DAMIC: A direct dark matter search with CCDs Alvaro E. Chavarria, Kavli Institute for Cosmological Physics The University of Chicago 1
Cold dark matter. CCDs as particle detectors. Calibration of nuclear recoil energy scale. SNOLAB installation. Exclusion limit with 2015 data. Radioactivity measurements. DAMIC0 status. Outline Toward a kg-scale CCD experiment. Other applications: Helioscopes for axions, etc. 2
It s great to be back! PRL 8, 051302 (2012) P H Y S I C A L R E V I E W L E T T E R S week ending 3 FEBRUARY 2012 First Evidence of pep Solar Neutrinos by Direct Detection in Borexino G. Bellini, 1 J. Benziger, 2 D. Bick, 3 S. Bonetti, 1 G. Bonfini, 4 D. Bravo, 5 M. Buizza Avanzini, 1 B. Caccianiga, 1 L. Cadonati, 6 F. Calaprice, 7 C. Carraro, 8 P. Cavalcante, 4 A. Chavarria, 7 A. Chepurnov, 9 D. D Angelo, 1 S. Davini, 8, A. Derbin, 11 12 13,4 14 7 4 4 1 15... 2011-2012 3
Dark matter The centripetal force exerted on the Sun cannot be explained by stars and gas. 4
Dark matter Cold dark matter is inferred from cosmology...... and formation of galaxies and clusters of galaxies...... and evident from gravitational lensing. Overall 5.5 times more dark than baryonic matter. Local density ~0.3 GeV cm -3. 5
This is a big problem for particle physics. The dark matter is all around us. It could be made of particles...... that could interact with Standard Model Particles. Look close enough, we might be able to notice its interactions with baryonic matter on Earth. 6
What is the signal? We do not know the mass of the dark matter particles. We do not know by what mechanism the dark matter abundance in the Universe arose. We do not know by what mechanism dark matter interacts with baryonic matter. Without a physics model it is not straightforward to compare constraints from different physical processes e.g. production at colliders, annihilation in dwarf galaxies, direct detection experiments. Very little guidance from theory and other experiments. 7
Signal WIMP: Weakly interacting massive particle is a leading dark matter candidate, σ < 1 nb ( -33 cm 2 ). Event rate from 0 to ~ 4 g -1 d -1 WIMP Lab Speed Distribution Recoil spectrum in Si target 0.0025 ] -1 dp/dv [(km/s) 0.002 0.0015 0.001 0.0005 0 0 0 200 300 400 500 600 700 800 900 WIMP speed [km/s] 8
Strategy Detector with lowest possible energy threshold (especially in nuclear recoil energies). Experiment with large exposure. Up to ton-year. Correspondingly low backgrounds from cosmic rays and natural radioactivity. Measured in dru (events kev -1 kg -1 d -1 ). Multiple targets and technologies to explore more parameter space, avoid unexpected nuclear effects. 9
WIMP-electron cross-section / pb 6 5 4 3 2 1 1 2 3 Strategy Spin-independent elastic 4 WIMP 90% exclusion limits 3 scattering. There are many other possibilities. Dark photon models 2 DAMIC0 XENON Electron recoils 1 1-2 WIMP Mass / GeV c WIMP-nucleon cross-section / pb 6 2 4 8 12 14 4 2 1 6 WIMP 90% exclusion limits ADM models Solar neutrinos 1 1 Will not make it to lab Low-mass search (DAMIC0) High-mass search (LUX) SUSY models 2 DSNB + Atmospheric 3 Nuclear recoils -2 WIMP Mass / GeV c 4 5
WIMP 90% exclusion limits WIMP-nucleon cross-section / pb DAMIC Charge-coupled devices (CCDs) as 1 1 2 3 4 5 6 7 8 low threshold, low background CRESST(2015) particle detectors. WIMP 90% exclusion limits DAMIC1K(2019) CDMSLite(2015) SUPERCDMS(2014) CDMSII-Si(2013) LUX(2015) DAMIC(2016) DAMIC0(2017) 1-2 WIMP Mass / GeV c WIMP mass Will directly probe the possible signal in CDMS II-Si. WIMP-nucleon cross-section 11
DAMIC Collaboration International collaboration: U Chicago, Fermilab, LPNHE-Paris, SNOLAB, U Zurich, Michigan, UNAM, FIUNA, CAB, UFRJ. Physicists with experience from low energy particle physics to cosmology (optical astronomy) to colliders to ultra high-energy cosmic rays. 12
Why CCDs? Lowest threshold ionization detector 50 ev ee. Tried-and-true technology (CCDs used extensively in astronomy since the 80s). Fully active device, no regions of partial charge/signal collection. Scalable: Manufactured at large volumes with standard CMOS technology. Inexpensive (3 k$ per 6 g device). Read-out can be multiplexed. Fabricated in highest-resistivity (purity) silicon in semiconductor fab. Expect very low radioactive contamination in device. Robust, easy to operate (great performance with commercial electronics), easy cryogenics (130 K). Smart investment: Further development of the technology has applications beyond dark matter searches. 13
Charge-coupled device 6 cm CCD Wire bonds Signal cable Copper frame CCDs revolutionized astronomy! 2009 Nobel Prize in Physics. 14
Detector y Charge drifted up and held at gates. pixel z x σxy ~ z 15 y x σxy z x Charged particles produce ionization in CCD bulk. 3.62 ev for e-h pair. Charge collected by each pixel on CCD plane is read out. ~2 e - RMS read-out noise.
CCDs are not triggered The CCD is exposed for a set period of time (usually hours or days). All charge produced in the silicon bulk is collected on the pixel array. There is no hardware energy threshold. Very weak signals (e.g. single electron ionization) can be integrated for long periods to be detectable above read-out noise. 16
CCD readout 3x3 pixels CCD P 1 P 2 P 3 P 1 P 2 P 3 e state H 3 1 H 2 horizontal register serial register H 1 H 3 H 2 H 1 H 3 channel stop channel stop 2 3 H 2...... H 1 sens node P 1 P 2 P 3 P 1 P 2 P 3 P 1 P 2 P 3 ampli er 7 17
Particle tracks 1330 1280 4180 4190 4200 42 4220 1320 50 pixels 13 1300 1290 1280 1290 X-ray? n, WIMP? e α 1300 13 1320 980 960 940 Diffusion limited 6 kev front 6 kev back image 920 5 Front 900 4180 4190 4200 42 4220 5 15 20 25 30 1 2 3 4 5 Energy measured by pixel / kev Energy measured by pixel / kev 18 880 1740 1760 1780 1800 18 5 image Diffusion limited μ 15 20 25 1330 Back 30 image 30 25 20 15
CCD Performance CCDs are manufactured with very high resistivity silicon: Low radioactive backgrounds. Low dark current (< -3 e - / pix / day). Very few defects in the silicon lattice. Distribution of pixel values in image 5 4 30 ks exposure blank 3 2 6.7 evee RMS noise! ±6% 1 0 50 0 150 200 Energy measured in pixel / ev ee >95% of the image good quality. 794 images acquired over 126 days. All good. 19
Reconstructed energy / kev 1 X-rays Calibration data to X-ray lines 55 Fe Si Kα O Kα Al Kα 1 C Kα (0.28 kev) 60 kev 241 Am Ca Kα Energy / kev E resolution: 53 ev at 5.9 kev from front! Fano = 0.13. Depth reconstruction. 20 5 4 3 2 Energy / kev 55 Fe source spectrum in Chicago chamber Noise Si K! Cl K! Mn K escape lines Mn K! Mn K " Cr K! 1 0 1 2 3 4 5 6 7 Energy / kev 7 6.8 6.6 6.4 6.2 6 5.8 5.6 5.4 5.2 Mn K α Front 5 0 0.2 0.4 0.6 0.8 1 1.2 1.4 / pixels σxy = 1.4 from front and back Back σ xy z = 675 μm 200 180 160 140 120 0 80 60 40 20 0
Nuclear recoil calibration a) Cross-section of setup 3 He counter b) 124 Sb- 9 Be source detail Vacuum chamber Table CCD Source 20 cm Lead shielding Table 2.75 cm BeO base BeO cap BeO cylinder Activated antimony rod 24 kev neutrons from 9 Be(γ,n) reaction Use MCNP to simulate neutron production in the source and neutron propagation in the lead. UChicago Simulation has ben validated by rate of 3 He counter in 7 positions around lead shield. 21
Event identification N e (E) Gaus(x, y, µ x,µ y, (z)) Number of ionized electrons Best estimate for mean of energy deposition Lateral spread Use 11x11 pixels moving window and fit to a 2D Gaussian distribution. Register LL of best-fit. Compared to LL of constant pixel values. Difference between the two LL (ΔLL) allows us to select physical events. 22
Data selection Acceptance 6 5 4 3 2 1 1 0.8 0.6 0.4 0.2 ΔLL for events with cdist < 1.5 and 80 60 ev < E < 250 ev 2 χ / ndf 152.2 / 58 0.01 events per Prob 2.175e- image p0 from 1.144e+08 noise ± 1.284e+06 p1 1.318 ± 0.002-0 -90-80 -70-60 -50-40 -30-20 - 0 ΔLL Acceptance for events in bulk 50% at 90 evee Noise From events simulated on real images. 0 0 0.05 0.1 0.15 0.2 0.25 0.3 Measured energy / kev ee -1 Counts from source / ( ev ee) 23 1200 00 800 600 400 200 0 300 250 200 150 0 50 Cr K α Mn K α Fe K α Fe K β 0 5 5.5 6 6.5 7 7.5 Al target (observed) Al target (corrected) BeO target (observed) BeO target (corrected) 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Measured ionization / kev ee Neutrons-on, neutrons-off measurements done by using BeO and Al targets, respectively.
Nuclear recoil spectrum -1 ) ee Recoil signal counts / ( ev 1200 00 800 600 400 200 Data - full BeO Best-fit with Monte Carlo spectrum Single-recoil spectrum very similar to signal from 3 GeV WIMP. End-point = 3.2 kevr χ 2 / ndf 142.5 / 153 Prob 0.718-1 f (0.06) 0.6742 ± 0.0125-1 f (0.3) 1.978 ± 0.019 f'(0) 0.01991 ± 0.01815 f(e ) 0.6059 ± 0.0196 y offset 1.427 ± 0.994 By comparing simulated and observed single recoil spectrum can extract ionization efficiency. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Measured ionization / kev ee As CCD is fully active and there is no multiple scattering, systematic uncertainties are relatively small. Dominated by 9% uncertainty in total predicted rate. 24
Calibration results Ionization energy / kev ee 1 Ionization efficiency in silicon 9 UChicago 124 Sb Be Antonella (systematic) Antonella Gerbier et al (1990) Lindhard, k=0.15 Lindhard, k=0.05 Calibrated entire DAMIC WIMP search ROI -1 0.77 ± 0. kevr at 60 evee 1 Recoil energy / kev r 25
2 km underground DAMIC 26
SNOLAB installation 16 Mpix CCD 5.8 g 6 cm VIB Polyethylene Lead Copper module Kapton signal cable Lead block Kapton signal cable Cu box with CCDs J. Zhou Cu vacuum vessel 27
Event rate at SNOLAB day -1 kg -1 Rate / kev -1 5 4 3 2 1-1 -2 Since 2013 we have decreased the background by > 3. About order of magnitude improvement per year. DAMIC spectrum AlN support AlN frame Si support Pb shield upgrade 222 Rn around vessel (simulation) Now 5 dru 5 15 20 25 Energy / kev 28 Not easy! In the last year: Seven interventions at SNOLAB. Nitrogen purge installation. Improvements in treatment of copper surfaces. Suppression of background from thermal neutron captures in copper. Mitigation of background from condensation e.g. 3 H.
Hardware binning In a CCD we can hardware bin, i.e. put the charge from many pixels into the output node before measuring. Less pixels but same pixel noise! 1x 55 Fe from back: Loss of x, y and 3x3 1x1 Data shows clear z information improvement in energy resolution. 5 6 7 8 Energy [kev] 55 5515 5520 5525 5530 5535 5540 5545 1640 1700 1680 1660 α-β coincidence 1720 1x1 1740 RUNID= 2222, EXTID= 5, cluster_id= 131 0 29 000 20000 30000 40000 50000
Dark matter search Last year (2015) was mostly devoted to decreasing radioactive backgrounds at SNOLAB. However, we took some small amount of DM search data. 0.385 kg d of 1x1 data and 0.234 kg d of 1x0 data. Total radioactive background rate ~30 dru (now 5 dru!). Exclusion limits presented at UCLA Dark Matter Conference in February 2016. 30
Event selection 5 ΔLL distribution for E<250 evee ΔLL distribution for E < 0.25 kev ee and cdist < 1.75 σ xy distributions 4 3 Uniform distribution (simulation) blank exposures 30 ks exposures 8 Front Data Simulation 2 6 4 Back 1 <0.01 events from readout noise -0-90 -80-70 -60-50 -40-30 -20-0 ΔLL 2 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 / pixels σ xy Perform a hard cut on the ΔLL to exclude random fluctuations from noise. Noise leakage <0.01 event in final sample. Evidence from surface events in data. Include in the signal extraction the likelihood that an event is from bulk or surface. 31
Detection efficiency Detection 1x1, efficiency Likelihood vs. depth Acceptance after noise rejection cut Depth / µm 600 500 1.2 1 Acceptance 1 0.8 Simulated 400 300 200 0.8 0.6 0.4 0.6 0.4 1x0 1x1 0 1x1 0.2 0.2 160 evee Simulated depth / µm 0 0 0.05 Acceptance 0.1 0.15 0.2 for 0.25 1x0, 0.3 0.353σ-seed 0.4 0.45 0.5 Energy / kev 600 500 400 300 0 1.2 1 0.8 0.6 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Energy / kev ee 70 evee Sensitive to events deeper in the bulk at low energies. 200 0 1x0 0 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.4 0.2 0 Less pixels read out: Less background from readout noise. Simulated energy / kev ee 32
Bulk spectra d -1 kg -1 60 50 0-1 kev ee kg -1 d -1 60 50 1-7 kev ee Compton scattering Data Event rate / kev -1 40 30 20 Event rate / kev -1 16 dru 40 30 20 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Energy / kev ee 0 1 2 3 4 5 6 7 Energy / kev ee Unbinned likelihood fit to 1x1 and 1x0 data done independently, combined in a single exclusion limit. Null (background-only) hypothesis consistent with both data sets. 33
Exclusion limits WIMP-nucleon cross-section / pb 1 1 2 3 4 5 6 7 CRESST(2015) 52 kg-d WIMP 90% exclusion limits CDMSLite(2015) 70 kg-d, not calibrated SUPERCDMS(2014) DAMA/Na(2009) LUX(2015) DAMIC(2016) 0.62 kg-d CDMSII-Si(2013) 8 1-2 WIMP Mass / GeV c 34
Radioactive contamination in CCD The ultimate sensitivity of CCD dark matter search will be determined by the intrinsic radioactivity of the devices. We have placed strong limits on the radioactive contamination of detector-grade silicon. Strategies include α spectroscopy and spatial coincidence searches. Could have applications beyond dark matter searches. 35
α particles σ x /σ y N pix 2 29 days 3 CCDs Plasma (675 µm) Plasma (500 µm) Bloomed (675 µm) Bloomed (675 µm) 2 Po (surface) 0 2 4 6 8 12 14 16 Energy / MeV α-β discrimination based on shape of track. Bound box Plasma (back or bulk) Bloomed (front) Limits on contamination: 238 U < 5 kg -1 d -1 = 4 ppt 4 2 232 Th < 15 kg -1 d -1 = 43 ppt 36
α-β coincidences RUNID= 2222, EXTID= 5, cluster_id= 131 1740 50000 β Bragg peak (end) β Bragg peak (end) 19 RUNID= 716, EXTID= 4, cluster_id= 421 25000 1720 40000 1900 20000 1700 30000 1890 15000 1880 Back 000 β emerges 1680 20000 1870 1860 Plasma α from decay 5000 point under α. 1660 Front 000 8280 8290 8300 83 8320 0 1640 Bloomed α 0 55 5515 5520 5525 5530 5535 5540 5545 37
β-β coincidences 64 kev 1.2 MeV 2 Pb 2 Bi 2 Po τ1/2 = 5 d 0.22 MeV 1.7 MeV 32 Si 32 P 32 S τ1/2 = 14 d 57 days of data in 1 CCD: 2 Pb < 37 kg -1 d -1 (95% C.L.) +1-65 32 Si = 80 kg -1 d -1 (95% C.L.) 32 Si - 32 P candidate 1360 1355 1350 1345 1340 1335 1330 30000 25000 20000 15000 000 5000 0 Decay E1 = 114.5 kev (xo, yo) Δt = 35 days Cluster #79 point E2 = 328.0 kev 38
DAMIC0 Already achieved radioactive background and low-noise performance for 0 g detector. Stack of 18, 16 Mpix CCDs in current SNOLAB vacuum vessel and shielding. Currently packaging CCDs. Installed 8 modules (48 g) last week. Installation of remaining modules in ~1 month. First results by the end of this year. 39
DAMIC0 Installation CCD in module with electroformed copper by PNNL. Seven (five shown) CCDs in commercial copper modules. Carefully machined and cleaned. Now inner 5 cm of ancient lead shielding. 40
DAMIC1K Fifty 20 g CCDs. Current CMOS manufacturing technology. Goal to be limited by 32 Si background after coincidence veto at 0.01 dru. 0.5 e- noise with new on-chip amplifiers and digital filtering. New vacuum vessel in current shield. Aim for results before 2020. Electroformed copper by PNNL for DAMIC testing. 41
Development of CMOS manufacturing technology on thicker wafers. Selection of sources of raw silica with low 32 Si content. Design of read-out circuits optimized for different applications. E.g. Skipper CCD for lower noise, lower energy threshold detectors. Goal: single electron/ photon detection. Beyond Great opportunities for collaboration with industry! 42
Nuclear recoils WIMP-nucleon cross-section / pb 1 1 2 3 4 5 6 7 8 CRESST(2015) 52 kg-d WIMP 90% exclusion limits DAMIC1K(2019) 300 kg-d 0.01 dru, 0.5 e - noise CDMSLite(2015) - 70 kg-d SUPERCDMS(2014) CDMSII-Si(2013) LUX(2015) DAMIC(2016) 0.62 kg-d DAMIC0(2017) 30 kg-d 1-2 WIMP Mass / GeV c 43
Electron recoils WIMP-electron cross-section / pb 6 5 4 3 2 1 1 2 3 4 WIMP 90% exclusion limits DAMIC1K(2019) 300 kg-d 0.01 dru, 0.5 e - noise 3 XENON S2-ONLY (2012) 2 44 DAMIC0(2017) 30 kg-d 1-21 WIMP Mass / GeV c
Axion-like particles Ae g 9 ALP CDM exclusion limits DAMIC0(2017) 11 12 DFSZ Axion models CoGeNT(2008) XENON0(2014) Red giant 13 KSVZ DAMIC1K(2019) 14 1 1 2 3 4 ALP mass / ev 5 c -2 45
Other Applications A single photon detector CCD with higher quantum efficiency in the infrared will have transformative applications in astronomy. Precise spatial localization and identification of radioactive particulates: Non-destructive nuclear forensics + low radioactivity physics. On-going collaboration with PNNL. Lowest background X-ray detectors for next generation axion helioscopes (e.g. IAXO). Detection of ν-nucleus coherent scattering from nuclear reactors. Trackers to search for fractionally charged particles (q~0.01 e - ) from particle colliders. 46
Conclusions CCDs are very low radioactivity, low threshold particle detectors whose response to ionizing radiation has been thoroughly characterized. Different particles can be identified, localized and tracked. Very useful for signal identification and background rejection. DAMIC0 will start taking dark matter search data in the next month. Results by end of 2016. DAMIC1K to be developed in the next three years. Expect to lead the Low-mass WIMP field in both nuclear and electron recoil channels. 47