Skip to content

Erdős Problem #386

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

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

Erdős Problem 386 asks whether a nontrivial binomial coefficient can repeatedly be made from one uninterrupted run of primes. The live statement takes integers satisfying \(2\leq k\leq n-2\) and asks whether \(\binom{n}{k}\) can be the product of consecutive primes infinitely often. In concrete terms, this means that there should be a fixed \(k\) and infinitely many \(n\) for which

\[ \binom{n}{k}=p_a p_{a+1}\cdots p_{b-1} \]

for some consecutive primes \(p_a,p_{a+1},\ldots,p_{b-1}\). The phrase “infinitely often” is the central difficulty: finding many examples, or checking every case below a large cutoff, does not settle it.

The reading with a fixed parameter \(k\) matters. The pinned formal file encodes an existential choice of \(k\) followed by “frequently” many \(n\) tending to infinity. It also records separate variants asking for every fixed \(k\) and for \(k=2\). Those are useful comparisons, but they are not automatically equivalent to the main question. By symmetry, \(\binom{n}{k}=\binom{n}{n-k}\), so one can often inspect \(k\leq n/2\) when doing a finite search.

There is also a difference between a product of consecutive primes and a number whose prime divisors merely form a consecutive set: the former requires each prime in the interval exactly once. A correct audit must therefore check both squarefreeness and the absence of gaps. This is why an equality of large integers alone is not enough to certify the intended property.

A small example

The page’s example is exact:

\[ \binom{21}{2}=\frac{21\cdot20}{2}=210=2\cdot3\cdot5\cdot7. \]

The four factors are consecutive primes. The following standard-library calculation was run as an exact integer check:

from math import comb, prod

assert comb(21, 2) == 210
assert prod([2, 3, 5, 7]) == 210
assert comb(21, 2) == prod([2, 3, 5, 7])

Other displayed examples also verify directly: \(\binom{7}{3}=35=5\cdot7\), \(\binom{10}{4}=210=2\cdot3\cdot5\cdot7\), \(\binom{14}{4}=1001=7\cdot11\cdot13\), and \(\binom{15}{6}=5005=5\cdot7\cdot11\cdot13\). These calculations establish examples, not an infinite family. For \(k=2\), the elementary identity

\[ 8\binom{n}{2}+1=(2n-1)^2 \]

is also an exact way to test a proposed value of \(n\); it does not prove that only finitely many such values exist.

What is known

The primary source is P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), printed page 74. The scanned page gives the consecutive-prime-product question, the example \(\binom{21}{2}\), and the authors’ assessment that proving finiteness for \(k=2\) seemed hopeless. It then presents as a probability or conjectural expectation the stronger belief that the phenomenon should not occur infinitely often in the interior range \(3\leq k\leq n-3\). This is historical mathematical context, not a proof.

The maintained Erdős Problems page currently labels #386 OPEN and says that a finite computation cannot resolve it. It records the four examples supplied by Desmond Weisenberg and lists \(n=4,6,15,21,715\) as the known values in the \(k=2\) case, with a link to OEIS A280992. The page also warns that its open status reflects its owner’s current belief and that unlisted literature may exist. Thus the safest status statement is that the problem remains open in the maintained source, while a literature search should not be mistaken for proof that no partial result exists.

GCL's current state

Original issue #1656 is the direct-editorial source-audit record for assignment ERDOS-CATALOGUE-0386-S01. Its protected packet identifies the catalogue snapshot teorth/erdosproblems commit 536408716724a2402bdfa7c1d764bf83d64a8c89, entry blob 4e60263a80af6207aca494d2e1388e44fa372889. That entry says informal status open as of 2025-08-31, formal status unformalized, and statement metadata formalized: yes dated 2026-01-11. The local Programme registry places the item in PROTECTED_READY_BACKLOG.

The pinned FormalConjectures file has commit 85f863718beeec7b58a3a1926ee92e3472bc2020 and blob 4061bb7920827d409d4206824f1529d7c765082f. It contains the main existential formulation and the forall and two variants, each with theorem proof sorry. Its index interval over Nat.Prime is a formal representation of consecutive primes, but the file’s presence is not a proof or a certification of semantic equivalence with every informal reading.

What remains

The smallest source-sensitive task is to freeze the quantifiers before doing a large computation: fixed \(k\) versus a \(k\) allowed to vary with \(n\), and “consecutive primes” versus an arbitrary finite prime interval. A bounded successor can then take an explicit cutoff \(N\), enumerate all \(n\leq N\) and \(2\leq k\leq n/2\), and return every exact equality with its prime interval. Acceptance is an independently recomputed complete list for that declared region; falsification is one missed equality, nonprime factor, or nonconsecutive interval. The task stops at \(N\) and makes no claim beyond it.

For the unresolved theorem, the missing step is not another example but a uniform argument controlling all sufficiently large \(n\) and the permitted \(k\). Any proposed reduction must state its hypotheses and prove that it preserves the product-of-consecutive-primes condition. A finite certificate can support the documentary record and test formal definitions, but it cannot decide the infinite question.

Sources and claim boundaries

These sources support a status-neutral account of the question, examples, and formalization boundary. V2 ledger state alone decides accepted coverage. This article supplies no mathematical certification, no Lean replay, and no claim that the open problem has been solved.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1656 — 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