AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (2025)

Amanda L. Steinhebel  Regina Caputo  Daniel P. Violette  Anthony Affolder  Autumn Bauman  Carolyn Chinatti  Aware Deshmukh  Vitaliy Fadayev  Yasushi Fukazawa  Manoj Jadhav  Carolyn Kierans  Bobae Kim  Jihee Kim  Henry Klest  Olivia Kroger  Kavic Kumar  Shin Kushima  Jean-Marie Lauenstein  Richard Leys  Forest Martinez-Mckinney  Jessica Metcalfe  Zachary Metzler  John W. Mitchell  Norito Nakano  Jennifer Ott  Ivan Peric  Jeremy S. Perkins  Max R. Rudin  Taylor (K.W.) Shin  Grant Sommer  Nicolas Striebig  Yusuke Suda  Hiroyasu Tajima  Janeth Valverde  Maria Zurek

Abstract

A next-generation medium-energy gamma-ray telescope targeting the MeV range would address open questions in astrophysics regarding how extreme conditions accelerate cosmic-ray particles, produce relativistic jet outflows, and more.One concept, AMEGO-X, relies upon the mission-enabling CMOS Monolithic Active Pixel Sensor silicon chip AstroPix.AstroPix is designed for space-based use, featuring low noise, low power consumption, and high scalability.Desired performance of the device include an energy resolution of 5555keV (or 10% FWHM) at 122keV and a dynamic range per-pixel of 257002570025-70025 - 700keV, enabled by the addition of a high-voltage bias to each pixel which supports a depletion depth of 500500500500μ𝜇\muitalic_μm.This work reports on the status of the AstroPix development process with emphasis on the current version under test, version three (v3), and highlights of version two (v2).Version 3 achieves energy resolution of 10.4±3.2%plus-or-minus10.4percent3.210.4\pm 3.2\%10.4 ± 3.2 % at 59.5keV and 94±6plus-or-minus94694\pm 694 ± 6μ𝜇\muitalic_μm depletion in a low-resistivity test silicon substrate.

1 Introduction

Over the past several decades, silicon strip detectors (SSDs) have been a key detector technology used in gamma-ray and cosmic-ray telescopes such as the Fermi𝐹𝑒𝑟𝑚𝑖Fermiitalic_F italic_e italic_r italic_m italic_i Large Area Telescope (LAT)[1],the Alpha Magnetic Spectrometer (AMS)[2],and DArk Matter Particle Explorer (DAMPE)[3].Breakthroughs in particle physics instrumentation have enabled the development of High Voltage-CMOS (HVCMOS) monolithic active pixel sensors (MAPS)[4], which have significant advantages over other silicon-based detectors (such as SSDs).MAPS have signal amplification and readout circuitsembedded in the sensor, enabling reductions in pixel power consumption without the need for a separate readout Application Specific Integrated Circuit (ASIC).The design reduces the overall mass of the detector system and limits the passive material in the active detector volume while also improving spatial resolution.

In particular for soft gamma-rays (<10absent10<10< 10 MeV), it is imperative to have two-dimensional hit location information in a single detector layer[5].These detector capabilities make them a particularly compelling technology for future gamma-ray telescopes (see Refs.6, 5 for more details).

AstroPix is a HVCMOS MAPS being developed for space-based mission concepts such as the All-sky Medium Energy Gamma-ray Observatory eXporer (AMEGO-X) [5].Inspiration is drawn from the ATLASPix chip, which was designed to detect charged particles in the inner detector of the ATLAS Experiment[7].The AstroPix project has subsequently coordinated incremental development away from the ATLASPix designs toward a final version which will be optimized for a space environment (see Table1).The current device under test and subject of this work is the third iteration, or AstroPix_v3.The performance of AstroPix_v1 and AstroPix_v2 has been documented in Refs.8, 9, 10, 11.

In addition to reviewing AstroPix_v3 operation and functionality, this work will overview noise and energy resolution performance.Current-voltage and capacitance-voltage curves are presented with discussion of sensor noise and depletion.AstroPix_v3 will be the first flight-tested AstroPix chip as the main component of a sounding rocket paylod, the Astropix Sounding rocket Technology dEmonstration Payload (A-STEP) (see Sec.7).As such, operation and charicterization results relevant for this upcoming flight will be emphasized.

AstroPix has been tested in multiple beam environments, including with a 120120120120GeV proton beam at the Fermilab Test Beam Facility.A detailed report of this testing and results including the measurement and identification of minimum ionizing particles, alternate energy calibration method, and alternate measurement of depletion depth will be reported in an upcoming independent publication.

The paper is outlined as follows:Section2 describes the AstroPix_v3 chips;Section3 illustrates the benchtop test setup;Section4 describes the amplitude and impact of noise and dark count rate;Section5 details the AstroPix_v3 characterization, including calibration, energy resolution, and depletion depth;Section6 summarizes radiation testing of AstroPix_v2 in a heavy ion beam;and finally Section7 summarizes the work and outlines future outlook and applications of AstroPix.

2 AstroPix_v3

HVCMOS MAPS detectors were developed by Ivan Peric more than a decade ago [4] primarily for particle physics applications.The Karlsruhe Institute of Technology (KIT) ASIC and Detector Laboratory (ADL) has continued this work, advancing the technology forward[12, 13, 14].ADL’s experience with chips such as MuPix and ATLASpix inform the development of AstroPix.The design evolution over the course of several iterations has culminated in the first full-scale flight prototype chip: AstroPix_v3.An illustration of key properties of each AstroPix version, including AstroPix_v3, is shown in Table1.For the first time, AstroPix_v3 uses the full 2×2222\times 22 × 2cm2 reticle and features a 35×35353535\times 3535 × 35 pixel matrix with a pixel pitch of 500×500500500500~{}\times~{}500500 × 500μ𝜇\muitalic_μm2.A 300×300300300300\times 300300 × 300μ𝜇\muitalic_μm2 high voltage deep n𝑛nitalic_n-well (DNW) protects the embedded CMOS circuits and creates a bias junction with the p𝑝pitalic_p-type bulk silicon, leading to a depletion region with HV application (see Fig.1).AstroPix_v3 uses a standard high voltage CMOS process with a deep n𝑛nitalic_n-well and unthinned bulk silicon wafers with a thickness of 720μ𝜇\muitalic_μm.

ATLASPixAstroPix_v1AstroPix_v2AstroPix_v3AstroPix
(Measured)[6, 15][8][8, 10][11, 16, 17](Goal)[5]
Eres(FWHM)7.3±1.2%plus-or-minus7.3percent1.27.3\pm 1.2\%7.3 ± 1.2 %20±7.4%plus-or-minus20superscriptpercent7.420\pm 7.4\%^{*}20 ± 7.4 % start_POSTSUPERSCRIPT ∗ end_POSTSUPERSCRIPT15±3plus-or-minus15315\pm 315 ± 3%10.4±3.2%plus-or-minus10.4superscriptpercent3.210.4\pm 3.2\%^{\lozenge}10.4 ± 3.2 % start_POSTSUPERSCRIPT ◆ end_POSTSUPERSCRIPT10%
at 30.1 keVat 30.1 keVat 30.1 keVat 59.5 keVat 122 keV
Pixel pitch [μ𝜇\muitalic_μm]150×5015050150\times 50150 × 50175×175175175175\times 175175 × 175250×250250250250\times 250250 × 250500×500500500500\times 500500 × 500500×500500500500\times 500500 × 500
Thickness [μ𝜇\muitalic_μm]100725725725525
Depletion [μ𝜇\muitalic_μm]48Not measuredNot measured94±6plus-or-minus94superscript694\pm 6^{\triangle}94 ± 6 start_POSTSUPERSCRIPT △ end_POSTSUPERSCRIPT500
Dynamic range5-3214-12214-8014-20025-700
[keV]
Analog power12014.73.41.061.0
[mW/cm2]
Digital power409.93.753.060.5
[mW/cm2]
AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (1)

The signal path is shown in Fig.2.Each pixel contains a charge-sensitive amplifier (CSA) and comparator.For testing and comparison purposes, the first three columns feature a PMOS amplifier whereas the rest implement the standard NMOS amplifiers.These PMOS columns were found to impair data quality with higher noise rates than NMOS columns, and are not considered in this work’s analysis or for future design iterations.A simplified schematic for pixel operation and timestamp generation is shown in Fig.3.

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (2)
AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (3)

The in-pixel comparator enables self-triggering.A Time Over Threshold (ToT) measurement gives proxy for the deposited charge by recording the duration of the signal amplitude while over a user-defined threshold.In AstroPix_v3, a global comparator threshold value is set for the full array111AstroPix_v4 and subsequent versions include Tune DACs to enable individual pixel threshold setting [18]..

When a pixel comparator passes a signal that exceeds the threshold, the chip processes this information and lowers an ‘interrupt’ signal.To keep the number of readout channels low, and therefore reduce power consumption, the pixel comparator outputs per row and column are OR wired (Fig.2).A low ‘interrupt’ signal indicates that data is ready for collection222AstroPix_v3 readout utilizes a Field Programmable Gate Array (FPGA) which receives and stores data by the software-based data acquisition (DAQ) system (see Sec.3)The comparator threshold is set to achieve maximal detector efficiency by setting the lowest possible detector threshold while still minimizing triggering from noise fluctuations.

The matrix digitisation is done via 12 bit counters driven by a 200200200200MHz clock, implemented both per row and column to measure ToT (Fig.2).Resulting ToT values fall within the μ𝜇\muitalic_μs range, therefore the ns-scale resolution from this clock is much smaller than the ToT noise.An 8 bit time of arrival (ToA) timestamp is driven by a 2222MHz clock (Fig.2).

Once collected at the pixel level, the digital signal is stored in a hit buffer and ultimately read out over SPI (Fig.3).Each digitized signal returns a 5-byte data packet - a 4 bit header for data integrity, 11 bit pixel address, 12 bit ToT, and 8 bit ToA.

In addition to digitized output, an “analog" output can also be accessed.This is sent from the in-pixel amplifier prior to the standard digitization readout path (including the in-pixel comparator) and can be read from a single pixel at a time from the bottom row of pixels (row0).The analog output signal provides an important cross-check on the energy resolution of the detector and front-end amplifier because it is not limited by the digital resolution of the readout.

Power consumption has decreased with each chip version (see Table1).Strategies such as limiting clock distribution to the matrix periphery, optimizing the bias circuits, and increasing pixel size account for some power saving.Analog power consumption has been the emphasis of past design iterations, where 97%percent9797\%97 % analog power reduction was achieved from AstroPix_v1 to the current 1.061.061.061.06mW/cm2 draw of AstroPix_v3.Future versions of AstroPix [18] will continue this trend of reduced power consumption with a renewed focus on the digital power draw.A brief discussion of the planned path forward is included in Sec.7.

Figure4 (left) shows a single AstroPix_v3 mounted to a carrier board.Chips are diced from the wafer as single arrays and in a 2×2222\times 22 × 2-array configuration called a quad chip, shown in Fig.4 (right) with a total area of 3.81×3.933.813.933.81\times 3.933.81 × 3.93cm2.This quad chip, utilizing four independent AstroPix arrays, will be the building block of larger AstroPix-based structures including A-STEP.18%percent1818\%18 % of fabricated AstroPix_v3 chips (180 chips) underwent Quality Control testing following fabrication, dicing from the wafer, mounting to a carrier board, and wirebonding.97.1%percent97.197.1\%97.1 % of 200400Ω200-400~{}\Omega*200 - 400 roman_Ω ∗cm resistivity (Okmetic) chips passed the Quality Control testing, with the few failures due mostly to low breakdown voltage.This exemplifies the scalability that AstroPix is designed for, where the formation of large format structures with straightforward integration, operation, and readout is achievable.Power is delivered to each chip in a quad-chip via custom-designed bus bars to supply the chips in parallel.Data are sent between arrays via wirebonds that enable a daisy chain and is delivered off-chip with a quasi-Serial Peripheral Interface (SPI).Quad chips of AstroPix_v3 are also under test with a suite of software and firmware developed for the readout of large arrays and multiple layers333https://github.com/AstroPix/astep-fw.

The 20202020MHz SPI interface clock transmits bytes in 200200200200ns over two MISO lines, leading to a single chip latency of 400400400400ns.Larger instrument designs can daisy chain up to 32323232 AstroPix chips to one SPI bus, limited by available chip identification bits, which increases the per-bus latency up to 8μ8𝜇8~{}\mu8 italic_μs.The 2222MHz timestamp clock turns over after 128128128128μ𝜇\muitalic_μs, so the latency does not impact timing resolution.

All iterations of AstroPix were fabricated using TSI Semiconductors’ 180180180180nm process and AstroPix_v3 was delivered in October 2022444TSI Semiconductors has subsequently been acquired by Bosch Semiconductor LLC and the foundry business has ceased. Subsequent iterations of AstroPix will be fabricated at a different foundry such as AMS in Austria or LFoundry in Italy..The final design must deplete 500500500500μ𝜇\muitalic_μm and thus a global high-voltage (HV) bias is applied to a substrate contact along the chip edge on the frontside (Fig.1).A backside bias is not utilized in order to simplify chip processing and eventual integration into large format structures.Full depletion will be achieved with a high-resistivity (>5000Ω>5000~{}\Omega*> 5000 roman_Ω ∗cm) silicon wafer thinned to 525525525525μ𝜇\muitalic_μm.Neighboring pixels are isolated by a p𝑝pitalic_p-stop isolation ring (blue in Fig.1), separated by a distance of 100100100100μ𝜇\muitalic_μm.This large separation between the DNW guard ring (dark green in Fig.1) and the isolation p𝑝pitalic_p-stop allows for operation at high voltages with breakdown occurring at 400400-400- 400V.The bulk substrate between implants sustains the high voltage, creating a smooth depletion layer through the bulk without introducing inter-pixel dead space.

AstroPix_v3 was fabricated on 25±8plus-or-minus25825\pm 825 ± 8, 200400200400200-400200 - 400, and 25,000±8,00025plus-or-minus000800025,000\pm 8,00025 , 000 ± 8 , 000Ω\Omega*roman_Ω ∗cm wafers provided by the TSI Foundry, Okmetic, and Topsil respectively.The highest-resistivity AstroPix_v3 chip was targeted for exploring the goal depletion depth, but the medium-resistivity chip (henseforth referred to by its manufacturer, Okmetic) will be the focus of this paper (see Sec.5.2 for more details).The depletion value stated in Table1 reflects this Okmetic wafer where full depletion is not expected.Further details of AstroPix characterization can be found in Refs. 11, 16, 17.

A current-voltage (IV) curve of the biasing HV used to deplete the substrate for a representative Okmetic chip is shown in Fig.5.Breakdown occurs between 380380-380- 380V and 400400-400- 400V depending on the wafer555Breakdown occurs around 200200-200- 200V with the 25±8Ω25\pm 8~{}\Omega*25 ± 8 roman_Ω ∗cm wafer..Applications of AstroPix require low power draw and a bias leakage current less than 1μ1𝜇1~{}\mu1 italic_μA, allowing operation of an Okmetic AstroPix_v3 chip up to 400400-400- 400V.A larger bias enables more complete depletion (more details in Sec.5.2), so the energy calibration reported in Sec.5.1 uses a bias of -350V which incorporates a safety margin with respect to the breakdown voltage.Otherwise, a stable bias of 150150-150- 150V can be assumed for the remainder of the studies in this paper unless explicitly stated.

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (6)

3 Experimental Setup

The equipment used to test AstroPix, as shown in Fig.6, includes a custom built GEneric Configuration and COntrol (GECCO [19]) Data Acquisition System, a NexysVideo (Xilinx Artix-7) FPGA, and a carrier board for the integration of the chips into the GECCO system (along with an oscilloscope and power supplies). Data is read off the sensor by a software-based DAQ system.

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (7)

In-pixel circuitry operation can be fine-tuned through user-definable values configured through Digital-to-Analog Converters (DACs).Currents such as that fed to the source follower which sets the amplifier operating point, and voltages such as comparator voltage baseline impact signal shaping and power consumption.Initial values are set from simulation and were further optimized for performance while on the bench.Shaping the pulse width via the in-pixel band pass filter was a focus, to not limit the dynamic range given by the maximum measureable value of 20.5similar-toabsent20.5\sim 20.5∼ 20.5μ𝜇\muitalic_μs by the 12121212bit 200200200200MHz ToT counter.Following optimization, all testing utilizes the same set of DAC values.This can be assumed for all the results explored in this work.

The chip can operate in “high gain" or “high dynamic range" mode, which impacts the in-pixel CSA gain.The studies in this work all operate in “high dynamic range" mode which causes bilinear output from the CSA.Therefore, a linear response through the full dynamic range is not expected.

Custom firmware666https://github.com/AstroPix/astropix-fw and software777https://github.com/AstroPix/astropix-python allow for interfacing with the chip, properly setting clocks and configuring chip operational settings, and executing data collection and storage.

All studies in this work exclusively consider this digital data.The OR wiring results in separate hits containing row and column information.Events are considered for data analysis only if a pair of row and column hits record a timestamp within 1 clock count (5555ns) and ToT values within 0.150.150.150.15μ𝜇\muitalic_μs. In this way, the response of individual pixels can be considered.

The work presented in this paper is a collection of results from a dedicated international collaboration with worldwide testing campaigns.As such, minor differences in experimental setup may be present for different tests presented here.Unless explicitly stated in a section, the settings in Table2 can be assumed for all studies in this work.

Wafer nameResistivityComparatorBiasMax. dark
[Ω\Omega*roman_Ω ∗cm]Thresh. [mV][-V]count rate [Hz]
Okmetic200-4002001502

Additional settings include operation in ‘high dynamic range‘ gain mode, optimized voltage and current DAC settings, and disabling of the first three pixel columns with PMOS amplifiers.

4 Noise Studies

‘Noise’ hits occur when particles from background sources interact with the detector, such as cosmic ray interactions or naturally occurring background radiation.Fluctuations in the electronics can also trigger a comparator readout, even without a particle interaction.This is a ‘dark count’ and is measured as a rate, the dark count rate (DCR).The goal of these studies are to determine the percentage of pixels which are sensitive within the dynamic range, understand the dark count rate at different thresholds, and set an optimized threshold voltage such that DCR is minimized.

The analog baseline, as read off the amplifier from selected pixels, shows random fluctuations up to 50505050mV over a nearly constant baseline.This constant noise floor is due to detector electronics and consists of shot noise from the sensor diode, and thermal and flicker noise from the electronics.The latter two sources are thought to be sub-dominant, as they carry a dependence on capacitance and the AstroPix_v3 pixel capacitance is roughly 1111pF.Large fluctuations with a low rate are likely noise from the environment in which the sensor is operated.The dark count rate can be mitigated during data collection by setting the comparator threshold higher than 50505050mV above baseline.A 22.122.122.122.1keV photopeak from cadmium-109 is visible in measurements made in Sec.5 with a 200200200200mV threshold, so the 50505050mV fluctuations can be disregarded and do not impact the dynamic range.The impact of increasing threshold on DCR is shown in Fig.7.

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (8)

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (9)

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (10)

Threshold values of 200200200200mV still allow the detection of 25252525 keV signals as required in Table1 (see Section5 for details).With nominal operational settings, <0.5absent0.5<0.5< 0.5% of pixels per array are too noisy for data collection and must be masked (their comparators disabled).

The variation in dark count rate also illustrates a degree of variation between pixels within the same array.Process variations and device mismatch lead to this natural variation in amplifier gain, leading to pixels responses varying 2035%20percent3520-35\%20 - 35 %.Individual-pixel calibration therefore is required to correct for this variation.Section5 outlines this calibration strategy and its results.

The in-pixel comparator does not collect sub-threshold hits.Studies in Ref.6 show that charge sharing between pixels does not induce above-threshold hits in neighboring pixels.However, the threshold is set to maximize the dynamic range and has been shown (in Sec.5) to enable measurement below the 25252525keV requirement.In this way, some degree of charge splitting can be accounted for provided that there is enough to trigger a neighboring comparator operating at this low global threshold.Simulation studies are planned to investigate the potential impact of charge sharing and its effect on energy resolution.

The studies of this section illustrate that DCR can be reduced to tolerable rates with AstroPix devices.This can be done by carefully setting a global threshold which minimizes DCR while still providing measurements within the required dynamic range of 257002570025-70025 - 700keV and identifying and masking noisy pixels with large DCRs.These noisy pixels have no geometric dependence within the array and have not shown induced charge sharing.With the optimal operation described here and standard run settings from Table2, a single AstroPix_v3 array measures a total DCR below 2222Hz.This measured rate is acceptable for future AstroPix applications including A-STEP and large-format future-observatories such as AMEGO-X.Additionally, future AstroPix versions are expected to decrease DCR with the ability to set thresholds at an individual pixel level [18].

5 Energy Resolution

A high bias voltage enables more complete depletion and in general more efficient charge collection.Here we consider an energy calibration conducted at 350350-350- 350V bias, as described in Ref.16.The current draw off the bias line is 50similar-toabsent50\sim-50~{}∼ - 50nA888The IV curve from this study is shown in Fig.5..

5.1 Energy Calibration and Resolution

The energy calibration procedure is described in detail in Ref.16, and is applied to one Okmetic chip.After masking 8 noisy pixels, measurements of radioactive sources including Cd-109, Ba-133, Am-241, and Co-57 were made.Photopeaks could be identified from individual pixel spectra between 22.212222.212222.2-12222.2 - 122keV.A feature indicative of a Compton edge at around 220220~{}220220 keV was present in the data.Calibration curves for each pixel relate the expected photopeak energy and the mean measured ToT with a function of the form

y=aE+b[1exp(E/c)]+d,𝑦𝑎𝐸𝑏delimited-[]1𝐸𝑐𝑑y=a*E+b*\left[1-\exp{(-E/c)}\right]+d~{},italic_y = italic_a ∗ italic_E + italic_b ∗ [ 1 - roman_exp ( - italic_E / italic_c ) ] + italic_d ,

where E𝐸Eitalic_E is the true photopeak energy and y𝑦yitalic_y is the uncalibrated ToT value.The functional form involves a low-energy linear component and higher-energy exponential decay.This form is motivated by the in-pixel charge-sensitive amplifier which operates in two different gain regimes, creating a bilinear gain structure.The amplifier saturation effect is also reflected.One example calibrated pixel is shown in Fig.8 and compared to an example pixel from AstroPix_v2.The difference in slope is due in part to differences in front-end amplifier tuning.Figure9 shows a calibrated Am-241 spectrum for all 996 calibrated pixels after individual pixel calibration.The 59.559.559.559.5keV photopeak is clearly visible.Further work is underway to determine its source, though it is expected to be artificial.

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (11)
AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (12)

When considering one 35×35353535\times 3535 × 35 array, 92.4%percent92.492.4\%92.4 % of pixels achieve the low-energy floor requirement of 25252525keV sensitivity.44%percent4444\%44 % of pixels meet the energy resolution requirement of 5.95.95.95.9keV at 59.559.559.559.5keV with a median full-width half-max (FWHM) of 6.26.26.26.2keV (10.4%percent10.410.4\%10.4 %).The lack of a confident feature identification at 220similar-toabsent220\sim 220∼ 220keV makes the measured dynamic range 142001420014-20014 - 200keV.

The high-end of the dynamic range is hindered in AstroPix_v3 due to incomplete depletion (see Section5.2) and amplifier saturation.Simulations of the amplifier in “high dynamic range" mode showed that its output amplitude saturates at 250similar-toabsent250\sim 250∼ 250keV.This is consistent with observations which saw maximum energies (that could not be fully calibrated) of 220similar-toabsent220\sim 220∼ 220keV.A subsequent iteration of AstroPix implements test pixels with a dynamic feedback capacitance realized with an NFET device [20], which displays no saturation at energies over 700700700700keV while maintaining performance through the full dynamic range.Additionally, more ToT bits are allotted to prevent a dynamic range loss due to overflow in the digitization step in future versions.

5.2 Impact of Depletion

The highest-resistivity AstroPix_v3 design (25000±8000Ω25000\pm 8000~{}\Omega*25000 ± 8000 roman_Ω ∗cm) was targeted for exploring the goal depletion depth of 500500500500μ𝜇\muitalic_μm within a 100100-100- 100V bias range.The two other resistivities are included for comparison with previous generations and consideration for further studies.The different depletion depths expected for each wafer resistivity is shown in Fig.10.

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (13)

However, this high-resistivity substrate could not be thoroughly tested due to a high leakage current.An inherent voltage gradient coupling through conductive mounting to the test board contributes to a low breakdown voltage of <1absent1<-1< - 1V, but isolating the backside does not resolve the high current.The cause of this current is an active area of investigation.

Detector capacitance plays an important role in the readout system performance.By design, it is connected to the input of the charge-sensitive amplifier.The amplifier noise is a monotonically increasing function of the capacitance.Therefore, reducing the capacitance through higher bias reduces noise, in addition to increasing the ionization efficiency by augmenting the depletion region.

A capacitance-voltage (CV) curve of the 200400Ω200-400~{}\Omega*200 - 400 roman_Ω ∗cm Okmetic substrate is shown in Fig.11.Notably, the function shape at low voltages is different from the linear dependence of 1/C21superscript𝐶21/C^{2}1 / italic_C start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT on bias voltage.We attribute this to the depletion area development.At near-zero bias the depleted regions starts to grow from the implants, therefore the depleted area is smaller than the full chip size.For this reason, initially 1/C21superscript𝐶21/C^{2}1 / italic_C start_POSTSUPERSCRIPT 2 end_POSTSUPERSCRIPT grows sub-linearly with the bias voltage until the depleted region between neighboring pixels reconnect.The other feature is a sharp capacitance change around 150150-150- 150V.It is dependent on the test frequency, indicating a possible relation to surface charge.Both features are under further investigation, however the functional shape of the CV curve suggests under-depletion and isolated depleted regions under each implant in the voltage range tested.

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (14)

Estimations of the depletion depth that the Okmetic AstroPix_v3 can achieve was tested in Ref.16 by measuring the count rates of X-ray sources.The method exploits a rather small path length of the daughter photo-electron in silicon, of less than 20202020μ𝜇\muitalic_μm, leading to proportionality of the rate and the depleted volume.An average depleted depth is extracted using the pixel area.The drawback of this method is the need to rescale the rate to account for low-energy part of the photon spectrum.This indirect measurement results in 60±3plus-or-minus60360\pm 360 ± 3μ𝜇\muitalic_μm depletion at 150150-150- 150V bias and 94±6plus-or-minus94694\pm 694 ± 6μ𝜇\muitalic_μm depletion at 350350-350- 350V bias, which agrees with the PN junction model curve with the uncertainty in the resistivity.Fitting a PN junction model to the data reproduces a resistivity of 236.8±2.4Ω236.8\pm 2.4~{}\Omega*236.8 ± 2.4 roman_Ω ∗cm.

A direct measurement of depletion depth from the same wafer using the Edge Transition Current Technique [22] was conducted at the Santa Cruz Institute for Particle Physics in November 2023.Current results from this study are detailed in Ref.21 and Ref.17.The results are consistent with Ref.16 over the voltage range tested, however further analysis is ongoing.

6 Heavy-Ion Radiation Testing

A previous iteration of AstroPix, AstroPix_v2, was tested for radiation hardness and single-event effects (SEEs).The digital periphery design was not changed between the two versions.Like AstroPix_v3, the AstroPix_v2 chips tested were fabricated on an Okmetic (200400Ω200400Ω200-400~{}\Omega200 - 400 roman_Ω*cm) substrate.The presence of the on-chip digital periphery motivates this testing, as that area may be the most susceptible.Two classes of SEE were monitored:

  1. 1.

    Single-event latchup (SEL), a catastrophic event leading to runaway power draws due to parasitic switching of the CMOS component transistors [23], and

  2. 2.

    Single-event functional interrupt (SEFI), a temporary state where radiation causes bit flips which may degrade data or configuration but a system reset can restore nominal operation.

6.1 Experimental Setup

The radiation tolerance of AstroPix_v2 was tested in June 2022 at the Lawrence Berkeley National Laboratory (LBNL) Berkeley Accelerator Space Effects (BASE) Facility 88” cyclotron [24].The beam provides a cocktail of ions with 16 MeV/amu tune shown in Table3.A range of ions with differing atomic masses provides a range of testable linear energy transfer (LET), and tilting the detector plane relative to the ion beam direction of propagation provides an effective LET (the surface-incident LET in silicon at normal incidence divided by the cosine of the detector tilt angle).Effective range is the ion penetration depth in silicon calculated perpendicular to the surface of the sensor (penetration range multiplied by the cosine of the detector tilt angle).

IonInitial EnergyAir gapSensor tiltSurface energyEffective LETEffective
[MeV][mm]angle [deg][MeV±1plus-or-minus1\pm 1± 1 STD][MeV*cm2/mg]range [μ𝜇\muitalic_μm]
40Ar+14642200555±1.0plus-or-minus5551.0555~{}\pm~{}1.0555 ± 1.08.0206
63Cu+221007200805±1.1plus-or-minus8051.1805~{}\pm~{}1.1805 ± 1.119141
4523737±1.9plus-or-minus7371.9737~{}\pm~{}1.9737 ± 1.921115
78Kr+2813172001021±2.9plus-or-minus10212.91021~{}\pm~{}2.91021 ± 2.928132
4523921±3.0plus-or-minus9213.0921~{}\pm~{}3.0921 ± 3.032107
124Xe+4319752001341±4.6plus-or-minus13414.61341~{}\pm~{}4.61341 ± 4.65798
35231216±4.9plus-or-minus12164.91216~{}\pm~{}4.91216 ± 4.964102

Spaceflight standards [25] require instruments to survive catastrophic latchup up to a LET of 6075607560-7560 - 75MeV*cm2/mg and fluence of 1×1071superscript1071\times 10^{7}1 × 10 start_POSTSUPERSCRIPT 7 end_POSTSUPERSCRIPTcm-2.Dosimetry is provided from BASE through calibration of a set of four photomultiplier tubes (PMTs) to a center PMT that is then removed during irradiation.

Figure12 shows AstroPix_v2 on the left, 20202020mm away from the beam pipe.The chip carrier board and auxiliary boards are covered with aluminum to protect active electronic components from recoil ions.Digital data was collected from the full array (with 35%percent3535\%35 % of pixels masked to reduce noise), as well as analog data from the bottom left corner pixel.Four input voltage rails and the high voltage bias were also monitored during irradiation for SEL.Non-destructive SEFI events could present as features in either power draw or returned data.

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (15)

High- and low-flux data sets were collected, where the average high flux 2.05×104absent2.05superscript104\leq 2.05\times 10^{4}≤ 2.05 × 10 start_POSTSUPERSCRIPT 4 end_POSTSUPERSCRIPT/cm2/s and average low flux 123.1absent123.1\leq 123.1≤ 123.1/cm2/s.The total fluence of high flux runs was 1×1071superscript1071\times 10^{7}1 × 10 start_POSTSUPERSCRIPT 7 end_POSTSUPERSCRIPT/cm2 unless an destructive event occurred (see Sec.6.2).

6.2 Radiation Results

No immediately destructive high-current events occurred, suggesting AstroPix_v2 has a destructive SEL threshold greater than 64646464MeV*cm2/mg.The device is susceptible to recoverable soft errors with a LET threshold above 19191919MeV*cm2/mg.It is therefore unlikely that proton-induced secondary ions will cause on-orbit SEEs, assuming a low-inclination Low-Earth Orbit.

Three potential classes of SEFI events were identified, and one ("Class B") has subsequently been identified as a software-based error in the DAQ code.The remaining two classes of events are potentially flipped bits in either the analog configuration ("Class A") or the returned digital data ("Class C") leading to loss of analog signal or digitized data corresponding to nonphysical pixel locations.Each class was corrected with a reset or power cycle of the chip.

Cross sections for each data collection run are calculated by fitting the data to a Weibull curve [26] of the form

F(x)=A(1e[(xx0)/w]s),𝐹𝑥𝐴1superscript𝑒delimited-[]𝑥subscript𝑥0𝑤𝑠F(x)=A\left(1-e^{-\left[(x-x_{0})/w\right]s}\right)~{},italic_F ( italic_x ) = italic_A ( 1 - italic_e start_POSTSUPERSCRIPT - [ ( italic_x - italic_x start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT ) / italic_w ] italic_s end_POSTSUPERSCRIPT ) ,(6.1)

where x𝑥xitalic_x is the effective LET, A𝐴Aitalic_A is the limiting or plateau cross section, x0subscript𝑥0x_{0}italic_x start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT is an onset parameter such that F(x)=0𝐹𝑥0F(x)=0italic_F ( italic_x ) = 0 for x<x0𝑥subscript𝑥0x<x_{0}italic_x < italic_x start_POSTSUBSCRIPT 0 end_POSTSUBSCRIPT, w𝑤witalic_w is a width parameter, and s𝑠sitalic_s is a dimensionless exponent.Figure13 shows conservative cross sections assuming all three event classes were SEFIs.Asterisks represent Class A events and blue filled circles represent Classes B and C.The light blue circles at high LET and low cross section are from runs with an average flux of 4×104absent4superscript104\leq 4\times 10^{4}≤ 4 × 10 start_POSTSUPERSCRIPT 4 end_POSTSUPERSCRIPT/cm2/s, whereas the dark blue points were collected with an average flux of 112absent112\leq 112≤ 112/cm2/s.The distinction is not made at lower LET, as this flux effect was not found with lower LET beams for similar fluences.To account for these flux effects when calculating an upper bound on-orbit SEFI rate, the data have been fit with two Weibull curves and event rates summed.Both curves consider all data with LET<35absent35<35< 35MeV*cm2/mg.Curve 1 additionally includes only dark blue (high flux) data above 35353535MeV*cm2/mg whereas curve 2 includes only light blue (low flux) high-LET data.The fits are intentionally worst-case with steep increases and saturation levels at the highest cross section data point as opposed to an average.Corresponding Weibull parameters are given in Table4.

AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (16)
CurveSaturation CrossOnset LETPowerWidth
Section [cm-2][MeV*cm2/mg](s)(w)
13.00×1073.00superscript1073.00\times 10^{-7}3.00 × 10 start_POSTSUPERSCRIPT - 7 end_POSTSUPERSCRIPT1851
26.00×1046.00superscript1046.00\times 10^{-4}6.00 × 10 start_POSTSUPERSCRIPT - 4 end_POSTSUPERSCRIPT3157

Using CREME96 [27], a worst-case galactic cosmic ray environment of interplanetary space at solarminimum was used to determine a SEFI rate.The sensitive volumes defined in CREME96 had x- and y-dimensions equal to the square root of the saturated cross section; each volume was assigned a depth of 2 μ𝜇\muitalic_μm.Although the pixel depletion volume is approximately 70 μ𝜇\muitalic_μm thick at the voltage utilized for testing, the CMOScircuitry that is potentially responsible for the SEFIs will have a shallower sensitive volume.Within CREME96, this shallower volume yields higher event rates than would a deeper sensitive volume designation, conservatively bounding the rate.Total event rates for the worst-caseassumptions described above are given in Table5.

SEFI rate per daySEFI rate per year
2.386×1042.386superscript1042.386\times 10^{-4}2.386 × 10 start_POSTSUPERSCRIPT - 4 end_POSTSUPERSCRIPT8.717×1028.717superscript1028.717\times 10^{-2}8.717 × 10 start_POSTSUPERSCRIPT - 2 end_POSTSUPERSCRIPT

This result is unique to AstroPix_v2, though the digital block of AstroPix_v2 is identical to that of AstroPix_v3.The satisfactory results raise confidence in future designs.Future AstroPix versions will be tested in an identical manner to ensure continued compliance with spaceflight standards.

7 Conclusion and Outlook

The AstroPix design program is strong and robust as advances in the design are consistently made toward realizing a final version.AstroPix versions one through three show consistent improvements in key metrics such as power consumption and energy resolution, as shown in Table1 however the design process is ongoing.

This paper has presented an overview of AstroPix_v3 HVCMOS design and benchtop operation.Current-voltage measurements illustrated chip properties, including high-voltage breakdown between 380400380400-380-400- 380 - 400V.An overview of energy resolution studies and results from Ref.16 were presented, where 10.4±3.2%plus-or-minus10.4percent3.210.4~{}\pm~{}3.2\%10.4 ± 3.2 % FWHM energy resolution at 59.559.559.559.5keV was achieved using a medium-resistivity 200400Ω200-400~{}\Omega*200 - 400 roman_Ω ∗cm substrate.The measured 94±6μplus-or-minus946𝜇94~{}\pm~{}6~{}\mu94 ± 6 italic_μm depletion of this substrate is discussed, and plans for increasing this depletion depth to the designed 500μ500𝜇500~{}\mu500 italic_μm were presented.Radiation testing with a cocktail of ions was performed with AstroPix_v2 which shares an identical digital bloc to AstroPix_v3, and no catastrophic events were detected.Single event functional interrupt rates were estimated to be at the order of 104superscript10410^{4}10 start_POSTSUPERSCRIPT 4 end_POSTSUPERSCRIPT per day in the planned AMEGO-X orbit, which is a tolerable rate.

The first space-based test of AstroPix will be the Astropix Sounding rocket Technology dEmonstration Payload (A-STEP), featuring three AstroPix_v3 quad chips to be flown on a sounding rocket in 2025 [28].The first large-format test of AstroPix is the AMEGO-X prototype tower ComPair2 [29], which is intended to fly on a high-altitude balloon.The other AMEGO-X subsystems and operations team take heritage from the 2023 ComPair flight [30].ComPair2 features 10 tracker layers with nearly 100 AstroPix_v3 chips per layer, which will interact with a unified trigger system to return science data.The results from this publication serve as a baseline for AstroPix_v3 operation, and will provide the basis for A-STEP design, optimization, calibration, and analysis.

At the time of writing, AstroPix_v4 has been fabricated and is undergoing preliminary testing [18].This testing has informed the submission of AstroPix_v5 during the 2025 calendar year.Continued improvements in power consumption and dynamic range are expected with the elimination of an external fast 200200200200MHz clock and addition of dynamic feedback capacitance.

The AstroPix project benefits from the expertise of international collaborators in multiple fields of physics and engineering.Though not currently a final design, each AstroPix version improves upon the previous.The ultimate design will revolutionize γ𝛾\gammaitalic_γ-ray astronomy, especially in the elusive MeV range.

Acknowledgments

The authors would like to acknowledge the contributions of engineers and technicians at all participating intuitions, including but in no way limited to Kenneth Simms, David Durachka, Ryan Boggs, Timothy Cundiff, and Kirsten Affolder.

This work is funded in part by 18-APRA18-0084 and 20-RTF20-0003 and is supported by the U.S. Department of Energy, Office of Science, Office of Nuclear Physics, and Laboratory Directed Research and Development (LDRD) funding from Argonne National Laboratory, provided by the Director, Office of Science, of the U. S. Department of Energy under Contract No. DE-AC02-06CH11357.

ALS and DV acknowledge that research was sponsored by NASA through a contract with ORAU.KK and ZM acknowledge that this material is based upon work supported by NASA under award number 80GSFC21M0002.YS’s work was supported by JSPS KAKENHI, Japan, Grant Numbers JP23K13127.

References

  • [1]W.B.Atwood etal., The Large Area Telescope on the Fermi Gamma-Ray Space Telescope Mission, ApJ 697 (2009) 1071 [0902.1089].
  • [2]J.L.Bazo Alba and AMS-02 Tracker Collaboration, In-flight performance of the AMS-02 silicon tracker, in Journal of Physics Conference Series, vol.409 of Journal of Physics Conference Series, p.012032, Feb., 2013, DOI.
  • [3]DAMPE collaboration, The DArk Matter Particle Explorer mission, Astropart. Phys. 95 (2017) 6 [1706.08453].
  • [4]I.Peric, A novel monolithic pixelated particle detector implemented in high-voltage CMOS technology, Nucl. Instrum. Meth. A 582 (2007) 876.
  • [5]R.Caputo, M.Ajello, C.A.Kierans, J.S.Perkins, J.L.Racusin, L.Baldini etal., All-sky Medium Energy Gamma-ray Observatory eXplorer mission concept, Journal of Astronomical Telescopes, Instruments, and Systems 8 (2022) 044003.
  • [6]I.Brewer, M.Negro, N.Striebig, C.Kierans, R.Caputo, R.Leys etal., Developing the future of gamma-ray astrophysics with monolithic silicon pixels, Nucl. Instrum. Meth. A 1019 (2021) 165795 [2109.13409].
  • [7]I.Peric, M.Prathapan, H.Augustin, M.Benoit, R.C.Mohr, D.Dannheim etal., A high-voltage pixel sensor for the ATLAS upgrade, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment (2018) .
  • [8]A.L.Steinhebel, H.Fleischhack, N.Striebig, M.Jadhav, Y.Suda, R.Luz etal., AstroPix: novel monolithic active pixel silicon sensors for future gamma-ray telescopes, in Space Telescopes and Instrumentation 2022: Ultraviolet to Gamma Ray, J.-W.A.den Herder, S.Nikzad and K.Nakazawa, eds., vol.12181, p.121816Y, International Society for Optics and Photonics, SPIE, 2022, DOI.
  • [9]A.LSteinhebel, R.Caputo, H.Fleischhack, N.Striebig, M.Jadhav, Y.Suda etal., AstroPix: CMOS pixels in space, PoS Pixel2022 (2023) 020.
  • [10]Y.Suda, R.Caputo, A.L.Steinhebel, H.Fleischhack, N.Striebig, M.Jadhav etal., Development of an HV-CMOS active pixel sensor AstroPix for all-sky medium-energy gamma-ray telescopes, PoS ICRC2023 (2023) 644.
  • [11]Y.Suda, R.Caputo, A.L.Steinhebel, N.Striebig, M.Jadhav, Y.Fukazawa etal., Development of a novel HV-CMOS active pixel sensor AstroPix for gamma-ray space telescopes, in Space Telescopes and Instrumentation 2024: Ultraviolet to Gamma Ray, J.-W.A.den Herder, S.Nikzad and K.Nakazawa, eds., vol.13093, p.130937P, International Society for Optics and Photonics, SPIE, 2024, DOI.
  • [12]A.Schöning, J.Anders, H.Augustin, M.Benoit, N.Berger, S.Dittmeier etal., MuPix and ATLASPix – Architectures and Results, 2020.
  • [13]I.Perić and N.Berger, High Voltage Monolithic Active Pixel Sensors, Nucl. Phys. News 28 (2018) 25.
  • [14]N.Striebig, Development of integrated sensors for gamma ray astronomy, Master’s thesis, Karlsruhe Institute of Technology, 2021.
  • [15]L.Meng, A.Andreazza, D.Muenstermann, E.Hutchinson, F.Wilson, H.Fox etal., First Results of an ATLASPix3.1 Telescope, in Proceedings of the 31st International Workshop on Vertex Detectors (VERTEX2022), JPSCP (2023), DOI [https://journals.jps.jp/doi/pdf/10.7566/JPSCP.42.011023].
  • [16]Y.Suda, R.Caputo, A.L.Steinhebel, N.Striebig, M.Jadhav, Y.Fukazawa etal., Performance evaluation of the high-voltage CMOS active pixel sensor AstroPix for gamma-ray space telescopes, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 1068 (2024) 169762.
  • [17]O.Kroger, “An Investigation of Depletion in AstroPix, a High Voltage Monolithic CMOS Sensor.” https://escholarship.org/uc/item/5g24799m, 2024.
  • [18]N.Striebig, R.Leys, I.Peric, R.Caputo, A.Steinhebel, Y.Suda etal., AstroPix4 — a novel HV-CMOS sensor developed for space based experiments, Journal of Instrumentation 19 (2024) C04010.
  • [19]R.Schimassek, Development and Characterisation of Integrated Sensors for Particle Physics, Ph.D. thesis, Karlsruhe Institute of Technology (KIT), 2021.
  • [20]M.Manghisoni, D.Comotti, L.Gaioni, L.Ratti and V.Re, Dynamic compression of the signal in a charge sensitive amplifier: Experimental results, IEEE Transactions on Nuclear Science 65 (2018) 636.
  • [21]A.L.Steinhebel, J.Ott, O.Kroger, R.Caputo, V.Fadeyev, A.Affolder etal., The path toward 500 μ𝜇\muitalic_μm depletion of AstroPix, a pixelated silicon HVCMOS sensor for space and EIC, 2024.
  • [22]Kramberger, G. and Cindro, V. and Mandić, I. and Mikuž, M. and Milovanović, M. and Zavrtanik, M. and Žagar, K., Investigation of Irradiated Silicon Detectors by Edge-TCT, IEEE Transactions on Nuclear Science 57 (2010) 2294.
  • [23]T.Aoki, Dynamics of heavy-ion-induced latchup in cmos structures, IEEE Transactions on Electron Devices 35 (1988) 1885.
  • [24]M.McMahan, Radiation effects testing at the 88-inch cyclotron, in 1999 Fifth European Conference on Radiation and Its Effects on Components and Systems. RADECS 99 (Cat. No.99TH8471), pp.142–147, 1999, DOI.
  • [25]D.Heynderickx, B.Quaghebeur and H.D.R.Evans, The ESA Space Environment Information System (SPENVIS), in IAF abstracts, 34th COSPAR Scientific Assembly, p.475, 2002, DOI.
  • [26]V.U.S.ofEngineering, “Weibull – creme-mc site.”
  • [27]A.Tylka, J.Adams, P.Boberg, B.Brownstein, W.Dietrich, E.Flueckiger etal., Creme96: A revision of the cosmic ray effects on micro-electronics code, IEEE Transactions on Nuclear Science 44 (1997) 2150.
  • [28]A.L.Steinhebel, N.Striebig, M.Jadhav, D.Violette, D.Durachka, R.Boggs etal., A-STEP for AstroPix : Development and Test of a space-based payload using novel pixelated silicon for gamma-ray measurement, PoS ICRC2023 (2023) 579.
  • [29]R.Caputo, C.Kierans, N.Cannady, A.Falcone, Y.Fukazawa, M.Jadhav etal., ComPair-2: a next-generation medium-energy gamma-ray telescope prototype, in Space Telescopes and Instrumentation 2024: Ultraviolet to Gamma Ray, J.-W.A.den Herder, S.Nikzad and K.Nakazawa, eds., vol.13093, p.130932L, International Society for Optics and Photonics, SPIE, 2024, DOI.
  • [30]L.D.Smith, N.Cannady, R.Caputo, C.Kierans, N.Kirschner, I.Liceaga-Indart etal., The 2023 balloon flight of the ComPair instrument, in Space Telescopes and Instrumentation 2024: Ultraviolet to Gamma Ray, J.-W.A.den Herder, S.Nikzad and K.Nakazawa, eds., vol.13093, p.130937Z, International Society for Optics and Photonics, SPIE, 2024, DOI.
AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement (2025)
Latest Posts
Recommended Articles
Article information

Author: Virgilio Hermann JD

Last Updated:

Views: 6587

Rating: 4 / 5 (41 voted)

Reviews: 88% of readers found this page helpful

Author information

Name: Virgilio Hermann JD

Birthday: 1997-12-21

Address: 6946 Schoen Cove, Sipesshire, MO 55944

Phone: +3763365785260

Job: Accounting Engineer

Hobby: Web surfing, Rafting, Dowsing, Stand-up comedy, Ghost hunting, Swimming, Amateur radio

Introduction: My name is Virgilio Hermann JD, I am a fine, gifted, beautiful, encouraging, kind, talented, zealous person who loves writing and wants to share my knowledge and understanding with you.