CALICO : Computing Annihilators from Linear Identities Constraining (differential) Operators Preprint: ZU-TH 44/25
Giuseppe Bertolini, Gaia Fontana, Tiziano Peraro · 2025
Read CALICO : Computing Annihilators from Linear Identities Constraining (differential) Operators Preprint: ZU-TH 44/25 by Giuseppe Bertolini, Gaia Fontana, Tiziano Peraro (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-13653.txt.
Introduction
Giuseppe Bertolini, Gaia Fontana, and Tiziano Peraro present a tool for finding relations among special functions in a 2025 study. High-energy physics work produces integrals whose kernels are multivalued. Operators that annihilate those kernels can generate useful identities. Finding such annihilators by hand is slow and error prone.
Their method computes parametric annihilators with fast linear solvers. Relations between broad classes of special functions then follow with little extra work. The authors show the approach across many integral forms rather than a single case. Machine identity finding speeds the core work behind particle physics predictions, where thousands of integrals must be reduced.
Catalog record
| CREATOR | Giuseppe Bertolini, Gaia Fontana, Tiziano Peraro |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 126 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-13653
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
CALICO : Computing Annihilators from Linear Identities Constraining (differential) Operators Preprint: ZU-TH 44/25 Giuseppe Bertolini, Gaia Fontana, Tiziano Peraro Abstract We elaborate on the method of parametric annihilators for deriving integral relations. Parametric annihilators are differential operators that annihilate multivalued integration kernels appearing in suitable integral representations of special functions. We illustrate this approach in a way that applies to a broad variety of integral representations. We describe a method for computing parametric annihilators based on efficient linear solvers and use them to derive relations between a wide class of special functions related to important problems in high-energy physics. We also formulate a similar method for deriving differential equations satisfied by the independent integrals within an integral family. We show applications to several classes of special functions, including hypergeometric functions, loop integrals in various representations (including Baikov, loop-by-loop Baikov, Lee-Pomeransky and Schwinger representations) and duals of loop integrals. We finally present the public Mathematica package CALICO for computing parametric annihilators and its usage in several examples of high relevance in theoretical particle physics. 1 Introduction Loop integrals are crucial ingredients for our theoretical understanding of quantum field theory. They contribute to precise phenomenological predictions about the fundamental interactions of nature, which are essential to match the uncertainty of modern experiments. They are also used for new approaches to classical theories — including gravity — which are increasingly important due to recent improvements in experimental observations. Due to the rich mathematical structure they exhibit, they are also an interesting subject for mathematical studies, which in turn we may exploit to improve our capability of making higher-order theoretical predictions in perturbation theory. The set of loop integrals which contribute to a theoretical prediction are generally not independent. They satisfy linear relations, such as symmetry relations and, most importantly, Integration By Parts ( IBP ) identities Tkachov:1981wb ; Chetyrkin:1981qh , which can be found via purely algebraic means. By finding and solving such relations, this set of integrals is reduced to a linear combination of an independent subset of them, called master integrals ( MIs ). The reduction to master integrals is an essential ingredient of most higher-loop theoretical predictions. Besides drastically reducing the number of integrals which need to be computed, it casts amplitudes in a form where many of its properties and symmetries are manifest, and it enables the method of differential equations Kotikov:1990kg ; Gehrmann:1999as for computing the master integrals themselves. IBPs are traditionally found using the momentum-space representation of loop integrals. However, many other representations of these integrals are known, including the Feynman, Schwinger, Baikov Baikov:1996iu (and its loop-by-loop variant Frellesvig:2017aai ) and Lee-Pomeransky Lee:2013hzt representations. Different representations offer different tradeoffs, hence their usefulness is highly dependent on the problem at hand. The Baikov representation, for instance, is used — among other things — in modern approaches to IBP reduction, combined with syzygy techniques, for finding simpler identities among loop integrals which do not contain higher-powers of denominators Larsen:2015ped . The Lee-Pomeransky representation has also been used in the context of integral reduction, most notably using the approach of parametric annihilators Baikov:1996iu ; Lee:2014tja ; Bitoun:2017nre . The latter proved to be extremely efficient in certain contexts, e.g. for reducing high powers of denominators (see e.g. vonManteuffel:2019wbj ) which often appear in integrals having desirable properties, such as quasi-finiteness Panzer:2014gra ; vonManteuffel:2014qoa or a pure functional form Henn:2013pwa . The method of parametric annihilator has been extensively described in Bitoun:2017nre for twisted Mellin integrals and in particular for the Lee-Pomeransky parametrization of loop integrals. It consists in finding differential operators in the integration variables which annihilate the multivalued integration kernel appearing in such integral representation. These operators are thus used to find linear identities obeyed by loop integrals. The main goals of this work are the following. We review and elaborate on the method of parametric annihilators for finding integral identities, with a focus on parametric representations of loop integrals. We formulate the method in more general terms, extending its range of applications to different integral representations — including Schwinger and loop-by-loop Baikov which had not been used in this context. In particular, we find the Schwinger representation yields very efficient systems of identities in some cases. We also illustrate a similar technique for finding differential equations satisfied by the master integrals. We provide a description of these methods in a language that is accessible to the scientific community of theoretical particle physics and phenomenology, in particular the one focusing on higher-order perturbative corrections, avoiding mathematical language that is not common knowledge in these fields. We describe and provide an implementation of annihilators and other important differential operators based on modern linear solvers which rely on cutting edge finite-field techniques vonManteuffel:2014ixa ; Peraro:2016wsq . This implementation is distributed as a Mathematica package named CALICO , which we believe can be a useful tool for studying a broad variety of integrals, including loop integrals in various parametric representations. Finally, we introduce a new way of characterizing and finding relations between duals of loop integrals Caron-Huot:2021xqj ; Caron-Huot:2021iev ; Fontana:2023amt — which play a crucial role in the study of loops within the framework of intersection theory Mizera:2017rqa ; Mastrolia:2018uzb — compatible with the approach discussed in this paper. The paper is structured as follows. In section 2 , we set some notation and describe the method of parametric annihilators in a way that covers a wide class of functions having a suitable integral representation. We also clarify the connection between this approach and others based on syzygy equations. In section 3 , we describe two methods for deriving differential equations, the second of which also relies on finding suitable differential operators in the integration variables. In section 4 , we illustrate a method for computing annihilators by solving linear constraints, which is suitable for applying modern linear solvers and finite-field methods. In sections 6 to 9 , we describe several applications of this approach, including different representations of loop integrals — namely the standard and loop-by-loop Baikov, Lee-Pomeransky and Schwinger representations — as well as duals of loop integrals. In section 10 , we describe the basic usage of the CALICO package. Finally, in section 11 we draw our conclusions and in appendix A we provide more details on the definitions of the polynomials which appear in the integral representations we use. 2 Integral families and parametric annihilators In this section we set up our notation and review the main concepts we use in this paper, namely parametric annihilators and their usage in finding linear identities within integral families. 2.1 Basic definitions and notation We use a multi-index notation. Given an n -dimensional multi-index \alpha=(\alpha_{1},\ldots,\alpha_{n}) and a list of n variables \mathbf{z}=(z_{1},\ldots,z_{n}) , we define the monomial \mathbf{z}^{\alpha}=\prod_{j=1}^{n}z_{j}^{\alpha_{j}} (2.1) and its total degree |\alpha|=\sum_{j=1}^{n}\alpha_{j}. (2.2) When this does not cause ambiguity, we define the derivative operator as \partial_{j}\equiv\frac{\partial}{\partial z_{j}}. (2.3) We consider an integral family , whose elements are linear combinations of integrals of the type I_{\alpha}=\int\differential^{n}\mathbf{z}\,\varphi_{\alpha}(\mathbf{z})\,u(\mathbf{z}), (2.4) where u(\mathbf{z}) is a multivalued function common to all integrals, also known as twist . In most of our applications, u takes the form u(\mathbf{z})=\prod_{j}B_{j}(\mathbf{z})^{\gamma_{j}}, (2.5) where B_{j}(\mathbf{z}) are polynomials and \gamma_{j} are generic exponents , which typically depend on regulators. Other