Chapter 11

Elementary Scattering Theory

As seen in the previous chapter, the bound state problem is characterized through the
stationary, normalizable states in the Hilbert space. Since quantum mechanical states in
principle always have to be normalized (because of the statistical interpretation of the
wave functions), scattering states should be non-stationary, normalizable solutions of
the Schrodinger equation.
An elementary, conceptually not quite satisfactory, however, in practical applications
extremely successful approach, consists of dropping the normalization condition. This
allows to introduce scattering states as suitably chosen stationary, non-normalizable
solutions of the Schrodinger equation. Stationary states are in principle eigenstates of
the Hamiltonian. If they are supposed to be non-normalizable, they have to be special
solutions from the continuous spectrum of H .

11.1 Free Motion

The force free, single particle motion (free motion) is given by a Hamiltonian, often called
free Hamiltonian,
~2
H0 = P2m :
(11.1)
Possible eigenvalues of H0 are the momentum eigenstates dened by
(11.2)
P~ j 'p~i = ~p j 'p~i :
251

With (11.2) and (11.1) follows

~2
H0 j 'p~i = P2m j 'p~i = Ep j 'p~i ;
the "norm" of those states can be chosen as
h'p~ j 'p~i = (p~ 0 ? ~p) :

(11.3)

(11.4)
The specic form of j 'p~i is obtained when the explicit representation of the operator P~
in coordinate space is employed:
h~x j P~ j 'p~i = hi r~ h~x j 'p~i = hi r~ 'p~(~x) = ~p 'p~(~x) ;
(11.5)
which has as solution of the dierential equation
(11.6)
'p~(~x) = (2h1)3=2 e hi p~
~x := h~x j p~i ;
which corresponds to a plane wave. The norm is explicitly given as
Z
1
h'p~ j 'p~i = (2h )3 d3x e? hi p~
~x e hi p~
~x = (p~ 0 ? ~p) ;
(11.7)
thus, the norm in a regular sense does not exist.
0

In general, the time evolution of solutions of the free Schrodinger equation is given
by
(11.8)
? hi dtd j '(t)i = H0 j '(t)i
with the solution
j '(t)i = e? hi H0 t j 'i :
(11.9)
With j 'p~i as starting vector follows with (11.3)

j 'p~(t)i = e? hi Ept j 'p~i :

(11.10)
This is a stationary solution with the typical time dependence given in a phase factor. It
is obviously not normalizable
(11.11)
h'p~ (t) j 'p~(t)i = h'p~ j 'p~i = (p~ 0 ? p~) :
Inserting (11.6) into (11.10) gives the explicit representation
'p~(~x; t) = e? hi Ept 'p~(~x) = (2h1)3=2 e hi (p~
~x?Ept) ;
(11.12)


252

which is just the De-Broglie wave. In coordinate space, the physical interpretation of
(11.12) can be readily studied. We have a plane wave. Positions with the same phase

p~
~x ? Ept = constant
(11.13)
spread with the phase velocity
x = Ep p^ = 1 p2 p^ = p p^ = v p^ ;
~vph = d~
(11.14)
dt
p
p 2m
2m
2
where p^ = p^= j ~p j indicates the direction of the spread. This leads intuitively to the
fact that plane waves are used to describe the motion of quanta with a denite starting
momentum p~.

11.2 Free Wave Packets

Especially at the beginning, it is useful to understand that in principle we should have
described the free motion with a normalizable state j 'ai. We should work with a packet

j 'a i =

d3p j 'p~ih'p~ j 'ai =

d3p j 'p~i '~a (p~) ;

(11.15)

where '~a (p~) is the experimentally given distribution of momenta, e.g., around a specic
value ~pa. Of course, we can assume that this distribution is nite, so that the integral
over j '~a (p~) j2 exists and can be set to one. This means, we can require

k 'a = h'a j 'a i =

k2

d3p

h'a j 'p~ih'p~ j 'ai =

d3p j '~a(p~) j2 = 1 : (11.16)

In contrast to j 'p~i this normalizable state j 'ai is not an eigenvector of H0 to a denite

energy E , since
Z
Z
2
p
3
3
6 E j 'ai : (11.17)
H0 j 'a i = d p H0 j 'p~i '~a (p~) = d p 2m j 'p~i '~a(p~) =
The general solution (11.9) of the time-dependent Schrodinger equation (11.8) reads with
the initial state j 'a i:

j 'a(t)i =

d3p e? hi H0 t

j 'p~i '~a(p~) =

d3p? hi

p2
2m

j 'p~i '~a (p~) ;

(11.18)

and similarly to (11.17) the time dependence cannot be pulled in front of the integral.
One has instead

j 'a(t)i 6= e? hi Et j 'a i :


253

(11.19)

Using (11.12) we obtain for the wave packet in coordinate space representation
Z
'a (~x; t) = (2h1)3=2 d3p e hi (p~
~x?Ept) '~a (p~) :
(11.20)
It can be shown that for a general initial state j '(0)i the explicit time evolution is given
by
3=2 Z

(11.21)
d3x0 e hi m2t (~x?~x )2 '(~x; 0) :
'(~x; t) = 2imh t
Estimating the absolute value gives

m 3=2 Z d3x0 j e 2imt (~x?~x )2 j j '(~x) j
j '(~x; t) j  j t 1j3=2  2i
h 
= constant
j t j3=2 :
(11.22)
0

Thus for the probability density, we obtain

(~x; t) = j '(~x; t) j2  constant

j t j3=2 :

(11.23)

However, the total probability, which corresponds to the conservation of the number of
particles is given by

k '(t) k2 = h'(t) j '(t)i = he? hi H t '(0) j e? hi H t '(0)i

= h'(0) j '(0)i = k '(0) k2 :


(11.24)

This means that the total probability is independent of t; however, in (11.21) one has to
integrate over increasingly larger areas in order to compensate the jtj13=2 behavior.

11.3 Scattering States

As we saw with respect to the free motion, the energy eigenvalues in the continuous
spectrum are innitely degenerate. In this simple case, it was possible to explicitly give
the physically realized solutions, the plane waves. If a potential V is present, this is in
general no longer the case. Here we need boundary conditions in order to pick the
physically relevant solutions among the innitely many possible solutions.

Asymptotic Conditions:

If we assume that the region of the interaction V is specially restricted (short-ranged

254

potential), then it seems plausible to assume that the scattering solution j p~(+) i consists
of a superposition of the free solution and a scattering piece j psci, which is given by
the potential. According to the previous discussion, one chooses as the free solution the
momentum state j 'p~i. We, therefore, assume that the solution of the full eigenvalue
equation, which is associated with an incoming particle with sharp momentum ~p,
2
H j p~(+) i = E j p~(+) i = 2pm j p~(+) i
(11.25)
can be written as
(11.26)
j p~(+) i = j 'p~i + j p~sci :
In principle, this is not yet a real statement, since one can always separate a piece j 'p~i
o any vector (which is normalized on a  function). The ansatz (11.26) becomes less
trivial if we assume that the piece j p~sci is caused by the potential, and thus vanishes if
the potential vanishes. This transition can be achieved by multiplying a given potential
with a factor  and one studies the limit  ! 0.
In order to obtain a detailed requirement for j p~sci, we use the coordinate space representation. Here (11.26) reads
1 e hi p~
~x + sc(~x) :
(+)
(11.27)
p~
p~ (~x) = (2h
 )3=2
As already mentioned, p~sc(~x) is in general in a complicated way given by the potential
V . If one considers how the in uence of a perturbation (here a short-ranged potential)
in uences a plane wave, it is reasonable to require as physically motivated boundary
condition that p~(+) (~x) behaves asymptotically as an outgoing spherical wave, i.e.,
1 f (^x; p^) e hi pr ;
!1
sc(~x) r?!
(11.28)
p~
(2h )3=2 E
r
where E = 2pm2 . With this we x the scattering solution as the specic solution of (11.25)
for which in the coordinate space representation the condition (11.28) holds. This means
they are dened by the following requirements
#
"
2
h

2
~
? 2m r + V (~x) p~(+) (~x) = E p~(+) (~x)
(11.29)
3

i
h pr
i ~
~x
1
e
!1
(+) (~x) r?!
4e h
 p
+ fE (^x; p^) r 5 :
(11.30)
p~
(2h )3=2
The factor fE (^x; p^) is a measure for the magnitude of the scattering part p~sc(~x) of the
scattering solution and is called scattering amplitude. According to our assumptions
fE (^x; p^) has to vanish if the potential is zero. The condition (11.30) is also known as
Sommerfeld Radiation Condition.

255

11.4 Cross Section

The scattering of a particle under the in uence of a potential can be described in the
following fashion. From the incoming current ~j0 of particles the scattering process creates
a current ~j sc of scattered particles.

n . dF

j0

Fig. 11.1

z-axis
scattering
center

Explanation of the dierential cross section.

Experimentally one measures at a distance r from the scattering center the probability
current which penetrates through an area
(11.31)
dF = r2 d
:
Here d
is a solid angle which is given by the resolution of the detector (ideally d

is innitesimally small). This current is then related to the incoming current, i.e., one
considers the quantity
2
~ sc
~ sc
d = j
~~n dF = j
~n~ r d
:
(11.32)
j j0 j
j j0 j
Here ~n characterizes the location of the areal element dF through which the scattered
particles penetrate. This denition (11.32) shows that d has the dimension of an area.
The current is given by the probability density multiplied by the velocity. For the incoming
plane wave, this means

2

i p~
~x 
1
~v ;

~j0 = 
h
e
(11.33)

~
v
=
 (2 h

 )3=2
(2h )3=2
whereas for the scattered wave
2



~j sc = j sc j2 v0
~n =  1 3=2 fE (^x; p^) 1 e hi pr  v0~n

 (2 h
)
r
256

0
2
= j fE (^x2; p^) j v 3 ~n :
r
(2h )

(11.34)

Here the current points in direction of ~n, i.e., it is perpendicular to the corresponding
spheres whose center is given by the scattering center. Inserting (11.33) and (11.34) into
(11.32) leads to
s

0
0
d = j fE (^x; p^) j2 vv d
= j fE (^x; p^) j2 EE d
:
(11.35)
The factor r12 in (11.34) is canceled through the multiplication with the factor r2 from the
area dF in (11.31). This is important since then the result (11.35) does not depend on
the distance r at which the detector is positioned. Experimentally one only has to make
sure that this detector is positioned outside the interaction region, where the asymptotic
conditions are realized. When considering short-ranged potentials, this is trivially fullled,
however, leads to complications when considering the quantum mechanical treatment of
Coulomb scattering.

In (11.33) and (11.34) v and v0 denote the velocity of the incoming and outgoing particles.
Correspondingly the factor
s

v 0 = p0 = E 0
(11.36)
v
p
E
occurs in (11.35). However, in the here considered case of potential scattering v = v0, a
condition which as been assumed in the asymptotic condition (11.30). Scattering with
v = v0 is also called elastic scattering. From (11.35) we derive that the dierential
cross section dd
is given through the scattering amplitude fE (^x; p^) according to
d = j f (^x; p^) j2 :
(11.37)
E
d

The amplitude fE has obviously the dimension of a length, which can be immediately
seen from (11.30). In (11.35) the factor (11.36) has been considered explicitly to point out
that in, e.g., inelastic processes energy can be lost in the scattering process, e.g., through
the excitation of composite targets, and then the factors (11.36) have to occur in the cross
section.

Spatial Symmetries
if the potential is rotationally symmetric or has an axial symmetry with respect to the
direction p^ of the incoming current, one has to expect that the scattering amplitude will
257

only depend on the azimuthal angle  with respect to the direction of the incoming particle.
Choosing the incoming direction as z-axis, we obtain for the asymptotic condition (11.30)
2

i
1
r!1
(+)
4e h

p~ (~x) ?! (2h
 )3=2

i
h pr
e
+ fE () r

pz

3
5

and the dierential cross section becomes

d = j f () j2 :
E
d

(11.38)

(11.39)

Total Cross Section

If one sums up the scattering into all dierent directions, i.e., integrates the dierential
cross section over all angles, one obtains the total cross section
Z
Z
d
tot = d
d
= d
j fE (^x; p^) j2 :
(11.40)
If the potential has the symmetries leading to (11.39), then this denition simplies to

tot =

d
j fE () j2 =

= 2

2

0
Z 1

?1

sin  d d' j fE () j2

d cos  j fE () j2 :

(11.41)

In the preceding discussion we introduce the probability current densities by multiplying

the probabilities with the corresponding velocities, as is customary in electro dynamics,
where charge density times velocity is introduced as current density. This is a quite
intuitive procedure and leads quickly to the desired results. However, we could also have
used the more formal denition of a current density
~ (~x; t)) :
~j (~x; t) = h =m(  (~x; t)r
(11.42)
m
The current of the incoming particles is then obtained by inserting the plane wave:

~ e hi p~
~x
~j0 (~x; t) = h =m 1 3 e? hi p~
~x r
m
(2h )
h 1 =m i ~p = ~p=m ;
= m
(2h )3
h
(2h )3
"

258

(11.43)

a result which coincides with (11.33). Similarly, we can determine the current of scattering
particles in direction ~n and obtain
i
h =m h sc(~x; t)~n
r
~ sc(~x; t)
jrsc = ~n
~j sc(~x; t) = m
"
#
h

@
= m =m sc(~x; t) @r sc(~x; t) :

(11.44)
~ = ~xr
r
~ = @r@ . Now we need the
Here we used that ~n = ~xr , from which follows ~n
r
scattered current at the place of the detector, i.e., large r, and thus can use (11.30). This
gives
"
 (^x; p^) i fE (^x; p^) #
1
f
h

r
!1
E
sc
sc
jr = ~n
~j (~x; t) ?! m (2h )3 =m
r h p r
 
+ higher orders in 1r :
(11.45)
Aside from higher orders, which are negligible for suciently large r, we obtain
1 j fE (^x; p^) j2 p ;
!1
~n
~j sc r?!
(11.46)
(2h )3
r2
m
which corresponds to (11.34). Inserting (11.43) and (11.45) into (11.32), where now
automatically p0 = p, we obtain the dierential cross section (11.37).

11.5 Phase Shifts

The task is to determine fE (; '), i.e., the solution of the Schrodinger equation
2
? 2hm r2 p~(~r) + V (j ~r j) p~(~r) = Ep~ p~(~r) :
(11.47)
If V is rotationally symmetric, i.e., [H; L2 ] = 0, then we can write
X
a^`(p) u`;pr(r) Y`m(; ') :
(11.48)
p~ (~r) =
`;m
For p~ parallel to the z-axis,  is the angle between p~ and the axis, and the problem is
symmetric around the z-axis, i.e., independent of '. Thus we use
1
X
(11.49)
a`(p~) u`;pr(r) P`(cos ) :
p~ (~r) =
`=0
259

Inserting in (11.47) gives as radial Schrodinger equation

d2 u (r) + p2 ? 2m V (r) ? `(` + 1) u (r) = 0

`;p
dr2 `;p
r2
h 2
#

"

(11.50)

where we used that Ep = 2pm2 . The requirement that only the solution which is regular
at the origin is physically allowed gives u`;p(0) = 0 and leads to a unique solution of
(11.50). Next, we need to consider the behavior of u`;p(r) for r ! 1. If the potential
V (r) vanishes suciently strong for r ! 1, then (11.50) goes asymptotically for large r
to
#
"
d2 u (r) + p2 ? `(` + 1) u (r) = 0 ;
(11.51)
`;p
dr2 `;p
r2
which has the general solution

u`;p(r) = B`;pr j`(pr) + C`;pr n`(pr) ;

(11.52)

where j`(pr) and n` (pr) are the Bessel and Neumann functions. They have the following
properties:

pr << 1 : j` (pr)  (pr)` =^ regular solution

n`(pr)  ?(pr)?`?1 =^ irregular solution
pr >> 1 : j` (pr) ; pr1 sin pr ? `2
!

n`(pr)

? pr1 cos pr ? `2 :

!

Thus, we have the following asymptotic behavior for the radial component of the scattering
solution:
`;p(r)

 u`;pr(r)

pr>>1

sin pr ?
B`;p
pr

`
2

cos pr ?
? C`;p
pr

`
2

(11.53)

or equivalently


sin pr ?

`
2

+ `(p)

u`;p(r) pr>>
1
:
(11.54)
;
A`;p
r
pr
Thus `(p) in (11.54) is a phase shift originating from the potential, which describes a
shift with respect to the free wave (solution). The phase shifts ` do not only depend
260

on the potential but also on the scattering energy Ep, thus `  `(p). (11.54) can be
rewritten as
"
!
!
1
`
`
1
A`;p
pr sin pr ? 2 + ` (p) = A`;p pr sin pr ? 2 cos ` (p)
!
#
`
+ cos pr ? 2 sin `(p) ;
(11.55)
and thus we have




u`;p(r) ; A cos  sin pr ? `2 + A sin  cos pr ? `2 :
(11.56)
`;p
`
`;p
`
r
pr
pr
A comparison of the coecients of (11.56) and (11.53) gives
B`;p = A`;p cos `
C`;p = ?A`;p sin `
(11.57)
with A`;p being a free parameter, which is determined by the overall normalization of the
wave function. Thus, the radial solution can be written as
u`;p(r) = A [cos  j (pr) ? sin  n (pr)] :
(11.58)
`;p
` `
` `
r
Through this equation (11.58) the phase shifts ` are dened. The radial solution R`;p(r)
can be further rewritten as
(11.59)
R`;p(r) = u`;pr(r) = A`;p [ j`(pr) ? tan ` n`(pr)]
with
A`;p = A`;p cos ` = B`;p
(11.60)
and
sin ` = ? C`;p :
tan ` = cos
(11.61)
`
B`;p
In practical applications, one obtains the phase shift tan ` by solving the radial Schrodinger
equation (11.50) up to a value of r where the potential has suciently fallen o (in case of
V being the nucleon-nucleon (nn) interaction typical values of r are 15fm). Eq. (11.59),
together with its derivative, gives two equations to determine tan ` as
0
(11.62)
tan ` = p j0`(pr) ? j`(pr)
p n`(pr) ? n`(pr)
261

where is the logarithmic derivative of the solution R`;p(r)

d R (r) :
= R 1(r) dr
(11.63)
`;p
`;p
Once the phase shift ` is obtained, the asymptotic behavior of the scattering solution is
given as (cp. 11.49)




2
` 3
`
1
cos
pr
?
sin
pr
?
X
2
2 5
+
sin

P`(cos ) :(11.64)
a`(p) 4cos `
`
p~ (~r) ;
pr
pr
`=0
Since we are looking for a solution with a denite asymptotic behavior (11.30), we have
to decompose (11.30) with respect to spherical waves. With
1

eipz =

fE () =

`=0

`=0

i` (2` + 1) j`(pr) P`(cos )

(2` + 1) f`(p) P`(cos )
(11.65)

we can write (11.30) as



sin pr ? `2
eipr 5 P (cos )
`
4
(2` + 1) i
+
f
p~ (~r) ;
` (p)
`
pr
r
`=0
"
1
?ipr
ipr #
ipr
X
e
e
e
`
=
(2` + 1) 2ipr ? (?1) 2ipr + f`(p) r P`(cos )
`=0 "
!
1
ipr #
?ipr
X
(2
`
+
1)
1
e
e
(?1)`+1
=
2ip
r + (2` + 1) 2ip + f`(p) r P`(cos ) :
`=0
(11.66)




`
`
1
ei(pr? 2 ) ? e?i(pr? 2 ) . In
Here we used that i` = ei` 2 and pr1 sin pr ? `2 = 2ipr
(11.64) the asymptotic form of the scattering solution was given. If one does not use this
form, the original wave function reads
2

p~ (~r)

`=0

a`(p) [cos `(p) j`(pr) ? sin `(p) n`(pr)] P`(cos ) :

(11.67)

Comparing the coecients in (11.67) and (11.66) leads to the determination of a` (p) and
f`(p) via
a` (p) = (2` + 1) i` eiL (p)
e2i` (p) = 1 + 2ip f`(p)
(11.68)
262

from which follows

f`(p) = 21ip [e2i` (p) ? 1]

or

(11.69)

f`(p) = p1 ei` (p) sin `(p) =: p1 t`(p) ;

where t`(p) is called the partial wave amplitude.

(11.70)

Thus, the solution of the Schrodinger equation for positive energies is given by
1
(+) (~r) = X (2` + 1) i` ei` (p) u`;p(r) P (cos )
(11.71)
`
p~
r
`=0
and
1
X
fE () = p1 (2` + 1) ei`(p) sin `(p) P`(cos )
`=0

1
X
1
= p
(2` + 1) t`(p) p`(cos ) :
`=0

The dierential cross section is then given by

d = j f () j2 = 1 X (2` + 1)(2`0 + 1)
E
d

p2 ``
ei(` ?` ) sin ` sin ` P`(cos ) P` (cos )

(11.72)

and the total cross section by

tot =

2

d'

(11.73)

dcos dd

?1
2 (2` + 1)2 sin2  (p)
`
2` + 1
Z

1
X
= 2p2
`=0
1
X
4

= p2
(2` + 1) sin2 `(p)
`=0

where us used

d P` P`0 =

2
2`+1)

p;`
tot

(11.74)
` `. If we dene a "partial wave" cross section as
:= 42 (2` + 1) sin2 `
p
1
X
=
p;`
0

`=0

263

(11.75)

max =
then we see that for each `; p;`

4
p2

(2` + 1) is the upper bound for p;`.

The expressions (11.73) and (11.74) for the dierential and total cross sections are only
practically useful if only a few partial waves contribute, i.e., are dierent from zero. This
is the case for short-ranged potentials and small energies, as we will see later. If too
many `'s have to be taken into account, the use of a partial wave expansion becomes
questionable.

Insert: Dierent view on boundary conditions:

Consider the free solution of the radial Schrodinger equation

ul;p(r) = Al;p [cos l rjl (pr) ? sin l rnl(pr)]

 eil [sin l Gl (pr) + cos l Fl (pr)]

(11.76)

Fl (pr) = prjl (pr)

Gl (pr) = ?prnl (pr):

(11.77)

with

Then
and

0 pr) ? Fl (pr) (R)

tan l = ? ??GFl0 ((pr
) ? G (pr) (R)
l

(11.78)


(11.79)
(R) = u1 dudrl; p 
l;p
r=R
Here one can establish a connection to bound states. A bound state is characterized by a
square integrable solution,
Z

dru2l (r) = 1
0
rlim
!1 ul (r) ! 0:

(11.80)

For scattering states one has the asymptotic behavior

ul (r)  ei(pr?l 2 );

(11.81)

with Ep = p2 =2. With p  i this becomes Ep = ?2 =2 and

ei(pr?l 2 ) = er il ;

(11.82)

where only p = i gives a normalizable wave function. Then the logarithmic derivative is
given by
= ?:
(11.83)
264

This gives an extra condition to the wave function which makes our articial bound state
problem into an eigenvalue problem.

End insert
The most simplest case occurs if only the lowest partial wave has to be considered (` = 0),
i.e., if we have s-wave scattering. Then
d = sin2 0 ;
(11.84)
d

p2
which is independent of the scattering angle . Thus, for s-wave scattering, the dierential
cross section is isotropic.
A dierent expression for the phase shift can be obtained in the following way. We consider
the dierential equations for the free solution
"
#
`
(
`
+
1)
00
2
(r j` (pr)) + p ? r2 (r j`(pr)) = 0
(11.85)
and for the full solution
"
#
`
(
`
+
1)
2
m
00
2
u`;p(r) + p ? r ? 2 V (r) u`;p(r) = 0 :
(11.86)
h
Multiplying (11.85) with u`;p(r) and (11.86) with r j` (pr) and subtracting both equations
leads to
d [(r j (pr))0 u (r) ? (r j (pr)) u0 (r)] = ?j (pr) 2m V (r)r u (r) : (11.87)
`
`;p
`
`
`;p
`;p
dr
h 2
Integrating over r and taking into account the boundary condition us;p(0) = j`(0) = 0 as
well as asymptotic forms
!
`
r!1 1
r j`(pr) ?! p sin pr ? 2
!
`
r!1 1
u`;p(r) ?! p sin pr ? 2 + `(p)
!
`
r!1
0
(r j`(pr)) ?! cos pr ? 2
(11.88)
leads to

sin `(p) = ? 2m2

h

dr r j`(pr) V (r) u`;p(r)

265

(11.89)

where

cos pr ? ` 1 sin pr ? ` + `(p)

2 p
2
!

? p1 sin pr ? `2
!
`
cos pr ? 2 + `(p) = 1p sin `(p)
!

has been used. The expression (11.89) is no simplication with respect to the form
(11.62). Though no asymptotic conditions enter, the integral of (11.89) has to be solved
for all values of r to give the exact solution for sin `(p). In addition, (11.89) requires the
calculation of the scattering solution u`;p(r). However, if one uses the approximation
(11.90)
u`;p(r)  pr j`(pr) ;
one obtained the so-called Born approximation for the phase shift
Z 1
2
m
1
dr V (r) [pr j`(pr)]2 :
(11.91)
sin `(p)  ? 2 p
0
h
The question is under which conditions can (11.91) provide a good approximation for the
phase shift. This should be expected if the right-hand side of (11.94) is small compared
to 1, or more precisely, if the function j` (pr) is small in the domain of V (r). For small
incident energies, one can argue as follows: The function  j`() increases
close to the
q
`
+1
origin ( = 0) from zero as  . It has a turning point
at  = `(` + 1). Thus, we
q
can assume that pr j` (pr) stays small up to r = 1p `(` + 1) and that, therefore, the
approximation (11.91) is good if the range of the potential fullls
q
(11.92)
R  p1 `(` + 1) :
Formulated in a dierent way: If the range R and the incident particle energy (given p)
are xed, the phase shifts ` are small for all ` values which fulll (11.92) and then (11.91)
is a good approximation. This result can also be understood semi-classically. The length
q

`(` + 1)
h `(` + 1)
angular momentum
b` =
=
=
p
h p
momentum
is the impact parameter of the particles incident with a specic angular momentum. From
classical mechanics, we know that there is no scattering if the impact parameter becomes
larger than the range. For large energies, we use that the function  j`( ) is bounded for
all values of the argument. Then the right-hand side of (11.91) will be small if p is so
large that
2m 1 j V (r) j dr  1 :
h 2 p 0
Z

266

(This estimate is, of course, only valid if the integral over j V (r) j exists.) Thus, for high
energies the Born approximation (11.91) should be valid for all values of `. However, at
high energies one should calculate relativistically.
A nal remark:
If we had started with a radial wave function normalized such that

u`;p(r) ?! j`(pr) + tan ` n`(pr) ;

(11.93)

then (11.89) becomes

tan ` = ? 2m2 1p
h

dr j`(pr) V (r) u`;p(r) :

(11.94)

From the Born approximation for the phase-shift (11.91) we can draw some conclusions
about the relation of the sign of the phase-shift ` and the behavior of V (r), since all
other terms under the integral are positive.

 If ` < 0, then V (r) > 0, i.e. repulsive.

 If ` > 0, then V (r) < 0, i.e. attractive.

Example: Hard Sphere Scattering

Imagine a hard sphere of radius R, i.e. a potential that is innitely repulsive for r < R
and zero outside the sphere. Since the wave function does not penetrate inside the sphere,
the wave function must vanish for r < R. That is we obtain
(
;
r<R
(11.95)
u0(kr) = 0sin(
kr ? kR); r > R
A comparison with the free solution

ul(kr)  eil sin(kr ? l 2 + l )

(11.96)

0 (E ) = ?kR:

(11.97)

gives for the phase shift obtained for scattering from a hard sphere

Here we see that a negative phase shift usually arises from repulsive potentials. Note also,
that even a very simple potentials produce energy dependent phase shifts.
267

11.6 The Low Energy Limit

Here the behavior of the phase shift at low energies will be considered. For potentials
that can be assumed to vanish beyond r = R, the earlier expressions for tan `, (11.62)
and (11.94) can be used to explore the low energy behavior of the phase shift. When
the external kinetic energy E is small compared to the depth of the potential, the wave
function inside the potential will not depend sensitively on E . The total kinetic energy
at any radius is E + j V (r) j, which is for small E nearly equal to j V (r) j. Thus, we can
to rst approximation consider the logarithmic derivative (11.63) to be independent of
energy.
If we introduce the low energy behavior of j`(pr) and n`(pr) as
`+1
?! 1
3
(5pr

)
(2` + 1)
pr j`(pr) pr<<`
?pr n`(pr) pr<<`
?! 1
3
5(
pr

)(2` ` ? 1)

(11.98)

then we obtain from (11.62) after multiplying numerator and denominator with pR
(pR)2`+1
p!0 (` + 1) ? R 0 (R)
tan ` ?!
` + R 0(R) [1
3
5


(2` ? 1)]2 (2` + 1)

(11.99)

where 0(R) is the zero energy logarithmic derivative. Thus, as the energy approaches
zero, the tangent of the phase shift also approaches zero as
tan `  a` p2`+1 :

(11.100)

For ` = 0 (s-waves) (11.99) yields

p!0 1 ? R 0 (R)
tan 0 ?!
0(R) p = ?pa0 :

(11.101)

The quantity a0 is usually called the zero energy scattering length, or simply the
scattering length and is dened by
268

a0 = R 0((RR))? 1 :
0

(11.102)


If the wave function is small at r = R, the quantity R 0(R) will be large at aR0 ' 1;
if, instead, u0`;p(R) is nearly zero, R 0(R) will be small and a ' ? 0 (1R) . The scattering
length a0 has a simple geometric interpretation. In the low energy limit, the wave function
p!0
u0(r) r>R
= ei0 sin(pr + 0 ) ?!
p(r ? a0) :

(11.103)

Thus a0 is the point nearest to the origin at which the external wave function, or its
extrapolation toward the origin, vanishes. Considering the radial Schrodinger equation
for p ! 0; ` = 0, i.e.,
"

d2 ? U (r) u (r) = 0
0
dr2
#

we can deduce
1. For a repulsive potential (U > 0), the curvature of u0(r) is always away from the r
axis so that a0 > 0.

U(r)

Fig. 11.2

Illustration of the geometrical meaning of the scattering length a0 .

269

2. For an attractive potential

1. incapable of producing an s-wave bound state:
a0 < 0 (Fig. 11.3.a)
2. capable of producing an s-wave virtual state or "zero energy resonance": a0 =
1 (Fig. 11.3.b)
3. capable of producing one s-wave bound state: a0 > 0 (Fig. 11.3.c)

u0(r)

u0(r)

u0(r)

a
U(r)

U(r)

(a)

Fig. 11.3

(b)

U(r)

(c)

Illustration of the scattering length a0 for various attractive potentials U (r).

The scattering length also has an important physical signicance. In the low energy limit,
only the s-wave makes a non-zero contribution to the cross section (11.100) so that the
angular distribution of the scattering is spherically symmetric and the total cross section
is
p!0
tot = 4p2 sin2 0 ?!
4a20 :
(11.104)
This is similar to the result one obtains for the low energy scattering of a hard sphere of
radius a. Thus, the scattering length is the "eective radius" of the target at zero energy.
However, there is a factor of 4 between the quantum mechanical cross section at low
270

energies and the classical cross section for scattering from a hard sphere (classical = a2).
This can be explained by the fact that in quantum mechanics one considers probabilities,
i.e. the square of wave functions. At low energies the wave length is considerably larger
than the size of the target, thus the scattering can be viewed as scattering by on the entire
surface of the sphere, not only the cross section area of the sphere.

11.6.1 Eective Range Expansion

Here we only consider the very low energy regime, and thus we only need to take l = 0
into account. For the scattering amplitude we have
f0 (k) = k1 ei0 sin 0
0 = 1
sin 0
= k1 sin
?
i
e 0 k cos 0 ? i sin 0
0 = 1 1
1
= k 1 ?tan
i tan 0 k cot 0 ? i
1
(11.105)
= k cot  ? ik :
0

Using the denition of the scattering length, k cot 0 = 1=a0, we can rewrite the scattering
amplitude as
f0 (k) = 1 ?1 ik = 1 +aa02 k2
0
a0
2
= a0 + ia2 02k :
(11.106)
1 + a0 k
For the dierent pieces of the scattering amplitude we thus get
k!0
<ef0(k) ?!
 a0
k!0
=mf0(k) ?!  a20 k:

(11.107)

From Eq. (11.104) we saw that the elastic cross section in the low energy limit is given
by  = 4a20. Thus in the low energy limit the optical theorem
(11.108)
 = 4k =mf0 (k)
is fullled.

Important consequence: A single, real number, a0 , completely parameterizes all low

energy scattering. This is an advantage for e.g. experiments that use low energy neutron
271

scattering to study solids. It is a disadvantage for deduction of specic properties of the

projectile-target interaction.
Consider the s-wave wave function

u0(r)  ei0 [sin 0 G0(kr) + cos 0 F0 (kr)]

k!0
?!
kr + tan 0
 (r + a0 )k
(11.109)
This shows that outside of the range of the potential the wave function for in the low
energy limit is proportional to a0 + r, and thus intercepts the axis at r = ?a0 .
Let's go back to Eq. (11.105) and insert the eective range expansion to the next order,
i.e.
(11.110)
k cot 0 = ? a1 + 21 kr02:
0

This leads to the scattering amplitude

f0 (k) = ? 1 ? ik1 ? 1 kr2

a0
2 0
a
0
= ?
1 + ia k ? 1 a kr2
0

(11.111)

2 0 0

Again, this expression for the scattering amplitude fullls the optical theorem:
4 =mf (k) = 4 ?a0 (?a0 k) =
4a20
(11.112)
0
k
k 1 ? 21 a0 kr02 + a20 k2 1 ? 21 a0 kr02 + a20 k2
and
Z
a20
 = dd
= 4 jf0(k)j2 = 1 ? 1 a 4kr
(11.113)
2 + a2 k 2
2 0

11.6.2 Relation to Bound States

Consider the scattering amplitude
0  a0 :
(11.114)
fl=0 ' k1 ei0 sin 0 = k1 1 ?tan
i tan 0 1 ? ia0 k
f is an analytic function and has a pole at tan 0 = 1=i. Here this means that k = ?1=a0.
Considering the energy we have
2
?1  ?E :
E = 2km = 2ma
(11.115)
B
2
0
272

Thus, the knowledge of the scattering lengths seems equivalent to the knowledge of the
scattering length. If there is a bound state (with a small binding energy), it determines
the low energy scattering:
  jf0j2  2m(E1+ E ) :
(11.116)
B

Conversely, knowledge of the scattering length determines the bound state energy. E.g.
from the knowledge of the deuteron binding energy one determines the triplet nucleonnucleon scattering length to be
at0  p 1 = +4:32fm
(11.117)
EB 2m
Experimentally a value at0 = 5:42fm is extracted. The dierence comes from the fact
that the bound state is not truly at zero energy and some higher order terms are needed.
While the scattering length approximation does not hold if the energy gets far from
zero, the relation of the poles of the scattering amplitude to bound states is general.

Note:

11.7 Levinson's Theorem

The scattering length a0 is not simply a measure of the potential size or strength. If U (r)
becomes stronger, then more than one bound state can exist, and the cycle going from
negative to positive scattering length can repeat. This is consistent with tan 0 = pa0,
where tan is a multi-valued function.
tan 0 varies periodically and discontinuously from ?1 to +1. If the potential is strong
enough to have a bound state at k = 0, we must be on the second branch of the tan
function, and the zero energy phase shift is  rather than zero.
The general result of this consideration is Levinson's Theorem

(k = 0) ? (k  1) = nB 

(11.118)

where nB is the number of bound states supported by the specic potential U (r). Levinson's theorem relates the phase shift as zero and innite energy to the number of bound
states nB .

273

11.8 Breit-Wigner Resonances

Negative energy state (bound states) are stationary states and obey the stationary Schrodinger
equation. States with positive energy, conned in a positive potential well are conned,
but will eventually through the potential barrier. Those state are called quasi-bound
states or resonances.

V(r)

Resonances

E>0
r

Bound

E<0

States

Fig. 11.6

Sketch of a potential supporting bound states and

resonance states.

Consider the expansion of the phase shift tan ` for p ! 0 as given in Eq. (11.99),
(pR)2`+1
p!0 (` + 1) ? R 0 (R)
(11.119)
tan ` ?!
` + R 0(R) [1
3
5


(2` ? 1)]2 (2` + 1)
The denominator vanishes for R 0(R) = ?`. Thus tan ` ! 1, i.e. tan ` = 2 + n,
This condition occurs for a specic momentum pR at a specic energy ER = pR=2.
Expanding (pR) around Er gives

R) j
(pR)R  ?l + (E ? ER ) d(dE
E =Er

Substitution of Eq. (11.120) into Eq. (11.119) leads to

2`+1
tan `  E ?1 E 21 2(pR) 2 d( R)
R [(2` ? q )!!] dE
= E ?1 E ?2l
R

274

(11.120)

(11.121)

with

?l =

2(pR)2`+1 :
R)
[(2` ? q)!!]2 d(dE

(11.122)

This leads to the Breit-Wigner resonance form of the amplitude

kf` = ei` sin ` = E ? ?El =?2 i? =2
(11.123)
R
l
and the Breit-Wigner cross section in a specic partial wave partial wave `.
2
(11.124)
` = 4(2p`2+ 1) (E ? E?l)=24+ ?2 =4
R
l
where ?l is called the width of the resonance. From the derivation it is apparent that
?l  1= (pR). Thus, if ?l is small, (pR) (and thus `) varies rapidly near ER . If this is
the case, the resonance will be sharp, and the total cross section tot will have a sharp
peak.
However, unless the resonance is very narrow, then the 1=p2 factor in tot will distort
the shape of the cross section and may shift the peak from ER to a lower energy.
Physically, a sharp peak in the energy dependence of the cross section indicates a dynamical origin, such as a strong attraction at that energy. If the phase shift passes rapidly
through =2 (modulo ), this probably means a resonance, i.e. beam and target particle
binding temporarily and then breaking up again.
Resonances are poles in f`(E ) at E = ER ? i?l =2. This means
q
q
q
q
p = 2E = 2Er ? i?l = 2E 1 ? i?l =2Er
 pr ? i ?pl =2 (for small ?l )
(11.125)
r
If ?l is small, the pole is right below the real positive p-axis. When taking E = p2=2,
bound state poles map to the rst Riemann sheet, resonance poles move into the lower
half of the second (unphysical) Riemann energy sheet.
The smaller ?l , the sharper tot peaks, the longer lived the resonance is, and the closer
the pole is to the real axis.

11.9 The Classical Limit

Let us now examine the scattering amplitude (11.69, 11.72) in the classical limit, that
is, at a suciently high energy that the particle may be localized. At these energies,
275

many partial waves will contribute so that the partial wave expansion for f`() may be
approximated by an integral. In addition, most of the scattering will result from high
partial waves so that we can assume that the phase shift for "normal" potentials will
behave as shown in Fig. 11.4.

Fig. 11.4

E.

The phase shift ` as function of ` for high energies

Fig. 11.4 The phase shift ` as function of ` for high energies E .

The region where ` is relatively constant will make little contribution, as we will see
below. As a result, we have from (11.72)
2i`
X
fE () =
(2` + 1) e 2ip? 1 P`(cos )
`=0
Z 1
2i`
' 0 d` ` e ip? 1 P`(cos ) :
(11.126)
For large ` and small angles, the Legendre polynomial may be replaced by a Bessel function
using the expansion


 
1
1
(11.127)
P`(cos ) ' j0 2 ` + 2 sin 2  + 14 sin2 21  +


:
This greatly simplies the integration since it avoids integrating over the order of the Legendre polynomial. It is also convenient to use b = p` and to recognize that the momentum
transferred to the scattered particle, which we shall denote by q, is
q = j p~ ? ~p 0 j = (p2 + p02 ? 2pp0 cos )1=2
= p(2(1 ? cos )) 21 = 2p sin 
2
(11.128)
276

for elastic scattering, i.e., j p~ j = j ~p 0 j. Then (11.126) may be rewritten to obtain the
semi classical approximation
Z 1
p
db b (e2i(b) ? 1) j0 (qb) :
fE () = i
(11.129)
0
This formula is reminiscent of very similar formulas in the classical theory of diraction.
Its physical implications can be seen most strikingly by assuming that (b) has a limiting
form as suggested by Fig. 11.4, namely that it is constant for b < R and zero for b > R.
In this simple case, the integral in (11.129) can be performed analytically giving
(qR)
(11.130)
fE ()  j0qR
where j1(qR) is the rst-order Bessel function. This result, familiar from the theory of
Fraunhofer diraction, gives a cross section as shown in Fig. 11.5.

|f( )|

3.8
qR

Fig. 11.5

Dierential cross section is the classical limit.

As expected, the cross section is sharply peaked in the forward direction and is concentrated within the region having  < ( pR1 ). At high energies, we can extract the physical
content of (11.129) by using the method of "steepest descent." For large q the integrand
will oscillate rapidly as b varies. Asymptotically, the Bessel function has the form

 




!1
(2qb)1=2 exp i qb ? 4 + exp ?i qb ? 4
j0 (qb) qb?!
(11.131)
so that the dominant contribution to the integral will come from those values of b for
which 2(b)  qb is nearly constant. The term in (11.129) that is independent of the phase
277

shift does not contribute to the scatteringPaway from the

 forward direction as can be seen
1 P (cos ) =  (1 ? cos ). Thus,
by returning to (11.126) and noting that 1
`
+
`
`=0
2
the important region of b is determined, for xed , by the relation
d(b)  1 q = p sin  :
(11.132)
db
2
2
It may be seen from this, as pointed out earlier, that values of b, for which (b) is a
constant, will not contribute to scattering out of the forward direction (q > 0).

278