Preliminary Results on n e / n t selection at DUNE FD. CP violation - - PowerPoint PPT Presentation

▶
preliminary results on n e n t selection at dune fd
SMART_READER_LITE
LIVE PREVIEW

Preliminary Results on n e / n t selection at DUNE FD. CP violation - - PowerPoint PPT Presentation

Sep 25, 2022 445 likes •584 views

Preliminary Results on n e / n t selection at DUNE FD. CP violation & n t physics perspectives. Thomas Kosc (kosc@ipnl.in2p3.fr) PhD student at IPNL (France) Supervisor : Dario AUTIERO Thesis : Development of the CP violation search

slide-1
SLIDE 1

1

Thomas Kosc / IPNL (Lyon, France)

Preliminary Results on ne/nt selection at DUNE FD.

CP violation & nt physics perspectives.

Thomas Kosc (kosc@ipnl.in2p3.fr) PhD student at IPNL (France)

Supervisor : Dario AUTIERO

Thesis : Development of the CP violation search analysis in DUNE and assessment of the related systematics.

slide-2
SLIDE 2

2

Thomas Kosc / IPNL (Lyon, France)

  • Some work on tau selection in previous meetings:

Herilala/Miriama thesis. ~80% of ne CC rejection and ~60% of nt CC (t—>e-) kept, based on GENIE (version ?) and the ROOT Toolkit for Multivariate Analysis. nt appereance at NOMAD (R. Petti). Likelihood approach, up to ~1% background contamination ?

  • Elaborate an analysis of nt detection, based on simulations (GENIE v3.0.2). Outlook of CP

violation analysis (nt background rejection) and nt physics (nt selection).

  • ne are restricted to beam contamination at production point (without oscillation). ne from
  • scillation will depend on PMNS parameters, included later.
  • Rely on GENIE only. t decay ? nt cross section ? Comparison with other generators ?

A priori discussion

  • This talk focuses only on CC nt with t—>e-+2n (easy channel and background of ne CC, but only

~17% of total branching ratio). Other channels require further analysis. The nt come from the

  • scillation of nµ.
slide-3
SLIDE 3

3

Thomas Kosc / IPNL (Lyon, France)

Tools & Method

  • DUNE FD setup is implemented only via the `lux. See the `ile I used:

histos_g4lbne_v3r5p4_QGSP_BERT_OptimizedEngineeredNov2017_neutrino_LBNEFD_fastmc.root. Results are

`lux dependent, any comparison should be careful with using the same `lux.

  • Neutrino cross sections, GENIE pre-computed `ile NULL_G1802a00000-k500-e1000 (see “Associated

data release” on http://www.genie-mc.org/).

  • Two `iles generated and used for the time being: oscillated nµ —> nt and unoscillated ne
  • contamination. Both at FD.
  • No reconstruction effects (smearing) implemented yet. Should come in a near future.

5 10 15 20 25 30 Energy (GeV) 0.05 0.1 0.15 0.2 0.25 0.3 0.35

e

ν

τ

ν

FSI_Emeas

nutau_fluxosc Entries 5603837 Mean 2.622 Std Dev 1.237 5 10 15 20 25 30 Energy (GeV) 1 2 3 4 5 12 − 10 × / POT 2 s / GeV / m τ ν Oscillated nutau_fluxosc Entries 5603837 Mean 2.622 Std Dev 1.237 nutau_fluxosc nue_flux Entries 346856 Mean 4.463 Std Dev 4.639 5 10 15 20 25 30 Energy (GeV) e ν 5 10 15 20 25 30 35 40 45 15 − 10 × / POT 2 s / m e ν Unosc nue_flux Entries 346856 Mean 4.463 Std Dev 4.639 nue_flux energy E (GeV) µ ν 1 2 3 4 5 6 7 8 9 )
  • 1
.GeV 2 cm
  • 38
/E (10 σ 0.2 0.4 0.6 0.8 1 1.2

CC Total cross-section on Argon

µ

ν

FLUX at FD ne/nt CROSS SECTION on Ar ne nt

Two `iles generated. Right : distributions

  • f

total kinetic energy in the `inal state interaction (neutrons removed).

  • Easier to start with ne from beam contamination rather than oscillated nµ—>ne
slide-4
SLIDE 4

4

Thomas Kosc / IPNL (Lyon, France)

Kinematics (neutrons removed !)

  • Basic idea: distinguish CC ne from CC nt with t- —> e-+ne+nt using kinematical criteria (NOMAD).
  • Transverse plane kinematics (TPK) (remove uncertainties due to incoming neutrino

momentum).

Jet phad Electron pe- Missing momentum pmiss

fhe fem fhm ne CC in TPK

Jet phad e- pe- Missing momentum pmiss

fhe fem fhm

n

nt CC in TPK with t-—>e-

  • Kinetic variables at play:

Ke- = kinetic energy of the electron. , , = tranverse momenta. Angles between transverse momentum fhe, fhm, fem.

pmiss

(tr)

pe−

(tr)

phad

(tr)

pasym = pe−

(tr) − phad (tr)

pe−

(tr) + phad (tr) Small missing momentum, so fhe close to 180°. Unseen neutrinos increase the missing m o m e n t u m a n d ch a n ge th e a n g l e s distributions.

slide-5
SLIDE 5

20 40 60 80 100 120 140 160 180 angle (deg) 0.05 0.1 0.15 0.2 0.25 0.3 0.35

e

ν

τ

ν

tr_ele-had

1 2 3 4 5 6 7 8 )

2

Momentum (GeV/c 0.05 0.1 0.15 0.2 0.25 0.3 0.35

e

ν

τ

ν

MissingP

5 10 15 20 25 30 Energy (Gev) 0.05 0.1 0.15 0.2 0.25 0.3 0.35

e

ν

τ

ν

ElectronEnergy

Distributions

Ke-

pmiss

(tr)

pe−

(tr)

phad

(tr)

pasym = pe−

(tr) − phad (tr)

pe−

(tr) + phad (tr)

fhe fem fhm

5

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 )

2

Momentum (GeV/c 0.05 0.1 0.15 0.2 0.25 0.3 0.35

e

ν

τ

ν

ElectronTrMomentum

0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 )

2

Momentum (GeV/c 0.05 0.1 0.15 0.2 0.25 0.3 0.35

e

ν

τ

ν

HadronicTrMomentum

20 40 60 80 100 120 140 160 180 angle (deg) 0.05 0.1 0.15 0.2 0.25 0.3 0.35

e

ν

τ

ν

tr_had-miss

20 40 60 80 100 120 140 160 180 angle (deg) 0.05 0.1 0.15 0.2 0.25 0.3 0.35

e

ν

τ

ν

tr_ele-miss

3 − 2 − 1 − 1 2 3 0.05 0.1 0.15 0.2 0.25 0.3 0.35

e

ν

τ

ν

PtrAsymmetric

slide-6
SLIDE 6

1 2 3 4 5 6 7 8 (GeV)

(tr) miss

P 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 (GeV)

(tr) had

P

(tr) miss

vs P

(tr) had

P Entries 924 Mean x 0.6605 Mean y 0.5329 Std Dev x 0.3869 Std Dev y 0.4076

(tr) miss

vs P

(tr) had

P Entries 924 Mean x 0.6605 Mean y 0.5329 Std Dev x 0.3869 Std Dev y 0.4076 (tr) miss

vs P

(tr) had

P

Correlations, two examples.

ne CC nt CC

φhe, pmiss

(tr)

⎡ ⎣ ⎤ ⎦

phad

(tr), pmiss (tr)

⎡ ⎣ ⎤ ⎦

6

1 2 3 4 5 6 7 8 (GeV)

(tr) miss

P 20 40 60 80 100 120 140 160 180 (deg)

he

φ

(tr) miss

vs P

he

φ Entries 924 Mean x 0.6628 Mean y 104 Std Dev x 0.3871 Std Dev y 50.87

(tr) miss

vs P

he

φ Entries 924 Mean x 0.6628 Mean y 104 Std Dev x 0.3871 Std Dev y 50.87

(tr) miss

vs P

he

φ

1 2 3 4 5 6 7 8 (GeV)

(tr) miss

P 20 40 60 80 100 120 140 160 180 (deg)

he

φ

(tr) miss

vs P

he

φ Entries 7593 Mean x 0.432 Mean y 146.1 Std Dev x 0.3404 Std Dev y 38.35

(tr) miss

vs P

he

φ Entries 7593 Mean x 0.432 Mean y 146.1 Std Dev x 0.3404 Std Dev y 38.35 (tr) miss

vs P

he

φ

1 2 3 4 5 6 7 8 (GeV)

(tr) miss

P 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 (GeV)

(tr) had

P

(tr) miss

vs P

(tr) had

P Entries 7593 Mean x 0.4286 Mean y 0.7196 Std Dev x 0.34 Std Dev y 0.4811

(tr) miss

vs P

(tr) had

P Entries 7593 Mean x 0.4286 Mean y 0.7196 Std Dev x 0.34 Std Dev y 0.4811 (tr) miss

vs P

(tr) had

P

slide-7
SLIDE 7

Likelihood analysis

  • Given an event (ne or nt) with a set of kinematic variables, we

compute the likelihood ratio L. LS (resp. LB) is the probability that a given kinematic variable (or a correlation of several of them) occurs for the signal (resp. background).

L = log LS LB ⎛ ⎝ ⎜ ⎞ ⎠ ⎟

Convention: nt CC = SIGNAL; ne CC (beam) = BACKGROUND.

LS for nt LB for ne

  • Comparison of the signal/background likelihood distribution informs about the separability

power of the kinematic variable (or of their correlation).

7

slide-8
SLIDE 8

Some likelihood plots

8

4 − 3 − 2 − 1 − 1 2 3 4 Log(Ls/Lb) 0.05 0.1 0.15 0.2 0.25 0.3

Cut at 0: 0.699 sig kept 0.250 bck cont.

Bck2dLH_Phihe vs Pmiss

4 − 3 − 2 − 1 − 1 2 3 4 Log(Ls/Lb) 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2

Cut at 0: 0.748 sig kept 0.276 bck cont.

WorkSigLHPhihe vs Pmiss x kele

4 − 3 − 2 − 1 − 1 2 3 4 Log(Ls/Lb) 0.05 0.1 0.15 0.2 0.25 0.3

Cut at 0: 0.808 sig kept 0.351 bck cont.

Sig2dLH_Pthad vs Pmiss

4 − 3 − 2 − 1 − 1 2 3 4 Log(Ls/Lb) 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22

Cut at 0: 0.764 sig kept 0.305 bck cont.

WorkSigLHPthad vs Pmiss x Phihe

L φhe, pmiss

(tr)

⎡ ⎣ ⎤ ⎦

( )

L φhe, pmiss

(tr)

⎡ ⎣ ⎤ ⎦.Ke−

( )

L phad

(tr), pmiss (tr)

⎡ ⎣ ⎤ ⎦.φhe

( )

L phad

(tr), pmiss (tr)

⎡ ⎣ ⎤ ⎦

( )

  • Using different combinations
  • f the 8 variables, hard to get

much improvement. Cuts at 0 reach ~[75;80]% of nt selection and some [25;30]%

  • f ne contamination.
  • The cut at 0 is shown as a

matter of indication. We use integral likelihood distributions for more quantitative results.

slide-9
SLIDE 9 1 2 3 4 5 6 7 8 20 40 60 80 100 120 140 160 180 (tr) miss vs P he φ Entries 332 Mean x 0.4321 Mean y 155.5 Std Dev x 0.2404 Std Dev y 16.93 (tr) miss vs P he φ Entries 332 Mean x 0.4321 Mean y 155.5 Std Dev x 0.2404 Std Dev y 16.93 (tr) miss

vs P

he

φ

1 2 3 4 5 6 7 8 20 40 60 80 100 120 140 160 180 (tr) miss vs P he φ Entries 6096 Mean x 0.3232 Mean y 163 Std Dev x 0.2522 Std Dev y 16.49 (tr) miss vs P he φ Entries 6096 Mean x 0.3232 Mean y 163 Std Dev x 0.2522 Std Dev y 16.49 (tr) miss

vs P

he

φ

1 2 3 4 5 6 7 8 20 40 60 80 100 120 140 160 180

Phihe vs Pmiss

A posteriori check : φhe, pmiss

(tr)

⎡ ⎣ ⎤ ⎦

Full distributions

Signal (nt) Background (ne)

Likelihood

Discriminating background region. Discriminating signal region.

1 2 3 4 5 6 7 8 20 40 60 80 100 120 140 160 180

Phihe vs Pmiss

4 − 3 − 2 − 1 − 1 2 3 4 Log(Ls/Lb) 0.05 0.1 0.15 0.2 0.25 0.3

Cut at 0: 0.699 sig kept 0.250 bck cont.

Bck2dLH_Phihe vs Pmiss Remove “blue” here

Apply cuts at +0.2, remove any contribution that’s above. If a signal or a background event gets a likelihood ratio above the arbitrary threshold 0.2, it is removed.

Truncated Distributions CONSISTENT

Background region remaining only

9

slide-10
SLIDE 10 1 2 3 4 5 6 7 8 20 40 60 80 100 120 140 160 180 (tr) miss vs P he φ Entries 766 Mean x 0.7288 Mean y 90.66 Std Dev x 0.3321 Std Dev y 45.81 (tr) miss vs P he φ Entries 766 Mean x 0.7288 Mean y 90.66 Std Dev x 0.3321 Std Dev y 45.81 (tr) miss

vs P

he

φ

1 2 3 4 5 6 7 8 20 40 60 80 100 120 140 160 180 (tr) miss vs P he φ Entries 3130 Mean x 0.6319 Mean y 115.3 Std Dev x 0.3356 Std Dev y 41.33 (tr) miss vs P he φ Entries 3130 Mean x 0.6319 Mean y 115.3 Std Dev x 0.3356 Std Dev y 41.33 (tr) miss

vs P

he

φ

A posteriori check :

Full distributions

φhe, pmiss

(tr)

⎡ ⎣ ⎤ ⎦

Signal (nt) Background (ne)

Likelihood

Discriminating background region. Discriminating signal region.

1 2 3 4 5 6 7 8 20 40 60 80 100 120 140 160 180

Phihe vs Pmiss

1 2 3 4 5 6 7 8 20 40 60 80 100 120 140 160 180

Phihe vs Pmiss

4 − 3 − 2 − 1 − 1 2 3 4 Log(Ls/Lb) 0.05 0.1 0.15 0.2 0.25 0.3

Cut at 0: 0.699 sig kept 0.250 bck cont.

Bck2dLH_Phihe vs Pmiss Remove “red” here

Truncated Distributions CONSISTENT

10

Signal region remaining

  • nly

Apply cuts at -0.2, remove any contribution that’s above. If a signal or a background event gets a likelihood ratio under the arbitrary threshold -0.2, it is removed.

slide-11
SLIDE 11

Outlook/Discussion

11

  • Preliminary results plausible/encouraging. Though hard to get much improvement (looking for

a good combination of variables isn’t obvious). Simple cuts at 0 for likelihood ratio leads to ~75% of signal (oscillated nµ—>nt) selection and ~30% of background (ne beam) contamination.

  • Add analysis with neutrons and see the difference as a crosscheck.
  • Reproduce analysis with oscillated ne.
  • Include a reconstruction (smearing at fewer extent), detector response and geometry.
  • Include a PMNS parametrisation of the analysis (i.e work on the `lux part).

Apply the ef`iciencies obtained to ~234 ne CC beam bck and ~37 nt CC for 3.5 years staged.

  • Results not so good… Why ? Distinguish QEL/RES/DIS as a `irst clue.

DUNE TDR, Vol 2, p91

slide-12
SLIDE 12

nutau_fluxosc Entries 5603837 Mean 2.622 Std Dev 1.237 5 10 15 20 25 30 Energy (GeV) 1 2 3 4 5

12 −

10 × / POT

2

s / GeV / m

τ

ν Oscillated nutau_fluxosc Entries 5603837 Mean 2.622 Std Dev 1.237

nutau_fluxosc

nue_flux Entries 346856 Mean 4.463 Std Dev 4.639 5 10 15 20 25 30 Energy (GeV)

e

ν 5 10 15 20 25 30 35 40 45

15 −

10 × / POT

2

s / m

e

ν Unosc nue_flux Entries 346856 Mean 4.463 Std Dev 4.639

nue_flux

12

Back up: FD \lux

ne from beam contamination. nt from nµ

  • scillation