{"id":"985ec7c5-ca3d-4bee-9102-dd91d2e256a4","arxiv_id":"1908.08427","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper gives quadratic-form reconstruction formulas for the boundary value and normal derivative of the conductivity from the Dirichlet-to-Neumann map, and deduces bulk uniqueness for conductivities in W^{1+(n-5)/(2p)+,p}, n>=5.","lead":"This math paper proposes a way to recover the conductivity and its normal derivative on the boundary of a body from boundary electrical measurements, and uses this to prove a uniqueness result for rough conductivities in high dimensions. A specialist should know that the main proof depends on a boundary estimate whose derivation appears to contain a false step for low-regularity conductivities.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reader's objection to inequality (20) misses the paper's dyadic summation convention; the estimate is valid by a geometric-series/Hölder argument, so Theorem 5 does not collapse.","rationale":"I read the paper in good faith and focused on the reader's weakest assumption: the boundary trace estimate in Theorem 5, specifically inequality (20). The reader's critique asserts that (20) is false for p>2 and 1/p<s<1-1/p, with a counterexample based on a Besov block sequence. However, the paper's convention is that all frequency sums are over dyadic λ=2^k. For dyadic sums, the tail in (20) satisfies a geometric decay bound: after Hölder, the norm of the decaying weight sequence is finite for every positive s−1/p, and the exponent matches M^{1/p−s}. The proposed counterexample has a convergent dyadic tail and is not a counterexample. The later arguments in Theorem 5 and Theorem 2 rely only on dyadic Littlewood-Paley blocks, so the reader's non-dyadic l^{p'} concern is irrelevant. I also scrutinized the reconstruction proof. The main estimates for A2 and the correction terms are dimensionally consistent and use Theorem 2/Collary 8 as stated. The only substantive gap I found is the parenthetical Gagliardo-Nirenberg reduction in §1.1 claiming one may assume p≥2; as written this is incorrect for p<2. But a one-line fix using boundedness of γ converts the L^2 average to the L^1 average, so the range 1<p<2 in Theorem 1(A) is recoverable. This does not undermine the central construction. Since the reader's load-bearing objection does not land, and the remaining issues are minor and repairable, the appropriate verdict is ACCEPT rather than REJECT.","tokens_in":10912,"tokens_out":49181,"duration_ms":484509,"concrete_test":"Take p=4, s=1/2, M=2^K with K=10, and define a_k=2^{-k/2}/(k+1) for k≥0, mirroring the reader's proposed counterexample. Compute L(K)=∑_{k≥K}2^{k/4}a_k and R(K)=2^{K(1/4-1/2)}(∑_{k≥K}2^{2k}a_k^4)^{1/4}. If L(K)/R(K) remains bounded as K grows (it decays like 1/K), then the dyadic version of (20) holds for this example; the same geometric/Hölder argument proves it for all Besov sequences, so the alleged failure of Theorem 5 does not occur.","verdict_should_be":"ACCEPT","load_bearing_attack":"The paper's Notations state that all sums over λ run over dyadic numbers λ=2^k. Under that convention, inequality (20) is true for every s>1/p, including the range 1/p<s<1-1/p singled out by the reader. Indeed, set a_k=|P_{2^k}f|_p and b_k=2^{ks}a_k, with δ=s-1/p>0. Then the tail is ∑_{k≥K}2^{k/p}a_k = ∑_{k≥K}2^{-kδ}b_k, and Hölder's inequality for sequences gives ≤ (∑_{k≥K}2^{-kδp'})^{1/p'} (∑_{k≥K}b_k^p)^{1/p} = C_δ 2^{-Kδ} ||f||_{B^{s,p}}, because ∑_{k≥K}2^{-kδp'} is a convergent geometric series for any δ>0. The proposed counterexample with ||P_λ f||_p=λ^{-s}(log λ)^{-1} corresponds to a_k=2^{-ks}/(k+1); its tail is ∑_{k≥K}2^{-kδ}/(k+1), which converges and is in fact bounded by C2^{-Kδ}/K, so it does not violate (20). Consequently the proofs of Theorem 5, Theorem 2, and the reconstruction formulas in Theorem 1 are not invalidated by the reader's objection. A minor gap remains in the terse Gagliardo-Nirenberg remark used for part A when p<2: the paper cannot simply replace p by p≥2. However, this is fixable either by using |γ−γ(0)|^2 ≤ C|γ−γ(0)| and the q=1 case of Theorem 2, or by interpolating to γ∈W^{ps/2,2}; thus it is not a load-bearing obstruction.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper addresses the Calderón inverse conductivity problem at the boundary. The main result, Theorem 1, shows that for a conductivity γ∈W^{s,p}(Ω) with s>1/p, the Dirichlet-to-Neumann map Λ_γ determines γ(y) for almost every boundary point y through a sequence of quadratic forms ⟨Λ_γ f_{0,h}, f_{0,h}⟩ = c_{0,h}γ(y)+o(1), and, under one additional derivative, determines the normal derivative B_ν log γ(y) through an expansion of order h. The reconstruction uses singular solutions in the spirit of Alessandrini, with error estimates based on a new boundary Lebesgue-point theorem (Theorem 2, proved via Theorem 5 and a Littlewood-Paley trace inequality). As a consequence, Theorem 3 states bulk uniqueness in W^{1+(n−5)/(2p)+,p}(Ω) for n≥5 and n≤p<∞, conditional on the companion result [11, Thm 4].","tokens_in":11259,"tokens_out":26571,"duration_ms":252184,"significance":"If correct, this is a substantial advance in rough-coefficient boundary determination: it recovers both the conductivity and its normal derivative at the boundary from the DN map under essentially minimal trace regularity, and it supplies the boundary step needed for the bulk uniqueness theorem in [11]. The proof is largely self-contained, and the new boundary Lebesgue-point estimate (Theorem 5) is a useful tool in its own right. I do not share the reader's rejection: the alleged failure of inequality (20) disappears once the paper's dyadic summation convention (Notations, p. 15) is respected. For dyadic λ=2^k, setting b_k=2^{ks}‖P_{2^k}f‖_p, the tail ∑_{k≥K}2^{k(1/p−s)}b_k is bounded by (∑ 2^{-k(s−1/p)p'})^{1/p'}‖f‖_{B^{s,p}}, a convergent geometric series for every p>1 and every s>1/p. The proposed continuous-λ counterexample is therefore not admissible. The only substantive weakness I found is a local, fixable gap in the treatment of 1<p<2 in Section 1.1.","major_comments":[],"minor_comments":[{"comment":"For 1<p<2, the sentence 'we can assume that γ∈W^{s,p}(Ω) for s>1/p and 2≤p<∞' is not justified as stated, because the integrability exponent cannot be increased without changing the regularity index. Since γ is also in L∞, interpolation gives γ∈W^{ps/2,2}(Ω) with ps/2>1/2, after which the p=2 case applies; alternatively, all estimates in this subsection can be run with the q=1 case of Theorem 2. Please make this reduction explicit.","section":"§1.1 (Gagliardo-Nirenberg reduction)"},{"comment":"The interpolation step in the trace inequality is too compressed: the endpoint estimate is proved for (q,p)=(1,r), and the reader must infer that r is the final p divided by the final q and that complex interpolation yields the constant λ^{1/p_final}. Spell out this step, since the current text appears to conclude the general case directly from the L^1 endpoint.","section":"Lemma 7"},{"comment":"The bulk uniqueness result depends on Theorem 4 of the companion preprint [11], whose proof is not reproduced here. The manuscript should state this dependence explicitly, and, if the journal requires it, either include a proof or mark the claim as conditional on [11]; Theorem 1 itself is independent of this dependence.","section":"Theorem 3"},{"comment":"The proof applies Theorem 2 to ∇logγ, whose regularity index is s−1, but Theorem 2 requires the index to be strictly below 1+1/p. Since W^{s,p}⊂W^{s',p} for s'<s, one should note before the application that s can first be reduced to a value below 2+1/p; without this observation the hypotheses of Theorem 2 appear incompatible with the condition s>1+1/p in Theorem 1(B).","section":"§1.2 / Theorem 2 compatibility"},{"comment":"The symbol p1 in Corollary 8 appears to be the Hölder conjugate p' but is never defined; please define it. Also, the displayed formula in Theorem 2 contains a spurious negative sign before 1/r^n, likely a typographical artifact.","section":"Corollary 8 / Notations"},{"comment":"The estimate (20) is valid under the paper's dyadic summation convention, but the proof should say at this display that the sum is over dyadic λ; a reader who ignores the convention will see a divergent continuous sum in the range 1/p<s<1−1/p for p>2.","section":"§2, estimate (20)"}],"recommendation":"minor_revision","confidential_remarks":"The reader's rejection rests on a misreading of the summation convention: inequality (20) is correct under the dyadic convention stated in the Notations, so the central proof does not collapse. The mathematical substance of the paper appears sound, and the only genuine proof gap I found is local and easily repaired. The main point to monitor is the dependence of Theorem 3 on the author's companion preprint [11]; if that preprint is not yet accepted, the bulk-uniqueness claim should be marked conditional. There are no concerns about novelty: the boundary reconstruction method using singular solutions for rough W^{s,p} conductivities is a clear contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The reader's central objection is wrong. Inequality (20) is valid because every sum over λ is over dyadic λ=2^k. Writing a_k=||P_{2^k}f||_p and δ=s-1/p>0, the tail is ∑_{k≥K}2^{k/p}a_k = ∑_{k≥K}2^{-kδ}(2^{ks}a_k) ≤ M^{-δ}||f||_{B^{s,p}} by Hölder. The proposed counterexample with a_k=2^{-ks}/k is in B^{s,p} and its tail actually decays like M^{-δ}/log M, so it is not a counterexample. The proof of Theorem 5, and hence Theorems 2 and 1, stands.\n\nThe paper's real contribution is extending boundary determination of the conductivity below the Lipschitz threshold (s>1/p, not just s≥1 as in Brown) and recovering the normal derivative for s>1+1/p (or s>3/p, p<2) using singular solutions instead of oscillatory ones. That is a genuine, useful step for Calderón theory, and the quadratic-form reconstruction formulas are explicit.\n\nThe soft spots are minor. The Gagliardo–Nirenberg comment used to reduce to p≥2 is telegraphic; the intended interpolation to W^{ps/2,2} works and gives the needed trace regularity. The bulk uniqueness Theorem 3 relies on the author's companion paper [11], which is not proved here—a referee should verify that result, but it is not circular. The dyadic-sum convention, once stated, is easy to miss, and the paper could make it louder.\n\nThis is a serious paper. I'd send it to a referee with confidence. It deserves to be published, probably after minor revisions for clarity.","headline":"The reader's objection to (20) is mistaken—dyadic summation makes it true—and the paper's boundary reconstruction is sound and worth refereeing.","tokens_in":11845,"tokens_out":12214,"would_cite":true,"duration_ms":110851,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R30","35J25","46E35"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper constructs explicit boundary functions whose Dirichlet-to-Neumann quadratic forms recover the conductivity and its normal derivative at almost every boundary point.","keywords":["electrical impedance tomography","Calderón problem","Dirichlet-to-Neumann map","conductivity boundary determination","Besov spaces","singular solutions","inverse boundary value problem"],"falsifier":"Inspect inequality (20) in the proof of Theorem 5 and test it with a Besov function built from Littlewood-Paley blocks of unit $L^p$ norm supported on disjoint thin slabs approaching the boundary. For $p>2$ and $1/p<s<1-1/p$, the left-hand side $\\sum_{\\lambda\\ge M}\\lambda^{1/p}\\|P_\\lambda f\\|_p$ grows like $M^{1-1/p-s}\\|f\\|_{s,p}$, whereas the right-hand side is a constant times $M^{1/p-s}\\|f\\|_{s,p}$; the inequality fails. If it fails, the proof of Theorem 2 has no replacement in the paper, and the error control in both reconstruction formulas collapses.","tokens_in":10627,"feed_emoji":"⚡","tokens_out":8331,"duration_ms":79679,"temperature":0.7,"pith_summary":"The paper claims that the Dirichlet-to-Neumann map of the conductivity equation determines both the conductivity and its normal derivative at almost every boundary point, through explicit boundary functions and explicit formulas. The proof builds highly concentrated singular solutions near a boundary point and shows that the quadratic form picks out the conductivity at that point, then uses a second family of solutions to pick out the normal logarithmic derivative at the next order in the scale parameter. Because knowing the normal derivative is exactly what is needed to extend a rough conductivity across the boundary, this boundary determination unlocks a bulk uniqueness theorem for conductivities in a specific Sobolev scale. If correct, the formulas give a direct, non-iterative route from boundary measurements to the first-order Taylor data of the unknown conductivity.","feed_headline":"Conductivity and its normal derivative read off from boundary data","feed_subtitle":"New explicit boundary potentials let current-potential pairs expose γ and its normal derivative.","key_machinery":"The central object is a family of approximate singular solutions built from $u_h(x)=(x_n+h)/|x+he_n|^n$ in a boundary chart with $0$ at the boundary, together with correction functions $r_h\\in H^1_0(\\Omega)$ solving $\\operatorname{div}(\\gamma\\nabla(u_h+r_h))=0$. The boundary traces $f_{0,h}=h^{n/2}u_h|_{\\partial\\Omega}$ concentrate like a delta at the boundary point as $h\\to 0$, and the quadratic form $\\langle\\Lambda_\\gamma f,f\\rangle$ naturally splits into a principal term carrying $\\gamma(y)$ and error terms controlled by the boundary Lebesgue-point estimate of Theorem 2. For the normal derivative, the test functions are replaced by $v_h=\\gamma^{-1/2}u_h$, which makes the leading correction term involve $\\nabla\\log\\gamma$ at $y$.","core_discovery":"Let $\\Lambda_\\gamma$ be the Dirichlet-to-Neumann map for $\\operatorname{div}(\\gamma\\nabla u)=0$ in a bounded Lipschitz domain $\\Omega$. Theorem 1 states that for almost every boundary point $y$ there exist boundary functions $f_{0,h}$ and conductivity-independent constants $c_{0,h}\\approx 1$ such that $\\langle\\Lambda_\\gamma f_{0,h}, f_{0,h}\\rangle = c_{0,h}\\gamma(y)+o(1)$ as $h\\to 0$; with one more derivative of regularity, further functions $f_{1,h}$ satisfy $\\langle\\Lambda_\\gamma f_{1,h}, f_{1,h}\\rangle - c_{0,h} = c_{1,h}\\,\\partial_\\nu\\log\\gamma(y)\\,h + o(h)$. The paper also proves the boundary approximation estimate (Theorem 2) that controls the error terms, and combines the boundary result with the author's companion theorem to obtain uniqueness of $\\gamma$ in $W^{1+\\frac{n-5}{2p}+,p}(\\Omega)$ for $n\\ge 5$ and $n\\le p<\\infty$.","pith_inferences":["Editorial extension: the same two-scale quadratic-form scheme could be iterated to read higher normal derivatives of $\\gamma$ at the boundary, giving a rough-conductivity analogue of the full Taylor expansion known for smooth conductivities.","Editorial extension: because the reconstruction is an explicit formula in $\\Lambda_\\gamma$ and the boundary geometry, the construction makes a stability analysis under operator-norm perturbations of $\\Lambda_\\gamma$ concrete, even though the paper does not quantify such stability.","Editorial extension: the companion uniqueness range $W^{1+\\frac{n-5}{2p}+,p}$ suggests that the boundary determination is the pacing step; if the supporting Besov trace estimate is strengthened, the derivative requirement in the bulk uniqueness theorem may improve."],"forward_implications":["At almost every boundary point, $\\gamma(y)$ and $\\partial_\\nu\\log\\gamma(y)$ can be computed from $\\Lambda_\\gamma$ by evaluating the quadratic form on the explicit families $f_{0,h}$ and $f_{1,h}$ and taking $h\\to 0$.","Because $c_{0,h}$ and $c_{1,h}$ do not depend on the conductivity, the same boundary functions work for every unknown conductivity; no adaptive choice is needed.","Knowing the normal derivative at the boundary gives the first-order matching condition needed to extend two conductivities outside $\\Omega$ with equal traces, the step required by the companion uniqueness proof.","For $n\\ge 5$ and $n\\le p<\\infty$, equality of Dirichlet-to-Neumann maps forces equality of conductivities in $W^{1+\\frac{n-5}{2p}+,p}(\\Omega)$."],"supporting_citations":[{"why":"Supplies the singular-solution technique the paper uses in place of oscillatory boundary data.","marker":"[1]"},{"why":"Gives the pointwise boundary recovery for the conductivity whose arguments are borrowed and extended to the normal derivative.","marker":"[3]"},{"why":"Establishes the smooth-case boundary determination that motivates recovering full boundary Taylor data from the Dirichlet-to-Neumann map.","marker":"[9]"},{"why":"Provides the pseudodifferential expansion of the Dirichlet-to-Neumann map showing where the normal derivative of the conductivity sits in the symbol.","marker":"[14]"},{"why":"The author's companion theorem converts the boundary determination proved here into uniqueness of the conductivity in the bulk.","marker":"[11]"},{"why":"Supplies the function-space estimates adapted for the boundary Lebesgue-point and trace arguments.","marker":"[12]"},{"why":"Provides the trace theory that fixes the lowest regularity at which boundary values of $\\gamma$ and $\\partial_\\nu\\gamma$ are well defined.","marker":"[10]"}],"fun_headline_variants":["Boundary data reveal conductivity and its normal derivative","Conductivity and its boundary slope from potential-current pairs","Boundary probes recover conductivity derivative for inverse problem","Uniqueness of conductivity from boundary measurements in n≥5"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire reconstruction rides on the boundary approximation estimate of Theorem 2: it is the only mechanism that turns the difference $\\gamma(x)-\\gamma(y)$ near a boundary point into a negligible error; if that estimate is not true in the stated range, the leading terms in the reconstruction formulas cannot be isolated.","fun_headline_variants_meta":{"raw":{"variants":["Boundary data reveal conductivity and its normal derivative","Conductivity and its boundary slope from potential-current pairs","Boundary probes recover conductivity derivative for inverse problem","Uniqueness of conductivity from boundary measurements in n≥5"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000211,"raw_usage":{"total_tokens":1359,"prompt_tokens":831,"completion_tokens":528,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":447,"completion_tokens_details":{"reasoning_tokens":464}},"tokens_in":447,"tokens_out":528,"duration_ms":5841,"temperature":1.0,"reasoning_tokens":464,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:42:53.233115+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Inspect inequality (20) in the proof of Theorem 5 and test it with a Besov function built from Littlewood-Paley blocks of unit $L^p$ norm supported on disjoint thin slabs approaching the boundary. For $p>2$ and $1/p<s<1-1/p$, the left-hand side $\\sum_{\\lambda\\ge M}\\lambda^{1/p}\\|P_\\lambda f\\|_p$ grows like $M^{1-1/p-s}\\|f\\|_{s,p}$, whereas the right-hand side is a constant times $M^{1/p-s}\\|f\\|_{s,p}$; the inequality fails. If it fails, the proof of Theorem 2 has no replacement in the paper, and the error control in both reconstruction formulas collapses.","supporting_citations":[{"cited_title":"Alessandrini","cited_arxiv_id":null,"evidence_quote":"Supplies the singular-solution technique the paper uses in place of oscillatory boundary data."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the pointwise boundary recovery for the conductivity whose arguments are borrowed and extended to the normal derivative."},{"cited_title":"Kohn and M","cited_arxiv_id":null,"evidence_quote":"Establishes the smooth-case boundary determination that motivates recovering full boundary Taylor data from the Dirichlet-to-Neumann map."},{"cited_title":"Sylvester and G","cited_arxiv_id":null,"evidence_quote":"Provides the pseudodifferential expansion of the Dirichlet-to-Neumann map showing where the normal derivative of the conductivity sits in the symbol."},{"cited_title":"The Bilinear Strategy for Calder\\'on's Problem","cited_arxiv_id":"1908.04050","evidence_quote":"The author's companion theorem converts the boundary determination proved here into uniqueness of the conductivity in the bulk."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the function-space estimates adapted for the boundary Lebesgue-point and trace arguments."},{"cited_title":"Marschall","cited_arxiv_id":null,"evidence_quote":"Provides the trace theory that fixes the lowest regularity at which boundary values of $\\gamma$ and $\\partial_\\nu\\gamma$ are well defined."}],"review_version":1}