Difference between revisions of "SoLID Ecal Weekly 20230518"

Latest revision as of 17:13, 1 June 2023

SoLID ECal Weekly May 18, 2023

Update

• Mike converted Darren's Cherenkov code to C++

Discussion on Tracking

• Presentation by Mike: Tracking Comparison
  • projected position on SC-D is outside the SC-D size (in y).
  • GEM x minus EC cluster x has resolution of 26 mm (same for y).
  • Xinzhan's suggestions:
    • latest cooking is still not using the latest tracking -> Mike will update cooking code and redo the analysis
    • if done correct, we expect GEM resolution to be 1mm, and total about 10mm (convolution of GEM and EC cluster, plus realistic factors)
    • We can get the true value of (x,y) of different detectors by examining the scintillator hit-required GEM position
    • not worth trying ECal- assisted because it only speed up the track-finding but our detector setup is so small that the effect will be none.

Discussion on Simulation

• Presentation by Ye: 18deg_rate_data_simulation_May182023.pdf
  • showed basic tree structure of simulation (rate is included).
    • For beam-on-target, the PID of secondary particles at each detector plane is not saved yet because there can be more than one particle/flux. Eendsum is currently the sum of all fluxes.
    • Cherenkov simulation used CO2, but very close to N2 (our test).
  • continued rate comparison. For data, used event vs. time to cut out beam trips
    • 18 deg LD2 10uA run, rates agree within 20%: data/sim~0.8, but for 5uA run data/sim becomes 0.5. ??
    • Boiling test, TS4 rate doesn’t look linear with run 4755 rate super high at 70uA. Tim mentioned the total rate did not go up much from 4754 to 4755. -> need to check rate extraction.
  • Other checks:
    • Run 4696 (60uA LD2), Shower_L seems to have a bump at 1500. Did MIP shift? -> Tim will help check this.

Discussion on Cherenkov

• First, it was pointed out that the cone size in arXiv is too big by factor 2 (figure had the wrong size). The cone should be filling up about 4 MAPMT channels.
• Presentation by Darren (simulation)
  • pi0 behaves very similar to e-, with Npe/Nch=4
  • pi+ and pi- mostly do not give Npe, but when they do, appears to be the same as e- (Npe/Nch=4)
• Presentation by Tim (data) Cherenkov Data
  • plotting the same Npe/Nch shows no separation between accidental background (supposedly Npe/Nch=1) and other events. Adding progressively high ShowerSum cuts seems to separate the electrons from the Npe/Nch=1 peak, but only barely.
  • This is perhaps consistent with our earlier observation that Cherenkov light yield is only half of what simulation shows. It means that while we can use CerSum=0 to select pions, we can't do a "high Cersum" cut to select electron samples.

Discussion on AI/ML for PID

• Presentation by Darren
• To do:
  • apply trigger cut and make the simulation sample as close to real data as possible
  • is AI/ML PID orthogonal to classical method?

Discussion on Charged Particle PID Analysis Plan

(Xiaochao)

• Starting point: 18 deg runs:
  • TS1 = CerSum
  • TS2= SC-B. and. SC-D
  • TS3 = SC-A .and. SC-D → SC-C.and.SC-D.and.ShowerSum
  • TS4 = ShowerSum
• Select “center-hit” events:
  • use GEM projected position, check that the event has hits in all detectors of the trigger
  • use GEM tracking to select events that hit center of ECal
• Setup:
  • Obtain preshower and Shower MIP position, convert data to Edep (in MeV)
    • Preshower:
    • Shower:
      • weekly notes say MIP is about FADCint=400, Ye’s slide shows 40 MeV (check: 194*1.5mm*2MeV/cm=58 MeV? Reduction due to Birk effect?). So 1GeV electron → Edep=1GeV*20%sample=200 MeV= 5*40 MeV = 5 * 400 → 2000 in peak integral
      • from Ye’s slide, run 4780 threshold (120mV in TS4) shows up at FADCint=2000; MIP=400 int = 40 MeV Edep = 200 MeV momentum; so 2000 threshold (120mV) → 1 GeV electron; 180 mV → 1.5 GeV/c electron momentum
      • if clear correlation exists between Edep and electron energy, convert FADC peak integral to electron energy: 1 GeV electron energy = 2000 in peak integral
• PID:
  • use TS3 (or TS4?) compare Psh, Sh, and Cherenkov spectra with simulation
  • the question is: if we apply cuts to select electrons at >95% efficiency, what is the pion rejection factor? Cuts include:
    • Cherenkov cut (1D Npe, or 2D Nch vs. Npe)
    • Preshower 1D cut
    • 2D cut on Preshower + Shower (after calibration)
  • Trigger cuts need to be in simulation
  • could use 1 GeV energy (electron-energy equivalent) bin in ShowerSum (better using PSH+SH sum if calibrated), as a rough momentum identification for understanding the signal particle composition. For example, we can select a momentum slice and study Preshower signal and compare with simulation. (Note that SoLID will have momentum info from GEM tracking.)
    • apply Cherenkov cut to see if it makes sense, but I suspect we can’t use Cherenkov to define “clean” electron samples.
    • either with or without Cherenkov cut, apply 1D preshower cut, and 2D preshower+shower (normalized) cut, study electron efficiency and pion rejection. -- need some innovation here