REVIEW 5 major objections 4 minor 4 references
The Fine-Structure Hierarchy of Prime Biases and the Universal Dominance of $-1 \pmod N$
T0 review · 5 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash
Pith's one-line read Under DRH, a Gaussian-mollified spectral formula fixes the fine-structure bias of primes in reduced residue classes to C_N log L(1,χ_{1,a}), yielding a deterministic hierarchy with -1 (mod N) uniformly dominant.
desk verdict The main theorem is not proven; Eq (2.5) cancels the claimed bias, though the virtual-character idea and the -1 dominance conjecture are worth a second look. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the spectrally normalized, Gaussian-mollified explicit formula: the test function $h(\gamma) = \frac{\cos(k\gamma \log p) e^{-(\gamma/T)^2}}{p^{k/2}}$ localizes the zero sum around prime powers with Schwartz-class decay, allowing the order of summation over prime powers and zeros to be interchanged. Together with the virtual character $\chi_{1,a}(x) = 1_{x\equiv 1} - 1_{x\equiv a}$, this makes the principal character cancel exactly, reduces the double sum to diagonal terms $p^k = q^m$, and connects the result to the linear combination $\log L(1, \chi_{1,a})$.
What would settle it
With $T$ fixed, compute the actual contribution of near-resonant pairs $(p^k, q^m)$ with $|m \log q - k \log p| \le 1/T$ at, say, $p^k, q^m \le 10^6$; if these off-diagonal contributions are not exponentially small in $T^2$, the reduction to the diagonal fails. A second check is to numerically evaluate $\tilde{S}_T(x, \chi_{1,7}) - C_8 \log L(1,\chi_{1,7})$ for increasing $x$ and test whether the discrepancy decays like $(\log x)/\sqrt{x}$ rather than persisting at a larger order.
Extended reading notes
Core claim
Under the Deep Riemann Hypothesis, for any fixed spectral scale $T$ the difference between the mollified prime-power sums for $1$ and $a$ (mod $N$), normalized by $\log x / \sqrt{x}$, equals $C_N \log L(1, \chi_{1,a}) + O((\log x)/\sqrt{x})$. The leading growth and the principal-character noise cancel identically because the virtual character $\chi_{1,a}$ has coefficient $1 - \chi_0(a) = 0$, leaving only non-principal L-series special values. Hence the bias between any two reduced residue classes is asymptotically a fixed, computable constant; for $N=8$ the paper derives $7 > 3 > 5 > 1 \pmod{8}$, and by Remark 1.6 the class $-1 \pmod{N}$ is universally dominant.
Load-bearing premise
The load-bearing step is the claim that, after interchanging the two sums, every matching of a prime power in the zero-sum with a different prime power in the spatial sum is exponentially negligible ($O(e^{-cT^2})$), leaving only exact diagonal matches; without that, the constant $C_N \log L(1,\chi_{1,a})$ is not reached.
Editorial extensions
If this is right
- Prime races among classes with identical quadratic character are not equiprobable; for example, asymptotically 7 > 3 > 5 > 1 (mod 8).
- The classical Chebyshev bias, such as 3 > 1 (mod 4), becomes a special case of the same formula.
- For any modulus N, the full ranking of reduced residue classes is computable from the special values L(1,χ), so the hierarchy is deterministic rather than probabilistic.
- The residue class -1 (mod N) is universally dominant, because odd characters align constructively (factor 2) while even characters cancel.
- Raw prime-count fluctuations at moderate scales, such as x = 1.3 × 10^13, can temporarily invert the ranking due to low-lying complex zeros; the regularized sums filter out this transient noise.
Reading between the lines
- The same mollified-explicit-formula construction might expose analogous fine-structure biases in other arithmetic functions that admit explicit formulas, such as divisor sums or prime values of polynomials.
- The proof only needs diagonal dominance and cancellation of the principal character; if off-diagonal control can be obtained under GRH rather than DRH, the same hierarchy would follow with weaker input.
- Since the universal dominance of -1 (mod N) is stated as Conjecture 1.7 and verified numerically, a direct stress test is to search for moduli where an even character's L(1,χ) is exceptionally small; such a case would challenge the conjecture.
- The constant C_N is asserted to be explicit but no closed form is displayed; extracting it would make the predicted ranking quantitatively testable.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes a 'fine-structure hierarchy' of prime biases among residue classes modulo N after removing the leading prime-square effect. It introduces a Gaussian-mollified prime power sum S_T(x,a), a spectral normalization, and a virtual character χ_{1,a} to isolate differences between the classes 1 and a (mod N). Under a novel 'Deep Riemann Hypothesis' (DRH), Theorem 1.4 claims that the normalized bias difference satisfies eS_T(x,χ_{1,a}) = C_N log L(1,χ_{1,a}) + O((log x)/√x), leading to deterministic rankings such as 7>3>5>1 (mod 8) and the universal dominance of -1 (mod N). The proof rests on Weil's explicit formula, a diagonalization of the double sum over prime powers, and an algebraic cancellation of the principal character.
Significance. If the main theorem were correct, it would be a striking refinement of Rubinstein–Sarnak prime number races, showing that secondary biases among residue classes with identical quadratic character are deterministic and governed by the special values L(1,χ). The paper also presents numerical tables in support of the conjectured hierarchy, and it correctly identifies a genuine gap in the classical Chebyshev-bias framework. However, the central proof is internally inconsistent: the only explicit diagonal evaluation in Eq. (2.5) is positive and vanishes after the spectral normalization, directly contradicting the claimed nonzero constant in Theorem 1.4. Moreover, DRH is never precisely formulated, the off-diagonal decay is asserted without proof, and the key constant C_N is never computed. No machine-checked proofs or reproducible code are provided. The numerical tables are not accompanied by algorithms or accuracy bounds, limiting their evidentiary value.
major comments (5)
- [§2.2, Eq. (2.5)] The diagonal contribution computed in (2.5) is incompatible with Theorem 1.4. For the virtual character χ_{1,a} ∈ {0,±1}, one has |χ_{1,a}(p)|^{2k} + χ_{1,a}(p)^{2k} = 2 when p ≡ 1 or a (mod N) and 0 otherwise. Thus the k=1 part of (2.5) is (T/(2√π)) Σ_{p≤x, p≡1,a} log p / p ∼ (T/(2√π φ(N))) log x. After the normalization log x/√x, this contribution is O((log x)^2/√x) → 0. For N=4, the theorem predicts the negative limit C_4 log(π/4); the diagonal is positive and tends to 0. Since §2.2 explicitly discards all off-diagonal terms, there is no remaining source for the asserted constant.
- [§2.2] The reduction to the diagonal rests on the claim that off-diagonal terms are O(e^{-cT^2}). This is not established and is generally false for fixed T. There are infinitely many pairs of prime powers p^k and q^m with |k log p − m log q| arbitrarily small (e.g., by density of the multiplicative subgroup generated by primes), so the Gaussian factors in Eq. (2.2) need not be exponentially small. The stated justification, 'exponential decay of h(γ)', concerns decay in the spectral variable γ, not uniform decay in the prime-power summation. No rigorous interchange of summation or tail estimate is provided.
- [§2.4, Eq. (2.11)] The main formula is announced rather than derived. The only explicit evaluation before (2.11) is Eq. (2.5), which is a positive sum with no dependence on L(1,χ). The text then jumps to eS_T(x,χ_{1,a}) = C_N log L(1,χ_{1,a}) after noting the principal-character cancellation (2.10). That cancellation removes χ_0 but does not produce the special-value term; no argument connects the diagonal sum or any surviving spectral term to log L(1,χ). Equation (2.11) is therefore unsupported by the preceding calculation.
- [§2.3, Lemma 2.1] Lemma 2.1's claim that the Archimedean contribution is O((log x)/√x) is false under the paper's own bound (2.9), which gives I_Γ(T,χ) = O(T log T / p^{k/2}). For k=1, Σ_{p≤x} 1/√p ∼ 2√x / log x. Substituting into (2.8) yields a normalized contribution of constant size O(T log T), not O((log x)/√x). DRH cannot alter the size of this sum over primes in fixed residue classes. Thus the Archimedean place is not negligible at the asserted order, contradicting the proof of Theorem 1.4's error term.
- [§1.1, Footnote 1] The Deep Riemann Hypothesis (DRH) is never stated as a precise mathematical hypothesis. Footnote 1 describes it only as a 'stronger assertion than GRH' that 'dictates bounded phase oscillations and asymptotic convergence of Euler products on the critical line' and cites only the author's prior work [1]. Because Theorem 1.4 and all corollaries are conditional on DRH, the theorem is not a well-defined conditional statement unless the hypothesis is formulated with sufficient rigor (e.g., as an explicit statement about Dirichlet L-functions and their Euler products). This is load-bearing, not a presentation issue.
minor comments (4)
- [§1.4 vs §1.5] The ranking for N=8 is internally inconsistent. Corollary 1.5 uses log L(1,χ_{1,7}) < log L(1,χ_{1,3}) < log L(1,χ_{1,5}) to obtain 7>3>5>1, but §1.5 lists log L(1,χ_{1,5}) = 0.21008 and log L(1,χ_{1,3}) = 0.24647, which gives log L(1,χ_{1,5}) < log L(1,χ_{1,3}) and hence 7>5>3. Table 4 repeats the 7>3>5>1 ordering. The numerical values contradict the stated hierarchy.
- [Definition 1.1 / Eq. (1.2)] The notation h(γ/(2π)) in (1.2) is inconsistent with h(γ) defined in (1.1). If h is as in (1.1), replacing γ by γ/(2π) changes both the Gaussian width and the cosine frequency. Section 2 silently drops the factor 2π. The convention should be fixed and used consistently throughout.
- [§1.5, Tables 1–2] The numerical values of log L(1,χ_{1,a}) are presented without explanation of how they were computed. No algorithm, code, or numerical precision bounds are given. Since Corollary 1.5 depends on the ordering of these values, reproducibility is limited.
- [§2.4, Remark 2.2] The phrase '100% analytical rigor' is not substantiated and is incompatible with the unproved off-diagonal bound in §2.2. It is also inconsistent with the paper's own admission that DRH is needed to control non-absolute convergence.
Circularity Check
DRH premise is sourced solely to the authors' own prior paper; the main formula is announced rather than derived, but no fitted-input circularity is present.
-
self citation load bearing
[§1.2 and footnote 1; used in Theorem 1.4]
"The Deep Riemann Hypothesis (DRH), as introduced and investigated in recent literature on prime biases (e.g., [1]), is a stronger assertion than the Generalized Riemann Hypothesis (GRH)... The inspiration for employing the Deep Riemann Hypothesis (DRH) to unravel prime distribution biases originates from the recent pioneering work of Aoki and Koyama [1]."
DRH is the central unproved premise underlying the regularized explicit formula and Theorem 1.4. The only source cited for DRH is [1], co-authored by the present author; the paper provides no independent statement, proof, or external reference. Every subsequent ranking (e.g., 7>3>5>1 mod 8) is conditional on an assumption whose authority traces back to the authors' own prior work, making the self-citation load-bearing for the paper's central premise.
full rationale
No fitted-input circularity is exhibited: the constants log L(1,chi) are externally computed special values, not parameters fit to the prime-bias data, and the claimed hierarchy is a genuine assertion about those values. The main derivation is nevertheless not self-contained: §2.2 evaluates a diagonal contribution (Eq. 2.5), §2.4 cancels the principal character by the algebraic identity 1-chi0(a)=0 (Eq. 2.10), and then states 'As x -> infinity, we arrive directly at the main formula' (Eq. 2.11) without showing how Eq. 2.5 or the Archimedean bounds produce C_N log L(1,chi_{1,a}). This is an omitted/unsupported derivation, and the skeptic's observation that the diagonal is positive while the N=4 limit is negative indicates a correctness problem, not a circular reduction. The only circularity-adjacent flaw is the self-citation of DRH to the authors' own paper; hence score 4.
Assumptions & free parameters
free parameters (1)
- Gaussian mollifier width T =
arbitrary fixed > 0
assumptions (3)
- ad hoc to paper Deep Riemann Hypothesis (DRH) for all Dirichlet L-functions modulo N
- ad hoc to paper Interchange of summation and exponential decay of off-diagonal terms in the explicit formula
- ad hoc to paper Final step equating the regularized sum to C_N log L(1,\chi_{1,a})
invented entities (1)
-
Deep Riemann Hypothesis (DRH)
Cite this review
Pith. "Pith review of The Fine-Structure Hierarchy of Prime Biases and the Universal Dominance of $-1 \pmod N$." pith.science (2026). https://pith.science/paper/5OEV7CUN
@misc{pith2026260728931,
author = {Pith},
title = {Pith review of: The Fine-Structure Hierarchy of Prime Biases and the Universal Dominance of $-1 \pmod N$},
year = {2026},
howpublished = {\url{https://pith.science/paper/5OEV7CUN}},
note = {Machine review of arXiv:2607.28931}
}
abstract
We investigate the deterministic hierarchy of prime distribution biases in arithmetic progressions modulo $N$ using a regularized spectral approach. Classical studies on Chebyshev's bias attribute prime races primarily to the accumulation of prime squares $p^2 \equiv 1 \pmod N$, which creates a systematic deficit in quadratic residue classes. However, this classical mechanism fails to explain or distinguish any bias among residue classes sharing identical quadratic residue status (e.g., $3, 5, 7 \pmod 8$). To overcome the long-standing analytical obstacles of jump discontinuities and non-convergent boundary fluctuations inherent in classical Perron-type step-function truncations, we introduce a smooth $C^\infty$ Gaussian mollifier into Weil's explicit formula for Dirichlet $L$-functions. By defining the spectrally normalized individual mollified sums $\widetilde S_T(x, a)$ and adopting the virtual character $\chi_{1,a}(x) := \mathbf{1}_{\{x \equiv 1 \pmod N\}} - \mathbf{1}_{\{x \equiv a \pmod N\}}$, the principal character component $\chi_0$ cancels identically since $1 - \overline{\chi_0}(a) = 0$. This automatic algebraic elimination erases both the universal logarithmic growth $\log x$ and the background noise $\log L(1, \chi^2)$. Under the Deep Riemann Hypothesis (DRH), we uncover a hitherto undetected \textbf{fine-structure bias} (or \emph{secondary bias}) strictly governed by the special values $\log L(1, \chi)$. We prove that $\widetilde S_T(x, \chi_{1,a}) := \widetilde S_T(x, 1) - \widetilde S_T(x, a) = C_N \cdot \log L(1, \chi_{1,a}) + \mathcal{O}((\log x)/\sqrt x)$ as $x \to \infty$, where $C_N > 0$ depends solely on $N$. Consequently, we establish a deterministic multi-way ranking (such as $7 > 3 > 5 > 1 \pmod 8$) that completely transcends the classical quadratic residue framework.
Reference graph
Works this paper leans on
-
[1]
Aoki and S
M. Aoki and S. Koyama,Chebyshev’s bias against splitting and principal primes in global fields, J. Number Theory245(2023), 233–262
2023
-
[2]
Feuerverger and G
A. Feuerverger and G. Martin,Biases in the prime number race, Experiment. Math.9(2000), no. 4, 535–570
2000
-
[3]
Iwaniec and E
H. Iwaniec and E. Kowalski,Analytic Number Theory, American Mathematical Society Colloquium Publications, vol. 53, American Mathematical Society, Providence, RI, 2004
2004
-
[4]
Rubinstein and P
M. Rubinstein and P. Sarnak,Chebyshev’s bias, Exp. Math.2(1994), no. 3, 173–197. Department of Mechanical Engineering, Toyo University, 2100 Kujirai, Kawagoe-shi, 350- 8585 Japan Email address:koyama@toyo.jp
1994
Reviewed August 3, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.