Abstract
In this paper, we study the Banach space \(\ell _{\infty }\) of the bounded real sequences, and a measure \(N(a,\Gamma )\) over \(\left( \textbf{R}^{\infty },\mathcal {B}^{\infty }\right) \) analogous to the finite-dimensional Gaussian law. The main result of our paper is a change of variables’ formula for the integration, with respect to \(N(a,\Gamma )\), of the measurable real functions on \(\left( E_{\infty },\mathcal {B}^{\infty }\left( E_{\infty }\right) \right) \), where \(E_{\infty }\) is the separable Banach space of the convergent real sequences. This change of variables is given by some \(\left( m,\sigma \right) \) functions, defined over a subset of \(E_{\infty }\), with values on \(E_{\infty }\), with properties that generalize the analogous ones of the finite-dimensional diffeomorphisms.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
In this paper, we aim to provide a theoretical basis for calculating the integrals of real measurable functions with respect to an infinite-dimensional Gaussian measure; an example is the following integral, shown with Example 42 in Sect. 4:
where
\(\gamma _{ii}\in \textbf{R}^{+}\), for any \(i\in \textbf{N}^{*}\), and \(\sum _{i=1}^{\infty }\gamma _{ii}\in \textbf{R}^{+}\).
With current knowledge of mathematical analysis, this calculation cannot be performed exactly, since the set A is not the product of Borel sets, and so we cannot apply Theorem 44 stated in Sect. 6; moreover, it is evident that, to calculate the integral (1), we need an infinite-dimensional change of variables’ formula for the integration of the measurable real functions, analogous to that used in the finite-dimensional case.
In the mathematical literature, some articles introduced infinite-dimensional measures analogue of the Lebesgue one: see for example the papers of Jessen [12], Léandre [16], Tsilevich et al. [23], Baker [9], and Accardi et al. [1]. Similarly, there is a vast literature about Gaussian measures and their transformations: for example, we can cite the books by Kuo [13] and Bogachev [11]; moreover, in the article of Ustunel et al. [24], the authors prove a change of variables’ formula on abstract Wiener spaces (Theorem 4.1). However, in the theorem proved in the last paper, the set of functions that define the change of variables is quite restricted.
In order to improve the results obtained in [24], in the articles [5,6,7,8], we introduce a family of infinite-dimensional measures that generalize the Lebesgue measure; in particular, in the first two papers, we consider the Banach space \(\ell _{\infty }\subset \textbf{R}^{\infty }\) of the bounded real sequences \(\left\{ x_{n}\right\} _{n\in \textbf{N}^{*}}\), the \(\sigma \)-algebra \(\mathcal {E}_{\infty }\) given by the restriction to \(\ell _{\infty }\) of \(\mathcal {B}^{\infty }\) (defined as the infinite product of the same Borel \(\sigma \)-algebra \(\mathcal {B}\) on \(\textbf{R}\)), and we define a class of functions over an open subset of \(\ell _{\infty }\), with values on \(\ell _{\infty }\), called \(\left( m,\sigma \right) \) standard. Moreover, for any strictly positive integer k, we introduce over the measurable space \(\left( \textbf{R}^{\infty },\mathcal {B}^{\infty }\right) \) a family of infinite-dimensional measures \(\mathcal {\lambda }_{N,a,v}^{(k,\infty )}\), dependent on appropriate parameters N, a, v, given by the product of Lebesgue measures on \(\textbf{R}\) or on compact subsets of \(\textbf{R}\). Furthermore, we provide a change of variables’ formula for the integration of the measurable real functions on \(\left( \ell _{\infty },\mathcal {E}_{\infty }\right) \); this change of variables is defined by particular \(\left( m,\sigma \right) \) standard functions, with properties similar to those of the finite-dimensional diffeomorphisms. In the paper [7], we introduce a class of functions, called \(\left( m,\sigma \right) \), that generalizes the set of the \(\left( m,\sigma \right) \) standard functions given in [6]. The main result is the definition of the determinant of a linear \(\left( m,\sigma \right) \) function, as the limit of a sequence of the determinants of some standard matrices. In the paper [8], we prove that the change of variables’ formula given by the standard finite-dimensional theory and in the papers [5, 6] can be extended by using particular \(\left( m,\sigma \right) \) functions. There are several applications of this result; for example, we can generalize the papers [2,3,4], by considering, for any \(p\in \textbf{R}^{+}\), a Markov chain \(\left\{ X_{n}\right\} _{n\in \textbf{N}}\) on \([0,p)^{\textbf{N}^{*}}\) defined recursively by an affine function. It is possible to prove that, with appropriate assumptions, the sequence \(\left\{ X_{n}\right\} _{n\in \textbf{N}}\) converges with geometric rate to a random variable with law \({\bigotimes }_{i\in \textbf{N}^{*}}\left( \left. \frac{1}{p}Leb\right| _{\mathcal {B}([0,p))}\right) \).
In this paper, we generalize the observation that, if U and V are open sets of the euclidean topology on \(\textbf{R}\), for any \(a,b\in \textbf{R}\), for any \(\Gamma ,\Lambda \in \textbf{R}^{+}\), for any diffeomorphism \(\varphi :U\longrightarrow V\), for any \(B\in \mathcal {B}(V)\) and for any measurable function \(f:\left( V,\mathcal {B}(V)\right) \longrightarrow (\textbf{R},\mathcal {B})\) such that \(f^{+}\) (or \(f^{-}\)) is integrable with respect to the Gaussian measure \(N(a,\Gamma )\), we have
An analogous formula is true by considering, for any \(m\in \textbf{N}^{*}\), a m-dimensional Gaussian measure with arbitrary parameters. Then, we consider the set \(E_{\infty }\) of the convergent real sequences, and the \(\sigma \)-algebra \(\mathcal {B}_{\infty }\) given by the restriction to \(E_{\infty }\) of \(\mathcal {B}^{\infty }\); moreover, we introduce over the measurable space \(\left( \textbf{R}^{\infty },\mathcal {B}^{\infty }\right) \) (and in particular over \(\left( E_{\infty },\mathcal {B}_{\infty }\right) \)) an infinite-dimensional Gaussian measure \(N(a,\Gamma )\), and we prove a change of variables’ formula similar to that given by (2). In this regard, observe that, in this new infinite-dimensional formula, we are forced to use the measure \(N(a,\Gamma )\), because the measure \(\bigotimes _{i\in \textbf{N} ^{*}}Leb\) does not exist.
In Sect. 2, we recall the definitions of the infinite-dimensional measurable spaces \(\left( \ell _{\infty },\mathcal {E}_{\infty }\right) \) and \(\left( E_{\infty },\mathcal {B}_{\infty }\right) \); moreover, we expose results on these spaces, and on the infinite-dimensional matrices. In Sect. 3, we recall some properties of the \(\left( m,\sigma \right) \) functions defined in [7]. In Sect. 4, we present the main theorem of our paper, that is a change of variables’ formula for the integration, with respect to \(N(a,\Gamma )\), of the measurable real functions on \(\left( E_{\infty },\mathcal {B}_{\infty }\right) \); this change of variables is defined by particular \(\left( m,\sigma \right) \) functions (Theorem 41). Moreover, in Example 42, we solve exactly the integral (1), and we compare Theorem 41 with Theorem 4.1 in [24]. In Sect. 5, we expose some ideas for further study in the probability theory, and Sect. 6 is an Appendix about the differentiation theory over \(\ell _{\infty }\), and about the theory of the \(\left( m,\sigma \right) \) functions.
2 The measurable spaces \(\left( \ell _{\infty },\mathcal {E}_{\infty }\right) \) and \(\left( E_{\infty },\mathcal {B}_{\infty }\right) \)
Definition 1
Let \(n\in \textbf{N}^{*}\) and let \(I\ne \emptyset \) be a set; indicate by \(\overline{\textbf{R}}\), by \(\tau \), by \(\tau ^{n}\), by \(\mathcal {B}\), by \(\mathcal {B}^{n}\), by \(\mathcal {B}^{I}\), by \(\mathcal {B} (\overline{\textbf{R}})\), and by Leb, respectively, the set \(\textbf{R} \cup \left\{ -\infty ,\infty \right\} \), the euclidean topology on \(\textbf{R} \), the euclidean topology on \(\textbf{R}^{n}\), the Borel \(\sigma \)-algebra on \(\textbf{R}\), the Borel \(\sigma \)-algebra on \(\textbf{R}^{n}\), the \(\sigma \)-algebra \(\bigotimes _{i\in I}\mathcal {B}_{i}\), with \(\mathcal {B}_{i}=\mathcal {B}\) for every i, the Borel \(\sigma \)-algebra on \(\overline{\textbf{R}}\), and the Lebesgue measure on \(\textbf{R}\). Moreover, for any \(A\subset \overline{\textbf{R}}\), for any \(B\subset \textbf{R}^{n}\), and for any \(C\subset \textbf{R}^{I}\), indicate by \(\mathcal {B}(A)\) the \(\sigma \)-algebra induced on A by \(\mathcal {B} (\overline{\textbf{R}})\), indicate by \(\mathcal {B}^{n}(B)\) the \(\sigma \)-algebra induced on B by \(\mathcal {B}^{n}\), and indicate by \(\mathcal {B} ^{I}(C)\) the \(\sigma \)-algebra induced on C by \(\mathcal {B}^{I}\). Furthermore, for any \(A\subset \textbf{R}\), for any countable and infinite set K, and for any \(B\subset \textbf{R}^{K}\), indicate the set \(A^{K}\) by \(A^{\infty }\), and the \(\sigma \)-algebra \(\mathcal {B}^{K}(B)\) by \(\mathcal {B} ^{\infty }(B)\). Moreover, for any \(\emptyset \ne H\subset I\), define \(x_{I}=\left( x_{i}:i\in I\right) \in \textbf{R}^{I}\) and define the projection \(\pi _{I,H}:\textbf{R}^{I}\longrightarrow \textbf{R}^{H}\) by \(\pi _{I,H}\left( x_{I}\right) =\left( x_{H}\right) \); in particular, in the case \(I=\textbf{N}^{*}\) indicate \(\pi _{I,H}\) by \(\pi _{H}\), and \(\pi _{\left\{ i\right\} }\) by \(\pi _{i}\), for any \(i\in \textbf{N}^{*}\). Furthermore, if \(\left\{ a_{i}:i\in I\right\} \) is a countable set whose elements are not ordered, define \(\sum _{i\in I}a_{i}=\sum _{i\in I}a_{i}^{+}-\sum _{i\in I}a_{i}^{-}\), if this difference exists in \(\overline{\textbf{R}}\) (and so \(\sum _{i\in I}a_{i}\in \textbf{R}\) if and only if \(\sum _{i\in I}\left| a_{i}\right| \in \textbf{R}\)). Moreover, if \(\left\{ f_{i}\right\} _{i\in I}:S\longrightarrow \textbf{R}\) is a countable family of functions, for any \(U\subset S^{I}\), indicate by \(\bigotimes _{i\in I}f_{i}:U\longrightarrow \textbf{R}\) the function defined by \(\left( \bigotimes _{i\in I}f_{i}\right) \left( x_{i}:i\in I\right) =\prod _{i\in I}f_{i}(x_{i})\), \(\forall \,\left( x_{i}:i\in I\right) \in U\). Finally, if E is a topological space, indicate by \(C^{0}(E)\) the set of the continuous functions from E to \(\textbf{R}\), and, if E is a topological vector space, indicate by \(E^{*}\) the set of the linear and continuous functions from E to \(\textbf{R}\).
Definition 2
Let \(I\ne \emptyset \) be a set; define the function \(\left\| \cdot \right\| _{I}:\textbf{R}^{I}\longrightarrow [0,+\infty ]\) by
and define
In particular, suppose that \(I=\left\{ i_{n}:n\in \textbf{N}^{*}\right\} \subset \textbf{R}\) is a countable set such that \(i_{n}<i_{n+1}\), for any \(n\in \textbf{N}^{*}\); define the vector space
Moreover, indicate by \(\left\| x\right\| _{\infty }\) the function \(\left\| x\right\| _{\textbf{N}^{*}}\), by \(\ell _{\infty }\) the set \(\ell _{\textbf{N}^{*}}\), by \(E_{\infty }\) the set \(E_{\textbf{N}^{*}}\), by \(\mathcal {E}_{\infty }\) the \(\sigma \)-algebra \(\mathcal {B}^{\infty }(\ell _{\infty })\), and by \(\mathcal {B}_{\infty }\) the \(\sigma \)-algebra \(\mathcal {B} ^{\infty }(E_{\infty })\); furthermore, for any \(n\in \textbf{N}^{*} \cup \left\{ \infty \right\} \), define the set
Finally, for any \(p\in \left[ 1,\infty \right) \), define the function \(\left\| \cdot \right\| _{p}:\textbf{R}^{\infty }\longrightarrow [0,+\infty ]\) by
and define the vector space
Definition 3
Let \(I\ne \emptyset \) be a set, let \(x_{0}\in \ell _{I}\) and let \(A\subset \ell _{I}\); indicate by \(\tau _{\left\| \cdot \right\| _{I}}\) the topology induced on \(\ell _{I}\) by the norm \(\left\| \cdot \right\| _{I}\), and indicate by \(\tau _{\left\| \cdot \right\| _{I}}(A)\) the topology induced on A by \(\tau _{\left\| \cdot \right\| _{I}}\); moreover, for any \(\varepsilon \in \textbf{R}^{+}\), indicate by \(B_{I}(x_{0},\varepsilon )\) the set \(\{x\in \ell _{I}:\left\| x-x_{0}\right\| _{I}<\varepsilon \}\), and define the set
In particular, if \(I=\textbf{N}^{*}\) and \(A\subset \ell _{\infty }\), indicate by \(\tau _{\left\| \cdot \right\| _{\infty }}\) the topology \(\tau _{\left\| \cdot \right\| _{\textbf{N}^{*}}}\), by \(\tau _{\left\| \cdot \right\| _{\infty }}(A)\) the topology \(\tau _{\left\| \cdot \right\| _{\textbf{N}^{*}}}(A)\), and by \(\mathcal {D}_{\infty }(A)\) the set \(\mathcal {D}_{\textbf{N} ^{*}}(A)\).
Remark 4
Let H and I be two sets such that \(\emptyset \ne H\subset I\); then:
-
1.
\(\pi _{I,H}:\left( \ell _{I},\tau _{\left\| \cdot \right\| _{I}}\right) \longrightarrow \left( \ell _{H},\tau _{\left\| \cdot \right\| _{H}}\right) \) is continuous and open.
-
2.
\(\pi _{I,H}:\left( \textbf{R}^{I},\mathcal {B}^{I}\right) \longrightarrow \left( \textbf{R}^{H},\mathcal {B}^{H}\right) \) is measurable.
Proof
The point 1 follows from Proposition 6 in Asci [6]; moreover, the point 2 can be easily proved. \(\square \)
Remark 5
\(\ell _{\infty }\) is a Banach space, with the norm \(\left\| \cdot \right\| _{\infty }\); moreover, one has \(\ell _{\infty }\in \mathcal {B} ^{\infty }\).
Proof
In order to prove that \(\ell _{\infty }\) is a Banach space, we can see, for example, the proof of Remark 2 in [5]. Moreover, \(\forall \,i\in \textbf{N}^{*}\), the function \(\left( x_{i}:i\in \textbf{N}^{*}\right) \longrightarrow |x_{i}|\) is \(\left( \mathcal {B}^{\infty },\mathcal {B}(\overline{\textbf{R}})\right) \)-measurable, and so the function \(\left( x_{i}:i\in \textbf{N}^{*}\right) \longrightarrow \left\| x\right\| _{\infty }\) is \(\left( \mathcal {B}^{\infty },\mathcal {B} (\overline{\textbf{R}})\right) \)-measurable too, from which \(\ell _{\infty } \in \mathcal {B}^{\infty }\). \(\square \)
Remark 6
\(E_{\infty }\) is a separable Banach space, with the topology \(\tau _{\left\| \cdot \right\| _{\infty }}\left( E_{\infty }\right) \); moreover, one has \(E_{\infty }\in \mathcal {B}^{\infty }\) and \(\mathcal {D} _{\infty }(E_{\infty })=E_{\infty }\).
Proof
Let \(\left\{ x_{n}\right\} _{n\in \textbf{N}}\subset E_{\infty }\) be a sequence such that \(\lim _{n\rightarrow \infty }x_{n}=a\), where \(a=\left( a_{i}:i\in \textbf{N}^{*}\right) \in \ell _{\infty }\); then, \(\forall \,\varepsilon \in \textbf{R}^{+}\), \(\exists \,n_{1}\in \textbf{N}\) such that, \(\forall \,n\in \textbf{N}\), \(n\ge n_{1}\), we have \(\sup _{i\in \textbf{N}^{*}}\left| \left( x_{n}\right) _{i}-a_{i}\right| =\left\| x_{n}-a\right\| _{\infty }<\frac{\varepsilon }{3}\); moreover, since \(x_{n_{1}}\in E_{\infty }\), \(\exists \,n_{2}\in \textbf{N}\) such that, \(\forall \,i,j\in \textbf{N}^{*}\) such that \(i,j\ge n_{2}\), we have \(\left| \left( x_{n_{1}}\right) _{i}-\left( x_{n_{1}}\right) _{j}\right| <\frac{\varepsilon }{3}\), from which
then, \(\left\{ a_{i}\right\} _{i\in \textbf{N}^{*}}\) is a Cauchy sequence, and so, since \(\textbf{R}\) is complete, we have \(a\in E_{\infty }\); thus, \(E_{\infty }\) is closed, and so it is complete, since \(E_{\infty }\subset \ell _{\infty }\); thus, \(E_{\infty }\) is a Banach space. Moreover, let \(I=\left\{ i_{n}:n\in \textbf{N}^{*}\right\} \subset \textbf{R}\) be a countable set such that \(i_{n}<i_{n+1}\), for any \(n\in \textbf{N}^{*}\), and such that \(\lim _{n\rightarrow +\infty }i_{n}\in \textbf{R}\); it is known (see, for example, the Rudin’s book [18]) that the set \(C^{0}\left( \overline{I}\right) \) is separable; moreover, since the set \(E_{I}\) can be identified with \(C^{0}\left( \overline{I}\right) \), \(E_{I}\) is separable too, and so this is true for the set \(E_{\infty }\), which can be identified with \(E_{I}\). Furthermore, \(\forall \,i\in \textbf{N}^{*}\), the function \(\left( x_{i}:i\in \textbf{N}^{*}\right) \longrightarrow x_{i}\) is \(\left( \mathcal {B}^{\infty },\mathcal {B}(\overline{\textbf{R}})\right) \)-measurable, and so
Finally, \(\forall \,x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\), \(\forall \,\varepsilon \in \textbf{R}^{+}\), by setting \(y=\left( x_{i}+\frac{\varepsilon }{2}:i\in \textbf{N}^{*}\right) \), we have \(y\in B_{\infty }(x,\varepsilon )\cap E_{\infty }\backslash \left\{ x\right\} \), and so \(x\in \mathcal {D}_{\infty }(E_{\infty })\), from which \(E_{\infty } \subset \mathcal {D}_{\infty }(E_{\infty })\); /0258since \(E_{\infty }\) is closed, we have \(\mathcal {D}_{\infty }(E_{\infty })\subset E_{\infty }\), and so \(\mathcal {D}_{\infty }(E_{\infty })=E_{\infty }\).\(\square \)
The following two results can be easily proved.
Remark 7
For any \(p\in \left[ 1,\infty \right) \), \(\ell _{p}\) is a Banach space, with the norm \(\left\| \cdot \right\| _{p}\), and one has \(\ell _{p} \in \mathcal {B}^{\infty }\). Moreover, \(\ell _{2}\) is a separable Hilbert space, with the inner product \(\left\langle \cdot ,\cdot \right\rangle :\ell _{2} \times \ell _{2}\longrightarrow \textbf{R}\) defined by \(\left\langle x,y\right\rangle =\sum _{i=1}^{\infty } x_{i}y_{i}\), \(\forall \,x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in \ell _{2}\), \(\forall \,y=\left( y_{i}:i\in \textbf{N}^{*}\right) \in \ell _{2}\).
Remark 8
For any \(p,q\in \left[ 1,\infty \right) \) such that \(p\le q\), one has \(\ell _{p}\subset \ell _{q}\).
Definition 9
Let \(k\in \textbf{N}^{*}\) and let \(B\subset \textbf{R}^{k}\); a function \(g:\left( C_{c}(B),\tau _{\left\| \cdot \right\| _{B}}\left( C_{c}(B)\right) \right) \longrightarrow \left( \textbf{R},\tau \right) \) linear and continuous is called Radon measure on B; moreover, if, for any \(f\in C_{c}(B)\) such that \(f\ge 0\), one has \(g(f)\ge 0\), then g is called nonnegative Radon measure on B.
Theorem 10
(Riesz) Let \(k\in \textbf{N}^{*}\), let \(B\subset \textbf{R}^{k}\) and let \(g:\left( C_{c}(B),\tau _{\left\| \cdot \right\| _{B}}\left( C_{c}(B)\right) \right) \longrightarrow \left( \textbf{R},\tau \right) \) be a nonnegative Radon measure on B; then, there exists a unique measure \(\mu \) on \(\left( B,\mathcal {B}^{k}(B)\right) \) such that
Moreover, every Radon measure on B can be expressed as a difference between two nonnegative Radon measures on B.
Proof
The proof can be found, for example, in the Rudin’s book [18]. \(\square \)
Proposition 11
Let \(I=\left\{ i_{n}:n\in \textbf{N}^{*}\right\} \subset \textbf{R}\) be a countable set such that \(i_{n}<i_{n+1}\), for any \(n\in \textbf{N}^{*}\), and such that \(i_{\infty }\equiv \lim _{n\rightarrow \infty }i_{n}\in \textbf{R}\); moreover, let \(g:\left( C^{0}\left( \overline{I}\right) ,\tau _{\left\| \cdot \right\| _{\overline{I}}}\left( C^{0}\left( \overline{I}\right) \right) \right) \longrightarrow \left( \textbf{R},\tau \right) \) be a linear and continuous function; then, there exists \(\left( a_{i}:i\in \overline{I}\right) \in \textbf{R}^{\overline{I}}\) such that \(\sum _{i\in \overline{I}} \left| a_{i}\right| \in \textbf{R}\) and such that \(g(f)=\sum _{i\in \overline{I}}a_{i}f_{i}\), for any \(f=\left( f_{i}:i\in \overline{I}\right) \in C^{0}\left( \overline{I}\right) \).
Proof
Since \(\overline{I}=I\cup \left\{ i_{\infty }\right\} \) is compact, we have \(C_{c}\left( \overline{I}\right) =C^{0}\left( \overline{I}\right) \); then, g is a Radon measure on \(\overline{I}\) and, from Theorem 10, there exist two Radon measures \(g_{1}\) and \(g_{2}\) nonnegative on \(\overline{I}\), and two measures \(\mu _{1}\) and \(\mu _{2}\) on \(\left( \overline{I},\mathcal {B}\left( \overline{I}\right) \right) \) such that, \(\forall \,f=\left( f_{i}:i\in \overline{I}\right) \in C^{0}\left( \overline{I}\right) \), we have
where \(a_{i}=\mu _{1}\left( \left\{ i\right\} \right) -\mu _{2}\left( \left\{ i\right\} \right) \in \textbf{R}\), \(\forall \,i\in \overline{I}\); moreover, if we choose \(f=1\), we obtain \(\sum _{i\in \overline{I}} \mu _{1}\left( \left\{ i\right\} \right) =g_{1}(1)\in \textbf{R}\) and \(\sum _{i\in \overline{I}}\mu _{2}\left( \left\{ i\right\} \right) =g_{2}(1)\in \textbf{R}\), from which
\(\square \)
Corollary 12
Let \(I=\left\{ i_{n}:n\in \textbf{N}^{*}\right\} \subset \textbf{R}\) be a countable set such that \(i_{n}<i_{n+1}\), for any \(n\in \textbf{N}^{*}\), and such that \(i_{\infty }\equiv \lim _{n\rightarrow +\infty }i_{n}\in \textbf{R}\); moreover, let \(g:\left( E_{I},\tau _{\left\| \cdot \right\| _{I}}\left( E_{I}\right) \right) \longrightarrow \left( \textbf{R},\tau \right) \) be a linear and continuous function; then, there exists \(\left( a_{i}:i\in \overline{I}\right) \in \textbf{R}^{\overline{I}}\) such that \(\sum _{i\in \overline{I}} \left| a_{i}\right| \in \textbf{R}\) and such that \(g(f)=\sum _{i\in \overline{I}}a_{i}f_{i}\), for any \(f=\left( f_{i}:i\in I\right) \in E_{I}\), where \(f_{i_{\infty }}\equiv \lim _{n\rightarrow \infty } f_{i_{n}}\).
Proof
\(\forall \,f=\left( f_{i}:i\in \overline{I}\right) \in C^{0}\left( \overline{I}\right) \), we have \(\left( f_{i}:i\in I\right) \in E_{I}\); then, let \(h_{g}:C^{0}\left( \overline{I}\right) \longrightarrow \textbf{R}\) be the linear function defined by \(h_{g}\left( f_{i}:i\in \overline{I}\right) =g\left( f_{i}:i\in I\right) \); \(\forall \,\varepsilon \in \textbf{R}^{+}\), \(\forall \,x=\left( x_{i}:i\in \overline{I}\right) \in C^{0}\left( \overline{I}\right) \), \(\exists \,\delta \in \textbf{R}^{+}\) such that, \(\forall \,y_{I}=\left( y_{i}:i\in I\right) \in E_{I}\) such that \(\left\| y_{I}-x_{I}\right\| _{I}<\delta \), we have \(\left| g\left( y_{I}\right) -g\left( x_{I}\right) \right| <\varepsilon \); then, \(\forall \,y\in C^{0}\left( \overline{I}\right) \) such that \(\left\| y-x\right\| _{\overline{I}}<\delta \), we obtain
then, \(h_{g}\) is \(\left( \tau _{\left\| \cdot \right\| _{\overline{I}} }\left( C^{0}\left( \overline{I}\right) \right) ,\tau \right) \)-continuous and, from Proposition 11, there exists \(\left( a_{i}:i\in \overline{I}\right) \in \textbf{R}^{\overline{I}}\) such that \(\sum _{i\in \overline{I}}\left| a_{i}\right| \in \textbf{R}\) and such that \(h_{g}(l)=\sum _{i\in \overline{I}}a_{i}l_{i}\), \(\forall \,l=\left( l_{i}:i\in \overline{I}\right) \in C^{0}\left( \overline{I}\right) \); thus, let \(f=\left( f_{i}:i\in I\right) \in E_{I}\) and let \(l\in C^{0}\left( \overline{I}\right) \) such that \(l_{I}=f\); we have
where \(f_{i_{\infty }}\equiv \lim _{n\rightarrow \infty }f_{i_{n}} \). \(\square \)
Corollary 13
Let \(I=\left\{ i_{n}:n\in \textbf{N}^{*}\right\} \subset \textbf{R}\) be a countable set such that \(i_{n}<i_{n+1}\), for any \(n\in \textbf{N}^{*}\), and let \(i_{\infty }=\lim _{n\rightarrow +\infty }i_{n}\in \textbf{R}\cup \left\{ \infty \right\} \); then:
-
1.
For any function \(g:\left( E_{I},\tau _{\left\| \cdot \right\| _{I}}\left( E_{I}\right) \right) \longrightarrow \left( \textbf{R},\tau \right) \) linear and continuous, there exists \(\left( a_{i}:i\in I\cup \left\{ i_{\infty }\right\} \right) \in \textbf{R}^{I\cup \left\{ i_{\infty }\right\} }\) such that \(\sum _{i\in I\cup \left\{ i_{\infty }\right\} }\left| a_{i}\right| \in \textbf{R}\) and such that
$$\begin{aligned} g(f)=\underset{i\in I\cup \left\{ i_{\infty }\right\} }{\sum }a_{i}f_{i},\quad \forall \,f=\left( f_{i}:i\in I\right) \in E_{I}, \end{aligned}$$(3)where \(f_{i_{\infty }}\equiv \lim _{n\rightarrow \infty }f_{i_{n}}\).
-
2.
For any \(\left( a_{i}:i\in I\cup \left\{ i_{\infty }\right\} \right) \in \textbf{R}^{I\cup \left\{ i_{\infty }\right\} }\) such that \(\sum _{i\in I\cup \left\{ i_{\infty }\right\} }\left| a_{i}\right| \in \textbf{R}\), the function \(g:\left( E_{I},\tau _{\left\| \cdot \right\| _{I}}\left( E_{I}\right) \right) \longrightarrow \left( \textbf{R},\tau \right) \) defined by (3) is linear and continuous.
Proof
-
1.
If \(i_{\infty }\in \textbf{R}\), the statement follows from Corollary 12. Conversely, suppose that \(i_{\infty }=\infty \), and let \(J=\left\{ j_{n}:n\in \textbf{N}^{*}\right\} \subset \textbf{R}\), where \(j_{n} =1-2^{-n}\), \(\forall \,n\in \textbf{N}^{*}\); moreover, let \(\rho :E_{J}\longrightarrow E_{I}\) be the function defined by \(\left( \rho \left( f\right) \right) _{i_{n}}=f_{j_{n}}\), \(\forall \,f=\left( f_{j_{n}}:n\in \textbf{N}^{*}\right) \in E_{J}\), \(\forall \,n\in \textbf{N}^{*}\); observe that \(\rho \) is linear and \(\left( \tau _{\left\| \cdot \right\| _{J} }\left( E_{J}\right) ,\tau _{\left\| \cdot \right\| _{I}}\left( E_{I}\right) \right) \)-continuous, and so the function \(\left( g\circ \rho \right) :E_{J}\longrightarrow \textbf{R}\) is linear and \(\left( \tau _{\left\| \cdot \right\| _{J}}\left( E_{J}\right) ,\tau \right) \)-continuous; then, let \(f=\left( f_{i}:i\in I\right) \in E_{I}\), and let \(h=\left( h_{j}:j\in J\right) =\rho ^{-1}(f)\in E_{J}\); since \(j_{\infty }\equiv \lim _{n\rightarrow \infty }j_{n}=1\in \textbf{R}\), from Corollary 12, there exists \(\left( b_{j}:j\in \overline{J}\right) \in \textbf{R}^{\overline{J}}\) such that \(\sum _{j\in \overline{J}}\left| b_{j}\right| \in \textbf{R}\) and such that
$$\begin{aligned} g(f)=\left( g\circ \rho \right) (h)=\underset{j\in \overline{J}}{\sum } b_{j}h_{j}=\underset{i\in I\cup \left\{ i_{\infty }\right\} }{\sum }a_{i} f_{i}, \end{aligned}$$where \(a_{i_{n}}=b_{j_{n}}\), \(\forall \,n\in \textbf{N}^{*}\cup \left\{ \infty \right\} \),and \(f_{i_{\infty }}\equiv \lim _{n\rightarrow \infty } f_{i_{n}}\).
-
2.
It is easy to prove that \(g:E_{I}\longrightarrow \textbf{R}\) is linear; moreover, let \(x_{0}=\left( x_{0,i}:i\in I\right) \in E_{I}\), let \(\varepsilon \in \textbf{R}^{+}\) and let \(\delta =\varepsilon \left( 2 \sum _{i\in I\cup \left\{ i_{\infty }\right\} }\left| a_{i}\right| \right) ^{-1}\in \textbf{R}^{+}\); \(\forall \,x=\left( x_{i}:i\in I\right) \in E_{I}\) such that \(\left\| x-x_{0}\right\| _{I}<\delta \), we have
$$\begin{aligned} \left| g(x)-g\left( x_{0}\right) \right|= & {} \left| \underset{i\in I}{\sum }a_{i}(x_{i}-x_{0,i})+a_{i_{\infty }}\underset{n\rightarrow \infty }{\lim }\left( x_{i_{n}}-(x_{0})_{i_{n}}\right) \right| \\\le & {} \delta \underset{i\in I\cup \left\{ i_{\infty }\right\} }{\sum }\left| a_{i}\right| =\frac{\varepsilon }{2}<\varepsilon , \end{aligned}$$and so g is continuous in \(x_{0}\); then, since \(x_{0}\in E_{I}\) is arbitrary, g is \(\left( \tau _{\left\| \cdot \right\| _{I}}\left( E_{I}\right) ,\tau \right) \)-continuous.
\(\square \)
Proposition 14
One has \(\mathcal {B}_{\infty }=\sigma \left( g:g\in (E_{\infty })^{*}\right) =\sigma \left( g:g\in C^{0}(E_{\infty })\right) =\sigma \left( \tau _{\left\| \cdot \right\| _{\infty }}\left( E_{\infty }\right) \right) \).
Proof
\(\forall \,i\in \textbf{N}^{*}\), let \(\pi _{i}:E_{\infty }\longrightarrow \textbf{R}\) be the projection defined by \(\pi _{i}(x)=x_{i}\), \(\forall \,x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\); \(\forall \,A\in \mathcal {B}\), \(\forall \,i\in \textbf{N}^{*}\), we have \(\pi _{i}^{-1}(A)=\left( \textbf{R}^{i-1}\times A\times \textbf{R}^{\infty }\right) \cap E_{\infty }\), from which \(\mathcal {B}_{\infty }=\mathcal {B}^{\infty }\left( E_{\infty }\right) =\sigma \left( \pi _{i}:i\in \textbf{N}^{*}\right) \), and so \(\mathcal {B}_{\infty }\subset \sigma \left( g:g\in (E_{\infty })^{*}\right) \). Moreover, from Corollary 13, \(\forall \,g\in (E_{\infty })^{*}\), there exists \(\left( a_{i}:i\in \textbf{N}^{*}\cup \left\{ \infty \right\} \right) \in \textbf{R}^{\textbf{N}^{*}\cup \left\{ \infty \right\} }\) such that \(g(x)=\sum _{i\in \textbf{N}^{*}\cup \left\{ \infty \right\} } a_{i}x_{i}\), \(\forall \,x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\), where \(x_{\infty }\equiv \lim _{i\rightarrow \infty }x_{i}\), and so
then, g is \(\left( \sigma \left( \pi _{i}:i\in \textbf{N}^{*}\right) ,\mathcal {B}\right) \)-measurable, from which
and so \(\mathcal {B}_{\infty }=\sigma \left( g:g\in (E_{\infty })^{*}\right) \); finally, since \(E_{\infty }\) is a separable Banach space, from Theorem A.3.7 in [11], we have \(\sigma \left( g:g\in (E_{\infty })^{*}\right) =\sigma \left( g:g\in C^{0}(E_{\infty })\right) =\sigma \left( \tau _{\left\| \cdot \right\| _{\infty }}\left( E_{\infty }\right) \right) \). \(\square \)
The following concept generalizes Definition 6 in [5] (see also the theory in the Lang’s book [15] and that in the Weidmann’s book [25]).
Definition 15
Let \(m,n\in \textbf{N}^{*}\cup \left\{ \infty \right\} \) and let \(A=\left( a_{ij}\right) _{i\in I_{m},j\in I_{n}}\) be a real matrix \(m\times n\) (eventually infinite); define the linear function \(A=\left( a_{ij}\right) _{i\in I_{m},j\in I_{n}}:\textbf{R}^{n}\longrightarrow \textbf{R}^{m}\), and write \(x\longrightarrow Ax\), in the following manner:
on condition that, for any \(i\in I_{m}\), the sum in (4) converges to a real number. Furthermore, if \(m=n\) and \(a_{ij}=0\) for any \(i\ne j\), the real matrix A and the corresponding function are called diagonal; in particular, indicate by \(\textbf{I}_{n}=\left( \delta _{ij}\right) _{i,j\in I_{n}}\) the diagonal real matrix defined by
Moreover, for any \(\emptyset \ne L\subset I_{m}\), for any \(\emptyset \ne N\subset I_{n}\), indicate by \(A^{(L,N)}\) the real matrix \(\left( a_{ij}\right) _{i\in L,j\in N}\), and, for any\(\,l,o\in \textbf{N}^{*} \cup \left\{ \infty \right\} \), indicate \(A^{(I_{l},N)}\) by \(A^{(l,N)}\), \(A^{(L,I_{o})}\) by \(A^{(L,o)}\), and \(A^{(I_{l},I_{o})}\) by \(A^{(l,o)}\). Furthermore indicate by \(^{t}A=\left( b_{ji}\right) _{j\in I_{n},i\in I_{m} }:\textbf{R}^{m}\longrightarrow \textbf{R}^{n}\), if it is defined, the linear function given by \(b_{ji}=a_{ij}\), for any \(j\in I_{n}\) and for any \(i\in I_{m}\). Furthermore, if \(m=n\) and \(A=\left. ^{t}A\right. \), the function A is called symmetric. Finally, if \(o\in \textbf{N}^{*}\cup \left\{ \infty \right\} \) and \(B=\left( b_{jk}\right) _{j\in I_{n},k\in I_{o}}\) is a real matrix \(n\times o\), define the \(m\times o\) real matrix \(AB=\left( \left( AB\right) _{ik}\right) _{i\in I_{m},k\in I_{o}}\) by
on condition that, for any \(i\in I_{m}\) and for any \(k\in I_{o}\), the sum in (5) converges to a real number.
Proposition 16
Let \(m,n,o\in \textbf{N}^{*}\cup \left\{ \infty \right\} \) and let \(A=\left( a_{ij}\right) _{i\in I_{m},j\in I_{n}}\) be a real matrix \(m\times n\); then:
-
1.
The linear function \(A=\left( a_{ij}\right) _{i\in I_{m},j\in I_{n} }:\ell _{I_{n}}\longrightarrow \textbf{R}^{m}\) given by (4) is defined if and only if, for any \(i\in I_{m}\), \( \sum _{j\in I_{n}}\left| a_{ij}\right| <\infty \).
-
2.
One has \(\sup _{i\in I_{m}}\sum _{j\in I_{n}}\left| a_{ij}\right| <\infty \) if and only if \(A(\ell _{I_{n}})\subset \ell _{I_{m}}\) and A is continuous; moreover, if \(A(\ell _{I_{n}})\subset \ell _{I_{m}}\), then \(\left\| A\right\| =\sup _{i\in I_{m}} \sum _{j\in I_{n}}\left| a_{ij}\right| \).
-
3.
If \(A(\ell _{I_{n}})\subset \ell _{I_{m}}\), then, for any \(\emptyset \ne L\subset I_{m}\), for any \(\emptyset \ne N\subset I_{n}\), one has \(A^{(L,N)}(\ell _{N})\subset \ell _{L}\).
-
4.
If \(B=\left( b_{jk}\right) _{j\in I_{n},k\in I_{o}}:\ell _{I_{o} }\longrightarrow \ell _{I_{n}}\) is a linear function, then the linear function \(A\circ B:\ell _{I_{o}}\longrightarrow \textbf{R}^{m}\) is defined by the real matrix AB.
Proof
See the proof of Proposition 2.13 in Asci [8]. \(\square \)
3 Theory of the \(\left( m,\sigma \right) \) functions
The following Definition 17 introduces a class of functions \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\), called \(\left( m,\sigma \right) \), that in the case \(U\subset E_{\infty }\) will be used to provide a change of variables’ formula for the integration, with respect to an infinite-dimensional Gaussian measures, of the measurable real functions on \(\left( E_{\infty },\mathcal {B}_{\infty }\right) \). In fact, the properties of some \(\left( m,\sigma \right) \) functions generalize the analogous ones of the standard finite-dimensional diffeomorphisms.
The axioms that define the \(\left( m,\sigma \right) \) functions serve to prove Proposition 38, which in turn is fundamental to proving Proposition 40 and Theorem 41. Essentially, the \(\left( m,\sigma \right) \) functions and their inverse functions must behave similar to infinite-dimensional diagonal matrices, which, when applied to an infinite-dimensional rectangle, generate a new one (see formula (19), where these rectangles are \(\left[ -M,M\right] ^{\infty }\) and \(\left[ -O,O\right] ^{\infty }\) respectively).
In order to understand the concept of a \(\left( m,\sigma \right) \) function \(\varphi \), the reader can consider the particular case where \(\varphi \) is linear, and so it is defined by Remark 20; moreover, if \(\varphi \) is a linear and \(\left( m,\sigma \right) \) function, it is possible define the determinant of \(\varphi \) (see Theorem 22 and Definition 23): a concept without sense, if \(\varphi \) is not \(\left( m,\sigma \right) \). The non linear case of Definition 17 can be seen as a generalization of the linear case. After Definition 23, we provide two examples: Examples 24 and 26; a further example (Example 42) will be exposed at the end of Sect. 4, in order to solve the integral (1).
Definition 17
Let \(m\in \textbf{N}\), let \(\emptyset \ne U^{(m)} \subset \textbf{R}^{m}\), let \(U^{(\infty )}\subset \ell _{\infty }\) such that \(\pi _{i}(U^{(\infty )})=\textbf{R}\) for any \(i\in \textbf{N}^{*}\), let \(U=U^{(m)}\times U^{(\infty )}\subset \mathcal {D}_{\infty }(U)\) (with the convention \(U=U^{(\infty )}\) if \(m=0\)), and let \(\sigma :\textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} \longrightarrow \textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} \) be an increasing function; a function \(\varphi =(\varphi _{i}:i\in \textbf{N}^{*}):U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) is called \(\left( m,\sigma \right) \) if, for any \(i\in \textbf{N}^{*}\) and for any \(j>m\), there exist some functions \(\varphi _{i}^{(\infty ,m)}:U^{(m)}\longrightarrow \textbf{R}\) and \(\varphi _{ij}:\textbf{R}\longrightarrow \textbf{R}\) such that:
-
1.
\(\forall \,i\in \textbf{N}^{*}\), \(\forall \,x=(x_{j}:j\in \textbf{N} ^{*})\in U\), one has
$$\begin{aligned} \varphi _{i}(x)=\varphi _{i}^{(\infty ,m)}(x_{1},\ldots ,x_{m})+\underset{j=m+1}{\overset{\infty }{ {\displaystyle \sum } }}\varphi _{ij}(x_{j}); \end{aligned}$$(6)moreover, \(\forall \,(x_{1},\ldots ,x_{m})\in U^{(m)}\), one has \(\lim _{i\rightarrow \infty }\varphi _{i}^{(\infty ,m)}\left( x_{1},\ldots ,x_{m} \right) =0\).
-
2.
\(\forall \,i\in \left\{ 1,\ldots ,m\right\} \), \(\forall \,x=(x_{j}:j>m)\in U^{(\infty )}\), one has
$$\begin{aligned} \underset{j=m+1}{\overset{\infty }{ {\displaystyle \sum } }}\left| \varphi _{ij}(x_{j})\right| <\infty . \end{aligned}$$ -
3.
\(\forall \,i>m\), \(\forall \,j\in \left\{ 1,\ldots ,m\right\} \), there exists \(\frac{\partial \varphi _{i}^{(\infty ,m)}}{\partial x_{j}}:U^{(m)} \longrightarrow \textbf{R}\); moreover, \(\forall \,(x_{1},\ldots ,x_{m})\in U^{(m)}\), one has \(\sum _{i=m+1}^{\infty }\left| \frac{\partial \varphi _{i}^{(\infty ,m)}(x_{1},\ldots ,x_{m})}{\partial x_{j}}\right| <\infty \).
-
4.
\(\forall \,i>m\), \(\forall \,j>m\) such that \(j\ne \sigma (i)\), \(\forall \,t\in \textbf{R}\), one has \(\varphi _{ij}(t)=0\); moreover, the function \(\varphi _{i,\sigma (i)}\) is constant or derivable such that \(\varphi _{i,\sigma (i)}^{\prime }(t)\ne 0\) \(\forall \,t\in \textbf{R}\), and, \(\forall \,\left( x_{j}:j\in \sigma (\textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} )\right) \in \ell _{\sigma (\textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} )}\), one has \(\sup _{i>m}\left| \varphi _{i,\sigma (i)}^{\prime }(x_{\sigma (i)})\right| <\infty \).
-
5.
By setting
$$\begin{aligned} \mathcal {I}(\varphi )=\left\{ i>m:\sigma (h)=h\text {, }\forall \,h\ge i\text {, and }\varphi _{ii}\text { is not constant}\right\} , \end{aligned}$$\(\forall \,x=(x_{i}:i\in \mathcal {I}(\varphi ))\in \ell _{\mathcal {I}(\varphi )}\), there exists \(\prod _{i\in \mathcal {I}(\varphi )}\varphi _{ii}^{\prime }\left( x_{i}\right) \in \textbf{R}^{*}\); moreover, there exists \(d(\varphi )\in \textbf{R}^{+}\) such that, \(\forall \,i\in \mathcal {I}(\varphi )\), \(\forall \,t\in \textbf{R}\), one has \(\left| \varphi _{ii}^{\prime }(t)\right| \ge d(\varphi )\).
Definition 18
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function; then:
-
1.
\(\varphi \) is called \(\left( m,\sigma \right) \) standard (or \(\left( m,\sigma \right) \) of the first type) if, for any \(i>m\) and for any \(\left( x_{1},\ldots ,x_{m}\right) \in U^{(m)}\), one has \(\varphi _{i}^{(\infty ,m)} (x_{1},\ldots ,x_{m})=0\).
-
2.
\(\varphi \) is called \(\left( m,\sigma \right) \) of the second type if, for any \(i\in \left\{ 1,\ldots ,m\right\} \), for any \(j>m\) and for any \(t\in \textbf{R}\), one has \(\varphi _{ij}(t)=0\).
-
3.
\(\varphi \) is called \(\left( m,\sigma \right) \) elementary if it is \(\left( m,\sigma \right) \)-standard and \(\left( m,\sigma \right) \) of the second type.
Definition 19
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function; for any \(\emptyset \ne L\subset \textbf{N}^{*}\) and for any \(\left\{ 1,\ldots ,m\right\} \subset N\subset \textbf{N}^{*}\), denote by \(\varphi ^{(L,N)}=\left( \varphi _{i}^{(L,N)}:i\in L\right) :\pi _{N}(U)\longrightarrow \textbf{R}^{L}\) the function given by
Furthermore, for any \(\emptyset \ne L\subset \textbf{N}^{*}\) and for any \(\emptyset \ne N\subset \textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} \), indicate by \(\varphi ^{(L,N)}=\left( \varphi _{i}^{(L,N)}:i\in L\right) :\pi _{N}(U)\longrightarrow \textbf{R}^{L}\) the function given by
In particular, suppose that \(m=1\), \(U^{(1)}=\textbf{R}\); then, indicate \(\varphi _{i}^{(\infty ,1)}\) by \(\varphi _{i1}\), for any \(i\in \textbf{N}^{*}\); moreover, for any \(\emptyset \ne L\subset \textbf{N}^{*}\) and for any \(\emptyset \ne N\subset \textbf{N}^{*}\), indicate by \(\varphi ^{(L,N)}:\pi _{N}(U)\longrightarrow \textbf{R}^{L}\) the function defined by formula (8).
Furthermore, for any \(l,o\in \textbf{N}^{*}\cup \left\{ \infty \right\} \), indicate \(\varphi ^{(I_{l},N)}\) by \(\varphi ^{(l,N)}\), \(\varphi ^{(L,I_{o})}\) by \(\varphi ^{(L,o)}\), and \(\varphi ^{(I_{l},I_{o})}\) by \(\varphi ^{(l,o)}\).
Moreover, for any \(\left\{ 1,\ldots ,m\right\} \subset L\subset \textbf{N}^{*}\) and for any \(\left\{ 1,\ldots ,m\right\} \subset N\subset \textbf{N}^{*}\), define the function \(\overline{\varphi }^{(L,N)}:U\longrightarrow \textbf{R}^{\infty }\) in the following manner:
Finally, for any \(l,o\in \left( \textbf{N}^{*}\backslash \left\{ 1,\ldots ,m-1\right\} \right) \cup \left\{ \infty \right\} \), indicate \(\overline{\varphi }^{(I_{l},N)}\) by \(\overline{\varphi }^{(l,N)}\), \(\overline{\varphi }^{(L,I_{o})}\) by \(\overline{\varphi }^{(L,o)}\), \(\overline{\varphi }^{(I_{l},I_{o})}\) by \(\overline{\varphi }^{(l,o)}\), and \(\overline{\varphi }^{(m,m)}\) by \(\overline{\varphi }\).
Remark 20
Let \(A=\left( a_{ij}\right) _{i,j\in \textbf{N}^{*}}:\ell _{\infty }\longrightarrow \ell _{\infty }\) be a linear function, let \(m\in \textbf{N}\) and let \(\sigma :\textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} \longrightarrow \textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} \) be an increasing function; then, A is \(\left( m,\sigma \right) \) if and only if:
-
1.
\(\forall \,j\in \left\{ 1,\ldots ,m\right\} \), one has \(\sum _{i=1}^{\infty }\left| a_{ij}\right| <\infty \).
-
2.
\(\forall \,i>m\), \(\forall \,j>m\) such that \(j\ne \sigma (i)\), one has \(a_{ij}=0\).
-
3.
There exists \(\prod _{i\in \mathcal {I}(A)}a_{ii}\in \textbf{R}^{*}\), where
$$\begin{aligned} \mathcal {I}(A)=\left\{ i>m:\sigma (h)=h, \, \forall \,h\ge i,\,\text {and}\,a_{ii}\ne 0\right\} . \end{aligned}$$(9)
Moreover, if the points 1, 2 and 3 hold, one has \(\sup _{i>m}\left| a_{i,\sigma (i)}\right| <\infty \). Finally, A is \(\left( m,\sigma \right) \) standard if and only if A is \(\left( m,\sigma \right) \) and \(a_{ij}=0\), for any \(i>m\), for any \(j\in \left\{ 1,\ldots ,m\right\} \), and A is \(\left( m,\sigma \right) \) of the second type if and only if A is \(\left( m,\sigma \right) \) and \(a_{ij}=0\), for any \(i\in \left\{ 1,\ldots ,m\right\} \), for any \(j>m\).
Proof
Since A is linear, formula (6) is alwais satisfied, with \(\varphi _{i}^{(\infty ,m)}(x_{1},\ldots ,x_{m})=\sum _{k=1}^{m} a_{ik}x_{k}\), \(\forall \,i\in \textbf{N}^{*}\), \(\forall \,(x_{1},\ldots ,x_{m})\in \textbf{R}^{m}\); analogously, the point 2 of Definition 17 is always satisfied, by Proposition 16. Moreover, it is easy to prove that the point 3 of Definition 17 is equivalent to the condition \(\sum _{i=1}^{\infty }\left| a_{ij}\right| <\infty \), \(\forall \,j\in \left\{ 1,\ldots ,m\right\} \), and the point 4 of Definition 17 is equivalent to the condition \(a_{ij}=0\), \(\forall \,i>m\), \(\forall \,j>m\), \(j\ne \sigma (i)\), and \(\sup _{i>m}\left| a_{i,\sigma (i)}\right| <\infty \); moreover, the point 5 of Definition 17 is equivalent to the condition that there exists \(\prod _{i\in \mathcal {I}(A)}a_{ii}\in \textbf{R}^{*}\), where \(\mathcal {I}(A)\) is defined by formula (9). Furthermore, if \(\sum _{i=1}^{\infty } \left| a_{ij}\right| <\infty \), \(\forall \,j\in \left\{ 1,\ldots ,m\right\} \), then \(\sup _{i\in \textbf{N}^{*}} \sum _{j=1}^{m}\left| a_{ij}\right| <\infty \), and so \(\varphi ^{(\infty ,m)}\left( \textbf{R}^{m}\right) \subset \ell _{\infty }\); then, from Remark 54, we have \(\sup _{i>m}\left| a_{i,\sigma (i)}x_{\sigma (i)}\right| <\infty \), \(\forall \,\left( x_{j}:j\in \sigma (\textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} )\right) \in \ell _{\sigma (\textbf{N} ^{*}\backslash \left\{ 1,\ldots ,m\right\} )}\), and so \(\sup _{i>m} \left| a_{i,\sigma (i)}\right| <\infty \); moreover, we have \(\lim _{i\rightarrow \infty }\varphi _{i}^{(\infty ,m)}\left( x_{1},\ldots ,x_{m} \right) =\lim {i\rightarrow \infty } \sum _{k=1}^{m} a_{ik}x_{k}=0\), \(\forall \,(x_{1},\ldots ,x_{m})\in \textbf{R}^{m}\). Thus, A is \(\left( m,\sigma \right) \), with \(U^{(m)}=\textbf{R}^{m}\), if and only if the points 1, 2 and 3 hold.
Finally, since \(\varphi _{i}^{(\infty ,m)}(x_{1},\ldots ,x_{m})=\sum _{k=1}^{m} a_{ik}x_{k}\), \(\forall \,i>m\), \(\forall \,(x_{1},\ldots ,x_{m})\in \textbf{R}^{m}\), and \(\varphi _{ij}(t)=a_{ij}t\), \(\forall \,i\in \left\{ 1,\ldots ,m\right\} \), \(\forall \,j>m\), \(\forall \,t\in \textbf{R}\), then A is \(\left( m,\sigma \right) \) standard if and only if A is \(\left( m,\sigma \right) \) and \(a_{ij}=0\), \(\forall \,i>m\), \(\forall \,j\in \left\{ 1,\ldots ,m\right\} \), and A is \(\left( m,\sigma \right) \) of the second type if and only if A is \(\left( m,\sigma \right) \) and \(a_{ij}=0\), \(\forall \,i\in \left\{ 1,\ldots ,m\right\} \), \(\forall \,j>m\). \(\square \)
The following definitions and results (from Proposition 21 to Definition 23) can be found in [7].
Proposition 21
Let \(A=\left( a_{ij}\right) _{i,j\in \textbf{N}^{*}}:\ell _{\infty }\longrightarrow \ell _{\infty }\) be a linear \(\left( m,\sigma \right) \) function; then, A is continuous.
Theorem 22
Let \(A=\left( a_{ij}\right) _{i,j\in \textbf{N}^{*}}:\ell _{\infty }\longrightarrow \ell _{\infty }\) be a linear \(\left( m,\sigma \right) \) function, and let \(A^{(n,n)}=\left( a_{ij}\right) _{i,j\in \left\{ 1,\ldots ,n\right\} }\), for any \(n\ge m\); then, the sequence \(\left\{ \det A^{(n,n)}\right\} _{n\ge m}\) converges to a real number.
Definition 23
Let \(A=\left( a_{ij}\right) _{i,j\in \textbf{N}^{*}}:\ell _{\infty }\longrightarrow \ell _{\infty }\) be a linear \(\left( m,\sigma \right) \) function; define the determinant of A, and call it \(\det A\), the real number
Example 24
Consider the linear function \(A=\left( a_{ij}\right) _{i,j\in \textbf{N}^{*}}:\ell _{\infty }\longrightarrow \ell _{\infty }\) given by
observe that A is expressed by the following infinite-dimensional matrix:
then, from Remark 20, A is \(\left( 2,\sigma \right) \), where \(\sigma (i)=i\), \(a_{ii}=2^{2^{-i}}\), \(\forall \,i>2\), and so \(\mathcal {I} (A)=\textbf{N}^{*}\backslash \left\{ 1,2\right\} \); moreover, \(\forall \,n>1\), we have
In order to calculate \(\det A\), observe that, \(\forall \,n>1\), we have \(A^{\left( \{2\},n\right) }=u_{n}+v_{n}\), where \(u_{n}=A^{\left( 1,n\right) }=\left( 2^{-j}:j\in \left\{ 1,\ldots ,n\right\} \right) \in \textbf{R}^{n}\), and \(v_{n}=\left( 1,0,\ldots ,0\right) \in \textbf{R}^{n}\). Then, we have \(\det A^{(n,n)}=\det U^{(n,n)}+\det V^{(n,n)}\), where \(U^{(n,n)}=\left( u_{ij}^{(n)}\right) _{i,j\in \textbf{N}^{*}}\) and \(V^{(n,n)}=\left( v_{ij}^{(n,n)}\right) _{i,j\in \textbf{N}^{*}}\) are the linear functions obtained by substituting the second row of \(A^{(n,n)}\) by \(u_{n}\) and \(v_{n}\), respectively; moreover, since \(\left( U^{(n,n)}\right) ^{\left( 1,n\right) }=\left( U^{(n,n)}\right) ^{\left( \{2\},n\right) } \), we have \(\det U^{(n,n)}=0\), from which
Finally, \(\forall \,n>2\), we have
since the second row of \(V^{\left( n-1,\{2,\ldots ,n\}\right) }\) is zero, and so \(\det V^{\left( n-1,\{2,\ldots ,n\}\right) }=0\). Then, by recursion, from (11) we obtain
and so formula (10) implies
Lemma 25
Let \(k\in \textbf{N}\) and let \(a=a_{i}:i>k\in \left( -1,+\infty \right) ^{\infty }\) such that \(\sum _{i=k+1}{\infty }\left| a_{i}\right| \in \textbf{R}\); then, one has \(\prod _{i=k+1}^{\infty }\left( 1+a_{i}\right) \in \textbf{R}^{+}\).
Proof
Since \(\sum _{i=k+1}^{\infty }\left| a_{i}\right| \in \textbf{R}\), we have \(\lim _{i\rightarrow \infty } a_{i}=0\); then, since \(\lim _{t\rightarrow 0}\frac{\ln (1+t)}{t}=1\), there exists \(i_{0}>k\) such that, \(\forall \,i\in \textbf{N}\), \(i\ge i_{0}\), we have \(\frac{\ln (1+\left| a_{i}\right| )}{\left| a_{i}\right| }\le 2\), \(\frac{\ln (1-\left| a_{i}\right| )}{-\left| a_{i}\right| }\le 2\), from which
then, we have \(\sum _{i=i_{0}}^{\infty }\left| \ln \left( 1+a_{i}\right) \right| \le 2\sum _{i=i_{0}}^{\infty }\left| a_{i}\right| <\infty \), from which \(\sum _{i=i_{0}}^{\infty }\ln \left( 1+a_{i}\right) \in \textbf{R}\), and so
\(\square \)
Example 26
Consider the function \(\varphi :\ell _{\infty }\longrightarrow \ell _{\infty }\) given by
moreover, \(\forall \,i>1\), let \(g_{i}:\mathbf {R\longrightarrow R}\) be the function defined by \(g_{i}(t)=\left( 1+\frac{1}{i^{2}}\right) t+\frac{\sin (t)}{i^{2}}\), \(\forall \,t\in \textbf{R}\); we have \(\left| g_{i}^{\prime }(t)\right| =g_{i}^{\prime }(t)=1+\frac{1}{i^{2}}+\frac{\cos (t)}{i^{2}}\ge 1\), \(\forall \,i>1\), \(\forall \,t\in \textbf{R}\); moreover, \(\forall \,x=(x_{j}:j>1)\in \textbf{R}^{\infty }\), we obtain
thus, since \(\frac{1}{i^{2}}+\frac{\cos (x_{i})}{i^{2}}\ge 0>-1\), \(\forall \,i>1\), and \(\sum _{i=2}^{\infty }\left( \frac{1}{i^{2}}+\frac{\cos (x_{i})}{i^{2}}\right) \in \textbf{R}\), from formula (12) and Lemma 25, we have \(\prod _{i=2}^{\infty }g_{i}^{\prime }\left( x_{i}\right) \in \textbf{R}^{+}\subset \textbf{R} ^{*}\).
Then, \(\varphi \) is \(\left( 1,\sigma \right) \), where \(\varphi _{1} ^{(\infty ,1)}(x_{1})=\frac{\sin (x_{1})}{2}\), \(\varphi _{i}^{(\infty ,1)} (x_{1})=2^{-i}\cos (x_{1})\), \(\varphi _{1j}(t)=2^{-j}\sin (t)\), \(\sigma (i)=i\), \(\varphi _{ii}(t)=\left( 1+\frac{1}{i^{2}}\right) t+\frac{\sin (t)}{i^{2}}\), \(\forall \,i,j>1\), \(\forall \,t\in \textbf{R}\), and so \(\mathcal {I} (\varphi )=\textbf{N}^{*}\backslash \left\{ 1\right\} \).
4 Gaussian change of variables’ formula
The following two definitions (Definition 27 and Definition 28) can be found, for example, in [11].
Definition 27
Let E be a vector space on \(\textbf{R}\) and let T be a topology on E; a probability measure \(\mu \) on \(\left( E,\sigma (T)\right) \) is called Gaussian measure if, for any function \(g:\left( E,T\right) \longrightarrow \left( \textbf{R},\tau \right) \) linear and continuous, the probability measure \(\mu \circ g^{-1}\) on \(\left( \textbf{R},\mathcal {B} \right) \) is Gaussian; moreover, a random variable \(X:\left( \Omega ,\mathcal {F},P\right) \longrightarrow \left( E,\sigma (T)\right) \) is called Gaussian random variable on E if its law is a Gaussian measure.
Definition 28
Let E be a Banach space, with the topology T induced by the norm of E, and let \(\mu \) be a Gaussian measure on \(\left( E,\sigma (T)\right) \); define
The vector space \(H(\mu )\) on \(\textbf{R}\) (that can be indicated simply by H) is called Cameron–Martin space associated to \(\left( E,\mu \right) \). In particular, if E is separable, the pair \(\left( E,\mu \right) \) (indicated by \(\left( E,H,\mu \right) \) too) is called abstract Wiener space.
Remark 29
Let \(a=\left( a_{i}:i\in \textbf{N}^{*}\right) \in \textbf{R}^{\infty }\), let \(A=\left( a_{ij}\right) _{i,j\in \textbf{N}^{*} }:\textbf{R}^{\infty }\longrightarrow \textbf{R}^{\infty }\) be a linear and continuous function, and let \(X=\left( X_{i}:i\in \textbf{N}^{*}\right) :\left( \Omega ,\mathcal {F},P\right) \longrightarrow \left( \textbf{R} ^{\infty },\mathcal {B}^{\infty }\right) \) be a Gaussian random variable; then, the random variable \(AX+a:\left( \Omega ,\mathcal {F},P\right) \longrightarrow \left( \textbf{R}^{\infty },\mathcal {B}^{\infty }\right) \) is Gaussian.
Proof
\(\forall \,g\in E^{*}\), we have \(g(AX+a)=(g\circ A)(X)+g(a)\); moreover, since \((g\circ A)\in E^{*}\), the random variable \((g\circ A)(X):\left( \Omega ,\mathcal {F},P\right) \longrightarrow \left( \textbf{R},\mathcal {B} \right) \) is Gaussian, and so the random variable \(g(AX+a):\left( \Omega ,\mathcal {F},P\right) \longrightarrow \left( \textbf{R},\mathcal {B} \right) \) is Gaussian too. \(\square \)
Definition 30
Let \(m\in \textbf{N}^{*}\) and let \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*}}:\textbf{R}^{\infty }\longrightarrow \textbf{R}^{\infty }\) be a linear and symmetric function, such that \(\gamma _{ij}=0\), for any \((i,j)\notin \left\{ 1,\ldots ,m\right\} ^{2} \bigcup \left( \bigcup _{k=m+1}^{\infty }\left\{ (k,k)\right\} \right) \); we say that:
-
1.
\(\Gamma \) is m-positive semidefinite if the matrix \(\Gamma ^{(m,m)}\) is symmetric positive semidefinite, \(\gamma _{ii}\in \left[ 0,\infty \right) \), for any \(i>m\).
-
2.
\(\Gamma \) is m-positive definite if the matrix \(\Gamma ^{(m,m)}\) is symmetric positive definite, \(\gamma _{ii}\in \textbf{R}^{+}\), for any \(i>m\).
Definition 31
Let \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*}}:\textbf{R}^{\infty }\longrightarrow \textbf{R}^{\infty }\) be a linear and diagonal function; we say that:
-
1.
\(\Gamma \) is 0-positive semidefinite if \(\gamma _{ii}\in \left[ 0,\infty \right) \), for any \(i\in \textbf{N}^{*}\).
-
2.
\(\Gamma \) is 0-positive definite if \(\gamma _{ii}\in \textbf{R}^{+}\), for any \(i\in \textbf{N}^{*}\).
Definition 32
Let \(m\in \textbf{N}\), let \(a=\left( a_{i}:i\in \textbf{N}^{*}\right) \in \textbf{R}^{\infty }\) and let \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*}}:\textbf{R}^{\infty }\longrightarrow \textbf{R} ^{\infty }\) be a linear, symmetric and m-positive definite function; define the probability measure \(N(a,\Gamma )\) on \(\left( \textbf{R}^{\infty },\mathcal {B}^{\infty }\right) \) by
Moreover, for any \(A\subset \textbf{R}^{\infty }\) such that \(\pi _{i} (A)=\textbf{R}\) for any \(i\in \textbf{N}^{*}\), indicate by \(N(a,\Gamma )\) the measure \(\left. N(a,\Gamma )\right| _{\mathcal {B}^{\infty }(A)}\) too.
Remark 33
Let \(m\in \textbf{N}\), let \(a=\left( a_{i}:i\in \textbf{N}^{*}\right) \in \textbf{R}^{\infty }\), let \(b=\left( b_{i}:i\in \textbf{N}^{*}\right) \in \textbf{R}^{\infty }\), let \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*}}:\textbf{R}^{\infty }\longrightarrow \textbf{R} ^{\infty }\) be a symmetric m-positive definite linear function, and let \(X=\left( X_{i}:i\in \textbf{N}^{*}\right) :\left( \Omega ,\mathcal {F},P\right) \longrightarrow \left( \textbf{R}^{\infty },\mathcal {B}^{\infty }\right) \) be a random variable with law \(N(a,\Gamma )\); then, the law of the random variable \(X+b\) is \(N(a+b,\Gamma )\).
Proof
Let \(\mathcal {A}=\left\{ \prod _{i=1}^{\infty } A_{i}:A_{i}\in \mathcal {B}, \forall \,i\in \textbf{N}^{*}\right\} \subset \mathcal {B}^{\infty }\); \(\forall \,A= \prod _{i=1}^{\infty } A_{i}\in \mathcal {A}\), we have
then, the law \(\mathcal {L}_{X+b}\) of \(X+b\) and \(N(a+b,\Gamma )\) coincide on \(\mathcal {A}\); moreover, we have \(\mathcal {L}_{X+b}(\textbf{R}^{\infty })=N(a+b,\Gamma )\left( \textbf{R}^{\infty }\right) =1<\infty \), \(\textbf{R} ^{\infty }\in \mathcal {I}\), and so \(\mathcal {L}_{X+b}\) and \(N(a+b,\Gamma )\) are \(\sigma \)- finite on \(\mathcal {A}\). Then, since \(\mathcal {A}\) is a \(\pi \)-system on \(\textbf{R}^{\infty }\) such that \(\sigma (\mathcal {A})=\mathcal {B}^{\infty }\), from Proposition 45, \(\mathcal {L}_{X+b}\) and \(N(a+b,\Gamma )\) coincide on \(\mathcal {B}^{\infty }\). \(\square \)
Lemma 34
\(\forall \,x\in \textbf{R}^{+}\), one has
Proof
See for example, Lemma 1.1.3 in [11]. \(\square \)
Theorem 35
Let \(\left\{ X_{i}\right\} _{i\in \textbf{N}^{*}}\) be a sequence of independent random variables such that \(E\left[ X_{i}\right] =0\), for any \(i\in \textbf{N}^{*}\), and \(\sum _{i=1}^{\infty } Var\left[ X_{i}\right] \in \left[ 0,\infty \right) \); then, there exists a real random variable Z such that \(\sum _{i=1}^{n}X_{i}\overset{a.\,s.}{\underset{n\rightarrow \infty }{\longrightarrow }}Z\).
Proof
See, for example, page 112 in the Williams’ book [26]. \(\square \)
Proposition 36
Let \(m\in \textbf{N}\), let \(a=\left( a_{i}:i\in \textbf{N}^{*}\right) \in \textbf{R}^{\infty }\) and let \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*}}:\textbf{R}^{\infty }\longrightarrow \textbf{R} ^{\infty }\) be a symmetric m-positive definite linear function, such that \(\sum _{i=1}^{\infty }\gamma _{ii}\in \textbf{R}^{+}\); then:
-
1.
\(\forall \,c\in \textbf{R}^{+}\), one has \(N(a,\Gamma )\left( \prod _{i=1}^{\infty }\left[ a_{i}-c,a_{i}+c\right] \right) \in \left( 0,1\right) \).
-
2.
If \(a\in E_{\infty }\), one has \(N(a,\Gamma )(E_{\infty })=N(a,\Gamma )(\ell _{\infty })=1\); in particular, \(N(a,\Gamma )\) is a probability measure on \(\left( E_{\infty },\mathcal {B}_{\infty }\right) \).
-
3.
If \(a\in \ell _{2}\), one has \(N(a,\Gamma )(\ell _{2})=1\).
-
4.
If \(a\in \ell _{1}\) and \(\sum _{i=1}^{\infty } \sqrt{\gamma _{ii}}\in \textbf{R}^{+}\), one has \(N(a,\Gamma )(\ell _{1})=1\).
Proof
We have
then, we can prove this proposition by supposing that \(\Gamma \) is 0-positive definite.
-
1.
\(\forall \,c\in \textbf{R}^{+}\), we have
$$\begin{aligned} N(0,\Gamma )\left( \left[ -c,c\right] ^{\infty }\right)\le & {} N(0,\Gamma )\left( \left[ -c,c\right] \times \textbf{R}^{\infty }\right) \\= & {} N(0,\gamma _{11})\left( \left[ -c,c\right] \right) <1; \end{aligned}$$moreover
$$\begin{aligned} N(a,\Gamma )\left( \left[ -c,c\right] ^{\infty }\right) =\underset{i=1}{\overset{\infty }{\prod }}\left( \overset{c}{\underset{-c}{\int }}\frac{1}{\sqrt{2\pi \gamma _{ii}}}\exp \left( -\frac{x^{2}}{2\gamma _{ii}}\right) dx\right) =\underset{i=1}{\overset{\infty }{\prod }}A_{i}, \end{aligned}$$(14)where
$$\begin{aligned} A_{i}=A_{i}(c)\equiv \overset{\frac{c}{\sqrt{\gamma _{ii}}}}{\underset{-\frac{c}{\sqrt{\gamma _{ii}}}}{\int }}\frac{1}{\sqrt{2\pi }}\exp \left( -\frac{t^{2} }{2}\right) dt\in \left( 0,1\right) ,\quad \forall \,i\in \textbf{N}^{*}; \end{aligned}$$(15)furthermore
$$\begin{aligned} A_{i}= & {} 2\overset{\frac{c}{\sqrt{\gamma _{ii}}}}{\underset{0}{\int }}\frac{1}{\sqrt{2\pi }}\exp \left( -\frac{t^{2}}{2}\right) dt=2\left( \frac{1}{2}-\overset{\infty }{\underset{\frac{c}{\sqrt{\gamma _{ii}}}}{\int }}\frac{1}{\sqrt{2\pi }}\exp \left( -\frac{t^{2}}{2}\right) dt\right) \nonumber \\= & {} 1-\sqrt{\frac{2}{\pi }}\overset{\infty }{\underset{\frac{c}{\sqrt{\gamma _{ii}} }}{\int }}\exp \left( -\frac{t^{2}}{2}\right) dt\ge 1\nonumber \\{} & {} -\sqrt{\frac{2\gamma _{ii}}{\pi c^{2}}}\exp \left( -\frac{c^{2}}{2\gamma _{ii}}\right) \quad \text {(from Lemma 34)}. \end{aligned}$$(16)Moreover, \(\forall \,x\in \textbf{R}^{+}\), we have \(\sqrt{x}\exp (-x)\le \frac{1}{\sqrt{2e}}\); then, since \(\lim _{i\rightarrow \infty }\gamma _{ii}=0\), there exists \(i_{0}=i_{0}(c)\in \textbf{N}^{*}\) such that, \(\forall \,i\ge i_{0}\), we have \(\gamma _{ii}\le c^{2}\), \(\sqrt{\frac{c^{2} }{2\gamma _{ii}}}\exp \left( -\frac{c^{2}}{2\gamma _{ii}}\right) \le \frac{1}{\sqrt{2e}}\), from which
$$\begin{aligned} \sqrt{\frac{2\gamma _{ii}}{\pi c^{2}}}\exp \left( -\frac{c^{2}}{2\gamma _{ii} }\right) \le \sqrt{\frac{2\gamma _{ii}}{\pi c^{2}}}\cdot \sqrt{\frac{\gamma _{ii}}{ec^{2}}}=\sqrt{\frac{2}{\pi e}}\cdot \frac{\gamma _{ii}}{c^{2} }\equiv x_{i}, \end{aligned}$$(17)where \(x_{i}=x_{i}(c)\in \left( 0,\sqrt{\frac{2}{\pi e}}\right] \subset \left( 0,1\right) \); then, formulas (14), (16) and (17) imply
$$\begin{aligned} N(0,\Gamma )\left( \left[ -c,c\right] ^{\infty }\right) \ge \left( \underset{i=1}{\overset{i_{0}-1}{\prod }}A_{i}\right) \underset{i=i_{0} }{\overset{\infty }{\prod }}\left( 1-x_{i}\right) ; \end{aligned}$$(18)finally, observe that \(\sum _{i=i_{0}}^{\infty }\left( -x_{i}\right) =-\frac{1}{c^{2}}\sqrt{\frac{2}{\pi e}} \sum _{i=i_{0}}^{\infty }\gamma _{ii}\in \textbf{R}\); then, from Lemma 25, we have \(\prod _{i=i_{0}}^{\infty }\left( 1-x_{i}\right) \in \textbf{R}^{+}\), and so formulas (15) and (18) imply \(N(0,\Gamma )\left( \left[ -c,c\right] ^{\infty }\right) >0\), from which
$$\begin{aligned} N(a,\Gamma )\left( \overset{\infty }{\underset{i=1}{\prod }}\left[ a_{i}-c,a_{i}+c\right] \right) =N(0,\Gamma )\left( \left[ -c,c\right] ^{\infty }\right) \in \left( 0,1\right) . \end{aligned}$$ -
2.
Suppose \(a\in E_{\infty }\), let \(a_{\infty }=\lim _{i\rightarrow \infty }a_{i}\in \textbf{R}\) and let \(\left\{ X_{i}\right\} _{i\in \textbf{N}^{*}}\) be a sequence of independent random variables such that \(X_{i}\sim N\left( a_{i},\gamma _{ii}\right) \), \(\forall \,i\in \textbf{N} ^{*}\); we have \(E\left[ X_{i}-a_{i}\right] =0\), \(Var\left[ X_{i} -a_{i}\right] =\gamma _{ii}\), \(\forall \,i\in \textbf{N}^{*}\), and \(\sum _{i=1}^{\infty }\gamma _{ii}\in \textbf{R}^{+}\); then, from Theorem 35, there exists a real random variable \(T_{1}\) such that \(\sum _{i=1}^{n}\left( X_{i}-a_{i}\right) \overset{a.\,s.}{\underset{n\rightarrow \infty }{\longrightarrow }}T_{1}\), from which \(\left( X_{i}-a_{i}\right) \overset{a.\,s.}{\underset{i\rightarrow \infty }{\longrightarrow }}0\), and so \(P\left( \lim _{i\rightarrow \infty }X_{i}=a_{\infty }\right) =1\); thus, since \(\left\{ \lim _{i\rightarrow \infty }X_{i}=a_{\infty }\right\} \subset \left\{ \left( X_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\right\} \), we obtain
$$\begin{aligned} P\left( \left( X_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\right) =1, \end{aligned}$$that is \(N(a,\Gamma )(E_{\infty })=1\); moreover, since \(E_{\infty }\subset \ell _{\infty }\), we have \(N(0,\Gamma )(\ell _{\infty })=1\).
-
3.
Suppose \(a\in \ell _{2}\); since \(\frac{X_{i}-a_{i}}{\sqrt{\gamma _{ii}} }\sim N(0,1)\), it is easy to prove that \(\left( X_{i}-a_{i}\right) ^{2}=\gamma _{ii}\left( \frac{X_{i}-a_{i}}{\sqrt{\gamma _{ii}}}\right) ^{2}\sim \Gamma \left( \frac{1}{2},\frac{1}{2\gamma _{ii}}\right) \) (the Gamma law with parameters \(\frac{1}{2}\), \(\frac{1}{2\gamma _{ii}}\)), \(\forall \,i\in \textbf{N}^{*}\), and so \(E\left[ \left( X_{i}-a_{i}\right) ^{2}-\gamma _{ii}\right] =0\), \(Var\left[ \left( X_{i}-a_{i}\right) ^{2}-\gamma _{ii}\right] =2\gamma _{ii}^{2}\), \(\sum _{i=1}^{\infty } 2\gamma _{ii}^{2}\in \textbf{R}^{+}\); then, from Theorem 35, there exists a real random variable \(T_{2}\) such that \(\sum _{i=1}^{n}\left( \left( X_{i}-a_{i}\right) ^{2}-\gamma _{ii}\right) \overset{a.\,s.}{\underset{n\rightarrow \infty }{\longrightarrow }}T_{2}\), from which \(\sum _{i=1}^{n} \left( X_{i}-a_{i}\right) ^{2}\overset{a.\,s.}{\underset{n\rightarrow \infty }{\longrightarrow }}T_{2}+ \sum _{i=1}^{\infty }\gamma _{ii}\), and so \(P\left( \sum _{i=1}^{\infty } \left( X_{i}-a_{i}\right) ^{2}\in \textbf{R}\right) =1\). Moreover, we have \(a_{i}X_{i}\sim N\left( a_{i}^{2},a_{i}^{2}\gamma _{ii}\right) \), \(\forall \,i\in \textbf{N}^{*}\), and so \(E\left[ a_{i}X_{i}-a_{i} ^{2}\right] =0\), \(Var\left[ a_{i}X_{i}-a_{i}^{2}\right] =a_{i}^{2} \gamma _{ii}\), \(\sum _{i=1}^{\infty } a_{i}^{2}\gamma _{ii} \in \left[ 0,\infty \right) \); then, from Theorem 35, there exists a real random variable \(T_{3}\) such that \(\sum _{i=1}^{n}\left( a_{i}X_{i}-a_{i}^{2}\right) \overset{a.\,s.}{\underset{n\rightarrow \infty }{\longrightarrow }}T_{3}\), from which \(\sum _{i=1}^{n} a_{i}X_{i}\overset{a.\,s.}{\underset{n\rightarrow \infty }{\longrightarrow }}T_{3}+\sum _{i=1}^{\infty } a_{i}^{2}\), and so \(P\left( \sum _{i=1}^{\infty } a_{i}X_{i}\in \textbf{R}\right) =1\); furthermore, \(\forall \,n\in \textbf{N}^{*}\), we have
$$\begin{aligned} \underset{i=1}{\overset{n}{\sum }}X_{i}^{2}=\underset{i=1}{\overset{n}{\sum } }\left( X_{i}-a_{i}\right) ^{2}+2\underset{i=1}{\overset{n}{\sum }}a_{i} X_{i}-\underset{i=1}{\overset{n}{\sum }}a_{i}^{2}, \end{aligned}$$from which \(P\left( \sum _{i=1}^{\infty } X_{i}^{2} \in \textbf{R}\right) =1\), that is \(N(a,\Gamma )(\ell _{2})=1\).
-
4.
Suppose \(a\in \ell _{1}\) and \(\sum _{i=1}^{\infty } \sqrt{\gamma _{ii}}\in \textbf{R}^{+}\); \(\forall \,i\in \textbf{N}^{*}\), we have
$$\begin{aligned} E\left[ \left| X_{i}-a_{i}\right| \right]= & {} 2\underset{a_{i}}{\overset{\infty }{\int }}\frac{x-a_{i}}{\sqrt{2\pi \gamma _{ii}}}\exp \left( -\frac{\left( x-a_{i}\right) ^{2}}{2\gamma _{ii}}\right) dx\\= & {} 2\underset{0}{\overset{\infty }{\int }}\sqrt{\frac{\gamma _{ii}}{2\pi }} \exp \left( -t\right) dt=\sqrt{\frac{2\gamma _{ii}}{\pi }}\\{} & {} \Longrightarrow E\left[ \left| X_{i}-a_{i}\right| -\sqrt{\frac{2\gamma _{ii}}{\pi }}\right] =0,\\{} & {} \quad Var\left[ \left| X_{i}-a_{i}\right| -\sqrt{\frac{2\gamma _{ii}}{\pi } }\right] =Var\left[ \left| X_{i}-a_{i}\right| \right] \\= & {} E\left[ \left( X_{i}-a_{i}\right) ^{2}\right] -\left( E\left[ \left| X_{i}-a_{i}\right| \right] \right) ^{2}\\= & {} \gamma _{ii}-\frac{2\gamma _{ii}}{\pi }=\left( \frac{\pi -2}{\pi }\right) \gamma _{ii}\\{} & {} \Longrightarrow \underset{i=1}{\overset{\infty }{\sum }}Var\left[ \left| X_{i}-a_{i}\right| -\sqrt{\frac{2\gamma _{ii}}{\pi }}\right] \\= & {} \left( \frac{\pi -2}{\pi }\right) \underset{i=1}{\overset{\infty }{\sum }}\gamma _{ii} \in \textbf{R}^{+}; \end{aligned}$$then, from Theorem 35, there exists a real random variable \(T_{4}\) such that \(\sum _{i=1}^{n}\left( \left| X_{i}-a_{i}\right| -\sqrt{\frac{2\gamma _{ii}}{\pi }}\right) \overset{a.\,s.}{\underset{n\rightarrow \infty }{\longrightarrow }}T_{4}\), from which \(\sum _{i=1}^{n}\left| X_{i}-a_{i}\right| \overset{a.\,s.}{\underset{n\rightarrow \infty }{\longrightarrow }}T_{4}+ \sum _{i=1}^{\infty }\sqrt{\frac{2\gamma _{ii}}{\pi }}\), and so \(P\left( \sum _{i=1}^{\infty }\left| X_{i}-a_{i}\right| \in \textbf{R}\right) =1\); furthermore, we have
$$\begin{aligned} \underset{i=1}{\overset{\infty }{\sum }}\left| X_{i}\right| \le \underset{i=1}{\overset{\infty }{\sum }}\left| X_{i}-a_{i}\right| +\underset{i=1}{\overset{\infty }{\sum }}\left| a_{i}\right| , \end{aligned}$$from which \(P\left( \sum _{i=1}^{\infty }\left| X_{i}\right| \in \textbf{R}\right) =1\), that is \(N(a,\Gamma )(\ell _{1} )=1\).
\(\square \)
Proposition 37
Let \(m\in \textbf{N}\), let \(a=\left( a_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\) and let \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*}}:\textbf{R}^{\infty }\longrightarrow \textbf{R} ^{\infty }\) be a symmetric m-positive definite linear function, such that \(\sum _{i=m+1}^{\infty }\gamma _{ii}\in \textbf{R}^{+}\); then, the probability measure \(N(a,\Gamma )\) on \(\left( E_{\infty },\mathcal {B} _{\infty }\right) \) is Gaussian, and one has \(H\left( N(a,\Gamma )\right) \supset \ell _{2}(\Gamma )\), where
Proof
Let \(X=\left( X_{i}:i\in \textbf{N}^{*}\right) :\left( \Omega ,\mathcal {F},P\right) \longrightarrow \left( E_{\infty },\mathcal {B}_{\infty }\right) \) be a random variable with law \(N(a,\Gamma )\); we can consider X as a random variable on \(\left( \textbf{R}^{\infty },\mathcal {B}^{\infty }\right) \); thus, we have \(X_{i}\sim N\left( a_{i},\gamma _{ii}\right) \), \(\forall \,i\in \textbf{N}^{*}\); morerover, the random variables \(\left( X_{1},\ldots ,X_{m}\right) \) and \(\left( X_{i}:i>m\right) \) are independent and such that \(\left( X_{1},\ldots ,X_{m}\right) \sim N\left( (a_{1},\ldots ,a_{m} ),\Gamma ^{(m,m)}\right) \); finally, the random variables \(\left\{ X_{i}\right\} _{i>m}\) are independent. Moreover, let \(g:\left( E_{\infty },\tau _{\left\| \cdot \right\| _{\infty }}\left( E_{\infty }\right) \right) \longrightarrow \left( \textbf{R},\tau \right) \) be a linear and continuous function; from Corollary 13, there exists \(\left( c_{i}:i\in \textbf{N}^{*}\cup \left\{ \infty \right\} \right) \in \textbf{R} ^{\textbf{N}^{*}\cup \left\{ \infty \right\} }\) such that \(\sum _{i\in \textbf{N}^{*}\cup \left\{ \infty \right\} }\left| c_{i}\right| \in \textbf{R}\) and such that \(g(x)=\sum _{i\in \textbf{N} ^{*}\cup \left\{ \infty \right\} }c_{i}x_{i}\), \(\forall \,x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\), where \(x_{\infty } \equiv \lim _{i\rightarrow \infty }x_{i}\), and so \(g(X)=\sum _{i=1}^{\infty } c_{i}X_{i}+c_{\infty } \lim _{i\rightarrow \infty }X_{i}\); furthermore, by setting \(a_{\infty }=\lim _{i\rightarrow \infty }a_{i}\in \textbf{R}\), from Theorem 35, there exists a real random variable T such that \(\sum _{i=m+1}^{n}\left( X_{i}-a_{i}\right) \overset{a.\,s.}{\underset{n\rightarrow \infty }{\longrightarrow }}T\), from which \(\left( X_{i} -a_{i}\right) \overset{a.\,s.}{\underset{i\rightarrow \infty }{\longrightarrow }}0\), and so \(X_{i}\overset{a.\,s.}{\underset{i\rightarrow \infty }{\longrightarrow }}a_{\infty }\); moreover, observe that \(\sum _{i=1}^{\infty } c_{i}X_{i}=\lim _{n\longrightarrow \infty }\sum _{i=1}^{n} c_{i}X_{i}\) and \(\sum _{i=1}^{n} c_{i}X_{i}\sim N\left( \sum _{i=1}^{n} c_{i}a_{i},\left\langle \Gamma ^{(m,m)}\left( c_{1},\ldots ,c_{m} \right) ,\left( c_{1},\ldots ,c_{m}\right) \right\rangle + \sum _{i=m+1}^{n} c_{i}^{2}\gamma _{ii}\right) \), \(\forall \,n>m\); then, we have
and so \(N(a,\Gamma )\) is Gaussian on \(\left( E_{\infty },\mathcal {B}_{\infty }\right) \). Finally, let \(x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in \ell _{2}(\Gamma )\), let \(\lambda \in \textbf{R}^{+}\) be the smaller of the eigenvalues of \(\Gamma ^{(m,m)}\), and let
we have \(x_{\infty }=0\), \(\sum _{i=1}^{\infty } \frac{x_{i} ^{2}}{\alpha _{i}}\in \textbf{R}\) and \(\left| x\right| _{H\left( N(a,\Gamma )\right) }=\sup \left\{ \left| \sum _{i=1}^{\infty } c_{i}x_{i}\right| :c\in \mathcal {A}\right\} \), where
moreover, \(\forall \,c=\left( c_{i}:i\in \textbf{N}^{*}\right) \in \mathcal {A}\), we have
then, since \(\left\langle \Gamma ^{(m,m)}\left( c_{1},\ldots ,c_{m}\right) ,\left( c_{1},\ldots ,c_{m}\right) \right\rangle + \sum _{i=m+1}^{\infty } c_{i}^{2}\gamma _{ii}\le 1\), we obtain \(\left| \sum _{i=1}^{\infty } c_{i}x_{i}\right| \le \sqrt{\sum _{i=1}^{\infty }\frac{x_{i}^{2}}{\alpha _{i}}}<\infty \), from which \(\left| x\right| _{H\left( N(a,\Gamma )\right) }<\infty \), and so \(x\in H\left( N(a,\Gamma )\right) \). \(\square \)
Proposition 38
Let \(\varphi :U=U^{(m)}\times E_{\infty }\longrightarrow E_{\infty }\) be a bijective and continuous \(\left( m,\sigma \right) \) function, such that \(\overline{\varphi }\) is bijective and continuous, \(\varphi ^{(m,m)}\) is open, \(\varphi ^{(n,n)}\) is continuous, for any \(n\in \textbf{N}^{*}\), \(n\ge m\), and \(\varphi _{ij}\) is continuous, for any \(i\in \left\{ 1,\ldots ,m\right\} \), for any \(j>m\); moreover, suppose that there exists \(\varepsilon \in \left[ 0,\infty \right) \) such that \(\left| \varphi _{i}^{(\infty ,m)}\left( x_{1},\ldots ,x_{m}\right) \right| \le \varepsilon \), for any \(i>m\), for any \(\left( x_{1},\ldots ,x_{m}\right) \in U^{(m)}\), and the sequence \(\left\{ \left( \overline{\varphi }^{(n,n)}\right) ^{-1}\right\} _{n>m}\) converges uniformly to \(\varphi ^{-1}\) over the closed and bounded subsets of \(E_{\infty }\); then, for any \(M\in \textbf{R}^{+}\), there exists \(O\in \textbf{R}^{+}\) and there exists a closed subset K of \(U\cap [-O,O]^{\infty }\) such that, for any \(n\in \textbf{N}^{*}\), \(n\ge m\), one has
Proof
Since \(\overline{\varphi }\) is bijective, from Proposition 65, \(\varphi ^{(m,m)}\) and \(\sigma \) are bijective; furthermore, \(\forall \,i>m\), from Remark 55, we have \(\sigma (i)=i\), and, from Proposition 65, \(\varphi _{ii}:\textbf{R}\longrightarrow \textbf{R}\) is bijective. Let \(M\in \textbf{R}^{+}\), \(x_{0}=(x_{0,j}:j\in \textbf{N}^{*})\in U\) and \(x=\left( x_{j}:j\in \textbf{N}^{*}\right) \in \varphi ^{-1}\left( \left[ -M,M\right] ^{\infty }\cap E_{\infty }\right) \subset U\); \(\forall \,i\in \left\{ 1,\ldots ,m\right\} \), we have
and so
where
furthermore, \(\forall \,i>m\), we have
where
moreover
then, since the function \(\varphi _{ii}^{-1}\) is derivable on \(\textbf{R}\), the Lagrange theorem implies that, for some \(\xi _{i}\in \left( \rho _{i},\tau _{i}\right) \), where
we have
thus, from (23), we obtain
Moreover, from Definition 17, we have
moreover, from Remark 54, we have \(\sup _{i>m}\left| \varphi _{ii}(x_{0,i})\right| <\infty \), and so formula (25) implies
analogously, we obtain
then, from formula (22), \(\forall \,i>m\), we have \(\left| x_{i}\right| \le O_{1}\), from which \(\left\| (x_{j}:j>m)\right\| _{\infty }\le O_{1}\).
Moreover, \(\forall \,i\in \left\{ 1,\ldots ,m\right\} \), \(\forall \,j>m\), the function \(\varphi _{ij}\) is continuous, and so there exists \(t_{ij} =t_{ij}\left( \varphi ,M,\varepsilon ,x_{0}\right) \in \left[ -O_{1},O_{1}\right] \) such that \(\max _{t\in \left[ -O_{1},O_{1}\right] }\left| \varphi _{ij}(t)\right| =\left| \varphi _{ij}(t_{ij})\right| \); then, we have
and so \(\left\| \left( w_{1},\ldots ,w_{m}\right) \right\| _{I_{m}}\le M+O_{2}\equiv O_{3}\in \textbf{R}^{+}\), from (21); then, since the function \(\left( \varphi ^{(m,m)}\right) ^{-1}\) is continuous, from (20), we have \(\left\| \left( x_{1},\ldots , x_{m}\right) \right\| _{I_{m}}\le O_{4}\), for some \(O_{4}=O_{4}\left( \varphi ,M,\varepsilon ,x_{0}\right) \in \textbf{R}^{+}\) such that
and so \(\left\| x\right\| _{\infty }\le \max \left\{ O_{1},O_{4}\right\} \equiv O\); thus
Furthermore, let \(n\in \textbf{N}^{*}\), \(n\ge m\), and let
\(\forall \,i\in \left\{ m+1,\ldots ,n\right\} \), since \(\varphi _{i}(z)=\overline{\varphi }_{i}^{(n,n)}(z)\), by repeating the previous arguments, we have \(z_{i}\in \left[ \alpha _{i},\beta _{i}\right] \subset \left[ -O_{1},O_{1}\right] \); conversely, \(\forall \,i>n\), we have
and so \(z_{i}\in \left[ \gamma _{i},\delta _{i}\right] \), where
then, since \(\left[ \gamma _{i},\delta _{i}\right] \subset \left[ \alpha _{i},\beta _{i}\right] \), we obtain \(z_{i}\in \left[ -O_{1},O_{1}\right] \) again; thus, we have \(\left\| (z_{j}:j>m)\right\| _{\infty }\le O_{1}\). Moreover, \(\forall \,i\in \left\{ 1,\ldots ,m\right\} \), we have
and so
where
furthermore, we have
and so \(\left\| \left( w_{1}^{(n)},\ldots ,w_{m}^{(n)}\right) \right\| _{I_{m} }\le O_{3}\), from (30); then, since the function \(\left( \varphi ^{(m,m)}\right) ^{-1}\) is continuous, from (29), we have \(\left\| (z_{1},\ldots ,z_{m})\right\| _{I_{m}}\le O_{4}\), and so \(\left\| z\right\| _{\infty }\le O\); thus
Moreover, since \(E_{\infty }\) is closed, the set \(\left[ -M,M\right] ^{\infty }\cap E_{\infty }\) is closed and bounded, and so the sequence \(\left\{ \left( \overline{\varphi }^{(n,n)}\right) ^{-1}\right\} _{n\ge m}\) converges uniformly to \(\varphi ^{-1}\) over \([-M,M]^{\infty }\cap E_{\infty }\); furthermore, since \(\varphi \) is continuous, the set \(\varphi ^{-1}\left( [-M,M]^{\infty }\cap E_{\infty }\right) \) is closed; then, there exist \(\delta =\delta (\varphi ,M)\in \textbf{R}^{+}\) and \(\overline{n}=\overline{n}(\varphi ,M)\in \textbf{N}\), \(\overline{n}\ge m\), such that, \(\forall \,h\in \textbf{N}^{*},h>\overline{n}\), \(\left( \overline{\varphi } ^{(h,h)}\right) ^{-1}\left( [-M,M]^{\infty }\cap E_{\infty }\right) \subset \varphi ^{-1}\left( [-M,M]^{\infty }\cap E_{\infty }\right) +\overline{B_{\infty }(0,\delta )}\subset U\); thus, \(\forall \,n\in \textbf{N}^{*}\), \(n\ge m\), we have
and so formula 31 implies
Analogously, from formula 27, we have
Moreover, by assumption, the functions \(\overline{\varphi }\) and \(\varphi ^{(h,h)}\), \(\forall \,h\in \left\{ m,\ldots ,\overline{n}\right\} \), are continuous; then, from Proposition 59, \(\overline{\varphi }^{(h,h)}\) is continuous too, and so formula (32) implies that K is a closed subset of \(U\cap [-O,O]^{\infty }\). \(\square \)
Definition 39
Let \(m\in \textbf{N}\), let \(n\in \textbf{N}^{*}\), \(n\ge m\), let \(a=\left( a_{i}:i\in \textbf{N}^{*}\right) \in \textbf{R}^{\infty }\), and let \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*} }:\textbf{R}^{\infty }\longrightarrow \textbf{R}^{\infty }\) be a linear and symmetric m-positive definite function; define the measurable function \(\Psi ^{\left( n,a,\Gamma \right) }:\left( \textbf{R}^{\infty },\mathcal {B} ^{\infty }\right) \longrightarrow (\left[ 0,+\infty \right) ,\mathcal {B} \left( \left[ 0,+\infty \right) \right) )\) by
Proposition 40
Let \(\varphi :U=U^{(m)}\times E_{\infty }\longrightarrow E_{\infty }\) be a bijective and continuous \(\left( m,\sigma \right) \) function, such that \(\overline{\varphi }\) is continuous and, for any \(n\in \textbf{N}^{*}\), \(n\ge m\), the function \(\overline{\varphi } ^{(n,n)}:U\longrightarrow E_{\infty }\) is a diffeomorphism; moreover, let \(a=\left( a_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\), let \(b=\left( b_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\) such that \(a_{i} =\varphi _{ii}(b_{i})\), for any \(i>m\), and let \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*}}:\textbf{R}^{\infty }\longrightarrow \textbf{R}^{\infty }\) and \(\Lambda =\left( \lambda _{ij}\right) _{i,j\in \textbf{N}^{*}}:\textbf{R}^{\infty }\longrightarrow \textbf{R}^{\infty }\) some symmetric m-positive definite linear functions, such that \(\prod _{i=m+1}^{\infty }\frac{\lambda _{ii}}{\gamma _{ii}}\in \textbf{R}^{+}\), \(\sum _{i=m+1}^{\infty }\gamma _{ii}\in \textbf{R}^{+}\), \(\sum _{i=m+1}^{\infty }\lambda _{ii}\in \textbf{R}^{+}\), and \(\sum _{i=m+1}^{\infty }\left| \frac{\left( \varphi _{ii}^{\prime }\left( x_{i}\right) \right) ^{2}}{\gamma _{ii}}-\frac{1}{\lambda _{ii}}\right| \in [0,\infty )\), for any \(\left( x_{i}:i>m\right) \in E_{\infty }\); moreover, suppose that there exists \(\varepsilon \in \left[ 0,\infty \right) \) such that \(\left| \varphi _{i}^{(\infty ,m)}\left( x_{1},\ldots ,x_{m}\right) \right| \le \varepsilon \), for any \(i>m\), for any \(\left( x_{1},\ldots ,x_{m}\right) \in U^{(m)}\), and the sequence \(\left\{ \left( \overline{\varphi }^{(n,n)}\right) ^{-1}\right\} _{n>m}\) converges uniformly to \(\varphi ^{-1}\) over the closed and bounded subsets of \(E_{\infty }\). Then, for any \(M\in \textbf{R}^{+}\) such that \(M\ge \left\| a\right\| _{\infty }\), there exists \(O\in \textbf{R}^{+}\), there exists a closed subset K of \(U\cap [-O,O]^{\infty }\), and, for any \(q>m\), there exists a measurable function \(g_{q}:\left( [-O,O],\mathcal {B} \left( [-O,O]\right) \right) \longrightarrow (\textbf{R}^{+},\mathcal {B} \left( \textbf{R}^{+}\right) )\) such that \(\left( \bigotimes _{q=m+1}^{\infty } g_{q}\right) (x)\in \textbf{R}^{+}\), for any \(x\in [-O,O]^{\infty }\), and such that, for any \(n\in \textbf{N}^{*}\), \(n\ge m\), and for any measurable function \(l:\left( E_{\infty },\mathcal {B}_{\infty }\right) \longrightarrow (\left[ 0,\infty \right) ,\mathcal {B}\left( \left[ 0,\infty \right) \right) )\) such that \(l(x)=0\) for any \(x\in E_{\infty }\backslash \left[ -M,M\right] ^{\infty }\), one has
Proof
The previous assumptions imply that \(\overline{\varphi }\) is bijective; then, from Proposition 65 and Remark 55, we have \(\sigma (i)=i\), \(\forall \,i>m\); moreover, \(\varphi _{ij}\) is continuous, \(\forall \,i\in \left\{ 1,\ldots ,m\right\} \), \(\forall \,j>m\), and the function \(\varphi ^{(m,m)}:U^{(m)}\longrightarrow \textbf{R}^{m}\) is open; thus, let \(a=\left( a_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\), let \(M\in \textbf{R}^{+}\) such that \(M\ge \left\| a\right\| _{\infty }\), and let \(O\in \textbf{R}^{+}\) and \(K\subset U\cap [-O,O]^{\infty }\) the closed set defined by Proposition 38; moreover, let \(n\in \textbf{N}^{*}\), \(n\ge m\), and let \(B=\prod _{i=1}^{\infty } B_{i}\in \mathcal {B}^{\infty }\left( \left[ -M,M\right] ^{\infty }\right) \); we have
Furthermore, from Proposition 64 and Proposition 66, \(\varphi ^{(n,n)}\) is a diffeomorphism, and so
Moreover, we have
furthermore, \(\forall \,q>m\), \(\varphi _{qq}\) is a diffeomorphism, from Proposition 66; then, formula (36) implies
since \(a_{q}=\varphi _{qq}\left( b_{q}\right) \), \(\forall \,q>m\), by assumption, and so \(b_{q}=\varphi _{qq}^{-1}\left( a_{q}\right) \in \left[ -O,O\right] \), \(\varphi _{qq}^{-1}\left( B_{q}\right) \subset \left[ -O,O\right] \), from Proposition 38. Moreover, let \(t\in \left[ -O,O\right] \) and let \(q>m\); from Lagrange theorem and since the function \(\varphi _{qq}\) is injective, there exists a unique \(\xi _{q}(t)\in (\min \left\{ t,b_{q}\right\} ,\max \left\{ t,b_{q}\right\} )\in \left[ -O,O\right] \) such that
then, consider the measurable function \(g_{q}:\left( \left[ -O,O\right] ,\mathcal {B}\left( \left[ -O,O\right] \right) \right) \longrightarrow (\textbf{R}^{+},\mathcal {B}\left( \textbf{R}^{+}\right) )\) defined by
from formula (37), we have
Furthermore, the set \([-O,O]\) is closed and bounded, and from Proposition 66 the function \(\varphi _{qq}^{\prime }\) is continuous; then, there exists \(z_{1}=\left( (z_{1})_{q}:q>m\right) \in [-O,O]^{\infty }\) such that, \(\forall \,p\in \left( \textbf{N}^{*}\backslash \left\{ 1,\ldots ,n\right\} \right) \cup \left\{ \infty \right\} \), \(\forall \,(x_{q}:q>m)\in [-O,O]^{\infty }\), we have
and in particular \(\prod _{q=m+1}^{\infty } g_{q}\left( x_{q}\right) \in \textbf{R}^{+}\); furthermore, since \(\mathcal {I} (\varphi )=\textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} \), from Definition 17, we have \(\prod _{q=m+1}^{\infty } \left| \varphi _{qq}^{\prime }\left( x_{q}\right) \right| \in \textbf{R} ^{+}\), from which
thus, formula (41) implies \(\lim _{p\rightarrow \infty } \prod _{q=p+1}^{\infty } g_{q}\left( x_{q}\right) \left| \varphi _{qq}^{\prime }\left( x_{q}\right) \right| =1\). Analogously, there exists \(z_{2}=\left( (z_{2})_{q}:q>m\right) \in \left[ -O,O\right] ^{\infty }\) such that, \(\forall \,p\in \left( \textbf{N}^{*}\backslash \left\{ 1,\ldots ,n\right\} \right) \cup \left\{ \infty \right\} \), \(\forall \,(x_{q}:q>m)\in [-O,O]^{\infty }\), we have
and so, \(\forall \,(x_{q}:q>m)\in [-O,O]^{\infty }\), we obtain
then, formula (42) implies that the measurable function
is \(\left( \bigotimes _{q=n+1}^{\infty } N\left( b_{q},\lambda _{qq}\right) \right) \)-integrable. Moreover, the sequence \(\left\{ \beta _{n,p}\right\} _{p>n}\) converges to \(\beta _{n,\infty }\), and so there exists \(\overline{\beta }_{n}\in \textbf{R}^{+}\) such that \(\beta _{n,p}\le \overline{\beta }_{n}\), \(\forall \,p\in \left( \textbf{N}^{*}\backslash \left\{ 1,\ldots ,n\right\} \right) \cup \left\{ \infty \right\} \), from which \(\left( \bigotimes _{q=n+1}^{\infty } g_{q}\left| \varphi _{qq}^{\prime }\right| \right) (x)\le \alpha \overline{\beta }_{n}\), \(\forall \,x\in \prod _{q=n+1}^{\infty } \varphi _{qq}^{-1}(B_{q})\).
Then, if \(\prod _{q=n+1}^{\infty } \varphi _{qq}^{-1}(B_{q})=\emptyset \), there exists \(q>n\) such that \(\varphi _{qq}^{-1}(B_{q})=\emptyset \), and so
conversely, if \(\prod _{q=n+1}^{\infty } \varphi _{qq}^{-1}(B_{q})\ne \emptyset \), from Theorem 44, there exists
such that \(\left( \bigotimes _{q=n+1}^{\infty } N\left( b_{q},\lambda _{qq}\right) \right) (D)=0\) and such that, \(\forall \,y_{\textbf{N}^{*}\backslash \left\{ 1,\ldots ,n\right\} }\in \left( \prod _{q=n+1}^{\infty } \varphi _{qq}^{-1}(B_{q})\right) \backslash D\), we have
Then, from formulas (39), (43) and (44), we obtain
thus, since \(\left( \overline{\varphi }^{(n,n)}\right) ^{-1}(B)=\left( \varphi ^{(n,n)}\right) ^{-1}\left( \prod _{p=1}^{n} B_{p}\right) \times \left( \prod _{q=n+1}^{\infty } \varphi _{qq}^{-1}(B_{q})\right) \cap E_{\infty }\) and \(\left( \bigotimes _{q=n+1}^{\infty } N\left( b_{q},\lambda _{qq}\right) \right) \left( E_{\infty }\right) =1\), from formulas (34), (35) and (45), we have
Consider the measures \(\mu _{1}\) and \(\mu _{2}\) on \(\Sigma \equiv \mathcal {B} ^{\infty }\left( \left[ -M,M\right] ^{\infty }\right) \) defined by
from formula (46), \(\mu _{1}\) and \(\mu _{2}\) coincide on the set
moreover, we have \(\mu _{1}\left( [-M,M]^{\infty }\right) =\mu _{2}\left( [-M,M]^{\infty }\right) <\infty \), \([-M,M]^{\infty }\in \mathcal {I}\), and so \(\mu _{1}\) and \(\mu _{2}\) are \(\sigma \)- finite on \(\mathcal {I}\). Then, since \(\mathcal {I}\) is a \(\pi \)-system on \([-M,M]^{\infty }\) such that \(\sigma (\mathcal {I})=\mathcal {B}^{\infty }\left( \left[ -M,M\right] ^{\infty }\right) \), from Proposition 45, \(\mu _{1}\) and \(\mu _{2}\) coincide on \(\mathcal {B}^{\infty }\left( \left[ -M,M\right] ^{\infty }\right) \), and in particular on \(\mathcal {B}_{\infty }\left( \left[ -M,M\right] ^{\infty }\cap E_{\infty }\right) \); thus, suppose that \(\psi :\left( E_{\infty },\mathcal {B}_{\infty }\right) \longrightarrow ([0,+\infty ),\mathcal {B}\left( [0,+\infty )\right) )\) is a simple function such that \(\psi (x)=0\) \(\forall \,x\in E_{\infty }\backslash \left[ -M,M\right] ^{\infty }\); \(\forall \,x\in E_{\infty }\), we have \(\psi (x)=\sum _{i=1}^{k} c_{i}1_{B_{i}}\), for some \(k\in \textbf{N}^{*}\), \(c_{1},\ldots ,c_{k}\in \textbf{R}\), \(B_{1},\ldots ,B_{k}\in \mathcal {B}_{\infty }\left( \left[ -M,M\right] ^{\infty }\cap E_{\infty }\right) \), and so
Then, if \(l:\left( E_{\infty },\mathcal {B}_{\infty }\right) \longrightarrow ([0,+\infty ),\mathcal {B}\left( [0,+\infty )\right) )\) is a measurable function such that \(l(x)=0\) \(\forall \,x\in E_{\infty }\backslash \left[ -M,M\right] ^{\infty }\), and \(\left\{ \psi _{i}\right\} _{i\in \textbf{N}}\) is a sequence of increasing positive simple functions over \(\left( E_{\infty },\mathcal {B}_{\infty }\right) \) such that \(\psi _{i}(x)=0\) \(\forall \,x\in E_{\infty }\backslash \left[ -M,M\right] ^{\infty }\), \(\forall \,i\in \textbf{N} \), and \(\lim _{i\longrightarrow \infty }\psi _{i}=l\), from Beppo Levi theorem we have
\(\square \)
Theorem 41
(Gaussian change of variables’ formula). Let \(\varphi :U=U^{(m)}\times E_{\infty }\longrightarrow E_{\infty }\) be a bijective, continuous and \(\left( m,\sigma \right) \) function, such that \(\overline{\varphi }:U\longrightarrow E_{\infty }\) is continuous and such that, for any \(n\in \textbf{N}^{*}\), \(n\ge m\), the function \(\overline{\varphi } ^{(n,n)}:U\longrightarrow E_{\infty }\) is a diffeomorphism; moreover, let \(a=\left( a_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\), let \(b=\left( b_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\) such that \(a_{i} =\varphi _{ii}(b_{i})\), for any \(i>m\), and let \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*}}:E_{\infty }\longrightarrow E_{\infty }\) and \(\Lambda =\left( \lambda _{ij}\right) _{i,j\in \textbf{N}^{*}}:E_{\infty }\longrightarrow E_{\infty }\) some symmetric m-positive definite linear functions, such that \(\prod _{i=m+1}^{\infty } \frac{\lambda _{ii}}{\gamma _{ii}}\in \textbf{R}^{+}\), \(\sum _{i=m+1}^{\infty } \gamma _{ii}\in \textbf{R}^{+}\), \(\sum _{i=m+1}^{\infty } \lambda _{ii}\in \textbf{R}^{+}\), and \(\sum _{i=m+1}^{\infty } \left| \frac{\left( \varphi _{ii}^{\prime }\left( x_{i}\right) \right) ^{2}}{\gamma _{ii}}-\frac{1}{\lambda _{ii}}\right| \in \left[ 0,+\infty \right) \), for any \(\left( x_{i}:i>m\right) \in E_{\infty }\); finally, suppose that there exists \(\varepsilon \in \left[ 0,\infty \right) \) such that \(\left| \varphi _{i}^{(\infty ,m)}\left( x_{1},\ldots ,x_{m}\right) \right| \le \varepsilon \), for any \(i>m\), for any \(\left( x_{1},\ldots ,x_{m}\right) \in U^{(m)}\), and suppose that the sequence \(\left\{ \left( \overline{\varphi }^{(n,n)}\right) ^{-1}\right\} _{n>m}\) converges uniformly to \(\varphi ^{-1}\) over the closed and bounded subsets of \(E_{\infty }\), and the sequences \(\left\{ -\Psi ^{\left( n,a,\Gamma \right) }\circ \overline{\varphi }^{(n,n)} +\Psi ^{\left( n,b,\Lambda \right) }\right\} _{n>m}\) and \(\left\{ \det J_{\overline{\varphi }^{(n,n)}}\right\} _{n>m}\) converge uniformly over the closed and bounded subsets of U. Then, for any \(B\in \mathcal {B}_{\infty }\) and for any measurable function \(f:\left( E_{\infty },\mathcal {B}_{\infty }\right) \longrightarrow (\overline{\textbf{R} },\mathcal {B}\left( \overline{\textbf{R}}\right) )\) such that \(f^{+}\) (or \(f^{-}\)) is \(N(a,\Gamma )\)-integrable, one has
In particular, assume that, for any \(x\in U\), there exists the function \(J_{\varphi }(x):\ell _{\infty }\longrightarrow \ell _{\infty }\); then, one has
Proof
From Proposition 40, \(\forall \,M\in \textbf{R}^{+}\) such that \(M\ge \left\| a\right\| _{\infty }\), there exists \(O\in \textbf{R}^{+}\), there exists a closed subset K of \(U\cap [-O,O]^{\infty }\), and, \(\forall \,q>m\), there exists a measurable function \(g_{q}:\left( [-O,O],\mathcal {B}\left( [-O,O]\right) \right) \longrightarrow (\textbf{R}^{+},\mathcal {B}\left( \textbf{R}^{+}\right) )\) such that \(\left( \bigotimes _{q=m+1}^{\infty } g_{q}\right) (x)\in \textbf{R}^{+}\), \(\forall \,x\in [-O,O]^{\infty }\), and such that, \(\forall \,n\in \textbf{N}^{*}\), \(n\ge m\), and \(\forall \,\)measurable function \(l:\left( E_{\infty },\mathcal {B}_{\infty }\right) \longrightarrow (\left[ 0,\infty \right) ,\mathcal {B}\left( \left[ 0,\infty \right) \right) )\) such that \(l(x)=0\) \(\forall \,x\in E_{\infty }\backslash \left[ -M,M\right] ^{\infty }\), we have \(\int _{E_{\infty }}ldN(a,\Gamma )=\int _{K} h_{n}dN\left( b,\Lambda \right) \), where, \(\forall \,n\in \textbf{N}^{*}\), \(n\ge m\), the function \(h_{n}:\left( K,\mathcal {B}^{\infty }\left( K\right) \right) \longrightarrow ([0,\infty ),\mathcal {B}\left( [0,\infty )\right) )\) is defined by
thus, we obtain
In particular, suppose that l is \(\left( \tau _{\left\| \cdot \right\| _{\infty }}\left( E_{\infty }\right) ,\tau \left( [0,\infty )\right) \right) \)-continuous and such that \(l\left( E_{\infty }\right) \subset \left[ 0,1\right] \), \(l(x)=0\), \(\forall \,x\in E_{\infty } \backslash \left[ -M,M\right] ^{\infty }\); by assumption, the sequences \(\left\{ -\Psi ^{\left( n,a,\Gamma \right) }\circ \overline{\varphi } ^{(n,n)}+\Psi ^{\left( n,b,\Lambda \right) }\right\} _{n>m}\) and \(\left\{ \det J_{\overline{\varphi }^{(n,n)}}\right\} _{n>m}\) converge uniformly over K; then, there exists \(\widehat{n}\in \textbf{N}^{*}\), \(\widehat{n}>m\), such that, \(\forall \,n\ge \widehat{n}\), \(\forall \,x\in K\), we have
and so, \(\forall \,n>m\), \(\forall \,x\in K\), the following inequalities hold:
Moreover, \(\forall \,h\in \left\{ m+1,\ldots ,\widehat{n}\right\} \), \(\forall \,x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in K\), we have
furthermore, from Proposition 71 and Proposition 61, \(\forall \,h\in \left\{ m+1,\ldots ,\widehat{n}\right\} \), \(\forall \,i\in \textbf{N}^{*}\backslash \left\{ 1,\ldots ,h\right\} \), the functions \(\det J_{\varphi ^{(h,h)}}\) and \(\varphi _{ii}^{\prime }\) are continuous; moreover, the sets \(\pi _{\left\{ 1,\ldots ,h\right\} }\left( K\right) \) and \(\pi _{i}\left( K\right) \) are closed and bounded, and from Definition 17 we have \(\prod _{i=h+1}^{\infty } \left| \varphi _{ii}^{\prime }(x_{i})\right| \in \textbf{R}^{+}\), \(\forall \,x=(x_{i}:i\in \textbf{N}^{*}\backslash \left\{ 1,\ldots ,h\right\} )\in \textbf{R}^{\infty }\); then, from formulas (50) and (51), there exists \(c\in \textbf{R}^{+}\) such that
finally, from formula (40), there exists \(\alpha \in \textbf{R}^{+}\) such that \(\prod _{q=n+1}^{\infty } g_{q}(x_{q})\le \alpha \), \(\forall \,n>m\), \(\forall \,x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in K\); thus, since \(l\left( E_{\infty }\right) \subset [0,1]\), we obtain \(\left| h_{n}\right| \le g\), \(\forall \,n>m\), where
is the measurable function defined by
and so
furthermore, \(\forall \,x_{I_{h}}\in \textbf{R}^{h}\), we have
and \(\left( \bigotimes _{i=h+1}^{\infty } N\left( b_{i},\lambda _{ii}\right) \right) \left( E_{\infty }\right) =1\) (from Proposition 36); then, formula (53) implies
where, \(\forall \,h\in \left\{ m+1,\ldots ,\widehat{n}\right\} \), the measurable function
is defined by
moreover, \(\forall \,h\in \left\{ m+1,\ldots ,\widehat{n}\right\} \), the function \(\varphi ^{(h,h)}\) is continuous, and so the function \(t_{h}\) is continuous too; then, since the set \(\pi _{\left\{ 1,\ldots ,h\right\} }(K)\) is closed and bounded, from formula (54), we obtain \(\int _{K} gdN\left( b,\Lambda \right) <\infty \). Moreover, we have \(\lim _{n\rightarrow \infty }\overline{\varphi }^{(n,n)}=\varphi \), from the point 1 of Definition 17, and \(\lim _{n\rightarrow \infty }\left( \prod _{q=n+1}^{\infty } g_{q}(x_{q})\right) =1\), \(\forall \,x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in K\), from which
consequently, from formula (49) and from the dominated convergence theorem, we obtain
Moreover, \(\forall \,y=(y_{i}:i\in \textbf{N}^{*})\in \ell _{\infty }\), \(\forall \,\delta =(\delta _{i}:i\in \textbf{N}^{*})\in \left( \textbf{R} ^{+}\right) ^{\infty }\cap \ell _{\infty }\) such that \(\inf _{i\in \textbf{N}^{*}}\delta _{i}>0\), \(\forall \,h\in \textbf{N}^{*}\), \(\forall \,t\in [0,1]\), let \(A_{y,\delta ,h,t}=\prod _{i=1}^{\infty }\left( y_{i}-\delta _{i}\left( 1-\frac{t}{2h}\right) ,y_{i}+\delta _{i}\left( 1-\frac{t}{2h}\right) \right) \subset \ell _{\infty } \), and let
where the interior parts of \(A_{y,\delta ,1,0}\) and \(\left( -M,M\right) ^{\infty }\) are considered with respect to the topology \(\tau _{\left\| \cdot \right\| _{\infty }}\), and so \(\overset{\circ }{\left( -M,M\right) ^{\infty }}=B_{\infty }(0,M)\); since \(\inf _{i\in \textbf{N}^{*}}\delta _{i}>0\), we have \(\overset{\circ }{A_{y,\delta ,h,t}}\cap E_{\infty } \in \tau _{\left\| \cdot \right\| _{\infty }}\left( \overset{\circ }{\left( -M,M\right) ^{\infty }}\cap E_{\infty }\right) \subset \tau _{\left\| \cdot \right\| _{\infty }}\left( E_{\infty }\right) \). Consider the function \(l_{y,\delta ,h}:E_{\infty }\longrightarrow [0,\infty )\) defined (with respect to the topology \(\tau _{\left\| \cdot \right\| _{\infty }}\)) by
observe that \(l_{y,\delta ,h}:E_{\infty }\longrightarrow [0,\infty )\) is a function such that \(l_{y,\delta ,h}\left( E_{\infty }\right) \subset [0,1]\), \(l_{y,\delta ,h}(x)=0\), \(\forall \,x\notin [-M,M]^{\infty }\); moreover, \(\forall \,t_{1},t_{2}\in [0,\infty )\) such that \(t_{1}<t_{2}\), we have
thus, \(l_{y,\delta ,h}\) is \(\left( \tau _{\left\| \cdot \right\| _{\infty } }\left( E_{\infty }\right) ,\tau \left( [0,\infty )\right) \right) \)-continuous, and so it is \(\left( \mathcal {B}_{\infty },\mathcal {B}\left( [0,\infty )\right) \right) \)-measurable. Then, \(\forall \,B=\overset{\circ }{A_{y,\delta ,1,0}}\cap E_{\infty }\in \mathcal {A}_{M}\), since \(\left\{ l_{y,\delta ,h}\right\} _{h\in \textbf{N}^{*}}\) is an increasing positive sequence such that \(\lim _{h\longrightarrow \infty }l_{y,\delta ,h}=1_{B}\), from Beppo Levi theorem and formula (55), we have
Consider the measures \(\mu \) and \(\upsilon \) on \(\left( E_{\infty },\mathcal {B}_{\infty }\right) \) defined by
from formula (56), \(\mu \) and \(\upsilon \) coincide on the set \(\mathcal {A}_{M}\); moreover, we have
and so \(\mu \) and \(\upsilon \) are \(\sigma \)- finite on \(\mathcal {A}_{M}\); furthermore, since \(\mathcal {B}^{\infty }\left( \overset{\circ }{\left( -M,M\right) ^{\infty }}\cap E_{\infty }\right) \supset \sigma (\mathcal {A} _{M})\supset \tau _{\left\| \cdot \right\| _{\infty }}\left( \overset{\circ }{\left( -M,M\right) ^{\infty }}\cap E_{\infty }\right) \) and since Proposition 14 implies
we obtain \(\sigma (\mathcal {A}_{M})=\mathcal {B}^{\infty }\left( \overset{\circ }{\left( -M,M\right) ^{\infty }}\cap E_{\infty }\right) \); then, since \(\mathcal {A}_{M}\) is a \(\pi \)-system on \(\overset{\circ }{\left( -M,M\right) ^{\infty }}\cap E_{\infty }\), from Proposition 45, \(\mu \) and \(\upsilon \) coincide on \(\mathcal {B}^{\infty }\left( \overset{\circ }{\left( -M,M\right) ^{\infty }}\cap E_{\infty }\right) \).
Moreover, \(\forall \,B\in \mathcal {B}_{\infty }\), \(\forall \,k\in \textbf{N}^{*}\), set \(B_{k}=B\cap \overset{\circ }{\left( -k,k\right) ^{\infty }} \in \mathcal {B}^{\infty }\left( \overset{\circ }{\left( -k,k\right) ^{\infty } }\cap E_{\infty }\right) \); we have \(B_{k}\subset B_{k+1}\), \(\varphi ^{-1}\left( B_{k}\right) \subset \varphi ^{-1}\left( B_{k+1}\right) \), \(\bigcup _{k\in \textbf{N}^{*}}B_{k}=B\) and \( \bigcup _{k\in \textbf{N}^{*}}\varphi ^{-1}(B_{k})=\varphi ^{-1}(B)\); thus, from the continuity property of \(\mu \) and \(\upsilon \), we have
Then, let \(B\in \mathcal {B}_{\infty }\) and let \(g:\left( E_{\infty },\mathcal {B}_{\infty }\right) \) \(\longrightarrow (\left[ 0,+\infty \right] ,\mathcal {B}\left( \left[ 0,+\infty \right] \right) )\) be a measurable function; by proceeding as in the proof of formula (47), formula (57) implies
Then, for any measurable function \(f:\left( E_{\infty },\mathcal {B}_{\infty }\right) \longrightarrow (\overline{\textbf{R}},\mathcal {B}\left( \overline{\textbf{R}}\right) )\) such that \(f^{+}\) (or \(f^{-}\)) is \(N(a,\Gamma )\)-integrable, we have
In particular, assume that, \(\forall \,x\in U\), there exists the function \(J_{\varphi }(x):\ell _{\infty }\longrightarrow \ell _{\infty }\); from Proposition 69, we have
and so formula (58) implies
\(\square \)
Example 42
Let \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*} }:E_{\infty }\longrightarrow E_{\infty }\) be a linear and diagonal function, such that \(\gamma _{ii}\in \textbf{R}^{+}\), for any \(i\in \textbf{N}^{*}\), and \(\sum _{i=1}^{\infty }\gamma _{ii}\in \textbf{R}^{+}\); then, one has
Proof
Consider the linear function \(\varphi :E_{\infty }\longrightarrow E_{\infty }\) defined by
observe that \(\varphi \) is \(\left( 1,\sigma \right) \) standard, where \(\sigma (i)=i\), \(\forall \,i>1\); thus, from Proposition 21, \(\varphi \) is continuous; moreover, the function \(\overline{\varphi }:E_{\infty }\longrightarrow E_{\infty }\) is given by
and so \(\overline{\varphi }\) is \(\left( 1,\sigma \right) \) standard, bijective and continuous; then, from Propositions 65 and 67, \(\varphi \) is bijective; analogously, \(\forall \,n\in \textbf{N}^{*}\), the function \(\overline{\varphi }^{(n,n)}:E_{\infty }\longrightarrow E_{\infty }\) is \(\left( 1,\sigma \right) \) standard, with \(\overline{\overline{\varphi }^{(n,n)}}=\overline{\varphi }\), and so \(\overline{\varphi }^{(n,n)}\) is bijective too; moreover, \(\varphi \) is \(C^{1}\) and Proposition 66 implies that \(\overline{\varphi }\) is a diffeomorphism; then, from Proposition 63, \(\overline{\varphi }^{(n,n)}\) is a diffeomorphism. Observe that,\(\forall \,y=\left( y_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }\), \(\forall \,n\in \textbf{N} ^{*}\), \(\forall \,i>1\), we have \((\varphi ^{-1}(y))_{i}=(\left( \overline{\varphi }^{(n,n)}\right) ^{-1}(y))_{i}=y_{i}\); then
and so
thus, from formula (60), we have
Then, let H a closed and bounded subset of \(E_{\infty }\), and let \(d\in \textbf{R}^{+}\) such that \(H\subset \left[ -d,d\right] ^{\infty }\); \(\forall \,y\in H\), formula (61) implies
and so the sequence \(\left\{ \left( \overline{\varphi }^{(n,n)}\right) ^{-1}\right\} _{n>1}\) converges uniformly to \(\varphi ^{-1}\) over H; moreover, \(\forall \,n\in \textbf{N}^{*}\), \(\forall \,x=(x_{i}:i\in \textbf{N}^{*})\in E_{\infty }\), we have \(\det J_{\overline{\varphi } ^{(n,n)}}(x)=2\), and
as we observed, the sequence of real functions \(\left\{ h_{n}\right\} _{n\ge 1}\) on \(E_{\infty }\) given by \(h_{n}(x_{i}:i\in \textbf{N}^{*})=\sum _{i=1}^{n} 2^{-i+2}x_{i}\), \(\forall \,n\in \textbf{N}^{*}\), \(\forall \,(x_{i}:i\in \textbf{N}^{*})\in E_{\infty }\), converges uniformly to h over the closed and bounded subsets of \(E_{\infty } \), where the real function h on \(E_{\infty }\) is defined by
then, from formula (63), the sequence \(\left\{ -\Psi ^{\left( n,0,\Gamma \right) }\circ \overline{\varphi }^{(n,n)}+\Psi ^{\left( n,0,\Gamma \right) }\right\} _{n>1}\) converges uniformly over the closed and bounded subsets of \(E_{\infty }\), and
moreover, \(\forall \,\left( x_{i}:i>1\right) \in E_{\infty }\), \(\forall \,i>1\), we have \(\frac{\left( \varphi _{ii}^{\prime }\left( x_{i}\right) \right) ^{2}}{\gamma _{ii}}-\frac{1}{\gamma _{ii}}=0\), and so \(\sum _{i=2}^{\infty } \left( \frac{\left( \varphi _{ii}^{\prime }\left( x_{i}\right) \right) ^{2}}{\gamma _{ii}}-\frac{1}{\gamma _{ii}}\right) =0\). Then, the assumptions of Theorem 41 are satisfied in the case \(a=b=0\in E_{\infty }\), \(\Lambda =\Gamma \), \(B=\left\{ (y_{i}:i\in \textbf{N}^{*})\in E_{\infty }:y_{1}>0\right\} \) and \(f(x)=1\), \(\forall \,x\in E_{\infty }\); thus, since \(\varphi ^{-1}\left( B\right) =\left\{ (x_{i}:i\in \textbf{N}^{*})\in E_{\infty }: \sum _{i=1}^{\infty } 2^{-i+1}x_{i}>0\right\} \), we obtain
from which (59) follows. \(\square \)
At this point, we can compare Theorem 41 with Theorem 4.1 in [24]. The last result proves a formula like (48), although written in a formally different way. The function that defines the change of variables is of the type \(\varphi :E\longrightarrow E\), where E is a separable Banach space with a Gaussian measure \(\mu \), such that \(\varphi (x)=x+F(x)\), for any \(x\in E\), for some function \(F:E\longrightarrow H\), where \(H=H(\mu )\) is the Cameron–Martin space of \(\left( E,\mu \right) \), with the norm \(\left| \cdot \right| _{H}\) defined by formula (13). The function F satisfies some conditions, including the hypothesis
for almost all \(x\in E\), for all \(h\in H\), where \(c\in (0,1)\) (see Theorem 3.3 in [24]).
In Theorem 41, we consider a \(\left( m,\sigma \right) \) function \(\varphi :U\subset E_{\infty }\longrightarrow E_{\infty }\), with some properties. In general, the two theorems cannot be compared, but, in some particular cases, it is possible to prove that the hypotheses of Theorem 4.1 in [24] imply those of Theorem 41, which can therefore be considered a generalization of the first. For example, consider the function \(\varphi :E_{\infty }\longrightarrow E_{\infty }\) and the measure \(N(0,\Gamma )\) defined by Example 42; as we have seen, this is a case where Theorem 41 can be applied; conversely, let \(h=(h_{j}:j\in \textbf{N}^{*})\) given by \(h_{j}=\delta _{j1}\), for any \(j\in \textbf{N}^{*}\); since \(h\in \left\{ x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }:\sum _{i=1}^{\infty } \frac{x_{i}^{2}}{\gamma _{ii}} \in \textbf{R}\right\} =\ell _{2}(\Gamma )\), and since Proposition 37 implies \(\ell _{2}(\Gamma )\subset H\left( N(0,\Gamma )\right) \), we obtain \(h\in H\left( N(0,\Gamma )\right) \); moreover
furthermore, we have \(\varphi (x)=x+F(x)\), where \(F_{1}(x)=x_{1}+\sum _{i=2}^{\infty } 2^{-i+2}x_{i}\), \(F_{j}(x)=0\), for any \(x=(x_{i}:i\in \textbf{N}^{*})\in E_{\infty }\), for any \(j>1\), and so \(F(x)\in \ell _{2}(\Gamma )\subset H\left( N(0,\Gamma )\right) \); then, since \(F(x+h)-F(x)=F(h)=h\), for any \(x\in E_{\infty }\), and since \(\left| h\right| _{H\left( N(0,\Gamma )\right) }>0\), the hypothesis given by formula (64) is not satisfied, and so we cannot use Theorem 4.1 in [24].
5 Problems for further study
A natural application of this paper, in the probability theory, is the generalization of Definition 32, by defining a random variable on \(\left( E_{\infty },\mathcal {B}_{\infty }\right) \) of the form \(X=AZ+c\), where \(A=\left( a_{ij}\right) _{i,j\in \textbf{N}^{*}}:E_{\infty }\longrightarrow E_{\infty }\) is a bijective and linear \(\left( m,\sigma \right) \) standard function, \(c\in \ell _{1}(\Gamma )\equiv \left\{ x=\left( x_{i}:i\in \textbf{N}^{*}\right) \in E_{\infty }:\sum _{i=m+1}^{\infty } \frac{\left| x_{i}\right| }{\gamma _{ii}}\in \textbf{R}\right\} \subset \ell _{2}(\Gamma )\cap \ell _{1}\subset H\left( N(0,\Gamma )\right) \), \(Z:\left( \Omega ,\mathcal {F},P\right) \longrightarrow \left( E_{\infty },\mathcal {B}_{\infty }\right) \) is a random variable with law \(N(0,\Gamma )\), where \(\Gamma =\left( \gamma _{ij}\right) _{i,j\in \textbf{N}^{*}}\) is a symmetric and m-positive definite linear function, such that \(\sum _{i=m+1}^{\infty } \gamma _{ii}\in \textbf{R}^{+}\) and \(\sum _{i=m+1}^{\infty } \frac{\left| a_{ii}^{2}-1\right| }{\gamma _{ii}}\in \left[ 0,+\infty \right) \). Since A is continuous and \(N(0,\Gamma )\) is a Gaussian measure, then X is a Gaussian random variable; moreover, the assumptions of Theorem 41 are satisfied, with \(\varphi (x)=Ax+c\), for any \(x\in E_{\infty }\), \(a=b=0\) and \(\Lambda =\Gamma \); consequently we can determine the density of X with respect to the probability measure \(N(0,\Gamma )\). In particular, if \(A=\textbf{I}_{\infty }\), this result provides an explicit expression of the Cameron–Martin formula (see Corollary 2.4.3 in [11]). Moreover, it is possible to develop some theoretical results (for example, the convolution between the laws of two independent Gaussian random variables with values in \(E_{\infty }\)) and some applications in the statistical inference.
Furthermore, in the statistical mechanics, it is possible to describe the systems of smooth hard particles, by using the Boltzmann equation (see, for example, the paper [22]), or the more general Master kinetic equation, described in the papers [19,20,21]. In order to study the evolution of these systems, we can consider the model of countable particles, such that their joint infinite-dimensional density can be determined by composing an infinite-dimensional Gaussian random variable with a \(\left( m,\sigma \right) \) function.
References
Accardi, L., Smolyanov, O.G., Smolyanova, M.O.: Change of variable formulas for infinite-dimensional distributions. Math. Notes 60, 212–215 (1996)
Asci, C.: Generating uniform random vectors in \(\textbf{Z} _{p}^{k}\): the general case. J. Theoret. Probab. 22, 791–809 (2009)
Asci, C.: Asymptotic behavior of an affine random recursion in \(\textbf{Z} _{p}^{k}\) defined by a matrix with an eigenvalue of size \(1\). Stat. Probab. Lett. 79, 1421–1428 (2009)
Asci, C.: Convergence in total variation of an affine random recursion in \([0, p)^{k}\) to a uniform random vector. Markov Process. Relat. Fields 19, 125–140 (2013)
Asci, C.: Integration over an infinite-dimensional Banach space and probabilistic applications. Int. J. Anal. (2014). https://doi.org/10.1155/2014/404186
Asci, C.: Differentiation theory over infinite-dimensional Banach spaces. J. Math. (2016). https://doi.org/10.1155/2016/2619087
Asci, C.: Theory of the \((m,\sigma )\)-general functions over infinite-dimensional Banach spaces. Rend. Istit. Mat. Univ. Trieste 50, 1–36 (2018)
Asci, C.: Change of variables’ formula for the integration of the measurable real functions over infinite-dimensional Banach spaces. Rend. Istit. Mat. Univ. Trieste 51, 61–103 (2019)
Baker, R.: “Lebesgue measure’’ on \(\textbf{R} ^{\infty }\). Proc. Am. Math. Soc. 113, 1023–1029 (1991)
Billingsley, P.: Probability and Measure, 3rd edn. Wiley, New York (1995)
Bogachev, V.: Gaussian measures. In: Mathematical Surveys and Monographs, vol. 62. American Mathematical Society, Providence, RI (1998)
Jessen, B.: The theory of integration in a space of an infinite number of dimensions. Acta Math. 63, 249–323 (1934)
Kuo, H.H.: Gaussian measures in Banach spaces. In: Lecture Notes in Mathematics, vol. 463. Springer, New York (1975)
Lang, S.: Analysis II. Addison-Wesley, Reading, MA (1969)
Lang, S.: Linear Algebra, 3rd edn. Springer, New York (1987)
Léandre, R.: Lebesgue measure in infinite dimension as an infinite-dimensional distribution. J. Math. Sci. 159, 833–836 (2009)
Rao, M.M.: Measure Theory and Integration. Wiley, New York (1987)
Rudin, W.: Real and Complex Analysis, 2nd edn. McGraw Hill, New York (1974)
Tessarotto, M., Cremaschini, C., Asci, C., Soranzo, A., Tironi, G.: Global validity of the Master kinetic equation for hard-sphere systems. Eur. Phys. J. Plus. 130 (2015). https://doi.org/10.1140/epjp/i2015-15169-2
Tessarotto, M., Asci, C.: Asymptotic orderings and approximations of the Master kinetic equation for large hard spheres systems. Phys. Lett. A 381, 1484–1489 (2017)
Tessarotto, M., Mond, M., Asci, C.: Microscopic statistical description of incompressible Navier–Stokes granular fluids. Eur. Phys. J. Plus. 132 (2017). https://doi.org/10.1140/epjp/i2017-11472-2
Tessarotto, M., Cremaschini, C., Mond, M., Asci, C., Soranzo, A., Tironi, G.: On the Boltzmann–Grad limit for smooth hard-sphere systems. Found. Phys. 48, 271–294 (2018)
Tsilevich, N., Vershic, A., Yor, M.: An infinite-dimensional analogue of the Lebesgue measure and distinguished properties of the Gamma process. J. Funct. Anal. 185, 274–296 (2001)
Ustunel, A. S., Zakai, M.: The change of variables formula on Wiener space. In: Séminaire de Probabilités XXXI, Lecture Notes in Mathematics, vol. 1655, pp. 24–39. Springer, Berlin (1997)
Weidmann, J.: Linear Operators in Hilbert Spaces. Springer, New York (1980)
Williams, D.: Probability with Martingales. Cambridge University Press, Cambridge (1991)
Funding
Open access funding provided by Università degli Studi di Trieste within the CRUI-CARE Agreement.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author declares that there is no conflict of interest regarding the publication of this paper.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
In this section, we expose some definitions and results about the differentiation theory over infinite-dimensional Banach spaces, and about the theory of the \(\left( m,\sigma \right) \) functions, introduced in Sect. 3. We start with the following three results (from Theorem 43 to Proposition 45) of the measure theory.
Theorem 43
Let \(I\ne \emptyset \) be a set and, for any \(i\in I\), let \(\left( S_{i},\Sigma _{i},\mu _{i}\right) \) be a measure space such that \(\mu _{i}\) is finite. Moreover, suppose that, for some countable set \(J\subset I\) and for any \(i\in I\backslash J\), \(\mu _{i}\) is a probability measure, and \(\prod _{j\in J}\mu _{j}(S_{j})\in \mathbf {[}0\mathbf {,\infty )}\). Then, over the measurable space \(\left( \prod _{i\in I} S_{i},\bigotimes _{i\in I}\Sigma _{i}\right) \), there is a unique finite measure \(\mu \), indicated by \(\bigotimes _{i\in I}\mu _{i}\), such that, for any \(H\subset I\) such that \(0<|H|<\infty \) and for any \(A=\prod _{h\in H} A_{h}\times \prod _{i\in I\backslash H} S_{i}\in \bigotimes _{i\in I}\Sigma _{i}\), where \(A_{h}\in \Sigma _{h}\) \(\forall \,h\in H\), we have \(\mu (A)=\prod _{h\in H}\mu _{h}(A_{h})\prod _{j\in J\backslash H}\mu _{j}(S_{j}).\) In particular, if I is countable, then \(\mu (A)=\prod _{i\in I}\mu _{i}(A_{i})\) for any \(A=\prod _{i\in I} A_{i}\in \bigotimes _{i\in I}\Sigma _{i}\).
Proof
See the proof of Corollary 4 in Asci [5]. \(\square \)
Theorem 44
For any \(i\in \textbf{N}^{*}\), let \(\left( S_{i},\Sigma _{i},\mu _{i}\right) \) be a measure space such that \(\mu _{i}\) is finite, and \(\prod _{i=1}^{\infty } \mu _{i}(S_{i})\in \mathbf {[} 0\mathbf {,\infty )}\); moreover, let \(\left( S,\Sigma ,\mu \right) =\left( \prod _{i=1}^{\infty } S_{i},\bigotimes _{i=1}^{\infty } \Sigma _{i}, \bigotimes _{i=1}^{\infty }\mu _{i}\right) \), let \(f\in L^{1}\left( S,\Sigma ,\mu \right) \) and, for any \(H\subset \textbf{N}^{*}\) such that \(0<|H|<\infty \), let the measurable function \(f_{H^{c}}:\left( S,\Sigma \right) \longrightarrow (\textbf{R},\mathcal {B})\) defined by
where \(\left( S_{H},\Sigma _{H},\mu _{H}\right) \) is the measure space \(\left( \prod _{i\in H}S_{i},\bigotimes _{i\in H}\Sigma _{i},\bigotimes _{i\in H}\mu _{i}\right) \). Then:
-
1.
If \(\prod _{i=1}^{\infty }\mu _{i}(S_{i})>0\), there exists \(D\in \Sigma \) such that \(\mu (D)=0\) and such that, for any \(x\in D^{c}\), one has \(\lim _{n\rightarrow \infty }f_{\left\{ 1,\ldots ,n\right\} ^{c} }(x)=\int _{S} fd\mu \).
-
2.
If \(\prod _{i=1}^{\infty } \mu _{i}(S_{i})=0\) and f is bounded, for any \(x\in S\), one has \(\lim _{n\rightarrow \infty } f_{\left\{ 1,\ldots ,n\right\} ^{c}}(x)=\int _S fd\mu \).
Proof
-
1.
Suppose that \(\prod _{i=1}^{\infty } \mu _{i}(S_{i})>0\); this implies \(\mu _{i}(S_{i})\ne 0\), \(\forall \,i\in \textbf{N}^{*}\); thus, \(\overline{\mu }_{i}=\frac{\mu _{i}}{\mu _{i}(S_{i})}\) is a probability measure, and also \(\overline{\mu }=\bigotimes _{i=1}^{\infty }\overline{\mu }_{i}\), \(\overline{\mu }_{\left\{ 1,\ldots ,n\right\} } =\bigotimes _{i=1}^{n}\overline{\mu }_{i}\), \(\forall \,n\in \textbf{N}^{*}\). We have \(\int _{S}\left| f\right| d\overline{\mu }=\frac{1}{\prod _{i=1}^{\infty }\mu _{i}(S_{i})}\int _{S}\left| f\right| d\mu <\infty \); then, define the function \(\overline{f}_{\left\{ 1,\ldots ,n\right\} ^{c}}:\left( S,\Sigma \right) \longrightarrow (\textbf{R},\mathcal {B})\) by
$$\begin{aligned} \overline{f}_{\left\{ 1,\ldots ,n\right\} ^{c}}(x)=\underset{S_{\left\{ 1,\ldots ,n\right\} }}{\int }f(\cdot ,x_{\left\{ 1,\ldots ,n\right\} ^{c}} )d\overline{\mu }_{\left\{ 1,\ldots ,n\right\} }; \end{aligned}$$from Theorem 3, page 349, in Rao [17], there exists \(D\in \Sigma \) such that \(\overline{\mu }(D)=0\) and such that, \(\forall \,x\in D^{c}\), we have
$$\begin{aligned} \underset{n\rightarrow \infty }{\lim }f_{\left\{ 1,\ldots ,n\right\} ^{c}}(x)= & {} \underset{n\rightarrow \infty }{\lim }\left( \underset{i=1}{\overset{n}{\prod }}\mu _{i}(S_{i})\right) \overline{f}_{\left\{ 1,\ldots ,n\right\} ^{c}}(x)\\= & {} \underset{i=1}{\overset{\infty }{\prod }}\mu _{i}(S_{i})\underset{S}{\int }fd\overline{\mu }=\underset{S}{\int }fd\mu ; \end{aligned}$$thus, since \(\mu (D)= \prod _{i=1}^{\infty } \mu _{i} (S_{i})\overline{\mu }(D)=0\), we obtain the statement.
-
2.
If \(\prod _{i=1}^{\infty } \mu _{i}(S_{i})=0\) and f is bounded, let \(M\in \textbf{R}^{+}\) such that \(\left| f\right| \le M\), and let \(x\in S\); we have
$$\begin{aligned} 0\le & {} \underset{n\rightarrow \infty }{\lim \inf }\left| f_{\left\{ 1,\ldots ,n\right\} ^{c}}(x)\right| \le \underset{n\rightarrow \infty }{\lim \sup }\left| f_{\left\{ 1,\ldots ,n\right\} ^{c}}(x)\right| \\\le & {} \underset{n\rightarrow \infty }{\lim \sup }\underset{\underset{i=1}{\overset{n}{\prod }}S_{i}}{\int }Md\left( \underset{i=1}{\overset{n}{ {\displaystyle \bigotimes } }}\mu _{i}\right) =M\underset{i=1}{\overset{\infty }{\prod }}\mu _{i}(S_{i})=0, \end{aligned}$$and so
$$\begin{aligned} \underset{n\rightarrow \infty }{\lim }f_{\left\{ 1,\ldots ,n\right\} ^{c} }(x)=0=\underset{S}{\int }fd\mu . \end{aligned}$$
\(\square \)
Proposition 45
Let \(\left( S,\Sigma \right) \) be a measurable space, let \(\mathcal {I}\) be a \(\pi \)-system on S, and let \(\mu _{1}\) and \(\mu _{2}\) be two measures on \(\left( S,\Sigma \right) \), \(\sigma \)- finite on \(\mathcal {I}\); if \(\sigma (\mathcal {I})=\Sigma \) and \(\mu _{1}\) and \(\mu _{2}\) coincide on \(\mathcal {I}\), then \(\mu _{1}\) and \(\mu _{2}\) coincide on \(\Sigma \).
Proof
See, for example, Theorem 10.3 in Billingsley [10]. \(\square \)
The following definitions and results (from Definition 46 to Definition 53), can be found in [8], and generalize the differentiation theory in the finite-dimensional case (see, e.g., the Lang’s book [14]).
Definition 46
Let \(m,n\in \textbf{N}^{*}\cup \left\{ \infty \right\} \), let \(U\subset \mathcal {D}_{I_{n}}(U)\), let \(x_{0}\in U\), let \(l\in \ell _{I_{m} }\) and let \(\varphi :U\longrightarrow \ell _{I_{m}}\) be a function; we say that \(\lim _{x\rightarrow x_{0}}\varphi (x)=l\) if, for any \(\varepsilon \in \textbf{R}^{+}\), there exists \(\delta \in \textbf{R}^{+}\) such that, for any \(x\in B_{I_{n}}\left( x_{0},\delta \right) \cap \left( U\backslash \{x_{0}\}\right) \), one has \(\left\| \varphi (x)-l\right\| _{I_{m} }<\varepsilon \).
Definition 47
Let \(m,n\in \textbf{N}^{*}\cup \left\{ \infty \right\} \), let \(U\subset \mathcal {D}_{I_{n}}(U)\) and let \(\varphi :U\longrightarrow \ell _{I_{m}}\) be a function; we say that \(\varphi \) is continuous in \(x_{0}\in U\) if \(\lim _{x\rightarrow x_{0}}\varphi (x)=\varphi (x_{0})\), and we say that \(\varphi \) is continuous in U if, for any \(x\in U\), \(\varphi \) is continuous in x.
Remark 48
Let \(m,n\in \textbf{N}^{*}\cup \left\{ \infty \right\} \), let \(U\subset \mathcal {D}_{I_{n}}(U)\), let \(V\subset \ell _{I_{m}}\) and let \(\varphi :U\longrightarrow V\) be a function; then, \(\varphi :\left( U,\tau _{\left\| \cdot \right\| _{I_{n}}}(U)\right) \longrightarrow \left( V,\tau _{\left\| \cdot \right\| _{I_{m}}}(V)\right) \) is continuous if and only if \(\varphi \) is continuous in U.
Definition 49
Let \(m,n\in \textbf{N}^{*}\cup \left\{ \infty \right\} \), let \(U\subset \ell _{I_{n}}\), let \(\varphi :U\longrightarrow \ell _{I_{m}}\) be a function and let \(\left\{ \varphi _{k}\right\} _{k\in \textbf{N}}\) be a sequence of functions such that \(\varphi _{k}:U\longrightarrow \ell _{I_{m}}\), for any \(k\in \textbf{N}\); we say that:
-
1.
The sequence \(\left\{ \varphi _{k}\right\} _{k\in \textbf{N}}\) converges to \(\varphi \) over U if, for any \(\varepsilon \in \textbf{R}^{+}\) and for any \(x\in U\), there exists \(k_{0}\in \textbf{N}\) such that, for any \(k\in \textbf{N}\), \(k\ge k_{0}\), one has \(\left\| \varphi _{k}(x)-\varphi (x)\right\| _{I_{m}}<\varepsilon \).
-
2.
The sequence \(\left\{ \varphi _{k}\right\} _{k\in \textbf{N}}\) converges uniformly to \(\varphi \) over U if, for any \(\varepsilon \in \textbf{R}^{+}\), there exists \(k_{0}\in \textbf{N}\) such that, for any \(k\in \textbf{N}\), \(k\ge k_{0}\), and for any \(x\in U\), one has \(\left\| \varphi _{k}(x)-\varphi (x)\right\| _{I_{m}}<\varepsilon \).
Definition 50
Let \(m,n\in \textbf{N}^{*}\cup \left\{ \infty \right\} \), let \(U\subset \mathcal {D}_{I_{n}}(U)\), let \(v\in \ell _{I_{n}}\) such that \(\left\| v\right\| _{I_{n}}=1\) and let a function \(\varphi :U\longrightarrow \textbf{R}^{m}\); for any \(i\in I_{m}\), the function \(\varphi _{i}\) is called differentiable in \(x_{0}\in U\) in the direction v if there exists the limit
This limit is indicated by \(\frac{\partial \varphi _{i}}{\partial v}(x_{0})\), and it is called derivative of \(\varphi _{i}\) in \(x_{0}\) in the direction v. If, for some \(j\in I_{n}\), one has \(v=e_{j}\), where \((e_{j})_{k}=\delta _{jk}\), for any \(k\in I_{n}\), then the function \(\varphi _{i}\) is called differentiable in \(x_{0}\in U\) with respect to \(x_{j}\); furthermore, indicate \(\frac{\partial \varphi _{i}}{\partial v}(x_{0})\) by \(\frac{\partial \varphi _{i} }{\partial x_{j}}(x_{0})\), and call it partial derivative of \(\varphi _{i}\) in \(x_{0}\), with respect to \(x_{j}\). Moreover, if there exists the linear function defined by the matrix \(J_{\varphi }(x_{0})=\left( \left( J_{\varphi }(x_{0})\right) _{ij}\right) _{i\in I_{m},j\in I_{n}}:\ell _{I_{n} }\longrightarrow \textbf{R}^{m}\), where \(\left( J_{\varphi }(x_{0})\right) _{ij}=\frac{\partial \varphi _{i}}{\partial x_{j}}(x_{0})\), for any \(i\in I_{m} \), \(j\in I_{n}\), then \(J_{\varphi }(x_{0})\) is called Jacobian matrix of the function \(\varphi \) in \(x_{0}\). Finally, if, for any \(x\in U\), there exists \(J_{\varphi }(x)\), then the function \(x\longrightarrow J_{\varphi }(x)\) is indicated by \(J_{\varphi }\).
Definition 51
Let \(n\in \textbf{N}^{*}\cup \left\{ \infty \right\} \), let \(U\subset \mathcal {D}_{I_{n}}(U)\), let \(i,j\in I_{n}\) and let \(\varphi :U\longrightarrow \textbf{R}\) be a function differentiable in \(x_{0}\in U\) with respect to \(x_{i}\), such that the function \(\frac{\partial \varphi }{\partial x_{i}}\) is differentiable in \(x_{0}\) with respect to \(x_{j}\). Indicate \(\frac{\partial }{\partial x_{j}}\left( \frac{\partial \varphi }{\partial x_{i} }\right) (x_{0})\) by \(\frac{\partial ^{2}\varphi }{\partial x_{j}\partial x_{i}}(x_{0})\) and call it second partial derivative of \(\varphi \) in \(x_{0}\) with respect to \(x_{i}\) and \(x_{j}\). If \(i=j\), it is indicated by \(\frac{\partial ^{2}\varphi }{\partial x_{i}^{2}}(x_{0})\). Analogously, for any \(k\in \textbf{N}^{*}\) and for any \(j_{1},\ldots ,j_{k}\in I_{n}\), define \(\frac{\partial ^{k}\varphi }{\partial x_{j_{k}}\ldots \partial x_{j_{1}}}(x_{0})\) and call it k-th partial derivative of \(\varphi \) in \(x_{0}\) with respect to \(x_{j_{1}},\ldots x_{j_{k}}\).
Definition 52
Let \(m,n\in \textbf{N}^{*}\cup \left\{ \infty \right\} \), let \(U\subset \mathcal {D}_{I_{n}}(U)\) and let \(k\in \textbf{N}^{*}\); a function \(\varphi :U\longrightarrow \textbf{R}^{m}\) is called \(C^{k}\) in \(x_{0}\in U\) if, in a neighbourhood \(V\in \tau _{\left\| \cdot \right\| _{I_{n}}}(U)\) of \(x_{0}\), for any \(i\in I_{m}\) and for any \(j_{1},\ldots ,j_{k}\in I_{n}\), there exists the function defined by \(x\longrightarrow \frac{\partial ^{k}\varphi _{i}}{\partial x_{j_{k}}\ldots \partial x_{j_{1}}}(x)\), and this function is continuous in \(x_{0}\); \(\varphi \) is called \(C^{k}\) in U if, for any \(x_{0}\in U\), \(\varphi \) is \(C^{k}\) in \(x_{0}\).
Definition 53
Let \(m,n\in \textbf{N}^{*}\cup \left\{ \infty \right\} \), let \(U\in \tau _{\left\| \cdot \right\| _{I_{n}}}\) and let \(V\in \tau _{\left\| \cdot \right\| _{I_{m}}}\); a function \(\varphi :U\longrightarrow V\) is called diffeomorphism if \(\varphi \) is bijective and \(C^{1}\) in U, and the function \(\varphi ^{-1}:V\longrightarrow U\) is \(C^{1}\) in V.
Henceforth, we will consider the \(\left( m,\sigma \right) \) functions.
Remark 54
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function; then, for any \(\left( x_{j}:j\in \sigma (\textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} )\right) \in \ell _{\sigma (\textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} )}\), one has \(\sup _{i>m}\left| \varphi _{i,\sigma (i)}\left( x_{\sigma (i)}\right) \right| <\infty \).
Proof
From Definition 17, we have \(\varphi (U)\subset \ell _{\infty }\) and \(\varphi ^{(\infty ,m)}(U^{(m)})\subset \ell _{\infty }\); then, \(\forall \,x=\left( x_{j}:j\in \textbf{N}^{*}\right) \in U\), we have
\(\square \)
The following results (from Remark 55 to Proposition 61) can be found in [8].
Remark 55
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function; then, \(\sigma \) is bijective if and only if, for any \(i>m\), one has \(\sigma (i)=i\).
Proposition 56
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function and let \(\left\{ 1,\ldots ,m\right\} \subset L\subset \textbf{N}^{*}\), \(\left\{ 1,\ldots ,m\right\} \subset N\subset \textbf{N}^{*}\); then, one has \(\varphi ^{(L,N)}(\pi _{N}(U))\subset \ell _{L}\), \(\overline{\varphi } ^{(L,N)}(U)\subset \ell _{\infty }\), and the function \(\overline{\varphi } ^{(L,N)}:U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) is \(\left( m,\sigma \right) \). Moreover, if \(\sigma \) is bijective, for any \(n\in \textbf{N}^{*}\), \(n\ge m\), \(\overline{\varphi }^{(L,N)}\) is \(\left( n,\tau \right) \), where the function \(\tau :\textbf{N}^{*}\backslash \left\{ 1,\ldots ,n\right\} \longrightarrow \textbf{N}^{*}\backslash \left\{ 1,\ldots ,n\right\} \) is defined by \(\tau (i)=\sigma (i)=i\), for any \(i>n\).
Proposition 57
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function, and let \(\emptyset \ne L\subset \textbf{N}^{*}\) and \(\emptyset \ne N\subset \textbf{N}^{*}\) such that \(\left\{ 1,\ldots ,m\right\} \subset N\) or \(N\subset \textbf{N}^{*}\backslash \left\{ 1,\ldots ,m\right\} \), and such that, for any \(i\in L\) and for any \(j\in N\backslash \left\{ 1,\ldots ,m\right\} \), the functions \(\varphi _{i}^{(\infty ,m)}:\left( U^{(m)},\mathcal {B}^{m}\left( U^{(m)}\right) \right) \longrightarrow \left( \textbf{R},\mathcal {B}\right) \) and \(\varphi _{ij}:(\textbf{R},\mathcal {B})\longrightarrow \left( \textbf{R},\mathcal {B}\right) \) are measurable; then:
-
1.
The function
$$\begin{aligned} \varphi ^{(L,N)}:(\pi _{N}(U),\mathcal {B}^{N}(\pi _{N}(U)))\longrightarrow \left( \ell _{L},\mathcal {B}^{L}\left( \ell _{L}\right) \right) \end{aligned}$$is measurable; in particular, suppose that, for any \(i\in \textbf{N}^{*}\) and for any \(j>m\), \(\varphi _{i}^{(\infty ,m)}\) and \(\varphi _{ij}\) are measurable functions; then, \(\varphi :(U,\mathcal {B}^{\infty } (U))\longrightarrow \left( \ell _{\infty },\mathcal {B}^{\infty }(\ell _{\infty })\right) \) is measurable.
-
2.
If \(\left\{ 1,\ldots ,m\right\} \subset L\) and \(\left\{ 1,\ldots ,m\right\} \subset N\), then the function \(\overline{\varphi }^{(L,N)}:(U,\mathcal {B} ^{\infty }(U))\longrightarrow \left( \ell _{\infty },\mathcal {B}^{\infty } (\ell _{\infty })\right) \) is measurable.
Corollary 58
Let \(A=\left( a_{ij}\right) _{i,j\in \textbf{N}^{*}}:\ell _{\infty }\longrightarrow \ell _{\infty }\) be a linear function, and let \(\emptyset \ne L\subset \textbf{N}^{*}\) and \(\emptyset \ne N\subset \textbf{N}^{*}\); then:
-
1.
The function \(A^{(L,N)}:(\ell _{N},\mathcal {B}^{N}(\ell _{N} ))\longrightarrow \left( \ell _{L},\mathcal {B}^{L}\left( \ell _{L}\right) \right) \) is measurable; in particular, \(A:(\ell _{\infty },\mathcal {B} ^{\infty }(\ell _{\infty }))\longrightarrow \left( \ell _{\infty },\mathcal {B} ^{\infty }(\ell _{\infty })\right) \) is measurable.
-
2.
If A is \(\left( m,\sigma \right) \), \(\left\{ 1,\ldots ,m\right\} \subset L\) and \(\left\{ 1,\ldots ,m\right\} \subset N\), then the function \(\overline{A}^{(L,N)}:(\ell _{\infty },\mathcal {B}^{\infty }(\ell _{\infty }))\longrightarrow \left( \ell _{\infty },\mathcal {B}^{\infty }(\ell _{\infty })\right) \) is measurable.
Proposition 59
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function such that \(\sigma \) is bijective and \(\overline{\varphi }:\left( U,\tau _{\left\| \cdot \right\| _{\infty }}(U))\longrightarrow (\ell _{\infty },\tau _{\left\| \cdot \right\| _{\infty }}\right) \) is continuous; then, for any \(n\in \textbf{N}^{*}\), \(n\ge m\), \(\varphi ^{(n,n)}:(U^{(m)}\times \textbf{R}^{n-m},\tau ^{n} (U^{(m)}\times \textbf{R}^{n-m}))\longrightarrow (\textbf{R}^{n},\tau ^{n})\) is continuous if and only if \(\overline{\varphi }^{(n,n)}:(U,\tau _{\left\| \cdot \right\| _{\infty }}(U))\longrightarrow (\ell _{\infty },\tau _{\left\| \cdot \right\| _{\infty }})\) is continuous.
Proposition 60
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function; then, \(\overline{\varphi }:U\longrightarrow \ell _{\infty }\) is bijective if and only if the functions \(\varphi ^{(m,m)}:\) \(U^{(m)}\longrightarrow \textbf{R}^{m}\), \(\varphi _{i,\sigma (i)}:\textbf{R}\longrightarrow \) \(\textbf{R}\), for any \(i>m\), and \(\sigma \) are bijective.
Proposition 61
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function. Then, \(\overline{\varphi }:U\longrightarrow \ell _{\infty }\) is a diffeomorphism if and only if the functions \(\varphi ^{(m,m)}:\) \(U^{(m)}\longrightarrow \textbf{R}^{m}\) and \(\varphi _{i,\sigma (i)}:\textbf{R}\longrightarrow \) \(\textbf{R}\), for any \(i>m\), are diffeomorphisms, and \(\sigma \) is bijective.
The following two results (Propositions 62 and 63) can be found in [6].
Proposition 62
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) standard function such that the functions \(\varphi ^{(m,m)}:\) \(U^{(m)}\longrightarrow \textbf{R}^{m}\), \(\varphi _{i,\sigma (i)}:\textbf{R}\longrightarrow \) \(\textbf{R}\), for any \(i>m\), and \(\sigma \) are bijective; then, \(\varphi \) is bijective.
Proposition 63
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) standard function such that \(\varphi _{i}:U\longrightarrow \textbf{R}\) is \(C^{1}\), for any \(i\in \left\{ 1,\ldots ,m\right\} \), and such that \(\overline{\varphi }:U\longrightarrow \overline{\varphi }(U)\) is a diffeomorphism; then, for any \(n\in \textbf{N} ^{*}\), \(n\ge m\), the functions \(\varphi ^{(n,n)}:U^{(m)}\times \textbf{R}^{n-m}\longrightarrow \varphi ^{(n,n)}\left( U^{(m)}\times \textbf{R}^{n-m}\right) \) and \(\overline{\varphi }^{(n,n)}:U\longrightarrow \overline{\varphi }^{(n,n)}\left( U\right) \) are diffeomorphisms.
Analogously, it is easy to prove the following results (from Propositions 64–67).
Proposition 64
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow E_{\infty }\) be a \(\left( m,\sigma \right) \) function and let \(\left\{ 1,\ldots ,m\right\} \subset L\subset \textbf{N}^{*}\), \(\left\{ 1,\ldots ,m\right\} \subset N\subset \textbf{N}^{*}\); then, one has \(\overline{\varphi }^{(L,N)}(U)\subset E_{\infty }\), and the function \(\overline{\varphi }^{(L,N)}:U\subset \ell _{\infty }\longrightarrow E_{\infty }\) is \(\left( m,\sigma \right) \). Moreover, if \(\sigma \) is bijective, for any \(n\in \textbf{N}^{*}\), \(n\ge m\), \(\overline{\varphi }^{(L,N)}\) is \(\left( n,\tau \right) \), where the function \(\tau :\textbf{N}^{*}\backslash \left\{ 1,\ldots ,n\right\} \longrightarrow \textbf{N}^{*}\backslash \left\{ 1,\ldots ,n\right\} \) is defined by \(\tau (i)=\sigma (i)=i\), for any \(i>n\).
Proposition 65
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow E_{\infty }\) be a \(\left( m,\sigma \right) \) function; then, \(\overline{\varphi }:U\longrightarrow E_{\infty }\) is bijective if and only if the functions \(\varphi ^{(m,m)}:\) \(U^{(m)}\longrightarrow \textbf{R}^{m}\), \(\varphi _{i,\sigma (i)}:\textbf{R}\longrightarrow \) \(\textbf{R}\), for any \(i>m\), and \(\sigma \) are bijective.
Proposition 66
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow E_{\infty }\) be a \(\left( m,\sigma \right) \) function. Then, \(\overline{\varphi }:U\longrightarrow E_{\infty }\) is a diffeomorphism if and only if the functions \(\varphi ^{(m,m)}:\) \(U^{(m)}\longrightarrow \textbf{R}^{m}\) and \(\varphi _{i,\sigma (i)}:\textbf{R}\longrightarrow \) \(\textbf{R}\), for any \(i>m\), are diffeomorphisms, and \(\sigma \) is bijective.
Proposition 67
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow E_{\infty }\) be a \(\left( m,\sigma \right) \) standard function such that the functions \(\varphi ^{(m,m)}:\) \(U^{(m)}\longrightarrow \textbf{R}^{m}\), \(\varphi _{i,\sigma (i)}:\textbf{R}\longrightarrow \) \(\textbf{R}\), for any \(i>m\), and \(\sigma \) are bijective; then, \(\varphi \) is bijective.
The following two results (Propositions 68 and 69) can be found in [7].
Proposition 68
Let \(A=\left( a_{ij}\right) _{i,j\in \textbf{N}^{*}}:\ell _{\infty }\longrightarrow \ell _{\infty }\) be a linear \(\left( m,\sigma \right) \) function, such that A is \(\left( m,\sigma \right) \) standard or \(\left( m,\sigma \right) \) of the second type. Then, if \(\sigma \) is bijective, we have
Conversely, if \(\sigma \) is not bijective, we have \(\det A=0\). In particular, if \(A=\textbf{I}_{n}\), we have \(\det A=1\).
Proposition 69
Let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function and let \(x_{0} =(x_{0,j}:j\in \textbf{N}^{*})\in U\) such that there exists the linear function \(J_{\varphi }\left( x_{0}\right) :\ell _{\infty }\longrightarrow \ell _{\infty }\); then, \(J_{\varphi }\left( x_{0}\right) \) is \(\left( m,\sigma \right) \); moreover, for any \(n\in \textbf{N}^{*}\), \(n\ge m\), there exists the linear \(\left( m,\sigma \right) \) function \(J_{\overline{\varphi }^{(n,n)}}\left( x_{0}\right) :\ell _{\infty }\longrightarrow \ell _{\infty }\), and one has
The following two results (Propositions 70 and 71) can be found in [8].
Proposition 70
Let \(m\in \textbf{N}\), let \(n\in \textbf{N}^{*}\), \(n\ge m\), and let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function such that, for any \(i\in \left\{ 1,\ldots ,n\right\} \), for any \(j_{1}\in \left\{ 1,\ldots ,m\right\} \) and for any \(j_{2}\in \left\{ m+1,\ldots ,n\right\} \), there exist the functions \(\frac{\partial \varphi _{i}^{(\infty ,m)}}{\partial x_{j_{1}}}:\left( U^{(m)},\mathcal {B}^{m}\left( U^{(m)}\right) \right) \longrightarrow \left( \textbf{R},\mathcal {B}\right) \) and \(\frac{\partial \varphi _{ij_{2}}}{\partial x_{j_{2}}}:(\textbf{R},\mathcal {B})\longrightarrow \left( \textbf{R},\mathcal {B}\right) \), and they are measurable; then:
-
1.
The function \(\det J_{\varphi ^{(n,n)}}:\left( U^{(m)}\times \textbf{R}^{n-m},\mathcal {B}^{n}(U^{(m)}\times \textbf{R}^{n-m})\right) \longrightarrow \left( \textbf{R},\mathcal {B}\right) \) is measurable.
-
2.
Suppose that, for any \(i>m\), the function \(\varphi _{i,\sigma (i)} ^{\prime }:(\textbf{R},\mathcal {B})\longrightarrow \left( \textbf{R},\mathcal {B}\right) \) is measurable; then, for any \(x\in U\), there exists the function \(J_{\overline{\varphi }^{(n,n)}}(x):\ell _{\infty }\longrightarrow \ell _{\infty }\), and it is \(\left( m,\sigma \right) \); moreover, the function \(\det J_{\overline{\varphi }^{(n,n)}}:\left( U,\mathcal {B}^{\infty }(U)\right) \longrightarrow \left( \textbf{R},\mathcal {B}\right) \) is measurable.
-
3.
Suppose that, for any \(x\in U\), there exists the function \(J_{\varphi }\left( x\right) :\ell _{\infty }\longrightarrow \ell _{\infty }\); moreover, suppose that, for any \(i\in \textbf{N}^{*}\), for any \(j_{1}\in \left\{ 1,\ldots ,m\right\} \) and for any \(j_{2}>m\), the functions
$$\begin{aligned} \frac{\partial \varphi _{i}^{(\infty ,m)}}{\partial x_{j_{1}}}:\left( U^{(m)},\mathcal {B}^{m}\left( U^{(m)}\right) \right) \longrightarrow \left( \textbf{R},\mathcal {B}\right) \end{aligned}$$and \(\frac{\partial \varphi _{ij_{2}}}{\partial x_{j_{2}}}:(\textbf{R},\mathcal {B})\longrightarrow \left( \textbf{R},\mathcal {B}\right) \) are measurable; then the function \(\det J_{\varphi }:\left( U,\mathcal {B}^{\infty }(U)\right) \longrightarrow \left( \textbf{R},\mathcal {B}\right) \) is measurable.
Proposition 71
Let \(m\in \textbf{N}\), let \(n\in \textbf{N}^{*}\), \(n\ge m\), and let \(\varphi :U\subset \ell _{\infty }\longrightarrow \ell _{\infty }\) be a \(\left( m,\sigma \right) \) function such that \(\varphi ^{(n,n)}\) is \(C^{1}\); then, the function \(\det J_{\varphi ^{(n,n)}}:\left( U^{(m)}\times \textbf{R}^{n-m},\tau ^{n}(U^{(m)}\times \textbf{R}^{n-m})\right) \longrightarrow \left( \textbf{R},\tau \right) \) is continuous.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Asci, C. Infinite-dimensional Gaussian change of variables’ formula. Ann Univ Ferrara (2024). https://doi.org/10.1007/s11565-024-00490-z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s11565-024-00490-z
Keywords
- Infinite-dimensional Banach spaces
- Infinite-dimensional differentiation theory
- \((m, \sigma )\) functions
- Gaussian change of variables’ formula