{"id":"b10f2fdc-8748-49cd-80e1-e583e6ad4c21","arxiv_id":"1908.05662","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"PISCO generates synthetic CMB time-ordered data by convolving a polarized sky with a fully polarized, arbitrary telescope beam model using GPU acceleration.","lead":"This paper introduces PISCO, a GPU-accelerated software tool that simulates what a cosmic microwave background telescope would measure, including the full polarization state of its beam. It is useful for testing how telescope imperfections and scanning patterns distort data in experiments searching for signs of inflation.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Generality to non-Gaussian beams is asserted but untested; every quantitative closure test uses smooth Gaussian-family beams, so the arbitrary-beam claim rests on unverified pixel-sampling assumptions.","rationale":"The reader correctly identifies that the versatility claim for arbitrary beams is validated only with smooth Gaussian beams, and I agree that this is the main soft spot. The reader's phrasing that 'all validation tests use circular Gaussian beams' is slightly too strong: Section 7 uses elliptical Gaussian beams for the CLASS simulations, which tests a broader but still smooth Gaussian family. The load-bearing concern is not that the pixel-space algorithm is wrong for what it demonstrates; the point-source, full-sky CMB, and CLASS systematics tests provide genuine support for the implementation. The gap is that the step from 'works well for smooth Gaussian beams' to 'arbitrary beams, sky models and scanning strategies' requires the pixel-center sampling and interpolation of Equation 4.5 to remain accurate for beams with sharp or extended structure, and no test or error analysis establishes that. The proposed concrete test would settle this by pushing the code into the advertised regime and comparing against an independent method. I do not see grounds to reject the paper, and the reader's CONDITIONAL verdict is appropriate; my read does not change it.","tokens_in":21300,"tokens_out":4181,"duration_ms":49208,"concrete_test":"Run the Section 6.2 full-sky closure test using a non-Gaussian beamsor, e.g., a 1.5-degree Airy pattern with a first sidelobe near 3-4 degrees and a sidelobe floor extending to 10 degrees, pixelated at NSIDE >= 2048. Compare the power spectra of maps made from PISCO TOD against spectra obtained from an independent harmonic-space convolution code (such as beamconv) using the same sky model and scanning strategy, after applying the analytic beam transfer function and pixel window. Require residuals comparable to the Gaussian case; as a second check, repeat with a 10-degree truncation radius to verify that the disc edge does not bias the result. If the residuals are substantially larger than in the Gaussian case, the 'arbitrary beams' claim is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim (Section 8) is that PISCO generates beam-convolved timestreams for arbitrary beams, sky models and scanning strategies. The load-bearing mathematical step is Equation 4.5, which replaces the continuous Mueller-formalism integral of Equation 4.3 with a discrete sum that evaluates the beamsor at sky-pixel centers after spherical-trigonometric re-pixelization (Appendix A). No sub-pixel integration or error bound is given for this approximation. For smooth Gaussian beams the error is demonstrably small: flux is preserved to 0.2% in Section 6.1 and full-sky spectra close to input in Section 6.2. But 'arbitrary beams' includes sharply varying sidelobes, non-symmetric polarized responses, and beams with significant support beyond the chosen truncation disc; for these, pixel-center sampling and beamsor interpolation can alias power, and the 5-degree cutoff justified in Section 7.1.3 only applies to a 1.5-degree Gaussian. No validation test probes this regime: Section 6 uses circular Gaussians and Section 7 uses elliptical Gaussians, still a smooth two-parameter family. The CLASS systematic tests in Section 7 are also qualitative demonstrations of leakage rather than quantitative comparisons against an independent reference simulator, so they do not close the loop for non-Gaussian beams. In addition, the Introduction and abstract promise time-varying beams, yet the described implementation accepts a beamsor as a static input and no time-indexed beamsor mechanism is specified. The demonstrated physics is sound, but the advertised generality to arbitrary beams is an extrapolation beyond the evidence.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper describes PISCO, a GPU-accelerated code that generates synthetic time-ordered data for CMB experiments by convolving a fully polarized sky model with a detector beamsor (a field of Mueller matrices) in pixel space, following the formalism of O'Dea et al. (2007). The algorithm evaluates, for each pointing, a discrete sum (Eq. 4.5) over HEALPix sky pixels after re-pixelizing the beamsor via spherical-trigonometric interpolation (Appendix A). The paper presents the CUDA implementation and benchmarks, validates the code against healpy.smoothing for point sources and against analytic beam-window-corrected input spectra for full-sky simulated CMB maps, and applies PISCO to a realistic CLASS-like simulation with elliptical Gaussian beams, reproducing T-to-P leakage from pointing and beam mismatch and P-to-P leakage from uneven intra-pixel coverage.","tokens_in":21529,"tokens_out":4675,"duration_ms":50289,"significance":"If the full claims hold, PISCO would be a valuable complement to existing harmonic-space convolution codes such as beamconv and FEBeCoP, providing a fully polarized pixel-space convolution engine that can, in principle, handle arbitrary beams and scanning strategies. The validation that is present is meaningful: the point-source test is checked against an independent external routine with flux preserved to about 0.2%; the full-sky CMB test is compared with an analytic beam window function and shows small residuals in TT, EE, and BB; and the CLASS tests reproduce physically motivated leakage signals without fitted parameters. The main limitation is that every quantitative test uses a smooth Gaussian-family beam, and the advertised time-varying beam capability is asserted but not demonstrated in the implementation.","major_comments":[{"comment":"The central 'arbitrary beams' claim rests on the step from the continuous Mueller-formalism integral (Eq. 4.3) to the pixelated sum (Eq. 4.5), where the beamsor is re-pixelized and evaluated at sky-pixel centers. No sub-pixel integration, interpolation error bound, or beam-bandwidth criterion is given. All quantitative tests use smooth Gaussian-family beams (circular in §6, elliptical in §7), and the 5-degree truncation justified in §7.1.3 is specific to a 1.5-degree Gaussian. Beams with sharp sidelobes, non-symmetric polarized responses, or significant support beyond the truncation disc could be aliased by pixel-center sampling. Please add at least one non-Gaussian validation (for example, an Airy pattern or a measured beam with sidelobes) against an independent reference, and quantify the interpolation and truncation error for such beams.","section":"§4.2, Eq. (4.5), and Appendix A"},{"comment":"The introduction and abstract promise 'arbitrary shaped, time-varying beams', and §8 states that PISCO can model time-dependent behavior of the optics, but the implementation described in §5.1 and §5.2 takes a single static beamsor as input and the convolution kernel (Eq. 4.5) has no time index. No test with a time-varying beamsor is presented. Either implement and demonstrate a time-indexed beamsor or revise the claims to reflect the current capability.","section":"§1, §5, and §8"},{"comment":"The CLASS simulations are presented as reproducing the expected effects of pointing mismatch, beam mismatch, and uneven intra-pixel coverage, but the comparison is qualitative: leakage amplitudes are described by eye relative to reference spectra, with no quantitative comparison against an independent simulator or an analytic prediction. This does not by itself invalidate the code, but it does not close the validation loop for the systematic effects that are central to the paper's application claim. Please add a quantitative comparison of the measured leakage levels with the analytic expectations from O'Dea et al. (2007) and Poutanen et al. (2005), or state explicitly that these tests are demonstrations only.","section":"§7, Figs. 12–14"}],"minor_comments":[{"comment":"There are typos: 'For simplicitly' should be 'For simplicity' in both places.","section":"§6.1 and §6.2.1"},{"comment":"The text says the scaling of wall-time with the maximum radial extent was 'roughly quadratic', then reports a linear fit in the number of pixels (Eq. 5.1). Since the number of pixels in a disc is approximately quadratic in the radius, both statements are consistent, but the connection should be stated explicitly for clarity.","section":"§5.3.1 and Fig. 5"},{"comment":"The statement that 'We are not aware of any other publicly available code that performs pixel space convolution' is difficult to verify and is somewhat in tension with FEBeCoP being described in the same paragraph; a softer comparison of features would be more appropriate.","section":"§8"},{"comment":"No repository, download URL, or version identifier for PISCO is provided. For a software methods paper, a public code release or at least a clear availability statement would substantially aid reproducibility.","section":"General availability"},{"comment":"The figure captions are inconsistent about whether the reference E/B-mode spectra are shown with a black dashed or purple dashed line; please harmonize the captions with the actual plotting colors.","section":"§7.2.2–§7.2.4, captions of Figs. 12–14"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the instrumentation/computation scope of JCAP, but as a software-oriented submission it would benefit from a public code release; without it, reproducibility is hard to assess. The novelty discussion should be more carefully scoped against FEBeCoP and beamconv, which already address related problems, and the time-varying beam claim should be either implemented or removed from the headline claims before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: PISCO is a real, GPU-accelerated pixel-space convolver that does what its name says for polarized CMB time streams, but the 'arbitrary beams' claim is broader than the evidence. The paper gives a clean implementation of the O'Dea et al. Mueller-beam formalism, with the new bits being CUDA acceleration and a careful treatment of fully polarized beamsors. The validation is genuinely useful: point-source flux preserved to 0.2%, ideal CMB spectra close to input with no T-to-P or E-to-B leakage, and the CLASS simulations reproduce the expected leakage from pointing mismatch, beam mismatch, and uneven intra-pixel coverage. The small residuals at l<=250 are encouraging.\n\nThe soft spots are in proportion to the gap between claims and tests. Every quantitative test uses a smooth Gaussian (circular or elliptical) beam. The load-bearing step, Equation 4.5, replaces a continuous integral with a sum over sky pixel centers after re-pixelating the beamsor; no sub-pixel integration or error bound is given. For sharply varying sidelobes or non-symmetric polarized responses, that approximation could alias power. The 5-degree cutoff is justified only for a 1.5-degree Gaussian, not 'arbitrary beams.' Also, the abstract and intro promise time-varying beams, but the implementation takes a static beamsor and no time-indexed mechanism is described. That is an overreach, not a fatal flaw.\n\nThe biggest practical issue is that no code or data is released. For a software paper that is a real problem: the reader cannot test the advertised generality or reuse the tool. The authors say they are unaware of other public pixel-space convolution code; FEBeCoP is not public, so PISCO could fill a genuine gap if the code were available.\n\nBottom line: this is a competent implementation paper with honest validation in the Gaussian regime. The physics content is not new, the generality is asserted rather than demonstrated, and the lack of release limits its immediate value. But it deserves a serious referee: with a revision that tempers the claims and either releases the code or adds a non-Gaussian test, it would be a useful contribution for CMB systematics studies.","headline":"PISCO is a useful GPU pixel-space convolver for CMB simulations, but its 'arbitrary beams' claim outruns the Gaussian-only validation.","tokens_in":22174,"tokens_out":2072,"would_cite":true,"duration_ms":21164,"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 pixel-space convolver reproduces beam leakage into CMB B-modes.","keywords":["cosmic microwave background","B-mode polarization","pixel space convolution","beam systematic effects","time-ordered data simulation","Mueller matrix","GPU computing","scanning strategy"],"falsifier":"Take an asymmetric or non-Gaussian beamsor with a visible sidelobe, point it at a polarized point source or a CMB-like sky, and compare the PISCO timestreams against an independent harmonic-space convolution of the same beam; if point-source flux or the recovered Stokes Q and U maps differ by more than a few tenths of a percent, the interpolation step does not preserve the convolution.","tokens_in":21023,"feed_emoji":"🔭","tokens_out":5458,"duration_ms":53305,"temperature":0.7,"pith_summary":"This paper introduces PISCO, a GPU-accelerated simulator that produces mock time-ordered data for cosmic microwave background experiments by convolving a fully polarized antenna response with a pixelated sky in real space. The central claim is that pixel-space convolution can handle arbitrary beams, sky models, and scanning strategies, and that it correctly reproduces the power-spectrum effects of pointing mismatch, beam mismatch, and uneven intra-pixel coverage. These effects matter because they create spurious temperature-to-polarization and E-to-B leakage that can mimic the primordial B-mode signal that inflation experiments are trying to detect. The paper demonstrates the method on point sources, an ideal full-sky experiment, and a realistic simulation of an ongoing ground-based survey.","feed_headline":"A pixel-space convolver reproduces beam leakage into CMB B-modes","feed_subtitle":"PISCO turns arbitrary beams and scans into mock timestreams, capturing leakage that could masquerade as primordial B-modes.","key_machinery":"The beamsor, a field of 4x4 Mueller matrices defined at every antenna-basis direction, carries the polarized electromagnetic response of the telescope. The central measurement equation sums the detector response vector, beamsor, polarization-basis rotation matrices, and Stokes sky over every sky pixel inside the beam footprint, and this sum is parallelized over pixels and pointings on a GPU. For matched co-polar beams, the beamsor reduces to the scalar beam times the identity matrix, and arbitrary pointing directions are handled by spherical-trigonometry re-pixelization.","core_discovery":"The discovery is methodological: PISCO evaluates each detector sample as a direct sum over sky pixels of the detector response vector, a beamsor (a field of Mueller matrices encoding the polarized antenna response), rotation matrices that carry the detector polarization basis to the sky basis, and the Stokes sky maps. Convolving in pixel space rather than harmonic space lets the same code handle arbitrary, time-varying beam shapes and arbitrary scanning strategies, which harmonic-space methods handle with difficulty. In validation runs, the flux of a polarized point source is preserved to better than 0.2%, an ideal full-sky CMB simulation shows no temperature-to-polarization or E-to-B leakage, and realistic simulations reproduce the expected leakage levels: pointing mismatch produces a spurious B-mode reaching about 0.02 $microK^{2}$ at l=190 and dominating an r=0.1 B-mode spectrum above l=160, beam mismatch dominates above l=80, and uneven intra-pixel coverage produces subdominant P-to-P leakage that becomes dominant above l=200.","pith_inferences":["If the re-pixelization and interpolation preserve non-Gaussian beams as they do circular Gaussians, the same machinery could quantify the B-mode bias from far sidelobes and near-sidelobe pickup for future surveys; all validation tests in the paper use circular Gaussian beams, so this remains a testable extension.","Because each time sample is computed independently, the architecture could be extended to place detector time constants, gain drifts, or correlated noise directly into the time-ordered data without adding convolution cost.","The observed linear scaling of wall time with number of sky pixels and pointing directions suggests that multi-GPU systems could push pixel-space convolution to the higher resolutions typical of next-generation small-arcminute-beam experiments."],"forward_implications":["PISCO timestreams can inject realistic beam, pointing, and scanning effects into map-making and power-spectrum pipelines, testing how well those pipelines recover the sky.","Effects that are hard to include in harmonic-space implementations, such as atmospheric turbulence, transient sky events, and time-dependent optics, can be added at modest extra cost because each sample is generated independently in pixel space.","Modulation techniques such as half-wave plates can be included by supplying a modulated version of the beamsor.","The matched-versus-mismatched simulations give concrete thresholds for when leakage dominates the B-mode signal: pointing mismatch above multipole 160 and beam mismatch above multipole 80 for an r=0.1 reference spectrum."],"supporting_citations":[{"why":"Supplies the beam-Mueller-field (beamsor) formalism for polarized antenna response and the expected temperature-to-polarization leakage from pointing and beam mismatch.","marker":"[18]"},{"why":"Provides the harmonic-space full-sky beam convolution implementation that PISCO complements and compares against in spirit.","marker":"[8]"},{"why":"Documents the only previously available pixel-space convolution code for CMB surveys, positioning PISCO's approach relative to it.","marker":"[17]"},{"why":"Defines the HEALPix equal-area pixelization used for both sky and beamsor maps, which the convolution is built on.","marker":"[10]"},{"why":"Describes map-making and intra-pixel gradient coupling that the paper invokes to explain the P-to-P leakage from uneven intra-pixel coverage.","marker":"[23]"},{"why":"Supplies the correlation-function power-spectrum estimator used to compute masked spectra in the realistic survey simulations.","marker":"[6]"},{"why":"Documents the experiment whose detector offsets, beam parameters, and scanning strategy drive the realistic simulation.","marker":"[12]"},{"why":"Provides the analytic beam window function for smooth symmetric beams used to deconvolve the ideal-experiment spectra.","marker":"[19]"},{"why":"Defines Ludwig's third definition of cross polarization, which underlies the co-polarization angle computation.","marker":"[15]"}],"fun_headline_variants":["PISCO: pixel-space convolution reproduces beam leakage into B-modes","Pixel-space tool simulates CMB beam leakage for arbitrary scans","GPU code PISCO maps beam leakage in pixel space","Pixel-space convolution predicts B-mode leakage from beam mismatches","PISCO: simulating CMB mock data streams with full polarization"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The beamsor formalism is assumed to be accurate for arbitrary, time-varying beams, and the interpolation and re-pixelization are assumed to preserve the convolution for non-Gaussian beams; all validation tests use circular Gaussian beams, so that advertised versatility is unverified.","fun_headline_variants_meta":{"raw":{"variants":["PISCO: pixel-space convolution reproduces beam leakage into B-modes","Pixel-space tool simulates CMB beam leakage for arbitrary scans","GPU code PISCO maps beam leakage in pixel space","Pixel-space convolution predicts B-mode leakage from beam mismatches","PISCO: simulating CMB mock data streams with full polarization"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000233,"raw_usage":{"total_tokens":1481,"prompt_tokens":923,"completion_tokens":558,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":539,"completion_tokens_details":{"reasoning_tokens":470}},"tokens_in":539,"tokens_out":558,"duration_ms":5727,"temperature":1.0,"reasoning_tokens":470,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:17:32.416214+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take an asymmetric or non-Gaussian beamsor with a visible sidelobe, point it at a polarized point source or a CMB-like sky, and compare the PISCO timestreams against an independent harmonic-space convolution of the same beam; if point-source flux or the recovered Stokes Q and U maps differ by more than a few tenths of a percent, the interpolation step does not preserve the convolution.","supporting_citations":[{"cited_title":"O’Dea, A","cited_arxiv_id":null,"evidence_quote":"Supplies the beam-Mueller-field (beamsor) formalism for polarized antenna response and the expected temperature-to-polarization leakage from pointing and beam mismatch."},{"cited_title":"Full-sky beam convolution for cosmic microwave background applications","cited_arxiv_id":"1809.05034","evidence_quote":"Provides the harmonic-space full-sky beam convolution implementation that PISCO complements and compares against in spirit."},{"cited_title":"Mitra, G","cited_arxiv_id":null,"evidence_quote":"Documents the only previously available pixel-space convolution code for CMB surveys, positioning PISCO's approach relative to it."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the HEALPix equal-area pixelization used for both sky and beamsor maps, which the convolution is built on."},{"cited_title":"Lawrence, Davide Maino, P Natoli, S Prunet, Radek Stompor, and Romain Teyssier","cited_arxiv_id":null,"evidence_quote":"Describes map-making and intra-pixel gradient coupling that the paper invokes to explain the P-to-P leakage from uneven intra-pixel coverage."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the correlation-function power-spectrum estimator used to compute masked spectra in the realistic survey simulations."},{"cited_title":"Harrington, T","cited_arxiv_id":null,"evidence_quote":"Documents the experiment whose detector offsets, beam parameters, and scanning strategy drive the realistic simulation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the analytic beam window function for smooth symmetric beams used to deconvolve the ideal-experiment spectra."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines Ludwig's third definition of cross polarization, which underlies the co-polarization angle computation."}],"review_version":1}