{"id":"725bdc7c-0291-44ff-a5a8-33a7ef478d33","arxiv_id":"2502.07521","paper_version":2,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"Convex integration yields nontrivial C^1 weak solutions of 2D elastodynamics from zero initial data under a weak null condition.","lead":"The paper constructs nonzero C^1 weak solutions of two-dimensional elastodynamics starting from zero initial data, even for materials satisfying a null condition. It uses convex integration with new wave-like building blocks, showing that weak formulations of nonlinear elasticity admit hidden non-uniqueness.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The closing estimate (8.16) appears incompatible with the frequency schedule (2.3) for b>5: the linear error term δ_{q+1}^{1/2}/(λ_{q+1}μ_q) is asymptotically larger than δ_{q+2}, matching the limitation admitted in Section 1.3.","rationale":"The reader identified the closing estimate (8.16) as the load-bearing premise. My stress test sharpens this into a concrete, checkable failure: the same linear term that Section 1.3 explicitly says limits the scheme is claimed in Proposition 8.3 to be controlled by δ_{q+2}, but the asymptotic comparison shows the opposite for the schedule (2.3) with any b>5. Because λ_q grows only subexponentially while δ_q decays exponentially, no choice of b or of the constants in Proposition 8.3 can make the term (λ_{q,i+1}μ_{q,i})^{-1}δ_{q+1}^{1/2} smaller than δ_{q+2} for all large q. Since Theorem 1.7 depends entirely on Proposition 2.6, which in turn uses Proposition 8.3, the central claim is not merely unverified but appears to contradict the paper's own stated limitation. I recommend REJECT rather than CONDITIONAL because the failure is an internal quantitative inconsistency, not an unproved technical detail. The concern is made in good faith: the construction is elaborate and the geometric lemma and building blocks are plausible, but the closing estimate is the gate through which the C^1 solution must pass, and it does not close.","tokens_in":107730,"tokens_out":7407,"duration_ms":59968,"concrete_test":"Independently compute the asymptotic order of the second term in (8.16) for q large. Take λ_q = ⌈ε^{-b(q+3)^{2/3}}⌉^3, δ_q = ε^{2(q+3)/3}, λ_{q,i} = λ_q^{1-i/3}λ_{q+1}^{i/3}, μ_{q,i} = (2⌈(λ_{q,i}λ_{q,i+1})^{1/2}δ_q^{1/4}/2⌉)^{-1}, and simplify λ_{q,i+1}μ_{q,i} ≈ (λ_{q+1}/λ_q)^{1/6}δ_q^{-1/4}. Then compare (λ_{q,i+1}μ_{q,i})^{-1}δ_{q+1}^{1/2} with δ_{q+2}; if the ratio tends to infinity as q increases (for example at q=10, 20, 40 with b=6), then (8.16) is false and Proposition 8.3 cannot be repaired by adjusting constants. This directly settles whether the internal contradiction with Section 1.3 is real.","verdict_should_be":"REJECT","load_bearing_attack":"The central proof relies on Proposition 8.3, specifically the estimate (8.16), which asserts with the schedule (2.3)-(2.4) that for b>5 and small ε, the new Reynolds error satisfies ‖∂t^r δR_{q,i+1}‖_N ≤ C λ_{q,i+1}^{N+r}(εδ_{q+1} + (λ_{q,i+1}μ_{q,i})^{-1}δ_{q+1}^{1/2}) ≤ (1/128)c_0^4 λ_{q,i+1}^{N+r}δ_{q+2}. The second summand is the linear wave error discussed in Section 1.3, where the authors state that this term 'limits our choice of δ_q' and 'prevents us from achieving the non-uniqueness of weak solutions in C^alpha spaces with alpha>0'. That admission is in direct tension with Theorem 1.7's C^1 conclusion, since C^1 regularity would imply C^alpha for every alpha<1. Quantitatively, using (2.3) and (2.4), one obtains λ_{q,i+1}μ_{q,i} ~ (λ_{q+1}/λ_q)^{1/6}δ_q^{-1/4} for each i=0,1,2. Since λ_{q+1}/λ_q ≈ ε^{-2b(q+3)^{-1/3}}→1 for fixed b, this gives (λ_{q,i+1}μ_{q,i})^{-1}δ_{q+1}^{1/2} ≈ ε^{(3q+11)/6}, while δ_{q+2}=ε^{(4q+20)/6}. Hence the quotient of the left-hand error term by δ_{q+2} grows like ε^{-(q+9)/6}, which diverges as q→∞. Thus inequality (8.16) cannot hold for all large q with the stated schedule, independent of the constants C_{b,σ,M} and c_0. This is an internal inconsistency: the inductive proposition that produces the nontrivial C^1 solution from zero data has a closing estimate that fails exactly at the linear wave error term, as the paper's own Section 1.3 indicates.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims to construct nontrivial weak solutions u ∈ C^1([0, T] × T^2) to the 2D elastodynamic system (1.9) emanating from zero initial data, under the weak null condition (1.13). The proof is a 66-page convex integration scheme: an induction produces approximate solutions (u_q, c_q, R_q) solving (2.1), with longitudinal-transverse building blocks (Lemma 7.1) exploiting the double wave speeds, a Lagrangian formulation of momentum and angular momentum constraints (1.17), a local inverse of the asymmetric divergence operator (Prop. 5.10), and an inductive Proposition 2.6 whose key closing estimate is (8.16). Theorem 1.7 asserts convergence of the scheme to a nonzero C^1 weak solution from zero data.","tokens_in":108374,"tokens_out":28043,"duration_ms":205083,"significance":"If the main theorem were correct, this would be a striking and important result: the first convex-integration-based non-uniqueness theorem for elastodynamics in Lagrangian coordinates, extending the method from fluid systems to quasilinear hyperbolic systems with a null condition. The paper contains genuinely new ideas and a large amount of careful technical work: the conserved quantities (1.17), the solution of the asymmetric divergence equation with time-derivative bounds in Proposition 5.10, the building block Lemma 7.1, and the energy-increment control (2.13). These components are creative and internally plausible in isolation. However, the significance of the paper hinges entirely on the quantitative closing of the induction, and that closing is not supported by the stated parameter schedule; the quantitative failure identified below is not a presentation issue but affects the central claim. Credit is due for the construction framework, but the main theorem is not established as written.","major_comments":[{"comment":"The closing estimate (8.16) of Proposition 8.3 is incompatible with the frequency-amplitude schedule (2.3)–(2.4). From (2.4), μ_{q,i} ≈ (λ_{q,i}λ_{q,i+1})^{−1/2}δ_q^{−1/4}, and for each i = 0, 1, 2 one obtains λ_{q,i+1}μ_{q,i} ≈ (λ_{q+1}/λ_q)^{1/6}δ_q^{−1/4}. The second summand in (8.16) therefore behaves as (λ_{q,i+1}μ_{q,i})^{−1}δ_{q+1}^{1/2} ≈ (λ_{q+1}/λ_q)^{−1/6}δ_q^{1/4}δ_{q+1}^{1/2} = ε^{(3q+11)/6}(λ_{q+1}/λ_q)^{−1/6}, while δ_{q+2} = ε^{(4q+20)/6}. Since λ_{q+1}/λ_q ≈ ε^{−2b(q+3)^{−1/3}} → 1 for any fixed b, the quotient of the linear-error term by δ_{q+2} grows like ε^{−(q+9)/6}, which diverges as q → ∞ independently of C_{b,σ,M} and c_0. Hence the asserted bound ≤ (1/128)c_0^4 λ_{q,i+1}^{N+r}δ_{q+2} cannot hold for all large q. This is not a missing computation but an internal inconsistency: the estimate the induction requires is false for the stated schedule. The paper's own admission in §1.3 that the linear wave error 'limits our choice of δ_q' and 'prevents us from achieving the non-uniqueness of weak solutions in C^α spaces with α>0' is exactly this failure. The same calibration problem invalidates (7.77) in the proof of Proposition 7.8, since Mμ_{q,i}λ_{q,i} ≈ M(λ_{q,i}/λ_{q,i+1})^{1/2}δ_q^{−1/4} ≈ Mε^{−(q+3)/6} cannot be bounded for large q by the claimed right-hand side ε^2δ_{q+1}δ_{q,i}^{−1/2} ≈ ε^{(2q+20)/6}. Thus the induction over q cannot be closed as written.","section":"§8.3, Eq. (8.16); §2, Eqs. (2.3)–(2.4)"},{"comment":"The mollification stage also fails under the stated schedule. Proposition 3.3 relies on (3.21), in particular on the inequality 120Mλ_{q,i} ≤ ℓ_{q,i}^{−1}. Using (2.3)–(2.4), λ_{q,i}ℓ_{q,i} = (λ_{q,i}/λ_{q,i+1})^{1/2}δ_q^{−1/4} = (λ_q/λ_{q+1})^{1/6}δ_q^{−1/4}, and since (λ_q/λ_{q+1})^{1/6} ≈ ε^{(b/3)(q+3)^{−1/3}}, we get λ_{q,i}ℓ_{q,i} ≈ ε^{(b/3)(q+3)^{−1/3} − (q+3)/6}. The exponent is negative and grows in magnitude like (q+3)/6 for large q, so λ_{q,i}ℓ_{q,i} ≫ 1 for all q beyond a fixed threshold (e.g., q ≥ 3 when b = 6). Consequently 120Mλ_{q,i} ≤ ℓ_{q,i}^{−1} is violated for large q uniformly in ε < ε_1^*(b,M). Since the estimates (3.11)–(3.17) are used at every step of the induction, the mollification step is not justified by the stated parameters. This is a second, independent calibration failure in the same core mechanism.","section":"§3, Eq. (3.21) and Proposition 3.3"},{"comment":"Theorem 1.7 claims u ∈ C^1([0,T] × T^2), and C^1 regularity implies C^α regularity for every 0 < α < 1. Section 1.3 states that the treatment of second-order and higher-order nonlinearities 'limits our choice of δ_q' and, in particular, that 'this limitation prevents us from achieving the non-uniqueness of weak solutions in C^α spaces with α>0.' Both statements cannot hold for the same construction. If the limitation in §1.3 is genuine, then the C^1 conclusion of Theorem 1.7 and the corresponding claim in Proposition 2.6 are false as stated; if the limitation is not genuine, the discussion in §1.3 misrepresents the scope of the method. Either way, the manuscript contains a direct self-contradiction on the regularity of the constructed solutions, and this contradiction is load-bearing because the final convergence argument in §9 rests on the C^1 bounds supplied by the induction.","section":"§1.3 and Theorem 1.7"}],"minor_comments":[{"comment":"The abstract contains spacing/typo artifacts such as 'NON-UNIQUENE SS' and 'elastodyna mic', and §1.3 contains 'Rlinaer' instead of 'R_linear'; these should be corrected in a final version.","section":"Abstract and §1.3"},{"comment":"The notational load is very heavy (superscripts (1),m,c, subscripts q,i+1, velocities A^υ_I, etc.). A summary table of the main symbols introduced in Sections 5–7 would substantially improve readability and reduce the risk of misreading the estimates.","section":"§5–§7 notation"},{"comment":"The version of the manuscript supplied for review breaks off inside §8.3.1 and does not display §8.4, §9, or the appendices. The final version must contain the complete proof of Proposition 2.6 and Theorem 1.7, because the closing of the induction depends on the estimates asserted in §8.3.","section":"§8.3.1, §8.4, §9"}],"recommendation":"reject","confidential_remarks":"This is a substantial technical manuscript with creative components, but the core quantitative claims do not survive scrutiny of the stated parameter schedule: the closing estimate (8.16) and the auxiliary bound (7.77), as well as the mollification inequality (3.21), fail for large q under (2.3)–(2.4). The contradiction between the claimed C^1 conclusion and the paper's own §1.3 limitation statement is direct. A reparameterization with much faster (e.g., super-exponential) frequency growth might salvage a weaker regularity statement, but that would require reworking large parts of Sections 3–8 and is beyond a local revision; as written, the main theorem is unsupported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things about this paper. First, it genuinely extends convex integration to elastodynamics in Lagrangian coordinates, with new building blocks that mix a longitudinal O(1) wave and a transverse O(ε) wave, exploiting λ+μ≠0, plus an asymmetric divergence inverse for symmetric stresses. That is the first construction of its kind, and the 66 pages of explicit estimates show real technical skill. Second, the central theorem — nontrivial C^1 weak solutions from zero data under a weak null condition — is not supported by the estimates in the paper, and the paper's own Section 1.3 says so.\n\nThe problem is the closing estimate (8.16). It requires the Reynolds error δR_{q,i+1} to be bounded by δ_{q+2}. The linear wave error term in that estimate is δ_{q+1}^{1/2}/(λ_{q,i+1} μ_{q,i}). With the schedule (2.3)-(2.4), λ_{q,i+1} μ_{q,i} ~ (λ_{q+1}/λ_q)^{1/6} δ_q^{-1/4} ≈ δ_q^{-1/4} = ε^{-(q+3)/6}, so the linear error is ~ ε^{(3q+11)/6}. But δ_{q+2} = ε^{(4q+20)/6}. The ratio is ε^{-(q+9)/6}, which blows up as q→∞. No constant can absorb that. So (8.16) cannot hold for all q.\n\nThe authors are aware of this: Section 1.3 states the linear error \"prevents us from achieving the non-uniqueness of weak solutions in C^α spaces with α>0.\" C^1 is strictly stronger than C^α for α<1, so the claim of Theorem 1.7 is in direct contradiction with that admission. Either the theorem's regularity should be downgraded substantially, or the parameter schedule has to change. As written, the proof does not close.\n\nThe new machinery — the mixed longitudinal-transverse building blocks and the asymmetric inverse divergence — is valuable and likely reusable. The construction of the perturbation and the angular momentum corrections is careful, and the citations to prior convex integration work are appropriate. If the authors can repair the closing estimate or adjust the claim, this could be a solid paper. But as it stands, the main result is unsupported.\n\nWho should read it: experts in convex integration and hyperbolic conservation laws, especially those working on non-uniqueness for quasilinear systems. It deserves a serious referee — the ideas are novel and the flaw is subtle enough that a referee's report could help the authors fix it. I would not cite it as a correct result in the next year, but I would want to see a revised version. Send it to peer review rather than desk-reject.","headline":"Novel first convex integration for Lagrangian elastodynamics, but the C^1 claim appears to contradict the paper's own admitted error estimates, so the main theorem is not supported as written.","tokens_in":108844,"tokens_out":11829,"would_cite":false,"duration_ms":83945,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35A02","35D30","35L05","35L15","35L72"],"pacs":[],"model":"deepseek-v4-flash","headline":"Under a weak null condition, 2D elastodynamics has nontrivial C^1 weak solutions from zero initial data, so weak solutions are not unique.","keywords":["convex integration","non-uniqueness","elastodynamics","weak null condition","C^1 weak solutions","Cauchy problem","hyperbolic systems"],"falsifier":"Compute the first Reynolds-error residual predicted by (8.16) with the paper's schedule $\\lambda_q=\\lceil \\varepsilon^{-b(q+3)^{2/3}}\\rceil^3$, $\\delta_q=\\varepsilon^{2(q+3)/3}$, $b>5$: the induction can start only if the residual after the first correction is at most the claimed multiple of $\\delta_2$. If that bound, or its uniform version for all $q$, fails, the approximate solutions need not converge in $C^1$ and the theorem's conclusion is not established.","tokens_in":107502,"feed_emoji":"🌊","tokens_out":8215,"duration_ms":75230,"temperature":0.7,"pith_summary":"The paper targets the two-dimensional elastodynamic system for hyperelastic materials whose stored energy satisfies a weak null condition. It claims that this system admits nontrivial weak solutions of class $C^1([0,T]\times\\mathbb{T}^2)$ starting from zero initial data. If that is right, the Cauchy problem from the zero datum has at least two weak solutions—the zero solution and the constructed one—so weak solutions are not unique. The proof is a convex integration construction: a sequence of approximate solutions is corrected by adding high-frequency waves, with the leftover error pushed to zero. The construction uses the double wave speeds of elasticity to design building blocks made of a longitudinal wave plus a small transverse correction.","feed_headline":"Elastodynamics: two C^1 weak solutions from one start","feed_subtitle":"A convex-integration construction shows that a weak null condition does not force a unique weak solution.","key_machinery":"The load-bearing mechanism is a convex integration iteration for the approximate system $\\partial_{tt}u-\\operatorname{div}((\\mathrm{Id}+G)\\Sigma_G)=\\operatorname{div}((\\mathrm{Id}+G)(c+R))$. At each sub-step the approximate solution is perturbed by high-frequency waves built from Lemma 7.1 building blocks: a longitudinal wave of order one along a direction $f_i$ plus a smaller transverse wave, which are approximate solutions of the linearized elastodynamic wave operator precisely because the two wave speeds $\\lambda+2\\mu$ and $\\lambda+\\mu$ are distinct. A geometric lemma (Lemma 2.2) represents any symmetric matrix close to the identity as a sum of three projection matrices along the directions $(4,3)$, $(4,-3)$, $(0,5)$, which lets the quadratic part of the perturbation cancel the current Reynolds error. The remaining non-symmetric error terms are absorbed through a new local right-inverse of the asymmetric divergence equation $\\operatorname{div}((\\mathrm{Id}+\\nabla u)\\tilde R)=U$, which is solvable when $U$ has zero Lagrangian momentum and angular momentum; that inverse is what keeps the new Reynolds error symmetric and gives the estimate (8.16) that closes the induction.","core_discovery":"The central claim is Theorem 1.7: for any stored-energy function of the form (1.2) satisfying (1.13) (positive Lam\\'e constants, convexity near the origin, and the weak null condition $3\\sigma_{11}+2\\sigma_{111}=0$, $\\sigma_*\\neq 0$), one can construct a non-zero weak solution $u\\in C^1([0,T]\\times\\mathbb{T}^2)$ to (1.9) that emanates from zero initial data. The proof runs through Proposition 2.6, an inductive statement that, starting from the zero tuple, produces approximate solutions $(u_q,c_q,R_q)$ with Reynolds error tending to zero, and then shows the sequence converges to a genuine weak solution. The solution is not an artifact of low regularity: it is continuously differentiable, even though the construction is of convex-integration type. The paper reads this as showing that a null condition—usually a mechanism for global existence and uniqueness of small classical solutions—does not prevent non-uniqueness once one works with weak solutions.","pith_inferences":["If Theorem 1.7 is right, then a null condition—usually a small-data global-existence device—does not restrict the weak-solution set: the weak formulation is far from well-posed even when the nonlinearity has good null structure.","The admitted obstruction to positive-H\\\"older regularity suggests that a different absorption of the non-principal quadratic terms might raise regularity; a natural next step is to try replacing the exponential decay of $\\delta_q$ with a polynomial one.","The Lagrangian momentum/angular-momentum inverse divergence is transportable: any hyperelastic system whose stress depends symmetrically on the deformation gradient, and for which a geometric lemma of the type (2.2) holds, should admit the same non-uniqueness construction.","A testable robustness check is to add a fourth-order term to the stored-energy function (1.2); the paper indicates third-order terms are handled directly, so this would probe whether the error absorption survives higher-order nonlinearities."],"forward_implications":["The Cauchy problem for the elastic wave system (1.9) has at least two weak solutions from the same zero data, so weak solutions are not unique under the assumptions (1.13).","The non-uniqueness occurs in $C^1$, which is smoother than the $L^\\infty$ or H\\\"older weak solutions typical of convex integration; it shows the non-uniqueness is not an artifact of very low regularity alone.","The characteristic double wave speeds $\\lambda+2\\mu>0$ and $\\lambda+\\mu\\neq 0$ are used to build perturbations as a longitudinal wave plus a small transverse wave; materials violating $\\lambda+\\mu\\neq 0$ fall outside this construction.","The iteration produces approximate solutions whose stored energy increases by a controlled amount at each step, so the limiting solution has positive energy despite zero initial data."],"supporting_citations":[{"why":"Supplies the convex integration iteration and the geometric lemma (2.2) that the induction uses to cancel the Reynolds error.","marker":"[22, 24]"},{"why":"Introduces the symmetric divergence equation and the momentum/angular-momentum conservation used for the local inverse of divergence.","marker":"[28]"},{"why":"Provides the multi-direction perturbation method used when adding waves in three spatial directions at each step.","marker":"[42]"},{"why":"The authors' companion work showing the higher-order obstacle; it sets the parameter restriction that the present paper's new error absorption must overcome.","marker":"[44]"},{"why":"States the null condition for elastic waves that yields the weaker condition (1.11) and motivates the structural assumptions.","marker":"[48]"}],"fun_headline_variants":["Null condition fails to ensure uniqueness in elastodynamics","Elastodynamic null condition still allows multiple weak solutions","Weak solutions in elastodynamics bypass null condition uniqueness","Convex integration shows null condition not enough for uniqueness","Null condition does not force uniqueness in 2D elastodynamics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole induction closes only if the leftover error after each correction is smaller than the next error level for the chosen exponentially growing frequencies and decaying amplitudes, at every step; the authors flag this absorption step as the one that blocks positive-H\\\"older regularity, so the $C^1$ claim rests entirely on that estimate holding uniformly.","fun_headline_variants_meta":{"raw":{"variants":["Null condition fails to ensure uniqueness in elastodynamics","Elastodynamic null condition still allows multiple weak solutions","Weak solutions in elastodynamics bypass null condition uniqueness","Convex integration shows null condition not enough for uniqueness","Null condition does not force uniqueness in 2D elastodynamics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000566,"raw_usage":{"total_tokens":2640,"prompt_tokens":858,"completion_tokens":1782,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":474,"completion_tokens_details":{"reasoning_tokens":1720}},"tokens_in":474,"tokens_out":1782,"duration_ms":11820,"temperature":1.0,"reasoning_tokens":1720,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T12:28:03.978361+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the first Reynolds-error residual predicted by (8.16) with the paper's schedule $\\lambda_q=\\lceil \\varepsilon^{-b(q+3)^{2/3}}\\rceil^3$, $\\delta_q=\\varepsilon^{2(q+3)/3}$, $b>5$: the induction can start only if the residual after the first correction is at most the claimed multiple of $\\delta_2$. If that bound, or its uniform version for all $q$, fails, the approximate solutions need not converge in $C^1$ and the theorem's conclusion is not established.","supporting_citations":[{"cited_title":"Isett and S","cited_arxiv_id":null,"evidence_quote":"Introduces the symmetric divergence equation and the momentum/angular-momentum conservation used for the local inverse of divergence."},{"cited_title":"H\\\"{o}lder Continuous Solutions to the Three-dimensional Prandtl System","cited_arxiv_id":"1804.04285","evidence_quote":"Provides the multi-direction perturbation method used when adding waves in three spatial directions at each step."},{"cited_title":"Mao and P","cited_arxiv_id":null,"evidence_quote":"The authors' companion work showing the higher-order obstacle; it sets the parameter restriction that the present paper's new error absorption must overcome."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"States the null condition for elastic waves that yields the weaker condition (1.11) and motivates the structural assumptions."}],"review_version":1}