First Dark Matter Results from the XENON100 Experiment

Phys. Rev. Lett. 105, 131302
Global fit to all Leff measurements with fixed neutron energies between 5keVnr and 100keVnr, together with 90% confidence contours (solid lines) and extrapolations to lower energies (dashed lines).

A vast array of observational evidence suggests that 83% of the matter in the Universe is in some unknown form called dark matter [1]. Extensions of the standard model of particle physics that aim at addressing some of the puzzles associated with the electroweak scale predict stable weakly interacting massive particles (WIMPs), that can be thermally produced in the early Universe and become ideal dark matter candidates [2]. One method to detect WIMPs is to measure the energy they deposit in a detector by scattering off target nuclei.

XENON100 is a new ultralow background detector developed to continue the XENON dark matter search program with liquid xenon (LXe) as WIMP target and detection medium. Like XENON10 [3], it is a three-dimensional (3D) position-sensitive dual-phase (liquid-gas) time projection chamber (TPC) filled with ultrapure LXe. Particle interactions in the sensitive LXe volume are measured with two arrays of photomultiplier tubes (PMTs), which simultaneously detect the primary scintillation ( S1) and the ionization signal ( S2), via the proportional scintillation mechanism [4]. The 3D event localization allows “fiducialization” of the TPC, keeping only an inner volume in which the background rate is drastically reduced. The different ionization density of nuclear recoils, from neutrons or WIMPs, and electronic recoils, from γ or β backgrounds, leads to a different S2/S1 ratio, which can be used to discriminate the two types of recoils.

Electronic (top) and nuclear (bottom) recoil bands from Co60 and AmBe241 calibration data, respectively, after data selection and the 40 kg fiducial volume cut. Colored lines correspond to the median log10(S2/S1) values of the electronic (blue [dark grey]) and nuclear (red [light grey]) recoil bands. The WIMP search energy window 8.732.6keVnr (vertical, dashed line) and S2 software threshold of 300 PE (long dashed line) are shown.Electronic (top) and nuclear (bottom) recoil bands from Co60 and AmBe241 calibration data, respectively, after data selection and the 40 kg fiducial volume cut. Colored lines correspond to the median log10(S2/S1) values of the electronic (blue [dark... Show more

The detector uses 161 kg of LXe, divided in two concentric cylindrical volumes. These are physically and optically separated from each other by polytetrafluoroethylene (PTFE) panels on the side, a PMT array on the bottom, and a stainless steel diving bell on the top. The bell allows the inner liquid level to be set independently of the outer one. The PTFE panels define the TPC with 30.5 cm diameter and 30.6 cm height, acting also as an efficient UV reflector. Four stainless steel meshes provide the electric field to drift ionization electrons in the liquid, extract them to the gas phase, and accelerate them to produce proportional scintillation. A drift field of 530V/cm has been used for the measurements reported here.

The PMTs are 2.5cm×2.5cm metal-channel type (R8520-AL) specifically developed in collaboration with Hamamatsu Co. for operation in LXe, with a quantum efficiency of about 30% at 178 nm and with low intrinsic radioactivity ( 1mBq/PMT in U238/Th232). The 80 PMT array at the bottom of the TPC is immersed in the liquid to efficiently detect the S1 signal, while another array of 98 PMTs is placed in the xenon gas above the anode so that the hit pattern of an S2 signal can be used to reconstruct the event position in ( x, y). The interaction depth ( z) in the detector can be computed from the time difference between S1 and S2 pulses with resolution <2mm. The outermost LXe volume is used as an active veto, instrumented with 64 PMTs. The energy threshold of the veto has been measured to be better than 200keVee (keV electron equivalent [5] as inferred from Cs137). The signals from all 242 PMTs are digitized at 100MS/s and 40 MHz bandwidth. The trigger is provided by the summed signal of 84 central PMTs, low-pass filtered with 1 MHz. Given the strong amplification in the gas proportional region, at low energies the trigger is given by the S2 pulse, with an efficiency >99% above 300 photoelectrons (PE).

Cut acceptance (top, not including 50% acceptance from S2/S1 discrimination) and log10(S2/S1) (bottom) as functions of nuclear-recoil energy for events observed in the 40 kg fiducial volume during 11.17 live days. Lines as in Fig. 2.

The detector has been deployed underground at the Laboratori Nazionali del Gran Sasso (LNGS), where the muon flux is reduced by a factor 106, thanks to the average 3600 m water equivalent of rock overburden. The LXe is contained in a double walled, vacuum insulated, stainless steel cryostat. A 200 W pulse tube refrigerator (PTR) continuously liquefies the gas circulated through a hot getter and maintains the liquid at -91°C. The PTR system is installed outside a passive shield to achieve a lower radioactive background in the target. This shield consists of a 20 cm thick layer of lead and a 20 cm thick layer of polyethylene within, to attenuate the background from external γ rays and neutrons, respectively. The shield structure rests on a 25 cm thick slab of polyethylene and is surrounded on the top and three sides by a 20 cm thick water layer to lower the background contribution from neutrons from the cavern rock. A 5 cm thick layer of copper covers the inner surface of the polyethylene to attenuate the gamma background due to its radioactivity. Calibration sources ( Co57, Co60, Cs137, AmBe241) are inserted through a copper tube which penetrates the shield and circles around the detector in the middle of the drift region.

The gas used in XENON100 has been processed through a distillation column to reduce the Kr85 background to 33μBq/kg, as measured with delayed β-γ coincidences [6]. Assuming an isotopic abundance of 10-11, this Kr85 contamination corresponds to 143-90+130ppt (mol/mol), at 90% confidence, of natural Kr.

Distribution of all events (dots) and events below the nuclear-recoil median (red circles) in the TPC (grey line) observed in the 8.732.6keVnr energy range during 11.17 live days. No events below the nuclear-recoil median are observed within the 40 kg fiducial volume (dashed line).Distribution of all events (dots) and events below the nuclear-recoil median (red circles) in the TPC (grey line) observed in the 8.732.6keVnr energy range during 11.17 live days. No events below the nuclear-recoil median are observed within the 4... Show more

PMT gains are measured under single PE conditions using light emitting diodes (LEDs) coupled to optical fibers which illuminate the TPC and veto volumes. The PMT gains, equalized to 1.9×106 at the beginning of the run, are regularly monitored and are stable within ±2% ( σ/μ).

Event positions are calculated using three independent algorithms, based on χ2 minimization, support vector machine regression, and a neural network. We take the PMT gains into account and correct for nonuniformities of the drift field as inferred from a finite element simulation. The three algorithms give consistent results for radii r<14cm with an ( x, y) resolution better than 3 mm, as measured with a collimated γ source. This motivated the choice, for the present analysis, of a 40 kg fiducial volume as a cylinder of radius 13.5 cm and height 24.3 cm.

90% confidence limit on the spin-independent elastic WIMP-nucleon cross section (solid and long dashed lines), together with the best limit to date from CDMS (dotted line) [15], recalculated assuming an escape velocity of 544km/s and v0=220km/s. Expectations from a theoretical model [16], and the areas (90% C.L.) favored by CoGeNT (green) [13] and DAMA (red/orange) [14] are also shown.90% confidence limit on the spin-independent elastic WIMP-nucleon cross section (solid and long dashed lines), together with the best limit to date from CDMS (dotted line) [15], recalculated assuming an escape velocity of 544km/s and v0=220km/s. ... Show more

Corrections for the spatial dependence of the S1 light collection in the TPC are obtained by irradiating the detector at different azimuthal positions with an external Cs137 source and computing the average light yield in 1cm×2.5cm ( r, z) cells. The average light yield of the whole TPC for Cs137 662keVee γ rays is 1.57PE/keVee at a field of 530V/cm. The spatial correction is also inferred using 40keVee γ rays produced via the inelastic reaction Xe129(n,nγ)Xe129, together with 80keVee γ rays from Xe131(n,nγ)Xe131, during the calibration of the detector with an external AmBe241 source. These γ rays are more uniformly distributed in the sensitive volume due to the larger neutron mean free path. In addition, 164keVee and 236keVee γ rays are produced following the same neutron calibration from the decay of metastable Xe131m and Xe129m, respectively. The 164keVee line is also used to infer the S1 spatial dependence. The corrections inferred from these independent calibrations differ by less than 3% and improve the energy resolution ( σ/E) at 662keVee from 24% to 13% using the scintillation signal alone.

Calibrations with Cs137 were taken daily during the data taking presented here, to infer the electron lifetime and to subsequently correct the S2 signal for its drift time dependence. The electron lifetime increased from 154μs to 192μs, resulting in the average S2 z correction decreasing from 75% to 60%. The S2 signal is also corrected for its ( x, y) variation, mostly due to light collection effects near the edge of the TPC. This dependence is determined using the 40keVee γ rays from the neutron calibration data and computing the proportional scintillation light yield in 2cm×2cm ( x, y) cells. Only insignificant differences ( <2%) were observed between corrections obtained using other calibration data sets of various γ-ray energies ( 164keVee, 662keVee). The energy resolution ( σ/E) at 662keVee using the S2 signal alone is improved from 7.3% to 6.5% after applying the S2 spatial corrections.

The nuclear-recoil equivalent energy Enr in LXe is conventionally computed from the scintillation signal, S1, using Enr=(S1/Ly)(1/Leff)(See/Snr), where Leff is the scintillation efficiency of nuclear recoils relative to that of 122keVee γ rays at zero field, and See and Snr are the electric field scintillation quenching factors for electronic and nuclear recoils, respectively, with measured values of 0.58 and 0.95 [7]. Since 122keVee γ rays cannot penetrate far in the sensitive volume, their light yield Ly at 530V/cm is calculated from a fit to all γ-ray lines mentioned above, yielding Ly(122keVee)=(2.20±0.09)PE/keVee. Leff data measured at fixed neutron energies [5,8,9], shown in Fig. 1, have less systematic uncertainty than those inferred from a comparison of neutron calibration spectra with Monte Carlo simulations. Therefore, the energy dependence of Leff and its uncertainty is determined here through a global cubic-spline fit to all data shown in Fig. 1 in the energy range with at least two measurements ( 5100keVnr). The spline knots are fixed at 5, 10, 25, 50, and 100keVnr. Below 5keVnr, a constant extrapolation of the global fit, consistent with the trend reported in Aprile et al. [5] and Sorensen et al. [10], is used in this analysis. A logarithmic extrapolation of the lower 90% confidence contour to zero scintillation near 1keVnr, following the trend in Manzur et al. [8], is also shown in Fig. 1.

Data selection criteria are motivated by the physical properties of xenon scintillation light, the characteristics of proportional light signals, and the expected WIMP-induced single-scatter nuclear-recoil signature. Cuts were developed and tested on calibration data, specifically on low energy electronic recoils from Compton scattered Co60 γ rays and nuclear recoils from AmBe241. In particular, a twofold PMT coincidence is required in a 20 ns window for the S1 signal and events which contain more than a single S1-like pulse are discarded. This allows true low energy events to be distinguished from events with random single photoelectrons from PMTs or accidental coincidences. For the S2 signal, a lower threshold of 300 PE is set, corresponding to about 15 ionization electrons, and events are required to contain only one S2 pulse above this threshold. This rejects events with multiple scatters at different z positions. In addition, the width of the S2 pulse is required to be consistent with what is expected from the inferred drift time due to diffusion of the electron cloud [11]. Events that deposit energy in the veto volume in coincidence with the S1 signal in the TPC are also discarded. The regions of the digitized waveform away from S1 or S2 pulses are required to be free of extraneous PMT signals or noise. Finally, events outside the predefined fiducial volume are rejected.

Background rejection in XENON100 is achieved through a combination of volume fiducialization and the identification of recoil species based on the ratio S2/S1 for electronic and nuclear recoils. Accurate knowledge of the response to both types of recoils is essential to define the signal region, to determine the signal acceptance, and to predict the expected leakage into the signal region. Statistics for the low energy electronic recoil calibration are accumulated at regular intervals with a 1 kBq Co60 source. The response of XENON100 to elastic nuclear recoils was obtained by irradiating the detector with a 220n/s AmBe241 source for 72 h. Figure 2 shows the log10(S2/S1) distribution of single-scatter electronic and nuclear recoils in the 40 kg fiducial volume, as function of nuclear-recoil energy. The energy window for the WIMP search is chosen between 8.732.6keVnr (4–20 PE). The upper bound is taken to correspond approximately to the one used for the XENON10 blind analysis [3], after recomputing the corresponding nuclear-recoil equivalent energy using the new Leff parametrization from the global fit, shown in Fig. 1. The lower bound is motivated by the fact that the acceptance of the S1 twofold coincidence requirement is >90% above 4 PE. The log10(S2/S1) upper and lower bounds of the signal region are, respectively, chosen as the median of the nuclear-recoil band and the 300 PE S2 software threshold.

A first dark matter analysis has been carried out, using 11.17 live days of background data, taken from October 20th to November 12th 2009, prior to the neutron calibration. Although this was not formally a blind analysis, all the event selection criteria were optimized based on calibration data only. The cumulative software cut acceptance for single-scatter nuclear recoils is conservatively estimated to vary between 60% (at 8.7keVnr) and 85% (at 32.6keVnr) by considering all single-scatter events in the fiducial volume that are removed by only a single cut to be valid events (Fig. 3). Visual inspection of hundreds of events confirmed that this is indeed a conservative estimate. Within the 8.732.6keVnr energy window, 22 events are observed, but none in the predefined signal acceptance region (Fig. 3). At 50% nuclear-recoil acceptance, the electronic recoil discrimination based on log10(S2/S1) is above 99%, predicting <0.2 background events in the WIMP region. The observed rate, spectrum, and spatial distribution (Fig. 4) agree well with a GEANT4 Monte Carlo simulation of the entire detector.

An upper limit on the spin-independent WIMP-nucleon elastic scattering cross section is derived based on the assumption of an isothermal WIMP halo with v0=220km/s, density 0.3GeV/c2, and escape velocity 544km/s [12]. We take a S1 resolution dominated by Poisson fluctuations into account and use the global fit Leff with constant extrapolation below 5keVnr. The acceptance-corrected exposure in the energy range considered, weighted by the spectrum of a 100GeV/c2 WIMP, is 172kg·days. Figure 5 shows the resulting 90% confidence upper limit, with a minimum at a cross section of 3.4×10-44cm2 for a WIMP mass of 55GeV/c2. The impact of assuming the lower 90% confidence Leff contour together with the extrapolation to zero around 1keVnr is also shown. Our limit constrains the interpretation of the CoGeNT [13] and DAMA [14] signals as being due to light mass WIMPs. These initial results, based on only 11.17 live days of data, demonstrate the potential of the XENON100 low-background experiment to discover WIMP dark matter.


  1. E. Komatsu et al., Astrophys. J. Suppl. Ser. 180, 330 (2009).
  2. G. Bertone, D. Hooper, and J. Silk, Phys. Rep. 405, 279 (2005).
  3. J. Angle et al., Phys. Rev. Lett. 100, 021303 (2008).
  4. A. Lansiart et al., Nucl. Instrum. Methods 135, 47 (1976).
  5. E. Aprile et al., Phys. Rev. C 79, 045807 (2009).
  6. E. Aprile et al., arXiv:1001.2834.
  7. E. Aprile et al., Phys. Rev. Lett. 97, 081302 (2006).
  8. A. Manzur et al., Phys. Rev. C 81, 025808 (2010).
  9. F. Arneodo et al., Nucl. Instrum. Methods Phys. Res., Sect. A 449, 147 (2000); ; R. Bernabei et al., Eur. Phys. J. direct C 11, 1 (2001); ; D. Akimov et al., Phys. Lett. B 524, 245 (2002); ; E. Aprile et al., Phys. Rev. D 72, 072006 (2005); ; V. Chepel et al., Astropart. Phys. 26, 58 (2006).
  10. P. Sorensen et al., Nucl. Instrum. Methods Phys. Res., Sect. A 601, 339 (2009).
  11. T. Doke, Nucl. Instrum. Methods 196, 87 (1982).
  12. J. D. Lewin and P. F. Smith, Astropart. Phys. 6, 87 (1996); ; M. C. Smith et al., Mon. Not. R. Astron. Soc. 379, 755 (2007).
  13. C. E. Aalseth et al., arXiv:1002.4703.
  14. C. Savage et al., J. Cosmol. Astropart. Phys. 04 (2009) 010.
  15. Z. Ahmed et al., Science 327, 1619 (2010).
  16. R. Trotta et al., J. High Energy Phys. 12 (2008) 024.

About the Authors

Image of
Image of E. Aprile
Image of K. Arisaka
Image of F. Arneodo
Image of A. Askin
Image of L. Baudis
Image of A. Behrens
Image of K. Bokeloh
Image of E. Brown
Image of J. M. R. Cardoso
Image of B. Choi
Image of D. B. Cline
Image of S. Fattori
Image of A. D. Ferella
Image of K.-L. Giboni
Image of A. Kish
Image of C. W. Lam
Image of J. Lamblin
Image of R. F. Lang
Image of K. E. Lim
Image of J. A. M. Lopes
Image of T. Marrodán Undagoitia
Image of Y. Mei
Image of A. J. Melgarejo Fernandez
Image of K. Ni
Image of U. Oberlack
Image of S. E. A. Orrigo
Image of E. Pantic
Image of G. Plante
Image of A. C. C. Ribeiro
Image of R. Santorelli
Image of J. M. F. dos Santos
Image of M. Schumann
Image of P. Shagin
Image of A. Teymourian
Image of D. Thers
Image of E. Tziaferi
Image of H. Wang
Image of C. Weinheimer

Related Articles

Focus: Surface Waves Store Bouncing Droplet’s History
Nonlinear Dynamics

Focus: Surface Waves Store Bouncing Droplet’s History

A droplet bouncing and wandering across a liquid surface can produce waves that store the history of its chaotic motion. Read More »

Synopsis: Get the Hydrogen Out
Industrial Physics

Synopsis: Get the Hydrogen Out

Using clean helium made with a two-step purification process could prevent damaging blockages in cryostats. Read More »

Synopsis: Emptiness Constrains the Universe

Synopsis: Emptiness Constrains the Universe

The distribution of galaxies around regions of relatively empty space can be used to constrain cosmological parameters. Read More »

More Articles