sPHENIX tracking discussion

[Anthony Frawley <afrawley AT fsu.edu>]

Hi All,

Following some discussion of clustering efficiency in the TPC, I had a look at how the truth/reco cluster matching is done for the SvtxEvaluator ntp_cluster and ntp_g4cluster ntuples. I think that what I found is useful to know for anyone who uses these.

1) The ntp_cluster (reco cluster) ntuple:

-- Loops over all reco clusters (when scan_for_embedded is off).

-- Gets a list of truth particles that contributed to this reco cluster.

-- Loops over all those truth particles and finds the one with the largest truth cluster energy in the same layer as the reco cluster. Chooses that truth cluster.

-- Checks that the truth cluster matches the reco cluster position with a 4 sigma cut.

This fails for (typically) a few clusters per track because there is a secondary particle that contributes SOME energy to the reco cluster, and happens to have the highest truth cluster energy in the layer. But that truth cluster centroid fails the position matching cut. So the matching failure is not because the clustering failed, it is because the matching logic failed to find the correct g4particle.

This matching logic could be improved.

2) The ntp_g4cluster (truth cluster) ntuple:

-- Loops over all truth clusters

-- Gets all contributing g4hits

-- Gets all reco clusters with any contribution from these g4hits

-- Checks for a 4 sigma position match to the truth cluster.

This logic is more robust, and almost always finds a matching reco cluster.

Bottom line: The ntp_g4cluster truth/reco matching is more reliable than the ntp_cluster matching. You should use ntp_g4cluster to check the clustering efficiency.

Tony