Вы находитесь на странице: 1из 114

Quantum Field Theory

(Shorter Version)

Pankaj Sharan
Formerly Professor
Physics Department, Jamia Millia Islamia
New Delhi

March 2019
2
Contents

1 Background and Introduction 1

§ 1. Hilbert spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

§ 2. Observables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

§ 3. Unitary and anti-unitary operators . . . . . . . . . . . . . . . . . 2

§ 4. Generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

§ 5. Symmetry and Wigner’s Theorem . . . . . . . . . . . . . . . . . 3

§ 6. Symmetry Groups and Lie algebras . . . . . . . . . . . . . . . . 5

§ 7. Observables as generators of symmetry . . . . . . . . . . . . . . 7

§ 8. Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

§ 9. Cross-section formulas . . . . . . . . . . . . . . . . . . . . . . . . 11

§ 10. Moving representation . . . . . . . . . . . . . . . . . . . . . . . 13

§ 11. Dirac’s bra-ket notation . . . . . . . . . . . . . . . . . . . . . . 14

§ 12. Useful Identities . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2 Introduction to Fock Space and Second Quantization 19

§ 13. Quantum mechanics of many identical particles . . . . . . . . . 19

§ 14. The Fock space . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

§ 15. Annihilation and Creation operators on F . . . . . . . . . . . . 21

§ 16. The symmetric Fock space . . . . . . . . . . . . . . . . . . . . . 22

§ 17. Annihilation and Creation operators on FS . . . . . . . . . . . 22

i
ii CONTENTS
§ 18. Properties of af and a†f . . . . . . . . . . . . . . . . . . . . . . 24

§ 19. Commutation Relations . . . . . . . . . . . . . . . . . . . . . . 24

§ 20. Orthonormal bases in Fock space . . . . . . . . . . . . . . . . . 25

§ 21. The anti-symmetric Fock space FA . . . . . . . . . . . . . . . . 26

§ 22. Properties of af and a†f . . . . . . . . . . . . . . . . . . . . . . 27

§ 23. Observables in Fock space . . . . . . . . . . . . . . . . . . . . . 29

3 Non-relativistic field theory 33

§ 24. Number Operator . . . . . . . . . . . . . . . . . . . . . . . . . 33

§ 25. Non-relativistic ‘Potential’ . . . . . . . . . . . . . . . . . . . . . 34

§ 26. Galilean invariance . . . . . . . . . . . . . . . . . . . . . . . . . 35

§ 27. Configuration space . . . . . . . . . . . . . . . . . . . . . . . . 36

§ 28. A and B ‘particles’ . . . . . . . . . . . . . . . . . . . . . . . . . 36

§ 29. Bare and Dressed Particles . . . . . . . . . . . . . . . . . . . . 37

§ 30. Single dressed B-particle . . . . . . . . . . . . . . . . . . . . . . 38

§ 31. Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

§ 32. B-B effective interaction . . . . . . . . . . . . . . . . . . . . . . 40

4 Relativistic spin zero particles and fields 43

§ 33. Mass m spin zero, neutral particle . . . . . . . . . . . . . . . . 43

§ 34. Representation of the Poincare Group . . . . . . . . . . . . . . 44

§ 35. Covariantly transforming states . . . . . . . . . . . . . . . . . . 44

§ 36. Klein-Gordon equation . . . . . . . . . . . . . . . . . . . . . . . 45

§ 37. States |xi are not orthogonal to each other. . . . . . . . . . . . 45

§ 38. {|xi} form a complete set for a fixed x0 . . . . . . . . . . . . . 45

§ 39. Klein-Gordon equation . . . . . . . . . . . . . . . . . . . . . . . 46


CONTENTS iii
§ 40. Inner product in |xi basis wave-functions . . . . . . . . . . . . . 46

§ 41. A note about the inner-product in |xi basis wave-functions . . . 47

§ 42. Observable matrices in |xi basis ‘wave-functions’ . . . . . . . . 48

§ 43. Pauli-Jordan Function . . . . . . . . . . . . . . . . . . . . . . . 49

§ 44. Fock space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

§ 45. Quantum Field . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

§ 46. Properties of the quantum field . . . . . . . . . . . . . . . . . . 52

§ 47. Energy and momentum . . . . . . . . . . . . . . . . . . . . . . 53

§ 48. Normal Ordering . . . . . . . . . . . . . . . . . . . . . . . . . . 53

§ 49. Energy and momentum as function of field . . . . . . . . . . . . 54

§ 50. Zero Point energy . . . . . . . . . . . . . . . . . . . . . . . . . 56

§ 51. Canonical equal time commutation relation . . . . . . . . . . . 57

§ 52. The Lagrangian density . . . . . . . . . . . . . . . . . . . . . . 57

§ 53. Parity and Time reversal . . . . . . . . . . . . . . . . . . . . . . 57

§ 54. Charge and U (1) internal symmetry . . . . . . . . . . . . . . . 61

§ 55. The “complex” scalar field . . . . . . . . . . . . . . . . . . . . . 62

§ 56. Charge Conjugation . . . . . . . . . . . . . . . . . . . . . . . . 63

5 Interacting boson fields 65

§ 57. Feynman Propagator . . . . . . . . . . . . . . . . . . . . . . . . 65

§ 58. Interaction picture . . . . . . . . . . . . . . . . . . . . . . . . . 66

§ 59. An illustrative model . . . . . . . . . . . . . . . . . . . . . . . . 67

§ 60. φ-φ scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

§ 61. Zeroth order . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

§ 62. First order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

§ 63. Second order . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69


iv CONTENTS
§ 64. Feynman Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . 72

§ 65. Momentum space . . . . . . . . . . . . . . . . . . . . . . . . . . 75

§ 66. Pairings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

§ 67. Wick’s theorem for ordinary products . . . . . . . . . . . . . . 77

§ 68. Wick’s theorem for time-ordered products . . . . . . . . . . . . 79

§ 69. Wick’s theorem for product of normal ordered expressions . . . 80

§ 70. Feynman Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

6 Spin 1/2 particle: Hilbert space and wave-functions 83

§ 71. Spin of a particle . . . . . . . . . . . . . . . . . . . . . . . . . . 83

§ 72. Momentum states for a spin half particle . . . . . . . . . . . . . 83

§ 73. Direct boost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

§ 74. SU (2) in the rest frame . . . . . . . . . . . . . . . . . . . . . . 84

§ 75. Construction of states in momentum space . . . . . . . . . . . . 85

§ 76. Configuration space states . . . . . . . . . . . . . . . . . . . . . 86

§ 77. Space and time inversion . . . . . . . . . . . . . . . . . . . . . . 87

§ 78. Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

§ 79. Time reversal . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

§ 80. Summary with matrix notation . . . . . . . . . . . . . . . . . . 90

§ 81. Dirac equation, 4-spinor, and γ matrices . . . . . . . . . . . . . 91

§ 82. ‘Configuration space’ wave functions . . . . . . . . . . . . . . . 92

§ 83. Properties of γ matrices . . . . . . . . . . . . . . . . . . . . . . 93

§ 84. Inner product and matrix elements . . . . . . . . . . . . . . . . 94

7 Spin 1/2 particle: Fock space and fields 95

§ 85. Fock space for neutral spin 1/2 (Majorana) Particles . . . . . . 95


CONTENTS v
§ 86. Fock space: Charged spin 1/2 field for (Dirac) Particles . . . . . 99

§ 87. Four-spinor formalism . . . . . . . . . . . . . . . . . . . . . . . 103

§ 88. Bilinear ‘invariants’ . . . . . . . . . . . . . . . . . . . . . . . . . 104


vi CONTENTS
Chapter 1

Background and Introduction

§ 1: Hilbert spaces

We assume familiarity with elementary facts about vector spaces.

A Hilbert space is a complex vector space, usually infinite dimensional,


with an inner product defined on it. If H is the space and φ and ψ vectors in
it and c a complex number then an inner product on H associates a complex
number (φ, ψ) satisfying the following conditions:

1. (φ, φ) > 0, and (φ, φ) = 0 if and only if φ = 0.

2. (φ, ψ + χ) = (φ, ψ) + (φ, χ)

3. (φ, cψ) = c(φ, ψ)

4. (φ, ψ) = (ψ, φ)∗ .

q
The positive number (φ, φ) is called the norm of φ and denoted by kφk. The
number kφ − ψk is a distance function between two vectors and that makes
a Hilbert space a metric space. A Hilbert space should be a complete metric
space with this distance function.

A linear operator on H is defined in the usual fashion. In general, the


domain of definition of a linear operator cannot be chosen to be the whole
space H.

§ 2: Observables

1
2 CHAPTER 1. BACKGROUND AND INTRODUCTION
All observables in quantum mechanics are self-adjoint operators. Adjoint
A† of an operator A, when it exists, is defined by the equation (φ, Aψ) =
(A† φ, ψ). A self-adjoint operator is one for which A = A† . It has the property
(φ, Aψ) = (Aφ, ψ). The eigenvalues of a self-adjoint operator are real numbers,
and eigenvectors corresponding to different eigenvalues are orthogonal.

There can be more than one linearly independent eigenvectors correspond-


ing to the same eigenvalue. If there are two observables A and B which com-
mute, that is [A, B] = 0, then there exist simultaneous eigenvectors of both
observables. For the same fixed eigenvalue of A there may be eigenvectors
belonging to different eigenvalues of B. Therefore for fixed eigenvalues of both
A and B there will be fewer independent eigenvectors. The degeneracy (that
is, the number of linearly independent vectors for a fixed set of eigenvalues) is
reduced by choosing two commuting observables.

In this way, by adjoining more and more observables, which all commute
with each other, we reach a complete set of commuting observables such that
there is only one simultaneous eigenvector for each set of eigenvalues. These
simultaneous eigenvectors are all orthogonal to each other. Such common
eigenvectors are very important in quantum mechanics because they form or-
thogonal basis elements.

§ 3: Unitary and anti-unitary operators

A unitary operator U on a Hilbert space H is a linear one-to-one invertible


linear mapping which satisfies
(U φ, U ψ) = (φ, ψ)
for all φ, ψ ∈ H

An anti-unitary operator A is also defined everywhere; and is a one-to-one


invertible mapping, but it is an anti-linear operator, that is,
A(φ + ψ) = Aφ + Aψ
for all φ, ψ ∈ H but, for a complex number c,
A(cφ) = c∗ Aφ
and moreover, it satisfies
(Aφ, Aψ) = (φ, ψ)∗ = (ψ, φ).

Both unitary and anti-unitary operators preserve the transition probability:


|(U φ, U ψ)|2 = |(φ, ψ)|2
3
and

|(Aφ, Aψ)|2 = |(φ, ψ)|2 .

§ 4: Generators

Given a self-adjoint operator G = G† and a real number s we can define a


unitary operator U (s) by

s2 G2
U (s) = eisG = 1 + isG − + ···
2!
It is easy to see that U (0) = 1 and

U (s)U (t) = U (s + t).

This makes the correspondence s → U (s) a unitary representation of the


one-parameter additive group of real numbers.

Note that the inverse of U (s) = exp(isG) is (U (s))−1 = U (−s) = exp(−isG).

Conversely, every unitary representation U (s) of one parameter group de-


fines a self-adjoint operator G, called the generator of the group representation,
so that U (s) = exp(isG).

This statement is known as Stone’s theorem.

§ 5: Symmetry and Wigner’s Theorem

All physical, that is, experimental, predictions of quantum mechanics de-


pend on the formula

an |(φn , ψ)|2
X
hAiψ = (ψ, Aψ) =
n

where ψ is the state of the system, A the observable, and an and φn the
eigenvalues and eigenstates of A. (This formula has been written for the case
when A has discrete eigenvalues with no degeneracies for illustrative purposes.
It can be suitably generalised for other cases.)

The physical state of a system therefore, is determined by a unit vector


ψ, (ψ, ψ) = 1, just as well as by exp(iθ)ψ, where θ is real. The so-called
‘phase factor’ exp(iθ), is a complex number of modulus unity. The two vectors
represent the same physical state. Thus, in quantum mechanics a physical
4 CHAPTER 1. BACKGROUND AND INTRODUCTION
state is represented by not just one unit vector ψ but by the whole infinite set
of vectors {exp(iθ)ψ}, (where θ is any real number). Such a set of vectors is
called a unit ray.

When a quantum transition takes place from a state represented by ψ into a


state represented by φ, what is physically relevant is the transition probability
|(φ, ψ)|2 . When a measurement of an observable is made, according to the
accepted interpretation of quantum mechanics, the state of the system makes
such a transition from that state, say ψ into one of the eigenstates, say φn of the
measured observable. This transition probability remains the same whether
we use φn and ψ or any multiples of them by two phase factors.

The transition probability depends only on the unit rays which rep-
resent the state of the physical system.

A symmetry transformation relates physical states for two ‘observers’ : that


is, there is a one-to-one mapping between physical states as seen by the two
observers. We expect that the number of times one physical state makes a
transition to another for one observer, should be the same as seen by the other
observer between the corresponding states for the other observer.

A one-to-one mapping of unit rays which preserves the transition


probability is called a symmetry transformation.

But unit rays are not easy to deal with.

For one thing, they do not even form a vector space! On the other hand,
given a unitary or an anti-unitary operator (which are mappings of vectors
into vectors), we can construct a ray mapping by associating the set of vectors
in one ray into the set of their images. This mapping will map unit rays into
unit rays.

Given a correspondence of unit rays, if a unitary (or anti-unitary) operator


maps the rays into the corresponding rays then we say that the operator is
compatible with the ray correspondence.

Wigner’s theorem says that

1. Given a symmetry correspondence of unit rays one can choose either a


unitary or an anti-unitary operator which is compatible with the given
ray correspondence.
2. This operator is unknown ‘up to a phase factor’.
5
3. The unitary or anti-unitary nature is fixed by the ray correspondence
itself.

The product of two unitary operator is a unitary operator, and the prod-
uct of two anti-unitary operators is also a unitary operator. Therefore any
symmetry transformation which can be written as a product of two symme-
try transformations of the same type will have to be represented by a unitary
operator.

Except for time reversal, all symmetry transformations are represented by


unitary operators. And there is a good physical reason why the time-reversal
operator is anti-unitary.

Symmetries do not occur individually, but as groups. If we think of a


symmetry transformation as a correspondence between two observers, then
the group character is implied by the fact that the descriptions of all observers
should match. But it is not possible to imagine the set of observers for internal
symmetries, or for the permutation symmetries of identical particles.

§ 6: Symmetry Groups and Lie algebras

A symmetry transformation followed by another such transformation is


also a symmetry. If we think of the operation of successive transformation as
a product, we conclude that :

all symmetries form a group of symmetries.

Most of the symmetry groups are Lie groups because each symmetry trans-
formation depends on a number of real parameters. For each value of these
parameters, (which can change continuously), there is a unitary operator. If
we change one parameter and fix all others then by a suitable choice of a
function of this parameter, we can get a one-parameter subgroup of trans-
formations represented by unitary operators. As discussed above, this will
identify a self-adjoint operator called the generator of these transformations.

Let U (t) and V (s) be two one-parameter groups of unitary transformations,


with generators A and B respectively. If A and B commute, that is, if [A, B] =
0, then the quantity
W = V −1 (s)U −1 (t)V (s)U (t),
is equal to identity:
W = 1.
6 CHAPTER 1. BACKGROUND AND INTRODUCTION
Let us find out what happens when the two generators do not commute. One
should picture the above products as a circuit of successive symmetry opera-
tions as in the following diagram:

end start
V −1 (s)U −1 (t)V (s)U (t) = W ? 1 → t → U (t)
↑ ↓
−s s
↑ ↓
U −1 (t)V (s)U (t) ← ← −t ← V (s)U (t).

When t and s are infinitesimally small, then we can write U (t) = 1 + itA −
t A2 /2! + . . . and V (s) = 1 + isB − s2 B 2 /2! + . . .. We can calculate W for
2

small t and s

W = (1 − isB − s2 B 2 /2! + . . .)(1 − itA − t2 A2 /2! + . . .)


×(1 + isB − s2 B 2 /2! + . . .)(1 + itA − t2 A2 /2! + . . .)
= 1 + ts(AB − BA) + . . .

obtained by carefully keeping terms up to the second order (that is upto t2 , s2


and ts).We find that for t, s small, W is an infinitesimal unitary operator a
one-parameter group C is the self-adjoint operator C = −i[A, B]. Therefore,
the commutator of A and B generates another infinitesimal symmetry trans-
formation and C is a new generator or observable.

We can then take a similar circuit with any pair of these observables and
obtain iD = [C, A], iE = [C, B] and so on. We can also take any (real) linear
combinations of these observables (generators) and generate new symmetry
transformations.

Fortunately, for physical applications the process of generating new ob-


servables by taking linear combinations and commutators closes and there are
only a finite number of them A1 , A2 , . . . , An such that commutator of any two
of them only gives a linear combination of these same observables. They are
said to form a Lie algebra. Since finite symmetry transformations can always
be constructed by successive application of infinitesimal ones, the Lie algebras
determines the representation of the Lie group of symmetries.

A real algebra is a real vector space V with elements X, Y, . . . on which is defined a


bilinear ‘product’ X.Y in addition to the vector space structure. If the product is asso-
ciative (X.Y ).Z = X.(Y.Z) it is called an associative algebra. If all elements ‘commute’
7
(X.Y = Y.X) it is called a commutative algebra. A Lie algebra is defined by a bilinear
product (written as) [X, Y ] which is antisymmetric [X, Y ] = −[Y, X] and which satisfies the
Jacobi identity [[X, Y ], Z] + [[Y, Z], X] + [[Z, X], Y ] = 0. This makes the Lie algebra non-
associative because [[X, Y ], Z] = [X, [Y, Z]]+[Y, [Z, X]]. For physical applications the Lie al-
gebra elements are linear operators whose Lie product is the commutator [X, Y ] = XY −Y X.
Actually, the real Lie algebras physicists deal with are made up of anti-Hermitian opera-
tors, but they prefer to write the Lie algebra elements as Hermitian or self-adjoint operators
because Hermitian operators represent observables. That is why in the Lie bracket or com-
mutator relations like [J1 , J2 ] = iJ3 the imaginary i occurs which should not be there in a
real vector space and algebra.

Summary: If U (t) = exp(itA) and V (s) = exp(isB) are two uni-


tary one-parameter subgroups of symmetry generated by self-adjoint
operators A and B respectively, then they determine a third unitary
one-parameter subgroup of symmetry generated by the self-adjoint
operator C given by [A, B] = iC.

We will have many examples of Lie groups and Lie algebras.

§ 7: Observables as generators of symmetry

We have seen that generators of any one-parameter unitary group (which, in


general, is a subgroup of a larger symmetry group) is a self-adjoint operator,
and hence an observable in principle. On the other hand, every observable A
will define a one-parameter group of symmetry transformations by the formula
s → exp(isA).

In a most general sense, once the Hilbert space for the physical system
has been specified, all self-adjoint operators are observables, and all unitary
(and anti-unitary) operators are symmetries. We have here an embarras de
richesses!

This situation occurs in classical physics also.

In the Hamiltonian formulation, any real function A(q, p) defined on the


phase space is an observable, whether or not we are interested in it. And any
observable A is a generator of a one-parameter group of canonical transforma-
tion (the equivalent of a symmetry transformation) (q, p) → (q(s), p(s)) using
the poisson bracket as follows

dq dp
= {q, A}, = {p, A}.
ds ds
8 CHAPTER 1. BACKGROUND AND INTRODUCTION
The canonical transformations also form a group, and like the quantum me-
chanical case, if A(q, p) and B(q, p) are two observables which generate one-
parameter groups, then their Poisson bracket C(q, p) = {A, B} is another
generator. These generators also form a Lie Algebra just as their quantum
mechanical counterparts form a Lie algebra through the commutator bracket.
This correspondence between Poisson brackets and commutators was discov-
ered by Dirac.

In classical mechanics the situation is somewhat under control because the


coordinates and momenta {q i , pi } themselves are observables. In classical me-
chanics, the specification of the state of a system through {q i , pi } is equivalent
to specifying the values of all its observables.

When we establish a quantum theory, we do not know, a priori, which self-


adjoint operators will represent which observables. All we have is a Hilbert
space, and,

unlike classical physics, the state vectors do not have any bearing
on the observables that may exist.

This is what makes the role of symmetries indispensable in quantum me-


chanics. All observables are generators of some symmetry transformations.
Moreover,

the symmetries are supposed to be known to us independently of the


quantum mechanical formulation.

For example, Lorentz and Poincare symmetries are our starting point in
setting up a relativistic quantum theory. We just ask: Is there a Hilbert space
which supports a representation by unitary operators of every member of the
Poincare group? This is purely a mathematical question. Then, once we have
the answer, we can go on to define energy, momentum, angular momentum
etc as the respective generators of time translations, space translations, and
rotations etc.

One word about parity and time-reversal operators. They are as defined
on space time points by 4×4 matrices whose squares are equal to unity. They
are represented by Wigner’s theorem as unitary and anti-unitary operators
respectively. But they are undetermined upto a phase factor. These phase
factors can be restricted by physical considerations and their values lead to
observable consequences. Therefore these parity and time reversal phases also
fall in the category of observables.
9
§ 8: Dynamics

Dynamics (that is time evolution) in quantum mechanics is determined by


the unitary operator exp(−iHt/h̄). In the Schrodinger picture, all observables
are time independent and the one special vector ψ(t) which represents the state
of the system at time t changes according to

∂ψ
ψ(t) = exp(−iHt/h̄)ψ(0), or ih̄ = Hψ.
∂t

However other descriptions are possible. What matters physically are the
experimentally measurable quantities, which are the expectation values of ob-
servables,

hAiψ = (ψ(t), Aψ(t)) = hψ(t)|A|ψ(t)i


= hψ(0)| exp(+iHt/h̄)A exp(−iHt/h̄)|ψ(0)i.

Therefore we can equally well calculate the expectation value of time-dependent


observables

AH (t) = exp(+iHt/h̄)A exp(−iHt/h̄)

in the unchanging state represented by the vector ψH = ψ(0). This way of


treating dynamics is called the Heisenberg picture. (The subscript H in AH or
ψH is for Heisenberg, not Hamilton!) The observables satisfy the equations of
motion,

dA
ih̄ = [A(t), H]
dt
in Heisenberg picture.

A third description, called the interaction picture, or Dirac picture is used


in perturbation theory when the Hamiltonian can be written i(in Schrodinger
picture) as

H = H0 + λV

with H0 , the ‘free’ hamiltonian, and λV the ‘interaction hamiltonian’. (And


not any kind of potential energy, as the notation might suggest!) λ is a ‘cou-
pling’ constant or parameter assumed small.

Here both the state vector as well as the observables change with time. The
time evolution is determined by the total hamiltonian always. Starting from
10 CHAPTER 1. BACKGROUND AND INTRODUCTION
the Schrodinger picture we define the (fictitious) free time evolution determined
by H0 for observables:

dAI
AI (t) = exp(+iH0 t/h̄)A exp(−iH0 t/h̄), ih̄ = [AI (t), H0 ].
dt
In order to keep the expectation values unchanged the state vectors in this
picture will also have to change according to

ψI (t) = exp(+iH0 t/h̄)ψ(t) = exp(+iH0 t/h̄) exp(−iHt/h̄)ψ(0).

The usefulness of the interaction picture is this: suppose the interaction is


‘switched on’ only for some finite time say from time −T to +T . This means
that V is effectively zero before −T as well as after T . A state φin before −T
will be a ‘free state’ evolving with H0 . During the period between (−T, T ) it
will evolve as
∂φI dh i
ih̄ = ih̄ exp(+iH0 t/h̄) exp(−iHt/h̄)ψ(0)
∂t dt

= λVI (t)φI (t), φI (−T ) = φin ,

where

VI (t) = exp(+iH0 t/h̄)V exp(−iH0 t/h̄).

Perturbation theory is developed by integrating the equation for φI (t),


iterating the equation assuming VI to be small. If there was no interaction
then states in the interaction picture will be constant in time and observables
will evolve with H0 . With V switched on, a free state φin before time −T will
become a state χ after time T where

λ2 Z T
!
λ ZT Z t2
χ = 1+ dt1 VI (t1 ) + dt V (t
2 I 2 ) dt1 VI (t1 ) + · · · φin
ih̄ −T (ih̄)2 −T −T

λ2 1 Z T Z T
!
λ ZT
= 1+ dt1 VI (t1 ) + dt2 dt1 T (VI (t2 )VI (t1 )) + · · · φin
ih̄ −T (ih̄)2 2! −T −T
" !#
iλ Z T
= T exp − dtVI (t) φin
h̄ −T
≡ U (T, −T )φin

where T is the time ordering or chronological ordering symbol and simply


means that if, in a term, there are factors dependent on time then those factors
must be re-arranged so that factors with later times are to the left of factors
11
with earlier times. The operator U (T, −T ), in the limit of T → ∞ is called
the Scattering matrix or, simply, S-matrix:

S = lim U (T, −T )
T →∞
" !#
iλ Z ∞
= T exp − dtVI (t)
h̄ −∞

The transition probability of a free state which starts with φin in remote past
(that is before −T , for large T ) into another free state φout (in remote future,
after a large time +T ) is

|(φout , χ)|2 = |(φout , Sφin )|2 .

Remark 1 It is clear that the time-ordering is not un-ambiguous if two factors


happen to be defined for the same value of time. This ambiguity is not trivial at all
and is the source of ‘divergences’ or infinities in quantum field theory. We will deal
with them when faced with it.

Remark 2 In practice the Schrodinger picture is never used in quantum field the-
ory. Mostly we use the perturbation theory, where we can start with interaction
picture. All observables, which depend on fields, are taken to be freely evolving
with the hamiltonian H0 . The interaction is introduced as a function VI (t) of these
free fields. The S-matrix is then calculated upto the order of λ required.

§ 9: Cross-section formulas

We give without proof the formulas for scattering cross sections in terms
of the matrix elements of the S-matrix.

1. Write the S-matrix in energy basis as

hE, α|S|E 0 , α0 i = δ(E − E 0 )δαα0 − 2πiδ(E − E 0 )T (E, α; E 0 , α0 ).

2. Non-relativistic scattering: Starting from the initial state with mo-


mentum k the differential cross-section for the particle to end up with
same energy but momentum k0 (|k0 | = |k|) within a solid angle dΩ is

dσ = (2π)4 h̄2 m2 |T (E, α; E, α0 )|2 dΩ


12 CHAPTER 1. BACKGROUND AND INTRODUCTION
3. Relativistic scattering The cross section comes out in this case as

(2π)4 h̄2 Z d3 k1 d3 kn 4
σ= 0 0 · · · δ (Pf − Pi )|T (k1 λ1 . . . kn λn ; p1 σ1 p2 σ2 )|2
2p1 2p2 vrel 2k10 2kn0

where the relative velocity of particles in one beam relative to those in


the other is

p p2
2 1
vrel = c 0 − 0
p1 p2

4. Relativistic decay

N (t) = N (0) exp[−t/τξ ]

where τξ is the decay constant (“lifetime”).

1 2 1
= Im hpσ|B|pσi
τξ h̄ 2p0
2π 1 Z
= d(ξ)δ 4 (Pξ − p)|T (ξ, pσ)|2
h̄ 2p0
where the T-matrix is defined by

hξ|S|pσi = hξ|pσi − 2πiδ 4 (Pξ − p)T (ξ, pσ)


13
§ 10: Moving representation

Although it is obvious, but it needs to be emphasised that whenever time


dependence is involved one must specify whether we are in Schrodinger, or
Heisenberg or interaction picture. In the Schrodinger picture only one vector,
namely the state vector depends on time. All observables are constants in time.
In the Heisenberg picture all observables are dependent on time and the state
vector is independent of time. But we can, if required, consider eigenstates of
some observable in the Heisenberg picture. Since the observable in question
is time-dependent, its eigenvalues and eigenvectors will depend on time too.
If we have a complete set of commuting observables in Heisenberg picture,
and use its common eigenstates for defining a ‘wave-function’, then the time
dependence of such a wave function comes not from the state vector (which is
time independent) but from the time dependence of the representation. Such
a representation is called a moving representation.

As a simple example, consider a single non-relativistic free particle in


Heisenberg picture. Let {|x, ti} be the eigenstates of the position operator
x̂(t), that is
x̂(t)|x, ti = x|x, ti.
Now we can legitimately ask the question:

If the state vector of the system is given by |x, t1 i what will be


state vector of the system at time t2 > t1 ?

The answer to this trivial question is not |x, t2 i, but the same vector |x, t1 i,
because in the Heisenberg picture states do not change!

What we can and do ask is the question:

given the state of the particle is |x1 , t1 i what is the probability


that when a measurement is made for the position of the particle,
at time t2 , it will be found to have value x2 ?

Now, position operator at time t2 is x̂(t2 ), and the probability amplitude will
therefore be
hx2 , t2 |x1 , t1 i.

Exercise 1 Show that for a free non-relativistic particle of mass m the amplitude
above is
3/2 " #
m i m(x2 − x1 )2

hx2 , t2 |x1 , t1 i = exp
2πh̄i(t2 − t1 ) h̄ 2(t2 − t1 )
14 CHAPTER 1. BACKGROUND AND INTRODUCTION
§ 11: Dirac’s bra-ket notation

History

Dirac’s bra-ket notation was developed to handle continuous eigenvalues.


The first indication of notation (such as (α0 /β 0 )) appears in the 1927 paper
“The Physical Interpretation of Quantum Dynamics” Proc Roy Soc. A 113,
621, 1927. This paper also introduces the generalized function, later called by
the name ‘Dirac delta function’. In another paper in 1929 (Proc Roy Soc A
123, 714) the notation (q 0 |H|q 00 ) appears. The notational convention published
in The proceedings of the Cambridge Philosophical Society, 35, 416, (1939), is
introduced by Dirac in the 1947 edition of the book “Principles of Quantum
Mechanics” with these words,

“The book has again been mostly rewritten to bring in various


improvements. The chief of these is the use of the notation of bra
and ket vectors, which I have developed since 1939.”

Connection with Hilbert space notation

We assume familiarity with basics of Hilbert space theory. For every φ ∈ H


the inner product (φ, ψ) can be interpreted as a bounded linear functional Fφ
(that is, a linear mapping of the Hilbert space H into complex numbers, which
is bounded over all unit vectors):

Fφ (ψ) = (φ, ψ).

In fact all bounded linear functionals arise in this way. For any bounded
linear functional F there exists a vector φ ∈ H such that F (ψ) = (φ, ψ).
Thus there is a one-to-one correspondence with such functionals and vectors
of H. If φ corresponds to F then c∗ φ corresponds to cF and so on. Such
linear functionals themselves form a Hilbert space H∗ called the dual of H.
The inner product (F1 , F2 ) in H∗ between two such functionals F1 and F2 is
defined as (F1 , F2 ) = (φ2 , φ1 ) (notice the order of factors) where φ1 and φ2 are
the vectors corresponding to F1 and F2 respectively.

In Dirac’s notation the vectors of H such as φ are written as ‘kets’ |φi and
the corresponding linear functional in Fφ ∈ H∗ is written as a ‘bra’ : Fφ = hφ|.
The action of Fφ on H is written

Fφ (ψ) = hφ|ψi
15
which is the complete ‘bra-c-ket’, a complex number.

It should be noted that in Dirac’s notation complex numbers can be multi-


plied to the left as well as to the right of the bras or kets. Usually multiplication
by a number in a vector space is defined as multiplication from left. Thus a
complex number c can be written on both sides of the bra or ket vector:

c|φi = |φic, hφ|c = chφ|.

The following rules can be worked out easily for this correspondence. (Here
c is a complex number, and A, B etc linear operators on H.)

H ↔ H∗
|φi ↔ hφ|
|φi + |ψi ↔ hφ| + hψ|
c|φi ↔ c∗ hφ|
A|φi ↔ hφ|A†

The last equation is to be understood as follows.

FAφ (ψ) = (Aφ, ψ) = (φ, A† ψ) = Fφ (A† ψ) = hφ|A† |ψi.

Thus the linear functional FAφ is again a linear functional denoted by hφ|A† .

We can regard the bra and ket vectors as adjoints of each other under the
mapping ↔. In usual Hilbert space theory only operators have adjoints. In
Dirac’s notation it is possible to write

(hψ|)† = |ψi, (|ψi)† = hψ|

but normally one doesn’t have to. In Dirac’s book bras and kets are called
‘conjugate complex’ of each other.

Here are some more relations,

AB|φi ↔ hφ|B † A†
hφ|ψi∗ = hψ|φi
hφ|AB|ψi∗ = hψ|B † A† |φi

and so on.

What is |φihψ| ?
16 CHAPTER 1. BACKGROUND AND INTRODUCTION
Given a pair of vectors φ and ψ in H we can define a linear operator Aφψ as
follows: acting on a vector ξ the result is

Aφψ ξ = (ψ, ξ)φ.

The operator Aφψ is written as |φihψ| and the defining equation as

(|φihψ|)|ξi = |φi(hψ|ξi)

the quantity in parentheses being the complex number hψ|ξi. Thus in Dirac
notation, we can simply write

(|φihψ|)|ξi = |φihψ|ξi.

The adjoint of this operator is defined by the equation

(χ, A†φψ ξ) = (Aφψ χ, ξ) = (ψ, χ)∗ (φ, ξ) = (χ, ψ)(φ, ξ) = (χ, Aψφ ξ),

therefore A†φψ = Aψφ . Since the operator Aφψ is written as |φihψ|, the above
equations are written elegantly as

(|φihψ|)† = (hψ|)† (|φi)† = |ψihφ|.

The most important class of such operators is the class of projection operators.
If φ is normalized (φ, φ) = 1, then the projection operator Pφ corresponding
to it is defined as

Pφ ψ = (φ, ψ)φ.

Is is easy to see that Pφ is self-adjoint and satisfies Pφ2 = Pφ . These relations


are obvious in Dirac’s notation

Pφ = |φihφ|.

If {φn }∞
n=1 is an orthonormal basis in the Hilbert space then
X X
P φn = 1 = |φn ihφn |.
n n

This holds for continuous set of basis vectors as well when we choose as basis
vectors the simultaneous eigenvectors of a ‘complete set of commuting observ-
ables’ and where some of the observables have continuous eigenvalues.

Dimensional analysis for bra-kets

We denote the physical dimension of a physical quantity A by [A].


17
In xµ = (x0 = ct, x) the time component has [x0 ] = L. Similarly for a
particle of rest mass m
h q i
µ 0
p = p = ωp = [ |p|2 c2 + m2 c4 ]/c, p

p0 = ωp will have the same dimensions as momentum, [p0 ] = M LT −1 .

Proper Hilbert space vectors are dimensionless, but that is not the case for
non-normalizable Dirac kets (or bra) which are labelled by continuous vari-
ables. For momentum eigenvectors |pi of a relativistic particle
Z
d3 p
hp|pi = 2ωp δ 3 (p − p0 ), |pihp| = 1,
2ωp

and so, in natural units,


1
[|pi] = .
momentum
The
R
Dirac delta function carries the inverse dimension [δ(x)] = [x]−1 because
of dx δ(x) = 1.

§ 12: Useful Identities

Identity 1:
Let A and B be two operators. Then
1 1
eA Be−A = B + [A, B] + [A, [A, B]] + [A, [A, [A, B]]] · · ·
2! 3!
1 1
= B − [B, A] + [[B, A], A] − [[[B, A], A], A] + · · ·
2! 3!
Proof:
Define

Bt = etA Be−tA , B0 = B, B1 = eA Be−A .

Then, we can write the Taylor expansion



dBt t2 d2 Bt
Bt = B0 + t + + ···
dt t=0 2! dt2 t=0

Now,
dBt
= etA [A, B]e−tA = [A, Bt ]
dt
18 CHAPTER 1. BACKGROUND AND INTRODUCTION
and
d2 Bt d
2
= [A, Bt ] = [A, [A, B]]
dt dt
and so on.

Identity 2:
Let A and B be two operators such that [A, B] commutes with both A and B.
Then

eA eB = e[A,B]/2 eA+B = e[A,B] eB eA .

Proof:
Define

F (t) = etA etB e−t(A+B) , F (0) = 1.

Differentiate,

dF (t)
= etA [A, etB ]e−t(A+B) .
dt
Now,
X tn
[A, etB ] = [A, B n ] = t[A, B]etB
n!
because [A, B] commutes with B. Therefore, since [A, B] commutes with A as
well,

dF (t)
= t[A, B]F (t).
dt
The solution to this operator differential equation is
2 [A,B]/2 2 [A,B]/2
F (t) = F (0)et = et .

Therefore , multiplying by eA+B on the right we get,

eA eB = e[A,B]/2 eA+B .

The second part of the identity is obtained by interchanging A and B, and


substituting for eA+B from the first part of the identity:

eB eA = e−[A,B]/2 eA+B = e−[A,B] eA eB .

Therefore eA eB = e[A,B] eB eA .
Chapter 2

Introduction to Fock Space and


Second Quantization

In order to understand quantum field theory you essentially need to understand


two concepts:

(1) quantum theory of many (actually, infinite number of) identical quanta
or particles;

(2) and how exchange of quanta produces interaction or force between other
quanta or particles.

§ 13: Quantum mechanics of many identical particles

A single quantum system is described by vectors f, g etc. in a Hilbert space


H characteristic of the system.

If there are two systems then their states are vectors in the product Hilbert
space H1 ⊗ H2 .
Usually Hilbert spaces in quantum mechanics are defined in terms of representations of
vectors by wave functions, either in configuration space or momentum space. For example
if in the spaces H1 , H2 vectors f ∈ H1 and g ∈ H2 are determined by wave functions f (x1 )
and g(x2 ) then the vector f ⊗ g ∈ H1 ⊗ H2 is determined by (f ⊗ g)(x1 , x2 ) = f (x1 )g(x2 ). A
general vector in H1 ⊗ H2 is a linear combination of ‘decomposable’ or factorizable vectors
such as f ⊗ g. The inner product in this space is defined first on factorizable vectors as
Z Z
(f ⊗ g, u ⊗ w) = (f, u)(g, w) = d x1 d3 x2 f ∗ (x1 )g ∗ (x2 )u(x1 )w(x2 ),
3

and then extended to all vectors using the linear property of the inner product.

If the two systems are identical, the states are members of H2 = H ⊗ H.

19
20CHAPTER 2. INTRODUCTION TO FOCK SPACE AND SECOND QUANTIZATION
In this case the vector f ⊗ g and g ⊗ f actually represent the same physical
state because the two sub-systems are identical. In other words, the trans-
formation U12 which maps all states f ⊗ g into g ⊗ f (for any f or g in H)
is a symmetry transformation. If we apply this transformation twice, we get
the same state that is, it is equivalent to the identity transformation. Thus,
identity 1 and U12 form a group of transformations. If there are N identical
particles, we get a symmetry group of all operators U (P ) of N ! permutations
P . (A permutation P here is a one to one correspondence or mapping of
the set {1, 2, . . . , N } of these natural numbers onto themselves. There are
N ! such mappings. The group product of mappings is defined by successive
application.)

Exercise 2 Show that the mapping U (P ) defined on Hn = H ⊗ · · · ⊗ H (n factors)


by
U (P )(f1 ⊗ · · · ⊗ fn ) = fP (1) ⊗ · · · ⊗ fP (n)
is unitary. Hint: Show that
(U (P )g1 ⊗ · · · ⊗ gn , U (P )f1 ⊗ · · · ⊗ fn ) = (g1 ⊗ · · · ⊗ gn , f1 ⊗ · · · ⊗ fn ).

As you know, symmetries of a quantum system are unitary (or anti-unitary)


operators in H. And, under a symmetry transformation, the states either
do not change (that is, they are invariant), or the physical states form a
subset of states which, under symmetry transformation, transform into linear
combinations of themselves. These subsets are said to provide a representation
of the symmetry group.

In the familiar example of rotational symmetry, a number j labels these


representations. j = 0 corresponds to invariant states, and other values of j
provide higher, 2j + 1 dimensional representations.

It is a fact of nature, for identical particles, only the two simplest represen-
tations of the permutation group are realised in nature. These are, those for
which the permutation operator U (P ) is either the identity 1 for all P . These
are states symmetric under U (P ). Or, U (P ) = ±1, depending on whether the
permutation P is even or odd. These are the antisymmetric states.

There are many other mathematically possible representations of the per-


mutation group, but they do not seem to be realised in nature.

§ 14: The Fock space

In field theory we deal with any number of identical particles, involving super-
position of states with different number of particles. The particles or quanta
which occur in nature only with symmetric states are called bosons.
21
The big Fock space F is a direct sum of all the following sub-spaces:

1. the one dimensional space H0 = {Φ0 } of zero- or no particles, with a


distinguished vector Φ0 called the vacuum,

2. the space H of one particles,

3. the subspace H2 = (H ⊗ H) containing states of two particles,

4. the subspace H3 = (H ⊗ H ⊗ H of states of three particles,

and so on...

A typical vector of the Fock space F will be like

φ = c0 Φ0 + c1 Φ1 + · · · + cn Φn + · · ·

where Φn are in HSn = (H⊗· · ·⊗H)S , the symmetric subspace of n-fold product
Hn = H ⊗ · · · ⊗ H.

§ 15: Annihilation and Creation operators on F

For any f ∈ H define the annihilation operator Af : Hn → Hn−1 for any n,


first on decomposable vectors as

Af Φ0 = 0,
Af g = (f, g)Φ0 ,

Af (f1 ⊗ · · · ⊗ fn ) = n(f, f1 )f2 ⊗ · · · ⊗ fn , n>1

and then extending the definition to all other states as a linear operator. we
call it annihilation operator because it reduces the number of particles by 1.

The adjoint of this operator is written A†f and is a mapping from Hn−1 to
Hn :

A†f Φ0 = f
A†f g = f ⊗ g

A†f (f1 ⊗ · · · ⊗ fn−1 ) = n f ⊗ f1 ⊗ · · · ⊗ fn−1 , n>2

Exercise 3 Show that A†f is the adjoint of Af .


22CHAPTER 2. INTRODUCTION TO FOCK SPACE AND SECOND QUANTIZATION
Schematically

Af Af
H0 ←− · · · Hn−1 ←− Hn · · ·

A†f A†f
H0 −→ · · · Hn−1 −→ Hn · · ·

§ 16: The symmetric Fock space

The symmetric Fock space FS ⊂ F is the symmetric subspace of the big Fock
space. It is composed of the following sub-spaces

1. the one dimensional space H0 = {Φ0 } of zero- or no particles, with a


distinguished vector Φ0 called the vacuum,
2. the space H of one particles,
3. the subspace HS2 = (H ⊗ H)S of all symmetric (under interchange) of
vectors of H2 = H ⊗ H,
4. the subspace HS3 = (H ⊗ H ⊗ H)S of all symmetric (under permutation)
of vectors of H3 = H ⊗ H ⊗ H,

and so on...

All these subspaces are orthogonal to each other.

§ 17: Annihilation and Creation operators on FS

Physical states are not in F, but they are either in symmetric subspace FS
if the identical particles are bosons, or in the antisymmetric subspace FA if
they are fermions. Therefore we have to use operators which connect these
subspaces.

Presently we only consider symmetric subspaces.

The annihilation operator Af defined on F, when operating on symmet-


ric vectors in HSn produces linear combination of symmetric vectors in HSn−1
because, for example,
Af (f1 ⊗ f2 ⊗ f3 + f1 ⊗ f3 ⊗ f2 + f2 ⊗ f1 ⊗ f3 + f2 ⊗ f3 ⊗ f1
+f3 ⊗ f1 ⊗ f2 + f3 ⊗ f2 ⊗ f1 )
=
√ √
3(f, f1 )(f2 ⊗ f3 + f3 ⊗ f2 ) + 3(f, f2 )(f1 ⊗ f3 + f3 ⊗ f1 )

+ 3(f, f3 )(f1 ⊗ f2 + f2 ⊗ f1 )
23
and similarly for higher n.

We denote the restriction of Af to the symmetric subspace FS by af .

On the other hand, a creation operator A†f does not map a symmetric vector
in Hn−1 into a symmetric vector in Hn . For example, in

A†f (f1 ⊗ f2 + f2 ⊗ f1 ) = 3(f ⊗ f1 ⊗ f2 + f ⊗ f2 ⊗ f1 )

the right hand side is not a vector of HS3 . Therefore we have to symmetrize the
right hand side after operating with A†f . Therefore we define the symmetrizing
operator first (on any Hn , n > 1) as

1 X
Sn = U (Pn ),
n! Pn

where Pn are the n! permutations of n objects.

With this definition any arbitrary vector of HSn can be written as

1 X
Sn (f1 ⊗ · · · ⊗ fn ) = fP (1) ⊗ · · · ⊗ fPn (n)
n! Pn n

for some vectors f1 , . . . , fn ∈ H.

The creation operator can now be defined as

a†f Sn−1 (f1 ⊗ · · · ⊗ fn−1 ) = Sn A†f Sn−1 (f1 ⊗ · · · ⊗ fn−1 ),

or, more compactly as

a†f = Sn A†f Sn−1 .

Notice we can also write, in a similar way the annihilation operator as

af = Sn−1 Af Sn

which automatically takes care of af being the restriction of Af to symmetric


spaces.

Exercise 4 Show that the symmetrizer Sn is a projection operator: that is, it


is self-adjoint, and Sn2 = Sn .

Hint: it is easier to first prove that operators U (P ) are unitary, and then
write Sn = (1/n!) P U (P ). On taking the adjoint, U (P )† = U (P )−1 and the
P

sum contains both U (P ) as well U (P )−1 .


24CHAPTER 2. INTRODUCTION TO FOCK SPACE AND SECOND QUANTIZATION
These expression also prove that a†f is the adjoint of af because the pro-
jection operators are self-adjoint:
(af )† = (Sn−1 Af Sn )† = Sn† A†f Sn−1

= Sn A†f Sn−1 = a†f

§ 18: Properties of af and a†f

The most important property of these operators are summarized below:

a†f is linear in f

a†f +g = a†f + a†g , (2.1)


a†cf = c a†f . (2.2)

af is anti-linear in f
af +g = af + ag , (2.3)
acf = c∗ af . (2.4)

For all f ∈ H,
af Φ0 = 0, (2.5)
af g = (f, g)Φ0 , (2.6)
a†f Φ0 = f. (2.7)

The annihilation and creation operators act as


1 X ∼
ag Sn (f1 ⊗ · · · ⊗ fn ) = √ (g, fi ) Sn−1 (f1 ⊗ · · · ⊗ fi ⊗ · · · ⊗ fn ) (2.8)
n i

a†g Sn (f1 ⊗ · · · ⊗ fn ) = n + 1 Sn+1 (g ⊗ f1 ⊗ · · · ⊗ fn ) (2.9)
(The ∼ on a vector in a tensor product is used to indicate that that factor is
to be omitted.)

§ 19: Commutation Relations

For all f, g ∈ H
[af , ag ] = af a†g − a†g af = 0, (2.10)
h i
a†f , a†g = a†f a†g − a†g a†f = 0, (2.11)
h i
af , a†g = af a†g − a†g af = (f, g)1. (2.12)
25
Here (f, g) denotes the inner product of the Hilbert space H.

Proof: Use (2.18) and (2.9)



a†g Sn (f1 ⊗ · · · ⊗ fn ) = n + 1 Sn+1 (g ⊗ f1 ⊗ · · · ⊗ fn )
af a†g Sn (f1 ⊗ · · · ⊗ fn ) = (f, g)Sn (f1 ⊗ · · · ⊗ fn )
X ∼
+ (f, fi )Sn (g ⊗ f1 ⊗ · · · ⊗ fi ⊗ · · · ⊗ fn )
i

On the other hand,


1 X ∼
a†g af Sn (f1 ⊗ · · · ⊗ fn ) = a†g √ (f, fi ) Sn−1 (f1 ⊗ · · · ⊗ fi ⊗ · · · ⊗ fn )
n i
X ∼
= (f, fi ) Sn (g ⊗ f1 ⊗ · · · ⊗ fi ⊗ · · · ⊗ fn )
i

Therefore

af a†g − a†g af = (f, g)1.

§ 20: Orthonormal bases in Fock space

An orthonormal basis {ei }, i = 1, 2, 3. . . . in H

(ei , ej ) = δij ,

can be used to construct a basis for the Fock space. First of all, the vacuum
is a normalised vector

(Φ0 , Φ0 ) = 1,

and already orthogonal to all subspaces containing 1, 2, 3 etc particles. {ei }


is the basis for H, the subspace of 1 particle. Next we define

eii = ei ⊗ ei , i = 1, 2, 3, . . .
1
ei1 i2 = √ (ei1 ⊗ ei2 + ei2 ⊗ ei1 ), i1 < i2 , i1 , i2 = 1, 2, 3, . . .
2!
Note that i1 < i2 and we do not define a vector ei2 i1 because that would
be equal to ei1 i2 and not linearly independent. It is easy to see that ei1 i2 is
normalised, and orthogonal to all other ej1 j2 with different values for the pair
of indices.

Bases for higher spaces is discussed as exercise below.


26CHAPTER 2. INTRODUCTION TO FOCK SPACE AND SECOND QUANTIZATION
Exercise 5 Basis for the Symmetric Subspaces Hsn
Start with the basis (1.3) above for Hn . Let
1 X
G = S(n) ei1 ⊗ · · · ⊗ ein = e ⊗ · · · ⊗ eP (in ) (2.13)
n! P P (i1 )

The norm square of G is easily seen to be,


nj1 ! . . . njr !
kGk2 = (2.14)
n!
where indices {i1 , . . . , in } are actually r different types of indices j1 , . . . , jr repeated
nj1 , . . . , njr times respectively, nj1 + · · · + njr = n. Therefore the normalised vectors
1
E(j1 , nj1 , . . . , jr , njr ) ≡ (n!/nj1 ! . . . njr !) 2 G (2.15)
− 12
X
≡ (n!nj1 ! . . . njr !) eP (i1 ) ⊗ · · · ⊗ eP (in ) (2.16)
P

form an orthonormal basis; two vectors being orthogonal unless they have the same
type of indices repeated the same number of times. We can further simplify the
notation and allow ni to take values zero as well. Then the basis vectors of Hsn are
labelled by a sequence n1 , n2 , . . . ni , . . . with ni = 0 for all i except for i = j1 , . . . jr
for some r ( 1 ≤ r ≤ n ) and nj1 + · · · + njr = n. We denote such a vector by
Φs (n1 , . . . , ni , . . .). It is given by the vector E(j1 , nj1 , . . . , jr , njr ) above :

1
Φs (n1 , . . . , ni , . . .) ≡ (n!nj1 ! . . . njr !)− 2
X
eP (j1 ) ⊗ · · · ⊗ eP (jr ) (2.17)
P

where the tensor product on the right consists of all the n vectors ej1 , (nj1 times) ,
ej2 , (nj2 times) etc.

§ 21: The anti-symmetric Fock space FA

The anti-symmetric Fock space FA ⊂ F is the anti-symmetric subspace of the


big Fock space. It is composed of the following sub-spaces

1. the one dimensional space H0 = {Φ0 } of zero- or no particles, with a


distinguished vector Φ0 called the vacuum,
2. the space H of one particles,
2
3. the subspace HA = (H ⊗ H)A of all anti-symmetric (under interchange)
2
of vectors of H = H ⊗ H,
4. the subspace HS3 = (H ⊗ H ⊗ H)A of all anti-symmetric (under permu-
tation) of vectors of H3 = H ⊗ H ⊗ H,
27
and so on...

All these subspaces are orthogonal to each other.

We denote the restriction of Af to the anti-symmetric subspace FA by af ,


as in the symmetric case. It maps anti-symmetric states into such states.

On the other hand, a creation operator A†f does not map an anti-symmetric
vector in Hn−1 into an anti-symmetric vector in Hn . Therefore we define the
anti-symmetrizing operator first (on any Hn , n > 1) as
1 X
An = (−1)Pn U (Pn ),
n! Pn

where Pn are the n! permutations of n objects.

With this definition any arbitrary vector of HSn can be written as


1 X
An (f1 ⊗ · · · ⊗ fn ) = (−1)Pn (fPn (1) ⊗ · · · ⊗ fPn (n) )
n! Pn

for some vectors f1 , . . . , fn ∈ H.

The creation operator can now be defined as

a†f An−1 (f1 ⊗ · · · ⊗ fn−1 ) = An A†f An−1 (f1 ⊗ · · · ⊗ fn−1 ),

or, more compactly as

a†f = An A†f An−1 .

Notice we can also write, in a similar way the annihilation operator as

af = An−1 Af An

which automatically takes care of af being the restriction of Af to anti-


symmetric spaces.

Exercise 6 Show that the anti-symmetrizer An is a projection operator: that is,


it is self-adjoint, and A2n = An .

These expression also prove that a†f is the adjoint of af because the pro-
jection operators are self-adjoint:

§ 22: Properties of af and a†f


28CHAPTER 2. INTRODUCTION TO FOCK SPACE AND SECOND QUANTIZATION
The most important property of these operators are summarized below:

a†f is linear in f

a†f +g = a†f + a†g ,


a†cf = c a†f .

af is anti-linear in f
af +g = af + ag ,
acf = c∗ af .

For all f ∈ H,
af Φ0 = 0,
af g = (f, g)Φ0 ,
a†f Φ0 = f.

The annihilation and creation operators act as


1 X ∼
ag An (f1 ⊗ · · · ⊗ fn ) = √ (g, fi )(−1)i−1 An−1 (f1 ⊗ · · · ⊗ fi ⊗ · · · ⊗ fn )
n i

a†g An (f1 ⊗ · · · ⊗ fn ) = n + 1 An+1 (g ⊗ f1 ⊗ · · · ⊗ fn )
(The ∼ on a vector in a tensor product is used to indicate that that factor is
to be omitted.)

bf b∗g A(n−1) f1 ⊗ · · · ⊗ fn−1



= nbf A(n) g ⊗ f1 ⊗ · · · ⊗ fn−1
= (f, g)A(n−1) f1 ⊗ · · · ⊗ fn−1
n−1
(−1)i+1 (f, fi )A(n−1) g ⊗ f1 ⊗ · · · f/i · · · ⊗ fn−1
X
+
1

while,
b∗g bf A(n−1) f1 ⊗ · · · ⊗ fn−1
n−1
1
= b∗g √ (f, fi )(−1)i A(n−2) f1 ⊗ · · · f/i · · · ⊗ fn−1
X
n−1 i
n−1
X
= (f, fi )A(n−1) g ⊗ f1 ⊗ · · · f/i · · · ⊗ fn−1
i
29
This is similar to the symmetric case except for the signs (−1)i+1 and (−1)i
the first one having one more power of (−1) because the creation operator had
increased the number of factors in the tensor product by one. This leads to
the anticommutation relation :

{bf , b∗g } ≡ bf b∗g + b∗g bf = (f, g) (2.18)

Exercise 7 Verify the anticommutation relations

{bf , bg } = 0 (2.19)

{b∗f , b∗g } = 0 (2.20)

Hint : bf bg acting on A(n) f1 ⊗ · · · ⊗ fn gives (aprt from constant factors)


X
(±)(g, fi )(f, fj )A(n−2) (f1 ⊗ · · · ⊗ fn )
i,j

with both fi and fj missing from the tensor product. The sign ± is actually equal
to (−1)i+j for j < i and (−1)i+j+1 for j > i. For bg bf the expression is the same
except that the signs are opposite.

Exercise 8 Basis for the Antisymmetric Subspaces Han


The orthonormal basis for the antisymmetric subspace is much simpler compared
to the symmetric case. A(n) ei1 ⊗ · · · ⊗ ein is actually zero if any of the indices is
repeated. The orthonormal basis can therefore be chosen
1
F (i1 , . . . , in ) ≡ (n!)− 2
X
(−)P ei1 ⊗ · · · ⊗ ein
P

for all different indices i1 , . . . , in .

Notice that for any P

F (P (i1 ), . . . , P (in )) = (−)P F (i1 , . . . , in )

so that in order to label only the independent basis vectors we must use some
convention like i1 < i2 < . . . < in .

We can also use a notation similar to Φs for the symmetric case. In the present
case ni ’s take values zero or one. If ni = 1 for i = i1 < i2 < . . . < in then
1
Φa (n1 , . . . , ni , . . .) ≡ (n!)− 2
X
(−)P ei1 ⊗ · · · ⊗ ein
P

§ 23: Observables in Fock space


30CHAPTER 2. INTRODUCTION TO FOCK SPACE AND SECOND QUANTIZATION
Recall that when two quantum systems with Hilbert spaces H1 and H2 form
a single system, the states of the joint system are vectors in H1 ⊗ H2 , which
are linear combinations of vectors of the type f1 ⊗ f2 . In order to define any
operators (and in particular, observables) on this space it is enough to define
them on these ‘factorisable’ vectors first and later on extend them on other
vectors in H1 ⊗ H2 using linearity.

Of importance are observables which refer to only one of the systems, for
example the energy H1 of one system alone. The operator corresponding to it
in H1 ⊗ H2 is H1 ⊗ 1 defined by
(H1 ⊗ 1)(f1 ⊗ f2 ) = (H1 f1 ) ⊗ f2 .
Similarly for the second particle and the sum of the individual energies will
then be written as H1 ⊗ 1 + H1 1 ⊗ H2 , where H2 is the energy operator of
the second system. These are observables of the two-particle system which are
sum of individual one-particle observables. Not all operators are of this type
though. The interaction between the two systems will be an operator which
cannot be decomposed as sum of such operators.

For a system of identical bosonic particles the sum of one particle observable
A is of the type
A ⊗ 1 ⊗ · · · ⊗ 1 + · · · + 1 ⊗ · · · ⊗ A ⊗ · · · ⊗ 1 + · · · 1 ⊗ · · · ⊗ A,
for the n-particle subspace.

The method of “second quantization” offers a very compact and power-


ful expression valid for all n. Let us choose an orthonormal basis {ei }, i =
1, 2, 3 . . .. Let
Aij ≡ (ei , Aej )
be the matrix elements of the one-particle observable A, and let us write
ai ≡ ae i , a†i ≡ a†ei .
Then the observable
X †
A= ai Aij aj
ij

has the right form for every subspace Hsn .

Proof: We calculate
X † 1 X ∼
ASn (f1 ⊗ · · · ⊗ fn ) = ai Aij √ (ej , fk )Sn−1 (f1 ⊗ · · · ⊗ fk ⊗ · · · ⊗ fn )
ij n k
31
X † 1 X ∼
= ai Aij √ (ej , fk )Sn−1 (f1 ⊗ · · · ⊗ fk ⊗ · · · ⊗ fn )
ij n k
X X ∼
= Aij (ej , fk )Sn (ei ⊗ f1 ⊗ · · · ⊗ fk ⊗ · · · ⊗ fn )
ij k
X  X ∼ 
= Sn (Aij (ej , fk )ei ) ⊗ f1 ⊗ · · · ⊗ fk ⊗ · · · ⊗ fn .
ij k

But
X X X X
Afk = A (ej , fk )ej = (ej , fk )Aej = (ej , fk )(ei , Aej )ei = Aij (ej , fk )ei .
j j ij ij

Therefore,
X  ∼ 
ASn (f1 ⊗ · · · ⊗ fn ) = Sn Afk ⊗ f1 ⊗ · · · ⊗ fk ⊗ · · · ⊗ fn
k
X  
= Sn f1 ⊗ · · · ⊗ Afk ⊗ · · · ⊗ fn
k

The simplest observable in Fock space is the number operator which is


generated by the constant operator number A = 1. As Aij = δij , in this case,
the Fock space observable corresponding to 1 is just
X †
N= ai ai . (2.21)
i

It is easy to see that it takes the value n on HSn :


(a†i ai )Sn (f1 ⊗ · · · ⊗ fn ) =
X X
Sn (f1 ⊗ · · · ⊗ 1.fk ⊗ · · · ⊗ fn )
i k
= nSn (f1 ⊗ · · · ⊗ fn ).
We will discuss more about it in the next chapter.

Exercise 9 Given f, g, h ∈ H as orthogonal and normalised vectors, calculate the


norm of the following vectors:
a†f Φ0 , a†f a†g Φ0 , a†h a†f a†g Φ0

Exercise 10 Repeat the previous exercise without assuming orthonormality. An-


swers:
||f ||2 , ||f ||2 ||g||2 + |(f, g)|2 ,
and
||f ||2 ||g||2 ||h||2 + ||f ||2 |(g, h)|2 + ||g||2 |(h, f )|2 + ||h||2 |(f, g)|2 + 2<[(f, g)(g, h)(h, f )].
(It is interesting to observe the symmetry of the last expression under permutation
of f, g, h)
32CHAPTER 2. INTRODUCTION TO FOCK SPACE AND SECOND QUANTIZATION
Chapter 3

Non-relativistic field theory

§ 24: Number Operator

The formula which raises a one-particle observable G to a Fock space observable


G is
X †
G= ai Gij aj (3.1)
i,j

where {ei }, i = 1, 2, . . . is an orthonormal basis in the one particle space H and

ai ≡ aei .

Exercise 11 Show that the definition of A is independent of the orthonormal basis


chosen to define it.

The simplest observable is the identity operator 1. For it (ei , 1ej ) = δij
and thus
X †
N ≡1= ai ai .
i

This is called the number operator because acting on Hsn it simply acts as

1 ⊗ · · · ⊗ 1 + . . . + 1 ⊗ · · · ⊗ 1 = n(1 ⊗ · · · ⊗ 1).


Exercise 12 Verify directly by acting on a vector in Hsn .
P
i ai ai

The operators

Nk = a†k ak , k = 1, 2, . . .

33
34 CHAPTER 3. NON-RELATIVISTIC FIELD THEORY
are the number operators for each basis state ek . They all commute with each
other, and can be simultaneously diagonalised. Their simultaneous eigenstates
are called the occupation number representation of the Fock space. We will
talk more about it (when and if) we need it.

Exercise 13 Let f ∈ H be a unit norm vector (||f || = 1), and let g1 , . . . gn be a


finite orthonormal set of distinct vectors. Define

Nf = a†f af

as the number operator corresponding to f . Show that the expectation value of Nk


in the state Φ = Sn (g1 ⊗ · · · ⊗ gn ) is
n
(Φ, Nf Φ) X
= |(f, gi )|2 .
(Φ, Φ) i=1

Exercise 14 Let G be a one particle observable (a self adjoint operator) and U =


(iG) the unitary operator corresponding to it. Define

U = exp(iG).

where G is defined (as above)


X †
G= ai Gij aj .
i,j

Prove that
(1)

[G, af ] = −aGf
h i
G, a†f = a†Gf

(2)

U af U −1 = aU f
U a†f U −1 = a†U f .

Hint: use linearity of a†f and anti-linearity of af in f , as well as the identity

1
exp(S)A exp(−S) = 1 + [S, A] + [S, [S, A]] + · · · .
2!

§ 25: Non-relativistic ‘Potential’


35
Let a boson of mass m be described by vectors in a Hilbert space H which
consists of wave function made from momentum basis |pi with

hp|p0 i = δ(p − p0 ).

Let the annihilation and creation operators corresponding to |pi be denoted


by

a(p) ≡ a|pi , a† (p) ≡ a†|pi

then the free Hamiltonian for these particles in the bosonic Fock space is given
by

p2
Z !
H0 = d3 p mc2 + a† (p)a(p)
2m

The particles interact with a background source which can absorb or emit
these particles.

A simple form of a term in the interaction can be for example


Z
d3 p d3 k v(p, k)a† (p)a(k)

where v(p, k) is a complex number representing the ‘amplitude’ for the process
which absorbs the particle with momentum k and emits the particle with
changed momentum p and takes away the balance of momentum and energy.
It is understood that the rest mass energy mc2 is so large that there is no
serious violation of energy.

The above term, as it stands, cannot be added to the Hamiltonian because


it is not self-adjoint. We must add its adjoint to get a hermitian interaction
Hamiltonian.
Z  
HI = d3 pd3 k v(p, k)a† (p)a(k) + v ∗ (p, k)a† (k)a(p)

we can rewrite it as
Z
HI = d3 pd3 k (v(p, k) + v ∗ (k, p)) a† (p)a(k)

§ 26: Galilean invariance

Seen from a moving frame with velocity v, k becomes k − mv and p becomes


p − mv. Physically we expect the amplitude to make a transition from k to p
to be the same as from k − mv to p − mv. HI can be invariant under Galilean
transformations if v(p, k) is a function of p − k.
36 CHAPTER 3. NON-RELATIVISTIC FIELD THEORY
Thus HI has the form
Z
HI = d3 pd3 k w(p − k)a† (p)a(k)

with w a function with property w∗ (q) = w(−q) .

§ 27: Configuration space

Let us define the position space eigenstates


1 Z
|xi = d3 p exp[−ip.x/h̄] |pi
(2πh̄)3/2

Using the anti-linear property of annihilation operators (af +g = af + ag and


acf = c∗ af ) we obtain
1 Z
a(x) ≡ a|xi = d3 p exp[ip.x/h̄] a(p)
(2πh̄)3/2
Note that we can differentiate and write
1 Z
−ih̄∇a(x) = d3 p exp[ip.x/h̄] pa(p)
(2πh̄)3/2
Similar formulas can be written for the creation operator.

We can use the position space operators to write the Hamiltonian using
this formula
1 2
Z  
3 † 2
H= d x a (x) mc − ∇ + V (x) a(x)
2m
where
Z
V (x) = d3 q w(q) exp[iq.x/h̄]

§ 28: A and B ‘particles’

In this section we introduce interaction between two species of particles


and illustrate the fundamental fact of quantum field theory that exchange of
a quantum between two particles leads to interaction between them.

We discuss a simple, exactly solvable model which shows the basic mech-
anism by which quanta interact. We take the grossly simplified model where
the particles have just one dimensional Hilbert space.
37
Let there be two types of ‘particles’ : a boson called B and described by
operators b, b† and a boson called A with a, a† .

In this simple sense they are just two oscillators

[b, b† ] = 1, [a, a† ] = 1, [a, b] = 0, [a, b† ] = 0

with a ‘free’ that is non-interacting Hamiltonian

H0 = Ωb† b + ωa† a (3.2)

The interpretation of the n-th energy nω or nΩ of the oscillators A or B is


that there are n quanta of the type A or B.

The total Hamiltonian is written as

H = H0 + λb† b(a + a† ) (3.3)

the second term being the interaction term containing observables from both
the particles.

So far we have dealt with the Fock spaces which describe the quantum
states of just one type particle. Now that we are discussing interaction, we
need Fock spaces which describe both the species of the particle. We should
use the tensor product of the two Fock spaces to describe the interacting states.

The vacuum state of free Hamiltonian is defined by

bΦ0 = 0, aΦ0 = 0 (3.4)

This just means that Φ0 is the tensor product of the two vacuum states :
Φ0 = ΦA ⊗ΦB . The vacuum state defined as the state on which all annihilation
operator act and give zero, is also the lowest energy state

H0 Φ0 = 0

This state also happens to be the eigenstate of the total Hamiltonian.

HΦ0 = 0

which follows because a† and b commute so that in the term b† ba† the operator
b can be taken to the right of a† and made to act on Φ0 . In the general case in
quantum field theory the ‘free’ vacuum state is not an eigenstate of the total
Hamiltonian.

§ 29: Bare and Dressed Particles


38 CHAPTER 3. NON-RELATIVISTIC FIELD THEORY
The number operators are NB = b† b and NA = a† a. We see that

[H, NB ] = 0

This means that under time evolution by the total Hamiltonian a state with
some fixed number of the number of B quanta retains this number. That is not
the case with A quanta because [H, NA ] 6= 0. Another way to look at it is as
follows. Under an infitesimal time evolution ψ → (1 − idtH)ψ the interaction
term λb† b(a + a† ) causes annihilation or creation of A-quanta but keeps the
number of B-quanta the same. As a result the state which is an eigenstate of
H containing one B-quantum acquires a ‘cloud’ of A-quanta around it. For
this reason it is called a dressed state. In contrast the eigenstates of H0 are
said to contain bare B- and A-quanta.

§ 30: Single dressed B-particle

We now calculate the one quantum dressed B-state. We know that it is a


state that contains one bare B quantum and will have any number of A-quanta
in the cloud surrounding it. Let

HΨ = EΨ

Ψ = (d0 + d1 a† + d2 (a† )2 + . . .)b† Φ0

The actual calculation is given below.The one B-particle dressed state is an


eigenstate of the total Hamiltonian given by

Ψ = exp[−λ2 /2ω 2 ] exp[−λa† /ω]b† Φ0 (3.5)

with eigenvalue

E = Ω − λ2 /ω (3.6)

§ 31: Calculation

We substitute the proposed form of Ψ into the eigenvalue equation and compare the
coefficients of various powers of a† .

(H − E)Ψ = (Ω − E)(d0 + d1 a† + d2 (a† )2 + . . .)b† Φ0


+ω(d1 a† + 2d2 (a† )2 + . . .)b† Φ0
+λ(d1 + 2d2 a† + 3d3 (a† )2 + . . .)b† Φ0
+λ(d0 a† + d1 (a† )2 + . . .)b† Φ0
39
This gives the sequence of equations, the first three of which are

(Ω − E)d0 + λd1 = 0
(Ω − E)d1 + ωd1 + λ(2d2 + d0 ) = 0
(Ω − E)d2 + ω2d2 + λ(3d3 + d1 ) = 0

If we argue from a perturbation theory point of view, then as λ → 0, the state Ψ must
become the free or bare one B-particle state proportional to b† Φ0 . Therefore all di except
d0 must go to zero as λ → 0. In fact we expect d1 = O(λ), d2 = O(λ2 ) etc.

The first equation of the sequence of equations then tells us that (Ω − E) = O(λ2 ).
A look at the second equation tells us that it consists of terms of O(λ3 ) and O(λ), which
should seperately be equated to zero. Therefore

ωd1 + λd0 = 0
(Ω − E)d1 + λ2d2 = 0

This determines
λ
d1 = − d0
ω
which determines the value of (Ω − E) from the first equation

E = Ω − λ2 /ω

as well as
 2
1 λ
d2 = − d0
2! ω
In the third equation there are second and fourth order terms in λ. The second order terms
ω2d2 + λd1 is identically zero, the fourth order give
 3
1 λ
d3 = − d0
3! ω
The general solution is


λ † †
Ψ = d0 exp − a b Φ0
ω
The normalization constant d0 can be determined as follows

1 = (Ψ, Ψ) = |d0 |2 (Φ0 , exp[−λa/ω] exp[−λa† /ω]bb† Φ0 )


= |d0 |2 exp[λ2 /ω 2 ]

where use the identity

exp(A) exp(B) = exp([A, B]) exp(B) exp(A)

Thus the one B-particle dressed state an eigenstate of the total Hamiltonianis given by

Ψ = exp[−λ2 /2ω 2 ] exp[−λa† /ω]b† Φ0 (3.7)

with eigenvalue

E = Ω − λ2 /ω (3.8)
40 CHAPTER 3. NON-RELATIVISTIC FIELD THEORY
§ 32: B-B effective interaction

Now we take two B-particle states. In perturbation theory the second order
term in interaction will involve creation of an A-quantum by one B-particle
and its absorption by the other. This back and forth exchange A by B particle
leads to an interaction between B particles whithout any reference to A. The
A-exchange can be integerated out and replaced by an effective B-B interation.
Let us see how this happens in the extremely simplified model we are studying.

First notice that the one B-particle dressed state is obtained from the bare
one particle state b† Φ0 by operating by exp[−λ2 /2ω 2 ] exp[−λa† /ω]. We do a
little manipulation which is worth observing carefully :
Ψ = exp[−λ2 /2ω 2 ] exp[−λa† /ω]b† Φ0
= exp[−λ2 /2ω 2 ] exp[−λa† /ω] exp[λa/ω]b† Φ0
because exp[λa/ω]Φ0 = Φ0 . Now use
exp(A) exp(B) = exp([A, B]/2) exp(A + B)
which holds whenever [A, B] commutes with both A and B to get
Ψ = exp[λ(a − a† )/ω]b†
The advantage of this manipulation is that the operator exp[λ(a − a† )/ω] is
unitary.

The two particle dressed-state will similarly involve exp[2λ(a−a† )/ω] acting
on the two particle bare state because there are two factors of exp[λ(a−a† )/ω].
Similarly the three particle dressed state will be a similar operator with 2
replaced by 3 in the exponent. Let us define a unitary operator
" #
λ
U = exp b† b(a − a† )
ω

so that it gives the right factor in the exponent because b† b is the number
oerator. This unitary operator changes the bare to dressed states. If we apply
it to all our observables, then the transformed operators are
" # " #
λ λ
ã = U aU = exp b† b(a − a† ) a exp − b† b(a − a† )

ω ω
λ
= a + b† b
ω
where we use the identity
1
exp[S]A exp[−S] = A + [S, A] + [S, [S, A]] + . . .
2!
41
Similarly
λ †
ㆠ= a† + bb
ω
and
" # " #
λ λ
b̃ = exp b† b(a − a† ) b exp − b† b(a − a† )
ω ω
!2
λ 1 λ
= b − (a − a† )b + (a − a† )2 b − . . .
ω 2! ω
λ
= exp[− (a − a† )]b
ω
and so
λ
b̃† = b† exp[ (a − a† )]
ω
We can transform the Hamiltonian in terms of these new operators, Rewriting
the inverted formulas
λ †
a = ã − b̃ b̃
ω
λ
a† = ㆠ− b̃† b̃
ω
λ
b = exp[ (a − a† )]b̃
ω
λ
b† = b̃† exp[− (a − a† )]
ω
Therefore,

H = Ωb† b + ωa† a + λb† b(a + a† )


λ2 † λ2
!
= Ω−2 b̃ b̃ + ωㆠã − 2 b̃† b̃† b̃b̃
ω ω

The Hamiltonian has separated into dressed particles created by b̃† which in-
teract with themselves by the b̃† b̃† b̃b̃ term and a species of free bosons which
are created by ㆠ.
42 CHAPTER 3. NON-RELATIVISTIC FIELD THEORY
Chapter 4

Relativistic spin zero particles


and fields

From now on we use natural units such that the values of c, the
speed of light and h̄, the Planck’s constant, have numerical values
equal to 1. This means there is only one fundamental unit which
can be taken to be either mass, or energy units (e.g. MeV), or
length units (e.g. fm or Fermi). One has to be careful however:
a particle of mass
√ 1 MeV, which has a momentum of 1 MeV has
energy equal to 2 MeV.

§ 33: Mass m spin zero, neutral particle

A single mass m spin zero particle is described by momentum space basis states
{|pi}, where we now use the 4-vector notation
q
p = (p0 = ωp , p), ωp = + p2 + m2 .

ˆ
If P̂ 0 and P~ are the operators in the one-particle Hilbert space H, then
q
P̂ 0 |pi = + p2 + m2 )|pi,
P̂|pi = p|pi.

These states define the inner product, in the Hilbert space H of the particle.

hp|p0 i = 2ωp δ 3 (p − p0 ),
Z 3
dp
hf |gi = f (p)∗ g(p), f (p) = hp|f i, g(p) = hp|gi.
2ωp

43
44 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
The choice of the inner product is dictated by relativistic invariance. That is,
if Λ is a 4 × 4 Lorentz matrix, then
Z 3~
Z
d3 p d Λp
= .
2ωp 2ωΛp
~

Exercise 15 Prove the relativistic invariance of d3 p/2ωp .

§ 34: Representation of the Poincare Group

For a Lorentz transformation matrix Λ

U (Λ)|pi = |Λpi,

and for a space-time translation given by the four vector a = (a0 = ct, a)

U (a)|pi = exp(iωp a0 − ip · a)|pi.

We can write this last equation as

U (a)|pi = exp(−ip · a)|pi

where

p · a = −p0 a0 + p · a = ηµν pµ aν , ηµν = diag(−1, 1, 1, 1)

§ 35: Covariantly transforming states

Let x = (x0 , x1 , x2 , x3 ) = (t, x) be any space time point. Define a state

1 Z d3 p
|xi = exp(−ip · x)|pi, (4.1)
(2π)3/2 2ωp

then,

U (Λ)|xi = |Λxi,

and

U (a)|xi = |x + ai.

Exercise 16 Prove above relations.


45
§ 36: Klein-Gordon equation

Let f ∈ H. Construct

1 Z d3 p
f (x) = hx|f i = exp(ip · x)hp|f i.
(2π)3/2 2ωp

Then
∂2
!
− 2 + ∇2 − m2 f (x) = 0.
∂t

Every vector |f i of H has such a ‘wave-function’ associated with it. These are
called ‘positive-frequency’ or ‘positive-energy’ solutions which is a historical
name.

Remark 3 We have used f (x) for hx|f i, and f (p) for hp|f i to avoid clumsiness. Of
course, these are not the same function of their arguments. This should not cause
any confusion.

§ 37: States |xi are not orthogonal to each other.

The states |xi are not the eigenstates of some observable! x is merely a
label. If x were eigenvalues of an observable then for distinct values of x and
y the states |xi and |yi would have been orthogonal. As it happens

1 Z d3 p
hx|yi = exp[ip · (x − y)]
(2π)3 2ωp

For simplicity let us take x0 = y 0 , then

1 Z d3 p
hx|yi = exp[ip · (x − y)]
(2π)3 2ωp
1 Z∞ √
= dkk k 2 + m2 sin(kr)
π2r 0

where r = |x − y|. This function is not zero for r 6= 0, but can be shown to
fall off exponetially with kr. This means that the significant non-zero part is
in a region of the Compton wavelength (r ∼ 1/m = h̄/mc ).

§ 38: {|xi} form a complete set for a fixed x0


46 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
The |xi restricted to the subset x0 fixed to some constant value form a complete
set, although they are not orthogonal. We can see this by expressing |pi in
terms of |xi with x0 fixed. Since every vector can be expanded in |pi’s and
|pi’s can be expanded in the |xi’s (for a fixed x0 ), it follows that |xi’s form a
complete set.

Multiply the defining equation (4.1) by (2π)−3/2 exp(ik.x) and integrate


over d3 x.
1 Z d3 p Z 3 i(p0 −k0 )x0 −i(p−k)·x
d xe e
(2π)3 2ωp
The x integration gives a delts function in p − k and momentum integration
makes p0 equal to k 0 . Thus
2k 0 Z 3 ik.x
|ki = d xe |xi.
(2π)3/2
Since every vector in H can be expanded in basis {|ki}, this formula can be
used to ecpress them as linear combination of |xi, x0 = constant.

§ 39: Klein-Gordon equation

The functions
f (x) = hx|f i
satisfy (with x = (x0 = t, x)),
∂2
!
− ∇2 + m2 f (x) = 0.
∂t2

Exercise 17 Prove the above equation.

§ 40: Inner product in |xi basis wave-functions

For any two vectors |f i, |gi ∈ H


∂g(x) ∂f ∗ (x)
Z !
3 ∗
i d x f (x) − g(x) = hf |gi.
x0 ∂x0 ∂x0
This can be seen by just evaluating the left hand side. Actually the two terms
are equal and one could equally well write
!
Z
∂g(x)
hf |gi = 2i d3 x f ∗ (x) ,
x 0 ∂x0
47
or,

∂f ∗ (x)
Z !
3
hf |gi = −2i dx g(x) .
x0 ∂x0

This allows us to write the completeness relation in the x-basis.


 → ← 
Z
∂ ∂ 
i d3 x |xi  0 − hx| = 1.
x0 ∂x ∂x0

Note that, despite its appearance, the inner product is independent of


which x0 is chosen to evaluate the integral. This follows from the Klein-Gordon
equation:

∂g(x) ∂f ∗ (x)
!
∂ ∂ Z 3 ∗
hf |gi = i 0 d x f (x) − g(x)
∂x0 ∂x x0 ∂x0 ∂x0
∂ 2 g(x) ∂ 2 f ∗ (x)
Z !
3 ∗
= i d x f (x) − g(x)
x0 ∂(x0 )2 ∂(x0 )2
Z  
= i d3 x f ∗ (x)∇2 g(x) − ∇2 f ∗ (x)g(x) .
x0

This is zero because if we do integration by parts and transfer the ∇ from one
to the other factor.

Exercise 18 Let P̂ µ = (P̂ 0 , P̂) be the energy and momentum operators in H.


Show that

hx|P̂ 0 |f i = i f (x)
∂t
hx|P̂|f i = −i∇f (x)

§ 41: A note about the inner-product in |xi basis wave-functions

The wave-functions f (x) = hx|f i are positive energy (or even ‘positive fre-
quency’) wave-functions. This is because the definition of |xi states is through
the basis |pi which all correspond to positive energy. And by construction
these wave functions satisfy the Klein-Gordon equation. Historically it was
the Klein-Gordon equation which was established first. Any function of x can
be written in the form
Z
0 x0 +ip·x
ψ(x) = d4 pe−ip f (p).
48 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
0 2
This will satisfy the Klein-Gordon
√ 2 equation provided (p√) = p2 + m2 . This
0 0
would include both p = + p + m2 as well as p = − p2 + m2 . The neg-
ative energy solutions are just the complex conjugate of some positive energy
solution.

We have constructed the inner product above for positive energy states. If
we use the same inner product for negative energy states then the inner product
is not positive definite. Moreover, the two types of solutions are orthogonal to
each other for this inner product.

Let f, g ∈ H be two positive energy solutions. Then f ∗ is a negative


energy solution. And the expression for the inner product will be (remember
(f ∗ )∗ = f ),
!

Z
3 ∂g(x) ∂f (x)
(f , g) = i d x f (x) − g(x)
x0 ∂x0 ∂x0
Z 3 Z 3
Z
dp dk 0 0 0 0
= d3 x f (p)e−ip x +ip·x (k 0 )e−ik x +ik·x g(k)
x0 2ωp 2ωk
Z 3 Z 3
Z
dp dk 0 0 0 0
− d3 x f (p)(p0 )e−ip x +ip·x e−ik x +ik·x g(k)
x0 2ωp 2ωk

The x integration will give a delta factor δ(p + k), which allows k-integration
replacing k with −p. The two terms then cancel.

§ 42: Observable matrices in |xi basis ‘wave-functions’

We prove the following important formulas:

∂f ∗ (x) ∂g(x)
Z " #
0
hf |P̂ |gi = dx 3
+ ∇f ∗ (x) · ∇g(x) + m2 f ∗ (x)g(x)
x0 ∂x0 ∂x0
∂g(x) ∂f ∗ (x)
Z " #
3 ∗
hf |P̂|gi =− d x ∇f (x) + ∇g(x)
x0 ∂x0 ∂x0

Proof:

 → ← 
Z
∂ ∂ 
hf |P 0 |gi = d3 x hf |xi  0 − hx|P 0 |gi
x0 ∂x ∂x0
 → ←  !
Z
3 ∂ ∂  ∂g(x)
= d x hf |xi  0 − i
x 0 ∂x ∂x0 ∂x0
49
using the result of the previous section. As g(x) satisfies the Klein-Gordon
equation , we can use it as
∂2
g(x) = ∇2 g(x) − m2 g(x).
∂(x0 )2
Moreover, an integration by part allows transfer of one ∇ from second factor
to first, after throwing away the surface term because the functions f (x) and
g(x) vanish at spatial infinity. Thus we obtain the expression given above.
The second equation is similarly proved.

§ 43: Pauli-Jordan Function

The function
1 Z d3 p ip.(x−y)
i∆m (x − y) = hx|yi − hy|xi = 3
[e − e−ip.(x−y) ]
(2π) 2ωp
is called the Pauli-Jordan invariant function. It has the following properties
which can be easily seen from the definition.

1. It is Lorentz invariant.

∆m (Λx − Λy) = ∆m (x − y)

This because
Z
d3 p ip.Λ(x−y) Z 3
d p iΛ−1 p.(x−y) −1
[e − e−ip.Λ(x−y) ] = [e − e−iΛ p.(x−y) ]
2ωp 2ωp
Now the integration measure is Lorentz invariant
Z
d3 p Z d3 p0
=
2ωp 2ωp0

where p and p0 are related by a Lorentz transformation. Choose p0 = Λ−1 p


so we can change the integral from p to p0 in the expression to get the
result.

2. It is zero if x and y are spacelike separated

∆m (x − y) = 0, (x − y).(x − y) < 0

This follows from Lorentz invariance. If events x and y are space-like


separated then there exists a frame of reference in which the two events
are simultaneous. That is there exists a Lorentz transformation Λ such
50 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
that (Λx)0 = (Λy)0 . Thus we only need to prove that ∆m (x − y) = 0 for
x0 = y 0 . But then
Z
d3 p ip.(x−y) Z 3
d p ip.(x−y)
[e − e−ip.(x−y) ] = [e − e−ip.(x−y) ]
2ωp 2ωp
Z 3
d p ip.(x−y)
= [e − eip.(x−y) ]
2ωp
= 0

The function remains zero in the limit if y approaches x along a space-


like curve. Another way to say it is that ∆m (0, x) = 0 for all values of x
including x = 0.

3. The “equal-time” value of the time derivative of the Pauli-Jordan func-


tion is the Dirac delta function

∂ ∂
∆ (x − y) = − ∆ (x − y) = δ(x − y)

0 m 0 m
∂x
x0 =y 0
∂y 0 0
x =y

Exercise 19 Prove the last equation.

§ 44: Fock space

The Fock space for H can be constructed with annihilation and creation opera-
tors corresponding to any of the bases: an orthonormal asis |ni, n = 1, 2, 3, . . .,
or the momentum basis {|pi} or the covariant basis {|xi}. We denote them as

an = a|ni , a†n = a†|ni ,


a(p) = a|pi , a† (p) = a†|pi ,
a(x) = a|xi , a† (x) = a†|xi .

The usual relations are: ( |Φ0 i is the vacuum state)

an |Φ0 i = 0, a(p)|Φ0 i = 0, a(x)|Φ0 i = 0.

[an , am ] = 0, [a(p), a(p0 )] = 0, [a(x), a(y)] = 0,

[a†n , a†m ] = 0, [a† (p), a† (p0 )] = 0, [a† (x), a† (y)] = 0.

[an , a†m ] = δnm , [a(p), a† (p0 )] = 2ωp δ 3 (p − p0 ), [a(x), a† (y)] = hx|yi.


51
All unitary operators U = exp(iG) with self-adjoint generators G can be raised
(“second-quantized”) by first defining

a†m hm|G|nian ,
X
G=
m,n

and then calculating

U = exp(iG).

Then

U af U −1 = aU f , U a†f U −1 = a†U f .
52 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
§ 45: Quantum Field

The relativistic quantum field corresponding to these particles is chosen as

φ(x) = a|xi + a†|xi = a(x) + a† (x).

If {|ni}, n = 1, 2, 3, . . . is an orthonormal basis, we can write

fn (x)∗ |ni,
X X
|xi = |nihn|xi =

Therefore, using the anti-linear and linear nature of a and a† ,

[an fn (x) + a†n fn∗ (x)].


X
φ(x) =
n

Or, for that matter, if we write expansion for |xi in the momentum basis, then

1 Z d3 p 0 0 0 0
φ(x) = 3/2
[a(p)e−ip x +ip·x + a† (p)e+ip x −ip·x ]
(2π) 2ωp

Exercise 20 Show that ∂µ φ transforms as a covariant 4-vector. That is,

U (Λ)∂µ φ(x)U (Λ)−1 = (∂ν φ)(Λx)Λν µ .

§ 46: Properties of the quantum field

φ(x) has been defined so that the following properties are satisfied:

1. It is Hermitian, as is obvious from construction.

2. φ obeys the Klein-Gordon equation

∂2
!
− ∇2 + m2 φ(x) = 0.
∂t2

3. It is relativistically covariant: since

U (a, Λ)a|xi U (a, Λ)−1 = aU (a,Λ)|xi = a|Λx+ai ,

and a similar relation for a†|xi , it follows that

U (a, Λ)φ(x)U (a, Λ)−1 = φ(Λx + a).


53
4. It commutes with itself at space-like distances. The commutator is

[φ(x), φ(y)] = i∆m (x − y),

which is zero as a property of the Pauli-Jordan function.

§ 47: Energy and momentum

The energy and momentum can be calculated using the result of §37. If
{fn }, , n = 1, 2, 3 . . . is an orthonormal basis and an = afn the annihilation
operator etc, the second quantized version of P 0 is

P0 = a†m hfm |P 0 |fn ian


X

m,n

∂a† (x) ∂a(x)


Z " #
= 3
dx + ∇a† (x) · ∇a(x) + m2 a† (x)a(x) .
x0 ∂x0 ∂x0

Here we have just used the fact that


X X
an fn (x) = hx|fn ia|fn i = aP |fn ihfn |xi = a|xi = a(x),
n

and a similar relation for a†n fn∗ .

Similarly the momentum can be written as

∂a(x) ∂a† (x)


Z " #
3 †
P=− d x ∇a (x) + ∇a(x) .
x0 ∂x0 ∂x0

It may seem as if the above expressions depend on the x0 chosen to pick


up the |xi states. But we have seen that any inner product is independent of
the x0 chosen.

§ 48: Normal Ordering

The Fock space observables obtained by ‘second quantization’ are normal or-
dered by definition.

Normal order of any product of annihilation and creation operators in sym-


metric Fock space is defined by bringing all creation operators to the left and
all annihilation operators to the right. The symbol for normal ordering of
operators A1 A2 . . . An is : A1 A2 . . . An :. For example,

: a1 a†2 a3 a4 a†5 := a†2 a†5 a1 a3 a4 .


54 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
In a symmetric Fock space all creation operators commute with each other,
and so do the annihilation operators, it does not matter what order is chosen
within the creation operators , or within the annihilation operators.

Normal ordering for antisymmetric Fock space is defined similarly, except


that since operators anti-commute, there is an overall ± sign in bringing the
initial oder of factors to the final normal order.

Because a|Φ0 i = 0 for every annihilation operator and hΦ0 |a† = 0 for every
creation operator,

the vacuum expectation value of a normal ordered expression is


always zero.

§ 49: Energy and momentum as function of field

Our expression for observables is written in terms of a† (x) and a(x). We try to
express them in terms of φ = a(x)+a† (x). Since P 0 is quadratic in annihilation
creation operators, we try
" #
Z
∂φ(x) ∂φ(x)
A= d3 x + ∇φ(x) · ∇φ(x) + m2 φ(x)2 .
x0 ∂x0 ∂x0

Writing φ(x) = a(x) + a† (x) we get four integrals :

∂a† (x) ∂a(x)


Z " #
A = dx 3
+ ∇a† (x) · ∇a(x) + m2 a† (x)a(x) (1)
x0 ∂x0 ∂x0
∂a(x) ∂a† (x)
Z " #
3 † 2 †
+ dx + ∇a(x) · ∇a (x) + m a(x)a (x) (2)
x0 ∂x0 ∂x0
∂a† (x) ∂a† (x)
Z " #
3 † † 2 † †
+ dx + ∇a (x) · ∇a (x) + m a (x)a (x) (3)
x0 ∂x0 ∂x0
" #
Z
3 ∂a(x) ∂a(x) 2
+ dx + ∇a(x) · ∇a(x) + m a(x)a(x) (4).
x0 ∂x0 ∂x0

The first term is exactly the expression for P 0 . The last integral (4) is zero
because if we write a(x) in momentum basis
∂a(x) ∂
0
= a|xi
∂x ∂x0
∂ 1 Z d3 p −ip0 x0 +p·x
= e a|pi
∂x0 (2π)3/2 2ωp
1 Z d3 p 0 −ip0 x0 +ip·x
= −i (p )e a|pi
(2π)3/2 2ωp
55
Similarly,
1 Z d3 p 0 0
∇a(x) = i 3/2
(p)e−ip x +ip·x a|pi .
(2π) 2ωp
Therefore
Z
∂a(x) ∂a(x)
d3 x
x0 ∂x0 ∂x0
Z
3 1 Z d3 p Z d3 k 0 0 −ip0 x0 +ip·x −ik0 x0 +ik·x
= − dx (p k )e e a|pi a|ki
x0 (2π)3 2ωp 2ωk
The integration over x gives the delta function δ(p + k), which allows k inte-
gration to be performed and k can be replaced by −p giving
Z 3
Z
∂a(x) ∂a(x) dp 1 0 0
d3 x 0 0
= − (p0 )2 e−2ip x a|pi a|ki |k=−p .
x0 ∂x ∂x 2ωp 2ωp
Similarly,
Z Z
d3 p 1 0 0
3
d x ∇a(x) · ∇a(x) = (p2 )e−2ip x a|pi a|ki |k=−p
x0 2ωp 2ωp
where the plus sign occurs because p · k becomes −p2 when k = −p. The
integral becomes zero due to the mass shell condition −(p0 )2 + p2 + m2 = 0.

The last-but-one integral (3) (with both a† in each term) is also zero for
the same reason.

This leaves the second integral (2) in which a and a† occur in opposite
order. Using a similar expansion in momentum basis like the first term we
get a similar expression except for the order of the annihilation and creation
operators:
∂a(x) ∂a† (x)
Z " #
dx 3
0 0
+ ∇a(x) · ∇a† (x) + m2 a(x)a† (x)
x 0 ∂x ∂x
Z 3 Z 3
Z
1 dp dk 0 0
= d3 x 3
(p k + p · k + m2 )
x0 (2π) 2ωp 2ωk
0 x0 +ip·x 0 x0 −ik·x
×e−ip eik a|pi a†|ki
Z
1 Z d3 p Z d3 k 0 0
= d3 x (p k + p · k + m2 )
x0 (2π)3 2ωp 2ωk
0 x0 +ip·x 0 x0 −ik·x
h i
×e−ip eik a†|ki a|pi + 2p0 δ(p − k)
∂a† (x) ∂a(x)
Z " #
= 3
dx + ∇a† (x) · ∇a(x) + m2 a† (x)a(x)
x0 ∂x0 ∂x0
1 Z 3 Z 3
+ d x d p ωp .
(2π)3 x0
56 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
Therefore, finally
" #
Z
∂φ(x) ∂φ(x)
3
dx + ∇φ(x) · ∇φ(x) + m2 φ(x)2
x0 ∂x0 ∂x0
0 1 Z 3 Z 3
= 2P + d x d p ωp .
(2π)3 x0

§ 50: Zero Point energy

The expression of energy calculated in the last section can be written as,
introducing the h̄ where it is hidden:
" #
1Z 3 ∂φ(x) ∂φ(x)
dx + ∇φ(x) · ∇φ(x) + m2 φ(x)2
2 x0 ∂x0 ∂x0
0 1 Z 3 Z 3 1
= P + d x d p h̄ωp .
(2πh̄)3 x0 2

Remember that our P 0 by definition is normal ordered, therefore its expecta-


tion value is zero. The vacuum expectation value for this expression of energy
in terms of fields is therefore
* Z " # +
1 ∂φ(x) ∂φ(x)
3 2 2
Φ0 dx + ∇φ(x) · ∇φ(x) + m φ(x) Φ0

2 x 0 ∂x0 ∂x0
1 Z Z
1
= d3 x d3 p h̄ωp .
(2πh̄)3 x0 2

This is the zero point energy equal to (1/2)h̄ωp for each oscillator labelled by
p. The number of modes is calculated by the formula
Z
d3 xd3 p
modes in phase space voiume = .
h3

The zero point energy is actually infinite, but it is just an infinite constant
added to the expression for energy written in terms of fields. We started from
particles and obtained a normal ordered expression in the Fock space. If we
start from a classical field and then ”quantize” by some set of rules, we get
this expression with the zero-point energy.

To avoid it, expressions for physical quantities are re-defined by normal


ordering.
" #
0 1Z 3 ∂φ(x) ∂φ(x)
P = dx : + ∇φ(x) · ∇φ(x) + m2 φ(x)2 :
2 x 0 ∂x0 ∂x0
57
Exercise 21 Show that we can write
1 ∂φ(x) ∂φ(x)
Z  
3
P= d x : 0
∇φ(x) + ∇φ(x) :
2 x0 ∂x ∂x0

Exercise 22 Is there a “zero-point momentum” ? In other words, is


∂φ(x) ∂φ(x) ∂φ(x) ∂φ(x)
Z   Z  
3 3
d x 0
∇φ(x) + ∇φ(x) − d x : ∇φ(x) + ∇φ(x) :
x0 ∂x ∂x0 x 0 ∂x0 ∂x0
equal to zero or not?

§ 51: Canonical equal time commutation relation

The time derivative of the field π(x) = ∂φ/∂x0 is called the momentum canon-
ically conjugate to φ(x).

Exercise 23 Show that


[φ(x), π(y)]|x0 =y0 = iδ(x − y).

Hint: use the last property of the Pauli-Jordan function.

§ 52: The Lagrangian density

Define
1h i
L = π(x)∂0 φ(x) − π(x)2 + ∇φ(x) · ∇φ(x) + m2 φ(x)2
" 2 #
1 ∂φ(x) ∂φ(x) 2 2
= − ∇φ(x) · ∇φ(x) − m φ(x)
2 ∂x0 ∂x0
1h i
= − η µν ∂µ φ∂ν φ + m2 φ2
2
as the Lagrangian density. The analogy is from L = pq̇ − H(q, p).

§ 53: Parity and Time reversal

The Lorentz transformation corresponding to parity (space-inversion) is the


4 × 4 matrix
1
 
 −1 
Is =  ,
 
 −1 
−1
58 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
and similarly for the time reversal (time-inversion)

−1
 
 1 
It = −Is =  .
 
 1 
1

Note that

Is2 = 1, Is−1 = Is

and similarly for the time reversal matrix.

To represent parity we must use the group property.

Is (a, Λ)Is = (Is a, Is ΛIs ).

Here, we can check that


!
Λ0 0 −Λ0 i
Is ΛIs = It ΛIt = , i, j = 1, 2, 3.
−Λi 0 Λi j

For the special case of pure translations it becomes

Is (a, 1)Is = (Is a, 1).

Remark 4 Wigner’s theorem tells us that each symmetry is represented by (1) a


unitary or anti-unitary operator, and (2) it is unknown upto a phase factor. Opera-
tors U (a) and U (Λ) have to be unitary and the phases can all be chosen to be equal
to ±1 by an argument of continuity. The phase ±1 can be converted to +1 for all
these operators if we choose the group SL(2, C) as the fundamental symmetry group
and not the Lorentz group. The Lorentz group was discovered by classical physics,
but the existence of spin half particles proves that the real symmetry group is the
group SL(2, C) of 2 × 2 complex matrices of determinant 1. This group has twice
as many elements: the SL(2, C) matrices A and −A both correspond to the some
Lorentz transformation which is connected to the identity in a continuous manner.
This is enough to remove the ±1 ambiguity of the phase factors. However that leaves
parity and time reversal are not connected to the identity continuously. The unitary
or anti-unitary nature of these symmetries is fixed by physical requirements like
positivity of energy, and their phases are fixed by convention, as well as consistency
of choice when interaction with other particles is involved.

For parity Is2 = 1 therefore the operator P = U (Is ) guaranteed by the


Wigner theorem and unknown upto a phase factor will satisfy

P 2 = PP = ωP 1, |ωP | = 1.
59
Moreover,

Is (a, 1)Is−1 = (Is a, 1)

which for our convenience we write as (a → Is a)

Is (Is a, 1)Is−1 = (a, 1)

therefore we expect the operator to satisfy

PU (Is a)P −1 = U (a)

or

PU (Is a) = U (a)P.

Now,

U (a) = exp(iP 0 a0 − iP · a),

where P 0 and P are energy and momentum operators. Acting on a state |pi,
we find

PU (Is a)|pi = U (a)P|pi

or
h i
P exp(iωp a0 + ip · a)P|pi = exp(iP 0 a0 − iP · a) [P|pi] .

If P is unitary, then the factor exp(iωp a0 + ip · a) come out to the left of P,


proving that P|pi is an eigenstate of momentum −p. If anti-unitary, then, the
factor exp(iωp a0 + iP · a) will become complex conjugate
√ 2exp(−iω
0
p a − ip · a)
and that would imply that the energy of the state is − p + m2 , which we do
not want.

Therefore, the parity operator P should be chosen to be unitary.

The action of P on states |pi is such that P|pi is proportional to |Is pi. The
proportionality factor can only be a phase ηP . So, finally

P|pi = ηP |Is pi, Is p = (ωp , −p).

By convention, η is defined to be real, then it can be a number 1 or −1


characteristic of the particle, called intrinsic parity.

For time reversal, a similar exercise can be done. The time reversal operator
T = U (It ) has to be chosen to be anti-unitary because

T U (It a)|pi = U (a)T |pi


60 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
or
h i
T exp(−iωp a0 − ip · a)|pi = exp(iP 0 a0 − iP · a) [T |pi] .

implies that a unitary choice will lead to negative energy eigenstate T |pi.

The time reversal operator T should be chosen to be anti-unitary.

For anti-unitary T ,
exp(+iωp a0 + ip · a) [T |pi] = exp(iP 0 a0 − iP · a) [T |pi] .
This shows that T |pi is an eigenstate with positive energy and momentum
−p. Similarly to parity, we can write
T |pi = ηT |Is pi.
(Note that it is Is p with momentum −p, and not It p on the right hand side.

It is interesting that the phase factor ηT can also be chosen to be real.


These phases do not seem to have any physical consequences.

Exercise 24 Show that


P|xi = ηP |Is xi,

and

T |xi = ηT |It xi.

(Hint: use the anti-linear property of T . )

We know that a symmetry transformation S acting on one particle states


when generalized to multi-particle states acts as
S(f1 ⊗ · · · ⊗ fn ) = (Sf1 ) ⊗ · · · ⊗ (Sfn ).
This means that for Af defined by

Af (f1 ⊗ · · · ⊗ fn ) = n(f, f1 )(f2 ⊗ · · · ⊗ fn )
application of S will require

SAf S −1 (Sf1 ) ⊗ · · · ⊗ (Sfn ) = nS(f, f1 )(f2 ⊗ · · · ⊗ fn )

= nS(f, f1 )(f2 ⊗ · · · ⊗ fn )

= n(f, f1 )(Sf2 ⊗ · · · ⊗ Sfn ), if S unitary
√ ∗
= n(f, f1 ) (Sf2 ⊗ · · · ⊗ Sfn ), if S anti-unitary

= n(Sf, Sf1 )(Sf2 ⊗ · · · ⊗ Sfn ), in both cases.
61
Therefore,
SAf S −1 = ASf ,
It is just one step to conclude that our creation and annihilation operators
satisfy for a symmetry transformation
Saf S −1 = aSf
Sa†f S −1 = a†Sf .

Therefore
Pa|xi P −1 = ηP∗ a|Is xi
Pa†|xi P −1 = ηP a†|Is xi .
If ηP is real then
Pφ(x)P −1 = ηP φ(Is x).

§ 54: Charge and U (1) internal symmetry

Let there be two spin-less particles with identical mass m, but with a charge
q and −q respectively. They have their respective Hilbert spaces Hq and H−q ,
with basis states |p, qi and |p, −qi as before. When we consider both the
particles together, then we are using a bigger space Hq ⊕ H−q where every
vector of the first space is orthogonal to every vector of the other.

The states |x, qi and |x, −qi can also be constructed accordingly.

The charge operator Q acts as


Q|x, qi = q|x, qi, Q|x, −qi = −q|x, qi.
The symmetry transformation generated by Q is an internal symmetry given
by
U (α) = exp(iαQ).
It commutes with the space-time symmetries U (Λ) and U (a) as is obvious from
the definition. The group of this symmetry is U (1), the unitary group of one
dimensional unitary matrices exp(iα.

The Fock space will be built out of annihilation and creation operators
corresponding to the vectors of the joint Hilbert space. We denote by af , a†f
the operators of f ∈ Hq , and by bg , b†g the operators of g ∈ H−q . Their
commutation relations are
[af1 , a†f2 ] = (f1 , f2 ), [bg1 , b†g2 ] = (g1 , g2 )
with the remaining all commutators zero.
62 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
§ 55: The “complex” scalar field

When we try to construct fields in a manner similar to for neutral particle we


realise that although a(x) and a† (x) transform in the same way under space-
time symmetries, under U (1) generated by charge they go as
U (α)a|x,qi U (α)−1 = exp(−iαq)a|x,qi
whereas
U (α)a†|x,qi U (α)−1 = exp(iαq)a†|x,qi .
Therefore we are forced to combine the annihilation operator of a particle with
the charge q with the creation operator of the particle with the charge −q ,
which are of the same mass and spin. Such a pair of particles is called a
particle-anti-particle pair.

Define,
φ = a(x) + b† (x)
so that
φ† = b(x) + a† (x).
This non-hermitian (or “complex”) scalar field has the right transformation
properties:
U (a, Λ)φ(x)U (a, Λ)−1 = φ(Λx + a)
U (α)φ(x)U (α)−1 = exp(−iαq)φ(x)
Under parity, if the intrinsic parity of particle with charge q is ηPq and of
that with charge −q is ηP−q , then the causal field has the right transformation
property only if
(ηPq )∗ = ηP−q .
This implies that ηPq ηP−q = 1.

For a spin less particle the intrinsic parities of the particle and the
antiparticle are the same.

We must also check the “causal nature” (or the micro-causality condition):
the fields must commute or anti-commute at space-like separation.
[φ(x), φ(y)] = 0,
h i
φ† (x), φ† (y) = 0,
h i
φ(x), φ† (y) = hx, q|y, qi − hy, −q|x, −qi = i∆m (x − y).
63
Exercise 25 Let {fn } be an orthonormal basis in Hq and {gn } be an orthonor-
mal basis in H−q constructed in the same way, that is with same momentum wave
functions as for fn . The only difference is that in place |p, qi, sates |p, −qi are used.
Call afn by an and similarly for bn = bgn . Show that the charge operator

(a†n an − b†n bn )
X
Q=q
n

can be written as
!
∂φ ∂φ†
Z
3 †
Q = iq : d x φ − φ :
∂x0 ∂x0

§ 56: Charge Conjugation

Any one-to-one unitary mapping between Hq and H−q can be called a


charge conjugation mapping. Usually however it is defined as

C|p, qi = ηC |p, −qi, C|p, −qi = ηC |p, qi, C 2 = 1, ηC = ±1.

In the Fock space

Can C −1 = ηC bn , Cbn C −1 = ηC an
Ca†n C −1 = ηC b†n , Cb†n C −1 = ηC a†n

For the causal field,

Cφ(x)C −1 = ηC φ† (x).

Exercise 26 Show that if ηC = 1 then we can write


πX †
 
C = exp i (an − b†n )(an − bn ) .
2
64 CHAPTER 4. RELATIVISTIC SPIN ZERO PARTICLES AND FIELDS
Chapter 5

Interacting boson fields

§ 57: Feynman Propagator

When discussing interactions we will often require vacuum expectation value


(VEV) of time-ordered product of two free fields. The simplest case is that of
a scalar field.

Let φ(x) be a neutral free scalar field discussed before. The time ordered
VEV is defined as

hΦ0 |T (φ(x)φ(y))|Φ0 i = hΦ0 |φ(x)φ(y)|Φ0 i = hx|yi, if x0 > y 0


= hΦ0 |φ(y)φ(x)|Φ0 i = hy|xi, if x0 < y 0

We calculate therefore,

hΦ0 |T (φ(x)φ(y))|Φ0 i = θ(x0 − y 0 )hx|yi + θ(y 0 − x0 )hy|xi.

Here θ is the step function, or Heaviside function equal to 1 for positive argu-
ment, and zero for negative. The following representations for the step function
will be used,
0 0
0 0 1 Z∞ eiλ(x −y )
θ(x − y ) = dλ
2πi −∞ λ − i
−iλ(x0 −y 0 )
−1 Z ∞
e
= dλ
2πi −∞ λ + i
The proof of these formulas is simple. For the first case we consider λ as a
complex variable. For x0 > y 0 we take a D-shaped very large counter clockwise
contour closing the real line with a semicircle in the upper half plane. Since
the imaginary part of λ is positive and large on the semi-circular loop there
is a factor exp(−(x0 − y 0 )=λ) which goes to zero when the contour is taken

65
66 CHAPTER 5. INTERACTING BOSON FIELDS
to infinity. The integral contributes the residue of the pole at λ = i, and
evaluates to 1. When x0 < y 0 the contour has to be taken in the lower half
plane where there is no pole.So the integral is zero. Similar arguments apply
to the next representation.

Let us now evaluate


0 0
0 0 −1 Z ∞ e−iλ(x −y ) 1 Z d3 p −iωp (x0 −y0 )+ip·x−y
θ(x − y )hx|yi = dλ e
2πi −∞ λ + i (2π)3 2ωp
0 0
i Z d3 p Z ∞ e−i(λ+ωp )(x −y )+ip·x−y
= dλ
(2π)4 2ωp −∞ λ + i

Change the λ integration variable to p0 = λ + ωp , (p0 has no connection to


ωp ), so that we can write
0 0 0
0 0 i Z d4 p e−ip (x −y )+ip·x−y
θ(x − y )hx|yi =
(2π)4 2ωp p0 − ωp + i
Similarly, using the other representation of the theta function,
0 0 0
0 0 −i Z d4 p e−ip (x −y )+ip·x−y
θ(y − x )hy|xi = .
(2π)4 2ωp p0 − ωp − i

Finally,

hΦ0 |T (φ(x)φ(y))|Φ0 i = θ(x0 − y 0 )hx|yi + θ(y 0 − x0 )hy|xi

0 0 0
−i Z 4 e−ip (x −y )+ip·x−y
= d p
(2π)4 −(p0 )2 + p2 + m2 − i
≡ −i∆F (x − y)

The function
1 Z 4 eip·(x−y)
∆F (x − y) = dp
(2π)4 p · p + m2 − i
is called the Feynman propagator.

§ 58: Interaction picture

A scalar like φ(x) is in the Heisenberg picture by construction. Its time de-
pendence is given by

exp(iP 0 x0 )φ(0, x) exp(−iP 0 x0 ) = φ(x0 , x).


67
where P 0 is the energy, or Hamiltonian
" #
0 1Z 3 ∂φ(x) ∂φ(x)
P = dx : + ∇φ(x) · ∇φ(x) + m2 φ(x)2 :
2 x 0 ∂x0 ∂x0

When we introduce interaction, adding a term Hint to P 0 as a function of


this (or other fields), then the picture becomes interaction picture, because
the fields continue to develop according to P 0 which now is the “free” or
unperturbed Hamiltonian.

The quantity of interest in the interaction picture is the scattering matrix,


given by
  Z ∞ 
S = T exp −i dt Hint (t)
−∞

where T denotes the time-ordering and the t = x0 dependence of Hint is coming


from the fields φ(x).

§ 59: An illustrative model

Take two neutral scalar fields φ and Φ with masses m and M respectively.
They interact locally, that is, the interaction terms containing the product of
fields refer to the same space-time point. We will write

φ(x) = a(x) + a† (x), Φ(x) = A(x) + A† (x).

The only non-vanishing commutators are [a, a† ] and [A, A† ]. In particular a, a†


both commute with both of A, A† .

The free Hamiltonian is


" #
1Z 3 ∂φ(x) ∂φ(x)
H0 = dx : + ∇φ(x) · ∇φ(x) + m2 φ(x)2 :
2 x0 ∂x0 ∂x0
" #
1Z 3 ∂Φ(x) ∂Φ(x) 2 2
+ dx : + ∇Φ(x) · ∇Φ(x) + M Φ(x) :
2 x0 ∂x0 ∂x0

Despite the fact that it is calculated at some fixed x0 = t, H0 is independent


of time because both φ and Φ satisfy the respective Klein-Gordon equations.

Let us choose an interaction Hamiltonian


Z
Hint (t) = κ d3 x : φ2 (x)Φ(x) :
x0 =t

Notice the normal ordering, which is done to avoid zero-point energy like in-
finities, coming from the commutators [a(x), a† (x)]. κ is a ‘coupling constant’
68 CHAPTER 5. INTERACTING BOSON FIELDS
assumed small so that a perturbation theory makes sense. For this case the
S-matrix is
  Z ∞ Z 
3 2
S = T exp −iκ dt d x : φ (x)Φ(x) :
−∞ x0 =t
  Z 
= T exp −iκ d4 x : φ2 (x)Φ(x) : .

We can now calculate the S-matrix amplitudes for any process, using the per-
turbation formula.

§ 60: φ-φ scattering

Let initial state be a two φ particle state with incoming states |f i and |gi
(having momenta sharply peaked around k1 and k2 respectively). And the
same two particle going out with states |f 0 i, |g 0 i having momenta p1 and p2
respectively. Such states can be written (calling |0i in place of |Φ0 i) as

|ini = a†f a†g |0i, |outi = a†f 0 a†g0 |0i.

These states are by definition symmetric for the two identical particles, and
are of unit norm if each of the vectors f, g, f 0 , g 0 are unit vectors.

The S-matrix element is simply


  Z 
4 2
hin| T exp −iκ d x : φ (x)Φ(x) : |outi.

§ 61: Zeroth order

The zeroth-order term in the expansion is

hin|outi = h0|ag0 af 0 a†f a†g |0i = (f 0 , f )(g 0 , g) + (g 0 , f )(f 0 , g).

We assume the final state to be different from the initial so that the zeroth-
order term of the S-matrix is zero. This is always the case, the final state
momenta are in a different direction than the incoming momenta. Usually
the initial momenta are in the opposite direction in the incoming beams, and
scatter off in a different direction.
69
§ 62: First order

The first order term looks like


Z
hin|S (1) |outi = h0|ag0 af 0 [−iκ d4 x : φ2 (x)Φ(x) :]a†f a†g |0i.

The first order term for φ-φ scattering is also zero because, (omitting (x) to
simplify writing) in

: φ2 (x)Φ(x) : = : (a + a† )(a + a† )(A + A† ) :


= A† [a† a† + aa + 2a† a] + [a† a† + aa + 2a† a]A

A in the second term can go past a†f a†g and act directly on |0i and give zero.
Similarly, A† in the first term can go to left of ag0 af 0 and give 0 = h0|A† .

§ 63: Second order

The second order term looks like


(−iκ)2 Z 4
hin|S (2)
|outi = d x1 d4 x2 h0|ag0 af 0 [ T : φ2 (x1 )Φ(x1 ) :: φ2 (x2 )Φ(x2 ) :]a†f a†g |0i.
2!
Let us write a1 for a(x1 ) etc. for convenience. The integrand then contains
 
T h0|ag0 af 0 A†1 [a†1 a†1 + a1 a1 + 2a†1 a1 ] + [a†1 a†1 + a1 a1 + 2a†1 a1 ]A1
 
× A†2 [a†2 a†2 + a2 a2 + 2a†2 a2 ] + [a†2 a†2 + a2 a2 + 2a†2 a2 ]A2 a†f a†g |0i

To manage time ordering we take the two cases separately.

case 1: x01 > x02


The first thing to notice is that we can drop the first term in the first factor
in parentheses (which has A† on the left), because this creation operator can
get past ag0 af 0 and act on h0| and give zero. Similarly, the second term in the
second factor in parentheses which has A2 going past a†f a†g and give zero acting
on |0i. Thus we have a total of 3 × 3 = 9 terms to deal with in

h0|ag0 af 0 [a†1 a†1 + a1 a1 + 2a†1 a1 ] A1 A†2 [a†2 a†2 + a2 a2 + 2a†2 a2 ] a†f a†g |0i.

The A1 A†2 in the middle is effectively a complex number because

A1 A†2 = [A1 , A†2 ] + A†2 A1 = hx1 |x2 iM + A†2 A1


70 CHAPTER 5. INTERACTING BOSON FIELDS
and the second term will give zero because A1 can go and act on |0i. So,
  
hx1 |x2 iM h0|ag0 af 0 a†1 a†1 + a1 a1 + 2a†1 a1 a†2 a†2 + a2 a2 + 2a†2 a2 a†f a†g |0i.

Again, since the matrix element is between the vacuum states, the number of
creation and annihilation operators should be the same in each term, otherwise
it will give zero. There are already two annihilation and two creation operators
due to initial and final particles. Therefore , (apart from the factor hx1 |x2 iM
outside) of the nine terms only the following three will survive:

h0|ag0 af 0 (a†1 a†1 a2 a2 ) a†f a†g |0i (1)


+h0|ag0 af 0 (a1 a1 a†2 a†2 ) a†f a†g |0i (2)
+4h0|ag0 af 0 (a†1 a1 a†2 a2 ) a†f a†g |0i (3).

Term (1):
we can take a2 a2 to the right past the creation operators of the incoming
particles:

a2 a2 a†f a†g |0i = a2 (hx2 |f i + a†f a2 )a†g |0i


= 2hx2 |f ihx2 |gi|0i
= 2f (x2 )g(x2 )|0i.

Similarly, we can simplify

h0|ag0 af 0 a†1 a†1 = 2h0| f 0∗ (x1 )g 0∗ (x1 )

Combining the two parts we calculate the first term:


∗ ∗
(1) = 4f 0 (x1 )g 0 (x1 )f (x2 )g(x2 )

Term (2):
Here, as

a1 a1 a†2 a†2 = 2hx1 |x2 i2 + hx1 |x2 ia†2 a1 + a†2 a†2 a1 a1

we get three terms,

2hx1 |x2 i2 h0|ag0 af 0 a†f a†g |0i (2a)


+hx1 |x2 ih0|ag0 af 0 a†2 a1 a†f a†g |0i (2b)
h0|ag0 af 0 a†2 a†2 a1 a1 a†f a†g |0i. (2c)

Easy calculation shows

(2a) = 2hx1 |x2 i2 [(f 0 , g)(g 0 , f ) + (f 0 , f )(g 0 , g)]


= 0, because initial and final states are assumed orthogonal
71
For the same reason,
∗ ∗
   
(2b) = f 0 (x2 )hg 0 | + g 0 (x2 )hf 0 | |gif (x1 ) + |f ig(x1 )
∗ ∗
= f 0 (x2 )f (x1 )(g 0 , g) + f 0 (x2 )g(x1 )(g 0 , f )
∗ ∗
+g 0 (x2 )f (x1 )(f 0 , g) + g 0 (x2 )g(x1 )(f 0 , f )
= 0.

The term (2c) is of the same form as term (1) with two creation operators
to the left of two annihilation operators sandwiched between the operators
of incoming and outgoing particles, with the role of x1 and x2 interchanged.
Therefore,
∗ ∗
(2c) = 4f 0 (x2 )g 0 (x2 )f (x1 )g(x1 )

Term (3):
The third term is
∗ ∗ ∗ ∗
(3) = f 0 (x1 )g 0 (x2 )f (x1 )g(x2 ) + f 0 (x1 )g 0 (x2 )f (x2 )g(x1 )
∗ ∗ ∗ ∗
+f 0 (x2 )g 0 (x1 )f (x1 )g(x2 ) + f 0 (x2 )g 0 (x1 )f (x2 )g(x1 ).

Therefore for x01 > x02 the total amplitude is,


(−iκ)2 Z 4 h
Amp(x01 > x02 ) = d x1 d4 x2 θ(x01 − x02 )hx1 |x2 iM
2!
0∗ ∗
4f (x1 )g 0 (x1 )f (x2 )g(x2 )
∗ ∗
+ 4f 0 (x2 )g 0 (x2 )f (x1 )g(x1 )
∗ ∗ ∗ ∗

+ 4 f 0 (x1 )g 0 (x2 )f (x1 )g(x2 ) + f 0 (x1 )g 0 (x2 )f (x1 )g(x1 )
∗ ∗ ∗ ∗
i
+f 0 (x2 )g 0 (x1 )f (x1 )g(x2 + f 0 (x2 )g 0 (x1 )f (x2 )g(x1 )

where we introduce the step function to remind that this expression is valid
for x01 > x02 .

case 1: x02 > x01


The calculation goes on exactly like the case x01 > x02 , with the interchange of
x1 and x2 . Thus,
(−iκ)2 Z 4 h
Amp(x01 < x02 ) = d x1 d4 x2 θ(x02 − x01 )hx2 |x1 iM
2!
0∗ ∗
4f (x2 )g 0 (x2 )f (x1 )g(x1 )
∗ ∗
+ 4f 0 (x1 )g 0 (x1 )f (x2 )g(x2 )
∗ ∗ ∗ ∗

+ 4 f 0 (x2 )g 0 (x1 )f (x2 )g(x1 ) + f 0 (x2 )g 0 (x1 )f (x1 )g(x2 )
∗ ∗ ∗ ∗
i
+f 0 (x1 )g 0 (x2 )f (x2 )g(x1 ) + f 0 (x1 )g 0 (x2 )f (x1 )g(x2 )
72 CHAPTER 5. INTERACTING BOSON FIELDS
The two amplitudes can be combined together. The first term in the big
square bracket or Amp(x01 > x02 ) will combine with the second of the Amp(x01 <
x02 ) so that there is the Feynman propagator for the particle of mass M
θ(x01 − x02 )hx1 |x2 iM + θ(x02 − x01 )hx2 |x1 iM = −i∆M
F (x1 − x2 )

as a common factor. Similarly, the second of the first amplitude combines with
the first of the second. The third term in both is the same any way. The result
is
(−iκ)2 Z 4 h
Amp = d x1 d4 x2 (−i)∆M F (x1 − x2 )
2!
∗ ∗
4f 0 (x2 )g 0 (x2 )f (x1 )g(x1 )
∗ ∗
+ 4f 0 (x1 )g 0 (x1 )f (x2 )g(x2 )
∗ ∗ ∗ ∗

+ 4 f 0 (x2 )g 0 (x1 )f (x2 )g(x1 ) + f 0 (x2 )g 0 (x1 )f (x1 )g(x2 )
∗ ∗ ∗ ∗
i
+f 0 (x1 )g 0 (x2 )f (x2 )g(x1 ) + f 0 (x1 )g 0 (x2 )f (x1 )g(x2 )
recall that the Feynman propagator is symmetric in its argument. This means
that interchanging x1 and x2 produces no effect. Therefore we can further
shorten the above expression as
(−iκ)2 Z 4
Amp = 8 d x1 d4 x2 (−i)∆M F (x1 − x2 ) ×
2!
∗ ∗
h
f 0 (x2 )g 0 (x2 )f (x1 )g(x1 )
∗ ∗ ∗ ∗
 i
+ f 0 (x2 )g 0 (x1 )f (x2 )g(x1 ) + f 0 (x2 )g 0 (x1 )f (x1 )g(x2 )
≡ Amp1 + (Amp2a + Amp2b)

§ 64: Feynman Graphs

A Feynman graph is simply a concise way of representing a normal


product.
S. S. Schweber,
An Introduction to Relativistic Quantum Field Theory, 1961, p.435

In the amplitude calculated in the previous section for two φ particles


coming in with states f and g and scattering off with states f 0 and g 0 the first
term Amp1, (apart for the common factors and integration) can be written
more explicitly as
∗ ∗
A: for x01 < x02 : f 0 (x2 )g 0 (x2 )hx2 |x1 iM f (x1 )g(x1 )
∗ ∗
B: for x01 > x02 : f 0 (x2 )g 0 (x2 )hx1 |x2 iM f (x1 )g(x1 ).
This has a simple graphical interpretation. Make a space-time diagram of two
events x1 and x2 . The first process, A, corresponds to
73
1. The two particles come from the past with wave-functions f, g,

2. get annihilated at x1 , with the creation of the Φ particle at the same


point,

3. the propagation amplitude hx2 |x1 iM of the Φ particle from x1 to x2 ,

4. annihilation of Φ particle at x2 , with simultaneous creation of a pair of


φ particles with wave functions f 0 , g 0 .

A 
A 
f’A g’
A 
A 
A 
6 A 
Aux2
hx2 |x1 iM 



x1 u  Process A
x01 < x02
A
 A
Time  A
 A
 A
f Ag
 A
 A

The other process is similar, with the role of x1 and x2 changed. The only
remarkable thing is that in the second case the propagator from x1 to x2 runs
backwards in time. This is typical to relativity, and if x1 and x2 are space-like
separated, then the order of events does not matter.
74 CHAPTER 5. INTERACTING BOSON FIELDS

f’ g’
A 
A 
6 A 
ux1
A 
A  hx1 |x2 iM  A
A  
   A
A  
u
 A
x2 A
  Process B  A
x01 > x02  A
Time 
A
A
 A
f g

The Feynman propagator takes care of the time orderings. There is no


need to write two diagrams. A single one like the one below is sufficient.
A 
A 
f’ A  g’
A 
A 
x1 Au

∆M
F (x1 − x2 )

x2 u
A
 A
 A
 A
 A
f Ag

Graph corresponding to Amp1

Similarly, the diagrams for the terms inside parentheses can be drawn as
the following two Feynman graphs:
75
 
f’ g’ g’ f’
 
A  A 
A  A 
sx1 s x
A A   1
A 
  A A
  A
A  A 
s  ∆M (x − x ) A s  ∆M (x − x ) A
A A
x2A

x2 A 
 F 1 2  F 1 2
A A
 A  A
 A  A
f  g f  g
 
 
Graph corresponding to Amp2a Graph corresponding to Amp2b

The integrations over d4 x1 and d4 x2 tell us that amplitudes for all these
processes have to summed up. Basically, that is Feynman’s method.

§ 65: Momentum space

Let us choose for initial and final states the sharp momentum states:
eik1 ·x
f (x) = hx|f i = hx|k1 i =
(2π)3/2
eik2 ·x
g(x) = hx|gi = hx|k2 i =
(2π)3/2
e−ip1 ·x
f 0 (x)∗ = hf 0 |xi = hp1 |xi =
(2π)3/2
0 ∗ 0 e−ip2 ·x
g (x) = hg |xi = hp2 |xi =
(2π)3/2
Then, recalling
1 Z 4 eip·(x−y)
∆M
F (x − y) = d p
(2π)4 p2 + M 2 − i
where p2 ≡ cdotp, we get
(−iκ)2 (−8i) 1 Z 4 4
Z
4 eip·(x1 −x2 )
Amp1 = d x 1 d x 2 d p
2! (2π)6 (2π)4 p2 + M 2 − i

×ei(k1 +k2 )·x1 e−i(p1 +p2 )·x2

The 4-dimensional integrations over x1 and x2 take away (2π)8 and produce
δ 4 (p + k1 + k2 ) and δ 4 (p + p1 + p2 ) respectively, one of which can be integrated
76 CHAPTER 5. INTERACTING BOSON FIELDS
with d4 p to produce the total energy-momentum conservation delta function
δ 4 (p1 + p2 − k1 − k2 ). Thus,

4κ2 1
Amp1 = i(2π)4 δ 4 (p1 + p2 − k1 − k2 ) .
(2π) (k1 + k2 ) + M 2 − i
6 2

It is written in this form so that we can read off the T-matrix, defined as

S = 1 − i(2π)4 δ 4 (Pf − Pi )Tf i

directly from the expression.

Similarly,

4κ2 1
Amp2a = i(2π)4 δ 4 (p1 + p2 − k1 − k2 )
(2π) (p1 − k1 ) + M 2 − i
6 2

4κ2 1
Amp2b = i(2π)4 δ 4 (p1 + p2 − k1 − k2 ) .
(2π) (p1 − k2 ) + M 2 − i
6 2

§ 66: Pairings

We have encountered how a product of several annihilation-creation operators


sandwiched between between the vacuum state, can be evaluated by bringing
annihilation operators to the right (or creation operators to the left) so that
they can act on the vacuum and give zero. The non-zero result corresponds
to the various inner products which result when an annihilation operator goes
past a creation operator.

Wick’s theorem gives us a recipe to do precisely that in a systematic man-


ner. The following discussion is for the boson fields. The changes required for
fermion fields will be discussed when we introduce those fields.

Let A1 A2 be the product of two operators which are either annihilation


or creation operators, and let : A1 A2 : be their normal ordered product. The
difference between the ordinary product and the normal ordered product can
only be a complex number, c :

A1 A2 =: A1 A2 : + c.

Take vacuum expectation value of both sides, and use the fact that the VEV
of a normal ordered quantity is always zero, we find that

c = h0|A1 A2 |0i.
77
There is a convenient notation for this quantity, called pairing:

h0|A1 A2 |0i ≡ A1 A2

It is important to realise that in a pairing the order of factors matters.

There can be multiple pairings in a product like A1 A2 A3 A4 A5 A6 . For


example,

A1 A2 A3 A4 A5 A6 = h0|A1 A3 |0ih0|A2 A5 |0iA4 A6

A normal product with pairings is defined in a similar fashion. For the


above example,

: A1 A2 A3 A4 A5 A6 : = h0|A1 A3 |0ih0|A2 A5 |0i : A4 A6 :

§ 67: Wick’s theorem for ordinary products

We can now state the Wick theorem for ordinary products.

Let A1 , . . . , An be n creation or annihilation operators. Then their product


can be expressed in the normal ordered form as

A1 . . . An = : A1 . . . An : (zero pairings)
P
+ : A1 A2 A3 . . . Ak . . . An : (all 1-pairings)
P
+ : A1 A2 A3 . . . Ak . . . An : (all 2-pairings)
+···

If the number of factors n is odd, the last terms will be single operators.
If it is even, (n = 2m), the last terms are just numbers with all possible
m-pairings. The number of pairings is given below:
n(n − 1)
1-pairings =
2!
78 CHAPTER 5. INTERACTING BOSON FIELDS
1 n(n − 1) (n − 2)(n − 3)
2-pairings =
2! 2! 2!
1 n(n − 1) (n − 2)(n − 3) (n − 4)(n − 5)
3-pairings =
3! 2! 2! 2!

···

This can be seen as follows. For example, the one pairings are n C2 = n(n−1)/2!
in number as we can choose two of the factors in so many ways. For two
pairings, we choose one pairings first out of n factors, and then choose further
1-pairings out of the remaining (n − 2) factors. This will be n C2 × n−2 C2
ways. But there is extra counting because the pairing chosen in the second
stage and the pairing chosen in the first stage could have been chosen in the
reverse order. In other words, permuting the pairings with each other does not
give new pairings. Thus we should divide by a factor of 2!. Similarly the three
pairings are n C2 × n−2 C2 × n−4 C2 divided by 3!.

It may be noted that most of the pairings are zero because h0|AB|0i can
be non-zero only if A is an annihilation operator and B is a creation operator.

Proof:
The theorem holds for n = 2.

Suppose it holds for n then we prove that it holds for n + 1 factors. Let
An+1 be the factor to be multiplied on the right of A1 . . . An .

If An+1 is an annihilation operator, then

A1 . . . An An+1 = : A1 . . . An : An+1 (zero pairings)


P
+ : A1 A2 A3 . . . Ak . . . An : An+1 (all 1-pairings)
P
+ : A1 A2 A3 . . . Ak . . . An : An+1 (all 2-pairings)
+···

Since An+1 is an annihilation operator, it can be pulled inside the normal


order symbol : :. All the pairings of An+1 with other operators are zero because
it comes on the right. Therefore the theorem is proved in this case.

If An+1 is an creation operator, then we have to take it across all the factors
creating pairings with each.
From : A1 . . . An : An+1 we will get n 1-pairings of A1 , A2 . . . , An with An+1
which can be written as
79

P
: A1 A2 A3 . . . . . . An+1 : (n such 1-pairings)

and, in addition, we get the zero-pairing term

An+1 : A1 . . . An : = : An+1 A1 . . . An : = : A1 . . . An An+1 : .

Similarly, each of the n C2 1-pairing terms will become (n − 2) 2-pairing terms


with An+1

: A1 A2 A3 . . . An+1 : (n − 2 such 2-pairings)


P

as well as the extra term with An+1 on the extreme left with all the old 1-
pairings. Since An+1 is a creation operator it can be taken from the extreme
left position inside the normal product and then it can travel to the last posi-
tion on the right because inside the normal product the order does not matter
for boson operators considered here.

The new n 1-pairings will combine with the previous n C2 1-pairings (in
which An+1 has been taken inside the normal product to give n+1 C2 1-pairings.
Similarly the old n C2 × n−2 C2 /2! 2-pairings will combine with the new n C2 ×
(n − 2) 2-pairings to give n+1 C2 × n−1 C2 /2! pairings. And so on.

The theorem is thus proved.

§ 68: Wick’s theorem for time-ordered products

Wick’s theorem for time-ordered products (of time dependent creation-


annihilation operators like those of fields) is just like that for ordinary prod-
ucts with the time-ordered or ‘chronological’ pairings replacing the ordinary
pairings. The chronological pairing of A1 A2 is defined as

h0|T (A1 A2 )|0i ≡ A1 A2 .

The theorem for this case states that


T (A1 . . . An ) = : A1 . . . An : (zero pairings)
P
+ : A1 A2 A3 . . . Ak . . . An : (all 1-pairings)
P
+ : A1 A2 A3 . . . Ak . . . An : (all 2-pairings)
+···
80 CHAPTER 5. INTERACTING BOSON FIELDS
The proof just uses the fact that the ordinary pairings are also ordered. If
x01 > x02 > · · · x0k · · · then

T (A1 A2 A3 . . . Ak . . . An ) = A1 A2 A3 . . . Ak . . . An .

Then the Wick theorem for ordinary products gives an expression like

: A1 A2 A3 . . . Ak . . . An :

where the pairings are already chronological:

h0|A1 A2 |0i = h0|T (A1 A2 )|0i, h0|A3 Ak |0i = h0|T (A3 Ak )|0i.

When time coordinates interchange their positions, the time ordering also
changes the positions of the factors and the ordinary pairing again is the same
as chronological pairing.

§ 69: Wick’s theorem for product of normal ordered expressions

Recall that we have used : φ(x)φ(x)Φ(x) : in the interaction Hamiltonian. In


second order we will encounter

T (: φ(x1 )φ(x1 )Φ(x1 ) :: φ(x2 )φ(x2 )Φ(x2 ) :).

Wick’s theorem can be applied to this product, as before except that we do


not pair any of the factors lying within the same normal ordered product. This
is so because all such pairings will have annihilation operator to the right (or
creation operator on the left).

§ 70: Feynman Rules

Feynman Rules are a procedure to draw all possible graphs and to write
down the amplitudes for a those graphs.

The above example shows that every occurrence of κ : φ(x)2 Φ(x) : produces
a vertex of the following kind
81

A
A
A 
A 
A 

x As







where the thick line corresponds to Φ and the two thin lines to φ. Each vertex
contributes a factor (−iκ). There will be a an integration d4 x for each vertex
also.

For an incoming particle in state f there is an external line coming from


below. If that lines reaches a vertex at point x, then there is a factor f (x).
Similarly for other incoming particles. For an outgoing particle in state f 0
emanating from a vertex at x, there will be a factor f 0 (x)∗ .

For every internal line connecting vertices x1 and x2 there will be a factor
∆F (x1 − x2 ) corresponding to that particle whose internal line it is.

We have to first decide


(1) the incoming and out going states.
(2) the perturbative order to which calculation is to be made
(3) then we make as incoming lines as the number of incoming particles, usu-
ally a lines coming from below
(4) make as outgoing lines as the number of outgoing particles, as lines going
up
(5) make as many vertices as the order of perturbation series, each vertex with
a coordinate x1 , x2 etc.
(6) complete the graph by joining the external lines and the lines from the
vertices.
(7) then use the above rules for calculating the amplitudes.
82 CHAPTER 5. INTERACTING BOSON FIELDS
Chapter 6

Spin 1/2 particle: Hilbert space


and wave-functions

The existence of spin one half particles (like electrons) is a proof of the fact
that the real symmetry of nature is not the Lorentz group, but another group
related to it, called SL(2, C). Some elementary facts about the group and its
close relation to the Lorentz group are given in the Appendix A.

§ 71: Spin of a particle

Spin is a relativistic concept. For any relativistic system, the eigenstates of


total momentum |p, αi may require other extra variables (denoted by α here),
for a complete description of the state. If it is a particle of mass m > 0, then
p · p = −m2 .

The subset of states with zero three-momentum (for which p = k =


(m, 0, 0, 0)) provide a representation of the group SU (2) (which corresponds to
rotations) because the momentum k does not change under rotations. There-
fore if U (R) is a unitary transformation for SU (2) matrix R, U (R)|k, αi is
again a linear combination of |k, αi’s:
X
U (R)|k, αi = |k, βiD(R)βα ,
β

where D(R)αβ matrices are a unitary representation of the SU (2) group. When
the representation is irreducible and labelled by s, then s is called the spin of
the particle.

§ 72: Momentum states for a spin half particle

83
84CHAPTER 6. SPIN 1/2 PARTICLE: HILBERT SPACE AND WAVE-FUNCTIONS
The procedure for construction is first to find simultaneous eigenvectors
of momentum √P and energy P 0 with eigenvalues p = (p0 , p1 , p2 , p3 ). For a
particle p0 = m2 + p2 . α = 1, 2 is the other discrete variables taking two
values for a spin half particle. We normalise these eigenstates as

hp, α|p0 , βi = 2ωp δ 3 (p − p0 )δαβ .

§ 73: Direct boost

The SL(2, C) matrix which corresponds to the Lorentz boost in the direction
of p to take four-momentum k = (m, 0, 0, 0) to p = (p0 , p) is called the direct
boost. We denote this Hermitian matrix by Bp . It is defined by the basic
relation connecting SL(2, C) with the Lorentz group:

Bp (k · σ)Bp† = p · σ.

Since
!
0 m 0
k · σ = k σ0 = ,
0 m

we can write

(Bp )2 = p · σ/m.

We find Bp to be

p.σ p.σ + m
r
Bp = =q , (6.1)
m 2m(p0 + m)

because a boost with velocity v = tanh α in the direction of the unit vector n
is given by cosh(α/2) + n.~σ sinh(α/2). Since this boost brings (m, 0, 0, 0) to
(p0 , p),
p p
p0 = m cosh α, |p| = m sinh α, n= =q ,
|p| (p0 )2 − m2

which determines
s s
p0 + m p0 − m
cosh(α/2) = , sinh(α/2) = .
2m 2m

§ 74: SU (2) in the rest frame


85
We start with the rest frame states |k, αi where α = 1, 2. There is only
one representation corresponding to spin s = 1/2, which is by these matrices
R ∈ SU (2) themselves:
X
U (R)|k, αi = |k, βiRβα .

§ 75: Construction of states in momentum space

Since Bp corresponds to transforming k to p, the state U (Bp )|k, αi has


momentum p. We choose to define the states |p, αi (without any change in α)
as

U (Bp )|k, αi = |p, αi.

For a general A ∈ SL(2, C) we know that U (A)|p, αi has momentum q =


L(A)p, and so
X
U (A)|p, αi = |q, βiXβα , q = L(A)p,

where the coefficients Xβα depend on A and p. We prove that X correspond


to the SU(2) matrix R(A, p)

X = R(A, p) = Bq−1 ABp , q = L(A)p.

We omit the summation in matrix multiplication for simplifying notation:

U (A)|p, αi = U (A) U (Bp )|k, αi


= U (Bq ) U (Bq−1 ) U (A) U (Bp )|k, αi
= U (Bq ) U (R(A, p))|kαi
= U (Bq )|kβiRβα (A, p)
= |q, βiRβα (A, p)

To make sure that R(A, p) is indeed SU(2)

RR† = (Bq−1 ABp )(Bq−1 ABp )† = (Bq−1 ABp2 A† Bq−1 )


1 −1 1
= Bq A(p.σ)A† Bq−1 = Bq−1 (q.σ)Bq−1
m m
1
= (k.σ) = I
m

This completes the construction of momentum space states. The SU(2)


matrix R(A, p) is called “Wigner rotation”.
86CHAPTER 6. SPIN 1/2 PARTICLE: HILBERT SPACE AND WAVE-FUNCTIONS
§ 76: Configuration space states

For a spinless particle the momentum states |pi which obey U (A)|pi = |qi, q =
L(A)p, the spacetime states
1 Z d3 p −ip.x
|xi = e |pi
(2π)3/2 2ωp
have the covariance property U (A)|xi = |L(A)xi. For spin 1/2 we cannot
construct
1 Z d3 p −ip.x
|x, αi = e |p, αi
(2π)3/2 2ωp
because of the p dependence of R(A, p). Instead, we define new states

|p, α) = |p, βi(Bp−1 )βα ,

so that

U (A)|p, α) = U (A)|p, βi(Bp−1 )βα ,


= |q, γiRγβ (A, p)(Bp−1 )βα , q = L(A)p
= |q, γi(Bq−1 ABp )γβ (Bp−1 )βα
= |q, γi(Bq−1 )γβ Aβα
= |q, β)Aβα

The normalization of these states is


(Is p).σ
(p, α|p0 , β) = 2ωp δ 3 (p − p0 )(Bp−2 )αβ = 2ωp δ 3 (p − p0 ) .
m
We can now define
1 Z d3 p −ip.x
|x, α) = e |p, α)
(2π)3/2 2ωp
so that

U (A)|x, α) = |L(A)x, β)Aβα

Since R(A, p) = Bq−1 ABp is unitary it also equal to its ‘inverse-dagger’

R(A, p) = (R(A, p))−1† = Bq A−1† Bp−1 .

This gives us another possibility to construct covariantly transforming states

|p, α)) = |p, βi(Bp )βα ,


87
which implies that

U (A)|p, α)) = |L(A)p, β))(A−1† )βα .

The normalization of these states is


p.σ
((p, α|p0 , β)) = 2ωp δ 3 (p − p0 )(Bp2 )αβ = 2ωp δ 3 (p − p0 ) .
m
And further, as before, define
1 Z d3 p −ip.x
|x, α)) = e |p, α))
(2π)3/2 2ωp
so that

U (A)|x, α)) = |L(A)x, β))(A−1† )βα .

§ 77: Space and time inversion

Parity (space-inversion) and time reversal (or time inversion) are defined only
on the 4-dimensional space-time. They do not have an analogue in 2 × 2
matrices. We know that

Is L(A)Is = It L(A)It = L(A−1† ),

and for a Poincare transformation we seek

Is U (a, A)Is = U (Is a, A−1† ).

If we call the Hilbert space operators for Is and It by P and T respectively,


then we expect for infinitesimal translations

P(iP 0 )P −1 = iP 0 , P(iP i )P −1 = −iP i i = 1, 2, 3

as well as

T (iP 0 )T −1 = −iP 0 , T (iP i )T −1 = iP i , i = 1, 2, 3.

These imply that parity can be represented by a unitary operator but time
reversal is to be represented by an anti-unitary operator to keep the states
with positive energy mapping into positive energy states.

For Lorentz transformations, the generators of Lorentz transformations


have the following relations:

P Jˆi P −1 = Jˆi , P K̂i P −1 = −K̂i


T Jˆi T −1 = −Jˆi , T K̂i T −1 = K̂i .
88CHAPTER 6. SPIN 1/2 PARTICLE: HILBERT SPACE AND WAVE-FUNCTIONS
§ 78: Parity

Since parity does not change the angular momentum, we can define

P|k, αi = ηP |k, αi.

Therefore, using

PU (Bp )P −1 = U (Is )U (Bp )U (Is )−1 = U (Bp−1† ) = U (Bp−1 ) = U (BIs p )

we find

P|p, α) = U (BIs p )ηP |k, βi(Bp−1 )βα = ηP |Is p, α)),

and,

P|x, α) = ηP |Is x, α)).

Similarly,

P|p, α)) = ηP |Is p, α), P|x, α)) = ηP |Is x, α).

§ 79: Time reversal

The time reversal operator T changes the sign of both the linear and the
angular momentum. Therefore time reversal is defined through a matrix. First
define it on the rest frame states. Let

T |k, αi = ηT |k, βiCβα .

Apply an R ∈ SU (2), ,

U (R)T |k, αi = ηT |k, γiRγβ Cβα .

On the other hand, write 1 = T T −1 to the left of U (R) and remember that

T −1 U (R)T = U (R)

because under T the generators of U (R) change sign, but the i accompanying
the generators also changes sign due to anti-unitary nature of T . Therefore,
 

U (R)T |k, αi = T U (R)|k, αi = T |k, βiRβα = ηT |k, γiCγβ Rβα

Comparing the two expressions for U (R)T |k, αi we find that


∗ −1T
Rγβ Cβα = Cγβ Rβα = Cγβ Rβα .
89
because R is unitary. Such a matrix C is already exists, and the conventional
choice is
!
0 −1
Cβα = −ε = .
1 0

On finite momentum states then


 
T |p, αi = T U (Bp )|k, αi = U (Bp−1 )T |k, αi = ηT |Is p, βiCβα
because under T the boost generators do not change while the i becomes −i
due to anti-unitary nature of T .

A point to note is that


 

T 2 |p, αi = T ηT |Is p, βiCβα = ηT∗ ηT |p, γiCγβ Cβα
= |ηT |2 |p, βi(ε2 )βα
so that
T 2 = −ηT2 = −1, T −1 = −T ,
because C = −ε is real with square −1, and we can choose ηT to be real
without upsetting anything else. Thus, briefly
T |pi = ηT |Is pi(−ε).

On states |p, α) and |p, α)) time reversal acts as follows. Omitting the
matrix indices,
 
T |p) = T |piBp−1 = ηT |Is piCBp−1∗ = ηT |Is pi(−ε)Bp−1∗ (ε)(−ε)
= ηT |Is piBp (−ε) = ηT |Is piBI−1
sp
(−ε)
= ηp |Is p)(−ε)
where we use the facts that
(1) −εAε = A−1T for any A ∈ SL(2, C),
(2) Bp is hermitian, and,
(3) Bp = BI−1
sp
.

Similarly,
T |p)) = ηp |Is p))(−ε).
Also using the anti-unitary nature of T it follows that
T |x) = ηp |It x)(−ε)
T |x)) = ηp |It x))(−ε)
90CHAPTER 6. SPIN 1/2 PARTICLE: HILBERT SPACE AND WAVE-FUNCTIONS
§ 80: Summary with matrix notation

The following equalities are written with self-explanatory matrix notation.


They are either definitions already given or a direct consequence of those def-
initions, whose details are to be filled in as exercises.

p.σ p.σ + m
r
Bp = = =q
m 2m(p0 + m)
Bp−1 = BIs p
hp|p0 i = 2ωp δ 3 (p − p0 )1
|p) = |piBp−1
|p)) = |piBp
Is p · σ
(p|p0 ) = 2ωp δ 3 (p − p0 )
m
0 3 0 p·σ
((p|p )) = 2ωp δ (p − p )
m
U (A)|p) = |q)A, q = L(A)p
U (A)|p)) = |q))A−1†
(q|U (A) = A−1† (p|, q = L(A)p
((q|U (A) = A((p|, q = L(A)p
P|pi = ηP |Is pi, |ηP | = 1.Do not assume ηP to be real.
P|p) = ηP |Is p))
P|p)) = ηP |Is p)
T |pi = ηT |Is pi(−ε)
T |p) = ηT |Is p)(−ε)
T |p)) = ηT |Is p))(−ε)
1 Z d3 p −ip·x
|x) = e |p)
(2π)3/2 2ωp
1 Z d3 p −ip·x
|x)) = e |p))
(2π)3/2 2ωp
1 Z 3
|p) = d x2ωp eip·x |x)
(2π)3/2 x0
1 Z 3
|p)) = 3/2
d x2ωp eip·x |x))
(2π) x0
Z 3
1 d p ip·(x−y) Is p · σ
 
(x|y) = e
(2π)3 2ωp m
Z 3
1 d p ip·(x−y) p · σ
 
((x|y)) = e
(2π)3 2ωp m
91
1 Z d3 p ip·(x−y)
(x|y)) = e = ((x|y)
(2π)3 2ωp
U (A)|x) = |L(A)x)A
U (A)|x)) = |L(A)x))A−1†
(Lx|U (A) = A−1† (x|
((Lx|U (A) = A((x|
P|x) = ηP |Is x))
P|x)) = ηP |Is x)
T |x) = ηT |It x)(−ε)
T |x)) = ηT |It x))(−ε)
Z 3 Z 3 Z 3
dp d p p·σ dp Is p · σ
1 = |pihp| = |p) (p| = |p)) ((p|
2ωp 2ωp m 2ωp m
Z 3 Z 3
dp dp
1 = |p)((p| = |p))(p|
2ωp 2ωp
Z h i
1 = i d3 x |x)∂0 ((x| − ∂0 |x)((x|
0
Zx h i
1 = i d3 x |x))∂0 (x| − ∂0 |x))(x|
x0

§ 81: Dirac equation, 4-spinor, and γ matrices

From the definition of |p) and |p))


|pi = |p)(Bp ) = |p))Bp−1 .
Therefore,
Is p · σ
|p) = |p))(Bp−1 )2 = |p))
m
2 p·σ
|p)) = |p)Bp = |p)
m
Is p · σ
(p| = (Bp−1 )2 ((p| = ((p|
m
p·σ
((p| = (Bp2 )(p| = (p|.
m
because Bp ’s are Hermitian matrices.

If |ψi is any vector in the Hilbert space then its representatives in these
can be written as two component column vectors
!
(p, 1|ψi
ψ(p) ≡ (p|ψi = ,
(p, 2|ψi
92CHAPTER 6. SPIN 1/2 PARTICLE: HILBERT SPACE AND WAVE-FUNCTIONS
!
∼ ((p, 1|ψi
ψ (p) ≡ ((p|ψi = .
((p, 2|ψi

They satisfy
" #" # " #
0 (Is p).σ ψα (p) ψα (p)
∼ =m ∼
p.σ 0 ψ α (p) ψ α (p)

where we use the relation Bp2 = p.σ/m and Bp−2 = (Is p).σ/m. This equation
is the celebrated Dirac equation in the momentum space. If the momentum
space wave function is defined as a column vector of size four, called a 4-spinor,
" #
ψα (p)
Ψ(p) = ∼
ψ α (p)

then it satisfies the more conventional form of Dirac equation in momentum


space:

(pµ γ µ + m)Ψ(p) = 0,

where pµ = (−p0 , p), and the 4 × 4 matrices γ µ are defined as


! !
0 0 σ0 i 0 σi
γ = , γ = , i = 1, 2, 3.
σ0 0 −σi 0

§ 82: ‘Configuration space’ wave functions

In the covariantly transforming space-time basis states (the so-called ‘configu-


ration space’), where
" #
ψα (x)
Ψ(x) = ∼
ψ α (x)
1 Z d3 p ip.x
ψα (x) ≡ (x, α|ψi = e ψα (p)
(2π)3/2 2ωp
∼ 1 Z d3 p ip.x ∼
ψ α (x) ≡ ((x, α|ψi = e ψ α (p)
(2π)3/2 2ωp

the Dirac equation looks like

(iγ µ ∂µ − m)Ψ(x) = 0,

for every vector |ψi ∈ H.


93
Remark 5 One must appreciate that we are dealing with three bases {|pi}, {|p)}
and {|p))} for constructing momentum space wave functions and similarly two bases
{|x)} and {|x))} for ‘configuration space’ wave-functions in the same Hilbert
space. For any given vector ψ ∈ H we can construct wave-functions in any of these
five bases. The bases |p) and |p)) are related by parity. Similarly the bases |x) and
|x)) are related by parity. The Dirac equation is simply a reminder that only one of
the bases is sufficient to describe the particle. As S. Weinberg put it, a ‘relativistic
wave equation’ just tells us which components of the wave function are superfluous.

It is proper to record here the avoidable confusion that the older notations carry.
In non-relativistic quantum mechanics the Schrodinger wave function is written as
ψ(x) as a generic name for any state ψ, φ, χ etc. Similarly, in relativistic quantum
mechanics of the electron, the 4 × 1 column vector ‘wave-function’ Ψ(x) stands for
any vector ψ, φ, χ of the corresponding Hilbert space. For the sake of clarity we will
denote Ψφ , Ψψ etc to denote the 4-spinor constructed from the wave functions of
the corresponding Hilbert space vectors.

§ 83: Properties of γ matrices

The constant matrices γ µ are made from the Pauli matrices, and have similar
properties which can be easily checked. (While verifying these properties one
should use the fact that when two square matrices are partitioned in a similar
fashion, one can multiply them as if these sub-matrices are individual matrix
elements.) Using the notation with 1 as the 4 × 4 identity matrix,
(γ 0 )† = γ 0 , (γ i )† = −γ i
and
(γ 0 )2 = 1, (γ 1 )2 = −I, (γ 2 )2 = −I, (γ 3 )2 = −I.
For for i, j = 1, 2, 3, i 6= j
γ 0 γ i + γ i γ 0 = 0, γ i γ j + γ j γ i = 0.
These are summarised in a single equation
γ µ γ ν + γ ν γ µ = −2η µν 1.
For later use we define a special matrix γ5 :
!
0 1 2 3 1 0
γ5 = −iγ γ γ γ = .
0 −1
It is hermitian matrix with square equal to identity, and it anti-commutes with
all the four matrices γ µ :
γ5 γ µ = −γ µ γ5 .
94CHAPTER 6. SPIN 1/2 PARTICLE: HILBERT SPACE AND WAVE-FUNCTIONS
§ 84: Inner product and matrix elements

We have already expressed the ‘resolution of identity’ formulas. For example


Z h i
1=i d3 x |x)∂0 ((x| − ∂0 |x)((x|
x0

gives
Z h ∼ ∼ i
hφ|ψi = i d3 x φ(x)† ∂0 ψ (x) − ∂0 φ(x)† ψ (x)
x0

The Dirac equation allows us to simplify it a little. Write


∼ ∼
i∂0 ψ = mψ − i~σ · ∇ ψ
∼†
−i∂0 φ† = m φ −i∇φ† · ~σ ,

so that
Z ∼† ∼ Z ∼

hφ|ψi = m 3
d x(φ ψ+ φ ψ) − i d3 x∇ · (φ†~σ ψ).
x0 x0

The second, surface at infinity term can be put equal to zero as wave functions
are supposed to vanish there. Thus
Z ∼† ∼ Z
hφ|ψi = m 3 †
d x(φ ψ+ φ ψ) = m d3 xΨ†φ (x)Ψψ (x).
x0 x0
Chapter 7

Spin 1/2 particle: Fock space


and fields

§ 85: Fock space for neutral spin 1/2 (Majorana) Particles

If there are no other symmetries other than the SL(2, C) and the transla-
tions, parity and time-reversal, then we can try to ‘second-quantize’ the spin
1/2 particles we have discussed so far. We do not decide if the Fock space
is symmetric or anti-symmetric, instead write a general commutator symbol
[, ]± where the plus sign denotes the anti-commutator, and the minus sign the
commutator.

First we will try to make a covariant space-time dependent field and then
apply locality or micro-causality.

Since U (a, A)|x) = |Lx + a)A, (we use L for L(A)),


U (a, A)a|x) U (a, A)−1 = A† a|Lx+a)
U (a, A)a†|x) U (a, A)−1 = AT a†|Lx+a) .
Similarly,
U (a, A)a|x)) U (a, A)−1 = A−1 a|Lx+a))
U (a, A)a†|x) U (a, A)−1 = A−1∗ a†|Lx+a) .

Remark 6 When we deal with column or row vectors of operators (that is, matri-
ces of operators with a single index) a confusion is bound to occur with the symbol
‘†’. If
!
A
φ= ,
B

95
96 CHAPTER 7. SPIN 1/2 PARTICLE: FOCK SPACE AND FIELDS
does φ† mean
!
† A†
φ = ?
B†
or

φ† = (A† , B † )?

That depends on whether we take the adjoint of just the operators or operators as
well as the matrix.

We use the following convention.

A matrix with only one index will be treated as a column matrix if it occurs to
the right of a square matrix (of appropriate size) so that the column index of the
square matrix is to be summed with that of the one-index matrix. If it occurs to
the left of a square matrix then it will be treated as a row. Thus if
!
C D
M=
E F

then
! ! !
C D A† CA† + DB †
M φ† = =
E F B† EA† + F B †
!
† † † C D
φM = (A , B ) = (A† C + B † E, A† D + B † F )
E F

Since operators do not commute, it is important to know the left-right order of


factors. For us the matrix notation is only a convenience of not having to write the
indices.

In our notation, indices of (one or two indexed) consecutive factors are to be


summed as in matrix multiplication.

In any case, when there is any possibility of doubt, it is safest to put back the
indices in full glory.


For notational simplicity write a(x) for a|x) and a (x) for a|x)) . These fields will
always be accompanied by their argument (x) etc. and will not be confused
with the space-time translation 4-vector a. The above equations will be written
as
U (a, A)a(x)U (a, A)−1 = A† a(Lx + a)
U (a, A)a† (x)U (a, A)−1 = AT a† (Lx + a)
∼ ∼
U (a, A) a (x)U (a, A)−1 = A−1 a (Lx + a)
∼† ∼†
U (a, A) a (x)U (a, A)−1 = A−1∗ a (Lx + a).
97
Under parity

Pa(x)P −1 = ηP∗ a (Is x),
∼†
Pa† (x)P −1 = ηP a (Is x)

P a (x)P −1 = ηP∗ a(Is x),
∼†
P a (x)P −1 = ηP a† (Is x)

And under time reversal,

T a(x)T −1 = ηT εa(It x),


T a† (x)T −1 = ηT εa† (It x).

To make local causal fields we have to combine both a and a† so that the combi-
nation commutes or anti-commutes with itself and its Hermitian adjoint. Un-
∼†
der Poincare transformations a(x) and a† (x) transform very differently, but a
can be made to transform like a(x) if we use the fact that the (real antisymmet-
ric SL(2, C)) matrix ε changes any SL(2, C) matrix into its inverse-transpose.
Therefore we try, a linear combination
∼†
ψ(x) = a(x) + λε a (x),

where λ is a complex number for the linear combination. (We can always
throw away an overall constant.) so that

U (a, A)ψ(x)U −1 (a, A) = A† ψ(Lx + a).

This will commute (or anti-commute) with itself at space like distances if
∼† ∼†
[ψ(x), ψ(y)]± = [a(x) + λε a (x), a(y) + λε a (y)]± = λ[(x|y))εT ± ε(y|x))]

is proportional to the Pauli Jordan function. As εT = −ε that will be the case


if we choose the anti-commutator.

Next, now that we have chosen the anti-commutator., let us check the
anti-commutator with

ψ † (x) = a† (x) + λ∗ ε a (x).

∼† ∼
[ψα (x), ψβ† (y)]+ = [aα (x) + λεαα0 aα0 (x), a†β (y) + λ∗ εββ 0 aβ 0 (y)]+
1 Z d3 p ip·(x−y) Is p · σ
 
= e
(2π)3 2ωp m αβ
Z 3
1 d p −ip·(x−y) p · σ
 
2
+|λ| εαα0 εββ 0 e
(2π)3 2ωp m β 0 α0
98 CHAPTER 7. SPIN 1/2 PARTICLE: FOCK SPACE AND FIELDS
In the second term,
p0 σ0 + pi σi
!
p·σ T Is p · σ
     
ε ε = [εBp2 ε−1 ]βα = (BIs p )αβ = =− .
m βα m αβ m αβ

Therefore, if we choose

|λ|2 = 1,

the right hand side can be written as


1 Z d3 p ip·(x−y)
!
1 ∂ ∂
−i σ0 0 + σi i [e − e−ip·(x−y) ].
m ∂x ∂x αβ
(2π)3 2ωp

which is a derivative of the Pauli-Jordan function. Since the space-like region


is an open set, therefore the Pauli-Jordan function and its derivatives with
respect to x (or y) are all zero when x and y are space-like separated.

We can now check if the field ψ(x) transforms covariantly by parity. We


know the transformation of the states |x) and |x)), which implies

Pψ(x)P −1 = ηP∗ a (Is x) + ληP εa† (Is x)
ηP∗ −1 ∼
" #

= ληP ε a (Is x) + ε a (Is x)
ληP
ηP∗ ∼
" #

= ληP ε a (Is x) − ε a (Is x) .
ληP

The right hand side becomes proportional to ψ † (Is x) if we choose


ηP∗
λ∗ = − .
ληP
Since |λ|2 = 1, it follows that

ηP∗ = −ηP ,

or ηP2 = −1.

The intrinsic parity of a Majorana particle is purely imaginary, equal to


either i or −i.

Under time-reversal the fields transform as

T a(x)T −1 = ηT∗ εa(It x)


T a† (x)T −1 = ηT εa† (It x)
∼ ∼
T a (x)T −1 = ηT∗ ε a (It x)
∼† ∼†
T a (x)T −1 = ηT ε a (It x)
99
Therefore,then
∼†
T ψ(x)T −1 = T [a(x) + λε a (x)]T −1
∼†
= ηT∗ εa(It x) + λ∗ ηT ε a (It x)
ηT∗
" #
∗ ∼†
= ηT ε a(It x) + λ a (It x) .
ηT

The right hand side will be related to ψ(It x) again if we choose

λ = λ∗ , and
ηT∗ = ηT ,

that is, both λ and ηT should be chosen to be real. Thus,

T ψ(x)T −1 = ηT εψ(It x).

From now on we choose λ = 1.

Although there is no real need, it is customary to make a 4-spinor field.


Define
∼ ∼
ψ (x) =a (x) − εa† (x).

Then

ψ(x)† = ε ψ (x)

ψ (x)† = −εψ(x)

Therefore if we define
!
ψ(x)
Ψ(x) = ∼ ,
ψ (x)

then
! ! !
† 0 ε ψ(x) 0 ε
Ψ(x) = ∼ = Ψ(x).
−ε 0 ψ (x) −ε 0

This is a kind of “reality condition” for the neutral Majorana particle.

§ 86: Fock space: Charged spin 1/2 field for (Dirac) Particles
100 CHAPTER 7. SPIN 1/2 PARTICLE: FOCK SPACE AND FIELDS
For a particle with charge q, in order to get a field covariant under the
U (1) corresponding to charge, we must combine the annihilation operator of
the particle with the creation operator of the anti-particle, just as we did for
the charge spin zero particles.

Denote the charge ±q states by the subscript ±. The states in the two
orthogonal Hilbert spaces H± are respectively

|x)+ , |x))+ ∈ H+ , |x)− , |x))− ∈ H− .

We denote the Fock space annihilation operators by (traditional)


∼ ∼
a(x) = a|x)+ , a (x) = a|x))+ , b(x) = a|x)− b (x) = a|x)+

and their adjoints as creation operators.

The Dirac field for the particle (of charge q) is the combination
∼†
ψ(x) = a(x) + λε b (x),

transforming with A† and


∼ ∼
ψ (x) = a (x) + λ0 εb† (x),

transforming with A−1 . Since a(x) commutes/anti-commutes with a(y) (both


being annihilation operators), and with b† (y) because vectors in the spaces H+
and H− are orthogonal there is no conclusion or condition. But to get

[ψ(x), ψ(y)† ]± = 0, for x and y space-like,

we are forced to choose the anti-commutator, |λ|2 = 1 as in the Majorana case.


Similarly, |λ0 |2 = 1. For
∼†
[ψ(x), ψ (y)] = 0, for x and y space-like,

we must require

λ
λ(λ0 )∗ = = −1.
λ0

For parity, let ηP ± be the intrinsic parities of the particle and the antipar-
ticle. Then
" #
∼ ∼ λη
Pψ(x)P −1 = ηP∗ + a (Is x) + ληP − εb† (Is x) = ηP∗ + a (Is x) + ∗P − εb† (Is x) .
ηP +
101

For the right hand side to be proportional to ψ (Is x),
ηP −
λ0 = λ = ληP − ηP + .
ηP∗ +

Similarly,
∼ ∼†
P ψ (x)P −1 = ηP∗ + a(Is x) + λ0 ηP − ε b (Is x)
λ0 ηP − †
" #

= ηP + a(Is x) + ∗ εb (Is x)
ηP +

so that
ηP −
λ = λ0 = λ0 ηP − ηP + .
ηP∗ +

Since λ = −λ0 , both the conditions require

ηP + ηP − = −1.

Thus the the intrinsic parity of a Dirac particle-antiparticle system is negative.

Under time-reversal the fields transform as

T a(x)T −1 = ηT∗ + εa(It x)


T a† (x)T −1 = ηT + εa† (It x)
∼ ∼
T a (x)T −1 = ηT∗ + ε a (It x)
∼† ∼†
T a (x)T −1 = ηT + ε a (It x)

and,

T b(x)T −1 = ηT∗ − εb(It x)


T b† (x)T −1 = ηT − εb† (It x)
∼ ∼
T b (x)T −1 = ηT∗ − ε b (It x)
∼† ∼†
T b (x)T −1 = ηT − ε b (It x)

Therefore, (bringing back indices for clarity, and omitting the space-time vari-
aible which is x on the left hand side and It x on the right),
∼†
T ψα T −1 = T (aα + λεαβ b β )T −1
∼†
= ηT∗ + εαγ aγ + λ∗ εαβ ηT − εβγ b γ

" #
λ η T − ∼†
= ηT∗ + ε a + ∗ ε b
ηT +
102 CHAPTER 7. SPIN 1/2 PARTICLE: FOCK SPACE AND FIELDS
Therefore , as |λ|2 = 1,
λ∗ ηT − ηT −
λ= ∗
= ∗ , that is λ2 ηT∗ + = ηT − .
ηT + ληT +

From the transformation of T ψ T −1 we get the same condition because λ2 =
(λ0 )2 .

All these restrictions can be taken care of by choosing

λ = 1, λ0 = −1, ηP + = real = −ηP − ≡ ηP , ηT + = real = ηT − ≡ ηT ,

and the fields are


∼†
ψ(x) = a(x) + ε b (x)
∼ ∼
ψ (x) = a (x) − εb† (x).

Finally, if we define the charge conjugation for the particle-antiparticle


states

C|p, α, qi = ηC+ |p, α, −qi, C|p, α, −qi = ηC− |p, α, qi

then
∼ ∼ ∼† ∼†
CaC −1 = ηC+

b, C a C −1 = ηC+

b, Ca† C −1 = ηC+ b† , C a C −1 = ηC+ b .

and
∼ ∼ ∼† ∼†
CaC −1 = ηC−

b, C a C −1 = ηC−

b, Ca† C −1 = ηC− b† , C a C −1 = ηC− b .

Therefore,
 
∼† ∼†
Cψ(x)C −1 = ηC+
∗ ∗
b + ηC− ε a = ηC+ b + ηC− ηC+ ε a .

Comparing that to
∼† ∼† ∼†
ψ =a −εb = −ε[b + ε a ]

we deduce that

ηC+ ηC− = 1,

Since there is no other restriction, we can take ηC to be real as well. Thus,


∼†
CψC −1 = ηC ε ψ

C ψ C −1 = −ηC εψ †
103
Remark 7 The above result (ηC+ ηC− = 1) leads to an important conclusion. The
intrinsic charge-conjugation parity of a Dirac particle-antiparticle pair is −1. A
state like b† a† Φ0 becomes

C(b† a† Φ0 ) = ηC+ ηC− a† b† Φ0 = a† b† Φ0 = −b† a† Φ0

because of the anti-commutation in the last step.

Summary

U (a, A)a(x)U (a, A)−1 = A† a(Lx + a)


U (a, A)a† (x)U (a, A)−1 = a† (Lx + a)A
∼ ∼
U (a, A) a (x)U (a, A)−1 = A−1 a (Lx + a)
∼† ∼†
U (a, A) a (x)U (a, A)−1 = a (Lx + a)A−1†
∼ ∼†
The formulas for b, b† , b , b are exactly the same as for corresponding a’s.

Pa(x)P −1 = ηP a (Is x)

P a (x)P −1 = ηP a(Is x)

Pb(x)P −1 = −ηP b (Is x)

P b (x)P −1 = −ηP b(Is x), ηP real.

∼ ∼† ∼ ∼†
T A(x)T −1 = ηT εA(It x), ηT = real, A = {a, a† , a, a , b, b† , b , b }

CA(x)C −1 = ηC B(x), CB(x)C −1 = ηC A(x),

where, ηC is real and the left and right hand sides have the corresponding
members from the sets
∼ ∼† ∼ ∼†
A = {a, a† , a, a }, B = {b, b† , b , b }.

§ 87: Four-spinor formalism

Almost all work, old or recent, on spin half fields is expressed in 4-spinor
formalism. Define
!
ψ(x)
Ψ(x) = ∼ .
ψ (x)
104 CHAPTER 7. SPIN 1/2 PARTICLE: FOCK SPACE AND FIELDS
With this definition,
!
A† 0
U (a, A)Ψ(x)U (a, A)−1 = Ψ(L(A)x + a)
0 A−1
!
† −1 † A 0
U (a, A)Ψ(x) U (a, A) = Ψ(L(A)x + a) −1†
0 A
PΨ(x)P −1 = ηP γ 0 Ψ(Is x),
PΨ(x)† P −1 = ηP Ψ(Is x)† γ 0 ,
!
−1 ε 0
T Ψ(x)T = ηT Ψ(It x)
0 ε
!
ε 0
T Ψ(x)† T −1 = −ηT Ψ(It x)†
0 ε
!
−1 0 ε
CΨ(x)C = ηC Ψ(x)†
−ε 0
!
0 ε
CΨ(x)† C −1 = ηC Ψ(x) , Ψ† and Ψ are rows here.
−ε 0

Note that in the last equation Ψ(x)† is a column vector and not a row because
it occurs to the right.

Define

Ψ(x) ≡ Ψ(x)† γ 0 .

The Dirac equations can be written down for Ψ(x), just as for wave functions.

(iγ µ ∂µ − m)Ψ(x) = 0,

Ψ(x)(iγ µ ∂µ +m) = 0.

§ 88: Bilinear ‘invariants’

Expressions of the type Ψ(x)(γ µ · · · γ ν )Ψ(x) are called bilinear invariants


because of their transformation properties.

Product of any number of γ matrices can always be reduced to a sum of


product of lower number of these matrices by bringing two matrices with the
same µ index together using the rule γ µ γ ν = −γ ν γ µ − 2η µν 1 and then the
square of the matrix gives ±1. So essentially there are only five independent
possibilities:
105

Scalar 1,
Vector γ µ, µ = 0, 1, 2, 3
µ ν
(Antisymmetric) Tensor γ γ , µ, ν = 0, 1, 2, 3, µ 6= ν
µ ν τ
Pseudovector γ γ γ , µ, ν, τ = 0, 1, 2, 3 all different,
0 1 2 3
Pseudoscalar γ γ γ γ proportional to γ5

Scalar:
The scalar combination is Ψ(x)Ψ(x):

U (a, A) (Ψ(x)Ψ(x)) U (a, A)−1 = Ψ(Lx + a)Ψ(Lx + a),


P (Ψ(x)Ψ(x)) P −1 = Ψ(Is x)Ψ(Is x)
T (Ψ(x)Ψ(x)) T −1 = Ψ(It x)Ψ(It x)
C (Ψ(x)Ψ(x)) C −1 = Ψ(x)Ψ(x)

Psuedo-scalar:

U (a, A) (Ψ(x)γ5 Ψ(x)) U (a, A)−1 = Ψ(Lx + a)γ5 Ψ(Lx + a)


P (Ψ(x)γ5 Ψ(x)) P −1 = −Ψ(Is x)γ5 Ψ(Is x)
T (Ψ(x)γ5 Ψ(x)) T −1 = Ψ(It x)γ5 Ψ(It x)
C (Ψ(x)γ5 Ψ(x)) C −1 = Ψ(x)γ5 Ψ(x)

Vector:

U (a, A) (Ψ(x)γ µ Ψ(x)) U (a, A)−1 = (L−1 )µ ν Ψ(Lx + a)γ ν Ψ(Lx + a)

Proof:
If we use the ‘parity reversed’ Pauli matrices

σ P = {σ0 , −σ1 , −σ2 , −σ3 }

then the gamma matrices are, (in terms of 2 × 2 component matrices)


!
µ 0 σµ
γ = .
σµP 0

The proof depends on the defining equation

(Lx) · σ = A(x · σ)A† , L = L(A)

which is equivalent to the following identities,

σL = AσA† , σ(L−1 )T = A−1† σA−1 , σ P (L−1 )T = Aσ P A† ,


106 CHAPTER 7. SPIN 1/2 PARTICLE: FOCK SPACE AND FIELDS
the last one of which is obtained by multiplying with the parity matrix Is from
the right to the first identity and using Is LIs = (L−1 )T .

Now, calling x0 = L(A)x + a, to reduce writing,

U (a, A) (Ψ(x)γ µ Ψ(x)) U (a, A)−1


! !
0 † A 0 A† 0
= Ψ(x ) 0 µ
γ γ Ψ(x0 )
0 A−1† 0 A−1
! ! !
A 0 σP 0 A† 0
= Ψ(x0 )† −1† Ψ(x0 )
0 A 0 σ 0 A−1
!
0 † Aσ P A† 0
= Ψ(x ) Ψ(x0 )
0 A−1† σA−1
!
σP 0
= Ψ(x0 )† Ψ(x0 )(L−1 )T
0 σ
µ
= Ψ(x0 )γ ν Ψ(x0 )(L−1T )ν
= (L−1 )µ ν Ψ(x0 )γ ν Ψ(x0 )

Axial-vector:
The three gamma matrices γ µ γ ν γ τ with µ, ν, τ all different, can always be
brought to the form γ µ γ5 . The bilinear invariant is:

U (a, A) (Ψ(x)γ µ γ5 Ψ(x)) U (a, A)−1 = (L−1 )µ ν Ψ(Lx + a)γ ν γ5 Ψ(Lx + a).

Under parity it changes sign.

Tensor:
Similarly in the ‘tensor’ case of two gamma matrices, γ µ γ ν , µ 6= ν we can use
1
Σµν = (γ µ γ ν − γ ν γ µ )
2
without having to specify that µ 6= ν.

The remaining details of these are invariants are left as exercise.

Вам также может понравиться