update on energy reconstruction in dp · introduction 2 andrea scarpelli for the tdr we would like...

Update on energy reconstruction in DP

Andrea Scarpelli (APC - Université Paris Diderot)

[email protected]

LBPWG meeting 09 April 2018

mailto:[email protected]

Introduction 2

Andrea Scarpelli

For the TDR we would like to have a complete reco chain for the DP

detector and we would like to be able to produce some spectra

• Testing, tuning and improving the reconstruction machinery already in

place

• Energy reconstruction for DP and study of DP specific impact on the

detector resolution

Today:

‣ Introduction on Dual Phase

‣ Issues with the reconstruction

‣ νμ CC energy reconstruction

Far detector geometry 3

Andrea Scarpelli

X (Drift)

Y

Z

+ 300 cm

-300 cm

-800 cm

+600 cm (Liquid)

-10cm

+ 1300 cm

+700 cm (Gas)

From the Jan. 2018 collab. meeting: https://indico.fnal.gov/event/14581/session/6/contribution/172/material/slides/0.pdf

3 x 3 m2 TPC 1 cm gap between TPC

Anode 50x50 cm2 module

Y pi

tch

Z pitch

Use of workspace geometry to have a similar fid. Volume to SP: dunedphase10kt_workspace4x2_geo

https://indico.fnal.gov/event/14581/session/6/contribution/172/material/slides/0.pdf

Dual-Phase specificities 4

Andrea Scarpelli

Drift length (fiducial volume) 12 m Electron lifetime and diffusion, reduce image quality

Drift Field 0.5 kV/cm

Wire pitch 3 mm Higher granularity

Pich direction Z and Y Topologically problematic for the far detector neutrino events

# of Planes 2 Ambiguities in reconstruction unresolved

Plane type Collection Larger image quality (no double pulse)

Gain Minimal CDR requirement is 20. Is supposed to compensate for all the flawns from the longer drift

Liquid-Gas interface / Effect of electron extraction on the signal waveform -> Detector resolution

Space charge / In liquid and at the liquid gas interface

• Because of gain, one might expect different performances between DP and SP in the energy

reconstruction: → Impact of gain on the energy resolution

• Smaller pitch might improve patter recognition

• Hard event topology and absence of a third plance could make reconstruction harder → Find an

estimate of that and solution to improve it

• Sensitivities for DUNE → DP dedicated studies have never been completed (as far as I know)

FD Simulation 5

Andrea Scarpelli

From the Jan. 2018 collab. meeting: https://indico.fnal.gov/event/14581/session/6/contribution/172/material/slides/0.pdf

https://indico.fnal.gov/event/14581/session/6/contribution/172/material/slides/0.pdf

Moving forward: reconstruction 6

Andrea Scarpelli

Standard approach for reconstruction so far:

✓Hit finder: raw hit finder fitting

✓Clustering: linecluster

✓Tracking: pmalg

Not yet in place, but planned:

‣ Clustering using Trajcluster

‣ CNN hit feature labelling

‣ Shower reconstruction: emshower

A long shot:

• Pandora …

νμ energy reconstruction 7

Andrea Scarpelli

Following the approach of the SP by N. Grant and T. Yang ‣ See the internal note https://docs.dunescience.org/cgi-bin/private/RetrieveFile?docid=2278&filename=NuEnergyReco_v2.pdf&version=2

Event Topology Reco feature Observable

νμ CC μ contained Longest track contained Momentum from track length

νμ CC μ not contained Longest track not contained Momentum from MCS

νe CC all Em shower with largest charge Cal. energy from charge

Hadronic all Charge collected Cal. energy from charge

Efforts to improve νμ CC events reconstruction:

• Correct longest track identification and reconstruction

https://docs.dunescience.org/cgi-bin/private/RetrieveFile?docid=2278&filename=NuEnergyReco_v2.pdf&version=2

Troublesome reconstruction 8

Andrea Scarpelli

Main problem in νμ CC events leading muon track reconstruction:

νμ

μ

p

νμ

μ

p

νμ

μ

p

νμμ

p

νμμ

p

νμμ

p

Case: Not contained

Case: Contained

Truth Reco

Truth Reco

Truth Reco

Truth Reco

Truth Reco

Truth Reco

Ok

Ok

No

No No

No

Use completeness and truth info to identify correctly reconstructed tracks and have an

idea of the reconstruction efficiency

Reconstruction efficiencies (preliminary) 9

Andrea Scarpelli

1− 0.5− 0 0.5 1xθLepton Cos

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

lineclusterStandard

lineclusterStandard

lineclusterTuned

1− 0.5− 0 0.5 1yθLepton Cos

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

lineclusterStandard

lineclusterStandard

lineclusterTuned

600− 400− 200− 0 200 400 600Lepton drift position (cm)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

lineclusterStandard

lineclusterStandard

lineclusterTuned

0 1 2 3 4 5 6 7 8 9Lepton Momentum (GeV)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

lineclusterStandard

lineclusterStandard

lineclusterTuned

1− 0.5− 0 0.5 1zθLepton Cos

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

lineclusterStandard

lineclusterStandard

lineclusterTuned

Possible efficiency degradation related to the direction due to momentum-direction correlation

Direction cos. along x (drift)

Direction cos. along y (drift)

Direction cos. along z (drift)

Lepton momentum Drift

2− 1.5− 1− 0.5− 0 0.5 1 1.5 2MCS momentum residuals

0

50

100

150

200

250

300

350

400

MCS

10

Andrea Scarpelli

Longest track momentum

2− 1.5− 1− 0.5− 0 0.5 1 1.5 2Momentum fractional residuals (Length)

0

100

200

300

400

500

600

700

Length

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 momentum (GeV)µTrue

100

200

300

400

500

600

Leng

th (c

m)

/ ndf 2χ 0.03178 / 3Prob 0.9985p0 69.26±23.72 − p1 66.59± 421

/ ndf 2χ 0.03178 / 3Prob 0.9985p0 69.26±23.72 − p1 66.59± 421

Length

0.5 1 1.5 2 2.5 3 momentum (GeV)µTrue

0.8

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

MC

S m

omen

tum

(GeV

)

/ ndf 2χ 0.002624 / 3Prob 1p0 0.2663± 0.1693 p1 0.1412± 0.8916

/ ndf 2χ 0.002624 / 3Prob 1p0 0.2663± 0.1693 p1 0.1412± 0.8916

MCS

‣ Contained tracks

‣ Exiting tracks

Bias: 0.00Resolution: 0.05


11

Andrea Scarpelli

Hadronic energySimple hit energy sum and correction for the invisible energy

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6True hadro. energy (GeV)

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9R

econ

stru

cted

had

ro. e

nerg

y (G

eV)

/ ndf 2χ 0.1344 / 3Prob 0.9874p0 0.09292± 0.03215 p1 0.098± 0.6168

/ ndf 2χ 0.1344 / 3Prob 0.9874p0 0.09292± 0.03215 p1 0.098± 0.6168

Hadro

2− 1.5− 1− 0.5− 0 0.5 1 1.5 2Fractional residuals

0

1

2

3

4

5

6

7

8

9

10

ene

rgy

(GeV

)µν

True

hResHadroNuEnergyEntries 8144Mean x 0.005313− Mean y 2.976Std Dev x 0.4562Std Dev y 1.495

0

2

4

6

8

10

12

14

16

18

20hResHadroNuEnergyEntries 8144Mean x 0.005313− Mean y 2.976Std Dev x 0.4562Std Dev y 1.495

Hadro2− 1.5− 1− 0.5− 0 0.5 1 1.5 2

Fractional residuals (Hadronic energy)0

50

100

150

200

250

300

Hadro


‣ Hadronic energy

• Assume linear dependence on the true

energy, while it is not.

• Ad hoc tweak to correct the resolution

and bias energy dependence..

12

Andrea Scarpelli

νμ energy (contained)

hNuResEntries 2498Mean 0.0626− Std Dev 0.1663

/ ndf 2χ 57.62 / 31Prob 0.002548Constant 6.4± 246.3 Mean 0.00331±0.06168 − Sigma 0.003± 0.158


0

50

100

150

200

250hNuRes

Entries 2498Mean 0.0626− Std Dev 0.1663

/ ndf 2χ 57.62 / 31Prob 0.002548Constant 6.4± 246.3 Mean 0.00331±0.06168 − Sigma 0.003± 0.158

Contained

0 1 2 3 4 5 6 7 8 true energyµν

0.3−

0.2−

0.1−

0

0.1

0.2

0.3


0

1

2

3

4

5

6

7

8

9

10

Ener

gy

hNuResEnergyEntries 2498Mean x 0.06304− Mean y 2.31Std Dev x 0.1662Std Dev y 1.229

0

2

4

6

8

10

12

14

16hNuResEnergyEntries 2498Mean x 0.06304− Mean y 2.31Std Dev x 0.1662Std Dev y 1.229

Contained

0 1 2 3 4 5 6 7 8 9 10 true energy (GeV)µν

0

1

2

3

4

5

6

7

8

9

10

reco

ene

rgy

µν

hNuScattEntries 2498Mean x 2.168Mean y 2.301Std Dev x 1.215Std Dev y 1.205

0

5

10

15

20

25

30

35


Contained

Bias: -0.06Resolution: 0.16

• Bias • Resolution

13

Andrea Scarpelli

νμ energy (not-contained)

hNuResEntries 5646Mean 0.003117Std Dev 0.2583

/ ndf 2χ 201.8 / 58Prob 18− 8.759eConstant 7.3± 389.1 Mean 0.003043±0.002369 − Sigma 0.003± 0.223


0

50

100

150

200

250

300

350

400

450 hNuResEntries 5646Mean 0.003117Std Dev 0.2583

/ ndf 2χ 201.8 / 58Prob 18− 8.759eConstant 7.3± 389.1 Mean 0.003043±0.002369 − Sigma 0.003± 0.223

Not Contained

0 1 2 3 4 5 6 7 8 true energyµν

0.3−

0.2−

0.1−

0

0.1

0.2

0.3


0

1

2

3

4

5

6

7

8

9

10

Ener

gy

hNuResEnergyEntries 5646Mean x 0.008981Mean y 3.285Std Dev x 0.2572Std Dev y 1.499

0

2

4

6

8

10

12

14

16

18

20

22

24

hNuResEnergyEntries 5646Mean x 0.008981Mean y 3.285Std Dev x 0.2572Std Dev y 1.499

Not Contained

0 1 2 3 4 5 6 7 8 9 10 true energy (GeV)µν

0

1

2

3

4

5

6

7

8

9

10

reco

ene

rgy

µν


0

2

4

6

8

10

12

14

16

18

20


Not Contained

Bias: -0.02Resolution: 0.22

• Bias • Resolution

14

Andrea Scarpelli

Energy reco performances

Event Topology Resolution SP* Resolution DP (at gain = 20)

νμ CC μ containedTrack length: 5%

Hadronic part: 39%Overall: 18%

Track length: 5%Hadronic part: Overall: 16%

νμ CCμ not

containedTrack MCS: 23%

Hadronic part: 39%Overall: 20%

Track MCS: 25%Hadronic part: 41%

Overall: 22%

✴ See the internal note https://docs.dunescience.org/cgi-bin/private/RetrieveFile?docid=2278&filename=NuEnergyReco_v2.pdf&version=2

• At gain 20 and selecting well-reconstructed tracks , SP and DP perform alike (as

predicted from CDR minimal requirement )

• A Gain increase might impact the hadronic component, yielding a better

estimate of the overall energy

https://docs.dunescience.org/cgi-bin/private/RetrieveFile?docid=2278&filename=NuEnergyReco_v2.pdf&version=2

Conclusion 15

Andrea Scarpelli

First look on reconstruction highlight major problems

• Better tuning of the existing algorithms

• Implement CNN to solve possible ambiguities

First look on energy

• Energy calibration and reconstruction for νμ CC events was done selecting

nicely reconstructed events

Questions, tips or feedbacks?

Outlook

• Improvement of the reconstruction

• Using truth info try to evaluate the impact of gain, purity and electric field on

the energy resolution (and event selection (?) )

• Export the results for physics studies

Thank you!

LBPWG Hack days 20-22 March 2018

Topology of the events 17

Andrea Scarpelli LBPWG Hack days 20/03/18

1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 1xθLepton Cos

0

100

200

300

400

500

Not contained

Contained

1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 1yθLepton Cos

0

50

100

150

200

250

300

350

400

450Not contained

Contained

1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 1zθLepton Cos

0

500

1000

1500

2000

2500

3000

3500

4000

4500

Not contained

Contained

0 1 2 3 4 5 6 7 8 9 10Lepton true momentum (GeV)

0

100

200

300

400

500

600

700

800

900Not contained

Contained

CRP wires in DP are oriented towards difficult directions in z

Considered 34421 numu events: 11125 contained, 23296 not contained

Direction cos. along x (drift)

Direction cos. along y

Direction cos. along z

True lepton momentum

Topology of the events - 2D 18


1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 1xθCos

1−

0.8−

0.6−

0.4−

0.2−

0

0.2

0.4

0.6

0.8

1yθC

os

0

2

4

6

8

10

12

14

16

18

20

22

1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 1xθCos

1−

0.8−

0.6−

0.4−

0.2−

0

0.2

0.4

0.6

0.8

1zθC

os0

50

100

150

200

250

1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 1yθCos

1−

0.8−

0.6−

0.4−

0.2−

0

0.2

0.4

0.6

0.8

1zθC

os

0

20

40

60

80

100

120

140

160

180

200

220

1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 1xθ

0

1

2

3

4

5

6

7

8

9

10

True

mom

entu

m (G

eV)

0

2

4

6

8

10

12

14

16

18

20

22

1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 1yθ

0

1

2

3

4

5

6

7

8

9

10

True

mom

entu

m (G

eV)

0

5

10

15

20

25

1− 0.8− 0.6− 0.4− 0.2− 0 0.2 0.4 0.6 0.8 1zθ

0

1

2

3

4

5

6

7

8

9

10

True

mom

entu

m (G

eV)

0

20

40

60

80

100

120

140

Observe correlations between directions and momentum

Issues with the FD Reconstruction - evd 1 19


Kink of the cluster In the vertex region cause ambiguous

track matching

Wrong matching of the leading track

Ambiguities are produced by long clusters kinks in the vertex region, causing ambiguous track matching within the TPC.

Track interpolation to match the other plane

Issues with the FD Reconstruction - evd 2 20


These should be two different clusters

Muon clusterProton cluster

Proton and muon clusters are reconstructed together

Rather common issue: two almost collinear clusters should be separated from their different charge, but they are not.

update on energy reconstruction in dp · introduction 2 andrea scarpelli for the tdr we would like...

Documents