{"id":"8590c912-b67e-48b5-ac67-81abfcaaf149","arxiv_id":"2608.02459","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Dirichlet Liouville Brownian motion has a.s. simple spectrum and absolutely continuous joint eigenvalue laws for all 0<γ<2, via a new Cameron–Martin response calculus.","lead":"An explicit open question about the spectrum of Liouville Brownian motion is resolved: on every bounded connected planar domain, for all subcritical disorder strengths, the random drum has almost surely simple spectrum and absolutely continuous joint eigenvalue laws. A new response calculus based on Cameron–Martin shifts of the Gaussian free field supplies the proof and also yields resolvent Ward identities and density results for averaged resolvent observables.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the central proof chain is coherent and the only delicate external input (exact-dimensionality of GMC) appears correctly cited and locally reduced.","rationale":"The reader's weakest-assumption identified the singularity and full quasi-support of M_h as the load-bearing premise. I agree that this is the most delicate step and that it depends on established GMC results, but I do not find it inadequately supported: the manuscript provides a local reduction to exact-dimensionality on smooth interior subdomains (Lemma 6.12), a full-quasi-support proof via the massive-GFF PCAF criterion (Lemma 3.3), and careful Borel/coherence bookkeeping for the slicing argument. No internal inconsistency or unsupported claim in the main theorem's proof was detected. The Ward-identity and density results in Part III are independent and appear correct. The strongest central claim, Theorem 1.1, is therefore accepted as coherent; the appropriate verdict remains ACCEPT with moderate confidence, so no change is needed. I mark agreement as 'partial' because I share the reader's identification of the singularity premise as the least secure point, though I do not believe it constitutes a flaw.","tokens_in":66625,"tokens_out":44501,"duration_ms":364933,"concrete_test":"Verify that the covariance G_U(x,y)=log^+(T/|x-y|)+g(x,y), with g continuous on W×W for W⋐U, satisfies the hypotheses of Rhodes–Vargas [20, Theorems 4.1–4.2] for exact dimensionality of the GMC measure, and confirm via Shamov's approximation independence [21, Corollary 18] that the heat-kernel regularization used in Lemma 3.1 produces the same measure as the regularization in [20]. If both checks pass, Lemma 6.12 is fully sound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I read the proof of Theorem 1.1 as resting on three pillars: (i) the Liouville measure M_h is almost surely finite, smooth, of full quasi-support, and singular with respect to Lebesgue; (ii) square transversality (Theorem 6.14) isolating the M_h-singular and Lebesgue-absolutely-continuous parts of the distributional equation; (iii) finite-dimensional Gaussian slicing converting pathwise splitting and submersion statements into almost-sure simplicity and joint absolute continuity. Each is supported by the manuscript: Lemma 3.3 reduces full quasi-support to the massive-GFF PCAF criterion of [9,14]; Lemma 6.12 reduces singularity to exact-dimensionality of GMC via the local covariance form (6.26) and the comparison (3.8); Theorem 6.14 legitimately uses the Lebesgue decomposition because M_h ⊥ dx almost surely; and the slicing arguments in Theorems 7.6–7.8 use standard zero-set and inverse-function theorems with careful Borel encodings (Lemmas 5.4, 7.4, 7.5). I did not find an internal inconsistency or a missing step in the transference from coherent families to canonical measures. The most delicate external input is the exact-dimensionality theorem [20, Thms 4.1–4.2]; its applicability to a zero-boundary covariance of the form (6.26) with a continuous additive term is plausible and standard, but this is where a referee should focus. This is a citation-check concern, not an identified gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a Cameron–Martin response calculus for Dirichlet Liouville Brownian motion on arbitrary bounded connected planar domains, with the Gaussian free field realized on a negative Sobolev space. Part I constructs Borel coherent Gaussian-multiplicative-chaos families, trace forms, Green potentials, and measurable spectral data, and places finite-dimensional perturbations on a fixed Hilbert space. Part II proves the main spectral theorem (Theorem 1.1): for 0<γ<2, the Dirichlet LBM generator has almost surely compact resolvent, strictly increasing eigenvalues, and absolutely continuous laws for every finite vector of ordered eigenvalues. The proof chain runs through cluster response compressions (Prop. 6.3), square transversality (Thm. 6.14), the existence of simple-splitting directions (Thm. 6.18), and finite-dimensional Gaussian slicing with zero-set/submersion arguments (Thms. 7.6 and 7.8). Part III establishes fixed-energy resolvent differentiability (Thm. 1.2), tested moving-measure and fixed-base resolvent Ward identities with explicit causal Laplace representatives (Thm. 1.3), and, for 0<γ<√2, absolute continuity results for averaged resolvent observables (Cor. 1.4). The paper is carefully explicit about what is not asserted: no time-domain heat-semigroup difference-quotient convergence, no simultaneous canonical agreement along uncountable parameter lines, and no high-energy quantum-unique-ergodicity control.","tokens_in":66965,"tokens_out":31218,"duration_ms":285807,"significance":"If correct, this resolves the open simple-spectrum problem for Dirichlet Liouville Brownian motion [2, Problem 4.2] and strengthens it to joint absolute continuity of finite ordered eigenvalue vectors, throughout the full subcritical range and without boundary regularity assumptions. The paper is strong methodologically: the pathwise spectral part is reduced to explicit local lemmas, the Borel measurability of spectral data is handled systematically, and the Gaussian slicing arguments are careful about full-measure fibers and countable localization. Part III is also transparent about its limitations, especially the deliberately weak sense in which causal Laplace representatives realize the Ward responses. The most delicate external input is the exact-dimensionality/singularity theorem for GMC used in Lemma 6.12; on inspection it appears correctly cited and locally reduced, and no circularity is present. The central claims are supported by a coherent and unusually detailed proof chain.","major_comments":[],"minor_comments":[{"comment":"The application of [20, Thms. 4.1–4.2] to the zero-boundary covariance in (6.26) is terse: the covariance is exact-log plus a bounded continuous additive term, and transfer to the killed-heat canonical measure uses [21, Cor. 18]. Since this singularity input is load-bearing for Theorem 6.14, a sentence explaining why the continuous perturbation g_j does not change the exact dimension, and why approximation independence applies to the specific heat-semigroup version, would make the step easier to verify. I do not see a substantive gap.","section":"Lemma 6.12"},{"comment":"The displayed inequality Cap_{1,D}(N_j) ≤ Cap_{1,U_j}(N_j) uses the compact-localized capacity comparison, but N_j need not be compact. The intended conclusion follows by applying the comparison to N_j ∩ V_j, which is compactly contained in U_j, and then using η_j=1 on V_j. Please adjust the line accordingly.","section":"Lemma 3.3 proof, after (3.25)"},{"comment":"The symbol \\tilde e appears in \\int_D \\tilde e f\\,dσ^h without definition. It appears to be a typographical artifact; the pairing should presumably be written \\int_D f\\,dσ^h.","section":"Appendix A, Corollary A.1, (A.7) and (A.9)"},{"comment":"The symbol W is overloaded: it denotes the isonormal Gaussian coordinate field in (5.10)–(5.11) and the Ward distributions W^h(a,b;f) in (1.15). This is confusing, especially near the appendix, and one of the two should be renamed.","section":"Sections 2 and 9"},{"comment":"Several key references are very recent arXiv preprints ([2], [3], [10]). If any have appeared in final peer-reviewed form by the publication date, the citations should be updated.","section":"References"}],"recommendation":"accept","confidential_remarks":"This is a strong and unusually complete manuscript. The central proof chain is coherent, the target conclusions are not assumed, and the delicate external input (exact dimensionality of GMC) appears correctly cited and locally reduced. The minor comments are expository and do not affect the correctness of the main results."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it. The paper proves simple spectrum for Dirichlet Liouville Brownian motion on arbitrary bounded connected planar domains, which is explicitly open in [2, Problem 4.2], and goes further to joint absolute continuity of ordered eigenvalue vectors. That joint-density result I also don't recall seeing. The proof is long but the architecture is clean: coherent GMC families under Cameron–Martin shifts, fixed-space analytic perturbation, then Gaussian slicing. The Borel encodings and full-probability events are handled with care; that's usually where these arguments get sloppy, and here it's solid.\n\nI don't see a load-bearing flaw. The square transversality argument separating the singular and absolutely continuous parts of the distributional equation is legitimate because M_h is almost surely singular. The most delicate external input is exact-dimensionality of GMC for the zero-boundary covariance in (6.26). The paper cites Rhodes–Vargas for that; the application is plausible and standard, but I'd want a referee to verify the local comparison and the continuous additive term. That's a citation-check concern, not an identified gap.\n\nThe Ward identities and causal Laplace representatives are a solid addition, and the paper is honest about what it doesn't assert—no time-domain heat-semigroup difference-quotient convergence, and the narrower range for averaged-resolvent densities is explained.\n\nMain soft spot is simply the length. A full referee check is a real time commitment, but I didn't find an internal inconsistency or a circularity. The response calculus will likely be reusable for related spectral questions. Specialists in LQG and Dirichlet forms should read it. I'd send it to peer review.","headline":"Long, careful proof of a real open problem; I think it's right, but the exact-dimensionality input deserves referee scrutiny.","tokens_in":67435,"tokens_out":2764,"would_cite":true,"duration_ms":26917,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J35","47A10","60G57","31C25"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that the Dirichlet Liouville Brownian motion generator has almost surely simple spectrum and absolutely continuous finite joint eigenvalue laws for every subcritical γ in (0,2), via a Cameron–Martin response calculus.","keywords":["Liouville Brownian motion","Gaussian multiplicative chaos","simple spectrum","eigenvalue densities","Cameron–Martin shift","resolvent Ward identities","Gaussian free field","absolute continuity"],"falsifier":"Find a bounded connected planar domain and a γ∈(0,2) for which the eigenvalue collision event {Λ^h_n = Λ^h_{n+1}} has positive probability, or for which some ordered eigenvalue vector (Λ^h_{n_1},...,Λ^h_{n_m}) has an atom. Concretely, compute the cluster discriminant on a finite-dimensional Gaussian slice: if that discriminant vanishes identically on a positive-measure slice, simplicity fails.","tokens_in":66486,"feed_emoji":"🔢","tokens_out":3865,"duration_ms":47161,"temperature":0.7,"pith_summary":"The paper establishes, for the Dirichlet Liouville Brownian motion generator on any bounded connected planar domain with no boundary regularity assumption, and for every subcritical γ in (0,2), that the spectrum is almost surely simple and that every finite vector of ordered eigenvalues is absolutely continuous with respect to Lebesgue measure. This resolves an open problem in the spectral geometry of Liouville quantum gravity. The proof works by developing a response calculus: Cameron–Martin shifts of the Gaussian free field produce coherent Gaussian multiplicative chaos measures, and the first variation of an eigenvalue cluster becomes a finite-dimensional matrix compression. Transversality of the resulting signed measures, combined with Gaussian slicing and analytic zero-set arguments, yields both non-degeneracy and densities.","feed_headline":"Almost surely simple spectrum for Liouville Brownian motion","feed_subtitle":"A response calculus shows every finite vector of ordered eigenvalues is absolutely continuous for every subcritical γ.","key_machinery":"The paper introduces a Cameron–Martin response calculus for the Liouville speed measure. Coherent Gaussian multiplicative chaos measures are built so that shifting the Gaussian free field by τf multiplies the measure by e^{γτf}. Transporting the varying L² spaces to one fixed Hilbert space makes the first variation of an isolated eigenvalue cluster a finite-dimensional compression on its eigenspace, with matrix entries −γΛ∫fφ_iφ_j dM_h. The key transversality step separates the distributional equation for (φ)² into its M_h-singular and Lebesgue-absolutely-continuous parts, proving that the response measures of distinct simple eigenvalues are linearly independent. Finite-dimensional Gaussian","core_discovery":"The central discovery is Theorem 1.1: for every bounded connected planar domain, without boundary regularity assumptions, and for every 0<γ<2, the Dirichlet Liouville Brownian motion generator has almost surely compact resolvent and its eigenvalues, ordered with multiplicity, satisfy 0<Λ_1<Λ_2<...↑∞. Moreover, every finite vector of ordered eigenvalues has an absolutely continuous law with respect to Lebesgue measure. This resolves the previously open simple-spectrum problem for Dirichlet LBM and adds the stronger joint absolute-continuity conclusion. The paper also proves a fixed-energy resolvent derivative formula, tested resolvent Ward identities, and, in the range 0<γ<√2, absolute contin","pith_inferences":["The same response calculus likely transfers to other singular Gaussian fields whose GMC measures are coherent under Cameron–Martin shifts, so simple spectrum and joint densities may hold beyond Dirichlet LBM; this is an extension the paper does not pursue.","Because the first-order eigenvalue response is an explicit finite matrix, the calculus can probably be iterated to obtain higher-order joint density formulas or correlation functions for eigenvalue clusters; the paper stops at first order for spectral non-degeneracy.","The absolute-continuity results for averaged resolvent observables are stated outside deterministic null parameter sets; it is plausible that a refined submersion argument removes these exceptional sets, since the response Gram matrix is positive definite at infinity in each parameter.","The pathwise identity relating the log-derivative of a normalized eigenvalue to the difference of the area measure and the eigenfunction-square measure gives a concrete route toward quantum unique ergodicity, provided the high-energy control that the paper explicitly leaves open can be supplied."],"forward_implications":["Almost surely, no eigenvalue of the Dirichlet LBM generator is repeated, so the ordered spectral labels Λ_n are well-defined individual random variables.","Every finite vector of ordered eigenvalues has a density with respect to Lebesgue measure, so finite-dimensional spectral laws have no atoms.","These conclusions hold on every bounded connected planar domain, including domains with irregular boundaries, without any boundary regularity condition.","The same theorems extend to the singular-scale LQG normalization, answering the previously open simplicity question in that setting.","The resolvent Ward identities give explicit formulas for the derivatives of moving-measure and fixed-base resolvent pairings for every λ>0 in the full subcritical range."],"fun_headline_variants":["Liouville Brownian motion has almost surely simple spectrum","Simple spectrum proven for all Liouville Brownian motion","Joint eigenvalue densities exist for Liouville Brownian motion","Response calculus yields simple spectrum and joint densities","Eigenvalues of LBM: simple and absolutely continuous"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The argument hinges on the Liouville measure M_h being almost surely singular with respect to Lebesgue measure, with full support and positive mass on every open set; if M_h ever had an absolutely continuous component, the separation of the distributional square-eigenfunction equation into singular and absolutely continuous parts would fail, and the linear independence of weighted eigenfunction-square measures could collapse.","fun_headline_variants_meta":{"raw":{"variants":["Liouville Brownian motion has almost surely simple spectrum","Simple spectrum proven for all Liouville Brownian motion","Joint eigenvalue densities exist for Liouville Brownian motion","Response calculus yields simple spectrum and joint densities","Eigenvalues of LBM: simple and absolutely continuous"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000299,"raw_usage":{"total_tokens":1608,"prompt_tokens":832,"completion_tokens":776,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":576,"completion_tokens_details":{"reasoning_tokens":703}},"tokens_in":576,"tokens_out":776,"duration_ms":7442,"temperature":1.0,"reasoning_tokens":703,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T06:53:48.903714+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a bounded connected planar domain and a γ∈(0,2) for which the eigenvalue collision event {Λ^h_n = Λ^h_{n+1}} has positive probability, or for which some ordered eigenvalue vector (Λ^h_{n_1},...,Λ^h_{n_m}) has an atom. Concretely, compute the cluster discriminant on a finite-dimensional Gaussian slice: if that discriminant vanishes identically on a positive-measure slice, simplicity fails.","supporting_citations":[],"review_version":1}