Erdős Problem #250
Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: proved (not automatically the live result).
Subject areas: number theory, 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
For a positive integer \(n\), let
be the sum of its positive divisors. Erdős Problem 250 asks whether the real number
is irrational; equivalently, whether \(S\notin\mathbb Q\). This is an ordinary convergent real series, not a question about a finite decimal approximation. Indeed, there are at most \(n\) divisors of \(n\), each at most \(n\), so \(\sigma_1(n)\leq n^2\). Comparison with \(\sum_{n\geq1}n^2/2^n\) proves convergence.
The quantifiers matter. The classical statement concerns the one real number defined by the displayed series. The pinned Lean declaration instead says, in effect,
That implication does not itself assert that a sum exists; it asserts irrationality for every real \(x\) which is a sum of the specified sequence. In standard analysis, convergence gives existence and uniqueness, so the two formulations align after that bridge is supplied. The indexing convention also needs care: the registered sum starts at \(n=1\), whereas a formal Lambert-series development may include an \(n=0\) term and rely on its coefficient being zero.
A small example
The first divisor sums are
Using the first eight terms gives the exact rational partial sum
The corresponding finite version of Nesterenko's normalization is
and the next omitted summand is \(\sigma_1(9)/2^9=13/512\). These figures were independently recomputed with exact rational arithmetic. A minimal reproduction is:
from fractions import Fraction
def sigma(n):
return sum(d for d in range(1, n + 1) if n % d == 0)
s8 = sum((Fraction(sigma(n), 2**n) for n in range(1, 9)), Fraction(0))
assert [sigma(n) for n in range(1, 10)] == [1, 3, 4, 7, 6, 12, 8, 15, 13]
assert s8 == Fraction(687, 256)
assert 1 - 24*s8 == Fraction(-2029, 32)
This checks normalization and arithmetic only. No finite list of partial sums can prove irrationality of the limiting value, and the displayed negative rational comes with no certified error estimate for \(1-24S\).
What is known
The answer is affirmative, and the strongest source-backed statement here is stronger than irrationality: \(S\) is transcendental. The key published result is Yuri V. Nesterenko, “Modular functions and transcendence questions,” Sbornik: Mathematics 187(9) (1996), 1319–1348. On English p. 1319 he defines Ramanujan's function
Theorem 1 on English p. 1320 states that for every complex \(q\) with \(0<|q|<1\), at least three of \(q,P(q),Q(q),R(q)\) are algebraically independent over \(\mathbb Q\). Corollary 2 on the same page states that if \(q\) is algebraic and \(0<|q|<1\), then \(P(q),Q(q),R(q)\) are algebraically independent over \(\mathbb Q\), hence each is transcendental. Taking \(q=1/2\) gives
Because \(1/2\) is algebraic and lies strictly between \(0\) and \(1\), Corollary 2 makes \(P(1/2)\) transcendental. The nonzero rational affine change
therefore makes \(S\) transcendental, and consequently irrational. This is a borrowed deep lemma from Nesterenko's published paper; this article does not claim to reprove it.
The original source packet records the problem in P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), p. 61, and in P. Erdős, “On the irrationality of certain series: problems and results,” in New Advances in Transcendence Theory (1988), p. 102. The maintained page cites these as [ErGr80] and [Er88c], and attributes the affirmative solution to Nesterenko. Its bibliography points to Nesterenko's 1996 C. R. Acad. Sci. announcement, pp. 909–914; the Sbornik article above is the detailed theorem source used for the reduction. These are classical sources and a classical solution, not a new GCL result.
GCL's current state
The authenticated source audit is issue #1519, RESULT/1, assignment ERDOS-CATALOGUE-0250-S01. Its disposition is STATUS_CONFLICT: it verifies the Nesterenko reduction but identifies a mismatch between several status surfaces. The pinned catalogue snapshot at teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89 labels the informal problem proved, reports formal status Lean, and says the statement is formalized. The current catalogue blob reported in the audit is 97ae243a7fc056bb9c6413ce665b575eadd27ff3; the entry was textually unchanged in that comparison. The maintained problem page says “The answer is yes, as shown by Nesterenko,” was last edited 28 September 2025, and identifies OEIS A066766. Direct HTML retrieval was 403 during this review, so that live-page wording is reported from the authenticated packet and indexed page record, not presented as a fresh GCL page certification.
The pinned Formal Conjectures declaration has blob 44c53cdb412bdb16a09a88346546303e4f6e3543 and contains the right-looking statement, but its proof is by sorry. The current file has blob f45e0be77e7f7c20949ee6138ce56f9f6b806213 and adds a formal_proof using lean4 pointer to plby/lean-proofs commit 8822f7d; the current declaration still contains sorry. A pointer to an external Lean proof is a reported formalization lead, not GCL certification. No Lean build, axiom report, or semantic-equivalence check is claimed here.
What remains
The mathematical question itself is classically settled by Nesterenko. What remains is source- and formalization-boundary work, not a reason to call the original problem open. A responsible formal acceptance would compile the exact external proof commit with its declared Lean and mathlib versions, inspect all dependencies, and report whether any sorry or unreported axiom remains. It would also prove the semantic bridge from
to the value \((1-P(1/2))/24\), including the \(n=0\) convention, the meaning of σ 1, the real-versus-complex coercions, and the denominator/exponent convention. Uniqueness of the sum should be explicit rather than silently inferred from a theorem whose hypotheses have not been inspected.
The authenticated return proposes exactly this bounded successor check and expressly stops short of reproving Nesterenko. Until that check is replayed, “formalized in Lean” and “proved (Lean)” remain upstream reported statuses. They must not be rewritten as GCL-certified formal proof. Likewise, the finite computation above is a reproducibility check, not partial evidence for transcendence.
Sources and claim boundaries
- Original problem sources: P. Erdős and R. L. Graham (1980), Old and New Problems and Results in Combinatorial Number Theory, p. 61, source scan; P. Erdős (1988), “On the irrationality of certain series: problems and results,” pp. 102–109, source scan, especially p. 102. The authenticated return records these page references; this manuscript does not claim a new transcription of either scan.
- Primary resolving paper: Yuri V. Nesterenko (1996), “Modular functions and transcendence questions,” Sbornik: Mathematics 187(9), 1319–1348, DOI 10.1070/SM1996v187n09ABEH000158, full text. Definition of \(P\) is p. 1319; Theorem 1 and Corollary 2 are p. 1320.
- Maintained problem record: Erdős Problems #250, including its reported last-edit date,
PROVEDlabel, source keys, and formalization note. Live status can drift; it is not mathematical certification by GCL. - Formal-provider evidence: pinned
250.leanand current250.lean. The pinned/current distinction and external pointer are evidence about provenance, not a verified proof result. - GCL evidence: authenticated RESULT/1. It supports the Nesterenko reduction, the exact finite arithmetic, and the status-conflict warning; it does not grant certification, publication authority, or a claim of original proof.
The article therefore distinguishes four things: an open original problem (not the status here), a classical published solution (yes, via Nesterenko), a reported but unreplayed formal proof (the external Lean pointer), and a source-limited GCL editorial draft. Only the second is used to explain why the mathematical answer is yes; the latter two remain explicitly unaccepted and uncertified.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1519 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: OUTSIDE_STRICT_OPEN_TRIAGE.
- 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.