Simulations of a Scintillator Compton Gamma Imager for Safety and Security

Transcription

1 PRESENTED AT SORMA WEST 8, BERKELEY, CALIFORNIA, JUNE 8 Simulations of a Scintillator Compton Gamma Imager for Safety and Security D.S.Hanna, A.M.L.MacLeod, P.R.B.Saull, and L.E.Sinclair arxiv:9.v [physics.ins-det] Sep 8 Abstract We are designing an all-scintillator Compton gamma imager for use in pre-event criminal and national security investigations involving radioactive threat material. The instrument will also be useful post-event, in the consequence management of an accidental or malicious dispersion of radioactivity. It has applications as well in decommissioning of contaminated facilities. To satisfy requirements for a rugged and portable instrument, we have chosen solid scintillator for the active volumes of both the scatter and absorber detectors. Using the BEAMnrc/EGSnrc Monte Carlo simulation package, we have constructed models using four different materials for the scatter detector: LaBr, NaI, CaF and. We have developed a measure to quantify detector performance and have used this to compare the performance of the different models. Using a fitting algorithm, we have reconstructed images from the simulated data. We find that LaBr provides the best performance. CaF and NaI provide similar performance, slightly worse than LaBr. All of the materials investigated for the scatter detector, including, have the potential to provide performance adequate for our purposes. I. INTRODUCTION A requirement for innovative detection technologies to assist investigators in intelligence gathering prior to or after a radiological or nuclear incident has been identified by Canada s Chemical, Biological, Radiological-Nuclear and Explosives Research and Technology Initiative (CRTI). To address this need, we are designing a Compton gamma imager. There are other groups investigating related imager designs, employing HPGe [], Si [] [], CZT [], or gaseous time projection chamber [] detector technologies. These techniques generally provide superior energy, and ultimately image, resolution on a per-event basis, to what can be achieved with scintillator materials. On the other hand, scintillators provide the benefit of a cost-effective way to produce a high-efficiency detector in a form which can readily be made compact and rugged, for deployment to the field. This portability constraint also introduces a need for an instrument which has low power consumption, and the cost-effectiveness of scintillator opens the possibility of deployment of more than one unit. There is an interesting approach to hybrid Compton and coded-aperture imaging ongoing, which also uses an allscintillator design [7]. Spare parts from an all-scintillator D.S.Hanna and A.M.L.MacLeod are with the Physics Department, McGill University, Montreal, Quebec, Canada P.R.B.Saull is with the Institute for National Measurement Standards, National Research Council, Ottawa, Ontario, Canada L.E.Sinclair, corresponding author, is with the Geological Survey of Canada, Natural Resources Canada, Ottawa, Ontario, Canada. Correspondence: laurel.sinclair@nrcan.gc.ca This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible. Fig.. Fig.. E γ m c E a) b) E Compton scatter diagram, a) initial-state and b) final-state. absorber detector E, P scatter detector E, P Schematic diagram of a Compton gamma imager. space-borne Compton telescope have been used to demonstrate a capability at ground level of identifying sources of radiation [8]. There has also been a study indicating that an all-scintillator design could be promising for the detection of highly enriched uranium [9]. Here, we present design studies conducted using the BEAMnrc/EGSnrc simulation package [,]. This work will proceed toward the development of a prototype. θ II. COMPTON IMAGING The process of Compton scattering is illustrated in Figure. An incoming photon of energy E γ scatters from an atomic electron, leading to a final state in which there is an outgoing electron of energy E and an outgoing photon of energy E. A sketch of a Compton gamma imager is provided in Figure. The energy E is deposited at some location in a pixellated scatter detector. The outgoing photon escapes the scatter detector to deposit its energy, E, at some location in a position-sensitive absorber detector. The scattering angle between the initial and final state photons, θ, can be determined from the two energy deposits, according to, cosθ = + m c ( E γ E ), () φ θ

2 PRESENTED AT SORMA WEST 8, BERKELEY, CALIFORNIA, JUNE 8 where E γ = E + E and m c is the electron rest energy. Thus, the position of the source may be reconstructed to lie somewhere on a cone of opening angle θ with its axis along the line joining the positions of the two energy deposits, and its apex at the first energy deposit. By back-projecting the cones from several events onto an image plane, an image may be reconstructed from the positions where the cones overlap. III. DETECTOR MODELS Using the BEAMnrc/EGSnrc Monte Carlo simulation packages we have constructed models of Compton gamma imagers. The models consist of layers of scintillator cm x cm in cross-section, with cm thickness in the scatter detector and.8 cm thickness in the absorber detector. Four different materials have been tested for the scatter detector, LaBr, NaI, CaF and polyvinyltoluene-based plastic scintillator, hereinafter referred to as. The number of scatter detector layers and the inter-layer spacing is dependent upon the material. All of the models feature an absorber detector consisting of five layers of LaBr. In the following, the different models will be referred to by the name of the scatter detector material. The schematic imager shown in Figure, represents the CaF detector. To determine the optimal thickness of the scatter detector for each material, we looked at the probability for an incoming gamma to Compton scatter or to undergo a photo-electric process, as a function of material thickness. For this study, we have generated, events from a kev. mm x. mm square parallel beam incident at the centre of the front face of a cm x cm x cm cube of material. The probability for interaction to occur is presented in Figure as a function of depth within the slab. For all materials, as the thickness of the material increases, the probability of at least one Compton scatter occuring, represented by the solid histogram in Figure, increases. However, the probability of the photon being absorbed in a photo-electric process, represented by the dotted histogram, increases with thickness as well. The probability of secondary Compton scatters also increases with material thickness (not shown). This means that there is some thickness at which the probability of exactly one Compton scatter (and no more), is maximized. This probability is represented in Figure by the dashed histograms. In the models discussed here, we have chosen the material thickness of the scatter detector according to the maximum of the dashed curve in Figure. Thus the LaBr, NaI, CaF and scatter detectors feature two, three, four and eight, cm-thick slabs of scintillator, respectively. It is a simple matter to convince oneself that the image resolution of a Compton imager will improve, the farther apart the energy depositions E and E. In order to not penalize the denser materials in the comparison of detector performance, we have therefore allowed the spacing between slabs to be larger where there are fewer slabs. Under the constraint of similar portability, we required the overall detector volume to Densities for these materials are taken from the technical data provided by the commercial supplier Saint Gobain. The used in these simulations corresponds to Saint Gobain s general purpose scintillator BC-8. be similar for all four models. The spacing between layers of the scatter detector was chosen to be cm, cm,. cm, and cm for the LaBr, NaI, CaF and scatter detectors respectively. IV. EVENT SIMULATION To compare the performance of the detectors, we simulated, gammas from a mono-energetic point-source situated on axis, m distant. This corresponds to min. of exposure to a mci source. We investigated the energy dependence of the detector performance, choosing point sources of energy kev, kev, MeV,. MeV and MeV. We investigated three different sources of image degradation: initial-state electron physics effects, energy resolution effects, and the effect of finite detector segmentation. In EGSnrc, binding effects and Doppler broadening are treated according to the relativistic impulse approximation []. These effects are controlled by an input parameter and may easily be turned on or off. Smearing of the energy measurement in the scintillator and readout was applied to energy deposits in the NaI scatter detector using energy resolutions determined by experiment []. For all other materials, the energy deposits E i were smeared by a Gaussian distribution about the true energy deposited, of width C E i, where the constant C was determined for LaBr, CaF and by the constraint that the resolution at kev should come to.9, and respectively. To get an idea of the possible effect of underestimation of this parameter, we have simulated an additional detector with a less optimistic energy resoluton of at kev. A segmentation of the detector into cm pixels was also simulated. To determine whether full containment of the event was achieved, the spectra of the sum of the energies deposited in the scatter and absorber detectors were examined. A clear fullenergy deposition peak was observed for all detector types. Good fits to the peaks with the sum of a Gaussian distribution and a straight line distribution were obtained. The one-sigma widths of the Gaussians fit to the full-energy deposition peaks are presented in Table I. Note that the energy resolutions on the total energy, as presented in Table I, are in some cases considerably better than one would expect from the smearing which has been applied to the energy deposits in the scatter detectors. This is due to the fact that most of the energy is actually deposited in the absorber detector, so the overall detector energy resolution is dominated by the energy resolution of LaBr..9 is a typical energy resolution at kev for LaBr quoted by suppliers. Note that suppliers do not quote typical energy resolutions at kev for CaF and, because for these materials the high Compton to photo-electric cross-section ratio means that no photopeak for Cs-7 may be observed. We chose the energy resolutions of CaF and based on the energy resolution of NaI, and consideration of the number of optical photons produced by these materials relative to NaI ( and for CaF and, respectively). Experiments with quite different geometries from ours have obtained energy resolutions in ranging from [] to []. Of course, these quantities should be determined by experiment, for the particular configuration of scintillator and light-collection device chosen, and this will be the next phase of our detector design program.

3 PRESENTED AT SORMA WEST 8, BERKELEY, CALIFORNIA, JUNE 8 a) b) 8 8 c) d) at least Compton exactly Compton photo-absorbed 8 8 Fig.. Percentage of kev events undergoing interaction as a function of slab thickness for a) LaBr, b) NaI, c) CaF and d) materials. The solid histogram indicates the percentage of incoming gammas for which at least one Compton scatter will occur. The dashed histogram shows the percentage which will have undergone exactly one Compton scatter. The dotted histogram shows the percentage of events undergoing a photo-electric process, as a function of material thickness. TABLE I WIDTH OF GAUSSIAN FIT TO FULL-ENERGY DEPOSITION PEAK (KEV) Energy of source (kev) Detector LaBr NaI CaF (σ E =) (σ E =) Golden events were defined as those which satisfy: > kev energy deposited in scatter detector, > kev energy deposited in absorber detector, no more than one energy deposit in scatter detector, and total of energies deposited in scatter and absorber detector lies within two sigma of full-energy deposition peak. For each energy,, events were generated. The number of those passing the criteria for golden events, after simulation of all sources of smearing, is shown in Table II. It is interesting to note that over most of the energy range, around four to seven percent of incoming gammas lead to golden events, independent of the scatter detector material. TABLE II NUMBER OF GOLDEN EVENTS FOR, GENERATED EVENTS Energy of source (kev) Detector LaBr NaI CaF (σ E =) (σ E =) 98 This similarity in the overall detector efficiencies could have been predicted from Figure. There it was shown that, despite large differences between the materials in the number of events which Compton scatter, or undergo a photo-electric process, careful choice of the total scatter detector thickness can result in a similar percentage of events undergoing exactly one Compton scatter. Table II also makes it clear that with LaBr or NaI for the scatter detector, at least with the detector geometries chosen in this study, Compton imaging at energies of around kev will be limited in effectiveness by poor

4 PRESENTED AT SORMA WEST 8, BERKELEY, CALIFORNIA, JUNE 8 detector efficiency. Note that several sources of uncertainty and inefficiency have been left out of our simulation. We have not included dead material for the readout within the detector, nor is the quantum efficiency of the light collection included. The simulation also does not include background due to other sources of radiation. V. RESULTS A. Cone Reconstruction Performance For each golden event, we calculated the Compton cone opening angle according to Equation. The Compton cones were back-projected onto an image-sphere. We then looked at the number of events for which the reconstructed Compton cone came within two degrees of the true source location. These are called well-reconstructed events. Figure shows the number of well-reconstructed events, as a percentage of the number of generated events, as a function of source energy, for four of the detectors. The degradation of image resolution due to initial-state electron effects, and the effects of energy resolution and position segmentation has been added successively to the simulation. The dashed lines in Figure show the percentage of generated events which are well-reconstructed allowing for effects such as backscattering, which could lead to misassignment of the scatter and absorption occurrences. No additional source of smearing is included in the dashed curves. The dotted curve shows the incremental image degradation due to including a detailed treatment of effects associated with the initial-state electron binding effects and Doppler broadening. The dash-dotted curve shows the effect of including also the smearing of energies in the scatter and absorber detectors. The solid curve is the final result for each detector. It shows the percentage of generated events which are well-reconstructed after every treated source of image degradation including segmentation of the scatter and absorber detectors into cm pixels. As the thickness of the scatter detector was chosen to optimize the acceptance of Compton events from Cs-7, which emits kev gammas, it is not surprising that overall, the number of well-reconstructed events peaks at some energy around. to MeV. More information can be gleaned about how the different detectors are performing by comparing their behaviour under the three different smearing conditions. Both the LaBr and NaI detectors exhibit significant image degradation due to initial-state electron effects, at all energies. For the CaF detector, the initial-state electron effects decrease with increasing energy. For the (σ E = ) detector, the initial-state electron binding and Doppler broadening effects are really only an issue below around kev. Initial-state electron effects in the the LaBr detector are larger effects than its energy resolution. For the NaI detector, energy resolution and initial-state electron physics are similarly-sized effects. For the CaF and (σ E = ) detectors on the other hand, the energy resolution is a much more severe effect than the initialstate electron binding and momentum, though the effect is lessened with increasing energy. The chosen detector segmentation of cm is a sensible choice for these designs. It does introduce an additional image degradation. However, for most of the designs the image degradation due to position segmentation is similar in magnitude to that caused by the other effects. In Figure, the four curves representing the fraction of generated events which are well-reconstructed after all sources of smearing from each detector are overlaid. Also included is the corresponding curve for the detector with σ E = at kev. Figure indicates that at an energy of around kev, a detector composed entirely of LaBr should perform the best. The CaF and NaI detectors come in second place with the CaF detector possibly exhibiting slightly superior performance. Although we have assigned CaF a poorer energy resolution, its superior Compton to photo-electric crosssection ratio means that a larger number of events can be well-reconstructed with the CaF detector than with the NaI detector, in the same period of time. Finally, despite their very low initial-state electron effects, the detectors, with their poor energy resolution, perform the worst of the four detectortypes examined in this study. B. Image Reconstruction Peformance To reconstruct images, we are developing an algorithm based on fitting procedures. In this code, a χ minimization based on the distance of closest approach to each backprojected Compton cone (including its estimated uncertainty) is employed to reconstruct the most likely position of the source. For the image reconstruction study, we looked at the equivalent of 7 s of accumulation of data for each detector, from a mci kev source m away. We then applied the golden event selection criteria described in Section IV. Figure shows the -σ, -σ and -σ confidence intervals returned by the fit for a) the LaBr detector, and b) the (σ E = ) detector, overlaid on sketches of typical objects which may be expected in the field of view. The images suggest that from a safe distance, even the poorer detector can likely distinguish a place harbouring a significant amount of radioactive material from one which is not, in a short amount of time. The approximate half-width of the -σ contours for 7 s of accumulation from the same source, are provided in Table III for each of the detector models studied. Again, we find the performance of the LaBr detector to be the strongest, with the NaI and CaF detectors a close second. With a good image reconstruction algorithm, it appears that even a scatter detector based on may have adequate localization ability. These results are strongly suggestive that a detector based on the designs presented in this paper, should be able to localize a mci source m away, to within a few metres, in under a minute.

5 PRESENTED AT SORMA WEST 8, BERKELEY, CALIFORNIA, JUNE 8 a)... b)... Smearing: None - + Initial-state e + Energy + Position c)... d)... Fig.. Percentage of generated events for which the back-projected cone comes within two degrees of true source location for a) LaBr, b) NaI, c) CaF and d) (σ E = ) detectors, as a function of source energy. The dashed curve shows the result including no smearing (other than misassignments). The dotted curve includes the smearing due to initial-state electron binding effects and Doppler broadening. The dash-dot curves show the effect of including energy smearing. The solid curve shows the total image degradation including the effect of segmentation of the scatter and absorber detectors. The shaded band on each of the curves indicates the statistical error... LaBr NaI CaF (σ E = ) (σ E = ) TABLE III INDICATIVE IMAGE RESOLUTION Detector LaBr NaI CaF (σ E = ) (σ E = ) Resolution Fig.. Percentage of generated events for which the back-projected cone comes within two degrees of true source location including all treated sources of physics and detector smearing. The solid curve shows the result for the LaBr detector. The dashed curve shows the result for the NaI detector. The dotted curve shows the result for the CaF detector. The dash-dotted and longdash-dotted curves show the results for the detectors with σ E = and σ E =, respectively. The statistical error on each of the curves is indicated by a shaded band. VI. CONCLUSION The performance of four different all-scintillator Compton gamma imager models based on different materials for the scatter detector, has been investigated. We have obtained encouraging results from all four of the scintillators looked at, LaBr, NaI, CaF, and. The all-labr detector is predicted to perform the best of the models studied, with NaI and CaF coming in second. Indications are that even the worst model studied, with a scatter detector composed of, may be able to provide an image of a mci kev source at m with a resolution of around two degrees, in under a minute. The next stage of this work will be to establish a test stand to validate the Monte Carlo studies experimentally.

6 PRESENTED AT SORMA WEST 8, BERKELEY, CALIFORNIA, JUNE 8 a) [] K. Hattori et al, Gamma-ray imaging with a large micro-tpc and a scintillation camera, Nucl. Instr. Meth., vo. A 8 p. 7, 7. [7] W. Lee, D. Wehe, Hybrid gamma ray imaging Model and results, Nucl. Instr. Meth., vol. A 79, p., 7. [8] J.M. Ryan, J. Baker, J.R. Macri, M.L. McConnell and R. Carande, A Compton telescope for remote location and identification of radioactive material, Proc. SPIE, vol. 9, art. 9, 8. [9] B.F. Phlips, E.I. Novikova, E.A. Wulf and J.D. Kurfess, Comparison of shielded uranium passive gamma-ray detection methods, Proc. SPIE, vol., art. H,. [] D.W.O. Rogers, B.A. Faddegon, G.X. Ding, C.-M. Ma and T.R. Mackie, BEAM: A Monte Carlo code to simulate radiotherapy treatment units, Med. Phys., vol. (), p., 99. [] I. Kawrakow and D.W.O. Rogers, The EGSnrc Code System: Monte Carlo Simulation of Electron and Photon Transport, PIRS-7, National Research Council of Canada, Ottawa, Canada,. [] R. Ribberfors, Relationship of the relativistic Compton cross section to the momentum distribution of bound electron states, Phys. Rev., vol. B, p. 7, 97. [] M. Moszyński et al, Intrinsic energy resolution of NaI(Tl), Nucl. Instr. Meth., vol. A 8, p. 9,. [] S.-B. Tang, Q. Ma, Z. Yin, Y. Cheng, D. Zhu, Numerical simulation of simple position-sensitive gamma-ray detector based on plastic scintillating fiber array (OFT--R), Opt. Fiber Tech., vol., p., 8. [] P.L. Reeder, D.V. Jordan, L.C. Todd, G.A. Warren, Advanced Large- Area Plastic Scintillator Project (ALPS): Final Report, PNNL-7, Pacific Northwest National Laboratory, Richland, Washington, USA, 7. b) Fig.. Reconstructed image of a mci kev source located in the doorway of a house m away from a) the LaBr detector and b) the (σ E = ) detector. The stick figure is cm tall. The -σ, -σ and -σ confidence intervals after 7 s of accumulation of data are indicated by the shaded contours. The true source position is indicated by the black dot. ACKNOWLEDGMENT The authors thank H. Seywerd and J. Carson for critical readings of the text. This work will be proceeding to the development of a prototype, supported through funding from the Chemical, Biological, Radiological-Nuclear and Explosives, Research and Technology Initiative (CRTI Project 7-9RD). REFERENCES [] L. Mihailescu, K.M. Vetter, M.T. Burks, E.L. Hull and W.W. Craig, SPEIR: A Ge Compton camera, Nucl. Instr. Meth., vol. A 7, p. 89, 7. [] E.A. Wulf et al, Compton imager for detection of special nuclear material, Nucl. Instr. Meth., vol. A 79, p. 7, 7. [] K. Vetter et al, High-sensitivity Compton imaging with positionsensitive Si and Ge detectors, Nucl. Instr. Meth., vol. A 79, p., 7. [] A.S. Hoover et al, Gamma-ray imaging with a Si/CsI(Tl) Compton detector, Appl. Radiat. Isot., vol., p. 8,. [] B. Smith et al, An Electronically-collimated Gamma-ray Detector for Localization of Radiation Sources, IEEE Nuclear Science Symposium Conference Record, p. 7,.