Further Quantum Physics Perturbation Theory 1

FURTHER QUANTUM MECHANICS

Part 1: Pertubation Theory

Dr C W P Palmer Hilary Term 2013

CWPP 15/2/2013
Further Quantum Physics Perturbation Theory 2

What is a perturbation?
A ‘small’ change H1 to the Hamiltonian of a solved problem H0 .
Notation: Perturbed problem H |ni = En |ni
(0)
But H = H0 + H1 and the unperturbed problem is solved: H0 |n0 i = En |n0 i.
When is a perturbation small?
The fundamental assumption of perturbation theory: The unperturbed ket is a good approxi-
mation to the exact ket. The unperturbed kets form a complete set so we can write
X X
|ni = cm |m0 i where |cm |2 = 1.
m m
The fundamental assumption means then implies:
there is one big term in this expansion: cn ≈ 1 and all other terms are small: cm ≪ 1 for
n 6= m.
If we don’t care about normalisation (we can always normalise at the end of the calculation,
and it doesn’t affect eigenvalues) then we can in fact divide by cn and define instead
X
|ni = |n0 i + cm |m0 i .
m6=n

The correction to the ket is orthogonal to the unperturbed ket.

CWPP 15/2/2013
Further Quantum Physics Perturbation Theory 3

Perturbation Parameters
In perturbation theory there is usually an identifiable physical parameter λ such that H1 =
λh1 . For example, a fruitful source of perturbation problems is application of external fields
to atoms. These give perturbations like qφ = −qEz and −B · µ where q is the electric charge,
and µ the magnetic dipole moment. Obviously the external field strengths E and B can play
the role of λ.
We can now attempt an expansion of the unknown ket and eigenvalue as a power series in λ:
X
En = λp e(p) (0) (1) (0) (1)
n = en + λen . . . = En + En . . .
p

where the first term is independent of the perturbation, the second term scales linearly with
the perturbation, and so on. In effect we are splitting up the corrections to the eigenvalue
and ket into parts that vary as different powers of λ:
En = En(0) + En(1) + En(2) . . . |ni = |n0 i + |n1 i + |n2 i + . . .
(2)
where En contains a factor λ2 and |n1 i a factor of λ etc.
The question of the convergence of these series is a delicate one (and depends on the prob-
lem). But even if the series don’t converge, we can still use the first few terms to improve our
approximate zeroth order solution.

CWPP 15/2/2013
Further Quantum Physics Perturbation Theory 4

Non-degenerate Perturbation Theory

We first apply this approach to a non-degenerate level, so that there is a single zeroth-
(0)
order ket |n0 i with eigenvalue En . We write down the Schrodinger equation:

(H0 + H1 ) [|n0 i + |n1 i + |n2 i] = (En(0) + En(1) + En(2) ) [|n0 i + |n1 i + |n2 i]

and identify zeroth-, first-, second-order terms from the number of factors of λ, and equate
them separately:
(0)
Zeroth Order: H0 |n0 i = En |n0 i
(1) (0)
First Order: H1 |n0 i + H0 |n1 i = En |n0 i + En |n1 i
(2) (1) (0)
Second Order: H1 |n1 i + H0 |n2 i = En |n0 i + En |n1 i + En |n2 i
(0)
We recognise the unperturbed equation, giving us the En and |n0 i we already know.
The other equations give us the corrections.

CWPP 15/2/2013
Further Quantum Physics Non-degenerate Perturbation Theory 5

First-Order Correction to Energy

The first-order equation contains two things we don’t know, the correction to the ket |n1 i,
(1)
and the correction to the energy En :

H1 |n0 i + H0 |n1 i = En(1) |n0 i + En(0) |n1 i .

We can isolate these by taking the inner product with one of the unperturbed kets:

Suppose we take the inner product with |n0 i:

hn0 | H1 |n0 i + hn0 | H0 |n1 i = En(1) h n0 | n0 i + En(0) h n0 | n1 i .

But on the left we can use the Hermitian property of H0 :

 ∗  ∗
hn0 | H0 |n1 i = hn1 | H0 |n0 i = En h n1 | n0 i = En(0) h n0 | n1 i
(0)

which cancels a term on the right. (In fact, both of these terms are zero since h n0 | n1 i = 0.)

CWPP 15/2/2013
Further Quantum Physics Non-degenerate Perturbation Theory 6

First-Order Correction to Energy

This leaves just two terms which we can immediately rearrange to give

hn0 | H1 |n0 i
En(1) = = hn0 | H1 |n0 i
h n0 | n0 i

since the denominator is 1. (But note we don’t have to do perturbation theory with nor-
malised kets — and if we don’t, then the theory still gives the correct result.)

The first-order energy correction is the expectation of the perturbation in the unperturbed state.

CWPP 15/2/2013
Further Quantum Physics Non-degenerate Perturbation Theory 7

First-Order Correction to ket

The correction to the ket |n1 i is found by taking the inner product with any other unper-
turbed ket |m0 i:

hm0 | H1 |n0 i + hm0 | H0 |n1 i = En(1) h m0 | n0 i + En(0) h m0 | n1 i .

(1)
The term involving En vanishes and the second term on the left becomes (same argument)
(0)
Em h m0 | n1 i. Thus

hm0 | H1 |n0 i
h m0 | n1 i = (0) (0)
m 6= n
En − Em

This defines all the amplitudes for |n1 i except h n0 | n1 i. But that is just zero because they
are orthogonal. So

X |m0 i hm0 | H1 |n0 i

|n1 i = (0) (0)
m6=n En − Em

CWPP 15/2/2013
Further Quantum Physics Non-degenerate Perturbation Theory 8

Reflections on First-Order Results

(1) That the energy correction is just the expectation is unforgettable — what else could it
be? How else do you get a number out of an operator (H1 ) and a ket (|n0 i)? (The answer of
course is an eigenvalue, but that’s exactly what we trying to avoid!) The result is much more
memorable than its derivation.

(2) To look at it another way, eigenvalues are special cases of expectation values:

H |ni = En |ni → hn| H |ni = En = hn| H0 |ni + hn| H1 |ni

If we make the fundamental assumption that the unperturbed ket is a good approximation to
(0)
the exact ket we immediately get the same result En ≈ En + hn0 | H1 |n0 i.

(3) If we rearrange the first-order equation we find out why we don’t determine h n0 | n1 i:
 
H0 |n1 i − En(0) (1)
|n1 i = − H1 − En |n0 i

The corresponding Schrodinger equation is a linear differential equation with an inhomoge-

neous term. The solution we have found is the particular integral, but we can add any multi-
ple of the complementary function, which is |n0 i!

CWPP 15/2/2013
Further Quantum Physics Non-degenerate Perturbation Theory 9

Second-Order Correction to Energy

We only find the second-order correction to the energy, not the ket. This is because we are
usually more interested in eigenvalues than kets, and in general a p’th-order wavefunction
enables us to find an eigenvalue correct to order 2p + 1. (So the first-order ket gives us the
third-order energy as well . . .)

H1 |n1 i + H0 |n2 i = En(2) |n0 i + En(1) |n1 i + En(0) |n2 i

(2)
As before there are two things we don’t know, En and |n2 i.
And as before we can eliminate all the unknown kets by taking the inner product with |n0 i:

En(2) = hn0 | H1 |n1 i

We can of course substitute for |n1 i:

X hn0 | H1 |m0 i hm0 | H1 |n0 i X | hm0 | H1 |n0 i |2

En(2) = (0) (0)
= (0) (0)
m6=n En − Em m6=n En − Em

CWPP 15/2/2013
Further Quantum Physics Non-degenerate Perturbation Theory 10

Second-Order Correction to Energy

(Alternatively we could eliminate the H1 matrix elements to get
X 
En(2) = En(0) − (0)
Em | h m0 | n1 i |2
m6=n

but this is by no means a normal move!)

These equations indicate the close link between the first-order correction to the ket and the
second-order correction to the energy: you can’t have one without the other! The first form
(2) (2)
shows that if |n1 i = 0 then so is En . The final form shows that En could only be zero by
some flukey cancellation in the sum over amplitudes.
∗
There is an interesting symmetry of the mixing amplitudes h m0 | n1 i: h m0 | n1 i = − h n0 | m1 i .
Thus if a bit of state m is mixed into n by the perturbation then there is an equal but oppo-
site mixing of n into m. This is exactly what we need to preserve orthogonality to first order.

Then we see that states that mix in first order repel in second order. And hence the second-
order correction to the ground state is always negative.

CWPP 15/2/2013
Further Quantum Physics Non-degenerate Perturbation Theory 11

Example 1: Box with a non-flat bottom

For our first example we take the particle in a box (between 0 and a) with a perturbation:
 
2πx
H1 = W cos .
a

n2 π 2 h̄2
The unperturbed eigenvalues are En(0)
= 2
= n2 E1 (where n = 1, 2, 3 . . .)
2ma r
2 h nπx i
and the eigenkets have a simple x-representation h x| n0 i = un (x) = sin .
a a
The first-order energy shift is then
Z a
En(1) = hn0 | H1 |n0 i = un (x) H1 un (x) dx
0

The integral is helped by the use of an addition formula:

r   h i r     
2 2πx nπx W 2 (n + 2)πx (n − 2)πx
H1 un (x) = W cos sin = sin + sin .
a a a 2 a a a

This simplifies to W
2 (un+2 − un−2 ) except for the special cases of n = 1 and n = 2. For n = 2,
sin (n − 2)πx/a = 0 while for n = 1 sin (n − 2)πx/a = − sin πx/a.

CWPP 15/2/2013
Further Quantum Physics Non-degenerate Perturbation Theory 12

Example 1: Box with a non-flat bottom

We can summarise the action of H1 on the eigenkets as follows:
  
 W

 |(n + 2)0 i + |(n − 2)0 i for n > 2

 2
W
H1 |n0 i = |40 i for n = 2

 2  

 W |3 i − |1 i

for n = 1
0 0
2

(1)
Thus almost all the En vanish because H1 |n0 i is orthogonal to |n0 i. The only exception is

(1) W  W
E1 = h10 | H1 |10 i = h 10 | 30 i − h 10 | 10 i = −
2 2

This also gives us all the matrix elements of H1 to evaluate the ket correction:
 
W |(n − 2)0 i |(n + 2)0 i
|n1 i = + (only 1 term for n=1,2)
2 [n2 − (n − 2)2 ]E1 [n2 − (n + 2)2 ]E1

CWPP 15/2/2013
Further Quantum Physics Non-degenerate Perturbation Theory 13

Finally this allows us to calculate the second-order correction to the energy:

 
W2 1 1
En(2) = + 2 (only 1 term for n=1,2)
4 [n2 − (n − 2)2 ]E1 [n − (n + 2)2 ]E1
h i h i
(2) W2 −1 (2) W2 1 1
For example E1 = 4 8E1 and E3 = 4 8E1 − 16E1 .

This problem can be solved exactly - the Schrodinger equation becomes the Mathieu equa-
tion, and the wavefunctions are Mathieu functions. There is a Mathematica notebook on the
FQM Website which compares the exact and perturbative solutions.

Example 2: Quadratic Stark Effect in Hydrogen

The Stark effect is a perturbation by an electric field E = −∇Φ. There are two charges in the
atom, so the change in the Hamiltonian is H1 = +eΦ(0) − eΦ(r).
For a uniform E in the z-direction, magnitude E, Φ(r) = Φ(0) − Ez so H1 = eEz.

CWPP 15/2/2013
Further Quantum Physics Non-degenerate Perturbation Theory 14

This Hamiltonian has a useful commutator with the Parity operator P .

[H0 , P ] = 0, so the hydrogen eigenkets can be chosen to be eigenkets of P .
(And these are the |nℓmi kets.)
But P H1 = −H1 P , so H1 anti-commutes with P : P H1 + H1 P = 0.
Consider a matrix element between two parity eigenstates (p = ±1):

hn′ ℓ′ m′ p′ | P H1 + H1 P |nℓmpi = (p + p′ ) hn′ ℓ′ m′ p′ | H1 |nℓmpi = 0

So either p = −p′ , the states have opposite parity, or the matrix element is zero.
Thus all diagonal matrix elements vanish and so does the first-order energy (unless there is
degeneracy). (This argument is perfectly general and applies to all atomic states — the Stark
effect is usually only second-order.) But in Hydrogen there is degeneracy. . . .

(2)
X | hn′ ℓ′ 0| eEz |100i |2
But in the non-degenerate ground state: E1 = (0) (0)
.
n′ ℓ′ 6=10 E1 − En

CWPP 15/2/2013
Further Quantum Physics Degenerate Perturbation Theory 15

Example of Degeneracy: Rigid Rotor

This is a system with moment of inertia I free to rotate about an axis.
L2
The Hamiltonian is thus H0 = .
2I
(0) n2 h̄2
The energy levels are En = .
2I
1
The ground state is non-degenerate: h φ| 00 i = √
2π
But all the excited states are two-fold degenerate, so any basis eigenkets represent an arbi-
trary choice. Two obvious choices are:

c,s basis: +,− basis:

1
1 h φ| (1, +)0 i = √ exp(+iφ)
h φ| (1, c)0 i = √ cos φ 2π
π
1
1 h φ| (1, −)0 i = √ exp(−iφ)
h φ| (1, s)0 i = √ sin φ 2π
π

but any linear combination of the degenerate eigenkets is an eigenket.

This means the fundamental assumption is no longer necessarily true!

CWPP 15/2/2013
Further Quantum Physics Degenerate Perturbation Theory 16

In general a perturbation will remove the degeneracy in En . Imagine what happens if we

could control the pertubation parameter and make it smaller. As the states approach each
other they are each some particular linear combination of our arbitrary basis — but we don’t
know what it is.
So the correct zeroth-order wavefunctions |(n, z)0 i are an unknown linear combination of our
original, arbitrary basis |(n, a)0 i:
X
|(n, z)0 i = |(n, a)0 i h (n, a)0 | (n, z)0 i .
a

We can now put this (correct but as yet unknown!) ket into the first-order equation:

(1)
H1 |(n, z)0 i + H0 |(n, z)1 i = Enz |(n, z)0 i + En(0) |(n, z)1 i

We take the inner product with one of our arbitrary basis kets for En :

(1)
h(n, a)0 | H1 |(n, z)0 i + h(n, a)0 | H0 |(n, z)1 i = Enz h (n, a)0 | (n, z)0 i + En(0) h (n, a)0 | (n, z)1 i

Two terms cancel in the same way as the non-degenerate case.

CWPP 15/2/2013
Further Quantum Physics Degenerate Perturbation Theory 17

This leaves just

(1)
h(n, a)0 | H1 |(n, z)0 i = Enz h (n, a)0 | (n, z)0 i
We simply insert the summation for the unknown state:
X
h(n, a)0 | H1 |(n, a′ )0 i h (n, a′ )0 | (n, z)0 i = Enz
(1)
h (n, a)0 | (n, z)0 i .
a′

We have as many equations like this as the order of the degeneracy.

Remembering that h (n, a)0 | (n, z)0 i are elements of a vector, and h(n, a)0 | H1 |(n, a′ )0 i ele-
ments of a matrix, we can write:

H1 c(z) = En(1) c(z)

An eigenvalue equation!! With a Hermitian matrix in the known (but arbitrary) a basis.
This gives us:
(1)
Eigenvalues: Each eigenvalue is one of the first-order energy shifts En ;
Eigenvectors: Each component of the eigenvector gives us one of the amplitudes
h (n, a)0 | (n, z)0 i which define the correct zeroth-order z basis in terms of the a basis.
Again, the result is more memorable than the derivation.

CWPP 15/2/2013
Further Quantum Physics Degenerate Perturbation Theory 18

Example: Perturbed Rotor

We perturb the rigid rotor with an angle-dependent potential V0 sin 2φ.
We choose the +,− basis and find the shifts in n = 1.
The integrals in the matrix elements are of the form
Z 2π  i2φ −i2φ

1   ∗ e −e
e±iφ V0 e±iφ dφ
2π 0 2i
This gives
    
0 −iV0 /2 c+ (1) c+
= E1
iV0 /2 0 c− c−
(1)
We find E1 = ±V0 /2 and eigenvectors
   
V0 1 1 V0 1 1
+ : c(+) = √ − : c(−) = √
2 2 i 2 2 −i
Thus the wavefunction for the state shifted up in energy is
1 iφ −iφ

1 + i 
√ e + ie = √ cos φ + sin φ .
2 π 2 π
If we had used these states to start with, the H1 matrix would have been diagonal, with the
eigenvalues as the diagonal elements.

CWPP 15/2/2013
Further Quantum Physics Degenerate Perturbation Theory 19

How to avoid Degenerate Perturbation Theory.

It’s obvious that, compared with just evaluating an expectation value in the non-degenerate
case, Degenerate Perturbation Theory is a big deal.
If we knew the correct zeroth-order wavefunctions to start with we would only have to evalu-
ate the expectation values.

Suppose there is an operator S which commutes with both H0 and H1 . Then it commutes
with the exact H, and the exact eigenstates are eigenstates of S. So we just use the mutual
eigenbasis of H0 and S, |(n, s)0 i.
In this basis the off-diagonal elements are all zero (as long as the degenerate states have dif-
ferent s eigenvalues):

h(n, s)0 | SH1 − H1 S |(n, s′ )0 i = (s − s′ ) h(n, s)0 | H1 |(n, s′ )0 i = 0

In the example given the relevant operator is reflection about the line φ = π/4, 5π/4 so that
the correct eigenfunctions are cos(φ − π/4) and sin(φ − π/4).

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 20

We can make the Hamiltonion time-dependent if we introduce an interaction with a time-

varying external field. We shall consider the interaction is weak enough to be treated by per-
turbation theory:
H = H0 + V (t)

where the eigenvalues and eigenkets of H0 are known.

Suppose we prepare the system in an eigenstate — say the ground state.
And V varies harmonically — say as cos ωt.
And there’s an excited state at an energy E above the ground state.
This is called absorption spectroscopy. What do we expect?

The example suggests that if E = h̄ω then the system responds and can end up in the excited
state, and if not, not.

So our physical intuition suggests that the result is discontinuous! This is a hint that we need
to be a bit careful. The combination of discrete energy levels and perfectly-defined frequen-
cies is obviously a tricky one. We shall proceed with caution!

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 21

Basic equations.
The eigenkets of H0 form a complete set, so we can always expand a state in terms of them.
If V = 0 then we have seen that the state is in general
X
|ψ(t)i = an exp (−iEn t/h̄) |ni
n

so a sensible place to start is to assume an expansion of this form but allowing an to vary
with t:
X
|ψ(t)i = an (t) exp (−iEn t/h̄) |ni .
n

Substitute into the time-dependent Schrodinger equation ih̄d/dt |ψ(t)i = (H0 + V ) |ψ(t)i:
X  dan  X  
ih̄ + an En exp (−iEn t/h̄) |ni = an exp (−iEn t/h̄) En + V (t) |ni .
n
dt n

As expected, there is a cancellation so that an only changes because of V :

X dan X
ih̄ exp (−iEn t/h̄) |ni = an exp (−iEn t/h̄)V (t) |ni .
n
dt n

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 22

Take inner product with |mi:

dam X
ih̄ exp (−iEm t/h̄) = an exp (−iEn t/h̄) hm| V (t) |ni .
dt n

or, taking the time-dependence across:

X  
dam i(Em − En )t
ih̄ = an exp hm| V (t) |ni .
dt n
h̄
which has an obvious vector/matrix shape:
da e
ih̄ = Va
dt
where Ve has matrix elements that oscillate due to the energy difference between states as
well as due to the time-dependence of V (t):
 
i(E m − E n )t
Vemn = exp hm| V (t) |ni
h̄
We haven’t so far made any approximations.
If the number of states is finite and fairly small we can solve these coupled linear equations
exactly — numerically if necessary.
But in many problems the number of states involved is too large and we must approximate.

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 23

Perturbative Expansion.

We expand am in powers of the size of V (t):

(0) (1)
am = a(0) (1) (2)
m + am + am + . . . where am is independent of V , am ∝ V etc

Substituting this into the equation we get

!
(0)
dam
(1)
dam
(2)
dam X  
ih̄ + + +... = e (0) (1) (2)
Vmn an + an + an + . . . .
dt dt dt n

Equating powers of V we get a series of equations for the rate of change of each coefficient in
terms of the next lower coefficient:
(0)
dam
Zeroth Order: ih̄ =0 → a(0) (0)
m (t) = am (t0 ).
dt
(1) X X Z t 
dam 1
First Order: ih̄ = Vemn a(0)
n → a (1)
m (t) = Vemn dt′ a(0)
n
dt n
ih̄ n t0
(2)
dam X 1 X Z t 
Second Order: ih̄ = Vemn a(1)
n → a(2)
m (t) = Vemn a(1) ′
n (t ) dt
′
dt n
ih̄ n t0

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 24

We shall only consider particularly simple cases:

(0)
The initial state is simple: |ψ(t0 )i = |ii, so ai = 1 and all the other a(0) are 0.
We also ignore the possibility of diagonal matrix elements Vemm = 0. (This is quite com-
mon in any case for parity reasons).
We work to first order only.
Then at a later time
Z t
1 (0) (0)
a(1)
m (t) = Vemi dt for m 6= i and ai (t) = ai (t0 ) = 1
ih̄ t0

Example 1: Turning on a perturbation.


Consider turning on a otherwise time-independent perturbation, V (t) = eαt V for t < 0
V for t > 0
We look at the state at t = 0:
Z 0  
hm| V |ii i(Em − Ei )t αt hm| V |ii
a(1)
m (0) = exp e dt → a(1)
m (0) =
ih̄ −∞ h̄ Ei − Em + ih̄α

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 25

Sudden and Adiabatic Approximations.

This has two interesting limits:

(1)
α → ∞: This corresponds to a very quick (‘sudden’) turn-on. In this case am → 0: the state
at t = 0 is unchanged from t = −∞ (apart from normal time-evolution of |ii).
But after t = 0 we have a Hamiltonian H0 + V , so the time evolution after t = 0 is in terms of
the eigenkets of the new Hamiltonian |n′ i:
X X ′
′ ′
|ψ(0)i = |n i h n | ii and so |ψ(t)i = e−iEn t/h̄ |n′ i h n′ | ii
n′ n′

α → 0: This corresponds to a very slow turn on, and at t = 0 the state is

X hm| V |ii
|ψ(0)i = |ii + |mi = |i′ i to first order.
Ei − Em
m6=i

In this case the wavefunction follows the evolving Hamiltonian in the corresponding eigenket.
This is known as the adiabatic approximation, and requires h̄α ≪ |Ei − Em |.

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 26

Example 2: Transition due to an Oscillating Perturbation.

We now consider an oscillating perturbation, such as we might get from applying EM radia-
tion to an atom:
V = −E(t) · D = −2EDz cos ωt
where D is the electric dipole moment of the atom, and I have made the radiation linearly
polarized in the z-direction with amplitude 2E.
We can split the cosine up: cos ωt = (1/2)[eiωt + e−iωt ].
Since the first-order theory is linear in the perturbation we can treat just one term, which we
choose to be e−iωt :
V (t) = −EDz e−iωt .
Then, if the perturbation is turned on at t = 0 when the atom is in state |ii, and turned off
at t: Z t  
′
hm| − ED z |ii i(E m − E i )t
a(1)
m (t) = exp exp −iωt′ dt′
ih̄ 0 h̄
exp (i ∆E t/h̄) − 1
= hm| − EDz |ii where ∆E = Em − h̄ω − Ei
−∆E
hm| − EDz |ii  
= −i exp (i ∆E t/2h̄) sin (∆E t/2h̄)
∆E/2

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 27

The probability of the atom being in state |mi at time t is

2 2
 
4| hm| E Dz |ii | ∆E t
|am (t)|2 = sin2
∆E 2 2h̄
The probability oscillates, and the amplitude of oscillation involves the square of the ratio of
the matrix element squared divided by ∆E. This ratio is typically tiny except when ∆E ≈ 0.
This has a natural interpretation in terms of the absorption of a quantum of energy h̄ω — a
photon in the case of light. But it will happen for any oscillating perturbation.
However ∆E = Ei + h̄ω − Em is the energy imbalance in the process — we have appar-
ently lost/gained an energy of ±∆E — but only for about one half-cycle of the oscillation
2πh̄/∆E. So the energy imbalance is constrained by
∆E ∆t ∼ 2πh̄.
This is the process of ‘borrowing energy’ mentioned in many popular accounts of QM.
If there happens to be a final state |mi exactly on resonance, ∆E = 0, then the amplitude
does not oscillate but grows linearly with t:
hm| − EDz |ii
a(1)
m (t) = t.
ih̄
This is the ‘discontinuity’ we naively expected.

CWPP 15/2/2013
Further Quantum Physics Continuum States 28

Continuum States — the Free Particle

For a free particle the Hamiltonian is H = p̂2 /2m and this obviously commutes with p̂.
So we can choose the eigenstates to be the eigenkets of p̂: p̂ |P i = P |P i. Let h x| P i = uP (x).

∂uP
−ih̄ = P uP → uP (x) = A exp[iP x/h̄]
∂x
This is the solution for any value of P , so the possible values of momentum are P = any-
thing, with energy eigenvalues P 2 /2m — not discrete eigenvalues but a continuous range.
A is a normalisation constant — but the eigenfunctions are not normalisable!
These two problems always occur together: eigenkets belonging to eigenvalues in a continuous
range are unnormalisable. The whole set of states are sometimes referred to as the contin-
uum.
′ ′
√
The standard normalisation is ‘δ-function normalisation’: h P | P i = δ(P − P ); A = 1/ 2πh̄.
But if we think about the uncertainty principle the situation is obvious: in an eigenket the
momentum is sharply defined, ∆p = 0 and so the position uncertainty must be infinite.
In fact the eigenket |P i is physically unrealisable. The same is true of the position eigenket
|Xi. They still form a basis but physical states must be formed by superposition to make a
wavepacket.

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 29

Example 3: Fermi’s Golden Rule

The ‘almost discontinuous’ result follows from the combination of discrete energy levels and a
single frequency perturbation. The theory looks a bit different when the final state belongs to
the continuum, because there is always a final state exactly on resonance.
We consider a transition from an initial discrete state |ii, energy Ei to a final continuum |pi,
energy E(p), induced by a perturbation V (t) = V e−iωt .
The first-order amplitude is as above (p42, last line):
(1) hp| V |ii  
ap (t) = −i exp (i ∆E t/2h̄) sin (∆E t/2h̄)
∆E/2
where ∆E = E(p) − h̄ω − Ei . These amplitudes define the first-order (∝ V ) ket correction:
Z ∞
|ψ(t)i = |ii exp(−iEi t/h̄) + a(1)
p (t) exp(−iEp t/h̄) |pi dp
0
(1)
The excitation probability is given by |ap |2 summed over the final states:
Z ∞ Z ∞
| hp| V |ii |2 t2
|a(1)
p (t)|
2
dp = 2 sinc2 (∆E t/2h̄) dp
0 0 h̄

where we have introduced the sinc function sinc x = sin x/x.

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 30

The sinc function has a peak at x = 0, falling to zero at x = ±π. Thus for any value of t the
sinc function gives particular weight to final states E(p) for which −2πh̄/t ≤ ∆E ≤ 2πh̄/t.
As time goes on this range gets steadily narrower, emphasising states around the energy con-
serving state ∆E = 0. It makes sense to change variable to the argument of the sinc function:
dE(p) t t
x = (E(p) − Ei − h̄ω)t/2h̄ and dx = dp = dp
dp 2h̄ 2h̄ρf (E(p))

ρf = dp/dE(p) is known as the density of states factor. It is a measure of the ‘quantity’ of

final states per unit energy range. (The lower limit xmin is large and negative.)
Z ∞ Z ∞
t
|a(1) 2
p (t)| dp = | hp| V |ii |2 ρf (Ep ) sinc2 (x) dx
0 2h̄ xmin
Fermi’s Golden Rule approximates this result in the limit that the rapid variation of the sinc
function takes place on a scale over which the other factors scarcely change: we can take
them out of the integral (evaluated at the peak, ∆E = 0 or Ep = Ei + h̄ω), and we can
formally extend the range to ±∞.
Z ∞ 2 Z ∞
| hp| V |ii | 2t
|a(1) 2
p (T )| dp ≈ ρf (E(p)) sinc2 (x) dx
0 h̄ −∞

CWPP 15/2/2013
Further Quantum Physics Time-dependent Perturbation Theory 31

The integral is π, so finally, expressing the result as a rate of excitation:

2π
Fermi’s Golden Rule: Rate = | hp| V |ii |2 ρ(E).
h̄

Simplest Possible Example of Fermi’s Golden Rule

The simplest system with a bound state as well as a continuum is a negative delta-function
potential V (x) = −W δ(x) — a sort of limiting narrow deep square well.
This has a single bound state:

mW 2 h̄
|ii with Ei = − and h x| ii = √ exp (−mW |x|/h̄2 )
2h̄2 mW

and two continuum states for every positive value of E, one even parity and one odd parity:
(

h̄2 p2 h x| p, −i = √1 sin px
|p, ±i with E(p) = and
πh̄  h̄ 
2m h x| p, +i = √1
πh̄
cos p|x|h̄ +φ where tan φ = πmW
ph̄

We perturb this system with −F xe−iωt , so that only the odd-parity states contribute. Thus
we use the above results with |pi replaced by |p, −i.

CWPP 15/2/2013
Further Quantum Physics Summary of Time-dep’t Pert Theory. 32

Schrodinger Equation:
d
ih̄ |ψi = H0 |ψi + V (t) |ψi
dt
so in a time δt the increment to the ket is
 
−iH0 −iV (t)
δ|ψi = |ψi + |ψi δt.
h̄ h̄
The first term provides the usual time-dependence for energy eigenstates |Ei, rotating them
(clockwise!) in the complex plane with angular frequency E/h̄.
The second term gives the transitions between eigenstates.
We can project the ket increment onto the energy eigenstates (just the second term):
X iV (t)
δ|ψi = |Ei hE| |ψi δt
h̄
E

so so to get to a particular final state Ef this must be non-zero: the ket increment must not
be orthogonal to |Ei (and the more similar they are the bigger the probability of transition).
The initial state is rotating in complex plane at Ei /h̄; the little bit of state |Ef i rotates at
Ef /h̄, so the successive increments cancel out unless the time-dependence of V (t) contains a
part which rotates at the correct difference frequency (Ef − Ei )/h̄.

CWPP 15/2/2013
Further Quantum Physics Trinity Term Outline 33

This gives the expected distribution of material in Trinity Term by lecture:

1 Review of Time-dep’t Perturbation Theory/ Selection Rules

2 Selection Rules / Identical Particles
3 Identical Particles / Atomic Hamiltonian
4 Atomic Units: Hydrogen, Helium
5 Helium: Exact Symmetries, Perturbation Theory
6 Helium: Variational Principle
7 Helium: quick look at modern calculations / Questions.

CWPP 15/2/2013
Further Quantum Physics Selection Rules 34

Selection Rules are Conservation Laws

Of all the possible transitions between states of an atom only a quite small subset occur at
any significant rate. These possible transitions are determined by Selection Rules.
One way of looking at selection rules is that they represent conservation laws — a transition
not obeying the selection rule would violate the conservation law.
The obvious conservation laws are:
Energy: Ei = Ef ±h̄ω (for emission/absorption). This is Bohr’s Frequency Rule, and we
have seen how it comes out of time-dependent perurbation theory. It’s not a constraint
on what transitions i → f can occur.
Momentum: This is also conserved in radiative processes, but the emitted or absorbed
momentum is taken up by the atom as a whole.
So more exactly, for emission of a photon with wavevector k = (ω/c)n:
p2i
Ei = Ef + h̄ω E i = Ei +
2M
where
pi = pf + h̄k p2f
E f = Ef +
2M
This does not constrain the transitions i → f either.

CWPP 15/2/2013
Further Quantum Physics Selection Rules 35

Angular Momentum: This turns out to give an important selection rule.

A.M of initial state = A.M of final state + A.M. of radiation Field

The Total A.M. of the electrons in the atom is usually given the symbol J .
(If the nucleus is included as well it’s called F , but strictly speaking the selection rule applies
to whatever is the total A.M. of the atom.)
The A.M. of the radiation field can’t be less than 1h̄, (the spin of the photon), but can be
more: spatial structure in the field can carry additional A.M.
Different interaction multipoles (dipole, quadrupole . . .) carry different amounts of A.M.
We shall just consider the lowest multipole (dipole) and the strongest case (electric).
P
In this case the A.M. is 1h̄, and the interaction Hamiltonian is −E · D, where D = −e i ri
summed over the electrons.

The addition of the angular momentum of the two parts of the system require the Rules for
the Addition of Angular Momenta:

CWPP 15/2/2013
Further Quantum Physics Selection Rules 36

Rules for the Addition of Angular Momenta

The possible angular momentum quantum numbers for a system composed of two subsystems
with angular momenta quantum numbers of J1 and J2 are given by J where:
The maximum value of J is Jmax = J1 + J2 ;
The minimum value of J is Jmin = |J1 − J2 |;
J goes from Jmin to Jmax in steps of 1.
The selection rule is simply that there must be a way of couple the two angular momenta of
the final state to give the same as the initial state:

So if the final state involves adding Jf and 1, the possible final angular momenta are from
Jf + 1 to |Jf − 1|. Obviously this depends on the value of Jf :

If Jf ≥ 1 then the range of possibilities is Ji = Jf + 1, Jf , Jf − 1.

If Jf = 12 then we can have Ji = 32 or 12 .
If Jf = 0 then the only possible Ji is 1.

We can summarize this as ∆J = 0, ±1 0 6→ 0.

This expresses the exchange of one unit of A.M. between the atom and the radiation field.

CWPP 15/2/2013
Further Quantum Physics Selection Rules 37

Selection Rules are Symmetries

Another way of looking at selection rules is that they follow from symmetries. For example
the atomic Hamiltonian always commutes with the Parity operator P :
[P, H] = 0
and hence we can choose atomic eigenstates to be eigenstates of P . Thus all atomic states
have a parity quantum number of p = ±1.
However the electric dipole interaction Hamiltonian anti-commutes with P :
P E · D = −E · D P or {P, Hint } = 0.
(This is because P acts on D and not E. If we enlarge the scope of P to act on the field vari-
ables as well then we recover the expected commutator, not anti-commutator.)
Take a matrix element of the anti-commutator (repeat of Stark effect argument):
hf | P D + DP |ii = (pf + pi ) hf | D |ii = 0
So either pi = −pf , the states have opposite parity, or the matrix element is zero.

So the Selection Rule is Parity Changes which is usually summarised as YES

(Conservation law view: Parity is conserved, but both and atom and field change.)

CWPP 15/2/2013
Further Quantum Physics Selection Rules 38

Selection Rules follow from Commutators

We have already seen this for P ; it is also true for J . The relevant commutator is

[Ji , Dj ] = iǫijk h̄Dk or [Jx , Dy ] = ih̄Dz and cyclically.

This is true simply because D is a vector operating on the atomic variables, and J is the to-
tal atomic angular momentum. This is a very powerful argument because of its generality:
S(α) = (1 − iαJz /h̄) rotates the atomic state by the infinitesimal angle α around the z-axis.
When we do this the matrix elements of a vector operator must behave as vectors:

hf | S † (α)Dx S(α) |ii = hf | Dx |ii − α hf | Dy |ii .

If we keep terms to first order on the left:

iα
hf | Dx |ii + hf | Jz Dx − Dx Jz |ii = hf | Dx |ii − α hf | Dy |ii ,
h̄

giving the zx version of the commutator.

CWPP 15/2/2013
Further Quantum Physics Selection Rules 39

We now use the fact that the free-atom states will be eigenstates of angular momentum:
replace |ii with |Ji Mi i and similarly for the final state. Then the three commutators with Jz
give the selection rule on M :

[Jz , Dz ] = 0 : hJf Mf | Jz Dz − Dz Jz |Ji Mi i = (Mf − Mi )h̄ hJf Mf | Dz |Ji Mi i = 0

The other two are [Jz , Dx ] = ih̄Dy and [Jz , Dy ] = −ih̄Dx or [Jz , Dx ± iDy ] = ±h̄(Dx ± iDy ).
Again we take a general matrix element:

hJf Mf | Jz D± − D± Jz |Ji Mi i = (Mf − Mi )h̄ hJf Mf | D± |Ji Mi i = ±h̄ hJf Mf | D± |Ji Mi i

(Mf − Mi ∓ 1)h̄ hJf Mf | D± |Ji Mi i = 0.

We can summarize this as ∆MJ = 0, ±1.

We can also derive the ∆J rule from a commutator.

CWPP 15/2/2013
Further Quantum Physics Selection Rules 40

We can use the commutator method to generate more selection rules:

X X
J = L + S where L = l i and S = si
electrons electrons

(Note the emerging typographical convention: we use upper case letters for whole atom quan-
tities and lower case for single electron quantities.)
Now the transition operator D does not involve S at all:

[Si , Dj ] = 0 and hence [Li , Dj ] = [Ji − Si , Dj ] = ih̄ǫijk Dk .

If the atomic state is an eigenstate of L2 and S 2 :



hLf Sf Jf Mf | S D−DS |Li Si Ji Mi i = Sf (Sf +1)−Si (Si +1) h̄2 hLf Sf Jf M f | D |Li Si Ji Mi i = 0
2 2

which means the rate is zero unless ∆S = 0.

The second commutator is the same as the J commutator so the selection rule is identical:
∆L = 0, ±1 0 6→ 0.

In multi-electron atoms these rules are only approximate because whereas the state of the
free atom is an eigenstate of J 2 , it is only approximately an eigenstate of L2 and S 2 .

CWPP 15/2/2013
Further Quantum Physics Identical particles and Pauli Principle 41

Suppose we have a state of two non-interacting particles in the same gravitational potential
(I’m only using gravity because if they are non-interacting they had better be electrically
neutral!):
p21 p22
H= + + mφ(r1 ) + mφ(r2 )
2m 2m
This acts on a ket |ψi, and there is a complete set of basis kets |r1 , r2 i = |r1 i |r2 i represent-
ing a state in which particle i is at ri .
What sort of eigenstates can we expect to find?
There is an obvious symmetry — both particles are identical so they appear in the Hamil-
tonian in identical ways. (If they didn’t we could use the difference to distingush between
them.) To make it more formal we introduce an exchange operator P12 which swaps particle
labels:
p21 p22
P12 = etc and hence [P12 , H] = 0.
2m 2m
Thus it appears that eigenstates of H can be chosen to be eigenstates of P12 .
2
P12 = 1 and so the eigenstates are either symmetric P12 |ψi = |ψi or antisymmetric P12 |ψi =
− |ψi. But the vanishing [P12 , H] commutator doesn’t just apply to this situation: it will be
true for any other interaction we care to introduce. It simply depends on the indistinguisha-
bility of the two particles.

CWPP 15/2/2013
Further Quantum Physics Identical particles and Pauli Principle 42

Thus neither this Hamiltonian nor any other possible Hamiltonian can distinguish between
the two particles so [P12 , H] = 0 for all situations.
But this means that two particles in a symmetric state will always remain in a symmetric
state (compare the argument for the Parity selection above). So at the foundation of the
world some of these identical particles were put in symmetric and some into antisymmetric
states . . . this is getting ridiculous!

It is not so! For any given particle the wavefunction is either always symmetric under P12 , or
always antisymmetric, depending on the particle. Particles with symmetric wavefunctions
obey Bose-Einstein statistics, and are known as bosons, and particles with antisymmetric
wavefunctions obey Fermi-Dirac statistics and are known as fermions.

How do we know which is which? The Spin-Statistics Theorem states:

Particles with integer spin are bosons, and particles with half-integer spin are fermions.

Thus electrons, protons, neutrons, 3 He nuclei are fermions, deuterium nuclei, hydrogen atoms
and alpha particles are bosons.
That electron wavefunctions are antisymmetric under P12 is known as the Pauli principle.
(Nomenclature complex here.)

CWPP 15/2/2013
Further Quantum Physics The Atomic Hamiltonian 43

The atom consists of N electrons (charge −e, mass m, labelled 1 to N ) and a nucleus (charge
Ze, mass M , labelled 0). The mass of the atom is M = M + N m.
The Hamiltonian is the energy function of the atom. The biggest terms are obvious:

X P2
i
H0 = Kinetic energy of the electrons
i
2m
X Ze2
− Electrostatic electron-nucleus potential energy
i
4πǫ0 Ri
X e2
+ Electrostatic electron-electron potential energy
i>j
4πǫ0 Rij

What does this leave out? There are at least five categories of omitted terms:
(1) Relativistic correction to non-relativistic kinetic energy
(2) Kinetic energy of the nucleus
(3) Finite size of the nuclear charge distribution
(4) Other corrections to the electron-nucleus interaction
(5) Corrections to the electron-electron interaction

CWPP 15/2/2013
Further Quantum Physics The Atomic Hamiltonian 44

1. Relativistic Correction to the Kinetic Energy

This takes the same form as in Hydrogen for each electron:

X Pi4
H1 = −
i
8m3 c2

2. Kinetic Energy of the Nucleus

Our starting point H0 treats the nucleus as fixed at the origin — and so, in effect, infinitely
massive. The configuration of the system is defined by the N positions R1 . . . RN .
We can easily give the nucleus a variable position R0 (with corresponding change Ri → Ri0
in the electron-nuclear distance) and then include a term

P02
H2 = .
2M

However it’s not so simple . . .

CWPP 15/2/2013
Further Quantum Physics The Atomic Hamiltonian 45

By giving the system three more degrees of freedom, it is no longer fixed to the origin - the
whole atom can now move! We need to remove this free-particle degree of freedom to dis-
cover what the energy of the system is in the centre-of-mass frame.
We do this with a co-ordinate change:
P
M R0 + j mRj
r0 = Centre of Mass
M
ri = R i − R 0 Electron i relative to Nucleus

We need to find out what the new conjugate momenta are — i.e. we have to find out what
the operators P = −ih̄∇0 and pi = −ih̄∇i represent.
In fact the algebra is shorter if we proceed backwards(!):

∂ ∂x0 ∂ X ∂xi ∂ ∂x0 ∂ X ∂xi ∂

PX0 = −ih̄ = −ih̄ − ih̄ PXi = −ih̄ − ih̄
∂X0 ∂X0 ∂x0 i
∂X0 ∂xi ∂Xi ∂x0 i
∂Xi ∂xi

Reading off the partial differentials

M X m X
P0 = P− pi Pi = P + pi and hence P = P0 + Pi
M i
M i

CWPP 15/2/2013
Further Quantum Physics The Atomic Hamiltonian 46

Thus P (the momentum conjugate to r0 ) represents the total momentum of the atom in the
original reference frame, and hence P /M represents the velocity of the centre of mass V .
Then pi = Pi − mV V = m(V − V ) is the momentum in the centre-of-mass frame.
Thus although our co-ordinate is ri = Ri − R0 , the position relative to the nucleus,
the conjugate momentum pi represents the mass × velocity relative to centre of mass.

The momentum of the nucleus in the centre-of-mass frame is P0 − M V ,

P
which we define to be p0 = P0 − M V = − i pi .

To complete the calculation we find the total kinetic energy in terms of the new momenta:

X P2 P02 X ( m P + p i )2 M
(M P + p 0 )2 P2 X p2 p20
i M i
+ = + = + +
i
2m 2M i
2m 2M 2M i
2m 2M

which nicely splits into the free-particle term, the electronic term (but relative to centre of
mass instead of fixed origin) and a new term representing the kinetic energy of the nucleus:
P 2
(− i pi )
H2 =
2M

CWPP 15/2/2013
Further Quantum Physics Atomic Units 47

Standard Atomic Units

If we adopt units in which Planck’s constant h̄, the elementary charge e, the mass of the elec-
tron me and 4πǫ0 are all 1 then we eliminate a lot of scaling factors in Hamiltonians, and
this is the standard set of atomic units. We can use these quantities to define units of mass,
length, time and current, and so they form a complete set of units:
Constant Dimensions
me [M ]
h̄ [M L2 T −1 ]
e2 /4πǫ0 [M L3 T −2 ]
Quantity Unit Name
h̄2
Length [L] Bohr radius a0
me e2 /4πǫ0
h̄3
Time [T ]
me (e2 /4πǫ0 )2
e2
Velocity [LT −1 ]
4πǫ0 h̄
me (e2 /4πǫ0 )2
Energy [M L2 T −2 ] Hartree H0
h̄2

CWPP 15/2/2013
Further Quantum Physics Atomic Units 48

Scaling the Hydrogen Hamiltonian

The Hamiltonian for the internal motion in any hydrogen-like atom (neglecting relativistic
terms) is
h̄2 2 h̄2 2 Ze2 h̄2 2 Ze2
H =− ∇ − ∇ − =− ∇ −
2me 2M 4πǫ0 r 2µ 4πǫ0 r
where the nucleus has charge Ze and mass M , and µ = me M/(me + M ) is the reduced mass.
There are quite a lot of different systems described by this Hamiltonian, with different values
of Z and µ, or ζ = µ/me :
System Z ζ
mp
Hydrogen 1 mp +me ≈ 1
md
Deuterium 1 md +me ≈ 1
4 mα
He+ 2 mα +me ≈ 1
mZ
Hydrogen-like Z Z mZ +me ≈ 1
mµ
Muonium 1 mµ +me ≈ 1
1
Positronium 1 2
mZ mµ
Muonic Z Z mZ +mµ me ≈ 207

CWPP 15/2/2013
Further Quantum Physics Atomic Units 49

Scaled Atomic Units

However the constants appearing in this Hamiltonion are not the ones we used to define the
units: we have µ instead of me and Ze2 instead of e2 . With these replacements we get:
h̄2 a0
Length: a(Z, ζ) = =
µZe2 /4πǫ0 Zζ
µZ 2 e4
Energy: H(Z, ζ) = 2 = ζZ 2 H0 .
(4πǫ0 )2 h̄
We switch to these units by defining r = a(Z, ζ)r E = H(Z, ζ)E and this strips out all
the constants from the Schrödinger equation:

1 1
− ∇2 ψ − ψ = Eψ.
2 |r|

1 e−|r|
This has eigenvalues Enlm = − 2 and the wavefunctions are, for example, ψ100 = √ .
2n π
This is a remarkable result: all these distinct systems have the same eigenvalues and wave-
functions — apart from the scaling factors. All the system properties for hydrogen can be
scaled appropriately and apply to positronium or He+ .

CWPP 15/2/2013
Further Quantum Physics Atomic Units 50

However if we add in the relativistic correction terms in H this is no longer exactly true:
they have system-dependent scaling factors even after the switch to these variables. (Also,
in the case of Ps, we may have to reconsider terms neglected because of their ‘nuclear’ mass-
dependence)

Consider the relativistic correction to the electron kinetic energy:

   
p4 h̄4 α 2
H1 = − =− p4 = (Z 2 ζ 3 )H(Z, ζ) − p4 .
8m3 c2 8m3 c2 a4 8

Thus at this level the system-dependent constants remain: we see that in He+ this term is
relatively Z 2 = 4 times bigger than in Hydrogen (neglecting the small change in ζ), and abso-
lutely Z 4 = 16 times bigger.

In positronium it is relatively smaller by ζ 3 = 1/8 and absolutely smaller by 1/16, and we

have to include the corresponding term for the positron which in an ordinary nuclear atom is
much smaller because of the M 3 dependence.

CWPP 15/2/2013
Further Quantum Physics Atomic Units 51

Application to Helium-like systems

An atom with two electrons has the Hamiltonian (neglecting all relativistic and magnetic
terms)
X  p2 Ze 2

e2 (−p1 − p2 )2
i
H= − + +
i=1,2
2m e 4πǫ0 r i 4πǫ0 r 12 2M

If we expand out the nuclear kinetic energy:

X  p2 Ze 2

e2 p1 · p2
i
H= − + +
i=1,2
2µ 4πǫ0 ri 4πǫ0 r12 M

It looks as though the same Z and µ-scaled atomic units will simplify H:
 
X p2  
 i 1 1 1 µ
H = H(Z, ζ) − + + p1 · p2  .
i=1,2
2 |ri | Z |r12 | M

But in this case we can’t eliminate the system-dependent factors from the scaled Hamiltonion
— the fact that we can for Hydrogen is (yet another) special result. Every He-like system is
a fundamenatally different problem. We shall therefore make the usual choice of mass-scaled
but not Z-scaled units:

CWPP 15/2/2013
Further Quantum Physics Helium 52

Exact Analysis: Symmetries

 
X p2  
 i Z 1 µ
H = H(1, ζ) − + + p1 · p2  .
i=1,2
2 |ri | |r12 | M

We start with some analysis on H to find out what properties its eigenkets will have.
A complete set of kets is provided by position and spin eigenstates |r1 , r2 , σ1 , σ2 i in which
particle i is at ri , and has z-component of spin σi which can be either ± 12 .
Thus h a, b, α, β| ψi where a, b are position vectors, and α, β are ±1, is the amplitude for the
configuration with electron 1 at a with spin α and electron 2 at b with spin β.
Obviously [P12 , H] = 0 and we require fermionic antisymmetry:

ha, b, α, β| P12 |ψi = h b, a, β, α| ψi = − h a, b, α, β| ψi

We define operators R12 and S12 which exchange position and spin labels (P12 = R12 S12 ):

ha, b, α, β| R12 |ψi = h b, a, α, β| ψi ha, b, α, β| S12 |ψi = h a, b, β, α| ψi .

H contains no spin operators, so that [R12 , H] = [S12 , H] = 0. Hence the eigenkets can be
eigenkets of both R12 and S12 .

CWPP 15/2/2013
Further Quantum Physics Helium 53

We can safely assume that [J, H] = 0, where J = L + S is the total angular momentum.
But since H contains no spin operators [S, H] = 0, and also [L, H] = 0.
However [ll1 , H] is non-zero because of the electron-electron interaction term:
   
1 µ 1 r1 ∧ r2
[ll1 , H] = r1 ∧ p1 , + p1 · p2 and r1 ∧ p1 , | = −i 6= 0
|r12 | M |r12 |r12 |3

So the eigenkets can be eigenkets of any component of L and S, but not of the single-electron
angular momenta.
Finally H commutes with parity P (as do all atomic Hamiltonians) [H, P ] = 0.

How many of these are simultaneously possible — that is, how many of these operators com-
mute with each other as well as with H?
All except for the fact that the different components of angular momenta do not commute
([Lx , Ly ] = iLz etc.) We therefore choose to make the states eigenstates of L2 and Lz etc.
We then find that [J 2 , Lz ] 6= 0 and [J 2 , Sz ] 6= 0.

Summary: The eigenkets of the approximate Helium H can be chosen to be eigenstates

of P12 (with eigenvalue −1), R12 , S12 , P , L2 , S 2 , and either Lz , Sz or J 2 , Jz .

CWPP 15/2/2013
Further Quantum Physics Helium 54

Symmetric and Antisymmetric Spin States

We can easily find the required eigenstates of S12 by considering just the spin part of our ba-
sis kets |σ1 , σ2 i:
S12 |+, +i = |+, +i S12 |+, −i = |−, +i
S12 |−, −i = |−, −i S12 |−, +i = |+, −i

Thus two of the four states are symmetric, and the other two have no definite symmetry. But
we can easily find normalised symmetric or anti-symmetric combinations:

1 1 1
S12 √ [|+, −i + |−, +i] = √ [|−, +i + |+, −i] = + √ [|+, −i + |−, +i]
2 2 2
1 1 1
S12 √ [|+, −i − |−, +i] = √ [|−, +i − |+, −i] = − √ [|+, −i − |−, +i]
2 2 2

So there are three symmetric spin states (S12 eigenvalue +1) and one antisymmetric
spin state (S12 eigenvalue −1). These must be combined with spatial states of the opposite
symmetry so that the overall P12 symmetry is antisymmetric (P12 eigenvalue −1).

CWPP 15/2/2013
Further Quantum Physics Helium 55

Spin properties of the symmetrised states

Having decided on symmetric and antisymmetric spin states we now expect to find eigen-
states of total spin S2 and Sz . The Sz part is easy:

1
Sz = s1z + s2z and we know s1z |+i = |+i .
2
Thus we can easily deduce
 
1 1
Sz |+, +i = + |+, +i = (+1) |+, +i Sz |+, −i = 0
2 2
 
−1 −1
Sz |−, −i = + |−, −i = (−1) |−, −i Sz |−, +i = 0
2 2

Thus Sz eigenvalues: the three symmetric states have +1, 0, −1, and the antisymmetric state
has 0.

The S2 part needs a bit more work: S2 = (s1 + s2 )2 = s21 + s22 + 2s1 · s2 . We can expand the
final term:

2s1 · s2 = 2s1x s2x + 2s1y s2y + 2s1z s2z = s1+ s2− + s1− s2+ + 2s1z s2z .

CWPP 15/2/2013
Further Quantum Physics Helium 56

We know the effect of the individual electron s2 and s+ or s− operators:

3
s2 |±i = s(s + 1) |±i = |±i s+ |+i = 0 s+ |−i = |+i .
4
Hence  
3 3 1 1
S2 |+, +i = + +0+0+2 |+, +i = 2 |+, +i
4 4 2 2
and similarly for |−, −i, while for the other two we note that s1+ s2− |−, +i = |+, −i so that
h i 3 3 1 −1
h i
2
S |+, −i ± |−, +i = + ±1+2 |+, −i ± |−, +i .
4 4 2 2
Thus all four states are eigenstates of S2 , with eigenvalues of 2 or 0. What did we expect?
We are adding two angular momenta of 12 so we expect the maximum S quantum number to
be 1 and the minimum to be 0. The eigenvalue is S(S + 1) which is thus 2 or 0 . . . perfect!
Thus making the spin states symmetric and antisymmetric under S12 also made coupled the
spins to 0 or 1. We label these states with S,MS for total spin and z-component:
1
|1, 1i = |+, +i |1, 0i = √ [|+, −i + |+, −i]
2
1
|1, −1i = |−, −i |0, 0i = √ [|+, −i − |+, −i] .
2

CWPP 15/2/2013
Further Quantum Physics Helium 57

Summary of exact properties of eigenstates

The approximate Hamiltonian containing just non-relativistic kinetic energies and electro-
2S+1
static interactions has eigenstates denoted by L with the following properties:
3
S, 3 P , 3 D. . . using the usual letter code, but now in capitals: 0 ≡ S, 1 ≡ P , 2 ≡ D etc.
• S12 eigenvalue +1 with R12 eigenvalue −1 ;
• P12 eigenvalue −1 (only)
• S2 eigenvalue S(S + 1) = 2 (S = 1, ‘triplets’) and Sz eigenvalue MS = 0, ±1;
• L2 eigenvalue L(L + 1) with L = 0, 1, 2 . . .
• an eigenstate of parity;
• NOT an eigenstate of ℓ21 + ℓ22 ;
1
S, 1 P , 1 D. . .
• S12 eigenvalue −1 with R12 eigenvalue +1 ;
• P12 eigenvalue −1 (only)
• S2 eigenvalue S(S + 1) = 0 (S = 0, ‘singlets’) and Sz eigenvalue MS = 0;
• L2 eigenvalue L(L + 1) with L = 0, 1, 2 . . .
• an eigenstate of parity;
• NOT an eigenstate of ℓ21 + ℓ22 ;

CWPP 15/2/2013
Further Quantum Physics Helium 58

Perturbation Theory
However these exact properties do not enable us to find the eigenkets or their energies.
The only approach we have to fall back on is Perturbation Theory .
We drop the small nuclear motion terms, and revert to standard atomic units to leave
 
X p2  
 i Z2 1 
H = H0 − + .
i=1,2
2 |r i | |r 12 |

The term that makes it hard is the 1/r12 term — without that it is just two uncoupled hy-
drogen Hamiltonians. But the coefficient of 1 is not exactly small in comparison with Z,
especially in Helium. But this approach might just give some insight in helium-like ions of
higher Z.

So we treat the final term using Pertubation Theory.

The unperturbed eigenkets are just products of hydrogenic states with any spin state:
|n1 , ℓ1 , m1 ; n2 , ℓ2 , m2 ; S, MS i = |n1 , ℓ1 , m1 i |n2 , ℓ2 , m2 i |S, MS i .

CWPP 15/2/2013
Further Quantum Physics Helium 59

However these don’t have all the symmetries we want (as well as some quantum numbers —
n and ℓ — we said we didn’t want). But the ground state (n1 = n2 = 1, ℓ1 = ℓ2 = 0 ≡ s) is
OK:
1s2 1 S = |1si |1si |0, 0i

where the spatial state is symmetric so the spin state is antisymmetric, and there is no angu-
lar momentum either orbital or spin, and even parity. The zeroth-order energy is two hydro-
genic energies:
(0) Z2 Z2
E1s2 =− − = −Z 2 in units of H0 .
2 2

The general state above has zeroth-order energy

Z2 Z2
En(0)
1 n2
=− 2 − 2 in units of H0 .
2n1 2n2

Taking n2 → ∞ gives a prediction for the 1st ionization energy of Z 2 H0 /2, or 54.4 eV for
helium — a rotten prediction! But it makes one useful prediction: the state with both elec-
trons in n = 2 has 1/4 of the binding energy of the ground state, whereas the first ionisation
is predicted to be a half: any doubly excited states are above the first ionisation level.

CWPP 15/2/2013
Further Quantum Physics Helium 60

So we consider only singly excited states 1snℓ:

Ground State: 1s2 1S = |1si |1si |0, 0i

1  i
1
Excited Singlet State: 1snℓ L = √ |1si |nℓi + |nℓi |1si |0, 0i
2
1  i
3
Excited Triplet State: 1snℓ L = √ |1si |nℓi − |nℓi |1si |1, MS i
2

These states have all the right eigenvalue properties: spin and space exchange symmetry, par-
ity, eigenstates of S2 , L2 , Sz and Lz . (No coupling required in orbital angular momentum be-
cause one of the ℓ is zero!) The only problem is that it is also an eigenstate of ℓ21 + ℓ22 .

The zeroth-order energies of these states are

(0) Z2 Z2
E1snℓ =− − 2 in units of H0 .
2 2n

Now we have constructed the correct zeroth-order eigenkets and energies we can do some per-
turbation theory.

CWPP 15/2/2013
Further Quantum Physics Helium 61

First-Order Perturbation Theory

We now evaluate the effect of the electron-electron interaction in perturbation theory:

1 1 1 1
= p2 = √
|r12 | r1 − 2r1 · r2 + r22 r> 1 − 2t cos θ12 + t2

where r> is the larger of |r1 |, |r2 |, and t = r< /r> . But this is just the generating function for
Legendre polynomials so
∞  ℓ 
X
1 r<
= ℓ+1
Pl (cos θ12 )
|r12 | r>
ℓ=0

If we only consider ℓ = 0, or s, states then we need only consider the first term in the series:
the others introduce angular factors that integrate to zero.
For the ground state we have
Z Z
(1) ∗ 1 3 3 Z6 1 −2Zr1 −2Zr2 3
E1s2 = ψ1s2 (r1 , r2 ) ψ1s2 (r1 , r2 ) d r1 d r2 H0 = 2 e e d r1 d3 r2 H0 .
r> π r>

(1)
The integral is 5Z/8 so E1s2 = (5Z/8) H0 .

CWPP 15/2/2013
Further Quantum Physics Helium 62

CWPP 15/2/2013
Further Quantum Physics Helium 63

In the excited states we now reap the benefits of putting some effort into the symmetries of
the zeroth-order kets: what would otherwise be a degenerate perturbation problem of large
order can be treated as non-degenerate because the kets are all eigenkets of operators com-
muting with H0 and H1 :
(
(1) 1s2s3 r1> 1s2s3 S
E1s2s =
1s2s1 r1> 1s2s1 S

In excited states this method yields two classes of integral:

Z Z
(1) 1 1
E1s2s = |φ1s (r1 )|2 |φ2s (r2 )|2 d3 r1 d3 r2 ± [φ1s (r1 )φ2s (r1 )] [φ1s (r2 )φ2s (r2 )] d3 r1 d3 r2
r> r>

known as direct and exchange integrals (upper sign for singlet, lower for triplet).

Enthusiasts may care to check that this gives

 
(1) 17 16
E1s2s = ± ZH0 .
81 729

CWPP 15/2/2013
Further Quantum Physics Helium 64

CWPP 15/2/2013
Further Quantum Physics Helium 65

Variational Theorem
To do better than this we must use a more powerful method. The variational theorem allows
us to construct very accurate ground states (or lowest states of given symmetry):
Theorem:
For any ket |αi the expectation value of the Hamiltonian hα| H |αi / h α| αi is an upper bound
for the ground state energy.
Proof:
P
Expand |αi in energy eigenstates |αi = i ci |Ei i. Then
P 2
P
hα| H |αi i |c i | E i i |ci |2 (Ei − E1 )
= P 2
= E1 + P 2
≥ E1 .
h α| αi |c
i i | |c
i i |
since the last term is positive definite. Equality is obtained iff |αi = |E1 i.
All accurate atomic wavefunctions are based in way or another on the variational theorem,
either constructing the wavefunction as a sum of analytic functions with variable coefficients,
or as a numerical approximation on a grid of points.
For excited states there is an extension, the HUM (Hylleraas-Undheim-MacDonald) Theorem:
for a set of trial wavefunctions the n’th eigenvalue of the Hamiltonian matrix gives an upper
bound for the n’th excited state.

CWPP 15/2/2013
Further Quantum Physics Helium 66

A very simple variational wavefunction is provided by an effective-Z model: we observe that

the electron-electron interaction counteracts the electron nulear interaction by tending to
push electrons apart rather than together. So we introduce a different splitting up of H:
 
X p2  ′
 
X Z − Z′ 
i Z 1
H= − + −  H0 .
i=1,2
2 |ri | |r12 | i=1,2 |ri |

The ground state eigenfunction of the first term is

Z ′3
exp[−Z ′ (r1 + r2 )],
π
and we use this as our variational wavefunction. The expectation value is given by
First term: −Z ′2
5Z ′
e-e interaction:
8
′
′2 (Z − Z ) ′ ′
Last term: −2Z = −2Z (Z − Z )
Z′
giving a total energy of Z ′2 − (2Z − 5/8)Z ′ . This has a minimum:
 2
′ 5 5
Zmin =Z− Emin = − Z − H0 .
16 16

CWPP 15/2/2013
Further Quantum Physics Helium 67

CWPP 15/2/2013
Further Quantum Physics Helium 68

Obviously this is somewhat better for Helium but for the higher Z cases it’s not really any
improvement — because we haven’t tackled the fundamental problem of the correlation in-
duced by the electron-electron interaction.
The real problem stems from the fact that the approximate kets we are using are eigenkets of
ℓi . For the groundstate this implies that the variational wavefunction is of the form ψ(r1 , r2 ):
the amplitudes
h r1 n, r2 n, α, β| ψi and h r1 n, −r2 n, α, β| ψi

are equal. What is needed is an explicit dependence on r12 as well.

From a perturbation perspective, the effect of the 1/r12 perturbation is to mix other configu-
rations, including a small amount of 2p2 1 S into 1s2 1 S .

The quantum numbers ℓi are thus approximate: giving a rough indication of the property of
the state but not exactly true: an important idea in atomic physics.

CWPP 15/2/2013