CIMBA: Fast Monte Carlo generation using cubic interpolation

Abstract

Monte Carlo generation of high energy particle collisions is a critical tool for both theoretical and experimental particle physics, connecting perturbative calculations to phenomenological models, and theory predictions to full detector simulation. The generation of minimum bias events can be particularly computationally expensive, where non-perturbative effects play an important role and specific processes and fiducial regions can no longer be well defined. In particular scenarios, particle guns can be used to quickly sample kinematics for single particles produced in minimum bias events. CIMBA (Cubic Interpolation for Minimum Bias Approximation) provides a comprehensive package to smoothly sample predefined kinematic grids, from any general purpose Monte Carlo generator, for all particles produced in minimum bias events. These grids are provided for a number of beam configurations including those of the Large Hadron Collider. Program summaryProgram title: CIMBA (Cubic Interpolation for Minimum Bias Approximation)CPC Library link to program files:http://dx.doi.org/10.17632/49m44md4ph.1Licensing provisions: GPL version 2 or laterProgramming language: Python, C++Nature of problem: generation of simulated events in high energy particle physics is quickly becoming a bottleneck in analysis development for collaborations on the Large Hadron Collider (LHC). With the expected long-term continuation of the high luminosity LHC, this problem must be solved in the near future. Significant progress has been made in designing new ways to perform detector simulation, including parametric detector models and machine learning techniques, e.g. calorimeter shower evolution with generative adversarial networks. Consequently, the efficiency of generating physics events using general purpose Monte Carlo event generators, rather than just detector simulation, needs to be improved.Solution method: in many cases, single particle generation from pre-sampled phase-space distributions can be used as a fast alternative to full event generation. Phase-space distributions sampled in particle pseudorapidity and transverse momentum are sampled from large, once-off, minimum bias samples generated with Pythia 8. A novel smooth sampling of these distributions is performed using piecewise cubic Hermite interpolating polynomials. Distributions are created for all generated particles, as well as particles produced directly from hadronisation. Interpolation grid libraries are provided for a number of common collider configurations, and code is provided which can produce custom interpolation grid libraries.Restrictions: Single particle generation