Sie sind auf Seite 1von 8

Lecture 12: Time-dependent perturbation

theory: emission and


absorption of radiation (10/27/2005)
In the previous lecture we have seen that physical systems may be tickled
by perturbations that oscillate with time. These harmonic perturbations
were capable to change one energy eigenstate into another energy eigenstate,
especially if the frequency of the perturbation is chosen to be close to ba .
For molecules (MASERs) or atoms, the external perturbation stimulates the
particle to change its state. We talk about stimulated emission or absorption
of energy.
Additional reading if you wish: Griffiths ch. 9.3
However, the excited states of the hydrogen atom (or any other atom or
molecule, for that matter) usually decays even in the absence of an external
perturbation. This seems to violate our rule that the energy eigenstates
should be exactly stationary. How do the atoms manage to fall into their
ground states? We will follow Albert Einstein and others.

Einsteins insights from 1917


One year after general relativity was completed, when Vladimir Lenin was
finalizing his plans for the Great October Revolution back in 1917, Albert
Einstein had very different problems. Eight years before quantum mechanics
was born, Einstein managed to discover the stimulated emission and its relationship with stimulated absorption and spontaneous emission. What was
his argument?
Consider a collection of atoms. Each of them may be in one of two states,
|ai and |bi. The total numbers of atoms in these two states are denoted as
Na and Nb , respectively. Let us assume Eb > Ea . Let A be the rate (the
probability per atom and per unit time) of spontaneous emission of radiation:
|bi |ai |
hi
where you may replace by + if the equation looks too confusing to you;
the same holds for similar equations in the future. Also, let Bba (ba ) be the
1

rate of stimulated emission. You can see that this rate is proportional to the
energy density of radiation per unit frequency (its density both in space
as well as in the spectrum of energies). This proportionality is what we have
seen: the probabilities were proportional to |Vba |2 and the energy density will
also be proportional to the squared perturbations the squared electric field,
for example. Once again, Bba (ba ) is the rate of stimulated emission
|bi |n
hi |ai |(n + 1)
hi.
In the very same way, Bab (ba ) is the rate of stimulated absorption:
|ai |n
hi |bi |(n 1)
hi.
Note that there cant be any spontaneous absorption if there is nothing to
absorb. Finally, we are getting to an equation. The atoms are in equilibrium
which implies
0=

dNb
= Nb A Nb Bba (ba ) + Na Bab (ba ).
dt

On the other hand, the Maxwell-Boltzmann distribution gives us


eEa /kT
Na
= E /kT = e(Eb Ea )/kT = ehba /kT .
Nb
e b
You may now eliminate Na from the differential equation:


0 = Nb A Bba (ba ) + Bab (ba )ehba /kT

And Einstein thus learned that

(ba ) =

A
eh/kT Bab

Bba

Another thing he had to figure out was that Bab = Bba . It follows from
Plancks formula (derived at the beginning of the course)
h
3
() = 2 3 h/kT
c (e
1)
because the ratios of the two terms agree for both denominators if Bab = Bba .
Finally, by matching the two equations for (), we may see that Einsteins
A, B coefficients are
Bab = Bba ,

A=

3
h
ba
Bba .
2 c3

Calculating Bba
We want to calculate Bba , the coefficient of the stimulated emission, for the
case of thermal environment with isotropic, uniform, incoherent, and unpolarized photons. More precisely, we want to calculate the leading contribution to Bba . Despite our goal, we must start with a monochromatic, linearly
polarized electromagnetic wave
~ r , t) = E0  cos(~k ~r t)
E(~

~ r , t) = E0 k  cos(~k ~r t)
B(~
c
where is the frequency, k = c is the momentum, k is its direction (unit
vector), and  is the polarization direction (for the electric field). Draw
a simple figure. How does this field act on charged particles? Recall the
Lorentz force law
F~ = e(E + ~v B)
which means that the magnetic force will be v/c times the electric force.
Recall that e < 0 for electrons. In physics of atoms we consider, |v/c|  1,
which is why we may use non-relativistic physics and neglect the magnetic
force. Positively speaking, we only want to consider the first-order effect of
the electric field.
The wavelength of the relevant electromagnetic waves we consider
=

2
k

will usually be longer than 100 nm which is much more than 0.1 nm, the
typical size of the atom. This allows us to assume that the electric field is
uniform over the atom. If we locate the (nucleus of the) atom at ~r = 0, the
previous formula for the electric field becomes
~ r , t) = E0  cos(~k ~r t) = E0  cos(t)
E(~
The Lorentz force is then simply
F~ = eE0  cos(t).
It may also be written in terms of the potential energy:
~
F~ = V

V = eE~ ~r

Vba = eE~ hb|~r|ai.

This is known as the dipole approximation. Define the matrix elements of


the dipole operator as
~ ba = ehb|~r|ai
P

~
Vba = E~ P.
3

For a given electric field, this can be plugged as the time-dependent perturbation of the Hamiltonian into the calculations from the previous lecture.
Because
~ 2
E~ = |E0| |Vba |2 = E02 |
 P|
and our first-order perturbative calculations give us
Pba

2
E02
2 sin [(ba )t/2]
~
 P|
= 2 |
(ba )2
h

Averaging
So far we considered a monochromatic, linearly polarized wave moving in a
specific direction. But we want to know how the atom responds to wideband, unpolarized, isotropic radiation because only in this case Einsteins
formula applies as originally written down. In other words, we must try to
average over directions, frequencies, polarizations, and so forth.
Let us start with the frequencies. The time-averaged energy density is
R
()d, the integrated contribution from all frequencies; and we eventually
want to write our results using (). But the time-averaged energy density
may also be written as the time average of
1
1 2
= 0 E 2 +
B .
2
20
The electric and magnetic components contribute the same amount. Therefore, we need the average of 0 E 2 . For an oscillating E, it is simply
1
0 E 2 =
2 0

()d

where the factor 1/2 appeared as the time-average of cos2 (t). The averaging of our probability formula over time and frequencies may therefore be
obtained by replacing E02 according to the identity above, leading us to

Pba

~ 2Z
sin2 [(ba )t/2]
2|
 P|
()
=
d
(ba )2
0
0h
2

Let us now assume that () does not depend on too much; it is slowly
varying. Only the frequencies near = ba will contribute significantly, as
we will see again, so you just require () to be slowly varying in this region.
Not a big assumption. But it allows you to replace () by (ba ) which
means
Z
~ 2
2|
 P|
sin2 [(ba )t/2]
Pba =
(
)
d
ba
(ba )2
0
0 h
2
The integral may be calculated easily. Let us switch our variables
x=

( ba )t
2

2
d = dx
t
4

and slightly extend the limits for x to (, ) (this wont bring any factor
of two because x could have both signs anyway). We get:

Pba

Z
~ 2
|
 P|
sin2 x
dx
(
)t
ba
x2
0 h
2

The integral over x equals as you can see using the residue techniques if
you wish. Replace the integral by and the rate i.e. the probability per unit
time is obtained from the previous formula by omitting t:
Rba =

~ 2 (ba ).
|
 P|
0 h
2

We also need to average over , more precisely over the relative orientation
~ which is random. Note that
between  and P
~ 2 = |
~ 2 cos2 ,
|
 P|
|2 |P|
and
2

hcos i =

|
| = 1

+1 2
z dz
d cos2
1
R
= 1
=
R +1
d
3
1 dz

With this result, the rate (probability per unit time) of transition for unpolarized, incoherent, isotropic radiation is given by
Rba =

~ 2
2 |P| (ba ).
30h

This formula is a special example of Fermis golden rule we will discuss later.
Recall that this quantity is what we used to call Bba (ba ) which implies
Bba =

~ 2 = Bab ,
|P|
30h
2

A=

3
3 ~ 2
h
ba
ba
|P|
B
=
ba
2
3
c
30 h
c3

Spontaneous emission: lifetime


Imagine that you have Nb atoms in the excited state |bi and you want to find
their time evolution. You see that
dNb
= A dt
Nb
ln Nb = At + ln Nb (0) Nb = Nb (0)eAt

dNb = Nb A dt

For a single atom that started as |bi, you may also say that the probabilities
will go like
Pb (t) = exp(At),

Pa (t) = 1 exp(At).

You may also define the lifetime


= 1/A

Pb (t) = exp(t/ ).
5

We have discussed a two-level system. If we consider a more general case in


which |ki may decay to many final states |ni for different values of n, the
total spontaneous decay rate is simply the sum
A=

Akn

1
.
n Akn

=P

The matrix elements of the dipole operator matter


The coefficients we have found depended on
~ ehb|~r|ai
P
where e is negative and where we would have to sum ~r over all electrons if
there were many. The fate of the atom depends on some matrix elements of
the position operator. Sometimes they are zero, sometimes they are not. We
want to understand these issues. Note that
~ 2 = e2 |hb|~r|ai|2 = e2 hb|~r|ai hb|~r|ai
|P|
= e2

|hb|x|ai|2 + |hb|y|ai|2 + |hb|z|ai|2

Here we must really sum these three terms. If you have a feeling that this
is like taking three times the squared dipole because only the dipole in the
direction relevant for a particular wave should be included, recall that we
have already included the factor 1/3 from the averaging over the angles, and
therefore our full formula agrees with your intuition. We may also express
the squared dipole using the spherical tensors
m=q
Tq k = Yl=k
(~r).

In terms of the components, the relevant ones have k = l = 1 and they are
T+11 = 12 (x + iy)
T01 = z
T11 = + 12 (x iy)

x = 12 (T11 T+11 )
y = i2 (T11 + T+11 )

z = T 1
0

Inserting these formulae for x, y, z into our formula for the dipole, we obtain
1
1
|hb|(T11 T+11 )|ai|2 + |hb|(T11 + T+11 )|ai|2 + |hb|T01 |ai|2
2
2
h
i
2
1
2
1
2
= e |hb|T1 |ai| + |hb|T0 |ai| + |hb|T+11 |ai|2

~ 2 = e2
|P|

Selection rules
We want to identify some necessary assumptions for the transition to occur.
Recall that the Wigner-Eckart theorem implies that
0

hn0 l0 m0 |Tq k |nlmi = clkl


mqm0

hn0 l0 ||T 0||nli

2l + 1
6

We see that our matrix elements depend on the Clebsch-Gordan coefficients


and they are only nonzero if
(l = l0

or l = l0 1) and (m = m0

or m = m0 1)

Moreover, the transitions between


(l = m = 0) and (l0 = m0 = 0)
are forbidden, too. By these statements we mean that if the initial state
|bi and the final state |ai have quantum numbers (l0 , m0 ) and (l, m) that do
~ 2 simply vanish and no
not satisfy the rules above, all contributions to |P|
transitions occur.

Parity
There is one more selection rule due to parity. It is the operator defined by
r)i = |(~r)i.
|(~
The operator squares to one which also means that its eigenvalues must be
1. Its commutation properties with ~r, ~p are given by
r = ~r

~
p = ~p

~r} = 0
{,
~p} = 0
{,

where the second line follows from the first since ~p = d~r/dt. However, two
minus signs cancel for
~ = ~r p~,
L

L
~ = +L
~

L]
~ = 0.
[,

Lz , L2 ,
all commute with each other, we can simultaneously
Because H,
diagonalize all of them. This also means that the spherical harmonics must
acts as
be eigenvalues of the parity operator. In spherical coordinates,
:

and by looking at the form of the functions, you will see that
m = (1)l Y m .
Y
l
l
Because parity is a hermitian operator,
ha|bi = 0
if |ai, |bi have different parity. The proof is simple,
= a ha|bi = b ha|bi = 0
ha||bi

for a 6= b .
7

This also means that


hn0 l0 m0 |~r|nlmi = 0
if |n0 l0 m0 i and ~r|nlmi have a different parity. In other words, the matrix
element vanishes if |n0 l0 m0 i and |nlmi have the same parity because ~r changes
the parity. This means that l = l0 is actually forbidden by parity conservation
and the allowed transitions must have
(l = l0 1) and (m = m0

or m = m0 1)

The subleading contributions beyond the simple dipole approximation we


made actually allow some other transition occur at well, albeit at much
smaller rates, which means that some of the so-called forbidden transitions
violating the condition above are not completely forbidden.

Magnetic dipole and electric quadrupole


We made the assumption that the electromagnetic field is essentially constant
over the atom. If we include its gradient, there will be other matrix elements
that will contribute to the transitions, although with smaller rates, especially:
~ |nlmi
# hn0 l0 m0 | L
# hn0 l0 m0 |Tq 2 |nlmi

(magnetic dipole)
(electric quadrupole)

The magnetic dipole transitions have selection rules


l = 0,

m = 0, 1,

l = l0 = 0 forbidden

while the electric quadrupole moment have selection rules


l = 0, 2

m = 0, 1, 2,

l = l0 = 0 forbidden

Note that in all three types of emissions, transitions with l = l0 = 0 were


forbidden. This means that the 2s state of the Hydrogen atom is stable.
Well, not quite. It is metastable because it can decay via multi-photon
emissions that we have neglected; it can also fall to 1s by colissions. Note
that the angular momentum and parity selection rules do not restrict n, n0
in any new way. In the case of the Hydrogen atom, it means that all n, n0
transitions are allowed which gives us the Balmer, Lyman, and other series
of the spectral lines.

Das könnte Ihnen auch gefallen