Zentralblatt MATH
Publications of (and about) Paul Erdös
Zbl.No: 476.10045
Autor: Erdös, Pál; Szemeredi, Endre
Title: Remarks on a problem of the American Mathematical Monthly. (In Hungarian)
Source: Mat. Lapok 28, 121-124 (1980).
Review: Let A = a1 < a2 < ... be a sequence of positive integers. Let F(A,x,i) denote the number of k's for which the least common multiple [ak,ak+1,...,ak+i-1] satisfies in the inequality [ak,ak+1,...,ak+i-1] \leq x. Some years ago P.Erdös formulated the problem in Am. Math. Mon. to prove that F(A,x,i) < ci x1/i, where ci is a constant depending only on i. This statement is false. This can be seen from the following results of the paper (see III).
I. For any A we have lim\supx > oo \frac{F(A,x,2)}{\sqrt{x}} \leq sumk = 1oo \frac{k ½-(k-1) ½}{k} and
liminfx > oo \frac{F(A,x,2)}{\sqrt{x}} = 0  (1) provided that in (1) the sign = holds.
II. For any A we have liminfx > oo \frac{F(A,x,2)}{\sqrt{x}} \leq ½ .
III. If i > 4, then there exists an \alphai > 0 such that for each sufficiently large x and suitable A we have F(A,x,i) > x 1/i +\alphai. The authors conjecture that III holds for i = 4, too. For i = 3 they have proved that for each sufficiently large x and any A F(A,x,3) < c0x1/3 log x (c0 > 0) and that there is such an A that F(A,x,3) > c1x1/3 log x (c1 > 0) for infinitely many x holds. The question whether there is such an A that for each x the inequality F(A,x,3) > c2x1/3 log x (c2) holds, remain open.
Reviewer: T.Salát
Classif.: * 11B83 Special sequences of integers and polynomials
11A05 Multiplicative structure of the integers
Keywords: sequence of positive integers; least common multiple
© European Mathematical Society & FIZ Karlsruhe & Springer-Verlag