Let $f: \mathbb{R}^n \rightarrow \mathbb{C}$ be in $L^2\cap L^1,$ then the Fourier transform is in $L^2 \cap L^\infty.$

Does this imply that we can take common norms in the sense that we can estimate

$$\sup_{x_i \in \mathbb{R}} \sqrt{\int_{\mathbb{R}^{n-1}} \left\lvert \hat{f}(x_1,...,x_i ,...,x_n) \right\rvert^2 dx_1...dx_{i-1}dx_{i+1}...dx_{n}}$$ or alternatively $$ \sqrt{\int_{\mathbb{R}^{n-1}}\sup_{x_i \in \mathbb{R}} \left\lvert \hat{f}(x_1,...,x_i ,...,x_n) \right\rvert^2 dx_1...dx_{i-1}dx_{i+1}...dx_{n}}?$$

Maybe this can even be bounded by mixed norms

$$\int_{\mathbb{R}} \sqrt{\int_{\mathbb{R}^{n-1}} \left\lvert f(x_1,...,x_i ,...,x_n) \right\rvert^2 dx_1...dx_{i-1}dx_{i+1}...dx_{n}}dx_i$$ or

$$ \sqrt{\int_{\mathbb{R}^{n-1}}\left(\int_{\mathbb{R}} \left\lvert f(x_1,...,x_i ,...,x_n) \right\rvert dx_i\right)^2 dx_1...dx_{i-1}dx_{i+1}...dx_{n}}?$$

However, I do not see yet whether this is true or not.


Yes, such bounds follow by considering partial Fourier transforms and then using the norm bounds you quoted. Let me take $n=2$ for ease of notation. Write $g(x,t)=\int f(x,k) e^{2\pi i kt}\, dk$. Then, for example (and as desired), $$ \sup_y \|\widehat{f}(x,y)\|_{L^2(dx)} =\sup_y \|g(x,y)\|_{L^2(dx)}\le \left\| \int |f(x,k)|\, dk \right\|_{L^2} \le \int\|f(x,k)\|_{L^2(dx)}\,dk , $$ by the integral version of Minkowski's inequality for the final step.


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.