Skip to content

Erdős Problem #261

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

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 text does not confer mathematical certification.

The question in plain language

For a positive integer \(n\), ask whether the small number \(n/2^n\) can be rebuilt as a finite sum of terms of the same general shape, using at least two different positive integer indices. In symbols, define

\[ P(n) :\Longleftrightarrow \exists t\geq 2,\ \exists a_1,\ldots,a_t\geq 1, \quad a_i\ne a_j\ \text{when }i\ne j, \quad \frac{n}{2^n}=\sum_{i=1}^{t}\frac{a_i}{2^{a_i}}. \]

Erdős asks whether \(P(n)\) is true for infinitely many \(n\), and whether it is true for every positive integer \(n\). The indices may be sorted, so “different” means a finite set of distinct positive integers rather than an order-sensitive tuple.

There is a second, genuinely different question. Can one rational number \(x\) have at least \(2^{\aleph_0}\) different infinite representations

\[ x=\sum_{k=1}^{\infty}\frac{a_k}{2^{a_k}}, \]

where \(a_1<a_2<\cdots\) are positive integers? The increasing-order convention prevents mere reorderings from being counted as new representations. Erdős’s 1988 source also records a weaker version asking only for two representations; that weaker question must not be confused with the continuum-cardinality question.

A small example

The first useful identity is already visible at \(n=4\):

\[ \frac{4}{2^4}=\frac{5}{2^5}+\frac{6}{2^6}=\frac{10}{64}+\frac{6}{64}=\frac14. \]

Thus \(P(4)\) holds with indices \(5\) and \(6\). The published Borwein–Loring construction generalizes this. If \(m\geq2\) and

\[ n=2^{m+1}-m-2, \]

then the indices \(n+1,n+2,\ldots,n+m\) give a witness. The calculation is short enough to reproduce. Since

\[ \sum_{i=1}^{m}\frac{n+i}{2^{n+i}} =\frac{1}{2^n}\left(n\sum_{i=1}^{m}\frac1{2^i}+\sum_{i=1}^{m}\frac{i}{2^i}\right), \]

and the finite geometric identities are

\[ \sum_{i=1}^{m}\frac1{2^i}=1-\frac1{2^m}, \qquad \sum_{i=1}^{m}\frac{i}{2^i}=2-\frac{m+2}{2^m}, \]

the bracket equals

\[ n+2-\frac{n+m+2}{2^m}=n+2-2=n, \]

because \(n+m+2=2^{m+1}\). This proves the displayed identity, and \(m\geq2\) supplies the required two or more distinct terms. Exact rational replay gives \(n=4,11,26,57,120,247,502\) for \(m=2,3,4,5,6,7,8\).

The weaker infinite question also has a concrete example. The standard sum \(\sum_{k\geq1}k/2^k=2\), together with the identity above, shows that \(7/4\) is represented by the positive integers with \(4\) removed and also by the positive integers with \(5\) and \(6\) removed. Both sets are enumerated increasingly and both omissions remove \(1/4\). This proves “at least two,” not “at least \(2^{\aleph_0}\).”

What is known

Erdős’s paper “On the irrationality of certain series: problems and results,” New Advances in Transcendence Theory (1988), pp. 102–109, printed p. 104, gives the problem and notes that Cusick had a simple proof of infinitely many finite solutions, without recording that proof. Borwein and Loring later supplied an explicit proof interface: their Proposition 1, pp. 382–383, gives the family above and conjectures that every \(n\) works. Their argument is a direct finite identity for this family; it does not prove the universal statement.

Tengely, Ulas, and Zygadło study the corresponding Diophantine equation with sorted indices. Their Theorem 2.1, p. 448, gives structural bounds for a fixed number of summands; Remark 2.2, pp. 448–449, records the Borwein–Loring family; and Theorem 2.5, p. 450, lists small-term solutions including \((n,(a_i))=(1,(3,6,8))\) and \((4,(5,6))\). Their Theorem 3.5, p. 455, reports a greedy computation finding a solution for every \(2\leq n\leq10^4\). Together with the displayed \(n=1\) witness, this supports the finite range \(1\leq n\leq10^4\), but a finite computation cannot establish \(P(n)\) for every positive integer.

The same paper’s Corollary 3.6, pp. 456–457, proves that infinitely many rationals have at least nine representations. That is meaningful progress on multiplicity, but it is not the requested rational with \(2^{\aleph_0}\) representations. Borwein and Loring’s uncountable-representation results concern a dense family of irrational numbers, so they do not settle the rational requirement either.

GCL's current state

CONSPICUOUS EXCEPTION — AUTHENTICATED STATUS: PARTIAL, WITH A SOURCE/STATUS CONFLICT. The source audit does verify a classical infinite family and the published bounded computation, while the full all positive \(n\) and continuum-rational questions remain open. This article preserves those distinct quantifiers and does not promote a solved sub-variant into a solution of Problem #261.

The controlling evidence is the authenticated RESULT/1 comment on original issue #1531. The issue remains an open documentary work item. The pinned catalogue snapshot at teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, entry 261 and blob 4e60263a80af6207aca494d2e1388e44fa372889, says open and unformalized, with a formalized statement. The current catalogue entry at commit 503d0a58ff3b6a12d7ff1665aa37fb953416d85f remains open, blob 97ae243a7fc056bb9c6413ce665b575eadd27ff3; no mathematical status change was found.

The current formal declaration at commit 0d5c14ee95b645274722b6c622c8f8f825aedca3 (Git blob 3e7ff49c9a6d781248c46921d78ec675444e58fa) captures the finite predicate, the infinite-representation predicate, the Borwein–Loring family, the \(n\leq10000\) statement, and the two open questions. Its theorem bodies are sorry. Thus it is useful statement infrastructure, not a GCL proof or MathCert certificate. The finite encoding is equivalent to the sorted formulation up to normalization; its infinite predicate explicitly requires strict increase, positivity, summability, and equality of the rational series.

What remains

The principal missing finite result is a proof of \(P(n)\) for every positive \(n\). The Borwein–Loring identity proves infinitely many carefully selected values, and the Tengely–Ulas–Zygadło search checks a large finite interval, but neither supplies the universal quantifier. Their structural theorems constrain fixed-term solutions; they do not show that some number of terms exists for every \(n\).

The principal missing infinite result is a rational \(x\) with continuum many increasing representations. The explicit \(7/4\) construction settles only the weaker two-representation variant. It is therefore incorrect to write that Problem #261 is solved merely because that example is formalizable.

A bounded next task could formalize the finite Borwein–Loring lemma and replay exact rational instances, while separately recording the hypotheses and stopping before the all positive \(n\) or continuum claims. The unresolved historical source claim is Cusick’s unrecorded “simple proof.” It should remain an attribution requiring source recovery, not be silently reconstructed or treated as a certified proof.

Sources and claim boundaries

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1531 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
  • GCL triage: FORMAL_SOURCE_REFRESH_REQUIRED.
  • Formal-statement coverage: No statement was confirmed in GCL's pinned formal-provider 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