Denominators of Egyptian fractions (Q5905811)

The authors obtain, by elementary methods, good upper and lower bounds for the size of the denominators of Egyptian expansions of fractions and also state several related conjectures. A fraction \(a/b\) is said to be written in Egyptian form if we write \(a/b=1/n_1+1/n_2+ \ldots +1/n_k\), \(n_1<n_2< \ldots <n_k\), where the \(n_i\) are positive integers. Let \(D(a,b)\) be the minimal value of \(n_k\) in all expansions of \(a/b\). Let \(D(b)\) be given by \(D(b)= \max \{D(a,b): 0<a<b \}\). In this work it is shown that \(D(b) \leq Kb(ln b)^3\) for some constant \(K\) and that for \(P\) a prime \(D(P) \geq P \{\{\log_2P \} \}\) where \(\{\{x \} \} =-[-x]\) is the least integer not less than \(x\). Both theoretical and computational evidence are given to indicate that \(D(N)/N\) is maximum when \(N\) is a prime. A number of special cases are dealt with, for example, the authors prove that \(D(P^n)<2P^{n-1}D(P)\). Among the conjectures stated the two of most general interest are, perhaps, (i) \(D(N)\) is submultiplicative, i.e., \(D(N \cdot M) \leq D(N) \cdot D(M)\). If true, relative primeness of \(M\) and \(N\) is probably irrelevant. (ii) Let \(n_1<n_2< \ldots\) be an infinite sequence of positive integers such that \(n_{i+1}/n_i>c>1\). Can the set of rationals \(a/b\) for which \(a/b=1/n_{i_1}+1/n_{i_2}+ \ldots +1/n_{i_t}\) is solvable for some \(t\) contain all the rationals in some interval \((\alpha , \beta)\). We conjecture not. The main results have been improved upon in a second paper by the same authors [Illinois J. Math. 20, 598-613 (1976; Zbl 0336.10007.]