Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 EMCal & PID Rikard Sandström Universite de Geneve MICE collaboration meeting 26/6-05.

Similar presentations


Presentation on theme: "1 EMCal & PID Rikard Sandström Universite de Geneve MICE collaboration meeting 26/6-05."— Presentation transcript:

1 1 EMCal & PID Rikard Sandström Universite de Geneve MICE collaboration meeting 26/6-05

2 2 Outline Introduction Definition of background & good event Setup –Beam, background, detectors. Performance & results Summary

3 3 Introduction Provided a clear definition of background. Using neural network to do PID in G4MICE. Increased difficulty by choosing larger p rms of beam. Now also using tracker, TOF. More aggressive filtering and cuts.

4 4 Definition of good event A good event is a event which leaves a hit in TOF2. T_min = light speed between TOF1 & TOF2 T_max = half light speed - “” -.

5 5 Definition of downstream background A downstream background is an good event where no hit in TOF2 is (in truth) an antimuon. –Hence, empty TOF2 events are not background nor signal -> lost.

6 6 The beam Background from decay of muon beam. Less monochrome beam in p_z than earlier: –10 mm pi rad, 4D emittance –Pz = 196.2 +- 30.7 MeV/c –Px = 0 +- 28.4 MeV/c –Z_start = -6011 mm (after diffuser) –T = 0 +- 70 ps (corresponding to TOF resolution) Causes worse PID than earlier beam. Considerable scraping in cooling channel. Muons are often off-phase in RF system.

7 7 Sources of background 1.Particles get lost in RF system and decay. TOF window. 2.Particles decaying in flight. Tricky… 3.Particles decay during the EMCal gate at rest. Short gate / TDC / don’t stop muons in calorimeter. 4.Particles decay in time window of other event. Needs to be studied. In addition, RF BG will hit TOF2.

8 8 Tracker and TOF Using Gaussian(truth,std) –P_t and P_z resolution from Ellis simulation of SciFi in presence of RF background. 1.75 resp 2.41 MeV/c –TOF resolution 70 ps. –Could be ignoring systematic effects! Forcing TOF window rejects mu decay at rest. Individually, p and tof not good variables for PID. Calculate tofError = tof-[expected tof using,m_mu] –Excellent for PID

9 9 Calorimeter Standard 4 layer KLOE light spaghetti. –Fiber, lead, glue. Amp(t) ~ (t/T)^2 exp(-t/T), T=8 ns best exp fit. Open gate 100 ns. 17 cm/ns transversal delay (along fiber to PMT). Still manually triggered. In future, could make use of expected muon range given p in tracker.

10 10 Two principle ideas of calorimeter Either range based calorimeter… –Given momentum, range of muon is well defined. …or avoiding muon decay in calorimeter –Will cause additional background in 1.own event (100 ns/2mus -> 3.4 % probability) 2.other event (600*0.1/1000 = 6 % probability) Best of both worlds? Have simulated two alternative designs (sandwich & smörgås), no time to analyze results yet.

11 11 Performance 1.Filtering on TOF2. Output-> Good/bad events 2.Using Neural Network to clean up. Good events -> signal/bg 10 pi mm rad gives –16.68% bad events Mostly scraping 7.5% of scraped events leave electrons at tof2. –99.557% input purity Only ~130k Events.

12 12 Inputs (1) Off phase in RF

13 13 Inputs (2) Due to mu decay at rest Sum[sqrt(adcL*adcR)]

14 14 Too few events

15 15 Efficiency Plots show efficiency for muon id and background id. Background ID is poor, caused by too small sample?

16 16 Purity Choosing purity and efficiency means choosing a cut value. Working under assumption target = –efficiency > 99.9 % –purity > 99.8 % –Achieved!

17 17 Performance for subsystems good (No Ckov2)

18 18 Comments on results Lower p_z rms gave better result, but... Target achieved, do we want to do better? –When are we happy? Room for improvements: –Expected muon range in EMCal. –Transverse size in EMCal. –Getting rid of decay at rest in EMCal (geometry or TDC?) Background detection efficiency expected to improve with larger sample. –Positive effect on purity. –Must find memory leak. Ignoring possible correlations in tracker –Ex. Resolution as function of p_t.

19 19 Summary G4MICE + Neural Net shows downstream PID works. –Also with scraping, huge p_rms and emittance. –Ckov2 can give further improvements. More data desired. Two major sources of background not included: –Overlap of muons. –Decay of muons in window of other events. –Need spill based simulation! Invite you to my talk in the software session!


Download ppt "1 EMCal & PID Rikard Sandström Universite de Geneve MICE collaboration meeting 26/6-05."

Similar presentations


Ads by Google