{"id":"16168f4d-b901-4379-ab00-b1dbedd3494a","arxiv_id":"2501.09865","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For ℓ-divisible extensions K/F, the ℓ-rank of the class group of K is at least the number of base primes with at least one ℓ-divisible ramification index, divided by a group invariant, corrected by unit ranks and the ℓ-part of the degree.","lead":"This paper proves new lower bounds on the size of certain algebraic symmetries, called ℓ-torsion, in the ideal class groups of number fields, using only primes where at least one above prime has ramification index divisible by ℓ. The bounds apply to a newly defined family of extensions and show that, in many nilpotent families, almost all fields have infinite class field towers.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified; Theorem 1.4 and its proof are internally coherent, and the ℓ-divisibility premise, while restrictive, is an hypothesis rather than a gap.","rationale":"The reader's weakest-assumption analysis and I both focus on Definition 1.3 as the part of the paper most likely to hide a problem. However, I do not regard that as a correctness risk: the proof of Theorem 1.4 is valid conditional on the hypothesis, and the appendix contains genuine group-theoretic verifications for all families used in the stated corollaries. I attempted to find a concrete failure mode in the proof, particularly in the passage from a ramified prime p to an intermediate prime q whose powers become ℓth powers in K. The distinctness of the q's, the use of Sylow ℓ-subgroups of inertia groups, and the final Kummer-rank bound all appear sound. I also tested the appendix's dihedral criterion on small cases; the criterion is consistent with direct computation. The paper does rely on substantial cited results (Connell–Sussman, Golod–Shafarevich, Klüners–Malle, Malle's parametrization), and there is no machine-checked formalization, but the reader's moderate-confidence acceptance already accounts for that. The minor textual slips I noticed are easily repaired and do not affect the mathematics. Hence the accepted verdict should stand unchanged.","tokens_in":22961,"tokens_out":37374,"duration_ms":392769,"concrete_test":"Use GAP to enumerate all subgroups H ≤ G_i ≤ D_8 and D_6, compute conjugacy classes of order-2 elements, and verify Definition 1.3 directly: this should give δ_2(D_8/⟨s⟩) = 2 and prove that D_6/⟨s⟩ is not 2-divisible, thereby checking the explicit δ computations on which Theorem 1.4's dihedral applications depend.","verdict_should_be":"UNCHANGED","load_bearing_attack":"After checking the central argument, I find no load-bearing flaw. The proof of Theorem 1.4 correctly converts each ramified prime counted by t_ℓ into a distinct prime q in some intermediate field whose qO_K is an ℓth power: distinctness follows because q∩F = p, and the contradiction at (2.8) is valid because the inertia group I(Q|q) lies in G_i. Lemma 2.2's rank inequality is a faithful adaptation of Connell–Sussman; the exact-sequence argument around (2.2)–(2.6) gives the kernel identification used in the injection. The final Kummer-exponent term is bounded by e_ℓ(K/F_{i0}) because the field generated by the relevant ℓth roots has ℓ-power degree dividing [K:F_{i0}]. The only restrictive point is Definition 1.3: ℓ-divisibility is nonconstructive and must be verified for each Galois pair. The appendix proves it for the families used (towers, nilpotent groups, semidirect products), and I found no counterexample to those verifications; for instance, the claimed non-divisibility of D_6/⟨s⟩ in Corollary 4.11 checks out because the conjugacy class of r^2s in any subgroup containing H and r^2s contains s. Minor issues such as 'Lemma 2.3' instead of Proposition 2.3 and the 'a such σ' phrasing in Lemma 4.6 are typos or ambiguities, not substantive gaps.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces a group-theoretic condition called ℓ-divisibility for Galois pairs G/H (Definition 1.3) and proves that for an ℓ-divisible extension K/F, each prime of F over which at least one prime of K has ℓ-divisible ramification index contributes at least 1/δℓ(K/F) to the ℓ-rank of the class group, up to unit-rank and ℓ-adic degree-error terms (Theorem 1.4). A stronger version, strong ℓ-divisibility (Definition 1.8), gives a bound supported on primes of K rather than of F (Theorem 1.9). The paper then combines Theorem 1.4 with the Golod–Shafarevich theorem to produce criteria for infinite class field towers (Corollaries 3.1 and 3.3) and an upper bound for the count of nilpotent extensions with finite class field towers (Theorem 3.6 and Corollary 1.15). The appendix verifies ℓ-divisibility and strong ℓ-divisibility for Galois towers, nilpotent groups, semidirect products, and composita, and gives explicit values for dihedral groups.","tokens_in":23223,"tokens_out":26575,"duration_ms":280761,"significance":"The main innovation is replacing the usual hypothesis that every ramification index at a prime is divisible by ℓ with the weaker condition that at least one is. This is a genuine strengthening, and the paper correctly identifies the group-theoretic obstruction with the example of cubic fields, where abundant two-prime splitting does not force 2-torsion. The proofs are detailed and mostly self-contained given standard inputs such as Connell–Sussman, Golod–Shafarevich, and Klüners–Malle; the appendix provides checkable criteria for the nonconstructive ℓ-divisibility hypothesis. There are no fitted constants, and the bounds depend only on explicit group-theoretic invariants. If Theorem 1.4 stands, it gives a new and fairly general mechanism for producing ℓ-torsion from ramification and yields a density statement for infinite class field towers in nilpotent families. The main caveat is that ℓ-divisibility is a strong premise that must be verified case by case, but the paper is transparent about this limitation.","major_comments":[{"comment":"The proof asserts that every term S_i of an arbitrary central series of the Sylow ℓ-subgroup S is a characteristic subgroup of G. This is false: central series terms of S need only be normal in S, and they need not be invariant under conjugation by G. For example, if G = C_p^2 ⋊ C_q with q acting irreducibly on C_p^2, then S = C_p^2 is the unique Sylow p-subgroup, but a line N in S gives a central series 1 < N < S for S while N is not normal in G. Since the proof subsequently forms quotients G/S_i and subgroups HS_i, an arbitrary central series is not usable. The argument can be repaired by choosing a central series of S consisting of G-invariant subgroups, for instance the upper central series of S, which preserves the stated bound on δℓ(G/H). This correction is needed because Theorem 4.8 underpins Corollary 4.9 and therefore the nilpotent density result in Theorem 3.6.","section":"4.2, proof of Theorem 4.8"}],"minor_comments":[{"comment":"The reference to “Lemma 2.3” should be to Proposition 2.3, and the expression “tℓ(K/Q)” should be “tℓ(K/F)”.","section":"3, proof of Corollary 3.1"},{"comment":"In the induction step, “the subgroup generated by a such σ” should read “the subgroup generated by all such σ”; as printed, the subgroup generated by a single element is not necessarily normal, and the proof of the induction step requires the subgroup generated by all elements of order ℓ^n.","section":"4.2, Lemma 4.6"},{"comment":"There are small typos: “There must be some σ” should be lowercase, and “pigeon-hole principal” should be “pigeon-hole principle”.","section":"2, proof of Theorem 1.9"},{"comment":"The inequality “rkℓ O*_K ≤ rkℓ O*_K + 1 ≤ n” is not coherent as printed; it should presumably be rk O*_K + 1 ≤ d (or a similar bound using the degree), and should be corrected.","section":"3, proof of Theorem 3.6"},{"comment":"The phrase “This has the affect” should be “This has the effect”.","section":"4.3, after Theorem 4.10"}],"recommendation":"major_revision","confidential_remarks":"The main theorem and its proof are sound, and I found no circularity or inappropriate self-citation. The only substantive issue is the proof of Theorem 4.8, which is patchable by choosing a G-invariant central series. I would support publication after that correction; rejection is not warranted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a solid, honest paper that earns its main theorem. Martin replaces the all-ramification-indices-divisible-by-ℓ condition in Roquette–Zassenhaus and Connell–Sussman with an at-least-one condition, under a new group-theoretic hypothesis he calls ℓ-divisibility. That is a genuinely different mechanism, and it pays off: the nilpotent density result—almost all nilpotent G/1-extensions have infinite class field towers, conditional on Klüners–Malle—is new and does not reduce to earlier genus-theory or Roquette–Zassenhaus arguments.\n\nWhat is actually new and good: Definitions 1.3 and 1.8 are new group-theoretic conditions, and the appendix proves them for a reasonable stock of examples—towers of Galois extensions, nilpotent groups, composita under a common base, and semidirect products including dihedral D_n with 4|n. Theorem 1.4 is the clean statement: t_ℓ(K/F) ramified primes, each with at least one index divisible by ℓ, force class group ℓ-rank at least t_ℓ/δ_ℓ minus unit ranks and the ℓ-adic degree exponent. The proof is structurally sound. The inertia-group step around (2.7)–(2.8) works: the Sylow argument correctly forces the intervening prime q to become an ℓth power in K, and Lemma 2.2 is a faithful adaptation of Connell–Sussman that drops the principal-ideal hypothesis they needed. The final Kummer-exponent bound is fine: the field generated by the relevant ℓth roots has ℓ-power degree dividing [K:F_{i0}]. Lemma 2.2's exact-sequence argument is correct.\n\nSoft spots, in proportion. The ℓ-divisibility condition is restrictive and nonconstructive. It must be verified for each pair G/H, and the paper does not show it holds generally; the appendix works for specific families. That is a real limitation, but it is a hypothesis, not a gap—the paper says so honestly. Theorem 1.9's gain over Theorem 1.4 is heuristic; the paper labels it as such. The density theorem leans on Klüners–Malle and Malle's weak conjecture as external inputs; the citation pattern is appropriate, no self-citation, and the dependence is explicit. There are minor typos ('Lemma 2.3' instead of 'Proposition 2.3', and the 'a such σ' phrase in Lemma 4.6), but nothing that affects the mathematics.\n\nWho it is for: number theorists working on class group bounds, genus theory, and class field towers. It deserves a serious referee; an editor should not desk reject it. My own verdict is accept after minor revisions, with the main request being a clearer discussion of how restrictive ℓ-divisibility is in practice.","headline":"A genuinely improved ℓ-rank bound under a new group-theoretic hypothesis, with a sound proof and a real density consequence; the main limitation is the restrictiveness of the ℓ-divisibility condition.","tokens_in":23773,"tokens_out":2273,"would_cite":true,"duration_ms":22225,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11R29","11R32"],"pacs":[],"model":"deepseek-v4-flash","headline":"For ℓ-divisible extensions, one prime with an ℓ-divisible ramification index below it forces a lower bound on the ℓ-rank of the class group.","keywords":["ideal class groups","ℓ-rank","ramification","ℓ-divisibility","class field towers","nilpotent extensions","dihedral extensions","genus theory"],"falsifier":"Take any $\\ell$-divisible extension $K/F$ with $t_\\ell(K/F)$ large enough to make the right-hand side of Theorem 1.4 positive, for example a dihedral quartic field with many ramified primes of type $\\mathfrak{p}_1^2\\mathfrak{p}_2$, compute $\\operatorname{rk}_\\ell \\mathrm{Cl}(K)$ exactly, and compare it with the theorem's lower bound; a smaller rank would refute the theorem.","tokens_in":22718,"feed_emoji":"🧮","tokens_out":10998,"duration_ms":96101,"temperature":0.7,"pith_summary":"This paper seeks to show that a single wildly ramified prime can force nontrivial $\\ell$-torsion in the ideal class group of a number field, provided the Galois group of the extension satisfies a new combinatorial condition it calls $\\ell$-divisibility. The main theorem bounds $\\operatorname{rk}_\\ell \\mathrm{Cl}(K)$ from below by the number of primes of the base field over which at least one prime of $K$ has ramification index divisible by $\\ell$, divided by a small group-theoretic constant, minus unit-rank and degree terms. Previous results required every prime above the counted prime to become an $\\ell$th power; here only one such prime is needed. The author verifies the $\\ell$-divisibility hypothesis for towers of Galois extensions, nilpotent extensions, and certain semidirect-product extensions, and uses the lower bound together with a classical class-field-tower criterion to prove new results on infinite class field towers, including a density statement for nilpotent extensions.","feed_headline":"Weak ramification forces ℓ-torsion in class groups","feed_subtitle":"One prime above a base prime is enough: the new ℓ-rank lower bound works where older bounds required all primes to ramify.","key_machinery":"The key machinery is the notion of $\\ell$-divisibility for a group pair $G/H$ (Definition 1.3): for every element $\\sigma\\in G-H$ of order $\\ell$, the pair must admit finitely many subgroups $H\\le G_1,\\ldots,G_\\delta\\le G$ such that some $G$-conjugate of $\\sigma$ lies in $G_i$ and has its $G_i$-conjugacy class disjoint from $H$. This condition is what lets the proof turn a prime with a single $\\ell$-divisible ramification index into a prime that becomes an $\\ell$th power inside an intermediate field, unlocking the classical relative lower bound. The appendix shows the condition holds for Galois towers, nilpotent groups, and certain semidirect products; in the dihedral case $D_n/\\langle s\\rangle$, it is $2$-divisible exactly when $4\\mid n$.","core_discovery":"The paper's central claim, Theorem 1.4, is that for an $\\ell$-divisible extension of number fields $K/F$, if $t_\\ell(K/F)$ is positive then\n$$\\operatorname{rk}_\\ell \\mathrm{Cl}(K) \\ge \\frac{t_\\ell(K/F)}{\\delta_\\ell(K/F)} - \\operatorname{rk}_\\ell O_K^\\times + \\operatorname{rk}_\\ell O_F^\\times - e_\\ell(K/F),$$\nwhere $t_\\ell(K/F)$ counts prime ideals of $F$ having at least one prime of $K$ above them with ramification index divisible by $\\ell$, $\\delta_\\ell(K/F)$ is the minimal number of intermediate groups required by Definition 1.3, and $e_\\ell(K/F)$ is the $\\ell$-adic exponent of $[K:F]$. The proof converts each counted prime $p$ of $F$ into a prime $q$ in some intermediate field $F_i$ whose powers become $\\ell$th powers in $K$, so that a relative version of the classical class-rank bound applies. The new feature is that the conversion works even when $p$ has several primes above it in $K$ and only one of them has ramification index divisible by $\\ell$; previous bounds required all of them to do so.","pith_inferences":["The theorem's value depends on how small $\\delta_\\ell(K/F)$ can be made; for a given small Galois group one can compute the optimal intermediate subgroups directly, and the appendix's catalogue is only a first step.","A natural computational test is to enumerate small dihedral quartic fields, compute the $2$-rank of their class groups, and see how often the new lower bound exceeds the genus-theory bound and by how much.","Because $\\ell$-divisibility is preserved under composita with a common base field, the theorem can be iterated to construct families of extensions with arbitrarily large $\\ell$-rank from staged ramification, suggesting quantitative lower bounds that grow with the discriminant.","The same mechanism—converting a prime with one $\\ell$-divisible index into an $\\ell$th power in an intermediate field—may transfer to narrow class groups or Selmer groups, though the paper only treats ideal class groups."],"forward_implications":["For a tower $F=F_0\\subseteq\\cdots\\subseteq F_n=K$ of Galois extensions, the lower bound becomes $t_\\ell(K/F)/n - \\operatorname{rk}_\\ell O_K^\\times + \\operatorname{rk}_\\ell O_F^\\times - e_\\ell(K/F)$, so ramification in early layers still contributes even when higher layers are unramified.","For nilpotent Galois groups, strong $\\ell$-divisibility gives an analogous bound counted at primes of $K$ rather than $F$, with the constant $\\Delta_\\ell(K/F)$.","If the relative discriminant has sufficiently many prime factors, the lower bound crosses the classical class-field-tower criterion threshold, so the $\\ell$-class field tower of $K$ is infinite.","Among nilpotent Galois extensions with fixed group over a number field, the proportion with finite class field towers tends to zero as the discriminant grows; the paper proves an explicit $O((\\log\\log x)^{e-1}/\\log x)$ upper bound.","Dihedral extensions of degree divisible by $8$ gain $2$-rank contributions from primes that split as $\\mathfrak{p}_1^2\\mathfrak{p}_2$ or $\\mathfrak{p}_1^2\\mathfrak{p}_2\\mathfrak{p}_3$, a situation where genus theory contributes nothing."],"supporting_citations":[{"why":"Supplies the relative class-rank inequality (2.1), which the proof's Lemma 2.2 strengthens by removing a coprimality condition.","marker":"[CS70]"},{"why":"The classical absolute lower bound on ℓ-rank from primes that become ℓth powers, which Theorem 1.4 generalizes to relative extensions with weaker ramification data.","marker":"[RZ69]"},{"why":"The Golod–Shafarevich criterion that Section 3 combines with Theorem 1.4 to detect infinite class field towers.","marker":"[GS64]"},{"why":"Provides the count of nilpotent Galois extensions used to turn the tower result into the density statement of Theorem 3.6.","marker":"[KM04]"},{"why":"Introduces the weak Malle conjecture and the exponent a(G/H) used to state the counting theorem and the density result.","marker":"[Mal02]"},{"why":"Landau's estimate for integers with a fixed number of prime factors underlies Lemma 3.5, which controls the number of discriminants in the density proof.","marker":"[Lan09]"}],"fun_headline_variants":["One ℓ-ramifying prime above is enough for lower bound","Class group ℓ-rank bound from a single ramified prime","Weakening ramification to at least one prime yields ℓ-rank bound","New ℓ-rank bound: only one prime needs ℓ-divisible index","Relaxed ramification gives stronger ℓ-rank lower bounds"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire bound rests on the $\\ell$-divisibility condition: for every element of order $\\ell$ that does not fix $K$, some conjugate of that element must have its whole conjugacy class inside an intermediate Galois group stay outside the subgroup of automorphisms fixing $K$; if this group-theoretic premise fails, the theorem gives nothing.","fun_headline_variants_meta":{"raw":{"variants":["One ℓ-ramifying prime above is enough for lower bound","Class group ℓ-rank bound from a single ramified prime","Weakening ramification to at least one prime yields ℓ-rank bound","New ℓ-rank bound: only one prime needs ℓ-divisible index","Relaxed ramification gives stronger ℓ-rank lower bounds"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00054,"raw_usage":{"total_tokens":2598,"prompt_tokens":963,"completion_tokens":1635,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":579,"completion_tokens_details":{"reasoning_tokens":1543}},"tokens_in":579,"tokens_out":1635,"duration_ms":13159,"temperature":1.0,"reasoning_tokens":1543,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:39:55.341568+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take any $\\ell$-divisible extension $K/F$ with $t_\\ell(K/F)$ large enough to make the right-hand side of Theorem 1.4 positive, for example a dihedral quartic field with many ramified primes of type $\\mathfrak{p}_1^2\\mathfrak{p}_2$, compute $\\operatorname{rk}_\\ell \\mathrm{Cl}(K)$ exactly, and compare it with the theorem's lower bound; a smaller rank would refute the theorem.","supporting_citations":[],"review_version":1}