Properties of silicon photon counting detectors and silicon photomultipliers

Transcription

1 Journal of Modern Optics Vol. 56, Nos. 2 3, 20 January 10 February 2009, Properties of silicon photon counting detectors and silicon photomultipliers A.G. Stewart, L. Wall and J.C. Jackson* SensL, Lee House, Riverview Business Park, Blackrock, Cork, Ireland (Received 28 January 2008; final version received 11 September 2008) Geiger-mode avalanche photodiodes (APD) or photon counting detectors (PCD) have become the basis of a range of new detectors and applications. Such devices are sensitive to single photons and are being adapted to create different detector technologies such as the silicon photomultiplier (SPM). A silicon photomultiplier, so-called because of its similarity in performance to conventional photomultiplier tubes, is based on an array of photon counting detectors but with a single common output. While the silicon photomultiplier does not provide positional information, it does allow photon number resolution and photon counting at higher count rates than are achievable with a solitary PCD. These new detectors are enabling a range of new applications in the fields of medicine, biology, high-energy physics and space exploration. In this paper we report the performance of PCD with dimensions ranging from 20 mm 20 mm to 200 mm 200 mm and on 1 mm 2 silicon photomultipliers and illustrate the use of an SPM in single photon counting mode. Keywords: photon counting; silicon photodetectors; silicon photomultiplier; photomultiplier tube; Geiger-mode; photon timing 1. Introduction Photon counting detectors (PCD), based on silicon Geiger-mode avalanche photodiodes (APDs) [1], have recently come to the market in a range of commercial products [2] and are used in a diverse range of applications from the life sciences to astronomy. In their simplest form, these detectors make excellent photon counters with high optical detection efficiency and fast timing response. However, the photon counting detectors can also be considered as building blocks for other detector concepts. For example, by monolithically assembling an N M array of individually addressed photon counting detectors, a photon counting camera or DigitalAPD can be realised. Such a configuration would allow two-dimensional imaging to be performed under extremely low-light level conditions [3]. Another Geiger-mode photodiode detector array concept is that of the silicon photomultiplier (SPM) [4 8]. An SPM detector consists of an array of photon counting microcells with a single common output. Each photon counting microcell consists of a Geigermode photodiode in series with a monolithically integrated quenching element. By connecting all the SPM microcells in parallel, the SPM can be considered to be an analogue detector in which the output photocurrent or number of Geiger pulses is proportional to the number of incident photons. The SPM detector has many of the attributes of conventional photomultiplier tube (PMT) detectors including high gain and single photon sensitivity at room temperature. In addition the SPM has all the benefits of a solid-state device such as compactness, high degree of robustness and low operating bias voltage. The SPM detector also has two key advantages over the PMT, these are insensitivity to magnetic fields and the SPM is not damaged by exposure to high photon flux. In addition, the SPM offers photon number resolution which is critical for a number of applications including linear-optics quantum computing [9]. The SPM is proving useful in a number of applications including microscopy and is particularly suited to the optical readout of scintillator crystals and fibres. Scintillator crystals and fibres are used to convert high-energy photons (Gamma and X-ray) and high-energy particles into short flashes of lower energy photons with wavelengths typically in the visible region of the spectrum. Scintillators are widely used in high-energy physics and nuclear medicine. Positron emission tomography (PET) is a medical imaging technique that produces three-dimensional images or maps of functional processes in the body. PET systems rely on the fast and efficient detection of anti-parallel 511 ke V gamma rays from positron annihilation and SPM detectors are ideally suited to the detection of the visible photons emitted by the scintillator when the gamma rays interact with the scintillator crystal [10,11]. *Corresponding author. cjackson@sensl.com ISSN print/issn online ß 2009 Taylor & Francis DOI: /

2 Journal of Modern Optics Silicon photon counting detectors Planar structure silicon photon counting detectors with wide active areas are required to obtain high photon collection efficiency without the need for complex and costly optical alignment components and for fibre pigtailing where the fibres typically have diameters of 100 mm. We have fabricated silicon PCD devices with a square geometry where the optically sensitive areas have the dimensions of 10, 20, 50, 100, 150 and 200 mm per edge. The devices are fabricated on 100 mm diameter silicon epitaxy on p-type wafers and are formed from planar n þ -p junctions. A selectively implanted region forms an enrichment area that defines the optically active area and determines the breakdown voltage of the device. The polysilicon n þ top contact layer is formed by diffusion from a phosphorous implant. The n þ layer extends beyond the enrichment region to form a guard ring. This helps to reduce the electric field at the edge of the diode and gives a uniform breakdown [12]. The process also provides for high value, compact resistors which are used for passive quenching circuits where required for PCD and SPM. Both PCD and SPM devices are characterised on wafer for pre-breakdown leakage current, breakdown voltage and dark count rate. This is achieved using an automated wafer probe station from Cascade and computer controlled test equipment. The data is used to create a wafer map that can be used to locate the position of the best performing detectors. Once the wafer is sawn, the individual silicon die are automatically picked and assembled Breakdown voltage The breakdown voltage of the devices is set by the dopant concentration on the low-doped side of the p-n junction. The uniformity in breakdown voltage across a wafer is therefore a good indicator of the process uniformity and is critical in the development of the SPM detectors since variation in the breakdown voltage across the array of microcells would result in variations in optical detection efficiency and gain. Figure 1(a) shows the room temperature breakdown voltage statistics for a total of 420 PCD devices with active areas of 50 mm 50 mm and 100 mm 100 mm. The figure shows a histogram of the breakdown voltages together with a cumulative probability plot. The data shows that 90% of the PCD devices on the wafer have a breakdown voltage between 27.8 and 28 V and that 40% of the devices have a breakdown voltage of 27.9 V. The extremely tight breakdown voltage distribution across the wafer is crucial for the development of silicon photomultipliers and results from the detector design and the uniformity afforded by modern batch processed semiconductor fabrication Dark count rate The dark count rate of the photon counting detectors is assessed at wafer level as a function of bias voltage and temperature. The bias voltage is typically expressed as the bias voltage applied in excess of the breakdown voltage and is referred to as the overbias. Since the breakdown voltage is a function of temperature [13] the same bias voltage at different temperatures results in different overbias values. As such the breakdown voltage should be determined at different temperatures in order to determine the overbias. On production wafers all the devices on the entire wafer are measured and the results used to create a statistical analysis of the wafer dark count rate level as well as allowing individual detectors to be binned and sorted for assembly. Figure 1(b) shows a cumulative probability plot of the dark count rate data obtained for all 150 mm 150 mm and all 200 mm 200 mm PCD devices on a 100 mm wafer. The dark count rate was measured at 4 V above the breakdown voltage and at 20 C. The figure shows that the typical median dark count value for a 150 mm 150 mm PCD device at 4 V above the breakdown voltage and at 20 C is 5.9 kcps while the median dark count rate for a 200 mm 200 mm PCD device under identical conditions is 10.3 kcps. Such probability plots are used to bin the detectors according to dark count rate. The ratio of the median dark count rates of the detectors is 0.57 while the ratio of the active areas is 0.56; showing that the dark count rate scales with the area of the device. Figure 2(a) below shows the bias dependence of the median dark count rate of 100 mm 100 mm devices across a 100 mm wafer at room temperature, 20 C and 40 C. The error bars represent the 25th and 75th percentile points of all the dark count rate data at each overbias point. Figure 2(b) shows the mean room temperature dark count rate at 4 V overbias across a 100 mm wafer of each size of PCD fabricated. The figure shows that the mean dark count rate of the PCD devices scales with the area of the detector. This is a very significant result as previous attempts to fabricate large area PCD devices resulted in considerably higher dark count rate values than would be expected from the relative increase in active area. Table 1 below shows the median room temperature dark count rate attained from each detector size from a 100 mm wafer.

3 242 A.G. Stewart et al. Figure 1. (a) Histogram (primary y-axis) and cumulative probability (secondary y-axis) plot of room temperature breakdown voltage of 50 mm 50 mm and 100 mm 100 mm PCD devices across a wafer and (b) cumulative plot of the dark count rate across a 100 mm wafer at 4 V above the breakdown voltage and at 20 C for 150 mm 150 mm (circles) and 200 mm 200 mm (squares) photon counting detectors Photon detection probability The PCD devices are mounted on thermoelectric coolers (TEC) and housed in hermetically sealed T08 cans. When the detectors have been packaged the optical efficiency and spectral response of the detectors is measured. The photon detection probability (PDP) is given by the quantum efficiency of silicon, which is a function of wavelength, and the carrier avalanche initiation probability. The PDP can be written as PDPð, VÞ ¼ ðþ "ðvþ, ð1þ where () is the quantum efficiency at a given wavelength and "(V ) is the carrier avalanche initiation probability and is a function of the applied bias V. To measure the optical efficiency or PDP of a detector, the detector is mounted in an integrating sphere and the optical efficiency measured relative to a calibrated large-area silicon PIN diode (Newport 1853-C). The PCD and Newport detector are mounted in adjacent ports of an integrating sphere and illuminated with a low-level continuous signal from a Bentham TMC300 monochromator and white light source through a third port in the integrating sphere.

4 Journal of Modern Optics 243 The PCD is passively quenched and the signal level adjusted such that the detected photon rate is typically less than 100 khz. The PCD dark count and signal count rates are then recorded at each wavelength together with the photon flux as measured by the Newport reference detector. Figure 3 shows the PDP as a function of wavelength of a 50 mm 50 mm PCD detector at room temperature at 1.5, 2, 3 and 4 V above the breakdown voltage. The detectors have a peak PDP of 42% at 480 nm and at 4 V above breakdown. 3. Silicon photomultipliers One limitation of photon counting detectors is the need to quench and reset after every detected photon. Figure 2. (a) Median dark count rate of 100 mm 100 mm photon counting detectors as a function of overbias across 100 mm wafer at room temperature, 20 C and 40 C and (b) median room temperature dark count rate at 4 V overbias as a function of PCD device area. (The colour version of this figure is included in the online version of the journal.)

5 244 A.G. Stewart et al. This typically limits the maximum count rate of a single detector to about 10 MHz. In addition, the PCD output is binary and independent of the number of photons that initiated the Geiger avalanche. These limitations can be overcome if an array of photon counters are connected in parallel such that the output is proportional to the incident number of photons. A silicon photomultiplier is a parallel array of photon counting microcells connected to a common output. Each microcell consists of a PCD and an integrated quench resistor. We currently fabricate both 1 mm 2 and 9 mm 2 SPM detectors with a variety of different microcell designs. The microcell design is a compromise between a number of conflicting parameters for a given SPM active area. The trade-off is between the microcell active area, the total number of microcells, which directly relates to the dynamic range, and the SPM fill factor. Different applications may require different SPM designs depending on the required Table 1. Median room temperature dark count rate for PCD devices at 4 V overbias from a 100 mm wafer. Detector size (mm) k k k k k Median dark count rate (cps) performance parameters. The SPMMini1020 has an active area of 1 mm 2, consists of 1144 microcells and has a fill factor of 43%. Each microcell consists of a 20mm square geometry PCD and a monolithically integrated quenched resistor and have a 29 mm pitch. Figure 4 shows two micrographs of the SPMMini microcells. The typical breakdown voltage of the microcells is 28 V. For free space and fibre coupled applications, the SPMMini1020 is mounted on a thermoelectric cooler (TEC) and housed in a hermetically sealed T08 can. With the T08 can mounted in a heatsink, the TEC allows the SPM detector to be operated at 20 C. Since the SPM dark count rate is thermally activated, cooling the detector results in a large reduction in the dark count rate. For applications where the SPM is directly coupled to the emitting facet of a scintillating crystal, fibre or optical waveguide the SPM is packaged in a T05 package and covered with a 300 mm layer of transparent epoxy [14] Photon detection efficiency The photon detection efficiency (PDE) of an SPM is given by the product of the photon detection probability (PDP) of the microcells and the geometrical efficiency or fill factor of the device and can be written as PDEð, VÞ ¼ðÞ"ðVÞF, where F is the SPM fill factor. The PDE is measured using a similar method to that described for the ð2þ Figure 3. Photon detection probability as a function of wavelength for a 50 mm 50 mm PCD at 1.5, 2, 3, 4 V above the breakdown voltage.

6 Journal of Modern Optics 245 Figure 4. Micrographs of SPMMini1020 detector. (The colour version of this figure is included in the online version of the journal.) Figure 5. Photon detection efficiency as a function of wavelength and at 1.5, 2, 3 and 4 V overbias for SPMMini1020 detector. The PDE was measured at 20 C. (The colour version of this figure is included in the online version of the journal.) PCD devices. The SPM detector is mounted in an integrating sphere together with the reference detector and a continuous low level optical signal from a monochromator and white light source is introduced into the sphere. The optical signal is such that the photon flux at the SPM surface is low enough that the probability of the SPM detecting two or more photons within the deadtime of the microcells is very low (photon counting mode). The dark count rate and signal count rates at different wavelengths are then recorded together with the optical signal power as recorded by the large area Newport reference detector. Figure 5 shows the SPMMini1020 PDE as a function of wavelength at a number of different bias values above the breakdown voltage measured at 20 C. The SPM has a peak PDE of 17% at 480 nm and at 4 V above the breakdown voltage Geiger pulse characterisation The output of an SPM consists of a Geiger pulse for each detected photon. Figure 6 shows a screen shot from an oscilloscope of three single Geiger pulses and a double pulse resulting from two microcells firing simultaneously. The timing characteristics of a Geiger pulse are defined by two parameters: the onset time and the recovery time. The onset time of the Geiger pulse is estimated to be of the order of hundreds of picoseconds [15] and requires fast amplifiers and electronics to accurately measure. Using a fast voltage amplifier and a 1 GHz oscilloscope, the measured 10% to 90% onset time of the Geiger pulse is typically 2 5 ns. The 90% to 10% recovery time of the Geiger pulse is set by the RC combination of the quench resistor and the junction and parasitic capacitances and is typically ns.

7 246 A.G. Stewart et al. Figure 6. Oscilloscope screen shot of Geiger pulses from an SPM after external amplification. The timebase is 500 ns per division while the vertical scale is 100 mv per division. Figure 7. Typical single photoelectron spectrum for a 1 mm 2 silicon photomultiplier at room temperature Single photo-electron spectrum The uniform high-gain of the SPM microcells results in a single photo-electron spectrum with well-resolved peaks and gives photon number resolution for short pulses of light. Figure 7 shows the typical single photo-electron spectrum from a 1 mm 2 SPM. The photo-electron spectrum is obtained by repeatedly illuminating the SPM detector with nanosecond flashes of light from a fast green LED. The optical signal results in a current pulse from the SPM composed of an integer number of Geiger pulses corresponding to the number of microcells that fire as a result of detecting photons. The current pulse is then amplified and read into a charge to digital converter (QDC). The output

8 Journal of Modern Optics 247 charge is integrated for a set period or gate to reduce the noise. The gate width is typically 100 ns and is triggered to open by the pulse generator driving the LED. A computer program then generates a histogram of the charge output for many optical pulses from the LED. The resulting histogram consists of a number of wellresolved peaks corresponding to the integer number of microcells that have detected photons for each optical pulse. The first peak of the spectrum, known as the pedestal, is a measure of the electronic noise in the system (amplifier noise etc.) and corresponds to instances when no photons were detected. In such cases, since there are no Geiger pulses only the noise of the electronics will be recorded. The second peak in the spectrum corresponds to a single microcell firing in response to the optical signal. The subsequent peaks correspond to two, three, four, etc., microcells firing in response to the optical signal. The variation in the number of photons detected per signal event is a result of the Poissonian nature of the number of photons and the statistical probability that the SPM detects a photon. The well-resolved peaks of the photo-electron spectrum demonstrate the excellent photon number resolution of the SPM SPM dark count rate Figure 8(a) shows the typical bias dependence of the SPMMini1020 dark count rate. The dark count rate is plotted as a function of overbias for both room temperature and 20 C operation. The dark count rate is a linear function of the bias voltage and decreases by over an order of magnitude when operated at 20 C. The breakdown voltage of a Geiger mode photodiode shifts with temperature and this shift in breakdown voltage must be taken into account when determining the overbias value at different temperatures. Figure 8(b) shows the typical temperature dependence of the breakdown voltage of the microcells of an SPM. The gradient of the plot gives the shift in breakdown voltage per unit temperature. The breakdown voltage decreases by 23 mv per C decrease in temperature [16]. Figure 9 shows a plot of the SPMMini1020 dark count rate as a function of the peak photon detection efficiency (480 nm) at 20 C. The optimum operating bias voltage for a given application is determined by the strength of the signal and the trade-off between the dark count rate and the photon detection efficiency. by the high electric field [17,18]. The photons, which cover a broad spectrum from the visible to the near infrared, can travel to and be absorbed in the active region of a neighbouring microcell resulting in an additional Geiger avalanche event. As the emitted photons can travel to the nearest neighbour microcells in around 0.12 ps, the optical crosstalk process can be considered to happen instantaneously and results in false counts or signal to be added to both the dark count and signal count of the detector. The optical crosstalk probability is defined as the probability that an avalanching microcell will trigger a second microcell to avalanche. The optical crosstalk is a function of the SPM bias voltage and the distance between neighbouring cells. The optical crosstalk probability can be estimated from the ratio of the single and double level count rates. The dark count rate at the single level is measured by setting the oscilloscope trigger threshold at half the peak height of the Geiger pulses while the dark count rate at the double pulse level is measured by setting the oscilloscope trigger threshold at 1.5 times the Geiger pulse peak height. The crosstalk probability can be determined from the plateaus in the plot which represent the dark count rates at the single, double, and triple photo-electron levels. Typical optical crosstalk probabilities are of the order of 2 to 10% depending on the pitch of the microcells and the applied overbias [6]. The optical crosstalk can be significantly reduced by blocking the optical path between adjacent microcells. This can be achieved by etching a trench between the neighboring microcells and filling it with an opaque material. Prototype SPM devices in which the microcells have been optically isolated by a trench have recently been fabricated and tested. Figure 10(a) shows a scanning electron microscope image of the trenches. The prototype devices have the same design as the SPMMini1020, 20 mm square microcells on a 29 mm pitch, but have an optical crosstalk probability of approximately 1% at 2 V above the breakdown voltage. Figure 10(b) shows the room temperature dark count rate as a function of threshold at 2 V above the breakdown voltage, i.e. the count rate at the double photo-electron level is 3 kcps while the count rate at the single photo-electron level is 300 kcps. The ratio of the count rates at the double and single photo-electron levels give an optical crosstalk probability of approximately 1% Optical crosstalk When a PCD undergoes Geiger breakdown, carriers near the junction emit photons as they are accelerated 4. Silicon photomultiplier in photon counting mode As discussed above, one of the main advantages of SPM detectors over PMT detectors is the gain

9 248 A.G. Stewart et al. Figure 8. (a) Dark count rate as a function of overbias for SPMMini1020 detector at room temperature and at 20 C and (b) SPM breakdown voltage as a function of temperature. (The colour version of this figure is included in the online version of the journal.) uniformity of the pulses in response to single photons. As the gain of a PMT in response to a single photon varies widely due to excess noise, complex electronics such as constant fraction discriminators (CFD) are required to precisely time the arrival of a pulse. However, the Geiger pulses of an SPM are near identical in terms of gain, onset time and recovery time. This means that the Geiger pulse arrival time can be accurately determined using simple level thresholding comparator circuitry. This provides a significant advantage over PMT detector based systems in reduced complexity and higher integration. To demonstrate the photon counting capability of an SPM, the output from a 1 mm 2 SPM was amplified and fed into a comparator circuit. The output of the comparator was then sent to a SensL HRMTime time module which records the arrival time of TTL pulses. An external amplification of 137 was used to produce single level Geiger pulses with a peak height of 120 mv. The onset and recovery time of the pulses

10 Journal of Modern Optics 249 Figure 9. SPMMini1020 dark count rate as a function of peak photon detection efficiency (480 nm) at 20 C. (The colour version of this figure is included in the online version of the journal.) was measured to be 6 and 30 ns, respectively. The threshold level of the comparator was then set to 60 mv. The width of the comparator (TTL) pulses was approximately 12 ns for a single Geiger pulse from the SPM. The SPM and HRMTime were then used to detect a heavily attenuated signal from an LED. The LED was driven with a clock pulse with a period of 8 ms and 50% duty cycle. The LED was attenuated using filters before passing into an integrating sphere. The SPM was attached to one of the ports on the sphere. The attenuation and integrating sphere ensure that the optical signal incident on the SPM is uniformly distributed and at the photon counting level. Figure 11(a) shows an oscilloscope screen shot of the TTL pulses from the comparator circuit with the SPM illuminated with the modulated LED optical signal. The signal from the SPM comparator electronics was fed into the STOP input of the HRMTime module. The START input of the HRMTime is connected to a clock signal in phase with the 8 ms signal that drives the LED but at 1 8 the frequency. Hence a START pulse occurs every 8 cycles of the LED driver signal. The HRMTime was operated in Multiscalar/Counter Histogram mode with a time-bin resolution of 27 ps. In this mode the module waits for a START pulse and records all STOP signals by incrementing time-bins in the memory. When a new START pulse arrives the timer is reset and the process is repeated. By running the process continuously, the HRMTime will build up a histogram of the STOP events that occur during the period of the START signal. Figure 11(b) shows a typical histogram produced by the SensL software used for interfacing with the HRMTime module and provides data acquisition and analysis. The histogram consists of eight 4 ms blocks consistent with the incident optical signal from the LED. The SPM as a single photon counter has applications in LIDAR, fluorescence detection and lifetime measurement and microscopy. 5. Conclusion A silicon Geiger-mode photodiode is the basic element of a photon counting detector which must also include either a passive or active quench circuit. Photon counting detectors have numerous applications including microscopy, fluorescence in biotechnology applications, optical time domain reflectometry and LIDAR. A silicon Geiger-mode photodiode can also be the basic element of more complex detection architectures such as a photon counting camera and silicon photomultiplier. A silicon photomultiplier is an array of photon counting microcells connected to a common output. Each microcell consists of PCD connected in series with a quenching element. While SPM detectors do not provide spatial information, they are well-suited to numerous applications including medical imaging and radiation detection. To reduce the dark count rate the dimensions of photon counting devices are typically a few tens of microns. However, this makes directly coupling light on such a small area a more intricate problem and induces losses when coupling to

11 250 A.G. Stewart et al. Figure 10. (a) Scanning electron microscope image of trenches etched between neighbouring microcells of an SPM and (b) room temperature dark count rate at 2 V above the breakdown voltage as a function of threshold position for a prototype SPM with trenches. optical fibres. There is therefore a strong need to produce low dark count, wide area photon counting detectors to improve the coupling efficiency to optic fibres and to reduce the need for complex and costly optics. To address this need, photon counting detectors with dimensions up to 200 mm per edge have been demonstrated and it has been shown that the dark count rate scales with the area of the PCD device. The PCD devices have a peak PDP of 45% at 490 nm and at 4 V above the breakdown voltage. The median room temperature dark count rate of a 200 mm 200 mm PCD at 4 V above the breakdown voltage across a wafer is 77.9 kcps. Silicon photomultiplier detectors, with 1 mm 1 mm active areas, have also been fabricated. These devices consist of 1144 microcells and have a fill factor of 43%. The SPM detectors have excellent photon number resolution at room temperature as demonstrated by the well-resolved peaks of the single photoelectron spectrum and have a peak photon detection efficiency of 17% at 490 nm

12 Journal of Modern Optics 251 Figure 11. Oscilloscope screen shot of TTL pulses from SPM with single photon counting circuitry in response to attenuated and diffuse LED signal. (a) The oscilloscope timebase is 5 ms per division while the vertical scale is 5 V per division and (b) histogram of recorded single photon events from HRMTime module. and at 4 V above the breakdown voltage. At 20 C the dark count rate is approximately 600 kcps at 4 V above the breakdown voltage. SPM detectors can also be used in single photon counting mode. This has been demonstrated using a simple comparator circuit and a SensL HRMTime timing module. References [1] Cova, S.; Ghioni, M.; Lotito, A.; Rech, I.; Zappa, F. J. Mod. Opt. 2004, 51, [2] (accessed Jan 2008). [3] Bellis, S.; Wilcock, R.; Jackson, J.C. Proc. SPIE. 2006, 6068, 60680D [4] Saveliev, V. Nucl. Instr. Meth. A 2004, 518, [5] Stewart, A.G.; Green-O Sullivan, E.; Herbert, D.J.; Saveliev, V.; Quinlan, F.; Wall, L.; Hughes, P.J.; Mathewson, A.; Jackson, J.C. Proc. SPIE. 2006, 6119, 61190A. [6] Stewart, A.G.; Saveliev, V.; Bellis, S.J.; Herbert, D.J.; Hughes, P.J.; Jackson, J.C. IEEE J. Quantum Electron. 2008, 44, [7] Collazuol, G.; Ambrosi, G.; Boscardin, M.; Corsi Dalla Betta, G.F.; Del Guerra, A.; Dinu, N.; Galimberti, M.; Giulietti, D.; Gizzi, L.A.; Labate, L.; Llosa, G.; Marcatili, S.; Morsani, F.; Piemonte, C.; Pozza, A.; Zaccarelli, L.; Zorzi, N. Nucl. Instr. Meth. A 2007, 581,

13 252 A.G. Stewart et al. [8] Gomi, S.; Hano, H.; Iijima, T.; Itoh, S.; Kawagoe, K.; Kim, S.H.; Kubota, T.; Maeda, T.; Matsumura, T.; Mazuka, Y.; Miyabayashi, K.; Miyata, H.; Murakami, T.; Nakadaira, T.; Nakaya, T.; Otono, H.; Sano, E.; Shinkawa, T.; Sudo, Y.; Takeshita, T.; Taguchi, M.; Tsubokawa, T.; Uozumi, S.; Yamaoka, M.; Yamazaki, H.; Yokoyama, M.; Yoshimura, K.; Yoshioka, T. Nucl. Instr. Meth. A 2007, 581, [9] Knill, E.; Laflamme, R.; Milburn, G.J. Nature 2001, 409, [10] McElroy, D.P.; Saveliev, V.; Reznik, A.; Rowlands, J.A. Nucl. Instr. Meth. A 2007, 571, [11] Herbert, D.J.; Stewart, A.G.; Hughes, P.J.; Jackson, J.C. IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS-MIC 2007). [12] Jackson, J.C.; Morrison, A.P.; Phelan, D.; Mathewson, A. IEEE International Electron Devices Meeting 2002, 32, [13] Sze, S.M. Physics of Semiconductor Devices, 2nd ed.; Wiley-Interscience: New York, 1981; pp [14] SPMScint1020 Datasheet, (accessed Jan 2008). [15] Kovaltchouk, D.V.; Lolos, G.J.; Papandreou, Z.; Wolbaum, K. Nulc. Instr. Meth. A 2005, 538, [16] Goetzberger, A.; McDonald, B.; Haitz, R.H.; Scarlet, R.M. J. Appl. Phys. 1963, 34, [17] Jackson, J.C.; Morrison, A.P.; Phelan, D.; Mathewson, A. IEEE Trans. Electron Devices. 2002, 32, [18] Akil, N. IEEE Trans. Electron Devices. 1999, 46,

14