Abstract

Light measurements are widely used in physics experiments and medical applications. It is possible to nd many of them in High{Energy Physics, Astrophysics and Astroparticle Physics experiments and in the PET or SPECT medical techniques. Two different types of light detectors are usually used: thermal detectors and photoelectric effect based detectors. Among the rst type detectors, the Bolometer is the most widely used and developed. Its invention dates back in the nineteenth century. It represents a good choice to detect optical power in far infrared and microwave wavelength regions but it does not have single photon detection capability. It is usually used in the rare events Physics experiments. Among the photoelectric effect based detectors, the Photomultiplier Tube (PMT) is the most important nowadays for the detection of low-level light flux. It was invented in the late thirties and it has the single photon detection capability and a good quantum efficiency (QE) in the near-ultraviolet (NUV) and visible regions. Its drawbacks are the high bias voltage requirement, the diculty to employ it in strong magnetic field environments and its fragility. Other widely used light detectors are the Solid-State detectors, in particular the silicon based ones. They were developed in the last sixty years to become a good alternative to the PMTs. The silicon photodetectors can be divided into three types depending on the operational bias voltage and, as a consequence, their internal gain: photodiodes, avalanche photodiodes (APDs) and Geiger-mode detectors, Single Photon Avalanche Diodes (SPADs). The first type detector does not have internal gain, thus its signal is proportional to the number of incoming photons that are converted in electron-hole pairs. The detector and the read-out circuit noises limit the detector sensitivity to, at least, some hundreds of photons. The APD exploits the impact ionisation effect to have an internal gain up to some hundreds or more. The internal gain allows the detector to improve the performance with respect to a similar area photodiode reducing the sensitivity limit to some tens of photons. Operating the APDs with a bias voltage larger than the breakdown one, the Geiger-mode operation range can be reached. In this case, the detectors have a very high gain, in the order of (10^5-10^7), but their signal is not proportional to the number of the incoming photons, it is always the same. The SPAD (or GM-APD) is a typical Geiger-mode silicon detector. It has the single photon detection capability as the PMT, due to the high gain, but its signal is digital: it is fired or not by the incoming photons. The capability to give a signal proportional to the number of photons is lost in a SPAD. To recover this property, a matrix of independent SPADs connected in parallel is built. These matrices are called Silicon Photomultipliers, SiPMs (or Multi-Pixel Photon Counters, MPPCs). In a typical SiPM, the output signal is the sum of the SPAD ones, thus, although the digital nature of the SPAD signal, it is analogue and ranges from zero to the maximum number of cells composing the matrix. The detector response can be considered linear if the number of incoming photons is much smaller than the total number of cells because the probability that two photons arrive on the same cell is negligible. Main features of the SiPMs are the low bias voltage (<100 V typically), a high Photo-Detection Efficiency (PDE) in NUV and visible range (usually larger than the PMT QE), compactness, insensitivity to magnetic elds and high internal gain. The noise sources are the Dark Count Rate (DCR), the optical Cross-talk (CT) and the Afterpulsing (AP). The first is called primary noise and mainly depends, at room temperature, on the thermal generation of electron-hole pairs that can travel to the high-field region triggering a cell. The others are collectively called correlated noise because they can happen only after a primary signal, caused both by a photon or by a spontaneous carrier generation. In this thesis, the focus is on the characterisation of one SiPM technology produced in Fondazione Bruno Kessler (FBK) in Trento, Italy, named NUV-SiPMs. This technology, implementing the p-on-n concept, showed, from the beginning, a high PDE spectrum peaked in the NUV to violet region exceeding the 30 %, a very low DCR, typically below 100 KHz/mm^2 in the operating voltage range, and a total correlated noise probability under the 50 %. In the last years, the technology was developed modifying the silicon wafer substrate properties, reducing the delayed correlated noise probabilities, and adopting the so-called High-Density concept. In this last version, named NUV-HD, a new layout, with a narrower border region around the cell active area and deep trenches to electrically and optically isolate the cells, is employed. The first improvement has a direct influence on the PDE because it increases the Fill Factor (FF), the ratio between the active to the total cell area. The second layout change reduces the probability that a secondary photon can travel from a cell to a neighbouring one, reducing the cross-talk probability. Possible applications of the NUV-HD devices are: medical application (e.g. the Time of Flight PET, ToF-PET, scanners), the rare events Physics experiments (e.g. NEXT and DarkSide) and Astroparticle Physiscs experiments (e.g. the Cherenkov Telescope Array Observatory, CTA). The ToF-PET scanners are promising medical techniques with the goal of improving the spatial resolution of the classical PET scanners measuring the gamma rays time of flight. For this reason, these scanners require very fast photodetectors with coincidence time resolution (CTR) less or equal to 100 ps. NEXT and DarkSide are low temperature experiments with the main goal of observe rare events as the neutrinoless double beta decay and the Dark Matter particles. Due to the signal rarity, the detector noise requirements are very stringent. CTA will be a future ground-based Imaging Atmospheric Cherenkov Telescope (IACT) observatory, with the goal to track very high energy cosmic gamma rays, up to hundreds of TeV, to their galactic sources. It will consist of two matrices of telescopes of different type and size, each one having a camera. In the small size telescopes, the possibility to use the SiPMs to build the camera is under investigation. Since the cosmic gamma rays are detected through the secondary Cherenkov photons, produced by the accelerated electrons in the Earth atmosphere, the camera photosensors must have a very high detection efficiency from 300 nm to 600 nm, and, possibly, a low sensitivity in the NIR region to reject the night sky background. In addition, they must have good timing properties and high granularity. To fully characterise the SiPM, measurements of signal properties, noise parameters and PDE are needed. The set-ups and analysis of the characterisation procedure are fully described. In particular, the optical set-up, with its calibration procedures, and the analysis methods, with the definition of the possible uncertainty sources, are the central point of this work. During the dark characterisation, the SiPM is enclosed in a light-tight climatic chamber. An oscilloscope acquires and sends to a PC software millisecond-long SiPM waveforms. The software implements a Differential Leading Edge Discriminator (DLED) algorithm to better distinguish the SiPM pulses with time separation larger than a few nanoseconds. This analysis allows to count the primary pulses, due to the thermal/tunnelling excitations, obtaining the DCR, and measure the correlated noise probabilities. In addition, signal parameters as amplitude, gain and cell recharge time are measured. The PDE measurements require a set-up in which the number of impinging photons to the device is precisely known. For this reason, a compact set-up, consisting of an integrating sphere inside a light-tight box, a series of LEDs with peak wavelength ranging from NUV to NIR, fully characterised before use, a monochromator, equipped with a tungsten lamp, and a transparent optical fibre, was developed. Along with the set-up, a light calibration procedure, taking into account different uncertainty sources (LED wavelength shift, light uniformity at the device position, etc.), was also developed. Three different analysis techniques can be used to obtain the technology PDE. Each technique has its own benets and error sources. The equivalence among the different methods is shown. Moreover, measuring the PDE on SPADs with the same layout of single SiPM cells, identical results are obtained. This fact shows the equivalence between the single cell device and its larger counterpart, opening the possibility to measure the PDE of a new technology on SPADs. This is a very important result because the SPAD is a simpler device, with lower correlated noise, because it has no CT, and negligible primary one, often less than 1 kHz. Measurements are more precise, faster and it is possible to apply larger bias voltage, obtaining more information on the technology in such conditions at which no SiPM can be tested any more. A rst version NUV-HD technology characterisation is shown. In this version, the NUV-HD SiPMs have cell pitch ranging from 25 um to 40 um. A typical primary noise lower than 100 KHz/mm^2 and a delayed correlated noise probability less than 5 % are measured, up to 10 V of overvoltage. In the same bias voltage range, a direct CT probability lower than 45 % is measured in the largest cell devices (25 % in the smallest ones). The PDE spectrum has the expected shape with the maximum in the NUV-violet region. A maximum value exceeding the 60 % is measured in the largest cell devices (45 % in the smallest ones). To investigate possible variations of the measured features on the wafer, devices taken from different wafer points are measured and compared finding no difference but the primary noise. This parameter shows a variation by a factor up to about three on the wafer level. To compare the different cell devices, all the measured parameters are plotted as a function of the peak PDE, about 400 nm. During this comparison, the smallest device reveals worse than the others having a larger noise, both primary and correlated, at the same PDE value. The other three devices are comparable within the measurement errors. From the PDE measurements, a comparison between the measured FF and the expected one, as dened by the design, is obtained. In the smallest cell device, this comparison shows an unexpected discrepancy leading to the possibility that the expected FF is larger than the effective one. This possibility is investigated in the last part of this thesis in which a complete study of the factors contributing to the PDE is shown. This study is performed on a new NUV-HD version employing a photodiode with equal dopant prole of the SiPMs, a circular SPAD having 100 % FF and a square one with 35 um cell size and a nominal FF equal to 81 %. A developed box model is used to describe the electric eld inside the cell. The calculated effective FF is always different from the expected one. The reason of the measured difference is the electric field transition from the constant high value to zero occurring at the active area border region. This partially efficient region has an effect similar to an added completely ineffective region of 1-1.5 um size inside the expected active area. The transition region effect is critical for the smallest cells because it strongly reduces the effective FF with respect to the design one. The study of the factors contributing to the PDE of the NUV-HD SiPMs is very important. Through the obtained results, it is confirmed that the technology QE is just maximised in the wavelength range of interest, NUV to blue, and, at the same wavelengths, the triggering probability saturation rate is very small allowing the detectors to reach the maximum PDE when biased with a few volts of overvoltage. This operating condition has also the effect to employ the detector having low noise, both primary and correlated one. The best solution to further improve the technology PDE is a redesign of the electric field border region to reduce the gap between the expected FF and the effective one. This is more important for the smallest cell devices in which the actual transition region effect reduces the PDE performance to about the 50-60 % of the expected values.