APPLgrid

APPLgrid
Name	APPLgrid
Developer	APPLgrid Collaboration
Released	2003
Latest release	2012
Programming language	C++
License	GPL
Operating system	Cross-platform

APPLgrid is a software library and format designed to accelerate repeated evaluations of next-to-leading order (NLO) perturbative calculations in high-energy physics by precomputing interpolation grids. It provides a bridge between fixed-order calculations and phenomenological applications by enabling fast reweighting and convolution with different parton distribution functions (PDFs) and scale choices. APPLgrid has been used in analyses related to the Large Hadron Collider, facilitating comparisons between theoretical predictions and experimental measurements and supporting global PDF fits and uncertainty estimates.

Overview

APPLgrid was developed to address challenges encountered when combining NLO calculations from codes like MCFM, NLOjet++, FEWZ, and MC@NLO with PDF extraction efforts associated with collaborations such as CTEQ, MSTW, NNPDF, and HERAPDF. By storing coefficient functions on a kinematic grid, APPLgrid enables rapid convolution with sets from providers like LHAPDF without re-running expensive matrix element integrations. Its workflow interacts with tools from the CERN computing ecosystem, including ROOT for histogramming and LHAPDF6 for PDF access, and supports phenomenology tasks connected to experiments such as ATLAS, CMS, LHCb, Tevatron, CDF, and DØ.

Design and Architecture

APPLgrid structures NLO information into multidimensional interpolation tables keyed by observables and partonic channels; these tables are produced by wrappers interfacing to generator codes like POWHEG BOX and Sherpa. The internal format emphasizes modularity and portability, storing grids in formats compatible with HDF5-like serialization and integration with GNU toolchains. The grid interpolation uses polynomial basis functions across nodes inspired by numerical techniques from projects associated with CERN Open Data initiatives and techniques similar to those used in fastNLO and aMCfast frameworks. Its design enables separation between matrix element computation performed by authors affiliated with institutions such as INFN, DESY, IHEP, and IPPP, and phenomenological analysis by research groups at universities like Oxford, Cambridge, MIT, and Stanford.

Applications and Use Cases

APPLgrid has been applied to a broad set of collider observables: inclusive jet production studied by ATLAS and CMS; Drell–Yan processes central to measurements by LEP legacy analyses and LHCb forward physics; top-quark pair production central to Tevatron legacy results; and electroweak boson production relevant to CDF and DØ precision studies. It supports PDF profiling efforts used by PDF4LHC working groups and has been incorporated into global fits conducted by CTEQ-TEA, MMHT, and NNPDF collaborations. Additional use cases include rapid theory predictions for new-physics searches undertaken at SLAC and FNAL, systematic uncertainty propagation for detector calibrations carried out by CERN experiment teams, and pedagogical exercises in academic groups at Princeton and UCL.

Performance and Accuracy

The principal performance benefit of APPLgrid is dramatic reduction in turnaround time: convolutions that would take hours with direct integration can be reduced to seconds, enabling large-scale replica fits and Monte Carlo error propagation used by HERAFitter and Professor. Accuracy depends on grid resolution, interpolation order, and binning choices; studies benchmarked by collaborations like ATLAS and CMS compared APPLgrid interpolations against reference calculations from MCFM and NLOjet++, demonstrating per-cent-level agreement for well-configured grids. Users from institutions such as DESY and KIT have quantified residuals introduced by interpolation and provided recipes to control systematic biases in precision observables like the W boson mass and jet cross sections measured in Run 1 and Run 2 datasets.

Development and Implementation

APPLgrid originated from collaborations among theorists and experimentalists affiliated with groups at University of Edinburgh, Universidad Autónoma de Madrid, CERN, and University of Torino. The codebase, primarily in C++, includes APIs for generating grids, reading stored tables, and performing convolutions with different PDF sets managed through LHAPDF. Implementation practices followed standards from software projects such as GitHub repositories and code review models used by ROOT-based analyses. Maintenance and extensions were driven by inputs from working groups like PDF4LHC and by requests from experimental analysis teams at ATLAS and CMS to support new observable definitions and differential distributions.

Adoption and Integration

APPLgrid gained adoption within the high-energy physics community through integration with analysis frameworks including Rivet, Professor, HERAFitter, and bespoke analysis pipelines at Pavia and DESY. Experimental collaborations used APPLgrid grids to provide theory tables alongside measured data in database formats compatible with HEPData records, enabling reproducibility and external validation by groups at Yale, Caltech, ETH Zurich, and Imperial College London. The tool also interacted with emerging fast-evaluation initiatives like fastNLO and converter projects such as aMCfast to harmonize interfaces between Monte Carlo generators and PDF-fitting machinery.

Limitations and Future Work

Limitations of APPLgrid include dependency on the availability of precomputed grids for specific observables and sensitivity to binning and interpolation artifacts when extrapolated beyond training ranges; this constrains direct application to novel observables proposed by groups at LAPP or IFIC. Future work envisioned by collaborations like PDF4LHC and research centers including SISSA and CERN Theory involves automation of grid generation via generator-level interfaces, improved support for higher-order corrections from efforts tied to NNLOJET and Geneva, and tighter integration with modern data formats promoted by HEPData and Zenodo. Continued community development aims to reduce interpolation bias for precision measurements relevant to programs at LHC Run 3 and potential future facilities such as Future Circular Collider.

Category:High energy physics software