Professional Documents
Culture Documents
Path Integrals by Wipf
Path Integrals by Wipf
Andreas Wipf
Theoretisch-Physikalisches-Institut
Friedrich-Schiller-Universität, Max Wien Platz 1
07743 Jena
5. Auflage WS 2008/09
1. Auflage, SS 1991 (ETH-Zürich)
I ask readers to report on errors in the manuscript and hope that the corrections
will bring it closer to a level that students long for but authors find so elusive.
(email to: wipf@tpi.uni-jena.de) September 18, 2015
Contents
1 Introduction 4
4 Perturbation Theory 28
4.1 Perturbation expansion for the propagator . . . . . . . . . . . . . . . . . . . . 28
4.2 Quartic potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1
CONTENTS Contents 2
7 Brownian motion 60
7.1 Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
7.2 Discrete random walk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.3 Scaling limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
7.4 Expectation values and correlations . . . . . . . . . . . . . . . . . . . . . . . 65
7.5 Appendix A: Stochastic Processes . . . . . . . . . . . . . . . . . . . . . . . . 66
8 Statistical Mechanics 72
8.1 Thermodynamic Partition Function . . . . . . . . . . . . . . . . . . . . . . . . 72
8.2 Thermal Correlation Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 73
8.3 Wigner-Kirkwood Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
8.4 High Temperature Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
8.5 High-T Expansion for D/2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
8.6 Appendix B: Periodic Greenfunction . . . . . . . . . . . . . . . . . . . . . . . 84
9 Simulations 87
9.1 Markov Processes and Stochastic Matrices . . . . . . . . . . . . . . . . . . . . 88
9.2 Detailed Balance, Metropolis Algorithm . . . . . . . . . . . . . . . . . . . . . 92
9.2.1 Three-state system at finite temperature . . . . . . . . . . . . . . . . . 93
10 Berezin Integral 95
10.1 Grassmann variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
————————————
A. Wipf, Path Integrals
CONTENTS Contents 3
————————————
A. Wipf, Path Integrals
Chapter 1
Introduction
These lectures are intended as an introduction to path or functional integration techniques and
their applications in physics. It is assumed that the participants have a good knowledge in
quantum mechanics. No prior exposure to path integrals is assumed, however.
We are all familiar with the standard formulations of quantum mechanics, developed by
H EISENBERG , S CHR ÖDINGER and others in the 1920s. In 1933, D IRAC speculated that in
quantum mechanic the classical action S might play a similarly important role as it does in
classical mechanics. He arrived at the conclusion that the amplitude for the propagation from
the initial position q 0 at time 0 to the final position q at time t,
is given by
where wcl is the classical trajectory from q 0 to q in time t. The exponent is dimensionless,
since the reduced Planck-constant h̄ has the dimension of an action. For a free particle with
Hamiltonian and Lagrangian
1 2 m 2
H0 = p and L0 = q̇ (1.3)
2m 2
the above formula is easily checked: free particles move on straight lines such that the trajectory
w(s) of a particle moving from q 0 to q and the corresponding action read
Z t
1 m
w(s) = {sq + (t − s)q 0 } and S = dt L0 (w, ẇ) = (q − q 0 )2 . (1.4)
t 0 2t
Following Diracs suggestion this leads to the amplitude
0 2 /2h̄t
K0 (t, q, q 0 ) ∼ eim(q−q ) . (1.5)
4
CHAPTER 1. INTRODUCTION 5
The amplitude for an individual path is ∼ exp (iS[path]/h̄) and the amplitudes for all paths are
added according to the usual rule for combining probability amplitudes,
K(t, q, q 0 ) ∼ eiS[path]/h̄ .
X
(1.10)
paths q 0 →q
Surprisingly enough, the same calculus (in the sense of a analytical continuation) was already
known to mathematicians due to the work of W IENER in the study of stochastic processes. This
calculus in functional space attracted the attention of other mathematicians, including K AC, and
was subsequently further developed. The standard reference concerning these achievements is
the review of G ELFAND and YAGLOM [5], where the early work was first critically discussed.
The path integral method had its great, early successes in the 1950s and its implications
have been beautifully expounded in Feynmans original review paper [3] and in his book with
H IBBS [4]. This book contains many applications and still serves as a standard literature on
path integrals.
————————————
A. Wipf, Path Integrals
CHAPTER 1. INTRODUCTION 6
Path integration provides a unified view of quantum mechanics, field theory and statistical
physics and is nowadays a irreplaceable tool in theoretical physics. It is an alternative to the
Hamiltonian method for quantizing classical systems and solving problems in quantum me-
chanics and quantum field theories.
These lectures should introduce you both into the formalism and the techniques of path
integration. We shall discuss applications that will convince you that path integrals are worth
studying not only for reasons of beauty but also for practical purposes.
Path integrals in quantum mechanics and quantum field theory are ideally suited to deal with
problems like
• Incorporating constraints
In the first part of these lecture we shall reformulate ordinary quantum mechanics in Feynmans
path integral language. We shall see how to manipulate path integrals and we shall apply the
results to simple physical systems: the harmonic oscillator with constant and time dependent
frequency and the driven oscillator. Then we consider the path integral for imaginary time
and give a precise meaning to the sum over all paths. Functional determinants show up in
many path integral manipulations and we devote a whole section to these objects. It follows a
chapter on the path integral approach to quantum systems in thermal equilibrium. We derive the
semiclassical and high-temperature expansions to the partition functions and conclude the part
on quantum mechanics with Monte Carlo simulations of discretized Quantum Mechanics.
In the second part these lectures a simple field-theoretical model, namely the Schwinger
model or QED in 2 dimensions, is introduced and solved. This model is interesting for vari-
ous reasons. Due to quantum correction the ’photon’ acquires a mass and the classical chiral
symmetry is broken like it is in QCD. These model allows us to introduce many relevant field
theoretical concepts like regularization, Berezin-integrals, gauge fixing and perturbation theory.
Then we deal with anomalies and effective actions. We shall see how to employ path integral
techniques to compute anomalies in gauge theories. We ’integrate’ certain anomalies and derive
————————————
A. Wipf, Path Integrals
CHAPTER 1. INTRODUCTION 7
the Casimir effect in external fields. Finally we shall compute the particle production in external
electromagnetic and gravitational fields.
In the last part of these lectures we study the lattice version of field theories. In particular
we introduce and discuss the symmetry breaking by means of effective potentials. Then the nu-
merical simulations of scalar theories on a finite lattice is discussed. Finally I shall explain how
to formulate gauge field theories with fermions on a space-time lattice and the some problems
of these lattice gauge theories.
There are many good books and review articles on path integrals. I have listed some ref-
erences which I suggest for further readings. In particular the references [1]-[9] contain in-
troductory material. These references are only a very small and subjective selection from the
extensive literature on functional integrals. In the bibliography at the end of these lectures you
find further references on particular topics of path integrals.
————————————
A. Wipf, Path Integrals
Chapter 2
Quantization is a procedure for constructing a quantum theory starting from a classical the-
ory. There are different approaches to quantizing a classical system, the prominent ones being
canonical quantization and path integral quantization1. In this course I shall assume that you
are familiar with the first one, that is the wave mechanics developed by S CHR ÖDINGER and the
matrix mechanics due to B ORN , H EISENBERG and J ORDAN. Here we only recall the important
steps in a canonical quantization of a classical system.
and this structure naturally extends to observables by the derivation rule {OP, Q} = O{P, Q}+
{O, Q}P and the antisymmetry of the brackets. The time-evolution of any observable is deter-
mined by its equation of motion
Now one may ’quantize’ a classical system by requiring that observables become hermitean
linear operators and Poisson brackets are replaced by commutators:
1
O(p, q) → Ô(p̂, q̂) and {O, P } −→ [Ô, P̂ ]. (2.3)
ih̄
1
Others would be geometric and deformation quantization.
8
CHAPTER 2. DERIVING THE PATH INTEGRAL 2.1. Recall of Quantum Mechanics 9
In passing we note, that according to a famous theorem of G ROENEWOLD [10], later extended
by VAN H OVE [11], there is no invertible linear map from all functions O(p, q) of phase space
to hermitean operators Ô in Hilbert space, such that the Poisson-bracket structure is preserved.
It is the Moyal bracket, the quantum analog of the Poisson bracket based on the Weyl corre-
spondence map, which maps invertible to the quantum commutator.
The evolution of observables which do not explicitly depend on time is determined by the
Heisenberg equation of motion
d i
Ô = [Ĥ, Ô] =⇒ Ô(t) = eitH/h̄ Ô(0)e−itH/h̄ . (2.4)
dt h̄
In particular the phase-space coordinates become operators and their equations of motion read
d i d i i
p̂i = [Ĥ, p̂i ] and q̂ = [Ĥ, q̂ i ] with [q̂i , p̂j ] = ih̄δ ij . (2.5)
dt h̄ dt h̄
For example, for a non-relativistic particle with Hamilton operator
1 X 2
Ĥ = Ĥ0 + V̂ , with Ĥ0 = p̂i (2.6)
2m
one finds the familiar equations of motion,
d d i p̂i
p̂i = −V̂,i and q̂ = . (2.7)
dt dt 2m
Observables are represented as hermitean linear operators acting on a separable Hilbert space
H (the elements of which define the states of the system)
Here we do not distinguish between an observable and the corresponding hermitean operator.
In the coordinate representation the Hilbert space for a particle on the line is the space L2 (R)
of square integrable functions on R and and the position- and momentum operators are
h̄
(q̂ψ)(q) = qψ(q) and (p̂ψ)(q) = ∂q ψ(q). (2.9)
i
In experiments we have access to matrix elements of observables. For example, the expectation
values of an observable in a given state is given by the diagonal matrix element2 hψ|O(t)|ψi.
The time-dependence of expectation values is determined by the Heisenberg equation (2.4). We
may perform a t-dependent similarity transformation from the Heisenberg- to the Schrödinger
picture,
————————————
A. Wipf, Path Integrals
CHAPTER 2. DERIVING THE PATH INTEGRAL 2.1. Recall of Quantum Mechanics 10
The values of observable matrix elements do not depend on the chosen picture. After the picture
changing transformation {O(t), |ψi} −→ {Os , |ψs (t)i} the states evolve in time according to
the Schrödinger equation
d
ih̄ |ψs i = H|ψs i. (2.13)
dt
The solution is given by the time evolution (2.10),
and depends linearly on the initial state vector |ψs (0). In the coordinate representation this
solution takes the form
Z
ψs (t, q) ≡ hq|ψs (t)i = hq|e−itH/h̄ |q ′ihq ′ |ψs (0)idq ′
Z
= K(t, q, q ′ )ψs (0, q ′)dq ′ , (2.15)
where we made use of the completeness relation for the position eigenstates,
Z
dq ′ |q ′ihq ′ | = 1 (2.16)
It is the probability amplitude for the particle to propagate from q ′ at time 0 to q at time t and is
occasionally denoted by
This evolution kernel (sometimes called propagator) will be of great importance when we
switch to the path integral formulation. It satisfies the time dependent Schrödinger equation
d
ih̄ K(t, q, q ′ ) = HK(t, q, q ′), (2.19)
dt
————————————
A. Wipf, Path Integrals
CHAPTER 2. DERIVING THE PATH INTEGRAL 2.2. Feynman-Kac Formula 11
where H acts on the coordinates q of the final position. In addition K obeys the initial condition
The propagator is uniquely determined by the differential equation and initial condition. For a
non-relativistic free particle in Rd with Hamiltonian H0 as in (2.6) it is a Gaussian function of
the initial and final coordinates q ′ , q ∈ Rd , ,
m
r
′ 2 /2h̄t
K0 (t, q, q ′ ) = hq|e−itH0 /h̄ |q ′ i = Adt eim(q−q ) , At = . (2.21)
2πih̄t
The factor proportional to t−d/2 infront of the exponential function is needed to recover the
δ-distribution in the limit t → 0, see (2.20). In one dimension one has
′ 2 /2h̄t
K0 (t, q, q ′) = At eim(q−q ) . (2.22)
After this preliminaries we now turn to the path integral representation of the evolution kernel.
To prove this simple formula we introduce the n’th roots of the matrices on both sides in (2.23),
namely Sn := exp[(A + B)/n] and Tn := exp[A/n] exp[B/n] and telescope the difference
Since the matrix-norms of the sum and product of two matrices X and Y satisfy kX + Y k ≤
kXk + kY k and kX · Y k ≤ kXk · kY k it follows at once that k exp(X)k ≤ exp(kXk) and
————————————
A. Wipf, Path Integrals
CHAPTER 2. DERIVING THE PATH INTEGRAL 2.2. Feynman-Kac Formula 12
Finally, using Sn − Tn = −[A, B]/2n2 + O(1/n3) this proves the product formula for matrices.
This theorem and its proof can be extended to the case where A and B are self-adjoint operators
and their sum A + B is (essentially) self-adjoint on the intersection D of the domains of A and
B:
n
e−it(A+B) = s − lim e−itA/n e−itB/n . (2.24)
n→∞
With the strong limit one means that the convergence holds on all states in D. The first for-
mulation is relevant for quantum mechanics and the second is needed in statistical mechanics
and diffusion problems. For a proof of the Trotter product formula for operators I refer to the
mathematical literature [21, 22].
Using the product formula (2.24) in the evolution kernel (2.17) yields
Inserting n − 1-times the resolution of the identity 1 = dwj |wj i hwj | associated with the
R
position eigenstates, we obtain for the matrix element on the right hand side
hq| e−itH0 /h̄n e−itV /h̄n 1 e−itH0 /h̄n e−itV /h̄n 1 . . . 1 e−itH0 /h̄n e−itV /h̄n |q ′ i
Z j=n−1
hwj+1| e−itH0 /h̄n e−itV /h̄n |wj i .
Y
= dw1 · · · dwn−1 (2.27)
j=0
In the last formula wn = q is the final position and w0 = q ′ the initial position of the particle.
Since the potential is diagonal in the coordinate representation we find
hwj+1| e−itH0 /h̄n e−itV /h̄n |wj i = hwj+1 | e−itH0 /h̄n |wj i e−itV (wj )/h̄n . (2.28)
Now we insert the evolution kernel (2.22) of the free particle and find for Kn the representation
Z
′ (n) (w)/h̄
K(t, q, q ) = = lim An dw1 · · · dwn−1 · eiS
n→∞ ǫ
m n−1 n−1
2
wj+1 − wj
(n)
X X
S (w) = ǫ − ǫV (wj ) (2.29)
2 j=0 ǫ j=0
where ǫ = t/n. This is the celebrated formula of Feynman and Kac and it is just the path
integral representation of the evolution kernel we have been aiming at.
————————————
A. Wipf, Path Integrals
CHAPTER 2. DERIVING THE PATH INTEGRAL 2.2. Feynman-Kac Formula 13
b w1
b w3 b
b
b
b
w2 b
wn =
b
q
b
w0 = q ′ b
0 ǫ 2ǫ 3ǫ nǫ
To see more clearly why (2.29) is called a path integral (or functional integral in field theory)
we divide the time interval [0, t] into n equidistant intervals with length ǫ = t/n and identify wk
with w(s = kǫ), see fig. (2.1). Now we connect every pair of points (jǫ, wj ) and (jǫ + ǫ, wj+1 )
by a straight line and obtain a broken line path from w0 = q ′ to wn = q
The exponent SL in (2.29) is a Riemann sum approximation to the classical action of a
particle moving along this broken line path,
Zt
m 2
(n) n→∞
S (w) −→ ds ẇ − V (w) ≡ S[w] (2.30)
2
0
The integrations dw1 . . . dwn−1 in (2.29) is to be interpreted as summing over all possible
R
broken line paths connecting q ′ with q. Since any continuous path from q ′ with q can be ap-
proximated by a broken line path and since finally we must take the continuum limit n → ∞ or
equivalently ǫ → 0, we may interpret the integral (2.29) as a sum over all paths from q ′ at time
0 and to q at time t. The ǫ-dependent constant
n/2
m
Anǫ = (2.31)
2πih̄ǫ
in the path integral (2.29) is required to obtain a unitary time evolution. It diverges in the
continuum limit ǫ → 0, but this divergence is harmless as we shall see later. In the continuum
limit we denote the path integral representation for the evolution kernel (2.29) by
w(t)=q
Z
′
K(t, q, q ) = Dw eiS[w]/h̄ , (2.32)
w(0)=q ′
with the formal ’measure’ Dw on the set of paths defined by the limit (2.29). Since the infinite
product of Lebesgue measures like ∞ 1 dwj fails to be a measure, the symbol Dw is mathemati-
Q
cally not well-defined. However, one can define a measure on the set of paths if one analytically
————————————
A. Wipf, Path Integrals
CHAPTER 2. DERIVING THE PATH INTEGRAL 2.3. Non-stationary systems 14
continues to imaginary time. For more general Lagrangian systems, for example for particles
propagating in 3 dimensions, a similar path integral representation for the evolution kernel can
be given. In same cases, for example for particle in external fields, ordering ambiguities in the
canonical approach translate into discretization ambiguities even in the continuum limit.
where T denotes the time ordering. The generalization to time-dependent Hamiltonians is useful
when on considers system under varying external conditions, for example in a time-dependent
external field. In a non-stationary situation the evolution operator depends on the initial and
final times and not only on the time-difference t − t′ . But the continuum path integral for the
evolution kernel K looks the same as in the stationary case,
w(t)=q
Z
′ ′
K(t, q, t , q ) = Dw eiS[w]/h̄, (2.34)
w(t′ )=q ′
where now the Lagrange function depends explicitly on time. For a system with Hamiltonian
H = H0 + V (t) the path integral is the continuum limit of
Z
′ ′ (n) (w)/h̄
K(t, q, t , q ) = lim An dw1 · · · dwn−1 eiS
n→∞ ǫ
m n−1 n−1
X wj+1 − wj 2
S (n) (w) = ǫV (t′ + jǫ, wj ),
X
− (2.35)
2 j=0 ǫ j=0
where ǫ = (t−t′ )/n. Note that now the potential depends on the (discretized) time. To establish
(2.34) we show that
Z
ψ(t, q) = K(t, q, t′ , q ′ ) ψ(t′ , q ′ ) dq ′ (2.36)
obeys the time-dependent Schrödinger equation. For that purpose we set t′ = t − ǫ with small
ǫ. The evolution for a infinitesimal time step ǫ is given by
im iǫ
Z
ψ(t, q) = Aǫ dq ′ exp (q − q ′ )2 − V (t − ǫ, q ′ ) ψ(t − ǫ, q ′ ).
2h̄ǫ h̄
Changing variables according to q → q ′ + u this reads
Z
2 /2h̄ǫ
ψ(t, q) = Aǫ du eimu e−iǫV (t−ǫ,q+u)/h̄ ψ(t − ǫ, q + u). (2.37)
————————————
A. Wipf, Path Integrals
CHAPTER 2. DERIVING THE PATH INTEGRAL 2.4. Greensfunctions 15
Due to the first Gaussian factor the u-integral gets its main contribution from the neighborhood
of u = 0 and thus we may expand the last two factors in powers of u. The resulting integrals
over u are computed with the help of the formula
!n
1 ih̄ǫ
Z
2n imu2 /2h̄ǫ
du u e = (2n − 1)!! (2.38)
Aǫ m
where 0!! = (−1)!! = 1 by definition. Of course, the integrals with odd powers of u vanish. We
only need the terms of order 1 and ǫ on the right hand side in (2.37) and thus it is sufficient to
expand ψ to second order in u. Up to terms of order ǫ2 we find
u2
!
iǫ
Z
imu2 /2h̄ǫ
ψ(t, q) = Aǫ du e 1 − V (t, q) ψ(t − ǫ, q) + ψ ′′ (t, q) + O(ǫ2 ).
h̄ 2
The integration over u finally leads to
iǫ ih̄ǫ ′′
ψ(t, q) = ψ(t − ǫ, q) − V (t, q)ψ(t, q) + ψ (t, q) + O(ǫ2 ). (2.39)
h̄ m
Note that for ǫ → 0 the right hand side converges to ψ(t, q) such that K converges to the identity
as t′ → t. Now we subtract ψ(t−ǫ, q) from both sides in (2.39) and divide the resulting equation
by ǫ. In the continuum limit ǫ → 0 we recover the time-dependent Schrödinger equation,
∂ψ h̄2 ′′
=−ih̄ ψ + V (t)ψ, (2.40)
∂t 2m
and this shows that even for a time-dependent Hamiltonian the propagator is given by the path
integral (2.34) or more accurately by (2.35).
2.4 Greensfunctions
In quantum field theory one is interested in vacuum expectation values of time-ordered products
of Heisenberg field operators since these objects are related to amplitudes of physical processes
such as scattering amplitudes or decay rates of particles. We look at the analogous objects in
quantum mechanics:
G(n) (t1 , t2 , . . . , tn ) = hΩ|T q̂(t1 )q̂(t2 ) · · · q̂(tn )|Ωi, (2.41)
where|Ωi represents the vacuum state and the position operator has the time dependence
q̂(t) = eitH/h̄ q̂e−itH/h̄ , (2.42)
see equation (2.4). The objects G(n) are known as Greensfunction or correlation functions. The
time ordering operator T orders its arguments such that the operator at earliest time acts first (is
the right-most), the operator at the second earliest time acts next etc. For example
(
q̂(t1 )q̂(t2 ) t1 > t2
T q̂(t1 )q̂(t2 ) = (2.43)
q̂(t2 )q̂(t1 ) t2 > t1 .
————————————
A. Wipf, Path Integrals
CHAPTER 2. DERIVING THE PATH INTEGRAL 2.4. Greensfunctions 16
Now will derive the path integral expression for the Greensfunction (2.41). Actually we shall
calculate correlation functions with fixed endpoints, for example
is the past-evolved position eigenstate, q̂(t)|t, qi = q|t, qi. Later we shall see how one recovers
the vacuum expectation values (2.41) from the correlation functions with fixed endpoints. We
assume t1 > t2 and insert twice the identity in (2.44), one after every position operator q̂,
Inserting the path integral representation for the three matrix elements we obtain
Z w(t)=q
Z w(tZ
1 )=w1 w(tZ
2 )=w2
′ iS/h̄ iS/h̄
hq, t| q̂(t1 )q̂(t2 )|q i = dw1dw2 w1 w2 Dw e Dw e Dw eiS/h̄ . (2.45)
q(t1 )=w1 w(t2 )=w2 q(0)=q ′
This expression consists of a first path integral from the initial position q ′ to the position w2 , a
second one from w2 to the position w1 , and a third one from w1 to the final position q. So we
are integrating over all paths from q ′ to q, subject to the restriction that the paths pass through
the intermediate points w2 and w1 at times t2 and t1 , respectively. Finally we integrate over the
two arbitrary positions w2 and w1 , so that in fact we are integrating over all paths. Thus we may
combine the three path integrals and the integrations over w1 and w2 into a single path integral.
The factors w1 and w2 in the integrand are just the values w(t1 ) and w(t2 ) of the paths at the
intermediate times. Hence we end up with
w(t)=q
Z
′
hq, t| q̂(t1 )q̂(t2 )|q i = Dw w(t1 )w(t2 ) eiS[w]/h̄ (t1 > t2 ). (2.46)
w(0)=q ′
A similar calculation reveals that the same result holds true for the matrix element of q̂(t2 )q̂(t1 )
when t2 > t1 . The path integral takes care of the time ordering. Thus we arrive at the following
formula for all pairs t1 , t2 :
w(t)=q
Z
hq, t| T q̂(t1 )q̂(t2 )|q ′ i = Dw w(t1 )w(t2) eiS[w]/h̄ . (2.47)
w(0)=q ′
————————————
A. Wipf, Path Integrals
CHAPTER 2. DERIVING THE PATH INTEGRAL 2.4. Greensfunctions 17
Now we relate the time ordered correlation functions for fixed endpoints to the vacuum expecta-
tion values in (2.41). Normalizing the Hamiltonian such that its groundstate|Ωi has zero energy,
in which case it is time-independent, we obtain
Z Z Z
hΩ| = dq hΩ| q, ti ht, q| = dq hΩ| qi ht, q| = dq Ω̄(q) ht, q| . (2.49)
Now we multiply (2.48) with Ω̄(q)Ω(q ′ ) and integrate over the arguments q and q ′ . This yields
Z w(t)=q
Z
′ ′
hΩ| T q̂(t1 ) · · · q̂(t1 )|Ωi = dqdq Ω̄(q)Ω(q ) Dw w(t1 ) · · · w(tn ) eiS[w]/h̄. (2.50)
w(0)=q ′
Generating functional for time ordered products: The Greenfunctions for time ordered
products of position operators at different times are generated by a functional depending on an
external source. It is given by the path integral in which a source term is added to the action,
Z t
S[w] −→ Sj [w] = S[w] + (j, w), (j, w) = dsj(s)w(s). (2.51)
0
is just the generating functional for the Greenfunctions (2.48). For example, its first variational
derivative with respect to the source is
h̄ δ Z
K(t, q, q ; j) = Dw w(t1 ) eiSj /h̄ .
′
(2.53)
i δj(t1 )
The n-fold differentiation of K at j = 0 yields the path integral with several w-insertions,
h̄ δ h̄ δ
Z
··· K(t, q, q ′ ; j)|j=0 = Dw w(t1 ) · · · w(tn ) eiS[w]/h̄, (2.54)
i δj(t1 ) i δj(tn )
which according to the result (2.48) is equal to the expectation value of the time-ordered product
of n position operators at different times
h̄ δ h̄ δ
··· K(t, q, q ′ ; j)|j=0 = hq, t| T q̂(t1 ) · · · q̂(tn )|q ′ i . (2.55)
i δj(t1 ) i δj(tn )
For an interacting system the generating functional cannot be calculated in closed form. But
with the result (2.48) we can easily set up a perturbative expansion for the Greenfunctions. This
will be done in chapter 4.
————————————
A. Wipf, Path Integrals
Chapter 3
To get acquainted with path integrals we consider the harmonic oscillator for which the path
integral can be calculated in closed form. We allow for an arbitrary time-dependent oscillator
strength and later include a time dependent external force. We begin with the discretized path
integral (2.29) and then turn to the continuum path integral (2.32).
m n−1
X h1 i
S (n) (w) = (wj+1 − wj )2 − ǫ ωj2 wj2 with ωj = ω(t′ + jǫ). (3.2)
2 j=0 ǫ
For the following manipulations is it convenient to introduce two n − 1-tupels, one with the
integration variables as entries and the other with the positions of the endpoints,
18
CHAPTER 3. THE HARMONIC OSCILLATOR 3.1. Solution by discretization 19
For vanishing ωj the square matrix C is proportional to the discretized second derivative (one-
dimensional lattice Laplacian) on the discrete time lattice. We are left with calculating the
Gaussian integral
Z 1/2
(n) (ξ,η)/h̄ m
Kω (t, q, t′, q ′ ) = lim Anǫ dn−1 ξ eiS , where Aǫ = , (3.6)
n→∞ 2πih̄ǫ
and the lattice action (3.4) is a quadratic function of the integration variables ξ. As a function
of these variables it is extremal at ξcl , given by
δS (n)
(ξ = ξcl ) = 0 or Cξcl = η. (3.7)
δξi
ξcl is the classical solution of the discretized equation of motion. Expanding the action about
this solution yields
m
S (n) (ξcl + ξ) = S (n) (ξcl ) + (ξ, Cξ) (3.8)
2ǫ
with the following action of the classical solution
mh 2 i 1
S (n) (ξcl ) = η − (η, C −1 η) − mω02 ǫq ′2 . (3.9)
2ǫ 2
Terms linear in ξ are absent since ξcl is an extremum of S. Inserting (3.9) into (3.6) leads to
Z
(n) (ξ )/h̄
Kω (t, q, t′ , q ′) = lim Anǫ eiS cl
dn−1 ξ e im/2ǫh̄ (ξ,Cξ) . (3.10)
n→∞
Here we encounter for the first time a Gaussian integral. Such integrals appear frequently in
path integral calculations. The one-dimensional Gaussian integral is
s
Z
−αξ 2 /2 2π
dξ e = . (3.11)
α
The generalization to multi-dimensional Gaussian integrals follows after a diagonalization of
the matrix defining the quadratic form in the exponent and is given by
(2π)p/2
Z
p 1
d ξ exp − (ξ, Bξ) = √ . (3.12)
2 det B
————————————
A. Wipf, Path Integrals
CHAPTER 3. THE HARMONIC OSCILLATOR 3.1. Solution by discretization 20
Here B is a p-dimensional symmetric matrix with non-negative real part. For a non-symmetric
B the antisymmetric part does not contribute to the integral and B is replaced by (B + B † )/2
on the right hand side. For an imaginary B the result (3.12) holds in the distributional sense.
Using this useful formula in (3.10) and performing the continuum limit yields
r
m 1 (n)
Kω (t, q, t′ , q ′ ) = lim √ eiS (ξcl )/h̄ . (3.13)
ǫ→0 2πih̄ ǫ det C
It remains to calculate the determinant of the matrix C and the matrix element (η, C −1η) enter-
ing the classical action in (3.9).
To find the determinant of the n − 1-dimensional matrix C in (3.6) we consider the p-
dimensional matrix
µ1 −1 0 ··· 0
−1 −1 ···
µ2 0
Cp =
.. .. .. ,
µj = 2 − ǫ2 ωj2, (3.14)
. . .
0 ··· −1 µp
and denote its determinant by dp . Expanding the determinant in the last row yields the recursion
relation dp = µp dp−1 − dp−2 with the initial conditions d1 = µ1 and d0 = 1. To solve this
recursion relation we write it in the form
and divide by ǫ2 . Furthermore we set dp = d(sp ), where sp = t′ + pǫ denotes the time after p
time-steps have passed since the initial time t′ . For ǫ → 0 we may approximate differences by
differentials such that the recursion relation turns into the differential equation,
¨ = −ω 2 (s)d(s).
d(s) (3.16)
The initial slope of d diverges in the continuum limit since d2 − d1 = 1 + O(ǫ2 ). Hence we
rescale d(s) → D(s) = ǫ d(s) in order to get a non-singular function. At initial time t′ the
rescaled function vanishes and has unit slope. Hence in the continuum limit we have
ǫ→0
ǫ det C = ǫdn−1 −→ D(t, t′ ), (3.17)
where the D-function solves the Gelfand-Yaglom initial value problem [5]
————————————
A. Wipf, Path Integrals
CHAPTER 3. THE HARMONIC OSCILLATOR 3.1. Solution by discretization 21
Besides the determinant we need the matrix element (η, C −1 η) in the classical action. It
only depends on the elements in the corners of the matrix C −1 . These are given by
cn−2 ··· 1
1
C −1 = . ··· . , (3.19)
dn−1
1 · · · dn−2
The elements on the diagonal are cn−2 = d(t, t′ + ǫ) and dn−2 = d(t − ǫ, t′ ). Expanding in ǫ
the classical action is now seen to depend only on the function D and its time derivatives as the
initial and final time as follows,
!
′ ′
(n) ǫ→0 m 2 dD(t, t ) ′ 2 dD(t, t )
S (wcl ) −→ S[wcl ] = q − q − 2qq ′ . (3.20)
2D(t, t′) dt dt′
Since the solution D of the initial value problem (3.18) determines both the classical action and
the determinantal factor, see (3.17), it determines the exact time evolution kernel
r
m 1
′
Kω (t, q, t , q ) =′
q eiS[wcl ]/h̄ . (3.21)
2πh̄i D(t, t′ )
Differentiating the action of the classical path S[wcl ] with respect to the initial and final position
we recover the D-function,
1 1
∂q ∂q′ S[wcl ] = − . (3.22)
m D(t, t′ )
We see that the classical action determines both the phase factor and the determinantel factor
infront of the phase factor. The evolution kernel of the time-dependent oscillator is completely
determined by the classical action,
s !1/2
1 ∂ 2 S[wcl ]
′ ′
Kω (t, q, t , q ) = − e iS[wcl ]/h̄ . (3.23)
2πh̄i ∂q∂q ′
It is not difficult to see that this kernel satisfies the Schrödinger equation and for t → 0 it
reduces to the free evolution kernel (2.22) and thus to the delta function. Hence it obeys the
————————————
A. Wipf, Path Integrals
CHAPTER 3. THE HARMONIC OSCILLATOR 3.2. Oscillator with external source 22
initial condition (2.20). The kernel Kω (t, q, q ′ ) is singular for ωt = nπ. This apparent problem
can be dealt with by integrating the kernel against wave packets. The Feynman path integral for
Z
ψ(t, q) = dq ′ hq|e−itH/h̄ |q ′ i ψ0 (0, q ′ ), (3.26)
has no singularities.
After this rather involved manipulation let us recapitulate the crucial steps in deriving the
evolution kernels. First we replaced the integration variables ξ by ξcl + ξ, where ξcl has been
an extremum of the classical ’action’. This shift eliminates the linear in ξ terms in the clas-
sical action. Without mentioning, we also assumed the measure to be translational invariant,
dn−1 (ξcl + ξ) = dn−1 ξ, which is of course correct for a finite product of Lebesgue measures.
The resulting Gaussian integral can be calculated and is given in (3.13).
and S denotes the action (3.1) of the oscillator without external force. By considering the forced
oscillator we shall encounter several problems which one comes across in various approxima-
tions to more realistic and complicated systems. In addition, the resulting path integral yields
the generating functional for the Greenfunctions and thus will be of use when we derive the
perturbation expansion for interacting quantum system.
Classical solutions are extremal points of the action and fulfill the equation of motion
δS[w]
− = mẅcl (s) + mω 2 (s)wcl (s) = j(s). (3.29)
δw(s) cl w
Similarly as for the discrete path integral considered in the previous section we expand an
arbitrary path about the classical trajectory,
w(s) −→ wcl (s) + ξ(s), where wcl (t′ ) = q ′ and wcl (t) = q. (3.30)
————————————
A. Wipf, Path Integrals
CHAPTER 3. THE HARMONIC OSCILLATOR 3.2. Oscillator with external source 23
The classical path wcl obeys the boundary conditions such that the fluctuations ξ vanishes at the
endpoints, ξ(t′ ) = ξ(t) = 0. With
Z ξ(t)=0
Z
iSj [w]/h̄ iSj [wcl ]/h̄
′ ′
Kω (t, q, t , q ; j) = Dw e =e Dξ eiS[ξ]/h̄ . (3.32)
ξ(t′ )=0
The path integral factorizes into a classical part depending on the source and the endpoints
and a path integral over the fluctuations. The latter is just the propagator Kω of the force-free
oscillator (3.21) for the propagation from q ′ = 0 to q = 0. For vanishing endpoints the action
S[wcl ] entering Kω in (3.21) is zero and we obtain the simple formula
r
m 1
Kω (t, q, t′, q ′ ; j) = q eiSj [wcl ]/h̄ , (3.33)
2πh̄i D(t, t′ )
After a partial integration in the integral of ẇh ẇcl0 the last two term can be written as
Z Z
t d 0 t
m (w ẇh ) − m wcl0 (ẅh + ω 2 wh ) = 0.
t′ ds cl t′
The first term is zero because wcl0 vanishes at the endpoints and the second term is zero because
wh obeys the homogeneous equation of motion. Thus we obtain
Z
Sj [wcl ] = Sj [wcl0 ] + S[wh ] + ds j(s)wh(s). (3.35)
————————————
A. Wipf, Path Integrals
CHAPTER 3. THE HARMONIC OSCILLATOR 3.2. Oscillator with external source 24
When the source is switches off then the action reduces to the source-independent term S[wh ]
and the propagator reduces to the kernel Kω in (3.21), such that
As Greenfunction of a selfadjoint and real operator GD is symmetric in its arguments and van-
ishes at the endpoints,
Now we can construct the solution wcl0 with the help of this Greensfunction as follows,
Z t
wcl0 (s) = GD (s, s′ )j(s′ )ds′ . (3.40)
t′
Inserting this result into (3.37) yields the following expression for the Schwinger functional,
Z Z
1
Wω [j] = ds j(s)wh(s) + dsds′ j(s)GD (s, s′ )j(s′ ). (3.41)
2
The first term is linear and the second is quadratic in the source. Note that according to (2.55)
and (2.52) the kernel in (3.36) generates all Greenfunctions of time-ordered products of the po-
sition operators at different times. For example, the correlator of two positions for the oscillator
without source is
!
δWω δWω h δ 2 Wω
hq, t| T q̂(t1 )q̂(t2 )|q ′ i = +
Kω (t, q, t′ , q ′ )
δj(t1 ) δj(t2 ) i δj(t1 )δj(t2 ) j=0
!
h̄
= wh (t1 )wh (t2 ) + GD (t1 , t2 ) Kω (t, q, t′ , q ′ ). (3.42)
i
Next we calculate the kernel and in particular the Schwinger functional for the free particle and
for the oscillator with constant frequency.
————————————
A. Wipf, Path Integrals
CHAPTER 3. THE HARMONIC OSCILLATOR 3.2. Oscillator with external source 25
Free particle
For simplicity we take t′ = 0 as initial propagation time of the free particle. The Greenfunction
and homogeneous solution are
1 1
GD (s > s′ ) = (s − t)s′ and wh (s) = [sq + (t − s)q ′ ]. (3.43)
mt t
The quadratic Schwinger functional (3.41) for the free particle has the explicit form
Z Z Z
1 t 1 t s
W0 [j] = ds (sq + (t − s)q ′ )j(s) + ds ds′ (s − t)s′ j(s)j(s′ ) (3.44)
t 0 mt 0 0
Again we take as initial time t′ = 0. For a constant frequency ω the Greenfunction and solution
of the source-free oscillator read
1
GD (s > s′ ) = sin ω(s − t) sin ωs′
mω sin ωt
1
wh (s) = {q sin ωs + q ′ sin ω(t − s)}. (3.47)
sin ωt
Hence the Schwinger function of the oscillator has the explicit form
Z
1 t
Wω [j] = ds (q sin ωs + q ′ sin ω(t − s)q)j(s)
ω sin ωt 0
Z t Z s
1
+ ds ds′ sin ω(s − t) sin ωs′ j(s)j(s′ ), (3.48)
mω sin ωt 0 0
and for a vanishing frequency is converges to the Schwinger functional of the free particle. The
functional Wω enters the formula for the propagator of the oscillator with constant frequency
For vanishing endpoints the evolution kernel for j = 0 on the right hand side simplifies further
and we obtain the simple formula
r ( )
sin ω(s − t) sin ωs′
Z Z
mω i t s
′
Kω (t, 0, 0; j) = exp ds ds j(s)j(s′ ) . (3.50)
2πih̄ sin ωt h̄ 0 0 mω sin ωt
It generates all correlations of time-ordered products of oscillator positions at different times.
————————————
A. Wipf, Path Integrals
CHAPTER 3. THE HARMONIC OSCILLATOR 3.3. Mode expansion 26
Since the map ξ(s) −→ {an } is a unitary map form L2 to ℓ2 the ’measure’ in Dξ is equal to the
Q
’measure’ dan . Inserting the expansion into the exponent in (3.32) we obtain
ξ(t)=0
Z Z Y !1/2
i(ξ,S ′′ ξ)/2h̄ iλn a2n /2h̄
Y 2πih̄
Dξ e = dan e = . (3.55)
λn
ξ(t′ )=0
The product of the eigenvalues λn is the determinant of the fluctuation operator S ′′ and thus the
path integral leads to an inverse square root of the determinant of S ′′ ,
N
Kω (t, q, t′ , q ′ ) = q eiS[wcl ]/h̄ . (3.56)
2 2
det(∂ + ω )
For simplicity we assumed that the external source has been switched off. The divergent nor-
malization factor N can be fixed a posteriori by considering the ratio of two path integrals. This
is sufficient in quantum mechanics where the ratio of two fluctuation determinants is finite. It
is not sufficient in field theory where an additional regularization may be necessary. Before
considering the ratio of determinants we quote a classical result of W EYL [23], according to
which the eigenvalues in (3.52) grow asymptotically as
2
n
|λn | ∼ const · , (3.57)
t − t′
————————————
A. Wipf, Path Integrals
CHAPTER 3. THE HARMONIC OSCILLATOR 3.3. Mode expansion 27
implying that the determinant does not exist. This is not surprising since already in the regular-
ized path integral on the time lattice (3.17) det C ∼ 1/ǫ also tends to infinity in the continuum
limit. The problem with this harmless divergence is resolved as follows: imagine that we repeat
the same steps leading to (3.56) for the free particle instead of the oscillator. We obtain
N
K0 (t, 0, t′ , 0) = q , (3.58)
det(∂ 2 )
since the classical trajectory starting and ending at the origin is just wcl (s) = 0 and hence the
action S[wcl ] in (3.56) vanishes in this case. On the other hand we know from (2.21) that
s
′ m
K0 (t, 0, t , 0) = . (3.59)
2πih̄(t − t′ )
Now we divide the evolution kernel in (3.56) by K0 as in (3.58) and multiply again by K0 as in
(3.59). The unknown constant N chancels in the quotient and we obtain
s !−1/2
m ∂ 2 + ω 2 (.)
′ ′
Kω (t, q, t , q ) = det eiS[wcl ]/h̄ . (3.60)
2πih̄(t − t′ ) ∂2
According to (3.17) the ratios of the determinants are given by the ratios of the D-functions of
the corresponding fluctuation operators. The D-function of ∂ 2 is D(s, t′ ) = s − t′ , such that
r
m 1
′ ′
Kω (t, q, t , q ) = q eiS[wcl ]/h̄ . (3.61)
2πih̄ D(t, t′ )
Alternatively we could divide and multiply (3.56) with the evolution kernel Kω of the oscillator
with constant ω, as given in (3.25). One finds
s !−1/2
mω ∂ 2 + ω 2 (.)
′ ′
Kω (t, q, t , q ) = det eiS[wcl ]/h̄ , (3.62)
2πih̄ sin ω(t − t′ ) ∂2 + ω2
where ω and ω(.) are the constant and time-dependent frequencies. Inserting the D-function
1/ω · sin ω(t − t′ ) of the oscillator with constant frequency again leads to the result (3.61).
————————————
A. Wipf, Path Integrals
Chapter 4
Perturbation Theory
H = H0 + λV (4.1)
is small and expands the eigenvalues and eigenfunctions of H in a power series in λ. Here we
perform an expansion of the evolution kernel in powers of the coupling constant. For H0 one
usually takes the Hamiltonian of the free particle or the harmonic oscillator such that for λ = 0
the problem is soluble. This way one obtains a non-convergent series which (at least in quantum
mechanics) has a good chance of being asymptotic.
The coupling constant λ measures the strength of the interaction. It is introduced for an easy
identification of terms contributing to a given order in the perturbative expansion. In order to
find this expansion for the propagator we use its path integral representation
w(t)=q
Z
′
K(t, q, q ) = Dw eiS[w]/h̄ , (4.3)
w(0)=q ′
where one integrates over all paths with fixed endpoints q ′ and q. Inserting the decomposition
(4.2) one immediately obtains a power series expansion for K in powers of λ. We assume a
28
CHAPTER 4. PERTURBATION THEORY 4.1. Perturbation expansion for the propagator 29
small coupling and expand exp(iSI /h̄) in powers of λ with the result
Z
K(t, q, q ′ ) = Dw eiS0 /h̄ eiSI /h̄
∞
!n Z n
Z
iS0 /h̄
X 1 λ
= Dw e V (w(s))ds . (4.4)
n=0 n! ih̄
The leading term is just the propagator of the free particle (2.22). The sub-leading term of order
O(λ) is given by by the path integral
w(t)=q
λ t
Z Z
K1 (t, q, q ) = ′
ds Dw eiS0 [w]/h̄ V (w(s)), (4.5)
ih̄ 0
w(0)=q ′
where we have interchanged the order of integrations and first did the path integral and then
the time-integration. To calculate the path integral at hand (prior to the s-integration) we first
integrate over all path from the initial position q ′ at time 0 to an intermediate event s, u and
then over all path from (s, u) to the final position q at time t. Finally we integrate over all
intermediate position u,
Z Z w(t)=q
Z w(s)=u
Z
iS0 [w]/h̄ iS0 [w]/h̄
Dw e V (w(s)) = du Dw e V (u) Dw eiS0 [w]/h̄ . (4.6)
w(s)=u w(0)=q ′
The two path integrals are given by the propagator K0 of the free particle (2.22). Hence we
arrive at the following expression for the first order perturbation K1 ,
λ Zt Z∞
K1 (t, q, q ′ ) = ds du K0 (t − s, q, u)V (u)K0 (s, u, q ′). (4.7)
ih̄ 0 −∞
Since K0 (s, u, v) is a Gaussian function of u and v the integral over the intermediate position u
can be calculated explicitly for a polynomial potential. This expression for K1 can be interpreted
as follows: first the particle propagates freely from q ′ to u, where at time s it is ’hit’ by the
potential. Then it again propagates freely to q during the time interval t − s. The total traveling
time being t. Then the amplitudes for all intermediate positions u and times s of possible hits are
summed. One of Feynman’s big achievements was to provide a pictorial representation of the
amplitude by a so-called Feynman diagram. The contribution of order O(λ2 ) to the propagator
reads
!2 Z
′ 1 λ iS0 [w]/h̄
Z t
K2 (t, q, q ) = Dw e dsds′ V (w(s))V (w(s′ )) (4.8)
2 ih̄ 0
!2 Z t Zs
λ
Z
= ds ds′ dudv K0 (t−s, q, u)V (u)K0 (s−s′, u, v)V (v)K0 (s′ , v, q ′),
ih̄
0 0
————————————
A. Wipf, Path Integrals
CHAPTER 4. PERTURBATION THEORY 4.1. Perturbation expansion for the propagator 30
time time
q q
t t
V (u)
V (u) s2
s
s1 V (v)
q′ u space q′ u v space
Figure 4.1: The Feynman graphs associated to first and second order perturbation theory.
which can be visualized as a particle propagating freely from q ′ to v, where at time s′ it is hit
by V , then moving freely to u, where it is hit by V at time s and finally propagates freely to q.
It arrives at the final position at time t. Then the amplitudes for all intermediate positions u and
v and intermediate times s′ and s are superimposed.
The perturbative expansion can easily be calculated with the help of the generating func-
tional for the Greenfunctions of the free particle. According to our result (3.45) this functional
reads
Z
K0 (t, q, q ′ ; j) = Dw eiS0j [w]/h̄ = K0 (t, q, q ′ ) eiW0 [j]/h̄ , (4.9)
where K0 (t, q, q ′) denotes the propagator without source and the Schwinger functional W0 [j]
depends quadratically on the source. Because of
!n Z
h̄ δ iS0j /h̄
Z
Dw e = Dw eiS0j /h̄ w n (s) (4.10)
i δj(s)
we may calculate the path integrals appearing in the perturbative expansion (4.4) as follows,
!Z
h̄ δ
Z
iS0j /h̄
V Dw e = Dw eiS0j /h̄ V (w(s)). (4.11)
i δj(s)
The final expansion for the kernel can be written in the concise form
" !#
λ h̄ δ
Z
′
K(t, q, q ) = K0 (t, q, q ) exp ′
ds V eiW0 [j]/h̄|j=0 . (4.12)
ih̄ i δj(s)
To calculate the moments in (4.10) we define the ’normalized’ n-point correlation functions of
the free theory with action S0 ,
Dw eiS0 /h̄ w(t1 ) · · · w(tn )
R
(n)
G0 (q, q ′ ; t1 , . . . tn ) = . (4.13)
Dw eiS0 /h̄
R
————————————
A. Wipf, Path Integrals
CHAPTER 4. PERTURBATION THEORY 4.1. Perturbation expansion for the propagator 31
In our notation we made the dependence on the end points for the path over which one integrates
explicit. Inserting the result (4.9) for the generating functional the normalized correlation func-
tions take the simple form
!
(n) h̄ δ h̄ δ
G0 (q, q ′ ; t1 , . . . , tn ) = ··· eiW0 [j]/h̄ . (4.14)
i δj(t1 ) i δj(tn ) j=0
(n)
Using the explicit form of W0 [j] in (3.44) the G0 can be calculated explicitly. Actually, since
W0 is a quadratic functional of j they can be expressed in terms of the 1 and 2-point correlation
functions. The formulas expressing the higher n-point functions in terms of the 1 and 2-point
functions is the celebrated Theorem of Wick.
In case q ′ = q = 0 the homogeneous solution wh vanishes for all times and the theorem
takes a much simpler form, since the generating functional simplifies to
∞ n
1 i t
Z
iW0 [j]/h̄
X
′ ′
e = j(s) GD (s, s ) j(s ) . (4.15)
n=0 n! 2h̄ 0
(n)
To simplify our notation we denote the Greenfunctions with q ′ = q = 0 by G0 (0, t1 , . . . , tn ).
(n)
Since the functional contains even powers of j only, the G0 vanish for odd n. The first non-
vanishing correlation function is
(2) h̄
G0 (0, t1 , t2 ) = GD (t1 , t2 ). (4.16)
i
For general even n the Greenfunction is given by a sum of products of the two-point function,
(2n) X (2) (2)
G0 (0, t1 , . . . , tn ) = G0 (0, ti1 , ti2 ) · · · G0 (0, ti2n−1 , ti2n ), (4.17)
pairs (i1 i2 )···(i2n−1 i2n )
where two indices in the sum are unequal and the pairs are ordered. This is the Wick theorem
found in most text books and it holds for all theories with quadratic actions. For example, the
4-point function contains 3 terms
(4) (2) (2)
G0 (0, t1 , . . . , t4 ) = G0 (0, t1 , t2 )G0 (0, t3 , t4 )
(2) (2)
+ G0 (0, t1 , t3 )G0 (0, t2 , t4 ) (4.18)
(2) (2)
+ G0 (0, t1 , t4 )G0 (0, t2 , t3 ).
For all theories with quadratic action the generating functional W [j] is quadratic in j and the
truncated or connected correlation functions
n
!
i Y h̄ δ
G(n) ′
c (q, q ; t1 , . . . , tn ) = W [j]|j=0 (4.19)
h̄ k=1 i δj(tk )
vanish for n > 2. This simple observation then just proves the theorem of Wick.
————————————
A. Wipf, Path Integrals
CHAPTER 4. PERTURBATION THEORY 4.2. Quartic potentials 32
where the 2-point function G(2) = h̄i GD and the homogeneous solution wh for the free particle
have been calculated earlier in (3.43),
h̄ 1
G(2) (s, s) = (s − t)s and wh (s) = [sq ′ + (t − s)q]. (4.23)
imt t
To compute K1 we just need to integrate the fourth order polynomial in (4.22) which results in
ih̄ t3 3 t2 2
(
2 4
K1 (t, q, q ′ ) = λ K0 (t, q, q ′ ) 2
+ (q + q ′ + qq ′ )
m 10 m 10 3
it 4
3 ′ 2 ′2 ′3 ′4
− (q + q q + q q + qq + q ) . (4.24)
h̄ 5
We can trust the perturbative expansion if K1 ≪ K0 , which is the case if
n m2 m h̄ o
λ ≪ max , , .
h̄t3 t2 q 2 tq 4
The expansions becomes reliable for short propagation times t and small q ′ and q. It breaks
down for small particle masses. According to Wicks theorem the higher order contributions in
the perturbative series (4.4) reduce to integrals of products of 1 and 2-point functions of the free
particle. Hence they can be calculated in closed form. However, the number of terms one must
include grows rapidly with increasing order n.
The perturbative expansion for the Greenfunction hq, t| Tq̂(t1 ) · · · q̂(tn )|qi is obtained sim-
ilarly as for the evolution kernel. Again we assume that S is the sum of a free part S0 and an
————————————
A. Wipf, Path Integrals
CHAPTER 4. PERTURBATION THEORY 4.2. Quartic potentials 33
interaction term SI , see (4.2). Now we expand the right hand side of (2.48) in powers of the
coupling constant λ. This leads to the expansion
The matrix elements on the right hand side are to be evaluated for the system without interaction
which means for the system with action S0 . Formally this series can be summarized as follows
with Schwinger function W0 [j] for the non-interacting system, see (3.44). Since W0 is quadratic
in the source j we may use Wick’s theorem to calculate the perturbative expansion on the right
hand side.
————————————
A. Wipf, Path Integrals
Chapter 5
In this section study the dynamics of a charged particle in a given external electromagnetic field.
In reality the field is modified by a moving charge, for example by the radiation emitted by the
particle. But here we shall neglect this backreaction. This is a reasonable approximation for
strong or/and almost constant fields.
If ψ fulfills the time-dependent Schrödinger equation with potentials ϕ and A then the gauge-
transformed wave function fulfills the Schrödinger equation with gauge-transformed potentials.
According to the general rules we expect that the path integral representation for the propagation
of a charged particle from (t′ , x ′ ) to (t, x ) in an electromagnetic field is given by
t m 2 e
Z Z
′ ′ iS[w ,A]/h̄
K(t, x , t , x ) = Dw e , S= ds ẇ + ẇ · A − eϕ , (5.8)
t′ 2 c
where the values of the potentials along the particle path enter, for example ϕ = ϕ(t, w (t)). To
prove that this propagator satisfies the time dependent Schrödinger equation we proceed simi-
larly as in section 2.3 and replace the time-integral (5.8) by a Riemann sum. In the discretisation
of the integral ds ẇ · A we must choose the midpoint rule,
R
n−1
wj+1 − wj sj+1 + sj wj+1 + wj
Z X
ds ẇ(s) · A(s, w (s)) −→ ǫ ·A , (5.9)
j=0 ǫ 2 2
with wj = w (sj ). This corresponds to the socalled Ito-calculus in the theory of stochastic
differential equations. If we would take the potential at wj instead of the midpoint between wj
and wj+1 then we would obtain a gauge non-invariant propagator.
Now we take a wave function at time t − ǫ and let it be propagated toward t. If u = x − y
denotes the difference between the final and initial position then we obtain up to terms of O(ǫ2 )
im 2 iǫ
Z
ψ(t, x ) ≈ lim A3ǫ d3 u exp
u exp Lint ψ(t − ǫ, x − u)
ǫ→0 2h̄ǫ h̄
eu ǫ u ǫ u
Lint = · A t − ,x − − eϕ t − , x − , (5.10)
cǫ 2 2 2 2
As earlier Aǫ = (m/2πih̄ǫ)1/2 enters as normalizing factor. Expanding the two last factors in
the first line up to terms linear in ǫ or quadratic in u. We obtain
im 2 1 ieǫ
Z
ψ(t, x ) = lim A3ǫ 3
d u exp u ψ(t − ǫ) + ui uj Di Dj ψ − ϕψ + . . . , (5.11)
ǫ→0 2h̄ǫ 2 h̄
————————————
A. Wipf, Path Integrals
CHAPTER 5. PARTICLES IN E AND B FIELDS 5.1. Charged scalar particle 36
∂ h̄2
ih̄ ψ(t, x ) = − (D 2 ψ)(t, x ) + eϕ(t, x )ψ(t, x ), (5.14)
∂t 2m
which is just the Schrödinger equation (5.6) in the position representation. It is a useful exercise
to show that if we do not take the midpoint rule in (5.9) then we would get a different result.
Actually for the scalar potential and for the time-integration no midpoint rule is needed. We
would still get the correct propagator in the continuum limit if we would take
eu u
Lint = · A t, x − − eϕ (t, x ) , (5.15)
cǫ 2
instead of Lint in (5.10). But with the choice (5.10) the convergence to the continuum limit is
faster. Under a gauge transformation (5.3) with gauge function λ(t, x ) the action changes by
path independent boundary terms,
!
eZ t ∂ e
∆S[w , A, ϕ] = − ds ẇ · ∇λ + λ = − {λ(t, x ) − λ(t′ , x ′ )} (5.16)
c t′ ∂s c
such that the propagator transforms covariantly under gauge transformations,
′ ′ )/h̄c
K(t, x ; t′ , x ′ ) −→ e−ieλ(t,x )/h̄c K(t, x , t′ , x ′ ) eieλ(t ,x . (5.17)
This agrees with the transformation rule (5.7) for the solutions of the Schrödinger equation
under gauge transformations.
————————————
A. Wipf, Path Integrals
CHAPTER 5. PARTICLES IN E AND B FIELDS 5.1. Charged scalar particle 37
Here we consider the Aharonov-Bohm effect due to a magnetic flux Φ confined to a solenoid.
We assume that the solenoid is straight and very long and choose the coordinate system such
that the z-axis is the symmetry axis of the solenoid. Outside the solenoid there in no magnetic
field and for an infinitely long solenoid the magnetic potential has the form
Φ xdy − ydx
A · dx = , ρ2 = x2 + y 2. (5.18)
2π ρ2
We assume that the particle can not penetrate into the solenoid. Let us consider a particle
trajectory w (s) defining a curve C. The term containing the magnetic vector potential in the
action (5.8) is proportional to
t dw (s) Φ xdy − ydx
Z Z Z
A(w (s)) · ds = A(x ) · dx = . (5.19)
t′ ds C 2π C ρ2
Transforming to cylinder coordinates (x, y, z) = (ρ cos ϕ, ρ sin ϕ, z) the line integral becomes
Φ
Z Z
A · dx = dϕ. (5.20)
C 2π C
A path Cn : x ′ → x outside the solenoid is characterized by its winding number n ∈ Z. For its
definition one takes some standard contour C0 : x ′ → x and counts the number of times that
the closed curve Cn − C0 winds around the solenoid. In figure 5.1 we have depicted a reference
x b
C1
∆ϕ C0
solenoid
x′ b
path C0 and a path C1 with winding number one. For a path with winding n one has
Φ
Z Z
A · dx = nΦ + A · dx = nΦ + ∆φ, (5.21)
Cn C0 2π
————————————
A. Wipf, Path Integrals
CHAPTER 5. PARTICLES IN E AND B FIELDS 5.2. Spinning particles 38
where ∆Φ is the angle shown in figure 5.1. In the path integral one admits all paths connecting
x ′ with x . We do the integration in two steps: first we integrate over the set paths {Cn } with
winding number n and then sum over all winding numbers. This yields
XZ
K(t, x , x ′ ) = Dw eiS[w ,A]/h̄ = eieΦ∆φ/hc eineΦ/h̄c Kn (t, x , x ′ ),
X
(5.22)
n {Cn } n
————————————
A. Wipf, Path Integrals
CHAPTER 5. PARTICLES IN E AND B FIELDS 5.2. Spinning particles 39
Here σ = (σ1 , σ2 , σ3 ) is the 3-tuple of Pauli matrices. The Pauli-Hamiltonian contains a cou-
pling of the electron spin to a magnetic field with the correct g-factor of 2. Indeed, with the help
of σi σj = iǫikj σk + 12 the Pauli-Hamiltonian can be rewritten as
1 e 2 e h̄
H= p − A(t, x ) + eϕ(t, x ) − B(t, x ) · s, s= σ, (5.27)
2m c mc 2
where the two first terms act as identity operator in spin space. The corresponding matrix-valued
Lagrange function
m 2 e e
L= ẋ + ẋ · A(t, x ) − eϕ(t, x ) + B(t, x ) · s (5.28)
2 c mc
should enter the path integral for a non-relativistic spin-1/2 particle. Although L is matrix
valued we could proceed as in the previous section and would end up with the result (5.10) with
interaction Lagrangian
eu e
Lint (t, x , u) = · A − eϕ + B ·s . (5.29)
cǫ mc midpoint
If the propagation is from (t − ǫ, x − u) → (t, x ) as it is in (5.10), then the midpoint rule means
evaluation of the potentials and magnetic field at time t − 12 ǫ and position x − 21 u. This way
one obtains for the propagator the representation
Z
′ ′
K(t, x , t , x ) = lim A3n d3 w1 · · · d3 wn−1 eiǫLn−1 /h̄ · · · eiǫL0 /h̄ ,
n→∞ ǫ
m uj2 e uj e
Lj = + · A(s̄ j , w̄ j ) − eϕ(s̄ j , w̄ j ) + B(s̄j , w̄j ) · s, (5.30)
2 ǫ2 c ǫ mc
where w0 = x ′ , wn = x and we have used the abbreviations
wj+1 + wj sj+1 + sj
uj = wj+1 − wj , w̄j = , s̄j = . (5.31)
2 2
As earlier the propagation time interval [t′ , t] is divided into n intervals of length ǫ = (t − t′ )/n
and sj = t′ + jǫ. For a time and/or space dependent magnetic field two Lj in (5.30) with
different j do not commute due to the B · s-term in the Lagrangian. In the (formal) continuum
limit we identify wj with the position w (sj ) of the particle at time sj . Then Lj is the value of
the Lagrangian at time sj . We see that the factors in (5.30) are time ordered: on the right we
have the factor exp(iǫL0 /h̄) at earliest time and on the left the factor exp(iǫLn−1 /h̄) at latest
time. Thus we are lead to the path ordered integral
iZt
Z
′ ′
K(t, x , t , x ) = Dw P exp ds L(s) ,
h̄ t′
L(s) = L w (s), A(s, w (s)), ϕ(s, w (s)) , (5.32)
————————————
A. Wipf, Path Integrals
CHAPTER 5. PARTICLES IN E AND B FIELDS 5.2. Spinning particles 40
where the time is ordered along the path w (s). The path ordered integral satisfies the differential
equation
∂ i t i i t
Z Z
P exp ds L(s) = L(t) P exp ds L(s) , (5.33)
∂t h̄ t′ h̄ h̄ t′
with cyclotron frequency ω = eB/mc. The particle moves freely in the z-direction and only
the propagation in the xy-plane is affected by the external field. Thus we may assume that x ′
and x are both in the plane with z = 0 such that the whole tracetory w (s) lies in this plane.
Without loss of information we may study the two-dimensional dynamics in the xy-plane and
in the following we assume that all vectors lie in the plane, for example w = wx ex + wy ey .
For a uniform magnetic field the spin-term does not depend on the trajectory and hence does
not enter the equation of motion. With the help of the rotation matrix
!
cos ωt sin ωt
R(ωt) = (5.37)
− sin ωt cos ωt
sin(ω̂s) ω
wcl (s) = x ′ + R (ω̂(s − t)) (x − x ′ ), ω̂ = , (5.38)
sin(ω̂t) 2
————————————
A. Wipf, Path Integrals
CHAPTER 5. PARTICLES IN E AND B FIELDS 5.2. Spinning particles 41
The kinetic energy and the term containing the orbital angular momentum diverge if the propa-
gation time is a multiple of 2π/ω. Both contributions to the action contain a term proportional
to t/ sin2 (ω̂t) and since they have different signs they cancel in the sum.
As earlier we decompose an arbitrary path as w (s) = wcl (s) + ξ(s), where the fluctuations
ξ vanish at initial and final time. With
m d2 d
S[w ] = S[wcl ] + (ξ, Mξ), M = − 2 + iωσ2 , (5.40)
2 ds ds
the path integral yields
N
K(t, x , x ′ ) = √ eiS[wcl ]/h̄ . (5.41)
det M
We remain with calculating the determinant of the matrix differential operator M. This can be
achieved by a generalization of the Gelfand-Yaglom initial value problem. One defines a matrix
S, the columns of which are linearly independent solutions of Mξ = 0 vanishing at s = 0,
MS = 0 with S(0) = 0, Ṡ(0) = 1. (5.42)
Any solution of Mξ = 0, ξ(0) = 0 is a linear combination of the columns of S. Let us now
assume that
det S(t) = 0. (5.43)
Then there is a linear combination of the columns of S which vanish at the final time t. It
is an eigenfunction of the fluctuation operator with zero energy such that det M must vanish.
Since the converse statement is also true, it is not surprising that the ratio of two fluctuation
determinants is given by
det M det S 1
= = 2 det S. (5.44)
det M0 det S0 t
Here S0 = t1 is the matrix of solutions of the fluctuation operator M0 with vanishing ω. In
particular for the fluctuation operator in (5.40) we have
S(t) = ω̂ −1 sin ω̂t (cos ω̂t + i sin ω̂t σ2 ) (5.45)
and this leads to the following ratio of determinants:
!2
det M sin ω̂t
= . (5.46)
det M0 ω̂t
Inserting this result into (5.41) yields the well known propagator for a spinning particle in a
uniform magnetic field
m 3/2 ω̂t im
′
K(t, x , x ) = exp (z − z ′ )2 + iω̂σ3
(
2πih̄t sin ω̂t 2h̄t !)
imω̂ cot ω̂t h i
× exp (x − x′ )2 + (y − y ′ )2 + (x′ y − xy ′ ) . (5.47)
h̄ 2
————————————
A. Wipf, Path Integrals
CHAPTER 5. PARTICLES IN E AND B FIELDS 5.2. Spinning particles 42
To obtain the propagator in 3 dimensions we have multiplied with the propagator for the free
motion in the z-direction. Similarly as for the harmonic oscillator the propagator is singular
at times tn = 2πn/ω after which a classical particle returns to its starting point in the plane
orthogonal to the B-field. Note that the two spin-components acquire different phases in a non-
vanishing magnetic field. The above result (without spin-term) has been obtained by G LASSER
[20] and by F EYNMAN and H IBBS [4].
————————————
A. Wipf, Path Integrals
Chapter 6
The oscillatory nature of the integrand eiS/h̄ in the path integral gives rise to distributions. If
the oscillations were suppressed, then it might be possible to define a sensible measure on the
set of paths. With this hope much of the rigorous work on path integrals deals with imaginary
time t → −iτ for which the Lagrangian density undergoes the so-called Wick rotation. One
analytically continues to imaginary times, calculates the corresponding Greensfunctions and
continuous back to real time by the inverse Wick rotation τ → it. For imaginary time the
measure on the set of paths can rigorously be defined and leads to the Wiener measure.
where the integral has its support on the spectrum of H. Here PE is the orthogonal projector
onto the subspace spanned by the eigenfunction with energy less or equal to E. We assume that
the Hamiltonian is bounded from below and add a constant to it, such that it has non-negative
spectrum. Then the above integral extends from 0 to ∞. Now we assume that t becomes
complex t → t − iτ ,
Z ∞
−(τ +it)H
e = e−E(τ +it)H dPE . (6.2)
0
{t − iτ ∈ C, τ ≥ 0}. (6.3)
If we would know the operator (6.2) for imaginary times (t = 0, τ ≥ 0) then we could analyt-
ically continue it to real times (t, τ = 0), the time domain of interest in quantum mechanics.
43
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.1. Quantum Mechanics for Imaginary Times 44
If we analytically continue t to −iτ then the Minkowskian metric η = diag(1, −1, −1, −1)
turns into a metric with Euclidean signature. This explains why this continuation is called the
transition from the Lorentzian- to the Euclidean sector of the theory.
The evolution operators U(t) are defined for all real times and form a one-parametric uni-
tary group. It satisfies a Schrödinger equation
d
i U(t) = HU(t), (6.4)
dt
and its kernel K(t, q ′ , q) = hq| U(t)|q ′ i is complex and of oscillatory character. For imaginary
times the ’evolution operators’
are positive (and hence hermitean instead of unitary) with non-negative eigenvalues. The U(τ )
exist for positive τ and form a so-called a semigroup. For almost all initial state vectors a
propagation backwards in imaginary time is impossible. U(τ ) satisfies a diffusion type equation,
d
U(τ ) = −HU(τ ), (6.6)
dτ
and its kernel
is real for real Hamiltonians1. This kernel is strictly positive as follows from the
Theorem: Suppose V is continuous and bounded from below, and let H = H0 + V be
essentially selfadjoint. Then the evolution kernel is positive,
If a (real) kernel would be negative for q ′ in an open set O then we could construct a function
ψ(q) ≥ 0 with support in O for which (ψ, U(τ )ψ) ≤ 0. This should not be possible for the
positive operator U(τ ). For a rigorous proof of (6.8) I refer to the textbook of Glimm and Jaffe
[9], page 50. The positive kernel for a free particle in d dimensions reads
m d/2 −m(q′ −q)2 /2τ
K0 (τ, q, q ′) = e , (6.9)
2πτ
and the kernel for an oscillator with constant frequency ω is given by Mehlers formula
mω d/2 n mω 2 2qq ′ o
Kω (τ, q, q ′) = exp − [(q + q ′2 ) coth ωτ − ] . (6.10)
2π sinh ωτ 2 sinh ωτ
1
If we couple charged particles to a magnetic field, then the Hamiltonian ceases to be real.
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.1. Quantum Mechanics for Imaginary Times 45
The strict positivity of these kernels is manifest. This positivity allows for a deep connection of
Euclidean quantum mechanics (and field theory) to probability theory, see below. The object
maybe interpreted as probability density for a particle starting at 0 to show up at q after a time
interval τ . The probability to end up anywhere should be one,
Z Z
C· dq hq, τ | 0, 0i = C · dqK(τ, q, 0) = 1, (6.12)
and this fixes the constant C. In particular, for the free particle
m d/2 −mq2 /2τ
P0 (τ, q) = e . (6.13)
2πτ
This is the probability density for the Brownian motion with diffusion coefficient D = 1/2m. It
fulfils the same diffusion equation (6.6) as U(τ ), in the present context called master equation.
Wightman functions
Of great interest in a relativistic quantum field theory are the Wightman functions. These are
expectation values of product of field operators in the vacuum state. In quantum mechanics
these are the expectation values
W (n) (t1 , . . . , tn ) = hΩ| q̂(t1 ) · · · q̂(tn )|Ωi , q̂(t) = eitĤ q̂ e−itĤ , (6.14)
where|Ωi denotes the ground state. We subtracted the groundstate energy from H such that|Ωi
has zero energy and H becomes a non-negative operator. The W (n) are not symmetric in their
arguments, since the position operators at different times do not commute. We may analytically
continue the W (n) to zi → ti − iτi ,
W (n) (z1 , . . . , zn ) = hΩ| q̂e−i(z1 −z2 )H q̂e−i(z2 −z3 )H q̂ · · · q̂ e−i(zn−1 −zn )H q̂|Ωi , (6.15)
provided the imaginary parts of the complex time-differences obey the inequalities
ℑ(zk − zk+1 ) ≤ 0.
In equation (6.15) we used that H annihilates the groundstate, exp(iζH)|Ωi = 0. From the
definition of the zi it follows that the W (n) are analytic in
The Wightman functions for real times may be reconstructed as boundary values of the analytic
Wightman functions with complex arguments,
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.2. The Euclidean Path Integral 46
For purely imaginary times the Wightman functions are called Schwinger functions,
Let us calculate the 2-point Wightman- and Schwinger functions of the harmonic oscillator. The
zero-point energy subtracted Hamiltonian reads
H = ωa† a,
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.3. Semiclassical Approximation 47
The kernels for the free particle and the harmonic oscillator are given in (6.9) and (6.10).
Here α replaces 1/h̄ in the path integral. We wish to find a good approximation to this integral in
the limit α → 0. Let us assume that the function S in the exponent possesses a minimum at wcl .
Then the main contribution to the integral comes from points near wcl . Expanding S(wcl + ξ)
about this minimum leads to
Z
1 ′′ (w 2 − αP (w
−αS(wcl )
K=e dξ e− 2 αS cl )ξ cl ,ξ)
, (6.25)
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.3. Semiclassical Approximation 48
where P = o(ξ 2) contains all higher order terms in the Taylor-expansion of S about its mini-
mum. Let us estimate the relative error we make when we approximate K by
n
αk
Z
1 ′′ (w 2
Kn = e−αS(wcl ) dξ e− 2 αS cl )ξ
(−)k P (wcl , ξ)k .
X
(6.26)
k=0 k!
We estimate the error relative to the quadratic approximation K0 ,
n k
( )
K − Kn kα
Z
−αP (ξ) k
X
∆n K = ≡ dµα (ξ) e − (−) P (wcl , ξ) , (6.27)
K0 k=0 k!
where we have introduced the probability measure defined by the quadratic term in the Taylor-
expansion of the function S in the exponent,
s
1 −αS ′′ (wcl )ξ2 /2 αS ′′ −α S ′′ ξ2 /2
dµα (ξ) = e dξ = e dξ. (6.28)
K0 2π
On the right hand side we abbreviated S ′′ = S ′′ (wcl ). The measure satisfies the scaling relation
√
dµα(ξ) = dµ1 ( αξ). (6.29)
We recall from the theory of Taylor series expansion, that for a C n+1 -function f (α) one has
n
αk f (n+1) (η) n+1
f (k) (0)
X
f (α) − = α (6.30)
k=0
k! (n + 1)!
for some value η in the interval [0, α]. Applied to the expression between the curly brackets in
(6.27) we conclude that
(−α)n+1 n+1 −ηP
{...} = P e , with 0 ≤ η ≤ α. (6.31)
(n + 1)!
Without loss of generality we may assume that P is non-negative such that exp(−ηP ) is less or
equal to one. Thus we can bound the relative error from above (6.27) as follows
αn
Z
|∆n−1 K| ≤ dµα (ξ)P n(wcl , ξ). (6.32)
n!
To extract the α-dependence of this bound we use the scaling relation (6.29) which implies
αn √
Z
|∆n−1 K| ≤ dµ1 (u) P n wcl , u/ α . (6.33)
n!
Recall that P (wcl , ξ) contains the higher order terms in the Taylor expansion of S and is of order
O(ξ 3). From this we conclude that the relative error in (6.33) tends to zero for large values of
the parameter α. For example, for a quartic function P (ξ) = λξ 4 the upper bound reads
!n Z
1 λ
|∆n−1 K| ≤ dµ1 (u) u4n (6.34)
n! α
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.3. Semiclassical Approximation 49
Note that for a non-convergent asymptotic series α(n) depends on n, contrary to the situation for
a convergent series. Asymptotic series is the best we can hope for in a perturbative expansion
of the propagator2 . For the quartic function S = w 2/2 + λw 4 with wcl = 0 the exact integral
(6.24) for α = 1 is given by a modified Bessel function,
√
K = 2 κ eκ K1/4 (κ), κ = 1/32λ. (6.37)
The saddle point approximations of order 3 is given by the polynomial
√
K3 = 2π 1 − 3λ + 105λ2 /2 − 3465λ3/2 (6.38)
The exact result together with the approximations K1 , K2 , K3 are depicted in figure 6.1.
2
there is a general argument due to Dyson which shows that perturbative expansion cannot be analytic at the
point where the coupling constant vanishes.
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.3. Semiclassical Approximation 50
K, Km
√
2π
m=2
exact
m=1
m=3
λ
0 0.05
Figure 6.1: The integral K for a quartic function and its lowest saddle point approximations.
An arbitrary path w(σ) is the sum of the classical path and a fluctuation ξ(σ) and the Euclidean
action has the expansion
1
SE [wcl + ξ] = SE [wcl ] + (ξ, SE′′ (wcl )ξ) + P [wcl , ξ], (6.41)
2
where P contains all terms of cubic or higher orders
1 Z τ ′′′
P [wcl , ξ] = V (wcl (σ))ξ 3 (σ) + · · · , (6.42)
3! 0
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.3. Semiclassical Approximation 51
d2
SE′′ (wcl ) = −m + V ′′ (wcl (σ)). (6.43)
dσ 2
Inserting these results into (6.23) yields
m
s Z
′
K(τ, q, q ) = e−SE [wcl ]/h̄ dµh̄ (ξ) e−P [wcl,ξ]/h̄ , (6.44)
2πh̄D(τ )
where one integrates over all fluctuating paths starting at the origin and returning to this point
after a time τ and the D-function in the first factor belongs to the fluctuation operator S ′′ . As
for the ordinary integrals we have introduced the Gaussian probability measure
1 −(ξ,S ′′ ξ)/2h̄
Z
′′
dµh̄ (ξ) = e E Dξ with K0 = Dξ e−(ξ,SE ξ)/2h̄ . (6.45)
K0
The relative errors are bounded by
n−1
( )
Z 1 1
Z
−P/h̄ k
dµh̄ P n .
X
|∆n−1 K| = dµh̄ e − ( − P/h̄) ≤ (6.46)
n
k! n! h̄
k=0
Note that the measure dµh̄ possesses the scaling property (6.29) with α replaced by 1/h̄. Using
this property for the quartic function P = λ ξ 4 (σ)dσ one obtains
R
λn h̄n
Z Z n
|∆n−1 K| ≤ dµ1 (ξ) dσ ξ 4(σ) . (6.47)
n!
Since the path integral is calculated with the Gaussian measure with dµ1 it is independent of h̄.
This shows that the relative error ∆n−1 K is of the order O(h̄n ) for a quartic P .
The first correction to the classical contribution exp(−S[wcl ]h̄) coming from the quadratic
fluctuation is called the semiclassical correction or one-loop correction. According to (6.44) the
one-loop approximation to the evolution kernel consisting of the classical part and the one-loop
correction is
m
s
′
K0 (τ, q, q ) = e−SE [wcl ]/h̄ . (6.48)
2πh̄D(τ )
The higher order terms can be computed systematically by calculating the corresponding mo-
ments of the Gaussian measure defined by the fluctuation operator.
To determine the function D appearing in the above expansions we first solve the classical
equation of motion with given initial position q ′ and initial momentum p′ . Next we differentiate
the equation for the classical solution wcl with respect to the initial momentum and obtain
d2 dwcl dwcl
m 2 ′
= V ′′ (wcl ) ′ . (6.49)
dσ dp dp
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.4. Functional Determinants 52
For very short times the particles moves freely such that wcl (σ) ∼ q ′ + p′ σ/m + O(σ 2 ) for small
σ. We conclude that
dwcl d dwcl
m ′ (0) = 0 and m (0) = 1. (6.50)
dp dσ dp′
Since m dwcl /dp′ obeys the same differential equation and the same initial condition as the D-
function the two functions must be identical. It follows that the 1-loop approximation to the
kernel can be written as
!−1/2
dwcl (τ )
K(τ, q, q ′ ) = 2πh̄ e−SE [wcl ]/h̄ . (6.51)
dp(0)
This means that the correction to the classical result is given by the spread in the final position
of the classical particle (area where a projective may land) in terms of the spread in the initial
momentum (spread in the angle of projection). Thus, for artillery computation at least, it is an
important quantity!
The result can further be simplified by noting that the variation of the action when one varies
the end points is
Z
′ ′
δS = pq − p q + (−mẅ + V ′ (w)) δw. (6.52)
Here U0 is either zero or a positive constant. The first determinant appears in ordinary quantum
mechanics and the second in Euclidean quantum mechanics. Such determinant show up in many
applications of path integrals and it is worth studying these objects in detail.
If λn and λ0n are the eigenvalues of M and M0 , respectively, then both the numerator and
denominator in
det M λn ∞
Q
=Q 0 =
det M0 λn ∞
are not well defined. Hence we consider the determinant of the ratio of these two operators and
define
M Y λn
det := , (6.56)
M0 λ0n
which in quantum mechanics at least leads to a finite and well-defined result. For example, for
the harmonic oscillator with frequency ω, the eigenvalues of the fluctuation operator on the time
interval [0, t] or [0, τ ] are
2 2
nπ nπ
2
λn (ω) = − +ω or λn (ω) = + ω2 (6.57)
t τ
so that the infinite products (6.56), in which we skip the divergent factor with n = 0, are
λn (ω) Y ωt 2 sin(ωt)
Y
real time: = 1− =
n6=0 λn (0) nπ ωt
λn (ω) Y ωτ 2 sinh(ωτ )
Y
imag. time: = 1+ = . (6.58)
n6=0 λn (0) nπ ωτ
We defined the determinant of the operator M/M0 as the product of the ratios of the eigenvalues.
These seems to be reasonable for commuting operators M and M0 which can be diagonalized
simultaneously. For non-commuting operators it needs some further justification and this will
be given below. In the remainder of this this section the explicit calculations are done for real
time. The corresponding results for imaginary time are gotten by an analytic continuation.
interpolating between the operator without potential M0 and the operator of interest M = M1 .
According to the Feynman-Hellman formula the variation of an eigenvalue λ(α) is given by
d
λn (α) = (ψn (α), Uψn (α)), (6.60)
dα
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.4. Functional Determinants 54
Integrating these inequality with respect to α from 0 to 1 yields the following two inequalities
for the eigenvalues λn of M and the eigenvalues λ0n of M0 :
For example, if we choose for M0 the operator ∂ 2 with eigenvalues λ0n = −(nπ/t)2 then these
inequalities yield the following upper and lower bounds for the infinite product,
1 Y λn 1
sin(Ωt) ≤ ≤ sinh(Ωt), (6.63)
Ωt λ0n Ωt
where one of the inequalities becomes an equality for constant potentials. In particular we see
that the infinite products are always convergent. Actually, one can prove a stronger statement,
namely that for any square-integrable potential
1 t
Z
λn − λ0n = ds U(s) + rn
t t′
holds true, where the sum of the rn2 exists. This inequality guarantees that the infinite product
in (6.56) converges.
From infinite products to determinants and traces: Now we wish to show that
Y λn
−1
= det 1 + UM0 (6.64)
λ0n
can be defined in terms of the spectrum of the operator UM0−1 alone and that the definition
agrees with the previous one. This statement is not empty, since for non-commuting M and M0
their eigenfunctions may be rather different.
The operator A = UM0−1 of interest is a socalled trace class operator. These are operators
with the property that tr |A| is finite, where |A| is the positive square root of A† A. Trace-class
operators are compact and possess a discrete spectrum {µ1 ≥ µ2 , . . .} and a representation
n−1
X
Aφn = µn φn + αnm φm , (φn , φm ) = δmn . (6.65)
m=1
In the adapted basis {φn } a compact operator is represented by a triangular matrix with eigen-
values on the diagonal. Now one defines the determinant of 1 + A similarly as for matrices as
the product of the eigenvalues
Y
det(1 + A) = (1 + µn ). (6.66)
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.4. Functional Determinants 55
It follows at once that the determinant of a trace-class operator is always finite and can easily
be bounded above as follows
X
|µn | ≤ etr |A| .
Y
det(1 + A) = (1 + µn ) ≤ exp (6.67)
We see that for a trace-class operator the infinite product (6.66) is always absolute convergent3 .
From the definition of the determinant and the spectral decomposition (6.65) it is clear that
This formula will play an important role when we discuss the quantization of fermions in exter-
nal fields or the one-loop effective actions for bosons.
Next we determine the variation of det(1 + A) if A is slightly perturbed by a trace class op-
erator ǫB. Without loss of generality we may assume that the perturbed operator is sufficiently
small so that we may expand as follows
which holds true for two trace-class operators A and B. To prove this product rule we multiply
A with a deformation parameter α and show that the α-derivatives of the logarithm of both sides
in (6.70) are the same.
After these general remarks let us now prove the identity (6.64). Again we introduce a
deformation parameter such that M = M0 + αU and compute the α-derivative of the logarithm
of the determinant appearing on the right hand side of (6.64). Using the identity (6.69) with
3
µ2n , µnk−1
P P P
More generally one can prove the following useful theorem: If for any k > 0 the series µn ,
and |µn |k are convergent, then the infinite product (1 + µn ) is convergent as well.
P Q
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.4. Functional Determinants 56
A = UM0−1 and the cyclicity of the trace (which holds for a trace-class and a bounded operator)
one sees at once that
d
log det 1 + α UM0−1 = tr UM −1 . (6.71)
dα
On the other hand, using again the Feynman-Hellman formula one can compute the variation of
the logarithm of the infinite product in (6.64) directly as follows:
Comparing (6.71) with (6.72) we see that both sides of (6.64) possess the same α-derivative.
Since they are equal for α = 0 they must be equal for α = 1 and this proves (6.64) as required.
In terms of these fundamental solutions the Dirichlet Greenfunction for M reads for s < s′ :
C(t)
GD (s, s′ ) = D(s′ )D(s) − C(s′ )D(s). (6.76)
D(t)
With this explicit expression for the Green function we can now calculate the trace in (6.71),
t C(t) t t
Z Z Z
tr UM −1 = GD (s, s)U(s) = D 2 (s)U(s) − C(s)D(s)U(s). (6.77)
t′ D(t) t′ t′
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.4. Functional Determinants 57
It depends on the initial time t′ as well, but here we are interested in its dependence on the final
time t and on α. By using (6.73) together with (6.75) one finds that W obeys the differential
equation
where the fluctuation operator M acts on s. On the other hand, when differentiating (6.74) with
respect to α (recall that M = M0 + αU) on obtains
which shows that W (t) and ∂α D(t) are solutions of the same differential equation. Finally
it follows from the initial conditions (6.74) and from the representation (6.77,6.78) that both
functions and their first derivatives vanish at t = t′ and thus we have
Integrating the identity (6.72) from α = 0 to α = 1 and noting that the infinite product is one
for α = 0 finally yields
Y λn
log = log D(t) − log D0 (t), (6.82)
λ0n
M Y λn D(t)
det = 0
= , (6.83)
M0 λn D0 (t)
for the functional determinants appearing in the path integrals. Nowhere did we assume that M0
is the free fluctuation operator such that the above formula holds true if U0 in M0 = ∂ 2 + U0 (t)
is time-dependent.
We could have anticipated (6.83) by noting that
M0 + αU
det
M0
is an analytic function of α with simple zeros at values of α for which one of the eigenvalues
λn (α) vanishes. For these α the function Dα (s) vanishes at t and is proportional to the cor-
responding eigenfunction. Hence, the infinite product and the analytic function Dα (t) share
the same zeros. Clearly, since for α = 0 the determinant is one, we must divide Dα by the
α-independent D0 , but this normalization does not remove or add zeros/poles. According to a
theorem by Weierstrass the quotient Dα (t)/D0 (t) must then be an analytic function exp(f (α))
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.4. Functional Determinants 58
without zeros times the determinant. Finally from the asymptotic growth of the zeros of Dα one
can deduce that f (α) = 0 and this proves (6.83).
What we have shown then is that the kernel (3.62) is given by
!1/2
m D0 (t, t′ )
s
′ ′
K(t, q, t , q ) = eiS[wcl ]/h̄ , (6.84)
2πih̄(t − t′ ) D(t, t′ )
where D0 (s, t′ ) and D(s, t′ ) are the Gelfand-Yaglom D-functions belonging M and M0 . They
fulfill the initial conditions
∂D(s, t′ ) ∂D0 (s, t′ )
D(t′ , t′ ) = D0 (t′ , t′ ) = 0 and ′ = ′ = 1. (6.85)
∂s s=t ∂s s=t
For the free particle D0 (s) = (s − t′ ) and its only role in (6.84) is to chancel the t − t′ in the
square root. We thus recover our result for the kernel of the oscillator with time dependent force
in (3.21).
ξ̈ = P ξ̇ + Qξ (6.87)
with ’boundary operators’ BL and BR . For Dirichlet boundary conditions at t′ and t we have
1 0
! !
0 0
BL = and BR = . (6.90)
0 0 1 0
————————————
A. Wipf, Path Integrals
CHAPTER 6. EUCLIDEAN PATH INTEGRAL 6.4. Functional Determinants 59
To calculate the determinant of the fluctuation operator M one solves the initial value problem
d
ψ = Kψ, ψ(t′ ) = 1. (6.91)
ds
The columns of the matrix ψ(s) form a complete set of linearly independent solutions of the first
order differential equation (6.88): any solution of this equation must be a linear combination of
the columns of ψ. Let us assume that the determinant of
(6.91)
BL ψ(t′ ) + BR ψ(t) = BL + BR ψ(t)
vanishes. Then there exists a linear combination of the columns of ψ which satisfies the bound-
ary condition (6.89). This linear combination gives rise a solution of the second order fluctua-
tion problem (6.87) fulfilling the boundary conditions. This means that det(M) vanishes when
the determinant of BL + BR ψ(t) vanishes. Since the converse statement holds as well, it is not
surprising that the ratio of two fluctuation determinants is given by [19]
det M1 det(−∂ 2 + P1 ∂ + Q1 ) det(BL + BR ψ 1 (t))
= 2
= . (6.92)
det M2 det(−∂ + P2 ∂ + Q2 ) det(BL + BR ψ 2 (t))
For one degree of freedom and Dirichlet boundary condition with boundary operators (6.90)
one recovers the original Gelfand-Yaglom formula for the ratio of functional determinants.
Zeta-function determinants
Finally we introduce another representation for determinants which is popular in field theory,
namely the zeta-function definition of determinants. The zeta-function of a (elliptic) selfadjoint
M is defined as
1 1
X −s Z Z
dt ts−1
X −tλ
ζM (s) = λn = − e n =− dt ts−1 tr e−tM , (6.93)
Γ(s) Γ(s)
where Γ(s) is the Riemann zeta-function. From the bounds (6.62) one infers that the difference
λn
+ O(s2)
X
ζM (s) − ζM0 (s) = −s log
λ0n
is analytic at the origin and that its derivative at his point is related to the determinant of M/M0
as follows:
M ′ ′
log det = −{ζM (0) − ζM (0)}. (6.94)
M0 0
The advantage of this representation is that one only needs to know the derivative of the zeta
function at s = 0 to compute the determinants of interest. In addition, the zeta-function defined
determinants have some nice properties. In particular they respect a local gauge symmetry.
Later we shall see that in certain field theoretical models (e.g. the massless Schwinger model)
′
the value ζM (0) can be calculated explicitly.
————————————
A. Wipf, Path Integrals
Chapter 7
Brownian motion
The well-known Brownian motion is a particular Gaussian stochastic process with covariance
E(wτ wσ ) ∼ min(τ, σ). There are many other known examples of Gaussian stochastic pro-
cesses, for example the Ornstein-Uhlenbeck Process or the oscillator process. They all belong
to a larger class of processes which are in general not even Gaussian and which we shall discuss
in the appendix.
The Brownian process describes the disordered motion of small particles suspended in a
liquid. It is believed that Brown studied pollen particles floating in water under the microscope.
He observed minute particles executing a jittery motion. The theory of this motion has been
invented by E INSTEIN and S MOLUDCHOWSKI. The mathematically rigorous construction of
the corresponding stochastic process has been developed by W IENER.
We have seen that contrary to the complex transition amplitude K(t, q, 0) in ordinary quan-
tum mechanics, its continuation K(τ, q, 0) defines a probability density. For the free particle
starting at the origin the probability to end up at q after a ’time’ τ is
m d/2 −mq2 /2τ
P0 (τ, q) = e , (7.1)
2πτ
and the probability to end up in the open set O ⊂ Rn is
Z
P0 (τ, O) = dq K0 (τ, q, 0) ≤ 1. (7.2)
O
P0 belongs to a Brownian motion, named after the botanist ROBERT B ROWN. Although the
mathematical model of Brownian motion is among the simplest continuous-time stochastic pro-
cesses it has several real-world applications. An example is stock market fluctuations.
7.1 Diffusion
Diffusion is described by Fick’s diffusion laws [25]. They were derived by A DOLF F ICK in
the year 1855. The first law relates the diffusive flux to the concentration field, by postulating
60
CHAPTER 7. BROWNIAN MOTION 7.1. Diffusion 61
that the flux goes from regions of high concentration to regions of low concentration, with a
magnitude and direction that is proportional to the concentration gradient,
J = −D∇φ. (7.3)
Here J is the diffusion flux, D the diffusion coefficient with dimension m2 /s and φ is the con-
centration of the diffusing substance. D is proportional to the squared velocity of the diffusing
particles, which depends on the temperature and viscosity of the fluid and the size of the parti-
cles according to the Stokes-Einstein relation
kB T
D= , (7.4)
γ
where γ is the drag coefficient, the inverse of the mobility. For spherical particles of radius
r in a medium with viscosity η the drag coefficient is γ = 6πηr. In applications the driving
force is a out of equilibrium concentration of particles, a spacial distribution of temperature or
a non-vanishing gradient of a chemical potential.
Fick’s second law predicts how diffusion causes the concentration field to change with time
τ . It follows from his first law and the continuity equation
∂φ
= −∇ · J (7.5)
∂τ
which expresses our expectation that the number of particles is conserved. The change of the
number of particles in a given region is equal to the number of particles leaving or entering the
region through its boundary. Inserting the continuity equation into (7.3) yields the second law
of Fick,
∂φ
= ∇ · (D∇φ). (7.6)
∂τ
For a constant diffusion coefficient D this law simplifies to
∂φ
= D△φ (7.7)
∂τ
and it has the same form as the heat equation. An important example is the equilibrium case for
which the concentration does not change in time, so that the left side of (7.7) is identically zero
and △φ = 0. This is Laplace’s equation, the solutions to which are harmonic functions.
If we start at time 0 with one particle at q ′ the solution of (7.7) in denoted by K0 (τ, q). With
the initial condition
————————————
A. Wipf, Path Integrals
CHAPTER 7. BROWNIAN MOTION 7.2. Discrete random walk 62
as can be verified by substitution. This has been known since the beginning of the last century
and forms the subject of several textbooks on Brownian motion [26]. This particular solution is
just the Euclidean propagator (7.1) if we identify D = 1/2m.
1 1
2 2
a b
Figure 7.1: The particle may jump with equal probabilities one step to the left or right.
steps are statistically independent. Hence, the probability for a transition from qj = ja to
qk = ka during a time ǫ is
1
if |k − j| = 1
Pkj = P (ǫ, qk , qj ) = 2 . (7.10)
0 otherwise.
This simple example of a stochastic process (actually a Markov chain) is homogeneous and
isotropic,
After n time-steps the probability to jump from qj to qk is given by the sum of the probabilities
of the possible ways of achieving that, which is just
The initial position of the particle may be uncertain and the probability to find it at lattice point
qj is pj . If it sits with certainty 1 at the origin then pj = δj0 . After n time-steps the system
has evolved and produced a new distribution P n p. The evolution operators P n determines the
change of the initial probability distribution after n time-steps.
————————————
A. Wipf, Path Integrals
CHAPTER 7. BROWNIAN MOTION 7.3. Scaling limit 63
It is not difficult to calculate the powers of P . The probability to hop from the lattice site qj
to the site qk after n time-steps is 1/2n times the number of paths on the lattice from qj to qk . If
n is even then k − j must be even and if n is odd then k − j must be odd. The particle must jump
r = 21 (n + k − j) steps to the right and ℓ = 12 (n + j − k) steps to the left. The number of paths
from qj to qk is then equal to the number of ways one can combine r steps to the right with ℓ
steps to the left to obtain a path of length n. This number is given by the binomial coefficient.
Hence one finds the following probability
! !
1 n 1 n
P (nǫ, qk − qj ) = n = n .
2 r 2 ℓ
1 a2 1
{P (τ + ǫ, q) − P (τ, q)} = {P (τ, q + a) − 2P (τ, q) + P (τ, q − a)}, (7.14)
ǫ 2ǫ a2
where τ = nǫ is the time during which the particle jumps.
a2
a → 0, ǫ → 0 with nǫ = τ, D= fixed. (7.15)
2ǫ
Other scaling limits are possible. For example a → 0 with fixed ǫ would lead to a situation
where the particle does not move anymore. The limit a, ǫ → 0 with fixed a/ǫ would lead to a
classical theory without fluctuations. But if we keep a2 /ǫ constant then the correlations tend to
finite values in this so-called diffusion limit. The constant D is the macroscopic diffusion con-
stant. In the macroscopic description q and τ become continuous variables and the difference
equation (7.14) converts into a one-dimensional continuous diffusion equation
∂ ∂2
P (τ, q) = D 2 P (τ, q). (7.16)
∂τ ∂q
————————————
A. Wipf, Path Integrals
CHAPTER 7. BROWNIAN MOTION 7.3. Scaling limit 64
At the initial time no diffusion has occurred and P (0, q) = δ(q). The solution of the diffusion
equation with this initial condition is just the Gaussian function
1 2
P0 (τ, q) = K0 (τ, q, 0) = √ e−q /4Dτ . (7.17)
4πDτ
The transition probability for the discrete random walk is replaced by the probability
1 q
Z
2 /4Dτ
lim P (nǫ, qj ) = √ du e−u . (7.18)
q <qj <q
′
4πDτ q′
The trivial matrix identity P n P m = P n+m turns into the Chapman-Kolmogorov equation
Z
du P0 (τ, q − u)P0(σ, u − q ′ ) = P0 (τ + σ, q − q ′ ). (7.19)
Higher dimensions
The extension to higher dimensions is not difficult. For that we note that the lattice-Laplacian
in one dimension acts on a function on the lattice as follows,
1
(△L f )(qj ) = 2 {f (qj + a) − 2f (qj ) + f (qj − a)} (7.20)
a
such that the probability for a transition (7.10) can be rewritten as
a2
P =1+△L . (7.21)
2
Now we calculate the n’th power of P for n → ∞ and use the scaling laws in (7.15)
a2 τ /n n Dτ n
n→∞
Pn = 1 + △L = 1 + △L −→ e τ D△ , (7.22)
2 ǫ n
where lima→0 △L = △ is the second derivative in the continuum. The kernel hq, τ | eτ D△ |0i
is just the above distribution P0 (τ, q). Now the generalization to d dimensions is natural. If j
enumerates the lattice points on a d-dimensional hypercubic lattice with lattice spacing a, then
the matrix P is given by
1
i, j nearest neighbors a2
Pij = 2d or P = 1 + △L , (7.23)
0 otherwise 2d
where △L is the lattice Laplacian in d-dimensions, given by
a2 △L f (qj ) =
X
f (qk ) − 2d · f (qj ). (7.24)
k:|k−j|=1
The factor 1/2d in (7.23) is needed such that the probability to go somewhere is 1. In the scaling
limit we end up with a similar result as in one dimension,
a2
lim P n = eτ D△ , nǫ = τ,
= D, (7.25)
n→∞ 2dǫ
and the Brownian motion tends to a Gaussian process with Laplacian △.
————————————
A. Wipf, Path Integrals
CHAPTER 7. BROWNIAN MOTION 7.4. Expectation values and correlations 65
The event wτ ∈ O simply means that the Brownian particle has passed the region O at time τ ,
as sketched in figure 7.2. The probability of finding the particle at time τ1 in the open set O1 , at
space
τ time
Figure 7.2: Brownian path starting at q = 0 and passing through the window O at time τ .
A stochastic process for which the finite dimensional distributions fulfills these conditions and
for which
1 if 0 ∈ O
P0 (wτ =0 ∈ O) = (7.28)
0 otherwise
is called a Wiener process. With the distribution P0 (q, τ ) at hand we can answer all possible
questions we may think of.
For example, it is not difficult to check that the expectation value (or mean value) of the
position of a Brownian particle is zero,
Z
E(wτ ) = du u P0(τ, u) = 0. (7.29)
The process recalls the starting position 0 since future positions are constrained by (7.29).
E(wτ ) is to be interpreted as a conditional expectation. It is the mean value of wτ , given the
information w0 = 0. Let us calculate the probability that the increment wτ2 − wτ1 of a Brownian
motion starting at the origin assumes some value within the regions O ∈ Rd . The answer is
Z
P (wτ2 − wτ1 ∈ O) = du2du1 P0 (u2 − u1 , τ2 − τ1 )P0 (u1 , τ1 )
u2 −u1 ∈O
————————————
A. Wipf, Path Integrals
CHAPTER 7. BROWNIAN MOTION 7.5. Appendix A: Stochastic Processes 66
where have already anticipated the result for τ < σ. For the Brownian motion in higher dimen-
sions the corresponding result reads
One can show that a typical trajectory w(τ ) of the Brownian motion is continuous. In dimension
one it is also recurrent, returning periodically to its origin. Indeed, one can prove the following
remarkable theorem:
Theorem: Let B(R, 0) ⊂ Rd be the ball with radius R centered at the origin. Then
1 for d = 1, 2
E (wτ ∈ B(R, 0) for one τ ) =
< 1 for d ≥ 3.
The times of return of a one-dimensional Brownian motion can serve as a sophisticated random
number generator. As a mathematical model it does not only describe the random movement of
small particles suspended in a fluid; it can be used to describe a number of phenomena such as
fluctuations in the stock market. Trajectories of a Brownian motion are self-similar, a term that
is often used to describe fractals. Self-similarity means that for every segment of a given curve,
there is either a smaller segment or a larger segment of the same curve that is similar to it.
————————————
A. Wipf, Path Integrals
CHAPTER 7. BROWNIAN MOTION 7.5. Appendix A: Stochastic Processes 67
• The second entry A of the triplet denotes a σ-algebra of subsets of Ω called events. A
σ-field is closed under complementation, countable intersections and unions,
∞
[
A, B, Ai ∈ A =⇒ A\B ∈ A, Ai ∈ A, Ω ∈ A. (A.1)
i=1
• The third entry P is a probability measure. To any event A ∈ A it assigns its probability
P (A) ∈ [0, 1]. The probability of the empty set ∅ is zero and that of the sample space Ω
is one. The measure has the following natural property
X
P ( ∪ Ai ) = P (Ai ) for Ai ∈ A, Ai ∩ Aj = ∅, i 6= j. (A.2)
We recall that the Borel sets is the largest σ-algebra containing the open sets in Rd . Let X be
Borel-measurable on (Ω, A). Then X is called P -integrable, if
∞
( )
X k k k+1
lim P w : < X(w) ≤ ≡ J+
n→∞
k=0 n n n
0
( )
X k k−1 k
lim P w: < X(w) ≤ ≡ J−
k=−∞ n n n
n→∞
————————————
A. Wipf, Path Integrals
CHAPTER 7. BROWNIAN MOTION 7.5. Appendix A: Stochastic Processes 68
Of particular importance are real-valued random variables PX . They define measures on Borel
sets in R. From the theorem one immediately concludes the
Lemma: If f is Borel measurable on R and X : Ω → R a real random variable, then
Z Z
E(f ◦ X) = (f ◦ X) dP = f dPX (A.7)
Ω R
In particular, the expectation value of a random variable is
Z
E(X) = w dPX (w), (A.8)
holds true. Important and often used random variables are the Gaussian ones. A random variable
is called Gaussian with variance Σ ≥ 0 and mean 0 if
1 2
dPX (w) = √ e−w /2Σ =⇒ E(X) = 0 and E(X 2 ) = Σ. (A.14)
2πΣ
For a vanishing variance Σ this simplifies to δ(w)dw. A set of Gaussain random variables Xi is
joint Gaussian with means E(Xi ) = 0 and covariance E(Xi Xj ) = Σij if
1 −1
PX1 ⊗...⊗Xn (w) = q e−(w,Σ w)/2 . (A.15)
(2π)n det Σ
After this preparations we introduce the notion of stochastic processes. A stochastic process is
a family of random variables labelled by a (continuous) real parameter. In most application this
parameter is time. More accurately:
————————————
A. Wipf, Path Integrals
CHAPTER 7. BROWNIAN MOTION 7.5. Appendix A: Stochastic Processes 69
P (τ, w) ∼ PXτ
in the present notation. The probability for finding the Brownian particle at time τ in the interval
between a and b is
Z b
P ({ω : a ≤ Xτ (ω) ≤ b}) = dPXτ (w). (A.16)
a
Similarly, the probability for finding the particle at times τi between ai and bi , where 1 ≤ i ≤ n,
is given by
Z b1 Z bn
P ({ω : ai ≤ Xτi (ω) ≤ bi }) = ... dPXτ1 ⊗...⊗Xτn (w1 , . . . , wn ). (A.17)
a1 an
If τ1 , . . . , τn and n are arbitrary one obtains the set of cylinder sets. They do not form a σ-
algebra. In the theory of stochastic processes one uses the σ-algebra generated by this set.
Let us now indicate how one constructs a unique probability measure on the set of paths,
that is how one performs the limit from joint distributions on a finite set of random variables to
a induced probability measure dPXI . The construction proceeds as follows:
Let J be a subset of the parameter set I and define
EJ = and let BJ = ⊗τ ∈J Bτ
Y
Eτ , Eτ = E (A.18)
τ ∈J
pJ : E J −→ E (A.19)
are measurable. For a finite set J the elements of E J may be identified with the corners of a
broken line path and for a continuous J as a path J → E. The BJ is then a σ-algebra on these
sets. In addition, let K be a subset of J. Then the projections
————————————
A. Wipf, Path Integrals
CHAPTER 7. BROWNIAN MOTION 7.5. Appendix A: Stochastic Processes 70
XJ = ⊗τ ∈J Xτ : Ω −→ E J , Ω ∋ ω −→ path Xτ (ω), τ ∈ J
and let PXJ denote the joint distribution of random variables Xτ ∈J . For example, for the set
J = (τ1 , . . . , τn ) this means
for two arbitrary finite subsets K and J with K ⊂ J, then the family is called projective. Now
one can prove the following important theorem
Theorem (Kolmogorov): Is E = Rn and B the σ-algebra of it Borel sets and if I is a non-empty
set, then to each projective family PXJ of probability measures with finite J on (E J , BJ ) there
exists exactly one probability measure PXI on (E I , BI ) with
One calls PXI the projective limit of the family PXJ . The following theorems are useful:
Theorem: Let Σ(τ, σ) be a continuous and real-valued function on I ×I, where I is a separable
topological space. If for all {τ1 , . . . , τn } the function Σ(τi , τj ) is positive semi-definite, then
there exists a Gaussian process (Ω, A, P, Xτ ) with covariance Σ, that is
To explain the following theorem, due to B OCHNER, we introduce the characteristic function
of a random variable,
Definition: Let X be a Rd -valued random variable. Then the Fourier transform of its measure,
Z
φX (j) = ei(j,w) dPX (w) = E ei(j,X) (A.22)
is called the characteristic function. The measure PX is uniquely determined by the character-
istic function φX of the random variable. Now we can state the
————————————
A. Wipf, Path Integrals
CHAPTER 7. BROWNIAN MOTION 7.5. Appendix A: Stochastic Processes 71
Theorem (Bochner): A function a(j) is the characteristic function of a random variable if and
only if a(j) is continuous, a(0) = 1 and
∀j1 , . . . , jn ∈ R; z1 , . . . , zn ∈ C.
X
z̄i a(ji − jj )zj ≥ 0
The following theorem states, that under certain conditions the measure lives on the set of
continuous paths:
Theorem (Kolmogorov-Prehorov): Let (Ω, A, P, Xj⊂I ) be a stochastic process. Then one may
change the random variables on a set of measure zero such that the new process (Ω, A, P, X̃τ ⊂I )
is continuous, provided that there exist real numbers a > 0, b > 1, c > 0 such that
One concludes that almost all (in the sense of measure theory) Brownian paths are continuous.
One can also prove that almost all Brownian paths are nowhere differentiable.
————————————
A. Wipf, Path Integrals
Chapter 8
The Feynman path integral formulation reveals a deep and fruitful interrelation between quan-
tum mechanics and statistical mechanics. The discretized Euclidean path integrals can be
viewed as partition functions for particular lattice spinmodels. Vice versa, the partition function
in statistical mechanics is given by a particular path integral with imaginary time. This unified
view of quantum and statistical mechanics allows us to apply many powerful methods known
in statistical mechanics to calculate correlation functions in Euclidean quantum mechanics.
Here one integrates over all paths starting at q ′ and ending at q. For imaginary times the inte-
grand is real and positive and contains the Euclidean action SE . Clearly, the object
Z
dqK(τ, q, q) = tr e−τ Ĥ/h̄ (8.2)
is just the quantum partition function for temperature kB T = h̄/τ . We conclude that the parti-
tion function Z(β) and free energy F (β) at inverse temperature1 β = 1/T are given by
Z
−βF (β) −β Ĥ
Z(β) = e = tr e = dq K(h̄β, q, q). (8.3)
1
We set kB = 1.
72
CHAPTER 8. STATISTICAL MECHANICS 8.2. Thermal Correlation Functions 73
Note that setting q = q ′ in the last term means that we integrate over periodic paths starting and
ending at q in (8.1). Integrating over q is then equivalent to integrating over all periodic paths
with period h̄β. Thus we end up with
I
Z(β) = Dw e−SE [w]/h̄ . (8.4)
w(0)=w(h̄β)
For example, for the Euclidean harmonic oscillator with evolution kernel (6.10)) we have
For a Hamiltonian with discrete energies En the partition function has the spectral resolution
where as orthonormal states |ni we used the eigenfunction of Ĥ. A comparison with equation
(8.6) immediately yields the energies of the harmonic oscillator with circular frequency ω,
1
En = h̄ω n + , n ∈ N. (8.8)
2
For low temperature β → ∞ the spectral sum (8.7) is dominated by the contribution of lowest
energy such that the free energy will tend to the ground state energy,
1 β→∞
F (β) = − log Z(β) −→ E0 . (8.9)
β
To perform the high-temperature limit is more tricky. We shall investigate this limit later in this
lecture. In applications one is also interested in the energies and wave functions of the excited
states. Now we shall discuss a method to extract these quantities from the path integral.
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.2. Thermal Correlation Functions 74
The normalizing factor in the denominator is the partition function (8.4). At temperature zero
these correlation functions are just the Schwinger functions we introduced in section 6.1.
The gap between the energies of the ground state and first excited state can be extracted
from the thermal 2-point function
1
hq̂E (τ1 )q̂E (τ2 )iβ = tr e−β Ĥ q̂E (τ1 )q̂E (τ2 )
Z(β)
1
= tr e−(h̄β−τ1 )Ĥ/h̄ q̂ e−(τ1 −τ2 )Ĥ/h̄ q̂ e−τ2 Ĥ/h̄ (8.12)
Z(β)
as follows: we calculate the trace with the orthonormal energy eigenstates |ni and insert the
identity |mi hm| after the first position operator. Denoting the energy of|ni by En we obtain
P
Now we consider the low-temperature limit of this expression. For β → ∞ the terms containing
the energies En>0 of the excited states are exponentially damped and we conclude
Note that the term exp(−βE0 ) chancels against the low temperatur limit of the partition function
in the denominator. At low temperature the thermal expectation values tend to the ground state
expectation values,
lim hq̂E (τ1 )q̂E (τ2 )iβ = hΩ| q̂E (τ1 )q̂E (τ2 )|Ωi (8.15)
β→∞
This just expresses the fact that thermal fluctuations freeze out at low temperature. Thermal
expectation values become vacuum expectation values at absolute zero. In particular the 1-
point function has this property,
hq̂E (τ1 )q̂E (τ2 )ic,β = hq̂E (τ1 )q̂E (τ2 )iβ − hq̂E (τ1 )ihq̂E (τ2 )iβ . (8.17)
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.2. Thermal Correlation Functions 75
It characterizes the correlations of the differences ∆q̂E = q̂E − hq̂E i at different times,
hq̂E (τ1 )q̂E (τ2 )ic,β = h∆q̂E (τ1 ) ∆q̂E (τ2 )iβ . (8.18)
The connected correlator is exponentially damped for large |τ1 − τ2 | since the term with m = 0
in (8.14) is missing in the spectral resolution for the connected 2-point function,
lim lim hq̂E (τ1 )q̂E (τ2 )ic,β = e−(E1 −E0 )(τ1 −τ2 )/h̄ | hΩ| q̂|1i |2 , (8.20)
τ1 −τ2 →∞ β→∞
and thus one can extract the energy-gap E1 − E0 and modulus of the matrix element hΩ| q̂|1i
from the large-time behavior of the connected 2-point function.
To find the path-integral representation for the thermal correlators we proceed similarly as
in section 2.4. To calculate the two-point function (8.12) we insert the identity 1 = du|ui|ui
R
and in terms of the kernel K(τ, u, v) for the imaginary time evolution we obtain for the integrand
in (8.21) the expression
Z
hq| . . .|qi = dvdu hq| K(h̄β − τ1 )|viv hv| K(τ1 − τ2 )|ui u hu| K(τ2 )|qi .
Now we use the path-integral representation for each evolution kernel separately, similarly as
we did for the Greensfunctions in section 2.4. Then the path-integral representation of the is
evident: First we sum over all path from q to u in time τ2 and then multiply with the position u
of the particle at time τ2 . Next we sum over all path from u to v in time τ1 − τ2 and multiply
with the position v of the particle at time τ1 . Finally we sum over all path from v to q in time
h̄β − τ1 . The integration over the intermediate positions u and v just means that we must sum
over all paths, not only over those which have positions u and v at times τ2 and τ1 , and include a
factor q(τ1 )q(τ2 ) in the integrand. Since the total traveling time of the particle is h̄β, we obtain
the representation
w(h̄β)=q
Z
−β Ĥ
hq| e q̂E (τ1 )q̂E (τ2 )|qi = Dw e−SE [w]/h̄ w(τ1 )w(τ2 ), τ1 > τ2 , (8.22)
w(0)=q
for the kernel of the thermal 2-point function. Clearly, to get the trace in (8.12) we must inte-
grate over q and then divide by Z(β) (whose path-integral representation we derived earlier).
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.2. Thermal Correlation Functions 76
Integrating over q means that we integrate over all periodic paths with period h̄β. When we
applied the Trotter product formula we have assumed that τ1 is bigger than τ2 . The path integral
representations for the higher ’time ordered’ thermal correlation functions is now evident,
1
I
hT q̂E (τ1 ) · · · q̂E (τn )iβ = Dw e−SE [w]/h̄w(τ1 ) · · · w(τn ). (8.23)
Z(β)
w(0)=w(h̄β)
or the corresponding generating functional, which is just the partition function in the presence
of an external source j(τ ),
Z I
Z[β, j] = dqK(h̄β, q, q; j) = Dw e−SEj [w]/h̄ (8.25)
w(0)=w(h̄β)
by differentiating with respect to the source. The source term (j, w) in (8.24) is the scalar
product on L2 [0, h̄β]. Note that Z(β, 0) = Z(β) is the previously introduced partition function
without source. For example, the thermal two-point function is
1 h̄2 δ 2
hT q̂E (τ1 )q̂E (τ2 )iβ = Z[β, j]|j=0, (8.26)
Z(β, 0) δj(t1 )δj(t2 )
where T denotes the time ordering.
The connected correlation functions are generated by the finite-temperature Schwinger func-
tional. It is proportional to the logarithm of the partition function and hence proportional to the
free energy with external source,
In many cases the source is just the applied magnetic field. The connected correlation functions
are gotten by functionally differentiating W [β, j] several times with respect to the source,
h̄n−1 δ n
hT q̂E (τ1 )q̂E (τ2 ) · · · q̂E (τn )ic,β = W [β, j]|j=0. (8.28)
δj(τ1 ) · · · δj(τn )
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.2. Thermal Correlation Functions 77
For later use we allow for a time-dependent frequency. We proceed similarly as in section 3.2
where we calculated the propagator for the driven oscillator. Not to repeat ourselves we just
recall the key ideas and point out the main differences as compared to the previous calculation.
First one splits w into the classical path wcl from q ′ to q and the fluctuation ξ(τ ) which vanishes
at the endpoints. The classical path is determined by the inhomogeneous equation of motion
d2
S ′′ wcl (τ ) = j(τ ), S ′′ = −m
+ mω 2 (τ ), (8.30)
dτ 2
and the boundary conditions wcl (0) = q ′ and wcl (h̄β) = q. Expanding the action SEj about the
dominant classical path yields
m
s
′ −SEj [wcl ]/h̄ −SEj [wcl ]/h̄
K(h̄β, q, q ; j) = e K(h̄β, 0, 0) = e , (8.31)
2πh̄D(h̄β)
where K(h̄β, 0, 0) is the propagator for the propagation without source from 0 to 0. The D-
function solves S ′′ D = 0 with initial conditions D(0) = 0 and Ḋ(0) = 1. This formula is just
the Euclidean version of the result (3.32). Again we decompose the classical trajectory in two
parts,
Z h̄β
wcl (τ ) = wh (τ ) + GD (τ, σ)j(σ)dσ, (8.32)
0
where wh is the classical path from q ′ → q without external source and GD (τ, σ) is the symmet-
ric Dirichlet Greenfunction of the fluctuation operator S ′′ . The second term on the right hand
side vanishes at initial and final times and solves the equation of motion with source. Using this
information the action of wcl can be rewritten as
1
SEj [wcl ] = SE [wh ] − (j, wh ) − (j, GD j) . (8.33)
2
Now we insert this decomposition into (8.31). The term SE [wh ] converts the source-free propa-
gator for the propagation from 0 → 0 in (8.31) into the source-free propagator for propagation
from q ′ to q such that
K(h̄β, q, q ′; j) = K(h̄β, q, q ′) · exp {(j, GD j)/2h̄ + (j, wh )/h̄} . (8.34)
In order to obtain the partition function we must set q ′ = q and integrate over q. To do the
integration we need the explicit q-dependence of the action SEj [wcl ] in (8.33). To that aim we
introduce a particular solution φ(τ ) of the homogeneous equation of motion,
S ′′ φ = 0 and φ(0) = φ(h̄β) = 1. (8.35)
The classical solution wh from q → q in the decomposition (8.32) is just wh (τ ) = qφ(τ ). Using
the equation of motion and boundary conditions for φ the position dependent contributions to
SEj in (8.33) can be isolated,
1 1
SEj [wcl ] = mβ q 2 − q(j, φ) − (j, GD j), (8.36)
2 2
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.2. Thermal Correlation Functions 78
We see that the resulting propagator K(h̄β, q, q; j) is a Gaussian function of q. Integrating over
q we obtain the partition function in the presence of an source
1 φ(τ )φ(σ)
W [β, j] = (j, GP j), where GP (τ, σ) = + GD (τ, σ). (8.40)
2 mβ
Actually GP is just the Greenfunction of S ′′ for periodic boundary conditions. The term pro-
portional to φ(τ )φ(σ) in (8.40) converse the Dirichlet Greenfunction into the periodic Green-
function. This fact is proven in the appendix to the present chapter.
To compute th connected 2-point function at finite temperature we differentiate W [β, j]
twice with respect to the source and then set the source to zero. We obtain
Note that the partition function Z(β) does not enter the result for the connected correlation
function. For the oscillator with constant frequency the periodic Greenfunction reads
1 cosh(ω|τ − σ| − h̄ωβ/2)
GP (τ, σ) = (8.42)
2mω sinh(h̄ωβ/2)
β→∞ h̄
h Tq̂E (τ1 )q̂E (τ2 )iβ = h̄GP (τ1 , τ2 ) −→ e−ω|τ1 −τ2 | . (8.43)
2mω
Comparing with (8.20) we can extract both the mass gap and modulus of the matrix element of
q̂ between ground state and first excited state,
h̄
E1 − E0 = h̄ω and | hΩ| q̂|1i |2 = , (8.44)
2mω
a familiar result from the quantum mechanics course.
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.3. Wigner-Kirkwood Expansion 79
The normalization of the (formal) path integral is fixed by the classical limit Zcl (β) in (8.45).
We rescale imaginary time such that the periodicity of the paths is β instead of h̄β,
τ −→ h̄τ. (8.47)
Observe that for a moving particle the kinetic energy dominates the potential energy for small h̄
whereas for a particle at rest the potential term is dominant. This suggest to split a path into its
constant part (for which the kinetic energy vanishes) and fluctuations about it. Thus we change
variables from w(τ ) → q + h̄ξ(τ ) such that ξ(0) = ξ(β) = 0. We obtain
( ) ( )
m β β
Z Z Z Z
Z(β) = dq Dξ exp − ξ dτ˙2 exp − V (q + h̄ξ) dτ . (8.49)
2 0 0
ξ(0)=ξ(β)=0
Now we expand the last exponential factor containing the potential in powers of h̄,
h̄2 ′ 2 2
(
−βV
exp { . . . } = e 1 − h̄V ′ I(ξ) + V I (ξ) − V ′′ I(ξ 2 )
2
h̄3 ′ 3 3
)
− V I (ξ) − 3V ′ V ′′ I(ξ)I(ξ 2) + V ′′′ I(ξ 3 ) + . . . , (8.50)
3!
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.3. Wigner-Kirkwood Expansion 80
where the argument of V and its derivatives is the constant path q and we abbreviated the time
integrals of powers of the fluctuation field by
Z
ξ n (τ )dτ ≡ I(ξ n ). (8.51)
The remaining path integral in (8.49) leads to correlators of the time integrated fluctuations, for
example
m
Z Z Z Z
2
hI(ξ )I(ξ)i = Dξ exp − ξ˙ 2 2
ξ (τ )dτ ξ(σ)dσ. (8.52)
2
ξ(0)=ξ(β)=0
The explicit form of the lowest order terms in the resulting series is
h̄2 ′ 2 2
Z ( )
βV ′
Z(β) = dq e h1i − h̄V hI(ξ)i + V hI (ξ)i − V ′′ hI(ξ 2)i + . . . (8.53)
2
The expectation values (8.52) before time-integration are generated by the generating functional
of the free particle with Dirichlet boundary conditions and h̄ = 1,
s ( )
m 1 Zβ Zτ
K(β, 0, 0; j) = exp dτ dσ (β − τ )σj(τ )j(σ) . (8.54)
2πβ mβ 0 0
where we have already anticipated that the odd powers of h̄ vanish in this expansion, since the
odd moments of the free measure vanish.2
The first term in this power series in h̄ is the classical partition function. The coefficients of
O(h̄n ) contain exactly n derivatives of the potential, e.g. the h̄4 coefficient contains terms like
We see that the h̄-expansion is actually a gradient expansion. A strongly coupled system be-
haves classically and we may expect that its behavior is described by a Wigner-Kirkwood gradi-
ent expansion. The first few terms in the semiclassical expansion have been derived by W IGNER
AND K IRKWOOD [27].
2
One needs to restore an additional factor of 1/h̄ to find the correct classical limit. This is due to the different
rescaling of the constant path and the fluctuations ξ(τ ).
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.4. High Temperature Expansion 81
The heat kernel K(β, q) is the path integral over the fluctuations in (8.49) (to simplify the
notation we set h̄ = 1). Now we rescale time such the fluctuation have periodicity 1 and
at the same time rescale the amplitude of the fluctuations such the kinetic term becomes β-
independent:
q
τ −→ βτ and ξ −→ β ξ. (8.59)
After these rescaling the β-dependence comes only from the temperature-dependent potential:
1 m 1
Z Z Z q
Z(β, q) = √ Dξ exp − ξ˙2 exp −β V (q + β ξ) dτ . (8.60)
β 2 0
ξ(0)=ξ(1)=0
Now we expand the last exponential in powers of β. Again we are lead to calculate time integrals
of correlators of the fluctuation field. Using hξ(τ )i = 0 we find the following expansion for the
heat kernel
β2 2 β2
*
q
Z(β, q) = V (q) − V ′′ (q) I(ξ 2)
β 1 − βV (q) +
2 2
β3 3 β3 ′ ′ 2 β3 3
+
β
− V + V V I (ξ) + V V ′′ I(ξ 2 ) − V ′′′′ I(ξ 4 ) . . . . (8.61)
3! 2 2 24
Inserting the correlators (8.55) with β = 1 we can compute the remaining correlators and time-
integrals. This way one obtains
β2 β 2 ′′
s (
m
Z(β, q) = 1 − βV + V 2 − V
2πβ 2 12m
β3 β3 ′ ′ β3 β3
)
− V3+ V V + V V ′′ − V ′′′ + . . . , (8.62)
6 24m 12m 240m2
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS /2
8.5. High-T Expansion for D 82
where the potential and its derivatives are evaluated on the constant path q. The partition func-
tion is the space integral of this kernel. If the potential is localized near the origin or alternatively
if space is compact we obtain the series expansion
β2 2 β3 3 β3 ′ 2
s ( )
m Z
Z(β) − Z0 (β) = dq −βV + V − V − V + ... (8.63)
2πβ 2 6 12m
For very high temperature the particles move almost freely and as expected Z(β) tends to the
partition function Z0 (β) of a gas of free particles with mass m. Note that besides the overall
factor β −1/2 the density and the partition function have an expansion in integer powers of β,
√
although the argument of V in (8.60) contains β. The corresponding expansion for a particle
propagating in d-dimensions reads
!d/2 Z
β2 β3 β3
( )
m
Z(β) − Z0 (β) = d q −βV + V 2 − V 3 −
d
(∇V )2 + . . . . (8.64)
2πβ 2 6 12m
In [28] the high temperature expansion was completely computerized using the algebraic lan-
guage FORM. The calculation was performed up to order β 11 . For the coefficient of β n in the
expansion of Z(β) the number of terms are:
n 7 8 9 10 11
# of terms 37 114 380 1373 5301
The high temperature expansion enters the inverse mass expansion of worldline path integrals
occuring in one-loop effective action.
/ = γ µ Dµ .
D (8.66)
contains the covariant derivative which couples the fermionic field to the gauge potential Aµ ,
Dµ = ∂µ − iAµ . (8.67)
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS /2
8.5. High-T Expansion for D 83
In Quantum Electrodynamics (QED) the electric and magnetic fields entering the field strength
tensor Fµν are given by Fµν = ∂µ Aν − ∂ν Aµ . In non-Abelian gauge theories the potential Aµ
and field strength Fµν are matrix values functions. They take their values in the Lie-algebra
of the underlying gauge group. Besides the gauge potential the Dirac operator contains the
Euclidean matrices γ µ obeying the anti-commutation relations
In d dimensions they are 2[d/2] -dimensional hermitian matrices. The explicit form of the second
order hermitian matrix differential operator D/ 2 is
1 1
/ 2 = {γ µ , γ ν }Dµ Dν + [γ µ , γ ν ][Dµ , Dν ] = D µ Dµ + Σµν Fµν ,
D (8.69)
2 4
where we have introduced the components Fµν of the field strength tensor which for non-
Abelian gauge theories contain additional commutators,
The squared Dirac operator also contains the d(d − 1)/2 matrices
1 µ ν
Σµν = [γ , γ ] (8.71)
4i
generating the spin rotations in Euclidean spacetime.
Similar as in section 5.2 one proves that the Euclidean evolution kernel has the following
path integral representation:
2
Z
/
βD ′
hq| e |q i = Dw P e−SE [w,A] , (8.72)
2
/ ,
where the Euclidean action is given by the Wick-rotated Legendre transform of D
Zβ
1 2
Z
SE = dτ dd−1 x LE (w, A), LE = ẇ − iẇA − ΣF, (8.73)
4
0
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.6. Appendix B: Periodic Greenfunction 84
As in the scalar case one expands the exponent in powers of β 1/2 and uses that ξ˙ = 0 and that
R
The least trivial term is the δ-function coming from a careful evaluation of hξ˙µ ξ˙ν i. Finally, after
integrating over τ (and σ) one ends up with the following heat kernel expansion
β2
( )
1 2 µν
Z(β, q) = 1 + β ΣF + 2∆(ΣF ) + 6(ΣF ) − Fµν F + ... . (8.78)
(4πβ)d/2 12
Note that all terms in this high-temperature expansion are gauge covariant, as required. Up to an
overall factor β −d/2 only integer powers of β occur, very much like in the scalar case. The reason
is that all odd moments of the free measure vanish. To obtain the high temperature expansion
of the partition function one takes the trace with respect to the Dirac indices and integrates
over the Euclidean spacetime. The coefficients in the expansion are finite for spacetimes with
finite volumes. For non-Abelian gauge fields the calculation is not much different and the result
(8.78) also holds in this case if we only replace the Laplacian by the gauge-covariant Laplacian
and in addition take the trace over internal indices.
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.6. Appendix B: Periodic Greenfunction 85
d2
S ′′ = −m + mω 2 (τ ). (B.2)
dτ 2
with respect to periodic boundary conditions. Here GD (τ, σ) is the Greenfunction for Dirichlet
boundary conditions. φ is a solution of S ′′ φ = 0 and takes the values 1 at initial and final time
0 and β. The normalizing constant mβ is3
mβ = m φ̇(β) − φ̇(0) . (B.3)
For the proof we introduce two fundamental solutions D(τ ) and E(τ ) with boundary conditions
Two typical fundamental solutions are depicted in figure 8.1. It is easy to prove that the Dirichlet
D(τ ) E(τ )
3
τ τ
0 h̄β 0 h̄β
Greenfunction is given by
(
D(τ )E(σ) for τ < σ
mGD (τ, σ) = (B.6)
D(σ)E(τ ) for τ > σ
D(τ )
φ(τ ) = E(τ ) + . (B.7)
D(β)
3
In this appendix we set h̄ = 1.
————————————
A. Wipf, Path Integrals
CHAPTER 8. STATISTICAL MECHANICS 8.6. Appendix B: Periodic Greenfunction 86
D(σ)
Ė(β)D(σ) − E(σ)) = − − E(σ) = −φ(σ), (B.10)
D(β)
where we used Ḋ(0) = 1 and that the Wronskian at final time is Ė(β)D(β) = −1. Comparing
with (B.9) we conclude that α = −1 and this proves that GP in (B.1) with mβ given in (B.3) is
the Greenfunction of S ′′ with respect to periodic boundary conditions.
————————————
A. Wipf, Path Integrals
Chapter 9
So far we have either dealt with exactly soluble systems like the (time-dependent) oscillator or
with various expansions like the semiclassical, perturbative and high-temperature expansion. In
this section I introduce a numerical method which plays an important role in modern develop-
ments in (gauge-) field theories - the Monte-Carlo simulations.
One conveniently starts from the lattice approximation to Z(β, q) = K(h̄β, q, q) in (6.21)
for the density of the partition function. In the finite dimensional integral over the points
w1 , . . . , wn−1 defining the broken line path one sets w0 = wn = q and integrates over q to
get the trace Z(β). More generally, the lattice approximation of expectation values reads
R Z∞ Y
n
dn w A(w) e−SE (w) Z
n
hAi = R , where d w= dwj , (9.1)
dn we−SE (w) 1
−∞
and SE (w) = SE (w1 , . . . , wn ) is the Euclidean lattice-action entering the discrete path integral
(6.21). One could use standard numerical integration routines and approximate an integral by a
sum with a finite number of terms
Z M
X
dn wf (w) ∼ f (wµ )∆wµ .
µ=1
Note that wµ denotes a configuration labelled by µ whereas wj denotes the j’th component
of the configuration w = (w1 , . . . , wn ). the If the space of integration is high-dimensional, it
is however advantageous to choose the points wµ at random instead of using a regular set of
points. Note, however, that the integrand exp(−S) will vary over many orders of magnitude
for different configurations wµ (we used that already in the semiclassical approximation). The
Monte Carlo method introduced by M ETROPOLIS , ROSENBLUTH , ROSENBLUTH and T ELLER
is based on the idea of important sampling [29]. Instead of selecting the points wµ at random
they are taken from that region where the dominant contributions to the integrals come from.
More precisely, we choose the points according to the Boltzmann distribution
Z
1 −SE (w)
P (w) = e , Z = e−SE (w) dn w (9.2)
Z
87
CHAPTER 9. SIMULATIONS 9.1. Markov Processes and Stochastic Matrices 88
such that the Monte Carlo estimate Ā for the average of A simply reduces to an arithmetic
average
M
1 X
hAi ∼ Ā = A(wµ ), (9.3)
M µ=1
where M is the total number of points (states) generated in the Monte Carlo simulation.
and normalized
X
W (s, s′ ) = 1. (9.5)
s′
Now one tries to construct a Markov process such that in the ’long-time’ limit n → ∞ the
configurations are distributed according to (9.2). The ’long-time’ behavior of the system is
determined by the limit of W n as n → ∞.
The set of stochastic matrices form a semigroup and every stochastic matrix transforms a
P
stochastic vector (a f -dimensional vector p with non-negative entries Ps obeying Ps = 1)
into a stochastic vector. The entry Ps of p is the probability to find the system in the state s.
————————————
A. Wipf, Path Integrals
CHAPTER 9. SIMULATIONS 9.1. Markov Processes and Stochastic Matrices 89
for a < 1. The stochastic matrices W n converges exponentially fast as n → ∞. Also note that
this matrix has the eigenvalue 1. Later we shall see that under certain assumptions on W the
W n always converge to a stochastic matrix with identical rows.
A second and less trivial example is the stochastic matrix for a system with 3 states,
1 1
a 2
(1 − a) 2
(1 − a)
W = 0 0 1 (9.9)
0 1 0
and
1 1
an 2
(1 − an ) 2
(1 − an )
Wn = 0
0 1 for odd n,
0 1 0
and which again possesses the eigenvalue 1. For a < 1 a stochastic vector p is mapped into
n p1 p1 p1 p1
p −→ p W −→ 0, p2 + , p3 + or 0, p3 + , p2 +
2 2 2 2
as n → ∞, depending on whether n is even or odd. Note that p W n approaches a periodic orbit
exponentially fast and hence the Markov process does not converge. As we shall see, the reason
for this lack of convergence is that the minima in all columns of W are zero.
A stochastic matrix has always the eigenvalue 1. The corresponding right-eigenvector is
(1, 1, . . . , 1)T . To find the left-eigenvector we consider the sequence
1 n−1
X
pn = p W j. (9.10)
n j=0
Since the set of stochastic vectors is compact this series must have a convergent subsequence
k −1
1 nX
p W j −→ P.
nk 0
————————————
A. Wipf, Path Integrals
CHAPTER 9. SIMULATIONS 9.1. Markov Processes and Stochastic Matrices 90
PW = P. (9.11)
and therefore any stochastic matrix W possesses at least one fixpoint P that is a eigenvector
with eigenvalue 1. Let us now assume that W possesses at least one column whose entries
are bigger or equal to a positive number δ which means that all states have a non-vanishing
probability for a transition in a certain state. Such W are called attractive stochastic matrices.
The W in the first example above is attractive whereas the second one is not. For an attractive
W all states s have a non-vanishing probability to end up in a given state s′ , that is W (s, s′ ) > 0
for all s. Now we prove that an attractive W is contractive on vectors of length 2. First we note
that for two real numbers p and p′ we have
|p − p′ | = p + p′ − 2 min(p, p′ ),
First we prove this statement for the difference of two cartesian basis vectors es , s = 1, . . . , f .
For that we apply the identity (9.12) to the stochastic vectors es W and es′ W , that is to the rows
s and s′ of W . For an attractive W we find for s 6= s′
X
kes W − es′ W k = 2 − 2 min {W (s, s′′ ), W (s′, s′′ )}
s′′
≤ 2 − 2δ = (1 − δ) kes − es′ k with 0 < δ < 1, (9.14)
| {z }
=2
and this proves that W is contractive of the difference vectors es − es′ . We used
————————————
A. Wipf, Path Integrals
CHAPTER 9. SIMULATIONS 9.1. Markov Processes and Stochastic Matrices 91
where s∗ belongs to the particular column of W the elements of which are bigger or equal to δ.
Now we prove the contraction property for all vectors ∆ in (9.13). Since
X X
△s − △s = k∆k = 2
s:△s ≥0 s:△s <0
X X
△s + △s = 0
s:△s ≥0 s:△s <0
we conclude that
X X
△s = 1 and △s = −1. (9.15)
△s ≥0 △s <0
To simplify our notation we denote the non-negative elements of ∆ by △s and the negative
elements by △s′ . Note that the index sets {s} and {s′ } have no common elements. Because of
(9.15) we have
X X X
k∆k = 2 = −2 △s △ s′ = − △s △s′ kes − es′ k, (9.16)
| {z }
=2
where we used the inequality (9.14). Comparing with (9.16) leads to the inequality
k∆ W k ≤ (1 − δ)k∆k (9.18)
which shows that W is contractive on vectors of the form (9.13). Since this inequality is linear
∆ we may drop the condition k∆k = 2. Hence W is contractive on all vectors the elements of
which add up to zero and in particular on differences of two stochastic vectors.
Iterating the inequality we obtain
k∆ W n k ≤ (1 − δ)n k∆ k. (9.19)
Now we apply this inequality to p − P, where P is the fixpoint in (9.11) and p is an arbitrary
stochastic vector. Since the elements of p − P add up to zero we conclude
n→∞
k(p − P)W n k = kp W n − Pk −→ 0
————————————
A. Wipf, Path Integrals
CHAPTER 9. SIMULATIONS 9.2. Detailed Balance, Metropolis Algorithm 92
or equivalently that
n→∞
p W n −→ P. (9.20)
For the stochastic vectors es the left hand side is just the row s of the limit lim W n = W eq
such that W eq has identical rows. This means that all elements in a column of W eq are equal,
similarly as in the example above.
W eq (s, s′ ) = n→∞
lim W n (s, s′ ) = Ps′ , (9.21)
where Ps′ is the element s′ of P. It follows that both the limit distribution P and matrix W eq
are unique. Else there would exist a second fixpoint P ′ of the Markov process with
X X X
Ps′′ = Ps′ W (s, s′) = n→∞
lim Ps′ W n (s, s′ ) = Ps′ Ps′ = Ps′ ,
s s s
The fixpoint condition for the equilibrium distribution P (w) takes the form
Z
P (w ′) = Dw P (w) W (w, w ′). (9.23)
————————————
A. Wipf, Path Integrals
CHAPTER 9. SIMULATIONS 9.2. Detailed Balance, Metropolis Algorithm 93
The detailed balance condition does not fix the stochastic process uniquely. Considerations
of computational efficiency dictates its specific form. The local Metropolis- and heatbath al-
gorithm are the most popular once since they can be applied to almost all statistical systems.
More recently nonlocal cluster algorithms are also used in cases where local algorithms fail
due to ’critical slowing down’. Unfortunately cluster algorithms are not known for lattice gauge
theories.
In the M ETROPOLIS algorithm one starts with a certain initial configuration. The better the
initial configuration is chosen the less computer time is needed. For example, for a statistical lat-
tice model at high temperature one should choose the variables at different lattice sites (having
the lattice approximation (6.21) in mind) randomly, since there is almost no correlation between
variables at neighboring sites (see the high temperature expansion). If we are simulating a quan-
tum mechanical system with a deep double well potential then (at low or zero temperature) one
better chooses a initial configuration which resembles an instanton. After preparing the initial
configuration one takes the ’first’ lattice variable w1 and changes it or leaves it according to the
following rules:
• If S decreases (P increases) then the variable at site 1 is set to the new value w1′ .
• If the action increases then a random number r with uniform distribution between 0 and
1 is generated and the variable is changed to w1′ only if exp(−∆S) > r. Otherwise the
lattice variable retains its previous value w1 .
• The other sites of the lattice are then probed in exactly the same way.
• When one arrives at the last lattice site then one has completed one Monte-Carlo sweep.
To test whether one is already near equilibrium one measures some expectation values as a
function of iteration time. After the measured values have settled (that is we are not too far from
the equilibrium distribution) then we may start ’measuring’ observables according to (9.3).
The Heat-bath algorithm is very similar as the Metropolis algorithm and I refer you to the
literature on Monte Carlo simulations for learning more about the various techniques to simulate
quantum- or spin system. For an introduction into Monte Carlo (MC) methods, numerical
studies and further references I refer you to the nice paper of Creutz and Freedman [30], the
text book of N EWMAN and BARKENNA [31] or the text book of B ERG [32].
————————————
A. Wipf, Path Integrals
CHAPTER 9. SIMULATIONS 9.2. Detailed Balance, Metropolis Algorithm 94
Since the energy decreases for the transitions|2i → |1i , |3i → |1i and|3i → |2i the correspond-
ing transition amplitudes are constant in the Metropolis algorithm. The inverse transitions cost
energy and the probabilities are proportional to the Boltzmann factors
bpq = eβ(Ep −Eq ) , p > q. (9.27)
We see that the Markov process has the stochastic matrix
2 − b21 − b31 b21 b31
1
W = 1 1 − b32 b32 , (9.28)
2
1 1 0
and its powers converge to
e−βE1 e−βE2 e−βE3
1 −βE1
W∞ e−βE2 e−βE3 ,
= e (9.29)
Z
e−βE1 e−βE2 e−βE3
where Z is the partition function, Z = exp(−βE1 ) + exp(−βE2 ) + exp(−βE3 ). Every initial
probability distribution converges to the Boltzmann distribution,
1 −βE1 −βE2 −βE3
P= e ,e ,e . (9.30)
Z
The following figure shows the convergence to equilibrium for different initial distributions. I
used the differences E2 − E1 = 0.5 and E3 − E2 = 0.3.
kalter
Start
W W W
p1 p2 p3 p1 p2 p3 p1 p2 p3 p1 p2 p3
warmer Start
For a cold start with ground state as initial state and a hot start with equal probabilities for the
different states we find good convergence to equilibrium. If we start with the most excited state
as initial state the the convergence is bad.
————————————
A. Wipf, Path Integrals
Chapter 10
Berezin Integral
In any field theory describing the elementary particles in nature there are bosonic and fermionic
fields. The latter describe the propagation of electrons, muons, neutrinos, quarks and so on. In
this chapter we introduce anticommuting Grassmann-variables and the Berezin integral [33].
These enter the path integral quantization of fermionic degrees of freedom.
√ √
! !
† 1 i 1 i
a =√ ωm q − √ p a= √ ωm q + √ p , (10.1)
2h̄ ωm 2h̄ ωm
[a, a† ] = 1. (10.2)
The creation and annihilation operators are represented on the anti-holomorphic functions f (z̄)
endowed with the scalar product
1
Z
(f1 , f2 ) ≡ f¯1 (z)f2 (z̄)e−z̄z dzdz̄, z = x + iy, dzdz̄ = 2idxdy. (10.3)
2πi
The normalization is such that the constant function f = 1 has unit norm. The creation- and
annihilation operators are represented as
∂
(af )(z̄) = f (z̄) and (a† f )(z̄) = z̄f (z̄). (10.4)
∂ z̄
95
CHAPTER 10. BEREZIN INTEGRAL 10.1. Grassmann variables 96
Using H = h̄ω(a† a + 1/2) and that the anti-holomorphic functions fn (z̄) = (n!)−1/2 z̄ n form
an orthonormal base in this space,
2
Z
2
(fm , fn ) = δmn e−r r 2n+1 dr = δmn , z = reiϕ ,
n!
we can calculate the matrix element hz ′ |e−itH/h̄ |zi of the evolution operator. One subtle point
in the bosonic case is the normal ordering. One starts with the normal ordered Hamiltonian,
that is the Hamiltonian with zero-point energy subtracted
In order to replace the operators by classical variables, H(a† , a) → h(z̄, z), one needs to normal
order the evolution operator :e−itH/h̄ : and not only the Hamiltonian. However, in the continuum
limit only the first order term 1 − iǫ:H(a† , a):/h̄ in the series expansion for the normal ordered
evolution operator contributes. But this term is assumed to be already normally ordered.
Now we turn to the fermions, that is we replace (10.2) by
As representation space we can choose the analytic functions depending on ᾱ only. Since
ᾱ2 = 0 such functions have a terminating series expansion
f (ᾱ) = f0 + f1 ᾱ.
G2 ≡ C ⊕ Λ1 (V ) ⊕ Λ2 (V )
and elements in G2 have the form f = f00 +f10 α +f01 ᾱ +f11 ᾱα. More generally, for n degrees
of freedom (10.6) generalizes to
Grassmann variables are nilpotent, αi2 = ᾱi2 = 0, and they generate the Grassmann algebra
Gn ≡ ⊕Λk (V ), k = 1, 2, . . . , 2n,
αi1 · · · αip ᾱj1 · · · ᾱjq with p + q = k and i1 < ... < ip , j1 < ... < jq
————————————
A. Wipf, Path Integrals
CHAPTER 10. BEREZIN INTEGRAL 10.1. Grassmann variables 97
Then the derivative is extended to polynomials and hence to all functions of the Grassmann
variables {αi , ᾱi }.
The fermionic creation and annihilation operators in (10.2) are represented by differential
operators acting on analytic functions f (ᾱ) = f (ᾱ1 , . . . , ᾱn ) as follows,
∂
(ai f )(ᾱ) = f (ᾱ) and (a†i f )(ᾱ) = ᾱi f (ᾱ) =⇒ [ai , a†j ] = δij 1. (10.9)
∂ ᾱi
We also would like to introduce a scalar product on the space of analytic functions f (ᾱ). For
that aim we introduce an integration over Grassmann variables. Such integrals have been intro-
duced by Berezin and they are defined by the following linear functional [33, 34]:
Z Z Z Z
dαi αj = dᾱi ᾱj = δij and dαi = dᾱi = 0. (10.10)
To integrate a monomial with respect to αi one first brings αi in the monomial to the left (using
the anti-commutation rules) and then drops this variable. For example,
Z
dαi αj αk = δij αk − δik αj , (10.11)
R
and similarly for higher monomials. We see that the Berezin integral dαi is equivalent to left
derivative with respect to ∂αi . For the integral over all Grassmann variables we choose the sign
convention such that
Z n
Y n
Y n
Y
DαD ᾱ (ᾱi αi ) = 1, where DαD ᾱ ∝ dαi dᾱi , (10.12)
1 1 1
and it is supposed that the dαi and dᾱj anticommute with each other and with αi and ᾱj . The
integral over Grassmann variables which are permutations of the α’s and ᾱ’s in (10.12) is then
given by the anti-commutation rules. The integral of less then 2n variables is always zero,
Z p
Y q
Y
DαD ᾱ αi ᾱj = 0 for p + q < 2n. (10.13)
1 1
From this property it follows that under a shift of the integration variables by Grassmann vari-
ables the Berezin integral is not changed,
Z Z
DαD ᾱ f (α + η, ᾱ + η̄) = DαD ᾱ f (α, ᾱ). (10.14)
————————————
A. Wipf, Path Integrals
CHAPTER 10. BEREZIN INTEGRAL 10.1. Grassmann variables 98
Actually, to prove this translational invariance one also uses (α + η)2 = αη + ηα = 0. Let us
now see how the Berezin integral changes under linear transformations
X X
βi = Uij αj and β̄i = Vij ᾱj (10.15)
j j
Note that only those terms contribute for which {j1 , . . . , jn } and {k1 , . . . , kn } are permutations
of {1, . . . , n}. These permutations are denoted by σ and σ̃. Thus we find
Z XY
... = Uiσ(i) Vℓσ̃(ℓ) sgn(σ)sgn(σ̃) = det U · det V. (10.16)
σ,σ̃ i,ℓ
For theories containing fermions the Gaussian Berezin integrals are as important as the ordinary
Gaussian integrals are for theories containing bosons. With the help of (10.16) it is not difficult
to compute the Gaussian integral
Z
Z= DαD ᾱ e−ᾱAα , where ᾱAα = ᾱi Aij αj . (10.17)
One just changes variables according to βi = Aij αj (and leaves the ᾱ’s) so that
1
Z Z Z
DαD ᾱ e−ᾱi βi = DαD ᾱ (βi ᾱi )n =
Y
Z= DαD ᾱ (βi ᾱi ) = det(A).
n!
We end up with the important formula
Z
DαD ᾱ e−ᾱAα = det(A), ᾱAα = ᾱi Aij αj . (10.18)
This should be compared with the corresponding bosonic Gaussian integral for which one ob-
tains the inverse square root of the determinant of A.
The generating function for Grassmann integrals can be computed by shifting the integration
variables in (10.18) according to
Using the translational invariance of the Berezin integral, see (10.14), one arrives at
Z
DαD ᾱ e−ᾱAα+η̄α+ᾱη = det(A) eη̄A
−1 η
, η̄α = η̄i αi . (10.19)
Now we define the scalar product of two analytic (in ᾱ) functions, similarly as in the bosonic
case, according to
Z
(g, f ) = DαD ᾱ g † (α)f (ᾱ)e−ᾱα , (10.20)
————————————
A. Wipf, Path Integrals
CHAPTER 10. BEREZIN INTEGRAL 10.1. Grassmann variables 99
where the adjoint of a function g = g0 + gi ᾱi + gij ᾱi ᾱj + . . . is given by g † = ḡ0 + ḡi αi +
ḡij αj αi + . . .. Inserting the expansions for g † and f yields
n
X
(g, f ) = ḡi1 ...ip fi1 ...ip (10.21)
p=0
for the scalar product of two functions g(ᾱ) and f (ᾱ). The last formula makes clear that the
scalar product is indeed sesqui-linear and positive as required. The space of analytic functions
f (ᾱ), endowed with this scalar product, forms the Hilbert space on which the linear operators
are represented. One can show that the operators a and a† are (formally) adjoint of each other
on this Hilbert space. A basis of the Hilbert space is defined by the orthonormal set of Fock
states a†i |0i, where|0i is represented by the constant function 1.
Q
A(ᾱ, β) = eᾱβ AN (ᾱ, β) with AN (ᾱ, β) = K00 + K01 β + K10 ᾱ + K11 ᾱβ.
This generalizes in an obvious way to more than one degree of freedom: a normally ordered
linear operator  has a kernel A which is obtained from  by replacing a, a† by β, ᾱ and
multiplying the resulting expression with exp(ᾱβ). Similarly one can show that
Z
(AB)(ᾱ, α) = A(ᾱ, β)B(β̄, α)e−β̄β dβ̄dβ.
With these formulas we can now derive the path integral representation for the kernel of the
normal ordered evolution operator K(t, â, ↠). As in the bosonic case we divide the time interval
[0, t] into n time steps of equal length ǫ = t/n and obtain for the kernel
Z n−1 n n−1 n
!
KǫN (ᾱi , αi−1 ) exp
Y Y X X
K(t, ᾱn , α0 ) = dᾱi dαi − ᾱi αi + ᾱi αi−1 , (10.24)
i=1 i=1 1 i=1
where the variables α0 and ᾱn at initial and final time are held fixed. In the continuum limit
n → ∞ or ǫ → 0 we may approximate
iǫ
KǫN (ᾱ, α) ∼ exp − H N (ᾱ, α)
h̄
————————————
A. Wipf, Path Integrals
CHAPTER 10. BEREZIN INTEGRAL 10.1. Grassmann variables 100
where one integrates over the Grassmann variables {ᾱi , ᾱi } with i = 1, 2, .., n − 1. The second
form follows from the first by a ’partial integration’ and shows, that the factors ᾱn αn and ᾱ0 α0
are surface terms which can be neglected in the continuum limit. Thus in the continuum limit
we end up with the following path integral
ᾱ(tZ
2 )=ᾱn
Zt2
h i
K(t, ᾱn , α0 ) = D ᾱDα exp − dt ᾱα̇ + iH N (ᾱ, α) . (10.25)
α(0)=α0 t1
Note that the function in the exponent is just the action corresponding to the (normal ordered)
Hamiltonian H. This means that the path integral for fermionic degrees of freedom is formally
the same as for bosonic systems. The crucial difference (which forbids a probabilistic interpre-
tation) is the replacement of c-numbers by ’Grassmann numbers’. Before turning to the field-
theoretical generalization we discuss an interesting application of (10.25) to supersymmetric
quantum mechanics.
————————————
A. Wipf, Path Integrals
Chapter 11
and a and b are bosonic and fermionic annihilation operators: [a, a† ] = 1 and {b, b† } = 1. The
Fockspace is generated by acting with the creation operators on the vacuum defined by
Using the commutation and anticommutation relations for the creation and annihilation oper-
ators one finds that besides the non-degenerate zero-energy ground state all excited states are
double degenerate since (a† )n |0i and b† (a† )n−1 |0i have both energy E = nω. Introducing the
fermion number operator NF = b† b we see that there is always a bosonic state (NF = 0) and
a fermionic one (NF = 1) with the same energy. This system is the simplest supersymmetric
quantum mechanical system, namely the supersymmetric harmonic oscillator (we have set the
mass to one and shall also set h̄ = 1 in what follows).
Let us now generalize the above Hamiltonian and consider
1 2
H = HB + HF , where HB = p + W2 and HF = W ′ b† b, (11.3)
2
101
CHAPTER 11. SUPERSYMMETRIC QUANTUM MECHANICS 102
where W (x) is an arbitrary function. Using the formula (10.25) and the corresponding bosonic
result (2.32) yields the following path integral representation for the evolution kernel
Z
K(t, q, q ′, ᾱ, α′) = DwDαD ᾱ eiS[w,α,ᾱ] , (11.4)
where one sums over all paths w(t), α(t), ᾱ(t) with
The action contains the familiar bosonic part SB and an additional term depending on the Grass-
mann values path,
Z
S= dt L = SB [w] + SF [w, α],
with constant anticommuting parameters ǫ, ǭ, the variation of the Lagrange function is a total
time-derivative,
d
δL = ẇ ᾱǫ − iW ǭα (11.7)
dt
and thus the action is invariant.
It has been observed by Nicolai [40] that the following transformation of the bosonic field
for which
!
1 2 1 2 1 2 δy(t) d
y = ẇ − W + iW ẇ and = + iW ′ δ(t − t′ ) (11.9)
2 2 2 δw(t′) dt
simplifies the analysis considerable, due to supersymmetry. To see that we first note that
!
d
Z
−(ᾱα̇+iW ′ ᾱα)
Dw DαD ᾱ e = Dw det + iW ′ = Dy, (11.10)
dt
which means that the Jacobian of the bosonic transformation is canceled by the fermionic inte-
gral. We have been a bit sloppy with the boundary conditions, for a more detailed analysis of
this point I refer you to the paper of Ezawa and Klauder [41]. Second we observe that
1Z 2 1Z 2 Z Z q
y dt = ẇ − W 2 dt + i W ẇdt = SB + i W (w)dw. (11.11)
2 2 q′
————————————
A. Wipf, Path Integrals
CHAPTER 11. SUPERSYMMETRIC QUANTUM MECHANICS 103
Inserting the last two identities into the evolution kernel (11.4) we see that this kernel is given
by a Gaussian integral in terms of the new variables,
Z q Z
2 /2
K(t, . . .) = exp W Dy eiy . (11.12)
q′
To obtain the partition function we continue to imaginary time t = −iτ such that the action
changes into the Euclidean action
1 1
Z
SE = dτ L with L = ẇ 2 + W 2 + ᾱα̇ + W ′ ᾱα, (11.13)
2 2
and the supersymmetric transformations are modified to
d
δα = (ẇ − W )ǫ δ ᾱ = ǭ(−ẇ − W ) with δL = (ẇ ᾱǫ − ǭαW ) . (11.14)
dτ
Note that the transformation of w is unchanged. The Nicolai map of the Euclidean model reads
To obtain the ’partition function’ one integrates over β-periodic paths w(τ ) and β-antiperiodic
paths ᾱ(τ ), α(τ ) (see below). Such finite temperature boundary conditions break supersymme-
try which transforms periodic bosonic fields into periodic fermionic fields. Physically this is
not surprising since a equilibrium state is not invariant under Lorentz transformation and hence
cannot be supersymmetric. After all, supersymmetry is an extension of Lorentzsymmetry. If we
instead integrate only over periodic paths then supersymmetry is not violated by the boundary
condition. This corresponds to the Euclidean model and for β → ∞ expectation values become
vacuum expectation values. For periodic boundary conditions we can transform to Nicolai vari-
ables and obtain
I
2 /2
Zper = Dy e−y , (11.16)
where one integrates over β-periodic paths y(τ ). When treating the boundary conditions more
carefully one can indeed show that for matrix elements the cancellation between the fermionic
determinant and the bosonic Jacobian occurs for a certain definition of the fermionic path inte-
gral and thus the above formal manipulations are justified.
Note that the correlation functions can now be evaluated as
Z
hw(τ1 ) . . . w(τn )i = w(τ1 ) . . . w(τn )dµ0 (y)
with the Gaussian measure as in (11.16) rather than with the complicated interaction measure.
However, the moments are not that easy to calculate because w(t) is generally a nonlinear and
nonlocal function of the fluctuating y-path as determined by the inverse Nicolai map.
————————————
A. Wipf, Path Integrals
Chapter 12
After introducing path integrals in quantum mechanics we now turn to the path integral rep-
resentation of field theories. In this chapter we discuss the fermionic sector of the Schwinger
model, which is probably the simplest non-trivial field theory. The Schwinger model is just
QED for massless fermions in 2 dimensions [42]. This model shows at least two (related)
striking features. First the classically massless ’photon’ acquires a mass due to its interaction
with the massless fermions and second the operator ψ̄(x)ψ(x) has a non-vanishing vacuum
expectation [43]. Clearly, since this model contains fermions we first must discuss the path
integral for fermionic, and in particular the path integral representation of the n-point functions.
The zero-temperature Schwinger model has been solved some time ago by using operator
methods [44] and more recently in the path integral formulation [45]. Some properties of the
model (e.g. the non-trivial vacuum structure) are more transparent in the operator approach
and others (e.g. the role of the chiral anomaly) are better seen in the path integral approach.
More recently the Schwinger model has been solved in the path integral approach on the 2-
dimensional sphere and the role of the fermionic zero modes has been emphasized [46].
The discrete index i becomes the continuous position in space and the summation is to be
replaced by an integration over space.
For the Dirac fermions minimally coupled to a gaute field Aµ the action reads
Z
S= L / − m)ψ.
with L = ψ̄(iD (12.1)
Ω
104
CHAPTER 12. FERMION FIELDS 12.1. Dirac fermions 105
Inserting this into the field-theoretical generalization of (10.25) we obtain the functional integral
representation
Z
Z= DψD ψ̄ eiS[ψ̄,ψ] , (12.4)
where S is the action for fields on a space-time region Ω. The boundary conditions for the fields
on the boundary ∂Ω must be specified. Here we choose for Ω the Minkowski space to avoid
boundary effects.
We are primarily interested in the generating functional in the presence of external currents,
which now is constructed by using two anticommuting sources η̄(x) and η(x):
Z Z
Z[η̄, η] = DψD ψ̄ exp iS[ψ̄, ψ] + i [η̄(x)ψ(x) + ψ̄(x)η(x)]dd x . (12.5)
We can simplify this path integral by expanding the exponent about its extremum. The exponent
is extreme for
/ − m)−1 η
ψcl = −(iD and / − m)−1 .
ψ̄cl = −η̄(iD
————————————
A. Wipf, Path Integrals
CHAPTER 12. FERMION FIELDS 12.1. Dirac fermions 106
Since Scl is independent of the integration variables, the path integral (12.5) reads
Z
/ − m) exp − i
Z[η̄, η] = det(iD η̄(x)GF (x, y)η(y)ddxdd y . (12.10)
Differentiating (12.5) with respect to the sources η and η̄ yields the correlation function
To prove this identity we use the generating functional (12.5) and expand the exponent contain-
ing the source-terms in a power series:
2n
i2n
Z Z
DψD ψ̄ eiS
X Y
Z[η̄, η] = dx1 ...dx2n (η̄(xi )ψ(xi ) + ψ̄(xi )η(xi ))
n (2n)! i=1
2n
(−)n Z Y (2n)! Z
dxi η̄ α1 (x1 )...η βn (x2n ) DψD ψ̄ eiS ψα1 (x1 )...ψ̄βn (x2n ),
X
= (12.13)
n (2n)! 1 n!n!
where we have used the anticommutation properties of the fields and sources and the fact that
the functional integral is nonzero only if there are as many fields as adjoint fields. On the other
hand using (12.10) we may expand the generating functional as
n
Z[η̄, η] X (−i)n
Z Y
= dx1 ...dxn dy1...dyn η̄(x1 )η(y1 )...η(yn ) GF (xi , yi ) (12.14)
Z[0] n n! i=1
————————————
A. Wipf, Path Integrals
CHAPTER 12. FERMION FIELDS 12.1. Dirac fermions 107
Comparing with (12.13) and using the fact that the sources are arbitrary, proves the Wick theo-
rem (12.12).
Finally we turn to the fermionic thermal Green’s functions. As we have already seen in in
quantum mechanics, the transition to the Euclidean sector is made by replacing t → −iτ such
that
(and keeping the other quantities fixed) or equivalently by replacing xj → ixj such that
Since we prefer to use a Minkowskian metric with signature (+, −, −, −) we continue accord-
ing to (12.17) rather than (12.16). In the case of Dirac fermions the exponent in (12.5) becomes
then
Z Z
iS + i (η̄ψ + ψ̄η) −→ −SE + (η̄ψ + ψ̄η),
Z
SE = LE , / + mψ̄ψ.
LE = −iψ̄ Dψ (12.18)
When calculating the partition function Z(β) at finite temperature we must choose antiperiodic
boundary conditions for the fields, in contrast to to the bosonic case (see (8.22)). The reason is
that the fermionic Green’s functions are β-periodic in imaginary time [48]. This is taken into
account if antiperiodic boundary conditions in the path integral are chosen and then the partition
function becomes
Z
Z(β) = const · DψD ψ̄ e−SE [ψ̄,ψ] , (12.19)
a.p.
where a.p. should indicate that we integrate over anti-periodic fields ψ(h̄β, x ) = −ψ(0, x ) and
analog for ψ̄. In analogy to (12.5) the generating functional for the thermal Green’s functions
reads
Z Zβh̄
Z[β, η̄, η] = DψD ψ̄ exp − S[ψ̄, ψ] + dd x [η̄(x)ψ(x) + ψ̄(x)η(x)] (12.20)
a.p. 0
and the thermal correlation functions are obtained by differentiation with respect to the external
current
(−)n δ 2n
T h0|ψα1 (x1 )ψ̄β1 (y1 )...ψ̄βn (yn )|0iβ = Z[η̄, η]|η̄=η=0 (12.21)
Z[0] δη βn (yn )...δ α1 η̄(x1 )
where T denotes the Euclidean time ordering. Note the presence of the factor (−1)n in contrast
to (12.11). This is due to the Wick rotation to imaginary time.
————————————
A. Wipf, Path Integrals
CHAPTER 12. FERMION FIELDS 12.2. The index theorem for the Dirac operator 108
Next we simplify (12.21). We could calculate a ’classical’ path with antiperiodic boundary
conditions, calculate the partition kernel and then integrate over the boundary conditions. This
approach analogous to is rather involved in the present situation. Therefore we choose a some-
what different (and more formal) approach which can be applied for quadratic actions (and for
simple boundary conditions). We just apply the Gauss integral formula to (12.20)
1
/ − m) e−(η̄(x),Gβ (x,y)η(y)) ,
Z[β, η̄, η] = det(iD Gβ (x, y) = hx| |yi. (12.22)
/
iD − m
/ − m) (that is the Green’s function on the space
Gβ (x, y) is the thermal Green’s function of (iD
of the functions antiperiodic in β). The formula (12.22) then implies in particular
1 δ2
T hψ(x)ψ̄(y)iβ = − Z[β, η̄, η]|η̄=η=0 = Gβ (x, y). (12.23)
Z[β, 0] δη(y)δ η̄(x)
More generally, the Wick-theorem (12.12) still holds if we drop the (−i)n and replace the
Feynman propagator by the thermal Green’s function (or Euclidean propagator) on the right
hand side of (12.12). This concludes our proof of the equivalence between the functional inte-
gral approach and the canonical approach for fermionic systems. We have seen that formally
there is a close analogy of fermionic path integrals with those in quantum mechanics. So far
we haven’t dealt with the inherent divergences of field theories, a feature with is not present
in ordinary quantum mechanics. Finally we have seen the path integral formalism allows for a
unified treatment of zero-temperature and finite temperature systems.
/
Γ = log Z = log det D (12.24)
(see (12.22)), where we assume the fermions to be massless. As we shall see later, this deter-
minant can be calculated explicitly in 2 dimensions by integrating the chiral anomaly. As a first
step we now determine the number of zero modes of D. / It will turn out that this number is a
physically and mathematically interesting number.
We use the notation and convention as in (8.67-8.70) and assume that space-time is even
dimensional so that we can introduce γ5 = (−i)n(n−1)/2 γ 1 γ 1 · · · γ n (the factor is chosen such
that γ52 = Id) which anti-commutes with all γ’s
{γ5 , γ µ } = 0 =⇒ {γ5, D} / 2 ] = 0.
/ = [γ5 , D (12.25)
————————————
A. Wipf, Path Integrals
CHAPTER 12. FERMION FIELDS 12.2. The index theorem for the Dirac operator 109
/ = hχ =⇒ iD(γ
iDχ / 5 χ) = −γ5 (iDχ)
/ = −h(γ5 χ) (12.26)
i.e. the γ5 -transform of an eigenmode has the opposite eigenvalue (note that γ5 χ has the same
norm as χ and hence cannot be zero). This implies that all excited states of −D / 2 are (at least)
double degenerate, more precisely to each left-handed eigenmode γ5 ψL = ψL there is a right-
handed partner γ5 ψR = −ψR with the same eigenvalue E = λ2 . In terms of the eigenfunctions
of iD/ they read ψL = 21 (1 + γ5 )χ and ψR = 21 (1 − γ5 )χ. This pairing need not and generally
does not occur for the zero-energy states. The ground states of −D / 2 are also eigenstates of iD /
with eigenvalue zero (this is not true for the excited states) and thus have fixed chirality. Now
we define the index of the Dirac operator as the number of left-handed minus the number of
right-handed zero modes of −D / 2 or iD:
/
/ = n+ − n− .
index (iD) (12.27)
/ 2)
This index can be computed quite differently. For that we note that the (super) trace tr γ5 exp(β D
can be computed via path integrals similarly to the partition function in (8.3) and (8.4). Using
/ 2 in evaluating the trace we find
the eigenfunction of −D
2 X
tr γ5 eβ D/ = e−βEL,n − e−βER,n = n+ − n− = index (iD),
/ (12.28)
n
where we have used that due to the pairing of the excited states only the zero-modes contribute
to the sum. Note in particular that the super-trace is β-independent.
The supertrace can now be calculated by using the density (8.65) of the partition function.
This way we find for the index
2
Z
n+ − n− = tr γ5 eβ D/ = dn x tr γ5 Z(β, x) (12.29)
where the last trace is over spin- and internal color indices and Z(β, x) possesses the path
integral representation (8.74). Since the super-trace is independent of β we may assume β to be
very small and use the high temperature expansion (8.78).
In two dimensions ΣF = γ5 F01 and we find
1
Z Z
tr γ5 Z(β, x) = tr (γ52 )F01 + O(β). (12.30)
4π
With our sign convention for γ5 we obtain in four dimensions tr (γ5 Σµν Σαβ ) = ǫµναβ and thus
1 1
Z Z
tr γ5 Z(β, x) = ǫµναβ Fµν Fαβ + O(β). (12.31)
(4π)2 2
————————————
A. Wipf, Path Integrals
CHAPTER 12. FERMION FIELDS 12.3. The Schwinger model, Part I 110
In a first step we treat the fermionic part of the path integral only and thus may assume the
photon field to be an external field. Integrating out the fermionic degrees of freedom according
to (12.22) yields
Z R R R
/ A )e− LB +i η̄(x)G(x,y)η(y)+ Aµ J µ
Z[J, η̄, η] = DA det(iD . (12.36)
The reason which allows the model to be solved exactly is that the electron propagator and the
fermionic determinant in an arbitrary external field can be found explicitly, as has been observed
by Schwinger. For that purpose we introduce
1
Φ= ∂α Aα + iΣαβ Fαβ (12.37)
∆
————————————
A. Wipf, Path Integrals
CHAPTER 12. FERMION FIELDS 12.3. The Schwinger model, Part I 111
where the matrices Σαβ have been introduced in (8.71). Note that under a gauge transformation
A → A + dΛ this functions transforms as Φ → Φ + Λ. Using the identity 2iΣαβ = γ α γ β − δ αβ
one sees that
1 µ α β 1
/ =
∂Φ γ γ γ ∂µ ∂α Aβ = γ β ∆Aβ = γ β Aβ . (12.38)
∆ ∆
In particular in 2 dimensions ΣF = γ5 F01 and
1
Φ= ∂A + iγ5 F01 . (12.39)
∆
Taking into account that γ5 anti-commutes with the Dirac operator we can rewrite the Dirac
operator as
†
/ = ∂/ − iγ µ Aµ = ∂/ − i∂Φ
D / = eiΦ ∂e
/ −iΦ . (12.40)
Note, that the last identity holds only in 2 dimensions since we have used that γ µ Φ = Φ† γ µ .
Now it is clear that the exact propagator which obeys the equation
/
iDG(x, y, A) = δ 2 (x − y) (12.41)
————————————
A. Wipf, Path Integrals
CHAPTER 12. FERMION FIELDS 12.3. The Schwinger model, Part I 112
such that
/ α = i(Φ† D
δD /α − D
/ α Φ). (12.47)
The variation of the zeta-function becomes then (we suppress the index α)
1 2i
Z 2
Z 2
δζ−D/ 2 (s) = ts−1 tr etD/ t(δ D
/D/ +D / δ D)
/ = ts tr etD/ D/ 2 (Φ† − Φ)
Γ(s) Γ(s)
4 1 4s 1
Z 2
Z 2
= ts tr etD/ D/ 2 γ5 F01 = − ts−1 tr etD/ γ5 F01 ,
Γ(s) ∆ Γ(s) ∆
where we have partially integrated to obtain the last equality. Since finally s → 0 only the
singular part (more precisely the single pole at s = 0) of the integral survives, because of the
factor s. We may split the integration over t into an integration from 0 to ǫ and from ǫ to ∞.
The second integral is finite for s = 0 (recall that the free heat kernel falls off like t−1 ) and we
need only consider the interval near 0. Here we may use the asymptotic expansion (8.78) for
/ 2α and find (Fα = αF ):
the heat kernel of D
Zǫ
s 1 1 i
δζ−D/ 2 (s) = − dt[ts−2 tr (γ5 F01 ) + ts−1 αtr (γ5 F01 γ5 F01 ) + O(ts ) . (12.48)
πΓ(s) ∆ ∆
0
Since the first term in (12.39) vanishes the integration over t yields
2α s 1
− ǫ F01 F01 + O(s) .
πΓ(s) ∆
e2 1 e2 δµν − ∂µ∂ν ν
Z Z
log det(iD) / =
/ − log det(i∂) F01 F01 = − Aµ ( )A , (12.50)
2π ∆ 2π ∆
where we have reinserted the electric charge e. In what follows we may drop the (divergent) de-
terminant of the free Dirac operator, since it is independent of the gauge potential and chancels
in expectation values. Hence the effective action in (12.36) entering the integration over the
remaining ’photon’-field is
1 e2 ∂µ∂ν ν
Z Z
µν
Γ[A] = Fµν F + Aµ (δµν −
)A . (12.51)
4 2π ∆
√
This effective action belongs to a free particle with mass e/ π. We see that, due to the in-
teraction of the photons with the electrons, the classically massless photons acquire a mass.
————————————
A. Wipf, Path Integrals
CHAPTER 12. FERMION FIELDS 12.3. The Schwinger model, Part I 113
This so-called Schwinger mechanism happens without gauge symmetry breaking. We see that
the statement that a mass-term for the photon field breaks the gauge symmetry is not true in
general, in particular if we allow non-local interactions like in (12.51). Note, however, that
in the Lorentz gauge ∂A = 0 the effective action becomes local in the gauge-potential. This
observation will simplify the remaining path integral considerably.
Let us now discuss two consequences of (12.51). From
Z R
/
(iψ̄ ∂ψ+A µ
/ = µ ψ̄γ ψ)
det(iD) DψD ψ̄ e (12.52)
we see that
1 δ e2 ∂ ∂
j µ = hψ̄γ µ ψi = / = − (δµν − µ ν )Aν
det(iD) (12.53)
/ δAµ
det(iD) π ∆
and hence
∂µ j µ = 0, (12.54)
that is that the vector current is conserved. This is just a consequence of the gauge invariance
of the effective action (or the gauge invariant zeta-function regularization). Using the identity
γ5 γ µ = iǫµν γ ν , valid in 2 dimensions, we can also calculate the axial current:
e2 ∂ν ∂α
j5µ = hψ̄γ5 γ µ ψi = iǫµν j ν = −i (ǫµα − ǫµν )Aα . (12.55)
π ∆
Hence we find
e2 e2
∂µ j5µ = −ih̄ ǫµν ∂µ Aν = −ih̄ ǫµν Fµν , (12.56)
π 2π
and thus the axial current, contrary to the vector current, is not conserved. We have reinserted
h̄ in order to see that this non-conservation is a quantum effect. Classically the axial current is
conserved since it is the Noether current belonging to the chiral transformation
ψ −→ eαγ5 ψ and ψ̄ −→ ψ̄ eαγ5 (12.57)
which in any even dimension (for which γ5 exists) leave the classical action invariant since γ5
anti-commutes with the Dirac operator. What we have shown then is that a classically con-
served current is not anymore conserved after quantization or that the classical axial symmetry
is broken due to quantum effects. Such a phenomena is called an anomaly.
Let us now return to the problem of computing the correlation functions of the Schwinger
model. We begin with the representation for the 2-point function
1 1
Z Z
hψ(x)ψ̄(y)i = DA e−Γ[A] G(x, y) = DA e−Γ+i[Φ(x)−Φ(y)] G0 (x, y) (12.58)
Z(0) Z(0)
which only involves a Gaussian integral over the photon field. The same is of course true for the
higher correlation functions. To proceed we must first study how one evaluates path integrals
over gauge potentials. Due to the gauge invariance of the action we have to extend the path
integral to systems subject to constraints (coming from the gauge-invariance).
————————————
A. Wipf, Path Integrals
Chapter 13
Constrained systems
In this section the implementation of constraints within the path integral formalism is discussed.
The study of constraints is quantum mechanics is subtle and significant, since constraints are
closely related to symmetries. All gauge theories are systems with constraints and conversely:
all systems with first class constraints are gauge theories.
We shall see that, with some important adjustments to the measure, the path integral quan-
tization for constrained system is very similar to the previously discussed path integral for un-
constrained systems.
In a classical mechanical system whose phase space consists of 2n degrees of freedom
{q , . . . , q n , p1 , . . . , pn }, a constraint consists of some relation between the coordinates. To
1
illustrate what may happen in such cases, we first study a simple mechanical system for two
point-’particles’, confined to a line and governed by a Hamiltonian
p21 p2
H= + 2 + V (q 1 − q 2 ). (13.1)
2m1 2m2
Since the interaction depends only on the distance of the two particles the total momentum is
conserved
d i
P = [H, P ] = 0. (13.2)
dt h̄
After a canonical transformation to the center of mass Q and the relative coordinate q,
m1 1 m2 2
Q= q + q , P = p1 + p2
M M
m2 m1
q = q1 − q2 , p= p1 − p2 (13.3)
M M
the inverse transformation of which reads
m2 m1
q1 = Q + q , p1 = P + p,
M M
m1 m2
q2 = Q − q , p2 = P − p, (13.4)
M M
114
CHAPTER 13. CONSTRAINED SYSTEMS 115
where M = m1 + m2 is the total mass of the system, the Hamiltonian takes the form
P2 p2 P2
H= + + V (q) = + HCM (p, q). (13.5)
2M 2µ 2M
We have introduced the reduced mass 1/µ = 1/m1 + 1/m2 . H does not depend on the position
Q of the center of mass and that is why it commutes with the total momentum. Since P is
conserved it can be simultaneously diagonalized with the Hamiltonian
h̄
Pψ = ∂Q ψ =⇒ ψ = eiP Q/h̄ ψ(q). (13.6)
i
Let us assume we would like to describe a system with P = p1 + p2 = 0. We cannot demand
this as an operator identity, since this would imply
or that the commutation relations are violated. However, we can enforce the constraint P = 0
on the physical states,
which is a consequence of demanding that physical states have a sharp value of P (which is
conjugate to Q). The solution to this problem is that we should not normalize with respect to Q.
However one should keep in mind that the physical states are not normalizable, else one could
run into formal contradictions as
Now we wish to implement the constraint into the path integral. For doing that it is convenient
to use the phase-space formulation of the path integral. This is similarly derived as the path
integral (2.29) in the coordinate space. One first introduces the eigenstates of the position and
momentum operators:
————————————
A. Wipf, Path Integrals
CHAPTER 13. CONSTRAINED SYSTEMS 116
Now we proceed as in the coordinate space and write the evolution kernel as (with the same
conventions as in (2.27), e.g. τ = it/h̄)
Z n−1
hqj+1 |e−itT /n e−itV /n |qj i.
Y
K(t, q ′ , q) = hq ′ |e−itH |qi = dq1 . . . dqn−1 (13.12)
j=0
hqj+1|pj i e−it/n(T (pj )+V (qj )) hpj |qj i = eipj (qj+1 −qj )/h̄−it/n H(pj ,qj ) .
where T (pj ) and V (qj ) are the values of the kinetic and potential energy in the momentum and
position eigenstates, respectively. Hence their sum H(pj , qj ) is just the classical energy of a
particle with momentum pj at position qj . If T also depends on the coordinate this is still true if
it is understood that the kinetic energy is normally ordered, that is the momentum on the left and
the coordinates on the right. When rewriting each factor this way and reinserting h̄ we finally
end up with
qn ′
Z=q n−1 h i n−1
Y dqi dpi X i
K(t, q ′ , q) = lim exp {pj (qj+1 − qj ) − ǫH(pj , qj )} (13.13)
n→∞
q0 =q 1 2πh̄ h̄ 1
dpj hi p2j i m n im
Z r o
exp (pj (qj+1 − qj ) − ǫ ) = exp (qj+1 − qj )
2πh̄ h̄ 2m 2πih̄ǫ 2h̄ǫ
————————————
A. Wipf, Path Integrals
CHAPTER 13. CONSTRAINED SYSTEMS 117
and thus we recover the representation (2.29) for the path integral in coordinate space,
m n/2
Z
K(t, q ′ , q) = lim dq1 · · · dqn−1
n→∞ 2πih̄ǫ
j=n−1
n iǫ X h m q
j+1 − qj 2
io
exp ( ) − V (qj ) . (13.15)
h̄ j=0 2 ǫ
Now we would like to express the evolution kernel for the system (13.5) subject to the constraint
that the total momentum vanishes, in the full phase space. Clearly, on the physical subspace we
have
Z
i
(Pj (Qj+1 −Qj )+pj (qj+1 −qj )−ǫH(Pj , pj , qj ))
Y P
DqDpDQDP δ(Pj )e h̄
diverges. This can be remedied by inserting another delta-function in the variables Qj conjugate
to the constraint, setting them to arbitrary constants Yj . Since the Jacobi Matrix of the canonical
transformation (13.3) has determinant one and since Pj (Qj+1 − Qj ) + pj (qj+1 − qj ) transforms
into the same expression with (P, Q, p, q) → (p1 , q 1 , p2 , q 2 ) we find
Z ni Z o
K(t, q ′ , q) = const · Dq i Dpi δ(P )δ(Q − Y ) exp (pi q̇ i − H[pi , q i ]) ,
h̄
where we have taken the continuum limit such that δ(Pj ) → δ(P (t)) ≡ δ(P ) and similarly
Q
for δ(Qj − Yj ). If it is not clear how to identify the variable conjugate to the constraint we
Q
————————————
A. Wipf, Path Integrals
CHAPTER 13. CONSTRAINED SYSTEMS 118
and thus we arrive at Faddeev’s formula for the functional integral on the full 2-body phase
space appropriate to a constrained quantum system [50]
Z δF (q i (t)) ni Z o
i
′
K(t, q , q) = Dq Dpi δ(P )δ(F ) det exp (pi q̇ i − H(pi , q i )
δQ(t′ ) h̄
Z ni Z o
= Dq i Dpi δ(P )δ(F ) det{F, P } exp (pi q̇ i − H(pi , q i ) . (13.19)
h̄
The first delta function enforces the constraint. Since the second one involves an arbitrary
function it is called a choice of gauge. It follows from our derivation that the path integral is
unaffected by a different choice of the auxiliary condition F (q i ) = 0. Note that the exponent is
just the classical action in terms of the canonical variables.
The expression has the following geometric interpretation: The constraint P = 0 defines a
3-dimensional sub-manifold C (in our simple example it is just a plane, since the constraint is
linear) of the 4-dimensional phase space. However, the constraint also generates a Hamiltonian
flow,
and J is the symplectic matrix, J = iσ2 ⊗ Id. Since Ṗ = {P, P } = 0, this is a flow on C.
Furthermore, from (13.2) we see that H is constant on the lines of flow in C. Now we can
identify two points if and only if they belong to the same trajectory of the flow (13.20). This
defines an equivalence relation which is independent of the choice of the constraint (we could
have taken an equivalent constraint a(p, q) · P = 0, where a(p, q) possesses no zeroes, instead
of P = 0) and is invariant under the time evolution. All observables commute (weakly) with
the constraint and thus are constant under the flow generated by the constraint. We see that
the constraint generates a (gauge) symmetry of the system. It is thus sufficient to choose a
representative in each equivalence class in a regular manner. The regularity condition means
that one chooses a submanifold of C by fixing a gauge F = 0 such that each flow trajectory
intersects this sub-manifold exactly once. Locally this is equivalent to demanding that the flow
generated by the constraint is never parallel to the gauge-fixing surface F = 0, or that the inner
product of the vector XP generating the flow and the gradient vector ∇F orthogonal to the
gauge fixing surface in C does not vanish
In particular, if one chooses for F the variable conjugate to the constraint, then these vectors
are parallel and the gauge fixing surface is orthogonal to the flow trajectories.
The described procedure can be generalized to a set of m independent first class constraints
in a 2n-dimensional phase space, that is a set of constraints
Cj (pi , q i ) = 0, j = 1, . . . , m, (13.22)
————————————
A. Wipf, Path Integrals
CHAPTER 13. CONSTRAINED SYSTEMS 119
which form a closed algebra and weakly commute with the Hamiltonian,
They define a 2n − m dimensional submanifold C of the phase space. The flows generated
by these constraints stay entirely in C and are symmetries of the system. Again one chooses a
regular gauge
which defines a 2(n − m) dimensional sub-space of the full phase space which may in turn
be considered as a phase space. Similar considerations as above lead to the same path integral
representation for K(t, q ′ , q) as given by (13.19), where now δ(P ) is replaced by δ(Ci ) and
Q
δ(F ) by δ(Fi ).
Q
————————————
A. Wipf, Path Integrals
Chapter 14
All fundamental theories in particle physics are gauge theories. These theories contain first class
constraints which generate the (time-independent) gauge transformations and hence must be
quantized along the lines outlined above. We shall first recall the classical canonical structure of
pure Yang-Mills theories with particular emphasis on the constraints. At the end we specialize
to the Abelian case and set some of the potentials and field strengths to zero to recover the path
integral for the Schwinger model.
dim
XG dim
XG
Aµ = Aµa Ta , F µν = Faµν Ta ,
a=1 a=1
where the (hermitian) generators Ta of the Lie algebra obey the commutation relations
120
CHAPTER 14. GAUGE FIELDS 14.1. Classical Yang-Mills Theories 121
with totally anti-symmetric and real structure constants fabc , we find the following formulae for
the components in group-space,
d 1
Ea = Aa − ∇A0a + fabc A0b Ac , Ba = −∇ × Aa − fabc Ab × Ac . (14.4)
dt 2
We have set A = (A1 , A2 , A3 ), E = (E1 , E2 , E3 ) and B = (B1 , B2 , B3 ). One would have the
usual sign convention [51] if one would take A = (A1 , A2 , A3 ) that is replace A by −A. In the
non-covariant notation the Lagrangian reads
1X 2
L= (Ea − Ba2 ). (14.5)
2 a
The non-covariant form of the Yang-Mills equations Dν F µν are the generalized Gauss- and
Ampere law
D · E = 0 ⇐⇒ ∇ · Ea + fabc Ab · Ec = 0
Dt E = (D × B) ⇐⇒ ∂t Ea + fabc A0b Ec = ∇ × Ba + fabc (Ab × Bc ). (14.6)
The corresponding identities in two dimensions for F01 = E are obtained by setting E =
(E, 0, 0), B = 0 and A2 = A3 = 0 in the above equations.
The canonical equal time commutation relations read (here we do not distinguish between upper
and lower indices, in particular Ai = Ai )
{Aia (t, x), Ebj (t, y)} = δab δij δ 3 (x − y), (14.11)
and hence the Hamiltonian equations reproduce Ampere’s law (14.8) and the definition of Ea
in terms of Ȧa . However, Gauss’s law has yet not emerged, since it is a fixed-time constraint
between canonical variables.
To understand the role of the Gauss constraints
more clearly let us calculate the commutator of these constraints with the canonical variables.
One finds
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.1. Classical Yang-Mills Theories 123
Now we shall see, that the constraints generate the time-independent gauge transformations
The corresponding small transformations of the gauge potential and field strengths are
which are identical with the corresponding commutation relations in (14.18,14.19) with the
smeared constraint Cθ . Hence the Gauss-constraints generate the time-independent gauge trans-
formations.
It follows then that the Hamiltonian commutes with the constraints since it is gauge invari-
ant. Finally, using the identity
one shows that the commutator of two different constraints follow the Lie algebra of the gauge
group,
and thus form a system of first class constraints. The transition from the classical Poisson
bracket to the corresponding commutators is as usual achieved by replacing Poisson brackets
{., .} by commutators −i[., .]/h̄ in the above relations.
The path integral for the Yang-Mills Hamiltonian (14.10) is given by analogy with the con-
strained quantum mechanical system (13.19) by
Z hi Z 1 1 i
Z= DEa DAa δ(Ca )δ(Fa ) det{Fa , Cb } exp (Ea Ȧa − Ea2 − Ba2 )dtd3 x , (14.26)
h̄ 2 2
where the Fa are the gauge fixing depending on Aa . We have seen that θa Ca generates in-
R
finitesimal gauge transformations, and hence {Fa , Cb} is just an infinitesimal gauge transfor-
mation with parameters θa stripped off
δ
{Fb (A(y)), Ca (x)} = δθ (Fb [A(y)]) ≡ δa Fb . (14.27)
δθa (x)
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.1. Classical Yang-Mills Theories 124
In our derivation the gauge conditions Fa depend only on the spatial components of the gauge
potential. Recall that det(δa Fb ) is the determinant of the scalar-products of the gradient vectors
∇A Fb (A) with the symmetry-generating vector-fields (generating the θa -gauge orbits). We
may now assume that Fb also depends on A0 as long as we guarantee that the determinant keeps
this geometric meaning in the enlarged space of the gauge potentials (and not only their spatial
components). But also in this enlarged space
δ δFb δ
δθ Fb = ( a δθ Acµ ) = (∇Fb , Xa ), (14.30)
δθa (x) c
δAµ δθ
where now the gauge transformation may depend on time as well, and hence in δa Fb we must
take the gauge variation of all components of Aaµ . We see that the gauge fixing functions Fa
in (14.29) may depend on all components of the gauge potential. Since the action is gauge-
invariant, (14.29) still holds and the second equation in (14.27) still defines the object δa Fb
appearing in the path integral.
We can derive a more general representation for the transition amplitude than (14.29) by
shifting Fa → Fa + ga , where the functions ga do not depend on the gauge potential and hence
δa (Fb − gb ) = δa Fb . Since (14.29) is independent of the gauge choice Fa it is also independent
of the functions ga . Hence (we suppress h̄)
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.1. Classical Yang-Mills Theories 125
At this point one can introduce Grassmann-valued fields, so-called Fadeev-Popov ghosts η, η̄ to
represent the determinant of the infinitesimal gauge transformations, so that finally
Z
i(S[A]+ η̄(δa Fb )η+ j µ Aµ )
R R
Z[j] = const · DADηD η̄ G(Fa ) e , (14.32)
where we have re-introduce the coupling to a conserved current. The constant in front of the
path integral is chosen such that Z[0] = 1.
Let us see apply this formalism to the Lorentz gauge
We strip of the gauge parameter and obtains the following Faddeev-Popov operator,
δ
2 µ
δa Fb = δ F
θ b (A(y)) = − δab ∂ + f A
abc c (x)∂µ δ(x − y).
δθa (x)
Let us further take
i
h Z i
G(Fa ) = exp Fa2 . (14.35)
2λ
Finally, writing
1 1
Z Z
a
S[A] = − Fµν Faµν = Aµa (ηµν ∂ 2 − ∂µ ∂ν )Aνa + Sint [A], (14.36)
4 2
where Sint [A] contains all the cubic and quartic (self-interacting) terms, the path integral takes
the form
Z
DADηD η̄ ei(Seff [A,η,η̄]+ j µ Aµ )
R
Z[j] = const · , (14.37)
where
0 int
Seff [A, η, η̄] = Seff + Seff . (14.38)
We have split Seff into a quadratic term and a term containing higher orders of the fields,
1 1
Z Z
0
Seff = Aµa ηµν ∂ 2 − (1 − )∂µ ∂ν Aνa + η̄a (−∂ 2 )ηa
2 Z λ
int
Seff = Sint [A] + η̄a (fabc Aµc ∂µ )ηb . (14.39)
Now we see the effect of the gauge fixing more clearly. Whereas S0 (the term quadratic in the
gauge potential) has zero modes, S0 [Aµ = ∂µ λ] = 0, and hence cannot be inverted, the effective
quadratic term in (14.39) has no zero mode and can be inverted.
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.2. Abelian Gauge Theories 126
where
0 1 1
Seff = (Aµ , Kµν Aν ), Kµν = ηµν ∂ 2 − (1 − )∂µ ∂ν . (14.41)
2 λ
Since the operator K has no zero modes we can calculate the Gaussian integral and find
h i i
Z[j] = exp − (j µ , Kµν
−1 ν
j ) (14.42)
2
for the partition function, where the propagator is easily found to be
−1 1 1
Kµν = ηµν − (1 − λ) ∂µ ν .
∂ (14.43)
∂2 ∂2
Common choices for λ are λ = 1 (Feynman gauge) and λ = 0 (Landau gauge).
The continuation to the Euclidean sector is achieved by replacing E → −iE , B → −B
and d3 x → −id3 x, so that
Z R
0
Z[j] = C · DA e−Seff [A]+ jA
, (14.44)
where now
0 1 1
Seff = (Aµ , Kµν Aν ) with Kµν = −δµν ∆ + (1 − )∂µ ∂ν , (14.45)
2 λ
so that
h1 i
Z[j] = exp (j µ , Kµν
−1 ν
j ). (14.46)
2
The Euclidean propagator reads
−1 1 1
Kµν = − δµν + (1 − λ) ∂µ ∂ν . (14.47)
∆ ∆
Aµ = ǫµν ∂ν φ + ∂µ λ, (14.48)
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.3. The Schwinger model, Part II 127
where λ is a pure gauge degree of freedom and drops in gauge invariant expressions. In partic-
ular
1 1
F01 = −∆φ =⇒ Fµν F µν = (∆φ)2 , (14.49)
4 2
and the effective action Γ in (12.51) becomes
1 e2
Z
Γ[A] = φ ∆2 − ∆ φ. (14.50)
2 π
The function Φ in (12.39) simplifies to
Φ = λ − iγ5 φ. (14.51)
Note that both the effective action and the Green function are local in the new fields φ and λ.
We shall use the representation (14.29) (or rather its Euclidean continuation) for the path
integral, where we choose the Lorentz gauge
F = ∂µ Aµ = ∆λ (14.52)
and transform variables from A to φ, λ. First we note that the Jacobian of the transformation
(14.48) is just
! !
∂1 ∂0 1/2 ∆ 0
J = det = det = det(∆) (14.53)
−∂0 ∂1 0 ∆
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.3. The Schwinger model, Part II 128
or functions of such bilinears. Here M is one of the four matrices Id, γ5 and γ µ . The phase
factor is needed for the bilinear expression to be gauge invariant (recall that ψ → exp(iλ)ψ
under gauge transformations). Using
one finds
1
R Z R
ie A ǫµν ∂ν φdxµ
he ψ̄(y)Mψ(x)i = − Dφe−Γ[φ] eie tr MG(x, y)|λ=0 . (14.57)
Z[0]
Recalling that ((12.42))
i ξ µ γµ
G(x, y)|λ=0 = eγ5 (eφ(x)−eφ(y)) G0 (x − y), where G0 (ξ) = − (14.58)
2π ξ 2
we see that the spinorial trace in (14.58) vanishes for M = Id and M = γ5 and thus
1
hJ± i = 0, where J± = ψ̄P± ψ, P± = (1 ± γ5 ). (14.59)
2
Similarly, using
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.3. The Schwinger model, Part II 129
D is just the difference of a massive and massless Klein-Gordon propagator. Whereas the Klein-
Gordon operator is ultra-violet divergent the effective propagator D is well behaved for x = y.
Comparing (14.60) and (14.61) we see that j(z) = 2eδ(y − z) − 2eδ(x − z) so that
1
e2e [D(x,x)+D(y,y)−2D(x,y)] ,
2
hJ− (x)J+ (y)i = − (14.63)
4π 2 (x− y) 2
where we have used that D is symmetric in its arguments. For large separations r = |x − y| →
∞ only the massless propagator contributes to D(x, y) and thus (see (12.43))
π 1 1
D(x, y) −→ − 2
hx| |yi = − 2 log[µr]. (14.64)
e ∆ 2e
The function exp(−4e2 D(x, y)) ∼ µ2 (x − y)2 grows sufficiently fast to cancel the decreasing
factor in (14.63) and thus makes the whole expression remain constant for large separations
µ2 4e2 D(0)
hJ− (x)J+ (y)i −→ − e . (14.65)
4π 2
To find the numerical value we must compute D(0). The exact massive propagator is just a
Bessel function
1 1 √ 1h √ i
hx| |yi = − K 0 (er/ π) ∼ log(er/2 π) + γ (14.66)
∆ − e2 /π 2π 2π
where γ = 0.577215. Together with the massless propagator (12.43) one finds then
π 1h √ i 1 h e i
hx|D|yi ∼ log(er/2 π) + γ − log(µr) = log √ + γ . (14.67)
e2 2π 2e2 2µ π
√
The only natural mass-scale is the mass of the ’photon’, hence we set µ = e/ π and then
so that finally
e2 2γ
hJ− (x)J+ (y)i −→ − e (14.68)
16π 3
(the overall sign does not agree with the result in the literature?). For completeness we also
write down the exact answer
e2 2γ h √ i
hJ− (x)J+ (y)i = − e exp 2K 0 (er/ π) . (14.69)
16π 3
Now there is a subtle problem with the result (14.68) or (14.69). For a system with a unique
vacuum state the linked cluster property should hold, which states that
hJ− (x)J+ (y)i −→ hJ− (x)i · hJ+ (y)i = hJ− (0)i · hJ+ (0)i (14.70)
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.3. The Schwinger model, Part II 130
for |x − y| → ∞. In other words the connected 2-point function of J− and J+ should decay for
large separations. From (14.70) we conclude that
e 1 γ −iθ e 1 γ +iθ
hJ− i = √ e e and hJ+ i = √ e e , (14.71)
4π π 4π π
where θ is an arbitrary parameter not fixed by our considerations. Summing the two expectation
values yields then
e 1 γ
hψ̄ψi = √ e cos(θ) (14.72)
2π π
that is a generically non-vanishing fermionic condensate. On the other hand, in (14.59) we
concluded that the expectation values (14.71) and hence the condensate must vanish. What
went wrong?
To see what are the problems with the above calculation let use study the zero-energy eigen-
states of the Dirac operator. Introducing spherical coordinates
e−iφ (Dr − ri Dφ )
!
0
/=
D
e (Dr + ri Dφ )
iφ
0,
so that the Dirac equation for the zero-energy states ψ = (ψ+ , ψ− ) can be rewritten as
Integrating this equations around a circle or radius R and introducing the electric flux 2πΦ(R) =
R Aφ dφ through the corresponding disk yields
H
I I
2πΦ(R) = −i ∂φ log(ψǫ ) − ǫr∂r log ψǫ , (14.74)
where we have chosen the spherical gauge Ar = 0 in the gauge invariant expression (14.72).
The first integral on the right hand is just the winding number m of the solutions, e.g. if ψ ∼
exp(imφ) then it coincides with the angular momentum.
Near the origin a normalizable ψ must be smaller then 1/r and since Φ(0) = 0 we find
For large radii the wave function must decay more rapidly than 1/r and setting Φ = Φ(∞) we
obtain
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.3. The Schwinger model, Part II 131
It follows that m and Φ possess the same sign and that 0 ≤ m < |Φ| − 1 and 1 − |Φ| < m ≤ 0
for ǫ = + and ǫ = − respectively. Given Φ, the conditions on ǫ and m can be summarized as
Note that there are only either right- or lefthanded zero-modes, depending on the sign of the
total flux, and that the total number of zero modes is just the biggest integer less than |Φ|. For
example, for a flux Φ = 3.1 there are 3 zero modes ψ+ , but for Φ = 1 there is no zero mode.
Now, for gauge fields for which the Dirac operator possesses zero modes (12.20) is not equal
to (12.22) as we shall see next. Lets assume that the Dirac operator has n zero-modes which
we denote by ψj , j = 1, . . . , n. The excited modes we denote by ψk , k = n + 1, . . . , ∞).
Decomposing the field operators as
n
X ∞
X
ψ(x) = αj ψj (x) + βk ψk (x)
1 n+1
Inserting this decomposition into (12.20) and using DψD ψ̄ = DαD ᾱDβD β̄ the integral over
the α’s can easily be done since the action does not depend on them. One finds
1 hX in
Z hX i Z
DαD ᾱ exp (η̄, ψj )αj + ᾱj (ψj , η) = DαD ᾱ (η̄, ψj )αj + ᾱj (ψj , η)
n!
Z Y Y n
Y
= DαD ᾱ αj ᾱj (η̄, ψj )(ψj , η) = (η̄, ψj )(ψj , η).
1
where det′ is the determinant with the zero-eigenvalues omitted and Ge is the Green function
of the excited states that is on the space orthogonal to the zero modes
ψj (x)ψj† (y).
X
/ e (x, y) = δ(x − y) −
iDG (14.78)
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.3. The Schwinger model, Part II 132
Inserting all this into the path integral for the partition function we end up with
n R
η̄Ge η
(η̄, ψk )(ψj , η)det′ (iD)
/ e−
Y
Z[η̄, η] = (14.79)
1
and this is the generalization of (12.24) when fermionic zero-modes are present.
Let us now come back to problem of computing the two point functions (14.56) with M =
Id and M = γ5 . We have already seen that the naive calculation, which is valid for gauge
fields with no zero-modes, that is for gauge fields with total flux less or equal to 1, gives no
contribution. The gauge field with 2 or more zero modes do not contribute either, since Z
is higher order in the fermionic current so that after differentiating twice with respect to these
currents and setting them afterward to zero on gets a zero-result. So the only contribution comes
from the gauge fields with flux between 1 and 2 or −1 and −2. Those have exactly one zero
mode ψ1 and thus
Z
D ψ̄Dψ ψ̄(x)Mψ(x) = det′ (iD)tr
/ (ψ̄1 (x)Mψ1 (x)). (14.80)
For M = P+ only the right-handed zero mode contributes and thus only gauge potentials with
1 < Φ ≤ 2. For M = P− only the left-handed zero mode contributes and thus only gauge
potentials with −2 ≤ Φ < −1.
Typical gauge configurations having fermionic zero-modes are the vortex potentials
Φ(r)
Aµ = − ǫµν xν (14.81)
r2
where Φ is a function which vanishes at the origin so that A is regular there and tends to a
constant value for large radii Φ(r) −→ Φ. The corresponding φ in the decomposition (14.48)
and field strength read
Zr
Φ(r ′ ) ′ Φ′ (r)
φ(r) = − dr ∼ Φ log(r) and F01 = −∆φ = (14.82)
r′ r
from which follows that the Φ’s in (14.82) and (14.74) are the same. For these vortex fields
both the primed determinant (after subtracting the determinant of the free Dirac operator) and
the classical Maxwell action are finite and so is then the effective action Γ appearing in the
bosonic path integral. Thus the functional integration over φ’s with a given vortex flux should
yield a non-zero answer for
where the effective action in the denominator has the form (2.87a) and the one in the numerator
contains the classical Maxwell term and the primed determinant. As far as I now, nobody has
————————————
A. Wipf, Path Integrals
CHAPTER 14. GAUGE FIELDS 14.3. The Schwinger model, Part II 133
so far attempted to calculate the remaining path integral over φ in the continuum. But we see
that our previous naive calculation missed this non-vanishing term.
Similar considerations show that in the correlation function (14.60) the zero-modes drop
completely, since for a given gauge potential these modes are either left- or right handed. This
is the reason why the naive calculation above yields the correct result for the expectation values
(14.68,14.69).
This finishes the technical part of our discussion of the Schwinger model. Most of the
results presented have been obtained by Nielsen and Schroer [52]. The Schwinger model on the
sphere and the torus have also been studied and the results of these refined calculations agree
with (14.71,14.72). So there is no doubt that the Schwinger model shows a breaking of the
chiral symmetry (the operator ψ̄ψ transforms non-trivially under global chiral transformations).
One may ask what happened to the celebrated Goldstone theorem since on the one hand a
continuous U(1) symmetry is broken and on the other hand there is no massless Goldstone
boson. The answer to this apparent contradiction comes from the fact that the axial current is
not conserved in the Schwinger model, and the derivation of the Goldstone theorem assumes a
conserved Noether current. The Schwinger model possesses another quiet interesting property.
If we couple the gauge potential to an external current L −→ L + j µ Aµ with j 0 (x) = ρ(x) =
q1 δ(x − x1 ) + q2 δ(x − x1 ), then the interaction decreases exponentially with the separation
|x1 −x2 | of the two charges, due to the mass of the photon. So the expected long range Coulomb
force does not appear. This can only happen if the charges q1 and q2 are shielded. The physical
mechanism responsible for this charge shielding is the spontaneous pair production. As soon
as one tries to separate two ’quarks’ (we call the fundamental field ψ quark field to emphasize
the analogy to QCD) it is favorable to create a quark pair out of the vacuum and then each of
the two created quarks shield one of the originally present quarks. The physical particles of the
theory are quark pairs, and not quarks.
————————————
A. Wipf, Path Integrals
Chapter 15
There are many interesting physical effects induced by external fields, e.g. the Coulomb scat-
tering of a charged electron by a heavy nucleus, the electron-positron pair production in strong
electric fields, the Hawking radiation emitted by a black hole and the Casimir effect induced
by external gauge- and gravitational fields to mention only a few of them. One of the central
objects to describe such phenomena is the S-matrix. So we shall first derive its path integral rep-
resentation and apply the result to the calculation of the pair creation in strong electromagnetic
fields.
Setting
134
CHAPTER 15. EXTERNAL FIELD PROBLEMS 15.2. Scattering in Quantum Mechanics 135
The generalized step function θ is 1 if its arguments are in decreasing order and else it is 0. In
other words, in the time ordered product of n operators the operator with the ’latest time’ stands
on the left, the one with the second-latest time follows and so on.
Between the asymptotic states there is the relation
and this defines the scattering matrix transforming asymptotic in-states in asymptotic out-states.
The path integral representation is most easily obtained be rewriting (15.2) as
where U and U0 are the full and free evolution operators in the Schrödinger picture.
————————————
A. Wipf, Path Integrals
CHAPTER 15. EXTERNAL FIELD PROBLEMS 15.3. Scattering in Field Theory 136
where D = D(t, t′ ) is the solution defined in (3.18) and Ḋ and D ′ denote the partial derivatives
with respect to t and t′ respectively. Let us take as an example a harmonic force which vanishes
exponentially for large times, e.g.
2a2
ω 2 (t) = . (15.11)
cosh2 (at)
For this interaction the D function reads
D(t, t′ ) = tanh(at′ ) t tanh(at) − 1/a − t ↔ t′ (15.12)
Assuming t′ = −t and letting t → ∞ one finds after expanding the D-function and its deriva-
tives to leading order in t and eat the result
s s
1 e2at ie2at i
h. . .i = exp − (p + p′ )2 exp ((p + p′ )2 + 2p2 + 2p′2 ) .
2πimh̄ 8a 16amh̄ 4amh̄
Using the identity
α iαξ2
r
e −→ δ(ξ) for α→∞
iπ
we end up with
2 /mh̄
hp|S|p′i = iδ(p + p′ )eip . (15.13)
for the exact S-matrix. One easily checks that SS † = I as it must be. Note that a particle subject
to a harmonic force with time-dependent coupling strength as defined in (15.11) reflected with
probability one. This is a particular feature of the chosen coupling.
For systems which are not exactly soluble one has to retreat to some approximation, e.g.
the ordinary perturbation theory in the coupling constant or the semiclassical approximation.
To find the perturbative expansion of the S-matrix one inserts the perturbation serie (4.12) into
(15.9) and this yields the well-known rules for the diagrammatic expansion of S-matrix ele-
ments. Similarly, the semiclassical expansion is obtained by inserting (6.40) into (15.9)
————————————
A. Wipf, Path Integrals
CHAPTER 15. EXTERNAL FIELD PROBLEMS 15.3. Scattering in Field Theory 137
and fulfills
and similarly for the corresponding momentum densities. Here H(t) = H(Φ(t), π(t), j(t))
may depend on an external current and H0 is the time-independent free Hamiltonian. It follows
from these formulae that
U(t)H(Φ(t), π(t), j(t))U −1 (t) = H(Φin (t), ψin (t), j(t)). (15.17)
Now we may use (15.17) for the second term on the right hand side to find
and similarly for the time derivative of ψin . Comparing this result with the time evolution
determined by (15.16) we see that
U̇ U −1 + i H(Φin , ψin , j) − H0 (Φin , πin ) ≡ U̇ U −1 + iHI (t)
commutes with all in-fields and hence must we a multiple of the identity operator. This central
operator will drop in normalized matrix elements and can be left out in the following. Thus the
time dependence of U is determined by the interacting Hamiltonian HI as follows
Zt
U(t) = T exp − i dt′ HI (t′ ) . (15.21)
−∞
Zt
S = lim T exp − i HI (t′ ) . (15.22)
t→∞
−∞
————————————
A. Wipf, Path Integrals
CHAPTER 15. EXTERNAL FIELD PROBLEMS 15.3. Scattering in Field Theory 138
This is a rather formal representation of the scattering matrix. When one tries to calculate S
(e.g. perturbatively) one encounters short-distance singularities which must be regularized. The
treatment of these singularities is the subject of renormalization theory.
Let us now consider the electron-positron field in interaction with an external gauge fields.
Its interaction Hamiltonian is given by (see (12.3) and below)
This should be compared with the perturbation expansion of the path integral,
Z Z R
/
DψD ψ̄eiS = DψD ψ̄eiS0 +ie ψ̄ Aψ
(ie)2
Z Z Z Z
iS0 / + / / + · · · .(15.27)
= DψD ψ̄e 1 + ie ψ̄ Aψ ψ̄ Aψ ψ̄ Aψ
2!
According to (12.12) the moments are just the corresponding expectation values of the time-
ordered fields. Hence we obtain the following simple looking path integral representation for
the expectation value of the S-matrix in the in-vacuum (omitting the subscript ’in’):
1 (ie)2
Z Z Z Z
h0in |S|0ini = DψD ψ̄eiS0 1 + ie / +
ψ̄ Aψ /
ψ̄ Aψ / + · · · , (15.28)
ψ̄ Aψ
Z[0] 2!
which according to (15.27) is, up to a A-independent normalization constant, just the full path
integral. Hence we conclude, that
1
Z
iD/ − m + iǫ
h0in|S|0ini = DψD ψ̄eiS = det = exp (iSeff [A]). (15.29)
Z[0] i∂/ − m + iǫ
————————————
A. Wipf, Path Integrals
CHAPTER 15. EXTERNAL FIELD PROBLEMS 15.4. Schwinger-Effect 139
This formula yields a direct physical interpretation of the fermionic determinant. Expanding
1
/
log iSeff [A] = log det I + eA (15.30)
i∂/ − m + iǫ
in powers of the electric charge reproduces the well-known perturbation expansion for the
vacuum-vacuum amplitude (External A-lines attached to a fermionic loop).
In the last section we have computed this determinant for massless two-dimensional fermions
exactly. Continuing the Euclidean result (12.50) back to Minkowski space-time (the inverse
transformation of (12.17) on finds
iD/ + iǫ h ie2 Z
1 i
h0in |S|0ini = det = exp F01 F01 (15.31)
i∂/ + iǫ 2π ∂2
for the vacuum to vacuum amplitude. Since this is a pure phase, no pairs are produced in the
Schwinger model. This is not true anymore for massive fields. Also, this conclusion only holds
for gauge-fields for which (12.50) is the correct formula for the fermionic path integral. We
have already seen that this formula is only correct for gauge fields for which the Dirac operator
has no zero modes, that is for gauge fields with flux less or equal to 1.
15.4 Schwinger-Effect
Let us now calculate the pair production rate of massive fermions in a constant electro-magnetic
/
field. To compute the determinant of iD−m we recall that the non-zero eigenvalues of iD/ come
/ 2 2
always in pairs {λ, −λ} so that in the determinant det(iD−m) they contribute −λ +m . Hence
the determinant of iD/ − m can be defined as the square root of the determinant of −D / 2 − m2
(for the zero-modes this is true anyway). To compute the logarithm of the determinant we use
the identity
Z∞
ds is(b+iǫ)
log(a/b) = e − eis(a+iǫ) (15.32)
s
0
which yields
ds −is(m2 −iǫ)
Z Z 2 2
− log (2iSeff [A]) = e d4 x hx|e−isD/ |xi − hx|e−is∂/ |xi , (15.33)
s
where we have used the (formal) identity log det(A) = tr log(A) and have represented the trace
in the |xi basis. For a constant electric field in the 3-direction the only non-vanishing field
strength components are
————————————
A. Wipf, Path Integrals
CHAPTER 15. EXTERNAL FIELD PROBLEMS 15.4. Schwinger-Effect 140
As potential we choose Aµ = (0, 0, 0, Ex0) with constant E. In the present case the square of
/ (see (8.69)) simplifies to
D
Since the Pauli term in D / 2 commutes with D 2 , its exponential can be computed separately.
Using (γ 0 γ 3 )2 = 1, one finds
Now we are left with computing the heat kernel of D 2 . For that purpose we observe that D 2 can
be written as the sum of two 2-dimensional commuting operators
2
D 2 = − ∂12 + ∂22 + ∂02 − (∂3 − iEx0 )2 = −∆12 + D03 (15.37)
and thus its heat kernel is just the product of the two corresponding two-dimensional heat ker-
nels
2 1
hx|e−isD |xi = hx1 , x2 |eis∆12 |x1 , x2 iK(s, x0 , x3 ) = K(s, x0 , x3 ), (15.38)
4iπs
2
where K the heat kernel belonging to D03 . To calculate this remaining heat kernel we first note
that ∂3 commutes with D03 . Thus they can be diagonalized simultaneously and the eigenfunc-
tions have the form
2 3
p3 2
D03 ψλ = λψλ =⇒ ψλ = eip3 x φλ , where ∂02 + E 2 (x0 − ) φλ = λφλ . (15.39)
E
It follows that the diagonal-elements of K are independent of the x3 . The remaining operator
on the right hand side in (15.39) is just a shifted harmonic oscillator with imaginary frequency
and thus has eigenvalues −i(2n+1)E (we assume E to be positive, else we would have to write
everywhere |E|. The minus sign is due to time ordering). Since the eigenvalues are independent
of p3 they are degenerate and apriori we can determine the trace of K only up to the multiplicity
C of the eigenmodes as
∞
Z
C
dx0 dx3 K(s, x0 , x3 ) = C e−s(2n+1)E =
X
. (15.40)
n=0 2 sinh(sE)
However, recalling that for a vanishing electric field K is the free heat kernel,
s
−is∂ 2 i −i 1
< x0 , x3 |e |x0 , x3 i = = ,
4πs 4πs 4πs
————————————
A. Wipf, Path Integrals
CHAPTER 15. EXTERNAL FIELD PROBLEMS 15.4. Schwinger-Effect 141
(since phases relevant, we have emphasized that due to the (+, −)-signature in D03 the diagonal
elements of h. . .i are real), we can now easily determine C and find
EV03
Z
dx0 dx3 K(s, x0 , x3 ) = , (15.41)
4π sinh(sE)
where V03 denotes the volume of the (0, 3) plane. Inserting now (15.36,15.38) and (15.41) into
the general formula (15.33), we find for the real part d4 x w(x) ≡ ℜ log(2iSeff ) the formula
R
Z∞
V 1 1 i ds
Z h
4 2 −iǫ)
d x w(x) = ℜ e−is(m E coth(sE) −
(2π)2 i s s2
0
V 1 i ds
Z h
= − e−ǫs sin(sm2 ) E coth(sE) − , (15.42)
(2π)2 s s2
where V = V03 V12 is the volume of the four-dimensional Minkowski space-time. Since
R
2 iSeff [A] 2 2ℜ(iSeff ) − d4 x w(x)
|h0in|S|0ini| = |e | =e =e
where α = e2 /4π is the fine structure constant. The analog calculation in two dimensions yields
∞
eE X 1 nπm2
w(x) = exp − . (15.45)
2π 1 n |eE|
————————————
A. Wipf, Path Integrals
CHAPTER 15. EXTERNAL FIELD PROBLEMS 15.4. Schwinger-Effect 142
In these exact formulae for the pair creation density in a constant electric field the essential
factor is non-perturbative ∼ exp(−πm2 /eE) and can be interpreted as Gamov factor for the
tunneling of an electron in an external electric field through a potential barrier. Such a factor
cannot be gotten by ordinary perturbation theory, since exp(c/e) cannot be expanded in powers
of the coupling constant. Unfortunately, pair creation in a constant electric field has not been
observed since |E| ≪ m2 for realistic electric fields. Due to the exponential suppression factor
the creation density is then too small.
————————————
A. Wipf, Path Integrals
Chapter 16
Effective potentials
We have already pointed out the difficulty with (local) mass terms in pure gauge theories. Ex-
plicit mass terms spoil the crucial gauge invariance of the massless theory (however, as we have
seen in the Schwinger model, non-local gauge invariant mass terms are possible). The problem
of generating masses in a manner consistent with gauge invariance was solved by Weinberg and
Salam. For a historical account and references see [53]. They used the idea of spontaneous sym-
metry breaking. A familiar example of this mechanism is the magnetisation of a ferro-magnetic
material below its Curie temperature.
In field theory the symmetry breaking is implemented by scalar fields which minimally
couple to gauge fields and interact with themselves. The electro-weak Lagrangian for the gauge,
scalar and fermion fields has the form
1
L = − Fµν F µν + ψ̄(iD)ψ
/ + (Dµ φ)† D µ φ − Γψ̄φψ − V (φ), (16.1)
4
/ the Dirac operator (8.66) acting on quarks and
where Fµν is the field strength tensor (8.71), D
leptons, Dµ φ = (∂µ − iAµ )φ the covariant derivative of the scalar field, Γψ̄φψ the Yukawa
interaction between the fermions and scalars and V (φ) the self-interaction of the scalars. All
fields transform under certain representation of the electro-weak gauge group SU(2)L × U(1).
If the scalar field acquires a non-vanishing vacuum expectation value, hφi = v, then both
the gauge bosons and fermions may become massive, mA = ev, mψ = Γv, due to the third and
fourth term on the right hand side of (16.1). In what follows we shall concentrate on the scalar
sector to understand how φ can acquire a non-vanishing vacuum expectation value. The proper
quantities to describe the spontaneous symmetry breaking mechanism are effective potentials.
143
CHAPTER 16. EFFECTIVE POTENTIALS 16.1. Legendre transformation 144
Zβ
1 1 h Z i
W (β, j) = log tr e−β(H−jq) = log c · Dx exp − S + j x(τ ) , (16.2)
β β
0
Here E0 (j) denotes the ground state energy of the shifted Hamiltonian H −jq. The conventional
effective potential is obtained from the Schwinger function by a Legendre transformation
h i
Γ(β, φ̄) = (LW )(φ̄) = sup j φ̄ − W (β, j) . (16.4)
j
The maximizing current (if it exists) is called the current conjugate to φ̄.
Since Legendre transformations play an important role in the classical mechanics, thermody-
namics and quantum field theory, let us first collect some relevant properties of these transfor-
mations. In the following φ̄ and j are elements of a convex set in n . R
1. The Legendre transform of a function which is convex for sufficient large arguments (here
we are not concerned with domain problems) is always convex.
To see that let
= (1 − α)Γ(φ̄1) + αΓ(φ̄2 ),
where we have used that the supremum of the sum is less or equal to the sum of the
suprema. The last expression is just the linear interpolation between the points (φ̄i , Γ(φ̄i )).
Thus we have shown that the graph of Γ is always below the segment between two points
on this graph and this proves the convexity of Γ.
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.1. Legendre transformation 145
It follows that
that is the double-Legendre transform is always greater or equal to the original function.
On the other hand
or that the double-Legendre transform is always less or equal to the original function.
Together with the above inequality we conclude that for any convex function
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.1. Legendre transformation 146
If we replace (j, φ̄) → (p, ẋ) and (W, Γ) → (H, L) this is the familiar Legendre transfor-
mation in classical mechanics from the Hamiltonian to the Lagrangian formulation.
After this excursion to the property of Legendre transformation we note that the Schwinger
function is always convex, since
Zβ
d2 1 Dh ih iE
2
W (β, j) = dsdτ x(s) − hx(s)ij x(τ ) − hx(τ )ij ≥ 0,
dj β j
0
where the expectation values are taken with respect to the shifted action S − j x, and thus
R
are current-dependent. To get a better intuition for its Legendre transform Γ we note that for
β→∞
h i
W (j) = suphjq − Hi = sup tr ρ[jq − H] = sup j φ̄ − inf tr (ρH) , (16.8)
ψ ρ φ̄ tr ρq=φ̄
where we have used that the set of density matrices {ρ|tr ρ = 1, ρ = ρ† > 0} is a convex and
compact set, and hence the infimimum of the linear functional tr ρ(jq − H) is attained for pure
states, ρ = Pψ .
On the other hand, the constraints (there may be more than one q and thus several constraints)
tr ρq = φ̄ define a plane and thus the density matrices obeying these constraints form again a
convex (and compact) set. It follows that the infimum of tr ρH on the constraint plane is attained
on the intersection of this plane with the boundary of the set of density matrices. Let us assume
that
inf tr ρH = tr ρi H, i = 1, 2
tr ρ=φ̄i
that is, ρ1 and ρ2 are the densities which minimize tr ρH under constraints tr ρi q = φ̄i . Defining
ρα = (1−α)ρ1 + αρ2 one easily sees that tr ρα q = φ̄α (see 16.5) and hence
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.2. Effective potentials in field theory 147
The infimum of Γ is
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.2. Effective potentials in field theory 148
in the presence of a constant external current j. The current is chosen constant so as to preserve
the translational invariance of Z(j). For finite volumes Ω, translational invariance is understood
to be with respect to periodic boundary conditions. Again the Schwinger function
1
W (Ω, j) = log Z(Ω, j) (16.14)
Ω
is strictly convex since its second derivative is (Ω times) the expectation value of the positive
quantity (M −hMij )2 , where M = (1/Ω) φ(x)dd x. The current-dependent expectation values
R
are to be computed with the shifted action as in (16.13). W (j) allows one to compute the
effective field, defined as
R
−S[φ]+j φ
Dφ φ(x) e dW
R
hφ(x)ij = R = . (16.15)
−S[φ]+j
Dφ e φ dj
R
Of course, in cases where W is non-differentiable (or equivalently Γ shows at least one plateau)
we must be cautious what we mean by formulae like (16.15). We shall come back to this point
later on.
The conventional effective potential Γ(Ω, φ̄) in (16.4) is the Legendre transform of W . If
the minimum of Γ occurs at a unique point φ̄ = φ̄0 , the point φ̄0 defines the vacuum state
of the theory, and the semiclassical expansion around φ̄0 generates the one-particle-irreducible
Feynman graphs [54]. The minimum is unique if either the volume is finite, or the classical
potential is convex (or both). When the classical potential is not convex, as happens in particular
for spontaneous broken potentials, the minimal points φ̄0 of Γ(φ̄) = Γ(∞, φ̄) are not unique but
lie on a plane in φ̄-space, as pointed out above. In this case the vacuum is not determined by
Γ(φ̄) but by Γ(φ̄) plus the direction from which a trigger current j approaches the value zero.
Such a trigger current forces the system into a pure state. As we have seen, the expectation
value φ̄ in a pure state lies on the edge of the plane of Γ. Furthermore, in the degenerate case
the naive semiclassical expansion for the effective potential breaks down and must be replaced
by some alternative approximation.
Since for Ω = ∞, V non-convex, the loop expansion (semiclassical expansion) has problems,
a computational approach is more desirable, and in that case Γ may not be the best quantity to
consider. Also note that we haven’t been able to write down an explicit path integral represen-
tation for the conventional effective potential Γ. A much more suitable and direct (at least in
the path integral approach) quantity is the effective potential defined by
Z
−S[φ] 1Z d
exp − ΩU(Ω, φ̄) = Dφ δ(M − φ̄) e , M= d x φ(x), (16.16)
Ω
which we called constraint effective potential in [55]. Clearly, if the classical potential is invari-
ant under the action of the symmetry group, then U(Ω, φ̄) is invariant as well. The constraint
effective potential is relates to similar definitions in statistical mechanics and spin systems and
e−ΩU (Ω,φ̄)
P (φ̄) ≡ R (16.17)
dφ̄ e−ΩU (Ω,φ̄)
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.2. Effective potentials in field theory 149
it to be interpreted as the probability density for the system to be in the state of ”magnetization”
φ̄. The probability for the occurrence of a state whose averaged field is not a minimum of
U then becomes less and less as Ω → ∞. Also, the constraint effective potential is a more
direct quantity to compute with Monte Carlo simulations, since an external current need not be
introduced.
Multiplying both sides of (16.16) by exp(Ωj φ̄) and integrating over φ̄, yields
Z
eΩ[j φ̄−U (Ω,φ̄)] dφ̄ = eΩW (Ω,j) . (16.18)
Hence W is related to U by a Laplace transformation. Note that since Γ is the Legendre trans-
form of W , the function Γ(Ω, φ̄) is uniquely defined by U(Ω, φ̄). Conversely, since W is the
Legendre transform of Γ, U can be recovered from Γ by an inverse Laplace transformation.
Thus there is a one-to-one correspondence between U(Ω, φ̄) and Γ(Ω, φ̄).
Now let us discuss what happens in the infinite-volume limit Ω → ∞. In this limit the saddle-
point approximation to the ordinary integral (16.18) becomes exact. Then
W (j) = sup j φ̄ − U(φ̄) = (LU)(j). (16.19)
φ̄
It follows that Γ = LW = L2 U. Thus Γ is the convex hull of U. Although U(Ω, φ̄) is in general
not convex for finite volumes, one can prove that it becomes convex for Ω → ∞ [55] so that
Thus in the infinite-volume limit the two potentials become identical. However, in a finite
volume the two potentials are not identical and U(Ω, φ̄) need not necessarily be convex.
The constraint effective potential is also useful for extracting information directly about the
gross properties of the system such as whether is suffers a spontaneous symmetry breakdown
or whether it has a finite correlation length. To see this ones notes that
h i
φ̄p exp Ω(j φ̄ − U(Ω, φ̄)) dφ̄
R
1
Z
h i = dd x1 . . . dd xp hφ(x1 ) · · · φ(xd )iΩ
j, (16.21)
exp Ω(j φ̄ − U(Ω, φ̄)) dφ̄ Ωp
R
h i
i.e. that the moments of N −1 exp Ω(j φ̄ − U(Ω, φ̄)) are the averaged Schwinger (correlation)
functions. For p = 1 this gives the vacuum expectation value
Z
hφ(x)iΩ
j = N
−1
φ̄eΩ[j φ̄−U (Ω,φ̄)] dφ̄. (16.22)
For any finite volume the symmetry of U leads to hφ(x)iΩ 0 = 0 To get a non-trivial result one
must keep a trigger current, and only after the infinite-volume limit has been taken can the
trigger be removed. If there remains a non-trivial expectation value after setting j = 0 then
there is a spontaneous symmetry breaking.
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.3. Lattice approximation 150
where φi = phi(xi ), (i = 1, 2, . . . , N = Ω/ad ) and hiji is the sum over all nearest neighbour
P
where the lattice potential V L is equal to the classical potential, but with rescaled parameters.
The masses and coupling constants are rescaled according to their dimensions, e.g. mL = a2 m
etc. By using the lattice field as new integration variable the constraint effective lattice potential,
which is related to the continuum potential as
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.3. Lattice approximation 151
value. The saddle point of ipφ̄ + log f (p) in the complex p-plane is the point p = ij, where j is
a solution of φ̄ = dW0 /dj and exp(W0 (j)) = dφ exp[jφ − V (φ)]. Hence we find that in the
R
limit N → ∞
where W0 is the Schwinger function of the zero-dimensional theory with potential V , i.e.
Z
eW0 (j) = ejφ−V (φ) dφ. (16.29)
Clearly, since we have neglected the positive kinetic energy in the action the incoherent potential
Γ0 yields a lower bound on the exact potential,
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.3. Lattice approximation 152
Also note that since W0 is a differentiable and strictly convex function the constraint potential
Γ0 is strictly convex as well.
Next we derive an upper bound for U. Since 12 (φi − φj )2 ≤ φ2i + φ2j we find
1X
(φi − φj )2 = T [φ − φ̄] ≤ 2d φ2i − 4dφ̄ φi + 2dN φ̄2 ,
X X
T [φ] = (16.31)
2 hiji i
where we have taken into account that in d dimensions every site has 2d nearest neighbours.
Inserting this inequality into (16.27) one obtains
Z
−N U (N,φ̄) 2dN φ̄2
e ≥e δ(M − φ̄)e−V2d [φ] ,
where V2d [φ] = 2d φ2i + V (φi ). This yields the upper bound
P P
i
where
Z
Γ2d (φ̄) = (LW2d )(φ̄) and W2d = log dφ ejφ−V2d (φ) (16.33)
is the incoherent constraint effective potential which corresponds to the classical potential with
shifted mass V2d (φ). In general the function on the r.h.s. of (16.32) is not convex. However,
since U(φ) is known to be convex, (16.32) actually implies that
U(φ̄) ≤ L2 −2dφ̄2 + Γ2d (φ) . (16.34)
We conclude our discussion of the analytic properties of U(φ̄) by deriving an Ehrenfest equation
which is very useful for Monte-Carlo simulations. For that purpose we shift the field by a
constant φi → φi + φ̄ in (16.27). Because of the translational invariance of the measure Dφ we
obtain
Z
e−N U (N,φ̄) = Dφ δ(M)e−S[φ+φ̄] .
Only the potential term in the action is affected by the shift and therefore
d 1
U(N, φ̄) = hV ′ [φ]iφ̄ , (16.35)
dφ̄ N
where
Dφ δ(M − φ̄)O[φ]e−S[φ]
R
hO[φ]iφ̄ = (16.36)
Dφ δ(M − φ̄)e−S[φ]
R
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.3. Lattice approximation 153
and this is the required Ehrenfest equation which relates the derivative of the quantum potential
to the expectation value of the derivative of the classical potential. For example, for the Higgs-
model
where we have used the translational invariance, i.e. that h φ3i iφ̄ = Nhφ3i iφ̄ . For the Higgs
P
model this equation can be used and has been used for the MC simulations. The following
figures show the two effective potentials Γ and U for the Higgs models (16.37) in various di-
mensions and for different ”volumes” N. Also the lower and upper bounds (16.30) and (16.34)
are plotted in the figures. The calculation has been done with a modified Metropolis algorithm
(see section 9.2).
So far we considered the regularized scalar theories, that is we kept the lattice constant a
fixed. At the end we wish to let the lattice constant tend to zero in order to remove the regu-
larization. Then the bar quantities have to be related to physical quantities by renormalization.
First one introduces a dimensionless lattice length ǫ(a = ǫΛ−1 , ) where Λ is a scale parameter
Z Z
with a mass dimension) and compares the lattices d and ǫ d when ǫ is allowed to take values in
the interval 0 < ǫ ≤ 1. The parameters and field are scaled so that the scaled potential becomes
U ǫ (φ̄, m, g) = ǫ−d U ǫ=1 Zǫ(d−2)/2 φ̄, m(ǫ), g(ǫ) (16.39)
has a continuum limit. We define a ”physical” mass µp and coupling constant gp in terms of U ǫ
by some typical equation such as (for d = 4)
Λ2 1
hφ2 iǫ = and hφ4 iǫ − hφ2 i2ǫ = , (16.40)
µ2p gp
where
ǫ
dφ̄ O(φ̄)e−U
R
hOiǫ = . (16.41)
dφ̄ e−U ǫ
R
Of course, µp and gp as defined in (16.40) will not necessarily be the physical mass and quartic
coupling constant for the scalar field, but just some related physical quantities.
Now the lattice renormalization consists in letting the bare mass and coupling constant m, g
depend on ǫ in such a way that the physical constants µp , gp do not depend on ǫ. Given the
dependence of U ǫ on ǫ, m, g and given m(1) = m, g(1) = g the ǫ-dependence of the bare pa-
rameters is then determined implicitly by (16.40). In other words, the renormalization consists
of constructing an ǫ-dependent map from (µp , gp ) to (m, g) by means of (16.40).
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.4. Mean field approximation 154
d2 ǫ
U ǫ (φ̄) = minimal for φ̄ = φp and U |φp = mp , (16.42)
dφ̄2
where the physical quantities (φp , mp ) are measured w.r. to some mass scale Λ. We have already
pointed out that in the broken phase Γ and hence U (recall that Γ = U in the thermodynamic
limit N → ∞) develop a plateau. As minimizing value φp in (16.42) we take the maximal
φp which minimizes U since this belongs to a pure phase of the theory. Also we evaluate the
second derivative near φp but a little bit away from the plateau.
φ2i −
X X X
S[φ] = d φi φj + V (φi ) (16.43)
i hiji i
and hence becomes an incoherent model. Analog to (16.30) and (16.33) we obtain
Z
UM F (φ̄) = −dφ̄2 + Γd (φ̄), where Γd = LWd , Wd = log dφ ejφ−Vd (φ) . (16.45)
Note that UM F (φ̄) is half way between the lower bound Γ0 (φ̄) in (16.30) and the upper bound
−2dφ̄2 + Γ2d (φ̄) in (16.33). One can actually prove that UM F is also an upper bound for the
exact potential. Note that UM F is differentiable and in the non convex case (the broken phase of
the MF-model) it displays no plateau. Hence the apparent problem with the conditions (16.42)
mentioned after (16.42) do not arise and we may take these conditions literally.
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.4. Mean field approximation 155
We shall need the minimum φ0 of UM F and the curvature at this minimum. Using j(φ̄) =
Γ′d (φ̄),
which relates the current to its conjugate field, one sees at once that the minimum con-
dition becomes
Since Γd is the Legendre transform of Wd the inverse relation reads φ(j) = Wd′ (j). By inserting
the minimum condition into that equation we find the self-consistency equation
dφ φ ej0φ−Vd (φ)
R
φ0 = R = hφij0 , where j0 = 2dφ0 , (16.47)
dφ ej0 φ−Vd (φ)
′′ ′′
for the expectation value of the Higgs field. To compute UM F (φ0 ) we use the relation Γ (φ̄) =
W ′′ [j(φ̄)]−1 between the curvatures of two Legendre-related functions. Together with the min-
imum conditions one obtains
D E−1
′′ 2
m0 = UM F = −2d + (φ − φ0 ) (16.48)
j0
for the Higgs-boson mass in the broken phase. Clearly the incoherent Schwinger function in
(16.45) is strictly convex and symmetric (if V is symmetric) and hence j(φ̄) vanishes when
φ̄ does. From (16.48) we conclude that the curvature of UM F at the origin is negative when
hφ2 i0 > 1/2d. Consequently the potential (16.45) is spontaneously broken in case where
φ2 e−Vd (φ) 1
R
−V (φ)
= hφ2 i0 > . (16.49)
e d 2d
R
Suppose, for example, that the mass m in V (φ) = mφ2 + gφ4 is less than −d. Since the m-
derivative of the expectation value hφ2 i0 decreases with increasing mass the expectation value
becomes smaller when m is replaces by −d. However, for m = −d the effective mass m + d
in Vd vanishes and the expectation value can be computed explicitly. In this way one finds from
(16.49) that UM F is spontaneously broken when
" #2
Γ(3/4)
m < −d and g < 2d . (16.50)
Γ(1/4)
The continuum limit for the mean-field theory: As physical parameters we take the expec-
tation value of the Higgs field φp and the Higgs-boson mass mp in the broken phase (see 16.42).
One can prove that in the mean field approximation the wave function renormalization constant
Z
Z in (16.39) is one [56] so that the potential on (ǫ )d becomes
ǫ −d d/2−1
UM F (φ̄) = ǫ UM F (ǫ φ̄) = −dǫ−2 φ̄2 + ǫ−d Γd m(ǫ), g(ǫ), ed/2−1 φ̄ ,
where the scaled bare parameters are to be determined by the renormalization conditions. Here
we take the conditions (16.42).
————————————
A. Wipf, Path Integrals
CHAPTER 16. EFFECTIVE POTENTIALS 16.4. Mean field approximation 156
ǫ d/2−1
Clearly, when φp minimizes UM F then ǫ φp minimizes UM F and satisfies the self-
consistency equation (16.47). Thus, the first renormalization condition reads
and the expectation values have been defined in (16.47). In the same way, using (16.48), one
obtains the second renormalization condition
d
2 −1
′′
ǫ2 mp = ǫ2 (UM
ǫ
F ) (φp ) = −2d + φ − ǫ 2 −1 φp . (16.52)
jp
The following asymptotic renormalization flows for ǫ → 0 in 2 and 3 dimensions can be de-
rived:
mp 3
d=2: g(ǫ) ∼ 2 ǫ2 , m(ǫ) ∼ − + 2φ2p g(ǫ)
8φp 2
mp
d=3: g(ǫ) ∼ 2 ǫ, m(ǫ) ∼ −g(ǫ). (16.53)
8φp
————————————
A. Wipf, Path Integrals
Bibliography
[2] L.S. Schulman, Techniques and Applications of Path Integration, John Wiley & Sons, Inc.,
1981.
[3] R.P. Feynman, Space-time approach to non-relativistic quantum mechanics, Rev. Mod.
Phys. 20 (1948) 367.
[4] R.P. Feynman and A.R. Hibbs, Quantum Mechanics and Path Integrals, McGraw-Hill,
New York 1965.
[5] I.M Gel’fand and A.M. Yaglom, Integration in functional spaces and its applications in
quantum physics, J. Math. Phys. 1 (1960) 48.
[6] E. Nelson, Feynman integrals and the Schrödinger equation, J. Math. Phys. 5 (1964) 332.
[7] S.G. Brush, Functional Integrals and Statistical Physics, Rev. Mod. Phys. 33 (1961) 79.
[8] H. Kleinert, Path Integrals in Quantum Mechanics, Statistics and Polymer Physics, and
Financial Markets, 4 nd edition 2006, World Scientific, Singapore
[9] J. Glimm and A. Jaffe, Quantum Physics, A Functional Integral Point of View, Springer,
1981
[11] L. van Hove, Sur certaines représentations unitaires d’un groupe infini de transformations,
Proc. Roy. Acad. Sci. Belgium 26 (1951) 1.
[12] R.P. Feynman, The Principle of Leat Action in Quantum Mechanics, Ph.D.Thesis, Prince-
ton University, May 1942.
[13] M. Kac, (1949). On Distributions of Certain Wiener Functionals, Transactions of the AMS
65 (1) (1949) 1.
157
BIBLIOGRAPHY Bibliography 158
[14] J. Schwinger, Selected Papers on Quantum Electrodynamics, Dover, New York, 1958.
[17] M. Peshkin, The Aharonov-Bohm Effect: Why it cannot be eliminated from Quantum Me-
chanics, Phys. Rep. 80 (1981) 375.
[19] K. Kirsten and A. McKane, Functional determinants for general Sturm-Liouville prob-
lems, J. Phys. A37 (2004) 4649.
[20] M.L. Glasser, Summation over Feynman histories: charged particle in a uniform magnetic
field, Phys. Rev. B 113 (1964) 831.
[21] P.R. Chernoff, Note on product formulas for operator semigroups, J. Funct. Anal. 2 (1968)
238.
[22] M. Reed and B. Simon, Methods of Modern Mathematical Physics I, Academic Press,
1972.
[23] H. Weyl, Das asymptotisher Verteilungsgesetz der Eigenwerte lineare partialler Differen-
tialgleichungen, Math. Ann. 71 (1912) 441.
[24] M. Reed and B. Simon, Methods of Modern Mathematical Physics IV, chap. XIII.17,
Academic Press, New York, (1978).
[25] A. Fick, Über Diffusion, Poggendorff’s Annalen der Physik 94 (1855) 59.
[26] A. Einstein, Investigations on the Theory of the Brownian Movement Dover, New York,
1956; S. Chandrasekhar, Stochastic Problems in Physics and Astronomy, Rev. Mod. Phys.
15 (1943) 1.
[27] E.P. Wigner, On the quantum correction for thermodynamic equilibrium, Phys. Rev. 40
(1932) 749; J.G. Kirkwood, Quantum Statistics of Almost Classical Assemblies, Phys.
Rev. 44 (1933) 31.
[28] D. Fliegner, P. Haberl, M. Schmidt and C. Schubert, An improved heat kernel expansion
from worldline path integrals, Discourses Math. Appl. 4 (1995) 87.
————————————
A. Wipf, Path Integrals
BIBLIOGRAPHY Bibliography 159
[29] N. Metropolis, A.W. Rosenbluth, M.N. Rosenbluth, A.H. Teller and E. Teller, Equations
of state calculations by fast computing machines, J. Chem. Phys. 21 (1953) 1087.
[30] M. Creutz and B. Freedman, A statistical approach to quantum mechanics, Ann. Phys.
132 (1981) 421.
[31] M.E.J. Newman und G.G. Barkenna, Monte Carlo Methods in Statistical Physics, Claren-
don Press, Oxford, 1999.
[32] Markov Chain, Monte Carlo Simulations and Their Statistical Analysis, World Scientific,
Singapore, 2006.
[33] A. Berezin, The Method of Second Quantization, Academic Press, New York, 1966.
[36] E. Witten, Supersymmetry and Morse Theory, J. Diff. Geom. 17 (1982) 661.
[37] F. Cooper, A. Khare and U. Sukhatme, Supersymmetry Quantum Mechanics, Phys. Rep.
251 (1995) 267 and World Scientific, Singapore, 2001.
[38] G. Junker, Supersymmetric Methods in Quantum Mechanics and Statistical Physics, texts
and monographs in physics, Springer, Berlin 1996.
[41] H. Ezawa and J.R. Klauder, Fermion Without Fermions: The Nicolai Map Revisited, Prog.
Theor. Phys. 74 (1985) 904.
[43] S. Coleman, R. Jackiw and L. Susskind, Ann. Phys. 93 (1975) 267; L.S. Brown, Nuovo
Cimento 29 (1963) 617.
[44] J.H. Loewenstein and J.A. Swieca, Ann. Phys. 68 (1971) 172; A.K. Raina and G. Wanders,
Ann. Phys. 132 (1981) 404; A.Z. Capri and R. Ferrari, Nuovo Cimento A62 (1981) 273. P.
Becher, Ann. Phys. 146 (1983) 223
————————————
A. Wipf, Path Integrals
BIBLIOGRAPHY Bibliography 160
[45] N.K. Nielsen and B. Schroer, Nucl. Phys. B 120 (1977) 62; K.D. Rothe and J.A. Swieca,
Ann. Phys. 117 (1979) 382; M. Hortascu, K.D. Rothe and B. Schroer, Phys. Rev. D 20
(1979) 3203; N.V. Krasnikov et.al, Phys. Lett Phys. Lett. B97 (1980) 103; R. Roskies and
F. Schaposnik, Phys. Rev. D 23 (1981) 558.
[47] I. Sachs and A. Wipf, Finite Temperature Schwinger Model, Helv. Phys. Acta 65 (1992)
652.
[48] C. Bernard Phys. Rev. D9 (3312) 1974; Itzykson Zuber (QFT) or Birrell and Davies (QFT
in curved space time)
[49] R. Musto, L. O’Raifeartaigh and A. Wipf, The U(1)-Anomaly, the Non-Compact Index
Theorem, and the (Supersymmetric) BA-Effect, Phys. Lett 175 B (1986) 433; P. Forgacs,
L. O’Raifeartaigh and A. Wipf, Scattering Theory, U(1)-Anomaly and Index Theorems for
Compact and Non-Compact Manifolds, Nucl.Phys. B 293 (1987) 559
[51] J.D. Jackson, Classical Electrodynamics, J. Wiley and Sons, Inc., 1975
[53] G. Glashow, S. Weinberg and A. Salam in Rev. Mod. Phys. 52 (1980) 515 and ibid, 53
(1980) 539
[54] K. Huang, Quarks, Leptons and Gauge Fields, World Scientific (1982)
[55] L. O’Raifeartaigh, A. Wipf and H. Yoneyama, The Constraint Effective Potential, Nucl.
Phys. B271 (1986) 653
[56] Y. Fujimoto, A. Wipf and H. Yoneyama, Symmetry Restoration of Scalar Models at Finite
Temperature, Phys. Rev. D38 (1988) 2625
————————————
A. Wipf, Path Integrals
Index
161
INDEX Index 162
————————————
A. Wipf, Path Integrals
INDEX Index 163
Theorem
of Bochner, 71
of Kolmogorov, 70
of Kolmogorov-Prehorov, 71
thermal correlation functions, 73
thermal de Broglie wavelength, 79
time evolution kernel, 10
time ordering, 15
trace class, 54
Trotter product formula, 11
variance, 68
vector potential, 34
Wick rotation, 43
Wick theorem, 31
Wiener measure, 43
Wiener process, 65
Wightman function, 45
Wightman functions, 45
Wigner-Kirkwood expansion, 79
winding number, 37
Wronskian, 56
zeta-function, 59
————————————
A. Wipf, Path Integrals