{"id":"346dd959-87aa-4906-b017-9792a8b196c4","arxiv_id":"2509.05450","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"A new analysis shows that internal crosstalk in silicon photomultipliers is mostly hole-triggered, with reflected light from the device surface contributing a significant electron-triggered component.","lead":"This paper measures crosstalk in two silicon photomultipliers and uses the differing voltage dependence of electron- versus hole-triggered avalanches to locate where crosstalk avalanches originate. It finds that surface-reflected secondary photons are a significant crosstalk source and proposes a diagnostic to guide future SiPM design.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Untested assumption that junction volume is independent of overvoltage could bias the extracted electron/hole crosstalk fractions in Eq. 4.","rationale":"The reader's weakest_assumption identifies the V-independence of the junction volume, which is indeed the most load-bearing assumption in the paper. The method in Section II defines λ* = αPe(V) + βPh(V) and then fits α and β as constants. If the depletion width changes with overvoltage, PA,p and PA,n change, making α and β functions of V. This would directly bias the central numerical claim. The paper explicitly states the assumption but gives no evidence or test. A C-V measurement or a voltage-dependent re-fit would settle the issue. I considered other potential concerns—such as the use of modeled Ph(V) for the FBK device or the reliance on DeCT for Ph in the HPK device—but those are secondary because the fit structure itself depends on the constant-α/β assumption, and the paper already notes the V-independence assumption without support. Therefore, the reader's conditional verdict is appropriate, and no change is needed.","tokens_in":10122,"tokens_out":2873,"duration_ms":32170,"concrete_test":"Measure the capacitance-voltage (C-V) characteristics of both SiPMs over the full operating overvoltage range (1–10 V). Use C(V) to extract the depletion width as a function of V. If the depletion width changes by more than ~5% over this range, the junction volume is not independent of V and the constant-α/β assumption fails. Alternatively, re-fit the data with α(V)=α0+α1·V and β(V)=β0+β1·V and test whether α1 and β1 are consistent with zero; if not, the original fit is biased.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that crosstalk is mostly hole-initiated (FBK 89%, HPK 92.6%) rests on fitting Eq. 4 with α and β as constants. This requires the photon absorption probabilities in the p and n regions, PA,p and PA,n, to be independent of overvoltage. However, as V increases, the depletion region widens, changing the volumes of the p- and n-side absorption regions and thus making α and β voltage-dependent. The paper states 'Assuming the junction volume is independent of V' (Section II) but provides no justification or sensitivity test. If α and β vary over the 1–10 V range, the fitted constants will be biased because Pe(V) and Ph(V) have different functional forms, so the decomposition would incorrectly attribute part of the V-dependence to carrier type rather than to changing absorption volumes. Given the tight uncertainties quoted (e.g., ±4%), even a modest V-dependence could shift the inferred fractions and the interpretation of surface reflections as a significant crosstalk source.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports measurements of internal crosstalk (iCT) probability as a function of overvoltage and temperature for two p-on-n SiPMs, the FBK VUV-HD3 and the Hamamatsu VUV4. The central methodological contribution is a decomposition of the reduced crosstalk rate λ* = αP_e(V) + βP_h(V), where P_e and P_h are electron- and hole-initiated avalanche triggering probabilities, allowing the electron- versus hole-triggered fractions of crosstalk to be extracted. The authors find that in both devices roughly 90% of crosstalk is hole-triggered and use the smaller electron-triggered component to argue that optical reflections from the device surface are a significant crosstalk source. For the HPK device, delayed crosstalk is used to directly extract P_h(V), a useful new measurement.","tokens_in":10485,"tokens_out":3182,"duration_ms":34551,"significance":"If the analysis is robust, the paper offers a practical diagnostic for SiPM development, particularly for backside-illuminated devices and for evaluating crosstalk mitigation structures. The experimental P_iCT(V) data are clean and the use of DeCT to measure P_h in a p-on-n device is a valuable, independent cross-check against companion modelling. However, the quantitative conclusions rest on an untested assumption about the voltage independence of the junction volume and on a fit whose electron-triggered fraction for HPK DiCT is statistically indistinguishable from zero. The diagnostic concept is promising, but the evidence for the surface-reflection mechanism in the HPK device is currently weaker than the text claims.","major_comments":[{"comment":"The decomposition λ* = αP_e(V) + βP_h(V) assumes that 'the junction volume is independent of V' so that PA,p and PA,n, the photon absorption probabilities in the p- and n-regions, are constant. If the depletion region widens with overvoltage, α and β are voltage-dependent and the fitted constants will be biased, because P_e and P_h have different functional forms. The paper states this assumption but provides neither a physics justification nor a sensitivity test. Given the small quoted uncertainties (e.g., ±4% for FBK), even a modest voltage dependence of PA could shift the inferred carrier fractions and alter the surface-reflection conclusion. Please provide a quantitative estimate of the depletion-width variation over the measured overvoltage range, or a sensitivity analysis using the same data.","section":"Section II, Eqs. (3)-(4)"},{"comment":"The HPK DiCT decomposition yields α = 36±26% (equivalently 37.2±26.8% in Table I). This is not statistically distinguishable from zero. Yet the Discussion section states that 'the significant number of electron-driven events in the VUV4' supports surface reflections as an important crosstalk path. That conclusion is not supported by the HPK data alone; it can only be claimed for the FBK device, where the electron fraction is 11±4%. Please revise the claim or provide additional constraints (e.g., a joint fit with the DeCT component or a prior on the expected electron fraction) to make the HPK result meaningful.","section":"Section IV, Table I and Fig. 7"}],"minor_comments":[{"comment":"There is a typographical error: the denominator should be N_{1.5PE}, not N_{0.5PE} as written in the text.","section":"Eq. (6)"},{"comment":"The placeholder '[xx need lidar reference]' should be replaced with a real citation for LIDAR applications.","section":"Introduction"},{"comment":"The legend entry 'CoH K, -110 P' appears to be a typo for 'CoHPK, -110'.","section":"Fig. 4 legend"},{"comment":"The functional form P_{e,h}(V) = 1 - e^{V/V_{e,h}} appears dimensionally unusual. If the intended expression is 1 - exp(-V/V_{e,h}), please correct the notation.","section":"Eq. (7)"}],"recommendation":"major_revision","confidential_remarks":"The analysis relies on data from a companion paper [10] (de St Croix et al., 2025), including P_e measurements and modelled P_h values. The editor should ensure that [10] is available and that the dependence on it is clearly disclosed. The HPK DiCT uncertainty is large enough that the headline surface-reflection conclusion for that device is currently overreaching."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I'll get to the point: the paper's real contribution is a method for decomposing internal crosstalk into electron- and hole-initiated avalanches, and it gives the first extraction of the hole avalanche triggering probability in a p-on-n SiPM from delayed-crosstalk data. That is new and useful. The measurements are careful, the temperature and overvoltage trends are clean, and the DeCT/DiCT separation using the first derivative of the waveform is a nice practical step. The cross-check between the DeCT-derived Vh (18.9±3.0) and the companion paper's value (16.62±1.2) gives some independent grounding, which is worth noting. The soft spot is exactly where the reader's stress test points: Eq. 4 assumes the junction volume is independent of overvoltage, so the absorption probabilities PA,p and PA,n are constant. That assumption is stated but not tested. As overvoltage increases, the depletion region widens, mainly into the n-side, so β (the hole-triggered weight) likely changes with V. If it changes, the fit will absorb that V-dependence into the Pe and Ph shapes, biasing the extracted α and β. The quoted uncertainties (e.g., ±4% for FBK) are tight enough that even a modest effect could move the interpretation. This is a real flaw, but it is not a load-bearing break unless the effect is large. The paper should either justify the assumption with device modeling or show a sensitivity test. The other caveat is the HPK DiCT electron fraction: 36±26% is so uncertain that the claim about surface reflections in that device is weak. The FBK result (11±4%) is more solid, and the overall hole dominance is likely correct, but the surface-reflection interpretation rests more on the FBK excess electron component plus the structural argument about metallized trenches than on the HPK DiCT fit. Who is this for? Anyone working on SiPM characterization or crosstalk mitigation in detector physics will get value from the method and the data. The paper deserves a serious referee; the central idea is worth pursuing even if the voltage-independence assumption needs more work. I'd send it to review and ask the authors to address that assumption directly.","headline":"A genuinely useful new decomposition of SiPM crosstalk into electron- and hole-triggered components, but the untested voltage-independence assumption in the model could shift the inferred fractions.","tokens_in":631,"tokens_out":983,"would_cite":true,"duration_ms":27406,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Measuring which carriers start SiPM crosstalk avalanches reveals surface reflections as a major path.","keywords":["silicon photomultiplier","internal crosstalk","direct crosstalk","delayed crosstalk","avalanche triggering probability","overvoltage","crosstalk mitigation","p-on-n SiPM"],"falsifier":"Measure λ* versus overvoltage over a wider range, or on a device whose depletion width is known to change, and check whether the fit yields constant α and β. A systematic drift of the fitted coefficients with voltage would mean the inferred carrier mix is an artifact; alternatively, inject photons at known depths using two-photon absorption to measure P_e and P_h directly and re-fit the same data.","tokens_in":10086,"feed_emoji":"🔬","tokens_out":5863,"duration_ms":58228,"temperature":0.7,"pith_summary":"This paper asks where the secondary photons that cause internal crosstalk in silicon photomultipliers are absorbed, and which type of charge carrier starts the resulting noise avalanche. It shows experimentally that in two p-on-n SiPMs, roughly nine out of ten crosstalk avalanches are initiated by holes, meaning the photons are absorbed in n-doped bulk or substrate regions; the remaining electron-initiated share is larger than geometry alone would suggest. The authors argue that this excess electron component is the signature of secondary photons reflecting off the device surface into neighboring pixels, a path that survives even in a device with metallized trenches that block direct optical transmission. They also measure, for the first time in a p-on-n SiPM, the hole avalanche-triggering probability directly from delayed-crosstalk rates. If the method holds, it gives device developers a diagnostic to identify and suppress the dominant crosstalk path in new SiPM designs.","feed_headline":"Holes start 89% of internal crosstalk avalanches in two SiPMs","feed_subtitle":"The new decomposition tells device makers where to cut noise, not just how much exists.","key_machinery":"The central object is the reduced internal crosstalk λ*, defined as the mean number of crosstalk avalanches per primary pulse divided by charge gain, and modeled as α·P_e(V) + β·P_h(V). Since electron and hole avalanche triggering probabilities have markedly different overvoltage dependencies — P_e saturates quickly, P_h rises slowly — the shape of λ*(V) reveals which carrier starts the avalanches, with α and β proportional to photon absorption in the p and n regions. Delayed crosstalk supplies a direct experimental measure of P_h, because delayed events are almost entirely hole-triggered once electrons drift to the junction in under 10 ps.","core_discovery":"The paper's central claim is that internal crosstalk in both tested SiPMs is dominated by hole-initiated avalanches — 89±4% for the FBK VUV-HD3 and 92.6±5.3% for the Hamamatsu VUV4 — but that the electron-initiated fraction is too large to be explained by direct transmission through trenches or simple volume ratios. The authors model the reduced crosstalk λ* = λ_iCT/gain as a linear combination α·P_e(V) + β·P_h(V), where P_e and P_h are the measured or modeled avalanche triggering probabilities for electrons and holes, and α and β reflect the fractions of secondary photons absorbed in p-type and n-type regions. Fitting this form to data over 1–10 V of overvoltage yields the carrier mix. Beca","pith_inferences":["The same decomposition could be applied to backside-illuminated SiPMs, where the optical paths differ and surface reflections may contribute an even larger share of crosstalk.","If the junction volume actually widens with overvoltage, the fitted α and β would drift with voltage; fitting over a wider overvoltage range and checking stability would reveal how much this biases the inferred carrier mix.","The measured electron/hole ratio depends on overvoltage because P_e saturates faster than P_h, so operating at lower overvoltage may suppress reflection-driven crosstalk more than bulk-driven crosstalk — a prediction that could be tested by repeating the analysis at several fixed voltages.","Because the authors find no significant temperature dependence over -39°C to -110°C, room-temperature operation might behave differently if thermally activated absorption or trapping alters the secondary-photon paths; this is an open extension, not a claim of the paper."],"forward_implications":["Crosstalk in p-on-n SiPMs is mostly hole-triggered because secondary photons are absorbed predominantly in the n-type bulk or substrate, not in the p-implant.","Even with metallized trenches blocking direct optical transmission, surface reflections can generate a substantial electron-triggered crosstalk component (11±4% in the FBK device; 36±26% of direct crosstalk in the HPK device).","Delayed crosstalk is a purely hole-triggered process and can be used to measure the hole avalanche-triggering probability directly, removing a model dependence for this device type.","Designing passivation layers or anti-reflection coatings to minimize surface reflections, and operating at lower overvoltage to suppress P_h/P_e, are concrete ways to reduce crosstalk in future devices.","The method can be generalized to other SiPM structures, and would become fully model-independent if two-photon absorption injection allowed direct measurement of P_e and P_h at arbitrary depths."],"supporting_citations":[{"why":"Provides measured electron triggering probabilities, modeled hole probabilities, and junction-depth data for both devices, used as inputs in the fit.","marker":"[10]"},{"why":"Describes the VERA vacuum setup and gain calibration on which all crosstalk measurements rely.","marker":"[11]"},{"why":"Supports the linear relation between secondary photon number and avalanche gain used to define reduced crosstalk.","marker":"[12]"},{"why":"Describes the SiO2/poly-Si trench composition of the first device, grounding the direct-transmission path.","marker":"[13]"},{"why":"Documents the metallized trench of the second device, which blocks direct trench crosstalk.","marker":"[17]"},{"why":"Supplies the distinct overvoltage trends of electron vs hole avalanche triggering probabilities used in the fit.","marker":"[18]"},{"why":"Gives the sub-10 ps electron drift time, justifying the attribution of delayed events to hole-triggered avalanches.","marker":"[19]"},{"why":"Provides the electron/hole ionization coefficient ratio used for the lower-field mitigation suggestion.","marker":"[20]"}],"fun_headline_variants":["Optical reflections drive SiPM internal crosstalk","Hole avalanches dominate SiPM crosstalk","New method pinpoints SiPM crosstalk sources","SiPM crosstalk traced to surface reflections","Hole-triggered avalanches cause most SiPM noise"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The analysis assumes the junction volume stays fixed as overvoltage changes, so α and β remain constant; if the depletion region widens with voltage, the inferred electron/hole split would be biased.","fun_headline_variants_meta":{"raw":{"variants":["Optical reflections drive SiPM internal crosstalk","Hole avalanches dominate SiPM crosstalk","New method pinpoints SiPM crosstalk sources","SiPM crosstalk traced to surface reflections","Hole-triggered avalanches cause most SiPM noise"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000469,"raw_usage":{"total_tokens":2193,"prompt_tokens":784,"completion_tokens":1409,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":528,"completion_tokens_details":{"reasoning_tokens":1341}},"tokens_in":528,"tokens_out":1409,"duration_ms":10280,"temperature":1.0,"reasoning_tokens":1341,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T05:23:43.402719+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure λ* versus overvoltage over a wider range, or on a device whose depletion width is known to change, and check whether the fit yields constant α and β. A systematic drift of the fitted coefficients with voltage would mean the inferred carrier mix is an artifact; alternatively, inject photons at known depths using two-photon absorption to measure P_e and P_h directly and re-fit the same data.","supporting_citations":[{"cited_title":"Mapping the photon detection efficiency of vuv sensitive sipms from the ultra-violet to the near infra- red,","cited_arxiv_id":null,"evidence_quote":"Provides measured electron triggering probabilities, modeled hole probabilities, and junction-depth data for both devices, used as inputs in the fit."},{"cited_title":"Measurements of the quantum yield of silicon using Geiger- mode avalanching photodetectors,","cited_arxiv_id":null,"evidence_quote":"Describes the VERA vacuum setup and gain calibration on which all crosstalk measurements rely."},{"cited_title":"Stimulated secondary emission of single-photon avalanche diodes,","cited_arxiv_id":null,"evidence_quote":"Supports the linear relation between secondary photon number and avalanche gain used to define reduced crosstalk."},{"cited_title":"NUV-Sensitive Silicon Photomultiplier Technologies Developed at Fondazione Bruno Kessler,","cited_arxiv_id":null,"evidence_quote":"Describes the SiO2/poly-Si trench composition of the first device, grounding the direct-transmission path."},{"cited_title":"Characterization of the Hamamatsu VUV4 MPPCs for nEXO,","cited_arxiv_id":null,"evidence_quote":"Documents the metallized trench of the second device, which blocks direct trench crosstalk."},{"cited_title":"Characterization of SiPM Avalanche Triggering Probabilities,","cited_arxiv_id":null,"evidence_quote":"Supplies the distinct overvoltage trends of electron vs hole avalanche triggering probabilities used in the fit."},{"cited_title":"Timing mea- surements with silicon single photon avalanche diodes: principles and perspectives,","cited_arxiv_id":null,"evidence_quote":"Gives the sub-10 ps electron drift time, justifying the attribution of delayed events to hole-triggered avalanches."},{"cited_title":"On the avalanche initiation probability of avalanche diodes above the breakdown voltage,","cited_arxiv_id":null,"evidence_quote":"Provides the electron/hole ionization coefficient ratio used for the lower-field mitigation suggestion."}],"review_version":1}