Deciphering the Critical Strip: Möbius Integral Bounds and the Reciprocal Zeta Function

Download Full Article

This article is available as a downloadable PDF with complete code listings and syntax highlighting.

Introduction

The Riemann Hypothesis (RH) remains the most significant unsolved problem in pure mathematics, asserting that all non-trivial zeros of the Riemann zeta function, ζ(s), lie on the critical line where the real part of s is equal to 1/2. While the hypothesis is often framed in terms of the distribution of prime numbers, an equivalent and analytically fertile formulation involves the Möbius function, μ(n), and its summatory function, the Mertens function, M(x). The research presented in arXiv:hal-01847199 explores the intricate relationships between the Mertens function, Dirichlet series, and the integral representations that bridge the gap between the half-plane of absolute convergence and the critical strip.

The core of the investigation in arXiv:hal-01847199 lies in establishing rigorous bounds for the remainder terms of the Dirichlet series for 1/ζ(s). By utilizing Abel's summation formula and the properties of square-free integers, the paper provides a framework for analyzing the growth of M(x). The significance of this approach is rooted in the fact that if M(x) grows no faster than x raised to the power of (1/2 + ε) for any ε > 0, the Riemann Hypothesis is true. This article provides a technical decomposition of these findings, examining the spectral-like properties of the Möbius function and the implications of the integral bounds derived in the source text.

This article situates the method of arXiv:hal-01847199 within the standard analytic framework, clarifies which steps are unconditional and which implicitly require RH-strength input, and proposes research pathways that preserve the paper’s core strategy while replacing fragile pointwise inequalities with averaged or mollified statements that are closer to established techniques.

Mathematical Background

To understand the contributions of arXiv:hal-01847199, we must first define the primary mathematical objects. The Möbius function, μ(n), is defined for all positive integers n. It is 1 if n = 1; (-1)^r if n is the product of r distinct primes; and 0 if n is divisible by a square other than 1. The Mertens function is the partial sum of this sequence: M(x) = sum for n ≤ x of μ(n).

The Riemann zeta function, ζ(s), where s = σ + it, is defined for σ > 1 by the series sum of n^-s. The reciprocal of the zeta function is given by the Dirichlet series: 1/ζ(s) = sum for n from 1 to infinity of μ(n) / n^s. This series converges absolutely for σ > 1. A critical tool in analytic number theory is the representation of this sum as an integral involving the Mertens function. Using Abel's summation formula, one can show that 1/ζ(s) = s * integral from 1 to infinity of M(u) / u^s+1 du.

The source paper arXiv:hal-01847199 focuses on the tail of this integral and its relation to the partial sums of the Möbius series. Specifically, it examines the term: |s * integral from x to infinity of M(u) / u^s+1 du| = |1/ζ(s) - sum for n ≤ x of μ(n)/n^s + M(x)/x^s|. Another essential structure is the distribution of square-free integers. The number of square-free integers up to x, denoted N(x), has a density of 6/π². The source paper relates the absolute value of the Möbius function to the zeta function via: sum of |μ(n)| / n^σ = ζ(σ) / ζ(2σ). This ratio plays a pivotal role in establishing the upper bounds for the fluctuations of the Dirichlet series as s approaches the critical line.

Main Technical Analysis: Spectral Properties and Zero Distribution

Analytic Continuation via Abel Summation

The technical core of arXiv:hal-01847199 is the derivation of precise inequalities that bound the difference between the reciprocal of the zeta function and its truncated Dirichlet series. By applying the formula for s * integral from x to infinity of M(u) / u^s+1 du, the paper establishes that this quantity is bounded by ζ(σ)/ζ(2σ) + m * x^β-σ. Here, β represents the supremum of the real parts of the zeros of the zeta function. If RH is true, β = 1/2.

The derivation starts with the identity: sum for n ≤ x of μ(n)/n^s = M(x)/x^s + s * integral from 1 to x of M(u)/u^s+1 du. As x tends to infinity, for σ > 1, the sum converges to 1/ζ(s). The paper focuses on the error term for finite x. The source demonstrates that the magnitude of the integral from x to infinity is constrained by the density of square-free numbers. This is significant because it suggests that the randomness of the Möbius function is sufficiently high to ensure that the integral does not grow uncontrollably as σ decreases toward 1/2.

The Key Difficulty: Lower Bounds and Phase Alignment

The most delicate displayed inequality in the source excerpt is of the form | sum μ(n)/n^σ+it | > (1 - ε θ(x)) sum |μ(n)|/n^σ. Conceptually, this is attempting to say: for suitable choice of x, the complex phases n^-it do not destroy the magnitude of the Möbius Dirichlet series, so the absolute value is almost as large as the sum of absolute values.

From the perspective of classical analytic number theory, this is the step that cannot hold uniformly in t without additional structure. The terms μ(n)/n^σ already oscillate in sign, and multiplication by n^-it adds a further phase rotation. A generic Dirichlet series behaves like a random walk in the complex plane as t varies. Lower bounds near the sum of absolute values require that the terms be nearly co-linear in argument, which is exceptional for arithmetic sequences. Thus, a pointwise inequality of the above "almost no cancellation" type is not consistent with standard distributional heuristics and would need a precise statement of exceptional sets of t, or an averaged version.

Asymptotic Estimates of the Remainder

A major contribution of arXiv:hal-01847199 is the refinement of the estimate for the sum of |μ(n)| / n^σ. The paper provides the asymptotic expansion: sum for n ≥ x of |μ(n)| / n^σ = (6 * x^1-σ) / (π² * (σ-1)) + O(x^1/2-σ). This formula is derived by treating the square-free count N(u) as 6u/π² + O(sqrt(u)). When substituted into the integral representation of the sum, it allows for a precise determination of how the mass of square-free integers is distributed as x increases. The fact that the remainder term is O(x^1/2-σ) is a strong hint toward the 1/2 exponent required by the Riemann Hypothesis.

Novel Research Pathways

Pathway 1: Weighted Integral Transforms and Moment Estimates

The research in arXiv:hal-01847199 utilizes the standard kernel u^-s-1 for the integral representation of 1/ζ(s). A promising extension would be to investigate weighted kernels, such as u^-s-1 * log(u)^k. These weights correspond to higher-order derivatives of 1/ζ(s). By analyzing the decay rates of these weighted integrals, one might establish a sequence of moments for M(x). Integrating |1/ζ(s)|^2k over horizontal strips using these bounds should yield improved exponents in zero-density theorems.

Pathway 2: Sieve-Theoretic Refinement of the Remainder Term

The error term O(x^1/2-σ) in the square-free sum is currently derived from the standard estimate for N(x). However, using modern sieve methods, it may be possible to refine the variance of |μ(n)| in short intervals. If the local density of square-free numbers can be shown to be more uniform than current bounds suggest, the resulting bound on the integral would be tightened. This would directly impact the "m" constant used in the source paper's primary inequality, potentially leading to a contradiction for any β > 1/2.

Pathway 3: Hybrid Analytical-Computational Zero Investigation

This direction combines theoretical insights from arXiv:hal-01847199 with computational exploration. The idea is to use the integral bounds to guide computational searches for zeros. When the bound |s * integral M(u)/u^s+1 du| becomes small, it indicates that 1/ζ(s) is large, suggesting proximity to a zero. Conversely, when the bound is large relative to the finite sum, it indicates regions where ζ(s) is bounded away from zero. This feedback loop could potentially lead to the verification of RH for zeros up to unprecedented heights.

Computational Implementation

The following Wolfram Language code provides a framework for testing the Abel-tail identity and visualizing the integral tail magnitudes as discussed in the source paper.

(* Section: Abel-Tail Decomposition for 1/Zeta(s) *)
(* Purpose: Test 1/Zeta(s) = Sum mu(n)/n^s + TailIntegral - M(x)/x^s *)

ClearAll["Global`*"];
mertensM[x_Integer] := Total[MoebiusMu[Range[x]]];
finiteMuSum[s_, x_Integer] := Sum[MoebiusMu[n]/n^s, {n, 1, x}];

(* Numerical approximation of the tail integral using a step function *)
tailIntegralApprox[s_, x_Integer, umax_Integer] := Module[
  {sigma = Re[s], t = Im[s], integrand},
  integrand[u_?NumericQ] := (s * mertensM[Floor[u]]) * u^(-s - 1);
  NIntegrate[integrand[u], {u, x, umax}, 
    Method -> {"GlobalAdaptive", "MaxErrorIncreases" -> 100}, 
    WorkingPrecision -> 20]
];

(* Parameters for evaluation *)
sVal = 1.1 + 14.1347 I; (* Near first zero *)
xVals = {10, 50, 100, 200};
umax = 1000;

results = Table[
  Module[{lhs, partial, mx, tail, error},
    lhs = 1/Zeta[sVal];
    partial = finiteMuSum[sVal, x];
    mx = mertensM[x]/x^sVal;
    tail = tailIntegralApprox[sVal, x, umax];
    error = Abs[lhs - (partial + tail - mx)];
    {x, error}
  ], {x, xVals}
];

Print["Error in tail identity vs x:", TableForm[results]];
ListLinePlot[results, AxesLabel -> {"x", "Error"}, PlotLabel -> "Identity Convergence"]

Conclusions

The analysis of arXiv:hal-01847199 provides a rigorous path for bounding the Mertens function through the lens of the Dirichlet series for 1/ζ(s). By establishing that the integral of M(u)u^-s-1 is fundamentally constrained by the density of square-free integers, the paper reinforces the connection between the randomness of the Möbius function and the location of the zeta zeros.

The most promising avenue for further research lies in the refinement of the constant "m" and the error term for the square-free count. If the error in N(x) can be suppressed further using localized sieve estimates, the resulting pressure on the growth of M(x) may eventually force the critical exponent to be exactly 1/2. Future work should focus on the transition of these inequalities from the region σ > 1 to the critical strip, utilizing the continuity of the integral representation to bridge the gap at the σ = 1 boundary.

References

Source Paper: arXiv:hal-01847199
G. H. Hardy & M. Riesz, The General Theory of Dirichlet’s Series, Cambridge Tracts no. 18.
E. C. Titchmarsh, The Theory of the Riemann Zeta-Function, Oxford University Press.
A. M. Odlyzko and H. J. J. te Riele, Disproof of the Mertens Conjecture, Journal für die reine und angewandte Mathematik.