Ntuples for NLO Events at Hadron Colliders
Z. Bern, L. J. Dixon, F. Febres Cordero, S. Hoeche, H. Ita, D. A. Kosower, D. Maitre
TL;DR
This work presents an event-file format (nTuple) for disseminating NLO QCD predictions at hadron colliders, enabling efficient reweighting for different scales and PDFs without recomputing matrix elements. It details the data structure, process collections, and conventions (W/Z+jets and pure jets up to 4 jets), as well as a C++/Python library to read and reweight the files. By storing Born, virtual, real, and subtraction components with extensive metadata, the approach supports flexible jet algorithms and cuts while providing robust methods for uncertainty estimation. The framework facilitates widespread use by theorists and experimentalists, offering practical tools to study scale and PDF uncertainties and to compare results across analyses at the LHC.
Abstract
We present an event-file format for the dissemination of next-to-leading-order (NLO) predictions for QCD processes at hadron colliders. The files contain all information required to compute generic jet-based infrared-safe observables at fixed order (without showering or hadronization), and to recompute observables with different factorization and renormalization scales. The files also make it possible to evaluate cross sections and distributions with different parton distribution functions. This in turn makes it possible to estimate uncertainties in NLO predictions of a wide variety of observables without recomputing the short-distance matrix elements. The event files allow a user to choose among a wide range of commonly-used jet algorithms and jet-size parameters. We provide event files for a $W$ or $Z$ boson accompanied by up to four jets, and for pure-jet events with up to four jets. The files are for the Large Hadron Collider with a center of mass energy of 7 or 8 TeV. A C++ library along with a Python interface for handling these files are also provided and described in this article. The library allows a user to read the event files and recompute observables transparently for different pdf sets and factorization and renormalization scales.