The STEREO Experiment a Search for a Sterile Neutrino ~1Ev at Short Baseline

The STEREO Experiment a Search for a Sterile Neutrino ~1eV at Short Baseline

Adrien Blanchet CEA Paris-Saclay/Irfu/DPhN

GDR Neutrino

the 26th of June 2019 Physics Motivations Motivation : Testing the Sterile Hypothesis

• 2011 : Re-evaluation of the neutrino reactor ﬂux prediction

• Reactor Antineutrino Anomaly (RAA) • All reactor short-baseline experiments are observing a deficit (~6%) • Confirmed by recent accurate measurements from Daya Bay, RENO & Double Chooz • Adding a fourth "sterile" neutrino may explain the deficit : Phys. Rev. D 83, 073006 (2011)

6% deﬁcit ➡ STEREO will be able to measure an oscillation pattern without prediction

2 eV2 sin2 Δm14 = 2.3 ; (2θ14) ≃ 12 % 3 Motivation : Nuclear Physics Biases

Shape Anomalies : Several experiments Bugey-3 Double Chooz Daya Bay RENO NEOS ROVNO Gösgen revealed a "bump" around 5 MeV w.r.t. Courtesy of G. Mention predicted spectrum

• Could be linked to underestimation of some isotopes of uranium/plutonium • Cannot explain the total deﬁcit • Everyone does not see the same spectral distortion

Rate Anomalies : Studies of decorrelation of isotopes contributions at commercial reactors

→ Claim that the deficit is carried by U235

➡ STEREO will measure the pure U235 neutrino spectrum

4 Worldwide Experimental Program

DANSS Neutrino 4 Solid NEOS STEREO PROSPECT

Research reactor (Pure Research reactor (Pure Power reactor (Mixed U235), Li-loaded U235) Gd-loaded Pu9/U5), Gd-loaded 5 The STEREO Experiment STEREO : Experimental Site

ILL research facility, Grenoble,Water channelFrance 15 m.w.e overburden

Research reactor core ~ 58MWth 19 → 10 ν¯e/s • No Damping • Compact core (∅40cm × 80cm) • Short baseline measurement 9.4 m < Lcore < 11.2 m • Pure U235 neutrino spectrum : Highly enriched fuel

93 tons moved on air cushions AutumnAutumn 2016 2016

Challenging Mitigation of the Background

• Gamma & Neutron Background from neighboring experiments • Surface-level experiment (15 m.w.e thanks to water channel)

7 STEREO : Detection Principle

Interaction Channel : Inverse Beta Decay (IBD)

Prompt Delayed

4.5 MeV < E < 10 MeV Time Interval Cuts :

1.5 MeV < E < 10 MeV

Δt < 70 μs

8 STEREO : The Detector

9 STEREO : Data Taking

Phase-I Phase-II ON = 66 days ON = 119 days OFF = 22 days OFF = 211 days

• Data taking efﬁciency: 98.5% • Calibrations: 3.3% • Nb of days in effective time : 14% dead time

10 Data Analysis Prompt Signal : E Reconstruction

Energy calibration • Weekly calibration with 54Mn sources • Dedicated algorithm to go from deposited Testing detector's response with charge to reconstructed energy several gamma sources

1.04 Cell 1 Cell 2 1.03 Cell 3 Cell 4 AmBe Zn K 1.02 Cell 5 AmBe n-H Cs Mn Co AmBe n-Gd

Na Cell 6 1.01

1.00

0.99

Data/MC Erec detector 0.98 0.97 Preliminary 0.96 0 2 4 6 8 10 Nominal Energy (MeV)

12 Prompt Signal : E Reconstruction

Evaluation of Systematics • Use the cosmic-induced n-H peak (Quasi-uniform vertex distribution, 3 similar to the neutrino distribution) Cell 1 Cell 2 2 Cell 3 Cell 4 Cell 5 Cell 6 1

−1

Relative Deviation to Mean (%) −2 Preliminary −3 31/12/17 01/04/18 02/07/18 01/10/18 31/12/18 Run Date

n-H Candidates in Cell 2 1.0 Crystall Ball Fit

0.8 • Monitoring time stability of energy Events (a.u.) response 0.6 Compare energy response with 0.4 • cosmic simulation 0.2

0.0 1.5 2.0 2.5 3.0 3.5 4.0 13 Reconstructed Energy (MeV) Delayed Signal : Detection Efﬁciency

AmBe source in cell 4, z=45cm 0.06 Dominant uncertainty in efﬁciency Simulation, GLG4sim comes from the neutron capture events (energy scale anchored to nH) Data 0.04 → Dedicated studies using Am-Be neutron source deployed in the detector → γ cascade of gadolinium not well described 0.02 in Geant 4 → Correcting Efﬁciency at the % level using

3D model of n-detection efﬁciency 123 45678910 2 → Prompt-Delayed Time interval well 1

described data/mc 0 123456789 10 Reconstructed Energy [ MeV ]

AmBe source at z=45cm 2D map of the data/mc neutron eciency data/mc 3000 900 Data Cell 1 Cell 2 Cell 3 Cell 4 Cell 5 Cell 6 0.95 2500 800 Simulation 2000 700 0.94 events

1500 z axis [mm] 600 0.93 1000 500 0.92 500 400 300 0.91 1.5 200 1.0 0.90 0.5 100 data/MC 0.89 10 20 30 40 50 60 70 0 correlation time [µs] 1000 500 0 500 1000 x axis [mm] 14 Delayed Signal : Improving Gd Cascade

Use of FIFRELIN's data for Gd cascade

3500 → Researchers from CEA Cadarache generated for us GLG4Sim 3000 a lot of Gd cascades → Very good agreement with data 2500 Data → Suppress data/MC discrepancies with associated 2000 systematics → Published on Arxiv (preprint) + Zenodo (data) 1500 → arXiv:1905.11967

Counts (a.u.) 1000 → Experiments already using it < → 500 DANSS → WATCHMAN 0 1 2 3 4 5 6 7 8 9 10 Reconstructed Energy (MeV)

3500

3500 3000 FIFRELIN GLG4Sim 3000 2500 Data 2500 Data 2000 2000

1500 1500

Counts (a.u.) 1000 < Counts (a.u.) 1000 500 500 0 1 2 3 4 5 6 7 8 9 10 0 Reconstructed Energy (MeV) 1 2 3 4 5 6 7 8 9 10 Reconstructed Energy (MeV) 3500 15 3000 FIFRELIN

2500 Data

2000

1500

Counts (a.u.) 1000

500

0 1 2 3 4 5 6 7 8 9 10 Reconstructed Energy (MeV) Remaining Background : Cosmic Induced

Using Pulse Shape Discrimination (= Qtail/Qtot)

PSD properties of the LS allow to discriminate neutrinos from dominant remaining cosmic background

• PSD distributions are corrected from temperature & evolution of optical properties • ON data ﬁtted using OFF model • Background model given by OFF data • Accidentals accurately measured and included in the model • Gaussian added for neutrino component

Neutrino rates are extracted for each cell and Energy bin

Extraction relies on two hypothesis only

• Gaussian shape of the neutrino component • Stability of the correlated background 16 Background Stability

Major Background Contribution : Cosmic Induced

• Clear correlation between atmospheric pressure and counting rate

PSD background shapes independent of atmospheric pressure

• Splitting data in 2 groups : Low and High pressure • Normalization factor = (93.3 ± 0.25)% • Expected for a 10 hPa difference = (93.8 ± 0.3)%

High stability of the shape of the background PSD distribution is demontrated for all cells 17 Extensive Biases Analysis

2 ON−(a × OFF + Gν + acc) χ = ∑ + . . . bins σ

Note: Here expressed under χ² for better understanding, but using binned Log Likelihood for minimization

Likelihood Estimator can be Biased A B ∼ + + . . . N N2 • Merging of all periods is necessary to minimize the bias on likelihood estimator • Bias << statistical uncertainty for all energy bin • Constrains our binning choice 18 New Results Results : 1/L² Law

STEREO Phase-II data only

• Measured neutrino rates corrected for detection efﬁciency of each cell found in perfect agreement with the expected 1/L² law:

N 1 ν ∝ 2 ϵdet, cell i L

20 Results : Oscillation Analysis

Prediction Independent - Shape Only

• Build model, adding nuisance parameters • No input from predicted spectrum: φb are free parameters for each energy bin but common to all cells

2 Δi = Δc=cell ; b=Ebin ≐ Dcb − ϕbMcb(Δm14, θ14)

Fully free parameters only constraint : common to all cells

χ2 = Δ V−1 Δ ∑ ∑ i ( cov)ij j i j 21 Results : Oscillation Analysis

Statistical Inference done with a χ² formalism χ2 = Δ V−1 Δ ∑ ∑ i ( cov)ij j i j Oscillation scenarios (H) rejected using Δχ² 2 2 2 Δχ = χH − χbest fit

Vcov = Vstat + ∑ Vs s=syst

Systematics • Cell-Correlated Escale • Cell-Uncorrelated Escale • Cell-Uncorrelated Norm

22 Results : Oscillation Analysis

Covariance matrices depend on the prediction True/God's 2 value 2 (D − M) Error of the instrument, χ ∼ Stereo's associated w/ the True σ(M)2 measurement value

here's yours ! 2 − + ϕbMcb(Δm14, θ14) (Mc,b + Mc,b−1)Eb − (Mc,b + Mc,b+1)Eb ΔM ∝ cb 2(E+ − E−) Statistical error estimated with b b pseudo experiments

Low statistic expected for some oscillation scenarios 23 Results : Oscillation Analysis

Very good agreement between Data and Model non-oscillated × φi

no-sterile hypothesis not rejected

24 Rejection Contours

• Due to detector maintenance phase I & II can be considered as independent measurements.

• Preliminary combination is done by summing the two χ² and assuming a standard χ² law for the C.L.

• Best-ﬁt value of the RAA rejected at C.L. ~ 99.8%

25 Conclusion and Outlook Conclusion & Outlook

STEREO demonstrates its high precision capability:

• Detection – 43.4 k neutrinos detected in phase-II, 65.5 k total • Good control of Prompt and Delayed Signal • Backgrounds – Extensive measurements (233 days OFF and 185 ON) show a very high stability of the background • Sterile ν exclusion contours – Major fraction of the initial RAA contour is now rejected with no sign of cell-to-cell systematics • n-Gd FIFRELIN paper already available (arXiv:1905.11967) with cascades on Zenodo

Perspectives in the (very) near future:

• Oscillation analysis paper is currently being written • Absolute measurement – Accurate measurement of the pure 235U neutrino rate, publication is planned by the end of the year • Spectrum shape – Very sensitive to Escale, studies are underway • x2 stat. until spring 2020

27 Conclusion & Outlook

Thanks for your Attention !

Spokesperson: David Lhuillier (CEA) Contact: [email protected] Website: http://stereo-experiment.org

L L

Photo: S. Schoppmann I

: o t

28 o The STEREO Collaboration h P Backup Slides PROSPECT @ HFIR, USA

Phys. Rev. Lett. 121, 251802

35 days of data RAA rejection @ CL>95%

arXiV:1812.10877

• 7-9m from a 85MW core • SNR=2 with only 1 m.w.e. • Good control of Escale • Pure 235U ﬁssion n spectrum • Disfavors 235U as the sole source of the 5 MeV bump at 3 s C.L. 30 NEOS @ Hanbit-5 Reactor, Korea

Phys.Rev.Lett. 118 (2017) no.12, 121802

• 25 m from a 2.8 GW core • SNR=22, ~2000 nu/day • Large size of the core damps down the oscillation signal at low E • Oscillation analysis relies on the comparison with an external prediction. Daya Bay spectrum is used as a non-oscillated reference. 31 DANSS @ Kalinin Nuclear Power Plant, Russia

Phys.Lett. B787 (2018) 56-63

• 11-13m from a 3.1 GW core • SNR>30, ~5000 nu/day • Large size of the core (3-4m high) damps down the oscillation signal at low E • Low E resolution (17%@4MeV) 32 Sterile Neutrino Discovery ?

JHEP 1808 (2018) 010

• NEOS + DANSS rejected large mixing angles in the RAA region

• Both analysis tend to the same best-ﬁt spot • Δm² = 1.3 eV² • sin(2θ) = 4% • Sub-percent oscillation implies a good control of systematics

★ Complementary measurements near research reactor are needed to get rid of the damping 33 Spectrum Shape Analysis

STEREO Phase-II data only

• Predicted spectrum • Huber 235U spectrum • + % level corrections in the ﬁrst 2 energy bins (n-Al capture, off-equilibrium effect, spent fuel).

• Good agreement with the data up to 6.375 MeV (χ²=14.9/18)

• Large deviation observed in the 3 highest energy bins (χ²=33.3/21) • Such localized large distortion cannot be explained by varying the 3 parameters of a quadratic model of the energy scale

• Further constraints from upcoming higher statistical accuracy and combination with other pure U235 spectra are required to draw pertinent tests of the spectrum shape

Next Step : • Reﬁned tuning of the MC • Complementary calibration observable (source at 6 MeV (Am-C), Boron 12 spectrum ...) • Improved background rejection (NN for cuts optimization) 34 From Δχ² to C.L. :

As neutrino statistic increase, normal χ² laws start being a good approximation to estimate Conﬁdence Level

Reﬁned Statistical Inference : • Use evolving statistical error with mixing • Correct covariance matrix volume in χ² • Unbinned best-ﬁt search

Sensitivity Contours for Phase 2 (Raster-Scan) ) 2 10 (eV 2 14 m Δ

90 C.L. with Normal Δχ2 Law

90 C.L. with PDFs 10−1

−1 10 2 1 sin (2θ14) 35 Prompt Signal : E Reconstruction

Monitoring time stability 3 Evaluation of Systematics Cell 1 Cell 2

• Use the cosmic-induced n-H peak (Quasi-uniform vertex distribution, 2 Cell 3 Cell 4 similar to the neutrino distribution) Cell 5 Cell 6 1

−1

Relative Deviation to Mean (%) −2 Preliminary −3 31/12/17 01/04/18 02/07/18 01/10/18 31/12/18 Run Date

n-H Peak Comparison Between Data and MC 3

n-H Candidates in Cell 2 1.0 1 Crystall Ball Fit

0.8 0 Events (a.u.) 1 0.6 − Data / MC Deviation (%) −2 0.4 Preliminary −3 0.2 1 2 3 4 5 6 Cell # 0.0 1.5 2.0 2.5 3.0 3.5 4.0 36 Reconstructed Energy (MeV) Cuts

37 PSD Correction

38 Neutrino Extraction Method

39 Estimator Bias

Neutrino rates are corrected from the bias of the likelihood estimator computed using pseudo-experiments (bias < 2%)

40 S/B and Background Shape

Signal / Background Ratio Background Spectrum