MiniBooNE Event Reconstruction and Particle Identification

Hai-Jun Yang

University of Michigan, Ann Arbor

(for the MiniBooNE Collaboration)

DNP06, Nashville, TN

October 25-28, 2006

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 2

Outline

• Physics Motivation

• MiniBooNE Event Types

• Event Reconstruction

• Particle Identification

• Summary

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 3

Physics Motivation LSND observed a positive signal, but not confirmed. The MiniBooNE is designed to confirm or refute LSND oscillation result at m2 ~ 1.0 eV2 .

PL m

Ee( ) s in ( ) s in (.

) ( . . . )% 2 22

21 2 7

0 2 6 4 0 0 6 7 0 0 4 5

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 4

MiniBooNE Flux8 GeV protons on Be target gives:

p + Be + , K+ , K0

from:

+ + K+ + K0 - +

Intrinsice from:

+ e+ e K+ 0 e+ e K0 - e+ e

L

L

L

The intrinsic e is ~0.5% of the neutrino Flux, it’s one of major backgrounds for e search.

L(m), E(MeV), m2(eV2)

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 5

Event Topology

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 6

Event Reconstruction• To reconstruct event position, direction, time, energy

and invariant mass etc.• Cerenkov light – prompt, directional• Scintillation light – delayed, isotropic• Using time likelihood and charge likelihood method to

determine the optimal event parameters.• Two parallel reconstruction packages

– S-Fitter is based on a simple, point-like light source model; – P-Fitter differs from S-Fitter by using more 0th

approximation tries, adding e/ tracks with longitudinally varying light source term, wavelength-dependent light propagation and detection, non-point-like PMTs and photon scattering, fluorescence and reflection.

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 7

Reconstruction Performance

Michel Electron

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 8

Particle Identification

Two complementary and parallel methods:

• Log-likelihood technique: – simple to understand, widely used in HEP data

analysis but less sensitive

• Boosted Decision Trees: – Non-linear combination of input variables– Great performance for large number of input

variables (about two hundred variables)– Powerful and stable by combining many

decision trees to make a “majority vote”

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 9

Boosted Decision Trees

How to build a decision tree ?For each node, try to find the best variable and splitting point which gives the best separation based on Gini index.Gini_node = Weight_total*P*(1-P), P is weighted purityCriterion = Gini_father – Gini_left_son – Gini_right_sonVariable is selected as splitter by maximizing the criterion.

How to boost the decision trees?Weights of misclassified events in current tree are increased, thenext tree is built using the same events but with new weights,Typically, one may build few hundred to thousand trees.

How to calculate the event score ?For a given event, if it lands on the signal leaf in one tree, it is

given a score of 1, otherwise, -1. The sum (probably weighted) of scores from all trees is the final score of the event.

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 10

Performance vs Number of Trees

Boosted decision trees focus on the misclassified events which usually have high weights after hundreds of tree iterations. An individual tree has a very weak discriminating power; the weighted misclassified event rate errm is about 0.4-0.45.

The advantage of using boosted decision trees is that it combines many decision trees, “weak” classifiers, to make a powerful classifier. The performance of boosted decision trees is stable after a few hundred tree iterations.

Ref1: H.J.Yang, B.P. Roe, J. Zhu, “Studies of Boosted Decision Trees for MiniBooNE Particle Identification”, Ref1: H.J.Yang, B.P. Roe, J. Zhu, “Studies of Boosted Decision Trees for MiniBooNE Particle Identification”, Physics/0508045, Nucl. Instum. & Meth. A 555(2005) 370-385.Physics/0508045, Nucl. Instum. & Meth. A 555(2005) 370-385.Ref2: B.P. Roe, H.J. Yang, J. Zhu, Y. Liu, I. Stancu, G. McGregor, ”Boosted decision trees as an alternative to Ref2: B.P. Roe, H.J. Yang, J. Zhu, Y. Liu, I. Stancu, G. McGregor, ”Boosted decision trees as an alternative to artificial neural networks for particle identification”, physics/0408124, NIMA 543 (2005) 577-584.artificial neural networks for particle identification”, physics/0408124, NIMA 543 (2005) 577-584.

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 11

Output of Boosted Decision Trees

Osc e CCQE vs All Background MC vs Data

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 12

Summary• MiniBooNE Event Reconstruction

– Position resolution ~ 23 cm– Direction resolution ~ 6o

– Energy resolution ~ 15%– Reconstructed 0 mass resolution ~ 20 MeV/c2

• MiniBooNE Particle Identification– For 0.1% eff., ~ 90% electron eff.– For 1% 0 eff., ~ 70% electron eff.– For 0.5% all background eff., ~ 80% electron eff.

• MiniBooNE Results are coming soon …

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 13

Backup Slides

10/26/2006 DNP06, H.J.Yang, MiniBooNE PID 14

Light Model

θc η

Directional Cherenkov light ρ

Isotropic Scintilation light φ

Point-like light source model

(x y z t)

(ux uy uz)

ri

• Predicted charge

• Cerenkov light - directional

• Scintillation light - isotopic

iCER

ii CER

i

F E fr

r

(co s , ) (co s )

ex p ( / )2

iSC I

ii sc i

i

fr

r

(co s )

ex p ( / )2

i iC ER

iSC I

1. Cerenkov angular distribution - F(cos)

2. PMT angular response – f(cos)3. Cerenkov attenuation length –

cer4. Scintillation attenuation length

- sci5. Relative quantum efficiency - i

6. Cerenkov light strength - 7. Scintillation light strength -

f(cosη

)

cosη

t t tr

ccorri

ii

n

( ) 0

T tE E

t t Ecer corr

c ccorr c( )

( , )ex p

( , )[ ( , )]

1

2

1

2 2 02

T tE

E

E

t t E

E

ErfcE

E

t t E

E

sc i corrs

s

s

corr s

s

s

s

corr s

s

( )( , )

ex p( , )

( , )

( , )

( , )

( , )

( , )

( , )

( , )

1

2 2

2 2

2

20

0

1. Corrected time

2. Cerenkov light tcorr(i) distribution

3. Scintillation light tcorr(i) distribution

4. Input: Cerenkov light – t0cer

,σcer

Scintillation light – t0sci

,σsci,τsci

5. Total negative log time likelihood

L t T t T tcorri c

c scer corr

ic

s

c ssc i corr

is( ) lo g ( ( , ) ( , ))( ) ( ) ( )

Light Model

exp