Noncommutative Lévy processes for generalized (particularly anyon) statistics (Q443968): Difference between revisions

Let \(T\) be a locally compact Polish space, \(\sigma\) a nonatomic measure on \(T\), \(D \subset T ^2\) the diagonal, \(A \supset D\) be measurable, \((\sigma\otimes\sigma) (A)=0\) and symmetric, \(T ^{(2)}=T^2 \backslash A\), \(Q:T^{(2)} \rightarrow S^1=\{| \cdot|=1\}\) be measurable, \(Q(t,s)=\overline{Q(s,t)}\), \((\Psi f) (s,t)=Q(s,t)f(t,s)\), \(T^{(n)}=\bigcap _{i \neq j}pr_{i,j}^{-1}(A) \subset T^n\), for \(f:T^{(n)} \rightarrow \mathbb C\), \(( \Psi_jf) (t_1,\dots,t _n)=Q(t_j,t_{j+1})f(t_1,\dots,t_{j+1},t_{j},\dots,t_n)\), \(P_n=n!^{-1} \sum_{\pi \in S _n}\Psi _\pi\) where \(\pi \rightarrow \Psi _\pi\) is the homomorphism extending \((j,j+1) \rightarrow \Psi _j\), \(H _\mathbb C=L _\mathbb C^2 (\sigma )\), \(H_\mathbb C^\odot n=P_n (H_\mathbb C^\otimes n)\), \(\mathcal F^Q(H)=\bigoplus _{n \geq 0} H _\mathbb C^{\bigodot n} n !\) (the \(Q\)-Fock space, \(\| \bigoplus _{n \geq 0 } u _{n } \| ^{2 } = \sum _{n \geq 0 } n ! \| u _{n } \| ^{2 }\)), \(\Omega = 1 \oplus 0 \oplus \dots \), \(a ^{+ } ( h ) = h\odot \cdot\), \(a ^{-} ( h ) = ( a ^{+ } ( h ) )^{ \ast }\) for \(h \in H _{{\mathbb C}}\) and, for a bounded \(h\), \(a^{0 } ( h ) f ( t _{1 }, \dots , t _{n } ) = (h(t_{1})+\dots+h(t_{n}))f(t_{1},\dots,t_{n})\), all three on \({\mathcal F} _{\mathrm {f i n} }^{Q } ( H )\) (topological direct sum of \(H _{{\mathbb C}} ^{\odot n } n !\)). Fix \(\lambda \in {\mathbb R}\) and let \(( h,\omega ) = a^{+ } ( h ) + \lambda a^{0 } ( h ) + a^{-} ( h )\), \({\mathcal P}\) be the \( \ast \)-algebra generated by all \(( h , \omega )\) and \(1\), \(\tau ( p ) = \langle p \Omega , \Omega \rangle \) for \(p \in {\mathcal P}\): \(p \rightarrow p \Omega\) follows a unitary \(L ^{2 } ( \tau ) \longleftrightarrow {\mathcal F}^{Q} ( H )\). The authors introduce the notations \(\partial _{t }\), \(\partial _{t }^{+ }\) and give a sense to \(a^{+ } ( h ) = \int h ( t ) \partial _{t }^{+ } d \sigma ( t )\), \(a^{-} ( h ) = \int \overline {h ( t )} \partial _{t} d \sigma ( t )\), \(a^{0 } ( h ) = \int h ( t ) \partial _{t }^{+ } \partial _{t } d \sigma ( t )\), denote \(\omega ( t ) = \partial _{t }^{+ } + \lambda \partial _{t }^{+ } \partial _{t } + \partial _{t }\) and define a Wick ordering \(: \cdot : _{W }\) which consists in replacing, in the product, each \(\partial _{s} \partial _{t }^{+}\) by \(Q ( s , t ) \partial _{t }^{+} \partial _{s }\) until all \(a^{+ }\) are to the left of all \(a^{-}\). In this symbolism, they introduce also \(\delta ( t _{1 } , \dots, t _{n })\) as \(\int f ( t _{1 }, \dots , t _{n }) \delta ( t _{1 }, \dots , t _{n }) d \sigma ( t _{1 }) \cdots d \sigma ( t _{n }) = \int f ( t , \dots, t) d \sigma ( t )\). The first result appearing as a theorem is Theorem 4.5: If \(Q ( t , s ) \in \{ \pm 1 \} \) and \(\lambda = 0\) then \({:} \cdot {:} _{W }\) is the usual \({:} \cdot{:}\). The second (4.7) is \(\omega (t _1) \cdots \omega (t _n) = \sum _\mathcal V Q ({\mathcal V}, t _{1 }, \dots , t _n) {:} \omega ( t _{1 }) \cdots \omega (t_n){:}_\mathcal V\) over all partitions \({\mathcal V}\) of \( \{ 1, \dots , n \} \), each element of \({\mathcal V}\) being marked \(+ 1\) or \(-1\), all singletons in \({\mathcal V}\) by \(+1\), \({:}\omega ( t _{1 }) \cdots \omega ( t _n){:}_{{\mathcal V}}\) being obtained from \({:} \omega ( t _{1 }) \cdots \omega ( t _{n }) {:}\) by removing all \(\omega ( t _{i })\) with \(i\) not in a singleton of \({\mathcal V}\) and multiplying, for every \(B = \{ t_{i _{1}},\dots,t_{i _{k}} \} \in {\mathcal V}\), \(k \geq 2\), by \(\delta ( t _{1 _{1 }}, \dots , t _{i _{k }})\) and also by \(\lambda ^{k - 1 }\) if \(B\) is marked by \(+ 1\) and by \(\lambda ^{k - 2 }\) if \(B\) is marked by \(-1\), while \(Q ( {\mathcal V} , t _{1 } , \dots , t _{n } )\) is the product of all \(Q ( t _{\min B _{2 }}, t _{\max B _{1 }} )\), for all pairs \(B _{1 } , B _{2 } \in {\mathcal V}\), for which \(B _{2 }\) is marked \(-1\) and either \(B _{1 }\) is marked with \(-1\) and \({\min B _{1 }} < {\min B _{2 }} < \max B _{1 } < \max B _{2 }\) or \(B _{1 }\) is marked \(+ 1\) and \(\min B _{2 } < \max B _{1 } < \max B _{2 }\). The independence is defined in a more general setup, namely having a linear subspace \({\mathcal D}\) of a separable complex Hilbert space \({\mathcal F}\) and an injective linear map \(f \rightarrow ( f, \xi )\) from the bounded measurable \(f\) to symmetric operators on \({\mathcal D}\), a \(\Psi \in {\mathcal D}\), \(\| \Psi \| = 1\) and complex measures \(m _{n }\) on \(T ^{n }\) such that \(\langle \Pi _{i = 1 }^{n } ( f _{i } , \xi ) \Psi , \Psi \rangle = \int \prod _{i = 1 }^{n } ( f _{i } \circ p r _{i } ) d m _{n }\). The cumulants \(C _{n } ( ( f _{1 }, \xi ), \dots , ( f _{n } , \xi ) )\) are defined as \(\int \prod _{i = 1 } ^{n } ( f _{i } \circ p r _{i } ) d c _{n }\) where \(c _{n }\) are complex measures defined recursively by \(c _{1 } = m _{1 }\), \(m _{n } = \sum _{{\mathcal V} } Q ( {\mathcal V}, \cdot ) d c _{{\mathcal V} }\), \(c _{{\mathcal V} }\) being the product of \(c _{B }\), \(B \in {\mathcal V}\), where, for \(B = \{ i _{1 }, \dots, i _{k } \} \), \(c _{B }\) is \(c _{k }\) on \(T ^{k } = \prod _{j = 1 } ^{k } T _{i _{j}}\) etc. \(( f _{i } , \xi )\), \(i = 1 , \dots, n\), are said to be independent if \(C _{k } ( ( f _{i _{1 }} , \xi ), \dots , ( f _{i _{k }} , \xi ) ) = 0\) for all \(k \geq 2\) and the indices \(i _{j }\) not all equal. The \(( f, \omega )\), \(f\) bounded measurable, introduced before, are independent. The family of all \(( f, \xi )\) is said to be a \(Q\)-Lévy process if \(( \chi _{\Delta _{i }} , \xi )\), \(i = 1, \dots, n\), are independent when the \(\Delta _{i }\) are disjoint and if \(\sigma ( \Delta _{1 } ) = \sigma ( \Delta _{2 } )\) implies \(\langle ( \chi _{\Delta _{1 }} , \xi )^{k } \Psi , \Psi \rangle = \langle ( \chi _{\Delta _{2 }} , \xi ) ^{k } \Psi , \Psi \rangle \) for all \(k \in {\mathbb N}\). The following general example is considered. \(\nu\) is a probability on \({\mathbb R}\), \(\int e ^{\varepsilon | x | } d \nu ( x ) < + \infty\) for some \(\varepsilon > 0\), \(Q ( t _{1 }, x _{1 , t{2 }}, x _{2 }) = Q ( t _{1 }, t _{2 } )\), \(( t _{1 }, t _{2 } ) \in T ^{( 2 ) }\), \(( x _{1 }, x _{2 } ) \in {\mathbb R}^{2 }\), \({\mathcal G} = L^{2 } ( \sigma \otimes \nu )\), \({\mathcal D} \subset {\mathcal F}^{Q } ( {\mathcal G} )\) being the set of all \(F \in {\mathcal F} _{\mathrm {fin}} ^{Q } ( {\mathcal G} )\), with the component \(k\) in \(P _{k } ( H _{{\mathbb C}}^{\otimes k } \otimes ' p _{k } )\), \(p _{k }\) being the space of polynomials in \(k\) variables \(x _{1 } , \dots, x _{n }\). Set \(( f , \xi ) = a^{+ } ( f \otimes 1 ) + a ^{0 } ( f \otimes x ) + a^{-} ( f \otimes 1 )\), \(f\) bounded measurable on \(T\), \(x\) interpreted as the polynomial. These constitute a \(Q\)-Lévy process. Defining \(X _{k } ( f ) = a^{+ } ( f \otimes x^{k - 1 } ) + a^{0 } ( f \otimes x^{k } ) + a^{-} ( f \otimes x^{k - 1 } )\), \(k \in {\mathbb N}\), and \(Y _{k }\) as the result of orthogonalizing this sequence, the paper finishes with an orthogonal representation of \(L ^{2 } ( \tau )\), the terms being indexed by finite sequences \(\alpha = ( \alpha _{0 },\dots , \alpha _{k } )\) of positive integers, the \(\alpha\)-term being an integral of functions in tensor products of \(H _{{\mathbb C}} ^{^{\odot \alpha } i }\) with respect to a product of \(\alpha _{0 }\) factors \(d Y _0\), \(\alpha_1\) factors \(d Y_1,\dots\), \(\alpha_k\) factors \(d Y_k\). The paper contains a lot of propositions necessary for working with the introduced concepts. It starts with three pages of history and motivations. The (anyon) case when \(T\) has a total ordering for pairs in \(T^{( 2 ) }\) and \(Q ( s , t ) = q\) for \(s < t\), \(\overline {q}\) for \(s > t\), is mentioned as having importance in physics.