DRICH: move PDU placements into per-sector sensorbox subvolumes

[wdconinc]

Summary

Group the 210 PDU assemblies per sector into dedicated per-sector gas subvolumes (DRICH_sensorbox_sec{N}) to reduce the number of direct daughters of DRICH_gas from 1,269 to 15. This improves TGeo voxelisation and navigation performance for optical photon tracking in Geant4.

Motivation

Perfetto-traced forward-θ simulation (θ ∈ [3°, 25°], 2 events) showed:

• DRICH_gas had 1,269 daughters (6 × 210 PDU assemblies + aerogel + airgap + filter + 6 mirrors)
• Navigation cost: 3 µs/call (3× higher than other volumes), totalling 61 ms = 49% of all navigation time
• Root cause: TGeo must voxelise and search 1,269 irregularly-placed children at every optical photon step

Geometry change

Before:

DRICH_gas: 1,269 daughters
  aerogel, airgap, filter  (3)
  mirror_sec0..5            (6)
  pdu_sec0[0..209]         (210)
  pdu_sec1[0..209]         (210)
  ...
  pdu_sec5[0..209]         (210)

After:

DRICH_gas: 15 daughters
  aerogel, airgap, filter  (3)
  mirror_sec0..5            (6)
  sensorbox_sec0..5         (6)  ← new per-sector subvolumes
    each: 210 PDU assemblies

Implementation

The gasvolSensorboxTube solid was already defined (and used in the gas union solid). Per-sector Volume objects are created using this solid with the same gas material, placed in gasvolVol at the same position as the union contribution. PDU placement transforms are adjusted to the sensorbox local coordinate frame via the composed inverse transform.

No changes to:

• physVolIDs (sector, pdu, sipm) — still set at the same levels
• Optical skin surfaces (on pssVol, unaffected by hierarchy)
• IRT2 reconstruction (uses pduAssemblyPlacement in gas coordinates for position calculations)
• Compact XML / detector constants

[Copilot]

Pull request overview

This PR restructures the dRICH geometry so that the 210 PDU assemblies in each sector are placed inside a dedicated per-sector “sensorbox” gas subvolume, reducing the number of direct daughters under DRICH_gas and improving TGeo voxelisation/navigation performance for optical photon tracking.

Changes:

• Add per-sector sensorbox gas subvolumes (DRICH_sensorbox_sec{N}) placed into DRICH_gas using the existing gasvolSensorboxTube solid.
• Move PDU placements from direct children of DRICH_gas to children of the corresponding per-sector sensorbox, transforming placements into the sensorbox local frame via sensorboxGasPlacement.Inverse() * pduAssemblyPlacement.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[github-actions]

Capybara summary for PR 1115

• rec_dis_10x100_minQ2=1000_epic_craterlake
• rec_dis_10x100_minQ2=1_epic_craterlake
• rec_dis_18x275_minQ2=1000_epic_craterlake
• rec_dis_18x275_minQ2=1_epic_craterlake
• rec_dis_5x41_minQ2=1_epic_craterlake
• rec_e_1GeV_20GeV_epic_craterlake
• rec_pi_1GeV_20GeV_epic_craterlake
• sim_chargedgeantino_1GeV_20GeV_epic_craterlake
• sim_dis_10x100_minQ2=1000_epic_craterlake
• sim_dis_10x100_minQ2=1_epic_craterlake
• sim_dis_18x275_minQ2=1000_epic_craterlake
• sim_dis_18x275_minQ2=1_epic_craterlake
• sim_dis_5x41_minQ2=1_epic_craterlake
• sim_e_1GeV_20GeV_epic_craterlake
• sim_geantino_1GeV_20GeV_epic_craterlake
• sim_pi_1GeV_20GeV_epic_craterlake
  ^{_{Last updated 2026-06-06T19:39-04:00 df2bf72}}

[wdconinc]

As in #1051, this introduces changes in the number of steps (boundary injected) so hits are in general going to be different.

[wdconinc]

    ┌────────────────────┬──────────────────────────┬──────────────────────┬─────────────────┐
    │ Metric             │ Before (1,269 daughters) │ After (15 daughters) │ Improvement     │
    ├────────────────────┼──────────────────────────┼──────────────────────┼─────────────────┤
    │ DRICH nav time     │ 61.1 ms                  │ 38.8 ms              │ −36%            │
    ├────────────────────┼──────────────────────────┼──────────────────────┼─────────────────┤
    │ DRICH nav/call     │ 3 µs                     │ 2 µs                 │ −33%            │
    ├────────────────────┼──────────────────────────┼──────────────────────┼─────────────────┤
    │ DRICH % of all nav │ 49%                      │ 30%                  │ —               │
    ├────────────────────┼──────────────────────────┼──────────────────────┼─────────────────┤
    │ fEcalWScFi nav     │ 44.7 ms                  │ 35.4 ms              │ −21% (paramvol) │
    ├────────────────────┼──────────────────────────┼──────────────────────┼─────────────────┤
    │ Total s/event      │ ~27 s                    │ ~20 s                │ −26%            │
    └────────────────────┴──────────────────────────┴──────────────────────┴─────────────────┘

(ignore the fEcalWScFi line which is the impact from #1051 in this forward region; also, time per event is inflated due to profiling instrumentation)