{"id":"bcdafde2-1231-4f38-9fda-ba77649c4059","arxiv_id":"2505.24602","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A Stefan problem with a nonlocal radiative term in the solid is shown to admit a unique local-in-time solution, and a global-in-time solution for a class of bounded initial data.","lead":"The paper proves local and global well-posedness for a one-dimensional, two-phase Stefan problem in which heat moves by conduction in both phases and by radiation inside the solid. The radiation is reduced to a nonlocal integral term, and the proofs combine a Banach fixed-point argument with parabolic regularity and maximum principles.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3.1 relies on an asserted but omitted regularity bootstrap: the differentiated equation for ∂yT2 is used to bound the derivative, yet the needed C^{1+δ/2,3+δ} regularity is dismissed with \"we omit the details.\"","rationale":"The reader's weakest_assumption correctly identifies the bootstrap regularity as the primary gap, and I agree that the global theorem hinges on it. I do not, however, treat the stationary RTE approximation as an equally load-bearing concern: the paper's headline claim is well-posedness of the resulting nonlocal Stefan problem, and the passage from the kinetic model to (1.9) is a standard quasi-static reduction, with the 1/c term neglected explicitly and consistently. Within the PDE argument, the local fixed-point proof and the maximum-principle structure are mostly explicit and coherent; the unresolved step is the asserted C^{1+δ/2,3+δ} bootstrap, which is exactly where the nonlocal term must be differentiated. The comparison arguments for ∂yT2 would be valid if that bootstrap were supplied, so the appropriate verdict remains conditional rather than accept or reject. Minor typos and notational slips (for example the Γ± formula missing the TM denominator and the L3/L2 label) do not affect the central concern.","tokens_in":1803,"tokens_out":1649,"duration_ms":244463,"concrete_test":"Provide a complete proof that the local solution of Theorem 2.2 satisfies Ti ∈ C^{1+δ/2,3+δ}_{t,y}([ε,t*] × R±) for every ε>0, by deriving the equation for the difference quotient v_h = [T2(t,y+h)−T2(t,y)]/h and passing to the limit, with explicit estimates for the term 4∫0∞ α/2 E1(α|y−η|) T2^3(η) v_h(η)dη and the boundary term α/2 T_M^4 E1(αy). If the limiting v is only a weak solution, check whether the comparison principle for L2^1(v) can be run directly on difference quotients; otherwise the derivative bound used to extend the solution in Theorem 3.1 is not justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The global well-posedness proof in Section 3 needs a uniform bound on ∂yT2 to extend the local solution past t*. To obtain this bound, the proof differentiates the T2 equation and applies a maximum-principle comparison for v=∂yT2 against φ±. But differentiating the nonlocal term Iα[T2] requires more regularity than is established in Theorem 2.2. The text states: \"One can prove that Ti ∈ C^{1+δ/2,3+δ}_{t,y}([ε,t*] × [±γ,±∞)) ... Since the computations are similar to the one in Proposition 2.1, Lemma 2.1 and Lemma 2.2 we omit the details.\" This omitted bootstrap is load-bearing: without a strong equation for v, the comparison argument for ∂yT2 has no object to act on, and without the resulting ∂yT2 bound there is no continuation past t*, so Theorem 3.1 is not established. The nonlocal term is not a cosmetic difficulty: its y-derivative contains a kernel convolution with T2^3 v plus a boundary contribution from η=0, namely α/2 T_M^4 E1(αy), and one must verify the integration-by-parts identity and the Hölder regularity of the resulting source before Schauder or maximum-principle arguments can be applied. The fixed-point local theory appears internally coherent, and the modelling reduction to the stationary RTE is standard for this kind of quasi-static approximation; the unresolved point is the global regularity step.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies a one-dimensional two-phase Stefan problem with radiative transfer in the solid phase. After a quasi-static reduction of the radiative transfer equation, the model becomes a parabolic free-boundary system with a nonlocal operator I_alpha[T] = T^4 - integral (alpha/2) E1(alpha|y-eta|) T^4 deta. The authors prove local existence and uniqueness for the free-boundary problem by a Banach fixed-point argument in a C^{0,1} space, then upgrade the solution to classical Holder regularity, and finally prove a global well-posedness result for a class of initial data satisfying an explicit smallness condition on the liquid temperature and a positivity condition on the solid temperature. The global theorem is proved by constructing explicit stationary sub- and supersolutions and applying maximum-principle comparisons to the temperature and its spatial derivative.","tokens_in":38831,"tokens_out":9461,"duration_ms":118804,"significance":"If completed, this would be the first rigorous well-posedness theory for this Stefan-radiation model, and the explicit barrier construction gives a quantitative, checkable global existence condition. The local fixed-point argument is written in enough detail to be verifiable, and the reduction from the radiative transfer equation to the nonlocal parabolic operator is standard and correctly derived. The main limitation is that the global theorem currently rests on an asserted but omitted regularity bootstrap; that bootstrap is load-bearing for the derivative comparison argument, so the paper is not yet self-contained in its claimed form.","major_comments":[{"comment":"The global theorem is not established as written because the proof of the uniform bound on derivative_y T_2 depends on a regularity bootstrap that is asserted but not proved. The text states: \"One can prove that T_i in C^{1+delta/2,3+delta}_{t,y}([epsilon,t*] x [gamma, infinity)) ... Since the computations are similar to the one in Proposition 2.1, Lemma 2.1 and Lemma 2.2 we omit the details.\" This is not a routine repetition: differentiating the nonlocal term I_alpha[T_2] produces a convolution of T_2^3 derivative_y T_2 with the kernel E_1 and a boundary term proportional to T_M^4 E_1(alpha y). Without a rigorous derivation of the equations L^1_2(derivative_y T_2) = (alpha/2) T_M^4 E_1(alpha y) and L^2_2(derivative_y T_2) = 0, the subsequent maximum-principle comparisons for derivative_y T_2 have no justified object to act on. Since the boundedness of derivative_y T_2 is exactly what allows extension past the local existence time t*, Theorem 3.1 is incomplete without a detailed proof of this bootstrap.","section":"Section 3, proof of Theorem 3.1"},{"comment":"In the convergence step for the approximating problems, the text states that the limiting integral equations hold \"where a_1 = kappa and a_2 = 0.\" This is inconsistent with the representation (2.4) and with the equation (2.22), where the coefficient in front of the second spatial derivative for u_2 is 1, not 0. The value should be a_2 = 1. As written, the limiting equation for u_2 would involve a heat kernel with zero diffusivity, which cannot be the intended statement. This appears to be a typographical error, but it should be corrected because Eq. (2.36) is the basis for identifying the limit with the fixed-point solution.","section":"Section 2, proof of Lemma 2.3, Eq. (2.36)"}],"minor_comments":[{"comment":"The statement says \"the temperature satisfies T_0(x) > T_M if x < s(t), T_0(s(t)) = T_M, T_0(x) < T_M if x > s(t)\"; the symbol T_0 should be T(t,x), since the inequalities are for the evolved temperature, not the initial data.","section":"Theorem 1.1"},{"comment":"After constructing the constants C_1, C_2, the text says \"we have found constants C_1, C_2 > 0 such that w(y) < T_0(y) for y < 0 and T_0(y) > 0 for y > 0.\" This is the reverse of the required inequalities. The preceding construction shows that T_0(y) < w(y) for y < 0 and T_0(y) > w(y) for y > 0, so this sentence should be corrected.","section":"Section 3, proof of Lemma 3.1"},{"comment":"The formula for Gamma_plus/minus(C_1) is written with C_2 inside the square root: \"Gamma_plus/minus(C_1) = (L C_1 +/- sqrt(L^2 C_2? 1 - T_M^5))/2.\" The quantity under the square root should depend on C_1^2, not on C_2, as is clear from the later derivation of the quadratic inequality 4 C_2^2 T_M - 4 L C_1 C_2 + T_M^4 < 0.","section":"Section 3, Theorem 3.1 and its proof"},{"comment":"In the statement of the lemma, the Holder regularity of u_2 is written as \"u_1 in C^{delta/2,1+delta}_{t,y}([0,t*] x [0,R])\"; this should be u_2. The same typo appears in the following line for u_1.","section":"Lemma 2.3"},{"comment":"Theorem 1.2 states only that a \"large class\" of initial data yields global well-posedness, while Theorem 3.1 gives explicit conditions. The introductory statement should refer to the explicit conditions in Theorem 3.1 so that the reader does not have to locate them later.","section":"Theorem 1.2 vs. Theorem 3.1"}],"recommendation":"major_revision","confidential_remarks":"The local well-posedness part appears sound and is presented in sufficient detail to be checked. The global theorem is the advertised main result, but its proof depends on the omitted C^{1+delta/2,3+delta} bootstrap and the associated differentiated nonlocal equation. This is fixable within the manuscript's scope by adding a technical section with the required estimates, but it is not a purely cosmetic omission. I would recommend asking the authors to supply the full bootstrap argument and to correct the small consistency and typographical errors listed above."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: this is the first rigorous well-posedness result I know of for a two-phase Stefan problem with a genuinely nonlocal radiative term, and the core fixed-point argument for local existence is written in enough detail to verify. But the global well-posedness theorem has a load-bearing gap: the C^{1+δ/2,3+δ} bootstrap for ∂yT2 is asserted with 'we omit the details,' and the maximum-principle comparison for ∂yT2 needs exactly that regularity to make sense. This is not a cosmetic omission. The y-derivative of the nonlocal term involves a kernel convolution with T2^3∂yT2 plus a boundary term, and you need Hölder regularity of that source before the differentiated equation is a legitimate object. Until that bootstrap is written out, Theorem 3.1 should be treated as conditional, not established.\n\nWhat the paper does well: the modelling chain from the stationary radiative transfer equation to the nonlocal operator Iα is standard and clearly executed. The reduction to the moving-coordinate system and the rescaling are careful. The local existence proof via Banach fixed-point on Green's functions follows the classical Rubenstein/Friedman strategy, and the estimates are mostly explicit. I worked through enough of the contraction estimates to believe they close. The sign and constant typos are annoying but not fatal; a reader can resolve them. The maximum principle argument for the local solution properties is a nice piece of work. The citation pattern looks appropriate, and the relevant numerical engineering literature is covered.\n\nThe global section is the soft spot. Besides the omitted bootstrap, the barrier function w is constructed ad hoc and the proof of Lemma 3.1 is compressed. The claim that the class of initial data is 'optimal' for this argument is fine but should be labelled as such, not as a statement about the problem's true threshold. There are also notation inconsistencies in the L2 operators near the end that make the comparison argument harder to follow than it should be.\n\nBottom line: this deserves a serious referee. The model is new and physically motivated, and the local theory looks solid enough to publish on its own. But the global theorem should not be accepted in its current form. If I were the editor, I would send it to review with a clear request that the bootstrap be completed or the global result weakened to a conditional statement.","headline":"A credible first rigorous treatment of a Stefan problem with volumetric radiation; the local theory is checkable, but the global theorem currently rests on a regularity bootstrap the paper explicitly omits.","tokens_in":39359,"tokens_out":2643,"would_cite":false,"duration_ms":28853,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R35","35K55","80A22","35A01","35B50"],"pacs":[],"model":"deepseek-v4-flash","headline":"A two-phase Stefan problem with radiation is well-posed, locally and globally.","keywords":["Stefan problem","radiative transfer","free boundary problem","nonlocal parabolic equation","well-posedness","Banach fixed point","maximum principle","phase change"],"falsifier":"Work out the differentiated equation for v = ∂yT2 and check whether the source term 4∫₀^∞ (α/2)E1(α(y−η))T2³(η)∂ηT2(η)dη really lies in $C^{{δ/2,δ}}$_{t,y} with the stated Hölder exponents; exhibiting initial data for which ∂yT2 is only Lipschitz in space at t = 0 would invalidate the omitted bootstrap and therefore the global well-posedness theorem.","tokens_in":38344,"feed_emoji":"❄️","tokens_out":4771,"duration_ms":64586,"temperature":0.7,"pith_summary":"This paper studies a one-dimensional two-phase Stefan problem in which heat moves by conduction in both phases and also by radiation inside the solid, while the liquid is transparent to radiation. The authors reduce the radiative transfer equation to a nonlocal integral term with an exponential-integral kernel, so the solid temperature obeys a parabolic equation with a nonlocal emission term. Their main result is that for bounded initial data with the liquid initially above and the solid initially below the melting temperature, a unique solution exists locally in time, becomes classical away from t=0, and preserves the correct phase ordering. A second theorem gives global existence and uniqueness whenever the initial liquid temperature stays below TM + κL²/(K TM) and the initial solid temperature stays above 0, proved with stationary sub- and supersolutions and the maximum principle. If the results are correct, this is a well-posedness theory for melting with combined conduction and radiation that also exposes an explicit quantitative condition on the initial data separating global solvability from possible failure.","feed_headline":"Melting ice with radiation has unique solutions","feed_subtitle":"A nonlocal radiative term in the solid is tamed by contraction mapping and barrier functions, giving local and global well-posedness.","key_machinery":"The moving-coordinate change y = x − s(t) fixes the interface at y = 0, and the radiative transfer equation is solved by characteristics, turning radiation into the nonlocal term Iα[T] = T⁴(y) − ∫₀^∞ (α/2)E1(α|y−η|)T⁴(η)dη, where E1 is the exponential integral. The local fixed-point operator uses the half-space heat kernel G(y,ξ,a(t−τ)) = Φ(y−ξ,·) − Φ(y+ξ,·) to represent the temperatures u1,u2 and the Stefan velocity ṡ as integrals over initial data and nonlinear sources. For the global result, the load-bearing object is the stationary barrier w defined piecewise on R± and depending on parameters C1,C2,α chosen so that the inequalities L1(w) > 0 and L2(w) < 0 hold relative to the bounds on ṡ; this barrier controls the solution through the maximum principle.","core_discovery":"The core claim is that the free boundary problem (2.1) — a heat equation in the liquid, a heat equation with a nonlocal radiative term in the solid, and the Stefan condition for the interface speed — is well-posed. Locally the proof rewrites the solution as the fixed point of an operator built from half-space Green's functions for the heat equation, estimates the nonlocal radiation term in $C^{{0,1}}$, and applies the Banach contraction theorem on a small time interval; the fixed-point solution is then upgraded to a classical Hölder solution using parabolic estimates. Globally, the paper constructs stationary barrier functions w: on y<0 the barrier solves κw'' − C1w' = 0 and on y>0 it satisfies w'' + C2w' ≥ w⁴, with w(0)=TM. If the initial temperature lies below w on the liquid side and above w on the solid side, comparison shows that these inequalities persist, which bounds the temperature, its spatial derivative, and the interface speed, so the local solution extends to all times.","pith_inferences":["The threshold sup_{R−} T0 < TM + κL²/(K TM) is likely sharp for the barrier method, and near that threshold the solid-side barrier approaches TM from below while the interface speed is allowed to grow without bound; testing initial data at or above the threshold may reveal finite-time blow-up of the interface velocity.","Because the model assumes a transparent liquid and no external radiation, radiation can leave the solid but never return; adding an incoming radiative boundary condition gν(n) > 0 at the interface would plausibly produce superheated solid, as the paper notes, and one could test whether the same barrier construction still works in that regime.","The reduction from the kinetic radiative transfer equation to the nonlocal parabolic model drops the (1/c)∂tIν term, treating the photon transit time as negligible; a quantitative error estimate comparing this model with the full kinetic system in the c → ∞ limit would be a natural companion test of the modeling premise."],"forward_implications":["For every initial profile satisfying the hypotheses of Theorem 2.1, there is a unique solution up to some positive time t*, and that solution is classical for t > 0 with the interface speed given by the Stefan condition.","If the initial liquid temperature is below TM + κL²/(K TM) and the initial solid temperature is bounded away from 0, the unique solution exists for all positive times.","Along the solution, the phase ordering is preserved: the liquid stays above the melting temperature and the solid stays between 0 and TM, so no supercooled liquid or superheated solid appears in this regime.","The interface speed remains bounded between two constants determined by the barrier parameters, and the spatial derivatives of the temperature stay bounded, growing at most linearly in time on the liquid side and exponentially in time on the solid side.","The global result applies to a class of initial data that is optimal for the barrier argument: taking the parameters to their limiting values makes the liquid-side bound converge exactly to the threshold TM + κL²/(K TM)."],"supporting_citations":[{"why":"Supplies the Green's-function fixed-point strategy for the Stefan problem that the local existence proof adapts.","marker":"[37]"},{"why":"Establishes the classical local well-posedness method for melting problems that the contraction argument mirrors.","marker":"[17]"},{"why":"Provides the parabolic Schauder estimates and double-layer potential results used to upgrade the fixed-point solution to a classical Hölder solution.","marker":"[30]"},{"why":"Contains the characteristic computation that reduces the radiative transfer equation to the nonlocal E1-kernel integral term used throughout the paper.","marker":"[13]"},{"why":"Uses the maximum-principle route for global solutions of the classical Stefan problem that Section 3 extends to the radiative case.","marker":"[18]"}],"fun_headline_variants":["Well-posedness proved for radiating Stefan problem","Two-phase Stefan with radiation: local and global well-posedness","Melting ice with radiation: unique solutions proven","Nonlocal radiation in Stefan problem: well-posed","Global well-posedness for a radiating Stefan problem"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The global existence proof rests on a regularity bootstrap asserting that the solutions lie in $C^{{1+δ/2,3+δ}}$_{t,y} on [ε,t*] × R±, stated in the paper with the sentence 'Since the computations are similar ... we omit the details'; if that bootstrap fails, the maximum-principle control of ∂yT2 and hence the global theorem is not established.","fun_headline_variants_meta":{"raw":{"variants":["Well-posedness proved for radiating Stefan problem","Two-phase Stefan with radiation: local and global well-posedness","Melting ice with radiation: unique solutions proven","Nonlocal radiation in Stefan problem: well-posed","Global well-posedness for a radiating Stefan problem"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00029,"raw_usage":{"total_tokens":1655,"prompt_tokens":861,"completion_tokens":794,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":477,"completion_tokens_details":{"reasoning_tokens":731}},"tokens_in":477,"tokens_out":794,"duration_ms":9812,"temperature":1.0,"reasoning_tokens":731,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:19:15.484151+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Work out the differentiated equation for v = ∂yT2 and check whether the source term 4∫₀^∞ (α/2)E1(α(y−η))T2³(η)∂ηT2(η)dη really lies in $C^{{δ/2,δ}}$_{t,y} with the stated Hölder exponents; exhibiting initial data for which ∂yT2 is only Lipschitz in space at t = 0 would invalidate the omitted bootstrap and therefore the global well-posedness theorem.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the parabolic Schauder estimates and double-layer potential results used to upgrade the fixed-point solution to a classical Hölder solution."}],"review_version":1}