First Dark Matter Search Results from a Surface Run of the 10-L DMTPC Directional Dark Matter Detector

The Dark Matter Time Projection Chamber (DMTPC) is a low pressure (75 Torr CF4) 10 liter detector capable of measuring the vector direction of nuclear recoils with the goal of directional dark matter detection. In this paper we present the first dark matter limit from DMTPC. In an analysis window of 80-200 keV recoil energy, based on a 35.7 g-day exposure, we set a 90% C.L. upper limit on the spin-dependent WIMP-proton cross section of 2.0 x 10^{-33} cm^{2} for 115 GeV/c^2 dark matter particle mass.


I. INTRODUCTION
Despite strong astrophysical evidence that dark matter comprises approximately 23% of our universe [1], the nature of this dark matter remains largely unknown. Weakly interacting massive particles (WIMPs) are a favored dark matter candidate [2]. Many indirect and direct detection experiments aim to discover and measure the properties of WIMPs [3].
Direct WIMP detection experiments search for the interaction of WIMPs with a nucleus in the detector, resulting in low-energy nuclear recoils [4]. Most experiments seek to detect the kinetic energy deposited by the recoiling nucleus; a handful of recent efforts, including this work, also seek to detect the direction of the nuclear recoil, and in this way, infer the direction of incoming WIMPs [5][6][7][8][9][10][11]. The arrival direction of WIMPs is predicted to peak in the direction opposite to the earth's motion around the galactic center in the simplest dark matter halo model, and have a time-varying asymmetry because the Earth's rotation gives angular modulation in time [12]. The angular signature of directional detection offers the potential for unambiguous observation of dark matter [13]. This paper presents the first dark matter limit from the DMTPC directional detection experiment, from a surface run at MIT.

II. THE DARK MATTER TIME PROJECTION CHAMBER EXPERIMENT
DMTPC is a dark matter detector designed to measure the direction and energy of recoiling fluorine nuclei. CF 4 is chosen as a target due to good scintillation characteristics [14] and the relatively large predicted axial-vector coupling for fluorine, allowing sensitivity to spin-dependent WIMP interactions [15,16].   signal and produces scintillation. The wavelength spectrum of the scintillation light peaks at ∼600 nm, with roughly two-thirds of the scintillation emission in the visible [14]. The gas gain is approximately 4 × 10 4 , measured with an 55 Fe calibration source. The operating anode voltage is chosen to maximize the gain while limiting the rate of electronic discharge between the anode and ground plane to <0.025 Hz. The drift electric field is chosen to minimize the transverse diffusion of the drifting electrons. For a more detailed discussion of the 10 liter detector amplification and diffusion, see [17] and [18].
Scintillation light produced in the amplification region is focussed by a Nikon photographic lens (f/1.2, 55 mm focal length) onto an Apogee Alta U6 camera containing a 1024×1024 element Kodak 1001E CCD chip with 24×24 µm 2 pixels. The CCD clock rate is 1 MHz with 16-bit digitization, and typical readout time is 0.2 seconds. With this camera we are read-noise limited. To improve the signal-to-noise ratio (and to reduce deadtime from CCD readout), pixels are binned 4×4 prior to digitization. In addition to optical readout, we also digitize the integrated charge induced on the anode, although we do not use the charge data in this analysis.
The surface run data set consists of 231,000 five-second CCD exposures from each camera, collected without trigger or camera shutter. Of these, 10.5% and 4.4% are rejected by analysis cuts as spark events in the top and bottom TPCs respectively, and 3.5×10 −3 % and 8.7×10 −4 % as the associated residual bulk image (RBI) background pixels (described in Section II B) respectively. After correcting the live time for these analysis cuts, the data  The energy response of the detector is obtained from the same data. The integral light yield of segments of alpha tracks at known distances from the source, in the arbitrary digital units (ADU) of the CCD, is compared to the SRIM simulation [19] prediction for the visible energy loss in that segment. The segment length is chosen such that the SRIM prediction for the energy loss in each segment is 100-1000 keV, depending on the location and size of the segment along the alpha track. This procedure is done in the region of the alpha track where the alpha energy is above 1 MeV (before the Bragg peak). According to SRIM, at these energies, the alpha energy loss is >97% electronic and so we are not sensitive to assumptions about the nuclear quenching in this calibration. This procedure gives the energy calibration scattered 122 keV photons. The track-finding algorithm does not identify distinct tracks in the 57 Co data, largely because these fail the requirement of having at least five contiguous pixels above threshold (described further below). This is consistent with the predicted low ionization density of electron-like tracks (see, for example Figure 16 in [5]). Rather, these events may have a few pixels above background in the entire field of view. To obtain high statistics, each gain non-uniformity measurement is integrated over 10,000 seconds; from calculation, the intensity and position of the source are such that the area imaged by each CCD pixel is covered by at least one electron recoil per second. The measurement yields a 10% variation of the total system gain, which is included as a position-dependent correction in the gain systematic study in Section II B. The stability of the gain vs. time was measured to be 1% over 24 hours using an 241 Am alpha source. To maintain 1% gain uniformity, the chamber is evacuated to 10 mTorr and refilled with CF 4 every 24 hours. in [17].

B. Surface Run Results
A major goal of the surface run was to identify detector backgrounds prior to underground operations. We found two broad categories: events which produce ionization outside the TPC drift volume, and events which occur inside it. A summary is given in Table I.
Background events producing ionization outside the fiducial volume are mostly interactions of cosmic rays or radioactivity in the CCD chip, which is a well documented phenomenon [23]. These may be removed in the future by requiring coincidence of CCD and charge or PMT readout; in this CCD-only analysis, we reject these events in software. Such tracks typically have a few bins with very high yields. We identify these events by the large ADU and RMS of the pixels comprising the track . Another type of outside event is associated with sparks in the amplification region. Sparks are identified by having an image mean which differs by >1% from the previous image. For comparson, images containing very bright alpha tracks differ in this metric by <0.01%. Sparks may induce residual bulk images (RBIs), which appear at the same spatial position for many subsequent images. RBIs are the result of the leakage of charge from the epitaxial/substrate interface of the CCD; these are a well-known background in front-illuminated CCDs associated with interactions of >600 nm photons in the chip [23,24]. We identify these events by their coincident positions.
Background events producing ionization inside the fiducial volume come primarily from alphas and neutrons. Alpha particles are emitted by radio-impurities in or on the materials of the detector; the majority are from the stainless steel drift cage. These are identified as CCD edge-crossing tracks. Another characteristic of alphas is their long range; we require nuclear recoils to have projected ranges <5 mm. This range vs. energy discrimination is unique to 8 tracking detectors. Figure 3 (left) shows events identified as alpha particles in comparison with the SRIM prediction for the maximum projected range vs. visible energy; tracks which are not parallel to the image plane have projected ranges less than this maximum. The ambient neutron flux comes from 238 U and 232 Th decays, and from cosmic ray spallation. There is no evidence for gamma-induced electron backgrounds [25]; the measured rejection is > 10 6 [31]. The events remaining after all background cuts are shown in Figure 4 (left), compared to WIMP Monte Carlo.
We set a limit on the spin-dependent WIMP-proton interaction cross section using the method described in [16].  (Figure 4, right). We do not take into account the building around the detector, and so assign 100% uncertainty to the neutron background and report the limit assuming zero expected events. Using the Feldman-Cousins method [27], we set a 90% confidence level limit on the spin-dependent WIMP-proton cross section, shown in Figure 5 (left). Following [16], we use the thin-shell spin-dependent form factor approximation, and the interaction factor C 2 W p = 0.46 for Higgsino-proton coupling. The 90% C.L. cross section upper limit is 2.0 × 10 −33 cm 2 at 115 GeV/c 2 WIMP mass. If we vary the gain non-uniformity by 100%, the limit is < 2.3 × 10 −33 cm 2 . If we include the estimated background of 74 events, the limit is < 8.0 × 10 −34 cm 2 .
We evaluate the probability that events passing the nuclear recoil selection cuts come from an isotropic background vs. anisotropic WIMP-induced recoil angle distribution. The Rayleigh statistic is a powerful tool to analyze the uniformity of a distribution of angles when looking for a preferred direction [28]. Using the Rayleigh statistic, we quantify the anisotropy in (φ − φ source ), which is the most sensitive variable to test for anisotropy in the 9 WIMP mass (GeV)  case of two dimensional readout [29]. (φ−φ source ) is the difference between the reconstructed φ and the projection of the expected dark matter direction at the time of each event onto the image plane. The (φ − φ source ) vs. E R distribution after nuclear recoil selection cuts in is shown in Figure 5 (right). We find no statistically significant deviation from a uniform distribution; 36% of the time uniformly distributed data have a Rayleigh value higher than that of our candidate events. The reconstructed angle of the 252 Cf calibration data relative to its source is also shown in Figure 5 (right). The 252 Cf calibration source is effectively a point source in the lab frame at φ source =0. The Rayleigh test applied to the 252 Cf calibration data after the nuclear recoil selection cuts, in the same recoil energy range (80-200 keV), gives a probability of <1% for a uniform distribution.

III. CONCLUSIONS
We present the first dark matter limit from DMTPC, σ χ−p < 2.0×10 −33 cm 2 at 90% C.L., from a 35.7 g-day surface exposure of a 10 liter detector. The 10 4 rejection of backgrounds using range vs. energy properties of nuclear recoils, from Table I, is an impressive demonstration of the low pressure directional time projection chamber concept. We find that the backgrounds in the analysis window of 80-200 keV are qualitatively consistent with the predicted neutron background. The 10L detector described here began running underground at the Waste Isolation Pilot Plant outside Carlsbad, NM in October 2010. The depth of the WIPP site is 1.6 km water-equivalent. The gamma, muon, and radon background levels have been measured, and the neutron background has been estimated at this site [30]. Based on these, we project that underground operation will lower the expected neutron background to <1 event/year. The projected zero background sensitivity of this detector at WIPP for a 1 year exposure is shown in Figure 5 (left). DMTPC has built a second-generation detector with radio-pure materials for operation at WIPP; this is expected to substantially reduce alpha backgrounds, and fiducial volume coverage by CCDs in coincidence with charge readout will eliminate CCD backgrounds. At the scale of a 1 m 3 detector (300 g target), which the collaboration is actively developing, this detector technology is competitive with the best current spin-dependent cross section limits from conventional dark matter detectors, also shown in Figure 5 (left).