Lecture 13: Time-dependent perturbation theory:

photoelectric effect and Fermis golden rule (11/1/2005)

Last time we discussed spontaneous and stimulated emission and stimulated absorption and their
relationships, using the insights from time-dependent perturbation theory. We also derived the rate
of the emission of electric dipole radiation.
Additional reading if you wish: Bransden and Joachain, ch. 11.7
At the beginning of our time-dependent adventures, we discussed the transition between two
states or a small number of states. But more typically, the number of states between which we find
transitions is infinite; the energy of the free particles is continuous, for example. This requires some
modificiations in our formulae. We will see what Fermis golden rule is and how these ideas are used
in a detailed analysis of the photoelectric effect.

Approximating the frequency-dependence by a delta-function

Consider the case of the harmonic perturbation
H 0 = V cos(t),

0
Hni
= hn|V |ii cos(t) = Vni cos(t)

A week ago, we derived that the first-order probability for the initial state |ii to change into one of
the final states |ni is
|Vni|2 sin2 [(ni )t/2]
Pin =
(ni )2
h
2

As time t , the second factor becomes strongly peaked around ni = . Using the result of the
integral
Z
sin2 x
dx =
x2
calculated by the residue techniques and the substitution x = ( ni )t/2, dx = (t/2)d, we may
also see that
Z
t
sin2 [(ni )t/2]
d
=
( ni)2
2

Because the integral becomes concentrated near = ni for large t, we can also generalize the
formula above to
Z
t
sin2 [(ni )t/2]
d
=
f (ni)
f ()
( ni )2
2

The result only depends on the value of f () at the right point which is allowed assuming that f
changes slowly enough, namely if
df
t
d

sin2 [(ni )t/2]

t
= ( ni )
2
( ni )
2

Such a new form of the ratio allows us to write

Pin =

|Vni |2
t( ni )
2 h
2

Does it make sense?

In other words, as time t goes to infinity, the transition is only allowed if the photons or other quanta
of energy have the right value of the energy required by energy conservation. But the formula has
two obvious problems
the probability can become infinite which seems unphysical (=physically meaningless)
the probability does not describe flopping we studied previously
However, both of these apparent problems evaporate if we the final states |ni form a continuum
because:
the total probabilities of well-defined outcomes will be obtained as integrals of this delta
function and will be finite
if the final states form a continuum, it is possible for the particles to escape to infinity and no
flopping occurs
How does it work quantitatively? Denote the number of states per unit energy as n (En ). To
find the probability of transition to any of these states, simply integrate the previous formula:
Pin

(any)

|Vni |2
t
=
2 h
2

n (En )( ni )dEn

In the integral we must change the variables of the delta-function to energy, gaining a factor of h
,
and the integral equals h
n (En ) where En is the energy of the correct final state that preserves
the energy. We therefore have
|Vni|2
Pin (t) =
n (En )t
2 h

Fermis golden rule

Let us make a final step and erase the factor of t, to find the rate of the transitions i.e. the number
of transitions per unit time (if you multiply it by the number of atoms you have):
Rni =

|Vni|2 n (E).
2
h

This simple formula is called Fermis golden rule and it is useful for problems involving a continuum of states. Note that it resulted from the first-order perturbation theory; when we deal with a
continuum, the reliability of perturbation theory also increases and most of the refinements we have
seen for the two-level systems become unimportant.
More typically, you will see the rule written in the following form
Rni =

2 0 2
|Hni | n (E)
h

in which the general time-dependence was recovered, gaining (surprisingly) a factor of four.

Photoelectric effect
Einstein received his 1921 Nobel prize for an explanation of this effect. Recall that the energy of
outgoing electrons depends on the frequency, not intensity, of the incident light because the situation
may be described as absorption of the photons whose energy is h
by the electrons and one must
use the energy conservation.
Let us now study this effect in detail using the time-dependent perturbation theory. The kick
of the photon only affects one electron. Let us choose the electron in the k-shell (klosest to the
nucleus) of the atom. This electron sees the full charge Ze of the nucleus and the other electrons do
not play much role. We are back to a Hydrogen-like problem. The initial state of our electron is
|ii =

Z3
a3

!1/2

exp(Zr/a),

40 h
2
.
a=
me2

Much like in the previous lecture, we are considering the transitions between two electronic states.
However, in the present case, we choose the final state to be a free electron far away from the nucleus.
It may be represented by a plane wave and these plane waves are normalized to a delta-function.
~

|ki =

eik~r
,
(2)3/2

hk 0 |ki = (3) (k k 0 )

Despite this change towards the continuum, the arguments work much like before: the leading
transitions are determined by the electric dipole effect interacting with the electric field of the
electromagnetic wave:
~ 2 = . . . = |
~ 2 E~02 = |P|
~ 2 E~02 cos2
|Vni|2 = |E~ P|
 P|
where is the angle between the matrix element of the electric dipole and the polarization vector .
Thus, using our first Fermis golden rule, we have
~ 2 ~2
Rni =
|P| E0 cos2 n (En )
2
h
What is the density of states? For a free particle, it is an easy question:
2

(E)dE = k dk d;

h
2k2
h
2k
p2
=
dE =
dk
E=
2m
2m
m

(E) =

mk 2
d.
h
2

Substituting this to our rule, we have

Rni =

mkn ~ 2 E~02
cos2 d
|P|
2
h
3

Cross sections
When we talk about the probabilities of transition in which the initial state contains two objects
for example, a beam of particles of one type directed to particles of other types a useful concept
is cross section. A cross section may be thought of as the area of a single target that one must hit
in order for a particular reaction to take place.
Take this definition and imagine that a layer of thickness d of atoms whose total number density
is Ntarget /V is being bombed by Nbeam /t particles per unit time. The number of interactions per
unit time is then, following our definition of the cross section, equal to
Rni =

Nbeam
N
Ntarget
=

d
t
t
V


Rni
(Nbeam /t)(Ntarget /V )d

To relate these formulae to our previous form of Rni , notice that the energy density may again be
written in two different ways:
1
/V
energy density = 0 E02 = Nbeam h
2
We dropped an index on . Also realize that in the previous section, our target was only made of
one atom
Ntarget = 1
and the speed of light relates the thickness with the time in which the interactions occur:
d
=c
t
With these substitutions, we obtain
=

Rni
(1/2)0 E02 c/
h

Using our previous form of Rni , our cross section is therefore

=

mkn
~ 2 d
cos2 |P|
2
h
c0

d
mkn
~ 2
cos2 |P|
= 2
d
h
c0

Switching to the momentum form

Up to some simple factors, we have expressed the cross section using the matrix element of the dipole
moment. Now we want to calculate this matrix element of the electric dipole moment between our
initial and final states:
~ = ehn|~r|ii
P

Actually, it will be helpful to express this matrix element in a form that superficially looks more
complicated, by inserting the Hamiltonian as follows:
e
~ =
0~r ~rH
0 |ii
P
hn|H
En Ei

Because

0 = h
2 + V (~r)
H
2m

and

we have

[V (~r), ~r] = 0,

e
h
2

hn|2~r ~r2 |ii

En Ei
2m
The commutator is easy to evaluate
!

~ =
P

~
2~r ~r2 = 2

h
~ = e
~
P
hn||ii
m

~ because ~p = i
The last equation is the momentum form of P
h. This form, as we have said,
is actually simpler for evaluation of the matrix elements. Finally, let us make the calculation of the
matrix element for our case:
h
~ = e
P
m

Z

dr

~ n (~r)
i (~r)

e
h
=
m
= +i

e
h
m
4

Z3
a3

!1/2

1
~ i~kn r d3 r =
eZr/a e
3/2
(2)
!1/2
~kn Z
Z3
~
eZr/a eikn r d3 r
a3
(2)3/2
Z

The integral is a standard Fourier transform:

Z

eZr/a eikn r d3 r =

8(Z/a)
[(Z/a)2 + kn2 ]2

It can be calculated if you realize that the Laplacian has a pretty simple action on eZr/a . Substi~ we obtain
tuting it to P,
 5/2
~kn
ie
h
2
2 Z
~ =
P
mn
a
[(Z/a)2 + kn2 ]2

Simplifying the cross section at high energies

Let us simplify this expression assuming that the energy of the incoming photon and the outgoing
electron is still non-relativistic but actually much higher than the binding energy of the atom.
Actually, we have already made this approximation when we pretended that the final state (the
plane wave) was an energy eigenstate. For
Z
kn2 
a


2

h
2 kn2
h

2m

Finally,
d
=
d

e2
40 h
c

32
h Z
m kn a


5

cos2 =

k=

32
h Z
m kn a


2m
h

5

cos2

where we introduced = 1/137.036..., the fine-structure constant that determines the squared
charge of the electron in natural units. It is illuminating to express the cross section as a multiple
of the squared Bohr radius
h

40 h
2
=
a=
2
me
mc
in the following way:
!
8 5 2 mc2 7/2
d
= 4 2 Z a
cos2
d
h

The angle between the (electric) polarization direction  and the direction of the outgoing electron
is not observable. We should try to switch to the angle between the incident photon and the
outgoing electron. Draw a picture one that should also imply
cos2 = sin2 cos2
Averaging this over the angle gives
cos2 =
This leads us to

1 2
sin
2

d
mc2
= 2 28 Z 5 a2
d
h

!7/2

sin2

The electron is usually emitted nearly = 90 degrees to the photon, because of the transverse
electric field. Note that all factors except for a2 are dimensionless and a2 has the correct dimension
of an area. You could expect that is O(a2 ) but dont forget that 8 8 1018 .
5

Graphs and errors in the periodic table

On the blackboard, I should be able to draw a graph showing how the cross section more precisely,
the absorption coefficient N/ for platinum depends on the wavelength in
A of the radiation
that must be in the X-ray region to eject the inner electrons. You should notice that there are sharp
edges. They correspond to ejecting of the electrons with energies
E = 13.6

(Z Sn )2
eV
n2

where Sn is an effective shielding by inner electrons. There is an edge for ejecting the k-shell
electrons; several edges for l-shell electrons, and so forth. In 1913, years before our derivations could
have been made, H. G. J. Moseley plotted the energy of k- and l-shell edges and showed that

EZ
This allowed him to determine Z of many elements he analyzed. Consequently, he could have
corrected some mistakes in the periodic table of that time. After the errors are fixed, the graph
looks very clean, and it should be drawn on the blackboard right now. Below, you find space to
draw the two graphs, too.