{"id":"a8138cbf-1a37-49ee-8cf6-67159f4dd417","arxiv_id":"2411.11573","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Log-type Hausdorff contents, chosen so that their inverse matches the heat kernel, give sharp sufficient conditions and a critical-scale counterexample for heat-equation observability.","lead":"This paper proves new observability inequalities for heat equations, showing that solutions can be recovered from measurements on very thin sets, including certain sets of dimension d-1. The key tool is a new family of log-type Hausdorff contents tuned to the heat kernel, and in one dimension the scale is shown to be optimal.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 3.8 applies Lemma 3.7 with the wrong set inclusion: (3.23) forces δ_n ≤ δ, so the bound is for a smaller level set, not the one needed in the Remez and observability chain.","rationale":"The paper's central positive results all rest on the quantitative upper bound for the log-type Hausdorff content of polynomial lemniscates, Lemma 3.8. This lemma is used to prove the Remez inequality, the propagation of smallness, the spectral inequality, and finally the observability inequalities. The proof of Lemma 3.8 invokes Lemma 3.7, which bounds the content of the level set at δ_n = (g_n(H)/4)^n. For this to give a bound at the desired level δ, one must have δ ≤ δ_n so that the target set is contained in the level set appearing in Lemma 3.7. However, the derivation in the paper establishes a lower bound on log(1/H) in (3.23), which makes δ_n exponentially smaller than δ, as shown by the substitution into the lower bound for δ_n. Thus the set inclusion is reversed: Lemma 3.7 controls a strictly smaller lemniscate, not the one required. The paper does not reconcile this direction nor prove the existence of a suitable H; the two necessary inequalities on log(1/H) move in opposite directions. This is a genuine, load-bearing gap in the proof, and it affects the 1D and d-dimensional theorems alike. The reader's verdict of CONDITIONAL is therefore appropriate: the statements may be true, but the current argument does not justify them. No independent verification or alternative proof of Lemma 3.8 is supplied. A concrete check on the model case P(z)=z^n would confirm whether the monotonicity issue is real and whether a repair is possible.","tokens_in":47916,"tokens_out":14018,"duration_ms":115852,"concrete_test":"Re-derive the application of Lemma 3.7 in Lemma 3.8 for the model polynomial P(z)=z^n with n=10, δ=e^{-100}, α=2. Compute δ_n(H) explicitly from (3.12) for H satisfying (3.23), and verify whether there exists any H ∈ (0,e^{-3}] such that δ ≤ δ_n(H). If no such H exists, the proof as written fails. Alternatively, write down the actual set inclusion obtained from (3.23): check whether it yields E(P;δ) ⊂ E(P;δ_n) or the reverse; only the former would justify the passage from Lemma 3.7 to (3.17).","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 3.8 is the foundational step: it feeds the Remez inequality (Lemma 3.10), propagation of smallness (Lemma 3.11), the spectral inequality (Lemma 4.1), and ultimately Theorems 2.1 and 2.8. Lemma 3.7 states chα(P;δ_n) ≤ h(H) for δ_n = (g_n(H)/4)^n, where (3.12) defines δ_n. To upper-bound chα(P;δ) for an arbitrary δ, one needs δ ≤ δ_n, because the level set {|P|≤δ^n} is increasing in δ and the Hausdorff content is monotone under set inclusion. The proof, however, derives (3.23): log(1/H) ≥ log(1/(4δ))/(2n A^α) − 1. Substituting this into the lower bound δ_n ≥ 4^{-n} exp{−n log(1/H)(2n A^α − 1)} gives δ_n ≥ 4^{-n} e^{−S} = δ/4^n, where S = log(1/δ). Thus δ_n is typically far smaller than δ, not larger. Consequently E(P;δ_n) ⊂ E(P;δ), and Lemma 3.7 only controls a smaller set; it cannot yield the desired upper bound on chα(P;δ). The text never proves the existence of H satisfying both δ ≤ δ_n and the h(H) bound used in (3.27); indeed the direction in (3.23) makes the two constraints incompatible for large n. Since every later statement depends on Lemma 3.8, the central observability theorems are not established by the present proof.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces log-type Hausdorff contents associated with gauges hα(t) = (log 1/t)^(-1/2) (log log 1/t)^(-α) in one dimension and Fα,β(t) = t^(d-1) (log 1/t)^(-β) (log log 1/t)^(-α) in higher dimensions. It claims that positive content at these gauges implies observability inequalities for the heat equation on bounded intervals and domains (Theorems 2.1 and 2.8), that thickness at these scales implies observability on R and Rd (Theorems 2.5 and 2.11), and that in one dimension the gauge (log 1/t)^(-1/2) is critical: for every ε>0 there is a non-observable set with positive (log 1/t)^(-1/2+ε)-content (Theorem 2.1(ii)). The proof strategy follows a Lebeau-Robiano chain: a Cartan/Lubinsky level-set estimate for monic polynomials (Lemma 3.8), a Remez inequality (Lemma 3.10), propagation of smallness for analytic functions (Lemma 3.11), a spectral inequality (Lemma 4.1) or Logvinenko-Sereda type uncertainty principle (Lemma 4.5), and a summation argument. The higher-dimensional case adds capacity-based slicing (Lemmas 5.8 and 5.14) to reduce to the one-dimensional propagation result.","tokens_in":48281,"tokens_out":35262,"duration_ms":309509,"significance":"If the results are correct, they would give a substantial extension of observability theory for heat equations: observation sets of Hausdorff dimension d-1 are covered, with an explicit log-log correction that the authors identify as sharp in one dimension. The paper is ambitious and detailed: it provides a complete chain of tools, including an explicit counterexample at the critical gauge, and it makes heavy but transparent use of classical results from fractal geometry and approximation theory. These are valuable strengths. However, the foundational polynomial level-set lemma has a proof gap that is load-bearing for the entire chain, and there is a significant ambiguity in the statement of the optimality result. Because the main theorems depend on Lemma 3.8, the paper cannot be accepted in its present form.","major_comments":[{"comment":"Lemma 3.7 gives the bound c_f(P; δ_n(H)) ≤ f(H), where δ_n(H) = 4^(-n) exp{ n(log H - (1/f(H)) ∫ ... ) }. Since the level sets E(P;ε) = {|P| ≤ ε^n} are increasing in ε, applying Lemma 3.7 to the level δ requires δ ≤ δ_n(H). The proof of Lemma 3.8, however, derives only the lower bound (3.23), log(1/H) ≥ log(1/(4δ))/(2nA^α - 1). Substituting this into the lower bound δ_n ≥ 4^(-n) exp{-(2n^2A^α - n) log(1/H)} gives δ_n ≥ 4^(-n) e^(-n log(1/δ)) = (δ/4)^n, which is much smaller than δ and does not imply δ ≤ δ_n(H). The condition needed for Lemma 3.7 has the opposite direction: one would need log(1/H) ≤ (log(1/δ) + n log 4)/(2n^2A^α - n). Thus Lemma 3.7 is applied to a smaller level set, and (3.17) is not established. Since Lemma 3.8 feeds Lemma 3.10, Lemma 3.11, Lemma 4.1 and hence Theorems 2.1 and 2.5, and via Corollary 3.14 also the higher-dimensional results, this gap affects the central claim of the paper.","section":"3.2, Lemma 3.8"},{"comment":"There is a mismatch between the stated optimality theorem and the gauge actually used in the proof. Proposition 4.8 defines f_ε(t) = (log 1/t)^(-1/2+ε), and the mass estimate in Case 1 yields μ(B(x,r)) ≲ (log 1/r)^(-(1-ε2)/(2+ε1)) = (log 1/r)^(-1/(2+ε)) by (4.45). This proves positivity of the (log 1/t)^(-1/(2+ε))-content, a gauge that is larger than (log 1/t)^(-1/2) because -1/(2+ε) > -1/2. It does not prove positivity of the F_{0,1/2+ε}-content, i.e. (log 1/t)^(-(1/2+ε)), which is a smaller gauge. If the intended statement of Theorem 2.1(ii) is the weaker exponent -1/2+ε, then the Introduction's reference to 'positive F_{0,1/2+ε}-Hausdorff content' should be corrected, for instance to F_{0,1/2-ε}. If the intended statement is the stronger one with F_{0,1/2+ε}, then Proposition 4.8 does not prove Theorem 2.1(ii) and a new construction would be needed. Since the critical-gauge claim is advertised as a main result, this ambiguity should be resolved.","section":"2.1 / 4.3 / Proposition 4.8"}],"minor_comments":[{"comment":"The exponent in (4.62) is typeset as -1/2+ε1, but the intended expression is almost certainly -1/(2+ε1); the missing parentheses make the displayed equivalence incorrect as written.","section":"4.3, Eq. (4.62)"},{"comment":"In (5.13), the term H^d(E_k) should be H^(d-1)(E_k), since the integration is over a ⊂ l⊥ and the measure is (d-1)-dimensional Hausdorff measure.","section":"5.1, Lemma 5.8 proof"},{"comment":"The proof of Lemma 5.14 says 'the application of the conclusion (ii) in Lemma 5.14'; this should refer to Lemma 5.11(ii), not to Lemma 5.14 itself.","section":"5.2, Lemma 5.14 proof"},{"comment":"There are several typos that should be corrected: 'when when' in Remark 4.6, 'moinc' in Remark 2.6, 'dicussed' in Remark 2.9, and a stray '/suppress' in Theorem 6.1.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The most serious issue is the direction gap in Lemma 3.8; it is not a cosmetic slip because the polynomial level-set estimate is the first step of the whole chain. If the authors can repair this lemma, or replace it by a correct estimate and propagate the resulting constants through Lemma 3.10, Lemma 3.11 and the spectral/uncertainty inequalities, the paper may become acceptable. The gauge ambiguity in Theorem 2.1(ii) also needs to be fixed. The paper is otherwise careful and detailed, and the explicit counterexample construction is a genuine asset."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it. Bottom line: this is a serious, substantial paper, and the central claim—observability for heat equations from sets of positive log-type Hausdorff content, including dimension d-1—is new and worth refereeing. The 1D sharp gauge (log 1/t)^{-1/2} with matching counterexample is the cleanest part; the Liouville/Cantor constructions in Section 3 and the counterexample in Proposition 4.8 look right to me. The overall strategy is coherent: Cartan/Lubinsky for polynomial level sets, Remez, propagation of smallness, adapted Lebeau-Robiano. Credit where due: the proofs are detailed, the comparison with Green et al., Burq-Moyano, Le Balc'h-Martin, Dolecki/Samb is honest, and the acknowledgment that [22] caught an error in Proposition 5.17 is a good sign, though it means the d-dimensional section has had at least one real bug.\n\nThe soft spot is the one the stress-test flagged, but I'd put it differently. Lemma 3.7 as printed says c_f(P;(g_n/4)^n)≤f(H), with c_f(P;ε) defined as c_f({|P|≤ε^n}). Read literally, that's a bound for level (g_n/4)^{n^2}; the Cartan proof actually gives a bound for level (g_n/4)^n. If that's a typo and Lemma 3.7 should read c_f(P; g_n/4), then Lemma 3.8 works: (3.23) implies δ_n≥δ^n, i.e. δ≤g_n/4, so the desired level δ^n sits inside the controlled lemniscate. The present text doesn't say this, and the direction of (3.23) confused even a careful reader. That needs to be fixed explicitly; as written, the chain from Lemma 3.8 to Theorem 2.1 has a hole.\n\nThe d-dimensional part is heavier and less transparent: the capacity-content transfer (Lemma 5.14), the integrability condition (5.28), and the slicing argument rely on constants and small print. Given the admitted previous error in Proposition 5.17, I'd want a referee to go through Section 5 line by line. I didn't find circularity; the gauge functions are tailored but the positive and negative results are independent.\n\nWho is this for? People working on quantitative unique continuation and parabolic control. It deserves a serious referee, not a desk reject. I'd send it out, with a specific request to check Lemma 3.7/3.8 notation and the d-dimensional capacity estimates.","headline":"Serious, substantial paper on heat-equation observability from log-type Hausdorff content; the central 1D result looks right, but Lemma 3.7/3.8 has a notation/direction problem that needs fixing before the chain is airtight.","tokens_in":48830,"tokens_out":13303,"would_cite":true,"duration_ms":115427,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35K05","35B60","93B07","28A78"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that log-type Hausdorff content, not ordinary dimension, marks the critical scale for observability of heat equations, with (log 1/t)^{-1/2} as the sharp boundary in one dimension.","keywords":["observability inequality","heat equation","log-type Hausdorff content","gauge function","spectral inequality","propagation of smallness","critical gauge","null controllability"],"falsifier":"Take the monomial $P(z)=z^n$ in Lemma 3.8: the sublevel set $\\{|z|\\le\\delta\\}$ has explicit $h_\\alpha$-content, so one can directly compare the left-hand side of (3.17) with the claimed right-hand side across $\\delta$, $n$, and $\\alpha$; a single violation would break the chain, and independently, verifying the direction of (3.23) numerically for arbitrary $H$ would settle whether the proof's step is valid.","tokens_in":47700,"feed_emoji":"🔥","tokens_out":11381,"duration_ms":106901,"temperature":0.7,"pith_summary":"This paper claims that observability of the heat equation is governed by a logarithmically corrected notion of size rather than by ordinary Hausdorff dimension. In one spatial dimension, any observation set with positive $h_\\alpha$-Hausdorff content, where $h_\\alpha(t)=(\\log 1/t)^{-1/2}(\\log\\log 1/t)^{-\\alpha}$ and $\\alpha>3/2$, determines the full solution at later times, and this is sharp: for every $\\varepsilon>0$ there is a set with positive $(\\log 1/t)^{-1/2+\\varepsilon}$-Hausdorff content for which observability fails. In $d\\ge2$, the corresponding gauge is $F_{\\alpha,\\beta}(t)=t^{d-1}(\\log 1/t)^{-\\beta}(\\log\\log 1/t)^{-\\alpha}$ with $\\beta=3/2$ and $\\alpha>5/2$, and observability holds on bounded domains and on $\\mathbb{R}^d$ under an $F_{\\alpha,\\beta}$-thickness condition. If the paper is right, observation sets of Hausdorff dimension exactly $d-1$, and even zero-dimensional sets in one dimension, can be used to recover heat solutions and, by duality, to control them.","feed_headline":"Log-type Hausdorff content sets the boundary for heat observability","feed_subtitle":"A logarithmic refinement of fractal size marks the exact threshold where observations of heat flow succeed or fail.","key_machinery":"The argument runs through a five-step chain. A Cartan-lemma-based estimate (Lemma 3.8) bounds the $h_\\alpha$-Hausdorff content of the set where a monic polynomial of degree $n$ is at most $\\delta^n$; this produces a Remez inequality at the $h_\\alpha$ scale, with constants $e^{Cn^2(\\log(n+e))^{-2\\alpha/3}/c_{h_\\alpha}(E)^2}$ replacing the usual $(C/|E|)^n$. The Remez inequality gives a quantitative propagation of smallness for analytic functions, which yields the spectral inequality in one dimension and a Logvinenko-Sereda-type uncertainty principle on the whole line; the adapted Lebeau-Robiano summation then converts these into observability, and convergence of that summation is what forces $\\alpha>3/2$. In higher dimensions the same chain is lifted by a capacity-based slicing lemma and a quantitative transfer between Hausdorff content and capacity, which is why the required exponent weakens to $\\alpha>5/2$.","core_discovery":"The central claim is that a heat-kernel-adapted log-type Hausdorff content, not the usual power-law content, is the correct yardstick for observability. The one-dimensional theorem states that $c_{h_\\alpha}(E)>0$ with $\\alpha>3/2$ implies the observability inequality $\\|u(T,\\cdot)\\|_{L^2(0,L)}\\le C_{obs}\\int_0^T \\sup_{x\\in E}|u(t,x)|\\,dt$ for every solution of the Dirichlet heat equation on an interval, with $C_{obs}$ depending only on $T$, $L$, and the content. The companion sharpness result shows that gauges of the form $(\\log 1/t)^{-1/2+\\varepsilon}$ are not sufficient for observability, so the exponent $-1/2$ is the critical log scale. In higher dimensions the paper establishes observability for sets of positive $F_{\\alpha,\\beta}$-content with $\\beta=3/2$ and $\\alpha>5/2$, and on the whole space for sets that are thick at the scale of the same content. Along the way it derives spectral inequalities and uncertainty principles at these log-type scales, including a spectral inequality whose growth rate differs from the classical $e^{C\\sqrt{\\lambda}}$ rate.","pith_inferences":["Beyond the paper, the same mechanism suggests that for any parabolic equation with a Gaussian fundamental solution, the critical observability gauge should be tied to the kernel's decay profile; the heat kernel's $e^{-x^2}$ tail is what fixes $(\\log 1/t)^{-1/2}$.","The gap between the one-dimensional threshold $\\alpha>3/2$ and the higher-dimensional $\\alpha>5/2$ appears to be an artifact of routing the proof through capacity and slicing; a direct higher-dimensional level-set estimate for monic polynomials would likely lower the threshold.","A testable extension is to replace the Laplacian by fractional or degenerate elliptic operators and check whether the same log-type gauge, with exponents adjusted by the kernel's heat decay, still marks observability.","The sharpness construction, which plants near-zeros of the eigenmodes $\\sin(n\\pi x)$ into the observation set, gives a template for producing failure sets for other self-adjoint parabolic semigroups."],"forward_implications":["In one dimension, observable sets can have Hausdorff dimension zero, as with the Liouville-type set $E_\\infty$; the classical threshold of positive Lebesgue measure is replaced by a log-content threshold.","The critical gauge is tied to the heat kernel: the inverse gauge satisfies $F^{[-1]}(1/x)=e^{-x^2}$, and adding the $(\\log\\log 1/t)^{-\\alpha}$ factor is exactly what makes the Lebeau-Robiano series converge.","In $d\\ge2$, the $F_{\\alpha,\\beta}$-content admits observable sets of Hausdorff dimension exactly $d-1$, while nodal sets such as $\\{x:\\varphi_\\lambda(x)=0\\}$ remain non-observable at dimension $d-1$; the log-type content separates these cases.","On $\\mathbb{R}^d$, observability holds for sets that are thick at the log-content scale, extending the classical thick-set characterization to zero-measure sensor sets.","By duality, each observability inequality yields null controllability with controls supported on these thin sets, with control cost bounded by the observability constant."],"supporting_citations":[{"why":"supplies the adapted Lebeau-Robiano strategy that converts the spectral inequality into the observability inequality.","marker":"[14]"},{"why":"supplies the estimate on small values of monic polynomials from which Lemma 3.8 is derived.","marker":"[35]"},{"why":"is the classical Cartan lemma underlying the polynomial sublevel-set covering.","marker":"[7]"},{"why":"established observability from positive (d-1+delta)-Hausdorff content and is the prior result this paper refines.","marker":"[5]"},{"why":"proved observability from positive (d-1+delta)-content for arbitrarily small delta and provides the comparison baseline for sharpness.","marker":"[22]"},{"why":"supplies the quantitative propagation-of-smallness technique and the Lebesgue-measure comparison used in the one-dimensional proof.","marker":"[27]"},{"why":"is the source of the capacity-based slicing theorem and mass-distribution tools used in higher dimensions.","marker":"[38]"},{"why":"provides the integral-geometric capacity slicing bound used in Lemma 5.8.","marker":"[37]"},{"why":"contains the earlier qualitative Hausdorff-content versus capacity connection that Lemma 5.11 makes quantitative.","marker":"[6]"},{"why":"supplies the local estimate for exponential polynomials used in proving the one-dimensional spectral inequality.","marker":"[40]"}],"fun_headline_variants":["Heat observability tied to log-type Hausdorff size","Sharp log-scale threshold for heat observations","Log-Hausdorff content decides heat observability","Optimal log gauge for heat equation observations","Critical log-exponent for heat observability"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is Lemma 3.8, the estimate that bounds the log-type Hausdorff content of the set where a monic polynomial is small; if that estimate fails, the Remez inequality, propagation of smallness, spectral inequality, and observability theorems collapse, and the proof as written does not explicitly reconcile the direction of the key inequality in the step leading to (3.23), while in $d\\ge2$ the capacity-content transfer (Lemma 5.14) with its integrability condition (5.28) is also load-bearing.","fun_headline_variants_meta":{"raw":{"variants":["Heat observability tied to log-type Hausdorff size","Sharp log-scale threshold for heat observations","Log-Hausdorff content decides heat observability","Optimal log gauge for heat equation observations","Critical log-exponent for heat observability"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000304,"raw_usage":{"total_tokens":1861,"prompt_tokens":1175,"completion_tokens":686,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":791,"completion_tokens_details":{"reasoning_tokens":616}},"tokens_in":791,"tokens_out":686,"duration_ms":6969,"temperature":1.0,"reasoning_tokens":616,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T18:26:13.467866+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the monomial $P(z)=z^n$ in Lemma 3.8: the sublevel set $\\{|z|\\le\\delta\\}$ has explicit $h_\\alpha$-content, so one can directly compare the left-hand side of (3.17) with the claimed right-hand side across $\\delta$, $n$, and $\\alpha$; a single violation would break the chain, and independently, verifying the direction of (3.23) numerically for arbitrary $H$ would settle whether the proof's step is valid.","supporting_citations":[{"cited_title":"Duyckaerts and L","cited_arxiv_id":null,"evidence_quote":"supplies the adapted Lebeau-Robiano strategy that converts the spectral inequality into the observability inequality."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the estimate on small values of monic polynomials from which Lemma 3.8 is derived."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"is the classical Cartan lemma underlying the polynomial sublevel-set covering."},{"cited_title":"Burq and I","cited_arxiv_id":null,"evidence_quote":"established observability from positive (d-1+delta)-Hausdorff content and is the prior result this paper refines."},{"cited_title":"Walton Green, K´ evin Le Balc’h, J´ er´ emy Martin, and Marcu-Antone Orsoni","cited_arxiv_id":null,"evidence_quote":"proved observability from positive (d-1+delta)-content for arbitrarily small delta and provides the comparison baseline for sharpness."},{"cited_title":"Kovrijkine","cited_arxiv_id":null,"evidence_quote":"supplies the quantitative propagation-of-smallness technique and the Lebesgue-measure comparison used in the one-dimensional proof."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"is the source of the capacity-based slicing theorem and mass-distribution tools used in higher dimensions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the integral-geometric capacity slicing bound used in Lemma 5.8."},{"cited_title":"Carleson","cited_arxiv_id":null,"evidence_quote":"contains the earlier qualitative Hausdorff-content versus capacity connection that Lemma 5.11 makes quantitative."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the local estimate for exponential polynomials used in proving the one-dimensional spectral inequality."}],"review_version":1}