221B Lecture Notes

Scattering Theory III

1 Partial Wave Analysis

1.1 Partial Wave Expansion
The scattering amplitude can be calculated in Born approximation for many
interesting cases, but as we saw in a few examples already, we need to work
out the scattering amplitudes more exactly in certain cases. The useful
method is the partial wave analysis. When the potential is central, i.e.,
spherically symmetric V (r) = V (r), angular momentum is conserved due
to Noether’s theorem. Therefore, we can expand the wave function in the
eigenstates of the angular momentum. Obtained waves with definite angular
momenta are called partial waves. We can solve the scattering problem for
each partial wave separately, and then in the end put them together to obtain
the full scattering amplitude.
The starting point is the asymptotic behavior of the wave function
eikr
ψ(x) ∼ eikz + f (θ)
. (1)
r
We use the formula shown in “Notes on Spherical Bessel Functions”
∞


eikz = (2l + 1)il jl (kr)Pl (cos θ). (2)
l=0

The plane wave contains all values of l. This can be understood intuitively
as follows. The plane wave is infinitely extended in all space. Therefore,
in classical terms, it contains all values of the impact parameter b. For a
fixed value of the momentum p = h̄k, the angular momentum is L = bp and
hence it contains all values of the angular momentum L = h̄l with l = bk.
It is useful later to write down the asymptotic behavior at large r using
jl (kr) ∼ sin(kr − lπ/2)/kr,
∞
1

eikz = (2l + 1)(eikr − (−1)l e−ikr )Pl (cos θ). (3)
2ikr l=0
It contains both the wave converging to the origin and the wave emerging
from the origin, as intuition suggests.

1
 Similarly, the 2nd term should be expanded in terms of the partial waves.
∞


f (θ) = (2l + 1)fl Pl (cos θ). (4)
l=0

All information on physics is contained in the complex numbers fl . The

factor 2l + 1 is for mere convenience.

1.2 Optical Theorem Constraint

The optical theorem places important constraints on fl . Recall the optical
theorem 
4π
σ = dΩ|f |2 = f (0). (5)
k
The l.h.s. is

σ = dΩ|f |2


= 2π d cos θ (2l + 1)(2l + 1)fl fl∗
Pl (cos θ)Pl
(cos θ)
l,l



= 4π (2l + 1)|fl |2 . (6)
l

Here, normalization and orthogonality of the Legendre polynomials

 1
2
Pl (t)Pl
(t)dt = δl,l
(7)
−1 2l + 1
was used. The r.h.s. of Eq. (5) is, on the other hand,
4π 4π

f (0) = (2l + 1)(f ). (8)
k k l

Here, I used the fact Pl (cos θ) = 1 for cos θ = 1. Comparing Eqs. (6) and
(8), we find
1
|fl |2 = fl (9)
k
This is an important result. This constraint can be rewritten as

|1 + 2ikfl |2 = 1. (10)

2
In other words, the combination 1 + 2ikfl is just a phase

e2iδl = 1 + 2ikfl (11)

or equivalently,
e2iδl − 1 1
fl = = eiδl sin δl . (12)
2ik k
The meaning of this phase δl becomes clearer shortly.

1.3 Phase Shifts

We write the asymptotic form Eq. (1) in terms of partial waves. It is
∞ ∞

eikr

ψ(x) ∼ (2l + 1)il jl (kr)Pl (cos θ) + (2l + 1)fl Pl (cos θ). (13)
l=0 r l=0

Using the asymptotic behavior jl (kr) ∼ sin(kr−lπ/2)/kr and fl = eiδl sin δl /k,
∞    
1
π
ψ(x) ∼ (2l + 1)Pl (cos θ) il sin kr − l + eikr eiδl sin δl . (14)
kr l=0 2

The terms in the square bracket can be combined as

1
∞  
ψ(x) ∼ (2l + 1)Pl (cos θ) e2iδl eikr − (−1)l e−ikr . (15)
2ikr l=0

Or in other words, the radial wave function behaves as

1 1  
Rl (r) ∼ e2iδl eikr − (−1)l e−ikr . (16)
2ikr 4π(2l + 1)

(Here I used the normalization factor Yl0 (θ, φ) = (2l + 1)/4πPl (cos θ), but
the rest of the discussions does not depend on this factor.) This is an inter-
esting equation. Compare it to the case of the plane wave without scattering
Eq. (3). What this equation says is that the wave converging on the scat-
terer e−ikr has the well-defined phase factor −(−1)l , the same as in the case
without scattering. This is because it comes from the expansion of the plane
wave part only. On the other hand, the wave that emerges from the scatterer
has an additional phase factor e2iδl . All what scattering did is to shift the

3
phase of the emerging wave by 2δl . The reason why this is merely a phase
factor is the conservation of probability. What converged to the origin must
come out with the same strength. But this shift in the phase causes the
interference among all partial waves differ from the case without the phase
shifts, and the result is not a plane wave but contains the scattered wave.
The phase factor is the so-called S-matrix element

Sl = e2iδl , (17)

This is related to the T -matrix discussed earlier by

S = 1 + iT, (18)

and hence
Tl = 2eiδl sin δl = 2kfl . (19)
The S-matrix includes the transmitted wave, while T -matrix removes it and
keeps only the scattered wave. This notation of S- and T -matrix is used
extensively in the time-dependent formulation of the scattering problem.
In terms of the phase shifts, the cross section is given by
4π

σ= 2
(2l + 1) sin2 δl . (20)
k l

Actual calculation of phase shifts is basically to solve the Schrödinger

equation for each partial waves,


1 d2 l(l + 1) 2m
− 2
r+ + 2 V (r) Rl (r) = k 2 Rl (r). (21)
r dr r2 h̄

After solving the equation, we take the asymptotic limit r → ∞, and write
Rl (r) as a linear combination of jl (kr) and nl (kr). The relative coefficients
of jl and nl determines the phase shift δl , and hence the cross section.

1.4 Unitarity Limit

It is interesting that there is a maximum possible value for the cross section
for each partial wave. From Eq. (20), the cross section for a given partial
wave l is
4π 4π
σl = 2 (2l + 1) sin2 δl ≤ 2 (2l + 1). (22)
k k
4
The maximum is obtained for the phase shift δl = ±π/2. This is called
unitarity limit, as it is a consequence of the unitarity of the S-matrix.
This limit can be qualitatively understood in the following semi-classical
argument. If you inject a particle with momentum p at the impact param-
eter b, it has angular momentum L = pb. On the other hand, the angular
momentum is quantized in quantum mechanics, L = h̄l. Let us say that the
quantized angular momentum l corresponds roughly to the impact parameter
b = L/p for h̄l < L < h̄(l + 1), i.e.,

l l+1
≤b≤ . (23)
k k
Assuming that the particle gets scattered with 100% probability when enter-
ing this ring, the classical cross section would be
 2  2
l+1 l 2l + 1
π −π =π . (24)
k k k2

The unitarity limit is roughly the same as this semi-classical argument except
for a factor of four.

2 Hard Sphere Scattering

As an example of partial wave analysis, we first look at the hard sphere
scattering, with the potential

0 (r > a)
V = (25)
∞ (r < a)

This potential represents an impenetrable ball, and mimics the classical im-
age of scattering. Because of infinite potential within the radius a, Born
approximation is clearly not appropriate. We resort to the partial wave anal-
ysis to work out cross sections. The infinite potential corresponds to the
boundary condition
Rl (a) = 0 (26)
for all l in solving the Schrödinger equation Eq. (21).

5
2.1 S-wave
At low momenta k 1/a, the centrifugal barrier inhibits the particle from
entering the region of the scatterer. Therefore the scattering occurs only for
the l = 0 partial wave, or S-wave. We first analyze the S-wave only, which
turns out to be particularly simple. The analysis below, however, applies
also when k is large.
The Schrödinger equation Eq. (21) is simply that of a free particle in one
dimension

d2 2m
− 2 + 2 V (r) (rR0 (r)) = k 2 (rR0 (r)), (27)
dr h̄
with the boundary condition rR0 = 0 at r = a. Therefore the solution is
uniquely
c  ikr−ika 
rR(r) = c sin(k(r − a)) = e − e−ikr+ika , (28)
2i
where c is an overall normalization factor (in general complex).
To determine the phase shift, we compare this solution to the general
expression Eq. (16)
1 1  2iδ0 ikr 
R0 (r) ∼ √ e e − e−ikr , (29)
2ikr 4π
and we find
δ0 = −ka. (30)
The reason behind the phase shift is obvious. Because the wave cannot
penetrate into r < a, the wave is shifted outwards, which is the shift in the
phase −ka.
The cross section from the S-wave scattering is obtained from Eq. (20),
4π 4π
σ0 = 2
sin2 δ0 = 2 sin2 ka. (31)
k k
The maximum cross section occurs at k = 0, where σ0 = 4πa2 . This is four
times larger than the classical geometric cross section πa2 , but at least of
the same order of magnitude. The partial wave cross section saturates the
unitarity limit where ka = nπ, and keeps oscillating to higher momenta.
The oscillating behavior merely signals the finite size of the target. How-
ever, the persistent oscillation up to infinite momentum is because of the
oversimplification of impenetrable sphere with a rigid surface.

6
 Because this analysis is so simple, let us generalize the discussion to the
case of a little bit penetrable potential

0 (r > a)
V = (32)
V0 (r < a)

We use the notation K 2 = 2mV0 /h̄2 . First consider the situation k > K. We
then find  √
sin( k 2 − K 2 r) r < a
rR = . (33)
sin(ka + δ0 ) r>a
By matching the logarithmic derivatives of the wave function at r = a, we
find
(rR) √ 2 √
= k − K 2 cot( k 2 − K 2 a) = k cot(ka + δ0 ), (34)
rR
or

−1 k √
δ0 = tan √ tan( k − K a) − ka.
2 2 (35)
k2 − K 2
For k  K, one can neglect K and the phase shift vanishes. The energy
is too large to care the slight potential and there is no scattering any more.
Therefore the partial wave cross section does not saturate the unitarity limit
at k  K and asymptotes to zero.
On the other hand, for small k < K, the wave function is
 √
sinh( K 2 − k 2 r) r < a
rR = . (36)
sin(ka + δ0 ) r>a

The phase shift is obtained as


−1 k √
δ0 = tan √ tanh( K 2 − k 2 a) − ka. (37)
K −k
2 2

At small k K, it can further be expanded as

 
1
δ0 ∼ ka tanh Ka − 1 . (38)
Ka
The quantity in the square bracket is always between 0 and −1, and hence the
cross section is at most that of the hard sphere. This makes sense intuitively.
Another interesting point is that the phase shift δ0 always starts linearly with

7
k at small momentum, and the slope is negative. This is a completely general
result for a repulsive potential, and a convenient quantity
dδ0
a0 = − lim (39)
k→0 dk

is called the scattering length, as it has the dimension of the length. This
quantity basically measures how big the scatterer is. The cross section at
k → 0 limit is then given by 4πa20 . For the hard sphere potential, the
scattering length is indeed the size of the sphere.
This example basically demonstrates that the cross section of α particle
off an atom, which we discussed within Born approximation before, cannot
be much larger than the geometric cross section given by the size of the atom
despite what Born approximation suggested for k a−1 .

2.2 Higher Partial Waves

For the hard sphere problem, the phase shifts for higher partial waves can
be worked out also easily. For r > a, the Schrödinger equation is again that
of the free one, and hence the solution is a linear combination of jl and nl ,

Rl ∝ jl (kr) cos θ + nl (kr) sin θ

     
π π π
∼ sin kr − l cos θ + cos kr − l sin θ = sin kr − l + θ (40)
.
2 2 2
Comparing this asymptotic form at r  a to the definition of the phase shift
 
2iδl ikr l −ikr iδl l π
Rl ∝ e e − (−1) e = e i 2i sin kr − l + δl , (41)
2
θ above is nothing but the phase shift δl .1 Then we require that the wave
function vanishes at r = a:

Rl (a) ∝ jl (ka) cos δl + nl (ka) sin δl = 0. (42)

Therefore we find
jl (ka)
tan δl = − , (43)
nl (ka)
1
I’ve just noticed that my notation for nl differs from Sakurai’s by a sign as seen in
Eq. (7.6.52) on page 409. I’m sorry for that, but to keep consistency with my notes on
spherical bessel functions, I stick with my convention, which was taken from Messiah.

8
or
(−)
hl (ka)
e2iδl = (+)
. (44)
hl (ka)
For small momenta k a−1 , it is straightforward to show that δl ∝ k 2l+1 ,
and hence is smaller for higher partial waves. This is easy to understand.
When k is small, the centrifugal barrier −h̄2 l(l +1)/r 2 does not allow particle
to reach r = a classically, and only the exponential tail of the wave function
reaches r = a. Therefore the effect of the potential is extremely suppressed.
On the other hand at high monenta, sin2 δl oscillates between 0 and 1 as
a function or l up to l ka. Above this value, the phase shift drops rapidly
to zero. This makes sense from the classical physics intuition. When l ≥ ka,
the impact parameter is larger than the size of the target b ≥ a and there
shouldn’t be any scattering. This behavior of the phase shift leads to the
total cross section or 2πa2 at high momenta.

3 Attractive Potential
The potential Eq. (32) can also be an attractive potential if V0 < 0. The
phase shift for this case can be easily obtained from Eq. (35) by changing
the sign of K 2 = −2mV0 /h̄2 ,


−1 k √
δ0 = tan √ tan( k 2 + K 2 a) − ka. (45)
k2 + K 2
The first interesting feature of this phase shift is that it can start with a
positive slope unlike the repulsive case. The scattering length is
  
dδ0  tan Ka
a0 = −  = a 1 − . (46)
dk k=0 Ka
For small K, the scattering length is negative, i.e., the opposite sign of the
repulsive case. This is easy to understand because the wave is pulled into
the potential rather pushed out unlike the repulsive case. However, once we
make the potential more attractive (larger K), the scattering length grows
and becomes even infinite at K = π/2! What is going on?
To answer this question, let us study the analytic structure of the scat-
tering amplitude more carefully. From Eq. (45), we can write
√
1 + i √ k tan( k 2 + K 2 a)
2
k +K 2
e2iδ0 = e−2ika √ (47)
1 − i √k2k+K 2 tan( k 2 + K 2 a)

9
This S-matrix element can have a pole if
k √
1 − i√ tan( k 2 + K 2 a) = 0. (48)
k2 + K 2
This equation appears impossible to satisfy, but it can be on the complex
plane of k. For a pure imaginary k = iκ, the equation becomes
√
K 2 − κ2
κ=− √ . (49)
tan( K 2 − κ2 a)

This is nothing but the condition for bound states. By decreasing K from
a sufficiently large value with bound state(s), the bound state energies E =
−h̄2 κ2 /2m move up. When Ka = (n + 12 )π, tan Ka = ∞, and we find a
bound state approaching κ = k = 0. This is when the scattering length
diverges in Eq. (46). In other words, the infinite scattering cross section at
k = 0 happens because there is a bound state exactly at k = 0. If you
further decrease K, the bound state completely disappears. However the
cross section for small k remains very large, not quite 4π 2 /k 2 as allowed by
unitarity, but much bigger than 4π 2 a2 .
This can also be seen on the complex k plane in the following manner. The
lower half plane is unphysical as it corresponds to an exponentially growing
wave function at the infinity for the scattered wave eikr . When there are
bound states, you see poles along the positive imaginary axis. By decreasing
K, the poles along the positive imaginary axis go down, and a pole reaches
the origin. By further decreasing K, the pole goes below the origin into the
unphysical region. However, the existence of a pole just below the origin
makes the scattering amplitude at k ∼ 0 large and results in an anomalously
large cross section.
On the other hand, the phase shift Eq. (45) can be multiples of nπ for
special values of k. Even though the particle passes through a potential,
the wave oscillates precisely integer of half-integer times in the potential
in addition to the free particle phase and there is no cross section. This
phenomenon is called Ramsauer–Townsend effect, which had been observed
in the scattering of electrons by rare gas atoms and was a great mystery
before Bohr proposed the wave description.
A good example of large cross sections close to the threshold is the neutron
scattering cross section off large nuclei. The cross section can be many orders
of magnitude larger than the geometric size of nuclei. This is why slow

10
Figure 1: Neutron cross section that shows a large enhancement close to the
threshold and many resonances above the treshold. Taken from Subatomic
Physics, Hans Frauenfelder and Ernest M. Henley, Prentice-Hall, Inc., 1974.

neutrons can be effectively absorbed by Uranimum in nuclear power plants

to cause further fission processes in a chain reaction, and similarly in atomic
bombs.

4 Resonances
The attractive spherical well potential discussed in the previous section led to
possible large cross sections close to the threshold k ∼ 0. Can one obtain large
cross sections away from the threshold? One can, if there are “resonances.”
Recall that the poles in the unphysical lower half plane in k were respon-
sible for large cross section near the threshold. This pole corresponds to a
would-be bound state. Similarly a pole just below the real axis would lead
to a large cross section. They must also correspond to some sort of “bound
states.” What are they?
There are few examples of potential that can be worked out simply and
exhibit resonances. Here we discuss an idealized potential called “delta-shell”

11
potential,
V (r) = γδ(r − a). (50)
This potential leads to a true bound state if γ is sufficiently negative. On
the other hand, for γ → ∞, the regions inside r < a and outside r > a the
potential are decoupled and one finds a tower of states confined inside the
shell. The fate of these states for finite γ is very interesting.
The phase shift for the S-wave can be worked out analytically,
2
h̄ k ika
2iδ0 1 + 2mγ
h̄2 k
e−ika sin ka −2ika
sin ka + 2mγ e
e = 2mγ ika = e 2
h̄ k −ika
. (51)
1 + h̄2 k e sin ka sin ka + 2mγ e

We now look for poles of the denominator, which can be rewritten as

h̄2 k
e2ika = 1 − 2i . (52)
2mγ
When γ is large, the second term is small, and we find 2ika = 2πin, or
ka = nπ. Given the correction, we want to solve

h̄2 k
2ika = log 1 − 2i + 2inπ. (53)
2mγ

Expanding the logarithm up to O(k 2) and solving the quadratic equation,

we obtain  2 2
nπ h̄ (nπ)2
k h̄2 − i 3
+ O(γ)−2 (54)
a + 2mγ 2mγ a
The poles are in the unphysical lower half plane. But when γ is large, the
poles are very close to the real axis, and the scattering amplitude receives a
large enhancement due to these poles.
What are these poles? Unlike the poles along the positive imaginary axis,
which represent real bound states, these poles are unphysical. However in
the limit of γ → 0, or in other words in the limit of small coupling between
the regions inside and outside the shell, they become poles along the real
axis. They are the discrete states inside the shell in this limit. By making
γ finite, you introduce coupling between the discrete states inside the shell
to the continuum states outside the shell. Therefore it is still useful to find
“physical” interpretations of the unphysical poles.

12
 For this purpose, it is instructive to solve Schrödinger equation for the
values of k which correspond to the location of poles satisfying Eq. (52). The
solutions can be found in a straight-forward way

sin kr (r < a)
rR0 (r) = (55)
sin ka eik(r−a) (r > a)

Because the factor eik(r−a) grows expontentially at large r due to the negative
imaginary part in k, the solution is not a regular normalizable solution.
But nonetheless let us proceed. In the large γ limit, polease are given in
Eq. (54) and sin ka ∼ O(γ)−1 is small. Therefore the wave function almost
vanishes at the shell. Out side the shell, the wave function oscillates at
the small amplitude sin ka, which however starts growing again due to the
eik(r−a) factor exponentially. We now put the time dependence in. The energy
eigenvalue is nothing but E = h̄2 k 2 /2m, where k is at the pole. If the pole
is at
k = k0 − iκ (56)
the energy eigenvalue is at

Γ h̄2 k02 h̄2 k0 κ

E = E0 − i = −i + O(κ2 ). (57)
2 2m m
For instance in the large γ limit, the poles given in Eq. (54) give
 2  3
Γ h̄2  nπ  h̄2 nπ 2
E = E0 − i ∼ h̄2
− i 2
+ O(γ)−3. (58)
2 2m a + 2mγ 2ma γ a

The time dependence of the wave function is simply

rR0 (r, t) = rR0 (r)e−iEt/h̄ = rR0 (r)e−iE0 t/h̄ e−Γt/2h̄ . (59)

Overall, the wave function is then


sin kr e−iE0 t/h̄ e−Γt/2h̄ (r < a)
rR0 (r, t) = ik(r−a) −iE0 t/h̄ −Γt/2h̄ (60)
sin ka e e e (r > a)

This is a very interesting result. Inside the shell, it shows an exponen-

tially decaying probability density |rR0 (r, t)|2 ∝ e−Γt/h̄ uniformly over space.
Outside the shell, the probability density is |rR0 (r, t)|2 ∝ e2κr e−Γt/h̄ , which

13
shows the probability flowing out to inifinity with speed Γ/2h̄κ = h̄k0 /m,
nothing but the velocity of the particle itself. In other words, the wave func-
tion describes a “bound state” inside shell decaying into a continuum state
outside the shell moving away at the expected velocity (a “run-away” wave).
Even though the pole is certainly in the “unphysical” region, this interpreta-
tion makes it quite physical at least in the limit of large γ or small coupling
between the discrete and continuum states. The resonances can be viewed as
quasi-bound states which decay into continuum states. The lifetime of the
quasi-bound states is τ = h̄/Γ.
Is the complex energy eigenvalue allowed? You have been repeatedly told
that a Hermitean operator, such as Hamiltonian, has only real eigenvalues.
However, this statement is true for normalizable wave functions, because
the proof crucial depends on the integration by parts and for unnormaliz-
able wave functions integral themselves are ill-defined. Once one allows an
exponentially growing wave function, the ordinary proof of real eigenvalues
breakd down, and one can find complex eigenvalues.
In fact, all excited states of an atom appear as resonances in the photon-
atom scattering. In the limit of turning off the coupling of photons to the
electron, the excited states are all stable bound states. But the coupling
(albeit small thanks to α = 1/137 1) lets the excited state decay into the
continuum states of photons.
In general, once we know that there is a pole just below the real axis,
we can approximate the S-matrix by the contribution from the pole only,
ignoring a continuum. Note that this approximation is good only if the pole
is close to the real axis. Then as a function of the energy, S-matrix element
is approximated as
g(E)
Sl = e2iδl . (61)
E − E0 + iΓ/2
Because of the unitarity |S|2 = 1, we immediately conclude
E − E0 − iΓ/2
Sl = e2iδl . (62)
E − E0 + iΓ/2
Then
Γ2
sin2 δl = . (63)
(E − E0 )2 + Γ2 /4
At E = E0 , it saturates the unitarity limit sin2 δl = 1, and it has a Lorenzian
shape in terms of the energy. Γ is nothing but the FWHM (Full-Width-Half-
Maximum) of the Lorentzian peak in sin2 δl . Cross section is σ = (4π/k 2 )(2l+

14
1) sin2 δl as usual. Comparing the discussion of the decaying probability
density with a run-away wave and the dependence of the cross section on the
energy, we established the relationship between the life time of the quasi-
bound state τ and the FWHM of the resonance Γ as τ = h̄/Γ. This is an
explicit manifestation of the energy-time uncertainty relation ∆E∆t ∼ > h̄.
Another interesting point is that the real part of the energy eigenvalue for
the resonances is shifted from the limit γ → 0. In other words, the fact that
the quasi-bound state can decay into continuum changes the energy of the
quasi-bound state due to the coupling to the continuum. In fact, the energies
of the excited states of an atom are different from the energies calculated
without considering the decay and the difference has to be included given
the high accuracy of atomic physics experiments.
Coming back to the real energy eigenvalue and the delta-shell potential,
the wave function is given by
 sin(ka+δ )
0
sin(ka)
sin(kr) r < a
rR0 (r) = . (64)
sin(kr + δ0 ) r>a
From Eq. (51), we find


h̄2 k ika
δ0 = −ka + arg sin ka + e . (65)
2mγ
For most values of ka, the second term in the square bracket is negligible and
the second term vanishes. Therefore the result is approximately the same as
the hard sphere. Inside the shell, the prefactor sin(ka + δ0 ) is basically zero.
In other words, the wave does not enter the shell. On the other hand, for
special values ka = 2nπ, the second term in the square bracket quickly moves
from 0 to π (and π to 2π for ka = (2n − 1)π). Only for these values of k, the
prefactor sin(ka + δ0 )/ sin(ka) can be sizable, but at most unity. The wave,
therefore, enters the shell only for the “resonant” values of k.

5 Coulomb scattering
The case of Coulomb potential is somewhat special because the potential turn
off at infinity rather slowly. In fact, the formalism we used so far assumed
that the potential dies quickly enough to justify the asymptotic form
eikr
ψ(x) ∼ eikz + f (θ) . (66)
r
15
This asymptotic behavior, however, is not valid for the Coulomb potential.
Coulomb pontential is long-ranged and distorts the wave function even at
large distances. In order to see this, we need to solve the equation exactly.
As usual, we go back to the Schödinger equation


h̄2 ZZ  e2
− ∆+ ψ(x) = Eψ(x). (67)
2m r

We introduce some notations:


h̄2 k 2 1 h̄k
E= = mv 2 v= , (68)
2m 2 m

ZZ  e2 1 h̄2
γ= = aB = . (69)
h̄v kaB ZZ  e2 m

The dimensionless parameter γ controls the impact of the Coulomb field on

the wave function. In terms of these quantities, the Schrödinger equation
can be written much more simply as


2γk
2
∆+k − ψ(x) = 0. (70)
r

To solve Eq. (70), we take the ansatz

ψ(x) = eikz f (u), u = r − z. (71)

By substituting the ansatz into Eq. (70), we find

d2 d
u 2
f + (1 − iku) f − γkf = 0. (72)
du du
Introducing yet another variable v = iku = ik(r − z), it becomes

d2 d
v 2 f + (1 − v) f + iγf = 0. (73)
dv dv
This is a differential equation of Laplace-type and hence its solution is given
in terms of a confluent hypergeometric function. Putting all pieces together,
the solution is given as

ψ(x) = Aeikz F ( − iγ|1|ik(r − z)). (74)

16
A is an arbitrary overall normalization factor. The exact solutions are some-
times called Coulomb harmonics.
Details of the hypergeometric functions are not of our interest. But we
are interested in the asymptotic behavior of the function. By choosing the
normalization factor A = Γ(1 + iγ)e−πγ/2 for convenience, the asymptotic
behavior is given as


i(kz+γ log k(r−z)) γ2
ψ ∼ e 1+ +···
ik(r − z)


γ Γ(1 + iγ) i(kr−γ log k(r−z)) (1 + iγ)2
− e 1+ + · · · . (75)
k(r − z) Γ(1 − iγ) ik(r − z)

The terms indicated by dots are suppressed by higher powers in 1/(r − z).
Clearly this expression is not useful when r = z, i.e., the extreme forward
region. But as we discussed in “Scattering I,” the scattering cross section
does not deal with the forward region because it ignores the interference
term. Therefore we will not worry about the subleading terms in 1/(r − z)
and keep only the leading term 1 in the asymptotic expansion.
The asymptotic form of the wave function in Eq. (75) is not quite that
in Eq. (66), but is similar enough to allow us to read off the scattering
amplitude. We slightly modify the definition of the scattering amplitude
from Eq. (66) as

ei(kr−γ log k(r−z))

ψ(x) ∼ ei(kz+γ log k(r−z)) + f (θ) . (76)
r
One can check, for example using wave packets, that this generalized defini-
tion still gives the probability of the particle to be scattered in a given solid
angle. Comparing Eqs. (76) and (75), we find
γ Γ(1 + iγ) r γ Γ(1 + iγ) 1
f (θ) = − =− (77)
k Γ(1 − iγ) r − z k Γ(1 − iγ) 2 sin2 θ/2
The scattering cross section is then obtained by the usual formula
 2
dσ γ2 ZZ  e2 1
= |f (θ)|2 = 2 4 = 4 . (78)
dΩ 4k sin θ/2 4E sin θ/2
It is a coincidence that the Born approximation and the classical calculation
both agree with this exact quantum mechanical result.

17
 The fact that we did not have to worry about logarithmic correction in
the exponents to obtain the scattering amplitude may make you wonder if
the distortion of the wave function is of any physical significance. For the
Rutherford scattering itself, certainly it does not matter. However, the dis-
tortion has importance consequences on other processes. One prime example
is the nuclear β-decay. As well-known, nuclear β-decay transforms one type
of nucleus with A = Np + Nn , Z = Np to another one with the same A
but a smaller atomic number Z − 1 by emitting an electron e− and anti-
electron-neutrino ν̄e . When the β-electron escapes the nucleus, it is subject
to the binding due to the Coulomb interaction. To calculate the decay matrix
element, it is important to use Coulomb harmonics rather than plane waves.
Coming back to the scattering problem, now that we have the scattering
amplitude, we can look for poles. Note that the Gamma function does not
have zeros, and hence we look for poles of the numerator Γ(1 + iγ). The
poles of Γ(z) are located at z = 0, −1, −2, · · ·, or in other words −n + 1 for
n = 1, 2, · · ·. Therefore the poles are at
1 + iγ = −n + 1 (79)
Recalling the definition of γ in Eq. (69), γ = 1/kaB , we clearly need a pure
imaginary k: bound states. To be in the physical region (upper half plane
k = iκ with κ > 0) to have an exponentially damping function at large radii,
and to satisfy the condition Eq. (79), we need
1
= −n, (80)
κaB
2
or in other words a negative aB = ZZh̄
e2 m . It is possible only when ZZ  < 0,
i.e. when the Coulomb potential is attractive. This is indeed what we expect.
The energy levels are then obtained as
h̄2 κ2 h̄2 Z 2 Z 2 e4 m Z 2 Z 2 α2 mc2
E=− =− = − = − . (81)
2m 2ma2B n2 2h̄2 n2 2n2
This is nothing but the Bohr levels of hydrogen-like atoms as expected.

6 Two-to-two Scattering
We have discussed only the scattering of a particle by a static potential. In
practice, a potential is generated by another particle, and we need to discuss

18
two-to-two scattering problems. As long as the scattering is elastic, namely
if the initial state particles and final state particles are the same, what we
have done applies directly to realistic problems.
The point is just the separation of the center-of-mass motion in the two-
body system. Starting from the two-particle Hamiltonian,

p21 p2
H= + 2 + V (|x1 − x2 |), (82)
2m1 2m2
we separate the center-of-mass motion by defining
1
P = p1 + p2 , p1 − p2 ),
p = (
2
 = m1x1 + m2x2 ,
X x = x1 − x2 . (83)
m1 + m2
Then the Hamiltonian becomes

P 2 p2
H= + + V (|x|), (84)
2M 2µ

with M = m1 + m2 and µ = m1 m2 /(m1 + m2 ). Then the problem reduces

to the potential scattering problem for a particle of mass µ.
There is, however, one interesting complication due to quantum statistics.
If two particles that scatter are idential particles, such as electron-electron
scattering or scattering of two idential atoms, symmetry of the wave function
needs to be considered. Under the interchange of two particles x1 ↔ x2 , p1 ↔
p2 , the center of mass motion is not affected, but the relative coordinates
change their signs x ↔ −x, p ↔ − p. If they have spins, their spins need to
be interchanged at the same time.
If two particles are idential spinless bosons, say two Helium atoms (as-
suming 4 He isotopes), there is no spin degrees of freedom and the interchange
of particles is simply x → −x in the wave function. Because they are bosons,
the wave function should not change under the interchange of particles, and
hence the wave function must be an even function of x. Therefore the asymp-
totic form of the wave function Eq. (66) must be changed to

eikr
ψ(x) ∼ eikz + e−ikz + [f (θ) + f (π − θ)] . (85)
r

19
The scattering amplitude f (θ) is calculated without the statistics in con-
sideration, and the combination in the square bracket symmetrizes it. The
differential cross section is then found to be
dσ
= |f (θ) + f (π − θ)|2 . (86)
dΩ
There is a constructive interference at θ = π/2 which can be experimentally
observed. Note that one should not integrate over the entire solid angle to
obtain the total cross section because (θ, φ) and (π − θ, φ + π) correspond to
an identical state:  2π  1
dσ
σ= dφ d cos θ . (87)
0 0 dΩ
For two spin 1/2 fermions, there are two possible spin wave functions,
symmetric S = 1 and anti-symmetric S = 0. Therefore depending on the
spin wave function, we either have a anti-symmetric or symmetric spatial
wave function, respectively. In particular, the differential cross section is
the same for the spinless bosons Eq. (86) for the anti-symmetric spin wave
function S = 0, while it is
dσ
= |f (θ) − f (π − θ)|2 (88)
dΩ
for the symmetric spin wave function S = 1. In the latter case, the differ-
ential cross section vanishes identially at θ = π/2. This is an interesting
observation, and one can actually isolate S = 1 combination by studying
θ = π/2 region.
Many scattering phenomena of interest are inelastic, i.e., the final state
particles are not the same as the initial state particles. For instance, when
an electron scatters off an atom, the final state atom may be in an excited
state. Or one of the electrons bound to the atom may be kicked out from
the atom. These are examples of inelastic scattering problems. We will not
discuss these problems, but obviously the combination of elastic and inelastic
processes will tell us a great deal about the nature of the object you study
by scattering processes.

20