Math RefereeIndependent analyses and adjudication provenance
These six readable mathematical analyses preserve the returned text. All six original deliveries failed native schema validation. Metadata-only normalized copies subsequently passed the strict source validator. Their existence does not change the original statuses or certify their conclusions. Current source checks, edition corrections and the new forward spectral repair in the panel report supersede provisional remarks here. In particular, the 19 October reports predate the completed repair and whole-paper Sol return.
The package also includes the accepted original whole-paper Sol assessment and 25 accepted local assessments as JSON. The whole-paper assessment reviewed arXiv v1; its objections are not automatically errors in the revised Acta text. Four local deliveries and three other whole-paper attempts returned no accepted assessment.
Visible independent arbitration — t0
Model: gpt-6-astra; effort max. Original delivery: rejected by schema validation. This readable record preserves the returned mathematical text. A separately logged metadata-normalized copy passes the strict source validator; this is not a new model call or a formal/human proof certificate.
Section 2, expressly reused for t0, contains an invalid pointwise expansion inference, an incomplete case split, and an unsupported tube-to-displacement inference. The full introduction supplies definitions omitted from the review packet, although t0's real-group formulation leaves discreteness implicit. The reverse implication remains dependent on an uninspected reference. No downstream theorem or whole-paper verdict is asserted.
Statement assessment
unresolved
The identified failures concern the supplied proof, not a counterexample to the intended equivalence for discrete groups and their finite covers.
Findings
finding-39d760b707b1a4c68ab1
Returned disposition: confirmed_error; returned severity: substantive.
Let e be the unit constant vector in the regular representation, and choose a nonzero smooth scalar function h on the base quotient with integral h·φ₀ equal to zero. Then F=h e satisfies the required equivariance and global ground-state orthogonality. Nevertheless, F(γz)=F(z) for every γ, contradicting the displayed inequality wherever h is nonzero. Global orthogonality does not imply that each F(z) lies in the orthogonal complement of the invariant fiber.
Hypothesis check: This countercheck works already in the intended discrete integer-group setting, for any finite deck group. It does not depend on verifying the cited expansion theorem. The source imposes no pointwise zero-sum condition and supplies no separate invariant-fiber treatment.
Original quotation:
for all $F \in H_0(q)$, we have \beq \label{e:exp} \|F(\gamma z)
-F(z)\| \geq \varepsilon \|F(z)\| \quad \text{for some}\, \,
\gamma \in S . \eeq
Later-edition disposition: persists.
Page 259 changes the orthogonality notation to ϕ₀⊗Id but retains global ground-state orthogonality and the same universal pointwise assertion. The example h e remains orthogonal to that product because its scalar integral against ϕ₀ vanishes.
Later-edition quotation:
Applying [3, Theorem 1.7] for each z∈F(1) implies that there is ε>0, depending
only on S, such that for all F ∈H0(q), we have
∥F(γz)−F(z)∥⩾ε∥F(z)∥ for some γ ∈S. (2.1)
finding-506bcbd22c1215945a9e
Returned disposition: confirmed_error; returned severity: substantive.
Writing x=∫|b|², the preceding calculation gives only Q(F)−λ₀≥c₁x. This provides a uniform improvement when x exceeds a fixed positive threshold, but no such improvement as positive x tends to zero. The subsequent calculation requires x=0 and uses the resulting pointwise identity ‖u‖=1. Neither a quantitative perturbation argument nor a justified limiting reduction covers 0<x≤ε₁. Allowing ε₁ to depend on F would not prove a uniform spectral gap.
Hypothesis check: Even granting the asserted ground state and c₁>0, the case split is incomplete. Finite-index normality and expansion do not themselves fill the missing quantitative step.
Original quotation:
Now consider the case of $\int_{\fd}|b|^2 d \mu =0$. We can assume
that $a=1$ and write $F(z) =u(z) \varphi_0(z)$
Later-edition disposition: persists.
Page 260 retains the exact-zero assumption. The following argument still uses unit pointwise norm and supplies no estimate for small positive remainders.
Later-edition quotation:
Now consider the case
∫
F |b|2 dµ=0. We may assume a=1 and write F(z)=u(z)ϕ0(z),
finding-77994208a782ffc56683
Returned disposition: confirmed_error; returned severity: substantive.
Fubini retains an integral over the transverse coordinate t; it does not produce 2δ times an unspecified slice. Selecting one favorable slice would not establish the subsequent average over the entire ball. Moreover, the tubes are declared to lie inside K, without a construction identifying their endpoints with w and γⱼw across paired sides. The one-dimensional inequality therefore does not establish e:grad4 as displayed. This identifies a gap in the derivation, not a disproof of an appropriately formulated compact-core estimate.
Hypothesis check: Even granting smoothness, positivity of φ₀ on the compact region, and finite-sided geometry, one still needs the transverse average and a compatible endpoint/path construction. Compactness can bound weights; it does not provide those identifications by itself.
Original quotation:
by Fubini's
Theorem we obtain \beq \int_{T_{j}(\delta)} \varphi_0^2 \|\nabla
u\|^2 d \mu \ge 2 \delta \int_{\rho_{1, j}}^{\rho_{2,
j}}\varphi_0^2 \|u'(\rho)\|^2 \cosh(\rho) d \rho. \eeq
Later-edition disposition: persists.
There is a genuine partial correction: page 262 retains the double integral in t and ϱ, fixing the unsliced Fubini formula. However, the tubes remain inside K, and page 263 still asserts the ball displacement estimate without supplying paired-side paths or endpoint matching. Thus the broader finding persists, not its corrected Fubini component.
Later-edition quotation:
we obtain that (2.7) implies that for all j=1, ..., k we have
∫
B(z,δ)
ϕ2
0∥u(γjz)−u(z)∥dµ(z) < 3
∫
B(z,δ)
ϕ2
0 dµ(z). (2.8)
finding-7c5300e2838c17082a51
Returned disposition: unresolved; returned severity: substantive.
The claim that δ and the spectral setting are absent is resolved by full-source context: the introduction defines δ and subsequently indexes Laplace eigenvalues on XΛ. The notation λⱼ(Λ/Nᵢ) is evidently intended to concern the cover Nᵢ\H, not the finite deck group itself. However, t0 changes from integer to real subgroups without explicitly assuming discreteness or stating a global discrete-only convention. The intended Fuchsian reading is apparent, but the literal real-group scope is not adequately specified. This compound objection is therefore only partly resolved by packet context.
Hypothesis check: Discreteness is automatic in the integer-group argument, not from finite generation in SL(2,R). Finite-index normality does not supply it. The substantive unresolved issue concerns literal scope, not the existence of the introductory definitions.
Original quotation:
Let $\Lambda$ be a finitely generated non-elementary subgroup of
$\SL_2(\zed)$; let $X_{\Lambda} =\Lambda \bs \hyp$ be the
corresponding hyperbolic surface (which is of infinite volume if
$\Lambda$ is of infinite index in
$\SL(2, \zed)$). Let $\delta(\Lambda)$ denote the Hausdorff
dimension of the limit set of $\Lambda$.
Later-edition disposition: unclear.
The later introduction again supplies δ and spectral definitions, but Theorem 1.2 retains the real-group wording without explicit discreteness and retains the finite-quotient spectral notation. It does not resolve whether discreteness is an implicit convention or an omitted hypothesis.
Later-edition quotation:
Theorem 1.2. Let Λ=⟨S⟩ be a finitely generated subgroup of SL(2, R) with δ(Λ)> 1
2 .
finding-0d441924df5b704cb972
Returned disposition: unresolved; returned severity: substantive.
A precise external citation can legitimately supply this implication; absence of its proof here is not a confirmed mathematical error. However, the supplied material contains only the citation and bibliography, not Gamburd's §7. Its exact hypotheses and applicability to t0 cannot be checked. Neither the article's integer-group title nor the introduction's separate δ>5/6 application establishes the scope of that section.
Hypothesis check: The required implication concerns arbitrary finite-index normal families in the intended discrete real-group setting with δ>1/2 and a fixed generating set. Those hypotheses must be checked against the actual cited argument.
Original quotation:
the implication $\ref{item2}
\Rightarrow \ref{item1}$ is proved using Fell's continuity of
induction in section 7 of \cite{ag}.
Later-edition disposition: unclear.
Page 256 repeats the same reference without supplying its argument. The applicability question remains unresolved; repetition is neither a correction nor evidence of an invalid citation.
Later-edition quotation:
The implication (ii) ⇒(i) is proved us-
ing Fell’s continuity of induction in [10, §7].
Visible independent arbitration — narrative:51014
Model: gpt-6-astra; effort max. Original delivery: rejected by schema validation. This readable record preserves the returned mathematical text. A separately logged metadata-normalized copy passes the strict source validator; this is not a new model call or a formal/human proof certificate.
The difference-measure convolution application has a substantive normalization gap, corrected locally in the later edition. The missing fixed-block factor is an index/normalization typo. The subgroup reduction remains unverified, and the original real-parameter uniformity is unspecified. The constant-choice, invariance and remainder objections do not establish additional substantive errors once the full preceding estimates are used. This is a local assessment, not a verdict on the entire paper.
Statement assessment
unresolved
Findings
finding-c89d1a5fd82a16a6927a
Returned disposition: unresolved; returned severity: substantive.
The fixed-real-parameter choice is checkable. Take φ to be the indicator of a cylinder of length k=⌊ηm⌋. Normalizing Ruelle's operator by its positive eigenfunction, primitivity and nontriviality give two predecessor blocks with positive lower probability bounds. Consequently ν(C_k)≲θ^k and ∥φ∥ρ≲ρ^−k. Small η makes both terms in e36 exponentially small. Thus the spectral-gap objection alone is insufficient. What remains unresolved is the requested uniformity: reviewed §8 specifies no bounded range of Re z, while tr supplies constants for each potential.
Hypothesis check: The construction uses real Hölder potential, primitive nontrivial coding and tr(1),(3),(5). Non-elementarity excludes the one-symbol exception. A bounded real-parameter range cannot be borrowed from the later edition.
Original quotation:
We may now choose $\phi$ suitably, so as to obtain an estimate
Later-edition disposition: corrected.
Read with the subsequent absence of restrictions on Im z, this supplies the missing bounded real-part convention. On a compact real range, the branch and Ruelle constants can be made uniform; the cylinder choice above verifies the abbreviated step. The later text also explicitly requires nonnegative φ.
Later-edition quotation:
Here z denotes a complex number with Re z< 0 which is bounded
in absolute value.
finding-7ef07aefa50a7f05925a
Returned disposition: unresolved; returned severity: substantive.
The full source still delegates both the reduction and polynomial coset counting to bg. Properness alone does not supply the asserted reduction, and the supplied material does not contain the exceptional-case treatment or the cited counting theorem. This is an unverified essential input, not a demonstrated false theorem.
Hypothesis check: The application requires uniform control for square-free q₁, proper projection at every prime divisor, admissible words and sufficiently short word lengths. The citation's exact hypotheses cannot be checked.
Original quotation:
Then we can assume the second commutator
of $\pi_p(H)$ to be trivial if $p|q_1$
Later-edition disposition: persists.
The later edition repeats the reduction and reference without supplying the missing external argument.
Later-edition quotation:
Then we can assume the second commutator of
πp(H) to be trivial if p|q1, and hence the second commutator of H to be trivial (mod q1).
finding-12e0f66345eff3a5d8b4
Returned disposition: confirmed_error; returned severity: substantive.
Even granting this estimate, write A for the displayed bound in e321. Bounds ∥ν∥₁≤A and coset mass≤q₁^−κA do not imply l1q's hypothesis normalized by ∥ν∥₁. The difference coefficients can have much smaller total variation. No comparability is proved. The prime-modulus mass split earlier in §7 does not itself establish the needed all-divisor majorant formulation. Thus the stated application yielding e323 is unjustified.
Hypothesis check: Square-freeness and φ∈E_q are appropriate; the missing condition is normalization, not membership in E_q.
Original quotation:
\frac{1}{\eqref{e321}} |||
\pi_{q_1}(\nu)|||_{\infty} < q_1^{-\kappa}
Later-edition disposition: corrected.
Later Lemma 7.2 explicitly accepts a majorant B. Its divisor condition only concerns q₁>q^(1/10); the established q₁^−κB′ bounds imply q^−κ/10B′ there, with exponent slack for constants. This corrects the local application, without independently certifying the imported square-free flattening theorem.
Later-edition quotation:
such that if µ
satisfies ∥µ∥1⩽B and
finding-0489cf6006078208f704
Returned disposition: confirmed_error; returned severity: cosmetic.
Division by e321 gives λ^−m exp(s(S_n−T_r)), not λ^−m exp(sS_n). Bounded distortion reduces the former to the normalized m-block weight. The correct cancellation already appears in e35; changing the displayed n-block to the m-block restores the intended calculation without a new estimate.
Hypothesis check: No original hypothesis permits simply discarding exp(−sT_r); in particular s may be negative.
Original quotation:
\leq \lambda^{-m} e^{\Re z(\tau(i_n, \dots, i_1, x)+ \dots
+ \tau(i_1, x))}
Later-edition disposition: corrected.
The later expression uses precisely the m-block weight to which (8.1) applies.
Later-edition quotation:
⩽λ−meRe z(τ(im,...,i1,x)+...+τ(i1,x))
finding-69477627f2f9e32093e6
Returned disposition: unsupported_objection; returned severity: none.
The source leaves m∼log q 'to be specified'. Choose its positive coefficient small enough that m≤(log q)/4, compatible with short-word injectivity. Then n>log q gives r−n/2=n/2−m≥(log q)/4. Taking κ′<|log ρ|/4 establishes the required inequality. The same choice handles e326. Explicit numerical constants are not an additional substantive input.
Hypothesis check: This uses large q and the freedom to decrease the positive decay exponent and short-word coefficient.
Original quotation:
Therefore it follows that if $n > \log q$
Later-edition disposition: corrected.
The strengthened wording explicitly accommodates a sufficiently large coefficient; it removes the literal quantitative ambiguity.
Later-edition quotation:
Therefore it follows that, if n≫log q, then
finding-993650936bbf92b1f44c
Returned disposition: unsupported_objection; returned severity: none.
Retain an exponent β smaller than the preceding decay exponent α. With t=1+|Im z| and w=ρ^(n/2), a suitable n makes wt≤1; the bounded prefactor is absorbed by q^(α−β). Literal preservation of one numerical κ′ is unnecessary for the existential estimate. Moreover every divisor-factoring subspace, and its orthogonal complement, is translation invariant. Thus E_q and hence F′ρ are preserved by the transfer operator.
Hypothesis check: The vector-valued norms and definition of E_q are supplied in §§6–7. This verification is conditional on the preceding convolution estimates.
Original quotation:
It follows from
\eqref{e311}, \eqref{e326}
Later-edition disposition: persists.
The same abbreviated implication remains; exponent slack and translation invariance justify it as above.
Later-edition quotation:
It follows from (8.3) and (8.8) that
finding-7c2fbd7f53ca60ad7f50
Returned disposition: unsupported_objection; returned severity: none.
The alleged norm conversion is unnecessary: applying e326 to the final n-block, and e329 to preceding blocks, gives an unweighted bound ≲tλ^(kn)q^−kβ after exponent slack. For s<n, e12 and the coefficient-difference estimate e318 give N(M_z^s f)≲λ^sN(f), where N(f)=∥f∥∞+w|f|ρ, since wt≲1. Writing the power as kn+s incurs at most q^β; choosing β<1 absorbs this into q. Powers below n satisfy the crude bound ≲tλ^s. Thus the earlier displayed estimates supply the omitted remainder calculation.
Hypothesis check: Translations are unitary in the supplied l² norm. This is conditional algebra using the preceding estimates, not validation of their unresolved upstream inputs.
Original quotation:
Thus for $m \geq 1$
Later-edition disposition: corrected.
The later all-power bound explicitly allows a fixed polynomial prefactor. The unchanged multiples estimate is also supported by applying the seminorm estimate to the final block, rather than using the questioned norm conversion.
Later-edition quotation:
(q(1+|Im z|))c.
Visible independent arbitration — narrative:31101
Model: gemini-4-argon; effort high. Original delivery: rejected by schema validation. This readable record preserves the returned mathematical text. A separately logged metadata-normalized copy passes the strict source validator; this is not a new model call or a formal/human proof certificate.
In Section 3 (narrative:31101), the elementary matrix-norm to hyperbolic-distance conversion and the introductory shift-space formalism in Section 4 check out, and the hypotheses on q in Theorem t4a are present in Section 1 of the full manuscript. However, the proof of Theorem t4a in Section 3 has a confirmed normalization error in the base eigenfunction formula varphi_{0,q} = |SL_2(Z/qZ)|^{-1} varphi_{0,1} (which should carry exponent -1/2 to have unit L^2 norm and to produce the factor |SL_2(q)|^{-1} in the quadratic main term). In addition, the derivation of Theorem t4a leaves unverified the q^3-uniform Lax-Phillips remainder in (e258a), the reduction of a general residue class gamma == g mod q via a lift endpoint w = gamma_g i, and the aggregate bound on the nonprincipal discrete spectral terms. All of these features persist unchanged in the supplied later edition.
Statement assessment
unresolved
The elementary hyperbolic-geometry conversion ||g||^2 = 4u(gi, i) + 2 = 2 cosh rho(gi, i) for g in SL_2(R) and the symbolic-dynamics definitions in Section 4 are valid, and the erroneous ground-state normalization varphi_{0,q} = |SL_2(Z/qZ)|^{-1} varphi_{0,1} can be locally repaired to varphi_{0,q} = |SL_2(Z/qZ)|^{-1/2} varphi_{0,1} (up to specifying the {-I} covering convention), which restores the claimed |SL_2(q)|^{-1} scaling in the bilinear product varphi_{0,q}(z) overline{varphi_{0,q}(w)} and is invariant under coset shifts w = gamma_g i because varphi_{0,1} is Lambda-invariant. However, the proof of Theorem t4a in Section 3 also relies on the uninspected external Lax-Phillips expansion (e258), its undemonstrated q^3-uniform congruence extension (e258a), and an omitted uniform bound for the sum of nonprincipal spectral terms at varying coset endpoints w = gamma_g i, leaving the full target statement unresolved from the supplied material alone.
Findings
finding-303e3b2a156574ceab31
Returned disposition: confirmed_error; returned severity: substantive.
With respect to the standard hyperbolic area measure d mu used in Section 2 on a fundamental domain F(q) of index d = [Lambda : Lambda(q)] = |SL_2(Z/qZ)| (ignoring the +/-I action factor), the pullback of a unit-norm base eigenfunction varphi_{0,1} on F(1) has squared L^2(F(q)) norm equal to d, so unit L^2 normalization requires varphi_{0,q} = d^{-1/2} varphi_{0,1}, not d^{-1} varphi_{0,1}. Moreover, substituting the displayed formula varphi_{0,q} = d^{-1} varphi_{0,1} into the quadratic ground-state term c_0 varphi_{0,q}(z) overline{varphi_{0,q}(w)} e^{delta T} of (e258a) produces a factor of |SL_2(Z/qZ)|^{-2} rather than the claimed |SL_2(q)|^{-1} in Theorem t4a. While replacing the exponent -1 by the correct L^2-normalization exponent -1/2 immediately restores the |SL_2(q)|^{-1} factor in the product, the displayed equation is mathematically incorrect as stated.
Hypothesis check: Under the hypotheses of Theorem t4a (q square-free and coprime to q_0), strong approximation gives [Lambda : Lambda(q)] = |SL_2(Z/qZ)|. If -I is in Lambda and q > 2, the surface covering degree is [Lambda : {+/-I}Lambda(q)] = |SL_2(Z/qZ)|/2 unless the measure or counting convention on SL_2 vs PSL_2 is adjusted accordingly.
Original quotation:
\varphi_{0, q} = \frac{1}{|\SL_2(\zed/
q\zed)|} \varphi_{0,1}.
Later-edition disposition: persists.
The later edition on page 265 retains the exact same un-square-rooted normalization factor 1/|SL_2(Z/qZ)|.
Later-edition quotation:
ϕ0,q = 1
|SL2(Z/qZ)|ϕ0,1.
finding-9fab21ae2b6ce413a484
Returned disposition: confirmed_error; returned severity: cosmetic.
Equation (e:b4) defines u(z, w) = |z - w|^2 / (4 Im z Im w), so literally setting z = w = i gives u(i, i) = 0. The intended substitution into (e:b4) is z = gi and w = i for g = [[a, b], [c, d]] in SL_2(R) (so ad - bc = 1). Under det(g) = 1, Im(gi) = 1/(c^2 + d^2) and |gi - i|^2 = (a^2 + b^2 + c^2 + d^2 - 2)/(c^2 + d^2), which verifies (e:b7): 4u(gi, i) + 2 = a^2 + b^2 + c^2 + d^2 = ||g||^2. Although the preceding paragraph defines ||g|| on all of M_2(R), the identity (e:b7) is only used for elements gamma in Lambda subset SL_2(Z), where det(gamma) = 1 holds.
Hypothesis check: All group elements gamma in Lambda lie in SL_2(Z), so det(gamma) = 1 is satisfied in every application of (e:b7).
Original quotation:
By taking $z=w=i$ in (\ref{e:b4}) we obtain that
Later-edition disposition: persists.
The later edition on page 264 keeps the same phrasing 'By taking z=w=i in (3.2)' without explicitly stating z=gi, w=i or restricting (3.3) to g in SL_2(R).
Later-edition quotation:
By taking z=w=i in (3.2), we obtain that
∥g∥2 = a2+b2+c2+d2 = 4u(gi, i)+2. (3.3)
finding-a7f7afec1e3db84fbbed
Returned disposition: packet_context; returned severity: none.
Combining (e:b3) with (e:b7) gives ||gamma||^2 = 4u(gamma i, i) + 2 = 2 cosh rho(i, gamma i) for gamma in Lambda. Hence ||gamma|| <= T is equivalent to rho(i, gamma i) <= R := arcosh(T^2 / 2) = 2 log T + O(T^{-4}). Substituting R into e^{delta R} gives (T^2)^{delta}(1 + O(T^{-4})), and substituting R into the hyperbolic remainder O(q^3 R^{5/6} e^{R/2}) gives O(q^3 (log T)^{5/6} T), which is O(q^3 T^{2 delta - varepsilon_1}) because delta > 1/2. The concluding sentence cites (e:b2) and (e:b7) while leaving (e:b3) implicit, which is a harmless exposition detail.
Hypothesis check: The assumption delta > 1/2 in Section 3 ensures 2 delta > 1, so the remainder O(q^3 T (log T)^{5/6}) is absorbed into O(q^3 T^{2 delta - varepsilon_1}) for sufficiently small varepsilon_1 > 0.
Original quotation:
\cosh \rho (z, w) = 1+2 u(z, w),
Later-edition disposition: persists.
The later edition on pages 264-265 retains the same equations and citations (citing (3.1) and (3.3) in the final sentence).
Later-edition quotation:
We will use the expression
cosh ϱ(z, w) = 1+2 u(z, w),
finding-c1a35e16092b8951e07c
Returned disposition: unresolved; returned severity: substantive.
Equation (e258a) asserts a q-uniform error term O(q^3 T^{5/6} e^{T/2}) for the principal congruence subgroup Lambda(q) 'using the methods of [LP82]', without providing a derivation or citing a specific theorem in [LP82] that tracks the dependence on the index [Lambda : Lambda(q)] = O(q^3). Furthermore, Theorem t4a counts elements in an arbitrary residue class gamma == g mod q. Reducing this coset count to N_{Lambda(q)}(R; z, w) requires choosing a lift gamma_g in Lambda with gamma_g == g mod q and evaluating N_{Lambda(q)}(R; i, gamma_g i) at a coset-dependent second argument w = gamma_g i. While gamma_g acts as an isometry of Lambda(q)\H (so the main term is invariant since varphi_{0,1}(gamma_g i) = varphi_{0,1}(i)), neither the text of [LP82] nor the q- and endpoint-uniformity of (e258a) at w = gamma_g i is supplied in the manuscript.
Hypothesis check: Under the hypotheses of Theorem t4a, surjectivity of Lambda -> SL_2(Z/qZ) guarantees the existence of a lift gamma_g in Lambda for every g in SL_2(q), and normality of Lambda(q) in Lambda gives gamma in Lambda(q) gamma_g.
Original quotation:
where implied constant is independent of $q$.
Later-edition disposition: persists.
The later edition on page 265 states (3.4) in the same form without added proof details or explanation of the coset lift w = gamma_g i.
Later-edition quotation:
Turning to congruence subgroups Λ( q) we have, using the methods of [18], that
⏐⏐⏐⏐NΛ(q)(T; z, w)−
∑
j
cjϕj,q(z)ϕj,q(w)eδj,qT
⏐⏐⏐⏐= O(q3T5/6eT/2), (3.4)
where the implied constant is independent of q.
finding-824b985f35172e347657
Returned disposition: unresolved; returned severity: substantive.
Theorem t4 provides a uniform spectral gap lambda_1(Lambda(q)) >= lambda_0(Lambda) + varepsilon for square-free q, which translates via lambda_{j,q} = delta_{j,q}(1 - delta_{j,q}) to an exponent gap delta_{j,q} <= delta - eta (with eta > 0 depending only on Lambda) for all nonprincipal discrete eigenfunctions j >= 1 of Lambda(q). However, passing from (e258a) to the error term O(q^3 T^{2 delta - varepsilon_1}) in Theorem t4a also requires bounding the sum over j >= 1 of |c_j varphi_{j,q}(i) overline{varphi_{j,q}(gamma_g i)}| (or the corresponding Bessel sum of spectral projections in the Lax-Phillips wave-equation proof) by O(q^3) uniformly in q and g. This spectral-sum estimate is not explained in the manuscript and depends on the uninspected mechanics of [LP82].
Hypothesis check: Theorem t4 applies to all square-free q >= 1 for finitely generated Lambda <= SL_2(Z) with delta(Lambda) > 1/2, matching the hypotheses of Theorem t4a.
Original quotation:
Combining Theorem \ref{t4} with \eqref{e258a}
Later-edition disposition: persists.
The later edition on page 265 makes the same one-sentence transition from Theorem 1.1 and (3.4) to Theorem 1.3 without additional details on the nonprincipal spectral sum.
Later-edition quotation:
Combining Theorem 1.1 with (3.4), (3.1) and (3.3), we obtain ε1>0 depending on
Λ such that for any g∈SL2(q) we have
finding-bf9c4b8ceac8830cc22f
Returned disposition: packet_context; returned severity: none.
In the full reviewed manuscript, the hypotheses that Lambda is a finitely generated non-elementary subgroup of SL_2(Z) with delta(Lambda) > 1/2 and that q is square-free with (q, q_0) = 1 are explicitly stated in Section 1 in the statement of Theorem t4a, which the final sentence of Section 3 explicitly concludes ('establishing Theorem \ref{t4a}'). Thus the hypotheses are in force from the theorem statement being proved, while the strong approximation theorem [mvw] supplying q_0 is a pre-cutoff external reference.
Hypothesis check: All hypotheses required for Theorem t4a are stated in Section 1 of the full reviewed source.
Original quotation:
Assume that $q$ is square free and $(q, q_0) =1$,
Later-edition disposition: persists.
The later edition states the same hypotheses in Theorem 1.3 on page 256 and concludes 'establishing Theorem 1.3' on page 265.
Later-edition quotation:
Assume that q is square-free and (q, q0)=1, where q0 is provided by the strong approx-
imation theorem [20].
finding-d4bf8d435b9e49f37d69
Returned disposition: packet_context; returned severity: none.
All elementary definitions and statements in the opening part of Section 4 (compactness of Sigma, completeness of F_rho under ||.||_rho, boundedness and positivity of the transfer operator L_f, and the iterate formula for L_f^n) are straightforward and hold as stated. The absence of Ruelle's theorem in the wave-000 slice is purely an artifact of packet slicing: in the full manuscript, Theorem tr (Ruelle's theorem) immediately follows this sentence.
Hypothesis check: The standing assumptions in Section 4 (finite alphabet {1, ..., l}, irreducible aperiodic transition matrix A, 0 < rho < 1, and f in F_rho) suffice for all stated properties.
Original quotation:
For each $\rho \in (0,1)$ and $f \in \mF_{\rho}$, $\mL_{f}:
Later-edition disposition: persists.
The later edition on page 266 contains the identical definitions and immediately states Ruelle's theorem as Theorem 4.1.
Later-edition quotation:
For each ϱ∈(0, 1) and f ∈Fϱ, Lf : Fϱ!Fϱ is a continuous linear operator. If f is real-
valued then Lf is positive.
Visible independent arbitration — p3
Model: gemini-4-argon; effort high. Original delivery: rejected by schema validation. This readable record preserves the returned mathematical text. A separately logged metadata-normalized copy passes the strict source validator; this is not a new model call or a formal/human proof certificate.
Proposition p3 contains two substantive defects in its displayed formulas, both persisting in the later Acta Mathematica edition (Proposition 10.1 and (10.3)): (1) a sign error in the residue of (1 - \lambda_z)^{-1} at z = -\delta (arising from confusing the parameter z in e^{z\tau} near -\delta with s = -z in e^{-s\tau} near +\delta), which gives the subtracted principal part in (e74) the wrong sign; and (2) stating the variable rank-one subtraction (e73) on the whole half-plane \Re z < -\delta + \varepsilon_5 even though \lambda_z, \nu_z, h_z are only defined on a local neighborhood U of -\delta in Theorem \ref{ln}(2). Conditional on Theorem \ref{ln}, flipping the sign in (e74) and performing the Laurent subtraction on U before combining with (e71) outside U repairs Proposition p3, and its qualitative consequence used in Section 7a (that (I - \mathcal{L}_z)^{-1} is holomorphic on \Re z < -\delta + \varepsilon_5 except for a simple pole at z = -\delta) remains intact. Of the remaining three findings, finding-83a48d2f2dcd4e48201a is resolved by full-manuscript context (equation (e44) in Section 5 has the negative exponent e^{-\varepsilon n}, so e^{\varepsilon n} in Section 7 is a cosmetic typo, and the n=0 projection term is holomorphic), while finding-4db9e5f572137f48df65 and finding-e9db0c5391fc385d5e23 are confirmed calculation errors in the downstream proof of Proposition p4 rather than Proposition p3.
Statement assessment
refuted
Under the manuscript's transfer-operator convention (\mathcal{L}z g)(x) = \sum{\sigma y = x} e^{z\tau(y)} g(y) from Sections 4 and 5 with \tau eventually positive, differentiating \mathcal{L}z h_z = \lambda_z h_z with respect to z at z = -\delta (where \lambda{-\delta} = 1) and integrating against the eigenmeasure \nu_z (satisfying \mathcal{L}z^* \nu_z = \lambda_z \nu_z and \int h_z d\nu_z = 1) yields (d\lambda_z/dz)|{z=-\delta} = \int \mathcal{L}{-\delta}(\tau h{-\delta}) d\nu_{-\delta} = +\int \tau h_{-\delta} d\nu_{-\delta} = +\int \tau d\mu_{-\delta} > 0. Consequently, the residue of (1 - \lambda_z)^{-1} at z = -\delta is -1/\int \tau d\mu_{-\delta}, not +1/\int \tau d\mu_{-\delta} as displayed in Proposition p3. Subtracting +\frac{\nu_{-\delta} \otimes h_{-\delta}}{\int \tau d\mu_{-\delta}} \frac{1}{z+\delta} in (e74) therefore doubles the simple pole at z = -\delta rather than cancelling it, refuting the literal printed formula (e74). Moreover, the intermediate formula (e73) writes \lambda_z, \nu_z, h_z across the half-plane \Re z < -\delta + \varepsilon_5, whereas Theorem \ref{ln}(2) defines them only on a neighborhood U of -\delta. New repair (self-contained conditional on Theorem \ref{ln}): Shrink U around -\delta so that z = -\delta is the sole zero of 1 - \lambda_z on \overline{U}. On U, by (e43)-(e44), (I - \mathcal{L}z)^{-1} = \frac{1}{1-\lambda_z}(\nu_z \otimes h_z) + (I - \mathcal{L}z'')^{-1}(I - \nu_z \otimes h_z) has a simple pole at z = -\delta with residue -\frac{\nu{-\delta} \otimes h{-\delta}}{\int \tau d\mu_{-\delta}}, so (I - \mathcal{L}z)^{-1} + \frac{\nu{-\delta} \otimes h_{-\delta}}{\int \tau d\mu_{-\delta}} \frac{1}{z+\delta} is holomorphic and bounded on U. On {z : \Re z < -\delta + \varepsilon_5} \setminus U, (e71) gives that (I - \mathcal{L}z)^{-1} is holomorphic and bounded by C|\Im z|^2, while \frac{\nu{-\delta} \otimes h_{-\delta}}{\int \tau d\mu_{-\delta}} \frac{1}{z+\delta} is holomorphic and O(1). Combining the two regions establishes directly (without extending \lambda_z, \nu_z, h_z outside U) that (I - \mathcal{L}z)^{-1} + \frac{\nu{-\delta} \otimes h_{-\delta}}{\int \tau d\mu_{-\delta}} \frac{1}{z+\delta} is analytic and bounded by C(|\Im z|^2 + 1) on \Re z < -\delta + \varepsilon_5.
Findings
finding-8a6940fd31edcf3bda9a
Returned disposition: confirmed_error; returned severity: substantive.
In Sections 4 and 5, the transfer operator is defined with weight e^{z\tau} by (\mathcal{L}z g)(x) = \sum{\sigma y = x} e^{z\tau(y)} g(y), where \tau is eventually positive and \lambda_{-\delta} = 1. Differentiating \mathcal{L}z h_z = \lambda_z h_z with respect to z at z = -\delta and integrating against \nu{-\delta} (with \mathcal{L}{-\delta}^* \nu{-\delta} = \nu_{-\delta} and \int h_{-\delta} d\nu_{-\delta} = 1) gives (d\lambda_z/dz)|{z=-\delta} = \int \tau h{-\delta} d\nu_{-\delta} = +\int \tau d\mu_{-\delta} > 0. Therefore the residue of (1 - \lambda_z)^{-1} at z = -\delta is -(\lambda'{-\delta})^{-1} = -1/\int \tau d\mu{-\delta}, opposite in sign to the displayed formula. Consequently, the principal part subtracted in (e74) has the wrong sign (the authors carried over the sign from the e^{-s\tau} parameterization near s = +\delta without adjusting for z = -s).
Hypothesis check: Checked against the full manuscript: \mathcal{L}f g(x) = \sum{\sigma y = x} e^{f(y)} g(y) in Section 4, abbreviation \mathcal{L}z = \mathcal{L}{z\tau} at the end of Section 4, (e59)-(e510) in Section 5, and normalization \int h_z d\nu_z = 1 in Theorem \ref{ln}(2).
Original quotation:
-\frac{1}{(\frac{d}{ d z}
\lambda_z)|_{z=-\delta}} = \frac{1}{\int \tau d \mu_{-\delta}}
Later-edition disposition: persists.
In Section 10 of the later Acta Mathematica edition (p. 281), the same sign error in the residue formula and in equation (10.3) appears verbatim immediately following Proposition 10.1.
Later-edition quotation:
− 1
(dλz/dz)|z=−δ
= 1∫
Σ τ dµ−δ
.
Consequently
(I−Lz)−1− ν−δ ⊗h−δ∫
Σ τ dµ−δ
1
z+δ (10.3)
is analytic for Re z< −δ+ε5.
finding-872376f1e78bb5c9642d
Returned disposition: confirmed_error; returned severity: substantive.
Theorem \ref{ln}(2) in Section 5 provides the holomorphic extensions z \mapsto \lambda_z, h_z, \nu_z and the decomposition (e43) only for z \in U, a small neighborhood of -\delta. Outside U, on the half-plane \Re z < -\delta + \varepsilon_5 with arbitrarily large |\Im z|, \lambda_z, \nu_z, and h_z are not defined by Theorem \ref{ln}. Thus the global statement (e73) on \Re z < -\delta + \varepsilon_5 is not justified as written. Instead, one must replace \frac{1}{1-\lambda_z}(\nu_z \otimes h_z) by its constant-coefficient principal part at z = -\delta on U (shrinking U so 1 - \lambda_z has no other zeros on \overline{U}) and then combine with (e71) on {z : \Re z < -\delta + \varepsilon_5} \setminus U to obtain the analyticity and C(|\Im z|^2 + 1) bound for the sign-corrected version of (e74).
Hypothesis check: Checked against Theorem \ref{ln}(1)-(2) in Section 5: part (1) applies to \Re z < -\delta + \varepsilon, z \notin U, and part (2) defines \lambda_z, \nu_z, h_z only for z \in U.
Original quotation:
(I-\mL_z)^{-1} - \frac{1}{1-\lambda_z}(\nu_z \otimes h_z)
Later-edition disposition: persists.
Proposition 10.1 of the later edition (p. 281) retains the exact same formulation (e73) over the half-plane Re z < -\delta + \varepsilon_5 without defining \lambda_z, \nu_z, h_z outside U.
Later-edition quotation:
Proposition 10.1. Consider Lz acting on Fϱ(Σ). Then , for Re z< −δ+ε5,
(I−Lz)−1− 1
1−λz
(νz ⊗hz)
is holomorphic and bounded by C(|Im z|2+1).
finding-83a48d2f2dcd4e48201a
Returned disposition: packet_context; returned severity: cosmetic.
In the full manuscript, the cited equation (e44) in Theorem \ref{ln}(2) (Section 5) explicitly states |(\mathcal{L}z'')^n|\rho < e^{-\varepsilon n} for z \in U (with a negative exponent). Thus e^{\varepsilon n} in Section 7 is merely a typographical omission of the minus sign when restating (e44). Furthermore, regarding the projection onto the complementary spectral subspace: for P_z = \nu_z \otimes h_z, \mathcal{L}_z^n = \lambda_z^n P_z + (\mathcal{L}_z'')^n holds for all n \ge 1, while at n = 0 summing over n \ge 0 on the whole space gives (I - \mathcal{L}_z)^{-1} = \frac{1}{1-\lambda_z} P_z + (I - \mathcal{L}_z'')^{-1}(I - P_z) = \frac{1}{1-\lambda_z} P_z + (I - \mathcal{L}z'')^{-1} - P_z. Whether (I - \mathcal{L}z'')^{-1} is restricted to the complementary subspace (I - P_z)\mathcal{F}\rho(\Sigma) or taken on \mathcal{F}\rho(\Sigma) (differing by the holomorphic projection P_z on U), the holomorphy of (I - \mathcal{L}_z)^{-1} - \frac{1}{1-\lambda_z}(\nu_z \otimes h_z) on U is unaffected.
Hypothesis check: Checked against Theorem \ref{ln}(2), equations (e43) and (e44) in Section 5 of the full source.
Original quotation:
where $\|(\mL''_z)^n\| < e^{\vge n}$ by \eqref{e44}.
Later-edition disposition: persists.
In the later edition (p. 281), the typographical omission of the minus sign in e^{\varepsilon n} when citing (5.9) (which has e^{-\varepsilon n} on p. 269) and the display (10.2) persist unchanged.
Later-edition quotation:
Ln
z = λn
z (νz ⊗hz)+(L′′
z )n,
where ∥(L′′
z )n∥<eεn by (5.9).
Hence for z∈U,
(I−Lz)−1 = 1
1−λz
(νz ⊗hz)+(I−L′′
z )−1 (10.2)
with ( I−L′′
z )−1 holomorphic (this is Proposition 7.2 in [16]).
finding-4db9e5f572137f48df65
Returned disposition: confirmed_error; returned severity: cosmetic.
This equation (e77) lies outside Proposition p3 and belongs to the downstream proof of Proposition p4. Regarding the two points raised by the checker: (1) by Theorem \ref{tr}(3) in Section 4, \nu_f is a Borel probability measure on \Sigma, so \nu_{-\delta}(1) = \int_\Sigma 1 d\nu_{-\delta} = 1 automatically and is not missing; (2) however, evaluating the prefactor 1/z at the pole z = -\delta contributes a factor -1/\delta (which, together with the missing minus sign from \int_0^\infty e^{az} da = -1/z in (e59) and the residue sign in (e74), leaves an omitted factor of 1/\delta in the explicit formula for C_0 in (e77) and (e79), along with a typo d\nu_{-\delta} for d\mu_{-\delta} in (e79)). Because Proposition p4 only states that C_0 is a fixed constant (and Theorem t5b absorbs C_0 into c_\Lambda), this omitted 1/\delta factor in the formula for C_0 does not affect Proposition p3 or the statement of Proposition p4.
Hypothesis check: Checked against Theorem \ref{tr}(3) in Section 4 (\nu_f is a probability measure) and equations (e77)-(e79) and Proposition p4 in Section 7.
Original quotation:
F(z) = \frac{1}{z}
(I-\mL_z)^{-1} 1 = \frac{h_{-\delta} (\xi \equiv w)}{\int \tau d
\mu_{-\delta}} \frac{1}{z+ \delta} + G(z),
Later-edition disposition: persists.
The formula on p. 282 of the later edition retains the same coefficient without the 1/\delta factor from 1/z at z = -\delta.
Later-edition quotation:
F(z) = 1
z (I−Lz)−11 = h−δ(ξ ≡w)∫
Σ τ dµ−δ
1
z+δ +G(z),
finding-e9db0c5391fc385d5e23
Returned disposition: confirmed_error; returned severity: substantive.
This step occurs in the downstream proof of Proposition p4, not in Proposition p3. By (e69) and (e611) in Section 6, \operatorname{supp}(k_\gamma) \subset [-\gamma/2, \gamma/2]. Consequently, for a > \gamma/2 and t \ge 0, t + a > \gamma/2, so the displayed integral \int_0^\infty k_\gamma(t+a) dt equals 0, not 1. The error arises from a sign slip in the argument of k_\gamma (writing t+a instead of t-a, for which \int_0^\infty k_\gamma(t-a) dt = \int_{-a}^\infty k_\gamma(u) du = 1 for a > \gamma/2 by (e610)) combined with writing a principal-value integral on \Re z = -\delta instead of shifting the Laplace inversion contour from \Re z < -\delta across the simple pole at z = -\delta to \Re z = -\delta + \frac{1}{2}\varepsilon_5 (where Cauchy's residue theorem directly gives \hat{k}\gamma(0) = \int k\gamma(t) dt = 1). This does not affect Proposition p3.
Hypothesis check: Checked against (e69)-(e611) in Section 6 and (e78)-(e711) in Section 7.
Original quotation:
\int_{0}^{\infty} k_{\gamma}(t+a) d
t \overset{\eqref{e610}}{=} 1.
Later-edition disposition: persists.
The displayed identity on p. 282 of the later edition is unchanged.
Later-edition quotation:
PV-
∫ ∞
−∞
e−iaθˆkγ(iθ) 1
iθ dθ =
∫ ∞
0
kγ(t+a) dt
(9.4)
= 1 .
Visible independent arbitration — pt1
Model: gemini-4-argon; effort high. Original delivery: rejected by schema validation. This readable record preserves the returned mathematical text. A separately logged metadata-normalized copy passes the strict source validator; this is not a new model call or a formal/human proof certificate.
Proposition pt1 is locally supported by the full reviewed source. Two of the four findings (finding-4af4e397abeb37220130 and finding-412360ba8d7106cebee1) arose solely because Proposition p4, equation (e82), and equation (e56) were outside the local dependency packet supplied to the reviewer; all are present in earlier sections of the manuscript. A third finding (finding-e6edf021555063084d2b) concerns a remark after Proposition pt1 about bounding the ratio (ratio), which is not part of the proof of pt1. The remaining finding (finding-251abf8b3bfaabc76e27) correctly notes that the intermediate condition (e812), \log(1/\gamma) \ll a, if read as O(a) with an arbitrary constant, does not ensure convergence of the comparison series in (e811); however, this is a cosmetic imprecision because the authors use \ll to mean a sufficiently small implied constant and state Proposition pt1 under the stronger hypothesis (e814), \log(1/\gamma) \ll a / \log a, which ensures convergence and allows the polynomial factor |a|^C in (e89) to be absorbed by reducing the generic constant c > 0.
Statement assessment
locally_supported
Fix a square-free modulus q with (q, q_0) = 1 and an element \xi \in SL_2(q). By the orthogonal decomposition L^2(SL_2(q)) = \mathbb{R} \oplus \bigoplus_{1 \ne q_1 | q} E_{q_1} of (e82), the indicator 1_{g=\xi} decomposes as |SL_2(q)|^{-1} + \sum_{1 \ne q_1 | q} \varphi_{q_1}, where \varphi_{q_1} = P_{q_1}(1_{g=\xi}) \in E_{q_1} satisfies |\varphi_{q_1}|{l^2(SL_2(q_1))} \le |SL_2(q_1)| / |SL_2(q)| = q_1^{O(1)} / |SL_2(q)|. Summing over y \in \Lambda with d_H(0, yw) \le a + t, convolving with k\gamma(t), and applying Proposition p4 (with a shifted by d_H(0,w)) to the constant component and Proposition p2 to each divisor 1 \ne q_1 | q gives (e87)-(e88). Splitting the divisor sum at q_1 < |a|/\gamma and q_1 \ge |a|/\gamma bounds (e88) by (\gamma^{-C} |a|^C / |SL_2(q)|) e^{\delta a} \sum_{1 < q_1 < |a|/\gamma} e^{-ca \log q_1 / \log(a/\gamma)} + \gamma^{-C} q^C e^{(\delta - c)a}. Under (e814), \log(1/\gamma) \ll a / \log a (with a -> \infty, 0 < \gamma \le 1, and a sufficiently small implied constant), the exponent u = ca / \log(a/\gamma) satisfies u \gg \log a -> \infty. Thus \sum_{q_1 | q, q_1 > 1} q_1^{-u} \le \exp(\sum_{s=2}^\infty s^{-u}) - 1 = O(2^{-u}) = O(e^{-(c \log 2) a / \log(a/\gamma)}), and for a sufficiently small implied constant in (e814), the polynomial factor |a|^C is absorbed by reducing c > 0, yielding (e813).
Findings
finding-4af4e397abeb37220130
Returned disposition: packet_context; returned severity: none.
Proposition p4 was omitted only from the automated candidate-dependency packet, not from the manuscript. In the full reviewed source, Section 8 (labeled s7) explicitly states and proves Proposition p4: \int k_\gamma(t) N(a+t) dt = C_0 e^{\delta a} + o(\gamma^{-3} e^{(\delta - \varepsilon_6)a}). Shifting a by d_H(0,w) to pass from d_H(0,yw) - d_H(0,w) \le a in (e75) to d_H(0,yw) \le a in (e84) and dividing by |SL_2(q)| directly yields (e87).
Hypothesis check: In the full reviewed source, Proposition p4 is stated and proved in Section 8 (s7), equation (e711), for the lattice counting function N(a) of (e75). Replacing a by a - d_H(0,w) as noted after (e81) converts C_0 e^{\delta(a - d_H(0,w))} + o(\gamma^{-3} e^{(\delta - \varepsilon_6)(a - d_H(0,w))}) into C_1 e^{\delta a} + o(\gamma^{-3} e^{-ca} e^{\delta a}) with C_1 = C_0 e^{-\delta d_H(0,w)} and c = \varepsilon_6 > 0, matching (e87) after division by |SL_2(q)|.
Original quotation:
We use Proposition \ref{p4} to evaluate the right-hand side of
\eqref{e84} and Proposition \ref{p2} to bound terms \eqref{e85}.
Later-edition disposition: persists.
In the later edition (Section 12, p. 284), Proposition p4 appears as Proposition 10.2 (Section 10, p. 282) and Proposition p2 appears as Proposition 9.2 (Section 9, p. 280), with the same deduction of the main term and error bound.
Later-edition quotation:
We use Proposition 10.2 to evaluate S and Proposition 9.2 to bound T.
finding-412360ba8d7106cebee1
Returned disposition: packet_context; returned severity: none.
The items flagged as missing are present in earlier sections of the full reviewed source. Equation (e82) and the definitions of E_{q_1} and P_{q_1} appear in Section 5 (s2) just before Lemma l1q; equation (e56) appears in Section 4; and the unnormalized l^2 counting norm on SL_2(q) is defined in Sections 4 and 5. For \xi \in SL_2(q) and 1 \ne q_1 | q, the projection of 1_{g=\xi} to functions mod q_1 is supported on the single element \pi_{q_1}(\xi) \in SL_2(q_1) with value |SL_2(q_1)|/|SL_2(q)| (fibers of \pi_{q_1} have size |SL_2(q)|/|SL_2(q_1)|). Even if \varphi_{q_1} = P_{q_1}(1_{g=\xi}) is measured in the unnormalized l^2(SL_2(q_1)) counting norm (where its norm is bounded by |SL_2(q_1)|/|SL_2(q)|) or with |SL_2(q_1)|^{1/2}/|SL_2(q)| as written in (e86), the polynomial factor |SL_2(q_1)|^{O(1)} is absorbed into q_1^C in (e88)-(e89) because C is a generic positive constant.
Hypothesis check: Section 5 (s2) assumes q is square-free, and Section 9 (s8) fixes q square-free with (q, q_0) = 1. Every divisor q_1 | q inherits these hypotheses. Equation (e56) in Section 4 (s1/s6) defines N(a) for functions on SL_2(q), and shifting a by d_H(0,w) matches the summation condition d_H(0,yw) \le a in (e84)-(e85).
Original quotation:
Recall the decomposition of the space $L^2(\SL_2(q))$ in
\eqref{e82}.
Later-edition disposition: persists.
The later edition (Section 12, p. 284) retains the identical decomposition reference (pointing to (7.11) on p. 273) and the same norm estimate.
Later-edition quotation:
Recall the decomposition of the space L2(SL2(q)) in (7.11).
finding-251abf8b3bfaabc76e27
Returned disposition: confirmed_error; returned severity: cosmetic.
If the intermediate condition (e812), \log(1/\gamma) \ll a, is read in standard Vinogradov notation as \log(1/\gamma) = O(a) with an unrestricted implied constant, then choosing \gamma = e^{-Ka} with K \ge c gives u = ca / \log(a/\gamma) \le c/K \le 1, for which the comparison series \sum_{s=2}^\infty s^{-u} in (e811) diverges. Furthermore, the last inequality of (e811) and the absorption of the |a|^C prefactor from (e89) into (e813) implicitly replace c by a smaller positive constant c' < c \log 2 and require \log(a/\gamma) \le \varepsilon_0 a / \log a with \varepsilon_0 small enough that |a|^C e^{-(c \log 2 - c') a / \log(a/\gamma)} = o(1). Because the authors use \ll in the sense of a sufficiently small implied constant (as in Section 6 above Lemma props3) and immediately state Proposition pt1 under the stronger condition (e814), \log(1/\gamma) \ll a / \log a, this imprecision in the intermediate condition (e812) and constant renaming is cosmetic and does not invalidate Proposition pt1.
Hypothesis check: Proposition pt1 explicitly imposes the stronger hypothesis (e814): \log(1/\gamma) \ll a / \log a (with a -> \infty and 0 < \gamma \le 1, where throughout Sections 6 and 9 the authors use \ll to denote a sufficiently small implied constant, cf. the case distinction before Lemma props3). Under (e814) with a sufficiently small implied constant, u = ca / \log(a/\gamma) \ge (c/\varepsilon_0) \log a -> \infty, so the series in (e811) converges and the polynomial factor |a|^C in (e89) is absorbed by reducing the constant c > 0.
Original quotation:
\label{e812} \log
\frac{1}{\gamma} \ll a.
Later-edition disposition: persists.
The later edition (Section 12, p. 284-285) retains the exact same intermediate condition \log(1/\gamma) \ll a after the series estimate and the stronger condition \log(1/\gamma) \ll a / \log a in Proposition 12.1.
Later-edition quotation:
assuming
log 1
γ ≪a.
finding-e6edf021555063084d2b
Returned disposition: packet_context; returned severity: none.
As the review notes, the quoted passage is part of a remark following Proposition pt1 (discussing an alternative bound for the smoothed ratio (ratio) in sieving applications) and was only linked to pt1 by the automated dependency map. It is not used in the proof of Proposition pt1, which instead relies on Propositions p2 and p4.
Hypothesis check: The quoted sentence occurs in the remark after the end of Proposition pt1 and its proof. It is not a hypothesis or proof step of Proposition pt1.
Original quotation:
To bound the ratio \eqref{ratio} it suffices to use Proposition
\ref{p2} (which builds crucially on the generalized expansion
result given by Lemma \ref{l1q}) combined with the result of
Lalley \cite{l89}.
Later-edition disposition: persists.
The later edition (Section 12, p. 285) keeps this post-proposition remark unchanged after Proposition 12.1.
Later-edition quotation:
To bound the ratio (12.1) it suffices to use Proposition 9.2 (which builds crucially on the
generalized expansion result given by Lemma 7.2) combined with the result of Lalley [16].
Visible independent arbitration — t:cs
Model: gemini-4-argon; effort high. Original delivery: rejected by schema validation. This readable record preserves the returned mathematical text. A separately logged metadata-normalized copy passes the strict source validator; this is not a new model call or a formal/human proof certificate.
Theorem 13.1 (t:cs) states an abstract combinatorial sieve bound of order X / (log X)^t under axioms (A_0), (A_1), and (A_2) with z = D^{1/s} and s > 9t, but omits the hypothesis that D is at least a positive power of X (or writes log X instead of log z in the conclusion). Under the intended prime-modulus condition beta(p) <= 1 - 1/c_1 in (A_0), an elementary counterexample with z = log log X and D = z^{10} satisfies (A_0)-(A_2) while S(A, P_z) >> X / log log X, refuting the literal upper bound X / (log X)^t. In addition, (A_0) contains a typographical error writing beta(d) instead of beta(p) when quantifying over p not in B, and no proof or specific theorem citation from Halberstam-Richert or Iwaniec-Kowalski is supplied in the packet to verify the threshold s > 9t. The attached Subsection 13.2 (Applying the sieve), which applies Theorem 13.1 to prove Theorems 1.6 and 1.7 (t2 and t22), also contains three confirmed errors: an inverted fraction in the exponent condition tau < 2delta / ((C+3)varepsilon_1) instead of tau < varepsilon_1 / (2delta(C+3)); an invalid uniform composite-modulus bound |Lambda_d^f| << d^2 (and beta(d) << 1/d) that fails when t(f) > 1 due to the multiplicative factor t(f)^{omega(d)}; and a coset-summation identity (consum1) that replaces the coset condition gamma = rho (mod d) with the principal congruence subgroup gamma in Lambda(d) and omits the constant c_Lambda in the intermediate display. All of these issues persist in the 2011 Acta Mathematica edition.
Statement assessment
refuted
Taken under the intended prime-modulus condition beta(p) <= 1 - 1/c_1 in (A_0) (fixing the literal typo beta(d) <= 1 - 1/c_1, which is impossible at d = 1 since beta(1) = 1 for any nonzero multiplicative function), the literal statement of Theorem 13.1 (t:cs) is refuted by an elementary counterexample because neither (A_0)-(A_2) nor Theorem 13.1 imposes a lower bound on log D (or log z) in terms of log X. Specifically, let a_n = 1 for 1 <= n <= N and a_n = 0 for n > N, so X = N. Take B = empty, beta(d) = 1/d, c_1 = 2, t = 1, s = 10 > 9t, z = log log N, and D = z^{10} = (log log N)^{10} < X. For every square-free d < X, sum_{n = 0 (mod d)} a_n = floor(N/d) = X/d + r(A, d) with |r(A, d)| <= 1, and beta(p) = 1/p <= 1/2 for all primes p, verifying (A_0). Moreover, sum_{d <= D} |r(A, d)| <= D = (log log X)^{10} << X/(log X)^B for every fixed B > 0, verifying (A_1). By elementary Chebyshev estimates (theta(u) = O(u) from binomial coefficients and log(m!) = sum_{k <= m} Lambda(k) floor(m/k), giving sum_{p <= u} (log p)/p = log u + O(1)), (A_2) holds with t = 1. However, by inclusion-exclusion, S(A, P_z) = sum_{d | P_z} mu(d) floor(N/d) >= N prod_{p <= z} (1 - 1/p) - 2^{pi(z)} >= N/floor(z) - 2^z ~ X / log log X, where prod_{p <= z} (1 - 1/p) >= prod_{m=2}^{floor(z)} (1 - 1/m) = 1/floor(z). This violates the claimed upper bound S(A, P_z) << X / (log X)^t = X / log X. (The conclusion in (13.5) holds only when log z asymp log X, i.e., when D >= X^tau for some fixed tau > 0, or if log X in the denominator of (13.5) is replaced by log z or log D.)
Findings
finding-3e4cd9947efd9c8b9625
Returned disposition: confirmed_error; returned severity: substantive.
Under the hypotheses (A_0), (A_1), and (A_2) with z = D^{1/s} and s > 9t, the main term produced by the combinatorial sieve is of order X / (log z)^t asymp_s X / (log D)^t, not X / (log X)^t. Although the narrative remarks parenthetically that 'in our applications z will be a small power of X', neither (A_0)-(A_2) nor Theorem 13.1 imposes any lower bound on D (such as D >= X^tau for a fixed tau > 0) relating log D to log X. The reviewer's self-contained counterexample (a_n = 1 for 1 <= n <= N, X = N, B = empty, beta(d) = 1/d, t = 1, s = 10, z = log log N, D = z^{10}) satisfies (A_0)-(A_2) for all large X, yet S(A, P_z) >= X / floor(z) - 2^z ~ X / log log X, contradicting S(A, P_z) << X / log X.
Hypothesis check: Checked all hypotheses of Theorem 13.1 and axioms (A_0), (A_1), and (A_2): (A_1) requires only D = D(X) < X and sum_{d <= D} |r(d, A)| << X / (log X)^B for all B > 0, which places an upper bound on D in practice but no lower bound on D relative to X.
Original quotation:
\frac{X}{(\log X)^t} \ll S(A, P_z) \ll
\frac{X}{(\log X)^t}
Later-edition disposition: persists.
Theorem 13.1 in the 2011 Acta Mathematica edition retains the exact same statement and denominator (log X)^t without adding a hypothesis that D >= X^tau or replacing log X by log z.
Later-edition quotation:
Theorem 13.1. Assume (A0), (A 1) and (A2). For s>9t, z=D1/s and X large we
have
X
(log X)t ≪S(A, Pz) ≪ X
(log X)t .
finding-7f3fec4e224b6b09fa7f
Returned disposition: confirmed_error; returned severity: cosmetic.
In axiom (A_0), the condition quantifies over primes p not in B but writes beta(d) <= 1 - 1/c_1 instead of beta(p) <= 1 - 1/c_1. Since beta(d) is a nonzero multiplicative function, beta(1) = 1, so the bound cannot hold at d = 1; it is a typographical error for the standard prime-modulus condition beta(p) <= 1 - 1/c_1.
Hypothesis check: Checked (A_0): beta(d) is declared multiplicative on square-free d < X with no prime factors in B, and the quantifier 'for p notin B' confirms that beta(p) was intended.
Original quotation:
\text{for}\, p \notin B, \beta(d) \leq 1-\frac{1}{c_1}
Later-edition disposition: persists.
The typographical error writing beta(d) instead of beta(p) after 'for p notin B' remains unchanged on page 286 of the later edition.
Later-edition quotation:
where β(d) is a multiplicative function of d and, for p /∈B,
β(d) ⩽1− 1
c1
for a fixed c1.
finding-7ffb4d924eb7c80486fb
Returned disposition: confirmed_error; returned severity: substantive.
In Subsection 13.2 (Applying the sieve, which applies Theorem 13.1 to prove Theorem 1.6), the bound sum_{d <= D} |r(A, d)| << X / (log X)^B requires D^{C+3} X^{1 - varepsilon_1 / (2delta)} << X / (log X)^B. Substituting D = X^tau requires (C+3)tau - varepsilon_1 / (2delta) < 0, i.e., tau < varepsilon_1 / (2delta*(C+3)). The printed condition tau < 2delta / ((C+3)varepsilon_1) inverts the factor varepsilon_1 / (2delta); for small varepsilon_1 > 0, values of tau near 2delta / ((C+3)varepsilon_1) make D^{C+3} X^{1 - varepsilon_1 / (2delta)} >> X, failing to establish (A_1). In addition, summing 1/d over d <= D introduces a factor O(log D) in the first remainder term, which was omitted from the display (though absorbed by X^{-1/(2deltalog log X)}).
Hypothesis check: Checked the remainder estimate r(A, d) << d^{-1} X^{1 - 1/(2deltalog log X)} + d^{C+2} X^{1 - varepsilon_1/(2delta)} in Subsection 13.2; the exponent of X in the second term is 1 - varepsilon_1/(2delta), so the power-saving threshold for D <= X^tau is tau < varepsilon_1 / (2delta(C+3)).
Original quotation:
\tau < \frac{2 \delta}{(C+3) \varepsilon_1}
Later-edition disposition: persists.
Page 288 of the later edition retains the inverted threshold tau < 2*delta / ((C+3)*varepsilon_1) and the omitted log D factor.
Later-edition quotation:
D ⩽Xτ with τ < 2δ
(C+3)ε1
.
finding-18f9cea00004dba2595b
Returned disposition: confirmed_error; returned severity: substantive.
In Subsection 13.2, equation (ontr) states |Lambda_d^f| / |Lambda_d| = t(f)/p + O(p^{-3/2}), mixing d on the left with a prime p on the right. For a prime p outside S(Lambda), |Lambda_p| = |SL_2(F_p)| = p(p^2 - 1), so |Lambda_p^f| = t(f) p^2 + O(p^{3/2}). By the Chinese Remainder Theorem and surjectivity of Lambda -> prod_{p | d} Lambda_p for square-free d coprime to S(Lambda), |Lambda_d^f| = prod_{p | d} |Lambda_p^f| = d^2 prod_{p | d} (t(f) + O(p^{-1/2})). When the number of irreducible factors t(f) is strictly greater than 1, this product grows like t(f)^{omega(d)}, which is unbounded over square-free d. Consequently, neither the uniform composite bound |Lambda_d^f| << d^2 in (ldfb) nor the subsequent bound beta(d) = |Lambda_d^f| / |Lambda_d| << 1/d used in r(A, d) holds for t(f) > 1 (one instead has |Lambda_d^f| <<_varepsilon d^{2+varepsilon} and beta(d) <<varepsilon d^{-1+varepsilon}, or |Lambda_d^f| <= d^3 and sum{d <= D} beta(d) << (log D)^{t(f)}).
Hypothesis check: Checked the hypotheses of Theorem 1.6 (t2) and Subsection 13.2: f in Z[x_{ij}] is weakly primitive with t(f) irreducible factors, where t(f) >= 1 is arbitrary and not restricted to t(f) = 1.
Original quotation:
|\Lambda_d^f| \ll d^2
Later-edition disposition: persists.
Page 287 of the later edition retains |Lambda_d^f| << d^2, the mixed d/p notation in (13.1), and the bound r(A, d) << (1/d) X^{1 - 1/(2deltalog log X)} + d^{C+2} X^{1 - varepsilon_1/(2*delta)} on page 288.
Later-edition quotation:
Using the Lang–Weil theorem [17] as in [3], we obtain
|Λf
d| ≪d2
and
|Λf
d|
|Λd| = t(f)
p +O(p−3/2). (13.1)
finding-9af7765b927585c38b04
Returned disposition: confirmed_error; returned severity: substantive.
In Subsection 13.2, equation (consum1) decomposes the sum over {gamma in Lambda : |gamma| <= T, f(gamma) = 0 (mod d)} into a sum over residue classes rho in Lambda / Lambda(d) with f(rho) = 0 (mod d), but writes the inner sum over {gamma in Lambda(d) : |gamma| <= T} instead of the coset {gamma in Lambda : gamma = rho (mod d), |gamma| <= T}. Because the archimedean norm |gamma| is not invariant under coset translation, the inner count depends on the coset rho and does not equal the count in the principal congruence subgroup Lambda(d). Furthermore, when applying Theorem 1.5 (t5b) in the next display, the authors drop the constant c_Lambda in the first line of the split equation and write the error term O(d^C T^{2*delta - varepsilon_1}) outside the sum over rho before multiplying it by |Lambda_d^f| in the second line. (Theorem 1.5 actually applies to each individual coset gamma = rho (mod d), which would immediately repair the decomposition once the inner sum and constant c_Lambda are written properly.)
Hypothesis check: Checked Theorem 1.5 (t5b) and (consum1): Theorem 1.5 provides the asymptotic count for any residue class gamma = g (mod q) with main term c_Lambda T^{2*delta} / |SL_2(q)|, assuming q is square-free and (q, q_0) = 1.
Original quotation:
\sum_{\substack{\gamma \in \Lambda(d) \\
|\gamma| \leq T}}1
Later-edition disposition: persists.
Page 287 of the later edition retains the inner sum over gamma in Lambda(d) instead of gamma = varrho (mod d) as well as the omitted c_Lambda in the intermediate display.
Later-edition quotation:
∑
n≡0 (mod d)
an(T) =
∑
γ∈Λ
|γ|⩽T
f(γ)≡0 (mod d)
1 =
∑
ϱ∈Λ/Λ(d)
f(ϱ)≡0 (mod d)
∑
γ∈Λ(d)
|γ|⩽T
1.
finding-dca9e3ab0b340b4f439f
Returned disposition: unresolved; returned severity: none.
The manuscript states Theorem 13.1 without a self-contained proof, citing [BGS], [iw] (Iwaniec-Kowalski, 2004), and [hr] (Halberstam-Richert, 1974) without specific theorem or page numbers. Because the texts of these external references are not supplied in the review packet, the exact hypotheses of the Fundamental Lemma and the sufficiency of the constant threshold s > 9t (even after correcting log X to log z or adding D >= X^tau) cannot be verified from the packet alone.
Hypothesis check: Checked the bibliography and Section 13.1: [hr] (1974) and [iw] (2004) predate the 2009-12-29 cutoff, while [BGS] is listed as 'to appear' without a pre-cutoff date in the reviewed source; none of the external texts are included in the packet.
Original quotation:
the simplest
combinatorial sieve which is turn is based on the Fundamental
Lemma in the theory of elementary sieve
Later-edition disposition: persists.
Page 285 of the later edition fixes the grammatical typo ('which in turn is based') and updates the bibliography entry [3] to Invent. Math. 179 (2010), 559-644, but still gives only unspecific book citations [12] and [13] without a self-contained proof.
Later-edition quotation:
As in [3], we will make use of the simplest combinatorial sieve which in turn is based
on the fundamental lemma in the theory of elementary sieves, see [12] and [13].
Additional supporting archive items are identified in the original report. View the manuscript hashes and repository snapshot.