Physical symmetries of quantum histories (Q1302005)

This interesting paper continues the author's investigation into various aspects of the quantum history approach to statistical physical systems. A decoherence functional \(d\) on the \(\text{W}^*\)-algebra \(A\) is a complex-valued function \((p,q)\to d(p,q)\) from the set \({\mathcal P}(A)\times{\mathcal P}(A)\) of pairs of projections in \(A\) such that: (i) for \(p\) and \(q\) in \({\mathcal P}(A)\), \(d(p,q)= \overline{d(q,p)}\); (ii) for \(p_1\) and \(p_2\) in \({\mathcal P}(A)\), such that \(p_1p_2= 0\) and all \(q\) in \({\mathcal P}(A)\), \(d(p_1+ p_2,q)= d(p_1,q)+ d(p_2,q)\); (iii) for all \(p\) in \({\mathcal P}(A)\), \(d(p,p)\geq 0\); (iv) \(d(1,1)= 1\). Observe that, for each state \(x\) of \(A\), the function \(d_x\), defined for \(p\) and \(q\) in \({\mathcal P}(A)\), by \[ d_x(p,q)= x(pq), \] is a decoherence functional. A pair \((A,D)\) consisting of a \(\text{W}^*\)-algebra \(A\) and a family \(D\) of decoherence functionals for which there exists a separating set \(S\) of states of \(A\) such that, for \(x\) in \(S\), \(d_x\) lies in \(D\), is said to be a full quantum history system. A physical symmetry \(\phi\) of \((A,D)\) is a bijection from \({\mathcal P}(A)\) onto itself, for which, for each \(d\) in \(D\), the mapping \((p,q)\to d(\phi(p),\phi(q))\) is a decoherence functional. The first main result shows that every physical symmetry extends to a Jordan *-automorphism of \(A\), provided that \(A\) has no weak\(^*\)-closed ideal of type \(I_2\). In general, the Bunce and Wright form of Gleason's theorem shows that a decoherence functional \(d\) extends to a bounded sesquilinear functional \(q_d\) on \(A\). Let \(\text{sym}(d)\) denote the group of Jordan *-automorphisms of \(A\) that leave \(q_d\) invariant, and let \(G\) be a subgroup of \(\text{sym}(d)\). Then \(d\) is said to be \(G\)-representable if there exists a Hilbert space \(H\), a bounded linear operator \(\rho\) from \(A\) to \(H\) and a symmetry \(2P- 1_H\) in \(B(H)\) such that, for all \(a\) and \(b\) in \(A\), \[ q_d(a,b)= \langle(2P- 1_H)\rho(a),\rho(b)\rangle, \] and, for each \(\phi\) in \(G\), there exists a unitary operator \(U_\phi\) on \(H\), commuting with \(P\), such that, for each \(a\) in \(A\), \[ U_\phi\rho(a)= \rho(\phi(a)). \] The second main result uses the Haagerup-Pisier-Grothendieck inequality to show that, for every \(G\)-invariant state \(x\) of \(A\) such that, for some constant \(M\), and all \(a\) and \(b\) in \(A\), \[ |q_d(a,b)|\leq Mx(a^*a+ aa^*)^{{1\over 2}} x(b^*b+ bb^*)^{{1\over 2}}, \] it follows that \(d\) is \(G\)-representable.