PLEASE SCROLL DOWN FOR ARTICLE

Transcription

1 This article was downloaded by: [Ghioni,] On: 2 April 2009 Access details: Access Details: [subscription number ] Publisher Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: Registered office: Mortimer House, Mortimer Street, London W1T 3JH, UK Journal of Modern Optics Publication details, including instructions for authors and subscription information: Resonant-cavity-enhanced single photon avalanche diodes on double siliconon-insulator substrates Massimo Ghioni a ; Giacomo Armellini a ; Piera Maccagnani b ; Ivan Rech a ; Matthew K. Emsley c ; M. Selim Ünlü d a Dipartimento di Elettronica e Informazione, Politecnico di Milano, Milano, Italy b IMM-CNR sezione di Bologna, Bologna, Italy c Analog Devices Inc., Wilmington, MA, USA d Department of Electrical and Computer Engineering, Boston University, Boston, MA, USA First Published:January2009 To cite this Article Ghioni, Massimo, Armellini, Giacomo, Maccagnani, Piera, Rech, Ivan, Emsley, Matthew K. and Ünlü, M. Selim(2009)'Resonant-cavity-enhanced single photon avalanche diodes on double silicon-on-insulator substrates',journal of Modern Optics,56:2, To link to this Article: DOI: / URL: PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

2 Journal of Modern Optics Vol. 56, Nos. 2 3, 20 January 10 February 2009, Resonant-cavity-enhanced single photon avalanche diodes on double silicon-on-insulator substrates Massimo Ghioni a *, Giacomo Armellini a, Piera Maccagnani b, Ivan Rech a, Matthew K. Emsley c and M. Selim U nlu d a Dipartimento di Elettronica e Informazione, Politecnico di Milano, Milano, Italy; b IMM-CNR sezione di Bologna, Bologna, Italy; c Analog Devices Inc., Wilmington, MA, USA; d Department of Electrical and Computer Engineering, Boston University, Boston, MA, USA (Received 31 January 2008; final version received 11 June 2008) We report the first resonant-cavity-enhanced single photon avalanche diode (RCE SPAD) fabricated on a reflecting silicon-on-insulator (SOI) substrate. The substrate incorporates a two period distributed Bragg reflector fabricated using a commercially available double SOI process. The RCE SPAD detectors have peak photon detection efficiencies ranging from 42% at 780 nm to 34% at 850 nm and an excellent photon timing resolution of 35 ps full width at half maximum. Despite the higher defectivity of double SOI substrates compared to standard silicon substrates, RCE SPADs with 20 mm diameter exhibit a fairly low dark count rate (DCR) of 3500 cs 1 at room temperature and a yield of 80%. A DCR less than 50 cs 1 can be attained with these detectors by reducing the temperature down to 15 C, while keeping the total afterpulsing probability below 9% with a dead-time of 80 ns. Keywords: single photon avalanche diode; resonant cavity enhanced; silicon on insulator; photon counting; time-correlated single photon counting 1. Introduction There is nowadays a widespread and steadily growing interest in single photon detectors, driven by the need for ultimate sensitivity in various scientific and industrial applications such as fluorescence spectroscopy in life and material sciences, quantum cryptography and computing, time-of-flight ranging and imaging, particle sizing etc. In the last four decades, single photon counting (SPC) and timecorrelated single-photon counting (TCSPC) techniques have been developed relying on photomultiplier tubes (PMTs), that is, vacuum tube detectors with high internal gain. PMTs offer wide sensitive area (cm 2 ) and they also attain picosecond time resolution with micro-channel plate (MCP) models [1]. However, due to the intrinsic limits of the photocathodes they have moderate or low photon detection efficiency (PDE), particularly in the red and near-infrared spectral range. The PDE of conventional bialkali and multialkali photocathodes reaches 20 25% between 400 and 500 nm [1]. GaAs photocathodes have an improved PDE in the red and near infra-red (NIR) range (15% up to 850 nm), but they can introduce a transit time spread of the order of 100 to 150 ps [2]. In more recent years, silicon single-photon avalanche diodes (SPADs) have emerged as a solid state alternative to PMTs [3 5]. Photon detection modules employing low-voltage SPAD detectors with diameter up to 100 mm have become commercially available from various manufacturers. Such detectors are fabricated in planar technology on silicon substrates and offer the typical advantages of solid state devices (miniaturization, ruggedness, low voltage, low power, low cost, etc.) along with photon timing resolution better than 35 ps full width at half maximum (FWHM) and high photon detection efficiency in the visible range (50% at 550 nm) [5]. The PDE of planar SPADs typically varies from 25% to 12% in the NIR range from 700 to 850 nm, being much higher than that of photomultiplier tubes equipped with standard S20 or S25 photocathodes. Nevertheless, a better PDE performance in the NIR range would be highly desirable in various fields, as for instance high bitrate, short-wavelength quantum key distribution [6] and in vivo molecular imaging for life sciences applications [7]. The most straightforward way of achieving higher PDE in SPAD detectors is to increase the depletion region thickness. However, the need of ensuring a high *Corresponding author. ghioni@elet.polimi.it ISSN print/issn online ß 2009 Taylor & Francis DOI: /

3 310 M. Ghioni et al. electric field strength (above the critical value for breakdown) over a thick depletion layer leads to high operating voltage, causing substantial power dissipation and self-heating problems. Furthermore, a thick depletion region would make the photonassisted propagation mechanism to dominate the jitter of the avalanche current leading-edge, resulting in worse photon timing performance [8]. Perkin Elmer Optoelectronics manufactures a popular Single Photon Counting Module (SPCM) based on a proprietary SPAD detector having a thick depletion layer (20 25 mm) [9]. These modules have very high PDE (above 50% between 500 and 800 nm) and low noise, but the photon timing resolution is typically a few hundreds of ps FWHM, whereas sub-100 ps values would be desirable in many applications. Therefore, it is desirable to enhance the PDE without increasing the absorption layer thickness. A Fabry Pe rot cavity can be exploited to enhance the optical field inside the SPAD detector at resonant wavelengths [10]. Such a resonant cavity can be formed using a buried reflector and the air/semiconductor top interface [11,12]. This approach enables higher PDE at the same depletion region thickness, thus avoiding adverse effects on photon timing resolution and power dissipation. Resonant-cavity-enhanced (RCE) photodetectors have been the focus of extensive research over the past decade in the design of high bandwidth-efficiency product p-i-n and avalanche photodetectors working in linear mode [10]. However, no SPAD detector with resonant structure has been reported so far. In this paper, we report the first planar SPAD detector grown atop a highly reflective two-period silicon-on-insulator (SOI) substrate. We demonstrate that these RCE SPADs provide a substantial PDE improvement in the NIR, while keeping excellent photon timing resolution and low operating voltage. In Section 2, we briefly describe the fabrication process of RCE SPAD detectors. In Section 3, we present and discuss experimental data obtained for RCE SPAD detectors pertaining to photon detection efficiency, dark count rate, afterpulsing, and time resolution. In Section 4, we summarize our results. Figure 1. Schematic cross-section of the RCE SPAD detector. (The color version of this figure is included in the online version of the journal.) tuned to achieve a reflectance in excess of 90% around 850 nm [11]. On top of these reflecting wafers, double epitaxial SPADs with active area diameter of 8, 20 and 50 mm were fabricated using the planar process described in [13]. Figure 1 shows a schematic crosssection of the RCE SPAD detector. The active n þ p junction is built in the upper low-doped p-epilayer. The buried p þ epilayer provides a low-resistance path to the side ohmic contact. The total thickness of the epilayers is about 5 mm. A boron implantation in the central part of the n þ p junction defines the high electric field region, that is, the active area of the detector. A deep, highly doped n þ diffusion provides electrical isolation between adjacent SPADs and acts as a gettering region for transition metals [14]. The silicon substrate acts as a control gate to induce an inversion or accumulation layer at the buried Si/SiO 2 interface. Al contacts were deposited by sputtering, and patterned photolithographically. The active area of SPAD detectors was coated with a single, 100 nm-thick SiO 2 layer to prevent reliability problems. The reflectivity of the coated surface is about half of that of bare silicon/air surface at 850 nm (14% instead of 32%), resulting in a lower finesse of the resonant cavity, that is, a lower enhancement of the PDE. Control SPAD detectors made on conventional n-type substrates were fabricated in the same batch for comparison purposes. 2. Device fabrication RCE SPAD fabrication started from 4 inch reflecting silicon wafers having an epitaxy ready single crystalline surface. These wafers incorporate a two period distributed Bragg reflector (DBR) fabricated using a commercially available double-soi process. The thickness of the DBR layers (437 nm for the SiO 2 layers and 174 nm for the Si layers) was specifically 3. Device characterization An average breakdown voltage of 32.5 V with a standard deviation of 0.6 V was measured at room temperature for RCE SPADs with 50 mm diameter randomly distributed on the wafer. Similar results were obtained for the 8 and 20 mm diameter devices, thus confirming the good uniformity of the fabrication process.

4 All experimental measurements described in the following sections were performed by operating the SPAD detectors at 5 V excess bias voltage with an external active quenching circuit (AQC) [15,16]. Journal of Modern Optics Photon detection efficiency Photon detection efficiency measurements were performed using a dedicated set-up including a halogen light source, monochromator, integrating sphere and a reference silicon photodetector. The PDE of RCE SPAD detectors was measured in two different conditions, namely with an inversion layer (INV) or an accumulation layer (ACC) formed at the buried Si/SiO 2 interface. These two conditions were easily obtained by applying a suitable voltage between substrate and anode contacts, that is, about 10 V for accumulation and þ10 V for inversion. Figure 2 shows the PDE of control and RCE SPADs as a function of wavelength, measured at 5V excess bias voltage. As expected no resonance peaks are observable in RCE SPADs up to 650 nm, since either most of the incident photons are absorbed before reaching the buried mirror or the buried mirror itself has a low reflectivity [11]. In contrast, resonant enhancement of PDE occurs between 750 and 950 nm where the buried mirror has its maximum reflectivity and the absorption lengths are greater than 10 mm. A remarkable PDE of 34% was measured at 850 nm for RCE SPADs with accumulation layer. The PDE reduces to 23% at the same wavelength by inducing an inversion layer at the buried Si/SiO 2 interface. This figure, however, still favorably compares with the PDE of control SPAD detectors (10%). The noticeable difference between the PDE of RCE SPADs with INV and ACC layers is due to minority carriers photogenerated in the neutral region beneath the active junction (see Figure 3). If an INV layer is formed, the reverse biased junction between n-isolation and p-epilayer is extended all over the buried Si/SiO 2 acting as a buried pn junction. Minority carriers diffusing toward the buried reflector are captured by the reverse-biased buried junction, giving no photon signal. A similar loss of minority carriers happens in conventional SPAD detectors, where the p-type epitaxial layers are grown over a n-type substrate. In contrast, if an ACC layer is formed there is no buried junction and the neutral region extends all the way down to the buried Si/SiO 2 interface, being therefore thicker than that of INV RCE SPADs. Furthermore, minority carriers diffusing toward the buried reflector may be backscattered and eventually captured by the active junction, where they may trigger an avalanche pulse. In practice, there is almost no loss Figure 2. Photon detection efficiency as a function of wavelength for control SPAD detector and RCE SPAD detector with inversion (INV) or accumulation (ACC) layer formed at the Si/SiO 2 buried interface. (The color version of this figure is included in the online version of the journal.) of minority carriers photogenerated in the neutral region, resulting in a higher PDE. It must be noted that this mechanism enhances the PDE at all wavelengths for which the optical absorption in the neutral layer is not negligible (i.e nm), independently of the optical reflectivity of the buried mirror. We may conclude that the PDE enhancement in INV RCE SPAD detectors is basically due to the optical resonance alone, while in ACC RCE SPAD detectors the PDE is enhanced by a combination of: (i) optical resonance, (ii) enhanced collection of diffusing electrons and, (iii) thicker neutral region. In order to assess the uniformity of the PDE, several measurements were performed on devices taken from different regions of the wafer. PDE data show a relative standard deviation of 5% and 6% at 550 nm and 850 nm wavelength, respectively. The fairly small PDE dispersion is mainly due to the non-uniform thickness of both the top oxide layer and the bottom DBR layers. In addition, we observed a small dispersion of the resonance peak position around 850 nm. This dispersion is comparable with the resolution limits of the experimental apparatus (5 nm), and it can be ascribed to the slightly non-uniform thickness of the double epitaxial layer over the wafer. The inherent wavelength selectivity of RCE SPADs is not an issue in fluorescence spectroscopy applications, since most of the commonly used fluorophores exhibit a rather broad emission spectrum (FWHM 50 nm) compared to the width and the spacing of resonance peaks. Therefore, the effective PDE of the RCE SPAD detector is given by the weighted local average of the actual PDE. In contrast, wavelength selectivity might be a drawback in applications where the incoming light has a narrow spectrum, since strict control of the epilayer thickness would be required to

5 312 M. Ghioni et al. Figure 3. Cross-section of the RCE SPAD detector showing the effect of the inversion layer (a) and of the accumulation layer (b) on diffusing minority electrons. (The color version of this figure is included in the online version of the journal.) accurately set the resonance wavelength. To avoid the need for complex spectral tuning procedures [11] and yet still benefit from the RCE structure, one can coat the top surface of the detector with an antireflection coating which will result in a two-pass detector where the light enters the photodetector and reflects off the buried mirror, thus doubling the effective absorption width. Table 1. Number of devices tested and yield of RCE and control SPAD detectors with different active area diameter. Diameter Devices tested Yield 8 mm % 20 mm % 50 mm % 50 mm (control) % 3.2. Yield and dark count rate Preliminary DCR measurements were performed at room temperature. The dead-time of the AQC was set to 300 ns in order to reduce afterpulsing effects to a negligible level [15]. We found out that several devices had a DCR so high as to cause saturation of the AQC count rate (i.e. DCR 4 1/dead-time ¼ 3.3 Mc s 1 ). These devices can not be used as SPADs and they were designated as not working SPADs. All other tested SPADs had a DCR well below the saturation limit of the AQC. These devices were designated as working SPADs. To determine the quality of the fabrication process we introduced a yield parameter given by: working SPADs yield ¼ tested SPADs : ð1þ Table 1 shows the yield of RCE and control SPAD detectors with different active area diameter. A 100% yield is achieved for the smallest RCE SPAD detectors.

6 Journal of Modern Optics mm diameter SPAD detectors still exhibit a good yield of 80%, while the yield of the 50 mm diameter SPADs drop to 30%. For comparison, the yield of 50 mm diameter control SPADs is 100%. We argue that the anomalous high DCR of not working SPADs is due to the occurrence of microplasmas [17,18] within the active area of the device. Microplasmas produce high field peaks localized in submicrometer zones resulting in a strongly enhanced field emission. It has been shown [19,20] that microplasmas are correlated with structural defects, most likely dislocations. According to Camassel et al. [21], dislocations can be induced in SOI layers by strain relaxation after high temperature processes. The stress originates from the considerable difference in thermal expansion coefficients between Si and SiO 2 ( K 1 and K 1 respectively at room temperature). DCR data for working SPADs are presented in the form of inverse cumulative distribution functions in Figure 4. The typical DCR is 450, 3500 and 100,000 cs 1 for RCE SPADs having 8, 20 and 50 mm active area diameter, respectively. The dotted line in Figure 4 represents the DCR distribution for the 50 mm diameter control SPADs. The typical DCR of control SPADs is between one and two orders of magnitude lower than that of RCE SPADs with the same active area diameter, showing that the defectivity of double SOI substrates is higher than that of standard Si substrates. It is well known that the DCR in SPAD devices is due to thermal and field-enhanced generation of carriers via deep energy levels localized in the depletion region. Transition metal impurities are the most common source of deep levels [22]. Metal contamination may occur during silicon handling, high-temperature heat treatments or ion implantations. As unintentional contaminants Fe, Cu, Ti, Mo or Ni are usually found in bulk silicon in concentrations of cm 3 [23]. The worse DCR performance of RCE SPADs is likely due to a higher concentration of metal impurities in these devices. This can be ascribed to the buried SiO 2 layers that act as a barrier preventing the diffusion of metal impurities from the top silicon layer into the bulk substrate [24,25]. In order to check the validity of this assumption, we performed DCR measurements on 20 mm diameter RCE SPADs having the highly-doped n þ diffusion placed at different distances from the active area. This distance was varied from 70 mm, corresponding to the default spacing in devices used to obtain data in Figure 4, down to 30 mm. As shown in Figure 5, the DCR reduces by a factor of 4 by decreasing the distance between the active area and the n þ diffusion acting as a gettering region, confirming that medium-fast-diffusing metal impurities play a central role in determining the DCR performance. Figure 4. Percentage of RCE and control SPAD detectors (horizontal scale) found within a given limit of the individual dark counting rate (vertical scale) measured at room temperature. The curve parameter is the diameter of the SPAD active area. (The color version of this figure is included in the online version of the journal.) Figure 5. Percentage of 20 mm diameter RCE SPAD detectors (horizontal scale) found within a given limit of the individual dark counting rate (vertical scale) measured at room temperature. The curve parameter is the distance between the SPAD active area and the highly doped n þ diffusion acting as a gettering region. (The color version of this figure is included in the online version of the journal.) Similar measurements performed on control SPADs having the same diameter showed that the DCR reduces by a factor of 1.5 by reducing the active-togetter region spacing from 70 to 30 mm. The weaker reduction factor can be explained by observing that metal impurities introduced at the silicon surface can freely diffuse into the thick (500 mm) substrate instead of being confined within the thin (5 mm) SOI layer. This makes the presence of lateral gettering sinks less effective. Figure 6 shows the DCR as a function of temperature for a 20 mm diameter RCE SPAD detector. The breakdown voltage varies between 31 and 33.3 V over the temperature range from 50 C to 25 C.

7 314 M. Ghioni et al. Figure 6. Dark count rate as a function of temperature for a20mm diameter RCE SPAD. (The color version of this figure is included in the online version of the journal.) Measurements were performed at a constant excess bias voltage of 5 V. The DCR decreases almost exponentially with temperature: at a relatively low temperature of 15 C (easily obtainable with a thermoelectric cooler) the typical DCR is about 50 cs 1. The steep decrease of the DCR with temperature ( a factor of 10 every 17 C) shows that the dominant generation centers responsible for the DCR are displaced from the midgap by about E G /4 [26] Afterpulsing Figure 7 shows the probability density in time for the occurrence of an afterpulse after an initial avalanche pulse for a 20 mm diameter RCE SPAD detector operated at 5V excess bias voltage. Measurements were performed at different temperatures by using the time-correlated carrier counting (TCCC) technique [27]. An AQC with a dead-time of 80 ns was exploited for the measurements. It is worth noting that the probability of having an afterpulse quickly decays, being negligible after about 1 ms from the initial avalanche pulse. The total afterpulsing probability, i.e. the integral over the entire probability density curve, is about 2% at room temperature and it increases up to 8.3% at 15 C. This increase is due to the dependence of the trap emission lifetime on temperature and to the presence of a fixed dead-time. At lower temperatures, the emission lifetime of a given trap gets longer: accordingly, the probability that a carrier is emitted after the dead-time (thus being able to trigger an avalanche) gets higher Time resolution Time resolution measurements were performed in a conventional time-correlated single photon counting (TCSPC) setup by using an ultrafast laser diode (Antel Figure 7. Afterpulsing probability density for a 20 mm diameter RCE SPAD, measured at different temperatures. The curve parameter P ap is the total afterpulsing probability obtained by integrating over the entire probability density curve. (The color version of this figure is included in the online version of the journal.) Figure 8. Time response of RCE and control SPAD detectors to a laser diode emitting 15 ps optical pulses at 820 nm. Tested SPADs have an active area diameter of 50 mm and are operated at 5 V excess bias voltage. (The color version of this figure is included in the online version of the journal.) MPL-820 laser module) emitting 15 ps FWHM optical pulses at 820 nm wavelength. Figure 8 shows the time response of 50 mm diameter RCE and control SPADs. For all devices, the time resolution FWHM is about 35 ps. The control and RCE SPAD with INV layer show clean exponential diffusion tails [13] with similar lifetimes (280 and 330 ps, respectively), whereas the RCE SPAD with ACC layer show a second exponential component in the diffusion tail, with a lifetime of 2.6 ns. According to Figure 3, the RCE SPAD with INV layer has a neutral region bounded at the top and bottom sides by depletion regions. As shown in [8], the diffusion current at the edge of the neutral region is given by: J n ðtþ ¼ 4qn od n w X þ1 k¼0 exp ð2k þ 1Þ2 p 2 D n w 2 t, ð2þ

8 Journal of Modern Optics 315 where n o is the electron concentration at time t ¼ 0, D n is the electron diffusion coefficient and w is the width of the neutral region. The first term in the series has a time constant d ¼ w 2 /( 2 D n ), which determines the slowest decay constant of the tail in the time response of SPAD detectors. Conversely, in the RCE SPAD with ACC layer the neutral region is bounded at the bottom side by the accumulated Si/SiO 2 interface. As a consequence: (i) the neutral region width w* is larger than w and, (ii) photogenerated electrons can make a double pass through the neutral zone after being backscattered at the buried interface. Based on simple symmetry arguments, we concluded that Equation (2) is still valid for describing the diffusion current at the edge of the neutral region, provided that the width w is replaced with an effective width w eff ¼ 2w*. The slowest component of the diffusion tail is therefore modeled with a simple exponential decay with lifetime d ¼ (w eff ) 2 / ( 2 D n ) ¼ (2w*) 2 /( 2 D n ). Widths w and w* were estimated by fitting experimental data with D n ¼ 20 cm 2 s 1, obtaining 2.6 and 3.6 mm, respectively. The faster exponential decay visible in the time response curve of the ACC RCE SPAD is well described by the second term of the series in Equation (2). This term has a time constant d ¼ (2w*) 2 /(9 2 D n ) ¼ 300 ps, which is in satisfactory agreement with the measured value of 330 ps. 4. Conclusion We presented the first RCE SPAD detectors with a buried distributed Bragg reflector fabricated by means of a commercially-reproducible double SOI technique. Compared to single-pass SPAD detectors, RCE SPADs provide a two- to three-fold improvement in photon detection efficiency in the near infrared range between 750 and 950 nm, while maintaining excellent photon timing resolution and low operating voltage. The fabricated RCE SPADs have peak photon detection efficiencies ranging from 42% at 780 nm to 34% at 850 nm along with a time resolution of 35 ps FWHM. Although the defectivity of double SOI substrates is higher than that of standard Si substrates, RCE SPADs with 20 mm diameter exhibit a fairly low dark count rate of 3500 c s 1 at room temperature and a yield of 80%. A DCR less than 50 c s 1 can be attained with these detectors by reducing the temperature down to 15 C, while keeping the total afterpulsing probability below 9% with a dead-time of 80 ns. We showed that a substantial improvement in DCR performance is obtained by reducing the distance between the SPAD active area and the gettering region. RCE SPAD detectors are suitable for demanding photon counting applications where both high photon detection efficiency and picosecond time resolution are required. Advances in custom engineered substrates are opening the way to more sophisticated RCE SPADs (e.g. back-illuminated SPADs on relaxed SiGe-on-insulator substrates) with enhanced spectral responsivity in the NIR up to 1.3 mm. Acknowledgements This work was supported by the European Commission FP6, Information Society Technologies (NANOSPAD project IST-NMP and SECOQC project IST ). References [1] Hamamatsu Photonics K.K. com (accessed Jan 21, 2008). [2] Becker, W. Advanced Time-Correlated Single Photon Counting Techniques; Springer: Berlin, [3] Dautet, H.; Deschampes, P.; Dion, B.; MacGregor, A.D.; MacSween, D.; McIntyre, R.J.; Trottier, C.; Webb, P. Appl. Opt. 1993, 32, [4] Cova, S.; Longoni, A.; Andreoni, A. Rev. Sci. Instrum. 1981, 52, [5] Ghioni, M.; Gulinatti, A.; Rech, I.; Zappa, F.; Cova, S. IEEE J. Sel. Topics in Quantum Electron. 2007, 13, [6] Gordon, K.J.; Fernandez, V.; Townsend, P.D.; Buller, G.S. IEEE J. Quant. Electron. 2004, 40, [7] Ntziachristos, V.; Ripoll, J.; Wang, L.V.; Weissleder, R. Nature Biotechnology. 2005, 23, [8] Spinelli, A.; Lacaita, A. IEEE Trans. Electron Devices. 1997, 44, [9] PerkinElmer Optoelectronics. perkinelmer.com (accessed Jan 21, 2008). [10] U nlu, M.S.; Strite, S. J. Appl. Phys. 1995, 78, [11] Emsley, M.K.; Dosunmu, O.; U nlu, M.S. IEEE J. Select. Topics Quantum Electron. 2002, 8, [12] Emsley, M.K.; Dosunmu, O.; U nlu, M.S. IEEE Photon. Techn. Lett. 2002, 14, [13] Lacaita, A.; Ghioni, M.; Cova, S. Electron. Lett. 1989, 25, [14] Myers, S.M.; Seibt, M.; Schro ter, W. J. Appl. Phys. 2000, 88, [15] Cova, S.; Ghioni, M.; Lacaita, A.; Samori, C.; Zappa, F. Appl. Opt. 1996, 35, [16] Zappa, F.; Lotito, A.; Giudice, A.C.; Cova, S.; Ghioni, M. IEEE J. Solid-State Circuits. 2003, 38, [17] Rose, D.J. Phys. Rev. 1957, 105, [18] Haitz, R.H.; Goetzberger, A.; Scarlett, R.M.; Shockley, W. J. Appl. Phys. 1963, 34, [19] Chynoweth, A.G.; Pearson, G.L. J. Appl. Phys. 1958, 29,

9 316 M. Ghioni et al. [20] Lesniak, M.; Holt, D.B. J. Mat. Sci. 1987, 22, [21] Camassel, J.; Falkovsky, L.A.; Planes, N. Phys. Rev. B 2000, 63, [22] Weber, E.R. Appl. Phys. A 1983, 30, [23] Graff, K. Metal Impurities in Silicon-device Fabrication, 2nd ed.; Springer-Verlag: Berlin, [24] Istratov, A.A.; Väino la, H.; Huber, W.; Weber, E.R. Semicond. Sci. Technol. 2005, 20, [25] Hong, S.Q.; Wetteroth, T.; Shin, H.; Wilson, S.R.; Werho, D.; Lee, T.-C.; Schroder, D.K. Appl. Phys. Lett. 1997, 71, [26] Ghioni, M.; Gulinatti, A.; Maccagnani, P.; Rech, I.; Cova, S. SPIE Optics East, Advanced Photon Counting Techniques, Vol. 6372, Boston, October 1 4, [27] Cova, S.; Lacaita, A.; Ripamonti, G. IEEE Electron. Dev. Lett. 1991, 12,