{"id":"6a34ae01-c4bc-4857-973b-fd19463a46f3","arxiv_id":"2511.00159","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Interferometric 21-cm observations of simulated Milky Way-mass galaxies miss 10-40% of diffuse hydrogen outside the disk because short-baseline data are absent, and no single correction factor works.","lead":"Radio interferometers can miss a large fraction of the faint hydrogen gas surrounding galaxies because they lack short-baseline data. This paper uses six simulated Milky Way-like galaxies to show that the missing fraction is 10-40% and depends on orientation and environment, so single-dish telescopes are needed to recover it.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Headline loss fractions rest on a Gaussian high-pass filter that omits real uv sparsity; actual interferometric recovery of diffuse CGM HI may differ materially from Table 2 and Fig 7.","rationale":"The reader's weakest assumption—that the Gaussian high-pass plus CLEAN pipeline faithfully represents missing short baselines—is indeed the load-bearing point. The paper is transparent about the simplification and even describes it as conservative by retaining more uv points than a real interferometer, so the qualitative conclusion is probably safe: real observations are unlikely to recover more diffuse CGM emission. But the central contribution is quantitative: 10–40% CGM HI loss, up to 50% in a worst case, and the '96–99%'/'15%' abstract values. A conservative bias of unknown size is precisely the kind of unvalidated numerical approximation that should be checked before accepting these percentages. The disk retention (≥98%) and the HI size-mass comparison provide independent support for the simulations themselves, but they do not validate the missing-short-baseline loss fractions. I also considered the abstract/body mismatch (20 Mpc vs 10 Mpc) and the missing FOGGIE XII reference; these are real presentation and reproducibility problems, but they are secondary to the fact that the headline numbers are produced by an idealized filter that has not been compared with actual uv-sampled mock observations. Because the reader already issued a CONDITIONAL verdict, this concern does not move the verdict; it sharpens the condition: the quantitative claim should either be tested with a realistic uv-sampling pipeline or restated as a model-dependent lower limit with quantified caveats.","tokens_in":20536,"tokens_out":12558,"duration_ms":117991,"concrete_test":"Run mock observations of the same six FOGGIE datacubes at 10 Mpc with a real uv-sampling tool (CASA simobserve or vis_sample) using the actual MeerKAT minimum baseline (29 m) and VLA D-config (35 m) configurations; image and CLEAN to 3σ as in §4, run the same SoFiA-2 source detection, and recompute the CGM HI ratios in Table 2. Accept the Gaussian-filter approximation if each halo's filtered/smoothed CGM ratio changes by <0.10 absolute (e.g., Blizzard MHONGOOSE-LR remains within 0.42–0.62). If shifts are larger, replace the 10–40% range with the corrected values and repeat the pipeline at 20 Mpc to test the abstract numbers.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's quantitative claim—interferometric 21-cm observations lose 10–40% of CGM HI, with up to 50% at bmin=29 m—rests entirely on the §4 synthetic filtering step. That step is an FFT of each spectral slice, a Gaussian high-pass filter whose FWHM is tied to the minimum baseline, an inverse FFT, and a Högbom CLEAN. The authors state in §4: 'we also did not consider any sparse sampling of the rest of the UV plane, meaning we are retaining more UV points than an actual interferometer would.' This is not a cosmetic detail: the amount of diffuse flux that survives into the 'filtered' cubes feeding Table 2 and Fig 7 depends on the full uv transfer function. A real array has incomplete, non-uniform uv sampling, calibration errors, and a dirty beam with sidelobes; CLEAN depth and SoFiA-2 masking interact with those artifacts and can either recover or suppress additional extended emission. Since the model knowingly retains more uv points than a real observation, the true loss is likely larger than quoted, but by how much is unquantified. The abstract's '96–99%'/'up to 15%' numbers are similarly built from this idealized filter, and are additionally stated at 20 Mpc even though the body's fiducial setup is 10 Mpc. Until the filter is validated against actual uv-sampled mock observations, the exact loss percentages are an uncalibrated model prediction, not a measured result.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses the six Milky Way-mass FOGGIE zoom-in simulations to study the observability of extended HI disks, warps, and circumgalactic HI in synthetic 21-cm observations. The authors construct synthetic datacubes, add noise, smooth to survey resolution, and apply a Gaussian high-pass spatial filter plus CLEAN deconvolution to mimic the missing short baselines of interferometers. They report that interferometric surveys recover essentially all HI within the central disk, but that 10-40% of CGM HI (up to ~50% in Fig. 7) can be lost, preferentially at low column density and low velocity dispersion. They also compare the simulated disks with the HI size-mass relation and discuss how inclination and position angle affect both the recovered HI mass and the observed kinematics of warps.","tokens_in":20864,"tokens_out":5190,"duration_ms":47818,"significance":"If the quantitative results hold, the paper addresses an important systematic in HI galaxy surveys: the missing short-baseline problem is often discussed qualitatively, but this is one of the first attempts to quantify its effect on extended CGM HI in realistic cosmological simulations. The strength of the paper is its detailed synthetic pipeline, the use of a suite with well-resolved CGM gas, and the comparison across multiple survey configurations (MHONGOOSE-LR/HR, THINGS, SKA). The conclusion that single-dish plus interferometric data are needed to recover diffuse CGM HI is timely and relevant to ongoing and planned surveys. However, the headline loss fractions rest on an idealized UV-plane filter that has not been validated against actual interferometric sampling, and the abstract numbers are inconsistent with the body. These issues currently prevent the quantitative claims from being fully accepted.","major_comments":[{"comment":"The arXiv abstract reports that observations at 20 Mpc retain ~96-99% of total HI and miss up to ~15% of HI outside the central disk, while the body abstract and §5 report that interferometric observations can miss ~10-40% of diffuse emission, and Fig. 7 shows up to ~50% loss at b_min=29 m. The fiducial setup in §4 is 10 Mpc, not 20 Mpc. These are mutually inconsistent numbers for the same central claim. Please reconcile the abstract with the body and specify whether the 20 Mpc/96-99%/15% numbers are new results or an error.","section":"Abstract and §4-5"},{"comment":"The quantitative loss fractions (Table 2, Fig. 7) are derived from an FFT, a Gaussian high-pass filter with FWHM set by the minimum baseline, and a Högbom CLEAN. The paper explicitly notes: \"we also did not consider any sparse sampling of the rest of the UV plane, meaning we are retaining more UV points than an actual interferometer would.\" This is a load-bearing approximation: real interferometers have irregular, incomplete UV coverage, sidelobes, and calibration artifacts that interact with CLEAN depth and SoFiA-2 source masking. The direction and magnitude of the resulting bias are not quantified. Please validate the filter against at least one realistic UV-sampled mock observation (e.g., CASA simobserve) for a representative configuration, or provide a quantitative estimate of the systematic uncertainty this approximation introduces.","section":"§4, synthetic filtering step"},{"comment":"The paper claims the spatial filtering step reduces recoverable CGM emission by ~10-40%, but Table 2 contains values inconsistent with that range. For example, Tempest in the SKA column goes from 0.722 (smoothed) to 0.0759 (filtered), a filtering loss of ~90%. Fig. 7's caption states that an observation with b_min=29 m \"potentially misses 50% of the CGM HI,\" also outside the 10-40% range. Additionally, Table 2 reports no uncertainties and no indication whether the values are single snapshot/single orientation; the strong orientation dependence shown in Fig. 8 suggests that one number per halo is insufficient. Please provide uncertainties or a range of values across orientations/time snapshots and reconcile the reported loss fractions with the table and figure.","section":"Table 2 and Fig. 7"}],"minor_comments":[{"comment":"The companion paper FOGGIE XII is cited as \"arXiv:2510.tbd\" in the reference list (Trapp et al. 2025). This is a placeholder, not a complete reference, and several key definitions (disk definition, population classes) rely on it. Please provide the full reference or state the status of the companion paper.","section":"References"},{"comment":"The text says \"All the Less Populated systems (Tempest, Maelstrom, and Hurricane)\" but Table 1 classifies Hurricane as More Populated; the Less Populated systems are Tempest, Maelstrom, and Blizzard. This appears to be a typo but is confusing in the discussion of the HI size-mass relation.","section":"§3, Table 1"},{"comment":"The discussion of Tempest's position-angle series refers to \"the second panel of Fig. 8\" for the velocity field; the first-moment maps are shown in Fig. 9 (and position-velocity diagrams in Fig. 10), while Fig. 8 shows the observable HI ratio. Please correct the figure reference.","section":"§7.2"},{"comment":"The units of the synthetic flux density are given as \"cm^-2 s\", which is nonstandard for spectral data cubes. If the intention is column density per channel, please define the relation to brightness temperature or Jy/beam explicitly. Also define m_HI or m_H consistently.","section":"§4, Eq. 1"},{"comment":"The caption says the column density sensitivity is 10^18 cm^-2 and that noise is added with standard deviation 0.2 times the sensitivity limit, so that 5σ detections correspond to the sensitivity limit. This is clear, but for reproducibility, the exact kernel sizes used for SoFiA-2 should be stated in the text or a table, since they are only described as \"adjusted for other surveys as necessary.\"","section":"Figures 2-3"}],"recommendation":"major_revision","confidential_remarks":"The paper's central idea is sound and the pipeline is described in unusual detail, but the headline quantitative claims need to be made consistent and the idealized UV filter needs validation against realistic interferometric sampling before the specific loss percentages can be trusted. I would not reject; this is a major-revision situation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth reading, and worth sending to referees. The paper makes a concrete claim with real consequences: interferometric 21-cm observations can miss roughly 10–40% of the HI mass in the CGM of Milky-Way-mass galaxies, and the effect is so dependent on galaxy, distance, and orientation that no simple correction factor will work. The body's tables and figures support that claim; the abstract's different numbers do cause confusion.\n\nThe genuinely new part is the application of a synthetic-interferometer pipeline to the CGM-resolved FOGGIE zoom-ins. The pipeline is careful and described in enough detail to follow: Voigt profiles, noise, beam smoothing, primary beam, SoFiA-2 source finding, and then a spatial high-pass filter plus CLEAN to mimic missing short baselines. The finding that the filtering preferentially removes low column density, low velocity dispersion gas—and that this brings the simulated N_HI distributions closer to what surveys like MHONGOOSE actually see—is a solid result with a useful message for observers.\n\nThe soft spots are real but not fatal. First, the abstract quotes 96–99% total HI retention and up to 15% loss outside the disk at 20 Mpc, while the body's fiducial is 10 Mpc and the CGM-only loss in Table 2 and Fig. 7 runs to ~40–50%. The authors need to align the abstract with the body. Second, the synthetic filtering step is a Gaussian high-pass filter that keeps all other uv points; the authors themselves note this is not sparse sampling. A real interferometer has irregular uv coverage, dirty beams, and calibration errors. That makes the exact percentages uncalibrated—likely underestimates, but by how much is unknown. Third, FOGGIE XII, which supplies the disk definition and population classes, appears with a placeholder arXiv ID. That has to be fixed. Table 2's lack of uncertainties is minor, since the pipeline is controlled and the sample is six galaxies.\n\nI'd accept this for peer review. The qualitative conclusion—that missing short baselines bias HI observations of the CGM—is robust, and the paper gives observers and simulators a useful warning. The quantitative loss fractions should be framed as estimates pending a more realistic uv-sampling validation. Recommend the authors test their filter against a CASA-style mock observation with actual uv coverage, and fix the abstract/body mismatch.","headline":"Useful and honest forward-modeling study; the 10–40% CGM-loss result is directionally right, but exact numbers rest on an idealized uv filter and the abstract/body mismatch needs fixing.","tokens_in":21471,"tokens_out":3853,"would_cite":true,"duration_ms":35044,"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":"Interferometric observations of Milky Way-mass galaxies can miss 10–40% of the diffuse hydrogen gas surrounding them.","keywords":["atomic hydrogen","HI 21-cm line","interferometry","missing short baselines","circumgalactic medium","galaxy warps","synthetic observations","disk galaxies"],"falsifier":"Measure total HI flux of a sample of nearby Milky Way-mass galaxies with a single-dish telescope and compare it with interferometric maps of the same galaxies. If single-dish fluxes are systematically higher by tens of percent in the outer regions, the missing-short-baseline loss is real and substantial; if they agree, the simulated diffuse CGM gas is likely overproduced.","tokens_in":20349,"feed_emoji":"📡","tokens_out":6081,"duration_ms":45089,"temperature":0.7,"pith_summary":"This paper argues that interferometric 21-cm observations of Milky Way-mass galaxies systematically undercount the diffuse atomic hydrogen that surrounds them, missing roughly 10–40% of the emission from the circumgalactic medium. The cause is not sensitivity or resolution but the absence of short baselines, which makes interferometers blind to spatially large, low-column-density gas. The authors demonstrate this by building synthetic 21-cm emission cubes from six simulated galaxies and then filtering out low spatial frequencies the way a real interferometer does. They find that the lost gas is preferentially the diffuse, low-velocity-dispersion component, and that the amount lost depends on the galaxy, its distance, and its orientation, so no single correction factor can fix the bias. The practical consequence is that fully measuring extended HI disks and their interface with the circumgalactic medium requires combining interferometers with single-dish telescopes.","feed_headline":"Missing baselines hide up to 40% of a galaxy's diffuse gas","feed_subtitle":"Simulated galaxies show interferometers lose the faint, spread-out hydrogen linking disks to their surroundings.","key_machinery":"The key mechanism is the synthetic 21-cm observation pipeline: the authors project simulated gas into datacubes, add noise, smooth with a Gaussian beam, and then mimic the missing short baselines by applying a fast Fourier transform, cutting out low spatial frequencies below the minimum baseline with a Gaussian high-pass filter, and cleaning the dirty image with the Högbom CLEAN algorithm. This spatial filter is what isolates the effect under study; it selectively removes the diffuse, large-scale emission that an interferometer cannot see. The pipeline also includes the SoFiA-2 source finder to identify significant emission, matching the procedures of real surveys like MHONGOOSE, THINGS, and","core_discovery":"The paper's central claim is that the missing short baselines of interferometric arrays remove a non-negligible and biased fraction of HI emission from the circumgalactic medium of Milky Way-mass galaxies. Using mock observations of six simulated galaxies, the authors show that while the inner disk is essentially unaffected (98% or more of its HI is recovered), the diffuse, spatially extended gas outside the disk loses 10–40% of its detectable mass, and in the worst case up to half. The filtering preferentially removes low column density (N_HI < 10^20 cm^-2) gas with low velocity dispersion, which is exactly the material that traces the disk–CGM interface and possible accretion. Because the","pith_inferences":["If the 10–40% loss applies to real galaxies, the cosmic HI mass density in the CGM may be underestimated by a similar factor, affecting models of galaxy accretion and baryon cycling.","The bias toward compact, high-column-density clumps means some of the 'missing baryons' problem could be partly an observational selection effect.","Applying the same filtering pipeline to other simulation suites with coarser CGM resolution could show that the inferred loss fraction is resolution-dependent, not a universal constant.","The position-angle-dependent velocity inversions seen in warped disks suggest some observed 'anomalous' kinematics in galaxies could be projection artifacts of warps rather than true kinematic features."],"forward_implications":["Interferometric HI surveys systematically underestimate the amount of diffuse gas in the circumgalactic medium, and the missing fraction varies from galaxy to galaxy.","Comparisons between simulations and observations must forward-model the missing short spacings before drawing conclusions about HI content or kinematics.","Measured disk sizes from interferometric maps may be underestimated by a few kiloparsecs for galaxies with diffuse extended emission.","The orientation dependence adds an uncalibratable scatter to statistical samples of galaxy HI properties.","Combining single-dish and interferometric data is not just an improvement but a requirement for measuring the full HI content of the CGM."],"fun_headline_variants":["Interferometers lose up to 40% of a galaxy's diffuse halo gas","Missing short spacings bias HI views of galactic halos","Galaxy surveys miss faint gas linking disks to their surroundings","Short-baseline gaps hide 40% of circumgalactic hydrogen"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The quantitative loss fractions assume that a Gaussian high-pass filter plus CLEAN deconvolution faithfully reproduces the effect of missing short baselines; real interferometers sample the uv-plane sparsely and irregularly, so the actual missing fraction could be higher or lower.","fun_headline_variants_meta":{"raw":{"variants":["Interferometers lose up to 40% of a galaxy's diffuse halo gas","Missing short spacings bias HI views of galactic halos","Galaxy surveys miss faint gas linking disks to their surroundings","Short-baseline gaps hide 40% of circumgalactic hydrogen"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000197,"raw_usage":{"total_tokens":1234,"prompt_tokens":812,"completion_tokens":422,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":556,"completion_tokens_details":{"reasoning_tokens":347}},"tokens_in":556,"tokens_out":422,"duration_ms":13785,"temperature":1.0,"reasoning_tokens":347,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T06:52:54.042305+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure total HI flux of a sample of nearby Milky Way-mass galaxies with a single-dish telescope and compare it with interferometric maps of the same galaxies. If single-dish fluxes are systematically higher by tens of percent in the outer regions, the missing-short-baseline loss is real and substantial; if they agree, the simulated diffuse CGM gas is likely overproduced.","supporting_citations":[],"review_version":1}