{"id":"e7ab544f-1adf-41cf-878d-07bfcd17f40b","arxiv_id":"1908.07014","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A subgroup of GL_n over the function field F_p(t), with simple connected Zariski closure and full adjoint trace field, has uniform spectral gap along congruence quotients by square-free polynomials whose factor degrees have no small prime factors.","lead":"This paper proves a new expansion result: matrix groups over the function field F_p(t), reduced modulo admissible square-free polynomials, yield families of expander graphs under Zariski-simplicity and a trace-field condition. It advances super-approximation beyond the SL_2 case to general linear groups in positive characteristic.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 60's subfield-type bound appears to rely on an unjustified quotient-of-free-group estimate: equation (87) needs injectivity of the projection on the relevant random-walk ball, and the final exponent in (90) is not shown to be uniform in the degree ratio.","rationale":"The reader's weakest assumption is the trace-field hypothesis, which is a genuine input condition but is stated as a hypothesis of the theorem and is not where I find the most load-bearing soft spot. The reader does list Lemma 60 in the rationale as terse, but the primary concern I identify is more specific: the proof of Lemma 60 needs to transfer Kesten's free-group decay through a quotient map, and the displayed inequalities as written do not justify that transfer or the uniformity of the resulting exponent. This concern is internal to the proof and directly affects the escape-from-arbitrary-subgroups step that completes the proof of Theorem 1. I do not claim the theorem is false; the gap may be repairable by an explicit injectivity argument on the random-walk ball and by a geometric-mean bound instead of the final min bound. For that reason the appropriate verdict remains the same as the reader's: CONDITIONAL, pending a rigorous repair of Lemma 60. The concern is concrete and checkable by redoing the size comparisons with explicit constants.","tokens_in":1090,"tokens_out":1088,"duration_ms":433036,"concrete_test":"Re-derive Lemma 60 with explicit constants in the extreme case deg f1 >> deg f2, say deg f1 = N and deg f2 = 2. Write out equations (87), (88), and (90) with the Kesten factor q^{-c deg f2} and the subfield-size factor |pi_{f2}|^{-1/c0'}; then compute whether the final exponent eta can be chosen independently of N. In the same calculation, verify the injectivity condition deg f2 > C_{Omega'} * l0 for the chosen l0 <<_Omega deg f2; if collisions in pi_f2 restricted to B_{2*l0}(Omega') can occur at the selected length, the Kesten transfer in (87) fails, and if the resulting eta tends to 0 as N grows, the uniform escape constant required by Theorem 58 does not follow from the displayed inequalities.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 1 is reduced by Theorem 58 to the escape estimate in Lemma 60. In Lemma 60, an arbitrary proper subgroup H is split into a structural part H1 and a subfield-type part H2. The subfield part is handled by equations (87)-(88): the probability that the l-step random walk lands in H2 is bounded by |pi_f1(Gamma)|^{-c1}|H2|, with Kesten's free-group maximal atom supplying the exponential factor. The problem is that the random walk has been pushed forward through the homomorphism pi_f2 from the free group Gamma' to the finite group pi_f2(Gamma); Kesten's bound controls the free-group measure, not its pushforward. Many reduced words can collapse to the same quotient element, so the maximal atom of the pushed measure can be much larger than the free-group atom. To justify (87), one must prove that pi_f2 is injective on the ball B_{2*l0}(Omega'), which follows only if deg f2 > C_{Omega'} * l0; the paper never states or proves this, and l0 is only said to satisfy l0 <<_Omega deg f2. Even granting injectivity, the constant c1 in (87) is not uniform: it depends on the ratio deg f2/deg f1, since it converts a decay of size q^{-c deg f2} into a power |pi_f1(Gamma)|^{-c1}. The final display in (90) then replaces the preceding geometric-mean bound by |pi_f1*f2(Gamma)|^{-min(delta0,c1/4)}; because c1 is not bounded below independently of H, the required single eta and C0 for Theorem 58 are not established. Thus the step where subfield-type subgroups are neutralized by the admissibility condition is not rigorously justified as written, and the central uniform spectral gap conclusion is not fully supported by the proof in the manuscript.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims a super-approximation theorem over a global function field: for a finitely generated subgroup Γ of GL_{n0}(F_p[t, 1/r0(t)]), p > 5, whose Zariski closure is connected, simply-connected, absolutely almost simple, and whose adjoint trace field is F_p(t), the Cayley graphs of Γ modulo square-free polynomials whose irreducible factors have degrees with no small prime factors form a family of expanders. The proof follows the Bourgain–Gamburd–SGV12–Varjú architecture: Weisfeiler strong approximation gives the product structure of congruence quotients; a Larsen–Pink dichotomy classifies proper subgroups as structural or subfield type; new positive-characteristic results (Propositions 9, 23, 28) refine the algebraic description; a modified Varjú multi-scale theorem (Proposition 33) supplies the product theorem; and Proposition 6 gives escape from purely structural subgroups. The final Lemma 60 is meant to pass from structural subgroups to arbitrary proper subgroups using the admissibility condition on degrees.","tokens_in":52608,"tokens_out":8237,"duration_ms":81998,"significance":"If the main theorem were correct, it would be the first higher-rank super-approximation result over F_p(t), going beyond Bradford's SL_2 result with prime degree factors, and it would extend the SGV12/Varjú machinery to positive characteristic with new subtleties (non-complete reducibility, Weil-restriction obstructions, subfield subgroups). The paper contains genuinely useful technical contributions that appear internally sound: the field-descent proof of Proposition 9, the subfield-intersection result Proposition 23/Corollary 25, and the invariant-theoretic description of positive-dimensional subgroups via affine representations in Proposition 28. However, the central final step that carries the theorem from structural subgroups to arbitrary subgroups is not justified as written, and the proof of the main theorem therefore currently fails.","major_comments":[{"comment":"The estimate (87) applies Kesten's free-group bound directly to the pushforward measure π_{f2}[P_{Ω'}^{(2l0)}] without any control of the quotient map. Many reduced words collapse to the same element in π_{f2}(Γ), so the maximal atom of the pushforward can be much larger than the free-group atom; to justify (87) one needs injectivity of π_{f2} on the relevant ball B_{2l0}(Ω'), which would require a bound such as deg f2 ≫ l0, but the paper neither states nor proves such an inequality, and the condition l0 ≪_Ω deg f2 does not imply it. Moreover, the passage from a free-group decay e^{-c l0} to a power |π_{f1}(Γ)|^{-c1} is dimensionally unmotivated unless deg f1 and deg f2 are related; the c0-admissibility condition imposes no relation between these two degrees, and f2 can have much larger degree than f1. Consequently the bounds (88)–(90) do not yield the uniform η needed for Theorem 58, and Lemma 60, which is precisely the step that converts escape from structural subgroups into escape from arbitrary proper subgroups, is not established.","section":"§5.2, Lemma 60, Eqs. (87)–(90)"},{"comment":"Proposition 31 is the only place where the exponential escape from proper algebraic subgroups is obtained, yet no proof is supplied: the text says 'See proof of [SGV12, Proposition 20]' and §3.4 explicitly states that the proofs of Varjú and SGV12 are not repeated. The present setting involves local fields of positive characteristic, a fresh list of representations from Proposition 30, and affine actions without fixed points; it is not automatic that the ping-pong argument from characteristic zero carries over verbatim. Since Proposition 31 feeds directly into Proposition 6 and hence into the main theorem, a self-contained proof or at least a precise reduction to the cited statement is required.","section":"§3.4, Proposition 31"},{"comment":"The verification of Varjú's condition (V3)L for the structural families is sketched rather than proved. The text defines families H_i for subgroups of fixed dimension and bounded complexity, then says that for smaller dimensions one 'allow[s] slightly larger complexity to include the connected components of the intersections of larger dimension connected proper subgroups.' No argument is given that the required index bound [H1 : H1∩H2] < L holds uniformly, that the families remain closed under conjugation with a uniform L, or that the hierarchy index j < i can be maintained. This is a load-bearing assumption for Theorem 58, and it needs a proof, not a heuristic.","section":"§5.1, verification of (V3)L"}],"minor_comments":[{"comment":"The abstract writes π_{f(x)}(Γ) while the body consistently uses π_f(t); please make the notation uniform.","section":"Abstract and §1.1"},{"comment":"The running header 'TOW ARDS SUPER-APPROXIMATION' contains a typo: 'TOWARDS' should be a single word.","section":"Title page"},{"comment":"Theorem 7 is stated for subgroups of G^{Che}_{0,p}(F_p), but Definition 8 refers to φ_ℓ(H) and G^{Che}_p(K(ℓ)) without defining this latter symbol; align the notation between the theorem and the definition.","section":"§2.2, Theorem 7 and Definition 8"},{"comment":"In the final paragraph, the expression F_p[s(t)/r(t)]/⟨ℓ⟩ is used, but the ring should be written with the same parentheses as elsewhere, e.g. F_p[s(t)/r(t)] or Fp[s/r]; the intended meaning is clear but the notation should be cleaned up.","section":"§3.5, proof of Proposition 6"},{"comment":"The text states that by Proposition 6 there is a symmetric subset Ω' of Γ and then, in (84), π_f(⟨Ω'⟩)=π_f(Γ). To pass from spectral gap for Ω' to Theorem 1' for Ω, the equivalence of spectral gap under change of generators should be explicitly invoked, since the Cayley graphs in Theorem 1' are defined using π_f(Ω).","section":"§5, beginning"}],"recommendation":"reject","confidential_remarks":"The paper contains valuable algebraic refinements (Propositions 9, 23, 28) and a coherent overall strategy, but the proof of Theorem 1 fails at Lemma 60: the Kesten-type estimate for the pushforward measure onto subfield-type factors is not justified, and the exponent c1 is not uniform in the degree ratio. This is not a cosmetic issue; it is the only place where arbitrary proper subgroups are handled, and the suggested fixes would require a genuinely new argument. Given the manuscript's own admission that escaping subfield subgroups is not known, I do not think the gap can be repaired within the current scope, which motivates rejection rather than major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper proves a real new theorem—first super-approximation at the GL_n level over a global function field, with expansion for c0-admissible moduli. It extends Bradford's SL_2 result and pushes toward Lubotzky's conjecture in positive characteristic. The new positive-characteristic tools (subfield containment in Proposition 9, adjoint subgroup rigidity in Proposition 23, affine representations forced by non-complete reducibility in Proposition 28, and the modified Varju product theorem) are valuable and likely of independent interest. Credit where due: the authors are also honest about what they cannot do—they neutralize subfield-type subgroups by admissibility rather than escaping them, and they flag the trace-field hypothesis as a genuine input condition.\n\nNow the soft spot. The stress-test note is on target. Lemma 60 is the step that turns escape from purely structural subgroups into escape from arbitrary proper subgroups, and its equation (87) does not hold as written. Kesten's bound controls the random walk on the free group; the measure pushed forward to π_{f2}(Γ) can have much larger atoms because many reduced words collapse to the same quotient element. To conclude ||P^{(2l0)}_{π_f2(Ω')}||_∞ ≤ |π_f1(Γ)|^{-c1}, you need the projection to be injective on the relevant ball, which requires deg f2 ≫ deg f1 relative to l0. The paper only states l0 ≪ deg f2, so if deg f1 is significantly larger than deg f2, the bound fails. And the constant c1 in (87) depends on the degree ratio; the final exponent min(δ0, c1/4) in (90) is not shown to be uniform, which is exactly what Theorem 58 needs. Without Lemma 60, the proof of Theorem 1 does not close. This is the main obstruction.\n\nLesser concerns: Proposition 31 is imported from SGV12 with a terse transfer, but that is standard practice and less worrying. The trace-field hypothesis is restrictive, but it is stated plainly and is a known failure mode; that is an assumption, not a flaw.\n\nWho is this for? Specialists in expander graphs, arithmetic groups, and positive-characteristic algebraic groups. It deserves a serious referee; the result is too important to desk-reject. But the referee should be asked to scrutinize Lemma 60 carefully. If the gap is real, a major revision is needed. If it is patchable, the paper could become a solid contribution.","headline":"Genuine new result, serious machinery, but the proof has a load-bearing gap in Lemma 60 that the stress-test correctly identifies.","tokens_in":53382,"tokens_out":6547,"would_cite":true,"duration_ms":68233,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["22E40","20G30","05C81"],"pacs":[],"model":"deepseek-v4-flash","headline":"Under a Zariski-density and trace-field condition, the Cayley graphs of a finitely generated subgroup of $\\mathrm{GL}_{n_0}(\\mathbb{F}_p(t))$ modulo square-free admissible polynomials form an expander family, equivalently the spectral gap…","keywords":["super-approximation","expander graphs","positive characteristic","function fields","Cayley graphs","spectral gap","strong approximation","subfield-type subgroups"],"falsifier":"Take any concrete $\\Gamma$ for which the trace-field condition can be verified, choose a generating set $\\Omega$, and compute the second-largest eigenvalue of the averaging operator on $\\mathrm{Cay}(\\pi_f(\\Gamma),\\pi_f(\\Omega))$ for square-free $f$ whose irreducible factors have pairwise distinct degrees $>1$ with no prime factor below a chosen $c_0$; if the eigenvalues tend to 1 along some sequence, the uniform spectral gap asserted by the theorem fails.","tokens_in":52021,"feed_emoji":"🎲","tokens_out":10655,"duration_ms":107618,"temperature":0.7,"pith_summary":"This paper establishes a super-approximation theorem for finitely generated subgroups of the general linear group over a rational function field $\\mathbb{F}_p(t)$ with $p>5$. Provided the Zariski closure is connected, simply connected and absolutely almost simple, and provided the traces of the adjoint representation of the group generate the whole field $\\mathbb{F}_p(t)$, the congruence quotients modulo square-free polynomials with suitably restricted irreducible factors form expander graphs. In spectral terms, the random-walk operator on each quotient has a second-largest eigenvalue uniformly bounded away from 1. This is the first function-field result of this kind beyond the $\\mathrm{SL}_2$ case, and it isolates what must be added to the characteristic-zero method when subfield-type subgroups appear.","feed_headline":"Cayley graphs of F_p(t)-subgroups expand uniformly","feed_subtitle":"A trace-field condition makes congruence quotients of F_p(t)-subgroups into expanders.","key_machinery":"The load-bearing mechanism is a dichotomy for proper subgroups of the finite quotients. A subgroup is structural if it lies in the $K(\\ell)$-points of a proper algebraic subgroup of bounded complexity, and subfield type if, after passing to the adjoint group, it sits between the commutator subgroup and the full points of a model over a proper subfield. The trace-field hypothesis makes strong approximation applicable, so $\\pi_f(\\Gamma)$ splits as a product of the $G_\\ell(K(\\ell))$, and this is where that hypothesis enters. Structural subgroups are escaped through invariant-theoretic descriptions: every proper positive-dimensional algebraic subgroup either fixes a line in one of finitely many irreducible representations or fixes a point in one of finitely many affine representations, the latter needed because wedge powers of the adjoint representation need not be completely reducible in positive characteristic. A ping-pong argument produces a free generating set whose random walk hits any proper algebraic subgroup with exponentially small probability. Subfield-type subgroups are kept small by the admissibility condition on degrees, and a modified multi-scale product theorem converts the escape bounds into a uniform spectral gap.","core_discovery":"The main result is Theorem 1. Let $\\Omega$ be a finite symmetric subset of $\\mathrm{GL}_{n_0}(\\mathbb{F}_p[t,1/r_0(t)])$ with $p>5$, and let $\\Gamma=\\langle\\Omega\\rangle$. If the Zariski closure of $\\Gamma$ is connected, simply connected and absolutely almost simple over $\\mathbb{F}_p(t)$, and if the field generated by $\\mathrm{Tr}(\\mathrm{Ad}(\\Gamma))$ equals $\\mathbb{F}_p(t)$, then there is a constant $c_0$ such that the supremum of $\\lambda(P_\\Omega; \\prod_i \\mathrm{GL}_{n_0}(\\mathbb{F}_p[t]/\\langle\\ell_i\\rangle))$ over all irreducible polynomials $\\ell_i$ not dividing $r_0$, with degrees strictly increasing, larger than 1, and with no prime factor below $c_0$, is strictly less than 1. The equivalent formulation Theorem 1' states that the Cayley graphs $\\mathrm{Cay}(\\pi_f(\\Gamma),\\pi_f(\\Omega))$ form an expander family as $f$ ranges over square-free $c_0$-admissible polynomials. The conclusion is explicitly about uniformity: the spectral gap does not deteriorate as the modulus grows.","pith_inferences":["The trace-field hypothesis is probably stronger than necessary; a condition stated in terms of the profinite closure of $\\Gamma$ might be the right replacement, and the paper's Question 5 points toward such a formulation.","If the affine representations in the invariant-theoretic step are never actually needed, as Question 29 asks, the escape argument would simplify and might allow the admissibility restriction to be relaxed.","The multi-scale product theorem could be tested independently of function fields on products of non-isomorphic finite simple groups, e.g. $\\mathrm{PSL}_2(q_i)$ with distinct $q_i$, to see whether the growth exponent can be made effective.","The paper suggests that subfield-type subgroups, rather than approximate subgroups, are the main obstruction to super-approximation over fields with nontrivial subfields; a similar phenomenon should appear in other global fields."],"forward_implications":["For every admissible modulus $f$, the Cayley graph $\\mathrm{Cay}(\\pi_f(\\Gamma),\\pi_f(\\Omega))$ has a spectral gap bounded below by a constant depending only on $\\Gamma$, not on $f$.","Choosing moduli whose irreducible factors have degrees with large prime factors yields explicit expander families of size growing like $|f|^{\\dim G}$.","The trace-field condition, not just the Zariski closure, is what determines whether super-approximation holds in this setting; equal Zariski closures are no longer sufficient.","The paper's Section 4 product theorem applies to any family of pairwise non-isomorphic finite groups satisfying the given axioms, giving a standalone growth statement for direct sums of such groups.","The admissibility restriction is a genuine feature of the proof: it is exactly the condition that neutralizes subfield-type subgroups."],"supporting_citations":[{"why":"Supplies the strong approximation theorem that identifies $\\pi_f(\\Gamma)$ with the product of the $G_\\ell(K(\\ell))$, the starting point of the proof.","marker":"[Wei84]"},{"why":"Provides the dichotomy classifying finite subgroups as structural or subfield type, which the paper refines for the quotients.","marker":"[LP11]"},{"why":"Gives the characteristic-zero expansion method and the example showing that equal Zariski closures do not force equal super-approximation, motivating the trace-field hypothesis.","marker":"[SGV12]"},{"why":"Supplies the multi-scale product theorem and entropy framework that Section 4 adapts to groups with subfield-type subgroups.","marker":"[Var12]"},{"why":"Classifies approximate subgroups of linear groups, used to verify the product-growth axiom for the finite quotients.","marker":"[BGT11]"},{"why":"Provides the growth theorem for approximate subgroups of finite simple groups of Lie type used to satisfy condition (V4).","marker":"[PS16]"},{"why":"Gives quasirandomness bounds for finite simple groups, used to turn return-probability estimates into spectral-gap estimates.","marker":"[Gow08]"},{"why":"Gives exponential decay of return probabilities for random walks on free groups, used in the escape-from-subgroups argument.","marker":"[Kes59]"}],"fun_headline_variants":["Super-approximation in char p from a trace field condition","Cayley graphs from F_p(t) subgroups are expanders","Trace field condition forces uniform expansion in char p","Function-field super-approximation yields expanders","Admissible moduli yield expanders for F_p(t) subgroups"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof depends on the assumption that the traces of the adjoint representation of $\\Gamma$ generate the whole field $\\mathbb{F}_p(t)$; if that fails, the congruence quotients may not have the product structure the argument needs.","fun_headline_variants_meta":{"raw":{"variants":["Super-approximation in char p from a trace field condition","Cayley graphs from F_p(t) subgroups are expanders","Trace field condition forces uniform expansion in char p","Function-field super-approximation yields expanders","Admissible moduli yield expanders for F_p(t) subgroups"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000861,"raw_usage":{"total_tokens":3795,"prompt_tokens":1062,"completion_tokens":2733,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":678,"completion_tokens_details":{"reasoning_tokens":2651}},"tokens_in":678,"tokens_out":2733,"duration_ms":18649,"temperature":1.0,"reasoning_tokens":2651,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:30:56.000450+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take any concrete $\\Gamma$ for which the trace-field condition can be verified, choose a generating set $\\Omega$, and compute the second-largest eigenvalue of the averaging operator on $\\mathrm{Cay}(\\pi_f(\\Gamma),\\pi_f(\\Omega))$ for square-free $f$ whose irreducible factors have pairwise distinct degrees $>1$ with no prime factor below a chosen $c_0$; if the eigenvalues tend to 1 along some sequence, the uniform spectral gap asserted by the theorem fails.","supporting_citations":[],"review_version":1}