Suppose $X$ and $Y$ are normed spaces over $\mathbb{R}$ and suppose $T: X \rightarrow Y$ is a bounded linear map. I want to prove that the adjoint map $T^\star : Y^\star \rightarrow X^\star$ is bounded as a linear map also, but I get super confused when trying to deal with the norms on the dual spaces since there are four norms, to deal with and I am unsure whether it is the operator norm of $T^\star$ or the dual norms i need to take?

I suppose the question is that since $\Vert T(x) \Vert_X \leq M \Vert x \Vert_Y$ it should be possible to find a $K$ s.t. $\Vert T^\star(l) \Vert_X^{\star} \leq K \Vert l \Vert_Y^{\star}$, but expanding these seems a little unfeasible.


Let $A \in L(X,Y)$, $\tilde y \in Y^*$, $x \in X$. Then from $\left(A^*(\tilde y)\right)(x) := \tilde y (A(x))$ you get

$$|A^*(\tilde y)(x)|= |\tilde y (A(x))|≤\|\tilde y\|_{Y^*} \|A(x)\|_Y ≤ \|\tilde y\|_{Y^*} \|A\|_{L(X,Y)} \|x\|_X$$

For $A^*$ to be bounded you must consider the following expression

$$\|A^*\|_{L(Y^*,X^*)}=\sup_{\tilde y \in Y^*, \ \|\tilde y\|_{Y^*}≤1} \left\{\|A^*(\tilde y)\|_{X^*}\right\}$$

Note that you have the following inequality, which follows from the first one

$$\|A^*(\tilde y)\|_{X^*}=\sup_{x \in X, \ \|x\|_X≤1}\{|A^*(\tilde y)(x)|\}≤\sup_{x \in X, \ \|x\|_X≤1}\{\|\tilde y\|_{Y^*} \|A\|_{L(X,Y)}\ \|x\|_X \}$$

Putting it together gives you


Note further that the construction gives you $\|A^{**}\|_{L(X^{**},Y^{**})}≤\|A^*\|_{L(Y^*,X^*)}≤\|A\|_{L(X,Y)}$. If you restrict $A^{**}$ to the subspace given by the embedding of $X$ into $X^{**}$ you get precisely $A$ again (composed with the embedding of $Y$ into $Y^{**}$). So $\|A^{**}\|_{L(X^{**},Y^{**})}≥\|A\|_{L(X,Y)}$ also holds, and the two inequalities are equalities, so also $\|A^*\|=\|A\|$.

  • $\begingroup$ I'm not sure of the adjoint of an operator $X \to Y$ (instead of $X \to X$) are you saying that sometimes $\|A^*\| \ne \|A\|$ ? it is when $Y$ is not isomorphic to $X$ ? $\endgroup$ – reuns Apr 24 '16 at 14:49
  • $\begingroup$ I have edited the answer to show $\|A^*\|=\|A\|$ in general. $\endgroup$ – s.harp Apr 24 '16 at 14:54

the definition of the adjoint of a bounded operator $T : X \to X$ is

$$ \text{for every }\ x \in X, y \in X^* : \qquad y(T(x)) = (T^* y)(x)$$ then note that for any $x \in X$ :$$\|x\|_{X} = \max_{y \in X^*, \ \|y\|_{X^*} =1} |y(x)|$$ hence $$\|T\| = \max_{\|x\|_X=1} \|Tx\| = \max_{\|x\|_X=1}\max_{\|y\|_{X^*}=1} |y(Tx)| = \max_{\|y\|_{X^*}=1} \max_{\|x\|_X=1}|y(Tx)|$$ $$ = \max_{\|y\|_{X^*}=1} \max_{\|x\|_X=1}|(T^*y)(x)| = \max_{\|y\|_{X^*}=1} \|T^*y\|_{X^*} = \|T^*\|$$

(this can easily be extended to the case $T : X \to Y$, if the adjoint exists)

  • $\begingroup$ (and when the normed space is not a Banach space, replace $\max$ by $\sup$ because the maximiser is in the Banach space, not always in the incomplete normed vector space) $\endgroup$ – reuns Apr 24 '16 at 14:54

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.