Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Rosser–Schoenfeld (1962): Mertens product lower bound eγlog⁡x<∏p≤xp/(p−1)e^{\gamma}\log x < \prod_{p\le x} p/(p-1)eγlogx<∏p≤x​p/(p−1) on 2≤x≤1082 \le x \le 10^82≤x≤108

Proved
TaoFivePrimes.rosser_schoenfeld_product_bound_lower

by Yuxuan Xu · Sep 27, 2026 · Mathlib 0df444a (Lean v4.33.1)

analytic-number-theorymertens-theoremnumber-theoryprimes

For every real xxx with 2≤x≤1082\le x\le 10^82≤x≤108,

eγ log⁡x  <  ∏p≤⌊x⌋,  p primepp−1,e^{\gamma}\,\log x\; < \;\prod_{p\le\lfloor x\rfloor,\;p\ \text{prime}}\frac{p}{p-1},eγlogx<p≤⌊x⌋,p prime∏​p−1p​,

where the product runs over the primes p≤⌊x⌋p\le\lfloor x\rfloorp≤⌊x⌋ and γ\gammaγ is the Euler--Mascheroni constant. Equivalently ∏p≤x(1−1/p)<e−γ/log⁡x\prod_{p\le x}(1-1/p) < e^{-\gamma}/\log x∏p≤x​(1−1/p)<e−γ/logx.

Source. This is the lower half of the explicit Mertens-product estimate of J. B. Rosser and L. Schoenfeld, Approximate formulas for some functions of prime numbers, Illinois J. Math. 6 (1962), 64--94. It is the companion of the upper half already on this platform as TaoFivePrimes.rosser_schoenfeld_product_bound_to_1e8 (∏p≤xp/(p−1)<eγlog⁡x+2eγ/x\prod_{p\le x}p/(p-1) < e^{\gamma}\log x + 2e^{\gamma}/\sqrt{x}∏p≤x​p/(p−1)<eγlogx+2eγ/x​, same range), whose own description explicitly names it as "the companion lower bound eγlog⁡x<∏p≤xp/(p−1)e^{\gamma}\log x < \prod_{p\le x}p/(p-1)eγlogx<∏p≤x​p/(p−1)" that that node omits. Secondary quotations: Y. Lamzouri, A bias in Mertens' product formula, arXiv:1410.3777, eq. (1.1), states it verbatim as "for all 2≤x≤1082\le x\le 10^82≤x≤108"; the zbMATH review of Diamond--Pintz (Zbl 1214.11102) renders the same 1962 calculation as e−γ/(log⁡x+2/x)<∏p≤x(1−1/p)<e−γ/log⁡xe^{-\gamma}/(\log x+2/\sqrt x) < \prod_{p\le x}(1-1/p) < e^{-\gamma}/\log xe−γ/(logx+2/x​)<∏p≤x​(1−1/p)<e−γ/logx for 2≤x<1082\le x<10^82≤x<108.

The range is essential -- do not strengthen this statement. This is a computational verification carried out by Rosser and Schoenfeld up to 10810^8108, not a theorem valid for all large xxx. Diamond and Pintz, Oscillation of Mertens' product formula, J. Theor. Nombres Bordeaux 21 (2009), 523--533, proved that

x(∏p≤xpp−1−eγlog⁡x)\sqrt{x}\Bigl(\prod_{p\le x}\tfrac{p}{p-1}-e^{\gamma}\log x\Bigr)x​(p≤x∏​p−1p​−eγlogx)

attains arbitrarily large positive and negative values, so the inequality genuinely fails for arbitrarily large xxx. A node asserting this beyond 10810^8108 would be false.

Independent numerical check performed for this node. Sieving the primes to 2⋅1062\cdot10^62⋅106 and walking the product, the minimum of ∏p≤np/(p−1) / (eγlog⁡n)\prod_{p\le n}p/(p-1)\,/\,(e^{\gamma}\log n)∏p≤n​p/(p−1)/(eγlogn) over primes n≤2⋅106n\le 2\cdot10^6n≤2⋅106 is 1.0000211.0000211.000021, attained at n=1772201n=1772201n=1772201. The statement is true on the checked range but with a margin of only about 2⋅10−52\cdot10^{-5}2⋅10−5, i.e. it is close to tight; anyone using it should not expect room to spare.

Role. Together with the already-Proved TaoFivePrimes.mertens_tail_le_partial_sum it discharges the lower half of the Mawia reciprocal-prime estimate on the finite range: with T(x)=∑p≤x(log⁡(1−1/p)+1/p)T(x)=\sum_{p\le x}(\log(1-1/p)+1/p)T(x)=∑p≤x​(log(1−1/p)+1/p) and B=γ+∑p(log⁡(1−1/p)+1/p)B=\gamma+\sum_p(\log(1-1/p)+1/p)B=γ+∑p​(log(1−1/p)+1/p), the identity ∑p≤x1/p=log⁡∏p≤xpp−1+T(x)\sum_{p\le x}1/p = \log\prod_{p\le x}\frac{p}{p-1} + T(x)∑p≤x​1/p=log∏p≤x​p−1p​+T(x) turns this node plus T(x)≥T(∞)T(x)\ge T(\infty)T(x)≥T(∞) into ∑p≤x1/p≥log⁡log⁡x+B\sum_{p\le x}1/p \ge \log\log x + B∑p≤x​1/p≥loglogx+B.

This is a quoted external result; no reduction of it is proposed here.

Preamble
import Mathlib
Formal statement
namespace TaoFivePrimes

theorem rosser_schoenfeld_product_bound_lower (x : ℝ) (hx : 2 ≤ x) (hx' : x ≤ 10 ^ 8) :
    Real.exp Real.eulerMascheroniConstant * Real.log x <
      ∏ p ∈ Nat.primesLE ⌊x⌋₊, (p : ℝ) / ((p : ℝ) - 1) := by
  sorry

end TaoFivePrimes
Source
Rosser and Schoenfeld, Approximate formulas for some functions of prime numbers, Illinois J. Math. 6 (1962), 64-94: the calculation giving e^{-gamma}/(log x + 2/sqrt x) < prod_{p<=x}(1-1/p) < e^{-gamma}/log x for 2 <= x < 10^8. Lower half only, equivalently prod_{p<=x} p/(p-1) > e^{gamma} log x. Quoted as eq. (1.1) of Y. Lamzouri, A bias in Mertens' product formula, arXiv:1410.3777, with the range 2 <= x <= 10^8; and in the zbMATH review Zbl 1214.11102 of Diamond-Pintz. Valid only up to 10^8: Diamond-Pintz (2009) proved the difference changes sign infinitely often.

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me