PERFORMANCE TESTING OF X-RAY AND GAMMA-RAY DETECTORS FOR IMAGING AND SPECTROSCOPY. Richard Giordmaina. Department of Physics. University of Surrey

Transcription

1 PERFORMANCE TESTING OF X-RAY AND GAMMA-RAY DETECTORS FOR IMAGING AND SPECTROSCOPY Richard Giordmaina A dissertation submitted to the Physics Department at the University of Surrey in partial fulfilment of the degree of Master in Physics. Department of Physics University of Surrey April 2008

2 ABSTRACT Spectroscopic detection of X and gamma (γ) radiation is of great importance in medical and security applications. Research was undertaken at the University of Surrey and the Dstl Fort Halstead laboratories to identify the imaging and spectroscopic capability with both existing and novel radiation detection technology. There is a real need for spectroscopic information to be provided by radiation detectors to offer the possibility for material discrimination, providing essential benefits that come with this knowledge. Testing of the imaging and spectroscopic detection capabilities took place mainly using a Cadmium Zinc Telluride (CZT) detector array, and novel Silicon Photomultiplier (SPM) detector technology. An initial review explored the various detection technologies which are currently available, populating a graphical model to illustrate how these link together. Preparation and characterisation of CZT radiation detectors was conducted to explore how raw detector materials can be tested and made ready for use, and research using these detectors continued throughout the year, involving many experiments to characterise and measure the performance of radiation detection equipment. A CZT detector array was characterised by measuring the uniformity and energy resolution using radioactive calibration sources to identify the spectroscopic capability of the detector. This was repeated after a hardware upgrade to higher quality CZT, showing a noticeable improvement. It was found that the CZT has energy resolutions of % at 59.5keV and % at 122keV (an increase with energy as expected). The spectroscopic performance was then tested using two calibration sources to determine the linearity of the counts recorded with time and inverse square law with distance. The efficiency was found to be 72 1% at 122keV, proving that most photons will be detected at 140keV, an energy commonly used in medical imaging. The CZT detector was then used as a pinhole imager of backscattered X-ray photons from a variety of objects, and the resulting images were compared to those obtained with an Intensified Charged Coupled Device (ICCD) detector. These demonstrated that backscatter X-ray images can be produced with both detectors. Analysis of the acquired images showed that the larger pixels of the CZT detector make a more sensitive and better quality imaging camera, although with coarser images than the ICCD. The majority of the research placement explored the characterisation and spectroscopic capability of Silicon Photomultiplier (SPM) detectors, a relatively new technology, which are marketed to be a possible replacement for Photomultiplier Tubes (PMTs) in many detection applications. A model exploring the expected efficiency and energy resolution when different single pixel SPM detectors (1mm and 3mm pixel sizes) are coupled to different scintillator crystals was produced to identify if spectroscopy is theoretically possible. It was found that in certain scintillator, γ-source and SPM combinations, spectroscopy would be possible as there would be enough photons remaining after losses for an energy resolution of less than 15% to be obtained. The SPMs were then tested by measuring the spectra produced when different scintillator crystals (CdWO 4, CsI(Tl), BGO and LYSO) were coupled to the three different SPMs and irradiated with radioactive γ-sources ( 241 Am, 57 Co, 22 Na and 137 Cs). Many results found were directly comparable to the results expected using the model. From the various measurements, energy resolutions were achieved, including % at 662keV. The practical use for SPM detectors as small, fast counting spectroscopic radiation detectors has been shown to be possible from the results obtained. ii

3 Acknowledgements The author would like to take this opportunity to thank Dr Regan and Dr Sellin for organising the placement which was a new, varied and exciting experience, which will no-doubt be of benefit in the future. In addition, their assistance during the initial placement delays was greatly appreciated. Dr Sellin has been very supportive as visiting tutor, by being available during the research year for advice and help during my time away from the University. The author would also like to show gratitude to his friends and colleagues at Dstl, especially Ian, Jane, Dave and Paul for providing guidance and assistance, especially at the beginning of the new role. Additionally thanks go to Paul for the use of portions of his existing programming code, which saved 're-inventing the wheel' in several cases, and for his help in checking through sections of new code. Finally, the author would like to thank his friends and family, for their support, patience and understanding during the research year and whilst writing this dissertation. Author Declarations Whilst preparing this dissertation, the author has not been registered for any other academic qualifications, and this dissertation has only been submitted for the Master of Physics academic award. iii

4 List of Abbreviations A list of all the abbreviations used throughout this dissertation is included here. Abbreviation APD BGO CCD CdWO 4 Cs CsI(Tl) CT CZT DAQ DGF Dstl ehp(s) et al. etc. ev FF FNA FWHM GAPD GMS IC ICCD IV LED LYSO MCP M.Phys. MCA MOD NaI NIST NQR PAB PDE PFNA PMT ROI SCA SPES SPM SNR TNA UK XCOM XIA Description Avalanche Photodiode Bismuth Germinate Charge Coupled Device Cadmium Tungstate Cesium Cesium Iodide (Thallium activated) Computerised Tomography Cadmium Zinc Telluride Data Acquisition System Digital Gamma Finder Defence Science and Technology Laboratory Electron-hole pair(s) and others et cetera Electron volt Fill Factor Fast Neutron Activation Full Width at Half Maximum Geiger mode Avalanche Photodiode Graphical Modelling System Integrated Circuit Intensified Charged Coupled Device Current-Voltage Light Emitting Diode Lutetium Yttrium Silicon Dioxide Micro Channel Plate Master in Physics Multi Channel Analyser Ministry of Defence Sodium Iodide National Institute of Standards and Technology Nuclear Quadrupole Resonance Probability to initiate Avalanche Breakdown Photon Detection Efficiency Pulsed Fast Neutron Activation Photomultiplier Tube Region of Interest Single Channel Analyser Single Photoelectron Spectrum Silicon Photomultiplier Signal to Noise Ratio Thermal Neutron Activation United Kingdom Attenuation Database X-ray Instrumentation Associates iv

5 List of Tables A list of all the tables used in this dissertation is included here. Table 2-1: Scintillator crystals and their key properties [12 (LYSO details from 13)] Table 3-1: Procured calibration source details. (*Used for CZT detector uniformity and energy resolution experiments and were original laboratory sources.) Table 4-1: The results for the uniformity (after flat fielding) and energy resolution of the CZT detector before and after the hardware upgrade Table 4-2: The uniformity of the ICCD detector for each source before and after applying the flat field corrections Table 4-3: The effect of flat fielding on the CZT Gaussian peaks, for the upgraded detector Table 4-4: A close match for the measured values to that expected from the model for energy resolution measurements Table 4-5: The average measured energy resolutions (%). (Errors at the 95% confidence level) Table 5-1: The best obtainable spectroscopic energy resolutions (nearest %) for the two detectors at the same energies Table 5-2: The best obtainable detection efficiencies (nearest %) for the two detectors at the same energies Table 6-1: Resistivity values and results from the conducting foam experiment v

6 List of Figures A list of all the figures in this dissertation is included here. Figure 1-1: The summed attenuation in two regions for various materials over 2cm Figure 2-1: The electromagnetic spectrum (left) [4] and a diagram of an X-ray tube (right) [5] Figure 2-2: A typical X-ray spectrum showing the characteristic peaks of the target material and the Bremsstrahlung spectrum which accounts for most of the energy [4] Figure 2-3: The energy level transitions in an atom Figure 2-4: The main photon interaction mechanisms in matter varying with energy [6] Figure 2-5: Possible photon interactions in a detector (left) [6] and the Compton scattering of an electron (right) [4] Figure 2-6: A typical X-ray transmission imaging arrangement Figure 2-7: An image of a brain tumour identified by 99 Tc m imaging [3] Figure 2-8: Pinhole imaging providing sharper images for a smaller entrance aperture [8] Figure 2-9: The operation of a semiconductor (CZT) radiation detector [6] Figure 2-10: The drift velocity (V d ) as a function of applied electric field [6] Figure 2-11: The stages in an ICCD detector [11] Figure 2-12: The readout in a charged coupled device [6] Figure 2-13: The emission of scintillation crystals and the response of PMT devices [6] Figure 2-14: The scintillation process from activated states [6] Figure 2-15: A photomultiplier tube with the amplification stages in the dynodes [3] Figure 2-16: A single pixel SPM (on a square base approximately 3x3cm) [20] Figure 2-17: A 1mm 2 SPM pixel with many microcells on the top (left) [17] and the GAPDs in the pixel are connected together to provide a photon proportional output [17] (right) Figure 2-18: A graph containing the PDEs for 4V over bias for various SPM products [19] Figure 2-19: The attenuation ratio I/I 0 for the three detector materials explored for the project Figure 2-20: The attenuation ratio over the energy range to be tested for the four scintillator crystals used at 3mm thickness and the CZT detector at 5mm for direct comparison Figure 2-21: Operation of an MCA, an extension of many single channel analysers (SCAs) [6] Figure 2-22: Typical pulse height spectra from γ-sources [6] Figure 2-23: An improved energy resolution is obtained with a thinner peak [6] Figure 2-24: The source emitting in 4π, where only a proportion of the activity is incident on the detector at some distance (d) away Figure 3-1: The CZT detector array Figure 3-2: The experimental setup for the uniformity and energy resolution measurements Figure 3-3: An image (based on detector area ~12.5cm 2 ) from a γ-source illumination (left) and a histogram from which the detector uniformity can be calculated (right) vi

7 Figure 3-4: The Gaussian fit to the 241 Am peak (at 0.5m for 10 minutes) Figure 3-5: The ICCD detector (with a 2mm pinhole plate attached) Figure 3-6: The housing applied to the CZT detector providing a pinhole aperture at 10cm from the centre of the detector Figure 3-7: Top view of the arrangement for X-ray backscatter pinhole imaging Figure 3-8: An image created using a pinhole receiving backscattered X-rays (left) and the effect chamfering has on the pinhole (right-above before and right-below after chamfering) Figure 3-9: The leakage obtained before the additional shielding (detector area ~7.9x10-3 m 2 ) Figure 3-10: The optimum shielding configuration (left) and the more uniformly distributed counts with this shielding Figure 3-11: The experimental arrangement for separation experiment (CZT pinhole covered for a background measurement) Figure 3-12: The positions of the tubes of salt and sugar, moving closer together in 1cm steps Figure 3-13: The line of data extracted for ICCD images (left) and CZT images (right) Figure 3-14: The transistor astable circuits produced to pulse an LED [30] (left) and a circuit diagram for the 555 IC used to pulse an LED [32] (right) Figure 3-15: The oscillator circuits produced to pulse an LED using transistors and capacitors (left) and the pre-packaged oscillator in the IC NE555 timer (right) Figure 3-16: Experimental arrangement for the SPM pulse linearity experiment Figure 3-17: The pulse exploration and SPM spectroscopy experimental arrangement Figure 3-18: The Xia Pixie-4 system used to acquire the spectrum from the SPM detectors Figure 3-19: A summary of the planned SPM testing, showing each SPM coupled to each scintillator crystal activated by each γ-source, and the background measurements Figure 3-20: The single photoelectron spectrum possible with an SPM [21 modified] Figure 4-1: The activity calculator with extrapolated activity over two months Figure 4-2: A screenshot from the shielding requirements spreadsheet for pinhole imaging Figure 4-3: An image (left) and histogram (right) from the 241 Am illumination (over 30 minutes) Figure 4-4: A histogram with a fitted Gaussian for the 30 minute integration of 241 Am Figure 4-5: Flat field corrections applied to the 10 minute data for 241 Am (left) and 57 Co (right) Figure 4-6: The energy resolution per pixel for the 57 Co 30 minute integration (left) and for the 241 Am 30 minute integration (right) Figure 4-7: Energy spectra using 57 Co for 30 minutes (left) and 241 Am for 30 minutes (right) Figure 4-8: Illumination of the 30 minute 57 Co (left) and 241 Am for 30 minutes (right) showing crystal artefacts Figure 4-9: Flat fielded image with 57 Co (left) and 241 Am (right) showing the artefact has now gone. Note the bright pixels (right) mask the true discontinuities image Figure 4-10: The effect of flat fielding on the 30 minute flood illuminations for both sources 241 Am (left) and 57 Co (right) Figure 4-11: A linear relationship successfully identified between increase in counts and integration time, for two γ-sources vii

8 Figure 4-12: The energy spectrum after just one second, showing 241 Am (59.5keV) and 57 Co source (122 and 136keV) peaks Figure 4-13: The effect of time and distance on the maximum number of counts received for two of the integrations times tested using the CZT detector, showing the inverse square law Figure 4-14: The measured energy spectrum (left) and one just published for 57 Co (right) [38] Figure 4-15: The efficiency of the CZT detector with increasing source distance Figure 4-16: A direct comparison for the same object (sugar) with the 4mm and 2mm pinholes (before flat field corrections) from the ICCD detector Figure 4-17: Flat fielded images of cotton in a case, using CZT (left) and ICCD (right) Figure 4-18: Talc and aluminium powder on a plastic base (flat fielded) using CZT (left), ICCD (middle), and before flat field corrections applied to the CZT (right) Figure 4-19: Images of the same object (but reversed) of the sections taken for analysis of the mean and standard deviation (CZT left and ICCD right) Figure 4-20: A graphical representation of the data collected for the ratio of object to background each image from both detectors Figure 4-21: The number of counts from the average of 10 pixels in the CZT and ICCD detector for all of the objects (error as the standard deviation) Figure 4-22: 1cm apart before flat field corrections (left) and after (right) for the CZT detector Figure 4-23: 2-4cm (left to right) after flat field corrections for the CZT detector Figure 4-24: 1cm apart before flat field corrections (left) and after (right) for the ICCD detector Figure 4-25: 2-4cm (left to right) after flat field corrections with 57 Co 30 minutes for the ICCD detector Figure 4-26: The separation seen using the CZT (top) and ICCD (bottom) detectors over 5cm, where the background for the ICCD is always higher than CZT Figure 4-27: The separation displayed as the change in normalised intensity for both detectors (error bars as smallest unit taken from the trough height) Figure 4-28: Results for the modelled energy resolutions for each crystal and SPM over the energy rage to be tested. Clockwise from top left: CsI(Tl), BGO, LYSO and CdWO Figure 4-29: The 1mm SPM pulses with no sources present seen in oscilloscope mode on the DAQ (left) and expanded on another oscilloscope showing distinguishable dark photons (right) Figure 4-30: The dark counts in the 3mm 20μm SPM (top) and 3mm 35μm SPM (bottom), showing noise photons not as clearly defined as the 1mm SPM Figure 4-31: 3mm 35µm SPM at 32V bias using the 555 IC timer (left) and 3mm 35µm 32V (right) for an LED pulse where the SPM response (pink) to the LED pulse (blue) Figure 4-32: The onset times of all three SPMs using pulsed LED circuits to show that the pulses are indeed produced in around 12ns Figure 4-33: A pulse caused by the pulsing of a red square LED, with black tape around the four exposed sides on the 1mm SPM Figure 4-34: The SPM response to an LED pulse not fully recorded Figure 4-35: The counting of the SPM to be at least 2x10 5 Hz viii

9 Figure 4-36: The effect of resistance (and therefore LED power from the 555IC timer) on the size of the SPM pulses produced Figure 4-37: A scintillation pulse from the 3mm 20µm SPM (at 32V bias) using CsI(Tl) with the 22 Na source Figure 4-38: A comparison of the voltage pulses from scintillator crystals and sources using the 3mm 20μm SPM at 32V Figure 4-39: The CdWO 4 response to 22 Na on the 3mm 20μm SPM, the decay lasting much less than the 20µs expected Figure 4-40: Comparing the response of the pulse preamplifier which cuts off the full pulse duration (left) with the transimpedance amplifier (right) Figure 4-41: The effect of bias on the SNR for each SPM Figure 4-42: The energy spectrum using CsI(Tl) with the 22 Na source, providing an energy resolution of 14% using the 3mm 35μm SPM Figure 4-43: The modelled energy resolutions (top) and measured average energy resolutions using CsI(Tl) for each SPM (bottom) Figure 4-44: An energy spectrum from the 3mm 20μm SPM using BGO and 137 Cs Figure 4-45: Two separate SPM linearity experiments; using the 3mm 20μm SPM with 137 Cs and CsI(Tl), and the 3mm 35μm SPM with 22 Na and BGO Figure 4-46: A ln-ln plot of energy vs. energy resolution using BGO on the 3mm 35μm SPM Figure 4-47: The average measured energy resolution results using BGO, LYSO and CsI(Tl) scintillator crystals on the 3mm 20µm SPM Figure 4-48: The average measured energy resolution results from several scintillator crystals coupled to the 3mm 35µm SPM Figure 4-49: The average measured energy resolution results for the CsI(Tl) scintillator crystal compared to the modelled result for the 1mm SPM Figure 4-50: The measured energy resolutions with the 1mm SPM and the modelled results based on a complete match in areas and at a 1/9 area match Figure 4-51: A comparison of the measured energy resolutions for the three SPMs using all of the sources on the CsI(Tl) crystal Figure 4-52: The extrapolated energy resolution (based on measured results) to 140keV for eight scintillator and SPM combinations Figure 4-53: A 137 Cs Spectrum for two minutes on the 3mm 35μm SPM (left) and the effect seen when 241 Am (further back at 5cm) from the crystal is added to the experiment (right) Figure 4-54: The 122keV peak clearly to the right of the cursors showing the position around the 59.5keV peak Figure 4-55: Illumination with 57 Co (left) and both 241 Am and 57 Co (right) showing a broadening due to the 59.5keV source from 33% to 52% at 150 counts Figure 4-56: Both high energy sources (511 and 662keV) when integrating for two minutes Figure 4-57: The energy spectrum for 22 Na (511keV) without the 662keV source, the peak is missing when integrating for two minutes ix

10 Figure 4-58: BGO with 137 Cs for two minutes on the 3mm 35μm SPM Figure 4-59: The addition of 22 Na to the 137 Cs source for two minutes Figure 4-60: Integrating 22 Na and 137 Cs for 30 minutes better defines the spectrum Figure 4-61: The beta spectrum from LYSO taken for 30 minutes with no additional radioactive sources using the 3mm 20μm SPM Figure 4-62: The linear increase of counts with integration time for the 3mm 35µm using 57 Co and CsI(Tl) Figure 4-63: A comparison of the system efficiencies for each SPM using CsI(Tl) Figure 5-1: An array of 16 3mm pixel SPMs [34] Figure 6-1: The system start-up screen Figure 6-2: The Pixie4 Run Control menu Figure 6-3: The positions of the cursors to find the energy resolution Figure 6-4: The cursors around the peak to provide the energy resolution and value of the peak Figure 6-5: The attenuation jumpers for each channel of the acquisition system Figure 6-6: GMS with no object selected Figure 6-7: Identifying who takes Mathematics by moving the mouse over Mathematics Figure 6-8: The Weld View of the Students Tutorial Figure 6-9: A schematic of the conducting foam used [35] Figure 6-10: The experimental arrangement to find the CZT sample resistivity Figure 6-11: The change of source energy from 662keV (top) to 59.5keV (bottom) x

11 Contents ABSTRACT... ii Acknowledgements... iii Author Declarations... iii List of Abbreviations... iv List of Tables... v List of Figures... vi Contents... xi Chapter 1 : Introduction and Background Background... 1 The Research Problem... 2 Detection Techniques Review... 3 Material Modelling... 4 Chapter 2 : Theory Production and Properties of X and Gamma-rays... 6 X-rays... 6 Radioactive Sources... 8 Photon Interactions in Matter... 8 Photon Attenuation Detection Technologies X-ray Transmission Imaging Medical Imaging Backscatter X-ray Imaging Radiation Detection Systems Cadmium Zinc Telluride (CZT) Detectors Intensified Charge Coupled Device Detectors Scintillator Crystal Properties The Silicon Photomultiplier (SPM) Radiation Measurement and Spectroscopy What Happens to the Detected Radiation? Pulse Height Spectra Detector Calibration How is Spectroscopic Capability Determined? Error Analysis xi

12 Chapter 3 : Experiments and Work Conducted Calculations and Modelling Radioactive Source Calculator CZT Shielding Requirements CZT Experiments The CZT Detector Uniformity and Energy Resolution CZT Spectroscopy CZT Efficiency Measurements The ICCD Detector CZT and ICCD X-ray Backscatter Pinhole Imaging CZT and ICCD Angular Resolution SPM Experiments Scintillator and SPM Energy Resolution Model Light Emitting Diode (LED) testing SPM Pulse Linearity SPM Pulse Observations SPM Spectroscopy SPM Detector Efficiency The Single Photo Electron Spectrum Chapter 4 : Results and Analysis Modelling Results Radioactive Source Calculator CZT Shielding Results CZT Experimentation Results Uniformity and Energy Resolution CZT Spectroscopy CZT Efficiency CZT and ICCD X-ray Backscatter Imaging Angular Resolution Scintillator and SPM Experiments Scintillator and SPM Energy Resolution Model Scintillator and SPM: Preliminary Test Results xii

13 LED Testing SPM Pulse Linearity SPM Pulse Observations SPM Spectroscopy Observing Complex Spectra using SPMs SPM Detection Time SPM Efficiency Results Chapter 5 : Review, Conclusions and Further Work Conclusions Further Research References Chapter 6 : Appendices APPENDIX I: OBTAINING A SPECTRUM APPENDIX II: GMS APPENDIX III: Initial Research: Introduction and CZT Resistivity APPENDIX IV: Experimental Equipment List xiii

14 Chapter 1 : Introduction and Background This dissertation describes the research conducted in 2007 into current and novel spectroscopic and imaging radiation detectors, which could have widespread applications. 1.1 Background The research placement at the Defence Science and Technology Laboratory (Dstl) started in April. Dstl is a trading fund of the Ministry of Defence (MOD), comprising over 3500 employees in several locations across the UK providing essential, impartial, high quality, timely advice on science and technology issues. Working for the UK Armed Forces, the MOD or other government departments, Dstl does not engage in work that can be done outside Government and therefore does not compete for business with Industry [1]. Dstl is divided into 14 departments (including Electronics, Biomedical Sciences and Physical Sciences) and teams [1]. The research placement was undertaken in the Physical Detection team of the Energetics Department. The broad aims of the team are to investigate, develop and refine the sensors used in non-invasive detection. Technologies using X-rays, neutrons and nuclear quadrupole resonance are some of those of interest. There is a vital need to detect materials non-invasively, and an obvious example of this is aviation security which has the requirement for rapid threat detection, due to a high volume of items. Frequently, imaging using transmitted X-rays is performed, which relies upon the attenuation of photons due to different absorption in different materials to produce an image. When materials overlap, differentiation is more difficult as a thick amount of one material could be similar to a thinner amount of another [2]. A key question is could detection systems provide any more information? For example, if spectroscopic information was obtained, there could be the possibility to identify materials based on their elemental composition, rather than just an image to be interpreted. The need for non-invasive detection in also vital for diagnostic medicine using passive γ-radiation detectors to identify tumours in Positron Emission Technology (PET) and Single Photon Emission Computerised Tomography (SPECT), where it is not always required to operate on a person to diagnose certain tumours. By inducing a γ-emitter which is taken up by a tumour, it is possible to build 1

15 up a picture of the inner body from the outside. The energy range of the diagnostic medicine includes 140keV and 511keV (higher for some radiological treatments). The possibility for improved detectors could have important benefits including rapid detection (to reduce exposure time and dose required) and imaging in medical and threat detection applications. The placement has explored three strands of detection: active transmission X-ray modelling to determine how well metals and organics can be separated; active backscatter X-ray imaging to identify imaging when there is only access to one side of an object; passive γ-ray spectroscopy and detector characterisation. The Research Problem The research has been driven by the requirement to improve the current sensors for radiation detection measurements, by exploring the characteristics of new detector materials and their ability to provide spectroscopy. There is the potential to improve the information provided by detection systems. Detailed detector characterisation was conducted by: identifying the detector performance of a CZT detector array (before and after a hardware upgrade), in terms of the uniformity (and comparing this to existing imaging technology using an ICCD detector) and spectroscopic energy resolution to determine the level of peak separation possible; exploring the detection efficiency of the CZT detector using single and multiple γ-sources (to include linearity of count rate over integration time and confirming the inverse square law with increasing distance from the detector); demonstrating backscatter X-ray imaging through a pinhole using the CZT detector with a specifically designed graded material housing, and comparing these images to ones taken with existing technology (an ICCD detector); 2

16 characterisation of three novel single pixel Silicon Photomultipliers (SPMs) to identify detection efficiency, proportionality of the light output and the noise of the detector to establish the performance possible; spectroscopic testing of the SPMs with several different scintillator crystals in combination with a large energy range of γ-sources; comparing the different detection systems tested in terms of detection efficiency, energy resolution, portability and linearity of the detectors. All of this research focuses on identifying the characteristics and capabilities of two main detector systems detectors (CZT, and scintillators with SPMs) as spectroscopic imagers to quantify their current characteristics and to determine the suitability of these in passive γ-ray detectors, active transmission or active backscatter detectors. Should these new detector systems have superior qualities, they could replace existing technology in many applications. Detection Techniques Review The need for the research was identified by conducting a review of current detection techniques used, providing an understanding of the systems that currently exist, and their method of operation. This included metal detectors, X-ray detection (such as transmission and backscatter), neutron-based detectors (such as Thermal, Fast and Pulsed Fast Neutron Analysis (TNA, FNA, PFNA)), vapour detection, millimetre wave radiation and Nuclear Quadrupole Resonance (NQR). Information was found using journals, texts and the Internet (product manufacturer web pages) for current and relevant information. Where a piece of equipment using the technique existed, it was explored for advantages and disadvantages. Using a graphical modelling system (GMS 4.2 produced by the Office of Naval Research) a model was produced summarising this information based on equipment identified, showing graphically the different detection techniques. The software allows a technique to be selected, to display equipment available using the technique, or select an item to be detected (e.g. metals) to display methods to detect this. The report and model were designed to give an understanding of the techniques available and how the research over the placement would aid the area of detection by adding spectroscopic sensors. 3

17 Material Modelling To create a complete picture of the three detection methods, X-ray transmission imaging was explored to identify the level of material discrimination possible. The attenuation data of nine common materials (aluminium, copper, iron, salt, iron, Perspex, cotton, water and sucrose) was investigated using XCOM (an Internet based attenuation calculator). The ratio of X-ray intensities through to that absorbed was calculated for a range of energies for a range of material thickness. This explored how the X-ray attenuation in these materials varies with X-ray energy and material thickness over a 1-140keV (every kev) range over a material thickness of 2mm up to 2cm thickness (every 2mm). A pre-existing modelled X-ray spectrum was normalised for intensity at each energy, and was multiplied into the material attenuation data in a spreadsheet to produce attenuation curves (or banana curves ) for each material. Attenuation curves were made using data at specific attenuation energies; 80 and 140keV, and 90 and 120keV. Then by dividing the 140keV energy range into sections (1-85keV and keV) the sum of the attenuations for each thickness in each section was plotted. The level of material discrimination possible is determined by how close the data points lie together. Multi-energy attenuation was explored by further dividing the 140keV energy range into three equal sections and sums and averages were taken of the attenuation in each material in each energy section, for thicknesses of 0 to 2cm. To visualise three-dimensional plots, the values of summed attenuation were read into a program, which showed organic materials, such as sucrose and water are very different to inorganic materials such as iron and aluminium when explored at these energies. By further dividing the energy 140keV range into 4, 5 and 6 energy ranges, multi-energy material discrimination could be explored. However attenuation curves cannot be created for more than three dimensions, so at the time of writing, various mathematical methods were being explored to analyse the data collected, including Principle Component Analysis. The results of the mathematical analysis of this should determine the ability for differentiation based on more than two energies used. 4

18 Figure 1-1: The summed attenuation in two regions for various materials over 2cm. Figure 1-1 shows the attenuation curves for common materials and that there is clear separation between metals and organic materials. The results of this study showed that material discrimination is possible based on the attenuation of X-rays through a material. With this introductory work completed, the next section of the project was to experiment with X-ray backscatter imaging and spectroscopy using newly acquired experimental equipment. 5

19 Chapter 2 : Theory This chapter contains relevant theory used over this research project. 2.1 Production and Properties of X and Gamma-rays X-rays X-rays were first discovered by Röntgen in 1895 [3], and applications for them has continued to increase since their ability to penetrate matter was identified. They are used in medical X-rays and Computerised Tomography (CT) scans, and industrial applications such as fault finding in structures. X-rays are part of the high energy section of the continuous electromagnetic spectrum comprising radio waves (wavelength (λ) ~ 10 8 m) through to X-rays (λ~10-10 m) and γ-rays (λ~10-15 m). Figure 2-1: The electromagnetic spectrum (left) [4] and a diagram of an X-ray tube (right) [5]. To produce X-rays, an electron beam can be produced by thermionic emission (heating off electrons) from a cathode. By applying a high potential difference between the anode and cathode, an electric field is produced (Equation 13) and the electrons are rapidly accelerated towards a target metal (commonly tungsten). In the metal, electrons are excited into higher energy levels which promptly decay, producing X-ray photons with energy directly proportional to the difference between energy levels. This process takes place under vacuum to remove the air in the path of the electrons which would otherwise cause a breakdown of the air inside the generator due to the high voltages applied (a 160kV potential produces 160keV photons). 6

20 X-rays are the result of atomic de-excitation, and are distinguished from γ-rays which are the result of nuclear de-excitation. There are two types of X-ray. Bremsstrahlung radiation is a range of energies caused by electrons scattering and changing velocity due to a nucleus, producing Bremsstrahlung photons. These are seen as a range of energies in a typical X-ray spectrum (Figure 2-2). Figure 2-2: A typical X-ray spectrum showing the characteristic peaks of the target material and the Bremsstrahlung spectrum which accounts for most of the energy [4]. Characteristic X-rays are material specific (in terms of energy) X-rays, generated from transitions in energy levels of the atom. By irradiating a material with high energy photons, enough energy can be applied to remove an electron in the target material from its orbit, creating a vacancy. Depending on which shell the electron falls from determines the energy of the resulting X-ray, and these can be detected and analysed. The levels have historically been assigned letters starting with K for n=1. Figure 2-3: The energy level transitions in an atom. 1. An electron decaying from the M to the L shell produces a L α X-ray. 2. An electron decaying from the M to the K shell produces a K β X-ray. 3. An electron decaying from the L to the K shell produces a K α X-ray. A branch of spectroscopy explores the peaks corresponding to these transitions which appear in the energy spectrum, which are specific to each element and can therefore be used to identify that element. 7

21 Radioactive Sources Radiation comes in many forms, from charged alpha and beta particles to radiation comprised of uncharged photons such as X-rays and γ-rays. In fact, the use of alpha particles helped physicists to formulate a model of the atom in the famous gold leaf experiment. X-rays can be artificially produced and switched on and off as required, making them a useful probing tool. Radioactivity from natural sources is a random process which has no such capability and is therefore stored in lead when not required. Radiation detection can be active or passive. Active imaging requires additional radiation to be supplied externally such as in transmission X-rays, whilst passive detection makes used of existing radiation being emitted, such as in thermal imaging, which makes use of radiation already emitted by a body. The decay constant (λ) in Equation 1 gives the probability of a decay per second, which is inversely proportional to the half life (t 1/2 ). Equation 1 ln 2 t 1/ 2 The activity (A) in Equation 2 is the number of decays per second in Becquerel (Bq) (where 1Bq is 1 decay per second). Activity decreases exponentially with time. Equation 2 A e A 0 t The half life (Equation 3) is a measure of the time it takes for half of a sample to decay, and this varies widely for different sources. For the sources to be used in the experiments, it is required to determine the change in source activity over the time of the project. The half life can be derived using the exponentially decaying number N 0 or activity A 0 as they are simply related by the decay constant, reducing to N 0 /2 or A 0 /2. Equation 3 t 1 2 ln 2 Photon Interactions in Matter A photon is defined as a quantum of electromagnetic radiation (Equation 4) with an energy E (where h is the Planck constant and υ is the frequency of the radiation [5]). There are three main photon 8

22 interactions in matter which are, in ascending order of energy, the Photoelectric effect, Compton scattering and Pair production. Equation 4 E hf hc The following interaction mechanisms have different probabilities of occurring (determined by the cross section (σ)) which relate to the atomic number (Z) of the detector material seen in Figure 2-4. The photoelectric effect and Compton scattering are most relevant to the energies explored. Figure 2-4: The main photon interaction mechanisms in matter varying with energy [6]. Figure 2-5: Possible photon interactions in a detector (left) [6] and the Compton scattering of an electron (right) [4]. 9

23 Photoelectric Effect (or Photoelectric Absorption) An incident photon surrenders its energy to eject an inner electron from an atom, and a photoelectron with kinetic energy T is produced from the energy of the photon (with energy hυ) once the energy required to eject the electron (the material s work function Ф) is gained (Equation 5). The vacancy from the ejected electron is filled from the deexcitation of a higher shell electron, with the energy released being an X-ray photon with energy equal to the difference between the two levels. Equation 5 h T The likelihood of an interaction via the photoelectric effect (P) occurring is approximated in Equation 6, where Z is the atomic number (to the power of between 4 and 5) of the material and E γ is the energy of the incident photon. An interaction via this method is more likely for higher Z materials. Equation 6 P Z E ~ Compton Scattering An incident photon collides with a stationary electron in a material and the photons scatters (now carrying a reduced energy E ' in Equation 7) with the remaining energy given to the electron, causing it to carry a momentum (Figure 2-5 right). The photons which interact in the detector by this method scatter at the different angles (θ) depositing only some of the energy. The probability of scatter is a function of the incident energy, depositing more energy with a larger angle of scatter. When only a portion of the photon energy stays inside the detector, energies less than the peak are recorded. The range of energies appears as the Compton continuum in Figure Equation 7 E ' 1 (1 E cos ) E m c 0 2 Pair Production An electron-positron pair is produced equally sharing the γ-ray energy in the presence of the nuclear electric field. The positron loses energy in scattering events and then produces two 511keV (almost) back-to-back photons following annihilation with a free electron. An initial photon energy of at least 1.022MeV is therefore required for pair production to occur and is unlikely to 10

24 be seen in these measurements. Equation 8 shows how the energy of the incident photon is distributed in pair production, where T and T are the kinetic energies of the electron and positron respectively, e e and 2m 0 c 2 is the rest mass energy of the electron and positron. Equation 8 E T e T e 2m c 0 2 When one annihilation photons leaves the detector, an escape peak appears 511keV less than the full energy peak, or 1.022MeV less than the full energy peak if both annihilation photons escape (Figure 2-22). Photon Attenuation Attenuation is the reduction in number of photons in a beam due to passing though a material and all materials (including air) attenuate photons. In Equation 9, I o is the initial number of photons, I is the number remaining after attenuation through thickness (x), of the material. Colour coding based on the attenuation detected is possible in some detector systems, separating metals into blue and organics into orange for example, to help with identification of materials. The linear attenuation coefficient, µ, is a measure of the attenuation in a medium, and it is constructed from the sum of the attenuations of the three main photon interactions (Equation 10), where µ varies with photon energy. I / I o usefully provides the ratio of photons through, regardless of the initial number of photons. Equation 9 I I 0 e x Equation 10 PE CS PP 11

25 2.2 Detection Technologies X-ray Transmission Imaging The technique of using X-rays to produce an image (a radiograph or just an X-ray) of an object by transmitting X-rays through it is well understood. In this standard transmission imaging, such as a bone X-ray, an X-ray generator is placed on one side of an object and a detector (commonly silver halide film [7]) is placed on the opposite side of the object. The X-ray photons pass through the object and due to attenuation of the photons in the beam, fewer pass through the object to reach the detector. Film darkens where a photon strikes it, so bones for example, which are more dense (higher µ) than the surrounding tissue will absorb more photons, so a film based detector on the other side will show bone as whiter as fewer photons have struck the detector. The attenuation of the X-ray photons is the main factor to determine the image obtained. The detection of these photons used in transmission imaging has advanced dramatically since photographic film, to include charge coupled devices (CCDs) and research is constantly conducted into finding new detector materials. Figure 2-6: A typical X-ray transmission imaging arrangement. Medical Imaging Common medical imaging techniques include Positron Emission Tomography (PET) and Single Photon Emission Computed Tomography (SPECT). In PET, a natural radionuclide is induced into the body (such as carbon oxygen or nitrogen depending on the region required to be imaged) which decay by emitting a positron. The positron travels a short distance (millimetres) losing energy within the body. When enough energy has been lost in scattering events, the positron annihilates with a free electron creating a pair of 511keV γ-ray photons approximately back-to-back, which are detected from 12

26 outside the body. Detecting the concentration of these photons allows an image to be constructed to identify abnormalities. In SPECT imaging, the collection of 140keV γ-ray photons emitted from inside the body is another method to non-invasively image the internal body structure. Molybdenum decays by β - emission into technetium (metastable or long-lived state), which then decays into technetium and a single 140keV γ- ray photon (Equation 11). Choosing a pharmaceutical which is taken up by a specific organ allows imaging of that region when the emitter is added to the solution. These photons can be detected after collimation by scintillator crystals and a bank of photomultiplier tubes [3]. Equation Mo 99 Tc m 99 Tc + γ Figure 2-7: An image of a brain tumour identified by 99 Tc m imaging [3]. New sensors for the detection of these photons could make lower power, light and portable radiation detectors a possibility, whilst retaining or improving the spectroscopic and imaging capability. Backscatter X-ray Imaging Compton scattered photons, cause a wide angle of scattered radiation, (determined from Equation 7) can also be used to provide X-ray images. Backscatter (or single-sided) X-ray can be used when there is only easy access to one side of an object, if for example the contents of a case are unknown and it therefore may not be safe to move the item to view it in a traditional transmission X-ray system. This can provide more information due to the increased photon flux. In this arrangement, a detector is placed on the same side as the X-ray source, which records X-rays that have returned from the object. The amount of energy from a backscattered X-ray is significantly less than the transmitted energy due 13

27 to the scattering. Higher Z materials will absorb more of the X-rays (Equation 6) whilst lower Z materials will scatter more. The pinhole effect is a well understood method for imaging light rays, commonly by exposing film. The recorded image will be a reflection of the true one as seen in Figure 2-8. When beams of light are passed through smaller apertures, sharper images are obtained but will take more time to form, as a lower photon flux is incident on the film. Larger aperture leads to blurred image Figure 2-8: Pinhole imaging providing sharper images for a smaller entrance aperture [8]. When the diameter of the pinhole aperture is of the order of the wavelength of the radiation, diffraction occurs, which is the spreading of waves from a source. Maximum diffraction occurs at the point when the size of the aperture is of the same order as the wavelength of the radiation (light). The wavelengths being used in this application are 7.8x10-12 m (Equation 4) and therefore it is determined that diffraction will not need to be compensated for, when using a 2mm pinhole. 2.3 Radiation Detection Systems Cadmium Zinc Telluride (CZT) Detectors Radiation detectors are broadly divided into two categories; charge-based and light-based. The proportional counter, Geiger-Muller tube and semiconductor detectors are charge-based detectors working on the principle of electron and holes charges; whilst light-based detectors such as a photomultiplier tubes (PMTs) or photodiodes operate by detecting and amplifying optical photons produced by light-emitting (scintillator) materials. CZT is a semiconductor based radiation detector which has gained a lot of attention over the past few years due to its room temperature (without the need for liquid nitrogen cooling to reduce the thermally created charges) detection of X- and γ-ray 14

28 photons. CZT is made from materials with high atomic numbers (Z) of 48, 30 and 52 respectively, which have the ability to attenuate photons better than lower Z materials, and therefore detect an interaction. Other materials which are used as direct semiconductor radiation detectors include silicon (Z=14, suited for low energy photons) and germanium (Z=32, requires cooling) and research into new materials for detecting radiation is constantly ongoing. CZT is more formally written as Cd 1-x Zn x Te where x is the blending fraction of zinc telluride in cadmium [6]. A higher zinc concentration increases the band gap of the material, allowing fewer thermally generated electrons to be detected as noise. CZT resistivity values between 2.5x10 10 Ωcm (4% zinc concentration) and 1.5x10 11 Ωcm (20% zinc concentration) are expected for CZT depending on the blending fraction of zinc in the sample [9]. The value of resistivity is an important feature in the detection of radiation photons in the material, as a higher resistivity implies a lower leakage current (as resistivity is inversely proportional to current) providing a better quality material. Incident radiation such as those from radioactive sources producing X- or γ-rays will deposit energy creating electron-hole pairs (ehp) proportionally to the amount of incident energy (Figure 2-9). The ehp pair creation energy (W) for a typical Cd 0.9 Zn 0.1 Te detector 4.64eV [10]. Figure 2-9: The operation of a semiconductor (CZT) radiation detector [6]. The capacitance (C) of the detector system is generally a constant, and includes the capacitance of the detector and all leads (which should therefore be kept short). The voltage (V) detected is then directly proportional to the charges produced by the radiation (Equation 12). 15

29 Q C Equation 12 V (V ) Inside a semiconductor detector, an applied voltage (V) between the two electrodes (at thickness d) creates an electric field (Equation 13) often 10 5 Vm -1, causing the charges created by radiation to drift with a velocity υ. The electrons migrate to the positive electrode (anode) and the holes to the negative electrode (cathode). The drift velocity (for electrons υ e and holes υ h in Equation 14) is proportional to the mobility, μ, of the electrons and holes (how easily they can move in the material), and the strength of the electric field. The charges created in the detector move between 10 6 and 10 7 ms -1 to the collection points at the electrodes of the detector material. Equation 13 illustrates how a smaller distance between the anode and cathode will linearly increase the electric field. CZT and SPM detectors can be made very thin providing a very large electric field (10 5 Vm -1 ). However, thinner detectors will be less able to attenuate radiation photons causing a lower detection efficiency. Figure 2-10 shows how the increase in the electric field causes the drift velocity to increase in silicon (in semiconductor detectors the electric field is approximately 10 3 Vcm -1 ). Equation 13 V E (Vm -1 ) d Equation 14 e e E and h E h (ms -1 ) Figure 2-10: The drift velocity (V d ) as a function of applied electric field [6]. 16

30 From Ohm s Law (Equation 15) for a given voltage (V) and current (I) the resistance (R) of a detector material can be found. By measuring the current at different voltages, a plot of current vs. voltage provides resistance from the reciprocal of the gradient. Equation 15 V R (Ω) I Resistivity ( ) is a measure of a material s ability to oppose the flow of an electric current and increases with resistance (R) and cross sectional area, A [5]. It is commonly expressed in units of Ωcm for CZT devices. An IV graph can be used to determine the sample s resistivity, where m in Equation 16 is the gradient of an IV graph is the reciprocal of resistance. Equation 16 RA L A ml (Ωm) Intensified Charge Coupled Device Detectors [11] describes the operation of Intensified Charge Coupled Device (ICCD) detectors. Figure 2-11 shows incident photons striking a photocathode, producing photoelectrons. A micro channel plate (MCP) receives these photoelectrons and through the electric field, many more secondary electrons are produced. These strike a fluorescent screen (such as phosphorus), producing optical light flashes which are usually transferred by fibre optic cable. A CCD receives these photons which are converted to charges, which are moved and read out row-by-row in the readout section (Figure 2-12). 17

31 Figure 2-11: The stages in an ICCD detector [11]. Figure 2-12: The readout in a charged coupled device [6]. This device is used for imaging only, as only counts and no energy information is recorded. Having a pixellated input phosphor allows for imaging, as variations in intensity over the pixels can be displayed graphically as an image. Scintillator Crystal Properties A scintillator is a light-based radiation detector which emits photons due to excitation from the incident radiation, and qualities such as high linear light yield (the number of photons created per unit of energy deposited (usually given in MeV)) and a short decay time are desired, such that fast and efficient detection is possible. Scintillators, such as cesium and sodium iodide, are commonly coupled to a 18

32 Photomultiplier tube (PMT) increasing a low level of light photons into a much higher one. The PMT turns the light produced from the scintillator into electrons in the photocathode by the photoelectric effect, and the efficiency of this process is determined by the quantum efficiency of the photocathode. There is a rapid increase the number of electrons using stages of dynodes [3]. According to [6], the charge creation energy is 3.6eV for silicon and around times that for scintillators, which only convert 5-10% of energy to light. Scintillators are therefore intrinsically less efficient than semiconductor detectors at stopping and detecting radiation photons. Radiation damage to scintillators and other detectors is possible, as the lattice can be altered from the regular periodic structure, compromising the ability to effectively detect and respond to radiation events [6]. Many materials scintillate in response to radiation. Scintillator materials are sub-divided into organic (liquids and plastics) and inorganic materials, the latter being used for the project. Using Equation 4, the energy of the optical photons from four scintillator crystals was determined. Scintillator Light Yield / MeV Decay Time (μs) Peak λ (nm) Photon Energy (ev) BGO 9, CdWO 4 13, LYSO 32, CsI(Tl) 52, Table 2-1: Scintillator crystals and their key properties [12 (LYSO details from 13)]. Based on these detector properties, these four scintillators were procured to provide a range of fast and slow decay times, and high and low light yields to fully test the range of the SPM. Hygroscopic crystals were not chosen to avoid degradation in crystal performance over the period of experiments. Figure 2-13: The emission of scintillation crystals and the response of PMT devices [6]. 19

33 Figure 2-13 shows the different responses of PMTs and scintillator crystals (Figure 2-18 shows the SPM detector response). The best response of a detector would be where the peak detector response matches the peak crystal emission, however, this is rarely the case for many crystal-detector combinations due to the range of scintillator wavelengths. The emission spectrum of light emission from scintillator crystals and the spectrum that detectors are sensitive to, allows detection of light from scintillator crystals using detectors which are not perfectly matched (albeit with lower sensitivity). Lutetium Yttrium Silicon Oxyorthosilicate (LYSO) is a scintillator crystal which was chosen for its very fast decay time and high light yield. However, Lutetium is itself radioactive decaying by β - emission, causing a level of background radiation in all measurements [14]. Using values from [14] and scaling these values for the scintillator size used in experiments, the intrinsic activity of the scintillator due to lutetium is about 13.5kBq, whilst the sources used at the energy of interest are much more active at approximately 370kBq. Activation of a scintillator using Thallium is common for sodium and cesium iodide which adds sites in the scintillator that can produce optical photons in the forbidden band [6]. This increases the probability that de-excitation of an optical state due to radiation will lead to optical photons (Figure 2-14). Having a fast, bright scintillator crystal which has a high attenuation of radiation photons over a large energy range (detection efficiency) with a peak emission wavelength of the detector used, is the ideal scintillator crystal. As this doesn't exit, a compromise of detection properties is usually made. Figure 2-14: The scintillation process from activated states [6]. The light yield of a scintillator material arises as more charges are successfully created from the incident radiation. By having a linear light output the amount of incident radiation can be determined to deduce the energy of the incident radiation. The rise time varies in a scintillator crystal is related to 20

34 the mobility of the charges in the material [15, 16] where more prompt decay times are due to a better charge mobility in the material. [16] describes the process of fluorescence in a scintillator crystals. The absorption of radiation causes electron-hole pairs to be created in the crystal at between two and seven times the band gap energy. Electrons scatter to reduce in energy until they can excite luminescent centres which produce optical photons. Due to the incident radiation, the scintillator emits photons in the visible part of the electromagnetic spectrum (~ nm). The light produced from a scintillator can be recorded by a photomultiplier, a photodiode, an avalanche photodiode and a new detector technology, the SPM. The Silicon Photomultiplier (SPM) The high gain, low noise and availability in large sizes are some of the reasons making the PMT system a popular detector for scintillation light. There are however, several disadvantages to the technology; fragility of the glass, large power requirements and sensitivity to magnetic fields [17]. For portable radiation detectors, the large sizes of the PMTs can be a disadvantage. Scintillator Photocathode Dynodes Photomultiplier tube Anode Figure 2-15: A photomultiplier tube with the amplification stages in the dynodes [3]. 21

35 Photodiodes have been used as an alternative to PMTs to detect optical photons from scintillator crystals, which enter the photodiode material and create electron-hole pairs. Photodiode gain is not as large as the PMT, so a better alternative to the PMT is the avalanche photodiode (APD) which can provide a much larger gain (several hundred is possible [6]) than a standard photodiode. A rapid increase in the number of electrons is created due to further collisions and the multiplicative process occurs due to the presence of a large electric field [6]. The gain is still a limiting factor and alternative methods to readout from scintillator crystals are being explored. Figure 2-16: A single pixel SPM (on a square base approximately 3x3cm) [20]. SPMs (Figure 2-16) combine the features of PMTs and APDs providing a high gain (~10 6 ) whilst requiring a low operating voltage [18] making portability a possibility. Geiger mode Avalanche Photodiodes are (GAPDs) are operated with a bias above the breakdown voltage such that only one carrier is required for breakdown. This breakdown is stopped or quenched by a large resistance in series with the GAPDs. By connecting the output of thousands of GAPDs together in parallel, a photon flux causes a current which is directly proportional to the number of incident photons [17]. Each GAPD is a microcell, any many are tightly packed such that there are thousands in a single pixel. 22

36 Figure 2-17: A 1mm 2 SPM pixel with many microcells on the top (left) [17] and the GAPDs in the pixel are connected together to provide a photon proportional output [17] (right). When coupled to a scintillator crystal, the SPM detects the optical photons produced by radiation incident on the scintillator. SPMs have intrinsic characteristics causing the loss of photons, including the active area (the area able to detect a photon interaction), photon detection efficiency (PDE) and fill factor (the ratio of the active area to the total area) which reduces the possible energy resolution as fewer photons are present. The PDE (Equation 17) is the product of quantum efficiency (QE), the probability an avalanche breakdown occurs (PAB), and the fill factor (FF). According to [19], the PDE is the probability that an incident photon produces a Geiger pulse from one of the microcells, and encompasses the probability of the photon initialising the avalanche breakdown and the quantum efficiency of the detector. Figure 2-18: A graph containing the PDEs for 4V over bias for various SPM products [19]. Equation 17 PDE QE*PAB* FF [22] 23

37 With the number of photons remaining after the losses caused by the scintillator and SPM system, the energy resolution (R) can be determined using Equation 18 and Equation 19. Equation 18 R FWHM No. photonsavailable Equation 19 R 2.35* No. photonsavailable No. photonsavailable It is known [3, 6] that the energy resolution improves (by reducing) with increasing energy (E) provided by Equation 20, where K is a constant of proportionality. Equation 20 R K E By taking logarithms of both sides, Equation 21 is produced, which should produce a straight line and finding the constant K allows the statistical broadening of the peak to be found [6]. 0.5 Equation 21 ln( R) ln(k) -ln(e ) A pre-amplification ( preamp ) board is connected to the SPM which increases the signal to hundreds of mv, with the brightest scintillators predicted to produce (negative) SPM pulses of the order of volts out. [22] states that the dead time of the detector to be 0.1µs, leading to a maximum count rate of 10MHz. Equation 22 shows how SPM gain is linear with the bias voltage (V o ) after breakdown voltage (V br ) (the electron charge (e) and the capacitance of the system (C) are constants). [22] also shows how the dynamic range (the maximum number of simultaneous photons which can be detected is limited to the number of microcells) increases with the number of microcells. As the microcell size reduces, the gain decreases, so there is an important trade-off between the gain and microcell size. Equation 22 G C( V V ) o br e 24

38 Two of the many possible areas where the SPMs have been proposed for use includes; [17] Medical: PET scanning is dependant on fast coincidence timing which the SPM should be capable of; Portable security applications: such as portable radiation detectors due to the small size and low power requirements. Silicon has a low Z (14) which has a lower stopping power for X- and γ-ray photons than CZT (Figure 2-19). However, in SPMs, radiation detection is indirect, as optical photons come from the scintillator crystals, not directly from the radiation photons. Scintillator materials have a good stopping ability of low energy photons (especially 0-100keV). The noise in an SPM takes the form of dark photons, which are thermally generated (there is the option of cooling the SPM to reduce these). [33] shows that the dark rate is 9 times higher in the 3mm SPM than the 1mm SPM and therefore the individual dark photon pulses cannot be as clearly distinguished. It also states that the 1mm SPM has a faster onset and decay time than the 3mm SPM. [23] describes how optical cross talk causes distinct levels of noise present. This is where a photon is recorded across two or three microcells, with the probability decreasing with the increased number of microcells. The onset time and dark photons were investigated in the project. Figure 2-19: The attenuation ratio I/I 0 for the three detector materials explored for the project. Using attenuation data collected from XCOM, Figure 2-19 shows how CZT attenuates far more photons over the energy range and should therefore have the better detection efficiency of all the detectors explored. (An I/I o value of 1 indicates that all of the photons pass through the material 25

39 unattenuated). Silicon as a direct radiation detector is seen to be very inefficient at energies greater than 100keV. When comparing the efficiency of the CZT to the scintillator materials (in the thicknesses used for the detector experiments) it is seen that the detection efficiency of CZT is comparable to the scintillator crystals. Figure 2-20: The attenuation ratio over the energy range to be tested for the four scintillator crystals used at 3mm thickness and the CZT detector at 5mm for direct comparison. 2.4 Radiation Measurement and Spectroscopy What Happens to the Detected Radiation? Imaging is possible when more than one pixel is present, so differences in the radiation detected over the pixels can be seen. At the detector, the radiation interaction of the photon in the material is analysed by two common methods; Photon counting is used for low numbers of photons where a background level is set and all counts above this are amplified and recorded. Charge integration is used where there is a much higher flux of photons. The number of counts incident on each pixel can be found with knowledge of the time each pixel was integrating for, and the current for that integration time per pixel. The electron-hole pair creation energy (W) for the detector material is known and therefore the number of counts per 26

40 pixel can be found. With this information, an image can be built up from the amount of counts incident on each pixel in the detector by assigning a greyscale which typically lightens with a higher number of counts in a pixel. The charge deposited is found by integrating the current pulses produced in a given time according to Equation 23. Equation 23 Q t 0 Idt Pulse Height Spectra The energy spectrum from the radiation is typically displayed using a Multi Channel Analyser (MCA). Here the channels (which are related to the energy of the radiation according to Equation 24) are displayed on the x axis, and the number of counts at this channel on the y axis. In this way, a distribution of counts over the number of available channels is produced, resulting in a pulse height spectrum or simply spectrum. The location of the peak determines the its energy, which can be used to identify the material from the energy value. For example, a peak at 59.5keV would show that 241 Am is present. Figure 2-21: Operation of an MCA, an extension of many single channel analysers (SCAs) [6]. A pulse height spectrum (Figure 2-22) shows features such as the Compton edge, the peak centroid (or full energy peak where the counts corresponding to the main γ-energy are), and escape peaks. 27

41 Figure 2-22: Typical pulse height spectra from γ-sources [6]. Detector Calibration When spectroscopic data is collected, the x axis is usually a channel which is not directly equivalent to an energy. The y axis displays the number of counts received in each channel. By irradiating the detector with calibration sources (with peaks at known γ-energies) the channels corresponding to the peaks can be identified with energies according to Equation 24. Equation 24 Energy = Calibration constant * Channel Number At least two peaks are required so an accurate calibration constant for the detector can be calculated. More points will provide a more accurate peak value when an unknown source is used. How is Spectroscopic Capability Determined? Energy resolution is the ability to distinguish between two close energy peaks seen in the energy spectrum as separate. This is a very important basis for comparison between spectroscopic detectors, and is found using Equation 27. Spectroscopy is possible when the energy resolution, commonly expressed as a percentage at a given energy, is low enough so the separation of close energy peaks can be identified. The following three equations show how the energy resolution is calculated to identify the possibility for spectroscopy from the counts in the energy peak, where σ is the standard deviation (and according to Poisson statistics is also the error) and FWHM is the Full Width at Half Maximum. 28

42 Equation 25 No. of counts Equation 26 FWHM 2.35* Equation 27 Energy Resolution FWHM Centroid 2.35 Centroid Figure 2-23: An improved energy resolution is obtained with a thinner peak [6]. Ideally, a single line would be produced corresponding to the photon's γ-ray energy, and this is approached with some very high resolution detectors, however, according to [6], the peak broadens (Figure 2-23) due to statistical fluctuations in how the pulse is recorded, as there are inherent fluctuations in the number carriers produced each time. For detectors with the same efficiency, the better energy resolution will be taller (the efficiency is not dependant on the energy resolution). As more charge carriers are produced, the error becomes less significant, so at higher energies, the energy resolution is improved (the peak is less broad) as, for an equivalent number of photons, more charge carriers are produced from higher energy photons, reducing the peak broadening. The radiation peak produced is assumed to follow a Gaussian distribution (Equation 28), where the parameters for A, B, and C correspond to the peak height, peak centroid and standard deviation respectively. Equation 28 y Aexp 1 2 ( x B) C 2 29

43 The area of Gaussian peaks (Equation 29) provides the number of counts in them (to be used with Equation 28) to find the detector s efficiency. Equation 29 Area A 2π Other properties of a radiation detector determine its effectiveness: Intrinsic efficiency of a detector (ε int ) (Equation 30) is a measure of how many photon interactions the detector has recorded from the number of photons which can interact with the detector, based on the area the detector occupies. Equation 30 int No. pulses recorded No. pulses incident on detector Figure 2-24 shows how the radioactive sources radiate in a sphere, where a detector usually occupies a small area of this. The further back the source is from the detector, the smaller the area of the sphere incident on the detector, where fewer counts are expected to reach the detector. Figure 2-24: The source emitting in 4π, where only a proportion of the activity is incident on the detector at some distance (d) away. The sources used are at different activities, and as for most sources, the γ-decays of interest for spectroscopy (59.5, 122, 511 and 662 kev) do not account for the whole source activity (% determined from the decay scheme). The activity (and therefore the number of counts received by a detector) reduces as the source is moved further away from a detector according to the inverse square law as 30

44 seen in Equation 31, where C is the number of counts, K is a constant of proportionality for the detector and d is the source to detector distance. Equation 31 C K 2 d Once a number of counts recorded at a given distance is known, it is possible to determine how many counts are expected at another source to detector distance by determining K. By using this method for several points, the change in counts with distance to the source can be identified. Saturation is the point at which no new counts can be detected due to a maximum number of counts a detector can record being reached. It is an important quantity to know so the sources which are used are kept below this level. Dead time is the time that the detector cannot detect radiation (or count) due to the previous counts being recorded and processed. This is equal to the difference between live time and real time where the live time is the time the detector has actually been counting for, whilst the real time is the time elapsed since the starting the experiment. Reducing the number of counts (placing the source further from the detector) will be reduce the dead time. Uniformity (Equation 32) is a measure of how similar the recorded radiation is over all the pixels in a detector for the same incident radiation. σ is the standard deviation and centroid is the peak energy. Equation 32 Uniformity σ Centroid Gain is the amplification of a signal to utilise the full dynamic range of the detector. By increasing the gain, the noise is also amplified, and therefore the level of the gain needs to be carefully checked to ensure that any benefit in signal is not at the expense of more noise. The signal to noise ratio (SNR) was determined for the SPM detector using Equation 33 at two different bias voltages. Equation 33 SNR PulseSize Noise Size 31

45 Error Analysis The standard error in the mean (s m ) (Equation 34) was commonly used to provide an uncertainty estimate where multiple measurements were taken, where σ n-1 is the standard deviation and n is the number of readings or samples. Equation 34 s m n 1 n Various techniques of error analysis were performed over the course of the project in order to give results with an estimate of the associated error. Applying Poisson statistics allows the error in a number of counts to be recorded simply as the square root of the value as seen in Equation 35, therefore at larger numbers of counts the statistical error in the count rate is lower. Equation 35 Error No. counts Background radiation is a random factor adding to the uncertainty. It is naturally occurring, however, in a radiation laboratory, levels of background are likely to be higher and therefore needed to be accounted for by talking measurements without the additional radiation used in the experiments. [6] describes the propagation of errors for a function (u) dependant on three variables (x, y and z). Equation 36 2 u u x 2 2 x u y 2 2 y u z 2 2 z When appropriate, a confidence level was applied to a measured value, the most common being 95%. The coverage factor (k) is used to give the confidence in a measurement and commonly the intervals used are k=2 (95% confidence) and k=3 (99% confidence) [24]. There are an almost infinite number of possible sources of uncertainty when taking a measurement including the temperature, humidity, the time of the reading, the ability of the measurer, ambient light, presence of drafts, fields and the list goes on. Therefore any uncertainty applied to a value is only an estimate of the error, and it is almost impossible to get a perfect measure of the uncertainty. 32

46 Chapter 3 : Experiments and Work Conducted Many programs, models and practical experiments were carried out to meet to aims of quantifying spectroscopic and imaging performance of different radiation detector systems. 3.1 Calculations and Modelling Radioactive Source Calculator A spreadsheet was created to calculate the current activity of the radioactive sources used for the experiments based on a known activity at some point in the past. By entering the half life of the source, and the activity at a known date, the current activity was calculated. Using Equation 1 for each source, the activity at any time was determined. By extrapolating the date over time from the current activity, the future activity was calculated by Equation 2. This was a useful spreadsheet to create, which was used several times over the project to determine which and when sources need replacing, and the effect of diminishing activity over the measurement period was determined. CZT Shielding Requirements The thickness of shielding required for the CZT detector housing needed to be calculated so the detector could be used as a pinhole imager with a pinhole (at 10cm) away which can be added and removed as required. All of the X-ray photons other than those from the pinhole needed to be attenuated. The thickness of lead to do this was calculated using Equation 9. With values of total attenuation with coherent scattering, from XCOM, which varies with different materials and energies. A spreadsheet was produced calculating the material thicknesses required (based on the absorption of different materials) to reduce the photon flux to a desired amount. 5mm of lead would sufficiently attenuate the X-rays, however, to reduce the fluorescence lines from the lead, the thickness of tin was found, and finally the thickness of copper was found to reduce the fluorescence lines from the tin. The exponentially decreasing intensity (I) was calculated as the material thickness (x) increases, reducing the initial number of photons (I 0 ) due to the type of material defined by the absorption coefficient (µ). 33

47 3.2 CZT Experiments The CZT Detector Figure 3-1: The CZT detector array. The CZT detector array (Figure 3-1) has a pixel size of 1.5mm on a 1.6mm pitch which is 5mm thick. An electric field is produced with a 500V bias, resulting in an electric field of 10 5 Vm -1. With electron mobility (µ e ) in CZT of 1000 cm 2 /Vs [25] this leads to a υ e of 10 6 ms -1. According to the CZT detector manufacturer, the detector can record at 100kcps, which corresponds to 32 cps / pixel, however, the system could equally handle all 100kcps in one pixel if all the others had zero counts [26]. Uniformity and Energy Resolution The CZT detector was irradiated with 241 Am and 57 Co γ-sources to determine the uniformity and energy resolution of the detector. The purpose of these experiments was to identify any improvement (or deterioration) in the CZT detector performance in terms of energy resolution and uniformity, after an upgrade, to determine whether to fully populate the detector with the upgraded material. 34

48 Figure 3-2: The experimental setup for the uniformity and energy resolution measurements. The sources were in turn placed 10cm from the centre of the detector, integrations of the flood illumination were taken over 10 and 30 minutes for each source as seen in Figure 3-2. The detector records energy spectrum information per pixel. The analysis required code to be written in IDL (a programming language similar to FORTRAN 90) to obtain a greyscale image of the pixel intensity. By producing histograms (such as Figure 3-3) from the data producing this image, an inbuilt Gaussian fit function was applied to provide values for standard deviation (C) and centroid (B). Using Equation 32, the uniformity was calculated for both sources, to provide the spread of this data. Binning is a useful tool, especially in with low count levels, which allows the available number of channels to be combined such that a count with a small accepted range of energies is added to only one channel. Figure 3-3: An image (based on detector area ~12.5cm 2 ) from a γ-source illumination (left) and a histogram from which the detector uniformity can be calculated (right). 35

49 Code was then written to correct for the artefacts and discontinuities at the CZT module boundaries by flat fielding which uses the data taken over a large time integration (30 minutes which is much longer than data would be collected for) to provide a data set high statistical quality data. By programming an iterative loop to calculate the average number of counts received per pixel for the 30 minute data (performed for each source in turn), a factor was found giving the difference from the average number of counts per pixel. This factor was multiplied to each pixel for any other image captured with the detector (this was initially applied to the 10 minute source flood illumination data to produce flat fielded images). A flat (uniform number of counts over the image) image should then be produced. Flat fields (using both calibration sources) were available to correct for other artefacts in images collected by the detector. Importantly, using the flood illumination data, the energy resolution was found per pixel (for each source separately), using Equation 27. The data was read out into a spreadsheet, where statistical analysis on the energy resolution could take place (principally the average energy resolution per pixel, standard deviation and error) to determine the spread of the data over the pixels. The detector was then returned to its manufacturer to have all 25 modules (5x5 array with 6400 pixels) of CZT replaced with 12 modules (3x4 array with 3072 pixels) of higher quality CZT (the exact details are unknown but it is assumed better leakage and uniformity across the material). Following the detector hardware upgrade, the CZT detector was again illuminated using the previous method with the same calibration sources in turn at 10cm from the centre of the face of the detector. Again, 10 and 30 minute integrations were taken to check the functionality of the flat field program. All of the previous IDL programs were modified to account for the change in size of the array. It was hoped that the CZT detector will be an improvement to the imaging quality of the ICCD detector, and as spectroscopic information can be obtained from the detector (the CZT records the charge and the energy per pixel) a much more compact and useful system will then be available. 36

50 Summed counts CZT Spectroscopy The spectroscopic capability of the detector was tested in experiments with the pinhole cover and housing removed (to use the full area) using the 241 Am and 57 Co sources both individually and separately. The time taken and the type of γ-spectrum which can be obtained from the CZT detector was determined. Using the 241 Am and 57 Co sources at 10cm and 12cm from the centre of the detector respectively, the simultaneous energy spectrum (and counts recorded by the detector at each energy) were recorded. The sources were then aligned at 15cm from the detector and were incrementally moved further from the detector (every 5cm from 15cm to 30cm) and the energy spectrum was recorded in increasing time increments from 1 to 10 seconds at each distance. This experiment was designed to quantify the maximum number of counts received by the detector over the energy range to identify if the detector counts linearly over time and that the inverse square law is followed. CZT Efficiency Measurements 241 Am and 57 Co sources were placed separately at three distances (0.5, 1 and 1.5m) from the CZT detector, without the housing and pinhole, and 10 minute integrations for each source at each distance were taken. To determine the intrinsic efficiency (Equation 30), the number of counts in the energy peak is required, which for the CZT detector was determined by calculation (Equation 29) using the height and standard deviation values from the fitted Gaussian peak. (The energy peak is assumed to be a Gaussian, where the fit is a good match to the data as seen in Figure 3-4). Energy (kev) Figure 3-4: The Gaussian fit to the 241 Am peak (at 0.5m for 10 minutes). 37

51 From the activity incident on the detector (based on the distance to the detector and source), the portion of the sphere occupied by the detector created by the source, and the activity at the energy peak of interest (59.5 and 122keV for the CZT detector) the detection efficiency was determined. The ICCD Detector The Intensified Charge Coupled Device (ICCD) detector has a thin (2.5mm thick) pixellated CsI(Tl) crystal array converting X- and γ-ray photons into optical photons. Figure 3-5: The ICCD detector (with a 2mm pinhole plate attached). Removing the cover and pinhole plate exposes the front of the detector, which allowed the detector area to be flooded with the same two calibration sources used for the CZT detector, in the same experimental arrangement. Software used with the detector allows use of the previously acquired data for flat field corrections. For a fair comparison to the CZT for the detector uniformity, flat field measurements were made using the ICCD detector. CZT and ICCD X-ray Backscatter Pinhole Imaging The use of the CZT detector as an imager using a newly designed graded shielding was tested and compared to the images obtained using an ICCD detector in the same experimental configuration. 38

52 Figure 3-6: The housing applied to the CZT detector providing a pinhole aperture at 10cm from the centre of the detector. CZT detector (with pinhole) Target material X-ray generator Figure 3-7: Top view of the arrangement for X-ray backscatter pinhole imaging. By facing an X-ray generator towards a target material and irradiating it with 160keVp (peak energy of 160keV) X-rays for 10 seconds, the CZT detector (with the 2mm pinhole) received backscattered photons from the target and surroundings. Figure 3-8: An image created using a pinhole receiving backscattered X-rays (left) and the effect chamfering has on the pinhole (right-above before and right-below after chamfering). 39

53 Figure 3-8 (right) shows that the area covered by the pinhole (exaggerated) which was not initially the entire active area of the detector (in Figure 3-8 (left) black indicates where there are fewer photon interactions due to the collimation effect). It was therefore necessary to chamfer behind the pinhole by calculating the angle required to allow the whole detector area to be used. With the chamfering successful, a background measurement with the pinhole covered and backscattered X-rays being incident on the detector was made. This found that X-rays were leaking in to the detector, causing the image shown in Figure 3-9. Figure 3-9: The leakage obtained before the additional shielding (detector area ~7.9x10-3 m 2 ). Before any backscatter images were taken, this leakage needed to be eliminated to avoid the additional effect on the measurements. The best method for additionally shielding the detector was established by applying lead shielding all around the detector (including the pinhole) and sequentially removing the lead in sections to identify where the leak was coming from, by measuring the average number of counts per second in the imaging section of the CZT software for short (10-60 second) integrations. It was found that the optimum arrangement (using the least additional lead to produce a uniform output) of lead shielding requires blocks between the X-ray generator and the side of the CZT detector, and additional blocks at the back of the detector to reduce X-rays getting in this way, as seen in Figure

54 Figure 3-10: The optimum shielding configuration (left) and the more uniformly distributed counts with this shielding. After warming up the X-ray generator (required to recondition the vacuum after a period of inactivity), various objects were illuminated with 160keVp X-rays from a continuous X-ray generator for 10 seconds. The detector recorded all the time X-rays were being produced. The objects were placed 55cm from X-ray generator and 30cm from the pinhole aperture of the detector. These distances and times were determined after looking at several settings and integration times, so the output images were centred, a good size in proportion to the background and not saturated. The same experimental arrangement (integration times, pinhole, distances and energies) was setup for the ICCD detector and images were taken using the 2mm pinhole, once removed from the CZT detector housing. Images were taken of several objects with different atomic numbers (Z) and densities (including salt, sugar, cotton, talc powder, aluminium powder and tungsten) which were on their own, and then in a typical luggage case. To elevate the objects that were not in a suitcase (for differentiation between the floor and the object and to reduce scatter from the floor), the objects were placed on a plastic block 5cm high when out of case measurements were made for a direct comparison between the images. The ICCD images had a best fit function for the brightness and contrast applied using Image Pro Plus software. The CZT data was processed with IDL code to display the images. The flat field corrections acquired prior to these measurements for both the CZT and ICCD detectors, (based on the 30 minute 57 Co illumination) were used to clean-up the images by removing the artefacts caused by the detector. 41

55 CZT and ICCD Angular Resolution This experiment was conducted to provide another basis of comparison between the ICCD and CZT detectors. A tube of salt and a tube of sugar (arbitrary objects 7.5cm long (up to the lid) with a diameter of 2.5cm), were placed at 40cm from the both detectors, and the X-ray generator was placed at 60cm from the objects to be imaged. Additional shielding was again added around the CZT detector. The 2mm pinhole was used for both detectors, and as there is only one pinhole, the experiments were run first for the CZT and then for the ICCD using the same experimental setup. X-ray Generator ICCD Detector Objects CZT Detector Figure 3-11: The experimental arrangement for separation experiment (CZT pinhole covered for a background measurement). The salt tube was placed on one circle and was fixed there, and the sugar was placed in decreasing distances in increments of 1cm closer to the salt tube. At distances of 1 and 2cm from the tube of salt, the sugar was placed further back from the salt (Figure 3-12) else both tubes would have overlapped. 1 2 Fixed salt position Figure 3-12: The positions of the tubes of salt and sugar, moving closer together in 1cm steps. 42

56 Backscatter X-ray images were taken for 10 seconds using 160keVp X-rays. The images from the CZT detector were then processed using IDL code to flat field each image using the 30 minute 57 Co data. The ICCD images were also flat fielded using a pre-existing function in the image processing software, also based on the 30 minute 57 Co illumination for the ICCD detector. Using IDL code for each flat fielded image from each detector, each image was modified so a line of data (or a profile), of counts (CZT) and intensity (ICCD) was stored to a file, recording the change through the objects and background. (This line was identified by setting the pixel values along a line to an arbitrary number, which was returned to its original value when the data was read out). Figure 3-13: The line of data extracted for ICCD images (left) and CZT images (right). This data was normalised for each image, and plots of the intensity with pixel value were made separately for each image, combined for individual detectors and combined for both detectors to numerically quantify how the images vary for each detector records over each image. 3.3 SPM Experiments Scintillator and SPM Energy Resolution Model A spreadsheet model was created to estimate the number of photons reaching each SPM and then being lost, to find the energy resolutions possible using Equation 18 and Equation 19 over the energy range. Starting from an initial energy deposited to each scintillator through the four calibration sources, this study was performed to indicate if spectroscopy is theoretically possible from the energy resolutions 43

57 obtained. For example, an energy resolution of 20% at 662keV would allow another peak to be resolved below 530 and above 795keV. According to [28] values for the scintillator efficiency are between 3 and 15%. The justification for using scintillator efficiencies for each of the scintillators of 5% comes from the fact that not all of this light produced will be transferred to the SPM (coupling losses between the scintillator and SPM with an imperfect match) which will reduce this efficiency. Numbers corresponding to each scintillator material and SPM are chosen by the user, and the spreadsheet automatically looks up the values for light yield, active areas, and photon detection efficiency from product manufacturer data. The number of photons remaining after each deduction can be seen for each scintillator and SPM combination over the energy range, resulting in the overall expected system efficiency and the expected energy resolution. Multiplying the PDE (obtained from Figure 2-18) at each wavelength by the scintillator by the active area (which varies between SPMs) gives a number of photons, which was used to find the energy resolution for that combination. According to the manufacturer [27], Figure 2-18 could be used to determine the PDEs for all of the SPMs by multiplying by a constant (1.5) from the 20µm PDE values to find the PDE for the 35µm SPM. These values were used in the model to calculate the energy resolution for each combination of scintillator, crystal and SPM when used with each source. The PDE technical note was released in August 2007 showing the recent development in the technology. Plots based on the model provided the expected energy resolution as a function of energy for each of the scintillator crystals with each SPM. These were compared to the measured spectra produced later on. The spreadsheet can easily be upgraded with new values or materials. Saturation of the SPM occurs when all of the microcells simultaneously detect photons, and there are 1144 microcells on the 1mm 20μm SPM, and 3640 and 8640 in the 3mm 35μm and 20μm SPM [37]. Therefore the SPM with the highest PDE is the 3mm 35μm SPM, which will however saturate before the 3mm 20μm SPM. 44

58 Light Emitting Diode (LED) testing Before the scintillator crystals arrived, experiments utilising LEDs were designed and built to simulate light flashes. This series of experiments was used to identify the speed of the SPM response from an off to an on state, and the size of the SPM response pulses for a given LED light pulse incident on the SPM. Astable multivibrator (pulsing) circuits were setup using various combinations of components (Figure 3-14) providing a selection of duty cycles, frequencies and durations to cause an LED to flash. Figure 3-14: The transistor astable circuits produced to pulse an LED [30] (left) and a circuit diagram for the 555 IC used to pulse an LED [32] (right). The 555 Integrated Circuit (IC) was used to pulse an LED for the SPM experiments. Usually a 555 operates with a duty cycle (the amount of time the unit is on) of over 50%. Using a diode in parallel with the R2 resistance allowed the duty cycle to be less than 50%, and the on time is then related to R1*C, and the off time by R2*C. These were used to find values of resistors and capacitors to keep the LED pulse on and off for the desired times (between 0.5 and 20µs). Figure 3-15: The oscillator circuits produced to pulse an LED using transistors and capacitors (left) and the pre-packaged oscillator in the IC NE555 timer (right). 45

59 A 555 Integrated Circuit (IC) was used for the majority of the measurements with the timer output pulsing an LED requiring fewer components to set the frequency than Figure 3-15 (left). An oscilloscope placed in parallel with the LED, was used to capture the pulses the LED was receiving to confirm the frequency, duration and size. The timer is capable of microsecond pulsing [31] and the LED was then pulsed with the maximum repetition frequency of the 555 (on for 0.5μs). The tests were initially conducted on the 1mm SPM at a bias of 30V. The LED was placed on top of the SPM in contact with the glass, and secured there. The equipment was all housed in an aluminium box and covered with black cloth to prevent as much extraneous light as possible reaching the SPM. The rise time of the SPMs was found by using LED pulses incident on the SPM to determine the time taken for the SPM pulse voltage to go from off to on (onset time). The SPM was connected to a channel of the oscilloscope and another channel of the oscilloscope was connected in parallel with the LED to allow the LED pulse and SPM response to be recorded simultaneously. Various amplitudes and frequencies of light pulses were sent to the LED and the SPM responses were captured using a printer connected to the oscilloscope (Lecroy 334AM). This was repeated for various frequencies for each SPM. SPM Pulse Linearity Tests were again conducted on the 1mm SPM at a bias of 30V, to determine how the SPM responds to the varying power of an LED (changing the number of photons produced). It is expected that the parallel arrangement of the SPM microcells will provide a proportional output to the amount of incident light. A variable resistor (potentiometer) was placed in series with the output of the 555 timer circuit and a red 3mm LED (which was in contact with the SPM) to adjust the power to the LED. An ohmmeter was placed across the resistor to determine its resistance, whilst simultaneously the current through the LED was determined by placing an ammeter in series with the LED. Readings were taken approximately every 10Ω, from 0 to 70Ω, however, it was extremely difficult to get 10Ω increments accurately, due to a very small adjustment on the potentiometer changing the 46

60 resistance by several ohms. Repeat measurements were therefore not possible using the same values of resistance, however, by taking many measurements over the range from the LED on to off (around 10 measurements) a graphical fit can be obtained. The data collected (the current and voltage for each resistance) was used to calculate the power going through the LED, so comparisons between the resistance and LED voltage, resistance and SPM voltage, and power and SPM voltage can be made. Figure 3-16: Experimental arrangement for the SPM pulse linearity experiment. SPM Pulse Observations To activate the scintillator crystals 57 Co, 22 Na and 137 Cs sources at high activity with different γ- energies were selected and procured, providing peak γ-energies from 59.5 to 662keV which were used with an existing 241 Am source. Source Peak γ-energy Activity (kbq) Half life % of Source (kev) (1/10/07) Intensity 241 Am* yr Co days Na yr Cs yr Co (original)* days Table 3-1: Procured calibration source details. (*Used for CZT detector uniformity and energy resolution experiments and were original laboratory sources.) 47

61 Four scintillator crystals (CsI(Tl), BGO, LYSO and CdWO 4 ) which produce light at different wavelengths with different light yields and decay times were supplied as 3x3x3mm cubes to couple to the SPMs. These cubes were tightly wrapped with reflective material (Tyvek paper) around 5 of the sides which was held in place with plumbers tape, to ensure as much light as possible leaves through the exposed aperture to the SPM. The free side was then coupled to the SPM surface using a tiny amount of silicone grease (just enough to cover the exposed side of the scintillator) allowing conduction of the light from the scintillator to the SPM, and providing enough friction to hold the crystal on top of the SPM. The scintillator and SPM combination was then placed in the aluminium box. Power and output to an oscilloscope were connected to the SPM through a small hole in this box, and each γ-source was placed in turn at approximately 2.5cm from the scintillator crystal. The box was then closed and covered with a black cloth (in the absence of a dark room). This setup can be seen in Figure The source caused the SPM to record pulses from the scintillator crystal light, the voltage pulse size and decay times from the scintillator were found for each combination and directly printed from the oscilloscope once a pulse had triggered (trigger set above the noise). To prevent radiation damage to the crystals, sources were placed only placed near the crystals when necessary. The printouts were used to find the decay time and size of pulses so the data acquisition system could be setup to look for decay times and voltages of the pulses from each experimental combination. These values for rise time and pulse height were entered to a spreadsheet, where the differences between the sizes of the pulses from the different SPMs and crystals were quantified. Using CsI(Tl) with 22 Na in the experimental arrangement above, the effect of the increase in the SPM bias voltage from 30 to 32V was determined by exploring the sizes of the scintillation pulses printed from an oscilloscope. Noise pulses were found by removing the source and printing the SPM responses. The sizes of the pulses were read from the oscilloscope printouts for both the noise (dark photons) and for the scintillation photons, and a comparison plot of the SNR (Equation 33) for all three SPMs at the two bias voltages was produced. 48

62 Output Wrapped crystal SPM Power Source Figure 3-17: The pulse exploration and SPM spectroscopy experimental arrangement. SPM Spectroscopy Using a similar experimental setup for the pulse measurements, the wrapped scintillator cubes were coupled to the SPMs using silicone grease (RS ). The SPM output was then connected to a DAQ system (Figure 3-18), which records and displays the energy spectrum produced from the SPM. Parameters on the DAQ were set based on the decay of the crystals found previously and all settings needed to be very carefully adjusted by collecting spectra for about 10 minutes using a range of rise times and decay times to provide the best energy resolution. Once correctly setup, the spectra were collected for 30 minutes for each radioactive source, and these measurements were repeated again to allow for the standard error in the mean (Equation 34) to be applied to the energy resolutions measured. Following these measurements, the source was changed such that all four sources were used on the same scintillator crystal and SPM. The system was then setup again using the second of the four scintillator crystals, and the four radioactive sources used in turn to activate them. The measurements using this crystal were then repeated to allow an average and error to be applied. Finally, the SPM was tested without any sources present, to quantify the background. 49

63 Figure 3-18: The Xia Pixie-4 system used to acquire the spectrum from the SPM detectors. The setup was then disassembled and the experiments repeated for the next SPM where all of the measurements were repeated, and then finally for the third SPM. A summary of all the measurements planned is included in Figure Figure 3-19: A summary of the planned SPM testing, showing each SPM coupled to each scintillator crystal activated by each γ-source, and the background measurements. For consistency, the same set of scintillator crystals (cubes 3x3x3mm) was used for each of the SPM experiments even though there was a mismatch in sizes when using the 1mm SPM. For best light transfer, the area of the detector and scintillator should match. It was initially estimated that 1 in 9 of the photons would be detected due to the difference in scintillator area to the SPM area. For 1mm spectra, another MCA channel (without a 7.5x attenuation) was used with a lower SPM bias voltage (30V compared to 32V for the 3mm SPMs), and a higher threshold was set to eliminate the noise pulses. MCA events seen in the MHz region (due to noise) were reduced to ~10Hz with the correct threshold set, indicating that only scintillation events are recorded. 50

64 SPM Detector Efficiency Previously (Figure 2-20) it was seen that the ability of the scintillator crystals to attenuate at 59.5keV is much higher than at 662keV. It was required to determine the SPM detector efficiency over a range of energies. Using several energy spectra found from in previous section, the number of counts in the photopeak for the SPM detector was easier to determine than for the CZT. Using the MCA software to select a region of interest (ROI) around the γ-peak to return the number of counts, Equation 30 was used to find the system efficiency (as for the CZT efficiency seen previously, to find the area of a sphere occupied). The Single Photo Electron Spectrum The SPMs are able to detect single photons and the Single Photo Electron Spectrum (SPES) is produced when the charge from the short light pulses are integrated [21]. Using the pulsed LED circuits placed on top of the SPMs, attempts to collect the SPES were made. It was however very difficult to perform this experiment due to the requirements to produce such a small number of photons. The fastest the circuits made could pulse was in microseconds, whilst for SPES nanoseconds are preferred to ensure fewer photons are created. The distance from the LED to the SPM was maximised, the power to the LED was reduced as much as possible (using a resistor in series with the LED) and varying thicknesses of paper were placed over the SPM to reduce the light. However, when exploring the spectrum, blurring of the peaks was found, indicating the increased number of photons from the LED. This spectrum has been successfully produced by others [21] (Figure 3-20) for the SPM devices, showing three clear photoelectron peaks (corresponding to 1, 2 or 3 microcells firing due to an incident photon) after the pedestal (integration of the noise of the system [17]) peak. Figure 3-20: The single photoelectron spectrum possible with an SPM [21 modified]. 51

65 Chapter 4 : Results and Analysis Vast amounts of data and spectra were collected from the experiments using the various detectors which produced graphs, images and numerical values for comparison. These were mainly analysed using IDL software, MCA software and spreadsheets. 4.1 Modelling Results Radioactive Source Calculator It can be seen from Figure 4-1, that the biggest decrease in radioactive source activity occurs with 57 Co due to its relatively short half life. Additionally, the activity on the date of the measurements could be found allowing the intrinsic detector efficiency calculations to be performed. Figure 4-1: The activity calculator with extrapolated activity over two months. CZT Shielding Results A spreadsheet using data from XCOM was used to determine the amount of attenuation provided by three different metals to attenuate the X-rays and the fluorescence lines from the materials. Figure 4-2: A screenshot from the shielding requirements spreadsheet for pinhole imaging. It can be seen that to reduce the number of photons to an acceptable level (taken as approximately 2%) from the fluorescence lines of lead (Pb), 2mm of tin (Sn) is required, and to reduce the fluorescence lines of tin, 1mm of copper (Cu) is required. At these thicknesses, the greatest contribution to a spectrum would be from the 85keV peak from lead, which is only 2.3% of its original value. The final design included an aluminium casing so that lead is not handled directly. 52

66 Number of pixels with this count intensity 4.2 CZT Experimentation Results Uniformity and Energy Resolution The results of the first 30 minute integrations for the original combination of the 5x5 CZT array clearly show the module edges (Figure 4-3 is 241 Am). A greyscale image for the intensity at each pixel (brighter pixels shows higher photon counts) and a histogram were produced for both sources. Number of counts Figure 4-3: An image (left) and histogram (right) from the 241 Am illumination (over 30 minutes) Figure 4-4: A histogram with a fitted Gaussian for the 30 minute integration of 241 Am. The CZT detector uniformity was found using Equation 32 using values from the fitted Gaussian in Figure 4-4, where C is the standard deviation and B is the peak centroid. It was found that the uniformity of the detector after applying flat fielding corrections was 8.7% for 241 Am and 9.8% for 57 Co. With this low spread in the data we can determine that after these corrections, any images produced should have a similar (within error from the flat field corrections) greyscale intensity for the same incident radiation as the number of counts from an object. Flat fielding the 10 minute illumination using the 30 minute flat field data for each illumination results in Figure 4-5 for both sources, where the majority of the defects (module edges) in Figure 4-3 are removed. It should be noted that the bright pixels (Figure 4-5 right) make the image appear more uniform or flat than they really are, numerically there is not as much difference between the two which is shown in Table

67 Energy Resolution (%) Energy Resolution (%) Bright pixels masking true discontinuities Figure 4-5: Flat field corrections applied to the 10 minute data for 241 Am (left) and 57 Co (right). Based on the data collected, the energy resolution was calculated to be % for 57 Co and % for 241 Am, which is very encouraging, showing that spectroscopy should be possible at these energies, with another energy peak able to be resolved above 127keV in the presence of a 122keV 57 Co peak. This was tested to determine if the 136keV peak in 57 Co was seen in the energy spectrum. Once upgraded, the CZT detector had 3072 pixels on a 3x4 array (reduced from 6400 on the 5x5 array). The energy resolution per pixel was plotted following the CZT detector upgrade based on the 30 minute integration data for 241 Am and 57 Co are shown in Figure 4-6. Channel Number Channel Number Figure 4-6: The energy resolution per pixel for the 57 Co 30 minute integration (left) and for the 241 Am 30 minute integration (right). When plotted, the summed energy spectrum from each pixel shows clear peaks at the energy of the sources used (Figure 4-7) demonstrating spectroscopic capability. The 136keV is also present at 10% the activity of the 122keV line as expected. 54

68 Summed Energy Spectrum Summed Energy Spectrum Energy (kev) (kev) Energy (kev) Figure 4-7: Energy spectra using 57 Co for 30 minutes (left) and 241 Am for 30 minutes (right). Plotting the intensity per pixel for the upgraded CZT detector clearly shows an artefact in the top middle module and the module boundaries (Figure 4-8) which appear in images for both sources before flat fielding. By applying the flat field corrections in IDL to the 10 minute integration image, these artefacts are removed, and the flat fields were then applied to the future images taken. Without applying the flat field, the artefacts will continue to be present in any images taken. Substantial artefact in this module Figure 4-8: Illumination of the 30 minute 57 Co (left) and 241 Am for 30 minutes (right) showing crystal artefacts. Figure 4-9: Flat fielded image with 57 Co (left) and 241 Am (right) showing the artefact has now gone. Note the bright pixels (right) mask the true discontinuities image. 55

69 Creating a histogram from these images and fitting a Gaussian curve in IDL produces values for the standard deviation and centroid used to find the upgraded detector uniformity. (The uniformity ratio encompasses an error as the standard deviation is the error in the centroid.) Source Before Upgrade After Upgrade Uniformity (%) Energy Resolution (%) Uniformity (%) Energy Resolution (%) 241 Am Co Table 4-1: The results for the uniformity (after flat fielding) and energy resolution of the CZT detector before and after the hardware upgrade. The 10 and 30 minute flood illuminations were also completed for the ICCD detector using the same experimental setup and sources as the CZT detector measurements, so flat field corrections would be available to correct for systematic artefacts in the images produced by the ICCD detector. Table 4-2 shows the effect the flat field corrections have on the 10 minute collected data compared to that collected for 30 minutes for the ICCD detector. For both sources, the flat field causes the spread in the data to be reduced by over a factor of 2. The results in Table 4-2 show that on average the CZT detector has an overall better uniformity after the flat field corrections, than the ICCD detector, which is thought to be due to the much larger pixel size (1.6mm compared to 0.5mm) of the CZT collecting more charges reduces the statistical error on the uniformity. Source Before flat field corrections After flat field corrections % Improvement 241 Am Mean SD SD/Mean 14.53% 6.51% 223% 57 Co Mean SD SD/Mean 17.30% 8.30% 209% Table 4-2: The uniformity of the ICCD detector for each source before and after applying the flat field corrections. The errors in the CZT detector energy resolution were found using the average energy resolution per pixel and calculating the standard error (Equation 34, where n is 6400 for the original detector and 56

70 3072 for the upgraded detector). It can be seen that the error in the energy resolution is reduced for the upgraded detector indicating there is less spread over all pixels. It can be seen that a 4% improvement in energy resolution has been gained from the upgrade with the 57 Co source, even with a more pronounced artefact in one of the top modules. With this performance increase, better γ-peak separation (as the energy resolution is improved) and more uniform images (more similar pixel values for the same incident radiation) should be obtained from this detector. It was found that both the energy resolution and the uniformity of the new detector is improved compared to the original detector. Initial observations showed that the number of obviously hot pixels (those with a number of counts substantially higher than the mean) was also reduced (however, this is also because there are fewer pixels available). To quantify the effect the flat field had in addition to Figure 4-8, normalised Gaussian curves were plotted based on the parameters (A, B and C) provided by IDL for both before and after the flat field for the upgraded CZT detector using both γ-sources. Using the four sets of Gaussian parameters (both sources before and both sources after the flat field corrections), four Gaussian curves were plotted using Equation 28. The curves were normalised so a comparison for each source of before and after flat field corrections were applied to determine the change in statistical spread in the data (as seen in Figure 4-10). Source FWHM before corrections FWHM after corrections Improvement (%) 241 Am % 57 Co % Table 4-3: The effect of flat fielding on the CZT Gaussian peaks, for the upgraded detector. Figure 4-10: The effect of flat fielding on the 30 minute flood illuminations for both sources 241 Am (left) and 57 Co (right). 57

71 CZT Spectroscopy To test the spectroscopic capability of the detector, in terms of linearity in the number of counts recorded over time and the energy resolutions possible, radioactive sources were placed near the detector. The results seen in Figure 4-11 show that for two different sources with different activities there is linearity with a high goodness of fit, in the detection of the counts (taken as the maximum number in the peak) at 59.5 and 122keV with time. The CZT detector background (cover off with no sources) is insignificant compared to these values providing 14 1 counts per second (error taken as nearest count from the average background measurements of 10 and 60 seconds). Figure 4-11: A linear relationship successfully identified between increase in counts and integration time, for two γ-sources. The expected number of counts for each source can be confidently extrapolated, based on the data collected due to the uniformity of the fit. Dividing the number of counts by the integration time, the count rate (in counts per seconds) was determined for each source over the 10 second integration period which was repeated for both sources. These were found to be for 57 Co and for 241 Am. When incrementally decreasing the integration time, it was possible to obtain a spectrum with over a thousand counts in only one second using this detector, with well defined peaks at the two γ- energies used. Figure 4-12 shows an energy spectrum with three clearly distinguishable peaks (59.5, 58

72 122 and 136keV). The detector uniformity is indicated by the hits / channel chart on the bottom showing nearly all pixels receive a similar number of counts (all but one pixel has approximately 10 counts). Figure 4-12: The energy spectrum after just one second, showing 241 Am (59.5keV) and 57 Co source (122 and 136keV) peaks. It was found that the CZT detector counts linearly for both sources over the experimented time range, (as the count rate remains constant when graphed they are within error bars). As the source to detector distance increases, the number of counts decreases following the inverse square law (Figure 4-13) as expected from Equation 31. The results in this section prove that the detector can make a high resolution and fast spectroscopic detector when integrating for short times, for passive γ-ray detection, currently operating in the range of ~20 200keV. 59

73 Figure 4-13: The effect of time and distance on the maximum number of counts received for two of the integrations times tested using the CZT detector, showing the inverse square law. Figure 4-14 shows the excellent correspondence to recent (in the final weeks of the placement) published results using 57 Co [38] also using a CZT detector array. Figure 4-14: The measured energy spectrum (left) and one just published for 57 Co (right) [38]. The energy resolutions have also recently been explored by others [38], where an excellent match to the measured results in addition to the measured spectra. It was found in the Uniformity and Energy Resolution section that the upgraded CZT detector provides an energy resolution of % at 59.5keV and % at 122keV. [38] presents an average energy resolution of 5.5keV FHWM at 122keV, equivalent to 4.5% which is strikingly similar to the results measured. 60

74 CZT Efficiency Efficiency is a measure of how well the detector records the incident radiation. The CZT detection efficiency was measured to be 72 1% for the 241 Am source and 58 2% for using the 57 Co source (calculated as the standard error in the measurements where error bars were found using the fractional errors based on the error in the distance from the detector, and the error in the number of counts). This corresponds to a 24% decrease in efficiency at the 122keV energy compared to at 59.5keV, where from the attenuation of the material, only a 6% decrease is expected due to the increase in energy. The CZT detector efficiency is less than expected, based on the attenuation of the CZT material for these energy photons and is possibly due to not all of the CZT thickness able to detect photons. Figure 4-15: The efficiency of the CZT detector with increasing source distance. CZT and ICCD X-ray Backscatter Imaging To determine the performance of the imaging quality using the CZT detector and for comparison with the ICCD detector, X-ray backscatter images through a pinhole were taken. The same experimental setup was used for both detectors using the newly created graded shielding with a 2mm pinhole aperture with the CZT detector (along with the additional lead shielding required). It was proved that imaging using both the ICCD and the CZT detector is clearly possible and 20 images were taken for objects both in and out of a case. The items in a case made no significant impact on the images using either detector, due to its thin size and low density. Due to the number of CZT pixels being substantially less than in the ICCD (3072 vs. approximately 42,000) and much larger pixels 61

75 being used in the CZT (CZT pixel size of 1.5mm (pitch 1.6mm) vs. 450μm (pitch 0.5mm) for the ICCD), the images are not as clear using the CZT detector compared to the ICCD. However, the difference in contrast initially appeared better in the CZT images, where there is an object and where there is no object, which was quantified later on. The ICCD images sharpen as the pinhole reduces in size from 4mm to 2mm for the same time integration as expected (Figure 4-16). Figure 4-16: A direct comparison for the same object (sugar) with the 4mm and 2mm pinholes (before flat field corrections) from the ICCD detector. To compare the images between the CZT and ICCD detectors in the rest of this section fairly, the images are shown with the peak of dynamic range selected for both, such that the contrast is not biased towards either detector. The images however, are quantified numerically later on, where the effect of the contrast range selected makes no difference. Figure 4-17: Flat fielded images of cotton in a case, using CZT (left) and ICCD (right). 62

76 Figure 4-18: Talc and aluminium powder on a plastic base (flat fielded) using CZT (left), ICCD (middle), and before flat field corrections applied to the CZT (right). After flat field correcting, ( flat fielding ) the images produced are much clearer (compare Figure 4-18 left and right). The blackened pixels in the CZT images are assumed to correspond to the pixels switched off due to erroneously high count rates. Future work on the image processing can eliminate the effect these pixels have on the images by programming code to identify these pixels (those with an extremely low count) and setting them equal to the average of the surrounding pixels. To compare the images produced from the backscattered X-rays on the CZT and ICCD images, it was necessary to quantify the data numerically. The number of counts (CZT) and intensity (ICCD) of the pixels were measured for the same images. Regions of the flat fielded images were selected to encompass a portion of the object and then the same size portion of the background, by identifying an area by eye which appears in images from both detectors, and selecting pixels in these areas. 63

77 Figure 4-19: Images of the same object (but reversed) of the sections taken for analysis of the mean and standard deviation (CZT left and ICCD right). For the flat fielded CZT detector images, 20 pixels were manually chosen in these regions of objects and the background and the count values were taken and recorded in a spreadsheet. This was repeated so that there were two target objects and two background objects. The same regions were selected in the ICCD images, where software automatically calculates the mean and standard deviation for all of the pixels in the selected region. Figure 4-20: A graphical representation of the data collected for the ratio of object to background each image from both detectors. By investigating the same two images for each detector, it was found that for each image, the ratio of object to background was higher for the CZT detector (Figure 4-20). This means that the objects of interest stand out better for this detector than for the ICCD detector. A surprisingly high fractional error was found for the CZT detector for these measurements which was attributed to the vast 64

78 Normalised average counts in the object difference in number of pixels selected for the CZT detector and ICCD detector, causing the standard deviation to have a much greater effect on the associated error. Based on the images produced and the improved energy resolution, it is recommended that the CZT detector be fully populated with the upgraded material, to provide a wider imaging area, and therefore a larger detection probability. In the future, reducing the CZT pixel size, clearer images will be obtained, making the CZT better competition for the smoothness provided by the ICCD detector. Figure 4-21: The number of counts from the average of 10 pixels in the CZT and ICCD detector for all of the objects (error as the standard deviation). Figure 4-21 shows how the highest Z element (tungsten) produces the fewest backscattered photons (as expected as more X-rays are absorbed in the photoelectric effect at higher Z). The effect of the case the objects were in was found to be negligible for the experiments. The error (taken as the standard deviation of the data for both detectors) was greater for each of the ICCD images than for the CZT detector, which reflects the larger pixel size of the CZT providing better sensitivity. Angular Resolution The resulting backscatter images from this experiment were processed and displayed using IDL (CZT detector) and Image Pro Plus software (ICCD detector) to compare the two detectors. The images from each detector are displayed in Figure 4-22 to Figure 4-25 (the 1cm images for both detectors being used to compare the flat fielded images to the non-flat fielded images). 65

79 Figure 4-22: 1cm apart before flat field corrections (left) and after (right) for the CZT detector. Figure 4-23: 2-4cm (left to right) after flat field corrections for the CZT detector. Figure 4-24: 1cm apart before flat field corrections (left) and after (right) for the ICCD detector. Figure 4-25: 2-4cm (left to right) after flat field corrections with 57 Co 30 minutes for the ICCD detector. 66

80 Based on the diameter of the two tubes, the first position where the two tubes are physically separate is at 3cm (the two radii of the tubes are 2.5cm therefore the first tested position where the objects should be resolved is the 3cm position). A change in intensity indicating separation was found at this distance for both detectors. For the ICCD detector, the overlapping of the lids in the 2cm image are an indication that more than one object is present whilst for the CZT detector, the overlapping at 2cm causes a wider object to be produced than at 1cm. The results from the profile for each separation image for both detectors are shown in Figure 4-26, which were normalised for a direct comparison. Figure 4-26: The separation seen using the CZT (top) and ICCD (bottom) detectors over 5cm, where the background for the ICCD is always higher than CZT. For both detectors, there is a trough in the normalised counts (CZT) and normalised intensity (ICCD) at the 2cm separation image, even though there is no physical gap between the objects; the difference is due to the thickness of the materials being detected due to a change in the number of photons. This is a success for both detectors. It should be noted that there is a much higher background using the ICCD detector (about 20% of the maximum intensity compared to less than 10% for the CZT). The 67

81 separation of the materials was measured by the size of the trough between the two peaks was plotted for each detector at each distance, and the results shown in Figure 4-27, giving a comparison between the two. The high ICCD background level makes the objects stand out less than for the CZT which follows the theory as smaller pixels give better spatial resolution but reduced sensitivity. Figure 4-27: The separation displayed as the change in normalised intensity for both detectors (error bars as smallest unit taken from the trough height). This experiment showed that both detectors measure a difference in intensity where the objects begin to overlap, which increases with the separation of the two objects. Importantly, the larger pixel size of the CZT provides better sensitivity seen as a much lower background where no object is present, than the ICCD detector which shows a steady 20% background (Figure 4-26 bottom). 4.3 Scintillator and SPM Experiments Scintillator and SPM Energy Resolution Model The model showed that coupling scintillators to SPMs would be an inefficient process with some combinations of scintillator, SPM and γ-sources, resulting in a total loss of over 95% of optical photons produced from the radiation photons, producing energy resolutions from around 145% to 11%. As the photon detection efficiency is a function of wavelength, the energy resolutions possible are dependant on the scintillator crystal used. The highest PDE for the SPMs is at a wavelength of approximately 470nm [19], corresponding best to BGO (480nm) and worst for CsI(Tl) (565nm). 68

82 Energy Resolution (%) Energy Resolution (%) Energy Resolution (%) Energy Resolution (%) However, the much lower light yield of BGO compared with CsI(Tl) means that the best modelled energy resolution still occurs using the CsI(Tl) crystal. Figure 4-28: Results for the modelled energy resolutions for each crystal and SPM over the energy rage to be tested. Clockwise from top left: CsI(Tl), BGO, LYSO and CdWO 4. These modelled results correspond well with knowledge of the materials and the system, as the scintillators with the highest light yield provide better energy resolutions. The model also shows how the energy resolution improves with increasing source energy (more light flashes are produced). Spectroscopy is at least theoretically possible based on these calculations especially at higher energies, as some energy resolutions are estimated to be less than 20%. In each case the energy resolution was modelled to be best for the 3mm 35μm SPM, as from the theory the PDE is highest with this SPM, followed by the 3mm 20μm SPM with the 1mm 20μm SPM consistently providing the lower energy resolutions as it has the lower PDE and smallest active area. The energy resolution was found to vary with the square root of the energy was successfully verified for the model, and the modelled results were compared to those measured experimentally. 69

83 Scintillator and SPM: Preliminary Test Results Figure 4-29: The 1mm SPM pulses with no sources present seen in oscilloscope mode on the DAQ (left) and expanded on another oscilloscope showing distinguishable dark photons (right). Figure 4-29 (left) shows levels of pulses are present, which correspond to optical cross talk and individual photons are clearly visible on the 1mm SPM Figure 4-29 (right). Figure 4-30: The dark counts in the 3mm 20μm SPM (top) and 3mm 35μm SPM (bottom), showing noise photons not as clearly defined as the 1mm SPM. The rate of the dark photons is known to depend on the SPM size, with the 3mm SPMs having a dark rate 9 times as high. From Figure 4-29 and Figure 4-30 the dark rate can be approximated as; 1mm SPM at 3MHz and 3mm SPM at 6MHz, proving the 3mm SPMs have a higher dark noise rate. LED Testing Preliminary tests explored the pulses produced by the SPM when subjected to light from LEDs. It was found that the circuit using transistors and capacitors instead of a 555 IC took much longer to switch on 70

84 (microseconds compared to nanoseconds with visible capacitor charging seen on the graphs) and therefore the SPM response was also much slower (average of 95ns) so these results were not included in Figure All three SPMs respond to the LED circuits with a large negative pulse in the order of volts (the maximum possible is 2V from the board [33]). Figure 4-31: 3mm 35µm SPM at 32V bias using the 555 IC timer (left) and 3mm 35µm 32V (right) for an LED pulse where the SPM response (pink) to the LED pulse (blue) The time for the SPMs to change state from off to maximum pulse voltage (Figure 4-32) is between 10 and 15 nanoseconds which corresponds well to the 12ns stated [33]. (The errors were found using the smallest measurable scale on the oscilloscope printouts). Figure 4-32: The onset times of all three SPMs using pulsed LED circuits to show that the pulses are indeed produced in around 12ns. 71

85 Figure 4-33: A pulse caused by the pulsing of a red square LED, with black tape around the four exposed sides on the 1mm SPM. The SPM pulses caused by the LED photons are much slower (as they are on for µs which is similar to the scintillator crystal decay time) compared to the dark noise photons which are present for approximately 50ns. The amplitude of the pulses are also quite different, as the dark noise photons are approximately 18mV at the 30V bias, whilst the LED pulse causes the SPM to produce voltages of over 1.5V where the noise is only 1%. The SPM pulse output was measured for LED pulse durations of 0.5, 1, 2, 4 and 12.5μs using a 3mm round red LED. Figure 4-34: The SPM response to an LED pulse not fully recorded. From these LED measurements, an important though unexpected result was found (Figure 4-34). The SPM response to the LED pulse does produce the maximum pulse voltage for the whole duration of the LED pulse. After discussions with the manufacturer this was found to be due to the Alternating Current (AC) coupling, a feature of the pulse pre-amplifier board, which is more evident for the larger pulse durations (>1μs), as these are not fully recorded. Using an alternative board, a transimpedance amplifier, would allow much longer light signals to be recorded. (One was received on 25/11/07.) 72

86 This was a useful experiment to conduct, as; It was confirmed that the SPM produces pulses in response to optical photons; The size and duration of the pulses was identified which helped locating the output from the scintillators; The AC coupling due to the preamp board was identified where the full light pulse is not completely recorded by the SPM. Components were chosen to allow the LED to pulse at the maximum speed of the IC, with the LED on for 0.2µs and off for 5µs. As each LED pulse was clearly registered by the SPM, by producing a large negative pulse, it is seen that the count rate of the detector is at least 2x10 5 Hz. However, this is severely limited by the maximum rate the circuit could pulse the LED at due to the IC used, and the detector is able to quoted to be able to count at 10MHz [22]. Figure 4-35: The counting of the SPM to be at least 2x10 5 Hz. SPM Pulse Linearity As the resistance to an LED was increased, the power was reduced, decreasing the number of photons emitted. Figure 4-36 shows the linear (with a high goodness of fit) decrease of the SPM voltage pulse with the increasing resistance in series with the LED. 73

87 Figure 4-36: The effect of resistance (and therefore LED power from the 555IC timer) on the size of the SPM pulses produced. This confirms that the SPM response is linear with resistance (directly affecting the power which is directly proportional to resistance) applied to the LED, which means for any future measurements, principally spectroscopy, the SPM response the amount of light incident (energy) is directly proportional to the SPM voltage produced. SPM Pulse Observations The output pulses from the SPMs were viewed on an oscilloscope and printed for many SPM, crystal and source combinations to provide the DAQ with the correct settings required for spectroscopy. Figure 4-37: A scintillation pulse from the 3mm 20µm SPM (at 32V bias) using CsI(Tl) with the 22 Na source. Figure 4-37 shows the pulse from a CsI(Tl) crystal using the energy from the 22 Na source with a very clear 1μs decay time as expected from the crystal details value in Table 2-1. The SPM (for these times) 74

88 responds for the correct duration. There is also a trend for the voltage of the SPM pulses to reduce with the decreasing source energy as expected from the decreased light yield produced. The exception to this are the pulses from 22 Na, which appear higher than the other sources. The cause of this is so far unknown. Figure 4-38: A comparison of the voltage pulses from scintillator crystals and sources using the 3mm 20μm SPM at 32V. LYSO pulses are not present in Figure 4-38 as the pulses from the natural radioactivity were not distinguishable from those caused by scintillation flashes on the oscilloscope. (Background measurements were made to confirm that the spectra obtained were from the γ-sources and not background from Lutetium). The pulses from the CdWO 4 crystal were found to have a shorter decay time than those stated when looking at the light pulses produced by the SPM. From the previous LED measurements, it is know that longer light pulses over about 1μs are not fully recorded due to the pulse amplifier board used for pulses, hence the incorrect display of the longer scintillation pulses. 2µs 200mV Figure 4-39: The CdWO 4 response to 22 Na on the 3mm 20μm SPM, the decay lasting much less than the 20µs expected. 75

89 A transimpedance amplifier will reduce this effect, and the longer pulses will be retained. A measurement was made using the 3mm 20μm SPM with a pulsed LED using this new amplifier (Figure 4-40). Figure 4-40: Comparing the response of the pulse preamplifier which cuts off the full pulse duration (left) with the transimpedance amplifier (right). The SPM responses to the pulsed LED using the transimpedance amplifier allow the full duration of a long (18μs) pulse to be recorded (Figure 4-40 right), where the negative well created by the SPM does not stop until the signal from the LED pulse stops. For comparison, the original board (Figure 4-40 left) shows the cut off effect on much shorter pulses. The use of different boards will therefore affect the energy resolution possible from the spectroscopic measurements for scintillator crystals with a longer decay time, as only a fraction of the light will be recorded. By measuring the size of the voltage pulses produced by the SPMs at two voltages (30 and 32V), it was found that an increase in SNR (Equation 33) was achieved in pulses from the 1mm to the 3mm (20μm) SPM and a further increase from the 3mm (20μm) to the 3mm (35μm) SPM. The SNR (Figure 4-41) is not as high with the bias increase, although the much larger signal pulses (~1V) arise from the photons from the scintillator crystal which are more easily detected by the acquisition system. (A threshold was set such that the system records values greater than or equal to this value, which can be observed by viewing the event rate and increasing the threshold to suppress the noise). 76

90 SPM Spectroscopy Figure 4-41: The effect of bias on the SNR for each SPM. Following the modelling and characterisation experiments using LEDs, the arrival of scintillator crystals allowed spectroscopic experiments to begin leading to dozens of measurements, the results of which are summarised graphically and in Table 4-5. Due to the ambitious number of measurements designed to fully test the SPM system, and the scintillator crystals and 3mm SPMs not arriving until mid-october, spectra were limited to 30 minute integrations to allow for as many tests as possible to be conducted. (This was in addition to the time required to trial a range of rise times and gains, and physically setting up each experiment by coupling the detector to the power boards, coupling the scintillator to the SPM and adjusting the bias where necessary). The acquisition system used was primarily a way of collecting the data, rather than for complex analysis of the spectrum. The energy of only one peak can be entered at a time (to provide the energy resolution of, or the number of counts in, the energy peak). However, when selecting a peak as a region of interest (ROI) it can be seen that the peak shifts left when decreasing energies are applied and spectra collected, and approximate values for the centre of the peak are displayed based on the energy of the one ROI. 77

91 Figure 4-42: The energy spectrum using CsI(Tl) with the 22 Na source, providing an energy resolution of 14% using the 3mm 35μm SPM. An energy spectrum from the SPMs is shown in Figure 4-42 for the 22 Na source and the CsI(Tl) crystal. Following a well defined peak at 511keV, the Compton continuum continues due to portions of the 1274keV peak not being fully recorded in the scintillator until 1274keV. Better energy resolutions were occasionally found using the 22 Na source in place of the 137 Cs source despite its lower energy, which corresponds with the larger pulses seen when using the oscilloscope in earlier measurements. Some of the experiments matched very well with the model, which predicted several outcomes: One of the best energy resolutions possible was found using the 3mm 35μm SPM (with the higher PDE) and the scintillator crystals with the highest light yield. This was measured and an energy resolution of % was found for CsI(Tl) (predicted value 10.98%); The best energy resolution comes from the 3mm 35μm SPM for all measurements (due to the higher PDE), which was found to be the case for most measurements (Table 4-5); 78

92 Figure 4-43: The modelled energy resolutions (top) and measured average energy resolutions using CsI(Tl) for each SPM (bottom). The best energy resolution occurs at 662keV. This was found to be the case for many measurements (see Table 4-5) as more charges are created per event reducing the statistical broadening of the energy peak, increasing the energy resolution. Some combinations produce values for energy resolution of over 100%. The model predicted that there would be some experimental combinations using BGO and CdWO 4 and the lower energy γ-sources that would provide such a low light output that energy resolutions would be as high as 145%. Poor energy resolutions were experimentally verified using these crystals such that BGO on the 3mm 20μm SPM gave an energy resolution of 126 2% when using the 241 Am (59.keV) source. Having an energy resolution of more than 126% at 59.5keV means that an energy range of 76.2keV either side of the energy peak is required to resolve additional peaks. Figure 4-44: An energy spectrum from the 3mm 20μm SPM using BGO and 137 Cs. 79

93 The linearity of the detector, in terms of number of counts received with increasing time, was explored for the SPMs by measuring the maximum number of counts of the energy peak in two separate experiments; using the 3mm 20μm SPM with CsI(Tl) and the 137 Cs source, and the 3mm 35μm SPM with BGO and the 22 Na source. Figure 4-45: Two separate SPM linearity experiments; using the 3mm 20μm SPM with 137 Cs and CsI(Tl), and the 3mm 35μm SPM with 22 Na and BGO. As can be seen in Figure 4-45, there is a linear increase with a high goodness of fit for the increase in the maximum number of counts with time for two different sources and SPMs. This shows the stability of the detector over time and that there is no measured random counting drift in the detector in this range. Energy (kev) Average Measured Energy Resolution (%) Root Energy Result Modelled Table 4-4: A close match for the measured values to that expected from the model for energy resolution measurements. In many cases, the measured energy resolutions were a close match to those modelled and those predicted using Equation 20, for the measured energy resolution varying with root of the energy, as seen in Table 4-4. In this example, at 511 and 662keV the energy resolutions were better than predicted, so the scintillation losses may not have been as poor as initially predicted. 80

94 Plotting Equation 21 (Figure 4-46) shows a linear relationship is present when selected energy resolutions are chosen, and therefore Equation 20, stating that energy resolution is proportional to root energy is verified. Figure 4-46: A ln-ln plot of energy vs. energy resolution using BGO on the 3mm 35μm SPM. The very poor energy resolutions obtained for the CdWO 4 crystal have been attributed to the pulse amplifier which has been used in place of a transimpedance amplifier (which would be more appropriate for the longer sized pulses). The decay time of CdWO 4 is approximately 20μs where nearly all of the light pulse from the crystal is not recorded using the pulse amplifier. Figure 4-47: The average measured energy resolution results using BGO, LYSO and CsI(Tl) scintillator crystals on the 3mm 20µm SPM. Due to very poor energy resolutions measured using CdWO 4, the values have been removed from Figure 4-47 to avoid distorting the values recorded using the other scintillator crystals. For the 3mm 20µm SPM, the best energy resolutions come from the LYSO crystal providing better energy 81

95 resolution at higher energies. This is thought to be due to the much faster decay time being better suited to the amplifier used with the SPM for these measurements. With the transimpedance amplifier, it is expected that the energy resolution for the CsI(Tl) crystal will provide the best energy resolutions overall based on the better light yield, as previously modelled. Figure 4-48: The average measured energy resolution results from several scintillator crystals coupled to the 3mm 35µm SPM. Figure 4-48 shows how the overall trend for the improvement in energy resolution (%) is followed with increasing energy. LYSO provides the better energy resolution for the 3mm 35µm SPM at the two lower energies, however, as predicted by the model, the best energy resolution on the 3mm 35µm SPM is found using CsI(Tl) crystal at 662keV. Comparing the 3mm 20μm and 3mm 35μm SPMs (to identify the effect of the number of microcells changing the PDE), shows that in nearly all of the measurements the average energy resolution (from two measurements) is improved for the 35μm SPM, which is due to the higher PDE in this SPM. The LYSO crystal is able to offer comparable energy resolutions to CsI(Tl) at higher energies even with its lower light yield, which has been attributed to LYSO s much quicker decay time. 82

96 Figure 4-49: The average measured energy resolution results for the CsI(Tl) scintillator crystal compared to the modelled result for the 1mm SPM. Figure 4-49 shows the energy resolution improving with source energy for the 1mm 20µm SPM for the modelled values and those found experimentally. The energy resolutions achieved are the poorest of all three SPMs using the 1mm 20µm SPM (as predicted by the model). This is likely to be a combination of the crystal size mismatch causing approximately 1 in 9 of the photons to be incident on the SPM area, the lower PDE and lower active area. It was found that the 3mm 20μm SPM provided better energy resolutions than the 1mm 20µm SPM (for all but the 122keV source energy). Due to the mismatch in scintillator and SPM area for the 1mm SPM, the model estimating the energy resolution was modified, allowing for only 1 in 9 scintillation photons reaching the SPM for the CsI(Tl) crystal. It was found that the measured energy resolution fits between both the modelled energy resolutions, where it is assumed that areas match completely, and the other model (labelled modelled 1/9 th ER) where only 1 in 9 photons reach the SPM. As the model has been proved to work well for other combinations of SPM and crystals, it is reasonable to assume that the mismatch has caused the loss of some, but not as much as 1/9 th of the light emitted by the crystal as seen in Figure Figure 4-50: The measured energy resolutions with the 1mm SPM and the modelled results based on a complete match in areas and at a 1/9 area match. 83

97 One of the project aims was to determine the difference in the energy resolution between the three SPMs, therefore comparing the different pixel sizes and the two fill factors. Figure 4-51 shows how the measured energy resolution varies for the same scintillator crystal (CsI(Tl)) using each of the three SPMs. It is clear for each SPM that the increase in energy resolution is gained with increased energy. Additionally, the best energy resolution is found using the 3mm 35µm SPM, followed by the 3mm 20µm and finally the 1mm 20µm SPM (in all but 1 case) matching with the predictions. Figure 4-51: A comparison of the measured energy resolutions for the three SPMs using all of the sources on the CsI(Tl) crystal. Using Equation 20, the energy resolutions of various combinations of scintillator and SPMs were extrapolated to provide the expected energy resolution at 140keV, the energy used in SPECT imaging. The results in Figure 4-52 show that the energy resolutions at this energy are unlikely to be better than 30% for any combination, which means that other energy peaks can be resolved above 180keV or below 100keV. With this large dynamic range, it is seen that using SPM technology at two important energies, 140 and 511keV, can be both simultaneously and separately recorded at a very high photon counting rate. 84

98 Figure 4-52: The extrapolated energy resolution (based on measured results) to 140keV for eight scintillator and SPM combinations. The measured SPM energy resolution never reached the current performance possible using PMTs, or those experimentally found using a different semiconductor detector (CZT). However, the testing of this technology has proved that spectroscopy is possible. Their size and gain make them a serious contender for low power, light weight, fast counting passive γ-ray detectors. 3mm 20µm 59.5keV 122keV 511keV 662keV CdWO BGO LYSO CsI(Tl) mm 35µm CdWO BGO LYSO CsI(Tl) mm 20µm CsI(Tl) Table 4-5: The average measured energy resolutions (%). (Errors at the 95% confidence level). A summary of the measured energy resolutions obtained and the corresponding error at the 95% confidence level is displayed in Table 4-5. (A value of 0 indicates that no measurement (or repeat) was possible for this experimental arrangement.) 85

99 Observing Complex Spectra using SPMs To observe complex spectra, the same experimental setup was used for the SPM spectroscopic measurements but with an additional source near the scintillator when required. A series of experiments was conducted to identify the effect on the energy spectrum produced when simultaneous sources are incident on the scintillator crystals coupled to the SPMs (CsI(Tl) on the 3mm 35μm was initially tested as it provided among the best energy resolutions in previous measurements). Integrating briefly for two minutes using 137 Cs in the standard experimental setup (Figure 3-17) produces the spectra in Figure 4-53 (left) and adding 241 Am (placed several cm further back from the crystal to avoid saturation) provides Figure 4-53 (right). Figure 4-53: A 137 Cs Spectrum for two minutes on the 3mm 35μm SPM (left) and the effect seen when 241 Am (further back at 5cm) from the crystal is added to the experiment (right). Higher attenuation of photons occurs at lower energies (Figure 2-20), causing many more counts to be recorded at lower energies (photons are more easily stopped). Counts at higher energies occur less frequently and are not visible on the energy spectrum if a low energy source is placed at a similar distance to a higher energy source. This was found experimentally to be the case. By replacing the 241 Am source by 57 Co a γ-peak should be found at higher energy than the 241 Am peak at 59.5keV. This was successfully found, as is displayed in Figure

100 Figure 4-54: The 122keV peak clearly to the right of the cursors showing the position around the 59.5keV peak. The 136keV peak from the 57 Co source is not visible in any of the SPM measurements, due to the poor energy resolutions obtained from the experimental equipment. (An energy resolution of 50% at 122keV means that another peak could be resolved at keV, where to resolve a peak at 136keV, an energy resolution of 11% at 122keV (or better) is required, hence it is visible on the CZT detector). Peaks corresponding to high and low energy sources are clearly distinguishable in the same energy spectrum based on Figure When using both the low energy sources (59.5 and 122keV) together, the peaks are not as distinguishable due to the large energy resolutions achieved (Table 4-5). Figure 4-55: Illumination with 57 Co (left) and both 241 Am and 57 Co (right) showing a broadening due to the 59.5keV source from 33% to 52% at 150 counts. 87

101 An increase in the broadening of the peak (by 57%) is detected (Figure 4-55) but the energy resolutions are so large that the peaks at 59.5 and 122keV cannot successfully be resolved. Using both high energy sources (511 and 662keV) together allows some indication that additional peaks are present due to better energy resolutions at higher energies (Figure 4-56 and Figure 4-57). Figure 4-56: Both high energy sources (511 and 662keV) when integrating for two minutes. Figure 4-57: The energy spectrum for 22 Na (511keV) without the 662keV source, the peak is missing when integrating for two minutes. Using BGO with 137 Cs for two minutes on the 3mm 35μm SPM provides a definite spectrum with a clear peak at 662keV from the 137 Cs source is visible. 88

102 Figure 4-58: BGO with 137 Cs for two minutes on the 3mm 35μm SPM. By adding 22 Na to the experiment, after two minutes neither the 511 nor 662keV peak is well resolved but the continuum due to the 1274keV peak continues from the 22 Na source in Figure Figure 4-59: The addition of 22 Na to the 137 Cs source for two minutes. By integrating for 30 minutes, a better defined spectrum is produced (Figure 4-60) which shows the 662keV peak sitting on the 511keV peak and again the continuum present until the peak at 1274keV from the 22 Na source. Figure 4-60: Integrating 22 Na and 137 Cs for 30 minutes better defines the spectrum. 89

103 Figure 4-61: The beta spectrum from LYSO taken for 30 minutes with no additional radioactive sources using the 3mm 20μm SPM. Figure 4-61 is the naturally occurring radioactive energy spectrum from the Lutetium in the LYSO scintillator showing a low level background always present with this crystal as predicted. The effect is negligible compared to the peak source energies. SPM Detection Time In most detection applications, the faster the detection time the better. The time taken to produce an energy spectrum using the 3mm 35μm SPM and the CsI(Tl) crystal with 57 Co (closest energy for SPECT imaging) was explored (Figure 4-62). This verifies the uniform counting of the detector over the time and energy ranges explored. In just 10 seconds, over one thousand counts are in present the 122keV energy peak allowing identification that the energy is present. Figure 4-62: The linear increase of counts with integration time for the 3mm 35µm using 57 Co and CsI(Tl). 90

104 Extrapolating the best fit line in Figure 4-62, an energy peak with 1000 counts at 122keV would be present in less than 10 seconds. The closeness of the energies between 122 and 140keV allows comparisons to be drawn to SPECT medical detection of 140keV, where rapid detection of this energy would be possible using SPMs and scintillator crystals. SPM Efficiency Results For the 3mm 35μm SPM, the trend was successfully identified that as the source energy increases, the detection efficiency decreases. This corresponds with Figure 2-20, where fewer photons are attenuated at higher energies. Figure 4-63: A comparison of the system efficiencies for each SPM using CsI(Tl). Figure 4-63 displays the measured detection efficiency of the three SPMs using the CsI(Tl) crystal, showing how the efficiency decreases with increasing source energy (as expected). The attenuation data (curve labelled CsI Att) shows how the ideal attenuation in the crystal over the energies should decrease with increasing energy. It is seen that the experimental observations follow the predicted outcome where the efficiencies reduce with increasing energy. However, the measured observations show lower efficiencies than the ideal curve. Possible reasons for this include the loss of photons from any slight mismatch of the alignment of the scintillator crystal to the SPM, and absorption of photons in the silicone grease between the scintillator and SPM, which would cause fewer photons to be detected by the SPM. To improve this setup the scintillators would be coupled directly using superior optical grease and the crystals would be perfectly matched to the SPM area to maximise the efficiency. 91

105 Chapter 5 : Review, Conclusions and Further Work 6.1 Conclusions Over the course of the M.Phys. placement year, research was performed mainly at the Dstl laboratory, exploring radiation detection. Modelling the attenuation of transmission X-rays for a range of materials was undertaken initially, showing the distinct difference between metals and organics. This followed a review of the different detection techniques which populated a graphical model to provide information on where the research would be useful, principally in identifying the spectroscopic performance of radiation detectors. CZT detector preparation and characterisation experiments were conducted to gain an understanding into detector theory and application. The main aims of this project were to determine and quantify the performance of the CZT and SPM detector systems as new detectors for spectroscopy and imaging. This is of great importance as spectroscopic information provided by γ-detection and could provide the possibility for material discrimination based on the energy of the material present. This could have an impact in medical and security applications. An ICCD detector was unable to provide any spectroscopic information, but was used as a basis of comparison to the CZT detector for the backscattered X-ray images. When directly comparing the ICCD detector to the CZT for X-ray backscatter imaging, it was seen (Figure 4-26) that the CZT detector provides lower noise and better object to background ratios (Figure 4-20). The ICCD detector is quite long about 0.6m, whereas the CZT detector is only 0.25m thick with the specifically produced shielding applied. An SPM could further reduce this thickness to less than 5cm including a scintillator crystal. An array of SPMs to cover the area of the current ICCD detector could provide a thin lightweight alternative to both the CZT and ICCD detectors, to provide imaging and spectroscopic information with the ability to be a fast counting passive γ-detector. The energy resolution of the CZT array, required to identify how well close peaks can be resolved, was quantified before and after a hardware upgrade, which identified an improvement in the energy resolution using two different γ-sources. It was found that the CZT detector provides excellent energy 92

106 resolution of 4% at 122keV allowing direct application as a γ-ray detector which is very close to the energy of the γ-line used in SPECT imaging for medical applications. The inverse square law for the decrease in radiation with increased distance was verified for the CZT detector array when several distances (15 to 30cm) were tested with a range of integration times (1 to 10 seconds). The linear increase of the counts with time was also verified, proving the detector s stability. To use the CZT detector array as a pinhole imager, the thickness of different Z materials was calculated to provide shielding only allow X-rays through the pinhole. Pinhole imaging using the CZT detector to detect X-ray backscattered photons demonstrated the concept worked for both the ICCD and CZT detectors. The images produced from the X-ray backscatter experiments show that without image processing (by flat fielding) the ICCD detector produces images which are initially clearer due to the large number of smaller pixels. After flat fielding, the CZT images are made clearer, with the effect of flat fielding being more pronounced for the CZT detector (as seen in the change in uniformity before and after the flat field corrections were applied). Unfortunately, the graded shielding manufactured, requires additional lead to prevent the bright spots seen in the CZT images. An optimum configuration using extra lead required to prevent these spots was used for the measurements and will be used until the shielding problem is fixed. From the backscatter measurements, the larger pixel size of the CZT was found to provide better sensitivity, shown as a lower background noise (at least 10% less than the ICCD detector where no object is present) and a better object to background ratio than the ICCD detector when using the same experimental setup. By determining the number of counts detected, the efficiency of the CZT detector was determined. The highest efficiency of the CZT detector was found to be 70 1 % (at 59.5keV) which is lower than expected based on the attenuation of the γ-ray photons over the 200keV energy range explored. This is possibly due to attenuation in the thin carbon window on the aperture, or the 5mm thickness not fully depleted and able to attenuate and detect radiation photons. Considering the spectroscopic energy resolution and detection efficiency, this detector was found to have many useful application areas, and has been shown to be successful as a high energy resolution passive γ-ray detector, and as an active imager for X-ray backscatter measurements. 93

107 Testing on novel SPMs initially involved modelling how the energy resolution varies when using different sized SPMs with four scintillator crystals and four radioactive sources to activate the crystals, to explore the possibility of spectroscopy. The model showed that energy resolutions of less than 15% (and more than 140%) would be possible with some combinations of scintillator crystals, SPMs and radioactive sources with improving energy resolutions at higher energies. A vast amount of testing was then performed on novel SPM detector systems providing several key findings: The pulse output is directly proportional to the amount of light incident on the SPM with direct application to radiation spectroscopy, where increased energy will linearly produce more scintillation light; The decay constant of the crystals can be accurately seen using the SPM when exploring the short scintillation (< 1µs) pulses on an oscilloscope; The pulse preamplifier board used for the SPM measurements is unsuitable for longer (>1μs) light pulses and at the time of writing a new amplifier arrived which was proved to work with longer light pulses; Spectroscopy was found to be possible for various combinations of SPM and scintillator sources over a wide range of γ-energies. There are direct comparisons between the model predicting the energy resolutions, and with the spectroscopic measurements made. Energy resolutions were measured to be from 11% up to approximately 175%; When more than one source is used on the scintillator and SPM combination, it was found that the peaks corresponding to two sources are clearly distinguishable when large source energy separations are used. The resolution of close peaks is more difficult, and in some cases impossible with similar energies due to the poor energy resolutions found previously; The SPM detectors were shown to count linearly with increasing integration time using two different SPMs and different scintillator crystals; 94

108 The detection efficiency was found to decrease with the increasing source energy as expected from the modelling of the crystal attenuation modelling. The best SPM detection efficiency was measured to be 65 1%. Currently the size limitation of the SPM being only several millimetres, ensures that PMTs are still the standard for detection of light from scintillator crystals for the majority of applications. However, the development of arrays of SPMs is underway [17]. According to [6] the energy resolutions achievable using a PMT are 8.5% at 122keV using a NaI scintillator. Using Equation 20, this extrapolates to an energy resolution of 3.6% at 662keV. The SPM spectroscopic measurements never matched or improved upon these values however, the best possible energy resolution did match closely to that modelled proving the SPM detector concept. Detector Comparison The pixel pitch of the CZT detector used is 1.6mm which produces coarser images than the ICCD detector (pitch is 0.5mm). An array of SPMs scaled up to the size of the CZT would have about 870 3mm pixels where the CZT has 6400, and therefore any possible images would be very pixellated compared to the CZT reducing the spatial resolution. A single 3mm pixel of SPM can count photons at 10MHz which means the 870 pixels able to fit in the area of the CZT would be able to count at about 9GHz, where the same area of CZT is able to currently count at only 100kHz. However, saturation occurs at fewer counts per pixel on the SPM as this is limited to the number of microcells. Based on the number of microcells in the 3mm 35μm SPM (providing the best energy resolution), an array of these SPMs made to the same area as the fully populated CZT detector would be able to detect just over 3 million simultaneous photon interactions; thirty times more than the CZT detector currently can. The energy range of the CZT detector used is currently limited from 0-200keV, whereas the SPM has so far successfully detected from 59.5keV up to 1274keV providing a much larger dynamic range. However a comparison between the best obtainable energy resolutions (Table 5-1) for two energies shows that the CZT is far superior (Table 5-1) to the SPMs at both energies. 95

109 Detector Best Energy Resolution at Best Energy Resolution 59.5keV (%) at 122keV (%) SPM CZT 8 4 Table 5-1: The best obtainable spectroscopic energy resolutions (nearest %) for the two detectors at the same energies. The detection efficiency of the CZT detector (Table 5-2) was found to be 6% above the value obtained for the SPM using CsI(Tl) at the using the same 59.5keV source, showing that both detectors are able to successfully attenuate photons in the keV energy range. This was predicted from the attenuation curves, where CsI is able to attenuate slightly less over the energy range with comparable detection efficiencies. Detector Best Efficiency at 59.5keV (%) SPM 65 CZT 71 Table 5-2: The best obtainable detection efficiencies (nearest %) for the two detectors at the same energies. Creating an array of 1mm SPMs could provide better spatial resolution that the current CZT provides are smaller pixels would be used, however, as seen from the spectroscopic results, this would be at the expense of energy resolution, as the 1mm SPM provided the poorer energy resolutions. 6.2 Further Research Many measurements were made using both the CZT and SPM detector systems. Due to time constraints with the equipment, it was not possible to explore as many combinations for spectroscopy as planned earlier in the year. With more time, a complete comparison with the modelled results (particularly with the 1mm SPM) would be possible, as with any imaging device, the smaller the pixel size, used the more coarse the image will be as seen in the images between CZT and ICCD. Experimentation with a newly available 16 3mm pixel SPM array (Figure 5-1) will allow a small area imaging device to be tested (to be connected to the 16-channel acquisition system) to determine the spatial resolution and sensitivity of a new spectroscopic imaging detector. This array would allow a 96

110 wider area detector, capable of imaging to be tested for sensitivity for comparison to the CZT and ICCD for image quality. Figure 5-1: An array of 16 3mm pixel SPMs [34]. Obtaining an LED pulser should allow the single photoelectron spectrum for each SPM to be collected to determine the SPM gain. This will provide a direct comparison with PMTs which have well researched gain of the order of Repeating all of the spectroscopic experiments with the new transimpedance amplifier (in place of the pulse preamplifier supplied) should improve the energy resolution for the slower scintillator crystals (CsI(Tl) and especially CdWO 4 ) and provide another comparison to the PMT energy resolutions. Further exploration of the lower than expected CZT detector efficiency is required to identify if the full 5mm thickness of the CZT detector is able to detect photons. Additionally, calculating the attenuation in a carbon window protecting the CZT, to identify if this is the cause of the efficiency loss. In conclusion to this research project, characterisation experiments of the CZT and the SPM detector systems was conducted providing key findings to the department. This research project has been exceptionally interesting to explore over the placement, and it is hoped that the research conducted will continue to be of use to both the department and the wider physics community. 97

111 References The sources, including books, papers and technical data sheets, which have been referred to throughout the research project are listed below January Explosive detection systems (EDS) for aviation security, Sameer Singh et al., Maneesha, Introductory Nuclear Physics, Kenneth Krane, X 4. Fundamentals of Physics, 6 th Edition, Halliday / Resnick / Walker, Oxford Dictionary of Physics, 2003, Oxford University Press, Radiation Detection and Measurement, 3rd Edition, G. F. Knoll, October Principles of Physics, M. Nelkon, 8 th Edition, Properties of Narrow Gap Cadmium-based Compounds, Inspect publication, IEE, 1994, p Evaluation of a CdZnTe pixel array for X- and γ-ray spectroscopic imaging, F. Quarati et al., Digital Camera Fundamentals, December December Physical properties of Common Inorganic Scintillators, Saint-Gobain Crystals, ( pdf) 14. High-Energy Photon Detection with LYSO Crystals, R.W. Novotny et al., Scintillation: mechanisms and new crystals, M.J. Weber, The quest for the ideal inorganic scintillator, S.E. Derenzo et al., Tiled Silicon Photomultipliers for large area, low light sensing applications, P J Hughes et al., SensL, First Results of Scintillator Readout With Silicon Photomultiplier, Deborah J. Herbert et al., SPM Photon Detection Efficiency Technical Note Rev 1.3, SensL, August SPMScint_High_Performance_SPM_for_Radiation_Detection.html, November Study of the Properties of New SPM Detectors, A G Stewart et al., Introduction to the Silicon Photomultiplier Technical Note, SensL, Rev1.0 August The Silicon Photomultiplier for application to high-resolution Positron Emission Tomography, D.J. Herbert et al, The Expression of Uncertainty in Testing UKAS Publication ref: LAB 12, Edition 1, October TCT characterization of different semiconductor materials for particle detection, J. Fink et al., , 04/10/ , 09/10/ November , 20/11/ Industrial Electronics, Noel Morris, 2nd Edition, NE555 precision timer datasheet, Texas Instruments, SLFS022, Revised Feb 1992, October October SPM Pulse Preamplifier Technical Note, SensL, Rev 1.3, August SPMArray_Position_Sensitive_Multi_Anode_High_Gain_APD.html, October April November , 22/10/ Study of Cadmium Zinc Telluride (CZT) Radiation Detector Modules under Moderate and Long-Term Variations of Temperature and Humidity, Gunnar Mæhlum, November

112 Chapter 6 : Appendices APPENDIX I: OBTAINING A SPECTRUM Due to the novelty of the acquisition system, a certain amount of feet-finding was required, and after much manual reading, contact with the product manufacturer and general trial and error, the best procedure for setting up the system was found and can be seen below. One Quick Start method is to load a pre-existing experiment, such as those on the desktop and load the settings (from the Load menu on the Settings tab). The spectrum can then be refined for the new device by adjusting key parameters. When starting a new series of experiments from a new SPM: 1. Explore the pulses from the detector using an oscilloscope and note the rise time and the decay time. 2. Once logged into the computer, open the "Shortcut to Pixie4" from the desktop. The following screen will appear, and click on 'Start Up System'. Figure 6-1: The system start-up screen. 3. Once started, the Pixie4 Run Control menu will be displayed, which contains the four tabs (circled in Figure 6-2) to setup the system, and to take and analyse the data. 99

113 Settings Tab Figure 6-2: The Pixie4 Run Control menu. Note: If these options are not all visible, you should press the 'More' button on the bottom of the menu. As the SPM produces negative pulses, it is necessary to ensure that the 'Trigger positive button' in the channel menu of the settings tab is unselected. The trigger filter is particularly important and should be set a value such that the MCA is not triggered on the noise pulses. (From the oscilloscope the number of real light pulses can be estimated (<50 per second) whilst the noise pulses are in the MHz region). Calibrate Tab Here, the gain can be adjusted to utilise the full 14-bit dynamic range of the instrument. Values of seem best for SPM spectra. Run Tab The integration time is set here and after a few minutes, a good indication of the general shape of the spectrum should be produced. Whilst running, you can periodically Update the MCA view to view the event rate (in the Analyse tab) and see how the energy spectrum progresses. Analyse Tab Here the number of counts per second is displayed, if this is in the hundreds or higher, for a detector which is known not to produce this number of pulses, the trigger and threshold need to be increased. The rise time value and gain should be adjusted and short (10 minute) runs in MCA mode should be obtained to ensure the peak can be distinguished above the noise. Once these values are optimised, take a longer run (30 minutes) to reduce statistical fluctuations, as more counts will be received. Peak Fitting On a peak of the observed spectrum, a Gaussian curve can be fitted to an energy peak, from which the energy resolution can be found, select a cursor (a square or a circle) and drag it to 100

114 the left of the peak, and drag the remaining cursor to the right of the peak. Click on the 'Fit down arrow and select the channel holding the data (0 or 1 usually). This will provide values for FWHM% and FWHM abs. The FWHM% is the energy resolution. (An automatic should be available in a newer version of the software.) Figure 6-3: The positions of the cursors to find the energy resolution. Figure 6-4: The cursors around the peak to provide the energy resolution and value of the peak. To calibrate the system, the value for this peak can be entered into the Peak box (shown in black) for that channel. Attenuation Jumper settings on the each channel can be used to provide an attenuation of 7.5x if required (for larger voltage pulses) by changing jumper combinations seen in Figure

115 A jumper on the 50 pins keeps the input at 50Ohms (else 5kOhms). A jumper on the Attn pins removes 7.5x attenuation (else there is 7.5x attenuation). [39] Figure 6-5: The attenuation jumpers for each channel of the acquisition system. 102

116 APPENDIX II: GMS GMS allows easily identifiable links between nodes so the items that are linked can be instantly identified. When the mouse pointer is held over a box (or node), the box and any links with that box are illuminated easily showing the links. This makes things easy to see when the screen gets busy with lots of information. Figure 6-6 and Figure 6-7 are screenshots from a GMS tutorial highlighting the effect of linking, showing nothing highlighted and then with the mathematics students shown. Similarly, it can be seen which subjects Mark takes by moving the mouse over his name. Figure 6-6: GMS with no object selected. Figure 6-7: Identifying who takes Mathematics by moving the mouse over Mathematics. Weld View is a piece of software which can easily show the relationships between the quantities in the GMS model such that they can be read from a grid (Figure 6-8). 103

117 Figure 6-8: The Weld View of the Students Tutorial. 104

118 APPENDIX III: Initial Research: Introduction and CZT Resistivity The following section is based on reports produced during the initial part of the research program. Gold contacting a CZT sample is performed to connect the sample to a circuit to determine the characteristics and for practical use of the device (applied voltage required for detector operation). An experiment explored the use of conducting foam in place of / before gold contacting CZT. It would be very useful to be able to find the resistivity of a detector sample before the contacting process as contacting is a costly procedure in both time (taking hours to obtain the correct vacuum for each contact, where two or three contacts are required in total) and financial benefits (cost of pure gold). Finding if a sample has a good enough resistivity to continue with experiments before contacting would therefore be very useful. Figure 6-9: A schematic of the conducting foam used [35]. A non-contacted CZT sample (about 5x5x1mm) was sandwiched between two separate pieces of conducting foam, and the current through the sample was measured by applying four voltage ranges in varying step sizes (Figure 6-10). Resistance was found using the inverse gradient of an IV graph. With known sample dimensions, the resistivity can be found from Equation 16 to find values shown in Table

119 Figure 6-10: The experimental arrangement to find the CZT sample resistivity. Voltage Range (V) Step Size (V) Readings / Sample Gradient (1/Ω) Resistivity (Ωcm) Default 7.44E E Default 8.85E E E E E E+12 Table 6-1: Resistivity values and results from the conducting foam experiment. Higher than expected values of resistivity were found using the non-contacted CZT sample and conducting foam. The increase in the resistivity is thought to come from a larger than true value being used for the area in contact with the needle, as the area of the sample was used in calculations, when only the area in contact with the foam should have been used. Using a smaller value for the area based on the size of the needle rather than the whole detector sample in Equation 16 will reduce the resistivity. The experiments are being performed again based on a known value for contact area of a larger needle, and a more reasonable (lower than previously found) value for resistivity should be found. This looks like it may be a promising way to find the resistivity of a sample without the need for gold contacting providing several key benefits, including the potential for money saving, by making faster resistivity measurements. 106

120 APPENDIX IV: Experimental Equipment List The research project required many of the items to be procured by the author taking time for the ordering (obtaining the specific requirements, specifications and quotations where necessary), in waiting for the equipment to arrive, and checking the correct equipment arrived. The testing required different SPMs, radioactive sources, scintillator crystals and an acquisition system to record the pulses produced. Silicon Photomultipliers: These were the main emphasis of the testing and three were ordered to compare both the physical area and the effect of the fill factor in the detector. These were available as single pixel detectors, in 1mm SPM 20µm fill factor, 3mm SPM 20µm fill factor and 3mm SPM 35µm fill factor. The SPM detectors of 1mm pixel size provide an area of 1mm 2 and for the 3mm the area is 9mm 2. Scintillator Crystals: Four different scintillator crystals were provided as cubes (3x3x3mm) to provide a range of light yields from 9,000 to 52,000 photons/mev in decay times from 0.04 to 1µs to fully test the response of the SPMs. Data Acquisition System / MCA: The Digital Gamma Finder (DGF) Pixie-4 is a 14-bit digitising module allowing 4 simultaneous acquisition and analysis of pulse height for each channel. The system fits in a standard PCI crate and more modules were added together for acquisition of 16 inputs. Pixie- 4 Viewer is a graphical user interface written using IGOR from Wavemetrics operating in the Windows XP (TM) environment makes the easier to use [36]. There are various settings required so an MCA run can be started, which include the signal rise time, decay time, gain, run time and signal thresholds. Some of these can be found by first exploring the pulses on a standard oscilloscope. APPENDIX I provides the details to allow an energy spectrum to be found. 107

121 APPENDIX V: Varying Spectra with Energy The following figures are the spectra produced from the CsI(Tl) crystal using the 3mm 35µm SPM using decreasing energies from 662keV (as the main source energy) down to 59.5keV. Figure 6-11: The change of source energy from 662keV (top) to 59.5keV (bottom). 108