{"id":"c0ffcccc-9c45-445b-ac6a-f801948c9d5b","arxiv_id":"2412.16912","paper_version":3,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":2.0,"correctness_risk":"low","formal_verification":"none","parameter_count":1,"one_line_summary":"Bouch's construction yields square-lattice trees that can be grown from a root in L!/C^L distinct ways for infinitely many bond counts L.","lead":"This paper presents a complete, simplified proof of Bouch's 2015 construction showing that certain rooted trees on a square grid can be assembled in almost L! different orders, up to an exponential factor. The construction matters because it is a key ingredient in rigorous arguments about how operators grow and how singularities form in quantum spin systems in two or more dimensions.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The unproved non-overlap condition (2.6) is the load-bearing step: if a rotated copy of T_{j-1} has width exceeding ℓ_{j-2} along the horizontal segment, adjacent branches collide, T_j is not a tree, and the weight recursion (2.12) collapses.","rationale":"The reader's weakest assumption matches the true soft spot of the paper. The proof of Theorem 1.1 is otherwise coherent: the weight recursion, the change of variables to E_j, and the final estimates are standard and check out, modulo a notational typo in (2.20) where '=' should be '<='. Lemma 2.1 is standard and a citation to Bouch's proof is acceptable in a review note, so the missing induction for Lemma 2.1 is less concerning than the unproved geometric non-overlap condition. The geometric condition is load-bearing because it is what guarantees that the recursively defined object is actually a tree and that the independent branch weights W_{j-1} multiply. The concern is not that the theorem is false; Bouch's original construction very likely works. It is that the present paper, despite claiming completeness, does not supply the one geometric lemma needed to make the recursive construction rigorous. This is an addressable gap rather than a fatal flaw, so the reader's CONDITIONAL verdict is appropriate; no verdict change is needed.","tokens_in":5474,"tokens_out":21677,"duration_ms":193604,"concrete_test":"Give an explicit recursive embedding rule (e.g., in T_j, place branches at x_i = iℓ_j/b_j for i=1,...,b_j, with every branch's subbranches oriented in the same direction) and prove by induction that a rotated copy of T_{j-1} occupies a strip of width ℓ_{j-2} in the direction parallel to T_j's horizontal segment; then (2.6) immediately implies disjointness. As a complement, generate the embeddings for a0=1 and j=1,...,6, and check by enumeration that no bond is occupied twice and no non-root vertex lies on the horizontal segment. If the induction cannot be carried out for the orientation described in Section 2.2, then (2.6) as stated is insufficient and the construction is incomplete.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.1 is proved by bounding W(T_j) through the recursion (2.12), which in turn assumes T_j really is a tree and that its b_j branches are disjoint copies of T_{j-1}. The only guarantee offered is the spacing condition (2.6), ℓ_j/b_j > ℓ_{j-2}, followed by the sentence 'to ensure that vertical branches do not overlap' and no proof. For this condition to suffice, every rotated copy of T_{j-1} attached to T_j must have width at most ℓ_{j-2} in the direction parallel to the horizontal segment of T_j. That is a genuine geometric lemma: T_{j-1} itself has subbranches that are rotated copies, and its width depends on the orientation chosen at each generation. The paper neither specifies the orientation explicitly nor proves the needed strip-width bound by induction. If an overlap occurred for some j, the bond set would fail to be a tree, W_j would be undefined, and the bound (2.12) would have no basis. Since the infinite set G in Theorem 1.1 uses all j, this is a central, not peripheral, gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a detailed exposition of Bouch's construction of rooted trees on the square lattice with growth count N(T) ≥ L!/C^L for infinitely many bond numbers L. The author defines a recursive family T_j, derives a recursion for the product weight W(T_j) = ∏ w(b), bounds log W(T_j) ≤ C L_j using a fast-growing parameter sequence, and then applies the Kupin–Bouch formula N(T)=L!/W(T) to conclude Theorem 1.1. The note is intended as a self-contained review and also discusses the implication for quantum operator growth.","tokens_in":5699,"tokens_out":18970,"duration_ms":138552,"significance":"If the geometric gap identified below is closed, the paper will be a valuable pedagogical contribution: it makes Bouch's construction accessible, gives explicit and checkable estimates, and clearly separates the algebraic weight recursion from the geometric packing. The derivation of the bound W_j ≤ C^{L_j} is transparent, the final constant is explicit, and the paper is honest about borrowing Lemma 2.1. It should be useful to researchers in quantum many-body physics and combinatorics.","major_comments":[{"comment":"The paper asserts that the condition ℓ_j/b_j > ℓ_{j−2} ensures that the b_j rotated copies of T_{j−1} do not overlap, but no proof is given and the relevant notion of \"width\" is never defined. For the weight recursion (2.12) to be valid, T_j must be an actual tree, which requires that each rotated copy of T_{j−1} intersect the horizontal segment only at its attachment site and that distinct copies be disjoint. One needs an explicit geometric induction: if h_k denotes the height of T_k (maximal extent perpendicular to its spine), then h_k = ℓ_{k−1} for the chosen orientation, so the width of a rotated T_{j−1} along the horizontal segment is at most ℓ_{j−2}; then (2.6) gives the needed separation. The paper should also state the orientation convention (e.g., all branches on one side of the segment, no sub-branches crossing the segment). This gap is load-bearing because the bound (2.12) is the basis for the final estimate (2.32).","section":"§2.2, Eq. (2.6)"}],"minor_comments":[{"comment":"The sentence \"The condition (2.6) is automatically satisfied\" is correct, since ℓ_j/b_j = 4E_{j−2} > ℓ_{j−2} = 4E_{j−2}E_{j−4}/E_{j−3}, but the computation should be displayed because the inequality is not immediately apparent from (2.13) and (2.14).","section":"§2.4"},{"comment":"Since the paper claims to give a complete presentation, include the short induction proof of N(T)=L!/W(T), or state more prominently that this is quoted from Bouch's Lemma 6.2 with a proof reference.","section":"Lemma 2.1"},{"comment":"The heading \"Descritpion\" is a typo; it should be \"Description\".","section":"§2.4 heading"},{"comment":"The caption says \"Bauch's sequence\"; it should say \"Bouch's sequence\".","section":"Figure 4 caption"},{"comment":"The phrase \"form (2.21) and (2.22)\" should be \"from (2.21) and (2.22)\".","section":"Just before Eq. (2.28)"},{"comment":"The claim N(T)=(L−1)!!∼√L! is not tied to a specific tree; please clarify which tree this refers to, or remove the sentence if it is not needed.","section":"Footnote 1"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is an exposition of a published theorem, and the algebraic core is sound. The missing geometric proof of the non-overlap condition is easily fixable, but it is essential to the construction; hence I recommend major revision rather than rejection. The paper is within the scope of math-ph as a pedagogical note."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Tasaki's note does exactly what it claims: it walks through Bouch's 2015 construction of trees on the square lattice with growth count at least L!/C^L for infinitely many L. The theorem is Bouch's, so the novelty is low, but the presentation is a real service. The new notation (E_j, the weight recursion in log form) and the cruder bound in (2.12) make the proof much easier to follow than the original. I checked the algebra: the recursion (2.20), the estimate (2.30), and the final bound (2.31) are correct. For someone who needs this result for operator growth or Lanczos coefficients, this is a good place to look.\n\nThe soft spots are minor but worth naming. The paper promises a complete presentation yet imports Lemma 2.1 (N(T)=L!/W(T)) from Bouch without proof; the pointer is fine for a review, but not for 'complete.' The bigger issue is the non-overlap condition (2.6). It is asserted, not proved, and the stress-test note is right that the orientation of the rotated branches is never spelled out. That said, the concern about it being load-bearing is overblown: for the explicit sequence, l_j/b_j = 4E_{j-2}, which is astronomically larger than l_{j-2}, so regardless of how the branches are flipped, the spacing is far more than enough to prevent collisions. What is missing is a sentence or a figure to back up the assertion, not a fix to the theorem. A referee would ask for that.\n\nThe paper is honest about what it is: a review, with the remaining open problem (all L, not just infinitely many) stated plainly. The physics motivation is cited appropriately. I'd send it to peer review; it deserves a referee, and the gaps are easy to close.\n\nFor me: I would cite this as a convenient reference for Bouch's theorem if I were writing about operator growth. Reading group? Maybe, if someone is working on that area.","headline":"A useful, honest review of Bouch's tree construction; the proof is correct, with two small gaps that are easy to fix.","tokens_in":6237,"tokens_out":8004,"would_cite":true,"duration_ms":64831,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05C05","05A16"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper gives a complete proof of Bouch's theorem: on the square lattice, infinitely many trees can be grown from a root in at least L!/C^L distinct orders.","keywords":["rooted trees","square lattice","growth orders","Bouch's construction","hierarchical construction","tree weight","operator growth","quantum spin systems"],"falsifier":"Take the paper's explicit parameters (for instance $a_0=1$ or $a_0=20$), write down the coordinates of every bond of $T_3$ or $T_4$, and check directly whether any two rotated copies of $T_{j-1}$ on the horizontal segment overlap. An overlap would mean the object is not a tree, so the weight recursion (2.12) would fail at that generation; verifying the absence of overlap for the first several generations would confirm the geometric condition in practice.","tokens_in":5238,"feed_emoji":"🌳","tokens_out":15868,"duration_ms":113990,"temperature":0.7,"pith_summary":"This paper reworks Bouch's 2015 construction into a complete and self-contained proof that on the square lattice $\\mathbb{Z}^2$ there are infinitely many bond counts $L$ for which a rooted tree $T$ can be grown from its root in at least $L!/C^L$ distinct ways, where $C>1$ is a fixed constant. Since $L!$ is the trivial upper bound on the number of growth orders, such trees come within a single exponential factor of the absolute maximum. The motivation is quantum many-body physics: as the paper notes (following the cited references), the existence of these trees is a key ingredient in results about operator growth and Lanczos coefficients in quantum spin systems in two or higher dimensions. The paper also records that the corresponding statement for every $L$, rather than only infinitely many $L$, remains open.","feed_headline":"Infinitely many square-lattice trees grow in nearly L! ways","feed_subtitle":"A hierarchical construction proves it for infinitely many tree sizes, feeding into quantum operator-growth bounds.","key_machinery":"Two devices carry the argument. The first is the tree weight $W(T)=\\prod_{b\\in T} w(b)$, where $w(b)-1$ counts the bonds downstream of $b$ (the 'flow' starts at the root); Lemma 2.1, attributed to Elizabeth Kupin, states that $N(T)=L!/W(T)$, turning the combinatorial growth count into a product estimate. The second is the hierarchical construction: $T_j$ consists of a horizontal segment of $\\ell_j$ bonds and $b_j$ rotated copies of $T_{j-1}$, with $\\ell_1=E_1$, $b_j=E_j/E_{j-1}$, $\\ell_j=4E_jE_{j-2}/E_{j-1}$, and $E_j=(a_j)^2$ where $a_j=2a_{j-1}$ (an iterated exponential, e.g. $a_2=2^{2^{20}}$ with the paper's $a_0=20$). The enormous growth of the $E_j$ makes the ratio $E_{k-2}/E_{k-1}$ so small that the series $\\sum_{k\\ge2}4(E_{k-2}/E_{k-1})\\log L_k$ converges; this yields $\\log W_j\\le C_2L_j$ and hence $W_j\\le C^{L_j}$. The spacing condition $\\ell_j/b_j>\\ell_{j-2}$ is asserted to keep the rotated branches from overlapping.","core_discovery":"The central claim is Bouch's theorem (Theorem 1.1): there exists a constant $C>1$ and an infinite set $\\mathcal{G}$ of positive integers such that, for every $L\\in\\mathcal{G}$, there is a rooted tree on the square lattice $\\mathbb{Z}^2$ with exactly $L$ bonds whose growth-order count $N(T)$ satisfies $N(T)\\ge L!/C^L$. The proof constructs a sequence of trees $T_1,T_2,\\dots$ hierarchically: $T_j$ is a horizontal segment of $\\ell_j$ bonds carrying $b_j$ rotated copies of $T_{j-1}$ as side branches. Using Kupin's lemma $N(T)=L!/W(T)$, where $W(T)$ is the product of downstream-bond count weights, the desired lower bound on $N(T_j)$ is equivalent to an upper bound $W(T_j)\\le C^{L_j}$. The paper proves this bound by choosing the branch parameters to grow so fast (a double exponential) that the per-bond logarithmic weight sums to a finite constant.","pith_inferences":["If the spacing condition ever failed, the construction could likely be repaired by lengthening the horizontal segment (increasing $\\ell_j/b_j$), since the convergence estimates only need the ratios $E_{j-2}/E_{j-1}$ to be small; the paper does not explore this repair.","A direct computational check for small $j$ (e.g. $j=2,3,4$) is feasible: with the explicit coordinates fixed, the exact value of $N(T_j)$ can be computed and tested against $N(T_j)\\ge L_j!/C^{L_j}$, providing an independent check of both the growth bound and the non-overlap condition.","The double-exponential parameter choice is likely far from minimal; trying slower growth (such as a fixed-height tower of exponentials) would test whether the same summability, and hence the theorem, survives with a smaller constant or a denser set of valid $L$."],"forward_implications":["For every $L$ in the infinite set $\\mathcal{G}$, the construction produces an explicit rooted tree on $\\mathbb{Z}^2$ with $N(T)\\ge L!/C^L$, giving a constructive counterpart to the non-constructive Bethe-lattice argument reported in the paper.","The weight bound $W(T_j)\\le C^{L_j}$ implies that the vast majority of bonds in $T_j$ lie in the youngest-generation side branches, with the bottom horizontal segment containing a vanishing fraction of the bonds as $j$ grows.","Via the cited works, the existence of such trees enters lower-bound results on operator growth and Lanczos coefficients in quantum spin systems in two or higher dimensions, which is the paper's stated motivation.","The theorem leaves open whether the same near-saturation holds for every bond count $L$; Bouch's construction covers only an infinite subsequence."],"supporting_citations":[{"why":"The original paper that introduced the hierarchical construction and proved the theorem; the present work is a complete reworking of that proof.","marker":"[1]"},{"why":"Cited as the work where the implication of the tree construction for operator growth in quantum spin systems is discussed.","marker":"[2]"},{"why":"Cited for a recent discussion (Appendix A.3) connecting the tree-counting result to Lanczos coefficients and local conserved quantities.","marker":"[3]"}],"fun_headline_variants":["Square-lattice trees that grow in almost L! ways","Bouch's trees: nearly factorial growth-order counts","Infinite family of trees with L!/C^L growth orders","Hierarchical trees achieve near-L! growth on square lattice","Reviewing Bouch's construction of high-growth trees"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole proof assumes that the side branches (which are rotated copies of the previous tree) attached along the horizontal segment never overlap each other or the segment; the paper asserts a spacing inequality guarantees this but does not prove it for all generations.","fun_headline_variants_meta":{"raw":{"variants":["Square-lattice trees that grow in almost L! ways","Bouch's trees: nearly factorial growth-order counts","Infinite family of trees with L!/C^L growth orders","Hierarchical trees achieve near-L! growth on square lattice","Reviewing Bouch's construction of high-growth trees"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00029,"raw_usage":{"total_tokens":1653,"prompt_tokens":861,"completion_tokens":792,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":477,"completion_tokens_details":{"reasoning_tokens":711}},"tokens_in":477,"tokens_out":792,"duration_ms":7641,"temperature":1.0,"reasoning_tokens":711,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T05:59:21.868684+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the paper's explicit parameters (for instance $a_0=1$ or $a_0=20$), write down the coordinates of every bond of $T_3$ or $T_4$, and check directly whether any two rotated copies of $T_{j-1}$ on the horizontal segment overlap. An overlap would mean the object is not a tree, so the weight recursion (2.12) would fail at that generation; verifying the absence of overlap for the first several generations would confirm the geometric condition in practice.","supporting_citations":[],"review_version":1}