Available online at www.sciencedirect.

com

ScienceDirect
Nuclear Physics B 960 (2020) 115183
www.elsevier.com/locate/nuclphysb

Path integral calculation of heat kernel traces with first

order operator insertions
Fiorenzo Bastianelli a,b,∗ , Francesco Comberiati a
a Dipartimento di Fisica e Astronomia, Università di Bologna, via Irnerio 46, I-40126 Bologna, Italy
b INFN, Sezione di Bologna, via Irnerio 46, I-40126 Bologna, Italy

Received 31 May 2020; received in revised form 23 July 2020; accepted 16 September 2020
Available online 21 September 2020
Editor: Stephan Stieberger

Abstract
We study generalized heat kernel coefficients, which appear in the trace of the heat kernel with an in-
sertion of a first-order differential operator, by using a path integral representation. These coefficients may
be used to study gravitational anomalies, i.e. anomalies in the conservation of the stress tensor. We use
the path integral method to compute the coefficients related to the gravitational anomalies of theories in
a non-abelian gauge background and flat space of dimensions 2, 4, and 6. In 4 dimensions one does not
expect to have genuine gravitational anomalies. However, they may be induced at intermediate stages by
regularization schemes that fail to preserve the corresponding symmetry. A case of interest has recently
appeared in the study of the trace anomalies of Weyl fermions.
© 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/). Funded by SCOAP3 .

1. Introduction

Heat kernel methods provide a useful tool for investigating QFTs. They were introduced by
Schwinger for studying QED processes [1] and extended to curved spaces and non-abelian gauge
fields by DeWitt [2]. There are many reviews and books dedicated to them, as [3–5].
One application of the heat kernel finds its place in the study of anomalies. The connection
is most easily seen by recalling Fujikawa’s method [6], which identifies the anomalies as arising

* Corresponding author.
E-mail address: [email protected] (F. Bastianelli).

https://doi.org/10.1016/j.nuclphysb.2020.115183
0550-3213/© 2020 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/). Funded by SCOAP3 .
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

from the non-invariance of the path integral measure under the symmetry transformations. In that
approach the anomalies are cast as regulated infinitesimal jacobians, limβ→0 Tr J e−β R , where J
is the generator of the anomalous symmetry and R a regulator, usually a second-order differential
operator. Once Wick-rotated to euclidean space the regulator R becomes an elliptic operator and
e−β R defines the associated heat kernel. Often the operator J depends only on the spacetime
coordinates, and it does not contain any differential operator. This is the case of the usual chiral
and trace anomalies.
Our interest in this paper is in traces that contain a first-order differential operator. This sit-
uation arises when one considers gravitational anomalies [7]. The latter are anomalies in the
conservation of the stress tensor, and the corresponding symmetry is the arbitrary change of coor-
dinates (diffeomorphisms). Diffeomorphisms are generated by the Lie derivative of the quantum
fields, and on scalars and Dirac spinors the Lie derivative Lξ takes the simple form
Lξ = ξ μ (x)∂μ (1)
where ξ μ (x) is the vector field due to an infinitesimal change of coordinates → (x μ = x μ
x μ − ξ μ (x)). The corresponding anomaly is then related to the regularization of an infinitesi-
mal Fujikawa jacobian of the form
 
J = ξ μ (x)∂μ + σ (x) δ D (x − y) (2)
where σ (x) is a function that depends on the regularization scheme adopted (often one takes
σ (x) = 12 ∂μ ξ μ (x) as a convenient choice, see [8,9]). An application of this type of traces has
recently appeared in [10].
In this paper, we shall study traces of the heat kernel with an insertion of a first-order dif-
ferential operator of the form given in (2) by using a quantum mechanical path integral. After
making explicit the relationship between the heat kernel traces and their path integral represen-
tation, we use the latter to evaluate the first three heat kernel coefficients for an elliptic operator
R containing a non-abelian gauge field Aμ (x) and an arbitrary matrix-valued scalar potential
V (x). These coefficients are a generalization of the standard heat kernel coefficients, also known
as Seeley-DeWitt coefficients, as they contain the insertion of a first-order differential operator.
We shall call them generalized heat kernel coefficients, for simplicity. Some of these coefficients
have been computed before in [11] and [12]. Here we shall reproduce some of those results with
the path integral method, and compute an additional one.
Our motivation for investigating these coefficients stems from a desire of addressing the
anomalies of a Weyl fermion in four dimensions by using a regularization scheme that induces
gravitational anomalies as well. This situation appeared in [13], where the trace anomaly of a
Weyl fermion in an abelian gauge background was computed to verify the absence of a parity
violating term, conjectured in [14] to be a possibility for CP violating theories. The use of a
Pauli-Villars (PV) regularization with a Majorana mass showed the absence of such a term, as
the PV Majorana mass preserves CP and diffeomorphism invariance. On the other hand, the ver-
ification of the same result with a PV Dirac mass could not be completed, as the latter induces
gravitational anomalies, which can be computed by using a generalized heat kernel coefficient in
the background of a non-abelian gauge field and flat spacetime. This justifies the use of flat space
that we consider in our analysis. The non-abelian background is however needed as the regula-
tor R contains gamma matrices, making the connection contained in R effectively non-abelian,
even for the simple case of a Weyl fermion coupled to a U (1) gauge field.
Thus, the problem we face in this paper is to study the path integral method to compute
generalized heat kernel coefficients, verifying the ones previously known and producing a new

2
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

one. In section 2 we review the path integral representation of the heat kernel and its traces.
We start by considering a simple elliptic operator R, interpreted as the quantum hamiltonian of
a non-relativistic particle in a scalar potential, and study how to insert an arbitrary function of
the particle coordinates inside the heat kernel trace. We discuss the role played by the propa-
gators defined either by the Dirichlet boundary conditions (DBC) or by the string inspired (SI)
method, which can be used equivalently for generating the perturbative expansion of the path
integral. Section 3 extends the previous set-up to include the insertion of a first-order differential
operator inside the heat kernel trace, and uses a more general hamiltonian R with a non-abelian
gauge potential Aμ and a matrix-valued scalar potential V . The corresponding particle action is
also matrix-valued, and the path integral contains a time ordering prescription to maintain gauge
covariance. In section 4 we present the first three generalized heat kernel coefficients, and in sec-
tion 5 we describe the calculation of the simplest one with the path integral method, reproducing
the result of [11]. Having verified the consistency of the method, in section 6 we proceed with
the calculation of two more heat kernel coefficients. The first one is the flat space limit of a more
general result originally obtained in [12], which is relevant for the gravitational anomalies in a
flat four-dimensional space. The last coefficient is new, and may be useful for the gravitational
anomalies of gauge theories in a flat six-dimensional space. After our conclusions, we report in
appendix A the worldline propagators defined by the Dirichlet boundary conditions and by the
string inspired method, in appendix B we use them for computing some simple Seeley-DeWitt
coefficients as a simple review of the path integral method, and in appendix C we report further
calculational details.

2. Path integral representation of heat kernel traces

The path integral representation was used to study the trace anomalies in [15] and [16], where
the object of interest was represented by a heat kernel trace of the form
 
Tr σ (x) e−β R (3)

with σ (x) an arbitrary function and R an elliptic differential operator. In this section we take as
guiding example the operator
1 1
R = − ∂ 2 + V (x) = p 2 + V (x) (4)
2 2
where ∂ 2 = ∂ μ ∂μ is the laplacian and pμ = −i∂μ the momentum operator in the coordinate
representation of quantum mechanics. R is directly interpreted as the hamiltonian of a non-
relativistic particle of unit mass in D dimensions, and the functional trace is understood as a
trace on the Hilbert space of the particle
   
dDx
∞

Tr σ (x) e−β R = d D x σ (x)x|e−β R |x = D
σ (x) an (x)β n (5)
(2πβ) 2
n=0

where x|e−β R |x is the transition amplitude for an euclidean time β with coinciding initial and
final points, i.e. the heat kernel at coinciding points. The evaluation of the latter for an arbitrary
potential V (x) is not known in closed form, but often one needs only its perturbative expansion
for small values of β, which gives rise to the Seeley-DeWitt coefficients an (x). The first few ones
are

3
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

a0 (x) = 1
a1 (x) = −V
1 1 (6)
a2 (x) = V 2 − ∂ 2 V
2 12
1 1 1 1 4
a3 (x) = − V 3 + V ∂ 2 V + ∂μ V ∂ μ V − ∂ V.
6 12 24 240
A way of computing them is to use the path integral representation of the transition amplitude
for the quantum mechanical model with hamiltonian R and euclidean action
β  
1 μ
S[x(t)] = dt ẋ ẋμ + V (x) . (7)
2
0
Then, using the equivalence of path integrals with canonical quantization, one may write
  
−β R
Tr σ (x) e = d D x0 σ (x0 )x0 |e−β R |x0 

  0
x(β)=x 
= D
d x0 σ (x0 ) Dx(t) e−S[x(t)] = Dx(t) σ (x(0)) e−S[x(t)]
x(0)=x0 P BC

(8)
where in the second line we have used the path integral representation of the transition ampli-
tude at coinciding points, and recognized that the additional integration over the point x0 , which
creates the trace, implements periodic boundary conditions (PBC). Thus one finds a path integral
on all loops with an insertion of the function σ (x(t)). The argument x μ (t) of the function σ is
μ
evaluated at t = 0, which corresponds to the base point x μ (0) = x0 of the parametrized loop,
μ
but it could be anywhere on the loop described by the function x (t) as a consequence of time
translational invariance. Now, given the relation (8), one can use the perturbative expansion of the
path integral in the euclidean time β to evaluate the heat kernel coefficients in (5) with worldline
propagators and Feynman diagrams.
This set-up was discussed in [15,16], where it was extended to curved space and non-abelian
gauge fields and used to rederive the trace anomalies of many field theories. Actually, the precise
observable available (the trace anomalies) could be used as a benchmark to construct well-defined
path integrals for particles in curved spaces, stressing the necessity of using precise regularization
schemes on the worldline, which must include well-defined but scheme dependent counterterms
[15–18,9]. A list of counterterms needed for sigma models with N supersymmetries in various
regularization schemes is given in [19], with the N = 4 case that has been applied in the recent
construction of the path integral for the graviton in first quantization [20–22].
A direct extension of the above construction to the case of an insertion of a first-order differ-
ential operator in the trace of the heat kernel may not seem immediate. A simple way of obtaining
the insertion is to exponentiate the corresponding operator, and view it as a source added to the
action. Then a derivative creates the required insertion. Let us check the formulae we get this
way for a scalar insertion and compare them with the set-up described above.
To start with, let us consider
   
∂ 
Tr σ (x) e−β R = Tr e−β R+λσ (x)  (9)
∂λ λ=0

4
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

Fig. 1. Loop with Dirichlet boundary conditions (DBC) at x0 .

where the trace guarantees that the insertion arising by acting with a derivative can be placed on
the left of the exponential. The exponentiation can be viewed as a deformation of the hamiltonian,
which in turns generates a modified euclidean action with V (x) → V (x) − βλ σ (x), so that

β  
1 μ λ
S[x] → Sλ [x] = dt ẋ ẋμ + V (x) − σ (x) . (10)
2 β
0
Using the path integral representation one finds
⎛ β ⎞
   1

Tr σ (x) e−β R = Dx ⎝ dt σ (x(t))⎠ e−S[x] . (11)
β
P BC 0
This formula is equivalent to the one obtained earlier in (8). The equivalence between the two
expressions is understood by invoking the time translational invariance of the one-point function
of the operator σ (x(t)), which may be substituted by its time average.
As a side result, this reformulation makes it clear how to use different worldline propagators
for obtaining the same heat kernel coefficients. In the set-up described by eq. (8), it is natural to
parametrize the quantum integration variables by
x(t) = x0 + q(t) (12)
with q(0) = q(β) = 0, thus defining Dirichlet boundary conditions (DBC) on the quantum fluc-
tuations q(t). They parametrize loops with a fixed base point x0 . The final integration over x0
produces all possible loops in target space, thus implementing the full periodic boundary con-
dition (PBC) prescription, see Fig. 1. The emerging quantum integration variables q(t) have a
perturbatively well-defined propagator, as fixed by Dirichlet boundary conditions. This was the
approach used in [15,16].
Alternatively, one may find it useful to employ the so-called “string-inspired” (SI) propagator
[23], obtained by setting again
x(t) = x0 + q(t) (13)
but now with the condition
β β
1
x0 = dt x(t) ⇒ dt q(t) = 0 (14)
β
0 0
where the zero mode x0 is the average position of the loop, see Fig. 2. The non-local constraint
on q(t) defines the SI propagator. Again, the final integration over x0 creates all loops in target
space.
As a preparation for our worldline calculations, we collect these propagators in appendix A,
and use them in appendix B for obtaining the Seeley-DeWitt coefficients of eq. (6) with a simple
perturbative path integral calculation.

5
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

Fig. 2. Loop with average position x0 (SI).

The previous set-up is easily generalized by coupling the model to curved space and to non-
abelian gauge fields. For the latter, the simplest strategy requires the use of a time ordering
prescription to exponentiate the action with the matrix-valued gauge field, a method already
employed in [11]. More elaborate methods that avoid the time ordering are also available [24],
and could be used as well. More general ways of factorizing the zero mode x0 of the periodic
functions x(t) can be found in [25] and [26].

3. Insertion of a first-order operator

In this section, we consider the insertion of a first-order differential operator inside the trace
of the heat kernel and construct a path integral representation for it.
To start with, let us consider a more general hamiltonian R, with a non-abelian connection
Aμ and a matrix-valued scalar potential V
1
R = − ∇2 + V , ∇μ = ∂μ + Aμ . (15)
2
The corresponding matrix-valued euclidean action for the point particle of coordinates x μ(t)
reads
β  
1 μ
S[x] = dt ẋ ẋμ + ẋ μ Aμ (x) + V (x) (16)
2
0
and its exponential appears in the path integral with a time ordering. The latter guarantees gauge
covariance as in the standard construction of Wilson lines. The heat kernel is thus computed by
the path integral on the particle coordinates x μ(t) as

−β R
e = Dx(t) T e−S[x(t)] (17)

where T denotes the time ordering along the worldline parametrized by t: upon the expansion
of the exponential one should place the matrices associated with earlier times on the right of
those associated with later times. The trace of the heat kernel is computed by periodic boundary
conditions with period β, x μ (β) = x μ (0), and further implementing a finite dimensional trace
(denoted by “tr”) on the vector space where the matrix-valued potentials Aμ and V act upon
  
Tr e−β R = tr Dx(t) T e−S[x(t)] . (18)
P BC

Next, we would like to insert on the left-hand side an operator of the form
 
J = ξ μ (x)∂μ + σ (x) (19)

where ξ μ (x) is an arbitrary vector field (we have in mind applications to diffeomorphism anoma-
lies) and σ (x) a matrix-valued function that we will choose appropriately to simplify the relation

6
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

with the path integral and keep gauge invariance manifest. The last contribution can be modi-
fied at will by adding a standard heat kernel trace with the insertion of a matrix-valued scalar
function.
As in the previous section, we modify the action and the hamiltonian by adding a source so
that a derivative on its coupling constant creates an insertion. The source term in the action must
have a coupling to ξ μ (x), which can be considered as an abelian gauge field, so we deform the
action as
β  
1 μ
S[x] → Sλ [x] = dt ẋ ẋμ + ẋ Aμ (x) + V (x) + λẋ ξμ (x)
μ μ
(20)
2
0
where λ is a coupling constant. By going through the canonical formalism, one finds that the
hamiltonian corresponding to the previous euclidean action is given by
1
H = π 2 + V (x) (21)
2
where the covariant momentum
πμ = pμ − iAμ (x) − iλξμ (x) (22)
becomes a covariant derivative ∇ μ upon quantization
πμ → −i∇ μ = −i(∂μ + Aμ (x) + λξμ (x)) . (23)
Fixing the ordering ambiguities to maintain gauge covariance, one finds a corresponding quan-
tum hamiltonian
 
1 2 1 λ2
Rλ = − ∇ + V = R − λ ξ (x)∇μ + (∂μ ξ (x)) − ξ 2 (x)
μ μ
(24)
2 2 2
and a deformed version of (18) may be written down
  
−β Rλ
Tr e = tr Dx T e−Sλ [x] . (25)
P BC

Taking a λ-derivative on both sides, and setting λ = 0, one finds on the left-hand side the insertion
of the operator
1
ξ μ ∇μ + (∂μ ξ μ ) (26)
2
and on the right-hand side its path integral realization
⎛ ⎞
    β
1 1
Tr ξ μ ∇μ + (∂μ ξ μ ) e−β R = tr Dx ⎝− dt ẋ μ ξμ (x)⎠ T e−S[x] (27)
2 β
P BC 0
which is the formula we were looking for.
The insertion on the path integral side may be simplified by using time translation invariance
on the worldline, and one may substitute
β
1 dx μ (t) dx μ (0)
− dt ξμ (x(t)) → − ξμ (x(0)) (28)
β dt dt
0

7
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

with the insertion evaluated at a point of the loop, chosen here as the initial point. In the DBC
μ
method for evaluating the path integral one can use x μ (0) = x0 , see Fig. 1. In the SI method one
μ
will have to set x μ (0) = x0 + q μ (0) instead, see Fig. 2.
To summarize, we have found that computing with the path integral the expectation value of
β
1 dx μ (0)
− dt ẋ μ ξμ (x) or − ξμ (x(0)) (29)
β dt
0

creates an insertion of the operator (19) with the matrix-valued function σ (x) fixed to be
1
σ (x) = ξ μ Aμ + (∂μ ξ μ ) . (30)
2
We will study generalized heat kernel coefficients corresponding to this particular insertion.
Other forms of σ (x) can be easily worked out.

4. Generalized heat kernel coefficients

Having found a path integral representation of the trace of the heat kernel with the insertion of
a first-order differential operator, we evaluate the corresponding heat kernel coefficients by using
the perturbative expansion in β of the path integral. It takes the form
    ∞
1 −β R dDx 
Tr ξ ∇μ + (∂μ ξ ) e
μ μ
= D
bn (x)β n (31)
2 (2πβ) 2 n=0
where the bn (x) are the generalized heat kernel coefficients which include at the linear order the
abelian vector field ξ μ . For the operator R in (15) we use the action in (16) and compute up to
order β 3 to find

b0 = 0
1
b1 = tr Gμν Fμν
24
1  
b2 = tr ∂ 2 Gμν Fμν − 20Gμν Fμν V (32)
480
1 3 5
b3 = tr ∂ 4 Gμν Fμν + Gμν Fμν F 2 + Gμν Fνρ F ρλ Fλμ + 30Gμν Fμν V 2
1440 28 4  
− 6Gμν ∇ 2 Fμν + 6∂ 2 Gμν Fμν + 8∂ λ Gμν ∇λ Fμν + 2∂μ Gμν ∇ λ Fλν + 6Gμν Fμλ F λ ν V

where Gμν = ∂μ ξν − ∂ν ξμ is the abelian field strength, Fμν = ∂μ Aν − ∂ν Aμ + [Aμ , Aν ] the non-
abelian field strength, and ∇μ the covariant derivative of Aμ . Of course, Gμν could be taken out
of the color trace “tr”. These coefficients are up to total derivatives, and we have freed V from
derivatives.
The coefficient b1 was given in [11] and b2 in [12], both including their coupling to gravity.
The coefficient b3 is new, as far as we know. In the next sections, we describe their explicit
evaluation through the perturbative expansion of the path integral.
We have used an abelian vector field ξ μ , which allows for simplifications in the above formu-
lae. For example, in b1 the term Gμν can be taken out of the color trace so that only the abelian
part of Fμν survives the trace. Similarly, one may simplify the other coefficients, or write them in

8
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

equivalent ways. These coefficients may also be generalized by considering a non-abelian vector
field ξ μ , as in [12], but we have chosen to keep it abelian for a direct application to the anomalies
in the conservation of the stress tensor.

5. Perturbative expansion

We now study the perturbative expansion of the path integral with PBC, i.e. considering world-
lines with the topology of a circle. Since the kinetic operator cannot be inverted on the circle, one
μ
has to factor out a zero mode x0 and split the path integration variable as
μ
x μ (t) = x0 + q μ (t) . (33)
This can be done using either the DBC method or the SI one, as explained earlier.
To start with we rescale the time t → τ = βt , so that τ ∈ [0, 1], and we find the following path
integral representation of the trace
  
1 −β R
Tr ξ ∇μ + (∂μ ξ ) e
μ μ
2
⎛ ⎞
  1
1
= d D x0 tr Dq ⎝− dτ q̇ μ ξμ (x0 + q)⎠ T e−S[x0 +q]
β (34)
DBC 0
 ⎛ 1
⎞ 
d D x0 1
= D
tr ⎝− dτ q̇ ξμ (x0 + q)⎠ T e int 0
μ −S [x +q]
(2πβ) 2 β
0

where dots indicate derivative with respect to τ , and angle brackets denote normalized averages
with the free action, 1 = 1. The expectation values are to be computed by Wick contracting
with the propagators in appendix A, and with the interaction vertex taking the form

1  
Sint [x0 + q] = dτ q̇ μ Aμ (x0 + q) + βV (x0 + q) . (35)
0

We now start computing at order β to get b1 . There are two contributions. The first one comes
from Taylor expanding ξμ and Aμ to first order in q μ , and gives for the right-hand side of (34)

 ⎛ 1
⎞⎛
1
⎞
d D x0 1
D
tr ⎝− ∂ν ξμ (x0 ) dτ q ν (τ )q̇ μ (τ )⎠ ⎝−∂β Aα (x0 ) dτ  q β (τ  )q̇ α (τ  )⎠ .
(2πβ) 2 β
0 0
(36)
Time ordering is not needed and the disconnected Wick contractions vanish

1
F1 = = dτ • (τ, τ ) = 0 . (37)
0

The remaining connected correlation function gives

9
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

 
q ν (τ )q̇ μ (τ )q β (τ  )q̇ α (τ  )c = β 2 δ νβ δ μα (τ, τ  ) • • (τ, τ  ) + δ να δ μβ • (τ, τ  ) • (τ, τ  )
(38)
where the first term corresponds to a worldline Feynman diagram of the form
1 1
1
F2 = = dτ dτ  (τ, τ  ) • • (τ, τ  ) = (39)
12
0 0
and the second one to a diagram of the form
1 1
1
F3 = = dτ dτ  • (τ, τ  ) • (τ, τ  ) = − . (40)
12
0 0
In drawing Feynman diagram we denote vertices by black dots and derivatives by white circles
on the legs. Integration by parts relates the two integrals, F2 = −F3 , hinting at gauge invariance.
The above values are obtained using equivalently the DBC or the SI propagators. In the latter
case one may use translational invariance to eliminate one integration. Thus, the trace inside (36)
reduces to
  β  
β tr ∂ ν ξ μ (x0 ) ∂ν Aμ (x0 )F2 + ∂μ Aν (x0 )F3 = tr ∂ ν ξ μ (x0 ) ∂ν Aμ (x0 ) − ∂μ Aν (x0 ) .
12
(41)
A second term of the same order in β arises from considering two interaction vertices and
has the effect of completing the non-abelian gauge invariance. Keeping the leading term of the
2 [x + q] one finds
Taylor expansion of the non-abelian potential inside 12 Sint 0

1 1
1
T dτ1 q̇ (τ1 )Aα (x0 ) dτ2 q̇ β (τ2 )Aβ (x0 ) .
α
(42)
2
0 0
The time ordering is implemented explicitly as

1 τ1 1 1
1 1
Aα (x0 )Aβ (x0 ) dτ1 dτ2 q̇ (τ1 )q̇ (τ2 ) + Aβ (x0 )Aα (x0 ) dτ1 dτ2 q̇ α (τ1 )q̇ β (τ2 )
α β
2 2
0 0 0 τ1
(43)
and simplified using a Heaviside step function (and renaming integration variables)
1 1
Aα (x0 )Aβ (x0 ) dτ1 dτ2 q̇ α (τ1 )q̇ β (τ2 )θ (τ1 − τ2 ) . (44)
0 0
Inserted into the right-hand side of (34) it leads to

1 1 1
(−1)
tr ∂ν ξμ (x0 )Aα (x0 )Aβ (x0 ) dτ dτ1 dτ2 q̇ μ (τ )q ν (τ )q̇ α (τ1 )q̇ β (τ2 ) θ (τ1 − τ2 )
β
0 0 0
(45)

10
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

with nonvanishing contractions that produce the following integrals

1 1 1
1
dτ dτ1 dτ2 • • (τ, τ1 ) • (τ, τ2 )θ (τ1 − τ2 ) =
12
0 0 0
(46)
1 1 1
1
dτ dτ1 dτ2 • • (τ, τ2 ) • (τ, τ1 )θ (τ1 − τ2 ) = −
12
0 0 0
independently of the propagator used. At the end one finds a commutator term
β
tr ∂ ν ξ μ (x0 )[Aν (x0 ), Aμ (x0 )] . (47)
12
These are all the terms of order β. Summing (41) and (47) one finds for (34)
 
d D x0 β ν μ d D x0 β
D
tr ∂ ξ (x )F
0 νμ 0(x ) = D
tr Gμν (x0 )Fμν (x0 ) (48)
(2πβ) 2 12 (2πβ) 2 24
which delivers the coefficient b1 of eq. (32). Gμν is the abelian field strength of ξμ and can be
taken out of the trace, showing that only the abelian part of Fμν contributes. The time ordered
diagram that leads to the commutator in (47) does not survive the trace, but we have presented it
to exemplify the role of the time ordering prescription.
As noted, the SI propagators are explicitly translational invariant and in the perturbative ex-
pansion one may eliminate an integration of the Feynman diagrams, fixing for example the
insertion at τ = 0. In the DBC method, translational invariance on the circle can be used as
well. However, the calculation proceeds somewhat differently. One uses translational invariance
μ
to fix the insertion at τ = 0 and identifies x μ (0) = x0 (since q μ (0) = 0 by DBC). Then, eq. (34)
simplifies to
   
1 d D x0 (−1)  μ 
Tr ξ μ ∇μ + (∂μ ξ μ ) e−β R = D μ
ξ (x0 ) tr q̇ (0)T e−Sint [x0 +q]
2 (2πβ) 2 β DBC

(49)
which delivers the result with the vector field ξμ (x0 ) explicitly factored out. To evaluate the
same coefficient b1 in this set-up, one needs to expand Aμ to higher orders. The calculation is
simplified by using the Fock-Schwinger gauge (more on this later)
1 1
Aν (x0 + q) = Aν (x0 ) + q ρ Fρν (x0 ) + q ρ q σ ∇σ Fρν (x0 ) + ... (50)
2 3
to find
(−1)  μ  1 
q̇ (0)T e−Sint [x0 +q] = q̇ μ (0)Sint [x0 + q] + · · ·
β DBC β DBC

 1
1
= ∇σ Fρν (x0 ) dτ q̇ μ (0)q̇ ν (τ )q ρ (τ )q σ (τ )DBC (51)
3β
0
β β
= ∇ν F νμ (x0 )(G1 − G2 ) = − ∇ν F νμ (x0 )
3 12
where the integrals with the DBC propagators give

11
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

1
1
G1 = X = dτ • • (0, τ ) (τ, τ ) = −
6
0
(52)
1
1
G2 = X = dτ • (0, τ ) • (τ, τ ) =
12
0
(the cross in the vertex singles out the vertex without integration). Considering the color trace in
(49) one may substitute the covariant derivative with the standard derivative and notice that only
the abelian part of F μν survives the trace, so that by setting b1 = b1μ ξ μ one finds
1
b1μ = − tr ∂ ν Fνμ (53)
12
that indeed reproduces b1 after integrating by part inside (49).
The coefficient b1 was known from ref. [11], where it was obtained by computing the heat
kernel trace with plane waves, but presented in a form that did not show manifest gauge invari-
ance. The previous calculation verifies the consistency of the path integral method, and one may
proceed with confidence to evaluate higher order coefficients.

6. Higher order coefficients

To get additional coefficients one needs to push the perturbative expansion in β to higher
orders. To proceed faster, we use gauge invariance and select the Fock-Schwinger (FS) gauge

q μ (τ )Aμ (x0 + q(τ )) = 0 (54)

which allows to expand the gauge potential in terms of its curvature (and derivatives thereof)
evaluated at x0 (see for example [27])
1 1 1
Aμ (x0 +q) = Aμ (x0 )+ q ν Fνμ (x0 )+ q ν q ρ ∇ρ Fνμ (x0 )+ q ν q ρ q σ ∇σ ∇ρ Fνμ (x0 )+... .
2 3 8
(55)
A similar gauge holds also for ξμ (x)
1 1 1
ξμ (x0 + q) = ξμ (x0 ) + q ν Gνμ (x0 ) + q ν q ρ ∂ρ Gνμ (x0 ) + q ν q ρ q σ ∂σ ∂ρ Gνμ (x0 ) + ... .
2 3 8
(56)
Next, we give some details on the calculation of the higher order coefficients b2 and b3 .
To identify b2 we need terms with one vertex (from Sint ) and two vertices (from Sint 2 ). Substi-

tuting the potentials in the FS gauge, both in the insertion and in the vertices, we get the following
contribution of order β 2 to eq. (34)

d D x0  
D
tr A1 (x0 ) + A2 (x0 ) + A3 (x0 ) + A4 (x0 ) (57)
(2πβ) 2
where the single vertex produces1

1 We now use a compact notation, indicating q ≡ q(τ ),

 1 1
0 0 01 ≡ 0 dτ0 0 dτ1 , etc.

12
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183


1 μ β γ
A1 (x0 ) = q̇0 q0ν q̇1α q1 q1 q1δ Gνμ (x0 )∇δ ∇γ Fβα (x0 ) (58)
16β
01

1 μ ρ β γ
A2 (x0 ) = q̇0 q0ν q0 q̇1α q1 q1 ∂ρ Gνμ (x0 )∇γ Fβα (x0 ) (59)
9β
01

1 μ ρ β
A3 (x0 ) = q̇0 q0ν q0 q0σ q̇1α q1 ∂σ ∂ρ Gνμ (x0 )Fβα (x0 ) (60)
16β
01

while the two interaction vertices contribute with


1 μ β
A4 (x0 ) = − q̇0 q0ν q̇1α q1 Gνμ (x0 )Fβα (x0 )V (x0 ) . (61)
4
01

In this last term we have used cyclicity of the trace in (57) to eliminate the time ordering. Then,
it describes a disconnected contribution that embeds b1 , so it is immediately evaluated to

β 2 μν
A4 = − G Fμν V . (62)
24
As for the remaining terms, since they enter eq. (57), we simplify them with integration by parts
(covariant derivatives acting on Fμν become usual derivatives acting on Gμν ). Then, renaming
the time variables we get
  
1 μ β γ
A1 + A3 = q̇0 q0ν q̇1α q1 q1 q1δ  ∂γ ∂δ Gνμ Fβα + ∂γ ∂δ Gβα Fνμ (63)
16β
01

and performing the Wick contractions (and also integrating by parts on the worldline to get rid
of • •01 in the Feynman diagrams) we get

β2 β2
A1 + A3 = H1 (∂ 2 Gμν Fμν + 2∂ α ∂β Gβν Fαν ) = H1 ∂ 2 Gμν Fμν (64)
4 2
where H1 is the Feynman diagram
 
1
• DBC
H1 = = 01 •01 11 = 60
1
(65)
144 SI .
01

We proceed similarly with A2 to get


1 μ ρ β γ
A2 = − q̇0 q0ν q0 q̇1α q1 q1 ∂γ ∂ρ Gνμ Fβα
9β
01 (66)
β 2 2 μν  9 1 1 
= ∂ G Fμν − H2 + H3 − H4 + 2H5
9 2 2 4
where

13
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

 
1
• DBC
H2 = = 01 01 •01 = 90
1
360 SI
01
 
− 180
1
DBC
H3 = = 00 • •01 11 =
0 SI
01
 (67)
 1
DBC
H4 = = 00 • 01 •11 = 360
0 SI
01
 
• − 720
1
DBC
H5 = = 00 01 •11 =
0 SI .
01

Adding all terms up we get

1 1
1 1 2  β 2 2 μν
A1 + A2 + A3 = β 2 ∂ 2 Gμν Fμν H1 − H2 + H3 − H4 + H5 = ∂ G Fμν
2 2 18 36 9 480
independently of the worldline propagators used. Including A4 we obtain the generalized coeffi-
cient b2
1  
b2 = tr ∂ 2 Gμν Fμν − 20Gμν Fμν V (68)
480
which correctly reproduces the one reported in [12] (with abelian ξ μ and in flat space).
Finally, we wish to compute the coefficient b3 , which did not appear in the literature so far. It
is more laborious, so we just present the calculation of a single term, i.e. the first one inside b3
of eq. (32), dumping details on the calculation of the remaining part into appendix C. This term
receives contributions from the Fμν dependence of a single Sint vertex insertion, which read

1 μ β γ η
B1 (x0 ) = q̇0 q0ν q̇1α q1 q1 q1δ q1 q1  Gνμ ∇η ∇ ∇δ ∇γ Fβα (69)
288β
01

1 μ ρ β γ
B2 (x0 ) = q̇0 q0ν q0 q̇1α q1 q1 q1δ q1  ∂ρ Gνμ ∇ ∇δ ∇γ Fβα (70)
90β
01

1 μ ρ β γ
B3 (x0 ) = q̇0 q0ν q0 q0σ q̇1α q1 q1 q1δ  ∂σ ∂ρ Gνμ ∇δ ∇γ Fβα (71)
64β
01

1 μ ρ β γ
B4 (x0 ) = q̇0 q0ν q0 q0σ q0τ q̇1α q1 q1  ∂τ ∂σ ∂ρ Gνμ ∇γ Fβα (72)
90β
01

1 μ ρ β
B5 (x0 ) = q̇0 q0ν q0 q0σ q0τ q0λ q̇1α q1  ∂λ ∂τ ∂σ ∂ρ Gνμ Fβα . (73)
288β
01

As they are integrated in spacetime, see eq. (34), we integrate by parts the covariant derivatives
from Fβα to Gνμ , where they become standard derivatives. Then collecting identical Wick con-
tractions we get

14
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

β 3 4 μν
B1 + B5 = − ∂ G Fμν I1
8
β3
B2 + B4 = − ∂ 4 Gμν Fμν (−15I2 − 3I3 + I4 − I5 ) (74)
30
β3
B3 = − ∂ 4 Gμν Fμν (2I6 + I7 + 2I8 )
8
where the integrals corresponding to the worldline Feynman diagrams are listed in appendix C.
Adding these terms together we get the following contribution to b3

1 dDx 3
D
tr ∂ 4 Gμν Fμν . (75)
1440 (2πβ) 2 28
In appendix C we report details on the calculation of the other terms contributing to b3 .

7. Conclusions

We have studied path integral methods to compute heat kernel traces with insertion of a first-
order differential operator. We have considered hamiltonians R with couplings to non-abelian
gauge fields and matrix-valued potentials only. The coupling to a curved space metric is however
straightforward, as path integrals on curved spaces are well-studied by now [9]. The insertion of
a first-order differential operator into the trace of the heat kernel has been obtained by modifying
the hamiltonian with a source coupled to the first-order operator, and then varying the source.
This procedure translates then into the path integral representation of the desired trace, which
we have used to calculate the first three generalized heat kernel coefficients. Alternatively, one
could have applied the variational procedure directly to the standard heat kernel coefficients, as
the source term is structurally similar to the gauge coupling already present in the hamiltonian.
Indeed, this was the method followed in [12]. In the present case our results can be checked, and
further extended to reach b4 and b5 , by a gauge variation of the coefficients already calculated
with worldline methods in [25]. We have verified the correctness of our calculation this way as
well.
Our interest in these particular traces stems from a desire to compute the anomalies in the
conservation of the stress tensor, which appear in four dimensions if one uses regularization
schemes that are not symmetric enough. Such a situation emerged in the study of a Weyl fermion
in a U(1) background once regulated with Pauli-Villars fields with Dirac mass [13]. The gravita-
tional anomaly emerging in this scheme was calculated using generalized heat kernel coefficients
in [28]. The study of the anomaly structure of chiral fermions in four dimensions has become
recently of renewed interest, in particular regarding the trace anomaly. The latter has been scruti-
nized from various perspectives [29,13,30–33] to verify the absence of the Pontryagin topological
density (in curved space) or Chern-Pontryagin topological density (for couplings to gauge fields).
The presence of these topological densities was conjectured to be a possibility in [14] (see also
[34] for a supersymmetric extension of the conjecture), and the analyses of refs. [35–37] claimed
their existence in the trace anomaly of a Weyl fermion in curved space. It seems useful to con-
sider these issues even within regularization schemes that induce anomalies in the conservation
of the stress tensor.
The methods presented here may be considered as part of a general strategy of using worldline
path integrals to obtain field theoretical results in flat [23] and curved space [38], a strategy often
referred to as the worldline formalism. These methods are quite efficient from a calculational
point of view, and it seems worthwhile to extend their development and applications.

15
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

CRediT authorship contribution statement

All authors have contributed with equal weight on the various steps involved in the research
reported in the manuscript.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal rela-
tionships that could have appeared to influence the work reported in this paper.

Appendix A. Worldline propagators

For perturbative computations in β we find it useful to rescale the time t → τ = βt , so that

τ ∈ [0, 1]. Then, β appear explicitly as a perturbative parameter multiplying suitably the various
terms of the action (7), which takes the form
1  
1 μ
S[x(τ )] = dτ ẋ ẋμ + βV (x) (76)
2β
0
μ
where now ẋ μ ≡ dx μ
dτ . We consider periodic boundary conditions for x (τ ), appropriate for cre-
ating a functional trace in the path integral. The kinetic term identifies the propagator, which then
carries a power of β, while the potential term is treated perturbatively. Setting
μ
x μ (τ ) = x0 + q μ (τ ) (77)
one finds a perturbative propagator for the quantum field q μ(τ ) of the form

q μ (τ )q ν (τ  ) = −βδ μν (τ, τ  ) (78)

d2
where (τ, τ  ) is the Green function of the operator dτ 2
that depends on the boundary condi-
μ d 2
tions and the way the zero mode is factored out (recall that the differential operator dτ
x0 2 is
not invertible on the space of periodic functions, the constant function has zero eigenvalue and
constitutes a zero mode of the operator).

Dirichlet boundary conditions

Using the Dirichlet boundary conditions (DBC), q μ (0) = q μ (1) = 0, one finds for (τ, τ  )
D (τ, τ  ) = (τ − 1)τ  θ (τ − τ  ) + (τ  − 1)τ θ (τ  − τ )
1 1 (79)
= |τ − τ  | − (τ + τ  ) + τ τ 
2 2
with the step function θ (τ ) defined such that θ (0) = 12 . It satisfies

d2
D (τ, τ  ) = δ(τ − τ  ) (80)
dτ 2
where the Dirac delta is the one appropriate for functions with vanishing boundary conditions.
For later use it is convenient to list the derivatives of the worldline propagator in DBC, where a
left/right dot indicates a derivative with respect to the first/second argument

16
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

•
D (τ, τ  ) = τ  − θ (τ  − τ )
•D (τ, τ  ) = τ − θ (τ − τ  )
• • (81)
D (τ, τ  ) = 1 − δ(τ − τ  )
••
D (τ, τ  ) = δ(τ − τ  )
with coincident points limits
D (τ, τ ) = τ 2 − τ
1 (82)
•
D (τ, τ ) = τ − .
2
String inspired propagator
The “string inspired” (SI) propagator for the quantum fluctuations q μ (τ ) satisfies periodic bound-
1
ary conditions but with the constraint 0 dτ q μ (τ ) = 0. Then, one finds for (τ, τ  )
1 1 1
SI (τ, τ  ) = SI (τ − τ  ) = |τ − τ  | − (τ − τ  )2 − (83)
2 2 12
which satisfies
d2
SI (τ − τ  ) = δ(τ − τ  ) − 1 . (84)
dτ 2
It has the useful property of being translational invariant. It is an even function of τ − τ  , and its
first derivative is odd which implies that its coincident points limit vanishes. Here is a list of its
properties

• 1
SI (τ − τ  ) = sgn(τ − τ  ) − (τ − τ  )
2
••
SI (τ − τ ) = δ(τ − τ  ) − 1

(85)
1
SI (0) = −
12
•
SI (0) = 0
where by sgn(x) we denote the sign function. Of course • SI (τ, τ  ) = − •SI (τ, τ  ).

Appendix B. Perturbative expansion and heat kernel coefficients

Here we compute the heat kernel coefficients given in (6) by evaluating the path integral in
(11). We present it as a review of worldline methods and to exemplify the equivalence of the
DBC and SI methods for treating the zero mode on the circle [25].
Let us first rewrite (11) by factoring out the zero mode integration, and set up the perturbative
expansion
⎛ β ⎞
   1

−β R
Tr σ (x) e = Dx ⎝ dt σ (x(t))⎠ e−S[x]
β
P BC 0
⎛ 1 ⎞
  
= d D x0 Dq ⎝ dτ σ (x0 + q(τ ))⎠ e−S[x0 +q] (86)
0

17
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

 1
d D x0  
= D
dτ σ (x0 + q(τ )) e−Sint [x0 +q]
(2πβ) 2
0

where we have rescaled the time in the insertion and action, with the latter taking the form
1 1
1 1
S[x0 + q] = S0 [q] + Sint [x0 + q] = dτ q̇ 2 (τ ) + β dτ V (x0 + q(τ )) . (87)
β 2
0 0

Normalized averages with the free path integral are denoted by angle brackets, 1 = 1, and we
 D
have extracted the overall normalization constant Dq e−S0 [q] = (2πβ)− 2 .
The perturbative expansion is implemented by Taylor expanding about x0 the function σ (x)
and the potential V (x), and further expanding the exponential of the interaction term. Next one
computes the correlation functions by Wick contractions. Keeping exponentiated the terms that
generate disconnected diagrams, and recalling that the propagators carry a factor of β, we find at
order β 3

 1 
dτ σ (x0 + q) e−Sint [x0 +q]
0
 
β2 β3 β3
= e−βV (x0 ) σ (x0 ) 1 + ∂ 2 V (x0 ) J1 − ∂ 4 V (x0 ) J2 − ∂μ V (x0 )∂ μ V (x0 ) J3
2 8 2
 3 
β μ 2
+ ∂μ σ (x0 ) β 2 ∂ μ V (x0 ) J3 − ∂ ∂ V (x0 ) J4
2
  
β β2 2
+ ∂ σ (x0 ) − J1 1 + ∂ V (x0 ) J1
2
2 2
 
β3
+ ∂μ ∂ν σ (x0 )∂ ∂ V (x0 ) − J5
μ ν
2
 3 
β
+ ∂μ ∂ 2 σ (x0 )∂ μ V (x0 ) − J4
2
2 
β
+ ∂ 4 σ (x0 ) J2 + O(β 4 )
8
(88)
where one should expand the exponential in front by keeping only the powers of β needed to
match the chosen perturbative order. The J’s denote the worldline Feynman diagrams, where
lines depict propagators and dots denote vertices which include an integration over the time
τ ∈ [0, 1]. They are as follows
 
− 16 DBC
J1 = = 00 =
− 12
1
SI
0
 
1
DBC
J2 = = 200 = 301
144 SI
0

18
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

 
− 12
1
DBC
J3 = = 01 = (89)
0 SI
01
 
1
DBC
J4 = = 01 11 = 60
0 SI
01
 
1
90 DBC
J5 = = 201 = 1
720 SI .
01
We have given their values both in DBC and SI. To verify explicitly the equivalence between
DBC and SI, we first plug in (88) in (86), perform an integration by parts to free the function σ
from derivatives, and drop total derivative terms. Comparing with (5) we recognize the following
Seeley-DeWitt coefficients

a0 (x) = 1
a1 (x) = −V
1
a2 (x) = V 2 + (J1 − J3 )∂ 2 V
2
1 1 1
a3 (x) = − V 3 + (J3 − J1 )V ∂ 2 V + (J3 − J1 )∂μ V ∂ μ V − (J2 + J21 − 4J4 + 2J5 )∂ 4 V
6 2 4
(90)
which reproduce those quoted in (6), independently of the propagator used. Note that the manifest
translational invariance of the SI method allows to get rid of one of the time integrations in
Feynman diagrams. One may use it to fix τ = 0 in the insertion, thus relating (11) to (8).
Alternatively, one could compute eq. (8) directly with the DBC method. The answer is en-
coded in the second line of eq. (88) (the one proportional to σ (x0 )), from which one extracts the
expected answer.

Appendix C. Evaluation of b3

Here we give additional details on the evaluation of b3 . First we list the diagrams needed for
evaluating the leading term discussed in section 6. The worldline diagrams have been computed
both in DBC and SI, which serves as a check on the final result
 
− 280
1
, DBC
I1 = 200 • 01 •01 =
− 1728 , SI
1
01
 
• • − 4201
, DBC
I2 = 01 01 01 11 =
− 4320 , SI
1
01
 
− 1260
1
, DBC
I3 = • 00 •01 211 =
0, SI
01
 
1
• • , DBC
I4 = 00 01 11 11 = 5040
0, SI
01

19
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

 
• • − 2520
1
, DBC
I5 = 00 01 11 11 = (91)
0, SI
01
 
• − 5040
1
, DBC
I6 = 00 •01 11 01 =
0, SI
01
 
− 5040
17
, DBC
I7 = 00 01 •01 11
•
=
− 1728
1
, SI
01
 
• − 560
1
, DBC
I8 = 01 •01 201 =
− 60480
11
, SI .
01

Let us now consider the other contributions. We made extensively use of the color trace and
partial integration (both in spacetime and on the worldline), as they allow to collect identical
Wick contractions. We start by considering the terms arising from the expansion of two vertices
Sint coupling Fμν to V

1 μ β γ
C1 (x0 ) = − q̇0 q0ν q̇1α q1 q1 q1δ Gνμ (x0 )∇δ ∇γ Fβα (x0 )V (x0 ) (92)
16
01

1 μ β γ
C2 (x0 ) = − q̇0 q0ν q̇1α q1 q2 q2δ Gνμ (x0 )Fβα (x0 )∇δ ∇γ V (x0 ) (93)
8
012

1 μ β γ
C3 (x0 ) = − q̇0 q0ν q̇1α q1 q1 q2δ Gνμ (x0 )∇γ Fβα (x0 )∇δ V (x0 ) (94)
6
012

1 μ ρ β γ
C4 (x0 ) = − q̇0 q0ν q0 q̇1α q1 q1 ∂ρ Gνμ (x0 )∇γ Fβα (x0 )V (x0 ) (95)
9
01

1 μ ρ β
C5 (x0 ) = − q̇0 q0ν q0 q0σ q̇1α q1 ∂ρ ∂σ Gνμ (x0 )Fβα (x0 )V (x0 ) (96)
16
01

1 μ ρ β γ
C6 (x0 ) = − q̇0 q0ν q0 q̇1α q1 q2 ∂ρ Gνμ (x0 )Fβα (x0 )∇γ V (x0 ) (97)
6
012

where, strictly speaking, the derivatives in the Taylor expansion of the potential are standard
derivative, but we have covariantized them anticipating the effect of the time ordering with inser-
tions of vertices with bare Aμ , as discussed in sec. 5 (see the example in (47)). In addition, we
have the terms with three vertices and containing the scalar potential

1 β γ μ
C7 (x0 ) = q̇0α q0 q̇1 q1δ q̇3 q3ν g012 Gνμ (x0 )Fβα (x0 )Fδγ (x0 )V (x0 ) (98)
8
0123

β μ β
C8 (x0 ) = q̇0 q0ν q̇1α q1  Gνμ (x0 )Fβα (x0 )V 2 (x0 ) (99)
8
01

where the function g012 contains step functions that take care of the time ordering

20
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

g012 = θ01 θ12 + θ12 θ20 + θ20 θ01 . (100)

In C8 the time ordering is not necessary, thanks to the color trace. It is then recognized as a
disconnected diagram that is calculated straightforwardly from previous results. Moving on to
the term with three non-abelian field strengths we find

1 μ β γ η
C9 (x0 ) = q̇0 q0ν q̇1α q1 q̇2 q2δ q̇3 q3 θ12 θ23 Gνμ (x0 )Fβα (x0 )Fδγ (x0 )Fη (x0 ) . (101)
16β
0123

Let us now evaluate these terms. For C1 we get

β 3 μν 2 β3
C1 = − G ∇ Fμν V H1 + Gμν Fνα F α μ V H1 (102)
4 4
For C2 we get

β3 
C2 = − ∂ 2 Gμν Fμν − Gμν ∇ 2 Fμν − ∂ α Gμν ∇α Fμν − 2∂α Gαν ∇ μ Fμν
2 
+ 2Gαν Fνμ F μ α V K1 (103)
β3  
− 2∂ α Gμν ∇α Fμν + ∂ 2 Gμν Fμν + Gμν ∇ 2 Fμν V F3 J1
4
with the new diagram
 
1
• • 360 , DBC
K1 = 01 12 02 = 1
(104)
012 720 , SI .

Next, C3 evaluates to
β 3  μν 2 
C3 = G ∇ Fμν + ∂ σ Gμν ∇σ Fμν V K2
2 (105)
β 3  μν 2 
+ G ∇ Fμν + 2∂α Gαμ ∇ σ Fσ μ − 4Gαμ Fμσ F σ α V K3
2
with
 
1
K2 = 01 •01 12
•
= 120 , DBC
0, SI
012
  (106)
1
, DBC
K3 = • 01 • 11 02 = 720
0, SI
012

while (95) and (96) produce

β3 ν β3
C4 = − ∂ Gνμ ∇α F αμ V H4 − ∂ λ Gμν ∇λ Fμν V H2
2 2 (107)
β 3 2 μν
C5 = − ∂ G Fμν V H1 .
4
For C6 we have

21
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

β 3  2 μν 
C6 (x0 ) = ∂ G Fμν + ∂ σ Gμν ∇σ Fμν V K2
2 (108)
β 3  2 μν 
+ ∂ G Fμν + 2∂λ Gλμ ∇ α Fαμ V K3 .
2
Finally, we consider C7 that mixes with the above terms

β 3 ρν
C7 = G Fνμ F μ ρ V (K4 − K5 + K6 − K7 ) (109)
4
and contains the following integrals
 
1
• 240 , DBC
K4 = 03 •01 •13 g012 = 1
0123 360 , SI
 
7
• • • 720 , DBC
K5 = 13 03 01 g012 = 1
0123 360 , SI
  (110)
1
• • 720 , DBC
K6 = 01 •03 13 g012 = 1
0123 720 , SI
 
• − 240
1
, DBC
K7 = 01 •03 • 13 g012 =
− 360
1
, SI .
0123

Collecting all the terms from C1 to C7 we find that they sum up to the entire second line of b3 in
(32), independently of the propagator used. We have performed integrations by parts to reduce
to a set of independent terms (in particular, we have left V free from derivatives).
As for C8 , since the time ordering can be neglected, we find that it is given by a disconnected
correlation functions that embeds the first piece of b2 and produces

β 3 μν β 3 μν
C8 = − G Fμν V 2 F3 = G Fμν V 2 (111)
4 48
which sits inside (32) as the last term of the first row of b3 .
Finally, we consider (101), that contains three non-abelian field strengths Fμν . In (101) we
kept the time ordering encoded in the step functions. However, one may note that the Wick
contractions produce terms that have no ordering ambiguities under the trace, implying that the
abelian limit contains precisely the same information. Thus, we are allowed to drop the step
functions and consider the equivalent (under the color trace) form

˜ 1 μ β γ η
C9 (x0 ) = q̇0 q0ν q̇1α q1 q̇2 q2δ q̇3 q3 Gνμ (x0 )Fβα (x0 )Fδγ (x0 )Fη (x0 ) . (112)
96β
0123

We compute it as

β 3 μν β3
C˜9 = G Fμν F 2 (F3 )2 + Gμν Fνρ F ρλ Fλμ K8 (113)
8 2
where the first term arises from disconnected diagrams with F3 = − 12
1
already given in (40), and

22
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183


• 1
K8 = 01 • 12 • 23 • 30 = (114)
720
0123
valid in DBC and SI, thus producing the second and third term of b3 in (32).

References

[1] J.S. Schwinger, On gauge invariance and vacuum polarization, Phys. Rev. 82 (1951) 664, https://doi.org/10.1103/
PhysRev.82.664.
[2] B.S. DeWitt, Dynamical theory of groups and fields, Conf. Proc. C 630701 (1964) 585; Les Houches Lect. Notes
13 (1964) 585.
[3] A.O. Barvinsky, G.A. Vilkovisky, The generalized Schwinger-DeWitt technique in gauge theories and quantum
gravity, Phys. Rep. 119 (1985) 1, https://doi.org/10.1016/0370-1573(85)90148-6.
[4] B.S. DeWitt, The global approach to quantum field theory. Vol. 1, 2, Int. Ser. Monogr. Phys. 114 (2003) 1.
[5] D.V. Vassilevich, Heat kernel expansion: user’s manual, Phys. Rep. 388 (2003) 279, https://doi.org/10.1016/j.
physrep.2003.09.002, arXiv:hep-th/0306138.
[6] K. Fujikawa, Path integral measure for gauge invariant fermion theories, Phys. Rev. Lett. 42 (1979) 1195, https://
doi.org/10.1103/PhysRevLett.42.1195.
[7] L. Alvarez-Gaume, E. Witten, Gravitational anomalies, Nucl. Phys. B 234 (1984) 269, https://doi.org/10.1016/0550-
3213(84)90066-X.
[8] K. Fujikawa, H. Suzuki, Path Integrals and Quantum Anomalies, Clarendon Press, Oxford UK, 2004.
[9] F. Bastianelli, P. van Nieuwenhuizen, Path Integrals and Anomalies in Curved Space, Cambridge University Press,
Cambridge UK, 2006.
[10] M. Kurkov, D. Vassilevich, Gravitational parity anomaly with and without boundaries, J. High Energy Phys. 1803
(2018) 072, https://doi.org/10.1007/JHEP03(2018)072, arXiv:1801.02049 [hep-th].
[11] F. Bastianelli, P. van Nieuwenhuizen, A. Van Proeyen, Superstring anomalies in the semilight cone gauge, Phys.
Lett. B 253 (1991) 67, https://doi.org/10.1016/0370-2693(91)91365-3.
[12] T.P. Branson, P.B. Gilkey, D.V. Vassilevich, Vacuum expectation value asymptotics for second order differential
operators on manifolds with boundary, J. Math. Phys. 39 (1998) 1040, https://doi.org/10.1063/1.532369, arXiv:
hep-th/9702178; Erratum: J. Math. Phys. 41 (2000) 3301.
[13] F. Bastianelli, M. Broccoli, On the trace anomaly of a Weyl fermion in a gauge background, Eur. Phys. J. C 79 (4)
(2019) 292, https://doi.org/10.1140/epjc/s10052-019-6799-z, arXiv:1808.03489 [hep-th].
[14] Y. Nakayama, CP-violating CFT and trace anomaly, Nucl. Phys. B 859 (2012) 288, https://doi.org/10.1016/j.
nuclphysb.2012.02.006, arXiv:1201.3428 [hep-th].
[15] F. Bastianelli, The path integral for a particle in curved spaces and Weyl anomalies, Nucl. Phys. B 376 (1992) 113,
https://doi.org/10.1016/0550-3213(92)90070-R, arXiv:hep-th/9112035.
[16] F. Bastianelli, P. van Nieuwenhuizen, Trace anomalies from quantum mechanics, Nucl. Phys. B 389 (1993) 53,
https://doi.org/10.1016/0550-3213(93)90285-W, arXiv:hep-th/9208059.
[17] J. de Boer, B. Peeters, K. Skenderis, P. van Nieuwenhuizen, Loop calculations in quantum mechanical nonlinear
sigma models with fermions and applications to anomalies, Nucl. Phys. B 459 (1996) 631, https://doi.org/10.1016/
0550-3213(95)00593-5, arXiv:hep-th/9509158.
[18] F. Bastianelli, K. Schalm, P. van Nieuwenhuizen, Mode regularization, time slicing, Weyl ordering and phase space
path integrals for quantum mechanical nonlinear sigma models, Phys. Rev. D 58 (1998) 044002, https://doi.org/10.
1103/PhysRevD.58.044002, arXiv:hep-th/9801105.
[19] F. Bastianelli, R. Bonezzi, O. Corradini, E. Latini, Extended SUSY quantum mechanics: transition amplitudes and
path integrals, J. High Energy Phys. 1106 (2011) 023, https://doi.org/10.1007/JHEP06(2011)023, arXiv:1103.3993
[hep-th].
[20] R. Bonezzi, A. Meyer, I. Sachs, Einstein gravity from the N = 4 spinning particle, J. High Energy Phys. 1810
(2018) 025, https://doi.org/10.1007/JHEP10(2018)025, arXiv:1807.07989 [hep-th].
[21] F. Bastianelli, R. Bonezzi, O. Corradini, E. Latini, One-loop quantum gravity from the N = 4 spinning particle, J.
High Energy Phys. 1911 (2019) 124, https://doi.org/10.1007/JHEP11(2019)124, arXiv:1909.05750 [hep-th].
[22] R. Bonezzi, A. Meyer, I. Sachs, A worldline theory for supergravity, arXiv:2004.06129 [hep-th].
[23] C. Schubert, Perturbative quantum field theory in the string inspired formalism, Phys. Rep. 355 (2001) 73, https://
doi.org/10.1016/S0370-1573(01)00013-8, arXiv:hep-th/0101036.
[24] F. Bastianelli, R. Bonezzi, O. Corradini, E. Latini, Particles with non Abelian charges, J. High Energy Phys. 10
(2013) 098, https://doi.org/10.1007/JHEP10(2013)098, arXiv:1309.1608 [hep-th].

23
F. Bastianelli and F. Comberiati Nuclear Physics B 960 (2020) 115183

[25] D. Fliegner, P. Haberl, M.G. Schmidt, C. Schubert, The higher derivative expansion of the effective action by the
string inspired method. Part 2, Ann. Phys. 264 (1998) 51, https://doi.org/10.1006/aphy.1997.5778, arXiv:hep-th/
9707189.
[26] F. Bastianelli, O. Corradini, A. Zirotti, BRST treatment of zero modes for the worldline formalism in curved space, J.
High Energy Phys. 01 (2004) 023, https://doi.org/10.1088/1126-6708/2004/01/023, arXiv:hep-th/0312064 [hep-th].
[27] U. Muller, C. Schubert, A.M. van de Ven, A closed formula for the Riemann normal coordinate expansion, Gen.
Relativ. Gravit. 31 (1999) 1759–1768, https://doi.org/10.1023/A:1026718301634, arXiv:gr-qc/9712092 [gr-qc].
[28] S. Montefiori, Anomalies in the stress tensor of a chiral fermion in a gauge background, tesi di Laurea Magistrale,
Università di Bologna, 2019.
[29] F. Bastianelli, R. Martelli, On the trace anomaly of a Weyl fermion, J. High Energy Phys. 11 (2016) 178, https://
doi.org/10.1007/JHEP11(2016)178, arXiv:1610.02304 [hep-th].
[30] H. Godazgar, H. Nicolai, A rederivation of the conformal anomaly for spin- 12 , Class. Quantum Gravity 35 (10)
(2018) 105013, https://doi.org/10.1088/1361-6382/aaba97, arXiv:1801.01728 [hep-th].
[31] M.B. Fröb, J. Zahn, Trace anomaly for chiral fermions via Hadamard subtraction, J. High Energy Phys. 10 (2019)
223, https://doi.org/10.1007/JHEP10(2019)223, arXiv:1904.10982 [hep-th].
[32] F. Bastianelli, M. Broccoli, Weyl fermions in a non-Abelian gauge background and trace anomalies, J. High Energy
Phys. 10 (2019) 241, https://doi.org/10.1007/JHEP10(2019)241, arXiv:1908.03750 [hep-th].
[33] F. Bastianelli, M. Broccoli, Axial gravity and anomalies of fermions, Eur. Phys. J. C 80 (3) (2020) 276, https://
doi.org/10.1140/epjc/s10052-020-7782-4, arXiv:1911.02271 [hep-th].
[34] K. Nakagawa, Y. Nakayama, CP-violating super Weyl anomaly, arXiv:2002.01128 [hep-th].
[35] L. Bonora, S. Giaccari, B. Lima de Souza, Trace anomalies in chiral theories revisited, J. High Energy Phys. 07
(2014) 117, https://doi.org/10.1007/JHEP07(2014)117, arXiv:1403.2606 [hep-th].
[36] L. Bonora, M. Cvitan, P. Dominis Prester, A. Duarte Pereira, S. Giaccari, T. Štemberga, Axial gravity, massless
fermions and trace anomalies, Eur. Phys. J. C 77 (8) (2017) 511, https://doi.org/10.1140/epjc/s10052-017-5071-7,
arXiv:1703.10473 [hep-th].
[37] L. Bonora, M. Cvitan, P. Dominis Prester, S. Giaccari, M. Paulišic, T. Štemberga, Axial gravity: a non-perturbative
approach to split anomalies, Eur. Phys. J. C 78 (8) (2018) 652, https://doi.org/10.1140/epjc/s10052-018-6141-1,
arXiv:1807.01249 [hep-th].
[38] F. Bastianelli, A. Zirotti, Worldline formalism in a gravitational background, Nucl. Phys. B 642 (2002) 372–388,
https://doi.org/10.1016/S0550-3213(02)00683-1, arXiv:hep-th/0205182 [hep-th].

24