When dealing with time-dependent PDEs, one often obtain that some quantity $E(t,x)$ belongs to a Lebesgue space $L^p_t(L^q_y)$, which means that $$\int_0^{+\infty}\|E(t,\cdot)\|_{L^q(\mathbb{R}^n)}^p dt<+\infty.$$ Sometimes, we even have $E\in L^{p_1}_t(L^{q_1}_y)\cap L^{p_2}_t(L^{q_2}_y)$ ; when this occurs, a simple interpolation, which involves only Hölder inequality, yields $E\in L^p_t(L^q_y)$ where $(\frac1p,\frac1q)$ is any point in the segment between $(\frac1{p_1},\frac1{q_1})$ and $(\frac1{p_2},\frac1{q_2})$.

I am interested in a slightly different situation, where $E\in L^{p_1}_t(L^{q_1}_y)\cap L^{q_2}_y(L^{p_2}_t)$ (the order of integrations differ in both spaces). What are the interpolation spaces ?

For instance, suppose that $E\in L^\infty_t(L^1_y)\cap L^\infty_y(L^1_t)$. What can be said of $E$ ?

  • 1
    $\begingroup$ This old answer on Math.SE might contain some bits of information. $\endgroup$ – Giuseppe Negro Nov 12 '18 at 11:12
  • $\begingroup$ @GiuseppeNegro. But it does not answer my (rather specific) question. $\endgroup$ – Denis Serre Nov 12 '18 at 11:21
  • 1
    $\begingroup$ No, it does not. I was just hoping it could be a bit useful. Sorry if it was not. $\endgroup$ – Giuseppe Negro Nov 12 '18 at 11:23
  • $\begingroup$ There is something not (completely) trivial that can be said by interpolation: namely that for every $f \in L^2_t$, the function $y \mapsto \int E(t,y) f(t) dt$ belongs to $L^2_y$. And conversely by replacing $t$ and $y$. In other words, $E$ is the kernel of a bounded operator from $L^2_y$ to $L^2_t$. Indeed, the assumption $E \in L^\infty_t(L^1_y)$ is equivalent to $E$ being bounded $L^1_t \to L^1_y$ and conversely. $\endgroup$ – Mikael de la Salle Nov 16 '18 at 15:10

$E \in L^\infty_t(L^1_y) \cap L^\infty_y (L^1_t)$ is insufficient to give any information about whether $E$ belongs to $L^p_t L^q_y$ when $p < \infty$.

Example: for simplicity, let's work with sequence spaces. Let $$ E(t,y) = \delta_{t,y} \, .$$ Then clearly $E \in \ell^\infty_t \ell^1_y \cap \ell^\infty_y \ell^1_t$. But $E$ doesn't belong in any $\ell^p_t \ell^q_y$ space where $p < \infty$.

| cite | improve this answer | |
  • $\begingroup$ If either $p_1 \leq q_1$ or $q_2 \leq p_2$ you can use Minkowski to swap the order, and the standard result applies. When $p_1 > q_1$ and $q_2 > p_2$, modifying the above shows that you cannot embed into $\ell^p \ell^q$ with $p < \mathrm{min}(p_1, q_2)$. $\endgroup$ – Willie Wong Nov 12 '18 at 16:23
  • $\begingroup$ Your example doesn't work, it is only $\ell^1\ell^1$. But you can take the measure $\delta_D$ supported by the diagonal $D=\{t=y\}$. $\endgroup$ – Denis Serre Nov 12 '18 at 17:43
  • $\begingroup$ @DenisSerre: my example is not $\ell^1, \ell^1$. ($\delta$ is the Kronecker, since we are working in sequence spaces). So $\sum_{y} E(t,y) = 1$ for every $t$ and vice versa. $\endgroup$ – Willie Wong Nov 12 '18 at 20:59

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.