Skip to content

Erdős Problem #257

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: irrationality.

Read this status correctly

The status above is reproduced from GCL's immutable external-catalogue snapshot. It is not a new claim of current mathematical resolution, a source-semantic review, an independent proof check, or GCL certification. A published proof and a GCL replay are separate events.

Documentary explanation. Editorial acceptance is determined by the protected V2 admissions ledger and the exact manuscript source hash. This article does not confer mathematical certification.

The question in plain language

Erdős Problem #257 asks whether every infinite set of positive integers \(A\) gives an irrational number

\[ S(A)=\sum_{a\in A}\frac{1}{2^a-1}. \]

The word “every” is the whole difficulty: \(A\) may be dense, sparse, made of primes, or have arbitrarily complicated common divisors. It is not enough to settle one natural set or one family with extra structure. The original source packet writes \(A\subseteq\mathbb N\), but \(a=0\) makes the displayed denominator zero. This article therefore uses \(\mathbb N_{>0}\). The pinned Lean file quantifies over Set ℕ and relies on Lean’s division-by-zero convention, where its \(a=0\) summand is \(0\); that is a formalization convention, not the ordinary mathematical domain.

For positive \(a\), expand the geometric series and collect the coefficient of \(2^{-m}\). If

\[ f_A(m)=\bigl|\{a\in A:a\mid m\}\bigr|, \]

then nonnegative-term reindexing gives

\[ S(A)=\sum_{m\geq 1}\frac{f_A(m)}{2^m}. \]

For \(A=\mathbb N_{>0}\), \(f_A(m)=\tau(m)\), the divisor-counting function. For the prime set, \(f_A(m)=\omega(m)\), the number of distinct prime divisors. These identities explain why a question about a subseries becomes a question about arithmetic functions, but they do not themselves prove irrationality.

A small example

Take the finite set \(A=\{1,2,3\}\). Its direct sum is exactly

\[ S(A)=1+\frac13+\frac17=\frac{31}{21}. \]

For \(M=12\), let \(R_M=\sum_{m=1}^{M}f_A(m)2^{-m}\). Exact rational arithmetic gives

\[ R_{12}=\frac{6045}{4096},\qquad S(A)-R_{12}=\frac{31}{86016}. \]

Every term arising from \(a\in A\) has the geometric expansion \(\sum_{k\geq1}2^{-ak}\). The terms with \(ak>M\) are bounded by the full geometric tail \(\sum_{j>M}2^{-j}=2^{-M}\) for each \(a\), so

\[ 0\leq S(A)-R_M\leq |A|2^{-M}=\frac{3}{4096}. \]

This independently checks both the divisor-count reindexing and a useful finite tail estimate. A short reproducible check is:

from fractions import Fraction
A = {1, 2, 3}
S = sum((Fraction(1, 2**a - 1) for a in A), Fraction())
M = 12
R = sum((Fraction(sum(m % a == 0 for a in A), 2**m)
         for m in range(1, M + 1)), Fraction())
print(S, R, S - R, Fraction(len(A), 2**M))

It prints 31/21 6045/4096 31/86016 3/4096. This is a finite arithmetic check, not evidence that the infinite sum is irrational and not a proof of the problem.

What is known

The full-support case is classical. In “On arithmetical properties of Lambert series” (1948), p. 63, Paul Erdős states the theorem that for every integer \(|t|>1\), the Lambert series \(f(1/t)=\sum_{n\geq1}1/(t^n-1)\) is irrational. Taking \(t=2\) proves the case \(A=\mathbb N_{>0}\). The same paper’s identity \(\sum_{n\geq1}d(n)t^{-n}=\sum_{n\geq1}(t^n-1)^{-1}\) is the source of the divisor-count formulation. This is a published classical solution of one special case, not the universal assertion.

Erdős’s “On the irrationality of certain series” (Math. Student 36, 1968, 222–226) gives another substantial special case. The unnumbered theorem on printed p. 222 assumes an increasing sequence \(n_1<n_2<\cdots\) with \(\gcd(n_i,n_j)=1\) for \(i\ne j\) and \(\sum_i1/n_i<\infty\); it concludes that, for every integer \(t\geq2\),

\[ \sum_{i\geq1}\frac{1}{t^{n_i}-1} \]

is irrational. Thus it applies at \(t=2\) to infinite pairwise-coprime \(A\) whose reciprocal sum converges. The proof occupies pp. 223–225. The paper says that the coprimality hypothesis is superfluous by more complicated arguments, but gives no such argument there. That sentence is a research remark, not a proof of #257.

The prime-support case is now a published-source result. Tao and Teräväinen, “Quantitative correlations and some problems on prime factors of consecutive integers,” arXiv:2512.01739v2 (2026), Theorem 1.3 on printed p. 4, proves unconditionally that

\[ \sum_{n\geq1}\frac{\omega(n)}{2^n}=\sum_p\frac{1}{2^p-1} \]

is irrational. The paper says that a similar modification also handles the prime-power set, but explicitly leaves those details to the reader. The maintained problem page calls both cases settled; the prime case matches the theorem, while the prime-power wording is currently source-limited and should not be promoted here to an independently checked proof.

Kovač and Tao’s “On several irrationality problems for Ahmes series” (arXiv:2406.17593v4, Theorem 2.5 and Corollary 2.6, printed p. 7) concerns sums with variable integer shifts in the denominators. Its rationality constructions and its negative result for a shifted “Type 3” property do not alter the fixed denominator \(2^a-1\) in Problem #257. Treating that neighboring result as a counterexample to #257 would change the problem.

GCL's current state

The protected catalogue entry at teorth/erdosproblems commit 5364087 records informal status open, last updated 2025-08-31, with formal solution unformalized and statement formalized yes. The GCL registry records the same problem as open and PROTECTED_READY_BACKLOG. The live maintained problem page, checked on 12 October 2026 local time (13 October UTC), still displays OPEN and says it was last edited 15 April 2026. Its later literature and commentary therefore cannot silently replace the pinned snapshot.

The authenticated RESULT/1 comment on GCL issue #1527 has disposition PARTIAL. It verifies the full-support, pairwise-coprime, and prime interfaces above, proposes the finite reindexing and tail lemma illustrated here, and identifies the unrestricted divisor-count irrationality criterion as the missing substantive step. That return is source-audit evidence, not a theorem acceptance.

The pinned Formal Conjectures file marks the main theorem research open. Its Lambert-series identity is filled in, but the universal irrationality theorem and the full-support irrationality theorem contain sorry. I did not compile Lean. Any external Lean pointer or provider file is therefore reported formalization material only, never GCL certification.

What remains

The original universal statement still asks for irrationality for every infinite \(A\subseteq\mathbb N_{>0}\). Existing theorems require either \(A=\mathbb N_{>0}\), pairwise coprimality plus reciprocal convergence, or the special prime structure. A future proof must control the unrestricted functions \(f_A(m)\) without silently importing one of those hypotheses. The finite identity and tail bound are useful infrastructure, but they do not supply an irrationality criterion.

There are also documentary obligations. The \(\mathbb N\) versus \(\mathbb N_{>0}\) convention should be normalized in any formal proof; the live page’s prime-power claim needs a semantic check against a complete proof; and a formal compilation at the pinned provider revision would still be a formal replay, not by itself GCL mathematical certification. The safe current classification is therefore: open original problem; classical solved subcases; an unconditional published prime subcase; a source-reported but detail-omitted prime-power extension; and no accepted universal proof.

Sources and claim boundaries

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1527 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
  • GCL triage: PROTECTED_READY_BACKLOG.
  • Formal-statement coverage: A provider statement was included in the pinned snapshot. Presence does not mean the statement or proof has been checked.
  • External mathematical status: consult current problem page and its dated bibliography.

Back to catalogue · Compendium introduction