Published on

Operators on Infinite Dimensional Vector Spaces 3

Authors

Convergence of Operators

Definition 1.18: Let Tn,T∈B(H)T_n,T \in \mathcal{B}(\mathcal{H}). We say that:

  • Tn→TT_n \rightarrow T in the operator norm, (or uniformly) if ∥Tn−T∥→0\|T_n - T\| \rightarrow 0 as n→∞n \rightarrow \infty;
  • Tn→TT_n \rightarrow T strongly, if for any x∈Hx \in \mathcal{H} one has ∥Tnx−Tx∥→0\|T_nx - Tx\| \rightarrow 0 as n→∞n \rightarrow \infty;
  • Tn→TT_n \rightarrow T weakly, if for any x,y∈Hx,y \in \mathcal{H} one has (Tnx,y)→(Tx,y)(T_nx,y) \rightarrow (Tx,y) as n→∞n \rightarrow \infty.

Theorem 1.19: Uniform convergence implies strong convergence, and strong convergence implies weak convergence.

Proof: Exercise. ■\blacksquare

The following examples show the implication does not go in the other direction.

Example 1.20: Consider operators on ℓ2\ell^2.

  • Let Tn=1nIT_n = \frac{1}{n}I. Then Tn→0T_n \rightarrow 0 uniformly.
  • Consider the powers of the backwards shift operator on ℓ2\ell^2,
    (S∗)n(x0,x1,...)=(xn,xn+1,...).(S^*)^n (x_0,x_1,...) = (x_n,x_{n+1},...).
    Then (S∗)n→0(S^*)^n \rightarrow 0 strongly, but not uniformly.
  • Consider the powers of the shift operator in ℓ2\ell^2, i.e.:
    Sn(x0,x1,...)=(0,0,...,0⏞n,x0,x1,...).S^n(x_0,x_1,...)=(\overbrace{0,0,...,0}^n,x_0,x_1,...).
    Then Sn→0S^n \rightarrow 0 weakly but not strongly.