{"id":"48bcf851-b4a8-4b42-b7bc-969f0b2bc358","arxiv_id":"2504.20807","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Global weak solutions of the 3D compressible semi-geostrophic equations exist for arbitrary compactly supported measure-valued initial data, obtained by semi-discrete optimal transport and particle discretization.","lead":"Mathematicians proved that the 3D compressible semi-geostrophic equations, a simplified model of large-scale atmospheric flows, admit global-in-time weak solutions for any compactly supported measure-valued initial data. The proof uses a particle discretization and semi-discrete optimal transport, and it provides a theoretical foundation for numerical schemes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 5.3 asserts a uniform W^{1,∞} bound on σ*[β^N] that is not established and may fail; the proof of Theorem 1.2 as written relies on it, although a weak-convergence replacement may repair the gap.","rationale":"The reader's CONDITIONAL verdict is appropriate. The most load-bearing weak spot is not Lemma 5.2: the transfer of compactness to the compressible ODE is mostly a combination of the a priori estimates in Lemma 4.6 (bounded velocities and supports) and an Arzelà-Ascoli argument in W1, so the quoted incompressible result should carry over. The genuinely new and under-supported step is the uniform W^{1,∞} bound on σ*[β^N] in Lemma 5.3. It is asserted in one sentence, with no proof that w*(z) is bounded uniformly in N; the scaling calculation above shows that the internal-energy term F does not automatically provide such a bound for γ<2. The theorem may still be true because the limit passage only needs weak convergence of the source measures, but that weaker argument is absent from the paper. Therefore the verdict remains CONDITIONAL, with the condition being a correct proof of Lemma 5.3 or a replacement weak-convergence argument in Section 5.","tokens_in":32392,"tokens_out":17277,"duration_ms":192829,"concrete_test":"Take X=[0,1]^3, γ=3/2, κ=1, and for each N choose a well-prepared β^N with N particles in a fixed compact K⊂Y whose masses and seeds are arranged so that one c-Laguerre cell has volume ∼ m_i^2 while carrying mass m_i (for example, place N-1 seeds in a tight cluster and one seed slightly separated vertically). Solve the finite-dimensional dual problem (3.1) numerically for the optimal w*(z), and compute sup_N ∥σ*[β^N]∥_{W^{1,∞}}. If this quantity diverges, Lemma 5.3 as stated is false and the §5 proof needs the weak-convergence replacement; if it stays bounded for the tested sequences, the uniform bound is at least plausible.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central limit passage in Theorem 1.2 uses Lemma 5.3 to justify convergence of the source term. The key step in Lemma 5.3 is the claim that the optimal source measures σ*[β^N] are uniformly bounded in W^{1,∞}, via the assertion that max_{z∈D^N∩K^N} ∥ζ_i(·,w*(z),z)∥_{C^1(X)} is independent of N. This reduces to a uniform bound on the optimal dual weights w*(z). The mass balance (3.2), m_i=∫_{L_i^c(w*(z),z)}(f*)'(w_i-c(x,z_i))dx, does not by itself give such a bound: a cell of arbitrarily small Lebesgue measure can carry a positive mass, forcing w_i and hence σ*=(f*)'(-w^c) to be arbitrarily large. The internal-energy penalty does not prevent this for γ∈(1,2), since a density peak of height 1/δ on a set of measure δ^2 has ∫σ^γ ∼ δ^{2-γ} → 0. Consequently the displayed maximum in the proof of Lemma 5.3 is not justified solely by Proposition 3.8 and Corollary 3.9. Since Theorem 1.2 uses this lemma to pass from γ[α^N_t] to γ[α_t] in (5.2)–(5.4), the proof as written has a real gap at this point. This is not necessarily fatal: weak convergence of σ*[α^N_t] would suffice for the limit, and that weaker statement can be obtained by a continuity/Γ-convergence argument for E without the uniform W^{1,∞} bound. But the paper does not provide that alternative argument.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves global-in-time existence of weak solutions of the three-dimensional compressible semi-geostrophic equations in geostrophic coordinates for arbitrary compactly supported measure-valued initial data. The proof strategy is to discretize the initial measure by N weighted particles, derive a concave dual problem for the internal-energy-regularized optimal transport problem (Theorem 3.4), prove that the resulting finite-dimensional ODE (1.6) has a unique global C^2 solution (Theorem 1.1), and then pass to the limit N to infinity using compactness of the discrete solutions and stability of the optimal source measures (Theorem 1.2). The paper also contains two explicit examples: a steady state solution and an elliptic orbit solution for a single particle with gamma=2.","tokens_in":32706,"tokens_out":12713,"duration_ms":136222,"significance":"If correct, Theorem 1.2 significantly extends earlier existence results for the compressible semi-geostrophic system, which were limited to L^r initial data, and it provides a theoretical foundation for numerical schemes based on semi-discrete optimal transport. The paper is particularly strong on the variational side: the duality theorem, the regularity of Laguerre-cell integrals in Lemma 3.6, the differentiability of the centroid map, and the energy-conservation argument are treated in detail. The two explicit examples are useful consistency checks. The main fragility is the limit passage in Section 5, which depends on two compactness statements that are either imported from [7] or proved via an under-justified uniformity claim; both appear repairable by short arguments.","major_comments":[{"comment":"The proof of Lemma 5.3 asserts that the maximum of ||zeta_i(.,w*(z),z)||_{C^1(X)} over z in D^N cap K^N is independent of N, and this uniformity is load-bearing for the pointwise convergence sigma*[alpha_t^N] -> sigma*[alpha_t] used in Theorem 1.2. As written, the assertion is not justified: equation (3.2) alone does not bound w_i, since a positive mass could in principle be supported on a Laguerre cell of arbitrarily small measure, forcing w_i and the density to be large. However, the missing uniform bound is true and can be supplied by a short argument: mass balance forces w_i > min_{X x K} c, because otherwise (f*)'(w_i - c(x,z_i)) vanishes on the whole i-th cell; and the coercivity estimate in the proof of Theorem 3.4, together with G(w*,z) >= G(0,z) and the boundedness of c on X x K, gives a uniform upper bound on the positive part of w*. The authors should add this argument; without it, Lemma 5.3 and the limit passage in (5.4) rest on an unproved claim.","section":"Lemma 5.3 (p. 19-20)"},{"comment":"Lemma 5.2 is imported verbatim from [7, Lemma 5.2] even though the compressible ODE (1.6) and the centroid map C differ from the incompressible setting, and the lemma provides the uniform convergence (5.1) on which Theorem 1.2 depends. The authors should either prove the lemma in the present setting (a short Arzela-Ascoli argument using the a priori estimates of Lemma 4.6 and the uniform Lipschitz bound in the W1 metric) or state explicitly which hypotheses of [7, Lemma 5.2] are verified by the compressible system. As it stands, this is a load-bearing gap in the exposition.","section":"Lemma 5.2 (p. 19)"}],"minor_comments":[{"comment":"There are small typographical errors: 'seee (5.1)' in equation (5.3) and 'discrete discrete' in the statement of Lemma 5.2; these should be corrected.","section":"Eq. (5.3) and Lemma 5.2 statement"},{"comment":"The bibliographic entry for [43] is corrupted: it concatenates the title of Villani's book with the title of [44]. This should be fixed.","section":"Reference [43]"},{"comment":"The verification that the constructed steady state is a weak solution cancels the two integrals in Definition 2.6, but this cancellation is only implicit; spelling it out would improve readability.","section":"Proposition 6.1"},{"comment":"The passage from weak convergence of the marginals to weak convergence of the optimal plans gamma[alpha_t^N] cites [43, Theorem 5.20]; the authors should add one sentence explaining why the cited theorem applies, in particular why the limit optimal plan is unique (absolute continuity of sigma*[alpha_t] and the twist condition).","section":"Proof of Theorem 1.2, p. 20"}],"recommendation":"major_revision","confidential_remarks":"I believe the main theorem is correct and the flagged issues are repairable without changing the argument's architecture. The key missing piece is a short proof of the uniform bound on w* in Lemma 5.3, and a proof or precise transfer statement for Lemma 5.2. I did not find circularity or hidden assumptions: the self-citation [7] is used only for standard compactness and quantization statements that are independent of the novel compressible results. I would be happy to see a revised version."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, this is a real extension: arbitrary compactly supported measure-valued initial data for the compressible semi-geostrophic equations, where prior work required L^r densities. The semi-discrete optimal transport route from [7] does carry over, but not for free—the non-standard cost and the time-dependent optimal source measure force new duality and regularity results, and the paper supplies them. Second, the main theorem is probably right, but one lemma in the limit passage is not proved as written. Lemma 5.3 asserts uniform W^{1,∞} convergence of the optimal source measures σ*[β^N]; the proof claims the dual weights are uniformly bounded because a certain maximum is independent of N. That reduction is not justified by mass balance alone: a small cell can carry positive mass and force the local weight to grow, and the internal energy penalty does not preclude that for γ∈(1,2). So the stated Arzelà-Ascoli argument has a gap.\n\nWhat is good: the duality theorem (Theorem 3.4) and the regularity of the centroid map are worked out carefully. The a priori estimates and energy conservation for the discrete ODE are clean. The paper gives explicit steady-state and single-seed solutions, and the appendices are serious. The citations to [7] for quantization and compactness are appropriate; the new results are not obtained by assuming the target.\n\nOn the soft spot: the gap is in a supporting lemma, not the core strategy. The paper itself notes that weak convergence of source measures would suffice for the limit passage, and a Γ-convergence or continuity argument for E should supply that without the uniform W^{1,∞} bound. So I would flag this for the authors to fix rather than reject. Minor caveat: the domain assumption (A.1) should be stated in Theorem 1.2 to avoid overclaiming generality.\n\nThis paper is for people working on existence theory for the SG equations and for anyone wanting a template for semi-discrete optimal transport with non-quadratic cost. It deserves a serious referee, and the referee should carefully check Lemma 5.3 and the transfer of Lemma 5.2 from [7] to the compressible ODE. My recommendation: send to peer review, with a request for a revised proof of Lemma 5.3 or a replacement weak-convergence argument.","headline":"Novel measure-valued existence result for compressible SG equations, mostly solid, but Lemma 5.3 has an unproved uniform W^{1,∞} bound that needs a repair.","tokens_in":33305,"tokens_out":4017,"would_cite":true,"duration_ms":42300,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35Q35","49Q22","35D30","86A10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that the three-dimensional compressible semi-geostrophic equations admit global weak solutions for any compactly supported probability-measure initial data, obtained as a uniform limit of particle discretisations.","keywords":["semi-geostrophic equations","compressible fluids","optimal transport","semi-discrete optimal transport","weak solutions","measure-valued initial data","Laguerre tessellations","particle discretisation"],"falsifier":"A concrete way to test the central claim is to construct a sequence $\\beta^N\\to\\beta$ in the $W_1$ metric for which the optimal source densities $\\sigma^*[\\beta^N]$ have unbounded $W^{1,\\infty}$ norm; that would contradict Lemma 5.3 and remove the equicontinuity used to pass the energy term to the limit. Alternatively, a numerical solution of the particle ODE showing $\\sup_t W_1(\\alpha^N_t,\\alpha_t)$ not tending to zero would falsify the approximation statement of Theorem 1.2.","tokens_in":32152,"feed_emoji":"🌀","tokens_out":15810,"duration_ms":140914,"temperature":0.7,"pith_summary":"The paper establishes that the three-dimensional compressible semi-geostrophic equations, a model for large-scale atmospheric front formation, have global-in-time weak solutions for any compactly supported probability measure as initial data, not only for absolutely continuous densities of class $L^r$. The route is a particle discretisation: with $N$ weighted particles the equations reduce to a finite-dimensional ODE whose coefficients are fixed by semi-discrete optimal transport, and taking $N$ to infinity produces a weak solution as a uniform limit in the $W_1$ Wasserstein metric. The discrete solutions are unique, twice differentiable, and conserve the geostrophic energy, so the construction also gives a theoretical basis for numerical schemes that solve the equations by optimal-transport-based particle methods.","feed_headline":"Atmosphere flow model solved for any compact initial data","feed_subtitle":"Particle discretisation plus optimal transport broaden solvability to all compactly supported measure data.","key_machinery":"The engine is the discrete ansatz $\\alpha^N_t=\\sum_{i=1}^N m_i\\delta_{z_i(t)}$ and the finite-dimensional ODE $\\dot z=J_N(z-C(z))$, where the centroid map $C(z)$ assigns to each seed the centroid of the corresponding cell in the c-Laguerre tessellation of the optimal source measure $\\sigma^*[\\alpha^N_t]$. A c-Laguerre cell is the set of points $x\\in X$ satisfying $c(x,z_i)-w_i\\le c(x,z_j)-w_j$ for all $j$; the cost $c(x,y)=\\frac{1}{y_3}\\left(\\frac{f_{\\mathrm{cor}}^2}{2}((x_1-y_1)^2+(x_2-y_2)^2)+gx_3\\right)$ is twisted, so the optimal transport map from the source measure to the discrete target is unique and piecewise constant on these cells. A concave dual functional $G(w,z)$ characterises both the source density via $\\sigma(x)=(f^*)'(-\\min_i(c(x,z_i)-w_i))$ and the optimal weights, and an implicit-function argument makes $C$ continuously differentiable; a priori estimates keep the particles bounded and in distinct horizontal planes. In the limit, the imported compactness lemma and the equicontinuity of the optimal source measures pass the discrete solutions to a weak solution.","core_discovery":"The paper's central result, Theorem 1.2, is that for every final time $\\tau>0$ and every compactly supported probability measure $\\alpha\\in P_c(Y)$ on the geostrophic domain, the compressible semi-geostrophic equations admit a weak solution $\\alpha_t\\in C^{0,1}([0,\\tau];P_c(Y))$ with $\\alpha_0=\\alpha$, in the sense of Definition 2.6. The weak formulation replaces the formal wind $W[\\alpha_t]=J(\\mathrm{id}-T_{\\alpha_t}^{-1})$ by an expression that only uses the optimal transport map $T_{\\alpha_t}$ from the energy-minimising source measure $\\sigma^*[\\alpha_t]$ to $\\alpha_t$. The proof constructs discrete weak solutions $\\alpha^N_t$ from the particle ODE and shows $\\sup_{t\\in[0,\\tau]}W_1(\\alpha^N_t,\\alpha_t)\\to0$; at the discrete level the solution is unique, $C^2$, and conserves $E(\\sigma^*[\\alpha^N_t],\\alpha^N_t)$.","pith_inferences":["One likely next step is to replace the vertical term $gx_3$ in the cost by the more general vertical dependence treated in the original theory; the main obstacle would be preserving the convexity and regularity structure used to differentiate the centroid map.","The paper's equi-Lipschitz bound on optimal source measures implies a stability property it does not state explicitly: the optimal source density is determined continuously from discrete approximations, so the source field cannot develop arbitrarily sharp fronts as the data varies in the $W_1$ metric.","A concrete numerical test of the construction would be to solve the concave dual maximisation and the particle ODE with a standard semi-discrete optimal-transport solver and compare energy drift against the predicted conservation; the paper stops at existence."],"forward_implications":["Compactly supported measure-valued initial data, including sums of point masses, are admissible in three dimensions; the solution exists for every finite time horizon $\\tau>0$.","The discrete approximations $\\alpha^N_t$ converge uniformly to the weak solution in the $W_1$ metric, so particle approximations genuinely represent the continuum dynamics.","For discrete initial data the solution is unique, $C^2$, and global in time, so the particle ODE is not merely a formal reduction.","The discrete solution conserves the geostrophic energy $E(\\sigma^*[\\alpha^N_t],\\alpha^N_t)$, giving numerical schemes an exact invariant to monitor."],"supporting_citations":[{"why":"Formulates the compressible SG equations in geostrophic variables as a transport equation and supplies the original existence theorem for $L^r$ initial data that this paper generalises.","marker":"[18]"},{"why":"Supplies the semi-discrete particle-discretisation strategy and the quantization and compactness lemmas reused in the limit passage.","marker":"[7]"},{"why":"Provides the c-Laguerre tessellation theory and the dual characterisation of the optimal transport map used to define the centroid map.","marker":"[38]"},{"why":"Supplies the underlying optimal-transport facts: twist condition, existence and uniqueness of transport maps, duality, and continuity of transport cost.","marker":"[39]"},{"why":"Gives the c-convexity assumptions and the convex-cell transformation that make the regularity argument for Laguerre cell integrals work.","marker":"[33]"},{"why":"Supplies the regularity theory for classical Laguerre cell volumes that underpins the differentiability of the centroid map in Appendix B.","marker":"[20]"}],"fun_headline_variants":["Optimal transport unlocks 3D atmospheric flow for any compact data","Particle method proves global flow solutions for all compact measures","Compressible SG equations now solvable for any compact measure","Global weak solutions for 3D atmospheric flow from any compact data","Semi-discrete optimal transport resolves compressible SG equations"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The compactness lemma that guarantees a convergent subsequence of discrete solutions (Lemma 5.2) is imported without proof from the incompressible case, and the whole limit passage depends on it holding for the compressible ODE with the new centroid map; if it failed, no limiting weak solution would be constructed.","fun_headline_variants_meta":{"raw":{"variants":["Optimal transport unlocks 3D atmospheric flow for any compact data","Particle method proves global flow solutions for all compact measures","Compressible SG equations now solvable for any compact measure","Global weak solutions for 3D atmospheric flow from any compact data","Semi-discrete optimal transport resolves compressible SG equations"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000874,"raw_usage":{"total_tokens":3777,"prompt_tokens":933,"completion_tokens":2844,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":549,"completion_tokens_details":{"reasoning_tokens":2758}},"tokens_in":549,"tokens_out":2844,"duration_ms":17772,"temperature":1.0,"reasoning_tokens":2758,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T05:19:40.508355+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete way to test the central claim is to construct a sequence $\\beta^N\\to\\beta$ in the $W_1$ metric for which the optimal source densities $\\sigma^*[\\beta^N]$ have unbounded $W^{1,\\infty}$ norm; that would contradict Lemma 5.3 and remove the equicontinuity used to pass the energy term to the limit. Alternatively, a numerical solution of the particle ODE showing $\\sup_t W_1(\\alpha^N_t,\\alpha_t)$ not tending to zero would falsify the approximation statement of Theorem 1.2.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Formulates the compressible SG equations in geostrophic variables as a transport equation and supplies the original existence theorem for $L^r$ initial data that this paper generalises."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the semi-discrete particle-discretisation strategy and the quantization and compactness lemmas reused in the limit passage."},{"cited_title":"M´ erigot and B","cited_arxiv_id":null,"evidence_quote":"Provides the c-Laguerre tessellation theory and the dual characterisation of the optimal transport map used to define the centroid map."},{"cited_title":"Santambrogio","cited_arxiv_id":null,"evidence_quote":"Supplies the underlying optimal-transport facts: twist condition, existence and uniqueness of transport maps, duality, and continuity of transport cost."},{"cited_title":"Kitagawa, Q","cited_arxiv_id":null,"evidence_quote":"Gives the c-convexity assumptions and the convex-cell transformation that make the regularity argument for Laguerre cell integrals work."}],"review_version":1}