{"id":"8918b6bd-a085-4e77-951d-0c43eda42882","arxiv_id":"2509.03770","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"SEAL 2.0 is a rebuilt reflective, visible and near-infrared adaptive optics testbed with a 98% Strehl infrared science channel and new distributed real-time control software.","lead":"SEAL is a lab testbed that mimics a large segmented telescope to develop adaptive optics for exoplanet imaging. This paper reports its 2024 rebuild to visible and infrared wavelengths, a 98% infrared Strehl ratio, and new control software.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 30 nm NCPA estimate in §3.3 rests on an unvalidated Gaussian/independence assumption; a direct IR phase retrieval would settle whether the estimate is meaningful.","rationale":"The reader correctly identifies the weakest assumption: the NCPA estimate in §3.3 relies on adding normally distributed random variables. I agree that this is the most fragile step in the central claim. However, I would phrase the consequence slightly differently: the measured 98% Strehl is a direct performance result, and the fragility lies in the inferred NCPA value and the architectural conclusion drawn from it, not necessarily in the Strehl measurement itself. The paper would be strengthened by a direct IR phase measurement or by presenting the actual phase maps rather than a variance-only reduction. The reader's CONDITIONAL verdict is appropriate: the paper presents plausible, useful instrumentation results, but the NCPA estimate and missing error bars prevent full acceptance. No reason to escalate to REJECT or downgrade to UNVERDICTED, because the direct Strehl measurement and the closed-loop residual are concrete and the rebuild is well described.","tokens_in":12730,"tokens_out":7533,"duration_ms":82637,"concrete_test":"Recover the IR-branch OPD map directly from the measured 1550 nm PSF using an independent phase-retrieval method (e.g., Gerchberg-Saxton or a focal-plane diversity algorithm). Register this map to the residual PWFS phase map obtained under the same closed-loop condition, and compute the pixel-by-pixel difference to obtain a direct NCPA map. Compare its rms and Zernike decomposition with the inferred 30 nm. If the direct rms differs by more than ~10 nm or is dominated by low-order modes (coma, astigmatism), the variance-addition model in §3.3 is invalid and the claim should be revised; if it matches, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim in §3.3 is that the measured 98% IR Strehl after closing the visible PWFS loop corresponds to ~30 nm rms NCPA between the IR PSF and the visible PWFS. The derivation, however, is not self-contained: it applies the Maréchal approximation to convert the measured Strehl to an IR OPD, separately estimates a PWFS-branch OPD, and then combines the two using 'error propagation and the principles of adding normally distributed random variables.' This implicitly assumes that the two OPDs are independent, zero-mean Gaussian random phase errors. That assumption is not justified in the text, and §3.1 actually identifies the dominant residual terms as BMC DM quilting and IrisAO segment tilts—deterministic, non-Gaussian, spatially structured errors. If the IR and PWFS OPDs are correlated or dominated by a few low-order modes, quadrature subtraction does not yield the true NCPA, and the claimed 30 nm number is unsupported. The measured 98% Strehl is a direct observable and remains valuable, but the paper's interpretation—that the visible-PWFS/IR-science architecture has only ~30 nm of NCPA—is load-bearing for the stated technological readiness. The lack of error bars on the Strehl and on the 18 nm residual further compounds the problem. This is a conditional-acceptance-level weakness, not a fundamental invalidation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports the 2024 rebuild of the Santa Cruz Extreme AO Lab (SEAL) from a refractive, visible-only testbed to a reflective, multi-wavelength bench. Thirteen custom off-axis parabolic mirrors replace the original lenses, enabling visible and near-infrared operation. The paper describes the OAP characterization and alignment, the wavefront sensor suite (Shack-Hartmann, transmissive four-sided pyramid, vector-Zernike, and a photonic lantern in the separate muirSEAL path), the visible and infrared science channels, the vector-vortex coronagraph, and the new control software (Catkit2 and CACAO). Headline performance results are a ~18 nm rms residual after closed-loop correction with the pyramid wavefront sensor, a measured 98% Strehl ratio in the infrared branch after closing the visible PWFS loop, and a corresponding estimated NCPA of ~30 nm rms between the IR PSF and the visible PWFS. The paper also describes ongoing and future work on photonic coronagraphy, photonic lantern wavefront sensing, Zernike wavefront sensing, and predictive wavefront control.","tokens_in":13044,"tokens_out":3902,"duration_ms":44191,"significance":"If the performance claims hold, the rebuilt SEAL provides a useful laboratory demonstration of a visible-PWFS/IR-science architecture, which is directly relevant to future high-contrast imaging instruments. The paper is honest about the likely sources of residual error (DM quilting and IrisAO segment tilts) and represents a significant capability expansion for the testbed. The integration of Catkit2 and CACAO is a practical contribution to the AO software ecosystem. However, the central quantitative interpretation---the ~30 nm NCPA estimate---rests on assumptions that are not validated in the manuscript, and the headline performance numbers are reported without uncertainties, repeat counts, or measurement details. These issues do not invalidate the direct observable (98% Strehl) but they do weaken the paper's central claim that the visible-PWFS/IR-science architecture has only ~30 nm of non-common path error.","major_comments":[{"comment":"The claim that the measured 98% IR Strehl 'corresponds to' a ~30 nm rms NCPA is not self-consistently derived. The text applies the Marechal approximation and then combines the IR-branch OPD and the PWFS-branch OPD via 'error propagation and the principles of adding normally distributed random variables.' This assumes independent, zero-mean, Gaussian-distributed phase errors. But Section 3.1 states that the dominant residual errors are BMC DM quilting and IrisAO segment tilts, which are deterministic, spatially structured, and likely low-order or repetitive. If the two OPDs are correlated or dominated by a few low-order modes, quadrature subtraction does not yield the true NCPA. The paper should either provide a direct IR phase retrieval measurement, or present the 30 nm number as an order-of-magnitude illustration and explicitly avoid making it a quantitative claim. If the full derivati","section":"Section 3.3, Eq. (none) and paragraph on NCPA estimate"},{"comment":"No measurement uncertainties, repeat counts, or systematic-error estimates are reported for the ~18 nm residual, the 98% Strehl ratio, or the ~30 nm NCPA. The Strehl measurement method is not described: no information is given on the PSF normalization, detector background subtraction, exposure time, source stability, or whether the AO loop was paused during the measurement. Without this information, the reader cannot assess whether 98% is meaningfully different from 95% or whether 18 nm is reproducible. The authors should provide repeated measurements, an uncertainty budget, and a clear statement of how the Strehl ratio was evaluated.","section":"Sections 3.1 and 3.3, headline performance numbers"},{"comment":"The IR branch is aligned using a visible source, and the loop is closed on the visible PWFS. The manuscript does not discuss how residual chromatic effects (e.g., the infrared lens's chromatic aberration or the beamsplitter's wavelength-dependent phase) are accounted for in the NCPA estimate. If the visible and infrared paths have different focus or aberration terms, the measured IR Strehl could be affected by effects not captured by the Gaussian quadrature model. This should be addressed explicitly, either by modeling the chromatic terms or by measuring the IR PSF directly.","section":"Section 3.3, IR branch alignment and chromatic correction"}],"minor_comments":[{"comment":"Typos: 'segmened' in the abstract/introduction and 'cornographs' in Section 2. Also, the phrase 'the testbed includes a vector-vortex coronagraph' is in the abstract but the main text correctly notes it is a visible path; consider a small clarification.","section":"Abstract and Introduction"},{"comment":"The sentence 'errors that surpassed this set tolerance were implemented into the as-built design' is unclear. Does this mean the Zemax model was re-optimized with measured parameters, or that the alignment was adjusted to compensate? Please clarify.","section":"Section 2, OAP tolerance discussion"},{"comment":"The caption reports mean latencies but no error bars or sample counts. Given that the comparison with CACAO/ImageStreamIO is a claimed improvement, include at least the standard deviation or percentile spread. Also state whether the benchmark conditions are representative of the actual closed-loop AO usage.","section":"Figure 6, Catkit2 latency benchmark"},{"comment":"Several references are to 'these proceedings' or to a senior thesis. For a journal-style paper, please ensure that all such references are either fully citable or that the relevant content is summarized in the main text.","section":"References [36]-[38] and [26]"}],"recommendation":"major_revision","confidential_remarks":"This is a testbed status and performance paper with a strong practical component. The main reason for major revision is the unsupported NCPA estimate and the absence of error bars on the headline numbers. The direct 98% Strehl measurement is valuable and should be preserved, but the paper should either substantiate the 30 nm NCPA with a direct measurement or downgrade the claim. The many 'in these proceedings' references may also be a concern for an archival journal; the authors should make the manuscript self-contained enough to be evaluated independently."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is a straightforward, honest instrumentation paper about rebuilding the SEAL testbed with reflective optics. The genuinely new bits are: the all-OAP rebuild, the infrared science channel with a measured 98% Strehl at 1550 nm using a visible PWFS, and the Catkit2 multi-node message broker with sub-microsecond latencies. Those are real results, not simulations dressed up as predictions. The paper does what an SPIE testbed paper should do: it describes the hardware, shows what was measured, and names the likely residual error sources (BMC quilting, IrisAO segment tilts). The 18 nm residual and the IR Strehl are direct bench measurements, so there is no fitted-model circularity in the main claims. Credit also for shipping code links for Catkit2 and CACAO and for citing the prior SEAL work rather than burying it.\n\nThe soft spots are real but not fatal. First, there are no error bars or repeat counts on the headline numbers. I do not doubt that the bench reached ~18 nm and 98% Strehl, but I would like to see whether these are single frames, averages, or best-of-run. Second, the 30 nm NCPA estimate in Section 3.3 rests on the Marechal approximation plus an unstated assumption that the IR and PWFS OPD errors are zero-mean, independent Gaussians. The stress-test note is right that Section 3.1 identifies deterministic, structured residuals (quilting, segment tilts), which do not obviously satisfy that assumption. The text says \"error propagation and the principles of adding normally distributed random variables\" without justifying the model. That is a genuine gap: the 98% Strehl is a solid observable, but the interpretation that NCPA is only ~30 nm is load-bearing for the architecture's readiness, and a direct IR phase retrieval would settle it. This is conditional-acceptance-level weakness, not invalidation.\n\nThe paper is not trying to be a physics discovery. It is a group updating the community on a testbed, and it does that well. The reader who gets value is anyone working on AO testbeds, segmented telescope co-phasing, or pyramid/ZWFS control for Keck, ELTs, or HWO. I would send this to peer review with a request for uncertainty quantification and a cleaner NCPA story, but I would not desk-reject it. If I were working on AO testbeds, I'd cite it and bring it to a reading group as a current-state reference.\n\nRecommendation: conditional accept after minor revision, with the NCPA derivation and error bars as the referee's main asks.\n\nBest,\n[You]","headline":"Solid instrumentation rebuild paper with real bench numbers; the 30 nm NCPA estimate needs more support, but the core claims are plausible and worth refereeing.","tokens_in":13614,"tokens_out":981,"would_cite":true,"duration_ms":12391,"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":"A rebuilt all-reflective AO testbed reaches 98% Strehl in its infrared science path while sensing wavefronts in visible light.","keywords":["adaptive optics testbed","off-axis parabolic mirrors","pyramid wavefront sensor","non-common path aberrations","infrared science path","high-contrast imaging","coronagraphy","real-time control"],"falsifier":"Point an infrared-capable wavefront sensor, or use phase diversity on the 1550 nm point-spread function, while the visible pyramid wavefront sensor loop is closed, and reconstruct the infrared wavefront directly. If the directly measured infrared NCPA is substantially above about 30 nm rms, or if the independently measured infrared Strehl is below 98%, then the Gaussian variance model underlying the estimate is wrong. Repeating the measurement after intentionally misaligning one off-axis parabolic mirror by a known amount would also reveal whether the error budget is dominated by coherent low-","tokens_in":1811,"feed_emoji":"🔭","tokens_out":2381,"duration_ms":98516,"temperature":0.7,"pith_summary":"The paper reports on the rebuilt SEAL adaptive-optics testbed, which replaced its refractive lenses with an all-reflective relay of thirteen off-axis parabolic mirrors so that the same bench can feed visible wavefront sensors, coronagraphs, and a newly added infrared science camera. The central performance claim is that, with the loop closed on a visible non-modulated pyramid wavefront sensor, the infrared branch reaches a measured Strehl ratio of 98% at 1550 nm, implying an estimated non-common-path error of about 30 nm rms between the infrared point-spread function and the visible sensor. If correct, the result validates the dual-wavelength architecture planned for high-contrast exoplanet instruments on large segmented ground-based telescopes: wavefront sensing in the visible, where sensors are sensitive, and science imaging in the infrared, where the corrected image is delivered. The paper also documents additional wavefront sensor arms, coronagraphs, and two new real-time control software packages, but the load-bearing measured result is the 98% infrared Strehl.","feed_headline":"Visible sensor drives IR path to 98% Strehl on rebuilt AO testbed","feed_subtitle":"An all-mirror rebuild lets a visible wavefront sensor correct a 1550 nm science path to near-diffraction-limited quality.","key_machinery":"The load-bearing element is the all-reflective optical relay built from thirteen custom off-axis parabolic mirrors, which replaces the original refractive lenses and relays pupil planes among a segmented primary-mirror simulator, two deformable mirrors, the wavefront sensor arms, and the coronagraph and science cameras. Because every fold is reflective, the relay introduces no chromatic aberration, so a wavefront measured with a visible pyramid sensor at 635 nm can be used to correct a science image at 1550 nm. The measured 98% Strehl and 30 nm rms non-common-path estimate is carried by this relay together with the pyramid wavefront sensor's closed-loop reconstruction (a non-linear Gerchberg","core_discovery":"After replacing the refractive optics with 13 custom off-axis parabolic mirrors, the SEAL testbed operates at both visible (635 nm) and near-infrared (1550 nm) wavelengths. The authors report a measured Strehl ratio of 98% in the infrared science branch after closing the adaptive-optics loop on the visible non-modulated pyramid wavefront sensor, with the residual alignment and manufacturing errors of the off-axis mirrors corrected by the deformable mirrors. Using the Strehl-to-wavefront-error approximation and adding the estimated optical path differences of the IR and pyramid-wavefront-sensor branches as normally distributed random variables, they estimate the non-common-path aberration bet","pith_inferences":["The 30 nm non-common-path estimate likely rests on optimistic statistics: if residual off-axis-mirror misalignments produce low-order aberrations such as coma or astigmatism, those errors would add coherently rather than as independent Gaussian variables, so a direct infrared wavefront measurement could find a larger true NCPA.","If the 30 nm value holds, then the bench has headroom to push contrast further by adding an infrared pupil-plane calibrator, since the relay is not the limiting error source.","The same visible-to-infrared correction architecture could eventually be replaced by a photonic-lantern wavefront sensor operating directly at the science wavelength, which would remove the non-common-path problem entirely; the paper's separate infrared photonic-lantern testbed is a natural place to test that idea.","The fact that millimeter-scale manufacturing deviations in the off-axis mirrors were corrected by alignment in the presence of deformable mirrors suggests that relaxed mirror tolerances may be acceptable for similar future testbeds, reducing cost."],"forward_implications":["If correct, a visible pyramid wavefront sensor can serve as the wavefront reference for a separately aligned infrared science path with only tens of nanometers of uncorrected non-common-path error.","The all-reflective bench can validate wavefront-sensing and coronagraphic technologies for future segmented-telescope high-contrast instruments in the same visible-sensing, infrared-science configuration they would use on sky.","The measured performance sets a baseline against which future infrared-branch upgrades, such as a broadband white-light source and a dichroic beamsplitter, can be judged.","With three real-time control packages and multiple wavefront sensor arms on one platform, the testbed allows direct comparison of sensing and control approaches on identical hardware.","The reported Strehl suggests that the current bench limit is set by deformable-mirror surface quilting, segmented-mirror residual tilts, and non-common-path aberrations rather than by the off-axis-parabola relay itself."],"supporting_citations":[{"why":"Defines the original SEAL testbed architecture and first-light components that the reflective rebuild extends.","marker":"[1]"},{"why":"Supplies the real-time controller design on which the pyramid-wavefront-sensor loop operation is based.","marker":"[2]"},{"why":"Provides the Gerchberg–Saxton reconstruction used to turn non-modulated pyramid wavefront sensor signals into phase commands for closed-loop correction.","marker":"[8]"},{"why":"Gives the four-sided double-rooftop pyramid wavefront sensor design adopted in the rebuilt bench.","marker":"[24]"},{"why":"Provides the measured infrared Strehl ratio and the non-common-path error estimate between the infrared branch and the visible pyramid wavefront sensor.","marker":"[26]"}],"fun_headline_variants":["Reflective rebuild lets visible sensor drive IR path to 98% Strehl","All-mirror AO testbed hits 98% Strehl in infrared via visible sensing","Visible wavefront sensor corrects IR to 98% Strehl on mirrored testbed","SEAL 2.0: Reflective optics enable 98% Strehl IR correction from visible sensor","Mirror-based testbed achieves 98% infrared Strehl with visible-only sensing"],"cache_read_input_tokens":15232,"weakest_assumption_plain":"The 30 nm non-common-path estimate rests on assuming that the wavefront errors separating the infrared science path and the visible wavefront sensor add in quadrature as independent, zero-mean Gaussian fluctuations; if those errors are concentrated in a few low-order modes, the true infrared wavefront error is larger than 30 nm and the 98% Strehl overstates the infrared image quality.","fun_headline_variants_meta":{"raw":{"variants":["Reflective rebuild lets visible sensor drive IR path to 98% Strehl","All-mirror AO testbed hits 98% Strehl in infrared via visible sensing","Visible wavefront sensor corrects IR to 98% Strehl on mirrored testbed","SEAL 2.0: Reflective optics enable 98% Strehl IR correction from visible sensor","Mirror-based testbed achieves 98% infrared Strehl with visible-only sensing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000895,"raw_usage":{"total_tokens":3705,"prompt_tokens":768,"completion_tokens":2937,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":512,"completion_tokens_details":{"reasoning_tokens":2824}},"tokens_in":512,"tokens_out":2937,"duration_ms":20744,"temperature":1.0,"reasoning_tokens":2824,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T10:40:13.364854+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Point an infrared-capable wavefront sensor, or use phase diversity on the 1550 nm point-spread function, while the visible pyramid wavefront sensor loop is closed, and reconstruct the infrared wavefront directly. If the directly measured infrared NCPA is substantially above about 30 nm rms, or if the independently measured infrared Strehl is below 98%, then the Gaussian variance model underlying the estimate is wrong. Repeating the measurement after intentionally misaligning one off-axis parabolic mirror by a known amount would also reveal whether the error budget is dominated by coherent low-","supporting_citations":[{"cited_title":"The Santa Cruz Extreme AO Lab (SEAL): design and first light,","cited_arxiv_id":null,"evidence_quote":"Defines the original SEAL testbed architecture and first-light components that the reflective rebuild extends."},{"cited_title":"A near-infrared pyramid wavefront sensor for Keck adaptive optics: real-time controller,","cited_arxiv_id":null,"evidence_quote":"Supplies the real-time controller design on which the pyramid-wavefront-sensor loop operation is based."},{"cited_title":"Using the Gerchberg-Saxton algorithm to reconstruct nonmodulated pyramid wavefront sensor measurements,","cited_arxiv_id":null,"evidence_quote":"Provides the Gerchberg–Saxton reconstruction used to turn non-modulated pyramid wavefront sensor signals into phase commands for closed-loop correction."},{"cited_title":"AO3000 at Subaru: combining for the first time a NIR WFS using First Light’s C-RED ONE and ALPAO’s 64x64 DM,","cited_arxiv_id":null,"evidence_quote":"Gives the four-sided double-rooftop pyramid wavefront sensor design adopted in the rebuilt bench."},{"cited_title":"Implementation of a Visible Vector Vortex Coronagraph and Imaging Infrared branch on the SEAL Testbed,","cited_arxiv_id":null,"evidence_quote":"Provides the measured infrared Strehl ratio and the non-common-path error estimate between the infrared branch and the visible pyramid wavefront sensor."}],"review_version":1}