{"id":"55cf94ee-2bd5-4706-8a19-f9eba7ae4fc9","arxiv_id":"2505.23423","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For scalar elliptic transmission problems with piecewise Lipschitz coefficients, strong unique continuation holds at a C^{1,1} interface, proved via a new Carleman estimate and doubling inequality.","lead":"This paper proves strong unique continuation for elliptic equations with a piecewise Lipschitz coefficient that jumps across a smooth interface: a solution that vanishes faster than any power at an interface point must be zero everywhere. It obtains the result by deriving a doubling inequality from a new Carleman estimate with a singular weight.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Assumption (2.4) is not justified for C^{1,1} interfaces: the flattening that makes the normal direction vertical also makes the metric non-Lipschitz, so the Carleman estimate does not cover the claimed setting.","rationale":"The reader's weakest_assumption is the same one I identify as load-bearing: the unproved flattening condition (2.4). My stress-test strengthens it from 'no construction supplied' to 'standard constructions do not obviously deliver both (2.4) and the Lipschitz condition (2.2) for C^{1,1} interfaces'. If this concern lands, the Carleman estimate does not apply to the original piecewise-Lipschitz transmission problem with a C^{1,1} interface, so Theorem 4.5 and the SUCP conclusion are not established as stated. The paper may be repairable by strengthening the interface regularity, or by reformulating the estimate with a metric that is only bounded and modifying the small-σ estimates, but those are nontrivial changes rather than typographical fixes. I do not see evidence that the doubling inequality is false; the concern is that the stated assumptions do not cover the claimed application. The reader's CONDITIONAL verdict is therefore appropriate and I recommend no change to it. Secondary issues (the N^3/r1^3 exponent, the unproved lower-order terms, and the passage from (4.21) to (4.9) involving an unresolved ∫_{B_{4r}} term) are also present, but they are either weaker than the flattening problem or more readily fixable by rewriting constants and using a sharper Caccioppoli estimate.","tokens_in":18006,"tokens_out":42539,"duration_ms":437874,"concrete_test":"Take in R^2 the C^{1,1} interface x_2 = (1/2)x_1|x_1|, whose unit normal ν is Lipschitz but has a discontinuous derivative at x_1=0. Write any C^{1,1} flattening Φ satisfying (2.4) and compute the interface conditions: (2.4) forces ∂_{y_2}Φ(·,0)=±ν(·). Differentiating in y_1 then forces the coefficient of y_2 in ∂_{y_1}Φ to be ±ν'(·), which is a step-type L∞ function, not Lipschitz. Compute g=DFDF^T explicitly as y_2→0 and test the modulus of continuity of g entries; if the Lipschitz seminorm blows up, (2.2) and (2.4) are incompatible for this C^{1,1} interface.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The entire Carleman estimate (Theorem 2.1) and the doubling inequality depend on assumptions (2.1)–(2.4), in particular on g^{nk}(x',0)=0 for k<n and g^{nn}(x',0)=1 on the interface. These conditions are used critically to remove the interface boundary terms in (3.51), (4.11), and Lemma 4.1. The paper states that (2.4) is 'not restrictive' for the flattening change of variables, but gives no construction. This is not a minor omission: for a genuinely curved C^{1,1} interface, the standard graph-flattening map y=(x', x_n - f(x')) satisfies (2.4) only at the single point where the tangent plane is horizontal, not on an interface neighborhood. The signed-distance/Fermi construction that does satisfy (2.4) has ∂_{y_n}Φ equal to the unit normal ν at the interface. Since ν is only Lipschitz for a C^{1,1} interface, the mixed partial ∂_{y_i}∂_{y_n}Φ is forced to be ∂_{y_i}ν, which is merely L∞ and generally not continuous or Lipschitz. Consequently the first derivatives of Φ are not Lipschitz away from the interface, and the metric g=DFDF^T (up to scalar absorption) is only bounded, not Lipschitz. Thus assumptions (2.2) and (2.4) appear mutually incompatible for general C^{1,1} interfaces with non-C^{1} normal fields. If the intended regularity is C^{2,1}, that must be stated; otherwise a construction or a modified estimate avoiding (2.4) is required before the transmission problem of §1 falls under Theorem 4.5.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a Carleman estimate for the operator div(g^{-1}∇u) in a half ball with an interface on {x_n=0}, under structural assumptions on g that include a normalization condition (2.4) killing the normal-tangential coupling on the interface. From this estimate the authors derive a piecewise doubling inequality, Theorem 4.5, for solutions of the divergence-form equation div(ã∇_g U)=0, and they claim that, via the standard Garofalo-Lin argument [11], this yields the Strong Unique Continuation Property at an interface point for the scalar transmission problem with piecewise Lipschitz coefficient a(x)I across a C^{1,1} interface. The proof is carried out in full detail and the algebraic core of the Carleman estimate is written out with explicit constants.","tokens_in":18318,"tokens_out":29428,"duration_ms":296960,"significance":"If the assumptions are justified, the paper gives a substantial quantitative step on a known open problem: SUCP across an interface for discontinuous elliptic coefficients. The Carleman estimate and the doubling inequality are derived self-contained with explicit dependence on the parameters, and the connection to size estimates for unknown inclusions is a useful motivation. The paper cites weaker prior results and does not assume its own conclusion, so the derivation itself is genuine. However, the advertised applicability to C^{1,1} interfaces rests on an unproved flattening assertion, and the role of lower-order terms is unclear; these two points determine whether the main claim is actually established.","major_comments":[{"comment":"The introduction defines the operator Lu = div(A∇u)+b·∇u+cu and states that the paper proves SUCP at an interface point for solutions of Lu=0 when A=a(x)I and a is piecewise Lipschitz. The proof in Section 4, however, is carried out only for the divergence-form equation (4.7), div(ã∇_g U)=0, with no lower-order terms. No reduction, absorption argument, or density argument is supplied that would turn a doubling inequality for (4.7) into one for solutions of the equation with b·∇u+cu. Unless the announced result is explicitly restricted to b=c=0, the introductory claim is not supported by the theorems proved in the paper.","section":"Section 1 vs Section 4, equation (4.7)"},{"comment":"The assertion that assumption (2.4) is not restrictive is load-bearing and is not proved. The standard graph flattening y=(x', x_n-f(x')) produces a transformed coefficient whose normal-tangential entries are proportional to ∂f, so (2.4) holds only where ∇f=0, not on a genuinely curved C^{1,1} interface. A Fermi-type construction that imposes (2.4) forces the coordinate vector in the normal direction to be proportional to the unit normal ν; since ν is only Lipschitz for a C^{1,1} interface, the mixed tangential derivatives of the coordinate map are merely L∞ and can have jumps, which makes the metric coefficients discontinuous at positive normal distance unless the normal field is C^1. Thus it is not shown that assumptions (2.2) and (2.4) can hold simultaneously for the claimed application, and the exact cancellations in (3.51) and (4.11), on which the whole argument depends, may fail. The authors should either provide a construction, with regularity verified, or a precise reference for the flattening in the C^{1,1} class, or state the geometric result under a stronger interface regularity and adjust the Introduction accordingly.","section":"Section 2, assumption (2.4); uses in (3.51), (4.10)-(4.11)"}],"minor_comments":[{"comment":"There are several typos and small inconsistencies: 'restricitve' after (2.4), 'eistes' in Proposition 3.4, 'Pieciewise' in the heading of Theorem 4.5, and 'B±r' used without prior definition in Lemma 4.4.","section":"Throughout"},{"comment":"The inequality |H(σ)-εσ^ε| ≤ Cσ is stated 'for every σ ≥ 1', but it is used for small σ near 0; the intended range appears to be 0 < σ ≤ 1, and the text should be corrected.","section":"Section 3.1, display (3.36)"},{"comment":"The quantity N is called the 'largest frequency in the half ball', which is misleading; it is a global ratio of L² integrals. The proof actually yields a bound with N appearing only linearly before the final constant is enlarged; the stated N³/r₁³ form is safe but the relation between the choice of τ and the displayed N³ should be clarified.","section":"Theorem 4.5 and its proof"},{"comment":"The abstract speaks of 'piecewise Lipschitz coefficients' generally, while the theorem in Section 4 treats only the scalar case A(x)=a(x)I after flattening; the abstract and introduction should state clearly that the scalar case is the one treated.","section":"Abstract and Section 1"}],"recommendation":"major_revision","confidential_remarks":"The key scientific question is whether the flattening condition (2.4) can be realized for arbitrary C^{1,1} interfaces while preserving the Lipschitz regularity of g required by (2.2). If it cannot, the paper's advertised result for C^{1,1} transmission problems is not established, and the authors will need to either supply a valid construction or weaken/rewrite the claimed interface regularity. The algebraic core of the Carleman estimate appears sound and the paper is not circular; I would not reject outright, but the revision must resolve this load-bearing issue and clarify the lower-order terms before the paper can be considered for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here is my read. The paper does prove something real: a singular-weight Carleman estimate for a transmission problem with a jump across a flat interface, plus a doubling inequality for the model operator div(ã∇_g U)=0. That part is derived in detail, with explicit constants, and I did not find a hole in the core algebra. It is a genuine extension of the Escauriaza–Vessella route to the jump setting, and the interface boundary terms are handled carefully. If the metric g truly satisfies (2.1)–(2.4), I believe the doubling inequality (4.22) follows in essence.\n\nThe trouble is the advertised reduction to a C^{1,1} interface. The paper says assumption (2.4) is not restrictive and gives no construction. For a curved interface, the natural coordinate system that makes g^{nk}=0 and g^{nn}=1 at the interface is the Fermi/normal coordinate system. There the metric coefficients involve the second fundamental form times the signed distance, so their tangential derivatives are only L∞, not Lipschitz. The metric is not Lipschitz in a neighborhood of the interface. So (2.2) and (2.4) appear mutually incompatible for a general C^{1,1} interface, and the Carleman estimate as stated does not cover the original problem. If the intended regularity is C^{2,1}, that needs to be said explicitly; otherwise a modified estimate that tolerates the lower regularity is needed. This is the load-bearing gap.\n\nThere are two smaller issues. The introduction claims SUCP for the operator with b·∇u+cu, but the body only treats the principal part. That is an overclaim, though easily fixed. In Lemma 4.4 the vanishing of the interface term [ã∇g(ξU)·ν_g] is argued a bit quickly for n≥3; a density/regularity sentence would help. And in Theorem 4.5 the stated N^3/r1^3 bound does not directly come out of the calculation; I get N/r1^2. Minor, but the final constant should be checked.\n\nWould I accept this for review? Yes. The model Carleman estimate is worth referee time, and the flattening issue is exactly what a referee should probe. The paper is serious: it engages the literature, gives a full proof, and the flaw is a missing construction rather than a hidden circularity. I would not cite the C^{1,1} theorem yet, but I would cite the flat-interface estimate once the regularity question is settled.","headline":"Genuine new Carleman estimate for a flat-interface transmission problem, but the advertised C^{1,1} interface application rests on an unproven and likely false flattening assumption.","tokens_in":18902,"tokens_out":13625,"would_cite":false,"duration_ms":137612,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B60","35J15","35R30"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves a doubling inequality for solutions of an elliptic transmission problem with piecewise Lipschitz coefficients and derives strong unique continuation at the interface, a first step toward estimating the size of unknown…","keywords":["unique continuation","doubling inequality","Carleman estimate","transmission problem","interface","elliptic equations","inverse problems","Lipschitz coefficients"],"falsifier":"If the central claim is wrong, there should exist a nonzero solution of the scalar transmission problem, with a piecewise Lipschitz coefficient jumping across a $C^{1,1}$ interface, that vanishes of infinite order at an interface point but is not identically zero. A concrete way to look for it is to solve the two-dimensional problem numerically with a simple $C^{1,1}$ interface and compute the ratio $\\int_{B_{4r}} U^2/\\int_{B_{2r}} U^2$ along a sequence $r_j\\to 0$; Theorem 4.5 forces this ratio to remain bounded by $C N^3/r_1^3$ with $N$ fixed, so an unbounded ratio would disprove the doubling inequality and the strong unique continuation conclusion.","tokens_in":17762,"feed_emoji":"📏","tokens_out":12818,"duration_ms":126892,"temperature":0.7,"pith_summary":"The paper establishes a quantitative growth bound, called a doubling inequality, for solutions of elliptic equations whose coefficient jumps across a smooth interface, and uses it to prove the Strong Unique Continuation Property (SUCP) at interface points. SUCP means that a solution which vanishes faster than every power of the distance to a point must be identically zero. Such a property was open for scalar piecewise Lipschitz coefficients at an interface. The authors present the result as the first step toward an inverse problem: estimating the size of an unknown, merely measurable inclusion inside a conductor from boundary measurements. The main theorem bounds the $L^2$ mass on a ball of radius $4r$ by that on a ball of radius $2r$, with a constant depending on the frequency ratio $N$.","feed_headline":"At an interface, doubling inequality gives strong unique continuation","feed_subtitle":"Any solution that vanishes too fast at an interface point must be zero; a first step toward sizing unknown inclusions.","key_machinery":"The load-bearing object is the Carleman estimate of Theorem 2.1 for the operator $\\Delta_g=\\operatorname{div}(g^{-1}\\nabla)$ with a jump coefficient $\\gamma$. The singular weight is $w=\\psi(\\sigma)$, where $\\sigma=|x|$ and $\\psi(s)=s\\exp(-\\int_0^s dt/(t^{1-\\epsilon}(1+t^\\epsilon)))$, chosen so that $w\\sim\\sigma$ near zero while powers of $w$ provide the needed large-parameter growth. The proof combines Rellich's identity, applied with $B_v=v\\nabla_g v/|\\nabla_g v|^2$, with a decomposition of the conjugated operator into a symmetric part and an antisymmetric part; the antisymmetric part integrates to zero. On the interface $\\{x_n=0\\}$, assumption (2.4) makes $\\nabla_g w\\cdot\\nu_g=0$, which removes the normal boundary term, and the transmission condition $[\\tilde a\\,\\nabla_g U\\cdot\\nu_g]=0$ removes the remaining interface contribution. Together these cancellations turn the Carleman estimate into the doubling inequality of Theorem 4.5.","core_discovery":"On the paper's own terms, the central claim is Theorem 4.5: if $U$ solves $\\operatorname{div}(\\tilde a\\,\\nabla_g U)=0$ in $B_1$, with $\\tilde a=\\tilde a_+\\chi_{B_+^1}+\\tilde a_-\\chi_{B_-^1}$ positive and Lipschitz on each side of the flat interface $\\{x_n=0\\}$, and $g$ is the metric arising from flattening a $C^{1,1}$ interface, then $\\int_{B_{4r}} U^2 \\le \\frac{C N^3}{r_1^3}\\int_{B_{2r}} U^2$ for $0<r<r_1/16$, where $N=\\int_{B_{r_1}}U^2/\\int_{B_{r_1/4}}U^2$. The inequality is proved by a Carleman estimate with a singular weight, and the authors then invoke the known bridge from doubling inequalities to strong unique continuation to conclude SUCP at interface points. They also note that the doubling inequality implies that $|U|^2$ is an $A_p$ weight, the mechanism behind inclusion-size estimates, and frame the whole argument as a first step toward estimating the measure of an unknown measurable inclusion inside a conductor.","pith_inferences":["A natural continuation, not done in the paper, is to verify assumption (2.4) by writing the flattening map explicitly for a $C^{1,1}$ interface given as a graph; the paper asserts the assumption is non-restrictive but supplies no construction.","Because the proof uses the scalar structure only through the transmission condition, the same Carleman estimate may extend to matrix-valued conductivities that are scalar multiples with different factors on the two sides; testing this is a direct next step.","The doubling inequality's constant grows only polynomially in the frequency ratio $N$, so the size-estimate scheme for inclusions from boundary energy may run with merely measurable inclusions; this connection is not established in the present paper."],"forward_implications":["For scalar elliptic equations with piecewise Lipschitz conductivity jumping across a $C^{1,1}$ interface, the Strong Unique Continuation Property holds at every interface point.","The quantitative result is a doubling inequality: the $L^2$ mass on a ball of radius $4r$ controls the mass on the ball of radius $2r$, with a constant growing at most like the third power of the frequency ratio $N$ defined on a fixed outer ball.","Combining the doubling inequality with standard local regularity estimates makes $|u|^2$ an $A_p$ weight near the interface, following the known bridge in [11].","The authors position the theorem as a first step toward estimating the Lebesgue measure of an unknown, merely measurable inclusion inside a conductor from boundary measurements, replacing stricter geometric conditions used in earlier three-sphere approaches.","The theorem applies in every dimension $n\\ge 2$, with constants depending only on ellipticity, Lipschitz norms, and the lower bound of the coefficients."],"supporting_citations":[{"why":"Provides the flattening change of variables that turns a curved $C^{1,1}$ interface into the flat one where the Carleman estimate is proved.","marker":"[1]"},{"why":"Establishes an earlier Carleman estimate for elliptic equations with Lipschitz coefficients jumping at an interface, the transmission setting this paper extends.","marker":"[7]"},{"why":"Supplies the singular-weight Carleman method and the Rellich-identity decomposition that the proof adapts.","marker":"[8]"},{"why":"The cited bridge from any doubling inequality to strong unique continuation, used to draw the paper's main conclusion.","marker":"[11]"},{"why":"Gives the regularity for solutions on each side of the interface, needed to apply the estimate to cut-off solutions.","marker":"[15]"},{"why":"Frames the inverse problem of size estimates for unknown inclusions via three-region inequalities, which the new doubling inequality is designed to improve.","marker":"[9]"}],"fun_headline_variants":["Doubling inequality proves strong unique continuation across interface","Carleman with singular weight yields doubling inequality for SUCP","First step: doubling inequality for strong unique continuation at jumps","From Carleman to doubling: SUCP at interface for elliptic transmission","Sizing unknown inclusions starts with a doubling inequality"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes that after flattening the interface one can choose coordinates in which, at the interface, the normal direction is completely decoupled from the tangential directions, with the normal-normal component scaled to 1; the paper asserts this is non-restrictive without giving the coordinate construction, and this decoupling is what kills the interface boundary terms.","fun_headline_variants_meta":{"raw":{"variants":["Doubling inequality proves strong unique continuation across interface","Carleman with singular weight yields doubling inequality for SUCP","First step: doubling inequality for strong unique continuation at jumps","From Carleman to doubling: SUCP at interface for elliptic transmission","Sizing unknown inclusions starts with a doubling inequality"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.0003,"raw_usage":{"total_tokens":1698,"prompt_tokens":878,"completion_tokens":820,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":494,"completion_tokens_details":{"reasoning_tokens":739}},"tokens_in":494,"tokens_out":820,"duration_ms":8827,"temperature":1.0,"reasoning_tokens":739,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:48:22.524196+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"If the central claim is wrong, there should exist a nonzero solution of the scalar transmission problem, with a piecewise Lipschitz coefficient jumping across a $C^{1,1}$ interface, that vanishes of infinite order at an interface point but is not identically zero. A concrete way to look for it is to solve the two-dimensional problem numerically with a simple $C^{1,1}$ interface and compute the ratio $\\int_{B_{4r}} U^2/\\int_{B_{2r}} U^2$ along a sequence $r_j\\to 0$; Theorem 4.5 forces this ratio to remain bounded by $C N^3/r_1^3$ with $N$ fixed, so an unbounded ratio would disprove the doubling inequality and the strong unique continuation conclusion.","supporting_citations":[{"cited_title":"Adolfsson L","cited_arxiv_id":null,"evidence_quote":"Provides the flattening change of variables that turns a curved $C^{1,1}$ interface into the flat one where the Carleman estimate is proved."},{"cited_title":"Di Cristo, E","cited_arxiv_id":null,"evidence_quote":"Establishes an earlier Carleman estimate for elliptic equations with Lipschitz coefficients jumping at an interface, the transmission setting this paper extends."},{"cited_title":"Escauriaza and S","cited_arxiv_id":null,"evidence_quote":"Supplies the singular-weight Carleman method and the Rellich-identity decomposition that the proof adapts."},{"cited_title":"Garofalo, F","cited_arxiv_id":null,"evidence_quote":"The cited bridge from any doubling inequality to strong unique continuation, used to draw the paper's main conclusion."},{"cited_title":"Ladyzhensaya and N","cited_arxiv_id":null,"evidence_quote":"Gives the regularity for solutions on each side of the interface, needed to apply the estimate to cut-off solutions."},{"cited_title":"Francini, C.-L","cited_arxiv_id":null,"evidence_quote":"Frames the inverse problem of size estimates for unknown inclusions via three-region inequalities, which the new doubling inequality is designed to improve."}],"review_version":1}