Lecture 14: Time-Dependent Perturbation Theory: Correspondence Principle + Sudden Approximation (11/10/2005)
Lecture 14: Time-Dependent Perturbation Theory: Correspondence Principle + Sudden Approximation (11/10/2005)
x =
1
(
a+ + a )
2m
q
q
h
~ = hn0 |(
P
a+ + a
)|ni = (n0 ,n+1 n + 1 + n0 ,n1 n)
2m
2m
The two terms are responsible for absorption (n of the harmonic oscillator goes up) or emission (n
goes down), respectively. You see that the matrix element of the dipole is only nonzero if n0 = n 1.
It means that we can only emit or absorb a single frequency in this dipole radiation approximation
h
rad = En0 En = h
(n0 + 1/2) h
(n + 1/2) =
h,
namely the frequency of the harmonic oscillator itself. We have calculated the rate A of the spontaneous emission in two different ways lectures 12 and 14. The general result can be specialized to
our case of the harmonic oscillator with n = n 1:
Rn,n1
3
3P 2
=
=
30h
c3
30h
c3
q 2h
n
2m
q 2 n 2
.
60 mc3
This is the probability per unit time, and therefore the average energy per unit time is
P = Rn,n1h
=
q22
q22
(n
h
)
=
(E 0.5
h)
60 mc3
60 mc3
1
If we want to compare it with the classical Larmor formula, we also need to know the acceleration
a. A harmonic oscillator has x = x0 cos(t) which implies a = 2 x. Therefore a2 = x20 4 cos2 (t).
If we average the cos over one period, we finally obtain
1
ha2 i = x20 4
2
and because E = m 2 x20 /2 the maximal potential energy when the kinetic energy is zero it can
also be written as
E 2
ha2 i =
m
2
If we now replace the factors E /m in our quantum mechanical result for P by a2 and neglect the
0.5
h term (the ground state energy, negligible compared to the total energy in our approximation),
we obtain our classical Larmor formula. You can always try these consistency checks that your
quantum theory gives you the expected classical results in the appropriate limit.
X
n
cn |ni
expressed in terms of the past energy eigenstates, and we want to find the final state
| 0 i =
X
n0
dn0 |n0 i
as a superposition of the future energy eigenstates. The simplest, sudden case is when H0 changes
to H00 instantaneously. Note that the Hamiltonian is still finite at all times, and the equation
i
h
(t)
= H(t)(t)
t
therefore implies that (t) must still be continuous because its derivative is always finite. This just
means that at the moment of the transition that we call t = 0 we have
X
n
cn |ni =
X
n0
dn0 |n0 i
cn hk 0 |ni =
X
n0
X
n
cn hk 0 |ni
is known as the sudden approximation. Is it always valid? Well, it is valid if the initial, finite-energy
state remains a finite-energy state even after the change. In other words, the space generated by
|n0 i must contain the space generated by |ni. For example, this is the case if a particle is in the box
and the box suddenly expands (energy mostly decreases because the infinite energy from the walls
partially disappears); in that case, the sudden approximation works. In the opposite case when
the box shrinks, the sudden approximation fails because the particle is generally in the region that
becomes forbidden after the transition at t = 0, and therefore it has suddenly a huge energy and
many processes are too fast and make the sudden approximation invalid.
Pn
i
X
n
cn hi|ni
X
n0
X
i
XX
i
Compare it with the previous, simpler, sudden formula for dk0 . Its clear that if all the phases
can be replaced by one, then the summation over i of |iihi| is nothing else than the completeness
relation, and we are back to the sudden approximation. The phases are equal to one if
En0 Ei
3
for all pairs of eigenvalues. This condition cant be satisfied literally because En0 can become
arbitrarily large, making the condition very difficult. However, these high-energy states have very
small probability amplitudes, and a reasonable replacement for the previous inequality is therefore
1
p2
+ m 2 x2 ,
2m 2
H00 =
p2
1
p2
1
1
+ m 2 x2 qxE0 =
+ m 2 (x a)2 m 2 a2
2m 2
2m 2
2
qE0
.
m 2
Note that the final Hamiltonian describes another harmonic oscillator with the same frequency but
a displaced origin. The eigenvalues of H00 are therefore simply
1
(n0 + 1/2) m 2 a2 .
En0 = h
2
In the sudden approximation therefore
d n0 =
n 0 (x a)0 (x)dx
This is the type of integrals that Roy Glauber had to master long before he could get a Nobel prize
for the coherent states. The integral may be evaluated in closed form:
0
2 a2 /4
(a)n e
0
d n0 =
2n n0 !
where =
m
h
The probability that we end up at n0 -th excited state is simply obtained by squaring this amplitude:
0
(
n)n en
Pn0 = |dn0 | =
n0 !
2
1
q 2 E02
where n
= (a)2 =
.
2
2m
h 3
Note that if you sum Pn0 over n0 , you obtain one because the exponential cancels against the opposite
exponential arising as a Taylor expansion. In fact, you know this sum. If you think for a while, you
realize that Pn0 defines the Poisson distribution with the mean n
.
If you turn on the electric field in classical physics, you add some potential energy, but because
of symmetry, it is equally likely that it is positive or negative: classically you dont add any energy
by turning on the field. What about quantum mechanics? Clearly, we dont have an eigenstate
anymore, but we may still evaluate the average energy. The average excitation energy is
hEi =
X
n0
ma2 h
m 2 a2
h
=
=
2
h
2
2
m 2 a2 m 2 a2
= hH0 i
2
2
So quantum mechanics reproduces our classical result for the average energies: they dont change
because of the symmetry.
While we were able to get the exact result for the sudden change of the Hamiltonian, for arbitrarily
strong electric fields, it is still interesting to compare the result with the first order perturbation
theory. If the electric field is weak enough more precisely if n
1 we can try to use the first
order perturbation theory, too. Again, the dipole term will only be able to change n by one and
the sudden approximation gave us
hH00 i = hH0 i +
P1 =
n
en
q 2 E02
n
=
1
2m
h 3
for n
1
qE0
h
= qE0 h1|
x|0i = h1|
a+ |0i = qE0
2m
2m
Substituting this constant to the formula for c1 above and by squaring it, we obtain
P01 (t) =
4q 2 E02
sin2 (t/2)
3
2m
h
For t > 0, this time-dependence is of course identical to the usual oscillations of the two-level
systems; see lecture 11. But you should ask why did we obtain a time-dependence result at all?
Note that the sudden approximation gave a constant. You should see that if you replace sin2 (t/2)
by its average value 1/2, you reduce the first-order calculation to the sudden approximation result.
The only reason why the probabilities in the first-order approximation oscillate is that they continue
to refer to the probability of eigenstates of H0 which differ from the eigenstates of H00 that become
more relevant for t > 0.
H 3 He + e + e .
One of the neutrons inside the tritium decays into a proton, an electron, and an antineutrino (that
will invisibly escape). Because it is so hard to observe the antineutrino, we would like to determine
its properties especially its rest mass indirectly. One method is to measure the electron that
orbits the Hydrogen nucleus at the beginning. It will probably stay in the atom and orbit the helium
nucleus afterwards. The other electron created by the reaction above is very energetic its kinetic
energy is like 17 keV and it leaves the atom almost instantly, compared to the slow timescales of
5
the atom. This is why it is legitimate to use the sudden approximation for the peaceful electron
that orbits the atom. The nuclear energy are much greater than atomic (electronic) energies and
the corresponding processes are much faster.
The peaceful electron starts in the (nlm) = (100) ground state of the Hydrogen atom. Suddenly,
the (very heavy) Hydrogen atom abruptly changes to a Helium ion. It means that the corresponding
Hamiltonians are
h
2 2
e2
h
2 2
2e2
0
H0 =
,
H0 =
.
2m
40 r
2m
40 r
Our general derivations of the sudden approximation tell us that the amplitudes to appear in a
particular Helium state are
dn0 l0 m0 = Z=2 hn0 l0 m0 |100iZ=1
You can see that the inner product vanishes unless the angular momentum is conserved. Consequently, l0 = m0 = 0. The only candidates are these states (n0 , 0, 0) and the squared inner products
can again be calculated in closed form:
Pn0 00 =
2
(Z=2)
(Z=1)
drRn0 0 (r)R10 (r)
=
=
=
=
..
.
0.702
0.250
0.013
0.004
Pion = 1
Although it looks like a simple calculation, and indeed, in the zeroth approximation it agrees with
experiments, people tried to determine the mass of the neutrino by a precise analysis of some
corrections to this simple story. The first result they obtained from the observed excess of electrons
near the endpoint of their energy spectrum was a shock: the squared neutrino mass was negative;
the neutrino looked like a tachyon, a particle faster than light, as argued by J. Ciborowski and J.
Rembielinski in 1998 and others. Obviously, according to many of us, there had to be some bug
with their experiment or calculation, but I cant tell you what exactly happened later and whether
the discrepancy with special relativity is already fixed.