kr
. (3)
As with onedimensional scattering, we do not need to worry about the normalization of the incident state.
Furthermore, the potential V (
)V (r
)
0
(r
) +
_
dV
dV
G
0
(r, r
)V (r
)G
0
(r
, r
)V (r
)
0
(r
) +. . . , (17)
where we have dened
0
(r) = r[
0
) and G
0
(r, r
) = r[G
0
[r
to the
detector. The integral over all r
then sums over all possible collision locations. The next terms describes
the particle scattering twice, summing over both collision locations. The third term would include all
possible paths with three collisions, and soon. Thus we see that the total amplitude is the sum over
all possible trajectories by which the particle could have made it to the detector, assuming straightline
propagation between pointcontact collisions. From this interpretation, we can guess that the Greens
function would have the form:
G
0
(r, r
) =
e
ikrr

[r r[
, (18)
as the e
ikrr

factor just adjusts the phase of the state to reect propagation with wavelength = 2/k,
over a distance d = [r r
n
[z
n
)z
n
[
1 z
n
. (20)
Note that the series expansion (1 z)
1
only converges for [z[ < 1. This is because a series expansion
converges only as far from the expansion point as the nearest singularity in the function being expanded.
Here the expansion point is z = 0 and the singularity is at z = 1. We can perform a valid series expansion
of (1 G
0
V )
1
via
(1 G
0
V )
1
=
n
u(1 [z
n
[) [z
n
)z
n
[
_
1 + z
n
+ z
2
n
+ . . .
_
+
n
u([z
n
[ 1)
[z
n
)z
n
[
1 z
n
, (21)
where u(x) is the unit step function. For the case [z
n
[ > 1, we can expand in powers of 1/z
n
, as
1
1 z
n
=
1
z
n
1
1
1
zn
=
1
z
n
_
1 + z
1
n
+ z
2
n
+ . . .
_
. (22)
These two series can be combined to give
(1 G
0
V )
1
=
n
[z
n
)z
n
[
m=0
_
z
m
n
+ u([z
n
[ 1)
_
z
m
n
+
_
z
m1
n
__
=
m=0
n
[z
n
)z
n
[
_
z
m
n
u([z
n
[ 1)
_
z
m
n
+ z
m1
n
_
=
m=0
[(G
0
V )
m
R
m
] (23)
where the m
th
renormalization term is
R
m
=
n
u([z
n
[ 1) [z
n
)z
n
[
_
z
m
n
+ z
m1
n
_
. (24)
The idea is then that the renormalized series will converge normally, so you can take only as many terms
as required for precision. Of course this renormalization is dicult in practice because the eigenvalues and
eigenvectors of G
0
V are not usually known, but it establishes the proofofprinciple of the renormalized
Born series.
4
1.4 The TMatrix
From the Bornseries, we see that the scattered wave [
s
) = [) [
0
) is given by
[
s
) = (G
0
V + G
0
V G
0
+ G
0
V G
0
V G
0
V + . . .)[
0
). (25)
Based on our pathintegral interpretation, we see that each term contains at least one scattering event.
For each term, the scattered wave then propagated freely from the last scattering point to the detector.
This nal freepropagation can be factored out giving
[
s
) = G
0
(V + V G
0
V + V G
0
V G
0
V + . . .)[
0
). (26)
Then the sum inside parentheses is just the story of all the possible ways the particle could have made it to
the location of the nal scattering event. If we put all of this story in a black box, and call it a Tmatrix,
we get
[
s
) = G
0
T[
0
), (27)
which denes the Tmatrix. Comparison with (9) and (14) shows that
[
s
) = G
0
V (1 G
0
V )
1
[
0
) (28)
from which we see immediately that
T = V (1 G
0
V )
1
, (29)
or equivalently
T = (1 V G
0
)
1
V. (30)
This equivalence can be proven by hitting both equations from the left with (1 V G
0
) and from the right
with (1 G
0
V ), which gives V + V G
0
V = V + V G
0
V .
Ignoring the issue of convergence for a moment, the series expansion (1A)
1
= 1+A+A
2
+. . ., gives
T = V + V G
0
V + V G
0
V G
0
V + V G
0
V G
0
V G
0
V + . . . . (31)
Projecting this onto position eigenstates results in the positionspace matrix elements of the Tmatrix:
T(r, r
) = V (r)(r r
) + V (r )G(r , r
)V (r
) +
_
dV
V (r )G(r , r
)V (r
)G(r
, r
)V (r
) + . . . , (32)
where T(r , r
) = r [T[r
) and G
0
(r , r
) = r [G
0
[r
).
1.5 Computing the Greens Function via the Calculus of Residues
Because G
0
= (E H
0
+ i), it follows that the eigenstates of G
0
are the eigenstates of H
0
. Thus, the
Greens function can be expressed in the basis of eigenstates of H
0
, giving
G
0
=
m
_
Ec
dE
[E
, m)E
, m[
E E
+ i
+
n,m
[n, m)n, m[
E E
n
+ i
, (33)
where E
c
is the continuum threshold energy, the summation over m accounts for any degeneracies, and
the summation over n includes any boundstates which lie below the continuum threshold, i.e. E
n
< E
c
.
Simplifying to the case with no degeneracies, no boundstates, and E
c
= 0 gives
G
0
=
_
0
dE
[E
)E
[
(E E
+ i)
. (34)
5
We can project onto position eigenstates to give
r[G
0
[r
) =
_
0
dE
r[E
)R
[r
)
E E
+ i
, (35)
or equivalently
G(r, r
)
_
0
dE
(r)
E
(r
)
E E
+ i
, (36)
which, if the energy eigenstate wavefunctions are known, can often be solved by changing integration
variables from energy to wavevector, and then using contour integration to obtain an analytic expression.
1.5.1 Free particle in 1dimension
For a free particle in onedimension we have E =
2
k
2
2M
, and x[E) = (2)
1/2
e
ikx
. With these substitutions,
equation (36) becomes
G(x, x
) =
1
2
_
dk
e
ik
(xx
)
E
2
k
2
2M
+ i
=
2M
2
2
_
dk
e
ik
(xx
)
k
2
k
2
i
. (37)
With k
2
k
2
i = (k
k
2
+ i)(k
k
2
+ i) and
k
2
+ i = k +
i
2k
= k + i, this becomes
1
G(x, x
) =
2M
2
2
_
dk
e
ik
(xx
)
(k
+ k + i)(k
k i)
. (38)
This can be solved by the method of contour integration, which means we extend k
)
e
ik
R
(xx
)k
I
(xx
)
. Because we want the function to vanish
at [k[ , we see that for x > x
) =
_
2M
2
2
2i
e
ik(xx
)
2k
; x > x
2M
2
2
(2i)
e
ik(xx
)
2k
; x < x
. (39)
These results can be combined into a single expression:
x[G
0
[x
) =
iM
2
k
e
ikxx

. (40)
This clearly shows that the current is owing in the direction from x
k)
k[
E
2
k
2
2M
+ i
. (41)
Projecting onto position eigenstates, and factorizing the denominator gives
r[G
0
[r
) =
2M
(2)
3
2
_
d
3
k
e
i
(rr
)
(k
+ k + i)(k
k i)
, (42)
where k =
gives
r[G
0
[r
) =
2M
(2)
2
2
_
1
1
du
_
0
k
2
dk
e
ik
rr
u
(k
+ k + i)(k
k i)
(43)
Performing the uintegration gives
r[G
0
[r
) =
2M
(2)
2
2
_
0
k
2
dk
e
ik
rr

e
ik
rr

(k
+ k + i)(k
k i)ik
[r r
[
(44)
The twoterms in the integrand can then be combined to extend the integral to , so that
r[G
0
[r
) =
2M
(2)
2
2
_
k
2
dk
e
ik
rr

(k
+ k + i)(k
k i)ik
[r r
[
(45)
Since [r r
[ > 0, we can close on the upper halfplane, so that the Residue theorem gives
r[G
0
[r
) =
2M
(2)
2
2
(2i)k
2
e
ikrr

2ik
2
[r r
[
. (46)
Thus the nal expression for the 3d Greens function becomes
G([r r
[) =
M
2
2
e
ikrr

[r r
[
, (47)
which agrees with our previous guess (18).
1.6 Example: 1d deltafunction scattering
1.6.1 The Tmatrix of a 1d deltafunction
Consider a free particle in one dimension incident on a deltafunction potential V (x) = g(x). We want
to solve the scattering problem and nd the reection and transmission probabilities T and R. We rst
compute the Tmatrix via Eqs. (32) and (40), giving
T(x, x
) = g(x)(x
) + g(x)
_
iM
2
k
_
e
ikxx

g(x
)
+ g(x)
_
dx
iM
2
k
_
e
ikxx

g(x
)
_
iM
2
k
_
e
ikx

g(x
) + . . .
= g(x)(x
)
_
1 +
_
i
gM
2
k
_
+
_
igM
2
k
_
2
+ . . .
_
=
g(x)(x
)
1 + i
gM
2
k
(48)
7
We can now compute the scattered wave via Eq. (27). Taking
0
(x) = e
ikx
then gives
s
(x) =
_
dx
dx
G(x, x
)T(x
, x
)
0
(x
)
=
_
dx
dx
iM
2
k
_
e
ikxx

g(
x)(x
)
1 + i
gM
2
k
e
ikx
=
e
ikx
1 i
2
k
Mg
. (49)
Introducing the scattering length
a =
2
Mg
, (50)
we arrive at the full solution
(x) = e
ikx
e
ikx
1 ika
. (51)
For x < 0 this gives
(x) = e
ikx
e
ikx
1 ika
, (52)
from which we can identify the reection amplitude
r =
1
1 ika
. (53)
The reection probability, R = [r[
2
is therefore
R = [r[
2
=
1
1 + (ka)
2
. (54)
Turning o the potential requires g 0, corresponding to a , in which case we have R = 0 as
expected. For x > 0 we nd
(x) = e
ikx
e
ikx
1 ika
=
ika
1 ika
e
ikx
, (55)
from which we can identify the transmission amplitude as
t =
ika
1 ika
. (56)
This leads to a transmission probability of
T = [t[
2
=
(ka)
2
1 + (ka)
2
, (57)
which goes to 1 as g 0, as required. We note also that R+T = 1, satisfying conservation of probability.
1.6.2 Direct solution of the LippmanSchwinger equation
The previous example was to illustrate the Tmatrix formalism. For 1d scatters made up entirely of
deltafunctions, by far the easiest approach is to solve the LippmanSchwinger equation (LSE) directly.
Hitting the LSE from the left with x[ gives
x[) = x[
0
) +xG
0
V [). (58)
8
Inserting the projector between the G
0
and the V , and making use of the diagonality of V leads to
(x) =
0
(x) +
_
dx
G
0
(x, x
)V (x
)(x
), (59)
which is an integral equation for (x). For the case where V (x) = g(xx
0
), we can perform the integral,
giving
(x) =
0
(x) + gG
0
(x, x
0
)(x
0
). (60)
The unknown quantity, (x
0
) can be found by setting x = x
0
,
(x
0
) =
0
(x
0
) + gG
0
(x
0
, x
0
)(x
0
), (61)
which has the solution
(x
0
) =
0
(x
0
)
1 gG
0
(x
0
, x
0
)
, (62)
which gives the solution to the LSE as
(x) =
0
(x) +
gG
0
(x, x
0
)
0
(x
0
)
1 gG(x
0
, x
0
)
. (63)
With
0
(x) = e
ikx
and G
0
(x, x
) = i
M
2
k
e
ikxx

, this becomes
(x) = e
ikx
i
Mg
2
k
e
ikx
1 + i
Mg
2
k
= e
ikx
e
ikx
1 ika
, (64)
where we have again introduced a =
2
Mg
.
9
2 Scattering probabiilities
2.1 The scattering amplitude
In general, we want to nd the probability to detect the particle at position r after it has left the scattering
region. If we choose the center of the scattering potential as the origin, and if the detector is suciently
far from the scatterer, then we can compute the probability amplitude at the detector from the larger
limit of the full wavefunction,
lim
r
(r) =
0
(r) + lim
r
_
d
3
r
d
3
r
G(r, r
)T(r
, r
)
0
(r
). (65)
Taking a plane wave for the incident state,
0
(r) = e
i
kr
, (66)
and putting in the 3d Greens function (47), we nd
lim
r
(r) = e
i
kr
M
2
2
lim
r
_
d
3
r
d
3
r
e
ikrr

[r r
[
T(r
, r
)e
i
kr
. (67)
Now with r = r e
r
(, ) = r(cos e
z
z + sin cos e
x
+ sin sine
y
) we nd
lim
r
[r r
[ =
_
(r r
)
2
=
_
r
2
2re
r
r
+ r
2
r
_
1 2
e
r
r
r
+
r
2
r
2
r
_
1
e
r
r
r
_
r e
r
r
, (68)
For convenience we choose e
z
along
k, so that = 0 corresponds to the forwardscattering direction.
Similarly we nd
lim
r
1
[r r
[
1
r
. (69)
This gives
lim
r
G(r, r
) =
M
2
2
e
ikr
r
e
iker(,)r
, (70)
which leads to
lim
r
(r) = e
ikz
M
2
2
e
ikr
r
_
d
3
r
d
3
r
e
iker(,)r
T(r
, r
)e
ikz
. (71)
It is conventional to dene the scattering amplitude, f(, [k), via
lim
r
(r) = e
ikz
+ f(, [k)
e
ikr
r
, (72)
so we see immediately that
f(, [k) =
M
2
2
_
d
3
r
d
3
r
e
iker(,)r
T(r
, r
) e
ikezr
. (73)
10
With the substitutions
k e
r
(, )
, (74)
and
k e
z
k, (75)
the scattering amplitude generalizes to
f(
k) =
M
2
2
_
d
3
r
d
3
r e
i
T(r
, r)e
i
kr
=
(2)
2
M
2
_
d
3
r
d
3
r
[r
)r
[T[r)r[
k)
=
(2)
2
M
[T[
k), (76)
which should be interpreted as the probability amplitude to scatter in the direction
k
, given an incident
wavevector
k. The function f(, [k) is valid only for the case of elastic scattering, [
[ = [
k), where
r[
k) = e
i
kr
. (77)
This state is deltanormalized so that
k[
) = (2)
3
3
(
k)[
2
is clearly dimensionless.
One possible way to give a probabilistic interpretation to [
0
(r)[
2
would be to assume a nite quantization
volume V , so that the probability density would be dened as (r) = [
0
(r)[
2 1
V
, so that
_
V
d
3
r(r) = 1.
A plane wave corresponds to a uniform probability ow at speed v = k/M, so that the current
density is
j(r) = (r)v(r), so that for a plane wave, [j[ = v/V . The probability current through a surface
(perpendicular to
k) of area A
0
is then given by
J = jA
0
=
A
0
V
k
M
J
in
. (78)
In dening this as the incident current, we should interpret A
0
as the crosssectional area of the incident
particle beam. The innitesimal scattered current through an innitesimal area element dA at distance
11
r
0
in the , direction is then given by
dJ(r
0
, , ) =
[
s
(r, , )[
2
V
v dA =
[f(, [k)[
2
r
2
0
1
V
k
m
dA. (79)
With the surface area element begin given by dA = r
2
0
d = r
2
sin dd, this simplies to
dJ
S
(, ) =
[f(, [k)[
2
V
k
M
d (80)
The scattered probability current into the solid angle region
0
would then be given by
J
S
(
0
) =
_
0
dJ(, ) =
k
MV
_
0
d[f(, [k)[
2
, (81)
which shows that the scattering probabilities are independent of the choice of quantization volume, V . In
analogy with the way we dened reection and transmission probabilities in 1d, the probability to scatter
into solid angle
0
would be the ratio J
S
(
0
)/J
in
, giving
P
S
(
0
) =
1
A
0
_
0
d[f(, [k)[
2
. (82)
That the scattering probability would decrease as the incident beam area increases has the classical
interpretation that with a wider beam, one would be more likely to miss a target with a xed cross
sectional area , assuming A
0
. This leads to the total scattering probability
P
S
=
1
A
0
_
d[f(, [k)[
2
, (83)
where the integration is over the entire 4 solid angle. We note that only [
s
) contributes to the scatter.
This is done deliberately to account for the nite crosssectional area of the incident beam. For pure plane
wave, there would be an interference between [
0
) and [
s
) at every point in space, but in reality, this
interference only occurs inside the beam volume, whereas P
S
calculated above applies only to a detector
located outside of the beam volume.
We can dene the eective crosssectional area of the scatterer based on an analogy with classical
scattering of particles from a simple solidobject. Consider a target of crosssectional area , which we are
trying to hit with particles that travel in a straight line along e
z
, but whose transverse position is random
within the crosssectional area A
0
. The probability to hit the target is therefore the ratio of the two areas
P
hit
= /A
0
. This leads us to dene the eective crosssection of a generalized scatterer via
tot
= P
S
A
0
. (84)
For quantummechanical scattering, this leads to
tot
=
_
d[f(, [k)[
2
. (85)
Likewise we can dene the fraction of the eective crosssection due to scattering into a solid angle
0
as
(
0
) = P
S
(
0
) A
0
=
_
0
d[f(, [k)[
2
. (86)
Hence the innitesimal element of the crosssectional area due to scattering in the direction indicated by
, must be
d(, ) = d[f(, [k)[
2
. (87)
12
Based on this it is common to dene the dierential crosssection, denoted by
, according to
d(, )
d
= [f(, [k)[
2
. (88)
Thus the dierential crosssection is the square modulus of the scattering amplitude. It may be helpful to
think of
d
d
as simply a symbol, rather than as some sort of derivative.
A given detector located at angular position
d
,
d
, can be characterized by its solidangle of acceptance
d
, as well as its detection eciency c (i.e. the fraction of particles entering the detector which are actually
detected), and its darkcount rate T (i.e. the probability that the detector red but no particle actually
entered the detector). If a particle is incident on the scatterer with momentum
tot
A
0
+T
=
c T
A
0
_
0
d
d(, )
d
+T, (89)
. If the solid angle subtended by the detector is suciently small, than we can pull the dierential cross
section outside of the integral, and evaluate it at the angular location of the detector, giving
P
detect
= (c T)
d
[f(
d
,
d
[k)[
2
A
0
+T. (90)
2.3 Example: The Yukawa potential in the rst BornApproximation
The Yukawa potential is given by
V (r) =
V
0
r
e
r
. (91)
It is essentially a Coulomb potential with an exponential dropo as r . In the limit 0 and
V
0
0 with V
0
/ held xed, we recover the Coulomb potential. Consider a metallic conductor, on a
coarse scale it is electrically neutral, since there are an equal number of electrons and protons. At short
ranges the conduction electrons would see each other, as well as neighboring crystal ions, but at long range
it really sees nothing, due to the overall neutrality. Thus if this range were known, the the electronelectron
and electronion interactions could be replaced by Yukawa potentials, with the eective screening length,
without changing the physics. The Yukawa potential has the advantage, relative to the Coulomb potential,
that certain important classes of integrals then converge to nite numbers. It is often assumed that the
pure Coulomb physics can be then obtained simply by taking the 0 limit only at the very end of the
calculation, i.e. after the integrals have converged to nite values.
In this example we will solve the problem of scattering from the Yukawa potential approximately, by
keeping only the rstorder term in the Born series. This means we replace T = V + V GV + V GV GV +
V GV GV GV + . . . with T V . This leads to
T(r
, r
) V (r
)
3
(r
). (92)
From Eq. (76) we see that this leads to
f(
k) =
M
2
2
_
d
3
r
e
i(
)r
V (r
). (93)
13
If we choose e
z
to lie along the
k
k) =
MV
0
2
2
_
2
0
d
_
0
d
sin
_
0
dr
r
2
e
i
r
cos
e
r
=
MV
0
_
1
1
du
_
0
dr
e
i
ur
(94)
Using
_
1
1
due
iau
=
2 sin a
a
(95)
gives
f(
k) =
2MV
0
2
[
[
_
0
dr
sin([
[r
)
r
e
r
=
2MV
0
2
+[
[
2
. (96)
Since [
[ = [
k[ = k, we can write
[
[
2
= 2k
2
2k
2
cos
= 2k
2
(1 cos ), (97)
which gives
f(, [k) =
2MV
0
2
+ 2k
2
(1 cos )
. (98)
This leads to the dierential crosssection
d
d
=
_
2MV
0
_
2
1
[
2
+ 2k
2
(1 cos )[
2
, (99)
which describes the angular distribution of the scattered probability.
If we let 0 and V
0
0, with
V
0
=
ZZ
e
2
4
0
(100)
we recover the Coulomb potential. This leads to the partial scattering amplitude
d
d
=
_
2MZZ
e
2
4
0
2
_
2
1
4k
4
(1 cos )
2
, (101)
which famously recovers the classical Rutherford scattering. That the rstBorn approximation would give
the classical result is not surprising, since the rstBorn result is valid in the highenergy limit, for which
classical and quantum results typically agree.
Returning the the Yukawa potential, we can compute the total crosssection via
tot
= 2
_
2MV
0
_
2
_
1
1
du
1
(
2
+ 2k
2
2k
2
u)
2
. (102)
14
With y =
2
+ 2k
2
2k
2
u, this becomes
tot
= 2
_
2MV
0
_
2
_
2
+4k
2
2
dy
1
y
2
=
_
2MV
0
_
2
4
2
(
2
+ 4k
2
)
. (103)
For Coulomb scattering this gives
tot
= . This means all incident probability current will be scattered,
no matter how large A
0
is. For this reason we say that the Coulomb interaction is an innite range
interaction. For nite , on the other hand, the Yukawa potential is eectively a nite range interaction.
15
3 Conservation of angular momentum
3.1 Scattering from spherically symmetric potentials
We now focus on the case where the scattering potential V (r) is spherically symmetric. This is typical of
twobody collisions in the absence of external elds. As we know, the main consequence of this symmetry
is that angular momentum will be conserved. Quantum mechanically, this means that L
2
and L
z
will be
constants of motion. It is most convenient to chose the zaxis along along the incident wave propagation
direction, so that
k = ke
z
With this choice, the incident wave is azimuthally symmetric, and thus contains
only m = 0 components. Due to the spherical symmetry, no m ,= 0 states can be created by the scatter,
so that the complete scattering problem can be treated in the m = 0 subspace, which will simplify our
calculations somewhat.
Up to now we have considered only the case where the incident (unperturbed) state is a plane wave,
which is a state with welldened kinetic energy. Because we are going to expand this state onto angular
momentum eigenstates, it is interesting to consider rst the angular momentum of a classical particle
moving with constant velocity along the zdirection. The classical angular momentum is
L = r p, for the
case p = ke
z
, we nd
L = det
e
x
e
y
e
z
x y z
0 0 k
= k(ye
x
xe
y
). (104)
Thus we see that classical the classical angular momentum vanishes only if x = y = 0, and otherwise, it
can have arbitrarily large [
k), dened by
P[
k) = k[
(r[k) and , [, m) = Y
m
(, ). Here
the Y
m
(, ) are the usual spherical harmonics. From the theory of central potentials, we know that the
radial eigenfunctions, R
(0)
2
k
2
2M
+
2
2M
1
r
2
r
r
2
r
2
( + 1)
2Mr
2
_
R
(0)
(r[k) = 0. (109)
The solutions to this equation are wellknown special functions called spherical Bessel functions of the rst
kind, denoted by j
(kr). Thus the freeparticle angular momentum eigenstates, called partial waves are
given by
r, , [k, , m
(0)
) =
_
2k
2
(kr)Y
m
(, ), (110)
which are normalized as
_
d
3
r
_
k, , m
(0)
[r, , )r, , [k
, m
(0)
)
_
= (k k
)
,
m,m
. (111)
The projector onto the basis of partial waves is then
I =
=0
m=
_
0
dk [k, , m
(0)
)k, , m
(0)
[. (112)
3.2.1 Threshold behavior
It is interesting to examine the behavior of these functions near the origin, i.e. inside the (nite) scattering
region. Their behavior as r 0 is given by
lim
r0
j
(kr) =
(kr)
(2 + 1)!
. (113)
This means that as k 0, all the partial waves go rapidly to zero except for the = 0 wave. For each ,
there is an energy scale below which the
th
partial wave is eectively zero inside the scattering region. This
means that the
th
component of the incident wave no longer sees the scatterer, and so its contrubution
to the scattering crosssection is eectively zero. This eect is known as Threshold behavior. This leads
to the well known result that for most potentials, there will be a critical kvalue, below which only Swave
scattering makes a signicant contribution to the scattering amplitude, known as the Swave regime.
17
3.3 The scattering phaseshift
The spherical Bessel functions, j
(kr), in addition to being solutions of the free particle radial wave equation,
also satisfy the freeparticle boundary condition at r = 0, which is derivable from the requirement that
(r) should be everywhere continuous and smooth. If this boundary condition is relaxed, then there is
a second linearly independent solutions, called the spherical Bessel function of the second kind, denoted
as y
(kr), that are singular at r = 0. These spherical Bessel functions can be dened by the Rayleigh
formulas,
j
() =
d
d
_
sin
, (114)
y
() =
d
d
_
cos
. (115)
If we wish, instead, to nd solutions with a welldened direction of probability ow (inwards or outwards),
we can use spherical Bessel functions of the third kind. The rst of these two spherical Bessel functions is
dened by
h
() = j
() + iy
(), (116)
while the second linearly independent solutions is simply h
() = j
() iy
(kr) is
h
() = i
d
d
_
e
i
. (117)
From the sign of the exponent, we see that h
()
has a purely incoming current. Thus h
(kr) and h
(kr)
and y
(kr) =
1
2
[h
(kr) + h
(kr)] , (118)
and
y
(kr) =
1
2i
[h
(kr) h
(kr)] . (119)
In the limit r , the spherical Bessel functions have the asymptotic forms,
lim
r
j
(kr) =
sin(kr /2)
kr
, (120)
lim
r
y
(kr) =
cos(kr /2)
kr
, (121)
lim
r
h
(kr) = ie
i(kr/2)
. (122)
If we like we can consider the case where, instead of a plane wave for [
0
), we have a pure incoming
partial wave, having well dened energy, E =
2
k
2
2M
, and angular momentum, so that
0
(r) = h
(kr)Y
(). If
we then let the ket [k, ) represent the full scattering solution, conservation of angular momentum requires
it to have the limiting form
lim
r
r, [k, ) = a
(kr) + b
(k)h
(kr), (123)
which says that the incoming and outgoing currents have the same angular momentum, . The incoming
probability current will be proportional to [a
[
2
. Conservation of probability then requires that the incoming
18
and outgoing current be equal, i.e. [b
[ = [a
[. Thus a
and b
i
_
h
(kr) + e
i2
(k)
h
(kr)
_
. (124)
where we have introduced the partialwave scattering phaseshift,
(kr), so that
(r), via (122), we nd that the limiting form of the full eigenfunction,
including the eects of the scatterer, must then be
lim
r
(r, ) = a
_
e
i(kr/2)
kr
e
i2
(k)
e
i(kr/2)
kr
_
Y
(), (125)
or equivalently,
lim
r
(r, ) = 2ia
e
i
(k)
sin(kr /2 +
(k))
kr
, (126)
The complete information about a scatterer, at least to the extent that can be obtained
by measurements at r , is therefore contained in the set of partialwave phaseshifts,
0
(k),
1
(k),
2
(k), . . ..
19
3.4 Relationship between the scattering amplitude and the scattering phaseshifts
We now return to our initial assumption of an incident planewave,
0
(r) = e
ikz
, rather than an incoming
partial wave. The incident plane wave is given in spherical coordinates by
r, , [
0
) = e
ikr cos
. (127)
Because there is no dependence, its expansion onto partial waves can contain only azimuthallysymmetric
m = 0 states. It must still, however, be a superposition of dierent states, and due to angular momentum
conservation, we know that a spherically symmetrjic scatterer will not mix states with dierent . It is not
necessary to work through the details of how to compute the expansion coecients, rather we just need
the result
0
(r) = e
ikr cos
=
r, , [k, , 0
(0)
) k, , 0[
0
)
=
=0
i
_
4(2 + 1)j
(kr)Y
0
()
=
=0
i
2
_
4(2 + 1) (h
(kr) + h
(kr)) , (128)
which we see consists of equal contributions from incoming, h
R) = V (R), the solutions still have eigenvalues k, and , hence we will again assign them
the ket [k). These eigenstates decompose into radial and angular parts, with the angular part given by a
spherical harmonic, r, , [k, , m) = R
(r[k)Y
m
2
k
2
2M
+
2
2M
1
r
2
r
r
2
r
2
( + 1)
2Mr
2
+ V (r)
_
R
(r[k) = 0. (130)
From equation (124), we know already that, as a result of conservation laws, the solution must take the
limiting form
lim
r
R
(r[k) =
a
(k)
i
_
h
(kr) + e
i2
(k)
h
(kr)
_
= a
_
e
i(kr/2)
kr
e
i2
(k)
e
i(kr/2)
kr
_
In the sphericallysymmetric case, the scattering amplitude, f(, [k), depends only on and k, and
was dened via
lim
r
(r) = e
i
kr
+ f([k)
e
ikr
r
. (131)
20
We can expand the incident planewave onto partial waves, giving
lim
r
(r) =
=0
e
i/2
2
_
4(2 + 1) (h
(kr) + h
(kr)) Y
0
() + f([k)
e
ikr
r
. (132)
We can also expand the scattering amplitude onto spherical harmonics as
f([k) =
=0
c
(k)Y
0
() =
=0
_
4(2 + 1)f
(k)Y
(), (133)
where f
(k) is the partial wave scattering amplitude. In terms of f([k), the partial amplitudes are given
by
f
(k) =
1
_
4(2 + 1)
_
dY
()f([k). (134)
This denition leads to the result
lim
r
(r, ) =
=0
_
4(2 + 1)
_
e
i/2
2
(h
(kr) + h
(kr)) + f
(k)
e
ikr
r
_
Y
() (135)
Replacing h
=0
_
4(2 + 1)
i
+1
2
_
e
i(kr/2)
kr
e
i(kr/2)
kr
2ikf
(k)
e
i(kr/2)
r
_
Y
0
()
=
=0
_
4(2 + 1)
(i)
2
_
e
i(kr/2)
kr
(1 + 2ikf
(k))
e
i(kr/2)
r
_
Y
0
(). (136)
By comparison with (125) we then see that the relationship between the partial amplitude and the partial
phaseshift is
1 + 2ikf
(k) = e
i2
(k)
. (137)
Inverting this formula gives
f
(k) =
e
i
(k)
sin(
(k))
k
. (138)
Plugging this into (133) then gives the scattering amplitude in terms of the partial phaseshifts,
f([k) =
=0
_
4(2 + 1)
e
i
(k)
sin(
(k))
k
Y
(). (139)
21
3.5 The partialwave scattering crosssection
For spherically symmetric potentials, the total scattering crosssection (85) becomes
tot
= 2
_
0
d sin [f([k)[
2
. (140)
Based on the result (139), this can be written in terms of the partial phaseshifts as
tot
=
4
k
2
=0
=0
_
(2 + 1)(2
+ 1) sin(
(k)) sin(
(k))e
i(
)
_
d sin Y
()Y
(),
=
4
k
2
=0
(2 + 1) sin
2
(
(k)),
where we have made use of the orthogonality relation for the spherical harmonics. From this expression,
we can identify the
th
partial crosssection as
= 4(2 + 1)
sin
2
(
(k))
k
2
= 4(2 + 1)[f
(k)[
2
.
The partial crosssection,
tot
=
4
k
Imf( = 0[k). (141)
This is useful because it allows us to calculate the total crosssection, which contains contributions from
scattering into all possible angles, by only evaluating the scattering amplitude at = 0, which indicates
the forwardscattering direction. Using the results we have derived so far, the proof of the optical theorem
is straightforward. From Eq. (139) we nd
f( = 0[k) =
=0
2 + 1 (cos
(k) + i sin
(k))
sin
(k)
k
Y
0
(0). (142)
Since
Y
0
(0) =
2 + 1
4
, (143)
we see that
f( = 0[k) =
(2 + 1) (cos
(k) + i sin
(k))
sin
(k)
k
, (144)
so that
Imf( = 0[k) =
1
k
(2 + 1) sin
2
(k). (145)
Comparing this with (141) we see that
tot
=
4
k
Imf( = 0[k). (146)
The reason this works, is because the probability to scatter is just the incident probability current minus
the forwardscattering probability current, divided by the incident current. Since the incident current is
presumably known, it is no surprise that the total cross section is directly related to the forwardscattering
amplitude.
23
3.7 The swave scattering length
The swave scattering length iand eective range are dened by the equation
k cot(
0
(k)) =
1
a
+
1
2
r
e
k
2
+ . . . . (147)
If we assume that
0
(k) is a small angle, then this can be approximated by
0
(k) = ak
_
1 +
1
2
ar
e
k
2
+ . . .
_
. (148)
In the case where r
e
k 1/(ak) we can make the approximation
0
(k) ak, which leads to f
0
(k) e
i2ka
for the case ak 1 this can be approximated to second order by
(1 + 2ikf
0
(k)) = e
i2
0
(k)
=
e
i
0
(k)
e
i
0
(k)
1 iak
1 + iak
+ O((ak)
3
). (149)
For the zerorange pseudopotential we found in HW7.5, that T(r, r
) =
g
3
(r)
3
(r
)
1+i
Mg
2
2
k
r
r. Applying Eq. (76)
and taking g = 2
2
a/M gives for the pseudopotential,
(1 + 2ikf
0
(k)) =
1 iak
1 + iak
. (150)
Thus, under the condition that k is small enough to satisfy ak 1 and r
e
k 1/(ak), we can replace the
true potential with the pseudopotential
V (r) =
2
2
a
M
3
(r)
r
r, (151)
and compute the scattering physics correctly to second order in k. In many cases, particularly with cold
atoms, this allows us to nd analytic solutions to manybody problems which agree quantitatively with
experiment. This shows that a is the parameter which governs low energy scattering, while r
e
is the
parameter which tells when the energy is low enough to be governed only by a.
24
3.8 Identical particles and scattering
If the spinstate of a pair of identical particles is known, then the symmetry of the wavefunction is fully
determined by the requirement that Bosons have totally symmetric states and Fermions have totally
antisymmetric states under exchange of particle labels. For a pair of Fermions with a symmetric (triplet
for spin1/2) spin state require an antisymmetric wavefunction (r) = (r) where r = r
1
r
2
is the
relative coordinate. For a a pair of Fermions an in antisymmetric (singlet for spin1/2) state, a symmetric
wavefunction is required. The situation is reversed for Bosons. The primary result from partial wave
analysis is that the exchange transformation is and ., so that the exchange properties
of the partial waves are governed strictly by the properties of the spherical harmonics Y
0
0
(r) =
e
ikr
kr
e
i2
0
(k)
e
ikr
kr
. (153)
Since this is an exact eigenstate of H
0
, we should take it as the state for all r > R. Using the usual
denition R
(r[k) =
u(r)
r
, and dropping any overall constant factors, we let
u
I
(r) = e
ikr
e
i2
0
(k)
e
ikr
, (154)
for r > R, which we can call region I.
3.10 Anzatz for innermost region
For region II, r < R, we also need a freespace swave state, but with k K =
_
k
2
2MV
0
/
2
. Thus
we start with two arbitrary constants, by taking
u
II
(r) = ae
ikr
be
ikr
. (155)
The point r = 0 needs to be handled carefully. The primary requirement is the (r) must be smooth and
continuous at r = 0. If we expand u in this region as
u(r) = u
0
+ u
1
r + u
2
r
2
+ . . . , (156)
then the requirement that (r) =
u(r)
r
be continuous is simply u
0
= 0, so that there is no 1/r singularity.
The condition of smoothness is that
r
(r) = 0, which gives
r
u(r)
r
r=0
=
_
u
(r)
r
u(r)
r
2
_
r=0
= 0. (157)
Plugging in the series expansion gives
(u
1
r
1
+ 2u
2
+ 3u
3
r + . . .)
r=0
(u
0
r
2
+ u
1
r
1
+ u
2
+ u
3
r + . . .)
r=0
= 0. (158)
With u
0
= 0 from continuity, and by eliminating terms which vanish for r = 0, this becomes
u
1
r
1
+ 2u
2
u
1
r
1
u
2
= u
2
= 0. (159)
In fact these conditions are always satised if a and b are chosen so that
u
II
(r) = Asin(Kr). (160)
26
3.10.1 Boundary condition at interface
At the point r = R we again require continuity and smoothness for (r). The continuity condition
I
(R) =
II
(R) requires
u
I
(R) = u
II
(R), (161)
or
Asin(KR) = e
ikR
e
i2
0
(k)
e
ikR
(162)
Smoothness at r = R requires
R
I
(R) =
R
I
I(R) leads to
u
I
(R)
R
u
I
(R)
R
2
=
u
II
(R)
R
u
II
(R)
R
2
. (163)
Since we already have u
I
(R) = u
II
(R), this just requires
u
I
(R) = u
II
(R), (164)
which gives
AK cos(KR) = ik(e
ikR
+ e
i2
0
(k)
e
ikR
). (165)
We can clearly eliminate A just by dividing (162) by (165), giving
k tan(KR) = iK
e
ikR
e
i2
0
(k)
e
ikR
e
ikR
+ e
i2
0
(k)
e
ikR
. (166)
Multiplying through by e
ikR
+ e
i2
0
(k)
e
ikR
then gives
k tan(KR)(e
ikR
+ e
i2
0
(k)
e
ikR
) = iK(e
ikR
e
i2
0
(k)
e
ikR
). (167)
Solving for e
i2
0
(k)
gives
e
i2
0
(k)
=
K + ik tan(KR)
K ik tan(KR)
e
i2kR
. (168)
Note that in general
z
z
=
re
i
re
i
= e
i2
, (169)
where
= tan
1
_
y
x
_
. (170)
this gives us
0
(k) = kR + tan
1
_
k tan(KR)
K
_
. (171)
If we wish to, we can use Eq. (147), which states
k cot(
0
(k)) =
1
a
+
1
2
r
e
k
2
+ . . . , (172)
to determine the swave scattering length. Expanding k cot(
0
(k)) in powers of k, with K =
k
2
v
2
, so
that v =
2MV
0
/, gives
k cot(
0
(k)) =
v
Rv tanh(Rv)
+
1
6
_
3R
R
3
v
2
(Rv tanh(Rv))
2
+
3
v(Rv tanh(Rv))
_
k
2
+ . . . . (173)
27
From this we can see immediately that the scattering length is given by
a = R
1
v
tanh(Rv), (174)
and the eective range is
r
e
=
1
3
_
3R
R
3
v
2
(Rv tanh(Rv))
2
+
3
v(Rv tanh(Rv))
_
. (175)
It is good to see that we recover the hard sphere scattering length a R and eective range r
e
2R/3
in the limit v .
28
Much more than documents.
Discover everything Scribd has to offer, including books and audiobooks from major publishers.
Cancel anytime.