11
$\begingroup$

Let $T:H_1\rightarrow H_2$ be a bounded linear operator where $H_1$ and $H_2$ are Hilbert spaces. The Hilbert-adjoint is defined to the the operator $T^*:H_2\rightarrow H_1$ such that $\langle Tx,y\rangle_{H_2}=\langle x,T^*y\rangle_{H_1}$ for all $x\in H_1$ and $y\in H_2$. It can be shown that $\ker T^*=\left(\operatorname{im}\, T\right)^\perp$ and $\left(\ker T^*\right)^\perp=\overline{\operatorname{im}\, T}$ (see, e.g., here). It follows that $$T \text{ surjective} \Rightarrow \operatorname{im}\, T = H_2 \Rightarrow \ker T^* = H_2^\perp = \{0\}\Rightarrow T^* \text{ injective}$$ In the finite-dimensional case, we have that $$T \text{ surjective} \iff T^* \text{ injective}$$ In the general infinite-dimensional case, does the same statement hold, or is there a counterexample where $T^*$ is injective but $T$ is not surjective?

$\endgroup$
13
$\begingroup$

The main problem, in those things, is that linear operators in infinite dimensional spaces need not have closed image. So even if $T^\star$ is injective, meaning that $\ker T^\star=\{0\}$, we can only infer that $\overline{\text{im}\ T}=H_2$, which is not necessarily the same as $\text{im}\ T=H_2$.

For example, consider the operator $$T\colon \ell^2\to \ell^2,\qquad T\mathbf{x}=\left(x_1, \frac{x_2}{2}, \frac{x_3}{3}\ldots \right).$$ This operator is self-adjoint, so $T^\star=T$, and it is not surjective even if it is injective.

$\endgroup$
  • $\begingroup$ Why would it not be surjective ? $\endgroup$ – Theorem Jan 2 '13 at 12:35
  • 3
    $\begingroup$ Thanks Giuseppe. @Theorem, consider $y=\{1/n\}_{n=1}^\infty\in \ell^2$ and $x^{(n)} = (1,1,...,1,0,...)\in\ell^2$ where $x^{(n)}$ is a sequence with $n$ ones. Then $||Tx^{(n)} - y||=\sum_{j=n+1}^\infty \frac{1}{j^2} \rightarrow 0$. Thus, $y\in\overline{\operatorname{im}\, (T)}$ but $y\not\in\operatorname{im}\, (T)$ since $y=Tx$ implies that $x=(1,1,1,...)\not\in\ell^2$. $\endgroup$ – CCL Jan 2 '13 at 12:55
  • $\begingroup$ @CCL : Thanks for the explanation $\endgroup$ – Theorem Jan 2 '13 at 15:25
  • 3
    $\begingroup$ As Giuseppe explained, what you asked is not true. What is true however is the following: 1) Range of $T$ is dense if and only if $T^{\star}$ is injective and 2) If $T^{*}$ is injective and has closed range, then $T$ is surjective. $\endgroup$ – Theo Jan 2 '13 at 23:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.