Presentation is loading. Please wait.

Presentation is loading. Please wait.

PID simulations Rikard Sandström University of Geneva MICE collaboration meeting 2005-10-22 RAL.

Similar presentations


Presentation on theme: "PID simulations Rikard Sandström University of Geneva MICE collaboration meeting 2005-10-22 RAL."— Presentation transcript:

1 PID simulations Rikard Sandström University of Geneva MICE collaboration meeting 2005-10-22 RAL

2 Outline Definitions PID objective Calorimeter –Comparison with KLOE data Time of flight as PID variable Performance –Before PID analysis –After PID analysis

3 Definition, signal Signal –An event which is a muon at TOF1 and at TOF2. Background –An event which does not fulfill the Signal requirement.

4 Definition, good event Good event –An event which gives hits in both trackers, TOF1 & TOF2. is within  = 15cm in both trackers. has a time of flight corresponding to average β z within 0.5 and 1. has positive p z in trackers. Bad event –An event which does not fulfill the Good event requirement Good/bad given independently by 1.MC truth 2.Reconstructed tracks Nota bene!: Good event  Signal event

5 PID objective For good events, correctly assign signal/background tag. –Can be expressed in efficiency & purity Assigning signal as background -> Low efficiency Assigning background as signal -> Low purity How: –Find/construct variables which separates signal from background. –Every event is then assigned a “muonness” weight. Now done by fitting the signal variable in a neural net. –The weight is added to input file. Can be used in emittance calculation. Easy to compare alternative methods and cuts.

6 Calorimeter, KLOE in G4MICE To validate EmCal simulations, G4MICE was used to reproduce KLOE situation. –KLOE geometry Cells: 4.4x4.4x400 cm 3 Lead thickness 0.5 mm –KLOE readout Fibers: Kuraray SCSF-81 –Long attenuation –3% light collection efficiency PMT: Hamatsu R5946/01 1.5” –Gain 1M @ 2kV Result: –amplitude/visible energy = 60.09±4.19 adc counts/MeV amplitude  2(adc L adc R )/(adc L +adc R )

7 Calorimeter, data vs G4MICE KLOE dataKLOE in G4MICE 195<|p|<250 MeV/c

8 TOF and PID (MICE note on this topic coming soon.) Idea: 1.Time of flight given by TOF1 and TOF2. 2.Momentum given by trackers. 3.Comparing the two gives estimate of particle mass! Practice: 1.Take momentum from trackers. 2.Assume mass = muon mass. 3.Calculate when the particle is expected to arrive at TOF2. 4.Compare with measured time.

9 Time of flight dt=dz/(β z c) Hence, to first order t.o.f. depends on p z. –A rough estimate is taking only p z measured in trackers, and expected energy loss into account. Second most important effect is momentum transfer induced by magnetic field. Other things –Energy loss fluctuations. –RF phase.

10 Magnetic field & TOF Principle: –Total momentum conserved, longitudinal momentum not conserved. Lorentz force F=qvxB –Longitudinal component F z ~ v x B y -v y B x Field has largest transversal components at field flips –B(z=0)  k , k is a constant. Treat classically -> F z ~ p x y- p y x –-> sin  tan , as beam goes to pencil beam. Result (most difficult case): –t.o.f. = 49.81±1.93 ns predicted to rms 0.28 ns (muons). –I.e. spread reduced to 14.6%. (Upper limit.)

11 MC truth purity Beam: –6 pi mm rad mu+ beam, starting at TOF1. –1ppm of p, K+, pi+, e+ contamination. Starting purity at TOF1 = 99.62%. What happens: –Particles may decay, and another particle might arrive at TOF2. At TOF2: –Proton tracks never give good event. –4% of K+ tracks give good event. –68% of pi+ tracksgive good event. –0.42% of good-event mu+ of tracks has different particle ID at TOF2. Decay! –5.3% of mu+ tracks give bad event. -> Resulting purity at TOF2 = 99.46%.

12 Time of flight cut Time of flight can be predicted to < 300 ps –Worst case beam  280 ps. Using MC truth tracker info, apply 5 ns time of flight discrepancy cut –Efficiency = 99.994%  100% –Purity = 99.68% Background from muon decay reduced by 44%. Positrons (starting at TOF1) reduced by 100%. Pion background reduced by 2%. Kaon background reduced by 40%. Tracker reconstruction gave suspicious values. –I have the program ready to analyze using reconstructed values once all OK.

13 Neural network performance Using neural net to fit is more powerful than square cuts, or multidimensional Gaussian fits. Calorimeter + MC truth tracker & TOF info: –Purity = ?@ 99.9% efficiency. Calorimeter alone: –Purity = ?@ 99.9% efficiency.

14 Summary Background at TOF2 –Heavy particles get lost prior to TOF2. –Pions at TOF1 gives significant background at TOF2. Time of flight is almost powerless. Hard to separate from muons with present calorimeter design. –Positrons at TOF1 very easy to reject with time of flight. –Positrons from muon in-flight decays harder to reject. Time of flight & tracker & calorimeter makes good combination. –Purity before analysis = 99.42% (default beam). –Purity after analysis with neural net = ? @ eff =99.9%.


Download ppt "PID simulations Rikard Sandström University of Geneva MICE collaboration meeting 2005-10-22 RAL."

Similar presentations


Ads by Google