Dark Matter Research

DRIFT

Overview and summary of recent major achievements

The DRIFT programme is a UK-US joint effort at Boulby formed over 9 years ago, that, uniquely in the world, develops and runs underground in realistic conditions, detectors designed with capability to determine the direction of the recoil tracks expected from WIMPs. Publications by us show that such a capability could determine definitively that signal recoil events are of non-terrestrial (galactic) origin and correlated with our galactic motion. The technology is hence in principle very powerful at proving the existence of galactic dark matter. Our publications also show that DRIFT technology is capable of parallel searches for specific classes of axions and other rare decays to two gammas [2]. The activity is widely endorsed by senior international researchers (for instance B. Sadoulet and C. Rubbia) and by the community, as evidenced by the UK-run IDM and CYGNUS directional workshops [3,4]. The UK has played a majority lead role, conceiving, designing and building most of the experiment components including the data acquisition systems, slow control, data-pipelines, much electronics, vessel, internal detector field cages and supports, gas handling and vacuum systems, shielding and underground site. The UK has led major data analysis and simulations efforts. Membership of DRIFT continues to grow, with new US funding provided by NSF in June 2007 and new membership from MIT.

The DRIFT II project at Boulby was funded by PPARC until Oct 2005. We offer here a summary final grant report for the DRIFT programme covering the 4 year grant period and also relevant unfunded work undertaken post Oct 2005 but reported to the OsC. Previous regular written reports to the OsC provide further details. During the relevant period the DRIFT programme achieved or exceeded all internal and PPARC milestones and deliverables on budget. Particular highlights include:

First measurement of the rock neutron flux with DRIFT II unshielded is included also in [6]. A vital back-up independent scintillator measurement has also been completed and published [12]. Agreement was found between the ultra-low flux measurements (1.72±0.61 x 10-6 cm-2s-1) made with the scintillator and the limit produced by DRIFT IIa, with these two very different techniques - a useful additional result. Numerous conference proceedings have also been produced and we expect two further UK driven journal papers from the current work: (1) on directional analysis, and (2) the end-to-end GEANT/Maxwell detector simulation. The following sections provide a history of the programme relevant to the full grant period and details of recent progress. Following sections provide details, including a summary of DRIFT I as relevant to subsequent DRIFT II efforts.

DRIFT I operations and history

First engineering runs of DRIFT I at Boulby occurred during 2002/3, showing stable operation for around 9 months. The collaboration was comfortable that these tests confirmed basic operation of what was a new technology in the field - a low pressure, negative ion TPC underground. A significant result was the realization underground that the rejection efficiency for gammas was as predicted so that Pb gamma shielding for DRIFT I would not be needed. Gamma rejection works in quite a different way to that of other technologies and is difficult to quantify in a meaningful way for comparison - DRIFT aims to be a zero background detector. A reasonable guide would be rejection >1 in 105 at threshold in DRIFT I. During 2003 (after around a year of operation) a series of issues developed in the detector that took considerable effort to identify and correct. These were compounded by faults in the SLAC-built data acquisition boards and in the MWPC protection electronics that were eventually found not to be adequate. The result was a higher instance of wire breakage than should have resulted and hence delays in diagnosis because of the time needed to repair broken wires before the next test.

Background

47 background events that passed analysis cuts (red) plus 252Cf neutron calibration data (black) - Kirkpatrick Thesis.

By early 2004 the causes of the behavior had been successfully identified and corrected and DRIFT I was operational again, taking data. Three key issues were identified: (i) ageing of the Lucite MWPC structure, (ii) gas quality changes due to a small vacuum leak which also compounded the MWPC ageing effect, and (iii) a fatigued high voltage feed through that resulted in discharges. The Lucite ageing, identified as due to slow stress release over 18 months, resulted in a slight warping of the detector and hence wire breakdown, resulting then in wire breakage because of the electronics protection circuits. The warping was successfully corrected by retrofitting tensioning battens. The vacuum leak was successfully traced to one of the DAQ soldered feedthroughs, which was replaced. The HV feed breakdown fault was finally identified optically and replaced. Various other improvements to the detector were also instigated including: new lower noise DAQ cables, improved detector trolley apparatus to help with MWPC operations, re-configured slow control to improve monitoring and safety, new hard-wired pressure cut to ensure switch off of high voltage in case of a leak, new HVV filter to reduce signal noise. Considerable progress was made in data analysis and simulation codes. The later focused on potential sensitivity of the technology to different halo models. Unshielded data from the first operational period was sufficient to allow preliminary identification of neutrons from the cavern walls in agreement with Fluka and GEANT 4 simulations of the expected neutron rate. The figure above shows typical unshielded events plotted on the standard NIPs (energy) vs. R2 (range) plot from 37 live days of unshielded operation from DRIFT I.

G4 Model

Geant 4 Model of DRIFT Array

Following collection of around 2 weeks unshielded but vetoed data the neutron shield (8 tons of poly- pellets) was successfully installed. All detector refurbishments made in 2004 continued to functioned well including the tensioning battens retro-fitted to correct the slow fatigue warping. This had greatly strengthened confidence in the DRIFT II design, then on-going. However, despite progress to identify and correct faults in the Temple-SLAC DAQ this continued to cause issues. In particular, the veto channels (needed to reject stray low energy side alphas) were found to be malfunctioning; with continued variations in noise and board malfunctions. Meanwhile, with a much superior design thanks to the lessons from DRIFT I, construction of the first DRIFT II modules had gone very well, with the first units already being successfully commissioned on the surface. DRIFT II uses much simpler, lower threshold and commercial-based DAQ. Given this the collaboration decided to focus resources on rapid deployment of DRIFT II, using DRIFT I now as a test-bed for DRIFT II components, including the new DAQ. DRIFT I was subsequently de-commissioned.

DRIFT II

DRIFT II design

The design of DRIFT II, including engineering drawings, DAQ and gas systems, was completed by the end of 2003. Space constraints in the JIF area eventually ruled out a vessel significantly larger than DRIFT I. Instead DRIFT II was envisaged as an array of detectors of 1 m3 active volume, starting with DRIFT IIa. Meanwhile, the slow fatigue issues revealed in the DRIFT I MWPCs, due to use of low background Lucite, also indicated improved design of the MWPCs, to make them more robust and easier to deal with. DRIFT II was thus based on a simpler and more robust vessel and detector construction but designed also for duplication for a possible modular array. Specific detector design details for DRIFT II modules are as follows:

Simulations and halo models

Extensive background and sensitivity simulations for DRIFT II were performed at this time, in particular of neutron backgrounds including muon neutrons (FLUKA/FAUST) and rock and vessel neutrons (GEANT IV) (Carson et al., and thesis of J. Davies). These revealed that no muon neutron active veto is needed for the full DRIFT II array, assuming stainless steel U/Th contents below 1 ppb. The muon neutrons dominate the recoil background at ~1.6 per year. Simulations were also completed on idealized detectors to assess prospects for differentiating different halo models in the event of a detection. Input from triaxial and other halo models were passed through detector models to generate the likely response. Results suggested that with around x10 more WIMP counts than needed to identify a directional signal it would be possible to distinguish between extreme cases of halo models.

DRIFT II construction

Construction of DRIFT II made rapid progress and, at the time of the October 2004 proposal DRIFT IIa was fully constructed with surface commissioning nearly complete and first underground data expected in January 2005. Indeed plans were already well advanced with US partners for the first 6 modules with tendering for 6 vessels completed the previous May and orders placed with Royal (US) and Wessington (UK). Two companies were chosen to maximize commissioning efficiency and minimize transport delays - the first US vessel for delivery to the testing base at Oxy, the first UK vessel for delivery to Boulby. Modules 1 and 2 were complete and a further 4 vessels were on schedule to meet the milestone set in 2002 of 03/05. At this time MWPC production was ahead of schedule with sufficient for 5 vessels complete. The first field cages, cathodes and 100 KV feeds had been shipped to Oxy from the UK and assembled in module 1. The gas system and CS2 apparatus for two units was nearing completion and on schedule for installation underground. The data acquisition grid and anode electronics for module 1 was complete and tested using DRIFT I. The DRIFT IIa concept used commercial electronics with grouping of grid wires down to 8 Amptec channels, with PCI ADCs and 4 low threshold veto channels. The remote operation slow control hardware and code were also completed and tested and the data pipeline installed with analysis code. Finally, erection of neutron shielding underground had been started for module 1.

First Module

First DRIFT II module, DRIFT IIa

DRIFT IIa underground operations

Underground installation of the first DRIFT II module, DRIFT IIa, proceeded smoothly at Boulby in February 2005, being completed over 14 days with no significant technical issues or faults and with first data taking started on schedule. All safety documents were completed and passed by CPL. Both MWPCs and the field cages worked well and were stable at the design operating voltage of 3200V and 40KV respectively. All slow control, CS2 input/output, remote monitoring, electronics, computing and data collection sub-systems ran as expected as did both anode (x direction) and cathode (y direction) DAQ and all veto channels on both x and y. The channel grouping electronics was shown to work and the first 55Fe calibrations showed a threshold 3 times lower than that obtained in DRIFT I as expected. Some spark events were observed on one side of the detector with crossover to the other side but these were eliminated, the cause being a poor HV connection on the right hand detector.

By July 2005 the commissioning of DRIFT IIa was complete. For the latter part of May and early June DRIFT IIa was continuously run to demonstrate stability and response to calibration sources and to acquire unshielded background data. Installation of neutron shielding then started, consisting first of a supported outer shell containing ~9 tonnes of polypropylene pellets, prior to a planned fully shielded run of ~60 days. The full shield (15 tons of poly pellets) was completed in early August. Minor delays were caused by additional infrastructure safety issues (fire proofing, rear facility wall modifications and fire escape passage) and manpower demands for ZEPLIN II. However, the subsequent operational stability allowed time recovery. All remote monitoring systems, remote operation, the data pipeline and data archiving systems were completed. A DRIFT watch rota with 24hr monitoring of real time dark matter data and detector parameters by collaboration members was operated smoothly from July. Automated trips were fully functional and tested and the detector configured automatically to phone the duty officer in event of any system failure.

The first fully shielded run proceeded through the autumn of 2005. A slight increase in temperature after installation of shielding was observed but was mitigated by addition of air cooling through the shield. An initial total of 16 kg.days (~10 Tbytes), of underground DRIFT IIa data was accumulated including 10 kg.days fully shielded (the PPARC milestone) plus an additional 3 kg.days of partially shielded data and 3 kg.days of unshielded calibration and rock neutron background data. The first DRIFT II technical paper was submitted at this time (well ahead of the milestone) and promptly accepted for publication. The detector ran smoothly, with ~90% continuous live time over 2 months accumulating dark matter data at a better than expected rate of ~ 5kg.days per month at 40 Torr.

Commissioning

Underground commissioning of DRIFT IIa in Feb/Mar 2005

DRIFT IIa analysis preparation and first results

DRIFT analysis is very complex, requiring reduction and cut development to deal with an effective >2000 simultaneous readout signals with 10-bit pulse shape and time information at a raw rate of ~2 Hz. To tackle this two independent home-built code packages have been developed for analysis: (1) the UK-Sheffield SQL database system, mainly for 3d discrimination, alpha track reconstruction and low background/radon studies; (2) the US-Oxy/UNM package, mainly used for energy calibration and background rejection. In addition, UK-Edinburgh produced detailed background simulations.

First analysis work concentrated on using incoming DM and 55Fe data to finalize the data reduction code, streamline archiving for 1 Tb/month, confirm that behavior was as expected, characterize all noise event types, assess efficiencies and finalize parameter cuts. Preliminary monitoring and analysis revealed that the raw trigger rate was as expected; the spark rate dropped as the gas improved as expected; the high-energy alpha rate was as expected with the alpha vetoes working fine. First pass analysis showed sensitivity to neutrons and gamma rejection consistent with DRIFT I at >105. Particular effort was focused on analysis to minimize energy thresholds including noise rejection algorithms using FFT and boxcars.

Housekeeping

DRIFT IIa with shield (left); example house-keeping data for the cathode current as function of time (middle); typical rejected high-energy alpha event (right)

By May 2006 a period of intense collaborative effort on data analysis development and further data collection had resulted in significant progress being made in the development and validation of core analysis tools, with the subsequent application to the DRIFT IIa data giving us a much-improved understanding of the performance and capabilities of the detector. Analysis of the slow control and event-based run time indicators confirmed the stability of the detector throughout the first 5 months operation period. Event characterization and data cuts and flags had been developed to the point where essentially all but nuclear recoil events were rejected from the data. The following section details the analysis results from Mar 2005 up to June 2007, including OsC-relevant work undertaken after PPARC funding ceased in Oct 2005 funded at a low level using local and EU sources.

DRIFT II analysis summary: calibration, background, limits, directional analysis results

Prime results relevant here (and in publications) come from the near continuous run indicated above, the detector operated largely remotely via Internet. We believe these results, with subsequent data, successfully demonstrate the base power of the technology.

Nuclear recoil detection efficiency: vital to all dark matter experiments (often missing) is a detailed understanding of detection efficiency for nuclear recoils before and after all background cuts. Analysis of 252Cf source exposures clearly demonstrated the sensitivity of the detector to neutrons - with source on/off rates in excellent agreement with simulations. Comparison of data recorded before and after installation of CH shielding showed significant difference in accepted event rates as expected. The full results in this area are a major success of DRIFTIIa, achieved through many tests comparing gamma-ray, x-ray and neutron source data with in-house GEANT4 Monte Carlo simulations, including a difficult absolute flux calibration of a 252Cf neutron source (Chang thesis) and first measurement of the Boulby intrinsic rock neutron flux [6,12]. This revealed an intrinsic efficiency to neutron-induced recoils in DRIFTIIa (that is with minimal cuts in place) of 94±2(stat.)±5(sys.)% and an efficiency response function with all background cuts in place. These results confirmed also the superb power of DRIFT as a low flux recoil (neutron) detector.

Neutron Efficiency

neutron efficiency response of DRIFT IIa

Gamma rejection efficiency: the DRIFT gas-based technology uniquely allows very powerful gamma rejection via event dE/dX and track range information. With all developed software-based cuts designed to remove non-nuclear recoil events in place and neutron detection efficiency, the rejection efficiency was measured in detail using multiple 60Co sources and found to be better than 8×10-6 above 1000 NIP threshold [6]. This ecellent rejection power justified the exclusion of Pb shielding from the design of DRIFT at the WIMP sensitivity level epected and bodes well for scale-up designs.

Energy threshold and noise: energy calibration was successfully achieved every 6 hours using internal shuttered 55Fe sources. Most analysis used a threshold of ~1000 NIPs, limited by electronic noise. However, new work using software filters demonstrates <100 NIPs across the 1m3 volume (<2 keVe) is achievable even without replacement of current commercial CREMAT pre-amps. Long-term operational stability was also confirmed this way.

Energy Threshold

55Fe (5.9keV) response and noise reduction

Dark matter runs and radon progeny recoils: with cuts and efficiencies well understood, analysis proceeded of DRIFTIIa data obtained with 60cm poly neutron shielding in place. A significant and unexpected finding here was the identification of class of recoil-like background events in shielded data below 2000 NIPs. While it was known that radon from outside the fiducial volume could produce alphas that would be easily rejected in analysis, it took a detailed study and new tests to demonstrate the origin here to be alpha-decay of 222Rn daughter nuclei attached to the central cathode, resulting in ~100 keV 218Po Radon Progeny Recoils (RPRs). The main evidence is that these events and related contained 222Rn alpha background events, have the expected rate correlation with time following Cs2 flushes. Much successful analysis effort has proceeded to deal with the RPRs. In particular, new fiducialisation cuts were developed via pulse shape analysis based on track diffusion (near-cathode events like RPRs undergo greater diffusion prior to arrival at MWPCs). This successfully cut RPRs by ~x20. Application of this result yielded a final background rate in DRIFT IIa of ~10 counts/day (1000-2500 NIPs, 40% recoil detection efficiency), sufficient to generate trial limits [10]. Results of the analysis were presented at DM2006 (LA, 22-24th Feb 06) and included in [6].

RPR Rate

RPR background and rate vs. time after CS2 flush

Radon progeny recoil reduction: Residual background events are now well understood as remaining RPRs. Whilst certainly an annoyance, our understanding of the issue and how to handle it has greatly improved and we have good reason to be confident that RPRs can be reduced to acceptable levels in both DRIFT IIa and future detectors. In addition to fiducial pulse shape cuts as above, workable modifications to detector cathode field structures have now been developed, termed radon “traps”. However, most important is simple removal of the primary sources of the radon emanation. For this a Rn emanation test system (using a Durridge Rad7 Rn detector and emanation vessel) was built to test components for Rn emanation levels. All major Rn emanation sources for DIIa have now been identified (mainly BNC and Ribbon cables) [8]. This gives an excellent independent confirmation of the analysis findings and a system that can now serve as a Rn emanation material screening system for later DIIa refits (and for other experiments). The first science paper describing the performance of the detector, backgrounds and discrimination was delayed (first submitted in December 2006) because of the need to understand the RPRs. However, the work is possibly now a benchmark for understanding RPRs, something that is vital to all dark matter experiments, directional or not. The paper is currently awaiting referee reports.

Directional and range reconstruction with alphas: alphas are easily rejected in DRIFT but if specifically selected for analysis they provide a vital tool for optimising 3D track reconstruction, range discrimination and for detector characterisation. Results from extensive analysis and MCs include [7]: full determination of drift velocity using alphas, found to be (59.8 ± 1.4 m/s); 3D track generation with Bragg curve reconstruction; determination of radon sources and behaviour; and confirmation of SRIM range predications (SRIM2003 as found to be 10% out). The figure below shows an exceptional success here, a unique measurement of the populations of background nuclei via measurement of the alphas produced and separated by their 3D ranges in the gas using our track-fitting algorithm (a track range spectrum, not an energy spectrum). This work has formed the basis of our second science paper [7].

3D reconstruction of nuclear recoils and production of 3D sky maps: of ultimate interest is 3D reconstruction of low energy recoil tracks and hence production of direction sky-maps in galactic coordinates for correlation with our motion. This is also well advanced now. Firstly, basic x-axis directional analysis has been completed using directed neutron runs and hence 1D directionality demonstrated (Muna Thesis) [9]. More importantly software is in place to reconstruct S recoils in 3D. The figure below (centre) shows an example 3D event (with dE/dx along the track indicated by the circle size) and the right figure shows a preliminary sky-map directional plot from directed neutron data, together with results of a typical simulation. Further, work is needed now to improve reconstruction of z and y detector coordinates, made harder than x due to the need to unscramble grid/anode crosstalk and electronics overshoot, plus study of the possibility of head-tail discrimination. Software reconstruction is being implemented to achieve these tasks.

Range Spectrum
3D range spectrum of radon chain alphas allowing identification of 222Rn, 218Po, 220Rn, 216Po		 3D Reconstruction
example ~100 keV, 3D Sulphur recoil track reconstruction from DRIFT IIa with dE/dx. This allows extraction of θ, φ orientation of events 		Sky Map
example 3D angular direction sky-map reconstruction of nuclear recoil events in DRIFTIIa and simulation (inset). Gaps are due to dead regions along x,y,z, axis.

Head-tail discrimination: finally new simulations by us (Majewski et al.) using SRIM2007, in agreement with independent work by Hitachi et al., indicate that head-tail discrimination of low energy S recoils should be feasible

Track
example 100 keV recoil track simulation in 40 Torr CS2 showing clear head-tail difference 		Installation
DRIFT IIb installation

DRIFT IIa impact and forward look

This UK-based UK-US programme features on the recent US DMSAG, EU ASPERA and ApPEC roadmap documents, is internationally recognised as important, world-leading and unique. DRIFT IIa has successfully demonstrated at the few m3 scale the possibility of building and operating underground a low maintenance, basic negative ion gas TPC dark matter detector. The results support earlier claims of DRIFT's potential for low (<20 KeV recoil) threshold detection with a raw (non-directional) sensitivity at 10-6 pb per year of operation per 1m3 of 40 Torr CS2, provided the RPR issue is resolved. Clear routes to achieving RPR removal have been identified. However more ssignificant new R&D is needed now, specifically: (1) to determine fully the fundamental angular resolution and application in a realistic scaled-up experiment; (2) to confirm experimentally new simulations indicating head-tail discrimination is possible, (3) to understand how newly developed micro-pattern charge readout technology such as micromegas and GEMs can be used to lower costs and improve sensitivity. UK progress has been made on the latter recently via first successful demonstration of micromegas with CS2 [13] and a new large UK-US-EU cooperation is forming to pursue all these goals [4]. An upgrade module, DRIFT IIb, with reduced radon and lower noise is now operational.

References

[1] B. Morgan et al. Phys. Rev. D71 103507 (2005)
[2] B. Morgan et al., Astropart. Phys. 23, (2005) 287
[3] N. Spooner et al., Proc. IDM2006 (Rhodes, 11-16th Sep., 2006) tbp World Sci. (2007)
[4] http://www.hep.shef.ac.uk/conferences/cygnus2007
[5] G.J. Alner et al., NIM. A, 555 (2005) 173
[6] S. Burgos et al., sub Astropart. Phys. (2007)
[7] S. Burgos et al., TBP Astropart. Phys. (2007)
[8] S. Burgos et al., TBP NIM. A (2007)
[9] N. Spooner. J, Phys. Soc. Japan http://arxiv.org/abs/0705.3345
[10] E. Tziaferi, PhD thesis, University of Sheffield (2007)
[11] C. Chang, PhD thesis, University of Edinburgh (2007)
[12] E. Tziaferi et al. Astropart. Phys. 27 (2007) 326
[13] P.K. Lightfoot et al., Astropart. Phys. (2007) doi:10.1016/

Valid HTML 4.01 Strict