Feynman integrals are completely determined by linear algebra
Zhi-Feng Liu,Yan-Qing Ma
2022-01-01
Abstract:Introduction. — Feynman integrals (FIs) encode key information of quantum field theories. Study of FIs is important both for exploring mysteries of quantum field theories and for phenomenological application of them. Integrating over some variables is found to be a necessary step to determine FIs in all known systematic methods. This seems to be a reasonable phenomenon, because FIs themselves are defined by integrating over loop momenta. However, to determine FIs, is it possible to totally bypassing integration? Systematic methods to compute FIs on the market can be divided into direct methods and indirect methods. Direct methods include sector decomposition [1–13], MellinBarnes representation [8–13], loop-tree duality [14–23], and so on, where one computes FIs by directly performing integration over some variables. Indirect methods compute FIs indirectly by solving corresponding equations, which include difference equations [24–27] and differential equations [28–40]. To uniquely determine the solution, boundary information are needed in these indirect methods. Unfortunately, the only known systematic way to obtain boundary information is to use direct methods to calculate them. Therefore, integration is still necessary in these indirect methods. The auxiliary mass flow (AMF) method [38–40] is a kind of differential equations method, which computes FIs by setting up and solving differential equations with respect to an auxiliary mass term η (called η-DEs). The virtue of AMF is that its boundary conditions at η → ∞ are simply vacuum bubble integrals, which can be much easily calculated by using other methods [41–50]. The observation in this Letter is as following. Boundary information for AMF, which can be casted to singlemass vacuum FIs, can always be related to propagator integrals (p-integrals) with one less loops. Then, p-integrals can again be calculated by using the AMF method with input of new boundary information with one less loops. By using this strategy iteratively, we eventually do not need any input for boundary information in the AMF framework. It is thus surprising to find that integration is totally bypassed in determining FIs. As a result of our observation, FIs are completely determined by linear relations between FIs, which are used to decompose all FIs to a small set of bases, called master integrals (MIs), and to set up η-DEs of these MIs. We note that numerically solving ordinary differential equations (like η-DEs) is a well solved mathematical problem [51]. Therefore, the problem of integrating over loop momenta is now conceptually changed to an linear algebraic problem of exploring the linear space of FIs. In the rest of the Letter, we first review the AMF method and emphasize its input. We then explain how to compute boundary conditions within the AMF framework, without any input of integration. Finally, we give some examples to verify our strategy. Before continue, let us first give a brief introduction to FIs. A family of FIs are defined by the following integrals with various ~ν,