Results from a low-energy analysis of the CDMS II germanium data.
ABSTRACT We report results from a reanalysis of data from the Cryogenic Dark Matter Search (CDMS II) experiment at the Soudan Underground Laboratory. Data taken between October 2006 and September 2008 using eight germanium detectors are reanalyzed with a lowered, 2 keV recoil-energy threshold, to give increased sensitivity to interactions from weakly interacting massive particles (WIMPs) with masses below ∼10 GeV/c(2). This analysis provides stronger constraints than previous CDMS II results for WIMP masses below 9 GeV/c(2) and excludes parameter space associated with possible low-mass WIMP signals from the DAMA/LIBRA and CoGeNT experiments.
arXiv:1011.2482v1 [astro-ph.CO] 10 Nov 2010
Results from a Low-Energy Analysis of the CDMS II Germanium Data
Z. Ahmed,1D.S. Akerib,2S. Arrenberg,19C.N. Bailey,2D. Balakishiyeva,17L. Baudis,19D.A. Bauer,3
P.L. Brink,7T. Bruch,19R. Bunker,15B. Cabrera,11D.O. Caldwell,15J. Cooley,10P. Cushman,18
M. Daal,14F. DeJongh,3M.R. Dragowsky,2L. Duong,18S. Fallows,18E. Figueroa-Feliciano,5J. Filippini,1
M. Fritts,18S.R. Golwala,1J. Hall,3R. Hennings-Yeomans,2S.A. Hertel,5D. Holmgren,3L. Hsu,3
M.E. Huber,16O. Kamaev,18M. Kiveni,12M. Kos,12S.W. Leman,5R. Mahapatra,13V. Mandic,18
K.A. McCarthy,5N. Mirabolfathi,14D. Moore,1, ∗H. Nelson,15R.W. Ogburn,11A. Phipps,14M. Pyle,11X. Qiu,18
E. Ramberg,3W. Rau,6A. Reisetter,18,8T. Saab,17B. Sadoulet,4,14J. Sander,15R.W. Schnee,12D.N. Seitz,14
B. Serfass,14K.M. Sundqvist,14M. Tarka,19P. Wikus,5S. Yellin,11, 15J. Yoo,3B.A. Young,9and J. Zhang18
1Division of Physics, Mathematics & Astronomy,
California Institute of Technology, Pasadena, CA 91125, USA
2Department of Physics, Case Western Reserve University, Cleveland, OH 44106, USA
3Fermi National Accelerator Laboratory, Batavia, IL 60510, USA
4Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA
5Department of Physics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
6Department of Physics, Queen’s University, Kingston, ON, Canada, K7L 3N6
7SLAC National Accelerator Laboratory/KIPAC, Menlo Park, CA 94025, USA
8Department of Physics, St.Olaf College, Northfield, MN 55057 USA
9Department of Physics, Santa Clara University, Santa Clara, CA 95053, USA
10Department of Physics, Southern Methodist University, Dallas, TX 75275, USA
11Department of Physics, Stanford University, Stanford, CA 94305, USA
12Department of Physics, Syracuse University, Syracuse, NY 13244, USA
13Department of Physics, Texas A & M University, College Station, TX 77843, USA
14Department of Physics, University of California, Berkeley, CA 94720, USA
15Department of Physics, University of California, Santa Barbara, CA 93106, USA
16Departments of Phys. & Elec. Engr., University of Colorado Denver, Denver, CO 80217, USA
17Department of Physics, University of Florida, Gainesville, FL 32611, USA
18School of Physics & Astronomy, University of Minnesota, Minneapolis, MN 55455, USA
19Physics Institute, University of Z¨ urich, Winterthurerstr. 190, CH-8057, Switzerland
We report results from a reanalysis of data from the Cryogenic Dark Matter Search (CDMS II) ex-
periment at the Soudan Underground Laboratory. Data taken between October 2006 and September
2008 using eight germanium detectors are reanalyzed with a lowered, 2 keV recoil-energy thresh-
old, to give increased sensitivity to interactions from Weakly Interacting Massive Particles (WIMPs)
with masses below ∼10 GeV/c2. This analysis provides stronger constraints than previous CDMS II
results for WIMP masses below 9 GeV/c2and excludes parameter space associated with possible
low-mass WIMP signals from the DAMA/LIBRA and CoGeNT experiments.
PACS numbers: 14.80.Ly, 95.35.+d, 95.30.Cq, 95.30.-k, 85.25.Oj, 29.40.Wk
A convergence of astrophysical observations indicates
that ∼80% of the matter in the universe is in the form
of non-baryonic, non-luminous dark matter . Weakly
Interacting Massive Particles (WIMPs) , with masses
from a few GeV/c2to a few TeV/c2, form a well-
motivated class of candidates for this dark matter [1, 3].
If WIMPs account for the dark matter, they may be de-
tectable through their elastic scattering with nuclei in
terrestrial detectors .
Although many models of physics beyond the Stan-
dard Model provide WIMP candidates, supersymmet-
ric (SUSY) models where the lightest superpartner is a
cosmologically stable WIMP are among the most pop-
ular [1, 3].In the simplest SUSY models, WIMPs
with masses ?40 GeV/c2are generally disfavored by
accelerator constraints (e.g., ). Interest in lower-
mass WIMPs has been renewed by recent results from
the DAMA/LIBRA  and CoGeNT  experiments,
which have been interpreted in terms of elastic scatters
from a WIMP with mass ∼10 GeV/c2and cross-section
∼10−40cm2[8, 9]. Although it is difficult to accommo-
date a WIMP with these properties in the Minimal Su-
persymmetric Standard Model (MSSM) , relatively
simple extensions to the MSSM can avoid existing con-
straints . More complicated models which allow for
∼10 GeV/c2WIMPs have also been proposed (e.g., ).
The CDMS II experiment attempts to identify nuclear
recoils from WIMPs in an array of particle detectors
by measuring both the ionization and non-equilibrium
phonons created by each particle interaction.
grounds can be rejected on an event-by-event basis since
they primarily scatter from electrons in the detector, de-
positing significantly more ionization than a nuclear re-
coil of the same energy. Previous analyses of CDMS II
data  imposed a recoil-energy threshold of 10 keV
to maintain sufficient rejection of electron recoils that
only ∼0.5 background events would be expected in the
signal region. At lower energies, the discrimination be-
tween nuclear and electron recoils degrades, leading to
higher expected backgrounds. Since WIMPs with masses
<10 GeV/c2primarily produce <10 keV recoils, this
analysis lowers the recoil-energy threshold to 2 keV, com-
parable to the hardware trigger threshold. This lower en-
ergy threshold increases sensitivity to low-mass WIMPs
at the cost of significant acceptance of backgrounds.
The data analyzed here were collected using all 30 Z-
sensitive Ionization and Phonon (ZIP) detectors installed
at the Soudan Underground Laboratory [13, 14]. The de-
tector array consisted of 19 Ge (∼230 g each) and 11 Si
(∼105 g each) detectors, each of which is a disk ∼10 mm
thick and 76 mm in diameter. Each detector was instru-
mented with four phonon sensors on one face and two
concentric charge electrodes on the opposite face. A small
electric field (3–4 V/cm) was applied across the detectors
to extract charge carriers created by particle interactions.
The detectors were arranged in five stacks, or “towers,”
and are identified by their tower number (T1–T5) and by
their ordering within the tower (Z1–Z6). The entire array
was cooled to ?50 mK and surrounded by passive lead
and polyethylene shielding. An outer plastic scintillator
veto was used to identify showers containing cosmogenic
muons which were not shielded by the rock overburden
above the Soudan laboratory (2090 meters water equiv-
The data were taken during six data runs from Oc-
tober 2006 to September 2008 . Only the eight Ge
detectors with the lowest trigger thresholds were used
to identify WIMP candidate events since they have the
best expected sensitivity to WIMPs with masses from
5–10 GeV/c2. All 30 detectors were used to veto events
which deposited energy in multiple detectors.
Stability throughout the data-taking periods was mon-
itored on a detector-by-detector basis using Kolmogorov-
Smirnov tests, and periods of abnormal detector per-
formance were removed. The ratio of ionization to re-
coil energy (“ionization yield”) for electron recoils was
also monitored in order to exclude data periods show-
ing evidence of reduced ionization collection. Data taken
within 20 days following exposure of the detectors to a
neutron calibration source were removed to reduce low-
energy electron-recoil backgrounds due to activation of
the detectors. The data were randomly divided into two
subsets before defining selection criteria at low energy.
One subset, consisting of one quarter of the data (the
“open” data), was reserved to study backgrounds at low
energy and was not used to calculate exclusion limits.
The remaining subset totaled 241 kg days raw exposure,
after removing the bad data periods described above.
The detector response to electron and nuclear recoils
was calibrated by regular exposures of the detectors to
γ-ray (133Ba) and neutron (252Cf) sources. The ioniza-
tion energy scale was calibrated using the 356 keV line
from the133Ba source. The phonon energy was then cali-
brated by normalizing the phonon-based recoil energy for
electron recoils to their mean ionization energy. In con-
trast to previous analyses, a position-dependent calibra-
tion was not applied since position-dependent variations
in the reconstructed phonon energies are less significant
than noise fluctuations at low energies. The energy scale
for electron recoils was verified using activation lines at
1.3 keV and 10.4 keV produced from exposure of the Ge
detectors to the252Cf source. For nuclear recoils, the re-
coil energy was reconstructed from the measured phonon
energy alone by subtracting the Neganov-Luke phonon
contribution  corresponding to the mean ionization
measured for nuclear recoils of the same phonon energy
from the252Cf source. The ionization yield for nuclear
recoils was measured down to ∼4 keV, below which a
power-law extrapolation was used.
Candidate events were required to pass basic recon-
struction quality cuts similar to the criteria used in pre-
vious analyses of these data . Due to the negligible
probability of a WIMP interacting more than once in
the apparatus, candidates were required to have energies
consistent with noise in all but one detector and have no
coincident activity in the plastic scintillator veto. They
were further required to have ionization signals consistent
with noise in the outer charge electrode. The ionization
energy was required to be within (+1.25,−0.5)σ of the
mean ionization energy for nuclear recoils measured from
calibration data, which defines the “nuclear-recoil band.”
This asymmetric band was chosen based on calibration
data and the observed low-energy backgrounds in the
open data, in order to maximize sensitivity to nuclear
recoils while limiting leakage from electron recoils and
zero-charge events. The recoil-energy range considered
for this analysis was 2–100 keV.
The hardware trigger efficiency was determined using
events for which at least one other detector triggered,
which provide an unbiased selection of events near thresh-
old. The data are well described by an error function,
with the mean trigger threshold varying from 1.5–2.5 keV
for the eight Ge detectors. Based on the selection criteria
above, the signal acceptance was measured using nuclear
recoils from the252Cf calibration data. The livetime-
weighted average of the individual detector selection ef-
ficiencies is shown in the inset of Fig. 1, with the largest
loss of efficiency coming from the requirement on ioniza-
The energy spectrum for the candidate events passing
all selection cuts is shown in Fig. 1. Although the shape
of the observed spectrum is consistent with a WIMP sig-
nal, the majority of the candidates appear to arise from
unrejected electron recoils. Figure 2 shows the distribu-
2510 15 20
Recoil energy (keV)
Event rate (keV−1kg−1day−1)
Recoil energy (keV)
FIG. 1. (color online). Comparison of the energy spectra
for the candidate events and background estimates, co-added
over the 8 detectors used in this analysis. The observed event
rate (error bars) agrees well with the electron-recoil back-
ground estimate (solid), which is a sum of the contributions
from zero-charge events (dashed), surface events (+), bulk
events (dash-dotted), and the 1.3 keV line (dotted).
gray band denotes the 1σ statistical errors on the background
estimate. The selection efficiencies have been applied to the
background estimates for direct comparison with the observed
rate, which does not include a correction for the nuclear-recoil
acceptance. The inset shows the measured nuclear-recoil ac-
ceptance efficiency, averaged over all detectors.
tion of candidates in the ionization-yield versus recoil-
energy plane for T1Z5. A band of events with ionization
energies consistent with noise is seen below the nuclear-
recoil band. Most or all of these “zero-charge” events
arise from electron recoils near the edge of the detec-
tor, where the charge carriers can be completely collected
on the cylindrical wall rather than on the readout elec-
trodes. At recoil energies ?10 keV, these events can be
rejected using a phonon-based fiducial-volume cut; how-
ever, at lower energies, reconstruction of the event radius
using phonon information is unreliable. To maintain ac-
ceptance of low-energy nuclear recoils, some zero-charge
events are not rejected at energies ?5 keV where the ion-
ization signal for nuclear recoils becomes comparable to
noise. By extrapolating the exponential spectrum ob-
served for zero-charge events above 5 keV, we estimate
that they contribute 40–60% of the candidate events.
A second source of misidentified electron recoils comes
from events interacting near the detector surfaces, where
ionization collection may be incomplete. These events
are primarily concentrated in a band above the nuclear-
recoil band but below the bulk electron recoils, with an
increased fraction leaking into the signal region at low
energies. For recoil energies ?10 keV, nearly all such sur-
face events can be rejected  because they have faster-
rising phonon pulses than nuclear recoils in the bulk of
the detector. This analysis does not use phonon timing
to reject these events since the signal-to-noise is too low
for this method to be effective for recoil energies ?5 keV.
Extrapolating the exponential spectrum of surface events
identified above 10 keV implies that 10–20% of the can-
Recoil energy (keV)
Ionization yield (keV)
FIG. 2. (color online). Events in the ionization-yield ver-
sus recoil-energy plane for T1Z5.
(+1.25,−0.5)σ nuclear-recoil band (solid) are WIMP candi-
dates (large dots). Events lying outside these bands (small
dots) pass all selection criteria except the ionization-energy
requirement. The widths of the band edges denote variations
between data runs. The recoil-energy scale assumes the ion-
ization signal is consistent with a nuclear recoil, causing elec-
tron recoils to be shifted to higher recoil energies and lower
Events lying within the
didates are surface electron recoils.
At recoil energies ?5 keV, the primary ionization-
based discrimination breaks down as the ionization sig-
nal becomes comparable to noise even for electron recoils
with fully collected charge. Extrapolating the roughly
constant electron-recoil spectrum observed above 5 keV
indicates that 10–15% of the observed candidates arise
from leakage of this background into the signal region.
As shown in Fig. 2, T1Z5 has less leakage from this
background than the average detector since it has the
best ionization resolution. Just above threshold, there is
an additional contribution to the constant electron-recoil
spectrum from the 1.3 keV line, which leaks above the
2 keV analysis threshold since our recoil-energy estimate
assumes the ionization signal is consistent with a nuclear
recoil. The measured intensity of this line at ionization
yields above the signal region indicates that the 1.3 keV
line accounts for 5–15% of the observed candidates.
Monte Carlo simulations indicate that neutrons, whose
nuclear recoils are indistinguishable from WIMPs, pro-
duce a negligible ?0.2 event background.
As shown in Fig. 1, the observed candidate spectrum
can be accounted for with known backgrounds, and there
is no evidence for a signal.
ground model involves sufficient extrapolation that sys-
tematic errors are difficult to quantify, we do not sub-
tract this background but instead set upper limits on the
allowed WIMP-nucleon scattering cross section by con-
servatively assuming all observed events could be from
WIMPs. Limits are calculated using the high statistics
version of Yellin’s optimum interval method . Data
from multiple detectors are concatenated as described
in  due to the presence of significant backgrounds,
which are expected to vary by detector. The candidate
However, since the back-
WIMP mass (GeV/c2)
WIMP−nucleon σSI (cm2)
WIMP mass (GeV/c2)
WIMP−neutron σSD (cm2)
independent (SI) exclusion limits from these data (solid) to
previous results in the same mass range (all at 90% C.L.).
Limits from a low-threshold analysis of the CDMS shallow-
site data  (dashed), CDMS II Ge results with a 10 keV
threshold  (dash-dotted), recalculated for lower WIMP
masses, and XENON100 with constant (+) or decreasing (?)
scintillation-efficiency extrapolations at low energy  are
also shown. The filled regions indicate possible signal regions
from DAMA/LIBRA [6, 8] (dark), CoGeNT (light) [7, 8], and
a combined fit to the DAMA/LIBRA and CoGeNT data 
(hatched). Bottom: comparison of the WIMP-neutron spin-
dependent (SD) exclusion limits from these data (solid),
CDMS II Ge results with a 10 keV threshold (dash-dotted),
XENON10  (△), and CRESST  (dotted). The filled
region denotes the 99.7% C.L. DAMA/LIBRA allowed region
for neutron-only scattering . An escape velocity of 544
km/s was used for the CDMS and XENON100 exclusion lim-
its, whereas the other results assume an escape velocity from
(color online). Top:comparison of the spin-
event energies and selection efficiencies for each detec-
tor are given in .For this analysis, energy inter-
vals from T1Z5 provide the strongest constraints in the
5–10 GeV/c2mass range. The standard halo model de-
scribed in  is used, with specific parameters given
in [16, 22].
The limits do not depend strongly on the extrapola-
tion of the ionization yield used at low energies since the
Neganov-Luke phonon contribution is small for recoil en-
ergies below 4 keV. Conservatively assuming 50% lower
ionization yield near threshold would lead to only ∼5%
weaker limits in the 5–10 GeV/c2mass range.
Figure 3 (upper panel) compares the 90% upper confi-
dence limit on the spin-independent WIMP-nucleon scat-
tering cross section from this analysis to previous re-
sults in the same mass range.
stronger limits than previous CDMS II Ge results for
WIMP masses below ∼9 GeV/c2, and excludes param-
eter space previously excluded only by the XENON100
experiment assuming a constant extrapolation of the liq-
uid xenon scintillation response for nuclear recoils below
5 keV .This parameter space is not excluded by
XENON100 when more conservative assumptions for the
scintillation response are used [8, 17, 24].
Spin-dependent limits on the WIMP-neutron cross sec-
tion are shown in Fig. 3 (lower panel), using the form fac-
tor from . XENON10 constraints, calculated assum-
ing a constant extrapolation of the scintillation response
at low energy [18, 24], are stronger than these results for
WIMP masses above ∼7 GeV/c2.
DAMA/LIBRA annual modulation signal in terms
WIMPs (e.g., [8, 26]).We ignore the effect of ion
channeling on the DAMA/LIBRA allowed regions
since recent analyses indicate channeling should be
negligible . These results are also incompatible with
a low-mass WIMP explanation for the low-energy events
seen in CoGeNT [7, 8].
The CDMS collaboration gratefully acknowledges the
contributions of numerous engineers and technicians; we
would like to especially thank Jim Beaty, Bruce Hines,
Larry Novak, Richard Schmitt and Astrid Tomada. In
addition, we gratefully acknowledge assistance from the
staff of the Soudan Underground Laboratory and the
Minnesota Department of Natural Resources. This work
is supported in part by the National Science Founda-
tion (Grant Nos. AST-9978911, PHY-0542066, PHY-
0503729, PHY-0503629, PHY-0503641, PHY-0504224,
0802575, PHY-0847342, and PHY-0855525), by the De-
partment of Energy (Contracts DE-AC03-76SF00098,
DE-FG02-91ER40688, DE-FG02-92ER40701, DE-FG03-
90ER40569, and DE-FG03-91ER40618), by the Swiss
National Foundation (SNF Grant No. 20-118119), and
by NSERC Canada (Grant SAPIN 341314-07).
This analysis provides
∗Corresponding author: email@example.com
 G. Bertone, D. Hooper, and J. Silk, Phys. Rep., 405,
 G. Steigman and M. S. Turner, Nucl. Phys. B, 253, 375
 G. Jungman, M. Kamionkowski, and K. Griest, Phys.
Rep., 267, 195 (1996).
 M. W. Goodman and E. Witten, Phys. Rev. D, 31, 3059
(1985); R. J. Gaitskell, Ann. Rev. Nucl. Part. Sci., 54,
 E. A. Baltz and P. Gondolo, JHEP, 10, 052 (2004);
A. Heister et al., Phys. Lett. B, 583, 247 (2004).
 R. Bernabei et al. (DAMA/LIBRA), Eur. Phys. J. C, 56,
333 (2008); R. Bernabei et al., ibid., 67, 39 (2010).
 C. E. Aalseth et al. (CoGeNT), arXiv:1002.4703v2
 D. Hooper et al., arXiv:1007.1005v3 (2010).
 A. L. Fitzpatrick, D. Hooper, and K. M. Zurek, Phys.
Rev. D, 81, 115005 (2010); S. Chang et al., JCAP, 08,
18 (2010); S. Andreas et al., Phys. Rev. D, 82, 043522
 E. Kuflik, A. Pierce, and K. M. Zurek, Phys. Rev. D,
81, 111701 (2010); D. Feldman, Z. Liu, and P. Nath,
ibid., 81, 117701 (2010); D. Hooper and T. Plehn, Phys.
Lett. B, 562, 18 (2003).
 A. V. Belikov et al., arXiv:1009.0549v1 (2010); A. Bot-
tino et al., Phys. Rev. D, 69, 037302 (2004); G. B´ elanger
et al., JHEP, 03, 12 (2004).
 J. L. Feng, J. Kumar, and L. E. Strigari, Phys. Lett. B,
670, 37 (2008); S. Andreas, T. Hambye, and M. H. G.
Tytgat, JCAP, 10, 34 (2008); K. M. Zurek, Phys. Rev.
D, 79, 115002 (2009).
 Z. Ahmed et al. (CDMS), Phys. Rev. Lett., 102, 011301
(2009); Science, 327, 1619 (2010).
 D. S. Akerib et al. (CDMS), Phys. Rev. D, 72, 052009
 B. Neganov and V. Trofimov, Otkryt. Izobret., 146, 215
(1985); P. N. Luke, J. Appl. Phys., 64, 6858 (1988).
 D. Akerib et al. (CDMS), arXiv:1010.4290v1 (2010).
 E. Aprile et al. (XENON100), Phys. Rev. Lett., 105,
 J. Angle et al. (XENON10), Phys. Rev. Lett., 101,
 G. Angloher et al. (CRESST), Astropart. Phys., 18, 43
(2002); C. Savage, P. Gondolo,
Rev. D, 70, 123513 (2004).
 C. Savage et al., JCAP, 04, 10 (2009).
 S.Yellin, Phys.Rev.
 See EPAPS Document No. XXX for
list ofthe candidate
tor selection efficiencies
these limits. For more information on EPAPS, see
 J. D. Lewin and P. F. Smith, Astropart. Phys., 6, 87
 J. I. Collar and D. N. McKinsey, arXiv:1005.0838v3
 V. I. Dimitrov, J. Engel, and S. Pittel, Phys. Rev. D,
51, R291 (1995).
 C. Savage et al., arXiv:1006.0972v2 (2010); N. Bozorg-
nia, G. B. Gelmini, and P. Gondolo, arXiv:1006.3110v1
and K. Freese, Phys.