The optional stopping theorem for quantum martingales (Q2504362)

Consider the symmetric Fock space \(\Phi =\Gamma (L^{2}([0,\infty )))\), generated by the exponential elements \(e(f)\), the linear hull \({\mathcal E}\) of all \(e(f)\), \(f\in L^{2}([0,\infty ))\), \(\Phi =\Phi_{t]}\otimes \Phi_{[t}\) for all \(t\), where \(\Phi_{t]}\) is \(\Gamma (L^{2}([0,t]))\), \(1=e(0)\), \({\mathcal U}=L(\Phi )\), the set \({\mathcal U}_{t}\) of all \(T\otimes 1_{\Phi_{[t}}\) in \({\mathcal U}\), \(E_{t}:\Phi \rightarrow \Phi_{t]}\) be the projection, \(M_{t}X=E_{t}XE_{t}\otimes 1_{\Phi_{[t}}\). A stopping time \(T\) is a spectral measure on \([0,\infty ]\) such that \(T([0,t])\in {\mathcal U}_{t}\). \(E_{T}\) is defined as the projector on the set \(\Phi_{T}\) of all \(x\in \Phi\) with \(T([0,t])x\in \Phi_{t]}\otimes 1_{[t}\) for all \(t\). The author proves first that it is impossible to attach to every stopping time \(T\) an algebra \({\mathcal U}_{T}\subset {\mathcal U}\), such that \(XE_{T}=E_{T}XE_{T}\) for all \(X\in {\mathcal U}_{T}\), that \(S\leq T\) (as operators) imply \({\mathcal U}_{S}\subset {\mathcal U}_{T}\) and that \({\mathcal U}_{T}={\mathcal U}_{t}\) when \(T=t\), thus indicating the problem that one has to generalize the theory of stopping times to the noncommutative case. Let also, for \(s<t\), \(E_{s,t}=1_{\Phi_{s]}}\otimes P_{1_{[s,t]}}\otimes 1_{\Phi_{[t}}\) and \(EX=\langle X1,1\rangle\). If \(T\) is a stopping time with \(T(N)=1\) for a finite \(N\), then \(X_{T}\) is defined as \[ \sum_{s,t\in N}E_{[s,\max(s,t)]}T(\{s\})(M_{\max(s,t)}X)T(\{t\})E_{t,\max(s,t)}, \] even for a bounded \(X_{t}\) with \(X_{\max(s,t)}\) instead of \(M_{\max(s,t)}X\). If \(X\in {\mathcal U}\), \(T\) is a stopping time, \(x\in {\mathcal E}\), then the limit of \(X_{T_{N}}x\), where \(N=\{0=t_{0}<\cdots<t_{n}<\infty \}\), with \(T(\{t\})=0\) for \(t\in N\{0\}\), \(T_{N}(\{t_{i}\})=T([t_{i-1},t_{i}])\), \(T_{N}(\{+\infty \})=T([t_{n},+\infty ])\), when \(t_{n}\rightarrow +\infty\), \(\max(t_{i}-t_{i-1})\rightarrow 0\), exists and and defines an \(M_{T}(X)\in {\mathcal U}\) (in other words: \(X\) can be stopped), expressed as \(E_{T}XE_{T}+\int T([0,s])M_{s}(E_{T}XE_{T})T([0,s])\,da_{s}^{0}\), the same holding for \(M_{s}(E_{T})M_{s}(X)M_{s}(E_{T})\) instead of \(M_{s}(E_{T}XE_{T})\) where \(a_{s}^{0}(e(f))= e(1_{[0,s]}f)\). Properties: for \(S\leq T\) we have \(M_{S}(X_{T})=X_{S}\); \(E(X_{T})=E(X_{0})\) and, if for a bounded adapted \(X_{t}\) this is true for all \(T\) with \(T=T(N)\) for some finite \(N\), then \[ X_{s}=M_{s}X_{t} \quad \text{for }s<t; \] \[ M_{t}E_{T}=1 -\int_0^tT([0,s])M_{s}(E_{T})\,da_{s}^{0}= 1+ \int_0^t(T((s,+\infty ])-M_{s}(E_{T}))\,da_{s}^{0}; \] \[ M_{t}(E_{T})\int_0^tT([0,s])M_{s}(E_{T})H_{s}M_{s}(E_{T})T([0,s])\, da_{s}^{0}=0 \] for a bounded \(H\), leading to the remark that \(M_{T}\) does not have the property \[ M_{T}(M_{T}(X)Y))= M_{T}(X)M_{T}(Y) \] of a conditional mean value; using a similar expression, a map \(M_{T}^{\sim }\) on the set of all bounded \((Z_{t})\) with \(Z_{t}\in {\mathcal U}_{t}\), \(T([0,t])Z_{t}T([0,t])=Z_{t}\), \(Z_{t}\rightarrow Z_{\infty}\) strongly for \(t\rightarrow +\infty\) is obtained having that property. The existence and expression of the stopped process is also obtained for a \(X_{t}=\int_0^t H_{s}\,ds+E_{t}(M)\) and for \(X_{t}=\sum_{\varepsilon }\int_0^t H_{s}^{\varepsilon }\,da_{s}^{\varepsilon }\), \(\varepsilon =0,+,-,x\), \(a_{s}^{x}=s\cdot 1\), where \(a_{s}^{\pm }\) are the creation and annihilation operators corresponding to \(1_{[0,s]}\), \(H_{s}^{\varepsilon }\in {\mathcal U}_{s}\), \(H^{0}\) is bounded, \(\| H^{\pm }\|\) is in \(L^{2}\), \(\| H^{x}\|\) in \(L^{1}\). Let \(R_{t}e(f)=e(f1_{(t,\infty )})\); it satisfies \(M_{s}R_{t}=R_{s}\) for \(s<t\). The author considers then the closure \(\Phi^{T}\) of \(\Im R_{T}\), the subspace \(\Phi_{T}^{c}\) of \(L^{\infty }([0,\infty),\Phi )\) consisting of all \((x_{t})\) with \(x_{t}\in \Phi_{\min(T,t)}\), \(T([0,t])x_{t}=x_{t}\) and \(x_{t}\rightarrow x_{\infty }\) for \(t\rightarrow +\infty\) and, for \(x=(x_{t})\in \Phi_{T}^{c}\), the maps \(L_{s}(x)\in {\mathcal U}_{s}\) defined by \[ L_{s}(x)E_{s}F= \langle T([0,s])1,F\rangle\| T([0,s])1\|^{- 1}X_{s}. \] She proves that \(x\otimes R_{T}F\rightarrow \langle 1,R_{T}F\rangle X_{\infty }+ (\int L_{s}(x)\,da_{s}^{0})F\) extends to a linear contraction \(J:\Phi_{T}^{c}\otimes \Phi^{T}\rightarrow \Phi\). The closure of \(\Im J\) is \(\Phi\) if \(T([0,t])1=0\) implies \(T([0,t])=0\), generally this closure contains \(T([d,\infty ])\Phi\), where \(d\) is the maximal satisfying \(T([0,d))1=0\). The following examples are studied: (1) \(\Phi\) represented as \(L^{2}(P)\), \(P\) being the set of all finite subsets of \([0,\infty )\), \(T(\sigma )\) being the first \(t\) with \(\text{card}(\sigma \cap [0,t])>n\). (2) First jumping time of \(a_{t}^{+}+a_{t}^{-}+a_{t}^{0}+t\cdot 1\) and stopping time of \(a_{t}^{+}(f)+a_{t}^{-}(f)\), (3) Stopping \(t(a_{t}^{+}+a_{t}^{-})\). (4) If \(T\) is a stopping time, \(T((\cdot,\infty ))\) stopped at a stopping time \(S\) with ``\(S<T\)''.