Notes on commutators on Herz-type spaces (Q1423793): Difference between revisions

Let \(B_k=\{x\in \mathbb R^n; 2^{k-1}\leq | x| < 2^k\}\) for \(k\in \mathbb Z\), and \(\chi_k\) denote the characteristic function of the set \(B_k\). The homogeneous Herz space \(\dot K_q^{\alpha, p}(\mathbb R^n)\) is defined in terms of \(\| f\| _{\dot K_q^{\alpha, p}(\mathbb R^n)}\) \(=\{\sum_{k=-\infty}^\infty 2^{k\alpha p}\| f\chi_{k}\| _{L^q(\mathbb R^n)}^p\}^{1/p}\) by letting \(\dot K_q^{\alpha, p}(\mathbb R^n) =\{f\in L_{\roman{loc}}^q(\mathbb R^n\setminus \{0\}); \| f\| _{\dot K_q^{\alpha, p}(\mathbb R^n)}<\infty\}\). Herz-type Hardy spaces are defined in terms of the grand maximal functions or radial maximal functions of distributions \(f\ast \varphi_t\) \((\varphi\in C_0^\infty(\mathbb R^n)\) with \(\int \varphi dx=1)\), like as in the classical Hardy spaces. \(\roman{CMO}_q\) is defined like \(\roman{BMO}\), i.e., \(\roman{CMO}_q\) is the set of locally integrable functions satisfying \(\sup_{r>0}(| B(0,r)| ^{-1}\int_{B(0,r)} | f(x)-f_{B(0,r)}| ^q\,dx)^{1/q}<\infty\). The author considers commutators \([b,T]\) between the multiplication operator by a locally integrable function \(b\) and a Carderón-Zygmund singular integral operator \(T\), whose kernel is a nondegenerate convolution one \(K(x)\), i.e., there exists \(N>1\) such that for any \(r>0\) there is \(x_0\in\mathbb R^n\) with \(| x_0| =Nr\) and \(K(x)\geq C| x| ^n\) for \(x\in B(x_0,r)\). He gives a factorization theorem in the Herz-type Hardy spaces: Let \(1<p, q_1,q_2<\infty\), \(-n/q_i<\alpha_i<n(1-1/q_i)\), \(n(1/q_1+1/q_2)+\alpha_1+\alpha_2=n\), and \(q_2\leq q\). Then, every \(f\in H\dot K_q\) can be factorized as \(f=\sum_{k=-\infty}^{\infty}(g_kT^*h_k-h_kTg_k)\), where \(g_k\in \dot K_{q_1}^{\alpha_1,p}\), \(h_k\in \dot K_{q_2}^{\alpha_2,p'}\), and \(\sum_{k=-\infty}^{\infty}\| g_k\| _{\dot K_{q_1}^{\alpha_1,p}} \| h_k\| _{\dot K_{q_2}^{\alpha_2,p'}}\leq C\| f\| _{H\dot K_q}\). This corresponds to the classical Hardy space \(H^1\) by Coifman-Rochberg-Weiss. The author states also the following: Let \(p, q_1, q_2, \alpha_1, \alpha_2\) be the same as above. Then, if \([b,T]\) is bounded from \(\dot K_{q_1}^{\alpha_1,p}\) to \(\dot K_{q_2'}^{-\alpha_2,p}\), it follows \(b\in \roman{CMO}_{q_2'}\). (This is the converse part of a result given \textit{L. Grafakos}, \textit{X. Li} and \textit{D. Yang} [Trans. Am. Math. Soc. 350, No. 3, 1249--1275 (1998; Zbl 0893.47044)].) However, I could not follow a part of his proof. Nevertheless, I think his statement may be true. When \(K(x)=\Omega(x)/| x| ^n\), \(\Omega\) is homogeneous of degree zero and \(\Omega^{-1}\) is sufficiently smooth on some open set of \(S^{n-1}\), one can prove it by the same method used by \textit{S. Janson} to prove the classical \(L^p\) cases [Ark. Math. 16, 263--270 (1978; Zbl 0404.42013)].