A generalization of Vinogradov's mean value theorem (Q5316044)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 2204369
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | A generalization of Vinogradov's mean value theorem |
scientific article; zbMATH DE number 2204369 |
Statements
A generalization of Vinogradov's mean value theorem (English)
0 references
12 September 2005
0 references
Vinogradov's mean value theorem
0 references
many variables
0 references
efficient differencing
0 references
multiple exponential sums
0 references
0.9495584
0 references
0 references
0.9389522
0 references
0.9339868
0 references
0 references
0 references
This paper considers the generalization of Vinogradov's mean value theorem to many variables. Let \(|{\mathbf i}|= i_1+\cdots+ i_d\) and let \[ F(x,\pmb\alpha)= \sum_{1\leq|{\mathbf i}|\leq k}\alpha_{{\mathbf i}} x^{{\mathbf i}}, \] be the general polynomial of total degree at most \(k\) in the variables \(x= (x_1,\dots,x_d)\), and having no constant term. Then the mean value to be estimated is \[ J_{s,k,d}(P)= \int_{\pmb\alpha\in [0, 1]^r} |f(\pmb\alpha,P)|^{2s} \,d\pmb\alpha, \] where \(r\) is the number of coefficients \(\alpha_i\) in \(F(x,\pmb\alpha)\) and \[ f(\pmb\alpha, P)= \sum_{x\in [1, P]^d\cap\mathbb{Z}^d} e(F(x,\pmb\alpha)), \] in which \(e(t)= \exp(2\pi it)\) as usual. There is a trivial lower bound \[ J_{s,k,d}(P)\gg_{s,k,d} P^{sd-K},\quad K= \sum^k_{l=1} l\begin{pmatrix} L+ d-1\\ l\end{pmatrix}. \] The main result is a corresponding upper bound \[ J_{s,k,d}(P)\ll_{s,k,d} P^{sd-K+\Delta_s}, \] where \[ \Delta_s= \begin{cases} rk^{2-2s/rk}\quad &\text{if }1\leq s\leq s_0,\\ r(\log k)^2 e^{3-3(s- s_0)/2rk}\quad &\text{if }s> s_0.\end{cases} \] Here \(s_0= rk(\log k- 2\log\log k)/2\). A related result has been obtained by \textit{G. I. Arkhipov}, \textit{A. A. Karatsuba} and \textit{V. N. Chubarikov} [Sov. Math., Dokl. 21, 715--717 (1980); translation from Dokl. Akad. Nauk SSSR 252, 275--276 (1980; Zbl 0478.10020)]. This relates to polynomials in which \({\mathbf i}\in [0,k]^d\), rather than \(1\leq|{\mathbf i}|\leq k\). Estimates for \(J_{s,k,d}(P)\) derived from this involve an exponent AS which decays more slowly than above, being roughly \(rke^{-s/rk}\). The author also proves an estimate \[ f(\pmb\alpha)\ll_{k,d,\varepsilon} P^{d-\sigma+\varepsilon}, \quad \sigma= \{\textstyle{{3\over 8}} rk\log rk\}^{-1}, \] when \(k\) is suitably large in terms of \(d\), and \(\alpha\) satisfies an appropriate Diophantine approximation constraint. Finally an asymptotic formula is derived for \(J_{s,k,d}(P)\) when \(k\) is large in terms of \(d\) and \[ s\geq rk(\textstyle{{2\over 3}}\log r+ \textstyle{{1\over 2}}\log k+ \log\log k+ 2d+ 4). \] When \(d= 1\) \textit{T. D. Wooley} [Monatsh. Math. 122, No. 3, 265--273 (1996; Zbl 0881.11043)] establishes an asymptotic formula for the better range in which the coefficient \({2\over 3}\) is replaced by \({1\over 2}\). The present paper explains why the case \(d= 1\) should behave differently. The principal results are proved by recourse to \textit{T. D. Wooley's} ``efficient differencing'' procedure, originating in [Mathematika 39, No. 2, 379--399 (1992; Zbl 0769.11036)].
0 references