{"id":"9fe85681-974b-408a-9a37-269147f1f3ce","arxiv_id":"1908.08844","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For perfect fields of characteristic p, any p-algebra of degree p^{ℓm} and exponent p^m has essential p-dimension at least ℓ+1.","lead":"This paper proves a lower bound on the essential dimension of certain p-algebras of degree p^{ℓm} and exponent p^m over perfect fields of characteristic p, showing that at least ℓ+1 parameters are needed. The result extends earlier work from the m=1 case to all exponents and also gives symbol-length bounds in Kato-Milne cohomology.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The lower bound Theorem 5.4 stands or falls with Proposition 3.1, which imports from [Kar95] an existence and prime-to-p stability statement that is not stated as a theorem and is not proved here.","rationale":"I focused on the paper's stated goal, Theorem 5.4. The internal machinery (Lemma 5.2, Lemma 5.3, Proposition 3.2) is coherent; the single point where the argument imports a nontrivial existence result is Proposition 3.1. The reader's weakest assumption points to the same step, and I agree. The secondary issue raised by the reader concerning Lemma 6.3 affects the upper-bound corollaries (6.7, 6.8) but not the lower-bound theorem, so I do not make it the primary concern. If a check of the cited literature confirms Proposition 3.1, the central theorem should stand; if not, the paper is conditional on supplying that proof. Since this agrees with the reader's CONDITIONAL verdict, no adjustment is needed.","tokens_in":9835,"tokens_out":25981,"duration_ms":281157,"concrete_test":"Check the exact statement used: retrieve Karpenko 1995 (and if needed Kar98/McK08) and identify the theorem that yields Proposition 3.1. Verify explicitly that it gives a division algebra of degree p^{ℓm} and exponent p^m over a field containing any prescribed k of char p, for all (ℓ,m) except p=ℓ=2,m=1, and that the same conclusion holds after replacing F by any prime-to-p extension L. If the reference only proves indecomposability over F, or only for m=1, or only over algebraically closed k, Proposition 3.1 is not supported and Theorem 5.4 needs a new proof.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central lower bound Theorem 5.4 depends on Proposition 3.1: for every (ℓ,m) except p=ℓ=2,m=1 there must exist a p-algebra A of degree p^{ℓm} and exponent p^m over a field F⊇k such that after every prime-to-p extension L/F the symbol length sl_{p^m}([A_L]) is at least ℓ+1. Proposition 3.1's proof is a citation: 'Such an algebra exists by [Kar95]', with no theorem number and no derivation. The cited work is not shown to provide the compound property required: (i) a division algebra of the exact degree p^{ℓm} and exponent p^m over a field containing the arbitrary base k, and (ii) stability of the lower symbol length bound under arbitrary prime-to-p scalar extensions. Karpenko's paper may prove indecomposability over the base field, but indecomposability over F does not by itself force sl_{p^m}([A_L])≥ℓ+1 over L; the proof needs the extension to stay outside the class of tensor products of ℓ cyclic p^m-algebras. Since the contradiction in Theorem 5.4 uses Proposition 3.1 exactly, an unverified or misread citation here would collapse the main theorem. This is a load-bearing gap, not a stylistic concern.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the essential dimension of central simple algebras of degree p^{ℓm} and exponent p^m over fields of characteristic p. Its main result (Theorem 5.4) asserts that if k has p-rank r, then ed(Alg_{p^{ℓm},p^m}; p) ≥ ℓ+1−r, and in particular ℓ+1 when k is perfect. The proof combines a bound on the symbol length of Brauer classes by the p-rank (Proposition 3.2 and Lemma 5.3) with the existence of p-algebras of degree p^{ℓm} and exponent p^m whose symbol length remains ≥ ℓ+1 after every prime-to-p scalar extension (Proposition 3.1). The paper also gives an upper bound on the symbol length in H^{n+1}_{p^m}(F) in terms of the p-rank (Corollary 3.4), and offers upper and lower bounds for the essential dimension of Brauer classes of a given symbol length (Section 6).","tokens_in":10144,"tokens_out":10803,"duration_ms":105012,"significance":"If Theorem 5.4 is correct, it is a significant strengthening of the earlier lower bound from [McK17], extending it from m=1 and algebraically closed base fields to all m and all perfect base fields, with a concise argument. The symbol-length bounds for Kato-Milne cohomology in Corollary 3.4 are natural and likely useful. However, the paper's central lower bound rests on an unproved existence statement imported from [Kar95], and the upper-bound section relies on a lemma that appears to be false as stated. These issues must be resolved before the paper's results can be accepted.","major_comments":[{"comment":"This proposition is the sole source of the contradiction in Theorem 5.4, yet it is not proved. The proof is a vague citation to [Kar95] (\"see the introduction... Sections 2 and 3\") with no theorem number. The required property is stronger than the existence of an indecomposable p-algebra over F: the algebra must have symbol length at least ℓ+1 after every prime-to-p scalar extension L/F. Indecomposability over F does not by itself imply that A_L is not Brauer-equivalent to a tensor product of ℓ cyclic p^m-algebras for such L. A precise statement and proof, or an accurate reference establishing this exact stability property, must be supplied for the main theorem to be supported.","section":"§3, Proposition 3.1"},{"comment":"Lemma 6.3 is false as stated. For m=1 and G=(Z/p)^ℓ with ℓ≥2, the lemma claims ed(G) ≤ 1. But if the base field k is perfect, any field E of transcendence degree at most 1 over k has p-rank at most 1 by Lemma 5.2, and a G-Galois extension with G=(Z/p)^ℓ corresponds to ℓ linearly independent classes in E/℘(E) under the Artin-Schreier correspondence. Thus ed((Z/p)^ℓ) ≥ ℓ, contradicting the lemma. This error invalidates the proof of Proposition 6.5 and the bounds in Corollaries 6.6–6.8, which are advertised in the abstract. The authors should verify the original statement in [Led04] and either replace Lemma 6.3 with a correct version or provide an alternative proof of the upper bounds.","section":"§6, Lemma 6.3"},{"comment":"Even beyond the falsity of Lemma 6.3, the descent argument in Proposition 6.5 is not justified: Remark 6.4 asserts that a subgroup G′ of W_m(F)/℘(W_m(F)) generated by ℓ elements descends to an intermediate field of transcendence degree ≤ m. For m=1 this would imply that any ℓ Artin-Schreier classes can be defined over a field of p-rank at most 1, which is impossible for ℓ≥2 by the same p-rank argument. The upper bound ed ≤ m+ℓn therefore needs a fundamentally different proof or a corrected statement.","section":"§6, Lemma 6.3 and Proposition 6.5"}],"minor_comments":[{"comment":"The text says \"The case of p = n = 2 and m = 1\" but the variable n is not otherwise used in the theorem; this should be p = ℓ = 2 and m = 1.","section":"§5, proof of Theorem 5.4"},{"comment":"The sentence \"If the symbol length of A is at most ℓ then A decomposes (as an algebra) as a tensor product of ℓ cyclic algebras of degree p^m\" is imprecise: symbol length is about Brauer equivalence, so one obtains a Brauer decomposition, not necessarily an algebra isomorphism.","section":"§3, proof of Proposition 3.1"},{"comment":"The manuscript contains several typographical artifacts (e.g., \"n /greaterorequalslantm\", \"F 4\", \"ed(Algpn,pm ; p) /greaterorequalslant3\") that should be cleaned up in the final version.","section":"General"},{"comment":"The reference to [Led04] should be checked carefully, since the stated lemma appears to be a misquotation of the original example.","section":"§6, Lemma 6.3"}],"recommendation":"major_revision","confidential_remarks":"The main lower bound Theorem 5.4 is interesting and potentially correct, but the support for it via Proposition 3.1 is currently only a vague citation, and the Section 6 upper bounds rely on a lemma that is demonstrably false. This is a load-bearing issue, not a stylistic one. If the authors can supply a rigorous proof or precise reference for Proposition 3.1 and correct or remove the invalid upper-bound results, the paper may be suitable for publication. The novelty of the lower bound relative to [McK17] should also be clarified in the revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The one thing to know: Theorem 5.4 is a genuine advance, but it rests on Proposition 3.1, which is a citation with no theorem number, and the upper-bound half of the paper (Section 6) contains a lemma that is plainly false.\n\nWhat is new and good: the observation that p-rank bounds symbol length (Proposition 3.2) and the corollary for Kato-Milne cohomology (Corollary 3.4) are clean and useful. Using Lemma 5.2 to turn essential dimension into a transcendence-degree statement is elegant. For perfect k, the lower bound ed >= ℓ+1 is a real extension of McKinnie's m=1 result to all m and to arbitrary perfect base fields. The proof of Lemma 3.3 is straightforward and correct.\n\nSoft spots, in order of severity. First, Proposition 3.1 imports a very specific existence-and-stability statement from Karpenko [Kar95]: a p-algebra of degree p^{ℓm} and exponent p^m over a field containing k whose symbol length stays >= ℓ+1 after every prime-to-p scalar extension. The proof just says \"such an algebra exists by [Kar95]\" and points to the introduction and sections. Indecomposability over F does not automatically survive field extension, and the argument needs survival under all prime-to-p extensions. If Karpenko's paper does not contain exactly that statement, Theorem 5.4 cannot be concluded. This needs to be pinned down with a precise reference and a derivation.\n\nSecond, Lemma 6.3 is false. It claims ed(G) <= m for a p-group G of exponent p^m minimally generated by ℓ elements, with |k| >= p^ℓ. For m=1, G = C_p^ℓ has essential dimension ℓ, so ed(G) <= 1 is wrong unless ℓ=1. This is not a minor typo: the lemma drives Proposition 6.5 and Corollaries 6.6-6.8, so all upper bounds in that section are unsupported. The authors likely misread Ledet's example; the statement should involve ℓ, not just m.\n\nThe lower-bound half is worth taking seriously. The paper deserves a referee, but only after the Karpenko citation is settled and Section 6 is repaired or removed. For a reading group, I would discuss Sections 2-5 and skip Section 6. I would cite the p-rank symbol-length bound, not the upper bounds.\n\nRecommendation: send to peer review with a request for major revision focused on Proposition 3.1 and Lemma 6.3.","headline":"Solid lower-bound theorem with a load-bearing citation; the upper-bound section is broken by a plainly false lemma.","tokens_in":10668,"tokens_out":3969,"would_cite":true,"duration_ms":40058,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["16K20","13A35","19D45","20G10"],"pacs":[],"model":"deepseek-v4-flash","headline":"In characteristic p, defining central simple p-algebras of degree $p^{\\ell m}$ and exponent $p^m$ costs at least $\\ell+1-r$ parameters, and at least $\\ell+1$ when the base field is perfect.","keywords":["essential dimension","symbol length","p-rank","central simple algebras","Brauer group","Kato-Milne cohomology","positive characteristic","indecomposable p-algebras"],"falsifier":"Find a perfect field k of characteristic p and positive integers $\\ell\\ge 2$, m such that some algebra A of degree $p^{\\ell m}$ and exponent $p^m$ over a field $F\\supseteq k$ has essential p-dimension at most $\\ell$; equivalently, show that after some prime-to-p extension of F, A becomes Brauer-equivalent to a tensor product of at most $\\ell$ cyclic algebras of degree $p^m$. Such an example would contradict Theorem 5.4, while proving that no such algebra exists for any perfect k would confirm it.","tokens_in":9650,"feed_emoji":"🧮","tokens_out":8242,"duration_ms":73669,"temperature":0.7,"pith_summary":"The paper proves a lower bound on how hard it is to define central simple algebras in characteristic p. Over a base field k of characteristic p and p-rank r, every algebra of degree $p^{\\ell m}$ and exponent $p^m$ has essential p-dimension at least $\\ell+1-r$; when k is perfect, this reads at least $\\ell+1$. The argument is short: the p-rank of a field bounds the symbol length in its Brauer $p^m$-torsion, and there exist indecomposable p-algebras whose symbol length stays large after every prime-to-p extension. The paper also bounds the symbol length in the higher Kato-Milne cohomology groups $H^{n+1}_{p^m}(F)$ by the binomial coefficient $\\binom{r}{n}$, and gives upper and lower bounds for the essential dimension of Brauer classes of fixed symbol length.","feed_headline":"Perfect characteristic-p fields need ℓ+1 parameters for p-algebras","feed_subtitle":"A symbol-length bound plus p-rank forces the lower bound, extending prior results to all exponent heights.","key_machinery":"The load-bearing objects are the p-rank and the symbol length. The p-rank of F is the integer r with $[F:F^p]=p^r$; the symbol length of a Brauer class is the smallest number of $p^m$-symbols, meaning cyclic algebras of degree $p^m$, whose tensor product is Brauer-equivalent to the class. Proposition 3.2 shows that a p-basis of F of size r lets every class in $\\mathrm{Br}_{p^m}(F)$ be written as a sum of at most r symbols. Lemma 5.3 converts this into a lower-bound engine: if a class descends to a finitely generated field E of transcendence degree t over k, then E has p-rank $r+t$, so its symbol length is at most $r+t$, contradicting the existence of algebras that keep symbol length at least $\\ell+1$ after prime-to-p extension.","core_discovery":"On the paper's own terms, the central discovery is Theorem 5.4: for a field k of characteristic p with $\\mathrm{rank}_p(k)=r$ and integers $\\ell\\ge 2$, $m\\ge 1$, the essential p-dimension of the functor $\\mathrm{Alg}_{p^{\\ell m},p^m}$ satisfies $\\mathrm{ed}(\\mathrm{Alg}_{p^{\\ell m},p^m};p)\\ge \\ell+1-r$. In the perfect-field case this is $\\mathrm{ed}\\ge \\ell+1$. The proof runs by contradiction: if an algebra could be descended over a prime-to-p extension to a field of transcendence degree below $\\ell+1-r$, then Lemma 5.3 plus the p-rank bounds on symbol length would force the algebra's symbol length to be at most $\\ell$; Proposition 3.1 supplies an algebra whose symbol length is at least $\\ell+1$ even after every prime-to-p extension. Along the way the paper proves that the p-rank r bounds the symbol length in $\\mathrm{Br}_{p^m}(F)$ by r, and in $H^{n+1}_{p^m}(F)$ by $\\binom{r}{n}$.","pith_inferences":["The method's bottleneck is the imported existence result for indecomposable algebras with stable high symbol length; improving that construction would raise the lower bounds directly.","If the conjectured lower bound $\\mathrm{sl}_{p^m}([A])+1\\le \\mathrm{ed}_{\\mathrm{Br}_{p^m}}([A])$ in Remark 6.9 holds, then over fields of transcendence degree d over an algebraically closed field every p-primary algebra of exponent $p^m$ would have index at most $p^{d-1}$, a special case of the period-index conjecture.","The same p-rank device may apply to other cohomological functors in characteristic p, not just Brauer groups: whenever a class is expressed by $p^m$-symbols and the base field's p-rank controls the size of a p-basis, the symbol-length upper bound yields an essential-dimension lower bound by descent."],"forward_implications":["For every perfect base field k of characteristic p, the essential p-dimension of the functor of central simple algebras of degree $p^{\\ell m}$ and exponent $p^m$ is at least $\\ell+1$, recovering the known $m=1$ bound and extending it to all exponent heights m.","For a base field of finite p-rank r, the lower bound degrades by r: $\\mathrm{ed}(\\mathrm{Alg}_{p^{\\ell m},p^m};p)\\ge \\ell+1-r$, so the same parameter-counting obstruction persists unless r is large.","Because $\\mathrm{ed}(\\mathrm{Alg}_{p^t,p^m};p)\\ge \\mathrm{ed}(\\mathrm{Alg}_{p^{\\ell m},p^m};p)$ for $\\ell=\\lfloor t/m\\rfloor$, the theorem gives lower bounds for degrees that are not pure powers $p^{\\ell m}$.","The p-rank symbol-length bound implies that any Brauer class in $\\mathrm{Br}_{p^m}(F)$ with symbol length s has essential dimension at most $s+m$ over an infinite perfect field, and combined with known symbol-length upper bounds this yields $\\mathrm{ed}_{\\mathrm{Br}_2}([A])\\le 5$ for degree-8 exponent-2 algebras.","The higher-cohomology bound says the generic sum of $\\ell$ symbols in $H^{n+1}_{p^m}$ has essential p-dimension at least $\\ell+n$, extending symbol-counting lower bounds beyond the Brauer group."],"supporting_citations":[{"why":"Supplies the existence of a p-algebra of degree $p^{\\ell m}$ and exponent $p^m$ whose symbol length stays at least $\\ell+1$ under every prime-to-p extension; this is the counterexample that drives Theorem 5.4.","marker":"[Kar95]"},{"why":"Gives the p-rank additivity formula $\\mathrm{rank}_p(E)=r+t$ for a finitely generated extension of transcendence degree t, used in Lemma 5.3 to convert essential dimension to p-rank.","marker":"[Bou90, Chapter V, Section 16.6, Corollary 3]"},{"why":"Provides the decomposition of Brauer classes split by a p-basis extension into a tensor product of $p^m$-symbols, the core of the symbol-length bound in Proposition 3.2.","marker":"[Alb68, Theorem 28]"},{"why":"Supplies the exact sequence linking $H^{n+1}_{p^{m-1}}$, $H^{n+1}_{p^m}$ and $H^{n+1}_p$, used in the induction proving the symbol-length bound for higher Kato-Milne cohomology.","marker":"[AJO18, Theorem 2.31]"},{"why":"Covers the exceptional case $p=\\ell=2$, $m=1$ in Theorem 5.4 and provides the earlier lower bound that this paper extends.","marker":"[Bae11, Corollary 2.2]"},{"why":"Gives the prior lower bound for $m=1$ and the generic-symbol result used in Theorem 6.1 for higher cohomology.","marker":"[McK17]"}],"fun_headline_variants":["Perfect p-fields demand ℓ+1 parameters for p-algebras","p-rank r forces essential dimension at least ℓ+1−r","Symbol length in Brauer group bounded by p-rank; yields ed lower bound","Perfect fields need ℓ+1 parameters for p-algebras of exponent p^m","p-rank caps symbol length, proving ed ≥ ℓ+1−r"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof rests on the imported fact that, for each relevant p, $\\ell$, m, there is an algebra of the required degree and exponent whose smallest number of cyclic factors stays at least $\\ell+1$ even after enlarging the field by any extension of degree prime to p; this existence result is cited, not proved here.","fun_headline_variants_meta":{"raw":{"variants":["Perfect p-fields demand ℓ+1 parameters for p-algebras","p-rank r forces essential dimension at least ℓ+1−r","Symbol length in Brauer group bounded by p-rank; yields ed lower bound","Perfect fields need ℓ+1 parameters for p-algebras of exponent p^m","p-rank caps symbol length, proving ed ≥ ℓ+1−r"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001171,"raw_usage":{"total_tokens":4850,"prompt_tokens":961,"completion_tokens":3889,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":577,"completion_tokens_details":{"reasoning_tokens":3788}},"tokens_in":577,"tokens_out":3889,"duration_ms":28215,"temperature":1.0,"reasoning_tokens":3788,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:32:30.286652+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a perfect field k of characteristic p and positive integers $\\ell\\ge 2$, m such that some algebra A of degree $p^{\\ell m}$ and exponent $p^m$ over a field $F\\supseteq k$ has essential p-dimension at most $\\ell$; equivalently, show that after some prime-to-p extension of F, A becomes Brauer-equivalent to a tensor product of at most $\\ell$ cyclic algebras of degree $p^m$. Such an example would contradict Theorem 5.4, while proving that no such algebra exists for any perfect k would confirm it.","supporting_citations":[],"review_version":1}