{"id":"c8bec80e-1f99-48cd-86f4-23d30eddac86","arxiv_id":"2501.03812","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Under true dark conditions, both tested silicon photomultipliers show a second, tunneling-like breakdown between 20 and 30 kelvin that limits operating voltage, and the usable range recovers below 5 kelvin.","lead":"Two commercial silicon light sensors were cooled from room temperature to 90 millikelvin and tested under true dark conditions. The key result is a second breakdown that appears at 20-30 kelvin and collapses the usable voltage range, with both chips recovering below 5 kelvin.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The identification of Vt as an intrinsic tunneling breakdown is not yet secure, because the evidence is total photocurrent only and self-sustaining afterpulse trains or thermal runaway in the quenching network are not excluded.","rationale":"The reader's weakest assumption and my concern are the same: the sharp photocurrent rise at Vt is interpreted as an intrinsic second breakdown, but the manuscript provides only total-current measurements and does not exclude afterpulse-runaway or thermal artifacts. This is the most load-bearing issue because the abstract's operating-range collapse near 20-30 K depends directly on Vt being a real device threshold. If the rise were instead caused by self-sustaining afterpulse trains or heating in the quenching network, the operating-range minimum would be an experimental artifact rather than a property of the SiPM. The paper's shutter control is a genuine strength and correctly distinguishes IR-leak-induced afterpulse trains from a dark condition, but it does not test what happens at Vt. The authors themselves note in Section 3.5 that afterpulse trains can last up to 1 ms and cause heating effects, and they recommend further afterpulsing analysis in the conclusions. Thus the concern is concrete and tied to their own reported observations. The proposed oscilloscope test at fixed temperature with pulse counting and thermometer monitoring would settle the issue. Because the reader already assigned a conditional verdict that explicitly asks for this check, I do not recommend changing the verdict; the concern is real but does not, by itself, invalidate the core observation if the test passes.","tokens_in":8617,"tokens_out":3882,"duration_ms":42178,"concrete_test":"Record the SiPM output waveform at T ≈ 23 K with the shutter closed while stepping the bias in 100 mV increments across Vt, using long acquisitions (e.g., 10 s per step) and simultaneously monitoring the board thermometer. Count resolved single-cell pulses and afterpulse trains at each step. If resolved pulses or self-sustaining trains appear at biases below Vt, or if Vt shifts with sweep direction or integration time, the 'tunneling breakdown' is not an intrinsic threshold. If no pulses occur below Vt and the first event at Vt is a step-like simultaneous cell breakdown with no preceding trains, the intrinsic interpretation is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim—that Vt marks a second, tunneling breakdown of the SPAD cells—is supported only by the shape of the total photocurrent-versus-voltage curve in Fig. 4. The paper does not show single-cell Geiger pulses, dark-count rates, or chip-temperature records in the Vt region. This matters because Section 3.5 itself reports that below 40 K afterpulsing probability increases, afterpulse trains can last roughly 1 ms, and these trains produce heating effects from afterpulse-train-induced photocurrent. A steep current rise at Vt is therefore equally compatible with a thermal/afterpulse runaway in the quenching network as with an intrinsic avalanche-independent tunneling breakdown. The open-versus-closed shutter control distinguishes IR-induced afterpulse trains from a supposedly true dark condition, but it does not rule out runaway initiated by the residual below-1-Hz dark counts, whose rate increases strongly with overvoltage. Moreover, the pronounced minimum in Vt near 23 K could reflect the temperature dependence of trap release times or quench-resistor behavior rather than a band-to-band tunneling threshold. Without pulse-level verification at Vt, the reported 0.3 V operating window near 25 K may be a setup artifact.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a cryogenic characterization of two commercial SiPMs (Onsemi MicroFJ-30035-TSV and Hamamatsu S13370-6050CN) from 90 mK to 40 K, including pulse shapes, breakdown voltages, dark count rates, afterpulsing, photon detection efficiency, and thermal cycling. The central new observation is a steep, temperature-dependent rise in photocurrent at a voltage Vt above the normal breakdown voltage Vbd under closed-shutter (true dark) conditions below 40 K, interpreted as a second tunneling breakdown that limits the usable overvoltage range, with a pronounced minimum operating range near 25 K. The authors also reproduce earlier reports of afterpulse-train-induced photocurrent when the shutter is open and find that both devices survive thermal cycling.","tokens_in":84,"tokens_out":4104,"duration_ms":90676,"significance":"If the identification of Vt as an intrinsic breakdown is correct, the paper provides an important practical constraint for cryogenic SiPM operation: the operating voltage window collapses near 20-30 K and recovers below 5 K, which is directly relevant to experiments using SiPMs in dilution refrigerators. The use of a cryogenic shutter to achieve a true dark condition is a methodological improvement over prior work, and the paper includes a useful dataset of pulse shapes, DCR, PDE, and thermal-cycling behavior, with code and data available on GitHub. However, the central interpretation of Vt rests on total-photocurrent measurements alone; the evidence does not yet exclude afterpulse-train or thermal-runaway mechanisms, so the main quantitative claim requires additional verification.","major_comments":[{"comment":"The claim that the steep photocurrent rise at Vt is a tunneling breakdown of the SPAD microcells is supported only by a total-current-versus-voltage curve. The paper does not show single-cell Geiger pulses, dark-count-rate versus overvoltage, or chip-temperature records in the Vt region. Section 3.5 itself reports afterpulse trains lasting up to 1 ms and heating from afterpulse-train-induced photocurrent, so a self-sustaining afterpulse or thermal runaway in the quenching network is a plausible alternative explanation. Please add pulse-level waveforms at Vt (or a DCR-vs-overvoltage measurement) and a chip-temperature monitor during the current rise, or explicitly rule out dark-count-initiated trains, before asserting that Vt is an intrinsic breakdown.","section":"Sec. 3.3, Fig. 4"},{"comment":"The quantitative operating-range claim (e.g., 0.3 V at 25 K for Hamamatsu) has no stated uncertainty or measurement statistics. No error bars appear in Figs. 4, 5, or 7, and the number of repeated measurements is not given. Because the operating range is a difference between two voltages determined from curve features, the uncertainty on Vt and Vbd is essential for assessing whether the 0.3 V window is significant. Please provide uncertainties and a description of how Vt was extracted from the data.","section":"Sec. 3.4, Fig. 5"},{"comment":"The paper asserts a 'true dark condition' with the closed shutter, but also states that residual dark counts below 1 Hz remain, possibly from scattered thermal photons or cosmic rays. Since dark count rate increases with overvoltage and Section 3.5 notes that afterpulsing probability increases with overvoltage, the sub-1 Hz dark counts could seed afterpulse trains that masquerade as a breakdown current. Please quantify the dark count rate as a function of overvoltage up to Vt, or otherwise show that the closed-shutter photocurrent rise is not initiated by these residual events.","section":"Sec. 3.3"},{"comment":"The interpretation of the pronounced minimum in Vt near 23 K as a tunneling threshold is not supported by any microscopic model or temperature-dependence comparison. Trap release times and quench-resistor behavior also vary strongly in this range, and the paper provides no evidence distinguishing these. If the pulse-level tests above confirm an intrinsic breakdown, the temperature dependence should be discussed in terms of the known physics of band-to-band or trap-assisted tunneling.","section":"Sec. 3.3"}],"minor_comments":[{"comment":"The UV wavelength is given as both 275 nm (Abstract, Table 1) and 270 nm (Sec. 2, Sec. 3.6); please make the notation consistent.","section":"Abstract, Sec. 2, Table 1"},{"comment":"The Hamamatsu model is referred to as S13371-6050CN in the Conclusions, while Sec. 2 and Table 1 use S13370-6050CN; please correct the discrepancy.","section":"Conclusions"},{"comment":"No error bars or discrete measurement points are visible in the operating-range plots; please clarify whether the curves are fits or interpolations and show the underlying data.","section":"Fig. 5"},{"comment":"The statement 'We did not do a quantitative analysis of the afterpulsing effects' leaves the afterpulse-trains claim qualitative; a quantitative bound on afterpulse probability or train duration would strengthen the paper.","section":"Sec. 3.5"},{"comment":"The sentence 'the dark count rate strongly depends on the position of the shutter above SiPM' is vague; specify the shutter geometry and the measured DCR values for open and closed positions.","section":"Sec. 3.2"},{"comment":"The author contributions contain a grammatical error ('Data collection and was done by Tom Kiilerich'), and the sentence 'The fiber has also been thermalized at several temperatures (4 K, 1 K, mixing chamber) along the way' is awkward; please rephrase.","section":"Sec. 2"}],"recommendation":"major_revision","confidential_remarks":"The paper is a useful experimental characterization with a potentially important finding. My main concern is the identification of Vt; if the authors can provide pulse-level or DCR evidence and temperature monitoring, the paper would be suitable for publication in physics.ins-det. I would not require a full theoretical model of the tunneling, but the current evidence is insufficient to exclude afterpulse runaway. Note also that the paper's claim that the effect 'has not been previously reported' should be checked against the literature on cryogenic SiPM breakdown, since other groups working at similar temperatures may have observed related phenomena."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a look if you work with SiPMs in dilution fridges. The genuinely new thing is the closed-shutter control: by blocking room-temperature IR from the fiber, they get a true dark condition below 100 K, and then they see a second, sharp photocurrent rise at a voltage Vt well above Vbd, with a pronounced minimum near 23 K and recovery below 5 K. That observation, plus the comparison with the open-shutter case reproducing Zhang's afterpulse-runaway behavior, is solid and useful. They also ship code and data on GitHub, which is more than most instrumentation papers do.\n\nWhat I'd push back on: they call it tunneling breakdown in the abstract, and that's speculation. The evidence is total photocurrent versus voltage. They don't show single-cell Geiger pulses, dark count rates, or chip temperature at Vt. Their own Section 3.5 says afterpulsing probability increases with overvoltage and can create roughly 1 ms trains that cause heating. So the same steep current rise could be an afterpulse/thermal runaway in the quenching network rather than an intrinsic avalanche-independent tunneling. The shutter removes IR-initiated trains, but residual dark counts below 1 Hz remain, and those could seed runaway at sufficiently high overvoltage. The temperature dependence of Vt with a minimum near 23 K is also consistent with trap-release or quench-resistor behavior. So the practical warning—operating range collapses near 23 K—might be real for this circuit, but the mechanism claim goes beyond the data. They also have no error bars, one sample per model, and no quantitative afterpulsing correction in the PDE curves, though the PDE result agrees with prior work.\n\nIf I were the editor, I'd send it to review. The empirical mapping of Vt(T) for two commercial devices is a useful data point, the shutter method is a step forward, and the concerns are addressable. The authors should show pulse waveforms or at least a dark-count histogram near Vt, add error bars, and soften the abstract. Credit where due: the paper is clearly written, the open/closed shutter control is a real control, and the self-citation is legitimate context. The central observation is probably real; the mechanism is not established.","headline":"Useful cryogenic SiPM characterization with a real methodological advance (dark shutter), but the headline 'tunneling breakdown' is an unproven mechanism, not a settled result.","tokens_in":9410,"tokens_out":2559,"would_cite":true,"duration_ms":25702,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["85.60.Gz","07.20.Mc"],"model":"deepseek-v4-flash","headline":"Under true dark conditions below 40 K, both tested silicon photomultipliers develop a second tunneling breakdown that sharply limits their usable bias range, with the narrowest window near 23 K; below 5 K they recover a 3-5 V operating…","keywords":["silicon photomultiplier","SiPM","cryogenic detector","tunneling breakdown","dark count rate","photon detection efficiency","afterpulsing","low-temperature characterization"],"falsifier":"At 23 K with the shutter closed, sweep the bias across $V_t$ while counting individual Geiger pulses with a fast oscilloscope. If $V_t$ is a true second breakdown, single-photon pulse rate should rise steeply with overvoltage and pulse shape should remain stable; if the photocurrent rise is a setup artifact, current will grow without countable pulses or the chip temperature will drift upward from afterpulse-train heating.","tokens_in":8359,"feed_emoji":"❄️","tokens_out":14179,"duration_ms":110600,"temperature":0.7,"pith_summary":"The paper reports a cryogenic limitation of silicon photomultipliers (SiPMs): below 40 K, under a true dark condition in which all room-temperature infrared radiation is blocked by a cold shutter, both tested devices show a sudden, steep growth of photocurrent at a voltage $V_t$ above the normal Geiger breakdown $V_{bd}$. The paper attributes this second breakdown to band-to-band or trap-assisted tunneling and shows that it defines the upper edge of the usable bias range. The operating window between $V_{bd}$ and $V_t$ is narrowest near 23 K, where the Hamamatsu chip allows only about 0.3 V of overvoltage, and recovers to 3-5 V below about 5 K for both chips. This matters for cryogenic single-photon detection because it gives a concrete bias ceiling for SiPMs in dilution refrigerators and explains why earlier measurements saw an apparent afterpulse-induced current rise: with the shutter open, infrared leakage masks the tunneling breakdown.","feed_headline":"Tunneling breakdown limits cryogenic SiPM bias near 23 K","feed_subtitle":"A second tunneling breakdown appears under true dark below 40 K; below 5 K both chips recover 3-5 V.","key_machinery":"The central object is the photocurrent-versus-bias characteristic of a SiPM measured with a closed cryogenic shutter, which yields two distinct thresholds: the usual Geiger avalanche at $V_{bd}$ and a second, sharply rising breakdown at $V_t$. The cryogenic shutter is the enabling apparatus because it blocks room-temperature infrared photons leaking through the optical fiber, removing the afterpulse-train photocurrent that otherwise masks $V_t$; the paper defines the cryogenic operating voltage range as the interval between $V_{bd}$ and $V_t$.","core_discovery":"Using a cryogenic shutter to create a true dark condition, the study establishes that the Hamamatsu S13370-6050CN and Onsemi MicroFJ-30035-TSV SiPMs both exhibit a normal Geiger breakdown at $V_{bd}$ followed at higher bias by a second breakdown $V_t$, a steep photocurrent rise whose value depends strongly on temperature. $V_t$ has a pronounced minimum near 23 K, shrinking the overvoltage range to about 0.3 V for the Hamamatsu device; below roughly 5 K the range recovers to 3-5 V for both devices, and the Onsemi chip retains more than 12 V of operating range near 40 K. The paper interprets $V_t$ as direct band-to-band or trap-assisted tunneling in the high-field SPAD junction, an effect observable only when thermal excitations and stray infrared photons are frozen out, which is why previous work attributed the same photocurrent growth to self-sustaining afterpulse trains.","pith_inferences":["The authors leave implicit that the usable bias window is non-monotonic in temperature, so a fixed bias cannot serve a cryogenic experiment that sweeps temperature; a working detector would need to track $V_t$ in real time.","A testable extension is to count single-cell Geiger pulses across $V_t$ at 23 K with the shutter closed; if $V_t$ is intrinsic tunneling breakdown, countable single-photon pulses should persist and grow with overvoltage, whereas a setup artifact would show photocurrent growth without countable pulses.","Because the second breakdown is visible only when stray infrared is blocked, earlier open-shutter characterizations of cryogenic SiPMs may have overestimated the usable overvoltage range; re-testing those devices with a cold shutter would clarify whether the effect is generic.","The strong temperature dependence of $V_t$ between 20 and 30 K suggests a thermally assisted tunneling process, so modelling $V_t(T)$ could extract trap energy levels and guide dopant engineering for cryogenic SiPMs."],"forward_implications":["If the paper is right, cryogenic SiPM users must determine $V_t$ as well as $V_{bd}$; the usable overvoltage window is $V_t - V_{bd}$, and near 23 K that window nearly closes for the Hamamatsu chip.","Below about 5 K, both tested chips regain a 3-5 V overvoltage window, so dilution-refrigerator experiments can operate at bias margins comparable to room-temperature recommendations.","The Onsemi chip at about 40 K has a usable window larger than 12 V, making it attractive for photocurrent measurements requiring a wide dynamic range.","The Onsemi fast-output channel fails below 110 K, but a room-temperature high-pass RC filter on the normal output restores clean pulse counting, so metal-quench-resistor chips lose their apparent low-temperature advantage."],"supporting_citations":[{"why":"Prior cryogenic SiPM study that reported a rapid photocurrent rise at high overvoltage and attributed it to self-sustaining afterpulse trains; the paper reproduces this with an open shutter and contrasts it with the tunneling breakdown seen in true dark.","marker":"[7]"},{"why":"Supplied the biasing-board design and the pulse-height method used to determine breakdown voltage at cryogenic temperatures.","marker":"[3]"},{"why":"Source for band-to-band tunneling as the dominant dark-count mechanism below about 200 K, underpinning the interpretation of the second breakdown.","marker":"[12]"},{"why":"Onsemi application note giving the square-root-current versus voltage linear-fit method for room-temperature breakdown-voltage calibration.","marker":"[10]"},{"why":"Hamamatsu datasheet providing room-temperature PDE values and device specifications used as the calibration anchor for relative photon detection efficiency.","marker":"[5]"},{"why":"Onsemi datasheet providing room-temperature PDE values and device specifications used as the calibration anchor for relative photon detection efficiency.","marker":"[6]"},{"why":"Previous cryogenic PDE measurements of the Hamamatsu VUV4 chip that the present PDE data agree with, anchoring the relative-efficiency comparison.","marker":"[8]"},{"why":"Master's thesis documenting the measurement setup and data-analysis methodology in detail, cited for the experimental procedure.","marker":"[11]"}],"fun_headline_variants":["Cryogenic SiPMs face tunneling limit near 23 K, recover below 5 K","Tunneling breakdown caps SiPM bias window at ~23 K, then recovers","Second breakdown in SiPMs at 23 K: overvoltage range shrinks to 0.3 V","True dark exposes SiPM tunneling breakdown, bias range bounces back under 5 K","SiPMs survive 90 mK: tunneling dip at 23 K gives way to 3-5 V range"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The sharp photocurrent rise at $V_t$ is an intrinsic breakdown of the SPAD microcells rather than an artifact of the bias circuit, amplifier chain, or self-heating from afterpulse trains; the paper measures only total photocurrent versus voltage and does not verify single-cell Geiger-pulse behavior at $V_t$.","fun_headline_variants_meta":{"raw":{"variants":["Cryogenic SiPMs face tunneling limit near 23 K, recover below 5 K","Tunneling breakdown caps SiPM bias window at ~23 K, then recovers","Second breakdown in SiPMs at 23 K: overvoltage range shrinks to 0.3 V","True dark exposes SiPM tunneling breakdown, bias range bounces back under 5 K","SiPMs survive 90 mK: tunneling dip at 23 K gives way to 3-5 V range"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000821,"raw_usage":{"total_tokens":3589,"prompt_tokens":934,"completion_tokens":2655,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":550,"completion_tokens_details":{"reasoning_tokens":2529}},"tokens_in":550,"tokens_out":2655,"duration_ms":18418,"temperature":1.0,"reasoning_tokens":2529,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:46:32.935773+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"At 23 K with the shutter closed, sweep the bias across $V_t$ while counting individual Geiger pulses with a fast oscilloscope. If $V_t$ is a true second breakdown, single-photon pulse rate should rise steeply with overvoltage and pulse shape should remain stable; if the photocurrent rise is a setup artifact, current will grow without countable pulses or the chip temperature will drift upward from afterpulse-train heating.","supporting_citations":[{"cited_title":"Journal of Instrumenta- tion 17(06), 06024 (2022) https://doi.org/10","cited_arxiv_id":null,"evidence_quote":"Prior cryogenic SiPM study that reported a rapid photocurrent rise at high overvoltage and attributed it to self-sustaining afterpulse trains; the paper reproduces this with an open shutter and contrasts it with the tunneling breakdown seen in true dark."},{"cited_title":"Review of Scientific Instruments 94(12), 123202 (2023) https://doi.org/10.1063/5.0170629","cited_arxiv_id":null,"evidence_quote":"Supplied the biasing-board design and the pulse-height method used to determine breakdown voltage at cryogenic temperatures."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Source for band-to-band tunneling as the dominant dark-count mechanism below about 200 K, underpinning the interpretation of the second breakdown."},{"cited_title":"OnSemiconductors Appli- cation Notes (2021)","cited_arxiv_id":null,"evidence_quote":"Onsemi application note giving the square-root-current versus voltage linear-fit method for room-temperature breakdown-voltage calibration."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Hamamatsu datasheet providing room-temperature PDE values and device specifications used as the calibration anchor for relative photon detection efficiency."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Onsemi datasheet providing room-temperature PDE values and device specifications used as the calibration anchor for relative photon detection efficiency."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Previous cryogenic PDE measurements of the Hamamatsu VUV4 chip that the present PDE data agree with, anchoring the relative-efficiency comparison."},{"cited_title":"Mas- ter’s thesis, University of Turku (2024)","cited_arxiv_id":null,"evidence_quote":"Master's thesis documenting the measurement setup and data-analysis methodology in detail, cited for the experimental procedure."}],"review_version":1}