{"id":"e7496195-f8e2-4c63-8d1b-0186433b54b7","arxiv_id":"1908.01246","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Random plane partitions with k-periodic weights develop up to k turning points near the vertical boundary, with correlated GUE-corners processes at each point and rational-slope facets between them.","lead":"This math paper derives the exact random patterns at the edge of a stack of cubes when the cube weights repeat with period k. It shows the edge hosts up to k turning points with known GUE-corners processes nearby, separated by new vertical facets of rational slope.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The slice-0 regularization is not provably invisible: γ enters the leading-order action (8), the turning-point heights (16), and the kernel (17), so Theorems 4.2 and 4.6 are proven for a model with a defect at τ=0, not for the claimed periodic weights.","rationale":"The reader's conditional verdict is appropriate, and the reader's weakest assumption is indeed the load-bearing concern. I agree with the reader that the q_0 modification is the least secured step: the paper verifies only k=2 against [Mkr14] and otherwise asserts locality. My stress-test sharpens this: the paper's own formulas show γ entering S_{τ,χ}, χ_j, and the kernel coefficient, so the defect is not visibly negligible at leading order. This is a real mathematical gap, but it is addressable in principle and does not by itself invalidate the computations for the regularized model. I would not escalate to REJECT because the theorems are explicit statements about the regularized measure and can be re-scoped; the paper has independent value in its kernel computations, new frozen facets of rational slope, and the k=2 check. I also do not lower the verdict, because the concern is already captured by the conditional assessment. No authorial criticism is intended: the issue is an unproven and possibly false universality assertion, not a matter of intent.","tokens_in":18228,"tokens_out":22489,"duration_ms":221134,"concrete_test":"Recompute the turning-point height for the same weights α_i and the same V with q_0 multiplied by γ^2 instead of γ in (2), following the steepest-descent computation in Sections 4.0.2–4.0.5. If the resulting χ_j differs by an O(1) amount from (16), the regularization is visible at leading order, falsifying the locality assertion. If it does not change, repeat the computation with the modification placed at a different fixed slice instead of slice 0; any O(1) shift of χ_j or change in the limiting kernel (17) settles the issue.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is the assertion (Section 1, p.4, and again in Section 2) that multiplying q_0 by γ in (2) has no effect on the turning point processes because the turning points are macroscopically far from slice 0. This is not a harmless locality assumption. The factor γ enters the leading-order asymptotic action S_{τ,χ} in (8), through the Φ^- term integrated over M∈(-∞,0). It therefore appears in the critical-point equation, in the turning-point heights χ_j in (16), and in the coefficient S''(z_j) in the kernel (17). Consequently the limiting model described by Theorems 4.2 and 4.6 is the regularized model with a line defect at τ=0, not a periodic-weight model. The only evidence that the defect is invisible is the k=2 comparison with [Mkr14]; for k>2 no such check is supplied, and the explicit γ-dependence of the limit itself makes it doubtful that the turning-point positions are regularization-independent. This matters because the central claim that arbitrary-period weights change the turning-point universality class is about the unmodified periodic weights, whose partition function is infinite and which are therefore not actually defined in the paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies random plane partitions with weights that are periodic of period k in the slice variable, in the scaling limit r→0 with the size of the box growing as V/r. Because the unmodified periodic weights make the partition function infinite, the author multiplies the weight at the 0-th slice by a factor γ (Eq. (2)), producing a model that is periodic except at one slice. Using the determinantal Schur process kernel of Okounkov–Reshetikhin, the paper derives leading asymptotics of the kernel (Eq. (8)), locates turning points near the vertical boundary τ=V (with heights χ_j in Eq. (16)), and computes the limiting correlation kernels in the bulk near the edge (Theorem 4.2) and near the turning points (Theorem 4.6). It then argues that each turning point yields a GUE-corners process on selected vertical slices (Section 4.1) and that the frozen regions between turning points are vertical facets with arbitrary rational slope (Section 4.2). It also analyzes a first-order phase transition at τ=0 in the k=2 case (Theorem 5.1).","tokens_in":18458,"tokens_out":6299,"duration_ms":62872,"significance":"If correct, the main results establish a new family of turning-point universality classes: several GUE-corners processes that are nontrivially correlated, together with frozen facets of arbitrary rational slope. The paper is careful in stating precise theorems with explicit contour-integral kernels, and the k=2 comparison with the author's earlier work [Mkr14] is a useful consistency check. However, the central claims are formally about the regularized model with a defect at slice 0, and the assertion that this defect does not affect the turning points is not proven for k>2. Thus the significance for genuinely periodic weights is conditional, and the paper would benefit from either a proof of locality or a reframing of the claims.","major_comments":[{"comment":"The regularization at slice 0 is not proven to be invisible for the turning-point processes. The paper states (p.4) that multiplying q_0 by γ 'should have no effect' because the turning points are macroscopically far away, and the only evidence is the k=2 comparison with [Mkr14]. However, γ appears explicitly in the leading-order action (8) through the Φ^- term integrated over M<0, in the turning-point heights χ_j in (16), and in the kernel (17) via S''. Since the unmodified periodic weights do not define a probability measure (as shown in Section 2), Theorems 4.2 and 4.6 are theorems about a model with a line defect at τ=0. This is load-bearing because the abstract and introduction claim results for 'periodic weights of arbitrarily high period'. The author should either provide a proof that the turning-point asymptotics are independent of the regularization (for instance, by showing that the γ-dependence cancels to the relevant order), or reframe the theorems and abstract as applying to the regularized model.","section":"1 (p.4), Eq. (2), (8), (16), (17)"},{"comment":"The step 'By setting Sp(zj,±)=0 and solving for χ from (9) we obtain' (Eq. (13)) is too compressed for a result that determines the turning-point heights used in Theorem 4.6 and Section 4.1. In particular, the text says that the O(ε) coefficient g_ε in zj,± cancels, but the reader cannot verify that no √ε contributions are lost without seeing the full expansion. Please expand the derivation or include an appendix with the intermediate algebra.","section":"4.0.2, Eq. (13)"},{"comment":"The claim that the frozen regions between turning points have arbitrary rational slope rests on the assertion 'It can inductively be shown, that after \"passing through\" a turning point, exactly one of the L's in the repeating pattern changes to an R.' This is not proved, and the section is written as a heuristic discussion rather than a formal theorem. Since arbitrary rational slope is prominently advertised in the abstract, this inductive argument should be made rigorous or explicitly labeled as a conjecture/consequence.","section":"4.2"}],"minor_comments":[{"comment":"There are two typos: 'z_{2,-1}' should be 'z_{2,-}' and 'χ_{2,-1}' should be 'χ_{2,-}' (appearing in the inequalities after Eq. (13)).","section":"4.0.2"},{"comment":"'decay exponentially fast as r→∞' should read 'as r→0', since the scaling limit throughout is r→0; as written it is inconsistent with the rest of the paper.","section":"Remark 4.4"},{"comment":"The notation θ(1) with the parenthetical 'means of constant order as ε→0' is nonstandard; consider using Θ(1) or writing explicit bounds instead.","section":"4.0.2"},{"comment":"In the discussion of ill-definedness of the measure, the phrase 'the measure P_bar_q is not well-defined' would be clearer if it read 'the partition function is infinite, so the normalized probability measure does not exist'.","section":"2"},{"comment":"The abstract's claim about 'periodic weights of arbitrarily high period' is misleading until the regularization is discussed; please add a sentence in the abstract noting that a weight modification at one slice is used.","section":"Abstract and Introduction"}],"recommendation":"major_revision","confidential_remarks":"This is a potentially valuable paper, and the k=2 comparison with [Mkr14] is reassuring, but the gap between the regularized model and truly periodic weights is central. If the author can close it (or honestly reframe the claims), the paper would be a strong contribution. The paper's scope is appropriate for the journal, and the novel predictions about multiple GUE-corners processes and rational-slope facets are worth publishing once the regularization issue is addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague—\n\nHere's the short version: this is a genuinely new piece of work on turning points in plane partitions with periodic weights, but the main universality claim rests on an unproven locality assumption. The paper proves its theorems for a regularized model with a modified weight at slice 0, and the modification is not invisible in the limit: γ shows up in the leading-order action (8), in the turning-point heights (16), and in the kernel (17). For k>2 there is no argument that the γ-dependent limit equals what the true periodic model would produce if it existed. The k=2 check against [Mkr14] is encouraging, but it doesn't establish the general claim.\n\nWhat's actually new: arbitrary-period turning point analysis with several correlated GUE-corners processes, vertical facets with rational slopes, and an explicit phase-transition kernel in the k=2 case. The saddle point analysis is standard, but the phenomena are new. The paper is careful with theorem statements, and the derivation of the kernel is mostly explicit. The description of the frozen-region profiles (Section 4.2) is a nice touch.\n\nSoft spots, in order of size. The γ-regularization is the big one. The authors state that the modification 'should have no effect' on turning points because they are macroscopically far from slice 0, but the action integral over M∈(−∞,0) contributes at leading order, so the turning point locations and the local kernel depend on γ. That is a real gap. It doesn't destroy the paper: the theorems are correctly proven for the regularized model. But it means the paper's title and abstract overclaim, and the reader should treat 'periodic weights of arbitrary period' as 'a specific regularized version of periodic weights.' The second soft spot is minor: the phase transition result is proven only for k=2 and V=∞, while the abstract presents it without that qualification. A revised version should either prove γ-independence or clearly frame the results as being about the regularized model.\n\nCitation pattern looks fine; the relevant literature is covered, and the self-citations are for results used directly.\n\nWho it's for: specialists in random plane partitions and dimer models. It deserves a serious referee. I would send it to review, but ask the referee to focus on whether the γ-dependence can be removed or whether the results should be reframed. If the authors do that, the paper becomes a solid contribution. As it stands, I'd give it conditional, not unconditional, acceptance.","headline":"New results on turning points for a regularized periodic-weight model, but the claimed universality for true periodic weights is not established because the regularization is visible in the limit.","tokens_in":18996,"tokens_out":5940,"would_cite":true,"duration_ms":58667,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60K35","82B20","05A17"],"pacs":[],"model":"deepseek-v4-flash","headline":"Periodic weights turn a plane partition's single turning point into several correlated GUE-corners processes.","keywords":["plane partitions","periodic weights","turning points","GUE-corners process","lozenge tilings","frozen facets","Schur process","determinantal point process"],"falsifier":"An independent computation of the $k=3$ or $k=4$ turning-point process in a model with fully periodic weights kept finite by a different boundary modification (for example, a skew-plane-partition cut-out analogous to the $k=2$ construction) would settle the point: if it reproduces the kernel (17) and the rational-slope facets, the locality assumption holds; if it does not, the $\\gamma$ modification is not innocuous and Theorem 4.6 describes a model with a defect at the 0-th slice.","tokens_in":18000,"feed_emoji":"🧊","tokens_out":8273,"duration_ms":76432,"temperature":0.7,"pith_summary":"This paper studies random plane partitions whose volume weights repeat with period k and asks what the system looks like near its right vertical boundary in the scaling limit. The answer it defends is that a single GUE-corners turning point, familiar from the homogeneous weight case, is replaced by up to k turning points, one for each distinct product of consecutive weights; these turning points are separated by vertical facets whose projections can have arbitrary rational slope. The paper proves that the local point process at each turning point is a GUE-corners process on suitably chosen vertical slices, and that the full collection of turning-point processes is a non-trivial interlacing of several GUE-corners processes. A separate claim is that the weight modification used to make the measure finite creates a first-order phase transition along the corner slice, with a limiting point process invariant only in the vertical direction. If correct, this changes the expected universality of boundary fluctuations for determinantal random tilings under periodic weights.","feed_headline":"Periodic weights split one turning point into up to k GUE corners","feed_subtitle":"Each of the k turning points is a GUE-corners process, and between them lie vertical facets of rational slope.","key_machinery":"The engine is the Okounkov-Reshetikhin double contour-integral formula for the Schur process correlation kernel (equation (3)), whose leading exponential behavior is controlled by the action $S_{\\tau,\\chi}(z)$ in (8). With $k$-periodic weights, the products $\\beta_i=\\alpha_{d-1}\\cdots\\alpha_{d-i}$ collapse to $l$ distinct values $\\tilde{\\beta}_1<\\cdots<\\tilde{\\beta}_l$ with multiplicities $m_i$; near $\\tau=V$ the double real critical points of the action sit at $z=\\tilde{\\beta}_i e^V+O(\\sqrt{\\varepsilon})$, and each such critical point is the saddle point of a turning point. The paper regularizes the infinite partition function by multiplying the 0-th slice weight by $\\gamma=\\prod_{\\alpha_j<1}\\alpha_j$ (equation (2)); this factor enters the action through the terms $\\tilde{\\beta}_i\\gamma$ and is responsible for the first-order phase transition at the corner slice. Steepest-descent deformation of the $z$- and $w$-contours through $z_j=\\tilde{\\beta}_j e^V$ produces the limiting kernel (17), whose discrete horizontal and continuous vertical coordinates identify each turning point, slice by slice, with the GUE-corners process.","core_discovery":"The central claim is Theorem 4.6: as $r\\to 0$ near the vertical boundary $\\tau=V$, the system develops $l$ turning points, one for each distinct value $\\tilde{\\beta}_i$ among the products $\\beta_i=\\alpha_{d-1}\\cdots\\alpha_{d-i}$; the $j$-th turning point sits at $(V,\\chi_j)$ with $\\chi_j$ given by (16), and the correlation functions near it are determinants of the kernel (17). Section 4.1 shows that on selected vertical slices this kernel is exactly the GUE-corners kernel: any collection of slices in which the $i$-th slice carries $i$ particles yields the GUE-corners process, so the full turning-point process is several GUE-corners processes interlaced in a non-trivial way. Section 4.2 claims the frozen regions between turning points are vertical facets whose projection has angle $(j-1)\\pi/(2k)$, giving arbitrary rational slopes, and each facet is a deterministic periodic pattern of two lozenge orientations. The paper also claims (Theorem 5.1, computed for $k=2$) that the modified weight at the corner slice produces a first-order phase transition with a limiting point process that is translation invariant in the vertical direction but not in the horizontal direction.","pith_inferences":["Inference: if the locality assumption holds for every $k$, the number and type of turning-point processes is determined entirely by the combinatorial data $(\\tilde{\\beta}_i,m_i)$, so the boundary universality class is indexed by the weight sequence's distinct consecutive products rather than by its period alone.","Inference: a numerical check for $k>2$ — exact sampling or direct kernel evaluation with the regularization placed at different slices far from the right boundary — should leave the turning-point kernel unchanged; agreement would confirm locality, disagreement would reveal that the defect at the 0-th slice reaches the boundary.","Inference: the vertical-facet construction suggests that other dimer models with periodic weights and a Newton polygon carrying several boundary lattice points will exhibit analogous countable families of tilted frozen facets, with the present rational-slope mechanism as a template."],"forward_implications":["For weights of period $k$, the right boundary hosts up to $k$ distinct turning points, one per distinct consecutive product $\\tilde{\\beta}_i$, instead of the single turning point of the homogeneous model.","Each turning point carries a GUE-corners process on any family of vertical slices whose $i$-th slice has $i$ particles, so the boundary process is a non-trivial interlacing of several GUE-corners processes.","The frozen regions between turning points are vertical facets whose projection angle can be any rational multiple $(j-1)\\pi/(2k)$, realized by periodic deterministic patterns of two lozenge orientations.","In the bulk near the edge, the limiting correlation kernel is a product over the distinct $\\tilde{\\beta}_i$'s, recovering the incomplete-beta kernel in the homogeneous case and breaking horizontal translation invariance to translations by $k\\mathbb{Z}\\times\\mathbb{Z}$.","The weight modification at the corner slice induces a first-order phase transition whose limiting point process (computed for $k=2$) is translation invariant vertically but not horizontally."],"supporting_citations":[{"why":"Supplies the determinantal Schur-process correlation kernel and the double contour-integral formula (3) on which all asymptotics are based.","marker":"[OR07]"},{"why":"Introduced the turning-point/GUE-corners conjecture and the steepest-descent technique that the paper adapts to periodic weights.","marker":"[OR06]"},{"why":"The k=2 periodic-weight model whose turning-point process must match the present results, and the skew-plane-partition regularization that motivates the locality check.","marker":"[Mkr14]"},{"why":"Provides Lemma 2.3, used to count real versus complex critical points of the action and to prove frozen/liquid asymptotics.","marker":"[Mkr11]"},{"why":"Studied the periodic Schur process; the bulk process obtained here is stated as a special case of processes from that work.","marker":"[Bor07]"},{"why":"The paper's non-translation-invariant phase-transition process is described as a special case of Gibbs ensembles of nonintersecting paths studied there.","marker":"[BS10]"},{"why":"Connects limit shapes to the complex Burgers equation and explains that integer boundary points of the Newton polygon produce new frozen facets.","marker":"[KO07]"},{"why":"Links dimer models to amoebae and Newton polygons, the mechanism invoked for the rational-slope facets.","marker":"[KOS06]"},{"why":"Classifies fully translation-invariant ergodic Gibbs measures, cited to distinguish the kZ x Z-periodic bulk process from fully translation-invariant ones.","marker":"[She05]"}],"fun_headline_variants":["k GUE-corners processes from periodic weights, with rational facets","Rational-slope facets separate k GUE-corners turning points","Periodic weights create up to k GUE-corner turning points","Multiple interlaced GUE-corner processes from periodic weights"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that multiplying the weight on the corner slice by $\\gamma$ only makes the partition function finite and does not change the turning-point processes, because those points are macroscopically far from that slice; the paper checks this only for $k=2$, and for $k>2$ it is an unproven locality assertion.","fun_headline_variants_meta":{"raw":{"variants":["k GUE-corners processes from periodic weights, with rational facets","Rational-slope facets separate k GUE-corners turning points","Periodic weights create up to k GUE-corner turning points","Multiple interlaced GUE-corner processes from periodic weights"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000831,"raw_usage":{"total_tokens":3634,"prompt_tokens":959,"completion_tokens":2675,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":575,"completion_tokens_details":{"reasoning_tokens":2601}},"tokens_in":575,"tokens_out":2675,"duration_ms":19055,"temperature":1.0,"reasoning_tokens":2601,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:18:48.414242+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An independent computation of the $k=3$ or $k=4$ turning-point process in a model with fully periodic weights kept finite by a different boundary modification (for example, a skew-plane-partition cut-out analogous to the $k=2$ construction) would settle the point: if it reproduces the kernel (17) and the rational-slope facets, the locality assumption holds; if it does not, the $\\gamma$ modification is not innocuous and Theorem 4.6 describes a model with a defect at the 0-th slice.","supporting_citations":[],"review_version":1}