Skip to content

Aegir

Aegir is the simulation framework for the SHiP experiment.

It is built on Phlex, ROOT, Geant4 and Pythia8.

Repository: ShipSoft/aegir

API reference: Doxygen

Quick start

git clone https://github.com/ShipSoft/aegir.git
cd aegir
pixi run build
pixi run phlex -c workflows/pythia8_mt.jsonnet

See Installation for details.

Requirements

pixi installs everything for you. For reference, aegir 0.4 needs:

Features

The following capabilities are already implemented in aegir or provided through the Geometry Service and Field Service:

Event generators

  • Configurable particle gun with Philox / Random123 RNG (particle_gun_source)
  • Pythia8 fixed-target p–p generation, serial or PythiaParallel (pythia8_source)
  • Dual-target generation with p–p / p–n interaction-point sampling based on the material Z/A (fixed_target_source)
  • Pre-generated events read back from an RNTuple file (file_source), the equivalent of FairShip's MuonBackGenerator. It accepts MCParticle and SimParticle input, so aegir output can be fed into a new job.
  • GENIE rootracker files read without linking GENIE (genie_reader_source)
  • GENIE neutrino interactions generated in-process by the separate aegir-genie plugin (genie_source)
  • Every source publishes an EventHeader (event weight and provenance) next to its particles, and can record the full generator history including all mothers

Geometry (selectable per workflow)

  • Built-in test geometry: W target + Si scoring planes (geometry_builtin_provider)
  • GDML file geometry (geometry_gdml_provider)
  • GeoModel .db geometry via the Geometry Service (geometry_geomodel_provider)
  • Configurable sensitive / scoring volumes

Magnetic field (via the Field Service)

  • Field maps in the SHiP field-map format (RNTuple), configured per magnet from the workflow (field_map_provider)
  • Field-off / null provider (field_null_provider)

Transport & simulation

  • Geant4 transport, single- and multi-threaded with configurable concurrency (geant4_module)
  • Energy-cut transport and scoring-plane hit saving
  • Built-in profiling: Chrome tracing, flamegraphs and benchmarks

Output

  • RNTuple output with parallel writing (sim_output_module): generator particles (mc_particles) and, for full simulation, hits (sim_hits) and transported particles (sim_particles)
  • Validation histograms (ROOT 7)

Workflows

  • Ready-to-run Phlex jsonnet workflows for gun, Pythia8, fixed-target and file-input setups in single-threaded / multi-threaded / generator-only variants (workflows/)

Three steps to target simulation

  1. Collide proton beam at 400 GeV on fixed protons and neutrons using Pythia8
    • Beam travels in the z direction only, but smearing and painting are applied
    • Random choice for beam to interact with protons or neutrons based on the Z/A of the material it traverses
  2. Transport particles with GEANT4 (and decay them, make them interact with matter, etc.)
    • Interactions with matter is time-consuming bit of target simulation
    • To save time, particles below a certain energy are not transported
  3. Save particles which reach a sensitive plane right after the hadron absorber (HA)
    • Only particle above a certain energy are saved

This simulation contains only target complex and hadron absorber. The magnetic field is also turned off.

Once the target simulation is produced, particles (such as muons) can be further propagated through the detector. This is much less time consuming, which is why there is a separation between "target simulation" and the rest of the simulation.

GENIE neutrino generation

Neutrino interactions come from the separate aegir-genie plugin. See GENIE neutrino generation.