{"id":"531b42ac-33ad-400e-8e2f-386ed56df05d","arxiv_id":"2502.04001","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors extend Falconer's dimension theorem to linear projections of self-affine sets, prove the exceptional projections form algebraic varieties, and construct new examples of non-exact-dimensional projected measures and small sumsets.","lead":"This paper proves that the dimensions of linear projections of typical self-affine fractals are almost always equal to a new computable quantity, the projected affinity dimension. It also constructs the first examples of self-affine sets whose projected equilibrium measures are not exact-dimensional, and of strongly irreducible self-affine sets with surprisingly small sumsets.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem A(b) rests on Lemma 5.2, an imported integral estimate for non-invertible projections stated without proof; verifying its hypotheses and constants is the main residual risk.","rationale":"The reader's weakest assumption already flags both the separation condition and the imported Lemma 5.2. I agree with that reading, but I want to isolate the exact place where the central claim could fail: it is not the algebraicity machinery (Theorem 5 is proved in the text), nor the upper bound (Theorem 2(a) is proved), but the lower bound in Theorem 2(b), which is the qualitative heart of Theorem A. The lower bound reduces to Lemma 5.2; because Q is non-invertible, the standard Falconer integral estimate does not literally apply, and the paper only asserts a 'simple extension.' There is no visible internal contradiction or circularity; the remaining proofs, including the delicate Theorem 5(b)-(e), appear coherent. Thus this is a verification concern, not a demonstrated error. I recommend CONDITIONAL acceptance: if Lemma 5.2 survives checking, the paper's claims are supported; if not, the almost-sure statements in Theorem A(b) and Theorem B need hypothesis adjustment.","tokens_in":51721,"tokens_out":32667,"duration_ms":343962,"concrete_test":"Independently prove Lemma 5.2 in the exact stated form: for singular Q of rank r, s∈(0,r)\\Z, and any contracting tuple A satisfying max_{i≠j}||Ai||+||Aj||<1, show that the translation integral is bounded by C φ_s(QA_{i∧j})^{-1} with C independent of i,j. Concretely, track Falconer's change-of-variables on the subspace transverse to ker Q; if a factor involving σ_1(QA_p)/σ_r(QA_p) appears, check whether it can be absorbed using the separation condition. Cross-check the statement against [8, §9.4] and [26, Lemma 2.2], and run a small numerical experiment for d=2, rank Q=1, A=(diag(0.6,0.3), diag(0.3,0.6)), comparing the integral against 1/φ_s(QA_p) over many p to verify uniformity of the constant.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central new content is the lower bound in Theorem 2(b), from which Theorem A(b) follows directly. In §5.2 that lower bound is reduced, via Lemma 5.3, to Lemma 5.2: for every distinct pair of codes i,j, ∫_{B(0,r)} ∥QΠ^v(i)-QΠ^v(j)∥^{-s} dv ≤ C φ_s(QA_{i∧j})^{-1}. This is the key place where the separation hypothesis max_{i≠j}||Ai||+||Aj||<1 enters the lower-bound proof, and the lemma is not proved in the preprint: it is said to follow from [8, §9.4] plus 'a simple extension' of [26, Lemma 2.2] to non-invertible Q. The classical statements are for Q=id, where the map from translation parameters to the difference is a submersion of full rank; for singular Q of rank r, the Jacobian degenerates in the kernel of Q, and the required bound must be exactly C/φ_s(QA_p), with no extra factor depending on the gap between singular values. If the correct constant carries an additional factor involving the singular values of Q A_p, the lower bound could fail precisely in the new phenomena of Theorems 3 and C. Since Theorem A(b), Corollary 2.3, and Theorem B all inherit this step, this is the most load-bearing unverified input in the paper.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends Falconer's 1988 theorem on self-affine fractals to the dimensions of linear projections of the attractor. For a contracting linearization A, it introduces a Q-projected affinity dimension, proves a variational formula and algebraicity of its sublevel sets (Theorems 1 and 5), and shows that for Lebesgue almost every affine IFS with linearisation A, the projected attractor has Hausdorff and box dimension equal to the Q-projected affinity dimension (Theorem 2, Theorem A). The same machinery is used to construct new examples: equilibrium measures on self-affine fractals whose projections are not exact-dimensional (Theorem C) and strongly irreducible self-affine sets with small sumsets (Theorem B). The proofs rest on an abstract thermodynamic formalism for semigroup potentials and on a projected version of Falconer's integral estimate.","tokens_in":51898,"tokens_out":16787,"duration_ms":168603,"significance":"If the results hold, this is a major advance. Theorem A strictly generalizes Falconer's theorem and Marstrand's theorem in the self-affine setting, and the algebraic stratification of exceptional projections is a new structural phenomenon. Theorem C provides the first general examples of non-exact-dimensional projections of equilibrium measures on self-affine sets, and Theorem B gives small sumsets without arithmetic resonance. The abstract Theorem 5 is a substantial contribution in its own right. The paper is also commendably explicit about which ingredients are imported from previous work, which makes the remaining gaps visible and verifiable.","major_comments":[{"comment":"The key integral estimate is stated with no proof: the text says it follows from the arguments of [8, §9.4] together with 'a simple extension' of [26, Lemma 2.2] to non-invertible Q. This lemma is load-bearing: it is the only place where the separation hypothesis max_{i≠j}||Ai||+||Aj||<1 enters Theorem 2(b), and the bound must hold with exactly C/φ_s(QA_{i∧j}), with no additional factor depending on the singular values of Q A_p. In the classical case Q=id the linear map from translation parameters to the difference is of full rank; for rank-deficient Q, the Jacobian degenerates on ker Q, so the stated extension is not a routine exercise. Since Theorem A(b), Corollary 2.3, and Theorem B all inherit this step, a complete proof or a precise published statement covering the non-invertible case is required.","section":"§5.2, Lemma 5.2"},{"comment":"The proof asserts that from ess sup_x dimloc((QΠ^v)_*μ, x) ≥ dim^Q_aff A it follows that dim_H QX^v ≥ dim_* (QΠ^v)_*μ ≥ dim^Q_aff A. The second inequality is false in general: dim_* is the essential infimum of local dimension, not the essential supremum. Moreover, the chosen (φ_s,Q)-equilibrium state may have nonconstant pointwise Lyapunov dimension, as the paper itself emphasizes in the Technical Remark after Theorem 1. A correct proof must argue via positive-measure superlevel sets where the local dimension is at least dim^Q_aff A − ε and then apply the mass distribution principle; this repair seems feasible, but the current text does not supply it.","section":"§5.2, set-dimension lower bound"},{"comment":"The lower-bound argument proves only dimloc ≥ sup{s : liminf (1/n log(φ_s(QA_{i|n})/μ([i|n]))) > 0}. The equality claimed in Theorem 2(b) with dim^Q_Lyap(A,μ)(i), defined as the infimum of s for which the pointwise limit is negative, is not established: if the pointwise limit equals zero on a nontrivial interval of s, the strict-positivity condition used to levy Lemma 5.3 fails, and the series ∑ μ([i|n])/φ_s(QA_{i|n}) need not converge. Since Theorem C is derived from this measure equality, the gap is load-bearing; the proof needs an argument covering the zero-limit plateau or a revised definition of the projected Lyapunov dimension.","section":"§5.2, measure clause of Theorem 2(b)"}],"minor_comments":[{"comment":"‘F alconer’ and ‘Marstand’ are typographical errors for ‘Falconer’ and ‘Marstrand’; the same misspelling recurs in the header and elsewhere in the Introduction.","section":"§1.1"},{"comment":"In the proof of Theorem 5(b), the line ‘let F₂ ⊂ Γ_I be a finite set such that ⋃_{j∈F₁} G^o g_j = G’ appears to use the wrong index set (F₁ should be F₂); the subsequent definition F := F₂F₁ suggests a typo.","section":"§3.4"},{"comment":"The statement of Lemma 5.2 does not specify the dependence of the constant C on r, Q, A, and the chosen norm; since the lemma is invoked with a fixed r, stating that C may depend on these parameters would avoid ambiguity.","section":"§5.2, Lemma 5.2"},{"comment":"The abstract and Introduction repeatedly refer to ‘Marstand’s theorem’ and ‘Marstand problems’; standard spelling is ‘Marstrand’, which is also the spelling used in the bibliography.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The two proof gaps in §5.2 are the main substantive obstacles. I would like to see Lemma 5.2 proved in full rather than cited as a ‘simple extension’, because the rank-deficient Jacobian issue is exactly where the new phenomena in Theorem 3 and Theorem C arise. If the non-invertible extension turns out to require an extra singular-value factor, the main dimension formula may fail. The authors should also repair the essential-supremum/essential-infimum implication and the zero-plateau issue before the paper can be accepted. The algebraic and thermodynamic parts of the paper appear sound and are likely to be influential."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a strong paper. It extends Falconer's theorem to projections, introduces a projected singular value pressure with a variational principle using an essential supremum, and proves that exceptional projections form algebraic varieties. Theorems B and C give genuinely new phenomena: small sumsets without arithmetic resonance, and non-exact-dimensional projections of equilibrium measures. The abstract Theorem 5 is proved in detail and is the technical core; the derivation of Theorem 1 from it is clear. I found no circularity.\n\nThe main weakness is the same one the stress-test flagged: the lower bound in Theorem 2(b) rests on Lemma 5.2, an integral estimate for non-invertible projections, imported from [8, §9.4] and [26, Lemma 2.2] with 'a simple extension' to singular Q. That lemma is load-bearing for Theorem A(b), Corollary 2.3, and Theorem B. I tried to see whether the singular Q causes a problem: it does not obviously. The integration over the kernel of Q is harmless, and the classical argument applied to the r-dimensional image should give φ_s(QA_p)^{-1} with a constant independent of p. But the paper does not show the details, and a referee should ask for them. The separation hypothesis max||Ai||+||Aj||<1 is clearly flagged and known to be necessary in this generality, so that is not a defect.\n\nThe exposition is dense but the logic is coherent. The paper acknowledges contemporaneous overlap with Feng-Xie [37], which is honest. If Lemma 5.2 checks out, the main results stand. Who is it for? Specialists in fractal geometry and thermodynamic formalism. It deserves a serious referee; I would send it out with a request to expand the proof of Lemma 5.2 or give a precise reference covering the non-invertible case.","headline":"Strong paper extending Falconer's theorem to projections, with a load-bearing imported lemma on non-invertible projections that should be checked before acceptance.","tokens_in":52512,"tokens_out":3681,"would_cite":true,"duration_ms":243654,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28A80","37C45","37H15"],"pacs":[],"model":"deepseek-v4-flash","headline":"For Lebesgue almost every affine iterated function system with linearisation A, the dimension of each projected attractor equals a projected affinity dimension computed from the singular values, and the exceptional projections form…","keywords":["self-affine fractals","affinity dimension","projection theorems","singular value pressure","exceptional projections","equilibrium states","sumsets","exact dimensionality"],"falsifier":"Compute the projected affinity dimension $\\dim^Q_{\\mathrm{aff}} A$ for a specific contracting tuple $A$ in $\\mathbb{R}^4$ whose linearisation generates a Zariski-dense subsemigroup of $\\mathbb{R}^*\\mathrm{SO}(2,2)$, choose a rank-two isotropic projection $Q$, and measure the Hausdorff dimension of $QX_v$ for Lebesgue-random translations; a positive-measure set of $v$ with $\\dim_H QX_v\\neq\\dim^Q_{\\mathrm{aff}} A$ would falsify Theorem A(b), while equality for a single generic translation would support it.","tokens_in":51424,"feed_emoji":"📐","tokens_out":7626,"duration_ms":78171,"temperature":0.7,"pith_summary":"This paper extends the 1988 theorem that for almost every self-affine set the affinity dimension gives the Hausdorff dimension, from the full set to every linear projection of it. The authors define a projected affinity dimension, $\\dim^Q_{\\mathrm{aff}} A$, computed from the singular values of the composites $QA_i$, and prove that under a pairwise contraction condition the Hausdorff dimension of $QX_v$ equals this number for Lebesgue almost every translation vector $v$, for each fixed projection $Q$. The projections split into finitely many algebraic families, each preserved by the group generated by the linear maps, and on each family the dimension is almost surely constant. The same machinery yields two further phenomena: equilibrium measures on self-affine sets whose projected measures are not exact-dimensional, and strongly irreducible self-affine sets with small sumsets but no arithmetic resonance.","feed_headline":"Self-affine fractal projections get an almost-sure dimension formula","feed_subtitle":"For typical affine fractals, projected dimensions are constant on algebraic families of projections.","key_machinery":"The carrying object is the projected singular-value potential and its pressure. For $s\\in[0,\\mathrm{rank}\\,Q]$, the potential is $\\varphi_s(QA)=\\sigma_1(QA)\\cdots\\sigma_{\\lfloor s\\rfloor}(QA)\\sigma_{\\lceil s\\rceil}(QA)^{s-\\lfloor s\\rfloor}$, and the projected pressure is $P_Q(A,s)=\\lim_{n\\to\\infty}\\frac1n\\log\\sum_{\\lvert i\\rvert=n}\\varphi_s(QA_i)$. The projected affinity dimension $\\dim^Q_{\\mathrm{aff}} A$ is the unique zero of this strictly decreasing function. Theorem 1 shows that the pressure limit exists without subadditivity, satisfies a variational formula whose optimisers are $(\\varphi_s,Q)$-equilibrium states, and has sublevel sets $\\{Q:P_Q(A,s)\\leq t\\}$ that are algebraic varieties invariant under right multiplication by every $A_i$. The proof runs through an abstract theorem for potentials $\\Psi_{U,b}$ built from operator norms on quotients of exterior-power representations of the Zariski-closed group generated by $A$, where quasi-multiplicativity yields unique $\\psi$-mixing equilibrium states and a module-theoretic lemma makes the sublevel sets algebraic.","core_discovery":"The central claim is that Falconer's almost-sure dimension formula extends from attractors to their linear images. For every contracting tuple $A=(A_i)_{i\\in I}$ and every $k=0,\\ldots,d$, there exist integers $s_m<\\cdots<s_1\\leq k$ and a finite filtration $\\varnothing=W_{m+1}\\subset W_m\\subset\\cdots\\subset W_1=\\mathrm{Gr}(k,d)$ of algebraic varieties, each invariant under the Zariski closure of the semigroup generated by $A$, such that every affine IFS with linearisation $A$ has $\\dim_H Q_U X\\leq s_j$ whenever $U\\in W_j\\setminus W_{j+1}$. Under the additional hypothesis $\\max_{i\\neq j}\\lVert A_i\\rVert+\\lVert A_j\\rVert<1$, equality $\\dim_H Q_U X=s_j$ holds for Lebesgue almost every affine IFS with that linearisation. The exponents $s_j$ are the zeros of a projected pressure function built from the singular values of $Q_UA_{i_1}\\cdots A_{i_n}$, and the equality is proved for upper box dimension as well as for the local dimensions of projected equilibrium measures, giving a stratified projection theorem for generic self-affine sets.","pith_inferences":["If Theorem A is correct, the exceptional projections of a typical self-affine set are organised by the representation theory of the Zariski closure of the semigroup generated by $A$, so classifying possible exceptional sets should reduce to decomposing exterior powers into simple $G^o$-modules.","Theorem B suggests that small sumsets of self-affine sets can arise from bundles of linear subspaces with nontrivial pairwise intersections rather than from arithmetic resonance; a testable extension would be to characterise the achievable dimension drop in terms of the intersection poset of those subbundles.","Theorem C implies that projected local Lyapunov dimensions need not be constant for ergodic measures, so any variational principle for dimensions of projected measures must treat essential suprema and essential infima separately rather than a single a.e.-constant value.","The finite bound $\\prod_{k=1}^d(1+2\\binom{d}{k})$ on the number of distinct level sets suggests that, for a fixed tuple $A$, the full list of exceptional-projection varieties is computable; testing small dimensions could reveal whether the bound is anywhere near sharp."],"forward_implications":["For every projection lying in a fixed algebraic level set of the projected affinity dimension, a Lebesgue-generic affine IFS with linearisation $A$ has $\\dim_H QX_v$ equal to the corresponding exponent.","When the group generated by $A$ acts irreducibly on every exterior power $\\wedge^\\ell\\mathbb{R}^d$, or when $A$ consists of similarities, the filtration is trivial, so for typical attractors every projection has the same dimension.","Theorem 2 implies a stratified projection theorem: on each algebraic variety $W_{t_j}$, for Lebesgue almost every translation vector the projected set has dimension $t_j$ for Lebesgue almost every projection in $W_{t_j}$.","The almost-sure statement cannot be upgraded to all projections, since the one-dimensional Sierpiński triangle has a countable dense family of exceptional projections from exact overlaps.","The method produces open sets of linearisations with nontrivial filtrations, giving a new general mechanism for constructing large families of exceptional projections of self-affine sets."],"supporting_citations":[{"why":"Supplies the theorem being generalised, the singular-value function $\\varphi_s$, and the original almost-sure dimension formula for attractors.","marker":"[26]"},{"why":"Provides the modern form of the contraction condition and the integral estimate, cited with section 9.4, that underpins Lemma 5.2.","marker":"[8]"},{"why":"Gives the current form of the almost-sure dimension theorem that the paper extends to projections.","marker":"[76]"},{"why":"Supplies the invariant-measure formulation of Falconer's lower-bound argument that the paper adapts to projections.","marker":"[48]"},{"why":"Contributes recent modifications of the classical covering and energy arguments used in the proof of Theorem 2.","marker":"[35]"},{"why":"Establishes equilibrium states for generalised singular-value potentials and strict monotonicity properties that Theorem 1 builds on.","marker":"[16]"},{"why":"Provides the quasi-multiplicative potential equilibrium-state theory invoked in Theorem 4.","marker":"[31]"},{"why":"Introduces strong quasi-multiplicativity and the $\\psi$-mixing property of equilibrium states used in Theorem 5.","marker":"[68]"},{"why":"Supplies the variational formula for subadditive potentials that grounds the pressure theory in Theorem 4.","marker":"[21]"}],"fun_headline_variants":["Self-affine projection dimensions now follow an almost-sure law","Falconer's projection theorem widened for self-affine attractors","Almost-sure projection dimension for self-affine fractals","Projection dimensions of self-affine fractals: generic law","Exceptional projections of self-affine fractals sit in algebraic varieties"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The almost-sure equalities in Theorem A(b) and Theorem 2(b) rest on the pairwise contraction assumption $\\max_{i\\neq j}\\lVert A_i\\rVert+\\lVert A_j\\rVert<1$, which the paper states cannot be removed entirely, and the proof also imports the integral estimate Lemma 5.2 from external sources rather than proving it.","fun_headline_variants_meta":{"raw":{"variants":["Self-affine projection dimensions now follow an almost-sure law","Falconer's projection theorem widened for self-affine attractors","Almost-sure projection dimension for self-affine fractals","Projection dimensions of self-affine fractals: generic law","Exceptional projections of self-affine fractals sit in algebraic varieties"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001338,"raw_usage":{"total_tokens":5408,"prompt_tokens":880,"completion_tokens":4528,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":496,"completion_tokens_details":{"reasoning_tokens":4454}},"tokens_in":496,"tokens_out":4528,"duration_ms":29272,"temperature":1.0,"reasoning_tokens":4454,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T23:54:28.570075+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the projected affinity dimension $\\dim^Q_{\\mathrm{aff}} A$ for a specific contracting tuple $A$ in $\\mathbb{R}^4$ whose linearisation generates a Zariski-dense subsemigroup of $\\mathbb{R}^*\\mathrm{SO}(2,2)$, choose a rank-two isotropic projection $Q$, and measure the Hausdorff dimension of $QX_v$ for Lebesgue-random translations; a positive-measure set of $v$ with $\\dim_H QX_v\\neq\\dim^Q_{\\mathrm{aff}} A$ would falsify Theorem A(b), while equality for a single generic translation would support it.","supporting_citations":[{"cited_title":"J.The Hausdorff dimension of self-affine fractals.Math","cited_arxiv_id":null,"evidence_quote":"Supplies the theorem being generalised, the singular-value function $\\varphi_s$, and the original almost-sure dimension formula for attractors."},{"cited_title":"Self-similar and self-affine sets and measures , vol","cited_arxiv_id":null,"evidence_quote":"Provides the modern form of the contraction condition and the integral estimate, cited with section 9.4, that underpins Lemma 5.2."},{"cited_title":"Measure and dimension for some fractal families.Math","cited_arxiv_id":null,"evidence_quote":"Gives the current form of the almost-sure dimension theorem that the paper extends to projections."},{"cited_title":"Hausdorff dimension for randomly perturbed self affine attractors","cited_arxiv_id":null,"evidence_quote":"Supplies the invariant-measure formulation of Falconer's lower-bound argument that the paper adapts to projections."},{"cited_title":"Dimensions of projected sets and measures on typical self-affine sets","cited_arxiv_id":null,"evidence_quote":"Contributes recent modifications of the classical covering and energy arguments used in the proof of Theorem 2."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes equilibrium states for generalised singular-value potentials and strict monotonicity properties that Theorem 1 builds on."},{"cited_title":"Equilibrium states for factor maps between subshifts","cited_arxiv_id":null,"evidence_quote":"Provides the quasi-multiplicative potential equilibrium-state theory invoked in Theorem 4."},{"cited_title":"The weak Bernoulli property for matrix Gibbs states","cited_arxiv_id":null,"evidence_quote":"Introduces strong quasi-multiplicativity and the $\\psi$-mixing property of equilibrium states used in Theorem 5."},{"cited_title":"The thermodynamic formalism for sub-additive potentials","cited_arxiv_id":null,"evidence_quote":"Supplies the variational formula for subadditive potentials that grounds the pressure theory in Theorem 4."}],"review_version":1}