{"id":"e791f9ff-7dd7-4c9e-ab4f-77ac1453ac69","arxiv_id":"2506.16657","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The piecewise linear surface signature is injective: it characterizes surfaces up to translation and thin homotopy, generalizing Chen's path signature theorem.","lead":"This paper proves that a new algebraic invariant, the surface signature, completely distinguishes piecewise linear surfaces up to a natural notion of cancellation called thin homotopy. It answers, for these surfaces, a question posed by Kapranov and gives explicit formulas for computing the invariant.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Injectivity theorem depends on the compatible triangulation lemma (Theorem 6.16 / Lemma D.4); if that lemma fails, the Hurewicz argument collapses.","rationale":"The reader's weakest assumption and my concern agree. The proof of Theorems 1.2 and 6.1 is otherwise coherent: the abelianization (Theorem 4.12, Corollary 4.13), the matching via compactly supported forms (Proposition 6.18), and the kernel comparison (Proposition 6.20) all appear logically sound. The only step that is both indispensable and insufficiently verified is the compatible triangulation theorem. Since a single counterexample to Lemma D.4 would invalidate the main result, acceptance should be conditional on an independent verification of that lemma. I therefore recommend CONDITIONAL rather than a flat ACCEPT, while acknowledging that no flaw was found.","tokens_in":66927,"tokens_out":24674,"duration_ms":253864,"concrete_test":"Implement the line/plane-addition algorithm of Lemma D.4 and apply it to adversarial configurations in R^3: (a) two triangles intersecting along a segment passing through both interiors; (b) an edge crossing the interior of a triangle in the same plane; (c) three triangles pairwise meeting along three concurrent segments; (d) two triangles in the same plane with overlapping interiors. Verify that the output PLSC satisfies Definition 6.8 for every pair of simplices. If any configuration fails, Theorem 6.16 (and hence Theorem 6.1) is false; if all pass, the construction should be formalized or given a detailed proof to close the gap.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim (Theorem 6.1) is proved by reducing an element X ∈ PL_cl^1(V) with trivial signature to an element Y ∈ π2(C) of a compatible PLSC C, showing H(Y) = 0 via integration, and then using ker(H) = ker(W1) from Proposition 6.20. This reduction requires Theorem 6.16: every X admits a representative r(X) whose associated simplicial complex satisfies the intersection property of Definition 6.8. The proof of Theorem 6.16 rests entirely on Lemma D.4, which asserts that any finite collection of edges and convex polygons in V admits a compatible triangulation. This lemma is combinatorial and is proved by a somewhat sketchy line- and plane-addition construction. If D.4 fails—for instance, if some configuration of triangles in three dimensions cannot be refined so that all intersections are common subsimplices—then Lemma 6.10's injectivity of W0 can fail, and the lift Y of Proposition 6.17 need not lie in π2(C). Consequently Proposition 6.18 and the matching argument would not imply H(Y) = 0, and Proposition 6.20 would be inapplicable. I did not find a concrete counterexample, and the construction appears plausible, but the whole injectivity proof is load-bearing on this unverified combinatorial premise.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops an algebraic crossed module PL(V) of piecewise linear surfaces and defines the piecewise linear surface signature S_PL as a natural transformation extending the path signature. The central result is Theorem 6.1: the map S_PL,1 : PL_1(V) -> hat(K)_1(V) is injective, so a piecewise linear surface with trivial signature is thinly homotopic to the constant surface. The proof combines an abelianization theorem for closed surfaces (Theorem 4.12), a reduction to simplicial complexes via compatible triangulations (Theorem 6.16 and Appendix D), and a Hurewicz argument using Whitehead's free crossed module theorem. The paper also proves uniqueness of the signature as a natural transformation, gives an explicit decomposition of the signature into boundary and abelian components (Theorem 5.36), proposes equivalent definitions of thin homotopy for surfaces (Theorem 6.26), and connects thin null homotopy to group homology in Section 7.","tokens_in":67167,"tokens_out":12718,"duration_ms":146638,"significance":"If the main theorem is correct, it answers Kapranov's question in the piecewise linear setting and gives a genuine two-dimensional analogue of Chen's injectivity theorem for path signatures. The proof architecture is strong and transparent: the abelianization step reduces the closed-surface case to vanishing of compactly supported 2-form integrals, the simplicial reduction cleanly builds a PLSC model for any element of PL_1(V), and the final Hurewicz step is well motivated. The paper is also explicit about the external results it imports, such as Chen's path signature injectivity, Whitehead's free crossed module theorem, and continuity results from the rough surfaces literature. The decomposition formula in Theorem 5.36 is a useful contribution in itself, as it gives concrete coordinates for computing surface signatures. The main reservations are two-fold: the compatible triangulation lemma that the injectivity proof rests on is only sketched, and one of the claimed equivalent thin-homotopy conditions in Theorem 6.26 is not equivalent as stated.","major_comments":[{"comment":"The proof of (F2) implies (A2) is incorrect for non-closed 2-forms. For a general compactly supported 2-form omega, the assignment that sends a 2-cycle to the integral of the pullback of omega does not define a map on H_2(C;Z) unless d omega = 0. Since condition (A2) requires vanishing of integrals of all compactly supported 2-forms, the hypothesis H_2(f)([S^2]) = 0 is insufficient. A concrete test is any embedded sphere X in R^3 and a compactly supported bump 2-form nu with integral 1 over X; such a nu cannot be closed, and the homology condition cannot force its integral to vanish. Thus the equivalence in Theorem 6.26 as stated is false. The factorization condition needs to be strengthened, for example by requiring vanishing of the induced pairing with all compactly supported 2-forms, or Theorem 6.26 must be modified.","section":"Theorem 6.26, implication (F2) implies (A2)"},{"comment":"The existence of compatible representatives is the single combinatorial premise on which Proposition 6.17, Proposition 6.18, and therefore Theorem 6.1 rest. The proof of Lemma D.4 is only a sketch: after adding all lines and planes, it asserts that the arrangement decomposes into convex polygons whose vertices lie in the constructed set C_0, but it does not fully prove that the union of the chosen triangulations is a simplicial complex with common-subsimplex intersections in the presence of non-transverse configurations, such as edges lying in polygon planes, overlapping edges, or multiple planes meeting along a common line. Likewise, Lemma D.5 assumes a factorization in the fundamental group of the 1-skeleton and does not spell out how the marked tail paths in FMon(V) are chosen for arbitrary spanning-tree paths. Since a failure of D.4 would invalidate the central injectivity theorem, these constructions require a complete proof or a precise reference.","section":"Theorem 6.16 and Appendix D, Lemma D.4"},{"comment":"The construction of the simply connected complex bC is load-bearing for Proposition 6.20, but as written it is not fully justified. The point x is chosen only so that each triple {x, p_i, p_j} is non-collinear; this does not prevent the new triangles [x, p_i, p_j] from intersecting the existing 2-simplices of C outside the 1-skeleton C_1. In that case the union C union D is not a PLSC in the sense of Definition 6.6, and van Kampen's theorem cannot be applied to the intended decomposition. This can likely be repaired by choosing x in general position away from finitely many planes, but the argument needs to be stated and proved.","section":"Lemma 6.19"}],"minor_comments":[{"comment":"The notation 'C := T(E, V)' should read 'C := T(E, P)', since the second argument is the collection of polygons P, not the vector space V.","section":"Proof of Theorem 6.16"},{"comment":"The sentence 'If two planes in H intersect at a point' is imprecise in higher-dimensional V; two distinct affine planes may intersect in a point, a line, or be disjoint, and the construction should specify which intersections are being added to C_0.","section":"Lemma D.4, Step 3"},{"comment":"The definition of a smooth piecewise linear surface should explicitly require that the affine maps f_sigma agree on common edges of adjacent simplices; this is implicit but should be stated for rigor.","section":"Definition 6.22"},{"comment":"The phrase 'after modifying the boundary using a thin homotopy' is vague; it should be specified how the boundary modification interacts with the factorization X : [0,1]^2 -> S^2 -> C -> V, since condition (F2) already assumes trivial boundary.","section":"Theorem 6.26, condition (F2)"},{"comment":"In the coordinate-free formula S^Gamma_1(X) = sum_{k >= 0} (1/k!) integral_{C(X)} (id_V)^k, the placement of the tensor and symmetric products should be clarified, since (id_V)^k is being viewed as an element of C^infinity(V) tensor S^k(V).","section":"Remark 5.37"}],"recommendation":"major_revision","confidential_remarks":"The main injectivity theorem is well structured and appears likely to be correct, but it depends on a combinatorial triangulation lemma whose proof is not complete, and the same chain of reasoning uses a simply connected extension whose construction needs a genericity argument. In addition, the claimed equivalence in Theorem 6.26 between the factorization condition and the analytic condition is false as stated. These issues are local and probably fixable, but they are load-bearing for the advertised equivalent-characterization result. I therefore recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The one thing you should know: this paper looks like the real thing. It proves injectivity of the PL surface signature (Theorem 6.1), thereby answering Kapranov's question in the PL setting, and the proof structure is credible. The second thing: the load-bearing spot is the compatible triangulation lemma (Theorem 6.16, built on Lemma D.4), which is plausible but handed a bit quickly; that's where I'd focus referee attention.\n\nWhat's new: the crossed module PL(V) with its universal property, the abelianization theorem (4.12) that identifies the closed-surface signature with integration of polynomial 2-forms, and the simplicial reduction that lets the authors use Whitehead's free crossed module and Hurewicz. The decomposition results in Section 5.4—boundary and abelian parts—are genuinely useful for computation and clarify what information the signature carries. The final section connecting thin null homotopy to H3(G) is a nice bonus.\n\nWhat it does well: the paper is carefully organized, definitions are justified, and the main proof is broken into clear reductions: Lemma 6.5 reduces to closed surfaces; Proposition 6.17 lifts to the fundamental crossed module; Proposition 6.18 uses the signature to get a matching; Proposition 6.20 exploits Hurewicz after killing pi_1. I didn't find a circular dependency or a disguised assumption. The imported black boxes—Chen's path signature injectivity, Whitehead's theorem, Hurewicz, continuity from the rough-surfaces literature—are independent and legitimate.\n\nThe soft spots are minor relative to the central claim. First, as noted, the compatible triangulation lemma D.4 is the least expandable step. The construction (arrangement of lines/planes, then triangulate each polygon using boundary vertices) is standard and I don't see a counterexample, but termination and the 'no new bad vertices' claim deserve a few more sentences. Second, Theorem 5.21's uniqueness of the smooth signature uses a density/continuity argument that is fine but relies on results from [38]; not a problem, just external. Third, the paper only treats PL surfaces; the smooth case remains open, which they say explicitly.\n\nWho should read this: algebraic topologists interested in crossed modules and 2-dimensional holonomy, and the rough-paths/signature community. It deserves a serious peer review. My recommendation: send it out; ask the referee to verify Lemma D.4 carefully, but I'd be surprised if it fails. If I were handling it, I'd accept after minor expansion of that lemma.","headline":"A strong, coherent proof of PL surface-signature injectivity that deserves a serious referee; main check is the compatible triangulation lemma.","tokens_in":67701,"tokens_out":3097,"would_cite":true,"duration_ms":33340,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The surface signature is a complete invariant for piecewise linear surfaces up to translation and thin homotopy.","keywords":["surface signature","thin homotopy","crossed module","piecewise linear surface","surface holonomy","2-connection","iterated integrals","group homology"],"falsifier":"A reader could falsify injectivity by finding a closed piecewise linear surface $X$ in $\\mathbb{R}^3$ whose signature, computed by the explicit formula of Theorem 5.36, is zero while the corresponding word in $PL_1(V)$ cannot be reduced to the empty word by fold cancellations and the Peiffer identity. Concretely, enumerate triangulated representatives whose 2-simplices cancel in homology and test whether any such class is nonzero in $PL_1(V)$; the theorem predicts none exists.","tokens_in":66714,"feed_emoji":"📐","tokens_out":10921,"duration_ms":116077,"temperature":0.7,"pith_summary":"This paper proves that the surface signature—a formal power series of tensors obtained as surface holonomy of a universal translation-invariant 2-connection—uniquely determines a piecewise linear surface up to translation and thin homotopy (a homotopy that sweeps out no volume, the 2-dimensional analogue of tree-like equivalence for paths). The main theorem states that if the signature of a piecewise linear surface is trivial, then the surface is thinly homotopic to a constant surface. The proof first shows that a closed surface with trivial signature integrates all compactly supported 2-forms to zero, then realizes the surface as an element of a compatible simplicial complex and uses algebraic homotopy theory to collapse it. If correct, the result gives a complete invariant for piecewise linear surfaces, generalizing the classical path-signature theorem and answering a question posed in the literature. It also supplies an explicit decomposition of the signature into a boundary component and a series of surface integrals, which makes the invariant computationally accessible.","feed_headline":"Equal signature forces thin homotopy for PL surfaces","feed_subtitle":"Extends the classical path-signature completeness theorem to two dimensions, with explicit formulas for computing the invariant.","key_machinery":"The load-bearing object is the crossed module $PL(V) = (\\delta: PL_1(V) \\to PL_0(V), \\triangleright)$, whose group $PL_1(V)$ is generated by kites $(w,b)$: a piecewise linear tail path $w$ carrying a planar loop $b$. The boundary $\\delta(w,b)=w b w^{-1}$ records the overall loop, and quotienting by fold relations and the Peiffer identity encodes local and non-local cancellations of surfaces. The argument also relies on two auxiliary mechanisms: a gauge transformation that abelianizes the universal 2-connection so that the signature of a closed surface equals its integrals against all polynomial 2-forms, and the existence of compatible triangulated representatives, which lets every element of $PL_1(V)$ be studied through a simplicial complex whose intersections are common faces.","core_discovery":"On the paper's own terms, the central discovery is the injectivity of the piecewise linear surface signature $S_{PL,1}: PL_1(V) \\to \\hat{K}_1(V)$ between the crossed module of piecewise linear surfaces and the completed free crossed module of formal surfaces. Equivalently, Theorem 1.2 says that if $S_1(X)=0$ for a piecewise linear surface $X$, then $X$ is thinly homotopic to the constant surface. The injectivity is proved by lifting a representative of $X$ to the fundamental crossed module of a compatible triangulation, using the classical description of second relative homotopy groups as free crossed modules, showing via the abelianized curvature that the homology class of the lifted surface vanishes, and then killing the fundamental group of the ambient complex to apply the standard comparison between $\\pi_2$ and $H_2$. A corollary is that the algebraic relations defining $PL_1(V)$—local fold cancellations plus the Peiffer identity—account for every thin homotopy among piecewise linear surfaces.","pith_inferences":["If the piecewise linear injectivity can be transported to smooth surfaces by approximation (the direction the paper points to as future work), the smooth surface signature would inherit completeness, making the two-dimensional signature as discriminating as the path signature.","The abelian component of the signature suggests a practical numerical feature for two-dimensional data: truncate the formal symmetric-power series and evaluate the monomial 2-form integrals over the coned surface, bypassing the holonomy differential equation.","The group-homology classification suggests a converse route: every class in $H_3(G)$ should be realizable by a thinly null piecewise linear surface over a complex with fundamental group $G$, so group homology could be searched by enumerating surfaces rather than classifying spaces.","The equality between the kernel of the algebraic realization map and the kernel of the homology comparison suggests a decision procedure for thin null-homotopy: build a compatible triangulation and test reduction modulo folds and the Peiffer identity, without computing the full signature."],"forward_implications":["Two piecewise linear surfaces with equal surface signature are thinly homotopic after translation, so the signature is a complete invariant for the classification problem.","The equivalence conditions in Theorem 6.26 show that word reduction, holonomy triviality, rank-one/rank-two homotopies, image containment, the vanishing of all compactly supported 2-form integrals, and signature triviality all characterize the same thin-null surfaces.","The signature splits into the path signature of the boundary and the integrals of monomial 2-forms over the surface obtained by coning off the boundary, giving explicit canonical coordinates for computation.","Non-local thin cancellations are not arbitrary: in the embedded case they are governed by the fundamental group of the image and by its group homology $H_3(G)$, giving a geometric interpretation of that homology.","The realization map from $PL_1(V)$ into thin homotopy classes of smooth surfaces is injective, so the algebraic crossed module faithfully describes genuine thin homotopy of piecewise linear surfaces."],"supporting_citations":[{"why":"Supplies the image condition for thin null-homotopy of paths, used to prove the fundamental-group map of a compatible complex is injective.","marker":"[3]"},{"why":"Provides the theory of crossed modules and free crossed modules, including the double-groupoid equivalence and pullback constructions used to build PL(V).","marker":"[8]"},{"why":"Introduced thin homotopy of surfaces, the equivalence relation whose classification is the paper's main goal.","marker":"[12]"},{"why":"Establishes the path-signature injectivity theorem that the paper generalizes and uses to reduce to closed surfaces.","marker":"[17]"},{"why":"Fixes the tree-like/analytic characterization of thin equivalence for paths, one of the definitions generalized to surfaces.","marker":"[32]"},{"why":"Introduces the surface signature as holonomy of a universal translation-invariant 2-connection and poses the uniqueness question answered here.","marker":"[34]"},{"why":"Develops the multiplicative surface signature in a rough-path setting and provides the continuity/naturality properties used in the smooth decomposition.","marker":"[38]"},{"why":"Proves surface holonomy is a crossed-module morphism invariant under thin homotopy, which makes the signature well-defined.","marker":"[44]"},{"why":"Gives equivalent definitions of thin homotopy for C1 paths, the template for several two-dimensional conditions proved equivalent here.","marker":"[52]"},{"why":"The theorem that the fundamental crossed module of a 2-dimensional complex is free; it supplies the map from relative homotopy groups into PL(V).","marker":"[55]"}],"fun_headline_variants":["Surface signature characterizes PL surfaces up to thin homotopy","2D signature: thin homotopy is the only ambiguity for PL surfaces","Answering Kapranov: surface signature determines thin homotopy type","Surface signature generalizes Chen's path result to 2D"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof requires Theorem 6.16: every element of $PL_1(V)$ has a compatible triangulated representative, meaning its simplices intersect only in common faces; if that triangulation construction fails, the reduction to a simplicial complex and the homology argument collapse.","fun_headline_variants_meta":{"raw":{"variants":["Surface signature characterizes PL surfaces up to thin homotopy","2D signature: thin homotopy is the only ambiguity for PL surfaces","Answering Kapranov: surface signature determines thin homotopy type","Surface signature generalizes Chen's path result to 2D"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000829,"raw_usage":{"total_tokens":3592,"prompt_tokens":885,"completion_tokens":2707,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":501,"completion_tokens_details":{"reasoning_tokens":2632}},"tokens_in":501,"tokens_out":2707,"duration_ms":17850,"temperature":1.0,"reasoning_tokens":2632,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:22:16.105677+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A reader could falsify injectivity by finding a closed piecewise linear surface $X$ in $\\mathbb{R}^3$ whose signature, computed by the explicit formula of Theorem 5.36, is zero while the corresponding word in $PL_1(V)$ cannot be reduced to the empty word by fold cancellations and the Peiffer identity. Concretely, enumerate triangulated representatives whose 2-simplices cancel in homology and test whether any such class is nonzero in $PL_1(V)$; the theorem predicts none exists.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the image condition for thin null-homotopy of paths, used to prove the fundamental-group map of a compatible complex is injective."},{"cited_title":"Brown, P .J","cited_arxiv_id":null,"evidence_quote":"Provides the theory of crossed modules and free crossed modules, including the double-groupoid equivalence and pullback constructions used to build PL(V)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced thin homotopy of surfaces, the equivalence relation whose classification is the paper's main goal."},{"cited_title":"Surface holonomy for non-abelian 2-bundles via double groupoids","cited_arxiv_id":null,"evidence_quote":"Proves surface holonomy is a crossed-module morphism invariant under thin homotopy, which makes the signature well-defined."},{"cited_title":"On the holonomic equivalence of two curves","cited_arxiv_id":null,"evidence_quote":"Gives equivalent definitions of thin homotopy for C1 paths, the template for several two-dimensional conditions proved equivalent here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The theorem that the fundamental crossed module of a 2-dimensional complex is free; it supplies the map from relative homotopy groups into PL(V)."}],"review_version":2}