{"id":"2ff0fb49-6fc5-4154-aab7-a96468154153","arxiv_id":"1908.11429","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Simulated H4RG detector noise sources translate to radial velocity errors of 0.32 to 3.5 m/s in quadrature, suggesting sub-m/s precision is possible if persistence and nonlinearity are controlled.","lead":"Using simulations, the authors estimate how noise from H4RG near-infrared detectors affects the precision of radial velocity measurements of stars. They find that with careful mitigation of persistence and nonlinearity, such detectors could support sub-meter-per-second Doppler precision.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Table 1's optimistic 0.32 m/s is a point estimate from a single H2RG-based read-noise realization, with no repeated-realization or H4RG-validation statistics; the sub-m/s 'could' claim is therefore not yet pinned down.","rationale":"The central claim is deliberately conditional: H4RG-based spectrographs could reach sub-m/s precision if detector noise dominates the error budget. The quantitative support is Table 1, and the optimistic 0.32 m/s quadrature is the specific evidence that makes the claim credible. The largest optimistic term is read noise, and §4.1 derives it from a single simulated frame generated by software whose PCA basis is H2RG-based (§3.1), with only settings adjusted for H4RG. The 15% IRS2 reduction is also a private-communication scalar applied outside the noise generator, since the generator does not model IRS2. A modest change in read-noise amplitude would not by itself break the sub-m/s conclusion—read noise would need to approach 1 m/s to do that—so the H2RG-versus-H4RG provenance issue is not automatically decisive. The more load-bearing problem is that no repeated-realization statistics or H4RG validation are provided anywhere in the paper; every reported RV error is a point value. If those point values are favorable draws, the optimistic column could overstate the expected performance. This concern does not overturn the reader's CONDITIONAL verdict; it sharpens the reason for the condition: the quantitative claims need uncertainty quantification and laboratory validation before being used in instrument requirements.","tokens_in":12664,"tokens_out":8671,"duration_ms":90678,"concrete_test":"Run the §2 nominal iLocater simulation 50–100 times, each with an independent residual read-noise draw from the NG, and report the mean, standard deviation, and 95% interval of the recovered σ_RV instead of a single value; then repeat with the zeroth principal component replaced by a PCA basis computed from an actual H4RG dark/up-the-ramp stack (e.g., SPIRou or WFIRST lab data) processed with the same reference-pixel subtraction. If the read-noise term's 95% interval or the H4RG-based mean exceeds about 0.5 m/s, Table 1's optimistic quadrature and the sub-m/s conclusion are not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline conclusion rests on the optimistic column of Table 1, whose largest term is read noise: 0.31 m/s after applying a 15% IRS2 discount to the 0.37 m/s result of §4.1. That result is obtained from one simulated residual read-noise frame from the HxRG Noise Generator, and §3.1 states that the generator's PCA basis is built from NIRSpec H2RGs, with H4RG represented only by adjusted settings. No error bars or independent-realization statistics are reported for any entry in Table 1. Since the RV impact of read noise depends on the spatial correlation pattern across spectral traces, not merely on the 5.4 e- rms, an H2RG-based basis could change the dominant optimistic term. A single favorable realization is insufficient to establish that sub-m/s operation is an attainable expectation rather than a chance outcome of one simulation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents an end-to-end simulation study of the impact of H4RG near-infrared detector noise on the radial velocity (RV) precision of a high-resolution echelle spectrograph, using the iLocater instrument model as a representative case. Five detector effects are treated one at a time: read noise (via the HxRG Noise Generator built on PCA of NIRSpec H2RG data), pixel nonlinearity (using Artigau's cubic mapping), residual dark current (Poisson draws at literature rates), persistence (fractional parasitic signals from M-star, G-star, and Fabry-Perot sources at various RV offsets), and interpixel capacitance (convolution with 3x3 and 9x9 kernels). For each effect, an RV error is recovered with the iLocater data-reduction pipeline and masked cross-correlation. The results are summarized in Table 1 as optimistic and pessimistic estimates, with quadrature sums of 0.32 m/s and 3.5 m/s, respectively. The authors conclude that sub-meter-per-second RV precision is attainable with H4RG arrays if detector noise dominates the error budget and persistence and nonlinearity are mitigated.","tokens_in":12828,"tokens_out":13977,"duration_ms":121702,"significance":"Assuming the quantitative results hold, this paper addresses a real gap: it connects specific H4RG detector characteristics to Doppler RV error in a form usable for instrument error budgets, and it has already informed the iLocater design reviews. The forward-simulation approach is a strength: noise coefficients, dark currents, persistence fractions, and IPC kernels are taken from the published literature rather than fit to the target RV errors, so Table 1 constitutes a set of falsifiable laboratory predictions. The paper is also transparent about its scope, explicitly listing omitted effects such as random telegraph noise, intra-pixel QE variations, and the brighter-fatter effect, and identifying laboratory validation as the next step. The main significance is as a design-stage error budget method rather than as a final performance demonstration; the optimistic column in particular should be read as an expectation pending H4RG-specific read-noise validation.","major_comments":[{"comment":"The read-noise term, which dominates the optimistic quadrature sum in Table 1 (0.31 of the 0.32 m/s total), is computed from a single simulated residual read-noise frame produced by the HxRG Noise Generator. Section 3.1 states that the generator's PCA basis is built from NIRSpec H2RG data, with H4RG represented only by adjusted settings, and the RV impact of read noise depends on the spatial correlation structure of the residual frame across spectral traces, not merely on its 5.4 e− rms. The manuscript reports no repeated-realization statistics for this or any other stochastic entry in Table 1, and no validation that the H2RG-based correlation structure is representative of H4RG readout. I ask for either multiple noise realizations with reported scatter, or an explicit sensitivity test showing that plausible changes in the residual noise correlation structure do not move the optimistic read-noise term above the sub-m/s budget.","section":"§3.1, §4.1, Table 1"},{"comment":"The optimistic read-noise value of 0.31 m/s is obtained by scaling the simulated 0.37 m/s result by a 15% IRS2 noise-reduction estimate taken from a private communication. The paper itself notes that IRS2 also reduces correlated noise and 1/f banding, which would affect the RV error in a way that is not purely proportional to total read noise. Because this scaling sets the largest term of the optimistic column, the manuscript should either give the 15% figure a stated uncertainty and show the sensitivity of the quadrature sum to it, or model the IRS2/interleaved readout mode in the noise generator so the claim does not rest on an unverifiable input.","section":"§4.1, Table 1 note [1]"},{"comment":"The optimistic linearity entry (0.8 cm/s = 0.008 m/s) is asserted by applying SPIRou's measured 0.3% residual linearity to the simulated 30-minute test case, but the mapping from residual nonlinearity fraction to RV error is not shown in the text; the reader cannot reproduce the 0.8 cm/s value from Figure 5 or from equation (1). Please make the scaling explicit (for example, RV error approximately equal to the residual linearity fraction times the uncorrected nonlinearity error), or provide the derivation. This is the one Table 1 entry whose value is assigned by a scaling statement rather than by the simulation pipeline described in Section 2.","section":"§4.2, Table 1 note [2]"},{"comment":"The headline range of 0.32–3.5 m/s is presented as a quadrature sum of detector effects, but the spread between the columns is dominated by assumptions about observing conditions and mitigation (persistence spans 0.003–2.1 m/s; linearity 0.008–2.6 m/s) rather than by detector physics alone. As written, a reader cannot tell which terms represent a detector-noise floor and which represent a scenario choice. I recommend an explicit decomposition stating that, after persistence and linearity mitigation, the detector-limited quadrature sum is set by read noise, dark current, and IPC, and reporting that sub-m/s subset separately from the scenario-dependent terms.","section":"§5, Table 1"}],"minor_comments":[{"comment":"In-text references to 'Figure 3.2', 'Figure 4.1', and 'Figure 4.2' do not match the caption numbering (Figures 3, 4, and 5 in the List of Figures); the linearity mapping is Figure 3, the read-noise result is Figure 4, and the nonlinearity result is Figure 5, so the cross-references should be corrected.","section":"§3.2, §4.1, §4.2"},{"comment":"The text states that persistence results 'are shown in the vertical panels of Figure 7', but the figure caption describes top, middle, and bottom rows, i.e., horizontal panels; the wording should be changed.","section":"§4.4"},{"comment":"The phrase 'some elements along the electronic patch' appears to be a typo for 'electronic path'.","section":"§3.2"},{"comment":"The sentence 'Examples of dark current frames in HxRGs are can be found in several detector studies' contains a duplicated verb ('are can be'); remove 'are'.","section":"§3.3"},{"comment":"The sentence 'experimentally measured kernels do not exactly follow the mathematical α model exactly' repeats 'exactly'; one instance should be removed.","section":"§4.5"},{"comment":"The methodology states a median S/N of 187 per collapsed pixel, while the colorbar of Figure 4 spans median S/N values of 2000–10000; the relationship between these two quantities (per-pixel versus collapsed, single order versus full spectrum) should be clarified.","section":"§2, Fig 4"}],"recommendation":"major_revision","confidential_remarks":"To the editor: the manuscript is within scope for JATIS as an instrumentation/detector assessment. The central forward-simulation claims are defensible, but the optimistic column of Table 1 rests on a single realization of an H2RG-derived noise generator and a private-communication 15% IRS2 scaling, so I recommend that the revision address these before acceptance. The paper cites the iLocater simulator papers appropriately given that it reuses that code, and the external literature (Artigau, Rauscher, Beletic, Kannawadi) is cited for the adopted noise parameters, so I see no citation-pattern problem. One archival concern: a load-bearing number (the IRS2 improvement) is sourced to a private communication; the editor may wish to ask the authors whether Rauscher's published IRS2 paper (Ref. 18) contains a usable estimate instead."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis paper is worth your time if you work on NIR Doppler spectrographs. It does something genuinely useful: it translates H4RG detector noise terms into RV error using an end-to-end spectrograph simulator plus the Rauscher noise generator, and gives a clean optimistic/pessimistic summary in Table 1. Read noise, nonlinearity, dark current, persistence, and IPC are all treated, and the authors are honest about what they left out (random telegraph noise, intra-pixel QE variations, brighter-fatter).\n\nThe main soft spot is exactly what the stress-test note flags. The optimistic sub-m/s conclusion hinges on the read noise entry, 0.31 m/s, which comes from a single simulated residual read-noise frame, scaled down 15% for IRS2. The noise generator is PCA-based on NIRSpec H2RGs; H4RG is represented only by settings, not by H4RG data. No repeated realizations are reported, so there are no error bars on any Table 1 entry. That does not make the paper wrong, but it means the quantitative floor is not pinned down. The linearity optimistic value (0.008 m/s) is also asserted by scaling SPIRou's residual nonlinearity rather than derived from their own simulated mapping. The pessimistic columns are probably robust, and the persistence parameter study is thorough, so the overall range is still informative.\n\nThe citation pattern is fine. The paper leans on Rauscher's noise generator and Artigau's H4RG characterization, which is appropriate. It is a forward simulation using literature noise values, not a circular fit to the desired answer.\n\nMy recommendation: yes, send it to peer review. A good referee will ask for error bars on Table 1 and preferably a lab-validation test, but the paper deserves the attention. I would cite it for instrument requirements work and I would bring it to a reading group focused on precision RV instrumentation.","headline":"A useful first simulation-based error budget linking H4RG detector noise to RV precision, but the sub-m/s optimistic numbers rest on a single H2RG-derived read-noise realization and need error bars and lab validation before being used as requirements.","tokens_in":13395,"tokens_out":2070,"would_cite":true,"duration_ms":21838,"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":"H4RG near-infrared detectors can support sub-meter-per-second radial-velocity precision, provided persistence and nonlinearity are controlled.","keywords":["radial velocity","H4RG detectors","near-infrared spectroscopy","detector noise","error budget","interpixel capacitance","persistence","Doppler spectrographs"],"falsifier":"Take a science-grade H4RG-10, operate it at its nominal temperature, acquire dark frames with reference-pixel subtraction and up-the-ramp sampling, and measure the residual noise rms and its principal-component amplitudes. If the measured rms deviates from the 5.4 e- emulated by the noise generator, or if the spatial correlation of the noise differs, the 37 cm/s read-noise contribution and the quadrature budget would need revision.","tokens_in":12439,"feed_emoji":"🔭","tokens_out":6462,"duration_ms":55204,"temperature":0.7,"pith_summary":"This paper asks whether the best commercially available near-infrared detectors, H4RG arrays, are good enough for the next generation of precise Doppler radial-velocity spectrographs. Using an end-to-end spectrograph simulation, it quantifies five detector noise sources—read noise, pixel nonlinearity, dark current, persistence, and interpixel capacitance—in units of velocity error. The central result is a detector-noise error budget between 0.32 and 3.5 m/s depending on optimistic or pessimistic assumptions, with read noise alone contributing 37 cm/s for a 30-minute exposure. The authors conclude that sub-meter-per-second near-infrared precision is achievable with HxRG arrays if persistence and nonlinearity are mitigated and detector noise dominates the error budget.","feed_headline":"H4RG infrared detectors can hit sub-meter-per-second Doppler precision","feed_subtitle":"A five-source noise budget of 0.32 to 3.5 m/s leaves room for sub-meter-per-second NIR velocimetry.","key_machinery":"The central mechanism is the pairing of an end-to-end echelle spectrograph simulator with the HxRG Noise Generator, a tool that emulates residual read noise as a sum of five principal-component patterns: white noise, correlated pink noise, uncorrelated pink noise, alternating column noise, and picture frame noise. Each detector effect is injected into simulated spectral frames and converted to a velocity error by extracting spectra and performing masked cross-correlation. Supporting models map pixel nonlinearity through a cubic correction term, dark current through Poisson residuals after mean subtraction, persistence through a fractional image that decays as 1/t, and interpixel capacitance through convolution with a coupling kernel.","core_discovery":"The central claim is that H4RG detectors do not inherently preclude sub-meter-per-second radial-velocity measurements. For an M0V, I = 10 star observed with iLocater-like parameters, the simulations yield a detector-noise error budget whose optimistic and pessimistic quadrature sums are 0.32 and 3.5 m/s. Read noise alone gives 37 cm/s at 30-minute integration, or 32 cm/s with an expected 15% IRS2 improvement; linearity ranges from 0.8 cm/s after correction to 2.6 m/s uncorrected; residual dark current is below 20 cm/s at temperatures at or below 100 K; persistence can reach 2.1 m/s under a 0.1% identical-spectrum remnant; and interpixel capacitance ranges from 5 to 87 cm/s depending on the kernel. The paper therefore concludes that precision radial-velocity spectrographs could reach sub-meter-per-second precision in the near infrared if HxRG arrays are used and errors are dominated by detector noise.","pith_inferences":["The noise generator's principal-component basis comes from H2RG data; if H4RG residual noise has different spatial correlations, the 37 cm/s read-noise figure could shift, so a laboratory comparison on an actual H4RG would settle this.","The persistence results suggest that alternating science and etalon calibration frames is risky for sub-meter-per-second instruments, and a dedicated calibration fiber or persistence-model subtraction may be required.","Because several detector effects produce opposite-sign velocity shifts, a full end-to-end simulation with all effects turned on simultaneously could yield a total error lower than the quadrature sum.","If IRS2 interleaved readout reduces correlated 1/f noise as well as total noise, the read-noise term could fall below 32 cm/s, strengthening the case for sub-meter-per-second performance."],"forward_implications":["Read noise contributes 37 cm/s at a 30-minute exposure, and a 15% IRS2 readout improvement lowers this to 32 cm/s, leaving room for other terms in a sub-meter-per-second budget.","Uncorrected pixel nonlinearity can produce multi-meter-per-second errors, but with SPIRou-level correction it drops to 0.8 cm/s, making linearity correction a prerequisite.","Persistence is a dominant risk: a 0.1% remnant of an identical spectral type at a 3.5 km/s offset yields 2.1 m/s error, so observing schedules and calibration exposures must be managed.","Interpixel capacitance contributes anywhere from 5 to 87 cm/s depending on the kernel, meaning detector choice and IPC characterization matter.","The quadrature sum of all detector effects spans 0.32 to 3.5 m/s, and since individual effects partially cancel, this sum is likely an upper limit on the detector-induced velocity error."],"supporting_citations":[{"why":"Supplies the end-to-end spectrograph simulator and data reduction pipeline used to turn detector noise into radial-velocity errors.","marker":"[13]"},{"why":"Supplies the PCA-based HxRG Noise Generator that produces residual read-noise frames.","marker":"[14]"},{"why":"Provides the principal-component decomposition of NIRSpec detector noise on which the noise generator is built.","marker":"[17]"},{"why":"Supplies the nonlinearity mapping, persistence model, and SPIRou IPC and dark-current values used in the simulations.","marker":"[23]"},{"why":"Sets the representative instrument parameters and the 30 cm/s detector allocation used as a comparison target.","marker":"[15]"},{"why":"Provides the expected 15% read-noise reduction from improved reference sampling and subtraction.","marker":"[18]"},{"why":"Supplies the H2RG dark-current versus temperature data used to simulate residual dark-current noise.","marker":"[29]"},{"why":"Provides the interpixel-capacitance kernel model and the alpha values appropriate for H4RG detectors.","marker":"[33]"}],"fun_headline_variants":["H4RG detectors clear path to sub-m/s radial velocities","Sub-meter-per-second RV feasible with H4RG near-infrared","H4RG noise budget: 0.32–3.5 m/s, enabling sub-m/s","Near-infrared H4RG arrays don't block precise Doppler","Simulations show H4RG detectors suit sub-m/s RV goals"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The read-noise simulations assume that the PCA-based HxRG Noise Generator, trained on H2RG data, faithfully reproduces the residual noise of an H4RG after reference-pixel subtraction and up-the-ramp sampling.","fun_headline_variants_meta":{"raw":{"variants":["H4RG detectors clear path to sub-m/s radial velocities","Sub-meter-per-second RV feasible with H4RG near-infrared","H4RG noise budget: 0.32–3.5 m/s, enabling sub-m/s","Near-infrared H4RG arrays don't block precise Doppler","Simulations show H4RG detectors suit sub-m/s RV goals"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000617,"raw_usage":{"total_tokens":2881,"prompt_tokens":981,"completion_tokens":1900,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":597,"completion_tokens_details":{"reasoning_tokens":1805}},"tokens_in":597,"tokens_out":1900,"duration_ms":12845,"temperature":1.0,"reasoning_tokens":1805,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:15:16.097557+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a science-grade H4RG-10, operate it at its nominal temperature, acquire dark frames with reference-pixel subtraction and up-the-ramp sampling, and measure the residual noise rms and its principal-component amplitudes. If the measured rms deviates from the 5.4 e- emulated by the noise generator, or if the spatial correlation of the noise differs, the 37 cm/s read-noise contribution and the quadrature budget would need revision.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the end-to-end spectrograph simulator and data reduction pipeline used to turn detector noise into radial-velocity errors."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the PCA-based HxRG Noise Generator that produces residual read-noise frames."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the principal-component decomposition of NIRSpec detector noise on which the noise generator is built."},{"cited_title":"Artigau , J","cited_arxiv_id":null,"evidence_quote":"Supplies the nonlinearity mapping, persistence model, and SPIRou IPC and dark-current values used in the simulations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Sets the representative instrument parameters and the 30 cm/s detector allocation used as a comparison target."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the expected 15% read-noise reduction from improved reference sampling and subtraction."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the H2RG dark-current versus temperature data used to simulate residual dark-current noise."},{"cited_title":"Kannawadi , C","cited_arxiv_id":null,"evidence_quote":"Provides the interpixel-capacitance kernel model and the alpha values appropriate for H4RG detectors."}],"review_version":1}