In beam performances of the MIMOSIS-2.1 CMOS Monolithic Active Pixel Sensor

Abdelkader Himmi; Ali-Murteza Altingun; Andrei Dorokhov; Auguste Besson; Benedict Arnoldi-Meadows; Benedikt Gutsche; Christian Joachim Schmidt; Christian M\"untz; Christine Hu-Guo; Christopher Braun

REVIEW 3 major objections 4 minor 1 cited by

In beam performances of the MIMOSIS-2.1 CMOS Monolithic Active Pixel Sensor

T0 review · 3 major / 4 minor · reviewed 2026-08-08 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read MIMOSIS-2.1, the full-feature prototype for the CBM Micro Vertex Detector, detects charged pions with an efficiency above 99.9% in all pixel variants and reaches the targeted 5 µm spatial precision when thresholds are tuned.

desk verdict A solid, honest beam-test report for the CBM MVD sensor program; the headline efficiency and resolution numbers need statistical uncertainties before the requirements claim is fully established. read the letter →

arxiv 2502.05303 v1 pith:SRJQOE2X submitted 2025-02-07 physics.ins-det

M. Deveaux , Ali-Murteza Altingun , Julio Andary , Benedict Arnoldi-Meadows , Jerome Baudot , Gregory Bertolone , Auguste Besson , Norbert Bialas

show 30 more authors

Christopher Braun Roma Bugiel Gilles Claus Claude Colledani Hasan Darwish Andrei Dorokhov Guy Dozière Ziad El Bitar Ingo Fröhlich Mathieu Goffe Benedikt Gutsche Abdelkader Himmi Christine Hu-Guo Kimmo Jaaskelainen Oliver Keller Michal Koziel Franz Matejcek Jan Michel Frederic Morel Christian Müntz Hung Pham Christian Joachim Schmidt Stefan Schreiber Matthieu Specht Joachim Stroth Eva-dhidho Taka Isabelle Valin Roland Weirich Yüe Zhao Marc Winter

This is my paper · ORCID

classification physics.ins-det

keywords CMOSMonolithicActivePixelSensorMIMOSISbeamtestdetectionefficiencyspatialresolutionclustermultiplicityepitaxiallayerCBMMicroVertexDetector

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

Micro Vertex Detector of the CBM experiment, MIMOSIS-2.1 is the first full-scale, full-feature prototype of the MIMOSIS CMOS Monolithic Active Pixel Sensor. The paper reports beam-test results showing that every pixel variant detects charged pions with an efficiency above 99.9%, and that with suitably chosen thresholds the spatial precision reaches the design target of 5 µm. It also finds that a newly introduced 50 µm epitaxial layer improves the non-irradiated sensor, mostly by increasing cluster multiplicity. A separate calibration result is that the on-chip charge-injection capacitance is about twice its design value, so earlier threshold and noise numbers expressed in electrons need to be rescaled by this factor. If these results hold, the sensor meets the efficiency and precision requirements of the CBM-MVD, with radiation tolerance still to be demonstrated.

What carries the argument

The central object is MIMOSIS-2.1, a CMOS Monolithic Active Pixel Sensor built in a 180 nm imaging process, with 30 µm × 27 µm pixels that each integrate an amplifier, shaper, discriminator, and in-pixel memory. Hits are grouped on-chip into clusters and read out through priority encoders. Two sensing-node options (standard and p-stop) are each combined with 25 µm and 50 µm epitaxial layers; the 50 µm layer is the new feature. The beam-test analysis uses a telescope of MIMOSIS-2.1 reference planes and an iterative telescope-optimizer procedure that folds residuals back into the estimated track precision until convergence. The key observable is the cluster multiplicity, which the paper shows controls spatial precision: precision improves up to a cluster multiplicity near 2.4, then degrades, then improves again near 3.8 and 6, a pattern the authors interpret as a geometrical effect of charge sharing.

What would settle it

Measure the spatial precision of a 50 µm epitaxial MIMOSIS-2.1 sensor using a reference telescope whose track precision is known independently (for example, a silicon strip telescope with well-calibrated residuals), and check whether the extracted DUT precision still comes out at 5 µm. If it is systematically worse, the iterative telescope-precision estimate has biased the result.

Watch

Extended reading notes

Core claim

The central claim is that MIMOSIS-2.1's beam performance already satisfies the CBM-MVD requirements for detection efficiency and spatial resolution. Using a 120 GeV/c pion beam at the CERN-SPS, all pixel flavours (standard and p-stop, each with 25 µm and 50 µm epitaxial layers) show detection efficiency above 99.9%, with the best efficiency seen for the 50 µm p-stop pixels. The spatial precision reaches about 5 µm for suitable thresholds; the precision depends on cluster multiplicity, with clear optima around cluster multiplicities of 2.4, 3.8, and 6, a pattern the paper attributes to a geometry-driven mechanism rather than pixel type. The paper also establishes that the charge-injection gain used for absolute calibration is 1.86 e/mV for DC pixels and 2.02 e/mV for AC pixels, roughly twice the design value, meaning previously published electron-scale thresholds and noise should be re-evaluated with this factor.

Load-bearing premise

The load-bearing premise is that the telescope's track precision is correctly extracted by iterating residuals on identical sensors, and that the injection-gain calibration measured on 25 µm epitaxial test structures applies unchanged to the 50 µm epitaxial pixels.

Editorial extensions

If this is right

If the beam-test results are representative, MIMOSIS-2.1 satisfies the CBM-MVD requirement of better than 99.9% detection efficiency with all pixel variants tested.
With thresholds set near the multiplicity optima, the sensor reaches the 5 µm spatial precision target on the shorter pixel side, and about 10% worse on the other side.
The 50 µm epitaxial layer is a useful improvement for non-irradiated operation; whether it survives the targeted radiation dose of $10^{14}~n_{eq}/cm^2$ remains an open question the paper does not claim to answer.
The factor-of-two correction to the injection gain means that previously published electron-scale thresholds and noise values should be rescaled, changing numbers but not the qualitative conclusions of those studies.
The observed cluster-multiplicity dependence of the spatial precision, if understood, could be used to tune operating thresholds for optimal resolution rather than just maximum efficiency.

Reading between the lines

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

If the cluster-multiplicity pattern (optima near 2.4, 3.8, and 6) is truly geometry-driven, it should persist across different pixel pitches and sensor thicknesses; measuring the same quantity on a future variant with a different pixel size would test this conjecture.
The gain-calibration factor found for 25 µm epitaxial test structures may not transfer to the 50 µm epitaxial pixels; a direct measurement of the injection capacitance on 50 µm structures would settle whether the quoted absolute thresholds are accurate.
The paper leaves radiation tolerance open; if the 50 µm p-stop pixel keeps its efficiency after irradiation, it could become the baseline for the CBM-MVD, but that is an extrapolation beyond the data shown.
The telescope-precision convergence method assumes the reference planes do not share correlated systematic errors; using a mixed-technology reference telescope (e.g., strip or hybrid pixels) would provide an independent cross-check of the 5 µm figure.

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.

Desk Editor's Note

read the letter

Quick take: this is a straightforward, honest beam-test paper from a mature sensor R&D program. The new information is real: first in-beam results for MIMOSIS-2.1, first use of the 50 um epitaxial layer, and a gain calibration that corrects earlier electron-scale numbers by roughly a factor of two. The efficiency and resolution measurements are consistent with the headline claims, and the authors are appropriately cautious about what they can and cannot attribute.

What it does well: the sensor description is clear, the test setup is standard for the field, and the iterative telescope-precision estimate is a reasonable way to handle unknown reference-plane resolution. The paper also flags its own limitations: no radiation-tolerance claim, the efficiency improvement attributed to the biasing network is not yet confirmed, and the cluster-multiplicity pattern is left as an open question. That is the right level of epistemic humility for a prototype report.

The soft spots are in proportion: the main issue is that the central numbers come without statistical uncertainties. The efficiency claim ">99.9%" is a binomial point estimate; with no track count or confidence interval, the reader cannot tell whether the true efficiency is 99.95% or 99.6%. Same for the spatial precision plots: no error bars, lines "to guide the eye." For a detector paper whose whole point is to show that CBM requirements are met, this is a genuine gap. It is not a fatal one, because the efficiency clearly saturates in a way that makes a large deficit unlikely, and the resolution values hover in the 4-5 um band across thresholds. But it should be fixed in revision: report track counts, a Wilson interval for the efficiency, and per-point uncertainties on the resolution.

Two smaller things: the gain calibration was done on 25 um epitaxial test structures and is transferred to the 50 um pixels without a stated cross-check, and the beam test ran at uncontrolled 30-40 C. Both are acknowledged or at least visible in the text, and neither undermines the qualitative results. The telescope precision iteration is self-consistent, though it can converge to a biased value if reference planes share correlated errors; the authors could mention this as a caveat.

Verdict: this deserves a serious referee. It is a competent, useful measurement for the CBM MVD community and for anyone working on CMOS pixel sensors with similar readout architectures. I would accept it for review with the expectation of a revision that adds uncertainties. I would cite it for the gain-recalibration result if I worked on that sensor family.

Referee Report

3 major / 4 minor

Summary. This paper reports in-beam performance measurements of the MIMOSIS-2.1 CMOS Monolithic Active Pixel Sensor prototype developed for the CBM Micro Vertex Detector. The authors claim that all pixel variants of MIMOSIS-2.1 achieve a detection efficiency above 99.9%, that the spatial precision matches the targeted ~5 micrometers for suitable threshold settings, and that the new 50-micrometer epitaxial layer improves performance of non-irradiated devices. The paper also presents a recalibration of the in-pixel charge injection gain using a dedicated CE18 test structure and a 55Fe source, finding C_inj values about a factor of two above the design value. The authors explicitly state that no radiation-tolerance claim is made for the new sensing element.

Significance. If the headline numbers are established within well-quantified errors, this would be an important milestone for the CBM-MVD sensor development: MIMOSIS-2.1 is described as the first full-scale, full-feature prototype, and the claimed >99.9% efficiency and ~5 um spatial precision are directly relevant for the experiment's detector requirements. The paper has several genuine strengths: the gain calibration uses an external 55Fe source and is therefore not circular with respect to the beam-test results, the authors openly state the uncontrolled sensor temperature and the lack of radiation-tolerance claims, and the iterative telescope-precision procedure is described in sufficient detail to be scrutinized. However, the central performance claims are presented without any statistical uncertainties, track counts, or confidence intervals, which currently prevents the reader from verifying that the CBM requirements are met within errors.

major comments (3)

[Section 4, Fig. 4] The central claim that "All pixels ... provide a detection efficiency of > 99.9%" is not supported by the reported statistics. The paper gives no number of incident tracks, no number of matched hits, and no confidence intervals for the efficiency. For a binomial proportion, a point estimate of 99.9% is insufficient: for example, with 3000 tracks and three misses, the 95% Wilson lower bound is about 99.6%, which would not establish the claimed "> 99.9%". Please report the number of tracks, the number of misses, and binomial confidence intervals for each pixel variant and threshold setting shown in Fig. 4.
[Section 3 and Fig. 5] The quoted spatial precision is presented without error bars on any data point in Fig. 5, and the telescope-precision estimate rests on an iterative procedure that assumes the reference-plane residuals can be folded out reliably. If the reference planes share correlated errors, or if the iterative convergence settles on a biased value, the quoted DUT precision could be systematically too good. Please report per-point statistical errors, the converged telescope precision value, and at least one systematic check, for example splitting the data sample or varying the track search window.
[Section 2] The gain calibration is performed with CE18 test structures on a 25-micrometer epitaxial layer, but the resulting C_inj values are then used to interpret measurements of pixels with a 50-micrometer epitaxial layer. The paper does not justify that the collection-node capacitance and the charge-to-voltage conversion are identical between the two epitaxial thicknesses, nor does it provide an uncertainty on the measured C_inj values of 1.86 e/mV and 2.02 e/mV. Please state whether the calibration is transferable, and if so, on what grounds, and quantify the associated systematic uncertainty.

minor comments (4)

[Abstract and throughout] There are several minor language issues: "circuts" in the abstract should be "circuits", and "All pixels to provide a detection efficiency of > 99.9%" in Section 4 should be rephrased, for example as "All pixel variants provide a detection efficiency of > 99.9%."
[Fig. 5 caption] The caption reports that the precision for the other pixel dimension is about 10% worse, but this is not shown; a table or a second panel with both dimensions would make the claim reproducible.
[Section 2] The statement that the individual gain of more than 90% of the pixels remains within plus or minus 10% around the mean would benefit from stating the number of pixels tested and the width of the observed distribution.
[Fig. 4 caption] The caption does not define the axes or the units of the efficiency plot; please ensure all axes and threshold units are labeled in the final version.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the efficiency and spatial-precision claims are direct beam measurements, the gain calibration is externally referenced, and the telescope-precision estimate is self-consistent rather than self-defining.

full rationale

The central claims are experimental measurements, not derivations from their own inputs. The pixel gain calibration in Section 2 uses an external 55Fe source on the CE18 test structure; the resulting Cinj values are not fitted to the beam-test efficiencies or spatial resolutions, and the factor-of-two correction is independent of the beam results. The detection efficiency and spatial precision in Section 4 and Figures 4-5 are obtained from residual distributions in a beam telescope; no equation in the paper defines the claimed >99.9% efficiency or ~5 um precision in terms of the fitted parameters. The iterative telescope-precision estimate in Section 3 determines the reference-plane precision from measured residuals with an identical DUT and then unfolds that precision from the DUT residuals; this is a self-consistent calibration of measured quantities, not a reduction of the quoted sensor resolution to an assumed input. The self-citations ([3], [4], [6], [9]) provide architecture, analysis framework, and test-structure details, but the load-bearing performance numbers are supported by the beam data, not by those citations as premises. The paper even flags its own unconfirmed attribution of the efficiency improvement ('has yet to be confirmed') and its open geometrical explanation ('remains to be studied'), showing that those interpretive claims are not dressed up as derived predictions. The absence of statistical uncertainties on the efficiency and precision point estimates is a legitimate statistical-reporting concern, but it is not a circularity: it does not make the claim identical to its input. No specific circular step can be exhibited.

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

The paper introduces no new physical entities or ad hoc parameters. The measured gain factors are calibration constants, not free parameters fitted to the central result. The main assumptions are about the applicability of the 25 micrometer calibration to thicker epitaxial layers and the validity of the iterative telescope-precision estimate.

assumptions (3)

domain assumption The 50 micrometer epitaxial layer in the p-stop layout is likely fully depleted.
Section 2 and Figure 3 refer to [7, 8] for the p-stop layout; the 'likely fully depleted' status is inherited from those references, not demonstrated in this paper.
domain assumption The iterative telescope-precision estimation converges to the true track precision.
Section 3 describes the iterative method with no proof of convergence or uniqueness; the final precision is used for all DUTs.
ad hoc to paper Charge collection and gain behavior measured on 25 micrometer epitaxial test structures (CE18) apply to 50 micrometer epitaxial pixels.
Section 2 reports gain values from 'pixels with 25 µm epitaxial layer'; the paper does not show a separate calibration for the 50 µm epitaxial sensors used in the beam test.

how reviews work

0 comments

Cite this review

Pith. "Pith review of In beam performances of the MIMOSIS-2.1 CMOS Monolithic Active Pixel Sensor." pith.science (2026). https://pith.science/paper/SRJQOE2X

@misc{pith2026250205303,
  author       = {Pith},
  title        = {Pith review of: In beam performances of the MIMOSIS-2.1 CMOS Monolithic Active Pixel Sensor},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SRJQOE2X}},
  note         = {Machine review of arXiv:2502.05303}
}

abstract

MIMOSIS is a CMOS Monolithic Active Pixel Sensor developed to equip the Micro Vertex Detector of the Compressed Baryonic Matter (CBM) experiment at FAIR/GSI. The sensor will combine an excellent spatial precision of $5~\mu m$ with a time resolution of $5~\mu s$ and provide a peak hit rate capability of $\mathrm{\sim 80~ MHz/cm^2}$. To fulfill its task, MIMOSIS will have to withstand ionising radiation doses of $\sim 5~ \mathrm{MRad}$ and fluences of $\sim 7 \times 10^{13}~\mathrm{n_{eq}/cm^2}$ per year of operation. This paper introduces the reticle size full feature sensor prototype MIMOSIS-2.1, which was improved with respect to earlier prototypes by adding on-chip grouping circuts and by improving the analog power grid. Moreover, it features for a first time a $50~\mu m$ epitaxial layer, which is found to improve the performances of the non-irradiated device significantly. We discuss the in beam sensor performances as measured during beam tests at the CERN-SPS.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Detectors and Electronics for the CBM experiment at FAIR
physics.ins-det 2025-06 conditional novelty 2.0 of 10

The CBM detector and electronics systems are reported as mature, with series production underway and key performance-validated through beam tests at mCBM and deployments in STAR, HADES, and E16.

Reference graph

Works this paper leans on

12 extracted references · 12 canonical work pages · cited by 1 Pith paper

[1]

Stroth et al

J. Stroth et al. (editors) for the CBM collaboration,Technical Design Report for the CBM: Micro Vertex Detector (MVD), GSI-2022-00549,https://fair-center.eu/user/publications/ experiment-collaboration-publications#c56056

work page 2022
[2]

Ablyazimov et al

T. Ablyazimov et al. (the CBM collaboration), Eur. Phys. J. A 53, 60 (2017)

work page 2017
[3]

Deveaux et al., NIM A958(2020) 162653

M. Deveaux et al., NIM A958(2020) 162653

work page 2020
[4]

Darwish et al., JINST, 2023, 18

H. Darwish et al., JINST, 2023, 18. Jg., Nr. 06, S. C06013

work page 2023
[5]

Mager, NIM A824 (2016)

M. Mager, NIM A824 (2016)

work page 2016
[6]

Arnoldi-Meadows et al 2023 JINST 18 C04002

B. Arnoldi-Meadows et al 2023 JINST 18 C04002

work page 2023
[7]

Snoeys et al., NIM A871 (2017),

W. Snoeys et al., NIM A871 (2017),

work page 2017
[8]

Freeman et al., 2020 JINST 15 C03019

P.M. Freeman et al., 2020 JINST 15 C03019

work page 2020

Show all 12 references

[9]

Deveaux, 2019 JINST 14 R11001

M. Deveaux, 2019 JINST 14 R11001

2019
[10]

Chapter 6.4.1. in M. Šuljic,Study of monolithic active pixel sensors for the upgrade of the ALICE inner tracking system., PhD Thesis, University Triest (2018), https://hdl.handle.net/11368/2918673

2018
[11]

IPHC-PICSEL group,TAF analysis framework, https://github.com/jeromebaudot/taf
[12]

Mager,https://mmager.web.cern.ch/telescope/tracking.html – 5 –

M. Mager,https://mmager.web.cern.ch/telescope/tracking.html – 5 –

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.

[1] [1]

Stroth et al

J. Stroth et al. (editors) for the CBM collaboration,Technical Design Report for the CBM: Micro Vertex Detector (MVD), GSI-2022-00549,https://fair-center.eu/user/publications/ experiment-collaboration-publications#c56056

work page 2022

[2] [2]

Ablyazimov et al

T. Ablyazimov et al. (the CBM collaboration), Eur. Phys. J. A 53, 60 (2017)

work page 2017

[3] [3]

Deveaux et al., NIM A958(2020) 162653

M. Deveaux et al., NIM A958(2020) 162653

work page 2020

[4] [4]

Darwish et al., JINST, 2023, 18

H. Darwish et al., JINST, 2023, 18. Jg., Nr. 06, S. C06013

work page 2023

[5] [5]

Mager, NIM A824 (2016)

M. Mager, NIM A824 (2016)

work page 2016

[6] [6]

Arnoldi-Meadows et al 2023 JINST 18 C04002

B. Arnoldi-Meadows et al 2023 JINST 18 C04002

work page 2023

[7] [7]

Snoeys et al., NIM A871 (2017),

W. Snoeys et al., NIM A871 (2017),

work page 2017

[8] [8]

Freeman et al., 2020 JINST 15 C03019

P.M. Freeman et al., 2020 JINST 15 C03019

work page 2020