Bootstrapping form factor squared in ${\cal N}=4$ super-Yang-Mills

TL;DR

This paper introduces a bootstrap approach for form factor squared in maximally supersymmetric Yang-Mills theory, enabling systematic calculation of multi-loop integrands and unifying different configurations through diagrammatic constraints.

Contribution

It develops a novel graphical bootstrap method for form factor squared, fixing integrands using soft and collinear limits without unitarity cuts, and unifies various loop and leg configurations.

Findings

Explicit construction of master diagrams for N=3,4,5,6

Automatic inclusion of lower-point form factors up to four loops

Potential for higher N form factor bootstrap similar to amplitude methods

Abstract

We propose a bootstrap program for the {\it form factor squared} with operator ${\rm tr}(\phi^2)$ in maximally supersymmetric Yang-Mills theory in the planar limit, which plays a central role for perturbative calculations of important physical observables such as energy correlators. The tree-level $N$-point form factor (FF) squared can be obtained by cutting $N$ propagators of a collection of two-point ``master diagrams" at $(N{-}1)$ loops: for $N=3,4,5,6$ there are merely $1, 2, 4, 13$ topologies of such diagrams respectively, and their numerators are strongly constrained by power-counting (including ``no triangle" property) and other constraints such as the ``rung rule". Moreover, these two-point diagrams provide a ``unification" of FF squared at different numbers of loops and legs, which is similar to extracting (planar) amplitude squared from vacuum master diagrams (dual to $f$-graphs): by cutting $2\leq n<N$ propagators, one can also extract the planar integrand of $n$-point FF squared at $(N-n)$ loops, thus our results automatically include integrands of 2-point (Sudakov) FF up to four loops (where the squaring is trivial), 3-point FF squared up to three loops, and so on. Our ansatz is completely fixed using soft limits of (tree and loop) FF squared and the multi-collinear limit which reduces it to the splitting function, without any other inputs such as unitarity cuts. This method opens up the exciting possibility of a {\it graphical bootstrap} for FF squared for higher $N$ (which contains {\it e.g.} planar Sudakov FF to $N{-}2$ loops) similar to that for the amplitude squared via $f$-graphs. We also comment on applications to the computation of leading order energy correlators where new structures are expected after performing phase-space integrations.