Analysis of vacuum and argon gas fill data from the MiniCLEAN dark matter experiment

Analysis of Vacuum and Argon Gas Fill Data from

the MiniCLEAN Dark Matter Experiment

by

Stephen H. Jaditz

Submitted to the Department of Physics

in partial fulfillment of the requirements for the degree of

w

Doctor of Philosophy

at the

MASSACHUSETTS INSTITUTE OF TECHNOLOGY

February 2015

Massachusetts Institute of Technology 2015. All rights reserved.

Author ..

Certified by.

..

Signature redacted

Department of Physics

January 16, 2015

Signature redacted

Joseph A. Formaggio

Associate Professor of Physics

Thesis Supervisor

Signature redacted

Accepted by

Krishna Rajagopal

Associate Department Head for Education

Analysis of Vacuum and Argon Gas Fill Data from the

MiniCLEAN Dark Matter Experiment

by

Stephen H. Jaditz

Submitted to the Department of Physics on January 16, 2015, in partial fulfillment of the

requirements for the degree of Doctor of Philosophy

Abstract

The existence of particle dark matter provides a consistent framework for understand-ing many astronomical observations. The rotation curves of galaxies and galaxy-clusters, for example, indicate the majority of mass in these structures is unseen. The existence of weakly-interacting massive particles (WIMPs) was proposed in the early 1980s to account for the anomalous rotation curves and provide a mechanism for producing the cold dark matter relic density, which along with dark energy is thought to dominate the current energy density of our universe.

Efforts to observe the rare interaction of WIMPs with normal matter have contin-ued since their proposal, and so far have set limits on the WIMP-nucleon interaction cross-section extending to 1 x 10-9 pb. Contemporary experiments seek to observe ~ one WIMP-nucleus scatter per year per 100 kg of detector mass. These experiments must be conducted deep underground with stringent cleanliness requirements.

The MiniCLEAN dark matter experiment is a single-phase liquid argon scintil-lation detector which uses the wavelength-shifting fluor tetraphenyl butadiene and cryogenic photomultiplier tubes for light detection. The active spherical region of the detector contains 500 kg of liquid argon at temperature 87 K. Background events which could mimic a WIMP signal are mitigated through pulse-shape discrimination and position reconstruction.

At an intermediate stage of ongoing detector assembly 2 km underground at SNO-LAB in Ontario, the complete instrumented inner vessel was commissioned by col-lecting photomultiplier waveform data for periods when the vessel was evacuated and when filled with warm argon gas. Alpha decay events from radon progeny on the wavelength-shifting surface occur in this data at a measured rate of 19.0 t 0.4 /h/m2_.

MiniCLEAN's projected sensitivity to spin-independent WIMP-nucleon scattering, derived from simulation of this surface rate, is -si < 1.5 x 10-8 pb.

Thesis Supervisor: Joseph A. Formaggio Title: Associate Professor of Physics

Contents

1 Direct Dark Matter Detection

1.1 Introduction . . . . 1.2 Dark Matter Halo Parameters . . . . 1.3 Basics of Terrestrial Scattering . . . .

1.4 Setting Scattering Cross-section Limits . . . .

1.5 Comparing Results of Different Experiments . 1.6 Detector Technologies . . . .

2 The MiniCLEAN Detector

2.1 Scintillation in Noble Liquids . . . . 2.1.1 Light detection: tetraphenyl butadiene 2.2 MiniCLEAN Design . . . . 2.2.1 Assembly of inner vessel . . . . 2.2.2 Gas processing system . . . . 2.2.3 Data acquisition . . . . 3 Radioactive Backgrounds 3.1 External Backgrounds . . . . 3.2 Internal Backgrounds . . . . 3.2.1 39A r . . . . 3.2.2 Gammas . . . . 3.2.3 Fast neutrons . . . . 3.2.4 Radon progeny on wavelength shifter .

31 . . . . 31 . . . . 35 . . . . 36 . . . . 38 . . . . 38 . . . . 39 47 48 50 53 56 57 61 63 . . . . 63 . . . . 64 . . . . 64 . . . . 66 . . . . 67 . . . . 69

3.3 Sum m ary . . . .

4 Simulation and Analysis Software 73

4.1 Likelihood Method for Particle Identification . . . . 75

4.2 Likelihood Method for Position Reconstruction . . . . 76

5 The MiniCLEAN Veto System 79 5.1 O verview . . . . 79

5.2 H ardw are . . . . 79

5.2.1 PMT mounts . . . . 79

5.2.2 PMT testing . . . . 83

5.2.3 Electrical connections and data acquisition . . . . 85

5.3 Sim ulation . . . . 89

5.3.1 Cosmic muons . . . . 89

5.3.2 Cosmogenic neutrons . . . . 98

5.3.3 Summary of Cosmogenic Backgrounds . . . . 100

6 Analysis of MiniCLEAN Vacuum Data 103 6.1 PMT and DAQ configuration . . . . 103

6.2 Data quality . . . . 104

6.3 2 2_{Na calibration source . . . 104}

6.4 Vacuum data . . . 105

7 Analysis of MiniCLEAN Gas Data 119 7.1 Run Summaries . . . 119

7.2 Triplet Quenching Overview . . . 122

7.3 22Na calibration source . . . . 125

7.4 Instrument Effects . . . . 128

7.4.1 Instrument Effects Summary . . . . 134

7.5 Surface alphas . . . . 135

7.5.1 Surface alpha rate . . . . 138

7.5.2 Light yield . . . . 149

8 Sensitivity Outlook and Conclusion 151

8.1 Conclusion ... ... 158

A Select Topics in Detector Design 161 A.1 Validation of photomultiplier and base . . . 161

A.2 Selection of light guide reflector . . . . 171

A.3 Acrylic length and PMT neutron simulation . . . 174

List of Figures

1-1 Mean velocities in the plane of the galaxy, as a function of linear

dis-tance from the nucleus for 21 Sc (spiral) galaxies, arranged according to increasing linear radius. From [5]. None of the curves fall off at large radii, contrary to the expectation for galaxies comprised of visible mass. 32

1-2 A composite optical, X-ray (red), and weak lensing (blue) image of the

Bullet Cluster (lE 0657-558) [10]. . . . . 33

1-3 From [24]: the rotation curve for the Milky Way for values of Ro =

7.1 kpc, v, = 185 km - s- , and RO = 8.5 kpc, vc = 220 km - s-I, where

RO is the distance to the galactic center and v, is the circular velocity

of our solar system about the galactic center. The figure also shows one of the ways in which the rotation curve can be decomposed into the contributions from different mass components: the bulge (dotted line);

the stellar disc (filled circles); the H1 layer (crosses, where negative

values mean that the force is directed outwards); the H2 layer (circles);

and the dark halo (dashed line). The best-fitting model, which is ob-tained by summing the individual components in quadrature, is shown

1-4 Solid lines: measured spin-independent WIMP-nucleon cross-section limits. Non-solid (dashed and dotted) lines: projected limits from fu-ture measurements. Shaded regions with solid-line delimiting: signal claims. Shaded regions with no delimiting: predictions for theoreti-cal dark matter candidates. The dual-phase liquid argon experiment DarkSide-50, for which a projected limit is shown in this figure, recently obtained a limit of 6.1 x 10-" cm2 _{for a WIMP mass of 100 GeV [51].}

45 1-5 Limits for SI and SD WIMP-nucleon cross-section from the LHC

ex-perim ent CM S [20]. . . . . 45

2-1 Photograph of the assembled MiniCLEAN inner vessel at SNOLAB. . 47

2-2 Measured Leff _{for liquid xenon as a function of nuclear recoil energy,}

from [63]. The open circles represent that work's measurement; other markers are from previous measurements and cited in [63]. The solid line is from a best-fit analysis of XENON10 AmBe source data and Monte Carlo, and the dashed line in the theoretical prediction of Hi-tachi [68]. . . . . 50

2-3 Measurement of Leff for liquid argon as a function of nuclear recoil energy, made with the micro-CLEAN detector [69]. The rapid rise below 20 keVr is from trigger effects and not physical. . . . . 51

2-4 Measurement of Leff for liquid neon as a function of nuclear recoil en-ergy, from [70]. Solid lines represent fits to quenching models based on different ion stopping-power models from SRIM [71] and the reference [66] of M ei et al . . . . 51

2-5 Normalized visible reemission spectrum from a TPB film illuminated with various UV wavelengths. From [72]. . . . . 52

2-6 Photon emission time probability density functions for electronic and nuclear recoils in liquid argon and alpha scintillation in TPB . . . . . 52

2-7 (Left) SolidWorks model of the MiniCLEAN detector by Jeff Griego (LANL). The liquid argon or neon (LAr, LNe) is contained in the cold inner vessel (radius 75 cm), which is housed by the evacuated outer vessel. One of the 92 optical modules is depicted on the right. .... 53

2-8 Components of an optical module. (Left) An 8-inch Hamamatsu R5912-02MOD PMT attached to a tophat. (Center) An irregular-hexagon type light guide. The guide is shown before electropolishing and with-out the reflective inner surface, which is 3M ESR foil. (Right) An irregular-hexagon acrylic plug, illuminated on the far side with a UV LED so that the blue fluoresence of the TPB can be seen . . . . 54

2-9 Model of the central detector housed in the muon veto, a 7.9 m tall by 2.8 m radius water tank instrumented with 42 8-inch Hamamatsu R1408 PMTs. Figure by Jeff Griego. . . . . 55

2-10 (Left) Photograph of the Cube Hall in SNOLAB, looking down onto the water tanks for MiniCLEAN (right) and DEAP3600 (left). The right photograph is taken from the bottom of the MiniCLEAN water tank, looking up to the (empty) outer vessel and beyond to the Cube Hall and yellow gantry crane. . . . . 55

2-11 (Left) Photograph from the top of the Cryopit, looking down 50 feet onto the temporary soft-wall cleanroom used for IV assembly. Detector components are in bags stacked along the right rear wall. (Right) The IV in the soft-wall cleanroom with some spools and their clear polycarbonate covers attached . . . . 56

2-12 (Left) Photograph looking into an IV port during the "dry-fit" of the light guides, when each guide was inserted to check for fit and to fix its orientation for final installation, before being removed again for as-sembly with reflective foil and the acrylic plug. Note the guides do not abut along their edges, but have several millimeter gaps through which scintillation light could penetrate the central detector from out-side the wavelength-shifting radius. Minimization of these gap sizes motivated tight tolerances on IV dimensions, which was a driving fac-tor in fabrication cost. (Right) Photograph looking into the IV during final cassette insertion. The TPB-coated acrylic faces are outlined by reflective flaps which cover the gaps between guides. The flaps were made by creasing the reflective foil lining the light guide walls. ... 58

2-13 Photograph of the assembled IV in the softwall cleanroom, with DAQ racks and gas bottles. . . . . 59

2-14 Diagram of the MiniCLEAN gas system for purification of warm argon

gas... ... 59

2-15 Photograph of the partial purification system. The SAES getter is the grey box on the right, with the multi-meter sitting atop. To the right of the getter is the charcoal trap submerged in the dewar of ethanol slu rry. . . . . 60

2-16 SolidWorks model of the cold head attachments from James Nikkel (RHUL). The cold head is housed in a nitrogen-filled canister attached to the OV, with helium lines extending out to the compressor on deck. The cold "finger" extends into the insulation vacuum space, where it

is bonded by copper braids to cold fingers which penetrate the IV. . . 60

2-17 Diagram (from Boston University) and picture of the DAQ configura-tion for warm argon gas data. This configuraconfigura-tion does not incorporate the veto system, which will be present for liquid argon data taking. . 62

3-1 (Top) Comparison of DEAP-1 22Na data and the collaboration's ana-lytic model in the region 120-240 photoelectrons (approximately 43-86

keVee). (Bottom) P1cak distribution for the same data, showing the

an-alytic model with and without PMT noise included. The lower curve shows the expected backgrounds from high-Fp events, mostly cosmo-genic neutrons. For 50% nuclear recoil acceptance, the statistical model with noise projects pulse shape discrimination better than parts in 10 billion. Figures from [78]. . . . . 65 3-2 Intensity of prominent gamma lines in the 2 38U and 232_{Th decay chains}

(assuming equilibrium), derived from a listing given in a project pro-posal by the Majorana neutrinoless double-beta decay experiment [81]. 67

3-3 (a,n) neutron energy spectrum, for 0.103 ppm 238_{U and 0.170 ppm}

232_{Th contamination in borosilicate glass. Generated using the online}

tool http://neutronyield.usd.edu, based on [83]. . . . . 68

3-4 (Left) Scenarios for the surface decay of 210_{Po, depicted as a black} dot, to an alpha particle and 201Pb nucleus, which could mimic a dark matter signal. (Right) Example of the fiducial volume cut on surface events. ... ... 69

3-5 From [84]: an example radon exposure requirement to achieve a desired surface activity of 0.1 and 1.0 a decay per m2 per day due to 210Po on acrylic. . . . . 70

4-1 Cutaway view of the inner MiniCLEAN detector as rendered by the RAT simulation. The PMTs, outer vessel, and water tank are included in the geometry but not shown. . . . . 73

4-2 (Left) Simulated differential leakage of electron events obtained using the discrimination variables L, and Fp. (Right) Integral leakage of 39Ar events as a function of energy threshold into a 150 kg fiducial volume. From [87]. . . . . 75

4-3 (Left) Diagram showing 0. (Right) Example results for E(ij), the average probability for the generation of a photoelectron in a PMT at location i after a simulated energy deposition, indexed by

j,

4-4 Average reconstruction resolution along the X-axis for uniformly dis-tributed 20 keVee events at radii ranging from the center of the detec-tor to the TPB surface. The resolution is defined to be the sigma of a Gaussian fit to the difference between true and reconstructed position.

From the November 2010 MiniCLEAN proposal to the DOE [88]. . . 77

5-1 PMT mount drawing. . . . . 80

5-2 PM T string drawing. . . . . 81

5-3 Veto PMT shown in mount. . . . . 81

5-4 A complete string of four veto PMTs, suspended from a crane at Bates lab. Pictured from left to right are the author, Bates engineer Jim Kelsey, and then MIT post-doc Kim Palladino . . . . 82

5-5 Dark rate as a function of time at 2100 V for PMT PHSS. Dark pulses are those which descend 3 mV below baseline. . . . . 83

5-6 Single photoelectron charge spectrum for PHSS at 2250 V. The fit model is given in equation 5.1. . . . . 84 5-7 Dark rate and gain for PHSS as a function of voltage. Pulses counted

as dark are those whose charges exceed 1.5 PE. The vertical line indi-cates the operating voltage for this PMT determined at SNO, which produced gain 1.6 pC. The SNO-measured dark rate is higher because

a 1/4 PE threshold was used. . . . . 85

5-8 Summary of the measured operating voltages for the veto PMTs. The operating voltages are chosen to produce 1.6 pC gain. . . . . 86

5-9 Ratio of measured gain to the gain measured by SNO, at SNO oper-ating voltage. The gain has degraded over time. The two tubes with

ratio ~ 1.4 have blown back-termination resistors, which causes the

gain to double. . . . . 87

5-10 TRIUMF connector assembly drawing. . . . . 87

5-11 TRIUMF connector for mating cable to the R1408. From top to bot-tom is shown the coupling nut, the plug body into which the cable is inserted, the teflon insulator which is also inserted into the plug body, and the pin with a stripped cable. . . . . 88

5-12 Block diagram of veto electronics. . . . . 88

5-13 Muon flux as a function of azimuthal angle of incidence, from param-eterization in [92] referenced from [77], and cross-sectional area of the veto water tank. Integrating the product of the two curves gives the result 9.8 0.3 muons per day incident on the veto. . . . . 90

5-14 dE/dx for muons in silicon dioxide, from RAT simulation and the PDG. 91 5-15 Number of photoelectrons created in veto PMTs vs length of water

traversed by muon. The black line is an estimate of the total number of Cherenkov photons produced by the muon. The red line is an estimate of the number PEs generated in the PMTs, made using a rule-of-thumb for typical PMTs given in the PDG Review of Particle Physics [14], and the approximate 1% photocathode coverage of the veto tank surface by the PMTs. The red line does not take reflections into account, which is why it lies below the bulk of the points. . . . . 92

5-16 Number of hit veto PMTs vs length of water traversed by muon. In the top plot, greater than 1/4 of the average single PE anode charge was required in order for a PMT to be registered as hit. The bottom plot shows the result for a threshold of 4 PE. . . . . 93

5-17 Number of photoelectrons per veto PMT per event and time distribu-tion of photoelectron creadistribu-tion relative to the first photoelectron cre-ation per PMT. The yellow line is a fit of a sum of two exponentials to the data... ... 94

5-18 Efficiency for muon tagging vs veto PMT NHit requirement, for various thresholds. As examples: requiring NHit > 2 and threshold 1/4 PE results in efficiency greater than 0.999 and 3 missed muons per year; an NHit > 5 and threshold 4 PE results in efficiency 0.97 and 107 missed m uons per year. . . . . 95

5-19 Location of closest approach for muons unvetoed with trigger require-ments NHit < 10 and threshold

Q

5-20 Distance of closest approach, or impact parameter, of unvetoed muons for different trigger requirements. The impact parameter decreases with more stringent trigger requirements. . . . . 96

5-21 Radial coordinate of location of closest approach for an unvetoed muon, with different trigger requirements. Although the impact parameter de-creases with more stringent trigger requirement, the radial coordinate remains close to the radius of the tank. . . . . 97

5-22 Left: Simplified cavern geometry, showing cavern walls and the Mini-CLEAN veto. Right: Neutron origins, looking at cavern from East to W est. . . . . 99

5-23 cos(O) distributions for cosmogenic-neutron trajectories: (red) for all neutrons (including those which do not hit the veto) with respect to parent-muon trajectory; (blue) for all neutrons with respect to vertical; and (green) for neutrons incident on the veto with respect to vertical. 0 = 0 points along the muon trajectory for (red), and downward for (blue) and (green). (red) is drawn from [77] Eq. 16. Note the peak at 0 = 7r/2 for (green). The origin of this peak is the sudden increase in the volume of possible neutron origins in the cavern floor as the

azimuthal angle of incidence on the veto descends past horizontal. . . 100

5-24 Argon recoil energy distribution due to elastic scatters with cosmogenic neutrons within fiducial radius (29.5 cm), for qucnching factor 0.25. The number of scatters per year with 20-100 keVee is 0.080 0.002 per year. The plot shows 76 years worth of simulated neutrons. . . . . 101

6-1 A typical event collected in vacuum data. The top trace is the sum of the nine PMTs which fire during the event. The integrated charge is 61pC, or 12PE... 105

6-2 (a) Photoelectron, (b) Fp, and (c) charge centroid distributions for vacuum runs 475, 483, and 488, normalized to counts per hour. . . . . 110

6-3 (a) Fp vs PE, (b) Fp vs R3/R3PB, and (c) R3/R3PB vs PE distributions

for vacuum runs 475, 485, and 488. . . . . 111

6-4 Reconstructed angular coordinates of vacuum data. Peaks are evident at the coordinates of the cassettes. PMT 53, at cos(O) = -0.2,

#

6-5 Distribution of PMTs with maximum charge per event. PMTs 0, 11-15, 61-65, and 91 are the pentagons, which have the smallest TPB surface area of the three types of cassettes. . . . . 112 6-6 PE and R3_{/R3PB distributions for simulated gammas overlaid on data}

6-7 (a) R3/R}PB and (b) F, vs R3/4Rp for vacuum events with greater than 80 PE. The high energy events are strongly peaked at the TPB radius and fall into two F, bands, one above and one below Fp = 0.6. 114 6-8 Reconstructed azimuthal and polar angles for vacuum runs with "Na

source present. There is an excess of events at the position of the source in the upper right corner. . . . . 115

6-9 Gammas with origin at the PMTs simulated at 400 Hz and 100 kHz. The 100 kHz source does not cause pileup significant enough to shift the PE spectrum . . . . 115

6-10 (a) Radial and (b) PE distributions for simulated bursts of 1000-1300 photons with origin on the surface of the reflective foil. The group of events with ~ 90 PE originate along the edge of the acrylic block, causing a lower light yield than events originating closer to the PMT, away from the acrylic. . . . . 116

6-11 The measured centroid distribution for vacuum data. The distribution is fit with a third order polynomial between 0.85 and 0.92. The extent of the polynomial beyond 0.92 defines the projected contributions of gammas (in white) and ESR scintillation events (in blue). The blue region contains 1.4 million events, corresponding to 720 events per hour

per square meter of ESR foil. The total area of the foil is 18.5 m2 117

7-1 A typical gas event, with 132 PE. The top panel shows waveform traces for all PMTs with charge. The bottom panel is the event display, where the area of a polygon corresponds to the charge in that PMT. .... 120

7-2 PE, Fp, R3/RTPB distributions for the first hour of gas run 281,

com-pared with vacuum data. . . . . 121 7-3 Photoelectron and Fp distributions for the first five hours of gas runs

281 -301. The average PE value decreases with time, while the average F, value for scintillation events increases. . . . . 122

7-4 Mean total, late, and prompt photoelectron values for events with Fp < 0.6 during the first five hours of runs 281- 301. . . . . 123

7-5 Sum of PMT waveforms with Fp < 0.6 collected during the first 11 minutes of run 281, normalized to unit integral. The triplet time con-stant is 1.58 ps, determined by fitting with equation 7.1 . . . . 124 7-6 Decay of the triplet time constant with time. . . . . 124

7-7 PE, Fp, and R3

/R B distributions for tagged 22Na events, overlaid on events anti-coincident with the tag. The tagged and untagged dis-tributions are very similar, indicating there was a problem with the

tag... ... 126

7-8 Reconstructed charge centroid coordinates for data taken when the

22_{Na source was present and not present. Distributions are normalized}

to unity. The source is located in a divot in the inner vessel, drilled at

(cos(0) = 0.85, # 2.1). . . . . 127

7-9 An afterpulsing event with 1000 PE, Fp = 0.08, R3/R}PB = 0.979.

Top: all PMT waveforms. Left: the summed waveform on top, with the brightest PMT (WFD 9, Ch. 5) below. The afterpulse occurs ~1 us after the primary pulse. Right: event display with brightest PMT in center foreground. . . . . 129

7-10 PMTs with greatest charge for high-energy, low-Fp events. PMT 2 was prone to discharge. . . . . 130

7-11 A PMT discharge event with 735 PE, Fp = 0.12, R3/RPB = 0.996. Top: all PMT waveforms. Left: the summed waveform on top, with the brightest PMT (WFD 1, Channel 2) below. Both summed and

in-dividual waveforms have amplitude - -500 mV. Right: event display

7-12 An baseline sag event with 5286 PE, Fp = 0.13, R3/RP =0.004. Top: all PMT waveforms. The baseline sag occurs around 6000 ns, with subsequent rebound after 8000 ns. Left: the summed waveform on top, with the brightest PMT (WFD 7, Ch. 4) below. Right: event display with brightest PMT in center foreground. . . . . 133

7-13 Fp vs R3/R3PB distribution for events with more than 500 PE. The cluster of events with Fp < 0.3 and R3/4PB = 1 occur outside the TPB radius and induce one of the instrumental effects described in the text. ... ... 134

7-14 Fp vs PE distributions for the first six hours of runs 281-301. The group of events circled in the top left panel migrate to higher Fp and lower PE values with time, due to the increasing water vapor content

in the argon gas which quenches the late triplet scintillation light. . . 136

7-15 An alpha event with 2568 PE, F = 0.37, R3

/R3B = 0.55. Top: all PMT waveforms. Notice the high occupancy. Left: the summed waveform on top, with the brightest PMT (WFD 4, Channel 1) be-low. The summed waveform has amplitude -6000 mV, the individual PMT -1200 mV. Right: event display with brightest PMT in center

foreground. PMTs with greater charge have larger area polygons. . . 137

7-16 (Top) R3/4PB vs PE distribution for 2000 simulated alpha events

which deposit 5.3 MeV in the argon gas along the surface of the TPB. The argon scintillation yield has been set to 1800 photons per MeV in order to match the PE yield of 1500 observed during the first hour of run 281 (see Fig. 7-19). All the simulated events reconstruct away

from the TPB surface with R3/RiPB < 0.7. Of the 2000 simulated

events, 1960 cause an event trigger, and 35 generate less than 800 PEs. (Bottom) Projection onto the PE axis. The cluster of events near 600 PE originate near or in front of the baffle, which absorbs the UV scintillation light. . . . . 139

7-17 Distributions of Fp values for alpha events which satisfy the centroid and PE cuts, for several hours of runs 281-301. The fit function is a gaussian plus a constant. The instrumental effects described in the text cause the events to fall outside the central Fp value. . . . . 140 7-18 Summary of the fitted mean and variance values for the Fp distributions

shown in Figure 7-17. The mean variance, averaged over is 0.024. . . 141

7-19 Photoelectron distributions for events with R3_{/R}rPB > 0.96 during the} first three hours of run 281. Each panel contains at least one hour of live time, and is labeled by the time since the beginning of the gas fill. The vertical lines represent the position of the "by-eye" cut described in section 7.5.1. The solid circles are surface alpha events which induce an instrument-effect, identified by their Fp value which falls outside a range defined in Figure 7-18. . . . . 142

7-20 Photoelectron distributions several hours after the gas fill. The hour labels are the time since the fill. . . . . 143

7-21 Average photoelectron yield for alpha events which do not induce an instrument effect, fit with a sum of three exponentials. . . . . 144

7-22 Photoelectron cut. The solid line is the alpha PE yield multiplied by half and is used to determine the alpha rate. The dashed line was used to select alpha events to accept for the PE yield determination. The dashed line is a fit to the points, which were identified by-eye for several PE spectra where there was an obvious gap between the falling gamma spectrum and alpha accumulation, for example the first two hours of Figures 7-19. . . . . 145

7-23 The photoelectron distribution for vacuum events, and the integral of this distribution above the PE cut of Figure 7-22. . . . . 146

7-24 Surface alpha rates. (Top) Runs 106-115, started Jan. 29 2014, with 17.1 hours livetime. (Middle) Runs 215-232, started Feb. 14, with 63.2 hours livetime. (Bottom) Runs 281-301, started Feb. 28, with 53.6 hours livetime. The livetime-weighted mean rate for the three runs is 42.9 0.7 events per hour. . . . . 147

7-25 Distribution of PMTs with greatest charge for alpha scintillation events. The exposure time for each cassette during assembly is represented by the blue line, with an overall normalization determined by the integral of the measured rate. . . . . 148 7-26 Photograph taken during assembly of the inner vessel, looking into the

vessel before cassette installation was complete. . . . 148 7-27 Argon gas light yield for alpha emission following neutron capture

on boron film, from [98]. The emitted alpha has energy of 2.79 or 2.31 MeV, with branching ratios of 6% and 94%, respectively. The mea-surement does not extend to MiniCLEAN's 180 kPa pressure. However, the plateau beyond 60 kPa suggests the light yield remains constant at 5600 photons per decay above the limit of the measurement. This corresponds to 2400 photons per MeV of energy deposition. . . . . 150

8-1 Schematic of 21OPo decays to alpha particle and 206_{Pb near the}

wavelength-shifting surface, drawn to approximate scale. The measured alpha rate, represented on the left, receives contributions from parent nuclei on the TPB-argon interface, in the TPB, on the TPB-acrylic interface, and

up to - 10 pm depth in the acrylic. Class-I and II events can mimic a

dark m atter signal. . . . 151 8-2 Reconstructed radius distribution for simulated class-I surface events,

displaying the effects of several cuts. . . . . 154 8-3 (Left) Event origins along the TPB surface for simulated type-I events.

8-4 Reconstructed radius using the Shellfit algorithm vs the charge centroid m ethod . . . 155

8-5 PE threshold required to ensure zero-background fiducial mass, for various class-I type surface rates. . . . . 156

8-6 Number of class-I surface events per year which leak into the WIMP ROI with 150 kg fiducial volume. . . . . 157

8-7 Number of class-II surface events per year which leak into the WIMP ROI with 150 kg fiducial volume. . . . . 157

8-8 Projected 90% confidence SI WIMP-nucleon cross-section limits for MiniCLEAN. All except the red curve assume 6 PE/keV. All except the green curve assume 150 kg fiducial volume. The blue and red curves assume a PE threshold of 120; the black-dashed curve assumes 95 PE threshold and 15 background events. . . . . 159

A-1 Signal response of PMTs vs temperature from [106]. The R1221 and R649 (triangles) have multialkali photocathodes, while the rest have bialkali photocathodes and exhibit steep response degradation below about - 100C . . . . 162

A-2 Photograph of several 8-inch Hamamatsu photomultipliers. The two tubes on the left have lost vacuum, which causes evaporation of the brown photocathode and the tubes' clear appearance compared to the SNO (R1408) tube. The increased opacity of the platinum-coated tube can be seen in the duller appearance of the dead R5912-02MOD com-pared to the R5912. MiniCLEAN uses the frosted variety of R5912-02M O D . . . . 163

A-3 MiniCLEAN R5912-02MOD base schematic. Positive bias voltage (- 1100 V at TLAr) is applied across the PMT lead to ground.

Sig-nal appears as negative pulses

(

(red) back termination. Ringing seen in the red trace is damped by the back term ination. . . . . 165

A-5 Diagram and picture of cold gas cryostat for PMT testing. The pic-ture shows the closed setup, with only the outer vessel of the dewar visible. The PMT is housed in the stainless inner vessel where it can be covered by nitrogen or neon gas. Low-mass cable from Gore is used for electrical connections inside the dewar, while RG58 provides the rest of the connection to the DAQ (CAEN V1720 digitizer) and power supply (HV). Light from a YAG laser is sent to the PMT with optical

fiber. Not depicted in the diagram is the piping for gas handling. . . 166

A-6 (Left) Single photoelectron charge spectrum for PMT 26 at 950 V (top) and 1000 V (bottom), at room temperature. (Right) Gain and noise measurements vs voltage for PMT 108 at 298 K (top) and 90 K (bot-tom). The bias voltages which produce 5 pC gain are 1043 V and 1031 V 167 A-7 Gain and noise results vs temperature for PMT. The top panel shows

the operating voltage required to obtain 5 pC gain. The middle panel shows the rate of noise pulses with charge greater than 3/4 single pho-toelectron, at the operating voltage from the top panel. The bottom panel shows the gain at 950 V operating voltage. . . . . 168

A-8 Time separation between dark pulses. The non-Poisson distribution, with a peak near time zero, suggests a mechanism other than thermal emission causes dark pulses at cryogenic temperatures. . . . . 169

A-9 Examples of single photoelectron pulses at liquid argon temperature. The left (right) panels are classified as DLN (TLN). An ADC count corresponds to 2V/212 _{= 4.88mV. . . . . 169}

A-10 Reduced y2 _{distributions for Eqn. A.1 fit to PMT calibration data} taken at liquid argon temperature. Each pulse is classified as DLN, TLN, or multiple PE (upper right). . . . . 170

A-11 Pulse time distributions for the different catagories of pulses, weighted by their relative probability. . . . . 170

A-12 Charge distributions for the different categories of pulses, weighted by their relative probability. . . . . 170

A-13 Simulated light collection vs reflectivity of light guide walls, based on 10 eV betas at detector center. . . . . 171

A-14 Measured reflectivities of coatings from two companies, JDecker In-dustries (red) and Silvex (blue). These are calibrated measurements performed by Angstrom Sun Technologies (Acton, MA) using a Varian Cary-500 spectrometer and, for total reflectivity, an integrating sphere. The silver-based coating from JDecker is proprietary; the Silvex sample is plated with silver per ASTM B700 Type 2 Grade C Class S. . .. . 172

A-15 Measured reflectivities at several angles of JDecker coating, 3M ESR film, and an Al/SiO front-face mirror. These uncalibrated measure-ments were made with a Cary-500 and VASRA attachment for angular manipulation. The result is that the JDecker coating performs better than the ESR and Al/SiO at all angles. The lack of calibrated reference mirror prevents a quantitative statement from being made. . . . . 173

A-16 Example determination of o for thin acrylic and energy 100-105 pho-toelectrons. (Left) F, for simulated electron and WIMP events, using 100 ns. (Right) Contamination and difference between median values of FP6 and Fpx, as a function of . The left (right) arrow indi-cates the value of which minimizes electronic recoil contamination (maximizes the difference between median Fp values). o is chosen to minimize contamination, and for this energy window takes the value 140 ns. ... ... 175

A-17 o for thin acrylic as a function of energy expressed in photoelectrons. The average value 135 ns minimizes contamination. . . . . 176

A-18 PMT neutron background as a function of reconstructed photoelectron number, before cuts, then with FI > 0.7 and R < 29.5 cm cuts. After cuts, 60.0 and 26.5 neutrons per year reconstruct in the energy region of interest 75-150 PEs for thin (5mm) and 10 cm acrylic. The results are normalized to 40000 neutrons per year generated in the PMT glass. 176

A-19 (Red) Number of neutrons per year which reconstruct in the energy region of interest 75-150 PEs after F, and R cuts, as a function of acrylic length. (Blue) Percentage light loss relative to thin acrylic as a function of acrylic length. The collaboration chose to use 10 cm length acrylic, coincident by chance with the cross on this plot. . . . . 177

A-20 (Top) Jeff Griego's (LANL) mechanical drawing of the spool from two perspectives, showing a standard conflat on the left and inverted conflat on the right. (Middle) Photograph of an IV port's inverted conflat, without a spool attached. The inside of the IV can be seen through the port. (Bottom) Close-up photograph of spool attached to IV. In the foreground (toward the outside edges of the photo) is shown a standard conflat, with bolt-holes outside the radius of the knife-edge, where the tophat is to be attached. The black heads of bolts fastening the spool to the IV via the inverted conflat are in the photo's middle ground. The pictures were taken at Winchester during test-assembly for pressure vessel certification. . . . . 180 A-21 Cassette test stand schematic (top). This stand was conceived to

benchmark the optical properties of a full cassette in liquid argon, with light guide, reflector, PMT, and TPB-coated acrylic. The schematic shows the configuration used to purify gaseous argon and condense it with liquid nitrogen providing cooling via a small condensing canister

attached to the main vessel. Fine temperature control of the condenser is accomplished by pressurizing with nitrogen gas. During the inverted seal test, assembled as depicted on the lower right, helium was substi-tuted for argon, and a helium leak sniffer was attached to the vacuum shroud... ... ... .... ... . . ... . 181 A-22 (Left) Alistair Butcher (RHUL) cleaning bolt holes. The blue and

white wipes were used to keep track of which holes had been marginally cleaned. (Right) Comparison of bolts before and after electropolishing. 182 A-23 X-ray fluorescence measurement of tape lifts from an Inconel bolt with

scaling and from an electropolished Inconel bolt. The listed metals are all present in Inconel 718 alloy. The scaling is likely high-temperature oxidation which occurred during annealing. . . . . 182

List of Tables

1.1 Spin values for some relevant nuclides. . . . . 40

2.1 Scintillation parameters for three noble liquids. Results for the prompt and late time constants, Leff, and F. are from the cited references. The other parameters are assembled in [65]. Below the given lower electron-equivalent energy bounds, the measurements for electronic and nuclear

Fp tend toward equal central values. The light yields for Fp

measure-ments in neon and argon were 3.5 and 4.85 photoelectrons per keVee, respectively. (*) Measurement from [64] was made with fission fragments. 49 3.1 Gamma fluxes from norite, measured during the installation of SNO

with a NaI(Tl) detector and various thicknesses of lead. "The calcu-lations are based on neutron capture in the elements of norite with neutron flux predicted from the measured Th and U concentrations in the rock." From the SNOLAB User's Handbook, [76]. . . . . 64 3.2 Summary of 23U, 232_{Th, and} 40_{K gamma properties. . . . . 66}

3.3 Gammas generated per year by the PMT glass and steel of the inner and outer vessels. Assays reported in mBq/kg, from [82]. . . . . 68 3.4 Summary of background sources for MiniCLEAN and their reduction

via energy, fiducial volume, Fp, and F, cuts, derived from simulation and tabulated in the internal document [85]. . . . . 71

5.1 Values for the parameters in equation 5.4 derived from measurements of muon flux at several underground sites. . . . . 89

Chapter 1

Direct Dark Matter Detection

1.1

Introduction

We hold that the movement of celestial bodies is governed by the law of gravitation, developed first by Isaac Newton in the 17th century and modified by Albert Einstein in the early 20th. The power of physical laws derives from their utility in making predictions. A dramatic success story of Newtonian gravity, among many, was the discovery in 1846 of the planet Neptune [1]. Urbain Joseph Le Verrier inferred Nep-tune's position from its pull on the planet Uranus, whose orbit about the Sun was known to deviate slightly from the Newtonian expectation. Le Verrier sent his pre-diction for Neptune's location to Johann Gottfried Galle at the Berlin Observatory, who immediately observed the planet. Encouraged by his success, Le Verrier later proposed the existence of another planet, Vulcan, to explain the anomalous precession of Mercury's orbit. Vulcan, however, was never observed. Mercury's orbit was incor-porated into our understanding by Einstein, who derived its anomalous precession as a consequence of his modified law of gravitation, the general theory of relativity, in

1916 [2].

In the 1930s Fritz Zwicky uncovered another tension between theory and observa-tion, this time beyond our solar system: the galaxies in the Coma Cluster orbited each other at velocities too high to be gravitationally bound

[3].

1970s and 80s [4, 5]. Figure 1-1, from Rubin, Ford, and Thonnard's influential paper of 1980 [5], depicts the rotation curves for 21 galaxies with large spirals (Sc galaxies). Rotational velocity is determined by measuring the redshift of Hot-lines, which have extreme red wavelength. The measured line-of-sight velocity is converted to rota-tional velocity by projection onto the galactic plane, using the assumption that the emission regions are moving in planar circular orbits about the center of their galaxy. Remarkably, none of rotation curves fall-off with distance from the galactic center, even at the faint outer extent of the optical images. The expectation for galaxies comprised of visible mass is that the rotation curve falls off at large distances.

0 5 10 15 200 NGC 4605 II NGC 1035 - NGC 4062 NGC 2742 NGC 701 C- NGC 2608 NGC 3495 q NGC 1087 UGC 3691 NGC 482 -I 10 20 30 40 kpc NGC 3672 NOC 1421 NGC 4321 NGC 2715 _,-I IC 467 NGC 7541 21 Sc GALAXIES 100 km s-NGC 7664 NGC 2998 NGC 753 0 10 20 30 40 50 0 60100 120

DISTANCE FROM NUCLEUS (kpc) EH- 50 km s' Mpc9

Figure 1-1: Mean velocities in the plane of the galaxy, as a function of linear distance from the nucleus for 21 Sc (spiral) galaxies, arranged according to increasing linear radius. From [5]. None of the curves fall off at large radii, contrary to the expectation for galaxies comprised of visible mass.

200 100 0 E 0 LIJ z -J z 0 LIi 200 100 0 NGC 801 - UGC 2885 1.A

Following Le Verrier in his prediction of a massive, unseen planet, the implication taken by Rubin et al. in [5] was that "Sc galaxies of all luminosities must have significant mass located beyond the optical image." However, modified theories of gravity were proposed, with modified force-laws [6, 7] or dynamics [8], to account for the anomaly. The notion that most of the mass in a galaxy is invisible can be disquieting, and Rubin herself came to express aesthetic preference for a modified theory of gravity in 2005 [9].

Figure 1-2: A composite optical, X-ray (red), and weak lensing (blue) image of the

Bullet Cluster (lE 0657-558) [10].

Compelling evidence for the existence of non-luminous, weakly-interacting matter came in 2006 with gravitational lensing observations of the Bullet Cluster [11]. Figure 1-2 shows the now-famous composite image of the two galaxy clusters which collided 100 million years ago. The image shows that the massive parts of the two clusters continued along their original trajectories after the collision, while their x-ray emitting gas clouds were violently deformed. The decoupling of the majority of the mass (in blue) from the baryonic matter (in red) implies that the bulk of the matter in the clusters interacts weakly with itself and the known particle constituents of the gas

clouds. Here is direct evidence for the existence of unidentified "dark matter", which makes up most of the mass of galaxies.

The presence of stable, particle dark matter in large quantities fits well with our current understanding of the development of the universe from the Big Bang. A cos-mological model must account for the structure of the cosmic microwave background; the abundance of the light elements hydrogen, helium, deuterium, and lithium; the distribution of galaxies on large scales; and the accelerating expansion of the universe. The six-parameter ACDM model is the simplest successful model. It is based upon a spatially-flat, expanding universe with dynamics governed by general relativity and constituents dominated by cold dark matter (CDM) and a cosmological constant (A) at the current time. The parameters of the model, including the dark energy, bary-onic matter, and dark matter density of the universe, can be extracted from analysis of the cosmic microwave background (CMB) temperature, which has been measured by three generations of space missions since the early 90s. Most recently, the Planck Collaboration [12] has found that matter constitutes 31.5-1.% of the energy density of the universe, with (84.5) 2.4% of this matter required to be non-relativistic (cold), pressureless, and non-interacting (dark).

There is an overwhelming body of evidence that the dark matter required by cosmology cannot be a particle of the Standard Model

[13,

Despite the success of the dark matter model in providing a consistent framework for understanding diverse astrophysical observations, all the widely-accepted evidence for its existence is gravitational, and its particle nature remains unknown. Efforts to characterize dark matter particles fall into three categories: indirect detection

experiments which aim to detect the products of dark matter annihilation or decay; dark matter production at accelerator facilities such as the Large Hadron Collider; and direct detection of dark matter scattering in dedicated, underground detectors. Dark matter annihilation or decay to high-energy photons at the center of our galaxy is a promising route to indirect detection, and recent analysis of data from the Fermi telescope indicates there may indeed an excess of such photons coming from the galactic center [18]. Another method of indirect detection is to observe the effect of dark matter annihilation early in the history of the universe on the CMB temperature and polarization anisotropies [19]. Efforts in this direction have not produced a positive signal [19].

Searches at the LHC, where the signature is large missing transverse momentum, have to date been negative [20], and are setting interaction cross-section limits compa-rable to contemporary direct detection experiments (see Figure 1-5). Direct searches are the topic of the following sections, which survey direct detection principles and techniques. Then in Chapter 2 comes description of the MiniCLEAN dark matter detector that is the subject of this paper.

1.2

Dark Matter Halo Parameters

Dark matter preserved the primordial fluctuations in cosmological density on galactic scales that were wiped out in baryonic matter by viscosity, when radiation decoupled from baryons hundreds of thousands of years after the big bang. Gravitational collapse produces dark matter halos, which provide most of the gravitation for formation of stable structures in the universe. Unlike the baryonic components of galaxies, which are supported from radial collapse by angular momentum, dark halos are supported by random velocity which serves as collisionless pressure. The simplest halo model is spherical and isothermal, with a Maxwellian velocity distribution for the constituent WIMPs.

For consistent interpretation of experimental results, most direct detection experi-ments assume the simplest halo model for our galaxy, following the standard reference

of Lewin and Smith [21]. The halo parameters of interest are those that determine the dark matter energy distribution and flux onto a terrestrial detector: the local dark matter density PD, the dispersion velocity of dark matter vo, and the escape velocity

vesc above which the dark matter velocity distribution is truncated. A halo density

which falls off as r-2 (where r is the radial coordinate with origin at the galaxy cen-ter) generates the flat rotation curve observed in many spiral galaxies. Estimates for

PD based on this spherical density profile and the Milky Way rotation curve (Figure 1-3) have been in the range 0.2 GeV < PD < 0.4 GeV, leading to adoption by many

dark matter experiments of the central value PD 0.3 GeV. Use of more realistic halo models which include flattening, or structures like rings in the galactic plane, can cause PD to change by as much as a factor of four (for example, [22]).

For an isothermal, isotropic halo in hydrostatic equilibrium and with the Maxwellian dark matter velocity distribution f(v) eM/v, Drukier et al. [23] use the equation0

of hydrostatic balance to show that vo = vc, the circular velocity of the solar sys-tem about the galaxy center. The calculation assumes vo and v, are independent of galactic radius. Lewin and Smith [21] use vo = 230 km -s-' 20 km -s-1.

Drukier et al. [23] note that v.c is bounded below by the highest observed stellar

velocity, 583 km - s-1, and estimate a local upper bound of 625 km - s-1. The standard

value chosen by Lewin and Smith [21] is v... = 600 km -s-1.

1.3

Basics of Terrestrial Scattering

A WIMP of mass M. and kinetic energy E. which scatters elastically from a nucleus with mass MA deposits energy

Er = 2E MAMX (1 - cos(6)) (1.1)

S(MA + MX) 2

where E, and 0 are the energy and scattering angle of the struck nucleus in the center-of-momentum frame. Neglecting the motion of our solar system through the dark matter halo, the recoil spectrum of the stuck nucleus is obtained by folding the

300 200 200 0 X 0 R,=8 5 pc 200 E 4 1 0 _{e g os1 2} Galactaocentric Radius [ kpc]

Figure 1-3: From [24]: the rotation curve for the Milky Way for values of R0 7.1 kpc,

oc=185 kmn - s, and R0 8.5 kpc, ve 22Okm -s- 1, where R0 is the distance

to the galactic center and 'vc is the circular velocity of our solar system about the galactic center. The figure also shows one of the ways in which the rotation curve can be decomposed into the contributions from different mass components: the bulge

(dotted line); the stellar disc (filled circles); the H1 layer (crosses, where negative

values mean that the force is directed outwards); the H2 layer (circles); and the

dark halo (dashed line). The best-fitting model, which is obtained by summing the individual components in quadrature, is shown as a full line.

Maxwellian kinetic energy distribution with equation 1.1. This gives an exponentially

falling recoil spectrum with the form dR/dE oc e-Er/Eor, where the average recoil

energy E0 = M0vo and

r =4M MA / (My +i MA)2 (1.2) As typical targets have atomic masses of several tens of GeV, detectors with keV energy threshold are required. Another difficulty is the low rate of interaction, which may be cast conveniently normalized:

_ NOP-D (1.3)

A M,

405 (GeV PD (_____ oo counts

A k Mx ) 0.3 GeV - cm- (230 km- s-1 pb kg - day

where A is the atomic number of the target, No the Avogadro number, and go the 'zero-momentum transfer' dark matter interaction cross-section per nucleus.1

1.4

Setting Scattering Cross-section Limits

Experimental efforts to detect WIMPs center on reducing background events in a detector large enough to register the rare low-energy nuclear scattering predicted by equation 1.4. A WIMP of mass 1 GeV and weak-scale cross-section ao=1 pb would interact several times per day per kg of detector mass. When detectors fail to register any WIMP scattering, an upper limit on ao may be set as a function of Mx. The first dark matter experiments (of the late 1980s) set upper limits for co near 10-' pb, while current limits extend to 10-9 pb (see section 1.6). In other words, contemporary experiments seek to register about one dark matter scattering event per 100 kg per year of exposure. Sources for such low-energy scattering events abound, and must be carefully eliminated by reducing radioactive content within the detector and shielding of external radiation incident on the detector. The characteristic nuclear scattering of WIMPs must also typically be discriminated from more common electronic recoils.

1.5

Comparing Results of Different Experiments

In order to compare the limits obtained with different detector materials, a model for WIMP-nucleus scattering is needed from extensions of the Standard Model. In these extensions, the standard coupling of a WIMP proceeds via a scalar current to the mass of a nucleus (known as the spin-independent interaction, SI), or via an axial vector current to the spin of a nucleus (spin-dependent interaction, SD). The

1oo _{omits suppression of the interaction rate due to the form-factor of the target nucleus. The}

expression for R omits corrections accounting for detector response, including energy detection efficiency, resolution, and threshold. These corrections are detailed in the standard reference of

distinction generates a natural division between SI detectors using high-A nuclei and SD detectors using nuclei with unpaired nucleon spins. Groups conducting SI searches typically report an upper limit on a WIMP-nucleon cross-section o-ule, which, for a target nucleus with atomic number A, amounts to [17, 25]:

2

2 A) = 00/ A2 _(1.5)

where PA and pnre are the reduced masses of the WIMP-nucleus and WIMP-nucleon systems. For experiments with more than one type of nucleus as target, the total cross-section a-i is obtained by averaging across all targets, obtaining 1/o-e

The normalization procedure for SD limits is more complicated. An SD exper-iment reports limits on WIMP-proton or WIMP-neutron cross-sections oSD. This requires input of experimentally-determined values for the proton and/or neutron

spin expectation values (Sp,n) for the target nuclide(s). The theoretical models for

SD scattering are more varied than for SI, leading to further complexity. One method by Giuliani [26] that has come into use2 is to calculate, for each sensitive nuclide in the target molecule, one of the following cross-sections:

os(A)

3

/ J +1 (S,,n) (1.6)

where u,,n are the reduced masses of the WIMP-proton and WIMP-nucleon systems, and J is the total nuclear spin. The total cross-section may then be reported as in the

SI case: 1/o,2 = ZA [1/ SD(A)]. Reference spin values from [26] for some relevant

nuclides are given in Table 1.1.

1.6

Detector Technologies

Dark matter searches have utilized three types of signatures to detect nuclear scatter-ing: ionization, scintillation, and heat. A number of experiments use a combination

2

Nucleus Z J (Sp) (Sn) 19F 9 1/2 0.441 -0.109 23_{Na 11 3/2 0.248 0.020} 27_{Al 13 5/2 0.343 0.030} 3 5_{C1 17 3/2 -0.059 0.011} 37_{CI 17 3/2 -0.178} 0 127I _{53 5/2 0.309 0.075} 129_{Xe 54 1/2 .0.028} 0.359

Table 1.1: Spin values for some relevant nuclides.

of two of these, to discriminate electronic from nuclear recoils.

Ionization: The first dark matter detectors used germanium diodes, carefully fab-ricated for low radioactivity, at liquid nitrogen temperature. No discrimination of electronic vs nuclear recoils can be done with Ge diodes. The low-energy thresh-old is set by microphonic and electronic noise. Crystal masses for these detectors ranged from about 250 g to 3 kg. Typical thresholds were a few keV. The event rates at threshold were a few counts/keV/kg/day. Notable experiments using Ge diodes were the collaborations at Homestake (1987[30]) and Oroville (1988[31]), the

Hei-delberg/Moscow group (1994[32], 1998[33]), and the IGEX collaboration (2000[34]).

These groups obtained upper limits on oa in the range of 10-4 to 10- pb for

Mx = 60 GeV. More recently, the CoGeNT collaboration deployed a 440 g P-type

point-contact Ge detector at the Soudan Underground Laboratory. CoGeNT had

a 2keV threshold and used pulse shape for discrimination against events occurring near the edges of the crystal. CoGeNT reported an excess of events over expected backgrounds near threshold, and an annual modulation of the excess consistent with

ic ~1 x 10-4 pb for Mx = 10 GeV [35]. The modulation was subsequently found to be an order of magnitude too high for realistic halo models, and inconsistent with the lack of modulation seen in CDMS-II [36].

In another effort to exploit the ionization signature, several groups are pursuing low-pressure gas time-projection chamber (TPC) technology as a means to measure the energy and direction of nuclear recoils. This is motivated by the observation that WIMP-induced nuclear recoils should be preferentially opposed to the direction of the