6
$\begingroup$

Consider a standard integral operator $T$ formally defined by $$ Tf(x):=\int_{K} k(x,y)f(y)dy,\qquad x\in K, $$ where $K$ is a locally compact metric measure space. It is immediate to see that the operator is bounded in $\infty$-norm if $\sup_{x\in K}\int_K |k(x,y)| dy<\infty$.

Are there any conditions implying this operator to be bounded on $C_b(K)$ and/or on $BUC(K)$ and/or on $C_0(K)$?

$\endgroup$
7
  • 1
    $\begingroup$ As I understand you are asking under what conditions $T(X)\subset X$ for each $X$ you mentioned, namely (i) bounded continuous functions on $\overline K$ (is here $\overline K$ the one-point compactification ?); (ii) bounded uniformly continuous functions on $\overline K$ ; 3) $C_0(K)$= continuous functions vanishing at infinity. Correct? $\endgroup$ Commented Nov 28, 2019 at 9:12
  • 1
    $\begingroup$ @PietroMajer Sorry, by $\overline{K}$ I mean the end compactification. I should have written it; I've edited my question accordingly. Concerning your other question: yes, that's correct. $\endgroup$ Commented Nov 28, 2019 at 9:17
  • $\begingroup$ What are $C_b$ and $BUC$? $\endgroup$ Commented Dec 4, 2019 at 13:17
  • 2
    $\begingroup$ @DelioMugnolo Since $\overline{K}$ is compact,continuous functions are bounded and uniformly bounded (whatever it means). What do you mean by a uuniformly bounded function? $\endgroup$ Commented Dec 4, 2019 at 22:36
  • 1
    $\begingroup$ In order that $T$ maps the space of bounded continuous functions into itself it is most natural to assume that all $x\mapsto k(x,y)$ are continuous and in addition some assumption allowing to apply the dominated convergence theorem, e.g., for each $x_0\in K$ there are a neighbourhood $A$ and $g\in L^1(\mu)$ such that $|k(x,y)|\le g(y)$ for all $x\in A$ and $y\in K$. $\endgroup$ Commented Dec 5, 2019 at 13:22

1 Answer 1

2
$\begingroup$

Too long for a comment. Your requirement is too stringent and it is quite likely that to get continuity from $L^\infty$ into itself, it is indeed necessary to have $$ \text{esssup}_x\int\vert k(x,y)\vert dy<+\infty. $$ On the other hand, if you are interested in $L^2$ or $L^p$ continuity, you have a much more interesting question. By the way, no iff condition is known for the $L^2(\mathbb R^N)$ continuity, although you have several quite refined sufficient conditions. For instance, Calder\'on-Zygmund singular integrals and zeroth order pseudo-differential operators are bounded on $L^p(\mathbb R^N)$ for $p\in (1,+\infty)$, but not in general on $L^1$ or $L^\infty$. If you relax your $L^\infty$ requirement and for instance look for $BMO(\mathbb R^N)$ continuity, I believe that singular integrals with odd kernels are indeed bounded on $BMO(\mathbb R^N)$: to give a simple example, I think that with $$ k(x,y)=\text{pv}\frac{1}{x-y}=\frac{d}{dx}\bigl(\ln \vert x-y\vert\bigr) $$ you get an operator which is not bounded on $L^\infty(\mathbb R)$, but is bounded on $BMO(\mathbb R)$.

You can raise also a similar question for the Sobolev continuity, say from $H^s(\mathbb R^N)$ into $H^{s'}(\mathbb R^N)$.

$\endgroup$
3
  • $\begingroup$ Thanks, but the question is related to a specific application I have in mind. I already know by other means that the operator is bounded on $L^2$; what I want is to deduce from properties of its kernel that it's also bounded on $BUC$ or $C_0$. $\endgroup$ Commented Dec 8, 2019 at 18:33
  • $\begingroup$ @DelioMugnolo Do you know if your operator is $L^p$ bounded for $p\in (1,+\infty)$? On the other hand, I believe that the condition on the essential supremum is necessary for $L^\infty$ boundedness, at least in an Euclidean framework with the Lebesgue measure. $\endgroup$
    – Bazin
    Commented Dec 9, 2019 at 15:53
  • $\begingroup$ Yes, I know that my operator is contractive on all $L^p$-spaces. My question really was about the way it acts on continuous functions. $\endgroup$ Commented Dec 10, 2019 at 12:46

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .