5 Time–dependence

5.1 Time–dependent Hamiltonians

Recall that for a system described by a Hamiltonian, Ĥ0 , which is time–independent, the most
general state of the system can be described by a wavefunction |Ψ, ti which can be expanded in
the energy eigenbasis {|ni} as follows:
X
|Ψ, ti = cn exp(−iEn t/h̄) |ni
n

where the coefficients, cn , are time-independent, and En denotes the eigenvalue corresponding to
the energy eigenstate |ni of Ĥ0 .
When we generalise to the case where the Hamiltonian is of the form

Ĥ = Ĥ0 + V̂ (t)

we can again expand in |ni, the time-independent eigenbasis of Ĥ0

X
|Ψ, ti = cn (t) exp(−iEn t/h̄) |ni
n

but the coefficients, cn , will now in general be time-dependent.

The wavefunction satisfies the time-dependent Schrödinger equation;
∂
ih̄ |Ψ, ti = Ĥ|Ψ, ti
∂t
so that we can substitute the expansion of |Ψ, ti to determine the equations satisfied by the
coefficients cn (t). Writing En = h̄ωn and denoting the time derivative of cn by ċn we obtain
X X
ih̄ (ċn − iωn cn ) exp(−iωn t)|ni = (cn h̄ωn + cn V̂ ) exp(−iωn t)|ni
n n

which simplifies immediately to give

X
(ih̄ċn − cn V̂ ) exp(−iωn t)|ni = 0
n

We now premultiply this equation with another eigenstate of Ĥ0 , hm|, to give
X
ih̄ċm exp(−iωm t) − cn Vmn exp(−iωn t) = 0
n

giving the following set of coupled, first–order differential equations for the coefficients:

P
ih̄ċm = n cn Vmn exp(iωmn t)

where ωmn = ωm − ωn and Vmn = hm|V̂ |ni.

This tells us how the coefficient cm varies with time, i.e. the probability that a measurement will
show the system to be in the mth eigenstate. It is exact, but not terribly useful because we must,
in general, solve an infinite set of coupled differential equations.
It is worth dwelling on the importance of the quantity Vmn . This ‘matrix element’ is an integral
which tells us how much the potential V̂ mixes states |mi and |ni. If it is zero (which it often is,
by symmetry) then V̂ cannot induce a transition between states |mi and |ni.
5.2 Time-dependent Perturbation Theory

Consider the Hamiltonian

Ĥ = Ĥ0 + V̂ (t)
where the time dependent part is small. We can write the time dependent coefficients cn

cn (t) = cn (0) + ∆cn (t)

Where cn (0) is the value of cn at t=0. We substitute in the equation for ċm derived above to give
X
ċm (t) = (ih̄)−1 [cn (0) + ∆cn (t)]Vmn exp(iωmn t)
n

We can assume that for a perturbation cn (0) >> ∆cn (t), and ignore the second term. This allows
us to obtain the coefficients cm (t) by integrating the first–order differential equation to give:
X Z t
cm (t) = (ih̄)−1 cn (0) Vmn exp(iωmn t) dt
n 0

In the special case where the system is known to be in an eigenstate of Ĥ0 , say |ki, at t = 0, then
ck (0) = 1 and all other cm (0) = 0, m 6= k, giving
Z t
−1
cm (t) = (ih̄) Vmk exp(iωmk t) dt
0

Thus a system starting in a known eigenstate of the unperturbed system may transform to a
different eigenstate through the action of the perturbing potential. Notice that cm (t) is an integral
over time, if we wait a long time, the transition may become more likely.
The probability of finding the system at a later time, t, in the state |mi where m 6= k is given by

pm (t) = |cm (t)|2

Since we have assumed a small perturbation, this result is only reliable if pm (t) ¿ 1. “Small” here
applies to both Vmk and its integral over time.

5.3 Time–independent Perturbations

The results obtained in the last section can also be applied to the case where the perturbation,
V̂ , is actually independent of time (strictly, ‘switched on’ at t=0).
Again, starting the system in eigenstate |ki of Ĥ0 we obtain,
Z t
cm (t)∆cm (t) = (ih̄)−1 Vmk exp(iωmk t) dt
0
Vmk
= [1 − exp(iωmk t)]
h̄ωmk
for m 6= k, giving for the transition probability
|Vmk |2 sin2 (ωmk t/2)
2
pm (t) = |∆cm (t)| = .
h̄2 (ωmk /2)2
For sufficiently large values of t, the function
sin2 (ωmk t/2)
f (t, ωmk ) ≡
(ωmk /2)2
 2 2 2 2
|cn1(t)| = Vn1t /h

ωn1−4π/t ωn1−2π/t ω ωn1+2π/t ωn1+4π/t

Figure 4: Transition probability as a function of applied harmonic perturbation frequency

consists essentially of a large peak, centred on ωmk = 0, of height t2 and width ≈ 4π/t, as
indicated in Fig. 4. Thus there is only a significant transition probability if Em ≈ Ek . That is, if
|ωmk | < 2π/t.
Note that we are assuming that the system was prepared in some eigenstate of Ĥ0 which is not
an eigenstate of V̂ : if it were, then the matrix element Vnm would be zero and pm (t) = 0. Thus
although the analysis treats the perturbation as time independent, it is applied to cases where the
perturbation is switched on at t = 0. Moreover only perturbations which are incompatible with
the Hamiltonian can induce transitions.

5.4 Harmonic Perturbation

This is generally useful since by Fourier analysis we can decompose any periodic perturbation into
harmonic components.
Let the perturbing potential be V (r, t) = V (r) cos ωt
If the initial state at t = 0 is k, and the final state m then

Z t à !
−i 1 Vmk ei(ωmk −ω)t − 1 ei(ωmk +ω)t − 1
cm ≈ Vmk eiωmk t (eiwt + e−iwt )dt = +
h̄ 0 2 2h̄ ωmk − ω ωmk + ω

where Vmk is the time independent part of the matrix element hm|V̂ |ki. This function is dominated
by the first term in the region around ωmk = ω, so we can consider only the first term to obtain
an estimate for the transition probability:

2
Vmk sin2 [(ωmk − ω)t/2] 1 2
|cm (t)|2 = = V f (t, ωmk − ω)
2
h̄ (ωmk − ω)2 4h̄2 mk

Where the function f is the same as we encountered earlier. Thus an external perturbation at a
given frequency most strongly induces transitions between energy levels separated by h̄ω.
This is another manifestation of an uncertainty principle. If the potential is electromagnetic, the
most probable transition is the absorption of a h̄ωmk photon as the system changes energy by
h̄ωmk . But if the transition happens very fast, the peak is broad and the photon could have a wide
range of energies, contrariwise, if the transition occurs after a long time the photon frequency is
well defined: ∆E∆t ≥ h̄/2. This uncertainty gives rise to the ‘natural linewidth’ of a particular
transition, and causes a limit to the accuracy of certain experiments. There is a slight difference
from the Heisenberg Uncertainty in non-relativistic quantum mechanics because time is not an
operator so one cannot define the commutator of time with the Hamiltonian.
³ ´
2
Note the extraordinary result that the transition probability at small times is Vmk /4h̄2 t2 . Con-
sider what happens if the state is measured frequently compared to if measurements are made
infrequently: frequent measurement tends to inhibit the transition!

5.5 Transitions to a group of states

We are often interested in the situation where transitions take place not to a single final state but
to a group, G, of final states with energy in some range about the initial state energy

Ek − ∆E ≤ Em ≤ Ek + ∆E

Then the total transition probability is obtained by summing the contributions of all the final
states. The number of final states in the interval between Em and Em + dEm is g(Em ) dEm , where
the function g(Em ) is known as the density of final states. The total transition probability for
transitions to G is then given by
1 Z Ek +∆E
pG (t) = 2 |Vmk |2 f (t, ωmk ) g(Em ) dEm .
h̄ Ek −∆E
For sufficiently large t, and ∆E À 2πh̄/t, we observe that essentially the only contributions to
the integral come from the energy range corresponding to the narrow central peak of the function
f (t, ωmk ). Within this range we can neglect the variation of g(Em ) and Vmk , which can therefore
be taken out of the integral to give
" # Z Ek +∆E
|Vmk |2
pG (t) = g(Em ) f (t, ωmk ) dEm .
h̄2 Em =Ek Ek −∆E

Furthermore, we can extend the limits on the integration to ±∞. Noting that dEm = h̄ dωmk and
using the result that
Z ∞
sin2 x
dx = π
−∞ x2
we obtain for the first–order transition probability
2πt h i
pG (t) = |Vmk |2 g(Em )
h̄ Em =Ek

The transition rate, R, is just the derivative of this with respect to t and is thus given by the
so–called Fermi Golden Rule:

2π h i
R= |Vmk |2 g(Em )
h̄ Em =Ek

The Fermi Golden Rule is probably the single most widely used result in quantum mechanics. The
factor of 2π
h̄
depends on the choice of perturbing potential, but the |Vn1 |2 g(Em ) term appears for
any applied perturbation. Be careful about the density of energy states - one sometimes encounters
density of frequency states (which differs by a factor of h̄) or of wavevector states.
It may appear that need to know the density of final states, g(Em ), but this is not always true.
In cases where |Vmk | = 0 transitions are forbidden, and in some cases we can deduce g(Em ) from
the relative rates of related transitions.
5.6 Example of Golden rule - beta decay

A nucleus decays via the reaction n → p e− ν. to form a electron and antineutrino, releasing
energy E0 .
The simplest form for the matrix element describing nuclear β-decay is given by the so-called Fermi
ansatz Vmk = GF M/Ω where Ω is the normalisation volume for the wavefunctions, |M |2 ≈ 1 is
the wavefunction overlap between initial and final nuclear states and GF is a constant.
We can work in the COM reference frame, so the kinetic energy of the nucleus is zero. Momentum
is conserved, so the final state has nuclear, electron and neutrino momentum P + p + q=0
while the energy released goes into the electron and neutrino, which for simplicity we treat as
massless: E0 = Ee + qc The proton and neutron are heavy compared with the electron and
neutrino. Given that momentum must be conserved, the kinetic energy must be concentrated in
the lighter particles.
The density of final states for the electron is given by the phase space volume
d3 pd3 r
dn =
(2πh̄)3
with a similar expression for the neutrino. Number of states in a volume of phase space is given by
the number of electron states, times the number of neutrino states, provided energy is conserved:
d3 pd3 r d3 qd3 r
dn = δ(Ee + qc − E0 )
(2πh̄)3 (2πh̄)3

dp E
Using the relativistic relation E 2 = p2 c2 + m2 c4 implies = 2
dE pc
R
the normalisation volume is just d3 r = Ω, and rotational invariance gives d3 p = 4πp2 dp.
All of which which simplifies the integral to
Ω2 q
dn = 4 6 6 Ee Ee2 − m2e c4 Eν2 δ(Ee + Eν − E0 )dEe dEν
4π h̄ c
where Ee is the electron energy and Eν is the neutrino energy. What can actually be measured is
the electron energy, so we integrate over the neutrino energies,
dn Ω2 q
= 4 6 6 Ee Ee2 − m2e c4 (E0 − Ee )2
dEe 4π h̄ c
This is the distribution of electron energies from beta decay: the rate fo emission of electrons at
a particular energy is given by the Golden Rule
2π G2F M 2 q 2 2 4 2
R= 6 6 Ee Ee − me c (E0 − Ee )
h̄ 4π h̄ c
4

Figure shows the simplest case of beta-emission: neutron

decay. Conservation laws tell us that the electron energy
must lie between its rest mass (0.51MeV) and the total
energy available (0.7823MeV). But the entire shape can be
deduced from geometry.