Table of Contents
Fetching ...

Caravel: A C++ Framework for the Computation of Multi-Loop Amplitudes with Numerical Unitarity

S. Abreu, J. Dormans, F. Febres Cordero, H. Ita, M. Kraus, B. Page, E. Pascual, M. S. Ruf, V. Sotnikov

TL;DR

Caravel introduces the first public C++17 framework for computing multi-loop scattering amplitudes via numerical unitarity, combining D-dimensional integrand decomposition, finite-field arithmetic, and functional reconstruction. The paper details a modular software architecture with core arithmetic, tree-level currents, integrand-space construction, and master-integral libraries, plus example programs spanning tree, one-loop, two-loop, and finite-remainder calculations in QCD and Einstein gravity. It also showcases analytic reconstruction of amplitudes from numerical data, using MPI for parallelization and finite-field rational-function reconstruction to obtain analytic results. While not yet fully automated for arbitrary multi-loop multi-leg processes, Caravel provides a flexible platform to extend master integrals and process-specific inputs, with practical impact for precision predictions at the LHC and gravity scattering studies.

Abstract

We present the first public version of Caravel, a C++17 framework for the computation of multi-loop scattering amplitudes in quantum field theory, based on the numerical unitarity method. Caravel is composed of modules for the $D$-dimensional decomposition of integrands of scattering amplitudes into master and surface terms, the computation of tree-level amplitudes in floating point or finite-field arithmetic, the numerical computation of one- and two-loop amplitudes in QCD and Einstein gravity, and functional reconstruction tools. We provide programs that showcase Caravel's main functionalities and allow to compute selected one- and two-loop amplitudes.

Caravel: A C++ Framework for the Computation of Multi-Loop Amplitudes with Numerical Unitarity

TL;DR

Caravel introduces the first public C++17 framework for computing multi-loop scattering amplitudes via numerical unitarity, combining D-dimensional integrand decomposition, finite-field arithmetic, and functional reconstruction. The paper details a modular software architecture with core arithmetic, tree-level currents, integrand-space construction, and master-integral libraries, plus example programs spanning tree, one-loop, two-loop, and finite-remainder calculations in QCD and Einstein gravity. It also showcases analytic reconstruction of amplitudes from numerical data, using MPI for parallelization and finite-field rational-function reconstruction to obtain analytic results. While not yet fully automated for arbitrary multi-loop multi-leg processes, Caravel provides a flexible platform to extend master integrals and process-specific inputs, with practical impact for precision predictions at the LHC and gravity scattering studies.

Abstract

We present the first public version of Caravel, a C++17 framework for the computation of multi-loop scattering amplitudes in quantum field theory, based on the numerical unitarity method. Caravel is composed of modules for the -dimensional decomposition of integrands of scattering amplitudes into master and surface terms, the computation of tree-level amplitudes in floating point or finite-field arithmetic, the numerical computation of one- and two-loop amplitudes in QCD and Einstein gravity, and functional reconstruction tools. We provide programs that showcase Caravel's main functionalities and allow to compute selected one- and two-loop amplitudes.

Paper Structure

This paper contains 32 sections, 29 equations, 2 figures, 3 tables.

Figures (2)

  • Figure 1: Illustration of the interplay of the different modules in Caravel in generic calculations of scattering amplitudes. A black arrow indicates a dependence, a blue arrow means input for a module, and a red arrow the capacity of a module to deliver a given component of the calculation. Modules surrounded by a red box rely on external input to operate. All modules depend on the Core module.
  • Figure 2: Specifying momentum routing for the double-box tensor integral.