Welcome to visit Communications in Theoretical Physics,
Particle Physics and Quantum Field Theory

Ordered exponential and its features in Yang–Mills effective action

  • A V Ivanov , 1, 2, ,
  • N V Kharuk , 1, 2
Expand
  • 1St. Petersburg Department of Steklov Mathematical Institute of Russian Academy of Sciences, 27 Fontanka, St. Petersburg 191023, Russia
  • 2Leonhard Euler International Mathematical Institute, 10 Pesochnaya nab., St. Petersburg 197022, Russia

Author to whom all correspondence should be addressed.

Received date: 2023-02-03

  Revised date: 2023-04-13

  Accepted date: 2023-06-14

  Online published: 2023-08-08

Copyright

© 2023 Institute of Theoretical Physics CAS, Chinese Physical Society and IOP Publishing

Abstract

In this paper, we discuss some non-trivial relations for ordered exponentials on smooth Riemannian manifolds. As an example of application, we study the dependence of the four-dimensional quantum Yang–Mills effective action on the special gauge transformation with respect to the background field. Also, we formulate some open questions about a structure of divergences for a special type of regularization in the presence of the background field formalism.

Cite this article

A V Ivanov , N V Kharuk . Ordered exponential and its features in Yang–Mills effective action[J]. Communications in Theoretical Physics, 2023 , 75(8) : 085202 . DOI: 10.1088/1572-9494/acde4e

1. Introduction

Ordered exponentials play a crucial role in mathematical and theoretical physics, see [1, 2], because they have explicit geometrical and physical meanings. Indeed, on the one hand, they appear naturally in the differential geometry, see formulae (10.13a) and (10.17) in [3], as a solution for the ‘parallel transport' equation on principle bundles. On the other hand side, they allow us to make a transition to the Fock–Schwinger gauge condition in the non-Abelian gauge theories [4], and, hence, they are elements of the gauge transformation group. Also, formally, from the mathematical point of view, the ordered exponential is a product integral, which appears in different applications quite frequently [5]. Of course, there are a lot of ways to apply the exponentials and an ordering itself, including the theory of integrable models [6], the theory of the heat kernel method [79], and many others [10].
In this work, we present some non-trivial properties of the ordered exponential on a smooth Riemannian manifold in the case of a compact semisimple Lie group, see [4, 8, 11]. We apply these properties to the pure four-dimensional Yang–Mills theory [1214] to make some useful conclusions about the properties of the effective action regarding its dependence on the special gauge transformation with respect to the background field. We show that actually, the action depends on the field stress tensor and its covariant derivatives. Also, we formulate several crucial questions about the structure of divergences for a special type of regularization in the presence of the background field formalism, described in section 4. In particular, this would be useful for the renormalization scenario [1517] in the case of a cutoff regularization.
The paper has the following structure. In section 2, we give the problem statement and formulate the basic properties of the ordered exponentials. Then, in section 3, we present some additional non-trivial properties and prove them. In section 4, we discuss the application of the obtained equalities to the four-dimensional Yang–Mills effective action and formulate actual problems for further work. In the conclusion, we give some useful remarks and discuss the text of the work and alternative proofs.

2. Ordered exponentials

Let $G$ be a compact semisimple Lie group, and ${\mathfrak{g}}$ is its Lie algebra, see [11]. Let ${t}^{a}$ be the generators of the algebra ${\mathfrak{g}},$ where $\left(a=1,\ldots ,\dim {\mathfrak{g}}\right),$ such that the relations hold
$\begin{eqnarray}[{t}^{a},{t}^{b}]={f}^{abc}{t}^{c},\,{\rm{tr}}({t}^{a}{t}^{b})=-2{\delta }^{ab},\end{eqnarray}$
where ${f}^{abc}$ are antisymmetric structure constants for ${\mathfrak{g}},$ and ‘${\rm{tr}}$' is the Killing form. We work with an adjoint representation. It is easy to verify that the structure constants have the following crucial property
$\begin{eqnarray}{f}^{abc}{f}^{aef}={f}^{abf}{f}^{aec}-{f}^{acf}{f}^{aeb}.\end{eqnarray}$
Let $(M,g)$ be a smooth Riemannian manifold of dimension $\dim M=d\in {\mathbb{N}}.$ Then, we introduce two elements $x,y\in U,$ where $U\subset M$ is a smooth convex open domain, and Greek letters denote the coordinate components. Of course, we assume that the metric components form the symmetric matrix-valued operator, so we have ${g}^{\mu \nu }(x)={g}^{\nu \mu }(x).$ The corresponding local formula for the Christoffel symbols of the second kind can be written as
$\begin{eqnarray}{{\rm{\Gamma }}}_{\mu \nu }^{\rho }(x)=\frac{{g}^{\rho \alpha }(x)}{2}\left(\frac{\partial {g}_{\alpha \mu }(x)}{\partial {x}^{\nu }}+\frac{\partial {g}_{\alpha \nu }(x)}{\partial {x}^{\mu }}-\frac{\partial {g}_{\mu \nu }(x)}{\partial {x}^{\alpha }}\right).\end{eqnarray}$
Further, by symbol ${B}_{\mu }(x)={B}_{\mu }^{a}(x){t}^{a},$ where ${B}_{\mu }(\cdot )\in {C}^{\infty }(U,{\mathfrak{g}})$ for all values of $\mu ,$ we define the components of a Yang–Mills connection 1-form, see [3]. The operator ${B}_{\mu }(x)$ as an element of the Lie algebra acts as a commutator according to the adjoint representation. Hence, from now we consider ${B}_{\mu }(x)$ as the matrix-valued operator with the components ${f}^{adb}{B}_{\mu }^{d}\left(x\right).$
For the purposes of the section, we need to introduce several additional objects. First of all, we describe a number of derivatives. Let ${h}_{{\nu }_{1}\ldots {\nu }_{n}}^{{\mu }_{1}\ldots {\mu }_{k}}(\cdot )\in {C}^{1}(U,{\mathfrak{g}})$ be a tensor-valued operator, and ${h}_{{\nu }_{1}\ldots {\nu }_{n}}^{{\mu }_{1}\ldots {\mu }_{k}\,ab}(x)$ be its matrix components at the point $x,$ then we define
$\begin{eqnarray}\begin{array}{l}{\nabla }_{{x}^{\rho }}{h}_{{\nu }_{1}\ldots {\nu }_{n}}^{{\mu }_{1}\ldots {\mu }_{k}}(x)={\partial }_{{x}^{\rho }}{h}_{{\nu }_{1}\ldots {\nu }_{n}}^{{\mu }_{1}\ldots {\mu }_{k}}(x)\\ \,+\displaystyle \sum _{i=1}^{k}{{\rm{\Gamma }}}_{\rho \alpha }^{{\mu }_{i}}(x){h}_{{\nu }_{1}\ldots \ldots ..{\nu }_{n}}^{{\mu }_{1}\ldots \alpha \ldots {\mu }_{k}}(x)-\displaystyle \sum _{i=1}^{n}{{\rm{\Gamma }}}_{\rho {\nu }_{i}}^{\alpha }(x){h}_{{\nu }_{1}\ldots \alpha \ldots {\nu }_{n}}^{{\mu }_{1}\ldots \ldots ..{\mu }_{k}}(x),\end{array}\end{eqnarray}$
where $\alpha $ stands instead of the corresponding $i$th index, and
$\begin{eqnarray}{\mathop{D}\limits^{\longrightarrow }}_{{x}^{\rho }}^{{ab}}{h}_{{\nu }_{1}\ldots {\nu }_{n}}^{{\mu }_{1}\ldots {\mu }_{k}\,{bc}}(x)={{\rm{\nabla }}}_{{x}^{\rho }}{h}_{{\nu }_{1}\ldots {\nu }_{n}}^{{\mu }_{1}\ldots {\mu }_{k}\,{ac}}(x)+{f}^{{adb}}{B}_{\rho }^{d}(x){h}_{{\nu }_{1}\ldots {\nu }_{n}}^{{\mu }_{1}\ldots {\mu }_{k}\,{bc}}(x),\end{eqnarray}$
$\begin{eqnarray}{h}_{{\nu }_{1}\ldots {\nu }_{n}}^{{\mu }_{1}\ldots {\mu }_{k}\,{ab}}(x){\overleftarrow{D}}_{{x}^{\rho }}^{{bc}}={{\rm{\nabla }}}_{{x}^{\rho }}{h}_{{\nu }_{1}\ldots {\nu }_{n}}^{{\mu }_{1}\ldots {\mu }_{k}\,{ac}}(x)-{h}_{{\nu }_{1}\ldots {\nu }_{n}}^{{\mu }_{1}\ldots {\mu }_{k}\,{ab}}(x){f}^{{bdc}}{B}_{\rho }^{d}(x).\end{eqnarray}$
The main object of the paper, as it was noted in the introduction, is a path-ordered exponential. Let us give an appropriate definition by the following formula
$\begin{eqnarray*}\begin{array}{l}{{\rm{\Phi }}}^{ab}(x,y)={\delta }^{ab}+\displaystyle \sum _{k=1}^{+\infty }{(-1)}^{k}{\int }_{{u}_{0}}^{{u}_{1}}{\rm{d}}{s}_{1}\ldots \\ \times \,{\int }_{{u}_{0}}^{{s}_{k-1}}{\rm{d}}{s}_{k}\,{\dot{\gamma }}^{{\mu }_{1}}({s}_{1})({f}^{a{d}_{1}{c}_{1}}{B}_{{\mu }_{1}}^{{d}_{1}}(\gamma ({s}_{1})))\end{array}\end{eqnarray*}$
$\begin{eqnarray}\ldots \,{\dot{\gamma }}^{{\mu }_{k}}({s}_{k})({f}^{{c}_{k-1}{d}_{k}b}{B}_{{\mu }_{k}}^{{d}_{k}}(\gamma ({s}_{k}))),\end{eqnarray}$
where $\gamma (\cdot ):[{u}_{0},{u}_{1}]\to U$ is the parameterized geodesic, see [2], such that $\gamma ({u}_{0})=y$ and $\gamma ({u}_{1})=x,$ and the dot $\dot{\gamma }$ denotes the derivative ${\rm{d}}\gamma (s)/{\rm{d}}s$ by the parameter of parametrization. We note that the geodesic satisfies the following differential equation
$\begin{eqnarray}{\mathop{\gamma }\limits^{{\unicode{x02025}}}}^{\rho }(s)+{{\rm{\Gamma }}}_{\mu \nu }^{\rho }(\gamma (s)){\dot{\gamma }}^{\mu }(s){\dot{\gamma }}^{\nu }(s)=0\,\,{\rm{for}}\,{\rm{all}}\,\,s\in [{u}_{0},{u}_{1}].\end{eqnarray}$
Such type of operators have some useful properties, which can be formulated in the form
$\begin{eqnarray}{{\rm{\Phi }}}^{ab}(x,z){{\rm{\Phi }}}^{bc}(z,y)={{\rm{\Phi }}}^{\mathrm{ac}}({\rm{x}},{\rm{y}}),\,{({{\rm{\Phi }}}^{-1})}^{ab}(x,y)={{\rm{\Phi }}}^{ab}(y,x)={{\rm{\Phi }}}^{ba}(x,y),\,{{\rm{\Phi }}}^{ab}(y,y)={\delta }^{ab},\end{eqnarray}$
where the points $x,y,z\in U$ lie on the same geodesic. In other words, it means there is such a point $s\in {\mathbb{R}}$ that the equality $\gamma (s)=z$ holds. The proofs of the properties described above can be found in [4, 7, 8].
Then, we need to introduce Synge's world function $\sigma (x,y),$ see [18], as half the square of the geodesic length from $x$ to $y.$ It is a symmetric two-point scalar. Let us define some convenient abbreviations, suggested in Synge's monograph [19], as follows
$\begin{eqnarray}{\sigma }_{\rho }(x,y)={\partial }_{{x}^{\rho }}\sigma (x,y),\,{\sigma }^{\rho }(x,y)={g}^{\rho \nu }(x){\sigma }_{\nu }(x,y),\end{eqnarray}$
$\begin{eqnarray}{\sigma }_{\rho ^{\prime} }(x,y)={\partial }_{{y}^{\rho ^{\prime} }}\sigma (x,y),\,{\sigma }^{\rho ^{\prime} }(x,y)={\sigma }_{\nu ^{\prime} }(x,y){g}^{\nu ^{\prime} \rho ^{\prime} }(y),\end{eqnarray}$
where the symbol ‘$^{\prime} $' means that we work with the use of the second argument of a two-point function. Let us note that if $U$ is a domain of ${{\mathbb{R}}}^{d}$ with ${g}^{\mu \nu }(x)={\delta }^{\mu \nu },$ then Synge's world function $\sigma (x,y)$ is equal to $| x-y{| }^{2}/2.$
In the rest of the paper, in all derivatives (4)–(6) we write $\rho $ and $\rho ^{\prime} $ instead of ${x}^{\rho }$ and ${y}^{\rho ^{\prime} },$ respectively, because it does not cause any confusion. Therefore, we can formulate differential equations for exponential (7) in the form
$\begin{eqnarray}{\sigma }^{\rho }(x,y){\overrightarrow{D}}_{{\rho }^{}}^{{ab}}{{\rm{\Phi }}}^{{bc}}(x,y)=0,{{\rm{\Phi }}}^{{ab}}(x,y){\overleftarrow{D}}_{{\rho ^{\prime} }^{}}^{{bc}}{\sigma }^{{\rho ^{\prime} }^{}}(x,y)=0.\end{eqnarray}$
The proof can be achieved by the direct differentiation of (7) and the integration by parts, see the papers mentioned above.

3. Additional properties

In this section, we present a number of formulae for representing the path-ordered exponential and some additional properties, which are useful in gauge theories. We start with two series representations, which actually generalize the covariant Taylor representation from [20] on the case of non-zero Yang–Mills connection components.

Let $x,y\in U$, and $1$ denote the unit function. Then, under the conditions described above, we have

$\begin{eqnarray}{{\rm{\Phi }}}^{{ab}}(x,y)={\delta }^{{ab}}+\displaystyle \sum _{k=1}^{+\infty }\displaystyle \frac{{\left(-1\right)}^{k}}{k!}{\sigma }^{{\mu }_{1}^{{\rm{{\prime} }}}}(x,y)\ldots {\sigma }^{{\mu }_{k}^{{\rm{{\prime} }}}}(x,y)(1{\overleftarrow{D}}_{{\mu }_{1}^{{\rm{{\prime} }}}}^{a{c}_{1}}{\overleftarrow{D}}_{{\mu }_{2}^{{\rm{{\prime} }}}}^{{c}_{1}{c}_{2}}\ldots {\overleftarrow{D}}_{{\mu }_{k}^{{\rm{{\prime} }}}}^{{c}_{k-1}b}),\end{eqnarray}$
$\begin{eqnarray}{{\rm{\Phi }}}^{{ab}}(x,y)={\delta }^{{ab}}+\displaystyle \sum _{k=1}^{+\infty }\displaystyle \frac{{\left(-1\right)}^{k}}{k!}{\sigma }^{{\mu }_{k}}(x,y)\ldots {\sigma }^{{\mu }_{1}}(x,y)({\mathop{D}\limits^{\longrightarrow }}_{{\mu }_{k}}^{a{c}_{k-1}}\ldots {\mathop{D}\limits^{\longrightarrow }}_{{\mu }_{2}}^{{c}_{2}{c}_{1}}{\mathop{D}\limits^{\longrightarrow }}_{{\mu }_{1}}^{{c}_{1}b}1).\end{eqnarray}$

It is obvious that the initial condition ${{\rm{\Phi }}}^{ab}(y,y)={\delta }^{ab}$ holds, because ${\sigma }^{\rho }(x,y)$ and ${\sigma }^{\rho }\left(x,y\right)$ tend to zero when x → y. Let us check that the right hand side of (13) satisfies both equations from (12). We start with the second one. Let us apply the operator ${\mathop{D}\limits^{\longrightarrow}}_{\rho }^{bc}{\sigma }^{\rho }\left(x,y\right).$ Due to the fact that the derivative acts at the point $y,$ we get the answer instantly after using the following additional property of Synge's world function

$\begin{eqnarray}\begin{array}{l}{\sigma }^{\rho ^{\prime} }(x,y){\partial }_{{y}^{\rho ^{\prime} }}{\sigma }^{{\mu }_{i}^{{\prime} }}(x,y)={\sigma }^{{\mu }_{i}^{{\prime} }}(x,y)\\ \,+\,{\sigma }^{\rho ^{\prime} }(x,y){{\rm{\Gamma }}}_{\rho ^{\prime} v^{\prime} }^{{\mu }_{i}^{{\prime} }}(y){\sigma }^{v^{\prime} }(x,y),\end{array}\end{eqnarray}$
which actually follows from differentiation of $2\sigma (x,y)\,={\sigma }^{\rho ^{\prime} }(x,y){\sigma }_{\rho ^{\prime} }(x,y),$ see formula (2.31) in [19].
Let us move on to the first equation from (12). In this case, we have the derivative at the point $x.$ So we are going to use the covariant Taylor expansion in the form
$\begin{eqnarray}\begin{array}{c}\,{\sigma }^{\rho }(x,y){B}_{\rho }^{d}(x)=\displaystyle \sum _{k=1}^{+\infty }\displaystyle \frac{{\left(-1\right)}^{k}}{\left(k-1)!\right.}{\sigma }^{{\mu }_{k}^{{\prime} }}(x,y)\ldots {\sigma }^{{\mu }_{1}^{{\prime} }}(x,y){{\rm{\nabla }}}_{{\mu }_{k}^{{\prime} }}\ldots {{\rm{\nabla }}}_{{\mu }_{2}^{{\prime} }}{B}_{{\mu }_{1}^{{\prime} }}^{d}(y).\end{array}\end{eqnarray}$
It does not follow from ordinary covariant expansion, because the last formula contains both functions, ${\sigma }^{\rho }(x,y)$ and ${\sigma }^{\rho ^{\prime} }(x,y),$ and we need to decompose the first function in terms of another one. The simplest way to do this is to use the formalism of geodesic lines, successfully applied to prove the covariant expansions, see appendix in [20]. First of all, let us rewrite the derivatives of Synge's world function as
$\begin{eqnarray}\begin{array}{l}{\sigma }^{\rho }(x,y)=({u}_{1}-{u}_{0}){\dot{\gamma }}^{\rho }({u}_{1}),\,{\sigma }^{\rho ^{\prime} }(x,y)\\ \,=\,-({u}_{1}-{u}_{0}){\dot{\gamma }}^{\rho ^{\prime} }({u}_{0}).\end{array}\end{eqnarray}$
Hence, using the standard Taylor expansion, we get the chain of equalities
$\begin{eqnarray}{\sigma }^{\rho }(x,y){B}_{\rho }^{d}(x)=({u}_{1}-{u}_{0}){\dot{\gamma }}^{\rho }({u}_{1}){B}_{\rho }^{d}(\gamma ({u}_{1}))\end{eqnarray}$
$\begin{eqnarray}=\displaystyle \sum _{k=1}^{+\infty }\displaystyle \frac{{({u}_{1}-{u}_{0})}^{k}}{(k-1)!}\left(\displaystyle \frac{{{\rm{d}}}^{k-1}}{{\rm{d}}{t}^{k-1}}{\dot{\gamma }}^{\rho }(t){B}_{\rho }^{d}(\gamma (t))\right){\Space{0ex}{3.25ex}{0ex}| }_{t={u}_{0}}.\end{eqnarray}$
The expression in the large parentheses can be transformed in the manner (an analog of (4.112) from [20])
$\begin{eqnarray}\left(\displaystyle \frac{{d}^{k-1}}{d{t}^{k-1}}{\dot{\gamma }}^{\rho }(t){B}_{\rho }^{d}(\gamma (t))\right){\Space{0ex}{3.25ex}{0ex}|}_{t={u}_{0}}={\dot{\gamma }}^{{\mu }_{k}^{{\prime} }}({u}_{0})\ldots {\dot{\gamma }}^{{\mu }_{1}^{{\prime} }}({u}_{0}){{\rm{\nabla }}}_{{\mu }_{k}^{{\prime} }}\ldots {{\rm{\nabla }}}_{{\mu }_{2}^{{\prime} }}{B}_{{\mu }_{1}^{{\prime} }}^{d}(y),\end{eqnarray}$
which can be proved by mathematical induction with the usage of (8). Thereby, we obtain statement (16) after applying the second relation from (17).
Further, after performing the differentiation of the right hand side of (13), using the last equality and one additional property of Synge's world function
$\begin{eqnarray}{\sigma }^{\rho }(x,y){\partial }_{{y}^{\rho }}{\sigma }^{{\mu }_{i}^{{\prime} }}(x,y)={\sigma }^{{\mu }_{i}^{{\prime} }}(x,y),\end{eqnarray}$
we get a number of relations for each degree of ${\sigma }^{\mu ^{\prime} }(x,y)$
$\begin{eqnarray}\begin{array}{rcl} & & \displaystyle \sum _{k=1}^{+\infty }\displaystyle \frac{{\left(-1\right)}^{k}{\sigma }^{{\mu }_{1}^{{\prime} }}\left(x,y)\ldots {\sigma }^{{\mu }_{k}^{{\prime} }}(x,y\right)}{\left(k-1)!\right.}\,\left(\left.1{\overleftarrow{D}}_{{\mu }_{1}^{{\prime} }}^{a{c}_{1}}{\overleftarrow{D}}_{{\mu }_{2}^{{\prime} }}^{{c}_{1}{c}_{2}}\ldots {\overleftarrow{D}}_{{\mu }_{k}^{{\prime} }}^{{c}_{k-1}c}+{{\rm{\nabla }}}_{{\mu }_{k}^{{\prime} }}\ldots {{\rm{\nabla }}}_{{\mu }_{2}^{{\prime} }}{f}^{{adc}}{B}_{{\mu }_{1}^{{\prime} }}^{d}(y\right)\right.\\ & & +\left.\displaystyle \sum _{n=1}^{k-1}{C}_{n}^{k-1}\left({{\rm{\nabla }}}_{{\mu }_{\mathrm{k}}^{{\prime} }}\ldots {{\rm{\nabla }}}_{{\mu }_{n+2}^{{\prime} }}{f}^{{ade}}{B}_{{\mu }_{n+1}^{{\prime} }}^{d}\left(y\right)\right)\left(1{\mathop{D}\limits^{\unicode{x027F5}}}_{{\mu }_{1}^{{\prime} }}^{e{c}_{1}}{\mathop{D}\limits^{\unicode{x027F5}}}_{{\mu }_{2}^{{\prime} }}^{{c}_{1}{c}_{2}}\ldots {\mathop{D}\limits^{\unicode{x027F5}}}_{{\mu }_{n}^{{\prime} }}^{{c}_{k-1}c}\right)\right).\end{array}\end{eqnarray}$
The last sum is equal to zero. Indeed, the factor ${\sigma }^{{\mu }_{1}^{{\prime} }}(x,y)\ldots {\sigma }^{{\mu }_{k}^{{\prime} }}(x,y)$ leads to the symmetrization of the tensor in large parentheses. This means, in particular, that we can change the order of derivatives. Hence, the first term cancels the others after factorizing the first degree of the Yang–Mills connection.
Therefore, we have obtained the first statement of the lemma. The second equality follows from the first one, the second property from formula (9), the following permutations $a,x\leftrightarrow b,y,$ and the relation ${f}^{abc}=-{f}^{cba}.$
Notice one more representation for the path-ordered exponential.

Let $x,y\in U$, and $1$ denote the unit function. Also, ${N}_{r}(x,y)$ and ${N}_{l}(x,y)$ denote matrix-valued operators ${\sigma }^{\rho }(x,y){\mathop{D}\limits^{\longrightarrow}}_{\rho }^{ab}$ and ${\mathop{D}\limits^{\longrightarrow}}_{\rho ^{\prime} }^{ab}{\sigma }^{\rho ^{\prime} }\left(x,y\right),$ respectively. Then, under the conditions described above, we have

$\begin{eqnarray}{\rm{\Phi }}(x,y)=\mathop{\mathrm{lim}}\limits_{t\to 1-0}\left({{\rm{e}}}^{{N}_{r}(x,y)\mathrm{ln}(1-t)}1\right)=\mathop{\mathrm{lim}}\limits_{t\to 1-0}\left(1{{\rm{e}}}^{{N}_{l}(x,y)\mathrm{ln}(1-t)}\right).\end{eqnarray}$

For simplicity, we work with matrix-valued operators. Then, using property (15), we can write the following relation

$\begin{eqnarray}{\sigma }^{{\mu }_{k}}(x,y)...{\sigma }^{{\mu }_{1}}(x,y)({\mathop{D}\limits^{\longrightarrow }}_{{x}^{{\mu }_{1}}}...{\mathop{D}\limits^{\longrightarrow }}_{{x}^{{\mu }_{k}}})=\displaystyle \prod _{i=0}^{k-1}({N}_{r}(x,y)-i).\end{eqnarray}$
It is obtained with the use of mathematical induction. Applying the change $x\leftrightarrow y$ and transposition of the matrices, we can get the relation for the left operators. Let us note, that the right hand side of (24) contains the product of commutative (with each other) operators.
Hence, we can rewrite the following chain of equalities for representation (14)
$\begin{eqnarray}\begin{array}{l}{\rm{\Phi }}(x,y)=\displaystyle \sum _{k=0}^{+\infty }\displaystyle \frac{1}{k!}\left(\displaystyle \prod _{i=0}^{k-1}\left(i-{N}_{r}\left(x,y\right)\right)\right)1\\ \,=\,\displaystyle \sum _{k=0}^{+\infty }\displaystyle \frac{{\rm{\Gamma }}\left(k-{N}_{r}\left(x,y\right)\right)}{{\rm{\Gamma }}\left(k+1\right){\rm{\Gamma }}\left(-{N}_{r}\left(x,y\right)\right)}\,1\end{array}\end{eqnarray}$
$\begin{eqnarray}\begin{array}{l}=\mathop{\mathrm{lim}}\limits_{t\to 1-0}\left(\displaystyle \sum _{k=0}^{+\infty }\displaystyle \frac{{t}^{k}\,{\rm{\Gamma }}(k-{N}_{r}(x,y))}{{\rm{\Gamma }}(k+1){\rm{\Gamma }}(-{N}_{r}(x,y))}\,1\right)\\ \,=\,\mathop{\mathrm{lim}}\limits_{t\to 1-0}({(1-t)}^{{N}_{r}(x,y)}\,1),\end{array}\end{eqnarray}$
from which we obtain the first relation of the lemma. The second one follows from the change $x,r\leftrightarrow y,l$ and the matrix transposition.
From statements (13) and (14) it follows that for smooth connection components we have a covariant Taylor expansion for the path-ordered exponential (7). Moreover, it allows us to write out the answer for the ordered exponential in terms of the field ${B}_{\mu }^{d}$ and its covariant derivatives. At the same time, we need to emphasize that the first ordinary derivatives (left and right) of Synge's world function are uniquely related to each other with the usage of covariant Taylor series, see formula (16) for ${B}_{\rho }^{d}(x)={\delta }_{\rho i}$ with fixed $i\in \{1,\ldots ,d\}.$ Hence, the degrees of ${\sigma }^{\rho }(x,y)$ and degrees of ${\sigma }^{\rho ^{\prime} }(x,y)$ lead to equivalent expansions for two-point functions, like the degrees of ${\left(x-y\right)}^{\rho }$ in ${{\rm{R}}}^{d}.$
Now we are ready to formulate the next relation.

Let $x,y\in U.$ Then, under the conditions described above, we have

$\begin{eqnarray}{f}^{abc}{{\rm{\Phi }}}^{ae}(x,y){{\rm{\Phi }}}^{bd}(x,y){{\rm{\Phi }}}^{cg}(x,y)={f}^{edg},\end{eqnarray}$
$\begin{eqnarray}{{\rm{\Phi }}}^{{a}_{1}d}(x,y){{\rm{\Phi }}}^{{a}_{2}c}(x,y){f}^{{a}_{1}b{a}_{2}}{f}^{{a}_{3}b{a}_{4}}{{\rm{\Phi }}}^{{a}_{3}e}(x,y){{\rm{\Phi }}}^{{a}_{4}g}(x,y)={f}^{dbc}{f}^{ebg}.\end{eqnarray}$

Let us apply the operator ${\sigma }^{\rho }(x,y){\partial }_{\rho }$ to the left hand side of formula (27). In this case, the derivative acts on every ${\rm{\Phi }}$-factor. So we get three terms.

Then we need to combine formulae (5) and (12) and substitute $-{\sigma }^{\rho }(x,y){f}^{adb}{B}_{\rho }^{d}(x){{\rm{\Phi }}}^{bc}(x,y)$ instead of ${\sigma }^{\rho }(x,y){\partial }_{{x}^{\rho }}{{\rm{\Phi }}}^{ac}(x,y).$ After that we obtain three new terms
$\begin{eqnarray*}-{f}^{abc}({\sigma }^{\rho }(x,y){f}^{aij}{B}_{\rho }^{i}(x){{\rm{\Phi }}}^{je}(x,y)){{\rm{\Phi }}}^{bd}(x,y){{\rm{\Phi }}}^{cg}(x,y)\end{eqnarray*}$
$\begin{eqnarray*}-{f}^{abc}{{\rm{\Phi }}}^{ae}(x,y)({\sigma }^{\rho }(x,y){f}^{bij}{B}_{\rho }^{i}(x){{\rm{\Phi }}}^{jd}(x,y)){{\rm{\Phi }}}^{cg}(x,y)\end{eqnarray*}$
$\begin{eqnarray*}-{f}^{abc}{{\rm{\Phi }}}^{ae}(x,y){{\rm{\Phi }}}^{bd}(x,y)({\sigma }^{\rho }(x,y){f}^{cij}{B}_{\rho }^{i}(x){{\rm{\Phi }}}^{jg}(x,y)).\end{eqnarray*}$
Therefore, if we apply property (2) to the first term, it can be converted into the second and the third terms, but with opposite signs. This means we have proven that
$\begin{eqnarray}{\sigma }^{\rho }(x,y){\partial }_{{x}^{\rho }}({f}^{abc}{{\rm{\Phi }}}^{ae}(x,y){{\rm{\Phi }}}^{bd}(x,y){{\rm{\Phi }}}^{cg}(x,y))=0.\end{eqnarray}$
Now we need to use the representation for the path-ordered exponential derived above, see (14). From this, it follows that we can expand the ordered exponential in the covariant Taylor series in powers of ${\sigma }^{\rho }(x,y),$ because the field ${B}_{\mu }(x)$ is smooth by definition.
Hence, formula (29) means that ${f}^{abc}{{\rm{\Phi }}}^{ae}(x,y){{\rm{\Phi }}}^{bd}(x,y){{\rm{\Phi }}}^{cg}(x,y)$ does not depend on the variable $x.$ So we can take any convenient value of $x\in U.$ If we choose $x=y$ and use the property from (9), then we get the first statement of the lemma. The second statement is the consequence of the first one.
Here we have presented the proof in terms of mathematical analysis. An alternative algebraic proof with an appropriate interpretation can be found in the conclusion.
The last lemma is devoted to a product of three path-ordered exponentials. For convenience, we need to define the field strength tensor, components of which are equal to
$\begin{eqnarray*}{F}_{\mu \nu }^{a}(x)={\partial }_{{x}^{\mu }}{B}_{\nu }^{a}(x)-{\partial }_{{x}^{\nu }}{B}_{\mu }^{a}(x)+{f}^{abc}{B}_{\mu }^{b}(x){B}_{\nu }^{c}(x),\end{eqnarray*}$
where $x\in U.$

Let $x,y,z\in U.$ Then, under the conditions described above, we have the following relation

$\begin{eqnarray}{\rm{\Phi }}(z,x){\rm{\Phi }}(x,y){\rm{\Phi }}(y,z)={\rm{\Phi }}(x,y){| }_{{B}_{\mu }(\cdot )\to {f}_{\mu }(\cdot -z,z)},\end{eqnarray}$
where
$\begin{eqnarray}{f}_{\sigma }^{ab}(x-z,z)={{\rm{\Phi }}}^{ac}(z,x)({\delta }^{cd}{\partial }_{{x}^{\sigma }}+{f}^{ced}{B}_{\sigma }^{e}(x)){{\rm{\Phi }}}^{db}(x,z).\end{eqnarray}$
Moreover, let ${g}^{\mu \nu }(\cdot )={\delta }^{\mu \nu }$ for all points from $U.$ Also, ${\bf{1}},$ ${B}_{\mu },$ and ${F}_{\mu \nu }$ denote the matrix-valued operators with the elements ${\delta }^{ab},$ ${f}^{adb}{B}_{\mu }^{d},$ and ${f}^{adb}{F}_{\mu \nu }^{d},$ respectively. Also, we define an additional derivative of a special type by the formula ${{\mathfrak{D}}}_{{x}^{\mu }}\cdot ={\partial }_{{x}^{\mu }}\cdot +[{B}_{\mu }(x),\cdot \,].$ So, we obtain
$\begin{eqnarray}\begin{array}{l}{f}_{\sigma }^{ab}(x-z,z)=\displaystyle \sum _{k=1}^{+\infty }\displaystyle \frac{{(x-z)}^{{\mu }_{1}\ldots {\mu }_{k}}}{(k-1)!(k+1)}\\ \,\times \,{{\mathfrak{D}}}_{{z}^{{\mu }_{1}}}^{a{c}_{1}}\ldots {{\mathfrak{D}}}_{{z}^{{\mu }_{k-1}}}^{{c}_{k-2}{c}_{k-1}}{f}^{{c}_{k-1}{c}_{k}b}{F}_{{\mu }_{k}\sigma }^{{c}_{k}}(z),\end{array}\end{eqnarray}$
where ${(x-z)}^{{\mu }_{1}\ldots {\mu }_{k}}={(x-z)}^{{\mu }_{1}}\cdot \ldots \cdot {(x-z)}^{{\mu }_{k}}.$ The first terms in decomposition (30) have the form
$\begin{eqnarray*}\begin{array}{l}{\rm{\Phi }}(z,x){\rm{\Phi }}(x,y){\rm{\Phi }}(y,z)=1+\displaystyle \frac{1}{2}{(x-z)}^{\nu }{(y-z)}^{\mu }{F}_{\nu \mu }(z)\\ \,+\,\displaystyle \frac{1}{6}{(z-y)}^{\nu }{(z-x)}^{\sigma }{(y-x)}^{\mu }{D}_{{z}^{\left(\nu \right.}}{F}_{\left.\sigma \right)\mu }(z)\end{array}\end{eqnarray*}$
$\begin{eqnarray*}\begin{array}{l}+\,\displaystyle \frac{1}{8}{(x-z)}^{\nu \sigma }{(y-z)}^{\mu \rho }{F}_{\nu \mu }(z){F}_{\sigma \rho }(z)\\ -\,\displaystyle \frac{1}{24}{(y-x)}^{\mu \nu \sigma }{(x-z)}^{\rho }{{\mathfrak{D}}}_{{z}^{\mu }}{{\mathfrak{D}}}_{{z}^{\nu }}{F}_{\sigma \rho }(z)\end{array}\end{eqnarray*}$
$\begin{eqnarray}-\displaystyle \frac{1}{16}{(x-z)}^{\mu \rho }{(y-x)}^{\sigma }{(y-z)}^{\nu }{D}_{{z}^{\left(\mu \right.}}{D}_{{z}^{\left.\nu \right)}}{F}_{\sigma \rho }(z)+\ldots ,\end{eqnarray}$
where ‘three dots' denotes terms, which have the total degree of a monomial more than four. Parentheses denote index symmetrization without division by the corresponding factorial.

Let us introduce two gauge transformed derivatives according to the formulae

$\begin{eqnarray}{\mathop{D}\limits^{\longrightarrow}}_{{x}^{\sigma }}^{ab}(z)={{\rm{\Phi }}}^{ac}(z,x){\mathop{D}\limits^{\longrightarrow}}_{{x}^{\sigma }}^{cd}{{\rm{\Phi }}}^{db}(x,z)={\delta }^{ab}{\mathop{\partial }\limits^{\longrightarrow}}_{{x}^{\sigma }}+{f}_{\sigma }^{ab}(x-z,z),\end{eqnarray}$
$\begin{eqnarray}{\mathop{D}\limits^{\longrightarrow}}_{{x}^{\sigma }}^{ab}(z)={{\rm{\Phi }}}^{ac}(z,x){\mathop{D}\limits^{\longrightarrow}}_{{x}^{\sigma }}^{cd}{{\rm{\Phi }}}^{db}(x,z)={\delta }^{ab}{\mathop{\partial }\limits^{\longrightarrow}}_{{x}^{\sigma }}-{f}_{\sigma }^{ab}(x-z,z),\end{eqnarray}$
and rewrite the equations from (12) in the following form
$\begin{eqnarray*}\begin{array}{c}{\sigma }^{\mu }(x,y){\mathop{D}\limits^{\longrightarrow }}_{\mu }^{{ab}}(z)({{\rm{\Phi }}}^{{be}}(z,x){{\rm{\Phi }}}^{{ed}}(x,y){{\rm{\Phi }}}^{{dc}}(y,z))=0,\\ \,({{\rm{\Phi }}}^{{ab}}(z,x){{\rm{\Phi }}}^{{be}}(x,y){{\rm{\Phi }}}^{{ed}}(y,z)){\overleftarrow{D}}_{{\mu }^{{\rm{{\prime} }}}}^{{dc}}(z){\sigma }^{{\mu }^{{\rm{{\prime} }}}}(x,y)=0.\end{array}\end{eqnarray*}$
Moreover, the combination ${{\rm{\Psi }}}^{ac}(x,y;z)={{\rm{\Phi }}}^{ab}(z,x){{\rm{\Phi }}}^{bd}(x,y){{\rm{\Phi }}}^{dc}(y,z)$ satisfies the following initial condition ${{\rm{\Psi }}}^{ac}(y,y;z)={\delta }^{ac},$ which is the consequence of (9). This means that ${{\rm{\Psi }}}^{ac}(x,y;z)$ is the ordered exponential for the connection components of the form ${f}_{\sigma }^{ab}(\cdot -z,z).$ Hence, we have obtained statement (30). Formula (32) is the particular case and has been derived in the paper [8]. Then, expansion (33) follows from the results of (30) and the definition from (7) with the use of the explicit form for geodesic $\gamma (s)=y+s(x-y).$
Let us write a small summary for this section. We have obtained four results, and it would be useful to emphasize their positions in the general theory. Lemmas 1 and 2 provide new representation formulae for the ordered exponential, where the first one generalizes the very well-known result, mentioned above. Then, lemma 3 is known in the context of the algebraic approach, see Conclusion, but we have presented its new derivation in terms of mathematical analysis with the use of the differential equation. Further, lemma 4 gives a description of the transition to the Fock–Schwinger gauge condition. Actually, this is intuitively and partially known result [8], which we have formulated in such form for clarity and simplicity of various applications.

4. Application to the Yang–Mills theory

Let us introduce an effective action for the four-dimensional quantum Yang–Mills theory [12, 14]. For this we use the main conditions of section 2, but with some additional restrictions: $M={{\mathbb{R}}}^{4},$ ${g}^{\mu \nu }(\cdot )={\delta }^{\mu \nu }$ for all points from ${{\mathbb{R}}}^{4},$ and the field ${B}_{\mu }(x)$ satisfies the quantum equation of motion, which follows from the use of the path integral formulation [14, 21] and the background field method [2226]. This field is called the background field.
For further work we need to define some additional constructions, such as the classical action of the Yang–Mills theory
$\begin{eqnarray}{W}_{-1}[B]=\displaystyle \frac{1}{4}\displaystyle {\int }_{{{\mathbb{R}}}^{4}}{d}^{4}x\,{F}_{\mu \nu }^{a}(x){F}_{\mu \nu }^{a}(x),\end{eqnarray}$
the following two Laplace-type operators
$\begin{eqnarray}\begin{array}{c}{M}_{0}^{{ab}}(x)=-{\mathop{D}\limits^{\longrightarrow }}_{{x}_{\mu }}^{{ae}}{\mathop{D}\limits^{\longrightarrow }}_{{x}^{\mu }}^{{eb}},\,{M}_{1\mu \nu }^{\,{ab}}(x)\\ ={M}_{0}^{{ab}}(x){\delta }_{\mu \nu }-2{f}^{{acb}}{F}_{\mu \nu }^{c}(x),\end{array}\end{eqnarray}$
and the corresponding Green's functions ${G}_{0}$ and ${G}_{1},$ which follow from the equalities
$\begin{eqnarray}\begin{array}{l}{M}_{1\mu \nu }^{\,ab}(x){G}_{1\nu \rho }^{\,bc}(x,y)={\delta }^{ac}{\delta }_{\mu \rho }\delta (x-y),\,\\ \,{M}_{0}^{ab}(x){G}_{0}^{bc}(x,y)={\delta }^{ac}\delta (x-y),\end{array}\end{eqnarray}$
with appropriate boundary conditions. These conditions have a physical nature and, actually, are not studied well enough. They should be such that the problem for the quantum equation of motion would be well posed.
Now we are ready to introduce the effective action of the Yang–Mills theory for small values of a coupling constant $g.$ This is a function of the background field ${B}_{\mu },$ which has the following asymptotic expansion for $g\to +0$
$\begin{eqnarray}\begin{array}{l}W\left[B,{\rm{\Lambda }}\right]=\displaystyle \frac{1}{{g}^{2}}{W}_{-1}[B]+\left\{\displaystyle \frac{1}{2}\,\mathrm{ln}\,\det \left({M}_{1}^{{\rm{\Lambda }}}/{M}_{1}^{{\rm{\Lambda }}}{| }_{B=0}\right)\right.\\ -\,\left.\mathrm{ln}\,\det \left({M}_{0}^{{\rm{\Lambda }}}/{M}_{0}^{{\rm{\Lambda }}}{| }_{B=0}\right)\Space{0ex}{2.75ex}{0ex}\right\}+\displaystyle \sum _{n=1}^{+\infty }{g}^{2n}{W}_{n}[B,{\rm{\Lambda }}],\end{array}\end{eqnarray}$
where the symbol ${\rm{\Lambda }}$ denotes some type of regularization, such that the Green's functions are deformed near the diagonal ($x\sim y$). Removing the regularization corresponds to the limit ${\rm{\Lambda }}\to +\infty .$ Using the language of quantum field theory, we can declare that ${W}_{n}[B,{\rm{\Lambda }}]$ is the $(n+1)$th quantum correction, corresponding to the $(n+1)$-loop contribution, see [15, 16]. Precise formulae for these corrections can be found in papers [21, 27]. Fortunately, explicit expressions do not matter in our calculations, that is why we have introduced the effective action in such a general form.
Also, we need to note that the parameter ${\rm{\Lambda }}$ symbolizes some type of regularization, not necessarily a cutoff one, despite the fact that it looks like the standard momentum of cut in the momentum representation. The assumption that the regularization is removed at the limit transition ${\rm{\Lambda }}\to +\infty $ actually does not matter, because this condition always can be chosen after an appropriate linear fractional transformation.
Let us draw attention that the regularization is necessary because all corrections to the classical action in (39) contain divergent integrals. Some first terms have been studied earlier. For example, the divergent part of ‘$\mathrm{ln}\,\det $' has been calculated explicitly in [2830]. The two-loop contribution also has been computed with the use of different regularizations, see [27, 3032].
Then, we should note that the corrections from the last sum can be constructed with the use of the integration operator over ${{\mathbb{R}}}^{4}$ and the following elementary blocks (with the regularization applied)
$\begin{eqnarray}\begin{array}{l}{G}_{0}^{ab}({x}_{i},{x}_{j}),{\mathop{D}\limits^{\longrightarrow}}_{{x}_{i}^{\mu }}^{ac}{G}_{0}^{cb}({x}_{i},{x}_{j}),{G}_{0}^{ac}({x}_{i},{x}_{j}){\mathop{D}\limits^{\longrightarrow}}_{{x}_{j}^{\sigma }}^{cb},\\ {\mathop{D}\limits^{\longrightarrow}}_{{x}_{i}^{\mu }}^{ae}{G}_{0}^{ec}({x}_{i},{x}_{j}){\mathop{D}\limits^{\longrightarrow}}_{{x}_{j}^{\sigma }}^{cb},{G}_{1\nu \rho }^{\,ab}({x}_{i},{x}_{j}),{\mathop{D}\limits^{\longrightarrow}}_{{x}_{i}^{\mu }}^{ac}{G}_{1\nu \rho }^{\,cb}({x}_{i},{x}_{j}),\\ {G}_{1\nu \rho }^{\,ac}({x}_{i},{x}_{j}){\mathop{D}\limits^{\longrightarrow}}_{{x}_{j}^{\sigma }}^{cb},{\mathop{D}\limits^{\longrightarrow}}_{{x}_{i}^{\mu }}^{ae}{G}_{1\nu \rho }^{\,ec}({x}_{i},{x}_{j}){\mathop{D}\limits^{\longrightarrow}}_{{x}_{j}^{\sigma }}^{cb},\end{array}\end{eqnarray}$
which are connected to each other only by the following combinations of the structure constants:
$\begin{eqnarray}{f}^{abc}\,\mathrm{and}\,{f}^{abe}{f}^{ecd}.\end{eqnarray}$
It is quite important to note two conditions. Firstly, all the variables should be under the integration. Secondly, if three (four) basic blocks are connected to each other with the usage of ${f}^{abc}$ (${f}^{abe}{f}^{ecd}$), then the corresponding ends of the blocks have the same variables. Other combinations do not exist. Also, we have noted above that the blocks from (40) should contain regularized Green's functions instead of ${G}_{0}$ and ${G}_{1}.$ Now we want to introduce one restriction on a type of regularization. We assume that regularized Green's functions respect covariance. This means that after applying the gauge transformation, we obtain
$\begin{eqnarray}{B}_{\mu }(x)\to {B}_{\mu }^{\omega }(x)={\omega }^{-1}(x){B}_{\mu }(x)\omega (x)+{\omega }^{-1}(x){\partial }_{{x}^{\mu }}\omega (x)\end{eqnarray}$
and the regularized Green's functions get the following changes
$\begin{eqnarray}\begin{array}{l}{G}_{0}^{ab}({x}_{i},{x}_{j}){| }^{{\rm{\Lambda }}\,-reg.\,}\to \left({G}_{0}^{ab}({x}_{i},{x}_{j}){| }_{{B}_{\mu }\to {B}_{\mu }^{\omega }}^{{\rm{\Lambda }}\,-reg.\,}\right)\\ =\,{\omega }^{-1}({x}_{i})({G}_{0}^{ab}({x}_{i},{x}_{j}){| }^{{\rm{\Lambda }}\,-reg.\,})\omega ({x}_{j}),\end{array}\end{eqnarray}$
$\begin{eqnarray}\begin{array}{l}{G}_{1\nu \rho }^{\,ab}({x}_{i},{x}_{j}){| }^{{\rm{\Lambda }}\,-reg.\,}\to ({G}_{1\nu \rho }^{\,ab}({x}_{i},{x}_{j}){| }_{{B}_{\mu }\to {B}_{\mu }^{\omega }}^{{\rm{\Lambda }}\,-reg.\,})\\ =\,{\omega }^{-1}({x}_{i})({G}_{1\nu \rho }^{\,ab}({x}_{i},{x}_{j}){| }^{{\rm{\Lambda }}\,-reg.\,})\omega ({x}_{j}),\end{array}\end{eqnarray}$
where $\omega (\cdot )\in {C}^{\infty }({{\mathbb{R}}}^{4},G).$ Of course, non-regularized functions have this property by construction. Using the fact that in further calculations we want to make a special type of gauge transformation, we conclude that for simplicity we can produce all manipulations with (40) instead of the regularized ones because the deformation does not break the covariance. Also, we need to draw attention that such regularizations exist. As an example, we can suggest the explicit cutoff regularization in coordinate representation, recently studied in [33].
For further study, we need to give some basic concepts and results on the heat kernel expansion, [20, 3436]. Let us introduce a Laplace-type operator $A,$ which has a more general view than one from (37). Locally, it has the following form
$\begin{eqnarray}{A}^{ab}(x)=I{M}_{0}^{ab}(x)-{v}^{ab}(x),\end{eqnarray}$
where $I$ is an $n\times n$ unit matrix with $n\in {\mathbb{N}},$ and ${v}^{ab}(x)$ is an $n\times n$ matrix-valued smooth potential, such that the operator $A$ is symmetric. If we take $n=4,$ ${(I)}_{\mu \nu }={\delta }_{\mu \nu },$ and ${({v}^{ab})}_{\mu \nu }(x)=2{f}^{acb}{F}_{\mu \nu }^{c}(x),$ then we obtain the operator ${M}_{1\mu \nu }^{ab}(x).$ Also, for convenience, we will not write the matrix $I$ in the rest of the text, because this does not make any confusion.
Then we move on to a definition of the local heat kernel. As we know, the standard heat kernel is the solution of the problem
$\begin{eqnarray}\begin{array}{c}({\delta }^{{ac}}{{\rm{\partial }}}_{\tau }+{A}^{{ac}}(x)){K}^{{cb}}(x,y;\tau )=0,\\ \,{K}^{{ab}}(x,y;0)={\delta }^{{ab}}\delta (x-y),\end{array}\end{eqnarray}$
paired with boundary conditions for (45). At the same time, the local heat kernel [37] is defined as a solution of (46), which for small values of the proper time $\tau \to +0$ has the following asymptotic expansion
$\begin{eqnarray}{K}^{ab}(x,y;\tau )={(4\pi \tau )}^{-2}{e}^{-| x-y{| }^{2}/4\tau }\displaystyle \sum _{k=0}^{+\infty }{\tau }^{k}{{\mathfrak{a}}}_{k}^{ab}(x,y).\end{eqnarray}$
We note that an asymptotic expansion for the ordinary heat kernel of $A$ can be presented as the sum of (47) and some additional corrections. The coefficient ${{\mathfrak{a}}}^{ab}(x,y)$ of expansion (47) is the Seeley–DeWitt coefficient (also, it is named after Hadamard, MinakshiSundaram [38], and Gilkey [35]), see [39, 40], and can be calculated recurrently [7, 8, 35], because they satisfy the following system of equations
$\begin{eqnarray}\begin{array}{c}{{\mathfrak{a}}}_{0}^{{ab}}(x,y)={{\rm{\Phi }}}^{{ab}}(x,y),\,(k+{\left(x-y\right)}^{\sigma }{\mathop{D}\limits^{\longrightarrow }}_{{x}^{\sigma }}^{{ac}}){{\mathfrak{a}}}_{k}^{{cb}}(x,y)\\ =\,-{A}^{{ac}}(x){{\mathfrak{a}}}_{k-1}^{{cb}}(x,y),\,k\geqslant 1.\end{array}\end{eqnarray}$
Then, we are going to make a substitution. Let $z\in {{\mathbb{R}}}^{4}$ be an auxiliary point, that will play a role of an additional parameter. Then, we define
$\begin{eqnarray}{{\mathfrak{a}}}_{k}^{ab}(x,y;z)={{\rm{\Phi }}}^{ac}(z,x){{\mathfrak{a}}}_{k}^{cd}(x,y){{\rm{\Phi }}}^{db}(y,z).\end{eqnarray}$
Further, we substitute formula (49) into relation (48) and introduce gauge transformed operators, according to the papers [8, 9], in the form
$\begin{eqnarray}\begin{array}{c}{M}_{0}^{{ab}}(x-z;z)={{\rm{\Phi }}}^{{ac}}(z,x){M}_{0}^{{cd}}(x){{\rm{\Phi }}}^{{db}}(x,z)\\ \,=\,-{\mathop{D}\limits^{\longrightarrow }}_{{x}_{\sigma }}^{{ac}}(z){\mathop{D}\limits^{\longrightarrow }}_{{x}^{\sigma }}^{{cb}}(z),\end{array}\end{eqnarray}$
$\begin{eqnarray}{A}_{z}^{ab}(x)={{\rm{\Phi }}}^{ac}(z,x){A}^{cd}(x){{\rm{\Phi }}}^{db}(x,z)\end{eqnarray}$
$\begin{eqnarray*}\begin{array}{l}={\delta }_{\mu \nu }{M}_{0}^{ab}(x-z;z)-\displaystyle \sum _{k=0}^{+\infty }\displaystyle \frac{{(x-z)}^{{\mu }_{1}\ldots {\mu }_{k}}}{k!}{{\mathfrak{D}}}_{{z}^{{\mu }_{1}}}^{a{c}_{1}}\ldots \\ \times \,{{\mathfrak{D}}}_{{z}^{{\mu }_{k}}}^{{c}_{k-1}{c}_{k}}{v}^{{c}_{k}b}(z),\end{array}\end{eqnarray*}$
where we have used the Fock–Schwinger connection components (32) and formula (37). Hence, the system of the equations from (48) can be rewritten in the following form
$\begin{eqnarray}{{\mathfrak{a}}}_{0}^{ab}(x,y;z)={{\rm{\Phi }}}^{ac}(z,x){{\rm{\Phi }}}^{cd}(x,y){{\rm{\Phi }}}^{db}(y,z),\end{eqnarray}$
$\begin{eqnarray*}\begin{array}{c}(k+{\left(x-y\right)}^{\sigma }{\mathop{D}\limits^{\longrightarrow }}_{{x}^{\sigma }}^{{ac}}(z)){{\mathfrak{a}}}_{k}^{{cb}}(x,y;z)\\ =\,-{A}_{z}^{{ac}}(x){{\mathfrak{a}}}_{k-1}^{{cb}}(x,y;z),\,k\geqslant 1.\end{array}\end{eqnarray*}$
Now we are ready to note a piece of very useful information. It follows from formulae (32), (34), (35), (50), and (51), that if the potential $v$ is a function only of the field strength and its ${\mathfrak{D}}$-derivatives, then the decomposition of the Seeley–DeWitt coefficients ${{\mathfrak{a}}}_{k}^{cb}(x,y;z)$ can be constructed by using only the field stress tensor and its ${\mathfrak{D}}$-derivatives. As we can see, the operators ${M}_{0}^{ab}$ and ${M}_{1\mu \nu }^{\,ab}$ for the Yang–Mills theory are suitable for this statement. Additionally, we emphasize that formulae (48)–(52) are valid for all dimensions.
Further, using the formulae introduced above, we can write out the following asymptotic behavior near the diagonal for the Green's function in the four-dimensional space [39, 41]
$\begin{eqnarray*}\begin{array}{l}{({A}_{z}^{-1})}^{ab}(x,y)={R}_{0}(x-y){{\mathfrak{a}}}_{0}^{ab}(x,y;z)\\ +\,{R}_{1}(x-y){{\mathfrak{a}}}_{1}^{ab}(x,y;z)\end{array}\end{eqnarray*}$
$\begin{eqnarray}+{R}_{2}(x-y){{\mathfrak{a}}}_{2}^{ab}(x,y;z)+{{\mathscr{N}}}_{z}^{ab}(x,y)+{\mathscr{Z}}{ {\mathcal M} }_{z}^{ab}(x,y),\end{eqnarray}$
where
$\begin{eqnarray}\begin{array}{l}{R}_{0}(x)=\displaystyle \frac{1}{4{\pi }^{2}| x{| }^{2}},\,{R}_{1}(x)=-\displaystyle \frac{\mathrm{ln}(| x{| }^{2}{\mu }^{2})}{16{\pi }^{2}},\,\\ \,{R}_{2}(x)=\displaystyle \frac{| x{| }^{2}(\mathrm{ln}(| x{| }^{2}{\mu }^{2})-1)}{64{\pi }^{2}},\end{array}\end{eqnarray}$
${{\mathscr{N}}}_{z}^{ab}$ is a non-local part [41], depending on the boundary conditions of a spectral problem, and ${\mathscr{Z}}{ {\mathcal M} }_{z}^{ab}$ is a number of local zero modes to satisfy the problem. Let us separately note, it was shown in the paper [42], that an infrared part in the second loop does not depend on ${\mathscr{Z}}{ {\mathcal M} }_{z}^{ab}.$ Moreover, in the calculation process, we can choose ${\mathscr{Z}}{ {\mathcal M} }_{z}^{ab}$ in such a way [27], that the non-local part ${{\mathscr{N}}}_{z}^{ab}$ would have the following behavior near the diagonal $x\sim y$
$\begin{eqnarray}{{\mathscr{N}}}_{z}^{ab}(x,y)=-\displaystyle \frac{| x-y{| }^{2}}{{2}^{7}{\pi }^{2}}{{\mathfrak{a}}}_{2}^{ab}(y,y)(1+o(1)).\end{eqnarray}$
Now we move on to the property mentioned in the introduction. For this we should show the possibility of the following replacement ${f}^{adb}{B}_{\mu }^{d}(\cdot )\to {f}_{\mu }^{ab}(\,\cdot -z,z)$ in the effective action, where $z\in {{\mathbb{R}}}^{4}$ is an auxiliary point and ${f}_{\mu }^{ab}$ is from (32). The procedure can be carried out with the application of the gauge transformation using the ordered exponential ${{\rm{\Phi }}}^{ab}(\,\cdot \,,z).$ In the case we have formulae (34), (35), and
$\begin{eqnarray}{G}_{0}^{ab}({x}_{i},{x}_{j})={{\rm{\Phi }}}^{ac}({x}_{i},z){G}_{0}^{cd}({x}_{i},{x}_{j};z){{\rm{\Phi }}}^{db}(z,{x}_{j}),\end{eqnarray}$
$\begin{eqnarray}{G}_{1\nu \rho }^{\,ab}({x}_{i},{x}_{j})={{\rm{\Phi }}}^{ac}({x}_{i},z){G}_{1\nu \rho }^{\,cd}({x}_{i},{x}_{j};z){{\rm{\Phi }}}^{db}(z,{x}_{j}).\end{eqnarray}$
Using the last substitutions and formulae (38), (50), and (51), we can conclude, that the new covariant derivatives ${\mathop{D}\limits^{\longrightarrow }}_{{x}_{i}^{\mu }}^{{ab}}(z)$ and ${\overleftarrow{D}}_{{x}_{j}^{\mu }}^{{ab}}(z)$ and the Green's functions ${G}_{0}^{ab}({x}_{i},{x}_{j};z)$ and ${G}_{1\nu \rho }^{\,ab}({x}_{i},{x}_{j};z)$ are constructed only with the use of the following connection components ${f}_{\mu }^{ab}(\,\cdot -z,z),$ see (32), which are functions of the field stress tensor ${F}_{\mu \nu }$ and its ${\mathfrak{D}}$-derivatives.
At the same time after the gauge transformation, every building block from (40) is multiplied by the ordered exponential ${{\rm{\Phi }}}^{ea}({x}_{i},z)$ on the left hand side and by ${{\rm{\Phi }}}^{bg}(z,{x}_{j})$ on the right hand side. Further, let us remember, that the top indices of the blocks should be connected to each other by the ${f}^{abc}$ or ${f}^{abe}{f}^{ecd}.$ Hence, using the results of lemma 3, we can exclude all additional factors ${{\rm{\Phi }}}^{ea}({x}_{i},z)$ and ${{\rm{\Phi }}}^{bg}(z,{x}_{j})$ from the blocks. Therefore, we obtain the following equalities
$\begin{eqnarray}{W}_{n}[B,{\rm{\Lambda }}]={W}_{n}[f(\,\cdot -z,z),{\rm{\Lambda }}]\,\,{\rm{for}}\,{\rm{all}}\,\,n\geqslant 1,\end{eqnarray}$
where the functionals actually do not depend on $z.$ In other words, the procedure described above can be understood as a re-expansion of the blocks near some auxiliary point. It should be noted that in the calculation we have used only covariance of the regularized Green's functions (43) and (44). We did not assume any additional gauge conditions. The so-called Fock–Schwinger gauge (32) was generated inside each ${W}_{n}$ by substituting the unit $I=\omega {\omega }^{-1}$ between all adjacent blocks. Additionally, we want to emphasize that we have proven a more general result (58). It is correct for each Feynman graph separately.
In the same manner, we replace the operators in ‘$\mathrm{ln}\,\det $' by the gauge transformed ones. Then, in particular, using the invariance of the classical action with respect to the gauge transformations, we get
$\begin{eqnarray}W[B,{\rm{\Lambda }}]=W[f(\,\cdot -z,z),{\rm{\Lambda }}].\end{eqnarray}$
In the last part of the paper we want to formulate some actual open questions and make several useful remarks about divergencies searching. It is known that infrared divergences in the coordinate representation (or ultraviolet in the momentum one) follow from the appearance of non-integrable densities in (39), because the behavior (53) of the Green's function near the diagonal ($x\sim y$) includes such functions as (54).
At the same time, decomposition (53) is applicable only in the vicinity of the diagonal, so we need to split the Green's function into two parts, in some neighborhood of the $x=y$ and out of it. Formally, we can write this as
$\begin{eqnarray}({A}^{-1})(x,y)=\mathop{\underbrace{({A}^{-1})(x,y){\chi }_{({\mu }_{1}| x-y| \lt 1)}}}\limits_{\, {``} \mathrm{in} {"} \,\mathrm{part}\,}+\mathop{\underbrace{({A}^{-1})(x,y){\chi }_{({\mu }_{1}| x-y| \geqslant 1)}}}\limits_{\, {``} \mathrm{out} {"} \,\mathrm{part}\,},\end{eqnarray}$
where ${\mu }_{1}\gt 0$ is an auxiliary fixed finite number of momentum dimension, and ${\chi }_{(\,\mathrm{inequality})}$ is the Heaviside step function of a domain from ${{\mathbb{R}}}^{4}.$ While for the first part, we can write the standard decomposition (53) near the diagonal, the second part, actually, is unknown to us, because it and its properties depend on the boundary conditions and the form of the background field. Anyway, we assume that the ‘out' part has a good enough behavior at infinity, so the effective action does not include other types of divergences. Otherwise, we need to introduce one more regularization and study these non-infrared singularities in the coordinate representation.
Now we are ready to formulate some questions.

Does the effective action contain the infrared singularity in coordinate representation, depending on the ‘out' part? For the positive answer, it is sufficient to provide an explicit expression for some ${W}_{n}(B,{\rm{\Lambda }}).$ We draw attention that the two-loop contribution ${W}_{1}(B,{\rm{\Lambda }})$ includes only singularities, related to the ‘in' part because in this case the quantum correction is split into a sum of ‘in' and ‘out' parts, without mixing. Hence, without any fundamental justifications and assumptions, we are interested in the three-loop calculation. Moreover, it would be meaningful to describe the structure of such contributions for a cutoff regularization.

Does the effective action contain the infrared singularity in coordinate representation, depending on the non-local part ${{\mathscr{N}}}^{ab}$? This is one more non-obvious issue. The fact is that in the two-loop contribution, a special combination of non-local terms occurs, but it is converted to the local part with the use of the Seeley–DeWitt coefficients. The appearance or absence of the non-local part in the multi-loop terms is quite an interesting challenge because it leads to the possibility of using simplifications in the calculations.

The last two questions are related to the form of the coefficients, their locality and dependence on the boundary conditions. However, we have a problem, corresponding to a structure of singularities itself. Indeed, in the quantum corrections after the regularization introduced, we have some dimensional parameters: ${\rm{\Lambda }}$ is the parameter of regularization, $\mu $ is the parameter from (54), controlling the presence of local zero modes in the Green's functions, and ${\mu }_{1}$ is the auxiliary parameter from (60), which actually does not lead to any new dependence. Hence, the answer contains some dimensionless combinations of them. The question is the following.

Let $i,j\in {\mathbb{N}}\cup \{0\}$ and $n\gt 0.$ What combinations of ${({\rm{\Lambda }}/m)}^{i}{\mathrm{ln}}^{j}({\rm{\Lambda }}/m)$ does the correction ${W}_{n}(B,{\rm{\Lambda }})$ contain, where $m\in \{\mu ,{\mu }_{1}\}$ is a dimension parameter? For example, it was shown [27, 32] that the two-loop correction includes only $\mathrm{ln}({\rm{\Lambda }}/\mu ).$ And it is expected that for other orders we have $i=0$ and $j\in \{1,\ldots ,n\}.$

If the last hypothesis is correct, then for the local singular contribution, depending on the ‘in' part, we have only one type of the coefficients with non-zero trace ${f}^{aec}{F}_{\sigma \rho }^{e}{f}^{cdb}{F}_{\mu \nu }^{d}.$

5. Conclusion

In this paper, we have studied some non-trivial properties of the path-ordered exponentials. Additionally, we have shown that such equalities give the ability to make some useful conclusions about the structure of the four-dimensional Yang–Mills effective action. We believe that the result can be useful both in the general theory of renormalization of the Yang–Mills theory and in the investigation of some other models with non-Abelian structures. For example, in the sigma-models.
Furthermore, we would like to emphasize the relevance of our study and comment on some ways of further development. It should be noted that lemmas in section 3 are new and they provide the possibility to analyze multi-loop calculations in gauge theories, including Yang–Mills theory on Riemannian manifolds. The results of section 4 are devoted to one example of applying the obtained formulae for the ordered exponential to the effective action. Shortly, it can be written using two relations (58) and (59). From a physical point of view, our result guarantees covariance of the singular parts and, in particular, shows the absence of the term ${B}_{\mu }{B}_{\mu },$ the so-called mass term, after the renormalization.
Also, we need to repeat one more time a very important observation. The covariance under the gauge transformations of the Feynman diagram blocks actually does not exclude non-logarithmic divergencies. To formulate the proposition in a stronger form, we need some additional investigations which are not from the scope of this work. They are formulated in the previous section as open questions. Also, we emphasize that the standard gauge invariance of the Yang–Mills action (with the ghost fields) is not connected with the transition to the special gauge, generated in the effective action by hand - they are different. Indeed, the first one is the symmetry of the action, which can be violated by the regularization, while the second one follows from the covariance of the deformed Green's functions with respect to the background field (not the Yang–Mills field).
In addition, we need to comment on another way to prove lemma 3. Let us introduce three auxiliary arbitrary elements ${h}_{i}={h}_{i}^{a}{t}^{a},$ where $i=1,2,3,$ of the Lie algebra ${\mathfrak{g}}.$ Then, an element $g$ of the Lie group $G$ acts on this algebra according to the adjoint representation as $A{d}_{g}({h}_{i})=g{h}_{i}{g}^{-1}.$ Let ${g}^{ab}$ denote the corresponding matrix elements, such that ${(A{d}_{g}({h}_{i}))}^{a}{t}^{a}={g}^{ab}{h}_{i}^{b}{t}^{a}.$ After that we can write the following chain of equalities
$\begin{eqnarray*}\begin{array}{l}{f}^{abc}({g}^{ai}{h}_{1}^{i})({g}^{bj}{h}_{2}^{j})({g}^{ck}{h}_{3}^{k})\\ =\,-\displaystyle \frac{1}{2}{\rm{t}}{\rm{r}}([A{d}_{g}({h}_{1}),A{d}_{g}({h}_{2})]A{d}_{g}({h}_{3}))\end{array}\end{eqnarray*}$
$\begin{eqnarray*}=-\displaystyle \frac{1}{2}{\rm{t}}{\rm{r}}(A{d}_{g}([{h}_{1},{h}_{2}])A{d}_{g}({h}_{3}))\end{eqnarray*}$
$\begin{eqnarray*}=-\displaystyle \frac{1}{2}{\rm{t}}{\rm{r}}([{h}_{1},{h}_{2}]{h}_{3})={f}^{abc}{h}_{1}^{a}{h}_{2}^{b}{h}_{3}^{c},\end{eqnarray*}$
from which we obtain an alternative proof. Despite the fact that this derivation looks shorter, we note that it includes the usage of some additional properties, presented in (9). Indeed, we need to make sure that there is such an element $g$ in $G$ that the relation ${g}^{ab}={{\rm{\Phi }}}^{ab}(x,y)$ holds. So, we believe that the proof of lemma 3, written above, is more elegant and instructive in the context of covariant Taylor expansions.
Let us give some comments about the Yang–Mills effective action. In the paper, we have used the definition without using the path ordered formulation, because the last construction contains a lot of mathematical questions not related to the main topic. Due to this, we did not face the discussion of the measurement properties and their existence. Also, we study the effective action itself, without any counterterms, because we are interested in the properties of ‘pure' effective action before applying the renormalization procedure. Of course, the main proposition $B\to f(\,\cdot -z,z)$ can be expanded on the standard counterterms, because, according to the general theory, they also have the covariance. Unfortunately, some regularizations break the gauge invariance, and this leads to the necessity of restoring this symmetry and, as a consequence, adding some extra terms. We cannot guarantee that these terms (additional building blocks of the diagram technique) have covariance because they depend on a concrete view of the deformation. It would be useful to conduct an additional investigation on the issue.

Both authors are supported by the Ministry of Science and Higher Education of the Russian Federation, agreement 075-15-2022-289. Also, A V Ivanov is supported in parts by the Foundation for the Advancement of Theoretical Physics and Mathematics ‘BASIS', grant ‘Young Russian Mathematics'.

1
Lichnerowicz A 1955 Théorie Globale des Connexions et des Groupes d'holonomie Roma Edizioni Cremonese 1 297

2
Polyakov A 1987 Gauge Fields And Strings London, UK Harwood Academic Publishers 1 312

3
Nakahara M 2003 Geometry, Topology and Physics 2nd Edn Boca Raton, FL CRC Press 1 573

4
Shore G M 1981 Symmetry restoration and the background field method in gauge theories Ann. Phys. 137 262 305

DOI

5
Slavík A 2007 Product Integration, Its History and Applications Prague Matfyzpress 1 155

6
Connes A Marcolli M 2007 Noncommutative Geometry, Quantum Fields and Motives vol 55 Providence, RI American Mathematical Society, Colloquium Publications 1 808

7
Ivanov A V 2019 Diagram technique for the heat kernel of the covariant Laplace operator TMF 198 113 132

Ivanov A V 2019 Diagram technique for the heat kernel of the covariant Laplace operator Theoret. Math. Phys. 198 100 117

DOI

8
Ivanov A V Kharuk N V 2020 Heat kernel: proper-time method, Fock–Schwinger gauge, path integral, and Wilson line TMF 205 242 261

Ivanov A V Kharuk N V 2020 Heat kernel: proper-time method, Fock–Schwinger gauge, path integral, and Wilson line Theor. Math. Phys. 205 1456 1472

DOI

9
Ivanov A V Kharuk N V 2019 Non-recursive formula for trace of heat kernel 2019 Days on Diffraction (DD) 74 77

DOI

10
Bogoliubov N N Shirkov D V 1959 Introduction to the Theory of Quantized Fields New York New York City Wiley 1 720

11
Hall B 2003 Lie Groups, Lie Algebras, and Representations: An Elementary Introduction Berlin Springer Graduate Texts in Mathematics 1 354

12
Yang C N Mills R 1954 Conservation of isotopic spin and isotopic gauge invariance Phys. Rev. 96 191 195

DOI

13
Faddeev L D Popov V 1967 Feynman diagrams for Yang–Mills field Phys. Lett. B 25 29 30

DOI

14
Faddeev L D Slavnov A A 1991 Gauge fields: an introduction to quantum theory Front. Phys. 83 Reading, MA Addison-Wesley 1 236

15
Faddeev L D 2016 Scenario for the renormalization in the 4D Yang–Mills theory Int. J. Mod. Phys. A 31 1630001

DOI

16
Derkachev S E Ivanov A V Faddeev L D 2017 Renormalization scenario for the quantum Yang–Mills theory in four-dimensional space–time TMF 192 227 234

Derkachev S E Ivanov A V Faddeev L D 2017 Renormalization scenario for the quantum Yang–Mills theory in four-dimensional space–time Theoret. Math. Phys. 192 1134 1140

DOI

17
Ivanov A V 2018 About renormalized effective action for the Yang–Mills theory in four-dimensional space-time 20th Int. Seminar on High Energy Physics (Quarks-2018), EPJ Web of Conf. 191 p 06001

DOI

18
Synge J L 1931 A characteristic function in Riemannian space and its application to the solution of geodesic triangles London Math. Soc. 32 241 258

DOI

19
Synge J L 1960 Relativity: The general theory Amsterdam North-Holland Publishing Company 1 505

20
Barvinsky A O Vilkovisky G A 1985 The generalized schwinger–dewitt technique in gauge theories and quantum gravity Phys. Rept. 119 1 74

DOI

21
Faddeev L D 2002 Mass in Quantum Yang–Mills theory (comment on a Clay millenium problem) Bull. Braz. Math. Soc. (N. S.) 33 201 212

DOI

22
DeWitt B S 1967 Quantum theory of gravity. 2. the manifestly covariant theory Phys. Rev. 162 1195 1239

DOI

23
DeWitt B S 1967 Quantum theory of gravity. 3. applications of the covariant theory Phys. Rev. 162 1239 1256

DOI

24
Hooft G 't 1976 The background field method in gauge field theories, (Karpacz, 1975), Proc., Acta Universitatis Wratislaviensis , Wroclaw 1 345 369

25
Abbott L F 1982 Introduction to the background field method Acta Phys. Polon. B 13 33 50

26
Ya. Aref'eva I Slavnov A A Faddeev L D 1974 Generating functional for the S-matrix in gauge-invariant theories Theor. Math. Phys. 21 1165 1172

DOI

27
Ivanov A V Kharuk N V 2020 Two-loop cutoff renormalization of 4D Yang–Mills effective action J. Phys. G: Nucl. Part. Phys. 48 015002

DOI

28
Gross D J Wilczek F 1973 Ultraviolet behavior of nonabelian gauge theories Phys. Rev. Lett. 30 1343

DOI

29
Politzer H D 1973 Reliable perturbative results for strong interactions? Phys. Rev. Lett. 30 1346 1349

DOI

30
Jack I Osborn H 1982 Two-loop background field calculations for arbitrary background fields Nucl. Phys. B 207 474 504

DOI

31
Bornsen J P van de Ven A E M 2003 Three-loop Yang–Mills β-function via the covariant background field method Nucl. Phys. B 657 257 303

DOI

32
Ivanov A V Kharuk N V 2022 Formula for two-loop divergent part of 4D Yang–Mills effective action Eur. Phys. J. 82 997 C

DOI

33
Ivanov A V 2022 Explicit cutoff regularization in coordinate representation J. Phys. A: Math. Theor. 55 495401

DOI

34
Fock V 1937 Die eigenzeit in der klassischen- und in der quanten- mechanik Sow. Phys. 12 404 425

35
Gilkey P B 1975 The spectral geometry of a Riemannian manifold J. Differ. Geom. 10 601 618

DOI

36
Vassilevich D V 2003 Heat kernel expansion: user's manual Phys. Rept. 388 279 360

DOI

37
Ivanov A V Kharuk N V 2022 Special functions for heat Kernel expansion Eur. Phys. J. Plus 137 1060

DOI

38
Gibbons G W 1979 Quantum field theory in curved spacetime General Relativity, An Einstein Centenary Survey 639 679

39
DeWitt B S 1965 Dynamical Theory of Groups and Fields New York Gordon and Breach 1 248

40
Seeley R T 1967 Complex powers of an elliptic operator Singular Integrals Providence, RI American Mathematical Society vol 10 288 307 Proc. Sympos. Pure Math.

DOI

41
Lüscher M 1982 Dimensional regularisation in the presence of large background fields Ann. Phys. 142 359 392

DOI

42
Kharuk N V 2021 Zero modes of the Laplace operator in two-loop calculations in the Yang–Mills theory, Questions of quantum field theory and statistical physics Part 28, Zap. Nauchn. Sem. POMI, POMI, St. Petersburg 509 216 226

Outlines

/