Let $n\in \mathbb{N}$, $n\geqslant2$, and $\mathbb{N}_0=\mathbb{N} \cup \{0\}$. For $x=(x_1,\dots,x_n)$ and $y=(y_1,\dots,y_n) \in \mathbb{R}^n$ we put $xy=x_1y_1+\cdots+x_ny_n$. In what follows $\mathcal{S}:=\mathcal{S}(\mathbb{R}^n)$ and $\mathcal{S}':=\mathcal{S}'(\mathbb{R}^n)$ are the Schwartz spaces of test functions and tempered distributions, respectively, $\widetilde{\mathcal{S}}':=\mathcal{S}(\mathbb{T}^n)$ is the space of all distributions $f\in\mathcal{S}'$ that are $1$-periodic in each variable (that is, such that $\langle f,\varphi(\cdot+\xi)\rangle=\langle f,\varphi\rangle$ for all $\varphi\in\mathcal{S}$ and $\xi\in\mathbb{Z}^m$), $\widetilde{\mathcal{S}}:=\mathcal{S}(\mathbb{T}^m)$ is the space of all infinitely differentiable functions on $\mathbb{T}^m$, $\widehat{\varphi}$ is the Fourier transform of $\varphi\in \mathcal{S}$, and $\widehat{u}$ is the Fourier coefficient of $u\in \widetilde{\mathcal{S}}$:
Operators (1) were introduced by Meyer and Coifman in the mid-1970s and play an important role in harmonic analysis and its applications (see [1]–[3]).
Let $X_1(\mathbb{I}^n),\dots,X_N(\mathbb{I}^n)$, and $Y(\mathbb{I}^n)$ (where $\mathbb{I} \in \{\mathbb{R},\mathbb{T}\}$) be function spaces on $\mathbb{I}^n$, with (quasi)norms $\|\cdot|\, X_1(\mathbb{I}^n)\|,\dots, \|\cdot|\, X_N(\mathbb{I}^n)\|$, and $\|\cdot\,|Y(\mathbb{I}^n)\|$, respectively. Assume that $\mathcal{S}(\mathbb{I}^n)\subset X_\nu(\mathbb{I}^n)$, $\nu=1,\dots,N$. Let $R_a:=T_a$ for $\mathbb{I}=\mathbb{R}$ and $R_a:=\widetilde{T}_a$ for $\mathbb{I}=\mathbb{T}$. If there is a positive constant $C$ such that the inequality $\|R_a(u_1,\dots,u_n) \mid Y(\mathbb{I}^n)\| \leqslant C\prod_{\nu=1}^N\|u_\nu \mid X_\nu(\mathbb{I}^n)|$ is valid for all $(u_1,\dots,u_N)\in \mathcal{S}(\mathbb{I}^n)^N$, then we say that the operator $R_a$ is bounded from $X_1(\mathbb{I}^n)\times \cdots \times X_N(\mathbb{I}^n)$ to $Y(\mathbb{I}^n)$ and write $R_a\colon X_1(\mathbb{I}^n)\times\cdots\times X_N(\mathbb{I}^n)\to Y(\mathbb{I}^n)$.
The boundedness of operators (1) has intensively been studied for various symbols and functions spaces; see, for instance, [1]–[6] and the references there.
The Hörmander class $S_{\rho\delta}^m(\mathbb{R}^n;N)$ ($m\in\mathbb{R}$, $0\leqslant\delta, \rho\leqslant 1$) consists of all symbols $a\in C^\infty(\mathbb{R}^{n(N+1)})$ for which the following condition is satisfied: for all $\alpha,\beta^1,\dots,\beta^N \in \mathbb{N}_0^n$ there exists $C=C(\alpha,\beta^1,\dots,\beta^N)>0$ such that the differential inequality
holds true. Here $\langle \Xi \rangle=(1+\sum_{\nu=1}^N\xi^\nu\xi^\nu)^{1/2}$.
The Hörmander class $S_{\rho\delta}^m(\mathbb{T}^n; N)$ ($m\in\mathbb{R}$, $0\leqslant\delta, \rho\leqslant 1$) of toroidal symbols $a \colon \mathbb{R}^{n}\times\mathbb{Z}^{nN}\to\mathbb{C}$ ($a(\,\cdot\,,\Xi) \in C^\infty(\mathbb{T}^{n})$ for each $\Xi\in\mathbb{Z}^{nN}$) is defined analogously — with $\Delta^{\beta^\nu}_{\xi^\nu}$ (a finite difference of order $\beta^\nu$ with step $1$) and $\mathbb{Z}^{nN}$ in place of $\partial^{\beta^\nu}_{\xi^\nu}$ and $\mathbb{R}^{nN}$.
We fix $\phi\in \mathcal{S}$ such that $\displaystyle\int_{\mathbb{R}^n} \phi=1$. Put $\phi_t(x):=t^{-n}\phi(x/t)$.
Below $\widetilde{X}_p:=\widetilde{H}_p$ for $0< p\leqslant 1$, $\widetilde{X}_p:=\widetilde{L}_p$ for $1< p\leqslant\infty$, $\widetilde{Y}_r:=\widetilde{L}_r$ for $0< r< \infty$, and $\widetilde{Y}_\infty:=\widetilde{\operatorname{BMO}}$. Here $\widetilde{L}_p:=L_p(\mathbb{T}^n)$ ($0< p\leqslant\infty$) is the Lebesgue space endowed with the standard (quasi)norm $\|\cdot |\, \widetilde{L}_p\|$, $\widetilde{\operatorname{BMO}}:=\operatorname{BMO}(\mathbb{T}^n)$ is the space of functions of bounded mean oscillation on $\mathbb{R}^n$ which are 1-periodic in each variable, and
is the (real) Hardy space, $0< p\leqslant1$ (the periodization $\widetilde{g}\colon \mathbb{T}^m \to \mathbb{C}$ of the function $g\colon \mathbb{R}^m \to \mathbb{C}$ is defined as the (formal) sum of the series $\sum_{\xi\in \mathbb{Z}^m}g(x+\xi)$).
For $0 \leqslant \rho < 1$ and $0 < p,q,r \leqslant \infty$ such that $1/p+1/q=1/r$, we defined the quantities $m_\rho(p,q)=(1-\rho)m_0(p, q)$ and $m_0(p,q)=-n\max\{1/2,1/p,1/q,1- 1/r,1/r-1/2\}$.
Theorem 1. Let $0\leqslant\rho<1$, and let $0< p, q, r\leqslant\infty$ satisfy $1/p+1/q=1/r$. Then $\sup\{ m\in \mathbb{R} \mid \widetilde{T}_a \colon \widetilde{X}_p \times \widetilde{X}_q \to \widetilde{Y}_r$ for all $a\in S^m_{\rho\rho}(\mathbb{T}^n; 2)\}=m_\rho(p,q)$.
Moreover, $\widetilde{T}_a\colon \widetilde{X}_p \times \widetilde{X}_q \to \widetilde{Y}_r$ for each $a\in S^{m_0(p,q)}_{0 0}(\mathbb{R}^n;2)$. If, in addition, $0\leqslant\rho<1$ and $\min\{p,q,r\}\geqslant1$, then $\widetilde{T}_a\colon \widetilde{X}_p \times \widetilde{X}_q \to \widetilde{Y}_r$ for any $a\in S^{m_\rho(p,q)}_{\rho\rho}(\mathbb{T}^n;2)$.
Theorem 2. Let $N\geqslant2$, $0< r,p_1,\dots,p_N\leqslant\infty$, $1/r \leqslant 1/p_1+\cdots+1/p_N$, and $m\in\mathbb{R}$. Then $\widetilde{T}_a \colon \widetilde{X}_{p_1}\times \cdots \times \widetilde{X}_{p_N} \to \widetilde{Y}_r$ for all $a\in S^m_{0 0}(\mathbb{T}^n;N)$ if and only if $m\leqslant n\bigl(\min\{1/p,1/2\}-\sum_{\nu=1}^N\max\{1/p_\nu,1/2\}\bigr)$.
Remark. Theorems 1 and 2 are natural analogues of an important result of Miyachi and Tomita [4], [5] and the main result in [6], respectively. Notice the (only known to us) result in [7] on toroidal multiplinear PDOs with symbols in Hörmander classes: if $1\leqslant r<\infty$ and $1\leqslant p_\nu \leqslant\infty$ satisfy $1/r=1/p_1+\cdots+1/p_N$ ($N\geqslant2$), then for each $a\in S_{10}^0(\mathbb{T}^n;N)$
The author would like to thank the Isaac Newton Institute in Cambridge, UK, for their support and hospitality during the programme “Discretization and recovery in high-dimensional spaces” (1–26 July 2024) where the work on this paper was undertaken.
Bibliography
1.
R. R. Coifman and Y. Meyer, Au delà des opérateurs pseudo-différentiels, Astérisque, 57, Soc. Math. France, Paris, 1978, i+185 pp.
2.
Y. Meyer and R. Coifman, Wavelets. Calderón–Zygmund and multilinear operators, Transl. from the French, Cambridge Stud. Adv. Math., 48, Cambridge Univ. Press, Cambridge, 1997, xx+315 pp.
3.
C. Muscalu and W. Schlag, Classical and multilinear harmonic analysis, v. 2, Cambridge Stud. Adv. Math., 138, Cambridge Univ. Press, Cambridge, 2013, xvi+324 pp.
4.
A. Miyachi and N. Tomita, Indiana Univ. Math. J., 62:4 (2013), 1165–1201
5.
A. Miyachi and N. Tomita, Ann. Inst. Fourier (Grenoble), 70:6 (2020), 2737–2769
6.
T. Kato, A. Miyachi, and N. Tomita, J. Funct. Anal., 282:4 (2022), 109329, 28 pp.
7.
D. Cardona and V. Kumar, J. Fourier Anal. Appl., 25:6 (2019), 2973–3017
Citation:
D. B. Bazarkhanov, “Boundedness of toroidal multilinear pseudodifferential operators with symbols in Hörmander classes”, Russian Math. Surveys, 80:1 (2025), 137–139