Physics in DUNE UK WP1 Steve Dennis DUNE UK Meeting University of - - PowerPoint PPT Presentation

Steve Dennis 1

Steve Dennis DUNE UK Meeting University of Manchester 05/03/2019

Physics in DUNE UK WP1

Steve Dennis 2

Introduction

• An important time for DUNE physics as much as anything

else.

• The physics TDR is due in May.
• Final analysis should be frozen by April 8th.
• Physics results will be produced by groups from the UK
• CAFAna (C. Backhouse et al.)

–

The primary analysis framework used for NovA physics

• VALOR (C. Andreopoulos, SD et al.)

–

Designed for T2K, now used for SBN etc

• Our work is on long-baseline physics, and work to justify a

near-detector design.

Steve Dennis 3

Release of new GENIE Software and Tunes

• GENIE is the standard Monte Carlo generator used by DUNE

(and most other GeV+ scale neutrino experiments)

• After a long time, we’re proud to announce the release of

GENIE v3.

• 10th October 2018 for v3.0.0, now at v3.0.2.
• This comes with a strategy change – rather than simply

having a single physics model available, we will ofger a number of difgerent confjgurations, all of which are offjcially supported.

• Each of these are ofgered tuned to a number of datasets.
• Currently released everything with a deuterium/bubble chamber tune.
• Nuclear data tunes will be released soon!

Steve Dennis 4

GENIE Confjgurations in v3.0

see releases.genie-mc.org for full list

Steve Dennis 5

Example of the GENIE tuning

T uning of CC-inclusive nucleon cross-section using free nucleon data.

Black is nominal prediction Red is tuned on inclusive data Green is tuned on exclusive data Model G18_01a (default+new processes)

Steve Dennis 6

TDR LBL Analysis

(from C. Marshall)

Steve Dennis 7

Changes to the Oscillation Analysis

• Previous attempts at performing the oscillation analysis and

evaluating near detector designs used a highly complex analysis.

• Fitting many difgerent samples to maximise oscillation sensitivity.
• This proved intractable to properly apply systematic uncertainties to.
• “Sensitivity” does not tell the whole story about the near detector.
• We have refocused our efgorts on “robustness”:
• A simple near detector design should be able to provide suffjcient

sensitivity if our model is right and fjts our data.

• But when has that ever been true?
• Now using a simpler ND analysis for the TDR and using “adversarial”

studies to justify a more capable ND design.

Steve Dennis 8

ND Design Concept

75 kt pixel-readout non-magnetised LArTPC

“Multi-purpose detector” High pressure gaseous argon TPC 1 t fjducial mass (10 Atm) 0.5 T magnetic fjeld Surrounded by ECals

Steve Dennis 9

ND Samples for the TDR

• Currently implemented/working:
• Just numu CC inclusive in liquid argon.

–

The basic rate/shape that will create our oscillation signal.

• On the way, should be ready for the TDR fjts:
• Nu-e elastic scattering on liquid argon.

–

Extract fmux measurement independent of cross-section.

• Samples in gaseous argon.

–

Better probing of nuclear efgects.

• For future ND designs, more “challenge” datasets.
• Particularly look for ways in which we could be plausibly wrong that

we cannot identify with an on-axis detector that a prism-like design would identify.

Steve Dennis 10

Expansion of Systematics Model

• Flux systematics remain pretty much unchanged (calculated for
• ptimised beam design using PPFX), in 104 bins of neutrino energy

and fmavour per detector (208 total).

• Detector systematics have been improved signifjcantly.
• See following talk by Seb Jones.
• I’ll talk about changes to the interaction model.
• Past versions used GENIE systematics.

–

Either transformed to an efgective correlated-normalisation based model for everything except FSI.

–

Or using the GENIE dials manually for everything.

• Still using pretty much all of the manual GENIE dials (including FSI)
• Also have a couple of others necessary for DUNE (eg Ar40 scaling).
• And obvious ones for appearance experiments (nue/numu, nuebar/numubar)

Steve Dennis 11

New Interaction Parameters

• Added a large number of additional neutrino interaction

uncertainties to the analysis.

• These are mostly “inspired by” the current T2K oscillation analysis.

–

The T2K Neutrino Interactions Working Group has put a great deal of efgort into coming up with an efgective set of parameters for a mature experiment – this is as close to state-of-the-art as it gets.

➔ BeRPA (parameterisation of RPA, lowQ2 suppression in CCQE) ➔ CCQE Pauli Suppression ➔ 2p2h shape dependence. ➔ Uncertainty from comparison to MK Model for Single Pion ➔ Low Q2 Suppression of Single Pion (ofg right now) ➔ Binding energy efgects on lepton momentum ➔ Nonresonant pion multiplicity normalisations (DUNE specifjc)  23 of these

Steve Dennis 12

TDR-Style Outputs

• I’ll give examples of the important output plots we’re going

to be putting in the Physics TDR.

• Note that these are all work-in-progress and should not be

used outside this session.

• The plots shown here were produced with CAFana.

Steve Dennis 13

CPV and MH Sensitivity

MH Sensitivity (including reactors)

Steve Dennis 14

Error constraints

Flux FD Interactions

Steve Dennis 15

Systematics Correlations (postfjt)

Steve Dennis 16

Near Detectors - Why go ofg-axis?

• It’s become increasing accepted by members of

the ND and LBL groups that we should try to aim for a near-detector that can be moved into an ofg- axis position.

• As we move further ofg-axis, we get a beam fmux

that is much more sharply peaked in neutrino energy.

• Using a linear combination of fmuxes at difgerent
• fg-axis positions, you can analyse a pseudo-

monochromatic beam.

DUNE-DOC-8106

→ Can produce a map of true energy to reconstructed energy that is far less dependent on our interaction model

This is obviously good for physics. But moving your TPC 30m is expensive

→ Need a strong physics justifjcation for inclusion.

And mere sensitivity does not show this, if you assume your model is correct.

Steve Dennis 17

Potential Biases with on-axis only

• Need to fjnd a plausible physics hypothesis where we get

the wrong answer with on-axis only, and the right answer including an ofg-axis sample.

• One such suggestion has been developed (outside the UK)
• Cristóvão Vilela (Stony Brook)
• I’ll quote his idea in his words on the next slide.
• This model will be pushed through the oscillation analysis

to see if it can be used to justify a PRISM study using the UK-based fjtting packages.

Steve Dennis 18

PRISM Fake Data Strategy

(from C. Vilela)

Steve Dennis 19

Remaining Goals for the Year

• Priority for the rest of the year after the Physics TDR is evaluating

further ND designs.

• Rough analysis approach will be similar to the TDR justifjcation of

the ofg-axis analysis.

1) Come up with a plausible alternative model (fjrst will probably be simply NuWro, then we’ll get more complex) 2) Generate fake data 3) Evaluate biases induced by the difgerent model. 4) Introduce additional ND sample. 5) It fjxes it! (we hope – if not, fjnd a new sample or warn the ND group)

• This new approach was heavily driven by the feedback coming out
• f management that we haven’t suffjciently difgerentiated difgerent

ND technologies and thus demonstrated that we actually need the expensive detector we all think we need.

Steve Dennis 20

Conclusions

• Our generator/modeling goals have mostly been met.
• Released and maintaining many useful, self-consistent sets of models.
• Each of these has been released with the old default model and a free nucleon tune.
• Nuclear-data tuning release is in the work.
• The long-baseline physics group is working hard towards the rapidly

approaching TDR.

• Using tools primarily developed by UK groups.
• There have been signifjcant improvements this year in the interaction

uncertainty modelling, using uncertainties motivated both purely by models and by comparisons with other experimental data.

• We have developed a new strategy to meet the demands from the LBNC and

collaboration that rather than simply evaluating the performance of a specifjc detector technology, we identify specifjc forms of realistic mismodeling that each technology would resolve.

• The remainder of the year will be spent developing and testing these adversarial

datasets.