On probability measures of the multidimensional complex plane (Q1901208)

Let \(\mathbb{C}\) denote the complex plane, \(\mathbb{C}^s= \undersetbrace s\to{\mathbb{C}\times\cdots \times \mathbb{C}}\), \(s\geq 1\), and let \({\mathfrak B}(\mathbb{C}^s)\) stand for the class of Borel subsets of the space \(\mathbb{C}^s\). The paper gives a criterion of the weak convergence of probability measures on \((\mathbb{C}^s, {\mathfrak B}(\mathbb{C}^s))\) based on the trigonometric form of complex numbers by terms of characteristic transforms. Let \(P\) be a probability measure on \((\mathbb{C}^s, {\mathfrak B}(\mathbb{C}^s))\), and let \(P_j\), \(j= 1,\dots, s\), \(P_{j,l}\), \(j,l= 1,\dots s\), \(j\neq l\), \(P_{1,\dots, j- 1, j+1,\dots, s}\), \(j= 1,\dots, s\), denote its one-dimensional, two-dimensional, \dots, \((s- 1)\)- dimensional marginal distribution, respectively. Then the functions \[ w_j(t_j, k_j)= \int_{\mathbb{C}} r_j^{it_j} e^{ik_j\varphi_j} dP_j,\qquad j= 1,\dots, s, \] \[ w_{j,l}(t_j, t_l, k_j, k_l)= \int_{\mathbb{C}^2} r_j^{it_j} r_l^{it_l} e^{ik_j \varphi_j+ ik_l \varphi_l} dP_{j, l},\quad j,l= 1,\dots, s,\;j\neq l, \] \[ w({\mathbf t}, {\mathbf k})= \int_{\mathbb{C}^s} {\mathbf r}^{{\mathbf i}{\mathbf t}} e^{i({\mathbf k},\underline {\varphi})} dP \] are called the characteristic transforms of the measure \(P\) provided that the integrands are equal to zero when \(r_j= 0\) for some \(j= 1,\dots, s\). Here \({\mathbf r}^{{\mathbf i}{\mathbf t}}= r_1^{it_1}\cdots r_s^{it_s}\), \(\underline {\varphi}= (\varphi_1,\dots, \varphi_s)\), \(r_j= |z_j|\), \(\varphi_j= \arg z_j\), \({\mathbf t}= (t_1,\dots, t_s)\), \({\mathbf k}= (k_1,\dots, k_s)\), \(t_j\in \mathbb{R}\), \(k_j\in \mathbb{Z}\), \(({\mathbf k}, \underline {\varphi})= k_1 \varphi_1+\cdots+ k_s\varphi_s\). It turns out that the measure \(P\) is uniquely determined by its characteristic transforms. The probability measure \(P_n\) converges weakly in the sense \(\mathbb{C}^s\) to \(P\) as \(n\to \infty\) if \(P_n\) converges weakly to \(P\) and if in addition \[ P_{j;n}(0)@> n\to\infty>> P_j(0),\qquad j= 1,\dots, s, \] \[ P_{j,l; n}(\mathbb{C}^2\backslash (\mathbb{C}\backslash \{0\})^2)@> n\to \infty>> P_{j,l}(\mathbb{C}^2\backslash (\mathbb{C} \backslash \{0\})^2),\quad j,l= 1,\dots, s,\;j\neq l, \] \[ P_n(\mathbb{C}^s\backslash (\mathbb{C}\backslash \{0\})^s)@> n\to \infty>> P(\mathbb{C}^s\backslash (\mathbb{C}\backslash \{0\})^s). \] If the characteristic transforms of the measure \(P_n\) converge to some functions as \(n\to \infty\), continuous in \(t\) at zero, then there exists a probability measure \(P\) on \((\mathbb{C}^s, {\mathfrak B}(\mathbb{C}^s))\) such that \(P_n\) converges weakly in the sense \(\mathbb{C}^s\) to \(P\) as \(n\to \infty\). On the contrary, if \(P_n@> n\to \infty>> P\) in the sense \(\mathbb{C}^s\), then its characteristic transforms converge to corresponding characteristic transforms of \(P\).