particle identification in the na48 experiment using neural networks
DESCRIPTION
L. Litov University of Sofia. Particle Identification in the NA48 Experiment Using Neural Networks. Introduction. NA 48 detector is designed for measurement of the CP-violation parameters in the K 0 – decays –successfully carried out. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/1.jpg)
Particle Identification in the NA48 Experiment Using
Neural Networks
L. Litov
University of Sofia
![Page 2: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/2.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Introduction
NA 48 detector is designed for measurement of the CP-violation parameters in the K0 – decays –successfully carried out.
Investigation of rare K0 s and neutral Hyperons decays – 2002
Search for CP-violation and measurement of the parameters of rare
charged Kaon decays – 2003
A clear particle identification is required in order to suppress the background In K – decays – and e Identification of muons do not cause any problems
We need as good as possible e/- separation
![Page 3: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/3.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
NA48 detector
![Page 4: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/4.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Introduction
2003 Program for a precision measurement of Charged Kaon Decays Parameters
Direct CP – violation in , Ke4 - Scattering lengths
Radiative decays
K 00K)( eK
20
00 ,aa
0,, KKK
![Page 5: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/5.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Introduction
![Page 6: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/6.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
E /p
K3 Background
The standart way to separate e andis to use E/p
energy deposited by particle in the EM calorimeter
p – particle momentumcut at E/p > 0.9 for ecut at E/p< 0.8 for
signal
K3 background
control region signal
Introduction
![Page 7: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/7.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Sensitive Variables
Difference in development of e.m. and hadron showers Lateral development EM calorimeter gives information for lateral development From Liquid Kripton Calorimeter (LKr)
E/p Emax/Eall, RMSX, RMSY Distance between the track entry point and the associated shower Effective radius of the shower
![Page 8: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/8.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Sensitive variables - E/p
MC simulationA correct simulation of the energy deposed by pions in the EM calorimeter - problem for big E/pIt is better to use experimental events
E/p distribution
![Page 9: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/9.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Sensitive variables - RMS
RMS of the electromagnetic cluster
![Page 10: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/10.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Distance
Distance between track entry point and center of the EM cluster
![Page 11: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/11.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Sensitive variables - Emax/Eall, Reff
![Page 12: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/12.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
MC
To test different possibilities we have used:Simulated Ke3 decays – 1.3 MSimulated single e and π – 800 K π and 200 K e
Using different cuts we have obtainedRelatively to E/p < 0.9 cut Keeping > 95 %
Using Neural Network it is possible to reach e/π separation:Relatively to E/p < 0.9 cut Keeping > 98%
The background from ~ 0.1%
2107.15 eeff
eeff
K
eeff
2100.2 eeff
![Page 13: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/13.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Experimental data
E/pi separation – to teach and test the performance of NN
We have used experimental data from two different runsCharged kaon test run # 1 2001
electrons frompions from
run 2001electrons frompions from
eeK 0
K
00 K
40eK eK 0
![Page 14: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/14.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Charged run
Pions
Track momentum > 3 GeV
Very tight selection
Track is chosen randomly
Requirement – E/p < 0.8 for the other two tracks
K
K
![Page 15: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/15.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
eeK 0
Electron selection
![Page 16: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/16.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Charged run
E/p and momentum distributions
![Page 17: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/17.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Charged run NN output
Out 0 for If out > cut – eIf out < cut -
NN output
![Page 18: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/18.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Charged run NN performance
Net: 10-30-20-2-1Input: E/p, Dist, Rrms, p, RMSx, RMSy, dx/dz, dy/dz, DistX, DistYTeaching: 10000 π - , 5000 e -
K eeK 0
![Page 19: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/19.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
)( eK
3effM
3effM
E/p
E/p
E/p > 0.9
Non symmetric E/p distribution
E/p > 0.9
out > 0.9
Symmetric E/p distribution
![Page 20: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/20.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
E/p
E/p
)( eK
out > 0.9
E/p distribution
out > 0.8E/p distributionThere is no significant change in the parameters
![Page 21: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/21.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
)( eK
There is a good agreement between MC and Experimental distributions
3effM
MCEXP
![Page 22: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/22.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
reconstruction with NN
Decay
Significant background comes from when one π is misidentified as an e
Teaching sample:Pions - from , 800 K events Electrons - from , 22 K events
00 eK
00 K
eK 0
00 K
00 eK
![Page 23: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/23.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
reconstruction with NN 00 eK
NN output E/p distribution
![Page 24: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/24.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
reconstruction with NN 00 eK
rejection factore identification efficiency
![Page 25: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/25.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Ke4 run NN performance
Net: 10-30-20-2-1Input: E/p, Dist, Rrms, p, RMSx, RMSy, dx/dz, dy/dz, DistX, DistYTeaching: 10000 π - , 5000 e - eK 000 K
![Page 26: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/26.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
reconstruction with NN 00 eK
222 )5.2
498()
7
6( 3
MeV
MeVM
MeV
MeVpR Kt
Ke3
Ke4
1/REllipse
m / GeV
222 )5.2
498()
7
6( 3
MeV
MeVM
MeV
MeVpR Kt
![Page 27: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/27.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
recognition with NN 00 eK
NN output versus 1/R
NN output versus 1/R
•the background from K3is clearly separated
![Page 28: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/28.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
00 eK
e/ Neural Networke/ Neural Network
Performance
• no bkg subtraction! • using nnout > 0.9 cut• works visibly very well• but what about bkg?
![Page 29: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/29.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
reconstruction with NN 00 eK
e/ Neural Networke/ Neural NetworkBackground
• extending range of 1/R to 5• obviously there is bkg!
Ke4 MC
E /p
1/R
1/R
without NN
with NN
E /p E /p
![Page 30: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/30.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
00 eK reconstruction with NN
![Page 31: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/31.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Performance
• background is fitted both with and without NN • ratio R (rejection factor) is measure of performance
00 eK
e/ Neural Networke/ Neural Network
reconstruction with NN
![Page 32: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/32.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Optimization
• goal: optimize the cut values for nnout and 1/R
1/R
nnout
nnout
NN rejection factor background
Signal NN efficiency 1/R
00 eK
e/ Neural Networke/ Neural Network
reconstruction with NN
![Page 33: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/33.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Optimization
e/ Neural Networke/ Neural Network
• value to minimize: combined statistical and systematical error• statistical error goes with N-½
• systematical error grows with background
sigbkg
N1 c
1/R
nnout
statistical limitssystematical limits
00 eK reconstruction with NN
![Page 34: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/34.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
00 eK
e/ Neural Networke/ Neural Network
Performance
• background can be reduced at level 0.3 %•Ke4 reconstruction efficiency at level 95%
reconstruction with NN
![Page 35: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/35.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Conclusions – e/pi separation
e/π separation with NN has been tested on experimental dataFor charged K run we have obtained:
Relatively to E/p < 0.9 cut At 96%A correct Ke4 analysis can be done without additional detector (TRD) Background can be reduced at the level of ~ 1% ~ 5 % of the Ke4 events are lost due to NN efficiency
For Ke4 run we have obtained:Rejection factor ~ 38 on experimental dataBackground ~ 0.3% at 95%
~eff2104.3~ e
eff
~eff
![Page 36: Particle Identification in the NA48 Experiment Using Neural Networks](https://reader035.vdocuments.mx/reader035/viewer/2022081501/568143d7550346895db0657f/html5/thumbnails/36.jpg)
L. Litov Particle Identification in the NA48 Experiment Using Neural Networks ACAT’ 2002
Conclusions – NN analysis
Additionally Neural Network for Ke4 recognition has been developedThe combined output of the two NN is used for selection of Ke4 decaysNN approach leads to significant enrichment of the Ke4 statistics ~2 times
This work was done in collaboration with
C. Cheshkov, G. Marel, S. Stoynev and L. Widhalm