Abstract. We define a subset Z of (1, +∞) with the property that for each α ∈ Z there is a nonzero real number ξ = ξ (α) such that the integral parts [ξα n ] are even for all n ∈ .ގ A result of Tijdeman implies that each number greater than or equal to 3 belongs to Z. However, Mahler's question on whether the number 3/2 belongs to Z or not remains open. We prove that the set S := (1, +∞) \ Z is nonempty and find explicitly some numbers in Z ∩ (5/4, 3) and in S ∩ (1, 2).2000 Mathematics Subject Classification. 11J71, 11R04, 11R06, 11R09.
Introduction.Suppose that α > 1 is a real number. Then either for any real number ξ = 0 the fractional parts {ξα n } are greater than or equal to 1/2 for infinitely many n ∈ ގ or there exists a real number ξ = ξ (α) = 0 such that {ξα n } < 1/2 for every n ∈ .ގ The set of α > 1 for which the first possibility holds will be denoted by S. Similarly, we will denote by Z the set of α > 1 for which the second possibility holds, so S ∩ Z = ∅ and S ∪ Z = (1, +∞). Equivalently, α ∈ Z if and only if there is a real number ξ = ξ (α) = 0 such that the integral parts [ξα n ], n = 1, 2, . . . , are all even. In 1968, Mahler [16] asked whether the number 3/2 belongs to S or to Z (see also [20]). There are many reasons to believe that 3/2 ∈ S. Although for a 'random' pair ξ, α the fractional parts {ξα n }, n = 1, 2, . . . , are uniformly distributed in [0, 1) (see [13], [21]), the behavior of the sequence {ξα n }, n = 1, 2, . . . , for almost any 'specific' pair ξ, α, where Trivially, {2, 3, 4, . . .} ⊂ Z, because, for any integer g ≥ 2, by taking ξ = 2 we see that the numbers [2g n ] = 2g n , n = 1, 2, . . . , are all even. In general, one may expect that 'large' α belong to the set Z (see, for instance, Tijdeman's result [19] stated in Theorem 1(i) below) whereas 'small' α lie in S. In connection with this, one can ask whether there exist α > α > 1 such that α ∈ S and α ∈ Z. If the answer to this question were negative then the sets S and Z would simply be two intervals. Unfortunately, the situation is not that simple, because such α and α do exist. We will show, for instance, that the set Z contains a number smaller than 1.26 and that the set S contains the golden mean (1 + √ 5)/2 = 1.61803 . . . . In order to state our theorems, we recall first that α > 1 is called a Pisot number if it is an algebraic integer whose conjugates over ޑ different from α itself lie in the open unit disc. A Pisot number is called a strong Pisot number if it is not a rational integer and if its second largest (in modulus) conjugate is positive (see [2] and [6]). Also,