English

Non-Commutative Integration

Operator Algebras 2012-08-28 v1 Functional Analysis

Abstract

We will show that if \sM\sM is a factor, then for any pair \f,\p\sMdsup\f, \p\in\sMdsup of normal positive linear functionals on \sM\sM, the inequality: \lrnorm\f\lrnorm\p \lrnorm{\f}\leq \lrnorm{\p} is equivalent to the fact that there exist a countable family \lrbrace\ffdi:iI\sMdsup\lrbrace{\ffdi: i\in I}\subset \sMdsup in \sMdsup\sMdsup and a family \lrbrace\udi:iI\i\sM\lrbrace{\udi: i\in I}\i\sM of partial isometries in \cM such that \f=\sumdiI\ffdi,\sumdiI\udi\ffdi\udius\p,and\udius\udi=s\lr\ffdi,iI, \f=\sumd{i\in I} \ffdi,\quad \sumd{i\in I} \udi{\ffdi}\udius\leq \p, \quad \text{and} \quad \udius\udi=s\lr{\ffdi}, i\in I, where s(ω),ω\sMdsups(\omega), \omega\in\sMdsup, means the support projection of ω\omega. Furthermore, if \lrnorm\f=\lrnorm\p\lrnorm{\f}=\lrnorm{\p}, then the equality replaces the inequality in the second statement. In the case that \sM\sM is not of type \threeonec the family of partial isometries can be replaced by a family of unitaries in \cMp One cannot expect to have this result in the usual integration thoery. To have a similar result, one needs to bring in some kind of non-commutativity. Let \lrbraceX,μ\lrbrace{X, \mu} be a \sig\sig-finite semifinite measure space and GG be an ergodic group of automorphisms of \linflrX,μ\linflr{X, \mu}, then for a pair ff and gg of μ\mu-integrable positive functions on XX, the inequality: Xf(x)\txdμ(x)Xg(x)\txdμ(x) \int_X f(x)\txd \mu(x)\leq \int_X g(x)\txd \mu(x) is equivalent to the existence of a countable families \lrbrace\fdi:iIL1(X,μ)\lrbrace{\fdi: i\in I}\subset L^1(X, \mu) of positive integrable functions and \lrbrace\gdi:iI\lrbrace{\gdi: i\in I} in GG such that f=\sumdiI\fdiand\sumdiI\gdi\lr\fdig, f=\sumd{i\in I} \fdi\quad\text{and}\quad \sumd{i\in I} \gdi\lr{\fdi}\leq g, where the summation and inequality are all taken in the oredered Banach space L1(X,μ)L^1(X, \mu) and the action of GG on \lonelrX,μ\lonelr{X, \mu} is defined through the duality between \linflrX,μ\linflr{X, \mu} and \lonelrX,μ\lonelr{X, \mu}, i.e., \lr{\g(f)}(x)&=f\lr{\g\inv x}\frac{\txd\mu\scirc \g\inv}{\txd\mu}(x), \quad f\in\lonelr{X, \mu}.

Keywords

Cite

@article{arxiv.1208.5197,
  title  = {Non-Commutative Integration},
  author = {Masamichi Takesaki},
  journal= {arXiv preprint arXiv:1208.5197},
  year   = {2012}
}
R2 v1 2026-06-21T21:55:21.688Z