{"id":"abdb6c43-9aae-4c12-a2c1-58ea0f48b407","arxiv_id":"2412.00014","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A procedural framework for embedding quadratically nonlinear PDEs into infinite-dimensional linear PDE systems via continuous Kronecker powers of the state.","lead":"This paper extends Carleman linearization, a technique for rewriting nonlinear ODEs as infinite linear systems, to partial differential equations with quadratic nonlinearities. It gives explicit formulas for the linearized hierarchy and applies them to Burgers and Vlasov equations.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Truncated Carleman system (20) lacks any convergence or well-posedness analysis; the formal solution (23) requires semigroup generation by a differential operator matrix, and inviscid Burgers violates the smoothness needed for the hierarchy.","rationale":"Proposition 3.1 is a formal identity; the algebra of the Leibniz rule plus variable relabeling is consistent, and the Burger's example verifies the hierarchy up to third order. The load-bearing weakness is the step from formal embedding to a numerical or analytical method: the truncated system (20) is a linear PDE in N*m variables, and its solution (23) presupposes a semigroup for the differential operator A_N. No existence, uniqueness, or error bound is provided; Section 5 defers this explicitly. For inviscid Burgers, finite-time shock formation means the products in (12) lose smoothness, so the formal Taylor series (33) cannot be expected to converge past the shock time. These are gaps in support, not demonstrated contradictions, so the paper is best read as a formal construction with an unverified convergence claim. The reader's weakest assumption identifies the same point; our verdict matches theirs.","tokens_in":10400,"tokens_out":26323,"duration_ms":209997,"concrete_test":"For the viscous Burgers equation with a known Cole-Hopf solution, implement the truncated system (20) for N=1,2,3,4 using a spectral discretization of the space X_N, evolve to a fixed time T, and compute the L2 error against the exact solution. If the error does not decrease with N (or diverges), the central convergence assumption fails. Also compute the inviscid case and compare the radius of convergence of (33) with the shock time.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central mathematical identity (Prop 3.1) appears formally correct: differentiating (12) via the Leibniz rule and relabeling variables reproduces (15), and both examples check out. However, the paper's practical claim that truncation to (20) yields a solvable finite linear system is insecure. The operator A_N(X_N) in (22) is a matrix of unbounded differential/integral operators on R^{mN}; the exponential in (23) is only meaningful if A_N generates a C0-semigroup on a suitable function space, which is not established. Section 5 explicitly states that no error analysis has been performed, so convergence as N→∞ is unverified. For inviscid Burgers (µ=0), solutions develop shocks in finite time, so the products u(x1)...u(xi) cease to be classical functions; the Taylor series (33) therefore cannot converge beyond the shock time. Without a rigorous truncation bound analogous to Ref. [10] or a numerical demonstration, the advertised advantage of eliminating discretization error is unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper generalizes Carleman linearization from ODEs to PDEs with quadratic nonlinearities of the form (11). It defines auxiliary variables y_i = u(x1) ⊗ ... ⊗ u(xi) and proves in Proposition 3.1 that their evolution is given by the tridiagonal infinite linear system (15) with operators (16a)-(16c). The system is then truncated at level N, yielding the finite linear PDE system (20)-(22) with a formal solution written as an exponential (23). Two examples are worked out, the Burgers equation and the Vlasov equation. The paper concludes by noting that an error analysis has not yet been performed and proposes this as future work.","tokens_in":10561,"tokens_out":8783,"duration_ms":72508,"significance":"If backed by a rigorous truncation and convergence analysis, the construction would provide a procedural, exact embedding of a class of quadratically nonlinear PDEs into a linear hierarchy, with potential applications to quantum simulation and numerical methods. The Leibniz-rule derivation in Proposition 3.1 is explicit and checkable by hand, and the two examples cover both a scalar conservation law and a kinetic equation. The paper is commendably honest in stating that no error analysis is performed, but this honesty also exposes the gap between the formal construction and the practical claims of solvability and elimination of discretization error.","major_comments":[{"comment":"The formal exponential solution z(X_N,t) = exp(A_N t) z0 + ... is only meaningful if the operator A_N(X_N), whose entries are unbounded partial differential and integral operators, generates a C0-semigroup on a suitable Banach space over R^{mN}. The manuscript specifies no such function space and proves no generation theorem, so (23) remains a formal expression rather than an analytical solution. Since the paper's stated practical advantage is the solvability of the truncated linear system, this gap is load-bearing.","section":"§3.2, Eq. (23)"},{"comment":"The paper explicitly states that 'an error analysis has not yet been performed' and that a truncation error bound 'is the next step to be done.' Without a bound on the difference between solutions of the infinite hierarchy (15) and the truncated system (20), or a convergence result as N→∞, the central practical claim that the truncation yields a usable computational method is not supported. The analogy to the ODE error bound of Ref. [10] is not a substitute for a PDE analysis, which must address unbounded spatial operators and the choice of function spaces.","section":"§5, first two paragraphs"},{"comment":"For the inviscid Burgers equation (µ=0), which is presented as the first toy example, smooth initial data generally develop shocks in finite time. The hierarchy (12) requires u(x1,t)...u(xi,t) to be a classical function for each i, so the hierarchy and the Taylor series (33) cease to be well-defined at the shock time. The paper calls (33) a formal solution and suggests it 'may form the basis of a time-integrator,' but this example illustrates a concrete obstruction to the method's global validity that should be discussed or excluded.","section":"§4.1, Eq. (33) and §3.1, Eq. (12)"}],"minor_comments":[{"comment":"The domain of z is written as R^{m·n} × R≥0; it should be R^{mN} × R≥0 (or R^{m·N}), since X_N contains N copies of the m-dimensional coordinate variable.","section":"§3.2"},{"comment":"For i=1, the operator A^1_0 as defined involves the variable-change condition x_j = x_{j+1} for j≥1, but no x_2 exists in the one-level state; this case should be treated separately, as is implicitly done in (20) by making F0 an inhomogeneity.","section":"Proposition 3.1 and Eq. (16a)"},{"comment":"The bottom-left block of the matrix A_N should be A^N_{N-1}, consistent with (20); as printed it appears to be A^{N-1}_N, which is the coupling in the second-to-last row.","section":"Eq. (22)"},{"comment":"The sentence 'The F2 operator is not unique generally' raises a question: different choices of F2 lead to different operators A^i_{i+1} in the hierarchy, so the manuscript should state whether the resulting linear systems are equivalent, or at least that the non-uniqueness is harmless.","section":"§4.1"},{"comment":"The claim that the method 'eliminate[s] coordinate discretization error completely' should be qualified in light of the admitted absence of an error analysis in Section 5; at present the claim is only conditional on the existence of a convergent truncation.","section":"Abstract and §1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is classified as math.GM and reads as a formal research announcement rather than a complete analysis. The core identity is correct, but the gap between the formal construction and the stated computational advantages is substantial. I recommend major revision rather than rejection because the missing analysis could in principle be supplied. Note also that Ref. [17] apparently already sketches the continuous Kronecker product idea; the paper's claimed novelty is the procedural operator construction, which is acceptable if the analysis gap is closed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nI read the Vaszary preprint on Carleman linearization of PDEs. The headline: the central mathematical identity is correct and the paper is honestly written, but it oversells the practical reach.\n\nThe Leibniz-rule derivation in Proposition 3.1 checks out. The explicit operator formulas (16a)-(16c) and the truncation scheme (20) are genuinely useful and are not in the cited prior work. The inviscid Burgers formal solution (31) is a nice touch. The paper also deserves credit for stating in Section 5 that no error analysis has been performed; that admission is present, though easy to miss.\n\nThe soft spots are exactly where the title claims bite. The paper advertises eliminating coordinate discretization error and being procedurally implementable. Both claims depend on the truncated linear system (20) being well-posed and its solution converging to the original PDE as N grows. None of that is shown. The formal exponential solution (23) requires A_N(X_N) to generate a semigroup on a suitable function space; the entries are unbounded differential and integral operators, so this is not automatic. For inviscid Burgers, classical solutions break down in finite time, so the hierarchy (12) ceases to make sense beyond the shock, and the Taylor series (33) cannot converge there. Without a truncation bound analogous to [10] or at least a numerical demonstration, the practical advantage is unsupported. None of this invalidates the formal embedding; it just means the paper is a formalism, not yet a method.\n\nWho is this for? Someone working on Carleman or Koopman methods, or on quantum simulation of nonlinear PDEs, who wants the explicit operator structure. That reader gets a clean, self-contained derivation and two consistent examples. The citation pattern is fair: it attributes the continuous Kronecker power idea to [17] and does not hide its debt.\n\nI would send this to a serious referee. The core construction is new and formally grounded, and a referee can ask the authors to either prove convergence under stated conditions or soften the claims and add numerical evidence. It should not be desk rejected.\n\nRecommendation: engage with it, but expect major revision.\n\nYours,\n[Name]","headline":"The core formal construction is correct and clearly exposed, but the advertised practical claims outrun the evidence, as the paper itself concedes.","tokens_in":11110,"tokens_out":2081,"would_cite":false,"duration_ms":18889,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35A22","35Q53","35Q83"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper extends Carleman linearization from ODEs to quadratically nonlinear PDEs, embedding them exactly into infinite-dimensional linear PDE systems via continuous Kronecker powers.","keywords":["carleman linearization","partial differential equations","quadratic nonlinearity","kronecker product","variable change operator","burgers equation","vlasov equation","quantum simulation"],"falsifier":"Solve the linearized hierarchy for a smooth analytic solution of inviscid Burgers with known exact solution, and compare the first component of the truncated system (23) as $N$ grows against the exact solution: if the difference does not shrink to zero on the time interval where the exact solution is smooth and analytic, the exact-embedding claim is false. Alternatively, for the Vlasov example, check whether the truncated hierarchy preserves total mass; a violation that does not vanish as $N$ grows would also falsify exact embedding.","tokens_in":10181,"feed_emoji":"🧮","tokens_out":4616,"duration_ms":39164,"temperature":0.7,"pith_summary":"Dynamical systems with polynomial nonlinearity can be rewritten as infinite linear systems by Carleman linearization, a procedural method whose auxiliary variables are the Kronecker powers of the state vector. This paper shows the same construction works for partial differential equations with quadratic nonlinearities: one uses different copies of the coordinate variable in each factor of the Kronecker product, so the new variables are products of the field at different spatial points. The paper proves that these product variables obey a tridiagonal linear PDE system, and that truncating that system at finite order yields a finite linear PDE system with an explicit formal solution. The authors demonstrate the construction on Burgers' equation and on the Vlasov equation, and argue that the approach removes coordinate discretization error, a step toward solving nonlinear PDEs on analog quantum simulators.","feed_headline":"Quadratic PDEs become linear PDE systems","feed_subtitle":"Continuous Kronecker powers embed nonlinear PDEs into infinite linear hierarchies, shown on Burgers and Vlasov.","key_machinery":"The continuous analogue of the Kronecker product: for vector fields $u(x)$, the $i$-th level variable is $u(x_1)\\otimes \\cdots \\otimes u(x_i)$, a function on $i$ copies of the coordinate space. The paper also introduces the variable-change operator that replaces one coordinate variable by another, expressed both abstractly and as an integral against a Dirac delta; these operators let the quadratic term $F_2(x;w)[u(x)\\otimes u(w)]$ be rewritten as a linear operator acting on the next level $y_{i+1}$ after suitable relabelling of coordinates. Together they give the tridiagonal linear structure and make the whole construction procedural.","core_discovery":"The central claim is Proposition 3.1: for a PDE of the form $\\partial u/\\partial t = F_0(x) + F_1(x)u(x) + F_2(x;w)[u(x)\\otimes u(w)]$, the auxiliary functions $y_i = u(x_1)\\otimes \\cdots \\otimes u(x_i)$ evolve according to the tridiagonal linear PDE system $\\partial y_i/\\partial t = A^i_{i-1}(X_i) y_{i-1} + A^i_i(X_i) y_i + A^i_{i+1}(X_i; x_{i+1}) y_{i+1}$, with the explicit linear operators (16a)-(16c) built from $F_0$, $F_1$, $F_2$ and variable-change operators. It follows that every quadratically nonlinear PDE of this form is exactly embedded into an infinite-dimensional linear system, and truncating at level $N$ gives a finite linear PDE system with formal exponential solution (23). The paper applies this to Burgers' equation, where the inviscid case yields a Taylor-series solution in time, and to a two-species Vlasov–Poisson system.","pith_inferences":["A natural testable extension is an error bound for the truncated hierarchy in the style of the ODE Carleman bounds, using a continuous analogue of the log-norm; the paper states this is the next step.","The framework may extend to polynomial nonlinearities of any degree by the same quadratization procedure used for ODEs, which the paper anticipates but does not prove.","If well-posedness of the linearized system can be established, the method could be compared numerically against spectral or finite-difference solvers on shock problems, where the Taylor-series form (33) is expected to have finite radius of convergence.","For kinetic equations like Vlasov, the exact linear embedding could provide a new route to deriving reduced models or closures by projecting the linear hierarchy."],"forward_implications":["Any quadratically nonlinear PDE that fits the form (11) can be recast as an infinite linear PDE system, so methods and intuition for linear PDEs become applicable to nonlinear ones.","Truncating at order $N$ gives a finite linear PDE system whose formal solution is an exponential, so time integration can in principle be done without discretizing the coordinate space.","The inviscid Burgers example produces an explicit Taylor-series representation of the solution, suggesting a time-integrator built directly from the linearized hierarchy.","The Vlasov example shows the framework covers integro-differential nonlinearities, not just local ones, as long as they can be written in the quadratic form with dummy variables.","Because the linear system lives on continuous coordinate copies, it is a candidate for analog quantum simulation, eliminating coordinate discretization error entirely."],"supporting_citations":[{"why":"Introduces Carleman linearization for ODEs, the method being generalized.","marker":"[8]"},{"why":"Supplies the explicit error-bound framework and notation for the ODE case that the paper expects to generalize.","marker":"[10]"},{"why":"Identifies the continuous Kronecker power idea using different copies of independent variables, which the paper makes procedural.","marker":"[17]"},{"why":"Previous attempt at infinite-dimensional Carleman linearization without the continuous Kronecker product, which the paper contrasts with its own approach.","marker":"[18]"}],"fun_headline_variants":["Quadratic PDEs become infinite linear hierarchies","Carleman linearization generalized to PDEs","Exact linear embedding for quadratic PDEs","Infinite linearization of quadratic PDEs via Carleman","Burgers and Vlasov linearized by Carleman method"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction assumes the infinite linear system can be cut off at a finite level without losing the solution of the original PDE: the truncated system must be well-posed and its solution must converge to the true solution as the cutoff $N$ grows, and the paper states that no error analysis has been performed.","fun_headline_variants_meta":{"raw":{"variants":["Quadratic PDEs become infinite linear hierarchies","Carleman linearization generalized to PDEs","Exact linear embedding for quadratic PDEs","Infinite linearization of quadratic PDEs via Carleman","Burgers and Vlasov linearized by Carleman method"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000431,"raw_usage":{"total_tokens":2147,"prompt_tokens":839,"completion_tokens":1308,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":455,"completion_tokens_details":{"reasoning_tokens":1231}},"tokens_in":455,"tokens_out":1308,"duration_ms":9886,"temperature":1.0,"reasoning_tokens":1231,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T20:31:37.207194+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Solve the linearized hierarchy for a smooth analytic solution of inviscid Burgers with known exact solution, and compare the first component of the truncated system (23) as $N$ grows against the exact solution: if the difference does not shrink to zero on the time interval where the exact solution is smooth and analytic, the exact-embedding claim is false. Alternatively, for the Vlasov example, check whether the truncated hierarchy preserves total mass; a violation that does not vanish as $N$ grows would also falsify exact embedding.","supporting_citations":[{"cited_title":"Infinite-dimensional Carleman linearization, the L ie series and optimal control of non-linear partial differential equations,","cited_arxiv_id":null,"evidence_quote":"Previous attempt at infinite-dimensional Carleman linearization without the continuous Kronecker product, which the paper contrasts with its own approach."}],"review_version":1}