[Paper Review] Treatment of top-quark backgrounds in extreme phase spaces: the "top $p_{T}$ reweighting" and novel data-driven estimations in ATLAS and CMS

L. Serkin|arXiv (Cornell University)|May 9, 2021

Particle physics theoretical and experimental studies2 references4 citations

TL;DR

This paper presents the 'top $p_{\mathrm{T}}$ reweighting' technique to correct mismodelling of top quark transverse momentum in Monte Carlo simulations, significantly improving agreement with LHC data in both ATLAS and CMS experiments. It also introduces novel data-driven methods to estimate $t\overline{t}$ backgrounds in extreme phase spaces—such as high jet multiplicities and high $p_{\mathrm{T}}$—reducing reliance on uncertain simulations and enhancing sensitivity to new physics beyond the Standard Model.

ABSTRACT

The top quark plays an important role in searches for physics beyond the SM, both as a dominant background and as a key signature for the signal. The most notable feature found in the top physics analyses in both ATLAS and CMS Collaborations: the disagreement between simulation and data of the top quark $p_{T}$ spectrum - is highlighted. A reweighting procedure which significantly improves the agreement between simulation and data, also known as the "top $p_{T}$ reweighting", is summarised. Commonly raised points concerning the reweighting to fixed-order predictions are discussed, and several refined approaches are mentioned. An overview of several data-driven methods developed and used to estimate the $t\bar{t}$ background in regions with large jet and b-jet multiplicities and/or high top quark $p_{T}$ is presented.

Motivation & Objective

• Address the persistent discrepancy between simulated and observed top quark $p_{\mathrm{T}}$ spectra in $t\overline{t}$ events at the LHC.
• Reduce systematic uncertainties in top quark measurements and searches by improving Monte Carlo simulation fidelity.
• Develop and validate data-driven techniques to estimate $t\overline{t}$ backgrounds in high-multiplicity and high-$p_{\mathrm{T}}$ regions where simulations are unreliable.
• Enable more robust searches for new physics beyond the Standard Model in extreme phase spaces by minimizing simulation-dependent systematics.

Proposed method

• Apply empirical reweighting of simulated $t\overline{t}$ events based on the ratio of measured differential cross sections to NLO+PS Monte Carlo predictions.
• Use fixed-order QCD and electroweak corrections (NNLO QCD and NLO EW) as reference predictions for reweighting to improve agreement with data.
• Implement data-driven background estimation techniques such as functional fitting of $t\overline{t}$ invariant mass spectra and kernel-based two-dimensional template construction.
• Utilize control regions and the ABCD method with correction factors to estimate $t\overline{t}$ yields in signal-like regions without relying on simulation.
• Perform systematic uncertainty evaluations by varying fit functions, template morphologies, and jet $p_{\mathrm{T}}$ spectrum assumptions.
• Test iterative reweighting on multiple observables (e.g., top $p_{\mathrm{T}}$ and $t\overline{t}$ system mass) to account for correlations in high-energy tails.

Experimental results

Research questions

• RQ1 To what extent does top $p_{\mathrm{T}}$ reweighting based on fixed-order calculations improve agreement between simulation and data in $t\overline{t}$ events?
• RQ2 How do data-driven background estimation techniques perform in high-multiplicity and high-$p_{\mathrm{T}}$ regions where MC simulations fail?
• RQ3 What are the dominant sources of uncertainty in data-driven $t\overline{t}$ background estimates, and how are they quantified?
• RQ4 Can two-dimensional reweighting on top $p_{\mathrm{T}}$ and $t\overline{t}$ system mass improve simulation accuracy compared to single-variable reweighting?
• RQ5 How do ATLAS and CMS differ in their treatment of top $p_{\mathrm{T}}$ reweighting and data-driven estimation in high-energy searches?

Key findings

• Top $p_{\mathrm{T}}$ reweighting based on NNLO QCD + NLO EW predictions significantly improves agreement between simulation and data, reducing modelling systematics.
• The reweighting procedure reduces the discrepancy in the top quark $p_{\mathrm{T}}$ spectrum observed in both resolved and boosted regimes across ATLAS and CMS.
• Data-driven methods such as functional fitting and kernel-based template construction successfully model $t\overline{t}$ backgrounds in extreme phase spaces with uncertainties validated by control samples.
• In the $Z^\prime \to t\overline{t}$ search, the background spectrum derived from data via smooth fitting agreed with SM expectations within uncertainties.
• The $t\overline{t}t\overline{t}$ search using the $b$-tagging independence assumption yielded consistent results across systematic variations, validating the method.
• The ABCD method with correction factors for residual correlations enabled reliable $t\overline{t}$ estimation in SUSY searches, confirmed by control region studies.