Improved WIMP-search reach of the CDMS II germanium data

R. Agnese et al. (SuperCDMS Collaboration)

Phys. Rev. D 92, 072003 – Published 12 October 2015

Abstract

CDMS II data from the five-tower runs at the Soudan Underground Laboratory were reprocessed with an improved charge-pulse fitting algorithm. Two new analysis techniques to reject surface-event backgrounds were applied to the 612 kg days germanium-detector weakly interacting massive particle (WIMP)-search exposure. An extended analysis was also completed by decreasing the 10 keV analysis threshold to $\sim 5 keV$ , to increase sensitivity near a WIMP mass of $8 GeV / c^{2}$ . After unblinding, there were zero candidate events above a deposited energy of 10 keV and six events in the lower-threshold analysis. This yielded minimum WIMP-nucleon spin-independent scattering cross-section limits of $1.8 \times 10^{- 44}$ and $1.18 \times 10^{- 41}$ at 90% confidence for 60 and $8.6 GeV / c^{2}$ WIMPs, respectively. This improves the previous CDMS II result by a factor of 2.4 (2.7) for 60 $(8.6) GeV / c^{2}$ WIMPs.

8 More

Received 23 April 2015

DOI:https://doi.org/10.1103/PhysRevD.92.072003

Authors & Affiliations

Click to Expand

Article Text (Subscription Required)

Click to Expand

References (Subscription Required)

Click to Expand

Issue

Vol. 92, Iss. 7 — 1 October 2015

Reuse & Permissions

Access Options

Article Available via CHORUS

Download Accepted Manuscript

Author publication services for translation and copyediting assistance advertisement

Images

Figure 1
Isometric representation of the CDMS II detector arrangement and tower occupation, with direction of north indicated. The ZIP detectors in each tower are numbered 1–6 and are either silicon (yellow) or germanium (brown). The identification numbers of each detector were given to track the raw material the detectors were produced from and the processing to which they were subjected. For example the detector “G9” is a germanium detector with identification number 9. Green stars indicate detectors that were used in this data analysis (see Sec. 2).
Reuse & Permissions
Figure 2
a) Inner and outer electrodes for measuring the ionization signal. The inner electrode extends from the center to a radius of 34.5 mm and the outer electrode extends from a radius of 35.5 mm to just before the edge (0.5–2 mm; the diameter of the detectors is 76 mm). b) Phonon channels (QETs) A–D arranged in quadrants on the opposite side of the detector. Each phonon channel is composed of approximately 1000 TESs formed from thin films deposited onto the crystal surface and then photolithographically structured. The TESs are connected in parallel to form the QETs and each QET is read out by its own circuit containing a SQUID array.
Reuse & Permissions
Figure 3
The OFX $χ^{2}$ as a function of delay for the T3Z4 WIMP candidate from the original analysis [33]. The delay chosen by the original ionization fit to this event (red dashed line) is not a global minimum in $χ^{2}$ ; rather it is displaced $\sim 4 μ s$ from the global minimum (green dot-dashed line). While this small error in the inferred delay for this and other events did not affect the measured energy significantly (a $\sim 2 %$ decrease), it caused an error in the surface-event-rejection timing parameter for low-energy events.
Reuse & Permissions
Figure 4
Detection efficiency combined across all detectors as a function of energy for different classes of cuts. The curves show the total efficiency as more cuts are added beyond the hardware trigger (pink triple-dot-dashed): event-level data-quality cuts (green solid), ionization-based fiducial-volume cut (magenta dotted), ionization-yield cut (maroon dot-dashed), and 5d- $χ^{2}$ (see Sec. 5c) phonon timing cut (blue dashed). The falloff of the data-quality cut efficiency below $\sim 20 keV$ is due to the interplay between the ionization threshold and the requirement that the ionization yield be $3 σ$ below the ER band (see main text). The vertical black dashed line is our standard 10 keV analysis threshold.
Reuse & Permissions
Figure 5
ER and NR bands defined in the plane of yield versus recoil energy for a representative detector (T1Z2). The ER band mean $\pm 3 σ$ curves (red hatched), NR band mean $\pm 2 σ$ curves (blue fine cross-hatched), ionization threshold $3.0 σ$ and $4.5 σ$ curves (green coarse cross-hatched), and the 10 keV threshold line (black dashed) are shown superimposed onto ${}^{252}{Cf}$ data (points). The widths of the bands represent the time-period variation of the fits that are used to define them.
Reuse & Permissions
Figure 6
Classic timing cut in the delay versus rise-time plane for a representative detector (T1Z5). ${}^{252}{Cf}$ NRSS and NRMS events (blue open circles) and ${}^{133}{Ba}$ WBMS events (red crosses) are shown. Accepted events lie within the “consistency” region (black dashed lines) and to the upper right of the discrimination cut (black solid line). Considering the ${}^{133}{Ba}$ WBMS events (red crosses), 2381 events fail the timing cut and four events pass.
Reuse & Permissions
Figure 7
Classic timing cut in the ionization-yield versus timing-parameter plane for a representative detector (T1Z5). ${}^{252}{Cf}$ NRSS and NRMS events (blue open circles), ${}^{133}{Ba}$ WBMS events (red crosses), and ${}^{133}{Ba}$ events in the ER $\pm 2 σ$ band (black filled circles) are shown. Accepted events lie to the right of the timing parameter line (black dashed line) and in the NR band (black solid lines). Considering the ${}^{133}{Ba}$ WBMS events (red crosses), 2381 events fail the timing cut and four events pass.
Reuse & Permissions
Figure 8
Neural-network training distributions of a representative detector (T1Z5), for the low-energy neural-network bin ( $< 30 keV$ ). (Top panel) Contours of constant neural-network output (grayscale solid curves; see text) in the plane of the first two principal components. ${}^{252}{Cf}$ NRSS and NRMS events (blue open circles) and ${}^{133}{Ba}$ WBMS events (red crosses) are also shown. (Bottom panel) Distributions of the “transformed” neural-network output: the distribution of NRs from the ${}^{252}{Cf}$ data mentioned above (blue solid); and the surface-event distribution from the ${}^{133}{Ba}$ data mentioned above (red dashed). The cut threshold (vertical black dashed line) with the WIMP-search data passing (green asterisk) and failing ( $black \times ’ s$ ) the timing cut are also plotted along the top of the plot. The transformed neural network is used to make the distribution separation more visible; it is computed by using the inverse of the neuron response function: $ϕ : (- \infty, \infty) \to (0, 1)$ with $ϕ (x) = 1 / (1 + e^{- x})$ .
Reuse & Permissions
Figure 9
Distributions of calibration data in the 5d- $χ^{2}$ parameters for a representative detector (T1Z2). (Top panel) ${}^{252}{Cf}$ events in the NR $\pm 2 σ$ band (blue open circles) are shown with the consistency cut that retains 90% of neutrons (vertical black dashed line) in the $χ_{B}^{2}$ versus $χ_{N}^{2}$ plane. (Bottom panel) Events passing the consistency cut—NRs from ${}^{252}{Cf}$ data (blue open circles) and surface events from ${}^{133}{Ba}$ data (red crosses)—in the $χ_{B}^{2} - χ_{N}^{2}$ versus recoil-energy plane. The energy-dependent 5d- $χ^{2}$ timing cut is also shown (black solid line) and events above the line pass the cut. Note the cut is tighter at low energies and looser at higher energies.
Reuse & Permissions
Figure 10
Total combined efficiencies for all timing analyses. (Top panel) Efficiencies for the 10 keV threshold analyses. The SAEs ( $60 GeV / c^{2}$ WIMP) are 219.1 kg days for the classic analysis (black solid), 216.4 kg days for the neural-network analysis (red dot-dashed), and 262.3 kg days for the 5d- $χ^{2}$ analysis (blue dashed). (Bottom panel) The efficiencies for the extended-threshold analyses with the addition of the 5d- $χ^{2}$ analysis optimized for a $8 GeV / c^{2}$ WIMP mass (magenta dotted). All of the analyses have a total exposure (before efficiency reductions) of 612.2 kg days.
Reuse & Permissions
Figure 11
Distributions of WIMP-search events in the detectors containing the two candidate events for the classic timing analysis: T1Z5 (top panel) and T2Z3 (bottom panel). The normalized yield is the distance from the average NR yield in units of the standard deviation of the yield distribution; the black horizontal lines on the right side of the plot indicate the $\pm 2 σ$ NR band. The normalized timing parameter is the standard timing parameter ( $t_{del} + τ$ ) minus the value of the cut boundary. WIMP-search events (red crosses) to the left of the timing cut (vertical black dashed line) pass all cuts except phonon timing and yield; events to the right of that line pass the timing cut, and those that are also in the NR band (solid black lines) pass the yield cut. The highlighted event (blue open circle surrounding red cross) in the top panel is candidate 1 and the highlighted event in the bottom panel is candidate 2 (see Table 2). All the events shown are in the recoil energy range 10–100 keV.
Reuse & Permissions
Figure 12
Schematic of the CDMS II shielding configuration. The outermost vacuum can surface is shown near the center and radially outward the layers are inner polyethylene (green), inner lead (light grey), outer lead (darker grey), outer polyethylene (green), and muon-veto scintillator counters (blue) with attached light guides and phototubes (white and black). Cooling of the detectors is achieved through the “cold stem” from a dilution refrigerator (dark blue), while cabling passes through the “electronics stem” on the other side of the setup.
Reuse & Permissions
Figure 13
Measured event-summed energy across all detectors (black solid line) compared to simulations of contaminants from common locations: lead shielding (blue diamonds), inner polyethylene shielding (magenta crosses), copper cans and tower components (cyan filled circles), and outermost can surfaces (brown stars). The sum of all simulated components (red inverted triangles) is also plotted and is in good agreement with the measured data.
Reuse & Permissions
Figure 14
This figure compares the main results from this analysis (the current 5d- $χ^{2}$ analysis combined with prior CDMS II exposures, black solid and the classic extended-threshold analysis, black dashed) with previously published results (all limits are at 90% C.L.): Darkside-50 [81] (orange triple-dot-dashed); XENON100 [82] (blue triple-dot-dashed); LUX [83] (red dot-dashed); SuperCDMS low-threshold [84] (green dotted); CDMS II/EDELWEISS combined [35] (purple long dashed); CRESST-II [85] (magenta long-dot-dashed); CDMS II Si [39] (90% and 68% C.L. contours, blue, with the best-fit point marked with a black dot); DAMA/LIBRA [43, 86] ( $3 σ$ region, light red) and CoGeNT [40] (90% C.L., brown).
Reuse & Permissions
Figure 15
Experimental upper limits (90% confidence level) derived from each of the analyses presented in this work compared with the originally published [33] (black dotted) limits. The CDMS II Si contour is shown with the best-fit point marked with a black dot (WIMP mass of $8.6 GeV / c^{2}$ and WIMP-nucleon cross section of $1.9 \times 10^{- 41} {cm}^{2}$ ) [39]. (Top panel) The 10 keV threshold analyses. The 5d- $χ^{2}$ limit (blue dashed) is the “primary” high-threshold result to be quoted from this work. The neural-network and classic limits are shown as red dot-dashed and black solid lines respectively. (Bottom panel) The extended-threshold limits, focused on the lower WIMP-mass region. The same color code applies except that all of the analyses from this work correspond to the extended-threshold versions. The classic limit (black solid) is the “primary” extended-threshold result to be quoted from this work. The extended 5d- $χ^{2}$ limit shown corresponds to the timing-cut optimization assuming a $60 GeV / c^{2}$ WIMP mass. For comparison, the previous CDMS II low-threshold limit is shown [34] (green triple-dot-dashed).
Reuse & Permissions

Physical Review D

covering particles, fields, gravitation, and cosmology