Kernel maps and operator decomposition (Q2175547)

Let \(\mathcal{H}\) be a separable complex Hilbert space and \(\mathcal{L}\) a lattice of orthogonal projections onto subspaces of \(\mathcal{H}\) that is strongly closed and contains \(0\) and \(I\). For \(P\) in \(\mathcal{L}\), define \( P_-=\vee\{Q\in\mathcal{L}:Q<P\}\). A subspace lattice \(\mathcal{L}\) that is totally ordered under range inclusion is called a nest. A~nest is continuous if, for all \(P\) in \(\mathcal{L}\), we have \(P=P_-\). For a nest \(\mathcal{L}\), the corresponding nest algebra, denoted \(\mathcal{T(L)}\), is the set of all bounded linear operators \(T\) on \(\mathcal{H}\) for which \(P^\perp TP=0\) for all \(P\) in \(\mathcal{L}\). A subset \(\mathcal{M}\) of bounded linear operators on \(\mathcal{H}\) is called a Lie \(\mathcal{T(L)}\) module provided that, for all \(M\in\mathcal{M}\) and all \(T\in\mathcal{T(L)}\), we have \(MT-TM\in\mathcal{M}\). The main result of the paper under review is as follows. Theorem 3.1. If \(\mathcal{L}\) is a continuous nest and \(\mathcal{M}\) is a norm closed Lie \(\mathcal{T(L)}\) module, then \(\mathcal{M}\) is decomposable, meaning that every finite rank operator in \(\mathcal{M}\) can be written as a sum of finitely many rank~1 operators in \(\mathcal{M}\). Examples are given to show that the continuity of the nest is required in general. A certain amount of machinery is introduced to build up to the proof of Theorem 3.1. For a subspace lattice \(\mathcal{L}\) and any element \(x\) in \(\mathcal{H}\), define \[ P_x=\wedge\{P\in\mathcal{L}:Px=x\}\text{ and }\widehat{P}_x=\wedge\{P\in\mathcal{L}:Px=0\}.\] Proposition 1.1 shows that, if \(\mathcal{L}\) is a nest and \(N<P\) in \(\mathcal{L}\), then there exists \(x\in\mathcal{H}\) such that \(P=P_x\) and \(N=\widehat{P}_x\). For a bounded operator \(T\) and a subspace lattice \(\mathcal{L}\), define the bilattice \[ \mathrm{BIL}(T,\mathcal{L})=\{(P,\,Q)\in\mathcal{L}\times\mathcal{L}^\perp\,:\,QTP=0\}.\] \(\mathrm{BIL}(T,\mathcal{L})\) is partially ordered by \((P_1,\,Q_1)\preceq(P_2,\,Q_2)\) if and only if \(P_1\leq P_2\) and \(Q_2\leq Q_1\). For a bounded operator \(T\) and a subspace lattice \(\mathcal{L}\), define the map \(\phi_T:\mathcal{L}\to\mathcal{L}\) by \[ \phi_T(P)=\vee\{N\in\mathcal{L}:(N,\,P^\perp)\in \mathrm{BIL}(T,\mathcal{L})\}.\] Then define maps \(\psi_T:\mathcal{L}\to\mathcal{L}^\perp\) and \(\sigma_T:\mathcal{L}\to\mathcal{L}\) by \[\psi_T(P)= \vee\{Q^\perp\in\mathcal{L}^\perp:Q^\perp T\phi_T(P)=0\}\text{ and } \sigma_T(P)= \vee\{P'\in\mathcal{L}:\phi_T(P')=\phi_T(P)\}.\] Both \(\phi_T\) and \(\sigma_T\) are order homomorphisms on \(\mathcal{L}\), while \(\psi_T\) is an anti-order homomorphism from \(\mathcal{L}\) to \(\mathcal{L}^\perp\) (see Lemma 2.3). The kernel map \(\omega_T:\mathcal{L}\to \mathrm{BIL}(T,\mathcal{L})\) is defined by \(\omega_T(P)= (\phi_T(P),\psi_T(P))\) for \(P\in\mathcal{L}\). The image of the kernel map \(\omega_T\) is called the kernel set, denoted by \(\Omega_T\). The proof of Theorem 3.1 draws on three other theorems from the paper, as follows. Theorem 2.8 shows that, if \(\mathcal{L}\) is a nest and \(T\) is a finite rank operator, then \(T\) can be written as the sum of finitely many rank~1 operators \(T=\sum_{r=1}^{m}{x_r\otimes y_r^*}\), where, for all \(P\in\mathcal{L}\) and all \(1\leq r\leq m\), \(\omega_T(P)\in \mathrm{BIL}(x_r\otimes y_r^*,\mathcal{L})\). (Note: the rank~1 operator \(x\otimes y^*\) is defined by \(z\mapsto \langle z,y\rangle x\).) If the nest \(\mathcal{L}\) is also continuous, then, for each \(r\), there is an element \((\Phi_r,\,\Psi_r)\) in the kernel set \(\Omega_T\) such that \(x\otimes y^*= \sigma_T(\Psi_r^\perp)(x_r\otimes y_r^*)\Phi_r^\perp\). Theorem 3.4 demonstrates that, when \(\mathcal{L}\) is a continuous nest and \(\mathcal{M}\) is a norm closed Lie \(\mathcal{T(L)}\) module containing the rank~1 operator \(x\otimes y^*\), then \(\mathcal{M}\) also contains all rank~1 operators in the set \(P_x\mathcal{B(H)}\widehat{P}_y^\perp\). For a continuous nest \(\mathcal{L}\) and a norm closed Lie \(\mathcal{T(L)}\) module \(\mathcal{M}\), Theorem 3.6 shows that, if \(T\) is a finite rank operator in \(\mathcal{M}\) and \((\Phi,\,\Psi)\) is in the kernel set \(\Omega_T\), then \(\mathcal{M}\) contains all rank 1 operators in the set \(\sigma_T(\Psi^\perp)\mathcal{B(H)}\Phi^\perp\).