Resonant Di-Higgs Production at Gravitational Wave Benchmarks: A Collider Study using Machine Learning

TL;DR

This paper explores the potential of combined gravitational wave and collider experiments to detect electroweak phase transitions, using machine learning to optimize di-Higgs signals at the LHC within the xSM model.

Contribution

It introduces a comprehensive analysis of gravitational wave signals and collider signatures for electroweak phase transitions, employing machine learning for optimized detection strategies.

Findings

Identifies promising benchmark points for gravitational wave detection.

Demonstrates machine learning effectiveness in collider signal discrimination.

Shows potential for combined gravitational wave and collider searches.

Abstract

We perform a complementarity study of gravitational waves and colliders in the context of electroweak phase transitions choosing as our template the xSM model, which consists of the Standard Model augmented by a real scalar. We carefully analyze the gravitational wave signal at benchmark points compatible with a first order phase transition, taking into account subtle issues pertaining to the bubble wall velocity and the hydrodynamics of the plasma. In particular, we comment on the tension between requiring bubble wall velocities small enough to produce a net baryon number through the sphaleron process, and large enough to obtain appreciable gravitational wave production. For the most promising benchmark models, we study resonant di-Higgs production at the high-luminosity LHC using machine learning tools: a Gaussian process algorithm to jointly search for optimum cut thresholds and tuning hyperparameters, and a boosted decision trees algorithm to discriminate signal and background. The multivariate analysis on the collider side is able either to discover or provide strong statistical evidence of the benchmark points, opening the possibility for complementary searches for electroweak phase transitions in collider and gravitational wave experiments.