Characterization of Thin p-on-p Radiation Detectors with Active Edges T. Peltolaa,∗, X. Wub, J. Kalliopuskab,c, C. Granjad, J. Jakubekd, M. Jakubekd, J. H¨ark¨onena, A. G¨addaa aHelsinki Institute of Physics, P.O. Box 64 (Gustaf Ha¨llstr¨omin katu 2) FI-00014 University of Helsinki, Finland bVTT, Microsystems and Nanoelectronics, Tietotie 3, Espoo, P.O. Box 1000, FI-02044 VTT, Finland 6 cAdvacam Oy, Tietotie 3, Espoo, FI-02150, Finland 1 dInstitute of Experimental and Applied Physics, Czech Technical University in Prague (IEAP-CTU), Horska´ 0 3a/22, CZ 12800 Prague 2, Czech Republic 2 n a J 1 Abstract 2 Active edge p-on-p silicon pixel detectors with thickness of 100 µm were fabricated on 150 mm ] t e floatzonesiliconwafersatVTT.BycombiningmeasuredresultsandTCADsimulations,adetailed d - study of electric field distributions and charge collection performances as a function of applied s n voltage in a p-on-p detector was carried out. A comparison with the results of a more conventional i . s active edge p-on-n pixel sensor is presented. The results from 3D spatial mapping show that at c i pixel-to-edge distances less than 100 µm the sensitive volume is extended to the physical edge of s y the detector when the applied voltage is above full depletion. The results from a spectroscopic h p measurement demonstrate a good functionality of the edge pixels. The interpixel isolation above [ full depletion and the breakdown voltage were found to be equal to the p-on-n sensor while lower 2 v charge collection was observed in the p-on-p pixel sensor below 80 V. Simulations indicated this to 1 7 be partly a result of a more favourable weighting field in the p-on-n sensor and partly of lower hole 1 1 lifetimes in the p-bulk. 0 . Keywords: Silicon radiation detectors; Pixel sensors; Electrical characterization; Charge 1 0 collection; TCAD simulations 6 1 : v i ∗Correspondingauthor X Email addresses: [email protected](T.Peltola),[email protected](X.Wu) r a Preprint submitted to Journal of LATEX Templates January 22, 2016 1. Introduction An advantage offered by thin silicon radiation detectors in spectroscopic applications is the good radiation differentiation which is particularly important for nuclear industry where short range particles need to be detected with high gamma ray background [1]. Also radiation hardness is important in many nuclear safeguard applications. Further benefits of thin detectors include reduced mass, fast charge collection and due to lower drift time for a given voltage, some possible advantage in charge collection and reverse current after high irradiation fluences [2]. The planar p-type silicon radiation detectors have been under extensive studies in the high energy physics (HEP) community due to their radiation tolerance and cost effectiveness, and have become a strong candidate to replace the conventional n-type detectors for the Large Hadron Col- lider(LHC)upgradeatCERN[3,4]. Ap-typedetectoristypicallyreferredasn-on-p(n+/p−/p+). The advantages of this configuration include a favourable combination of weighting and electric fields after irradiation due to the absence of type inversion. The readout at n-type electrodes en- ables the collection of electrons that have three times higher mobility and longer trapping times than holes, resulting in high speed readout and higher radiation hardness. Further asset of the p-type sensor is the reduced dependence of the charge collection efficiency (CCE) from the reverse annealing of the effective space charge in highly irradiated detectors [4, 5]. However, the challenge of n-on-p design lies in the deteriorated isolation between the collection electrodes due to the electron accumulation layer induced by the positive charges in the SiO 2 passivation layer. Two methods, known as p-stop and p-spray, are widely used to improve the electrode isolation. Both methods, however, increase the process complexity and might lead to localized high electric fields which increase the likelihood of early breakdowns. Thin p-on-p pixel detector addresses this problem without compromising CCE and spatial res- olution excessively. Compared with the n-on-p detector, the p-on-p detector collects holes and the pn-junction remains on the unsegmented side of the detector. The relatively low hole mobility and long drift distance to the collection electrodes leads to a deteriorated CCE and make the thick p-on-p detector difficult to use. Themotivationtoimplementthethinp-on-ppixeldetector,exceptfortheadvantagesmentioned above, is the expected good CCE after irradiation and an improved spatial resolution which make the thin p-on-p pixel detector distinguished from its thicker counterparts. The sensitive volume of the sensor can be further extended by using active edge or edgeless design that minimizes the 2 regions where the signal cannot be collected [6, 7, 8, 9]. This concept could be usable also for high energy physics (HEP) tracking applications. For instance, to maintain low material budget and achieve high position resolution the implementation ofthinpixeldetectorsintheinnerdetectorlayersoftheLHCexperimentsisforeseenforthefuture upgrades [10]. The thin active edge p-on-p configuration has not been studied before in the HEP community. 2. Device and fabrication The investigated active-edge sensors were fabricated at VTT Technical Research Centre of Fin- landon150mmFloatzonesiliconwafersfromTopsilSemiconductorMaterialsA/Swithathickness of 100 µm. The geometry of the sensor is a matrix of 256 x 256 pixels with a 55 µm pitch and a pixel implant diameter of 30 µm. The layer thicknesses and implantation depths as well as a cross-sectional view of the pixel implant and its metallization are presented in table 1 and figure 7, respectively. Both n-type (resistivity of 5 kΩ·cm) and p-type (resistivity of 10 kΩ·cm) wafers were used to fabricate sensors with different polarities. The basic electrical characterization (IV, CV, etc.) of both p-on-p and p-on-n sensors has been reported in the reference [11]. The electrical contacts of the sensors are DC-coupled. 3. Spectroscopic results The p-on-p edgeless sensor was hybridized on the Timepix readout chip [12] at VTT with the lead-tin solder at a temperature of 210◦ C. The assembly was then wire bonded on the stacked PCB board designed by the Institute of Experimental and Applied Physics (IEAP-CTU), Czech Technical University in Prague. The FITPix [13] USB and Pixelman [14] software were used for the data acquisition and analysis. The detector was operated in the ToT (Time-over-Threshold) mode and with the bias voltage of 100 V. IntheTimepixdetector,eachpixelisconnectedtoitsindividualpreamplifier,discriminatorand digital counter integrated on the readout chip. The detector works in one of three modes: Medipix mode (the counter counts incoming particles), Timepix mode (the counter works as a timer and measures the time when the particle is detected) and the ToT mode (the counter is used as a Wilkinson type ADC allowing direct energy measurement in each pixel). The Timepix detector 3 Figure 1: Global energy calibration of the p-on-p edgeless detector hybridized on the Timepix readout chip. working in ToT mode measures the charge collected from each pixel. As the device contains 65536 independent pixels and their response can never be identical, it is necessary to perform an energy calibration for each of them. Theenergycalibrationandthespectroscopiccharacterizationwereperformedwiththeradioac- tivesource241AmandX-rayfluorescenceemittedbyaminiX-raytubewithvarioustargetmaterials (Cu, Zn, Zr, Mo, Cd, In). The p-on-p detector was first energy-calibrated with seven monoenergetic radiations. Only single pixel clusters were recorded to avoid the pile-up of signals. If the effect of charge sharing with the neighbouring pixels is sizeable, this will generate clusters including more than one pixel. Forourinvestigation,allclusterscomprisingofmorethanonepixelwereexcluded. Forthe100µm sensor with the bias above V , the charge sharing is not a sizeable effect. fd The spectral peak positions were found by Gaussian fitting method. Then the calibration was performedbyfittingthesevenToTpeakstotheknownenergies,asreportedin[15]. Figure1shows the global energy calibration curve of the investigated p-on-p detector. In reality, the calibration process was performed for each pixel individually which results in 65536 calibration curves. The energy calibration was performed for both p-on-n and p-on-p detectors. The edgeless technology used in the fabrication minimizes the inactive regions at the edges of the detector. The vicinity of n+ doped edge, however, distorts the electric field distribution and thus might influence the CCE of the nearest pixels. To study the edge pixel dependence on the 4 pixeltophysicaledgedistance,thefourdetectorbordersweredesignedtohavevariouspixel-to-edge distances (50 µm, 100 µm, 150 µm and 200 µm). Figure 2 shows the spectral responses of the edge pixels compared to the center ones. The spectral responses of the outermost pixels were summed along the edge and compared with the responseofthecenterpixels. Theresponsesofthecornerpixelanditsfourneighboringpixelswere excluded from the data analysis to eliminate the corner effect due to the wider depletion volume of the corner pixel and the electric field distortion due to the two edges. Since the center pixels have a smaller charge collection region this results in a smaller mean response than for the pixels at the edges. To reach sufficient statistics the data acquisition for e.g. Am exposure took about 30 minutes (168579 frames at 0.01 s per frame). Totally 20 million clusters were recorded resulting in ∼300 events for each pixel. As shown in the figure, the peaks of all the edge pixels and the center pixels are well aligned (sampling was done with the energy interval of 1 keV and all energy peaks appeared in the same location) and the heights of the peaks are proportional to the effective volumes of the pixels, indicating good functionality of the edge pixels. 4. Three-dimensional spatial mapping A 3D spatial mapping system at IEAP-CTU [16, 17] was used to investigate the influence of the active edge on the charge collection volumes of the edge pixels. The principle of the scanning systemistouseacollimatedtungstenX-ray(40kV)penetratingthedetectorinasharpangle(70◦), allowingthebeamwithadiameteroflessthan1mmtointeractwithseveralpixelsatvariousdepths. When the detector is shifted perpendicularly to the beam direction, the interactions at all pixel depths are recorded and a 3D map of the detector charge collection volume is obtained. Figure 3 shows the spatial mapping results of the p-on-p edgeless detector at a bias voltage of 20 V. The voltage was selected to be above the V ≈ 10 V with some margins. Thus, the fd depletionvolumewillreachthebackplaneofthe100µmthicksensorbutthelateralexpansionwill have different distances to the edge. Ten pixels adjacent to the detector edges were investigated. The X-ray enters the detector with a certain angle. Therefore the charges collected from different interaction depths of the pixel can be recorded. These regional signal responses are corresponding to the sensitive volumes of the pixels at certain depths which are indicated by different colors in figure 3. It can be seen that the outmost edge pixels collect charges from wider volumes than the other pixels. The pixels having wider pixel-to-edge distances received correspondingly more 5 (a) (b) Figure 2: (a) The 256×256 pixel topology of the detector (not to scale). The filled-in pixels were used for the response analysis of the given edge. The pixels in the center highlight the position of the pixels used to calculate the center response. For clarity a part of the pixels are not pictured, i.e. alltheblankregionscontainpixelsandallthepixelswithinthecentralsquarewereincludedin thecenterresponse. (b)Spectralresponseofthep-on-pedgelessdetectortothecopperandindium fluorescence and americium irradiation. Only the mean values of single pixel events were studied. 6 charges, indicating that the electrical distortion happened near the doped edges that extended the depletion volume towards the edges. When the measurement results from the three sensors in figure 3 are compared, it can be observed that the amount of charge registered in the outmost edge pixel is not proportional to the physical edge width of the sensor. This is usually due to the existence of certain non-depleted volume in the sensor bulk, e.g. at the corner on the segmented side of the chip, where the electric field is too ”weak” to build up a depleted region. 5. Proton beam irradiation The proton beam tests were carried out at the Van de Graaff accelerator laboratory of the InstituteofExperimentalandAppliedPhysics1 inPrague. Acceleratedprotonsofdifferentenergies from the beam line were used for the test. One 100 µm thick p-on-n edgeless detector and one 100 µm thick p-on-p edgeless detector were chosen for the test. The detectors and readout electronics were placed into a vacuum chamber into which the beam was guided. A gold foil (thickness in the range of 0.5 mm) was positioned in the vacuum chamber towards the beam direction to scatter the high intensity proton beam to the entire area of sensor, i.e. the tilted angle of the foil with respect to the beam allows the proton scattering to the sensor. The detectors were irradiated by the scattered beam from the non-segmented backplane and the signal was transmitted to the DAQ with coaxial cables. Figure 4 shows the deposited energies on two edgeless detectors as a function of applied bias voltage. It can be seen that the energies collected by the two 100 µm thick detectors increase with the bias voltages and the correct energies are registered only when the bias voltage is above 80 V. Also the p-on-n detector is collecting more charge for the most part of the voltage range. This behaviour is further investigated in section 6. Figure 5 shows a simulation using the SRIM tools2. The 300-800 keV protons are mostly absorbed within the 12 µm depth from the incident silicon surface. 1http://aladdin.utef.cvut.cz/projekty/VdG 2www.srim.org 7 Figure3: Spatialmappingresultsoftheedgepixelsatthreeborders. Thecolorsrepresentdifferent depths with respect to the segmented detector surface. The outermost pixels in the diagram have 50, 100 and 200 µm distance to the physical edge. 8 Figure 4: Collected proton energies (charges) as a function of the applied bias voltage. Figure 5: The SRIM simulated absorption depths in silicon with a layer of aluminum as in the backplane of the measured detectors. 9 6. TCAD simulations and modelling The simulations presented in this paper were carried out using the Synopsys Sentaurus3 finite- element Technology Computer-Aided Design (TCAD) software framework. 6.1. Simulation set-up Forthesimulationstudyoftheobservedchargecollectionbehaviourinfigure4the3-dimensional structure presented in figure 6 was applied. This was deemed necessary since in a 2-dim. structure onlyinteractionsbetweenasinglecolumnofpixelswouldbemonitored. Alsothecorrectreproduc- tion of the local electric fields due to the circular shape of the pixels required a 3-dim. structure. Sincemostofthepixelsinthedetectordonotexperienceanyinfluencefromtheactiveedges,these were not included in the simulated structure, as can be seen from figure 6. Thesimulatedpixelsensorconfigurations,p-on-pandp-on-n,weredesignedwithparametersas closetotherealsensorsaspossibleandthebothsensortypeshadaphysicalthicknessof100µm,a pitchof55µmandapixelimplantdiameterof30µm. Thelayerdimensionsanddopingparameters are given in table 1, in which the bulk dopings were estimated by using the resistivity data of the two sensor substrates (∼10 kΩ·cm for the p-on-p and ∼5 kΩ·cm for the p-on-n). The lateral diffusion of the pixel implants was set to 0.8×depth. The aluminum metallizations above the pixel implantsandtheirviasthroughtheoxidelayerhadthediametersof36µmand24µm,respectively. Detailedcross-sectionalsliceofthepixelispresentedinfigure7. Sincetheonlydifferencesbetween the sensor types (both had p+ implantations at the pixels and n+ at the non-segmented side) were the pixel implant diffusion depths and the type and concentration of the bulk doping, the figures 6 and 7 can be considered to represent both of the simulated sensor structures. Each pixel had a DC-coupled electrode at zero potential with sufficiently low resistance for charge collection. The reverse bias voltage was provided by the backplane contact. 6.2. Simulation results 6.2.1. Electrical characteristics Eventhoughthebulkdopingconcentrationofthep-on-psensorwassettoaconsiderablyhigher valuethanforthep-on-n, asshownintable1, thecapacitance-voltage(CV)simulationsofthetwo 3http://www.synopsys.com 10