On a sum involving the Euler function (Q5920209)

In this surprising and well-written paper, the authors aim at estimating the sum \(S(x)=\sum_{n\le x}\varphi([x/n])\), where \([y]\) is the integer part of \(y\). An equivalent question is to determine an asymptotic for \(\sum_{n\le x}\varphi(n)(|\frac{x}{n}]-[\frac{x}{n+1}])\). It is shown in this paper, by using exponential pair techniques, that \(S(x)/(x\log x)\) lies asymptotically between \(\frac{2629}{4009\zeta(2)}\) and \(\frac{2629}{4009\zeta(2)}+\frac{1380}{4009}\) and the question is asked whether \(S(x)\) is equivalent to \((x\log x)/\zeta(2)\). The authors do not venture into stating a conjecture as this simple looking sum remains mysterious. As consequences of their query, the authors solve a number of similar problems concerning sums \(\sum_{n\le x}f([x/n])\) for arithmetical functions that grow more slowly than the Euler \(\varphi\)-function, for instance when \(f(n)\ll n^{\varphi-1}/(1+\log n)^A\) for some positive \(A\) and where \(\varphi=\frac{1+\sqrt{5}}{2}\). They also show that the growth condition may be replaced by the average condition \(\sum_{n\le x}|f(n)|^2\ll x^\alpha\) for some \(\alpha\in(0,2)\). The proofs are combinatorial and rely at their beginning on the Fourier development of the integer part. As noted by the authors, the sum \(\sum_{n\le x}[x/n]\) is well-known and \(S(x)\) may be thought as a simple modification of it. This paper is a very good read for anyone interested in this attracting question.