Pith. sign in

REVIEW 2 major objections 6 minor 10 references

Evaluation of gamma-ray response of the AstroPix4 HV-CMOS active pixel sensor

T0 review · 2 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read AstroPix4 extends its gamma-ray sensitivity down to 14 keV and up to about 250 keV, with an energy resolution of 16.6 keV at 122 keV and a depletion depth near 90 micrometers.

desk verdict A solid, honest AstroPix4 characterization with a plausible 14–250 keV dynamic range, but the upper end rests on an unvalidated injection-linearity assumption that a referee should press on. read the letter →

arxiv 2507.21618 v1 pith:7ZHV4EAZ submitted 2025-07-29 astro-ph.IM

classification astro-ph.IM
keywords AstroPix4HV-CMOSactivepixelsensorgamma-raydetectorTime-over-ThresholdenergycalibrationdepletiondepthAMEGO-XMeVastronomy
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

AstroPix4 is the latest iteration of a high-voltage CMOS active pixel sensor being developed for the proposed AMEGO-X gamma-ray space telescope, and this paper sets out to measure how well it actually detects gamma rays. The central result is that the chip's reduced input capacitance lowers the noise floor enough to see the 14 keV photopeak of $^{57}$Co, while the charge-sensitive amplifier saturates near 250 keV, so the estimated dynamic range is roughly 14 to 250 keV. At 122 keV the energy resolution is 16.6 keV (about 14% FWHM), which does not yet meet AMEGO-X's $<10\%$ requirement, and the measured depletion depth is about 90 $\mu$m at $-240$\,V on a substrate whose resistivity cannot support full depletion. These numbers matter because they quantify how close AstroPix4 comes to the AMEGO-X requirements and directly motivate the planned AstroPix5 changes: a higher-range amplifier, fewer metal layers, and higher-resistivity wafers.

What carries the argument

The load-bearing object is the AstroPix4 pixel: each 500 $\mu$m-pitch pixel collects ionization electrons in a deep N-well, reads them out through an intra-pixel charge-sensitive amplifier and comparator, and digitizes the pulse duration as a Time-over-Threshold value. Two analysis techniques carry the results: an empirical calibration function $\mathrm{ToT} = aE + b(1-e^{-E/c}) + d$ fitted to six photopeaks, then extended above 122 keV by charge-injection data corrected under the assumption $Q = CV \propto E$; and a photopeak-counting method using $^{241}$Am 60 keV gamma rays to derive depletion depth, with corrections for readout dead time and for photoelectron path length. The central mechanism being tested is whether the sensor's depletion layer and signal chain behave as a simple PN junction with linear charge collection.

What would settle it

Measure a known calibration line between 122 and 250 keV such as the 244.7 keV line of $^{152}$Eu, reconstruct its energy with the injection-corrected calibration curve, and compare; a mismatch beyond the quoted 68% confidence interval would show that the linear-injection assumption breaks down and the dynamic-range ceiling is not reliable.

Watch

Extended reading notes

Core claim

The paper's claim is that AstroPix4 has a gamma-ray response spanning from 14 keV to roughly 250 keV, set at the low end by noise (now reduced thanks to lower input capacitance) and at the high end by saturation of the in-pixel charge-sensitive amplifier. Using a modified calibration curve that extends photopeak fits with charge-injection data, the authors reconstruct spectra from five radioisotopes and confirm the upper cutoff in $^{137}$Cs. They report an energy resolution of $16.6^{+4.0}_{-2.6}$ keV FWHM at 122 keV (about 14%), a 30% improvement at 60 keV over AstroPix3, and a depletion depth of about 90 $\mu$m at $-240$\,V following a PN-junction model with inferred substrate resistivity of about $299\pm5\ \Omega\cdot\text{cm}$.

Load-bearing premise

The assumption that injection voltage is linearly proportional to deposited energy ($Q = CV \propto E$) is load-bearing: it is used to correct injection data above 122 keV and to set the upper end of the stated dynamic range at roughly 250 keV, so if that linearity fails the high-energy calibration falls with it.

Editorial extensions

If this is right

  • The chip's 14 keV sensitivity beats the 25 keV lower bound AMEGO-X asks for, but the roughly 250 keV saturation cutoff is far short of the 700 keV requirement, so the next sensor needs a higher-saturation amplifier.
  • A 14% FWHM resolution at 122 keV misses the required 10%, and the paper attributes this to 122 keV photoelectrons not being fully absorbed in the roughly 90 $\mu$m depletion layer; deeper depletion should improve it.
  • Power consumption falls to about 2 mW/cm$^2$, half of AstroPix3, but still above the 1.5 mW/cm$^2$ AMEGO-X target.
  • The depletion depth grows with bias as a PN-junction model predicts, implying that fabricating AstroPix5 on higher-resistivity wafers should yield substantially larger depletion rather than requiring a new operating principle.
  • The injection-data calibration extension offers a way to map the sensor's upper response without waiting for rare high-energy photopeaks, and it is what locates the CSA saturation at about 250 keV.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the injection-voltage linearity survives direct testing, the same calibration technique could be reused for AstroPix5's high-dynamic-range test columns, letting the team characterize saturation behavior with electronic pulses instead of radioactive sources in the 300-700 keV gap.
  • The hypothesis that 122 keV resolution is limited by photoelectron escape from a 90 $\mu$m layer is directly testable by simulating photoelectron transport for that active depth and comparing the predicted 122 keV line width with the observed 16.6 keV.
  • A natural follow-up would be to check whether the 88 keV and 122 keV line widths scale with the inverse square root of collected charge; if they do not, the excess width at 122 keV has a source other than counting statistics, such as incomplete charge collection.
  • An independent anchor for the dynamic-range ceiling could come from a source like $^{152}$Eu, whose 244.7 keV gamma line lies just below the suspected saturation point, providing a real photopeak to test the injection-corrected calibration curve.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The manuscript reports a laboratory characterization of AstroPix4, a 16×13-pixel HV-CMOS active pixel sensor developed for the proposed AMEGO-X mission. Using Time-over-Threshold (ToT) spectra from five radioisotopes, the authors identify six photopeaks, fit an empirical calibration curve, and use charge-injection data to extend the calibration beyond 122 keV under an explicit linearity assumption. They report a gamma-ray dynamic range from 14 keV to about 250 keV, a FWHM energy resolution of 16.6 keV (about 14%) at 122 keV, a depletion depth of about 90 micrometers at -240 V, and a power consumption of about 2 mW/cm^2. The depletion-depth data are compared with a simple PN-junction model.

Significance. If the reported performance figures hold, this is a useful milestone for the AstroPix development program: it demonstrates detection of the 14 keV photopeak in a 500-micrometer-pitch HV-CMOS pixel, quantifies the improvement in 60 keV resolution relative to AstroPix3, and provides a depletion-depth-versus-bias dataset checked against a physical model. The six-photopeak calibration is anchored to external isotope lines, which is a clear strength, and the paper openly states the assumption used for the injection-based extension. The main weakness is that the upper end of the dynamic range, a central claim of the paper, depends on an unvalidated linearity assumption for the charge injector and on a saturation-edge interpretation that is not independently confirmed. These issues need to be addressed before the dynamic range claim is fully supported.

major comments (2)
  1. [Section 3, Figs. 3 and 4] The claim that the dynamic range extends to about 250 keV is load-bearing and currently rests on two unvalidated steps. First, the injection data above 122 keV are corrected under the assumption that Q = CV is proportional to deposited energy (stated in Section 3 and Fig. 3), but no high-amplitude validation of the injector's charge-versus-voltage response is shown. Second, the sharp cutoff near 140 microseconds in the 137Cs spectrum is identified as CSA saturation and mapped to about 250 keV using the extended calibration curve, but the paper does not rule out other origins for the cutoff, such as comparator saturation or ToT digitization/clamping effects. The 133Ba Compton edges at 164 and 207 keV are only used qualitatively, so there is no validated calibration point between 122 keV and the reported saturation boundary. Please provide a direct calibration point above 122 keV (for example, a monoenergetic source or beam measurement) or an independent characterization of the injector linearity at the relevant amplitudes, and propagate the resulting systematic uncertainty into the reported 250 keV upper limit.
  2. [Section 3, Fig. 4 and Section 5] The interpretation of the 137Cs cutoff as CSA saturation would be more convincing if supported by a direct measurement of the CSA response or by a controlled scan of injection amplitude and comparator threshold in the saturation region. As written, the paper asserts the saturation interpretation after the fact, and the alternative that the cutoff arises from the readout chain rather than the CSA is not quantitatively excluded. Since the upper dynamic range determines the comparison with the AMEGO-X requirement of up to 700 keV, this ambiguity is consequential for the paper's central characterization.
minor comments (6)
  1. [Abstract and Section 2] The power consumption figure of about 2 mW/cm^2 is stated without measurement details, operating conditions, or uncertainty. Please specify how this value was obtained and define the active area used in the normalization.
  2. [Fig. 3 caption and Section 3] The injection-voltage axis label 'mV / 0.08' is not self-explanatory. Please define the conversion factor between injection voltage and injected charge, and state whether it is a design value or a measured value.
  3. [Section 3] The dynamic range lower bound is presented as 14 keV based on the visibility of the 57Co photopeak, but the paper does not report the distribution of noise thresholds or the fraction of pixels that can detect 14 keV. Clarifying the pixel-to-pixel spread would make the lower-end claim more robust.
  4. [Fig. 7 caption] The caption 'Gauss ± Gauss' is unclear; it likely means the Gaussian fit mean plus/minus the fitted sigma. Please rewrite for clarity.
  5. [References] Reference [8] is incomplete: 'J. of Instrum., 19 (2024)' lacks an article number. Please complete the bibliographic information.
  6. [Abstract] There is a LaTeX artifact in the abstract: 'mW/{cm}^2' should be 'mW/cm^2'.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: energy calibration is anchored to external isotope photopeaks; the injection-based high-energy extension is an explicitly stated assumption rather than a renamed input.

full rationale

The energy calibration is anchored to six measured photopeak energies from external radioisotopes (14, 22, 31, 60, 88, and 122 keV), so the low-energy response is not derived from the quantities it predicts. The reported 122 keV FWHM and 60 keV resolution come from Gaussian fits to those photopeaks. Above 122 keV, the paper uses injection data to extend the calibration curve, but it explicitly states the assumption 'linearity between injection voltage and energy (Q = CV ∝ E)' and uses that assumption to assign energies to the injection points; this is a transparent modeling assumption rather than a hidden re-use of the target result. The 137Cs cutoff near 140 microseconds is an independent observable, and the interpretation that it corresponds to CSA saturation at about 250 keV is conditional on the stated calibration extension. The depletion-depth measurement follows the procedure of the authors' prior work [6], but this is a methodological self-citation rather than a load-bearing uniqueness or equivalence argument, and the result is cross-checked against a simple PN junction model using the independently specified (200-400) Ohm-cm wafer resistivity. No step in the derivation reduces by construction to its own input; the main caveat is the unvalidated high-energy injection linearity, which is a correctness risk rather than circularity.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central characterization claims rest on calibrated photopeaks (external standards), a standard depletion model, and one ad hoc linearity assumption for the injection-based calibration extension. No new entities are postulated.

free parameters (4)
  • Calibration curve parameters (a, b, c, d) = medians a=0.12, b=207, c=28, d=-64
    Empirical ToT(E)=aE+b[1-exp(-E/c)]+d fitted to six photopeaks per pixel (Section 3, Fig. 3).
  • Substrate resistivity (rho) = 299 +/- 5 Ohm-cm
    Fitted from depletion depth vs bias using the PN junction model (Section 4, Fig. 7).
  • Dead-time correction factor = 0.91
    Applied to the observation time when deriving photopeak event rates for depletion depth (Section 4), stated without a derivation in this paper.
  • Effective volume correction factor = 0.83 to 0.87 depending on bias
    Geant4-based correction for the 60 keV photoelectron path length in the depleted volume (Section 4).
assumptions (3)
  • domain assumption PN junction depletion model d = sqrt(2 epsilon mu rho (V_bias + V_built_in))
    Used to interpret depletion depth data and fit the substrate resistivity (Section 4, Fig. 7).
  • ad hoc to paper Linearity between injection voltage and deposited energy (Q = CV proportional to E)
    Basis for extending the calibration curve above 122 keV using injection data (Section 3, Fig. 3).
  • domain assumption Geant4 simulations accurately model photoelectron path lengths and effective detection volume
    Used to compute the effective volume correction factor for depletion depth measurements (Section 4).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Evaluation of gamma-ray response of the AstroPix4 HV-CMOS active pixel sensor." pith.science (2026). https://pith.science/paper/7ZHV4EAZ

@misc{pith2026250721618,
  author       = {Pith},
  title        = {Pith review of: Evaluation of gamma-ray response of the AstroPix4 HV-CMOS active pixel sensor},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7ZHV4EAZ}},
  note         = {Machine review of arXiv:2507.21618}
}
abstract

AstroPix is a novel high-voltage CMOS active pixel sensor being developed for a next generation gamma-ray space telescope, AMEGO-X. To meet AMEGO-X instrument requirements, AstroPix must achieve full depletion of its $500~\rm{\mu m}$ thick, $500~\rm{\mu m}$-pitch pixels. It must be sensitive to gamma rays in the range of $25-700$ keV, with the energy resolution at 122 keV of $<10$%. Furthermore, given the space-based nature of AMEGO-X, the power consumption of AstroPix needs to be lower than $1.5~\rm{mW/{cm}^2}$. We report the gamma-ray response of the latest version of AstroPix, AstroPix4. The chip contains $16\times 13$ array of $500~\rm{\mu m}$-pitch pixels. The power consumption is estimated to be about $2~\rm{mW/{cm}^2}$, which is approximately half the power of the previous AstroPix version. The input capacitance is reduced, allowing for the detection of the 14 keV photopeak from $\rm{^{57}Co}$ and a moderate energy resolution of 14% at 122 keV. The dynamic range is estimated to be in the range from 14 keV to $\sim250$ keV. We found that the sensor depletion layer expands as expected and the measured depletion depth is approximately $90~\rm{\mu m}$ when biased at $-240$ V.

Figures

Figures reproduced from arXiv: 2507.21618 by the authors.

Figure 1
Figure 1. AstroPix4 chip and its carrier board. Designed by Karlsruhe Institute [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 5
Figure 5. Energy calibration curves for all calibrated pixels. Green points and [PITH_FULL_IMAGE:figures/full_fig_p003_5.png] view at source ↗
Figure 6
Figure 6. Energy dependence of energy resolutions (FWHM). The inset dis [PITH_FULL_IMAGE:figures/full_fig_p003_6.png] view at source ↗
Figures from the paper (1 more)
Figure 7
Figure 7. Figure 7: Depletion depths as a function of bias voltage. The lines in blue show [PITH_FULL_IMAGE:figures/full_fig_p004_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 9 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    Caputo et al., ``All-sky Medium Energy Gamma-ray Observatory eXplorer mission concept,'' J

    R. Caputo et al., ``All-sky Medium Energy Gamma-ray Observatory eXplorer mission concept,'' J. of Astron. Telesc., Instrum., and Syst., 8, 4 (2022) 044003

  3. [3]

    Brewer et al., ``Developing the future of gamma-ray astrophysics with monolithic silicon pixels,'' Nucl

    I. Brewer et al., ``Developing the future of gamma-ray astrophysics with monolithic silicon pixels,'' Nucl. Instrum. Methods A 1019 (2021) 165795

  4. [4]

    Peri\' c , ``A novel monolithic pixelated particle detector implemented in high-voltage CMOS technology,'' Nucl

    I. Peri\' c , ``A novel monolithic pixelated particle detector implemented in high-voltage CMOS technology,'' Nucl. Instrum. Methods A 582 (2007) 876-885

  5. [5]

    Peri\' c and N

    I. Peri\' c and N. Berger, ``High Voltage Monolithic Active Pixel Sensors,'' Nucl. Phys. News 28(1) (2018) 25–27

  6. [6]

    A. L. Steinhebel et al., ``AstroPix: novel monolithic active pixel silicon sensors for future gamma-ray telescopes,'' Proc. SPIE 12181 (2022)

  7. [7]

    Suda et al., ``Performance evaluation of the high-voltage CMOS active pixel sensor AstroPix for gamma-ray space telescopes,'' Nucl

    Y. Suda et al., ``Performance evaluation of the high-voltage CMOS active pixel sensor AstroPix for gamma-ray space telescopes,'' Nucl. Instrum. Methods A 1068 (2024) 169762

  8. [8]

    A. L. Steinhebel et al., ``AstroPix: A Pixelated HVCMOS Sensor for Space-Based Gamma-Ray Measurement,'' Nucl. Instrum. Methods A (submitted), arXiv:2501.11698v1 (2025)

Show all 10 references
  1. [9]

    Striebig et al., ``AstroPix4 - a novel HV-CMOS sensor developed for space based experiments,'' J

    N. Striebig et al., ``AstroPix4 - a novel HV-CMOS sensor developed for space based experiments,'' J. of Instrum., 19 (2024)

  2. [10]

    Zhang et al., ``Breakdown performance of guard ring designs for pixel detectors in 150 nm CMOS technology,'' Nucl

    S. Zhang et al., ``Breakdown performance of guard ring designs for pixel detectors in 150 nm CMOS technology,'' Nucl. Instrum. Methods A 1063 (2024) 169287

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.