{"id":"c83a77e9-8a94-4861-a6c1-3cc34853ba11","arxiv_id":"2505.05709","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For a Kakeya set in R^n whose line-segment midpoints lie in a set A of packing dimension at most s, the Hausdorff dimension is at least n-s, with bush-argument improvements, e.g. max{19/5-3s/5, 4-s} for n=4.","lead":"Kakeya sets are needle sets containing a line segment in every direction. This paper shows that when all segment midpoints must lie in a small set A of dimension at most s, the needle set still has Hausdorff dimension at least n-s, and often more via a bush argument.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The improved n−g_n(s) bound rests on the n-dimensional parallelogram maximal estimate (3.22), imported from [B91] without proof for n≥4; if that estimate is not literally in Bourgain, Theorem 2.6 is unsupported.","rationale":"The paper's central claim is a conditional improvement: from a Kakeya maximal estimate in dimension n−1, restricted Kakeya sets in R^n should have dimension at least max{n−s, n−g_n(s)}. The only step that connects the lower-dimensional estimate to the n-dimensional bush argument is Lemma 3.5. The authors flag the vulnerability themselves: only the n=3 case is explicit in [B91], and the extension to R^n is asserted as implicit in [B91, Page 158, (2.8)]. This is a missing proof/support issue that the review rules require flagging. If Lemma 3.5 holds, the main argument is coherent and plausible; if it fails or requires a stronger assumption, Theorem 2.6 collapses and the abstract's improved bound is not established. The reader's weakest_assumption identifies the same lemma, so I agree. The other issues mentioned by the reader, such as the mis-stated n=4 piecewise formula and the direction-rotation gap in Corollary 2.12, are real but secondary; they can be patched without changing the architecture of the proof. Since the reader already assigned CONDITIONAL and the primary unresolved point remains Lemma 3.5, no change in verdict is needed.","tokens_in":16734,"tokens_out":18042,"duration_ms":180616,"concrete_test":"Consult [B91] and check whether (2.8) on page 158 states or directly implies (3.22) for every n≥3; then independently re-derive (3.22) for shell-supported f from the assumed estimate (2.3), tracking all constants and δ losses. If the re-derivation goes through unchanged, the concern is resolved. If it requires an additional hypothesis or an extra δ factor, recompute the exponents in §3.3 and in Corollaries 2.7 and 2.11. If no proof can be supplied, the n≥4 improved bounds should be explicitly conditional on the parallelogram estimate (3.22).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The main new content of the paper is the improved lower bound n−g_n(s), and its only route is Theorem 2.6. In the proof of Theorem 2.6, equation (3.26) applies Lemma 3.5 to bound ∥M_δ(χ_{B_i^k})∥_{L^{p_{n−1}}(S^{n−1})}. Lemma 3.5 is not proved here; the text explicitly says that only the n=3 case is explicit in [B91, Lemma 1.52] and that the extension to R^n is “implicit (and used)” at [B91, Page 158, (2.8)]. For n≥4, the n-dimensional parallelogram maximal estimate (3.22) is therefore an unverified imported input. If (3.22) requires a stronger or different Kakeya-maximal hypothesis than (2.3), or if the cited page does not actually contain the all-dimension statement, the bush iteration in §3.3 does not establish the improved bound. The elementary n−s bound from Proposition A and Theorem 2.4 survives independently, but the advertised n−g_n(s) improvement, the n=4 example, and Corollaries 2.7–2.11 are unsupported without a proof of Lemma 3.5.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies restricted Kakeya sets in R^n: compact sets containing a unit line segment in every direction whose midpoints are constrained to lie in a prescribed set A. The main results give lower bounds for the Hausdorff dimension of such sets in terms of the upper box or packing dimension s of A. The elementary bound dim_H K_A ≥ n−s is proved by a covering/pigeonhole argument and a product argument. The main new contribution is an improved bound dim_H K_A ≥ n−g_n(s) obtained by adapting Bourgain's bush argument to the restricted setting, where g_n(s) is built from known Kakeya maximal estimates in dimension n−1. Explicit consequences are worked out for R^3, R^4, and R^{10}, and maximal-function analogues of the dimension statements are given. A final section extends the statements from midpoints to arbitrary selected points on each segment and from upper box dimension to packing dimension.","tokens_in":17034,"tokens_out":7769,"duration_ms":74481,"significance":"If the improved bounds are correct, the paper provides a genuinely new family of dimension estimates for restricted Kakeya sets, including an explicit bound in R^4 that beats both the trivial n−s bound and the current general Kakeya lower bound on part of the parameter range. The elementary Proposition A is clean and the maximal-function framework is well chosen. The paper is honest about its external inputs: it builds on Wolff, Hickman–Rogers–Zhang, and Bourgain rather than claiming a proof of the Kakeya conjecture. However, the central improved bound rests on an imported n-dimensional parallelogram maximal estimate that is not proved here, and the packing-dimension upgrade contains a logical gap. These issues are load-bearing, so the paper is not ready in its present form.","major_comments":[{"comment":"The n-dimensional parallelogram maximal estimate is imported from [B91] without proof, and the text itself states that only the n=3 case is explicit in [B91, Lemma 1.52], with the extension to all n asserted as implicit in [B91, p.158, (2.8)]. This estimate is the only route by which the bush iteration in Theorem 2.6 produces the improved exponent n−g_n(s); without it, Corollary 2.7 and all subsequent improved bounds, including the n=4 example, are unsupported. Please provide a complete proof of (3.22) or a precise reference containing the full statement with proof for all n≥4.","section":"Section 3.3, Lemma 3.5 / Eq. (3.22)"},{"comment":"The proof constructs K0 whose line segments cover only the direction set E0=∪ r_i(E_k), which has measure >1/2, but then applies Corollaries 2.5, 2.7, and 2.11, which require an A-restricted Kakeya set in the sense of Definition 2.1, i.e. a segment in every direction of S^{n-1}. Since K0 does not satisfy that definition, the application is invalid. The argument needs a genuinely proved positive-measure-direction version of the main theorems; as written, the packing-dimension upgrade and Corollary 2.13 are not established.","section":"Section 2.4, proof of Corollary 2.12"},{"comment":"The displayed formula after Corollary 2.7 misstates the branch for 3≤s≤4. For s≥3, the maximum of 4−(3−s)/p−(s−1) over 1≤p≤5/2 is attained at p=1 and equals 2, not 19/5−3s/5. The piecewise expression should have a separate branch 2 for 3≤s≤4, with the improved branch restricted to 1/2≤s<3. This error affects the stated theorem and Figure 1.","section":"Section 2.2, n=4 piecewise formula"}],"minor_comments":[{"comment":"Reference [C77] is misprinted: the American Journal of Mathematics entry gives volume 9, (2023), pages 1–22; it should be volume 99 (1977), pages 1–22.","section":"References"},{"comment":"The symbol E0 is used both for the initial measurable set E and for a 10δ/λ-separated subset of D0; the resulting notational collision is confusing and should be resolved.","section":"Section 3.3, proof of Theorem 2.6"},{"comment":"The proof sketch says all arguments go through for endpoints and for positive-measure direction sets, but no formal verification is supplied; since this extension is also needed to repair Corollary 2.12, it should be proved explicitly or the statement should be marked conditional.","section":"Corollary 2.13"},{"comment":"The hypothesis is stated with a universal quantifier over ε>0 in the preceding display, but the proposition text says 'for some ε>0'; the mismatch should be corrected.","section":"Proposition 3.2"},{"comment":"The abstract and Definition 2.1 discuss midpoints, while Corollary 2.13 later allows arbitrary selected points on each segment; the introduction would benefit from an early sentence signalling this intended generalization.","section":"Introduction/Abstract"}],"recommendation":"major_revision","confidential_remarks":"The main technical risk is the unverified n-dimensional parallelogram maximal estimate in Lemma 3.5. If the authors cannot supply a proof or a precise citation with a complete proof for n≥4, the advertised improved bounds for n≥4 should be either withdrawn or explicitly marked as conditional. The packing-dimension upgrade (Corollary 2.12) is a genuine logical gap, not merely a presentation issue. I would not recommend acceptance before these two points are resolved."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper introduces a genuinely new restricted Kakeya problem: the segments must have their midpoints (or, later, any fixed point on each segment) in a prescribed set A of packing or box dimension at most s. The elementary bound dim_H K_A ≥ n−s is clean and correct, via both a pigeonhole/covering argument and a product argument. The genuinely new content is the bush-argument improvement, dim_H K_A ≥ n−g_n(s), built from Kakeya maximal estimates in one lower dimension. For n=4 this yields max{19/5 − 3s/5, 4−s} on part of the range, beating current general Kakeya bounds there, and the same mechanism operates in higher dimensions using Hickman–Rogers–Zhang. The proofs are mostly transparent and the maximal-function analogues are a nice bonus.\n\nThe soft spots are real but local. First, the piecewise formula in Section 2.2 mis-states the s≥3 branch: maximizing over p gives 2, not 19/5−3s/5, because the p=1 term dominates there. The same type of issue appears in the general piecewise corollary; the saturated regime should be a constant 2, not a decreasing line. This does not damage the theorems but the examples need correcting. Second, Corollary 2.12 claims the full-direction results extend from upper box dimension to packing dimension. The proof covers A by pieces, selects one piece whose direction set has positive measure, rotates that direction set to fill most of the sphere, and then applies the full-direction corollaries to the rotated union. The corollaries as stated require segments in every direction. The proof sketch in Corollary 2.13 says the arguments go through for positive-measure direction sets, but that extension is not actually proved. A short lemma would fix this. Third, and most importantly, the improved bound depends on Lemma 3.5, a parallelogram maximal estimate imported from Bourgain. The authors note that only the n=3 case is explicit in [B91, Lemma 1.52] and that the n-dimensional version is only implicit. Since Theorem 2.6 uses this lemma at exactly one step, the n−g_n(s) bound for n≥4 is only as solid as that citation. If the estimate is not literally in Bourgain, the proof needs to supply it. This is the one thing I would require before publication.\n\nThe citations and attributions are otherwise appropriate, and the paper is honest about what is new and what is not. Researchers working on Kakeya-type problems, restricted maximal functions, or dimension theory of Besicovitch sets will find it useful. The central idea is sound and the flaws are addressable. I would send this to a serious referee rather than desk-reject it, and expect a revision that corrects the example formulas and tightens the two proof gaps.","headline":"A new restricted Kakeya framework with a clean elementary bound and a plausible bush-argument improvement, held up mainly by one imported parallelogram estimate that needs a proof or a precise reference.","tokens_in":17522,"tokens_out":7093,"would_cite":true,"duration_ms":74163,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["28A80","28A78"],"pacs":[],"model":"deepseek-v4-flash","headline":"Restricting the midpoints of the unit segments in a Kakeya set to a set of dimension at most $s$ forces the set to have Hausdorff dimension at least $n-s$, and the bush argument improves this to $n-g_n(s)$.","keywords":["Kakeya conjecture","Kakeya set","restricted Kakeya set","Kakeya maximal function","Hausdorff dimension","packing dimension","box dimension","bush argument"],"falsifier":"Construct, for some $s$ in the range where the improvement is claimed, an $A$-restricted Kakeya set in $\\mathbb{R}^4$ with $\\dim_B A\\le s$ but $\\dim_H K_A<19/5-3s/5$; equivalently, exhibit functions supported in annuli $B(0,2r)\\setminus B(0,r)$ for which the parallelogram maximal estimate (3.22) fails on $\\mathbb{R}^4$ at the stated exponents.","tokens_in":16547,"feed_emoji":"📐","tokens_out":20893,"duration_ms":196126,"temperature":0.7,"pith_summary":"This paper studies Kakeya sets whose defining unit segments are forced to pass through a prescribed set $A$: for every direction $e$ there is a unit segment in direction $e$ whose midpoint lies in $A$. If $A$ has packing dimension at most $s$, the authors prove that any such $A$-restricted Kakeya set in $\\mathbb{R}^n$ has Hausdorff dimension at least $n-s$; the simple argument is that $K-A$ contains a ball and the product inequality transfers the dimension. Their main theorem converts a Kakeya maximal estimate in dimension $n-1$ into a weak-type estimate for the $A$-restricted maximal function in $\\mathbb{R}^n$ via the bush argument, yielding the improved bound $n-g_n(s)$ for a function $g_n$ built from the lower-dimensional exponent. In $\\mathbb{R}^4$ this gives $\\max\\{19/5 - 3s/5, 4-s\\}$, and in higher dimensions it gives explicit bounds from known maximal estimates. The same lower bounds hold if $A$ only contains an arbitrary point on each segment rather than the midpoint, and they force full dimension $n$ when such a set has very small packing dimension.","feed_headline":"Restricted midpoints force Kakeya sets to be large","feed_subtitle":"In 4D the bound max(19/5 - 3s/5, 4 - s) beats general Kakeya bounds for some s.","key_machinery":"The central object is the $A$-restricted Kakeya maximal function $K_{\\delta,A}(f)(e)=\\sup_{a\\in A}|T_\\delta^e(a)|^{-1}\\int_{T_\\delta^e(a)}|f|$, where $T_\\delta^e(a)$ is the $\\delta$-neighbourhood of the unit segment in direction $e$ with midpoint $a$. Weak-type estimates for this operator are converted into Hausdorff dimension lower bounds by Lemma 2.3. The improved estimates are carried by the bush argument: iteratively find a point where many $\\delta$-separated tubes overlap, remove that bush, and control the remaining directions with an $n$-dimensional parallelogram maximal estimate that is imported from the lower-dimensional Kakeya maximal hypothesis. The elementary $n-s$ bound is carried instead by the fact that $K-A$ contains a ball of radius $1/2$, together with product dimension inequalities.","core_discovery":"The central claim is Theorem 2.6: if the Kakeya maximal function in $\\mathbb{R}^{n-1}$ satisfies an estimate of the form $\\|(f)^*_\\delta\\|_{L^{p_{n-1}}(S^{n-2})} \\lesssim_\\varepsilon \\delta^{-h_{n-1}-\\varepsilon}\\|f\\|_{L^{p_{n-1}}}$, then for any $A$ with upper box dimension at most $s$, the $A$-restricted Kakeya maximal function in $\\mathbb{R}^n$ satisfies a weak-type $L^p$ bound with $p=(p_{n-1}+n(p_{n-1}-1)+1)/p_{n-1}$ and exponent $\\beta=(h_{n-1}p_{n-1}+sp_{n-1}-s)/(p_{n-1}+n(p_{n-1}-1)+1)$. A reduction lemma converts this into the Hausdorff dimension lower bound $n-g_n(s)$, where $g_n(s)=h_{n-1}+s-s/p_{n-1}$; the elementary bound $n-s$ always holds, so the combined lower bound is $\\max\\{n-s,n-g_n(s)\\}$. The paper records packing-dimension versions and shows the same bounds hold when $A$ is merely met by every segment at some point, and it proves matching statements for the restricted Kakeya maximal function.","pith_inferences":["The transfer principle suggests that the restricted problem is the quantitative core of the Kakeya conjecture: because the restriction can only help, any counterexample to a restricted bound would also be a counterexample to the unrestricted conjecture, so the new bounds show how much midpoint freedom is needed to break the trivial product bound.","The authors' remark that the improved bound does not beat $n-s$ for very small $s$ leaves open a possible second transition near $s=0$; testing this in $\\mathbb{R}^4$ with near-extremal restricted sets around $s=1/2$ would clarify whether the bush bound is sharp.","The same bush and parallelogram machinery should extend to families of $k$-dimensional disks in place of unit segments, replacing the sphere of directions by a Grassmannian and using a base Kakeya estimate for $k$-planes, yielding analogous $n-s$ and bush-improved bounds.","In three dimensions, where the general Kakeya conjecture has been settled, the restricted problem is automatically solved; the meaningful new information from these bounds lies in four and higher dimensions, where the gap between the two bounds suggests where a sharper argument might begin."],"forward_implications":["In $\\mathbb{R}^4$, a Kakeya set whose segment midpoints lie in a set of upper box dimension $s$ has Hausdorff dimension at least $\\max\\{19/5-3s/5,4-s\\}$, which exceeds the general four-dimensional lower bound for a range of $s$.","The same dimension bounds hold when the prescribed set contains an arbitrary point of each segment, not necessarily its midpoint (Corollary 2.13).","If a Kakeya set contains a set $P$ of packing dimension less than $\\varepsilon$ meeting every direction's segment, then the set has Hausdorff dimension $n$ (Corollary 2.14).","Any future improvement of the Kakeya maximal estimate in dimension $n-1$ feeds through Theorem 2.6 to improve the restricted dimension bound in $\\mathbb{R}^n$.","The corresponding estimates for the $A$-restricted Kakeya maximal function hold as weak-type $L^p$ bounds, giving maximal-function analogues of each dimension statement."],"supporting_citations":[{"why":"This supplies the bush argument and the parallelogram maximal estimate Lemma 3.5 that transfer Kakeya maximal bounds from dimension $n-1$ to dimension $n$.","marker":"[B91]"},{"why":"This provides the three-dimensional Kakeya maximal estimate used with Theorem 2.6 to obtain the explicit four-dimensional bound $19/5-3s/5$.","marker":"[W95]"},{"why":"This supplies the higher-dimensional Kakeya maximal estimates used by Corollary 2.11 for dimensions five and higher.","marker":"[HRZ22]"},{"why":"This gives the reduction from maximal estimates to Hausdorff dimension used in Lemma 2.3 and the geometric tube-intersection facts used in the bush proof.","marker":"[W03]"},{"why":"This supplies the discrete equivalence for Kakeya maximal estimates used to apply the higher-dimensional results.","marker":"[M15]"},{"why":"This supplies the product dimension inequality underlying the elementary $n-s$ bound through the containment of a ball in $K-A$.","marker":"[F14,M95]"}],"fun_headline_variants":["Midpoint restriction raises Kakeya dimension floor to n - s and beyond","Kakeya dimension bound improves when midpoints lie in a small set","In 4D, restricted Kakeya sets have dimension at least max(19/5-3s/5,4-s)","Bush argument lifts Kakeya dimension bound for midpoint-constrained sets","New lower bound: Kakeya dimension at least max(n-s, n-g_n(s))"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The improved bound in dimensions $n\\ge 4$ rests on an $n$-dimensional parallelogram maximal estimate imported from an earlier work whose explicit case is only three-dimensional; if that estimate fails, the stronger $n-g_n(s)$ bound is not established.","fun_headline_variants_meta":{"raw":{"variants":["Midpoint restriction raises Kakeya dimension floor to n - s and beyond","Kakeya dimension bound improves when midpoints lie in a small set","In 4D, restricted Kakeya sets have dimension at least max(19/5-3s/5,4-s)","Bush argument lifts Kakeya dimension bound for midpoint-constrained sets","New lower bound: Kakeya dimension at least max(n-s, n-g_n(s))"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001087,"raw_usage":{"total_tokens":4598,"prompt_tokens":1053,"completion_tokens":3545,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":669,"completion_tokens_details":{"reasoning_tokens":3430}},"tokens_in":669,"tokens_out":3545,"duration_ms":27540,"temperature":1.0,"reasoning_tokens":3430,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T22:59:22.938044+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct, for some $s$ in the range where the improvement is claimed, an $A$-restricted Kakeya set in $\\mathbb{R}^4$ with $\\dim_B A\\le s$ but $\\dim_H K_A<19/5-3s/5$; equivalently, exhibit functions supported in annuli $B(0,2r)\\setminus B(0,r)$ for which the parallelogram maximal estimate (3.22) fails on $\\mathbb{R}^4$ at the stated exponents.","supporting_citations":[],"review_version":1}