{"id":"3abcd1e9-86fc-4d26-8129-00e5d299f850","arxiv_id":"2411.16562","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Dwork's conjecture on solution growth is proved for p-adic differential modules with n=m-1 and for modules satisfying a subsidiary generic radius condition.","lead":"This paper proves new cases of Dwork's conjecture on the logarithmic growth of solutions of p-adic differential equations near the boundary of an open disc. The new cases cover equations whose solution space has dimension one less than the rank, plus a broader family identified by a gap in generic convergence radii.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unstated stronger solvability of the summand Q in Thm 4.5: the proof needs a global horizontal basis over K⟨α/t,t]]_0, not just radii equal to 1.","rationale":"The reader's weakest assumption was the uncited Kedlaya–Xiao decomposition theorem. I agree that theorem is a black box, but the more load-bearing issue is an additional, unstated property: the proof needs Q to be solvable in the strong sense of possessing a global horizontal basis over the bounded annulus. The paper's own Lemma 4.7 only establishes solvability over K{t} from radii equal to 1, which is insufficient for the gluing step. Because the rank-one example shows the general implication is false, the author must supply a citation or proof for this stronger property. If it is already contained in [Ohk19, Thm 3.5], the paper should say so; then my concern is resolved and acceptance is justified. As written, the most delicate construction rests on an unstated assumption, so I would move to conditional acceptance pending this check.","tokens_in":12567,"tokens_out":31369,"duration_ms":313734,"concrete_test":"Open [Ohk19, Theorem 3.5] (and the passage of [KX10, Thm 2.3.9] used there) and verify whether it explicitly asserts that the rank-n summand Q admits a basis {g_1,...,g_n} with D(g_i)=0 over K⟨α/t,t]]_0. If it does not, add this as a lemma: for a finite free module over K⟨α/t,t]]_0 coming from the decomposition, all subsidiary radii equal to 1 implies triviality over that ring. A minimal check of the gap: take K with a uniformizer λ satisfying |λ|=|p|^{1/(p-1)}, λ∉Z, and compute H^0(K⟨α/t,t]]_0 e, D(e)=λ t^{-1}e); it is 0 despite R_1=ρ, so the unstated implication fails and a special argument is required.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In the proof of Theorem 4.5, after obtaining the decomposition Q⊕Q′ from [KX10, Thm 2.3.9] / [Ohk19, Thm 3.5], the text asserts: 'Let {f_1,...,f_n} (resp. {g_1,...,g_n}) be a basis of P (resp. Q) such that D(f_1)=...=D(f_n)=0.' For P this is definitional, but for Q no justification is given. The cited decomposition, as quoted, only guarantees that Q′ realizes the first m−n subsidiary generic radii; it does not state that Q is solvable over the bounded annulus K⟨α/t,t]]_0. This is not a free consequence of having all n subsidiary radii equal to 1. Lemma 4.7 gives solvability over K{t} from R_1(M,1)=1, but the argument needs horizontal sections that lie in Q itself, i.e., bounded on the annulus. Over a ramified coefficient field, the rank-one module Q=K⟨α/t,t]]_0 e with D(e)=λ t^{-1}e, λ∉Z and |λ|=|p|^{1/(p-1)}, has all subsidiary generic radii equal to ρ for ρ∈(α,1) yet has H^0(Q)=0. Thus the implication 'all radii 1 ⇒ bounded horizontal basis' is false in general. The subsequent construction of e∈∧^n M by gluing Ψ(g_1)∧...∧Ψ(g_n) across the exact sequence 0→K[[t]]_0→K{t}×K⟨α/t,t]]_0→K⟨α/t,t}→0 depends on the existence of these g_i; without a proof or an explicit citation of this stronger property, the central construction is not self-contained.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves new cases of Dwork's conjecture on the logarithmic growth of horizontal sections of p-adic differential modules over the bounded unit disc. The main result, Theorem 1.4, gives two sufficient conditions under which the conjecture holds: (i) lim_{ρ→1-0} R_{m-n}(M⊗K(t)_ρ) < 1, and (ii) n = m-1. The proof strategy is to construct a solvable rank-n submodule L of M and then apply Dwork's theorem for solvable modules. The construction uses a decomposition theorem of Kedlaya-Xiao over a bounded annulus, an exterior-power argument, and a gluing exact sequence. The paper also proves a Dwork transfer theorem over K[[t]]_0 and establishes Condition (D) in the rank-2 case, recovering the author's previous result.","tokens_in":12955,"tokens_out":36915,"duration_ms":319165,"significance":"Assuming the proof can be completed, this is a substantial advance: it goes beyond the previously known rank-2 and trivial cases, and the exterior-power/gluing technique is elegant and potentially reusable. The paper is careful in setting up the necessary algebraic machinery (Property (P), elementary divisor rings, subsidiary radii) and gives reasonably self-contained proofs, with the main external input being the Kedlaya-Xiao decomposition. However, the central construction currently relies on an unproven assertion about the existence of a horizontal basis for the summand Q, which is load-bearing. The significance is high if the gap is fixable; otherwise the main theorem is not established.","major_comments":[{"comment":"The proof asserts that there exists a basis {g_1,...,g_n} of Q with D(g_i)=0, but this is not justified by the cited decomposition. The cited Kedlaya-Xiao theorem and [Ohk19, Theorem 3.5] only guarantee that Q' realizes the first m-n subsidiary generic radii; they do not assert that the complementary summand Q is solvable over the bounded annulus K⟨α/t,t]]_0. The condition that all subsidiary generic radii of Q equal 1 does not imply a horizontal basis for modules over K⟨α/t,t]]_0: for example, over a ramified coefficient field, the rank-one module Q=K⟨α/t,t]]_0 e with D(e)=λ t^{-1}e, λ∉Z and |λ|=|p|^{1/(p-1)}, has R_1(Q,ρ)=ρ for ρ∈(α,1) yet has H^0(Q)=0. The subsequent gluing step that produces e∈∧^n M depends crucially on the chosen g_i lying in Q (not merely in Q⊗K⟨α/t,t}). Without a proof or a precise citation of a stronger decomposition statement, the central construction in Theorem 4.5 is incomplete.","section":"§4, proof of Theorem 4.5"}],"minor_comments":[{"comment":"The displayed formula for F_i(M,r) repeats R_1 in every term; it should be R_1,...,R_i.","section":"§3, definition of F_i(M,r)"},{"comment":"The notation R_{m-n}(M⊗K(t)_ρ) is used in the introduction before the subsidiary generic radii are defined in §3; consider moving the definition earlier or adding a forward reference.","section":"§1, Theorem 1.4(i)"},{"comment":"The claim that there exists α∈(0,1) such that F_m(M,r)=mr for r∈[α,+∞) is stated without proof or reference. A short justification or a precise citation to [Ked22, Proposition 9.3.3] would improve readability.","section":"§4, proof of Lemma 4.7"},{"comment":"The exact sequence 0→K[[t]]_0→K{t}×K⟨α/t,t]]_0→K⟨α/t,t}→0 is used without proof; it would be helpful to cite a standard reference (e.g., [Ked22]) for this Mayer-Vietoris type sequence.","section":"§4, proof of Theorem 4.5"}],"recommendation":"major_revision","confidential_remarks":"The main gap concerns an unproven assertion about the solvability of the summand Q in the decomposition theorem. I recommend asking the author to either prove that Q can be chosen to have a horizontal basis over K⟨α/t,t]]_0, or to modify the argument to avoid this need (e.g., by constructing the gluing element e using sections of Q⊗K⟨α/t,t} and then proving they are bounded via the filtration). If the author can supply such a proof, the paper would be a strong contribution. Otherwise, the main theorem is not established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Hi [Name],\n\nShort version: the paper is real. It extends Dwork's conjecture on logarithmic growth of p-adic solutions to two new families: any rank when n=m-1, and any rank satisfying lim R_{m-n}<1. The method is an exterior-power descent that reduces the problem to constructing a solvable rank-n submodule; it's a genuine extension of Ohkubo's earlier rank-2 result, not a rehash.\n\nThe proof's spine is Theorem 4.5. Given the assumption R_{m-n}(M,1)<1, it pulls out a Kedlaya-Xiao decomposition Q⊕Q' over the bounded annulus, builds an isomorphism Γ between P and Q, and glues the n-th exterior powers to get an e∈∧^nM with D(e)=0. The gluing uses the Milnor square for K[[t]]_0, K{t}, and K⟨α/t,t]]_0. I checked the step the stress-test flagged: the sentence claiming a basis {g_i} of Q with D(g_i)=0. That claim is indeed unsupported and usually false—a rank-one module with D(e)=λ/t e, |λ|=|p|^{1/(p-1)}, λ∉Z, has all radii 1 but no horizontal sections. But the proof never actually uses D(g_i)=0. It only needs a basis of Q to form the wedge; the argument for D(e)=0 goes through τ(e), the P-side, not the Q-side. So the stress-test note misses the mark. The sentence should be deleted or fixed, but it's a cosmetic flaw, not a gap.\n\nThe only genuine soft spot is the reliance on Kedlaya-Xiao as a black box. That's standard practice and the cited theorem is published, so it's not a weakness in any serious sense. The rest of the proof—the vanishing of Hom to get Γ, the wedge-image equality, the exact sequence gluing—is coherent and self-contained.\n\nWho's this for? Specialists in p-adic differential equations, G-functions, and Dwork's conjecture. They should read it. It doesn't resolve the full conjecture but it's the first progress in years on the non-solvable side.\n\nSend it to peer review. The math looks solid, the exposition is mostly clean, and the result is worth having on record.\n\nBest,\n[Your name]","headline":"Proves Dwork's conjecture for two new families by an exterior-power descent trick; the stress-test worry about Q's horizontal basis evaporates on reading—the proof never uses it.","tokens_in":13485,"tokens_out":8552,"would_cite":true,"duration_ms":73364,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["12H25"],"pacs":[],"model":"deepseek-v4-flash","headline":"New cases proved for Dwork's p-adic solution-growth conjecture","keywords":["Dwork conjecture","p-adic differential equations","logarithmic growth","horizontal sections","subsidiary generic radii","differential modules","Dwork transfer theorem","solvability"],"falsifier":"Construct a rank-$2$ module $M$ over $K[[t]]_0$ with $n=1$ and $\\lim_{\\rho\\to 1-0} R_1(M,\\rho)<1$ whose unique horizontal section has logarithmic growth order strictly greater than $0$; the conjecture predicts such a section is bounded in logarithmic order by $0$. More generally, take any module covered by Theorem 1.4 with $n<m-1$ and compute the logarithmic order of its horizontal sections: a single section of order strictly larger than $n-1$ would disprove Conjecture 1.2 in the new cases.","tokens_in":12361,"feed_emoji":"📈","tokens_out":10398,"duration_ms":93208,"temperature":0.7,"pith_summary":"Solutions of $p$-adic differential equations on the open unit disc can grow logarithmically near the boundary; Dwork proved that, when the equation is solvable, every horizontal section has logarithmic order at most $m-1$. This paper establishes Dwork's conjecture for a broader class of equations that may not be solvable: the conjecture asserts that the bound should be $n-1$, where $n$ is the actual number of horizontal sections. The main theorem proves the conjecture for every nonzero finite free differential module whose $(m-n)$-th subsidiary generic radius of convergence has limit strictly below $1$ at the boundary, and in particular whenever $n=m-1$. The proof produces a solvable rank-$n$ submodule whose horizontal sections coincide with those of the original module, then applies Dwork's solvable theorem to that submodule. If correct, the conjecture is now certified for these new families and the growth filtration is determined without assuming solvability.","feed_headline":"New cases proved for Dwork's p-adic solution-growth conjecture","feed_subtitle":"Dwork's boundary-growth bound now reaches non-solvable equations: a radius gap suffices, and so does n=m−1.","key_machinery":"The central device is Condition (D): a triple $(L,\\phi,\\theta)$ consisting of a rank-$n$ finite free differential submodule $L\\subset M$ over $K[[t]]_0$, with $L\\otimes K\\{t\\}\\cong P:=H^0(M\\otimes K\\{t\\})\\otimes K\\{t\\}$ compatibly with the inclusion into $M$. When Condition (D) holds, Proposition 4.4 shows Conjecture 1.2 follows by applying Dwork's theorem to the solvable module $L$. To produce $L$, the paper uses the Kedlaya–Xiao decomposition $Q\\oplus Q'$ over $K\\langle\\alpha/t,t]]_0$ separating the top $n$ subsidiary generic radii from the smaller ones, an exterior-power descent argument analogous to Kedlaya's trick for $\\phi$-modules to obtain a horizontal section $e\\in\\wedge^n M$, and then defines $L$ as the kernel of $x\\mapsto x\\wedge e$.","core_discovery":"Conjecture 1.2 predicts that for a rank-$m$ finite free differential module $M$ over $K[[t]]_0$ with $n=\\dim_K H^0(M\\otimes K\\{t\\})$, the whole space of horizontal sections lies in $\\mathrm{Fil}^{n-1}$, i.e. every horizontal solution has logarithmic order at most $n-1$. The paper proves two new cases. If $\\lim_{\\rho\\to 1-0} R_{m-n}(M\\otimes K(t)_\\rho)<1$, then Conjecture 1.2 holds for $M$; and if $n=m-1$, it holds as well. The proof uses the Kedlaya–Xiao decomposition of $M\\otimes K\\langle\\alpha/t,t]]_0$ into a rank-$n$ summand $Q$ carrying the top $n$ subsidiary generic radii, all equal to $1$, and a complementary summand $Q'$ carrying the smaller radii, and shows that the horizontal-section module $P=H^0(M\\otimes K\\{t\\})\\otimes K\\{t\\}$ maps isomorphically into $Q$. An exterior-power descent then yields a horizontal section $e\\in\\wedge^n M$, and the kernel $L$ of $x\\mapsto x\\wedge e$ is a solvable rank-$n$ differential submodule whose horizontal sections are exactly those of $M$. Dwork's theorem 1.1 applied to $L$ gives the desired filtration statement.","pith_inferences":["The full conjecture will require a different argument exactly when $\\lim_{\\rho\\to 1-0} R_{m-n}(M,1)=1$ with $n<m$, because that is where the Hom-vanishing step that constructs the isomorphism $P\\otimes K\\langle\\alpha/t,t\\}\\cong Q\\otimes K\\langle\\alpha/t,t\\}$ breaks down.","The exterior-power descent that produces the horizontal section $e\\in\\wedge^n M$ may carry more information than the single bound $n-1$; tracking the growth of the sections that build $e$ could refine the log-growth filtration beyond its top index inside the new cases.","Since Lemma 4.7 identifies $n=m$ with $R_1(M,1)=1$, the conjecture could be restated as a purely radius-theoretic statement about the first subsidiary radius at $\\rho=1$ and the dimension of the horizontal-section space."],"forward_implications":["For every nonzero $M$ with $n=m-1$, Conjecture 1.2 holds, so the horizontal sections are governed by the bound $m-2$ and are realized by a solvable rank-$(m-1)$ submodule.","The radius criterion $\\lim_{\\rho\\to 1-0} R_{m-n}(M,1)<1$ is checkable from the subsidiary generic radii, so the theorem gives a practical sufficient condition for the filtration statement whenever $n<m$.","Combined with the previously known cases $m\\le 2$ and $n=m,0$, the conjecture is now established for all modules in the new families, including the previously open rank $3$ case with $n=2$.","The horizontal sections of $M$ and of the constructed submodule $L$ coincide, so Dwork's solvable theorem transfers the top logarithmic-growth bound from $L$ to $M$."],"supporting_citations":[{"why":"Supplies the decomposition theorem over $K\\langle\\alpha/t,t]]_0$ that splits the module into the rank-$n$ summand with top radii and the complement with smaller radii; this is the starting point for constructing the submodule $L$.","marker":"[KX10]"},{"why":"Provides the theory of subsidiary generic radii, the limit theorem used to pass from $\\rho<1$ to $\\rho=1$, and Dwork's transfer theorem used in Lemma 4.7.","marker":"[Ked22]"},{"why":"States Dwork's fundamental theorem that solvable modules satisfy the $m-1$ bound; this is the base case $n=m$ and the theorem applied to the constructed submodule $L$.","marker":"[And08]"},{"why":"Proved the rank-$2$ case of the conjecture and gave a detailed proof of the Kedlaya–Xiao decomposition used here; the present paper extends its method to arbitrary rank and codimension.","marker":"[Ohk19]"},{"why":"Contributes the exterior-power trick for descending modules, whose analogue is used to construct the horizontal section $e\\in\\wedge^n M$.","marker":"[Ked05]"}],"fun_headline_variants":["Dwork conjecture gains cases for non-solvable p-adic equations","Dwork's growth bound proved for n=m-1 and radius-gap cases","Two new cases of Dwork's p-adic solution growth conjecture","Radius gap or n=m-1: Dwork conjecture now proved","Descent proves Dwork's growth bound for two new equation cases"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof rests on Kedlaya and Xiao's decomposition theorem, cited as a black box: it assumes that whenever the subsidiary generic radii satisfy the required gap, the module over $K\\langle\\alpha/t,t]]_0$ splits as $Q\\oplus Q'$ with exactly those radii, and if that splitting fails for some $M$, the construction of the submodule $L$ collapses.","fun_headline_variants_meta":{"raw":{"variants":["Dwork conjecture gains cases for non-solvable p-adic equations","Dwork's growth bound proved for n=m-1 and radius-gap cases","Two new cases of Dwork's p-adic solution growth conjecture","Radius gap or n=m-1: Dwork conjecture now proved","Descent proves Dwork's growth bound for two new equation cases"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000552,"raw_usage":{"total_tokens":2658,"prompt_tokens":998,"completion_tokens":1660,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":614,"completion_tokens_details":{"reasoning_tokens":1565}},"tokens_in":614,"tokens_out":1660,"duration_ms":11134,"temperature":1.0,"reasoning_tokens":1565,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T12:59:34.590266+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Construct a rank-$2$ module $M$ over $K[[t]]_0$ with $n=1$ and $\\lim_{\\rho\\to 1-0} R_1(M,\\rho)<1$ whose unique horizontal section has logarithmic growth order strictly greater than $0$; the conjecture predicts such a section is bounded in logarithmic order by $0$. More generally, take any module covered by Theorem 1.4 with $n<m-1$ and compute the logarithmic order of its horizontal sections: a single section of order strictly larger than $n-1$ would disprove Conjecture 1.2 in the new cases.","supporting_citations":[],"review_version":1}