Search for black holes and sphalerons using novel machine learning techniques at CMS

TL;DR

This paper presents a novel machine learning approach to search for microscopic black holes and sphalerons in proton-proton collision data, setting new limits on their production and constraining models with large extra dimensions.

Contribution

It introduces a new event identification tool based on phase-space distance and provides model-independent limits, significantly improving sensitivity to black holes and sphalerons at the CMS experiment.

Findings

Black holes with masses below 9.0-11.4 TeV are excluded.

Upper limit of 0.0025 on sphaleron transition fraction.

Enhanced constraints on models with large extra dimensions.

Abstract

A comprehensive search for microscopic black holes and electroweak sphalerons is presented, using proton-proton collision data collected by the CMS detector during 2016-2018, corresponding to an integrated luminosity of $138~\mathrm{~fb}^{-1}$. A novel tool has been developed to identify collider events with distinct kinematic features, based on the phase-space distance between events. Model-independent limits are set on the cross section of new physics signals producing multiple jets and leptons, which are further interpreted as constraints on black hole and sphaleron production. In the context of models with large extra dimensions, semiclassical black holes with masses below 9.0-11.4 TeV are excluded by this search, significantly extending previous sensitivity. Additionally, a dedicated search for electroweak sphaleron transitions has been performed. An upper limit of 0.0025 is set at 95% confidence level on the fraction of quark-quark interactions with center-of-mass energy above the nominal threshold of 9 TeV that result in sphaleron transitions.