Skip to content

o2-sim: Fix event aborts and the exit block in VecGeom navigation - #16133

Merged
sawenzel merged 2 commits into
AliceO2Group:devfrom
sawenzel:swenzel/o2sim-vecgeom-update
Nov 16, 2026
Merged

sawenzel merged 2 commits into
AliceO2Group:devfrom
sawenzel:swenzel/o2sim-vecgeom-update

Conversation

@sawenzel

Copy link
Copy Markdown
Collaborator

This is fixing a bug where o2-sim with G4.navmode=kVecGeom (default kRelocating navigator) aborted events when an e± ran tangent to a cylindrical face in the RB24 beam-pipe region.

Such a track stays within the surface tolerance of the face for microns. With a field-free medium on one side and a field on the other, the two sides disagree about which of them holds the track, and the stuck-track push of 100 kCarTolerance was too short to leave that band before the abort after 25 zero steps. The push now grows with every further zero step, up to about 65 µm before the abort.

In addition, the volume blocked after a step that leaves a daughter and its mother at once is now the one G4Navigator blocks. This is the fix of https://gitlab.cern.ch/VecGeom/g4vecgeomnav/-/merge_requests/28 for the same code.

With these changes, 1500 Pythia8 pp events (ITS + PIPE) and 20 Pb-Pb events (full detector without ZDC) run without aborted tracks, where there were 5 and, in earlier tests, 4–8 per 20 Pb-Pb events. The ITS hit count relative to TGeo navigation is 0.999 ± 0.005.

Assisted by Claude Code.

sawenzel and others added 2 commits November 15, 2026 22:38
This fixes the volume blocked after a step that leaves a daughter and its mother at once.

- The blocked volume was the deepest volume left, a grandchild of the new volume, which VecGeom's navigators never match.
- It is now the volume left that is a daughter of the deepest volume the old and new paths share, as in G4Navigator.
- This is the fix of G4VecGeomNav for the same code.

https://gitlab.cern.ch/VecGeom/g4vecgeomnav/-/merge_requests/28

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
This fixes events aborted with kRelocating when a track runs tangent to a curved face.

- Such a track stays within the tolerance of the face for about sqrt(2 R tolerance), microns, and a field-free medium on one side and a field on the other make the two sides disagree about which of them holds it.
- The push after ten zero steps was 100 kCarTolerance, as in G4Navigator, which is ten VecGeom tolerances and too short to leave that band before the abort at twenty-five.
- It now grows fourfold per try up to a micron and doubles from the eleventh try on, about 65 um before the abort.
- The abort stays for a track that is stuck for another reason.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@sawenzel
sawenzel requested a review from a team as a code owner November 16, 2026 11:00
@sawenzel
sawenzel merged commit 3289589 into AliceO2Group:dev Nov 16, 2026
11 checks passed
@sawenzel
sawenzel deleted the swenzel/o2sim-vecgeom-update branch November 16, 2026 14:19
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

1 participant