{"id":"8d99baca-2606-45f4-af44-52a74a1eed7d","arxiv_id":"2608.06249","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Locally supported flux feedback exponentially stabilizes multidimensional continuity equations whenever every relevant characteristic accumulates enough damping over a finite time horizon.","lead":"Control designers usually steer densities by acting on the whole domain, but many applications only allow localized action. This paper gives a checkable geometric condition under which a small active region can exponentially stabilize a transported density toward a target profile, and it proves input-to-state bounds in the presence of noise.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem IV.1's explicit constants are invalid: the iteration proof requires C_T ≥ e^{2κm_T}, not e^{M_A^- T}, and the stated estimate fails for divergence-free fields before the entrance time.","rationale":"The reader's weakest assumption concerned the regularity of the velocity field. My stress-test identified a different, more concrete defect: the explicit constants in Theorem IV.1 do not follow from the paper's own proof. The algebra shows that the initial interval of length r<T requires a constant at least e^{2κm_T}, not e^{M_A^- T}. Since Theorem IV.1 is the central stability theorem, this is a load-bearing flaw. However, the qualitative exponential stability is not disproved: the iteration argument works as soon as the constant is corrected to e^{2κm_T} (or any bound accounting for the 'rough estimate' on each subinterval). Therefore the paper needs a correction and some re-verification of downstream statements, but not rejection. The verdict remains conditional, as the reader already recommended, though for a different reason. The quantitative example with a divergence-free rotation field makes the point concrete: before characteristics first reach the active region, energy cannot have decayed, so any claimed uniform estimate with a small C_T and positive rate is impossible for early times. No ad hominem is intended; the proof strategy is sound, but the printed constants are inconsistent with it.","tokens_in":809,"tokens_out":5050,"duration_ms":278785,"concrete_test":"Let Ω be the unit disk, u(x,y)=(−y,x) so ∇·u=0 and M_A^-=0. Let σ=1 on a small disk V_d centered at (0,0.3) with radius r_d=0.2, and σ=0 outside. Take initial error supported on a small arc of the circle of radius 0.5 centered at the origin, chosen so that the arc lies outside V_d initially but the full circle intersects V_d. For T=2π, the characteristic damping condition holds on the forward-invariant circle with m_T>0, and the gain condition holds for κ large. At t = half the time until first entry into V_d, the true L2 energy equals ∥e0∥², because no damping has acted yet, whereas the theorem's claimed estimate with C_T=1 gives a strictly smaller upper bound. Recompute the constants as C_T = e^{2κm_T} to confirm the corrected version.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In Theorem IV.1 the authors claim that one may take C_T = e^{M_A^- T} and α_T = 2κm_T/T − M_A^-. The iteration proof gives ∥e(nT+r)∥² ≤ e^{M_A^- r} q_T^n ∥e0∥² for r∈[0,T], where q_T = exp(M_A^- T − 2κm_T). Writing e^{-α_T(nT+r)} = q_T^n e^{-α_T r}, the required constant must satisfy C_T ≥ e^{(M_A^- + α_T)r} for all r∈[0,T], hence C_T ≥ e^{2κm_T}. The stated C_T = e^{M_A^- T} is strictly smaller whenever the gain condition 2κm_T > M_A^- T holds. The error is not merely cosmetic: take a divergence-free field (M_A^- = 0) and an initial error supported on a characteristic that has not yet reached the active region. For t < T_in, no damping acts and the L2 norm stays exactly at ∥e0∥², while the stated bound gives e^{-2κm_T t/T}∥e0∥² < ∥e0∥². Thus Theorem IV.1 as stated is false, although the qualitative exponential stability claim survives with a corrected constant.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies stabilization of multidimensional continuity equations on bounded domains by interior flux feedback prescribed through the divergence of the flux: ∇·J_d = κσe − r⋆, where e = ρ − ρ⋆, σ is a nonnegative localization function, and r⋆ is the transport residual of a reference profile. The closed-loop error satisfies ∂_t e + ∇·(ue) = −κσe + w. The central result is Theorem IV.1, which asserts exponential L2 stability under a finite-time characteristic damping condition D_T(t,x) ≥ m_T and a gain condition 2κm_T > M_A^-T. The paper also develops Lyapunov-type entrance conditions (Assumption III.4) to verify the damping condition, a weighted Lyapunov functional giving ISS estimates (Theorem IV.8), elliptic right-inverse constructions for realizing the feedback flux, an extension to density-dependent velocity fields under an admissible-class assumption, and a two-dimensional radial example.","tokens_in":24046,"tokens_out":14475,"duration_ms":144553,"significance":"If the main theorem is correct, the paper gives a clean and geometrically checkable sufficient condition for localized stabilization of continuity equations, with an explicit balance between accumulated damping and compressive amplification of the transport field. The proof is self-contained: the characteristic formula, Liouville's change of variables, and the entrance argument are stated explicitly, and the example verifies all hypotheses analytically. The separation of the divergence-level feedback law from the flux realization is a useful structural contribution, and the ISS estimate is a valuable addition. The nonlinear extension is honestly presented as conditional on an admissible class A, with well-posedness explicitly left open in Remark VI.7. The main issue found is a wrong explicit constant in Theorem IV.1; this is fixable and does not destroy the qualitative stability claim, but the theorem as stated is false.","major_comments":[{"comment":"The constants stated in Theorem IV.1 are not supported by the proof. After the one-step contraction ∥e(t+T)∥² ≤ q_T∥e(t)∥² and the rough bound ∥e(t+r)∥² ≤ e^{M_A^- r}∥e(t)∥² for r ∈ [0,T), iterating gives ∥e(nT+r)∥² ≤ e^{(M_A^-+α_T)r} q_T^n ∥e0∥². Since e^{−α_T(nT+r)} = q_T^n e^{−α_T r}, the multiplicative constant must be at least e^{(M_A^-+α_T)T} = e^{2κm_T}. The claimed C_T = e^{M_A^- T} is strictly too small whenever (28) holds. For a divergence-free field (M_A^- = 0) and an initial error whose characteristic has not yet reached the active region, the stated bound predicts decay e^{−2κm_T t/T}∥e0∥² before any damping acts, contradicting exact conservation of the L2 norm; hence the statement as written is false. The qualitative exponential stability is recovered with C_T = e^{2κm_T}, so the theorem and the sentence 'We may take C_T = e^{M_A^- T}, α_T = ...' should be corrected accordingly.","section":"§IV-A, Theorem IV.1 and its proof"}],"minor_comments":[{"comment":"In the displayed definition of ω[ρ](t), the numerator and denominator are identical, so ω[ρ] ≡ ω0 and the density dependence is vacuous. The tangency argument remains valid for any scalar functional, but the demonstration of a genuinely density-dependent perturbation should be corrected.","section":"§VII.E"},{"comment":"The symbol α is used both for the admissible Lyapunov rate from Theorem IV.5, where the homogeneous estimate has e^{−2αt}, and for the ISS decay rate in (38), which has e^{−αt}. The notation should be aligned to avoid confusion.","section":"§IV-B, Theorem IV.8"},{"comment":"In the anisotropic-field remark, the displayed inequality D_uΦ(x) ≤ −2µ(1−ε0R0²)a_d appears to use a_d where r_d² (or |x|²) is intended; please correct the notation for dimensional consistency.","section":"§VII, Remark VII.1"},{"comment":"The regularity assumption is described as 'minimal', but W^{1,∞} regularity of the velocity field is essential to the Lipschitz-flow/Liouville argument and excludes BV or discontinuous flows. The authors may wish to state this scope limitation explicitly, since the finite-time characteristic damping argument does not apply to shock solutions.","section":"Assumption II.4"}],"recommendation":"major_revision","confidential_remarks":"The constant error in Theorem IV.1 is real but localized: replacing C_T by e^{2κm_T} repairs the proof and preserves the qualitative exponential stability claim. The rest of the paper reads as technically sound, and the example is genuinely illustrative. I would not reject on this basis; the authors should be asked to correct the theorem statement and proof, and to fix the typos in the example. The paper is within scope for a control/PDE journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The core of the paper is genuinely useful: a finite-time characteristic damping condition plus a gain condition gives L2 exponential stability for multidimensional continuity equations with localized flux-divergence feedback, and the support-restricted formulation is a real improvement over global damping assumptions. The characteristic calculus in Theorem IV.1 is clean and correct, using Liouville's formula and the change of variables properly. The weighted Lyapunov criterion and ISS estimate are also sound, and the entrance-time argument in Proposition III.5 is valid. Separating the stabilization argument from the flux realization is thoughtful and opens a clean path to actuator design. The paper is worth a careful read.\n\nThat said, the stress-test note is right, and it lands on a stated result. In Theorem IV.1, the claimed constants C_T = e^{M_A^- T} and alpha_T = 2 kappa m_T / T - M_A^- are inconsistent with the iteration proof. After n iterations plus the rough bound on the final remainder, the prefactor must be at least e^{2 kappa m_T}, not e^{M_A^- T}. The example of a divergence-free field with an error that has not yet reached the active region shows the stated estimate is literally false: the norm stays constant while the stated bound decays with rate 2 kappa m_T / T. The good news is that the qualitative exponential stability claim survives if you take C_T >= e^{2 kappa m_T}; the proof goes through with a one-line change. So this is a fixable error, but it should be corrected before publication.\n\nThere is also an outright typo in Section VII.E: the formula for omega[rho](t) is omega_0 times a fraction whose numerator and denominator are identical, so omega[rho](t) is constant, not a density-dependent functional. The surrounding claim that it is a bounded nonlinear functional is wrong as written. This is minor and does not affect the linear theorems, but the authors should fix it.\n\nOther soft spots are minor. The nonlinear extension is honestly labeled as conditional, with well-posedness deferred, which is acceptable for a stability paper. The citation pattern is clean; the few self-citations are context only, and nothing is fitted to data. The lack of code is normal for this kind of theory paper.\n\nWho this is for: people working on stabilization of transport and continuity equations, and more broadly on geometric conditions for localized damping in PDE control. It deserves a serious referee. I recommend sending it to peer review, asking for the constant correction, the example formula fix, and perhaps a check of the remaining constants in Corollary IV.4 and Theorem VI.6 for the same prefactor issue.","headline":"Solid geometric framework for localized damping of transport PDEs, with a real but fixable error in the explicit decay constants of Theorem IV.1 and a typo in the illustrative example.","tokens_in":24514,"tokens_out":2373,"would_cite":true,"duration_ms":20959,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["93C20","93D15","93D30","35L65"],"pacs":[],"model":"deepseek-v4-flash","headline":"Localized interior flux feedback can exponentially stabilize a multidimensional transport error whenever every relevant characteristic accumulates a uniform amount of damping in finite time.","keywords":["continuity equations","transport equations","interior flux feedback","localized stabilization","finite-time characteristic damping","Lyapunov methods","input-to-state stability","control of partial differential equations"],"falsifier":"Simulate the closed-loop error equation $\\partial_t e+\\nabla\\cdot(ue)=-\\kappa\\sigma e$ on the unit disk with inward radial field $u(x)=-\\mu x$ and with $\\sigma$ supported in an annulus that the shrinking family of characteristics never reaches, for instance with $\\operatorname{supp}\\sigma$ outside $\\{|x|\\le e^{-\\mu t}R_0\\}$; then $D_T(t,x)=0$ on the relevant support, the gain condition $2\\kappa m_T>M_A^-T$ fails for every finite $\\kappa$, and the $L^2$ norm should grow like $e^{2\\mu t}$ rather than decay—directly contradicting any uniform exponential stabilization claim in that configuration.","tokens_in":23489,"feed_emoji":"🎯","tokens_out":11005,"duration_ms":112723,"temperature":0.7,"pith_summary":"This paper tries to establish that a multidimensional continuity equation with source terms can be driven towards a prescribed reference profile by a feedback that acts only through the divergence of an interior flux, localized to part of the domain. The mathematical claim is that the error $e=\\rho-\\rho_\\star$ obeys a transport equation with localized damping, $\\partial_t e+\\nabla\\cdot(ue)=-\\kappa\\sigma e$, and that this equation is exponentially stable in $L^2$ provided a finite-time characteristic damping condition holds: along every relevant flow line, the accumulated damping $\\int_t^{t+T}\\sigma(s,X(s;t,x))\\,ds$ is bounded below by $m_T>0$. The stability theorem then requires the feedback gain to beat the worst compressive amplification of the velocity field, $2\\kappa m_T > M_A^- T$, and yields an explicit decay rate. If true, this gives a geometric criterion for localized stabilization of transport and separates the stability argument from the engineering problem of realizing the feedback flux, which is handled by elliptic right-inverse constructions. A two-dimensional inward-flow example verifies the conditions and the gain threshold in a simple explicit setting.","feed_headline":"Transport PDEs stabilize if every flow line meets the damped zone","feed_subtitle":"A finite-time characteristic damping condition turns localized interior flux feedback into exponential decay, with an explicit gain…","key_machinery":"The central object is the finite-time characteristic damping condition of Definition III.1: for the characteristic flow $X(s;t,x)$ of $u$, the accumulated integral $D_T(t,x)=\\int_t^{t+T}\\sigma(s,X(s;t,x))\\,ds$ must be bounded below by $m_T>0$ on the relevant support family $A(t)$. This quantity is what makes localized damping effective; the proof inserts it into the characteristic formula for $e(t+T,X(t+T;t,x))^2 J(t+T;t,x)$ and uses Liouville's formula to convert compression and expansion into the constants $M_A^-$ and the damping into $\\kappa m_T$. The constructive check is a Lyapunov-type function $\\Phi$ whose sublevel sets define the active region $V_d(t)=\\{x\\in\\Omega:\\Phi(t,x)\\le a_d\\}$; the entrance condition $D_u\\Phi\\le-\\gamma$ outside $V_d$ and $\\sigma\\ge\\sigma_{\\min}$ on $V_d$ guarantee every relevant characteristic spends enough time in the damped region, giving $m_T=\\sigma_{\\min}(T-T_{\\rm in})$.","core_discovery":"The central claim is Theorem IV.1: for a bounded $C^2$ domain with an impermeable velocity field $u$ satisfying Assumption II.4, if the localization function $\\sigma$ is such that $D_T(t,x)=\\int_t^{t+T}\\sigma(s,X(s;t,x))\\,ds\\ge m_T>0$ on the relevant forward-invariant support family $A(t)$, and if the feedback gain satisfies $2\\kappa m_T>M_A^-T$ (with $M_A^-$ the essential supremum of $(\\nabla\\cdot u)^-$ on $A(t)$), then the unperturbed closed-loop error equation is exponentially stable in $L^2(\\Omega)$: $\\|e(t)\\|^2\\le C_T e^{-\\alpha_T t}\\|e_0\\|^2$ with $\\alpha_T=2\\kappa m_T/T-M_A^-$. The proof obtains a one-step contraction by evaluating the error along characteristics, squaring, and using Liouville's formula to compare the $L^2$ norm at $t+T$ with the norm at $t$; the factor $e^{M_A^-T}$ bounds worst-case compression and $e^{-2\\kappa m_T}$ accounts for damping. The paper also shows that these conditions are checkable: a Lyapunov-type function $\\Phi$ whose sublevel sets form the active region and which decreases outside it gives a uniform entrance time $T_{\\rm in}$, hence $m_T=\\sigma_{\\min}(T-T_{\\rm in})$; a weighted Lyapunov functional $W_\\ell=\\tfrac12\\int e^{\\ell\\Phi}e^2\\,dx$ converts this into a differential Lyapunov inequality and an ISS estimate for additive perturbations; and the feedback flux realizing $\\nabla\\cdot J_d=\\kappa\\sigma e-r_\\star$ is constructed as a right inverse of the divergence on an actuator region. The same characteristic-damping argument is extended to velocity fields of the form $u[\\rho]=u_0+\\varepsilon b[\\rho]$ satisfying uniform finite-time entrance conditions.","pith_inferences":["Read as a design principle, the finite-time damping condition suggests choosing the active region to maximize the worst-case integral $D_T$ over the reachable support family; the gain threshold $2\\kappa m_T>M_A^-T$ then tells the minimum actuator strength a given geometry allows.","The framework's mass-nonpreserving feedback is naturally interpreted as pickup, drop-off, or depot exchange in logistics; a concrete extension would be to couple the transported density with an external buffer and test whether the same exponential rate survives with bounded exchange rates.","The regularity assumption excludes discontinuous velocity fields, so applying the idea to shock-forming conservation laws would require a fundamentally different argument: BV flows lack the bi-Lipschitz change of variables on which the characteristic contraction relies."],"forward_implications":["Exponential stabilization with explicit constants: under the characteristic damping and gain conditions, $\\|e(t)\\|^2\\le e^{M_A^-T}\\exp(-(2\\kappa m_T/T-M_A^-)t)\\|e_0\\|^2$, so the decay rate is computable from the geometry and the gain.","Support-restricted errors: if the initial error and perturbations live in a forward-invariant family $K(t)$, only the characteristics in that family need to accumulate damping; the global condition can be replaced by the local one.","Lyapunov entrance gives design rules: with $\\Phi$, $a_d$, $\\gamma$, and $\\sigma_{\\min}$, the active region $V_d=\\{\\Phi\\le a_d\\}$ guarantees $m_T=\\sigma_{\\min}(T-T_{\\rm in})$, and the gain threshold reduces to $2\\kappa\\sigma_{\\min}>M_K^-$, making the design checkable from $u$, $\\sigma$, and $K(t)$.","ISS and robustness: the same weighted Lyapunov functional provides an input-to-state estimate, so bounded additive perturbations produce bounded error with exponential decay after the perturbation stops.","Realizability: the required divergence can be generated by solving local Neumann problems on an actuator region $U_d$, with an interface flux satisfying the compatibility condition, so the closed-loop equation is actually implementable."],"supporting_citations":[{"why":"Supplies prior stability estimates for the continuity equation in $L^p$, providing the baseline that the present localized-damping result refines.","marker":"[14]"},{"why":"Establishes localized controllability of the continuity equation via velocity fields, the actuation alternative against which this paper's divergence-flux feedback is positioned.","marker":"[22]"},{"why":"Provides the weighted Lyapunov functional technique for hyperbolic systems that Theorem IV.5 adapts with the weight $e^{\\ell\\Phi}$.","marker":"[16]"},{"why":"Supplies the method-of-characteristics well-posedness theory and the solution representation used to define mild solutions and the characteristic flow.","marker":"[25]"},{"why":"Supports the transport-equation treatment and the characteristic and volume-transformation arguments used throughout the proof.","marker":"[26]"},{"why":"Provides the finite-time-decay to uniform-exponential-stability iteration step used to turn the one-step $T$-contraction into the global decay estimate.","marker":"[28]"},{"why":"Provides the continuous right inverse of the divergence operator used for flux realization in Section V.","marker":"[30]"}],"fun_headline_variants":["Characteristic damping yields exponential L2 stability","Localized flux feedback stabilizes transport PDEs exponentially","Finite-time damping condition ensures exponential transport decay","Flux feedback with damping along flow lines proves stabilization","Transport PDEs: interior damping leads to exponential control"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the carrying velocity field is regular enough—with bounded spatial derivatives and bounded divergence—and impermeable at the boundary, so that the characteristic flow is a reversible Lipschitz transformation of the domain and volumes transform by Liouville's formula; if the field is discontinuous or only of bounded variation, as happens for shocks, this geometric machinery and the theorem's conclusion stop applying.","fun_headline_variants_meta":{"raw":{"variants":["Characteristic damping yields exponential L2 stability","Localized flux feedback stabilizes transport PDEs exponentially","Finite-time damping condition ensures exponential transport decay","Flux feedback with damping along flow lines proves stabilization","Transport PDEs: interior damping leads to exponential control"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000642,"raw_usage":{"total_tokens":3052,"prompt_tokens":1145,"completion_tokens":1907,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":761,"completion_tokens_details":{"reasoning_tokens":1834}},"tokens_in":761,"tokens_out":1907,"duration_ms":15144,"temperature":1.0,"reasoning_tokens":1834,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T11:25:42.010352+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the closed-loop error equation $\\partial_t e+\\nabla\\cdot(ue)=-\\kappa\\sigma e$ on the unit disk with inward radial field $u(x)=-\\mu x$ and with $\\sigma$ supported in an annulus that the shrinking family of characteristics never reaches, for instance with $\\operatorname{supp}\\sigma$ outside $\\{|x|\\le e^{-\\mu t}R_0\\}$; then $D_T(t,x)=0$ on the relevant support, the gain condition $2\\kappa m_T>M_A^-T$ fails for every finite $\\kappa$, and the $L^2$ norm should grow like $e^{2\\mu t}$ rather than decay—directly contradicting any uniform exponential stabilization claim in that configuration.","supporting_citations":[{"cited_title":"Stability results for the continuity equation,","cited_arxiv_id":null,"evidence_quote":"Supplies prior stability estimates for the continuity equation in $L^p$, providing the baseline that the present localized-damping result refines."},{"cited_title":"Approximate and exact controllability of the continuity equation with a localized vector field,","cited_arxiv_id":null,"evidence_quote":"Establishes localized controllability of the continuity equation via velocity fields, the actuation alternative against which this paper's divergence-flux feedback is positioned."},{"cited_title":"Bastin and J.-M","cited_arxiv_id":null,"evidence_quote":"Provides the weighted Lyapunov functional technique for hyperbolic systems that Theorem IV.5 adapts with the weight $e^{\\ell\\Phi}$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the method-of-characteristics well-posedness theory and the solution representation used to define mild solutions and the characteristic flow."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supports the transport-equation treatment and the characteristic and volume-transformation arguments used throughout the proof."},{"cited_title":"Uniform asymptotic stability of evolutionary processes in a banach space,","cited_arxiv_id":null,"evidence_quote":"Provides the finite-time-decay to uniform-exponential-stability iteration step used to turn the one-step $T$-contraction into the global decay estimate."},{"cited_title":"Solution of the first boundary value problem for the equation of continuity of an incompressible medium,","cited_arxiv_id":null,"evidence_quote":"Provides the continuous right inverse of the divergence operator used for flux realization in Section V."}],"review_version":1}