杨勇 上海交通大学 On Behalf of the Collaboration 粒子物理数据分析基础和前沿研讨会,2016/9/13 PandaX-II 物理数据分析 杨勇 上海交通大学 On Behalf of the Collaboration 粒子物理数据分析基础和前沿研讨会,2016/9/13 2018/12/4 杨勇，上海交通大学

Run history Series of Engineering runs Distillation and re-started (Run9, 80 days) 2014.10 2015 2016 Detector installation at CJPL Nov 22 – Dec 14, Physics commission (Run8, 19 days, stopped due to high Krypton background) The pandax-2 detector was installed in the lab in Oct 2014, and we had a couple of engineering runs in 2015. in the end of last year, we had one physics commissioning run about 19 days of DM data . We call this run as run8. it was stopped because we found an air leakage which leads to a high krypton contamination. So we paused the data taking and send the detector back to shanghai for krypton distillation. After that we resumed the data taking. Till the end of June, we have collected about 80 days of data. This is what we call run9. Run8+Run9=98.7 days Exposure:3.3x104 kg-day Phys. Rev. D. 39, 122009 (2016) 2018/12/4 杨勇，上海交通大学

Analysis chains overview RAW data: (PMT pulse) Reconstruction: S1, S2 and Vertex, Energy Calibration: Electron recoil Nuclear recoil Selections: Trigger Fiducial volume S1&S2 Results: Background and signal estimate, Limit calculation 2018/12/4 杨勇，上海交通大学

Calibration program Internal/external ER peaks: Detector uniformity corrections Light/charge collection parameters Low rate AmBe neutron source:  Simulate DM NR signal CH3T injection: tritium beta decays  Simulate ER background 2018/12/4 杨勇，上海交通大学

Single electron gain Identify smallest S2 in the data S2 <100 pE (mainly from one electron) fit with F-D fuction x two Gaussian Identify smallest S2 in the data Fitted with two Gaussians (with S2 rec. efficiency taken into account) SEG: 24.4 0.7 PE/e 2018/12/4 杨勇，上海交通大学

Extracting detector parameters W = 13.7 eV Gaussian fits to all ER peaks in data Linear fit in S1/E vs S2/E to extract PDE and EEE 2018/12/4 杨勇，上海交通大学

Comparison with NEST-0.99 NEST (Noble Element Simulation Technique): simulation of the excitation, ionization, and corresponding scintillation and electroluminescence processes in liquid noble elements. http://nest.physics.ucdavis.edu/site/?q=nest) NEST ER model Reasonable agreement observed between NEST ER model and values extracted from our data 2018/12/4 杨勇，上海交通大学

Energy nonlinearity and resolution < 2% nonlinearity achieved for [33, 1332] keV 129Xe (n,) 131Xe (n,) Resolution approaching 5% @ high energy 127Xe 2018/12/4 杨勇，上海交通大学

NR calibration Data medians ±2s MC medians ±2s ER median NR 99.9% acceptance S2>100 PE Ambe calibration data in log_10(S2/S1) vs S1. Medians of data: red solid circles : MC simulation: green square. 2-sigma boundaries, red (data) and green (MC)….. 162.4 hours of AmBe data taken, with ~3400 low energy single scatter NR events collected NR median curve and NR detection efficiency determined 2018/12/4 杨勇，上海交通大学

ER calibration with CH3T ER median NR median Events 18.0 hours of tritium data taken, with ~2800 low energy ER events collected. (<E>=5.7keV) 9 events leaked below NR median, (0.32  0.11)% Consistent with Gaussian expectation 2018/12/4 杨勇，上海交通大学

Event Selections Summary Notes Trigger Mostly triggering on s2 signal, threshold about 80 PE Data quality S1, S2 Identification, bad files removal, etc Skin Veto Suppress ambient gamma background Single S2 Remove double-scattering events Fiducial volume Suppress ER&NR background BDT cut Suppress accidental background 3<S1<45 PE, S2(raw)>100 && S2<10000 PE Low energy selection 2018/12/4 杨勇，上海交通大学

Background Overview Background Run8 Run9 Estimate Method Electron Recoil Mostly Kr-85 Mostly Kr-85, Xe-127 Normalization from data, distribution from MC&Data (eff. applied) Accidental Single S1 and single S2 random coincidence From Data Nuclear Recoil Neutrons in detector materials Neutron spectra calculated with SOURCES4A package, distri. from MC (eff. applied) Neutron background mostly from PTFE 2018/12/4 杨勇，上海交通大学

Krypton-85 background krypton is intrinsically present on the ppb (parts per billion level=10−9 Kr/Xe ) in commercially available xenon. (In atmosphere, ppm=10-6 level) Kr85/Kr = 2x10-11 85mRb 1.015 ms 2018/12/4 杨勇，上海交通大学

KITPC WORKSHOP, BEIJING 85Kr Estimated from delayed b-g coincidence analysis Uniformly distributed Significantly reduced after distillation Waveform: one b one g Commissioning run (Run8) 1st Physics run (Run 9) 2018/12/4 KITPC WORKSHOP, BEIJING

ER background Normalization of Xe127&Kr-85 extracted from data 127Xe (408 keV) 127Xe/129mXe 127Xe from the 408 keV (375+33.2 keV) peak Normalization of Xe127&Kr-85 extracted from data ~0.4 mDRU(10-3/keV/kg/day) for <10 KeVee 2018/12/4 杨勇，上海交通大学

Accidental background Isolated S1 before an S1 Isolated S2 pair Isolated S1 and S2 were selected and randomly paired to simulate accidental events Significantly reduced using a multivariate technique (BDT). x3 rejection for background 90% efficiency for signals 2018/12/4 杨勇，上海交通大学

BDT discrimination power Accidental S1 and S2 lack intrinsic correlations Single S2 likely originated from the gate grid (small width, large asymmetry) 2018/12/4 杨勇，上海交通大学

Overall background rate in run9 Events selected with energy <10 keV ~2 mDRU on average (15.3 mDRU in Run 8) Decrease over time due to 127Xe decay 2018/12/4 杨勇，上海交通大学

WIMP-Xenon Recoil Energy and Rate Total recoiling events / ton xenon / year sc-p=10-45 cm2 sc-p=10-45 cm2 Xenon 2018/12/4 杨勇，上海交通大学

WIMP signal efficiency > 80% efficiency for recoil energy 10-20 keV DM detection efficiency 2.5% @ 10 GeV, 52% @ 100 GeV 2018/12/4 杨勇，上海交通大学

Results Run8 Run9 Below NR med. ER Accidental NR Total bkg. Data Run8 ER Median NR Median NR 99.9% acceptance S2>100 PE Below NR med. ER Accidental NR Total bkg. Data Run8 2.0 0.33 0.09 2.4 ±0.8 2 Run9 1.2 0.84 0.35 2.4 ±0.7 1 2018/12/4 杨勇，上海交通大学

Final candidates in Run9 Gray: all Red: below NR median Green: below NR median and in FV 389 total candidates found in the FV 1 below NR median Outside FV, edge events more likely to lose electrons, leading to S2 suppression 2018/12/4 杨勇，上海交通大学

Selected Data/Bkg in 1-D comparison log(s2/s1) distribution for all s1 bins Run8+9 Good agreement between data and background prediction 2018/12/4 杨勇，上海交通大学

CLs method for upper limit Derive a upper limit on signal strength m, given a desired CL. (1) Construct Likelihood function binned unbinned (2) Construct test statistics (profiled likelihood ratio) (3) (4) Calculate two p-values 2018/12/4 杨勇，上海交通大学

CLs method for upper limit Derive a upper limit on signal strength m, given a desired CL. (5) Calculate (6) Adjust m until CLs(m) = 0.05, for 95% CL , or 0.1 for 90% CL 2018/12/4 杨勇，上海交通大学

WIMP-nucleon SI upper limits 90% CL Spin-independent limits calculated with CLs method X10 X2 2018/12/4 杨勇，上海交通大学

Comparison with LUX latest result PandaX-II 33 ton –day, LUX 33.5 ton-day 2018/12/4 杨勇，上海交通大学

Summary and outlook PandaX-II has reached the forefront of the DM search! 79.6 live-day of dark matter data with much suppressed background (152 mDRU) In combination with commissioning run (19.1 day), ~33,000 kg-day exposure in total. No DM signals are observed The WIMP-nucleon elastic scattering cross sections are constrained to 2.5x10-46 cm2 at 40 GeV DM mass. Continue PandaX-II data taking till end of 2017. In preparation for PandaX-xT! 2018/12/4 杨勇，上海交通大学

谢谢！ 2018/12/4 杨勇，上海交通大学