1

HW6.nb

HW 6
1. WKB harmonic oscillator
a) Energy levels
(Hitoshi does this problem in his WKB notes, on page 8.)
The classical turning points a, b are where KE=0, i.e.
1
E = VHxL = 2 m2 x2
so
2E
a, b = "############
# .
m 2

We apply or VHxL and the turning points to take the energy condition
b !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1
a 2 m@E - VHxLD x = In + 2 M @
and integrate the l.h.s.:
2 E0
a0 = - $%%%%%%%%%%%%%%
% ; b0 = -a0;
m 2
m 2 x2
V0@x_D = ;
2
Integrate@Sqrt@2 m HE0 - V0@xDLD, 8x, a0, b0<, Assumptions 8m > 0, > 0, E0 > 0<D
E0

= In + 2 M @

such that
We find

E = In + 2 M @
as desired.
1

b) SHO WKB wavefunctions

(Let us proceed in the units @ = m = = 1.)
The trick here is that we want to use the classicallyallowed WKB wavefunction for SHO in between the turning points,
and the "tunneling" WKB wavefunctions in the forbidden regions, which extend to . This can be seen easily on a plot
of the ground state, where the turning points are at 1:

HW6.nb

-x2
x2
PlotA9-14 ExpA E, =, 8x, -2, 2<, PlotStyle 8Dashing@8<D, Dashing@8.05, .05<D<E;
2
2

2
1.5
1
0.5

-2

-1

Our SHO energies (from part (a)), potential function, and classical turning points are:
1
En@n_D = n + ;
2
x2
V@x_D = ;
2
!!!!!!!!!!!!!!!!!!!!
a@n_D = - 2 En@nD ;
b@n_D = -a@nD;

We construct the wavefunction in allowed region using Eq. 30 in the notes. We split up the integral around x = 0 and
consider the two turning points a and b separately, so Mathematica doesnt give a trivial result:
13
i
y
j H2 Abs@V @a@nDDDL
z
z
j
z
a1@x_, n_D = j

j
j
z
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! z
k 2 HEn@nD - V@xDL {

12

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
CosAIntegrateA 2 HEn@nD - V@tDL ,

8t, a@nD, x<, Assumptions -> 8n >= 0, a@nD < x, x < 0<E - E;
4

i H2 Abs@V @b@nDDDL13 z
y
j
z
j
z
a2@x_, n_D = j
CosAIntegrateA
j
j
z
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! z
k 2 HEn@nD - V@xDL {

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2 HEn@nD - V@tDL , 8t, x, b@nD<, Assumptions -> 8n >= 0, x > 0, x < b@nD<E - E;
4
12

Similarly, for the forbidden regions we use Eq. 31:

13
y
i
1
z
j H2 Abs@V @a@nDDDL
z
z
j

f1@x_, n_D = j
j
z
j
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! z
2
k 2 HV@xD - En@nDL {
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
ExpA-IntegrateA 2 HV@tD - En@nDL , 8t, x, a@nD<, Assumptions 8n >= 0, x < a@nD<EE;
12

13
i
y
1
j H2 Abs@V @b@nDDDL
z
z
j
z
f2@x_, n_D = j

j
j
z
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! z
2
k 2 HV@xD - En@nDL {
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
ExpA-IntegrateA 2 HV@tD - En@nDL , 8t, b@nD, x<, Assumptions 8n >= 0, x > b@nD<EE;
12

HW6.nb

Ok, now what about the regions right around the classical turning points a and b? We know that the wavefunctions above
blow up near these points. So, lets define a radius about the turning points inside which we treat specially:
= 0.5;

Next, we construct a piecewise function from the asymptotic wavefunctions above outside these regions (using UnitStep):
wkbasy@x_, n_D = H-1Ln f1@x, nD H1 - UnitStep@x - Ha@nD - LDL +
H-1Ln a1@x, nD UnitStep@x - Ha@nD + LD H1 - UnitStep@xDL +
a2@x, nD UnitStep@xD H1 - UnitStep@x - Hb@nD - LDL + f2@x, nD UnitStep@x - Hb@nD + LD;

Now, in the special regions, we use the Airy functions themselves, and again construct a piecewise function:
wkbtpa@x_, n_D = H-1Ln AiryAi@-H2 Abs@V @a@nDDDL13 Hx - a@nDLD UnitStep@x - Ha@nD - LD
H1 - UnitStep@x - Ha@nD + LDL + AiryAi@H2 Abs@V @b@nDDDL13 Hx - b@nDLD
UnitStep@x - Hb@nD - LD H1 - UnitStep@x - Hb@nD + LDL;

Finally, we add these two chunks to get the whole function:

wkb@x_, n_D = wkbasy@x, nD + wkbtpa@x, nD;

WAIT: If were using the asymptotic wavefunctions mated to the actual Airy functions in the transition regions, why not
just use the Airy functions everywhere, expanding about the turning points? A moments reflection gives the answer: this
would work great coming in from and passing through the turning points, but as you move into the classicallyallowed
region, the Airy function would not reach its cosineshaped asymptote fast enough! The reason we do it piecewise is that
we want accurate results where E p V , and live with nastiness in the transition regions. (In the classicallyforbidden
regions, the Airy function should decay fast enough to match the asymptotic solution.)
To see this, lets construct this allAiry function wavefunction by expanding around the two turning points:
air@x_, n_D = H-1Ln AiryAi@-H2 Abs@V @a@nDDDL13 Hx - a@nDLD H1 - UnitStep@xDL +
AiryAi@H2 Abs@V @b@nDDDL13 Hx - b@nDLD UnitStep@xD;

Finally, we wish to compare to the exact results:

1
-x2
-14
exa@x_, n_D =

ExpA

E HermiteH@n, xD;
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!
2
2n Factorial@nD

Lets compare the exact and WKB solutions for n = 1 :

HW6.nb
Plot@8exa@x, 1D, wkb@x, 1D<, 8x, -4, 4<, PlotStyle 8GrayLevel@0D, Hue@0D<D;

0.6
0.4
0.2
-4

-2

2
-0.2
-0.4
-0.6

Our WKB solutions looks great at x and x 0 where E ` V and E p V respectively, as expected. The Airy
function matches nicely with the decaying wavefunctions in the forbidden regions, but does indeed mismatch the wavefunction in the allowed region.
Now, lets compare the exact and allAiry solutions:
Plot@8exa@x, 1D, air@x, 1D<, 8x, -4, 4<, PlotStyle 8GrayLevel@0D, Hue@0D<D;

0.6
0.4
0.2
-4

-2

-0.2
-0.4
-0.6
As expected, we get the opposite behavior. Coming in from , the wavefunction matches the exact solution nicely.
However, as we move toward x = 0, the discrepancy builds up until you get a mismatch.
Now lets look at n = 10:

HW6.nb
Plot@8exa@x, 10D, wkb@x, 10D<, 8x, -7, 7<, PlotStyle 8GrayLevel@0D, Hue@0D<D;

0.4
0.2

-6

-4

-2

-0.2
-0.4
Fantastic! Because the energy is much higher, the approximation is far more accurate (i.e, the state is much more classical),
and on this plot the mismatch between the Airy function in the transition region and the wavefunction in the allowed region
is not even visible.
Now lets look at the allAiry solution for n = 10:
Plot@8exa@x, 10D, air@x, 10D<, 8x, -7, 7<, PlotStyle 8GrayLevel@0D, Hue@0D<D;

0.4
0.2

-6

-4

-2

-0.2
-0.4
A bit better than n = 1. The discrepancy builds up again coming toward x = 0 from the turning points; however, because n
is even, there is no mismatch at x = 0, just a sharp point.
Finally, let us look at n = 20:

HW6.nb
Plot@8exa@x, 20D, wkb@x, 20D<, 8x, -10, 10<, PlotStyle 8GrayLevel@0D, Hue@0D<D;

0.4
0.2

-10

-5

10

-0.2
-0.4
We see that the WKB solution has a bit of overshoot due to the higher energy; otherwise the agreement is quite good.
Finally, let us compare the exact and allAiry solutions for n = 20:
Plot@8exa@x, 20D, air@x, 20D<, 8x, -10, 10<, PlotStyle 8GrayLevel@0D, Hue@0D<D;

0.4
0.2

-10

-5

10

-0.2
-0.4
The overshoot is still there, but the discrepancy is less approaching x = 0 as one would expect with the higher energy.
N.B.: Of course, one could just use polynomial to handle the transition regions. But, if were going to use numerical
methods, we might as well use the variational method instead of WKB. :P

HW6.nb

2. Classical limit of hydrogen atom

a) Correspondence between orbital motion and emitted photon
One can derive the Bohr model by taking the energy of an electron circling a proton (CGS units)
1
Z e2
E = 2 m v2 -

r
and applying the quantization condition
Ln = m v r = n @.
(This justification for this quantization is just units; we know that the ground state in fact has no angular momentum. One
can also derive the Bohr model by requiring whole wavelengths to fit on a circle of radius r .)
Substituting for v, we get
E = 2 m I
M -
.
mr
r
We require one more condition to eliminate r . Let us use the force balance
m v2
Z e2
F =
=

r
r2
which gives
Z e2
Z e2
mr 2
r =
=
H
L
m v2
m
n@
and finally
n2 @2
rn =
.
Z e2 m
Substituting this back in to the energy,
n@ 2

Z e2

En = 2
-
=
I
M - Z e2 I
M
m rn 2
rn
2m
n2 @2
n2 @2
to obtain
Z 2 e4 m
En = -
.
2 n2 @2
Similarly,
n@
n @ Z e2 m
Z e2
vn =
=
=
.
m rn
m n2 @2
n@
1

n2 @2

Z e2

n2 @2

Z e2 m

Z e2 m

Having derived these basic quantities, we can express the frequency of orbital motion in state n as
vn
1 Z e2 Z e2 m
1 Z 2 e4 m
o =
=
=
.
2 rn
2 n @ n2 @2
2 n3 @3
On the other hand, we know that the frequency of a photon due to a change from state n to n - k is
1
En -En-k
1
Z 2 e4 m
1
1
1
Z 2 e4 m i
zy
=
= -

I
-
M =
jj
2 - 1z .
2@
2
2 @3
n2
2 2 n2 @3
Hn-kL2
{
k I1- nk M
Approximating k ` n ,
1
Z 2 e4 m
2k
1 Z 2 e4 m
>
I1 +
- 1M = I
M HkL.
2 2 n2 @3
n
2 n3 @3
As indicated in the hint, we find that the frequency of the photon is an integer multiple of the frequency of orbital
motion o .

HW6.nb

b) Correspondence between classical radiation and mean lifetime

Under classical dynamics the electron is being continually accelerated, so it should emit radiation. To know the mean
lifetime of a state n, we must know how fast it loses power. Since the model is semiclassical, we can use a classical
1
radiation formula. We found above that the velocity goes like n , so the motion at large n is nonrelativistic and so we can
just use the Larmor power formula (again in CGS)
2 e2 a2
P =
.
3 c3
From above,
n
an =
=
=
I
M =
.
m
m rn 2
m
n2 @2
n4 @4
Substituting,

Z e2

Z e2

Z e2 m

Z 3 e6 m

2 Z 6 e14 m2

P =
.
3 c3 n8 @8
The mean lifetime should be
n
n-1
=
= I
M H1L
=
.
P
n3 @2
2 Z 6 e14 m2
Z 4 e10 m

E -E

3 c3 n8 @8

Z 2 e4 m

3 c3 n5 @6

Rearranging,
= 3
I
M

@c
@c
@
n5
2 e2

Z e2

m c2

as expected.

c) Comparison of classical and quantum lifetimes

We implement the correct expression above (with @ in units of eV s):
lifetimec@n_D =
2
mec2 1 y-1
i
j
j HZ L4 z
z . 8 137-1 , Z 1, mec2 511000, hbar 6.582 * 10-16 <;
hbar n5 {
k3

Computing for n = 2, 4, 6 we find respectively

82, 4, 6< lifetimec
82.98388 10-9 , 9.54841 10-8 , 7.25082 10-7 <

These are accurate to within an order of magnitude of the given quantum values, and seem to improve with increasing n as
we would expect.

HW6.nb

3. Spin precession in magnetic field

a) Time evolution of eigenstates
The Schrdinger equation is

i @ t \ = H \ = H-g
S BL \ = I-g
S M ] = - Sz ]
2mc
2mc z
e

eB

eB

where we defined = g
.
2mc

Then, for \ = Sz ; + \ we have

i @ t Sz ; + \ = -
Sz ; + ]
2
@

and for \ = Sz ; - \ we have

i @ t Sz ; - \ =
Sz ; - ].
2
@

So we have for the solutions

Sz ; + , t \ = e+i 2 t Sz ; + ]

Sz ; - , t \ = e-i 2 t Sz ; - ].

For a state aligned with the quantization axis, the "precession" is just a rotating phase.

b) Dynamics of Sx ; + \
First, define the spin matrices:

@
Sx = 880, 1<, 81, 0<<;
2
@
Sy = 880, -I<, 8I, 0<<;
2
@
Sz = 881, 0<, 80, -1<<;
2

Find the eigenstates of Sx :

eigenx = Eigensystem@Sx D
@
@
99- , =, 88-1, 1<, 81, 1<<=
2
2

Our state Sx ; + \ is then

Sx ; + \ =
!!!!! H Sz ; + \ + Sz ; - ] M
1
2

and so

Sx ; +, t \ =
!!!!! H Sz ; +, t \ + Sz ; - , t ] M.
1
2

In component form on the timeindependent basis 8 Sz ; + \, Sz ; - \<,

Our state Sx ; + \ is then

HW6.nb
and so

Sx ; + \ =
!!!!! H Sz ; + \ + Sz ; - ] M
1
2

10

Sx ; +, t \ =
!!!!! H Sz ; +, t \ + Sz ; - , t ] M.
1
2

In component form on the timeindependent basis 8 Sz ; + \, Sz ; - \<,

Sx ; +, t \ U

!!!!!
2

ij e+i 2 t
jj
j -i t
ke 2

yz
zz .
z
{

c) Showing precession
Let us implement the above state in component form, as well as its conjugatetranspose:
1

sxp@t_D =
tE=, 9ExpA-I
!!!! 99ExpAI
2
2
1

sxpct@t_D =
tE, ExpAI
!!!! 99ExpA-I
2
2

tE==;
2

tE==;
2

We calculate Z Sx ; +, t S Sx ; +, t ^ and put it into vector 8x, y, z< form:

ExpToTrig@8Hsxpct@tD.Sx .sxp@tDL@@1, 1DD,

Hsxpct@tD.Sy .sxp@tDL@@1, 1DD, Hsxpct@tD.Sz .sxp@tDL@@1, 1DD<D
1
1
9 @ Cos@t D, - @ Sin@t D, 0=
2
2
eB

This is the clockwise precession about the zaxis with angular frequency = g
of a vector of length 2 . As
2mc
expected, there is no zcomponent. (The classical analogue is a gyroscope precessing clockwise in the plane; of course,
here we cant simultaneously have components of x/y and z.)

4. HamiltonJacobi for light [optional]

a) Apply WKB approximation to equation of motion
Following the same steps we did for the Schrdinger equation, we first write the Maxwells equation with A0 = ei S :
2

2 y
ij n2 2
jj i S ji i S zy yzz i S ijj i S yzz 2 yzz i S
n2 i
zz ei S = ijjj
jj
z
j
=0 .

zz e

z
j
z
2
2
2
z
j c t

jc j k { z
k { {
{
k
{
k k
2

In the limit S p , we can drop terms of OHS L and keep those of OHS L :
2

S - J SN
c2
n2

= 0.

This is the "HamiltonJacobi" equation.

11

HW6.nb

(Of course, in the Maxwell equation, there is no notion of . What we are doing is a valid approximation when the variation
of the phase is very fast compared to the variation of the index of refraction. It is called "eikonal approximation" in optics.)

b) Equivalence to particle in given potential

The Hamiltonian of free particle in potential VIx M = -
nIx M is
2m

n2

H =
-
.
2m
2m
Applying this to the HamiltonJacobi equation, we obtain

J SN -
+
= 0.
2m
2m
t
1

n2

Since the action has no explicit time dependence, we can make the Legendre transformation SIt, x M = SIE, x M - E t to get
2

1
n2

J SN -
=E=0
2m
2m
and finally

n2 - J SN = 0 .
2

If n in part (a) has no time dependence, we can do the same Legendre transformation and obtain the equivalent "HJ"
equation

2
E
I

nM
J
SN = 0 .
c
2

c) Separate variables, integrate "action" variable

Assuming nIx M = nHxL, the "HamiltonJacobi" equation does not have an explicit dependence on y or t . Writing

SHx, y, tL = SHx, py , EL + py y - E t ,
~

the equation becomes

E2 - J
SN
c2
dx
n2

~ 2

- py 2 = 0.

Therefore,

= "##############################

E 2 - py 2
dx
c2
~

dS

n2

and hence

S=
~

nHxL
$%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

E2 - py 2 d x.
c2
2

Here, the possible xdependence of the index of refraction is emphasized.

12

HW6.nb

S=

nHxL
$%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

E2 - py 2 d x.
c2
2

Here, the possible xdependence of the index of refraction is emphasized.

d) Integral expressions for "angle" variables

~

, we find
Using t =
, y = -
E
py
t=

y=

nHxL2

c2

d x,
nHxL
$%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

E2 -py 2 %
c2
2

d x.
nHxL2
$%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

E2 -py 2 %
c2

e) Work to Snells law

Assuming that nHxL = n1 for x < 0 and nHxL = n2 for x > 0, and choosing the lower end of the integration at x = 0,
y
y< HxL =
x for x < 0

n1 2
$%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

E2 -py 2
c2

and

y
y> HxL =
x for x > 0.

n2 2
$%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

E2 -py 2
c2

Using the trigonometric relation sin =

!!!!!!!!!!!!!!!!!!!!!!!
! , we find
2
1

cot +1

cp

y
sin < = =
,
n1 E
n1 2 2

E -py 2
c2
&'''''''''''''''''''''''''''''''''''''''''

+1'
py 2

cp

y
sin > = =
,
n2 E
n2 2 2

E -py 2
c2
&'''''''''''''''''''''''''''''''''''''''''

+1'
py 2

and hence
sin

<
2

=
,
sin >
n1

which is nothing but the Snells law of refraction.