An extended Vinogradov's mean value theorem

In this paper, we provide novel mean value estimates for exponential sums related to the extended main conjecture of Vinogradov's mean value theorem, by developing the Hardy-Littlewood circle method together with a refined shifting variables argument. Let $d\geq 2$ be a natural number and $\boldsymbol{\alpha}=(\alpha_d,\ldots, \alpha_1)\in \mathbb{R}^d.$ Define the exponential sum \begin{equation*} f_d(\boldsymbol{\alpha};N):=\sum_{1 \leq n \leq N}e(\alpha_d n^d + \cdots+ \alpha_1 n). \end{equation*} For $p>0$, consider mean values of the exponential sums \begin{equation*} \mathcal{I}_{p,d}(u;N):=\int_{[0,1)\times [0,N^{-u})\times [0,1)^{d-2}}|f_d(\boldsymbol{\alpha};N)|^pd\boldsymbol{\alpha}, \end{equation*} where we wrote $d\boldsymbol{\alpha}=d\alpha_1 d\alpha_2\cdots d\alpha_{d-1}d\alpha_d.$ By making use of the aforementioned tools, we obtain the sharp upper bound for $\mathcal{I}_{p,d}(u;N)$, for $d=2,3$ and $0<u\leq 1$. Furthermore, for $d \geq 4$, we obtain analogous results depending on a small cap decoupling inequality for the moment curves in $\mathbb{R}^d.$