Let $p, q \in {]1,\infty[}$ where $\frac{1}{p}+\frac{1}{q}=1$ and define

$J\colon L^{q} \to (L^{p})^{*}, f \mapsto \ell_{f}: L^{p} \ni g\mapsto \ell_{f}(g)=\int_{X}fg\,d\mu$

My question is related in general to the notion of well definedness for operators. In particular, why is it sufficient for an operator to be considered well-defined, when we can show that for any $f \in L^{p}$ that $\|\ell_{f}\|_{*}<\infty$. In my previous, elementary versions of well-definedness, we defined it as for any $f \in L^{q}$ there exists exactly one $\ell\in (L^{p})^{*}$ so that $Jf=\ell$. Now showing that $\| Jf\|_{*}<\infty$ definitely shows that $Jf \in (L^{p})^{*}$ but it does not show that there is a unique $\ell \in (L^{p})^{*}$ so that $Jf=\ell$.

In other words, there may exist another $m \in (L^{p})^{*}$ where $m\neq \ell$ so that $m = Jf$ and $\ell = Jf$. I thought that this was the point of proving well-definition.

Why is it always reduced to the idea that well-definition depends only on whether $\|\ell_{f}\|_{*}<\infty$?

Any clarity on this topic will help me greatly.

  • 2
    $\begingroup$ What's the definition of "well-defined operator" that you're using? From what I know, the phrase "well-defined" is an abuse of language—when we say "$J$ is a well-defined operator", we don't actually mean "$J$ is an operator, and it has the property of being well-defined", we mean "the given definition of $J$, which is purported to be the definition of an operator, actually is the definition of an operator". $\endgroup$ – Tanner Swett Jun 21 at 22:01

There are two important notes here to take:

  1. The space $L^p(X,\mu)$ is actually a quotient space (by the subspace of almost everywhere zero functions), and the formal definition $g\mapsto \int_X|fg|\,d\mu$ of the mapping $\ell_f$ refers to actual functions and not their cosets in the quotient.
    So, well-definedness here effectively means that if $g$ is almost everywhere zero, then $\ell_f(g)=0$.

  2. Boundedness of a linear operator is rather connected to the fact that it is everywhere defined.
    For a typical example, it's usual to consider the differentiation operator $f\mapsto f'$ in e.g. $L^2(\Bbb R)$ or $C(\Bbb R)$, which is not bounded, and is defined only on a dense subspace.


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.