{"id":"3e9026d3-dca0-48b7-9346-089d547952b8","arxiv_id":"1909.01935","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For two-dimensional Zeckendorf lattice decompositions, the probability that a gap vector equals (v1, v2) converges to 1/2^(v1+v2), a bivariate geometric distribution.","lead":"A number theory paper computes the gaps between terms in a two-dimensional version of Zeckendorf decompositions and proves they follow a simple geometric rule. This generalizes a known one-dimensional result and gives a clean mathematical answer for lattice-based decompositions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified.","rationale":"The reader's verdict is ACCEPT with high confidence, and I agree. The reader's weakest assumption concerned the possibility of axis points in simple jump paths, but the path definition in Section 2 explicitly requires strict decrease in every coordinate at every step, which rules out all axis points except the origin. I independently re-derived the main counts: Lemma 3.1 is a clean product count of prefix and suffix paths; Lemma 3.2's three cases are exhaustive for the same reason; Lemma 3.3 follows from the elementary identities for sums of squared binomial coefficients; and the Stirling limit in Theorem 1.3 evaluates to 2^{-(v1+v2)}. I also verified the combinatorics by hand for n=3, where the formula gives 18 gaps equal to (1,1) and 5 gaps equal to (2,2), matching direct path enumeration. No circularity, hidden parameter fitting, or unjustified normalization appears. The result is an internally consistent extension of the one-dimensional gap theorems to the two-dimensional lattice path model, and the paper is appropriately careful about the limitations of the d=2 analysis.","tokens_in":13226,"tokens_out":25598,"duration_ms":237274,"concrete_test":"For n=3 and n=4, enumerate all simple jump paths from (n+1,n+1) to (0,0) directly, tally gap-vector counts for all positive v1,v2, and compare the tallies with the closed form in Lemma 3.2.","verdict_should_be":"UNCHANGED","load_bearing_attack":"No load-bearing flaw found. The only potentially delicate step is the boundary case analysis in Lemma 3.2, where the count assumes a gap's lower endpoint can touch the axes only at (0,0) and its upper endpoint can touch the start only at (n+1,n+1). This is guaranteed by the strict-decrease definition of simple jump paths in Section 2: a point with one coordinate zero and the other positive cannot be followed by further strict decreases, and a point with one coordinate equal to n+1 other than the start cannot be reached from the start. The Stirling approximation step for fixed v1,v2 is also sound, and the total count in Lemma 3.3 is consistent with an independent telescoping check of the mean step count. An independent small-n enumeration confirms the formula in Lemma 3.2.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the distribution of gap vectors between consecutive points in legal decompositions (simple jump paths) on the two-dimensional Zeckendorf lattice. Working with paths from (n+1,n+1) to (0,0), the authors count the total number of occurrences of a fixed gap vector (v1,v2) by splitting on the location of the gap's lower endpoint, obtaining a closed-form binomial expression. Dividing by the total number of gap vectors, which is computed in two independent ways, and applying Stirling's formula, they prove Theorem 1.3: the probability that a uniformly chosen gap vector equals (v1,v2) tends to 1/2^{v1+v2}, i.e., a product of two independent geometric(1/2) distributions. Theorem 1.4 derives the limiting gap-sum distribution (v-1)/2^v. The paper also discusses the difficulty of extending the argument to d>=3 and outlines conjectures about gap distances and longest gaps.","tokens_in":13351,"tokens_out":16364,"duration_ms":138773,"significance":"If correct, this is a natural and nontrivial extension of the one-dimensional gap-decay results to a higher-dimensional lattice setting, and the clean bivariate geometric limit is a nice addition to the Zeckendorf literature. The proof is elementary but careful: the boundary case analysis in Lemma 3.2 is the delicate point, and it is justified by the strict-decrease property that rules out axis points except the origin; the concern that additional boundary terms might enter therefore does not land on reading the paper. I also checked the Stirling computation in the proof of Theorem 1.3 and the two proofs of Lemma 3.3 for internal consistency, and the combinatorial identities are sound. The paper is explicit about the limitations of the method for d>=3 and provides a self-contained second proof of the key total-count lemma, which strengthens the exposition.","major_comments":[],"minor_comments":[{"comment":"The remark that the extra origin gap 'can thus be safely ignored' is confusing because the proof actually counts it as Case (2) in Lemma 3.2; please clarify that the origin gap is included in the counting and that its contribution vanishes in the limit, or alternatively state that one may subtract it from both numerator and denominator without changing the limit.","section":"Section 3, near (3.1)"},{"comment":"The total count is denoted g(n+1;(v1,v2)) in Lemma 3.2 but is written as G(n+1;(v1,v2)) in the proof, which collides with the per-placement notation G((x,y),(x+v1,y+v2)) introduced in Lemma 3.1; please use consistent notation.","section":"Proof of Theorem 1.3, (3.22)"},{"comment":"Please state explicitly the probability model: a gap vector is chosen uniformly among all gap vectors occurring in all simple jump paths, or equivalently a path is chosen uniformly and then one of its gap vectors is chosen uniformly; without this, 'the probability that a gap vector equals (v1,v2)' is ambiguous.","section":"Theorems 1.3 and 1.4"},{"comment":"The sentence 'for any choice of starting point (a1,...,ad) in R^d' should say N^d or N_+^d, since the coordinates are integer lattice points.","section":"Section 2, Definition of simple jump paths"},{"comment":"For fixed positive integers (v1,v2), the binomial expressions implicitly assume n is large enough (roughly n>v1+v2); stating this explicitly would avoid evaluating binomial coefficients with negative entries.","section":"Lemma 3.2"},{"comment":"There are several typographical errors, for example 'shoud l' in the first proof of Lemma 3.3 and 'LA TTICE' in the paper header; a careful proofreading pass is advised.","section":"Throughout"}],"recommendation":"minor_revision","confidential_remarks":"The mathematical content is sound and the central claim is correct. The paper is a coherent extension of earlier work by Chen, Guo, Jiang, Miller, Siktar, and Yu, with appropriate credit to that paper and to [FJLLLMSS]. The remaining issues are purely expository, so I recommend minor revision rather than full acceptance as-is."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: this paper proves that gap vectors in 2D simple jump paths converge to a bivariate geometric distribution. The proof is correct, the boundary case analysis holds up, and the result is new as far as I can tell. It's a modest extension of the lattice Zeckendorf program, not a breakthrough, but it is a clean, honest piece of work.\n\nWhat's actually new: previous work by Chen, Guo, Jiang, Miller, Siktar, and Yu established Gaussian behavior for the number of summands in d-dimensional lattice decompositions, but no gap distribution. This paper fills that gap for d=2. The counting argument in Lemma 3.2 is the core: it correctly identifies the only valid boundary placements, (x,y)=(0,0) and (n-v1+1,n-v2+1), and the Vandermonde step gives a closed form. I checked the algebra in the Stirling limit; it works. They also provide a self-contained second proof of the total gap count, so no hidden dependence on the prior mean-step lemma.\n\nThe soft spots are mostly expositional. The title says 'Zeckendorf lattice,' but the actual result doesn't depend at all on the specific values y_{i,j} constructed in Section 1; it's a statement about lattice paths with strict coordinate decreases. That's fine, but the reader should know the real subject is path combinatorics, not number theory. The remark about the origin gap is also a bit sloppy: they say the contribution is negligible and can be ignored, but they actually count it in the finite-n normalization; what they mean is that it doesn't affect the limit. And the 'bivariate geometric' limit is really a product of two independent geometric(1/2) variables; pointing that out would make the result more transparent.\n\nThe bigger limitation is that they don't generalize to d>2, and they are honest about why: the Vandermonde identity doesn't have an obvious analogue. So the value is as a stepping stone, not a final word.\n\nBottom line: this is a sound paper, appropriately modest, and worth publishing in a specialized combinatorics or number theory journal. I'd accept it for review without hesitation. I wouldn't cite it in my own work, but I'd send it to a student looking for a clean project in this area.","headline":"Clean, correct limit theorem for 2D lattice gap vectors; modest but genuine extension of the Zeckendorf program.","tokens_in":13847,"tokens_out":3476,"would_cite":false,"duration_ms":31637,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11B02","05A02"],"pacs":[],"model":"deepseek-v4-flash","headline":"In the two-dimensional Zeckendorf lattice, a randomly chosen gap vector $(v_1,v_2)$ between consecutive summands has limiting probability $1/2^{v_1+v_2}$, a bivariate geometric law.","keywords":["Zeckendorf decompositions","simple jump path","two-dimensional lattice","gap vectors","bivariate geometric distribution","Vandermonde's identity","legal decomposition","binomial coefficients"],"falsifier":"Enumerate all simple jump paths from $(11,11)$ to $(0,0)$ and compare the number of gap vectors equal to $(1,1)$ divided by the total number of gap vectors with the value $316,030/1,108,536$ predicted by Lemmas 3.2 and 3.3; if the counts disagree, the boundary-case counting is wrong, since the theorem says the ratio converges to $1/4$ (and generally to $1/2^{v_1+v_2}$).","tokens_in":13061,"feed_emoji":"📐","tokens_out":13348,"duration_ms":112529,"temperature":0.7,"pith_summary":"Zeckendorf's theorem says every positive integer has a unique decomposition into non-adjacent Fibonacci numbers, and for generalized one-dimensional recurrences the gaps between summands decay geometrically. This paper asks whether the same gap behavior survives when decompositions are two-dimensional lattices, where a legal decomposition is a chain of lattice points moving strictly down and left. The central result is that for fixed positive integers $v_1$ and $v_2$, among all such simple jump paths from $(n+1,n+1)$ to the origin, the probability that a randomly chosen gap vector equals $(v_1,v_2)$ converges to $1/2^{v_1+v_2}$ as $n$ tends to infinity. That is a bivariate geometric limit, the exact two-dimensional analogue of the one-dimensional geometric decay, and it yields a limiting gap-sum distribution of $(v-1)/2^v$. The proof is combinatorial, using Vandermonde's identity to count paths and Stirling's formula to extract the limit, so the result extends the known one-dimensional theory to a genuinely multidimensional setting.","feed_headline":"Gap vectors in the Zeckendorf lattice follow a geometric law","feed_subtitle":"For a 2D legal decomposition, the chance a gap equals (v1,v2) tends to 1/2^(v1+v2).","key_machinery":"The central object is the simple jump path: a sequence of lattice points from a starting corner to the origin in which every coordinate strictly decreases at each step, so each step is a legal decomposition move in the $d$-dimensional Zeckendorf lattice. In two dimensions the number of such paths from $(a_1,a_2)$ to the origin is the binomial coefficient $\\binom{a_1+a_2-2}{a_1-1}$, obtained from Vandermonde's identity. The gap count splits each path at the two endpoints of a proposed gap: the number of paths containing a gap vector $(v_1,v_2)$ is a product of the number of paths from the lower endpoint to the origin and the number from the upper endpoint to the far corner, summed over all interior positions, with separate corner terms at $(0,0)$ and $(n-v_1+1,n-v_2+1)$. Stirling's formula evaluates the resulting ratio.","core_discovery":"The authors prove Theorem 1.3: start at $(n+1,n+1)$ and consider every simple jump path to $(0,0)$, where each step strictly decreases both coordinates; adding the origin as a final point just adds a negligible extra gap. For fixed positive integers $v_1,v_2$, the probability that a uniformly chosen gap vector among all such paths equals $(v_1,v_2)$ satisfies $P(v_1,v_2)=g(n+1;(v_1,v_2))/g_2(n+1)\\to 1/2^{v_1+v_2}$. The numerator is counted in Lemma 3.2 as $g(n+1;(v_1,v_2)) = (2n-v_1-v_2-1)\\binom{2n-v_1-v_2-2}{n-v_1-1} + 2\\binom{2n-v_1-v_2}{n-v_1}$, built from interior placements of the gap plus two corner placements at $(0,0)$ and the far corner, and the denominator is $g_2(n+1)=(n/2+1)\\binom{2n}{n}$, the total number of gap vectors. Stirling's approximation then drives the ratio to $1/2^{v_1+v_2}$. The immediate corollary (Theorem 1.4) is that the limiting probability that a gap sum equals $v\\ge 2$ is $(v-1)/2^v$, since exactly $v-1$ positive pairs sum to $v$.","pith_inferences":["If the same limit were sought for compound paths (steps allowed to move down, left, or diagonally), paths may touch an axis away from the origin; the additional boundary terms are exactly where the gap law could deviate from $1/2^{v_1+v_2}$.","For Euclidean gap distance, the limiting squared-distance probability would be $\\sum_{v_1^2+v_2^2=g} 2^{-(v_1+v_2)}$, a sum over representations of $g$ as a sum of two squares; the paper only discusses this qualitatively.","A relaxed model with weak decrease ($i'_j \\le i_j$) would allow axis-touching paths and could change the limiting distribution, so the strict-decrease convention is not a mere technical convenience but part of what the theorem asserts."],"forward_implications":["For fixed $v_1,v_2$, the limiting gap-vector probabilities are $2^{-(v_1+v_2)}$, so the two coordinates converge to independent geometric variables with parameter $1/2$.","The limiting gap-sum distribution is $P(v)=(v-1)/2^v$ for $v\\ge 2$, with $P(0)=P(1)=0$.","In the limit the expected gap sum is $4$, since each coordinate of the bivariate geometric law has mean $2$; small gaps dominate, with gap vector $(1,1)$ occurring with probability $1/4$.","Because the proof of the path count relies on Vandermonde's identity, the same argument does not directly extend to dimension $d\\ge 3$; the authors note that new combinatorial identities would be needed.","The result matches the one-dimensional geometric decay established for generalized Zeckendorf decompositions, showing that the two-dimensional lattice preserves this phenomenon even though unique decomposition is lost."],"supporting_citations":[{"why":"supplies the simple jump path model, the path-count formulas, and the mean number of steps that give the denominator in the gap probability.","marker":"[CCGJMSY]"},{"why":"proves the one-dimensional geometric decay of gap distributions that the two-dimensional result generalizes.","marker":"[BILMT]"},{"why":"states the original Zeckendorf unique decomposition theorem that motivates legal decompositions.","marker":"[Ze]"},{"why":"gives the one-dimensional average gap distribution, the setting in which the paper's gap-vector question is first posed.","marker":"[BBGILMT]"}],"fun_headline_variants":["Zeckendorf lattice gaps: bivariate geometric decay proven","2D Zeckendorf gaps converge to 1/2^(v1+v2)","Bivariate geometric law for Zeckendorf gap vectors","Zeckendorf lattice: gaps follow 1/2^(v1+v2) limit","2D Zeckendorf gap distribution tends to 1/2^(v1+v2)"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The counting assumes that a simple jump path never visits a lattice point with one coordinate zero except the final origin, so the only boundary placements of a gap are the two corners; if paths could touch an axis, additional boundary terms would enter the gap count and could change the limiting distribution.","fun_headline_variants_meta":{"raw":{"variants":["Zeckendorf lattice gaps: bivariate geometric decay proven","2D Zeckendorf gaps converge to 1/2^(v1+v2)","Bivariate geometric law for Zeckendorf gap vectors","Zeckendorf lattice: gaps follow 1/2^(v1+v2) limit","2D Zeckendorf gap distribution tends to 1/2^(v1+v2)"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000758,"raw_usage":{"total_tokens":3443,"prompt_tokens":1098,"completion_tokens":2345,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":714,"completion_tokens_details":{"reasoning_tokens":2245}},"tokens_in":714,"tokens_out":2345,"duration_ms":15543,"temperature":1.0,"reasoning_tokens":2245,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:03:54.334671+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Enumerate all simple jump paths from $(11,11)$ to $(0,0)$ and compare the number of gap vectors equal to $(1,1)$ divided by the total number of gap vectors with the value $316,030/1,108,536$ predicted by Lemmas 3.2 and 3.3; if the counts disagree, the boundary-case counting is wrong, since the theorem says the ratio converges to $1/4$ (and generally to $1/2^{v_1+v_2}$).","supporting_citations":[],"review_version":1}