Tailored PDFs for New Physics searches

Abstract

Given the non-negligible interplay between parton distribution functions (PDFs) at large x and potential New Physics (NP) effects in the high-energy tails of hadron collider observables, a central question is which PDFs can be reliably employed in beyond-the-Standard-Model (BSM) analyses. In this work, we examine the fine balance between using PDF sets with small uncertainties in the large-x region -- crucial for maximising BSM sensitivity -- and adopting conservative PDF fits that exclude high-energy data potentially contaminated by unaccounted NP contributions. We systematically assess a range of conservative PDF fitting strategies designed to mitigate such biases and provide a recommendation for the class of PDFs best suited for robust BSM searches. In addition, we investigate the alternative approach of performing simultaneous fits of Standard Model Effective Field Theory (SMEFT) Wilson coefficients and PDFs, thereby consistently accounting for their mutual correlations. Starting from a toy model to illustrate the underlying mechanisms, we then analyse two realistic NP scenarios: one modifying high-mass Drell-Yan production and another affecting the high-invariant-mass tail of top-quark pair production. Both cases are representative of measurements that will be probed with high precision during the High-Luminosity phase of the LHC.

Figures (15)

• Figure 2.1: One--dimensional posterior distributions for the PDF parameter $f$ (left) and the EFT coefficient $c$ (right) obtained in the toy model. The red vertical lines indicate the true values $(f_{\rm true},c_{\rm true})$. The conservative fit is unbiased but exhibits inflated uncertainties, the BSM-biased fit yields artificially precise yet shifted posteriors, and the simultaneous fit correctly recovers both parameters in this closure--test configuration.
• Figure 3.1: Kinematic coverage of the data points included in the PDF fit. The points highlighted with a black edge are our HL-LHC projections, while the other data points correspond to existing measurements. The values of $x$ have been computed using a leading order approximation.
• Figure 3.2: Impact of the HL-LHC projections on the uncertainties of the quark-antiquark (left) and gluon-gluon (right) PDF luminosities.
• Figure 3.3: Projected impact of a flavour universal $W'$ with $m_{W'} = 13.8$ TeV ($\hat{W} = 8 \times 10^{-5}$) on high-mass Drell-Yan (DY) distributions measured by ATLAS and CMS in the HL-LHC phase (top row) and in the forward bosons production measured by LHC in the HL-LHC phase(bottom row) for charged current DY (left column) and neutral current DY (right column). The BSM signal is compared to the SM prediction and its total uncertainty, which is split into component due to PDF uncertainty and a component due to experimental uncertainties in the lower insets. In the case of high-mass DY the experimental uncertainty is split into a statistic and a systematic components, while in the case of forward DY distributions the statistic uncertainty is subdominant and is not displayed separately.
• Figure 3.4: SMEFT fit performed on the synthetic Drell-Yan high-mass data and HL-LHC pseudodata presented in Sect. \ref{['subsec:settings']} with $\hat{W} = 8 \times 10^{-5}$ injected. The 95% C.L. (grey) and 68% C.L. bands are displayed for a SMEFT fit in which the true PDFs are used as input set (top) and a fit in which the BSM-biased PDFs are used (bottom).