NNLO Computational Techniques: the Cases H -> gamma gamma and H -> g g

TL;DR

This paper develops and systematizes advanced two-loop computational techniques for Higgs decay processes, providing practical methods, numerical results, and insights into electroweak and QCD corrections relevant for precise theoretical predictions.

Contribution

It introduces general rules and recipes for two-loop renormalization, handling unstable particles, and extracting collinear logarithms, with applications to Higgs decay calculations.

Findings

Two-loop corrections to H -> gamma gamma and H -> g g are quantified.

Electroweak corrections to Higgs production are between -4% and +6%.

Proper implementation of the computational scheme stabilizes two-loop correction estimates.

Abstract

A large set of techniques needed to compute decay rates at the two-loop level are derived and systematized. The main emphasis of the paper is on the two Standard Model decays H -> gamma gamma and H -> g g. The techniques, however, have a much wider range of application: they give practical examples of general rules for two-loop renormalization; they introduce simple recipes for handling internal unstable particles in two-loop processes; they illustrate simple procedures for the extraction of collinear logarithms from the amplitude. The latter is particularly relevant to show cancellations, e.g. cancellation of collinear divergencies. Furthermore, the paper deals with the proper treatment of non-enhanced two-loop QCD and electroweak contributions to different physical (pseudo-)observables, showing how they can be transformed in a way that allows for a stable numerical integration. Numerical results for the two-loop percentage corrections to H -> gamma gamma, g g are presented and discussed. When applied to the process pp -> gg + X -> H + X, the results show that the electroweak scaling factor for the cross section is between -4 % and + 6 % in the range 100 GeV < Mh < 500 GeV, without incongruent large effects around the physical electroweak thresholds, thereby showing that only a complete implementation of the computational scheme keeps two-loop corrections under control.