{"id":"786a1d29-fbf9-4552-9c48-aff3f02ea333","arxiv_id":"2607.10636","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.5,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Under L1 perimeter the isoperimetric double-tiling profile is 2√x + 2√(1-x), uniquely realized by the Pythagorean tiling of two axis-aligned squares except at equal area.","lead":"The paper finds the cheapest way to tile the plane with two repeating cells under Manhattan (L1) length cost, for rectangular lattices and then for all lattices. The global optimum is two axis-aligned squares that meet at a corner (the Pythagorean tiling), unique except when the areas are equal.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"No significant objection identified","rationale":"The reader correctly notes that the density-estimate and connectedness arguments are only sketched by reference to the isotropic literature and that their constants depend on the lattice. Those statements are load-bearing for the full classification of rectangular-lattice minimizers (Theorem 1.1) and for the local-isoperimetric partitions of Section 6. They are not, however, used in the proof of the paper’s strongest claim (Theorem 1.2). That proof is a direct application of the classical Wulff inequality plus a short geometric argument that forces the lattice generators once both cells are known to be squares. Because the strongest claim therefore stands independently of the flagged regularity gap, the reader’s ACCEPT verdict with high confidence remains appropriate; no adjustment is required. The partial agreement simply records that the regularity concern is real for other parts of the manuscript but does not undermine the central formula or its uniqueness statement.","tokens_in":20807,"tokens_out":545,"duration_ms":6226,"concrete_test":"Independently re-derive the uniqueness paragraph of the proof of Theorem 1.2 (page 16) without any appeal to Corollary 3.6 or Lemma 3.7: start from equality in (2.2), conclude that both generators are axis-aligned squares, then verify that the only lattice vectors compatible with a tiling are those of the Pythagorean configuration (or the equal-area strip configurations). If the argument closes without those lemmas, the concern is confirmed to be non-load-bearing for the strongest claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of Theorem 1.2 (I_ℓ₁(x)=2√x+2√(1-x) with uniqueness of Pythagorean double tilings for x\neq1/2) rests on the Wulff inequality (2.2) plus an explicit competitor, followed by a short geometric rigidity argument that forces the generators to be axis-aligned squares whose lattice vectors must be of the form (a,h),(w,a) with w=h=√x. The reader’s flagged weakest assumption (density estimates and connectedness via Corollary 3.6 / Lemma 3.7) is used only for the rectangular-lattice classification (Theorem 1.1) and for the local-minimality statements of Section 6; it is not invoked in the proof of Theorem 1.2. The uniqueness case distinction in that proof is elementary and does not rely on the sketched density constants. Consequently the strongest claim is internally secure.","agreement_with_reader":"partial"},"referee_report":{"model":"grok-4.5","summary":"The paper studies periodic double tilings of the plane that minimize the anisotropic ℓ₁-perimeter. For a fixed rectangular lattice G of area 1, Theorem 1.1 gives an explicit formula for the (G,ℓ₁)-isoperimetric profile IG,ℓ₁(x) and classifies all minimizers: a square plus chipped rectangle when one cell is small, and two adjacent rectangles otherwise. Minimizing further over all planar lattices, Theorem 1.2 shows that Iℓ₁(x)=2√x+2√(1-x), attained uniquely (for x\neq1/2) by the Pythagorean double tiling of two axis-aligned squares sharing a vertex; when x=1/2 the minimizers are alternating strips of equal squares. Theorem 1.4 then proves that certain limiting non-periodic partitions obtained by sending one volume to infinity are locally ℓ₁-isoperimetric. The arguments combine the classical Wulff inequality for the ℓ₁-norm with elementary projection estimates and a finite case analysis on how horizontal and vertical lines meet the generators.","tokens_in":21019,"tokens_out":812,"duration_ms":10175,"significance":"The work supplies a complete, explicit solution of the anisotropic double-tiling problem for the Manhattan perimeter, both for rectangular lattices and after free minimization over all lattices. The profile formula of Theorem 1.2 is parameter-free and is realized by a classical geometric construction (the Pythagorean tiling), while uniqueness follows from the rigidity of the Wulff shape. The classification for rectangular lattices and the local-minimality statements for infinite-volume partitions extend the authors’ earlier isotropic results and fit naturally into the ongoing program on anisotropic clusters and periodic partitions. The proofs are elementary once the Wulff inequality and the projection lemmas are in place, and the geometric models are fully explicit.","major_comments":[],"minor_comments":[{"comment":"In the proof of Theorem 3.5 the density constants are obtained by covering a fundamental domain with finitely many balls and taking the worst-case constants from Theorem 3.4; a one-sentence remark that the resulting c,ρ₀ depend only on G and φ (and not on the particular tiling) would make the uniformity claim fully transparent.","section":null},{"comment":"Lemma 3.7 asserts connectedness of E₂ whenever R^{2}\\(Ē₁+G) is connected. The argument is short and correct, but a brief parenthetical note that the same conclusion holds after replacing E₁ by any finite union of its G-translates would clarify why the construction (3.3) preserves the property.","section":null},{"comment":"Figure 4 and the accompanying text in the introduction describe a continuum of energetically equivalent placements of the small square; it would help the reader if the precise range of admissible (h,l) were stated once in the body of Theorem 1.1 rather than only in the caption.","section":null},{"comment":"A few typographical inconsistencies appear (e.g., “prescrived” on p. 15, “horizonal” on p. 15, and the occasional missing space before parentheses). A light copy-edit pass would remove them.","section":null}],"recommendation":"accept","confidential_remarks":"The manuscript is a clean, self-contained contribution that sits comfortably within the journal’s scope. The only soft spot flagged by the reader (density estimates) is used solely for the rectangular-lattice classification and the local-minimality statements; it is not needed for the strongest claim (Theorem 1.2). I see no reason to delay acceptance."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The headline result is Theorem 1.2: the L1-isoperimetric profile over all planar lattices is exactly I(x)=2√x+2√(1-x), attained uniquely (when x≠1/2) by the Pythagorean double tiling of two axis-aligned squares that share a vertex. That formula is immediate from the classical Wulff inequality plus the explicit competitor, and the uniqueness argument is a short geometric rigidity check that forces the generators to be squares whose lattice vectors must satisfy w=h=√x. The stress-test note is right: this part does not lean on the density estimates that the reader flagged.\n\nWhat the paper does well is the rectangular-lattice analysis (Theorem 1.1). They get an explicit piecewise formula for IG,ℓ1(x) and a complete classification: square-plus-chipped-rectangle when one area is small, two adjacent rectangles otherwise. The lower bounds come from elementary projection estimates (Lemmas 4.1–4.5) plus a finite case distinction on how horizontal and vertical lines meet the generators (Theorem 5.1). Equality cases are characterized cleanly. The local-minimality statements in Section 6 for the infinite-volume limits (including the vortex) are a natural and useful add-on.\n\nThe soft spots are minor and standard for the field. Existence of essentially bounded open representatives and the connectedness lemma (Corollary 3.6, Lemma 3.7) rest on density estimates whose constants depend on the lattice and whose details are only sketched by reference to the isotropic literature. Those tools are used for the rectangular classification and the local-minimality claims, not for the global profile. Self-citations to the authors’ earlier isotropic work (NNP26) are appropriate; the L1 case is genuinely more rigid and cleaner.\n\nThis is for people working on anisotropic isoperimetric problems, crystalline surface energies, or periodic partitions. The math is solid, the constructions are explicit, and the citation pattern is honest. I would send it to a serious referee without hesitation; it is a clean, self-contained extension of their program and deserves to be published.","headline":"Clean L1 double-tiling classification: Pythagorean squares give the global profile, uniqueness holds for unequal areas, and the rectangular case is fully explicit.","tokens_in":21595,"tokens_out":536,"would_cite":true,"duration_ms":6713,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["49Q05","52C20","58E12"],"pacs":[],"model":"grok-4.5","headline":"The cheapest way to tile the plane with two repeating cells under ℓ1 length is the Pythagorean arrangement of two axis-aligned squares that share a vertex.","keywords":["tilings","anisotropic perimeter","isoperimetric profile","ℓ1-perimeter","Pythagorean tiling","periodic partitions","Wulff inequality"],"falsifier":"Exhibit any lattice of area 1 and any pair of generators of areas x and 1-x whose total ℓ1-perimeter is strictly smaller than 2√x + 2√(1-x), or show that a non-Pythagorean configuration saturates the bound for some x ≠ 1/2.","tokens_in":21726,"feed_emoji":"⬛","tokens_out":951,"duration_ms":12737,"temperature":0.7,"pith_summary":"The paper asks how to partition the plane into two families of identical cells so that the total length of interfaces, measured with the Manhattan (ℓ1) norm, is as small as possible for given cell areas. When the cells must repeat according to a fixed rectangular lattice, the optimal shapes switch from a square sitting inside a chipped rectangle (when one area is small) to a pair of adjacent rectangles (when the areas are comparable). Once every possible lattice is allowed, the global minimum is always achieved by two axis-aligned squares that share a single vertex—the classical Pythagorean tiling—and this arrangement is unique except when the two areas are equal. The same constructions remain locally optimal even after one cell area is sent to infinity, producing infinite-area “vortex” or “cross” partitions that cannot be improved inside any large box. The result supplies an explicit anisotropic counterpart to classical honeycomb and double-bubble theorems and shows that the ℓ1 geometry forces far more rigidity than the Euclidean perimeter.","feed_headline":"Two squares sharing a vertex beat every other double tiling","feed_subtitle":"Under Manhattan length the Pythagorean packing is the unique cheapest way to tile the plane with two cell sizes","key_machinery":"The (G,ℓ1)-isoperimetric profile IG,ℓ1(x) together with its global envelope Iℓ1(x); both are evaluated by combining the Wulff inequality for the ℓ1-perimeter with exhaustive case analysis on which generators are crossed by horizontal and vertical lines.","core_discovery":"For every area fraction x the global ℓ1-isoperimetric profile equals 2√x + 2√(1-x) and is realized uniquely (except at x = 1/2) by the Pythagorean double tiling of two axis-aligned squares sharing a vertex; for a fixed rectangular lattice the profile is piecewise explicit and the minimizers are completely classified as either a square-plus-chipped-rectangle or a pair of adjacent rectangles.","pith_inferences":["The same Wulff-plus-line-crossing method should classify triple or higher anisotropic tilings once the appropriate combinatorial cases are enumerated.","Because ℓ1 forces axis-alignment, the uniqueness statements are stronger than their Euclidean counterparts and may extend to other crystalline norms whose Wulff shapes are polygons.","Local minimality of the infinite partitions suggests they could serve as building blocks for free-boundary anisotropic cluster problems with mixed finite and infinite chambers."],"forward_implications":["Any periodic double tiling that is not Pythagorean (or a rectangular pair when the lattice is fixed) can be improved by a pure lattice change or a shape change.","The infinite-area “vortex” and “cross” partitions obtained by sending one volume to infinity are locally ℓ1-minimizing, so they cannot be bettered inside any finite window.","When the two areas are equal the minimizers form infinite families of striped square packings, all sharing the same perimeter cost.","The squared excess (IG,ℓ1(x) - IG,ℓ1(0))² is concave, giving a quantitative concavity statement for rectangular lattices."],"fun_headline_variants":["Two squares sharing a vertex uniquely minimize ℓ1 double tilings","Pythagorean packing of squares yields the global ℓ1 isoperimetric profile","Axis-aligned squares beat every other lattice for two-cell ℓ1 partitions","ℓ1 profile equals 2√x+2√(1-x) via unique Pythagorean double tiling","Square-plus-chipped-rectangle or adjacent rectangles for fixed lattices"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"The classification rests on the claim that every minimizer admits open, essentially bounded representatives whose topological boundaries coincide with their reduced boundaries, and that connectedness of one cell forces connectedness of the other.","fun_headline_variants_meta":{"raw":{"variants":["Two squares sharing a vertex uniquely minimize ℓ1 double tilings","Pythagorean packing of squares yields the global ℓ1 isoperimetric profile","Axis-aligned squares beat every other lattice for two-cell ℓ1 partitions","ℓ1 profile equals 2√x+2√(1-x) via unique Pythagorean double tiling","Square-plus-chipped-rectangle or adjacent rectangles for fixed lattices"]},"model":"grok-4.5","effort":"low","cost_usd":0.003322,"raw_usage":{"total_tokens":1087,"prompt_tokens":704,"num_sources_used":0,"completion_tokens":106,"cost_in_usd_ticks":33220000,"prompt_tokens_details":{"text_tokens":704,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":277,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":704,"tokens_out":106,"duration_ms":3600,"temperature":1.0,"reasoning_tokens":277,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-14T10:17:37.508625+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Exhibit any lattice of area 1 and any pair of generators of areas x and 1-x whose total ℓ1-perimeter is strictly smaller than 2√x + 2√(1-x), or show that a non-Pythagorean configuration saturates the bound for some x ≠ 1/2.","supporting_citations":[],"review_version":1}