Geant4 Geometry and Transport Working Group Meeting

Tuesday 14 Jan 2025, 16:00 → 17:40 Europe/Zurich

32/1-A24 (CERN)

16:00 → 17:00 Status and Plans for 2025: Brief Reports and Discussion

Round-table discussion of 2024 achievements and proposed 2025 deliverables / work topics

• Proposed objectives / deliverables
  • Improved GPU navigation using surfaces (VecGeom) - Severin & Andre
    • Improvements in stability
    • Merge into VecGeom master (targetting release 2.0)
    • Optimisation
• Progress reports and items of ongoing contribution
  • Raman S. : looking to resume contributing to VecGeom, including attending bi-weekly meetings
  • Guilherme L. : in 2024 fixed a few solids' implementations, created implementations of shapes for surface model, and investigated issues; intend to contribute in similar manner in 2025.
  • Ben M. : have & will tackle issues related to CMake configuration ; investigating use of HepNumerics (geant4/global category) use and optimisation in Geant4 in general - so any relevant to geometry will be included in this.

  • Ivana H. : reconsile the defaults of the field module with the default of the new Field ‘Builder’ classes

  • Seth J. : discussed the use of geometry debugging tools for potential contribution, 

    • a material scanner, for identifying problems with the different geometry modellers and implementations of solids (e.g. identified issue with twisted solid later reported by CMS )

    • a second tool provides a graphical representation of the isotropic safety distance on a place (or other 2-d surfaces?); useful for debugging this.

  • Evgueni: (2024 & future) implroving the robustness and performance of various G4Solids.  In 2024 these were: twisted Tubs, Generic Trap, poly-cone and polyhedra.  Also extending their testing suites.

 

New topics for investigation (Stewart B.)

1. Symplecticity of navigation can be destroyed in the case of a multiple volume geometry by repeated geometry transformation, in inaccuracies due to the composition of rotations.  For simpler use cases (likely including a g-2 type experiment) a simple solution is to create a single large volume with the field and geometry of the setup.  Potentially a geometry with just surfaces, that does not require transformations will help? For complex cases where multiple volumes are required, a new approach may be required.
   • To note as topic for future consideration: "Symplecticity and the impact of inaccuracies due to repeated composition of transformations (or from errors due simply to displacements from the inaccuracies going around a single revolution being amplified by repeated application)"
   • Stewart will gives us a reference or excerpt to document this issue
2. Interest in having a geometry modeller that allows the user to define a geometry setup using intersections of planes and quadratic (quadric?) half-spaces (‘half-space representation’)  to address use cases from accelerator application - ranging from LHC to medical applications.
   • Many geometries are described for simulation by FLUKA / MCNP / PHITS, and for simulation in Geant4 currently the geometry must be recreated, which is a significant burden
     • potential for performant implementation of these geometries with a compatible geometry modeller mentioned
   • Potential approaches discussed
     • developing Orange further for use by MCNP derived, OpenMC and related (FLUKA / PHITS) geometry descriptions ; an approach that adapted this to the G4 geometry state (a la ALICE / TGeo to G4 )  - Seth
     • consulting with Fluka team on their plans for a Geant4-embedded navigation system
     • [ recalled later / John ] a way to describe solids using half-spaces created by a student from Germany ( we should still have access to his repository - to be confirmed. )
   • Agreed to organise a dedicated meeting to discuss this in more depth.

 

16:00 Status and Outlook

Speaker: John Apostolakis (CERN)

Geometry-WG-Meeting-January2025.pdf

16:10 Reports, Open Topics and Plans - roundtable

2024 Deliverables

16:30 Key goals for 2025

Recall topics mentioned in 2024:

Navigation / modelling
- Vecgeom surface-based modeller
Performance / interfaces
- Reduction of initialisation time and overlap checking using Multi-Threading
Field:
- Robustness improvement of QSS3
- High order symplectic stepper (needs person)

Impact / Coupling with other developments (in Geometry & Transport and beyond)

See Deliverables (2023 & 2024) and potential deliverables (future)

2023 Deliverables (for reference)

Potential future development

16:50 Discussion

17:00 → 17:30 VecGeom bi-weekly progress meeting: Status Updates, open issues, plans

17:30 → 17:40 AOB: Other topics