{"id":"5f99fea8-2398-446c-b191-b4640a4da36b","arxiv_id":"2412.16066","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For almost every obstacle, the degenerate part of the free boundary in the fractional obstacle problem vanishes up to dimension 3 for every s in (0,1).","lead":"This paper proves that, for almost every obstacle, the free boundary in the fractional obstacle problem has no degenerate points in dimensions up to 3, for every fractional power s in (0,1). It also establishes new gaps in the possible growth rates of solutions, which sharpen the classification of singularities.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of the crucial frequency gap (2m, 2m+2s) in Proposition 5.10 is delegated to [FS24a] and only sketched; if the sign claim (5.3) fails for some s, the cleaning of Γ∗ and hence Theorem 1.3 collapse.","rationale":"The reader's weakest-assumption is the classification A_{1,s} from [FS18], which is indeed load-bearing for the dimension-reduction arguments in Proposition 6.1. I do not dispute that classification: it is a standard result. The more fragile point in the proof of the central claim is the new frequency-gap theorem, specifically Proposition 5.10 excluding frequencies in (2m, 2m+2s). The paper itself identifies this gap as crucial for the case s>1/2, and Theorem 1.6 relies on it. The proof of Proposition 5.10 is short and delegates the core sign claim to [FS24a], a preprint treating only s=1/2. The zero-counting argument for the s-dependent ODE (5.1) is plausible but not fully written, so an honest stress test should flag it. If the sign claim failed, the set Γ∗ would contain frequencies just above 2 (for s>1/2), invalidating the cleaning bound (7.3) and the tables in Proposition 8.2; the n≤3 statement would then not follow from the given proof. This does not show the theorem is false; it shows the conditional verdict is right and that a fuller proof of Proposition 5.10, or an independent verification of claim (5.3), is needed before the argument can be considered complete.","tokens_in":55414,"tokens_out":26633,"duration_ms":234310,"concrete_test":"Independently re-derive the endpoint sign claim (5.3) from the ODE (5.1) for general a∈(-1,1): show via Sturm comparison that for α∈(2m, 2m+2s), p_α(π/2) and ∂^a_θ p_α(π/2) are both nonzero and of opposite signs, and that the signs are constant on the whole interval. As a direct numerical check, integrate (5.1) for s∈{0.25, 0.5, 0.75}, m=1, at midpoints such as α=2.25, 2.5, and 2.75, and verify p_α(π/2) ∂^a_θ p_α(π/2) < 0 with the convention in (5.2); a violation would disprove the gap, while agreement would support Proposition 5.10 but not replace the analytical proof.","verdict_should_be":"UNCHANGED","load_bearing_attack":"For the main theorem when s>1/2, the introduction states that a crucial step is to know A_{n,s}∩(2, 2+2s+σ)=∅ for some small σ>0. In Theorem 1.6 this is obtained from Proposition 5.10 (no admissible frequencies in (2m, 2m+2s)) together with Proposition 5.1 and [Car24]. The proof of Proposition 5.10 is not fully self-contained: the key assertion is the sign claim (5.3), which is justified by saying 'arguing exactly as in [FS24a, Lemma 2]' and by a sketch about zero counting for solutions of the ODE (5.1). It is not proved that the number of zeros of p_α and ∂^a_θ p_α in [0,π/2] stays constant as α varies through (2m, 2m+2s) for the present a-dependent operator, nor that the endpoint signs change exactly when α crosses 2m and 2m+2s. Since [FS24a] treats s=1/2 (a=0), the extension to all a∈(-1,1) requires an independent verification. If an extra admissible frequency existed in (2, 2+2s), then for s>1/2 the set Γ∗ would contain frequencies below 2+2s+σ, so the cleaning estimate (7.3) would not apply; the splitting (1.13) has no other component to absorb such frequencies, and the generic dimension bound in Proposition 8.2 for n=3 would no longer be negative. Thus Proposition 5.10 is a load-bearing external dependency with a nontrivial sign claim that is only sketched.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves generic regularity for the fractional obstacle problem: for a family of solutions v(·,t) with obstacle φ−t, for almost every t the degenerate set is empty in dimensions n≤3 and has Hausdorff dimension at most n−3−α in dimensions n≥4, for every s∈(0,1) and general nonzero C^{4,γ} obstacles. The proof proceeds through the Caffarelli-Silvestre extension to the thin obstacle problem, a fine analysis of quadratic and (2m+2s)-frequency points, new explicit frequency gaps for admissible homogeneities, dimension reduction arguments, and cleaning estimates for each component of the degenerate set. The paper also establishes a full stratification of the contact set at (2m+2s)-frequency points and several new uniform frequency gaps, summarized in Theorem 1.6.","tokens_in":1529,"tokens_out":1559,"duration_ms":91492,"significance":"If the main results are correct, this is a substantial advance: it extends the generic regularity theorem of Fernández-Real–Torres-Latorre from s=1/2 to all s∈(0,1) and from zero obstacles to general nonzero obstacles, and it provides new explicit frequency gaps (Theorem 1.6) that are uniform in dimension and frequency. The paper is detailed and follows the established strategy of Figalli–Ros-Oton–Serra and Fernández-Real–Torres-Latorre, with new epiperimetric inequalities at frequencies 2m+2s and a new second blow-up analysis at quadratic points. The main theorem is falsifiable and the architecture of the proof is transparent. However, two load-bearing inputs are not fully proved in the manuscript: the sign claim in Proposition 5.10 and the spectral stability lemma (Lemma 4.7) whose proof is delegated to an unpublished preprint.","major_comments":[{"comment":"The proof of Proposition 5.10 rests on the sign claim (5.3), which is not proved in the paper. The sentence 'Claim (5.3) is proved arguing exactly as in [FS24a, Lemma 2]' is not sufficient because [FS24a] treats only s=1/2 (a=0), whereas the ODE (5.1) and the boundary flux ∂^a_θ p_α depend on a∈(−1,1). The sketched zero-counting argument does not establish that the number of zeros of p_α and ∂^a_θ p_α in [0,π/2] remains constant as α varies through (2m,2m+2s), nor that the endpoint signs change exactly at α=2m and α=2m+2s. This gap is load-bearing: the exclusion A_{n,s}∩(2m,2m+2s)=∅ enters Theorem 1.6, then Proposition 7.5(b) and Proposition 8.2 for s>1/2, and ultimately the generic dimension bound for n=3 in Theorem 1.3. Please provide a complete proof of (5.3), or a published reference that covers the present a-dependent case.","section":"§5.3, Proposition 5.10"},{"comment":"Lemma 4.7 is stated without proof; the text says only that the conclusion follows from the stability of eigenvalues and eigenfunctions under domain variation, citing [CV24, Proposition 3.1, Lemma 3.2]. Since [CV24] is a preprint (arXiv:2409.12110) and Lemma 4.7 is the key spectral input in the construction of the competitor in Proposition 4.5, which in turn underpins Theorem 1.4 and Proposition 5.1, the manuscript should either include a proof of Lemma 4.7 or reference a published, fully documented source. As it stands, a central tool of the paper is not self-contained.","section":"§4.2, Lemma 4.7"}],"minor_comments":[{"comment":"The notation a− in the condition γ∈(a−,1] is undefined. Since a=1−2s is already defined, please clarify whether the intended condition is γ>max{a,0} or γ∈(0,1] together with an additional lower bound when s<1/2.","section":"§1.1 and throughout"},{"comment":"Several typos should be corrected: Section 3 heading 'qadratic' should be 'quadratic'; Section 5 heading 'freqency' should be 'frequency'; Remark 2.7 'contronominal' should be 'contrapositive'; in the proof of Proposition 6.4, 'Haussdorf' should be 'Hausdorff'.","section":"§3, §5, §6"},{"comment":"In the last sentence of the proof, the sets '2N' and '2N+2s' should be denoted with the standard notation for the natural numbers; the current typesetting could be misread as the number 2N.","section":"§5.3, proof of Proposition 5.10"},{"comment":"In Lemma 7.2(b), the statement that for δ=0 the first eigenfunction is (x_1^2+x_{n+1}^2)^{-a/2} is unclear; please specify the normalization and explain how the exponent −a/2 arises, since the subsequent limit μ(δ)↓2s−1 is used in the proof.","section":"§7.1, Lemma 7.2"},{"comment":"The classification A_{1,s}={2m}∪{2m+2s}∪{2m+1+s} is quoted from [FS18, Proposition A.1] and is used critically in the dimension reduction arguments of Proposition 6.1. Please give a more precise reference to the location of this statement in [FS18], or include a proof in an appendix.","section":"§2.2, Proposition 2.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is technically impressive and the overall strategy is convincing, but it relies on two very recent preprints ([FS24a] and [CV24]) for load-bearing steps. I recommend requesting a complete proof of Proposition 5.10's sign claim and of Lemma 4.7, or evidence that those preprints have been accepted for publication in refereed venues, before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I read Carducci and Colombo's paper on generic regularity of free boundaries in the fractional obstacle problem. The paper is in good shape. It extends generic regularity from the thin obstacle case s=1/2 with zero obstacle (FT23) and from n=1 for all s (FR21) to all s in (0,1), nonzero obstacles, and dimensions up to 3: for a prevalence-generic obstacle, the degenerate set is empty when n<=3 and has dimension at most n-3-alpha when n>=4. That is a real step forward. The blow-up analysis at frequency 2m+2s (Theorem 1.4) with polynomial rate of convergence, the stratification of the contact set, and the new frequency gaps (Theorem 1.6) are genuinely new and go beyond the cited literature. The paper is carefully structured, follows established strategies, and I found no internal contradictions or hidden circularity.\n\nThe soft spot is exactly where the stress-test note points. Proposition 5.10, which rules out admissible frequencies in (2m,2m+2s) for every s, is load-bearing for the s>1/2 case: it supplies the gap (2,2+2s+sigma) used to clean the set Gamma_* in Prop 7.5 and hence to close Theorem 1.3. The proof of Prop 5.10 reduces to a sign claim, (5.3), about the homogeneous solutions p_alpha of the ODE (5.1). The claim is plausible, but it is not proved in the paper. The text says 'arguing exactly as in [FS24a, Lemma 2]' and gives a sketch about zeros of p_alpha and d^a_theta p_alpha being isolated and intertwined, with the number of zeros constant as alpha varies and increasing by one at alpha=2m and alpha=2m+2s. Since [FS24a] treats only s=1/2 (a=0), the extension to all a in (-1,1) requires an independent verification. If (5.3) fails for some s, the cleaning of Gamma_* does not go through and the main theorem for s>1/2 collapses. I do not think this is a death blow; the claim is likely true and the adaptation is probably routine, but it is not routine enough to skip.\n\nThe reader's weakest assumption was the classification of A_{1,s} in dimension 2, quoted from [FS18]. That is not a real concern; that classification is well-established and the dimension reduction uses it in a standard way. The paper depends on recent preprints ([FS24a], [CV24]) and same-author work ([Car24]); that is common in this area but adds to the verification burden.\n\nWho is this for? Anyone working on free boundary regularity in obstacle problems, especially the fractional/Signorini community. It deserves a serious referee: the main theorem is important and the proof is mostly detailed and honest. My recommendation is to send it to review, with the referee explicitly asked to verify the sign claim in Prop 5.10 or require a self-contained proof before acceptance.","headline":"Substantial and carefully written paper that very likely proves the stated generic regularity theorem; the one genuinely load-bearing dependency is the frequency gap (2m,2m+2s) in Prop 5.10, where the key sign claim is only sketched and cited to a preprint.","tokens_in":56385,"tokens_out":6551,"would_cite":true,"duration_ms":50583,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R35","47G20"],"pacs":[],"model":"deepseek-v4-flash","headline":"For almost every obstacle, the free boundary of the fractional obstacle problem has no degenerate points in dimensions up to 3, and the degenerate set has Hausdorff dimension at most $n-3-\\alpha$ in higher dimensions.","keywords":["fractional Laplacian","obstacle problem","free boundary","generic regularity","thin obstacle problem","frequency gaps","epiperimetric inequality","degenerate points"],"falsifier":"Find a nonzero global solution of the thin obstacle problem in $\\mathbb{R}^2$ with homogeneity inside $(2,3)$ different from $2+2s$, or more generally not among $A_{1,s}$; such a solution would refute the classification used in the dimension-reduction step and could allow degenerate points to accumulate, invalidating the $n\\le 3$ generic regularity conclusion.","tokens_in":55138,"feed_emoji":"","tokens_out":6133,"duration_ms":50448,"temperature":0.7,"pith_summary":"The paper proves a generic regularity statement for the obstacle problem for the fractional Laplacian $(-\\Delta)^s$: inside a one-parameter family of solutions driven by obstacles $\\varphi-t$, almost every member has a free boundary with no degenerate points when the spatial dimension $n$ is at most 3, and a degenerate set of Hausdorff dimension at most $n-3-\\alpha$ when $n\\ge 4$. This extends earlier generic-regularity results, which covered dimension 1 for every $s\\in(0,1)$ or the case $s=1/2$ with zero obstacle, to all $s\\in(0,1)$ and to general nonzero obstacles. The proof works through the Caffarelli--Silvestre extension, turning the nonlocal problem into a thin obstacle problem, and then combines a new frequency-gap theorem, a blow-up analysis at points with frequency $2m+2s$, dimension-reduction bounds, and cleaning results. A corollary states that for $C^\\infty$ obstacles in dimensions $n\\le 3$, the free boundary is a $C^\\infty$ manifold of dimension $n-1$ for almost every time.","feed_headline":"Almost every obstacle gives regular free boundaries up to dimension 3","feed_subtitle":"For every s in (0,1) with nonzero obstacles, the first generic-regularity proof for the fractional obstacle problem.","key_machinery":"The load-bearing mechanism is a splitting of the degenerate set by Almgren frequency together with the classification of admissible frequencies in dimension $2$, namely $A_{1,s}=\\{2m\\}_{m}\\cup\\{2m+2s\\}_{m}\\cup\\{2m+1+s\\}_{m}$, which the paper quotes and uses to control the non-admissible frequency set $\\Gamma_*$. The new frequency-gap theorem, Theorem 1.6, shows that no other frequencies occur in explicit intervals around each admissible frequency, which feeds into the cleaning results. The other central object is the Weiss energy and its epiperimetric inequalities at frequencies $2m+2s$, which yield polynomial rates of convergence to the blow-up at those points and also make the frequencies $2m+2s$ isolated. A truncated Almgren frequency function adapted to nonzero obstacles, together with the second blow-up analysis at quadratic points, separates quadratic points into ordinary and anomalous pieces and gives the dimensional bounds for $\\Gamma_2^a$.","core_discovery":"The central assertion, Theorem 1.1, is that for every $s\\in(0,1)$ and every obstacle $\\varphi\\in C^{4,\\gamma}$ with $\\gamma\\in(a_-,1]$, a monotone one-parameter family of solutions with obstacle $\\varphi-t$ has, for almost every $t$, an empty degenerate set $\\operatorname{Deg}(v(\\cdot,t))$ when $n\\le 3$, and $\\dim_H \\operatorname{Deg}(v(\\cdot,t))\\le n-3-\\alpha$ when $n\\ge 4$. The equivalent statement for the extended thin obstacle problem is Theorem 1.3, and the paper also proves new explicit frequency gaps for admissible homogeneities, summarized in Theorem 1.6. These gaps, together with a classification of two-dimensional admissible frequencies, are the engine that lets the authors isolate the degenerate set into pieces of controlled dimension and controlled cleaning rate. The paper further establishes uniqueness and polynomial convergence of blow-ups at points with frequency $2m+2s$, and a stratification of the corresponding contact set.","pith_inferences":["If the frequency-gap classification is sharp, the dimension bound $n-3-\\alpha$ in $n\\ge 4$ may be the natural barrier for the GMT-based method: new mechanisms would be needed to push generic regularity to dimension 4 and beyond.","The structure of the proof suggests that the same combination of frequency gaps and cleaning, once available, could apply to other nonlocal obstacle-type problems, such as fractional obstacle problems with drift or variable coefficients.","Because the only low-dimensional classification entering the argument is quoted from the literature, a direct proof of the two-dimensional frequency classification inside the same framework would make the generic-regularity result and the frequency-gap theorem self-contained.","The paper does not assert sharpness; one could test whether there exist obstacles for which the degenerate set in $n\\ge 4$ actually achieves dimension $n-3-\\alpha$, which would show the generic bound is optimal."],"forward_implications":["For $n\\le 3$ and every $s\\in(0,1)$, almost every obstacle has a free boundary with only regular points, so the free boundary is locally a $C^{1,\\alpha}$ manifold; with $C^\\infty$ obstacles it is a $C^\\infty$ manifold of dimension $n-1$.","For $n\\ge 4$, the degenerate set has Hausdorff dimension at most $n-3-\\alpha$, so singularities form a set of codimension at least $3+\\alpha$.","The explicit frequency gaps imply that admissible frequencies concentrate around integers as $s\\downarrow 0$ and around even integers as $s\\uparrow 1$.","Points with frequency $2m+2s$ have unique blow-ups with polynomial convergence and their contact sets are stratified into at most countably many $C^{1,\\alpha}$ manifolds.","For $s=1/2$ and zero obstacle, the main theorem recovers the previously known generic regularity for the thin obstacle problem up to dimension 3."],"supporting_citations":[{"why":"Provides the previous generic-regularity result in dimension 1 and the cleaning framework for one-parameter families that this paper extends.","marker":"[FR21]"},{"why":"Gives generic regularity for the thin obstacle problem in the case $s=1/2$ and zero obstacle, the statement that Theorem 1.3 generalizes.","marker":"[FT23]"},{"why":"Supplies the geometric measure theory lemma that converts dimension bounds and cleaning rates into generic dimension estimates.","marker":"[FRS20]"},{"why":"Quoted for the full classification of admissible frequencies in dimension 2, which is load-bearing for the dimension-reduction bounds.","marker":"[FS18]"},{"why":"Provides the Weiss-type monotonicity and epiperimetric tools for nonzero obstacles, including isolation of even frequencies, used throughout the paper.","marker":"[Car24]"},{"why":"Gives the epiperimetric inequality strategy for odd frequencies in the thin obstacle problem, which the paper adapts to frequencies $2m+2s$.","marker":"[CV24]"},{"why":"Builds the truncated Almgren frequency function for nonzero obstacles and establishes existence of blow-up limits at free boundary points.","marker":"[GR19]"},{"why":"Contains the second blow-up analysis at quadratic points and the very thin obstacle problem, used to control anomalous quadratic points.","marker":"[FJ21]"},{"why":"Introduces the Caffarelli--Silvestre extension that translates the fractional obstacle problem into the thin obstacle problem studied in the proofs.","marker":"[CS07]"}],"fun_headline_variants":["Almost every obstacle yields regular free boundaries up to 3D","First generic regularity proof for fractional obstacle problem","Fractional Laplacian: free boundaries regular for generic obstacles","For all s in (0,1), generic obstacles give regular boundaries in 3D","Almost all obstacles: regular free boundaries for fractional Laplacian"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes that the only possible growth rates of two-dimensional blow-up limits are the known list $2m$, $2m+2s$, and $2m+1+s$; if an additional rate existed, the dimension bounds for the bad sets and the final generic dimension estimate could fail.","fun_headline_variants_meta":{"raw":{"variants":["Almost every obstacle yields regular free boundaries up to 3D","First generic regularity proof for fractional obstacle problem","Fractional Laplacian: free boundaries regular for generic obstacles","For all s in (0,1), generic obstacles give regular boundaries in 3D","Almost all obstacles: regular free boundaries for fractional Laplacian"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001197,"raw_usage":{"total_tokens":4893,"prompt_tokens":863,"completion_tokens":4030,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":479,"completion_tokens_details":{"reasoning_tokens":3942}},"tokens_in":479,"tokens_out":4030,"duration_ms":30906,"temperature":1.0,"reasoning_tokens":3942,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T10:49:56.758699+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a nonzero global solution of the thin obstacle problem in $\\mathbb{R}^2$ with homogeneity inside $(2,3)$ different from $2+2s$, or more generally not among $A_{1,s}$; such a solution would refute the classification used in the dimension-reduction step and could allow degenerate points to accumulate, invalidating the $n\\le 3$ generic regularity conclusion.","supporting_citations":[],"review_version":1}