Skip to content

Erdős Problem #367

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

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

For a positive integer \(n\), write its prime factorization as \(n=\prod_p p^{v_p(n)}\). The 2-full part is

\[ B_2(n)=\prod_{v_p(n)\geq 2}p^{v_p(n)}. \]

Thus \(B_2(n)\) keeps the complete prime-power contribution from primes that occur at least twice, and discards primes occurring only to the first power. Erdős Problem 367 asks how large a product of these parts can be over a short interval. For each fixed integer \(k\geq 1\), set

\[ P_k(n)=\prod_{n\leq m<n+k}B_2(m). \]

The main question is whether \(P_k(n)\leq n^{2+o(1)}\) as \(n\) tends to infinity. A stronger proposed bound is \(P_k(n)\ll_k n^2\), with a constant allowed to depend on \(k\). The wording “fixed \(k\)” matters: \(k\) is chosen first and is not allowed to grow with \(n\).

The page also asks about higher full parts \(B_r\), where only prime powers of exponent at least \(r\) are kept. Its literal higher r wording quantifies over every positive \(\epsilon\), but the maintained formal statement now records the nontrivial reading in which some \(\epsilon=\epsilon(r,k)>0\) may depend on \(r\) and \(k\). That distinction is mathematical, not editorial.

A small example

The Pell recurrence

\[ x_0=1,\quad y_0=0,\qquad x_{j+1}=3x_j+8y_j,\quad y_{j+1}=x_j+3y_j \]

satisfies \(x_j^2-8y_j^2=1\). At \(j=7\) it gives \(x_7=114243\) and \(y_7=40391\). Put \(n=8y_7^2=13051463048\). Exact factorization gives

\[ n=2^3\cdot13^4\cdot239^2,\qquad n+1=3^2\cdot113^2\cdot337^2, \]

and

\[ n+2=2\cdot5^2\cdot29\cdot269\cdot33461. \]

Therefore \(B_2(n)=n\), \(B_2(n+1)=n+1\), and \(B_2(n+2)=25\). The exact three-term product is

\[ P_3(n)=13051463048\cdot13051463049\cdot25=4258517192659022833800. \]

This verifies the arithmetic certificate directly: \(x_7^2-8y_7^2=1\) and \(x_7^2\equiv-1\pmod {25}\). It does not prove an infinite family, a limsup statement, or any theorem about all \(k\). It is one finite check of the Pell mechanism and of the definition of \(B_2\).

What is known

The original source cited by the maintained page is P. Erdős and R. L. Graham, Old and New Problems and Results in Combinatorial Number Theory (1980), p. 68, recorded as `ErGr80`. The accessible live page for Problem 367 was last edited on 23 March 2026 and still labels the problem open. It records the elementary bound \(P_k(n)\ll n^2\) for \(k\leq 2\) and reports Wouter van Doorn’s observation that this fails for every \(k\geq 3\).

Van Doorn’s public November 2025 comment gives the Pell construction behind the finite example. Along a subsequence, \(n_j=8y_j^2\) and \(n_j+1=x_j^2\) are both 2-full, while a congruence forces \(5^t\mid n_j+2\). Since the Pell sequence grows exponentially in the index and the selected index is exponential in \(t\), this yields the reported lower bound \(P_3(n_j)\gg n_j^2\log n_j\) infinitely often. This is strong evidence against the proposed \(O_k(n^2)\) bound for \(k\geq 3\), but the public discussion is not a GCL-certified proof record.

For related primary literature, David T. Walker’s “Consecutive Integer Pairs of Powerful Numbers and Related Diophantine Equations,” The Fibonacci Quarterly 14 (1976), pp. 111–116, develops the Pell-equation classification of consecutive powerful pairs. Its Theorems 2.1 and 2.2 explain how powers of a fundamental Pell solution produce the relevant square and powerful factors. That paper supplies context for the construction, not a solution of the short-interval product problem. A later public comment claims affirmative higher r bounds and a stronger \(k=3\) limsup result, but those claims remain unverified external reports.

GCL's current state

The source packet in original issue #1637 pins `teorth/erdosproblems` at commit `536408716724a2402bdfa7c1d764bf83d64a8c89`, entry `number: "367"`, with Git blob `4e60263a80af6207aca494d2e1388e44fa372889`. That snapshot says informal status `open`, formal solution status `unformalized`, and statement-formalized metadata `yes`, last updated 31 August 2025. Its protected triage class is `FORMAL_SOURCE_REFRESH_REQUIRED`.

The same packet says that the admitted FormalConjectures snapshot `FC-GDM-85F86371` at commit `85f863718beeec7b58a3a1926ee92e3472bc2020` has no #367 file. A later current-provider file exists at commit `817d8a74e6bcd505ac8407388cc12ac3cfda3d0d`, blob `25ed6675ec64d04f8fcb78ec26680794878a283e`. It defines \(B_r\) using factorization support and states the main, stronger, \(k\leq 2\), lower-bound, and higher-full-parts interfaces. Every theorem body ends in `by sorry`; this is formal statement infrastructure, not a checked Lean proof. The V2 ledger alone decides accepted coverage. No GCL mathematical certification is supplied here.

What remains

The primary-source lock should first preserve the original Erdős–Graham wording and page reference while recording the later van Doorn clarification. A bounded mathematical successor could then formalize and prove the Pell congruence \(x_{j_t}^2\equiv-1\pmod {5^t}\) for \(j_1=1\) and \(j_{t+1}=5j_t+2\), together with the exact growth estimate needed to turn \(5^t\) into a logarithmic factor. Acceptance would require a proof for every \(t\) and a separate check that the induced \(B_2\) divisibility is stated correctly; failure of either implication would falsify the proposed route without deciding the original \(n^{2+o(1)}\) question.

The first main question therefore remains open on the evidence reviewed here. The stronger \(n^2\) claim is contradicted in the maintained page’s discussion for \(k\geq 3\), while the \(k\leq 2\) case is elementary. The higher r extension has a quantifier discrepancy and only reported, not accepted, progress.

Sources and claim boundaries

The documentary source set is issue #1637, the pinned catalogue, the maintained live entry, and Walker’s primary paper PDF, whose observed SHA-256 is `d838fee02ac7fd8341f0eaa2bc9b7e44573c3ab527bdf38fe4ecc9d18382f025`. The original `SOURCE RESULT/1` citation is authenticated original RESULT/1; no future comment or link is fabricated.

The catalogue and live page provide dated status reports, not mathematical authority. The Walker paper is primary evidence for Pell methods in consecutive powerful pairs, not for the full Erdős 367 claim. The current FormalConjectures file records quantified interfaces with `sorry`, not a replayed proof. The finite factorization above is independently verified exact arithmetic, not an asymptotic theorem. All acceptance, certification, and coverage decisions remain with the protected GCL controls and the V2 ledger.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #1637 — 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