PHYS 4011, 5050:

Atomic and Molecular Physics

Lecture Notes

Tom Kirchner1

Department of Physics and Astronomy

York University
April 7, 2013

1 [email protected]

Contents
1 Introduction: the field-free Schro
dinger hydrogen atom
1.1 Reduction to an effective one-body problem . . . . . . . .
1.2 The central-field problem for the relative motion . . . . .
1.3 Solution of the Coulomb problem . . . . . . . . . . . . . .
1.4 Assorted remarks . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.

.
.
.
.

2
2
4
5
13

2 Atoms in electric fields: the Stark effect

16
2.1 Stationary perturbation theory for nondegenerate systems . . 17
2.2 Degenerate perturbation theory . . . . . . . . . . . . . . . . . 22
2.3 Electric field effects on excited states: the linear Stark effect . 24
3 Interaction of atoms with radiation
3.1 The semiclassical Hamiltonian . . . .
3.2 Time-dependent perturbation theory
3.3 Photoionization . . . . . . . . . . . .
3.4 Outlook on field quantization . . . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

29
29
32
41
44

4 Brief introduction to relativistic QM

58
4.1 Klein-Gordon equation . . . . . . . . . . . . . . . . . . . . . . 58
4.2 Dirac equation . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5 Molecules
70
5.1 The adiabatic (Born-Oppenheimer) approximation . . . . . . . 71
5.2 Nuclear wave equation: rotations and vibrations . . . . . . . . 74
5.3 The hydrogen molecular ion H+
2 . . . . . . . . . . . . . . . . . 77

Chapter 1
Introduction: the field-free
Schr
odinger hydrogen atom
The starting point of the discussion is the stationary Schrodinger equation
= E
H

(1.1)

for the two-body problem consisting of a nucleus (n) and an electron (e).
The Hamiltonian reads
2e
2n
p
Ze2
p

H=
2mn 2me 40 |re rn |

(1.2)

with
mn = N 1836me ;

me 9.1 1031 kg

and N being the number of nucleons (N = 1 for the hydrogen atom itself,
where the nucleus is a single proton).
The first step is to separate this two-body problem into two effective
one-body problems.

1.1

Reduction to an effective one-body problem

Consider the (classical) coordinate transformation

(re , pe , rn , pn )

(R, P, r, p)

3
definitions :

M = me + mn mn
me mn
me
=
me + mn

R = mn rnM+me re
rn

center of mass
motion
pn
P = pe + pn = M R

r = re rn

relative
motion

me pn
pe
p = r = mn peM

QM transformation analogously
e , rn , p
n)
(re , p

P,
r, p
)
(R,

(1.3)

insertion into Eq. (1.2) yields

2
2
P
p
+
+ V (r) ,
2M
2
CM + H
rel
= H

=
H

Ze2 
V (r) =
40 r

(1.4)

Eq. (1.4) is the Hamiltonian of a non-interacting two-(quasi-)particle

system can be separated into two one-particle problems:
ansatz :

(r, R) = CM (R)rel(r)

(1.5)

Schrodinger equations (SEs)

2
CM CM (R) = ~ 2 CM (R) = ECM CM (R)
(1.6)
H
R
 2M2

~ 2
Ze2

Hrel rel (r) =

r
rel (r) = Erel rel (r) (1.7)
2
40 r
with
E = ECM + Erel
(1.8)

Equation (1.6) can be solved without difficulty:

CM (R) = AeiKR

free particle
1
K = ~P
motion

2 K2

ECM = ~2M

4
Equation (1.7) can be solved analytically, but before we sketch the
solution we consider some general properties/features of the quantum
central-field (V (r) = V (r)) problem.

1.2

The central-field problem for the relative

motion
2
rel = p
H
+ V (r)
2

Consider

(1.9)

One can show that Hrel is invariant with respect to rotations, and therefore
commutes with the angular momentum operator
l = r p
.

(1.10)

This is a manifestation of angular momentum conservation. In particular,

rel , l2 , lz form a complete set of compatible operators, i.e.,
the operators H
rel , l2 ] = [H
rel , lz ] = [l2 , lz ] = 0
[H

(1.11)

they have a common set of eigenstates. The eigenstates of l2 , lz are the

spherical harmonics Ylm .

ansatz

rel (r) = Rl (r)Ylm (, )

insertion into Eq. (1.7) for Hamiltonian (1.9) yields the radial SE
(
)
p2r
~2 l(l + 1)
+
+ V (r) Erel Rl (r) = 0
2
2r 2

(1.12)

(1.13)

~2
r (r 2 r )
r2
l2
= p2r + 2
r

with

p2r =

and operator identity

2
p

(can be proven in spherical coordinates in coordinate space).

Useful definition :

yl (r) = rRl (r)

(1.14)

(1.13)

yl(r)

h
l(l + 1) i
yl (r) = 0
+ U(r)
r2


~2
,
Erel =
2

1.3

V (r) =

(1.15)


~2
U(r)
2

Solution of the Coulomb problem

The radial Eq. (1.15) is very similar to the one-dimensional SE. There are,
however, two important differences. First, the total (effective) potential consists of two parts
Ulef f (r) = U(r) +

l(l + 1) r
0
r2
angular momentum barrier

(cf. classical central-field problem) Second, the boundary conditions are different.
a) Boundary conditions
r 0
|rel (r)|2 = |Rl (r)|2|Ylm (, )|2 <
in particular for r = 0
regularity condition
yl (0) = 0

(1.16)

r
1. Erel < 0 (bound spectrum)
Z
Z
Z
2 3
2 2
|rel (r)| d r =
r Rl (r) dr |Ylm (, )|2 d
Z0
=
yl2 (r) dr <
0

(square integrable solutions required)

yl (r) 0

( strong boundary condition)

6
2. Erel > 0 (continuous spectrum)
oscillatory solutions yl (r) for r


note:

for Erel > 0 the solution leads to Rutherfords scattering formula

(which is identical in classical mechanics and QM)

b) Bound-state solutions
2 = > 0
40 ~2
0.53 1010 m
a =
e2
me , a a0
is the Bohr radius

definition :

for
radial Eq. (1.15):

!
d2
l(l + 1) 2Z

+
2 yl (r) = 0
dr 2
r2
ar

(1.17)

transformation: x = 2r

d2
1 d2
=
dx2
42 dr 2

!
l(l + 1) 1
d2
yl (x) = 0

+
dx2
x2
x 4

asymptotic solutions:
1. x

because of yl (x ) = 0

!
d2
1
yl (x) = 0

dx2 4
x

yl (x) = Ae 2 + Be 2
B = 0

(1.18)

7
2. x 0

!
l(l + 1)
d2
yl (x) = 0

dx2
x2

because of yl (0) = 0

yl (x) =

A
+ Bxl+1
xl

A = 0 .

This consideration motivates the following ansatz

x

yl (x) = xl+1 e 2 vl (x).

(1.19)

Insertion into Eq. (1.18) yields a new differential equation for vl (x):
(
)
d2
d

x 2 + (2l + 2 x) (l + 1 ) vl (x) = 0.
(1.20)
dx
dx
The square integrable solution of (1.20) (Kummers or Laplaces differential eq.) are known; they are the associated Laguerre polynomials:
Lkp (x)


2
(p + k)!
()
=
xj .
(p

j)!(k
+
j)!j!
j=0
p
X

More specifically:
vl (x) =

L2l+1
nl1 (x)

nr = n l 1 0
n 1 l

Z
,
n = 1, 2, ...
(1.21)
n a
The detailed treatment shows that the integrability of the solutions requires
Z
= a
to be positive, integer numbers quantization of (i.e. quantization of the energy)1
with

n n =

2l+1
ynl (r) = Anl r l+1en r Lnl1
(2n r)

1
finds the square integrable solutions of (1.20) explicitly by using the ansatz vl (x) =
P One
l i
i bi x and by taking the boundary (and regularity) conditions into account.

8
and properly normalized wave functions take the form
rel (r) nlm (r) =

(n l 1)! l+ 1 l+2
a

3 2 2 n
(n + l)!

r l en r L2l+1
nl1 (2n r)Ylm (, )

(1.22)

n0
l n1
l m l

Rnl (r)Ylm (, )
The quantization condition (1.21) yields

Erel En

~2 Z 2
=
2a2 n2

n = 1, 2, ...

(1.23)

Z2
13.6 eV 2 .
n

(1.24)

The lowest-lying hydrogen eigenfunctions (orbitals):

n

nr = n l 1

1s

2s

2p0

1 1

2p1

with R = 13.6 eV

nlm (r)
  23
1

Z
a

Zr

e a
  32 
 Zr
Z
Zr
1
2

e 2a
a
4 2 a
  32   Zr
Z
Zr
1
e 2a cos
a
4 2 a
  23   Zr
Zr
81 Za
e 2a sin ei
a

En
RZ 2
RZ 2
4
RZ 2
4
RZ 2
4

s
E

d
p

0
n = 3
n = 2

n = 1

3 s

3 p

-1

3 p

2 s

2 p

-1

2 p

3 p
0
0

2 p

3 d
1

-2

3 d

-1

3 d
0

3 d
1

3 d
2

1 s

Figure 1.1: Energy spectrum of the Schrodinger-Coulomb problem. Note that

n
the Coulomb potential supports infinitely many bound states (En 0).
Degeneracy of energies (which depend only on n)
given
given

n
l

l = 0, 1, ..., n 1
m = l, ..., l

n1
X
(2l + 1) = n2
l=0

each energy level En is n2 -fold degenerate. Note that all central-field

problems share the (2l + 1)-fold degeneracy which originates from rotational
invariance. The fact that the energies do not depend on nr , l separately, but
only on n = nr + l + 1 is specific to the Coulomb problem (one names n
the principal quantum number and nr the radial quantum number, which
determines the number of nodes in the radial wave functions).

10
In QM, the wave functions themselves are (usually) not observable, but
their absolute squares are
2
nlm (r) = |nlm(r)|2 = Rnl
|Ylm (, )|2
y 2 (r)
= nl 2 |Ylm (, )|2 .
r

(1.25)

R
If nlm (r)d3 r = 1 one interpretes nlm (r)d3 r as the probability to find the
electron in the volume [r, r + dr]. For spherically symmetric potentials it is
useful to also define a radial probability density by
Z
2 2
nl (r) = r Rnl (r) |Ylm (, )|2 d
2
= ynl
(r)

(1.26)

nl (r)dr is the probability to find the electron in the interval [r, r + dr]
Momentum space representation
So far we have worked in coordinate space, in which states are wave functions (r). It is also possible and insightful to look at the problem
in another, e.g., the momentum space representation, which is connected to
the coordinate space representation by a (three-dimensional) Fourier transformation. Using the Dirac notation one can obtain momentum space wave
functions by considering
Z

nlm (p) = hp|nlmi = hp|rihr|nlmid3r

Z
i
1
=
e ~ pr nlm (r)d3 r.
3
[2~] 2

(1.27)

To work out the three-dimensional Fourier transform one uses the expansion of a plane wave in spherical coordinates
ikr

= 4

X
L
X

(i)L jL (kr)YLM (k )YLM

(r )

L=0 M =L

(1.28)

11
with p = ~k and the spherical Bessel functions jL .
1

X
L
X

r 2 jL (kr)Rnl (r)dr
[2~] L=0 M =L
0
Z

YLM
(r )Ylm
(r )dr YLM (k )
Z
1
l
r 2 jl (kr)Rnl (r)dr Ylm (k )
= 4
3 (i)
[2~] 2
0
=: Pnl (p)Ylm (k ).
(1.29)

nlm (p) = 4

3
2

(i)

Probability densities can be defined in the same way as in coordinate space

nlm (p) = |nlm (p)|2
2
nl (p) = p2 Pnl
(p).

(1.30)

Figure 1.2: Radial hydrogen 1s, 2s, 2p wave functions (blue) and probability
densities (red) in coordinate space.

The maximum of the 2p probability density is shifted to smaller r compared to the 2s state. We can understand this qualitatively in the following
way. Both states 2s,2p correspond to the same eigenenergy. The 2s state has
a contribution at small r (the first lobe), for which the potential energy is
rather large as the nucleus is close. By contrast, the 2p state approaches 0
for small distances (remember the angular momentum barrier: only s states

12
do not approach zero for r 0). To compensate for the stronger binding
energy of the 2s state at small r the 2p state has to have its only maximum
at smaller r compared to the second maximum of the 2s state.

Figure 1.3: Radial hydrogen 3s, 3p, 3d wave functions (blue) and probability
densities (red) in coordinate space.

Figure 1.4: Radial hydrogen 1s, 2s, 2p wave functions (blue) and probability
densities (red) in momentum space.

The number of nodes in momentum space and coordinate space is the

same. Note that the momentum profile of 2s is restricted to rather small
momenta. Loosely speaking, the inner lobe of the momentum distribution
corresponds to the outer lobe of the distribution in coordinate space: When
the electron is far away from the nucleus the momentum is relatively small

13
(and vice versa). This phenomenon is related to the uncertainty principle.
The inner lobe in coordinate space is rather sharp, whereas the outer lobe in
momentum space extends over a relatively wide range of momenta (and v.v.).
But note that this is not a rigorous argument, because the radial momentum
is NOT the canonical momentum of the radial coordinate, i.e., they do not
fulfill standard commutation and uncertainty relations.

Figure 1.5: Radial hydrogen 3s, 3p, 3d wave functions (blue) and probability
densities (red) in momentum space.

Check the Maple file hydrogen.mw for more details!

1.4

Assorted remarks

a) More (mathematical) details about the Coulomb problem can be found

in any QM textbook, in particular in [Gri], Chap. 4.14.3.
b) Hydrogen-like ions
We have solved not only the (Schrodinger) hydrogen problem (Z = 1),
but also the bound-state problems of all one-electron atomic ions (e.g.
He+ , Li2+ ,...) for Z = 2, 3, ... Note that En Z 2 .
c) Exotic systems
... are also solved
(a) positronium (e+ e )
(b) muonium (+ e )
(c) muonic atom (p )

14
In these cases one has to take care of the different masses compared to
m2
.
the hydrogen problem. Note that En = mm11+m
2
d) Corrections
The spectrum determined by Eq. (1.23) is the exact solution of the
Schrodinger-Coulomb problem, but not exactly what one sees experimentally. The reason is that the Schrodinger equation is not the ultimate answer, e.g., it has to be modified to meet the requirements of
the theory of special relativity. Therefore, corrections show up, which
lead to a (partial) lifting of the degeneracy. This will be discussed later
on.
e) Atomic units
So far, we have used SI units (as we are supposed to). In atomic and
molecular physics another set of units is more convenient and widely
used: atomic units. The starting point for their definition is the Hamiltonian
~2 2
e2

HSI =

(1.31)
2me r 40 r
(i.e., the one of Eq. (1.7) for Z = 1 and me ). Four constants show
up in this Hamiltonian way too many and so they are all made
to disappear!
Recipe
measure mass in units of me

measure charge in units of e

measure angular momentum in units of ~

measure permittivity of the vacuum in units of 40

In other words, atomic units (a.u.) are defined by setting me = e =

~ = 40 = 1.
Consequences
a.u. = 1 2
H
2 r

1
r

15
length: lets look at Bohrs radius
a0 =

40 ~2
= 0.53 1010 m = 1a.u.
me e2

energy: lets look at the hydrogen ground state H(1s)

E1s =

~2
= 13.6 eV = 0.5 a.u. = 0.5 hartree = 1 Rydberg
2me a20

time: lets do a dimensional analysis

time =
t0 :=

distance
distance mass
distance2 mass
=
=
speed
momentum
angular momentum

a20 me
~

= 2.4 1017 s = 1 a.u.

fine structure constant (dimensionless)

e2
1
=

.
40 ~c
137
In atomic units we have = 1/c, i.e. c 137 a.u. Thus, one
atomic unit of velocity corresponds to 2.2 106 m/s. This is also
obtained by using v0 = a0 /t0 .

Chapter 2
Atoms in electric fields: the
Stark effect
What happens if we place an atom in a uniform electric field? One observes
a splitting and shifting of energy levels (spectral lines). This was first discovered by Johannes Stark in 1913, i.e. in the same year, in which Bohr
developed his model of the hydrogen atom. Later on, this problem was one
of the first treated by Schrodinger shortly after the discovery of his wave
equation. Schrodinger used perturbation theory, and this is what we will do
in this chapter.
The first step is to figure out what kind of modification a classical electric
field brings about. Lets assume the field E is oriented in positive z-direction:
E = F k

(2.1)

The associated electrostatic potential reads

(r) = E r = F z

(2.2)

potential energy of an electron

W (r) = e(r) = eF z

(2.3)

has to be added to the Hamiltonian. The task then is to solve

i = E | i
H|

(2.4)

for (using atomic units from now on)

= 1 2 1 + F z
H
2
r
16

(2.5)

17
Taking a look at the total potential (Coulomb + Stark) one finds that
tunnelling is possible i.e. in a strict sense the Hamiltonian (2.5) does not
support stationary states. Eventually, a bound electron will tunnel through
the barrier and escape from the atom. In practice, however, the Stark potential is weak compared to the Coulomb potential and the tunnel effect is
unimportant unless one studies highly-excited states. This is why we can
apply stationary perturbation theory (PT).

2.1

Stationary perturbation theory for nondegenerate systems

a) General formalism
Task: solve stationary SE
i = E | i
H|

(2.6)

decompose

=H
0 + W

H
0 is known
and assume that the eigenvalue problem of H

(2.7)

0 |0 i = E (0) |0 i,
H

(2.8)

h0 |0 i =

We seek solutions of. Eq. (2.6) in terms of a Taylor (like) expansion based on
the (nondegenerate) eigenvalues and eigenstates of the unperturbed problem
be small. Lets
Eq. (2.8). Therefore, we require that the perturbation W
introduce a smallness parameter :

(2.6)

w
W

with 1


0 + w
H
| ()i = E ()| ()i

(2.9)
(2.10)

Taylor expansions about = 0:



dE () 
1 d2 E () 
E () =
+
2 + . . .
+
d =0
2 d2 =0
d
| ()i|=0 + . . .
| ()i = |0 i +
d
E(0)

(2.11)
(2.12)

18
We need to find expressions for the derivatives in Eqs. (2.11), (2.12):
consider derivative of. Eq. (2.10):

d 
H 0 + w
E () | ()i = 0
d



0 + w E () | ()i + w E () | ()i = 0
H

dE
etc.)
d

h ()|H()
E ()| ()i + h ()|w E ()| ()i = 0
(E =

i) =
=

E () = h ()|w|
()i

(2.13)

ii) 6=
=

h ()| ()i =

h ()|w|
()i
E () E ()

(2.14)

In order
the orthonormal
 to use
Eq. (2.14) for an expansion of | i in terms of
basis | i we have to consider the coefficient h ()| ()i in addition.
If we assume that h ()|()i = h ()| ()i (i.e. we choose real states
which is not a restriction) we can show that

h ()| ()i = 0
d
h ()| ()i = h ()| ()i + h ()| ()i
{z
}
d |

proof :

=1

= 2h ()|()i = 0

| ()i =
=

| ()ih ()| ()i

X h ()|w|
()i
6=

E () E ()

| ()i

(2.15)

19
Lets also consider the 2nd derivative term in Eq. (2.11):
d2
E ()
d2

d
(2.13) d
E () =
h ()|w|
()i
d
d
h ()|w|
()i + h ()|w|
()i
X |h ()|w|
()i|2
2
E () E ()
6=

=
=
(2.15)

(2.16)

Higher order terms can be calculated by differentiating expressions (2.15),

(2.16) successively. We stop here and insert (2.13)(2.16) in the Taylor expansions (2.11), (2.12):
E () = E(0) + h (0)|w|
(0)i
X |h (0)|w|
(0)i|2
+ ...
+ 2
E
(0) E (0)
6=
=
| ()i =

E(0)
|0 i

|0 i
h0 |W

|0 i|2
X |h0|W

(0)

6=

|0 i
X h0 |W

(0)
6= E

(0)
E

(0)

E E

|0 i + . . .

+ . . . (2.17)
(2.18)

Eqs. (2.17), (2.18) are the standard expressions for the lowest-order corrections the glorious result of this section!
Remarks:
(0)

(0)

1. Derivation and result are valid only if E 6= E (i.e. no degeneracies)

2. Convergence of perturbation series?
This cannot be answered in general. In some cases, perturbation expansions do converge, in some they do not, and in some other cases the
perturbation series turns out to be a so-called semi-convergent (asymptotic) series.
Consistency criterion for convergence (cf. Eq. (2.18))


 h0 |W
0 

|
i


(for 6= )
 (0)
1,
(0)
 E E 

20
3. In practice, exact calculations beyond 1st order in the energy are often
not feasible due to (infinite) sums over all basis states (cf. Eq. (2.17)).
4. Literature: [Gri], Chap. 6.1; [Lib], Chap. 13.1
b) Application to H(1s) in an electric field
Ingredients (in atomic units):
1
01s (r) = er

(0)

E1s = 0.5 a.u.

W = Fz
1st -order energy correction:
(1)

E1s

= h01s |W |01s i
Z
F
=
e2r r cos d3 r

Z
Z 1
Z 2
F 3 2r
=
r e dr
cos d cos
d
0
1
0
1
Z

= F
r 3 e2r drx2  = 0
0

(with x = cos )

(2.19)

(2.20)

The 2nd -order energy correction

(2)

E1s =

X |h01s |F z|0 i|2

(0)

6=1s

(0)

E1s E

is hard to calculate due to the infinite sum (which actually also involves
an integral over the continuum states). Let us content ourselves with an
estimate.
(0)

(0)

(2)

Note that E1s E < 0, i.e., E1s < 0.

21
Consider
(2)
|E1s |

= F
=

X |h01s |z|0 i|2

(0)
6=1s E

F2

(0)
En=2
2

(0)
E1s

(0)
E1s

h01s |z

<

F2
(0)
En=2

(0)
E1s 6=1s

h01s |z|0 ih0 |z|01s i

|0 ih0 |z|01s i h01s |z|01s ih01s |z|01s i

2
8F
(2.20) 8F
(h01s |z 2 |01s i h01s |z|01s i2 ) =
h01s |z 2 |01s i
3
3

One finds h01s |z 2 |01s i = 1 and obtains

8
(2)
|E1s | < F 2
3

quadratic Stark effect

(2.21)

The exact result (see [Sha], Chap. 17) is

9
(2)
E1s = F 2
4

(2.22)

Interpretation:
Consider a classical charge distribution in an electric field. The associated
potential energy is
Z
Z
3
W =
(r)(r)d r = F (r)zd3 r
= F dz

where dz is the z-component of the dipole moment.

Link this to QM by recognizing that (r) = |(r)|2 (atomic units!).
Z
dz = |1s (r)|2 zd3 r = h1s |z|1s i
Now use Eq. (2.18) to obtain
dz = 2F

X |h01s |z|0 i|2

(0)

6=1s

(0)

E1s E

2
16
(2)
+ O(2 ) E1s < F
F
3

22
Summary:
(1)

E1s = 0 expresses the fact that the unperturbed hydrogen ground

(0)
state has no static dipole moment dz (this is in fact true for any
spherically symmetric charge distribution).
(2)

E1s 6= 0 expresses the fact that it has a nonzero induced dipole

(1)
(1)
moment dz , i.e., a nonzero dipole polarizability D := dz /F . We
have found D < 16/3 a.u. (the exact result being D = 9/2 a.u. )

2.2

Degenerate perturbation theory

Problem:
with

=H
0 + w
H
0 |0 i = E (0) |0 i,
H
j

j = 1, . . . , g

(2.23)
(2.24)

where g is the degeneracy level. The set {|0j i, j = 1, . . . , g } spans a g (0)

dimensional subspace of Hilbert space associated with the eigenvalue E .
0
This implies that any linear combination of these states is an eigenstate of H
(0)
for E . When the perturbation is turned on, the degeneracy is (normally)
lifted:
j i = Ej |j i.
H|
(2.25)
The question arises which of the degenrate states is approached by a given
state |j i in the limit 0. At this point we cant say more than that it
can be any linear combination, i.e.
0

|j i |0j i =

g
X
k=1

akj |0k i.

(2.26)

The unknown states |0j i are the 0th -order states of the pertubation expansion:
(1)

Ej = E(0) + Ej +

|j i = |0j i + |1j i +

(2.27)
(2.28)

23
Note that this expansion is of the same type as the previous Taylor series of
Eqs. (2.11), (2.12) if one identifies

dE () 
(1)
(1)
etc.
E = E =
d 
=0

Now proceed as follows: Insert (2.27), (2.28) into the Schrodinger equation
(2.25) to obtain


0 + w)
(H
|0j i + |1j i +


(1)
= (E(0) + Ej + ) |0j i + |1j i + ,
sort this in terms of powers of
0 |0 i = E (0) |0 i
0 :
H
j

j
(1)
1
1
0

:
H0 |j i + w|
j i = E(0) |1j i + Ej |0j i,

and project the second equation onto an undisturbed eigenstate of the same
subspace:
(1)
0 E (0) |1 i + h0 |w
0 = h0 |H
E |0 i
l

0 =

0 =

j
l
(1) 0
0
hl |w Ej |j i
g
X
(1)
h0l |w Ej |0k iakj
k=1

l = 1, . . . , g

(2.29)

Eq. (2.29) is a standard matrix eigenvalue problem of size g g . Its solution is what (lowest-order) degenerate perturbation theory boils down to in
practice. Here is how to do it:
(1)

(i) Solve secular (characteristic) equation det(w E 1) = 0 obtain

(1)
eigenvalues {Ej , j = 1, . . . , g}, which happen to be the first-order
energy corrections, i.e. the quantities we are after!
(ii) Find mixing coefficients {akj , k, j = 1, . . . , g } by inserting eigenvalues
into the matrix equations (2.29).
(1)

(ii) Check that h0l |w|

0k i = Ej lk (i.e. perturbation matrix is diagonal
with respect to the states {|0k i}).
Note that higher-order calculations are possible, but tedious and not discussed in our textbooks. Refs. [Gri], Chap. 6.2 and [Lib], Chap. 13.2, (13.3)
merely paraphrase the material provided in this section.

24

2.3

Electric field effects on excited states: the

linear Stark effect

Let us now apply degenerate perturbation theory to the problem of the excited hydrogen states in a uniform electric field. To this end we have to set
up and diagonalize the perturbation matrix. Lets first look at
a) Matrix elements and selection rules

wlk
= h0l |w|
0k i

explicitly 0
hnlm|z|0nl m i

with (cf. Eq. (1.22))

F
(smallness parameter)
nlm (r) = Rnl (r)Ylm ()
r
4
rY10 ()
z = r cos =
3
r Z
Z
4
0
0

3
hnlm |z|nl m i =
()Y10 ()Yl m ()d
r Rnl (r)Rnl (r)dr Ylm
3 0

The angular integral is a special case of a more general integral over three (arbitrary) spherical harmonics, the result of which is known (Wigner-Eckart
theorem):
r
Z
4

Ylm
()YLM ()Yl m ()d
2L + 1



p
l L l
l
L l
m

(2.30)
(2l + 1)(2l + 1)
= (1)
0 0 0
m M m

with

j1 j2 j3
m1 m2 m3

Wigners 3j-symbol

The 3j-symbols are closely related to the Clebsch-Gordan coefficients. We

dont need to know much detail here other than that they fulfill certain
selection rules, i.e., they are zero in many cases:

25


j1 j2 j3
6= 0 iff m1 + m2 + m3 = 0 |j1 j2 | j3 j1 + j2
m1 m2 m3


j1 j2 j3
6
=
0 iff j1 + j2 + j3 = even |j1 j2 | j3 j1 + j2
0 0 0

These relations imply that Eq. (2.30) is nonzero only if

m = m

and l = l l = 1

(2.31)

These conditions are called electric dipole selection rules.

Literature on angular momentum, Clebsch-Gordan coefficients, and 3j-symbols:
[Lib], Chap. 9; [Mes], Vol II, Chap. 13 (and appendix); [CT], Chaps. 6, 10
(and complements)
b) Linear Stark effect for H(n = 2)
Lets do an explicit calculation for the four degenerate states of the hydrogen
L shell: {02s , 02p0 02p1 02p+1 }
matrix elements:
due to the dipole selection rules (2.31) there is only one nonzero matrix
element we need to calculate:
Z
Z 1
Z 2
1
0
0
4 r
2
w12 = h2s |r cos |2p0 i =
(2 r)r e dr
cos d cos
d
32 0
1
0
 Z

Z
1
1
4 r
5 r
2
r e dr
r e dr = (2 4! 5!) = 3 (a.u.)
=
24
24
0
0
= w21
perturbation matrix:

0 w12
w12 0
w=
0
0
0
0

0
0
0
0

0
0

0
0

26
secular equation:

E (1) w12
0
0
w12 E (1)
0
0
det
(1)
0
0
E
0
0
0
0
E (1)

= 0

(2.32)

2
(E (1) )4 (E (1) )2 w12
= 0
(1)
E
= {0, 0, w12, w12 }

mixing coefficients
(1)
(i) E1,2 = 0:

0 w12
w12 0

0
0
0
0

(1,2)
a2s
0
(1,2)

a2p0
0

0 a(1,2)
2p1
(1,2)
0
a2p+1

0
0
0
0

(1,2)

a2s

=
0

(1,2)

= a2p0

(1,2)

= 0

(1,2)

|0E (1) i = a2p1 |02p1 i + a2p+1 |02p+1 i

1,2

We can choose the nonzero coefficients as we please so lets pick

|0E (1) i |02p1 i
1

|0E (1) i
2

|02p+1 i

(1)

(ii) E3 = w12 :

w12 w12
w12 w12

0
0
0
0
(3)

0
0
0
0

(3)
a2s
0
(3)

a2p0
0

0 a(3)
2p1
(3)
0
a2p+1

(3)

(3)

=
0

(3)

a2s = a2p0 , a2p1 = a2p+1 = 0

27
(1)

(iii) E4 = w12 :

w12 w12
w12 w12

0
0
0
0
(4)

(4)
a2s
0 0
(4)

a2p0
0 0

0 0 a(4)
2p1
(4)
0 0
a2p+1
(4)

(4)

=
0

(4)

a2s = a2p0 , a2p1 = a2p+1 = 0

normalized Stark states:

1
|0E (1) i = |02s i + |02p0 i
3
2


1
|0E (1) i = |02s i |02p0 i
4
2

(2.33)
(2.34)

Summary
1. The (weak) electric field results in a splitting of the energy level the
degeneracy is (partly) lifted. The energy shifts are linear in the electric
field strength:
(1)

E1,2 = 0
(1)

E3

(1)

E4

= w12 = 3F

= w12 = 3F

2. Note that the original L-shell states have no static dipole moment
since
h0nlm |z|0nlm i = 0
3. The Stark states (2.33), (2.34) do have nonzero static dipole moments
(calculate them!). Check the Maple worksheet starkstates.mw to see
how these states look like.
4. Cylindrical symmetry is preserved by the Stark potential W = F z,
since
[lz , W ] = 0,
i.e., m is still a good quantum number, but l is not.

28
5. The diagonalization procedure can be simplified by recognizing that the
perturbation matrix is of block-diagonal structure (with three blocks
corresponding to the magnetic quantum numbers m = 0, m = 1, m =
+1. Consider three blocks Ai :

A1 0 0
det 0 A2 0 = 0
0 0 A3
detA = 0 detAi = 0 for i = 1, . . . , 3
The only nontrivial secular equation for the L-shell problem then is (cf.
Eq. (2.32)):


E (1) w12
= 0
det
w12 E (1)
In a similar fashion, one can study the Stark problem for the nine
M shell states. The perturbation matrix can be decomposed into five
blocks corresponding to the states with magnetic quantum numbers
m = 2 to m = 2.

Chapter 3
Interaction of atoms with
radiation
The first question to be addressed when discussing the interaction of atoms
with radiation concerns the level on which the electromagnetic (EM) field
shall be described. It seems natural to aim at a quantum theory. It is only
on this level that photons come into the picture. We will take a look at them
a bit later, but start off with coupling the classical EM field to our quantal
description of the (hydrogen) atom. It turns out that this is sufficient for the
description of a number of processes including the photoelectric effect, which
prompted Einstein to introduce the notion of photons in the first place.

3.1

The semiclassical Hamiltonian

The goal is to derive a Hamiltonian, which accounts for the interaction of an

atom with a classical EM field. Lets start completely classically.
a) Classical particle in an EM field
The action of a classical EM field on a classical particle is described by
the Lorentzian force
FL = q E + (v B)

where q and v are the charge and the velocity of the particle and E and
B the electric and magnetic fields. However, if we want to construct a
Hamiltonian we need EM potentials instead of EM fields:
29

30
EM potentials
the scalar potential and the vector potential A are defined via
B = A
E =

(3.1)
A
t



A
+ (v ( A))
FL = q
t

(3.2)

(3.3)

this equation can be rewritten by introducing a generalized (i.e., velocitydependent) potential energy:
Generalized potential energy
U := q( A v)

(3.4)

one can show that Eq. (3.3) can be written as

FL = U +

d
v U
dt

(3.5)

i.e.,

 U
d U 

i = 1, 2, 3
FLi =
xi dt x i
the generalized potential paves the way to set up the Lagrangian:

Lagrangian

m 2
v q + qA v
(3.6)
2
if one works out the Lagrangian equations of motion one finds ma = FL ,
i.e., Newtons equation of motion with the Lorentzian force. This shows
that the construction is consistent. Now we are only one step away from
the Hamiltonian:
L= T U =

Hamiltonian
H =pvL

(3.7)

31
with the generalized momentum
p = v L = mv + qA
1
v =
(p qA)
m

(3.8)
(3.9)

if one uses Eq. (3.9) in Eq. (3.7) one arrives at

H=

1
(p qA)2 + q
2m

(3.10)

for details see [GPS], Chaps. 1.5 and 8.1

b) Quantum mechanical Hamiltonian for an electron
add a scalar potential V (to account, e.g., for the Coulomb potential
of the atomic nucleus)
q = e
= i~
quantization: p p
1
(
p + eA)2 e + V
2m
1
=
(~2 2 i~e A(r, t) i~eA(r, t) + e2 A2 (r, t))
2m
e(r, t) + V (r)
~2 2
e~
e~
=
+ V (r) +
A(r, t) +
( A(r, t))
2m
mi
2mi
e2 2
A (r, t) e(r, t)
+
2m
0 + W
(t)
H
(3.11)

=
H

question: how do A and look like?

assume EM field without sources (neither charges nor currents). The
homogeneous Maxwell equations for A and take a simple form if one uses

32
the so-called Coulomb gauge defined by the requirement A = 0. One is
then left with
2 A

1 2
A = 0
c2 t2
2 = 0

(3.12)
(3.13)

The only solution to Eq. (3.13) which is compatible with the requirement
that free EM waves are transverse is the trivial one = 0. A monochromatic
(real) solution of Eq. (3.12) reads
A(r, t) = |A0 | cos(k r t + )

(3.14)

with the unit vector

being orthogonal to the wave vector k. This ensures
that indeed A = 0. If one inserts (3.14) into (3.12) one obtains the
dispersion relation = ck. For details on the solutions of the free Maxwell
equations see [Jac], Chap 6.5.
Using the gauge conditions we arrive at the perturbation
e2 2
(t) = e A(t) p
+
W
A
m
2m

(3.15)

= eAp
is the usual
For weak fields one can neglect the A2 term. W
m
starting point for a perturbative treatment of atom-radiation interactions in
the semiclassical framework. The perturbation depends on time. We need a
time-dependent version of perturbation theory to deal with it.

3.2

Time-dependent perturbation theory

a) General formulation
The Hamiltonian under discussion is of the generic form

0 + W
(t)
H(t)
= H
0 + w(t)
H

(3.16)

task: solve the time-dependent Schrodinger equation (TDSE)

i~t |(t)i = H(t)|(t)i

(3.17)

33
(t t0 ) = 0
assume that W

0 |j i = j |j i
H
|(t0 )i = |0 i ,

t t0 :
assume

ansatz :

|(t)i =
=

X
j

j = 0, 1, . . .
(initial state)

(3.18)

cj (t)e ~ j t |j i

(3.19)

cj (t)|j (t)i

(3.20)

insertion of Eq. (3.19) into Eq. (3.17):

 i
X
X
i

cj e ~ j t H(t)|

i~cj + j e ~ j t |j i =
ji
j

i~ck =

X
j

| hk (t)|

e ~ (k j )t cj hk |w(t)|

ji

(3.21)

coupled-channel equations. (still exact if basis is complete)

(t > T ) = 0
If W
ck (t > T ) = const and



2
2
2
pk = |ck | 
= |hk |i| 
= |hk |i| 
= const
t>T

t>T

t>T

(3.22)

transition probabilities 0 k

note that

X
k

pk =

X
h|k ihk |i = h|i = 1
k

as it should.
Ansatz for solution of Eq. (3.21): power series expansion
(0)

(1)

(2)

ck (t) = ck (t) + ck (t) + 2 ck (t) + . . .

insertion into Eq. (3.21) yields


(0)
(1)
2 (2)
i~ ck + ck + ck + . . .
 i
X  (0)
(1)
(2)

=
cj + cj + 2 cj + . . . e ~ (k j )t hk |w(t)|
ji
j

(3.23)

34

0 :

(0)

i~ck = 0
X (0) i
(1)

i~ck =
cj e ~ (k j )t hk |w(t)|
ji

1 :

(2)
i~ck

X
j

(1)

cj e ~ (k j )t hk |w(t)|

ji

these equation can be solved successively:

(0)

0 :

ck (t) = const = k0
(cf. Eq. (3.18))
X
i
(1)
i~ck
=
j0 e ~ (k j )t hk |w(t)|

ji

1 :

(3.24)

j
i

= e ~ (k 0 )t hk |w(t)|
0i

(1)
ck (t)

=
| {z }
~
(1)
ck (t0 )
=0

t0

e ~ (k 0 )t hk |w(t
)|0 i dt

(3.25)

(as = 0 at t = t0 )
accordingly:
(2)
ck (t)

Z
Z
1 X t t i (k j )t i (j 0 )t
dt e ~
dt
= 2
e~
~ j t0
t0
hk |w(t
)|j ihj |w(t
)|0 i

(3.26)

Comments:
(i) exact calculations beyond 1st order are in general impossible due to
infinite sums
(ii) interpretation
t0
st

1 order
nd

|0 i

|k i

direct transition

2 order
|0 i |j i |k i
transition via virtual intermediate states (two steps)
(iii) further reading (and better visualization in terms of generic diagrams):
[Mes] II, Chap. 17

35
b) Discussion of the 1st -order result
To 1st order time-dependent perturbation theory we have (cf. Eqs. (3.24)
and (3.25)):
Z
i t ik0 t
e
Wk0 (t ) dt
(3.27)
ck (t) k0
~ t0
with

k 0
~
transition frequency
|0 i = hk |w|
= hk |W
0i
transition
matrix element
Z t
1

eik0 t Wk0 (t ) dt |2
= 2|
~ t0
transition probability

k0 = k 0 =
Wk0
p0k

What about the elastic channel (k = 0)?

the calculation of the probability that the system remains in the initial state
is a bit trickier due to the occurrence of the 1 in c0 . To be consistent in the
orders of the smallness parameter one has to consider
(0)

(1)

(2)

p00 = |c0 + c0 + 2 c0 + . . . |2
(1)

(1)

(2)

(2)

= 1 + (c0 + c0 ) + 2 (c0 + c0
(2)

(1)

(1)

+ |c0 |2 ) + O(3)

= 1 + 2 (2Re(c0 ) + |c0 |2 ) + O(3 )

X
= ... = 1
p0k

(3.28)

k6=0

the latter result can be worked out explicitly and is not surprising: it expresses probability conservation.

36
Examples
(i) Slowly varying perturbation
Lets assume that the perturbation is turned on very gently after t = t0 .
It might then stay constant for a while and/or is turned off equally
is a very small
gently. This is to say that the time derivative of W
quantity for all times.
k 6= 0

ck (t)

=
=
k0 0
W

Z
i t ik0 t
e
Wk0 (t ) dt

~ t0
Z t
t
i
i h 1 ik0 t
1
ik0 t

e
Wk0 (t ) dt
e
Wk0 (t )
t0
~ ik0
ik0 t0
(t)|0 i
hk |W
eik0 t

k 0

lets plug this into the time-dependent state vector:

X
i
ck (t)e ~ k t |k i
|(t)i =
k

= c0 (t)e ~ 0 t |0 i +

X hk |W
(t)|0 i
k6=0

0 k

e ~ 0 t |k i

the amplitude c0 can be set equal to 1. This introduces at most a small

phase error (up to first order)

|(t)i =

|0 i +

X hk |W
(t)|0 i
k6=0
i

|0 (t)ie ~ 0 t

0 k

 i
|k i e ~ 0 t
(3.29)

comparing this to the general results of stationary PT one can interpret

0 + W
(t) with eigenen|0 (t)i as the first-order eigenstate of H(t)
=H
(1)
(t)|0 i. The system is in the ground state
ergy 0 (t) = 0 + h0 |W

of the instantaneous Hamiltonian H(t)

at all times. This situation is
called adiabatic.

37
Comments:
(i) The argument can be generalized to strong perturbations. The
general adiabatic approximation then results in the statement: if
the perturbation varies slowly with time, the system is found in

an eigenstate of H(t)
at all times.
(ii) Adiabatic conditions are realized, e.g., if atom beams are directed
through slowly varying magnetic fields ( Stern-Gerlach experiment) and in slow atomic collisions. In the latter case the electrons
adapt to the slowly varying Coulomb potentials of the (classically)
moving nuclei and do not undergo transitions. They are in socalled quasimolecular states during the collision and back in their
initial atomic states thereafter.
(iii) further reading: [Boh], Chap. 20; [Gri], Chap. 10
(ii) Sudden perturbation
V

(t) =
W
t

t t0 0
t > t0

k 6= 0

Z
i t ik0 t
e
Wk0 (t ) dt
ck (t) =
~ 0
|0 i Z t
hk |W

=
eik0 t dt
i~
0

|0 i 
hk |W
=
eik0 t 1
~k0

transition probability

4|Wk0|2
f (t, k0)
~2
2
k0 0 t

p0k (t) = |ck (t)|2 =

f (t, k0) =

t
sin2 k0
2
2
k0

(3.30)
(3.31)

38

Figure 3.1: y = f (t, k0)

Significant transitions occur only around k0 = 0 within the width

= 2t . This is a manifestation of the energy-time uncertainty relation: if one waits long enough, transitions can only occur into states
that have the same energy as the initial state. One can state this more
precisely (mathematically) by considering the limit t :
f (t, k0)

t
(k 0 )
2

(3.32)

2t
|Wk0|2 (k 0 )
~
The sudden perturbation sounds academic, but it has an important
application (see later).

p0k

(iii) Periodic perturbation

(t) =
W

it it
Be + B e

t t0 = 0
t > t0

(3.33)

39
=W
)
(note that W

1
ck (t) =
i~
=

1
~

eik0 t Wk0 (t ) dt

0
(


0i 
hk |B|
ei(k0 +)t 1
k0 +

)

|0 i 
hk |B
ei(k0 )t 1
+
k0
if t

(i.e. ):

o
4|Bk0 |2 n
f
(t,

+
)
+
f
(t,

)
k0
k0
~2
n
o
2t
t
2
|Bk0| (k 0 + ) + (k 0 )

~
0 i and Eq. (3.31)
with Bk0 = hk |B|
p0k (t) =

(3.34)

Comments:

(i) Resonances at k0 = : significant transitions occur only around

these frequencies
k0 =

e
0

k = 0 ~

stimulated emission (of energy)

(only possible if 0 is not the ground state)

h w
e
k

k0 = +

e
k

absorbtion (of energy)

h w
e

k = 0 + ~

(ii) Since 0 is not the ground state in the case of stimulated emission
it makes sense to change the notation and use i for the initial
and f for the final state. Transition frequencies are then denoted
as f i etc.

40
(iii) We were careful enough to write that energy is emitted or absorbed
if the resonance conditions are met. We cant tell in which form
this happens. Later we will see that the energy is carried by
photons if the perturbation is exerted on the atom by the quantized
radiation field. The perturbation derived in Sec. 3.1 corresponds
to a classical radiation field. Lets stick to this case first and
convince ourselves that it can be written in the form (3.33).
(iv) Connection to atom-radiation interaction
In Sec. 3.1 we found that for weak EM fields the perturbation
takes the form
= e Ap

(3.35)
W
m
If we use Eq. (3.14) for a monochromatic field we obtain


= e A0 ei(krt)
+ A0 ei(krt)

W
p
p
2m
it + B
eit
= Be
with

= e A eikr

p
(3.36)
B
2m 0
(v) Validity of first-order time-dependent perturbation theory
We have two conditions to fulfill to be able to apply our results
for the periodic perturbation:
criterion to avoid overlap of the resonances
2

t
validity criterion on resonance
=

4|Bf i |2
|Bf i |2 2
f
(t,

=
0)
=
t 1
f
i
~2
~2
~
t
|Bf i |

pif

combine

2
2
~
=

|f i |
|Bf i |
|f i | |Bf i |

41

3.3

Photoionization

Let us elaborate on the case of absorption. If the energy, i.e., the field
frequency is high enough, the atom will be ionized. The final state of the
electron will then be a continuum state and this requires some additional
considerations, since such a state is not normalizable. This implies that
pif = |hf ||(t)i|2 is a probability density rather than a probability. A
proper probability is obtained if one integrates over an interval of final states
a) Transitions into the continuum1 .
e k+ D e
e k-D e
e

Pif :=

f +

pif (f )(f ) df (3.37)

f



with (f ): density of states in interval [f ; f + ]
Lets be a bit more specific and calculate the density of states for free-particle
continuum states
f (r) = hr|pi =

i
1
exp(
p r)
[2~]3/2
~

These states are normalized with respect to -functions:

Z
Z
1
i

3
hp|p i = hp|rihr|p i d r =
exp[
(p p) r]d3 r = (p p)
3/2
[2~]
~
Nevertheless, they fulfill a completeness relation (like the position eigenstates
|ri) such that for any (normalized) state |i
Z
Z
3
1 = h|i = h|pihp|i d p = |(p)|2 d3 p
and the density of states with respect to momentum is simply (p) = 1. We
need to transform this from momentum to energy. For free particles we have
1
Unfortunately, the figure doesnt correspond to the i f notation, but I cannot
change it easily!

42
the simple relation f = p2 /2m such that
dp
df dp
d3 p = p2 dpdp = p2
df
p
=
2m3 f df dp
Z
Z p
Z

2 3
2
3
1 =
|(p)| d p =
2m f
|(p)| dp df
Z
=
(f )pif (f )df

with
(f ) =

p
2m3 f

Lets go back to the probability (3.37) and insert the first-order result
(3.34):
Pif

4
= 2
~

f +

n
o
|Bf i |2 (f ) f (t, f i + ) + f (t, f i ) df

Z f +
4
2
f (t, f i )df
2 |Bf i | (f )
~
f
Z
sin2 ( 2t )
4
2

|Bf i | (f )
d

2
|Bf i |2 (f )t
=
~

R sin2 ( t
)
where we have used 2 2 d
= t
. A similar result is obtained for the
2
case of emission. One defines a transition rate wif = dtd Pif to obtain
abs
Pif

Fermis Golden Rule (FGR)

e,a
wi


2

|Bf i |2 (f )
f =i ~
~

(3.38)

43
b) Dipole approximation
In a typical photoionization experiment the wavelength of the applied EM
field is large compared to the size of the atom. This allows for a simplification,
which is called the dipole approximation:
2
a0
eikr 1
k
(3.36)
e
|i i

Bfdip
A hf |
p
i =
2m 0
this can be rewritten by using
if =

=
p
0 =
for H

p2
2m

im
[H0 , r]
~

(3.39)

+ V (r)
|i i =
hf |
p

im
0
0 |i i
hf |H
r
rH
~

im
(f i )
hf |r|i i
~

The matrix elements in questions are the well-known dipole matrix elements
(cf. Sec. 2). For
= z we have the standard selection rules m = 0, l =
12 . They result in characteristic angular dependencies of photoionized electrons, e.g., (see assignment # 3)
e2
|A0 |2 (f i )2 (f )|hf |z|i i|2
2~3
cos2

dip,
z
wi
f =

if the initital state is an s-state (l = 0).

Literature: [CT], Chap. XIII; [Lib], Chap. 13.5-13.9; [Sch], Chap. 11; [BS],
Chap. IV
Note that the notion of photons for the interpretation of photoionization
(and stimulated emission) has no significance as long as the analysis is based
on classical EM fields. From a theoretical point of view photons come into
the picture only if the EM field is quantized. This does not change the final
2

for other polarizations the m-selection rule changes, but the l-selection rule does not.

44
expressions for stimulated emission and absorbtion, but it shows that there
is another process which cannot be described in our semiclassical framework:
spontaneous emission, i.e., the emission of a photon (and transition to a
lower-lying state) without any external EM field. So, lets take a look at
field quantization.

3.4

Outlook on field quantization

In the semiclassical framework we represented the EM field by classical potentials (A, ), which act on the wave function of the electron(s). Now, we
want to derive a Hamiltonian that acts on the electron(s) and the EM field.
Accordingly, we need a wave function that includes the degrees of freedom
of the field.
Lets start with the Hamiltonian. The generic ansatz for a system that
consists of two subsystems (atom and EM field in our case) which may interact is as follows:
= H
1 + H
2 + W

H
= H
+H
+W

H
| A {z F}

(3.40)

0
=H

A is the Hamiltonian of the atom, H

F the (yet unknown) Hamiltonian of
H
the interaction. We aim at a description of the problem
the field, and W
within first-order perturbation theory. So we know which steps we have to
take:
Steps for a first-order TDPT treatment
0
determine H
A =
(a) H

p
2
2m

F =?
(b) H

+V

45
0
solve eigenvalue problem of H
A |j i = j |j i
H
F |k i = k |k i ?
H


0 |jk i = H
A + H
F |j i|k i
H
= (j + k ) |j i|k i
= (j + k ) |jk i

(3.41a)

(3.41b)

0 ) is a physics problem, the second

While the first step (determining H
0 s eigenvalue problem) is a math problem.
one (solving H

determine W
here it seems natural to start from the semiclassical expression (3.35)
and replace the classical vector potential by an operator that acts on
the degrees of freedom of the EM field
= eA
p

W
m

(3.42)

F
Not surprisingly, it turns out that consistency with the form of H

determines the form of A

apply FGR
the main issue here is to calculate the transition matrix elements
|i i
Wif = hf |W

(3.43)

F
a) Construction of H
Loosely speaking, Hamiltonians are the quantum analogs of classical energy
expressions. Lets look at the energy of a classical EM field in a cube of
volume L3
Z
1
WEM =
(E D + B H)d3 r
2 L3
Z
1
0
(E2 + c2 B2 )d3 r
c2 =
(3.44)
=
2 L3
0 0

46
Recall that free EM waves within Coulomb gauge ( A = 0) are obtained
from
2 A

1 2
A = 0
c2 t2
= 0

Instead of a monochromatic solution of the wave equation we consider the full

solution in L3 for periodic boundary conditions A(x, y, z = 0) = A(x, y, z =
L) etc. The latter are a convenient means to obtain a set of discrete modes
and discrete sums in all expressions instead of integrals:
X
A (r, t)
(3.45a)
A(r, t) =


3 A (r, t) = 3 q ei(k r t) + q ei(k r t)
L2

k =

mode index

unit polarization vectors

2
(nx , ny , nz )
L

= ck

k = 0

(3.45b)

|
| = 1

ni Z wave vectors and numbers

field frequencies
transverse waves

Each mode is thus defined by a wave vector (with three components) and a
polarization direction3 . If one uses this explicit form of A to calculate
E(r, t) = t A(r, t)
B(r, t) = A(r, t)
and plugs this into Eq. (3.44) one arrives at
X
2 q q
WEM = 20

(3.46)

this is to say that is actually a quadruple index. Note that for the general solution
of the wave equation we need two linearly independent polarizations per wave vector.
This can be made more explicit by a somewhat more elaborate notation, see, e.g. [Sch],
Chap. 14.

47
Obviously, the EM energy doesnt change with timethis is another (convenient) consequence of using periodic boundary conditions.
The next step is a transformation to real amplitudes:

0 (q + q )
(3.47)
Q :=

P := i 0 (q q )
(3.48)

WEM



1
P
Q + i

2 0



1
P
Q i

2 0



P
P
1 X 2
(Q + i ) Q i
=
2

1 X
=
P2 + 2 Q2
2

(3.49)
(3.50)

(3.51)
(3.52)

Doesnt this look familiar? The EM field has the algebraic form of a collection
of harmonic oscillators. We know how to quantize the harmonic oscillator,
so here is the quantization recipe:
P
Q
with

P = P
= Q

, P ] = i~
[Q


X 
2 = H

F = 1
P2 + 2 Q
WEM H

F
2

(3.53)
(3.54)
(3.55)
(3.56)

F together with the commutation relations (3.55)

The algebraic form of H
determine its spectrum:


X
1
~ n +
En1 ,n2 ,... =
n = 0, 1, 2, ...
2

The standard QM interpretation would be to associate each mode with a

particle trapped in a parabolic potential. The energies En = ~ (n + 21 )

48
would then correspond to the ground- and the excited-state levels of that
particle. However, it is not clear what kind of particle this should be, since
Q and P do not correspond to usual position and momentum variables
(and actually there is no parabolic potential around).
The fact that the spectrum is equidistant allows for an alternate interpretation, in which each mode is associated with n quanta 4 , each of which
carries the energy ~ . In this interpretation a mode does not have a ground
state and excited states, but is more like a (structureless) container that can
accomodate (any number of) quanta of a given energy. The quanta are called
photons. At this point their only property is that they carry energy, but
we will see later that there is more in store. Note that the photon interpretation is only possible because the spectrum of the harmonic oscillator is
equidistant!
n : occupation
number of mode
P
N = n : total number of photons in the field

However, even without any photon around there is energy, unfortunately

even an infinite amount:
X ~
zero-point energy
E0 =

(3.57)
2

A proper treatment of this infinite zero-point energy requires a so-called

renormalization procedure. We will not dwell on this issue, but (try to) put
our minds at ease with the remark that normally only energy differences have
physical significance5 . These are always finite since the zero-point energies
cancel.
b) Creation and annihilation operators

For the further development it is useful to introduce creation and annihilation

operators:
b = 1
+ iP ) anniliation op.
( Q
(3.58)
2~
b = 1
iP ) creation op.
( Q
(3.59)

2~
4

you may say particles instead of quanta, but these particles turn out to have an odd
property: zero rest mass.
5
In fact, renormalization is not much more than a more formal way of adopting this
viewpoint.

49
Lets play with them and look at their commutators, e.g.
i
i
h
1 h

Q + iP , Q iP

b , b =
2~

h
i
h
i
1

, P +i P , Q
=
i Q
2~

| {z }
| {z }

i~

1
=
{~ + ~ } = 1
2~

i~

h
i
b , b =

h
b
h
b

, b
, b

(3.60)

= 0

(3.61)

= 0

(3.62)

rewrite the Hamiltonian


X 
F = 1
2
H
P2 + 2 Q

2


~ 2 ~ 2
1X

=
(b b ) +
(b + b )
2
2
2


1X
=
~ b b + b b
2

(3.63)

1
~ (b b + )
2
1
~ (
n + )
2

(3.64)

n
= b b is called occupation number operator.
h
i

Obviously HF , n
= 0 , i.e., they have the same eigenstates. Lets

50
first consider a single mode only:
F =
H

F
H

|n i = ~ (n + 1 ) |n i
H
F

2
with n N0

(3.65)

Usually, one uses a short-hand notation for the eigenstates and writes
|n i |n i. These states are called (photon) number or Fock states. Note
that |0 i |0i is not a vector of zero length, but the (ground) state associated with the statement that there are no photons in the mode (but
energy E0 = ~ /2). As eigenstates of a hermitian operator the |n i fulfill
an orthonormality relation
hn |n i = n n

n |n i = n |n i
n |0i = 0|0i = 0

1
H
n
= F 2
~

(3.66)
!

this is a real zero!

(3.67)
(3.68)

The eigenvalue n is the number of photons in the mode. This justifies the
name occupation number operator for n
.
Lets operate with creation and annihilation operators on these photon
number states. To do this we need a few relations that can be proven without
difficulty.
i
h
b , n
= b
i
h
b , n

= b




n
b |n i = b n
|n i + b |n i


= (n + 1) b |n i

(3.69)
(3.70)

51
obviously, the vector b |n i is an eigenstate of n with the eigenvalue
(n + 1). On the other hand:
n
|n + 1i = (n + 1) |n + 1i
combine:

b |n i = |n + 1i

2
|| hn + 1|n + 1i = hn | b b |n i
{z
}
|
=1

= hn | n
+ 1 |n i
= n + 1

b |n i =

n + 1

n + 1 |n + 1i

(3.71)

This justifies the notion creation operator: operating with it on a number

state creates one additional photon. Similarly one finds
b |n i = n |n 1i
(3.72)
to be consistent with Eq. (3.68) we stipulate
b |0i = 0
Generalization:
F |n1 , n2 , ...i = E |n1 , n2 , ...i
H
(3.73)

 X
X
1
E =
~ n +
=
~ n + E0 (3.74)
2

|n1 , n2 , ...i = |n1 i |n2 i...

product states

n |... n ...i = n |... n ...i

b |... n ...i =
n + 1 |... n + 1 ...i

b |... n ...i =
n |... n 1 ...i

(3.75)

(3.76)
(3.77)
(3.78)

52
c) Interaction between electron(s) and photons
We start from Eq. (3.42) and require that the quantization of the vector
potential be consistent with the quantization of the EM energy. The latter
implies the following quantization rule for the amplitudes:
! r

P
1
~
+ i
Q
q q =
=
b
(3.79)
2 0

20
which yields
t) =
A(r,



~
b ei(k r t) + b ei(k r t)

2 0 L3

(3.80)

is the time dependence of an operator in the

The time dependence of A
Heisenberg picture. The corresponding Schrodinger operator reads
r


X
~
b eik r + b eik r

(3.81)
A(r)
=

3
2
0 L

The latter is used in the interaction potential:

= eA
p
W

m
s
n
o
X
e2 ~3
ik r
ik r

b e
+ b e

(3.82)
= i
20 m2 L3

Not surprisingly, the interaction has a part that acts on the electron and a
part that acts on the photons. The electronic part is the same as in the semiclassical treatment. The photonic part involves the creation or annihilation
of a photon in the mode .
d) Transitions matrix elements
We need
|i i =
Wif = hf |W
s

|i i = i
hf |W

|i i
hf |W

e2 ~3
2 0 m2 L3

{hf | eik r
|i i hnf1 ... nf ...| b |ni1 ... ni ...i
+hf | eik r
|i i hnf ... nf ...| b |ni ... ni ...i}
1

53
The electronic matrix elements are the same as in the semiclassical framework. We have discussed them in dipole approximation in Sec. 3.3. Lets look
at the photonic matrix elements. Basically, they result in selection rules:
q
f
f
i
i

hn1 ... n ...| b |n1 ... n ...i = ni hnf1 ... nf ...|ni1 ... ni 1 ...i (3.83)
noting that hnf |ni i = f i in each mode one obtains
q
f
f
i
i

hn1 ... n ...| b |n1 ... n ...i = ni nf ni nf ni ...nf ni 1 ...

1

(3.84)

and similarly
hnf1

...

nf

...| b |ni1 ... ni ...i =

ni + 1 nf ni nf ni ...nf ni +1 ...
1

(3.85)

We obtain a nonvanishing contribution only if the number of photons in

one mode differs by one in the initial and final states, but is the same in
all other modes. Obviously, the annihilation process (3.84) corresponds to
photon absorption (one photon less in the final state than in the initial state)
and the creation process (3.85) to photon emission (one photon more in the
final state than in the initial state).
Summary:

|i i =
hf |W
6 0

iff

1. the photon numbers in |i i and |f i differ exactly by one in one single

mode. Otherwise orthogonality will kill the transition amplitude. This
is to say that first-order perturbation theory accounts for one-photon
processes only.
2. dipole selection rules are fulfilled for the crucial mode (assuming that
eik r 1).
3. energy conservation is fulfilled: Ei = Ef . This follows from the fact
that the total Hamiltonian is time independent and is also consistent
with the application of TDPT for a time-independent perturbation (cf.
example (ii) of Sec. 3.2 b)).

54
Elaborate on energy conservation: noting that the energy eigenvalues of electronic and photonic parts add (cf. Eq. (3.41b)) we find


X
1
i
(3.86)
Ei = i +
~ n +
2



X
1
f
~ n +
Ef = f +
(3.87)
2

Ei = Ef f = i ~

(3.88)

with the plus sign for absorption and the minus sign for emission.
e) Spontaneous emission
|i i = |i i|0, 0, . . .i
Even if there is no photon around in the initial state we can obtain a
nonvanishing transition matrix element by operating with a creation operator
on the vacuum state |0, 0, . . .i. This corresponds to the process of spontaneous
emission: an excited electronic state decays by (one-) photon emission even
though no external radiation field is present. One may say that it is the
zero-point energy in the given mode that triggers the transition.
For the investigation of the quantitative aspects of spontaneous emission
one uses FGR (3.38). The transition rate can be written as

s.e.
wif
=

2
|Wf i |2 (ef )
~s

Wf i = i

(3.89)

e2 ~3
hf |eik r
|i i
20 m2 L3
f i
with =
~

(3.90)

The density of states in FGR refers to the photon density in the final states.
To calculate it recall
k =

2
(n , n , n )
L x y z

ni Z

nx ny nz
N
=
=
(k) =
V
kx ky kz

L
2

3

55
We need = (ef ), i.e., we need to transform to energy space:
dV = d3 k = k 2 dkdk = k 2
photons: ef = ~ = ~ck

dk
def dk
def

1
dk
=
def
~c

3 2
 3 2
k

L
L
def dk =
def dk (ef )def dk
dN = (k)d k =
2
~c
2
~c3
Inserting this into Eq. (3.89) one obtains for the (differential) spontaneous
emission rate (for a given polarization
):
3

s.e.,

dwif
e2 ~
= 2
|hf |eik r
|i i|2
dk
8 0 m2 c3

In dipole approximation (using once again Eq. (3.39)) this reduces to

s.e.,

dwif
e2 3
=
|
rif |2
dk
8 2 0 ~c3
rif = hf |r|i i

(3.91)
(3.92)

The total rate is obtained by integrating (3.91) over dk and summing over
two perpendicular (transverse) polarizations. The final result for the inverse
s.e.
lifetime 1/if = wif
is
 dip
1
e2 3
=
| rif |2 .
(3.93)
if
30 ~c3
Examples:
Lifetime of H(2p)
dip
1.6 109 s
2p1s

This looks like a short lifetime, but compare it to the classical revolution
time of an electron in the hydrogen ground state!
st

1 order
2s1s

For the transition rate 2s 1s one finds a strict zero for the first-order
rate even beyond the dipole approximation. Experimentally one finds
2s1s 0.12s. Theoretically, one obtains this number in a secondorder calculation, in which two-photon processes are taken into account.

56
This expresses to a general pattern: an N-photon process corresponds to
an N-th order TDPT amplitude.
f) Concluding remarks on photons
What have we learned about photons (associated with a given mode)?
they can be created and annihilated (i.e. they dont last forever)
they carry energy ~
they carry momentum ~k
this can be shown by starting from the classical expression for the total
momentum of the free electromagnetic field
Z
PEM = 0
E B d3 r
L3

f one obtains
using similar arguments as for the translation of WEM to H
X
PF =
~k n

and

PF |n1 , n2 , . . .i =

~k n |n1 , n2 , . . .i

they carry angular momentum (called spin) ~

this can be shown by starting from the classical expression for the total
angular momentum of the free electromagnetic field
Z
LEM = 0
r E B d3 r
L3

f . However,
using similar arguments as for the translation of WEM to H
this calculation requires a more explicit consideration of different possible polarization directions (which are not obvious in our condensed
-notation).
photons are bosons!
The spin-statistics theorem shows that particles with integer spin fulfill Bose-Einstein statistics. There is no restriction on the number of

57
bosons (photons) which can occupy a given state (mode)we havent
encountered an upper limit for the occupation numbers n . One can
set up a similar (occupation number) formalism for many-electron systems and finds that in this case the occupation numbers can only be 0
or 1 due to the Pauli principle for fermions.
Finally, we add that photons travel with v = c (because free EM waves
do) and thus must have zero mass (to avoid conflicts with Einsteins
theory of relativity).
literature on field quantization:
relatively simple accounts on the quantization of the EM field can be
found in [Fri], Chap. 2.4 and [SN], Chap. 7.6
higher formulations: [Sch], Chap. 14; [Mes], Chap. 21 (and, of course,
QED textbooks)

Chapter 4
Brief introduction to
relativistic QM
Literature: [BD]; [BS], Chap. 1.b; [Mes], Chap. 20; [Sch], Chap. 13; [Lib],
Chap. 15; [SN], Chap. 8
The first reference is a classic textbook. The latter book chapters provide
condensed accounts on relativistic quantum mechanics.

4.1

Klein-Gordon equation

a) Setting up a relativistic wave equation

starting point: classical relativistic energy-momentum relation1
E 2 = p2 c2 + m2 c4

quantization :
(standard rules)

E i~t
~

p
i
p2 ~2 2
E 2 ~2 t

In the following, m always denotes the rest mass m0 .

58

(4.1)

59
obtain (free) wave equation (Klein-Gordon equation (KGE))
~2 t2 (r, t) = ~2 c2 (r, t) + m2 c4 (r, t)

(4.2)

The KGE was first obtained by Schrodinger in the winter of 1925/26,

but abandoned due to problems. Schrodinger then concentrated on
the nonrelativistic case and found his equation, while the KGE was
rediscovered a bit later by Klein and (independently) by Gordon.
b) Discussion
1. KGE is second-order PDE wrt. space and time
2. KGE is Lorentz covariant
3. Time development is determined from initial conditions (t0 ),
which is at odds with evolution postulate of QM

(t ),
t 0

4. Continuity equation?
one can derive t + divj = 0
with standard current j =
but :

i~
(
2m

i~
( t t )
2mc2

(4.3)

problem: (rt) 0 (i.e. not positive definite)

probabilistic interpretation is not possible (or at least not obvious)
5. Ansatz (i)
(r, t) = Aei(krt)
insertion into Eq. (4.2) yields together with de Broglie relations

E = ~ = c2 ~2 k2 + m2 c4 0
(4.4)
and j are okay (check!), but what does E < 0 signify?
Ansatz (ii)
(r, t) = Aei(krt)
results in the same expression for E, but corresponds to < 0 (check!).
Ansatz (iii)
(r, t) = A sin(kr t)

60
also results in the same expression for E, but corresponds to = 0,
j = 0 (check!).
6. Add Coulomb potential to free KGE and solve it (in spherical coordinates)
yields wrong fine structure of hydrogen spectrum (i.e., contradicts
experimental findings)
7. In 1934, the KGE was recognized as the correct wave equation for
spin-zero particles (mesons).

4.2

Dirac equation

In 1928, Dirac found a new wave equation which is suitable for electrons
(spin 21 -particles): the Dirac equation (DE)
a) Free particles
ansatz :

D
i~t = H

(4.5)

st

i.e. stick to the form of the TDSE; a PDE of 1 order in t such that (t0 )
is the only initial condition
requirements (Diracs wish list):
1. DE must be compatible with energy-momentum relation (4.1)
2. DE must be Lorentz-covariant
3. Obtain continuity equation with probabilistic interpretation
4. Stick to the usual quantization rules!
Dirac recognized that these requirements cannot be satisfied by a single scalar
equation, but by a matrix equation for a spinor wave function with N components.
ansatz :

D = c p
+ mc2
H
3
c~ X
=
i xi + mc2
i j=1

(4.6)

61
with N N matrices x , y , z , and spinor wave function

1 (r, t)

..
=
as solution of (4.5)

.
N (r, t)

requirement (1) is met if each component i solves KGE (4.2)

iterate Eq. (4.5):
D (H
D )
i~t (i~t ) = H
~2 t2 =

 c~ X
i

j xj + mc2

2 2

~ c

= ~2 c2

 c~ X

j k xj xk

jk

X j k + k j
jk


k xk + mc2

o


~ 3X
2 2 4
j + j xj + m c
+ mc
i
j

x2j xk + 2 m2 c4 +

~mc3 X
(j + j )xj
i
j

comparison with KGE yields conditions for i , :

j k + k j = 2jk
j + j = 0
j2 = 2 = 1

(4.7)
(4.8)
(4.9)

further conditions and consequences:

D shall be hermitian)
j , hermitian (because H
= real eigenvalues
4.9
= eigenvalues are 1
from (4.7)-(4.9) it follows that j , are traceless, i.e.
tr j = tr = 02
together with eigenvalues 1 this implies that dimension N is even
2

The trace of a matrix A is defined as the sum over P

the diagonal elements. The trace
does not change when A is diagonalized. Hence tr A =
eigenvalues.

62
N = 2 is too small as there are only three anti-commuting (Eqs. (4.7)
and (4.8)) matrices for N = 2 (the Pauli matrices). Dirac needs four!
try N = 4
derive explicit representations from these conditions

1
0



1
0 i

i =
,
=

1
i 0
0
1
with Pauli matrices i


0 1
x =
,
1 0

= free DE takes the form

1
2
i~t
3
4

y =

0 i
i 0

z =

1 0
0 1

2 2
= (c p

+
mc
)

3
4

(4.10)

(4.11)

and one can derive a meaningful continuity equation:

t + divj = 0

with

= =

4
X

(r, t)(r, t)

i=1

and

j = c

( i.e. jk =

c(1 , 2 , 3 , 4 )

0
k

b) Solutions of the free DE

Ansatz :

j (r, t) = uj ei(krt) ,

after some calculation one finds:

1


k
2 )
3
0
4

j = 1, ..., 4

63
there are 4 linear independent solutions.
p
Two correspond to
E = + p2 c2 + m2 c4
p
and two to
E = p2 c2 + m2 c4
they have the form (E > 0):

0
u(1) =
1
2
and for E < 0:

u(3)
with

1
2

=
1 ,
0

u(2)

u(4)

0
1

=
1
2

1
2

=
0
1

cpz
c(px + ipy )
c(px ipy )
,
2 =
, 1 =
, 2 = 1
2
2
E + mc
E + mc
E + mc2
c(px + ipy )
c(px ipy )
cpz
,
2 =
, 1 =
, 2 = 1
1 =
2
2
2
E mc
E mc
E mc
note that all these small components approach zero for v c.
u(1) , u(3) are interpreted as spin up
u(2) , u(4) are interpreted as spin down solutions
1 =

e
E

c
0

5 1 1 k e V fo r e
0
-m
0

h o le

D ira c s e a

Figure 4.1: Energy spectrum of the free DE

64
Diracs interpretation(1930):
In the vacuum all negative energy states (in the Dirac sea) are occupied.
Hence, if electrons are present at E > mc2 they cannot fall down into the
Dirac sea because of the Pauli principle (electrons are fermions).
On the other hand, one can imagine that it is possible to excite one
electron from the Dirac sea to E > mc2 . Such an excitation corresponds
to a hole in the Dirac sea, which can be interpreted as the presence of a
positively charged particle an anti-particle (i.e. a positron). This process
electron-positron pair creation has indeed been observed, and also the
reversed process destruction of electron-positron pairs and -ray emission
(the latter to balance the total energy).
In fact, the first experimental detection of positrons in 1932 was considered a strong proof of Diracs theory.
c) Throw in (classical) EM potentials
use the minimal coupling prescription
~
+ eA
+ eA = p
i
E i~t + e
o
n
2
(
i~t = c
p + eA) e + mc
p

(4.11)

(4.12)

one can show that Eq. (4.12) is Lorentz-covariant.

d) The relativistic hydrogen problem
Consider Eq. (4.12) with A = 0 and
=
ansatz :

Ze
40 r
i

(r, t) = (r)e ~ Et
(4.13)

n
Ze2 o
p
+ mc2
(r) = E(r)

c
40 r
this stationary DE can be solved analytically!
yields

(4.14)

65
Result for the bound spectrum ( fine structure):
# 21
"
2
(Z)
n = 1, 2, ...
(4.15)
Enj = mc2 1 +
(n j )2
r
1
1 3
1 2
1
(4.16)
(Z)2 ,
j = , , ...n
j = j +
j+
2
2
2 2
2
n (still) is the principal quantum number, while j can be identified as quantum number of total angular momentum.
~
e2
1
=

mca0
40 ~c
137
fine-structure constant

(4.17)

expansion of Eq. (4.15) in powers of (Z)2 1:

"
#

2
4
3
(Z)
(Z)
1

Enj = mc2 1
...

2n2
2n3 j + 21
4n
1st term:
2nd term:
3rd term:

(4.18)

rest energy
non-relativistic binding energy (1.23)
lowest order relativistic corrections fine structure splitting
of energy levels

E = m
0

n = 3
n = 2

3 s3 p 3 d
3 d
3 p
3 s

2 s2 p
2 p
2 s

n = 1

1 s

1 .8 x 1 0

-4

3 /2
1 /2

3 d
3 p

3 /2
1 /2

3 /2
1 /2

2 p

1 /2

e V
1 s

S c h r d in g e r

5 /2

1 /2

D ira c

Figure 4.2: Energy spectrum of the Coulomb problem

66
Further corrections (beyond DE)3 :
hyperfine structure (coupling of magnetic moments of electron(s) and
nucleus)
106 eV
QED effects (Lamb shift): further splitting of levels with same j, but
different l quantum numbers
106 eV
e) Nonrelativistic limit of the DE
Instead of solving Eq. (4.14) exactly and subsequently expanding the exact eigenvalues (4.15) it is useful to consider the non- (or rather: weak-)
relativistic limit of the stationary DE (4.14) and to account for the lowestorder relativistic corrections obtained in this way in 1st -order perturbation
theory. This procedure yields the same result (4.18) once again, but this
time it comes with an interpretation regarding the nature of the relativistic
corrections.
Starting point: stationary DE
n
o
2
p
+ mc + V (r) (r) = E(r)
c

(4.19)

group the 4-component spinor according to

 





1
1
=
, with =
, =

2
2

insert into (4.19) (using similar groupings of the Dirac matrices in terms of
Pauli matrices):

 o 


 
n
1 0

(4.20)
c
p
= E V (r) mc
0 1

= (E V (r) mc2 )
c p
= (E V (r) + mc2 )
c p
3

for details see [BS]

(4.21)
(4.22)

67
Solve (4.22) for and insert into (4.21):

c2
= (E V (r) mc2 )
p
E V (r) + mc2

(4.23)

define = E mc2 mc2

assume V (r) mc2
expand
1
c2
c2
1

=
=
V
(r)
2
2
E V (r) + mc
+ 2mc V (r)
2m
2m(1 + 2mc2 )



V (r)
1
2mc2

use all this in (4.23) to obtain





V (r)
1
1
= ( V (r))
p
p
2m
2mc2
[V (r)]
apply the product rule for p
use the following identity for Pauli matrices and arbitrary vector oper B

ators A,

=A
B
+ i (A
B)

( A)(
B)
use for a central potential
V (r) =

1 dV
r
r dr

to obtain
with

(T1 + T2 + T3 ) = ( V (r))
T1
T2
T3

 2


V (r) p
=
1
2
2mc
2m
~ 1 dV
( l)
=
4m2 c2 r dr
~ 1 1 dV
)
=
(r p
i 4m2 c2 r dr

(4.24)
(4.25)
(4.26)

68
Interpretation of terms
For the interpretation of T1 note that

2
p

( V (r))
2m
2
4
p
p
KE
T1 =

TN R + H
(4.27)
2m 8m3 c2
KE represents the lowest-order relativistic correction to the kinetic
H
energy (as it appears without hats in a classical treatment).

Introducing the spin operator s = ~2 , which due to the properties

of the Pauli matrices fulfills the standard commutation relations of an
angular momentum operator4 , T2 is identified as the spin-orbit coupling
term
SO .
T2 H
Hence, spin and spin-orbit coupling are automatically included in a
relativistic treatment (which is why some authors insist that electron
spin is a relativistic property).

T3 is not hermitian. Consider its hermitian average




1
~
1
~
1
dV
dV
T
+
T
3
3
) (
=
(r p
p r)
T3 =
2
8m2 c2 i r dr
i
r dr
2
~
Darwin .
2 V (r) H
=
8m2 c2
The Darwin term doesnt have a nonrelativistic or classical counterpart.
It is usually associated with the Zitterbewegung (trembling motion)
of the electron due to the nonzero coupling of electrons and positions
(or: large and small components of the Dirac spinor) [BD].
Now apply perturbation theory to the problem
= H
0 + W

(4.28)
2

Ze
0 = TN R + V (r) = p
H

2m 40 r
= H
KE + H
SO + H
Darwin
W
4

(4.29)
(4.30)

for a recap of angular momentum and spin operators consult a QM textbook, e.g.
[Gri], Chaps. 4.3, 4.4

69
One obtains for the first-order energy correction
E (1) =

 1
mc2
4
(Z)
2n3
j+

1
2

3
,
4n

(4.31)

i.e., the same results as in Eq. (4.18), which shows the consistency of the
approach.

Chapter 5
Molecules
A simple mans definition of a molecule says that it is an aggregate of atoms
which cling together by bonds. The big question then is: why are the atoms
doing this, i.e., what is the nature of the chemical bond? It turns out that
this can only be answered satisfactorily by quantum mechanics.
We should thus start by insisting that molecules are quantum-mechanical
many-body systems consisting of M 2 atomic nulcei and N 1 electrons,
which interact with each other (mainly) by Coulomb forces. Lets assume
that the Coulomb force is the only force present and that everything can be
described nonrelativistically.
Hamiltonian:
= Tnuc + Tel + V
H
M
M
X
2
~2 X 1 2
P

Tnuc =
2M
2 =1 M R
=1
Tel =
V
Vnn

M
N
X
2
p
~2 X 2

=
2me
2me i=1 ri
i=1

= Vnn + Vne + Vee

M
X
Z Z e2
=
40 |R R |

(5.1)
(5.2)
(5.3)
(5.4)
(5.5)

<

Vne =

M X
N
X
=1 i=1

Z e2
40 |R ri |

70

(5.6)

71

Vee =

N
X
i<j

e2
40 |ri rj |

(5.7)

The task is to solve the stationary Schrodinger equation

|i = E |i
H

(5.8)

for this Hamiltonian. This would be a complicated problem, but one doesnt
really have to do it in most situations of interest.

5.1

The adiabatic (Born-Oppenheimer) approximation

The adiabatic (or: Born-Oppenheimer) approximation forms the basis for

the discussion of molecular structure and properties. It exploits the large
mass difference between electrons and nuclei.
a) Prelude
Lets consider the (drastic) approximation of infinitely heavy nuclei which
cannot move:
M Tnuc 0 .

We are left with the simpler electronic SE



Tel + V (R) n (r, R) = En (R)n (r, R)

(5.9)

with r = (r1 , r2 , ..., rN ) and R = (R1 , R2, ..., RM ).

Remarks
1. The nuclear coordinates R are now external parameters (no independent variables). The electronic wave functions n and the energy
eigenvalues depend parametrically on R.
2. Actually, we should use a better notation that accounts for the electrons spins, e.g., r x = (r1 s1 , ...rN sN ), since the spin coordinates
play a role for the permutation symmetry of many-electron systems
(even though spin-dependent interactions are neglected). We dont do
it for the sake of simplicity, since the arguments used in this section
are independent of symmetry and spin.

72
3. Since Tel +V (R) is hermitian, the n form a complete (and orthogonal)
basis for any configuration {R} of the nuclei. We can write, e.g.,
hm | n ir = h m (R)| n (R)ir
Z
=
m (r1 , r,2 , ...rN , R)n (r1 , r,2 , ...rN , R)d3r1 d3 r2 ...d3 rN
= mn

b) Adiabatic approximation: separation of electronic and nuclear motion

Back to the full problem.
Ansatz:

(r, R) =

n (R)n (r, R)

(5.10)

Sub into Eq. (5.8), multiply everything by n (r, R), use Eq. (5.9) and integrate over electronic coordinates choosing n (r, R) = n (r, R) such that
R hm | m ir = h R m | m ir + hm |R | m ir
0 = 2 hm |R | m ir
to obtain


X

Cmn (R)n (R)

Tnuc + Em (R) + m (R) E) m (R) =

(5.11)

n6=m

with definitions
m (R) =



E
 
m (R) Tnuc  m (R)

X ~2

Cmn (R) =
(R))
(Amn (R) R + Bmn
M

Amn (R) = hm |R | n ir
 

1

Bmn
(R) =
m 2R  n r
2

The adiabatic approximation consists in neglecting all the coupling terms:

X
Cmn (R)n (R) 0 .
(5.12)
m6=n

73
Hence, we arrive at two SEs; one for the electronic and one for the nuclear
motion:


Tel + V (R) m (r, R) = Em (R)m (r, R)
(5.13)


Tnuc + Um (R) m (R) = Em (R)
(5.14)
Um (R) = Em (R) + m (R) .

(5.15)

The third equation provides the connection of the two SEs. This set of
equations is the standard framework for the discussion of molecular structure
and properties.
Discussion
,,Born-Oppenheimer approximation (BOA): adiabatic approximation
plus assumption m (R) 0.
potential energy surfaces (PES) Em (R) determine nuclear motion:

"antibonding"

R
E 0 (R)

De

"bonding"

Re

Figure 5.1: Typical PESs for M = 2 (Em (R) = Em (R)) as functions of the
internuclear distance R.
For R all PESs become horizontal (vanishing Coulomb forces),
for R 0 all PESs increase like R1 because Vnn dominates. The minimum of the lower (ground-state) curve corresponds to the equilibrium
distance (bond length) of the molecule.
me
Gist and justification of BOA: M
104

The nuclei are so slow that they feel only an average electronic field

74
(via the PESs). The fast electrons follow the slow nuclear motion adiabatically, i.e., without undergoing transitions.
A closer inspection of the BOA shows that the neglected terms m (R)
and Cmn (R) are proportional to the smallness parameter 1/M .

5.2

Nuclear wave equation: rotations and vibrations

Lets consider a diatomic molecule (M = 2). The structure of the problem

is that of a two-body central-field problem. Hence, we can reduce it to an
effective one-body problem using CM and relative coordinates (cf. Chaps. 1.1
and 1.2). The nontrivial SE for the relative (internal) motion reads
!
2

PR
+ Un (R) n (R)
2
!
2
PR2
J
+
+ Un (R) n (R)
2 2R2

En (R) =
=
with the reduced mass =

(5.16)

M1 M2
.
M1 +M2

Simplest model: rigid rotor

The rigid rotor is defined by setting R = Re , i.e., by assuming that the
internuclear distance is held fixed at the equilibrium distance. In this
model, n depend only on angular coordinates and Eq. (5.16) reduces
to
Erot JM (, ) =
Erot =
I =
JM (, ) =

2
J
JM (, )
2Re2
~2 J(J + 1)
~2 J(J + 1)
=
2Re2
2I
2
Re moment of inertia
YJM (, )

(5.17)
J = 0, 1, . . .(5.18)
(5.19)
(5.20)

Measurement of rotational absorption spectra (typically in the microwave regime) give information on I, i.e., on Re .

75
Allow for radial motion: vibrations
Equation (5.16) can be separated by using the ansatz
n (R) =

PnJ (R)
YJM (, )
R

radial equation:
 2 2

~ d
~2 J (J + 1)

+ Un (R) +
PnJ (R) = EPnJ (R)
2 dR2
2R2

(5.21)

Taylor-expand bonding ground-state PES about equilibrium distance:

1
U0 (R) U0 (Re ) + U0 (Re )(R Re ) + U0 (Re )(R Re )2 + ...
2
ke 2
(5.22)
= U0 (Re ) + x
2
with
ke = U0 (Re ) molecular force constant
x = R Re
use this to obtain the approximate radial equation


~2
1
~2 J(J + 1)
2
PnJ (x) + ke x PnJ (x) =
PnJ (x)
E U0 (R)
2
2
2Re2
Evib PnJ (x)
(5.23)
which is the SE of the harmonic oscillator with eigenenergies


1
Evib = ~e +
= 0, 1, 2, ...
2
s
s
ke
U0 (Re )
=
and e =

76

E
Re
De D0 D1

R
U0 (R)

Figure 5.2: Typical ground-state PES and harmonic oscillator model of vibrational motion. De = U0 (Re ) is the binding energy and D0 = De ~2 e the
dissociation energy of the molecule.
Summary:
E = U0 (Re ) + Evib + Erot
with
Eel = U0 (Re ) = E0 (Re ) + 0 (Re )


1
Evib = ~e +
2
2
~ J(J + 1)
Erot =
2I
note that
E0 (Re ) is independent of
Evib

Erot

Erot Evib Eel

Literature: [Lev], Chaps. 4.3 and 6.4; [Hur], Chap. 1

77

5.3

The hydrogen molecular ion H+

2

The hydrogen molecular ion is the simplest molecular system and plays a
similar role in molecular physics as the hydrogen atom in atomic physics.
The electronic SE can be solved exactly (but not in closed analytical form).
Lets just take a look (without going into technical details).
The electronic SE (atomic units)


1
1
1
1 2
n (r, R) = En (R)n (r, R)
+

2
ra rb R

(5.24)

is separable in confocal-elliptical coordinates (, , ):

1
(ra + rb ) 1 <
R
1
(ra rb ) 1 1
=
R
y
= arctan
0 < 2
x
=

with the standard azimuthal angle :

(, , ) = M()N()()
for () one obtains



d2
2
+ () = 0
d2
() = ei

require uniqueness:
() = ( + 2)
i2 i
= e|{z}
e
= 0, 1, 2, ...
1

lz () = i () = ()

h
i

H, lz = 0

(5.25)

78
However:
use to classify states

h
i
2

H, l 6= 0

H atom l = 0
notation
s
+
H2 = 0
notation

1
p
1

2
3
d
f
2 3

Table 5.1: Classification of atomic vs. molecular one-electron states

Eigenfunctions (, , ) of Eq. (5.25)
point reflections:

r r =

are also parity eigenstates

+

two types of solutions

(r) =
=
=
=
(r) =
=
=
=

(, , )
(, , + )
(r)
g (r)
gerade (even) parity
(, , )
(, , + )
(r)
u (r)
ungerade (odd) parity

classification of molecular orbitals (MOs)

1g , 1u , 2g , 2u , 3g , 3u , . . .
1g , 1u , 2g , 2u , 3g , 3u , . . .
..
.

79
Discussion of exact PESs En (R) and correlation diagrams Enel (R) = En (R)
Note that
2
n2
1
Enel (R ) = 2
2n
Enel (R 0) =

united-atom limit
separated-atom limit

Since the Enel are continuous functions of R, each united-atom state (e.g.
1sg ) is correlated with a separated-atom state (e.g. g (1s)).

Figure 5.3: Calculated correlation diagram (left) and PESs (right) of H+

2

1
R

80
g (1s)
R
u (1s)
R

g (1s) + u (1s)

Figure 5.4: Sketch of the electron density distributions of the lowest states
of gerade and ungerade parity in the separated-atom limit and how they add
to form an atomic state, e.g., on the left centre
Enel (R 0)
-2.0
-0.5
-0.5
-0.5
-0.222
-0.222
-0.222
-0.222
-0.222
-0.222
-0.125

AO
1s
2p0
2p1
2s
3p0
3d0
3d1
3d2
3p1
3s
4f0

MO
1g
1u
1u
2g
2u
3g
1g
1g
3u
4g
3u

AO Enel (R )
1s
-0.5
1s
-0.5
2p1
-0.125
2s
-0.125
2s
-0.125
2p0
-0.125
2p1
-0.125
3d2
-0.056
3p1
-0.056
3s
-0.056
2p0
-0.125

Table 5.2: Energies (in atomic units) and united- and separated-atom limits
of a few MOs of H+
2

81
bond length

binding energy

Re
1.9972
E0 (Re ) -0.6026
E0el (Re ) -1.1033
De
-0.1026

Table 5.3: Ground state properties of H+

2
Literature: [Lev], Chap. 13

Bibliography
[BD]

Bjrken, J.D. and Drell, S.D.: Relativistic Quantum Mechanics.

McGraw-Hill.

[Boh] Bohm, D.: Quantum Theory. Prentice Hall.

[BS]

Bethe, H. A. and Salpeter, E. E.: Quantum Mechanics of oneand two-electron atoms. Plenum.

[CT]

Cohen-Tannoudji, C.: Quantum Mechanics I, II. Wiley Interscience.

[Fri]

Friedrich, H.: Theoretical Atomic Physics. Springer.

[GPS] Goldstein, H., C. Poole and J. Safko: Classical Mechanics.

Addison Wesley.
[Gri]

Griffiths, D. J.: Introduction to Quantum Mechanics. Prentice

Hall.

[Hur] Hurley, A. C.: Introduction to the Electron Theory of Small

Molecules. Academic Press.
[Jac]

Jackson, J. D.: Classical Electrodynamics. John Wiley & Sons.

[Lev] Levine, I. N.: Quantum Chemistry. Prentice Hall.

[Lib]

Liboff, R. L.: Introductory Quantum Mechanics. Addison-Wesley.

[Mes] Messiah, A.: Quantum Mechanics I, II. Dover Publications.

[Sch] Schiff, L. L.: Quantum Mechanics. McGraw-Hill.
[Sha] Shankar, R.: Principles of Qunatum Mechanics. Kluwer Academic.
82

83
[SN]

Sakurai, J. J. and J. Napolitano: Modern Quantum Mechanics.

Addison-Wesley.