Towards the ultimate differential SMEFT analysis

TL;DR

This paper introduces a comprehensive differential analysis method for SMEFT that leverages full angular distributions to improve bounds on Higgs-gauge boson couplings, surpassing previous approaches.

Contribution

The paper presents a novel, generic approach using angular moments from differential distributions to maximize information extraction and avoid blind spots in SMEFT parameter constraints.

Findings

Achieves stronger bounds on Higgs couplings than previous methods.

Demonstrates the effectiveness of full differential data in constraining SMEFT parameters.

Provides a transparent alternative to machine learning techniques for SMEFT analysis.

Abstract

We obtain SMEFT bounds using an approach that utilises the complete multi-dimensional differential information of a process. This approach is based on the fact that at a given EFT order, the full angular distribution in the most important electroweak processes can be expressed as a sum of a fixed number of basis functions. The coefficients of these basis functions - the so-called angular moments - and their energy dependance, thus form an ideal set of experimental observables that encapsulates the complete multi-dimensional differential information of the process. This approach is generic and the observables constructed allow to avoid blind directions in the SMEFT parameter space. While this method is applicable to many of the important electroweak processes, as a first example we study the $pp \to V(\ell\ell)h(bb)$ process ($V \equiv Z/W^{\pm}, \; \ell\ell \equiv \ell^+\ell^-/\ell^\pm\nu$), including QCD NLO effects, differentially. We show that using the full differential data in this way plays a crucial role in simultaneously and maximally constraining the different vertex structures of the Higgs coupling to gauge bosons. In particular, our method yields bounds on the $h V_{\mu \nu}V^{\mu \nu}$, $h V_{\mu \nu}\tilde{V}^{\mu \nu}$ and $h Vff$ ($ff \equiv f\bar{f}/f\bar{f}'$) couplings, stronger than projected bounds reported in any other process. This matrix-element-based method can provide a transparent alternative to complement machine learning techniques that also aim to disentangle correlations in the SMEFT parameter space.