{"id":"b7e1ee3f-72b9-49f5-a70d-a0616e9fc024","arxiv_id":"2607.24733","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"1D Parker winds work only for weak tides; a simulation-calibrated Mixture Model of spherical plus L1/L2 nozzle escape predicts 3D mass-loss rates better across Roche-filling and thermal regimes.","lead":"3D hydro simulations show planetary atmospheric escape turns from spherical winds into two tidal tails as Roche filling and cooling increase. A calibrated Mixture Model then beats standard 1D Parker estimates for mass-loss rates across that range.","discovery_kind":"new_method","skeptic_critique":{"model":"moonshotai/kimi-k3","headline":"The Mixture Model's most dramatic application result — PW overpredicting escape by up to ~10^5 for hot, low-gravity planets — is generated by extrapolating the empirical g(λ_p)=exp(-2.47/λ_p²) correction far below the calibrated λ_p range, and the model is never tested out-of-sample.","rationale":"The reader's verdict (CONDITIONAL, correctness_risk medium) already hedges in the right place: accept the 3D regime map and 1D failure modes, treat Mixture Model coefficients as provisional pending held-out or independent tests. My stress test confirms that conditioning rather than overturning it. The reader's named weakest assumption was the thermodynamic idealization (γ=1.01, single sound speed, no wind/MHD); I agree that limits absolute-rate realism, but the authors are transparent about it and it does not threaten the internal comparison of 3D vs 1D models under identical physics. The sharper load-bearing weakness is the in-sample calibration plus the quantitatively dominant extrapolation of g(λ_p) to λ_p~0.5 in the population application, which the reader noted only as generic \"in-sample calibration (circularity ~5)\" without identifying that the headline population-level correction factors (down to 10^-5) are produced by that extrapolation specifically. Hence \"partial\" agreement. The proposed cross-validation and one additional simulation are cheap (the grid already demonstrates the pipeline) and would directly settle whether the Mixture Model's predictive claim survives out-of-sample. If both checks pass, the paper's claims stand essentially as written; if the second fails, the fix is modest — restrict §5.1 corrections to 2.5≤λ_p≤7.5 or temper the low-λ_p extrapolation — not a rejection-level revision.","tokens_in":27934,"tokens_out":2492,"duration_ms":89629,"concrete_test":"Two checks. (1) Leave-one-slice-out cross-validation: refit all five parameters using only the λ_p=2.5 and 7.5 rows of Table 1, then predict the 15 held-out λ_p=5 Ṁ_sim values; if held-out errors exceed the in-sample scatter by more than ~50%, the \"accurately predicts across parameter space\" claim is overstated. (2) Run one additional 3D simulation at λ_p≈1.0, f_ϕ≈1 (hot, weakly bound, quasi-spherical regime, well within code capability) and compare measured Ṁ_sim to the blind Ṁ_mix prediction; if the model underpredicts by more than a factor of ~2, the g(λ_p)=exp(-2.47/λ_p²) extrapolation — and hence the up-to-10^5 PW-overprediction claimed for low-gravity planets in §5.1 — is an artifact, and population corrections should be clipped to the calibrated λ_p range.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline claim that the Mixture Model \"accurately predicts mass-loss rates\" and \"outperforms\" the 1D Tidal model is established entirely in-sample: five free parameters (k_sph=2.47, k_tail=0.75, β0, β1, β2) are least-squares fit to the same 45 simulations (Table 1) against which the model is then validated in Fig. 4. That alone would be a calibration caveat. The more load-bearing problem appears in §5.1: the model is applied to observed planets with inferred λ_p from ~0.5 to 15 (only 81% lie inside the calibrated 2.5–7.5 window), and the most newsworthy correction — Ṁ_mix/Ṁ_PW as low as 10^-5 for loosely bound hot outflows (Fig. 6 bottom) — is produced mechanically by extrapolating g(λ_p)=exp(-k_sph/λ_p²) below its calibration domain. For λ_p~0.5, exp(-2.47/0.25)≈5×10^-5, i.e., the claimed orders-of-magnitude PW overprediction is exactly the value of an unvalidated exponential correction at λ_p values never simulated. Since for small λ_p the sigmoid drives w→0, Ṁ_mix≈g(λ_p)Ṁ_PW there, so the population-level takeaway for precisely the planets where escape matters most (close-in sub-Neptunes below the desert) rests on this extrapolation, not on any simulation. Nothing in the paper demonstrates that g(λ_p) retains its fitted form, or even the correct sign of the correction, for λ_p<2.5; a shallower base-flow sensitivity near the sonic point could plausibly flatten g at small λ_p. The in-grid agreement (Fig. 4) is good and the morphology story is independently supported by the sonic-surface and angular-flux diagnostics, so the concern is specifically about quantitative rate predictions outside the grid and the absence of any held-out check within it.","agreement_with_reader":"partial"},"referee_report":{"model":"moonshotai/kimi-k3","summary":"The paper presents a suite of 45 three-dimensional hydrodynamic simulations (Athena++), on a grid of Roche-lobe-filling factor f_ϕ ∈ [0.01, 5] and hydrodynamic escape parameter λ_p ∈ {2.5, 5, 7.5}, of thermal atmospheric outflows from a Jupiter-mass planet in the corotating Roche potential of a solar-mass star. The authors document a smooth morphological transition from quasi-spherical, isotropic winds (weak tides, hot outflows) to anisotropic two-tail escape through the L1/L2 channels (strong tides, cool outflows), supported by density maps, Mach-number/sonic-surface maps, and angular mass-flux distributions on the Hill sphere (Figs. 1–3). They then compare measured mass-loss rates with five analytic models — Parker wind, an empirically corrected \"Modified Spherical\" model, an L1/L2 \"Nozzle\" model, a \"Modified Tidal Two-Tail\" model with a fitted L2 attenuation, and the p-winds 1D tidal model — and combine the spherical and tidal limits into a \"Mixture Model\" (a logistic-weighted geometric mean) calibrated on the same simulation grid. Finally, they apply the Mixture Model to the observed exoplanet population using sunset-catalog outflow temperatures, finding that the Parker wind agrees with the Mixture Model for ~84% of planets but can deviate strongly for tide-dominated tightly bound systems and for hot, loosely bound ones.","tokens_in":28422,"tokens_out":4745,"duration_ms":117169,"significance":"If the simulation results hold, this is a useful and timely contribution. Hydrodynamic escape rates remain uncertain at the order-of-magnitude level in population studies, and a systematic 3D benchmark isolating tidal effects is directly relevant to interpreting the radius valley, the hot-Neptune desert, and He 1083 nm / Lyα transit observations. Notable strengths: a 45-run grid with a resolution convergence test (Appendix A) and a surface-integral Ṁ diagnostic verified to be radius-independent over 0.2–0.5a; explicit, equation-level comparisons against Parker, modified-spherical, nozzle, modified-tail, and 1D tidal (p-winds) models; and an open-source implementation of the Mixture Model with the full mass-loss table archived on Zenodo, making the results reproducible and the model falsifiable against future simulations. The morphology diagnostics (spherical-to-two-tail transition, sonic-surface deformation, angular flux maps) are robust, non-circular outputs of the hydro runs and constitute, in my view, the paper's most durable contribution. The credibility of the population-level application in §5.1, however, depends on resolving the extrapolation and in-sample-validation issues","major_comments":[{"comment":"The most dramatic application result — the bottom panel of Fig. 6 and the statement in §5.1 that 'Ṁ_mix/Ṁ_PW reaches values as low as 10^-5, implying that the Parker wind model may overestimate escape rates by several orders of magnitude' — is generated entirely by extrapolating the empirical correction g(λ_p)=exp(−k_sph/λ_p²) with k_sph=2.47 below the calibrated range. The simulated grid spans only 2.5≤λ_p≤7.5 (Table 1), over which g varies modestly (0.67–0.96). For the small-λ_p, low-η planets in question, the fitted logit (β0=−7.00, β1=6.13, β2=−4.03) drives w→0, so Ṁ_mix→g(λ_p)Ṁ_PW; at λ_p≈0.5 this gives exp(−2.47/0.25)≈5×10^-5 — i.e., the headline correction is numerically identical to the value of an unvalidated exponential at λ_p values never simulated. Worse, for λ_p<2 the isothermal Parker sonic radius r_s=R_p(λ_p/2) (§3.2) lies inside the planet, so no transonic Parker solution","section":"§5.1, Fig. 6 (bottom); Eq. (13)"},{"comment":"The central claim that the Mixture Model 'accurately predicts mass-loss rates across the parameter space explored and outperforms 1D model with tidal corrections' (Abstract; §4.4; Fig. 4) is established entirely in-sample. Five free parameters (k_sph, k_tail, β0, β1, β2) are least-squares fit to the same 45 Ṁ_sim values listed in Table 1 and then plotted against those same values in Fig. 4. With only three λ_p rows and smooth monotonic trends in f_ϕ, a five-parameter smooth interpolant is nearly guaranteed to agree, so Fig. 4 is a demonstration of fitting, not of predictive performance. A minimal fix that would substantially strengthen the paper: a hold-out test, e.g., calibrate on two of the three λ_p rows (or a subset of f_ϕ values) and predict the withheld runs. At minimum the authors should report per-point residuals, soften 'accurately predicts'/'outperforms' to reflect in-sample ca","section":"§4.1–4.4, Fig. 4, Table 2"},{"comment":"The functional form g(λ_p)=exp(−k_sph/λ_p²) is introduced without justification beyond the qualitative statement that the sonic point lies close to the surface at small λ_p. Within the grid, k_sph=2.47 is constrained almost entirely by the λ_p=2.5 row: at λ_p=5 and 7.5 the unmodified PW already under- or matches-predicts (Table 1), so the λ_p-dependence of g is essentially unconstrained by the data — yet it is this unconstrained λ_p-dependence that dominates the low-λ_p extrapolation behind Major Comment 1. The same concern applies, more mildly, to h(λ_p)=exp(−k_tail/Ro²): the claimed Ro^-2 scaling of the L2 suppression is motivated by a physical argument (§4.2) but the functional form is one of many consistent with the grid. Please report the fit residuals as a function of λ_p for both corrections, and either demonstrate that alternative forms (e.g., g constant, g∝exp(−k/λ_p)) are disfa","section":"§4.1, Eq. (13); §4.2, Eq. (17)"}],"minor_comments":[{"comment":"First data row: for (f_ϕ=0.010, λ_p=2.50), Ṁ_sim is listed as 1.00×10^-4 while the adjacent rows are ~9.9×10^-4 and Ṁ_mix=9.86×10^-4 exceeds it by an order of magnitude. This appears to be a typo for 1.00×10^-3; since Table 1 is the benchmark dataset (and is distributed on Zenodo), please verify.","section":"Table 1"},{"comment":"The text twice refers to 'Figure 3 (blue curve)' and 'Figure 3, left' where Figure 7 is clearly meant; the Fig. 7 caption also says 'shown in Figure 3 (blue curve)'. Please correct the cross-references.","section":"§5.2 / Fig. 7 caption"},{"comment":"In computing c_s=√(γk_BT/μm_p) for observed planets, γ=5/3 is adopted, whereas the simulations use γ=1.01. The mapping of a sunbather/sunset temperature onto an effective λ_p is therefore convention-dependent at the factor-of-~1.6 level in c_s², which shifts planets appreciably in λ_p. The qualitative caveat is noted, but a sentence quantifying the sensitivity (or showing the population statistics of §5.1 for both conventions) would be useful.","section":"§5.1"},{"comment":"The convergence test is shown for a single (unstated) model. Since the smallest effective planet is resolved by only ~11 zones across its diameter, please state which (f_ϕ, λ_p) case is tested and, ideally, show convergence for the most extreme case (smallest R_eff, e.g., f_ϕ=0.01, λ_p=7.5), where resolution effects should be strongest.","section":"Appendix A / Fig. 8"},{"comment":"Fig. 4 would benefit from a quantitative summary (RMS and maximum log-space residual per model), particularly to substantiate the Table 2 entry 'Matches simulations' for the Mixture Model relative to the ~1.4–1.6× systematics quoted for the 1D Tidal Model.","section":"Fig. 4"},{"comment":"The nozzle curvature values (ϕ^1_yy=3.4538, etc.) are specific to q=M_p/M_⋆=10^-3; a brief note of how they scale with mass ratio (or a pointer to an analytic formula) would help users of the released code apply the model to other systems.","section":"§4.2, Eq. (15)"},{"comment":"Uncertainties are reported for β0, β1, β2 but are not propagated into the population-level ratios Ṁ_mix/Ṁ_PW in Figs. 5–6. Even an indicative uncertainty band on the ratio for a few representative planets would clarify which of the deviant systems are significant.","section":"§4.4 / §5.1"},{"comment":"The Abstract phrase 'outperforms 1D model with tidal corrections' needs an article ('the 1D model') and, per Major Comment 2, an in-sample qualification.","section":"Abstract (typography)"}],"recommendation":"major_revision","confidential_remarks":"The simulation component of this paper is solid and well within the journal's scope, and the open code/data release is commendable. My main reservation is that the population-level conclusions in §5.1 — the part most likely to be quoted — are partially driven by unvalidated extrapolation of a fitted correction factor, and the headline \"outperforms\" claim is an in-sample statement. Both are fixable within the manuscript's scope (either by additional low-λ_p runs, which the authors' machinery can produce, or by honest re-scoping of the application figures and language). I therefore see this as a potentially strong contribution after one substantive revision round rather than a fundamental problem."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The real contribution here is the controlled 45-run Athena++ grid that isolates Roche filling (f_ϕ) from thermal binding (λ_p) with no stellar wind. That cleanly shows when outflows stay quasi-spherical versus when they collapse into L1/L2 tails, and it documents where plain Parker and even 1D-tidal rates go wrong. Morphology, sonic surfaces, and angular flux on the Hill sphere (Figs. 1–3) are directly supported by the hydro and look solid; the resolution check in Appendix A is adequate.\n\nWhat is new is not the existence of anisotropic Roche escape—that was already in Bourrier, Debrecht, MacLeod, Nail, etc.—but the quantitative side-by-side against Parker, nozzle, and 1D-tidal formulas, plus a public Mixture Model that stitches a modified spherical limit to a modified two-tail limit. Inside the simulated box the fit is good and beats 1D-tidal on integrated Ṁ (Fig. 4, Table 2). Code and table are shipped; that helps.\n\nSoft spots are real but localized. Five free parameters (k_sph, k_tail, three βs) are least-squares fit to the same 45 Ṁ_sim values used as the benchmark, so “outperforms” is in-sample. More importantly, the dramatic population claim—Ṁ_mix/Ṁ_PW down to ~10^{-5} for hot, low-gravity planets—comes from pushing g(λ_p)=exp(-2.47/λ_p²) well below the calibrated floor λ_p=2.5. At λ_p~0.5 the exponential alone supplies the orders-of-magnitude suppression, and the paper never shows that functional form (or even its sign) holds there. The isothermal γ=1.01, fixed-ρ_s, pure-hydro setup is an intentional idealization; absolute rates and the λ_p mapping for real planets inherit it. The authors flag the validation window, but the §5.1 figures still lean on the extrapolation.\n\nFor people who need geometry-aware escape rates or who want a clean benchmark of when 1D fails, this is worth reading and citing for the regime map and the failure modes. Treat the Mixture coefficients as provisional outside 2.5–7.5. I would send it to referees; the 3D results and the honest comparison to 1D are enough to deserve the time.","headline":"Solid 3D regime map of tidal vs thermal escape; the Mixture Model is a useful in-grid fit but its flashiest population corrections rest on unvalidated extrapolation of g(λ_p).","tokens_in":28978,"tokens_out":604,"would_cite":true,"duration_ms":11914,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"A mixture of spherical and tidal-tail escape formulas, trained on 3D runs, beats 1D Parker models for planetary mass-loss rates.","keywords":["atmospheric escape","photoevaporation","core-powered mass loss","Roche lobe","Parker wind","3D hydrodynamics","exoplanet mass loss","tidal tails"],"falsifier":"Run the same 3D setup at a handful of points inside the grid with a realistic energy equation (photoionization heating, radiative cooling, multi-species chemistry) and check whether the Mixture Model still recovers the measured mass-loss rates to the same accuracy.","tokens_in":28579,"feed_emoji":"🪐","tokens_out":922,"duration_ms":17807,"temperature":0.7,"pith_summary":"Close-in exoplanets lose atmosphere through thermal winds, and almost every population study still uses one-dimensional Parker-type formulas that assume the flow is spherical. This paper runs a grid of three-dimensional hydrodynamic simulations of a Jupiter-sized planet in a solar-type star’s tidal field, varying how full the Roche lobe is and how hot the wind is. The runs show a clean morphological change: weak tides and hot winds stay nearly spherical, while strong tides and cooler winds funnel into two tails through the L1 and L2 points. Standard Parker winds only match the simulations in the weak-tide limit; even a tidally corrected 1D model gets the total mass-loss rate roughly right but cannot capture the angle-dependent structure. The authors therefore build a Mixture Model that interpolates between a spherical formula and a two-nozzle formula, calibrate it on the 3D grid, and show that it recovers the simulated rates across the whole parameter space better than existing 1D tools.","feed_headline":"3D winds beat 1D Parker models for exoplanet mass loss","feed_subtitle":"A mixture of spherical and tidal-tail formulas, trained on simulations, recovers rates across all regimes","key_machinery":"The Mixture Model: a weighted geometric mean of a modified spherical Parker wind and a modified tidal two-tail (nozzle) rate, with the weight set by a sigmoid of the hydrodynamic escape parameter and the dimensionless L1 barrier height, calibrated directly on the 3D simulation grid.","core_discovery":"Three-dimensional hydrodynamic simulations of thermally driven planetary outflows demonstrate that mass-loss morphology and rate transition smoothly from nearly isotropic Parker-like winds (weak tides, hot atmospheres) to anisotropic two-tailed streams through the Lagrange points (strong tides, cooler atmospheres). A calibrated Mixture Model that geometrically averages a modified spherical Parker rate and a modified L1/L2 nozzle rate reproduces the simulated mass-loss rates across the explored grid and outperforms both pure Parker and tidally corrected 1D models.","pith_inferences":["Coupling the Mixture Model into long-term orbital and structural evolution codes would change predicted lifetimes most for the hottest, lowest-mass close-in planets and for near-Roche-lobe gas giants.","Transit spectroscopy of leading versus trailing tails could observationally flag the strong-tide regime where the Mixture weight approaches unity.","Extending the calibration grid below λ_p ≈ 2.5 and above ≈ 7.5 would immediately widen the model’s safe domain for super-Earths and cooler giants."],"forward_implications":["Most observed close-in planets lie where Parker and Mixture rates differ by less than 30 percent, so existing demographic calculations remain first-order safe.","A minority of strongly Roche-filling, tightly bound systems can have true mass-loss rates several times higher than pure Parker estimates.","A larger minority of hot, loosely bound small planets can have true rates far lower than pure Parker estimates, affecting radius-valley and desert models.","Mean radial density profiles from 1D tidal codes remain usable even when the flow is anisotropic; only the angular structure is lost."],"fun_headline_variants":["3D sims show tides turn exoplanet winds from spheres into two-tailed streams","Mixture model beats 1D Parker and tidal corrections for mass-loss rates","Weak tides yield isotropic Parker winds; strong tides form L1/L2 tails","1D Parker holds only for weak tides; 3D structure demands mixture rates","Calibrated spherical-plus-nozzle mix recovers 3D escape rates across regimes"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The entire thermodynamic state of the atmosphere is collapsed to a single nearly isothermal sound speed with fixed surface density and no stellar wind, magnetic fields, or detailed heating and cooling.","fun_headline_variants_meta":{"raw":{"variants":["3D sims show tides turn exoplanet winds from spheres into two-tailed streams","Mixture model beats 1D Parker and tidal corrections for mass-loss rates","Weak tides yield isotropic Parker winds; strong tides form L1/L2 tails","1D Parker holds only for weak tides; 3D structure demands mixture rates","Calibrated spherical-plus-nozzle mix recovers 3D escape rates across regimes"]},"model":"grok-4.5","effort":"low","cost_usd":0.004237,"raw_usage":{"total_tokens":1344,"prompt_tokens":848,"num_sources_used":0,"completion_tokens":90,"cost_in_usd_ticks":42368000,"prompt_tokens_details":{"text_tokens":848,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":406,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":848,"tokens_out":90,"duration_ms":10161,"temperature":1.0,"reasoning_tokens":406,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-31T06:29:10.337809+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Run the same 3D setup at a handful of points inside the grid with a realistic energy equation (photoionization heating, radiative cooling, multi-species chemistry) and check whether the Mixture Model still recovers the measured mass-loss rates to the same accuracy.","supporting_citations":[],"review_version":1}