{"id":"59c6a375-ed05-4801-acff-a99554b9779d","arxiv_id":"2502.03926","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":1.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"This survey shows how the Assouad spectrum, intermediate dimensions, and Fourier spectrum yield sharper projection theorems for fractal sets.","lead":"Dimension interpolation inserts continuous families of dimensions between familiar fractal dimensions, and this survey shows how three such families improve what is known about orthogonal projections of fractals. A generalist should read it as a compact map of a young programme that has already sharpened several results in fractal geometry.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 6.3 rests on unproved estimate (6.4); the Fourier-spectrum application is conditional on the cited [FdO24+].","rationale":"The survey's central assertions are Theorems 4.1, 5.3, and 6.1/6.3. I found no internal inconsistency in the Assouad-spectrum argument of Section 4: the dyadic potential estimate in the proof of Theorem 4.3 is sound, with the upper Assouad spectrum providing the r^{-(α-s)} and r^{-(β-s)(1-θ)} terms. Section 5's capacity framework is used correctly, and Corollary 5.6 is a genuine application of intermediate-dimension continuity; the well-definedness of profiles is explicitly cited to [BFF21], which is acceptable for a survey. The only load-bearing weakness is in Section 6: the proof of Theorem 6.3, which is otherwise self-contained, hinges on the unproved estimate (6.4), with (6.5) also stated without proof. The reader's weakest_assumption identified exactly this point. Since the cited preprint is available and the estimate is plausible, this is a completeness issue that warrants a conditional verdict or a simple label as a proof sketch, rather than a rejection. Thus the reader's CONDITIONAL verdict stands unchanged.","tokens_in":20430,"tokens_out":27064,"duration_ms":258104,"concrete_test":"Open arXiv:2404.11179, locate the proof of (6.4), and verify the dyadic estimate: for a Frostman measure ν of exponent τ on G(d,k), confirm that ν{V : d(z,V) ≤ r} ≲ (r/|z|)^{τ-(k-1)(d-k)} (the survey's (6.5)) and that splitting the y-integral into annuli around z yields the constant |z|^{u/θ + k(d-k) - d - τ} with N chosen > τ + 1. Reproduce the k = 1, d = 2 case by hand; if the exponent matches, Theorem 6.3 is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 6.3 in Section 6 is the part of the survey that carries the Fourier-spectrum application, and it stops at the unproved estimate (6.4). Everything after (6.4) is a routine comparison of exponents using s < dim^θ_F µ and the choice of τ, but (6.4) is the analytic heart: it converts the Frostman bound (6.5) on the Grassmannian into the |z| power needed to make the averaged Fourier integral finite. The text explicitly defers the proof to [FdO24+], and (6.5) is also asserted without derivation. If (6.4) were false or had a different exponent, Theorem 6.3's bound (6.1) and Corollary 6.4 would not follow. This is a completeness gap in the preprint as written, not evidence that the theorem is false; the heuristic dyadic splitting with shells {2^{j-1} < |yV - z| ≤ 2^j} is plausible and likely correct in [FdO24+]. But because the preprinted proof does not establish it, the central Fourier-spectrum claim is conditional on an external, not-yet-refereed source (arXiv:2404.11179).","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This survey paper presents the program of dimension interpolation—viewing Hausdorff, box, Assouad, and Fourier dimensions as endpoints of parameterized spectra—and concentrates on three applications to the dimension theory of orthogonal projections. The three spectra treated are the Assouad spectrum (Section 4), the intermediate dimensions (Section 5), and the Fourier spectrum (Section 6). For each, the paper states theorems from recent work and supplies proofs or proof sketches: Theorem 4.3 is proved in detail and yields the lower bound for box dimension profiles in Theorem 4.1; Section 5 states and partially proves results from Burrell–Falconer–Fraser on intermediate dimension profiles and a Marstrand-type theorem, with consequences relating box dimensions of projections to continuity of intermediate dimensions at zero; Section 6 states and attempts to prove Theorem 6.3 on exceptional-set estimates for Hausdorff dimension using the Fourier spectrum, with Corollary 6.1 as the headline application.","tokens_in":20598,"tokens_out":24882,"duration_ms":209693,"significance":"If the results presented are correct, the survey provides a useful unified view of a growing body of work, and it highlights a concrete theme: interpolation spectra can yield projection theorems that are stronger than what the endpoint notions alone provide. The underlying theorems are published or accepted in peer-reviewed venues, and the paper gives substantial value by organizing them, offering a worked example (X = {1/n} × [0,1]) where all three spectra are computed exactly, and stating two open questions about Marstrand theorems for the Assouad and Fourier spectra. The paper also includes machine-checkable detailed proofs of several auxiliary results, specifically Theorem 4.3, Lemma 5.2, and Corollaries 5.4–5.6, which enhances its usefulness for readers new to the area.","major_comments":[{"comment":"The proof of Theorem 6.3 is not complete as written. The decisive estimate (6.4), which converts the Frostman bound on the Grassmannian into the |z| power needed to make the averaged Fourier integral finite, is asserted without proof and is deferred to [FdO24+] with only a dyadic-splitting sketch. The related bound (6.5) is also asserted without derivation. Since the entire Fourier-spectrum application rests on (6.4), the paper should either provide a proof of (6.4) (for example, as a separate lemma with a complete argument) or explicitly label Theorem 6.3 as a result proved in [FdO24+] and the present argument as a proof sketch. As it stands, the phrase 'we now state and prove the main result' overstates the completeness of the exposition, even though the cited source is reliable.","section":"Section 6, proof of Theorem 6.3"}],"minor_comments":[{"comment":"The proof states that the lower bound follows from Theorem 4.1 by letting θ → 0, but the quasi-Assouad dimension dim_qA X appears in the limit θ → 1, not θ → 0 (since dim_qA X = lim_{θ↗1} dim_A^θ X). The displayed limit should be corrected from θ → 0 to θ → 1.","section":"Corollary 4.2, proof"},{"comment":"The proof refers to 'Recalling (5.1)' when bounding the potential of the measure µ, but the kernel ϕ_r^s was defined in (3.3), not in (5.1). Please update the cross-reference.","section":"Section 4, proof of Theorem 4.3"},{"comment":"The final sentence of the proof says 'taking α and β arbitrarily close to dim_A^θ F and dim_A F respectively,' but 'F' should be 'X' in both occurrences.","section":"Section 4, proof of Theorem 4.3"},{"comment":"The proof refers to 'Corollary 6.1' when applying the exceptional-set estimate; the correct reference is Theorem 6.1.","section":"Section 6.1, proof of Corollary 6.4"},{"comment":"In the display after (6.2), the notation 'cµV (y) = bµ(yV)' appears to be a typographical corruption of the intended '\\widehat{µ_V}(y) = \\hat{µ}(y_V)'. Please fix the typesetting.","section":"Section 6, proof of Theorem 6.3"},{"comment":"The example of the Lebesgue measure restricted to [0,1] says its Fourier and Sobolev dimensions 'exceed the Hausdorff dimension of the ambient space.' The intended comparison is with the Hausdorff dimension of the support, not the ambient space; please rephrase for clarity.","section":"Section 2.3"},{"comment":"The upper Assouad spectrum and the standard Assouad spectrum are denoted by nearly identical symbols (dim_A^θ X vs. \\overline{dim}_A^θ X). Please ensure the typesetting clearly distinguishes them, especially in Theorem 4.1, Theorem 4.3, and Corollary 4.2, where the difference is mathematically relevant.","section":"Notation throughout"}],"recommendation":"minor_revision","confidential_remarks":"The survey is heavily self-referential: the author introduced the Fourier spectrum and co-authored the three application papers [FFS21], [BFF21], and [FdO24+]. This is not a mathematical problem, but the editors may wish to consider whether the balance of the survey is sufficiently broad; several other recent papers on these spectra are cited, but the showcased applications are all from the author's own group. The main revision requested—clarifying the status of the proof of Theorem 6.3—should be straightforward to address."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this is a survey, not a research paper, and it is a good one. Fraser's stated goal is to show that dimension interpolation—the Assouad spectrum, intermediate dimensions, Fourier spectrum—has concrete payoffs for Marstrand projection theory. He succeeds: the narrative is clear, and the three applications genuinely show the spectra doing work that the classical dimensions alone cannot.\n\nWhat is actually new? Not much, and the paper does not pretend otherwise. Every theorem is attributed to earlier papers, mostly the author's own: Theorem 4.1 from Falconer–Fraser–Shmerkin, Theorem 5.3 from Burrell–Falconer–Fraser, Theorems 6.1/6.3 from Fraser–de Orellana. The one un-attributed item, the closed-form box dimension for projections of F_p × F_p, is explicitly left as an exercise. So novelty is not the point; synthesis is. That synthesis is well done. Section 4's proof of Theorem 4.3 is complete and gives a genuine flavour of the capacity method. Lemma 5.2 is also proved in full. The expository sections comparing the three spectra on the example {1/n} × [0,1] are illuminating.\n\nThe soft spot is Section 6, and it is exactly where the reader's eye lands. The text announces it will prove Theorem 6.3, but the proof stops at estimate (6.4), the analytic heart that converts the Frostman bound (6.5) on the Grassmannian into the |z| power needed for the averaged Fourier integral. We get 'the proof of (6.4) is technical' and a referral to [FdO24+]. That makes the Fourier-spectrum exceptional-set theorem conditional, in this preprint, on an external source. I believe the heuristic—dyadic splitting into shells—is plausible and I have no reason to doubt the result, but as written the proof is incomplete. If the paper is going to say 'we prove', it should either include (6.4) or call it a proof sketch. This is a fixable presentation problem, not a sign the mathematics is wrong.\n\nThe self-referential caution is worth stating but not overblown. Yes, the author is surveying his own programme, and the three application papers are co-authored by him. But those papers are published or accepted in respectable venues, and the survey's internal proofs are honest. Self-citation is not a flaw when the results are real.\n\nBottom line: this is a useful survey for anyone wanting an entry point to dimension interpolation and its projection applications. It deserves a serious referee; a good referee will ask for the Section 6 label to be corrected or the estimate included. I would accept it after minor revision.","headline":"A clear, honest survey of the dimension-interpolation programme, with complete proofs for the Assouad and intermediate-dimension applications but a real gap where the Fourier-spectrum theorem defers its key estimate.","tokens_in":21202,"tokens_out":3399,"would_cite":true,"duration_ms":29724,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28A80","42B10","28A75","28A78"],"pacs":[],"model":"deepseek-v4-flash","headline":"Dimension interpolation sharpens what we can say about almost all orthogonal projections of fractals.","keywords":["dimension interpolation","orthogonal projections","Marstrand projection theorem","Hausdorff dimension","box dimension","Assouad spectrum","intermediate dimensions","Fourier spectrum"],"falsifier":"Extract the deferred proof of (6.4) from [FdO24+] and test it on a concrete compactly supported measure with known Fourier spectrum: compute the integral over $G(d,k) \\times \\mathbb{R}^k$ of $(1+|y_V - z|)^{-N} |y|^{u/\\theta - k}$ with respect to a Frostman measure $\\nu$, and check whether it is bounded by a constant times $|z|^{u/\\theta + k(d-k) - d - \\tau}$ for $|z| \\geq 2$; if a single measure or direction set violates this bound, the averaging argument in Theorem 6.3 breaks and the exceptional-set theorem is false as stated.","tokens_in":20153,"feed_emoji":"📐","tokens_out":7241,"duration_ms":64449,"temperature":0.7,"pith_summary":"Dimension interpolation treats Hausdorff, box, Assouad and Fourier dimension not as separate notions but as endpoints of continuous spectra indexed by a parameter $\\theta$. This survey argues that these intermediate spectra carry geometric information the endpoints hide, and demonstrates the payoff in the dimension theory of orthogonal projections. Concretely, the paper claims the Assouad spectrum yields new lower bounds for box dimension profiles, the intermediate dimensions satisfy a Marstrand-type theorem with almost sure values given by interpolated profiles, and the Fourier spectrum produces exceptional-set estimates for the Hausdorff dimension of projections that recover and refine the classical Peres–Schlag bound. These results matter because they turn a philosophical unification into testable inequalities about how large a typical projection is and how large the set of bad directions can be.","feed_headline":"Spectra between fractal dimensions sharpen projection theorems","feed_subtitle":"Assouad, intermediate, and Fourier spectra improve almost-sure and exceptional estimates for orthogonal projections.","key_machinery":"The machinery is the three interpolation spectra plus the potential-theoretic dimension profiles. The Assouad spectrum fixes the relationship between localisation scale $R$ and covering scale $r = R^{1/\\theta}$; the intermediate dimensions restrict all cover sets to diameters between $r$ and $r^\\theta$; the Fourier spectrum replaces uniform Fourier decay by an $L^{2/\\theta}$ average, interpolating between Fourier dimension at $\\theta = 0$ and Sobolev/Hausdorff dimension at $\\theta = 1$. On top of these spectra, the box dimension profiles defined by the kernels $\\phi^s_r(x) = \\min\\{1, (r/|x|)^s\\}$ carry the projection theory, and Section 5 introduces a family of kernels $\\phi^{s,k}_{r,\\theta}(x)$ whose capacities define intermediate dimension profiles. The Fourier-spectrum proof works by assuming the exceptional set is large, building a Frostman measure $\\nu$ on it, and showing the averaged projected energy is finite via the estimate (6.4), a decay bound on an integral over directions that is deferred to [FdO24+].","core_discovery":"The central claim is that the spectra ‘live in between’ familiar dimensions and often behave differently from either endpoint, and that this behaviour has direct consequences for projections. In the projection setting the survey establishes three families of results. Theorem 4.3 bounds the upper box dimension profile $\\dim^{s}_{B} X$ below by $\\dim_{B} X - \\max\\{0, \\dim^{\\theta}_{A} X - s, (\\dim_{A} X - s)(1-\\theta)\\}$, which Theorem 4.1 converts into a lower bound for the upper box dimension of $P_V(X)$ for almost every $V$. Theorem 5.3 shows that the intermediate dimensions of projected sets are almost surely constant and equal to the intermediate dimension profile $\\dim^{k}_{\\theta} X$ for all $\\theta \\in (0,1]$ simultaneously, interpolating between $\\min\\{k, \\dim_H X\\}$ and the box dimension profile. Theorem 6.3 bounds the dimension of directions $V$ for which the Fourier spectrum of the projected measure drops below $u$, namely $\\dim_H\\{V : \\dim^{\\theta}_{F} \\mu_V < u\\} \\le \\max\\{0, k(d-k) + (u - \\dim^{\\theta}_{F} \\mu)/\\theta\\}$, and Theorem 6.1 uses the infimum over $\\theta$ to control the Hausdorff dimension exceptional set in the Marstrand–Mattila theorem. These are presented as genuine applications, not illustrations: the spectra supply information that the individual dimensions do not.","pith_inferences":["The survey notes that an affirmative answer to the open Marstrand questions for the Fourier or Assouad spectra would upgrade, by continuity, to almost sure constancy for all $\\theta$ simultaneously; that upgrade is an immediate corollary of the continuity results quoted here.","A pattern visible across the three applications is that endpoint continuity of an interpolated spectrum is what converts interpolation data into projection theorems; the same mechanism may give projection-type statements for other spectra constructed between other endpoint dimensions.","The identical Fourier-spectrum strategy of Hölder splitting, Schwartz localisation, and a direction-average decay estimate resembles arguments used for the Falconer distance problem and restriction estimates, so the exceptional-set theorem may transfer to radial projections or distance-set variants rather than only orthogonal projections.","Because the derivative condition $D \\ge k(d-k)$ in Corollary 6.4 forces $k(d-k) \\le d$, the current Fourier-spectrum method is sharpest for hyperplane projections and 4-dimensional projections into 2-planes; a spectrum with faster growth near $0$ would extend the continuity conclusion to all $k$."],"forward_implications":["Theorem 4.1 gives the almost sure lower bound $\\dim_B P_V(X) \\ge \\dim_B X - \\max\\{0, \\dim^{\\theta}_A X - k, (\\dim_A X - k)(1-\\theta)\\}$; in particular, if the quasi-Assouad dimension is at most $\\min\\{k, \\dim_B X\\}$, the typical projection has box dimension exactly $\\min\\{k, \\dim_B X\\}$.","Theorem 5.3 establishes a Marstrand theorem for intermediate dimensions: for almost all $V$ the equality $\\dim^{\\theta} P_V(X) = \\dim^k_{\\theta} X$ holds for every $\\theta \\in (0,1]$, so the whole curve of projected intermediate dimensions is almost surely constant, not just a single number.","Corollary 5.6 says that for sets whose intermediate dimension is continuous at $\\theta = 0$, the typical projection has full box dimension $k$ exactly when $\\dim_H X \\ge k$; this yields the surprising conclusion that Bedford–McMullen carpets and sets $F_p \\times F_p$ with $\\dim_H < 1$ have typical projections of box dimension strictly below $1$.","Theorem 6.1 bounds the exceptional set for Hausdorff dimension projections by $\\max\\{0, k(d-k) + \\inf_{\\theta} (u - \\dim^{\\theta}_F X)/\\theta\\}$, recovering the Peres–Schlag bound at $\\theta = 1$ and improving it when the Fourier spectrum rises steeply from $0$.","Corollary 6.4 gives continuity of the exceptional-set dimension at $u = \\dim_F X$ when the lower right semi-derivative $D$ of the Fourier spectrum at $0$ is at least $k(d-k)$, eliminating the jump that can occur using Fourier dimension alone."],"supporting_citations":[{"why":"Introduces the capacity approach to box and packing dimensions of projections and proves Theorem 3.1, the almost sure projection theorem for box dimension profiles that the survey feeds with spectra.","marker":"[F21]"},{"why":"Source of Theorem 4.3 and Theorem 4.1, showing the Assouad spectrum bounds the box dimension profiles from below.","marker":"[FFS21]"},{"why":"Source of the intermediate dimension profiles, the Marstrand theorem for intermediate dimensions (Theorem 5.3), and Corollaries 5.4–5.6.","marker":"[BFF21]"},{"why":"Source of the Fourier-spectrum exceptional set estimates (Theorems 6.1 and 6.3) and the analytic estimate (6.4) whose proof is deferred.","marker":"[FdO24+]"},{"why":"Provides the baseline exceptional-set bound (3.1) that Theorem 6.1 recovers at $\\theta = 1$ and introduced the Sobolev dimension used to define the Fourier spectrum.","marker":"[PS00]"},{"why":"Introduces intermediate dimensions and supplies the continuity-at-0 results for Bedford–McMullen carpets and sequence sets used in the applications of Corollary 5.6.","marker":"[FFK20]"},{"why":"Introduces the Assouad spectrum and the interpolation philosophy on which Section 4 rests.","marker":"[FY18]"},{"why":"Introduces the Fourier spectrum and proves the continuity and concavity properties needed in Section 6.","marker":"[F24]"},{"why":"Establishes Marstrand's planar projection theorem, the classical result the survey's applications extend.","marker":"[M54]"},{"why":"Generalises the projection theorem to higher dimensions via potential theory, together with Kaufman's method.","marker":"[M75]"}],"fun_headline_variants":["Dimension spectra sharpen projection estimates","Interpolating dimensions improves projection theorems","Fourier, Assouad, and intermediate spectra refine projections","New dimension interpolation advances orthogonal projections","Spectra between dimensions enhance projection results"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumptions are the deferred analytic estimate (6.4) behind the Fourier-spectrum theorems and, in Section 5, the standing assumption from [BFF21] that the intermediate dimension profiles are well defined and continuous at $\\theta = 0$; if either fails, the corresponding projection theorems lose their support.","fun_headline_variants_meta":{"raw":{"variants":["Dimension spectra sharpen projection estimates","Interpolating dimensions improves projection theorems","Fourier, Assouad, and intermediate spectra refine projections","New dimension interpolation advances orthogonal projections","Spectra between dimensions enhance projection results"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000277,"raw_usage":{"total_tokens":1688,"prompt_tokens":1024,"completion_tokens":664,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":640,"completion_tokens_details":{"reasoning_tokens":600}},"tokens_in":640,"tokens_out":664,"duration_ms":7370,"temperature":1.0,"reasoning_tokens":600,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T00:12:48.440298+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Extract the deferred proof of (6.4) from [FdO24+] and test it on a concrete compactly supported measure with known Fourier spectrum: compute the integral over $G(d,k) \\times \\mathbb{R}^k$ of $(1+|y_V - z|)^{-N} |y|^{u/\\theta - k}$ with respect to a Frostman measure $\\nu$, and check whether it is bounded by a constant times $|z|^{u/\\theta + k(d-k) - d - \\tau}$ for $|z| \\geq 2$; if a single measure or direction set violates this bound, the averaging argument in Theorem 6.3 breaks and the exceptional-set theorem is false as stated.","supporting_citations":[],"review_version":1}