THE LAMB SHIFT

by

PHILIP 0. KOCH

SUBMITTED IN PARTIAL FULFILLMENT OF THE

REQUIREMENTS FOR THE DEGREE OF

BACHELOR OF SCIENCE

at the

MASSACHUSETTS INSTITUTE OF TECHNOLOGY

(1956)

Signature of Author*... . .0.. . .0 . .

Certified by. . . . . . . .~ .
Thesis Supervisoi*
 ABSTRACT

The Lamb shift is studied in this paper. First,

the interaction between light and matter is discussed

and it is pointed out how this interaction should shift

the energy levels of an atom. The electromagnetic field

is then quantized and the mechanism of the interaction

of the electron with the zero-point radiation field is

described. A perturbation procedure is developed and

applied to the calculation of the Lamb shift in the same

approximation that Bethe used in his first calculation:

to the order A non-relativistic, and in dipole approxi-

mation. The idea of renormalization is stressed as it
arises during the calculation. The work of Kramers in
building a structure-independent theory of the interaction

of an electron with radiation is then described. The paper

ends with a discussion of the idea of renormalization which

has enabled the Lamb shift to be described successfully in

the present theory.

 TABLE OF CONTENTS

Section Title Page

I Introduction

II Hamiltonian 3

III Quantization of Electromagnetic Field 7

IV Perturbation Procedure 12

V Calculation of Energy Shift 15

VI Structure-independent Theory 21

VII Discussion 25

Bibliography 27
 I

I. INTRODUCTION

For many years, the spectrum of hydrogen agreed,

within experimental accuracy, with the prediction of Dirac's
relativistic electron theory. In this theory the S--t and

aP. states of hydrogen are degenerate. In 1947 Lamb and

Retherford used an ingenious microwave technique for the
comparison of these lines, publishing their first results
in August of the same year. They found that the 'Sir

state is higher than the 2AP. state by about 1000 megacycles

per second. This shift is about 9% of the difference between
the aP* and 9P, states. With improved experimental
techniques, the shift in the energy levels has been found

to high precision. It is 1057.8 megacycles per second. The

same effect has since been observed in the spectra of

deuterium and singly-ionized helium.

Schwinger, Weisskopf, and Oppenheimer first suggested

that the shift might be due to the interaction of the

electron with the zero-point fluctuations of the radiation

field. The possibility of such an effect must have been

considered before this time; but, without experimental

evidence to suggest it, no one had bothered to publish a
 2

calculation. If anyone had made a rough calculation, he

would have found the shift to be infinite. The calculation

was first made by Bethe and published two weeks after the
results of Lamb and Retherford. Bethets calculation was
non-relativistic, correct to the first order of the fine

structure constant, and done in dipole approximation. At

first the shift diverged linearly as the energy of inter-

acting radiation. By making a mass renormalization, he

was able to reduce this to a logarithmic divergence. Bethe

disregarded light of energy greater than the rest energy of

the electron, thus obtaining a finite shift. He guessed
that a relativistic calculation would give a convergent

shift without the necessity of arbitrarily cutting off at

high energy, and later work has proved him right. According

to his calculation, the & Si and aLP states should be

1040 megacycles per second apart. The calculation has

since been done relativistically, to higher order, and with

other small corrections. The result of all this is a value

1057.2 megacycles per second, in close agreement with experi-

ment.
 3

II. HAMILTONIAN

The energy (non-relativistic) of an electron moving

in the Coulomb field of the hydrogen atom is

Io~~j+V>
where r. is the reduced mass and

The eigenfunctions of this Hamiltonian are the well-known

hydrogen atom wave functions,

where the sth state is characterized by the integers n, 1,

and m and has an energy Ed.x

The energy in the electromagnetic field is

14, = -L ([Link]
This is the result of classical electromagnetic theory and

is not a quantum mechanical operator. When the electro-

magnetic field has been quantized, H1 is an operator having
the eigenvalues Ea and the eigenfunctions Fp. which com-
pletely describe the state of the field.

If the hydrogen atom and the radiation field exist

together and interact, it is improper to consider them

separately, as is done above. If there were no interaction,

 4

the Hamiltonian would be a sum,

H = Ho Ht-
,
and would have the eigenfunctions,

and the eigenvalues,

EA& = E +4E.
Spectra would not be affected by the field, for the energy
of every electron state is shifted by a constant amount,
regardless of the binding of the electron* In this case,
the energy in the field is not a physical observable and
its magnitude cannot be determined experimentally.

However, this is not the case, for there is an inter-

action between the electron and the field. The energy of
an electron moving non-relativistically in the electro-
magnetic field is given by

H= (A- )a+v,
where is the vector potential of the field. This is the
correct Hamiltonian because it leads to the Lorentz force

law. The energy of the total system of field and particle

is then

H = - . ( +- ) + i (Ea0 T

go + 4 -W
 5

where 14' is the energy due to the interaction of electron

and field.

It is this energy of interaction which gives rise to the

Lamb shift. Because the shift is known, from experiment,
to be very small, the use of perturbation theory to deter-
mine the effect of 1' is justified.
However, the Lamb shift is not explainable in terms
of classical electromagnetic theory. It is not an effect
like the Zeeman or Stark effects which occurs when a mag-
netic or electric field is turned on and disappears. when
the field is removed. In their experiments, Lamb and his
collaborators have gone to great lengths to determine the

effects of stray electric and magnetic fields and to sub-

tract such effects from their results. They have been able
to conclude that the shift is not caused by stray electric

or magnetic fields. From the classical point of view, a

radiation-free region is physically possible, although
difficult to obtain. In such a region, the above Hamiltonian
would reduce to N-o and no Lamb shift would be observed. In
order to understand the Lamb shift, one must use quantum

electrodynamics in place of the classical theory. Just as

in quantum mechanics the position and momentum of a particle

cannot be measured simultaneously with exactness, it is a

result of quantum electrodynamics that the measurement of

one component of the field introduces an uncertainty into

 6

another component. It is impossible to know that all com-

ponents of Z and 8 are simultaneously zero in a certain
region. It is not proper to consider a region of space as
completely free of radiation, for the fields continually
fluctuate about their zero value. It is these fluctuations
which interact with the electron and give rise to the Lamb
shift, when the effects of all externally applied fields
have been accounted for.
 7

III. QUANTIZATION OF THE ELECTROMAGNETIC FIELD

The success of quantum mechanics has suggested that

the quantum mechanical commutator relation might well be

applied to the electromagnetic field. The problem at hand

can not be solved when half of it is expressed quantum
mechanically and the other half classically. Therefore,

in this section, the electromagnetic field will be quantized.

Only an outline of the procedure will be given, since the
details can be found in texts.

A convenient gauge for this problem is the Coulomb

gauge:
=0 0 -
The vector potential then satisfies the equation:

and the electric and magnetic fields are given by:

The solution for A can be expanded in a sum of plane waves:

%
.-of . d AI

A Lk
 8

k points in the direction of motion of the plane wave

and has a magnitude

-A i and G:ka. are two unit vectors which form a right-

handed system with k and serve to determine the polariza-

tion of each plane wave in the sum. A takes on only

two values, but the k-sum is over all directions and magni-

tudes of .
The 41 ! are to obey periodic boundary conditions
on the opposite walls of a large box of side I. (running

waves are desired). This is satisfied by putting

L. Mi = 0L.
and, as a result, the 4AA 's are orthogonal:

The amplitudes are now redefined so that

Then

Since, in a plane wave, E z= the energy in the

field is

Because of the orthogonality of the AA 1

Is this reduces to

L
O ~ A IL ( & o 'a4J.~
 9

The symetrized form is used because, after quantization,

a1 Ag and A*A are not the same.

Now real combinations of the M& S are introduced:

Then

k AG

Thus the electromagnetic field can be represented by an

infinite number of pairs of canonical coordinates, &A and

.6A , each set having a Hamiltonian which is the same

as that of a simple harmonic oscillator. The field is

quantized by quantizing each of these "oscillators." For-

mally, the quantization is performed by setting

[P.3laiQAlI = -k
The method is the same as that used in the quantization of
the simple harmonic oscillator, as is the result:

Because the total Hamiltonian of the field is a sum, the

total wave function is a product; the following results are
obvious:

IIIF, nf+4t Fr=Ep r

F, (F .
 10

/YkA is called the number of photons of a certain

and A in the field. The numbers of photons in each

state are contained in the wave function of the field f-.

The operation of Alk.A on p reveals the number of photons
of that particular I and A that is present in the

field. By algebraic manipulations, it can be shown that:

a-u F(----- --- )= i + - .

These relations will be useful later. AA is called
the annihilation operator because, when it operates on the

field, the number of photons of type _rA is reduced by

one, as that one is absorbed by matter for example. 4Ap

is called the creation operator because it adds a photon

of type WA to the field, in the process of emission.

It will be observed that there is a zero-point energy

in the field.

4F 0 Ea F =( ( L) F
The fact that this zero-point energy is infinite is not

pleasing. It isntt a serious difficulty in itself, for

this energy is not a physical observable. Fe is the

zero-point wave function. It contains the information that

Malpj 0 for all states. Since

4fti
F. = F( ---- * --
it is not possible for a photon to be absorbed from the

zero-point field; a photon may be emitted. It is this

emission of a photon by the electron (and its immediate

 11

reabsorption) that is regarded as the mechanism by which

the electron interacts with the zero-point field.
It must be emphasized that this mechanism of photon
emission and absorption is a very convenient way of handling
radiation problems. It is the exact analogue of the classi-
cal methods which employ energy conservation in problem
solving. Just as, in classical physics, these same problems
can be solved in principle by using a force law, it is
possible in quantum electrodynamics to work directly with
the field fluctuations. It is a result of this theory that
the components of the fields obey comutation rules, not
all of Which are zero, This very fact that all field com-
ponents are not simultaneously measurable, gives rise to
the fluctuations in the field strengths. Welton has done
a non-relativistic calculation of the Lamb shift on this

basis. Considering the field fluctuations, he determined

the mean-square displacement of the electron from its un-

perturbed orbit and showed how this "Brownian motion" of

the electron reduced the effective potential of the nucleus.
As would be expected, he arrived at the same Lamb shift
expression that is obtained by the more common prodedure.
 12

IV. PERTURBATION PROCEDURE

Now that the electromagnetic field has been quantized,

recall

H ~ ~~
H H~- =
E A V +H 1-A
---- +- .
i-I 'will be regarded as a small perturbation and the
eigenfunction of 14 will be expanded in the
eigenfunctions of the Hamiltonian i-I, *14

A
4'aQ P FIL
Restricting the problem to the zero-point field and a
particular state mt- of the atom:

H P =-- E -- (E-,
+ -AE
1o + 1 1) 49-F =-. (E4u-+ CE ) An. F
T and ; are expanded in terms of increasingly higher
order, with a constant A , whose degree labels the order
of a term, as follows:

-'- use..

E~~~ -0 E.* EA 01 4' ----

E '+-
and the various orders of are defined by

q). Fm
. (2
1
(
W L
) 4 i
etc.
 13

In the Hamiltonian itself, the terms 14o +9 j are of

the zeroth order; and the perturbation is, of course, of
the first order in itself*

4 =_A$ + 1, +/-A '.

The expansions are then inserted into the equation

HE= EP
and terms of the same degree in A (thus of the same
order) are equated, giving a series of relations:

(I +11) = E '') E''

S+ a ) T ('"' 10 E '' 1" + E " )
lH4,14i ) HE
'1 E'*) 4'- T 4*+E' *
The first term says, as expected, that to the zeroth order
there is no perturbation:

1,~ F E'= E,.+E .o

When the expansion for and the now known values

of and E 0 are substituted into the second rela-

tion and certain manipulations are made, it is found that

E )= <n /II1fmo > .

Thus to first order (the parameter A is put equal to 1)

E = E" +-= EO+ 4+-

+- - lol" O>
A E - n o/H'y
1Lo> .
The third relation gives, upon substitution and manipulation,

A. e
 14

where Ae f ,y" a .

Thus, to second order

0c) + IE
-I-

+ (m o |H '|rwi ' 7> + zI 4 II <tj1 'IeM ><IM 0 I0 'IAL>

AE.a E0 - E. - E)

A-- (ama/|'b~no> czz (4l I'#Im0><PanI'| gtAP>

.Em+ Eo -EA.-- Eit)
0
 15

V. CALCULATION OF ENERGY SHIFT

The interaction of the electron with the zero-point

field was first described as the action of the zero-point
fluctuations of LE and 6 . For the purpose of
mathematical simplicity, this has been replaced by the
mechanism of emission and reabsorption of a photon by the

electron. Thus, in the sum over states of the field, the

only states which occur are the ground state F and the

states F, which contain one photon. It will be clear

that this is required mathematically in this calculation
to the order .a. , This will not be true for calculations
to higher orders; and, there, a more elaborate interaction

mechanism must be used.

Since the second term of the perturbation is already

of the order .A , its contribution to the Lamb shift

will be calculated in first order only. This term is

In the first order, the orthogonality of the F .s requires

that 4Z' and be equal to k and X . The above

product contains terms in 4 . aA# and &Ak aj. These

represent double emission and absorption and make no contri-

bution because they cannot connect F. to itself. The term
 16

Aa. aA also contributesnothing because the action of

the annihilation operator 4AA on F' gives 0. The
energy shift reduces to

When the sum on A is replaced by an integral (as is

done in a similar case later) this shift is seen to be
infinite, diverging quadratically with k . A divergent
shift like this is not very pleasing and points to the

necessity for an improved theory. However, it does not

stand in the way of the calculation. This shift, it will
be noticed, is the same for all electrons, regardless of
how they are bound and of how they are moving. Since it
is always present in every situation and at every time,
it is not a physical observable. For these reasons it

need not be thought of in terms of the interaction but

may be thought of as a self-energy which every electron

possesses. This process by which an infinite quantity is

"1subtracted" from a physically significant situation and

ascribed to the (unknown) structure of the electron is

called renormalization. It plays a very important part

not only in the understanding of the Lamb shift but of

the whole of quantum electrodynamics.

The first term of the perturbation is of the order r.

and so its effect should be calculated to second order.

 17

It is

0.A 0-. 0A-. JeaqA-2pE

It is clear that this gives no contribution to first order

because

AE C O 4ptAs .L + a .o
L

is zero by the orthogonality of the F 'S . The energy

shift to second order is

AE =Z.
~~ f F-r- + E0
x g
Once again, by orthogonality, the only state F that is
possible is F, which is characterized by one photon of
type WA . As before, the A.s should be expanded over
two sets of summation indices, ItA and WAT ; but
AE 0 unless

The energy difference in the denominator is then

Er - Ea= E-E = += : t kA
Then the energy shift is

A4 A-Et

First A are allowed to operate and

<O aA Il
>< laA*KO> - <0I- rO/1><KIr I.
Then the dipole approximation is made:

and the shift reduces to a.

 18

is replaced by 1m and the sum over is

replaced by an integral:

A.E qqri, ie dA d EA">

The integration is carried out over all directions of i

using ? as a polar axis and adjusting the CeS in

each case so that one of them lies in the plane of I and

*r and the other is perpendicular to it. The result is

.k- - >
For large k , this expression diverges linearly. How-
ever, Bethe pointed out that this is not the true expression
for the Lamb shift. The energy of the electron is shifted
by this amount for two reasons: one, that it is bound to the

hydrogen atom and twothat it has a certain average kinetic

energy. A free electron with the same average kinetic

energy will also undergo an energy shift because of its
interaction with the zero-point field. This shift should

properly be regarded as an electromagnetic mass effect,

because it is an energy which the particle possesses because
of its motion, regardless of how it may be bound. Thus it

is not physically observable and does not contribute to the

Lamb shift. The Lamb shift as it is observed in spectra
is the difference between the shift for the bound electron
and the shift for a free electron which has the same average

kinetic energy. For a free electron, the shift is given by

 19

the same formula; and, since the free electron wave

functions are eigenfunctions of the momentum, only the

diagonal element of &r remains:

Z.,>' I-el .' = 0 , A. +$-t1 E4m, - ExA.-- OW16

and,as a result)

AE =
3 -r J~<-.f Od
a; |< i rl 0*.
This is to be subtracted from the similar expression for

the bound electron to give, the correct expression for the

Lamb shift, E

E,.J (Z ACE
This expression diverges only logarithmically with

increasing I! . When this calculation is done relativisti-

cally, the shift converges. From relativistic considerations,

Bethe decided to cut off the energy of the interacting

photons at
-(.<. = /Kw-- .n
This cut-off can be appreciated if the electron is regarded
as having an extent of the order of its Compton wavelength,

0 .1The cut-off then means that all light of wave-

length smaller than the size of the particle should be

disregarded. In this picture, such short wavelengths do

not act upon the electron as a whole but affect only its
internal structure, of which nothing is known. Thus the

cut-off is justified.
If the integration over E is now performed,

-L I
F.
 20

and since /yyI.. C. S- >> E

AEL. -~
7rit I2flkLM4>/aI (E-E) - ~-
which is Bethe's result.
This is, of course, the shift for only one state. If

it were the same for all states, there would be no notice-

able effect. However, the part which the state type plays

in the formula makes it clear that different states are

shifted in different ways. Calculations have shown this

to be the case. S states are affected far more than

others. It is with S states in mind (particularly the
c2-J state of hydrogen) that the preceding work has
been done. To make the theory applicable to other states,
the perturbation theory must be developed more carefully,
keeping in mind the degeneracy of the unperturbed state.
 21

VI. STRUCTURE-INDEPENDENT THEORY

Many- of the problems which have arisen in the calcula-

tion of the Lamb shift and other radiative corrections were

studied by Kramers and discussed by him at the Solvay Con-

ference in 1948. Kramers was one of the firs.t to suspect

that the divergences which plagued quantum electrodynamics
were due, in good part, to the point model of the electron.
This model has, from the first, introduced an infinite self-

energy. The self-energy is that energy in the field of a

single electron: 0
ii= I (- -

and is clearly infinite if the radius is put equal to zero.

Since quantum electrodynamics has been formulated with the

correspondence principle always in mind, difficulties in
the classical theory have been carried over into the

quantized description. Kramers went to the root of the

problem and attempted, in non-relativistic approximation,

to make a better classical description of the interaction

of matter and radiation.

Kramers was guided in his work by Lorentzf classical

electron theory. Lorentz assumed the electron to be a

rigid sphere of finite radius but was able to describe many
 22

physical effects, using only two constants, the experi-

mental mass and charge of the electron. The radius of
the electron entered only into the electromagnetic mass,
a constant in a non-relativistic approximation. It is
the total mass, the sum of the inertial and electromagnetic
masses, which is a physical observable. By introducing
the total mass, Lorentz constructed a theory which depended

only upon physically measurable constants and was inde-

pendent of a knowledge of the structure of the electron.

Kramers' objective was to separate that part of

classical electromagnetic theory which is (approximately)
independent of the structure of the electron from that
part which depends upon the unknown details of electronic

structure. Although this is not, technically, a renormali-

zation, it is effectively the sane thing - and in a much

more appropriate place. The need for renormalization in

an advanced theory is often a result of a fault in the
basic theory. If a structure-independent theory is used
as a classical basis for quantum electrodynamics, those
divergences which arise from a point model or some other

structural assumption will not appear; and the corres-

ponding renormalizations will not be necessary.

In his address, Kramers considered the motion of a

single electron in the electromagnetic field* The electron

was considered to be a rigid sphere of finite extent; thus
 23

the treatment was, of necessity, non-relativistic. He

divided the vector potential A into two parts: Ae ,

the vector potential of the "proper" field,, and 747

the potential of the external field, v is approximately
the potential of a uniformly moving electron in radiation-

free space. It is structure-dependent in the near vicinity

of the electron. At may be regarded as the external
field which interacts with the electron in a physical
process. Although A depends upon Ao :

it was shown-to be approximately structure-independent.

The equation of motion of the electron was then brought
into the form ( r is the coordinate of the center of
mass of the electron):

where A. and V are evaluated at the center of mass.

If the field At is properly prescribed, this equation
represents the motion of an electron in an external field
in a manner which is approximately independent of the
electron's structure. Kramers constructed the Hamiltonian
which corresponds to this equation of motion. It is some-
what more complicated than the usual non-relativistic
Hamiltonian which is used in preceding sections. Kramers
showed that, when the Lamb shift is calculated in the same
elementary manner used in the preceding section, but with
his Hamiltonian, Bethe's final expression is obtained
 24

without the necessity of renormalization at this point.

This is a great triumph for the structure-independent

theory.
In a very interesting digression, Kramers gave a semi-

classical derivation of a Lamb shift for a harmonically

bound electron. The external field t , of a given fre-

quency, was expressed as a spherical standing wave which

was phase-shifted by an angle ?U which depended upon the

frequency of the light and the binding of the electron.

By enclosing the entire system in a large reflecting
sphere, the AIS were limited to a discrete, but infinite

set. The set of A 'S is clearly dependent upon the phase

shift fl and therefore upon the binding of the electron.

Kramers departed from a classical description only to con-

sider the zero-point energy in the field. Although it is

infinite for a harmonically bound electron or a free

electron, Kramers showed that the difference in the zero-

point energies in these two cases is finite and bears--a

close resemblance to Bethe's formula. The harmonic oscillator

was a simple example. It appears that it is not mathematically

feasible to make a semi-classical description of the hydrogen

Lamb shift in the same way.

Van Kampen has continued the work of Kramers. He has

derived Kramers' structure-independent Hamiltonian by a

series of canonical transformations and has applied it to

various radiation problems.

 25

VII. DISCUSSION

Although Kramers' work has led to a better understanding

of some of the problems of quantum electrodynamics, his
method does not seem capable of providing a complete solution.

For one thing, his theory is not quantum mechanical; yet

all radiative calculations are. For this reason, all of

the work done since 1948 has been directed toward perfecting
the process of renormalization rather than toward making a
structure-independent relativistic classical theory.

The difficulties which arise when the interaction of

radiation and matter is considered have been traced to the

use of a point electron. The divergent quantities have

been reduced to two: an infinite self-mass and an infinite

self-charge of the electron. The infinite self-mass is

just that self-energy, calculated in a previous section,
which is caused by the electronts interaction with the

zero-point field. The infinite self-charge is a more

sophisticated matter, which arises from Dirac's "hole"

theory and has no non-relativistic analogue. However, its

divergent nature is also due to the point model of the

electron. In the process of renormalization, this mass and

charge are considered as part of the experimental mass and
 26

charge. Once the necessary renormalizations have been

made finite calculations of physically observable effects

are possible. Resolution of the problem of the self-mass

and self-charge will require some knowledge about the

electron's structure. If it is really possible to separate

the problem of the behavior of the electron from that of
its structure, the theory is in good shape. It would seem
that such a separation would not be possible in an exact

treatment of the problem. However, as in much of physics,

the interaction problem has proved far too difficult to
be solved in any but an approximate way.

The situation is not quite as good as this, for

quantum electrodynamics contains other divergent quantities

which cannot be traced to a lack of understanding of elec-

tronic structure. A simple example is the infinite energy

of the zero-point field, which is obviously independent of

the presence of matter.' Any idea that this might be due

to the large size of the universe is seen to be false when
it is recalled that quantization in a box or sphere gives

rise to an infinite zero-point energy, regardless of the

volume enclosed. One may say that there is no difficulty

because the zero-point energy is not physically observable.

However, a theory is not closed simply because it can pre-

dict all observable effects accurately. A theory cannot

be wholly satisfactory if it contains implausible, though

unobservable, elements.
 27

BIBLIOGRAPHY

1. H. Bethe, Phys. Rev. 72 (1947).

2. W. Heitler, Qpantum Theory oIf Radiation (Oxford, 1954).
3. N. G. Van Kampen, Dan. Mat. Fys. Medd. 26 (1951).
4. H. A. Kramers, Report of the Solvay Conference 1948
(Brussels, 195OTo -
5. W. E. Lamb and R. C. Retherford, Phys. Rev. 72 (1947).

6. E. E. Salpeter, Phys. Rev. 89 (1953).

7. L. I. Schiff, Quantum Mechanics (New York, 1955).

8. V. F. Weisskopf, Rev. Mod. Phys. 21 (1949).
9. T. Welton, Phys. Rev. 74 (1948).