On fractional sum of the von Mangoldt function (Q7008060)

The paper under review belongs to a long line of articles dealing with sums of the shape \N\[\NS_f (x) := \sum_{n \leqslant x} f \left( \lfloor x/n \rfloor \right),\N\]\Nwhere \(\lfloor t \rfloor\) is the integer part of \(t\) and \(f\) is any usual arithmetic function. The present article focuses on \(f = \Lambda\) the von Mangoldt function, for which it has been proved in [the reviewer et al., J. Number Theory 202, 278--297 (2019; Zbl 1454.11009)] that \N\[\NS_\Lambda (x) = Cx + O \left( x^{1/2+\varepsilon}\right),\N\]\Nwhere \(C \approx 0.4498\). Subsequently, a large number of specialists have broken the \(\frac{1}{2}\)-barrier in the error term, until this work in which the authors show that \N\[\NS_\Lambda (x) = Cx + O \left( x^{22/47+\varepsilon}\right).\N\]\NThe ingredients of the proof are well known to the experts: Vaughan's identity, Vaaler's approximation of the \(\psi\)-function where \(\psi(x):=x-\lfloor x \rfloor - \frac{1}{2}\), Weyl's shift and Van der Corput's inequality, and a new estimate for a triple exponential sum of type II under a constant perturbation, which allows the authors to derive the bound \N\[\N\sum_{n \sim N} \Lambda(n) \psi \left( \frac{x}{n+\delta}\right) \ll x^{22/47+\varepsilon}\N\]\Nfor \(\delta \not \in \mathbb{Z}_{\leqslant -1}\) fixed, and for all large \(x \geqslant 3\) and \(x^{22/47} \leqslant N \leqslant x^{25/47}\), which is the key part of the proof.