Pith. sign in

REVIEW 3 major objections 5 minor 12 references

Enhancing Radiation Hardness and Granularity in HV-CMOS: The RD50-MPW4 Sensor

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Pith's one-line read Raising the bias voltage restores an irradiated HV-CMOS pixel sensor to near-full detection efficiency.

desk verdict Useful incremental sensor data; the recovery-with-bias result is likely real, but the paper leaves threshold settings off the efficiency curves, which weakens the headline claim. read the letter →

arxiv 2504.15730 v1 pith:DUBFZ7BS submitted 2025-04-22 physics.ins-det hep-ex

classification physics.ins-dethep-ex
keywords RD50-MPWHV-CMOSDMAPSdepletedmonolithicactivepixelsensorradiationdamageneutronirradiationbiasvoltage
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

This paper reports that the RD50-MPW4, a monolithic high-voltage CMOS pixel sensor, remains a functioning particle tracker after neutron irradiation up to $1\times10^{15}\,\mathrm{neq\,cm^{-2}}$, provided the bias voltage is raised from the standard 190 V to about 580 V and the chip is cooled. In that regime the hit detection efficiency recovers to roughly 99%, close to the >99.9% seen before irradiation. The authors take this as evidence that HV-CMOS is a viable path to radiation-hard tracking detectors for future collider experiments. Unirradiated samples also show a spatial resolution near $16\,\mu\mathrm{m}$ and a timing resolution of 9.8 ns, with more than 96% of hits inside a 25 ns window.

What carries the argument

The load-bearing mechanism is electric-field-driven charge collection in a high-voltage diode: biasing the sensor to several hundred volts enlarges the depleted volume and shortens charge drift, counteracting the charge trapping and elevated leakage current introduced by neutron damage. The device itself is a $64\times64$ pixel monolithic HV-CMOS DMAPS (depleted monolithic active pixel sensor) with $62\,\mu\mathrm{m}$ pitch, produced in a 150 nm CMOS process, with separated power domains, an improved guard ring, and an optional thinned, backside-metalized bias contact. The efficiency-recovery measurement is made in test beams using a six-plane reference telescope, a timing layer, and standard track-reconstruction software.

What would settle it

Record the hit detection efficiency of a $1\times10^{15}\,\mathrm{neq\,cm^{-2}}$ RD50-MPW4 sample as a function of bias with the threshold fixed at the pre-irradiation value (about 200 mV, roughly 5000 e−). If the 580 V efficiency stays near 99%, the recovery claim holds; if it drops sharply, the recovery in the paper is largely an artifact of threshold adjustment.

Watch

Extended reading notes

Core claim

The central discovery is that bias voltage can be used as a post-irradiation tuning knob: at a common operating point (190 V, threshold near 5000 e−) efficiency falls from 99.8% unirradiated to 99.5% at $1\times10^{14}$, 85.5% at $3\times10^{14}$, and 9.2% at $1\times10^{15}\,\mathrm{neq\,cm^{-2}}$, but increasing the bias restores the most heavily irradiated sample to about 98.9% at roughly 580 V. The same bias sweep shows cluster size and time-over-threshold growing with voltage, consistent with a larger depleted volume and more complete charge collection, while for fluences above $3\times10^{14}$ full depletion is no longer reached. The paper concludes that the HV-CMOS approach, with its large collection electrode and high-voltage operation, enables radiation-hard detectors.

Load-bearing premise

The efficiency-recovery curves in Figure 6 do not state the comparator threshold used for each sample, so part of the apparent recovery could be caused by threshold choice rather than by higher bias alone.

Editorial extensions

If this is right

  • Inner tracking layers at future hadron colliders could be built from HV-CMOS chips whose operating bias is raised over the detector lifetime to compensate accumulating radiation damage.
  • Cooling to about $-20\,^{\circ}\mathrm{C}$ becomes a requirement rather than an option: it lowers leakage current enough to allow the high bias voltages needed after irradiation.
  • Spatial resolution degrades after irradiation because cluster sizes shrink, but remains better than the binary pixel resolution of about $17.9\,\mu\mathrm{m}$.
  • Backside biasing offers no significant efficiency advantage over topside biasing at high fluence, so system integration can choose whichever scheme is mechanically simpler.
  • Timing resolution near 10 ns means the same sensor technology could provide both position and time information in a 25 ns bunch-crossing clock.

Reading between the lines

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

  • If the recovery is truly electrical, a fixed-threshold bias scan at each fluence should reproduce the same efficiency curves; running that scan would settle whether any part of the reported recovery is an artifact of per-sample threshold settings.
  • The same bias-recovery logic could apply to other large-collection-electrode CMOS sensors, suggesting a generic radiation-hardness strategy: design for high voltage headroom and use bias as a lifetime extension knob.
  • At fluences above $3\times10^{14}\,\mathrm{neq\,cm^{-2}}$ the absence of ToT and cluster-size saturation implies the sensor is no longer fully depleted, so the observable efficiency recovery may come from a partially depleted but still functional volume; simulation of the field profile would clarify which.
  • Tracking performance at future colliders could be limited by the need to cool the sensor to $-20\,^{\circ}\mathrm{C}$, so power density and cooling, not just efficiency, may set the practical fluence limit.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper characterizes the RD50-MPW4, a 64x64-pixel HV-CMOS DMAPS sensor fabricated in a 150 nm LFoundry process with 62 um pitch, separated power domains, optional backside metallization, and a design target of up to 800 V bias. It reports IV characteristics as a function of temperature for neutron-irradiated samples up to 1e16 neq/cm2, and test-beam results using the DESY telescope and the Corryvreckan analysis framework. For unirradiated sensors, the timing resolution is about 9.8 ns and the in-time hit-detection efficiency exceeds 96% within a 25 ns window. At a common bias of 190 V and a threshold of 200 mV above baseline (about 5000 e-), the hit-detection efficiency degrades from 99.8% unirradiated to 9.2% at 1e15 neq/cm2. The central claim is that raising the bias voltage restores efficiency to near pre-irradiation levels, e.g., 98.9% at about 580 V for the 1e15 sample, confirming that HV-CMOS is a viable radiation-hard tracking technology.

Significance. Demonstrating that a monolithic HV-CMOS DMAPS can remain efficient at 1e15 neq/cm2 when biased to several hundred volts and cooled is a practically important result for future tracking detectors at HL-LHC and FCC. The paper has real strengths: it uses an external telescope reference, covers multiple fluence levels, includes a temperature-controlled IV study, and explicitly compares top and backside biasing. The efficiency-recovery curves are the key evidence, and the central conclusion would be significant if the measurement conditions were fully documented. At present, however, the reported efficiency values lack matching threshold documentation and uncertainty estimates, so the strength of the claim is somewhat ahead of the supporting data.

major comments (3)
  1. [Section 3.3, Figure 6] The efficiency-versus-bias curves in Figure 6 do not state the comparator threshold settings used for each sample, while the caption of Figure 7 explicitly notes that 'different threshold settings were applied for the various fluence levels.' Since the hit-detection efficiency is a steep function of threshold, especially for irradiated sensors with reduced collected charge, the recovery of the 1e15 neq/cm2 sample from 9.2% at 190 V to 98.9% at about 580 V could be partly due to a lower threshold rather than to bias-induced restoration of charge collection. Within a single bias scan the threshold may have been fixed, but the cross-sample claim that efficiency is restored 'to levels comparable to those before irradiation' requires matched thresholds or an explicit threshold scan. Please report the threshold for every point in Figure 6 and, if possible, add a common-threshold bias scan or an efficiency-versus-threshold measurement at fixed bias.
  2. [Sections 3.2 and 3.3, Table 2 and Figure 6] The dataset contains only one sample per fluence (apart from the second 1e15 wafer listed in Table 1, which is not used as a reproducibility check), and the quoted efficiency values have no uncertainties or event counts. The summary statement that recovery reaches 'levels comparable to those before irradiation' therefore rests on point estimates; for example, 98.9% at 1e15 differs from the >99.9% unirradiated value, and without uncertainties one cannot tell whether this is a statistically significant residual loss. Please state the number of tracks, evaluate binomial and systematic uncertainties on each efficiency, and, where possible, include the W8 1e15 sample to demonstrate sample-to-sample consistency.
  3. [Sections 2 and 3.3] The IV characteristics show a very strong temperature dependence of the leakage current, yet the text only says that test-beam cooling reached 'approximately -15 C' without stating the actual sensor temperature during the efficiency scans of Figure 6 or the common-condition runs of Table 2. Because the maximum reachable bias before thermal runaway depends on temperature, the comparison between samples and the recovery claim require the operating temperature to be reported for each data set. Please provide the measured temperature for each efficiency scan and state whether it was constant across the bias ramp.
minor comments (5)
  1. [Table 1 and Section 2] Table 1 and the text disagree about the origin of the 1e15 neq/cm2 sample: the introduction says this sample is from W3 without backside processing, while Table 1 also lists a 1e15 sample from W8 and Section 2 says 'All samples are taken from W3.' Please clarify the wafer and processing status of each sample.
  2. [Section 3.1] The timing resolution is defined as the standard deviation of the innermost 99% of the main peak; please justify this truncation, for example by explaining the overflow-counter tail, and quote the number of tracks used or provide a Gaussian fit to the core.
  3. [Section 3.2] The text preceding Figure 4 contains the incomplete sentence 'The comparison of the (a) Comparison of the normalized cluster size.'; please rephrase this passage.
  4. [Section 3.2] The threshold setting 'VThr = 200 mV above baseline' should define what is meant by baseline, since this quantity is essential for interpreting the thresholds quoted throughout the paper.
  5. [Section 4] The phrase 'confirming that the HV-CMOS approach enables the development of radiation-hard detectors' is stronger than the single-sample, threshold-documented evidence supports; suggest replacing 'confirming' with 'supporting' or 'indicating'.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: efficiency recovery is a direct measurement against an external telescope and a non-irradiated baseline; the threshold caveat is an experimental limitation, not a circular derivation.

full rationale

The paper contains no fitted model, no derived prediction, and no reduction of a purported result to its own inputs. Hit detection efficiency and spatial/timing resolution are measured directly with the Corryvreckan framework using an external Adenium telescope and TelePix2 timing layer, and the post-irradiation recovery is compared with a non-irradiated sample at common bias and threshold settings (Table 2, Figure 6). The central claim that raising bias restores detection capability follows from measured efficiency-versus-bias curves, not from any parameter fitted to those curves. The Figure 7 caption's note that 'different threshold settings were applied for the various fluence levels' is a real experimental comparability limitation, but it is not circularity: it concerns whether cross-sample comparisons are apples-to-apples, not whether the reported efficiencies were manufactured from the assumptions. Self-citations [1], [2], and [4] describe prior sensor/test-structure characterization and are not load-bearing for the radiation-hardness conclusion, which rests on the new test-beam data presented here. No uniqueness theorem, ansatz-smuggling citation, or renaming of a known result is invoked. The conclusion is self-contained against external measurements and a non-irradiated baseline, so the appropriate circularity score is 0.

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

No free parameters or invented entities appear; this is a measurement paper. The listed axioms are standard assumptions about dosimetry, track reconstruction, and signal interpretation that the experimental conclusions rely on.

assumptions (4)
  • domain assumption Neutron fluences in Table 1 are accurate and spatially uniform across each sensor.
    Fluence is the independent variable for every radiation-damage comparison in Sections 2 and 3; no dosimetry uncertainty is reported.
  • domain assumption Telescope residual distributions and the geometric mean method give unbiased spatial resolution.
    Spatial resolution in Figure 8 is extracted as sqrt(sigma_biased * sigma_unbiased), following reference [12], which assumes no bias from including the DUT in track fitting and no misalignment.
  • domain assumption Time-over-threshold is proportional to deposited charge and cluster size reflects charge sharing.
    The interpretation of Figures 4 and 7 as charge-collection and depletion behavior relies on this proportionality.
  • domain assumption At high fluence, bulk damage dominates leakage current over surface effects.
    Section 2 uses this model (references [5,6]) to explain why top and backside biasing differ only at 1e14 neq/cm2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Radiation Hardness and Granularity in HV-CMOS: The RD50-MPW4 Sensor." pith.science (2026). https://pith.science/paper/DUBFZ7BS

@misc{pith2026250415730,
  author       = {Pith},
  title        = {Pith review of: Enhancing Radiation Hardness and Granularity in HV-CMOS: The RD50-MPW4 Sensor},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DUBFZ7BS}},
  note         = {Machine review of arXiv:2504.15730}
}
read the original abstract

The latest HV-CMOS pixel sensor developed by the former CERN-RD50-CMOS group, known as the \mpw, demonstrates competitive radiation tolerance, spatial granularity, and timing resolution -- key requirements for future high-energy physics experiments such as the HL-LHC and FCC. Fabricated using a \SI{150}{nm} CMOS process by \emph{LFoundry}, it introduces several improvements over its predecessor, the \emph{RD50-MPW3}, including separated power domains for reduced noise, a new backside biasing scheme, and an enhanced guard ring structure, enabling operation at bias voltages up to \SI{800}{V}. Tests with non-irradiated samples achieved hit detection efficiencies exceeding \SI{99.9}{\%} and a spatial resolution around \SI{16}{\mu m}. Neutron-irradiated sensors were characterized using IV measurements and test-beam campaigns, confirming the sensor's robustness in high-radiation environments. The results highlight the ability of HV-CMOS technology to restore hit detection efficiency post-irradiation by increasing the applied bias voltage. Details of these measurements and timing performance are presented in this paper.

Figures

Figures reproduced from arXiv: 2504.15730 by the authors.

Figure 1
Figure 1. IV characteristics of the various irradiated samples at di [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 4
Figure 4. Direct comparison of the various fluence levels recorded with shared [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗
Figure 3
Figure 3. In time efficiency for various timing cuts of a non-irradiated sample. The cut range from [60 ns, 6.3 µs] is not shown as no major in-time efficiency increase is observed in this region. ≈ 3 %, corresponds to several hits which got a wrong over￾flow counter value assigned to and are thereby displaced by 256 × 25 ns = 6.4 µs. The DAQ system assigns the overflow counter to the various hits off-chip and can cause a jum… view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: In-pixel efficiency maps of differently irradiated samples at common bias and threshold settings. The hit detection efficiency starts to drop at an irradiation level of 3×1014 neqcm−2 to 85.5 % and at 1×1015 neqcm−2 even to 9.2 %. The reason for these losses can be und…
Figure 8
Figure 8. Figure 8: Spatial resolution as a function of the bias voltage of the various [PITH_FULL_IMAGE:figures/full_fig_p004_8.png]
Figure 7
Figure 7. Figure 7: Charge-related characteristics of the various samples as a function of [PITH_FULL_IMAGE:figures/full_fig_p004_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

12 extracted references · 6 canonical work pages

  1. [1]

    Sieberer, et al., RD50-MPW3: a fully monolithic digital cmos sensor for future tracking detectors, JINST 18 (2023) C02061

    P. Sieberer, et al., RD50-MPW3: a fully monolithic digital cmos sensor for future tracking detectors, JINST 18 (2023) C02061. doi: 10.1088/ 1748-0221/18/02/C02061

  2. [2]

    Pilsl, et al., Characterization of the RD50-MPW4 HV-CMOS pixel sensor, NIM - A 1069 (2024) 169839

    B. Pilsl, et al., Characterization of the RD50-MPW4 HV-CMOS pixel sensor, NIM - A 1069 (2024) 169839. doi: https://doi.org/10. 1016/j.nima.2024.169839

  3. [3]

    Moll, Radiation damage in silicon particle detectors: Microscopic de- fects and macroscopic properties, Ph.D

    M. Moll, Radiation damage in silicon particle detectors: Microscopic de- fects and macroscopic properties, Ph.D. thesis, Hamburg U., 1999

  4. [4]

    Vilella, et

    E. Vilella, et. al, RD50-MPW4: a thin backside-biased high voltage cmos pixel chip for high radiation tolerance 20 (2025) C03044. URL: https: //dx.doi.org/10.1088/1748-0221/20/03/C03044. doi:10.1088/ 1748-0221/20/03/C03044

  5. [5]

    Lindstroem, et al., Radiation hard silicon detectors—developments by the RD48 (ROSE) collaboration, NIM - A 466 (2001) 308–326

    G. Lindstroem, et al., Radiation hard silicon detectors—developments by the RD48 (ROSE) collaboration, NIM - A 466 (2001) 308–326. doi: 10. 1016/S0168-9002(01)00560-5

  6. [6]

    H. L. Hughes, Surface e ffects of space radiation on silicon devices, IEEE Transactions on Nuclear Science 12 (1965) 53–63. doi: 10.1109/TNS. 1965.4323924

  7. [7]

    Diener, et al., The DESY II test beam facility, NIM - A 922 (2019) 265–286

    R. Diener, et al., The DESY II test beam facility, NIM - A 922 (2019) 265–286. doi:https://doi.org/10.1016/j.nima.2018.11.133

  8. [8]

    Yi, et al., ADENIUM — a demonstrator for a next-generation beam telescope at desy, JINST 18 (2023) P06025

    L. Yi, et al., ADENIUM — a demonstrator for a next-generation beam telescope at desy, JINST 18 (2023) P06025. doi: 10.1088/1748-0221/ 18/06/p06025

Show all 12 references
  1. [9]

    Huth, et al., TelePix2: Full scale fast region of interest trigger and timing for the eudet-style telescopes at the desy ii test beam facility, 2025

    L. Huth, et al., TelePix2: Full scale fast region of interest trigger and timing for the eudet-style telescopes at the desy ii test beam facility, 2025. URL: https://arxiv.org/abs/2503.08177. arXiv:2503.08177

  2. [10]

    Baesso, et al., The AIDA-2020 TLU: a flexible trigger logic unit for test beam facilities, Journal of Instrumentation 14 (2019) P09019–P09019

    P. Baesso, et al., The AIDA-2020 TLU: a flexible trigger logic unit for test beam facilities, Journal of Instrumentation 14 (2019) P09019–P09019. doi:10.1088/1748-0221/14/09/p09019

  3. [11]

    Dannheim, et al., Corryvreckan: a modular 4d track reconstruction and analysis software for test beam data, JINST 16 P03008 (2021)

    D. Dannheim, et al., Corryvreckan: a modular 4d track reconstruction and analysis software for test beam data, JINST 16 P03008 (2021). doi: 10. 1088/1748-0221/16/03/p03008

  4. [12]

    Alexopoulos, et al., Examining the geometric mean method for the extraction of spatial resolution, JINST 9 (2014) P01003–P01003

    T. Alexopoulos, et al., Examining the geometric mean method for the extraction of spatial resolution, JINST 9 (2014) P01003–P01003. doi:10. 1088/1748-0221/9/01/p01003. 5

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.