AA new approach to achieving high granularity for silicon diode detectors withimpact ionization gain
S. Ayyoub b , C. Gee a , R. Islam b , S. M. Mazza a ,B. Schumm a , A. Seiden a , Y. Zhao aa The Santa Cruz Institute for Particle Physics and the University of California, Santa Cruz,California, 95064 b CACTUS Materials, Inc., Tempe Arizona, 85284
Abstract
Low Gain Avalanche Diodes (LGADs) are thin (20-50 µm ) silicon diode sensors withmodest internal gain (typically 5 to 50) and exceptional time resolution (17 ps to 50 ps ).However, the granularity of such devices is limited to the millimeter scale due to the needto include protection structures at the boundaries of the readout pads to avoid prematurebreakdown due to large local electric fields. In this paper we present a new approach –the Deep-Junction LGAD (DJ-LGAD) – that decouples the high-field gain region from thereadout plane. This approach is expected to improve the achievable LGAD granularity to thetens-of-micron scale while maintaining direct charge collection on the segmented electrodes. Low Gain Avalanche Diodes (LGADs) [1, 2, 3] are a type of thin silicon diode sensor, typicallyimplemented with an n-on-p architecture with the addition of a highly doped p+ layer just belowthe n-type implants of the electrodes. This additional layer, called the “gain” or “multiplication”layer, generates a high field region where controlled charge multiplication is possible, and canbe up to a few microns thick. The remainder of the sensor, which is typically composed ofhigh-resistivity silicon, is referred to as the “bulk”. Owing to the intrinsic impact-ionizationgain (which is typically between 5 and 50) these devices can be very thin (20-50 µm ) whileachieving charge collection levels greater than their much thicker conventional counterparts.This allows a short collection time with a fast rising edge that results in very precise timinginformation (17-50 ps ) [4, 5, 6, 7, 8]. LGADs were first developed by the Centro Nacional deMicroelectr´onica (CNM) Barcelona, with significant participation by the RD50 Collaboration [9].Due to their precise timing capability, LGADs offer a prospective new paradigm for space-timeparticle tracking [10].The first application of LGADs is planned for the High Luminosity LHC (HL-LHC [11]),where extreme beam-collision pileup conditions will lower the efficiency for tracking and ver-texing for the inner tracking detector in the region close to the beam pipe. To restore thisperformance, LGAD-based timing layers that can time-stamp collision vertices are being devel-oped for the forward region of both the ATLAS and the CMS experiments. The ATLAS andCMS projects are called, respectively, the High Granularity Timing Detector (HGTD) [12] andthe End-Cap Timing Layer (ETL) [13]. The pixellated LGADs that will be used in these timinglayers will make use of an inter-channel protection structure referred to as the “Junction Termi-nation Extension” (JTE). These structures, which avoid breakdown between adjacent readoutchannels by terminating the high field in the gain layer (see Fig. 1), create “dead regions” oforder 50-100 µm between channels, in which the charge collection is severely limited. As a result,the granularity of the LGAD sensors under development for use at the HL-LHC is limited tothe millimeter scale. Although this degree of granularity is acceptable for HL-LHC applications,many possible future applications of LGADs, such as four-dimensional tracking at future collid-ing beam facilities, X-ray imaging, and medical physics applications, will require granularity atthe tens-of-micron scale [14].In this paper an innovative LGAD design, referred to as the “Deep Junction LGAD” (DJ-LGAD), will be presented. The Deep Junction approach permits granularity on the same scale1 a r X i v : . [ phy s i c s . i n s - d e t ] J a n s that of conventional silicon diode sensors, while maintaining a direct coupling of the signalcharge to the readout electrodes. This new design features a multiplication zone that is de-coupled from the readout plane by burying a high-field diode junction several microns belowthe surface of the device, separated from the surface readout plane by a region of lower fieldthat is still high enough to maintain drift-velocity saturation. In this way the high field areais kept sufficiently far from the segmented area of the silicon so that inter-channel breakdownis avoided, allowing for the standard pixelization of the readout plane, and the achievement ofgranularity as fine as tens of microns. LGADs were first proposed as an approach to achieving precise timing resolution for minimum-ionizing-particles (mips) through the internal amplification of the intrinsic ionization signalthrough impact-ionization gain in the silicon bulk. Typically, an LGAD includes a highly-dopedregion just below the junction of a silicon-diode sensor, producing strong electric fields (withpeak value of around 3 × V cm − ) that induces controlled electron impact ionization as thecharge is collected. This signal amplification enhances the signal-to-noise ratio, significantlyreducing the contribution of readout noise to the timing measurement and allowing for the useof thin sensors with rapid charge collection. In this way, LGAD sensors have achieved timingresolution of 17 ps for mip timing measurements [15] and frame rates of 500 MHz [16].However, when such an LGAD sensor is spatially segmented, the strong electric field in themultiplication layer will cause an early breakdown at the edge of the readout electrode if notterminated properly. To prevent this early breakdown, a “Junction Termination Extension”(JTE) structure, shown in Figure 1, is introduced in order to reduce the electric field betweenthe segmented implants. However the JTE region, which generally extends for ∼ µ m betweenneighboring channels, has no signal amplification or collection and is thus inefficient for detectingincident radiation. This limits the granularity of practical sensors to the 1 mm scale, since a finergranularity would dramatically decrease the sensor active area. However, to make use of LGADtechnology to fully exploit the capabilities of future accelerator facilities – to accomplish “4Dtracking” for colliding beam detectors [17] or X-ray imaging at next-generation light sources [14]– granularity of better than 100 µ m will be required.Figure 1: Schematic design of the LGAD Junction Termination Extension (JTE) [18].To achieve higher granularity by overcoming the limitations of the JTE, several variations ofLGAD design have been proposed. For the “AC Coupled” LGAD (AC-LGAD) [19], the highlydoped n++ implant of conventional LGADs is replaced with a more resistive n+ layer, with the2lectrodes AC coupled to the n+ layer through a thin dielectric layer. This approach eliminatesthe need for a JTE structure (except at the very edge of the sensor), and can achieve a nearly100% fill factor without complicating the fabrication process. However, AC-LGADs might notbe suitable for applications that involve high repetition rate and high signal density due to theAC coupled signal and the intrinsic charge sharing in the resistive n+ layer.Another approach, referred to as the “Inverse” LGAD (iLGAD) [20], features a gain layerthat is fabricated on the opposite side of the sensor relative to that of the conventional LGAD,permitting conventional segmentation techniques to be used at the readout surface, which forthe iLGAD is thus on the opposite side of the sensor to the gain layer. However, iLGADs haveyet to achieve a temporal resolution competitive with conventional LGADs. Addtionally, theapproach complicates the fabrication process by requiring wafer processing on both sides. Thisprocess also makes it difficult to produce sensors with an active bulk as thin as the 20-50 µ mthickness characteristic of LGADs.Recently, a third approach to achieving fine granularity, referred to as the “Trench-Isolated”LGAD (TI-LGAD) [21], has been proposed that inserts narrow (several micron wide) physicaltrenches filled with insulator at the boundary of each pixel. For this approach, R&D is still inan early phase and the efficacy of the approach remains unknown.Here, we introduce a new variant of LGAD sensor design geared towards higher granularity:the “Deep-Junction” LGAD (DJ-LGAD). As discussed above, the DJ-LGAD design producesthe high-field gain layer within a semiconductor junction that is buried several microns belowthe readout surface, thereby allowing the electric fields in the bulk just below the sensor to beof similar magnitude to that of a conventional silicon diode sensor. In this way the readoutplane on top of the device can be spatially segmented without the use of the JTE, allowing forgranularity on the scale of tens of microns.In the following sections, two levels of simulation making use of the Sentaurus TCAD softwarepackage [22] from the Synopsys corporation are presented. In the first of these (Section 3),performed by the Santa Cruz Institute for Particle Physics (SCIPP), the basic idea of theDJ-LGAD was explored, making use of idealized doping profiles and disregarding potentiallimitations imposed by the fabrication process. For the second of these (Section 4), performedcollaboratively between SCIPP and Cactus materials, the basic elements of the fabricationprocess were incorporated into the simulation, resulting in the design of a practical device. Afirst, planar (unsegmented) device was subsequently fabricated at the Cactus Materials facility;rudimentary characteristics of this initial prototype are presented in Section 5. The “Deep-Junction LGAD” (DJ-LGAD), described here for the first time, is a new approachto the application of controlled impact-ionization gain within a silicon diode sensor. The term“deep-junction” arises from the use of a p-n semiconductor junction buried several micronsbelow the surface of the device. The buried junction is formed by abutting thin, highly-dopedp+ and n+ layers, with the doping density chosen to create electric fields large enough togenerate impact ionization gain in the narrow buried junction region. Additionally, the dopingdensities chosen for the p+ and n+ layers are balanced so that when the sensor is fully depleted,the electric field outside of the junction region, while large enough to saturate the carrier driftvelocity, is significantly less than that require to create impact ionization gain. This preservesthe electrostatic stability at the segmented surface of the detector, thus in principle permittingthe production of DC-coupled LGADs with arbitrarily fine granularity.3igure 2: Schematic depiction of the DJ-LGAD concept.Element Doping Level (N / cm ) Extent In DepthN isolation layer (Nbulk) constant doping of3 × From 0 µ m tobeginning of N+ “gainplate” layerN+ gain plate (upperhalf of gain layer) Gaussian, peak doping3 × peak at 4 µ m, Gaussianwidth of 0 . µ mP+ gain plate(lowerhalf of gain layer) Gaussian, peak doping3 × Peak at 5 . µ m,Gaussian width of0 . µ mP drift region (P bulk) constant doping of3 × End of P+ “gain plate”layer to 50 µ mP stop constant doping of1 × µ m deep at surface,1 µ m wideN++ implant (underelectrode) constant doping1 × at surfaceGain layer dopingtolerance (N+ and P+varied together) effective operation peakdoping between2 . × and3 . × Table 1: Doping profile assumed for the initial realization of the DJ-LGAD device.The bulk of the sensor is formed from high-resistivity n-type and p-type silicon. Above theburied junction, between the junction and the segmented readout plane, a several-micron-thicklayer of high-resistivity n-type material (the “isolation layer”) is used so that, upon depletion, itcontributes very little fixed charge in the isolation layer, leading to a relatively uniform, moderate4lectric field in the region between the junction and readout plane. Below the junction – betweenthe junction and the planar biasing electrode – a thicker layer of high-resistivity p-type materialserves as the charge generation medium for through-going charged particles (mips) or absorbedX-rays or heavy ions. A schematic of the proposed doping profile for an initial DJ-LGAD design,showing the bulk, readout, and junction regions described above, is provided in Figure 2.Making use of idealized doping profiles, the Sentaurus simulation package [22] was used toexplore the possibility of developing such a device, and to specify the doping levels needed tosatisfy the design goals stated above. The doping profile of a model device that arose fromthese studies, featuring a 50 µ m thickness and a transverse segmentation (“channel”) pitch of20 µ m, is displayed in Table 1, which provides a numerical summary of the doping characteristicsassumed in the various regions of the simulated sensor.Figure 3: Electric field strengths along a meridian of the DJ-LGAD device described by Table 1that passes through the center of a channel, over a range of applied reverse bias voltages.The magnitude of the electrostatic field for the doping profile of Table 1, along a longitudinalmeridian through the center of a channel, is shown as a function of applied bias voltage inFigure 3. For all displayed bias voltages, the 50 µ m thick sensor is fully depleted. The peak inthe electric field strength arises in the highly-doped buried junction region, and reaches valueslarge enough to induce impact ionization gain for free electron carriers. The field strength inthe regions outside of the gain region is large enough to saturate the drift velocity, but remainssmall enough to protect the device from breakdown between neighboring channels, without theinclusion of a JTE structure. Since the mean-free path for impact ionization depends sensitivelyon the electric field, the device gain can be adjusted through control of the bias voltage, withoutcompromising the functionality in the bulk and electrode regions.In order to study the signal response and gain of this DJ-LGAD model, the heavy ion featureof Sentaurus is used to simulate a mip passing through the device. A uniform energy transferof 0 . µ m − , typical for mip energy deposition in thin silicon, is defined in a vertical linebetween the upper and lower surfaces of the device in order to simulate the deposited energyfrom a mip injection. Figure 4, for a bias voltage of 300 V, shows the signal response froma mip injected at the center of a channel. The electrode of the channel for which the mip isinjected experiences a non-zero integral response with a rise time of order 100 ps, while nearby5igure 4: Simulated signal arising from a mip injection along a meridian of the DJ-LGADdevice described by Table 1 that passes through the center of a channel. The red trace is fromthe readout channel collecting the mip charge, and the blue and green traces are from the twoneighboring readout channels.electrodes have signals of significantly smaller magnitude, initially opposite in polarity but thenintegrating to zero within 1 ns.To characterize the gain of the device, the mip injection simulation is performed on a sim-ulated reference diode (pin diode) with the same structure of DJ-LGAD but without the gainlayer, and the gain is determined as ratio of the integrated signal charge of the DJ-LGAD rel-ative to that of the pin diode. As seen in Figure 5, impact ionization gain in excess of 10 isachieved for bias voltages greater than 280 V, with the gain varying smoothly with applied biasvoltage.As discussed in Section 2, the primary advantage of the DJ-LGAD approach is the elimina-tion of the JTE structure, which creates “dead regions” of limited detector response over manytens of microns between neighboring channels. In contrast, no JTE structure is required for aDJ-LGAD array, as the buried gain layer ensures a uniform gain performance across channels.The resulting gain uniformity can be studied by exploring the simulated sensor response as afunction of the transverse location of the mip injection, with the device held at constant biasvoltage. Figure 6 shows the DJ-LGAD response uniformity for an aggressive channel pitch of20 µ m, for the device of Table 1 biased to approximately 200 V. To account for the sharing ofthe signal when a mip is injected in the region midway between the center of two channels, thesum of gain from all enabled channels is formed for each transverse location of the mip injection.The variation in the gain is seen to be within ±
5% across the device, with the loss in gain in theregion between channels caused by a small distortion of the gain-layer fields that is introduced bythe narrow inter-channel p-stop insertion. At this level of simulation, the DJ-LGAD approachis seen to hold significant promise towards the development of a highly-pixellated DC-coupledsilicon diode sensor with substantial internal gain and precise temporal resolution.6
80 200 220 240 260 280 300 320 340Reverse Bias [V]0246810121416 G a i n Figure 5: Simulated DJ-LGAD gain as a function of applied reverse bias voltage for the concep-tual version of the DJ-LGAD, as described in Table 1.
50 60 70 80 90Mip position [um]012345678910 T o t a l ga i n f r o m a ll pad s Figure 6: Sum of the integrated signal charge over all channels, as a function of the transverseposition of incidence of a simulated mip, for the 20 µm pitch sensor of Table 1, biased atapproximately 200 V. Uniformity within ±
5% is observed
As a first realization of the DJ-LGAD concept, several closely-related versions of a planar (non-segmented) device have been designed and fabricated, based on a wafer-to-wafer bonding ap-proach to establish the buried junction interface. The goals of this effort were to demonstrate7he feasibility of the deep junction idea, and to confirm several details of the design and fabrica-tion process essential to the development of the device, including the implantation and bondingprocesses, and the junction termination and guard-ring structures at the edge of the active area.In moving from an idealized doping profile to one arising from the simulation of the fabricationprocess, the performance characteristics might be expected to change somewhat from those ofthe conceptual devices described in the previous section. Thus the basic performance char-acteristics are revisited in this section, after the description of the augmented simulation andoptimization procedure.In the wafer-to-wafer bonding approach, ion implantation is used to establish the highly-doped n+ and p+ regions, respectively, near the surface of two high-resistivity (1,0,0) wafers.The surfaces of these two wafers are abutted against one another, and the interface barrierarising from surface roughness and other imperfections is then removed by high-temperatureannealing, leading to a clean surface-to-surface bond through which charge (electrons and holes)can flow freely.The ion implantation energy for the p-type dopant (boron) was set to 735 keV, the maximumenergy achievable by the implantation apparatus. To avoid excessive ion channeling along crystalplane boundaries, the wafers are canted at an angle of 7 ° relative to the normal axis of the waferduring the implantation process. This leads to a most-likely ion penetration depth, and acorresponding peak in the doping profile, at a depth of approximately 0.9 µ m.To establish the 1.5 µ m separation between the peaks of the n+ and p+ doping profiles (seeTable 1), the implantation energy of the n-type dopant (phosphorus) was then tuned within theSentaurus simulation to produce a doping profile with a peak at a depth of 0.6 µ m (again fora cant of 7 ° ), corresponding to an implantation energy of 450 keV. Simulation of the resultingn-type and p-type doping profiles, as a function of annealing temperature (180 minute annealingepisode) is shown in Figure 7. For the simulation in this figure, an ion-beam fluence of 1 . × (1 . × ) particles per cm was assumed for the n-type (p-type) implantation process. Whilethe simulation suggests that some channeling effects remain that produce tails in the depthprofile, the bulk of the distribution is adequately peaked at the desired depth. The simulationsuggests that modifications to the doping profile during the annealing process are expected tobe minimal for temperatures below 950 ° C. 8igure 7: Gain layer doping profile as a function of annealing temperature (180 minute annealingepisode)The impact-ionization gain of a device making use of these implantation-beam energies wassimulated for a buried-junction depth of 5 µ m, and was found to be sufficiently insensitive tothe precise value of the implantation beam fluence. For example, Figure 8 shows the impact-ionization gain as a function of bias voltage for a range of boron ion-beam fluences, assuminga phosphorus ion-beam fluence of 1 . × . A total detector thickness of 50 µ m is assumed.The gain is seen to be relatively stable over a range of ±
20% in the corresponding doping level.A beam fluence of 1 . × ions per cm was found to produce the most slowly-varying,9nd therefore controllable, gain, and was chosen as the p-type implantation doping level for theprototype device.Figure 8: Deep-junction gain curve for various p+ implantation ion-beam fluences, assuming ann+ implantation fluence of 1 . × .A simulation of the charge-collection response (current as a function of time), for a planarDJ-LGAD fabricated as described above, is displayed in Figure 9. The pulse is shown for twovalues of the applied bias voltage (300 V and 400 V), for which the impact-ionization gainfactor is approximately 5 and 15, respectively. For this more realistic model of the DJ-LGAD, arise time of approximately 250 psec is observed, consistent with that observed for conventionalsurface-junction LGADs.In producing an operable device, it is essential to properly terminate the high-field semicon-ductor junction at the end of the active area in a manner that avoids early (low bias voltage)breakdown of the sensor. Simulation studies suggested that this can be achieved by terminatingthe n+/p+ semiconductor junction in the region immediately below the gap between the planarreadout electrode and a conductive guard ring encircling the readout electrode at a distanceof 30 µ m. Two types of junction termination are considered: one for which the upper (n+)and lower (p+) implants terminate at the same point (“symmetric termination”) and one forwhich the upper implant extends 15 µ m further towards the guard ring than does the lowerimplant (“asymmetric termination”). Figure 10 shows the two-dimensional doping profile of theasymmetric termination scheme; for the symmetric scheme, the n+ and p+ regions terminateat the same transverse position, half-way between the edges of the readout electrode and guardring.The electric field configuration expected for these symmetric and asymmetric doping pro-files, assuming an applied reverse bias of 300 V and an overall detector thickness of 50 µ m, isshown in the region of the junction termination in Figure 11. The asymmetric scheme is seento provide a lower and more spatially uniform field, especially in the region near the electrodeboundaries. The corresponding expectations for the dependence of leakage current upon biasvoltage is shown in Figure 12, for both the guard ring and readout electrodes. While both termi-nation configurations are expected to allow for operation well above 300 V before breakdown, asanticipated from Figure 11, the asymmetric termination scheme is expected to provide a signifi-cantly greater operating range than that of the symmetric scheme. A final junction-terminationstudy showed no difference in the detector breakdown properties when a 1-2 µ m thick p-spray10igure 9: Simulated charge collection response to a mip injection for bonded wafers dopedaccording to the procedures described in the text. The total sensor thickness was set to 50 µ m.structure with a doping density of approximately 5 × cm − was embedded in the surface ofthe device, interposing between the readout electrode and guard ring.11igure 10: Doping profile cross section for the asymmetric termination(top) and symmetrictermination(bottom) scheme. For the symmetric termination scheme, the n+ and p+ regionsterminate at the same point in the x coordinate, half-way between (15 µ m from) the right edgeof the guard ring and left edge of the readout pad.12igure 11: Electric field map in the region of the gain layer termination, assuming a bias voltageof 300 V and a sensor thickness of 50 µ m. The upper (lower) plot shows the expected electricfield for the asymmetric (symmetric) termination scheme. The electric field lines are shownin black, while the magnitude of the electric field is indicated by the color. For the upperplot (asymmetric termination), the region enclosed by the thin white contour is undepleted.However, this narrow undepleted region lies outside the active area of the sensor, and thusdoes not degrade the performance of the device relative to that of the device with symmetrictermination. 13igure 12: Expected dependence of the linear leakage current density vs reverse-bias voltage(I-V curve) for the readout electrode (upper) and guard ring (lower). The expectation for thesymmetric termination scheme is shown in red, while that for the asymmetric scheme is shownin blue. • Whatever we can have - first say what configurations and splits were fabricated. Wasp-apray used as simulated? 14
Acknowledgements
This work was supported by ... WRITE SBIR NUMBER. This work was supported by theUnited States Department of Energy, grant DE-FG02-04ER41286, and partially performedwithin the CERN RD50 collaboration. We also acknowledge helpful conversations with andconsideration by SCIPP faculty member Hartmut F.-W. Sadrozinski.
References [1] G. Pellegrini et al.,
Technology developments and first measurements of Low GainAvalanche Detectors (LGAD) for high energy physics applications , Nucl. Instrum. Meth.
A765 (2014) 12 – 16.[2] M. Carulla et al.,
First 50 µ m thick LGAD fabrication at CNM, 28th RD50 Workshop,Torino, Italy, June 7th 2016 , .[3] H. F. W. Sadrozinski et al., Ultra-fast silicon detectors (UFSD) , Nucl. Instrum. Meth.
A831 (2016) 18–23.[4] R. Padilla, C. Labitan, Z. Galloway, C. Gee, S. Mazza, F. McKinney-Martinez et al.,
Effect of deep gain layer and carbon infusion on LGAD radiation hardness , Journal ofInstrumentation (oct, 2020) P10003–P10003.[5] Y. Jin, H. Ren, S. Christie, Z. Galloway, C. Gee, C. Labitan et al., Experimental study ofacceptor removal in ufsd , Nuclear Instruments and Methods in Physics Research SectionA: Accelerators, Spectrometers, Detectors and Associated Equipment (2020) 164611.[6] S. Mazza et al.,
Proprieties of FBK UFSDs after neutron and proton irradiation up to ∗ neq/cm , JINST (2020) T04008, [ ].[7] Y. Zhao et al., Comparison of 35 and 50 µ m thin hpk ufsd after neutron irradiation up to × n eq /cm , .[8] Z. Galloway et al., Properties of HPK UFSD after neutron irradiation up to 6e15 n/cm , submitted to NIM A (2017) , [ ].[9] RD50 collaboration, “ https://rd50.web.cern.ch/rd50 .”[10] H. Sadrozinski, A. Seiden and N. Cartiglia, , Reports on Progress in Physics (2017) , [ ].[11] HL-LHC, “http://dx.doi.org/10.5170/CERN-2015-005.”[12]
ATLAS Collaboration collaboration,
Technical Design Report: A High-GranularityTiming Detector for the ATLAS Phase-II Upgrade , Tech. Rep. CERN-LHCC-2020-007.ATLAS-TDR-031, CERN, Geneva, Jun, 2020.[13] CMS collaboration,
A MIP Timing Detector for the CMS Phase-2 Upgrade , Tech. Rep.CERN-LHCC-2019-003. CMS-TDR-020, CERN, Geneva, Mar, 2019.[14] Z. Wang,
On the single photon counting (spc) modes of imaging using an xfel source , JINST (2015) C12013.[15] N. Cartiglia, A. Staiano, V. Sola, R. Arcidiacono, R. Cirio, F. Cenna et al., Beam testresults of a 16ps timing system based on ultra-fast silicon detectors , Nuclear Instrumentsand Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors andAssociated Equipment (2017) 83 – 88.1516] Z. Galloway, C. Gee, S. Mazza, H. Ohldag, R. Rodriguez, H.-W. Sadrozinski et al.,
Use of“lgad” ultra-fast silicon detectors for time-resolved low-kev x-ray science , NuclearInstruments and Methods in Physics Research Section A: Accelerators, Spectrometers,Detectors and Associated Equipment (2019) 5 – 7.[17] H. F.-W. Sadrozinski, A. Seiden and N. Cartiglia,
4d tracking with ultra-fast silicondetectors , Reports on Progress in Physics (dec, 2017) 026101.[18] G. Pellegrini et al., “Status of LGAD production at CNM.” Contribution to the 30 th RD50 Workshop, Krakow, Poland, https://indico.cern.ch/event/637212/contributions/2608652/attachments/1470919/2276240/pellegrini_rd50.pdf , 2017.[19] A. Apresyan, W. Chen, G. D’Amen, K. F. Di Petrillo, G. Giacomini, R. Heller et al.,
Measurements of an AC-LGAD strip sensor with a 120 GeV proton beam , JINST (2020) P09038, [ ].[20] E. Curr ˜A¡s, M. Carulla, M. Centis Vignali, J. Duarte-Campderros, M. Fern ˜A¡ndez,D. Flores et al., Inverse low gain avalanche detectors (ilgads) for precise tracking andtiming applications , Nuclear Instruments and Methods in Physics Research Section A:Accelerators, Spectrometers, Detectors and Associated Equipment (2020) 162545.[21] G. Paternoster, G. Borghi, M. Boscardin, N. Cartiglia, M. Ferrero, F. Ficorella et al.,
Trench-isolated low gain avalanche diodes (ti-lgads) , IEEE Electron Device Letters (2020) 884–887.[22] Synopsis Corporation, “Sentaurus Device: An Advanced Multidimensional DeviceSimulator.”