SuperCDMS: Recent Results for low-mass WIMPS

Transcription

1 SuperCDMS: Recent Results for low-mass WIMPS David G. Cerdeño Institute for Theoretical Physics Universidad Autónoma de Madrid for the SuperCDMS Collaboration

2 Hints for low-mass WIMPs in direct detection experiments CDMS II Ge DAMA/LIBRA XENON S CDMSlite CoGeNT LUX CDMS II Si CRESST II CDMS II Ge EDELWEISS (LT) CDMS II Si: Phys.Rev.Le. (03) 530 CDMSlite: Phys.Rev.Le. (04) 0430

3 Particle Physics models provide candidates for light DM Among other possibilities: Supersymmetry neutralino in the MSSM or NMSSM sneutrino in extended models Asymmetric DM Are these theoretical predictions within the reach of our detectors? Sneutrino DM (Cerdeño, Peiró, Robles) It is an appealing window of the DM parameter space that is essential to explore 3

4 The search for low-mass WIMPs is challenging The signal is expected at very low recoil energies Favours light targets Low-threshold searches Ge is relatively heavy so the threshold has to be just above the noise to be sensitive to 5 GeV WIMPs WIMP scatters / kg / d in Ge SuperCDMS! analysis range lower recoil energy! =! sensitivity to lighter WIMPs trigger threshold.6 kevnr Backgrounds are more difficult to discriminate (this is not a background free search) 4

5 SuperCDMS at SOUDAN Operational since March 0 izip interleaved Z-sensitive Ionization & Phonon detectors Instrumented on both sides with charge+ 4 phonon sensors 3 Diameter.5 cm Thick 9.0 kg Ge (5 izips x 600g) Data for this analysis: 577 kg-days taken from Mar 0 July 03 7 izips with lowest trigger threshold Side Side 5

6 izip discrimination of surface events In the new izips the ionization lines (±V) are interleaved with phonon sensors (0V) on a ~mm pitch Bulk events: charges (e,h) drift to both sides of the crystal Surface events: charges (e,h) drift to only one side of the crystal e - h + e - h + Z-PARTITION: The resulting symmetry/asymmetry in charge collection in sides and 6

7 izip discrimination of surface events In the new izips the ionization lines (±V) are interleaved with phonon sensors (0V) on a ~mm pitch Bulk events: charges (e,h) drift to both sides of the crystal Surface events: charges (e,h) drift to only one side of the crystal Sidewalls Surface events on the sides of the detector leave more energy in the outer sensors. Z-PARTITION: The resulting symmetry/asymmetry in charge collection in sides and RADIAL PARTITION: division of energy between inner and outer sensors 7

8 The rejection of surface events with the new izips using Z-partition has been demonstrated with data from exposure to betas from Pb sources In ~800 live hours, no events leaked into the 8-5 kev signal region Rejection at >.7x -5 This could allow a background free search for 5 yr of operation in SNOLAB (~0 kg) Appl.Phys.Lett. 3 (03) 645 (the low threshold analysis corresponds to smaller energies and some leakage is expected) 8

9 Measurement of the recoil energy V IonizaYon Energy [kev] 5 - Ionization for nuclear recoils, measured from 5 Cf data: Detector: TZ C.L. < 68% 68% < C.L. < 95% 95% < C.L. < 99% C.L. > 99% Measurement Lindhard Best fit Measured Lindhard Best Fit ptnf [kev] Total phonon energy [kev] prompt phonons Luke phonons Total phonon energy = E total = E luke + E recoil E total is measured with phonons NR equivalent energy = E total E Luke NR E Luke NR estimated from mean NR ionization, varies with E total 9

10 Sources of background Bulk electron recoils Compton background.3 kev activation line Yield = Ionization/phonon helps discriminating NR from ER Sidewall & surface events betas and x-rays from Pb, Bi, recoils from 06 Pb, outer radial Comptons, ejected electrons from Compton scattering Z-Partition and Radial partition define a fiducial volume Neutrons (cosmogenic & radiogenic) Use active and passive shielding. Simulation determines remaining irreducible rate

11 Analysis: Selection criteria and efficiencies We carry out a blind analysis, with all singles in energy range following 5 Cf calibration due to activation removed from study, except data Data Quality: Reject periods with poor detector performance Remove misreconstructed and noisy pulses Measure efficiency with pulse MC Quality + Thresholds Trigger and analysis threshold: Select periods with stable well-defined trigger threshold Measure efficiency from 33 Ba calibration data Preselection: Single-detector scatter Remove events coincident with muon veto Ionization fiducial volume Ionization and phonon partitions consistent with NR Boosted Decision Tree: Optimised cut on the phonon fiducial volume and ionization yield at low energy Efficiency estimated from fraction of 5 Cf passing + Preselection + BDT Efficiencies: measured for neutrons from 5 Cf. Corrected for multiple scattering with Geant4

12 ecision Tree DT inputs [kev] [kev] gy [kev] 3 rgy [kev] 3 tition artition 0.4 partition 0.4 Number Number of of events events Number of of events events simulation" ents reweighted ev WIMP Boosted Decision Tree (BDT) Number of events Boosted Decision Tree 5 5 Total phonon energy [kev] Total phonon energy [kev] total total phonon phonon energy energy [kev] [kev] Number of events Inputs (per detector) Number Number of events of events Number of of events / 0.04 / Number of events Ionization energy [kev] ionization energy [kev] Boosted Decision Tree BDT inputs summed over detectors summed over detectors BDT BDT inputs BDT output inputs Number of events Number of events ionization energy [kev] Ionization energy ionization energy [kev] 5 Total phonon energy total phonon energy [kev] Number of events Number of events GeV GeV WIMP" summed WIMP" -4 over # = 6 x -4 cm cm detectors Number of events / 0.04 BDT output Output 5 Total phonon energy [kev] total phonon energy [kev] Number of events / 0.04 BD BD summed over dete summed over dete GeV WIMP" # = 6 x -4 cm BDT score BDT score per BDT score Construction: BDT per detector" Construction: BDT per detector" Optimization: set cuts 0. Vertical 0. phonon Vertical phonon to WIMP ( GeV)" phonon z-partition Radial phonon WIMP (-0.4GeV)" -0. Radial 0 phonon Optimization: 0.4 set cuts phonon r-partition phonon Vertical r-partition simultaneously phonon to minimize Vertical phonon phonon z-partition WIMP ( GeV)" phonon z-partition simultaneously Sidewall WIMP to minimize (06Pb" GeV)" Sidewall 06Pb" Background model: pulse expected simulation" 90% CL upper limit from" from" Sidewall 06Pb" expected 90% Sidewall CL upper!" 06Pb" limit Background model: pulse simulation" Signal Background Sidewall!" model: model: 5 Cf NR pulse events on WIMP-nucleon simulation" reweighted cross from" Face!" from" Pb Sidewall!" on WIMP-nucleon Sidewall cross!" Pb Signal model: 5 to Signal Face!" match model: 5, 7,, 5 Cf NR events Background: reweighted Modelled Cf and NR 5 events section with simulated GeV WIMP reweighted Face!" data Pb on sidebands section Pb.3 Face kev!" to match 5, 7,, and 5 GeV and line" to.3 WIMP match kev calibration. line" 5, 7,, and 5 GeV WIMP.3 kev line"!! Gammas.3 kev line"! Gammas Gammas Gammas WIMP Signal: Modelled with NR data from 5 Cf, then rescaled for WIMPs with mass 5, 7,, 5 GeV

13 Unblinding: Before BDT cut Events passing all the cuts prior to applying BDT Ionization energy [kev] 4 3 Lindhard nuclear-recoil energy [kevnr] kev line Approx NR band 0 Outer radial events Total phonon energy [kev] 3

14 Unblinding: After BDT cut candidates ( expected) Ionization energy [kev] 4 3 Lindhard nuclear-recoil energy [kevnr] Total phonon energy [kev] 4

15 Unblinding: After BDT cut candidates ( expected) Ionization energy [kev] 4 3 Lindhard nuclear-recoil energy [kevnr] % CL intervals for WIMPS with m=5 GeV m=7 GeV m= GeV m=5 GeV Total phonon energy [kev] 5

16 Post-unblinding discussion Events are high in quality. Only the lowest energy candidate looks like spurious noise For most of the detectors there is good agreement with predicted background GeV BDT P-value = 0.4 6

17 Post-unblinding discussion Events are high in quality. Only the lowest energy candidate looks like spurious noise For most of the detectors there is good agreement with predicted background However, T5Z3 observes the 3 highest-energy events (Poisson p-value is 0.04%) Events passing BDT range Range of counts of counts with p>0.05 with p-value>0.05! σ background expectation expectation observed observed shorted ionization guard 0 TZ TZ TZ T4Z T4Z3 T5Z T5Z3 Detector $ T5Z3 has a shorted ionization guard. This may have affected the background model performance. Additional studies are undergoing. 7

18 New limit for low-mass WIMPs 90% C.L. optimal interval method (no background subtraction) systematics (efficiency, energy scale, trigger efficiency) This work arxiv: Difference with expectation due to events in T5Z3 8

19 Conclusions First result using the background rejection capability of SuperCDMS 7 izips analysed (threshold.6 kev nr) Exposure: 577 kg day σ SI >. x -6 pb at 8 GeV New limit for WIMPs with masses in the range 4-6 GeV This work (below 4 GeV CDMSlite dominates) arxiv: CoGeNT interpretation of WIMP signal disfavoured in model-independent way CDMS-II (Si) disfavoured assuming standard WIMP interactions and for the standard halo model. High threshold analysis of SuperCDMS ongoing SuperCDMS Soudan detectors are a vast improvement over CDMS II 9

20 SuperCDMS-SNOLAB (with ~0 kg Ge and ~kg Si) will extend the sensitivity by over an order of magnitude with an excellent coverage of the light mass window. Si lite Ge lite Si Ge 0

21 The SuperCDMS collaboration

22 Extra material & Backup slides

23 Limit without T5Z3 The limit extracted without T5Z3 only differs for m> GeV and is approx. a factor better This work This work without T5Z3 3

24 candidates that pass all the cuts 4

25 Efficiencies 5

26 SuperCDMS Soudan 6

27 CDMSlite Amplification of the Luke-Neganov phonons V EQ = N eh ε = IonisaYon energy prompt phonons Luke phonons ε = 3 ev (in Ge) = (e- h) pair creayon mean energy Erecoil = E. of prompt phonons E Luke = V e Neh Ptot = Erecoil + ELuke If the bias voltage is increased, the work done in drifting charge carriers, ELuke, increases The signal is amplified, allowing a substantial reduction in energy threshold and a better energy resolution The phonon instrumentation thus measures the ionization energy (no signal is used from the ionization channels) à No Yield-based discrimination of electron and nuclear recoils 7

28 CDMSlite data taking Data were taken in three periods in 0 with 33 Ba calibration data interspersed throughout There were two neutron exposures ( 5 Cf) to determine energy scale and monitor stability August, and August 3 One izip was used, IT5Z 0.6 kg Selected for its low trigger threshold and low leakage current Signal gain at operating voltage of 69 V was 4 E T = E r ( + ev ) b ε γ Average excitation energy per charge pair ε γ = 3 ev. Raw exposure is 5 live days, 9.6 kg days Optimized based on a flat extrapolation of known electron recoil backgrounds in the -7 kev window 8

29 CDMSlite low-energy spectrum and efficiency Event selection criteria Recoil spectrum of WIMP search events Not coincident with signals in the muon veto Single detector energy depositions 7 evee Not consistent with high frequency noise ( glitches ) Not consistent with low frequency noise (microphonics) Time periods selected with low leakage and where the gain could be well calibrated Exposure =.3 live days = 6.3 kg day The spectrum shows features at.36 kev and.3 kev due to K- and L- shell electron captures in 7 Ge. The energy resolution of the.3 kev line is 43 ev (σ). 9

30 CDMSlite Spectrum of nuclear recoils E nr = E ee ( + ev b ε γ ) ( + ev b ε γ Y (E nr )) Resulting NR spectrum The yield is not measured: use theoretical (Lindhard) model 84 evnr A number of experiments have measured nuclear recoils in germanium over the relevant energy range. 30

31 CDMSlite Results This analysis limits new WIMP parameter space for low masses and rules out portions of CDMS II and CoGeNT contours The choice of a different model for the yield affects the reconstruction of the nuclear recoil energy and thus the interpretation as a limit on the WIMP-nucleon cross-section 3

32 Resulting experimental situation for low-energy WIMPs DAMA CoGeNT (03) CDMS II Si (03) DAMA 3

33 izip discrimination of surface events In the new izips the ionization lines (±V) are interleaved with phonon sensors (0V) on a ~mm pitch Bulk events: charges (e,h) drift to both sides of the crystal e - Surface events: h + charges (e,h) drift to only one side of the crystal e - h + The resulting symmetry/asymmetry in charge collection is an extremely effective discriminant for surface events 33

34 Pb izip calibration ~900 live hours in T3Z with a Pb Source β γ source on side 06Pb 7,55 electrons 6,58 06Pb recoils No events leaking into the signal region (8-5 kev) For 300 kg yr (00 kg Ge SNOLAB) the estimated leakage is < 0.6 events 90% cl 3 γ e-! 5Cf (n)! 06Pb! (a) (b) FIG. 3. (color online) All panels show the same data from 900 live hours of detector T3Z with the (Ionization threshold kevee) (c) Pb source facing side. Clearly 34 visible are the symmetric charge events (large blue dots) in the interior of the crystal, and the events that fail the symmetric charge cut

35 CDMS II 4.6 kg Ge (9 x 40 g). kg Si ( x 6g) 3 Diameter cm Thick SuperCDMS Soudan 9.0 kg Ge (5 x 600g) 3 Diameter.5 cm Thick charge + 4 phonon charge + charge 4 phonon + 4 phonon 35