e identification in the NO A Near Detector events Ciro Riccio - PowerPoint PPT Presentation

ν e identification in the NO ν A Near Detector events Ciro Riccio Supervisors: Xuebing Bu and Pat Lukens September 25 th , 2014 1 Thursday, September 25, 14

The NO ν A experiment • NOvA NuMI Off-Axis ν e Appearance is optimized for the detection of ν μ →ν e and ν μ →ν e oscillations • NOvA includes: 810 km Medium Energy Tune • Main Injector now @ 360 kW used to produce on-axis 80 the beam 7 mrad off-axis 14 mrad off-axis ν CC events / kt / 1E21 POT / 0.2 GeV 21 mrad off-axis • A 14 kt “totally active” tracking liquid scintillator 60 calorimeter sited 14.6 mrad off the NuMI beam axis at a distance of 810 km (Far Detector, FD) 40 • A 300 ton Near Detector (ND) identical to the far detector sited 14.6 mrad off the NuMI beam 20 axis at a distance of 1 km and 105 m μ underground. It is used to study the background 0 compositions and contributions for oscillation 0 2 4 6 8 10 E ν (GeV) analysis 60 m 16 m 4 cm 16 m 6 cm 15 m 4 m 4 m 2 Thursday, September 25, 14

APDs Quality Assurance Test Visual test Pressure and flow test Electrical test 3 Thursday, September 25, 14

ν e identification in the ND In order to identify ν e events I used Boosted Decision Trees (BDT): • BDT is a classifier implemented in TMVA; • The BDT was trained and tested using well known signal and background samples; • The BDT was applied to 1779 MC files for a total of 8.9 x 10 19 POT to identify ν e events in ND 4 Thursday, September 25, 14

List of variables used to train and test BDT and for PID • Σ E cells is the summed energy of all cells associated to the slice with the maximum number of associated cells; • N cells is the number of cells associated to the slice with the maximum number of cells; • L track is the lenght of the track; • The ratio of number of cells associated to the longest track over N cells ; • Number of MIP cells (Nmip defined requiring 100 < PECorr < 245, PECorr is corrected photo-electrons); • The ratio N cells over N mip ; • Fraction of energy in first 20 planes; • Maximal fraction of energy in 2 planes. Reflects the condensity of the longitudinal shower; • Maximal fraction of energy in 6 planes; • Fraction of energy in 2 σ ( σ = 2 cm) road. The ν e should have relatively narrower transverse shower than the π 0 ; • Fraction of energy out 3 σ road; • Number of 2D prongs; • Number of 3D prongs; • Energy balance between 2 most energetic 2D prongs; • Energy balance between 2 most energetic 3D prongs. 5 Thursday, September 25, 14

Input Variables 6 Thursday, September 25, 14

7 Thursday, September 25, 14

8 Thursday, September 25, 14

9 Thursday, September 25, 14

TMVA Output Background rejection versus Signal efficiency Overtraining check plot 10 Thursday, September 25, 14

Correlation Matrices for signal and background Correlation Matrix (signal) Linear correlation coefficients in % 100 energy balance for 2D prongs 16 33 36 24 -33 -1 -20 -41 -34 36 -28 -18 47 -26100 Correlation Matrix (background) 80 # of 2D prongs 10 6 -15 9 -8 -3 -13 5 -5 100-26 energy balance for 3D prongs -5 26 36 50 -45 -29 -18 -38 -33 71 -55 -52100 -5 47 60 Linear correlation coefficients in % # of 3D prongs 29 3 -16 -50 22 40 8 16 13 -56 36 100-52 5 -18 100 energy balance for 2D prongs 11 14 23 16 9 10 -16 -28 -28 17 -10 -3 25 -32100 40 Eiso of 3 sigma 22 -6 -18 -43 31 39 17 18 19 -74100 36 -55 -28 80 # of 2D prongs 26 25 -10 -41-34 19 -13 -29 7 16 -11100-32 Efrac of 2 sigma -33 4 22 62 -40 -53 -15 -23 -22100-74 -56 71 -13 36 20 energy balance for 3D prongs -17 -9 18 44 20 -20-14 -16 -23 58 -38 -46100 -11 25 Efrac of 6 slides -46 -50 -65 -15 21 -33 56 89 100-22 19 13 -33 -3 -34 60 # of 3D prongs 55 50 -9 -65 -41 49 -5 -10 -1 -60 38 100-46 16 -3 0 Efrac of 2 slides -50 -55 -61 -11 31 -32 45 100 89 -23 18 16 -38 -41 Eiso of 3 sigma 40 38 27 -2 -41 -16 40 10 -8 1 -69100 38 -38 7 -10 Efrac of 20 -33 -41 -48 -8 14 -24100 45 56 -15 17 8 -18 -8 -20 -20 Efrac of 2 sigma -60 -51 8 76 45 -55 -7 7 -6 100-69 -60 58 -29 17 # of mip cells 85 50 39 -54 32 100-24 -32 -33 -53 39 40 -29 9 -1 20 Efrac of 6 slides -40 -32 -49 -2 -18 -46 52 86100 -6 1 -1 -23 -28 -40 mip fraction -16 -47 -19 -11100 32 14 31 21 -40 31 22 -45 -33 0 -47 -40 -41 10 -5 -49 41100 86 7 -8 -10 -16 -28 Efrac of 2 slides ratio of Ncells -49 -25 26100 -11 -54 -8 -11 -15 62 -43 -50 50 -15 24 -60 -32 -31-36 5 -6 -36100 41 52 -7 10 -5 -14 -13 -16 Efrac of 20 Ltrack 53 55 100 26 -19 39 -48 -61 -65 22 -18 -16 36 36 -20 # of mip cells 91 78 44 -52 -9 100-36 -49 -46-55 40 49 -20 19 10 -80 Ereco 82 100 55 -25 -47 50 -41 -55 -50 4 -6 3 26 6 33 -40 mip fraction -43 -52 37 69 100 -9 -6 -5 -18 45 -16 -41 20 -34 9 100 82 53 -49 -16 85 -33 -50 -46 -33 22 29 -5 10 16 Ncells -100 -66 -63 33100 69 -52 5 10 -2 76 -41-65 44 -41 16 ratio of Ncells mip fraction # of mip cells Efrac of 2 sigma Eiso of 3 sigma # of 3D prongs energy balance for 3D prongs # of 2D prongs energy balance for 2D pro Ncells Ereco Ltrack ratio of Ncells Efrac of 20 Efrac of 2 slides Efrac of 6 slides -60 Ltrack 28 20100 33 37 44 -36 -41-49 8 -2 -9 18 -10 23 -80 Ereco 94100 20 -63 -52 78 -31 -40 -32 -51 27 50 -9 25 14 100 94 28 -66 -43 91 -32 -47 -40 -60 38 55 -17 26 11 Ncells -100 m # E E # e # e N E L r E E E a f i n n c r t r t i o f f f r s o o e e i p f r r r a f e f e c a o a a a o r r l c f m c c c c 3 g 2 g l o r o s k o a o D y D y i o o o f f c p f N f f f 3 p b p b t i c 2 2 6 2 r a r a o s c e 0 s o l o l e n l s s i n a n a l l l i g n n l s i i g m g g l s d d m c c s s e e a e e a s s f f o o r r 3 2 D D p r o n g s Some variables are correlated Some variable are correlated 11 Thursday, September 25, 14

BDT Output 12 Thursday, September 25, 14

Significance Vs BDT Output S √ S + B S √ B Requiring BDT Output larger than 0 and 11 variables = 26 % Requiring BDT Output larger than 0 and 11 variables S √ S + B S = 39 % √ B 13 Thursday, September 25, 14

11 Variables 14 Thursday, September 25, 14

Correlation Matrix 11 Variables Reducing the number of correlated variables we can reduce sources of systematic errors 15 Thursday, September 25, 14

Conclusions • BDT was been trained, tested and then it are applied to MC files using 15 variables; • The number of variables are reduced; • and are evaluated varying the BDT S S √ S + B √ B Output between -1 and 1; • Requiring BDT output > 0 and using 11 variable = 26 % S = 39 % S √ √ S + B B 16 Thursday, September 25, 14

Thank you for your attention! 17 Thursday, September 25, 14