Compactness properties of commutators of bilinear fractional integrals (Q2349908)

If \(T\) is a Calderón-Zygmund (CZ) operator on \(\mathbb R^n\), the commutator of \(T\) with appropriate symbols \(b\) is given by \[ [T,b]f:= T(fb) - b Tf. \] A classical result of \textit{R. R. Coifman} et al. [Ann. Math. (2) 103, 611--635 (1976; Zbl 0326.32011)] states that \([T,b]\) is bounded on \(L^p(\mathbb R^n)\), \(1<p<\infty\), whenever \(b\in \text{BMO}\). A result of \textit{A. Uchiyama} [Tohoku Math. J. (2) 30, 163--171 (1978; Zbl 0384.47023)] states that the commutator \([T,b]\) is actually compact whenever \(b\in\text{CMO}\), the closure of \(C^\infty _c\) in the \(\text{BMO}\) norm. For special classes of CZ operators, the previous implications can also be reversed. Let \(T:X\times Y\to Z\) be a bilinear operator, where \(X,Y,Z\) are normed spaces, a typical case being that of \(T:L^p\times L^q \to L^r\) for appropriate \(1\leq p,q,r \leq \infty\). The bilinear operator \(T\) is (jointly) compact if the set \(\{T(x,y):\, \|x\|_X,\|y\|_Y \leq 1\}\) is precompact in \(Z\). The commutators of \(T\) with appropriate symbols \(b\in \text{BMO}\) are defined as \[ [T,b]_1 (f,g) := T(bf,g) - b T(f,g), \quad\quad [T,b]_2 (f,g):= T(f,bg)-b T(f,g). \] If \(T\) is a bilinear CZ operator and \(b\in\mathrm{CMO}\) it was proved in [\textit{Á. Bényi} and \textit{R. H. Torres}, Proc. Am. Math. Soc. 141, No. 10, 3609--3621 (2013; Zbl 1281.42010)] that \([T,b]_i:L^p\times L^q \to L^r\), \(1/p+1/q=1/r\), is compact for \(i=1,2\). In the paper under review the authors consider two classes of bilinear operators. The first family \(\{T_\alpha\}_\alpha\) is parametrised by \(0<\alpha<2n\) and its typical member is the bilinear Riesz potential operator \[ \mathcal I_\alpha (f,g) (x):= \int_{\mathbb R^{2n}} \frac{f(y)g(z)}{(|x-y|+|x-z|)^{2n-\alpha}}dydz. \] Other members in that family have kernels with similar properties as the kernel in the Riesz potential while the endpoint \(\alpha=0\) corresponds to bilinear CZ operators. The first main theorem of the paper shows for example that for \(i=1,2\) the commutators \([T_\alpha,b]_i :L^p\times L^q \to L^r\) are compact for the natural values of \(p,q,r\). The second family \(\{BI_a\}_a\) consists of bilinear fractional integrals that correspond to the bilinear Hilbert transform: \[ BI_a(f,g):= \int_{\mathbb R^n} \frac{f(x-y)g(x+y)}{|y|^{n-1}}dy. \] These operators were introduced by \textit{L. Grafakos} [Stud. Math. 102, No. 1, 49--56 (1992; Zbl 0808.42014)]. The authors show that, for \(i=1,2\), the commutators \([BI_\alpha,b]_i\) map \(L^p\times L^q\) into \(L^r\) for \(1/r=1/p+1/q-\alpha/n\), \(1/p+1/q<1\) whenever \(b\in \text{BMO}\). Furthermore, if \(b\in\mathrm{CMO}\) then the commutators are separately compact in each entry (meaning, for example, that for each \(f\in L^p\), the operator \([BI_\alpha,b]_1(f,\cdot):L^q\to L^r\) is compact, and symmetrically when we interchange the roles of \(f,g\).) This is a weaker property than the joint compactness shown for the operators of the family \(T_\alpha\). The authors state a couple of interesting questions at the end of the paper, one of which is whether the commutators \([BI_\alpha,b]_i\) are jointly compact.