Factorisation of Hamiltonians:

theory and some applications

HEP Theory

Patrick Meessen
March 15, 2011
 In these notes we shall try to explain the method of factorisation of Hamiltonians as it is a
potent technique for finding spectra of 1-dimensional quantum mechanical systems. For some
odd reason this technique is not dealt with in the major text books on quantum mechanics,
whence the envisaged need for these notes.1
The outline of these notes, is to start with Dirac’s treatment of the quantum harmonic
oscillator as it was the first system to be solved by factorisation. Then in sec. (2) we shall
outline the idea of factorisation and introduce the notion of a dual Hamiltonian and its
implications. In secs. (2.1,2.3) and (2.4) we shall apply the factorisation method together
with a recurrence relation to obtain the spectrum of some simple 1-dimensional systems. The
last section, sec. (3), is devoted to the hydrogen atom and how the method of factorisation is
used to obtain the spectrum.
Finally: please have in mind that these notes are work in progress and that comments are
welcome.

1 Harmonic Oscillator
As is well-known the harmonic oscillator is a simple physical systen whose classical Hamilto-
nian is given by2
p2 mω 2 2
H(p, q) = + q . (1)
2m 2
Before plunging into the quantum harmonic oscillator, we will quickly discuss the classical
harmonic oscillator.
In order to find the classical trajectories it is easier to go from the phase-space description
in terms of the p and q variables to the configuration space variables q̇ and q. This is easily
done by the Legendre transform L(q̇, q) = q̇p − H(p, q), which leads to the perhaps better
known Lagrangean
m 2
q̇ − ω 2 q 2 .

L = (2)
2
The Euler-Lagrange equations that follow from the above Lagrangean are

q̈ = −ω 2 q −→ q(t) = q0 cos(ωt) + (v0 /ω) sin(ωt) , (3)

where q0 , resp. v0 , is the position, resp. velocity, at the time t = 0. If we impose the,
physically reasonable,3 boundary condition v0 = 0, then q0 is the amplitude of the oscillation.
Independently of the chosen boundary conditions, however, it is paramount that the period
of the oscillation is 2π/ω, and therefore follows Galileo’s isochronous law: the period of
oscillation of a pendulum with small amplitude is independent of said amplitude. Finally, the
energy of the trajectory is
mω 2 q02
E = ≥ 0, (4)
2
and the zero energy trajectory is clearly the one that is not moving at all, i.e. q(t) = 0.
1
One exception to this rule is R. Robinett’s “Quantum Mechanics” [1], Faculty of Science library code
K-53-645, who has a small section about factorisation, and you are advised to read it. Needless to say, should
you find books dealing with the factorisation of Hamiltonians, I would be very pleased to hear about it.
2
The factor of m in the potential term is not usual in classical treatments but will greatly simplify the
formulas in the quantum case.
3
Imagine the situation of a pendulum being held fixed at a position q0 until at t = 0 we release it.

1
 Let us then go on and consider the quantum version of the harmonic oscillator described
by the Hamiltonian operator
P2 mω 2 2
H = + Q . (5)
2m 2
The first observation is that in the quantum system the energy, i.e. the possible eigenvalues
of the Hamiltonian operator, cannot be zero: this is easily seen by considering an arbitrary
normalised state |Ψi to calculate

1 mω 2
E[Ψ] ≡ hΨ|H|Ψi = | P|Ψi |2 + | Q|Ψi |2 , (6)
2m 2
which clearly is bigger than or equal to zero. But, in order to attain E[Ψ] = 0, the state
|Ψi must satisfy P|Ψi = 0 and Q|Ψi = 0, which is incompatible with the fundamental
commutation relation [Q, P] = i~. We therefore reach the conclusion that in the quantum
harmonic oscillator the energy is bounded from below and is strictly bigger that zero. The
question then should be: What is the lowest energy of this system?
There is an algebraic approach to this system, invented by Dirac, which plays a prominent
rôle in 20th century physics and it is this rôle that prompts us to re-discuss it.
The gist of Dirac’s approach resides in the introduction of the following 2 operators
r
1 mω
A ≡ √ P − i Q, (7)
2~mω 2~
r
† 1 mω
A = √ P + i Q. (8)
2~mω 2~

Using these operators we can rewrite the Hamiltonian as

~ω  † 
H = A A + A A† , (9)
2
which in itself is not too exciting. The excitement starts when we calculate
h i
A , A† = Id . (10)

We can use the above commutation relation to rewrite the Hamiltonian as

h i
H = ~ω A† A + 12 . (11)

If we use this last expression to calculate E[Ψ], we see that

E[Ψ] = ~ω | A|Ψi|2 + ~ω/2 ≥ ~ω/2 !!! (12)

This result indicates that the lowest energy that can be achieved in this system holds for a
normalised state, denominated |0i, that satisfies

~ω
A|0i = 0 , H|0i = |0i , h0|0i = 1 . (13)
2
As any other state necessary has a higher energy we will call the state |0i the groundstate or
the vacuum state.

2
 The reader may feel ill-at-ease with the abstract definition of the groundstate, especially
w.r.t. its existence, a situation we propose to ameliorate by discussing the wave-function
corresponding to the groundstate: define the corresponding wave-function as Ψ0 (q) ≡ hq|0i
and use the differential representation of the operators, i.e. Q = q and P = −i~∂q , then you
can see that the normalised wavefunction is
 mω 1/4  mω 
Ψ0 (q) = exp − q2 , (14)
π~ 2~
where, as usual, we chose a possible phase to be unity.
Given that we found the state |0i, can we use it to find other states? Well, consider the
state
A† |0i , (15)
and ask yourself whether it is an eigenstate of the Hamiltonian H. This question is readily
answered by calculating
 
† † 1
HA |0i = ~ω A A + A† |0i
2
 
† † 1
= ~ω A AA + |0i
2
 h i 1
† † †
= ~ω A A A + A, A + |0i
2
 
1
= ~ω A† 1 + |0i
2
 
1
= ~ω 1 + A† |0i . (16)
2

So the state in eq. (15) is indeed an energy eigenstate with an energy ~ω above the groundstate.
At this point then we should try to find out whether it can be normalised and whether it
is orthogonal to |0i. The orthogonality issue is readily clarified once we see that energy
eigenstates of different energies are always orthogonal: consider the states |Ei and |E 0 i and
calculate

hE|H|E 0 i = E 0 hE|E 0 i (17)

0
= hE 0 |H|Ei = E hE|E i , (18)

from which we see that

E0 − E hE|E 0 i = 0 whence if E 0 6= E: hE|E 0 i = 0 .

(19)

A small calculation of |A† |0i|2 shows4 that the state can also be normalised and we must
conclude that we found another normalised energy eigenstate, denoted by |1i, whose relation
to the groundstate is given by
|1i ≡ A† |0i . (20)
Observe that in the normalisation of the above state, we took a possible phase to be unity.
4
To wit: |A† |0i|2 = h0|AA† |0i = h0| A, A† |0i = h0|0i = 1.
 

3
 Of course we can use the same trick to create state with higher energies. In order to see
this, it is handy to introduce an operator N = A† A whose commutation relation with A and
A† are readily calculated to be

[N, A] = −A , (21)
h i
N, A† = A† . (22)

It is then an easy task to see that

 n  n
N A† = A† (N + n) , (23)

n
which when coupled to H = ~ω N + 12 means that the states A† |0i are energy eigenstates
 

with energy ~ω(n + 1/2). Normalising these states, and labeling them with the index n, we
end up with a countable infinite set of energy eigenstates

1   n  H|ni = ~ω (n + 1/2) |ni ,
†
|ni = √ A |0i −→ (24)
n! 
hn|mi = δn,m .

Given this set of states we can calculate the action of the operators on them, only to find

N|ni = n|ni , (25)

√
A|ni = n |n − 1i , (26)
†
√
A |ni = n + 1 |n + 1i . (27)

We can imagine the energy spectrum as a ladder with the energy levels being the spokes;5
the operator N then gives the number of spokes that one had to climb up from the ground to
get to the desired spoke, and correspondingly N receives the name number operator. Using
the same analogy, the operator A† pushes you up one spoke and receives the name ladder-up
operator. Unsurprisingly, the operator A is baptised with the name ladder-down operator.6
In eq. (14) we defined the wave-function of the groundstate as Ψ0 (q) = hq|0i, and in this
case we want to find the wave-functions for all energy eigenstates. In order to do so, we follow
the same steps as needed for the derivation of eq. (14) and start by defining Ψn (q) ≡ hq|ni
and introducing the abbreviation µ ≡ mω/~. Then it is an easy calculation to show that

in  µ 1/4 √
exp −µ/2 q 2 Hn ( µq) ,


Ψn (q) = √ (28)
2n n! π

where Hn is the nth Hermite polynomial, one representation of which is

2 2
Hn (x) = (−1)n ex ∂xn e−x . (29)

The painful properties of the Hermite polynomials such as the orthogonality and completeness
relations can be derived quite painlessly from the algebraic approach to the quantum harmonic
oscillator.
5
Ladder = escalera; spoke = peldaño.
6
Similar operators appear in Quantum Field Theory and there they go by the names particle number
operator or occupation number (N), (particle) creation operator (A† ) and (particle) annihilation operator (A).

4
 Summarising, we have found a countable infinite number of energy eigenstates, all of
which can be created using the operator A repeatedly on the groundstate |0i. The remaining
question to be answered is whether there are energy eigenstates that were left out in the
above analysis. The answer is no, and in order to see this suppose there exists an energy
eigenstate |Ei, that doesn’t belong to the set we obtained above, i.e. @n∈N E = ~ω(n + 12 ).
The argument of A|Ei being an energy eigenstate with energy E − ~ω, however, goes through
and we can apply A repeatedly on the state |Ei as to end up with a negative energy, leading to
a contradiction with the fact that the energy should be positive. We must therefore conclude
that the only energy eigenstates of the harmonic oscillator are the ones that were obtained
above.

2 Factorisation of Hamiltonians
Dirac’s treatment of the harmonic oscillator by factorising the Hamiltonian, leads to the
obvious question, one supposedly posed first by Schödinger in ref. [2], whether all quantum
mechanical system can be factorised: for one dimensional systems the answer to this question
is, modulo important titbits, affirmative.
Suppose we are given a Hamiltonian operator H that is bounded from below, which is a
sound restriction as otherwise the system would be unstable, and that the groundstate |Ψ0 i
has energy E0 . Given this situation we will redefine the Hamiltonian as H− ≡ H − E0 Id,
which implies that the energy eigenvalues of H− is positive semi-definite, i.e. hH− i ≥ 0. The
question posed by Schrödinger can then be formulated as:
Is there an operator A such that H− = A† A ? (30)
As the groundstate of the Hamiltonian H− has zero energy we have that 0 = hΨ0 |H− |Ψ0 i =|
A|Ψ0 i |2 , implying that the groundstate satisfies the equation
A|Ψ0 i = 0 . (31)
In the modern technical language equations such as eq. (31) are called Bogomol’nyi-Prasad-
Sommerfield equations, which is usually abbreviated to BPS equations.
In order to advance a bit in the direction of demonstrating the feasibility of factorisation,
consider the following position-representation of the operators A and A† :
A ≡ i (∂q + W(q)) , A† = i (∂q − W(q)) , (32)
where W is a real function of q, called the superpotential. Using the above representation and
defining Ψ0 ≡ hq|Ψ0 i, we can then use the BPS equation, eq. (31), to determine
W = −∂q [ log (Ψ0 ) ] , (33)
which shows that given a system with a well-defined groundstate, we can always find W.
Conversely, given W we can always find Ψ0 as
Ψ0 (q) ∼ exp (−χ(q) ) where ∂q χ = W . (34)
Obviously, for the groundstate to be part of the spectrum of H− , it must be normalisable
which we suppose to be the case:7 so we see that the factorisation of Hamiltonians is always
possible.
7
Observe that if the spectrum of Q, S(Q), is equivalent to R then this implies drop-off conditions for Ψ0 .

5
 A related question is whether given a Hamiltonian, the factorisation can be used to find
the groundstate: in order to see whether this is possible let us calculate

H− Ψ = A† AΨ = −∂ 2 Ψ + W2 − ∂W Ψ ,
 
(35)

which we can compare to the Schrödinger equation (which we have rescaled with a factor
2m~−2 )
− ∂ 2 Ψ + V− (q)Ψ = EΨ . (36)
Now, before going over to identify eq. (35) with eq. (36), let us point out that the quantum
mechanical ground energy can be redefined by summing a finite piece without affecting the
physics: we used this relation in the beginning of this section when we related H to H− .
This then means that in order to make the match we must consider for our factorisation
programme the following constraint

W2 − ∂W ≡ V− (q) = V (q) + V0 , (37)

where V0 is some real constant that we can add to the potential such that the factorisation
works: seeing that the factorised Hamiltonian’s ground-state energy is zero, you can easily
work out that V0 = −E0 . The equation (37) is the Ricatti equation and its relation to the 2nd
order Schrödinger equation is well-known.
Even though the above is a nice titbit, which may be helpful to find the groundstate and
its energy, it may be that actually finding the function W for a given potential is actually a
tremendous, daunting or even impossible task. And even if we are able to find the function
W, what can we say about the rest of the energy spectrum?
Well, given eq. (30) we might be curious about the spectrum of the so-called dual Hamil-
tonian
H+ ≡ AA† = −∂ 2 + V+ (q) −→ V+ ≡ W2 + ∂W . (38)
Clearly the possible energy eigenvalues of H+ , denoted by Ẽ, must be positive semidefinite,
and the zero energy solution must satisfy A† |Ψ̃0 i = 0; the solution to the resulting differential
equation exists and reads

A† Ψ̃0 (q) = 0 −→ Ψ̃0 (q) ∼ eχ(q) , (39)

where χ was defined in eq. (34). Now, if we take S(Q) ∼ R then, as we imposed the groundstate
of H− to be normalisable, the state |Ψ̃0 i is not normalisable, and hence not part of H+ ’s
spectrum.
Consider then an arbitrary H− -eigenstate |Ei with energy E > 0, i.e. any energy eigen-
state except the groundstate. A small calculation then shows that

H+ (A|Ei) = A (H− |Ei) = E (A|Ei) , (40)

whence the state A|Ei is an eigenstate of the Hamiltonian H+ with the same energy. Of
course, for the state A|Ei to be physical it must be normalisable which is luckily no problem:
it is easy to see that the corresponding normalised states are given by

|Ẽi = E −1/2 A|Ei (E > 0) , (41)

which means that S(H− ) /{0} ⊆ S(H+ ).

6
 S(H− ) : |E0 i |E1 i |En i

A A† A A†

S(H+ ) : |Ẽ1 i |Ẽn i

Figure 1: A graphical depiction of the mapping of states derived in this section.

Turning the argument in eq. (40) around we have

   
H− A† |Ẽi = A† H+ |Ẽi = Ẽ A† |Ẽi (Ẽ > 0) , (42)

so that S(H+ ) ⊆ S(H− ). Putting then both things together we must conclude that

S(H+ ) ∼
= S(H− ) /{0} , (43)

or said in words: the spectrum of H+ consists of all the non-groundstates of H− .

Summarising the results obtained thus far, we see that given a system with a groundstate,
we can always factorise the Hamiltonian, H− , and define the “dual Hamiltonian” H+ which
is also factorised. The spectrum of the dual Hamiltonian is equal to that of the original
Hamiltonian but for the groundstate. As one can imagine, this mapping is useful if the
spectrum of the dual Hamiltonian is known as then the problem of finding the spectrum of
the original Hamiltonian is solved.
Of course, we might have a small problem if the dual Hamiltonian’s spectrum is not
known. But, as the dual Hamiltonian does admit a groundstate, namely the state dual to
the first excited state in the original system, the dual Hamiltonian itself must allow for a
factorisation and we can imagine using the process of factorisation recursively till we end up
with something known. In the following sections we shall apply the technique of factorisation,
together with the recurrence, to deduce the spectrum of some easy Hamiltonians.

2.1 The simplest of the pedagogical Pöschl-Teller potentials

Consider the superpotential8
W = tanh (q) , (44)
which using eq. (37) leads to the potential
2
V− (q) = 2 tanh2 (q) − 1 = 1 − , (45)
cosh2 (q)
8
One can consider of course the general function W = A tanh(µq) and follow the factorisation reasoning
to find the spectrum &c. for the resulting family of potentials, which are called Pöschl-Teller potentials. The
version we are using here gives easy formulas, is easy to understand and is therefore especially suited for
pedagogical purposes.

7
which is a simple potential whose absolute minimum lies at q = 0 with value V− (0) = −1
and asymptotes to V− (|q| → ∞) = 1. So the potential has all the necessary ingredients for
allowing a discrete spectrum and the question is how many discrete states there are?
From the general discussion, we see that it is worthwhile to investigate the dual Hamilto-
nian: a small calculation show that

V+ (q) = 1 (!!!) (46)

So we see that the dual-Hamiltonian is that of a particle in a constant potential, which is

equivalent to a free particle whose spectrum is continuous. This then means that the above
simple potential has only one discrete energy level whose energy w.r.t. the potential in eq. (45)
is zero; Had we chosen the more physically reasonable definition of the potential, namely the
one for which V (|q| → ∞) = 0, which can easily be arranged by adding a constant, then the
minimum value of the potential would be −2 and the energy of the bound state would be −1.
The groundstate is then easily found and is given by the normalised wavefunction
1
Ψ0 = √ . (47)
2 cosh(q)

What happens to the naive groundstate of the dual system given by A† Ψ̃ = 0? Clearly we
can solve the differential equation to give Ψ̃ = ψ̃ cosh(q), where ψ̃ is a complex constant, but
can readily convince oneself that it is non-normalisable.
Having dealt with the discrete part of the spectrum, it is then time to turn to its continuous
part: from eq. (42) we can see that the correctly normalised state in the original theory is
given in terms of the dual theory as

|Ei = √1 A† |Ẽi which for the wavefunctions leads to: ΨE (q) = √1 A† Ψ̃Ẽ (q) , (48)
Ẽ Ẽ

where in the last equation A† is the operator in the coordinate representation, see eq. (32), and
we defined ΨE = hq|Ei &c. As we know, for a free particle the eigenstates are, due to parity
invariance of the Hamiltonian, 2-fold degenerate and correspond to momentum eigenstates;
in the case at hand this means that the energy eigenstates of the dual theory are given by
Ψ̃Ẽ(k) = (2π)−1/2 eikq and the corresponding energy is Ẽ(k) = k 2 + 1. A straightforward
application of eq. (48) then leads to the following expression for the wavefunctions of the
energy eigenstates in the continuous spectrum

Ψk (q) = ( 2π(k 2 + 1) )−1/2 [ tanh(q) − ik ] eikq , (49)

whose energy is k 2 + 1. As usual for states in the continuous part of the spectrum, the Ψk (q)
form a δ-orthonormal set, i.e. hΨk0 |Ψk i = δ(k − k 0 ).

2.2 The pedagogical Pöschl-Teller potentials, self-similarity and recursion

The foregoing example was chosen such that the dual theory’s spectrum was known. Normally
one in not that lucky and one ends up with a dual theory whose spectrum is also unknown.
In that case one can use the fact that the dual Hamiltonian is bounded from below, factorise
the dual Hamiltonian and obtain the dual of the dual Hamiltonian. There is of course no
guarantee that the thus obtained dual-dual Hamiltonian’s spectrum is known, which is how-
ever no problem as we know how to deal with such situations: just factorise the dual-dual

8
Hamiltonian! This just-factorise-the-dual-Hamiltonian approach to solving problems has the
obvious drawbacks of being time/work-consuming and of having no guarantee whatsoever
that at some point one will end up with a known system.
The time/work-consuming part of the drawbacks can sometimes be overcome, as is the case
in some of the standard examples in quantum mechanics. The key to overcome these problems
lies in the similarity of the dual Hamiltonian to the original one. It is this self-similarity, as
it is called in the literature, that allows one to write down hierarchies of Hamiltonians and
to derive recursion relations for the energies of the states and the wave-functions. We shall
explain this using a generalisation of the Hamitonian studied in the foregoing section.
Consider the following generalisation of the prepotential in eq. (44)

W(n) = n tanh(q) (n ∈ N+ ) , (50)

for some given n ≥ 2. With this choice we can calculate the potential V− and also the
corresponding dual potential V+ to be

(n) n(n + 1)
V− (q) = n2 − = n(n + 1) tanh2 (q) − n , (51)
cosh2 (q)
(n) n(n − 1)
V+ (q) = n2 − = n(n − 1) tanh2 (q) + n . (52)
cosh2 (q)

By direct comparison we can derive the relation9

(n) (n−1) (n) (n)
V+ = V− + 2n − 1 −→ H± ≡ P2 + V± , (53)

What does this recursion relation mean? Well, consider for the moment the case of n = 2: in
(2)
that case the potential V+ is up to a trivial shift in groundstate energy exactly the one in
eq. (45) , which is the case with n = 1, whence the hierarchy of factorisation ends quickly as
we already know the spectrum of the dual theory. Since we loose one state in the step to the
dual Hamiltonian and the dual theory has 1 state in the discrete spectrum, we must conclude
(2)
that H− ’s discrete spectrum consists of 2 states.
(n)
For general n, the discrete spectrum of H− consists of n states: indeed, starting from the
Hamiltonian, we can use eq. (53) to derive the sequence

(n) step no. 1 (n) (n−1) step no. n−1 (1)

H− −−−−−−−−→ H+ = H− + 2n − 1 −→ . . . −−−−−−−−−−→ H− , (54)
(1)
where it effectively stops as we know the spectrum of the Hamiltonian H− . From the general
discussion we know that every −→ means the eradication of one state, namely H− ’s ground-
state, and since there are (n − 1) −→’s needed in order to get from n to 1, we see that the
(n)
spectrum of H− consists of n discrete states. Of course, there is a continuous spectrum on
top of that.
Having given the number of states in the discrete part of the spectrum, we would also like
(n)
to know the associated energy eigenvalues. These can be derived as follows: let Ep denote
(n)
the energy eigenvalue of the pth state in the discrete spectrum of the Hamiltonian H− . As
(n) (n)
we start counting from 0, the groundstate energy is E0 which by construction is E0 = 0.
9
Remember that we are comparing factorised Hamiltonians to ordinary, albeit rescaled, Hamiltonians of
the type displayed in eq. (36). This means that in this formula we are using P = −i∂q !

9
 (n)
Suppose then that we want to know the energy E1 , i.e. the energy of the first excited state
(n)
of H− . As we know from the general theory in sec. (2), this energy is exactly the same as the
(n) (n) (n)
one of the corresponding state for the dual Hamiltonian H+ , whence we have E1 = Ẽ1 .
(n) (n−1)
Eq. (53) then means that the state |Ẽ1 i is an eigenstate of the Hamiltonian H− , in fact
(n−1) (n−1)
its groundstate. W.r.t. to the Hamiltonian H− we would call this state |E0 i and a
(n) (n−1)
straightforward application of eq. (53) leads to the identity Ẽ1 = E0 + 2n + 1 = 2n + 1,
(n)
where we used the fact that E0 = 0 for all n. Putting all ingredients together we see that
(n)
E1 = 2n + 1.
(n)
We can repeat the above steps for an arbitrary energy in the discrete spectrum of H− .
(n) (n) (n)
Consider the pth state with energy Ep . By duality we have that Ep = Ẽp , and by the
(n) (n−1)
refactorisation of the dual Hamiltonian, eq. (53), we have that Ẽp = Ep−1 + 2n + 1; the
(n−1)
fact that the energy of the (p − 1)th state appears in Ep−1 is due to the fact that the state
(n) (n−1)
|Ẽ1 i = |E0 i, i.e. we have to take into account that we start counting states from 0 in
(n−1)
H− ’s spectrum. Putting then everything together, we find the recursion relation
(n−1)
(p = 1, . . . , n − 1) : Ep(n) = Ep−1 + 2n − 1 , (55)

The above recursion relation can be immediately integrated by using the boundary condition
(n)
E0 = 0 to give
Ep(n) = p(2n − p) . (56)
(n)
Observe that the possible energy eigenvalues lie beautifully between the extrema of V− ,
namely −n and n2 .
Given that we know the discrete part of the spectrum,10 we can then ask ourselves about
(n) (n)
the wavefunctions, which we shall define through Ψp (q) ≡ hq|Ep i. The groundstate wave-
functions are determined as usual by means of the BPS equation (31), but as we are dealing
with a family of Hamiltonians, we must take care to use the appropriate As. Introduce the
annihilation operators
  appropriate BPS condition (n)
A(n) = i ∂q + W(n) (q) −−−−−−−−−−−−−−−−−−−−−−→ A(n) Ψ0 (q) = 0 . (57)

The normalised solution to the BPS equation reads

s
(n) (2n − 1)!! 1
Ψ0 = , (58)
2 (n − 1)! coshn (q)
n

where the normalisation was obtained by using

Z ∞
n (n−1)!
dq cosh−2n (q) = 2(2n−1)!! (n ≥ 1) , (59)
−∞

and choosing a possible phase to be unity.

10
Of course, as in the case for n = 1, the continuous spectrum is two-fold degenerate and is given by states
|ki (|k| > 0) with energy E (n) (k) = k2 + n2 . The reader is invited to figure this out.

10
 Using these groundstate wavefunctions we can then derive a recursion relation for the
wavefunctions: remember that there is mapping taking normalised states of the dual system
to normalised non-groundstates in the original theory, see eq. (48). If we combine this with
the fact that the refactorisation of the dual system does not change the state, but only the
Hamiltonian that acts on them, we must infer that for p 6= 0
r r
(n) 1 † (n) 1 † (n−1)
|Ep i = (n) A(n) |Ẽp i = (n) A(n) |Ep−1 i . (60)
Ep Ep

Hitting the above relation with hq| and using eq. (56), we find the following recursion relation
for the wave functions

A†(n) Ψp−1 (q)

−i (n−1)
Ψ(n)
p (q) = √ (p 6= 0) , (61)
p(2n−p)

where we have introduced an extra phase-factor, −i, as to ensure that the resulting wave-
functions are real. This recursion relation, together with the groundstate wavefunctions in
eq. (58) is enough to derive all wavefunctions.
As an example consider the case of n = 2, then the two states in the discrete spectrum
(2) (2)
are Ψ0 , given by eq. (58), and Ψ1 , which can be generated using eqs. (61) and (47), lead to
r
(2) 3 1
Ψ0 = , (62)
4 cosh2 (q)
r
(2) 3 sinh(q)
Ψ1 = − , (63)
2 cosh2 (q)

which are easily seen to be correctly normalised.

2.3 The trigonometric Pöschl-Teller potential

As a variation on the foregoing subsection, consider the family of prepotentials

W(n) = n tan(q) (n ∈ N+ ) . (64)

This prepotential leads automatically to the potential

(n)
V− = −n + n(n − 1) tan2 (q) , (65)

which is naturally defined on the interval q ∈ (−π/2, π/2) which we shall take to be the
domain of our problem: clearly for n = 1 we find a problem with a constant potential on an
interval, which is equivalent to the problem of a particle in an infinite square well. From now
on we shall consider the case n ≥ 2.
The potential for n = 2 has a minimum at q = 0 which lies at V− (0) = −2 and blows up at
q = ±π/2, which is what we meant when we said that the potential was naturally defined on
the interval (−π/2, π/2). As this potential, for any n ≥ 2, has no point where it becomes flat
or constant, there is an infinite number of discrete states, just as in the case of the particle in
a box or the harmonic oscillator. We shall use the method of factorisation to find the energy
eigenstates of the associated Hamiltonian. To this end calculate
(n)
V+ = n + n(n + 1) tan2 (q) . (66)

11
As before we can use this form of the potential to derive a recursion relation
(n) (n+1)
V+ = V− + 2n + 1 , (67)

which in itself leads to the recursion relation for the energies

(n) (n) (n+1)
Ek = Ẽk = Ek−1 + 2n + 1 . (68)

(n)
This recursion relations can be integrated immediately by using E0 = 0, to give
(n)
Ek = k (2n + k) , (69)

which as we suspected has no upper-limit and is, as it should be, bounded from below.
(n)
The wavefunction of the groundstate of the Hamiltonian H− follows easily from eq. (34)
and results in11
s
(n) (2n)!!
Ψ0 = cosn (q) ; (70)
(2n − 1)!! π
the normalisation of the wavefunctions being fixed by using the integral
π/2
(2n − 1)!!
Z
dq cos2n (q) = π, (71)
−π/2 (2n)!!

and taking a possible phase to be unity.

From the wavefunction of the groundstates, we can deduce the wavefunctions using rela-
tions like (48); for the model with n = 2 the normalised wavefunctions of the 4 lowest energy
eigenstates are
q
(2) 16
sin(q) cos2 (q) 8 cos2 (q) − 5


Ψ3 = 3π

q
(2) 32
cos2 (q) 6 cos2 (q) − 5


Ψ2 = 15π

q
(2) 16
Ψ1 = π sin(q) cos2 (q)

q
(2) 8
Ψ0 = 3π cos2 (q)
(72)

In the above figure we overlaid the potential, the outer line, with the energy levels, the straight
horizontal lines; the wiggly lines around the energy levels are the (real) wavefunctions.
11 (n)
Observe that the wave functions naturally satisfy the boundary conditions Ψ0 (q = ±π/2) = 0.

12
 V(r)

De

re

Figure 2: A graphical representation of the Morse potential, indicating the meaning of the
parameters a, re and De .

2.4 Di-atomic molecules and vibrational modes: the Morse Potential

When thinking of a di-atomic molecule, we can envisage the molecule as two bodies connected
by some force, and clearly the first image that comes to mind is a spring connecting them.
When we then start thinking about the possible states of the molecule we first should think
about vibrational modes, perhaps influenced by the image of the molecule as being two bodies
connected by a spring. Clearly, we can also think about the rotational modes that are known
to occur, but we shall restrict our attention to vibrational modes: seeing the spring analogy,
we might be inclined to use the harmonic oscillator potential to approximate the true force
between the atoms. Even though this might be a correct approximation for small vibrations,
it means that there is no way to break the molecule apart: as we saw, the spectrum of the
harmonic oscillator is purely discrete and unbounded from above. In true molecules, however,
there is an energy which breaks up the molecule, in the same way atoms have an ionization
energy, called the dissociation energy. Also, measurements of true di-atomic energy spectrum
shows that the energy of the states does not grow linearly with a possible mode number,
indicating that the harmonic oscillator is a very bad approximation to the nature of the force
between the atoms in di-atomic molecules.
In order to remedy the above, Morse [3] proposed a phenomenological potential of the
form
V = De (1 − exp (−a(r − re )))2 − De , (73)
a graphical representation of which is given in fig. (2). As said before we are only interested
in vibrational modes, whence we are only interested in wavefunctions that depend only on
the distance between the atoms. This means that the problem we have to solve becomes
effectively 1-dimensional and in fact can be written as

~2 2
− ∂ Ψ + V (r)Ψ = EΨ . (74)
2m r

13
If we introduce the following redefinitions
√
2mDe 2m
ρ = ar , λ = , E = 2 2 E, (75)
~a ~ a
we can rewrite the problem in eq. (74)
 
− ∂ρ2 Ψ + λ2 e−2(ρ−ρ0 ) − 2e−(ρ−ρ0 ) Ψ = EΨ . (76)

Let us then try to use the factorisation method in order to find the eigenenergies: after some
educated guesses one can see that the solution to eq. (37) for the case at hand, is given by

W = λ− 1
2 − λe−(ρ−ρ0 ) , V0 = (λ − 21 )2 . (77)

This means that the groundstate energy of the system is given by

~2 a2 1 2


E0 = − λ− 2 . (78)
2m
True to the letter of the factorisation approach, we then consider the dual Hamiltonian and
calculate

V+ = W2 + ∂W = (λ − 21 )2 − 2λ(λ − 1) e−(ρ−ρ0 ) + λ2 e−2(ρ−ρ0 ) , (79)

which is quite similar to the original Morse potential and we can use the same recursion trick
as in sec. (2.2): consider the following family of superpotentials

W(n) = λ − n − 1
2 − λe−(ρ−ρ0 ) , (80)

where n ∈ N. n = 0 corresponds to the Morse’s potential, albeit with a shifted groundstate

energy.
Given the familiy of superpotentials in eq. (80), one can straightforwardly calculate the
associated potentials to be
(n)
V− = (λ − n − 12 )2 − 2λ(λ − n)e−(ρ−ρ0 ) + λ2 e−2(ρ−ρ0 ) , (81)

(n)
V+ = (λ − n − 21 )2 − 2λ(λ − n − 1)e−(ρ−ρ0 ) + λ2 e−2(ρ−ρ0 ) . (82)

We can then relate the two potentials by the relation

(n) (n+1)
V+ = V− + (λ − n − 12 )2 − (λ − n − 23 )2 , (83)

which automatically leads to the discrete spectrum and the associated energy values for
Morse’s Hamiltonian, namely

~2 a2 1 2


En = − λ−n− 2 . (84)
2m
Are we then finished? Well, taking eq. (84) at face-value, the energy spectrum would be
unbounded from below, which is clearly impossible. This means that we should restrict n to

0 ≤ n ≤ bλ − 12 c , (85)

14
as then there is a finite number of states in the discrete part of the spectrum, all of which
have a negative energy.
Why did the recursion argument fail? The argument failed because we did the recurrency
relation without thinking about the spectrum of the dual theories! In fact, one can see that
when n > bλ − 12 c the associated potentials have no global minimum except for the region
r −→ ∞, whence for those values of n the spectrum is purely continuous and our hierarchy
ends. Taking this into account, the method works out just fine.

3 The Hydrogen atom and Hydrogen-like systems

As one knows from classical mechanics the 2-body system with an interaction that depends
only on the relative distance between the 2 bodies can be split into two unrelated physical
systems, namely one describing the free motion of the centre of mass, and the other one
describing the relative motion. Given then the masses m1 and m2 of the two bodies, we know
that the centre of mass system is governed by the systems total mass M = m1 + m2 , whereas
the relative motion is governed by the so-called reduced mass µ given by
m1 m2
µ ≡ . (86)
m1 + m2
Clearly, if one of the masses is far bigger than the other one then the reduced mass is approx-
imately equal to the smaller one: take for instance m1  m2 then one sees from the above
formula that µ = m2 .12
As the dynamics of the centre of mass system is that of a free particle of mass M , its
physics is under control and we will from now on only deal with the system of the relative
motion. The relevant Hamiltonian is given in terms of the operators P ~ and X
~ and reads

~2
P  
H = ~
+ V |X| , (87)
2µ

which is clearly invariant under rotations in R3 : we can use the rotational invariance to reduce
this system to an effectively 1-dimensional system, which can then be dealt with using the
factorisation approach.
In order to reduce the above problem to a 1-dimensional problem, remember that the an-
gular momentum operators Li in the coordinate-representation are given by Li = −iεijk xj ∂k .
Using this expression one can see that
 
~L2 Ψ = −~x2 ∂~ 2 Ψ + ~x · ∂~ ~x · ∂Ψ
~ ~ .
+ ~x · ∂Ψ (88)

If we then introduce the standard spherical coordinate system with ~x2 = r2 , we see that
~x · ∂ = r∂r and can then reshuffle and rewrite the above expression as
~
L2
− ∂~ 2 Ψ = − 1r ∂r2 (rΨ) + r2
Ψ. (89)
12
If you want to see numbers, you can try to calculate the reduced mass for the sun-earth and the earth-
moon system using m = 1, 99 1030 kg, m♁ = 5, 97 1024 kg and m$ = 7, 34 1022 kg. If you’d like to see
things a bit more atom-like you can use mp = 938, 3 MeV/c2 , mn = 939, 6 MeV/c2 and me = 511 keV/c2 . Of
course, you might also calculate the reduced mass for the positronium which is a bound state of an electron
and its anti-particle, the positron, or for muonium, which is a bound state of an anti-muon and an electron
and was experimentally found in 1960 [4]. For this last calculation you’ll need mµ = 105, 7 MeV/c2 .

15
If we use this result in the eigenvalue problem associated to the Hamiltonian (86), i.e. HΨ =
EΨ, and multiply it by 2m~−2 in order to have nicer formulae, we find that
2µ 2µ
~2
E Ψ = −∂~ 2 Ψ + ~2
V (r) Ψ
h i
~
L2
= − 1r ∂r2 (rΨ) + 2µ ~2
V (r) + r2
Ψ. (90)

The key point of the reduction then lies in the fact that the dependency on the spherical
coordinates θ and φ is hidden in the Casimir operator ~L2 , and we will just split

Ψ = r−1 Φl (r) Ym
l
(θ, φ) and use ~L2 Ym
l l
= l(l + 1) Ym . (91)

Doing so we end up with the following Schrödinger equation for the function Φl
l(l + 1) 2µ 2µ
− ∂r2 Φl + Φl + ~2
V (r)Φl = ~2
EΦl , (92)
r2
where one should observe that the resulting system does not depend on the magnetic moment
quantum number m, indicating that for a given l, the energy state will be (2l + 1)-fold
degenerate; this degeneracy is due to the fact that the system is spherically symmetric. For
the moment we will obviate this degeneracy.
Up to this point our discussion was really only limited by the fact that the potential only
depended on the distance between the 2 bodies, implying immediately that in the relative
motion system, the potential was going to be spherically symmetric. At this point then,
we shall consider the case of an electron ‘circling’ a nucleus of charge Z, meaning that the
potential can be written as
Z e2
V (r) = − . (93)
r
A way to avoid errors during calculations is to deal with the simplest possible expressions,
and in this case simpler expressions can be obtained by redefining r = λρ and E = ER E
where
~2
λ = which is called the Bohr radius, (94)
µe2
and
µe4
ER = is called the Rydberg energy. (95)
2~2
2
The Rydberg energy can also be written as ER = µc2 α, where we introduced the fine structure
constant
e2 1
α = ' . (96)
~c 137
In terms of the new variables ρ and E, we can rewrite eq. (92) as
 
2 2Z l(l + 1)
H(l) Φl ≡ −∂ρ − + Φl = EΦl , (97)
ρ ρ2
which can be thought of as being an ordinary 1-dimensional quantum mechanical system.
In order to find the spectrum of the above Hamiltonian, we shall apply the technique of
factorisation of Hamiltonians outlined in sec. (2). To this end consider the prepotential
Z l+1
Wl ≡ − . (98)
l+1 ρ

16
Given this prepotential, the calculation of the Hamiltonian (35) and its dual (38) is straight-
forward and leads to

(l) 2Z l(l + 1) Z2
H− = −∂ρ2 − + + , (99)
ρ ρ2 (l + 1)2

(l) 2Z (l + 1)(l + 2) Z2
H+ = −∂ρ2 − + + , (100)
ρ ρ2 (l + 1)2
(l)
There are two thing to observe about the Hamiltonians H∓ : first of all we have the relation
(l)
of H− with Hl . This relation is

(l) Z2
H− = Hl + , (101)
(l + 1)2

which, as the ground-state energy of a H− -type factorised Hamiltonian is zero, implies that
the groundstate energy of the Hamiltonian Hl is

Z2 Z2
E0,l = − which in the original variables reads: E0,l = − ER . (102)
(l + 1)2 (l + 1)2

Clearly, the lowest value of the energy is attained for l = 0, which then corresponds to the
true groundstate energy of the hydrogen-like atoms. For the hydrogen atom we have Z = 1
and µ = me = 511 keV/c2 , which leads to

E0,0,0 = −ER ' −13, 6 eV . (103)

(l) (l+1)
The second thing to observe is the relation of the dual Hamiltonian H+ to H− ; this
relation is
(l) (l+1) Z2 Z2
H+ = H− + − . (104)
(l + 1)2 (l + 2)2
Consider for clarity’s sake, first the case of l = 0. Then as we know the ground-state energy of
(0) (0) (0)
the Hamiltonian H− is E0 = 0. The energy of the first excited state is E1 , which we know
(0) (0) (0)
to be equal to the energy of the same state of the dual Hamiltonian H+ , whence E1 = Ẽ1 .
(0) (1) 2 2
The recurrency relation in eq. (104) then means that Ẽ1 = E0 + Z 2 − Z22 = Z 2 − Z22 , where
in the last step we used the fact that the groundstate of any factorised Hamiltonian H− has
zero energy. Applying the same reasoning to the second excited state we find
(0) (0) (1) Z2
E2 = Ẽ2 = E1 + Z 2 − 22
(1) Z2 (2) Z2
= Ẽ1 + Z 2 − 22
= E0 + Z2 − 32
Z2
= Z2 − 32
. (105)

It is then a small step to deduce that for the k th state we have

(0) Z2
Ek = Z2 − . (106)
(k + 1)2

17
 (l)
At this point the deduction of the energy level of the Hamiltonian H− should pose no problem
and in fact the energy of the k th state is given by

(l) Z2 Z2
Ek = − . (107)
(l + 1)2 (k + l + 1)2

The above energy levels fully deplete the discrete part of the spectrum and we can use
them to write down the possible energy levels of the original system by remembering eqs. (101)
and (95) and that the Hamiltonian Hl lead to an (2l + 1)-fold degeneracy in the full system:
we see that there are energy levels given by integers k = 0, 1, . . . , ∞, l = 0, 1, . . . , ∞ and
m = −l, . . . , l whose energy is given by

Z2
E(k, l, m) = − ER , (108)
(k + l + 1)2

from which the announced (2l+1)-fold degeneracy is obvious as the energies do not depend m.
What is perhaps less obvious, but surely well-known, is that there are further degeneracies.
Indeed, as one can see we have that E(1, 0, 0) = E(0, 1, m) indicating that there are further
degeneracies. The full degeneracy can be highlighted by the introduction of the so-called
principal radial quantum number n (n = 0, 1, . . . , ∞) to write the energy simply as13

Z2
En,l,m = − ER , (109)
(n + 1)2

meaning that for a given n the possible pairs of integer number (k, l) have to satisfy k+l = n; it
immediately implies that for a given n the possible angular momentum quantum numbers are
l = 0, 1, . . . , n. The total degeneracy of the level n can then be easily calculated by observing
that there is no degeneracy due to the number k, leading to an (n + 1)2 -fold degeneracy.14
Using the BPS equation (31) and the relations (41,42), it is a reasonably straightforward
yet tedious task to calculate the explicit forms of the wavefunctions. The easiest ones are
 l+3/2
−1/2 2Z  
Ψl,l,m = [(2l + 2)!] rl exp − (l+1)λ
Z l
r Ym (θ, φ) . (110)
(l + 1)λ

13
Observe that in the literature the principal radial quantum number does not start at 0, but rather at
n = 1. Please keep this in mind when you are comparing books with the present notes.
14
One can ask oneself the following question: Since the (2l + 1)-fold degeneracy was due to the spherical
symmetry present in the original system, what is the symmetry reason we end up with an even bigger degen-
eracy? The clue to finding the responsable symmetry can be found in classical mechanics. In the study of
the earth-sun system, or Kepler problems in general, one finds that there exists an extra conserved current
due to the so-called Laplace-Runge-Lenz vector, which renders the Kepler problems integrable. The quantum
mechanical version of the Laplace-Runge-Lenz vector combines with the spherical symmetry of the hydrogen
atom, mathematically this symmetry is described by a group denoted by SO(3), into the bigger symmetry
group SO(4), which causes the above degeneracy [5].

18
References
[1] R. W. Robinett: Quantum mechanics; Oxford U.P., 1997.

[2] E. Schrödinger: “A method of determining quantum-mechanical eigenvalues and eigen-

functions”, Proc. Roy. Irish Acad. (Sect. A) 46(1940), 9–16.

[3] P. M. Morse: “Diatomic molecules according to the wave mechanics. 2. vibrational levels”,
Phys. Rev. 34(1929), 57–64.

[4] V. W. Hughes, D. W. McColm, K. Ziock, R. Prepost: “Formation of Muonium and

Observation of its Larmor Precession”, Phys. Rev. Lett. 5(1960), 63–65.

[5] L. A. Takhtajan: Quantum mechanics for mathematicians; American Mathematical So-

ciety, 2008.

19