{"id":"d42c9f0b-10bf-445e-9b19-cb5afee0c9bf","arxiv_id":"1908.03563","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A complete toric surface admitting an additive action has exactly one such action if its fan is wide, and two non-isomorphic actions otherwise.","lead":"This paper classifies the additive actions, equivariant completions of the affine plane, that can exist on a complete toric surface. It shows that every such surface carries either one or two non-isomorphic actions, with a simple condition on the defining fan separating the two cases.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 9 annihilator claim is false: Ann_V(M0)=V for M0=∏ x_j^{α_j2}, so the non-isomorphism proof in Theorem 3 is not valid as written.","rationale":"The paper's central claim is a complete classification of additive actions on complete toric surfaces. The most load-bearing step is the proof that the normalized and non-normalized actions are non-isomorphic (Lemma 9), because without it Theorem 3 reduces to a description of at most two actions but not a classification. That proof contains a concrete false statement: for the monomial M0 in the degree component C, the full Lie algebra V annihilates M0, so Ann_V(M0)=V. The lemma asserts the annihilator is always KD2 or 0. This is not a missing citation but an internal error visible from the paper's own equations (8): the case λ=0, λ_k=0 for all k≥1 is omitted. The intended invariant may still work after including V, since the two actions differ in the number of one-dimensional annihilators (the normalized action has infinitely many, the non-normalized action appears to have one), but that corrected argument is not supplied. The reader's identified weakness, the reduction of all additive actions to subgroups of a single maximal unipotent subgroup U, is a reasonable concern about standard structure theory, but it is less acute: for complete toric surfaces Aut(XΣ) is an algebraic group and maximal unipotent subgroups are conjugate, so the reduction is standard and likely fillable. The Lemma 9 error is a concrete defect in the proof as written and therefore the most load-bearing concern. The recommended disposition is unchanged from the reader's CONDITIONAL: the theorem is probably true, but the proof needs repair and re-verification of Lemma 9.","tokens_in":11215,"tokens_out":27304,"duration_ms":295053,"concrete_test":"In the notation of Lemma 9 for the non-normalized action, take f = M0 = ∏_{j≥3} x_j^{α_j2} ∈ C. Direct differentiation shows D1(M0)=D2(M0)=0, so Ann_V(M0)=V, contradicting the lemma's assertion that the annihilator is KD2 or 0. Then recompute the full collection {Ann_V f : f∈C∖{0}} for both actions and check whether it still separates them: the normalized action should have infinitely many one-dimensional annihilators, one for each (λ:λ1)∈P^1, while the non-normalized action should have only KD2 as a one-dimensional annihilator plus the common V. If the corrected collections remain non-isomorphic, Theorem 3 survives; if not, the non-isomorphism claim is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In Lemma 9 (Section 6), the proof that actions (5) and (6) are not isomorphic studies the homogeneous component C = ⟨x2⟩ ⊕ span{x1^k ∏_{j≥3} x_j^{α_j2−kα_j1} : 0≤k≤d} and claims that for the non-normalized action the annihilator Ann_V f is either KD2 or 0 for every nonzero f∈C. This claim is false. Take f = M0 = ∏_{j≥3} x_j^{α_j2}, the k=0 monomial in C. Both generators of V, D1=δ+∂d and D2=∂0, annihilate M0: δ differentiates only x1, while ∂d and ∂0 differentiate only x2, and M0 involves neither x1 nor x2. Hence Ann_V(M0)=V, the whole two-dimensional Lie algebra. The paper's own equations (8) confirm this: when λ=0 and all λ_k=0 for k≥1, no condition forces s1=s2=0, so the annihilator is V. Thus the dichotomy 'KD2 or 0' asserted in Lemma 9 is wrong. The underlying invariant may still distinguish the two actions, since the normalized action has infinitely many one-dimensional annihilators in C while the non-normalized action appears to have only KD2, but that corrected argument is not what the paper gives. This is load-bearing because non-isomorphism of the two actions is an essential half of Theorem 3, and the annihilator computation is the only proof offered.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper classifies additive actions (effective regular actions of G_a^n with an open orbit) on complete toric surfaces. The main result, Theorem 3, states that a complete toric surface admitting an additive action has exactly one additive action up to isomorphism if and only if its fan is wide (equivalently, it has rays in both regions A_I and A_II defined in Section 6), and otherwise it has exactly two non-isomorphic additive actions, one normalized by the acting torus and one not. The proof uses the Cox ring description of toric varieties, Demazure roots, and a normal-form analysis of commuting locally nilpotent derivations corresponding to two-dimensional commutative unipotent subgroups of the automorphism group.","tokens_in":11478,"tokens_out":10615,"duration_ms":108361,"significance":"If correct, this gives a complete and effectively checkable classification of additive actions on complete toric surfaces, extending the known cases of the projective plane, Hirzebruch surfaces, and some weighted projective planes. The paper is generally clearly written and relies on standard, well-established machinery (Demazure root theory, Cox rings, the Arzhantsev-Romaskevich classification of normalized actions). The main result is a natural and valuable contribution to the literature on additive actions and equivariant compactifications of the affine plane. However, the proof as written contains a specific error in the non-isomorphism argument (Lemma 9) and an unproved reduction to a single maximal unipotent subgroup, so the classification is not yet fully established by the present text.","major_comments":[{"comment":"The proof of Lemma 9 is incorrect: the claim that for the non-normalized action (6) the annihilator Ann_V(f) is either KD2 or 0 for every nonzero f in C fails. Taking f = ∏_{j≥3} x_j^{α_j2} (the k=0 monomial, so λ=0, λ_0=1, λ_k=0 for k≥1), both D1=δ+∂d and D2=∂0 annihilate f, because f involves neither x1 nor x2. Equations (8) impose no condition on s1 and s2 in this case, so Ann_V(f)=V, the full two-dimensional Lie algebra. Since Lemma 9 is the only argument proving that the normalized action (5) and the non-normalized action (6) are non-isomorphic, the proof of the 'otherwise there exist two non-isomorphic additive actions' half of Theorem 3 is incomplete as written. A correct invariant can likely be recovered by comparing the full collection of one-dimensional annihilators (the normalized action admits K D1, which never occurs for the non-normalized action), but this corrected argument does not appear in the paper.","section":"Section 6, Lemma 9"},{"comment":"The proof begins by declaring that additive actions on XΣ are classified by describing two-dimensional subgroups of a fixed maximal unipotent subgroup U of Aut(XΣ) generated by the positive Demazure roots R+ (with positivity determined by a vector u from Proposition 2). No justification is given for the reduction to this particular U: the authors do not prove (or cite a theorem stating) that every additive action on a complete toric surface is conjugate in Aut(XΣ) to a subgroup of U. This is load-bearing, because if some additive action were not conjugate into U, the classification in Theorem 3 would omit it. The gap is likely repairable using standard facts on unipotent subgroups of algebraic groups (the image of G_a^2 is connected and unipotent, hence lies in a maximal unipotent subgroup of Aut(XΣ)^0), but the argument must be supplied.","section":"Section 6, proof of Theorem 3"}],"minor_comments":[{"comment":"The assertion that 'the weights of the remaining m-2 coordinates with respect to the Cl(X)-grading form a basis of the lattice of characters of the torus H_X' is stated 'by construction' but not proved. Since p1 and p2 form a basis of N, the exact sequence 0→M→Z^m→Cl(X)→0 for a complete toric surface implies that the classes [D_3],...,[D_m] form a basis of Cl(X)≅Z^{m-2}; this should be spelled out to justify the open-orbit claim.","section":"Section 6, Lemma 6"},{"comment":"The proof uses exp(δ+∑η_k∂k) as an automorphism of the Cox ring, but the local nilpotence of the derivation δ+∑η_k∂k is not justified. It follows from the triangular form of the derivation with respect to the variables, but this point should be mentioned explicitly.","section":"Section 6, Lemma 7"},{"comment":"In the proof of Lemma 2, the expression '−p_i^* + ∑_{l≠i} Z_{\\ge0} p_j^*' contains a typo: the summation index is l but the basis vector is written as p_j^*; it should be p_l^*.","section":"Section 5, Lemma 2"},{"comment":"There are several typographical errors, including 'Surf ac es' and 'tor ic' in the title and abstract, 'In there term' in Section 6, and 'Propostion' in Example 4.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The main classification statement appears likely to be true, and the issues identified in the report (the annihilator computation in Lemma 9 and the reduction to a fixed maximal unipotent subgroup) are repairable without changing the statement of Theorem 3. The paper fits the journal's scope and the references are appropriate. I recommend major revision rather than rejection because the non-isomorphism half of the main theorem is not proven as written."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe headline is that Dzhunusov proves a natural classification: every complete toric surface admits at most two additive actions, and exactly one iff the fan is wide. That is a real step beyond the known special cases (P^2, Hirzebruch surfaces, weighted projective planes), the criterion is a simple numerical condition on rays, and the Demazure-root machinery is used cleanly. The bracket computations, the normal form reduction in Lemma 7, and the torus-conjugation argument in Lemma 8 are all coherent.\n\nThe trouble is Lemma 9, the only proof that the normalized and non-normalized actions are non-isomorphic. The stress-test note is right: take f = M0 = product_{j≥3} x_j^{α_j2}. Both D1 = δ + ∂d and D2 = ∂0 kill M0, so Ann_V(M0) = V, not KD2 or 0. The paper's own equations (8) confirm this: when λ = 0 and all λ_k for k≥1 are zero, no condition constrains s1 or s2. So the asserted dichotomy is false, and the non-isomorphism half of Theorem 3 is not proved as written. The underlying invariant may still separate the actions, but the argument needs a new idea or a more careful choice of test subspace.\n\nTwo smaller points. First, the proof of Theorem 3 starts by saying it will classify two-dimensional subgroups of a single maximal unipotent subgroup U of Aut(XΣ), but I do not see where it is shown that every additive action is conjugate into that U. This may be standard or derivable from [5], but it should be stated. Second, in Lemma 6 the claim that the degrees of x3,...,xm form a basis of the character lattice of H_X is asserted \"by construction\"; this is true for complete toric surfaces, but the argument belongs in the text.\n\nOverall: the theorem is likely correct and the paper is worth engaging, but it is not in publishable shape. A serious referee should send it back for a corrected Lemma 9 and a small clarification of the reduction step. I would not cite the main theorem until the proof is repaired, but I would bring the paper to the reading group—the wide-fan criterion is attractive and the flaw is instructive.","headline":"The wide-fan classification is a nice new result, but the non-isomorphism proof has a false annihilator claim that needs fixing.","tokens_in":12036,"tokens_out":5464,"would_cite":false,"duration_ms":53914,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14L30","14M25","13N15","14J50","14M17"],"pacs":[],"model":"deepseek-v4-flash","headline":"Complete toric surfaces carry either one or two additive actions, and a simple fan condition decides which.","keywords":["additive action","toric variety","complete surface","Demazure root","locally nilpotent derivation","Cox ring","unipotent group","fan classification"],"falsifier":"Compute, for the complete toric surface with ray vectors $p_1=(1,0)$, $p_2=(0,1)$, $p_3=(-1,-2)$, $p_4=(-2,-1)$, all pairs of commuting homogeneous locally nilpotent derivations of degree zero on its Cox ring; the theorem predicts that every pair is conjugate to the displayed normalized pair. If any pair not conjugate to that pair exists, Theorem 3 is false. Equivalently, any complete toric surface with a wide fan carrying two non-isomorphic additive actions would refute the dichotomy.","tokens_in":10960,"feed_emoji":"📐","tokens_out":12964,"duration_ms":121295,"temperature":0.7,"pith_summary":"An additive action on a surface is a regular, effective action of the two-dimensional additive group $\\mathbb{G}_a^2$ with an open orbit; it views the surface as an equivariant completion of the affine plane. This paper proves a complete classification for complete toric surfaces. Whenever such an action exists, there is either exactly one additive action up to isomorphism, or exactly two non-isomorphic ones, one normalized by the acting torus and one not. The alternative is determined by the fan: if the fan is wide, meaning the two root sets $R_1$ and $R_2$ each consist of a single element, the action is unique; otherwise both classes occur. This turns the classification into a finite combinatorial check on the defining fan and explains the known two-action pattern on $\\mathbb{P}^2$ and Hirzebruch surfaces.","feed_headline":"Toric surfaces admit one or two additive actions, never three","feed_subtitle":"A simple geometric condition on the fan tells which: a unique action when the fan is wide, two otherwise.","key_machinery":"The machinery is the dictionary between additive actions and pairs of commuting homogeneous locally nilpotent derivations of degree zero on the Cox ring $R(X_\\Sigma)$, with roots indexed by Demazure roots. After fixing an ordering so that the primitive ray vectors $p_1,p_2$ are a basis, the root sets are $R_1=\\{(-1,k):0\\le k\\le d_1\\}$ and $R_2=\\{(k,-1):0\\le k\\le d_2\\}$ for some nonnegative integers; the fan is wide exactly when both sets are singletons, i.e. $d=0$. The load-bearing identity is the commutator relation $[\\delta,\\partial_k]=k\\,\\partial_{k-1}$, which forces the coefficient of $\\partial_d$ to be invariant up to torus conjugation, so all nonzero values of that coefficient collapse into one non-normalized isomorphism class.","core_discovery":"The central claim is Theorem 3 of the paper. Let $X_\\Sigma$ be a complete toric surface admitting an additive action. Then $X_\\Sigma$ has exactly one additive action up to isomorphism if and only if the fan $\\Sigma$ is wide; otherwise it has exactly two non-isomorphic additive actions, one normalized and one not. The proof classifies all additive actions by describing the two-dimensional commutative subgroups of the maximal unipotent subgroup $U$ of $\\mathrm{Aut}(X_\\Sigma)$ generated by the one-parameter additive root subgroups. In Cox-ring coordinates the two classes are explicit: the normalized action is generated by derivations $\\delta$ and $\\partial_0$, while the non-normalized action is generated by $\\delta + \\partial_d$ and $\\partial_0$ with $d>0$, and conjugation by the torus reduces every non-normalized action to the same normal form with coefficient $1$.","pith_inferences":["Beyond the paper's claims: the same normal-form computations suggest a testable extension to complete toric threefolds, where the paper notes $\\mathbb{P}^3$ already has four non-isomorphic additive actions, so the one-or-two pattern is special to surfaces.","Beyond the paper's claims: the annihilator-line invariant used to separate the two surface actions could serve as a distinguishing invariant for additive actions on singular non-toric normal surfaces, where the Cox ring is still available but not polynomial; comparing it on the known singular del Pezzo examples would be a natural check.","Beyond the paper's claims: because the theorem reduces existence and uniqueness to integer comparisons among the coefficients $\\alpha_{ji}$ of the excess rays, the classification is effectively checkable by a finite algorithm for any explicitly presented complete toric surface."],"forward_implications":["Every complete toric surface that admits an additive action has at most two isomorphism classes of additive actions.","The unique-action case is characterized combinatorially: a fan is wide exactly when $R_1$ and $R_2$ are singletons, equivalently when primitive rays occur in both open regions of the negative quadrant separated by the diagonal.","When two actions exist, one is normalized by the torus and the other is not, and the non-normalized one has the explicit normal form $\\delta+\\partial_d,\\partial_0$; the parameter $\\mu_d$ is a fake modulus because torus conjugation identifies all nonzero values.","The classification reproduces the known cases: $\\mathbb{P}^1\\times\\mathbb{P}^1$ and the fan with $p_3=-p_1-2p_2$, $p_4=-2p_1-p_2$ have a single action, while $\\mathbb{P}^2$ and the Hirzebruch surface $F_1$ have two.","The two actions are distinguished by an explicit invariant computed in the Cox ring: the collection of annihilator lines $\\operatorname{Ann}_V f$ on the homogeneous component $C$ containing $x_2$ is a continuous family of lines for the normalized action and a two-point set for the non-normalized one."],"supporting_citations":[{"why":"Supplies the theorem that normalized additive actions on toric varieties correspond to complete collections of Demazure roots, and that existence of any additive action implies existence of a normalized one; the main proof builds directly on this.","marker":"[5]"},{"why":"Establishes the Cox ring presentation of toric varieties and the correspondence between regular $\\mathbb{G}_a$-actions and homogeneous locally nilpotent derivations of degree zero, the language in which all actions are written.","marker":"[9]"},{"why":"Introduces Demazure roots and the normalized one-parameter subgroups attached to them, the objects whose positive part generates the maximal unipotent subgroup $U$.","marker":"[11]"},{"why":"Provides the earlier classifications on projective space and Hirzebruch surfaces that Theorem 3 extends, including the two-action phenomenon and the explicit non-normalized action on $F_1$.","marker":"[17]"},{"why":"Supplies the annihilator-subspace method used in Lemma 9 to prove that the normalized and non-normalized actions are not isomorphic, together with the weighted projective plane cases.","marker":"[2]"},{"why":"Provides the Cox ring construction, the grading, and the good quotient formalism used to pass from derivations on the Cox ring to actions on the surface.","marker":"[3]"}],"fun_headline_variants":["Additive actions on toric surfaces: one or two, never three","Fan width tells if a toric surface has a unique additive action","One or two additive actions: fan width decides","Toric surfaces: unique additive action when fan is wide","Complete toric surfaces: additive action count from fan shape"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The classification assumes that every additive action on a complete toric surface is isomorphic to a two-dimensional commutative subgroup of the subgroup of automorphisms generated by the one-parameter additive subgroups attached to the positive Demazure roots, so that classifying those subgroups is exhaustive; the paper uses this step without writing out a full proof.","fun_headline_variants_meta":{"raw":{"variants":["Additive actions on toric surfaces: one or two, never three","Fan width tells if a toric surface has a unique additive action","One or two additive actions: fan width decides","Toric surfaces: unique additive action when fan is wide","Complete toric surfaces: additive action count from fan shape"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000173,"raw_usage":{"total_tokens":1184,"prompt_tokens":759,"completion_tokens":425,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":375,"completion_tokens_details":{"reasoning_tokens":341}},"tokens_in":375,"tokens_out":425,"duration_ms":4313,"temperature":1.0,"reasoning_tokens":341,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:34:32.135882+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, for the complete toric surface with ray vectors $p_1=(1,0)$, $p_2=(0,1)$, $p_3=(-1,-2)$, $p_4=(-2,-1)$, all pairs of commuting homogeneous locally nilpotent derivations of degree zero on its Cox ring; the theorem predicts that every pair is conjugate to the displayed normalized pair. If any pair not conjugate to that pair exists, Theorem 3 is false. Equivalently, any complete toric surface with a wide fan carrying two non-isomorphic additive actions would refute the dichotomy.","supporting_citations":[{"cited_title":"Additive actions on tor ic varieties","cited_arxiv_id":null,"evidence_quote":"Supplies the theorem that normalized additive actions on toric varieties correspond to complete collections of Demazure roots, and that existence of any additive action implies existence of a normalized one; the main proof builds directly on this."},{"cited_title":"The homogeneous coordinate ring of a toric variety","cited_arxiv_id":null,"evidence_quote":"Establishes the Cox ring presentation of toric varieties and the correspondence between regular $\\mathbb{G}_a$-actions and homogeneous locally nilpotent derivations of degree zero, the language in which all actions are written."},{"cited_title":"Sous-groupes algebriques de rang maximum d u groupe de Cremona","cited_arxiv_id":null,"evidence_quote":"Introduces Demazure roots and the normalized one-parameter subgroups attached to them, the objects whose positive part generates the maximal unipotent subgroup $U$."},{"cited_title":"Geometry of equivariant compactifications of Gn a","cited_arxiv_id":null,"evidence_quote":"Provides the earlier classifications on projective space and Hirzebruch surfaces that Theorem 3 extends, including the two-action phenomenon and the explicit non-normalized action on $F_1$."},{"cited_title":"Commutative algebraic monoid structures on affine spaces","cited_arxiv_id":"1809.05291","evidence_quote":"Supplies the annihilator-subspace method used in Lemma 9 to prove that the normalized and non-normalized actions are not isomorphic, together with the weighted projective plane cases."},{"cited_title":"Cox rings","cited_arxiv_id":null,"evidence_quote":"Provides the Cox ring construction, the grading, and the good quotient formalism used to pass from derivations on the Cox ring to actions on the surface."}],"review_version":1}