Notes 3: Hydrogen Fine Structure

High-resolution spectra of atoms show lines that are not single but have closely spaced multiple
components. This fine structure is from the splitting of energy levels due to effects that we have not
included in the simple model of the hydrogen atom. A proper analysis of fine structure requires solving
the relativistic Dirac equation. Here we consider an approach using three perturbations to the Schrödinger
Hamiltonian. The first perturbation is a first-order relativistic correction to the kinetic energy operator.
We then consider spin - orbit coupling arising from the interaction between the electron magnetic moment
and the magnetic field generated by the motion of the proton relative to the electron. The third
perturbation, called the Darwin term, only effects the l = 0 eigenstates. The Darwin term arises because
the electron cannot be confined to a point and must be treated as fuzzy ball of radius equal to the electron
Compton wavelength (i.e. the wavelength corresponding to the mass of the electron).

Relativistic correction
The kinetic energy operator in the Schrödinger equation is derived by assuming that the motion of the
particle(s) is non-relativistic, i.e. the relation between kinetic energy and momentum for a particle of mass
M is 1

1 p2
T= . (3.1)
2M

The relativistic expression is

 p2 
T= c 2 p 2 + M 2 c 4 − Mc 2 = Mc 2  1 + 2 2 − 1 , (3.2)
 M c 
 

where c is the speed of light in vacuum. We can find the first order correction to the Hamiltonian by
( Mc )
2
expanding T as a power series in p 2

 p2   1 p2 1 p4  1 p2 1 p4
T= Mc 2  1 + 2 2 − 1=
 Mc 2  − + = − + . (3.3)
 M c   2 M 2 2
c 8 M 4 4
c  2 M 8 M 3 2
c
 

The ratio of the first two terms in the series has magnitude p 2 ( 4M c ) , which is of order the electron’s
2 2

kinetic energy divided by its rest mass energy. Numerically, this ratio is ~10-5. Relativity effects are small
and difficult to detect in light atoms but become sizable for the inner electrons of massive atoms.

1
I use M do avoid confusion with the magnetic quantum number m in application to the hydrogen atom. In many
cases, M should be the reduced mass µ but GS do not make this distinction.

1
Since p2 is a Hermitian operator and commutes with itself, p4 is also a Hermitian operator (see problem
3.4 c). It is easy to show that the operators p2, L2 and Lz commute. Hence, by using the commutator
identity from problem (3.14),

=ˆ ˆ ˆ ˆ  ˆ ˆ  ˆ ˆ ˆ
 AB, C  A  B, C  +  A, C  B,

with Aˆ= Bˆ= p 2 and Cˆ = L2 or Lz , we can show that p4 commutes with L2 and Lz. Since L2 and Lz have
distinct eigenvalues, the good state theorem allows us to use the n2 degenerate hydrogen wave functions
ψ nlm in non-degenerate perturbation theory to find the first-order energy corrections due to the
perturbation

1 p4
H1 = − . (3.4)
8 M 3c 2

The correction to the energy is

1 1
E1 =
− 3 2
ψ 0 p 4ψ 0 =
− 3 2
p 2ψ 0 p 2ψ 0 , (3.5)
8M c 8M c

where we has used the Hermitian property of p2. For the unperturbed eigenstates

p 2ψ 0 2m ( E 0 − V 0 )ψ 0 ,
= (3.6)

and so

1
(E −V 0 ) 2 (
1  0 2
E ) − 2 E 0 V 0 + (V 0 )  .
2 2
E1 =
− 0
=
− (3.7)
2 Mc 2 2 Mc  

For the hydrogen atom

1 e2
V 0 (r ) = − , (3.8)
4πε 0 r

and so equation (3.7) gives

1  2 1 
2
 e2  1  e2 
E =
−
1
 E − 2 En  +  . (3.9)
4πε 0  r  4πε 0 
n
2 Mc 2   r2 
 

An elegant way to evaluate the expectation values is to use the so-called Hellmann-Feynman theorem
(Problems 7.38 and 7.42), which states that if the Hamiltonian H depends on a parameter β and has
(good state) orthonormal eigenfunctions ψ n ( β ) and eigenvalues En ( β ) , then

2
 ∂En ∂H
= ψn ψn . (3.10)
∂β ∂β

The results are

1 1 1 1
= = , , (3.11)
r 2
na r 2
 1 3 2
l + n a
 2

where a is the Bohr radius. Hence the relativistic correction to the energy is

 
1   e  1  e 
2 2 2
1 
E1rel =
−  E 2
+ 2 En  2 +   . (3.12)
 4πε 0  n a  4πε 0   l + 1  n3 a 2 
n
2 Mc 2 
   
 2

The energy of the unperturbed state is

1 e2 1
En = − . (3.13)
2n 2 4πε 0 a

Using this to eliminate the Bohr radius from equation (3.12), we obtain

E 2  4n 
E1rel =
− n 2 − 3 . (3.14)
2 Mc l + 1
 
 2 

The degeneracy has been partially lifted as this correction depends on l but not on the magnetic quantum
number m.

Problem 7.17 Find the lowest order relativistic correction to the energy levels of the one-dimensional
harmonic oscillator.

Solution: The energy states of the one-dimensional harmonic oscillator are non-degenerate. The energy
corrections are given by equation (3.7) but with

1
V 0 ( x) = mω 2 x 2 .
2

Thus we need to calculate x 2 and x 4 . The operator x is in terms of the raising and lowering
operators is


=x ( a+ + a− ) ,
2mω

3
Hence

1 ω ω
V 0= mω 2 x 2= ( a+ + a− ) = a+ a+ + a+ a− + a− a+ + a− a−
2

2 4 4
ω 2H ω  1
= 0+ + 0=  n + .
4 ω 2  2

Also

2
  
x4 = x 2ψ n x 2ψ n =  ( a+ + a− ) ψ n ( a+ + a− ) ψ n
2 2
 .
 2mω 

Using

a+ψ n =
n + 1ψ n +1 , a−ψ n =
nψ n −1 ,

we find

( a+ + a− ) ψ n = ( a+ a+ + a+ a− + a− a+ + a− a− )ψ n
2

= ( n + 2 )( n + 1)ψ n + 2 + ( 2n + 1)ψ n + n ( n − 1)ψ n − 2 .

Therefore

 4  ω    ω 
2 2 2

(V )  12 mω 2=
= 0 2
 x


 4 
 ( n + 2 )( n + 1) + ( 2n + 1) + n (=
2
n − 1)  
  4 
 ( 6n + 6n + 3) .
2

The energy correction is

1  
2 2
1 2 2  1 2 2 1 2 2
E1 =
− 2 
+  ω − +   ω +  ω ( 6n + 6n + 3 ) 
2
n   n
2mc  2  2 16 
( ω )
2
6n 2 + 6n + 3
= − .
mc 2 32

For both the hydrogen atom and the quantum harmonic oscillator the relative energy correction is

E1 E0
 .
E 0 Mc 2

For the hydrogen atom

2
E 0  e2 
  =α ,
2

Mc 2  4πε 0 c 

4
where α is the fine structure constant,

e2 1
=α ≈ .
4πε 0 c 137.036

The fine structure constant is a fundamental dimensionless quantity and has many physical
interpretations. One is that it is a coupling constant for the strength of the electron - photon interaction.

In terms of α , the zeroth-order hydrogen energies are

1 2 2
En = − α Mc , (3.15)
2n 2

and the Bohr radius is

1 
a= . (3.16)
α Mc
The first order relativistic corrections are from equation (3.14)

E1rel 1  4n 
−α 2 2 
= − 3 . (3.17)
E 0
4n l + 1
 
 2 

Spin – Orbit Coupling

The electron has an intrinsic magnetic moment that contributes to the total energy of the hydrogen atom
due to its interaction with the magnetic field from the orbital motion of the nucleus in the electron rest
frame.

The magnetic field from the proton’s motion is estimated by treating it as a current loop of radius r with
current equal to the charge divided by orbital period

e
=
I Ω, (3.18)
2π

where Ω is the orbital angular velocity. The magnetic field at the center of the loop (i.e. at the electron’s
location) is obtained from the Biot-Savart Law, which gives

µ0 I µ0 eΩ
=B = zˆ zˆ . (3.19)
2 r 4π r

We now eliminate the orbital angular velocity by using the orbital angular momentum of the electron
(since the angular velocity is the same for the proton and electron)

=
L Mr 2 Ωzˆ . (3.20)

5
The magnetic field is

µ0 e 1 e
=B = L L. (3.21)
4π Mr 3
4πε 0 Mc 2 r 3

We estimate the magnetic moment of the electron in a similar manner. The magnetic moment of a current
loop is equal to its area times the current

eΩ
μ = −π r 2 zˆ . (3.22)
2π

The angular momentum of the charge is

=
S Mr 2 Ω.zˆ (3.23)

Hence

e
μ= − S. (3.24)
2M

The ratio of a particle’s magnetic moment to its spin is called the gyromagnetic ratio.

The electron is a structureless point particle and treating it as a current loop is hard to justify. The
deviation from the ‘classical’ result for the magnetic moment is accounted for by introducing a
gyromagnetic ratio factor (usually called a g-factor) so that

e
μ = −g S. (3.25)
2M

For the electron, g is very close to 2.

Also in estimating the magnetic field, the electron was taken to be at rest. However, the electron rest
frame is a non-inertial frame and rotates relative to an inertial frame. This deficiency is accounted for by
including what is called Thomas precession.

When these additional considerations are included, the perturbation to the Hamiltonian due to spin – orbit
coupling is

e2 1 e2 1
H 1
=−μ ⋅ B =( g − 1) S⋅L ≈ S ⋅ L. (3.26)
8πε 0 M c r 8πε 0 M c r
SOC 2 2 3 2 2 3

This perturbation does not commute with L nor with S. Hence, the energy and Lz or Sz are not
1
simultaneously observable. However, H SOC does commute with L2, S2 and the total angular momentum J
= L + S. In particular, the Hamiltonian commutes with Jz, which has distinct eigenvalues. Thus non-
degenerate perturbation theory can be used.

6
Problem 7.19 Evaluate the following commutators: (a) [ L ⋅ S, L ] , (b) [ L ⋅ S, S ] , (c) [ L ⋅ S, J ] , (d)

L ⋅ S, L2  , (e) L ⋅ S, S 2  , (f) L ⋅ S, J 2  .

Solution: We make use of the commutator relation  Lx , Ly  = iLz and its permutations, and that L and
L2 commute.

(a) We have

[L ⋅ S, Lx ] = Lx S x + Ly S y + Lz S z , Lx  =S y  Ly , Lx  + S z [ Lz , Lx ] =i ( − S y Lz + S z Ly ) =i ( L × S ) x .

By also considering the other two components of the commutator, we conclude

[ L ⋅ S, L ] = i  ( L × S ) .
(b) By exchanging L and S in the result of part (a), we obtain

[ L ⋅ S, S ] =
−i ( L × S ) .

(c) Since J = L + S, from parts (a) and (b), we see that [ L ⋅ S, J ] =

0.

(d) Now

L ⋅ S, L2  = ( L ⋅ S ) L2 − L2 ( L ⋅ S ) = ( LL ) ⋅ S − ( L L ) ⋅ S =
2 2
L, L2  ⋅ S = 0.

(e) Applying the same method as in part (d), we obtain L ⋅ S, S 2  =

0.

(f) Since J 2 = (L + S) ⋅ (L + S) = L2 + 2S ⋅ L + S 2 , using the results from parts (d) and (e) gives
L ⋅ S, J 2  =
0.

To summarize, L ⋅ S commutes with L2, S2, J2 and J but not with L or S. Hence ml and ms are no longer
‘good’ quantum numbers but n, l, s, j and mj are.

Note that there is no spin – orbit coupling for spherically symmetric states, i.e. those for which l = 0.

To evaluate the first-order correction to the energy, we need to evaluate S ⋅ L and r −3 . The easiest
way to get the former is to use

J 2 = L2 + 2S ⋅ L + S 2 , (3.27)

so that

7
 J 2 − L2 − S 2
S⋅L = . (3.28)
2

Then

2
S=
⋅L  j ( j + 1) − l ( l + 1) − s ( s + 1)  .
2 
(3.29)

This is zero for l = 0, since then j = s.

The results of problem 7.43 can be used to find that, for l > 0,

2
r −3 = . (3.30)
n l ( l + 1)( 2l + 1) a 3
3

Putting the pieces together, we obtain

e2 1 −3 e2  2 1 j ( j + 1) − l ( l + 1) − s ( s + 1)
1
=
E SOC H 1
≈ r S ⋅=
L
8πε 0 M 2 c 2 8πε 0 M 2 c 2 a 3 n3l ( l + 1)( 2l + 1)
SOC

(3.31)
E 2 j ( j + 1) − l ( l + 1) − s ( s + 1)
= n2 n .
Mc l ( l + 1)( 2l + 1)

and so

E1SOC j ( j + 1) − l ( l + 1) − s ( s + 1)
=α2 . (3.32)
E 0
2nl ( l + 1)( 2l + 1)

The relativistic and spin-orbit corrections together are called fine structure. By considering separate cases
(problem 7.20), a concise expression for the fine structure energy correction can be obtained:

 
3 n  α2
= E  −
1 0
1
E fs . (3.33)
n2 4 j+ 
 2

The Darwin term

Strictly speaking, the semi-classical derivation of this result is valid only for l > 0. However, we have yet
to include the Darwin term, which has Hamiltonian

π 2 e2 3
= 1
H Dar = δ ( r ) πα 2 n 2 En a 3δ 3 ( r ) . (3.34)
2 M c 4πε 0
2 2

8
Because the l > 0 eigenfunctions are zero at the origin, they do not ‘see’ the delta function. Hence, the
Darwin term only affects the l = 0 states.

The energy corrections due to the Darwin term are

1
H Dar = πα 2 n 2 En a 3 ψ ( 0 ) . (3.35)

(
For l = 0, ψ ( 0 ) = 1 π ) ( na ) −3 2
. Hence

1
1
H Dar = α 2 En . (3.36)
n

Adding this correction to the relativistic correction leads to the same result as given by equation (3.33).

Splitting of Lyman alpha and Balmer alpha lines

The Lyman alpha spectral line is produced in the transition from the n = 2, l = 1 state to the n = 1, l = 0
state. The table shows the energy of these states with and without the fine structure correction. The energy
unit is the ionization potential of hydrogen.

n l j En Enj Enj-En
1 0 1/2 -1 -1.000013313 -0.000013313
2 1 3/2 -0.25 -0.250000832 -8.321 10-7
2 1 1/2 -0.25 -0.250004160 -4.160 10-6

The fine structure correction splits the Lyman alpha line into a doublet of wavelengths 121.5668 and
121.5674 nm. The separation between components is only 0.0006 nm.

The Balmer alpha spectral line is produced in the transition from the n = 3 state to the n = 2 state. Since
the orbital angular momentum can change by ±1, the allowed transitions are 3s to 2p, 3p to 2s and 3d to
2p. The table shows the energy of the states with and without the fine structure correction.

n l j En Enj Enj-En
2 0 1/2 -0.25 -0.250004160 -4.160 10-6
2 1 3/2 -0.25 -0.250000832 -8.321 10-7
2 1 1/2 -0.25 -0.250004160 -4.160 10-6
3 0 1/2 -0.111111111 -0.111112590 -1.479 10-6
3 1 1/2 -0.111111111 -0.111112590 -1.479 10-6
3 1 3/2 -0.111111111 -0.111111604 -4.931 10-7
3 2 3/2 -0.111111111 -0.111111604 -4.931 10-7
3 2 5/2 -0.111111111 -0.111111216 -1.644 10-7

9
The next table shows the possible photon wavelengths (in vacuum) for a given transition. The
wavelengths are in nm.

Transition
3s to 2p 656.4726663 656.4569364
3p to 2s 656.4569364 656.4522763
3d to 2p 656.4680060 656.4661719 656.4522763 656.4504427

The wavelength shown in red requires an electron spin flip. The wavelengths in italics are the same as
other wavelengths in the table. We see that we expect there to be five distinct peaks in a high-resolution
spectrum, spread over a wavelength interval of ~ 0.02 nm. When an additional effect called the Lamb
shift is included the number of expected peaks increases to seven. In atomic hydrogen gases, because of
the Doppler effect, usually only two peaks are observed. When saturation spectroscopy of gas cells or
collimated beams of hydrogen atoms are used, four peaks can be resolved. For more complicated atoms,
fine structure can be more easily visible. For example, the sodium D lines have a separation of 0.6 nm

The Lamb shift has some similarity to the Darwin term in that it affects only the l = 0, i.e. S states. Due to
the quantum fuzziness of the electron it has a slightly weaker interaction with the nucleus in the l = 0 state
than in l > 0 states. Hence, the 2S1/2 state has a higher energy than the 2P1/2 state.

10