Gaitskell
“Whither WIMPS?”
Results from Cryogenic Dark MatterSearch (CDMS) Experiment
Rick Gaitskell
Center for Particle AstrophysicsUC Berkeley
source at http://cdms.berkeley.edu/gaitskell/
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Summary
• Clean determination of nuclear recoil event (NR) rate in detectorsu 10.6 kg-day exposure of Ge
u Detectors specifically designed for WIMP direct detection search• Energy resolution better than 1 keV(recoil) FWHM• NR vs γ discrimination is very clean
• WIMP Detection Resultsu Summarise statistical analysis
• Outline assumptions made.
— Residual systematic errors smaller than Poisson fluctuations
• Can CDMS interpretation be wrong?
u Compare sensitivity to signal implied by DAMA annual modulation• DAMA and CDMS results are incompatible for scalar (SI) interactions
— (σ ~ A2)— DAMA ann. mod. signal predicts 40 events during Ge 10.6 kg-day
– (actually observe 13 NR single scatters)
• Conclusions are robust (i.e. even if CDMS does not rely on neutron hypothesis)
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Nuclear Recoil Discrimination - Event by Event
• Nuclear recoils arise from
uu WIMPsWIMPsu Neutrons
• Electron Recoils arise fromu photons
u electrons
u alphas
(Typical Background)
• Ionization yieldu ionization/recoil energy strongly
dependent on type of recoil
• Recoil energyu Phonons give full recoil energy
•
Neutrons (external source)
Gammas (external source)
Phonon Trigger Threshold
NOT A SIMULATION!1334 gamma events, 616 neutron events
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Run Overview
• CDMS-Iu This presentation will cover a full analysis of event data that
was taken as part of Cryogenic Dark Matter Searchexperiment based at the Stanford Underground Facility(*)
u 1998 (2 months)• 33 live days 100 g Si ZIP -> 1.6 kg-days after cuts
• (4 nuclear recoil events observed)
u 1999 (12 months)• 96 live days 4x165 g Ge BLIP -> 10.6 kg-days after cuts
• (17 nuclear recoil events observed)
(*) Used to be called Stanford Low Radioactivity Facility but the radiation safety paperwork became too onerous
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Ge BLIP Ionization & Phonon Detectors
NTD Gethermistors
ionizationmeasurementcircuit
ionizationcollectionelectrodes
lockinamplifier
165 g p-type Ge
QinnerQouter
(electronics not shown)
Vpb
Rb
Vqb
• Four 165 g Ge detectors, for total massof 0.66 kg during 1999 Run
• Calorimetric measurement of total energy
• ENERGY Resolution• σ = Ionisation 220 eV, Phonons 250 eV
Inner Ionization Inner Ionization ElectrodeElectrode
Outer IonizationOuter IonizationElectrodeElectrode
Passive Ge shieldingPassive Ge shielding
(NTD Phonon Sensors on underside)
TowerTower•• WiringWiring•• heat sinkingheat sinking•• holds coldholds cold FETs FETs for for
amplifiersamplifiers
BLIP
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Detector Environment
Stanford Underground Facility• 17 mwe of rock• hadronic component down by >1000• muon flux down by ~5
Low-Background Environment• 25 cm polyethylene reduces muon-
induced neutron flux from rock andlead by factor >100
• 15 cm Pb reduces photon flux byfactor >1000
• radiopure cold volume (10 kg)• additional internal (ancient) lead
shielding
Active Scintillator Muon Veto• muon veto >99.9% efficient• reject ~22 “internal” neutrons/ day
produced by muons within shield polyethyleneouter moderator
detectors inner Pbshield
dilutionrefrigerator
Iceboxouter Pb shieldscintillator
veto
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Photograph of Shielding
Neutron moderatorPb shielding
Copper cold volume
Neutron moderator
Scintillator muon veto
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Radiopure Cold Volume (“Icebox”)
Cold electronics
To internallead shield
anddetectors
Cutaways of lids ofconcentric copper cans
Electronics stem forsignal cables
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
4 x BLIP Tower Schematic (1999 Run)
• 1999 Run (net 10.6 kg-days)
• Low-radioactivity Cu and Ge housing
• Self-shielding through close packingu Spacing 3 mm vertically
Ge shielding
BLIP 4
BLIP 3
BLIP 5
BLIP 6
60 mm
60 mmOuter IonizationElectrodes• less shielded from electrons(surface events)
• imperfect ionization yield(electrode break)
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Anticoincident Ionisation Energy/Recoil Energy Plots
• Events over 10.6 kg-days,anticoincident with muonveto and other detectors
• Gamma Band
• Surface Electron Band
• Nuclear Recoil Bands(dashed lines indicate 90%acceptance region)
• B3 is clearly contaminated(discard in subsequentanalysis)
B6
B4B3
B5
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
60 mm
BLIP 4
BLIP 3
BLIP 5
BLIP 6
Contamination
Contamination
4 x BLIP Tower Schematic (1999 Run)
• Self shielding through close pack - spacing 3.5 mm vertically
Fiducial Region - contains ~0.25 kg of Ge
Outer IonizationElectrodes
•less shielded fromelectron-inducedsurface events•imperfect ionizationyield (particularly forevents in top-bottombreak)
60 mm
Ge shielding
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
1999 Run Ge BLIP Data Set Combined
all single-scattersNR candidates
Entire 96 live days operation GeBLIPs = 10.6 kg-days
u Gamma and electron bands wellseparated from NR band
• NR candidates are truly NR’s
u See a total of 13 events > 10 keVÜ ~ 1.2 events/kg/day
NR Band asymmetric (-3σ,+1.28σ)= 90% efficient
Expect 40 events for scalarWIMP σ given by the DAMAAnn Mod Signal
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Nuclear Recoil Events
• 13 single NR ~1/3 that expected for the DAMA Ann Mod
• However, we have strong evidence that these events are causedby neutrons
u 4 multiple scatter nuclear recoils observed in Ge in same data set (**)
u 4 nuclear recoil events seen during 1.6 kg-days Si data
Observation of Ge multiple/Observation of Ge multiple/Si Si single events is consistentsingle events is consistentwith all single nuclear recoil events being due to neutronswith all single nuclear recoil events being due to neutrons
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Neutron Multiple Scatters
• Observe 4 neutron multiple scattersin 10-100 keV multiple events
u 3 neighbors, 1 non-neighbor
• Calibration indicates negligiblecontamination by electron multiples
Ioni
zatio
n Yi
eld
B6
Ionization Yield B4
photons
neutronneutrons
Ioni
zatio
n Yi
eld
B5,
6
Ionization Yield B4,5
surfaceelectrons photons
Neighborinteraction
B4
B3
B5
B6
Non-Neighborinteraction
Neighbors Non-Neighbors
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Muon-Coincident & Calibration Neutrons
• Agreement between MC and data is good - no free parametersu (i) Simulate neutrons generated in Pb/Cu shielding by muons
u (ii) Simulate neutrons from Am/Be source on top of Pb shield (poly out)
Singles/multiples ratios & shape in MC match data well ~15% syst much smaller than statistical fluctuation in final result
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Consistency of Neutron Interpretation with MC
• Predicted ratios of numbers of events set by Monte Carlo simulations
• Ge multiples and Si singles imply large expected neutron background with largestatistical uncertainty
• Likelihood-ratio test indicates we should expect worse agreement 6% of the time
• Energy spectra consistent with expectations for neutrons (also with WIMPs)
Nuclear Recoil Events
4
13 16
1
3.4
Data w/ 68% confidence intervalPrediction based on Ge mult, SiPredictions based on most likely
+
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
• Follow ‘Unified Approach’ of Feldman and Cousins: ordering bylikelihood ratio R with parameters constrained to lie inside thephysical region:
u Large value of R(M, ) means the observed data is likely for the model (M, )
u Use Monte Carlo simulation to determine critical parameter R90% (which issmaller than 90% of simulated R’s). A model (M, ) is rejected by the data atthe 90% confidence level if R
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
mean over 2-6 keVee(22 – 66 keV recoil)
DAMA 2000 paper Figure 2
DAMA 15,000 kg-day
DAMA 215,000 kg-day
DAMA 3 + 438,000 kg-day
New CDMS Limit
CDMS 2000(90% CL)
• CDMS Limit (Published PRL17June2000) incompatiblewith DAMA (4σ) annualmodulation signal at 99.98%
MinimumDAMA NaI/1-4 (3σ)
Best fit to Ann Moddata alone
Best fit to Ann Moddata alone
Best FitDAMA NaI/1-4
DAMA 1996(90% CL)
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Combined Likelihood CDMS limit & DAMA signal
DAMA (3 year)58,000 kg-days
NaI
Ann. Mod. Signal (fit to Fig 2)m = 50 GeV
σ = 14.4 10-42 cm2
+Background limit in2-3 keV binm = 52 GeV
σ = 7.2 10-42 cm2CDMS 1999
10.6 kg-daysGe
= 40 = 202.3σ away from best
fit to Ann. ModGe singles +
Ge multiples +Si singles
≤ 8(90% CL)
99.98% 99.8%CDMS excludes all points in3σ (~99% CL) region at 75%
Ignore Gemultiples 99.93% 96.8%
Ge singles only(No Subtraction)
NWIMPs = 13≤ 19
(90% CL)99.5% 59%
•Expected sensitivity for CDMS with multiples subtraction (when exposure would see 1 multiplerather than 4, and 16 singles rather than 13) raises NWIMPs by 50%: ≤ 12 (90% CL)
•Combined likelihoods based on CDMS calculated fit to ann. mod. & an inferred likelihood for datawhen constraint of 2-3 keV bin included (DAMA actual likelihood function not yet made available)
NWIMPs is number of WIMPs in 10.6 kg-days Ge in CDMS run (Er>10 keV)
000919.3.rjg
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Compatibility of CDMS and DAMA
• CDMS resultsincompatible withDAMA Figure 2 data(left) at > 99.98% CL
• Estimate full DAMAlikelihood function:
u Two experiments areincompatible at 99.8%CL
u Ignore multiplescatters: 96.8% CL
CDMS dataBest simultaneous fitto CDMS and DAMApredicts too littleannual modulation inDAMA, too manyevents in CDMS
predictedspectrum
DAMA 2000 paper Figure 2
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Can CDMS be wrong? (1)
• The detector technology is too new to trust, inadequate technical write upu Yes this will be true of any new detector - we do have to convince you of
performance
u Within collaboration (25 active members) handling of systematics has beenextremely thorough
u Calibrations have been discussed
u PRD in preparation / Thesis (pp 500) also made available
• Completeness of Error Analysisu Because 13 Ge singles, 4 multiples, 4 Si Singles statistical fluctuations dominate
completely over systematic errors in final exclusion limit analysis
• Nuclear Recoil Efficiencyu Determined in situ using calibration-source neutrons; comparison to the
simulation of neutrons from this source indicates this efficiency is accurate to <20%. The stability of this efficiency is demonstrated by the veto-coincidentneutron rate, which is stable to 15%, consistent with statistical fluctuations, andagrees with the simulation of these neutrons to < 20%.
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
CDMS: Summary (I) and New Experiment (II)
• CDMS I probing supersymmetry regionu Results for scalar-interacting (σ~A2) WIMPs
(CDMS: 10.6 kg-day Ge) are incompatible with DAMA(58,000 kg-day NaI) signal at very high confidence
u PRL published / PRD to follow
u Will release additional analysis: 40% increase exposure (larger fiducial volume)
u Signal due to external neutron background from muons in rock
• 2000 Run underway at SUF shallow siteu 3 Ge and 3 Si detectors - better background subtraction
u Additional neutron moderator to cut background by ~2.3x
• CDMS II full funding (99-05) approved by NSF & DOE
R. Abusaidi et al., Phys. Rev. Lett. 84, 5699 (17June2000) astro-ph/0002471PRD in preparation + Thesis Golwala
000904.2.rjg
http://cdms.berkeley.eduhttp://dmtools.berkeley.edu
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Inner Ionization Electrode
Increasing the 1999 Run BLIP Fiducial Volume
•Results I show today are based on the more restrictive cut --keep only events fully contained in inner ionization electrode
•Events near inner-outer gap haveionization energyshared between innerand outer electrodes
TopView
Region ofSharedEvents
Outer IonizationElectrode
•Internal multiple scattersalso appear as shared events
•Including "shared" regionwill increase exposure by~40% for WIMPs, by ~60% forneutrons.
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Direct Detection: History & Future
Oroville (88)[m=20 GeV]
Homestake (87)
Heidelberg-Moscow (94)
Heidelberg-Moscow (98)
[m = ?? GeV - if significantlybetter limit obtained atdifferent mass]
DAMA (96)
UKDMC (96)[m=100 GeV]
DAMA (98/00)
90% CL Limit on Cross section for 60 GeV WIMP (scalar coupling)
CDMS SUF (T)
CDMS Soudan (T) 7 kg Ge+Si Cryodet
GENINO (T) 100 kg Ge Diode
GENIUS (T)100 kg Ge Diode
CryoArray (T)0.1-1 tonne Cryodet
~1 event kg-1 day-1
~1 event kg-1 yr-1
~1 event 100 kg-1 yr-1
LHC
Not meant to be a complete list - see http://dmtools.berkeley.edu
Different Different ColoursColoursIndicate DifferentIndicate Different
TechnologiesTechnologies
NOWNOWCDMS SUF (99)CDMS SUF (00)
Edelweiss (98)
000904.1.rjg
GeNaICryodet
(T) TargetSignal
Liq Xe
CDMS Gaitskell IDM2000 September 2000 Rick Gaitskell
Answers toAnswers toTricky Questions ….Tricky Questions ….