4 Time-ind.

Perturbation Theory II
We said we solved the Hydrogen atom exactly, but we lied. There are a
number of physical eects our solution of the Hamiltonian H = p
2
/2m
e
2
/r left out. We already said motion of proton can be accounted for
by nearly trivial reduced mass correctionnot important, since spectra of
original problem unchanged except for tiny overall energy scaling.
In fact there are a hierarchy of other eects we left out, each of which
can be treated in perturbation theory on top of the other. In the most
complete theory, almost all degeneracies are split. Classify formally in
powers of ne structure constant e
2
/ hc 1/137.
1. Bohr levels: E
n
= R/n
2
O(
2
mc
2
)
2. Fine structure:
R
2

2
n
4
_
n
j+1/2

3
4
_
O(
4
mc
2
)
(relativistic eectskinetic energy & spin-orbit coupling)
3. Hyperne structure:

0
ge
2
3m
p
ma
3
0
S
p
S
e
) O(m/m
p
)(
4
mc
2
)
(spin-spin coupling of e

& p
+
magnetic moments)
4. Lamb shift:
1
3
R
n
2
_
2
n
_
3
ln
mc
2
([E
m
E
n
[)
avg
O(
5
mc
2
)
(quantization of photon eldreally larger than hyperne)
4.1 Fine Structurespin-orbit coupling
1st eect of interest is coupling of i) spin of electron to ii) its motion, 2
things which we certainly treated independently before. Full derivation of
proper

V is hairy, give only hand-waving derivation here, misses factor of
2.
Force on electron is eE = e
2
r/(4
0
r
3
). Recall that observer moving
in electric eld E sees magnetic eld
1
B =
v
c
2
E v c (1)
So electron with magnetic moment
= g
e
2m
S, g 2 (2)
should have energy

V = B (3)
= g
e
2m
S
e
4
0
r
3
r
v
c
2
(4)
so with L r mv, get

V =
ge
2
2m
2
c
2
_
_
1
4
0
r
3
_
_
L S (5)
O by factor of 2 from full relativistic calculation (neglected Thomas pre-
cession!):

V =
ge
2
4m
2
c
2
1
4
0
r
3
L S (6)
(7)
= f(r)L S (8)
Now dene total angular momentum
J = L + S (9)
whose square is

J
2
=

L
2
+

S
2
+ 2L S, (10)
so

V =
f(r)
2
(

J
2

L
2

S
2
) (11)
Verify:

J
2
,

L
2
,

S
2
, and

J
z
now commute with H = H
0
+

V .

L
z
and

S
z
no longer do. We choose therefore to label states as
2
Thus we arent quite following prescription we set out, by looking at ex-
pectation value of

V in original eigenstates of H
0
. Why? Because we can
nd the new exact eigenstates with no work. So energy shift calculation
is exactexcept for higher order relativistic eects. In this basis we have
(

J
2

L
2

S
2
)[nsjm
j
)
= h
2
(j(j + 1) ( + 1)
3
4
)[nsjm
j
) (12)
so energy shift due to spin-orbit coupling is
E
so
= nsjm
j
[

V [nsjm
j
)
= h
2
(j(j + 1) ( + 1)
3
4
)f(r)) (13)
For f(r)) we need

1
r
3
) =
1
( + 1/2)( + 1)n
3
a
3
0
( ,= 0) (14)
(see Griths Probs. 6.34, 6.35). Spin-orbit energy shift is therefore
E
so
=
E
2
n
mc
2
_
_
_
n(j(j + 1) ( + 1)
3
4
)
( +
1
2
)( + 1)
_
_
_ (15)
=
E
2
n
mc
2
2n
_
_
_
1
+
1
2

1
j +
1
2
_
_
_ (check for j = 1/2!) (16)
Note although (15) looks singular at 0, proper relativistic treatment
3
by Dirac shows nonsingular result (16) is correct for all not obvious
though!
Relativistic kinetic energy corrections. Up to now we have neglected rel-
ativistic eects in kinetic energy term

T. Properly we should probably
write

T =
_
p
2
c
2
+ m
2
c
4
mc
2

p
2
2m

p
4
8m
3
c
2
+ . . . (17)
Then we need to nd the energy shift in [) to leading order in the per-
turbation

V =
p
4
8m
3
c
2
:
E
rel
=
1
8m
3
c
2
[ p
4
[) =
1
8m
3
c
2
p
2
[ p
2
) (18)
The rhs supposed to be evaluated using unperturbed wave fctns, so use
( p
2
/2me
2
/(4
0
r))[) = E
n
[) to nd
E
rel
=
1
2mc
2
[E
2
n
+ 2E
n
e
2
_
1
4
0
r
_
+ e
4
_
1
(4
0
)
2
r
2
_
] (19)
The expectation values depend only on a
0
, n and as usual:
_
1
r
_
=
1
n
2
a
0
;
_
1
r
2
_
=
1
( + 1/2)n
3
a
2
0
(20)
So full relativistic kinetic energy shift is
E
rel
=
E
2
n
2mc
2
_
_
_
4n
+ 1/2
3
_
_
_ (21)
and full ne structure energy splitting is
E
fine
= E
so
+ E
rel
(22)
yielding an energy spectrum for hydrogen up to ne structure corrections
of
E = E
n
+ E
fine
=
R
n
2
_

_1 +

2
n
2
_
_
_
n
j + 1/2
3/4
_
_
_
_

_ (23)
4
Figure 1: Hydrogen ne structurespin-orbit coupling
4.2 Hyperne Interaction
Hyperne splitting of atomic spectral lines arises from interaction of elec-
tron & nuclear dipole moments. Whereas ground state in hydrogen not
split by ne structure, it is split by hyperne interaction, allowing for
transition between two levels, producing radiation of wavelength 21 cm
( 10
7
Ryd, allows tracking of hydrogen by astronomers, measurements
of proper motion of celestial objects (Doppler shifts of line), & construc-
tion of stable atomic clocks. Calculation is lengthy, but can be solved by
elementary methods to high accuracy (real quantum mechanics prob-
lem!).
Magnetic eld due to dipole moment :
B =
1
4
0
_
_
_
3r( r)
r
5


r
3
_
_
_ , (24)
whereas the interaction of a magnetic moment with external eld is
U = B (25)
5
Figure 2: If
p
| z, r | x, B | z, then
e
|
p
in lowest energy conguration. If
p
| z, r | z, B | z, then
e
|
p
for lowest energy. Compare with (25).
so interaction is

V =
1
4
0
_
_
_
3(
e
r)(
p
r)
r
5
+
(
e

p
)
r
3
_
_
_ (26)
where r = r
e
r
p
Magnetic moment operators:

e
= g
e
e
2M
e
S
e
;
p
= g
p
e
2M
p
S
p
(27)
with gyromagnetic ratios g
e
2.00, g
p
5.59.
Perturbation theory
Want to use

V in (25) as perturbation, write H = H
0
+ H
fine
+

V , look
for hyperne energy shift E
hf
in ground state [
0
).
1
Can write ground
state as
[
0
) = [
100
(r))[m
e
m
p
) (28)
where 1st factor is spatial part, old friend
100
= e
r/a
0
/
_
a
3
0
, 2nd factor
represents spin state of electron and now proton, too (before we ignored
proton spin since it didnt contribute to energy, now we must include it as
1
Note we will focus exclusively on the ground state of the H-atom, where the spin-orbit coupling has already been accounted for.
Although in general states of denite m
p
, m
e
are not eigenstates because of the L S coupling, for the S
1/2
ground state, m
j
(evalue of

J
z
/h) is the same as m
s
(evalue of

S
z
/h) always since = 0. We call m
s
of electron m
e
here, and m
s
for proton m
p
.
6
dynamical variable). Note there are four degenerate states before taking

V into account, since both m

2], and
singlet state [ ) [ ))/

2.
So we want to evaluate
3
E
hf
=
/
0
[

V [
/
0
) (29)
=
g
e
g
p
e
2
16
0
M
e
M
p

100
(r)[SM[
_
3(S
e
r)(S
p
r)
r
5

(S
e
S
p
)
r
3
_
[SM)[
100
(r))
=
g
e
g
p
e
2
16
0
M
e
M
p
I

SM[

S
e

S
p
[SM) (30)
where the [SM) are estates of total spin

S
2
and total

S
z
, and
I

_
d
3
r[
100
(r)[
2
_
_
_
3r

r
5

r
3
_
_
_ (31)
is a tensor we have to evaluate using the wave function
100
. Tedious but
straightforward to show that I

= I

, with I = 8[
100
(0)[
2
/3 (see
below!). I diagonal = only spin matrix elements we need are
SM[S
e
S
p
[SM) = SM[
1
2
(

S
2

S
2
e


S
2
p
)[SM)
=
h
2
2
_

_
1(1+1)
1
2
(1+
1
2
)
1
2
(1+
1
2
) S = 1
0
1
2
(1+
1
2
)
1
2
(1+
1
2
) S = 0
=
_

_
h
2
/4 S = 1
3 h
2
/4 S = 0
(32)
So energy shifts for singlet and triplet states are
2
Check that

S
2
commutes with

V by rst writing

S
2
= (S
e
+ S
p
)
2
=

S
2
e
+

S
2
p
+ S
e
S
p
. First of all,

S
2
e
commutes with

V
since [

S
2
e
,

S
e,
] = 0, similarly for

S
2
p
. Now S
e
S
p
obviously commutes with second term of

V (see (25)), harder part is to see
[S
e
S
p
, (S
e
r)(S
p
r)] = 0. To see this, use relation
i

j
=
ij
+ i
ijk

k
.
3
Need to use states

0
> in which

V is diagonalthese are precisely the |SM!
7
E
hf
=
g
e
g
p
e
2
16
0
M
e
M
p
_
_
_
8[
100
(0)[
2
3
_
_
_
_

_
h
2
/4 S = 1
3 h
2
/4 S = 0
|
_
_
_
8M
3
e
e
6
3 h
6
_
_
_
=
g
e
g
p
e
8
M
2
e
24
0
M
p
h
4
_

_
1 S = 1
3 S = 0
(33)
So splitting between triplet and singlet (energy of photon emitted in triplet
decay) is
h = E
hf
(S = 1) E
hf
(S = 0)
=
g
e
g
p
e
8
M
2
e
6
0
M
p
h
4
(34)
and frequency of transition is
= /2 =
g
e
g
p
e
8
M
2
e
12
2

0
M
p
h
5
(35)
= 1420405751.800 0.028Hz (36)
or = c/ = 21.1 cm. Value given is current measurement!
Pf. of I

= 8[
100
(0)[
2

/3
Recall denition
I

_
d
3
r[
100
(r)[
2
_
3r

r
5

r
3
_
(37)
Quantity I

is singular integral and must be handled carefully. Note

= I

.
2. Diagonal terms funky, since angular part of, e.g. I
zz
gives
_
d (3z
2
r
2
) = 2r
2
_
1
1
d cos (3 cos
2
1) = 0 (38)
but radial integral is proportional to
_
dr
e
r/a
0
r
= (39)
So we have to nd way to take limit 0 sensibly. Note singularity in (38) comes entirely
from r = 0; if we were to exclude r = 0 from range of integration integral would vanish!. So
consider very small sphere around r = 0, of radius r
0
so small we can approximate smoothly
varying fctn.
100
(r) by
100
(0). Consider diagonal elt. I
zz
(others must be same by rotation
symmetry), express in cylindrical coords:
I
zz
= [
100
(0)[
2
_
r
0
0
d
3
r
_
3z
2
r
5

1
r
3
_
= [
100
(0)[
2
_
r
0
r
0
dz
_

r
2
0
z
2
0
d
2

_
3z
2
(
2
+ z
2
)
5/2

1
(
2
+ z
2
)
3/2
_
= [
100
(0)[
2
_
r
0
r
0
dz
_
2z
2
(
2
+ z
2
)
3/2
+
2
(
2
+ z
2
)
1/2
_

2
=r
2
0
z
2

2
=0
= [
100
(0)[
2
_
r
0
r
0
dz 2
_
1
r
0

z
2
r
0
_
=
8
3
[
100
(0)[
2
(40)
9