{"id":"69b841e4-e3ca-44b2-b487-7b547099dd9e","arxiv_id":"2412.12022","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A finite subgroup of the plane Cremona group is linearizable if and only if its G-Mori fibre space model is one of a short list of del Pezzo or Hirzebruch surface types, completing the linearization problem.","lead":"This paper gives the complete answer to a long-open question: which finite groups of birational transformations of the plane are conjugate to ordinary linear transformations. The answer is a short, explicit list of group types acting on specific surfaces, and the paper proves every other finite group action cannot be linearized.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Main Theorem's only-if for conic bundles relies on an unproved reduction to elementary transformations in §6.3 and an unwritten converse for Theorem 6.1.","rationale":"The paper is a serious and largely convincing solution of a major problem, and the main theorem is the right kind of result to build on the equivariant Sarkisov classification. My concern is not that the cited classification is wrong, but that two load-bearing reductions are asserted rather than proved: the restriction in §6.3 of a linearization of a rank-2 quadric to a chain of elementary transformations and type IV links, and the converse direction of Theorem 6.1. Both are likely fillable by a careful reader, so I do not recommend rejection; they are exactly the sort of gap that makes the present version conditionally acceptable. The reader's weakest assumption identified the external completeness of Iskovskikh's link classification and the missing converse of Theorem 6.1; I agree in substance, though I would locate the more immediate internal soft spot in the §6.3 'crucial observation' and the implicit final reduction for Theorem 6.1. If those two steps are written out and checked against the cited classification, the main theorem would be fully supported.","tokens_in":41478,"tokens_out":41671,"duration_ms":391289,"concrete_test":"Derive the §6.3 observation from [DI09, Propositions 7.12, 7.13] by enumerating all Sarkisov G-links starting from F0 with Pic(S)^G=Z^2; if any type I or type III link occurs with an odd-length center, the rank-2 classification is incomplete. Then complete Theorem 6.1's only-if: cite Corollary 6.9 to reduce even n to F0 and check, via Lemma 6.17 and Theorem 6.14, that the excluded groups are exactly those with \\hat G neither cyclic nor D_odd.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 6.3 begins with the assertion that for a rank-2 quadric S=F0, G is linearizable if and only if there is a chain F0 -> ... -> F1 -> P2 whose links are only G-elementary transformations or type IV, followed by a type III contraction. This 'crucial observation' is said to follow from [DI09, Propositions 7.12, 7.13], but the verification is omitted. It is the precise place where an unlisted link type—for instance a type I or III link centered at an odd-length G-orbit—would invalidate Theorem 6.14 and hence the Main Theorem's quadric rows. Likewise, Theorem 6.1 is stated as an iff, yet §6.1 supplies only the sufficiency direction: Corollary 6.8 covers cyclic and odd-dihedral base groups, and Corollary 6.9 covers odd n. The necessity for even n with \\hat G neither cyclic nor D_odd is never written; it must be inferred by reducing F_n to F0 and applying the later rank-2 classification. Until that reduction and the §6.3 observation are made explicit, the completeness half of the central claim is not fully demonstrated.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims a complete classification of linearizable finite subgroups of the plane Cremona group over an algebraically closed field of characteristic zero. The Main Theorem states that a finite group G ⊂ Cr2(k) is linearizable if and only if a regularization on a G-Mori fibre space is one of the following: a G-conic bundle over P1 of a specified Hirzebruch-surface or quadric type, or a G-del Pezzo surface of degree 5, 6, 8, or 9 with the specified group actions (including Blichfeldt's list for P2). The proof uses the equivariant Sarkisov program, reduces to G-del Pezzo surfaces of degree at least 4 and to conic bundles with no singular fibres, and then analyzes each case. The paper also contains a self-contained classification of finite subgroups of Aut(P1 × P1) with supporting Magma code.","tokens_in":41711,"tokens_out":6706,"duration_ms":60987,"significance":"If the Main Theorem is fully established, this is a major result: it solves a long-standing open problem and provides a complete, explicit list of linearizable finite subgroups of Cr2(k). The systematic use of the equivariant Sarkisov program, together with the detailed group-theoretic case analysis, is appropriate and the paper is clearly a substantial contribution to the field. The authors provide reproducible Magma code for the quadric-automorphism classification, and the overall strategy of reduction to K2 ≥ 4 del Pezzo surfaces and to no-singular-fibre conic bundles is sound. The main caveats are two places where the text states or relies on load-bearing implications without giving the full verification: the converse direction of Theorem 6.1 and the 'crucial observation' in Section 6.3. These are fixable but are necessary for the claimed completeness.","major_comments":[{"comment":"Theorem 6.1 is stated as an if-and-only-if, but the proof in §6.1 establishes only the sufficiency direction: Corollary 6.8 covers cyclic and odd-dihedral base actions, and Corollary 6.9 covers odd n. The necessity direction — that for even n with \\hat G neither cyclic nor isomorphic to D_{2m+1}, the G-conic bundle F_n is not G-birational to F1 or P2 — is never written. This is load-bearing for the first two rows of the Main Theorem. Please add a proof of this necessity, or state and justify a reduction to the rank-2 quadric classification (Theorem 6.14) and explain how the hypotheses of that theorem apply.","section":"§6.1, Theorem 6.1"},{"comment":"The paragraph after Proposition 5.2 in Section 6.3 asserts that for a rank-2 quadric S = F0, G is linearizable if and only if there is a sequence of G-elementary transformations and type IV links from F0 to F1 followed by a type III contraction, and that this follows from [DI09, Propositions 7.12, 7.13]. No derivation is given. This observation is the starting point of Theorem 6.14 and therefore underlies the rank-2 quadric rows of the Main Theorem. Please provide the missing verification, in particular explaining why no other Sarkisov link types (e.g., type I or type II links centred at orbits of length 1, 2, 3, or 5) can occur or are already accounted for in the stated chain.","section":"§6.3, 'crucial observation'"}],"minor_comments":[{"comment":"In the case |Σ| = 2, the sentence 'the points E2 ∩ E3 and E5 ∩ E6 are unique G-fixed points on T' and the later conclusion about T' are somewhat compressed; a short explanation of why no other G-fixed points appear after elementary transformations would improve readability.","section":"§6.2, Proposition 6.13"},{"comment":"The phrase 'the lengths of the orbits is preserved under G-fibrewise transformations' is not literally true for orbits on the total space; what is preserved is the length of the induced orbit on the base P1. The parity argument in the lemma is convincing once this is clarified, but the wording should be corrected.","section":"§6.2, Lemma 6.17"},{"comment":"In the table row for the quadric F0, the entry 'D_{2n+1} ×Q D_{2m+1} is dihedral' uses n,m without explicitly stating the range; Section 6.3 uses n,m ≥ 3, but the table would benefit from restating this condition.","section":"Main Theorem, table"},{"comment":"The table entries labelled 'No id' could confuse readers; it would be helpful to state explicitly that these families are infinite and therefore have no single GAP ID.","section":"Section 5, Theorem 5.6"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does what it says: it gives a complete list of linearizable finite subgroups of the plane Cremona group over an algebraically closed field of characteristic zero. That is a major result, and the list is clean enough to be a standard reference. The route through G-Mori fibre spaces is standard but executed carefully: del Pezzo surfaces of degree 5 and 6 are handled with explicit linearizations for C5, D5, C6, S3, and the non-linearizable cases are blocked by rigidity and the Burnside-type examples. Section 5, the classification of finite subgroups of Aut(P1×P1) with matrix generators, is a useful standalone contribution, especially with the Magma code in the appendix. The self-citations to [Yas24] and [Pin24b] are for imported lemmas, not for the main conclusion; that is fine.\n\nThe soft spots are real but mostly organizational. Theorem 6.1 is stated as an iff, but Section 6.1 explicitly proves only the sufficiency direction. The necessity for even n is supposed to follow by reducing to F0 and then applying the later rank-1 and rank-2 classifications, but that reduction is not spelled out. Similarly, the 'crucial observation' opening Section 6.3 — that linearizability from a rank-2 quadric forces a chain of elementary transformations/type IV links ending at F1 — is asserted with a citation to [DI09] and no verification. I believe the observation is correct and follows from the classification of Sarkisov links, but the paper should show the argument. There are also a few case checks in Section 5 that are left to the reader, though the Magma code compensates. None of this looks like a load-bearing error; it is unfinished exposition at exactly the points where a skeptic would want detail.\n\nThe dependence on Iskovskikh's classification of equivariant Sarkisov links is standard in this area and not a flaw. The abstract and intro are honest about what is new and what is borrowed. This is a paper for people working in birational geometry, Cremona groups, and finite transformation groups; it will be cited and used.\n\nMy recommendation: send it to a serious referee. The referee should ask the authors to expand the converse of Theorem 6.1 and the Section 6.3 observation into explicit proofs. Those are fixable gaps, and the result deserves to be in the literature in fully rigorous form.","headline":"A complete-looking solution to the linearization problem for finite subgroups of Cr2(k); the main theorem is very likely correct, but two key converse arguments are only implicit.","tokens_in":42252,"tokens_out":10965,"would_cite":true,"duration_ms":107280,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14E07","14E05","14E30","14J45","14M22"],"pacs":[],"model":"deepseek-v4-flash","headline":"A finite subgroup of the plane Cremona group is linearizable exactly when its minimal model appears on a short explicit list.","keywords":["plane Cremona group","linearization","G-Mori fibre space","Sarkisov program","del Pezzo surfaces","Hirzebruch surfaces","finite group actions","birational geometry"],"falsifier":"Test the converse of Theorem 6.1 directly: take a $G$-conic bundle $F_n$ with $n$ even whose base action has orbit lengths with greatest common divisor 1 but is neither cyclic nor dihedral of odd order (for instance a base image isomorphic to $A_4$), and search for a chain of $G$-elementary transformations to $F_1$; the theorem predicts none exists, so producing one would refute the Main Theorem.","tokens_in":41277,"feed_emoji":"📐","tokens_out":13478,"duration_ms":105815,"temperature":0.7,"pith_summary":"This paper claims a complete solution of the linearization problem for finite subgroups of the plane Cremona group over an algebraically closed field of characteristic zero. It asserts that after regularizing a finite subgroup $G \\subset \\mathrm{Cr}_2(k)$ on a two-dimensional $G$-Mori fibre space, $G$ is linearizable — conjugate in the Cremona group to a subgroup of $\\mathrm{PGL}_3(k)$ — if and only if that surface with its $G$-action appears on a short explicit list. The list covers $G$-conic bundles with $K^2=8$ satisfying parity conditions, and $G$-del Pezzo surfaces of degrees $5,6,8,9$ with specified cyclic, dihedral, or symmetric groups. If the theorem is correct, the long-standing classification question reduces to a membership test.","feed_headline":"Finite plane Cremona groups: linearizable exactly on a short list","feed_subtitle":"A complete list now separates the linearizable finite plane Cremona actions from the rest.","key_machinery":"The machinery is the equivariant Sarkisov program in dimension 2: every $G$-birational map between $G$-Mori fibre spaces decomposes into Sarkisov $G$-links of types I, II, III, and IV, each a blow-up followed by a contraction. The paper combines this with orbit-length arithmetic on $\\mathbb{P}^1$ (Klein's classification) and on $\\mathbb{P}^1 \\times \\mathbb{P}^1$ (Goursat fibre products) to decide which chains of links can reach $\\mathbb{P}^2$, and with explicit $G$-elementary transformations on Hirzebruch surfaces to construct linearizations when possible. The Euclidean algorithm in Proposition 6.19 is the constructive core for the dihedral fibre-product cases.","core_discovery":"The central claim is the Main Theorem. Let $k$ be algebraically closed of characteristic zero and let $G \\subset \\mathrm{Cr}_2(k)$ be finite. Choose a regularization of $G$ on a two-dimensional $G$-Mori fibre space $S$ over the base $B$. Then $G$ is linearizable if and only if $(S,G)$ is one of the following: a $G$-conic bundle over $B \\simeq \\mathbb{P}^1$ with $K_S^2 = 8$, namely a Hirzebruch surface $F_n$ with $n$ odd (any $G$), or $F_n$ with $n>0$ even where $G$ acts on $B$ cyclically or as $D_{2m+1}$, or the quadric $F_0 \\simeq \\mathbb{P}^1 \\times \\mathbb{P}^1$ with $G$ equal to $C_n \\times_Q C_m$, $C_n \\times_Q D_{2m+1}$, or a dihedral $D_{2n+1} \\times_Q D_{2m+1}$; or a $G$-del Pezzo surface, namely the quintic with $C_5$ or $D_5$, the sextic with $C_6$ or $S_3$, the quadric $\\mathbb{P}^1 \\times \\mathbb{P}^1$ with $(C_n \\times_Q C_n)\\bullet C_2$, or $\\mathbb{P}^2$ with Blichfeldt's list. Here $\\times_Q$ denotes a fibre product over a common quotient and $\\bullet$ denotes an extension. The list is both necessary and sufficient, so every finite subgroup not represented by one of these models is non-linearizable.","pith_inferences":["One testable extension is to automate the criterion: given generators of a finite subgroup of $\\mathrm{Cr}_2(k)$, a computer algebra system could compute the $G$-Mori fibre space and check membership in the list; the paper does not provide such an implementation.","The Euclidean-algorithm step used for dihedral fibre products is constructive, so the birational maps it produces could be assembled into explicit conjugating transformations for the linearizable cases.","Over non-closed fields such as $\\mathbb{R}$ or $\\mathbb{Q}$, the list is not expected to survive unchanged: the orbit-counting arguments rely on algebraic closure, and fixed-curve rigidity is already known to behave differently over $\\mathbb{R}$."],"forward_implications":["The linearization problem for the plane Cremona group is completely solved in the stated setting.","Every finite subgroup whose minimal model is a $G$-conic bundle with $K^2 \\in \\{1,2,4\\}$ is non-linearizable; the only linearizable conic bundles in the list have $K^2 = 8$.","On del Pezzo surfaces of degrees 5 and 6, only the listed cyclic and small symmetric or dihedral groups are linearizable; the $D_6$-action on the sextic is not linearizable.","Linearizability can be decided by computing a regularization, running the $G$-minimal model program, and comparing the resulting model with the list.","Stable linearizability is genuinely weaker than linearizability in dimension 2: the $D_6$-action on the sextic is stably linearizable but not linearizable."],"supporting_citations":[{"why":"Supplies the classification of Sarkisov G-links in dimension 2 (Theorem 2.6) that every G-birational map is assumed to decompose into.","marker":"[Isk96]"},{"why":"Provides the equivariant link classification (Propositions 7.12, 7.13) and the reduction from subgroups of the Cremona group to G-Mori fibre spaces.","marker":"[DI09]"},{"why":"Classifies finite subgroups of PGL2(k), the base groups whose orbit lengths drive the parity arguments for Hirzebruch surfaces.","marker":"[Kle19]"},{"why":"Gives Blichfeldt's classification of finite subgroups of PGL3(k), which is exactly the linearizable P2 case in the Main Theorem.","marker":"[Bli17]"},{"why":"Supplies Goursat's lemma, used to enumerate fibre-product actions on P1×P1.","marker":"[Gou89]"},{"why":"Establishes the structure and G-minimality criteria for G-conic bundles, restricting the possible K^2 values to {1,2,4,8}.","marker":"[Isk80]"}],"fun_headline_variants":["Complete linearization list for finite plane Cremona groups","Finite plane Cremona groups: full linearizability classification","Exact list of linearizable finite plane Cremona groups","All linearizable finite plane Cremona groups now classified"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the classification of equivariant Sarkisov links in dimension 2 is exhaustive, so that every $G$-birational map from a $G$-Mori fibre space to $\\mathbb{P}^2$ decomposes into links of types I, II, III, and IV.","fun_headline_variants_meta":{"raw":{"variants":["Complete linearization list for finite plane Cremona groups","Finite plane Cremona groups: full linearizability classification","Exact list of linearizable finite plane Cremona groups","All linearizable finite plane Cremona groups now classified"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001051,"raw_usage":{"total_tokens":4396,"prompt_tokens":906,"completion_tokens":3490,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":522,"completion_tokens_details":{"reasoning_tokens":3425}},"tokens_in":522,"tokens_out":3490,"duration_ms":22485,"temperature":1.0,"reasoning_tokens":3425,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:23:04.224860+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Test the converse of Theorem 6.1 directly: take a $G$-conic bundle $F_n$ with $n$ even whose base action has orbit lengths with greatest common divisor 1 but is neither cyclic nor dihedral of odd order (for instance a base image isomorphic to $A_4$), and search for a chain of $G$-elementary transformations to $F_1$; the theorem predicts none exists, so producing one would refute the Main Theorem.","supporting_citations":[],"review_version":1}