{"id":"b4f5a2a4-b7b6-460c-a22c-f0d3689bd86a","arxiv_id":"2501.17378","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For diagonal self-affine measures with distinct Lyapunov exponents and exponentially separated one-dimensional projections, Hausdorff dimension equals the minimum of d and the Lyapunov dimension.","lead":"This paper proves that the Hausdorff dimension of any diagonal self-affine measure equals its Lyapunov dimension whenever the Lyapunov exponents are distinct and the coordinate systems are exponentially separated. It settles a conjecture of Rapaport, removing an extra homogeneity condition on the linear parts.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reduction's polynomial-growth bound on |Γ| is elementary and valid; the real gap is the omitted proofs of entropy-increase lemmas 5.3, 5.4, 7.2 in the random-measure setting.","rationale":"The reader's weakest_assumption was the unproved polynomial growth of |Γ| in the reduction of Theorem 1.3. This is not a true obstacle: because the linear parts are diagonal and hence commute, the product Aφ_{x|N} is determined by the composition vector of x|N, giving at most binomial(N+|Λ|-1, |Λ|-1) distinct values, which is polynomial in N. Thus the reduction's estimate is valid and the approximation h_RW(Φ,A)→H(p) is justified. However, the paper does omit proofs of several lemmas that are central to the entropy increase machinery, and the setting here genuinely differs from Rapaport's by dropping the 1-dimensional subgroup assumption. Those omitted proofs are a legitimate completeness concern, and the verdict CONDITIONAL remains appropriate, but for a different reason than the one highlighted by the reader. The concrete check of supplying and auditing the omitted proofs would settle whether the central claim is fully established.","tokens_in":44839,"tokens_out":15081,"duration_ms":144419,"concrete_test":"Write out full proofs of Lemmas 5.3, 5.4, and 7.2 in the present random-measure notation. For Lemma 5.4, compare step-by-step with [46, Lemma 3.5] and confirm that no inequality uses the existence of cut-sets U_n satisfying (1.19), which required the linear parts to lie in a 1-dimensional subgroup. If any such use appears, Theorem 7.1 and hence Theorem 1.3 would not follow for the general diagonal case.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader's flagged assumption is not a genuine gap: since the diagonal matrices {A_i} commute, Aφ_{x|N} depends only on the count vector of symbols in x|N. There are at most binomial(N+|Λ|-1, |Λ|-1) such count vectors, so |Γ| = O(N^{|Λ|-1}), polynomial in N. Hence the reduction's bound n log|Γ| ≤ 2n|Λ| log N holds for large N (with a loose constant), and h_RW(Φ,A) → H(p) as N → ∞ is sound. The actual load-bearing risk is the incomplete proof of the entropy increase theorem: Theorem 7.1 is the engine behind Theorem 8.2 and Theorem 1.12, and its proof invokes Lemmas 5.3, 5.4, and 7.2, all stated without proof and justified only as 'almost identical' to lemmas in [46] that were proved under the 1-dimensional subgroup assumption. Since the paper's central claim is precisely the removal of that assumption, these adaptations must be checked not to secretly rely on the cut-sets U_n satisfying (1.19). If any omitted lemma does rely on (1.19), the entropy increase result and consequently Theorem 1.3 would lack support for general diagonal IFS.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves Rapaport's conjecture that a diagonal self-affine measure on R^d has dimension equal to min{d, dim_L(Φ,p)} whenever the Lyapunov exponents are distinct and each one-dimensional coordinate IFS is exponentially separated. The proof strategy disintegrates the Bernoulli measure β according to the linear parts of the IFS, obtaining random measures μ_ω with a convolution structure, establishes exact dimensionality and a Ledrappier-Young type formula for these disintegrations (Theorem 3.2), and then proves a central entropy increase result (Theorem 7.1) for the random measures via non-conformal partitions. This entropy increase is the engine behind Theorem 8.2 and the main Theorem 1.12, from which Theorem 1.3 is derived by a limiting argument. The paper also derives several applications, including a typical dimension result and a characterization of full-dimension measures on certain diagonal self-affine carpets.","tokens_in":45141,"tokens_out":7180,"duration_ms":62997,"significance":"If the proof is fully valid, the result is a substantial advance: it removes the 1-dimensional subgroup assumption that was essential in Rapaport's original argument, thereby confirming a conjecture and extending the dimension theory of diagonal self-affine measures to a natural general setting. The introduction of random measures with convolution structure and the adaptation of Hochman–Rapaport entropy techniques to this setting are likely to be influential. The paper also contains useful applications, such as Corollary 1.6 for algebraic parameters and Corollary 1.9 on measures of full dimension. However, the significance is presently qualified by the fact that several load-bearing lemmas in the entropy increase argument are stated without proof; the central claim is credible but not yet fully verified as written.","major_comments":[{"comment":"The proof of Theorem 7.1, which is the core entropy increase result, depends on Lemmas 5.3, 5.4, and 7.2, whose proofs are omitted with justifications such as 'the proof is almost identical to [26, Lemma 4.4]' or 'is therefore omitted.' Since the paper's main contribution is precisely the removal of the 1-dimensional subgroup assumption used in [46], these adaptations must be shown in detail. In particular, the author should verify explicitly that the proofs of Lemmas 5.3 and 5.4 do not secretly rely on the existence of cut-sets satisfying (1.19) or on the subgroup condition. If any of these lemmas do require that condition, then Theorem 7.1 and consequently Theorem 1.3 would lack support in the full diagonal case. I request full proofs or a detailed appendix spelling out the modifications needed in the random-measure setting.","section":"Section 5 and Section 7, Lemma 5.3, Lemma 5.4, Lemma 7.2"},{"comment":"Lemma 4.2 is used in the proof of Theorem 7.1 at equation (7.2), where it provides the asymptotic relation between n^{-1}H(θ∗k_j∗μ_ω, E^ω_n) and an average of conditional entropies over scales. The lemma is an adaptation of [26, Lemma 3.4] to the non-conformal partitions E^ω_n, but no proof is supplied. Since the partition family depends on ω and the convolution structure of the random measures is essential, the adaptation is not purely mechanical; the author should provide a complete proof or at least a detailed indication of how the original proof transfers to this setting.","section":"Section 4, Lemma 4.2"},{"comment":"The estimate H(β, ∨_{i=0}^{n-1} T^{-i}Γ) ≤ n log|Γ| ≤ 2n|Λ| log N is used to show that h_RW(Φ,A) → H(p) as N→∞. The second inequality relies on the fact that |Γ| = O(N^{|Λ|-1}), which holds because the diagonal matrices {A_i} commute and Aφ_{x|N} depends only on the count vector of the symbols in x|N. This elementary bound is not stated or proved, and it is load-bearing for the reduction. I recommend adding a short explicit derivation of the polynomial bound on |Γ|.","section":"Section 1.3, Reduction of Theorem 1.3"}],"minor_comments":[{"comment":"The phrase 'the j-th the Lyapunov exponent' contains an extra 'the'; this typo appears in the abstract and in the introduction.","section":"Abstract and Section 1.1"},{"comment":"The statement reads 'Φj is Diophantine and for 1 ≤ j ≤ d'; the word 'for' appears twice and the intended statement is 'Φj is Diophantine for 1 ≤ j ≤ d'.","section":"Theorem 1.12"},{"comment":"In the proof of Lemma 6.5, the displayed integral uses π_{[d-1]}μ_{T^nω}, but the lemma is stated for an arbitrary J ⊂ [d]. The integrand should be π_Jμ_{T^nω}; this appears to be a typographical error, since the surrounding text uses π_J.","section":"Lemma 6.5, proof"},{"comment":"In reference [5], the author name is corrupted as 'Micha/suppress l Rams'; this should be corrected.","section":"References"},{"comment":"The relation 'Aϕ_u ≈ Aϕ_v' is described as 'entrywise comparable', but the exact meaning of ≈ (with implicit constants) is not made precise; please define it.","section":"Section 1.3, equation (1.19)"}],"recommendation":"major_revision","confidential_remarks":"This is a strong paper on an important problem, and the main theorem is likely correct. However, the current version leaves too many central technical lemmas unproved, including the key adaptations to the random-measure setting. The omitted proofs are not merely cosmetic: the entire entropy increase argument rests on them. I would encourage the editor to invite a revision in which the author supplies full proofs of Lemmas 4.2, 5.3, 5.4, and 7.2, and also explicitly states the polynomial growth of |Γ| in the reduction. If those proofs are supplied and no hidden dependence on the subgroup condition appears, the paper would be suitable for a top-tier journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main theorem is genuinely new and significant: it removes the 1-dimensional subgroup assumption from Rapaport's dimension formula for diagonal self-affine measures, confirming his conjecture under distinct Lyapunov exponents and coordinate-wise exponential separation. The proof strategy is sensible — disintegrate into random convolution measures, prove exact dimensionality and entropy increase for those, then pass back to the original measure. The paper is long and mostly careful, and the applications (full-dimension measures, projections, typical parameter results) are reasonable and not padded.\n\nOne weakness flagged by the reader is a false alarm. The bound H(β, ∨T^{-i}Γ) ≤ n log|Γ| ≤ 2n|Λ| log N is fine: since the diagonal matrices commute, Γ depends only on the count vector of symbols in the first N digits, so |Γ| = O(N^{|Λ|-1}). The reduction of Theorem 1.3 from Theorem 1.12 does not have a combinatorial gap.\n\nThe real soft spot is the unexplained reliance on omitted proofs. Lemmas 4.2, 5.3, 5.4, and 7.2 are all stated with proofs 'almost identical' to lemmas in [26] or [46], but [46] worked under the 1-dimensional subgroup assumption, and the whole point here is to remove that assumption. The reader is entitled to see why the adapted arguments do not secretly use the cut-sets satisfying (1.19). In particular, Lemma 5.3 and Lemma 5.4 feed directly into Proposition 5.1, which is the engine behind Theorem 7.1 and hence Theorem 1.12. If any of those adaptations fails in the general diagonal setting, the main theorem loses its support. This is a genuine burden on the referee, not a stylistic quibble. The rest of the proof, including the exact dimensionality and Ledrappier-Young formula for the random conditional measures, appears coherent.\n\nThe citation pattern is honest: the author cites Rapaport, Hochman, Feng, and Jordan-Pollicott-Simon as the actual foundations, and the self-citation to [18] appears only in a footnote and a conjectural remark. No circularity problem.\n\nThis paper deserves a serious referee. My recommendation: send it to peer review, and require the author to either include the omitted proofs or provide a detailed appendix explaining exactly how each adapted lemma avoids the 1-dimensional subgroup condition. With that filled in, I would expect the result to be accepted.","headline":"Strong resolution of Rapaport's conjecture with a real but localized gap: several load-bearing lemmas are deferred to 'almost identical' arguments from the 1-dimensional-subgroup setting.","tokens_in":45636,"tokens_out":1295,"would_cite":true,"duration_ms":15441,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28A80","37C45"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that a diagonal self-affine measure has dimension equal to the smaller of $d$ and its Lyapunov dimension whenever its Lyapunov exponents are distinct and each coordinate IFS is exponentially separated.","keywords":["self-affine measures","Hausdorff dimension","Lyapunov dimension","exponential separation","diagonal iterated function systems","entropy increase","measure disintegration","exact dimensionality"],"falsifier":"Take the explicit two-map example from Example 1.7 with distinct prime ratios $a=q_1/q_2$, $b=q_2/q_3$ and weight $p_1\\in(0,1/2)$; the theorem predicts $\\dim\\mu=\\min\\{2,\\dim_L(\\Phi,p)\\}$, so a rigorous numerical or analytic computation giving a strictly smaller local dimension on a positive-measure set would falsify Theorem 1.3.","tokens_in":44604,"feed_emoji":"📐","tokens_out":15359,"duration_ms":139631,"temperature":0.7,"pith_summary":"The paper proves a dimension formula for a wide class of diagonal self-affine measures: if the Lyapunov exponents are pairwise distinct and each one-dimensional coordinate system is exponentially separated, then the Hausdorff dimension of the measure equals the smaller of the ambient dimension $d$ and the Lyapunov dimension built from the contraction rates and the entropy of the weight vector. This settles a conjecture from [46], where the same conclusion was known only under an extra assumption that all linear parts lie in a one-parameter subgroup. A sympathetic reader should care because the dimension of self-affine measures with overlaps is generally open, and this gives an explicit, checkable answer for an entire class of overlapping diagonal systems. The proof works by disintegrating the measure into random measures with a convolution structure, proving they are exact dimensional, and establishing an entropy-increase estimate for repeated self-convolutions.","feed_headline":"Diagonal self-affine measures reach their Lyapunov dimension.","feed_subtitle":"Even with overlaps, distinct exponents and exponential separation pin down the dimension.","key_machinery":"The central mechanism is a disintegration of the Bernoulli coding measure $\\beta$ according to diagonal linear parts at scale $N$: two sequences share a $\\Gamma$-atom when $A_{\\phi_{x|N}}=A_{\\phi_{y|N}}$, and the tail $\\sigma$-algebra is $\\mathcal{A}=\\bigvee_{n\\ge0}T^{-n}\\Gamma$. This produces random measures $\\mu_\\omega=\\Pi\\beta_\\omega$ on $\\mathbb{R}^d$ satisfying $\\mu_\\omega=\\nu_\\omega^n*A_{\\omega|n}\\mu_{T^n\\omega}$, a dynamical self-affinity that gives each $\\mu_\\omega$ a convolution structure at every scale. The paper proves an entropy-dimension formula asserting that the projections $\\pi_J\\mu_\\omega$ are exact dimensional, and an entropy-increase theorem: convolving any measure with positive $E_\\omega^n$-entropy against $\\mu_\\omega$ raises entropy by at least $N\\kappa_{\\mathcal{A}}+\\delta$ along the nonconformal partitions $E_\\omega^n=A_{\\omega|n}D_0$ (images of the unit dyadic grid under the current linear part). Together these imply the intermediate formula $\\dim\\mathcal{A}=\\min\\{d,f_\\Phi(h_{RW}(\\Phi,\\mathcal{A}))\\}$, and letting $N\\to\\infty$ yields the main theorem.","core_discovery":"On the paper's own terms, the central claim is Theorem 1.3. Let $\\Phi$ be a finite diagonal affine IFS on $\\mathbb{R}^d$, with maps $\\phi_i(x)=\\mathrm{diag}(r_{i,1},\\dots,r_{i,d})x+t_i$, let $p$ be a probability vector, and let $\\chi_j=\\sum_i -p_i\\log|r_{i,j}|$ be the $j$-th Lyapunov exponent. If $\\chi_1<\\dots<\\chi_d$ and each one-dimensional coordinate IFS $\\Phi_j$ is exponentially separated, then $\\dim\\mu=\\min\\{d,\\dim_L(\\Phi,p)\\}$, where $\\dim_L$ is the Lyapunov dimension defined by the piecewise-linear function $f_\\Phi$ in (1.4)--(1.5). Exponential separation means the minimal gap between distinct length-$n$ compositions of each coordinate system decays no faster than $c^n$ for some $c>0$. Earlier work in [46] proved the formula only when the diagonal linear parts lie in a one-dimensional multiplicative subgroup; the paper removes that restriction by replacing cut-set constructions with a disintegration of the Bernoulli coding measure into random measures $\\mu_\\omega$ that are dynamically self-affine, exact dimensional, and satisfy an entropy-dimension formula and an entropy-increase theorem. The main theorem is obtained by letting the block length tend to infinity.","pith_inferences":["The success of the random-measure disintegration suggests that the one-parameter subgroup restriction in [46] was not a genuine barrier; block-diagonal or simultaneously triangulable affine systems may be approachable by the same route, though the entropy-increase estimates would need to be rebuilt.","The paper notes but does not pursue the possibility that its typical dimension result, combined with known Fourier decay estimates, could imply typical absolute continuity of diagonal self-affine measures; that implication remains unproved.","Because Remark 1.4 exhibits an example with equal Lyapunov exponents where the dimension formula fails, any extension to non-distinct exponents must introduce a saturation mechanism rather than merely weakening the separation assumptions."],"forward_implications":["For every diagonal self-affine measure satisfying the two hypotheses, the Hausdorff dimension is now explicitly computable from the contraction rates and weights, even when the pieces overlap.","If the coordinate systems are given by algebraic parameters and have no exact overlaps, Corollary 1.6 yields the same dimension formula without further case checks.","For Lebesgue-typical parameter choices, Corollary 1.8 shows the dimension equals $\\min\\{d,\\dim_L(\\Phi,p)\\}$ outside a small exceptional set, so the formula is not confined to specially constructed examples.","In the planar case, Corollary 1.9 characterizes ergodic measures of full dimension on certain overlapping diagonal self-affine sets as exactly the Bernoulli measures with singular-value weights, and Corollary 1.10 gives the dimensions of their orthogonal projections in non-axis directions."],"supporting_citations":[{"why":"Supplies the theorem being generalized, the nonconformal-partition technique, and the conjecture this paper confirms.","marker":"[46]"},{"why":"Gives the exact-dimensionality theorem for self-affine measures whose disintegration version underlies Theorem 1.11.","marker":"[19]"},{"why":"Provides the entropy and transverse-dimension argument used to prove exact dimensionality of the random measures.","marker":"[17]"},{"why":"Introduces exponential separation and the inverse entropy theorems that motivate the entropy-increase step.","marker":"[26]"},{"why":"Contributes the reduction idea that lets the entropy-increase result be applied in the general case.","marker":"[5]"},{"why":"Introduced random-measure disintegration by linear parts, the construction adapted to define the measures $\\mu_\\omega$.","marker":"[24]"}],"fun_headline_variants":["Confirmed: diagonal self-affine measures reach Lyapunov dimension","Self-affine measures: dimension formula without subgroup restriction","Exponential separation fully determines diagonal self-affine dimension","Rapaport's conjecture proved for diagonal self-affine measures","Lyapunov dimension attained: overlaps don't prevent it"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The main theorem rests on the stated assumptions of distinct Lyapunov exponents and exponential separation of each coordinate IFS, and additionally on the combinatorial bound that only polynomially many distinct diagonal linear parts occur among words of length $N$, i.e. $|\\Gamma|\\le O(N^{2|\\Lambda|})$; this unstated bound is what lets the block length tend to infinity and recover the Lyapunov dimension.","fun_headline_variants_meta":{"raw":{"variants":["Confirmed: diagonal self-affine measures reach Lyapunov dimension","Self-affine measures: dimension formula without subgroup restriction","Exponential separation fully determines diagonal self-affine dimension","Rapaport's conjecture proved for diagonal self-affine measures","Lyapunov dimension attained: overlaps don't prevent it"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000569,"raw_usage":{"total_tokens":2806,"prompt_tokens":1173,"completion_tokens":1633,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":789,"completion_tokens_details":{"reasoning_tokens":1552}},"tokens_in":789,"tokens_out":1633,"duration_ms":11748,"temperature":1.0,"reasoning_tokens":1552,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T04:43:56.889935+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the explicit two-map example from Example 1.7 with distinct prime ratios $a=q_1/q_2$, $b=q_2/q_3$ and weight $p_1\\in(0,1/2)$; the theorem predicts $\\dim\\mu=\\min\\{2,\\dim_L(\\Phi,p)\\}$, so a rigorous numerical or analytic computation giving a strictly smaller local dimension on a positive-measure set would falsify Theorem 1.3.","supporting_citations":[{"cited_title":"Dimension theory of iterated fu nction systems","cited_arxiv_id":null,"evidence_quote":"Gives the exact-dimensionality theorem for self-affine measures whose disintegration version underlies Theorem 1.11."},{"cited_title":"Dimension of invariant measures for affine i terated function systems","cited_arxiv_id":null,"evidence_quote":"Provides the entropy and transverse-dimension argument used to prove exact dimensionality of the random measures."},{"cited_title":"On self-similar sets with overlaps an d inverse theorems for entropy","cited_arxiv_id":null,"evidence_quote":"Introduces exponential separation and the inverse entropy theorems that motivate the entropy-increase step."},{"cited_title":"On the dimension of self-affine sets and measures with overlaps","cited_arxiv_id":null,"evidence_quote":"Contributes the reduction idea that lets the entropy-increase result be applied in the general case."},{"cited_title":"Lq dimensions and projections of random measures","cited_arxiv_id":null,"evidence_quote":"Introduced random-measure disintegration by linear parts, the construction adapted to define the measures $\\mu_\\omega$."}],"review_version":1}