{"id":"703d1fb8-0742-4a14-abdb-f5a72f58a146","arxiv_id":"2504.13098","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Neutron and proton irradiation of RSD/AC-LGAD sensors shows lower gain-layer removal coefficients than standard LGADs, attributed to donor removal in the lightly doped n+ layer, with donor removal coefficients measured for the first time in this doping range.","lead":"R&D groups at KIT, INFN, and FBK irradiated resistive AC-coupled silicon detectors (RSDs) with neutrons and protons up to 3.5e15 cm^-2 and measured how their gain layers degrade. The measurements show RSDs lose gain doping more slowly than standard LGADs, and the authors propose donor removal in the resistive n+ layer explains the difference.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Donor-removal attribution is not quantitatively established: measured cD values are never shown to produce the observed c reduction; a TCAD or analytic bridge is needed.","rationale":"The reader's conditional verdict is appropriate, and my independent read agrees. The direct measurements are valuable and the paper is transparent about limitations, especially the preliminary single-wafer TCT result. The main weakness is not internal inconsistency but inferential distance: c and cD are measured on different structures (sensors versus Van der Pauw test structures) and connected only by a verbal geometric argument. The 300 pF proxy issue raised by the reader is real and contributes to the problem, but even if the proxy is accepted, the missing quantitative link between cD and the VGL shift is the load-bearing assumption. A TCAD or analytic calculation would settle whether the measured donor-removal coefficients are large enough to explain the observed reduction in c. The charge-sharing conclusion is explicitly preliminary and not central; it does not alter the verdict. I recommend keeping the manuscript as a conditional R&D study: accept the electrical data as measurements, but reframe the donor-removal interpretation as a hypothesis requiring quantitative confirmation.","tokens_in":5958,"tokens_out":5858,"duration_ms":57216,"concrete_test":"Run a calibrated TCAD model (or an analytic Poisson solver) of the FBK RSD production for wafers W3, W4, W6, and W14. For each fluence point, reduce the n+ layer donor concentration according to the measured cD values (converting sheet-conductance loss to active donor loss with a fixed or fitted profile), apply the standard LGAD acceptor-removal coefficient (c ≈ 5e-16 cm2 for neutrons, c ≈ 12e-16 cm2 for protons) to the p+ gain implant, and compute the CV curves. Then extract Vth at 300 pF with the same protocol and fit c. Compare the simulated c values to Table 2; if they do not match within uncertainties, the donor-removal mechanism is insufficient and the attribution fails. Running the same simulation with donor removal switched off would isolate the effect.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The claim that donor removal in the n+ resistive layer causes the lower c coefficients in RSDs is the central result, but the paper never supplies a quantitative connection between the two measured quantities. Table 2 shows c from CV fits; Table 3 shows cD from Van der Pauw sheet-conductance fits. The working hypothesis (Sec. 4) states that donor removal 'shrinks' the n+ layer, increasing d in Eq. (2) and raising VGL, thereby partially compensating acceptor removal. However, d and the depletion width inside the n+ layer are not measured before/after irradiation, and no calculation converts the measured cD values into a predicted change in VGL or c. The qualitative ordering (W3 has the highest cD and lowest c) is suggestive, but wafer-to-wafer variations in the gain implant, or the Vth proxy shifting with bulk damage and the 1 kHz-to-10 Hz measurement change, could produce the same ordering. The paper itself concedes in Sec. 3.2 that c tracks total sensor depletion rather than the gain layer alone, so the LGAD comparison is already indirect. The donor-removal explanation is therefore plausible but unproven; the conclusion 'both results are attributed to donor removal' overstates the evidence.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports irradiation studies of FBK RSD/AC-LGAD sensors from the second production. Sensors from wafers W3, W4, W6, and W14 were irradiated with neutrons and 23 MeV protons to fluences of 1.0, 2.0, and 3.5 x 10^15 cm^-2 (test structures also to 5.0 x 10^15 cm^-2). The sensors were characterized with CV measurements, from which the voltage at a fixed 300 pF capacitance threshold was used as a proxy for the gain-layer depletion voltage; the fraction Vth/VGL was fit to f(Phi)=B exp(-c Phi) to extract the coefficient c. Donor removal was measured directly with Van der Pauw sheet-conductance structures, giving coefficient cD. TCT laser scans were used to study charge sharing after irradiation. The central claim is that the c coefficients (2.6-3.2 x 10^-16 cm^2 for neutrons, 8.1-9.9 x 10^-16 cm^2 for protons) are lower than typical LGAD acceptor-removal coefficients, and that this is caused by donor removal in the n+ resistive layer, which widens the multiplication region and partially compensates acceptor removal. The paper also reports an inverse correlation of cD with initial n+ doping and preliminary TCT evidence that charge sharing is unchanged by irradiation.","tokens_in":6224,"tokens_out":4595,"duration_ms":42885,"significance":"If the proposed donor-removal compensation mechanism is correct, the result is significant: RSD gain layers would be intrinsically more radiation tolerant than standard LGADs, with implications for 4D tracking at future colliders and for LGAD design. The paper has clear strengths: direct Van der Pauw measurements of donor removal in a doping range not previously studied, cross-checked at KIT and Perugia with reproducible values; complete fluence series for electrical data; and falsifiable predictions that can guide follow-up work. However, the central mechanistic attribution is not quantitatively established. The coefficients c and cD are measured independently, but no calculation connects the measured cD values to the observed reduction in c or to changes in the depletion geometry described by Eq. (2). The 300 pF CV proxy is also not validated against gain-layer depletion across wafers and fluences, particularly because the measurement frequency is changed after irradiation. These issues are fixable with additional analysis or a more careful framing, but they currently leave the main conclusion unsupported.","major_comments":[{"comment":"The central claim that donor removal in the n+ layer causes the lower c coefficients is not quantitatively established. Table 2 reports c from CV fits and Table 3 reports cD from Van der Pauw fits, but the paper never converts the measured cD values into a predicted change in VGL or c, and neither the depletion depth d nor the gain-layer width w in Eq. (2) is measured before and after irradiation. The qualitative ordering (W3 has the highest cD and lowest c) is consistent with the hypothesis, but wafer-to-wafer variations in the gain implant and in the CV proxy could produce the same ordering. Please add a quantitative bridge, such as a TCAD simulation or an analytic estimate using Eq. (2), showing that the measured cD values change d/w sufficiently to explain the observed 20-40% reduction in c relative to the LGAD reference values, or explicitly rephrase the conclusion as a hypothesis rather than an established attribution.","section":"Section 5 and Conclusions"},{"comment":"The extraction of c relies on a single fixed capacitance threshold, Vth = 300 pF. The manuscript does not demonstrate that this point consistently tracks gain-layer depletion rather than total sensor depletion across all wafers and fluences; indeed Section 3.2 states that c is extracted from CV measurements that 'follow the evolution of sensor depletion, not just the gain layer.' In addition, the CV frequency is changed from 1 kHz before irradiation to 10 Hz after irradiation, which can shift the measured capacitance level and therefore the inferred Vth. Please validate the proxy by showing that the 300 pF point lies in the gain-layer depletion region for representative curves before and after irradiation, and by checking the sensitivity of the fitted c values to the chosen capacitance threshold and to the measurement frequency.","section":"Section 3.1 and 3.2"},{"comment":"The wafer-to-wafer comparison is weakened by the grouping of W3 and W4. Table 2 reports a combined c value for 'W3+4', while Table 3 reports cD for W3 and W6 and W14 separately. If W3 and W4 have different gain implants or n+ layer doping, the correlation between c and cD is not evaluated on the same samples. Please report c for W3 and W4 separately, or justify the grouping with data showing that the two wafers are electrically equivalent before irradiation.","section":"Section 3 and Table 2"}],"minor_comments":[{"comment":"Please state whether the normalization parameter B was constrained to 1 or treated as a free parameter. Since Vth/VGL should equal 1 at zero fluence, a free B changes the interpretation of the extracted c values and should be documented.","section":"Equation (1)"},{"comment":"The TCT charge-sharing study currently includes only one wafer (W14), and the distributions in Fig. 6 are normalized with no uncertainties or statistical treatment. The conclusion that 'irradiation does not alter the sharing mechanism' is stronger than the data support; please present this as an encouraging preliminary result until additional wafers and quantitative metrics are available.","section":"Section 6"},{"comment":"The LGAD comparison values are taken from reference [4], a self-authored textbook, and reference [9] is an overlapping study. Independent LGAD acceptor-removal data would strengthen the comparison; at minimum, please clarify the overlap and the provenance of the typical values quoted.","section":"Section 3.2 and References"},{"comment":"The caption says 'Figure based on [9]' but the data shown appear to be new measurements from this campaign. Please clarify the relationship to reference [9].","section":"Figure 3 caption"}],"recommendation":"major_revision","confidential_remarks":"The paper reports useful and reproducible irradiation data on a detector type for which such data are scarce. The main technical issue is the missing quantitative link between the measured cD values and the observed reduction in c; this is a load-bearing gap in the central conclusion. If the authors can add a TCAD or analytic demonstration, or alternatively soften the conclusion to a clearly labeled working hypothesis, I would support acceptance. The self-citation pattern is mild and not, in my view, a reason for concern."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nShort version: this is a useful dataset from an irradiation campaign on the second FBK RSD production, and the Van der Pauw donor-removal coefficients are the genuinely new piece. The paper's central claim—that donor removal in the lightly doped n+ layer explains the lower c coefficients—is plausible but not quantitatively demonstrated. I'd send it to review, but the authors should be pushed to close the gap between cD and c.\n\nWhat's new and good: first irradiation study of this specific RSD production; first cD measurements in this doping range, with values reproduced at KIT and Perugia (Tables 3 and 5 are consistent). The trend that lower donor density gives higher cD mirrors acceptor removal and is a clean, falsifiable result. The CV-based c coefficients per wafer are also new and show a consistent ordering. The paper is also candid that the TCT charge-sharing study is preliminary, one wafer, no error bars. Credit where due: the direct measurements are reproducible and the writing is honest about limitations.\n\nSoft spots: the mechanistic conclusion leans on a proxy. The 300 pF threshold is a reasonable device, but the paper itself notes (Sec. 3.2) that CV tracks total sensor depletion, not just the gain implant, and the measurement frequency changes from 1 kHz to 10 Hz after irradiation. More importantly, the paper never connects the measured cD values to the observed c reduction. The hypothesis in Eq. (2) says donor removal shrinks the n+ layer, increases d, and compensates acceptor removal, but d is not measured and no calculation or TCAD is supplied. The qualitative ordering (W3 has highest cD, lowest c) is suggestive, but wafer-to-wafer gain-implant variations or the Vth proxy shifting with bulk damage could produce the same ordering. So the conclusion 'both results are attributed to donor removal' goes beyond what the data establish. The c and cD measurements themselves stand; the attribution is the soft underbelly.\n\nAudience: detector R&D people working on LGAD/RSD radiation hardness. The dataset is worth having in the literature. For a journal like NIM A, this is appropriate after the authors either add a quantitative link (e.g., a simple TCAD or analytic estimate converting cD into ΔVGL) or soften the causal language in the conclusions.\n\nMy recommendation: accept with serious peer review, but request the quantitative bridge or a more cautious wording. The paper is not methodologically broken; it just overreaches in interpretation.","headline":"First irradiation data for FBK RSDs with solid donor-removal measurements, but the donor-removal explanation for the lower c coefficients is plausible rather than proven.","tokens_in":6777,"tokens_out":1725,"would_cite":true,"duration_ms":15240,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Irradiating RSDs removes donors from their resistive n+ layer, compensating acceptor removal and making the gain layer appear more radiation tolerant than a standard LGAD's.","keywords":["resistive AC-coupled silicon detector","AC-LGAD","irradiation","acceptor removal","donor removal","charge sharing","transient current technique","4D tracking"],"falsifier":"Measure the gain-layer depletion voltage directly on the same irradiated sensors, for example from Transient Current Technique gain-versus-bias curves, and compare it with the $c$ values extracted from the fixed 300 pF capacitance threshold; if the two disagree as a function of fluence, the donor-removal compensation explanation loses its measured basis.","tokens_in":5777,"feed_emoji":"⚛️","tokens_out":10856,"duration_ms":89917,"temperature":0.7,"pith_summary":"This paper reports an irradiation study of Resistive AC-coupled Silicon Detectors (RSDs), the sensors proposed for 4D tracking at future hadron colliders. The authors measure how the active doping of the gain layer falls with proton and neutron fluence using capacitance-voltage curves, and they find the extracted rate coefficients $c$ are lower than the acceptor-removal coefficients of standard LGADs with similar gain implants: $2.6$–$3.2\\times10^{-16}$ cm$^2$ for neutrons and $8.1$–$9.9\\times10^{-16}$ cm$^2$ for protons. They attribute this difference to donor removal in the lightly doped $n^+$ resistive layer, a mechanism that is negligible in standard LGADs. If correct, the result identifies a compensation effect that could be engineered to make LGADs more radiation hard for hadron-collider trackers.","feed_headline":"Irradiated RSDs retain gain-layer doping longer than LGADs","feed_subtitle":"Donor removal in the resistive layer offsets acceptor removal, offering a handle for harder LGAD designs.","key_machinery":"The central object is the lightly doped $n^+$ resistive layer that distinguishes RSDs from standard LGADs. The argument runs on two measurements: the capacitance-voltage threshold $V_{th}$ at a fixed 300 pF capacitance, taken as a proxy for gain-layer depletion and fit with $f(\\Phi)=B e^{-c\\Phi}$ to obtain the apparent gain-loss coefficient $c$; and the four-point sheet conductance of the $n^+$ layer, fit with the same exponential to obtain the donor-removal coefficient $c_D$. The load-bearing identity is $V_{GL}\\propto (1+2d/w)N_A w^2$, which converts the measured shrinkage of the $n^+$ layer into the prediction that $V_{GL}$ should fall more slowly than in a standard LGAD.","core_discovery":"The paper's central claim is that irradiation removes donors from the $n^+$ resistive layer of an RSD, and that this donor removal is why the gain layer appears to lose its active doping more slowly than a standard LGAD's. In a standard LGAD, acceptor removal from the $p^+$ gain implant lowers the gain-layer depletion voltage $V_{GL}$ with fluence; in an RSD, the donor concentration of the lightly doped $n^+$ layer falls as well, changing the geometry of the multiplication region. Using the relation $V_{GL}\\propto (1+2d/w)N_A w^2$, the authors argue that as the $n^+$ layer shrinks, the depth $d$ and width $w$ of the gain region increase, partially offsetting the acceptor-removal effect on $V_{GL}$. Direct four-point sheet-conductance measurements on irradiated test structures quantify donor removal with coefficients $c_D$ of about $0.6$–$1.6\\times10^{-16}$ cm$^2$ for neutrons and $1.7$–$4.7\\times10^{-16}$ cm$^2$ for protons, with lower-doped layers showing larger $c_D$, the same trend as acceptor removal. Preliminary Transient Current Technique measurements on one wafer suggest that charge sharing between AC pads is essentially unchanged after irradiation.","pith_inferences":["A direct test of the mechanism would compare two RSD wafers with identical $p^+$ gain implants but different $n^+$ doping: donor removal predicts the lighter-doped wafer shows a larger $c_D$ and a smaller apparent $c$, whereas a pure bulk-damage explanation predicts both coefficients move together.","If the effect is real, a natural design extension is to grade or reserve extra donor doping in the $n^+$ layer so that donor removal keeps widening the multiplication region over a chosen fluence range, an optimization the paper does not attempt.","The 300 pF CV proxy could be validated independently by extracting gain-layer depletion from TCT gain-versus-bias curves on the same irradiated sensors; disagreement would force a reinterpretation of the quoted $c$ values."],"forward_implications":["If the compensation holds, RSD gain layers will retain useful gain to higher fluences than standard LGADs, strengthening the case for RSD-based 4D tracking at the HL-LHC and future colliders.","Proton irradiation remains scale-wise more damaging than neutron irradiation, so proton-dominated environments will set the radiation-hardness requirement for any RSD tracker.","Because the apparent gain-loss coefficient is lower for wafers with more resistive $n^+$ layers, the radiation response of RSDs becomes a tunable design parameter.","The preliminary TCT result that charge sharing is unchanged after irradiation, if confirmed, means a tracker would not need continuous recalibration of its position reconstruction."],"supporting_citations":[{"why":"Supplies the RSD/LGAD design context, the relation $V_{GL}\\propto(1+2d/w)N_Aw^2$, and the standard LGAD acceptor-removal coefficients used as the comparison baseline.","marker":"[4]"},{"why":"Explains the operating principle of resistive AC-coupled silicon detectors, including how the resistive $n^+$ layer controls charge sharing.","marker":"[6]"},{"why":"Identifies the second production of RSD sensors from which the irradiated devices and test structures come.","marker":"[7]"},{"why":"Documents the wafer-to-wafer differences in substrate and $n^+$ doping that the comparison across wafers W3+4, W6, and W14 relies on.","marker":"[8]"},{"why":"Provides the $V_{th}/V_{GL}$ measurement approach that the fluence-dependence plot is based on.","marker":"[9]"},{"why":"Supplies the charge-sharing reconstruction methods and the TCT protocol used for the preliminary irradiated-sensor measurements.","marker":"[10]"}],"fun_headline_variants":["RSD radiation study reveals donor removal offsetting acceptor loss","Donor removal in RSDs softens radiation damage to gain layer","RSDs resist radiation better: donor removal explains slower gain decay","Irradiated RSDs keep gain doping via donor removal, unlike LGADs","Donor removal counteracts acceptor removal in irradiated RSDs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole comparison rests on treating the voltage at a fixed 300 pF capacitance as a consistent proxy for gain-layer depletion before and after irradiation, even though the CV curves shift with bulk damage and the measurement frequency is lowered from 1 kHz to 10 Hz after irradiation.","fun_headline_variants_meta":{"raw":{"variants":["RSD radiation study reveals donor removal offsetting acceptor loss","Donor removal in RSDs softens radiation damage to gain layer","RSDs resist radiation better: donor removal explains slower gain decay","Irradiated RSDs keep gain doping via donor removal, unlike LGADs","Donor removal counteracts acceptor removal in irradiated RSDs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000716,"raw_usage":{"total_tokens":3285,"prompt_tokens":1080,"completion_tokens":2205,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":696,"completion_tokens_details":{"reasoning_tokens":2113}},"tokens_in":696,"tokens_out":2205,"duration_ms":14446,"temperature":1.0,"reasoning_tokens":2113,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T12:14:16.078781+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the gain-layer depletion voltage directly on the same irradiated sensors, for example from Transient Current Technique gain-versus-bias curves, and compare it with the $c$ values extracted from the fixed 300 pF capacitance threshold; if the two disagree as a function of fluence, the donor-removal compensation explanation loses its measured basis.","supporting_citations":[],"review_version":1}