Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Statement

Printed p. 128. For integers n≥0n\ge0 and kk, consider the block n+1,n+2,…,n+kn+1,n+2,\ldots,n+k (the paper's display (9), here with only n≥0n\ge0 assumed, "and not n≥kn\ge k").

Theorem 3. "Amongst the integers (9) there are at least (12+o(1))klog⁡k(\frac12+o(1))\frac{k}{\log k} which do not divide the product of the others."

The paper adds that the theorem is best possible: for n=0n=0 and k>5k>5 the block contains exactly π(k)−π(k/2)=12klog⁡k+o(k/log⁡k)\pi(k)-\pi(k/2)=\frac12\frac{k}{\log k}+o(k/\log k) members that do not divide the product of the others.

Source. P. Erdős, On consecutive integers, Nieuw Arch. Wisk. (3) 3 (1955), 124--128; Theorem 3 with its proof and the best-possible remark on printed p. 128.

Read depth. Claims checked: the statement and the remark were read clause by clause on the page image. The proof was read for the sketch below; it is not verified.

Proof pointer

For n≥kn\ge k it follows from Theorem 2: a prime greater than kk divides at most one member of a block of kk consecutive integers, so a member with such a prime factor does not divide the product of the others. For n<kn<k, each prime pp with n+k/2<p<n+kn+k/2<p<n+k divides only one member of the block, and there are 12k/log⁡k+o(k/log⁡k)\frac12k/\log k+o(k/\log k) such primes.

Dependencies

Theorem 2; the prime number theorem.

Bears on

No problem page of this corpus.