MANY existing and planned experiments in high energy

Transcription

1 Prospects of Using Silicon Photomultipliers for the Astroparticle Physics Experiments EUSO and MAGIC A. Nepomuk Otte, Boris Dolgoshein, Jürgen Hose, Sergei Klemin, Eckart Lorenz, Gerhard Lutz, Razmick Mirzoyan, Elena Popova, Rainer H. Richter, Lothar W. J. Strüder and Masahiro Teshima Abstract We discuss the silicon photomultiplier (SiPM) as a novel photon detector for the Major Atmospheric Gamma Imaging Cerenkov Telescope (MAGIC) and the Extreme Universe Space Observatory (EUSO). For these astroparticle experiments we pursue two different developments of the SiPM. One development is pursued at MEPhI, where prototypes are available. In the course of this paper we will describe some characteristics of these devices. The second development, which is using the back illumination principle is at present in its design phase. Index Terms high energy astroparticle experiments, silicon photomultiplier (SiPM), avalanche photo diode (APD), air shower experiments. I. INTRODUCTION MANY existing and planned experiments in high energy astroparticle physics rely on photon detectors with fast time response, high photon detection efficiency and single photon counting capabilities. In this context we investigate a novel photon detector concept the silicon photomultiplier for the experiments MAGIC and EUSO. MAGIC is the worlds largest air Cherenkov telescope for ground based γ-astronomy. It is located on the island La Palma on top of the Roque de Los Muchachos at an altitude of 2200 m. The experiment detects very high energy gammas (VHE-γ) from a few tens of GeV up to several TeV. After entering the atmosphere, a high energy γ-ray initiates an electromagnetic shower in which many thousands of electrons are produced. A certain fraction of them propagates through the atmosphere faster than light and therefore emits Cherenkov light. The sum of all Cherenkov photons result in a light flash lasting a few nanoseconds which is detected by the MAGIC telescope. The flux of Cherenkov photons arriving on the ground is about 100 photons (300 nm-550 nm) per square meter for a 1 TeV-γ and scales in first order linear with the γ-energy. In order to detect such a small flux, large collection areas in combination with efficient and fast photon detectors are needed. Moreover, the photon detector has to operate in A. Nepomuk Otte ( otte@mppmu.mpg.de), Jürgen Hose, Eckart Lorenz, Gerhard Lutz, Razmick Mirzoyan, Rainer Richter and Masahiro Teshima are with the Max-Planck-Institut für Physik (Werner-Heisenberg- Institut), München 80805, Germany Boris Dolgoshein and Elena Popova are with the Moscow Engineering and Physics Institute (MEPhI), Moscow , Russia Sergei Klemin is with Pulsar Enterprise, Moscow, Russia Lothar W. J. Strüder is with the Max-Planck-Institut für Extraterrestrische Physik, Garching 85748, Germany the presence of a large light background from the night sky (stars, airglow, etc.). The physics objective of the MAGIC experiment is the study of the most violent processes known so far in our universe. Objects of interest are potential VHE-γ sources like e.g. active galactic nuclei, supernova remnants, gamma ray bursts and pulsars; fundamental physics questions as extragalactic background, quantum gravity and dark matter are also investigated. A detailed description of the MAGIC physics program can be found in [1]. Orbiting the Earth at about 500 km, EUSO looks down to the atmosphere and searches for air showers initiated by extreme energetic particles (> ev). The flux of these cosmic rays ( 1 particle per year per km 2 ) asks for huge detection volumes like the atmosphere which can easily be viewed from space. In an extended air shower, initiated by a cosmic ray, nitrogen molecules are excited. When falling back into the ground state, the molecules may emit fluorescence photons of distinct energies. The most prominent emission lines are in the wavelength region between 330 nm and 400 nm. This fluorescence light is the signature of a cosmic ray which EUSO is looking for. By recording the arrival time and position of each fluorescence photon in the EUSO-detector, the impact direction and energy of the primary cosmic ray can be reconstructed. For an extended air shower initiated by a cosmic ray with an energy of about ev, EUSO will detect several hundred photons within a time window of about 100 µs. Scientific objectives of the EUSO-mission are amongst others to study the high energy ( ev) part of the cosmic ray spectrum and trace back the origin of these high energy cosmic rays. For a more detailed discussion of the EUSOexperiment, the interested reader is pointed to [2] and [3]. Both experiments now use (MAGIC) or plan to use (EUSO) conventional photomultiplier tubes (PMTs) as photon detectors. From the physics point of view this is a far from optimal choice, as the photon detection efficiency (PDE) of availabe PMTs is limited to about 20%. A significant increase in signal to noise ratio (SNR) could be obtained, if a detector with a much larger PDE were be available comprising the high gain, speed and single photon counting capabilities of PMTs. For the experiments this means a lower energy threshold and improved energy resolution at higher energies. For EUSO the threshold must be as low as possible so results can be

2 sensibly compared to those produced by experiments like AGASA, HiRes and Auger. The motivation for MAGIC is to close the so far unobserved energy region between 30 GeV down to some GeV. Summarizing one can conclude that better photon detectors will produce better physics results in both experiments. For several years mainly Russian groups have developed the SiPM concept, e.g. [4] [6]. In the following we will introduce the idea behind the so called silicon photomultiplier (SiPM). We then focus on the development realized at the Moscow Engineering and Physics Institute (MEPhI) and Pulsar Enterprise before we report about a new development in the semiconductor laboratory (HLL) attached to the Max- Planck-Institut für Physik and Max-Planck-Institut für Extraterrestrische Physik. In that new development we want to combine for the first time the back illumination principle with the SiPM concept. II. THE SILICON PHOTOMULTIPLIER CONCEPT A SiPM is composed of an array of small avalanche photodiodes combined to form a macroscopic unit (typically 500 to 4000 diodes per mm 2 ). In the following we call such an avalanche photodiode a cell and the macroscopic unit SiPM. Each APD cell is independent from the rest of the unit and operates in limited Geiger mode, i.e. a few volts above breakdown voltage (see Fig. 1). In this mode of operation a single photoelectron can initiate a diverging avalanche confined to the APD cell (the photoelectron triggers the cell). The photoelectron can easily be detected, because in the breakdown a large and fast current pulse is generated. After the cell has been triggered, the breakdown is quenched by limiting the current through the cell. In the SiPM this is done by polysilicon resistors attached to the cells (s. Fig. 1). p- * Si Resistor V bias Al-conductor n + SiO 2 p+ Substrate p+ Guardring n - many small cells which are connected to the same readout line (marked as Al-conductor in the sketch). The main area of applications is in average light fluxes 1 photon/cell/recovery time. In this way the number of fired cells is in first order proportional to the photon flux, thus compensating for the missing dynamic range of a single Geiger mode APD. In reality the process is more complex because of the recovery time of the cells and the influence of dark rate. The main advantages of SiPMs are: standardized output pulses for single photoelectrons large intrinsic gain no need for sophisticated preamplifiers fast rise time pulses (about 1 ns) very low operating voltages V very low time jitter for single photoelectrons (about 100 ps) no sensitivity to magnetic fields very low sensitivity to pickup, EMI no aging no damage when exposed to large photon fluxes compactness low power consumption (about 10 µw/mm 2 ) expected low costs because of production process high radiation hardness A trade-off of available SiPM prototypes is a relative high dark rate (0.2 2) MHz/mm 2 at room temperature. The dark rate originates from thermally generated electrons in the active volume and tunnelling assisted charge carrier generation in the avalanche region. The former contribution can in principle be reduced by active cooling and the later one by a careful design of the avalanche region. Besides the dark rate another intrinsic feature of the SiPM has to be taken into account. Photons generated in the avalanche propagate within the SiPM and eventually are absorbed in the active volume of a neighboring cell. This gives rise to optical crosstalk. The efficiency of the photon generation is about per avalanche charge carrier [12]. Therefore, a straightforward measure to reduce this kind of crosstalk is to lower the gain as much as tolerable in the final application. Another approach is to absorb the crosstalk photons between neighboring pixels. Both methods can have an impact on the PDE. Because the probability of a Geiger breakdown is dependent on the overvoltage above breakdown voltage and therefore on the gain of a given device and grooves between pixels for the absorbtion of crosstalk photons might introduce additional dead area. Fig. 1. Cross section through the topology of one SiPM APD cell. Photons are incident from the top (from [7]) The fast and large output signal of single Geiger APDs is very attractive for highly efficient photoelectron detection purposes and has been studied by several authors [8] [11]. On the other hand, one is by no means able to attain information about the number of photoelectrons that initiated the breakdown, as the Geiger-APD output signal charge is fixed. This problem is circumvented in the SiPM, as the sensor is divided into III. SIPM PROTOTYPES BY MEPHI AND PULSAR We have studied prototype SiPMs produced by MEPhI and Pulsar enterprise to test the general use of this photon detector in experiments like MAGIC or EUSO. In this prototype, 576 APD cells are integrated on a total sensor area of (1 1)mm 2. A single cell has an active area of (20 20)µm 2. This is about four times smaller than the total cell size. A summary of the most important characteristics of the tested SiPM is given in Tab. I. In the following we will discuss some of the features in more detail.

3 TABLE I SPECIFICATIONS OF THE SIPM PROTOTYPE BY MEPHI AND PULSAR ENTERPRISE UNDER INVESTIGATION. parameter value Sensor area (1 1) mm 2 Nr. of individual Geiger cells 576 active area of the sensor 25% peak PDE (at 540 nm) 20% averaged over SiPM [5] bias voltage 52V-60V breakdown voltage 50V signal charge (gain) APD cell recovery time 1 µs typ. noise rate at room temperature 10 6 counts/mm 2 /s Fig. 2. Gain dependence on the bias voltage for the SiPM operating at room temperature In Fig. 2 the gain dependence of the SiPM as a function of the the applied bias voltage is plotted. It would be more correct to speak of signal charge rather than gain as the output signal of a single cell is independent of the generated photoelectrons. But as long as the probability is small that more than one photoelectron is generated in a cell which is the case for applications of SiPMs signal charge and gain has the same meaning. From the graph one can deduce the linear dependence of the gain on the applied voltage. This is due to the intrinsic structure of a single cell. In the process of a Geiger breakdown, the intrinsic cell capacitance and parasitic capacitance of the quenching resistor is discharged. Due to the linear relationship between charge Q and change of potential on a capacitance C, the charge that is externally flowing is therefore linearly dependent on the applied voltage U Bias Q = C (U Bias U cell@breakdown ) U Bias. (1) From the slope of the gain vs. bias one can deduce the total capacitance that is contributing to the output signal. Considering the above given equation we estimate a capacitance of about 40 ff per cell. After a cell has fired, it needs a certain time to recover to its original state. The recovery time constant is given by the value of the quenching resistor and a capacitance. The capacitance in one part comes from the pn junction and in another part from the parasitic capacitance parallel to the quenching resistor. For the measurement of the recovery time we have used a fast pulsed diode laser. The laser intensity was set to a level somewhat below at which a further increase did not show an increase in signal amplitude of the SiPM. We then reduced the time between subsequent laser pulses stepwise and recorded the resulting signal amplitude. From the measurement we derive a recovery time of 1 µs, i.e. the change in time needed between subsequent laser pulses that results in a change of 63% in a signal amplitude. The recovery time has an impact on the PDE of the SiPM as a certain amount of cells will always be dead due to dark noise or background light. Thus the active area of the SiPM is reduced. The largest limitation in PDE of the tested device is due to the currently small fraction of the cell area which is sensitive to photons (25%). This explains the peak 20% PDE of this device at a wavelength of 540 nm [5], i.e. the active area alone has a PDE of 80%. Below 450 nm the efficiency of the SiPM drops rapidly. This can be explained by the intrinsic structure of the cells. As can be seen from Fig. 1, the avalanche structure is implanted on a p-doped substrate (hereafter named n-on-p-structure). The pnjunction is reverse biased. Therefore only electrons that are generated in the depleted p-doped bulk (white in the drawing) below the structure drift into the avalanche region. UV-photons are absorbed within a short distance (< 100 nm) compared to the extension of the high field structure. Because of this the electrons generated by UV-photon absorbtion will not drift into the high field region. The hole that is also generated and which is drifting into the avalanche region has a much lower probability to initiate the avalanche breakdown. A high recombination rate close to the surface reduces further the UV sensitivity It is planned to invert the structure from n-on-p to p- on-n in the next prototypes in order to enhance the UVsensitivity. Then the photoelectrons generated after a UVphoton absorbtion close to the surface drift into the high field region as the electric field has changed its sign. Only inverting the structure will not improve the PDE much beyond 20% because of the dead area between cells. That is why we also plan to increase the overall cell sizes to ( ) µm 2. In doing so we want to increase the active area to 60%. If this development is successful we think about applying optical micro structures like micro lenses or micro Winston cones to the SiPM. This will further increase the effective active area of this type of SiPM. The optical crosstalk in this SiPM is 40%, if the sensor operates at a gain of This 40% is the fraction in which more than 1 cell fires at the same time while recording dark noise. After lowering the gain to , the crosstalk is reduced to 4%, a value which is acceptable for both EUSO and MAGIC. In future developments it is planned to introduce grooves between the cells of the SiPM. Photons that are generated in an avalanche and propagate into a neighboring cell will then be partially inhibited from doing so as they are either absorbed or scattered away at the grooves. Power consumption of the SiPM is almost negligible and dominated by dark rate. Operating the SiPM at room temper-

4 ature at a bias voltage of 55 V, corresponding to a gain of 10 6, the dark rate is about 1 MHz. This results in an average power dissipation P per mm 2 of P = 55 V q MHz 10 µw/mm 2, where q denotes the elementary charge. The power consumption due to the photon detectors in EUSO would then be about 50 W if equipped with this kind of SiPM. On the other hand, the SiPMs will need moderate cooling if used in EUSO, thus increasing the overall power consumption. shallow p+ drift rings p+ photoelectron driftpath photon depleted bulk deep n deep p 50µm 450µm avalanche region n+ electrode output line 100µm quenching resistor photoelectron path depleted bulk photon avalanche region Fig. 4. Blowup of one cell of the back illuminated SiPM principle. Please note that the drawing is not to scale. Si 50µm...450µm output Fig. 3. Principle of the back illuminated SiPM. The box marks one cell of which the blowup is shown in Fig. 4 well defined drift field between the two surfaces the backside is biased to -500 V. The avalanche region consists of a deep p- implantation which is enhanced in the middle by an additional low doped p-implantation. This ensures a homogenous electric field smoothly falling off to the sides (s. Fig. 5). IV. SIPM WITH BACK ILLUMINATION The design of the SiPM as introduced in the previous section is conceptionally limited in PDE due to the dead space between adjacent cells and shadowing from aluminium contact lines. In the MPI Semiconductor Laboratory in Munich we pursue a different approach in which we want to combine the SiPM principle with the back illumination principle that is already used with great success, e.g. with CCDs. The basic idea of the back illuminated SiPM is sketched in Fig. 3. Photons are incident on the detector side opposite to the avalanche regions (hereafter named backside) and drift into the avalanche region. As in the front side illuminated SiPM the sensor itself is divided into quasi independent cells. In Fig. 4 we sketch the cross section of such a cell. The cell can be divided into a completely depleted drift volume and the avalanche region, which is much smaller than the rest of the cell. On both sides of the drift volume, contacts at different potentials shape the electric field in such a way that photoelectrons drift into the avalanche region. Upon entering the avalanche region, a photoelectron will initiate a Geiger breakdown as in the front illuminated SiPM. Device simulations with the TOSCA program were performed to show the functionality of the back illuminated SiPM. We investigated a cylinder symmetric shaped cell with a diameter of 100 µm and a thickness of 50 µm and 450 µm respectively as shown in Fig. 4. First test structures and prototypes will be fabricated using 450 µm thick silicon. If necessary, further developments will be done using 50 µm thick silicon detector material. The drift field is shaped by three rings on the front side, including the deep p-implantation which is also part of the avalanche region. The applied voltages to these rings starting with the innermost is -50 V, -55 V and -60 V, respectively. For the full depletion of the bulk and a Distance from the frontside *10-3 cm *10 3 V/cm *10-2 cm 0.06 Distance from center of the Avalanche cell Fig. 5. Absolute electric field distribution of the Avalanche region. The values on the lower axes give the coordinates in cm, and the corresponding electric field is given on the vertical axis in V/cm. We simulated the position dependent charge collection efficiencies of this device by subdividing the avalanche region into six boxes of equal volumes and generated charges at different positions in the drift volume. For these simulations the avalanche mechanism was switched off. From the results we can deduce that the charge is nicely collected in the middle of the avalanche region. From the difference in drifttimes we estimate a time jitter below 1 ns. In Fig. 6 we simulated an avalanche breakdown in one cell. The quenching resistor was 1 MΩ with a parallel parasitic capacitance of 10 ff. The current pulse width is 100 ps with an amplitude of 4 ma.

5 ACKNOWLEDGMENT We would like to thank P. Liebig and S. Rodriguez for carefully reading this manuscript. current [ma] time [ns] Fig. 6. Current pulse of a simulated avalanche breakdown. The value of the quenching resistor was 1 MΩ with a parallel parasitic capacitance of 10 ff. V. SUMMARY AND CONCLUSION We have investigated the SiPM as a photon detector for the experiments MAGIC and EUSO. The SiPM shows features very similar to those of conventional photomultiplier tubes (PMTs). Moreover, it has many advantages compared to PMTs, e.g. it can be exposed to a strong light source without degrading the SiPMs properties, insensitivity to magnetic fields or pickup, etc.. The tested prototypes from MEPhI can be used as a photon detector in MAGIC and EUSO if further developments will succeed in enhancing the PDE beyond 50% in the blue wavelength region and in increasing the sensor size to (4 4) mm 2 for EUSO and at least (10 10) mm 2 for MAGIC. The focal plane of both experiments is in the order of a few square meter. A coverage of these surfaces with existing one square millimeter SiPMs is therefore not feasible. The PDE enhancement is pursued by increasing the cell size and changing the structure to p-on-n in future SiPMs. Optical crosstalk is in principle no problem if the SiPM operates at a gain of Additional efforts in decreasing the optical crosstalk will be made by introducing grooves between cells to absorb crosstalk photons. The SiPMs under investigation have a fairly high dark noise rate. As both experiments operate in the noisy environment of the night sky, only moderate cooling down to -50 C is needed to bring the dark noise down to an acceptable level. The front side illuminated SiPM is limited in PDE, as some areas will always be shadowed or otherwise insensitive to photons. For this reason we develop a new kind of SiPM at the MPI semiconductor laboratory that uses the back illumination principle. We have performed simulations to find a structure in which photons enter the detector from the backside and the photoelectrons drift into an avalanche region which is much smaller than the rest of the cell. The simulation results are very promising and meet the timing requirements needed in both experiments. We will now do test implantations to verify our technology simulation and in parallel develop test structures to pinpoint important parameters of the device like crosstalk, dark rate and breakdown behavior. REFERENCES [1] J. A. Barrio, et al., Technical Design Report of MAGIC, Max-Planck- Institut für Phyisk, Munich, MPI-PhE/98-5, March 1998, p (see also [2] L. Scarsi, et al., EUSO Report on the Phase A Study, EUSO-PI-REP- 002, Aug. 31, 2003 (see also [3] M. Teshima, et al., EUSO (The Extreme Universe Space Observatory)- Scientific Objectives, in Proc. ICRC, 2003, pp [4] V. Golovin, V. Saveliev, Novel type of avalanche photodetector with Geiger mode operation, NIM A, vol. 518, pp , [5] B. Dolgoshein, P. Buzhan, L. Filatov, A. Ilyin, V. Kantzerov, V. Kaplin, A. Karakash, F. Kayumov, S. Klemin, E. Popova and S. Smirnov, Silicon photomultiplier and its possible applications, NIM A, vol. 504, pp , [6] D. Bisello, Yu. Gotra, V. Jejerc, V. Kushpil, N. Malakhov, A. Paccagnella, Z. Sadygov, I. Stavitsky and E. Tsyganovc, Silicon avalanche detectors with negative feedback as detectors for high energy physics, NIM A, vol. 367, pp , [7] B. Dolgoshein, P. Buzhan, A. Ilyin, V. Kantserov, V. Kaplin, A. Karakash, A. Pleshko, E. Popova, S. Smirnov and Yu. Volkov. (2001, Fall). An advanced Study of Silicon Photomultiplier, ICFA [Onine]. Available: [8] R. J. McIntyre, Multiplication noise in uniform avalanche diodes, IEEE Trans. on Electron Dev. ED 13 pp 164. Jan [9] Gramsch, M. Szawlowski, S. Zhang, and M. Madden, Fast, high density avalanche photodiode array, IEEE Trans. Nucl. Sci., vol. 41, pp , [10] H. J. Kindt, N. H. Shahrjerdy, H. W. van Zeijl, A silicon avalanche photodiode for single optical photon counting in the Geiger mode, Sensors and Actuators A, vol. 60, pp , [11] A. Lacaita, M. Ghioni, F. Zappa, G. Ripamonti and S. Cova, Recent advances in the detection of optical photons with silicon photodiodes, NIM A, vol. 326, pp , [12] A. Lacaita, F. Zappa, S. Bigliardi and M. Manfredi, On the Bremsstrahlung Origin of Hot Carier Induced Photons in Silicon Devices, IEEE Trnas. on Electron Dev., vol. 40 No. 3, pp , Mar