Skip to content

Erdős Problem #4

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

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

Prime numbers are irregularly spaced. Let \(p_n\) be the \(n\)-th prime and write \(g_n=p_{n+1}-p_n\) for the gap to the next prime. Erdős asked whether gaps are infinitely often larger than any fixed multiple of a particular slowly growing benchmark:

\[ L(n)=\frac{\log n\,\log_2 n\,\log_4 n}{(\log_3 n)^2}, \qquad \log_1 n=\log n,\quad \log_{j+1}n=\log(\log_j n). \]

More precisely, for every \(C>0\), are there infinitely many sufficiently large indices \(n\) with \(g_n>C L(n)\)? The logarithms are natural logarithms, and the formula is used only on the tail where the iterated logarithms are defined and positive.

The two quantifiers matter. This is not the claim that all gaps grow large; many consecutive primes remain relatively close together. It asks whether exceptionally long gaps beat every fixed constant times the specified scale, over and over again.

A small example

The primes \(23\) and \(29\) are consecutive: \(24,25,26,27,28\) are all composite. Their prime gap is \(29-23=6\).

There is a simple construction of arbitrarily long runs of composite numbers. For any integer \(m\geq 2\), each number

\[ m!+2,\;m!+3,\;\ldots,\;m!+m \]

is composite, because \(m!+k\) is divisible by \(k\) and is larger than \(k\). So prime gaps have no absolute upper bound.

Why is that elementary trick not a solution? The factorial \(m!\) grows extremely quickly. A composite run of length \(m-1\) near \(m!\) does not by itself show that its enclosing prime gap beats the more demanding scale \(C L(n)\), where \(n\) is the corresponding prime's index. The challenge is to build long prime-free stretches efficiently relative to their location. That quantitative distinction is the substance of Erdős #4.

What is known

The question was answered yes in published work in 2016. Earlier, Rankin had established an inequality at this scale with some positive constant. Independently, James Maynard and Kevin Ford, Ben Green, Sergei Konyagin and Terence Tao proved that the constant can be taken arbitrarily large along an infinite sequence of prime gaps. Their papers appear consecutively in Annals of Mathematics 183 (2016), pages 915–933 and 935–974.

A 2018 joint paper by Ford, Green, Konyagin, Maynard and Tao strengthened the known lower bound. Writing \(G(X)\) for the largest gap between consecutive primes below a large threshold \(X\), the paper gives

\[ G(X)\gg \frac{\log X\,\log_2 X\,\log_4 X}{\log_3 X}. \]

The notation \(\gg\) hides a fixed positive multiplicative constant. This is stronger than the old Rankin-scale bound by a factor of \(\log_3 X\), which itself grows without bound. These published statements are formulated using prime sizes or the threshold \(X\); comparing them with the index \(n\) in the registered question requires careful normalization.

The central idea is not to predict every prime. It is to choose congruence obstructions so that a long interval is covered by residue classes forcing its entries to be composite, and then use sophisticated sieve/distribution arguments to make the construction quantitative. This is a conceptual guide, not a substitute for the published proofs.

GCL's current state

GCL's source-audit return for Problem #4 identifies the original published theorem family and the pinned Formal Conjectures statement. It proposes a modest formal bridge between a lower bound written with \(L(p_n)\) and the registered target written with \(L(n)\). On a sufficiently large tail, \(L\) is positive and increasing; since \(p_n\geq n\), a large gap exceeding \(C L(p_n)\) also exceeds \(C L(n)\).

A subsequent GCL critical review found the conditional mathematical reduction plausible, but explicitly returned REVISE: the source theorem's precise wording and index conventions still require source-semantic reconciliation, and there is no pinned, compiled Lean replay of the bridge. The review did not accept a formal proof or confer mathematical certification. The protected formal-statement file was reported to contain unfinished proof placeholders.

Therefore published resolution of the Erdős question and GCL verification of a formal proof are different facts. GCL has source/review evidence and an unaccepted formalization proposal, not an independently certified theorem.

What remains

The original infinitely-often question is settled in the mathematical literature, so it must not be advertised as an open problem for workers to solve again. Two real challenges remain distinct:

  • Proof provenance and formal replay: reconcile the 2016/2018 primary theorem statements with the exact registered index-scale question and formally verify the bounded source-to-index implication. Full machine verification of the deep sieve arguments would be a much larger undertaking.
  • Stronger quantitative prime-gap bounds: mathematicians still seek a sharper picture of how the largest gaps grow. Heuristics suggest gaps on scales approaching \((\log X)^2\), substantially beyond the established lower bounds; this is an open quantitative direction, not an unresolved part of the original yes/no question.

An editorial explanation, even when source-checked, does not establish new lower bounds or turn a conditional bridge into a proved formal theorem.

Sources and claim boundaries

Editorial status: first-pass, evidence-qualified manuscript awaiting a separate non-authoring mathematical/source check and normal protected publication. No assertion of independent verification, formal certification, source acceptance, or mathematical novelty is made.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1273 — 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.

Back to catalogue · Compendium introduction