{"id":"894a0463-817c-4282-8a5a-ef7428ab4ffc","arxiv_id":"2608.12933","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"Gal3D fits flexible superellipsoid shapes to density surfaces inside simulated galaxies and finds systematic differences in disks, bars, and boxy bulges between TNG and EAGLE simulations.","lead":"This paper introduces Gal3D, a tool that fits 3D superellipsoid shapes to the surfaces of equal density inside galaxy simulations, capturing disks, bulges, bars, and boxy structures. A smart generalist might read it because it offers a new way to compare how different galaxy simulations build 3D structure and to connect morphology to galaxy formation physics.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified","rationale":"The reader's weakest_assumption identifies the single-valued-per-ray inversion under Section 2.4.1 as the most fragile step for the box/peanut claims. I agree with the mechanism: an X-shaped box/peanut bulge has multiple density peak radii along some lines of sight, and the monotonic regularization collapses that to one radius per ray, so the fitted S_c and S_a are not direct measurements of the X-shape but rather summary parameters of the envelope. However, the paper itself says in Section 2.5 that 'strongly X-shaped box/peanut bulges are not fully described by this model, although their presence can still be reflected by large values of S_a and S_c,' which is an explicit scope limitation and matches the reader's conditional framing. A stress test on synthetic shapes would settle whether the S-indices recover the ground truth in the non-monotonic regime, which is the honest load-bearing unknown. That test is cheap with the public code and would raise confidence. The absence of error bars on Figure 7 and of significance statements for the TNG-EAGLE differences is a reporting weakness that the reader flagged correctly, but it is not an internal inconsistency. The paper's comparative claims might be overstated relative to their statistical support, though this is a presentation issue rather than a demonstrated flaw. I would not move the verdict from the reader's CONDITIONAL; the conditions (synthetic validation, error bars) already constitute the right level of qualified acceptance.","tokens_in":28512,"tokens_out":1374,"duration_ms":13254,"concrete_test":"Run a synthetic recovery test: construct a known superellipsoid density field with axis ratios and shape indices spanning the claimed ranges (S in 0.2 to 2), sample it with particles at the resolution of TNG-like and EAGLE-like runs, apply the Gal3D pipeline, and verify the recovered (epsilon_ab, epsilon_bc, S_a, S_b, S_c) match the input within the reported fit covariance. This directly tests the inversion and objective function without the ambiguity of real galaxy profiles.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No significant objection identified. The central methodological claim, that Gal3D's superellipsoid indices recover higher-order boxiness and diskiness, is internally coherent and supported by the numerical sensitivity tests in Appendix A. The reader's weakest_assumption correctly notes that the star-shaped single-valued-per-ray inversion in Section 2.4.1 suppresses non-monotonic density structure such as X-shaped box/peanut bulges. However, the paper explicitly acknowledges this limitation in Section 2.5 (a single superellipsoid cannot represent strongly X-shaped bulges exactly, though their presence is reflected in large S_a and S_c). This is an honest scope statement rather than a hidden flaw. The application claims lack error bars and significance tests, and synthetic-shape validation is absent, but these are completeness gaps rather than demonstrable errors in the central argument. The method's robustness to k and N_ray is established in Appendix A, code is publicly available, and the TNG versus EAGLE differences follow from the presented profiles.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents Gal3D, an open-source framework that reconstructs a smoothed stellar density field from particle data using adaptive kernel density estimation and fits superellipsoids to iso-density surfaces, extracting radial profiles of axis ratios, orientation angles, center offsets, and superellipsoid shape indices S_a, S_b, S_c. The method is applied to z=0 galaxy samples from IllustrisTNG (TNG50-1, TNG50-2, TNG100-1) and EAGLE100, and the resulting mean profiles are used to compare disk extents, bar strengths, box/peanut bulge signatures, and outer triaxiality as functions of stellar mass. The paper also includes numerical sensitivity tests for the fitting objective and for the smoothing and angular-sampling parameters, and makes the code and shape-profile data publicly available.","tokens_in":28758,"tokens_out":6011,"duration_ms":60639,"significance":"If the method performs as claimed, Gal3D offers a substantially more flexible description of intrinsic 3D galaxy structure than the standard iterative inertia tensor, with the potential to separate boxiness and diskiness in a single radial profile. The reported TNG versus EAGLE differences in box/peanut bulge signatures and outer triaxiality are of direct interest for galaxy formation comparisons. Strengths of the paper include the public code and data releases, the careful sensitivity analysis in Appendix A, and the explicit acknowledgment in Section 2.5 of the limitation imposed by the single-valued isodensity assumption. However, the absence of synthetic recovery tests and the lack of uncertainties on the population maps currently leave the central quantitative claims under-supported.","major_comments":[{"comment":"The population conclusions in Sections 5.1-5.3 rest on the mean maps in Figure 7, which are shown without any estimate of the uncertainty in the mean (for example, bootstrap or jackknife over galaxies) or a significance test for the TNG versus EAGLE differences. Please add per-bin uncertainties and perform a two-sample significance test in the relevant mass-radius bins, especially for the claims that EAGLE has less extended disks, weaker bars, and weaker box/peanut signatures; without these, the reported offsets may be within sampling noise.","section":"Section 5 / Figure 7"},{"comment":"The numerical tests in Appendix A compare the two objective functions and vary k and N_ray, but they never test whether the fitting recovers known input parameters. Because the core claim of the paper is that the superellipsoid indices S_a, S_b, S_c are recovered from simulated galaxies, the manuscript should include synthetic recovery tests in which density fields with known superellipsoid shape parameters (including S<1 and S>1, and with additional particle noise) are fitted and the input values are recovered. This would directly support the interpretation of S_a>1 and S_c>1 as box/peanut signatures and S_c<1 as a vertically disky morphology used in Sections 4.3 and 5.2.","section":"Appendix A"},{"comment":"The monotonic regularization of the radial density profile in Section 2.4.1 enforces a single isodensity radius per ray, which suppresses genuine non-monotonic structures such as X-shaped box/peanut bulges. The paper acknowledges this in Section 2.5, but the comparison in Section 5.2 between TNG and EAGLE box/peanut signatures relies on the S_a and S_c values obtained from these regularized profiles. Please quantify the impact of this approximation on the fitted indices for representative X-shaped or multi-component morphologies, or add an explicit caveat that the measured box/peanut differences refer to the regularized single-valued description.","section":"Section 2.4.1 / 2.5"}],"minor_comments":[{"comment":"The statement that the eigenvalues scale as lambda_i proportional to a^2, b^2, c^2 is made for a uniform ellipsoidal shell; please state the assumed surface density and the normalization convention so that readers can reproduce the relationship.","section":"Section 2.3 / Equation (5)"},{"comment":"The area-weighting factor (r'_i)^2 in the objective function is described as approximate; a brief derivation of why this factor corresponds to an equal-area weighting for the Fibonacci-sampled rays would improve transparency.","section":"Section 2.4.3 / Equation (10)"},{"comment":"The terms 'boxiness' and 'diskiness' are used to describe the superellipsoid shape indices, but these could be confused with the standard isophotal shape coefficients a_4/a; please add a sentence clarifying that S>1 and S<1 refer to the superellipsoid's departure from a purely ellipsoidal surface, not to Fourier isophote coefficients.","section":"Section 4.3"},{"comment":"The iterative shrinking-sphere centering method is cited but not described; adding a one-sentence summary of the algorithm and its convergence criterion would aid reproducibility.","section":"Section 3.3"}],"recommendation":"major_revision","confidential_remarks":"The manuscript makes a useful methodological contribution and the TNG-EAGLE comparison is interesting, but the central claims currently exceed the evidence: there is no synthetic recovery validation and the population maps lack error bars or significance tests. Both issues are fixable within a revision, so I do not recommend rejection, but the paper is not ready for acceptance in its present form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a look. The genuinely new thing here is Gal3D's superellipsoid fit to isodensity surfaces, with the S_a,S_b,S_c indices giving a 3D handle on boxiness and diskiness that plain inertia tensors cannot provide. The geometry is coherent: Eq. 5 gives the right eigenvalue scaling for ellipsoidal shells, and the projected residual maps (Fig. 3) show the superellipsoid beats the ellipsoid. The method is also reproducible: code on GitHub, data on OSF, and a sensible sensitivity study for the KDE neighbor count and ray number.\n\nThe TNG-EAGLE comparison is the weaker half of the paper. Section 5 presents galaxy-averaged maps in the mass-radius plane without error bars, significance tests, or a properly mass-matched sample. The claims that EAGLE disks saturate at lower mass and that TNG bars show stronger box/peanut signatures come out of those maps. They may well be right—the profiles in Figure 7 show coherent ordering—but nothing rules out selection effects or small-sample noise in the high-mass bins. The \"bar region\" also relies on an epsilon_ab=0.4 threshold presented as a visual guide, so the population-level statistics inherit that fuzziness.\n\nTwo methodological gaps matter more. First, there is no synthetic recovery test: Appendix A shows robustness to k and N_ray but never fits a known superellipsoid and checks that S_a, S_b, S_c come back correctly. The mapping from S values to physical morphology is asserted, not calibrated. Second, the ray-based inversion in Section 2.4.1 imposes single-valued, monotonic profiles per direction. Strongly X-shaped box/peanut bulges are therefore flattened into a single radius; the authors explicitly acknowledge this, so it is not hidden, but it means S indices could be biased for exactly the structures that Section 5 claims to distinguish between TNG and EAGLE. This is a completeness problem, not a demonstrated error.\n\nWho this is for: anyone working with simulated galaxy morphology who wants a practical radial 3D shape description. The method deserves a serious referee and publication as a methods paper. The comparative science claims should be backed with error bars and synthetic-shape validation before they appear as results. I would not desk-reject this; I would send it out with a request for those additions.","headline":"Gal3D is a genuine methodological step forward for measuring 3D boxiness/diskiness in simulated galaxies; the TNG-EAGLE comparison is plausible but presently under-supported by statistics.","tokens_in":29284,"tokens_out":2523,"would_cite":true,"duration_ms":27340,"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":"Gal3D fits superellipsoids to simulated galaxies' density surfaces, and its shape indices $S_a$, $S_b$, $S_c$ separate disks, bars, bulges, and triaxial components, exposing a TNG–EAGLE difference in box/peanut bulge prevalence.","keywords":["galaxy structure","superellipsoid","three-dimensional shape","box/peanut bulge","IllustrisTNG","EAGLE","cosmological hydrodynamical simulations","stellar morphology"],"falsifier":"Build a synthetic galaxy whose density is a known superellipsoid plus an X-shaped or ring-like perturbation, run Gal3D's ray-inversion pipeline on it, and check whether the fitted $S_a$ and $S_c$ recover the input values; if they are systematically biased when the radial density profile is non-monotonic, the box/peanut claims derived from TNG would be called into question.","tokens_in":28326,"feed_emoji":"🌌","tokens_out":12885,"duration_ms":101877,"temperature":0.7,"pith_summary":"The intrinsic three-dimensional shapes of galaxies are normally hidden behind projection effects; Gal3D sidesteps this by working on simulated particle data, reconstructing a smooth density field, and fitting superellipsoids to surfaces of constant density. The central claim is that the shape indices $S_a$, $S_b$, and $S_c$, which the fit returns alongside the usual axis ratios and orientations, quantify boxiness and diskiness that standard ellipsoidal fits miss, so a single radial profile can separate disks, bars, bulges, and triaxial spheroids. Applying the method to the IllustrisTNG and EAGLE simulations, the paper finds that the radial extent of flattened disk structure grows with stellar mass up to roughly $10^{11}\\,M_\\odot$ and then drops sharply, that bar-related elongation strengthens above $10^{10.5}\\,M_\\odot$ but is weaker in EAGLE, and that box/peanut bulge signatures are common in TNG but weak or absent in EAGLE. If these measurements are right, Gal3D offers a practical, open-source tool for comparing intrinsic morphology across cosmological simulations and for connecting three-dimensional structure to projected observables.","feed_headline":"Boxy galaxy bulges are more common in TNG than EAGLE","feed_subtitle":"Higher-order shape indices separate disks, bars, and boxy bulges in one radial profile","key_machinery":"The load-bearing object is the superellipsoid isodensity surface $f(x,y,z)=[(x/a)^2]^{S_a}+[(y/b)^2]^{S_b}+[(z/c)^2]^{S_c}=1$, with semi-axes $a\\ge b\\ge c$ and shape indices $S_a$, $S_b$, $S_c$ controlling boxiness ($S>1$) versus diskiness/pointedness ($S<1$). The paper constructs the density field with adaptive kernel density estimation, samples it along a golden-ratio Fibonacci lattice of rays, forces each ray's radial density profile to be monotonic via envelope interpolation, inverts to get one isodensity point per ray, and fits the superellipsoid by minimizing an area-weighted surface-radius-ratio mismatch $(r'^2 (D-1)^2)$. This parameterization allows each density level to have its own center, orientation, and higher-order shape, which is what lets one radial profile separate nuclear disks, bulges, main disks, bars, box/peanut bulges, and triaxial spheroids.","core_discovery":"The central discovery is that superellipsoid fitting of isodensity surfaces recovers higher-order, non-ellipsoidal structure that iterative shape-tensor methods miss, and that this structure is physically informative. Fitting the implicit surface $(x/a)^{2S_a} + (y/b)^{2S_b} + (z/c)^{2S_c} = 1$ to each density level returns axis ratios, Euler angles, center offsets, and shape indices; values $S>1$ correspond to boxy surfaces and $S<1$ to pointed, disky ones. In the TNG and EAGLE samples, the method shows the outer parts of TNG bars carry elevated $S_a$ and $S_c$, the signature of box/peanut bulges, while EAGLE bars do not, and it traces the expansion and collapse of disk regions with stellar mass. The paper presents these results as evidence that Gal3D can serve as a practical standard for quantifying intrinsic radial 3D structure in simulations.","pith_inferences":["One could test whether the same superellipsoid indices, applied to dark matter or gas components, trace assembly history more directly than the stellar shapes alone.","The monotonicity regularisation used in ray inversion could be relaxed (e.g., allowing two radii per direction for X-shaped bulges) to see whether the TNG box/peanut signatures become even stronger.","If Gal3D were applied to observed galaxies via deprojection of edge-on images, the inferred $S_c$ might serve as a new observational discriminator between boxy and disky bulges.","The resolution dependence seen within TNG (thinner disks at higher resolution) warns that the EAGLE–TNG differences could be partly numerical; a resolution-matched EAGLE run would settle this."],"forward_implications":["A single Gal3D radial profile can flag the presence of a bar, a box/peanut bulge, a disk, and a spheroid without separate component-by-component decompositions.","The result that flattened disk extent peaks near $M_* \\sim 10^{11}\\,M_\\odot$ and then declines sets a quantitative benchmark that galaxy formation models should reproduce.","The systematic TNG–EAGLE difference in box/peanut bulge strength implies that subgrid feedback implementations shape the vertical structure of bars, a prediction that can be compared with edge-on observations of barred galaxies.","Because Gal3D can project its 3D models, it offers a route to translating intrinsic shape indices into predicted isophotal boxiness/diskyness, linking simulations to observable quantities."],"supporting_citations":[{"why":"Establishes the iterative shape-tensor framework that Gal3D adapts to a smoothed density field.","marker":"M. Zemp et al. 2011"},{"why":"Provides the cubic-spline kernel used for the adaptive density estimation.","marker":"J. J. Monaghan & J. C. Lattanzio 1985"},{"why":"Supplies the EAGLE simulation data set used for the galaxy sample and comparisons.","marker":"J. Schaye et al. 2015"},{"why":"Supplies the IllustrisTNG simulation data set used for the galaxy sample and comparisons.","marker":"A. Pillepich et al. 2018"},{"why":"Identifies the specific TNG50 box/peanut galaxy used to validate the $S_a$–$S_c$ signature in Figure 5.","marker":"S. R. Anderson et al. 2023"},{"why":"Earlier measurement of bar fractions in EAGLE that the bar-strength comparison extends.","marker":"D. G. Algorry et al. 2017"},{"why":"Earlier comparison of bar fractions and strengths between TNG and EAGLE that this work builds on.","marker":"D. Zhao et al. 2020"}],"fun_headline_variants":["Superellipsoids unveil boxy bulges in TNG and EAGLE","3D shape fit separates boxy TNG from rounder EAGLE","Gal3D pinpoints box/peanut structure in simulated bars","Superellipsoid indices track disk flattening and boxiness","Boxiness signature in TNG bars, missing in EAGLE"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The method assumes that every ray from the galaxy center hits any given density level exactly once, so non-monotonic structures like X-shaped box/peanut bulges are forced into a single radius per direction and the fitted $S_a$ and $S_c$ could be biased.","fun_headline_variants_meta":{"raw":{"variants":["Superellipsoids unveil boxy bulges in TNG and EAGLE","3D shape fit separates boxy TNG from rounder EAGLE","Gal3D pinpoints box/peanut structure in simulated bars","Superellipsoid indices track disk flattening and boxiness","Boxiness signature in TNG bars, missing in EAGLE"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000283,"raw_usage":{"total_tokens":1760,"prompt_tokens":1120,"completion_tokens":640,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":736,"completion_tokens_details":{"reasoning_tokens":544}},"tokens_in":736,"tokens_out":640,"duration_ms":6527,"temperature":1.0,"reasoning_tokens":544,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:20:11.932750+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Build a synthetic galaxy whose density is a known superellipsoid plus an X-shaped or ring-like perturbation, run Gal3D's ray-inversion pipeline on it, and check whether the fitted $S_a$ and $S_c$ recover the input values; if they are systematically biased when the radial density profile is non-monotonic, the box/peanut claims derived from TNG would be called into question.","supporting_citations":[{"cited_title":"The interplay between accretion, galaxy downsizing and the formation of box/peanut bulges in TNG50","cited_arxiv_id":"2302.12788","evidence_quote":"Identifies the specific TNG50 box/peanut galaxy used to validate the $S_a$–$S_c$ signature in Figure 5."}],"review_version":1}