{"id":"2f67c542-889f-4131-80de-eb267bee56cd","arxiv_id":"2507.00167","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Potential density of integral points is established for relative SL2 and PGL2 character varieties of all smooth quasi-projective complex varieties with snc compactification.","lead":"This paper proves that for SL2 and PGL2 local systems on any smooth quasi-projective complex variety, integral points are Zariski dense after a finite extension of the number field. The result confirms a conjecture of Simpson and Campana for these rank-2 groups, building on the rigidity classification of Corlette-Simpson.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Orbicurve reduction in §7.4 is false for SL2 with order-2 stacky points: Prop 3.1.3 makes X_{1,-I} a point while X_{1,1,-2} is a surface.","rationale":"The surface-group theorems (Theorem 5.0.4 and Theorem 6.0.1) appear detailed and give strong positive-dimensional surface cases; I am not objecting to them. The fragile point is the passage from arbitrary Y to orbicurves and from orbicurves to surfaces. The external classification theorems may be fine, but the internal claim that orbicurve relative character varieties are disjoint unions of surface-group relative character varieties is false in the SL2 case: an order-2 orbifold point imposes monodromy -I, which is strictly stronger than fixing the boundary trace to -2. Proposition 3.1.3 itself records that the -I locus in Σ_{1,1} is a point, while the trace -2 locus is a surface. Hence Theorem 6.0.1 cannot imply density on the orbicurve variety. This is a concrete, checkable gap. It does not disprove the main theorem, but it means the proof as written does not establish Theorem 1.1.2 for SL2 over arbitrary quasi-projective Y. I therefore keep the reader's conditional posture but sharpen the condition to a specific missing case. Agreement is partial because the reader also flagged the external classification theorems; my concern is the orbicurve decomposition, which is not merely unproven but false as stated.","tokens_in":21141,"tokens_out":24125,"duration_ms":281539,"concrete_test":"Construct the orbicurve Z obtained from Σ_{1,1} by imposing c^2 = 1, where c is the loop around the puncture and [α,β]c = 1. Compute the SL2 relative character variety X_{SL2,C'}(Z) with C' forcing the order-2 inertia to map to -I. By Proposition 3.1.3 this is a single point (0,0,0). Compare with the surface relative character variety X_{SL2,-2}(Σ_{1,1}), the Markoff surface x^2 + y^2 + z^2 - xyz = 0, which is 2-dimensional. The two differ, so the §7.4 assertion that orbicurve relative character varieties are disjoint unions of surface-group relative character varieties fails for SL2; a separate argument for order-2 orbifold points is needed to complete Theorem 1.1.2.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing step is the reduction in §7.4 from a quasi-projective Y to punctured surfaces. After using [CS08, Thm 1] / [LPT16, Thm A] to make the generic representation ρ factor through an orbicurve Z, the proof says potential density for X_{G,C'}(Z) follows from Theorem 6.0.1, 'using that relative character varieties of orbicurves are disjoint unions of relative character varieties of surface groups.' This assertion is false for G = SL2 when Z has an orbifold point of order 2. Such a point forces the local monodromy to be the central matrix -I. But in a surface relative character variety, fixing the trace of a boundary loop to -2 does not force -I: for the once-punctured torus, X_{1,1,-2} is the Markoff surface x^2 + y^2 + z^2 - xyz = 0, whereas the locus with monodromy -I at the puncture is the single point (0,0,0) by Proposition 3.1.3, and the paper's own X_{1,-I} is a point. Thus the orbicurve character variety is generally a proper subvariety of the surface relative character variety, not a union of components. Zariski density of integral points in the larger variety does not imply density in a proper subvariety, so Theorem 6.0.1 cannot be invoked as written. The proof of Theorem 1.1.2 for SL2 and dim Y > 1 therefore has a gap unless order-2 stacky points are treated separately, for example via the X_{g,-I} technology of §6, which the paper does not do in this step.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves potential density of integral points in the relative SL2 and PGL2 character varieties of a smooth quasi-projective complex variety Y, with fixed algebraic-integer trace data along the boundary of a simple normal crossings compactification. The proof combines a detailed surface-group theorem (explicit integral P-good representations with Zariski-dense pure mapping class group orbits, constructed via the matrix families M_K, N_K, L_K and gluing along pants decompositions) with a reduction from arbitrary quasi-projective Y to curves using the Corlette–Simpson and Loray–Pereira–Touzet classification of rank-two local systems. The main theorem (Theorem 1.1.2) would resolve Conjecture 1.1.1 for G = SL2 and PGL2 and would give the first positive-dimensional cases of Simpson's integrality conjecture for rank 2.","tokens_in":21484,"tokens_out":21301,"duration_ms":221735,"significance":"If the result is correct, it is a substantial advance: it settles Campana's conjecture for these SL2/PGL2 relative character varieties and provides the first positive-dimensional evidence for Simpson's integrality conjecture beyond rigid local systems. The surface-group part is a strong technical contribution: the explicit matrix families, the P-good point constructions, the X_{g,-I} technology, and the careful bookkeeping of field extensions are detailed and largely self-contained. The paper also clearly identifies which ingredients come from external classification theorems. However, the quasi-projective reduction in §7.4 is the most fragile part of the manuscript, and two load-bearing gaps there and in Proposition 3.1.10 currently prevent the main theorem from being proven as written.","major_comments":[{"comment":"Lemma 3.1.8 is false as stated for N > 1. The proof asserts that the orbit of any point of G_m^{3g-3+n} under the group generated by the diagonal transformations T_{z_i} is Zariski dense when each t_i lies in A^1(Q) \\ E. This is only true if the λ_i are multiplicatively independent; for example, if N = 2 and λ_1 = 2, λ_2 = 4, the orbit is contained in the subtorus x_2 = x_1^2 and is not Zariski dense in G_m^2. The lemma is used in the final step of Proposition 3.1.10 to pass from Zariski density of tr_P(Γ_{g,n} · p) in A^{3g-3+n} to Zariski density of Γ_{g,n} · p in X_{g,n,k,Q}. That step requires the lemma for the full pants decomposition, i.e. for N = 3g-3+n, which can be arbitrarily large. Without an additional multiplicative-independence hypothesis (or a different argument, e.g. citing the known orbit-closure results of Golsefidy–Tamam), the proof of Proposition 3.1.10, and hence of Theorem 5.0.2 and Theorem 5.0.4, is incomplete.","section":"§3.1, Lemma 3.1.8"},{"comment":"The assertion that relative character varieties of orbicurves are disjoint unions of relative character varieties of surface groups is false for SL2 when the orbicurve has a stacky point of order 2. Such a point forces the local monodromy of an SL2 local system to be the central matrix -I. But in a surface relative character variety, fixing the boundary trace to -2 does not force -I: by Proposition 3.1.3, X_{1,1,-2} is the Markoff surface x^2 + y^2 + z^2 - xyz = 0, while the locus with monodromy -I at the puncture is the single point (0,0,0). Thus the SL2 relative character variety of the orbicurve is a proper subvariety, not a union of components, of the natural surface relative character variety. Consequently Theorem 6.0.1 cannot be invoked to conclude potential density in the orbicurve variety. The proof needs to treat order-2 stacky points separately, for example by using the X_{g,-I} technology developed in §6.2–§6.3, which is not done in this step.","section":"§7.4, proof of Theorem 1.1.2"},{"comment":"The lifting step from PGL2 back to SL2 does not control the boundary trace conditions. After proving potential density of integral points in the PGL2 relative character variety W, the proof invokes Lemma 7.3.1 to lift each OK'-point of W to a point of the SL2 relative character variety W'. But Lemma 7.3.1 only produces some SL2 lift; it does not guarantee that the lifted representation satisfies the prescribed tuple C' of boundary traces. For example, if C'_i = -2 at a boundary component, the corresponding PGL2 boundary datum is f = 4, which also contains unipotent monodromy. A Zariski-dense set of PGL2 points with f = 4 may avoid the locus that lifts to a representation with boundary trace -2. Therefore the density statement for W does not by itself imply density for W', and the claim that 'all these points lift to W'' needs a separate argument that respects the boundary traces, presumably again involving the X_{g,-I} loci.","section":"§7.4, final SL2 lifting step"}],"minor_comments":[{"comment":"In the proof, the system is said to be solved for 'x, y, z, w ∈ Z', but the context is a number field K and its ring of integers; this should read O_K (or O_L after passing to the quadratic extension).","section":"§4.1, Lemma 4.1.1"},{"comment":"In the case n = 1, g > 1 and k ∈ E, the text reads 'we pick M ∈ N with tr M /∈ E'; the symbol N should be N_K.","section":"§5, proof of Proposition 5.0.1"},{"comment":"The remark asserts that relative character varieties are independent of the choice of simple normal crossings compactification, but says only 'we leave verifying the details to the reader'. Since this is used freely in §7, a short proof or a precise reference would improve the exposition.","section":"§7.1, Remark 7.1.1"},{"comment":"The notation γ for a pants curve in the definition of X_{g,-I} conflicts with the earlier use of γ for a loop around a puncture in §3.1; this is a minor notational clash that could confuse readers.","section":"§6.2, Definition 6.2.1"}],"recommendation":"major_revision","confidential_remarks":"The two main gaps are both load-bearing and both occur in the passage from the surface case to the quasi-projective case or in the core density statement for surfaces. The surface-group technology in Sections 4–6 is detailed and promising; I believe the authors can repair the proof, e.g. by adding a multiplicative-independence condition in Lemma 3.1.8 (or invoking the Golsefidy–Tamam orbit-closure results) and by handling the -I monodromy case directly in §7.4 via the X_{g,-I} construction rather than through the flawed 'disjoint unions' assertion. I recommend requesting a revision rather than rejecting, because the main ideas are likely correct and the explicit constructions are valuable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the main thing to know: this is a serious paper. It proves potential density of integral points for relative SL2 and PGL2 character varieties of any smooth quasi-projective variety, confirming a conjecture of Simpson/Campana for these groups. The proof combines the Corlette–Simpson and Loray–Pereira–Touzet classification for rank-2 local systems on quasi-projective varieties with a genuinely new construction of integral local systems on surfaces whose mapping class group orbit is Zariski-dense.\n\nThe surface-group part is the real meat. The authors define explicit matrix families M_K, N_K, L_K and glue local systems along pants decompositions to get P-good integral points. Proposition 3.1.10 then gives Zariski density of the pure mapping class group orbit. I found this detailed and convincing. The results for X_{g,-I} in Section 6 are also new and needed for closed surfaces in PGL2.\n\nNow the soft spots. The reduction from quasi-projective Y to curves in Section 7.4 uses the assertion that relative character varieties of orbicurves are disjoint unions of relative character varieties of surface groups. The stress-test is right that this is false for SL2 with order-2 stacky points: the condition \"monodromy equals -I\" is strictly stronger than \"trace equals -2\". But the paper only uses this assertion for PGL2, where the boundary invariant does determine the conjugacy class, so the assertion is actually true there. The SL2 case is then obtained by reducing to PGL2 and lifting via Lemma 7.3.1. So the main theorem is not endangered. However, the phrasing is misleading and should be fixed. A referee should ask the authors to clarify that the orbicurve decomposition is only used for PGL2, or to handle the SL2 case directly with the X_{g,-I} technology they already developed.\n\nThe lifting step also claims that the cohomological obstruction to lifting a PGL2 point to SL2 is constant on connected components. This is plausible but not proven in the text; it would be good to justify it or give a reference.\n\nOverall, the central argument holds up. The paper deserves a serious referee. I would recommend engaging with it.","headline":"Strong new theorem on potential density of integral points for rank-2 character varieties; the orbicurve reduction is fine for PGL2, and the paper should clarify the status of the SL2 case.","tokens_in":22034,"tokens_out":7766,"would_cite":true,"duration_ms":78555,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["14G05","14D20","14H60","20H10"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that for $G=\\mathrm{SL}_{2,\\mathbb{Z}}$ or $G=\\mathrm{PGL}_{2,\\mathbb{Z}}$, integral points are potentially Zariski-dense in the relative character variety $X_{G,C}(Y)$ of any smooth quasi-projective complex variety with…","keywords":["potential density","integral points","character varieties","Betti moduli","mapping class group","SL2 local systems","quasi-projective varieties","Dehn twists"],"falsifier":"Exhibit a smooth quasi-projective variety $Y$ and a connected component of $X_{\\mathrm{SL}_2,C}(Y)$ of positive dimension whose generic representation has Zariski-dense image, has some boundary trace outside $E=\\{\\zeta+\\zeta^{-1}\\}$, and does not factor through any orbicurve; the proof in Section 7.4 asserts that no such component exists, so one explicit example would falsify the reduction. Alternatively, on the surface side, find a boundary tuple $k$ for which the relative character variety $X_{g,n,k}$ has no Zariski-dense set of $\\mathcal{O}_L$-points after any finite extension $L$, contradicting Theorem 5.0.4.","tokens_in":20920,"feed_emoji":"📐","tokens_out":11658,"duration_ms":119096,"temperature":0.7,"pith_summary":"The paper proves Conjecture 1.1.1 in the two lowest-rank cases: for $G=\\mathrm{SL}_{2,\\mathbb{Z}}$ or $G=\\mathrm{PGL}_{2,\\mathbb{Z}}$, integral points are potentially Zariski-dense in the relative character variety $X_{G,C}(Y)$ of any smooth quasi-projective complex variety $Y$, for any choice of algebraic-integer boundary traces $C$. This is the first positive-dimensional confirmation of the integrality conjecture for rank-two local systems, which previously was known only for isolated rigid representations. The proof proceeds by reduction to punctured Riemann surfaces: a classification theorem for rank-two local systems on quasi-projective varieties says that a Zariski-dense representation either factors through an orbicurve or is rigid, quasi-unipotent, and integral, so the whole problem collapses to relative character varieties of surface groups. There the authors construct, by gluing integral local systems on pairs of pants, a single integral representation whose orbit under the pure mapping class group is Zariski-dense in the relevant variety. The interest is that these Betti moduli spaces sit on the arithmetic side of non-abelian Hodge theory, where potential density of integral points was open in positive dimension except in special families of cubic surfaces.","feed_headline":"Integral points are potentially dense in rank-2 Betti moduli","feed_subtitle":"Reduces to punctured surfaces, where one integral local system has a Zariski-dense orbit.","key_machinery":"The motor of the proof is the pair consisting of pants-decomposition trace coordinates and $P$-good integral points. A pants decomposition $P=a_1\\cup\\dots\\cup a_{3g-3+n}$ of $\\Sigma_{g,n}$ gives a trace map $\\mathrm{tr}_P:X_{g,n,k}\\to\\mathbb{A}^{3g-3+n}$. A point is $P$-good when each coordinate of $\\mathrm{tr}_P$ lies outside $E=\\{\\zeta+\\zeta^{-1}:\\zeta\\text{ a root of unity}\\}$ and the fiber is perfect, meaning traces avoid $\\pm 2$ and the restriction to each pair of pants is irreducible; then a parametrization identifies the fiber with an algebraic torus $\\mathbb{G}_m^{3g-3+n}$ on which Dehn twists act by independent coordinatewise multiplications, so the twist orbit, and hence the full pure mapping class group orbit, is Zariski-dense. The arithmetic half is a gluing toolbox: the sets $M_K$ and $N_K$ of integral matrices with unit off-diagonal entries are used to build integral $\\mathrm{SL}_2(\\mathcal{O}_L)$-representations on pairs of pants, once-punctured tori, and two-holed tori that are $P$-good, and these are glued along common boundary monodromy to cover every $\\Sigma_{g,n}$ with $3g-3+n>0$. For $\\mathrm{PGL}_2$ and the closed-surface case, the same dynamics is run on $X_{g,-I}$, the relative variety of representations with monodromy $-I$ around a puncture, where the fiber is a smaller torus $\\mathbb{G}_m^{3g-3}$.","core_discovery":"The central result is Theorem 1.1.2: for $G=\\mathrm{SL}_{2,\\mathbb{Z}}$ or $\\mathrm{PGL}_{2,\\mathbb{Z}}$, a number field $K$, and a tuple $C\\in (G/\\mathrm{ad}G)(\\mathcal{O}_K)^n$ fixing the traces of monodromy around each boundary component of a simple normal crossings compactification, there is a finite extension $L/K$ such that $\\mathcal{O}_L$-points are Zariski-dense in $X_{G,C}(Y)$. For a surface $\\Sigma_{g,n}$, potential density is obtained by exhibiting an integral representation $\\rho$ that is $P$-good for a pants decomposition $P$: the traces of $\\rho$ along the pants curves avoid the set $E=\\{\\zeta+\\zeta^{-1}:\\zeta\\text{ a root of unity}\\}$ and the corresponding fiber of the trace map is perfect. A dynamical lemma then shows that the orbit of $\\rho$ under the pure mapping class group is Zariski-dense in the whole relative character variety, and since the action preserves integrality, the density of $\\mathcal{O}_L$-points follows. For arbitrary $Y$, the same statement is forced by the rank-two classification: any Zariski-dense local system is either rigid and quasi-unipotent, hence integral, or pulled back from an orbicurve, and relative character varieties of orbicurves are disjoint unions of those of surface groups. For $\\mathrm{PGL}_2$, the proof first handles the locus of representations with monodromy $-I$ around a puncture, then lifts the resulting density from $\\mathrm{PGL}_2$ to $\\mathrm{SL}_2$ through a finite morphism.","pith_inferences":["The gluing construction depends on infinitely many units in the ring of integers; a natural test is whether enlarging a field with finite unit group, such as an imaginary quadratic field, defeats the method, and whether some other source of integral matrices then supplies the same density.","If analogous unit-matrix sets exist for $\\mathrm{SL}_n$, the same pants-decomposition dynamics would suggest potential density for higher-rank Betti moduli on punctured surfaces, with the main difficulty being an $n$-dimensional analogue of the $P$-good conditions.","One could test the orbicurve reduction directly on a concrete orbicurve with nontrivial generic stabilizer, such as a weighted projective line, to see how the disjoint-union structure of its relative character varieties behaves; the paper treats this case only through the classification theorems."],"forward_implications":["Conjecture 1.1.1 is settled for $G=\\mathrm{SL}_{2,\\mathbb{Z}}$ and $G=\\mathrm{PGL}_{2,\\mathbb{Z}}$: every relative character variety of either group, with arbitrary algebraic-integer boundary data on any smooth quasi-projective $Y$, has potentially Zariski-dense integral points.","This supplies the first positive-dimensional cases of the rank-two integrality conjecture for local systems, going beyond isolated rigid representations.","For surfaces, the construction works over a biquadratic extension of the field generated by the boundary traces, and the cubic $x^2+y^2+z^2-xyz=3$ shows that some field extension is genuinely necessary.","Because these relative character varieties are log Calabi-Yau, the theorem confirms the expected potential density of integral points for log Calabi-Yau varieties in these cases."],"supporting_citations":[{"why":"Supplies the classification input: Zariski-dense rank-two local systems on quasi-projective varieties either factor through an orbicurve or are rigid with quasi-unipotent monodromy, and the latter are integral.","marker":"[CS08]"},{"why":"Extends the classification to the non-quasi-unipotent boundary case, closing the reduction to orbicurves in Theorem 1.1.2.","marker":"[LPT16]"},{"why":"Provides the torus parametrization of perfect fibers and the Dehn-twist action that makes $P$-good points have Zariski-dense mapping class group orbits.","marker":"[Wha20a]"},{"why":"Used to show that the map from the $\\mathrm{SL}_2$ relative character variety to its $\\mathrm{PGL}_2$ quotient is finite, so density lifts from $\\mathrm{PGL}_2$ to $\\mathrm{SL}_2$.","marker":"[Cot24]"}],"fun_headline_variants":["Integral points potentially dense in rank-2 Betti moduli","Potential integral density for SL2 local systems on quasi-projective varieties","Reduction to punctured surfaces proves potential integral density","Zariski-dense integral orbit under mapping class group","Quasi-projective SL2 character varieties: potential density of integral points"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof rests on a cited classification theorem saying that every Zariski-dense rank-two local system on a smooth quasi-projective variety either factors through an orbicurve or is rigid, quasi-unipotent, and already integral; if that classification has a counterexample, the reduction of the whole problem to punctured surfaces fails.","fun_headline_variants_meta":{"raw":{"variants":["Integral points potentially dense in rank-2 Betti moduli","Potential integral density for SL2 local systems on quasi-projective varieties","Reduction to punctured surfaces proves potential integral density","Zariski-dense integral orbit under mapping class group","Quasi-projective SL2 character varieties: potential density of integral points"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000941,"raw_usage":{"total_tokens":4041,"prompt_tokens":982,"completion_tokens":3059,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":598,"completion_tokens_details":{"reasoning_tokens":2972}},"tokens_in":598,"tokens_out":3059,"duration_ms":25769,"temperature":1.0,"reasoning_tokens":2972,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T21:34:23.950277+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a smooth quasi-projective variety $Y$ and a connected component of $X_{\\mathrm{SL}_2,C}(Y)$ of positive dimension whose generic representation has Zariski-dense image, has some boundary trace outside $E=\\{\\zeta+\\zeta^{-1}\\}$, and does not factor through any orbicurve; the proof in Section 7.4 asserts that no such component exists, so one explicit example would falsify the reduction. Alternatively, on the surface side, find a boundary tuple $k$ for which the relative character variety $X_{g,n,k}$ has no Zariski-dense set of $\\mathcal{O}_L$-points after any finite extension $L$, contradicting Theorem 5.0.4.","supporting_citations":[],"review_version":1}