{"id":"bb47f03f-4b0a-4c03-90ad-a7ab7dae119e","arxiv_id":"2504.18996","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For ghost-free pairs of generalised tree modules, homomorphisms are spanned by combinatorially defined generalised graph maps, yielding a sufficient condition for indecomposability.","lead":"This paper defines a broader class of modules called generalised tree modules and, under a ghost-free condition, gives a combinatorial description of all homomorphisms between two of them. A specialist might read it for new criteria for indecomposability of modules over quiver algebras, with an application to Dynkin type D quivers.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The unproved Proposition 3.6 is the load-bearing hinge: if complete R[2]-free subnetworks can violate exactly-one local choices, generalised graph maps may not be homomorphisms and Theorem A is not established.","rationale":"The reader correctly identifies both the ghost-free hypothesis and the unproved Proposition 3.6 as weak points. I focus on Proposition 3.6 because it is an internal, unproved assertion on which the definition of a generalised graph map and Proposition 3.9 directly depend: without it, the maps whose linear combinations Theorem A claims to exist are not known to be homomorphisms. The ghost-free hypothesis is an explicit condition of the theorem; whether it can be removed is an open question, so it limits the theorem's scope but does not by itself threaten the conditional statement. Proposition 4.5 and Lemma 5.7 are also terse, especially the completeness claim in Lemma 5.7, but they cannot be assessed independently of Proposition 3.6, since the objects they output are complete R[2]-free subnetworks whose homomorphism property is supplied by Proposition 3.9 via Proposition 3.6. The recommended verdict remains CONDITIONAL, which is the same as the reader's verdict; hence no change is proposed. The concern is substantive enough to justify a request for a complete proof of Proposition 3.6 or a machine-checked verification, but not enough to reject the paper, because no counterexample to the proposition has been exhibited in the text.","tokens_in":25538,"tokens_out":11957,"duration_ms":125774,"concrete_test":"Independently re-derive Proposition 3.6 from Definitions 3.2 and 3.5. Concretely: take a small 2-covering network with one R[2]-system, such as the hexagon arising in Example 3.1 or Example 3.13, enumerate all complete subnetworks and all R[2]-free subnetworks, and check whether any complete R[2]-free subnetwork has a vertex with two arrows satisfying Condition (1a) or Condition (2a), or with both an arrow and an edge. If such a subnetwork exists, compute H_M(alpha * v_n) and alpha * H_M(v_n) for the relevant arrow alpha; if these differ, Proposition 3.6 is false and Theorem A's generating set may contain non-homomorphisms. If every complete R[2]-free subnetwork in the enumeration satisfies exactly-one local conditions, the hinge holds and the remaining concern is the open scope of the ghost-free assumption.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem A is proved by induction on support: Proposition 5.5 shows that the support of a homomorphism is complete; Proposition 4.5 carves a complete R[2]-free subnetwork with the same vertex set; Proposition 3.9 converts such subnetworks into homomorphisms; and Lemma 5.7 extracts a generalised graph map. The only step stated without proof is Proposition 3.6, which asserts that a complete subnetwork is R[2]-free if and only if, at every vertex, exactly one of Conditions (1a)/(1b) and exactly one of Conditions (2a)/(2b) holds, with at most one arrow and at most one edge. Proposition 3.9's proof explicitly invokes Proposition 3.6 to obtain the uniqueness of arrows and edges, and Definition 3.10 defines generalised graph maps as complete R[2]-free subnetworks. If Proposition 3.6 is false, a complete R[2]-free subnetwork could have two incoming arrows satisfying Condition (1a) at a vertex; then the coefficient-counting argument in Proposition 3.9 can break, so the maps H_G produced by Theorem A need not be homomorphisms. The carving arguments in Proposition 4.5 and Lemma 5.7 are underdetailed, but they do not bypass Proposition 3.6: they produce complete R[2]-free subnetworks, and only Proposition 3.6 guarantees that these objects have the uniqueness properties that make Proposition 3.9 valid. The ghost-free hypothesis is an explicit condition whose necessity is left open in Question 5.8, so it is a scope limitation rather than an internal gap; the unproved Proposition 3.6 is more directly load-bearing for the correctness of the central construction.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies generalised tree modules over zero-relation algebras, obtained by dropping the injectivity condition in Crawley-Boevey's definition of tree module. The main result (Theorem A) claims that, when char(K) differs from 2 and the pair (M1,M2) is 'ghost-free', the space Hom(M1,M2) is spanned by homomorphisms attached to explicit finite combinatorial objects called generalised graph maps, defined as non-empty connected involution-free complete R[2]-free subnetworks of the 2-covering network N[2]. The proof proceeds by induction on the support of a homomorphism: Proposition 5.5 shows the support is complete, Proposition 4.5 carves out a complete R[2]-free subnetwork with the same vertices, Proposition 3.9 turns such subnetworks into homomorphisms, and Lemma 5.7 extracts a generalised graph map under the ghost-free hypothesis. Theorem B gives a sufficient condition for indecomposability of a generalised tree module, and Section 7 applies the results to construct indecomposable generalised tree modules for Dynkin quivers of type D, complementing Ringel's theorem that exceptional modules are generalised tree modules.","tokens_in":1513,"tokens_out":1949,"duration_ms":153423,"significance":"If the main theorem is correct, it is a meaningful extension of Crawley-Boevey's graph-map basis to a setting where basis elements may collide, giving a finite explicit generating set for Hom-spaces between generalised tree modules. The paper also provides a checkable indecomposability criterion and an explicit construction for type D Dynkin quivers, together with instructive examples of sign-flip homomorphisms and ghosts. The authors are transparent about the ghost-free scope and pose the natural open question whether it can be removed. However, the proof as written is not complete: Proposition 3.6 is stated without proof and is load-bearing for Proposition 3.9, while the carving arguments in Proposition 4.5 and Lemma 5.7 are underdetailed. The result is plausible, but the manuscript needs substantial revision before the central claim is established.","major_comments":[{"comment":"Proposition 3.6 is stated with the words 'we mention without proof', yet it is used essentially in the proof of Proposition 3.9. Specifically, it supplies the uniqueness of arrows and edges needed for the partition into I2 and J2, for the existence and uniqueness of the arrow in Condition (2a), and for the injectivity of the maps F and F' that yield the equality of the sums over I3 and J3. The asserted equivalence between R[2]-freeness and the local exactly-one conditions is not a formal consequence of the definitions as far as the text shows; one must prove, for example, that two arrows satisfying Condition (1a) for the same incoming arrow of T1 would produce a traversal in R[2], and conversely that absence of all R[2]-traversals forces the exactly-one conditions at every vertex. A complete proof, or a precise reference, is required before Theorem A can be considered established.","section":"Section 3, Proposition 3.6"},{"comment":"The proof of Proposition 4.5 does not, as written, establish the claim. The construction partitions the vertices of each R[2]-system and chooses perfect matchings locally, but the text never verifies that the union of these choices is R[2]-free when hexagons overlap in antipodal vertices, a situation explicitly allowed by Remark 4.4. The completeness of the resulting subnetwork is also asserted rather than proved: the final sentence claims that for every removed link alpha there are links beta and gamma in M' with beta alpha and alpha gamma in R[2], but no construction or argument is supplied. Since Corollary 5.6 and hence the induction in Theorem A depend directly on this proposition, this is a load-bearing gap.","section":"Section 4, Proposition 4.5"},{"comment":"The proof of Lemma 5.7 is underdetailed at the two points that matter for the induction in Theorem A. First, the modification of M to M' is described by cases, but the argument does not show that the modified links do not create new blocked traversals together with links from adjacent R[2]-systems. Second, the completeness of G is asserted by saying that any link between two vertices of G0 present in M is already present in G; this is not immediate because M' was explicitly allowed to delete links from M, and a deleted link between two vertices of G0 would not be present in G. A direct verification of the completeness conditions at each vertex of G for every arrow of T1 and T2 is needed. As the inductive step of Theorem A, this lemma is load-bearing.","section":"Section 5, Lemma 5.7"}],"minor_comments":[{"comment":"In Condition (2b), the displayed edge should be between (n,m,j) and (n,m'',-j); the printed version with (n'',m,-j) is inconsistent with the preceding vertex (n,m'',-j) and with the intended T2-sign-flip situation.","section":"Definition 3.2(2b)"},{"comment":"In the definition of J2, the set is written with set-minus I1; it should be set-minus J1, since J1 was already defined as the set of indices with zero coefficient.","section":"Proposition 3.9 proof"},{"comment":"The line 'there exists (n0,m0,j0) in M'_0' refers to M' before M' has been defined; it should refer to the connected component M0 of the previously constructed subnetwork.","section":"Lemma 5.7"},{"comment":"In the sentence discussing the coefficient of w_m, the subscript 0 is dropped; the intended coefficient is that of w_{m0}.","section":"Proposition 3.9 proof"}],"recommendation":"major_revision","confidential_remarks":"The paper is a serious contribution if the identified gaps can be closed. The main risk is the unproved Proposition 3.6, which is not merely a presentation issue: Proposition 3.9 and hence Theorem A depend on it. I would ask the authors to supply a complete proof of Proposition 3.6 and to rewrite Proposition 4.5 and Lemma 5.7 with full arguments. No concerns about attribution, citation practice, or novelty disclosure."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is worth reading. It takes Crawley-Boevey's theorem on Hom-sets between tree modules and tries to extend it to generalized tree modules, where the tree condition is relaxed and you can get sign-flip homomorphisms and ghosts. The main new content is the network formalism—pullback networks, 2-covering networks, generalized graph maps—and the claim that these give a finite generating set for the Hom-space when the ghost-free condition holds. The indecomposability criterion in Theorem B is a natural application, and the Dynkin D section is honest: it gives explicit constructions and correctly notes that the classification statement already follows from Ringel. There is real work here, and the exposition is mostly careful. The soft spot is exactly where the reader put it. Proposition 3.6 is stated without proof and is load-bearing. It asserts that a complete subnetwork being R[2]-free is equivalent to a set of \"exactly one of (1a)/(1b), at most one arrow, at most one edge\" uniqueness conditions at every vertex. Proposition 3.9 uses that uniqueness in the coefficient-counting argument to show the associated linear map is a homomorphism. If Proposition 3.6 is false, or even just not proved, then the central construction may not produce homomorphisms, and Theorem A collapses. The stress-test note is right: Proposition 4.5 and Lemma 5.7 carve out complete R[2]-free subnetworks, but they do not bypass this lemma. The paper asks in Question 5.8 whether the ghost-free hypothesis can be dropped, and that is a real limitation, not a flaw—but Proposition 3.6 is a gap, not a limitation. This is a case where the overall idea is plausible and the paper is honest about its weak points, but the main theorem is not yet fully supported. The proof of Proposition 3.6 might be a routine verification; if so, the authors just forgot to include it. But as it stands, the referee needs to see it. The paper deserves a serious referee. It is written by people who know the literature, the problem is a natural one, and the combinatorial machinery is new even if the final payoff is still conditional. I would not desk-reject it. I would send it to a representation theorist and ask specifically: prove or disprove Proposition 3.6, and make the carving arguments in Section 4 precise. If Proposition 3.6 survives, this is likely a solid paper; if it fails, Theorem A may fail too. Either way, the authors have done enough to earn engagement.","headline":"A genuine extension of Crawley-Boevey's Hom basis to a broader class, but the central theorem currently depends on an unproved combinatorial lemma that needs to be either proved or made explicit before the result is established.","tokens_in":736,"tokens_out":957,"would_cite":false,"duration_ms":21123,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["16G20"],"pacs":[],"model":"deepseek-v4-flash","headline":"For ghost-free pairs of generalised tree modules, every Hom-space is spanned by finitely many generalised graph maps.","keywords":["zero-relation algebra","generalised tree module","generalised graph map","graph map","Hom-space","indecomposability","Dynkin quiver type D","2-covering network"],"falsifier":"Check the unproved Proposition 3.6 by exhaustive search over small trees: find a complete subnetwork of a 2-covering network that is $R^{[2]}$-free yet admits, at some vertex $(n,m,j)$, two distinct incoming arrows $(n',m',j)$ and $(n'',m'',j)$ both satisfying Condition (1a) of Definition 3.2. Such a configuration would falsify the uniqueness characterisation and break the proof of Proposition 3.9, hence of Theorem A.","tokens_in":25322,"feed_emoji":"🕸️","tokens_out":18387,"duration_ms":141555,"temperature":0.7,"pith_summary":"For a zero-relation algebra, presented as a quiver in which certain paths are declared zero, this paper studies modules built from a tree by pushing forward along a map that sends the tree into the quiver. The classical tree-module condition requires distinct arrows with a common source or target to be sent to distinct arrows; the paper drops this condition and shows that, as long as a certain ghost-free hypothesis holds, the space of homomorphisms between two such generalised tree modules is spanned by finitely many explicitly constructed combinatorial maps called generalised graph maps. These maps are signed subnetworks of a 2-covering pullback network, and they generalise the graph maps that form a basis in the classical tree-module case, where they may become linearly dependent. The paper applies this spanning result to give checkable sufficient conditions for indecomposability and decomposability of generalised tree modules, and it constructs explicit generalised tree modules realising every indecomposable module over a Dynkin quiver of type D.","feed_headline":"Finitely many generalised graph maps span Hom-spaces","feed_subtitle":"Under a ghost-free condition, every homomorphism between generalised tree modules is a combination of these explicit maps.","key_machinery":"The load-bearing object is the 2-covering network $N^{[2]}$ associated with the pair of generalised tree modules: a two-sheeted cover of the pullback quiver of the two defining trees, in which each pullback vertex $(n,m)$ appears in two signed copies $(n,m,1)$ and $(n,m,-1)$, and each undirected edge that records a possible sign flip connects $(n,m,j)$ to $(n',m',-j)$. A subnetwork of $N^{[2]}$ is complete when it satisfies the existence conditions needed for the associated linear map to commute with the quiver action, and $R^{[2]}$-free when it satisfies the corresponding uniqueness conditions, expressed by forbidding certain length-two traversals inside triangles of the pullback network. A generalised graph map is a connected, involution-free, complete, $R^{[2]}$-free subnetwork; the machinery of the paper shows that the support of any homomorphism is complete, that completeness can be made $R^{[2]}$-free without changing the vertex set, and that under the ghost-free hypothesis a non-zero homomorphism always contains a generalised graph map, enabling an induction on support size that proves Theorem A.","core_discovery":"The central claim is Theorem A: for a ghost-free pair of generalised tree modules $M_1,M_2$, every homomorphism $H:M_1\\to M_2$ is a finite linear combination of homomorphisms $H_G$ attached to generalised graph maps $G$. A generalised graph map is a non-empty, connected, involution-free subnetwork of the 2-covering network $N^{[2]}$ built from the pullback of the two defining trees, satisfying two conditions called completeness and $R^{[2]}$-freeness; the attached linear map sends each basis vector $v_n$ to the signed sum of basis vectors $w_m$ appearing as vertices $(n,m,j)$ with sign $j$ in the subnetwork. The proof takes the support of $H$, carves it into a complete $R^{[2]}$-free subnetwork with the same vertex set, extracts a generalised graph map whose support lies inside the support of $H$ under the ghost-free hypothesis, and then subtracts a scalar multiple to shrink the support, giving an induction. The ghost-free hypothesis rules out non-empty, connected, involution-invariant complete $R^{[2]}$-free subnetworks, which are algebraically invisible because they define the zero homomorphism. For ordinary tree modules the construction recovers the classical graph maps, and the generating set is in fact a basis.","pith_inferences":["If the answer to the paper's Question 5.8 is affirmative, the ghost-free condition is an artefact of the support-carving induction rather than a genuine obstruction; a direct test is to take the ghost of Example 5.4 and check whether the zero homomorphism it defines can be written as a combination of generalised graph maps with strictly smaller supports.","Because the generalised graph maps are finite in number but not linearly independent in general, the dimension of the Hom-space can be computed by evaluating the finite set of homomorphisms $H_G$ on a basis and taking the rank, yielding a purely combinatorial dimension formula for ghost-free pairs.","The explicit construction for type $D$ suggests a pattern for other Dynkin quivers: choose the defining tree so that completeness conditions fail at exactly the vertices that would create off-diagonal support, then apply Theorem B; since every indecomposable over a Dynkin quiver is exceptional, such explicit trees should exist, and finding them would turn the known existence theorem for exceptiona","In characteristic 2 the 2-covering network collapses to the pullback network, so signs disappear and the distinction between graph maps and generalised graph maps vanishes; any statement that relies on the sign structure, including the ghost phenomenon, needs a separate formulation in characteristic 2."],"forward_implications":["Hom-spaces between ghost-free pairs of generalised tree modules have a finite, explicitly described generating set: the homomorphisms attached to the generalised graph maps, which can be read off directly from the 2-covering network.","A generalised tree module $M$ satisfying the ghost-free hypothesis is indecomposable whenever no generalised graph map from $M$ to itself contains a vertex $(n_1,n_2,j)$ with $n_1\\neq n_2$ coming from a subtree of the form $n_1\\leftarrow n\\rightarrow n_2$ or $n_1\\rightarrow n\\leftarrow n_2$ with equal arrow images, and no two generalised graph maps contain $(n_1,n_2,j)$ and $(n_2,n_1,j')$ respecti","For trees shaped like a central vertex with several branches satisfying a uniqueness condition on arrows, the existence of a generalised graph map containing $(n_1,n_2)$ in its support forces $M$ to decompose, and the paper conjectures that this decomposability criterion holds for every generalised tree module.","Every indecomposable module over a Dynkin quiver of type $D$ is isomorphic to a generalised tree module; the paper constructs an explicit tree for each dimension vector and shows the correct tree choice is indecomposable while wrong choices give decomposable modules."],"supporting_citations":[{"why":"This reference defines tree modules and graph maps and supplies the basis theorem for Hom-spaces between tree modules that the paper generalises.","marker":"[CB89]"},{"why":"This reference establishes the indecomposability of tree modules through universal covers, the classical result that the paper's indecomposability criteria extend.","marker":"[Gab81]"},{"why":"This reference shows that exceptional modules over finite quiver path algebras are tree modules, which the paper uses to conclude that all type-D indecomposables are generalised tree modules.","marker":"[Rin98]"},{"why":"This reference supplies the fact that all indecomposable modules over a Dynkin quiver are exceptional, which the paper needs for the type-D conclusion.","marker":"[Rin06]"},{"why":"This reference provides the idempotent criterion for indecomposability that the proof of Theorem B uses.","marker":"[ASS06]"},{"why":"This reference classifies indecomposable representations of Dynkin quivers by dimension vectors, giving the list used in Section 7's explicit constructions.","marker":"[Gab72]"}],"fun_headline_variants":["Ghost-free tree modules: homomorphisms spanned by graph maps","Generalised graph maps generate all Hom-spaces when ghost-free","Indecomposable Dynkin D modules are generalised tree modules","Homomorphisms between generalised trees: explicit finite generators","No ghosts: Hom-spaces spanned by generalised graph maps"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the pair of modules is ghost-free: the signed two-cover network contains no non-empty connected sign-invariant complete and $R^{[2]}$-free subnetwork, since the key lemma that extracts a generalised graph map from a homomorphism's support breaks down exactly when every connected component of the carved network is sign-invariant.","fun_headline_variants_meta":{"raw":{"variants":["Ghost-free tree modules: homomorphisms spanned by graph maps","Generalised graph maps generate all Hom-spaces when ghost-free","Indecomposable Dynkin D modules are generalised tree modules","Homomorphisms between generalised trees: explicit finite generators","No ghosts: Hom-spaces spanned by generalised graph maps"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000313,"raw_usage":{"total_tokens":1821,"prompt_tokens":1030,"completion_tokens":791,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":646,"completion_tokens_details":{"reasoning_tokens":705}},"tokens_in":646,"tokens_out":791,"duration_ms":7337,"temperature":1.0,"reasoning_tokens":705,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T10:04:23.992143+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check the unproved Proposition 3.6 by exhaustive search over small trees: find a complete subnetwork of a 2-covering network that is $R^{[2]}$-free yet admits, at some vertex $(n,m,j)$, two distinct incoming arrows $(n',m',j)$ and $(n'',m'',j)$ both satisfying Condition (1a) of Definition 3.2. Such a configuration would falsify the uniqueness characterisation and break the proof of Proposition 3.9, hence of Theorem A.","supporting_citations":[],"review_version":1}