Lecture 19: Scattering: Born approximation (11/29/2005)

The partial wave analysis was the first important technique to study scattering. We started with it
last week and continued today.
Additional reading if you wish: Bransden & Joachain, ch. 13.5.
The second important technique to study scattering is the Born approximation and we will look
at it today. This approximation is useful especially if the potential is weak. But it does not require
any conditions on the energy which can be low or high.

Time-dependent vs. time-independent approach

Griffiths discusses scattering in the Born approximation using time-independent perturbation theory
does not depend on time. We will however use some
which is legitimate because the Hamiltonian H
advantages offered by the time-dependent perturbation theory.
Our strategy will be to imagine that the interaction is turned off for t < 0. Then it turns on
at t = 0 when the incident particle is still approaching the region with a nonzero potential. The
perturbation turns off much later at t = when the particle is already escaping and far away from
the interaction region.
Our solution must start with a good choice of the basis of the Hilbert space. Not surprisingly,
we choose the plane waves which are the eigenstates of the free Hamiltonian (with the appropriate
time dependence to solve the unperturbed Schrodinger equation):
1
~
(~k, ~r, t) = q
eik~riEk t/h
(2)3 V
where the volume V was inserted to make the inner product of these wavefunctions defined as an
integral over the volume to be canonically normalized. In the integrals over the whole space, 1/V
survives
Z
(~k ~k 0 )
d3 r (~k 0 , ~r, t)(~k, ~r, t) =
V
but the right hand side can be written as Kronecker delta. At any rate, every initial wavefunction
may be expanded in terms of these basis vectors:
(~r, t) =

c(~k, t)(~k, ~r, t)d3 k

Well, thats nothing else than a Fourier transformation. Without the potential, the coefficients
c(~k, t) would be t-independent. With the perturbation V (~r) added, we must allow them to depend
on time. The dependence is determined by the Schrodinger equation. If we plug (~r, t) into the
equation

h
2 2
i
h
=
+ V (~r)(~r, t),
t
2m
we obtain
i
h

Z
c(~k, t) ~
(~k, ~r, t) 3
3
(k, ~r, t)d k + i
h c(~k, t)
dk=
t
t
Z
h
2
=
c(~k, t)2 (~k, ~r, t)d3 k + V (~r)(~r, t)
2m

The terms adjacent to the = symbol cancel because of the unperturbed equation, so we have
i
h

c(~k, t) ~
(k, ~r, t)d3 k = V (~r)(~r, t)
t

Multiply it by (~k 0 , ~r, t) and integrate over space.

i
h

Z Z

Z
c(~k, t) ~ 0
(k , ~r, t)(~k, ~r, t)d3 r d3 k = (~k 0 , ~r, t)V (~r)(~r, t)d3 r
t

The r-integral on the left-hand side has already been calculated as (~k ~k 0 )/V which simplifies
the equation to
Z
c(~k 0 , ~t)
i
h
= V (~k 0 , ~r, t)V (~r)(~r, t)d3 r
t
On the right hand side, we already know the simple time-dependence of . Also, when we do the
approximation, we may approximate by the unperturbed incoming wave which has also a simple
separated dependence on space and time:
1
eEk t/ih ~k (~r)
(~r, t) = q
3
(2) V
The equation for c therefore says
i
h

c(~k 0 , ~t)
1 i(Ek0 Ek )t/h i~k~r
=
e
e
t
(2)3

Solving the equations

The usual tricks of the first-order time-dependent perturbation theory require that we set the correct
initial conditions c(~k 0 , t) = 0 for ~k 0 6= ~k and that we integrate the equation from t = 0 to t = ,
giving us the usual formulae from the 13th lecture:
i
hc(~k 0 , t) =

1
it/2 sin(t/2)
2e
(2)3

~0

d3 r eik ~r V (~r)k (~r)

where = (Ek0 Ek )/
h. The transition rate to an infinitesimal volume d3 k 0 around the point ~k 0 in
the momentum space is
1 1 ~0 2 3 0
1
4 sin2 ( /2) 
R(~k 0 )d3 k 0 =
|c(
k
,
t)|
d
k
=

V2
2
(2)6 V 2h
2

i~k 0 ~
r

V (~r)k (~r)d

2

r  d3 k 0

The fraction starting with 4 goes to 2() as ; see lecture 13. Also, the measure is
d3 k 0 = k 02 dk 0 d
and because
Ek0 =

h
2 k 02
2m

dEk0 h
d =

2
h2 k 0 dk 0
,
2m

we may also write

dk 0 =

m
d
h
k0

d3 k 0 =

mk 0
d d
h

The transition rate per unit solid angle is thus


2

1
m Z Z i~k0 ~r
0 d
3 
~
R(k )
=
d  e
V (~r)k (~r)d r  2()k 0d
2
6
2
V

(2) V h
h

The factor of () guarantees energy conservation i.e. |k| = |k 0 |. You may also recall from previous
discussions that the cross section d may be obtained from R(~k 0 )d if we multiply it by the incident
flux
v
1 h
k
=
3
3
(2) V
(2) V m
which tells us that
Z
2
1 m2  i~k0 ~r
d
2
3 
~
|f (k, )| =
e
V (~r)k (~r)d r 
d
(2)2 h
4 
You can see that we can actually unsquare this equation and express the scattering amplitude
directly (although you must believe that the phase (minus sign) below is the correct one to agree
with other definitions):
Z
m
~0
eik ~r V (~r)k (~r)d3 r
f (~k 0 , ) =
2
2
h

Plugging plane waves: approximation

When you look at the derivation, the result we obtained is still exact, in fact. The first approximation
we are going to make is that we choose the incident wavefunction to be
~

k (~r) = eik~r .
You know it is not exact because the potential is neglected. With this approximation, the scattering
amplitude becomes
Z
m
~ ~0
0
~
f (k , ) =
ei(kk )~r V (~r)d3 r.
2
2
h
You see that the scattering amplitude is a Fourier mode of the potential corresponding to the
~ = ~k ~k 0 . Then a simple
momentum difference between the initial and the final state. Define K
picture (draw it!) implies that

|K| = 2k sin( )
2
~ goes in the positive z axis; no changes are
We can always choose a system of axes such that K
required for central potentials that we want to focus on. Then
Z
Z +1
m Z 2
2m Z
2
iKr cos
f (~k 0 , ) =
rV (r) sin Kr dr
d
r
dr
d
cos

V
(r)e
=

0
1
2
h2 0
h
2K 0
which is the final form of the Born approximation for the central potentials. When is it valid? We
used an eigenstate of the free particle Hamiltonian instead of the full Hamiltonian, and the result is
therefore only valid when these two are similar, i.e. for V (~r) much smaller than the kinetic energy.
In other words, for weak potentials.
Recall the optical theorem from the last lecture: the imaginary part of f is proportional to the
total cross section. This means, among other things, that if f is real, the total cross section vanishes.
Of course this conclusion is just an artifact of the approximation in which V (~r) is small.
Consider the case where kR KR (generic angle) and kR  1. Then sin Kr Kr and we have
Z
2m 2
0
~
f (k , ) 2
r V (r)dr.
h
0
As you see, the result is independent of ~k 0 . Only the total integral of the potential over space
influences the amplitude.
3

Spherical well
Recall that the spherical well has V (r) equal to V0 for r < a and 0 for r > a. The previous
supersimplified formula gives
f () =

2m
h
2

(V0 )r 2 dr =

2mV0 a3
3
h2

As always, d/d = |f ()|2 and

2mV0 a2
h
2

4
d
= a2
= 4
d
9

!2

Its not too surprising that the result agrees with the partial wave analysis of the same problem in
the same limit for momenta. Again, only the s-wave contributes.

Yukawa potential
In 1934, the Japanese physicist Yukawa proposed that much like the electromagnetic force is connected with photons, the strong force that holds nuclei together is connected with a new (predicted)
particle that was called pion or meson. Sure, the particle was later observed and Yukawas idea is
essentially correct although today we talk about quarks to describe the situation more accurately.
The main difference between pions and the photon is that pions are massive, and this leads to a
new kind of attractive potential between particles such as protons that interact with each other by
exchanging the pions:
A exp(r/a)
V =
r
Yukawa argued that this force defeats the repulsive electromagnetic force and keeps protons and
neutrons together in the nuclei. Note that it differs from the Coulomb potential by the extra
exponential (a consequence of the nonzero mass of the pions, as it turns out) in the numerator.
And the people who dislike either particle physics or Japan dont call it the Yukawa potential
but rather the screened Coulomb potential. Well, not all of them. Condensed matter physicists
often call it the Thomas-Fermi potential while Debye potential is used in plasma physics. When
you evaluate this data, you will see that you should call it the Yukawa potential.
What is the scattering amplitude for this potential?
2mA
f () = 2
h
K

er/a sin(Kr) dr =

2mA
1
2
2
h
K + (1/a)2

After the standard algebra,

d
4m2 A2
= |f ()|2 =
d
h
4

1
2
K + (1/a)2

!2

4m2 A2
=
h
4

a2
4k 2 a2 sin2 (/2) + 1

Coulomb potential
We picked the Yukawa potential because it exponentially decreases which makes the integrals nicely
convergent. The Coulomb potential itself is different. The integral of
V =

q1 q2
40 r
4

over the space diverges. The interaction simply does not decrease sufficiently quickly as the distance
grows. Nevertheless, you may study the Coulomb potential as the limit a of the Yukawa
potential with
q1 q2
A=
40
When we send a in the Yukawa results and rewrite the results using E = h
2 k 2 /2m, we obtain
d
q1 q2
=
d
40


2

1
16E 2 sin4 (/2)

Observations:
it turns out that this result for d/d is actually exact; the phase of f () in the exact result
is however slightly different than you would naively expect
the cross section does not depend on h
and therefore it looks classical; in fact, it is also the
exact classical result; please realize that such exact agreements between classical physics and
quantum physics are always special, rare, and a matter of good luck
at small , we have d/d 1/4 . The total cross section is therefore proportional to d/3
and diverges. The Coulomb field extends everywhere and for any impact parameter, at least
some interaction will occur. (This is shockingly not true in quantum mechanics for the Yukawa
potential and similar potentials: there is a large probability that the particle does not interact
at all even though the potential is strictly speaking nonzero also everywhere.) We call the
Coulomb potential a long-range field while the Yukawa potential is a short-range field.
The definition is according to whether or not the total cross section diverges.
R

The Born approximation is morally related to the Feynman diagrams in quantum field theory.
The individual vertices of the Feynman diagram are analogous to the potential energy, and the
more vertices you include in your Feynman diagram, the more tiny and accurate contributions to
the total cross section you calculate.

Scattering of identical particles

So far we considered a particle scattering off a fixed potential. This is a good approximation if the
source of the potential is an infinitely heavy particle or object. Often we need to study scattering of
two comparable particles whose masses are m1 , m2 . In that case, the dynamics of these two particles
can be reduced to the motion of the center of mass which behaves as a free particle of mass m1 +m2
and the nontrivial relative problem with the potential V (~r) V (~r1 ~r2 ) and the reduced mass m
where
1
1
1
=
+
m
m1 m2
Recall that we have checked the kinetic energy operators agree. Whenever this approach with the
reduced mass is used for scattering, we always use the scattering angle for the angle as measured
in the center-of-mass frame.
Also, you should recall the beginning of the course when we discussed identical particles. The
full scattering amplitude f () which is what we were secretly talking about must be the
(anti)symmetrized combination
f () = f () f ( )
5

where the sign + is chosen for bosons and for fermions. Note that there is no factor of 1/2 in this
formula. When you want to determine the cross section, you just square the total amplitude which
is obtained from the two pieces:
d
= |f () f ( )|2
d
Its interesting to look at = /2. You will get
d
= 0 for fermions and/or 4|f ()|2
d

for bosons

For distinguishable particles, you would get

d
= 2|f ()|2
d
because you would first have to compute d/d from a single f , and then you would have to sum the
two terms in the cross section. Once again, for the identical particles, you first sum up or subtract
the two pieces in the scattering amplitude, and then you square the total scattering amplitude to
get the cross section. For identical particles, interference (and cancellations, such as for = /2 in
the fermionic case above) may occur. For distinguishable particles,
it cannot.
Note that above, we did not include anything like 1/2 or 1/ 2 in the sum f f , and you may
therefore think that we overestimate the cross section. But you should not forget that because the
particles are equal, we will not distinguish 12 from 21 as the final state, and therefore if we
calculate the total cross section, it will only be
=

d cos

d
.
d

Note that the lower limit for cos is chosen to be zero (the equator) instead of 1 (the South pole).
If we changed 0 to 1 in the formula above, we would be double counting the cross section.
One more comment. When you calculate the f f combinations above, it is often useful to use
the partial wave analysis. One half of the partial waves will then cancel. Which ones? Recall a
property of the Legendre polynomials:
Pl (cos ) = (1)l Pl [cos( )]

fl () = (1)l fl ( )

You see that the bosonic scattering amplitude cancels for odd l while the fermionic scattering
amplitude cancels for even l.