Erdős Problem #414
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 \(\tau(n)\) be the number of positive divisors of \(n\), including \(1\) and \(n\). Define an arithmetic function by \(h(n)=n+\tau(n)\). Erdős Problem 414 asks whether all two positive starting integers eventually reach the same number after repeatedly applying \(h\), allowing a different number of applications for each start. If \(h_0(n)=n\) and \(h_{k+1}(n)=h(h_k(n))\), the problem is
This is an intersection problem about the eventual paths of a deterministic discrete dynamical system. It does not require \(i=j\) or that a path approaches a fixed numerical limit. In fact, because \(\tau(n)\geq1\), the paths increase without bound. For two paths to merge, they must hit exactly the same integer; being close or having similar growth does not suffice.
The proposal is attributed to Christopher Spiro in the original discussion. It resembles the preceding van Wijngaarden divisor-sum question, Erdős #412, but counts divisors instead of adding them. These are distinct operations, and evidence for one conjecture is not evidence for the other.
A small example
Starting at \(2\), exact integer arithmetic gives
The first step is \(2+\tau(2)=2+2=4\). At \(4\), the divisors are \(1,2,4\), so \(4+\tau(4)=7\). At \(7\), \(\tau(7)=2\), giving \(9\). At \(9\), the divisors are \(1,3,9\), so \(9+\tau(9)=12\). At \(12\), there are six divisors, yielding \(18\).
Starting at \(3\), we find \(3,5,7,9,12,18,24,\ldots\). The two paths therefore meet at \(7\), reached in two steps from each seed. The path starting at \(5\) reaches \(7\) in one step. Another pair also meets: starting at \(4\) gives \(4,7,9,12,18\), whereas starting at \(6\) gives \(6,10,14,18\). Thus the paths from \(4\) and \(6\) meet at \(18\), after four steps and three steps respectively. These are exact finite witnesses of intersection, not proofs about every possible pair.
All values were calculated by enumerating divisor pairs up to the square root of the current integer, counting each distinct divisor once. A useful independent check is the factorization formula
For example, \(12=2^2\cdot3\) gives \((2+1)(1+1)=6\) divisors and hence \(h(12)=18\). This formula provides a second verification route for every step in a finite table.
What is known
The original printed source is Erdős and Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), page 82. The historical scan attributes this particular question to Spiro and expresses it as the conjecture that there is essentially only one infinite iteration sequence. The authors say an affirmative answer seemed certain to them. Their confidence should not be mistaken for a theorem: the printed passage does not present a proof.
The maintained Problem 414 entry continues to identify it as OPEN and specifically connects it to Problems #412 and #413. The page also points to OEIS A064491, a sequence related to repeated applications to the smallest starting value. Such an observed sequence describes one orbit; it is not a universal intersection theorem.
One rigorously elementary observation is that \(h(n)\geq n+1\). This yields strict increase and also proves that no finite cycle or fixed point can occur among positive integers. The stronger common-orbit assertion does not follow: two strictly increasing integer sequences need not intersect at all. In particular, knowing that \(h\) is increasing as a map would not by itself settle this question, and \(\tau\) is not a monotone function of \(n\) anyway.
The number of divisors has an irregular dependence on prime factorization. For a prime \(p\), \(h(p)=p+2\), whereas for \(p^2\), \(h(p^2)=p^2+3\). These exact observations help explain why a simple constant-step or monotonic-increment argument will not provide a general meeting theorem.
GCL's current state
Original source-audit issue #1684 pins the frozen teorth catalogue at commit 536408716724a2402bdfa7c1d764bf83d64a8c89, file blob 4e60263a80af6207aca494d2e1388e44fa372889. It records an open informal status as of August 2025, with no formalized solution. The maintained mathematical page remains open as of this source inquiry on 13 October 2026.
The source issue also pins FormalConjectures/414.lean at commit 85f863718beeec7b58a3a1926ee92e3472bc2020, blob 7e4a2a159da20a72b335fb8bcc2210be24e07c32. It defines \(h(n)\) as \(n\) plus the cardinality of the natural-number divisor set. The theorem carries an answer placeholder, a specialized eventual-quantifier binder, and the proof body is literally a placeholder. That is a formal statement interface, not mathematical certification. In particular, concordance between the formal binder and the historical universal claim has not been established by compiling a theorem.
The authenticated source-return link remains authenticated original RESULT/1. This manuscript is descriptive evidence only; editorial acceptance is governed separately by the V2 register.
What remains
An appropriately bounded next task is to enumerate the \(h\) orbit from every seed \(1\) through \(100\) for twelve iterations, using exact integer arithmetic. For every pair with a collision in the recorded window, record the first common value and its two indices. For every other pair, record a bounded nonobservation instead of labeling the pair disjoint. Independently replay the sample \(4\) versus \(6\) and check divisor counts by prime factorization rather than the same enumeration algorithm.
A second, separate task is to compare the exact mathematical quantifiers in the original p. 82 prose with the pinned formal source. In particular, identify the meaning of the special binder and its treatment of zero, since the historical question concerns positive starts. The finite algorithm is accepted only if each claimed meeting has an executable witness and the missing cases are carefully qualified. A scope mismatch, false divisor count, or statement that finite testing proves the conjecture constitutes a failed interface check. These tasks stop before a general proof is asserted.
Sources and claim boundaries
Primary statement: Erdős and Graham (1980), printed page 82. Current research status: Erdős Problem 414, accessed 2026-10-13. Computational context: OEIS A064491. Protected provenance: GCL issue #1684, frozen catalogue, and pinned formal statement.
The finite collisions and divisor identities are verified examples, not an answer for all positive starts. The open status is a current-maintained-source description, subject to future discoveries. No complete mathematical proof, Lean replay, or protected editorial admission is claimed.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #1684 — 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.