Let HH be a . The strong-∗* operator topology on is the locally convex topology generated by the seminorms

qξ(T)=(∥Tξ∥2+∥T∗ξ∥2)1/2,ξ∈H.q_\xi(T)=\bigl(\|T\xi\|^2+\|T^*\xi\|^2\bigr)^{1/2}, \qquad \xi\in H.

Thus a net TiT_i converges strong-∗* to TT exactly when

Tiξ→TξandTi∗ξ→T∗ξT_i\xi\to T\xi \quad\text{and}\quad T_i^*\xi\to T^*\xi

in norm for every ξ∈H\xi\in H. A concrete inherits this topology from B(H)B(H). It refines the by requiring pointwise control of adjoints as well as operators.

Characterizing convergence

Strong-∗* convergence is equivalent to simultaneous strong convergence of TiT_i and Ti∗T_i^*. Consequently the involution T↦T∗T\mapsto T^* is continuous. Multiplication is jointly strong-∗* continuous on norm-bounded subsets: if Si→SS_i\to S and Ti→TT_i\to T strong-∗*, with both nets bounded, then SiTi→STS_iT_i\to ST strong-∗*.

Comparison with other topologies

Strong-∗* convergence implies strong convergence, but the converse can fail because adjoints need not preserve strong limits. For example, on ℓ2(N)\ell^2(\mathbb N), the powers S∗nS^{*n} of the backward shift converge strongly to zero, while their adjoints SnS^n do not. Hence S∗nS^{*n} does not converge strong-∗* to zero.

On norm-bounded sets, strong-∗* and convergence agree. Globally, the ultrastrong-star topology is strictly finer in infinite dimension because it can test square-summable families of vectors at once.

Conventions and scope

The symbols “strong-∗*,” “strong star,” and “SOT-star” all refer here to the topology generated by individual vectors. They should not be used for the σ\sigma-strong-star topology on an unbounded set. For abstract , the ultrastrong-star formulation is intrinsic, whereas a strong-∗* topology is attached to a concrete representation.

References
  1. Masamichi Takesaki, Theory of Operator Algebras I, Springer, 1979. Springer DOI record. Relevant: Chapter I, §3 on strong, strong-star, weak, and related operator topologies.