Quick guide to perturbation theory

Charles W. Clark
May 13, 2015
Why bother with perturbation theory ?
Perturbation theory is an essential tool for solving real-world problems.
There are no simple solutions to real-world problems
Just consider the motion of the Earth about the Sun. To a reasonable degree of approximation,
this can be described by two point particles interacting via the inverse-square law gravitational
force, which is a problem that we can solve exactly (and did so in Lecture 9).
However, it is nowhere close to describing the motion of the Earth as actually observed within
the accuracy of our measurements, due to the presence of other massive bodies in the Solar System.
There is no known exact solution to the equations of motion that accounts for even the effects of
the Moon, alone.
Thats no reason not to try !
Perturbation theory gives us a way of starting from an idealized model of a physical system
and and treating non-ideal aspects in a systematic way. An example from my own research work
is the calculation of the effects of temperature on the ticking rate of atomic clocks. These effects
are small, maybe affecting the ticking rate in the 17th digit. However they are among the largest
effects that limit our ability to measure time, at the current limits of precision. There is no way
to compute to 17 digits of accuracy any property of an atom like strontium, with its 38 electrons.
Perturbation theory allows us to separate off the effects of temperature dependence (which are due
to the irreducible presence of black-body radiation), and compute them directly.
The standard approach as introduced by Schrdinger
In 1926, Erwin Schrdinger introduced perturbation theory into quantum mechanics in order to
treat the effect of external electric fields on the Balmer series of spectral lines of atomic hydrogen.
His formalism remains in standard use today, and here is how we portray it in Lecture 13.
0 , for which we can calculate the eigenenergies, n ,
We start with a reference Hamiltonian, H
and eigenfunctions, n , defined by
0 n = n n
H

(1)

0 was the nonrelativistic Hamiltonian of the hydrogen

For the example treated by Schrdinger, H
atom that we have used throughout the course.

Now suppose that we wish to calculate the spectrum of a different Hamiltonian, given by
=H
0 + V ,
H

(2)

which involves the presence of another potential, V , with being a dimensionless parameter. In
is nearby H
0 for small values of . For the
a deliberately imprecise sense, one can say that H
example treated by Schrdinger, V is the potential due to an additional electric field applied to the
hydrogen atom : V = eF~ ~r, where e is the elementary charge, F~ the electric field and ~r the electron
position vector. The role of the dimensionless parameter is as follows.
that is, find solutions to the equation
We wish to find the spectrum of H,
E = 0.
H

(3)

Towards this end, we represent and E as power series in ,

=

n n

(4)

En n .

(5)

n=0

and
E=

X
n=0

Then, by combining eqs. (2) through (5), we obtain

h

H0 + V

iX

p p

p=0

Eq q

q=0

r r = 0.

(6)

r=0

We are now going to solve eq. (6) by expanding in powers of and requiring that the coefficient
of each such power vanish. At this point you should pause and review the steps taken through eq.
(2) to eq. (6) and reconstruct them several times until you understand the the goal of this exercise
and the approach to it. If you can remember the goal and the approach then you can easily rederive
any of the formulas that now follow.
Term-by-term solution in powers of
We now solve for the coefficients of each power p that occurs when eq. (6) is expanded, starting
with p = 0.
Power : 0
0 0 E0 0 = 0.
H

(7)

This is just a specific case of eq. (1), and it hints at the power of this approach : we can see
when the potential V is turned on slowly, for H
H
0
what happens to any eigenfunction of H

as 0. For specificity, lets assume we are interested in the ground state of H. In this case,
E0 = 0 , 0 = 0 .

(8)

Power : 1
0 1 + V 0 E0 1 E1 0 = 0.
H

(9)

From this we get the first non-trivial result, by projecting onto h0 |, and recognizing that
0 |1 i = E0 h0 | |1 i,
h0 | H
E1 = h0 | V |0 i = h0 | V |0 i .

(10)

Eq. (10) is always the first thing to compute, especially because h0 | V |0 i will sometimes vanish due to symmetry reasons. It also leads directly to another key result : rewriting eq. (9), we
get
h
i
h
i
0 E0 1 = E1 V 0 .
H
(11)
This equation is inhomogeneous in 1 , and it shows that we can require that h0 |1 i = 0 without
loss of generality, since
h
i
h
i
h
i
0 E0 (1 h0 |1 i 0 ) = H
0 E0 1 = E1 V 0 .
H
(12)
Here is a useful trick for calculating 1 : always try this first. Expand 1 in terms of the n ,
X
1 =
a n n ,
(13)
n

and insert in eq. (11) to find

(n 0 ) = hn | V0 |0 i , a0 = 0.

(14)

Now eq. (13) can now be expressed in computable form :

1 =

X hn | V |0 i
n .
0 n

(15)

n6=0

In many elementary problems, hn | V |0 i =

6 0 for only a few values of n. In such cases, eq. (15)
is highly efficient. Always check for this.
Power : 2
We conclude with the second-order results since this order is where most of the basic questions
get answered, for example the blackbody radiation shift in atomic clocks that was mentioned above.
Note again that the expansion of eq. (6) is straightforward ; the coefficient of 2 is
0 2 + V 1 E0 2 E1 1 E2 0 = 0.
H

(16)

Projecting this onto h0 | results in an almost magical simplification :

h0 | V |1 i = E2 ,

(17)

in which we can substitute eq. (15) to obtain the computable form

E2 =

|
V
|
i
h
X
n
0 
0 n

n6=0

(18)

For the same reasons stated after eq. (15), always check this first. This shows that the second-order
energy of the ground state always decreases.
up to second order in perturbation theory,
When calculating an actual value of an energy of H
2
E = E0 + E1 + E2 , just set to the appropriate value. It is convenient to choose = 1, for
0 + V .
then you are calculating an approximate energy of H