Neutron discrimination using waveform information. Yasuyuki - - PowerPoint PPT Presentation

Neutron discrimination using waveform information.

Yasuyuki Sugiyama (Yamanaka Taku Lab.)

Year-end annual report. 2015 Dec. 25(Fri)

1

J-PARC KOTO Exp. O

K T

ν ν s d

KOTODetector

• search for 


(B.R.=2.4x10-11@S.M.)
 @J-PARC

• Strategy
• Detect by

measuring 2γ on CsI Calorimeter.

• Require Pt π0>0 


with no other particle.

• First physics run in 


2013 May.

• More physics run in 2015

Spring, Autumn.

FB NCC MB CV CsI calorimeter CC03 OEV CC04 CC05 CC06 BHCV LCV BCV HINEMOS

KL

π0→2γ ν ν

vertexZ [mm]

• reconstructed

1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 [MeV/c]

t

P

• reconstructed

50 100 150 200 250 300 350 400 450 500

:Signal
 Box

2

calorimeter measures the the K0

L → π0νν deca

• f 576 pure CsI crystals.

measures π0ν pure CsI

KOTO CsI Calorimeter

3

• measure γ to reconstruct from KL decay.
• Use 2716 Un-doped CsI crystals from KTeV
• Waveform is recorded by 125MHz ADC with Bessel Filter

measures π0ν pure CsI

(27X0)

2.5cm□ 5cm□

500mm

つくりなお

Time[Clock=8ns] 5 10 15 20 25 30 35 40 45 50 55 60 65 Pulse Height [ADC counts] 500 550 600 650 700 750 800 850 900 950

waveform recorded by ADC KOTO CsI Calorimeter CsI Crystal

Neutron Background

• Single neutron (1~2GeV) hits on CsI Calorimeter may cause Background events.
• considered to be dominant in the first physics run.
• Need a new method to distinguish neutrons from photons.

4

CsI CsI

Opened the signal box

31 Rec. z [mm]

2000 3000 4000 5000 6000 [MeV/c]

t

P 50 100 150 200 250 300 350 400 450 500 0.5 1 1.5 2 2.5 3 3.5 4

0.36± 0.16 1

• Nobs=1, Nexp=0.36±0.16

Observed Expected

年 月 日金曜日

neutron neutron

from K.Shiomi’s talk in JPS 2014 Autumn

Analysis Result for 2013 Physics Run

How to reject Neutron Background?

• Current Cut Methods use the transverse shower shape.
• Longitudinal characteristics or Timing structure of Hadron Shower

may be new point of view to reject the background.

• These characteristics may affect pulse shape.

CsI Calorimeter CsI Calorimeter

Gamma Neutron Smaller Size


• f Cluster

Shower only in Upstream Shower exists 
 also in Downstream

Larger Size


• f Cluster

X Y X Y Z

5

6

Fitting waveform with Asymmetric gaussian.

• Pulse shape of CsI calorimeter has Gaussian-shape due to Bessel Filter

in ADC module.

• Fit waveform with “asymmetric Gaussian” to get information about the

change of pulse shape.

• Larger tail component will make “a” bigger.

Time [Clock = 8ns] 5 10 15 20 25 30 35 40 45 50 55 60 65 Pulse Height [ADC counts] 500 550 600 650 700 750 800 850 900 950

Pulse fitted by Asymmetric Gaussian

A(t) = |A| exp ✓ −(t − t0)2 2σ(t)2 ◆ , σ(t) = σ0 + a(t − t0)

Asymmetric Gaussian:

Procedure

• Check the difference of fit parameters from Gamma’s waveform
• Make Template of waveform parameters for Gamma events
• Compare acquired fit parameters (pfit) with Template(pexp(H)).
• Calculate χ2 value for γ and π0

7

χ2

γ

NDF = 1 2NCrystal

crystal

X

i

X

p=σ0,a

pfit

i

− pexp

i

(Hi) σpi(Hi) !2 χ2

γ

Hi : Pulseheight of i th crystal Precision template fit
 result

χ2

π0

NDF = 1 2 X

γi

χ2

γi

NDFγi

Fit parameter Template

• Make templates of fit parameter from KL->3π0 events taken in

2013 Physics Run.

• Get Profile (Mean values) pexp(H) and Its RMS σp(H).
• Use TSpline for interpolation between bins.

8

Log10(Pulse Height)[ADC count] 1.5 2 2.5 3 3.5 4 Asymmetric Parameter

• 0.1
• 0.05

0.05 0.1 0.15 0.2 0.25 0.3

50 100 150 200 250 300 350

h_asym_height_0

pexp(H) σp(H)

χ2/NDF for each γ,π

0 in KL->3π 0 events

• Calculate χ2/NDF for each γ clusters on the CsI Calorimeter and π0.
• γ’s χ2/NDF distributes around 1 and has small Energy dependence.
• π0’s χ2/NDF distributes around 1.

9

Gamma Energy [MeV] 500 1000 1500 2000 2500 3000

• /NDF for

2

• 0.5

1 1.5 2 2.5 3 3.5 4 4.5 5

1 10

10

10

10

asym chi2 for gamma vs GammaE

small E dependence

h1 Entries 631302 Mean 0.9621 RMS 0.3378

• 2/NDF for
• 0.5

1 1.5 2 2.5 3 3.5 4 4.5 5 10000 20000 30000 40000 50000

h1 Entries 631302 Mean 0.9621 RMS 0.3378

Pi0_AsymChi2 {Iteration$<3}

X:χ2/NDF of π0 Cut Threshold 5%
 loss

Analysis for Neutron-rich data

• Data taken with Inserting Aluminum target during 2015 April/May Run
• Only neutrons can arrive at Calorimeter from Aluminum target.
• Analyze events with 2 clusters on the CsI Calorimeter
• Apply all cuts used in 1st Physics Run (2013) Analysis


except for Transverse Cluster Shape Cut.

10

reconstructed Z vertex[mm]

• 3000

3500 4000 4500 5000 5500 6000 Pt[MeV/c]

• 50

100 150 200 250 300 350 400 450 500

h_PtZ_cut Entries 1915 Mean x 4810 Mean y 221.9 RMS x 438.2 RMS y 64.09

1 2 3 4 5 6 7

h_PtZ_cut Entries 1915 Mean x 4810 Mean y 221.9 RMS x 438.2 RMS y 64.09

All cut except for 
 Transverse Pulse Shape Cut

FB NCC MB CV CsI calorimeter CC03 OEV CC LCV BCV HINEMOS

t=10mm
 Al Target

χ2/NDF of Rec.π

0 with gamma template

• neutron-like events in Al target data have big χ2/NDF.
• KL->3π0 data distribute around 1.
• if we reject events with χ2/NDF>1.5, 


75% of neutron-like events can be rejected with 5% loss for π0->2γ

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 20 40 60 80 100 120 140 160 180

Chi2/NDF for Pi0 with gamma template

h_chi2_pi_gamma

Entries 5523 Mean 1.974 RMS 0.897

Chi2/NDF for Pi0 with gamma template

Al target with
 all cut except
 for ClusterShape

11

75% χ2/NDF

χ2/NDF(γ) X:Energy[MeV]

Energy[MeV]

• 500

1000 1500 2000 2500 3000

• 2/NDF for
• 0.5

1 1.5 2 2.5 3 3.5 4 4.5 5

Entries 3830 Mean x 511.3 Mean y 1.822 RMS x 268.6 RMS y 1.009

2 4 6 8 10 12 14

Entries 3830 Mean x 511.3 Mean y 1.822 RMS x 268.6 RMS y 1.009

12

Template of neutron-like events

• Make template of fitting parameters using Al target run data
• nly with Veto Cut.
• Use only a half of data to make template


and use another half of data to estimate performance.

• Calculate Chi2/NDF as same as the case with 6gamma

template.

• Calculate Likelihood for 6gamma template and Z0Al

template and determine which template is more closer to data.

Ln =

γ

• i=0,1

ch.

• j

par.

• k=0,1

Gaussian(sobs.

ijk |sexp. ijkn, σk(eij)) ,

Likelihood ratio

signal → (from upstream) ← background (from downstream)

MC

signal backward π0

Pulse shape Difference in each crystal

50 100 150 200 250 300 350 400 450 500

• 3
• 2
• 1

1 2 3 4

h1

Entries 653839 Mean x 32.14 Mean y -0.009284 RMS x 60.63 RMS y 0.6891 1 10

10

10

10

h1

Entries 653839 Mean x 32.14 Mean y -0.009284 RMS x 60.63 RMS y 0.6891

GamClusCsi_AsymChi2_Al-GamClusCsi_AsymChi2:GamClusCsiE {GamClusCsiE>3}

Energy in CsI crystal[MeV]

χ2/NDF(Al target template)


• χ2/NDF(3pi0 template)
• Template difference become larger


in Pulse Height >~500count (~50MeV).

• χ2/NDF difference in KL->3pi0 data 


becomes larger for E deposit > 50MeV/crystal.

Template for CsI Ch.800

log10(PulseHeight)

Al target
 3pi0

asym.par

13

Chi2/NDF for limited number of channel

• Calculate chi2/NDF and Likelihood ratio only with information from

channels with Edep >50MeV.

• If #(CsI with Edep>50MeV) is 0, use Chi2 and Likelihood of the CsI

Crystal with Maximum Edep.

χ2

γ

NDF = 1 2NCrystal

crystal

X

i

X

p=σ0,a

pfit

i

− pexp

i

(Hi) σpi(Hi) !2 χ2

γ

Hi : Pulseheight of i th crystal Precision template fit
 result

χ2

π0

NDF = 1 2 X

γi

χ2

γi

NDFγi

Ln =

γ

• i=0,1

ch.

• j

par.

• k=0,1

Gaussian(sobs.

ijk |sexp. ijkn, σk(eij)) ,

Likelihood ratio

signal → (from upstream) ← background (from downstream)

MC

signal backward π0

14

χ2/NDF distribution with 50MeV threshold/Crystal

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 10 20 30 40 50 60 70 80

Chi2/NDF for Pi0 with gamma template

h_chi2_pi_gamma

Entries 5523 Mean 2.49 RMS 1.217

Chi2/NDF for Pi0 with gamma template

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 50 100 150 200 250

Chi2/NDF for Pi0 with Z0Al template h_chi2_pi_al

Entries 5523 Mean 0.9017 RMS 0.5917

Chi2/NDF for Pi0 with Z0Al template

Al target data KL->3π0 data

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 100 200 300 400 500

10 ×

Chi2/NDF for Pi0 with gamma template

h_chi2_pi_gamma

Entries 1.158191e+07 Mean 0.8097 RMS 0.5906

Chi2/NDF for Pi0 with gamma template

3π0
 template

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 50 100 150 200 250 300 350

10 ×

Chi2/NDF for Pi0 with Z0Al template h_chi2_pi_al

Entries 1.158191e+07 Mean 1.287 RMS 0.8414

Chi2/NDF for Pi0 with Z0Al template

Al target
 template broad narrow narrow broad

15

h_LH_ratio_pi_gamma

Entries 830676 Mean 0.7045 RMS 0.3202

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

10

10

h_LH_ratio_pi_gamma

Entries 830676 Mean 0.7045 RMS 0.3202

LHratio for Pi0 with gamma template

Likelihood ratio for Rec.π0

• Likelihood ratio=Lγ/(Lγ+Lneutron)
• If we cut events with Likelihood ratio <0.1
• 7% loss for π0->2γ
• 92% reduction for 2 cluster in Neutron-like events


with Veto and all cut except for cluster shape cut

7%

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 10

10

10

LHratio for Pi0 with gamma template

h_LH_ratio_pi_gamma

Entries 5523 Mean 0.03468 RMS 0.1275

LHratio for Pi0 with gamma template

92%

Al target data
 (not used to 
 make template)

π0->2γ in KL->3π0

16

Likelihood ratio Likelihood ratio

→ π0->2γ n-like ← → π0->2γ n-like ←

Summary

• Developed new method to discriminate the incident particle between

γ and neutron using pulse shape information.

• The discrepancy between the typical pulse shape of γ and that of

neutron seems to be significant if E deposit >50~100MeV/ crystal

• By comparing fit result with the typical pulse shape of γ and neutron,


more than 90% of neutron-like events can be rejected.

17