Introduction

The COnfigurable Calorimeter simulatiOn for Ai (COCOA) is a nearly-hermetic calorimeter simulated with Geant4 and interfaced to the Pythia8 event generator. This open-source simulation is aimed to support the development of machine learning algorithms in high energy physics that rely on realistic particle shower modeling, such as reconstruction, fast simulation, and low-level analysis.

The COCOA calorimeter comprises a barrel and endcap system with configurable granularity, and with nearly uniform material depth distribution in pseudorapidity. An inner tracker system consisting of silicon and iron layers immersed in a magnetic field can be included optionally, along with basic tracking emulation. Output data are processed using on-board algorithms for topological clustering of calorimeter cells, graph creation, and jet clustering. The COCOA geometry is also provided in a format supporting event visualization with Phoenix.