Choose timezone
Your profile timezone:
Reconstruction Group Meeting
→
US/Eastern
Description
Zoom coordinates:
- URL: https://lbnl.zoom.us/j/94910574347?pwd=c01YQlg5bG04dFR6UWs3NjZkS3ZSdz09
- Meeting ID: 949 1057 4347
- Password: recon
Meeting Notes
- See the google doc
ePIC Reconstruction WG Meeting Minutes
Concerted by ChatGPT
Topic: Connecting central and far-forward detectors for physics reconstruction
Discussion participants: Chandradoy Chatterjee, Alexander Jentsch, Derek Anderson, Shujie Li
1. Bringing far-forward particles into a common collection
Alexander introduced two issues: handling the crossing angle consistently and allowing users to identify the detector source of each reconstructed particle. The immediate integration discussion concerned the Roman Pots and off-momentum detectors, rather than all far-forward detectors.
Their matrix reconstruction produces vectors in a local frame aligned with the hadron beam. Alexander explained that a 25 mrad rotation about the y-axis puts these vectors into the standard ePIC laboratory frame.
There was agreement on the overall direction: retain the existing detector-specific collections, rotate their outputs, and combine them with the central detector particles. The longer-term aim is for analyzers to use a common reconstructed-particle collection, with the standalone collections remaining available for detector studies and debugging.
2. Rotation algorithm and collection handling
Alexander initially suggested performing the rotation when copying particles from the existing reconstruction algorithms, since the rotation itself is a single operation. Derek said that writing directly to the combined collection from those algorithms was technically possible but not preferable.
Derek proposed the following sequence:
-
A small algorithm rotates the far-forward output.
-
The existing collection collector combines it with the central detector output.
-
The cloner converts the resulting subset collection into a full collection.
The links must be handled carefully, particularly MC associations. Derek noted that related work was already being considered alongside developments to remove the match-clusters algorithm.
Existing transformation tools were discussed as possible places to reuse functionality. Alexander pointed out that Undo Afterburner already reverses the afterburner transformations, but it operates on MC particles and has its own output branch. Derek noted that, even if the rotation were provided by a utility, an algorithm would still be needed to produce the transformed collection.
Outcome: Alexander agreed to add a small rotation algorithm and update the Roman Pot and off-momentum steering files. A more general transformation algorithm was discussed as a possible future extension; Alexander cautioned that generalization could introduce additional complications.
The output collection name was left open. Derek suggested a name that makes clear that the far-forward particles have been rotated.
3. Making detector-source information accessible
Alexander emphasized that users need to know which detector, or set of detectors, contributed to a reconstructed particle. He considered this necessary for making the combined collection useful.
The main reasons discussed were:
-
Pseudorapidity alone is not a reliable way to identify detector origin, especially for poorly reconstructed tracks.
-
Detector-specific information is needed for quality checks.
-
Users should be able to distinguish a Roman Pot proton from an unusual central tracking result.
-
Exclusive analyses need to identify the relevant scattered proton.
-
Straightforward access would reduce debugging work for analyzers and working group conveners.
Derek noted that users could, in principle, follow existing relations back to hits and extract detector IDs, but this requires several levels of traversal.
Two approaches were discussed. Derek favored a near-term utility returning the contributing subsystems, because it could be provided without changing the data model. Both supported eventually making the information directly accessible in reconstructed particles.
Possible representations included a bit field with helpers, an enum, a numeric identifier, a character, or a string. No representation was selected. Alexander was open to any format that makes the information accessible and favors a small, straightforward change.
4. Introducing the new collection without disrupting users
Alexander raised a concern about adding far-forward particles to the official collection before users can readily identify their source. He preferred having a solution to that issue when the change is deployed.
Derek proposed keeping the existing reconstructed-particle collection unchanged and giving the new combined collection a separate name during development. This would allow the work to be merged and checked by CI before an official switch.
Outcome: Alexander agreed with this approach.
Shujie suggested eventually renaming the existing collection to indicate that it contains central particles. Derek agreed with that direction. Alexander reiterated that collection naming alone would not replace accessible detector-source information.
5. Reconstruction status of the different far-forward detectors
Roman Pots and off-momentum detectors: In response to Shujie’s question, Alexander clarified that event reconstruction does not use MC-particle truth. It uses transport matrices and tabulated chromaticity information to extract momentum loss and transverse momentum from detector hits. Simulation is used to generate the tables. Both detector outputs would use the same rotation algorithm.
ZDC: Alexander said its outputs should already be in the standard laboratory frame and could, in principle, be combined largely as they are. He would need to check the available collections and appropriate contents with the ZDC experts.
B0: Alexander explained that B0 uses ACTS with truth seeding. Tracking with four layers remains an issue, while recent effort has focused on mechanical design. Neither Alexander nor Derek could confirm whether B0 output is already included in the official reconstructed-particle collection. Alexander raised the possible need to distinguish truth-seeded from realistically seeded particles.
Alexander also clarified that Roman Pot and off-momentum particles are reconstructed under a proton assumption: these detectors provide tracking without independent PID. For B0, some identification may be possible by combining calorimeter and tracking information.
Agreed next steps
-
Alexander: Implement the rotation algorithm, update the two steering files, and open a draft PR during the week of the meeting. Add Derek as a reviewer.
-
Derek: Help review the output interface and try to work on the collection-link issue during the same week.
-
Derek: Continue work on the subsystem-extraction utility, with time available the following week at the earliest.
The detector-source representation, collection names, ZDC inputs, and current inclusion of B0 in reconstructed particles remained unresolved.
There are minutes attached to this event.
Show them.