Skip to content

Erdős Problem #410

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: open (not automatically the live result).
Subject areas: number theory, iterated functions.

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.

The question in plain language

Let \(\sigma(n)\) be the sum of the positive divisors of \(n\). Start with an integer \(n\) and apply this operation repeatedly: \(\sigma_1(n)=\sigma(n)\) and \(\sigma_{k+1}(n)=\sigma(\sigma_k(n))\). Erdős Problem 410 asks whether every starting value \(n\ge 2\) produces an orbit whose typical multiplicative growth per step becomes arbitrarily large:

\[ \text{for every }n\ge 2,\qquad \lim_{k\to\infty}\sigma_k(n)^{1/k}=\infty. \]

The quantifier matters. This is not a claim about most \(n\), about a subsequence of \(k\), or merely about the sequence \(\sigma_k(n)\) being unbounded. It asks for one asymptotic statement for each fixed starting integer. The root with exponent \(k\) measures the average multiplicative scale of the first \(k\) iterates; saying that it tends to infinity is much stronger than seeing occasional large values.

The maintained Erdős Problems entry gives this formulation and tags it as number theory and iterated functions. The original issue for this audit is MATH-PROGRAMME #1680.

A small example

Here is an exact finite check for \(n=2\). Direct divisor enumeration gives the following orbit prefix:

\[ 2\to3\to4\to7\to8\to15\to24\to60\to168\to480\to1512. \]

For example, the divisors of \(480\) are \(1,2,3,4,5,6,8,10,12,15,16,20,24,30,32,40,48,60,80,96,120,160,240,480\), whose sum is exactly \(1512\). The preceding checks similarly give \(\sigma(168)=480\) and \(\sigma(60)=168\); enumerating all divisors verifies every displayed arrow. This is a reproducible exact-arithmetic calculation, not a floating-point estimate.

It shows that the orbit can grow substantially over a short window, and it supplies useful test data for implementations. It does not establish the limit. A finite prefix cannot decide an asymptotic statement, and even rapid early growth does not rule out later cycles, slowdowns, or other behavior. In particular, the example is not a proof for \(n=2\), still less for all \(n\).

What is known

The pinned catalogue packet records, at teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89 and the entry blob 4e60263a80af6207aca494d2e1388e44fa372889, informal status open, last updated 2025-08-31. It records formal status unformalized, while separately recording that the statement has been formalized. Those are catalogue metadata, not a proof or a certification decision.

The live page was accessed 2026-10-13 and says it was edited 18 January 2026. It retains the same \(n\ge2\) statement, reports zero proof expositions and zero proof claims, and displays two ordinary comments: one asks whether a particular value such as \(n=2\) is known, and an earlier comment prompted the explicit \(n>1\) condition. Nothing on that page is an externally verified solution of the problem.

The page's bibliography points to Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), while the protected formal source identifies the question as problem (iii) in Erdős, Granville, Pomerance, and Spiro, “On the Normal Behavior of the Iterates of Some Arithmetic Functions,” Analytic Number Theory (1990), p. 169. The latter chapter is the primary reference named by the formal source, with DOI 10.1007/978-1-4612-3464-7_13, pp. 165–204. Its publisher metadata and abstract are accessible; the full chapter text was not accessible during this bounded audit, so no theorem beyond the pinned page reference is attributed to it.

For historical context, the primary paper P. Erdős, “Some remarks on the iterates of the \(\phi\) and \(\sigma\) functions,” Colloquium Mathematicum 17 (1967), 195–202, DOI 10.4064/cm-17-2-195-202, is directly relevant. Its bibliographic record confirms the paper and pages; it is used here as context for the iterated-function subject, not as evidence that the exact Problem 410 limit has been proved.

GCL's current state

The protected formal provider is google-deepmind/formal-conjectures at commit 85f863718beeec7b58a3a1926ee92e3472bc2020. Its file FormalConjectures/ErdosProblems/410.lean has blob c0167d602bcab55a5c91e68b51c54ab2833ca210. It defines the iterates using function iteration and states, for every n > 1, convergence of the real sequence ((sigma 1)^[k] n : ℝ) ^ (1 / (k : ℝ)) to atTop.

The Lean declaration is a formal statement interface, not a completed derivation: its proof is by sorry, and the source itself warns that formalization does not establish semantic equivalence with the primary wording. The source and catalogue use the repaired lower bound \(n>1\), matching the live page. No Lean replay, mathematical certification, accepted coverage, or publication decision is supplied here. V2 ledger state, rather than this article or the catalogue metadata alone, decides accepted coverage.

What remains

The smallest useful successor is a source-and-semantics review of the primary page-169 formulation followed by a bounded orbit investigation. First, compare its quantifiers, definition of the iterate index, and meaning of the real root with exponent \(k\) with the live and Lean statements. Then choose a finite bound \(N\) and depth \(K\), publish an exact-arithmetic certificate for all \(2\le n\le N\) and \(0\le k\le K\), and separately test whether the verified values support a rigorously stated lemma such as positivity or a monotonicity property on that finite domain.

Acceptance should require the source quotation or scan location, a deterministic certificate, and an independently reproducible checker. Falsification is equally concrete: a source mismatch, a divisor-sum error, a counterexample to the proposed finite lemma, or an orbit entering a verified cycle stops that package. The stopping rule is reached at the chosen \(N,K\), at the first falsifying witness, or when the primary text needed to normalize the target cannot be obtained. None of these finite outcomes settles the asymptotic conjecture.

Sources and claim boundaries

The source lock for this article is the original issue #1680, the frozen catalogue commit/blob above, the live Problem 410 page, and the pinned Lean file/commit above. The primary references are the 1990 chapter DOI and the 1967 paper DOI. The original SOURCE RESULT/1 citation is authenticated original RESULT/1.

Established here: the normalized statement, the dated catalogue/live-page observations, the formal file's exact text and sorry status, and the displayed finite divisor calculation. Not established here: a proof, a counterexample, a complete reading of the paywalled 1990 chapter, semantic equivalence certified by GCL, or current accepted coverage. Informal website comments and metadata are reports and provenance aids, not mathematical certification. This article preserves those boundaries so that a later source review can strengthen or narrow the claims without silently changing the problem.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1680 — 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