fastNLO

fastNLO
Name	fastNLO
Programming language	C++, Python, Fortran
Operating system	Linux, macOS

fastNLO

fastNLO is a computational framework for fast evaluation of perturbative quantum chromodynamics cross sections used in high-energy physics analyses. It provides precomputed interpolation tables enabling rapid recalculation of next-to-leading order and higher-order predictions for jet and hadron collider observables, allowing experimental collaborations and theory groups to compare data from facilities such as the Large Hadron Collider, Tevatron, HERA, CERN, and Fermilab with a variety of parton distribution functions and scale choices.

Overview

fastNLO implements a tabulation strategy to decouple the expensive fixed-order matrix-element integration from repeated convolutions with parton distribution functions associated with collaborations like ATLAS, CMS, ALICE, LHCb, CDF, and DØ. By providing interpolation grids, fastNLO enables rapid re-evaluation needed in fits performed by groups such as NNPDF, CTEQ-TEA, MMHT, HERAPDF, and ABM. The framework interfaces with PDF libraries and fitting tools employed by institutions including Rutherford Appleton Laboratory, DESY, SLAC, Brookhaven National Laboratory, and Max Planck Institute for Physics to support analyses of observables measured at experiments like UA1, UA2, ALEPH, DELPHI, OPAL, and L3.

History and development

fastNLO originated from concerted efforts by theorists and experimentalists associated with collaborations around CERN and DESY to accelerate phenomenological studies after the commissioning of the LHC. Early methodological inspirations trace back to interpolation strategies used in programs like APPLgrid and matrix-element tools such as MCFM, NLOJet++, and POWHEG. Development involved contributors from universities and laboratories including Université de Genève, University of Oxford, University of Cambridge, University of Manchester, ETH Zurich, and University of Hamburg, and draws upon numerical techniques refined in projects like LHAPDF and standards set by HEPData and ROOT.

Technical architecture and methodology

The core architectural concept is numerical factorization of the hadronic cross section into kernel tables and convolution with PDFs. fastNLO stores pre-integrated coefficient functions on multidimensional interpolation grids that sample phase-space variables analogous to methods found in SHERPA, MadGraph, and Herwig. Tables are produced by fixed-order generators and then accessed by runtime code written in C++, with bindings to Python and interfaces compatible with Fortran workflows. The methodology supports renormalisation and factorisation scale variations and systematic eigenvector variations from PDF sets from providers like CTEQ, NNPDF Collaboration, MSTW, and HERAPDF Collaboration to enable fast propagation of theoretical uncertainties in global fits led by consortia such as PDF4LHC.

Applications and use cases

fastNLO is used to compute jet cross sections, inclusive jet spectra, dijet mass distributions, and event-shape observables for comparison to measurements published by ATLAS Collaboration, CMS Collaboration, ALICE Collaboration, LHCb Collaboration, H1 Collaboration, and ZEUS Collaboration. It is employed in global PDF fits performed by NNPDF Collaboration, CTEQ-TEA, MMHT Collaboration, and ABMP to include collider data, and in studies of strong coupling constant determinations by collaborations and working groups such as LEP Electroweak Working Group and LHC Higgs Cross Section Working Group. fastNLO also facilitates phenomenological studies on parton shower matching and tuning involving Pythia, Herwig, and matrix-element merging schemes developed by groups at CERN Theory Department and universities worldwide.

Validation and performance

Validation of fastNLO tables is carried out by cross-comparing to reference calculations from codes like NLOJet++, MCFM, POWHEG-BOX, and analytic results documented by collaborations including ATLAS and CMS. Performance benchmarks demonstrate orders-of-magnitude speedups for repeated convolutions compared to direct fixed-order runs, enabling large-scale Monte Carlo scans and PDF uncertainty propagation by consortia such as PDF4LHC Working Group and analysis groups at Fermilab. The framework is stress-tested against experimental phase-space coverage provided by detectors at CERN, Fermilab, and DESY to ensure stability across kinematic ranges probed in measurements like inclusive jet production and deep-inelastic scattering from HERA.

Software implementation and interfaces

Implementations include core libraries in C++, Python bindings used by analysis frameworks at ATLAS Collaboration and CMS Collaboration, and data formats compatible with tools such as HEPData, Rivet, and ROOT. Integration with PDF access libraries like LHAPDF and fitting platforms used by NNPDF and CTEQ-TEA facilitates inclusion into global analysis pipelines. Packaging and distribution practices align with software management at institutions such as CERN IT, KIT, and collaboration software stacks maintained by ATLAS Computing and CMS Computing.

Adoption and impact in particle physics

fastNLO has been adopted by experimental collaborations and theory groups to accelerate phenomenological studies, enabling rapid re-evaluation of theory predictions during analyses by ATLAS Collaboration, CMS Collaboration, and PDF fitting groups like NNPDF Collaboration and CTEQ-TEA. Its impact includes tighter constraints on parton distribution functions, improved determinations of the strong coupling constant reported in publications by Particle Data Group summaries, and enhanced ability for working groups such as PDF4LHC Working Group and LHC WG on PDFs to assess theoretical uncertainties across multiple datasets. The tool has influenced subsequent developments in grid-based approaches exemplified by APPLgrid and continues to be a component in reproducible analysis ecosystems used across major laboratories and universities.

Category:Computational physics software