Silicon photomultipliers and their bio-medical applications

ARTICLE IN PRESS

Nuclear Instruments and Methods in Physics Research A 571 (2007) 130–133 www.elsevier.com/locate/nima

Silicon photomultipliers and their bio-medical applications Eugene Grigorieva,b,, Alexander Akindinova, Marco Breitenmoserb, Stefano Buonoc, Edoardo Charbond, Cristiano Niclassd, Iris Desforgesb, Roberto Roccac a

Institute for Theoretical and Experimental Physics (ITEP), B.Cheremushkinskaya 25, Moscow 117218, Russian Federation b Forimtech S.A., Route de Malagnou 32, Geneva 1208, Switzerland c Advanced Accelerator Applications S.A (AAA), 20 rue Diesel, 01630 St. Genis Pouilly, France d Ecole Politechnique Federale de Lausanne (EPFL), AQUA Group, Lausanne 1015, Switzerland Available online 7 November 2006

Abstract Single Photon Avalanche Diodes (SPADs) have been used for photon counting since the 1960s, but only in the recent decade multipixel structures based on SPAD—arrays and silicon photomultipliers have been developed. These devices are finding more and more applications in many fields, where detection of light at the level of a single photon is needed. Due to their exclusive properties (fast response, low operating voltage, single photon sensitivity at room temperature, extremely high gain, stability, compactness, robustness and low price), such sensors are successfully replacing traditional vacuum photomultipliers in many devices. The paper briefly describes the state of the art and suggests some new applications in biology and medicine. r 2006 Published by Elsevier B.V. Keywords: SPAD arrays; SiPM; Single photon detection; Time-resolved fluorescence detection

1. Introduction: SPAD operating principle Single Photon Avalanche Diodes (SPADs) have been known and extensively used as single channel photon counters for decades [1]. The principle of operation is simple. A p–n junction is designed in such a way that the electric field reaches breakdown at relatively low voltage applied to the diode (20–50 V). When the operating voltage is set above breakdown, each charge carrier reaching the junction triggers a self-sustaining avalanche multiplication process, similar to Geiger–Mueller process in gases. It can be stopped only by reducing the applied voltage down to breakdown value. This is done by a series resistor in the circuit. The voltage drop on the resistor quenches the avalanche. Consequently, the amplitude of the signal equals to the product of the pixel capacitance and the over-voltage value. In ‘‘passive recharge’’ mode the recharging current flows through the same series resistor, thus recharging the diode in about 100–200 ns. This time Corresponding author. Institute for Theoretical and Experimental Physics (ITEP), B.Cheremushkinskaya 25, Moscow 117218, Russian Federation. Tel.: +41 22 840 0660; fax: +41 22 735 5288. E-mail address: [email protected] (E. Grigoriev).

0168-9002/$ - see front matter r 2006 Published by Elsevier B.V. doi:10.1016/j.nima.2006.10.046

can be reduced to 30–40 ns by introduction of an ‘‘active recharge’’ circuit. This scheme is used in SPAD arrays [2] in order to minimize the dead time. With this limitation SPADs can be used for counting photons, the frequency of signals being proportional to the intensity of incoming light. In order to detect multiple photons arriving simultaneously or distributed in space, for instance—to measure intensity of scintillation pulse, or to perform imaging or spectroscopy—one needs a multi-SPAD structure, either an SPAD array or Silicon Photomultiplier (SiPM), which is sometimes called Multi-pixel Avalanche Photo-Diode (MAPD) or Avalanche Photo-Diode operating in Geiger mode (APDG). 2. Two families of multi-SPAD structures SPAD arrays and SiPMs use the same basic element, but differ in architecture. The former have active recharge and individual pixel addressing circuit included in every pixel of the matrix, while the latter—only individual quenching resistor per pixel and one common electrode for signal readout. As a result, SiPM has high fill factor and dynamic range, but also high dark count rate and optical cross-talk.

ARTICLE IN PRESS E. Grigoriev et al. / Nuclear Instruments and Methods in Physics Research A 571 (2007) 130–133

SPAD arrays have poor fill-factor (1–6%) but a very low optical cross-talk. It is more compatible with standard CMOS process than SiPM, which need special measures for inter-pixel optical isolation, for instance—shown in Fig. 1 trenches in MESA-technology [3]. Typical characteristics of currently available [4–6] SiPMs are summarized in Table 1. 3. SPAD arrays for time-of-flight measurements Until recently, fully integrated SPADs have not been actively pursued. Quenching and recharge circuits have been typically implemented as external ancillary circuits [7]. Due to parasitic components introduced in this solution a higher number of carriers are involved in an avalanche discharge, thus degrading performance. High number of trapped charges and the intensity of electroluminescence increase the probability of after-pulses and cross-talk [1]. Only recently, researchers have successfully integrated SPADs in CMOS technology [8]. The level of miniaturization reached in more recent designs has enabled a significant reduction in the parasitic capacitance of the diode. As a result, Geiger mode of operation has become a viable option even with very large SPAD arrays. This option has been actively pursued since 2003 at EPFL where several SPAD arrays have been demonstrated in a number of CMOS technologies, including deep-submicron processes [2].

131

Recently, over 1 k-pixel arrays have been proposed for imaging applications where very high precision timing measurement is essential. Time-of-flight (TOF) rangefinders are an example. In order to achieve millimetric depth accuracy in three-dimensional (3-D) imaging, a timing accuracy of a few picoseconds is necessary, while SPADs achieve a timing jitter of about 50 ps. In order to ensure the desired accuracy, a pulsed laser with a repetition rate of several tens of megahertz was used, in combination with averaging techniques, Fig. 2 shows the photomicrograph of a 32  32 SPAD pixel array fabricated in 0.8 um CMOS technology and designed for 3-D imaging. Fig. 3 shows a plot of the depth map of a face measured with this array by illuminating the subject with an intentionally uncollimated laser beam. After passing through an objective, the photons reflected by all unoccluded points on the surface of the subject, impinge upon the surface of the detector at slightly different times. TOF, and thus depth, is computed for each pixel independently, thanks to an architecture that enables to random pixel access across the array. Table 2 lists some of the performance measures of an SPAD and of the overall rangefinder. The total power dissipation of the device in operation is less than 6 mW with an average optical emitted power less than 1 mW and an optical range of sensitivity 420–750 nm. 4. Scintillation detectors based on SiPM Due to high dark count rate, optimal size of SiPM is 0.5–1.0 mm2; therefore it performs best being coupled to an

Fig. 1. Layout of one pixel of SiPM of CPTA. (Center for Perspective Technology and Apparatus, Moscow, Russia.)

Table 1 SPAD array performance summary Performance

Symbol

Photon detection probability Dark count rate Overall timing uncertainty Depth accuracy Dead time Cross-talk Distance range Repetition rate

Z DCR s(t) s(d) td D fR

Min

Typ

Max

Unit

26

% Hz ps mm ns % m MHz

350 350 1.3 40 0.005 3.75 40

Fig. 2. Photomicrograph of 32  32 SPAD pixel array.

ARTICLE IN PRESS 132

E. Grigoriev et al. / Nuclear Instruments and Methods in Physics Research A 571 (2007) 130–133

for cancer diagnostic and surgery [12]. Schematic layout of the device is shown in Fig. 4. The probe according to the design has several novelties: utilization of SiPM (1) directly coupled to the scintillating fiber; additional back-side flat detector (4) for background reduction. Fiber diameter should be smaller than the range of the particle—in this case, as shown in Fig. 5, the device obtains the feature of ‘directional sensitivity’. This method of ‘‘active shielding’’ allows reducing significantly weight of traditional passive shield of the device. Laboratory tests with FDG (FluoroDeoxyGlucose F18) have shown very high sensitivity of the probe. 5. Single photon bio-imaging

Fig. 3. Depth map obtained with SPAD array.

Table 2 SiPM performance summary Operating voltage Power consumption Gain Photon detection efficiency (520 HM) Timing resolution (single photon) Typical sensitive area Typical dynamic range (number of pixels) Operating temperature Dark count rate (at 300 K) Sensitivity to magnetic field Sensitivity to ionizing particles Pixel recharge time Fill factor

20–60 V 50 mW/mm2 105 107 25% 120 ps 1 mm2 1000/mm2 300 K 105–106/mm2 s Very low Low 100–200 ns 50–80%

optical fiber, micro-lens or Micro Electro-Mechanical Systems (MEMS). The problem of high dark count rate is normally solved either by time correlation of several signals (laser pulse, coincidence channel, etc.) or by signal discrimination at the level of several photo-electrons (though dark counts have single-electron origin, higher amplitudes are present due to optical cross-talk). The first applications, demonstrating superior performances of SiPM compared to standard vacuum PM, involved highenergy physics in scintillation detectors for tracking and calorimetry of charged particles [9–11]. In all cases, the SiPM is optically coupled to 1 mm either scintillating fiber or wavelength one embedded in bulk scintillator. Similar approach is pursued by AAA-Forimtech collaboration in development of intra-operative beta/gamma probe

Bioimaging research activity has recently focused on applications, where assay, bio-luminescence and photon scattering methods are among the most efficient ones. In assay methods, one may determine the potential on a given cell in vitro or in vivo by means of Voltage-Sensitive Dyes (VSDs), for example, to evaluate the neural activity in the brain [13]. The challenge in these methods is generally to detect small photo-signals while maintaining high contrast in the presence of massive background illumination. Fluorescence is utilized in many techniques, such as Fluorescence Correlation Spectroscopy (FCS) [14], single and multi-photon Fluorescence Lifetime Imaging (FLIM) [15,16], Fo¨rster Resonance Energy Transfer (FRET) [17], etc. All these methods need detectors with very high timing resolution or, conversely, low time uncertainty. Bioluminescence methods on the contrary, exploit the fact that light is emitted as the result of a chemical reaction [18]. Extremely low emission intensity, typically in the mlux range, need be detected with sufficient signal-to-noise ratio. Speed is also a concern if complex, possibly parallel, evaluations are needed in a reasonable time frame, as it is

Fig. 4. Schematic layout of SiPM-based intra-operative probe.

ARTICLE IN PRESS E. Grigoriev et al. / Nuclear Instruments and Methods in Physics Research A 571 (2007) 130–133

0.2 mm fiber

tional photo-detectors and miniaturization of bio-medical devices. Small size, low voltage and power consumption, room temperature operation, single-photon sensitivity, sub-nanosecond resolution and low cost open possibilities of a breakthrough in developing new procedures using compact multi-channel systems for bio-medical and pharmaceutical research.

1.0 mm fiber

20.00 18.00 16.00 Spatial resolution, cm

133

14.00 12.00 10.00

Acknowledgments

8.00 6.00

This research was supported, in part, by the Swiss National Science Foundation and in part by the Interreg IIIa (France-Suisse) Program (Rhone-Alpes/Bassin Lemanique, Grant 49/BL/9.3/3).

4.00 2.00 0.00 0.00

50.00

100.00

150.00

200.00

Energy threshold, keV Fig. 5. Position resolution of fiber probe versus energy cut.

the case for example in DNA sequencing or protein identification. Scattering may be used, for example, to discriminate vessels full of oxygenated hemoglobin from those lacking it, thus enabling functional analysis in the brain or other organs [19]. The main challenge is again the faintness of scattered signals, subject to significant background noise. The discrimination between parallel optical channels is also a major problem due to the number of interfering optical signals in any given location, even when various coding/ decoding techniques are used. Owing to the exceptional combination of numerous advantages, SiPMs and SPAD arrays open new perspectives in further development of these methods and miniaturization of respective devices. To summarize, they can serve as photo-detectors in the following applications: DNA analysis, measuring protein dynamics using light scattering, two-photon fluorescence microscopy, automated DNA sequencing machines, particle and droplet sizing, optical biopsy, study of cancer processes using fluorescent proteins, fluorescence or diffuse optical tomography, implantable or endovascular fluorescence probes. Their commercial availability and customization open numerous wide opportunities for replacement of tradi-

References [1] R.H. Haitz, Solid-State Electron. 8 (1965) 417. [2] C. Niclass, M. Sergio, E. Charbon, Design and Test in Europe (2006) pp. 81–86. [3] N. Basharuli, G. Bondarenko, et al., Advanced Technology and Particle Physics, World Scientific, Singapore, 2002, p. 627. [4] G. Bondarenko, V. Golovin, M. Tarasov, Patent No. 2142175. [5] A.N. Otte, J. Barral, B. Dolgoshein, et al., Nucl. Instr. and Meth. A 545 (2005) 705. [6] Z.Ya. Sadygov, V.N. Jejer, Yu.V. Musienko, et al., Nucl. Instr. and Meth. A 504 (2003) 301. [7] C. Niclass, A. Rochas, P.A. Besse, E. Charbon, IEEE J. Solid-State Circuits 40 (9) (2005) 1847. [8] F. Zappa, A. Gulinatti, P. Maccagnani, S. Tisa, S. Cova, IEEE Photon. Technol. Lett. 17 (3) (2005) 657. [9] A. Akindinov, V. Golovin, E. Grigoriev, et al., Nucl. Instr. and Meth. A 539 (2005) 172. [10] Z. Sadygov, et al., Nucl. Instr. and Meth. A 550 (2005) 212. [11] B. Dolgoshein, M. Danilov, et al., Nucl. Instr. and Meth. A 540 (2005) 368. [12] Patent Application PCT/FR2006/001060. [13] A. Grinvald, et al., Modern Techniques in Neuroscience Research, in: U. Windhorst, H. Johansson (Eds.), Springer, Berlin, 2001. [14] M. Go¨sch, A. Serov, T. Anhut, T. Lasser, A. Rochas, P.A. Besse, R.S. Popovic, J. Biomed. Opt. 9 (5) (2004) 913. [15] P. Schwille, F.J. Meyer-Almes, R. Rigler, Biophys. J. 72 (1997) 1878. [16] A.V. Agronskaia, L. Tertoolen, H.C. Gerritsen, J. Biomed. Opt. 9 (6) (2004) 1230. [17] W. Becker, K. Benndorf, A. Bergmann, C. Biskup, K. Ko¨nig, U. Tirplapur, T. Zimmer, in: Proceedings of SPIE 4431, ECBO, 2001. [18] H. Eltoukhy, K. Salama, A. El Gamal, M. Ronaghi, R. Davis, IEEE ISSCC, February 2004, pp. 222–223. [19] A. Maki, IEEE International Conference on Solid-State Sensors, Actuators, and Microsystems, vol. 2, June 2005, pp. 1103–1105.