Interactions of Neutrons with Silicon CCD Chips MASSA CHUSETTS INSTITUTE F TECHNOLOGY by O I JN 0 8 2011 Timur Cemal Sahin Submitted to the Department of Physics in partial ffnlfillment of the requirements for the degree of LIBRARIES ARCHIVES Bachelor of Science in Physics at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY February 2011 0 Timur Cemal Sahin, MMXI. All rights reserved. The author hereby grants to MIT permission to reproduce and distribute publicly paper and electronic copies of this thesis document in whole or in part. A u th or ............................... Department of Physics 7 May 18, 2010 Certified by......... Denis Dujmic Research Scientist Certified by.................. Gabriella bcioia Associate Professor of Physics supervisor Accepted by ...... Professor David E. Pritchard Senior Thesis Coordinator, Department of Physics 2 Interactions of Neutrons with Silicon CCD Chips by Timur Cemal Sahin Submitted to the Department of Physics on May 18, 2010, in partial fulfillment of the requirements for the degree of Bachelor of Science in Physics Abstract Dark matter makes up approximately 22% of the energy density of the universe and as much as 83% of its matter composition. Despite its ubiquitous nature, it remains incredibly difficult to detect due to the weakness of its interaction with the regular matter. The Dark Matter Time Projection Chamber (DMTPC) is an experiment that searches for traces of ionization created by Weakly Interacting Massive Particles (WIMP's). The detector uses a charge coupled device (CCD) camera to image the ionization signal created in the detector gas. The CCD chip itself is also sensitive to interactions with the WIMP's and the background radiation. In this thesis I explore the contributions these interactions may have on the DMTPC experiment. First, I develop an algorithm that filters out the electronic noise in the CCD chip such that the remaining images contain true ionization events in the chip. Second, I study insidious effects of neutron interaction with the CCD chip. I develop a GEANT4 based Monte Carlo simulation and set up an experiment that uses a neutron source with a known energy and measure the energy deposition in the chip. The energy spectrum agrees with the prediction based on the elastic scattering kinematics and the silicon ionization rate thus providing an energy calibration. Finally, I measure the level of background in the CCD chip coming from the inside of the camera, I set up an experiment in which outside neutrons are shielded with layers of plastic material and x-rays are suppressed with lead bricks surrounding the camera. Cosmic muons create particle showers that can also interact with the CCD chip so I build an active shield using a pair of scintillating paddles. I find that background interactions with silicon produce CCD signals that are small enough to be eliminated by existing DMTPC cuts on interaction range. Thesis Supervisor: Denis Dujmic Title: Research Scientist Thesis Supervisor: Gabriella Sciolla Title: Associate Professor of Physics 4 Acknowledgments I would like to thank my thesis advisor Dr. Denis Dujmic for his substantial assistance and mentorship in this project from its inception through its completion, and for his guidance in personally and professionally trying times. I would also like to thank my advisor Gabriella Sciolla for her support and contributions to this work. I also thank Shawn Henderson, Jeremy Lopez, and Cosmin Deaconu for their assistance with matters of ROOT, Linux/Unix, and C++ programming. Finally, I would like to extend my gratitude to the John S. Reed Foundation for providing the grant for my original undergraduate research, from which this thesis has evolved. 6 Contents 13 1 Introduction 1.1 The Dark Matter Problem . . . . . . . . . . . . . . . . . . . . . . . . 1.2 DMTPC Experiment..................... 1.3 CCD Principles...................... .. . . . . . ... 14 . . ... 15 19 2 Interactions in CCD 2.1 2.2 3 13 . . . ... 19 2.1.1 Scattering Rate and Cross Section.......... . . . ... 19 2.1.2 Elastic Scattering Kinematics.............. . . ... 20 2.1.3 Stopping Power.................. . . ... 21 . . . . ... 24 Neutron Interactions in Silicon................. ... Cosmic Ray Interactions in Silicon............. 2.2.1 Direct Ionization by Cosmic Muons . . . . . . . . . . . . . . . 2.2.2 Ionization by Muon-Induced Showers.......... 2.2.3 Cosmic Ray Coincidence Setup........... . ... 25 . . . . ... 25 27 Data Reduction and Analysis 3.1 Processing of Scintillator Signals............... 3.2 Image Processing............ . . . . . . 24 . . ... 27 . . . . . . . . ... 27 . 3.2.1 Hot Pixels and Background Cleaning . . . . . . . . . . . . . . 27 3.2.2 Recognition of Significant Pixels... . . . . . . . . . . . . . 30 . . . . . . . . . 32 . . . . . 35 . . . . . . . . . 36 3.3 Measurement of Neutron Interaction Spectrum. . 3.4 Measurement of Background Interaction Spectrum... . . 3.4.1 Unshielded Spectrum......... . . . . . 3.4.2 Passive Neutron and Gamma Shielding . . . . . . . . . . . . . 36 3.4.3 Active Cosmic Ray Vetoing 36 . . . . . . . . . . . . . . . . . . . 4 Conclusions 37 A The McDark Simulation 39 A.1 Detector Geometry and Hit Recording....... . . . . . . . . .. A.2 Physics Process Selection . . . . . . . . . . . . . . . . . . . . . . . . . A.2.1 Electromagnetic.................. A .2.2 O ptical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.2.3 Hadronic........................ A.2.4 Decay. . . . . .. . ......................... A.3 Event Generation.................... . . . . . . .. 39 40 40 41 . . . . 41 . . .. 42 . . . . . 42 List of Figures 1-1 The DMTPC 1OL detector in successive stages of assembly. . . . . . . 1-2 Monte Carlo simulation of the 10L detector that I developed as a part 14 of this thesis. Plots show sensitive volumes of the TPC's and CCD ch ip s. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 1-3 Diagram of two adjacent CCD cells. Every third electrode is connected. 16 1-4 Diagram of CCD readout.................. 1-5 Electron motion through CCD cells. . . . . . . . . . . . . . . . . . . . 18 2-1 Neutron gun opposite CCD camera . . . . . . . . . . . . . . . . . . . 20 2-2 o- for elastic scattering of neutrons on silicon v.s. energy of neutrons. 21 2-3 Quenching Factor Q(E). 22 2-4 Stopping power and ionization rate for silicon ion in silicon. ..... 23 2-5 Scintillator coincidence circuit........... . . . . ... 26 3-1 Examples of scintillator PMT response............. . . ... 28 3-2 Distribution of significant clusters across CCD bins..... . . ... 29 3-3 Example images of hot pixels in background runs. . . . . . . . . . . . 3-4 Frequency with which significant clusters were located in a given pixel for one run. ... . . . . ... .................. . .... . . . . . .................. .... ... . ........ 17 30 31 3-5 Neutron Interactions........... . . . . . . . . . . . . 33 3-6 Example images from neutron interactions. . . . . . . . . . . . . . . . 34 3-7 Unshielded and shielded reconstructed background spectra . . . . . . 35 3-8 Ratio of events from shielding and muon cuts. . . . . . . . . . . . . . 36 . . . . 10 List of Tables 2.1 Parameters used in rate calculation. 3.1 Cluster event rate with shielding. 12 Chapter 1 Introduction 1.1 The Dark Matter Problem Current astronomical observations seem to indicate that visible light and matter constitute a relatively small fraction of the total energy density in the universe. The discrepency was first discovered by analyzing the velocities of galaxies in the Coma cluster and concluding that the calculated mass exceeded the observed mass by a large factor [1]. Further studies have shown that the Coma cluster is by no means unique and that similar mechanics govern the majority of observed spiral galaxies, including the Milky Way. This unseen "dark matter" seems to comprise approximately 23% of the total energy density or approximately 80% of the mass of the universe[2]. The dark matter has gravitational interaction and small coupling to with regular matter, but little else is known about it. Recent studies have focused on an attractive dark matter candidate known as the Weakly Interacting Massive Particle (WIMP). Current direct WIMP detection experiments attempt to identify WIMPs interacting with nuclei to produce a low energy recoil. Most experiments look for the energy deposited by these recoils. Some detectors, such as the Dark Matter Time Projection Chamber (DMTPC), also attempt to identify the the direction of the incoming WIMP in the collision[3]. 1.2 DMTPC Experiment The Dark Matter Time Projection Chamber (DMTPC) detector consists of a stainless steel vessel filled with CF 4 gas kept at approximately 75 Torr. The vessel is divided into two optically isolated volumes, each housing a TPC (with fiducial volumes 14.6 x 14.6 x 19.7 cm3 and 15.9 x 15.9 x 19.7 cm3 ). A WIMP interacts with a, gas nucleus that recoils and ionizes the residual gas molecules. In order to detect the ionization electrons, we apply an electric field throughout the gas that drifts the electrons toward an amplification plane where they create electron and photon avalanches. The cathode and ground plane of the TPC are meshes with diameter 27 cm and 256im pitch [4]. The electric field is approximately uniform (0.25 kV/cm) throughout the drift volume and sharply approaches higher values (14.4 kV/cm) near the amplification region. The produced photons are captured optically by a Charge Coupled Device (CCD) camera located at the top and bottom of the vessel. This setup is presented in Fig. 1-1. Part of the Geant4 rendering is visible in Fig. 1-2. The CCD camera images photons created in the TPC, but the CCD chip is also sensitive to radiation that interacts with the chip directly. The goal of this thesis is to investigate how neutron interactions with silicon CCDs can contribute to the background in the DMTPC experiment. Figure 1-1: The DMTPC 10L detector in successive stages of assembly. Figure 1-2: Monte Carlo simulation of the 1OL detector that I developed as a part of this thesis. Plots show sensitive volumes of the TPC's and CCD chips. 1.3 CCD Principles The image area of the CCD is a silicon wafer that operates by using the photoelectric effect to convert incoming photons into an organized electron charge. A photon striking a pixel in a CCD will excite electrons from the valence band into the conduction band of the silicon material. An electric field pattern applied across the cells keeps the electrons localized. The charge collected in a potential well is linearly proportional to the amount of incident light and non-linearly dependent on wavelength[5]. Through this process, a pixel of a CCD can collect charges directly proportional to the light falling on that pixel. However, anything that can free electrons from the valence band can contribute to this collection of charge. Indeed, we can measure the processes described above by exploiting this property. In addition, thermal excitations can also be a noise source that becomes more dramatic at higher temperatures. The signal from thermally generated electrons is known as "dark current." The CCD collects charge during the exposure period, during which the shutter may be left open or closed. For all these measurements, we are interested in only the background sources so the shutter is left closed. After the exposure period, the charges are shifted from adjacent cell to adjacent cell, with the furthest column depositing its charge into a serial register which digitizes the signal from each pixel, one by one. This shifting is accomplished by modulating the voltages on the electrodes connected to each individual pixel After a row is read out, the charge again shifts one pixel towards the register, until the entire image has been read out. This process is illustrated in Fig. 1-4 and Fig. 1-5. It is possible for radiation interactions to occur during this read-out phase as well as the exposure phase, so it is best to keep this read-out time to a minimum compared to the exposure time. Au Electrodes - 5pm Si 24 pm Figure 1-3: Diagram of two adjacent CCD cells. Every third electrode is connected. For the DMTPC experiment, we make use of an Apogee CCD camera (S/N A80333 and A80334) in order to image a region of CF 4 gas. The particular CCD cameras we use consist of a silicon chip divided into 1024 by 1024 sections. We rebin the pixels into a 256 by 256 matrix, such that each virtual pixel represents 16 pixels. We are particularly concerned with nuclear recoils on the order of several keV which produce a track length of about or less than 2 mm. These recoils can easily produce tracks that occupy only one or two virtual pixels (particularly if their momentum is mostly perpendicular to the CCD chip). In principle, there are a number of sources that can deposit a great deal of energy into one virtual pixel in the absence of a nuclear recoil within the CF 4 . The silicon CCD chip can interact with with cosmic rays, neutrons, and x-rays and produce an image containing a track when no event occurs within the TPC. Neutrons and x-rays can come from external sources or from impurities within the camera itself. Silicon CCD Al A2 A3 B1 B2 B3 B4 A4 C1 C2 C3 A12 Al A2 A3 12 B1 B2 B3 B4 C6 C7 C8 C9 C10 C11 C12 C1 C2 A5 A6 A7 A8 B6 B7 B5 C4 C5 Silicon CCD SerialRegister A9 A101 1 B8 B9 BlOB11 A4 Serial Register A5 A6 A7 A8 A9 A10 Al1 B8 B9 B1 B11 B5 B6 B7 C3 C4 C5 C6 C7 C8 C9 C10 C11 D7 D8 D9 Dl D11 Dl D2 D3 D4 D5 D6 D7 D8 D9 DIOD11 D12 D1 D2 D3 D4 D5 D6 El E2 E3 E4 E5 E6 E7 E8 E9 E10 Ell E12 El E2 E3 E4 E5 E6 E7 E8 E9 E10 Ell Fl F2 F3 F4 F6 F7 F8 F9 Fl F2 F3 F4 F5 F6 F7 F8 Gl G2 G3 G4 G5 G6 G7 Gl G2 G3 G4 G5 G6 HI H2 HI H2 H3 H4 H5 F5 H3 H4 H5 F10 Fl1F12 G8 G9 G10G11G12 H6 H7 H8 H9 H10H1Il H12 18 19 110 ill 112 -> 12 113 L 11 12 13 14 F9 F1O F1l G7 G8 G9 GlG11 H6 H7 H8 H9 Hi H11 IS 16 17 18 19 li0 111 Ouput amplifier Ouput amplifier (a) An image is recorded onto the CCD silicon. (b) Extreme pixels are transferred to serial register. Silicon CCD Al A2 A3 A4 Silicon CCD SerialRegister Serial Register A7 A8 A9 A10 -- A2 A3 B6 B7 B8 89 B10 811 B1 B2 B3 B4 B5 B6 B7 B8 B9 B10 - 811 - 011 A7 A8 A9 Al0 All A4 A5 A6 A1 A5 A6 All B1 B2 B3 B4 C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C1 C2 C3 C4 C5 C6 C7 C8 C9 D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 D11 D1 D2 D3 D4 D5 D6 D7 D8 D9 D10 -- D11 El E2 E3 E4 E5 E6 E7 E3 E4 E5 E6 E7 E8 E9 E10 -- Ell F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 -> Fll G1 G2 G3 G4 G5 G6 G7 G8 Hi H2 H3 H4 H5 11 12 13 B5 El E2 E8 E9 E10 E11 F7 F8 F9 F10 Fli H6 H7 G9 Gl G11 H8 H9 H1 Hi1 F2 F3 F4 F5 G1 G2 G3 G4 G5 G6 G7 G8 G9 H1 H2 H3 H4 H5 H6 14 15 6 11 1101Ill Ouput amplifier Fl 12 13 F6 10 10 -- H7 H8 H9 Hl 17 - 18 19 I10j- Gll H1l 111 Ouput amplifier 1 (c) Signal in register is digitized by output amplifier. (d) The process repeats to completion. Figure 1-4: Diagram of CCD readout 17 J ~ -f ............................. ---------.......... .......... ........... ................ ................. (d) ................... ............ ................ ................. ............. ................ ... ............ ................. ........... .......... ----------- %% (g) L %% 411 (h) Figure 1-5: Electron motion through CCD cells. Red coloring indicates a positive potential on the gold electrodes. Chapter 2 Interactions in CCD 2.1 Neutron Interactions in Silicon Unlike charged particles, neutrons can travel long distances without interacting in matter. Neutrons will tend to interact with matter through either scattering or absorption. Neutron scattering can occur elastically (where energy is conserved and transferred into the kinetic energy of a recoiling nucleus) or inelastically (where the recoil nucleus enters an excited state). Absorption can result in the emission of a wide variety of radiation or a fission reaction. 2.1.1 Scattering Rate and Cross Section We produce neutrons through the reaction 2 H + 3 H - He + n, producing monoener- getic neutrons with energy 14.1 MeV[6]. The Apogee CCD camera is placed on end opposite the neutron beam as depicted in Fig. 2-1. We can express the elastic scattering rate R in number of reactions per unit volume target as: R where <D = <}, X JEL X PSi X (2.1) is the neutron flux from the source (in 1/area/time)), oEL is the elastic scattering cross section (area), psi is the density of atoms in Silicon (1/volume), and Q is the solid angle of the target at a distance from the source. Evaluating these Figure 2-1: Neutron gun opposite CCD camera quantities for our silicon CCD chip at a distance of 1.4 meters, we find a neutronsilicon interaction rate of R ~ 12 cm-3s, or on the order of 0.035 Hz for a block of silicon with the dimensions of our CCD chip. 0EL psi Q Vsi 3.6 x 106 cm-2s-1) 0.68 b 5 x 1022 cm-3 105pm x 24mm x 24mm Table 2.1: Parameters used in rate calculation. 2.1.2 Elastic Scattering Kinematics For a neutron (mass mn and velocity v,) striking a nucleus at rest with mass M, we can write the following nonrelativistic conservation of energy and momentum relationships: 1 2 2 1 2 '2+ 1MV2 2 mnevn =mnv + MV (2.2) (2.3) E-5 E-4 E-3 E-2 E-1 E+O E+1 E+2 E+3 E+4 E+5 Incident Neutron Energy (eV) Figure 2-2: o for elastic scattering of neutrons on silicon v.s. energy of neutrons. Where ' is the velocity of the neutron after collision and V is the velocity of the recoil nucleus. We can solve these simple equations for v: , A - m,, (2.4) M + mi, and then use the conservation of energy to find the maximum energy transfer from the neutron to the recoil nucleus: 1 AEnax = -m 2 1 2 - /2 m,v,= 2 4Mm,, £ (M + mn) 2 (2.5) where E =mo2/2 is the neutron's initial kinetic energy. For a 14.1 MeV neutron hitting a silicon-28 nucleus, AEmax ~ 1.10 MeV. 2.1.3 Stopping Power We also need to understand the rate at which charged particles lose energy in silicon. The stopping power (-dE/dx) is the instanteneous energy loss per unit length of a particle travelling through a medium. In general, the quantity dE/dx depends on the energy already lost, so we are interested in the graph dE/dx vs. E. Though in principle one can utilize the Bethe equation to calculate the stopping power, it suffices to use the databases provided by software like SRIM and Geant4. The energy lost by charged particles travelling through the silicon in this way is deposited locally resulting in ionization. We see in Fig. 2-4 for recoil energies above 1 MeV, the elastic component of the stopping power is the dominating factor. We are also interested in the quenching factor, or the efficiency of conversion of nuclear recoil energy into light relative to electrons. The form of this quenching factor was extrapolated from Dougherty's data using an empirically determined functional form Q(E) = (1 - exp (-aE))[8]. This form allows for the slow expontential rise and has an upper bound of unity. We find a = 0.164 and b = 0.302. A plot of the quenching factor is available in Fig. 2-3. 0.6 0.4 0.2k 0 Energy (keV) Figure 2-3: Quenching Factor Q(E) E E x 106 10 102 103 104 Energy (keV) Figure 2-4: Stopping power and ionization rate for silicon ion in silicon. 2.2 Cosmic Ray Interactions in Silicon In order to identify any correlation between incident cosmic rays and the appearance of one-pixel artifacts, we isolate the CCD camera from the TPC and construct a simple coincidence setup. We place a small plastic Hamamatsu scintillator above our TPC and a large liquid scintillator undernearth. The Hamamatsu scintillator has a width and depth just greater than the side length of the CCD chip, and a length that far exceeds that of the chip and the camera itself. The face of the liquid scintillator is much greater than the area of the chip (Fig. 2-5). We synchronize digitizers and the camera such that the signals from the scintillators are digitized and recorded during any particular CCD exposure. In order to study the effects of external neutrons and x-rays, we also construct a shield out of polyethylene boards and lead bricks. A cavity holds the CCD camera and optionally the scintillators, with approximately one inch of polyethylene and two inches of lead on any side of the setup. 2.2.1 Direct Ionization by Cosmic Muons Because muons and ions are heavy charged particles, they can interact electromagnetically with the silicon target atoms. For muons with high velocities (high enough to reach the surface of the earth), the majority of their energy loss will occur through ionization and excitation of atoms. The equation governing the maximum energy transfer kinematically identical to Eq. 2.5. We can evaluate this equation for a muon striking an electron: A Emax 4mem 22 E ~ 0.019E (me + mP) (2.6) Which tells us that the muon loses only a small fraction of its energy per electronmuon collision. If we wish to consider the relativistic effects, we can expand Eq. 2.6 and recover AEmax - 2V 2- 2 mc2 2 (2.7) where in the last step we used the approximation y}m/I << 1. For a typical 1 GeV muon travelling at 99% the speed of light, AEmax/E a 0.05. 1/ 1 -/32 and /3 where -y = V/c. This low fractional energy deposition per collision indicates that these particles will, in the absence of any external electric fields, travel in fairly straight lines through the medium. 2.2.2 Ionization by Muon-Induced Showers here is also the possibility of ionization from pt showers rather than direct ionization. The expected rate of muons at sea level is approximately 1 cm-2 min-. For our CCD chip this equals approximately 0.1 per second. Assuming the branching fraction for muons to neutrons is approximately the same as the fraction for muons to charged particles, we expect something on the order of 10-' g-1 cm rate of neutron production from muons R,-,,, 2 neutrons per muon. The in a material is then: R,~,~R~x10c 2 xP x Ad R(_.n = R xP (g/cm2) (2.8) where RP is the rate of muons entering the medium, p is the medium's density and Ad is the penetration depth. For a silicon chip (p = 2.3 g/cm 3 , Ad = 5 x 10-4 cm11), R-~0/s. For a lead brick (p = 11.9 g/cm3 , Ad = 5 cm), R, 2.2.3 5 x 10/s. Cosmic Ray Coincidence Setup A long, thin Hamamatsu scintillator rests atop the Apogee camera located above the CCD chip. Underneath we have placed a large oil scintillator. PMTs are connected to amplifiers and digitizers which read out and record the amplified output signal. At sea level, we expect a muon to penetrate the top scintillator at a rate of approximately one per second. The actual rate taking into account the possibility of oblique muons is closer to 1.3 per second, however these oblique muons are less likely to penetrate the Figure 2-5: Scintillator coincidence circuit. ccd chip after striking the scintillator. A muon passing through the top scintillator will trigger readout on the bottom scintillator. Due to the large size of the oil scintillator, it is almost guaranteed that muons penetrating the top scintillator will go on to penetrate the bottom. The area of the top scintillator far exceeds the area of the CCD chip. Only about 7% of the cosmic rays responsible for coincidence triggers should penetrate the CCD chip. Chapter 3 Data Reduction and Analysis 3.1 Processing of Scintillator Signals We scan the waveforms on the scintillators for photopeaks and determine the time difference between peak locations on the top and bottom scintillators. We collect all waveforms that appear in the scintillator during the exposure. The top scintillator acts as a trigger which begins the readout and digitization of signals from both top and bottom scintillators. Examples coinciding and noncoincident peaks are available in Fig. 3-1. 3.2 3.2.1 Image Processing Hot Pixels and Background Cleaning Impurities in the silicon chip can cause the presence of "hot pixels," or pixels that consistently record a high number of counts regardless of photon absorption or other external energy deposition. We employ a number of methods in order to account for these hot pixels and remove them from our data set. As a relatively simple first cut, we can attempt to identify hot pixels by studying images taken with the camera shutter closed. In these "background runs" we expect the mean number of counts collected in each CCD pixel to be relatively low. If this Time (s) Time (s) 0 0 -0. - -Wm -0.2 - -0.2-0.4 -- .04 0 0.05 0.1 S10 02.3-f - 0 0.05 Time(s) (a) Coincident case 0.1 10 Time(s) (b) Anticoincident case Figure 3-1: Examples of scintillator PMT response. Top and bottom graphs represent top and bottom PMTs, respectively. is the case, we can look for hot pixels by taking all the exposures in a background run and summing them pixel for pixel. As the appearance of true one-pixel events should be distributed evenly throughout the area of the CCD chip, the vast majority of pixels in the summed image tend to have approximately identical number of ADU counts. Hot pixels, however, can accumulate orders of magnitude more counts in this process. We find the mean of number of counts in the histogram, as well as the RMS deviation from the mean, and identify pixels whose values are more than 5 x RMS greater than the mean. These pixels are discounted from analysis. Although this method is effective at removing the majority of hot pixels, it does not completely eliminate hot pixels. The remaining hot pixels were discerned using frequency analysis on the significant clusters. They get recognized as significant onepixel clusters and have their location recorded. If we assume that the appearance of such events is again statistically uniform across the CCD, we expect to see a Poisson distribution with most pixels having zero events, a few having one, even fewer having two, and so on. The expected value for the number of counts is calculated by dividing the number of worms surviving the first cut by the total number of pixels. We then cut out all events located in clusters which recorded more than 5c- events from this mean. Though the vast majority of pixels follow this distribution, hot pixels can have as many as several hundred such recorded events. These pixels are discounted from the analysis. For example, Fig. 3-2 shows the number of worms found in each pixel across 5000 1-second exposures. Each pixel is uniquely numbered between 0 and 2582 (256 pixels + 2 underflow/overflow bins that are an artifact of the format used to store the histogram). Un E I I 20000 40000 I 0 z L_ 10 3 102 10 60000 Pixel No. Figure 3-2: Distribution of significant clusters across CCD bins Identifying hot pixels is necessarily a multi-pass process, in part because the threshold for identifying these pixels is in part influenced by their presence. Discarding outliers while setting the threshold will keep the worst offenders from interfering, but hot pixels with far fewer counts are still present after the first pass. In addition to the threshold-setting component, a hot pixel is not always in an "active" state. A hot pixel can "turn on" and "turn off" during the course of a run. Fig. 3-4 illustrates the frequency with which a given pixel of the CCD camera was identified as being -r250 .200 3150 -100 - 50 0 290 300 8 0 830 310 320 330 -_50 . . . 50 , , 8 ,60 840 80 80 90 100 110 120 310 320 330 340 Figure 3-3: Example images of hot pixels in background runs. part of a significant cluster throughout one given run of 5000 one-second exposure. Hot pixels are easily identified as the long horizontal streaks with high counts. As stated, they have periods of high and low activity even over the course of a few hours. Because there is no guarantee that a hot pixel is in its active state during any particular run (or even part of a run), we must perform a frequency analysis to ensure that we account for hot pixels beneath the threshold. 3.2.2 Recognition of Significant Pixels For each exposure in each run, we must determine which pixels or clusters of pixels are significant and should be recorded as events or worms. We first remove all hot pixels using the method above then recalculate the mean and RMS deviation of the image histogram (discounting outliers). We use as our threshold the mean plus five times the RMS. -O 60000 ---. --.-------------- .. --. - C~ 40000 - .---- -1000D030013131:11.1.1 an00. . 20000 -3 00000. O0n a o o. n. .0. - - 0 - 0 0..... 0 0 - , - 0 o .0 ' .O' 1000 a 0000 o .00~~ . .00.000.00000000000000.0 .00000 .000. . .. 0. 00 I"' o .'"'O" ,""'","" 2000 00000. 0000000000 . . 00.0000.00. n~. . ..... . . . "1 ""1" 3000 4000 5000 Exposure number Figure 3-4: Frequency with which significant clusters were located in a given pixel for one run. Starting from the pixel with the highest number of ADU counts, we check to see if this pixel exceeds our threshold. If it does, we register it as part of a cluster and check its neighbors (the adjacent eight pixels in each cardinal and ordinal direction) to see if they pass the threshold. We perform this operation recursively for every neighbor passing the thresholds, adding each such pixel to the cluster. When we have exhausted all the neighbors, we record the total number of pixels and the sum of the counts in each pixel. For clusters made up of more than two pixels, we attempt to calculate the principal axes of inertia of the cluster. The cluster is then eliminated from the image and then identification starts again from the next pixel with the next highest counts. We repeat this process until we have exhausted all clusters in one exposure, and again for each exposure in each run. 3.3 Measurement of Neutron Interaction Spectrum Neutron data was collected using the method described in Chapter 2. The exposures were run through the hot pixel process and cluster recognition algorithm. Using the sum of the counts from the clusters present in those images, we can rebuild the energy spectrum. We use empirically measured ADU to energy calibration and quenching factor to recover the following spectra. For a given number of ADU, we work backwords knowing that E1 NADU X Q(E) x (3.1) W x G Where NADU is the number of ADU recorded in the cluster, Q is the quenching factor which, in general, depends on the energy, W is the work function of the electron in silicon and G is the empirically measured gain for our particular camera. W 3.6eV/e~ and we measure our gain for this CCD camera to be G = 1.1e~/ADU. We fired deuterons with an energy of 14.1 MeV at the CCD camera for 5000 consecutive one-second exposures. The reconstructed spectrum agrees with GEANT4 monte carlo. The Monte Carlo simulation measures the energy deposition of 14.1 MeV neutrons into a five micron thick sheet of silicon. Here we use the Q(E) found in Fig. 23 in our conversion from ADU to energy. We scale the Geant4 energy profile such that it has the mean rate calculated by Eq. 2.1. The measured energy spectra agree quantitatively well with monte carlo data above 100 keV with the exception of an aberrant peak in the measured event rate that occurs at approximately 230 keV. This peak is present only in the measured data when the neutron source is activated. It is unclear as to whether this is a true neutron-silicon interaction, or whether neutrons are activating or exciting some short-lived isotope which also interacts with the CCD. The interaction produces ADU corresponding very sharply to a 230 keV neutron event. GEANT4 simulations of neutrons interacting with aluminum and gold (present in the CCD camera) do not reproduce this peak. There is a qualitative agreement in the lower-energy spectra, though measured event rates are about an order of magnitude higher than what one calculates from GEANT4. - ... - - ......... I - 10-1 DT (data) DT (Geant4) 10-2 - I ~ 10-3 9 10~44 200 400 600 800 1000 Er (keV) (a) Reconstructed Neutron Spectrum 3000 (D LL 2000 1000 H 2 0, 0. 4 6 8 10 No. pixels in cluster (b) Reconstructed Neutron Energy vs. Npi, Figure 3-5: Neutron Interactions 200 - 530 540 550 560 1190 20 200 210 20 220 20 600 610 620 630 140- 130- 120 110- 100 490 500 510 Figure 3-6: Example images from neutron interactions. 520 6z 150 3.4 Measurement of Background Interaction Spectrum In order to measure the level of background radiation in the CCD, I take data with all sources removed from the vicinity of the camera. The remaining events in the camera come from eitheT surrounding radiation in nature, or from the camera material itself. In order to distringuish the two sources, I apply shielding against the environmental radiation. The results are summarized in Table 3.1 and Figure 3-7. Unshielded Shielded Shielded with rejection Rate (s- 1 ) 30.78 x 102 6.88 x 10-2 0.19 x 10-2 Table 3.1: Cluster event rate with shielding. .1-11 1 - 0 - - *- - Unshielded - Shielded ShieldedwithCoincidence Rejection -- - ~ t 4i 10-3 I I 10~4 10-6 10~5 I I . I ,, 102 103 Er (keV) Figure 3-7: Unshielded and shielded reconstructed background spectra 3.4.1 Unshielded Spectrum The unshielded spectrum is shown in Figure 3-7. The rate in the graph is normalized taking into account the total exposure time. The total rate is given in Table 3.1. 3.4.2 Passive Neutron and Gamma Shielding I surround the camera with 50 cm of Ricorad (boron-rich plastic) as a passive shielding against neutrons, and 20 cm of lead bricks as a shielding against the gammas. The rate normalized to the exposure time is shown in Figure 3-7, and the total rate is give in Table 3.1. The ratio between shielded and unshielded event rates is shown in Figure 3-8. 3.4.3 Active Cosmic Ray Vetoing A cosmic-ray veto is set up with two scintillator paddles above and below the camera. Rejecting those events with activity in both scintillators within approximately a microsecond, we recover the blue spectrum in Fig. 3-7. with and without coincidence rejection are available in Fig. 3-8. The ratio between shielded and no-coincidence event rates is shown in Figure 3-8. The measured ratio is consistent with the geometric considerations from Section 2.2.3 600 800 101 Energy (keV) Figure 3-8: Ratio of events from shielding and muon cuts. Chapter 4 Conclusions I have built a Monte Carlo model in GEANT4 that accounts for particle interactions in both the TPC and the CCD chip of the DMTPC experiment. I have tested experimentally that neutron interactions in the silicon chip agree with our Monte Carlo simulation. The majority of background interactions with silicon result in lowenergy (< 200 keV) clusters with a range of one to several pixels (Fig. 3-5). This overlaps partially with the spectrum we expect from dark matter candidate particle interactions in CF 4 . However, running this neutron silicon background data through the DMTPC cuts ([4]) reveals that no background events survive in the 80 - 200 keV range. This is due principally to the DMTPC's range requirement and the relatively small clusters produced by silicon CCD chip interactions. While other studies have used radioactive sources to study interactions within the TPC gas, this thesis is the first study on interactions directly with the CCD chip. 38 Appendix A The McDark Simulation A.1 Detector Geometry and Hit Recording The geometry of the detector is described in the McDarkDetectorConstruction class. The detector geometry includes most of the TPC vessel as well as the enclosed CF 4 gas and the silicon chip of the imaging camera. The silicon chip is set up as a "sensitive detector," as described in McDarkTpcSD. The inner part of the drift cage is also set up as a sensitive detector. Particles travelling through these "sensitive detector" region will automatically have have their track information and energy deposition information recorded when they interact. These interactions are saved as hits (McDarkHit). Electrons and photons have their quenching factor set to 1.0, whereas nuclei have their quenching factor calculated from the relationship Q (0.3 xSn Se) Sn +Se (A.1) where Sn is the quenching factor calculated from the nuclear stopping model from ICRU report 49 and Se is the quenching factor for an electronic loss model calculated by the SRIM2000 software. All hits in these sensitive regions are converted into a representation of a digital signal (McDarkTpcDigi). The McDarkTpcDigitizer govern the details of this conversion taking into account the area imaged by the CCD camera. The McDarkTpcDigitizer accesses the geometry and identifies the energy deposited towards electrons and photons and the location of the ionization. The McDarkTpcDigitzer uses the McDarkTpcGas class to perform calculations of the electron transport. This information is used to produce an equivalent charge readout, PMT readout, and CCD signal response for the ionization event. The virtual digitizer will loop through all PMTs and CCDs and record this information for each one. A.2 Physics Process Selection The various physics processes implemented in the McDark simulation can be found in the McDarkPhysicsList class. Our own process models are in the McDarkIonMultipleScattering and McDarkUrbanMscModel classes, as described below. A.2.1 Electromagnetic We have chosen to use the various G4LowEnergy packages for electromagnetic processes, being that low energy details could affect our recovered energy spectra. For the G4LowEnergyPhotoElectric class, this means that the simulation takes into account the relative cross-sections of all sub-shells when deciding which should release an electron in a photoelectric interaction. The G4LowEnergyIonisation calculates the continuous energy loss due to electron ionisation and simulates the production of electrons from secondary ionizations. We are especially interested in the production and energy loss of these secondaries as they contribute to the various signals in our experiment. This energy loss is calculated with the G4LowEnergyBremsstrahlung class, which simulates the continuous energy loss from low energy gamma emission. We set a lower production threshold of 250 eV, which represents the lower end of the spectrum for which the reference data is available and reliable. Gamma interactions also use the G4LowEnergyRayleigh and G4LowEnergyCompton processes which take into account Hubbel's form factor for scattering, and G4LowEnergyGammaConversion which uses the Bethe-Heitler differential cross-section with Coulomb correction for sampling the probability of a photon to produce a given pair of particles. Elec- trons utilize the standard G4eMultipleScattering process in addition to the low energy ionisation and low energy bremsstrahlung. Muons and antimuons use the standard G4MuMultipleScattering, G4MuIonisation, G4MuBremmstrahlung, and G4MuPairProduction packages. In addition, the G4MuonMinusCaptureAtRest process accounts for the possibility of negative muon capture. Light ions such as protons, alphas, and deutrons use the G4MultipleScattering and G4hLowEnergyIonisation processes. Other nuclei, such as carbon and fluorine, use the McDarkIonMultipleScattering particle written specifically for this simulation which incorporates the McDarkUrbanMscModel. This class is an implementation of the multiple scattering model from H. W. Lewis in Phys. Rev. 78 (1950). The model takes into account corrections on cross sections and path lengths for the nuclei of interest. A.2.2 Optical The most important optical process for our simulation is scintillation, for which we use the G4Scintillation class, adjusting its parameters as necessary for alphas and heavy nuclei. In addition, optical photons- use the G40pAbsorption and G40pBoundaryProcess processes to govern their interactions. A.2.3 Hadronic The various mesons (pions, kaons) utilize both the appropriate low to high energy inelastic processes specified by their own GEANT4 classes (e.g. G4PionMinusInelastic, G4LEPionMinusInelastic, and G4HEPionMinusInelastic. Protons and antiprotons also follow this scheme. In neutrons, we consider the low to high energy processes for both elastic and inelastic cases. We also allow for the possibility of neutron capture with G4HadronCaptureProcess and G4NeutronHPCapture. Deutrons, alphas, and tritons also utilize their appropriate inelastic scattering processes. A.2.4 Decay All particles utilize the standard G4Decay process if decay is applicable. In addition, ions that engage in radioactive decay use the G4RadioactiveDecay process. A.3 Event Generation Per GEANT4 standards, event generation is handled through the McDarkPrimaryGeneratorAction class. This class contains the various distributions that are relevant to the experiment. The class currently contains spectra for a particle gun with variable en- ergy and direction, an isotropic source of variable energy, a Cf-252 source, a Co57 source, and a DT source. Additionally, the "Spergel distribution" is defined in McDarkSpergelDistribution which implements the dark matter energy and direction dustribution described by Spergel in 1988. Bibliography [1] F. Zwicky. Helvetica Physica Acta, 6:110-127, 1933. [2] D. N. Spergel et al.* A strophys. J. Suppl., 170:377, 2007. [3] S. Ahlen et al.* Int. J. Mod. Phys. A, 25:1-51, 2010. [4] S. Ahlen et al.'2010. hep-ex/1006.2928. [5] KAF-1001E. Eastman Kodak Company, Rochester, New York, March 2004. [6] D. Dujniie. Observations of head-tail effect in nuclear recoils. DMTPC Collaboration, Cambridge, MA, November 2010. [7] P. F. Rose. BNL-NCS-17541 (ENDF-201), 1991. Plots produced using the Online Service retrieval code package written by C. L. Dunford. [8] B. L. Dougherty. Physical Review A, 45:2104-2107, 1992.