-----------------------0 .- EN P ERGY I ~i~c~f :: ! Fermi lab - - PowerPoint PPT Presentation

SLIDE 1

Neutrino Experiment Simulation Overview

Michael Kirby, Fermilab/Scientific Computing Division Mar 20, 2019 Thomas Jefferson National Accelerator Facility

FERMILAB-SLIDES-19-008-CD This manuscript has been authored by Fermi Research Alliance, LLC under Contract No. DE-AC02-07CH11359 with the U.S. Department of Energy, Office of Science, Office of High Energy Physics

• ----------------------0

Fermi lab

:: !

.- ENP

ERGY I ~i~c~f

SLIDE 2

Outline

• utlook on precision measurements in neutrino oscillations
• where simulations come into the picture
• simulation of neutrino beam fluxes and systematics
• neutrino interaction event generators and cross sections
• detector simulation with GEANT4
• slight diversion about other IF experiments at Fermilab

2

Big thanks to Laura Fields, Alex Himmel, Mary Bishai, Tao Lin, Rob Kutschke, Krzysztof Genser, Jen Raaf, Gabe Perdue, Renee Fatemi, and Leah Welty-Reiger for help with this talk. They deserve the

• credit. All mistakes are mine.

Sanford Unde~~round Research Fac1hty Fermilab

C= Fermilab

SLIDE 3

Neutrino Simulations in the era of oscillations

• Many neutrino experiments have measurement of neutrino oscillation parameters as their

primary goal

3

Neutrino Flux Interaction Cross Section Efficiency / Smearing Function Oscillation Probability

140 35 DUNE v. appearance DUNE v. appearance 3.5 years (staged) 3.5 years (staged) 120 Normal MH, 6cP=0 30 Normal MH, 6cP=0 100

• Signal (v0+v0 ) CC

25

• Signal (v0+v0 ) CC

>

• Beam (v0+v0 ) CC

>

• Beam (v0+v0 ) CC

Q) Q)

C,

• NC

C,

• NC

LO

80

• (v,+v,) CC

LO

20

• (v,+v,)CC

"! "!

• (vµ+vµ)CC
• (vµ+vµ)CC
• en

en

15

c c

Q) Q)

> >

w w

2 3 7 8 2 3

Reconstructed Enerqy (GeV) Reconstructed Energy (GeV)

• ------------------------------ CFermilab

SLIDE 4

Neutrino Simulations in the era of oscillations

• Mass ordering, CP-violation, mixing matrix unitarity
• precision measurements of oscillation


parameters requires accurate simulation


• f detector response and efficiency
• interaction cross sections & detector uncertainties


have significant impact on the potential reach


• f oscillation experiments
• beam fluxes dominant uncertainty for measurements

• f interaction cross sections and event yields

4

Neutrino Flux Interaction Cross Section Efficiency / Smearing Function Oscillation Probability

K2K @ Neutrino 2002 Koichiro Nishikawa

Laura Fields

C ~ 9

Normalized by area

8 7 6 5 4 3 2 O O 0 .5 1 1.5 2 2.5 3 3 .5 4 4 .5 5 E

• ------------------------------ CFermilab

SLIDE 5

Neutrino Simulations in the era of oscillations

• Mass ordering, CP-violation, mixing matrix unitarity
• precision measurements of oscillation


parameters requires accurate simulation


• f detector response and efficiency
• interaction cross sections & detector uncertainties


have significant impact on the potential reach


• f oscillation experiments
• beam fluxes dominant uncertainty for measurements

• f interaction cross sections and event yields

5

Neutrino Flux Interaction Cross Section Efficiency / Smearing Function Oscillation Probability

Laura Fields

II tl 2 3 DUNE vµ disappearance 3.5 years (staged)

• Signal v" CC
• NC
• (v,+v,)CC
• Bkgd vµ CC

4 5 6 7

Reconstructed Energy (GeV)

CP Violation Sensitivity DUNE Sensitivity

14 Normal Ordering

sin2201 3 = 0.085 ± 0.003

12 sin2023 ;;; O

.441 ± O .O42

• ◊c

p=

• r1

2

• 50% of S

C P values

• 75% of OcP values

. ..... 5% $ 1%

• Nominal: 5% EB 2%

"'"'"'"' 5°/4 $ 3%

200 400 600 800 1000 1200 1400

Exposure (kt-MW-years) 8

CFermilab

SLIDE 6

Beam simulations and neutrino flux

• simulate proton beam (8-120 GeV) incident on targets (thin, thick, C, Be)
• T2K Beam Simulation: FLUKA 2011.2c -> GEANT3+GCALOR
• MINERvA Beam Simulation: GEANT4 for production and simulation

– developed the ppfx package for hadron production uncertainties

• Booster Neutrino Beam: GEANT4 based tools developed by MiniBooNE


used by MicroBooNE, SBND, ICARUS

• near detectors help minimize uncertainty for oscillation measurements - not a silver bullet

6

• Hadron production:

based up production models from hadrons exiting the targets and from secondary and tertiary interactions in the horn, decay pipe, etc

• horn focusing: particle

propagation through magnetic fields and alignment of the horn elements

• ther subdominant

effects: gas in the decay pipe, absorber material, decay pipe windows

NuMI Beam

Target Hall Target Muon Monitors

r ___

• _eca

_~_

P_

ipe

___

~ bsorber

l l l

• -\ ___ _
• -- - - - - -

►

u

,,

Protons from

• - - -+

Main Injector

µ• uµ

Hom 1 Hom 2

• - --- +

10m 30m 675 m

u

,,

S m Hadron Monitor ii-;;

18m 210m

Rock

CFermilab

SLIDE 7

CADMesh utilized by the Muon g-2 Experiment

• Translates CAD files into GDML for

simulation in GEANT

• allows for precise shape and location of

detector components without recreation in GDML by hand

• does require greater precision than

engineers are sometimes focused on

• gaps in volumes and overlapping volumes

can be serious problems in GEANT

7

Leah Welty-Reiger, Renee Fatemi

100 ~

Q)

10 E

F

1

• CPU
• ClockTime

0.1 1 10 100 1000 10000 Number of Events

____::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::i C Ferm

ilab

SLIDE 8

CADMesh utilized by the Muon g-2 Experiment

• Translates CAD files into GDML for

simulation in GEANT

• allows for precise shape and location of

detector components without recreation in GDML by hand

• does require greater precision than

engineers are sometimes focused on

• gaps in volumes and overlapping volumes

can be serious problems in GEANT

8

Leah Welty-Reiger, Renee Fatemi

100 ~

Q)

10 E

F

1

• CPU
• ClockTime

0.1 1 10 100 1000 10000 Number of Events

____::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::i C Ferm

ilab

SLIDE 9

T2K Beam simulation

• hadron interactions are the greatest source of uncertainty for the flux prediction
• constrain the pion production using NA61/SHINE, HARP datasets
• neutrino flux uncertainty dominates the measurement of CC0π production
• this is the golden channel for measuring oscillation parameters - cleanest incident neutrino

energy determination

9 ND280: Neutrino Mode, vµ

• Hadron Interactions
• - Material Mcxleling

0.3 -- ProtonBeamProfile&Off-axis Angle -- NumberofProtons

• - Hom Current & Field
• - Horn & Target Alignment

Cl>xEv, Arb. Norm. 0.2 0.1

0.0

• - Thin Tuning Total

v-mode

10

Ev (GeV)

• T. Vlaclisavl1ev1c

Nulnt2018

....

t=10-1 Q)

"iii

5 10-2

n

~ U..10-3

10-'

10-" 0.94 < cos0 < 0.98 0.98 < cos0 < 1.0

• Data Statistical Uncertainty
• V
• Flux+Background+Pion FSI

1 ~ ~

• Detector

II,,

• Signal Modeling
• ProtonFSI
• Mass

1()-' L... ............

._

................................. ..... 10 10-" 10-' 10

pµ

[GeV/c]

true

pµ

[GeV/c]

true

CFermilab

SLIDE 10

MINERvA Beam Simulation

• Hadron production model variation dominant effect for

determining the uncertainty – use data from NA49, MIPP to tune the flux – scale 158 GeV proton data to 120 GeV using FLUKA – incorporated into Package to Predict the Flux (PPFX)

• still dominate uncertainty in most of the MINERvA

measurements

• pioneering measurements to constrain the flux using

neutrino scattering off electrons measurements - but suffers from limited statistics

• cross sections measurements important inputs for

improving neutrino interaction models and predictions

10

Deepika Jena, NuINT 2018

0.16 - J: I ~

P ~ h~ ys ~- ~ R ~ e ~ v~ . D ~ 9

=:

4

:::

,

:::

0= 9= 20 =0

=.5 ===

{2 = = 1

=~=~-

~

• :t:

he~

r

~ab~•-~-7

• meson inc.
• target abs.

0.14 1/)

~0.12

C "iii "li3 0.1

u

§0.00

• pC

➔ r X

• • •
• pC

➔ KX

• nC

➔ n X

• •. nucleon-A
• •• • pC ➔

nucleonX -

• thers
• total

HP

........ :

• ••••••••• ·---·· •• ·-••••• _, •••

1

• • ·•
• .....
• ~r--

··-. ......................................

,. :,. ■-■- 2 4 6 . 8 .. ,-

;~-

:'

."";"~

• ----i:----~
• ~
• - 18

20

Neutrino Energy (GeV) V + A ➔

µ· + 7t+ + A

x10·39 µ 2

• Total Sys. Error
• Detector Model

=

~n~:;;:ti~:s::~:~

=

::~:band Model

• Tracking Eff
• Vertex ~

En ~•=

• ~==j

5 10 15 20

Neutrino Energy (GeV)

CFermilab

SLIDE 11

Neutrino Interactions Simulation

11

(GeV)

ν

E

• 1

10 1 10

2

10 / GeV)

2

cm

• 38

(10

ν

cross section / E ν 0.2 0.4 0.6 0.8 1 1.2 1.4

• A. Schukraft, G. Zeller

TOTAL QE DIS RES

W" n" p+" μ'" νμ" Quasi'elas0c" (QE)" W" n" π+" μ'" νμ" Resonance" (RES)" Δ+" n" W" n" X" μ'" νμ" Deep"inelas0c" (DIS)" Lots%of%interes+ng%(nuclear)%physics%over%all%energy%ranges.%

Many%open%ques+ons% need%experimental%&% theore+cal%input!%

T2K NOvA DUNE MicroBooNE

SLIDE 12

Neutrino Interactions Simulation

• Determination of the incident neutrino is based upon interpretation through nuclear model of

reconstructed final-state objects - but there is no complete theory from first principles that describes the interaction and subsequent states

• the weak interactions at same scale as the


final state strong interactions

• various nuclear models
• potential nucleon correlations
• final-state interactions
• GENIE, NEUT vs NuWRO, GiBUU
• Tuning models to data samples is a common


goal for many experiments but limited ability
 to turn the knobs within generators and models

• understanding systematic uncertainty from the model is a critical component to using


these models and need tools to determine those uncertainties

12 Gollapini, arXiv:1602.05299

https://hepsoftwarefoundation.org/cwp/hsf-cwp-017-CWP_neutrino_event_generators.pdf

Charge Exchange Elastic

µ

V

Pion Production

• ------------------------------ CFermilab

SLIDE 13

Neutrino Interactions Simulation

• Determination of the incident neutrino is based upon interpretation through nuclear model of

reconstructed final-state objects - but there is no complete theory from first principles that describes the interaction and subsequent states

• the weak interactions at same scale as the


final state strong interactions

• various nuclear models
• potential nucleon correlations
• final-state interactions
• GENIE, NEUT vs NuWRO, GiBUU
• Tuning models to data samples is a common


goal for many experiments but limited ability
 to turn the knobs within generators and models

• understanding systematic uncertainty from the model is a critical component to using


these models and need tools to determine those uncertainties

13

https://hepsoftwarefoundation.org/cwp/hsf-cwp-017-CWP_neutrino_event_generators.pdf

20 − 20

)

• 3

10 × (

23

θ

2

Uncertainty in sin

Statistical Uncertainty Systematic Uncertainty Beam Flux Near-Far Differences Detector Response Normalization Muon Energy Scale Neutrino Cross Sections Detector Calibration Neutron Uncertainty NOvA Preliminary NOvA Preliminary 120

> 100

Q) (!) I!) 80 N

ci

60 (/)

c

Q)

>

40 UJ 20 00

NOvA Preliminary

NOvA 6.05x1020 POT-equiv.

• - Best fit prediction
• --- ·· · Unoscillated prediction
• +- Data

1 2 3 4

Reconstructed Neutrino Energy (GeV)

CFermilab

SLIDE 14

Tuning model to match interactions like this and determine neutrino energy

14

KE proton cand #1 = 154.6 MeV (559 MeV/c) KE proton cand #2 = 88.9 MeV (417 MeV/c) KE proton cand #3 = 123.4 MeV (497 MeV/c) KE proton cand #4 = 172.9 MeV (595 MeV/c)

p1 p2 p3 p4 µ

• ,/7

: .

.

• ··

....

• ·-:..

..

·~

.....

_ _ _

......

: : '·

• .. _

"

·.

....._

·.•

• ;; ---

··-=-

BNB DATA RUN 5211 EVENT 1225. FEBRUARY 29, 2016

• ----------------------------------- C Fermilab

SLIDE 15

Neutrino Detector Simulations

• Vast array of detector designs, materials, and energy scales
• Liquid Argon Time Projection Chambers, Liquid Scintillator,


Water Cherenkov, Lead, Steel, Scintillator

• Based mostly upon GEANT4 simulation with various different

physics lists used for both tuning and systematic uncertainties

• detector response needed to determine response and

efficiency along with uncertainties

• modeling particle response important:

– non-static detector simulation – primary electron vs primary photon separation – scintillation photon propagation – neutron propagation and response

15

ProtoDUNE Super-Kami

• ka n

de

• ------------------------------ CFermilab

SLIDE 16

NOvA Oscillation Measurements

• GEANT simulation of efficiency and energy smearing


important to estimation of energy reconstruction

16

20 − 20

)

• 3

10 × (

23

θ

2

Uncertainty in sin

Statistical Uncertainty Systematic Uncertainty Beam Flux Near-Far Differences Detector Response Normalization Muon Energy Scale Neutrino Cross Sections Detector Calibration Neutron Uncertainty NOvA Preliminary NOvA Preliminary

Alex Himmel, FNAL JETP

; .:;! H I J

w

3

CD

• CD

.....

(J) .::IHI ::,

''"' ~Top view

~ IIH.1 :::,

Beam direction ~ Side view

,111111 NOvA - FNAL E929 Rull 18670 11 Lvent 1 /B402 U IC f n Jan 9 201 ~> 00 11 ~] 087141608 ~)11

• \11

1 .:;, H I

14 meters Color denotes deposited charge

.:;.:;,"' / ( 1..'lll) q I \])( I

NOvA Preliminary

120 NOvA 6.05x1020 POT-equiv.

• - Best fit prediction

> 100

(1) ....... Unoscillated prediction (!)

• +- Data

I!) 80 C\I ,

....

ci

60 Cf)

c

(1)

>

40

w

, .... 20 00 1 2 3 4 5

Reconstructed Neutrino Energy (GeV)

CFermilab

SLIDE 17

LAr detector modeling

• Modeling important for LBL and SBL - muons, pions,

protons, neutrons, electrons, photons at < GeV scale

• LAr TPCs on the surface have occupancy dominated

by cosmic rays generates space charge effect (SPE)

• flow model of the liquid combined with electron

recombination needed to model the generation of areas of static charge

• important for alignment, calorimetry, and muon

momentum measurements

• Muon g-2 experiment has injection/kicker magnets

that generate varying magnetic fields inside the detector volume - tracking spin-dependent particles

• ver 100km of path length (one of first exp to use

G4EqEMFieldWithSpin) with precision greater than 1 mm

17

ProtoDUNE simulation SPE

100

"Neutrinos+ cosmics"

cryostat

1

With-Flow MC: Top Face t.Y [cm] 200 300 400 500

'- ,'

' '

1.6 µs beam

spill time

500 400 300 200 100 600 700

Z,eco

; I I

'

• -

I I

2.3 ms drift time

With-Flow MC: Upstream Face t.Z [cm]

• 300
• 200
• 100

100 200 300 X reco

• 5
• 10
• 15
• 20
• 25
• 30
• 35
• 40

SLIDE 18

Photon Simulation in Neutrino Detectors

• scintillation in LAr is 10000 photons/MeV
• 2.2 MeV/cm for a MIP
• 20000 photons for every cm a muon traverses
• simulating 6000000 photons in FD on a CPU
• resorted to using photon lookup tables from


dedicated photon bomb simulations (memory
 problems)

• DUNE - development of trigger, final detector


design, shower reconstruction, and energy resolution
 depends upon photon simulations

• DUNE wants “natively GPU-accelerated optical photon tracking built into a fully-featured

detector simulation”

18

"Neutrinos+ cosmics"

cryostat

1

1.6 µs beam

spill time

; I I

'

• I

I

• 2.3 ms drift time
• ------------------------------ CFermilab

SLIDE 19

19

Simulation challenges in JUNO

• JUNO: Jiangmen Underground Neutrino Observatory
• Measuring Mass Hierarchy with Reactor neutrinos
• Nominal experiment setup
• 700 m deep underground
• 53 km baseline; 36 GW reactor power
• 20 kton LS detector; 18,000 20inch PMTs
• 3% energy resolution@1MeV
• Major backgrounds: Cosmic ray Muons (3 Hz)
• dE/dx: 2 MeV/cm, Light Yield: 10,000/MeV

=> Need to simulate millions of photons. => Several hours per event. Need >4GB memory.

Central Detector Muon Top Tracker Water Cherenkov Detector

Mean: 215 GeV

Note: The photoelectrons in the same 1 ns time window are merged in order to save memory.

Thanks Tao Lin for slides

lo' F---~~--.. lo' 10 1 0' .~ 10 ,li '

~11

I ' II Iv~•

• IOOGcV
• 21.SGeV
• SOOGcV

I TcV I io-•

L

50

1=...-1-::

100 W,1!1.1~1 is s~

• u:..i~

2~

• ii""''"

. ""i 250

s':i'- x

300 F'ull Simulation Timdmin 1 0' lo' 1 0' 10 1 PE 1 ..574.201 > I PE 13.952.473 >I PE (mc:rgc) 2..514.850

CFermilab

SLIDE 20

20

Voxel method: parameterized fast simulation

• Parameterized PMT’s response for the

scintillation photons.

• Generate scintillation photons in a certain voxel.
• Simulate the response with Geant4.
• Save the histograms of nPE and hit time.
• Use the response into simulation at runtime.
• Speed up the optical photons’ propagation in

LS by a factor of 50 using CPU.

• Several minutes per event.
• I/O is not included in the measurement.
• Status: a GPU version is under development.
• Speedup: ~200x CPU version.
• The pre-generated response are loaded into GPU

global memory only once.

• The collected steps information are copied to

GPU memory in each event.

• The results are copied back to CPU memory.

Thanks Tao Lin for slides

muon Water vertex. L

···.Jl

··•

• •• 0

P~IT

':\

....................

·►• cenlcr

.. '

♦ .,. .....

: ,r

.. _;

• E.5-2680 vJ

♦ ;

7 ~-

50 Get Track Get parameter and Calculate R Traversing Voxel 100 150 200 250 30( Full Simulation Time/min Event Action Traversing Energy(2nd)

,------.

Sampling to

• btain nPE

Calculate Theta Sampling t o

• btain hit

time Save data Traversing Voxel(lst) PMT(GPU parallol)

CFermilab

SLIDE 21

Mu2e is Using G4MT

21

• Pilot project: look at cosmic ray air shower events that produce

signal like tracks (with loose cuts)

– Use the CRY air shower event generator – What is the rate of events that cannot be vetoed because the particle crossing the cosmic ray veto system is neutral?

Rob Kutschke

reconstructed as e+ with -87 MeV

+nan ns +nan ns +nan ns +nan ns +nan ns +nan ns SimParticle:detectorfilter: Run#: 2701 Sub Run#: 405 Event#: 1313768

SLIDE 22

Mu2e G4MT Technology Roadmap

• Pilot project:

– Runs within single threaded art v2 – Event generator produces ~50K events – G4 processes all of the events MT and saves interesting

• nes
• Most events are small

– Drain the stash of processed events into art events – Late 2018: 750K hours on ANL BeBop (KNL and Haswell) – Now: now running 2.5 M hours on ANL BeBop and Theta

• Starting April

– Port art v3 which supports MT – Use G4MT within art using cmssw as a model – Have requested 100M hours on Theta at ANL to do 3 studies one of which is the CRY cosmic ray air shower study

22

Scaling with Pilot Technology ( ~35% serial )

2500 2000 ~ 1500

I

C:

f

LU 1000 50

, , ,

Event Simulation Rate versus Thread Utilization 70 60

• -----

Q) 50 i§

• c

..

Q) .l: 40 -; .9

s

..

• 32 processes

30 oc

• 16 processes

,g

..

8 processes

• c

20

• 4 processes
• 2 processes

10

• 1 process

100 150 200 250 300 3~ Total Number of Threads Utilized

CFermilab

SLIDE 23

simulation of neutron response in detectors

• one of the biggest challenges facing neutrino experiments
• final-state neutrons produced in most anti-neutrino charged current interaction
• energy deposit displaced from interaction vertex, separate de-excitation photons, inelastic

scatters, etc.

• as important as proton identification for complete event reconstruction and incident

neutrino energy determination - typically deposit only a fraction of their energy

23

MINERvA CC candidate event Muon Neutron Laura Fields

..

111 11,

p n

• ------------------------------ CFermilab

SLIDE 24

Neutron simulation detectors

• big challenge facing all neutrino experiments
• final-state neutrons produced in every anti-

neutrino charged current interaction

• energy deposit displaced from interaction vertex,

separate de-excitation photons, inelastic scatters, etc.

• as important as proton identification for complete

event reconstruction and incident neutrino energy determination - typically deposit only a fraction of their energy

24

AgroNeuT arXiv:1810.06502v1

Ivan Lepetic, FNAL JETP

2000 1800 1 600 1400 1200 1 000 800 600

u, 400

I

200 n a, 2000

E

i= 1800

1 600 1400 1200 1 000 800 600 400 200 50 50 100 150 100

Wire

150 200 200 a, Cl

...

Ill .c: ()

CFermilab

SLIDE 25

Neutron simulation detectors

• big challenge facing all neutrino experiments
• final-state neutrons produced in every anti-

neutrino charged current interaction

• energy deposit displaced from interaction vertex,

separate de-excitation photons, inelastic scatters, etc.

• as important as proton identification for complete

event reconstruction and incident neutrino energy determination - typically deposit only a fraction of their energy

25

AgroNeuT arXiv:1810.06502v1

Ivan Lepetic, FNAL JETP

100 90

• +- Data

80

• FLUKA

.l!l 70 C Q) 60 >

• Stat. Error
• Total Error

UJ

0 50

cu .0 40 E ::,

z 30

10 2 3 4 5 5.5+

Total Energy (MeV)

350 300 f-

• +- Data

I

• GENIE

250

Ill

I

• Stat. Error

c

Q)

Jj200 f-

Ji 150

E I f- I ::,

z

100 f- I

I

I

50 f-

I I I

2 3 4 5+

Number of Clusters

CFermilab

SLIDE 26

Conclusions

• Neutrino simulations have impact on neutrino oscillation measures at various levels and through

multiple analysis pathways

• Drive to precisely measure oscillation parameters have driven increases in the beam intensity and

detector size, but strong need for improvements in determination of systematic uncertainties - lean heavily on simulation with no “standard candle”

• event generation is based upon multiple models with limited ability to tune or establish uncertainties

since there is no complete theory

• simulation involves wide range of scales from 100 keV to 10 GeV
• overwhelming reliance on GEANT4 tunes and physics lists that are tuned to external data (mostly

designed for collider experiments) – branching of custom versions of GEANT4 and tunes can create difficulties for new versions – need to understand how we can systematically vary GEANT4 to give confidence in detector response

• need to utilize vectorized algorithms to tackle several specific problems (scintillation photons) and

want to build upon the experiences and tools of other experiments

26

• -----------------------------CFermilab