A Prototype Detector for Directional Measurement of the Cosmogenic Neutron Flux
AA Prototype Detector for Directional Measurement ofthe Cosmogenic Neutron Flux
J. Lopez, K. Terao, J.M. Conrad, D. Dujmic, L. Winslow
Physics Dept., Massachusetts Institute of Technology, Cambridge, MA 02139
Abstract
This paper describes a novel directional neutron detector. The low pressuretime projection chamber uses a mix of helium and CF gases. The detectorreconstructs the energy and angular distribution of fast neutron recoils. Thispaper reports results of energy calibration using an α source and angular re-construction studies using a collimated neutron source. The best performanceis obtained with a 12.5% CF – 87.5% He gas mixture. At low energies thetarget for fast neutrons transitions is primarily helium, while at higher ener-gies, the fluorine contributes as a target. The reconstruction efficiency is bothenergy and target dependent. For neutrons with energies less than 20 MeV,the reconstruction efficiency is ∼
40% for fluorine recoils and ∼
60% for heliumrecoils.
1. Introduction
Precise measurement of the energy and direction of cosmogenic-muon-inducedfast neutrons, with various levels of underground shielding, will be valuable forfuture neutrino and dark matter experiments [1]. In this paper, we presenta first-generation detector for this purpose. The detector design is based onthat for the directional dark matter search DMTPC [2]. We introduce modestmodifications for directional neutron detection[3]. Our first application will bein the Double Chooz (DC) neutrino experiment [4], and it is, therefore, calledDCTPC.DCTPC is a low pressure time projection chamber (TPC) filled with twogasses: (1) He, the fast neutron target, and (2) CF , a scintillant and a quencherfor the electron avalanche. This detector is blind to x-rays and minimum ionizingparticles (MIP), like muons, because the density of primary ionization is toosmall to be detected. The ionization electrons from a non-MIP particle trackdrift down to a stainless steel ground mesh. Between the ground mesh andthe anode, an avalanche of electrons occurs. A charge-coupled device (CCD)camera installed in the detector images the visible scintillation light from theavalanche. The charge created in the avalanche is readout from the anode plateand ground mesh. The track direction and energy can be reconstructed usingthe CCD image and the charge information. The reconstructed track from the Preprint submitted to Elsevier November 5, 2018 a r X i v : . [ phy s i c s . i n s - d e t ] J a n ecoiling nucleus can then be correlated to the incoming neutron direction andenergy. In this article we present on-surface measurements from a small 2.8 liter,first-generation detector for an α -source run and a Cf neutron source run aswell as an on-surface cosmogenic run.
2. Motivation for DCTPC
DCTPC provides a benchmark for the Double Chooz fast neutron simula-tion at the specific sites of the two Double Chooz neutrino detectors [4]. Thedetectors sites have 114 and 300 meters water equivalent (m.w.e) shielding, re-spectively. The final DCTPC design calls for a large detector located at eachsite. The small first-generation detector described here will be installed in the300 m.w.e site for initial studies. The neutron flux energy and angle measure-ments from the rock can be used to tune the Monte Carlo and demonstrate aclear understanding of this background in the Double Chooz neutrino oscillationanalysis. However, the results will provide wider benefits as backgrounds fromfast neutrons produced in rock are an issue for all low background experiments.Low energy neutrons arise from ( α , n) interactions due to natural radioactiv-ity in the shielding rock. At locations with low shielding, this neutron source isoverwhelmed by neutron production from cosmic ray muons. However, muonsare attenuated with depth and the ( α , n) background becomes significant at300 m.w.e. [5]. DCTPC, with its very low energy threshold, can sample these( α , n) events. The comparison of the 114 m.w.e. to 300 m.w.e. depths at DoubleChooz will be useful in isolating the ( α , n) from the muon-induced neutrons.The remaining two broad categories of neutron production are caused by cos-mic ray muon interactions with rock: stopped muon capture and deep inelasticinteractions. The rates depend upon the depth of the final experiment becauselow energy muons are attenuated with shielding. DCTPC sensitivity will reachto >
20 MeV in fast neutron reconstruction, overlapping with the KARMEN2(surface) and LVD (3200 m.w.e.) data sets. As we show below, the maximumreconstructed recoil energy depends upon the size of the detector.The FLUKA [6, 7] and GEANT4 [8, 9] simulations are generally used topredict >
10 MeV fast neutrons that are produced by muons. DCTPC addsdata points at two depths which will complement the existing measurements.Information on the energy spectra of fast neutrons is scarce, coming only fromKARMEN2 experiment on the surface [10], KamLAND at 2700 m.w.e. [11]and the LVD experiment at 3200 m.w.e. [12]. DCTPC data will fill the gapin depth between KARMEN2 and deeper experiments. DCTPC also providesan angular distribution of events. In this paper, we show the quality of thisangular reconstruction. The unique directional information can help disentangledifferent neutron sources.
3. The First Generation Detector
We have developed a first generation of DCTPC. The TPC has a drift volumeabove an amplification plane that has two active regions. The amplification2CD (cid:1)(cid:1) (cid:65)(cid:65) -HV cathodemesh (cid:0)(cid:0)(cid:64)(cid:64) (cid:113)
Route2ElectronicsHS-AMP-CF-2nF groundmesh (cid:0)(cid:1)
RRRRRRRRRR (cid:64)(cid:64) (cid:0)(cid:0)
CrematCR-112veto veto +HV R B R B (cid:64)(cid:64) (cid:0)(cid:0) CrematCR-113anode (cid:81)(cid:81)(cid:81)(cid:81)(cid:81)(cid:81)(cid:81)(cid:81)(cid:81)(cid:81)(cid:81)(cid:81)(cid:107) neutron (cid:104) (cid:16)(cid:16)(cid:16)(cid:16)(cid:16)(cid:41) He (cid:113) (cid:113)(cid:113) (cid:113)(cid:113)(cid:113) (cid:113)(cid:113)(cid:113) (cid:113) Figure 1: A schematic of the detector: the drift field is created by a cathode mesh, field-shapingrings attached to a resistor chain, and a ground mesh. Primary ionization from a recoilingnucleus drifts down to the ground mesh. The high-field amplification region is formed by theground mesh and the anode plane. The ground mesh is read out with a fast amplifier and theveto and anode are readout with charge-sensitive preamplifiers. Scintillation light from theamplification region is recorded with the CCD camera. plane is imaged with one CCD, and the charge is read out for each active region.The redundancy in the readout of the amplification plane is powerful for eventreconstruction and background rejection. The detector schematic is shown inthe Fig. 1 and a photograph of the detector vessel is shown in Fig. 2.
The detector follows the classic design of TPCs. We have chosen a mixture ofCF and He as our operating gas. We study several gas ratios below, but mostdata were taken with the gasses mixed at 75 Torr and 525 Torr respectively.Helium and fluorine both have large cross sections for fast neutron scattering[13]. The He nucleus is closer to the neutron mass, and therefore the recoilswill produce longer tracks than the fluorine nucleus.Within the vessel, a uniform drift field is molded by a series of copper ringswith an inner diameter of 26.9 cm. The rings are connected with resistors, seeFig. 1. The top ring carries a high transparency stainless-steel mesh with wirepitch of 512 µ m and wire diameter of 31 µ m. This top mesh is maintained ata negative high voltage. The total height of the field cage is 10 cm. Primaryionization electrons are drifted down to a grounded stainless-steel mesh withwire pitch of 256 µ m and wire diameter of 30 µ m.3 igure 2: Left: The first generation detector. The detector vessel is sitting on top of thecart which holds the readout electronics. The CCD camera is mounted on the top flange.Right-top: A diagram of the anode plane. Right-bottom: A photograph of the field cageand amplification plane of the TPC. The ground mesh is difficult to see, however the spacersseparating it from the anode plane are visible as horizontal lines. The amplification region is created by the ground mesh and the copper anodethat are separated with insulating tubes, called spacers, 440 µ m in diameter,placed 2.56 cm apart [14]. This high-field amplification gap multiplies the pri-mary ionization charge by a factor of approximately 10 , as we discuss below.The anode bias scheme is shown in Fig. 1, where resistor R B is 210 MΩ.The anode of the amplification plane is divided into two regions. The innerpart of the anode, diameter 24.7 cm, is used for the detection of tracks. A vetoring, with inner diameter 24.8 cm and outer diameter 26.8 cm, surrounds thecentral plane and is used to veto charged particles originating from decays fromwithin the detector components. A central square region, 16.7 cm × In order to select the gas mixture, we studied three gases: 100%CF , 12.5%CF – 87.5% He, and 6.25% CF – 93.75% He. Operating voltages, listed inTable 1, were optimized for each gas. However, we had difficulty finding suitableoperating conditions for the 6.25% mixture. Small variations in voltage causedvery large gain differences. We also found that the absolute and relative energycalibrations were inconsistent over several gas fills with this mixture, likely due4F P V anode V drift G CCD G anode Gas Threshold % [Torr] [V] [ADU/keV α ] [mV/keV α ] gain [mV]
100 75 680 -1200 16.3 0.346 1 . × > . × > . × Table 1: Summary of detector calibrations for running in the various gas mixtures. We reportthe run conditions for the 6.25% mixture here, however due to instability, we do not report onthis mixture further (see text). The trigger threshold is that used for neutron and backgroundrunning. Higher thresholds are needed for the energy calibration runs. to the much steeper increase of gain with respect to anode voltage. Therefore,we rejected this mixture and do not report the results of the 6.25% CF below. The CCD camera records scintillation light created in the amplification re-gion, see Fig. 1. The camera is the U6 model made by Apogee InstrumentsInc., with the Kodak KAF-1001E CCD chip. The chip is kept at − ◦ C tominimize the dark current. An image from the amplification plane is focused tothe CCD chip using a Nikon lens, of focal length 55 mm and set to f/1.2. TheCCD chip consists of 1024 by 1024 square pixels where the side length of a pixelis 24 µ m. The active area on the anode plate is a square of 16.7 cm × µ m. The unit for one pixel count returned by the CCD readout is an ADU(Analog-to-Digital Unit).The camera does not use a shutter and the CCD sensor is always live, evenduring the CCD readout. This leads to two types of exposures: a nominalexposure lasting 1 second, and a shorter (typically 250 ms) ‘parasitic’ exposuretaken during the CCD readout and event processing. Track images recordedduring the short parasitic exposure can be shifted in the direction of readout ofCCD rows, and do not have corresponding signal in the charge readout channels.Therefore, they are easy to remove during analysis. The charge readout is very powerful in suppressing CCD-specific backgrounds( e.g., direct hits in CCD chip and residual bulk images). This is the major im-provement over the CCD-only approach described in Ref. [3]. The anode platecontains three regions separated by electrically insulated gaps, as shown in Fig 2.The outermost circular region is connected to a common ground, onto which theground mesh is attached. The inner disk region is the actual anode, maintainedat 680 V. The circular-ring region between is used as a veto and is kept at thesame potential as the anode.When a particle enters from outside, its path must cross the veto region.Therefore, we can detect this type of event by inspecting the charge signal from5he veto region. The veto is not applied at the trigger level. The veto signal isstored for use in the analysis.The anode signal is collected using a charge-sensitive preamplifier (CSP),the Cremat CR-113 with a gain of 1.5 mV/pC. The pulse height of the outputof the CSP corresponds to the total charge collected on the anode. Therefore,the pulse height is used to reconstruct the total energy of the event. The vetoelectrode is read out through the more sensitive CSP, the Cremat CR-112 with again of 15 mV/pC. The ground mesh is read out through a fast current amplifier,the Route2Electronics HS-AMP-CF-2nF with 1 ns rise time and gain of 80. Thefast amplifier has a built-in protection that guards against discharges from thishigh-capacitance detector. The CSPs are protected with 300 Ω resistors. Allthree signals are digitized using the ATS860, a 250 MS/s 8-bit flash ADC madeby AlazarTech. The raw unit for charge readout in the the following discussionis mV. The digitization of all three channels is triggered by a signal on theanode and stored with the current CCD image. The trigger threshold used isgas dependent.
The combination of the four measurements, CCD image, charge deposited onthe inner anode and veto anode and the current recorded at the ground mesh, isused for reconstruction. The CCD track reconstruction starts with subtractingthe bias image that is obtained at the beginning of every run. At this time, anyunusually high-valued pixels are removed. The image is then smoothed. Withthe smoothed image, a search is conducted for clusters of adjacent pixels withvalues 3.7 σ larger than the mean value of the image [2]. If enough such pixelsare found, the cluster is determined to be a possible track. An additional ring ofpixels around the main cluster is added to account for diffusion, and the trackis the corresponding pixels from the original un-smoothed image. For each ofthese tracks, the energy, range, angle, maximum value, and several momentsare calculated.We require that the track be contained within the TPC. The veto removessignals from radiation emitted from the drift rings or the vessel wall. However,this also vetoes long tracks produced within the active volume that enter theveto. Additional inefficiency comes from the requirement that the track be fullycontained in the CCD view-field. We show below that these are not a significantconstraints, even in the first generation detector.Tracks imaged by the CCD camera are matched to charge waveforms usingthe reconstructed energy. The rate of events in the detector, even with theneutron and α sources, is relatively low, and most exposures contain no tracks.
4. Calibration
Calibration of the detector was performed using a Am α source. A windowon the source degrades the energy of the α particles leading to a mean energyof 4 . ± .
04 MeV [15]. Example α events in two different gas mixtures areshown in Fig. 3. 6 .1. Spatial Calibration The spatial calibration is obtained by measuring the positions of the wirespacers separating the anode plane from the ground mesh, shown in Fig. 2. Thespacers create a region of low gain. A
Cs source is used to generate electronicrecoils so that a small but diffuse amount of light is seen in each image. Byintegrating many images, the spacers become visible as regions with low lightlevels. The spacers are placed 2.5 cm apart, and the fit gives a calibration of163 µ m per pixel. The total area of the 1024 by 1024 pixel region imaged by theCCD is then 16.7 cm × Figure 3: For α particle tracks from the Am source, images of the CCD and charge read-out system are shown. Left: CCD image of α particle tracks. Middle: Waveforms from theVeto and Anode charge read-out. Right: Waveforms from the Ground-Mesh charge read-out.Top: 100% CF ; Bottom: 12.5% CF + Helium. The total ionization left by the track is measured in three ways: (1) inte-grating the light in the CCD track, (2) measuring the pulse peak height of thecentral anode signal and (3) integrating the current signal from the mesh chan-nel. At the drift voltage used in these data, the α tracks saturate the digitizer ofthe ground mesh and possibly the anode preamp. Using an 8-bit ADC, it is notpossible to have both a fine energy granularity and a large range. The gain isoptimized to have good energy resolution for low-momentum recoils created inthe neutron scattering rather than the α studies. For this reason only methods7 igure 4: Energy gain calibration using the charge read-out (left) and CCD (right). for 100%(top) and 12.5%, (bottom) CF mixtures. (1) and (2) are used for determining track energy. Electron attachment wasmeasured to be negligible at our drift fields [16].The energy distributions from the anode and CCD are shown for two gasmixtures in Fig. 4. We have fit the peak position for the energy calibration. Theresults of the calibration are given in Table 1. The energy is quoted in units ofalpha equivalent energy to denote that the detector response is calibrated withan alpha source. We do not apply corrections due to ionization quenching atlower recoil momenta due to lack of experimental data.The gas gain shown in Table 1 is calculated from the charge energy calibra-tion, the preamplifier gain of 1.5 mV / pC, and the gas work function. Thework function of CF gas was measured to be 33.8 eV [17]. This is used forcalculating the gain of all mixtures.
5. The Neutron Source Run
The response of the detector to neutrons was characterized using a 1.24 mCi
Cf source.
Cf decays via spontaneous fission 3% of the time emittingmultiple neutrons with a mean energy of 2.35 MeV [18]. Given the activity ofthis source, we expect 5 . × neutrons per second in all directions. The sourceis placed in a collimator made of borated plastic, 2.1 m from the detector. Thehigh flux of both neutrons and x-rays causes an increase in the rate of sparksin the TPC from ∼ igure 5: To validate the α source calibration, we use CCD light versus charge read-outsignal for neutron scatters from a Cf source. Left: 100% CF . Right: 12.5% CF + He. % of CF RequirementAll cases Two adjacent pixels with counts > σ above mean [2];Tracks are contained within 24 ≤ x, y ≤ <
250 ADU and <
25% of total signal;Only one track in the image;No two tracks within 12 pixels in 1 run100% Tracks shorter than 80 pixels (1.3 cm)12.5% Tracks shorter than 160 pixels (2.6 cm)
Table 2: Cuts applied to CCD images for the study. exposures of a spark are excluded from analysis to allow sufficient time for thehigh voltage to recover.
The cuts to isolate good events are applied to the CCD image, the charge-readout signal and the charge-light matching. For each run condition describedin Table 1, we identified a set of optimized cuts. Table 2 describes the cuts placedon CCD images. Only cuts on track-length are gas-specific. The other cuts aredesigned to restrict tracks to the fiducial volume and remove backgrounds fromionization within the CCD chip, noise artifacts, and hot pixels.The charge-readout cuts are more gas dependent than the CCD cuts. Theycan be divided into rise time cuts and pulse height cuts, summarized in Tables 3and 4 respectively. For the veto we define two variables: the time is takes thepulse to rise from 10% to 90% of the peak pulse height, t veto , and the ratioof the mesh peak pulse height to the veto peak pulse height, R veto . For themesh and anode, we define pulse rise times, t mesh and t anode . The pulse heightrange over which t mesh and t anode are defined is tuned for each gas mixture.9 of CF t veto t anode and t mesh Pulseheight Range100% <
800 ns t mesh <
30 ns for pulseheight <
100 mV 25% to 75% t mesh <
25 ns for pulseheight >
100 mV 25% to 75%12.5% <
800 ns t anode < Table 3: Rise time requirements. Variables are defined in the text. Events are accepted if theypass the requirements listed. The pulseheight range over which t veto is measured is always10% to 90%. The pulseheight range for the definition of t mesh and t anode vary as listed incolumn 4. % of CF R veto R elec R ion R mesh > . < R elec < . . < R ion < . > . < R mesh < Table 4: Pulse height ratio requirements. The ratios are defined in the text. Events areaccepted if they pass the requirements listed in this table. N/A indicates that this variable isnot used.
For the mesh we also define two ratios for the pulse height: R elec the ratio ofthe electron peak mesh pulse height to the height of the anode pulse and R ion the ratio of the ion peak mesh pulse height to the anode pulse height. R mesh isa simple ratio of the anode pulse height to the mesh pulse height.The CCD and charge readout run autonomously. Matching the signals pro-ceeds by comparing the CCD track energy to the charge energy, where the bestenergy match is identified as the true signal from the track. The requirementsfor matching are shown in Table 5. Optical effects are estimated by measuringthe mean ratio of the energies of the light and charge signals in neutron calibra-tion running as a function of the distance from the image center. This is used tocorrect the energy of the light signal to obtain a linear relationship between thetwo energy measurements. The energy from the anode channel is more accuratethan the CCD energy, so it is used as the final track energy. Neutron sourcedata shown in Fig. 5 confirms a linear relationship between the summed anodeenergy and the CCD energy, demonstrating a good match between a CCD trackand an anode waveform can be found for most of the events.The efficiency of the cuts, including charge-light matching for neutron events% of CF Requirement100% | ∆ E | <
40 keV α if E anode <
150 keV α ; | ∆ E | <
75 keV α if E anode >
150 keV α ;12.5% | ∆ E | <
75 keV α for E anode <
250 keV α ; | ∆ E | <
125 keV α for 250 < E anode <
500 keV α ; | ∆ E | <
150 keV α for E anode >
500 keV α Table 5: Requirements for CCD and charge event matching in this study, where | ∆ E | = | E CCD − E anode | . igure 6: Efficiency of charge analysis in accepting tracks identified by the CCD analysis.The efficiency plateaus at 70-80% are due to exposure difference between the charge and lightreadout and depend on the event readout time. The decrease in efficiency at higher energiesis from the mesh pulse reaching the maximum value read out by the digitizer. Left: 100%CF gas. Right: 12.5% CF + He mixture. The errors on the data points reflect the statisticsof the calibration run. is given in Fig. 6. This data is dominated by nuclear recoils rather than CCDartifacts, so is ideal for the efficiency determination. The efficiency is defined asthe ratio of the number of tracks passing all CCD cuts, charge cuts and charge-light matching compared to the number of CCD tracks found without using thecharge signals. The neutron source generates a large number of nuclear recoils, with O (1)neutron-induced nuclear recoils detected per 1 s exposure. CCD images andcharge waveforms of neutron-induced recoils in two events with different gasmixtures are shown in Fig. 7. A large number of electronic recoils from the high γ flux are observed in the charge signals. However the CCD remains blind tothese events due to their low primary ionization density. The source createsa much higher number of CCD-induced artifacts, likely from electronic recoilsinside the active volume of the CCD chip. Even with these large source activityissues, the cuts are sufficient for this analysis.After applying all cuts, we use the measured track angles and ranges, Fig. 8and Fig. 9, to evaluate the performance of the detector and the analysis. Themeasured two-dimensional track range generally falls just below the expectedmean three-dimensional range calculated with the publicly available programSRIM [19]. For the 12.5% CF mixture, the measured two-dimensional trackrange falls close to the expected range for helium tracks. This follows fromthe fact that helium recoils are kinematically favored across most of the energyrange measured here. At lower energies, where both helium and fluorine recoilsare expected, the track ranges appear to be consistent with either nucleus.11 igure 7: Example figures of neutron recoils from Cf source showing graphical imagesof CCD and charge read-out system. Left: CCD image of He recoil. Middle: Waveformfrom the Veto and Anode charge read-out. Right: Waveform from the Ground-Mesh chargeread-out. Top row is 100% CF and bottom is 12.5% CF . We use the track angles to reconstruct the location of the neutron source.Tracks along the direction of the incoming neutrons should have a decreasinglight profile, and this “head-tail effect” can be used to reconstruct the sense oftrack direction [20]. Two peaks are seen in the angular distribution of Fig. 8.The larger peak at −
90 degrees is along the true mean direction of the nuclearrecoils. The second peak at +90 degrees is primarily from events where thetrack direction was mis-reconstructed. The large width of the distribution isexpected from elastic scattering kinematics. Additionally, some nuclear recoilsare from neutrons that have scattered several times within the laboratory, andno longer appear to originate at the source position.A model of helium and fluorine recoils is fitted to the measured recoil spec-trum in Fig. 10. The fit uses the likelihood approach with the probabilitydensity functions for helium and fluorine based on simulation, and a
Cf spec-trum based on ENDF tables [13]. We find the fraction of helium recoil eventsto be 61 ± Cf source that isbased on ENDF tables [13]. In order to improve the agreement between the dataand the simulation of the recoil and neutron spectra, we need to improve thesimulation to include the efficiency loss due to sparks. This requires additionalmeasurements, and a full simulation of the detector material which modifies the12 igure 8: Energy-Angular distribution for calibration and background runs for two gasmixtures. Left: Calibration run with
Cf source. Right: Background run. Top: 100% CF gas. Bottom: 12.5% CF . The direction of the neutrons from the source is indicated by a redsolid line at −
90 degrees. The population at +90 degrees is due to incorrect reconstruction ofthe track direction. energy spectrum of neutrons entering the sensitive region.
6. Surface Run
With the DCTPC prototype, we have also taken data with no sources tomeasure cosmogenic, environmental and detector backgrounds in a surface labon the MIT campus. The lab is on the ground floor of a two-story building.The lab has approximately 1 m concrete walls on all sides. After removingspark events, we obtain 112713 exposures in pure CF and 103822 exposuresin the 12.5% CF mixture. As in the other runs, each exposure is 1 s, so thiscorresponds to 1.3 and 1.2 days of exposure, respectively.In the pure CF run, we identified 128 possible tracks in the CCD analysis.Most of these tracks appear to be from CCD backgrounds such as noise artifacts13 igure 9: Energy-Range distribution, calculated in two dimensions, for calibration and back-ground runs with two different gas mixtures. Left: Calibration run with Cf source. Right:Background run. Top: 100% CF gas. Bottom: 12.5% CF mixture. Three-dimensionalanalytical prediction of the dE/dx curve for different nuclear elements of the gas mixture isshown in colored lines. a (keV recoil E
500 1000 1500 2000 2500 E v en t s / ( k e V ) (keV) neutron E Figure 10: Left: A fit to a recoil spectrum (markers) using a likelihood model based onhelium and fluorine recoils. The helium contribution (dashed) and total recoil distribution(full line) are shown as histograms. Right: Unfolded neutron spectrum from a calibration runwith
Cf source (markers), and a spectrum based on the
Cf table [13] (histogram). or ionization in the CCD. Of these 128 tracks, only 10 were matched to acharge trigger. Several additional tracks had too much energy to be accuratelymeasured by the charge channels, although the pulses were collected. Thesetracks are mostly low energy (
E <
100 keV α ) and the ranges generally appearto be consistent with the expected values from SRIM for fluorine recoils. Asingle event has a range more consistent with helium or hydrogen tracks.In the 12.5% mixture run, we identified 71 potential tracks in the CCD anal-ysis, with 16 having a matching charge signal. As with the pure CF run, mostof the potential tracks seen in the CCD are likely noise artifacts or tracks leftdirectly in the silicon of the CCD chip. As expected, we see more tracks in thisrun due to the presence of helium. The tracks seen here are also generally athigher energies than were seen in pure CF . This is to be expected if most ofthe tracks come from elastic scattering from neutrons due to the favorable kine-matics of neutron-helium scattering compared to neutron-fluorine scattering.Fewer events at lower energies are seen due to the differing energy thresholds ofthe two gas mixtures. There does not appear to be a favored direction of thesetracks, although much higher statistics would be necessary to confirm this. Theevents with E >
300 keV α here have ranges consistent with helium recoils butnot carbon or fluorine, while most of the other tracks have ranges consistentwith either carbon or fluorine nuclei or near vertical helium nuclei.
7. Future Underground Measurement
The first generation detector is being installed in the far-hall of Double Choozat 300 m.w.e. We can use the 13 neutrons per day measured in the surface runto make an estimate for our event rate in the far-hall. For the surface run,we measure a muon rate of 100 muons per m s . We assume the mean muonenergy is ∼ s with15 (keV) n E E ff i c i en cy FFHeHe
Figure 11: The efficiency for reconstructing helium and fluorine recoils as a function ofneutron energy. The geometric efficiency (full line) requires full track containment in the CCDview-field. The full efficiency (dashed line) assumes a step-function for the recoil reconstructionefficiency with thresholds given in the text. a mean muon energy of 60 GeV[4]. Using the power law scaling from Ref. [11],we estimate ∼ α , n) and muoninteractions. Fluorine and helium recoils are efficiently reconstructed above 150and 100 keV of recoil energy, respectively. Since fluorine recoils are shorter,they have a high probability to fit within CCD view-field. Geometric acceptanceefficiencies are shown as full lines in Fig. 11. Helium recoils receive more energyfrom a collision with neutrons and leave longer tracks – a combination thatis favored by reconstruction algorithms. Helium recoils are more efficient atenergies (cid:46)
20 MeV. Fluorine appears more efficient if the detector is being usedfor neutrons with the energy of tens of MeV’s due to more compact tracks.
We have modified the DMTPC dark matter detector design to produce a2.8 liter, first-generation directional neutron detector. The primary change tothe DMTPC design was to add helium as a target for the neutrons. We havedemonstrated that a gas mixture of 75 Torr of CF and 525 Torr of helium iscapable of extracting the energy and angular distributions of the neutrons froma Cf source.The sensitivity to higher energy neutrons and an estimate of 0.5 events perday at the Double Chooz far site indicates that the first generation detectorcould make an interesting measurement at the far site on its own. The successof this detector motivates the construction of a full-sized DCTPC system. Thiswill consist of two 60 l detectors, located at the near and far halls of DoubleChooz, which can be interchanged for comparison of systematics. The goal is16o measure the neutron flux, with information on both energy and direction, asa function of time, in order to tune the fast neutron simulation.
Acknowledgments
The authors thank Prof. Peter Fisher, of MIT, and the Double Chooz col-laboration for valuable input. JMC, KT and LW are supported by the NationalScience Foundation. DD is supported by the Department of Energy. JL is sup-ported by a Massachusetts Institute of Technology Lyons Fellowship and theInstitute for Soldier Nanotechnology.
References [1] J. A. Formaggio and C. J. Martoff. Backgrounds to sensitive experimentsunderground.
Ann. Rev. Nucl. Part. Sci. , 54:361–412, 2004.[2] S. Ahlen et al. First Dark Matter Search Results from a Surface Run of the10-L DMTPC Directional Dark Matter Detector.
Phys. Lett. , B695:124–129, 2011.[3] Alvaro Roccaro et al. A Background-Free Direction-Sensitive Neutron De-tector.
Nucl. Instrum. Meth. , A608:305–309, 2009.[4] F. Ardellier et al. Double Chooz: A search for the neutrino mixing angle θ . 2006.[5] Dongming Mei and A. Hime. Muon-Induced Background Study for Under-ground Laboratories. Phys. Rev. , D73:053004, 2006.[6] A. Ferrari, P. R. Sala, A. Fasso, and J. Ranft. FLUKA: A multi-particletransport code (Program version 2005). CERN-2005-010.[7] A. Fasso et al. The Physics Models of FLUKA: Status and Recent Devel-opment. 2003.[8] John Allison et al. Geant4 developments and applications.
IEEE Trans.Nucl. Sci. , 53:270, 2006.[9] S. Agostinelli et al. GEANT4: A simulation toolkit.
Nucl. Instrum. Meth. ,A506:250–303, 2003.[10] B. Armbruster et al. Upper limits for neutrino oscillations ¯ ν µ to ¯ ν e frommuon decay at rest. Phys. Rev. , D65:112001, 2002.[11] S. Abe et al. Production of Radioactive Isotopes through Cosmic MuonSpallation in KamLAND.
Phys. Rev. , C81:025807, 2010.[12] M. Aglietta et al. Measurement of the neutron flux produced by cosmic-raymuons with LVD at Gran Sasso. 1999.1713] M.B. Chadwick et al. ENDF/B-VII.0: Next Generation Evaluated NuclearData Library for Nuclear Science and Technology.
Nuclear Data Sheets ,107:2931–3060, 2006.[14] D. Dujmic et al. Charge Amplification Concepts for Direction-SensitiveDark Matter Detectors.
Astropart. Phys. , 30:58–64, 2008.[15] H. Yegoryan. Study of Alpha Background in a Dark Matter Detector.
Senior Thesis, Massachusetts Institute of Technology , 2010.[16] T. Caldwell et al. Transport properties of electrons in CF(4). 2009.[17] I.C. Wolfe. Measurement of Work Function in CF Gas.
Senior Thesis,Massachusetts Institute of Technology , 2010.[18] K. Nakamura et al. The Rev. of Part. Physics.
J. Phys. , G37:075021, 2010.[19] J.F. Ziegler, J.P. Biersack, and M.D. Ziegler. SRIM: The Stopping andRange of Ions in Matter. , 2011.[20] D. Dujmic et al. Observation of the ’head-tail’ effect in nuclear recoils oflow-energy neutrons.