥ 0 .
I(α) = (α∆K − i∆F (1.120)

Using the fact that F̂ and K̂ are Hermitian operators, and hence obey (1.44), we can cast this function into a different
form(3 :
Z Z
∗ ∗
I(α) = (α∆K d + i∆F d )ψ ∗ (α∆K
d − i∆F d )ψ = ψ ∗ (α∆K d + i∆Fd )(α∆K d − i∆F d )ψdq
Z



d 2 + (∆F
= ψ ∗ [α2 (∆K) d )2 + αĜ]ψdq = α2 (∆K)2 + (∆F )2 + α hGi ≥ 0 . (1.121)

I(α) is a parabola with its branches pointing up (because α2 > 0). Therefore, (1.121) is satisfied iff the minimum
value of I is positive. Minimum of I occurs at α = α0 given by
hGi
α0 = − , (1.122)
2 h(∆K)2 i
hence the minimum of I is
2

hGi
I(α0 ) = (∆F )2 − ≥ 0. (1.123)
4 h(∆K)2 i
We arrived at the Heisenberg uncertainty relation between the uncertainties of two Hermitian non-commuting opera-
tors:


hGi2
(∆F )2 (∆K)2 ≥ . (1.124)
4

The minimal possible product of uncertainties of two physical quantities – which do not simultaneously have definite
values – is achieved for a coherent state with a wave function ψc satisfying the following condition
Z   2
hGi
I(α0 ) = − d − i∆F
∆K d ψc dq = 0 . (1.125)
2
2 h(∆K) i

This implies that the coherent state wave function obeys equation
 
hGi d d
∆K + i∆F ψc = 0 . (1.126)
2 h(∆K)2 i

As a state that minimizes the uncertainty relation, turning inequality (1.124) into equation, the coherent state is a
closest approximation of the corresponding classical system.

∗ Example.

(3 Caution: operator i∆F


d is not Hermitian.

15
§9 Classical limit of quantum mechanics I BASIC PRINCIPLES OF QUANTUM MECHANICS

Let K̂ = x̂, F̂ = px , then Ĝ = ~ since [x̂, p̂x ] = i~. The uncertainty relation (1.124) reads



~2
(∆x)2 (∆px )2 ≥ . (1.127)
4
Denote hxi = x0 and hpx i = p0 , so that

d = x − x0 ,
∆K d = −i~∂x − p0 .
∆F (1.128)

Plugging into (1.126) yields


 
~(x − x0 )
+ i(−i~∂x − p0 ) ψc (x) = 0 . (1.129)
2 h(x − x0 )2 i

Look for a solution in form ψc (x) = N eiS(x)/~ , where N is a constant. We obtain an equation for S:

(x − x0 ) i dS ip0
2
+ − = 0. (1.130)
2 h(∆x) i ~ dx ~

Its solution (modulo an additive constant)

iS ip0 x (x − x0 )2
= − . (1.131)
~ ~ 4 h(∆x)2 i

Constant N is fixed by the normalization condition. The final result is


 
1 ip0 x (x − x0 )2
ψc (x) = exp − . (1.132)
(2π h(∆x)2 i)1/4 ~ 4 h(∆x)2 i

It is not difficult to verify that for this state (1.127) turns into equation.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §16.

§9. Classical limit of quantum mechanics

We know that macroscopic objects consists of microscopic ones: molecules, atoms, electrons etc. Microscopic
systems are governed by quantum mechanics, while the macroscopic ones by classical physics. Therefore, one should
be able to obtain classical equations of motion as a limiting case of quantum mechanics. This statement is known as the
correspondence principle. Consider application of this principle to the wave packet discussed in I §3. The uncertainties
of particle position and its momentum (1.14) decouple (i.e. become independent as in classical mechanics) in the formal
limit ~ → 0. In other words, in this limit one can in principle measure the position and momentum with any accuracy.
To better understand what is meant by this limit (after all ~ is a physics constant) note that ~ has units of action S.
So when for a given system S  ~, quantization can be neglected.
Consider the wave function of a free particle with given energy and momentum (1.7):

ψ(r, t) = N ei(p·r−Et)/~ . (1.133)

In classical mechanics, action of a free particle S = p · r − Et, so we can write

ψcl (r, t) = N eiS/~ . (1.134)

The classical limit corresponds to large values of the phase in (1.134). This is similar to the geometrical optics limit
of electromagnetic theory. There any component of electromagnetic wave can be written as u = N eiφ with real N
and φ. In geometrical optics limit phase φ becomes large corresponding to the limit of very short wave length λ → 0.
In such a case propagation of electromagnetic wave can approximated by straight lines (rays). Moreover, both S and
φ satisfy the variational principle that says that they must take the least possible value. Since any wave function can
be expanded in plane waves (1.133), it seems reasonable that (1.134) represents the form of the wave function in the
classical limit. We will discuss a more accurate derivation later in this course.

16
§9 Classical limit of quantum mechanics I BASIC PRINCIPLES OF QUANTUM MECHANICS

For future reference, recall a few facts form the classical mechanics. A mechanical system described by action S(q, t)
satisfies the Hamilton-Jacobi equation
∂S
− =H, (1.135)
∂t
where H = H (q, p, t) is the Hamiltonian function with the canonical momenta given by

∂S
p= . (1.136)
∂q
If the Hamiltonian function does not explicitly depend on time, then energy E = H is a conserved quantity. Depen-
dence of E on coordinates and momentum of a free particle follows from general physical requirements of uniformity
and isotropy of space and time and invariance under Galileo transformations, which leads to

p2
Hfree = (1.137)
2m
for one free particle.

17
§1 Hamiltonian II SCHRÖDINGER EQUATION

II. SCHRÖDINGER EQUATION

§1. Hamiltonian

Wave function ψ contains all information about a given physical system at any time (see I §2). Therefore, ∂ψ/∂t
at any given time must be determined by the value of ψ at the same time. The most general such relation reads
∂ψ
i~ = Ĥψ . (2.1)
∂t

Due to the superposition principle Ĥ must be a linear operator and i~ is extracted from Ĥ for convenience. Eq. (2.1)
describes the time-evolution of the wave function.(4
To find out the physical meaning of Ĥ we employ the correspondence principle. Consider the classical limit of (2.1)
by substituting the wave function (1.134) into (2.1). We have

∂S
Ĥψcl = − ψcl = Hψcl , (2.2)
∂t

i.e. operator Ĥ in the classical limit becomes the Hamiltonian function, see (1.135). On account of this correspondence
operator Ĥ is called the Hamiltonian operator or simply the Hamiltonian. Its operator form for a free particle follows
from (1.137):

p̂2 ~2 2
Ĥfree = =− ∇ . (2.3)
2m 2m
Interaction of a system of non-relativistic particles is described by potential energy U which is a function coordinates
of particles. Thus, the Hamiltonian of a system of non-relativistic particles reads
X p̂2
a
Ĥ = + U (r1 , . . . , ra ) , (2.4)
a
2m a

where index a runs over all particles. Substituting (2.4) into (2.1) we obtain the Schrödinger equation. For a single
particle it reads

∂ψ ~2 2
i~ =− ∇ ψ + U (r)ψ , (2.5)
∂t 2m
where U is the potential energy of the particle in an external field. For a free particle it has solution

ψ(r, t) = N ei(p·r−Et)/~ , (2.6)

which agrees with (1.7).


A possible time-dependence of the wave function ψ is restricted by the requirement of the total probability conser-
vation
Z
d
|ψ(q, t)|2 dq = 0 . (2.7)
dt

Let us verify that Schrödinger equation satisfies (2.7). From (2.1) it follows that

∂ψ ∂ψ ∗
i~ψ ∗ = ψ ∗ Ĥψ − i~ψ = ψ Ĥ ∗ ψ ∗ . (2.8)
∂t ∂t

(4 This time-evolution of the wave function is a result of the influence of forces acting on the system. It has nothing to do with the changes
in the system introduced by the measurement process. The later replaces one wave function by another, which is known as the wave
function collapse.

18
§2 Stationary states II SCHRÖDINGER EQUATION

Subtracting these equations gives


∂ ∗
i~ (ψ ψ) = ψ ∗ Ĥψ − ψ Ĥ ∗ ψ ∗ . (2.9)
∂t
Integrating over q and using the fact that Ĥ is a Hermitian operator we derive (2.7).
While the total probability is conserved, the probability density ρ = |ψ(q, t)|2 is not. Indeed, substituting (2.4) (for
a single particle) into (2.9) yields
∂ ~2 
i~ ρ=− ψ ∗ ∇2 ψ − ψ∇2 ψ ∗ . (2.10)
∂t 2m
Introduce the probability current density
~
j= (ψ ∗ ∇ψ − ψ∇ψ ∗ ) . (2.11)
2mi
Then, (2.10) can be cast in the form of the continuity equation:
∂ρ
= −∇ · j . (2.12)
∂t
Since ψ is a complex function, it can represented in terms of two real functions f and ϕ as
ψ(r, t) = f (r, t)eiϕ(r,t) . (2.13)
In this notation we have
 
2 ~ϕ
ρ=f , j = ρ∇ . (2.14)
m
In particular, if the phase ϕ is independent of coordinates, then the probability current vanishes j = 0.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §8,17, Merzbacher,
“Quantum Mechanics” 3rd edition,

§2. Stationary states

If the Hamiltonian function of a classical system is not explicitly time-dependent, its energy is conserved. To see
how energy conservation is expressed in quantum mechanics, consider a system describe by a Hamiltonian that does
not explicitly depend on time, i.e.

∂ Ĥ
= 0. (2.15)
∂t
For such a system time-dependence of the wave function can be computed from (2.1) employing the method of
separation of variables:
ψ(q, t) = T (t)ψE (q) . (2.16)
This yields

∂T (t) 1 ĤψE (q)


i~ = =E, (2.17)
∂t T (t) ψE (q)
where E is a constant. Indeed, a function of t can be equal a function of q only if both are constants. As a result we
have two equations
ĤψE (q) = EψE (q) , (2.18)
∂T (t)
i~ = ET (t) (2.19)
∂t
(2.20)

19
§2 Stationary states II SCHRÖDINGER EQUATION

Eq. (2.19) immediately gives

T (t) = e−iEt/~ . (2.21)


States with definite energy are called the stationary states. The time dependence of their wave functions reads
ψ(q, t) = ψE (q)e−iEt/~ . (2.22)
Eq. (2.19) becomes the time-independent Schrödinger equation:
~2 2
− ∇ ψ + (U − E)ψ = 0 . (2.23)
2m
In fact, since one is usually interested in stationary states, (2.23) as simply referred to as the Schrödinger equation.

Properties of the stationary states:

1. Time-dependence of a stationary state is uniquely determined by its energy.


2. ρ = const(t) and j = const(t).

3. If F̂ does not explicitly depend on time, then its expectation value in a stationary state ψ with energy E is also
time-independent:
Z
hF iE = ψ ∗ (q, t)F̂ ψ(q, t)dq = const(t) , (2.24)

which means that F is a conserved.


Note, that although hF iE is time-independent, the wave function of a stationary state does depend on time
according to (2.21).

a stationary state ψ(q, t) and let {ψn (q)} be a complete set of eigenfunctions of F̂ . We
4. Consider a system in P
can expend ψ(q, t) = n an (t)ψn (q). The probability to find the system in the n’th eigenstate is
Z 2

|an | = ψ(q, t)ψn (q)dq = const(t)
2 ∗
(2.25)

5. Functions {ψE (q)}, being eigenfunctions of Hamiltonian, formP a complete set. Therefore, any solution of the
Schrödinger equation can be expanded in them as ψ(q, t) = E TE (t)ψE (q). Owing to (2.21) we obtain a
general form of the expansion of an arbitrary wave function ψ(q, t) in a complete set of stationary states:
X
ψ(q, t) = an ψn (q)eiEn t/~ , (2.26)
n

for discrete spectrum and


Z
ψ(q, t) = aE ψE (q)eiEt/~ dE , (2.27)

for the continuous one.


Eqs. (2.26),(2.27) indicate that in order to calculate the time-evolution of a wave function one needs to (i) find
the spectrum of the Hamiltonian, (ii) fix coefficients an using the initial condition at t = 0 and (iii) sum or
integrate over the energy spectrum according to (2.26), (2.27).

6. Since Ĥ is a Hermitian operator, its eigenvalues E are real, see (1.64). This is also seen in (2.23) which is a real
differential equation. Moreover, the wave functions of the stationary states can be chosen to be real (except for

a system in magnetic field). Indeed, ψE and ψE satisfy the same equation. For a non-degenerate energy level
it implies that they are the same functions up to a non-important phase. For a degenerate level, one can form
real linear combinations of the wave functions.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §8,10,17,19,
Merzbacher, “Quantum Mechanics” 3rd edition, 3.1.

20
§3 Equations of motion in operator form II SCHRÖDINGER EQUATION

§3. Equations of motion in operator form

A. Time derivative of operators

Time derivative of a physical quantity F cannot be defined as in classical mechanics since in quantum mechanics the
value of F may not be known at two arbitrarily close time instants. Instead we will consider an observable quantity:
the expectation value of F . I pointed out in (2.24) that if F does not explicitly depend on time, its expectation
value in a stationary state isR time-independent. Consider now the expectation value of a Hermitian operator F in
an arbitrary state ψ: hF i = ψ ∗ F̂ ψdq. Taking its time derivative and using the Schrödinger equation ψ̇ = Ĥψ/i~,
ψ̇ ∗ = −Ĥ ∗ ψ ∗ /i~ we derive
Z ( )
d hF i ∗ ∂ F̂ ∂ψ ∗ ∗ ∂ψ
= ψ ψ+ F̂ ψ + ψ F̂ dq (2.28)
dt ∂t ∂t ∂t
Z ( )
∗ ∂ F̂ 1 ∗ ∗ 1 ∗
= ψ ψ − (Ĥ ψ )F̂ ψ + ψ F̂ Ĥψ dq (2.29)
∂t i~ i~
Z ( )
∗ ∂ F̂ 1
= ψ + [F̂ , Ĥ] ψdq , (2.30)
∂t i~

where in the last step I used (1.44) with the replacements ϕ → F̂ ψ and F̂ → Ĥ. Define now an operator dF̂ /dt such
that its expectation value obeys the following condition
* + Z
d hF i dF̂ dF̂
= = ψ∗ ψdq . (2.31)
dt dt dt

In view of (2.30) we obtain an operator expression of the time derivative

dF̂ ∂ F̂ 1
= + [F̂ , Ĥ] . (2.32)
dt ∂t i~
This implies that if ∂∂tF̂ = 0 (no explicit time dependence) and [F̂ , Ĥ] = 0, then hF i does not change with time, i.e.
F is a conserved quantity. We also conclude that any conserved quantity can be measured concurrently with the
Hamiltonian.

B. Ehrenfest theorem

Consider application of (2.32) to the motion of a single particle. Replacing F̂ first by p̂ and then by r̂ we get two
operator equations
dp̂ 1
= [p̂, Ĥ] , (2.33)
dt i~
dr̂ 1
= [r̂, Ĥ] . (2.34)
dt i~
Commutators are computed as follows:
[p̂, Ĥ]ψ = [p̂, U ]ψ = −i~[∇, U ]ψ = −i~[∇(U ψ) − U ∇ψ] = −i~∇U ψ , (2.35)
1 1 p
[r̂, Ĥ]ψ = [r̂, p̂2k ]ψ = {ei [r̂i , p̂k ]p̂k + p̂k [ei r̂i , p̂k ]} = i~ , (2.36)
2m 2m m
where ei is a unit vector. In the last line summation over k and i is implied and (1.55) is used. Plugging (2.35),(2.36)
into (2.33) and (2.34) gives
dp̂
= −∇U , (2.37)
dt
dr̂ p̂
v̂ = = , (2.38)
dt m

21
§3 Equations of motion in operator form II SCHRÖDINGER EQUATION

in agreement with the correspondence principle (which justifies our definition (2.31) a posteriori). These equations
are knows as the Ehrenfest theorem. Taking another time derivative of (2.38) we derive using (2.32)

dv̂ 1 dp̂ 1
= = − ∇U . (2.39)
dt m dt m
Formally, this formula has the same form as the classical equation of motion. However, it must be understood as a
relationship between the corresponding expectation values:
Z Z
d2
m 2 ψ ∗ r̂ψd3 r = − ψ ∗ ∇U ψd3 r . (2.40)
dt

To better understand the difference between the operator formula (2.39) and its classical counterpart, it is worth
examining the classical limit at this point. Namely, we would like to know what are the conditions that allow us to
replace operators in (2.39) with the corresponding classical quantities. For the sake of simplicity we consider only the
x-components. Denote x̄ = hxi, ∆x = x − x̄ and expand at small ∆x (in the classical limit ∆x vanishes):

∂U (x) ∂U (x̄) ∂ 2 U 1 ∂3U


= + 2
∆x + (∆x)2 + . . . (2.41)
∂x ∂ x̄ ∂ x̄ 2 ∂ x̄3
R
Substituting into (2.40) and noting that |ψ|2 ∆xdx = hxi − hx̄i = 0 we have

d2 x̄ ∂U (x̄) 1 ∂ 3 U

m 2
= − − 3
(∆x)2 + . . . (2.42)
dt ∂ x̄ 2 ∂ x̄
In the classical approximation we must require
3
∂U

 1 ∂ U (∆x)2 (2.43)
∂ x̄
2 ∂ x̄ 3

In addition to the requirement that uncertainty in x be small, we also have to find a similar requirement for the
momentum. Momentum uncertainty squared is defined as


2
(∆px )2 = p2x − hpx i , (2.44)

2
It is small when p2x ≈ hpx i , i.e. when the following condition is satisfied

2
~2
hpx i  (∆px )2 ≥ . (2.45)
4 h(∆x)2 i

Inequalities (2.43) and (2.45) indicate that the classical limit applies when external field is slowly varying and momenta
are large. We can combine (2.43) and (2.45) into one condition:

2 ~2 U 000
hpx i  . (2.46)
8 U0

C. Virial theorem

As in classical mechanics, we can prove the quantum virial theorem. On the one hand, in a stationary state
expectation value of p · r is time-independent, see (2.24), i.e.

d
hp · riE = 0 . (2.47)
dt
On the other hand,

d 1 1 p̂2
(p · r) = [p̂ · r̂, Ĥ] = p̂ · [r̂, Ĥ] + [p̂, Ĥ] · r̂ = − r̂ · ∇U . (2.48)
dt i~ i~ m

22
§4 Schrödinger and Heisenberg pictures of time-evolution II SCHRÖDINGER EQUATION

Employing the definition (2.31) we cast this in form


 
d p2
hp · riE = − hr · ∇U iE (2.49)
dt m E

We thus arrive at the required relation


 
p2
= hr · ∇U iE . (2.50)
m E

For illustration, suppose U ∝ rn . Then, hr · ∇U i = n hU i and the virial theorem implies a relation between the
expectation values of kinetic and potential energies:
 2
p n
= hU iE . (2.51)
2m E 2

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §9,19, Merzbacher,
“Quantum Mechanics” 3rd edition, 3.3,3.5,3.6.

§4. Schrödinger and Heisenberg pictures of time-evolution

In practical applications it is sometimes convenient to use different representations or “pictures” of time-evolution.


Observed quantities do not depend on a choice of a picture, of course.

A. Schrödinger picture

Thus far we have followed the Schrödinger picture of time-evolution in which wave functions depend on time,
whereas the operators corresponding the physical quantities usually do no explicitly depend on time. We can express
the time-evolution of a wave function ψ(q, 0) (at t = 0) as a result of action of the time-evolution operator Ŝ(t) as
follows

ψ(q, t) = Ŝ(t)ψ(q, 0) . (2.52)

The normalization of the wave function should not change with time, i.e.
Z Z
ψ ∗ (q, t)ψ(q, t)dq = ψ ∗ (q, 0)ψ(q, 0)dq = 1 . (2.53)

On the other hand,


Z Z Z
∗ ∗ ∗
ψ (q, t)ψ(q, t)dq = [Ŝ ψ (q, 0)][Ŝψ(q, 0)]dq = ψ ∗ (q, 0)Ŝ † Ŝψ(q, 0)dq , (2.54)

where I used the definition (1.43) with F̂ → Ŝ † and ϕ → Ŝψ. We conclude that

Ŝ † Ŝ = 1̂ . (2.55)

Operators satisfying (2.60) are unitary operators. To find the explicit form of operator Ŝ, substitute (2.52) to the
Schrödinger equation:

∂ Ŝ(t)
i~ ψ(q, 0) = Ĥ Ŝ(t)ψ(q, 0) . (2.56)
∂t
In operator form:

∂ Ŝ(t)
i~ = Ĥ Ŝ(t) . (2.57)
∂t

23
§4 Schrödinger and Heisenberg pictures of time-evolution II SCHRÖDINGER EQUATION

For closed systems Hamiltonian does not explicitly depend on time. In such a case one can integrate (2.57) to obtain
i
Ŝ(t) = e− ~ Ĥt . (2.58)

How does the operator (2.58) act on a wave function? Since Ŝ is a function of the Hamiltonian, it is helpful to
expand ψ(q, 0) in a set of stationary states {ϕn (q)}n and act on the result by the evolution operator Ŝ. Namely, using
(2.52) we have
!k
X X X 1 iĤt
ψ(q, t) =Ŝ an ϕn (q) = an − ϕn (q)
n n
k! ~
k
X X 1  iEn t k X
= an − ϕn (q) = an ϕn e−iEn t/~ , (2.59)
n
k! ~ n
k

in agreement with (2.26). One can think of (2.58) as a compact way of expressing the time evolution of (2.26) and
(2.27).

B. Heisenberg picture

In the Heisenberg picture wave functions do not evolve with time, but the operators do. Let ψ(q, t) be a wave
function in the Schrödinger picture describe a certain state and ψH (q, t) a wave function in the Heisenberg picture
describing the same state. According to (2.52), in order that ψH (q, t) be time-independent it must satisfy

ψH (q, t) = Ŝ −1 ψ(q, t) , (2.60)

where the inverse operator Ŝ −1 is defined by the condition Ŝ −1 Ŝ = 1̂. Consider now an expectation value of a physical
quantity F . It must be the same regardless of the picture we choose to work in. Thus,
Z Z Z
∗ ∗ ∗ †
hF i = ψ F̂ ψdq = (ŜψH ) F̂ ŜψH dq = ψH Ŝ F̂ ŜψH dq . (2.61)

Let F̂ be an operator in the Schrödinger picture. Then the same operator in the Heisenberg picture, which we denote
by F̂ , reads
F̂H = Ŝ −1 F̂ Ŝ , (2.62)

where I used (2.55) to write Ŝ † = Ŝ −1 .


Suppose that at some initial time t F̂H (t) = F̂ . After evolving over a short time-interval ∆t with (2.62) we have

F̂H (t + ∆t) = Ŝ −1 (∆t)F̂H (t)Ŝ(∆t) . (2.63)


Expanding in Taylor series we get
   
dF̂H i i
F̂H (t) + ∆t = 1+ Ĥ∆t F̂H (t) 1 − Ĥ∆t . (2.64)
dt ~ ~
Keeping only linear terms in ∆t we obtain
dF̂H 1
= [F̂H , Ĥ] . (2.65)
dt i~
This formula is similar to (2.32), but has a different meaning. Eq. (2.32) is a definition of the time-derivative of
F̂ , while in dF̂H /dt in (2.65) is a time-derivative of a physical quantity (whose operator is time-dependent in the
Heisenberg picture).

∗ As an example, consider motion of a free particle. Its Hamiltonian is Ĥ = p̂2 /2m. In the Schrödinger picture
p̂ = −i~∇. In the Heisenberg picture
i i it
p̂H = e ~ Ĥt (−i~∇)e− ~ Ĥt = p̂ + [Ĥ, p̂] = p̂ , (2.66)
~

24
§4 Schrödinger and Heisenberg pictures of time-evolution II SCHRÖDINGER EQUATION

where I used the following identity:


1 1
e B̂e− = B̂ + [Â, B̂] + [Â, [Â, B̂]] + . . . (2.67)
1! 2!
Position operator can be calculated similarly using (2.36):

i i it (it)2 p̂
r̂H = e ~ Ĥt r̂e− ~ Ĥt = r̂ + [Ĥ, r̂] + [Ĥ, [Ĥ, r̂]] + . . . = r̂ + t . (2.68)
~ 2~2 m
Alternatively, we can solve equations of motion
d i p̂ d i
r̂H = [Ĥ, r̂H ] = , p̂H = [Ĥ, p̂H ] = 0 (2.69)
dt ~ m dt ~
which leads to (2.68).

C. Interaction picture

Another popular representation of the time-evolution is the interaction picture. Suppose we can split the Hamilto-
nian into two parts:

Ĥ = Ĥ0 + V̂ , (2.70)

where Ĥ0 describes a system without taking interactions into account, while V̂ describes interactions. Introduce a
unitary operator(5 :

Sˆ = eiĤ0 t/~ (2.71)

and define the wave function in this picture as follows

ψI (q, t) = Sˆψ(q, t) . (2.72)

Substituting this into the Schrödinger equation i~ψ̇ = Ĥψ yields

i~ψ̇I = SˆV̂ Sˆ−1 ψI . (2.73)

Define operator V̂I in the interaction picture

V̂I = SˆV̂ Sˆ† = eiĤ0 t/~ V̂ e−iĤ0 t/~ . (2.74)

Then we can rewrite (2.73) as

i~ψ̇I = V̂I ψI . (2.75)

In general, operator F̂ evolves in time as

F̂I = eiĤ0 t/~ F̂ e−iĤ0 t/~ . (2.76)

To summarize, in the interaction picture, wave functions ψI and operators F̂I evolve in time only according to the
non-interacting part of the Hamiltonian as given by (2.72) and (2.76).

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §13, Merzbacher,
“Quantum Mechanics” 3rd edition, Ch.14.

(5 Note that I use a different font for the time-evolution operator Sˆ in order to emphasize that it is not the same as in (2.58).

25
§5 Symmetries and conserved quantities II SCHRÖDINGER EQUATION

§5. Symmetries and conserved quantities

In classical mechanics uniformity of space implies momentum conservation, its isotropy implies the orbital angular
momentum conservation and uniformity of time intimates energy conservation. Since the Hamilton function of a
system carries all information about its dynamics, it must exhibit the same symmetry properties as the system. The
same is true in quantum mechanics were we have to replace the Hamilton function with the corresponding operator Ĥ.
Let Ŝ be an operator performing a symmetry transformation. Invariance of Hamiltonian under such transformation
means that [Ŝ, Ĥ] = 0 (6 .

A. Translations

Consider translation (displacement as a whole) of a closed system of particles by a small constant vector δ, viz.
ra → ra0 = ra + δ. Under this transformation the wave function transforms as follows
X X 
ψ(r1 + δ, r2 + δ . . .) = ψ(r1 , r2 , . . .) + δ · ∇a ψ(r1 , r2 , . . .) = 1 + δ · ∇a ψ(r1 , r2 , . . .) . (2.77)
a a
P
Operator 1 + a δ · ∇a is an operator of an infinitesimal translation. Since the Hamiltonian is invariant (i.e. do not
change) under translations we arrive at the statement
X 
∇a , Ĥ = 0 . (2.78)
a

Recalling
P that p̂a = −i~∇a , we conclude that uniformity of space implies the total momentum conservation
[ a p̂a , Ĥ] = 0. In fact, we guessed the form of the momentum operator in I §4. Instead, we could have defined
momentum as the infinitesimal translation operator (up to a constant factor). Consider the operator of finite trans-
lations
T̂a ψ(r) = ψ(r + a) , (2.79)
where a is a finite constant. Expanding the right-hand-side in Taylor series we get
"  2 #
i 1 i
T̂a ψ(r) = 1 + a · p̂ + a · p̂ + . . . ψ(r) = eia·p̂/~ ψ(r) . (2.80)
~ 2 ~
Thus,
T̂a = eia·p̂/~ . (2.81)
The generator of translations is the following operator

∂ i
ĜT = T̂a = p̂ = ∇ . (2.82)
∂a a=0 ~
Notice that the generator does not depend on a.

B. Rotations

Invariance of a closed system under rotations follows from the isotropy of space. The corresponding operator must
be proportional to the operator of the total angular momentum. To see that this is indeed the case, consider an
infinitesimal rotation by angle δφ:
ra → ra0 = ra + δφ × ra . (2.83)

(6 Although this is pretty evident, let me illustrate this in the case of infinitesimal continuos transformation. Such
R a transformation can
be written as Ŝ = 1 + iŝ, where ŝ is Hermitian operator. This follows from the requirement that ψ ∗ ψdq = (Ŝψ)∗ Ŝψdq up to small
R

terms of order s2 . It is easy to check that in order that hHi stay invariant under ψ → Ŝψ we need to require that [ŝ, Ĥ] = 0, implying
that [Ŝ, Ĥ] = 0. This equation is true for discrete transformations as well.

26
§5 Symmetries and conserved quantities II SCHRÖDINGER EQUATION

Under rotation the wave function ψ(r) changes as follows



ψ ri + [δφ × r]i = ψ(ri + ijk δφj rk ) = ψ(ri ) + ∂i ψ(ri )ijk δφj rk = ψ(ri ) + δφ · (r × ∇)ψ(ri ) , (2.84)
X 
ψ(r1 + δφ × r1 , r2 + δφ × r2 , . . .) = (1 + δφ · ra × ∇a ) ψ(r1 , r2 , . . .) (2.85)
a
P
Operator 1 + δφ · a ra × ∇a ) is the operator of infinitesimal rotations. Since the closed system is invariant under
rotations it commutes with the Hamiltonian. In particular,
X
ra × ∇a ), Ĥ] = 0 . (2.86)
a
P
It follows that a ra × ∇a ) must be proportional to the total angular momentum of the system, which justifies our
definition (1.56).
Operator of rotation by a finite angle φ around axis with direction n can be computed by analogy with (2.79)-(2.81)
with the result
R̂φn = eiL̂·nφ/~ . (2.87)
Generator of rotations around the n-direction is

∂ n i
Ĝn = R̂φ = L̂ · n . (2.88)
∂φ φ=0 ~
If represent vectors as columns

Ax
A =  Ay  , (2.89)
Az
then rotation operators can be represented as 3 × 3 matrices. For example, operator of rotation around x-axis reads
 
1 0 0
R̂φx =  0 cos φ − sin φ  (2.90)
0 sin φ cos φ
The corresponding generator of rotations around x-axis is
 
0 0 0

Ĝx = R̂x =  0 0 −1  (2.91)
∂φ φ φ=0 0 1 0
By the same token,
   
0 0 1 0 −1 0
Ĝy =  0 0 0  , Ĝz =  1 0 0  (2.92)
−1 0 0 0 0 0
It is easy to verify that [Ĝx , Ĝy ] = Ĝz etc. Since Ĝx = iL̂x /~, Ĝy = iL̂y /~, Ĝz = iL̂z /~, we confirm the commutation
relations (1.57). We will see later that this is only one of infinite number of possible representations of angular
momentum.

Operators of all physical quantities can be classified according to their transformational properties under rotations.
Scalar operator Ŝ is an operator that commutes with the operator of rotation:
[Ŝ, R̂φn ] = 0 ⇒ [Ŝ, L̂i ] = 0 . (2.93)
For example, r̂ 2 , p̂2 , p̂ · r̂, L̂2 are all scalar operators.
Vector operator V̂ is an operator whose components satisfy
[L̂i , V̂k ] = iikl V̂l . (2.94)
Examples of vector operators: r̂, L̂, (p̂ · r̂)p̂.
Tensor operator of second rank T̂ik is an operator whose components satisfy
[L̂i , T̂kl ] = i(ikp δnl + iln δkp )T̂pn . (2.95)
Examples of tensor operators: r̂i r̂k , r̂i p̂k .

27
§5 Symmetries and conserved quantities II SCHRÖDINGER EQUATION

C. Time evolution

For infinitesimal time shift t → t0 = t + τ


 
∂ ∂
ψ(t + τ ) = ψ(t) + ψ(t)τ = 1+τ ψ(t) . (2.96)
∂t ∂t
Due to uniformity of time, the Hamiltonian of a closed system must commute with the operator ∂t . Therefore we
identify the time derivative operator as the energy operator (up to a multiplicative factor). But energy operator is
the Hamiltonian, see (2.1). It clearly commutes with itself and so energy is conserved.
Operator of finite time shift can is defined and calculated as in (2.79) and (2.80). Using ∂t = −iĤ/~ we have

Ŝτ ψ(t) = ψ(t + τ ) = e−iĤτ /~ ψ(t) . (2.97)

Operator Ŝτ evolves quantum system forward in time. This agrees with (2.58).
Translation, rotation and time evolution are examples of continuous transformations. Invariance of a system
(and hence the Hamiltonian) under such transformations leads to conservation laws both in classical and quantum
mechanics.

D. Space inversion

Unlike the continuous symmetry transformations that we have discussed thus far, space inversion is a discrete
transformation. In classical mechanics it does not lead to any conservation laws, however in quantum mechanics it
does. Space inversion reflects the position vector r → r 0 = −r, i.e. the right-handed coordinate system goes into the
left-handed one. The corresponding operator is called the parity operator P̂ and is defined as
P̂ ψ(r) = ψ(−r) . (2.98)
Consider an eigenvalue problem for the parity operator:
P̂ ψ(r) = P ψ(r) . (2.99)

Applying P̂ on both sides again and noting that P̂ 2 = 1̂ (unit operator) we derive ψ(r) = P 2 ψ(r), implying that
P = ±1 (2.100)
are the eigenvalues of the parity operator. The corresponding eigenstates ψ± satisfy

P̂ ψ± = ±ψ± . (2.101)
State ψ+ (ψ− ) is said to be an even (odd ) parity state.
Suppose that the Hamiltonian of a system is an even function of its coordinates, viz. H(r) = H(−r). Then,

[Ĥ, P̂ ]ψ(r) = Ĥ(r)P̂ ψ(r) − P̂ Ĥ(r)ψ(r) = Ĥ(r)ψ(−r) − Ĥ(−r)ψ(−r) = 0 ⇒ [Ĥ, P̂ ] = 0 . (2.102)


The fact that the parity operator commutes with such a Hamiltonian means that parity is conserved. Among systems
conserving parity are those that interact through electromagnetic and strong forces, and systems in central potentials.
Wave functions of the stationary states of a system invariant under the space inversion can be chosen to have a definite
parity.
In spherical coordinates spatial inversion corresponds to the transformation θ → θ0 = π − θ and φ → φ0 = π + φ.
This is easy to verify by writing down the spherical coordinates of r and observing that r → r 0 = −r. Under such
transformation
P̂ Ylm (θ, φ) = Ylm (π − θ, φ + π) = (−1)l Ylm (θ, φ) . (2.103)

Thus, spherical harmonics are eigenfunctions of P̂ .

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §15,26,30, Merzbacher,
“Quantum Mechanics” 3rd edition, 17.1,17.2,17.9.

28
§5 Symmetries and conserved quantities III MOTION IN ONE-DIMENSION

III. MOTION IN ONE-DIMENSION

§1. General properties of one-dimensional motion

Many problems in Quantum Mechanics can be reduced to motion in one dimension. In this section we discuss
general properties of one-dimensional motion. In the following sections we consider a number of applications. We
start with the Schrödinger equation

d2 ψ 2m
+ 2 [E − U (x)]ψ = 0 . (3.1)
dx2 ~

1. Energy levels of discrete spectrum are non-degenerate. Indeed, suppose the contrary is true, viz. there are two
linearly independent eigenfunctions corresponding to the same energy levels:

ψ100 2m ψ 00
= 2 [U (x) − E] = 2 ⇒ ψ100 ψ2 = ψ200 ψ1 . (3.2)
ψ1 ~ ψ2
Integating by parts
Z Z Z
(ψ100 ψ2 − ψ200 ψ1 )dx = ψ10 ψ2 − ψ10 ψ20 dx − ψ20 ψ1 + ψ20 ψ10 dx = ψ10 ψ2 − ψ20 ψ1 . (3.3)
| {z }
const

Recall that for a bound state ψ1,2 → 0 as x → ±∞, hence const=0 and

ψ10 ψ0
= 2. (3.4)
ψ1 ψ2
Integrating again
Z   Z  
ψ10 ψ0 d d
const = − 2 dx = ln ψ1 − ln ψ2 dx = ln ψ1 − ln ψ2 . (3.5)
ψ1 ψ2 dx dx

Thus, ψ1 ∝ ψ2 and upon normalization we obtain that ψ1 = ψ2 .


2. Let the discrete spectrum of the Hamiltonian be E0 < E1 < . . . and the corresponding eigenfunctions ψ0 , ψ1 , . . .
correspondingly. The oscillation theorem states that eigenfunction ψn has n zeros, i.e. it vanishes n times, not
including the boundary values. You can find a proof in the book of Messiah. As an illustration to this theorem,
in Fig. 1 I plotted wave functions ψ0 , ψ1 and ψ2 of the infinitely deep square potential well, see equations (3.53)
and (3.56) in III §3.
Ψn

1.0

0.5

x
-0.4 -0.2 0.2 0.4

-0.5

-1.0

p
FIG. p
1: Wave functions of the threeplowest states in the infinitely deep square potential well: ψ0 = 2/a cos(πx/a) (black),
ψ1 = 2/a sin(2πx/a) (blue), ψ2 = 2/a cos(3πx/a) (red). a = 1.

3. Consider a potential well of a general form shown in Fig. 2.

29
§1 General properties of one-dimensional motion III MOTION IN ONE-DIMENSION

UHxL

1.0 U0

0.5

x
-3 -2 -1 1 2 3

-0.5 Umin

FIG. 2: An arbitary potential well U (x).

(a) Discrete spectrum corresponds to motion in


a finite region of space. Moreover, motion with energies below
Umin is not possible because hEi = hU i + p2 /2m ≥ hU i ≥ Umin . Thus, for a potential in Fig. 2 the
discrete spectrum lies in the interval Umin < E < 0.
(b) In the region 0 < E < U0 the spectrum is continuous because particle can go to x → ∞. The eigenstates
are not degenerate, which can be proven as before using the fact ψ1,2 → 0 as x → −∞.

00 2m 2mE
at x → +∞ : ψ + 2 Eψ = 0 , ⇒ ψ = A cos(kx + δ) , k = . (3.6)
~ ~ p
2m 2m(U0 − E)
at x → −∞ : ψ 00 − 2 (U0 − E)ψ = 0 , ⇒ ψ = Beκx + Ce−κx , κ = . (3.7)
~ ~
The boundary condition requires that C = 0. Therefore, there is only one function that satisfies the
boundary conditions at given E.
(c) At E > U0 the spectrum is continuous and two-fold degenerate:

0 0 2mE
at x → +∞ : ψ = Aeik x + Be−ik x , k 0 = , (3.8)
p ~
2m(E − U0 )
at x → −∞ : ψ = Ceikx + De−ikx , k = . (3.9)
~
There are two solutions that satisfy the boundary conditions – one with particles incident from the left
and another from the right. A typical problem is that given a flux of particles incident on the potential
from, say, x = −∞, find what fraction of particles is reflected from the potential and what fraction is
transmitted. In such a case at x > 0 there are only particles that move toward x = +∞, meaning that
B = 0. It is also convenient to normalize the incident flux of particles to one particle per unit volume. To
this end, we compute the probability current density of incident particles using (2.11):
~k
j = |C|2 = |C|2 v . (3.10)
m
The required normalization is C = 1. Fluxes of transmitted j 0 and reflected j 00 particles are
j 0 = |A|2 v , (3.11)
00 2
j = |D| v . (3.12)
The corresponding fractions are transmitted and reflected
j0
T = = |A|2 , (3.13)
j
j 00
R= = |D|2 . (3.14)
j
T and R are called the transmission and reflection coefficients respectively. The probability conservation
implies R + T = 1.

30
§2 Delta-potential III MOTION IN ONE-DIMENSION

4. Consider two potentials U (x) and Ũ (x) = U (x) + δU (x) with δU (x) ≥ 0. Let En (λ) and ψn (λ) be eigenvalues
and eigenfunctions of the discrete spectrum of the following Hamiltonian
p2
Ĥ(λ) = + U (x) + λδU (x) , (3.15)
2m
where λ is a parameter. In one of the home assignments you proved the following theorem:
Z
∂En (λ) ∂ Ĥ
= ψn∗ (λ) ψn (λ)dx (3.16)
∂λ ∂λ
Plugging (3.15) into (3.16) we have
Z
∂En (λ)
= |ψn (λ)|2 δU (x)dx ≥ 0 . (3.17)
∂λ

Note that En (0) = En is an eigenvalue of Ĥ = p̂2 /2m+U , while En (1) = Ẽn is an eigenvalue of Ĥ = p̂2 /2m+ Ũ .
We conclude that Ẽn ≥ En . This means that a more shallow potential has higher lying energy levels. This
statement can be generalized to a multidimensional potentials.
As δU increases, energy levels go up, so that eventually the highest energy level (the one that has the smallest
|En |) becomes En = 0. A further increase of δU will push this level into the continuous spectrum. Thus, in
order to determine at what values of parameters of the Hamiltonian a new bound states appears/disappears one
has to solve the Schrödinger equation with E = 0.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §21, Messiah, “Quan-
tum Mechanics”, Sec. 12.

§2. Delta-potential

A. Continuity of the wave function

Wave function must be a continuous function of coordinates in order that its probabilistic interpretation make
sense. Moreover, its first derivative must also be a continuous function, provided that the potential is a non-singular
function. To prove this statement, consider a particle of mass m moving in the following singular one-dimensional
potential

U (x) = V (x) − αδ(x − x0 ) , (3.18)

where α and x0 are real numbers, V (x) is a non-singular function. Schrödinger equation reads

~2 00
− ψ (x) + [V (x) − α(x − x0 )]ψ(x) = Eψ(x) . (3.19)
2m
Integrating over x0 − ∆x ≤ x ≤ x0 + ∆x:
Z x0 +∆x   Z x0 +∆x
~2 00
dx − ψ (x) + [V (x) − α(x − x0 )]ψ(x) = E dxψ(x) . (3.20)
x0 −∆x 2m x0 −∆x

The discontinuity of the wave function at x is

∆ψ 0 (x) = ψ 0 (x + ∆x) − ψ 0 (x − ∆x) . (3.21)

Taking ∆x → 0 we derive from (3.20)


2mα
∆ψ 0 (x0 ) = − ψ(x0 ) . (3.22)
~2
This indicates, in particular, that if the potential is continuous, the first derivative of the wave function is continuous
as well.

31
§2 Delta-potential III MOTION IN ONE-DIMENSION

B. Discrete spectrum

Consider a particle moving in the delta potential U = −αδ(x). The corresponding Schrödinger equation

~2 00
− ψ (x) − α(x)δ(x)ψ(x) = Eψ(x) . (3.23)
2m
Suppose E < 0 and introduce a parameter κ that has dimension of inverse distance (same as wavenumber):
r
2mE
κ = − 2 > 0. (3.24)
~
Solution to the Schrödinger equation that is finite at x → ±∞ is

ψ(x) = Ae−κx x > 0, (3.25)


ψ(x) = Beκx , x < 0. (3.26)

To fix constants A and B we need two conditions. First, we use (3.22):


2mα
∆ψ 0 (0) = A(−κ) − Bκ = − A. (3.27)
~2
Second, the continuity of the wave function itself at x = 0 implies A = B. Thus,

κ= (3.28)
~2
There are two possibilities: (i) if α < 0, then according to (3.28) κ < 0 which contradicts (3.24). In this case there
are no bound states. (ii) If α > 0, there is one bound state

κ2 ~2 mα2
E0 = − =− 2 . (3.29)
2m 2~
The corresponding wave function is
√ mα
ψ0 (x) = κ0 e−κ0 |x| , κ0 = . (3.30)
~2

C. Continuous spectrum

Continuous spectrum lies at E > 0. It is convenient to introduce


r
2mE
k= > 0. (3.31)
~2
General solution to the Schrödinger equation is

ψ(x) = A1 eikx + A2 e−ikx , x > 0, (3.32)


ikx −ikx
ψ(x) = B1 e + B2 e , x < 0. (3.33)

A solution describing particles incident on the barrier from x = −∞ corresponds to A2 = 0. Normalization of one
particle per unit volume is fixed by setting B1 = 1. Using the continuity of ψ and (3.22) at x = 0 we can now compute
A1 and B2 :
1
A1 = mα , (3.34)
1+ i~2 k
1
B2 = − i~2 k
. (3.35)
1+ mα

32
§2 Delta-potential III MOTION IN ONE-DIMENSION

Transmission and reflection coefficients (3.13) and (3.14) are given by


1
T = |A1 |2 = mα2
, (3.36)
1+ 2~2 E
1
R = |B2 |2 = 2~2 E
. (3.37)
1+ mα2

They satisfy the probability conservation R + T = 1, as can be readily verified. Notice that the transmission and
reflection coefficients do not depend on the sign of α.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §25, Merzbacher,
“Quantum Mechanics” 3rd edition, 6.1.

§3. Rectangular potential well

U (x)
a
− a2 2
x

1 2 3

−U0

FIG. 3:

A. Discrete spectrum

In the case of the rectangular potential shown in Fig. 3 all solutions have E ≥ −U0 , see III §1. Thus, we can define
p
2m(E + U0 )
k= (3.38)
~
in place of E inside the well −a/2 ≤ x ≤ a/2. Furthermore, bound states correspond to energies E ≤ 0, at which the
particle cannot escape to x → ±∞.
Another important observation is that the potential is invariant under the inversion U (−x) = U (x) implying that
Ĥ(−x) = Ĥ(x). In this case the stationary states can be chosen to have definite parity either even or odd, which will
be denoted by plus or minus superscript, see (2.102). As a consequence, it is sufficient to find the wave functions only
at x ≥ 0. Schrödinger equation in the region 2 reads

ψ200 (x) + k 2 ψ2 = 0 , 0 ≤ x ≤ a/2 . (3.39)

It has two independent solutions with definite parity

ψ2+ = B cos(kx) , P = +1 , (3.40)


ψ2− = C sin(kx) , P = −1 . (3.41)

In the region 3,

ψ300 − κ2 ψ3 = 0 , x ≥ a/2 , (3.42)

where
√ p
−2mE 2m|E|
κ= = . (3.43)
~ ~

33
§3 Rectangular potential well III MOTION IN ONE-DIMENSION

Its solution finite at x → ∞ is

ψ3 (x) = Ae−κx , x ≥ a/2 . (3.44)

Boundary conditions for even states:

ψ2+ (a/2) = ψ3 (a/2) , ψ20+ (a/2) = ψ30 (a/2) , ⇒ (3.45)


−κa/2 −κa/2
B cos(ak/2) = Ae , −kB sin(ak/2) = −Aκe . (3.46)

Taking the ratio gives


r
2mU0
k tan(ka/2) = κ = − k2 . (3.47)
~2
Bound states exist only for those values of k = kn , n = 0, 1, . . . , N that satisfy (3.47). There is a finite number of
solutions because according to (3.47)

2mU0 a2
0 ≤ ka ≤ . (3.48)
~
The larger is U0 a2 , i.e. the deeper and wider the the well, the more bound states it contains. This can be seen in
Fig. 4 where I illustrated a numerical solution to this equation. Intersection of blue and red lines give the values of k
that solve (3.47). The red line terminates at the maximal ka satisfying (3.48). Using (3.38) they can be converted to
the corresponding energy levels.

8 6

6 4

4 2

ka
2 π 2π 3π 4π 5π 6π
-2
ka
π 2π 3π 4π 5π 6π
-4
-2
-6
-4
-8
-6

2
FIG.
p 4: Values of parameters are a = 1 and mU0 /~ = 100. Left panel (even solutions): p red line is tan(k/2), blue panel
is 2mU0 /k ~ − 1. Right panel (odd solutions): red line is cot(k/2), blue panel is − 2mU0 /k2 ~2 − 1. Intersections are
2 2

solutions to (3.47) and correspondingly. In this figure there are three even and two odd states.

Similarly, for odd states, boundary conditions yield


κ
C sin(ak/2) = Ae−κa/2 , C cos(ak/2) = − Ae−κa/2 . (3.49)
k
Thus,

k cot(ka/2) = −κ . (3.50)

Solution to this equation is illustrated in the right panel of Fig. 3.

 Consider the limit U0 a2 → ∞ at fixed ka, which corresponds to infinitely deep well. In such case, (3.47) for even
states becomes
r
2mU0
tan(ka/2) ≈  1. (3.51)
k 2 ~2
The corresponding solutions
~2 kn2 π 2 (2n + 1)2 ~2
kn a = π(2n + 1) , ⇒ En+ = − U0 = − U0 . (3.52)
2m 2ma2

34
§3 Rectangular potential well III MOTION IN ONE-DIMENSION

Even eigenfunctions:
r  
2 π(2n + 1)x
ψn+ (x) = cos , 0 ≤ x ≤ a/2 . (3.53)
a a

At x > a/2 eigenfunction vanishes because its argument is



2mU0
κn a ≈ a → ∞. (3.54)
~

For odd states


π 2 (2n)2 ~2
− cot(ka/2)  1 ⇒ kn a = 2πn ⇒ En− = − U0 . (3.55)
2ma2
The corresponding eigenfunctions are
r  
2 π(2n)x
ψn− (x) = sin , 0 ≤ x ≤ a/2 . (3.56)
a a
Thus, there is an infinite number of even and odd levels in the infinitely deep potential well.
Notice that ψ ± (±a/2) = 0, which means that particle cannot penetrate the region |x| ≥ a/2 because of the infinite
potential. At x = ±a/2 the derivative ψ 0 is discontinuous.
√ √
√ limit a U0  ~/ m, which corresponds to shallow and/or narrow well there is only one even
 In the opposite
level because k ≈ 2mU0 /~ and ka  2. The corresponding energy is

~2 k 2
E= − U0 ≈ 0 . (3.57)
2m

∗ Example: application of the oscillation theorem to the rectangular potential well.

Bound states lie at E ≤ 0, while continuous spectrum at E > √ 0. Thus E = 0 is the highest possible energy a bound
state can have. The corresponding values of k and κ are k0 = 2mU0 /~ and κ0 = 0. Imagine that one can change
the width a and the depth U0 of the well. Using the result (3.17) and the ensuing discussion, we come to the following
conclusions. As the width and/or depth of the well increase more bound states can be accommodated in the well. A
new bound state appear with E = 0 and sink as a and/or U0 increase.
To see how this happens consider first even states. Eq. (3.47) implies
 
k0 a
k0 tan = κ0 = 0 , ⇒ k0 a = 2πn , n = 0, 1, 2, . . . (3.58)
2
The ground state exists if k0 a ≥ 0, which is always satisfied. The first excited even state exists if k0 a > 2π, etc. The
n’th excited even state exists if k0 a > 2πn. The total number of even states
 
ak0
N+ = + 1. (3.59)

Square brackets indicate the integer part. Similarly, odd states exist when k0 a = π(1 + 2n), n = 0, 1, . . .. If k0 a < π
there are no odd states. The total number of odd states is
   
− ak0 1 ak0 1
N = − +1 = + . (3.60)
2π 2 2π 2

B. Continuous spectrum

Continuous spectrum lies at E > 0. In this case we are interested to know which fraction of the flux incident on the
well is reflected and which one is transmitted. Accordingly we write down the solution to the Schrödinger equation

35
§3 Rectangular potential well III MOTION IN ONE-DIMENSION

in the following way


0 0 √
ψ1 = Aeik x + Be−ik x , x < −a/2 , k0 = 2mE/~ , (3.61)
ikx −ikx
p
ψ2 = αe + βe , −a/2 < x < a/2 , k = 2m(E + U0 )/~ , (3.62)
0
ψ3 = Ceik x , x > a/2 . (3.63)

Matching conditions at x = a/2 are


0 k 0 ik0 a/2
Ceik a/2 = αeika/2 + βe−ika/2 , Ce = αeika/2 − βe−ika/2 . (3.64)
k
Solving for α and β we find
  
k0 0
α= C
2 1+ k ei(k −k)a/2
  (3.65)
β= C
1− k0 0
ei(k +k)a/2
2 k

Matching conditions at x = −a/2 are


0 0
Ae−ika/2 + Beika/2 = αe−ik a/2 + βeik a/2 , (3.66)
k 0  −ik0 a/2 0

Ae−ika/2 − Beika/2 = αe − βeik a/2 , (3.67)
k
which together with (3.65) yield
      
C ik0 a k k0 −ika k k0 ika
A= e 1+ 0 1+ e + 1− 0 1− e (3.68)
4 k k k k
       
C k k0 k k0
B= 1− 0 1+ e−ika + 1 + 0 1− eika . (3.69)
4 k k k k
Transmission coefficient
1
T =  , (3.70)
k0 k 2
1+ 1
4 k − k0 sin2 (ka)

and the reflection coefficient R = 1 − T .


When sin(ka) = 0, i.e. ka = πn, the flux of particles is completely transmitted T = 1, R = 0. The corresponding
values of energy

π 2 ~2 n2
Ẽn = − U0 , where n is integer, (3.71)
2ma2
are called the resonance energies or virtual levels. Since Ẽn ≥ 0

2a2 mU0
n2 ≥ . (3.72)
π 2 ~2

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §22,25,50; Merzbacher,
“Quantum Mechanics” 3rd edition, 6.4,7.4; Messiah, “Quantum mechanics”, VI.10.

§4. Infinite one-dimensional crystal

Consider motion of a particle in a periodic one-dimensional potential



X
U (x) = α δ(x − na) . (3.73)
n=−∞

36
§4 Infinite one-dimensional crystal III MOTION IN ONE-DIMENSION

30

25

20

15

10

-3 -2 -1 1 2 3

FIG. 5: a = 1. Delta functions are approximated by (B12) with  = 0.01.

that serves as an elementary model of an ideal infinite crystal. Points xn = na correspond to the lattice sites. To
determine the wave function we need to solve the Schrödinger equation between any two lattice sites, say an < x <
a(n + 1) and apply the matching conditions at the points x = a(n − 1) and x = an. Solution to the Schrödinger
equation reads

2mE
ψ(x) = An eik(x−na) + Bn e−ik(x−na) , k = , an < x < a(n + 1) . (3.74)
~
Here An , Bn are constants and constant phases e±ina are written for further convenience (they could have been
absorbed into the An and Bn ). Since the potential is a periodic function of x with the period a, the wave function ψ
must also be periodic with the same period:

ψ(x + a) = µψ(x) , (3.75)

where µ is a complex constant such that |µ|2 = 1. In other words, ψ(x + a) and ψ(x) can differ at most by a phase(7

µ = eiqa , −π ≤ qa ≤ π . (3.76)

States with fixed q are called the Bloch states, while ~q is called the quasi-momentum (don’t confuse it with the
particle momentum ~k).
In the interval a(n − 1) < x < an, the wave function reads

ψ(x) = An−1 eik(x−(n−1)a) + Bn−1 e−ik(x−(n−1)a) , a(n − 1) < x < an . (3.77)

On the other hand, we can obtain the wave function in that interval be translating (3.74) as follows

ψ(x + a) = An eik(x−(n−1)a) + Bn e−ik(x−(n−1)a) , a(n − 1) < x < an . (3.78)

Using (3.75) we find the recursion relations

An = µAn−1 , Bn = µBn−1 . (3.79)

Now we apply the boundary conditions at x = na. For δ → 0 we have

ψ(na + δ) = ψ(na − δ) , (3.80)


2mα
ψ 0 (na + δ) − ψ 0 (na − δ) = ψ(na) . (3.81)
~2
The wave functions on the each side of x = na are

ψ(na − δ) = An−1 eik(x−(n−1)a) + Bn−1 e−ik(x−(n−1)a) , (3.82)


ψ(na + δ) = An eik(x−na) + Bn e−ik(x−na) , (3.83)

(7 Condition |µ|2 = 1 is also required to insure that the wave function is not divergent: if |µ|2 > 1, then ψ diverges at x → ∞, if |µ|2 < 1,
it diverges at x → −∞

37
§4 Infinite one-dimensional crystal III MOTION IN ONE-DIMENSION

Using (3.80),(3.81) we get

An−1 eika + Bn−1 e−ika = An + Bn , (3.84)


2mα
ik(An−1 eika − Bn−1 e−ika ) − ik(An − Bn ) = 2 (An + Bn ) , (3.85)
~
Replacing An−1 and Bn−1 with An and Bn according to (3.79) we arrive at the following system of equations

An (eika − µ) + Bn (e−ika − µ) = 0 , (3.86)


     
ika 2mαi −ika 2mαi
An e − µ 1 + 2 − Bn e −µ 1− 2 = 0. (3.87)
~ k ~ k
Dividing the second of these equations by the first one we obtain a condition on µ:
h mα i
µ2 − 2µ cos(ka) + 2 sin(ka) + 1 = 0 . (3.88)
~ k
Being a quadratic equation it has two solutions µ1 and µ2 :
p mα
µ1,2 = f (E) ± f 2 (E) − 1 , f (E) = cos(ka) + 2 sin(ka) . (3.89)
~ k
In view of (3.76) f 2 ≤ 1 for otherwise µ is not a pure phase. It is easy to verify that µ1 µ2 = µ1 µ∗1 = 1.
We can rewrite (3.88) using (3.76) as follows

µ2 + 1
f= = cos(qa) . (3.90)

Now employing (3.89) we obtain

cos(qa) = cos(ka) + sin(ka) . (3.91)
~2 k
Solutions to this equations is a set of discrete values kn and the corresponding energy levels En . They are depicted
in Fig. 6.
fHEL
3

cos(qa)

ka
Π 2Π 3Π

1st zone 2nd zone 3rd zone

-1

FIG. 6: Solution to (3.91) with cos(qa) = 0.6 (red line) and mα/~2 = 2. f (E) is a blue line. Intersection of blue and red lines
are solutions to (3.91). Physical values of f satisfy f 2 ≤ 1 and lie between the two horizontal black lines which divide possible
values of ka into the Brillouin zones. In each zone there is one solution kn a (red circles) corresponding to the energy level En .

The wave function ψ can be represented in a general form

ψ(x) = eiqx u(x) , (3.92)

38
§5 Harmonic oscillator III MOTION IN ONE-DIMENSION

where u(x) is a periodic function with period a. This statement is known as the Bloch theorem. We can prove it as
follows

u(x + a) = e−iqx e−iqa ψ(x + a) = e−iqx µ−1 ψ(x + a) = e−iqx ψ(x) = u(x) , (3.93)

where (3.75) has been used.


Solution to the Schrödinger equation have form (3.74) or (3.92) even at x → ±∞ indicating that a particle is not
localized and can move in the entire crystal with quasi-momentum ~q.

• Additional reading: Ch. Kittel, “Introduction to Solid State Physics”.

§5. Harmonic oscillator

A. Energy spectrum

In many physical problems a physical system moves near a minimum of potential well U (x). Suppose that this
minimum is at x = 0 and expand
1
U (x) ≈ U (0) + U 00 (0)x2 . (3.94)
2
The corresponding force is −U 00 (0)x. In classical mechanics motion of a particle under such force is governed by the
equation mẍ = −U 00 (0)x describing periodic motion with frequency ω = (U 00 (0)/m)1/2 . In quantum mechanics we
are interested to solve the following Schrödinger equation
 
~2 d2 ω2 m 2
− + x − E ψ(x) = 0 . (3.95)
2m dx2 2

Introduce dimensionless variables


r
mω 2E
ξ=x , = . (3.96)
~ ~ω

We have
 
d2 2
− ξ +  ψ(ξ) = 0 . (3.97)
dξ 2
2
At ξ → ∞, we can neglect . The resulting equation ψ 00 = ξ 2 ψ has asymptotic solution ψ ∼ e−ξ /2
. Thus, we can
look for a solution in the form
2
ψ(ξ) = v(ξ)e−ξ /2
. (3.98)

Substituting and denoting  − 1 = 2n we get

v 00 − 2ξv 0 + 2nv = 0 . (3.99)

We can find a solution to this equation in the form of the power series

X
v(ξ) = ak ξ k . (3.100)
k=0

Upon substitution

X ∞
X ∞
X
ak k(k − 1)ξ k−2 − 2 kak ξ k + 2n ak ξ k = 0 . (3.101)
k=2 k=0 k=0

39
§5 Harmonic oscillator III MOTION IN ONE-DIMENSION

Changing the summation variable in the first sum as k 0 = k − 2 and then renaming it again to k 0 → k we obtain

X ∞
X ∞
X
ak+2 (k + 2)(k + 1)ξ k − 2 kak ξ k + 2n ak ξ k = 0 . (3.102)
k=0 k=0 k=0

This equations is satisfied if coefficients in front of same powers vanish, viz.

2(k − n)
ak+2 (k + 2)(k + 1) − 2kak + 2nak = 0 ⇒ ak+2 = ak . (3.103)
(k + 2)(k + 1)

If we start with a finite a0 and a1 = 0 we get a polynomial containing only even powers of ξ. However, if we start
with a finite a1 and a0 = 0 we get a polynomial containing only odd powers of ξ. This reflects the fact that the
Hamiltonian of the harmonic oscillator is invariant under the spatial inversion, and thus we can choose its stationary
states to have definite parity.
It can be shown (see the Merzbacher’s book) that ψ(ξ) diverges at ξ → ±∞ unless the polynomials are finite, which
requires n to be integer. Since E ≥ 0, n ≥ −1/2 implying that n = 0, 1, . . .. For illustration consider n = 2:

2 · (−2)
a2 = a0 = −2a0 , a4 = a6 = . . . = 0 . (3.104)
2·1
The corresponding polynomial is H2 (ξ) = −(2ξ 2 − 1)a0 . Consider n = 3.

2 · (1 − 3) 2
a3 = a1 = − a1 , a5 = a7 = . . . = 0 . (3.105)
3·2 3
The corresponding polynomial is H3 (ξ) = (− 32 ξ 3 + ξ)a1 . Coefficients a0 and a1 are chosen for each n in such a way
that the resulting polynomials satisfy the following formula
2
dn e−ξ
n ξ2
Hn (ξ) = (−1) e . (3.106)
dξ n
These are known as the Hermit polynomials. The lowest polynomials are

H0 = 1 , H1 = 2ξ , H2 = 4ξ 2 − 2 , H3 = 8ξ 3 − 12ξ , (3.107)

Let me note for the future reference the following recurrence relations
1
ξHn (ξ) = nHn−1 (ξ) + Hn+1 (ξ) , (3.108)
2
dHn
= 2nHn−1 (ξ) . (3.109)

The wave functions normalized as
Z +∞
ψn (x)ψm (x)dx = δmn (3.110)
−∞

read
 mω 1/4  r 
1 mω 2
ψn (x) = √ Hn x e−mωx /(2~) . (3.111)
π~ n
2 n! ~

The corresponding energy levels


 
1
E = ~ω n + . (3.112)
2

The harmonic oscillator spectrum is equidistant, meaning that the energy difference between the nearby levels is
a constant ~ω independent of n. The ground state energy E0 = ~ω/2 is known as the zero-point energy and is
interpreted as the energy of vacuum fluctuations. Since there is only one eigenfunction for each n, the spectrum is
non-degenerate as expected, see III §1.

40
§5 Harmonic oscillator III MOTION IN ONE-DIMENSION

Since the Hamiltonian of the harmonic oscillator is invariant under spatial inversion, the stationary states have
definite parity, viz. states with even n are even functions of x, while states with odd n are odd.

∗ Example. Expectation value of the particle position in the n’th stationary state can be computed as follows
Z ∞
hxin = ψn2 xdx = 0 (3.113)
−∞

because x is an odd function, while ψn2 is even one. Variance of the position is
Z ∞

2
2
∆x n = x n = ψn2 x2 dx . (3.114)
−∞

The easiest way to take this integral is to employ the orthonormality condition (3.110). To this end we substitute
(3.111) into (3.108) and obtain after some simple algebra
r r
n n+1
ξψn = ψn−1 + ψn+1 . (3.115)
2 2
Using this formula we can reduce the integrand of (3.114) to a sum of products of the wave functions of different
states:
p !
~ ~ n n + 1 n(n + 1)
(xψn )2 = (ξψn )2 = ψ2 + 2
ψn+1 + ψn−1 ψn+1 . (3.116)
mω mω 2 n−1 2 2

Using (3.110) the integral over x is now trivial:


 

2
~ 1
∆x n
= n+ . (3.117)
mω 2

For the future reference let me record another useful relation that follows from (3.109):
∂ψn 1 √ √ 
=√ nψn−1 − n + 1ψn+1 . (3.118)
∂ξ 2

B. Ladder operators

Eigenvalue problem for the harmonic oscillator can be reformulated in terms of the ladder operators (8
r  
mω i
â = x̂ + p̂ , (3.119)
2~ mω
r  
mω i
↠= x̂ − p̂ . (3.120)
2~ mω
Properties of the ladder operators:

1. â and ↠are not Hermitian, i.e. they do not correspond to any physical quantities. Only their certain combina-
tions, such as (3.124),(3.125) and (3.126) are Hermitian.
2. The ladder operators satisfy the following commutation relations:

[â, ↠] = 1 , [â, â] = 0 , [↠, ↠] = 0 , (3.121)

which can be derived using (3.119),(3.120) and the commutators [x̂, p̂] = i~ etc.

(8 In fact, this problem can be solved entirely employing these operators without solving first the Schrödinger equation, see Merzbacher’s
“Quantum Mechanics”, 10.6.

41
§6 Integral form of Schrödinger equation III MOTION IN ONE-DIMENSION

3. To verify the physical meaning of the ladder operators, apply them to the wave functions of the stationary
states. Using (3.111) and (3.109) we derive
r  
mω ~ d √
âψn = x+ ψn = nψn−1 , (3.122)
2~ mω dx
r  
mω ~ d √
↠ψn = x− ψn = n + 1ψn+1 . (3.123)
2~ mω dx

The raising operator ↠moves the particle to the next higher excited state, while the lowering operator â moves
the particle one state lower.

Equations (3.119),(3.120) can be used to express x̂ and p̂x through the ladder operators:
r
~
x̂ = (â + ↠) , (3.124)
2mω
r
mω~ †
p̂ = i (â − â) , (3.125)
2
In terms of the ladder operators the Hamiltonian reads
 
p̂2 ω 2 mx2 ~ω  2  ~ω  2  1
Ĥ = + =− â + (↠)2 + â↠+ ↠â + â + (↠)2 − â↠− ↠â = ↠â + ~ω , (3.126)
2m 2 4 4 2

where I used (3.121). Evidently, operator ↠â commutes with the Hamiltonian. Employing (3.122) and (3.123) we
find its eigenvalue in the n’th stationary state:

↠âψn = nψn . (3.127)



Thus we reproduce the spectrum (3.112): E = ~ω n + 12 .
The ladder operators can be applied to calculate the eigenfunctions ψn as follows. Let ψ0 be the wave function of
the ground state. Then, according to (3.122)
r  
mω ~ d
âψ0 = x+ ψ0 = 0 . (3.128)
2~ mω dx

Solution to this differential equation is


2
ψ0 = N0 e−mωx /2~
. (3.129)

Constant N0 is fixed by the normalization condition. It is easy to see from (3.123) that the wave functions of the
excited states can be obtained as follows
1
ψn = √ (↠)n ψ0 . (3.130)
n!
You can verify that it gives the same result as (3.111).

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §23, Merzbacher,
“Quantum Mechanics” 3rd edition, Ch. 5, 10.6.

§6. Integral form of Schrödinger equation

Eigenvalue problem for the Hamiltonian can be written down as an integral equation for the wave function. I will
consider separately the discrete and continuous spectra.

42
§6 Integral form of Schrödinger equation III MOTION IN ONE-DIMENSION

A. Discrete spectrum

We start with the Schrödinger equation

~2 00
− ψ (x) + U ψ(x) = Eψ(x) , (3.131)
2m
and assume that U is a potential well such that U → 0 as x → ±∞ and consider the bound states corresponding to
E < 0. Introduce the Green’s function GE (x, x0 ) satisfying

~2 d2
− GE (x, x0 ) − EGE (x, x0 ) = δ(x − x0 ) . (3.132)
2m dx2
Suppose that we solved (3.132) and found GE (x, x0 ). Then a solution to (3.131) reads
Z ∞
−κx
ψ(x) = Ae + Be −κx
GE (x, x0 )U (x0 )ψ(x0 )dx0 , (3.133)
−∞

where A, B are constants and


r
2mE
κ= − . (3.134)
~2
Indeed,
   
~2 d 2 ~2 d 2
− − E ψ(x) = − − E (Ae−κx + Beκx )
2m dx2 2m dx2
Z +∞  
~2 d2 GE (x, x0 )
− − − EGE (x, x ) U (x0 )ψ(x0 )dx0
0
−∞ 2m dx2
Z +∞
=− δ(x − x0 )U (x0 )ψ(x0 )dx0
−∞
= − U (x)ψ(x) . (3.135)

Note, that although (3.133) is a formal “solution” of the Schrödinger equation, it is still an equation for ψ. Since we
are interested in bound states which have wave functions localized at x → ±∞ we set A = B = 0.
The Green’s function can be calculate as described in III §2B (9 . We write
 0
C(x0 )eκ(x−x ) , x < x0 ,
GE (x, x0 ) = 0 (3.136)
D(x0 )e−κ(x−x ) , x > x0 .

and use boundary conditions (continuity of GE (x, x0 ) at x = x0 and discontinuity its derivative given by (3.22)) to fix
C and D:
2m m
C = D, κ(D + C) = , ⇒ C=D= (3.137)
~2 ~2 κ
Thus,
m −κ|x−x0 |
GE (x, x0 ) = e . (3.138)
κ~2
Finally, Schrödinger equation in the integral form is
Z +∞
m 0
ψE (x) = − 2 e−κ|x−x | U (x0 )ψE (x0 )dx0 . (3.139)
κ~ −∞

(9 The boundary condition (3.22) has to be slightly modified: ∆ψ 0 (x0 ) = −2mα/~2 because of the difference between (3.19) and (3.132).

43
§6 Integral form of Schrödinger equation III MOTION IN ONE-DIMENSION

Subscript E indicates that we used the boundary condition for the bound states with E < 0.

∗ Example. For a particle moving in the delta-potential well U = −αδ(x) we have


m
ψE (x) = αψE (0)e−κ|x| , (3.140)
κ~2
which is the eigenfunction of the only bound state. This equation must be consistent at x = 0 implying that
m
1= α, (3.141)
κ~2
which coincides with the result of III §2B.

B. Continuous spectrum

To obtain the Green’s function at E > 0 we notice that


r r
2mE 2mE
κ = − 2 = ∓ik , k= . (3.142)
~ ~2
Therefore, instead of (3.138) we obtain

im ±ik|x−x0 |
G± 0
E (x, x ) = ± e . (3.143)
~2 k

Schrödinger equation for scattering problem where particles with momentum p = ~k are incident from the left is
equivalent to the following integral equation
Z
0 0 + 0 0
ψk+ (x) = eikx − G+
E (x, x )U (x )ψk (x )dx . (3.144)

The first term in the right-hand-side corresponds to the incident particles, while the second term describes the reflected
and transmitted waves. Indeed, U (x0 ) in the integrand is non-vanishing only at x . a, where a is the range of the
potential. At |x|  a we can approximate |x − x0 | ≈ |x| so that asymptotic form of the solution to the Schrödinger
equation becomes:
( ikx  im R
e − ~2 k U (x0 )ψk+ (x0 )dx0 eikx = A1 eikx , x → +∞ ,
+
ψk (x) ≈  R (3.145)
eikx − ~im2k U (x0 )ψk+ (x0 )dx0 e−ikx = eikx + B2 e−ikx , x → −∞ .

Compare this with (3.32),(3.33). Schrödinger equation in the integral form (3.144) is the starting point of the scattering
theory.

∗ Example. In potential U = αδ(x) (3.144) reads

imα ik|x| +
ψk+ (x) = eikx − e ψk (0) . (3.146)
~2 k
At x = 0 this implies

~2 k
ψk (0) = . (3.147)
~2 k + imα
This coincides with the wave functions given by (3.32),(3.33), (3.34),(3.35).

44
§1 Representations of states IV REPRESENTATION THEORY

IV. REPRESENTATION THEORY

§1. Representations of states

A. Hilbert space

Consider a quantum system described by a set of quantum numbers a. Let the corresponding wave function be
ψa (q). Position q is an eigenvalue of a Hermitian position operator q̂. For a single particle q is the same as r. From
a mathematical standpoint there is nothing special about the operator q̂ and its eigenvalues q. Therefore, one can
describe the same quantum system by different functions Ψa (p), Φa (En ) etc., which depend on particle momenta,
energy, etc. These are called representations and the mathematical framework that describes the relationships between
different representations is called the representation theory.
In any representation, expectation values of physical quantities must be the same. It is therefore convenient to
think about an abstract state vector |ai, called ket, that incorporates the physical information about the state of the
system but does not dependent on a particular representation. All possible state vector form an abstract vector space.
Superposition principle dictates that α|ai + β|bi is also a possible vector state. In other words, the vector space is
linear.
For every ket |ai we introduce a dual state vector bra ha| such that |ai† = ha|. We also define the scalar product of
two state vectors |ai and |bi as hb|ai. Clearly, ha|bi† = hb|ai. Linear vector space with defined scalar product is called
the Hilbert space.
Similarly to the more familiar Euclidean space, we can project a state vector |ai onto any complete set of linearly
independent basis vectors. In Euclidean space such projection are called coordinates; any vector A is completely
described by a set of ordered coordinates. It is convenient to choose the orthonormal basis vectors. By the same
token, the infinite dimensional Hilbert space can be span by a set of orthonormal basis state vectors. Since spectrum
of any Hermitian operator is complete and orthonormal, it can be chosen as a set of basis state vectors. Unlike the
Euclidean space where projections are real numbers, projections in Hilbert space are functions.
For example, let’s take spectrum of the position operator q̂ as a basis set in the Hilbert space. Denote the basis
vectors by |qi. These vectors form a complete set, therefore we can expand any state vector as follows
Z
|ai = ψa (q)|qidq , (4.1)

where ψa (q) are coefficients. Orthonormality condition requires that

hq|q 0 i = δ(q − q 0 ) . (4.2)

Thus, forming a scalar product of (4.1) with hq 0 | yields

ψa (q) = hq|ai . (4.3)

From a mathematical standpoint, ψa (q) is a projection of a state vector |ai onto the basis vector |qi. It’s physical
interpretation is that for a physical system initially in a state |ai, |ψa (q)|2 is the probability density that a measurement
of position yields q. In other words, ψa (q) is the wave function of the system in the coordinate q-representation.
In particular, for a system that consists of just one particle, the wave function is depicted by the scalar product
hr|ai = ψa (r). Coordinate representation is an example of a continuous representation.
Consider now a Hermitian operator having a discrete spectrum. Denote its eigenstate vectors as |En i, where
n = 0, 1, 2, . . .. For example, Hamiltonians of harmonic oscillator and infinitely deep square well. Choose |En i as a
basis set of state vectors. The orthonormality condition reads

hEm |En i = δmn . (4.4)

We can expand any state vector |ai in a complete set of state vectors |En i as follows:
X
|ai = Φa (En )|En i . (4.5)
n

Forming a scalar product of this expression with hEm | and using (4.4) we find that Φa (En ) = hEn |ai. For a physical
system initially in a state |ai, |Φa (En )|2 is the probability that a measurement of physical quantity E yields En .

45
§1 Representations of states IV REPRESENTATION THEORY

Therefore, we can interpret Φa (En ) as a wave-function of a system described by quantum numbers a. Since Φa (En )
is a function of En it is called the wave function in the discrete E-representation.
In general, in a Hilbert space span by the eigenstate vectors |F i of a Hermitian operator F̂ , the wave function of
a system in a state |ai is depicted by the scalar product hF |ai. Obviously, physical observables do not depend on a
choice of the representation (i.e. a set of the basis state vectors). So, in practice one chooses a representation that is
the most convenient for a particular problem.

B. Transformation between a continuous and a discrete representations

What is the relationship between the wave functions of a physical system in the same state |ai but in different
representations? Consider first a continuous q-representation and a discrete E-representation. Denote by ϕn (q) =
hq|En i the n’th eigenfunction in the coordinate representation; its complex conjugate ϕ∗n (q) = hq|En i† = hEn |qi.
Orthonormality of the set {ϕn (q)}n means that
Z Z

ϕm (q)ϕn (q)dq = dqhEm |qihq|En i = δmn . (4.6)

Comparing (4.6) and (4.4) we notice the following identity


Z
|qihq| dq = 1 , (4.7)

reflecting the completeness of the set of |qi states.


To change from the coordinate q-representation ψa (q) = hq|ai of a state vector |ai to the discrete E-representation
we form a scalar product of (4.5) with the state vector |qi:
X
hq|ai = hq|En ihEn |ai . (4.8)
n

The same statement can be written using notations of the previous chapters:
X
ψa (q) = ϕn (q)Φa (En ) . (4.9)
n

Eq. (4.8) implies the following identity:


X
|En ihEn | = 1 , (4.10)
n

Transformation inverse to (4.9) reads in two equivalent notations:


Z
Φa (En ) = dq ϕ∗n (q)ψa (q) , (4.11)
Z
hEn |ai = dq hEn |qihq|ai . (4.12)

Forming scalar products of (4.10) with ha| and |ai we recover a known renormalization condition:
X X
ha|En ihEn |ai = |Φa (En )|2 = 1 . (4.13)
n n

C. Transformation between two continuous representations

Consider now a Hermitian operator p̂ having a continuos spectrum |pi. For example, momentum operator of a free
particle. We choose its eigenstates as a complete basis. Similarly to (4.1),(4.2),(4.3) we write
Z
|ai = ψa (p)|pidp , hp|p0 i = δ(p − p0 ) . (4.14)

46
§2 Representations of operators IV REPRESENTATION THEORY

where ψa (q) = hp|ai is the wave function in the p-representation. Forming a scalar product of (4.14) with hq| we get
Z Z
hq|ai = ψa (p)hq|pidp = hp|aihq|pidp , (4.15)

where hq|pi is the wave function of a system with a given momentum p in the q-representation. Now, forming a scalar
product of (4.1) with hp| yields inverse transformation:
Z Z
hp|ai = ψa (q)hp|qidq = hp|qihq|aidq , (4.16)

which can also be achieved by inserting the identity (4.7) in the left hand side of (4.16).
In particular, if a system consists of just one unbound particle, a set of eigenfunctions of momentum operator in
the coordinate representation reads (see (1.93)):
1
ϕp (r) = hr|pi = eip·r/~ . (4.17)
(2π~)3/2

A set of eigenfunctions of position operator r in the momentum representation is


1
hp|ri = e−ip·r/~ . (4.18)
(2π~)3/2

• Additional reading: Merzbacher, “Quantum Mechanics” 3rd edition, 9.1,9.2,3.2.

§2. Representations of operators

Action of a Hermitian operator F̂ on the state vector |ai is defined as

F̂ |ai = |bi , (4.19)

where |bi is another state vector from the same Hilbert space.

hb| = (F̂ |ai)† = ha|F̂ † = ha|F̂ (4.20)

Thus, we define F̂ to act on bra from the right.


In the coordinate representation F̂ is a function of q’s and ∂/∂q’s.

ψb (q) = F̂ ψa (q) , or hq|bi = F̂ hq|ai . (4.21)

A. Discrete representation

Let’s determine the form of F̂ in a discrete E-representation. Noting that


X
hq|ai = hq|En ihEn |ai , (4.22)
n
X
hq|bi = hq|En ihEn |bi , (4.23)
n

and substituting into (4.21) we get


X X
hq|En ihEn |bi = F̂ hq|En ihEn |ai (4.24)
n n

47
§2 Representations of operators IV REPRESENTATION THEORY

Now, multiply both sides by hEm |qi and integrate over all q using (4.4):
XZ XZ X
dqhEm |qihq|En ihEm |bi = dqhEm |qiF̂ hq|En i hEn |ai = Fmn hEn |ai , (4.25)
n n n

where
Z
Fmn = hEm |F̂ |En i = dqhEm |qiF̂ hq|En i . (4.26)

Finally, action of operator F̂ in the E-representation reads


X
hEm |bi = hEm |F̂ |En ihEn |ai . (4.27)
n

Thus, if spectrum En is not degenerate, then operator F̂ is represented by an infinite dimensional square matrix with
elements Fmn . Indexes m and n label rows and columns correspondingly. The state vector hEm |bi is represented by
a column, so that the action of operator becomes equivalent to matrix multiplication.
R R
If operator F̂ is Hermitian the corresponding matrix Fmn is Hermitian as well. Indeed if ψ ∗ F̂ ϕdq = ϕF̂ ∗ ψ ∗ dq
then,
Z Z

Fmn = ψm F̂ ψn dq = ψn F̂ ∗ ψm
∗ ∗
dq = Fnm . (4.28)

In particular, diagonal matrix elements are real Fnn = Fnn . Any operator is represented by a diagonal matrix in its
own representation. For example, if energy spectrum is discrete, then its matrix element in the energy representation
is hEn |Ĥ|En i = En δmn .

B. Continuous representation

The form of operator F̂ in momentum representation can be obtained in a similar way. We start with the expansions
Z
hq|ai = dphq|pihp|ai , (4.29)
Z
hq|bi = dphq|pihp|bi . (4.30)

which we substitute into the definition


hq|bi = F̂ hq|ai . (4.31)
0
Multiplying both sides with hp |qi and integrating over q using (4.2) yields
Z
hp0 |bi = dphp0 |F̂ |pihp|ai , (4.32)

with
Z
0
hp |F̂ |pi = dqhp0 |qiF̂ hq|pi . (4.33)

Eqs. (4.32),(4.33) describe action of F̂ on state vectors in the p-representation.

It follows from (4.26) and (4.33), that in general, an operator in the bra-ket notation can be represented as
Z
F̂ = dq |qiF̂ hq| , (4.34)

in a continuous q-representation or as
X
F̂ = |En iF̂ hEn | , (4.35)
n

in a discrete E-representation

∗ Examples.

48
§2 Representations of operators IV REPRESENTATION THEORY

1. In the coordinate representation momentum operator is p̂ = −i~∂x . In momentum representation

hp0 |p̂|pi = php0 |pi = pδ(p0 − p) , (4.36)

since |pi is an eigenstate vector of p̂. Eq. (4.36) means that the momentum operator is diagonal in its own
representation as expected. Using (4.34) we get
Z
hp |bi = dp p δ(p0 − p)hp|ai = p0 hp0 |ai ,
0
(4.37)

which indicates that action of p̂ in momentum space is simple multiplication.


2. In the momentum representation the form of the position operator x̂ is found from
Z
hp0 |x̂|pi = dxhp0 |xix̂hx|pi . (4.38)

Recalling (4.17) and using its one-dimensional analogue


1
hx|pi = √ eipx/~ (4.39)
2π~
we derive x̂hx|pi = −i~∂p hx|pi. Thus, the matrix element of momentum operator is
Z
hp0 |x̂|pi = −i~∂p dxhp0 |xihx|pi = −i~∂p δ(p0 − p) . (4.40)

Its action on a state vector reads


Z
∂ ∂
hp0 |bi = −i~ dphp|ai δ(p0 − p) = i~ 0 hp0 |ai . (4.41)
∂p ∂p
Therefore,

x̂ = i~ . (4.42)
∂p
It is easy to check that

[x̂, p̂x ] = i~ (4.43)

holds in momentum representation.


3. Hamiltonian in momentum representation reads
p̂2
Ĥ = + U (i~∇p ) . (4.44)
2m
The corresponding matrix element
p2
hp0 |Ĥ|pi = δ(p0 − p) + U (i~∇p )δ(p0 − p) . (4.45)
2m

4. Consider a wave function on a unit sphere ψ(θ, φ) = hθ, φ|ψi. We can change to the angular momentum
representation in which L2 and Lz have definite values. Let’s define the state vector |l, mi to be an eigenstate
of the corresponding operators, i.e. L̂2 |l, mi = ~2 l(l + 1)|l, mi and L̂z |l, mi = ~m|l, mi. Then,
X
hθ, φ|ψi = hθ, φ|l, mihl, m|ψi (4.46)
l,m

We have seen functions hθ, φ|l, mi = Ylm (θ, φ) before in I §6C. Functions hl, m|ai are the wave functions in the
angular momentum representation. They are given by
Z
hl, m|ψi = dΩhl, m|θ, φihθ, φ|ψi . (4.47)

49
§3 Eigenvalue problem in discrete represenation IV REPRESENTATION THEORY

As a more specific example, let


r
3
ψ(θ, φ) = sin θ sin φ. (4.48)

We can expand it in spherical harmonics as follows
i
ψ(θ, φ) = √ (Y1,−1 + Y1,1 ) . (4.49)
2
Hence, function ψ in the angular momentum representation is
i
h1, −1|ψi = h1, 1|ψi = √ , (4.50)
2
and all other hl, m|ψi’s vanish. A compact way to represent this state vector is to depict it as a column in an
abstract the three-dimensional space span by the orthonormal state vectors |1, −1i, |1, 0i, |1, 1i:
 
i 1
hl, m|ψi = √  0  . (4.51)
2 1

• Additional reading: Merzbacher, “Quantum Mechanics” 3rd edition, 9.3–9.4.

§3. Eigenvalue problem in discrete represenation

Eigenvalue problem for operator F̂ reads

F̂ |F i = F |F i , (4.52)

where F are eigenstates and |F i are the corresponding eigenstate vectors. In a continuous representation F̂ is a
differential operator, so that the eigenvalue problem takes form of a differential equation. For example, in the q-
representation

F̂ ψF (q) = F ψF (q) . (4.53)

To derive the eigenvalue problem in a discrete representation, start with (4.27) and comparing (4.52) with (4.19)
make the following substitutions in (4.27): |ai = |F i and |bi = F |F i. We have:
X
F hEm |F i = Fmn hEn |F i . (4.54)
n

This can be cast in a slightly more convenient form


X
{Fmn − F δmn } hEn |F i = 0 , (4.55)
n

which alludes to the fact that (4.55) is a system of algebraic linear homogeneous equations. Here ψF (En ) = hEn |F i
is the eigenfunction of F̂ in the E-representation. System of equations (4.55) has a non-trivial solution iff

det {Fmn − F δmn } = 0 . (4.56)

Since this determinant is infinite-dimensional, it has infinitely many solutions F1 , F2 , . . . which are the eigenvalues of
F̂ . Once Fn ’s are known, substitute them into (4.55) to determine the wave functions hEn |F i.
It is convenient to represent functions hEn |F i as a column
 
hE1 |F i
 hE2 |F i  . (4.57)
...

50
§4 Unitary operators IV REPRESENTATION THEORY

Then (4.56) can be written as



F11 − F F12 F13 ...

F21 F22 − F F23 ...
=0 (4.58)
F31 F32 F33 − F ...
... ... ... ...
Thus, the eigenvalue problem in a discrete representation is equivalent to the matrix diagonalization problem.
As has been mentioned above, any operator is diagonal in its own representation:
Fmn = hFm |F̂ |Fn i = Fn hFm |Fn i = Fn δmn . (4.59)

∗ Example: We know the harmonic oscillator Hamiltonian and its eigenfunctions ψn (x) = hx|ni in the coordinate
representation. Let us write the Hamiltonian in the energy representation. To this end we need to compute the matrix
elements (p2 )mn and (x2 )mn , where
Z +∞
2 ∗
(p )mn = ψm (x)p̂2 ψn (x)dx , (4.60)
−∞
Z +∞

(x2 )mn = ψm (x)x2 ψn (x)dx . (4.61)
−∞

To compute the integral in (4.60) we use an identity


r
∂ψn mω √ √ 
= nψn−1 − n + 1ψn+1 , (4.62)
∂x 2~
which follows from (3.118) and (3.96). Using it the second time we get
∂ 2 ψn mω p p 
= n(n − 1)ψ n−2 − (2n + 1)ψ n−2 ψ n + (n + 1)(n + 2)ψ n+2 , (4.63)
∂x2 2~
Thus,
Z +∞

(p2 )mn = −~2 ψm (x)ψn00 (x)dx (4.64)
−∞
~mω hp p i
=− n(n − 1)δm,n−2 − (2n + 1)δm,n + (n + 1)(n + 2)δm,n+2 . (4.65)
2
Similarly, using (3.115) we get
~ hp p i
(x2 )mn = n(n − 1)δm,n−2 + (2n + 1)δm,n + (n + 1)(n + 2)δm,n+2 . (4.66)
2mω
The matrix elements of the Hamiltonian are
 
(p2 )mn mω 2 (x2 )mn 1
Hmn = + = ~ω n + δm,n = En δmn . (4.67)
2m 2 2
Eq. (4.67) is a particular case of (4.59): the Hamiltonian is diagonal in the energy representation.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §11,23.

§4. Unitary operators

Transformation of the wave function between the representations can be thought of as an action of an certain oper-
ator Û (10 . Consider, for example a transformation from a discrete F -representation to the discrete E-representation:
X X
ψ(F ) = hFm |ψi = hFm |En ihEn |ψi = Umn hEn |ψi = Û ψ(E) . (4.68)
n n

(10 Don’t confuse it with the operator of potential energy.

51
§4 Unitary operators IV REPRESENTATION THEORY

The second equation is a definition of Umn ’s. In particular, we can take |ψi = |Fk i in which case (4.68) becomes
X X X †

hFm |Fk i = δmk = Umn hEn |Fk i = Umn Ukn = Umn Unk , (4.69)
n n n

In matrix notation this reads as

Û Û † = 1̂ , (4.70)

or

Û † = Û −1 . (4.71)

Such operators are called unitary, see II §4. This definition can be generalized to continuous representations as well,
in which case matrix elements Umn become a kernel of an integral transformation.
To find how an arbitrary operator P̂ transforms from E to F -representation consider its action on an arbitrary
state: ψ 0 (E) = P̂ (E)ψ(E). Substituting (4.68) we get

Û −1 ψ 0 (F ) = P̂ (E)Û −1 ψ(F ) ⇒ ψ 0 (F ) = Û P̂ (E)Û −1 ψ(F ) . (4.72)

So, if we define operator P̂ (F ) in F -representation as ψ 0 (F ) = P̂ (F )ψ(F ), then

P̂ (F ) = Û P̂ (E)Û −1 . (4.73)

This guarantees that the expectation values of physical quantities are the same in different representations:

hP (F )i = hψ(F )|P̂ (F )|ψ(F )i = hψ(E)|Û † P̂ (F )Û |ψ(E)i = hψ(E)|P̂ (E)|ψ(E)i = hP (E)i . (4.74)

We see, that every physical quantity can be represented by an infinite number of operators, differing from one
another by unitary transformations.

 Transformation between different representations in the Hilbert space is of course not the only example of unitary
operators. For example, transformation between difference reference frames in Euclidean space and time evolution
are also described by the unitary operators as we saw in II §5 and II §4. The main property of a unitary operator is
that it preserves the “lengths” of vectors and the “angles” between them. In the Euclidean space these are literally
lengths and angles, in the Hilbert space these are normalization and scalar products of the state vectors.
Here are some properties of the unitary transformations.

1. Invariance of the scalar product:

hψ(E)|ψ 0 (E)i = hψ(F )|ψ 0 (F )i . (4.75)

2. Invariance of matrix elements

hψm (F )|P̂ (F )|ψn (F )i = hψm (E)|P̂ (E)|ψn (E)i (4.76)

can be proved as (4.74).


3. Invariance of trace:

tr{P̂ (F )} = tr{Û P̂ (E)Û −1 } = tr{P̂ (E)Û −1 Û } = tr{P̂ (E)} . (4.77)

4. Invariance of the eigenvalue problem P̂ (E)ψ(E) = P ψ(E)

Û −1 P̂ (F )Û Û −1 ψ(F ) = P̂ Û −1 ψ(F ) ⇒ P̂ (F )ψ(F ) = P ψ(F ) . (4.78)

Notice that eigenvalues P are the same in all representations.


5. If operator P̂ (E) is Hermitian, then operator P̂ (F ) is also Hermitian. Indeed,

P̂ (F ) = Û P̂ (E)Û −1 ⇒ P̂ † (F ) = (Û −1 )† P̂ † (E)Û † = Û P̂ (E)Û † = P̂ (F ) . (4.79)

52
§5 Schrödinger equation in momentum representation IV REPRESENTATION THEORY

6. Transformation of commutation relations:


−1 −1 −1 −1 −1 −1
| {z Û} P̂ (E) Û
[P̂ (E), Q̂(E)] = Û | {z Û} − |Û {z Û} Q̂(E) |Û {z Û} P̂ (E) |Û {z Û}
| {z Û} Q̂(E) Û
1̂ 1̂ 1̂ 1̂ 1̂ 1̂
−1 −1 −1
= Û P̂ (F )Q̂(F )Û − Û Q̂(F )P̂ (F )Û = Û [P̂ (F ), Q̂(F )]Û . (4.80)

7. Any unitary operator can be represented as

Ŝ = eiF̂ , with F̂ † = F̂ , (4.81)

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §12, Merzbacher,
“Quantum Mechanics” 3rd edition, 9.4-9.6, 10.1-10.4.

§5. Schrödinger equation in momentum representation

Eigenvalue problem for the Hamiltonian can be formulated in the momentum representation. Formally, we can use
(4.44). However, it is not very convenient, for although the kinetic term p̂2 /2m is a constant, U may be a complicated
function of the operator ∂p . We need an explicit form of the U operator in the momentum space. I will show how to
derive it in one dimension. Generalization to three dimensions is straightforward.
We start with the Schrödinger equation in the form that does not refer to any particular representation:
Ĥ|ψi = E|ψi . (4.82)
R
Expand |ψi = dp0 |p0 ihp0 |ψi:
Z Z
0 0 0
dp Ĥ|p ihp |ψi = E dp0 |p0 ihp0 |ψi . (4.83)

Now multiply on the left by the bra hp|:


Z Z
dp0 hp|Ĥ|p0 ihp0 |ψi = E dp0 hp|p0 ihp0 |ψi . (4.84)

The matrix element of the Hamiltonian reads


p2
hp|Ĥ|p0 i = δ(p0 − p) + hp|Û |p0 i . (4.85)
2m
where I used hp|p0 i = δ(p − p0 ). We calculate the matrix element of U as follows
Z Z Z Z
hp|Û |p0 i = dx0 dxhp|xihx|Û |x0 ihx0 |p0 i = dx0 dxhp|xiU (x)δ(x0 − x)hx0 |p0 i
Z
1 0
= U (x)ei(p −p)x/~ dx . (4.86)
2π~
Assembling everything together and recalling that ϕ(p) = hp|ψi is the wave function in the momentum representation
we arrive at the Schrödinger equation in the p-representation
Z +∞
p2
ϕ(p) + hp|Û |p0 iϕ(p0 )dp0 = Eϕ(p) . (4.87)
2m −∞

∗ Examples.

1. Find bound state in the delta-well U = −αδ(x).


Note that the potential is an even function of coordinate, so we expect that the eigenstates can have definite
parity.
Z
1 0 α
hp|Û |p0 i = eix(p −p)/~ (−α)δ(x)dx = − . (4.88)
2π~ 2π~

53
§5 Schrödinger equation in momentum representation IV REPRESENTATION THEORY

Substituting into (4.87) we get

p2 α
ϕ(p) − C = Eϕ(p) , (4.89)
2m 2π~
where I denoted
Z +∞
C= ϕ(p)dp . (4.90)
−∞

Solution to (4.89) is
mα C
ϕ(p) = . (4.91)
π~ p2 + 2m|E|
I used that E = −|E| < 0. Now, inserting this into (4.90) and taking the integral yields
r
α m
C= C. (4.92)
~ 2|E|
Hence, the only bound state is
mα2
E=− . (4.93)
2~2
R
Normalization is fixed from ϕ2 (p)dp = 1 as
r
2πmα
C= . (4.94)
~

2. Find bound states in the following potential well:

U = −α[δ(x − a) + δ(x + a)] , α > 0. (4.95)

First we compute the Fourier image of the potential


Z
1 0 α h ia(p0 −p)/~ 0
i
hp|Û |p0 i = eix(p−p )/~ (−α)[δ(x − a) + δ(x + a)] = − e + e−ia(p −p)/~ . (4.96)
2π~ 2π~
Then write down the Schrödinger equation
p2 α h iap/~ i
ϕ(p) − e C+ + e−iap/~ C− = Eϕ(p) , (4.97)
2m 2π~
where
Z +∞
0
C± = e∓iap /~ ϕ(p0 )dp0 . (4.98)
−∞

Denote
2mE mα
κ2 = − > 0, α̃ = . (4.99)
~2 ~2
Then
α̃~  iap/~  1
ϕ(p) = e C+ + e−iap/~ C− 2 . (4.100)
π p + ~2 κ2
Substituting in (4.98) we derive a system of two linear equations with respect to C± :
α̃ 
C+ = C+ + e−2κa C− , (4.101)
κ
α̃ −2κa 
C− = e C+ + C− . (4.102)
κ

54
§6 Occupation number representation of harmonic oscillator IV REPRESENTATION THEORY

There is a non-trivial solution iff



1 − α̃κ − α̃κ e−2κa

= 0. (4.103)
− α̃ e−2κa 1 − α̃
κ κ

This is equivalent to the requirement that


α̃ α̃ 
1− = ± e−2κa , ⇒ κ = α̃ 1 ± e−2κa . (4.104)
κ κ
Solution to this equation gives the energy levels of the system.
Consider first the 0 +0 solution:

κ = α̃ 1 + e−2κa . (4.105)

We have
1 
C+ = −2κa
C+ + e−2κa C− ⇒ C+ = C− ⇒ ϕ(p) = ϕ(−p) . (4.106)
1+e
We thus found an even stationary state, which is the only even state of this system. It is also the ground state,
because the ground state cannot be odd.
In the limit κa  1 the two δ-wells are very close:

2mα2
κ ≈ 2α̃ ⇒ E+ = − . (4.107)
~2
In the opposite limit κa  1 we have e−2κa  1, so that in the first approximation κ ≈ α̃. In the second
approximation κ = α̃(1 − e−2α̃a ). Therefore,

mα2 
E+ = − 1 + e−2α̃a . (4.108)
2~2

The 0 −0 solution is a problem in one of the home assignments.

• Additional reading: Merzbacher, “Quantum Mechanics” 3rd edition, 2.2, 2.5.

§6. Occupation number representation of harmonic oscillator

In III §5B we represented the harmonic oscillator in terms of the ladder operators. The raising operator ↠moves
the oscillator up to the next excited state, while the lowering operator â moves it down. We can give a different
interpretation to these operators. Instead of counting the quantum states (ground state, first excited state, etc.) we
will count the number of quantum excitations, or simply quanta. Each quantum excitation of harmonic oscillator has
energy E = ~ω, see (3.112). For example, harmonic quantum excitation of the lattice structure of a solid body is
phonon, harmonic excitation of the electromagnetic field is photon. We will say that the n’th state contains n quanta
and denote it by the ket |ni. The physical meaning of the ladder operators is adding or removing a single quantum
to the harmonic oscillator:

â|ni = n|n − 1i , (4.109)


â |ni = n + 1|n + 1i . (4.110)

For this reason â and ↠are also called the annihilation and creation operators.
We can write the excitation number operator n̂ in terms of the ladder operators:

n̂ = ↠â . (4.111)

55
§6 Occupation number representation of harmonic oscillator IV REPRESENTATION THEORY

Indeed, it is easy to check that

n̂|ni = n|ni . (4.112)

Hamiltonian of the harmonic oscillator is

Ĥ = ~ω(↠â + 1/2) = ~ω(n̂ + 1/2) . (4.113)

Thus, states |ni are eigenstates of the Hamiltonian and the excitation number operator. State |0i is the ground state
which we normalize as usual h0|0i = 1. The excited states can be obtained by a successive application of the creation
operator
1
|ni = √ (↠)n |0i . (4.114)
n!
You can verify that hn|ni = 1.
State vectors |ni form a complete set of states because they are eigenstates of the Hamiltonian. Therefore, instead
of the coordinate representation that we discussed in III §5 one can introduce the occupation number representation.
For any state
X
|ψi = |nihn|ψi , (4.115)
n

where |hn|ψi|2 is interpreted as a probability to have n excitations. If a wave function is known in the coordinate
representation, then (4.115) implies that
X
hq|ψi = hq|nihn|ψi , (4.116)
n

In particular, a wave function describing a single particle in one dimension can be expanded as

X ∞
X
ψ(x) = hx|ψi = hx|nihn|ψi = Cn ψn (x) , (4.117)
n=0 n=0

where ψn (x) are eigenfunctions of the harmonic oscillator given by (3.111). The probability to have n excitations in
the state |ψi is |Cn |2 .
In matrix notation the state vectors |ni can be represented as follows
     
1 0 0
 0   1   0 
     
|0i =  0 , |1i =  0 , |2i =  1 , etc (4.118)
 0   0   0 
... ... ...

Ladder operators become matrices amn = hm|â|ni and a†mn = hm|↠|ni:
 
 √  0 0 0 ...
0 1 √0 0 . . . √
 1 0 0 ... 
 0 0 2 √0 . . .   √ 
â = 
 0 0 0
,
 â†
=  0
 2 √0 . . . 
 (4.119)
3 ...  0 0 3 ... 
... ... ... ... ...
... ... ... ...

By the way, we see that (↠)† = â as required. The number operator is a diagonal matrix
 
0 0 0 ...
 0 1 0 ... 
n̂ =  (4.120)
0 0 2 ... 
... ... ... ...

56
§6 Occupation number representation of harmonic oscillator IV REPRESENTATION THEORY

 Although the ladder operators are not Hermitian and therefore do not correspond to any physical quantity, their
eigenstates have clear physical interpretation. The eigenvalue problem for harmonic oscillator reads

â|αi = α|αi . (4.121)

In the coordinate representation:

âψα (x) = αψα (x) . (4.122)

Employing (3.122) we can write it as a differential equation


r
~ 0 2~
ψ + (x − x0 )ψ = 0 , with x0 = α . (4.123)
mω mω
Notice, that α is a complex number. Replacing x by x0 = x − x0 we obtain the same equation as in (3.128). Its
solution is given by (3.129):
 mω 1/4 ∗ 2
ψα (x) = emω(x0 −x0 )x0 /4~ e−mω(x−x0 ) /2~
. (4.124)
π~
R

It satisfies the normalization condition |ψα |2 dx = 1.(11 We can write it in terms of the expectation value (∆x)2 =
~/(2mω) as follows:
1
ψα (x) = e(x0 −x0 )x0 /8h(∆x) i /4h(∆x)2 i
∗ 2 2

1/4
e−(x−x0 ) . (4.125)
(2π h(∆x)2 i)

This is precisely the wave function of a coherent state generalized to complex x0 , see (1.132) with p0 = 0. We observe
that the eigenstates of the annihilation operator are the coherent states.
What is the probability that a coherent state contains n excitations of the harmonic oscillator? To answer this
question we can use (4.117) and expand (4.125) into the harmonic oscillator eigenstate basis:

X
ψα (x) = Cn ψn (x) , (4.126)
n=0

where ψn are given by (3.111). Coefficients Cn can be computed using the orthonormality of the ψn ’s:
Z +∞
Cn∗ = ψn∗ (x)ψα (x)dx (4.127)
−∞
 mω 1/2 Z +∞  r 
∗ 1 mω 2 2
= emω(x0 −x0 )x0 /4~ √ Hn x e−mωx /(2~) e−mω(x−x0 ) /2~ dx
π~ 2n n! −∞ ~
Z +∞ √
∗ 1 2
/2−(ξ− 2α)2 /2
= e(α−α )α/2
√ Hn (ξ)e−ξ dξ (4.128)
n
2 n!π −∞
Z +∞ 2
∗ 1 2 dn e−ξ −ξ2 +√2αξ−α2
= e(α−α )α/2
√ (−1)n eξ e dξ , (4.129)
2n n!π −∞ dξ n

where I used (3.106). Integrating by parts n times we get


2
√ Z +∞
∗ e−α 2

Cn∗ = e(α−α √ )α/2
( 2α)n e−ξ e 2αξ dξ (4.130)
n
2 n!π −∞
2
∗ e −α √ 2 √ αn 2
= e(α−α )α/2 √ ( 2α)n eα /2 π = √ e−|α| /2 . (4.131)
n
2 n!π n!

(11 Function (4.124) can be multiplied by any phase. I used this freedom to fix the phase so that the coefficients Cn in (4.126) have simple
form (4.131).

57
§6 Occupation number representation of harmonic oscillator IV REPRESENTATION THEORY

Thus, the probability to find a coherent state |αi with n harmonic oscillator excitations is
n
|α|2n −|α|2 hni −hni
Pn = |Cn |2 = e = e , (4.132)
n! n!
where hni = hα|n̂|αi = |α|2 is the average number of excitations in the given coherent state. Distribution (4.132)
is the Poisson distribution, which gives the probability to have n independent excitations in a coherent state with
average excitation number hni. Independence of excitations is closely related to the fact that the coherent state is as
close to the classical limit as a quantum system can be.

• Additional reading: Merzbacher, “Quantum Mechanics” 3rd edition, 10.6.

58
§1 General properties of motion in central potential V MOTION IN CENTRAL POTENTIAL

V. MOTION IN CENTRAL POTENTIAL

§1. General properties of motion in central potential

Motion in central potential U (r) is described by the Hamiltonian:

~2 ∇2
Ĥ = − + U (r) , (5.1)
2M
where r is a distance form the center, M is mass of the particle. Laplacian in spherical coordinates
     
2 1 ∂ 2 ∂ 1 1 ∂ ∂ 1 ∂2
∇ = 2 r + 2 sin θ + (5.2)
r ∂r ∂r r sin θ ∂θ ∂θ sin2 θ ∂φ2

Using (1.105) and (5.2) in (5.1) we have


 
~2 ∂ ∂ L̂2
Ĥ = − r2 + + U (r) . (5.3)
2M r2 ∂r ∂r 2M r2

Since U (r) does not depend on angles [Ĥ, L̂2 ] = [Ĥ, L̂z ] = 0. Thus, systems in central potential can be in stationary
states with given values of L2 and Lz . Eigenfunctions of L̂2 and L̂z in coordinate representation are spherical harmonics
Ylm (θ, φ). The corresponding eigenvalues are L2 = ~2 l(l+1), l = 0, 1, 2 . . . and Lz = ~m, m = 0, ±1, . . . , ±l. Therefore,
we can look for a solution to the Schrödinger equation equation Ĥψ = Eψ in the form

ψElm (r, θ, φ) = REl (r)Ylm (θ, φ) . (5.4)

In place of the radial function R it is convenient to introduce a different function

χ(r) = rR(r) . (5.5)

Schrödinger equation becomes


 
~2 d 2 χ ~2 l(l + 1)
− + U (r) + χ = Eχ . (5.6)
2M dr2 2M r2

Since R(r) must be finite at r → 0, we conclude that

lim χ(r) = 0 . (5.7)


r→0

Normalization condition for the wave function of the discrete spectrum:


Z Z Z

ψElm (r, θ, φ)ψE 0 lm (r, θ, φ)d3 r = δEE 0 REl
2 2
r dr = δEE 0 χ2El (r)dr = δEE 0 , (5.8)

while for the continuous spectrum:


Z Z ∞
∗ 3
ψklm (r, θ, φ)ψk0 lm (r, θ, φ)d r = χk0 (r)χk (r)dr = δ(k − k 0 ) , (5.9)
0

where k 2 = 2M E/~2 ≥ 0 as usual.


Some properties of the Schrödinger equation with the central potential:

1. For each l there is (2l + 1)-fold degeneracy corresponding to the values of m. States with l = 0, 1, 2, 3 . . . are
denoted by letters s, p, d, f, . . ..
2. Since [Ĥ, P̂ ] = 0, all stationary states can be chosen to have definite parity. In fact the spherical harmonics
have definite parity, see (2.103). Thus, states with even (odd) l are parity-even (odd).

59
§1 General properties of motion in central potential V MOTION IN CENTRAL POTENTIAL

3. Eq. (5.6) looks like the Schrödinger equation for one dimensional motion in the effective potential

~2 l(l + 1)
Ul = U (r) + , (5.10)
2M r2
with boundary condition (5.7). Hence, we can use the results of III §1 for one dimensional motion. In particular,
the energy spectrum is non-degenerate.
4. Suppose that the potential is such that U → 0 as r → ∞ and consider particle with negative energy. Such
particle is constrained to move in a volume around the origin and its energy spectrum is discrete. Multiplying
(5.6) by χ and integrating
Z ∞ Z ∞ Z ∞
~2
− χ00 χdr + Ul χ2 dr = Eχ2 dr . (5.11)
2M 0 0 0

Since at χ → 0 as r → ∞ we can integrate the first term by parts


Z ∞ Z ∞
− χ00 χdr = (χ0 )2 dr , (5.12)
0 0

which yields for the average energy


Z ∞    
~2 0 2 l(l + 1) 2M 2
E= (χ ) + + 2 U (r) χ (r) dr . (5.13)
2M 0 r2 ~

Suppose that the potential is U (r) = −A/rn at r ≤ a and vanishes otherwise; A > 0 and n > 0. Let us make
an order of magnitude estimate of the integral in (5.13). By uncertainty relation χ0 ∼ χ/a, implying that
 
~2 1 + l(l + 1) 2M A
E∼ − 2 n . (5.14)
2M a2 ~ a

If n > 2, then E → −∞ is at a = 0, which describes “falling onto the center”. However, if n < 2, then the
minimum of E is at a finite a, implying that no falling onto the center occurs. In this case the energy spectrum
starts with the finite negative value. Recall that in the classical mechanics falling onto the center happens at
any n > 0.

FIG. 7: falling onto the center happens at n < 2.

5. According to the oscillation theorem, if for a given l we arrange the energy eigenvalues in order of increasing
magnitude El0 < El1 < El2 < . . . and assign the corresponding states the radial quantum number nr = 0, 1, 2, . . .,
then the number of zeros of χ(r) in the interval 0 < r < ∞ is nr , not counting zero at r = 0. Since 1/r > 0,
R(r) has the same number of zeros as χ(r).

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §32,18, Merzbacher,
“Quantum Mechanics” 3rd edition, 12.1.

60
§2 Spherical waves V MOTION IN CENTRAL POTENTIAL

§2. Spherical waves

Free motion of a particle with given l and m is called the spherical wave. The corresponding solutions of Schrödinger
equation can be found form (5.6) with U = 0:
d2 χl l(l + 1)
− χl + k 2 χl = 0 . (5.15)
dr2 r2
Consider first s-states l = 0:
χ000 + k 2 χ0 = 0 ⇒ χ0 (r) = A sin(kr) + B cos(kr) . (5.16)
In view of the boundary condition (5.7), B = 0. Substituting into (5.9) we get
Z ∞ Z 0 0
A2 +∞ eik r − e−ik r eikr − e−ikr
A2 sin(k 0 r) sin(kr)dr = dr
0 2 −∞ 2i 2i
A2 πA2
= 2
{−2δ(k 0 − k) + 2δ(k 0 + k) }2π = δ(k 0 − k) = δ(k 0 − k)
2(2i) | {z } 2
=0 (k>0, k0 >0)
r
2
⇒ A= . (5.17)
π
Therefore,
r
χk0 (r) 2 sin(kr)
Rk0 (r) = = . (5.18)
r π r
In general, for l 6= 0 (5.15) has a solution in terms of the spherical Bessel and Neumann functions jl and ηl :
Rkl (r) = Ajl (kr) + Bηl (kr) . (5.19)
Spherical Bessel functions can be expressed through the Bessel functions of half-integer order:
r  l
π l d sin z
jl (z) = Jl+1/2 (z) = (−1) , (5.20)
2z dz z
r
π
ηl (z) = (−1)l+1 J−l−1/2 (z) , (5.21)
2z
Asymptotic behavior of these functions:

 zl
,
(2l+1)!! zl
jl (z) ≈   (5.22)
 1 cos z − π (l + 1) , z  l
z 2

 − (2l−1)!!
z l+1
, zl
ηl (z) ≈   (5.23)
 1 sin z − π (l + 1) , z  l
z 2

Evidently, only jl satisfies the boundary condition (5.7). Therefore, the general solution to the Schrödinger equation
for a free particle with given l and m reads
ψklm (r) = Ajl (kr)Ylm (θ, φ) . (5.24)
2
To understand the qualitative features of the radial wave function seen in (5.22),
p note that k can be neglected as
2
compared to the effective potential Ul = l(l + 1)/2M r in (5.15) when r  rl = l(l + 1)/k. In this case
d2 χl l(l + 1)
− χl = 0 ⇒ χl (r) ≈ Arl+1 , r  rl . (5.25)
dr2 r2
In the opposite case r  rl , we can neglect the effective potential:
χ00l (r) + k 2 χl (r) = 0 ⇒ χl (r) ≈ B sin(kr + δl ) , r  rl . (5.26)

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §33, Merzbacher,
“Quantum Mechanics” 3rd edition, 12.2.

61
§2 Spherical waves V MOTION IN CENTRAL POTENTIAL

§3. Spherical potential well

Consider motion of a particle of mass M in an infinitely deep spherically symmetric potential well:
(
0, r ≤ a,
U (r) = (5.27)
∞, r > a.

Solution inside the well is given by (5.24). At r ≥ a the wave function vanishes. Thus ψl (ka) = 0. Let me denote zeros
of the spherical Bessel function of the l’th order by Xlnr . The radial quantum number nr = 0, 1, 2, . . . enumerates
zeros, starting from the smallest one. For example, the lowest four zeros are

Xs0 ≡ X00 = π , Xp0 ≡ X10 = 4.493 , Xd0 ≡ X20 = 5.763 , Xs1 ≡ X01 = 6.283 . (5.28)

The energy levels are follow from ka = Xlnr :

~2 Xln
2
Enl = r
. (5.29)
2M a2

FIG. 8: Spherical well of finite depth.

Energy spectrum of a particle in an infinitely deep potential well consists of infinite number of discrete energy levels
(5.29). However, a well of finite depth contains only a finite number of bound states. A spherical well of finite depth
is described by the potential
(
−U0 , r ≤ a ,
U (r) = (5.30)
0, r > a,

and is depicted in Fig. 8. Bounds states have energies in the interval −U0 ≤ E ≤ 0. Schrödinger equation equation
for the s-state reads (l 6= 0 states can be discussed along the same lines but involve a more tedious algebra):
2M
χ00 + (U0 − |E|)χ = 0 , r ≤ a , (5.31)
~2
2M
χ00 − 2 |E|χ = 0 , r ≥ a (5.32)
~
Denote
2M 2M
k̃ 2 = (U0 − |E|) , κ2 = |E| . (5.33)
~2 ~2
Solution to (5.31) and (5.32) read:

χ = A sin(k̃r) , r < a, (5.34)


−κr
χ = Be , r > a. (5.35)

The logarithmic derivative χ0 /χ must be continuous across r = a implying that

k̃ cot(k̃a) = −κ . (5.36)

62
§3 Spherical potential well V MOTION IN CENTRAL POTENTIAL

Solutions to this transcendental equation are a set of discrete numbers k̃n corresponding to the energy levels in the
well. Notice that k̃ and κ are related by (5.33) as follows:
2M U0
k̃ 2 + κ2 = . (5.37)
~2
So, geometrically, k̃n0 s lie at the intersections of curves given by (5.36) and (5.37), see Fig. 9.

κa
2

0
0 π
π 3π 2π 5π
2 2 2
˜
ka

FIG. 9: Solution to (5.36) is shown in blue. Orange (no bound states), green (single bound state) and red (two bound states)
lines correspond to solution of (5.37) with 2M U0 /~2 = 1, 5, 25.

It is seen that there exist a solution only if k̃a ≥ π/2. Since |E| ≥ 0 this condition translates into a2 U0 ≥ π 2 ~2 /8M .
The first excited state exists if k̃a ≥ 3π/2 etc.

• Additional reading: Merzbacher, “Quantum Mechanics” 3rd edition, 12.3.

§4. Spherical harmonic oscillator

Schrödinger equation for a particle moving in the three-dimensional harmonic oscillator potential
M ω2 r2
U (r) = , (5.38)
2
can be solved by separation of variables in Cartesian, spherical and parabolic coordinates.(12 Spherical harmonic
oscillator in Cartesian coordinates is one of the home assignments; here we will discuss it in spherical coordinates.
As has been explained in IV §4, energy eigenvalues, being an observable quantity, are the same in all representations,
while eigenfunctions depend on a representation.
Schrödinger equation for a particle in potential (5.38) in spherical coordinates reads
 
~2 d2 M ω2 r2 ~2 l(l + 1)
− + + − Enl χnl (r) = 0 . (5.39)
2M dr2 2 2M r2
Introduce a length parameter
r
~
a= (5.40)

and dimensionless quantities ξ = r/a and ε = E/~ω. Then (5.39) takes the following form that contains no dimensional
parameters:
 2 
d 2 l(l + 1)
−ξ − + 2ε χnl (ξ) = 0 . (5.41)
dξ 2 ξ2

(12 Parabolic coordinates are helpful when discussing spherically symmetric systems in electric field.

63
§5 Coulomb potential V MOTION IN CENTRAL POTENTIAL

Its solution can be expressed in terms of the confluent hypergeometric function F as follows
2
χnl (ξ) = Nnl e−ξ /2 l+1
ξ F (−nr , l + 3/2, ξ 2 ) , (5.42)
where
 
ε 1 3
nr = − l+ . (5.43)
2 2 2
The only property of the confluent hypergeometric function that we need to know is that χ → 0 when ξ → ∞ only if
nr = 0, 1, 2, . . .. This gives energy spectrum as
 
3
Enr l = ~ω 2nr + l + , nr , l = 0, 1, 2, . . . (5.44)
2
with the corresponding eigenfunctions
1
ψnr lm = χn l (r)Ylm (θ, φ) . (5.45)
r r
Energy levels (5.44) depend only on a combination of quantum numbers 2nr + l, not on nr and l independently.
Quantum number n = 2nr + l is called the principal quantum number. Eigenstates of the spherical harmonic oscillator
are often labeled as (n + 1)l. For example, 1s state has n = 0, l = 0, 1p state has n = 0, l = 1, 2s state has n = 1,
l = 0 etc.
Energy levels are labeled as En indicating that they are degenerate with respect to m and with respect to nr and
l that add up to the same n. The ground state n = 0 is the only non-degenerate state. All excited states n ≥ 1 are
degenerate with the following degree (home assignment):
1
g(n) = (1 + n)(2 + n) . (5.46)
2
For instance, E2 = (7/2)~ω is six-fold degenerate. One s state with l = 0, nr = 1 and five d-states l = 2, nr = 0, m =
±2, ±1, 0 have this energy. While degeneracy of states with different m and the same l is due to spherical symmetry,
degeneracy with different l’s is due to a symmetry between the coordinates and momentum in the Hamiltonian, see
V §7.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §, Merzbacher, “Quan-
tum Mechanics” 3rd edition,

§5. Coulomb potential

Schrödinger equation with Coulomb potential


e2 Z
U (r) = − (5.47)
r
has special physical significance as it describes motion of electron in the field of atomic nucleus. In such case −e is
electron charge and Z is the nucleus electric charge in units of e.

A. Discrete spectrum

It is conventional to introduce the atomic units: atomic unit of length


~2
a= , (5.48)
M e2
also called the Bohr radius and atomic unit of energy
e2 M e4
Ea = = 2 . (5.49)
a ~

64
§5 Coulomb potential V MOTION IN CENTRAL POTENTIAL

Since mass of proton is about two thousand times larger than that of electron, the reduced mass of hydrogen atom
equals electron mass M with very good accuracy. Using the known values of electron mass M = 0.51 MeV and the
fine structure constant (see XIV §7) e2 /~c ≈ 1/137 we find that a = 5.29 · 10−9 cm and Ea = 27.21 eV.
Introducing dimensionless quantities ρ = r/a and ε = E/Ea , Schrödinger equation for the radial component takes
form
 2 
d 2Z l(l + 1)
+ 2ε + − χ(ρ) = 0 . (5.50)
dρ2 ρ ρ2

Bound states have E < 0 so we can denote α2 = −2ε > 0:


 2 
d 2 2Z l(l + 1)
−α + − χ(ρ) = 0 . (5.51)
dρ2 ρ ρ2
At ρ → ∞ the terms containing the inverse powers of ρ can be neglected yielding the asymptotic solution
χ(ρ) ≈ Ae−αρ + Beαρ , ρ→∞ ⇒ B = 0. (5.52)
Now we are looking for a solution in the form χ(ρ) = e−αρ w(ρ), where w is unknown function that we expand in
power series as follows

X
w(ρ) = ργ βν ρ ν . (5.53)
ν=0

We also need to check the boundary condition at ρ → 0. In this case χ(ρ) ≈ ργ β0 which must satisfy (5.51). This
gives
:0
2Z 

γ(γ − 1)ργ−2 − α2
ργ
+ργ − l(l + 1)ργ−2 = 0 , (5.54)
 ρ
where I dropped terms that vanish faster than others. Equation γ(γ − 1) = l(l + 1) has two solutions
γ = l + 1, −l . (5.55)
Only the first of them is finite at ρ → 0. Thus, after applying the boundary conditions we have
X
χ(ρ) = e−αρ ρl+1 βν ρ ν . (5.56)
ν

Loading this in (5.51) produces the following recurrence relation


2[α(ν + l + 1) − Z]
βν+1 = βν . (5.57)
(ν + l + 2)(ν + l + 1) − l(l + 1)
In order that χ be finite at large ρ the power series in (5.56) must terminate at some ν = nr . This happens if
Z Z
α(nr + l + 1) − Z = 0 ⇒ α= = , (5.58)
nr + l + 1 n
where the principal quantum number is given by
n = nr + l + 1 . (5.59)
Energy levels read
α2 M (e2 Z)2 1
En = − Ea = − , n = 1, 2, 3, . . . (5.60)
2 ~2 2n2
It is evident from (5.60) that for a given n states with l = 0, 1, . . . , n − 1 and m = 0, ±1, ±2, . . . ± l are degenerate.
The degeneracy g(n) can be computed by summing the arithmetic series
n−1
X 1 + 2(n − 1) + 1
g(n) = (2l + 1) = n = n2 . (5.61)
2
l=0

65
§5 Coulomb potential V MOTION IN CENTRAL POTENTIAL

The normalized wave functions can be expressed through the associated Laguerre polynomials L as follows
" 3 #1/2  l  
χnl (r) 2 (n − l − 1)!  r  2r 2r
Rnl (r) = = exp − L2l+1
n−l−1 , (5.62)
r nā 2n(n + l)! nā nā nā

where I denoted ā = a/Z. A few lowest order radial functions:


2  r
R10 (r) = √ exp − , (5.63)
ā3 ā
1  r  r 
R20 (r) = √ 2− exp − , (5.64)
8ā3 ā 2ā
1 r  r 
R21 (r) = √ exp − , (5.65)
24ā3 ā 2ā
8 1 r r   r 
R31 (r) = √ 1− exp − . (5.66)
27 6ā3 ā 6ā 3ā

The normalization condition is


Z ∞
2 2
Rnl r dr = 1 . (5.67)
0

For the future reference let me list here several useful formulas involving expectations values in states (5.62)

1
2 n2
hρi = [3n2 − l(l + 1)] , ρ = [5n2 + 1 − 3l(l + 1)] , (5.68)
2Z 2Z 2

2 n2
−3 Z3
−1 Z
ρ = [5n2 + 1 − 3l(l + 1)] , ρ = 3 , ρ = 2. (5.69)
2Z 2 n (l + 1)(l + 1/2)l n

B. Continuous spectrum

Continuous spectrum lies at E > 0. Introducing k 2 = 2ε ≥ 0 we rewrite the Schrödinger equation (5.50) as follows
 2 
d 2 2Z l(l + 1)
+ k + − χ = 0. (5.70)
dρ2 ρ ρ2

At ρ → ∞ the asymptotic behavior of the solution is

χ(ρ) ≈ Aeikρ + Be−ikρ . (5.71)

Thus we are looking for a solution in the form


X
χ(ρ) = e±ikρ ρl+1 βν ρ ν . (5.72)
ν

Plugging into (5.71) we obtain

2[i(ν + l + 1)k − Z
βν+1 = βν . (5.73)
(ν + l + 2)(ν + l + 1) − l(l + 1)

This power series describes a confluent hypergeometric function. The final solution to the Schrödinger equation reads
 
±ikρ l+1 2
χkl (ρ) = e ρ F l + 1 ± ; 2l + 2, ∓2ikρ . (5.74)
ik

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §36.

66
§6 Effective electric potential of the hydrogen atom V MOTION IN CENTRAL POTENTIAL

§6. Effective electric potential of the hydrogen atom

In the previous section we discussed motion of electron in the Coulomb potential neglecting the fact that electron,
being electrically charged particle, itself creates electromagnetic field. On the one hand, electron wave function ψ and
hence electric charge density of atom ρ is determined by the Schrödinger equation

p̂2
ψ + U (r)ψ = Eψ . (5.75)
2M
On the other hand, potential U obeys the Poisson equation

∇2 U = −4π(−e)ρ . (5.76)

In the case of the hydrogen atom Z = 1, we can write the electric charge density of atom as a sum of two terms

ρ(r) = eδ(r) − e|ψ(r)|2 . (5.77)

The first term on the right-hand-side of (5.77) describes the charge density of proton (whose size is negligible compared
to a), while the second term describes the electron “cloud”. In order to compute ψ we need to know U in (5.75), while in
order to compute U we need to know ψ in (5.76). We thus arrived at the system of coupled equations (5.75),(5.76) that
must be solved simultaneously. In this section I will show how to solve this problem using consecutive approximations
in the case of the hydrogen atom.
At first, we will neglect the electron cloud contribution in (5.77). Then, solution to the Poisson equation is
U = −e2 /r, viz. the Coulomb potential. Solution to (5.75) with such potential has been discussed in the previous
section. The bound state solutions are a system of wave functions ψnlm (r). The second step is to substitute these
wave functions into (5.76) and calculate the potential U . Clearly, the result depends on electron state. For illustration,
suppose that electron is in the ground state, viz.

ψ100 (r) = 2a−3/2 e−r/a Y00 (θ, φ) . (5.78)

Poisson equation
 
2 1 ∂ 2 ∂
∇ U= 2 r U = −4π(−e)ρ ⇒ (5.79)
r ∂r ∂r

In place of U it is convenient to introduce an auxiliary function f defined such that U = −ef /r:
 
1
f 00 = −4πr eδ(r) − 4ea−3 e−2r/a . (5.80)

At r > 0, the delta-function does not contribute, so that
4re −2r/a
f 00 = e . (5.81)
a3
R
At large distances U must decrease at least as fast as 1/r2 . This is because atom is neutral ( ρd3 r = 0) and thus the
monopole contribution to the multipole expansion vanishes. Thus, the boundary conditions at r → ∞ are f → 0 and
f 0 → 0. Integrating (5.81) twice we get
Z Z Z
4e ∞ 0 ∞ 00 −2r00 /a 00 4e ∞ 0 1 0 e
f (r) = 3 dr r e dr = 3 dr a(a + 2r0 )e2r /a = (a + r)e−2r/a . (5.82)
a r r0 a r 4 a
Finally,
 
1 1
U (r) = −e2 + e−2r/a . (5.83)
r a
We see that the Coulomb potential is modified at distances comparable to or larger than the Bohr radius a. Electron
cloud screens the proton potential.
One can continue the iteration process and solve the Schrödinger equation with potential (5.83) to find the modified
wave functions etc.

67
§7 Conservation laws and degeneracy of energy spectrum V MOTION IN CENTRAL POTENTIAL

§7. Conservation laws and degeneracy of energy spectrum

We have seen many times that there is a connection between the degeneracy of energy levels of a quantum system and
its symmetry. To make a general statement connecting degeneracy and symmetry consider two Hermitian operators
F̂ and K̂. Let [F̂ , K̂] 6= 0, but [F̂ , Ĥ] = [K̂, Ĥ] = 0, i.e. F̂ and K̂ correspond to conserved quantities, but are not
simultaneously measurable. Let ψE,F be a wave function of a stationary state: ĤψE,F = EψE,F and an eigenfunction
of F̂ : F̂ ψE,F = F ψE,F . This is possible because F̂ commutes with Ĥ.
Now, ψE,F is not an eigenfunction of K̂ because [K̂, F̂ ] 6= 0, i.e. K̂ψE,F ∝ψ
/ E,F . On the other hand,

Ĥ(K̂ψE,F ) = K̂(ĤψE,F ) = E K̂ψE,F . (5.84)

This indicates that K̂ψE,F is an eigenfunction of Ĥ. Thus, we come to conclusion that there are at least two different
eigenfunctions of Ĥ corresponding to the same level E: ψE,F and K̂ψE,F .
In summary, energy spectrum is degenerate if there are two or more operators that commute with the Hamiltonian
Ĥ, but not with each other. Symmetry implies degeneracy and degeneracy implies symmetry. Sometimes, symmetry
implied by a degenerate energy spectrum is not easily identifiable. In such cases it is sometimes misleadingly called
the accidental degeneracy (as in examples 3,4,5 below).

∗ Examples.

1. Free one-dimensional motion: Ĥ = p̂2 /2m. [Ĥ, p̂] = 0 and [Ĥ, P̂ ] = 0, where P̂ is the inversion (parity) operator.
However [p̂, P̂ ] 6= 0. Therefore, levels are two-fold degenerate. Indeed, there are two independent solution to the
Schrödinger equation: ψ = e±ikx .
2. For a spherically symmetric potential U (r) the Hamiltonian is invariant under rotations:

[Ĥ, L̂x ] = [Ĥ, L̂y ] = [Ĥ, L̂z ] = 0 . (5.85)


However,
[L̂i , L̂j ] = i~ijk L̂k 6= 0 . (5.86)
This is the source of the 2l + 1 degeneracy of l’th level. The corresponding energy levels Enl do not depend on
quantum number m.
3. Energy levels of spherical harmonic oscillator En do not depend on quantum number l (in addition to the 2l + 1
degeneracy of all central potentials). This is because operator
p̂i p̂k
T̂ik = + M ωr̂i r̂k (5.87)
M
commutes with the Hamiltonian
p̂2 M ωr2
Ĥ = + , (5.88)
2M 2
but does not commute with L̂2 . Conservation of T̂ik reflects symmetry between the coordinate and momentum
dependence of the Hamiltonian.
4. Energy spectrum of a particle in Coulomb potential also depends only on the principal quantum number n, not
on l and m. This is because operator
1   r̂
K̂ = 2
L̂ × p̂ − p̂ × L̂ + , (5.89)
2M e r
commutes with the Hamiltonian
p̂2 e2
Ĥ = − , (5.90)
2M r
but does not commute with other operators [L̂i , K̂j ] 6= 0 and [K̂i , K̂j ] 6= 0 if i 6= j. Symmetry corresponding to
(5.89) becomes explicit in the momentum representation.

68
§7 Conservation laws and degeneracy of energy spectrum V MOTION IN CENTRAL POTENTIAL

5. Three dimensional infinite rectangular potential well U (x, y, z) with sides a, b, c is given by
(
0 , x ∈ [−a/2, a/2], y ∈ [−b/2, b/2], z ∈ [−c/2, c/2];
U (x, y, x) = (5.91)
∞ , otherwise .

Solution to the Schrödinger equation can be found by separating variables in Cartesian coordinates:

ψn1 ,n2 ,n3 (r) = ψn1 (x)ψn2 (y)ψn3 (z) . (5.92)

The corresponding energy levels are, see III §3,


 
π 2 ~2 n21 n22 n23
En1 ,n2 ,n3 = + 2 + 2 , n1 , n2 , n2 = 1, 2, 3, . . . (5.93)
2m a2 b c

and the wave functions


q 
 2 cos πnx
, n = 1, 3, 5, . . .
ψn (x) = qa a
 (5.94)
 2 sin πnx
, n = 2, 4, 6, . . .
a a

If a 6= b 6= c each energy value corresponds to one wave function, i.e. there is no degeneracy. Actually, potential
U is invariant under rotations by π about the three coordinate axes and inversion. However, the corresponding
operators commute with each other. If a = b = c, then

π 2 ~2 2
En1 ,n2 ,n3 = (n + n22 + n23 ) , n1 , n2 , n2 = 1, 2, 3, . . . (5.95)
2ma2 1
Additional symmetry of this case is the symmetry of a cube, generated by rotations by π/2. Since rotations
about different axes do not commute, energy levels are degenerate. For example, quantum numbers (n1 , n2 , n3 ) =
2 2
(5, 1, 1), (1, 5, 1), and (1, 1, 5) correspond to the same level 27π
2ma2 . Interestingly, there is another level (3, 3, 3)
~

that has the same energy. This is another example of accidental degeneracy caused by a particular way in which
potential depends on the coordinates.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §36,§10, Merzbacher,
“Quantum Mechanics” 3rd edition, 12.5.

69
§1 Angular momentum operator VI ANGULAR MOMENTUM

VI. ANGULAR MOMENTUM

§1. Angular momentum operator

In I §4 and I §6 I introduced the orbital angular momentum operator L̂ and discussed some of its properties in the
coordinate representation. In particular, we have seen that its components satisfy the commutation relations (1.57).
In fact, a vector operator satisfying such commutation relations has much reacher mathematical structure than is
reflected by the coordinate representation, which can be used to describe only those physical quantities that have
classical analogues. This is why in this section we are going to study a generic operator satisfying the commutation
relations (1.57) without a reference to any particular representation (apart from a few examples at the end of this
section).
Define the angular momentum operator Jˆ whose components satisfy the commutation relation
[Jˆi , Jˆk ] = i~ikl Jˆl . (6.1)

Orbital angular momentum operator L̂ is a particular example of such an operator. In mathematical physics relations
(6.1) are referred to as the the Lie algebra. Introduce a Casimir operator
X
Jˆ2 = Jˆk2 = Jˆx2 + Jˆy2 + Jˆz2 . (6.2)
k

This operator commutes with all components of Jˆ (summation over the repeated indexes is implied):

[Jˆ2 , Jˆi ] = [Jˆk Jˆk , Jˆi ] = Jˆk [Jˆk , Jˆi ] + [Jˆk , Jˆi ]Jˆk = i~(kil Jˆk Jˆl + kil Jˆl Jˆk ) = i~ kil (Jˆk Jˆl + Jˆk Jˆk ) = 0 . (6.3)
|{z} | {z }
ant-sym sym

Since [Jˆ2 , Jˆz ] = 0 physical quantities J 2 and Jz can simultaneously have definite values. Therefore, there exist a
complete set of states |jmi such that

Jˆ2 |jmi = Jj2 |jmi , (6.4)


Jˆz |jmi = ~m|jmi . (6.5)

In other words, we denote the eigenvalues of Jˆ2 and Jˆz operators as Jj2 and ~m respectively; j denotes the largest value
of m. To determine the possible values of these eigenvalues it is convenient to introduce the following non-Hermitian
operators, which are similar to the ladder operators of harmonic oscillator:
1
Jˆ± = √ (Jˆx ± iJˆy ) . (6.6)
2
Their properties:
[Jˆ2 , Jˆ± ] = 0 , (6.7)
1
[Jˆz , Jˆ+ ] = √ (i~Jˆy + i(−i~)Jˆx ) = ~Jˆ+ , (6.8)
2
[Jz , J− ] = −~Jˆ− ,
ˆ ˆ (6.9)
[Jˆ+ , Jˆ− ] = ~Jˆz , (6.10)

We can express Jˆ2 via Jˆ± and Jˆz . To this end write

2Jˆ+ Jˆ− = (Jˆx + iJˆy )(Jˆx − iJˆy ) = Jˆx2 + Jˆy2 + iJˆy Jˆx − iJˆx Jˆy = Jˆx2 + Jˆy2 + ~Jˆz = Jˆ2 − Jˆz2 + ~Jˆz . (6.11)
Thus,
Jˆ2 = 2Jˆ+ Jˆ− + Jˆz2 − ~Jˆz . (6.12)
Simiarly,
Jˆ2 = 2Jˆ− Jˆ+ + Jˆz2 + ~Jˆz . (6.13)

70
§1 Angular momentum operator VI ANGULAR MOMENTUM

Consider

Jˆz Jˆ+ |jmi = (~Jˆ+ + Jˆ+ Jˆz )|jmi = ~Jˆ+ |jmi + ~mJˆ+ |jmi = ~(m + 1)Jˆ+ |jmi ⇒ Jˆ+ |jmi ∝ |j, m + 1i . (6.14)

Note, that |j, m + 1i is an eigenstate of Jˆz corresponding to the eigenvalue ~(m + 1). By the same token,

Jˆz Jˆ− |jmi = ~(m − 1)Jˆ− |jmi ⇒ Jˆ− |jmi ∝ |j, m − 1i . (6.15)

We see that operators Jˆ+ and Jˆ− are diagonal with respect to j and increase/decrease m by one. Since −j ≤ m ≤ j
it follows that 2j must be an integer. Therefore, j can be either a non-negative integer j = 0, 1, 2, . . . or a positive
half-integer j = 1/2, 3/2, 5/2, . . .. Evidently, for an integer j, m is integer, while for a half-integer j, m is half-integer
as well.
We are ready to compute Jj2 . Start from |j, m + 1i ∝ Jˆ+ |jmi and set m = j:

Jˆ+ |jji = 0 , (6.16)

because m ≤ j. Now, using (6.13)


1 1
Jˆ− Jˆ+ |jji = (Jˆ2 − Jˆz2 − ~Jˆz )|jji = (Jj2 − ~2 j 2 − ~2 j)|jji = 0 . (6.17)
2 2
Thus,

Jj2 = ~2 j(j + 1) (6.18)

∗ Examples.

1. In applications it is often convenient to know the matrix elements of operators Jˆx and Jˆy . To compute them we
start with the operator equation (6.12) and take its matric element:

hjm|Jˆ2 |jmi = ~2 j(j + 1) = 2hjm|Jˆ+ Jˆ− ||jmi + hjm|Jˆz2 − ~Jˆz |jmi


X
=2 hjm|Jˆ+ |j 0 m0 ihj 0 m0 |Jˆ− |jmi + ~2 (m2 − m)
m0 j 0

= 2hjm|Jˆ+ |j, m − 1ihj, m − 1|Jˆ− |jmi + ~2 m(m − 1) (6.19)

Since Jˆx and Jˆy are Hermitian operators

1 1
hj, m − 1|Jˆ− |jmi = hj, m − 1| √ (Jˆx − iJˆy )|jmi = hj, m| √ (Jˆx + iJˆy )|j, m − 1i∗ (6.20)
2 2
Substituting this into (6.19) we get

2|hj, m|Jˆ+ |j, m − 1i|2 = ~2 j(j + 1) − ~2 m(m − 1) = ~2 (j 2 + j − m2 + m) (6.21)

Thus, the only non-vanishing matrix elements are


~ p
hj, m|Jˆ+ |j, m − 1i = hj, m − 1|Jˆ− |j, mi = √ (j + m)(j − m + 1) . (6.22)
2
Consequently,
~p
hj, m ± 1|Jˆx |j, mi = (j ∓ m)(j ± m + 1) , (6.23)
2
i~ p
hj, m ± 1|Jˆy |j, mi = ∓ (j ∓ m)(j ± m + 1) . (6.24)
2

71
§1 Angular momentum operator VI ANGULAR MOMENTUM

2. The choice of z-axis as a quantization direction is a matter of convention. We can instead choose, say, x-axis.
For example, a state with the orbital angular momentum l = 1 and its projection on z-axis lz = ±1 is described
by the wave function
r r
3 ±iφ 3 x ± iy
ψl=1,lz =±1 = ∓i sin θe = ∓i . (6.25)
8π 8π r
After a cycle permutation of x, y, z: x → y, y → z, z → x, we get a wave function
r r
3 y ± iz 3
ψl=1,lx =±1 = ∓i = ∓i (sin θ sin φ ± i cos θ) , (6.26)
8π r 8π
describing a state with orbital angular momentum l = 1 and its projection on x-axis lx = ±1.
In general, transformation from one complete set |jmi1 to another complete set |jmi2 can be done by the
rotation operator
ˆ
R̂γn = eiJ·nγ/~ , (6.27)
which is a generalization of (2.87), as follows:

|jmi2 = R̂γn |jmi1 . (6.28)


P
Acting on the left-hand-side with a unit operator 1̂ = j 0 m0 |j 0 m0 i1 1 hj 0 m0 | we can write this in a matrix form
X X (j)
|jmi2 = |j 0 m0 i1 1 hj 0 m0 |R̂γn |jmi1 = Dm0 m |jm0 i1 , (6.29)
j 0 m0 m0

where I took into account that the rotation operator is diagonal with respect to j and introduced matrix elements
(j)
Dm0 m = 1 hjm0 |R̂γn |jmi1 . (6.30)
(j)
Dm0 m is called the Wigner matrix. This is the matrix element of the rotation operator (6.27) about the vector
n through the angle γ. It can be expressed in terms of three Euler angles. In the case of the orbital angular
momentum, projecting (6.30) onto the coordinate representation we have (see e.g. IV §2):
l
X (l)
Ylm (θ2 , φ2 ) = Dm0 m Ylm0 (θ1 , φ1 ) . (6.31)
m0 =−l

In a particular case of rotation around the y-axis the Wigner matrix depends only on one angle and is denoted
by the lower-case d:
(j) ˆ
dm0 m (γ) = hjm0 |e−iγ Jy |jmi . (6.32)
(Extra minus sign is a convention). Some of these functions are listed at the end of these notes.
3. In the case j = 1 (6.23) can be represented as a 3 × 3 matrix:
 √ 
√0 2 √0
~
Jˆx =  2 √0 2. (6.33)
2
0 2 0
Suppose that a particle is in a state with Jx = 0. Denote this state as
 
a
ψ0 =  b  . (6.34)
c

Because Jx is definite, ψ0 must be an eigenstate of Jˆx , i.e.


 √    
√0 2 √0 a √ b
Jˆx ψ0 = 0 ⇒  2 √0 2  b  = 2 a + c  = 0. (6.35)
0 2 0 c b

72
§1 Angular momentum operator VI ANGULAR MOMENTUM


Thus, b = 0 and a = −c. Normalization requires that a = 1/ 2, so finally
 
1  1 
ψ0 = √ 0 . (6.36)
2 −1

4. Consider a rotator (such as a diatomic molecule) with the moment of inertia I rotating in plane around its
symmetry axis. The Hamiltonian is

L̂2z
Ĥ = . (6.37)
2I

Since [Ĥ, L̂z ] = 0, the wave functions of stationary states can be chosen to be also eigenfunctions of L̂z = −i~∂φ .
The corresponding complete set is

1 ~2 m 2
ψm = √ eimφ , Em = , m = 0, ±1, ±2, . . . . (6.38)
2π 2I

All states except the ground state (m = 0) are two-fold degenerate. This is because the Hamiltonian commutes
with the operator of inversion (reflection) with respect to the x-axis defined as P̂x ψ(x, y) = ψ(x, −y) (which is
equivalent to φ → −φ), whereas [P̂x , L̂z ] 6= 0.
Instead of the states (6.38) that have definite Lz , one can choose states with definite Px = ±1:

+ 1 − 1
ψm = √ cos(mφ) , ψm = √ sin(mφ) . (6.39)
π π

Suppose that the rotator is in a state described by the wave function


r
2 4
ψ = A cos φ , A = . (6.40)

We can expand
X 1 A 2iφ 
ψ= Cm √ eimφ = e + 2 + e−2iφ . (6.41)
m
2π 4

It follows that the only three non-vanishing coefficients are


r r
2 1
C0 = , C±2 = (6.42)
3 6
Expectation values of physical quantities
X
hLz i = |Cm |2 ~m = 0 , (6.43)
m

X 4~2
L2z = |Cm |2 ~2 m2 = , (6.44)
m
3
X 2~2
hEi = |Cm |2 Em = . (6.45)
m
3I

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §26-28,58, Merzbacher,
“Quantum Mechanics” 3rd edition, Ch. 11,17.

73
§2 Spin VI ANGULAR MOMENTUM

§2. Spin

Thus far we considered only one example of the angular momentum operator, viz. the orbital angular momentum
L̂. The corresponding quantum number l can be only integer, which as we will see, is intimately related to the fact
that L̂ has a classical analogue. In fact, in I §4 we constructed L̂ as a generalization of its classical expression.
I pointed out in VI §1 that j can be either integer or half-integer. The first evidence of half-integer angular
momentum came from the Stern-Gerlach experiment. In this experiment a beam of atoms passes through a magnetic
field with fixed direction but varying absolute value in the direction x perpendicular to the beam velocity. After
passing through the magnetic field the beam is projected onto a screen. Atom in magnetic field is subject to a force
of magnitude
∂B
Fx = µB
∂x
along the x-direction. Coefficient µB is a projection of the magnetic moment of a particle in the bean onto the
magnetic field direction. If we compute the magnetic moment using the classical formula
e
µ=− L, (6.46)
2M c
where M is electron mass, and then quantize the orbital angular momentum along the magnetic field direction, then
the possible values of the magnetic moment are
e~
µB = − m, m = 0, ±1, ±2, . . . , ±l .
2M c
Since l is integer, we expect to see 2l + 1 traces on the screen, which is an odd number. However, the experiment
indicated that the incident beam of Ag and Na atoms produces an even number of traces. This can be explained only
by existence of an additional half-integer contribution to the total angular momentum.
All empirical evidence points to the existence of an intrinsic angular momentum, or spin, of particles. Unlike the
orbital angular momentum of a particle, which describes its motion in space, spin is an intrinsic property of a particle.
Spin has no classical analogy and cannot be visualized as a rotation of a particle around its axis. Spin characterizes
transformational property of the wave function under rotations.
Since spin Ŝ is an angular momentum operator, we can apply to it the formalism developed in VI §1. In particular,
it satisfies the commutation relations

[Ŝi , Ŝj ] = i~ijk Ŝk . (6.47)

Let |sσi be a complete set of eigenstates of operators Ŝ 2 and Ŝz ; the corresponding eigenvalues are S 2 = ~2 s(s + 1)
and Sz = ~σ, σ = −s, −s + 1, . . . , s − 1, s. It serves as a basis of the spin representation. Spin s can be either integer or
half-integer: s = 0, 21 , 1, 32 , . . .. In the spin representation, a wave function of a particle with spin s can be represented
by a column of 2s + 1 components called the spinor :
 
ψs (q)
 ψs−1 (q) 
 ...  . (6.48)
ψ−s (q)

The spin operator is represented by a square matrix that can obtained from (6.23),(6.24) after replacements J → S,
j → s and m → σ:
~p
hs, σ ± 1|Ŝx |s, σi = (s ∓ σ)(s ± σ + 1) , (6.49)
2
~p
hs, σ ± 1|Ŝy |s, σi = ∓i (s ∓ σ)(s ± σ + 1) , (6.50)
2
hs, σ|Ŝz |s, σi = ~σ . (6.51)

Spin operator is a generator of rotations in the spinor space (see (2.87)):

R̂φn = eiŜ·nφ/~ (6.52)

74
§2 Spin VI ANGULAR MOMENTUM

z
Consider a rotation through 2π about the z-axis R̂2π = e2πiŜz /~ . The eigenvalues of this operator are

2πiσ 1 for integer σ
e = cos(2πσ) + i sin(2πσ) = = (−1)2s (6.53)
−1 for half-integer σ

Thus, rotation of a system with half-integer spin in the spinor space by 2π generates an overall minus sign in front of
it wave function.

Of particular importance is the case of s = 1/2. The corresponding representation consists of two basis states with
σ = ±1. In this representation
D E D E ~  
1 1 1 1 1 1 1 1 ~ 0 1
, − Ŝx , = , Ŝx , − = ⇒ Ŝx = , (6.54)
2 2 2 2 2 2 2 2 2 2 1 0

and similarly for the other spin components. Introduce the Pauli matrices σ̂ such that(13
~
Ŝ = σ̂ . (6.55)
2
From (6.54) we see that in the spinor representations they have the following form:
     
0 1 0 −i 1 0
σ̂x = , σ̂y = , σ̂z = . (6.56)
1 0 i 0 0 −1

Properties of the Pauli matrices:

1. Square of any Pauli matrix is a unit matrix:


 
1 0
σ̂x2 = σ̂y2 = σ̂z2 = 1̂ = . (6.57)
0 1

2. Product of two different Pauli matrices is proportional to a third Pauli matrix

σ̂y σ̂z = iσ̂x , σ̂z σ̂x = iσ̂y , σ̂x σ̂y = iσ̂z . (6.58)

Eq. (6.57),(6.58) can be written together in a compact form

σ̂i σ̂j = iijk σ̂k + δij 1̂ . (6.59)

3. Pauli matrices anti-commute:

{σ̂i , σ̂j } = σ̂i σ̂j + σ̂j σ̂i = 2δij . (6.60)

This property follows from (6.59).


4. Other useful relations (home assignment):

σ̂ 2 = 3 1̂ , Tr σ̂i = 0 , Tr(σ̂i σ̂k ) = 2δik , (6.61)

(σ̂ · a)(σ̂ · b) = a · b 1̂ + iσ̂ · (a × b) , (6.62)


(
an 1̂ n even
(a · σ̂)n = (6.63)
n−1
a (a · σ̂) , n odd

(13 Do not confuse the Pauli matrices σ with the eigenvalues of Ŝz operator ~σ.

75
§3 Addition of angular momenta VI ANGULAR MOMENTUM

∗ Example. Prove that the rotation operator for s = 1/2 can be written as

R̂φn = cos(φ/2) 1̂ + i sin(φ/2)(n · σ̂) . (6.64)

Indeed, using (6.63) we can write (6.52) as


   
1 1 1
R̂φn = ei 2 σ̂·nφ = cos σ̂ · nφ + i sin σ̂ · nφ
2 2

X 1   2k X∞  2k+1
k 1 1 k 1
= (−1) σ̂ · nφ +i (−1) σ̂ · nφ
(2k)! 2 (2k + 1)! 2
k=0 k=0
X∞  2k X∞  2k+1
1 φ 1 φ
= (−1)k 1̂ + i (−1)k (n · σ̂) ,
(2k)! 2 (2k + 1)! 2
k=0 k=0

which yields (6.64). In particular, rotation about z-axis is given by


 iφ/2 
e 0
R̂φz = cos(φ/2) 1̂ + i sin(φ/2)σ̂z = (6.65)
0 e−iφ/2

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §54-56, Merzbacher,
“Quantum Mechanics” 3rd edition, 16.1-16.4.

§3. Addition of angular momenta

Consider a system consisting of two non-interacting parts each having angular momentum Jˆ1,2 :

[Jˆ1i , Jˆ2k ] = 0 . (6.66)

In particular, this applies to a system possessing orbital angular momentum and spin. Eq. (6.66) means that the
system can be in states with definite values of

J12 = ~2 j1 (j1 + 1) , and J22 = ~2 j2 (j2 + 1) (6.67)

and their z-projections

J1z = ~m1 , and J2z = ~m2 . (6.68)

Denote the corresponding state vectors by |j1 m1 j2 m2 i = |j1 m1 i|j2 m2 i. At fixed j1 and j2 there are (2j1 + 1)(2j2 + 1)
eigenfunctions with different m1 and m2 .
Define the operator of the total angular momentum as Jˆ = Jˆ1 + Jˆ2 . It satisfies the same commutation relations as
J1 and Jˆ2 :
ˆ

[Jˆi , Jˆk ] = i~ikl Jˆl . (6.69)

The Casimir operator can be written as

Jˆ2 = Jˆ12 + Jˆ22 + 2Jˆ1 · Jˆ2 . (6.70)

Since Jˆ2 is the Casimir operator of the algebra (6.69), it commutes with all components of Jˆ, in particular, [Jˆ2 , Jˆz ] = 0.
It also commutes with the Casimir operators of subsystems:

[Jˆ2 , Jˆ12 ] = [Jˆ2 , Jˆ22 ] = 0 , (6.71)

because [Jˆ12 , Jˆ1k ] = 0. Note, however, that [Jˆ2 , Jˆ1z ] 6= 0 since [Jˆ1 · Jˆ2 , Jˆ1z ] 6= 0.
Since Jˆz = Jˆ1z + Jˆ2z commutes with operators Jˆ12 , Jˆ22 , Jˆ1z , Jˆ2z , states |j1 m1 j2 m2 i are also its eigenstates corre-
sponding to the eigenvalues Jz = ~m = ~(m1 + m2 ). However, they are not eigenstates of the operator Jˆ2 , because

76
§3 Addition of angular momenta VI ANGULAR MOMENTUM

Jˆ2 does not commute with Jˆ1z and Jˆ2z . Operator Jˆ1 · Jˆ2 mixes states differing by one in m1 and m2 . In practice, it
is advantageous to have states with definite values of Jˆ2 and Jˆz . Since Jˆ12 and Jˆ22 commute with them we can choose
as a basis set of states, eigenstates of these four operators: Jˆ2 , Jˆz , Jˆ12 and Jˆ22 . I will denote such states by |j1 j2 jmi.
In view of completeness of both sets we can write the following unitary transformation
X
|j1 j2 jmi = hj1 m1 j2 m2 |j1 j2 jmi|j1 m1 i|j2 m2 i . (6.72)
m1 ,m2

The matrix elements hj1 m1 j2 m2 |j1 j2 jmi are called the Clebsch-Gordan coefficients. Often, when j1 and j2 are fixed
they are denoted simply by hm1 m2 |jmi to simplify the notation.
Properties of the Clebsch-Gordan coefficients:

1. Clebsch-Gordan coefficients are non-vanishing only if m = m1 + m2 . Indeed, on the one hand


X
Jˆz |j1 j2 jmi = hm1 m2 |jmi~(m1 + m2 )|j1 m1 i|j2 m2 i = ~(m1 + m2 )|j1 j2 jmi . (6.73)
m1 ,m2

On the other hand, Jˆz |j1 j2 jmi = ~m|j1 j2 jmi implying that m1 + m2 = m.
2. At fixed j1 and j2 number j can take only the following values:

|j1 − j2 | ≤ j ≤ j1 + j2 . (6.74)

A derivation of this property can be found in the book by Merzbacher, 17.5. Below I will illustrate it with an
example (Example 1).
To each value of j, at fixed j1 and j2 , there correspond 2j + 1 values of m: m = −j, −j + 1, . . . , j − 1, j. The
total number of states with all possible values of j:
1 +j2
jX
(2j + 1) = (2j1 + 1)(2j2 + 1) . (6.75)
j=|j1 −j2 |

According to (6.72) instead of the (2j1 + 1)(2j2 + 1) basis states

{|j1 m1 j2 m2 i}m1 =−j1 ,...,j2 ; m2 =−j2 ,...,j2 , (6.76)

which form a complete set in the Hilbert space, we can use another complete set

{|j1 j2 jmi}j=|j1 −j2 |,...,j1 +j2 ; m=−j,...,j (6.77)

spanning the same Hilbert space (and containing the same number (2j1 + 1)(2j2 + 1) of basis states).
3. By convention, Clebsch-Gordan coefficients are real. This implies that the inverse transformation to (6.72) is
X
|j1 m1 i|j2 m2 i = hj1 j2 m1 m2 |j1 j2 jmi|j1 j2 jmi . (6.78)
jm

4. Orthogonality and normalization


X
hj1 j2 m1 m2 |j1 j2 jmihj1 j2 m01 m02 |j1 j2 jmi = δm1 m01 δm2 m02 , (6.79)
jm
X
hj1 j2 m1 m2 |j1 j2 j 0 m0 ihj1 j2 m1 m2 |j1 j2 jmi = δjj 0 δmm0 , (6.80)
m1 ,m2

5. Symmetry under particle permutations:

hj1 j2 m1 m2 |j1 j2 jmi = (−1)j1 +j2 −j hj2 j1 m2 m1 |j2 j1 jmi , (6.81)

which implies that

|j1 j2 jmi = (−1)j1 +j2 −j |j2 j1 jmi . (6.82)

77
§3 Addition of angular momenta VI ANGULAR MOMENTUM

∗ Examples.
1. Consider two systems with j1 = 3 and j2 = 3/2. The total number of basis states is (3 · 2 + 1)( 32 · 2 + 1) = 28.
Let us list and count all these states in two different representations: |j1 j2 m1 m2 i and |j1 j2 jmi. I will organize
the list starting from the largest m and decreasing it by one. Keep in mind that for a given m, m1 + m2 = m
and j ≥ |m|.

|j1 j2 m1 m2 i |j1 j2 jmi


m1 m2 # of states m j # of states
3 3/2 1 9/2 9/2 1

3 1/2 9/2
2 7/2 2
2 3/2 7/2

3 −1/2 9/2
2 1/2 3 5/2 7/2 3
1 3/2 5/2

3 −3/2 9/2 (6.83)


2 −1/2 4 3/2 7/2 4
1 1/2 5/2
0 3/2 3/2

2 −3/2 9/2
1 −1/2 7/2
0 1/2 4 1/2 5/2 4
−1 3/2 3/2
no state with j=1/2
− − 1/2

>


··· ··· ··· ··· ··· ···


The table continues symmetrically to the values m = −1/2, −3/2, −5/2, −7/2, −9/2. Going down the table I
included states with j = 9/2, . . . , 3/2 in order that the number of states in both sets be the same. It is seen
that for m = 1/2, the state with j = 1/2 does not exist. This agrees with the rule that 3 − 3/2 ≤ j ≤ 3 + 3/2.
2. Consider expectation value of J = J1 +J2 in a state |j1 m1 j2 m2 i. Since Jˆ2 = Jˆ12 +Jˆ22 +2Jˆ1 ·Jˆ2 and hJx i = hJy i = 0
(home assignment) we get

2
J = ~2 j1 (j1 + 1) + ~2 j2 (j2 + 1) + 2~2 m1 m2 . (6.84)

3. Using the table of the Clebsch-Gordan coefficients.


(a) j1 = 3/2, j2 = 1/2.

1 3
|m1 = 3/2, m2 = −1/2i = |j = 2, m = 1i + |j = 1, m = 1i . (6.85)
2 2
(b) j1 = 2, j2 = 1.
r r
3 2
|m1 = 0, m2 = 0i = |j = 3, m = 0i − |j = 1, m = 0i . (6.86)
5 5
(c) j1 = 2, j2 = 1/2.
1 2
|j = 5/2, m = 3/2i = √ |m1 = 2, m2 = −1/2i + √ |m1 = 1, m2 = 1/2i . (6.87)
5 5

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §, Merzbacher, “Quan-
tum Mechanics” 3rd edition,

78
§4 Matrix elements of vector operators VI ANGULAR MOMENTUM

§4. Matrix elements of vector operators

We have noticed in I §4 that commutation relations of vector operators with the operator of orbital angular momen-
tum are similar, see (1.59) and (1.60). In II §5 we attributed this to the fact that L̂ is proportional to the generator of
rotations: all vector quantities transform under the rotations in the same way. (This is true for any tensor quantity,
but we will consider vectors for simplicity). This is why we defined a vector operator by (2.94). We can generalize
this definition as follows:

[Jˆi , V̂j ] = i~ijk V̂k . (6.88)

Eq. (6.88) allows us to relate the matrix element of the operator V̂ in the basis |jmi to the matrix elements of
operators Jˆ and Jˆ · V̂ in the same basis. To derive the corresponding expression, start with the following commutator

[Jˆ2 , Jˆ × V̂ ] = 2i~{Jˆ2 V̂ − (Jˆ · V̂ )Jˆ} , (6.89)

which can be proved using (6.88),(6.1) and (C7). Let us compute the matrix elements hjm0 | . . . |jmi on the both sides
of (6.89). Firstly,

hjm0 |[Jˆ2 , Jˆ × V̂ ]|jmi = hjm0 |Jˆ2 (Jˆ × V̂ ) − (Jˆ × V̂ )Jˆ2 |jmi = ~2 hjm0 |(Jˆ × V̂ ){j(j + 1) − j(j + 1)}|jmi = 0 . (6.90)

Secondly, consider the following matrix element


X
hjm0 |(Jˆ · V̂ )Jˆ|jmi = hjm0 |(Jˆ · V̂ ) |j 0 m00 ihj 0 m00 |Jˆ|jmi (6.91)
j 0 ,m00

It is easy to check that [Jˆ2 , Jˆ · V̂ ] = 0 and [Jˆz , (Jˆ · V̂ )] = 0 (because Jˆ · V̂ is a scalar). Therefore, state vectors |jmi
are eigenstates of operator Jˆ · V̂ implying that

hjm0 |Jˆ · V̂ |j 0 m00 i ∝ δm0 m00 δj 0 j 0 . (6.92)

It follows from (6.91) and (6.92) that

hjm0 |(Jˆ · V̂ )Jˆ|jmi = hjm0 |Jˆ · V̂ |jm0 ihjm0 |Jˆ|jmi . (6.93)

Lastly,

hjm0 |Jˆ2 V̂ |jmi = ~2 j(j + 1)hjm0 |V̂ |jmi . (6.94)

Combining (6.89),(6.90),(6.93) and (6.94) we finally obtain


1
hjm0 |V̂ |jmi = hjm0 |Jˆ · V̂ |jm0 ihjm0 |Jˆ|jmi . (6.95)
~2 j(j + 1)

Equivalently, replace m ↔ m0 , take the complex conjugation on both sides and use the fact that all involved operators
are Hermitian to write (6.95) in a slightly different way:
1
hjm0 |V̂ |jmi = hjm|Jˆ · V̂ |jmihjm0 |Jˆ|jmi , (6.96)
~2 j(j + 1)

which indicates that the matrix element hjm|Jˆ · V̂ |jmi does not depend on m at all.
Eq. (6.96) is a particular case of the Wigner-Eckart theorem. The Wigner-Eckart theorem separates the geomet-
ric and symmetry-related properties of the matrix elements (in our example hjm0 |Jˆ|jmi) from the other physical
properties that are contained in the reduced matrix element (in our example hjm|Jˆ · V̂ |jmi).

∗ Examples.

1. Suppose that Jˆ = Jˆ1 + Jˆ2 and let us compute the matrix elements hj1 j2 jm|Jˆ1,2 |j1 j2 jmi.

79
§4 Matrix elements of vector operators VI ANGULAR MOMENTUM

According to (6.95)
1
hjm0 |Jˆ1,2 |jmi = hj, m|Jˆ · Jˆ1,2 |jmihjm0 |Jˆ|jmi . (6.97)
~2 j(j + 1)
Decompose
1 1
Jˆ = √ (ex − iey )Jˆ+ + √ (ex + iey )Jˆ− + ez Jˆz . (6.98)
2 2
and calculate the matrix element
1 p
hjm0 |Jˆ|jmi = ~(ex − iey ) j(j + 1) − m(m + 1)δm0 ,m+1
2
1 p
+ ~(ex + iey ) j(j + 1) − m(m − 1)δm0 ,m−1 + ez ~mδmm0 . (6.99)
2
Next, using the identities
1 1
Jˆ · Jˆ1 = (Jˆ2 + Jˆ12 − Jˆ22 ) , Jˆ · Jˆ2 = (Jˆ2 + Jˆ22 − Jˆ12 ) . (6.100)
2 2
we have
1 ~2
hjm|Jˆ · Jˆ1 |jmi = hjm| (Jˆ2 + Jˆ12 − Jˆ22 )|jmi = [j(j + 1) + j1 (j1 + 1) − j2 (j2 + 1)] . (6.101)
2 2
Notice that the reduced matrix element (6.101) does not depend on m and m0 as expected. Plugging (6.101)
and (6.99) into (6.97) we derive
 
~m j1 (j1 + 1) j2 (j2 + 1)
hjm|Jˆ1 |jmi = ez 1+ − , (6.102)
2 j(j + 1) j(j + 1)
 
~m j1 (j1 + 1) j2 (j2 + 1)
hjm|Jˆ2 |jmi = ez 1− + . (6.103)
2 j(j + 1) j(j + 1)

2. Consider electron in a state with given angular orbital momentum l, and z-component m of its total angular
momentum j. The possible values of the total angular momentum are j = l + 1/2 and j = l − 1/2. Using (6.102)
and (6.103) with replacements Jˆ1 → L̂ and Jˆ2 → Ŝ we have the following diagonal matrix elements
D E  
1 1 1 1 ~m l(l + 1) 3 ~m
l + 2 , m; l, 2 Ŝ l + 2 , m; l, 2 = ez 1− 1 3 + 1 3 = ez , (6.104)
2 (l + 2 )(l + 2 ) 4(l + 2 )(l + 2 ) 2l + 1
D E  
~m l(l + 1) 3 ~m
l − 12 , m; l, 12 Ŝ l − 21 , m; l, 12 = ez 1− 1 1 + 1 1 = −ez , (6.105)
2 (l − 2 )(l + 2 ) 4(l − 2 )(l + 2 ) 2l + 1
D E  
~m l(l + 1) 3 ~m 2l
l + 21 , m; l, 12 L̂ l + 12 , m; l, 12 = ez 1+ − = ez , (6.106)
2 (l + 12 )(l + 32 ) 4(l + 12 )(l + 32 ) 2l + 1
D E  
1 1 1 1 ~m l(l + 1) 3 ~m(l + 1)
l − 2 , m; l, 2 L̂ l − 2 , m; l, 2 = ez 1+ 1 1 − 1 1 = ez . (6.107)
2 (l − 2 )(l + 2 ) 4(l − 2 )(l + 2 ) l + 12
We will see later that the magnetic dipole moment of electron is given by
e
µ̂ = (L̂ + 2Ŝ) . (6.108)
2M c
Its expectation value in two states with different j are
e D E e ~m(l + 1)

hµij=l+ 1 = l + 12 , m; l, 12 (L̂ + 2Ŝ) l + 12 , m; l, 12 = ez , (6.109)
2 2M c 2M c l + 12
e D E e ~ml

hµij=l− 1 = l − 12 , m; l, 12 (L̂ + 2Ŝ) l − 12 , m; l, 12 = ez , (6.110)
2 2M c 2M c l + 12

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §106,107, Merzbacher,
“Quantum Mechanics” 3rd edition, Ch. 17.

80
§1 Approximations for a discrete non-degenerate spectrum VII STATIONARY PERTURBATION THEORY

VII. STATIONARY PERTURBATION THEORY

§1. Approximations for a discrete non-degenerate spectrum

In most cases it is impossible to find the exact solution to the Schrödinger equation and we have to employ
approximate methods. Consider a system described by a time-independent Hamiltonian Ĥ that can we written down
as Ĥ = Ĥ0 + V̂ where Ĥ0 is the Hamiltonian with known spectrum and V̂ can be considered as a small perturbation.
The exact meaning of the term “small perturbation” will be discussed later. The goal of the perturbation theory is
to construct series in small perturbation that approximate the spectrum of Ĥ with a given accuracy.
We start with the “non-perturbed” Hamiltonian Ĥ0 which is taken to be the zeroth order in the perturbation
(0) (0)
theory. I denote it by En and the corresponding eigenfunctions by ψn . In this section we consider the simplest
case of discrete non-degenerate spectrum. The eigenvalue problem for Ĥ0 is

Ĥ0 ψn(0) = En(0) ψn(0) . (7.1)

We would like to solve the eigenvalue problem for Ĥ, which reads

(Ĥ0 + V̂ )ψ = Eψ . (7.2)
(0)
Since functions ψn is a complete set we can write
X
ψ= an ψn(0) , (7.3)
n

where the coefficients an constitute the energy representation of ψ. It is convenient to write (7.2) in the energy
(0)
representation. Substituting (7.3) to (7.2), multiplying by (ψm )∗ and integrating over all relevant degrees of freedom
(coordinates, spins etc.) we get
X
(0)
(E − Em )am = Vmn an , (7.4)
n

(0) (0)
where Vmn = hψm |V |ψn i.(14 This is the Schrödinger equation in the energy representation.
As mentioned above, in most cases one cannot solve the exact equation (7.4). Instead we are looking for a solution
in a form of series. To find an approximation to state l we write
(0) (1) (2) (2) (1) (0)
E = El + El + El + ... , where El  El  El , (7.5)
am = δml + a(1)
m + a(2)
m + ... , where a(2)
m  a(2)
m  1. (7.6)

Number in brackets indicates the order of the perturbation theory, i.e. how many factors of Vmn are taken into
account. It is important to realize, that eigenfunctions of Ĥ are generally not the same as those of Ĥ0 . Therefore,
the set of eigenfunctions at the n’th each oder of the perturbation theory is different from the set of eigenfunctions at
(1) (0)
the (n − 1)’th order. Particularly, ψn can be expressed as a linear superposition of ψn ’s. This is reflected in (7.6)
– had the wave functions were the same at each order, am ’s would have been equal one, see (7.3). Eq. (7.6) takes
(0)
into account mixing of the zeroth order wave functions ψn , that are used as the building blocks of the perturbation
theory. Substituting (7.5) and (7.6) into (7.4) and collecting terms of the same order in the perturbation theory we
obtain, for m = l:
(1)
El = Vll , (7.7)
(2) (1) (1)
X
El + El al = Vln a(1)
n , (7.8)
n

(14 Throughout this chapter there is no summation over repeated indexes. The summation sign Σ is explicitly indicated where needed.
Thus, there is no summation over m in the left-hand-side of (7.4).

81
§1 Approximations for a discrete non-degenerate spectrum VII STATIONARY PERTURBATION THEORY

and for m 6= l:
(0)
a(1)
m (El
(0)
− Em ) = Vml , (7.9)
(1) (0)
X
El a(1)
m + (El
(0) (2)
− Em )am = Vmn a(1)
m . (7.10)
n

From (7.7) it follows that energy of the l’th level is given by


(0)
E = El + Vll (7.11)

to the first order in the perturbation theory. At the same order


Vml
a(1)
m = (0) (0)
. (7.12)
El − Em

Thus, the eigenfunction of the l’th level is given by

(0)
X0 Vml (1) (0)
(0)
ψl = ψl + (0) (0)
ψm + al ψl . (7.13)
m El − Em
P0 (1)
Symbol m means summation over all m except m = l. Coefficient al must be chosen to satisfy the normalization
(0)
condition of ψl . Since ψl is also normalized we have

(0) (0) (1) (0) (0) (1) ∗ (0) (0) (1) (1) ∗
1 = hψl |ψl i = hψl |ψl i +al hψl |ψl i + al hψl |ψl i + o(V 2 ) ⇒ al + al = 0, (7.14)
| {z }
=1

(1) (1)
implying that al can be either imaginary or zero. The simplest normalization is to set al = 0. We finally obtain

(0)
X0 Vml (0)
ψl = ψl + (0) (0)
ψm . (7.15)
m El − Em

To calculate the next-to-the leading order (NLO) correction, i.e. the second order in the perturbation theory, we
substitute (7.12) into (7.8):

(2)
X0 X0 |Vlm |2
El = Vlm a(1)
m = (0) (0)
. (7.16)
m m El − Em
(2) (0) (2)
Notice that the second order correction to the ground state is always negative E0 < 0 because E0 < Em . Together
with (7.11) we obtain the following approximation to the energy levels:

(0)
X0 |Vlm |2
E = El + Vll + (0) (0)
. (7.17)
m El − Em

(0) (2)
Eqs. (7.17) and (7.12) indicate that the perturbation theory is applicable if Vlm  |El − Em | for any l 6= m.
This is however only a necessary condition. In some cases perturbation changes the character of the solution so that
(0)
in the mathematical limit Vlm → 0, energy levels El and eigenfunctions ψl do not continuously approach El and
(0)
ψl . We illustrate this below in example 3.

∗ Examples.

1. A charged particle moves in one-dimensional infinite potential well at 0 ≤ x ≤ a. Its energy spectrum is
r
(0) ~2 π 2 (n + 1)2 (0) 2 π(n + 1)x
En = , ψn = sin , n = 0, 1, 2 . . . (7.18)
2ma2 a a

82
§1 Approximations for a discrete non-degenerate spectrum VII STATIONARY PERTURBATION THEORY

In electric field E its acquires potential energy V = −qEx which can be considered a perturbation in a weak
field. The first order correction to the energy levels reads, see (7.17),
Z
2 a π(n + 1)x qEa
En(1) = Vnn = V (x) sin2 dx = − . (7.19)
a 0 a 2
(1)
The first order correction is the same for all energy levels. It is interesting to note that at large n  1, En
does not depend on n for any perturbation V . Indeed, due to rapid oscillations of the sinus, we can replace its
square by the average value sin2 (· · · ) ≈ 1/2 which implies that
Z
(1) 1 a
En ≈ V (x)dx , n  1 . (7.20)
a 0
To calculate the NLO correction, we need to know the off-diagonal matrix elements of the perturbation. In
particular, for the ground state l = 0:
Z
2 a π(n + 1)x πx 4[(−1)n − 1](n + 1)a
V0n = −qE x sin sin2 dx = −qE , n 6= 0 . (7.21)
a 0 a a π 2 n2 (n + 2)2
Using (7.21) and (7.18) in (7.16) and changing the summation variable n = 2k + 1 we derive
∞  2 ∞
(2) 2ma2 X 4[(−1)n − 1](n + 1)a 1 512q 2 E 2 a4 X (k + 1)2
E0 = q 2 E 2 2 2 2 2 2 2
= − 2 6
~ π n=1 π n (n + 2) 1 − (n + 1) ~ π (2k + 1)5 (2k + 3)5
k=0
2 2 4
1 2 q E a
=− π (15 − π 2 ) 6 2 . (7.22)
24 π ~
2. Consider particle in the Hulthen potential
U0
U =− , a > 0. (7.23)
−1
er/a
If the particle is in a state such that hri  a, then we can approximate (7.23) by the Coulomb potential:
U0 U0 a
U =− ≈− . (7.24)
er/a − 1 r
In such a case, we can choose the Coulomb potential as the zeroth order approximation of the perturbation
theory and try to calculate the energy levels in (7.23) using the perturbation theory. Since in the Coulomb field
~2
hri = 23 n2 a0 (for n  l), see (5.68), where a0 = maU 0
this approximation applies when
ma2 U0
λ≡  n2 . (7.25)
~2
Assuming that (7.25) is satisfied, we can write
ma2 U02 (0)
En(0) = − , ψnlm (r) = ψnlm Coulomb
(r) . (7.26)
2~2 n2
Next, we need to determine the perturbation operator:
   
U0 a U0 a 1 U0 a 1 r
V (r) = U (r) − − ≈− r r2
+ = −U0 − + , (7.27)
r r 1 + 2a + 6a 2
r 2 12a
where I dropped terms of the order of U0 r2 /a2 . Employing now (5.68) we derive the first order correction
 
(1) 1 1
Enl = U0 − [3n2 − l(l + 1)] . (7.28)
2 24λ
Notice that it depends on n and l, i.e. the Coulomb degeneracy is lifted by the perturbation.
Actually, energy spectrum of the Hulthen potential can be calculated exactly for l = 0. The result is
~2  2 2
Enexact
r0
=− 2 2
λ − (nr + 1)2 , (7.29)
8ma (nr + 1)
(0) (1)
where n = nr + l + 1. It is amusing that Enexact
r0
= En + En0 , i.e. the first order of the perturbation theory
reproduces the exact result.

83
§2 Approximations for a discrete degenerate spectrum VII STATIONARY PERTURBATION THEORY

3. Energy spectrum of a particle in harmonic oscillator potential U = 12 mω 2 x2 is discrete, because it cannot go to


x → ±∞. Consider the effect of the perturbation V = λx3 , with λ > 0 on the energy spectrum. The matrix
elements of the perturbation operator are
 3/2
~ √
Vmn = λ|hm|x3 |ni| = λ (2n + 3) n + 1δm,n+1 . (7.30)
2mω
The perturbation theory applies if
 3/2
~ √ (0)
λ (2n + 3) n + 1  |Em − En(0) | = ~ω|m − n| . (7.31)
2mω
However this condition is not sufficient. Notice that whereas the harmonic oscillator energy spectrum is discrete,
the spectrum of the perturbed Hamiltonian is continuous. This is because U + V → −∞ when x → −∞
indicating that particle eventually goes to x → −∞. No matter how small is λ, perturbation V converts discrete
spectrum into continuous, a feature that is completely missed by the perturbation theory. Nevertheless, the
perturbation theory is still useful since the probability that the particle escapes to infinity (by tunneling through
the potential barrier) is exponentially small. See also IX §4.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §38, Merzbacher,
“Quantum Mechanics” 3rd edition, 18.1-18.4.

§2. Approximations for a discrete degenerate spectrum

The perturbation theory developed in the previous section assumes that the energy spectrum of the unperturbed
(0)
Hamiltonian, which is the zeroth order approximation, is not degenerate. Suppose now that level El is f -fold
(0)
degenerate, viz. there are f linearly independent functions ψlk , k = 1, . . . , f corresponding to this level. For any two
(0)
wave functions ψlk , the denominator in (7.16) vanishes indicating the breakdown of the perturbation series. This
(0)
problem can be avoided if instead of the wave functions ψlk we choose f different linearly independent combinations
(0) (0) (0)
ϕlk as the zeroth order wave functions, and require that hϕlk |V |ϕlk0 i = 0 for any k 6= k 0 . This guarantees that the
(0)
problematic terms will not appear at all in the sum in (7.16). Functions ϕlk are often called the correct functions of
the zeroth order.
To determine the correct functions of the zeroth order, we repeat the derivation (7.1)–(7.4) now focusing on the
degenerate level l. The eigenvalue problem for Ĥ0 is
(0) (0)
Ĥ0 ψlk = El ψlk . (7.32)

The eigenvalue problem for Ĥ reads


(Ĥ0 + V̂ )ϕl = Eϕl . (7.33)
(0)
were ϕl are linear combinations of ψlk :
f
X (0)
ϕl = ak ψlk , (7.34)
k=1

Substituting (7.34) to (7.33), multiplying ψlk 0 and integrating over all relevant degrees of freedom we get

f
X (0)
[Vkk0 + (El − E)δkk0 ]ak0 = 0 . (7.35)
k0 =1

(0) (1) (0) (1)


Expanding E = El + El + . . . and ak = ak + ak + . . . we obtain
f
X (1) (0)
[Vkk0 − El δkk0 ]ak0 = 0 . (7.36)
k0 =1

84
§2 Approximations for a discrete degenerate spectrum VII STATIONARY PERTURBATION THEORY

This set of linear equations has a non-trivial solution iff


(1)
det |Vkk0 − El δkk0 | = 0 . (7.37)

This equation is called the secular equation. It has f real roots, since Ĥ is a hermitian opetator. If all roots are
different, then the l’th level split into f different levels Elk and for each of them we can find the corresponding wave
functions
f
X
(0) (0) (0)
ϕlk = ak ψlk (7.38)
k=1

by solving (7.36). In such a case we say that the perturbation V̂ has completely lifted the degeneracy. If one or more
roots of (7.37) are multiple, the lifting of the degeneracy is not complete. The wave functions corresponding to the
multiple roots are not uniquely determined, but can be chosen to be mutually orthogonal.
(0)
To summarize our discussion, we can apply (7.15) and (7.17) once the wave functions ψlk are replaced with the
(0)
correct wave functions ϕlk .

∗ Examples.

1. Particle is moving in 3D potential U = 21 mω 2 r2 + αxy. For α  1 we can use the perturbation theory to
calculate its energy spectrum. In this example I will focus on the first excited state.
As the zeroth order approximation (i.e. unperturbed) wave functions it is natural to choose the eigenfunctions of
the three-dimensional isotropic harmonic oscillator which can be written as a product of three one-dimensional
harmonic oscillators (by means of the separation of variables in Cartesian coordinates):

ψn(0)
x ny nz
= |nx i|ny i|nz i . (7.39)

The corresponding energy levels are


 
3
En(0) = n+ ~ω , n = nx + ny + nz . (7.40)
2

The first excited state is three-fold degenerate. The corresponding states are

|1i = |1i|0i|0i , |2i = |0i|1i|0i , |3i = |0i|0i|1i . (7.41)

The matrix elements of the perturbation that are needed for the secular equation are

Vkk0 = αhk|xy|k 0 i , k, k 0 = 1, 2, 3 . (7.42)

To calculate them recall, that the only non-vanishing matrix elements of operator x̂ for 1D harmonic oscillator
are
r
(n + 1)~
hn|x|n + 1i = hn + 1|x|ni = . (7.43)
2mω
Then

V11 = αh100|xy|100i = αh1|x|1ih0|y|0i = 0 . (7.44)

By the same token V22 = V33 = V13 = V31 = V23 = V32 = 0. The only non-vanishing matrix elements are
α~
V12 = αh100|xy|010i = αh1|x|0ih0|y|1i = = V21 . (7.45)
2mω
In matrix form
 
0 1 0
α~  
V̂ = 1 0 0. (7.46)
2mω
0 0 0

85
§2 Approximations for a discrete degenerate spectrum VII STATIONARY PERTURBATION THEORY

Secular equation now reads



−E (1) α~ 0
1 2mω "  2 #
α~
α~ (1) (1) 2
2mω −E1(1) 0 = −E1 (E1 ) − = 0. (7.47)
2mω
(1)
0 0 −E1

Its solutions are


(1) (1) α~ (1) α~
E1,0 = 0 , E1,+ = , E1,− = − . (7.48)
2mω 2mω
(0)
Thus, the first excited level E1 = 52 ~ω splits into three levels: E1,0 = 25 ~ω, E1,± = 52 ~ω ± α~
2mω . To compute
the corresponding correct wave functions write
 
a1
 
ϕ1 =  a2  . (7.49)
a3
(1)
Then, using (7.36) for E1 = 0 we have
    
0 V12 0 a1 a2 V12
    
 V12 0 0   a2  =  a1 V12  = 0 , (7.50)
0 0 0 a3 0

implying that a1 = a2 = 0. Thus,


ϕ1,0 = |3i . (7.51)
Since state |3i is not shifted by the perturbation its wave function is the correct wave function of the zeroth
(1) α~
order. Similarly for E1 = 2mω
    
−V12 V12 0 a1 −a1 + a2
    
 V12 −V12 0   a2  = V 12  a1 − a2  = 0, (7.52)
0 0 0 a3 0

implying that a1 = a2 , a3 = 0. The corresponding normalized wave function is


1
ϕ1,+ = √ (|1i + |2i) . (7.53)
2
(1) α~
Repeating this procedure again for E1 = − 2mω we find the third correct wave function:
1
ϕ1,− = √ (|1i − |2i) . (7.54)
2

2. Consider a plane rotator with electric dipole moment p in external weak electric field E. Free plane rotator has
been discussed in (6.37)–(6.39).
The perturbation operator is V̂ = −p̂ · E = −pE cos φ. Using (6.38) we obtain
Z 2π
dφ i(n−m)φ 1 iφ  pE
hm|V̂ |ni = −pEhm| cos φ|ni = −pE e e + e−iφ = − (δn,m−1 + δn,m+1 ) . (7.55)
0 2π 2 2
Thus the diagonal matrix elements vanish, implying that the perturbation theory starts with the second order.
All states except the ground state are degenerate. We will discuss them separately.
(a) The ground state.

pE X |hm|V̂ |0i|2 p2 E 2 I 2
hm|V̂ |0i = − (δm,1 + δm,−1 ) ⇒ E0 = =− . (7.56)
2 m=±1
(0)
E0 − Em
(0) ~2

86
§3 Spectrum with two close levels VII STATIONARY PERTURBATION THEORY

(b) The excited states. The perturbation operator V̂ commutes with the parity operator P̂x . Therefore, the
± ±
correct zero order wave functions are ψm of (6.39). I will denote µ = |m| = 1, 2, 3, . . . and ψm = |µ±i.
Calculation of the second order correction greatly simplifies if we note that since the perturbation operator
is parity-even function, only transitions between the same parity states are allowed: hµ ± |V̂ |µ0 ∓i = 0. In
particular, there are no transitions between the degenerate states |µ+i and |µ−i. This implies that we can
still use the non-degenerate perturbation theory in spite of the two-fold degeneracy of the excited levels.(15
The matrix elements of the perturbation for even states read
Z
1 2π
hµ + |V̂ |µ0 +i = cos(µφ) cos(µ0 φ)(−pE) cos φ dφ
π 0
√
 2, if µ = 0 , µ0 = 1 ,
pE  √
=− 2, if µ0 = 0 , µ = 1 , (7.57)
2  δ 0 + δ 0 , otherwise .
µ,µ +1 µ,µ −1

We have
(1)
Eµ+ = 0 , (7.58)
(2)
X |hµ + |V̂ |1+i| 2
|h0 + |V̂ |1+i| 2
|h2 + |V̂ |1+i| 2
E1+ = (0) (0)
= (0) (0)
+ (0) (0)
µ=0,2,3,... E1− Eµ − E1 E1 − E2
E0
 2  2
pE 2I pE 2I 5 p3 E 2 I
= √ + = , (7.59)
2 ~2 2 ~2 (1 − 4) 6 ~2
(2) p2 E 2 I
Eµ+ = , µ ≥ 2. (7.60)
(4µ2 − 1)~2
Similarly, for odd states
pE
hµ − |V̂ |µ0 −i = − (δµ,µ0 +1 + δµ,µ0 −1 ) , (7.61)
2
which yields
(1)
Eµ− = 0 , (7.62)
3 2
(2) 1p E I
E1− = − , (7.63)
6 ~2
2 2
(2) p E I
Eµ− = , µ ≥ 2. (7.64)
(4µ2 − 1)~2
(1) (1)
We observe that the degeneracy of the first excited level µ = 1 is lifted: Eµ+ 6= Eµ− , while the degeneracy
of all other levels µ ≥ 2 remains.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §39.

§3. Spectrum with two close levels

It is instructive to consider a system with two close, but not necessarily coinciding, levels and determine its energy
levels and correct eigenfunctions. In this case it is possible to obtain an exact solution to the system (7.35). (To solve
it in VII §2 we used the perturbative expansion that yielded (7.36)).

(15 Seemingly, one can use the same argument for the wave functions ψm of (6.38). Indeed, the matrix elements for transitions between the
degenerate states vanish: hm|V̂ | − mi = 0, see (7.55). Thus, the motivation to reorganize the perturbation theory, which we indicated
in the beginning of this section, does not apply! Nevertheless, it turns out that using ψm instead of the correct wave functions leads to
an incorrect result for the energy levels, see M.P. Silverman, Phys.Rev.A24, p.339, 1981.

87
§3 Spectrum with two close levels VII STATIONARY PERTURBATION THEORY

(0) (0) (0) (0)


Let the two close levels be E1 and E2 , and the corresponding eigenfunctions ψ1 and ψ2 . According to (7.34)
we are looking for the correct two eigenfunctions ϕ1,2 :
(0) (0)
ϕ = a1 ψ1 + a2 ψ2 . (7.65)

I did not append the superscript (0) to ϕ’s because we are going to determine a1,2 exactly. As in (7.33) we write the
SE

(Ĥ − E)ϕ = 0 , (7.66)

and plug in it (7.65) with the result


(0) (0) (0) (0)
a1 Ĥψ1 + a2 Ĥψ2 − Ea1 ψ1 − Ea2 ψ2 = 0 . (7.67)
(0)∗ (0)∗
Multiplying it by ψ1 or ψ2 and integrating we obtain the following two equations for a1,2 (cp. (7.35)):

a1 (H11 − E) + a2 H12 = 0 , (7.68)


a1 H21 + a2 (H22 − E) = 0 , (7.69)

where I denoted
(0) (0) (0) (0)
Hkk0 = hψk |H|ψk0 i = hψk |Ĥ0 + V |ψk0 i = E (0) δkk0 + Vkk0 . (7.70)

A non-trivial solution to (7.68),(7.69) exists if

(H11 − E)(H22 − E) − |H12 |2 = 0 . (7.71)

This is the secular equation for the two level system. Note, that unlike (7.37), it has not been expanded in the
perturbation series. Solution to (7.71) reads
1 1p
E1,2 = (H11 + H22 ) ± (H11 + H22 )2 − 4(H11 H22 − |H12 |2 ) (7.72)
2 2
1 1p
= (H11 + H22 ) ± (H11 − H22 )2 + 4|H12 |2 . (7.73)
2 2
This is the exact solution for the two-level system. Below we will calculate the exact wave functions. First, however,
consider the following two interesting limiting cases.

1. If |H11 − H22 |  |H12 | we can expand (7.73) in powers of |H12 | yielding


 
1 1 2|H12 |2
E1,2 = (H11 + H22 ) ± (H11 − H22 ) 1 + + ... . (7.74)
2 2 (H11 − H22 )2

In particular,

|H12 |2 (0) |V12 |2


E1 = H11 + = E1 + V11 + (0) (0)
. (7.75)
H11 − H22 E1 + V11 − E2 − V22
(0) (0)
This coincides with the perturbation theory result (7.17) if |V11 |, |V22 |  |E1 − E2 |.
2. In the opposite case |H11 − H22 |  |H12 |, viz. when the levels are close, we get by expanding (7.73):
 
1 (H11 − H22 )2
E1,2 = (H11 + H22 ) ± |H12 | + + ... . (7.76)
2 8|H12 |

We observe that the two close levels split by 2|H12 | = 2|V12 | (see (7.70)). This is sometimes referred to as the
level repulsion. This means that under the perturbation V two close levels shift in the opposite directions.

88
§4 Variational method VII STATIONARY PERTURBATION THEORY

The exact correct wave functions can be computed by substituting (7.73) into one of the equations (7.68) or (7.69)
(the other equation is not linearly independent as is easy to verify) and solving for a1 or a2 . We find
a1 H12 H12
= = 1
p 1
. (7.77)
a2 E1,2 − H11 2 (H11 + H22 ) ± (H11 − H22 )2 + 4|H12 |2 − H11
2

Eq. (7.77) can be cast in a more friendly form by introducing parameter β such that
2H12
tan β = . (7.78)
H11 − H22
Employing the identity
p
2 tan β2 β 1 + tan2 β − 1
tan β = β
⇒ tan = (7.79)
1 − tan2 2
2 tan β

we can write
   
a1 β a1 β
= cot , = − tan . (7.80)
a2 E=E1 2 a2 E=E2 2

Referring to (7.65) we find two correct normalized eigenfunctions:


β (0) β (0) β (0) β (0)
ϕ1 = cos ψ1 + sin ψ2 , ϕ2 = − sin ψ1 + cos ψ2 . (7.81)
2 2 2 2
(0) (0)
If tan β  1, which occurs for β  1, (7.81) reduces to ϕ1 = ψ1 , ϕ2 = ψ2 . This indicates that the original
(0)
eigenfunctions ψ1,2 are the correct ones. In the opposite limit tan β  1 (degenerate levels), which occurs when
β ≈ π/2, (7.81) implies the maximum mixing:
1 (0) (0) 1 (0) (0)
ϕ1 = √ (ψ1 + ψ2 ) , ϕ2 = √ (−ψ1 + ψ2 ) . (7.82)
2 2
Compare (7.82) with (7.53),(7.54).

• Additional reading: Merzbacher, “Quantum Mechanics” 3rd edition, 8.3, 8.4.

§4. Variational method

In this section we discuss a different approximation technique for a discrete spectrum called the variational method.
This method is based upon the following theorem. For any normalized state |ψi in a Hilbert space, the expectation
value of the Hamiltonian Ĥ is not smaller than its lowest eigenvalue E0 , i.e.

E0 ≤ hψ|H|ψi , where hψ|ψi = 1 . (7.83)

To prove it expand ψ in a set of Hamiltonian eigenfunctions ϕn :



X ∞
X
ψ= an ϕn , |an |2 = 1 . (7.84)
n=0 n=0

Then,

X ∞
X
hψ|H|ψi = |an |2 En ≥ E0 |an |2 = E0 . (7.85)
n=0 n=0

Moreover, the equality in (7.83) is achieved only when E = E0 , therefore

E0 = minhψ|H|ψi , provided that hψ|ψi = 1 . (7.86)


ψ

89
§4 Variational method VII STATIONARY PERTURBATION THEORY

Suppose now that we take a set of different normalized ψ 0 s, and compute the corresponding expectation values
hψ|H|ψi’s. Eq. (7.86) tells us that the smallest expectation value is the best approximation for the ground state
energy.
In practice, one chooses a trial function containing N unknown parameters {αn }, n = 1, 2, . . . , N , viz. |ψ{αn }i =
ψ(q, {αn }) and computes the following expectation value
J({αn }) = hψ{αn }|H|ψ{αn }i . (7.87)
Minimum of J({αn }) is then computed as a solution of a set of N algebraic equations
∂J({αn })
= 0, n = 1, 2, . . . , N . (7.88)
∂αn
Let {αn0 } be N solutions to (7.88). The number J({αn0 }) gives an approximation of the ground state energy E0 , while
ψ(q, {αn0 }) gives an approximation to the ground state wave function. This method is know as the variational or Ritz
method.
The same idea can be used to find the first excited level. Let ψ0 be the ground state calculated as described above,
i.e. ϕ0 ≈ ψ0 . Energy of the first excited state E1 can be computed as a solution to the following minimization problem:
E1 = minhψ1 |H|ψ1 i , provided that hψ1 |ψ1 i = 1 , hψ1 |ψ0 i = 0 . (7.89)
ψ1

To prove this statement write



X ∞
X
ψ1 = bn ϕn , |bn |2 = 1 . (7.90)
n=1 n=1

Note absence of b0 due to the orthogonality condition hψ1 |ψ0 i = 0. To compute the second excited state, identify
ϕ1 ≈ ψ1 and solve the following problem
E2 = minhψ1 |H|ψ1 i , provided that hψ2 |ψ2 i = 1 , hψ2 |ψ0 i = hψ2 |ψ1 i = 0 . (7.91)
ψ1

A computation error accumulates as one proceeds to higher excited states. Taming this error is an important problem
in computational physics.
The variational method gives the upper limit of the energy levels. They coincide with the exact levels only if the
trial functions coincide with the exact eigenfunctions.

∗ Example. Consider one-dimensional harmonic oscillator


~2 d2 1
Ĥ = − + mω 2 x2 . (7.92)
2m dx2 2
2
We established in (3.98) that the stationary state wave functions behave at large x as ψ ∼ e−αx /2 . Since the ground
state wave function should not have any zeros (oscillation theorem), we may choose the trail wave function for ground
state as
1 2
ψ0 (x, α) = Ae− 2 αx , (7.93)
where A and α are unknown constants. Normalization condition fixes A:
 α 1/4
A= . (7.94)
π
Now, following (7.87) we compute
Z +∞  
1 ~2 α mω 2
J(α) = ψ0∗ (x, α)Ĥψ0 (x, α)dx = + . (7.95)
−∞ 4 m α
Minimum of J(α) can be computed from the condition
∂J(α)
= 0, (7.96)
∂α α=α0

90
§4 Variational method VII STATIONARY PERTURBATION THEORY

yielding

α0 = . (7.97)
~
Thus, we obtain the following approximation of the ground state energy and the wave function:

~ω  mω 1/4 mω 2
E0 = J(α0 ) = , ϕ0 (x) = ψ(x, α0 ) = e− 2~ x . (7.98)
2 ~π
This happens to be the exact energy and the wave function of the ground state.
Since the wave function of the first excites state must have one zero at x = 0 (because it must be odd), a reasonable
choice of the trial function is
1 2
ψ1 (x, β) = Bxe− 2 βx . (7.99)

Normalization condition fixes B:


 1/4
4β 3
B= . (7.100)
π

It is easy to check that hψ1 |ψ0 i = 0. Next,


Z +∞  
3 ~2 β mω 2
J1 (β) = ψ1∗ (x, β)Ĥψ1 (x, β)dx = + . (7.101)
−∞ 4 m β

Minimum of J1 is at β0 = mω/~. Hence


 1/4 
3 4 mω 3/4 − mω x2
E1 = J1 (β0 ) = ~ω , ϕ2 (x) = ψ2 (x, β0 ) = xe 2~ . (7.102)
2 π ~

• Additional reading: Arfken at. al., “Mathematical methods for physicists”, 7th edition, 22.4.

91
§1 Schrödinger equation in magnetic field VIII MOTION IN MAGNETIC FIELD

VIII. MOTION IN MAGNETIC FIELD

§1. Schrödinger equation in magnetic field

In classical mechanics motion of a charged particle in electromagnetic field is described by the following Hamiltonian
1  q 2
H= p − A + eΦ , (8.1)
2m c
where A and Φ are the vector and scalar potentials correspondingly and p is the canonical momentum. I am using
the Gauss units. Quantum mechanical generalization of this expression is know as the Pauli Hamiltonian:
1  q 2
Ĥ = p̂ − Â + q Φ̂ − µ̂ · B . (8.2)
2m c
The first two terms on the right-hand-side appear due to the requirement that in the classical limit (8.2) reduce to
(8.1). The third term is proportional to particle spin and thus has no classical analogue. In fact, as the classical
equation (6.46) implies, it is a relativistic correction because it is proportional to 1/c. In order to derive this term from
the first principles we need an equation respecting relativistic (boost) invariance. We will derive such an equation,
known as the Dirac equation in the case of spin-1/2 particle, later in this course and show how (8.2) can be obtained
as a non-relativistic limit of that equation, see XV §8.
Generally, operator of magnetic moment µ̂ of a particle with spin s is defined as follows
µ
µ̂ = ŝ , (8.3)
~s
where µ is a constant that depends on particle species. The eigenvalues of its z-projection are
µ s−1 s−1
µz = σ = −µ, −µ ...,µ ,µ. (8.4)
s s s
Here area few examples for spin-1/2 particles:
e~
electron: µe = − ≡ −µB , (8.5)
2me c
e~
proton: µp = −2.79 , (8.6)
2mp c
e~
neutron: µn = 1.91 . (8.7)
2mp c
Constant µB is called the Bohr magneton.
The gyromagnetic ratio is defined as a proportionality coefficient between the magnetic moment and the angular
momentum µ = gJ . For example, the gyromagnetic ratio of electron due to its spin is g = |µB |/(~/2) = e/(me c).
This is twice as big as the contribution of its orbital angular momentum.
Throughout this chapter I will always use the coordinate representation, which allows us to drop the operator sign
over the potentials. Keeping in mind that [p̂, Â] = −i~∇ · A 6= 0 we write
 q 2  q  q  q2 q
p̂ − A = p̂ − A p̂ − A = p̂2 + 2 A2 − (p̂ · A + A · p̂) . (8.8)
c c c c c
Using this in (8.2) yields

1 2 q q2 µ
Ĥ = p̂ − (p̂ · A + A · p̂) + 2
A2 − ŝ · B + qΦ . (8.9)
2m 2mc 2mc ~s

A meaningful physical quantity must be gauge-invariant. The means in particular, that although (8.2) and (8.9)
explicitly depend only on potential A and Φ, energy spectrum can depend only on the fields E and B. The following
gauge transformation does not change the fields:
1 ∂f
A → A0 = A + ∇f , Φ → Φ0 = Φ − , (8.10)
c ∂t

92
§2 Motion in constant magnetic field VIII MOTION IN MAGNETIC FIELD

where f is an arbitrary function. However, it does change the Hamiltonian unless the wave function ψ is simultaneously
transformed according to

ψ → ψ 0 = eiG ψ , (8.11)

where G is related to f . You will compute G in a home assignment. Since ψ and ψ 0 defer only by a phase, no physical
quantities change under the gauge transformation (8.10),(8.11).

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §111, Weinberg, “Lec-
tures on Quantum Mechanics”, 10.1,10.2.

§2. Motion in constant magnetic field

A. Landau levels

Gauge invariance allows us to choose the potentials in a most convenient way. There are two such gauges for a
constant magnetic field: the Landau gauge and the symmetric gauge. We will discuss the former here, whereas the
latter you will study in a home assignment. The Landau gauge is

Ax = −By, Ay = Az = Φ = 0 . (8.12)

It is easy to verify that ∇ · A = 0 and B = Bez . Schrödinger equation reads


" #
1  q 2 p̂2y p̂2z µ
p̂x + By + + − ŝz B ψ = Eψ . (8.13)
2m c 2m 2m ~s

Since [Ĥ, ŝz ] = [Ĥ, p̂x ] = [Ĥ, p̂z ] = 0 we can choose ψ’s to be eigenfunctions of ŝz , p̂x and p̂z . Namely,

ψ(x, y, z) = (2π~)−1 ei(px x+pz z)/~ ϕ(y) χσ , (8.14)

where χσ is a spinor. Then eq. (8.13) becomes


" #
1  q 2 p̂2y p2z µ
px + By + + − σB ϕ = Eϕ . (8.15)
2m c 2m 2m s

Denoting the y-derivative by prime we re-write it as follows


  
00 2m µσ p2z m 2 2
ϕ + 2 E+ B− − ω0 (y − y0 ) ϕ = 0 , (8.16)
~ s 2m 2

where we introduced a notation


cpx |q|B
y0 = − , ω0 = . (8.17)
qB mc

ω0 is called the cyclotron frequency. Mathematically, (8.16) is identical to the Schrödinger equation for one-dimensional
harmonic oscillator (3.95). Eq. (3.112) in our case becomes
 
µσ p2z 1
E+ B− = n+ ~ω0 . (8.18)
s 2m 2

Thus, we obtain the Landau levels of a particle in external constant magnetic field:
 
1 p2 µσ
E = n+ ~ω0 + z − B. (8.19)
2 2m s

93
§2 Motion in constant magnetic field VIII MOTION IN MAGNETIC FIELD

Note that only motion in the xy-plane is quantized (the first term), while motion along the z-direction (magnetic field
direction) is continuous. This is easy to understand if you recall that the Lorentz force vanishes in z-direction. The
normalized y-dependent part of the eigenfunctions is
(y−y0 )2
 
1 −
2a2
y − y0
ϕn (y) = 1/2 √ n
e 0 Hn , (8.20)
π 1/4 a0 2 n! a0
p
where a0 = ~/(mω0 ).
e~
An important particular case of (8.19) is when the charged particle is electron. In this case µ = −µB = − 2mc and
s = 1/2. Writing
µσ
− B = ~ω0 σ , (electron) , (8.21)
s
we can cast the Landau levels in the form
 
1 p2
E = n + + σ ~ω0 + z , (electron) . (8.22)
2 2m

B. Equations of motion

In classical mechanics a particle with pz = 0 moves in constant magnetic field along a circle of radius RL = v/ω0 ,
e
known as the Larmor orbit. Here v = vx ex +vy ey . Let r0 be the position of the orbit center and denote ω = − |e| ω0 ez .
We can then write

v = ω × (r − r0 ) , (8.23)

where r is a vector in the xy-plane. This is equivalent to two equations of motion


vy vx
x0 = x − , y0 = y + . (8.24)
ω ω
Their quantum analogue is
v̂y v̂x
x̂0 = x̂ − , ŷ0 = ŷ + . (8.25)
ω ω
The velocity operator reads (see HW)
1  q 
v̂ = p̂ − A . (8.26)
m c
Thus,
 
1  q  1 ∂ qB cp̂x
ŷ0 = ŷ + p̂x − Ax = ŷ + −i~ + y =− . (8.27)
mω c mω ∂x c qB

Operators of the orbit center and the orbit radius are r̂02 = x̂20 + ŷ02 and R̂L
2
= (v̂x2 + v̂y2 )/ω 2 respectively. It can be
shown that

[Ĥ, x̂0 ] = [Ĥ, ŷ0 ] = [Ĥ, r̂02 ] = [Ĥ, R̂L


2
] = [r̂02 , R̂L
2
] = 0. (8.28)

On the other hand,


i~c
[x̂0 , ŷ0 ] = − 6= 0 . (8.29)
eB
We conclude that x0 , y0 , r0 and RL are all conserved as in the Classical Mechanics. However, x0 and y0 cannot
simultaneously have definite values.

94
§2 Motion in constant magnetic field VIII MOTION IN MAGNETIC FIELD

C. Degeneracy

As we have seen px , pz and σ are conserved, while py is not. However, the spectrum (8.19) does not depend on px .
Thus it is degenerate with respect to the continuous variable px . Since it is cumbersome to deal with an infinite-fold
degenerate system, suppose that the entire system is in a big box with edges Lx , Ly , Lz . Then, px is a discrete variable
∆px
with the spectrum px = 2π~nx /Lx . The number of its different values in the interval ∆px is Lx2π~ . Now, the orbit
center must satisfy 0 < y0 < Ly , i.e. it must be inside the box. In view of (8.27), this implies that ∆px = eB c Ly .
Thus, the number of states with fixed n and pz is
Lx qB
Ly . (8.30)
2π~ c
The number of states with fixed n:
Lz ∆pz Lx qB ∆pz
Ly = qB V . (8.31)
2π~ 2π~ c (2π~)2 c
In the case of electron, there is an additional two-fold degeneracy with respect to its spin direction because E(n, σ =
1/2) = E(n + 1, σ = −1/2).

∗ Example.
A particle of spin-0, mass M and charge q constrained to move in the harmonic oscillator potential U = 21 M ω 2 r2 ,
is placed in constant magnetic field B. To compute its energy spectrum, it is convenient to employ a symmetric
gauge is A = 21 (B × r) and choose the cylindrical coordinates with axis z being the direction of magnetic field. The
Hamiltonian reads
   
~2 1 1 qB M ω2 q2 B 2
Ĥ = − ∂ρ ρ ∂ρ + 2 ∂φ2 + 2 L̂z + + ρ2 + Ĥz , (8.32)
2M ρ ρ c~ 2 8M c2
where
~2 2 M ω 2 2
Ĥz = − ∂ + z . (8.33)
2M z 2
Since [Ĥ, L̂z ] = [Ĥz , L̂z ] = [Ĥ, Ĥz ] = 0 we can choose eigenfunctions as
1 √
ψ(z, ρ, φ) = √ eimφ ϕn2 (z) ρf (ρ) , (8.34)

where ϕn2 (z) are the harmonic oscillator eigenfunctions with n2 = 0, 1, 2, . . . and m is the magnetic quantum number.
Schrödinger equation for f (ρ) takes form
 
00 2 0 2M E⊥ qBm m2 − 1/4 ρ2
f + f + + − − 4 f = 0. (8.35)
ρ ~2 c~ ρ2 4a
I denoted
r
|q|B ~
E⊥ = E − ~ω(n2 + 1/2) , ω0 = , a= (4ω 2 + ω02 )−1/4 . (8.36)
Mc M
All these algebraic manipulations were worthwhile because (8.35) is mathematically equivalent to the Schrödinger
equation for the radial part of the spherical harmonic oscillator (5.39). Thus we can write the spectrum right away:
q
~ω0 mq 2n1 + |m| + 1
E⊥,nm = − + ~ ω02 + 4ω 2 . (8.37)
2|q| 2

It is instructive to consider two limits. First, in the weak field limit ω0  2ω:
 
3 q~m q 2 ~(2n1 + |m| + 1) 2
En1 n2 m = ~ω 2n1 + |m| + n2 + − B+ B (8.38)
2 2M c 8M 2 c2 ω
q 2 ~(2n1 + |m| + 1) 2
= ~ω(N + 3/2) − µ̂ · B + B (8.39)
8M 2 c2 ω

95
§3 Time dependence of states VIII MOTION IN MAGNETIC FIELD

For the ground state N = 0

1 q2 ~
∆E0 = − χ0 B 2 , χ0 = − . (8.40)
2 4M 2 c2 ω
The diamagnetic susceptibility χ0 of the oscillator is an important experimental observable.
In the opposite limit of strong field ω0  2ω:

~ω 2
En1 n2 m ≈ E⊥,n + (2n1 + |m| + 1) + Ez,n2 . (8.41)
ω0
|m| qm
E⊥,n are the Landau levels with n = n1 + 2 − 2|q| . Ez,n2 is the energy of free vibrations along z-axis.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §112, Weinberg, “Lec-
tures on Quantum Mechanics”, 10.3.

§3. Time dependence of states

Motion of a charged particle with spin in an external electromagnetic field is usually complicated because spatial
and spin degrees of freedom are entangled. There is however one important case where separation of spatial and spin
variables is possible: when magnetic field B depends only on time, while electric field E is arbitrary. In this case we
seek a solution to the Pauli equation
 
∂ψ 1  e 2 µ
i~ = p̂ − A − B · ŝ + qΦ ψ (8.42)
∂t 2m c ~s

as a product of a single component function ϕ(r, t) and a spinor χ(t):

ψ(r, t) = ϕ(r, t)χ(t) . (8.43)

Notice, that the spinor depends only on time. We get


 
∂ϕ ∂χ 1  q 2 µ
i~ χ + i~ ϕ = χ p̂ − A + qΦ ϕ − ϕ B · ŝχ . (8.44)
∂t ∂t 2m c ~s

This equation can be satisfied if ϕ and χ obey the following equations:


 
∂ϕ 1  e 2
i~ = p̂ − A + eΦ ϕ , (8.45)
∂t 2m c
∂χ µ
i~ = − B · ŝχ . (8.46)
∂t ~s
Thus, spatial and spin variables decouple (separate) and we can consider dynamics of a particle spin independently
of its motion in the configuration space.

∗ Examples.

1. A spin-1/2 particle of magnetic moment µ moves in constant magnetic field B = Bez . Its spin part is governed
by (8.46) which in this case reads

∂χ
i~ = −µB σ̂z χ . (8.47)
∂t
a

Denote χ = b and ω = µB/~. Spinor is normalized so that |a|2 + |b|2 = 1. We have
( (
ȧ = iωa , a = a0 eiωt ,
⇒ (8.48)
ḃ = −iωb . b = b0 e−iωt .

96
§3 Time dependence of states VIII MOTION IN MAGNETIC FIELD

Expectation value of spin is


~σ̂
hs(t)i = χ† χ. (8.49)
2
Using the Pauli matrices (6.56) and (8.48) we arrive at
~ ∗ −2iωt 
hsx (t)i = a b0 e + a0 b∗0 e2iωt (8.50)
2 0
~ 
hsy (t)i = −ia∗0 b0 e−2iωt + ia0 b∗0 e2iωt (8.51)
2
~
hsz (t)i = (|a|2 − |b|2 ) . (8.52)
2
At t = 0
~ ∗
hsx (0)i = (a b0 + a0 b∗0 ) (8.53)
2 0
~
hsy (0)i = (−ia∗0 b0 + ia0 b∗0 ) (8.54)
2
~
hsz (0)i = (|a|2 − |b|2 ) . (8.55)
2
Excluding a∗0 b0 and a0 b∗0 in favor of hsx (0)i and hsx (0)i we an write
hsx (t)i = hsx (0)i cos(2ωt) + hsy (0)i sin(2ωt) , (8.56)
hsy (t)i = hsx (0)i cos(2ωt) − hsy (0)i sin(2ωt) . (8.57)
Spin precesses around z-axis with frequency 2ω.

2. Suppose now that the particle of the previous problem is subject to the following spatially uniform, periodic
magnetic field:
Bx = B1 cos(ω0 t) , By = B1 sin(ω0 t) , Bz = B0 . (8.58)
In this case as well we can consider only the spin part of the Hamiltonian:
!
B1 B0 e−iω0 t
Ĥ = −µB · σ = −µ . (8.59)
B0 eiω0 t −B1

Shrödinger equation (8.46) for each spinor component χ = ab reads

i~ȧ = −µB1 a − µBo e−iω0 t b , (8.60)


iω0 t
i~ḃ = −µB0 e a + µB1 b . (8.61)
To get rid of the time-dependent coefficients, make a substitution
a = e−iω0 t/2 a0 (8.62)
iω0 t/2 0
b=e b (8.63)
We arrive at
 
0 ~ω0
i~ȧ = − µB1 + a0 − µB0 b0 (8.64)
2
 
0 0 ~ω0
i~ḃ = −µB0 a + µB1 + b0 (8.65)
2

We are looking for a solution in the following form


a0 = CeiΩt , (8.66)
0 iΩt
b = De , (8.67)

97
§3 Time dependence of states VIII MOTION IN MAGNETIC FIELD

where C and D are constants. (It is evident that functions a0 and b0 must have the same frequency dependence,
for otherwise they won’t solve a system of linear equations.) Upon substitution we get
 
~ω0
µB1 + − ~Ω C + µB0 D = 0 , (8.68)
2
 
~ω0
µB0 C − µB1 + + ~Ω D = 0 . (8.69)
2

As have been discussed many times in this course, a non-trivial solution to the system of equations (8.68),(8.69)
exists iff:
 2
~ω0
− µB1 + + ~2 Ω2 = (µB0 )2 (8.70)
2
There are two solutions:
s 2
1 ~ω0
Ω=± µB1 + + (µB0 )2 ≡ ±ω . (8.71)
~ 2

Introduce a shorthand notation


µB1 ω0 µB0
γ1 = + , γ2 = . (8.72)
~ 2 ~
Thus,
q
ω= γ12 + γ22 . (8.73)

From (8.68) we find that


±ω − γ1
D= C. (8.74)
γ2
Correspondingly, we can write the general solution as a superposition of two solutions

a0 (t) = C1 eiωt + C2 e−iωt , (8.75)


ω − γ1 ω + γ1
b0 (t) = C1 eiωt − C2 e−iωt . (8.76)
γ2 γ2

Now we recall the initial condition: a(0) = 1, b(0) = 0. Solving for C1 and C2 we arrive at the following spinor
!
1 [(ω + γ1 )eiωt + (ω − γ1 )e−iωt ]eiω0 t/2
ψ(t) = . (8.77)
2ω 2iγ2 sin(ωt) eiω0 t/2

Spin-flip probability is given by

w = |hψ↓ |ψ(t)i|2 . (8.78)

where ψ↓ is the eigenfunction of ŝz with sz = −~/2. Finally, we find

γ22 2 B02
w= sin (ωt) = sin2 (ωt) . (8.79)
ω2 B02 + (B1 + ~ω0 /2µ)2
It is interesting to note that when B0  B1 the spin-flip probability is small for all ω0 except in a narrow range
around
2µB1
ω0res = − . (8.80)
~
This formula is used for experimental measurements of magnetic moment µ.

98
§1 Transition probability IX TIME-DEPENDENT PERTURBATION THEORY

IX. TIME-DEPENDENT PERTURBATION THEORY

§1. Transition probability

If a system is described by the time-dependent Hamiltonian Ĥ(t), then it does not have stationary states (since
energy is not conserved). Suppose that Ĥ(t) can be written as a sum of two terms

Ĥ(t) = Ĥ0 + V̂ (t) , (9.1)

where Ĥ0 is time-independent and V̂ (t) is a small time-dependent perturbation. If perturbation V̂ (t) is neglected,
then spectrum of Ĥ(t) coincides with that of Ĥ0 , which I will denote as En and the corresponding eigenfunctions as
ϕn . This is the zeroth order approximation of the time-dependent perturbation theory. Suppose that the perturbation
turns on at some initial time ti and switches off at tf . Since energy is not conserved, the initial state |ii is generally
not the same as the final state |f i. Therefore, the main problem of the time-dependent perturbation theory is to
compute the transition probabilities wi→f from the initial state |ii to the final state |f i. To this end one needs to
determine the time evolution of the wave function ψ from ti to tf , and then expand it in a set of ϕn ’s.
Time evolution of a quantum system is governed by the Schrödinger equation:
∂ψ
i~ = (Ĥ0 + V̂ (t))ψ . (9.2)
∂t
It is convenient to seek for a solution in the interaction picture, viz.
X En t
ψ(q, t) = an (t)ϕn (q)e−i ~ , (9.3)
n

where q stands for all degrees of freedom (coordinates, spins). Suppose that at t = ti the system was in a state with
Ei . Its wave function reads
Ei t
ψ(q, ti ) = ϕi (q)e−i ~ ⇒ an (ti ) = δni . (9.4)

Then the probability that the system has made a transition from state |ii to state |f i during time ∆t = tf − ti is
given by

wi→f = |af |2 . (9.5)

To calculate an (t) at t > ti substitute (9.3) into (9.2). We have


X iEn −i En t
 X X
−i E~n t En t En t
i~ ȧn (t)ϕn e − an (t)ϕn e ~ = an (t)ϕn En e−i ~ + an (t)V̂ ϕn e−i ~ . (9.6)
n
~ n n

Ef t R
Multiplying by ϕf ei ~ and integrating over all degrees of freedom dq we derive an equation for an ’s:

daf (t) X
i~ = an (t)eiωf n t Vf n , (9.7)
dt n

where

~ωf n = Ef − En , (9.8)

and
Z
Vf n = hϕf |V̂ |ϕn i = ϕ∗f V̂ ϕn dq . (9.9)

Treating V̂ as a small perturbation means that transition probabilities are small compared to one. We thus can
expand

an = δni + a(1) (2)


n + an + . . . (9.10)

99
§1 Transition probability IX TIME-DEPENDENT PERTURBATION THEORY

Loading this into (9.7) yields an equation for the first order correction to the amplitude
(1)
daf X
i~ = δni eiωf n t Vf n = eiωf i t Vf i . (9.11)
dt n

Its solution is
Z t Z t
(1) i 0 i 0
af (t) = − Vf i eiωf i t dt0 = − Vf i eiωf i t dt0 . (9.12)
~ ti ~ −∞

The total transition probability is


Z 2
(1) 1 +∞
iωf i t0 0
wi→f = 2 Vf i e dt , (9.13)
~ −∞

where I extended the limits of integration since the perturbation is non-vanishing only at ti ≤ t ≤ tf . The second
order correction satisfies the equation
(2)
daf X
= a(1)
n e
iωf n t
Vf n . (9.14)
dt n

Its solution is
Z Z t0
(2) 1 X t 0 0 iωf n t0 00
af (t) = 2 dt Vf n (t )e dt00 Vni (t00 )eiωni t . (9.15)
~ n −∞ −∞

A general solution can be written down using the time-ordering operator T̂ defined for any two wave functions Ψ(q, t)
and Φ(q, t) as follows:

T̂ {Ψ(q, t1 )Φ(q, t2 )} = Ψ(q, t1 )Φ(q, t2 )θ(t1 − t2 ) + Φ(q, t2 )Ψ(q, t1 )θ(t2 − t1 ) , (9.16)


where θ(t) is the Heaviside step-function. This definition can be generalized to a product of any number of wave
functions. Using this definition a formal solution to (9.7) reads
 Z 
i t 0 0 i i
af (t) = hf |T̂ exp − V̂int (t )dt |ii , V̂int (t) = e ~ Ĥ0 t V̂ e− ~ Ĥ0 t . (9.17)
~ −∞

V̂int (t) is the perturbation operator in the interaction picture.

∗ Examples.

1. Charged one-dimensional harmonic oscillator of mass m and frequency ω initially in state n is placed in electric
2 2
field E = E0 e−t /τ . Transition probability to a state n0 is given by (9.13) with
r r r !
~ n+1 n
Vn0 n (t) = −eE(t)xn0 n = −eE(t) δn0 ,n+1 + δn0 ,n−1 , (9.18)
mω 2 2

and ωn0 n = (En0 − En )/~ = ω(n0 − n). We have


Z 2
e2 ∞ iω(n0 −n)t

wn→n0 = E(t)e dt [(n + 1)δn0 ,n+1 + nδn0 ,n−1 ]
2~mω −∞
e2 πτ 2 −ω2 τ 2 (n0 −n)2 /2
= e [(n + 1)δn0 ,n+1 + nδn0 ,n−1 ] . (9.19)
2~mω
2. Charged plane rotator with electric dipole moment p in a stationary state with definite projection of its orbital
angular momentum |mi is placed in coplanar electric field E(t). Transition probability to a state with a different
state |m0 i is given by (9.13) with
pE
Vm0 m = hm0 | − pE cos φ|mi = − [δm0 ,m+1 + δm0 ,m−1 ] , (9.20)
2

100
§3 Periodic perturbations IX TIME-DEPENDENT PERTURBATION THEORY

and
~ ~
ωm0 m = [m2 − (m ∓ 1)]2 = (1 ± 2m) . (9.21)
2I 2I
where I used (6.38). Thus,
Z 2
p2 +∞ iωm0 m t

wm→m0 = 2 E(t)e dt . (9.22)
4~ −∞

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §40,41; Messiah, “Quan-
tum mechanics”, XVII, 1-3.

§2. Perturbations finite at t → ∞

In the previous section we considered a perturbation that acted only during the finite time-interval ti ≤ t ≤ tf .
Now consider perturbations that start at some time ti , but approach a constant at t → ∞. Eq. (9.13) cannot be
applied since the integral diverges at the upper limit. However, (9.12) applies even in the present case. We can cast
it in a more convenient form by integrating over time by parts:
Z t Z t
(1) i t Vf i eiωf i t ∂Vf i eiωf i t
af (t) = − Vf i eiωf i t dt = − + dt . (9.23)
~ −∞ ~ωf i −∞ −∞ ∂t ~ωf i

At large t the first term on the right-hand-side describes mere distortion of the wave function ϕi due to the fact that
at t → ∞ the Hamiltonian is Ĥ0 + V̂ (∞). Indeed, it coincides with the first order correction to the wave function
in the time-independent perturbation theory, see (7.15). Additional time-dependent factor eiωf i t appears in (9.23)
because we use here full time-dependent wave-functions, whereas in (7.15) the wave functions are time-independent.
Thus, the transition probability is determined only by the second term in (9.23):
Z 2
(1) 1 +∞ ∂Vf i iωf i t
wi→f = 2 2 e dt . (9.24)
~ ωf i −∞ ∂t
At the higher orders of the perturbation theory, the wave function distortion at t → ∞ must also be taken into
account.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §41.

§3. Periodic perturbations

An important class of time-dependent perturbations are periodic perturbations that can be written as
V̂ (t) = v̂± e±iωt , (9.25)
where T = 2π/ω is the period. By definition, the periodic perturbations are eternal (this is of course a mathematical
idealization). Here v̂ is time-independent. The transition probability from initial state i to final state f is given by
(9.13). Integral over time can be taken explicitly using (B11)
Z ∞
hf |v̂± |iie±iωt eiωf i t dt = 2πhf |v̂± |iiδ(ωf i ± ω) . (9.26)
−∞

There is however a problem how to interpret the square of the delta-function that comes about upon substitution into
(9.13). To make sense of it, we write it as the following limit:
Z 2 Z +τ /2
+τ /2
±iωt iωf i t
lim hf |v̂± |iie e 2
dt = 2π|hf |v̂± |ii| δ(ωf i ± ω) lim e±iωt eiωf i t dt
τ →∞ −τ /2 τ →∞ −τ /2
Z +τ /2
2
= 2π|hf |v̂± |ii| δ(ωf i ± ω) lim dt = 2π|hf |v̂± |ii|2 δ(ωf i ± ω) lim τ . (9.27)
τ →∞ −τ /2 τ →∞

101
§3 Periodic perturbations IX TIME-DEPENDENT PERTURBATION THEORY

This indicates that the transition probability is proportional to the observation time τ . This is not surprising because
periodic perturbation repeats itself infinite number of times. A physically meaningful quantity is transition probability
per unit time ẇi→f = wi→f /τ which is also called the transition rate. It reads

ẇi→f = |hf |v̂± |ii|2 δ(Ef − Ei ± ~ω) . (9.28)
~
We observe that under the perturbation V̂+ ∼ eiωt the system looses a quantum of energy Ef = Ei − ~ω, whereas
under the perturbation V̂+ ∼ e−iωt it gains a quantum of energy Ef = Ei + ~ω.
If the final state is a part of the continuous spectrum (i.e. system is not bound), then we are interested to know
the transition rate into one of the dνf states with energies in the interval Ef ± dEf /2. We can express the number
of states of continuous spectrum in terms of the density of states ρ defined as

dνf = ρ(Ef )dEf . (9.29)

Thus, the transition rate into a continuum reads


Z
2π 2π
ẇi→f = |hf |v̂± |ii|2 δ(Ef − Ei ± ~ω)ρ(Ef )dEf = |hf |v̂± |ii|2 ρ(Ei ∓ ~ω) . (9.30)
~ ~
Formulas (9.28) and (9.30) are called the Fermi golden rule.
It is important to remember that the wave functions ψν (q) of the final state are normalized as follows:
Z
ψν∗ (q)ψν (q 0 )dν = δ(q − q 0 ) , ν continuous , (9.31)
X
ψn∗ (q)ψn (q 0 ) = δ(q − q 0 ) , n discrete . (9.32)
n

∗ Example. Particle in the delta-well U = −αδ(x) (α < 0) is subject to the periodic perturbation

V̂ (x, t) = −xF cos(ωt) , (9.33)

For instance, charged particle in electric field, in which case F = qE. Suppose that the particle is in the ground state
(this is actually the only state). The transition rate into continuum reads
Z

ẇ0→f = |hf |v̂± |ii|2 δ(Ef − E0 ± ~ω)dνf . (9.34)
~
The initial and final states are discussed in III §2:
√ ~2 κ2 mα
ψi ≡ ψ0 = κ e−κ|x| , E0 = − , κ= 2 , (9.35)
2m ~
2 2
1 ~ k
ψf ≡ ψk = √ eikx , Ef = , νf = k . (9.36)
2π 2m
The matrix elements
Z Z +∞

1 1 √ i 2kκ3/2 F
hf |v̂± |ii = ψk∗ v̂± ψ0 dx = √ eikx (−xF ) κ e−κ|x| dx = − √ . (9.37)
2π −∞ 2 π(k 2 + κ2 )2

Since E0 < 0 and Ef > 0, Ef > E0 so that the only possibility is excitation Ef = E0 + ~ω0 described by v̂− . We have
Z
2π ∞ 2k 2 κ3 F 2
ẇ0→f = δ(Ef − E0 − ~ω)dk . (9.38)
~ 0 π(k 2 + κ2 )4
Using

1 p 2m
dk = d 2mEf = p dEf (9.39)
~ 2~ Ef

102
§4 Quasi-stationary states IX TIME-DEPENDENT PERTURBATION THEORY

in (9.38) and integrating over Ef yields

~F 2 p
ẇ0→f = ~ω − |E0 ||E0 |3/2 . (9.40)
m(~ω)4
This formula is valid for high frequencies ~ω  |E0 | because we neglected interaction in the final state. Indeed, since
mα2 /~2 ∼ |E0 |  Ef the transmission coefficient T ≈ 1 in (3.36).
In fact, the transition rate is non-zero even at ~ω < |E0 |, although exponentially suppressed. In static field (ω = 0)
we will discuss this in the next Chapter in X §4.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §42,77, Messiah, “Quan-
tum mechanics”, XVII,6.

§4. Quasi-stationary states

Stationary states, i.e. state with given energy, is an idealization that neglects interaction of a system with its
environment. In reality, even weak interaction leads to transition probability growing with time, so eventually any
state decays (if allowed by the conservation laws). If perturbation is periodic, then the life-time τ of a state i is
defined as inverse of the total decay rate:
X −1
τi = ẇi→f . (9.41)
f

States that are observed over times much shorter than τ are approximately stationary or quasi-stationary. Quasi-
stationary states do not have definite energy, as the ideally stationary states do, but have a small uncertainty ∆E  E.
It is intuitively clear that the larger is the life-time, the smaller is the uncertainty. Let’s determine a more precise
relationship by following the time-evolution of a quasi-stationary state.
Consider a quasi-stationary state described by the wave function ψ(q, t). At t = 0 we can expand it in complete set
of stationary states ψE (q) (assumed to be continuous for simplicity):
Z
ψ(q, 0) = CE ψE (q)dE , (9.42)

and normalized by
Z

ψE (q)ψE 0 (q)dq = δ(E − E 0 ) . (9.43)

The probability that the system has energy in the interval between E and E + dE is |CE |2 dE. If the state were
stationary with energy E0 then |CE |2 = δ(E − E0 ). Energy distribution in a quasi-stationary state can be described
using (B12) as follows:
1 Γ
|CE |2 = , (9.44)
2π (E − E0 )2 + Γ2 /4
where Γ/2  E0 . It is normalized as required
Z ∞
|CE |2 dE = 1 . (9.45)
−∞

This so-called Lorentzian or Breit–Wigner distribution (see the home assignment for more discussions). The wave
function has time dependence
Z
ψ(q, t) = CE ψE (q)e−iEt/~ dE . (9.46)

The probability to find the system in the approximately stationary state ψ(q, 0) at a later time is
Z 2 Z Z Z 2

∗ ∗ ∗
P (t) ≈ ψ (q, t)ψ(q, 0)dq = dq CE ψE (q)e iEt/~ 0
dE CE 0 ψE 0 (q)dE . (9.47)

103
§5 Sudden interactions IX TIME-DEPENDENT PERTURBATION THEORY

Using (9.43) we obtain


Z ∞ 2 Z ∞ 2
1 Γ
P (t) = 2 iEt/~
|CE | e dE = 2 2
e iEt/~
dE = e−tΓ/~ . (9.48)
−∞ −∞ 2π (E − E0 ) + Γ /4

At early times P (t) ≈ 1 − tΓ/~ implying that the life-time is τ = ~/Γ. According to (9.44) uncertainty in energy is
∆E ∼ Γ. Hence,

∆E τ = ~ . (9.49)

This is the relationship we were looking for. It means that the longer is the lifetime of a system the smaller is the
energy uncertainty of its quasi-stationary states, the closer they are to a true stationary state. Eq. (9.49) reflects a
general principle, that the longer the observation time, the better one can ascertain energy of the system, see references
below a derivation and detailed discussion of this statement.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §10,144, Merzbacher,
“Quantum Mechanics” 3rd edition, 3.5.

§5. Sudden interactions

Consider now perturbations that change rapidly with time. I am going to call them “sudden perturbations”.

A. Weak perturbations

The transition probability under a weak perturbation can be computed using the perturbation theory. Suppose that
the perturbation potential suddenly turns on and stays constant. In this case we can use Eq. (9.24). The perturbation
is sudden if the time derivative of Vf i is much faster function of time than eiωf i t . Hence, the exponent can be taken
out of the integral and we find that the transition probability under a weak sudden perturbation

|Vf i |2
wi→f = . (9.50)
~2 ωf2 i

We see that a perturbation is sudden if the time interval over Vf i changes is much shorter than 1/ωf i , i.e.

|V̇f i |
 |ωf i | . (9.51)
|Vf i |

B. Strong interactions

Formula (9.50) relies on the perturbation theory that assumes that the matrix elements Vf i are small. However,
there is a simple method to evaluate the transition probabilities even for strong (and sudden) interactions.
Suppose that a system at t < 0 is in a stationary state ψi of Hamiltonian Ĥ. At time t = 0 the Hamiltonian
suddenly changes and becomes Ĥ 0 . Denote the eigenfunctions and eigenvalues of the new Hamiltonian by ψf0 , Ef0 .
During a sudden interaction satisfying the condition (9.51), the initial wave function ψi does not significantly change.
Therefore, at t = 0+ the system is still in the state ψi , which however, is not an eigenstate of the new Hamiltonian
Ĥ 0 . We can expand it in the complete set of eigenfunctions
X
ψi = cf i ψf0 . (9.52)
f

The transition probabilities are determined as

wi→f = |hψf0 |ψi i|2 . (9.53)

104
§6 Adiabatic perturbations IX TIME-DEPENDENT PERTURBATION THEORY

∗ Example. β-decay is emission of electron or positron by an atomic nucleus in the ground state. For example:
14 −
6 C → 14 23 23 +
7 N + e + ν̄e , 12 M g → 11 N a + e + νe . From experimental data it is known p that typical kinetic energy of
2 2
emitted electron is E ' 0.2 MeV. Its full relativistic energy is ε = E + mc = mc / 1 − v 2 /c2 , from which velocity
3a
is v ' 0.7c. Now, the atomic radius (in the ground state n = 1) is 2Z (see V §5), so the time it takes electron to leave
3a
atom – we can call it the interaction time – is t ' 2vZ . This time should be compared to the characteristic time of
2 2
orbital motion 1/ω12 . Since ω12 = e 2a
Z
(1 − 1/4) (see (5.60)) we estimate

9 e2 Z Z
ω21 t ' = . (9.54)
16 v 171
For light nuclei ω21 t  1 and we can consider β-decay as a sudden interaction.
The ground state wave function of a hydrogen-like atom before the β-decay is
 3/2
Z
ψ1s =2 e−Zr/a Y00 . (9.55)
a

β-decay changes nuclear charge by unity Z → Z ± 1. So the new eigenfunctions are ψ 0 = Rnl
0
(r)Ylm , where e.g.
 3/2  
0 Z ±1 (Z ± 1)r −(Z±1)r/2a
R20 (r) = 2− e . (9.56)
2a a

Using (9.53) we can compute the probability of 1s → 2s transition:


Z ∞
Z 3/2 0 211 Z 3 (Z ± 1)3
c20,10 = 2 R20 (r)e−Zr/a r2 dr ⇒ w1s→2s = |c20,10 |2 = . (9.57)
a3/2 0 (3Z ± 1)8

We can compute the same probability using (9.50) because of large enough Z the perturbation V̂ = Ĥ 0 − Ĥ = ±e2 /r
is small:

|h1s|V̂ |2si|2 211


w1s→2s = 2 = , (9.58)
~2 ω12 94 Z 2

which agrees with (9.57) at Z  1.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §41.

§6. Adiabatic perturbations

Interactions are slow or adiabatic if the interaction time is large compared to ωf−1
i . If Vf i were time-independent we
would have used the stationary perturbation theory. This has been done in Ch. VII. In (9.13) adiabatic approximation
(1)
allows us taking the matrix elements Vf i out off the integral, which leads to a result that af ∝ δ(Ef − Ei ) implying
that no transition is possible. This is a trivial result since the perturbation is treated as constant in time and thus
energy is conserved. Since we are interested in time-dependent, though slow, perturbations, we need to develop a
more systematic approach.
Consider a system described by a Hamiltonian Ĥ(q, λ(t)) that depends on time only through a slowly varying
parameter λ(t) (e.g. external field). For simplicity we start with only one time-dependent parameter. Let the
spectrum of Ĥ(q, λ(t)) be discrete and non-degenerate for any fixed λ. The corresponding eigenfunctions are real and
satisfy the Schrödinger equation

[Ĥ(q, λ(t)) − En (λ(t))]ϕn (q, λ(t)) = 0 , (9.59)

and the orthonormality condition


Z
ϕn (q, λ)ϕk (q, λ)dq = δnk . (9.60)

105
§6 Adiabatic perturbations IX TIME-DEPENDENT PERTURBATION THEORY

Time-evolution of the wave function ψ(q, t) is governed by the Schrödinger equation

∂ψ(q, t)
i~ = Ĥ(q, λ(t))ψ(q, t) . (9.61)
∂t
In view of slow time-dependence of λ, we can try to solve this equation in the same way as we did for the stationary
states, see (2.26). Namely, if the time-dependence of the Hamiltonian is slow enough, the functions ϕn are a complete
set at a given λ, so that we can expand
X  Z t 
0 0
ψ(q, t) = ak (t)ϕn (q, λ) exp −i ωk (t )dt , (9.62)
k 0

where ωk (t) = Ek (t)/~. Substituting (9.62) into (9.59) we obtain


X ∂ak (t) X ∂ϕk (q, λ) −i R t ωk (t0 )dt0
ωk (t0 )dt0
Rt
i~ ϕk (q, λ)e−i 0 + i~ ak (t) e 0 = 0. (9.63)
∂t ∂t
k k

ωn (t0 )dt0
Rt
Multiplying by ϕ∗n ei 0 , integrating over q and using (9.60) we derive an equation for an (t):
∂an X ∂ Rt 0 0
i~ + i~ ak (t)hn| |kiλ̇ e−i 0 ωkn (t )dt = 0 , (9.64)
∂t ∂λ
k

where

ωkn (t) = ωk (t) − ωn (t) . (9.65)

Consider the diagonal matrix element


Z
∂ ∂
hn| |ni = ϕ∗n (q, λ) ϕn (q, λ)dq . (9.66)
∂λ ∂λ
Since ϕn is real we can write
Z Z
∂ 1 ∂ϕ2n ∂
hn| |ni = dq = ϕ2n dq = 0 . (9.67)
∂λ 2 ∂λ ∂λ
Eq. (9.68) takes form
∂an X0 ∂ Rt 0 0
+ λ̇ ak (t)hn| |ki e−i 0 ωkn (t )dt = 0 . (9.68)
∂t ∂λ
k


To calculate hn| ∂λ |ki for m 6= n take derivative with respect to λ of (9.59):
" #
∂ Ĥ(q, λ) ∂Ek (λ) ∂ϕk (q, λ)
− ϕk (q, λ) + [Ĥ(q, λ) − Ek (λ)] = 0, (9.69)
∂λ ∂λ ∂λ

Multiplying by ϕ∗n and integrating over q we get


Z Z Z
∂ Ĥ ∂ϕk ∂ϕk
ϕ∗n ϕk dq + ϕ∗n Ĥ dq − Ek ϕ∗n dq = 0 . (9.70)
∂λ ∂λ ∂λ

Since Ĥ is hermitian operator we can write the second term as


Z Z Z
∂ϕk ∂ϕk ∂ϕk
ϕ∗n Ĥ dq + (Ĥϕn )∗ dq = En ϕ∗n dq . (9.71)
∂λ ∂λ ∂λ
Using the bracket notation (9.70) becomes

∂ hn| ∂∂λ

|ki
hn| |ki = . (9.72)
∂λ Ek − En

106
§6 Adiabatic perturbations IX TIME-DEPENDENT PERTURBATION THEORY

In the first order of the perturbation theory, the transition amplitude from state i to state f is determined by
exactly same method as in IX §1:
Z Z
tf
∂ Rt 0 0
tf
hf | ∂∂λ

|ii −i R t ωf i (t0 )dt0
af i = − λ̇ hf | |ii e−i 0 ωf i (t )dt dt = λ̇ e 0 dt . (9.73)
ti ∂λ ti Ef − Ei

∗ Examples.

1. One-dimensional charged harmonic oscillator in slowly varying electric field E(t) is described by the Hamiltonian

p̂2 mω 2 x2
Ĥ = + − eE(t)x . (9.74)
2m 2
Electric field plays the role of parameter λ. Suppose that initially the oscillator is in the ground state. The
probability that it is excited to state f is
Z +∞
dE hf | ∂∂E

|0i −i R t ωf i (t0 )dt0
af 0 = e 0 dt . (9.75)
−∞ dt Ef − Ei

To find the eigenfunctions of the Hamiltonian at fixed t we write (9.74) in a different form
 2
p̂2 mω 2 eE(t) e2 E 2
Ĥ = + x− − . (9.76)
2m 2 mω 2 2mω 2

Let the eigenfunctions of the free oscillator be ϕosc


n (x). Since the last term in (9.76) is independent of x, the
eigenfunctions and eigenvalues of Ĥ are
 
osc eE(t) e2 E 2
ϕn x− , En = ~ω(n + 1/2) − . (9.77)
mω 2 2mω 2

∂ Ĥ
Using ∂E = −ex we get:
Z ∞     r
∂ Ĥ eE(t) eE(t) ~
hf | |0i = ϕosc
n x− (−ex)ϕosc
0 x− dx = −ehf |x|0i = −e δf 1 . (9.78)
∂t −∞ mω 2 mω 2 2mω

The only non-vanishing transition amplitude becomes


r Z +∞
~ 1 dE −iωt
a10 = −e e dt . (9.79)
2mω ~ω −∞ dt

If electric field vanishes at t → ±∞, one can integrate by parts which finally yields
Z 2
e2 a2 +∞ −iωt

w0→1 = |a10 |2 = 2 E(t)e dt . (9.80)
2~ −∞

Compare this with (9.19).


2. A heavy particle of charge q passes at distance b from an atom in the ground state. As the result of the Coulomb
interaction atom can be excited. We can calculate the excitation probability using (9.13).
The collision geometry is depicted in Fig. 10 where the trajectory of the charge q is shown as the straight line.
This is because I assumed that its kinetic energy is much larger than the interaction energy. Vector R = vtex +ey
is the particle position. Let ra be positions of Z electrons in atom. The interaction energy is given by
qeZ X qe
V̂ (t) = − . (9.81)
R a
|R − ra |

107
§6 Adiabatic perturbations IX TIME-DEPENDENT PERTURBATION THEORY

R b

x
Ze

FIG. 10:

If the impact parameter is much bigger than the Bohr radius b  a we cab expand
qeZ X qe X eq(vtxa + bya ) X eq(vtxa + bya )
V̂ (t) ≈ − − = − . (9.82)
R a
R a
R3 a
R3

Its matrix element is


eZq(vthf |x̂|0i + bhf |ŷ|0i)
hf |V̂ |0i = − . (9.83)
R3
where
Z Z
hf |x̂|0i = ϕ∗f xϕ0 d3 r , hf |ŷ|0i = ϕ∗f yϕ0 d3 r , (9.84)

with ϕn (r) being stationary states of electron in atom. The transition probability is given by (9.13) as
Z 2
(Zeq)2 +∞ (vthf |x|0i + bhf |y|0i) iωf 0 t
w0→f = e dt . (9.85)
~2 −∞ [(vt)2 + b2 ]3/2
The t-integral can be taken with the result

(Zeq)2 4ωf2 0  
w0→f = |hf |x|0i|2 K02 (bωf 0 /v) + |hf |y|0i|2 K12 (bωf 0 /v) , (9.86)
~2 v4
where K0 and K1 are the modified Bessel functions.
Let’s discuss the two limiting cases. Inspecting (9.85) we see that at large t the transition probability decreases
at least as t−4 . The largest contribution to the t-integral comes from times for which the denominator in the
integrand is smallest, i.e. when the heavy particle is at the shortest distance from the atom. This implies that
−1
the interaction time is tint ∼ b/v. Thus, the interaction is adiabatic if tint  ω0f , i.e. ω0f b/v  1. If this
condition is satisfied, then the integrand is a rapidly oscillating function implying that the probability to excite
the atom is exponentially small. Indeed, using the asymptotic form of the modified Bessel functions at large
values of it argument, one can approximate (9.86) as

(Zeq)2 4ωf2 0 vπ −2bωf 0 /v  


w0→f ≈ e |hf |x|0i|2 + |hf |y|0i|2 . (9.87)
~2 v 4 2bωf 0

In the opposite case ω0f b/v  1 interaction happens very rapidly, hence eiωf 0 t ≈ 1 during the collision. Thus,
taking the t-integral in (9.86) yields
 2
2Zeq
w0→f ≈ |hf |ŷ|0i|2 . (9.88)
~bv

The same result is reached by expanding (9.86).

• Additional reading: Messiah, “Quantum mechanics”, XVII,7-14.

108
§7 Born-Oppenheimer approximation IX TIME-DEPENDENT PERTURBATION THEORY

§7. Born-Oppenheimer approximation

Consider two interacting systems #1 and #2 described by the Hamiltonian

Ĥ(q, ξ) = Ĥ1 (q) + Ĥ2 (ξ) + V̂ (q, ξ) , (9.89)

where Ĥ1 (q) and Ĥ2 (ξ) are Hamiltonians of free systems, q and ξ are coordinates of all particles in respective systems,
and V̂ (q, ξ) is interaction operator. The eigenvalue problem is

Ĥ(q, ξ)ΨN (q, ξ) = EN ΨN (q, ξ) . (9.90)

Suppose that the system #2 is “slow”, meaning that the periods of motion of particles in system #2 are much
larger than those in system #1. The spectrum of Ĥ can be determined using the method inspired by the adiabatic
approximation of the previous section and known as the Born-Oppenheimer approximation.
Consider the following Hamiltonian

Ĥ 0 = Ĥ1 (q) + V̂ (q, ξ) . (9.91)

The main idea of the the Born-Oppenheimer approximation is to treat the coordinates ξ of system #2 as slow varying
parameters of Ĥ 0 . At fixed ξ we consider the eigenvalue problem (as in (9.59)):

Ĥ 0 (q, ξ)ϕn1 (q, ξ) = En1 (ξ)ϕn1 (q, ξ) . (9.92)

The exact eigenfunctions of the Ĥ can be expanded in terms of ϕn1 (q, ξ). Moreover, since under adiabatic perturbation
by Ĥ2 (ξ) transition probabilities are small, we can approximate

ΨN (q, ξ) ≈ Φn1 n2 (ξ)ϕn1 (q, ξ) . (9.93)

We will see below, in (9.97), that the coefficients Φn1 n2 (ξ) are eigenfunctions of effective Hamiltonian for adiabatic
degrees of freedom. Substituting into (9.90) yields

[Ĥ1 (q) + V̂ (q, ξ) + Ĥ2 (ξ)]Φn1 ,n2 (ξ)ϕn1 (q, ξ) = En1 n2 Φn1 ,n2 (ξ)ϕn1 (q, ξ) . (9.94)

Because ξ is a slow variable in ϕn1 (q, ξ) we have in the same approximation



Ĥ2 (ξ) Φn1 ,n2 (ξ)ϕn1 (q, ξ) ≈ Φn1 ,n2 (ξ)Ĥ2 (ξ)ϕn1 (q, ξ) . (9.95)

Using (9.95) in (9.94) leads to

En1 (ξ)Φn1 ,n2 (ξ)ϕn1 (q, ξ) + Φn1 ,n2 (ξ)Ĥ2 (ξ)ϕn1 (q, ξ) = En1 n2 Φn1 ,n2 (ξ)ϕn1 (q, ξ) . (9.96)

Multiplying by ϕ∗n1 and integrating over q we derive an eigenvalue problem for the “slow” part of the wave function
 
Ĥ2 (ξ) + En1 (ξ) Φn1 n2 (ξ) = En1 n2 Φn1 n2 (ξ) . (9.97)

Solving this equation we determine the energy levels of the entire system. Notice that in (9.97) energy levels of the
“fast” system En1 (ξ) play the role of the effective potential for the “slow” system.

∗ Example. Consider two coupled oscillators of very different masses M  m:

p̂2x p̂2y k
Ĥ = + + (x2 + y 2 ) + αxy , |α| < k . (9.98)
2m 2M 2
Oscillator with small mass m is fast. Consider the Hamiltonian for the fast degrees of freedom (9.91):

p̂2x kx2
Ĥ 0 = + + αxy . (9.99)
2m 2
Treating y is a slow varying parameter we solve the eigenvalue problem for (9.99) as we did in (9.74),(9.76),(9.77):
 αy  α2 y 2
ϕn1 (x) = ϕosc
n1 x− , En1 = ~ω(n1 + 1/2) − , (9.100)
k 2k

109
§8 Berry phase IX TIME-DEPENDENT PERTURBATION THEORY

where ω 2 = k/m. The “slow” degrees of freedom are described by the Hamiltonian

p̂2y ky 2
Ĥ2 (y) = + . (9.101)
2M 2
and are governed by (9.97) which reads using notation of this example:
" #
p̂2y ky 2 α2 y 2
+ + ~ω(n1 + 1/2) − Φn1 n2 (y) = En1 n2 Φn1 n2 (y) ⇒ (9.102)
2M 2 2k
"   #
p̂2y k α2
+ 1 − 2 y 2 Φn1 n2 (y) = [En1 n2 − ~ω(n1 + 1/2)] . (9.103)
2M 2 k

Finally, spectrum of the entire system is


s  
m α2
En1 n2 = ~ω(n1 + 1/2) + ~ω 1 − 2 (n2 + 1/2) . (9.104)
M k

• Additional reading: S. Weinberg, “Lectures on Quantum Mechanics”, 5.6.

§8. Berry phase


We argued before that if the eigenfunctions ϕn are real, then the matrix element hn| ∂λ |ni vanishes. This is not
always the case, the simplest example being motion in magnetic field (see property 6 in II §2 and examine (8.1),(8.13)).
In this section we investigate its contribution to the wave function. In light of our discussion in the previous section,

the matrix element hn| ∂λ |ni contributes to the wave function of the n’th state, rather than describing the transition
probability. In fact, since the transition probability in the adiabatic case is small I am going neglect it in this
section. (Smallness of the transition probabilities stems from the presence of rapidly oscillating factors e−iωnm t ).
Thus, coefficient an satisfies the following equation

ȧn + λ̇an hn| |ni = 0 , (9.105)
∂λ
which is solved by
|nidt0
Rt ∂
an (t) = an (0)e− 0
λ̇hn| ∂λ
. (9.106)

Thus, the wave function in the presence of the adiabatic perturbation reads
X Rt 0 0
ψ= an (0)ϕn (q, λ)e−i 0 ωn (t )dt eiγn , (9.107)
n

where I defined the geometric phase (known also as the topological phase) γn :
Z t

γn = i λ̇hn| |nidt0 . (9.108)
0 ∂λ
Rt
In contrast, 0 ωn (t0 )dt0 is called the dynamical phase. A key property of γn is that it does not depend on time, but
rather on the path through the parameter space of the Hamiltonian from the initial point λ(0) to the final point λ(t).
Indeed, we can write (9.108) as
Z λ(t) Z
∂ ∂
γn = i hn| |nidλ = i hn| |nidλ , (9.109)
λ(0) ∂λ C(t) ∂λ

where C is a path connecting the initial and final points. The geometric phases are real (and therefore the term
“phase” is appropriate). This can be seen from the following equations:

0 = ∇λ hn|ni = h∇λ n|ni + hn|∇λ ni = hn|∇λ ni∗ + hn|∇λ ni = 2Re hn|∇λ ni . (9.110)

110
§8 Berry phase IX TIME-DEPENDENT PERTURBATION THEORY

For a closed path C, the geometric phase is called the Berry phase. I will denote it as γn [C]:
I

γn [C] = i hn| |nidλ . (9.111)
C ∂λ

We can generalize (9.105)-(9.109) for a system with N adiabatic parameters λ1 , . . . λN . To this end it is useful to
introduce N -dimensional parameter space, in which vector λ = {λ1 , . . . λN } denotes the instantaneous values of all
parameters at some time. Then, for example, the Berry phase reads
I
γn [C] = i hn|∇λ |ni · dλ . (9.112)
C

Referring to (9.107) we conclude that the Berry phase is determined modulo 2π, i.e. adding to it 2πk, with any integer
k, won’t change the wave function. Defining the Berry connection

An (λ) = ihn(λ)|∇λ |n(λ)i , (9.113)

we can write (9.112) as


I
γn [C] = An (λ) · dλ . (9.114)
C

An (λ) is a vector in the parameter space.


Seemingly, one can eliminate the geometric phases by transforming the eigenfunctions

ϕn (q, λ) → ϕ0n (q, λ) = eifn (λ) ϕn (q, λ) , (9.115)

where fn (λ)’s are chosen to cancel γn ’s in (9.107). If this were the case, then the geometric phases would have been
merely a matter of convention. However, the geometric phases γn can be eliminated only if the system is transported
along an open path in the parameter space. If it is transported along the closed path, in which cases the geometric
phases are the Berry phases γn [C], they do not change under the transformation (9.115) and thus cannot be eliminated.
Indeed, the Berry connection transforms as

An (λ) → A0n (λ) = ihn(λ)|e−ifn ∇λ eifn |n(λ)i = An (λ) − ∇λ fn (λ) , (9.116)

while the geometric phases as


Z
γn → γn0 = γn − ∇λ fn (λ) · dλ = γn − [fn (λ(T )) − fn (λ(0))] , (9.117)
C(t)

where T is the time at which the system arrives at the final state. For a closed path, the initial and the final state
are identical implying that
I
γn [C] → γn0 [C] = γn [C] − ∇λ fn (λ) · dλ = γn [C] . (9.118)
C

Thus, the Berry phases cannot be eliminated through the transformation (9.115).(16 This is the main observation of
this section. The “gauge” transformation (9.115) transforms the wave function to another equivalent one within a
certain class of γn .
There is an intimate analogy between the equations (9.115),(9.116) and the gauge transformations in electrody-
namics. If the potentials ϕ, A (17 and the wave function ψ are simultaneously transformed as
ie
ϕ → ϕ0 = ϕ + f˙/c , A → A0 = A − ∇f , ψ → ψ 0 = e ~c f ψ , (9.119)

(16 In derivation of (9.118) I assumed that f is a single-valued function. In fact the Berry phase is gauge-invariant even if f is not a
single-valued function but such that fn (λ(T )) − fn (λ(0)) = 2πN , N ∈ Z. See Sec. IX §9.
(17 Do not confuse ϕ, A and ϕn , An !

111
§8 Berry phase IX TIME-DEPENDENT PERTURBATION THEORY

where f as an arbitrary function, then equations of motion of electrodynamics do not change. This is because ϕ, A
and ψ are not observable quantities, unlike the fields E, B and expectation values of operators. Only gauge invariant
quantities (which do not depend on a choice of f ) can be experimentally measured. Similarly, ϕn and An (λ) are not
observable because they are gauge-dependent, whereas γn [C] is observable because it is gauge-independent. As in
electrodynamics, where magnetic field B = ∇ × A is observable, we can define the Berry curvature
∂An,j ∂An,i
Ωn = ∇λ × An (λ) , or Ωn,ij = − . (9.120)
∂λi ∂λj

which is a physical observable because it is gauge-independent.

FIG. 11:

The Berry curvature satisfies the Chern theorem that states that for any closed surface S
I
Ωn · dS = 2πN , N ∈ Z . (9.121)
S

To prove it divide the surface S into two open parts S1 and S2 that connect along the curve C, see Fig. 11. According
to the Stokes theorem
Z I
Ωn · dS = An (λ) · dλ = γn [C] , (9.122)
S1 C

where the surface S1 is bounded by the curve C that runs counter-clock-wise with respect to the outward normal
vector to S1 . Applying the Stokes theorem to the surface
R S2 and noting that the outward normal vector to it points
in the direction opposite to that of S1 , we find that S2 Ωn · dS = −γn [C]. Bearing in mind that γn [C] is defined
modulo 2πN , we obtain
I Z Z
Ωn · dS = Ωn · dS + Ωn · dS = 2πN , N ∈ Z . (9.123)
S S1 S2

N is also called the winding number due to the role it plays in topology.

∗ Example. Find the adiabatically stationary states of a spin-1/2 particle in magnetic field slowly precessing with
frequency ω, see Fig. 12.
In Cartesian coordinates the magnetic field is given by

B(t) = B{sin θ cos(ωt)ex + sin θ sin(ωt)ey + cos θez } . (9.124)

Interaction of spin with magnetic field is given by


!
cos θ e−iωt sin θ
Ĥ(t) = µB(t) · σ̂ = iωt
, (9.125)
e sin θ − cos θ

where µ is a constant (magnetic moment). For a slowly rotating magnetic field, particle spin has sufficient time to
turn either parallel or anti-parallel to the direction of magnetic field at a given time. Hence, in accordance with

112
§8 Berry phase IX TIME-DEPENDENT PERTURBATION THEORY

FIG. 12:

the adiabatic approximation, we find the instantaneous eigenvalues E± = ±µB and eigenstates of the Hamiltonian
corresponding to spin parallel and anti-parallel to magnetic field:
! !
cos(θ/2) − sin(θ/2)
|n+ (t)i = , |n− (t)i = . (9.126)
eiωt sin(θ/2) eiωt cos(θ/2)

These expressions are correct up to a phase that we proceed to calculate now (and show that it is non-trivial).
To calculate the Berry connection (9.113) we need the gradient operator in spherical coordinates ∇ = er ∂r +
eθ r−1 ∂θ + eφ (r sin θ)−1 ∂φ . The parameter space λ is a 3D space span by the components of vector B. In other words,
θ and φ are angles in the B-space and r = B. We have
! !
1 − sin(θ/2) 1 0
∇B |n+ (t)i = eθ + eφ , (9.127)
2r eiωt cos(θ/2) r sin θ ieiωt sin(θ/2)
! !
1 cos(θ/2) 1 0
∇B |n− (t)i = − eθ + eφ . (9.128)
2r eiωt sin(θ/2) r sin θ ieiωt cos(θ/2)

Then,
sin2 (θ/2) cos2 (θ/2)
A+ = − eφ , A− = − eφ (9.129)
r sin θ r sin θ
Finally, the Berry phases are
Z
γ+ [C] = A+ · eφ r sin θdφ = −2π sin2 (θ/2) = −π(1 − cos θ) , (9.130)
Z
γ− [C] = A− · eφ r sin θdφ = −2π cos2 (θ/2) = −π(1 + cos θ) = π(1 − cos θ) − 2π . (9.131)

This result can be expressed in terms of the solid angle o swept by the vector B:
Z 2π Z θ
o= dφ dθ0 sin θ0 = 2π(1 − cos θ) (9.132)
0 0

as follows
o
γ± [C] = ∓ mod 2π . (9.133)
2
Eqs. (9.130)-(9.133) show that the Berry phase depend only on the geometry (angle θ) of the problem, but not on the
dynamical quantities (magnetic field, time, etc.).
Let us now verify the Chern theorem (9.121)
 
1 ∂ 1 ∂ 1
Ω+ = ∇ B × A + = (sin θA+ )er − (rA+ )eθ = − 2 er . (9.134)
r sin θ ∂θ r ∂r 2r

113
§8 Berry phase IX TIME-DEPENDENT PERTURBATION THEORY

This is known as the magnetic monopole field due to the analogy between the Berry curvature and the magnetic field.
Flux of the curvature through a sphere of radius r reads
I
4πr2
Ω+ · dS = − 2 = −2π (9.135)
2r
H
as required. One can check that Ω− · dS = 2π.
To experimentally verify that the Berry phases are indeed observable, one can prepare a beam of polarized neutrons
and the split it into two beams: one travels in the adiabatically rotating magnetic field and another in a constant
field. The two beams are then recombined on a screen. The resulting wave function is
1
ψ = √ (|n+ i + |n+ ieiγ+ )e−iµBt/~ . (9.136)
2
The corresponding intensity (at equal distance from both fields) is
1 2 nπ o
I = I0 1 + eiγ+ = I0 cos2 (1 − cos θ) . (9.137)
2 2
If θ = 0 the phase difference between the two beams is the same, implying the maximum I0 of the interference pattern.
However, when θ 6= 0, there is a phase difference between the beams implying that I < I0 . This indicates shift of the
interference picture. This is illustrated below in Fig. 14 for the Aharonov-Bohm effect, but the idea is the same.

• Additional reading: S. Weinberg, “Lectures on Quantum Mechanics”, 6.6, 6.7, B. R. Holstein, “The adiabatic
theorem and the Berry’s phase”, Am. J. Phys. 57, 1079, 1989.

§9. Aharonov-Bohm effect

Geometric/topological phases similar to the Berry phase discussed in the previous section appear in many areas of
physics. Historically, the first such phase was discovered by Aharonov and Bohm.
Consider constant magnetic field restricted to a cylindrical region of size a, see Fig. 13. Since B = 0 outside the

FIG. 13:

cylindrical region, the most general expression for the vector potential there is A = ∇f , where f is an arbitrary
function. Indeed, B = ∇ × (∇f ) = 0. In such a case one says that magnetic field is a pure gauge. To determine f
note that flux of magnetic field Φ through the circular area in the yz-plane depicted in Fig. 13 is Φ = πa2 B 6= 0. On
the other hand, since the boundary of the circle lies in the region of vanishing B,
Z I I
Φ = B · dS = A · d` = ∇f · d` = f (2π) − f (0) . (9.138)

One possible chose of f satisfying (9.138) is


φ
f= Φ, (9.139)

114
§9 Aharonov-Bohm effect IX TIME-DEPENDENT PERTURBATION THEORY

where φ is a polar angle in the yz-plane. The corresponding vector potential is


Φ
A= ∇φ . (9.140)

Note, that although f is not a single-valued function, the difference f (2π) − f (0) is gauge-invariant because flux Φ is
obviously gauge-invariant, see (9.138). Therefore, taking this phase factor in (9.141) around the closed loop gives a
gauge-invariant result, see footnote (16.
Now we can determine the wave function of incident electrons moving along the z-axis with momentum ~k. If
A = 0 then the wave function is ψ = eikz . Performing the gauge transformation (9.119) with f given by (9.139) gives
the vector potential (9.140) and simultaneously transforms the wave function to
ie Φ
ψin = eikz e ~c 2π φ . (9.141)
The wave function of the outgoing electrons can be obtained from (9.141) in two ways: either by taking it clockwise
or anti-clockwise around the magnetic field region. This gives two wave functions
ieΦ
ψout = eikz e∓ 2~c , y > 0 (y < 0) . (9.142)
Evidently, y = 0 is a singularity of the wave function. The intensity of the outgoing wave at y = 0 (at the center of
the screen in Fig. 14) is
   
1 ieΦ − ieΦ
2
1 eΦ 2 eΦ
I = I0 e 2~c +e 2~c = I0 1 + cos = I0 cos . (9.143)
4 2 ~c 2~c
We see that although B = 0 at x2 + y 2 > a2 , presence of the finite potential A has an observable effect, although A
itself is of course not an observable quantity.

FIG. 14:

∗ Example: Dirac charge quantization. We can replace the magnetic field shown in Fig. 13 by a field of a magnetic
monopole of charge g:
g
Bm = er , (9.144)
r2
where r is a distance from the monopole in the Euclidean space. Flux of Bm through a sphere of radius r is
g
Φ= 4πr2 = 4πg . (9.145)
r2
If electron happen to travel by the magnetic monopole, its wave function would acquire a phase factor
ie Φ
e ~c 2π φ . (9.146)

115
§9 Aharonov-Bohm effect IX TIME-DEPENDENT PERTURBATION THEORY

The phase difference between electron traveling clockwise and counterclockwise is


e 4πge
Φ= . (9.147)
~c ~c
4πge
Since magnetic monopoles have not been observed, we must require that ~c = 2πn, n ∈ Z. Thus, we obtain the
quantization consition
ge n
= , n∈Z, (9.148)
~c 2
which Dirac suggested as a possible explanation of the discreteness of electric charge. We can view (9.148) as an
equation for the minimum magnetic charge
~c
gD = , (9.149)
2e
known as the Dirac magnetic charge.
Whether or not magnetic monopoles exist in nature is still an open question.

• Additional reading: S. Weinberg, “Lectures on Quantum Mechanics”, 10.4.

116
§1 The classical limit of quantum mechanics X WKB METHOD

X. WKB METHOD

§1. The classical limit of quantum mechanics

We have discussed the classical limit of quantum mechanics in I §9 and II §3. There we showed that this limit
is achieved in slowly varying fields and at large values of momentum. Our goal in this chapter is to develop an
approximate method of solving quantum mechanical problems for such almost classical systems.
Similarly to (1.134) we write the wave function of a particle moving in arbitrary external field as
i
ψ(r, t) = e ~ S(r,t) . (10.1)

In the classical limit the function S(r, t) becomes the classical action. Substitute (10.1) into the Schrödinger equation

∂ψ ~2 2
i~ =− ∇ ψ + Uψ (10.2)
∂t 2m
to obtain an equation for S:

∂S (∇S)2 i~ 2
− = +U − ∇ S. (10.3)
∂t 2m 2m
Let’s compare this equation with the classical Hamilton-Jacobi equation

∂Scl (∇Scl )2
− = +U, (10.4)
∂t 2m
where Scl is the classical action. The meaning of (10.4) is that the particle trajectory is orthogonal to the surface
Scl = const, it’s momentum being p = ∇Scl .
Comparing (10.3) and (10.4) we conclude that transition from the quantum to classical mechanics happens when
the term proportional to ∇2 S is small. Equivalently, one can say that the classical limit corresponds to ~ → 0.(18 In
a stationary state we can write

S(r, t) = σ(r) − Et , (10.5)

where σ is a new unknown function. Eq. (10.3) now reads

(∇σ)2 i~ 2
+U −E− ∇ σ = 0. (10.6)
2m 2m
We are going to seek a solution to this equation in a form of expansion near the classical limit

σ = σ0 + σ1 + σ2 + . . . , σ 0  σ1  σ2 . . . (10.7)

where σ0 satisfies the classical equation

(∇σ0 )2
+U −E =0 (10.8)
2m
and hence σ0 ∼ O(~0 ), while σ1 ∼ O(~1 ), σ2 ∼ O(~2 ), etc. The classical momentum is p = ∇σ and thus can be
written as
p
p = 2m(E − U ) . (10.9)

In order that the expansion (10.7) be convergent the following condition must be satisfied

(∇σ0 )2  ~|∇2 σ0 | . (10.10)

(18 Keep in mind though that in reality ~ is a constant. Saying that ~ → 0 means that the terms it multiplies are small.

117
§2 The wave function in the quasi-classical approximation X WKB METHOD

In terms of p (10.10) reads p2  ~|∇ · p|, while for the de Broglie wavelength λ = ~/p we get simply

|∂x λ|  1 , (10.11)

where x stands for any Cartesian coordinate. This means that the de Broglie wavelength must be a slow function of
coordinates. In optics this limit corresponds to the geometric optics. For a given potential and using (10.9), we can
express this condition as

p3  ~m|∇U | . (10.12)

If these inequalities hold, we can develop an approximation method knows as the quasi-classical or the semi-classical
or the WKB (Wentzel-Kramers-Brillouin) approximation.

§2. The wave function in the quasi-classical approximation

Under the conditions outlined in the previous section we can compute approximations σ1 , σ2 etc. Substituting
(10.7) into (10.6) in expanding to second order in ~2 we obtain
1   i~  2 
(∇σ0 )2 + (∇σ1 )2 + 2∇σ0 · ∇σ1 + 2∇σ0 · ∇σ2 + U − E − ∇ σ0 + ∇2 σ1 + ∇2 σ2 = 0 . (10.13)
2m 2m
Collecting terms of the zeroth (classical) order in ~ we recover (10.8). Terms of the first and second order give

2∇σ0 · ∇σ1 − i~∇2 σ0 = 0 , (10.14)


2 2
(∇σ1 ) + 2∇σ0 · ∇σ2 − i~∇ σ2 = 0 . (10.15)

To simplify the derivations we will proceed with the one-dimensional case in which case Eqs. (10.8),(10.14),(10.15)
read

(σ00 )2 + 2m(U − E) = 0 , (10.16)


2σ00 σ10 − i~σ000 = 0 , (10.17)
(σ10 )2 + 2σ00 σ20 − i~σ200 = 0 . (10.18)

Solution to (10.16) is
Z x
σ0 = ± p(x0 )dx0 , (10.19)

where p(x) is given by (10.9). I will use a notation p(x) = ~k(x). Writing (10.17) as
i~ σ000 i~ p0 i~ d ln p
σ10 = 0 = = , (10.20)
2 σ0 2 p 2 dx
and integrating we find

σ1 = i~(ln p + ln C) , (10.21)

where C is a constant. Once σ0 and σ1 are known, we can solve (10.18) for σ2 etc. Now, collecting formulas
(10.19),(10.21),(10.7),(10.5) in (10.1) we derive the quasi-classical wave function (the overall time-dependent factor
e−iEt/~ is omitted):
C Rx
k(x0 )dx0 C0 Rx 0 0
ψ(x) = p ei a +p e−i a k(x )dx , (10.22)
p(x) p(x)

where C, C 0 and a are arbitrary constants.


In the classically allowed region E > U (x) momentum p is a real and the wave function (10.21) can be written as
Z x 
A 0 0
ψ(x) = p sin k(x )dx + α . (10.23)
p(x) a

118
§2 The wave function in the quasi-classical approximation X WKB METHOD

The values xi at which p(x) = 0, i.e. E = U (x) are called the turning points. At these points a classical particle
rebounds off the potential. The WKB approximation breaks down near a turning point because p becomes small and
(10.12) cannot be satisfied. Thus, although the wave function (10.21) formally diverges at the turning points, this is
merely an indication that the quasi-classical analysis is not valid there.
To understand at what values of x near a turning point x0 the WKB approximation breaks down, we will use
(10.12) in the form

p2  mλ|U 0 | . (10.24)

Expanding the potential near x0 as U (x) ≈ U (x0 ) + U 0 (x0 )(x − x0 ) and recalling that U (x0 ) = E we estimate

p2 = 2m(E − U ) ≈ 2m|U 0 (x0 )||x − x0 | . (10.25)

Plugging this into (10.24) yields

|x − x0 |  λ/2 . (10.26)

The WKB approximation breaks down at |x − x0 | ∼ λ or smaller because the de Broglie wavelength becomes compa-
rable with characteristic distance in the problem.
In the classically forbidden region E < U , the momentum p is imaginary. In its place we will use a real function κ
defined as
1p
κ(x) = −ik(x) = 2m(U − E) . (10.27)
~
The wave function in this region reads

C Rx 0 0 C0 R x 0 0
ψ(x) = p e− a κ(x )dx + p e a κ(x )dx . (10.28)
|p| |p|

§3. Bohr-Sommerfeld quantization rules

So far we have determined the approximate form of the wave function in the classically allowed (E > U ) and
forbidden (E < U ) regions. The wave function in the entire region −∞ < x < ∞ must be continuous. Therefore,
the different approximations are not independent, but must be matched near the turning points. We cannot simply
require that the quasi-classical expressions be equal each other at the turning points, since, as I explained above, the
WKB approximation fails there. To solve this problem we need to find an exact solution to the Schrödinger equation
in the vicinity of each turning point and match the WKB expressions with this solution.

FIG. 15:

Suppose there are only two turning points in the problem x1 and x2 as shown in Fig. 15. The WKB is not applicable

119
§3 Bohr-Sommerfeld quantization rules X WKB METHOD

in regions a1 < x < b1 and b2 < x < a2 . In other regions we have

C0 Rx 0 0
ψ1 (x) = p e− x κ(x )dx , x < a1 ;
1
(10.29)
|p|
Z x 
A 0 0
ψ2 (x) = √ sin k(x )dx + α , b1 < x < b2 ; (10.30)
p x1
C − R x κ(x0 )dx0
ψ3 (x) = p e x2 , x > a2 . (10.31)
|p|

To match the wave function ψ1 and ψ2 we need an exact solution in the interval a1 < x < b1 . To this end, expand

U (x) ≈ U (x1 ) + U 0 (x1 )(x − x1 ) = E − F (x − x1 ) , (10.32)

where F = −U 0 (x1 ) > 0 and E = U (x1 ), and substitute into Schrödinger equation

~2 d2 ψ
+ (E − U )ψ = 0 . (10.33)
2m dx2
The result is the Airy equation

~2 d2 ψ
+ F (x − x1 )ψ = 0 . (10.34)
2m dx2
It solution which vanishes at x → −∞ is
 1/3
√ 2mF
ψ(x) = B 0 πAi(ξ) , ξ= (x1 − x) . (10.35)
~2

Where B 0 is a constant and π is inserted for future convenience.(19 Using (10.25) we can express ξ as follows
 2/3
|x1 − x|
ξ= sign(x1 − x) . (10.36)
λ

Since the WKB applies when |ξ|  1, we have to match (10.35) with (10.29) and (10.30) in that region. In
particular, we do not need the exact expression (10.34), but only its asymptotic behavior at large |ξ|. These can
be found in handbooks on special functions, e.g. “Handbook of Mathematical Functions” by M. Abramowitz and
I.A. Stegun. We find there the following expressions
1 2 3/2
ψ(ξ) ≈ B 0 ξ −1/4 e− 3 ξ , ξ  1 , (10.37)
2  
2 3/2 π
ψ(ξ) ≈ B |ξ|−1/4 sin
0
|ξ| + , ξ  −1 . (10.38)
3 4

To express ξ in terms of k and κ, consider first x  x1


√ Z x1
2 3/2 2(x1 − x) 2 2mF
ξ = = (x1 − x)3/2 = κ(x0 )dx0 , (10.39)
3 3λ 3~ x
p p
where κ(x) = 2m(U − E)/~ = 2mF (x1 − x)/~. Similarly, at x  x1
Z x
2 3/2 2(x − x1 )
|ξ| = = k(x0 )dx0 , (10.40)
3 3λ x1

(19 Function ψ(x) in (10.35) is not normalized because it represents only a small part of the particle wave function.

120
§3 Bohr-Sommerfeld quantization rules X WKB METHOD

p p
where k(x) = 2m(E − U )/~ = 2mF (x − x1 )/~. Using these in (10.39) and (10.40) we derive the required
asymptotic expressions of the exact solution near the turning point x1 :
B Rx 0 0
ψ(x) ≈ p e− x κ(x )dx , x ≈ a1 ,
1
(10.41)
2 |p|
Z x 
B π
ψ(x) ≈ √ sin k(x0 )dx0 + , x ≈ b1 , (10.42)
p x1 4

where B is a new constant. Comparing with (10.29) and (10.30) we find that the continuity of the wave function
implies the following matching conditions
B π
C0 = , A=B, α= . (10.43)
2 4

The matching procedure near the right turning point x2 is similar. The asymptotics of the exact solution near x2
read
D − R x κ(x0 )dx0
ψ(x) ≈ p e x2 , x ≈ a2 , (10.44)
2 |p|
Z x2 
D π
ψ(x) ≈ √ sin k(x0 )dx0 + , x ≈ b2 . (10.45)
p x 4

We need to match these to (10.30) and (10.31). To this end, we rewrite (10.30) as
Z x  Z x2  Z x2 
A 0 0 π A 0 0 π 0 0 π
ψ2 (x) = √ sin k(x )dx + = √ sin k(x )dx + − k(x )dx + (10.46)
p x 4 p x1 2 x 4
 Z1 x2  Z x2  Z x2  Z x2 
A 0 0 π 0 0 π 0 0 π 0 0 π
=√ sin k(x )dx + cos k(x )dx + − cos k(x )dx + sin k(x )dx +
p x1 2 x 4 x1 2 x 4
(10.47)

Now we see that in order that ψ2 coincide with (10.45) and (10.44) with (10.31) we have to require
Z x2
D π
C= , k(x0 )dx0 + = π(n + 1) , D = A(−1)n . (10.48)
2 x1 2

where n = 0, 1, . . .. One often introduces the loop integral over the closed classical trajectory of a particle with energy
E
I Z x2 Z x1 Z x2
p(x)dx = pdx + (−p)dx = 2 pdx . (10.49)
x1 x2 x1

In view of (10.48) we derive the Bohr-Sommerfeld quantization rule


I
p(x)dx = 2π~(n + 1/2) , n = 0, 1, 2, . . . (10.50)

This is the quasi-classical quantization condition for a particle in the potential well of Fig. 15.

Some properties of the quasi-classical solution:


Rx
1. The wave function (10.23) oscillates. When x = x1 , its phase φ(x) = x1 k(x0 )dx0 + π4 equals φ(x1 ) = π/4.
When x = x2 it changes to φ(x2 ) = π/4 + π(n + 1/2) = π(n + 3/4). Therefore, ψ vanishes n times between
the turning points x1 and x2 , i.e. n is the number of nodes of the wave function. According to the oscillation
theorem it also enumerates the states in the potential well.
2. The applicability of the WKB approximation requires λ  x2 − x1 . It implies, in view of (10.50), that n  1.
In other words, highly excited quantum states are approximately quasi-classical.

121
§3 Bohr-Sommerfeld quantization rules X WKB METHOD

3. Since the wave function is exponentially suppressed in the classically forbidden region, we can neglect it alto-
gether when computing the wave function normalization. Thus, the normalization condition is
Z x2
|ψ(x)|2 dx ≈ 1 . (10.51)
x1

Since in the region x1 < x < x2 the wave function rapidly oscillates (viz. n  1), one can replace sin2 φ ≈ 1/2.
Using (10.23) we obtain

Z !1/2
x2
|A|2 dx 2
=1 ⇒ A= R x2 dx
. (10.52)
2 x1 p x1 p

Notice that
Z x2 Z x2 Z t(x2 )
dx dx 1 T
= = dt = . (10.53)
x1 p x1 mv m t(x1 ) 2m

Here T is the period of motion. In terms of ω = 2π/T we write (10.52) as


 1/2
2mω
A= . (10.54)
π

So, finally we have


 1/2 Z x 
2mω 0 π 0
ψ(x) = sin k(x )dx + . (10.55)
πp(x) a 4

4. In Statistical Mechanics on introduces the phase space that is span by the coordinates and momenta of all
particle in a system. The phase space volume element of a single particle is d3 p d3 x, or dp dx per degree of
freedom. Eq. (10.50) tells us is that the number of particle states per degree of freedom in a unit element of the
phase space is dp dx
2π~ . Actually, we used this fact in derivation of (8.31).

∗ Examples.

1. Quasiclassical quantization of one-dimensional harmonic oscillator U = mω 2 x2 /2 is accomplished by the Bohr-


Sommerfeld rule
Z x2
p(x)dx = π~(n + 1/2) , (10.56)
x1

where E = U (x1,2 ) determines the turning points


r
2E
x1,2 = ± ≡ ±a . (10.57)
mω 2
We have
Z x2 Z a p πa2 E
p(x)dx = mω a2 − x2 dx = mω = π = π~(n + 1/2) , (10.58)
x1 −a 2 ω

Thus,

En = ~ω(n + 1/2) . (10.59)

This is one of rare cases where the WKB approximation gives the exact result.

122
§3 Bohr-Sommerfeld quantization rules X WKB METHOD

FIG. 16:

2. Consider a potential U (x) such that U = ∞ for x < 0, see Fig. 16. Near the turning point x = b one has to do
the same matching as for the potential in Fig. 15. This yields
Z b !
C π
ψ(x) = p sin k(x0 )dx0 + , x < b. (10.60)
p(x) x 4

It is clear from the figure that the WKB approximation is valid all the way to the left turning point x = 0
(because p 6= 0 for x > 0). Thus, we can apply the boundary condition ψ(0) = 0 directly to the quasi-classical
function (10.60) with the result
Z b I
π
k(x)dx + = π(n + 1) , n = 0, 1, 2, . . . ⇒ p(x)dx = 2π~(n + 3/4) . (10.61)
0 4
This is the Bohr-Sommerfeld quantization rule for potential of Fig. 16.
3. As a particular case of Fig. 16 consider the linear potential U = F x when x > 0 and U = ∞ when x < 0.
Applying (10.61) we obtain
Z E/F p √
23/2 E 3/2 m
2m(E − F x)dx = = π~(n + 3/4) . (10.62)
0 3F
Thus, the quasiclassical approximation of the particle spectrum is

(3π~F 2 )1/3
En = (n + 3/4)2/3 . (10.63)
2m1/3
The first two levels: E0 = 1.842(~2 F 2 /m)1/3 and E1 = 3.240(~2 F 2 /m)1/3 . Actually, this problem can be solved
exactly in terms of the Airy functions. The result for the two lowest levels is E0exact = 1.856(~2 F 2 /m)1/3 and
E1exact = 3.245(~2 F 2 /m)1/3 . A very good agreement should not take us by surprise because the linear potential
is very smooth and thus the quasiclassical approximation is very accurate. Also notice that the accuracy of the
WKB improves as n increases, as expected.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §46-48, Merzbacher,
“Quantum Mechanics” 3rd edition, Ch. 7.

§4. Tunneling through potential barriers

In Chapter III we have already discussed the tunneling problem. There we considered a few cases where the
energy spectrum can be calculated exactly and therefore the transmission and reflection coefficients can be calculated
exactly as well. Usually, however, the exact solution to Schrödinger equation is not known and one has to rely upon
approximate techniques. In this section we will discuss how to compute the transmission and reflection coefficients in
the quasiclassical approximation.

123
§4 Tunneling through potential barriers X WKB METHOD

FIG. 17:

Consider potential barrier U (x) shown in Fig. 17 as a solid line. Suppose that the incident particle moves from
x = −∞ to the right with momentum ~k. To solve the scattering problem, one needs to write the quasiclassical wave
function such that in the region 1 it describes the incident and reflected waves, while in the region 3 the transmitted
wave. Since the WKB breaks down at the turning points one needs to use the matching procedure described in
the previous section to obtain the scattering coefficients. One is referred to the book by Merzbacher who solves
this problem. Here, for the purpose of illustration, we will simplify the problem by approximating the slopes of the
potential as shown in Fig. 17 by the dashed lines. In this case, the wave function in the regions 1 and 3 reads
ψ1 = Aeikx + Be−ikx , x < 0, (10.64)
ikx
ψ3 = Ce x > l. (10.65)
In the entire interval 0 < x < l the wave function is a smooth function of x, implying that (10.12) is satisfied,
(10.29)–(10.31) can be used and the matching conditions are merely the requirements that ψ and ψ 0 be continuous.
In order to determine the unknown coefficients we have to match (10.64),(10.65) onto the wave function in the region
0 < x < l.
In the case (b) we have
1 n R x κ(x0 )dx0 Rx 0 0
o
ψ2 = p αe 0 + βe− 0 κ(x )dx , 0 < x < l . (10.66)
κ(x)
We now need to require that ψ and its derivative ψ 0 be continuous at x = 0 and x = l. In general, the derivate of a
quasiclassical function ψ = p−1/2 eS has form
p0 S S0 S S0 S
ψ0 = − e + e ≈ e . (10.67)
2p3/2 p1/2 p1/2
Indeed, since dλ/dx  1 (see (10.11)), the first term is much smaller than the second one (recall that S 0 ∝ p). In
other words, when taking the derivative of a quasiclassical wave function one should treat the pre-exponential factor
as a constant. This observation simplifies our calculation a lot.
Applying the boundary conditions and introducing notations
Z
1p 1p 1 lp
a = κ(0) = 2m(U (0) − E) , b = κ(l) = 2m(U (l) − E) , γ = 2m(U (x) − E)dx . (10.68)
~ ~ ~ 0
we have
1
ψ1 (0) = ψ2 (0) ⇒ A + B = √ (α + β) , (10.69)
a

ψ10 (0) = ψ20 (0) ⇒ ik(A − B) = a(α − β) , (10.70)

ψ2 (l) = ψ3 (l) ⇒ αeγ + βe−γ = C beikl , (10.71)

ψ20 (l) = ψ30 (l) ⇒ b(αeγ − βe−γ ) = ikCeikl . (10.72)
Solving the last two equations for α and β we get
 
1 √ ik
α= b+ √ Ceikl−γ (10.73)
2 b
 
1 √ ik
β= b− √ Ceikl+γ (10.74)
2 b

124
§4 Tunneling through potential barriers X WKB METHOD

Notice that in the WKB approximation γ  1. Therefore α  β. Thus, neglecting α in (10.69) and (10.70) we derive
 √ 
1 β aβ
A= √ − . (10.75)
2 a ik

Finding C from (10.74) we obtain

C 4e−ikl−γ
= √  √ . (10.76)
A √1 − a b− ik

a ik b

Finally, the transmission coefficient reads


2
C 16e−2γ
T = = 1
 k2
. (10.77)
A a + ka2 b + b

It can be shown (see e.g. the book by Landau and Lifshitz) that the quasiclassical transmission coefficient for a
smooth barrier has a general form

− 2 x2
R
2m(U −E)dx
T ∝ e−2γ = e ~ x1 . (10.78)

The functional form of the pre-exponential factor depends on potential. However, in the WKB method the leading
term is the one in the exponent.

∗ Example 1. Suppose now that the barrier is such that at x > l it is U = U0 > 0. In this case the wave function
at x > l is
0 1p
ψ3 = Ceik x , k 0 = 2m(E − U0 ) . (10.79)
~
The reflection and transmission coefficients are
2 2
B k0 C
R = , T = ,
A (10.80)
A k
where e.g.
0
C 4e−ik l−γ
= √  √ . (10.81)
A √1 − a b− ik0

a ik b

The leading exponential behavior is still given by (10.78).

Returning to the potential on Fig. 17 let us consider the case (a). The function in 0 < x < l reads
Z x 
α 0 0
ψ2 = p sin k(x )dx + β . (10.82)
k(x) 0

The boundary conditions yield



a(A + B) = α sin β , (10.83)
√ √
ik(A − B) = α a = α a cos β , (10.84)

α sin(φ + β) = C beikl , (10.85)

bα cos(φ + β) = ikCeikl . (10.86)

I used the following notations:


Z lp
1
a = k(0) , b = k(l) , φ= 2m(E − U (x)) . (10.87)
~ 0

125
§4 Tunneling through potential barriers X WKB METHOD

The solution to these equations is


B ik(b − a) + (k 2 − ab) tan φ
= . (10.88)
A ik(b + a) + (k 2 + ab) tan φ
Thus, the reflection coefficient reads
2
B k 2 (b − a)2 + (k 2 − ab)2 tan2 φ
R = = 2 . (10.89)
A k (b + a)2 + (k 2 + ab)2 tan2 φ
Consider a particular case of U (0) = U (l). Setting a = b in (10.89) gives
(k 2 − a2 )2 tan2 φ
R= . (10.90)
4k 2 a2 + (k 2 + a2 )2 tan2 φ
We observe that R vanishes when
Z lp
1
φ= 2m(E − U (x)) = nπ , n = 1, 2, . . . (10.91)
~ 0

At such energies the barrier becomes completely transparent. Recall that in the classical scattering theory, R(a) = 0
and T(b) = 0.

∗ Example 2: Cold field electron emission from metals. The potential energy of electrons near the flat surface of

FIG. 18:

a metal sample is sketched in Fig. 18 (solid line). Electrons are approximately free inside the sample so that U = 0
at x < 0. Since electron energy E is smaller than the value of the potential outside the sample U0 , electrons cannot
escape the sample. In the electric field E the potential becomes U = U0 − eEx (dashed line). We see that now the
potential barrier is of finite width and therefore there is a finite probability that electron tunnels through and escapes
to x → ∞. We can compute the emission probability using the WKB. Noting that
U (x) − E = ϕ − eEx , (10.92)
where ϕ is called the work function (it is a known parameter), we estimate with “the exponential accuracy”
 Z  ( √ )
2 ap 4 2m
T ≈ exp − 2m(ϕ − eEx)dx = exp − . (10.93)
~ 0 3~eEϕ3/2
This formula indicates that the cold field emission is possible when the electric field exceeds the threshold value of

4 2m
Eth = . (10.94)
3~eϕ3/2

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §50,25, Merzbacher,
“Quantum Mechanics” 3rd edition, 7.4, Messiah, “Quantum mechanics”, VI.10.

126
§4 Tunneling through potential barriers X WKB METHOD

§5. Double-well potential

Consider one dimensional double-well potential U (x) shown in Fig. 19. It is an even function of coordinate U (x) =

FIG. 19:

U (−x). If the value of the potential at x = 0 is very large U0 → ∞, the two wells are independent. Let E0 be energy
of the corresponding ground state and ψ0 (x) its wave function, say, in the right well. It satisfies the Schrödinger
equation

~2 00
− ψ + U ψ0 = E0 ψ0 (10.95)
2m 0
and is normalized as usual
Z ∞
ψ02 dx = 1 . (10.96)
−∞

Note, that for infinitely high U0 , ψ0 vanishes at at x ≤ 0, so that particle can move only at x > 0.
At finite (but large) U0 , a particle can tunnel through the barrier from one well to another one. Therefore, it can
move in the entire region −∞ < x < ∞ (although its wave function exponentially decays outside the double-well).
The Hamiltonian is an even function of the coordinate and hence eigenstates are states of definite parity. Using ψ0
we can construct the following even and odd functions
1
ψ1 (x) = √ [ψ0 (x) + ψ0 (−x)] , (10.97)
2
1
ψ2 (x) = √ [ψ0 (x) − ψ0 (−x)] , (10.98)
2
were we neglected an exponentially small contribution due to the tunneling. However, this exponentially small
contribution leads to splitting of the corresponding energy levels E1 and E2 . We would like to compute the magnitude
of the splitting E2 − E1 using (10.97),(10.98) as the zeroth order approximation.
Functions ψ1,2 are normalized to unity. Indeed,

Z ∞ Z ∞ Z ∞ Z :0
1 ∞

2 1 1 
ψ1,2 (x)dx = ψ02 (x)dx + ψ02 (−x)dx ± ψ (x)ψ (−x)dx = 1 .
 (10.99)
0 0

−∞ 2 −∞ 2 −∞ 2 −∞

Actually, the integral that we neglected is not identically zero because U0 is finite. It is however exponentially
suppressed. ψ1 obeys Schrödinger equation with a certain energy E1 :

~2 00
− ψ + U ψ1 = E1 ψ1 . (10.100)
2m 1

127
§5 Double-well potential X WKB METHOD

Multiplying (10.100) by ψ0 and (10.95) by ψ1 and taking the difference we obtain

~2 00
− [ψ ψ − ψ100 ψ0 ] = (E0 − E1 )ψ0 ψ1 . (10.101)
2m 0
Integrating this over 0 < x < ∞ we get
Z ∞ Z ∞
~2 ~2 0 ∞ 1
− 00 00
[ψ0 ψ1 − ψ1 ψ0 ]dx = − 0
[ψ ψ1 − ψ1 ψ0 ] 0 = (E0 − E1 ) ψ0 ψ1 dx = (E0 − E1 ) √ . (10.102)
2m 0 2m 0 0 2

I used ψ1 ≈ ψ0 / 2 at x > 0. Notice,

ψ0,1 (∞) = 0 , ψ1 (0) = 2ψ0 (0) , ψ10 (0) = 0 . (10.103)

Plugging this into (10.102) we get

~2 0 1 ~2 0
ψ0 (0)ψ1 (0) = √ (E0 − E1 ) ⇒ E0 − E1 = ψ (0)ψ0 (0) . (10.104)
2m 2 m 0

By the same token we can derive

~2 0
E0 − E2 = − ψ (0)ψ0 (0) . (10.105)
m 0
Taking difference of (10.105) and (10.104) we find the total splitting between the energy levels in the double-well:

2~2 0
E2 − E1 = ψ (0)ψ0 (0) . (10.106)
m 0

Wave function under the barrier reads, see (10.28):


 Z a 
C0 0 0
ψ0 (x) = p exp − κ(x )dx . (10.107)
|p| x

Normalization:
r
A
0 1 2mω
C = = , (10.108)
2 2 π
Thus,
r  Z a 
ω 0 0
ψ0 (0) = exp − κ(x )dx , (10.109)
2πv0 0

where
r
|p(0)| 2(U0 − E0 )
v0 = = . (10.110)
m m
Now,
 Z a 
0 C0 0 0
ψ (0) = p κ(x) exp − κ(x )dx = κ(0)ψ0 (0) . (10.111)
|p| 0

We also observe that κ(0) = |p(0)|/~ = mv(0)/~. Finally,


 Z a 
~ω 0 0
E2 − E1 = exp − κ(x )dx . (10.112)
π −a

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §50.

128
§5 Double-well potential X WKB METHOD

§6. Quasiclassical approximation in central field

The WKB method can be generalized to motion in central field U (r). Schrödinger equation for the radial part χ(r)
of the wave function reads
 
~2 d 2 ~2 l(l + 1)
− + + U (r) χ = Eχ . (10.113)
2m dr2 2mr2
2
l(l+1)
It seems that one can apply the formulas of X §2 and X §3 with the replacement U (x) → U (r) + ~ 2mr 2 and with
the boundary condition χ(0) = 0 as in Fig. 16. However, closer inspection reveals a problem near r = 0. Consider a
potential that near the origin behaves as U ∼ 1/rn . The classical momentum p(r), given by
 
2 ~2 l(l + 1)
p (r) = 2m E − U − , (10.114)
2mr2
diverges at the origin. Nevertheless, the WKB method is valid if n > 2 since dλ/dr also vanishes at the origin.
However, if n < 2 the centrifugal term dominates at small r and the corresponding classical momentum becomes
r
~2 l(l + 1) ~p dλ d r 1
|p| ≈ 2m 2
≈ l(l + 1) ⇒ = p =p . (10.115)
2mr r dr dr l(l + 1) l(l + 1)
We see that for small l’s condition (10.11) is not fulfilled. Another manifestation of the same problem can be seen if
we substitute p(r) from (10.115) into the expression for the quasiclassical wave function:
 Z r1  n p o √
C √ 1
χ(r) = p exp ± κ(r)dr ∼ r exp ± l(l + 1) ln r = r 2 ± l(l+1) , as r → 0 . (10.116)
κ(r) r

Here r1 is a turning point. This differs from the exact solution χ ∼ rl+1 at r → 0 (see V §2).

FIG. 20:

One can fix this problem using the Langer transformation. The idea is to change the radial variable r and the
function χ in such a way that l(l + 1) transforms into (l + 1/2)2 in (10.116). This can be accomplished by the following
substitutions:
r = a0 ex , ψ(x) = e−x/2 χ(x) , −∞ < x < ∞ . (10.117)
Here x is a dimensionless variable (not a coordinate), a0 is a constant length scale (it will disappear from the final
result and introduced here only to have correct dimensions for r), ψ(x) is an auxiliary function. The point is that the
Schrödinger equation for ψ(x) can be treated quasiclassically! To see this, substitute (10.117) into (10.113) and after
simple calculations derive
 
00 1 2 2 2x 2ma20 x 2x
ψ + − − l(l + 1) + a0 k e − U (a0 e )e ψ = 0. (10.118)
4 ~2
Prime means a derivative with respect to x and I use a usual notation E = ~2 k 2 /2m. Notice now that l(l + 1) + 1/4 =
(l + 1/2)2 as desired. An analogue of the classical momentum that we can read off (10.118) is
(    2 )
2 2 2 2 2m x 2x 1
p̃ (x) = ~ a0 k − 2 U (ae ) e − l + (10.119)
~ 2

129
§6 Quasiclassical approximation in central field X WKB METHOD

so that (10.118) reads

~2 ψ 00 (x) + p̃2 (x)ψ(x) = 0 . (10.120)

Now, for λ̃ = ~/p̃ we have (remember that n < 2)

dλ̃ d 1
= q    ∼ ex(2−n) → 0 , x → −∞ . (10.121)
dx dx 2m 1 2
a20 k 2 − x
~2 U (ae ) e2x − l + 2

Thus, the problem at r → 0, which is the same as x → −∞ is solved. Applying the WKB method to equation (10.120)
yields
 Z 
A 1 a
ψ(x) = p exp − |p̃(x)|dx . (10.122)
|p̃| ~ x

Using dx = dr/r = e−x dr/a0 we write


r √  Z 
r A/ a0 1 r1
χ(r) = ψ(ln(r/a0 )) = p exp − |pL (r)|dr , (10.123)
a0 |pL (r)| ~ r

where r1 is the left turning point and


 1/2
2 2m (l + 1/2)2
pL (r) = ~ k − 2 U (r) − . (10.124)
~ r2

Evidently, χ(r) ∼ rl+1 as r → 0.


Finally, the Bohr-Sommerfeld quantization rule in this case reads
I
pL (r)dr = 2π~(nr + 1/2) . (10.125)

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §.

§7. Quasiclassical approximation for Coulomb potential

The Langer transformation of the previous section cannot be used in an importantpcase of thepCoulomb field
U = −α/r √(α > 0) for a particle in the s-state, i.e. l = 0. Indeed, at r → 0, p ∼ 2m|U | = 2mα/r, hence
dλ/dr = ~/ mαr → ∞. In this case we can still use the quasiclassical approximation, but not in a vicinity of the
origin.

FIG. 21:

Our strategy will be as follows: Firstly, we compute the exact solution near the origin. Secondly, we employ the
WKB method to determine the wave function far away from the origin (but to the left of the turning point a). Finally,

130
§7 Quasiclassical approximation for Coulomb potential X WKB METHOD

we match the two solutions. It is critical to determine the regions of applicability of various approximations that we
are going to make. (In practice, the problems of this type, when one has to “assemble” a solution from various pieces,
arise quite often).
Consider Schrödinger equation for the radial part of the s-wave:
 
~2 α
− − − E χ = 0 , χ(0) = 0 . (10.126)
2m r
At r → 0, it can be approximated as
β 2mα
χ00 + χ ≈ 0, β≡ , r → 0. (10.127)
r ~2
Solution to this equation is
√ p
χ(r) = A rJ1 (2 βr) . (10.128)

Approximation to this solution at distances satisfying βr  1 can be obtained from the well-known asymptotic of
the Bessel function:
 1/4  p
r π p
χ(r) ≈ A sin 2 βr − , βr  1 . (10.129)
π2 β 4
Now, at such distances
dλ ~ ~ p
=√ √ β ∼ 1, (10.130)
dr mαr mα
implying that the WKB approximation is applicable! However, we must be cautious and check when the assumptions
leading to (10.129) are consistent. We made two assumptions: (i) α/r  |E| to get (10.127) and (ii) r  1/β to get
(10.129), or, in other words,
1 α
r . (10.131)
β |E|
2
Such a region exists only if |E|  αβ ∼ mα
~2 , i.e. if the energy is low enough.
At r > 0 the WKB solution has a general form (10.23) (remember that E − U > 0 at r < a)
 Z r 
B 1
χWKB = p sin p(r0 )dr0 + γ . (10.132)
p(r) ~ 0

In the region r  α/|E|


r √
p 2mα ~ β
p(r) = 2m(E − U ) ≈ = √ , (10.133)
r r
and the wave function (10.132) becomes
 1/4  p 
r
χWKB = B sin 2 βr + γ . (10.134)
β~2
Eq. (10.134) and (10.129) must coincide implying that γ = −π/4. On the other hand, the WKB solution at r < a is
given by (10.45) as
 Z a 
C 1 0 0 π
χWKB = p sin p(r )dr + . (10.135)
p(r) ~ r 4

Here a is the turning point satisfying −|E| = −α/a. Therefore, sum of the phases of the quasiclassical functions
(10.134) and (10.135) must be a multiple of π (cp. (10.42),(10.45) and (10.50))
Z Z
1 r π 1 a π
p(r0 )dr0 − + p(r0 )dr0 + = π(nr + 1) , (10.136)
~ 0 4 ~ r 4

131
§7 Quasiclassical approximation for Coulomb potential X WKB METHOD

which means that


Z a
1
p(r0 )dr0 = π(nr + 1) . (10.137)
~ 0

Taking the integral on the left hand side:


Z α/E
r α √ 
2mπα
2m − |E| dr = p (10.138)
0 r 2 |E|

we arrive at the following quasiclassical spectrum:

mα2
Enr = − . (10.139)
2~2 (n
r + 1)
2

Accidentally, this formula coincides with the exact expression.

132
§1 Spin-statistics theorem XI IDENTICAL PARTICLES

XI. IDENTICAL PARTICLES

§1. Spin-statistics theorem

Consider a system on N particles described by the Hamiltonian Ĥ(q1 , q2 , . . . , qN ), where qa stands for the coordinates
and spin of particle a. Its wave function ψ(q1 , q2 , . . . , qN , t) satisfies the Schrödinger equation with this Hamiltonian.
If all particles are of the same kind, then the uncertainty principle requires that the particles be indistinguishable, or,
identical. Mathematically, it means that Ĥ will not change if we permute (swap) any pair of particles.
Introduce the permutation operator P̂ab of particle a and b. Then, the fact that the particles are identical can be
expressed as a conservation law

[P̂ab , Ĥ] = 0 . (11.1)

Let ψ(. . . , qa , . . . , qb , . . . , t) be an eigenfunction of P̂ab with an eigenvalue λ. Since operator P̂ab is Hermitian, λ is
real. Namely,

P̂ab ψ(. . . , qa , . . . , qb , . . . , t) = λψ(. . . , qa , . . . , qb , . . . , t) . (11.2)

Since operator P̂ab is Hermitian, λ is real. Applying P̂ab twice gives


2
P̂ab ψ(. . . , qa , . . . , qb , . . . , t) = λ2 ψ(. . . , qa , . . . , qb , . . . , t) . (11.3)

On the other hand,

P̂ab ψ(. . . , qa , . . . , qb , . . . , t) = ψ(. . . , qb , . . . , qa , . . . , t) . (11.4)

We conclude that λ2 = 1, hence λ = ±1. Eigenfunctions ψs corresponding to λ = 1 are symmetric, while eigenfunctions
ψa corresponding to λ = −1 are antisymmetric with respect to the permutation of particles a and b. Since a and
b are arbitrary, we conclude that the wave function of a system of identical particles must be either symmetric or
antisymmetric with respect of permutation of any two particles.
We learned before that any linear combination of solutions to Schrödinger equation is a possible physical state.
This is true only for a system of distinguishable particles. For a system of N identical particles, only those linear
combinations that posses definite symmetry (i.e. either symmetric or antisymmetric) qualify as a possible physical
state.
Particles described by symmetric wave functions are called bosons, while those described by antisymmetric wave
functions are called fermions. The spin-statistic theorem states that particles with integer spin are bosons, while
particles with half-integer spin are fermions. This theorem can proved in the relativistic theory. In fact, in the
Quantum Field Theory it naturally follows from the relativistic (Lorentz) invariance.

To construct a wave function of a system of identical particles, consider first the simplest case of two (identical)
particles. Let ψ(q1 , q2 ) be a solution to the Schrödinger equation. There are two combinations with the definite
symmetry properties:
1
ψa (q1 , q2 ) = √ [ψ(q1 , q2 ) − ψ(q2 , q1 )] , (11.5)
2!
1
ψs (q1 , q2 ) = √ [ψ(q1 , q2 ) + ψ(q2 , q1 )] , (11.6)
2!

In a system of N identical particles there are N ! possible permutations of coordinates qa . Denote by P̂ν an operator
that performs successive permutations of ν pairs qa ,qb . If ψ(q1 , . . . , qN ) is a solution to the Schrödinger equation, then
N!
X
ψs (q1 , . . . , qN ) = As P̂ν ψ(q1 , . . . , qN ) , (11.7)
ν=1
N!
X
ψa (q1 , . . . qN ) = Aa (−1)ν P̂ν ψ(q1 , . . . , qN ) , (11.8)
ν=1

133
§2 Second quantization of bosons XI IDENTICAL PARTICLES

In a system of non-interacting particles we can represent the Hamiltonian as


N
X
Ĥ = Ĥa , (11.9)
a=1

where Ĥa is a Hamiltonian of a’th particle. Let {ϕna (qa )} be a complete set of eigenfunctions of Ĥa , and Ena the
corresponding eigenvalues. Wave function ϕna (qa ) describes a one-particle state. The wave functions of N non-
interacting bosons or fermions read
N!
X
ψ s = As P̂ν ϕn1 (q1 )ϕn2 (q2 ) . . . ϕnN (qN ) , (11.10)
ν=1
N!
X
ψ a = Aa (−1)ν P̂ν ϕn1 (q1 )ϕn2 (q2 ) . . . ϕnN (qN ) . (11.11)
ν=1

Suppose there are N1 bosons with energy En1 , N2 bosons with energy En2 etc. Any permutation of N1 bosons with
energy En1 will not change the wave function (11.10). There are total N1 ! such permutations. Similarly, there are
N2 ! ways of permuting bosons with energy En2 which also will not change the way function, etc. Thus the normalized
boson wave function reads
r N!
N1 !N2 ! . . . X
ψs = P̂ν ϕn1 (q1 )ϕn2 (q2 ) . . . ϕnN (qN ) . (11.12)
N! ν=1

For fermions, (11.11) can be represented as the Slater determinant:



ϕ (q ) ϕ (q ) . . . ϕ (q )
n1 1 n1 2 n1 N
1 ϕ (q ) ϕn2 (q2 ) . . . ϕn2 (qN )
ψa = √ n2 1 (11.13)
N ! · · · ··· ··· ···

ϕn (q1 ) ϕn (q2 ) . . . ϕn (qN )
N N N

Note, that if two fermions have exactly same quantum numbers, then the determinant vanishes. In other words, a
system of identical non-interacting fermions cannot contain even two fermions in the same one-particle state. This
statement is known as the Pauli (Exclusion) principle. Note that Pauli principle for strongly interacting fermions
only requires their wave function to be anti-symmetric.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §61, Weinberg, “Lec-
tures on Quantum Mechanics”, 4.5.

§2. Second quantization of bosons

It is convenient to describe a system of identical particles using the occupation number representation, also known
as the second quantization. A state of a system of N identical non-interacting bosons can be represented by a ket
|N1 , N2 , . . .i = ψs , where the notations are the same as in the previous section. We would like to find the form of
various operators in this representation.
(1)
Consider a one-particle operator fˆa that acts only on qa . For example, one-particle momentum operator p̂a =
−i~∇a . One-particle operator of the entire system is
N
X
F̂ (1) = fˆa(1) . (11.14)
a=1

Its generic matrix element between two states in the occupation number representation

hN10 , N20 , . . . |F̂ (1) |N1 , N2 , . . .i (11.15)

134
§2 Second quantization of bosons XI IDENTICAL PARTICLES

is related to the matrix elements


Z
(1)
fik = ϕ∗ni (qa )fˆa(1) ϕnk (qa )dqa (11.16)

(1)
of the coordinate representation. Note that fik do not depend on a. There are two possibilities. (i) If functions
(1) (1)
ϕni (qa ) are eigenfunctions of fˆa , then fik is non-vanishing only if i = k. Thus, the only non-vanishing matrix
element of F̂ (1) is
X (1)
hN1 , N2 , . . . |F̂ (1) |N1 , N2 , . . .i = Ni fii . (11.17)
i

(1) (1)
(ii) If ϕni (qa ) are not eigenfunctions of fˆa , then the wave function fˆa ϕnk (qa ) is a linear combination of ϕni (qa )’s
with different ni ’s:
X
fa(1) ϕnk (qa ) = fik ϕni (qa ) . (11.18)
i

(1)
This means that fˆa removes a boson from an initial state k and transfers it to another state i. In this case the only
non-vanishing matrix elements are
p p
(1) Ni !(Ni − 1)! Nk !(Nk − 1)! (1) p (1)
h. . . Ni . . . Nk − 1 . . . |F̂ | . . . Ni − 1 . . . Nk . . .i = fik = Ni Nk fik . (11.19)
(Ni − 1)!(Nk − 1)!

See Example 3 at the end of the next section for an illustration of how the counting in (11.19) works.
Define now the annihilation and creation operators as in IV §6
p
âi | . . . , Ni , . . .i = Ni | . . . , Ni − 1, . . .i . (11.20)
p
â†i | . . . , Ni , . . .i = Ni + 1| . . . , Ni + 1, . . .i . (11.21)

Their non-vanishing matrix elements read


p
h. . . , Ni − 1, . . . |âi |, . . . , Ni , . . .i = Ni . (11.22)
p
h. . . , Ni , . . . |â†i | . . . , Ni − 1, . . .i = Ni + 1 . (11.23)

Using these definitions we obtain


p
h. . . , Ni , . . . |â†i âi | . . . , Ni , . . .i = Ni h. . . , Ni , . . . |â†i | . . . , Ni − 1, . . .i = Ni . (11.24)

Thus, the particle number operator

N̂i = â†i âi (11.25)

is diagonal in the second quantization. It is easy to verify that âi â†i = Ni + 1. Comparing with (11.25) we conclude
that

[âi , â†k ] = δik , [â†i , â†k ] = 0 , [âi , âk ] = 0 . (11.26)

Employing the operators âi and â†i we can represent the one-particle operator of a bosonic system as
X (1) †
F̂ (1) = fik âi âk . (11.27)
i,k

To prove (11.27) note that its matrix elements are the same as (11.17) and (11.19).
A two-particle operator have the general form
X (2)
F̂ (2) = fˆab , (11.28)
a>b

135
§3 Second quantization of fermions XI IDENTICAL PARTICLES

(2)
where fˆab acts on qa and qb . For instance, the potential energy operator U (|ra − rb |) is a two-particle operator. In
the second quantization, F̂ (2) can be represented as
1 X (2) † †
F̂ (2) = fik,lm âi âk âm âl , (11.29)
2
iklm

where
Z Z
(2) (2)
fik,lm = ϕ∗i (q1 )ϕ∗k (q2 )fˆ12 ϕl (q1 )ϕm (q2 )d2 q1 d2 q2 . (11.30)

Using the second quantization formalism we can write the Hamiltonian


X X X
Ĥ = Ĥa(1) (ra ) + Û (2) (ra , rb ) + Û (3) (ra , rb , rc ) + . . . , (11.31)
a a>b a>b>c

(1) p̂2a
where Ĥa = 2m + Û (1) (ra ), in the form
X (1) † 1 X (2) † †
Ĥ = Hik âi âk + Uik,lm âi âk âl âm + . . . (11.32)
2
ik iklm
(1)
If ϕn (q) are eigenfunctions of the Hamiltonian Ĥ (1) , then its matrix elements are Hik = Ei δik .

§3. Second quantization of fermions

Second quantization of identical non-interacting fermions can be done along the same steps as in the previous
section. We start with the state vector |N1 , N2 , . . .i = ψa . Now, however, the quantum numbers can only take two
values Ni = 0, 1 because otherwise the Pauli principle dictates that ψa = 0. To fix the overall sign of ψa introduce
the following convention for the Slater determinant: if (i) the particle coordinates q1 , q2 , . . . appear successively in
columns 1, 2, . . . and (ii) the quantum numbers n1 < n2 < . . . appear successively in rows 1, 2, . . ., then the overall
sign is +.
∗ To understand how the one-particle operator fˆ(1) acts on ψa , consider a couple of examples.
1. Let the wave function of a three-fermion system be
1
ψa = √ [ϕ1 (q1 )ϕ2 (q2 )ϕ3 (q3 ) + permutations with proper signs] = |11 12 13 04 05 . . .i . (11.33)
3!
Suppose, that fˆ(1) acts on q1 and moves a particle from the state #1 to the state #4. The resulting wave
function is
1
ψa0 = √ [ϕ4 (q1 )ϕ2 (q2 )ϕ3 (q3 ) + permutations with proper signs]
3!
1
= √ [+ϕ2 (q1 )ϕ3 (q2 )ϕ4 (q3 ) + permutations with proper signs] = |01 12 13 14 05 . . .i . (11.34)
3!
In the second line the first term we wrote is the one with the ordered coordinates and quantum numbers
n2 < n3 < n4 because it let us fix the overall sign of the ket. The overall sign can be found from the first line
in (11.34) by two permutations of the coordinate indexes 123 → 132 → 312 which yields overall plus sign.
2. Suppose now that the wave function is
1
ψa = √ [ϕ1 (q1 )ϕ2 (q2 )ϕ4 (q3 ) + permutations with proper signs] = |11 12 03 14 05 . . .i . (11.35)
3!
Suppose now, that fˆ(1) acts on q1 and moves a particle from the state #1 to the state #3. The resulting wave
function is
1
ψa0 = √ [ϕ3 (q1 )ϕ2 (q2 )ϕ4 (q3 ) + permutations with proper signs]
3!
1
= √ [−ϕ2 (q1 )ϕ3 (q2 )ϕ4 (q3 ) + permutations with proper signs] = −|01 12 13 14 05 . . .i . (11.36)
3!

136
§3 Second quantization of fermions XI IDENTICAL PARTICLES

In these examples we see that the overall sign of ψa0 depends on how many states are occupied between the
initial state and the final one. Denote the total number of occupied states between k and l as

X l
X
(k, l) = Ni . (11.37)
i=k

Then, assuming that i < k, we derive for the non-diagonal matrix elements
(1)
h. . . , 1i , . . . 0k , . . . |fˆa(1) | . . . , 0i , . . . 1k , . . .i = fik (−1)
P
(i+1,k−1)
, i < k. (11.38)
(1)
Once one of fˆa ’s moved a fermion from i to k others do not contribute due to the Pauli principle. Hence,
X (1) P
hN10 , N20 , . . . |F̂ (1) |N1 , N2 , . . .i = fik (−1) (i+1,k−1) , i < k . (11.39)
ik

For a diagonal matrix element


X (1)
h. . . , Ni , . . . |F̂ (1) | . . . , Ni , . . .i = fii Ni , Ni = 0, 1 . (11.40)
i

As for bosons, we introduce the creation and annihilation operators of fermions b̂i and b̂†i whose non-vanishing
matrix elements are

h. . . 0i . . . |b̂i | . . . 1i . . .i = h. . . 1i . . . |b̂†i | . . . 0i . . .i = (−1) (1,i−1) .


P
(11.41)
P
The physical meaning of (1, i − 1) is a number of occupied states below i. With these definitions we can write
X (1) †
F̂ (1) = fik b̂i b̂k . (11.42)
ik

Indeed, if i < k

h. . . 1i . . . 0k . . . |b̂†i b̂k | . . . 0i . . . 1k . . .i
= h. . . 1i . . . 0k . . . |b̂†i | . . . 0i . . . 0k . . .ih. . . 0i . . . 0k . . . |b̂k | . . . 0i . . . 1k . . .i
P P P P P
(1,i−1) (1,k−1) (1,i−1) (1,i−1)+Ni + (i+1,k−1)
= (−1) (−1) = (−1) (−1)
P P P
= (−1)2 (1,i−1)
(−1)0 (−1) (i+1,k−1)
= (−1) (i+1,k−1)
. (11.43)

In the case i = k the matrix b̂†i b̂k is diagonal, so that b̂†i b̂i = Ni , as required.
Eq. (11.41) can be used to express the creation and annihilation operators in the form
P
(1,i−1)
b̂i | . . . Ni . . .i = (−1) Ni | . . . 1 − Ni . . .i (11.44)
b̂†i | . . . Ni
P
(1,i−1)
. . .i = (−1) (1 − Ni )| . . . 1 − Ni . . .i (11.45)

Consider now

h. . . 1i . . . 0k . . . |b̂k b̂†i | . . . 0i . . . 1k . . .i
= h. . . 1i . . . 0k . . . |b̂k | . . . 1i . . . 1k . . .ih. . . 1i . . . 1k . . . |b̂†i | . . . 0i . . . 1k . . .i
P P P P P P
(1,k−1)+ (1,i−1) (1,i−1)+Ni + (i+1,k−1)+ (1,i−1) (i+1,k−1)
= (−1) = (−1) = −(−1) . (11.46)

Comparing (11.43) and (11.46) we conclude that

b̂†i b̂k + b̂k b̂†i = 0 , i 6= k . (11.47)

For i = k it is not difficult to see that b̂i b̂†i = 1 − Ni implying that

b̂i b̂†i + b̂†i b̂i = 1 . (11.48)

137
§3 Second quantization of fermions XI IDENTICAL PARTICLES

In short, b̂i and b̂†i satisfy the anti-commutation relations:


{b̂i , b̂†k } = δik , {b̂i , b̂k } = 0 , {b̂†i , b̂†k } = 0 . (11.49)
In a system of bosons the creation and annihilation operators of different particles are independent because there is
no restriction on the occupation numbers Ni . In contrast, in a system of fermions, once a state ni is occupied, no
other fermion can have the same quantum numbers.
In the second quantization formalism the Hamiltonian describing fermions has the following form
X (1) † 1 X (2) † †
Ĥ = Hik b̂i b̂k + Uik,lm b̂i b̂k b̂l b̂m + . . . (11.50)
2
ik iklm

∗ Examples.
1. The two-particle state of identical bosons is described by the ket
|1i 1k i = N â†i â†k |0i . (11.51)
To compute the normalization constant consider first i 6= k. We have
0
h1i 1k |1i 1k i = |N |2 h0|âk âi â†i â†k |0i = |N |2 h0|âk (1 + 
â†i â
> † 2
i )âk |0i = |N | h0|0i = 1 ⇒ N = 1. (11.52)


If i = k, then using the same algebra we conclude that


1
|2i i = √ (â†i )2 |0i . (11.53)
2
These functions in the coordinate representation read
1
|1i 1k i = √ {ϕni (qi )ϕnk (qk ) + ϕnk (qi )ϕni (qk )} , i 6= k , (11.54)
2
|2i i = ϕni (qi )ϕni (qk ) . (11.55)

2. The two-particle state of identical fermions is described by the ket


|1i 1k i = N b̂†i b̂†k |0i . (11.56)
For i 6= k
0
h1i 1k |1i 1k i = |N |2 h0|b̂k b̂i b̂†i b̂†k |0i = |N |2 h0|b̂k (1 − 
b̂†i b̂
>)b̂† |0i = |N |2 h0|0i = 1
i k ⇒ N = 1. (11.57)
If i = k, then it is easy to see that |2i i = 0, i.e. such a state does not exist in accordance with the Pauli principle.
In the coordinate representation (11.56) read
1
|1i 1k i = √ {ϕni (qi )ϕnk (qk ) − ϕnk (qi )ϕni (qk )} , i 6= k . (11.58)
2
(11.59)

3. Consider a system of N = 4 distributed such that there are N1 = 1 and N2 = 3 particles in the states #1 and
#2, described by the ket |11 32 i. Operator F̂ (1) can move one boson from the state #2 to state the #1 so that
the system is described by the ket |21 22 i. The corresponding matrix element (11.19) reads
Z r
2!2!
dq1 dq2 dq3 dq4 [ϕ1 (q1 )ϕ1 (q2 )ϕ2 (q3 )ϕ2 (q4 ) + ϕ1 (q1 )ϕ2 (q2 )ϕ1 (q3 )ϕ2 (q4 )
4!
+ ϕ1 (q1 )ϕ2 (q2 )ϕ2 (q3 )ϕ1 (q4 ) + ϕ2 (q1 )ϕ2 (q2 )ϕ1 (q3 )ϕ1 (q4 )
+ ϕ2 (q1 )ϕ1 (q2 )ϕ2 (q3 )ϕ1 (q4 ) + ϕ2 (q1 )ϕ1 (q2 )ϕ1 (q3 )ϕ2 (q4 )]∗
4 r
X 3!1!
× ˆ
fa(1)
[ϕ1 (q1 )ϕ2 (q2 )ϕ2 (q3 )ϕ2 (q4 ) + ϕ2 (q1 )ϕ1 (q2 )ϕ2 (q3 )ϕ2 (q4 )
a=1
4!
+ ϕ2 (q1 )ϕ2 (q2 )ϕ1 (q3 )ϕ2 (q4 ) + ϕ2 (q1 )ϕ2 (q2 )ϕ2 (q3 )ϕ1 (q4 )]
r r 4 Z r r r r
3!1! 2!2! X 3!1! 2!2! 3!1! 2!2! 4! √
= 3 ϕ1 (qa )fˆa ϕ2 (qa )dqa =
∗ (1)
4 · 3f12 = f12 = 3 · 2f12 .
4! 4! a=1 4! 4! 4! 4! 2!1!

138
§4 Supersymmetric harmonic oscillator XI IDENTICAL PARTICLES

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §64.

§4. Supersymmetric harmonic oscillator

A system containing bosons and fermions is described by a Hamiltonian that contains creation and annihilation
operators of both bosons âi , â†i and fermions b̂i , b̂†i . A Hamiltonian may posses a symmetry with respect to inter-
changing bosons and fermions âi ↔ b̂i . This is knows as the supersymmetry. As a simple example of such a system
consider one-dimensional supersymmetric oscillator whose Hamiltonian is given by

Ĥ = ĤB + ĤF = ~ω(↠â + b̂† b̂) (11.60)

Its eigenstates are |nB nF i and eigenvalues EN = ~ω(nB + nF ) where N = nB + nF . The (anti-)commutation relations
read

[â, ↠] = 1 , {b̂, b̂† } = 1 . (11.61)

Introduce the operators Q̂ and Q̂† such that


√ √
Q̂ = ~ω ↠b̂ , Q̂† = ~ω âb̂† . (11.62)

They create (annihilate) fermion and annihilate (create) boson. Due to the exclusion principle

Q̂2 = (Q̂† )2 = 0 . (11.63)

It is straightforward to check that the Hamiltonian can be written as

Ĥ = Q̂Q̂† + Q̂† Q̂ . (11.64)

Mathematically, the supersymmetry means that the operators Q̂ and Q̂† commutate with the Hamiltonian:

[Ĥ, Q̂] = [Ĥ, Q̂† ] = 0 , (11.65)

where (11.63) has been used. Note that [Q̂† , Q̂] 6= 0 implies degeneracy of the energy levels (see below).
Eqs. (11.63),(11.64),(11.65) can be more compactly written in terms of the operators

Q̂1 = Q̂ + Q̂† , Q̂2 = i(Q̂ − Q̂† ) (11.66)

as follows

{Q̂i , Q̂k } = 2δik Ĥ , [Q̂i , Ĥ] = 0 , i, k = 1, 2 . (11.67)

Some properties of the spectrum:

1. All energy levels are non-negative E ≥ 0. Indeed, in any state hQ̂Q̂† i ≥ 0 and hQ̂† Q̂i ≥ 0 implying that hĤi ≥ 0.

2. All energy levels, except E = 0, are two-fold degenerate. To see this, introduce a Hermitian operator Ŝ = Q̂† Q̂.
Because it commutes with the Hamiltonian we can choose the complete set of states ψES that are eigenstates
of Ĥ and Ŝ. Namely

ŜψES = SψES , ĤψES = EψES . (11.68)

Acting on both sides with Q̂† and recalling that (Q̂† )2 = 0, hence S Q̂† ψES = 0, we conclude that either
S = 0, or Q̂† ψES = 0. In the latter case, from ĤψES = EψES we find that S = E. Thus Ŝ has two
possible eigenvalues: S = 0 and S = E. When E 6= 0 both possibilities S = 0 and S = E are realized; the
corresponding eigenfunctions are transform into each other by Q̂ and Q̂† . This is the origin of the two-fold
degeneracy of levels with E 6= 0. To prove that Q̂† ψE0 = ψEE consider ŜψE0 = Q̂† Q̂ψE0 = 0. This is the
same as (Ĥ − Q̂Q̂† )ψE0 = (E − Q̂Q̂† )ψE0 = 0. Acting with Q̂† on the left we get E Q̂† ψE0 = Ŝ(Q̂† ψE0 ), which
indicates that Q̂† ψE0 is the eigenfunction of Ŝ with eigenvalue S = E.

139
§4 Supersymmetric harmonic oscillator XI IDENTICAL PARTICLES

3. If level E = 0 exists, it is non-degenerate. Indeed, from Ĥψ0 = 0 it follows that Q̂ψ0 = 0 and Q̂† ψ0 = 0. Thus
the supersymmetry operators do not generate any new state.

As an example of a supersymmetric system consider charged spin-1/2 particle in constant magnetic field. We
discussed this in detail in VIII §2. Energy spectrum is given by (8.19) and consists of the kinetic energy of motion in
the direction parallel to the field (continuous spectrum) and motion in a plane transverse to the field (discrete Landau
levels). The transverse part of the spectrum reads

Enσ = (n + 1/2)~ω0 − 2µσB . (11.69)

For electron µ = −e~/2mc yielding (see (8.22))



Enσ = (n + 1/2 + σ)~ω0 ≡ N ~ω0 , N = 0, 1, 2, . . . (11.70)

All energy levels, except the ground state N = 0, are two-fold degenerate because energy of the level with n = N ,
σ = −1/2 is the same as the one with n = N − 1 and σ = 1/2. The ground state with n = 0 and σ = −1/2 is clearly
not degenerate.
Let’s write the transverse Hamiltonian (remember that the electron charge is −e)

1  e 2 ~ω
0
Ĥ⊥ = p̂⊥ + A⊥ + σ̂z (11.71)
2m c 2

in the form of (11.60) and identify the supersymmetry operators Q̂ and Q̂† . To this end introduce an operator
describing the spin degree of freedom
!
1 0 0
b̂ = (σ̂x − iσ̂y ) = . (11.72)
2 1 0

It is easy to verify that b̂† b̂ = 12 (1 + σ̂z ) and {b̂, b̂† } = 1. For the orbital degree of freedom we introduce an operator

1
â = √ (π̂y + iπ̂x ) , (11.73)
2m~ω0

where π̂ = mv̂ = (p̂ + eA/c). One can check that [â, ↠] = 1. With these definitions we cast the Hamiltonian (11.71)
in the form

Ĥ⊥ = ~ω0 (↠â + 1/2) + ~ω0 (b̂† b̂ − 1/2) . (11.74)

We observe that spin is a fermionic degree of freedom with nF = σ + 1/2 (where σ = ±1/2), whereas the orbital
motion is a bosonic degree of freedom with nB = n. Spectrum of (11.74) is given by

En⊥B nF = ~ω0 (nB + nF ) . (11.75)

The supersymmetry operators can be introduced as in (11.62) and read


!
† π̂y + iπ̂x 0 1
Q̂ = √ , Q̂ = Q̂† . (11.76)
2m 0 0

1
Q̂1 = Q̂† + Q̂ = √ (σ̂x π̂y − σ̂y π̂x ) , (11.77)
2m
1
Q̂2 = −i(Q̂† − Q̂) = √ (σ̂x π̂x − σ̂y π̂y ) , (11.78)
2m
In terms of these operators the Hamiltonian reads

Ĥ⊥ = Q̂21 = Q̂22 = {Q̂, Q̂† } . (11.79)

140
§5 Field operators ψ̂ XI IDENTICAL PARTICLES

§5. Field operators ψ̂

A more compact form of the second quantization formalism is achieved if we introduce the field operators
X X
ψ̂(q) = ϕni (q)âi , ψ̂ † (q) = ϕ∗ni (q)â†i (11.80)
i i

for a system of bosons and


X X
ψ̂(q) = ϕni (q)b̂i , ψ̂ † (q) = ϕ∗ni (q)b̂†i (11.81)
i i

for a system of fermions. The ψ̂-operators act on states in the occupation number representation, while q is considered
a parameter. In the rest of this section we will discuss only bosons. Equations for fermions can be obtained by replacing
âi by b̂i and the commutation relations by anti-commutation ones.
The field operators ψ̂ satisfy the following commutation relations
X
[ψ̂(q), ψ̂(q 0 )] = ϕni (q)ϕnk (q 0 )[âi , âk ] = 0 . (11.82)
ik
X X
[ψ̂(q), ψ̂ † (q 0 )] = ϕni (q)ϕ∗nk (q 0 )[âi , â†k ] = ϕni (q)ϕ∗ni (q 0 ) = δ(q − q 0 ) . (11.83)
ik i

This is the completeness relation. We can represent a one-particle operator F̂ (1) in the form
Z
F̂ (1) = ψ̂ † (q)fˆ(1) ψ̂(q)dq . (11.84)

Indeed,
Z X X X (1)
F̂ (1)
= ϕ∗ni (q)â†i fˆ(1) ϕk (q)âk dq = â†i âk fik . (11.85)
i k ik

Similarly,
ZZ
1
F̂ (2) = ψ̂ † (q)ψ̂ † (q 0 )fˆ(2) ψ̂(q 0 )ψ̂(q)dqdq 0 . (11.86)
2

∗ Examples.
1. Hamiltonian can be cast in the form
Z   Z Z
~2 † 1
Ĥ = − ψ̂ (q)∇2 ψ̂(q) + ψ̂ † (q)U (1) (q)ψ̂(q) dq + ψ † (q)ψ̂ † (q 0 )Û (2) (q, q 0 )ψ̂(q 0 )ψ̂(q)dqdq 0 . (11.87)
2m 2

2. Operator ψ̂ † (q)ψ̂(q) is the particle density operator. Indeed, if we integrate it over all coordinates q we get the
total number of particles:
Z Z X X X †
ψ̂ † (q)ψ̂(q)dq = ϕ∗ni (q)â†i ϕnk (q)âk dq = âi âi = N̂i , (11.88)
i k i

due to the orthogonality condition.


3. Operator of momentum of a system of identical bosons in the second quantization reads (see (11.84))
Z

P̂ = −i~ ψ̂ † (q 0 ) 0 ψ̂(q 0 )dq 0 . (11.89)
∂r
Using (11.82) and (11.83) we can compute the following commutator
Z Z
∂ ∂
[P̂ , ψ̂(q)] = P̂ ψ̂(q) − ψ̂(q)P̂ = −i~ ψ̂ † (q 0 ) 0 ψ̂(q 0 )dq 0 ψ̂(q) + i~ψ̂(q) ψ̂ † (q 0 ) 0 ψ̂(q 0 )dq 0
∂r ∂r
Z h i ∂ Z
† 0 † 0 ∂ ∂
= −i~ ψ̂ (q )ψ̂(q) − ψ̂(q)ψ̂ (q ) ψ̂(q )dq = i~ δ(q − q 0 ) 0 ψ̂(q 0 )dq 0 = i~ ψ̂(q) .
0 0
(11.90)
∂r 0 ∂r ∂r
In a system of identical fermions the same result hold for the anti-commutators.

141
§5 Field operators ψ̂ XI IDENTICAL PARTICLES

4. Consider a system of N identical spinless particles in volume V , known as the Bose-gas, interacting via a short-
range repulsive potential U (r) ≥ 0. If we neglect particle interactions, then the ground state energy is zero
because all bosons can be in their respective one-particle ground states. Interactions, however, give a finite
contribution to the ground state energy that can be computed in the perturbation theory as

E0 = hN0 |Û |N0 i , (11.91)

where |N0 i denotes the ground state of the entire Bose-gas with all N particles having zero momentum.
To compute the matrix element (11.91) we use (11.86) to write
ZZ
1
Û = ψ̂ † (r)ψ̂ † (r 0 )U (|r − r 0 |)ψ̂(r 0 )ψ̂(r)d3 rd3 r0 . (11.92)
2

Expanding the ψ̂-operator in the plane waves (states with definite momentum)
1 X ik·r
ψ̂(r) = √ e âk . (11.93)
V k

and plugging into (11.92) we arrive at


1 X
Û = â†k4 â†k3 âk2 âk1 Uk3 k4 ,k1 k2 , (11.94)
2
k1 k2 k3 k4

as in (11.29). Here
ZZ
1 0
Uk3 k4 ,k1 k2 = 2 U (|r − r 0 |)ei(k1 −k3 )·r ei(k2 −k4 )·r d3 rd3 r0 . (11.95)
V
We have

hN0 |â†k4 â†k3 âk2 âk1 |N0 i = N (N − 1) ≈ N 2 , if k1 = k2 = k3 = k4 = 0 , (11.96)

and vanishes otherwise. I used (11.23). The corresponding matrix element is


Z
1
U00,00 = U (r)d3 r ≡ hU i . (11.97)
V
Therefore, the ground state energy is
1
E0 = hU i N 2 . (11.98)
2

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §64,65; E. Merzbacher,
“Quantum Mechanics” 3rd edition, Ch. 21.

142
§1 Ground state of helium-like atoms XII ATOMS

XII. ATOMS

§1. Ground state of helium-like atoms

Atoms is one of the most important examples of a system of identical particles. It consists of a nucleus with the
electric charge eZ and Z electrons each with the electric charge −e and mass m. The nucleus and the electrons
interact through the attractive Coulomb potential. Since the nucleus is much heavier than the electrons, it’s motion
can be neglected. It is therefore convenient to study motion of the electrons in a reference frame with the origin at
the nucleus. In this section we consider the helium-like atoms, which is the simplest multi-particle system containing
only two electrons. Besides the He atom itself it includes the ionized lithium atom Li+ , double-ionized beryllium atom
Be++ and the like. The Hamiltonian of this system is
p̂21 Ze2 p̂2 Ze2 e2
Ĥ = − + 2 − + , (12.1)
2m r1 2m r2 |r1 − r2 |
If, as the zero order approximation, we neglect mutual electron interaction described by the last term, then the two
electrons are independent. The corresponding spectrum is given by
 
Z 2 e2 1 1
Ψ(0) (r ,
σ, σ2 1 2 r ) = ψ(r ,
1 2r )χ σ1 σ2 = ϕ (r
n1 l 1 m 1 1 )ϕ (r
n2 l2 m2 2 )χσ1 σ2 , E (0)
= − + . (12.2)
2a n21 n22
where the stationary states of a single electron are described by the wave functions

ϕnlm (r) = Rnl (r)Ylm (θ, φ) (12.3)

and χσ1 σ2 is the spin wave function (i.e. spinor) with σ =↑, ↓ being the spin projection on z-axis. Let me emphasize
that such separation of coordinate and spin dependence is possible because the Hamiltonian does not depend on
spin.(20
The ground state of a helium-like atom corresponds to both electrons being in 1s states, i.e. n = 1, l = m = 0. One
says that the electronic configuration is (1s)2 . Its coordinate wave function and energy read

Z 3 − Z (r1 +r2 ) (0) Z 2 e2


ψ0 = ϕ1s (r1 )ϕ1s (r2 ) = e a , E0 =− . (12.4)
πa3 a
The coordinate wave function ψ0 is symmetric under the permutation of the coordinates r1 ↔ r2 . On the other hand,
Ψ must be antisymmetric under the permutation of both the coordinates r1 ↔ r2 and the spins σ1 ↔ σ2 . Therefore,
the spin wave function must be anti-symmetric under the spin permutation viz.
1
χσ1 σ2 = √ (η1↑ η2↓ − η2↑ η1↓ ) ≡ χa , (12.5)
2
where η1σ1 and η2σ2 are one-particle spin states.
According to the perturbation theory, the first correction to the ground state energy is given by
ZZ
e2 e2
C = hψ0 | |ψ0 i = ϕ21s (r1 ) ϕ2 (r2 )d3 r1 d3 r2 . (12.6)
|r1 − r2 | |r1 − r2 | 1s
The integrals are easily taken using the well-known formula (see (D12))

1 X 1 l
r< ∗
= 4π l+1
Ylm (θ1 , φ1 )Ylm (θ2 , φ2 ) . (12.7)
|r1 − r2 | 2l + 1 r>
lm

We get
Z ∞ Z r1 Z ∞ 
Z6 1 − Z (r1 +r2 ) 1 − Z (r1 +r2 ) 5Ze2
C = 4πe2 r12 dr1 dr2 r22 e a + dr2 r22 e a = . (12.8)
πa6 0 0 r1 r1 r2 8a

(20 We will see later that in fact there are small relativistic corrections that couple spins and coordinates.

143
§1 Ground state of helium-like atoms XII ATOMS

Thus, the ground state energy with the account of the electron-electron coupling (i.e. interaction) is
 
(0) Z 2 e2 5Ze2 Ze2 5
E0 = E0 + C = − + =− Z− . (12.9)
a 8a a 8
C is called the Coulomb integral. Its measures the average interaction energy.

E0(He+)

E0(He)
(0)
E0 (He)

FIG. 22:

(0)
In Fig. 22 we show the ground state energy of the helium-like atom in the zeroth order approximation E0 (He) and
in the first order approximation E0 (He) given by (12.9). To ionize an atom one needs to remove one of its electrons.
In our case, removal of an electron produces the He+ ion that has one electron and whose spectrum is the same as
that of the hydrogen-like atom. Denote the ground state of the ionized helium-like atom by E0 (He+ ). The energy
cost of ionizing the helium-like atom is called the ionization energy I and can be computed as
   
+ Z 2 e2 Ze2 5 Ze2 5
I = E0 (He ) − E0 (He) = − + Z− = Z− . (12.10)
2a a 8 2a 4

∗ Example. Calculate the ionization energy of the helium-like atom using the variational method.
Let’s start with the following trial function
1 β 3 − β(r1 +r2 )
Φ0 = e a , (12.11)
π a3
where β is a parameter that has physical meaning of an effective ion charge. According to the variational method we
first compute the intergal
Z    
e2 2 5
J(β) = Φ0 ĤΦ0 d3 r1 d3 r2 = β − 2Z − β , (12.12)
a 8

where Ĥ is given by (12.1). Next we require that ∂J/∂β = 0 at some β = β0 . This gives the effective ion charge
5
β0 = Z − . (12.13)
16
The ground state energy is
 
5 25 e2
E0 = J(β0 ) = − Z 2 − Z + (12.14)
8 256 a
and the corresponding wave function
1 β03 − β0 (r1 +r2 )
Φ0 = e a , (12.15)
π a3
We see that compared to (12.9), (12.14) takes into account the mutual electron screening (note that an additional
term in (12.14) is independent of Z). Finally, the ionization energy is
 
Z 2 e2 Ze2 5 25
I=− − E0 = Z− + . (12.16)
2a 2a 4 128

144
§2 Lowest excited states of helium-like atoms XII ATOMS

§2. Lowest excited states of helium-like atoms

In the ground state both electrons are in 1s states, i.e. the electronic configuration is (1s)2 . The first excited
state has the configuration (1s)1 (2s)1 (21 . The corresponding coordinate wave functions, at the zeroth order of the
perturbation theory, are
1
ψs = √ [ϕ1s (r1 )ϕ2s (r2 ) + ϕ1s (r2 )ϕ2s (r1 )] , (12.17)
2
1
ψa = √ [ϕ1s (r1 )ϕ2s (r2 ) − ϕ1s (r2 )ϕ2s (r1 )] , (12.18)
2
where the one particle wave functions read
 3/2
1 Z
ϕ1s =√ e−Zr/a , (12.19)
π a
 3/2  
1 Z Zr
ϕ2s = √ 2− e−Zr/2a . (12.20)
4 2π a a
Possible spin wave functions are: one antisymmetric (12.5) (called a singlet) and three symmetric ones (called a
triplet):

 1
 √2 (η1↑ η2↓ + η2↑ η1↓ )
χs = η1↑ η2↑ (12.21)


η1↓ η2↓
Since the total wave function must be anti-symmetric, the possible atomic states are
ΨP = ψs χa , para-state , (12.22)
ΨO = ψa χs , ortho-state . (12.23)

At the first order of the perturbation theory we need to take into account interaction of electrons. Due to this
interaction, energy of the para-state is higher than that of the ortho-state. Indeed, if r1 = r2 , ΨO = 0, while ΨP
is maximal, implying that in the ortho-state electrons tend to be further away from each other compared to the
para-state. Thus, the average Coulomb repulsion energy (which is positive!) is smaller in the ortho-state than in the
para-state: EP > EO . More precisely, energies of the para- and ortho-states are
ZZ  
3 3 Z 2 e2 1
EP = ψs Ĥψs d r1 d r2 = − 1+ +C +X, (12.24)
2a 4
ZZ  
Z 2 e2 1
EO = ψa Ĥψa d3 r1 d3 r2 = − 1+ +C −X, (12.25)
2a 4
where
ZZ
e2 17 e2 Z
C= ϕ21s (r1 )ϕ22s (r2 ) d3 r1 d3 r2 = , (12.26)
|r1 − r2 | 81 a2
ZZ
e2 16 e2 Z
X= ϕ1s (r1 )ϕ2s (r2 ) ϕ1s (r2 )ϕ2s (r1 )d3 r1 d3 r2 = . (12.27)
|r1 − r2 | 729 a2
X is called the exchange integral. The exchange integral is a measure of the electron-electron correlation that occurs
because of the symmetry properties of the fermionic system. The total energy of the first excited states in He is
e2 e2
EP = −2.27 , EO = −2.31 . (12.28)
a a
The experimental data confirms that EP > EO :
e2 e2
EPexp = −2.146 , exp
EO = −2.175 . (12.29)
a a

(21 The p-states have higher energy due to the relativistic corrections, see (14.76).

145
§3 Hartree-Fock method XII ATOMS

§3. Hartree-Fock method

In systems with larger number of particles than hydrogen and helium atoms analytical calculations becomes more
complicated. These systems can be handled using numerical methods (provided that the number of degrees of freedom
is not too large). In this section I will outline the basic idea of a method known as the Hartree-Fock method.
We start with the Hamiltonian
X 1 X0
Ĥ = Ĥa + V̂ab , a, b = 1, 2, . . . , Z . (12.30)
a
2
ab

Here Ĥa = p̂2a /2m − e2 Z/ra is one-particle Hamiltonian of a’th electron in the Coulomb field of an ion of charge eZ,
V̂ab = e2 /|ra − rb | is electron-electron interaction energy. To calculate the ground state we will employ the variational
method according to which the wave function is determined from the minimizing functional J:
Z
δJ = δ ψ ∗ ĤψdΓ = 0 (12.31)

under the condition


Z
ψ ∗ ψdΓ = 1 . (12.32)
Q
I am using a shorthand notation for the integration volume dΓ = a d3 ra .
In this method it is important to start with a good trial function. Inspired by the perturbation theory calculations
of the previous sections, we choose it to be

ψ(r1 , r2 , . . . , rZ ) = ϕ1 (r1 )ϕ1 (r2 ) . . . ϕ1 (rZ ) . (12.33)

This ansatz assumes that electrons are independent. To take into account the symmetry properties of electrons the
wave function (12.33) must be converted into the Slater determinant as described in XI §1. However, for the sake of
simplicity, I will neglect the symmetry properties in this section. Substituting (12.33) into J we obtain
Z Z Y !
X 1 X0 Y
∗ ∗
J = ψ ĤψdΓ = ϕa (ra ) Ĥb + V̂cb ϕ(rd )dΓ
a
2
b bc d
XZ 1 X0
ZZ

= 3
ϕb Ĥb ϕb d rb + ϕ∗b ϕ∗c V̂cb ϕb ϕc d3 rb d3 rc . (12.34)
2
b bc

Its variation
Z ( )
X X0 Z
δJ = 2 δϕ∗b Ĥb + ϕ∗c V̂cb ϕc d3 rc ϕb d3 rb = 0 . (12.35)
b bc

Variation of (12.32) produces


Z
δϕ∗b ϕb d3 rb = 0 . (12.36)

One can combine (12.35) and (12.36) using the Lagrange multipliers method. It consists of introducing parameters
known as the Lagrange multipliers, which I will denote by Eb and considering a new functional
X Z 
J0 = J − Ea ϕ∗a ϕa d3 ra − 1 . (12.37)
a

The original problem (12.31),(12.32) is reduced to minimizing J 0 with respect to ϕa ’s and Ea ’s. In particular, in place
of (12.35) we write
( )
XZ X0 Z
0 ∗ ∗
δJ = 2 δϕb Ĥb + ϕc V̂cb ϕc d rc − Eb ϕb d3 rb = 0 ,
3
(12.38)
b c

146
§3 Hartree-Fock method XII ATOMS

which implies
" #
X0 Z
Ĥb + ϕ∗c V̂cb ϕc d3 rc − Eb ϕb = 0 , b = 1, 2, . . . , Z . (12.39)
c

In this set of equations the second term in the square brackets


X0 Z
Ub (rb ) = ϕ∗c V̂cb ϕc d3 rc (12.40)
c

can be interpreted as the effective potential energy called the self-consistent field. The Lagrange parameters Eb play
the role of the particle energies.
To solve the system of Z integro-differential equations (12.40) we will employ the method of consecutive approxi-
mations (a.k.a. iterative method). To begin, let us set the one-particle wave functions to be the hydrogen-like wave
(0)
functions ϕb . The self-consistent field then reads

(0)
X0 Z
Ub (rb ) = ϕ(0)∗
c V̂cb ϕ(0) 3
c d rc . (12.41)
c

Substituting into (12.39) yields the following system of equations


h i
(0) (1)
Ĥb + Ub − Eb ϕb = 0 , b = 1, 2, . . . , Z (12.42)

(1)
for functions ϕb . Its much easier to solve (12.42) than (12.39) because it is a set of ordinary differential equations.
Once it is solved, usually numerically, we calculate the next iteration of the self-consistent field:

(1)
X0 Z
Ub (rb ) = ϕ(1)∗
c V̂cb ϕ(1) 3
c d rc (12.43)
c

and plug it in (12.39) to obtain a set of equations


h i
(1) (2)
Ĥb + Ub − Eb ϕb = 0 , b = 1, 2, . . . , Z (12.44)

(2) (2)
for functions ϕb . Solving this set we can compute the self-consistent field Ub etc. This yields a series of functions
(0) (1) (2)
ϕb , ϕb , ϕb , . . . (12.45)
(n) (n−1)
The procedure stops when ϕb ≈ ϕb with the required accuracy. The last iteration is the approximate solution
for the one-particle wave function ϕb .
From (12.38) it follows that the one-particle energies are
Z X0 Z Z
Eb = ϕ∗b Ĥb ϕb d3 rb + ϕ∗b ϕ∗c V̂bc ϕb ϕc d3 rb d3 rc . (12.46)
c

The total energy of all one-particle states is


Z
X ZZ
1 X0
E= Eb − ϕ∗b ϕ∗c V̂bc ϕb ϕc d3 rb d3 rc , (12.47)
2
b=1 bc

where the last term accounts for the symmetry b ↔ c, see (12.34).

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §69, Arfken at. al.,
“Mathematical methods for physicists”, 7th edition, Ch. 22.

147
§4 Thomas-Fermi model XII ATOMS

§4. Thomas-Fermi model

The Hartree-Fock method works well for not too large Z. In this section we discuss the statistical model of Thomas
and Fermi that explains many general properties of heavy atoms without actually solving the Schrödinger equation.
The idea of the Thomas-Fermi method is to use the quasiclassical approximation because most of electrons are in
states with large principle quantum number. Indeed, electrons are fermions and cannot be in the same quantum state.
In the ground state of an atom electrons occupy all energy levels starting from the lowest one. The total number of
electrons N in the ground state equals approximately the volume of the phase space, see X §3 for the definition of the
phase space. Thus,
∆px ∆py ∆py V 4π 3 V
N =2 =2· p . (12.48)
(2π~) 3 3 F (2π~)3
The factor 2 in front accounts for two possible spin states of each electron. pF is the maximal momentum, called the
Fermi momentum. For a given particle number density n = N/V (12.48) implies

p3F
n= . (12.49)
3π 2 ~3
Keep in mind that n and pF are functions of r. We can write (12.49) as a statement about the maximal kinetic energy
of electron EF known as the Fermi energy:

p2F (3π 2 n)2/3 ~2


EF = = . (12.50)
2m 2m

Let now φ(r) be the electrostatic potential of atom satisfying φ(r) → 0 at r → ∞. The total energy of electron
with momentum pF is

p2F (r)
− eφ(r) ≡ −eφ0 . (12.51)
2m
(remember that e > 0, so that electron charge is −e). Note that φ0 > 0 because electron is bound in atom. Moreover,
φ0 is a constant for otherwise electrons would move to positions with larger φ0 (which would contradict the fact that
it is a stationary state). Employing (12.49) we derive
1
n(r) = [2me(φ(r) − φ0 )]3/2 . (12.52)
3π 2 ~3
Since p2F ≥ 0 it follows that φ ≥ φ0 . At φ = φ0 , the electron density n vanishes indicating the position of the atom
boundary.
Consider electrically neutral atoms (as opposed to ions). According to the Gauss’ law electric field of a spherically
symmetric atom of radius R vanishes when r > R. Together with the boundary condition mentioned before (12.51),
this implies that φ0 = 0 (for ions φ0 6= 0). Thus (12.52) becomes
1
n(r) = [2meφ(r)]3/2 . (12.53)
3π 2 ~3
The electrostatic potential satisfies the Poisson law

∇2 φ = 4πne . (12.54)

Substituting (12.53) we arrive at the main equation



8 2 3/2 5/2 3/2
∇2 φ = m e φ , (12.55)
3π~3
called the Thomas-Fermi equation. This equation requires two boundary conditions. These are the follows
eZ
φ→ , r → 0, (12.56)
r
rφ → 0 , r → ∞ . (12.57)

148
§4 Thomas-Fermi model XII ATOMS

Eq. (12.56) states that at small r the electrostatic potential of atom must coincide with the electrostatic potential of
its nucleus, while (12.57) states that the total charge of atom is zero.
It is convenient to introduce a dimensionless variable x and function χ as follows
 2/3
a 3π
r = xbZ −1/3 , with b = , (12.58)
2 4
 1/3 
Ze rZ
φ(r) = χ , (12.59)
r b

where a = ~2 /me2 is the Bohr radius. The advantage of the new notation is that the Thomas-Fermi equation (12.55)
becomes completely dimensionless:
d2 χ
x1/2 = χ3/2 , (12.60)
dx2
with boundary conditions

lim χ = 1 , lim χ = 0 . (12.61)


x→0 x→∞

Therefore, χ(x) is a universal function independent of nuclear parameters. Numerical solution to the Thomas-Fermi
equation is shown in Fig. 23. Once (12.60) is solved numerically we can obtain φ and n for a particular nucleus

FIG. 23:

through (12.58),(12.59) and (12.53). It can be shown that χ(x) rapidly decreases at x & 1 so that x = 1 indicates
that atomic boundary. Hence, the atom radius in this model is R = bZ −1/3 ≈ aZ −1/3 .
The total kinetic energy of electrons T can be computed as follows:
Z pF (r) 2 Z
p p2 dp 3
T = 2 3
d r ∝ [n(r)]5/3 d3 r . (12.62)
0 2m (2π~)
Potential energy due to the electron-nucleus interaction reads
Z
Ze2
UeZ = − n(r)d3 r , (12.63)
r
while the potential energy due to electron-electron interaction is
ZZ 2
1 e n(r 0 )n(r) 3 3 0
Uee = d rd r . (12.64)
2 |r − r 0 |

Finally, we need to verify at what r’s the quasiclassical approximation assumed by the Thomas-Fermi model breaks
down. Ignoring variation of χ in the interval 0 < x < 1 (for the sake of estimate), we have

~ ~ r
λ= = √ . (12.65)
2m(−eφ0 + Ze2 /r) mZe2

149
§4 Thomas-Fermi model XII ATOMS

The quasiclassical approximation is valid when

dλ ~ ~2 a
=√ 1 ⇒ r = . (12.66)
dr mZe2 r mZe2 Z

In Fig. 24 the density of the electron cloud in Mercury is calculated using the Hartree-Fock and Thomas-Fermi
methods. Although the Thomas-Fermi method is missing fine details of the distribution it gives correct overall
dependence on the radial coordinate with dramatically smaller computational cost.

FIG. 24:

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §70, §71.

§5. Atom in magnetic field

Consider an atom in magnetic field. Neglecting the Coulomb interactions of electrons we can write its Hamiltonian
as a sum over one-particle Hamiltonians of electrons. The later is given by

1  e 2 e2 Z e~ Ze2
Ĥa = p̂a + A(ra ) − + (σ̂a · B) + L̂a · Ŝa . (12.67)
2m c ra 2mc 2m2 c2 ra3

The last term describes the spin-obit interaction and is one of the relativistic corrections that can be derived from
the Dirac equation. A convenient gauge choice in a constant magnetic field is A = B × r/2. In this gauge ∇ · A = 0
implying [p̂, A] = 0. Using this in (12.67) we obtain

p̂2a Ze2 Ze2 e e2 e~


Ĥa = − + 2 2 3
L̂ a · Ŝ a + (B × ra ) · p̂ a + 2
(B × ra )2 + (σ̂a · B̂) . (12.68)
2m ra 2m c ra 2mc 8mc 2mc

150
§5 Atom in magnetic field XII ATOMS

The last three terms can be simplified as follows


e e2 e~
(B × ra ) · p̂a + (B × ra )2 + (σ̂a · B̂)
2mc 8mc2 2mc
2
e e e
= (ra × p̂a ) · B + 2
(B × ra )2 + (Ŝa · B̂)
2mc 8mc mc
e e2
= (L̂a + 2Ŝa ) · B + (B × ra )2 .
2mc 8mc2
Thus,
p̂2a Ze2 Ze2 e2
Ĥa = − + L̂a · Ŝa − µ̂a · B + (B × ra )2 , (12.69)
2m ra 2m2 c2 ra3 8mc2
where electron’s magnetic moment is
e µB
µ̂a = − (L̂a + 2Ŝa ) = − (L̂a + 2Ŝa ) , (12.70)
2mc ~
see (8.5). In practice, the term proportional to B 2 is small and can be neglected. Magnetic moment of an atom is
defined as
µB
µ=− (L̂ + 2Ŝ) , (12.71)
~
P P
where the angular momentum of atom is L̂ = a L̂a and its spin Ŝ = a Ŝa . In a state with given angular
momentum and spin |lsi (but not necessarily their projections on z-axis), the Hamiltonian of an atom can be written
as (see the references at the end of this section for more details)
Ĥ = Ĥ0 + γ L̂ · Ŝ − µ̂ · B , (12.72)

where γ is a phenomenological constant (i.e. it is fixed by fitting the experimental data) and Ĥ0 is the Hamiltonian
in the absence of the field.
The spin-orbit interaction causes the fine structure splitting of energy levels with respect to the total angular
momentum j. We denote the corresponding energy levels Enj . This phenomenon will be discussed in detail in XIV §7
and V §5. Depending on the relative strength of the last two terms in (12.72) we can distinguish two limiting cases.

A. Weak field. Zeeman effect.

If µB  |Enj − Enj 0 | then −µ̂ · B can be considered as a perturbation of the other terms. For example, in the
hydrogen atom the fine structure splitting of n = 2 levels is 5 · 10−5 eV. Since µB = 5.8 · 10−5 eV/T, the weak field
approximation applies when B  1T .
The stationary states of the non-perturbed Hamiltonian
Ĥ (0) = Ĥ0 + γ L̂ · Ŝ (12.73)
are |njlsmj i (recall (6.70)). The corresponding eigenvalues Enj are degenerated with respect to mj because the
atomic potential is spherically symmetric (22 . Magnetic field possesses only axial symmetry, therefore the degeneracy
with respect to mj is lifted.
According to the perturbation theory, to calculate corrections to the energy spectrum, we need to compute the
expectation values of the operator −µ̂ · B in the states |njlsmj i. We showed in VI §4 that due to the Wigner-Eckart
theorem the following relations hold, see (6.102),(6.103):
 
~mj l(l + 1) s(s + 1)
hjmj |L̂|jmj i = ez 1+ − , (12.74)
2 j(j + 1) j(j + 1)
 
~mj l(l + 1) s(s + 1)
hjmj |Ŝ|jmj i = ez 1− + , (12.75)
2 j(j + 1) j(j + 1)

(22 Essentially, this is an experimental observation. See e.g. Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §68.

151
§6 Atom in electric field XII ATOMS

where I chose z-axis as the direction of the magnetic field: B = ez B. We therefore derive
µB
−hjmj |µ̂ · B|jmj i = Bhjmj |L̂z + 2Sz |jmj i
 ~ 
mj µB l(l + 1) s(s + 1) 2l(l + 1) 2s(s + 1)
= B 1+ − +2− +
2 j(j + 1) j(j + 1) j(j + 1) j(j + 1)
 
mj µB B j(j + 1) − l(l + 1) + s(s + 1)
= 2+ = gµB Bmj , (12.76)
2 j(j + 1)
where
j(j + 1) − l(l + 1) + s(s + 1)
g =1+ (12.77)
2j(j + 1)
is called the Lande factor. Finally, we find

Enjlmj = Enj + gµB Bmj . (12.78)

For example, for electrons s = 1/2, j = l ± 1/2, l = 0, 1, . . ., mj = ±j, ±(j − 1), . . .. Hence the (2j + 1)-fold degeneracy
is lifted. This effect is known as the Zeeman effect. For spinless particles s = 0, thus g = 1.

B. Strong field. Paschen-Back effect.

Strong field limit corresponds to µB  |Enj − Enj 0 |. However, we also assume that µB  |En − En+1 | ∼ e2 /a (in
hydrogen this is about 2 eV for n = 2). Now, the non-perturbed Hamiltonian reads

Ĥ (0) = Ĥ0 − µ̂ · B , (12.79)

while γ L̂ · Ŝ is a perturbation. Eigenfunctions of the Hamiltonian Ĥ0 are states with definite values of L̂2 , L̂z , Ŝ 2 ,
Ŝz , i.e. |nlml sms i = ψnlml (r)χsms . (Note, that in the weak field only J 2 and Jz have definite values, i.e. they are
conserved). Thus,

∆Enml ms = µB B(ml + 2ms ) . (12.80)

This splitting is called the Paschen-Back effect. The sum ml + 2ms can have 2l + 3 possible values, i.e. l + 1, l, l −
1, . . . , −l − 1. The two lowest and two highest components are not degenerate while others are two-fold degenerate
because
(
ml + 1 , if ms = 21 ,
ml + 2ms = (12.81)
ml + 2 − 1 , if ms = − 12 .

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §72,113, Messiah,
“Quantum mechanics”, Part IV, Ch. XVI, §12.

§6. Atom in electric field

A. Atoms other than hydrogen-like

Interaction of an atom with electric field is described by the following term in the Hamiltonian

V̂ = −E · dˆ, (12.82)

where d is the dipole moment. For a single electron d = −er. Choose z-axis along the direction of electric field. The
Hamiltonian of atom in electric field reads

Ĥ = Ĥ0 + eEz , (12.83)

152
§6 Atom in electric field XII ATOMS

where Ĥ0 is the Hamiltonian in the absence of the field. In view of the spherical symmetry of Ĥ0 the eigenstates are
(0) (0) (0) (0) (0)
ψnlm (r) = Rnl (r)Ylm (θ, φ). States with definite Enl are 2l + 1 functions ψ−l , ψ−l+1 , . . ., ψl . (I indicated only the
m-number because n,l are the same). To apply the perturbation theory we need to compute the matrix elements
Z Z ∞ Z π Z 2π
0 (0)∗ ˆ (0) 3 2 3 m m0 0
hm |dz |mi = ψm dz ψm0 d r ∝ −e Rnl (r)r dr Pl Pl cos θ sin θdθ ei(m−m )φ dφ = 0 . (12.84)
0 0 0
That the matrix element vanishes for m 6= m0 is evident. For m = m0 ,
Z 1
(Plm (cos θ))2 cos θ d cos θ = 0 , (12.85)
−1

because Pl (x) has definite parity. Consequently, there is no splitting of energy levels proportional to E.
(0)
Note, that the states ψnml have parity (−1)l , d is a polar vector, i.e. odd under parity. (In contrast, µ is an axial
vector, i.e. even under parity.) Hence, its matrix elements are finite only for transitions between different nearby
parity levels, i.e. l → l ± 1. This can be also readily seen using the following recurrence formula
m
(l − m + 1)Pl+1 (x) = (2l + 1)xPlm (x) − (l + m)Pl−1
m
(x) . (12.86)
According to the perturbation theory, the leading correction to the energy levels reads
(0)
X hnlm|z|n0 l0 m0 ihn0 l0 m0 |z|nlmi
Enlm = Enlm + e2 E 2 (0) (0)
. (12.87)
n0 l 0 Enl − En0 l0

B. Hydrogen-like atoms

The situation is different in a hydrogen atom because in addition to the degeneracy with respect to m, there is also
an additional “accidental” degeneracy with respect to l (see V §7). This means that there are different parity states
with the same energy. Nevertheless, since the ground state n = 1 is not degenerate, the corresponding expectation
value vanishes: h1s|V̂ |1si = 0.
(0)
The first excited state n = 2 is four-fold degenerate and has energy E2 . The corresponding wave functions are
ψ1 ≡ ψ200 = R20 (r)Y00 (θ, φ) , l = 0 (12.88)

ψ2 ≡ ψ210 = R21 (r)Y10  
ψ3 ≡ ψ211 = R21 (r)Y11 l = 1. (12.89)


ψ4 ≡ ψ21,−1 = R21 (r)Y1,−1
The secular equation (7.37) reads

V − E (1) V12 V13 V14
11 2
(1)
V21 V22 − E2 V23 V24
(1) = 0. (12.90)
V31 V32 V33 − E2 V34

V41 V42 V43 V44 − E2
(1)
The only finite matrix elements are
V12 = V21 = −eEh200|z|210i = −eEh200|r cos θ|210i = −3eEa , (12.91)
where the Bohr radius a = ~2 /me2 . This leads to the following equation
  2  2 
(1) (1) 2
E2 E2 − (3eEa) = 0 , (12.92)

(1)
which is solved by E2 = 0, ±3eaE. Thus,
(0) (0)
E = E2 , E = E2 ± 3eaE . (12.93)
This means that the electric field splits the first excited level (and similarly all other excited energy levels) even at the
leading order, partially lifting the degeneracy: states with m = ±1 are still degenerate. Splitting ∆E ∝ E is called the
linear Stark effect in contrast to (12.86) knows as the quadratic Stark effect. It is observed only in the hydrogen-like
atoms that have Coulomb potentials (and therefore degenerate in l). In all other atoms potentials are different from
the Coulomb one, hence levels with different l have different energy.

153
§6 Atom in electric field XII ATOMS

C. Strong fields

Strong field can ionize atom. The probability of such process can be computed using the quasiclassical approximation
as the electron tunneling probability T through the potential barrier. Namely,
 Z z2 
2
T ≈ exp − Im p(z)dz , (12.94)
~ z1

where z1,2 are the turning points and the classical momentum can satisfies

p2
− zeE = −|E0 | , (12.95)
2m
with E0 < 0 being the binding energy. We have
( Z ) ( √ )
2 |E0 |/eE p 4 2m|E0 |3/2
T ≈ exp − 2m(|E0 | − zeE)dz = exp − . (12.96)
~ 0 3~eE

This formula holds when eE  m1/2 |E0 |3/2 /~.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §76,77.

154
§1 Scattering amplitude XIII ELASTIC SCATTERING

XIII. ELASTIC SCATTERING

§1. Scattering amplitude

The scattering problem in quantum mechanics as in the classical one, is to evaluate for a given incident particles
and their interaction, the distribution of outgoing particles. In this course we focus on the elastic scattering in which
identities and internal states of the colliding particles do not change. Elastic scattering of two particles of masses m1
and m2 in their center-of-mass frame can be described by the Schrödinger equation

~2 2
− ∇ ψ(r) + U (r)ψ = Eψ , (13.1)
2m
where r is a relative position of two particle, m = m1 m2 /(m1 + m2 ) is reduced mass and U is the potential energy
of their interaction. It will be convenient to say that particle of mass m scatters on the potential U . I am going to
neglect particle spins until section XIII §9. Let the interaction range be a, so that U is very small at r & a. Since at
large distances r  a particles are free, their energies are positive E > 0 and the spectrum is continuous. Introducing
k 2 = 2mE/~2 we write the Schrödinger equation as

2mU (r)
(∇2 + k 2 )ψ(r) = ψ(r) . (13.2)
~2
Since the incident particles start at r  a their motion is described by a solution to the Schrödinger equation (13.2)
with a definite relative momentum p = ~k:(23

ϕk (r) = eik·r . (13.3)

ϕk is normalized to one particle per unit volume, see III §2, i.e. the probability current density of incident particles is
~ ~k
j= (ϕ∗ ∇ϕk − ϕk ∇ϕ∗k ) = = v. (13.4)
2mi k m
In experiment, one scatters bunches containing macroscopic number of particles. Density of the bunches is chosen
in such a way that only one pair of particles can collide simultaneously. The main observable quantity is the scattering
cross section defined as a ratio of the number of particles scattered per unit time into the solid angle dΩ (i.e. the
transition rate) to the number of incident particles crossing the unit area per unit time. Let j 0 be the probability
current density of the scattered particles and jr0 its radial component. Then, the differential cross section can be
written as follows
jr0 r2 dΩ
dσ = . (13.5)
j
Equivalently, it can be expressed in terms of the transition rate, as
dẇi→f
dσ = , (13.6)
j
see (9.28). The total cross section σ is obtained by integrating (13.5) over the solid angle. Dimension of the cross
section is area. To calculate the cross section we need to find a solution to the Schrödinger equation that at large
distances (r  a) is a superposition of plane wave (13.3) describing the incident particles and a spherical expanding
wave describing the scattered particles. Let p0 = ~k0 be the (relative) momentum of the scattered particles. Energy
conservation implies that k = k 0 . However, k 6= k0 because momentum is not conserved in external field. Angle
θ between the vectors k and k0 is called the scattering angle. Thus, we are seeking a solution with the following
asymptotic form
f (Ω) ikr
ψ ≈ eik·r + e , r  a, (13.7)
r

(23 More precisely, both incident and outgoing particles are described by wave packets. However, in a realistic scattering experiments in
quantum realm, plane waves are a very good approximation of wave packets. See Messiah, “Quantum mechanics”, X.4.

155
§1 Scattering amplitude XIII ELASTIC SCATTERING

where f (Ω) is a certain function of angles called the scattering amplitude. If U (r) is spherically symmetric it depends
only on the scattering angle i.e. f = f (θ). The probability current density of scattered particles described by the
wave function ψ 0 = f eikr /r reads

~ ~k
j0 = (ψ 0∗ ∇ψ 0 − ψ 0 ∇ψ 0∗ ) = |f (Ω)|2 , r  a. (13.8)
2mi mr2
Therefore, (13.5) implies that

dσ = |f (Ω)|2 dΩ . (13.9)

The problem of computing the cross section is thus reduces to the problem of determining the scattering amplitude
from the wave function (13.7).

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §123; Merzbacher,
“Quantum Mechanics” 3rd edition, 13.1-13.4; Messiah, “Quantum mechanics”, X.

§2. Free-particle Green’s function

To solve (13.2) we will employ the Green’s function method that we discussed before in III §6 in a similar context.
Define the Green’s function G(r, r 0 ) as a solution to the follows equation

(∇2 + k 2 )G(r, r 0 ) = δ(r − r 0 ) . (13.10)

Once the Green’s function is found, solution to equation

2mU (r)
(∇2 + k 2 )ψ(r) = ψ(r) (13.11)
~
can be written as
Z
2mU (r0 )
ψ(r) = ϕk (r) + G(r, r 0 ) ψ(r 0 )d3 r0 . (13.12)
~

As we noted before in III §6, this representation has the desired asymptotic form (13.7).
To find the Green’s function we represent it as the Fourier integral expansion (see Appendix A) and use the
three-dimensional generalization of (B11) to obtain
Z
1 0 1
G(r, r 0 ) = 3
eiq·(r−r ) 2 d3 q . (13.13)
(2π) k − q2

Doing the angular integrations we find


Z 2π Z π Z ∞ 0 Z +∞ 0
1 eiq|r−r | 2 1 qeiq|r−r |
G(r, r 0 ) = q dq sin θdθdφ = dq . (13.14)
(2π)3 0 0 0 k2 − q2 4π 2 i|r − r 0 | −∞ k2 − q2

I formally extended integration over q to the negative real axis. The integral over q is evidently divergent at q = ±k.
Actually, these singularities is a mere indication of the fact that we have not yet specified whether the second term on
the right-hand-side of (13.12) is expanding or collapsing spherical wave. The correct asymptotic can be chosen using
a mathematical procedure known as the analytic continuation. The idea is to consider the integrand of (13.14) as an
analytic function of complex variable q and distort the integration contour away from the singularities. One possible
way of doing this is shown in Fig. 25. It is equivalent to adding an infinitely small imaginary part to k → k + i0.
Denote the Green’s function defined in such way as G+ (r, r 0 ). We can now do the integration over q by closing the
contour in the upper-half plane and using the residue theorem as follows
I 0 0 0
0 1 qeiq|r−r | 1 keik|r−r | eik|r−r |
G+ (r, r ) = − 2 dq = − 2 2πi = − . (13.15)
4π i|r − r 0 | C+ (q − k)(q + k) 4π i|r − r 0 | 2k 4π|r − r 0 |

156
§2 Free-particle Green’s function XIII ELASTIC SCATTERING

q
G+

k + i0
k
−k
−k − i0

FIG. 25:

If we choose another prescription k → k − i0 and do similar calculation we get


0
e−ik|r−r |
G− (r, r 0 ) = − . (13.16)
4π|r − r 0 |

Evidently, G+ and G− represent the expanding and collapsing waves respectively and are known as the retarded and
advanced Green’s functions.
We can now write the Schrödinger equation in the form convenient for the scattering problem. Substituting (13.15)
into (13.11) we arrive at
Z 0
m eik|r−r |
ψ(r) = ϕk (r) − U (r 0 )ψ(r 0 )d3 r0 . (13.17)
2π~2 |r − r 0 |

At large distances r  a we can expand


p  
r · r0
k|r − r 0 | ≈ k r2 − 2r · r 0 ≈ kr 1 − 2 = kr − k0 · r 0 , (13.18)
r

where k0 = k rr . Thus, the asymptotic expression for (13.17) has precisely the form of (13.7) with the scattering
amplitude given by
Z
m 0 0 m
f (Ω) = − e−ik ·r U (r 0 )ψ(r 0 )d3 r0 = − hϕk0 |U |ψi . (13.19)
2π~2 2π~2

Eq. (13.19) relates the scattering amplitude to the wave function ψ but it is not a solution to the scattering problem
because ψ itself is not known. It is however a convenient starting point for further development.

∗ Derivations of this section can be represented in a compressed form using symbolic notation that is often used in
the the literature. Consider Schrödinger equation

(Ek − Ĥ0 )ψ = V̂ ψ , (13.20)

where Ĥ0 = p̂2 /2m. A representation containing the expanding spherical wave (13.17) can be formally written as
1
ψ = ϕk + V̂ ψ . (13.21)
Ek + i0 − Ĥ0
Here

Ĥ0 ϕk = Ek ϕk , (13.22)

i.e. ϕk are a complete set of eigenfunctions of Ĥ0 , and Ek = ~2 k 2 /2m are the corresponding eigenvalues.
To show that (13.21) is equivalent to (13.17), expand function V̂ ψ can be expanded in a complete set of eigenfunc-
tions of Ĥ0 :
Z
V̂ ψ = ϕq hϕk |V̂ |ψid3 q , (13.23)

157
§3 Born approximation XIII ELASTIC SCATTERING

where
Z
1 0
hϕk |V̂ |ψi = e−iq·r V̂ (r 0 )ψ(r 0 )d3 r0 . (13.24)
(2π)3
Using (13.23) in (13.21) and recalling (13.22) yields
Z Z
1 3 ϕq hϕk |V̂ |ψi 3
ψ(r) = ϕk + ϕq hϕk |V̂ |ψid q = ϕk + d q. (13.25)
Ek + i0 − Ĥ0 Ek + i0 − Eq

Now, substituting (13.24) and recalling (13.13) we obtain (13.17) again.

• Additional reading: Merzbacher, “Quantum Mechanics” 3rd edition, 13.3, 20.2.

§3. Born approximation

If the potential U (r) can be considered a small perturbation, we can solve the integral equation (13.17) by successive
approximations. Namely, we seeking for a solution in the form
ψ(r) = ϕk (r) + ψ (1) (r) + ψ (2) (r) + . . . . (13.26)
For the first and the second order correction we get
Z 0
m eik|r−r |
ψ (1)
(r) = − U (r 0 )ϕk (r 0 )d3 r0 , (13.27)
2π~2 |r − r 0 |
Z 0
m eik|r−r |
ψ (2) (r) = − U (r 0 )ψ (1) (r 0 )d3 r0 . (13.28)
2π~2 |r − r 0 |
Expanding (where possible) at large distances using (13.18) and substituting into (13.19) we derive
Z  m 2 Z ik|r−r 0 |
m ∗ ∗ 0 e
f (θ) = − 3
ϕk0 (r)U (r)ϕk (r)d r + ϕk0 (r ) U (r 0 )U (r)ϕk (r 0 )d3 rd3 r0 + . . . (13.29)
2π~2 2π~2 |r − r 0 |
These are known as the first and second Born approximations. The reason I did not expand in the integrand of the
second term is that r and r 0 there are not the same as in (13.18); rather they both of the order of a. The first term
in (13.29) can be written in a compact form as
m
f B (Ω) = − hϕk0 |U |ϕk i . (13.30)
2π~2
I assigned the superscript B to indicate the (first) Born approximation. Denote q = k0 −k so that ~q is the momentum
transfer in the scattering process. Then, the scattering amplitude reads
Z
m
f B (Ω) = − e−iq·r U (r)d3 r . (13.31)
2π~2
Notice that the scattering amplitude is proportional to the Fourier image of the potential energy. To find a relationship
between the momentum transfer and the scattering angle write
q 2 = k 2 + k 02 − 2k · k0 = 2k 2 (1 − cos θ) = 4k 2 sin2 (θ/2) ⇒ q = 2k sin(θ/2) . (13.32)
Clearly the range of q is 0 ≤ q ≤ 2k. The differential cross section (13.9) in the Born approximation reads
 m 2
dσ B = |hϕk0 |U |ϕk i|2 dΩ . (13.33)
2π~2

∗ In a particular case of the spherically symmetric potential U (r) (13.31) can be simplified by integrating over the
directions of vector r:
Z Z 1 Z
m 2 iqr cos θ 2m ∞
f B (q) = − U (r)r dr 2π e d cos θ = − U (r) sin(qr)rdr . (13.34)
2π~2 −1 ~q 0

158
§3 Born approximation XIII ELASTIC SCATTERING

It is often convenient to express the differential cross section in terms of the momentum transfer q rather than the
scattering angle θ. Writing dq 2 = 2k 2 sin θdθ we have
π B2 2
dσ B = |f B |2 dΩ = 2π|f B |2 sin θdθ = |f | dq . (13.35)
k2
The total cross section can be computed as follows
Z 4k2
π
σ B
= 2 |f B |2 dq 2 . (13.36)
k 0

The total cross section is a function of the collision energy E which is related to k by E = ~2 k 2 /2m.

A. Conditions of applicability of the Born approximation

Born approximation assumes that the incident wave function is not significantly perturbed by the potential so that
ψ ≈ ϕk . In view of (13.26) and (13.27) this implies that

m Z eik|r−r0 |
0 0 3 0
|ϕk |  U (r )ϕk (r )d r . (13.37)
2π~2 |r − r 0 |

This integral decreases at large r as 1/r, therefore the main contribution to it comes from the region where r ∼ r0 .
Keeping this in mind consider two asymptotic cases.
∗ Low energy scattering ka  1. Hence k|r − r 0 |  1 implying that
Z
eik|r−r0 |
0 0 3 0
U (r )ϕk (r )d r ∼ |U | |ϕk | a2 , ka  1 . (13.38)
|r − r 0 |

Together with (13.37) this yields a condition 1  m|U |a2 /~2 , i.e.
~2
|U |  , ka  1 . (13.39)
ma2
For a spherically symmetric potential we can be a bit more precise by evaluating the integral in (13.37) at r = 0
(where U is usually largest). We have
Z 1 Z ∞
eikr(1+cos θ)
2 2
2πm d cos θ r drU (r)
r  2π~ , (13.40)
−1 0

which yields upon integration over the angle


Z ∞

m U (r) e 2ikr
− 1 dr  k~2 . (13.41)
0

At low energies ka  1 we expand kr  1 to get


Z ∞ Z ∞
1 ~2

2m U (r)krdr  k~2 ⇒ |U (r)|rdr  (13.42)
0 a2 0 ma2
as in (13.39).
∗ High energy scattering ka  1. In this case e2ikr is rapidly oscillating averaging to zero. Dropping the oscillating
term in (13.41) we get
Z ∞ Z
1 ∞ k~2 ~v
m |U (r)|dr  k~2 ⇒ |U |dr  = . (13.43)
0 a 0 ma a
Finally, the Born approximation is applicable at high energies if
~v
|U |  , ka  1 . (13.44)
a

159
§3 Born approximation XIII ELASTIC SCATTERING

2 2
~2
Note, that at high energies k~
ma = (ka) ma2 is much bigger than
~
ma2 . Therefore, the Born approximation at high
energies has wider range of applicability than at low energies.

∗ Examples.

1. The Born approximation can be derived directly from the time-dependent perturbation theory. According to
(9.30), the transition rate is

ẇk→k0 = |hϕk0 |U |ϕk i|2 ρ(Ek0 ) . (13.45)
~
The number of the final states in a unit volume (remember the normalization of ϕk in (13.3)) is
p02 dp0 dΩ p0 m
ρ(Ek0 )dEk0 = 3
= dΩdEk0 , (13.46)
(2π~) (2π~)3

where I used p0 = 2mEk0 . Reading the density of states from (13.46) and putting it into (13.45) we get
2π p0 m
dẇk→k0 = |hϕk0 |U |ϕk i|2 dΩ . (13.47)
~ (2π~)3
Dividing this by j = v = p/m and using (13.6) we find the cross section
dẇk→k0 m2 v 0
dσk→k0 = = |hϕk0 |U |ϕk i|2 dΩ . (13.48)
v 2π~2 v
For elastic scattering v = v 0 and we end up with (13.33) again.

2. Elastic scattering off the Yukawa potential U = − αr e−r/a .


Since the potential is spherically symmetric we can use (13.34) to calculate the scattering amplitude:
Z Z
B 2m ∞ 2mα ∞ −r/a 2mαa2
f (q) = − 2 U (r) sin(qr)rdr = e sin(qr)dr = − . (13.49)
q~ 0 q~2 0 ~2 (1 + q 2 a2 )
The total cross section reads
Z 4k2  2
π~2 mαa2 1
σ(E) = |f B (q)|2 dq 2 = 16π . (13.50)
2mE 0 ~2 1 + 4k 2 a2
In particular, when a → ∞ the Yukawa potential becomes the Coulomb potential in which case (13.50) reduced
to the Rutherford formula
dσ 4m2 α2
= 4 4 . (13.51)
dΩ ~ q
Note, that the total cross section for elastic scattering off the Coulomb potential is divergent at small q corre-
sponding to small scattering angles θ. This happens because the Coulomb potential drops off slowly with r. In
practice, however, either the potential is screened by other charges at large r (yielding the Yukawa potential)
or θ is restricted from below by the detector resolution.
Let us investigate for what values of α and a can the Born approximation be used. Since |U | ∼ α/a, conditions
(13.39) and (13.44) are satisfied when
~2
α , ka  1 , (13.52)
ma
α  ~v , ka  1 . (13.53)
A more accurate estimate can be obtained from (13.41). Since |U | is maximal at r = 0 we can write
Z ∞  p 2 1/2
α −r/a 2ikr 
m e e − 1 dr = mα ln 1 + 4k 2 a2 + φ2  k~2 , tan φ = 2ka . (13.54)
0 r
Since φ ≤ π/2 and logarithm is a slow function of ka we simplify (13.54) for ka & 1 as α  k~2 /m = ~v. For
ka  1 we expand (13.54) to find that mαka  k~2 , i.e. α  ~2 /ma. These agrees with (13.52) and (13.53).

160
§3 Born approximation XIII ELASTIC SCATTERING

3. Elastic scattering of fast electrons on a neutral atom with Z  1.


Potential of a heavy atom can be approximated using the Thomas-Fermi model, see (12.58),(12.59):
 1/3 
e2 Z rZ
U (r) = −eφ(r) = − χ . (13.55)
r b
The scattering amplitude
Z ∞ Z ∞  
B 2m 2me2 Z rZ 1/3
f (q) = − 2 U (r) sin(qr)rdr = χ sin(qr)dr . (13.56)
q~ 0 q~2 0 b
Although exact analytical integration cannot be done, we can derive a rather accurate approximation by taking
into account the properties of the Thomas-Fermi model. To this end, introduce two dimensionless variables
rZ 1/3 qa
y= , ξ= , (13.57)
a Z 1/3
where a = ~2 /me2 is the Bohr radius. Recall that a/Z 1/3 is the atom radius. We then get for the scattering
amplitude
Z
2me2 a2 Z 1/3 ∞  ay  me2 a2 1/3
fB = χ sin(ξy)dy = Z Φ(ξ) , (13.58)
ξ~2 0 b ~2
where we introduced a new universal (viz. same for all atoms) function
Z
2 ∞  ay 
Φ(ξ) = χ sin(ξy)dy . (13.59)
ξ 0 b
From (12.58) it follows that b/a ≈ 0.89.
At ξ  1 only small y’s contribute owing to the rapidly oscillating sinus. We can thus approximate
Z Z ∞
2 ∞ 2
Φ(ξ) ≈ χ(0) sin(ξy)dy = 2 sin y 0 dy 0 , ξ  1 . (13.60)
ξ 0 ξ 0
I used χ(0) = 1, see (12.61). At first sight the integral over y 0 linearly diverges. However, because sinus rapidly
oscillates it actually converges. To make sense of it, one uses the following trick
Z ∞
2 0 2 1 2
Φ(ξ) ≈ 2 lim e−λy sin y 0 dy 0 = 2 lim = 2 , ξ  1. (13.61)
ξ λ→0 0 ξ λ→0 1 + λ2 ξ
In this limit,
2me2 Z
f B (q) = , q  Z 1/3 /a . (13.62)
~2 q 2
This is exactly the Rutherford formula. Thus, if the momentum transfer is much larger than the inverse atom
size, the incident electron does not “see” the atomic electron cloud. Rather it is sensitive only to the Coulomb
field of the nucleus.
At ξ  1 we have
Z ∞  
ay
Φ(ξ) ≈ Φ(0) = 2 χ ydy = const. , q  Z 1/3 /a , (13.63)
0 b
which a Z-independent number. The scattering amplitude is independent of q in this limit. This is because the de
Broglie wavelength associated with the momentum transfer is much large than the atom size λ = ~/q  a/Z 1/3 ,
so that the incident electron does not resolve the atom structure at all.
The total cross section involves integration over all q. However, the main contribution comes from small q, i.e.
ξ  1 as can be seen from (13.61) and (13.63). Indeed, at larger ξ the squared amplitude falls off as 1/ξ 4 . Thus,
Z 4k2  2 Z ∞
π~2 π me2 π
σB = |f B (q)|2 dq 2 ≈ 2 Z 4/3 2
a Φ2 (ξ)dξ 2 = 7.14 2 Z 4/3 . (13.64)
2mE 0 k ~2 0 k

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §126, 139; Messiah,
”Quantum Mechanics”, XIX, 4-8.

161
§4 Partial waves XIII ELASTIC SCATTERING

§4. Partial waves

Many important properties of the scattering amplitude and the cross section can be derived without relying on the
perturbation theory. In this section we develop the partial wave method which is applied to the spherically symmetric
potentials. The idea is that since the orbital angular momentum l is conserved, states with definite values of l will
scatter independently. Therefore, it is advantageous to expand the scattering amplitude into sum of contributions
with fixed angular momenta.
Let z-axis be the direction of the incident wave. We are looking for a solution to the Schrödinger equation (13.2)
which has asymptotic form (13.7):

f (θ) ikr
ψ ≈ eikz + e , r  a, (13.65)
r
In view of the spherical symmetry of the potential we are looking for a solution as an expansion in Legendre polynomials

X
ψ(r, θ) = Cl Rl (r)Pl (cos θ) (13.66)
l=0

with unknown coefficients Cl . We need to fix Cl ’s in such a way that at large r the wave function behaves as (13.65).
The radial wave function χl = rRl satisfies the following equation
 2 
d l(l + 1) 2 2mU
− + k χl (r) = χl (r) . (13.67)
dr2 r2 ~2

Suppose that U → 0 as r → ∞. Then, at large r (13.67) reduces to


 2 
d 2
+ k χl (r) = 0 . (13.68)
dr2

General solution to (13.68) reads


 
Al πl
χl = sin kr − + δl , r → ∞, (13.69)
k 2

where Al and δl are two unknown constants. Constant phase −πl/2 is inserted for future convenience. Since each
term of the sum in (13.66) is proportional to an unknown constant Cl we can set Al = 1. Substituting (13.69) into
(13.66) yields

X  
eikr−iπl/2+iδl e−ikr+iπl/2−iδl
ψ(r, θ) = Cl Pl (cos θ) − , r → ∞. (13.70)
2ikr 2ikr
l=0

This formula gives expansion of the left-hand-side of (13.65). On the right-hand-side we expand the plane wave using
the following formula:

X
eikz = (2l + 1)eiπl/2 jl (kr)Pl (cos θ) , (13.71)
l=0

where jl (x) are spherical Bessel functions, see V §2. Since


 
1 πl
jl (kr) ≈ sin kr − , kr  1 , (13.72)
kr 2

we cast (13.71) in the form



X  ikr−iπl/2 
e e−ikr+iπl/2
eikz = (2l + 1)eiπl/2 Pl (cos θ) − , r → ∞. (13.73)
2ikr 2ikr
l=0

162
§4 Partial waves XIII ELASTIC SCATTERING

It follows now from (13.65),(13.70) and (13.73) that

eikr
ψ − eikz = f (θ)
r
Pl (cos θ) h ikr−iπl/2  iδl   i

X
= e e Cl − (2l + 1)eiπl/2 − e−ikr+iπl/2 e−iδl Cl − (2l + 1)eiπl/2 . (13.74)
2ikr
l=0

Comparing the two sides of the equation (13.74) we conclude that

Cl = (2l + 1)eiδl eiπl/2 , (13.75)


so that the scattering amplitude becomes

i X 
f (θ) = (2l + 1)Pl (cos θ) 1 − e2iδl . (13.76)
2k
l=0

Thus, the problem of calculating the scattering amplitude as a continuous function of the scattering angle reduces
to the problem of computing the phase shifts δl for the infinite number of possible values of the orbital angular
momentum l. As we will see in the following sections, in many cases only few phase shifts contribute to the scattering
making expansion (13.76) very useful.
One often defines another quantity
Sl = e2iδl , (13.77)
which is closely related to the scattering matrix that we will discuss in sec . Notice also a useful relationship
Sl − 1 = 2ieiδl sin δl (13.78)
which makes it easy to separate the real and imaginary parts of the scattering amplitude. Sl is a periodic function of
δl hence the phase shifts are not uniquely fixed. We will require that δl → 0 when U → 0.
The differential cross section reads
dσ 1 X
= |f (θ)|2 = 2 (2l + 1)(2l0 + 1)(1 − Sl )(1 − Sl∗0 )Pl (cos θ)Pl0 (cos θ)
dΩ 4k
ll0
1 X
= 2 (2l + 1)(2l0 + 1)[cos(δl − δl0 ) + i sin(δl − δl0 )] sin δl sin δl0 Pl (cos θ)Pl0 (cos θ) . (13.79)
k 0
ll

To compute the total cross section use the orthogonality of the Legendre polynomials, viz.
Z

Pl (cos θ)Pl0 (cos θ)dΩ = δll0 . (13.80)
2l + 1
Thus we can write the total cross section as a sum over the partial waves, i.e. over terms with given orbital angular
momentum l:

X ∞ ∞
π X 4π X
σ≡ σl = (2l + 1)|1 − Sl |2
= (2l + 1) sin2 δl . (13.81)
k2 k2
l=0 l=0 l=0

σl is called the partial scattering cross section.


It follows from (13.76) that the value of the scattering amplitude at θ = 0, known as the forward elastic scattering
amplitude is

i X 
f (0) = (2l + 1) 1 − e2iδl , (13.82)
2k
l=0

where I used Pl (1) = 1. The imaginary part of this quantity is


∞ ∞
i X  1X
Im f (0) = (2l + 1) 1 − e2iδl = (2l + 1) sin2 δl . (13.83)
2k k
l=0 l=0

163
§5 Properties of phase shifts XIII ELASTIC SCATTERING

Comparing with (13.81) we derive the optical theorem:



σ= Im f (0) . (13.84)
k
This is a remarkable result, as it states that the total cross section is determined by the value of the scattering
amplitude at one point! As we will see later, optical theorem is a consequence of unitarity.
As I have already mentioned, the partial wave method does not rely on the perturbation theory. Indeed, nowhere
in this section I assumed a scattering parameter is small.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §123.

§5. Properties of phase shifts

The partial waves method reduces the scattering problem to determination of the phase shifts δl . In practice it
may turn out that not all phase shifts are equally important. Consider, for example, a potential of range a that falls
off at large distances faster than the centrifugal potential U = ~2 l(l + 1)/2mr2 . Motion of a particle in a classically
allowed region satisfies the inequality

~2 l(l + 1) ~2 k 2
2
≤ =E. (13.85)
2mr 2m
p
The distance of the closest approach to the origin is rl = l(l + 1)/k. The probability to find the particle at r < rl
is exponentially suppressed. The partial waves with a < rl do not reach the region where U is significant and thus
they do not contribute to the scattering cross section. Only partial waves that satisfy
p
l(l + 1) < ka (13.86)

are important at given k. In particular, we realize that s-wave is always important. Also, at low energies ka  1 only
lowest l’s contribute.
To make this analysis more quantitate, let us derive an equation satisfied by δl . The radial wave function χl obeys
(13.67). Let gl the radial wave function for free particle, viz.
 2 
d l(l + 1) 2
− + k gl (r) = 0 . (13.87)
dr2 r2

Both functions satisfy the boundary condition χl (0) = gl (0) = 0. Solution to (13.87) is gl (r) = krjl (kr). Taking the
difference between (13.67) and (13.87) and integrating we obtain
Z r Z
2m r
[gl χ00l − χl gl00 ] dr0 = [gl χ0l − χl gl0 ]r0 =r = 2 U (r0 )χl (r0 )gl (r0 )dr0 . (13.88)
0 ~ 0

Choose r so large that kr  1. This allows us to expand the spherical Bessel function
 

gl (r) ≈ sin kr − , kr  1 . (13.89)
2

Asymptotic from of χl is given by (13.69). Substituting these asymptotics into (13.88) and using the identity
sin α cos β − sin β cos α = sin(α − β) we derive the desired equation
Z
2m ∞
k sin δl = − 2 U (r)χl (r)gl (r)dr . (13.90)
~ 0
Note that δl enters on both sides of this equations.
An approximate solution to (13.90) can be derived using the Born approximation χl ≈ gl . This yields
Z
2mk ∞
sin δlB = − 2 U (r)jl2 (kr)r2 dr . (13.91)
~ 0

164
§5 Properties of phase shifts XIII ELASTIC SCATTERING

Since the Born approximation assumes mU a2 /~2  1, (13.91) implies sin δlB ≈ δlB . Therefore,
Z
2mk ∞
δlB = − 2 U (r)jl2 (kr)r2 dr . (13.92)
~ 0

For a repulsive potential U > 0, the phase shifts are negative δl < 0, whereas for an attractive potential they are
positive. Formula (13.92) is valid for any ka.
In particular case of low energy scattering ka  1, we can expand
(kr)l
jl (kr) ≈ , kr  1 (13.93)
(2l + 1)!!
which gives
Z a  r 2l+1
2m(ka)2l+1
sin δlB ≈− 2 U (r) rdr . (13.94)
~ [(2l + 1)!!]2 0 a
We see that the phase shifts rapidly decrease with l, so at low energies the dominant contribution comes from the
s-wave, while p-wave is a small correction. The corresponding scattering amplitude is
i 1 iδ0 
f (θ) = [(1 − S0 ) + 3 cos θ(1 − S1 )] = e sin δ0 + 3 cos θeiδ1 sin δ1 , (13.95)
2k k
where d and higher waves are neglected. The differential cross section is
dσ 1  
= |f (θ)|2 = 2 sin2 δ0 + 6 cos θ sin δ0 sin δ1 cos(δ0 − δ1 ) + 9 cos2 θ sin2 δ1 . (13.96)
dΩ k
The first and the third terms in the square brackets represent contributions of the s and p-waves respectively, while
the second term is an interference between the s and p waves. The interference term vanishes from the total cross
section.
Experimental measurements give the cross section as a function of the scattering angle θ. The phase shifts can be
found from a fit of (13.96) to the data. This procedure is known as the phase shift analysis. Ideally, once all δl are
known, one can reconstruct the potential U .

∗ Examples.

1. It was found in a scattering experiment that (i) |δ0 | is the largest phase shift and (ii) |δ0 |  1. We wish to
determine the potential. To this end we can use the Born approximation and write
Z Z Z ∞
2mk ∞ 2m ∞ m
δ0 (k) = − 2 U (r)j02 (kr)r2 dr = − 2 U (r) sin2 (kr)dr = − 2 U (r)(1 − cos(2kr))dr , (13.97)
~ 0 k~ 0 k~ 0
were I took into account that j0 (x) = sin x/x. To calculate U in terms of δ0 , write this formula in a form of the
Fourier integral by considering the following quantity
Z ∞ Z
~2 d i ∞
− [kδ0 (k)] = rU (r) sin(2kr)dr = rU (|r|)e−2ikr dr . (13.98)
2m dk 0 2 −∞
I formally extended r to negative values and defined the potential there to be U (−r) when r < 0. (Of course,
the physical values are r ≥ 0). Now, we can easily invert the Fourier integral using (A3),(A4):
Z
i~2 ∞ 2ikr d
rU (|r|) = e [kδ0 (k)] . (13.99)
πm −∞ dk
d
Note that δ0 is an odd function of k, hence dk[kδ0 (k)] is also an odd function of k. Thus,
Z ∞
2~2 d
U (r) = − sin(2ikr) [kδ0 (k)] . (13.100)
πmr 0 dk
For instance, if δ0 (k) = C = const., then, using the same trick as in (13.61), we obtain
Z
2~2 C ∞ ~2 C
U (r) = − sin(2kr)dk = − . (13.101)
πmr 0 πmr

165
§5 Properties of phase shifts XIII ELASTIC SCATTERING

2. In some (rare) cases the phase shifts δl (k) can be calculated exactly. Consider, for instance, the delta-potential

U (r) = −αδ(r − a) , α > 0. (13.102)

At r 6= a the wave function satisfies the free Schrödinger equation with solution for the radial wave function
given by
(
Ajl (kr) , r < a;
R(r) = (13.103)
B 0 jl (kr) + C 0 nl (kr) , r > a ,

where A,B 0 ,C 0 are constants to be determined from matching conditions. In (13.103) we have already used a
boundary condition that requires R(r) to be finite at r = 0.
To calculate the phase shifts we need to represent this solution in the asymptotic form (13.69). To the end, use
the following asymptotic expressions
   
1 πl 1 πl
jl (kr) ≈ sin kr − , nl (kr) ≈ − cos kr − , kr  1 . (13.104)
kr 2 kr 2
If we substitute (13.104) into the second equation of (13.103), the asymptotic form (13.69) is not evident. So,
let us replace the constants B 0 and C 0 with two other constants as follows: B 0 = B cos δl and C 0 = −B sin δl .
Indeed, we have at r > a:
      
B πl πl B πl
R(r) = cos δl sin kr − + sin δl cos kr − = sin kr − + δl . (13.105)
kr 2 2 kr 2
as required. Eq. (13.103) takes form
(
Ajl (kr) , r < a;
R(r) = (13.106)
B[cos δl jl (kr) − sin δl nl (kr)] , r > a ,

Now we are ready to apply the boundary conditions at r = a (see III §2):

Rr<a (a) = Rr>a (a) , (13.107)


0 0 2mα
Rr<a (a) − Rr>a = − 2 R(a) . (13.108)
~
The first equation yields

Ajk (ka) = B[cos δl jk (ka) − sin δl nl (ka)] , (13.109)

while the second one becomes


 
2mα
A jl (ka) + 2 jl (ka) = B [cos δl jl0 (ka) − sin δl n0l (ka)]
0
(13.110)
~ k
The system of equations (13.109) and (13.110) has a solution only if
jl0 (ka) 2mα cos δl jl0 (ka) − sin δl n0l (ka)
+ 2 = . (13.111)
jl (ka) ~ k cos δl jk (ka) − sin δl nl (ka)
Solving this for δl we get the final result
2mα [jl (ka)]2
tan δl =   . (13.112)
~2 k nl (ka) −jl0 (ka) + 2mα 0
~2 k jl (ka) + jl (ka)nl (ka)

α ~2 αma
Consider now the Born approximation of this formula. It applies when U0 ∼ a  ma2 , i.e. when ~2  1. In
this approximation
jl0 (ka) ~2 k
2mα ∼  1. (13.113)
~2 k jl (ka)
2mα ka

166
§6 Scattering at low energies XIII ELASTIC SCATTERING

Using the identity (see Abramowitz, Stegun , “Handbook of Mathematical Functions”, 10.1.6)
1
−nl (z)jl0 (z) + jl (z)n0l (z) = , (13.114)
z2
(13.112) reduces to
2mα
δlB ≈ (ka)2 [jl (ka)]2 . (13.115)
~2 k
The same result can be easily obtained from (13.92).

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §126, see problems 4,5.

§6. Scattering at low energies

At low energies ka  1 scattering amplitude is dominated by the s-wave (l = 0). The corresponding Schrödinger
equation reads
2mU
χ000 (r) + k 2 χ0 (r) = χ0 (r) , χ0 (0) = 0 . (13.116)
~2
We are interested to find the phase in the asymptotic solution

χ0 (r) ≈ C sin(kr + δ0 ) . (13.117)

Many essential features of low energy scattering can be seen in the following two instructive examples.

1. Consider spherical potential well, which is an example of an attractive potential:


(
−U0 , r ≤ a ,
U (r) = (13.118)
0, r ≥ a.

This potential is used to model the nuclear force. Introduce notations:


2mU0
k02 = , κ2 = k 2 + k02 . (13.119)
~2
Eqs. (13.116) read

χ000 (r) + κ2 χ0 (r) = 0 , χ0 (0) = 0 , r ≤ a. (13.120)


χ000 (r) 2
+ k χ0 (r) = 0 , r > a. (13.121)

It’s general solution is

χ0 (r) = C1 sin(κr) , r ≤ a . (13.122)


χ0 (r) = C2 sin(kr + δ0 ) , r > a . (13.123)

Matching at r = a is performed by requiring that the logarithmic derivative of χ0 be continuous. This gives

χ00 (a)
= κ cot(κa) = k cot(ka + δ0 ) . (13.124)
χ0 (a)

Now using identity


tan α + tan β
tan(α + β) = . (13.125)
1 − tan α tan β

167
§6 Scattering at low energies XIII ELASTIC SCATTERING

we get (recalling that ka  1)


k
κ tan(κa) − tan(ka) k
tan δ0 = k
≈ tan(κa) − ka . (13.126)
1+ κ tan(κa) tan(ka) κ

In the limit k → 0, δ0  1. Thus, for the cross section we obtain


 2  2
4π 4π 4π k 1
σ ≈ σ0 = 2 sin2 δ0 ≈ 2 δ02 = 2 tan(κa) − ka = 4πa2 tan(κa) − 1 . (13.127)
k k k κ κa

Often instead of the phase shift δ0 one introduces the scattering length a0 as follows
1
a0 = − lim . (13.128)
k→0 k cot δ0
Replacing δ0 with a0 in (13.127) we write

4π tan2 δ0 k→0 4π (ka0 )2


σ0 = = 2 ≈ 4πa20 . (13.129)
k 2 1 + tan2 δ0 k 1 + (ka0 )2
Thus, a0 characterizes effective size of the target. In our case,
   
1 k tan(κa)
a0 = − lim tan(κa) − ka = −a −1 . (13.130)
k→0 k κ κa

tan(κa)-κa

κa
π 3π
π
2 2

-5

To better understand the properties of the cross section, I plotted tan(κ − 1) − κa in Fig. 1. We can distinguish
several key features:
π
- Energies satisfying 0 < κa < 2, correspond to δ0 > 0, a0 < 0 and describes scattering off an attractive
potential.
π
- Energies satisfying 2 < κa < π, correspond to δ0 < 0, a0 > 0 and describes scattering off an effectively
repulsive potential.
- When κa → π2 the scattering length a0 and the cross section diverge: σ0 → ∞. This phenomenon is called
the zero energy resonance. In practice there is not infinity there but rather a sharp maximum.
- When tan(κa) = κa, a0 = 0 implying that at the corresponding energies the total cross section vanishes
and the target becomes invisible. This is called the Ramsauer-Townsend effect.
Note, that at k → 0, κ ≈ k0 . Also, k0 a = π2 + πn are the conditions for the presence of s-levels (with E = 0) in
the spherical well. Thus, σ0 → ∞ exactly when the first, second etc. s-levels appear in the well.

2. Consider low energy scattering on a spherical potential well give by


(
U0 , r ≤ a ,
U (r) = (13.131)
0, r ≥ a.

168
§6 Scattering at low energies XIII ELASTIC SCATTERING

This is an example of a repulsive potential. Introduce notations:


2mU0
k02 = , κ2 = k02 − k 2 > 0 . (13.132)
~2
At low energies k  k0 , implying κ ≈ k0 . Eqs. (13.116) read

χ000 (r) − κ2 χ0 (r) = 0 , χ0 (0) = 0 , r ≤ a. (13.133)


χ000 (r) 2
+ k χ0 (r) = 0 , r > a. (13.134)

It’s general solution is

χ0 (r) = C1 sinh(κr) , r ≤ a . (13.135)


χ0 (r) = C2 sin(kr + δ0 ) , r > a . (13.136)

Continuity of the logarithmic derivative at r = a:


k
tan δ0 = tanh(κa) − ka . (13.137)
κ
Now, since tanh(k0 a) ≤ 1 it follows that δ0 → 0 when k → 0. If U0 → ∞ (infinitely high barrier), then κ → ∞
and δ0 = −ka0 .
The cross section at k → 0 reads
 2
4π tanh(k0 a)
σ0 = sin2 δ0 = 4πa2 −1 , (13.138)
k2 k0 a

and displayed in Fig. 2.

σ0 /4πa2

0.8

0.6

0.4

0.2

k0 a
2 4 6 8 10

As k → 0 the asymptotic form of χ0 can be further simplifies:


r→∞ C C
χ0 (r) ≈ sin(kr + δ0 ) = [sin(kr) cos δ0 + cos(kr) sin δ0 ]
k k
k→0 C C
≈ [kr cos δ0 + sin δ0 ] ≈ (kr + δ0 ) = C(r − a0 ) , (13.139)
k k
where I used (13.128). Eq. (13.139) implies that by expanding χ0 (r) at large r and taking the limit k → 0 (which is
the same as E → 0) we can obtain a0 as a coefficient in front of the term r0 (with minus sign).

∗ Example. Calculate the scattering amplitude a0 for the potential U (r) = −U0 e−r/a .
To solve this problem, consider s-wave at E = 0 which satisfies the following Schrödinger equation:
2mU0 2mU0
χ000 (r) = χ0 (r) = − 2 e−r/a χ0 (r) ≡ −k02 e−r/a χ0 (r) . (13.140)
~2 ~

169
§7 Scattering at high energies XIII ELASTIC SCATTERING

To solve it make a replacement x = e−r/2a . Now we consider χ0 to be a function of x and render (13.140) to be

d2 χ0 1 dχ0
+ + 4a2 k02 χ0 = 0 . (13.141)
dx2 x dx
It’s (exact) general solution is

χ0 (x) = C1 J0 (2ak0 x) + C2 N0 (2ak0 x) . (13.142)

Now we apply the boundary condition

χ(x = 1) = 0 = C1 J0 (2ak0 ) + C2 N0 (2ak0 ) . (13.143)

Thus,

χ0 (x) = C[N0 (2ak0 )J0 (2ak0 x) − J0 (2ak0 )N0 (2ak0 x)] . (13.144)

Next can find the asymptotic of (13.142) at r → ∞, which corresponds to x → 0, by using the following expressions

z2 2 z 2γ 1
J0 (z) ≈ 1 − , N0 (z) ≈ J0 (z) ln + + (1 − γ)z 2 , z  1, (13.145)
4 π 2 π 2π
where γ = 0.577 . . . is the Euler’s constant. We have
   
2 2γ
χ0 ≈ C −J0 (2ak0 ) ln(ak0 x) + + N0 (2ak0 ) (13.146)
π π
   
2 2γ r
= C −J0 (2ak0 ) ln(ak0 ) + − + N0 (2ak0 ) (13.147)
π π aπ
"  #

CJ0 (2ak0 ) −J0 (2ak0 ) π2 ln(ak0 ) + π + N0 (2ak0 )
= + r = C̃(r − a0 ) , (13.148)
πa J0 (2ak0 )/πa

where the last equation is the same as (13.139). We thus immediately obtain the scattering length
  
πa 2 2γ
a0 = − N0 (2ak0 ) − J0 (2ak0 ) ln(ak0 ) + . (13.149)
J0 (2ak0 ) π π

The total cross section is σ = 4πa20 .

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §132.

§7. Scattering at high energies

A. Wave function

High energy scattering is when ka  1. A free particle moving in the positive z direction is described by the wave
function ψ = eikz . which satisfies the Schrödinger equation:
2mU
∇2 ψ + k 2 ψ = ψ, (13.150)
~2
with U = 0. To find the effect of U , look for a solution in the form

ψ = eikz ϕ . (13.151)

To obtain an equation for ϕ, substitute (13.151) into (13.150) (I use a notation k = kez ):

∇ψ = eikz ∇ϕ + ikϕeikz , (13.152)

∇2 ψ = eikz ∇2 ϕ + 2ik · ∇ϕ − k 2 ϕ . (13.153)

170
§7 Scattering at high energies XIII ELASTIC SCATTERING

Before proceeding further, let us estimate the relative magnitude of various terms in (13.153). Since k · ∇ϕ ∼ ka |ϕ|,
and |∇2 ϕ| ∼ |ϕ|/a2 we have(24

k · ∇ϕ

∇2 ϕ ∼ ka  1 . (13.154)

This indicates that the second order derivatives can be dropped resulting in the following first order equation
∂ϕ 2mU
2ik = ϕ. (13.155)
∂z ~2
It can be easily integrated with the result
 Z z 
im 0 0
ϕ(r) = A exp − 2 U (r )dz . (13.156)
~ k −∞

Finally,
 Z z 
ikz i 0 0
ψ(r) = e exp − U (r )dz . (13.157)
~v −∞

where we set A = 1 so that when U = 0 we obtain the plane wave solution normalized to one particle per unit volume.
Introduce a convenient notation r = zez +b, such that b·ez = 0. Transverse vector b is called the impact parameter.
The wave function far away from the potential can be written as

ψ(r) = eikz S(b) , z  a, (13.158)

where
Z ∞
1
S(b) = e 2iδ(b)
, δ(b) = − U (r 0 )dz 0 . (13.159)
2~v −∞

It is important to emphasize, that (13.157) is a general result valid for a potential of any strength. In the Born
~2
approximation, |U |  ~v
a = ma2 ka, so that we can expand
 Z z 
B ikz i 0 0
ψ (r) ≈ e 1− U (r )dz . (13.160)
~v −∞
In the following we are not assuming the Born approximation unless indicated otherwise.

B. Scattering amplitude

To calculate the scattering amplitude, start with (13.19)


Z
m 0 0
f (Ω) = − 2
e−ik ·r U (r 0 )ψ(r 0 )d3 r0 . (13.161)
2π~
Using (13.151) and (13.155) we can write

~2 ∂ϕ
U ψ = U eikz φ = 2ik eikz . (13.162)
2m ∂z
Substituting (13.162) into (13.161) we derive
Z
m ~2 ik 0 0 0 ∂ϕ 3 0
f =− e−ik ·r eikz d r . (13.163)
2π~2 m ∂z 0

(24 Compare this with the WKB approximation.

171
§7 Scattering at high energies XIII ELASTIC SCATTERING

Momentum transfer is introduced in a usual way q = k0 − kez .


Let us assume that in addition to ka  1 the potential itself is much smaller than the collision energy E, i.e. |U |  E
(this condition is more general than the Born approximation and is called sometimes the eikonal approximation)(25 .
In this case the trajectory of the scattering particle is approximately the straight line. In other words, momentum
transfer is small q  k. Then, q · k = (k0 − k) · k = k0 · k − k 2 ≈ k 2 − k 2 = 0, i.e. q at high energies is approximately
transverse to the collision axis z. With this observation in mind we write (13.163) as
Z Z Z
ik 0 ∂ϕ ik ∂  −iq·b0  2 0 0 ik 0
f =− e−iq·b d3 0
r = − e ϕ d b dz = − e−iq·b (ϕ|z=∞ − ϕ|z=−∞ ) d2 b0 dz 0 (13.164)
2π ∂z 0 2π ∂z 0 2π
Now,
 Z ∞ 
i 0 0
ϕ|z=−∞ = 1 , ϕ|z=∞ = exp − U (b, z )dz = S(b) . (13.165)
~v −∞

Therefore, finally
Z
ik 0
f (q) = − e−iq·b [S(b) − 1]d2 b . (13.166)

The forward scattering amplitude at θ = 0 is


Z
ik
f (0) = − [S(b) − 1]d2 b . (13.167)

Thus, the total cross section
Z Z Z

σ= Im f (0) = 2 Re [1 − S(b)]d2 b = 2 [1 − cos(2δ)]d2 b = 4 sin2 δ(b)d2 b . (13.168)
k

|U |a
a . In this case δ ∼ ~v  1. Thus, S − 1 ≈ 2iδ and
The Born approximation applies when |U |  ~v
Z Z
ik 0 m
f = − 2i e−iq·b δd2 b = − U (r)e−iq·b d3 r0 . (13.169)
2π 2π~2

∗ Examples.

1. Calculate the total cross section for elastic scattering off spherical well
(
−U0 , r < a ,
U= (13.170)
0, r > a.

Suppose a particle moves along the z-axis at impact parameter b (it is also a√distance from the origin). The
trajectory of the particle is approximately the straight line. At r = a, z1,2 = ± a2 − b2 which are the entrance
and exit points. We have from (13.159)
Z ∞ Z z2
1 1 U0 p 2
δ(b) = − U dz = − U0 = a − b2 . (13.171)
2~v −∞ ~v z1 ~v

The cross section (13.168) reads


Z a  p 
U0 1 + 2λ2 − cos(aλ) − 2λ sin(2λ) U0 a
σ = 8π sin2 a2 − b2 bdb = 8πa2 , λ= . (13.172)
0 ~v 8λ2 ~v

(25 ~2 ~2
|U |  E implies that |U |  ma2
(ka)2 , which still allows for |U | & ~v
a
∼ ma2
ka.

172
§7 Scattering at high energies XIII ELASTIC SCATTERING

At λ  1 (Born approximation) we have to expand the numerator up to and including terms proportional to
λ4 . This yields:

σ = 2πa2 λ2 , λ  1, (13.173)

whereas in the strong field limit λ  1 the cross section becomes energy independent

σ = 2πa2 , λ  1. (13.174)

2. Consider scattering on the Yukawa potential U = g(E)e−r/a , where g(E) = g0 (E/E0 )n is a function of energy
E, with g0 , E0 and n being positive constants.(26 Let us determine the asymptotic behavior of the total cross
section at high energies.
According to (13.159) the total phase shift can be computed as follows:
Z ∞ Z ∞ √
1 1
e− z +b /a dz .
2 2
δ(b) = − U dz = − g(E) (13.175)
2~v −∞ ~v −∞

This integral is rapidly decreasing function of b. As we will see below the main contribution to the cross section
arises from large impact parameters b  a. Bearing this in mind, we expand the square root in the exponent of
(13.175) treating z/b as a small parameter:
Z ∞ √
1 z2 1
 
1
−a b+ 2b
δ(b) = − g(E) e 2
dz = − g(E)e−b/a 2πba . (13.176)
~v −∞ ~v

Let b0 be such that |δ(b0 )| = 1, i.e.


 n p
1 E
− g0 e−b0 /a 2πb0 a = 1 . (13.177)
~v E0

Since at b  b0 the phase shift is exponentially suppressed, (13.168) implies


Z b0 Z ∞ Z b0
2 2
σ=4 2
sin δ d b + 2 2
sin δ d b ≈ 4 sin2 δ d2 b ≈ 2πb20 . (13.178)
0 b0 0

I used the fact that average value of sin2 δ is 1/2. b0 can be found from (13.177) by taking logarithm of both
sides:
E b0 1
n ln − + ln b0 ≈ 0 , (13.179)
E0 a 2

where I assumes that E  E0 and therefore the ln EE0 is a big number. Thus,

E
b0 ≈ an ln , E  E0 . (13.180)
E0
We see that b0  a which is consistent with our previous assumption after (13.176) that typical b is much larger
than a. (This can be directly seen in (13.177): when E  E0 this equation is satisfied if b0  a). Finally,
E
σ = 2πa2 n2 ln2 . (13.181)
E0
This formula indicates the fastest possible increase of the total cross section with energy, known as the Froissart
theorem.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §45,131.

(26 Dependence of g on energy if referred to as the running coupling constant.

173
§8 Elastic scattering in Coulomb potential XIII ELASTIC SCATTERING

§8. Elastic scattering in Coulomb potential

The partial wave method is based on the observation that the asymptotic expression for the radial wave function is
(13.69) with constant phase shifts δl . This is true only if the potential falls off steeply enough with distance. As the
we will see, the radial wave function for motion in the Coulomb potential cannot be represented in the form (13.69)
due to its slow decrease with r.
2
To understand the problem with applying the partial wave analysis to the Coulomb potential U = ± Zer , we can
employ the quasiclassical approximation because at large r, U is small and changes smoothly. The radial wave function
for the s-wave reads (see (10.132)):
 Z r 
A 1
χ0 = p sin p(r0 )dr0 + γ , (13.182)
p(r) ~ 0
p
where γ is a constant and p = 2m(E − U ). Let r0 be such that E  U (r0 ). At r > r0 we can expand
p mU
≈k− 2 . (13.183)
~ ~ k
Thus,
 Z r0 Z r  
A 1 mU A
χ0 = p sin p(r0 )dr0 + k− 2 dr0 + γ ≡ p sin(kr + δ0 ) , (13.184)
p(r) ~ 0 r0 ~ k p(r)
where
Z r0 Z r
1 0 0 mU 0
δ0 = p(r )dr − kr0 + γ − dr . (13.185)
~ 0 r0 ~2 k
While the first three terms are constants, the fourth one is logarithmic in r, so that at larger r the phase shift δ0 ∼ ln r
is not constant. Evidently, if the potential were of the type U ∼ 1/r1+n with positive n, then δ0 we have been a
constant a large r. This problem with the Coulomb potential persists to higher partial waves as well. Indeed, the
centrifugal potential falls off as 1/r2 , i.e. faster than the Coulomb potential, and thus does not impede with the
logarithmic behavior at large r:
δl ∼ ln r , r → ∞. (13.186)

Fortunately, scattering problem in the Coulomb potential can be solved exactly. To this end we introduce the
parabolic coordinates ξ,η,φ via their relation to the Cartesian ones(27 :
p p 1
x= ξη cos φ , y= ξη sin φ , z= (ξ − η) . (13.187)
2
Solving for ξ,η,φ we can write
y
ξ = r+z, η = r−z, φ = arctan , (13.188)
x
p
where r = x2 + y 2 + z 2 as usual. It can also be expressed as r = 21 (ξ + η). We need the Laplacian in parabolic
coordinates:
    
2 4 ∂ ∂ ∂ ∂ 1 ∂2
∇ = ξ + η + . (13.189)
ξ + η ∂ξ ∂ξ ∂η ∂η ξη ∂φ2

Suppose that z is the collision axis, so k = kez . Due to the axial symmetry the wave function ψ is a function of
only ξ and η. (Similarly, ψ in (13.65) is a function of r and θ, but not φ). We can write the Schrödinger equation as
    
~2 4 ∂ ∂ ∂ ∂ 2Z1 Z2 e2 ~2 k 2
− ξ + η ± ψ= ψ. (13.190)
2m ξ + η ∂ξ ∂ξ ∂η ∂η ξ+η 2m

(27 Don’t be confused: there are many different conventions to define the parabolic coordinates.

174
§8 Elastic scattering in Coulomb potential XIII ELASTIC SCATTERING

This equation can be solved by separation of variables. The final result is


ξ−η
ψ(ξ, η) = Aeik 2 F (−iλ, 1, ikη) , (13.191)

where I introduced a parameter

Z1 Z2 e2 m Z1 Z2 e2
λ=± = ± . (13.192)
~2 k ~v
Function F is a confluent hypergeometric function and A is a constant (overall normalization).
To find the scattering amplitude we have to consider the asymptotic of (13.191) at r → ∞. To this end we need
the following formula that can be found in mathematical handbooks:
 
Γ(β)(−u)−α α(α − β + 1) Γ(β) u α−β
F (α, β, u) = 1− + e u , u  1. (13.193)
Γ(β − α) u Γ(α)

After some simple algebra keeping in mind that ξ − η = 2z and η = r − z = r(1 − cos θ) we arrive at
πλ   
e2 λ2 ikz+iλ ln[k(r−z)] f (θ) ikr−iλ ln(2kr)
ψ(r, θ) = A 1− e − e , (13.194)
Γ(1 + iλ) 2ikr sin2 (θ/2) r

where the scattering amplitude

λΓ(1 + iλ)e2iλ ln sin(θ/2)


f (θ) = . (13.195)
2kΓ(1 − iλ) sin2 (θ/2)

Terms proportional to λ describe distortion of the incident wave due to the Coulomb interaction at large r.
To compute the cross section we need to know the current density of incident and scattered waves j and j 0 . The
incident wave is given by
πλ  
e2 λ2
ϕ=A 1− eikz+iλ ln[k(r−z)] . (13.196)
Γ(1 + iλ) 2ikr sin2 (θ/2)

Its current density at r → ∞ is


2
e πλ
~ ~k 2
j= (ϕ∗ ∇ϕ − c.c.) = |A|2 ez . (13.197)
2mi m Γ(1 + iλ)

I used the fact that ∂r ϕ ∼ ϕ/r, while ∂z ϕ ∼ kϕ, thus only z derivative contributes. The scattered wave reads
πλ
e2 f (θ) ikr−iλ ln(2kr)
ψ 0 = −A e , (13.198)
Γ(1 + iλ) r

implying that the scattered particle flax into dΩ is


2
e πλ
0 2
2 ~k
2
j · er r dΩ = |A| |f (θ)|2 dΩ . (13.199)
m Γ(1 + iλ)

The cross section thus becomes


 2
dσ j 0 · er r2 dΩ Z1 Z2 e2
= = |f (θ)|2 = . (13.200)
dΩ j 2mv 2 sin2 (θ/2)

Amazingly, this exact result coincides both with the classical Rutherford formula and with the quantum mechanical
Born approximation.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §135.

175
§8 Elastic scattering in Coulomb potential XIII ELASTIC SCATTERING

§9. Exchange effects for identical spin-0 particles

Consider a collision of two identical spin-0 particles. Their wave function must be symmetric under an interchange
of the particles. This restriction leads the exchange effects. Let r1 and r2 be the positions of two particles. Their
relative position in the center-of-mass frame is r = r1 − r2 . Suppose that particles collide along the z-axis. If we
neglect the exchange effects (as we have done thus far), then the wave function asymptotic has form

f (θ) ikr
ψ(r) = eikz + e , r → ∞. (13.201)
r
To symmetrize this function note that interchange r1 ↔ r2 leads to r → −r. As the result r → r, θ → π − θ,
φ → φ + π. Thus, the symmetrized wave function is

f (θ) + f (π − θ) ikr
ψs (r) = eikz + e−ikz + e , r → ∞. (13.202)
r
We normalized this function such that the flux density of each particle equals v = ~k/m (one particle per unit volume),
where m is the reduced mass. As seen on Fig. 26 there are two possible processes that connect the same initial and
final states.

θ θ

FIG. 26:

Flux of the scattered particles is given by

j 0 · er r2 dΩ = v|f (θ) + f (π − θ)|2 dΩ . (13.203)

Consequently, the cross section is

dσ = |f (θ) + f (π − θ)|2 dΩ . (13.204)

Had we neglected the quantum symmetry requirement, we would have written the cross section as a sum of the two
cross sections represented the two possible channels depicted in Fig. 26, viz.

dσcl = {|f (θ)|2 + |f (π − θ)|2 } dΩ . (13.205)

Unlike the classical formula (13.205) with sums the cross sections, the quantum one (13.204) sums the amplitudes.
As a result, the quantum expression contains an interference term.
As a specific example consider scattering of two identical spin-0 particles. For instance, αα (α is a helium nucleus)
or π ± π ± scattering at distances larger than the range of the nuclear force (∼ 1 fm). Interaction between these
particles is electromagnetic and is described by the Coulomb potential. The scattering amplitude is given by (13.195).
Substituting into (13.204) we obtain
 2  
Z 2 e2 1 1 2 cos[λ ln tan2 (θ/2)]
dσ = + + dΩ . (13.206)
~mv 2 sin4 (θ/2) cos4 (θ/2) sin2 (θ/2) cos2 (θ/2)

176
§9 Exchange effects for identical spin-0 particles XIII ELASTIC SCATTERING

The third term on the right-hand-side describes the characteristic quantum mechanic interference effect. Scattering
of two particles interacting through the Coulomb force with account of the symmetry effects is knows as the Mott
scattering.
The exchange (interference) term is largest at θ = π2 . At this angle
 2  
Z 2 e2 1 2
dσ|θ=π/2 = √ ·2+ √ √ = dσcl |θ=π/2 dΩ , (13.207)
~mv 2 (1/ 2)4 (1/ 2) (1/ 2)2
2

i.e. the cross section is twice as large as predicted by the classical formula. Fig. 27 and Fig. 28 display the classical
and quantum cross section for scattering of two spin-0 particles.
The fact the exchange effects are quantum can be also seen mathematically: take the classical limit ~ → 0 (see
X §1), implying λ → ∞ (see (13.192)). In this limit the exchange term rapidly oscillates and thus vanishes when
averaged even over small (but finite) solid angle.
20
10

-100 -50 50 100


-10
-20

FIG. 27: Coulomb scattering of identical particles without exchange effects. Plotted are the first two terms of the expression
in the curly brackets of (13.206). In each denominator I replaced sin2 (θ/2) → sin2 (θ/2) + 0.1, cos2 (θ/2) → cos2 (θ/2) + 0.1 to
model detector resolution.

20
10

-100 -50 50 100


-10
-20

FIG. 28: Coulomb scattering of identical particles with even total spin. Plotted is the expression in the curly brackets of
(13.206). Detector resolution is taken into account in the same way as in Fig. 27. λ = 0.2.

§10. Exchange effects for identical particles of arbitrary spin

If the colliding particles have spin, the state of the system is described by a function of coordinates and spins.
Generally, the total angular momentum is conserved. Sometimes spin-flip transitions can be neglected. In such cases
the total angular momentum and total spin are conserved separately. The corresponding wave function Φ(q1 , q2 ) is a
product of the coordinate ψ(r1 , r2 ) and spin χ(s1 , s2 ) wave functions. It is this case that we are going to discuss in
this section.
Notations: each particle spin is denoted by s, while the total spin of their state by S.

A. s = 1/2

Consider first a collision of two spin-1/2 particles. The total spin can be either zero, corresponding to the singlet
state, or one, corresponding to the triplet of states, see XII §2:
1 
χ0 = √ | ↑i1 | ↓i2 − | ↓i1 | ↑i2 (13.208)
2
 
 1
 √2 | ↑i1 | ↓i2 + | ↓i1 | ↑i2
χ1 = | ↑i1 | ↑i2 (13.209)


| ↓i1 | ↓i2

177
§10 Exchange effects for identical particles of arbitrary spin XIII ELASTIC SCATTERING

In the singlet state, ψ(r) must be symmetric. Thus,


dσ0 = |f (θ) + f (π − θ)|2 dΩ , S = 0. (13.210)
In the triplet case, ψ(r) must be antisymmetric, viz.
f (θ) − f (π − θ) ikr
ψa (r) = eikz − e−ikz + e , r → ∞. (13.211)
r
Thus,
dσ1 = |f (θ) − f (π − θ)|2 dΩ , S = 1. (13.212)
This yield the following cross section
 2 2 2  
Z e 1 1 2 cos[λ ln tan2 (θ/2)]
dσ1 = + − dΩ . (13.213)
~mv 2 sin4 (θ/2) cos4 (θ/2) sin2 (θ/2) cos2 (θ/2)
Notice the negative sign in front of the exchange term. It implies that dσ1 = 0 when θ = π/2. This cross section is
plotted in Fig. 29.
15
10
5
-100 -50 50 100
-10
-15

FIG. 29: Coulomb scattering of identical particles with odd total spin. Plotted is the expression in the curly brackets of
(13.206). Detector resolution is taken into account in the same way as in Fig. 27. λ = 0.2.

If one studies scattering of unpolarized particle beams, then the cross section must be averaged over all possible
spins of the initial particles and summed over all possibles spins of the final particles. Since there are 3 states with
spin 1 and 1 state with spin 0, we have on average
1 3
dσ = dσ0 + σ1 . (13.214)
4 4

B. Any s

In the case of arbitrary spin s, there are (2s + 1)2 spin states with different total spin and its z-component. The
total spin S can have the following values:
S = 2s, 2s − 1, 2s − 2, . . . , 0. (13.215)
Let |smi be one-particle spinors. Then the wave function of the two-particle system with the total spin S reads (see
VI §3)
X
χSM (s1 , s2 ) ≡ |s1 s2 SM i = hs1 m1 s2 m2 |s1 s2 SM i|s1 m1 i|s2 m2 i , (13.216)
m1 ,m2

where hs1 m1 s2 m2 |s1 s2 SM i are the Clebsch-Gordan coefficients. One of the properties of states with fixed total S
and M that we need here is (6.82):
|s1 s2 SM i = (−1)s1 +s2 −S |s2 s1 SM i = (−1)2s−S |s2 s1 SM i . (13.217)

In other words, permutation of the particles yields a factor of (−1)2s−S in the spin wave function of the system. The
wave function of the two-particle system
ΦSM (q1 , q2 ) = ψ(r1 , r2 )χSM (s1 , s2 ) (13.218)
must be either symmetric or antisymmetric depending on whether the particles are bosons or fermions. This fact can
be expressed mathematically as
ΦSM (q2 , q1 ) = (−1)2s ΦSM (q1 , q2 ) . (13.219)

178
§11 Quasi-elastic scattering of electron by atom XIII ELASTIC SCATTERING

If s is integer, ΦSM is symmetric, if s is half-integer it is antisymmetric. We can now deduce the symmetry properties
of the coordinate wave function:

ΦSM (q2 , q1 ) = ψ(r2 , r1 )χSM (s2 , s1 )


= (−1)2s ΦSM (q1 , q2 ) = (−1)2s ψ(r1 , r2 )χSM (s1 , s2 ) = (−1)2s ψ(r1 , r2 )(−1)2s−S χSM (s2 , s1 ) . (13.220)

Since (−1)4s = 1 we conclude that

ψ(r1 , r2 ) = (−1)S ψ(r2 , r1 ) . (13.221)

This means that the symmetry properties of the coordinate wave function are determined only by the total spin of a
state.
Consequently, the cross sections for particles in a state with even and odd total spins are

dσe = |f (θ) + f (π − θ)|2 dΩ , S = 0, 2, 4, . . . (13.222)


2
dσo = |f (θ) − f (π − θ)| dΩ , S = 1, 3, 5, . . . (13.223)

In the case of Coulomb interaction these are depicted in Fig. 28 and Fig. 29.

Important to remember: all results in this section are derived assuming that there is no spin-orbit interaction and
consequently no spin-flip.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §137.

§11. Quasi-elastic scattering of electron by atom

Thus far we discussed only elastic scattering when neither internal states nor identities of colliding particles change.
In this section we consider scattering of an electron by an atom with in which the internal state of atom changes from
a stationary state n to another stationary state n0 . This is an inelastic process. However, since the final state still
contains the same particles as the initial one: electron and atom (although atom is possibly excited), the scattering
process is similar to elastic scattering. This is why it is called quasi-elastic scattering. If momentum transfer is too
small to kick the atomic electron out of atom ((~q)2 /m  Z 2 e2 /a, see (12.10)), the exchange effects can be neglected
since the scattering electron is always distinguishable. At high momentum transfers we will have to take the exchange
effects into account.

A. No exchange effects

Consider an electron incident on an atom at rest. Suppose that atom is hydrogen-like, i.e. contains only one electron
whose position we will label by r2 . Position the incident is labeled by r1 . Hamiltonian of electron in atom is given by

~2 2 Ze2
Ĥ2 = − ∇ − . (13.224)
2m 2 r2

Let ϕn (r2 ) be the functions of the stationary states with energies En , i.e. Ĥ2 ϕn = En ϕn . Hamiltonian of the incident
electron is
~2 2
Ĥ1 = − ∇ . (13.225)
2m 1
It’s eigenfunctions are eik·r1 . Interactions between the electrons and atom are described by

e2 Ze2
U (r1 , r2 ) = − . (13.226)
r12 r1

179
§11 Quasi-elastic scattering of electron by atom XIII ELASTIC SCATTERING

At r → ∞, one can neglect interactions, hence the wave functions of the entire system long before the collision 0
can
be described by the wave function ϕn (r2 )eik·r1 , while its wave function long after the collision by ϕn0 (r2 )eik ·r1 . The
wave functions are normalized as follows
Z
1 X 0
ϕn (r2 )ϕ∗n (r20 )eik·(r1 −r1 ) d3 k = δ(r1 − r10 )δ(r2 − r20 ) . (13.227)
(2π)3 n

The total energy E is conserved meaning that

~2 k 2 ~2 k 02
En + = En0 + =E. (13.228)
2m 2m
Final states with different n0 and k 0 satisfying this condition are called open scattering channels. We need to find a
solution to the Schrödinger equation
 
E − Ĥ2 − Ĥ1 ψ(r1 , r2 ) = U (r1 , r2 )ψ(r1 , r2 ) (13.229)

~2 k 2
that contains incident and scattered waves. The total energy is fixed by the initial state E = 2m + En . As before,
we introduce the Green’s function as a solution to the following equation

(E − Ĥ1 − Ĥ2 )G(r1 , r2 ; r10 , r20 ) = δ(r1 − r10 )δ(r2 − r20 ) . (13.230)

Using the symbolic notation of (13.21) we can write

δ(r1 − r10 )δ(r2 − r20 )


G+ (r1 , r2 ; r10 , r20 ) = . (13.231)
E − Ĥ1 − Ĥ2 + i0
The meaning of this notation becomes clear if we replace the delta-functions using (13.227) and note that the
ϕn (r2 )eik·r1 is an eigenfunction of the operator in the denominator, so we can replace the operators by the cor-
responding eigenvalues:
X Z 0
1 ϕn0 (r2 )ϕ∗n0 (r20 )eiq·(r1 −r1 )
G+ (r1 , r2 ; r10 , r20 ) = ~2 q 2
d3 q (13.232)
(2π)3 E − En0 − + i0
n0 2m
X Z 0
2m 1 eiq·(r1 −r1 )
= ϕn0 (r2 )ϕ∗n0 (r20 ) ~2
d3 q . (13.233)
~2 (2π)3 (E − E n 0 ) − q 2 + i0
n0 2m

~2 k 2
Bearing in mind that E = 2m + En and introducing

02 2 ~2
knn 0 = k + (En − En0 ) (13.234)
2m
we cast (13.233) in the form

X Z 0
2m 1 eiq·(r1 −r1 ) 3
G+ (r1 , r2 ; r10 , r20 ) = ϕn0 (r2 )ϕ∗n0 (r20 ) 2 d q (13.235)
~ (2π)3 k 02 − q 2 + i0
n0
X 0
m eiknn0 |r1 −r1 |
0

=− ϕn0 (r2 )ϕ∗n0 (r20 ) , (13.236)


2π~2 |r1 − r10 |
n0

where I used (13.13) and (13.16). Thus we obtain the following equation
Z
ψn (r1 , r2 ) = ϕn (r2 )e ik·r1
+ G+ (r1 , r2 ; r10 , r20 )U (r10 , r20 )ψn (r10 , r20 )d3 r10 d3 r20 (13.237)
Z
m X
0 0
iknn0 |r1 −r1 |
e
= ϕn (r2 )eik·r1 − ϕn0 (r2 ) ϕ∗n0 (r20 ) U (r10 , r20 )ψn (r10 , r20 )d3 r10 d3 r20 . (13.238)
2π~2 0 |r1 − r10 |
n

This formula is called the Lippmann-Schwinger equation.

180
§11 Quasi-elastic scattering of electron by atom XIII ELASTIC SCATTERING

At large r we expand using (13.18)

X 0
eiknn0 r
ψn (r1 , r2 ) = ϕn (r2 )eik·r1 + fnn0 ϕn0 (r2 ) , r → ∞, (13.239)
r
n0

where
Z
m 0 0
fnn0 =− ϕ∗n0 (r20 )e−iknn0 ·r1 U (r10 , r20 )ψn (r10 , r20 )d3 r10 d3 r20 , (13.240)
2π~2
0 0 0
with knn 0 = knn0 r1 /r1 . In a particular case of elastic scattering n = n . The current density of scattered particles for

a given final state n0 of atom is


0
~knn 0
j= |fnn0 |2 . (13.241)
mr2
Since current density of incident electrons is v = ~k
m we find the cross section in a particular channel where atom
undergoes transition n → n0 :
0
knn 0
dσnn0 = |fnn0 |2 dΩ . (13.242)
k
The scattering amplitude is given by (13.240). In the Born approximation it reads
Z
m 0 0
B
fnn0 = − ϕ∗n0 (r20 )ei(k−knn0 )·r1 U (r10 , r20 )ϕn (r20 )d3 r10 d3 r20 , (13.243)
2π~2
This formula can be written in a more compact form if we introduce matrix elements of the potential between to
stationary states of atom:
Z
Unn0 (r1 ) = ϕ∗n0 (r20 )U (r10 , r20 )ϕn (r20 )d3 r20 .
0
(13.244)

Then
Z
m 0 0
B
fnn 0 = − Unn0 (r10 )ei(k−knn0 )·r1 d3 r10 . (13.245)
2π~2

∗ Examples.

1. Elastic scattering of electron on atom in the ground state n = 0. This process is described by the diagonal
matrix elements:
Z Z
m −iq·r
f00 (q) = − e U00 (r)d r , U00 (r) = |ϕn (r 0 )|2 U (r, r 0 )d3 r0 .
3
(13.246)
2π~2

To compute U00 (r) we actually do not have to take an integral of U (r, r 0 ). Instead we notice that U00 (r) is the
average potential electrostatic energy of atom. It must therefore satisfy the Poisson equation

∇2 U00 (r) = −4πeρ(r) , (13.247)

with the average electric charge density ρ given by

ρ(r) = Zeδ(r) − en(r) , (13.248)

where n(r) is the electron density in the ground state. Now, expanding in Fourier integral
Z Z
1 iq·r 3 1
U00 (r) = Uq e d q , ρ(r) = ρq eiq·r d3 q , (13.249)
(2π)3 (2π)3

181
§11 Quasi-elastic scattering of electron by atom XIII ELASTIC SCATTERING

and using (13.247) we find


4πeρq
Uq = . (13.250)
q2
Thus,
Z
4πe 4πe
Uq = ρ(r)e−iq·r d3 r = [Z − F (q)] , (13.251)
q2 q2
where
Z
F (q) = n(r)e−iq·r d3 r . (13.252)

is called the atomic form factor. It is Fourier image of its charge distribution. Since the ground state is
spherically symmetric we can easily integrate over angles to obtain
Z
4π ∞
F (q) = n(r)r sin(qr)dr . (13.253)
q 0
Since for elastic scattering q = 2k sin(θ/2) and k = k 0 , the scattering amplitude becomes
m m 4πe
f00 = − Uq = − [Z − F (q)] . (13.254)
2π~2 2π~2 q 2
Finally, the cross section
 2  2
2me2 me2
dσ = [Z − F (q)] dΩ = [Z − F (2k sin(θ/2))] dΩ . (13.255)
~2 q 2 2~2 k 2 sin2 (θ/2)

Of course, in order to calculate the form factor one needs to know the electron density n(r). Consider the
hydrogen atom Z = 1 in the ground state. The corresponding wave function and the charge density are

e−r/a e−2r/a
ϕ0 = √ 3/2 ⇒ n(r) = . (13.256)
πa πa3

The form factor is computed using (13.253) as follows


1
F (q) = h  i2 . (13.257)
qa 2
1+ 2

In particular, at small scattering angles qa = 2ka sin(θ/2)  1, F (q) ≈ 1 − 21 q 2 q 2 implying that


 ea 4
dσ ≈ m2 dΩ , qa  1 , (13.258)
~
which is independent of angles. In the opposite limit qa  1 (large angle scattering), we recover the Rutherford
formula
 4
m2 e
dσ = dΩ . (13.259)
4 ~k sin(θ/2)

2. Inelastic scattering of electron on atom in the ground state n = 0. In this case we are interested in off-diagonal
matrix elements of U0n0 (n0 > 0).
Z Z  2 
e Ze2
U0n (r1 ) = ϕ∗0 (r2 )U (r1 , r2 )ϕn (r2 )d3 r2 = ϕ∗0 (r2 ) − ϕn (r2 )d3 r2
r12 r1
Z
e2
= ϕ∗0 (r2 ) ϕn (r2 )d3 r2 . (13.260)
|r1 − r2 |

182
§11 Quasi-elastic scattering of electron by atom XIII ELASTIC SCATTERING

0
Denoting the momentum transfer by q = k − k0n we get for the scattering amplitude
Z Z Z
m iq·r1 3 m ∗ 3 e2
f0n = − U0n (r1 )e d r1 = − ϕ 0 (r2 )ϕn (r2 )d r2 eiq·r1 d3 r1 . (13.261)
2π~2 2π~2 |r1 − r2 |

Now taking the Fourier transform of the Poisson equation, similarly to (13.250),
Z
2 e2 2 e2 4πe2
∇1 = −4πe δ(r1 − r2 ) ⇒ eiq·r1 d3 r1 = 2 (13.262)
|r1 − r2 | |r1 − r2 | q
we get
Z
m 4πe2
f0n = − ϕ∗0 (r2 )ϕn (r2 )e−iq·r2 d3 r2 . (13.263)
2π~2 q 2

For example, to calculate scattering of an electron with excitation of a hydrogen atom from n = 0 to n = 1 we
need to compute the scattering amplitudes f0→200 , f0→210 , f0→211 , f0→21,−1 , where I explicitly indicated the
quantum numbers n, l, m of the final state of atom. The cross section then is
0
k01
dσ01 = (|f0→200 |2 + |f0→210 |2 + |f0→211 |2 + |f0→21,−1 |2 )dΩ . (13.264)
k

B. Exchange effects

Now consider the case of electron scattering off an atom. In this case we have to take into account a possibility
that an atomic electron escapes from atom and its place is taken by the incident electron. Namely, along with the
processes discussed above in which the incident electron (at r1 ) scatters at some angle θ while exciting atom n → n0 ,
there is another process, having the same initial and final states, in which incident electron is captured into the n0
state of atom, while atomic electron (at r2 ) is emitted at angle θ.
In the Born approximation the corresponding scattering amplitude can be obtained from (13.243) by swapping the
electron positions in the final state:
Z
˜ m 0
B
fnn0 = − 2
ϕ∗n0 (r1 )e−iknn0 ·r2 U (r2 , r1 )eik·r1 ϕn (r2 )d3 r1 d3 r2 , (13.265)
2π~
where
e2 Ze2
U (r2 , r1 ) = − . (13.266)
r12 r2
If the two electrons are in the singlet state with the total spin S = 0, then the corresponding coordinate wave function
is symmetric implying that the cross section is given by
0
knn
|fnn0 (θ) + f˜nn0 (θ)|2 dΩ .
0
dσ0 = (13.267)
k
If electrons in the triplet state with the total spin S = 1, their coordinate wave function is antisymmetric meaning
that
0
knn
|fnn0 (θ) − f˜nn0 (θ)|2 dΩ .
0
dσ1 = (13.268)
k

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §139,148, Messiah,
”Quantum Mechanics”, XIX, 8.

183
§12 Scattering matrix XIII ELASTIC SCATTERING

§12. Scattering matrix

The fundamental quantity of the scattering theory that encodes information about all possible scattering channels
is the scattering matrix. In this section we will define it and review some of its properties.
Suppose that the incident particles at t → −∞ are described by the wave function ϕa (plane waves) and follow the
time evolution of this system using the Schrödinger equation. As particles interact they scatter and at t → +∞ are
described by another wave function ψa . We define the scattering operator Ŝ such that
i 0
ψa = Ŝϕa = lim lim e− ~ Ĥ(t−t ) ϕa . (13.269)
t→∞ t0 →−∞

Ŝ transforms the initial wave function at t → −∞ to the final wave function at t → +∞. The wave function ψa
includes all possible processes – elastic and inelastic. We would like to know the probability to find the system in a
particular state ϕb after scattering. To this end we note that a set of eigenfunctions {ϕa } is complete and therefore
we can expand the final state as follows
X
ψa = ϕb hϕb |ψi . (13.270)
b

The probability to find the system in state ϕb at t → ∞ is

wab = |hϕb |ψa i|2 = |hϕb |Ŝ|ϕa i|2 . (13.271)

Matrix elements Sab = hϕb |Ŝ|ϕa i make up the scattering matrix.


The scattering operator is unitary, as is evident from its definition (13.269). It implies that the corresponding
matrix satisfies Ŝ † Ŝ = 1̂, or, equivalently
X

Sab Sbc = δac . (13.272)
b

In particular, for a = c unitarity yields the probability conservation


X
wab = 1 . (13.273)
b

Scattering matrix Sab is diagonal in those quantum numbers which correspond to the conserved quantities. This is
because Ŝ commutes with Ĥ and thus also with operators of conserved quantities. Moreover, Ŝ can depend only on
conserved quantities. Indeed, the transition probabilities are observable quantities, therefore Ŝ cannot depend on the
reference frame choice. For instance, we have seen that the elastic scattering matrix elements Sl are diagonal with
respect to the orbital momentum l (in central field), but do not depend on ml .
If no scattering happened, then the incident wave is the same as the scattered one, i.e. Sab = δab . This trivial
contribution is included in the scattering matrix but is obviously not very interesting. One therefore introduces a new
matrix Tab defined as follows

Sab = δab − 2πiTab δ(Eb − Ea ) . (13.274)

The delta function is explicitly written down because it comes about in any calculation of the scattering matrix due
to energy conservation (except when the potential depends on time). Factor −(2πi) is a convention. We can express
the scattering probabilities as follows

wab = |Sab |2 = [δab − 2πiTba δ(Eb − Ea )][δab + 2πiTba δ(Eb − Ea )] (13.275)
2 2 2
= δab − 2πiδab (Tba − Tba )δ(Eb − Ea ) + (2π) [δ(Eb − Ea )] |Tba | (13.276)

= δab + δab 2Im Tba + 2π|Tba |2 δ(Eb − Ea ) 2πδ(Eb − Ea ) . (13.277)

To make sense of the square of the delta function, recall that


Z +∞ Z +τ /2
1 i 1 i
δ(Eb − Ea ) = e ~ (Eb −Ea )t dt = lim e ~ (Eb −Ea )t dt , (13.278)
2π~ −∞ τ →∞ 2π~ −τ /2

184
§12 Scattering matrix XIII ELASTIC SCATTERING

The physical meaning of τ is the observation time. It is a big macroscopic quantity, but nevertheless finite. We write
therefore
Z
 2 2π +τ /2 i (E −Ea )t
w = δab + δab Im Tba + |Tba |2 δ(Eb − Ea ) e~ b dt (13.279)
~ ~ −τ /2
 2 2π
= δab + δab Im Tba + |Tba |2 δ(Eb − Ea ) τ . (13.280)
~ ~
Transition probability per unit time (i.e. transition rate) is
2 2π
ẇab = δab Im Tba + |Tba |2 δ(Eb − Ea ) . (13.281)
~ ~
The first term on the right-hand-side describes modification of the incident wave (owing to the δab ). This is what we
called the forward (θ = 0) elastsic scattering. The second term on the right-hand-side includes all other states (elastic
and inelastic). The scattering cross section for a 6= b can be computes as usual
2π 2
ẇab ~ |Tba | δ(Eb − Ea ) 2πma
σba = = = |Tba |2 δ(Eb − Ea ) , b 6= a . (13.282)
ja va ~2 ka
Since the final particles are free, their spectrum is continuous. Hence, we have to multiply (13.282) by the number of
states dνb in the interval dEb is (see (13.46))
mb ~kb
dνb = ρ(Eb )dEb = dΩdEb . (13.283)
(2π~)3
We have
2πma
dσba = |Tba |2 δ(Eb − Ea )dνb . (13.284)
~2 ka
Upon integration over Eb we finally derive
dσba ma mb kb
= |Tba |2 . (13.285)
dΩ (2π)2 ~4 ka
We see that the matrix elements Tba are closely related to the scattering amplitude fba . To find the exact relationship,
recall that
dσba kb
= |fba |2 . (13.286)
dΩ ka
Comparing now (13.286) and (13.285) we can identify

ma mb
fba = − Tba . (13.287)
2π~2
The negative sign is a convention. Thus, operator T̂ (and Ŝ) determines the scattering amplitude.
In fact, (13.281) contains more information. In view of probability conservation (13.273) we have
X 2X 2π X
ẇab = 0 = δab Im Tba + |Tba |2 δ(Eb − Ea ) , (13.288)
~ ~
b b b

which implies that


2π X 2
|Tba |2 δ(Eb − Ea ) = − Im Taa . (13.289)
~ ~
b

~ka
Dividing both sides by ja = va = ma and using (13.287) we get
X 4π
σ(a → anything) = σb = Im faa . (13.290)
ka
ab

185
§12 Scattering matrix XIII ELASTIC SCATTERING

This is the optical theorem that relates the total cross section (which includes all elastic and inelastic processes) to
the imaginary part of the forward elastic scattering amplitude. This relation is a direct consequence of unitarity. We
have derived the optical theorem before in (13.84), but there we disregarded all inelastic processes.
Scattering matrix has many other important properties that are not related to a specific scattering potential, see
references below for more details.

• Additional reading: Landau and Lifshitz, “Quantum Mechanics: Non-relativistic theory”, §125, Weinberg, “Lec-
ture nites on Quantum Mechanics”, 8.1-8.3.

186
§2 Klein-Gordon equation XIV RELATIVISTIC SCALAR PARTICLES

XIV. RELATIVISTIC SCALAR PARTICLES

§1. Uncertainty principle in relativistic theory

State of a particle in the non-relativistic Quantum Mechanics is described by a state vector that depends only on
quantum numbers of this particle, see Sec. I. Such states are called one-particle states. This description breaks down
when the particle velocity becomes comparable to the speed of light c. Indeed, we know from experiment that the
number of relativistic particles is not a conserved quantity. The particles can be created and destroyed. Therefore,
the notion of one-particle states is not applicable in the relativistic theory. If the particle velocity is not too close to c,
then the one-particle state description is approximately valid. Sometimes this is referred to as the mildly relativistic
regime. The main problem in the mildly relativistic regime is to compute the relativistic corrections to the one-particle
states.
In the Classical Mechanics, energy ε and momentum p of a free particle of mass m are related as
p
ε = c p 2 + m 2 c2 . (14.1)

When the particle is at rest p = 0, its energy is finite ε = mc2 . Consider a state with energy 2mc2 . Since
√ the number
of particles is not conserved, it can be realized either as a one-particle state with momentum p = 3mc or as a
two-particle state with the particles at rest. Because of this ambiguity, the relativistic Quantum Mechanics is not a
self-consistent theory as we will see later in this section. A consistent approach is the Quantum Filed Theory based
on the second quantization method that takes into account creation and annihilation of particles. Nevertheless, the
relativistic Quantum Mechanics is a useful tool in the mildly relativistic regime. In the above example, it corresponds
to energies ε  2mc2 .
The uncertainty principle has a number of important consequences for the relativistic particles. Consider a particle in
the mildly relativistic regime. The one-particle state description holds for momenta up to the the maximal momentum
of the order of mc. This means that the uncertainty of particle momentum in such a state is smaller than

(∆p)max ∼ mc , (14.2)

According to the uncertainty principle then, the particle position uncertainty is at least
~ ~
(∆x)min ∼ ∼ = λC . (14.3)
(∆p)max mc

This is knows as the Compton wavelength. Thus, the particle position cannot be localized better than λC . This is in
contrast to the non-relativistic theory were we can describe the particle position with any accuracy, provided that its
momentum is completely uncertain. The position uncertainty ∆x > λC translates into the time uncertainty of how
long it takes a light signal to propagate across ∆x:
∆x ~
∆t ∼ > . (14.4)
c mc2
Since both x and t are fundamentally uncertain, the coordinate wave function ψ(x, t) is not as useful quantity in the
relativistic theory as it is in the non-relativistic one.
On the other hand, uncertainty of momentum of a particle in a box of size L is ∆p = ~/L. By taking large enough
box ∆p can be made as small as desired. In other words, the particle momentum can be measured with any accuracy.
Thus, the momentum space wave function is more appropriate for description of the one-particle states in the mildly
relativistic regime.

• Additional reading: Berestetskiy, Lifshitz, Pitaevskiy, “Quantum Electrodynamics”, §1.


For this chapter see also Messiah, “Quantum Mechanics”, Ch. XX; Schiff, “Quantum mechanics”, Ch.13.

§2. Klein-Gordon equation

A relativistic equation for spin-0 particle can be derived in the same way as we derived the Schrödinger equation
in (XIV §2). Starting with the relativistic relation between energy and momentum (14.1) we make the replacements:

187
§2 Klein-Gordon equation XIV RELATIVISTIC SCALAR PARTICLES

ε → i~∂t , p → −i~∇ to obtain the Klein-Gordon equation

~2 ∂ 2 ψ
= (~2 ∇2 − m2 c2 )ψ . (14.5)
c2 ∂t2
This equation can be written in the relativistic from using the four-vector notation
ε 
pµ = ,p (14.6)
c
and the corresponding operator
 
µ µ ∂ i~ ∂
p̂ = i~∂ = i~ = , −i~∇ , (14.7)
∂xµ c ∂t

where xµ = (ct, r). From pµ pµ = ε2 /c2 − p2 , (14.5) and (14.7) we obtain the covariant form of the Klein-Gordon
equation

(p̂µ p̂µ − m2 c2 )ψ = 0 . (14.8)

Expression in brackets is Lorentz scalar, i.e. it does not change under the Lorentz transformations. Therefore, ψ is a
Lorentz scalar as well, and describes a state of a scalar particle.
Consider now a spatial inversion r → −r and the corresponding parity operator P̂ , i.e. P̂ ψ(r) = ψ(−r). Let its
eigenvalues be λ, viz. P̂ ψ(r) = λψ(r). They can be computed as follows

ψ(r) = P̂ 2 ψ(r) = λ2 ψ(r) ⇒ λ = ±1 . (14.9)

If the wave function does not change under the inversion, the corresponding particle is called simply a scalar particle,
if it changes sign, it is called a pseudoscalar particle.

 As in II §1 one can use the Klein-Gordon equation to compute the probability and probability current density.
Multiplying (14.5) by ψ ∗ and subtracting from the resulting equations its complex conjugated we find
 
~2 ∂ ∗ ∂ψ ∂ψ ∗
ψ −ψ = ~2 ∇(ψ∇ψ ∗ − ψ ∗ ∇ψ) . (14.10)
c2 ∂t ∂t ∂t

This is equivalent to the continuity equation (see (2.11))

∂ρ
+ ∇ · j = 0, (14.11)
∂t
if we identify
~
j= (ψ ∗ ∇ψ − ψ∇ψ ∗ ) , (14.12)
2mi  
i~ ∗ ∂ψ ∂ψ ∗
ρ= ψ −ψ . (14.13)
2mc2 ∂t ∂t

The constant factor in front of ρ is chosen to ensure the correct non-relativistic limit (see below). Notice that the
expression for the current density j (14.13) coincides with the non-relativistic limit (2.11). The continuity equation
(14.11) implies the conservation of the following “charge”(28 :
Z
dQ
= 0 , Q = ρ d3 r , (14.14)
dt
where the integral runs over the entire space and I used the Gauss’ theorem.

(28 Don’t confuse Q with the electric charge. In the non-relativistic limit Q is the total probability to find a particle anywhere in space.

188
§3 Non-relativistic limit of Klein-Gordon equation XIV RELATIVISTIC SCALAR PARTICLES

The two equations (14.12) and (14.13) can be written as a components of the four-current defines as follows
 
µ ~ ∗ ∂ψ ∂ψ ∗
j = ψ −ψ . (14.15)
2mi ∂xµ ∂xµ
The continuity equation in the manifestly covariant form reads
∂ µ jµ = 0 . (14.16)

The Klein-Gordon equation (14.5) is of the second order in time-derivatives. Thus, in order to solve it one needs to
provide two initial conditions: one for the function ψ itself and another one for its time derivative ∂t ψ. Any pair of
initial conditions describes a certain state of spin-0 particle. In particular, one can choose the initial conditions such
that ρ < 0 owing to the relative minus sign in (14.13). This means that ρ is not a positive-definite unction. Therefore,
it cannot be interpreted as a probability density, as in the non-relativistic case. This is because the one-particle states
interpretation of the wave function is not consistent anymore in the relativistic region as explained in XIV §1. In
XIV §5 we discuss how this difficulty can be resolved in the second quantization framework.

• Additional reading: T. Ohlsson “Relativistic Quantum Physics”, 2.0, 2.1.

§3. Non-relativistic limit of Klein-Gordon equation

It is important to establish the non-relativistic limit of the Klein-Gordon equation. First, expand (14.1) at p  mc:
p2 p4
ε ≈ mc2 +− 2 3 + ... (14.17)
2m 8c m
Thus the quantity E that we called energy in the non-relativistic limit is related to ε appearing in the relativistic
formulas by
E = ε − mc2 . (14.18)
Therefore, the wave function ϕ satisfying the Schrödinger equation is related to ψ satisfying the Klein-Gordon equation
as follows
imc2
ψ(r, t) = ϕ(r, t)e− ~ t
. (14.19)
The Klein-Gordon equation for ϕ can be derived by calculating the following time derivatives
 
∂ψ ∂ϕ imc2 imc2
= − ϕ e− ~ t , (14.20)
∂t ∂t ~
"  2 #
∂2ψ ∂ 2 ϕ 2imc2 ∂ϕ imc2 imc2

2
= 2
− − ϕ e− ~ t , (14.21)
∂t ∂t ~ ∂t ~

In the non-relativistic limit



∂ϕ
i~ ∼ Eϕ  mc2 |ϕ| . (14.22)
∂t

Therefore, neglecting the second time derivative in (14.21), but keeping the first one and substituting into (14.5) we
get
∂ϕ ~2 2
i~ =− ∇ ϕ. (14.23)
∂t 2m
Thus, the Klein-Gordon equation indeed reduced to the Schrödinger equation for the function ϕ. The non-relativistic
limit of (14.12) and (14.13) follows from (14.19),(14.20),(14.21):
~
j= (ϕ∗ ∇ϕ − ϕ∇ϕ∗ ) , (14.24)
2mi
ρ = ϕ∗ ϕ . (14.25)

• Additional reading: T. Ohlsson “Relativistic Quantum Physics”, 2.2.

189
§4 Scalar particle states with definite momentum XIV RELATIVISTIC SCALAR PARTICLES

§4. Scalar particle states with definite momentum

Consider motion of a scalar particle with a definite momentum. The derivation of the Klein-Gordon equation (14.5)
suggests that a solution can be found in the form of the plane wave:
i
ψp = Ap e ~ (p·r−εt) . (14.26)

Substituting this into (14.5) we find a relationship between ε and p:


p
ε = ±εp , with εp = p2 + m2 c2 . (14.27)

Thus, unlike the non-relativistic case, there are two solutions


i
ψp,± = Ap e ~ (p·r∓εp t) , (14.28)

called the positive/negative energy solutions depending on the sign of ε. Plugging (14.28) into (14.13) we find that
the probability density is
εp
ρ± = ± |ψp,± |2 . (14.29)
mc2
We see that the negative energy solutions are the ones that cause the probability density to be not positive definite.
In the next section we will give a physical interpretation of these solutions.
In the non-relativistic limit, we normalized the wave functions of the continuous spectrum, such as (14.26), by
requiring that
Z
ψp∗ 0 ,± (r)ψp,± d3 r = δ(p − p0 ) . (14.30)
R
This
R normalization is meaningful because the probability conservation |ψ|2 d3 r = 1 requires that the integral
|ψ|2 d3 r be a conserved quantity. It simply means that there is a unit probability to find a particle described
by ψ somewhere in space. In the relativistic case the corresponding conserved quantity is
Z Z  
i~ ∗ ∂ψ ∂ψ ∗
Q = ρ d3 r = ψ − ψ d3 r , (14.31)
2mc2 ∂t ∂t

where the integral goes over the entire space, see (14.14). Thus, the correct normalization in the relativistic case is
Z  
i~ ∗ ∂ψp,± ∂ψp∗ 0 ,± (r)
ψ 0
p ,± (r) − ψ p,± d3 r = ±δ(p − p0 ) , (14.32)
2mc2 ∂t ∂t
and
Z  
i~ ∂ψp,± ∂ψp∗ 0 ,± (r)
ψp∗ 0 ,∓ (r) − ψp,∓ d3 r = ±δ(p − p0 ) , (14.33)
2mc2 ∂t ∂t

Using these normalization we can find the normalization constant Ap appearing in (14.28):
s
1 mc2
Ap = . (14.34)
(2π~)3/2 εp

Thus, the wave functions of states of a scalar particle with definite momentum are
s
1 mc2 ± i (p·r−εp t)
ψp,± = e ~ . (14.35)
(2π~)3/2 εp

We can use the solutions (14.35) to build wave packets of positive or negative energy solutions:
Z
ψ± (r, t) = ap,± ψp,± (r, t)d3 p , (14.36)

190
§5 Second quantization of scalar particles XIV RELATIVISTIC SCALAR PARTICLES

where ap,± is the amplitude of a wave with given momentum p. Sometimes it is convenient to consider motion in a
big but finite volume. In the case of motion in a box of side L the wave packets take form
1 X 2π~
Ψ± (r, t) = ap,± ψp,± , p= n, nx , ny , nz = 0, ±1, ±2, . . . (14.37)
L3/2 p
L

Using (14.36) and (14.32) we can express the “charge” Q in terms of ap,±
Z  ∗  Z
i~ ∗ ∂ψ± ∂ψ± 3
Q= ψ ± − ψ ± d r = |ap,± |2 d3 p (14.38)
2mc2 ∂t ∂t
R
We can interpret ap,± as the wave function in the momentum representation normalized by |ap,± |2 d3 p = 1. We
observe that while the wave function in the coordinate representation is not well-defined for a relativistic particle
(because the integrand in the left-hand-side of (14.38) is not a positive definite function), in momentum representation
it is a well-defined quantity in agreement with our discussion in XIV §1.

∗ Example. Consider the position operator r̂. In momentum space it reads r̂ = i~∇p . Its eigenfunction describing
a state of a particle at r = r0 reads
1 i
ar0 (p) = 3/2
e− ~ p·r0 . (14.39)
(2π~)
It is normalized as we discussed after (14.38). Let’s compute this wave function in the coordinate representation using
(14.36):
Z s
1 − ~i p·r0 1 mc2 i p·r 3
ψr0 (r) = 3/2
e 3/2
e~ d p . (14.40)
(2π~) (2π~) εp

In spherical coordinates integration over the angles is easy, while the remaining radial integral can be expressed in
terms of the modified Bessel function. The result is
 5/4  
(2π 2 )1/4 1 λC |r − r0 |
ψr0 (r, t) = K5/4 . (14.41)
2π 2 Γ(1/4) λ3C |r − r0 | λC

At |r − r0 |  λC it behaves like
1 mc
ψr0 (r, t) ∼ 7/4
e− ~ |r−r0 | . (14.42)
|r − r0 |
We observe that in contrast to the non-relativistic approximation ψr0 (r) 6= δ(r − r0 ) that would indicate the precise
knowledge of the particle position. Due to the square root in the normalization factor (14.34) the relativistic wave
function is smeared around r0 in agreement with our discussion in XIV §1.

§5. Second quantization of scalar particles

For a consistent description of relativistic systems one has to take into account all possible multi-particle states
compatible with the conservation laws. The most convenient way of doing this is to employ the second quantization
method and, in particular, the field operators ψ̂, see XI §5. However, unlike in XI §5 now we have to deal with the
negative-energy solutions, which together with the positive-energy solutions, comprise the complete basis. Eq. (11.80)
has to modified to reflect this fact. To this end, we start with the expansion of a general solution to the Klein-Gordon
equation in plane waves (in a big box):
X X
ψ(r, t) = ap,+ ψp,+ + ap,− ψp,− . (14.43)
p p

(To simplify the notation I set L = 1, i.e. consider a unit volume). In the second quantization method ψ becomes the
ψ̂-operator and coefficients ap,+ in the first sum become the annihilation operators âp as in (11.80). The second sum
runs over the negative-energy solutions.

191
§6 Scalar particle in electromagnetic field XIV RELATIVISTIC SCALAR PARTICLES

Recall that in the Heisenberg picture, the matrix elements of operators between the initial i and final f states
i
contain the time-dependent factor e− ~ (Ei −Ef )t . Therefore, if we compute the matrix elements of ψ̂, the negative
energy solutions will correspond to particles in the final states (in contrast to the positive energy solutions that
correspond to particles in the initial state). We conclude that coefficients ap,− should be replaced by the creation
operators b̂†−p of some other particles, called antiparticles. Changing the summation variable in the second sum
i
p → −p (in order that the plane wave take the standard form e− ~ (p·r−εp t) ), we can write (29
s s
X mc2 X mc2
−ip·x/~ † ip·x/~ †
ψ̂ = (âp e + b̂p e ) , ψ̂ = (â†p eip·x/~ + b̂p e−ip·x/~ ) . (14.44)
p
ε p p
εp

This is known as the Feynman-Stückelberg interpretation and is the basis for the modern Quantum Field Theory.
Operator ψ̂(r, t) describing satisfying the Klein-Gordon equation is called the scalar field and describes spin-zero
bosons. According to (11.26) creation and annihilation operators satisfy the commutation relations (for a particle in
a box)
[âp , â†p0 ] = δp,p0 , [b̂p , b̂†p0 ] = δp,p0 , (14.45)
while all other commutators vanish. These commutation relation in the infinite volume are
[âp , â†p0 ] = δ(p − p0 ) , [b̂p , b̂†p0 ] = δ(p − p0 ) (14.46)
Substituting (14.44) into (14.31) and using (14.45) we find that
X X
Q̂ = (â†p âp − b̂p b̂†p ) = (â†p âp − b̂†p b̂p − 1) . (14.47)
p p

Apart from the infinite additive constant the expectation value of this operator in the Fock space is
X
Q= (Np − N̄p ) , (14.48)
p

where Np and N̄p are numbers of particles and antiparticles. We thus resolved the problem with non-positivity of ρ
in (14.13): it describes the difference of particle and antiparticle densities, which should not be necessarily positive.
Owing to the relativistic invariance, particles and antiparticles are described as quantum excitations of the same field.
In contrast, in the non-relativistic case, particles decouple from antiparticles; each is described by its own Schrödinger
equation.
If particles are charged, then Q determines the total electric charge of the system in units of e. In this case,
as is evident from (14.48), particles and antiparticles carry opposite charges. The continuity equation signifies the
conservation of electric charge. Particles can carry other charges besides the electric charge or can be neutral, i.e.
carry no charge.
Thus far we discussed free particles. If particles interact, then the charge Q is conserved if [Q̂, Ĥ] = 0. For example,
electric charge is conserved in all known interactions. This implies that particles and antiparticles can be created or
annihilated only in particle-antiparticle pairs.

• Additional reading: Berestetskiy, Lifshitz, Pitaevskiy, “Quantum Electrodynamics”, §11.

§6. Scalar particle in electromagnetic field

Interaction of a charged non-relativistic particle with the electromagnetic field has been discussed in in (VIII §1).
The relativistic generalization is straightforward and follows from the correspondence of the quantum and classical
versions of the canonical formalism(30 . Introduce the four-potential
Aµ = (Φ, A) , (14.49)

(29 We can say that the negative energy solutions are interpreted as negative energy particles which propagate backward in time. Physically,
i i
solutions correspond to positive energy anti-particles which propagate forward in time: e− ~ εt = e− ~ (−ε)(−t) .
(30 You may consult my E&M lecture notes for details.

192
§6 Scalar particle in electromagnetic field XIV RELATIVISTIC SCALAR PARTICLES

and make a replacement


q q
p̂µ → p̂µ − µ = i~∂ µ − µ . (14.50)
c c
Eq. (14.8) reads
h q i
(p̂µ − Aµ )2 − m2 c2 ψ = 0 . (14.51)
c
Using the three-dimensional vectors this is equivalent to
"  2  #
1 ∂ q 2 2 2
i~ − qΦ − p̂ − A − m c ψ = 0 . (14.52)
c2 ∂t c

Repeating the by now familiar procedure we derive (see e.g. (14.10)–(14.13))


~ qA ∗
j= (ψ ∗ ∇ψ − ψ∇ψ ∗ ) − ψ ψ, (14.53)
2mi   mc
i~ ∗ ∂ψ ∂ψ ∗ qΦ ∗
ρ= ψ − ψ − ψ ψ. (14.54)
2mc2 ∂t ∂t mc2
The last terms in (14.53) and (14.54) are contributions of the electromagnetic field to the current and charge density.
In a stationary state ψ ∼ e−iεt/~
ε − qΦ 2
ρ= |ψ| . (14.55)
mc2
The one-particle interpretation is valid as long as ε  |qΦ|. In a very strong fields such that |qΦ| > ε density of
particle changes sign implying the breakdown of the one-particle interpretation due to the pair production. In such
a case one has to employ the second quantization formalism that includes antiparticles.

 Eq. (14.51) is meaningful only if it is gauge-invariant, i.e. it does not change its form when the potential is
transformed as
Aµ → A0µ = Aµ + ∂ µ f , (14.56)
where f is an arbitrary function, and simultaneously the wave function is transformed as
ψ → ψ 0 = eiG ψ , (14.57)
ie
where G depends on f . Eq. (14.56) is equivalent to (8.10). The requirement of gauge invariance fixes G = − ~c f (see
µ
your home assignment). The freedom to choose function f implies that the potential A can be restricted by one
arbitrary condition know as a gauge. For example, the Lorentz gauge is a condition such that
1 ∂Φ
∂ µ Aµ = 0 ⇒ + ∇ · A = 0. (14.58)
c ∂t

 To derive the non-relativistic limit of (14.52) we replace the wave function ψ by ϕ according to (14.19) and expand
assuming that

∂ϕ
i~ , |qΦϕ|  mc2 |ϕ| , (14.59)
∂t

which extends (14.22) to include a requirement that energy of particle in electric field be small compared to mc2
(magnetic field does not do work). The derivation is similar to (14.19) – (14.23) and is a subject to one of the home
assignments. The final result is
 2 
∂ϕ p̂ q q2 2
i~ = − A · p̂ + A + qΦ ϕ. (14.60)
∂t 2m mc 2mc2
This equation describes motion of a mildly relativistic spin-zero particle in the electromagnetic field.

• Additional reading: T. Ohlsson “Relativistic Quantum Physics”, 2.4.

193
§7 Scalar particle in Coulomb field XIV RELATIVISTIC SCALAR PARTICLES

§7. Scalar particle in Coulomb field

Stationary states of a scalar particle in the Coulomb field (the so-called pionic atom) are solutions of (14.52) with
i~∂t → ε and

Ze2
qΦ = − , A = 0. (14.61)
r
Namely,
" 2 #
Ze2 2 4 2 2 2
ε+ −m c +~ c ∇ ψ(r) = 0 . (14.62)
r

Since the potential is spherically symmetric, the angular depends separates in the following way:
1
ψ(r) = χl (r)Ylm (r) . (14.63)
r
Equation for χ reads
 
d2 l(l + 1) − (αZ)2 2αZε m2 c4 − ε2
− + − χl (r) = 0 . (14.64)
dr2 r2 ~cr ~2 c2

Here α = e2 /~c is the fine structure constant. From the mathematical standpoint this equation is the same as the
non-relativistic equation (5.50). Denoting

4(m2 c4 − ε2 ) 2αZε
β2 = , ρ = βr , λ= , (14.65)
~2 c2 ~cβ
we obtain
 
d2 λ l(l + 1) − (αZ)2 1
+ − − χl (ρ) = 0 . (14.66)
dρ2 ρ ρ2 4

Note that λ > 0 because ε > 0. A solution to this equation is


1
χl (ρ) = ρs+1 F (−λ + s + 1, 2s + 2, ρ)e− 2 ρ , (14.67)

where F is the confluent hypergeometric function and

s(s + 1) = l(l + 1) − (αZ)2 ⇒ (s + 1/2)2 = (l + 1/2)2 − (αZ)2 . (14.68)

In order that χl → 0 at ρ → ∞ we need that

λ − s − 1 = nr = 0, 1, 2 . . . (14.69)

Now, solving (14.68) we obtain


1 p
s = − ± (l + 1/2)2 − (αZ)2 . (14.70)
2
As noted above λ > 0 implying that only the positive sign in (14.70) is physical. Thus, we derive
1 p
λ = nr + s + 1 = nr + + (l + 1/2)2 − (αZ)2 . (14.71)
2
From (14.65) we find energy spectrum

mc2
ε= p . (14.72)
1 + (αZ/λ)2

194
§7 Scalar particle in Coulomb field XIV RELATIVISTIC SCALAR PARTICLES

Consider the weak field limit αZ  1. Expanding λ in powers of (αZ), substituting into (14.72) and expanding
again we derive
   
(αZ)2 mc2 (αZ)4 n 3
ε = E + mc2 = mc2 1 − − − + . . . , (14.73)
2n2 2n4 l + 1/2 4
where we denoted

n = nr + l + 1 . (14.74)

Clearly, n is the principal quantum number and nr is the radial quantum number. The second term in the curly
brackets represents the non-relativistic Coulomb spectrum
(αZ)2
En(0) = −mc2 . (14.75)
2n2
As expected the spectrum is degenerate with respect to the quantum numbers l and m. The first relativistic correction
is given by the third term:
 
(1) (αZ)4 n 3
Enl = −mc2 − . (14.76)
2n4 l + 1/2 4
It explicitly depends on l lifting the “accidental” degeneracy and splitting the levels with the same n but different l
(know as the fine structure). The relative splitting of s and p levels is

Enp − Ens 4(αZ)2


(0)
=− . (14.77)
En 3n

It decreases in the inverse proportion to n. At n = 1, only one value l = 0 is allowed, hence there is no splitting. At
n = 2, there is splitting with l = 0 or l = 1 which is the largest splitting.
It follows from (14.71) and (14.72) that in strong fields αZ > 21 , energy ε becomes complex indicating instability
of such a heavy atom. Taking into account the finite size of atomic nucleus one finds that the levels are unstable at
Zcr ≈ 170.

• Additional reading: T. Ohlsson, “Relativistic Quantum Physics”, 2.4,2.6; W. Greiner “Relativistic Quantum
Mechanics”, 1.9.

§8. Tunneling through potential barrier

U (x)

U0

FIG. 30:

Scattering of relativistic particles on potential barriers has a number of important features that distinguish it from
the non-relativistic scattering. As an illustration, we will consider a simple one dimensional step potential shown in
Fig. 30. Using qΦ = U (x) and A = 0 in (14.52) and considering stationary states with energy ε we write
 2

2 2 4 2 2 d
(ε − U0 ) − m c + ~ c ψ2 = 0 , x > 0 , (14.78)
dx2
 
d2
ε2 − m2 c4 + ~2 c2 2 ψ1 = 0 , x < 0 . (14.79)
dx

195
§8 Tunneling through potential barrier XIV RELATIVISTIC SCALAR PARTICLES

0
Solutions with definite momentum have form ψ2 ∼ e±ik x at x > 0 and ψ1 ∼ e±ikx at x < 0. Since these wave
functions must satisfy the Klein-Gordon equation (14.78),(14.79) we find that
1p 1p
k0 = (ε − U0 )2 − m2 c4 = (ε − U0 − mc2 )(ε − U0 + mc2 ) , (14.80)
~c ~c
1p 2
k= ε − m 2 c4 . (14.81)
~c
Suppose that a beam of particles is incident from x = −∞. In this case, the wave function reads

ψ = ψ1 = eikx + Be−ikx , x < 0, (14.82)


ik0 x
ψ = ψ2 = Ce , x > 0. (14.83)

Continuity of the logarithmic derivative ψ 0 /ψ at x = 0 implies that

k − k0 2k
B= , C= . (14.84)
k + k0 k + k0
Coefficients B and C seems to be the same as in the non-relativistic case. However, they are not, because the
relationships between k and ε and k 0 and ε (often called the ‘dispersion relations’) are relativistic.
The goal of the scattering theory is to compute the ratio between the outgoing and incoming fluxes. The incident
current density is, using (14.12),

~ ~k
j= (e−ikx ∂x eikx − eikx ∂x e−ikx ) = . (14.85)
2mi m
2
Similarly, current density of the reflected flux is m |B| . We have to be more cautious with the transmitted flux
~k

because k 0 can become imaginary. Thus,

~|C|2 0 −ik0∗ x ik0 x 0 0∗ ~|C|2 0 0 0∗


j0 = (ik e e + ik 0∗ eik x e−ik x ) = (k + k 0∗ )ei(k −k )x . (14.86)
2mi 2m
The transmission and reflection coefficients are
j0 Re k 0 0
T = = |C|2 e−2Im k x , R = |B|2 , (14.87)
j k

where B and C are given by (14.84). You can verify that R + T = 1.


There are three qualitatively different cases depending on the particle energy ε and the potential barrier height U0 :

1. If ε > U0 + mc2 , then k and k 0 are real and positive. According to (1.15), the group velocity of the transmitted
particles is

dω dε 1 ~c2 k 0
u0 = = = = > 0. (14.88)
dk 0 ~dk 0 ~dk 0 /dε ε − U0

The transmission and reflection coefficients are


 2
k 0 4k 2 4kk 0 k − k0
T = = , R= . (14.89)
k (k + k 0 )2 (k + k 0 )2 k + k0

We observe that R < 1 and T < 1, i.e. some particles are reflected and some are transmitted. This case closely
resembles the non-relativistic one albeit with different k and k 0 .
2. If U0 − mc2 < ε < U0 + mc2 , then k 0 is imaginary, see (14.80). Eq. (14.87) implies that T = 0. Denote k 0 = iκ,
where κ is real. Then,

(k − iκ)(k + iκ)
R= = 1. (14.90)
(k + iκ)(k − iκ)

This is the case of the total internal reflection. No particles are transmitted at all.

196
§8 Tunneling through potential barrier XIV RELATIVISTIC SCALAR PARTICLES

3. If ε < U0 − mc2 (very strong field), k 0 is real. Since ε < U0 , the group velocity computed in (14.88) is positive
only if k 0 < 0. In other words, the particles move in the positive x-direction only if k 0 < 0. (Out of two
0 0
solutions to the Klein-Gordon equation at x > 0 e±ik x we must choose e−ik x ). The transmission and reflection
coefficients are
 2
4k|k 0 | k + |k 0 |
T =− < 0 , R = > 1. (14.91)
(k − |k 0 |)2 k − |k 0 |

Still T + R = 1, of course. Thus more particles are reflected that are incident at the barrier! This strange result
is known as the Klein paradox. It appears because the one-particle state description breaks down in strong fields
as has been repeatedly mentioned before. It is resolved when ψ is upgraded to quantum field, that contains
negatively charged antiparticles. Strong field can produce particle-antiparticle pairs with antiparticles moving
in the negative x-direction.
The Klein paradox appears also in the case of spin-1/2 particles.

• Additional reading: T. Ohlsson, “Relativistic Quantum Physics”, 2.5.

197
§1 Dirac equation XV RELATIVISTIC SPIN-1/2 PARTICLES

XV. RELATIVISTIC SPIN-1/2 PARTICLES

§1. Dirac equation

Before it was realized that a consistent theory of relativistic Quantum Mechanics requirers promotion the wave
functions to quantized fields fields, the Klein-Gordon equation had been considered unphysical because particle density
derived from it is not positive-definite, see (14.13). In an attempt to find an equation that would lead to a positive ρ
Dirac derived his famous equation describing a particle of spin-1/2 and its antiparticle. In this section we are going
to follow his arguments (slightly modified to better fit in the course).
A non-relativistic spin-1/2 particle is described by a two-component spinor. Since we already know that the
relativistic wave function must also include antiparticles, it must contain at least four components(31 . Such four-
component wave functions are called bispinors and can be represented in a matrix form (similar to usual spinors):
 
ψ1 (r, t)
 
 ψ (r, t) 
Ψ(r, t) =  2 . (15.1)
 ψ3 (r, t) 
ψ4 (r, t)
We can also introduce its Hermitian conjugate
 
Ψ† (r, t) = ψ1∗ (r, t) ψ2∗ (r, t) ψ3∗ (r, t) ψ4∗ (r, t) . (15.2)

We noticed in XIV §2 that ρ of scalar particles is not positive-definite because along with the wave function it also
contains its time derivative and each has independent initial condition (since the Klein-Gordon equation is of the
second order in time derivatives). Thus, in order that the density be positive-definite it must have form
4
X
ρ = Ψ† Ψ = ψn∗ ψn (15.3)
n=1

and Ψ must satisfy an equation that contains only its first time-derivative. Such an equation must also be linear in
space derivatives because otherwise it cannot be Lorentz-invariant. The most general equation for Ψ can be written
as follows
3 4 4
1 ∂ψn X X (i) ∂ψl mc X
+ αnl +i βnl ψl = 0 . (15.4)
c ∂t i=1
∂xi ~
l=1 l=1
(i)
Here αnl and βnl are complex constants, index l enumerates components of bispinor, index i enumerates space
coordinates xi and all other constants m, c, ~ are inserted for future convenience. Eq. (15.4) can be compactly
(i)
written using the matrix notation: let elements of the four 4 × 4 matrices α̂i and β̂ be αnl and βnl respectively.
Moreover, we will denote the three matrices α̂ as a three-vector α̂. Then (15.4) reads
1 ∂Ψ mc
+ α̂ · ∇Ψ + i β̂Ψ = 0 . (15.5)
c ∂t ~
To determine the unknown matrices we impose two conditions: probability conservation and relativistic invariance.
Probability conservation means that
1 ∂ρ
+ ∇ · j = 0, (15.6)
c ∂t
where j is the current density to be determined. From (15.5) we obtain
1 † ∂Ψ mc †
Ψ + Ψ† α̂ · ∇Ψ + i Ψ β̂Ψ = 0 , (15.7)
c ∂t ~
1 ∂Ψ† mc † †
Ψ + ∇Ψ† · α̂† Ψ − i Ψ β̂ Ψ = 0 . (15.8)
c ∂t ~

(31 This is not an obvious statement, see e.g. Berestetskiy, Lifshitz, Pitaevskiy, “Quantum Electrodynamics” §17.

198
§1 Dirac equation XV RELATIVISTIC SPIN-1/2 PARTICLES

Adding (15.7) and (15.8) and using (15.3) we have


1 ∂ρ mc †
+ Ψ† α̂ · ∇Ψ + ∇Ψ† · α̂† Ψ + i (Ψ β̂Ψ − Ψ† β̂ † Ψ) = 0 . (15.9)
c ∂t ~
In order that (15.6) be satisfied, we must require that

α̂ = α̂† , β̂ = β̂ † . (15.10)

In other words, matrices α̂ and β̂ are Hermitian. The current density is given by
j = cΨ† α̂Ψ . (15.11)

Now we need to ensure that a particle described by (15.5) with (15.10) has relativistic relationship between its
energy and momentum: ε2 /c2 = p2 + m2 c2 . This is the case if Ψ satisfies the following second order equation

~2 ∂ 2 Ψ
− ~2 ∇2 Ψ + m2 c2 Ψ = 0 . (15.12)
c2 ∂t2
To obtain a second order equation of this form consider
  
1 ∂ mc 1 ∂ mc
− α̂ · ∇ − i β̂ + α̂ · ∇ + i β̂ Ψ = 0 . (15.13)
c ∂t ~ c ∂t ~

Opening the brackets, remembering that α̂ and β̂ are matrices (so their order is significant), and noting that the
operator ∂t commutes with all other operators we write
 
1 ∂2 mc m2 c2 2
− (α̂ · ∇)(α̂ · ∇) − i (α̂ · ∇β̂ + β̂ α̂ · ∇) + 2 β̂ Ψ = 0 . (15.14)
c2 ∂t2 ~ ~
Let us look more closely at the following term
X 1X
(α̂ · ∇)(α̂ · ∇)Ψ = α̂i α̂k ∂i ∂k Ψ = [(α̂i α̂k + α̂k α̂i ) + (α̂i α̂k − α̂k α̂i )]∂i ∂k Ψ
2 | {z } | {z }
ik ik
sym antisym
1X 1X
= (α̂i α̂k + α̂k α̂i )∂i ∂k Ψ = {α̂i , α̂k }∂i ∂k Ψ . (15.15)
2 2
ik ik

I used the anti-commutator sign {. . .}. We see now that in order that (15.14) reduce to (15.12) we have to require
that
β̂ 2 = 1 , α̂k β̂ + β̂ α̂k = 0 , α̂k α̂i + α̂i α̂k = 2δik , (15.16)

where 1 is a unit 4 × 4 matrix. Equation (15.5) with matrices α̂ and β̂ satisfying conditions (15.10) and (15.16) is
called the Dirac equation.
There is not unique choice of matrices α̂ and β̂ satisfying (15.10) and (15.16). In the standard representation these
matrices are choses as follows:
! !
0̂ σ̂ 1̂ 0̂
α̂ = , β̂ = , (15.17)
σ̂ 0̂ 0̂ −1̂

where 0̂, 1̂ and σ̂ denote zero, unit and Pauli 2 × 2 matrices. In other representation of α̂ and β̂ can be obtained by
a unitarity transformation

α̂0 = Û α̂Û −1 , β̂ 0 = Û β̂ Û −1 , (15.18)

where Û is a unitary matrix. The physical properties of the Dirac equation are of course independent of a represen-
tation.
From now on I will drop the hat sign above matrices α̂, β̂, σ but will keep the hat sign over operators.

• Additional reading:

199
§3 Covariant form of the Dirac equation XV RELATIVISTIC SPIN-1/2 PARTICLES

§2. Spin and the Dirac equation

Total angular momentum of a particle Jˆ = L̂+ Ŝ must be conserved. In this section we discuss how this conservation
law is realized for a particle satisfying the Dirac equation. It is convenient to cast the Dirac equation (15.5) in the
Hamiltonian form:
∂Ψ
i~ = ĤD Ψ , (15.19)
∂t
where

ĤD = cα · p̂ + mc2 β . (15.20)

The physical quantities whose operators commute with the Hamiltonian are conserved. Consider first the orbital
angular momentum

:0
[ĤD , L̂] = [cα · p̂, r̂ × p̂] + mc2

[β, ×
r̂ p̂] = cαl ijk [p̂l , x̂j p̂k ]ei
= cαl ijk [p̂l , x̂j ]p̂k ei = −i~cαj ijk p̂k ei = −i~c(α × p̂) (15.21)

Thus, L̂ is not conserved. However, since we know that Jˆ is conserved, we conclude that the (yet unknown) spin
operator has the following commutation relation with the Hamiltonian:

[ĤD , Ŝ] = i~c(α × p̂) . (15.22)

Let me show that that the required operator is given by


!
~ σ 0
Ŝ = Σ̂ , Σ̂ = . (15.23)
2 0 σ

Indeed, noting that


! ! ! ! !
0 σi σj 0 σj 0 0 σi 0 [σi , σj ]
[αi , Σj ] = − = = 2iijk αk . (15.24)
σi 0 0 σj 0 σj σi 0 [σi , σj ] 0

we derive

[ĤD , Σj ] = c[α · p̂, Σ̂j ] = cp̂i [αi , Σ̂j ] = 2ciijk p̂i αk = −2ic(p̂ × α)j , (15.25)

which implies (15.22). Note, that operator L̂ acts on the spatial coordinates r of the bispinor whereas operator Ŝ,
being a 4 × 4 matrix, acts on its spin components. Yet because of the special form of the Hamiltonian ĤD their sum
is conserved. This is a consequence of a non-trivial way that bispinors transform under the Lorentz transformations.
We can easily compute the eigenvalues of Ŝ 2 and Ŝz operators:
~2 2 3
Ŝ 2 = (Σ + Σ2y + Σ2z ) = ~2 1 . (15.26)
4 x 4
Since Ŝ 2 |s, ms i = ~2 s(s+1)|s, ms i, (15.26) implies that s = 1/2. Thus the Dirac equation describes spin-1/2 particles.

• Additional reading: M. Thomson “Modern Particle Physics”, 4.4.

§3. Covariant form of the Dirac equation

To write the Dirac equation in a manifestly covariant form recall the definition of the four momentum (14.7) and
introduce the following Dirac gamma-matrices:
! !
0 σ 0 1 0
γ = βα = , γ =β= . (15.27)
−σ 0 0 −1

200
§4 States of spin-1/2 particle with definite momentum XV RELATIVISTIC SPIN-1/2 PARTICLES

These matrices form a four-vector γ µ = (γ0 , γ). Indeed we can write the Dirac equation
 
i~ ∂
+ i~α · ∇ − mcβ Ψ = 0 (15.28)
c ∂t
as

(p̂0 − α · p̂ − mcβ)Ψ = 0 . (15.29)

Multiplying by matrix β on the left

(β p̂0 − βα · p̂ − mcβ 2 )Ψ = 0 (15.30)

and using (15.16) we write obtain the covariant form of the Dirac equation

(γ µ p̂µ − mc)Ψ = 0 . (15.31)

Clearly γ µ is a four-vector.
Some properties of γ-matrices:

γ µ γ ν + γ ν γ µ = 2g µν 1 , (15.32)
† 0† 0
γ = −γ , γ =γ , (15.33)

where g µν is the metric tensor


 
1 0 0 0
 
0 −1 0 0 
g µν = . (15.34)
0 0 −1 0 
0 0 0 −1

Consider the Hermitean conjugate of the Dirac equation (15.31):

Ψ† (γ 0† p̂0∗ − γ † · p̂∗ − mc) = 0 ⇒ Ψ† (−γ 0 p̂0 − γ · p̂ − mc) = 0 , (15.35)

where I used (15.33). Multiplying by γ 0 and using (15.32) we find that

Ψ† (−γ 0 γ 0 p̂0 + γ 0 γ · p̂ − γ0 mc) = −Ψ† γ 0 (γ 0 p̂0 − γ · p̂ + mc) = 0 . (15.36)

We see the adjoint spinor Ψ̄ = Ψ† γ0 obeys the following equation

Ψ̄(γ µ p̂µ + mc) = 0 . (15.37)

Notice the sign difference between (15.31) and (15.37). With this notation we can write (15.3) and (15.11) as

ρ = Ψ† Ψ = Ψ̄γ0 Ψ , j = cΨ† αΨ = cΨ̄γΨ , (15.38)

or, employing the four-vector notation as

j µ = (cρ, j) = cΨ̄γ µ Ψ . (15.39)

The continuity equation reads as usual ∂µ j µ = 0.

• Additional reading:

§4. States of spin-1/2 particle with definite momentum

To solve the free Dirac equation is convenient to use the Hamiltonian form (15.19). It is instructive to consider first
the particle rest frame p = 0. The stationary states must depend on time as
i
Ψ = uε e− ~ εt , (15.40)

201
§4 States of spin-1/2 particle with definite momentum XV RELATIVISTIC SPIN-1/2 PARTICLES

where uε is a constant bispinor. Plugging (15.40) into (15.19) we obtain an equation

εuε = mc2 βuε ⇒ εβuε0 = mc2 uε . (15.41)

Explicitly:
    
1 0 0 0 c1 c1
    
0 1 0 0   c2   c2 
ε   = mc2  . (15.42)
0 0 −1 0   c3   c3 
0 0 0 −1 c4 c4

There are two positive-energy solutions with ε = mc2


  

1 0
   
(1) 0 1
uε = N   , u(2)
ε =N . (15.43)
0 0
0 0

and two negative-energy solutions with ε = −mc2


  

0 0
   
0 0
u(3)
ε =N , u(4)
ε =N . (15.44)
1 0
0 1
(1) (2)
Note that all u(n) ’s are eigenfunctions of operator Ŝz . Therefore, uε and uε represent spin-up and spin down
(3) (4)
positive-energy solutions, while uε and uε represent spin-up and spin-down negative energy solutions.
Now consider a free particle with momentum p. We are looking for a solution to the Dirac equation in a form of a
plane wave, viz.
i
Ψ(r, t) = upσ e ~ (p·r−εt) . (15.45)

For the bispinor upσ , where index p stands for (ε, p) and σ = ±1/2 for spin projections, we find

εupσ = (cα · p + mc2 β)upσ . (15.46)

Introduce the two-component spinors ϕ and χ as follows


!
ϕ
u= . (15.47)
χ

We have
! ! ! ! ! !
ϕ 0 σ·p ϕ 1 0 ϕ 1 0
ε =c + mc2 . (15.48)
χ σ·p 0 χ 0 −1 χ 0 −1

This is equivalent to the system of equations

(mc2 − ε)ϕ + cσ · pχ = 0 , (15.49)


2
cσ · pϕ − (mc + ε)χ = 0 , (15.50)

which has a non-trivial solution iff




mc2 − ε cσ · p
=0 ⇒ (mc2 )2 − ε2 + c2 (σ · p)(σ · p) = 0 . (15.51)
cσ · p −(mc2 + ε)

202
§4 States of spin-1/2 particle with definite momentum XV RELATIVISTIC SPIN-1/2 PARTICLES

Using property (6.59) of Pauli matrices we simplify (σ · p)2 = p2 , so that


p
ε = ±εp = ±c p2 + m2 c2 . (15.52)

These are the negative and positive-energy solutions. Eigenvectors of the Hamiltonian corresponding to solutions
(15.52) can be found from any of the two equations
cσ · p cσ · p
ϕ= χ, χ= ϕ, (ε = ±εp ) . (15.53)
ε − mc2 ε + mc2

One possible choice of two linearly independent solutions is to take states with definite z-projection of spin in the
rest frame:
! !
1 0
ϕ= and . (15.54)
0 1

Using (15.53) this leads to


! ! ! !
cσ · p 1 c pz px − ipy 1 c pz
χ= = = . (15.55)
εp + mc2 0 εp + mc2 px + ipy −pz 0 εp + mc2 px + ipy

Similarly for the other spin state. The final result for the positive energy solutions is
   
1 0
   
 0   1 
u(1) = N 1
  , u(2)
= N  c(p −ip ) . (15.56)
p cp z  p 2 x y 
 εp +mc2   εp +mc2 
c(px +ipy ) cpz
εp +mc2 − εp +mc 2

(1) (2)
I denoted up↑ = up , up↓ = up . By the same token, for the negative-energy states we choose
! !
1 0
χ= and (15.57)
0 1

and compute ϕ using (15.53). The corresponding negative-energy bispinors are


   
cpz c(px −ipy )
εp −mc2 εp −mc2
   
 c(px +ipy )   cpz
− εp −mc 
u(3) = N 
3 εp −mc2 , u(4) = N  2 . (15.58)
p  p 4 
 1   0 
0 1

(1)
The upσ -bispinors may be normalized as ūpσ upσ0 = δσσ0 . For upσ we obtain
!
(1) (1) 2 c2 p2z c2 (p2x + p2y ) 2εp
ūp up = |N1 | 1 + 2 2
+ 2 2
= |N1 |2 , (15.59)
(εp + mc ) (εp + mc ) εp + mc2

implying that
s
εp + mc2
N1 = . (15.60)
2εp

All other normalizations constants have the same value.

• Additional reading: M. Thomson “Modern Particle Physics”, 4.6.

203
§5 Second quantization of spin-1/2 particles XV RELATIVISTIC SPIN-1/2 PARTICLES

§5. Second quantization of spin-1/2 particles

As in the case of scalar particles in XIV §5, negative energy states indicate existence of antiparticles.(32 Repeating
the same arguments as in XIV §5, and noting that Ψ̄ satisfies the same equation as Ψ with p → −p (see (15.31) and
(15.37)) we arrive at the following expressions for the field operators:
X
Ψ̂ = (âpσ upσ e−ip·x/~ + b̂†pσ u−p−σ eip·x/~ ) , (15.61)

X
ˆ =
Ψ̄ (â†pσ ūpσ eip·x/~ + b̂pσ ū−p−σ e−ip·x/~ ) . (15.62)

Field operator Ψ̂(r, t) describes spin-1/2 particles and their antiparticles. Therefore, the creation and annihilation
operators satisfy the anti-commutation relations for fermions

{âpσ , â†p0 σ0 } = δpp0 δσσ0 , {b̂pσ , b̂†p0 σ0 } = δpp0 δσσ0 , (15.63)

while all other commutators vanish. If p is discrete (in a box) and

{âpσ , â†p0 σ } = δ(p − p0 )δσσ0 , {b̂pσ , b̂†p0 σ0 } = δ(p − p0 )δσσ0 (15.64)

in the infinite volume.


In place of the particle bispinors upσ given by (15.56) and (15.58) we can introduce a set of four linearly independent
(3) (4)
antiparticle bispinors vpσ in such a way that the negative energy bispinors up and up become the positive energy
(1) (2)
bispinors vp and vp as follows
(4) (3)
vp(1) = u−p , vp(2) = u−p . (15.65)

The reason for such definition is that under p → −p spin changes sign S → − S, see below XV §6. Introducing a new
pair of two-spinors
!
ϕ̃
u= . (15.66)
χ̃

we have (cp. (15.53))


cσ · p cσ · p
ϕ̃ = χ̃ , χ̃ = ϕ̃ , (ε = ±εp ) . (15.67)
ε + mc2 ε − mc2
The explicit form of the v-bispinors is
   
c(px −ipy ) cpz
εp +mc2 εp +mc2
   
 cpz
− εp +mc   c(px +ipy ) 
vp(1) =N

2 ,
 vp(2) =N

εp +mc2 ,
 ε = +εp . (15.68)
 0   1 
1 0
   
1 0
   
 0   1 
vp(3) =N
 cpz ,
 vp(4) =N
 c(px −ipy ) ,
 ε = −εp . (15.69)
 εp −mc2   εp −mc2 
c(px +ipy ) cpz
εp −mc2 − εp −mc 2

(32 To explain why positive energy particles do not spontaneously fall into negative energy states, Dirac suggested that vacuum is filled with
negative energy states, so that a positive energy fermion cannot fall into it due to the Pauli exclusion principle. Totality of all negative
energy states is called the Dirac sea. In Dirac’s picture production of a particle-antiparticle pair corresponds to a particles jumping
from the Dirac sea to acquire some positive energy simultaneously creating a “hole” interpreted as antiparticle. This interpretation is
only of historical interest and I mention it here only because the terms like “the Dirac sea” are still occasionally used.

204
§6 Spin and helicity states XV RELATIVISTIC SPIN-1/2 PARTICLES

(1) (2) (1) (2)


As a complete set of bispinors we can choose {up , up , vp , vp }, so that (15.61) and (15.62) read
X
Ψ̂ = (âpσ upσ e−ip·x/~ + b̂†pσ vpσ eip·x/~ ) , (15.70)

X
ˆ =
Ψ̄ (â†pσ ūpσ eip·x/~ + b̂pσ v̄pσ e−ip·x/~ ) . (15.71)

In order to compute the Hamiltonian in the second quantization formalism, which I will denote by Ĥ, we employ
(11.84) to write
Z Z
∂Ψ 3
Ĥ = Ψ ĤD Ψd r = i~ Ψ†
† 3
d r, (15.72)
∂t

where I used (15.19). Substituting (15.61) and (15.62), or (15.70) and (15.71), we find
X
Ĥ = εp (â†pσ âpσ − b̂pσ b̂†pσ ) . (15.73)

Using the anti-commutation relations (15.63) we arrive at


X
Ĥ = εp (â†pσ âpσ + b̂†pσ b̂pσ − 1) . (15.74)

Its eigenvalues (up to an infinite constant) are


X
E= εp (Npσ + N̄pσ ) . (15.75)

It is crucial operators âpσ and b̂pσ obey the anti-commutation relations, for otherwise energy (15.75) would have not
been positive-definite.
Operator of the field “charge” is given by
Z X X
Q̂ = Ψ̄ ˆ γ 0 Ψ̂d3 r = (â†pσ âpσ + b̂pσ b̂†pσ ) = (â†pσ âpσ − b̂†pσ b̂pσ + 1) . (15.76)
p pσ

Apart from the infinite additive constant the expectation value of this operator in the Fock space is
X
Q= (Np − N̄p ) , (15.77)
p

• Additional reading: Berestetskiy, Lifshitz, Pitaevskiy, “Quantum Electrodynamics”, §23,25, M. Thomson “Modern
Particle Physics”, Ch. 4.

§6. Spin and helicity states

(1) (2)
At p = 0 bispinors up and up are eigenstates of operator Ŝz :
!
~ ~ σz 0
Ŝz = Σ̂z = . (15.78)
2 2 0 σz

This indicates that spin-up and spin-down states are eigenstates just like in the non-relativistic case. However, at
(1) (2)
p 6= 0, generally, up and up are not eigenstates of Ŝz . Indeed, we have seen before in XV §2 that

[ĤD , Ŝz ] = ic~(α × p̂) · ez 6= 0 . (15.79)

205
§7 Charged spin-1/2 particle in electromagnetic field XV RELATIVISTIC SPIN-1/2 PARTICLES

Nevertheless, if particle moves in z-direction p = pez and [ĤD , Ŝz ] = 0. In this case
   
1 0
 0   1 
   
u(1)
p =N cp  , u(2)
p =N , (15.80)
 εp +mc2   0 
cp
0 − εp +mc2

   cp

0 εp +mc2
 − cp   
   0 
vp(1) = N  εp +mc vp(2) = N 
2
, , (15.81)
 0   1 
1 0

Clearly,
~ (1) ~
Ŝz u(1) = u , Ŝz u(2) = − u(2) . (15.82)
2 2
Under momentum flip p → −p, spin transforms as S → −S. Thus, spin operator for antiparticles is Ŝ A = −Ŝ. It
follows then that in states v (1) and v (2) antiparticle has spin projection on z-axis given by
~ (1) ~
ŜzA v (1) = (−Ŝz )v (1) = v , ŜzA v (2) = (−Ŝz )v (2) = − v (2) . (15.83)
2 2

Because only projection of spin on the direction of momentum is conserved, it is convenient to define the helicity
operator

Ŝ · p
ĥ = . (15.84)
p

Its conservation means that [ĤD , ĥ] = 0. We can thus label free particles by values of ε,p and h. Helicity operator
has two eigenvalues h = ± ~2 . These are called the right-handed and left-handed states. One of your home assignments
deals with the eigenvalue problem for ĥ. You will show that the right and left-handed states of a particle are given
by the following bispinors
   
s cos(θ/2) s − sin(θ/2)
εp + m 
 sin(θ/2)eiφ 
 εp + m 
 cos(θ/2)eiφ 

upR =  cp , upL =  cp . (15.85)
2εp  εp +mc2 cos(θ/2)  2εp  εp +mc2 sin(θ/2) 
cp cp
εp +mc2 sin(θ/2)eiφ − εp +mc 2 cos(θ/2)e

• Additional reading: M. Thomson “Modern Particle Physics”, 4.8.

§7. Charged spin-1/2 particle in electromagnetic field

As in the case of the scalar particle, interaction of the spin-1/2 particle with electromagnetic field is described by
making a replacement
q
p̂µ → p̂µ − µ (15.86)
c
in the Dirac equation:
h q i
γ µ (p̂µ − µ ) − mc Ψ = 0 . (15.87)
c
Gauge invariance can be verified in the same way as in XIV §6.

206
§8 Non-relativistic limit of Dirac equation XV RELATIVISTIC SPIN-1/2 PARTICLES

Eq. (15.87) can be written as a second order equation that is sometimes more convenient for practical applications.
Consider
h q ih q i
γ µ (p̂µ − µ ) + mc γ µ (p̂µ − µ ) − mc Ψ = 0 . (15.88)
c c
This is equivalent to
h q q i
γ ν γ µ (p̂ν − Âν )(p̂µ − µ ) − m2 c2 Ψ = 0 . (15.89)
c c
We can write identically
1 ν µ 1
γν γµ = (γ γ + γ µ γ ν ) + (γ ν γ µ − γ µ γ ν ) = g νµ + σ νµ , (15.90)
2 2
where g µν is symmetric and σ νµ is antisymmetric in µ, ν. Since σ νµ Sνµ = 0 for any symmetric tensor Sµν , see (C2),
we can write
q q 1n q q q q o
σ νµ (p̂ν − Âν )(p̂µ − µ )Ψ = σ νµ (p̂ν − Âν )(p̂µ − µ ) − (p̂µ − µ )(p̂ν − Âν ) Ψ , (15.91)
c c 2 c c c c

where expression in the curly brackets is antisymmetric. Using the commutator [p̂µ , Âν ]Ψ = i~∂µ Aν Ψ we obtain

q q q iq~ νµ
σ νµ (p̂ν − Âν )(p̂µ − µ )Ψ = σ νµ i~{(∂µ Aν ) − (∂ν Aµ )}Ψ = σ Fµν Ψ . (15.92)
c c 2c 2c
Using this in (15.89) we derive the desired equation
 
q iq~ νµ
(p̂ν − Âν )2 − m2 c2 − σ Fµν Ψ = 0 . (15.93)
c 2c

In the three-dimensional notations


" 2  #
i~ ∂ q q 2 2 2 q~ iq~
− Φ − i~∇ + A − m c + Σ · B − α · E Ψ = 0. (15.94)
c ∂t c c c c

Terms proportional to B and E describe interaction of the electromagnetic field with spin. We discuss their physical
meaning in the next few sections.

• Additional reading: Berestetskiy, Lifshitz, Pitaevskiy, “Quantum Electrodynamics”, §32.

§8. Non-relativistic limit of Dirac equation

A. Free particle

In the non-relativistic limit the two-component spinors ϕ and χ of the bispinor u in (15.47) are of different order of
magnitude. To see this, write ε = mc2 + E. In the non-relativistic limit E  mc2 , p  mc. For the positive-energy
solutions ε = εp (15.53) implies that
cσ · p σ·p
χ= 2
ϕ≈ ϕ ⇒ |χ|  |ϕ| . (15.95)
εp + mc 2mc

For the negative energy solutions ε = −εp

cσ · p cσ · p 2mcσ · p
χ= ϕ≈ ϕ= ⇒ |χ|  |ϕ| . (15.96)
−εp + mc2 −E −p2

Thus spinor ϕ (χ) is dominant for is particles (antiparticles) in the non-relativistic limit. This is why particles and
antiparticles decouple in the non-relativistic limit meaning that they satisfy independent Schrödinger equation.

207
§8 Non-relativistic limit of Dirac equation XV RELATIVISTIC SPIN-1/2 PARTICLES

Consider now a stationary state of energy ε but not necessarily a state with definite momentum p. We will redefine
the two-component spinors as follows
!
ϕ
Ψ= . (15.97)
χ

They still satisfy (15.53) where p should be replaced by the momentum operator

cσ · p̂ σ · p̂ i~
χ= ϕ≈ ϕ=− σ · ∇ϕ . (15.98)
ε + mc2 2mc 2mc
Particle densities become

ρ = Ψ† Ψ = ϕ† ϕ + χ† χ = |ϕ|2 + O(v 2 /c2 ) , (15.99)

while the current density reduces to


i~ †
j = cΨ† αΨ = c(ϕ† σχ + χ† σϕ) ≈ − [ϕ σ(σ · ∇ϕ) − (∇ϕ† · σ)σϕ]
2m
~ ~
= [ϕ† ∇ϕ − (∇ϕ† )ϕ] + ∇ × (ϕ† σϕ) + O(v 2 /c2 ) . (15.100)
2mi 2m
The last equation is obtained using the following identities

σ(σ · ∇)ϕ = ∇ϕ − i(∇ × σ)ϕ , (15.101)


† † †
(∇ϕ · σ)σ = ∇ϕ − i(∇ × σ)ϕ (15.102)

B. Particle in external field

Particles and antiparticles appear symmetrically in the Dirac equation, so it is sufficient to concentrate our attention
on particles. Our goal is to reduce the Dirac equation to the Schrödinger equation for a spin-1/2 particle. The result
will contain physically interesting relativistic corrections.
Start with the Dirac equation in the Hamiltonian form:
∂Ψ n q o
i~ = cα · (p̂ − A) + βmc2 + qΦ Ψ . (15.103)
∂t c
As explained in XIV §3 we have to redefine the wave function to remove the rest energy
2
Ψ = Ψ0 e−imc t~
. (15.104)

This yields
  n o
∂ q
i~ + mc2 Ψ0 = cα · (p̂ − A) + βmc2 + qΦ Ψ0 . (15.105)
∂t c

Writing Ψ0 as
!
ϕ0
Ψ0 = , (15.106)
χ0

we obtain for a system of equations for two-component spinors


 
∂ q
i~ − qΦ ϕ0 = cσ · (p̂ − A)χ0 , (15.107)
∂t c
 
∂ q
i~ − qΦ + 2mc2 χ0 = cσ · (p̂ − A)ϕ0 . (15.108)
∂t c

208
§8 Non-relativistic limit of Dirac equation XV RELATIVISTIC SPIN-1/2 PARTICLES

Now, similarly to (14.22), in the non-relativistic limit is achieved when


i~|∂t χ0 | = E|χ0 |  mc2 |χ0 | , |qΦ|  mc2 . (15.109)
Eq. (15.108) then allows us to eliminate χ0 in favor of ϕ0 :
σ · (p̂ − qc A) 0
χ0 ≈ ϕ . (15.110)
2mc
Using this in (15.107) we obtain an equation for ϕ0 . From now on, to simplify the notation, I drop the prime sign
near ϕ:
 
∂ q 1 q
i~ − qΦ ϕ = cσ · (p̂ − A) σ · (p̂ − A)ϕ . (15.111)
∂t c 2mc c
Employing identity (6.62) and using (A × ∇ + ∇ × A)ϕ = ∇ × Aϕ we derive
   
∂ 1 2 q q q2 2 q q
i~ − qΦ ϕ = p̂ − A · p̂ − p̂ · A + 2 A − iσ · (A × p̂) − σ · (p̂ × A) ϕ (15.112)
∂t 2m c c c c c
  
1 q 2 q~
= p̂ − A − σ · B ϕ . (15.113)
2m c c
Finally,
 
∂ϕ 1  q 2 q~
i~ = p̂ − A + qΦ − σ · B ϕ. (15.114)
∂t 2m c 2mc
This is of course the Pauli equation that we discussed in detail in VIII §1. The term −µ · B is the leading order
relativistic correction to the Schrödinger equation.
In the presence of electromagnetic field the current density receives an additional contribution as compared to
(15.100)
~ ~ eA †
j = cΨ† αΨ = [ϕ† ∇ϕ − (∇ϕ† )ϕ] + ∇ × (ϕ† σϕ) − ϕ ϕ + O(v 2 /c2 ) (15.115)
2mi 2m mc

∗ Examples.
2
1. Consider a scalar particle with charge −e moving in the Coulomb potential U = − Zer . In state ψ its current
density is
~
j= (ψ ∗ ∇ψ − c.c.) . (15.116)
2mi
The corresponding magnetic field (Biot-Savart law)
Z
e j(r 0 ) × (r − r 0 ) 3 0
B(r) = − d r . (15.117)
c |r − r 0 |3

In all s-states ψ is real, thus j = 0. In 2p-states:


r
1
r −r 3
ψ21,±1 = R21 Y1,±1 = ∓√ e 2a sin θe±iφ ≡ ψ± , (15.118)
24a a
3 8π
~2
where a = Ze2 m . Substituting into (15.116):
 ∗ 
~ ∗ 1 ∂ψ± 1 ∂ψ± ~ sin θr −r/a
j± = ψ± − ψ± eφ = ± e eφ . (15.119)
2mi r sin θ ∂φ r sin θ ∂φ 64πma5
Using this in (15.117) and recalling that
eφ × er = eθ = cos θ cos φex + cos θ sin φey − sin θez (15.120)
we obatin
Z Z ∞ Z
e j(r 0 ) × r 0 3 0 e ~ π
sin2 θ −r/a e~
B(0) = ± 03
d r = ∓ ez e 2π sin θdθr2 dr = ∓ ez . (15.121)
c r c 54πma5 0 0 r 24mca3

209
§8 Non-relativistic limit of Dirac equation XV RELATIVISTIC SPIN-1/2 PARTICLES

2. If the particle has spin-1/2, then its current is non-vanishing even in the ground state due to the spin-dependent
term

js = c∇ × (ψ † µψ) , µ = µB σ . (15.122)

The ground state wave function


1 −r/a
ψ0 = e χ, (15.123)
πa3
where χ is spinor. We have
1 −2r/a
j = cµB ∇ × (ρχ† σχ) , ρ= e . (15.124)
πa3
Denote hσi = χ† σχ. The vector potential reads
Z Z
1 j(r 0 ) 3 0 ∇0 × (ρ hσi) 3 0
A(r) = d r = µB d r . (15.125)
c |r − r 0 | |r − r 0 |

It is convenient to proceed using tensor notations


Z
∂j0 ρ 3 0
Ai (r) = µB ijk hσk i d r (15.126)
|r − r 0 |

Writing identically
   
∂j0 ρ ρ 1 ρ 1
= ∂j0 − ρ∂j0 = ∂j0 + ρ∂j , (15.127)
|r − r 0 | |r − r 0 | |r − r 0 | |r − r 0 | |r − r 0 |

and using the Gauss’ theorem


Z   I
ρ ρ
∂j0 d3 r0 = dSj = 0 (15.128)
|r − r 0 | Sj →∞ |r − r 0 |

(because ρ falls off exponentially with r) we arrive at


Z
ρ(r 0 ) 3 0
Ai = µB ijk hσk i ∂j d r ≡ ijk Mk ∂j Φ(r) , (15.129)
|r − r 0 |

where we denoted the magnetization M = µB hσi and


Z
ρ(r 0 ) 3
Φ(r) = d r (15.130)
|r − r 0 |

which coincides up to a factor −e with the scalar potential. We can compute it by expanding |r − r 0 |−1 in
spherical harmonic series using (D12). Only the first, spherically symmetric term survives integration over
angles. The result is
 
1 1 1
Φ(r) = − + e−2r/a . (15.131)
r r a

Magnetic field B = ∇ × A:

Bi = ijk ∂j Ak = −ijk ∂j klm Ml ∂m Φ = −(δil δjm − δij δlm )∂j Ml ∂m Φ = −Mi ∂ 2 Φ + Ml ∂i ∂l Φ , (15.132)

where I used (C7).


At r  a:
1 2r2 4ri 4
Φ≈ − 3 + ... ⇒ ∂i Φ = − ⇒ ∂k ∂i Φ = − δik . (15.133)
a 3a 3a3 3a3

210
§9 Relativistic corrections to motion in spherically symmetric potential
XV RELATIVISTIC SPIN-1/2 PARTICLES

Thus,
4 8
B= 3
(3M − M ) = 3 M , r  a. (15.134)
3a 3a

At r  a,
1 ri δik 4ri rk
Φ≈ ⇒ ∂i Φ = − ⇒ ∂i ∂k Φ = − + 5 . (15.135)
r r3 r 3 r
Thus, ∂ 2 Φ = 0 (Laplace equation) and

Mi 3r · M 3(M · r)r − M r2
Bi = − + ri ⇒ B= , r  a. (15.136)
r3 r5 r5
This is the magnetic dipole M field.

• Additional reading: Berestetskiy, Lifshitz, Pitaevskiy, “Quantum Electrodynamics”, §33.

§9. Relativistic corrections to motion in spherically symmetric potential

In the previous section we develop a method of calculating the relativistic correction to the Schrödinger equation
and considered the leading corrections that lead us to the Pauli equation. In this section we apply this method to
compute relativistic corrections in the case of the spherically symmetric potential U = qΦ(r), A = 0. Consider a
stationary state with energy ε = mc2 + E. Eqs. (15.107),(15.108) read (I drop the primes near ϕ and χ)

(E − qΦ) ϕ = cσ · p̂χ , (15.137)



E − qΦ + 2mc2 χ = cσ · p̂ϕ . (15.138)
E−qΦ
We treat 2mc2 as a small parameter. From (15.138) we exclude χ keep the next-to-leading term:
 
1 E − qΦ
χ≈ 1− (σ · p̂)ϕ . (15.139)
2mc 2mc2

Using it in (15.137) we get


 
1 E − qΦ
(E − qΦ)ϕ = (σ · p) 1 − (σ · p̂)ϕ . (15.140)
2m 2mc2

Since we are interested in corrections of order v 2 /c2 we need to examine the probability density to the same accuracy:

1 ~2
ρ = ϕ† ϕ + χ† χ ≈ ϕ† ϕ + 2 2
(σ · p̂ϕ)† (σ · p̂)ϕ = ϕ† ϕ + (∇ϕ† ) · σ σ · ∇ϕ . (15.141)
4m c 4m2 c2
There is a problem here: a wave function ψ satisfying Schrödinger equation should be such that the probability density
is ρ = ψ † ψ. Clearly, ϕ does not satisfy this condition. Therefore, we wish to replace ϕ by a new function ψ such that
the total probability does not change, viz.
Z Z Z  
~2
ρd3 r = ψ † ψd3 r = ϕ† ϕ + (∇ϕ †
) · σ σ · ∇ϕ d3 r . (15.142)
4m2 c2

The requirement of probability conservation ensures that ϕ → ψ transformation is unitary. The right-hand-side of
(15.142) can be integrated by parts and using the identity (a · σ)2 = a2 yields
Z Z  
† 3 † ~2 † 2 2 †
ψ ψd r = ϕ ϕ− (ϕ ∇ ϕ + ϕ∇ ϕ ) d3 r . (15.143)
8m2 c2

211
§9 Relativistic corrections to motion in spherically symmetric potential
XV RELATIVISTIC SPIN-1/2 PARTICLES

It is not difficult to see that the required transformation is


   
p̂2 p̂2
ψ = 1+ ϕ , ϕ = 1 − ψ. (15.144)
8m2 c2 8m2 c2

Remember that terms smaller than ∼ v 2 /c2 are neglected. Replacing ϕ with ψ in (15.139) we obtain with the required
accuracy
     
p̂2 p̂2 p̂2 p̂2 p̂2 q
E 1− ψ = qΦ 1 − ψ + 1 − ψ − Eψ + (σ · p̂)Φ(σ · p̂)ψ , (15.145)
8m2 c2 8m2 c2 2m 8m2 c2 4m2 c2 4m2 c2

We can cast this equation in the form of the Schrödinger equation Eψ = Ĥψ with

p̂2 q p̂2 p̂2 p̂2 p̂2


Ĥ = + qΦ + 2 2
(σ · p̂)Φ(σ · p̂) − 2 2
E − qΦ 2 2 − . (15.146)
2m 4m c 8m c 8m c 2m 8m2 c2
To simplify it note, that
 
p̂2
Eψ = + qΦ ψ + O(v 2 /c2 ) , (15.147)
2m
implying that

p̂2 p̂4 q
2 2
Eψ = 3 2
ψ+ p̂2 Φψ . (15.148)
8m c 16m c 8m2 c2
Thus,

p̂2 q p̂4 q
Ĥ = + qΦ + (σ · p̂)Φ(σ · p̂) − − (Φp̂2 + p̂2 Φ) . (15.149)
2m 4m2 c2 8m3 c2 8m2 c2
We can further simplify

(σ · p̂)Φ(σ · p̂) = σi p̂i Φ(σ · p̂) = σi [p̂i , Φ](σ · p̂) + Φσi p̂i (σ · p̂)
= −i~σi ∇i Φ(σ · p̂) + Φp̂2 = i~(σ · E)(σ · p̂) + Φp̂2 . (15.150)

The Hamiltonian takes form


 
p̂2 p̂4 q 2 1 2 2
Ĥ = + qΦ − + i~(σ · E)(σ · p̂) + Φp̂ − (Φp̂ − p̂ Φ) . (15.151)
2m 8m3 c2 4m2 c2 2
Now

(p̂2 Φ − Φp̂2 )ψ = −~2 ∇2 (Φψ) − Φp̂2 ψ = −~2 (ψ∇2 Φ + 2∇ψ · ∇Φ + Φ∇2 ψ) − Φp̂2 ψ
= (−~2 ∇2 Φ + 2i~E · p̂)ψ . (15.152)

(σ · E)(σ · p̂) = E · p̂ + iσ · (E × p̂) . (15.153)

Finally, we derive the following Hamiltonian with relativistic correction to the order v 2 /c2 :

p̂2 p̂4 e~ e~2


Ĥ = + qΦ − − σ · (E × p̂) − ∇·E. (15.154)
2m 8m3 c2 4m2 c2 8m2 c2
The first correction
p̂4
V̂1 = − (15.155)
8m3 c2
in (15.154) comes from the expansion of the relativistic dispersion relation:
p p2 p4
E − mc2 = p2 c2 + m2 c2 − mc2 ≈ − + ... (15.156)
2m 8m3 c2

212
§10 Relativistic corrections in hydrogen-like atom XV RELATIVISTIC SPIN-1/2 PARTICLES

In practical applications it is sometimes convenient to write using (15.147) as

(E − qΦ)2
V̂1 = − . (15.157)
2mc2
The second correction is
e
V̂2 = − Ŝ · (E × p̂) . (15.158)
2m2 c2
In spherically symmetric potential E = − ∂Φ
∂r er . Thus

e ∂Φ e ∂Φ
V̂2 = 2 2
Ŝ · (r × p̂) = 2 2
Ŝ · L̂ . (15.159)
2m c r ∂r 2m c r ∂r
This is the spin-orbit operator that we already discussed in XII §5.
The third relativistic correction simplifies in the Coulomb field U = qΦ = −Ze2 /r satisfying the Poisson equation
∇ · E = −4πeZδ(r):

πe2 ~2
V̂3 = δ(r) . (15.160)
8m2 c2
This is the “contact” term that simply shifts energy of s-states by a constant.

• Additional reading: Berestetskiy, Lifshitz, Pitaevskiy, “Quantum Electrodynamics”, §33.

§10. Relativistic corrections in hydrogen-like atom

In this section we apply the results of the previous section to compute the relativistic correction to the spectrum
of the hydrogen-like atom. The corresponding potential is U = qΦ = −Ze2 /r and the Schrödinger equation reads

(Ĥ0 + V̂1 + V̂2 + V̂3 )ψ = Eψ . (15.161)

Where Ĥ0 is the non-relativistic Hamiltonian and


(E − Ze2 /r)2 Ze2 πe2 ~2
V̂1 = − , V̂2 = − Ŝ · L̂ , V̂3 = δ(r) (15.162)
2mc2 2mc c2 r3 8m2 c2
are the corrections. We can write the spin-orbit coupling term as
Ze2
V̂2 = − (Jˆ2 − L̂2 − Ŝ 2 ) . (15.163)
4mc c2 r3
The (full) Hamiltonian commutes with operators Jˆ2 , L̂2 , Ŝ 2 , hence the stationary states can be chosen as eigenstates
of those operators with eigenvalues Jˆ2 = ~2 j(j + 1), L̂2 = ~2 l(l + 1), Ŝ 2 = 3~2 /4:

ψ(r) = fnlj (r)hθφ|lsjmj i (15.164)

At the zeroth order in the perturbation theory the Schrödinger equation

Ĥ0 ψ (0) = E (0) ψ (0) (15.165)

is solved by ψ (0) = Rn (r)Ylm (θ, φ), given in V §5, with the corresponding spectrum

Z 2 e2 ~2
En(0) = − , n = 1, 2, . . . , a= . (15.166)
2n2 a me2
According to the stationary perturbation theory rules, the first order correction to the spectrum is
Z ∞
(1)
Enj = [Rn (r)]2 (V̂1 + V̂2 + V̂3 )r2 dr , (15.167)
0

213
§10 Relativistic corrections in hydrogen-like atom XV RELATIVISTIC SPIN-1/2 PARTICLES

where I took into account that the perturbations are spherically symmetric (i.e. depend only on r). Integration can
be down using (5.68)-(5.69). The calculation is rather straightforward though laborious. The final result is
 
(1) α 2 Z 4 e2 1 3
Enj = − − . (15.168)
2n3 a j + 1/2 4n

With the account of spin the total number of levels with given n is 2n2 . We observe that the n2 -fold degeneracy
(0) (1)
of the non-relativistic spectrum En is partially lifted. Since the relativistic corrections Enj do not depend on l, the
levels with the same n and j but different l = j ± 1/2 are still degenerate. Such degeneracy remains even in the exact
solution of the Dirac equation (see below).
The lowest levels of hydrogen atom are (in notation nlj ):

1s1/2 , 2s1/2 , 2p1/2 , 2p3/2 , 3s1/2 , 3p1/2 , 3p3/2 , 3d3/2 , 3d5/2 , . . . (15.169)

where I underlined states that have the same energy (i.e. degenerate) at the the order O(α2 ) of the perturbation theory.
Levels with the same n but different j’s are called the fine structure and is the result of the spin-orbit coupling.
Additional splitting of levels (e.g. 2s1/2 and 2p1/2 ) is caused by interaction of atomic electron with electron-positron
pairs that populate vacuum and is knows as the Lamb shift. This correction is of the order α5 .
Interaction of small nuclear magnetic moment with electron magnetic moment results in hyperfine splitting of energy
levels.

. Dirac equation with Coulomb potential can be solved exactly with the following result:
 −1/2
  (
 (αZ)2 
2 2 −l − 1 , j = l + 1/2 ,
ε = E + mc = mc 1 + p 2 , κ= (15.170)

 
 l, j = l − 1/2 .
κ2 − (αZ)2 + nr

Expanding this formula at small αZ we have


  
mc2 (αZ)2 (αZ)2 1 3
E≈− 1 + − , n = nr + |κ| . (15.171)
2n2 n |κ| 4n

Noting that |κ| = j + 1/2 we reproduce (15.168).


The exact solution indicates that for large Z such that when αZ > 1/2 the hydrogen becomes unstable because E
is complex. This is similar to the effect observed in (14.72).

• Additional reading: Berestetskiy, Lifshitz, Pitaevskiy, “Quantum Electrodynamics”, §33-36; Schiff, “Quantum
mechanics”, 53; Messiah, “Quantum mechanics”, V.28,29,34.

214
§1 Second quantization of E&M field XVI QUANTIZATION OF ELECTROMAGNETIC FIELD

XVI. QUANTIZATION OF ELECTROMAGNETIC FIELD

§1. Second quantization of E&M field

In the previous chapters we discussed quantum mechanics of relativistic particles of spins s = 0 and s = 1/2. We
showed that a self-consistent approach requires upgrading the wave function ψ to the ψ̂ field operator that described
multi-particles states. One can similarly develop a theory of particles of higher spins. In this chapter we will focus
on quantization of electromagnetic field which describes massless particle of spin one called photon.
Consider electromagnetic field in a big box with side L. As before we set L = 1 to simplify the notations. A
convenient gauge that we employ is Φ = 0 and ∇ · A = 0. We start by expanding the electromagnetic potential in
Fourier series
X 
A(r, t) = bk (t)eik·r + dk (t)e−ik·r , k = {kx , ky , kz } , kx = 2πnx , . . . , nx = 1, 2, . . . (16.1)
k

Since A is real, A∗ = A implying that dk = b∗k . Thus,


X 
A(r, t) = bk (t)eik·r + b∗k (t)e−ik·r . (16.2)
k

Time-dependence of the coefficients follows from the wave equation

1 ∂2A
− ∇2 A = 0 ⇒ b̈k − c2 k 2 bk = 0 , (16.3)
c2 ∂t2
implying that

bk ∝ e−iωk t , ωk = ck . (16.4)

This means that (16.2) is an expansion in plane waves. The gauge condition imposes a constraint on the components
of ak :

bk · k = b∗k · k = 0 , (16.5)

which means that the wave is transverse.


Before promoting the coefficients bk and b∗k to annihilation and creation operators, we need to figure out their
normalization, so that the commutation relations can be written down as (11.26). To this end, compute the total
energy of electromagnetic field in the box
Z Z  
1 1 1 2
E= (E 2 + B 2 )d3 r = Ȧ + (∇ × A)2
d3 r . (16.6)
8π 8π c2
RL 0
Substituting (16.2) and using 0
ei(kx −kx )x dx = δkx kx0 L etc. we find
( )
1 2 X ∗ 2
X

E= bk · bk ωk + 2 (k × bk )(k × bk ) . (16.7)
8π c2
k k

Using a vector algebra identity (k × bk )(k × b∗k ) = k 2 (bk · b∗k ) we arrive at

1 X 2
E= k bk · b∗k . (16.8)

k

This formula implies that the total energy of electromagnetic field is a sum of excitations of elementary harmonic
oscillators with energies k 2 /2π, see (4.113). Introducing new coefficients
r
2π~ωk
bk = ak , (16.9)
k2

215
§2 Photon radiation by an atom XVI QUANTIZATION OF ELECTROMAGNETIC FIELD

eq. (16.8) takes form of


X
E= ~ωk ak · a∗k . (16.10)
k

To quantize this system, we need to write (16.10) in terms of the canonical coordinates and momenta qk and πk ,
which we can be defined as follows (see (3.124),(3.125))
r r
~ωk ∗ ~
qk = (ak + ak ) , πk = −i (ak − a∗k ) . (16.11)
2 2ωk

Indeed, using these formulas (16.10) can be written in the canonical form
X1
E= (πk2 + ωk2 qk2 ) . (16.12)
2
k

Promoting the canonical coordinates and momenta to operators, we write the quantization conditions as

[π̂k , q̂k0 ] = −i~δkk0 , [π̂k , π̂k0 ] = 0 , [q̂k , q̂k0 ] = 0 . (16.13)

Using (16.11), for the corresponding operators we obtain a quantized analogue of the classical equation (16.10)
X
Ĥ = ~ωk (â†k âk + 1/2) . (16.14)
k

Finally, the secondly quantized vector potential reads


r
2π~ωk X  
Â(r, t) = 2
âk (t)eik·r + â†k (t)e−ik·r . (16.15)
k
k

Operators âk contain two independent components corresponding to two independent polarizations of electromagnetic
wave. Let α be the polarization vectors, α = 1, 2. We can write
s
2π~c2 X  
Â(r, t) = âkα kα (t)eik·r + â†kα (t)∗kα e−ik·r . (16.16)
ωk

where âkα is the annihilation operator of a photon with momentum ~k and polarization α. The commutation relations:

[âkα , â†k0 α0 ] = δkk0 δαα0 , [âkα , âk0 α0 ] = 0 , [â†kα , â†k0 α0 ] = 0 . (16.17)

• Additional reading: Berestetskiy, Lifshitz, Pitaevskiy, “Quantum Electrodynamics”, §2.

§2. Photon radiation by an atom

In this section we employ the second quantization formalism to calculate photon radiation by a heavy atom. Heavy
atom is a non-relativistic system whose interaction with the electromagnetic field can be described by the Pauli
equation (15.114). If electromagnetic field is weak and spin spin-related effects are small, than the interaction is
described by the following operator
e
V̂ = Â · p̂ . (16.18)
mc

Suppose that the non-perturbed Hamiltonian of atom is ĤA and that of electromagnetic field is ĤF , given by
(16.14). Let ψn be the eigenstates of ĤA with the corresponding eigenvalues En . The eigenstates of ĤF are |nkα i.
Eigenstates of the complete system are ψn |nkα i.

216
§2 Photon radiation by an atom XVI QUANTIZATION OF ELECTROMAGNETIC FIELD

Radiation of photon is described by the creation operator in (16.16). Therefore, the corresponding interaction term
is
r
e 2π~ X †
V̂+ = âkα (t)(∗kα · p̂) e−ik·r . (16.19)
m ωk

Let the initial and the final states of the system be |ii = ϕi |nkα i and |f i = ϕf |nkα + 1i. Using

â†kα |nkα i = nkα + 1|nkα + 1i , (16.20)
we obtain
r
e 2π~ √
hf |V̂+ |ii = nkα + 1 ∗kα · hϕf |e−ik·r p̂|ϕi i . (16.21)
m ωk
The emission rate is given by

ẇωα = |hf |V̂+ |ii|2 ρ(εf ) . (16.22)
~
where ρ(εf ) density of states of the final state photons. Since εf = ~ω and pf = εf /c, the number of states in a unit
volume (V = 1) with given momentum p and polarization α is

d3 p f ε2f dpf dΩ ω 2 dΩ
dνf = ρ(εf )dεf = 3
= 2 3
= dεf . (16.23)
(2π~) c (2π~) (2πc)3 ~

The problem thus reduces to calculation of the matrix element in (16.21). It involves integration over the distance
for the center of atom r. The wave functions decrease exponentially when r  a, where a is atomic radius. Therefore,
the main contribution to the matrix element come about from distances of the order of shorter than a which for atoms
is a ∼ 10−8 cm. Consider emission of light with wavelengths λ  a. For example, for visible light λ ∼ 10−5 cm, for
UV light λ ∼ 10−6 cm. For such wavelengths ka = 2πa λ  1 and we can expend the exponent in (16.21). Similarly we
can discuss emission of γ-rays by a nucleus of radius ∼ 10−12 cm. In the following we assume that ka  1 and write
in the leading approximation
hϕf |e−ik·r p̂|ϕi i ≈ hϕf |p̂|ϕi i , λ  a. (16.24)

Now, ĤA = p̂2 /2m + U , hence [r̂, ĤA ] = i~p̂/m. Taking matrix element
i~
hϕf |r̂ ĤA |ϕi i − hϕf |ĤA r̂|ϕi i = hϕf |p̂|ϕi i . (16.25)
m
Thus,
m
hϕf |p̂|ϕi i = (Ei − Ef )hϕf |r̂|ϕi i . (16.26)
i~
Substituting into (16.21) and introducing the matrix element of the electric dipole model
df i = −ehϕf |r̂|ϕi i (16.27)
we write
r
2π~ √
hf |V̂+ |ii = −iωf i nkα + 1 ∗kα · df i . (16.28)
ωk
Such radiation is called the electric dipole radiation. Combining (16.22),(16.23),(16.28) we find the rate
ω 3 (nkα + 1) ∗
dẇωα = |kα · df i |2 dΩ . (16.29)
2πc3 ~
This formula gives rate of emission of polarized photons with energy ω in the solid angle dΩ (around the direction of
k). Radiation power can be computed as
dPωα dẇωα ω 4 (nkα + 1) ∗
= ~ω = |kα · df i |2 . (16.30)
dΩ dΩ 2πc3

217
§2 Photon radiation by an atom XVI QUANTIZATION OF ELECTROMAGNETIC FIELD

An important feature of (16.30) is that it is finite even when the number of photons in the initial state vanishes
nkα = 0 (i.e. when there is no electromagnetic field). This phenomenon is known as the spontaneous emission. The
part of the power proportional to nkα is called the induced emission.
Summation over all photon polarizations can be done using the following formula known the theory of classical
electromagnetic radiation
X ki kj
∗kα,i kα,j = δij − . (16.31)
α=1,2
k2

The result is
dPω X dPωα ω 4 (nkα + 1)
= = 3
|df i |2 sin2 θ , (16.32)
dΩ α
dΩ 2πc

where θ is the angle between k and d. The total power of spontaneous radiation is obtained by integrating (16.32)
over the photon emission angles

2ω 4
Pω = |df i |2 . (16.33)
3c3
This is the quantum analogue of the classical formula for the dipole radiation (in the classical formula the matrix
element df i is replaced by the electric dipole vector d).

• Additional reading:

218
B DIRAC DELTA FUNCTION

Appendix A: Fourier analysis

Any integrable function f (x) can be expanded in a complete set of functions {eikn x }n in the interval −L/2 ≤ x ≤
L/2, known as Fourier series, as follows

X 2πn
f (x) = fn eikn x , kn = . (A1)
n=−∞
L

The Fourier coefficients fn are given by


Z L/2
1
fn = f (x)e−ikn x dx . (A2)
L −L/2

In the limit L → ∞ (A1),(A2) become continuous Fourier integral expansion


Z +∞
dk
f (x) = fk eikx , (A3)
−∞ 2π
Z +∞
fk = f (x)e−ikx dx , (A4)
−∞

Three-dimensional generalization of (A3),(A4) is


Z +∞
d3 k
f (r) = fk eik·r , (A5)
−∞ (2π)3
Z +∞
fk = f (r)e−ik·r d3 r , (A6)
−∞

• Additional reading: Arfken at. al. , “Mathematical methods for physicists”, 7th edition, Ch.19, 20.1-20.4.

Appendix B: Dirac delta function

Dirac delta function δ(x) is a singular function such that


(
0 , if x 6= 0 ,
δ(x) = (B1)
∞ , if x = 0 ,

and
Z +∞
δ(x)dx = 1 . (B2)
−∞

This definition implies the main property of the delta function: if f (x) is a smooth function and a < 0, b > 0, then
Z b
f (x)δ(x)dx = f (0) . (B3)
a

In particular,
Z +∞
f (x)δ(x)dx = f (0) . (B4)
−∞

Shifting the origin x → x − a we get in place of (B4)


Z +∞
f (x)δ(x − a)dx = f (a) . (B5)
−∞

219
C LEVI-CIVITA SYMBOL

Often one needs do deal with a delta function of the form δ[ϕ(x)], where ϕ is a smooth function. Suppose that
ϕ(x) has one and only zero at x = x0 . Then integrand in
Z +∞
f (x)δ[ϕ(x)]dx (B6)
−∞

is non-vanishing only at x = x0 . Thus, we can replace ϕ with the first non-vanishing term of its Taylor expansion
ϕ(x) = ϕ0 (x0 )(x − x0 ). We have
Z +∞ Z +∞
f (x)δ[ϕ(x)]dx = f (x)δ[ϕ0 (x0 )(x − x0 )]dx. (B7)
−∞ −∞
0
Making change of variables y = ϕ (x0 )(x − x0 ) we get using (B4)
Z +∞ Z +∞
1 f (x0 )
f (x)δ[ϕ(x)]dx = 0 f [y/ϕ0 (x0 ) + x0 ]δ(y)dy = 0 . (B8)
−∞ |ϕ (x0 )| −∞ |ϕ (x0 )|
Generally, if ϕ(x) has zeros at x = xn , were n is any integer number bigger than zero, then (B8) becomes
Z +∞ X f (xn )
f (x)δ[ϕ(x)]dx = . (B9)
−∞ n
|ϕ0 (xn )|

In a particular case ϕ(x) = cx, with constant c:


Z +∞
1
f (x)δ(cx)dx = f (0) . (B10)
−∞ |c|

Delta function can be represented in different forms that are often very useful in applications. As a Fourier integral:
Z +∞
dk
δ(x) = eikx . (B11)
−∞ 2π
As a limit of elementary functions:
1 
δ(x) = lim (B12)
π →0 x2 + 2
1 sin(x/)
= lim (B13)
π →0 x
1 1 2 2
= √ lim 2 e−x /4 (B14)
2 π →0 
Note that (B11) and (B13) are mutually consistent as follows:
Z 1/
1 1 sin(x/)
lim eikx dk = lim (B15)
2π →0 −1/ π →0 x

Three-dimensional delta function is defines as a product of three one-dimensional delta-functions


δ(r − r 0 ) = δ(x − x0 )δ(y − y 0 )δ(z − z 0 ) . (B16)

• Additional reading: Arfken et. al. “Mathematical methods for physicists”, 7th edition, 1.11.

Appendix C: Levi-Civita symbol

Levi-Civita symbol ijk is defined as




 +1 if (i, j, k) is (1, 2, 3), (3, 1, 2), (2, 3, 1) ,
ijk = −1 if (i, j, k) is (1, 3, 2), (3, 2, 1), (2, 1, 3) , (C1)

 0 if i = j or j = k or k = i

220
D LEGENDRE POLYNOMIALS

Levi-Civita symbol is completely antisymmetric: ikj = −ijk , jik = −ikj etc. That is, permutation of any two
nearby indices produces the minus sign. The nearby pairs of indices of ikj are i and k, k and j, j and i. With this
notation the i’th component of vector A = B × C can be written as Ai = ijk Bj Ck , where summation over the
repeated indices is implied(33 . The scalar product of any two vectors A and B reads A · B = Ai Bi .
We can use this, so called index notation to easily prove vector identities. For example,

A × (B × C) = ei ijk Aj klm Bl Cm = ei (δil δjm − δim δjl )Aj Bl Cm = B(A · C) − C(A · B) ,

where ei is a unit vector in i’th direction.


It is helpful to remember that if matrix [S]ij is symmetric, i.e. [S]ij = [S]ji and [A]ij is antisymmetric, i.e.
[A]ij = −[A]ji , then

[S]ij [A]ij = −[S]ji [A]ji = −[S]ij [A]ij = 0 (C2)

where we first permuted the indices i → j, j → i and than relabeled i as j and j as i.



The differential operator nabla is defined as ∇ = {∂x , ∂y , ∂z }, where ∂i ≡ ∂x i
. Some operations with nabla:

∇ · A ≡ δij ∂i Aj = ∂i Ai (C3)

∇φ = ei ∂i φ (C4)

∇ × A = ijk ei ∂j Ak (C5)

Examples of differential identities:

∇ · (A × B) = ∂i (ijk Aj Bk ) = ijk (∂i Aj )Bk + ijk (∂i Bk )Aj


= ijk (∂i Aj )Bk − ikj (∂i Bk )Aj = (∇ × A) · B − (∇ × B) · A

∇ · (∇ × A) = ∂i ijk ∂j Ak = 0, (C6)

since ∂i ∂j = ∂j ∂i is symmetric, whereas ijk is asymmetric in any pair of its indices.


A useful property of the Levi-Civita symbol:

ijk ilm = δjl δkm − δjm δkl . (C7)

It helps simplify multiple cross products.

Appendix D: Legendre polynomials

When separating variables in Schrödinger equation for spherically symmetric potential, one arrives at the following
equation
   
1 d dP m2
sin θ + `(` + 1) − P = 0. (D1)
sin θ dθ dθ sin2 θ
Denoting x = cos θ we rewrite it as follows:
   
d dP m2
(1 − x2 ) + `(` + 1) − P = 0. (D2)
dx dx 1 − x2

(33
P
In other words, we do not write explicitly the summation sign j,k . This is called the Einstein summation convention. The repeated
indices i, j are called dummy indices.

221
D LEGENDRE POLYNOMIALS

Finite solutions to (D2) with m = 0 are called the Legendre Polynomials Pl (x) and form a complete set on interval
D = {−1 ≤ x ≤ x}. They are normalized such that Pl (1) = 1. Orthogonality property:
Z 1
2
P` (x)P`0 (x)dx = δ`,`0 . (D3)
−1 2` + 1

Completeness of the set {P` }∞


`=0 implies that any function can be expanded in −1 ≤ x ≤ 1 as follows


X
f (x) = A` P` (x) . (D4)
`=0

Multiplying both sides of (D4) by P`0 and integrating over the interval −1 ≤ x ≤ 1 using (D3) yields
Z 1
2` + 1
A` = f (x)P` (x)dx . (D5)
2 −1

Legendre Polynomials have definite parity P` (−x) = (−1)` P` (x).


Solution of (D2) in the case m 6= 0 are denoted by P`m (x) and are called the generalized Legendre polynomials.
Values of m are restricted to m = −`, −` + 1, . . . , ` − 1, `. For each m they form a complete orthogonal set. Their
explicit expressions can be found in the references at the end of this section. In practice, it is convenient to introduce
a set of functions that is complete not only in the interval 0 ≤ θ ≤ π but also in 0 ≤ φ ≤ 2π (θ and φ are the polar and
the azimuthal angle correspondingly). In other words, we would like to have a complete orthonormal set of functions
on a unit sphere. These functions are called the spherical harmonics and defined as
s
2` + 1 (` − m)! m
Ylm (θ, φ) = P (cos θ)eimφ . (D6)
4π (` + m)! `

Orthonormality property means that


Z
Y`m (θ, φ)Y`∗0 m0 (θ, φ) dΩ = δ`0 ` δmm0 . (D7)

Here the solid angle is dΩ = d cos θ dφ. A few lowest spherical harmonic:
r r
1 3 iφ 3
Y00 = √ , Y11 = − sin θe , Y10 = cos θ . (D8)
4π 8π 4π

Note also the following properties



Y`,−m (θ, φ) = (−1)m Y`,m (θ, φ) , (D9)

and
r
2` + 1
Y`0 (θ, φ) = P` (cos θ) . (D10)

Completeness of spherical harmonics means that we can expand an arbitrary function on a unit sphere
∞ X̀
X Z

f (θ, φ) = A`m Y`m (θ, φ) , A`m = Y`m f (θ, φ)dΩ . (D11)
`=0 m=−`

Coulomb potential that we expanded as


∞ X̀
X `
1 1 r<
= 4π Y ∗ (θ0 , φ0 )Y`m (θ, φ) .
`+1 `m
(D12)
|r − r 0 | 2` + 1 r>
`=0 m=−`

222
E BESSEL FUNCTIONS

Appendix E: Bessel functions

Consider the Bessel equation:


 
00 1 0 ν2
R (x) + R (x) + 1 − 2 R(x) = 0 . (E1)
x x

For non-integer ν the two linearly independent solutions are the Bessel functions of the first kind, or simply the
Bessel functions, denoted by Jν (x) and J−ν (x). To obtain two linearly independent solutions even for integer ν, one
introduces the Bessel functions of the second kind, known also as the Neumann functions, denoted by Nν (x). The pair
Jν (x) and Nν (x) are linearly independent for all ν. Another convenient choice of the linearly independent solutions
(1) (2)
are the Bessel functions of the third kind, known also as the Hankel functions, denoted by Hν (x) and Hν (x).
In the following we will need asymptotic expansions of various Bessel functions at small and large x:
r 
2 νπ π 
Jν (x) → cos x − − , x → ∞, (E2)
πx 2 4
r 
2 νπ π 
Nν (x) → sin x − − , x→∞ (E3)
πx 2 4
and
1  x ν
Jν (x) ≈ , x  1, (E4)
Γ(ν + 1) 2
( 2 x

π ln 2 + γ , ν = 0 ,
Nν (x) ≈  x  1. (E5)
2 ν
− Γ(ν)
π x , ν=6 0,

If constant k is imaginary, it is useful to introduce its stead a real constant κ as k = iκ. Then, in place of (E1) we
get the modified Bessel equation
 
00 1 0 ν2
R (x) + R (x) − 1 + 2 R(x) = 0 , (E6)
x x

where now x = κr. The two linearly independent solutions are the modified Bessel functions Iν (x) and Kν (x) which
(1)
are analytical continuations of Jν (x) and Hν (x) correspondingly. Kν (x) is also known as the McDonald function.
Their asymptotic forms are
1
Iν (x) → √ ex , x → ∞ , (E7)
2πx
r
π −x
Kν (x) → e , x→∞ (E8)
2x

and
1  x ν
Iν (x) ≈ , x  1, (E9)
Γ(ν + 1) 2
(
− ln x2 − γ , ν = 0 ,
Kν (x) ≈ x  1. (E10)
Γ(ν) 2 ν
π x , ν 6
= 0 ,

Spherical Bessel functions are defined as


r
π
jν (x) = Jn+1/2 (x) , (E11)
2x
r
π
nν (x) = Nn+1/2 (x) . (E12)
2x

223
F LIST OF BOOKS USED IN PREPARATION OF THESE NOTES

For integer ν = ` they can be related to trigonometric functions. For example


sin z cos z
j0 (z) = , n0 (z) = − . (E13)
z z
Asymptotic expressions
1
j` (x) ≈ sin(x − π`/2) , x  1 , (E14)
x
1
n` (x) ≈ − cos(x − π`/2) , x  1 . (E15)
x

x`
j` (x) ≈ , x  1. (E16)
(2` + 1)!!

Appendix F: List of books used in preparation of these notes

- L. Landau and E. Lifshitz, “Quantum Mechanics: Non-relativistic theory”.


- A. Messiah, “Quantum Mechanics”.
- A. Davydov, ”Quantum Mechanics”.
- S. Weinberg, ”Lectures on Quantum Mechanics”.
- E. Merzbacher, “Quantum Mechanics”, 3rd edition.
- V. Galitski, B. Karnakov and V. Kogan, “Problems in quantum mechanics”.
- K. Tamvakis, “Problems and Solutions in Quantum Mechanics”.
- S. Flugge, “Practical Quantum Mechanics”.
- L. I. Schiff, “Quantum mechanics”.

For sections XIV–XVI:

- Berestetskiy, Lifshitz, Pitaevskiy, “Quantum Electrodynamics”


- T. Ohlsson, “Relativistic Quantum Physics”
- W. Greiner “Relativistic Quantum Mechanics”
- M. Thomson “Modern Particle Physics”

224
F LIST OF BOOKS USED IN PREPARATION OF THESE NOTES

34. Clebsch-Gordan coefficients 010001-1

34. CLEBSCH-GORDAN COEFFICIENTS, SPHERICAL HARMONICS,


AND d FUNCTIONS
p J J ...
Note: A square-root sign is to be understood over every coefficient, e.g., for −8/15 read − 8/15. Notation:
M M ...
r m1 m2
1/2 × 1/2 1
+1 3
1 0 Y10 = cos θ 2× 1/2 + 5/2 m 1 m 2 Coefficients
+ 1/2 + 1/2 1 0 0 4π 5/2 5/2 3/2
r + 2 +1/2 1 + 3/2 + 3/2
. .
+ 1/2 − 1/2 1/2 1/2 1 . .
− 1/2 + 1/2 1/2 − 1/2 − 1
3
Y11 = − sin θ eiφ + 2 − 1/2 1/5 4/5 5/2 3/2 . .
− 1/2 − 1/2 1 8π + 1 + 1/2 4/5 − 1/5 + 1/2 + 1/2
r  
5 3 1 + 1 − 1/2 2/5 3/5 5/2 3/2
Y20 = cos2 θ − 0 + 1/2 3/5 − 2/5 − 1/2 − 1/2
1 × 1/2 3/2
+ 3/2 3/2 1/2
4π 2
r
2
0 − 1/2 3/5 2/5 5/2 3/2
+ 1 + 1/2 1 + 1/2 + 1/2
15 − 1 + 1/2 2/5 − 3/5 − 3/2 − 3/2
Y21 = − sin θ cos θ eiφ
+ 1 − 1/2 8π 2 − 1 − 1/2 4/5 1/5 5/2
1/3 2/3 3/2 1/2
r 3/2× 1/2 + 2 2 1 − 2 + 1/2 1/5 − 4/5 − 5/2
0 + 1/2 2/3 − 1/3 − 1/2 − 1/2
2 1 15 2 2iφ + 3/2 +1/2 1 + 1 + 1
0 − 1/2 2/3 1/3 3/2 Y2 = sin θ e − 2 − 1/2 1
− 1 + 1/2 1/3 − 2/3 − 3/2
4 2π + 3/2 − 1/2 1/4 3/4 2 1
+ 1/2 + 1/2 3/4 − 1/4 0 0
2 × 1 +3 3 2
3 − 1 − 1/2 1
3/2×1 5/2
+ 1/2 − 1/2 1/2 1/2 2 1
+ 5/2 5/2 3/2
+2 +1 1 +2 +2 + 3/2 + 1 1 + 3/2 + 3/2 − 1/2 + 1/2 1/2 − 1/2 −1 −1
+ 2 0 1/3 2/3 3 2 1 + 3/2 0 2/5 3/5 5/2 3/2 1/2 − 1/2 − 1/2 3/4 1/4 2
+ 1 + 1 2/3 −1/3 +1 +1 +1 + 1/2 + 1 3/5 − 2/5 + 1/2 + 1/2 + 1/2 − 3/2 + 1/2 1/4 − 3/4 − 2
+2 −1 1/15 1/3 3/5 + 3/2 − 1 1/10 2/5 1/2 − 3/2 − 1/2 1
1 × 1 + 22 2 1 + 10 + 01 8/15 1/6 − 3/10
2/5 − 1/2 1/10
3 2 1 + 1/2 0 3/5 1/15 − 1/3 5/2 3/2 1/2
0 0 0 − 1/2 + 1 3/10 − 8/15 1/6 − 1/2 − 1/2 − 1/2
+1 +1 1 +1 +1
+ 1 − 1 1/5 1/2 3/10 + 1/2 − 1 3/10 8/15 1/6
+ 1 0 1/2 1/2 2 1 0 0 0 3/5 0 − 2/5 3 2 1 − 1/2 0 3/5 − 1/15 − 1/3 5/2 3/2
0 + 1 1/2 − 1/2 0 0 0 − 1 + 1 1/5 − 1/2 3/10 −1 −1 −1 − 3/2 + 1 1/10 − 2/5 1/2 − 3/2 − 3/2
+ 1 − 1 1/6 1/2 1/3 0 − 1 2/5 1/2 1/10 − 1/2 − 1 3/5 2/5 5/2
0 0 2/3 0 − 1/3 2 1 − 1 0 8/15 − 1/6 − 3/10 3 2 − 3/2 0 2/5 − 3/5 − 5/2
− 1 + 1 1/6 − 1/2 1/3 − 1 −1 − 2 + 1 1/15 − 1/3 3/5 − 2 − 2 − 3/2 − 1 1
0 − 1 1/2 1/2 2 − 1 − 1 2/3 1/3 3
Y`−m = (−1)m Y`m∗ − 1 0 1/2 − 1/2 − 2 r − 2 0 1/3 − 2/3 − 3 hj1 j2 m1 m2 |j1 j2 JM i
−1 −1 1 4π −2 −1 1
d `m,0 = Y m e−imφ = (−1)J−j1 −j2 hj2 j1 m2 m1 |j2 j1 JM i
2` + 1 `
j 0 j j
d m0 ,m = (−1)m−m d m,m0 = d −m,−m0 3/2× 3/2 3
+3 3 1/2 θ 1 + cos θ
2 d 10,0 = cos θ d 1/2,1/2 = cos d 11,1 =
+ 3/2 + 3/2 1 +2 +2 2 2
2× 3/2 + 7/2 + 3/2 + 1/2 1/2 1/2 3 2 1 1/2 θ 1 = − sin θ
7/2 7/2 5/2
+ 1/2 + 3/2 1/2 − 1/2 + 1 + 1 +1 d 1/2,−1/2
= − sin d 1,0 √
+ 2 + 3/2 1 + 5/2 + 5/2 2 2
+ 3/2 − 1/2 1/5 1/2 3/10
+ 2 + 1/2 3/7 4/7 7/2 5/2 3/2 + 1/2 + 1/2 3/5 0 − 2/5 1 − cos θ
+ 1 + 3/2 4/7 − 3/7 + 3/2 + 3/2 + 3/2 − 1/2 + 3/2 1/5 − 1/2 3/10
3 2 1 0
d 11,−1 =
0 0 0 0 2
+ 2 − 1/2 1/7 16/35 2/5 + 3/2 − 3/2 1/20 1/4 9/20 1/4
+ 1 +1/2 4/7 1/35 − 2/5 7/2 5/2 3/2 1/2
+ 1/2 − 1/2 9/20 1/4 − 1/20 − 1/4
2×2 4
+4 4 3
0 +3/2 2/7 − 18/35 1/5 + 1/2 + 1/2 + 1/2 + 1/2 − 1/2 + 1/2 9/20 − 1/4 − 1/20 1/4 3 2 1
+ 2 − 3/2 1/35 6/35 2/5 2/5 − 3/2 + 3/2 1/20 − 1/4 9/20 − 1/4 − 1 − 1 −1
+2 +2 1 +3 +3
+ 1 − 1/2 12/35 5/14 0 − 3/10 + 1/2 − 3/2 1/5 1/2 3/10
+2 +1 1/2 1/2 4 3 2 0 +1/2 18/35 − 3/35 − 1/5
+1 +2 1/2 − 1/2 + 2 +2 +2
1/5 7/2 5/2 3/2 1/2 − 1/2 − 1/2 3/5 0 − 2/5 3 2
− 1 +3/2 4/35 − 27/70 2/5 − 1/10 − 1/2 − 1/2 − 1/2 − 1/2 − 3/2 + 1/2 1/5 − 1/2 3/10 − 2 − 2
+ 2 0 3/14 1/2 2/7 + 1 − 3/2 4/35 27/70 2/5 1/10
+ 1 +1 4/7 0 − 3/7 − 1/2 − 3/2 1/2 1/2 3
4 3 2 1 0 − 1/2 18/35 3/35 − 1/5 − 1/5
0 +2 3/14 − 1/2 2/7 +1 +1 +1 +1 − 1 +1/2 12/35 − 5/14 0 3/10 7/2 5/2 3/2
− 3/2 − 1/2 1/2 − 1/2 − 3
+ 2 − 1 1/14 3/10 3/7 1/5 − 2 +3/2 1/35 − 6/35 2/5 − 2/5 − 3/2 − 3/2 − 3/2 − 3/2 − 3/2 1
+ 1 0 3/7 1/5 − 1/14 − 3/10 0 − 3/2 2/7 18/35 1/5
0 +1 3/7 − 1/5 − 1/14 3/10 4 3 2 1 0 − 1 − 1/2 4/7 − 1/35 − 2/5 7/2 5/2
− 1 +2 1/14 − 3/10 3/7 − 1/5 0 0 0 0 0
− 2 + 1/2 1/7− 16/35 2/5 − 5/2 − 5/2
+ 2 − 2 1/70 1/10 2/7 2/5 1/5 − 1 − 3/2 4/7 3/7 7/2
+ 1 − 1 8/35 2/5 1/14 − 1/10 − 1/5 − 2 − 1/2 3/7 − 4/7 − 7/2
0 0 18/35 0 − 2/7 0 1/5
− 1 +1 8/35 − 2/5 1/14 1/10 − 1/5 4 3 2 1 − 2 − 3/2 1
3/2 1 + cos θ θ − 2 +2 1/70 − 1/10 2/7 − 2/5 1/5 −1 −1 −1 −1
d 3/2,3/2 = cos
2 2 + 1 − 2 1/14 3/10 3/7 1/5
√ 1 + cos θ  1 + cos θ 2 0 − 1 3/7 1/5 − 1/14 − 3/10
3/2 θ d 22,2 = − 1 0 3/7 − 1/5 − 1/14 3/10 4 3 2
d 3/2,1/2 =− 3 sin 2
2 2 − 2 +1 1/14 − 3/10 3/7 − 1/5 −2 −2 −2
√ 1 − cos θ θ 1 + cos θ 0 − 2 3/14 1/2 2/7
3/2
d 3/2,−1/2 = 3 cos d 22,1 = − sin θ − 1 − 1 4/7 0 − 3/7 4 3
2 2 2
√ 1 + cos θ − 2 0 3/14 − 1/2 2/7 − 3 − 3
3/2 1 − cos θ θ 6 d 21,1 = (2 cos θ − 1)
d 3/2,−3/2 = − sin d 22,0 = sin2 θ − 1 − 2 1/2 1/2 4
2 2 4 2 − 2 − 1 1/2 − 1/2 − 4
r
3/2 3 cos θ − 1 θ 1 − cos θ 3 −2 −2
d 1/2,1/2 = cos d 22,−1 = − sin θ d 21,0 = − sin θ cos θ 1
2 2 2 2
3 cos θ + 1 θ  1 − cos θ 2 1 − cos θ 3 1
3/2
d 1/2,−1/2 = − sin d 22,−2 = d 21,−1 = (2 cos θ + 1) d 20,0 = cos2 θ −
2 2 2 2 2 2
Figure 34.1: The sign convention is that of Wigner (Group Theory, Academic Press, New York, 1959), also used by Condon and Shortley (The
Theory of Atomic Spectra, Cambridge Univ. Press, New York, 1953), Rose (Elementary Theory of Angular Momentum, Wiley, New York, 1957),
and Cohen (Tables of the Clebsch-Gordan Coefficients, North American Rockwell Science Center, Thousand Oaks, Calif., 1974). The coefficients
here have been calculated using computer programs written independently by Cohen and at LBNL.

225

Anda mungkin juga menyukai