MCC11 Smearing Matrices and Energy Resolution Erin Conley January - - PowerPoint PPT Presentation

MCC11 Smearing Matrices and Energy Resolution

Erin Conley January 23, 2019 SNB/LE Working Group Meeting

2

Outline

• Introduction
• MCC11 smearing matrices:

– Using truth deposited energy – Using different charge quantification – Applying drift correction

• Fractional energy resolution
• Takeaways

3 1/23/19

Introduction

• Hack Days 2018: produced the

right-hand smearing matrix for MCC11 MARLEY with strange behavior

• Many issue(s):

– Comparing to MCC10 shows biggest differences in low-energy bins – Charge distributions show double bumps, shoulders, negative charge – Determined that different files contain different charge distributions

• See SNB meeting from November

14 for more information

4 1/23/19

Different Charge Quantification

• Calculated reconstructed

energy for smearing matrix using reco hits (Hit::Integral)

• Smoothed, columns

normalized

• Filled in missing energy level

by averaging neighboring columns

• Less wiggly!

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

MARLEY Smearing Matrix (No Drift Correction)

10 20 30 40 50 60 70 80 90 100

Truth Neutrino Energy (MeV)

10 20 30 40 50 60 70 80 90 100

Reconstructed Energy (MeV) MARLEY Smearing Matrix (No Drift Correction)

1/23/19 5

Applying Drift Correction

No Drift Correction Reco Drift Correction

Less spread when drift correction is applied!

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

MARLEY Smearing Matrix (No Drift Correction)

10 20 30 40 50 60 70 80 90 100

Truth Neutrino Energy (MeV)

10 20 30 40 50 60 70 80 90 100

Reconstructed Energy (MeV) MARLEY Smearing Matrix (No Drift Correction)

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

MARLEY Smearing Matrix, Drift Corrected (Reco)

10 20 30 40 50 60 70 80 90 100

Truth Neutrino Energy (MeV)

10 20 30 40 50 60 70 80 90 100

Reconstructed Energy (MeV) MARLEY Smearing Matrix, Drift Corrected (Reco)

6

Fractional Energy Resolution

• Considered five different smearing matrices with

different “reconstructed energy” quantifications (see backup)

• For a given !true:

– Distribution of !reco values – Calculated the fractional difference, !reco − !true /!true – Found standard deviation, RMS of fractional difference values – Plotted vs neutrino energy

7

Fractional Energy Resolution Plot

Notes:

• Truth kinetic energy at

zero due to cut-off energy in MARLEY modeling; also see backup

• Truth information has

better resolution than reco

• Our resolution is worst for

reco charge, but improves with drift correction

8

Fractional Energy Resolution Plot

Notes:

• Same overall trends as

standard deviation plot

• Truth information has

better resolution than reco

• Our resolution is worst for

reco charge, but improves with drift correction

• See backup for more

information

9

Takeaways

• We now have MCC11 smearing matrices!

– Alternative charge quantification with reco Hits resolves the wiggly behavior seen using RawDigit objects

• Applying drift correction reduces the spread in the

smearing matrix

• RMS and standard deviation of fractional energy

quantifies the resolution/spread in the smearing matrices

Backup Slides

10

11 1/23/19

Toy Smearing Matrix (MARLEY + %)

• Used truth deposited energy

from the LArSoft simulation for the reconstructed energy

• Smeared with different

Gaussian resolutions using SNOwGLoBES (0-20%)

• Filled in missing energy level

by averaging neighboring columns

• Use these matrices to

reproduce plots in spectral parameter fitting

1/23/19 12

Applying Drift Correction

No Drift Correction Truth Drift Correction

Less spread when drift correction is applied!

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

MARLEY Smearing Matrix (No Drift Correction)

10 20 30 40 50 60 70 80 90 100

Truth Neutrino Energy (MeV)

10 20 30 40 50 60 70 80 90 100

Reconstructed Energy (MeV) MARLEY Smearing Matrix (No Drift Correction)

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2

MARLEY Smearing Matrix, Drift Corrected (Truth)

10 20 30 40 50 60 70 80 90 100

Truth Neutrino Energy (MeV)

10 20 30 40 50 60 70 80 90 100

Reconstructed Energy (MeV) MARLEY Smearing Matrix, Drift Corrected (Truth)

13 1/23/19

Reco Drift Correction

• Applied drift correction:

! = !# exp '0 )*

– !: Truth charge – !#: Observed charge – )*: Electron lifetime – '0: Reco interaction start time

• Find '0 using photon flash, reco

hit information (used longest track as reco electron track)

• Effect of correction: narrower reco

energy distribution

14

Comparing Event Rates: MCC10 vs. MCC11

Notes:

• Used GVKM flux
• Cutoff at 5 MeV due

to post-smearing efficiency model

• We see energy loss

between MCC11,

• ther two smearing

matrices

• See backup for more

event rates!

15

Comparing Event Rates: MCC11

Notes:

• Used GVKM flux
• Cutoff at 5 MeV due

to post-smearing efficiency model

• Applying drift

correction reduces energy loss in MCC11

• See backup for more

event rates!

16

Standard Deviation Plot

• For a given !true:

– Distribution of !reco values – Calculated the fractional difference for each !reco,i value using Efrac,i = !reco,i − !true /!true for event i

• Standard deviation:

0frac = ∑234

5

!frac,i − 6 !frac

7

8 − 1

– 8: number of events for a given !true – 6 !frac: mean of all Efrac,i values

• Plotted standard deviation vs neutrino energy

17

RMS Plot

• For a given !true:

– Distribution of !reco values – Calculated the fractional difference for each !reco,i value using Efrac,i = !reco,i − !true /!true for event i

• RMS:

Erms = 1 3 4

567 8

Efrac,i

9

– 3: number of events for a given !true

• Plotted RMS vs neutrino energy

18

How the two quantities are related

• (From Wikipedia)
• For !

"frac as the mean and 'frac as the standard deviation: "rms

*

= ! "frac

* + 'frac *

• If !

"frac = 0, then standard deviation and RMS are ~same

• If !

"frac ≠ 0, then we don’t expect the values to be the same

• I believe the truth kinetic energy curve for RMS is nonzero below 10

MeV due to relationship between 'frac and Erms

19

Example fractional energy distributions

Notes:

• The reason why 6.25 and

12.25 MeV peaks are lower than the others is due to less simulated events for those particular energy levels

• We see a nonzero mean

for all the energy levels

• All events for !true <

10 MeV are contained in a singular peak; we expect %frac = 0 for those energy levels

• Energy levels above 10

MeV are not a singular peak (see red arrows)

SMEARING MATRICES

Backup Slides

20

21

22

23

24

25