Supplement

8-A

A Useful Theorem
The following useful result appears in Paulis 1930 Handbuch Article on Quantum Theory: Consider eigenvalues and eigenfunctions of a Hamiltonian that depends on some parameterfor example, the mass of the electron, or the charge of the electron, or any other parameter that may appear in more complicated problems. The Schrdinger eigenvalue equation may then be written with the parameter explicitly indicated as, H( )un(r, ) E( )un(r, ) (8A-1)

It follows that with the eigenfunctions normalized to unity, d 3ru*(r, )un(r, ) n that E( ) d 3ru*(r, )H( )un(r, ) n . We get (8A-3) 1 (8A-2)

Let us now differentiate both sides with respect to E( ) d 3r u*(r, ) n H( )un(r, ) un(r, )

d 3ru*(r, )H( ) n

d 3ru*(r, ) n

H( )

un(r, )

Consider now the rst two terms on the right-hand side. Using the eigenvalue equation and its complex conjugate (with hermiticity of H), we see that they add up to E( ) d 3r u*(r, ) n un(r, ) E( ) d 3ru*(r, ) n un(r, ) E( ) We are therefore left with E( ) d 3ru*(r, ) n H( ) un(r, ) H( ) (8A-4) d 3ru*(r, )un(r, ) n 0

The utility of this result is somewhat limited, because it requires knowing the exact eigenvalues and, for the calculation on the right-hand side, the exact eigenfunctions.1 Nevertheless, the theorem does allow us certain shortcuts in calculations.
1

The extension of this to certain approximate solutions is due to R. P. Feynman and H. Hellmann. See Problem 10 in Chapter 14.

W-35

W-36

Supplement 8-A

A Useful Theorem

Consider, for example, the one-dimensional simple harmonic oscillator, for which the Hamiltonian is H The eigenvalues are known to be En (n
1 2

p2 2m

1m 2 2 x 2

(8A-5)

(8A-6)

If we differentiate En with respect to , and if we note that H we can immediately make the identication (n or x2
n
1 2

m x2

m 1 2

x2

En m 2

(8A-7)

Examples of relevance to the hydrogen atom are of particular interest. In the Hamiltonian, the factor e2 4 appears. The eigenvalue has the form Enl 1 mc2 2 n2
2 0r

c r

If we take as our parameter to be , then we get c 1 r so that 1 r

nl

n,l

Enl

mc2 n2

(8A-8)

mc n2

1 a0n2

(8A-9)

In the radial Hamiltonian, there is a term

2

l(l r
2

1) nr 1 2 mc 2 l
2

2m If we treat l as the parameter and recall that n

2

1, we get 2 n3 (8A-10)

2l r2

2m which is equivalent to 1 r2

1
nl

a2 n3(l 0

1 2

(8A-11)

A Useful Theorem

W-37

Using an observation of J. Schwinger that the average force in a stationary state must vanish, we can proceed from F dV(r) dr e2 4 to F(r) 0 and thus obtain 1 r3 e2 m 1 2 4 0 r2 l(l 1) 1
nl 0r 2

d dr
2

e2 4
0r

l(l 1) 2mr 2

(8A-12)

l(l 1) mr 3

nl

a3n3l(l 0

1 2

)(l

1)

(8A-13)

Supplement

8-B

The Square Well, Continuum Solutions

We saw in eq. (8-69) that asymptotically the free-particle solution has the form Rl(r) l 1 [e 2ikr
i(kr l /2)

ei(kr

l /2)

We now assert that the rst term is an incoming spherical wave, and the second is an outgoing spherical wave. The description is arrived at in the following way. Consider the three-dimensional probability ux j 2i ( *(r) (r) *(r) (r))

We shall see that it is only the radial ux that is of interest for large r. The radial ux, integrated over all angles, is thus d ir j(r) For a solution of the form (r) with d Ylm( , ) 2 nd that e ikr C r Ylm( , ) (8B-2) 2i d * r * r (8B-1)

1 the right-hand side of (8B-1) can easily be evaluated, and we kC

2

d jr

1 r2

(8B-3)

The sign describes outgoing/incoming ux. The factor 1/r 2 that emerges from our calculation is actually necessary for ux conservation, since the ux going through a spherical surface of radius r is r 2 d jr kC
2

independent of r

We therefore see that for the free-particle solution the incoming ux is equal in magnitude to the outgoing ux, which is what it should be, because there are no sources of ux. We now note that in the presence of a potential ux is still conserved. Any solution will asymptotically consist of an incoming spherical wave and an outgoing spherical W-38

The Square Well, Continuum Solutions

W-39

wave, with the constraint that the magnitude of the incoming ux and the outgoing ux be equal. Thus if the asymptotic solution has the form Rl(kr) l then it is required that Sl(k) We write Sl(k) in the standard form Sl(k) e2i
l(k)

1 (e 2ikr

i(kr l /2)

Sl(k)ei(kr

l /2)

(8B-4)

(8B-5)

(8B-6)

The real function l(k) is called the phase shift, because the asymptotic form of the radial function (8B-4) may be rewritten in the form Rl(r) l ei
l(k)

sin(kr

l /2 kr

l(k))

(8B-7)

Aside from the irrelevant phase factor in front, this differs from the asymptotic form of the free-particle solution only by a shift in phase of the argument. We note parenthetically that with a solution that has a 1/r behavior, the ux in any direction other than radial goes to zero as 1/r 2, and we were therefore justied in only considering the radial ux at large values of r. Let us now consider the special case of a square well. The above argument shows us that we only need to consider the phase shift, since at large distances from the well the only deviation from free particle behavior is the phase shift. We again consider the well V(r) 0 We again use the notation
2

V0

for r for r

a a

(8B-8)

2 (E
2

V0)

(8B-9)

Now the solution for r

a must be regular at the origin, so that it has the form Rl(r) Ajl( r) r a (8B-10)

The solution for r

a will contain an irregular part, so that we have Rl(r) Bjl(kr) Cnl(kr) r a (8B-11)

dRl(r) The matching of 1 at r Rl(r) dr djl ( )/d jl ( )

a yields an expression Bdjl/d Bjl( ) Cdnl /d Cnl ( )

k
a

(8B-12)
ka

from which the ratio C/B can be calculated. The ratio can be related to the phase shift. We do this by looking at the asymptotic form of the larger r solution, which has the form Rl(r) l B sin(kr kr l /2) C cos(kr kr l /2)

W-40

Supplement 8-B

The Square Well, Continuum Solutions

/k

(a)

u(r)

Solution for V = 0

/k

Solution for V = 0

(b) u(r)

Figure 8B-1 Continuum solution u(r) potential.

rR(r) for l

0 (a) attractive potential; (b) repulsive

Comparison with the form in (8B-7), which has the form Rl(r) l sin(kr kr l /2) cos l(k) cos(kr kr l /2) sin l(k)

shows that once we know C/B we can nd the phase shift from C B tan l(k) (8B-13)

The actual calculation of C/B is very tedious, except when l 0. In that case, using ul(r) rRl(r), we just have to match A sin r to B sin kr C cos kr (and the derivatives) at r a. Figure (8B-1) shows the shape of the wave functions for attractive and repulsive potentials for l 0. We conclude with the proof of a useful relation eikr cos
l 0

(2l

1)iljl(kr)Pl(cos )

(8B-14)

which is of great importance in scattering theory.

The Square Well, Continuum Solutions

W-41

The Plane Wave in Terms of Spherical Harmonics

The solution of the free-particle equation
2

(r)

k2 (r)

(8B-15)

can be written in two ways. One is simply the plane wave solution (r) eikr (8B-16)

The other way is to write it as a linear superposition of the partial wave solutionsthat is, Alm jl(kr) Ylm( , ) (8B-17)

We may therefore nd Alm such that (r) eikr in (8B-17). Note that the spherical angles ( , ) are the coordinates of the vector r relative to some arbitrarily chosen z-axis. If we dene the z-axis by the direction of k (until now an arbitrary direction), then eikr eikrcos (8B-18)

Thus the left side of (8B-18) has no azimuthal angle, , dependence, and thus on the right side only terms with m 0 can appear; hence, making use of the fact that Yl0( , ) 2l 1 4
1/2

Pl(cos )

(8B-19)

where the Pl(cos ) are the Legendre polynomials, we get the relation eikrcos
l 0

2l 1 4

1/2

Al jl(kr) Pl(cos )

(8B-20)

We may use the relation

1

1 2

d(cos ) Pl(cos ) Pl (cos )

1

ll

2l

(8B-21)

which is a direct consequence of the orthonormality relation for the Ylm and (8B-19) to obtain
1

Aljl(kr)

1 [4 (2l 2

1)]

1/2 1

dzPl(z) eikrz

(8B-22)

Compare the two sides of the equation as kr l 0. The rst term on the left-hand side is Al (kr)l 1, 3, 5, . . . , (2l
l

1)

and the corresponding power of (kr) on the right-hand side has

1

1 [4 (2l 2

1)]

1/2

(ikr)

l 1

dz Pl(z)zl/l!

The integral can be evaluated by noting that Pl(z) is an lth-degree polynomial in z. The coefcient of the leading power, zl, can be easily obtained from eq. (7-47) as the power of zl in d ( 1)l 1 2ll! dz
l

(1

z2)l

2l(2l

1)(2l

1) 2ll!

(l

1)

zl

0(zl 1)

W-42

Supplement 8-B

The Square Well, Continuum Solutions

We can rewrite this in the form zl 2l(2l 1)(2l 2ll! 2) (l 1) Pl(z) terms involving Pl 1(z) and higher

With the help of (8B-21) we nally get Al (kr)l 1, 3, 5, . . . , (2l 1) 1 [4 (2l 2 1)]1/2 (ikr)l 1 l! 2l(2l 2ll! 1)(2l 2) (l 2 1) 2l 1 (8B-23)

What results is the expansion eikrcos

l 0

(2l

1) iljl(kr) Pl(cos )

which we will nd exceedingly useful in discussions of collision theory.