The ZEPLIN II Detector
The ZEPLIN II detector is a two phase xenon detector using 40kg of xenon, constructed in collaboration with UCLA and Texas A&M. Increased discrimination over single phase operation is achieved through the simultaneous measurement of the primary scintillation and ionisation. The ionisation is observed through charge extraction into the gas phase leading to electroluminescence. Nuclear recoil signals are characterised much smaller ionisation signals. ZEPLIN II was designed to fit into the ZEPLIN I veto and castle, which allowed re-use of these systems.
The vacuum and target vessels were delivered to RAL from our US colleagues during Spring, 2003. Initial integrity and vacuum tests proved successful, although the casting process used for these components showed material inclusions. During final machining of the vessels in the UK in Summer/Autumn 2003, these inclusions proved difficult to control and ultimately led to the development of a vacuum leaks on both vessels. These leaks were successfully sealed for cooldown tests, but the expectation was that the technique used would contaminate the xenon. A new target chamber was therefore ordered, using electron beam welding and plate machining.
Cooldown tests, using argon within the existing chamber, successfully reached below liquid xenon temperatures (-126°C) and characterisation of the detector heat load was done. Manufacturing of the connecting neck, inlet port structure and associated pipework and flanges was completed and all vacuum systems constructed. The passive xenon dump chambers were constructed, cleaned, baked and tested to a vacuum of better than 10-9 torr. The components for the purifier were sourced and it was then assembled together with coolable xenon bottles and associated liquid nitrogen delivery. All photomultipliers, 9390QFL 5 inch hemispherical fast tubes with platinum underlay for low temperature performance, were delivered from ETL. These PMTs were extensively tested at Sheffield, and all passed initial cooldown gain and integrity tests and were fully characterised for uniformity. The DAQ system was used in tests at Texas A&M and is based on Acquiris PCI digitisers, as had been used for ZEPLIN I and NaIAD. The code for the DAQ system was extended from that developed for ZEPLIN I and tested on a xenon gas chamber at Texas. Slow control systems were constructed for pressure and temperature monitoring, and integrated into the slow control software. A bake-out control system was developed for both purifier and chamber.
The transfer of the ZEPLIN I infrastructure from the old facility into the new JIF laboratory was underway at the time of the October 2004 application. The shield and veto had been decommissioned, although transfer to JIF required CPL to mill out a roadway which had been blocked by a conveyor system. The detector itself was still being tested in the RAL surface laboratory.
By April 2005 the ZEPLIN II internal target chamber components had all been transferred to the new copper target chamber. Following this the complete inner target chamber has been installed into the outer vacuum jacket and the cold head refitted together with all the cabling. The complete assembly had been connected up and a first surface run of the whole system had been carried out. That run achieved a full load of 40kg of xenon liquefied in the target very efficiently, the thermal system design was validated, the gas delivery system worked above expectation, filling in 6 hours, the cooling system was fully validated, as expected from the earlier cooling tests, the photomultipliers worked to specification in the cold rig, scintillation signals were seen from a number of radioactive sources, the light collection efficiency for primary scintillation was measured as ~1 p.e./keV which was consistent with simulations and the xenon was recovered safely and efficiently from the target with no losses. The target was opened just after this test to rectify a wire disconnect which had occurred on that cool-down and which had prevented a full test of two-phase operation. A second test was planned to be done in the following weeks along with a pre-transport review to approve shipment to Boulby. A few additional ancillary items were in manufacture, including an electron lifetime monitor and a calibration source delivery system and a recirculation system as risk mitigation against not achieving sufficient xenon purity.
Surface testing of ZEPLIN II and underground preparation of the shield and veto
Preparations at Boulby for deployment of ZEPLIN II were completed at this stage to a level to allow deployment. An Underground Project Implementation Plan and associated health and safety reviews had been completed, leading to CPL and Boulby Management Board approval for deployment. Clean room operation of the Boulby facility had commenced, following some required refurbishment of the building structure due to roof movement. The ZEPLIN II castle floor section had been excavated and completed, incorporating new under-floor shielding. The castle construction itself was completed as far as possible before deployment using lead from the ZEPLIN I castle, including a specialised cleaning process. The veto system from ZEPLIN I had also been installed in the castle, with PMT refurbishment underway.
By July 2005 the wire disconnect in the HV supply to the grids had been fixed and a second surface run completed. Two-phase signals from the bulk liquid were seen from both γ-ray and neutron induced events. The time delay histogram between the primary and secondary signals confirmed that the secondary signals were coming from ionisation that had drifted up to several millimetres through the liquid. This verified that the most critical criterion had been achieved. In addition it was seen that when using a neutron source an additional population of events appeared with lower levels of ionisation. Unfortunately the HV tests also revealed some problems with the grid support/insulator and feedthrough/cabling designs. These problems resulted in surface leakage breakdown and gas breakdown respectively which limited the applied voltage differentials to ~4kV. The design specification was 20kV and this was needed to ensure useful discrimination between γ events and nuclear recoils. Rectification of these HV problems required two actions. Firstly the PTFE structure supporting the grips needed to be redesigned and remade. Secondly the HV connectors and cabling had to be replaced. These changes required opening up the target chamber again and stripping out the internals. The ability to achieve two-phase operation despite the HV problems meant that it was possible to proceed with the xenon purification whilst the new HV components were being manufactured. In addition the electron lifetime monitor from ZEPLIN III was also working and had been modified for use with ZEPLIN II. A decision was thus made to go ahead and install ZEPLIN II in the underground laboratory on the grounds that this made best use of the time. ZEPLIN II was transported and reassembled in the mine. All vacua had been re-established and the whole system was now ready for xenon purification to begin.
The first underground xenon liquefaction and target fill had been achieved, with HHV delivery and two phase operation validated.
Electron lifetime monitor viewport showing the liquid xenon level. On the right is shown a signal from an α-particle interaction in the gas phase. The change of slope just before the first horizontal division happens as the electrons enter the liquid phase. The amplitude of the signal change in the liquid phase is a measure of xenon purity.
ZEPLIN II transported and re-assembled in the underground laboratory
By November 2005 the HHV feedthroughs and extraction field grids had been redesigned, remanufactured and installed. The first cooldown and liquefaction attempts, following the HHV upgrades failed to liquefy a full charge. This problem was diagnosed as a cooling power problem on the commercial PolyCold cryogenerator, which was subsequently fixed during September. The electron lifetime monitor (ELM) was integrated underground, but required repair following a heater failure during operation. These repairs were completed, the ELM re-installed and operation validated.
Following these changes the first complete liquefaction cycle was achieved during early October. The fill process required four days, rather than the 36 hours required in the surface tests, to allow the target to stabilise between xenon fills. This is thought to be due to changes in the internal volumes, and has implications for the timescale of target cleansing cycles. During this fill the lifetime of the xenon gas being delivered was measured at 70µs, which corresponds also to the drift time required from the bottom of the target. The HHV systems were validated, with 22kV delivered to the bottom cathode, in excess of the original design requirement. Additionally the extraction grids were held at 8kV, showing that the surface charge leakage issue observed in the original design has been resolved. With the grids energised full two phase operation was observed, with single phase and two phase commissioning data samples taken over a period of five days. These runs validated the operation of the DAQ system, PMT operation and trigger hardware. After the run the full xenon charge was successfully extracted from the target, through the purification getter, validating the extraction process had not changed with this new configuration. Plans were put in place to start recirculation to aid the purification process. Two designs were being pursued, a passive system for which the internal target pipes and heaters were already installed, with additional pipework constructed, and an active system that would require an impellor pump. The design of a recirculator pump had been completed, with assembly underway at UCLA.
The ZEPLIN II underground assembly, castle roof and side shielding not mounted and first two phase signals from full underground chamber.
By May 2006 the ZEPLIN II detector was fully operational, with sufficient xenon purity to drift charge throughout the chamber with negligible loss. Repeated xenon fills and stable operation had been demonstrated, validating hardware, data pipeline and data reduction procedures. Operational parameters had been fixed for the first blinded background runs. The detector was routinely collecting background and operational calibration data. A recirculation pump had indeed been introduced along with control and alarm systems. The failsafe pump constructed at UCLA had failed, leading to the adoption of a commercial pump with a failure mode which risked dumping the xenon to atmosphere. A 24hr shift operation was temporarily introduced to manage this potential failure mode. Operation of this pump, with a high flow rate of >10 l/min, dramatically improved the xenon purity to >500µs levels, beyond which it was impossible to measure. Coupled with the improvement in purity, the light yield from the detector also improved. The zero-field light yield is consistent with light collection simulations, at 1.4 p.e./keV.
All PMT gains were matched to better than 10%, and the relative quantum efficiency values were calculated for the outer six symmetrical tubes. The detector was levelled to within 0.2mm, using the drift time measured from the data itself to determine the distance from the bottom grid to the liquid surface level. Similarly the distance from the liquid level to top grid was measured using the secondary pulse width. In addition the level of the liquid had also been characterised by monitoring the capacitance between the grids during a deliberate overfill of the detector, which matched the expected capacitive profile between the extraction grids. Voltages for the extraction grids and cathode were selected for the first background runs, to provide near unity charge extraction from the surface, and a conversion gain in the electroluminescence region of >100 photons/electron. The cathode voltage was held to provide 1kV/cm in the drift volume, a safe operational voltage that had been already demonstrated.
Customised radioisotope sources were delivered and installed into the automatic source dropper during December. This allowed full castle closure to provide a fully shielded detector. Operation of the source dropper had been validated, with delivery of 57Co sources beneath the target used for daily energy calibrations. This also provided a monitor of charge collection from the bottom of the chamber, as the 122keV gamma interaction length is only a few mm into the xenon volume.
The liquid scintillator Compton veto that surrounds the target had also been fully commissioned. The scintillator had been purged with argon, and operation of all PMTs validated with matched gains. The light yield from the veto was 0.03p.e./keV, which provided a 100keV threshold using a 3 from 10 trigger. This is identical to the operational parameters when the veto was first commissioned for ZEPLIN-I. The data stream for the veto comprised a summed output for calorimetery and a NIM trigger pulse, both recorded on the Acquiris digitisers.
Left: evolution of liquid xenon purity as a function of mass recirculated. The operational objective was 200µs. Right: first AmBe source calibration run, with populations due to gammas, elastic neutron scatters, inelastic neutron scatters and PMT noise events clearly identified
Additional hardware modifications to the system had been made to improve the operational procedures and stability of the detector. These include replacement and testing of the controlled pressure relief leak valve to the passive dump chambers, replacement of a pipe on the purifier that had developed a leak, the addition of an isolation valve to one of the xenon bottles, replacement of the HHV control potentiometers to improve voltage stability and subsequent introduction of automatic safety cut-out valves and control system on the recirculator pump, to ensure a failsafe operation.
Two unplanned power shutdowns to the JIF facility occurred. One was due to CPL, inadvertently shutting down the entire Southern district, the other due to a choke failure in a light fitting. ZEPLIN II continued operation during both failures, the latter not affecting experiment power due to the planned separation of supplies. These incidents, although unfortunate, showed that the safety margin is as expected (~10hrs without power) and safety mechanisms were operational.
The data pipeline has been re-engineered with the installation of the tape robot underground. In addition an analysis machine has been incorporated underground, to provide immediate off-line diagnostics capability. This machine allows the purity and light yield to be measured routinely on a daily basis, and a 'quick-look' at the background and calibration data to be performed to provide data integrity checks. Data are written simultaneously to two tapes within the robot, one remaining at Boulby for security, the other distributed to Sheffield/RAL for full off-line analysis.
With the achievement of required purity and light yield a series of commissioning data runs were taken, to assess the detector performance, data management and analysis procedures. Data runs taken included 60Co Compton, 57Co energy calibration, AmBe neutron calibration and background data runs. These runs have been used to provide an initial look at detector performance, showing clear discrimination between electron and nuclear recoils, stability of the detector during a two week run, position sensitivity and an initial estimate of the detector background. The analysis chain is split into five groupings - data reduction, stability, efficiencies, discrimination and backgrounds. The groups are using these data to determine analysis procedures and parameters.
In parallel, the detector was collecting first operational science data, with 10% unblinded for monitoring purposes. Extended gamma and neutron calibrations will be undertaken as part of this science data taking.
By October 2006 the first dark matter run had been completed, with a total exposure of ~1.5 tonne.days, which fiducialised down to ~200kg.days of background data when all cuts were applied. These data yielded a sensitivity of ~1.6x10-6pb from the unblind 10%. However unblinding the rest of the data highlighted an unexpected event population in the box, which limited the overall sensitivity from the run. These events have been identified as radon daughter ions migrating to the PTFE walls, and plans for their removal in a second data run are underway. Detailed efficiency, detector stability and event population analyses were completed, providing an excellent understanding of the detector and ancillary systems, including the Compton veto.
The final analysis of the data from the first dark matter run of ZEPLIN II has been completed and yielded a 90% confidence limit sensitivity at the 6.6x10-7pb level at the minimum of the sensitivity curve. The result has now been published in a fully refereed journal.
Simulation studies have been performed in ANSYS, Guideit, FAUST and GEANT4 to assess the ZEPLIN II target characteristics, including drift field characteristics, primary light yield and light collection uniformity, and the expected neutron background when ZEPLIN II at Boulby. These studies showed that the main background (80 recoils/yr out of 120 expected in the shielded configuration) arose from U/Th contamination in the PMT array, which was reduced by conversion of the Compton veto to a neutron veto through Gd loading. GEANT4 simulations were used to characterise the detector performance for cross comparison against the laboratory data.
A comprehensive background budget was compiled. This brought together and consolidated previous different studies made using a variety of techniques. Consistency was established and predictions verified that ZEPLIN II would, in the absence of the unexpected Rn induced background, achieve competitive sensitivity to CDMS II within a few months of running.
By July 2005 preparations for data analysis were well underway. A first version of a new off-line data analysis software had been released and tested on surface commissioning data. Version 2 was about to be released. In addition a first release of a fully featured simulation data set had enabled further software testing.
The data pipeline required the installation of the tape robot underground. In addition an analysis machine was incorporated underground, to provide immediate off-line diagnostics capability. This machine allowed the purity and light yield to be measured routinely on a daily basis, and a 'quick-look' at the background and calibration data to be performed to provide data integrity checks. Data were written simultaneously to two tapes within the robot, one remaining at Boulby for security, the other distributed to Sheffield/RAL for full off-line analysis.
With the achievement of required purity and light yield a series of commissioning data runs were taken, to assess the detector performance, data management and analysis procedures. Data runs taken included 60Co Compton, 57Co energy calibration, AmBe neutron calibration and background data runs. These runs were used to provide an initial look at detector performance, showing clear discrimination between electron and nuclear recoils, stability of the detector during a two week run, position sensitivity and an initial estimate of the detector background. The analysis chain was split into five groupings – data reduction, stability, efficiencies, discrimination and backgrounds. The groups were using these data to determine analysis procedures and parameters.
In parallel, the detector was collecting first operational science data, with 10% unblinded for monitoring purposes. Extended gamma and neutron calibrations were taken as part of this science data taking.
The dataset from the first extended background run was analysed with a blinded analysis. 10% of the data were extracted into a 'golden' dataset, with the remaining 90% designated as the 'platinum' dataset. Significant progress was made with the data reduction and analysis software, leading to the completion of the analysis of the golden dataset and the ability to study analysis procedure, selection rules, detector efficiencies and detector performance and stability. The analysis procedure was based on the extraction of events which may be dark matter candidates, and used cuts such as only allowing single scatters within the detector, selection based on the secondary to primary signal ratio and position reconstruction into a fiducial volume away from the walls. In addition an extensive set of efficiencies were quantified, including detector response, Compton veto response, analysis efficiency and physics efficiency. The culmination of these analyses was to define an acceptance region within the S2/S1 vs S1 parameter space, for the selected events, which has a 25% nuclear recoil acceptance efficiency. For the golden dataset this generated a background free acceptance region, for an exposure within the fiducial volume of ~200kg.days, and an associated sensitivity of ~1.6x10-6pb.
Based on the golden data selection rules the platinum data box was then opened, on September 10th. This revealed a population of events with low secondary signal, which had been statistically suppressed in the golden dataset due to the fiducialisation of the detector. Analysis of this population has allowed us to determine that these events are nuclear recoils off the PTFE walls of the chamber, where the liberated charge is suppressed either during production or through transit. These low S2 events distribute the secondary signal across the seven PMTs and have a finite probability of being placed in the central fiducial volume through statistical fluctuations, thereby entering the acceptance box. Analysis of the low S2 event population indicates that there is insufficient information to remove them directly from the dataset, and the sensitivity of the detector for this first run will therefore not be in a "background free" mode. Although the limit has been constrained by this low S2 event population, the benefit of the blind analysis in determining the acceptance box is clear, as it defines the gamma background acceptance region. The expected number of events within the acceptance box can be clearly determined from both the gamma and low S2 population, using higher energy data.
A final iteration on the stability corrections database was done, refining the time dependence of the electron lifetime within the xenon volume. This database also corrects the S2 signal for pressure, detector temperature, xenon level and associated field changes. On completion of this final iteration the expected numbers of events within the acceptance box was calculated and a sensitivity derived.
Left: Scatter plot of event populations for the golden data set. The red dots are events from the AmBe neutron calibration, the black dots are events from the data run. The acceptance band for 25% nuclear recoils is shown in blue.
Right: Detailed efficiency calculation for the platinum dataset, comprising the combination of the 31 efficiencies studied within the analysis chain. Shown are three different trigger settings.
A re-analysis of the stability database used for the first background run was done. This database has two components - the electron lifetime within the xenon liquid, and the variation in electroluminescence yield (S2). The electron lifetime was calculated by comparing the relative yield between recoil events from the bottom grid and the cathode. The electroluminescence stability was assessed by monitoring the variation of S2/S1 for nuclear recoils arising from the cathode. These corrections account for variations in S2 due to effects such as charging of the liquid surface, pressure and temperature fluctuations within the target, and hence changes in the xenon liquid level between the two extraction grids. This correction database was generated with a finer resolution than previously, allowing event by event corrections.
Application of this database tightened the background data distribution slightly, allowing the acceptance box to be extended to 50% nuclear recoil acceptance. The boundary was recalculated from the AmBe neutron calibration data, following reapplication of the corrections database. An issue with the application of one of the data cuts was also identified during this process, which mis-labelled certain primary signals, and this was rectified. The final (platinum) analysis of background data and calibration data was then completed.
Left: Science data from the complete 225 kg.day exposure of ZEPLIN-II. The upper plot shows events that also have a signal recorded in the liquid scintillator veto, the lower plot has these events removed. The nuclear recoil acceptance window used for the dark matter analysis is shown, with the 50% nuclear recoil acceptance boundary extended across the energy range. Also shown are two contours of constant S2, showing that the radon progeny background events observed in the lower S2/S1 population have a fixed S2 distribution.
Right: Calibration using neutrons from an AmBe source (upper) and Compton scattered 60Co gamma-rays (lower). These calibrations were performed at a high trigger rate, leading to a small, uniform, population of coincidental events distributed throughout the S2/S1 parameter space. These coincidentals are verified by comparison with events with unphysical drift times. Also shown are the S2/S1 boundary for 50% nuclear recoil acceptance and the acceptance window used in the dark matter analysis.
The acceptance box for nuclear recoil signals is also shown. Table 1 details the exposure of the background run and the effects of different data cuts applied, showing that the final fiducial volume exposure is limited by the requirement to remove wall events through the application of a radial cut.
The population of events at low S2/S1 which rises at low energy into the acceptance box is the radon daughter-initiated events from the PTFE walls. These events have small secondary signals and hence a poor position reconstruction accuracy. Although clearly from the PTFE walls - the radial distribution peaks on the wall - the position reconstruction uncertainty allows a small fraction into the acceptance volume (r < 7cm).
A detailed analysis of the expectation counts within the acceptance box for the gamma population and the radon-derived noise population has been completed. The gamma expectation was calculated through two routes: analysis of the 60Co gamma calibration data, and analysis of the background data itself, extrapolating the population from higher S2/S1 parameter space. The 'radon' noise population expectation was calculated by extrapolation of events in the box as the radial cut is varied. The table below shows the final values for the observed and expected counts within the acceptance box.
Left: Final spin independent WIMP-nucleon scattering cross section limit for the May-July '06 science run.
Right: Final spin-dependent WIMP-neutron scattering cross section limit for the May-July '06 science run. Separate limits on two isotopes sensitive to spin-dependent interactions are also shown.
The observed and expected counts were converted to a limit on the number of nuclear recoils using the Feldman-Cousins (F-C) technique, where the values are calculated using the F-C routine within the ROOT analysis environment. This lead to an upper limit on nuclear recoils of 10.4 events, which when converted to a cross section limit using the detector response matrix and related efficiencies, and the canonical WIMP halo model, yields a 90% confidence level upper limit to spin-independent WIMP-nucleon interactions of 6.6x10-7 pb at the minimum of the sensitivity curve. Similar analysis performed in the framework of WIMP-neutron spin-dependent interactions gives the second limit curve shown with a minimal value of 7x10-2 pb. This limit is comparable to the CDMS result.
Exposure summary for first science run
(31 kg target mass)
|Milestone exposure for first run|
(31 kg target mass)
| 57 days |
| Overall length of run following operational|
|Science data run|
(31 kg target mass)
| 44.2 days |
| Background data exposure, removing|
calibrations and system maintenance
(31 kg target mass)
| 31.2 days |
| Removing days experiencing field |
|Fiducial cuts (drift time) |
(26 kg target mass)
| 31.2 days |
| Fiducial cut in z to remove cathode and grid |
|Fiducial cuts (radial) |
(7.2 kg target mass)
| 31.2 days |
| Fiducial cut in x, y to remove PTFE wall|
Overall expectation values compared to observations for platinum.v8.
|Energy Range||Observed (v8)||Radon derived events||Gammas (data)||Gammas (60Co)||Total|
|5-10 keV||14||10.2 ± 2.2||5.6 ± 4.6||4.2 ± 2.4||14.4 ± 3.3|
|10-20keV||15||2.3 ± 0.5||13.0 ± 6.0||11.9 ± 2.7||14.2 ± 2.7|