{"id":"ab4d577d-83d7-4f8f-8625-5d2b03d279cf","arxiv_id":"2608.00755","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Finite simple groups of Lie type of bounded rank are uniformly expanding except for a density-zero set of primes, and the exceptional generating pairs have zero dimension for every prime power.","lead":"This paper proves that the Cayley graphs of almost all finite simple groups of Lie type of bounded rank are expanders, with tight control on the exceptional groups. It also proves a 'dimension zero' theorem: for every prime power q, all but q^δ conjugacy classes of generating pairs give expander Cayley graphs.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Uniform expansion hinges on the characteristic-uniform anti-concentration theorem of companion [5]; if its constants degrade with p, (4.1) fails and the O(q^δ) exception count in Theorem 1.2 collapses.","rationale":"The reader's weakest assumption and my most load-bearing concern coincide: the uniform, characteristic-independent anti-concentration bounds from the companion preprint [5] are essential to both Theorem 1.1 (via §3.4) and Theorem 1.2 (via §4 Step 2). I read the paper in good faith and did not find an internal inconsistency in the reduction arguments as presented; the exceptional-prime and exceptional-pair limitations are honestly stated in Remark (a), and the method's reliance on effective bounds from [4] is also disclosed. However, the dimension-zero theorem's quantitative conclusion O(q^δ) depends quantitatively on the constant c_G in (4.1). If that constant were characteristic-dependent, the fat-type count in Step 5 would no longer be controlled by q^δ, and the theorem's central claim would fail. Since [5] is not part of this submission and no proof of it is reproduced, the correct epistemic status is CONDITIONAL: plausible but unverified. This does not change the reader's verdict, so I recommend UNCHANGED. The decisive verification is to check, inside [5], whether the exponential rate is uniform in characteristic; the numerical test I propose would be a quick falsification probe in the simplest case, while a full re-derivation of [5, Theorem 1.3] is the analytic settlement.","tokens_in":22882,"tokens_out":7566,"duration_ms":91858,"concrete_test":"Obtain [5, Theorem 1.3] and trace its proof for an absolutely almost simple algebraic group G over F_p^al, isolating every constant that enters the exponential rate c in P(w(a1,...,am)∈V) ≪_deg V 2^{-cℓ}. Determine whether c can be chosen independent of p when V = V_M^(2) and M is fixed. In particular, check whether the diophantine/height-gap estimates used in [5] acquire any factor depending on p; if c_p → 0 as p → ∞, then (4.1) has p-dependence and Theorem 1.2's exponent δ is not uniform. A secondary computational check: estimate the empirical rate for a fixed word walk on SL2(F_p) into a fixed proper curve for p ≈ 1009 and p ≈ 10^6; if the rate decays like 1/log p, characteristic-uniformity is already false in the simplest case.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 1.2 and, in the prime case, Theorem 1.1 depends on the companion result [5, Theorem 1.3], quoted in §3.4 and §4 Step 2 as (4.1): for a Zariski-dense pair (a,b), P((w1(a,b),w2(a,b)) ∈ V_M^(2)) ≤ C_M 2^{-c_G ℓ}, with c_G > 0 independent of the characteristic. This estimate is applied to force every positive-dimensional fat ℓ-type to lie in V_M^(2) (Lemma 4.5), to handle structural and subfield subgroups (Lemmas 4.6 and 4.8), and then to bound the number of exceptional conjugacy classes by 2^{10drℓ}·log(q)/log(p). If the constant c_G in [5] turned out to depend on p, say c_p ≈ 1/log p, then the choice ℓ ≍ κ c^{-1} log q in Step 5 would only control fat types up to q^{O(δ c/c_p)}, which can exceed q^δ for large p. The claimed O(q^δ) count of exceptions would fail, and so would the non-concentration step that feeds Bourgain–Gamburd. The paper itself acknowledges in Remark (d) that this anti-concentration is a necessary step, so the dependency is explicit. No internal inconsistency in the present manuscript is apparent; the concern is that the central claim rests on an unverified external theorem with exactly the uniformity that the argument requires.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper addresses the Lubotzky–Weiss uniform expansion conjecture for finite simple groups of Lie type. Theorem 1.1 proves uniform expansion for all finite simple groups G(p^k) of bounded rank with k bounded, after excluding a density-zero (indeed X^δ-sized) set of primes p. Theorem 1.2 proves a “dimension zero” refinement: for every prime power q, all but O_{r,δ}(q^δ) conjugacy classes of generating pairs of G(q) are ε-expanding. The proof combines the Bourgain–Gamburd method with quasirandomness, product theorems, Larsen–Pink/Nori subgroup structure, and a new counting of “types” of subvarieties cut out by word-map equations. The two principal new inputs are a bad-reduction control theorem (Section 2, relying on the companion paper [4]) and a characteristic-uniform anti-concentration bound for random walks on algebraic subvarieties (Sections 3.4 and 4, relying on the companion preprint [5]).","tokens_in":23216,"tokens_out":14254,"duration_ms":167459,"significance":"If the result is valid, it constitutes major progress on a long-standing conjecture: it reduces the uniform expansion conjecture to at most O(q^δ) exceptional conjugacy classes, and proves uniform expansion outside a density-zero set of primes for all bounded-rank groups, extending the earlier SL_2 results. The paper is carefully structured and explicit about parameter dependencies, and it honestly flags the two companion dependencies. Strengths include the type-counting lemma (Lemma 2.3), the clean reduction to structural and subfield subgroups, and the explicit use of characteristic-uniform anti-concentration. The main caveat is that the two external dependencies are load-bearing and one is labelled “in preparation”; as it stands, the theorems are conditional on unverified companion results.","major_comments":[{"comment":"The non-concentration estimate (4.1) is quoted from the companion preprint [5, Theorem 1.3] and is used to prove Lemmas 4.5–4.8 and hence the O(q^δ) exception count in Step 5. The constant c_G is asserted to be independent of the characteristic, and this uniformity is exactly what the proof requires. If c_G depended on p, say c_G ~ 1/log p, then the choice ℓ ≍ 4κ c^{-1} log q in Step 5 would only control fat ℓ-types up to q^{O(δ c/c_p)}, which can exceed q^δ; Lemma 4.5 would no longer give the required dichotomy, and the Bourgain–Gamburd non-concentration step would collapse. The same theorem is used in §3.4 to obtain the non-concentration estimate (3.2) for Theorem 1.1. The paper’s Remark (d) acknowledges this is a necessary step, but the main theorems are not self-contained. Please either include a proof of [5, Theorem 1.3] or restate the main theorems as conditional on it.","section":"§3.4 and §4 Step 2, Eq. (4.1)"},{"comment":"Proposition 2.4, the bad-reduction control that defines the exceptional set B_δ in Theorem 3.1, is proved using Theorem 2.1 and Proposition 2.2, both quoted from the companion paper [4], labelled “in preparation”. The bounds on log Δ and the resulting count of bad primes are the numerical engine behind the density-zero statement (3.1). Without the proof of [4], the argument for Theorem 1.1 is incomplete. The dependency is admitted, but a referee cannot verify the central claim from the text. The authors should either include the needed results from [4] in this paper or clearly state that the theorems are conditional on [4].","section":"§2, Theorem 2.1 and Proposition 2.2"},{"comment":"To rule out fat diagonal ℓ-types lying in W_M, the text asserts that if ((a,a^{F0}),(b,b^{F0})) ∈ W_M while ⟨a,b⟩ = G, then G ≤ {g : g^{F0} = α(g)} for some α ∈ Aut G, and that “this is easily seen to be impossible” because F is a proper power of F0. This is load-bearing for the subfield-subgroup exception count, but the impossibility is not demonstrated. A proof can likely be supplied (for example, using Zariski density of G(F_q) and the fact that a nontrivial Frobenius is not an inner automorphism of the algebraic group), but as written the reader is asked to accept a nontrivial assertion. Please expand this argument.","section":"§4 Step 4, “Further note” after Lemma 4.8"}],"minor_comments":[{"comment":"There is a typo: “bytheabove” should be “by the above”. Also the notation d_0^{1+2+···+s} is hard to parse; please clarify the exponent.","section":"Lemma 2.3 proof"},{"comment":"The phrase “super-exponential bound exp(ℓ^{O_r(1)})” appears to understate the double-exponential bound on bad primes that follows from Prop. 2.4 (log Δ ≪ |E|^{r+1} d_0^{O_r(1)} h_0 with |E| exponentially large in ℓ). Since only the number of prime divisors of Δ matters for the density-zero count, this is cosmetic, but the statements should be aligned.","section":"Introduction, Remark (a)"},{"comment":"The reduction to [23, Corollary F] is used without stating the corollary. Please include its precise hypotheses, since the reader otherwise has to consult another paper to verify the conditions on the unipotent radical and the semisimple quotient.","section":"Theorem 3.11 / Remark 3.12"}],"recommendation":"major_revision","confidential_remarks":"Both major concerns are about external dependencies rather than internal errors. In this area it is acceptable to rely on companion papers, but here the two companions carry two of the three key ingredients (bad reduction and anti-concentration) and one is still “in preparation”. I would advise the editor to require that the authors supply those manuscripts, or clearly mark the main theorems as conditional on them, before publication. The mathematics that is included in the present text is coherent and promising."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Let me give you the short version: this is a serious paper by serious people, and the main theorems are exactly what the field has been waiting for since [14]. Theorem 1.1 extends strong uniform expansion from SL(2,p) to all bounded-rank finite simple groups of Lie type, excluding a density-zero prime set. Theorem 1.2 is genuinely new even for SL2: all but q^δ conjugacy classes of generating pairs give ε-expanders, for every prime power q. The dimension-zero phrasing is the right way to state it, and the exception count is uniform in q and p. The treatment of subfield subgroups—including Suzuki–Ree and triality groups—is uniform rather than case-by-case, which is a real improvement over [16].\n\nThe proof has a clear architecture: reduce non-concentration on subgroups to a statement about word maps and algebraic subvarieties, control the number of 'types' exponentially via Lemma 2.3, then use bad-reduction control and anti-concentration to rule out all but few classes. The exponential type bound is a neat trick, and the paper is honest about its own limitations—Remark (a) explicitly says exceptional primes cannot be ruled out with current bounds on bad reduction.\n\nThe soft spot is exactly what the stress-test note identifies. The argument hinges on the uniform-in-characteristic anti-concentration theorem from companion [5], quoted as (4.1). If its constant c_G degraded with p, the choice of ℓ in Step 5 would only give control up to q^{O(δ c/c_p)}, and the O(q^δ) count would collapse. The paper itself flags in Remark (d) that these anti-concentration bounds are necessary, so no one is hiding the dependency. But 'necessary' and 'verified' are different things. [4] is 'in preparation', [5] is a preprint; until those are publicly available and checked, the main theorems should be treated as conditional. I don't see an internal inconsistency in the present manuscript; the argument is coherent and the cited external results are plausible. But this is a textbook case where the verdict depends on companion papers.\n\nWho is this for: anyone working on expanders in finite groups, super-strong approximation, or random walks on algebraic groups. It deserves a serious referee, but the referee should be explicitly asked to check [5] and [4] as part of the process, or the companion papers should be posted and verified first.","headline":"Proves the expected uniform expansion theorem for bounded-rank groups of Lie type outside a density-zero prime set, plus a dimension-zero exception count; the catch is that the main load-bearing estimate is delegated to an unpublished companion.","tokens_in":23730,"tokens_out":1919,"would_cite":true,"duration_ms":21927,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["20G40","05C48","20F65","11B30"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that, for finite simple groups of Lie type of bounded rank, all but a zero-dimensional set of generating pairs produce expander Cayley graphs with a uniform expansion constant, and for prime fields all except a density-zero","keywords":["uniform expansion","Cayley graphs","finite simple groups of Lie type","expanders","Bourgain-Gamburd method","anti-concentration","character varieties","dimension zero"],"falsifier":"Take SL_2(p) (or any fixed-rank family) and, for a sequence of primes p, enumerate all conjugacy classes of generating pairs (a,b) and compute the non-concentration probability µ^ℓ_S(H) for the largest proper subgroup H at ℓ = floor(C log p) for some fixed C. If for some C and infinitely many p there are more than p^δ conjugacy classes with µ^ℓ_S(H) > e^{-cℓ} for every fixed c>0, then the key estimate (4.2) fails and the dimension-zero theorem collapses. A direct check of the characteristic-uniformity assumption would be to compute, for the subvariety x^p-x=0 inside SL_2 over F_p, whether the","tokens_in":22759,"feed_emoji":"🎲","tokens_out":8072,"duration_ms":88791,"temperature":0.7,"pith_summary":"The paper takes up the question, posed by Lubotzky and Weiss, whether the full family of Cayley graphs of finite simple groups of Lie type of bounded rank forms a family of expanders with a uniform spectral gap. It answers this up to a negligible set of exceptions: for every fixed rank r and every δ>0, all but O(q^δ) conjugacy classes of generating pairs of any such group G(q) yield ε-expander Cayley graphs, with ε depending only on r and δ; and when q=p is prime, expansion holds for every prime outside a set of density zero. This reduces the uniform expansion conjecture to ruling out a 'zero-dimensional scar' set. The proof extends the Bourgain–Gamburd method to all characteristics and all Lie-type groups by combining new uniform anti-concentration estimates for random walks on algebraic subvarieties with effective arithmetic-nullstellensatz control of reduction modulo primes, plus a counting argument showing that only exponentially many algebraic 'types' can obstruct expansion. If correct, the result makes expansion the generic phenomenon for these groups and opens the door to positive-characteristic super-strong approximation.","feed_headline":"All but q^δ pairs of generators of Lie-type groups expand","feed_subtitle":"A uniform spectral gap for Lie-type simple groups, with exceptions confined to a zero-dimensional 'scar' set.","key_machinery":"The engine is the Bourgain–Gamburd method, which proves a spectral gap from three inputs: quasi-randomness, a product theorem, and exponential non-concentration of the random walk on proper subgroups. The first two are already known for these groups; the missing piece is uniform non-concentration. The paper obtains it by (i) importing uniform anti-concentration bounds for random walks on algebraic subvarieties over arbitrary algebraically closed fields, which give exponential decay of the probability that a word of length ℓ lands in a proper algebraic subgroup with constants independent of the characteristic; and (ii) a counting lemma (Lemma 2.3) showing that the family of 'types' — the geom","core_discovery":"The central claim is Theorem 1.2 (the dimension zero theorem): for every rank bound r and every δ>0, there is ε=ε(r,δ)>0 such that for every prime power q and every finite simple group G(q) of Lie type of rank at most r, all but O_{δ,r}(q^δ) conjugacy classes of generating pairs a,b∈G(q) give Cayley graphs satisfying the ε-expansion inequality |AB| ≥ (1+ε)|B| for all A,B with |B|≤|G|/2 and A generating and containing the identity. A companion result, Theorem 1.1, says that for fixed r and m the family of groups G(p^k) with k≤m is uniformly ε-expanding for all primes p outside a set of density zero, i.e., a set with at most X^δ primes below X for any δ>0. The paper also extends both statement","pith_inferences":["A natural testable extension is to identify the exceptional 'scar' pairs explicitly: they should be exactly the conjugacy classes whose representative pair is fixed by a Frobenius-type automorphism, and the paper's explicit bounds on degrees and Bezout constants could in principle be used to enumerate them for small rank.","The exponential bound on types suggests that the number of exceptions may in fact be much smaller than q^δ; a plausible strengthening would be a fixed polynomial bound O(q^c) for some c<1, or even polylogarithmic in q for fixed rank, though the paper's method only gives O(q^δ) for arbitrary δ.","Because the only obstacle to removing the density-zero exceptional primes is the super-exponential bad-reduction bound inherited from general effective nullstellensatz estimates, one might hope that the special structure of word maps (e.g., their ideals having small Groebner bases) could push the bound down to exponential in ℓ, eliminating the exceptions entirely.","The dimension-zero approach via character varieties G^m//G may generalize to other families of finite groups where a similar counting of 'types' applies, such as other algebraic-group-valued word maps in positive characteristic."],"forward_implications":["If Theorem 1.2 holds, the uniform expansion conjecture for finite simple groups of Lie type reduces to excluding at most O(q^δ) conjugacy classes per group, so any failure would have to be an extremely sparse 'scar' phenomenon.","For prime fields, the density-zero exceptional set means that for each fixed rank and exponent bound m, the family G(p^k) with k≤m is a family of expanders on all but a density-zero set of primes.","The dimension-zero statement upgrades the earlier power-saving result of [16] from random pairs to all pairs except O(q^δ) conjugacy classes.","The method yields super-strong approximation in positive characteristic modulo prime ideals of prime degree or modulo 'most prime ideals', as the paper explicitly notes.","The proof applies uniformly to all Lie-type groups, including twisted and Suzuki–Ree groups, without requiring the special subcases that previous work needed."],"supporting_citations":[{"why":"Supplies the uniform anti-concentration bounds for random walks on algebraic subvarieties over fields of arbitrary characteristic, the key input needed for the non-concentration-on-subgroups phase.","marker":"[5]"},{"why":"Provides the effective bad-reduction and discriminant bounds for ideals over number fields, used to transfer characteristic-zero statements to almost all primes.","marker":"[4]"},{"why":"Originates the Bourgain–Gamburd method for proving spectral gaps in Cayley graphs of SL2(p), which this paper extends to all Lie-type groups.","marker":"[9]"},{"why":"The earlier result establishing uniform expansion for SL2(p) except a zero-density set of primes, the direct precursor this paper generalizes.","marker":"[14]"},{"why":"Establishes the product theorem and quasi-randomness setup for all finite simple groups of Lie type and proves a power-saving random-pair expansion result that Theorem 1.2 improves.","marker":"[16]"},{"why":"Larsen–Pink classification of M-dense finite subgroups as subfield subgroups, used in the subgroup analysis (Lemmas 3.7 and 4.5).","marker":"[34]"},{"why":"Nori's theorem on finite subgroups of GL_n(F_p), used to approximate the finite subgroup by a connected algebraic subgroup in Lemma 3.7.","marker":"[40]"},{"why":"Lubotzky–Weiss, the paper that formulated the uniform expansion question for finite simple groups.","marker":"[38]"}],"fun_headline_variants":["Lie-type groups: all but a zero-dim scar of pairs expand","Uniform expansion for Lie-type groups, except a sparse exceptional set","Bounded-rank Lie groups expand almost everywhere","Almost all generator pairs of Lie-type groups expand","Lie-type groups expand uniformly outside a zero-dim scar"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The argument assumes that the uniform anti-concentration theorem for random walks on algebraic subvarieties (stated as Theorem 1.3 of the companion paper [5]) holds with constants independent of the characteristic of the field, including in positive characteristic; if that uniformity fails, the non-concentration-on-subgroups step feeding the Bourgain–Gamburd machine collapses, and with it both main theorems.","fun_headline_variants_meta":{"raw":{"variants":["Lie-type groups: all but a zero-dim scar of pairs expand","Uniform expansion for Lie-type groups, except a sparse exceptional set","Bounded-rank Lie groups expand almost everywhere","Almost all generator pairs of Lie-type groups expand","Lie-type groups expand uniformly outside a zero-dim scar"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000598,"raw_usage":{"total_tokens":2591,"prompt_tokens":657,"completion_tokens":1934,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":401,"completion_tokens_details":{"reasoning_tokens":1856}},"tokens_in":401,"tokens_out":1934,"duration_ms":16396,"temperature":1.0,"reasoning_tokens":1856,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T00:19:29.199812+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take SL_2(p) (or any fixed-rank family) and, for a sequence of primes p, enumerate all conjugacy classes of generating pairs (a,b) and compute the non-concentration probability µ^ℓ_S(H) for the largest proper subgroup H at ℓ = floor(C log p) for some fixed C. If for some C and infinitely many p there are more than p^δ conjugacy classes with µ^ℓ_S(H) > e^{-cℓ} for every fixed c>0, then the key estimate (4.2) fails and the dimension-zero theorem collapses. A direct check of the characteristic-uniformity assumption would be to compute, for the subvariety x^p-x=0 inside SL_2 over F_p, whether the","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the effective bad-reduction and discriminant bounds for ideals over number fields, used to transfer characteristic-zero statements to almost all primes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Originates the Bourgain–Gamburd method for proving spectral gaps in Cayley graphs of SL2(p), which this paper extends to all Lie-type groups."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The earlier result establishing uniform expansion for SL2(p) except a zero-density set of primes, the direct precursor this paper generalizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes the product theorem and quasi-randomness setup for all finite simple groups of Lie type and proves a power-saving random-pair expansion result that Theorem 1.2 improves."},{"cited_title":"Larsen and Richard Pink,Finite subgroups of algebraic groups, J","cited_arxiv_id":null,"evidence_quote":"Larsen–Pink classification of M-dense finite subgroups as subfield subgroups, used in the subgroup analysis (Lemmas 3.7 and 4.5)."},{"cited_title":"Nori,On subgroups of GLn(Fp), Invent","cited_arxiv_id":null,"evidence_quote":"Nori's theorem on finite subgroups of GL_n(F_p), used to approximate the finite subgroup by a connected algebraic subgroup in Lemma 3.7."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Lubotzky–Weiss, the paper that formulated the uniform expansion question for finite simple groups."}],"review_version":1}