Choose timezone
Your profile timezone:
HCal Muon ID discussion
→
US/Eastern
Today's meeting evolved around a kick-off discussion of muon ID with the ePIC calorimeters. It was suggested to absorb future such discussion into the general ePIC calo meeting every 2nd Wednesday at 12:30 EST and we will proceed like this.
Caroline gave a brief overview of existing muon ID studies with the ePIC LFHCal, nHCal and BHCal and anticipated that there are also studies of muon ID with the 2nd EIC detector. The ePIC muon ID studies currently mainly focus on single particle production and do not yet make use of track-cluster matching.
Elke recommended we detail our studies about how well we see the MIPs: what is the light output? Are the MIPs absorbed? And how good is the MIP/pedestal separation? The answers are expected to be quite different for the 3 ePIC HCals. Moreover (also pointed out by OlegE), single particle studies won't give us the final answer because of the overwhelming pion flux (as shown by Jihee later, a factor of 500 more pions than muons at 1GeV). Realistic background has to be folded in. The situation will be better with decay channels like JPsi—>mumu, which give a clearer signature.
OlegT asked what efficiencies we are aiming for. This will depend on the eventual physics channel. Daniel offers to come back to this meeting and give an overview of processes for that muon ID will be useful. Stephen Katy wrote in the chat: "From Exclusive/Diffractive/Tagging - We have VM (J/Psi as the focus currently) and TCS - J/Psi is the "easier" of the two. Alex and Gary can probably provide some truth level distributions of the muons for each, but they're ~50/50 barrel/endcaps (barrel tends to be lower momentum than Andrew has looked at with the single particle studies)." Elke commented that it is important to precisely formulate our goals since we are not a HEP experiment and have other physics goals. Moreover, we have an asymmetric collision system, thus asymmetric detector with non-uniform momentum distributions across the different regions.
Andrew detailed his muon ID studies with the BHCal. He mixes 1:1 single muons and pions and creates one log-likelihood-based PID discriminator L_particle from three estimators: E/p(HCal), E/p(EMCal), shower shape. By cutting L_mu - L_pi > 0, he creates a much muon-enhanced sample with some pion contamination. The DeltaL cut can later be adjusted depending on the specific needs of the analysis. It was suggested to also check the transverse cluster size (MIP doesn't shower and leaves signal in just one tile). This estimator is probably closely related to the shower shape = energy-weighted radius, but its effectiveness should be checked.
Leszek had prepared slides about the nHCal but couldn't join the meeting today. He studies di-muons in pythia photoproduction and extracted residuals and efficiencies. We did not explicitly go through the slides this time and should do it at a future meeting.
Jihee presented studies for the 2nd EIC detector: muon ID in the forward including a machine learning approach. One value is created - a MLP response - and cut on to distinguish between signal and background. She showed a particularly instructive plot: DeltaE(HCal) vs. DeltaE(EMCal). While muons are concentrated in one MIP-like hot spot, pions can be grouped in three categories: those that shower in the EMCal, in the HCal, or those that don't shower (MIP-like). Elke commented that the ePIC EMCals have at least one hadronic interaction length or even more, depending on the incident angle, and in addition there is energy loss in the magnet. Jihee then extracts a mis-identification rate (i.e., when a muon is wrongly ID'ed as a pion) and corrects it for the pion flux by reweighting with the cross section from pythia. While the uncorrected mis-ID rate is 1% (40%) for low (high) momentum muons, it blows up to almost 100% (70%) after the correction. This demonstrates that it is essential to include background in the estimate of the purity and as well the efficiency.
Anselm presented the hKLM for the 2nd EIC Detector based on 2511.08432 [physics.ins-det]. The hKLM is a compact combined barrel HCAL, muID and ToF. ML/AI-guided design optimization was used (AID2E, JINST 19 (2024) 07, C07001), as was done for the dRICH in ePIC. He offers advise if the calorimeter groups at ePIC should be interested in this tool.
There are minutes attached to this event.
Show them.
