Sunteți pe pagina 1din 8

Lecture 18: Scattering: partial wave analysis (11/22/2005)

How do we study the structure of matter and the character of the physical laws? We either try to
look at the energy spectrum and observe or calculate the transitions. This is what we have been
doing for quite some time.
Additional reading if you wish: Bransden & Joachain, ch. 13.2-13.4.
Alternatively, we can collide things and study scattering. The bigger momentum p you use,
the shorter is the corresponding distance h
/p, and therefore just by increasing the momentum (and
energy) of particles used in scattering, we may study the architecture of matter at very short distance
scales. This is the strategy associated with the words high-energy physics that the physicists
have followed for many decades. High-energy accelerators are the textbook example of this strategy
in practice.

Classical scattering: impact parameter


Imagine a bullet that is approaching a target. Extrapolate the line of the initial velocity at t =
into a straight line. The distance of this straight line from the center of mass of the target is known
as the impact parameter and it is denoted b. Draw a picture. As a vector, ~b is a two-vector, the
projection of the position of your particle into the two-dimensional plane orthogonal to the initial
velocity.
Example: hard sphere
We would like to figure out the relation between the scattering angle, the impact parameter, and
the cross section. Let us a pick a simple example. Imagine a particle that approaches a sphere of
radius R from the left. If denotes the angular distance of the epicenter from the direct impact,
then
b = R sin
The scattering angle is defined as the total angle by which the direction of motion changes
between t = and t = +. You can see on the picture above that = 2, and therefore
b = R sin[( )/2] = R cos(/2)
What is the cross section d/d per unit solid angle? Well, you should rewrite d and d using the
impact parameter and the scattering angle. Note that d is measured using the scattering angle
(and from ) i.e. from the momenta at infinity not from the detailed geometry of the hard
sphere. On the other hand, d is nothing else than the infinitesimal area in the two-dimensional
plane of the impact parameter parameterized by polar coordinates (b, ):
d = b db d,

d = sin d d

We therefore have

d
b |db|
R cos(/2)(R/2) sin(/2)d
R2
=
=
=
d
sin d
sin d
4
I hope that you remember that sin 2x = 2 sin x cos x which we used for 2x = . At any rate, you
see that the angular -dependence has canceled. In other words, classical hard sphere scattering is
isotropic; the probability that the projectile is deflected to some direction does not depend on the
1

direction. It is a special property of the hard sphere only and does not hold for potatoes. You can
also easily compute the total cross section
=

R2
d
=
4 = R2
d
4

and see that it agrees with the geometric cross section. This agreement is a general result for classical
hard scattering that holds not only for spheres: the calculation is nothing else than slicing of the
geometric cross section according to different values of the scattering angle .

Back to quantum mechanics


In quantum mechanics, the impact parameter and the scattering angle cannot be used in the same
way as in classical physics because of the uncertainty principle. For example, if you know that a
projectile whose momentum is p hits a point with accuracy R (the error of the impact parameter),
the uncertainty principle dictates that the transverse momentum must have uncertainty of order
pt = h
/R, and therefore the error of the scattering angle will have to be at least
pt
h

=
p
pR

If you take an electron with energy 200 MeV scattering off a proton (R = 1015 m), you obtain
1 which means that the scattering angle will be essentially unknown if you decide to have
a sharp value of the impact parameter. But of course, the probability of scattering into any given
solid angle must be calculable. The probability of any well-defined measurable process is calculable
in quantum mechanics.
Plane waves
Because the beam is typically much broader than the target, it is an excellent approximation to
describe the beam as a plane wave. We did something like that in one-dimensional scattering. In
the three-dimensional case, we may decide to denote the axis of the initial momentum as z + . The
incoming (in) plane wave is then
in = Aeikz
The outgoing wave can be approximated by a simple function far away from the center. Because it
is outgoing, it must be proportional to exp(ikr). The result is
"

ikz

out = A e

eikr
+ f ()
r

Well, yes, this also includes the incoming wave. We allowed an arbitrary dependence on . The
denominator 1/r is necessary for conservation of probabilities. Its because the probability density
|out |2 must decrease as 1/r 2 . There was no
dependence on because we assumed azimuthal symmetry. Recall that the wavenumber k = 2mE/
h as usual. All formulae above seem completely
transparent except for the unknown function f (). It is called the scattering amplitude and we want
to solve for it.
Before we will look for its values, we should understand the relation between the scattering
amplitude and the cross section. To do so, we must compare the probabilities that the incoming
and outgoing particle appears in certain regions. The probability that the incident particle passes
through an area d is
dP = |in |2 dV = |in |2 d v dt = |A|2 d v dt
2

On the other hand, the probability that the outgoing particle emerges in a solid angle d is
dP = |out |2 dV = |out |2 r 2 dr d = |out |2 r 2 (v dt) d =

|A|2 |f |2
(v dt) r 2 d
r2

We used the same symbol dP because these probabilities must actually be equal, by conservation of
probabilities. Comparing the two expressions, we find that
d
= |f ()|2 .
d
Its the simplest relationship between the amplitude and the differential cross section that you could
have imagined. Finally, we know enough to ask how much f () actually is.

Partial wave analysis


Next week, we will use a different approach (Born approximation). But today we will focus on one of
two most important techniques to calculate f (), the so-called partial wave analysis. It is based on
a decomposition into angular momentum eigenstates (essentially the spherical harmonics). This is
very useful when only one or a few of these angular momentum states is participating in the process.
There are two important cases when it is so:
~ = ~r p~, the product Rp of the target size R
low energy: because the angular momentum is L
and the initial momentum p controls the relevant spherical harmonics. Only those for which
l(l + 1)
h2 (Rp)2

will be important for the process. For example, if Rp  h


2, only l = 0 will be important
resonances: particular values of l may dominate the scattering at particular values of energy
Satisfying the Schr
odinger equation at infinity
How do we determine f () that appears in the wavefunction? Of course, the only equation that the
wavefunction must satisfy is the Schrodinger equation. We will try to solve it at r = only; it is
definitely a necessary condition for the wavefunction to satisfy the equation at least far away from
the target. Recall that our outgoing wavefunction is
(r, ) = eikz + f ()

eikr
r

where only the relative coefficient between these two terms matters. For a central potential V (r),
the solution may be decomposed as
(r, ) =

c0l Rl (r)Yl0(, )

l=0

Because neither the initial wavefunction nor the potential depends on , we have only included
the Ylm functions with m = 0. Because these spherical harmonics may be written in terms of the
Legendre polynomials:
s
2l + 1
Pl (cos )
Yl0(, ) =
4
3

we have everything we need to rewrite the wavefunction using new coefficients cl


(r, ) =

cl Rl (r)Pl (cos ).

l=0

The coefficients cl may a priori be arbitrary, and we will determine them later. However, Rl (r) must
satisfy the usual radial time-independent Schrodinger equation (simply substitute to the usual
Schrodinger equation!)
"
#
l(l + 1)
2mV (r)
1 d2
2
r
+k
Rl = 0
r dr 2
r2
h
2
Note that the differential operator is dressed as usual. There is a centrifugal potential coming
from the angular part of the Laplacian. There is also a term k 2 from the energy eigenvalue. Even
for V (r) = 0, the solution to this equation is subtle and involves the spherical Bessel functions jl (x):
Rl (r) = Al jl (kr)
We are fortunately looking at the r region only where the Bessel functions may be approximated as follows:
!
l
1
sin kr
jl (kr)
kr
2
One of the critical points will come right now. The point is that the wavefunctions cant differ
too much from this Bessel function. Note that even for V (r) = 0, the second linearly independent
solution of the differential equation that behaves as cos instead of sin would be divergent somewhere
in the center. This is qualitatively true even for other potentials: one of the solutions is convergent
and allowed and the other is not. The only difference from the V (r) = 0 case is a phase shift l that
makes the wavefunction be equal to
l
1
Rl (r) =
sin kr
+ l
kr
2

Rayleighs formula
The sum of these functions over l should equal to our asymptotic form
ikz

eikr
+ f ()
r

and therefore we also need to expand the latter wavefunction into the angular momentum states.
The result is called Rayleighs formula and we also show how it simplifies for r :
eikz =

X
l=0

il (2l + 1)jl (kr)Pl (cos )

il (2l + 1)

l=0

sin(kr l/2)
Pl (cos )
kr

Finally, we can set the two forms of the wavefunction equal:

X
l=0

cl

X
sin(kr l/2)
eikr
sin(kr l/2 + l )
Pl (cos ) =
il (2l + 1)
Pl (cos ) + f ()
kr
kr
r
l=0

Expand sin x as [exp(ix) exp(ix)]/(2i) and compare the coefficients in front of exp(ikr)/kr and
exp(ikr)/kr. You immediately get
c l = il (2l + 1)eil ,

f () =
4

1X
(2l + 1)eil sin l Pl (cos )
k l=0

The scattering amplitude f () is therefore completely determined by the phase shifts l . They
contain all the information about the potential. Recall that we started with the cross section;
showed that it was the squared absolute value of the scattering amplitude; and now we expressed
the scattering amplitude in terms of the phase shifts l .
Incidentally, for weak fields one can determine the sign of l qualitatively. Attractive potentials
have l > 0; the waves of the sin are squeezed because the negative attractive potential has a
similar effect as lowering of the values of l (it reduces the centrifugal potential). On the contrary,
repulsive potentials have typically l < 0 for the opposite reason i.e. the same reason. Draw three
graphs of sin.

Cross sections from the phase shifts


Because we now view the phase shifts as the best package for the information about scattering, we
must be able to express the differential cross section. We have
X

1 X
d
= |f ()|2 = 2
(2l + 1)(2l0 + 1)ei(l l0 ) sin l sin l0 Pl (cos )Pl0 (cos )
d
k l=0 l0 =0

which contains a lot of mixed terms with l 6= l0 . However, the total cross section is much simpler.
Because of the orthogonality relations for the Legendre polynomials
Z

Pl (x)Pl0 (x)dx =

2
ll0 ,
2l + 1

x cos

the total cross section is nothing else than


=

d(cos ) |f ()|2 =

4 X
(2l + 1) sin2 l
k 2 l=0

Pretty simple formula, right? The total cross section is not the only quantity in which the formula
simplifies. You may also calculate the forward scattering i.e. f ( = 0):
f ( = 0) =

1X
(2l + 1)eil sin l
k l=0

Dividing exp(il ) = cos(l ) + i sin(l ) into the real and imaginary part, you see that the imaginary
part is more than similar to the total cross section. In fact we have
=

4
Im f (0)
k

which is known as the optical theorem. Physically, you can see some interference effects in the
forward scattering that will tell you how big a part of the beam has been deflected. Practically, it has
been used to measure the total section of the proton-proton scattering in high-energy experiments.
Mathematically, it is the most famous consequence of the unitarity of the matrix of scattering
amplitudes. If you write the evolution operator S as 1 + iT , then the unitarity condition SS = 1
implies i(T T ) + T T = 0 which relates the imaginary part of T and the modulus T squared.

Low energy scattering


As we said, the two main applications of partial wave analysis involve low-energy
scattering and
h, then
resonances. Lets start with the low-energy scattering. If kR  1 where k = p/
h = 2mE/
5

only one partial wave, namely the l = 0 wave called the s-wave, gives a significant contribution to
the sum. In that case:
ei0
d
sin2 0
f ()
sin 0 ,
= |f ()|2
k
d
k2
The scattering is isotropic, much like for the classical hard sphere we started with, and
=

d
4 sin2 0
d
d = 4
=
d
d
k2

Scattering problem
The scattering problem is how to determine l from V (r). The inverse scattering problem is how
to determine V (r) from l and it is harder. For most potentials and most ensembles of the phase
shifts, the problem must be solved numerically. However, there are a few important simple enough
examples that can be solved in closed form.
Spherical well
Thats our simplest example. A potential such that
V (r) =

V0 r < a
0 ra

We want to avoid the Bessel functions, so let us study the case in which l = 0 is dominant. The
centrifugal potential in the equation for Rl (r) drops and the Bessel function simplifies to
R(r) = A

sin k 0 r
,
k0r

k 02 = k 2 +

2mV0
h
2

for r < a

sin(kr + )
,
for r > a
kr
Both as well as d/dr need to be continuous at r = a. These two conditions require
R(r) = B

Aka sin(k 0 a) = Bk 0 a sin(ka + ),


A[kk 0 a cos(k 0 a) k sin(k 0 a)] = B[kk 0 a cos(ka + ) k 0 sin(ka + )]

(1)

We dont really care about A, B but rather about , k, k 0 . So we divide these two equations in a
standard maneuver that imposes the continuity of the so-called logarithmic derivative d(ln )/dr =
(d/dr)/. The ratio of the two equations above is, in this case,
k 0 cot(k 0 a) = k cot(ka + )

tan(ka + ) =

k
tan(k 0 a)
k0

We wont simplify it too much more except for a few special situations:
k  k 0 i.e. V0  h
2 k 2 /2m. Because we still keep k 0 a  1 and of course ka  1, we can expand
both the the left-hand side and the right-hand side (to higher order) of the tan-equation to get
(k 0 a)3
k
ka + = 0 k 0 a +
k
3

After a simple cancellation,

kk 02 a3
k 2mV0 3
=

a
3
3
h
2
The total cross section is the geometric cross section times a small dimensionless number:


4 2
4
4
2 sin2 () 2
k
k
9

2mV0
h
2

2

4
a6 = a2
9

2mV0 a2
h
2

!2

V is an even simpler limit because k/k 0 0 and the tan-equation requires tan(ka+) =
0. You can easily see that = ka and thus = (4/k 2 )(ka)2 = 4a2 which is four times the
geometric cross section. The same result actually holds also for V +, the (quantum) hard
sphere. Indeed, quantum mechanics gives you 4 times the classical probability of scattering
at very low energies. Of course, in the classical limit at very high energies the total cross
section must agree with classical physics.
let us see a very nice and special effect. Assume ka  1 it is not necessary for the effect but
it simplifies our math. Consider general k 0 a without any restriction. But rewrite the equation
tan(ka + ) ka + =

k
tan(k 0 a)
k0

in the following way:

k
[tan(k 0 a) (k 0 a)]
0
k
What is it good for? We see that the difference vanishes for k 0 a = 1.43 or 2.46 or 3.47 or
infinitely many other values. At these special values of energy, = 0, = 0, and the target
becomes invisible! This is known as the Ramsauer effect. It has been observed on scattering
of electrons on noble gas atoms (He, Ne, . . . ).
=

consider a special case in which k 0 a = /2 or 3/2 or 5/2 and so on. In these cases, tan(k 0 a) =
. This means that also tan(ka + ) must be which implies that ka + = /2 or 3/2 or
5/2 and so on. Since ka  1, it also means that = /2 or 3/2 or 5/2 and so on. At any
rate, sin2 = 1 and = 4/k 2 . Note that
sin2 = 1 is the maximum value of any sin2 (x) you can have. The corresponding cross
section = 4/k 2 is thus also the maximum allowed cross section for the s-wave, and it
is known as the unitarity bound.
this cross section only depends on k but not on a or V0
the values of k at which the cross section is maximized exactly correspond to zero-energy
(meaning zero binding energy) bound states (also known as threshold bound states) of the
potential. A level for these values of k is just bound to the well.

Resonances
The existence of resonances is always associated with the existence of bound states. Choose the
exactly correct energy that is necessary to create a bound state, and you will do so. If you change
the energy a little bit, you will still have a huge cross section. The projectile will spend a lot of time
near the target before it escapes again. In terms of phase shifts, the resonances appear if one of the
7

numbers sin2 l approaches one, i.e. if one of the phase shifts l approaches an odd multiple of /2.
When it occurs, the particular l dominates the cross section and the total cross section
l =

4
(2l + 1)
4
2
(2l
+
1)
sin

l
k2
k2
1 + cot2 l

can grow arbitrarily large as cot l goes to zero. Because it goes to zero when the energy E approaches
the resonant energy ER , we can expand
cot l =

ER E
(E)/2

where the numerator guarantees that the quantity vanishes at E = ER and the denominator is a
slowly varying function of E. Using this parameterization,
l =

4
2 (E)/4
(2l
+
1)
k2
(E ER )2 + 2 (E)/4

For sharp, narrow resonances, (E) can be approximated by a constant and the previous formula
simplifies to the so-called Breit-Wigner resonance formula:
l =

4
2 /4
(2l
+
1)
k2
(E ER )2 + 2 /4

We have seen the same function of the type 1/(1 + x2 ) in our discussion of harmonic time-dependent
perturbations; it appears at many places in physics. Also in this case, the parameterization is chosen
in such a way that measures the width along the energy direction measured at half-height of the
maximum l (E). Draw a picture.
If l > 0, the resonance is associated with a metastable state formed by the well (attractive,
negative potential) and the angular momentum barrier (the positive centrifugal potential). Draw a
typical dependence of the total potential on r for some typical value of l. The state would become
stable in the limit in which the well is infinite; then the energy would be sharp. The width
determines the lifetime via the uncertainty principle:
=

S-ar putea să vă placă și