Translationally invariant measure on the infinite dimensional vector space (Q1159780)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Translationally invariant measure on the infinite dimensional vector space |
scientific article |
Statements
Translationally invariant measure on the infinite dimensional vector space (English)
0 references
1980
0 references
On an infinite dimensional vector space, any measure can not be invariant nor quasi-invariant under all translations. However, for some dense subspace \(E_1\), there exist such measures that are \(E_1\)-quasi-invariant, i.e. quasi-invariant under the translations: \(x\to x+\xi\) where \(\xi\in E_1\). For instance, Gaussian measure \(g\) on \(R^\infty\) is \((\ell^2)\)-quasi-invariant. We have many other quasi-invariant measures, but the known examples including the measure \(g\) above have no equivalent invariant measure. Historically the translational quasi-invariance of probability measures has been discussed in detail, while the study of translational invariance has been neglected, perhaps because of difficulties of infinite measures which have less connection with the theory of probability. In this paper, we shall construct directly a \(\sigma\)-finite invariant measure. First we consider the product of Lebesgue measure and uniform probability measures on \(R^n \times [ -1/2,1/2]^\infty\), and in the limit of \(n\to\infty\) we obtain the requested measure \(\mu\) on \(R^\infty\). We can easily show that \(\mu\) is \(R_0^\infty\)-invariant, and a detailed study shows that \(\mu\) is \((\ell^1)\)-invariant. We can modify \(\mu\) to get an \((\ell^2)\)-invariant measure, and other modifications can give invariant measures with respect to a larger subspace of \(R^\infty\). The measure \(\mu\) thus constructed is translationally ergodic, rotationally invariant, and rotationally ergodic. It is singular with Gaussian measures, so that the uniqueness of rotationally invariant measure does not hold for infinite measures. Finally we shall prove the existence of a measure which is invariant both under translations and homotheties. This fact, which is false for the finite dimensional case, arises from the singularity of the infinite dimensional Lebesgue measure with respect to homotheties. Similar studies from somewhat different standpoints are: \textit{D. G. B. Hill} [Trans. Am. Math. Soc. 153, 347--370 (1971; Zbl 0212.08303)] and \textit{T. Hamachi} [Mem. Fac. Sci. Kyushu Univ., Ser. A 27, 335--341 (1973; Zbl 0286.60002)].
0 references
translational invariance
0 references
Gaussian measures
0 references