Proof of the main conjecture in Vinogradov's mean value theorem for degrees higher than three (Q350559)

This is a most important paper, proving essentially optimal exponents for the estimation of Vinogradov's mean value. \par Let $s,n$ and $N$ be positive integers and let $J_{s,n}(N)$ be the Vinogradov mean value \[ \int_0^1\ldots\int_0^1 \left|\sum_{j\le N}e(x_1j+\ldots+x_nj^n)\right|^{2s}\,dx_1\ldots dx_n. \] Then it is proved that \[ J_{s,n}(N)\ll_{s,n,\varepsilon}N^{s+\varepsilon}+N^{2s-n(n+1)/2+\varepsilon} \] for any $\varepsilon>0$. This estimate is optimal, apart from a possible factor $N^{\varepsilon}$, which can be removed when $s>n(n+1)/2$. It would be good to be able to prove such a bound with a strong dependence on $s$ and $n$. \par The cases $s=1$ and $s=2$ are completely elementary, while the case $s=3$ was recently resolved by \textit{T. D. Wooley} [Adv. Math. 294, 532--561 (2016; Zbl 1365.11097)]. In contrast to previous approaches to the problem, the current paper uses techniques from harmonic analysis, and in particular builds on the recent developments in decoupling theory, see [\textit{J. Bourgain} and \textit{C. Demeter}, Ann. Math. (2) 182, No. 1, 351--389 (2015; Zbl 1322.42014)], for example. Also crucial are results on multilinear Kakeya problems, as in the work of \textit{J. Bennett} et al. [Acta Math. 196, No. 2, 261--302 (2006; Zbl 1203.42019)], and an ``induction on scales'' method, taken from the work of \textit{J. Bourgain} and \textit{C. Demeter} [J. Anal. Math. 133, 279--311 (2017; Zbl 1384.42016)].