Define Hilbert Transform (HT) as the convolution with the function $1/x$. E. Stein proves in his book Singular Integrals and Differentiability Properties of Functions that HT, when understood as a singular integral operator, is a bounded operator on $L^p(\mathbb{R})$ for $p\in (1, \infty)$.

I am wondering if HT has compact commutator with multiplication by $C_0(\mathbb{R})$ on $L^p(\mathbb{R})$?

More precisely, if $T \in \mathscr{L}(L^p(\mathbb{R}))$ denotes the Hilbert transform, and $f \in C_0(\mathbb{R})$, is it true that $Tf - fT \in \mathbb{K}(L^p(\mathbb{R}))$? If it is true, would you please give me a reference? Thank you!

P.S: cross-posted from MSE here: https://math.stackexchange.com/questions/676833/does-hilbert-transform-commute-with-function-multiplication-modulo-compact-on-l

  • $\begingroup$ Does not look likely to me $\endgroup$ Feb 16, 2014 at 11:27

1 Answer 1


The problem reduces to the case of smooth functions with compact support, since they are norm dense in $C_0$.

Now let $f$ be a smooth function with compact support. Then $[T,f]$ is an integral operator with smooth kernel $k(x,y) := (f(x)-f(y))/(x-y)$.

There is an easy to check sufficient condition for compactness of integral operators from $L^p$ to $L^q$ in terms of iterated norms: namely, if an integral operator with kernel $k$ has finite norm $\left[ \intop \left( \intop |k(x,y)|^{p^\ast} dy \right)^{q/p^\ast} dx \right]^{1/q}$, $1/p+1/p^\ast = 1$, then the operator is compact from $L^p$ to $L^q$. This condition is obviously satisfied by our kernel.

  • $\begingroup$ Are smooth functions dense in the operator norm? $\endgroup$ Feb 18, 2014 at 12:30
  • $\begingroup$ Eh, this was not a very good comment. $\endgroup$ Feb 18, 2014 at 15:01
  • $\begingroup$ By the way, I am convinced now. $\endgroup$ Feb 18, 2014 at 15:03
  • $\begingroup$ Hi Alex, Thanks for your reply. Two questions: a) Is the operator norm of "multiplied by $f$" operator bounded by its sup (or $L^\infty$ norm)? b) For the case of $f$ smooth and cptly supported, can we use the fact that $k(x, y)$ can be approximated in sup norm by $\sum_i g_i(x)h_i(y)$? If so, is the operator norm of the "difference kernel" bounded by its sup norm? Thanks! $\endgroup$ Feb 18, 2014 at 21:27
  • 1
    $\begingroup$ a) The operator norm of multiplication by $f$ equals the $L^\infty$ norm of $f$. b) The sup norm of the kernel on $\mathbb{R}$ doesn't bound the operator norm, since not all bounded kernels correspond to bounded operators. However, the decomposition into a sum of rank-ones is certainly useful for the "$\mathrm{supp} f \times \mathrm{supp} f$ block" of the kernel, if you view it as a "block matrix" corresponding to $L^p(\mathbb{R}) \simeq L^p(\mathrm{supp} f) \oplus L^p(\mathbb{R} \setminus \mathrm{supp} f)$. On that block this decomposition actually converges fast enough to ensure that $\endgroup$ Feb 19, 2014 at 0:48

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.