[proofplan] We follow the standard proof from Stein, *Singular Integrals and Differentiability Properties of Functions*, Chapter IV §6: combine the [Marcinkiewicz Multiplier Theorem in Dimension One](/theorems/3190) with a product Littlewood--Paley decomposition along each coordinate axis. The key tools are (i) the product Littlewood--Paley square-function characterisation of $L^p(\mathbb{R}^n)$ for $1 < p < \infty$ (a consequence of iterating the one-dimensional Littlewood--Paley inequality, which is itself a corollary of the vector-valued Mihlin multiplier theorem on the UMD space $\ell^2$, cited), and (ii) summation by parts on each coordinate, applied to the dyadic-product decomposition of $m$. The full $n$-dimensional Marcinkiewicz hypothesis (variation bounds on every $|S|$-dimensional dyadic face, uniform in the frozen coordinates) is exactly what is needed to control the iterated summation by parts. We do **not** induct on dimension via an operator-valued one-dimensional Marcinkiewicz theorem; instead we treat all $n$ coordinates symmetrically inside a single product Littlewood--Paley estimate, reducing the multiplier bound to repeated applications of one-dimensional summation by parts and the (scalar) Littlewood--Paley inequality. [/proofplan]

[step:Set up the product dyadic decomposition along each coordinate axis] For $j \in \mathbb{Z}$, let $I_j^+ := [2^j, 2^{j+1})$ and $I_j^- := (-2^{j+1}, -2^j]$, so $\mathbb{R} \setminus \{0\} = \bigsqcup_{j \in \mathbb{Z},\,\pm} I_j^\pm$. For each coordinate $k = 1, \ldots, n$ and each "dyadic-signed index" $\jmath_k = (j_k, \pm) \in \mathbb{Z} \times \{+, -\}$, write $I_{\jmath_k}$ for the corresponding interval in the $\xi_k$-axis. A **dyadic rectangle** is a product \begin{align*} R_{\jmath} := I_{\jmath_1} \times I_{\jmath_2} \times \cdots \times I_{\jmath_n} \subset \mathbb{R}^n, \qquad \jmath = (\jmath_1, \ldots, \jmath_n). \end{align*} The collection $\{R_{\jmath}\}$ partitions $\mathbb{R}^n$ up to a Lebesgue-null set. For each axis $k$ and dyadic-signed index $\jmath_k$, let $\Delta_{\jmath_k}^{(k)}$ be the Littlewood--Paley projection in the $k$-th coordinate restricted to $I_{\jmath_k}$: \begin{align*} \Delta_{\jmath_k}^{(k)} : L^p(\mathbb{R}^n) &\to L^p(\mathbb{R}^n) \\ f &\mapsto \mathcal{F}^{-1}\bigl(\mathbb{1}_{I_{\jmath_k}}(\xi_k)\, \hat f(\xi)\bigr). \end{align*} The product projection \begin{align*} \Delta_{\jmath} := \Delta_{\jmath_1}^{(1)}\,\Delta_{\jmath_2}^{(2)}\,\cdots\,\Delta_{\jmath_n}^{(n)} \end{align*} is the Fourier-multiplier operator with symbol $\mathbb{1}_{R_{\jmath}}(\xi)$, and $\sum_{\jmath} \Delta_{\jmath} f = f$ for $f \in \mathcal{S}(\mathbb{R}^n)$ (interpreted as a partial-sum limit in $L^p$). [/step]

[step:Cite the product Littlewood--Paley square-function inequality] The **product Littlewood--Paley inequality** for $L^p(\mathbb{R}^n)$, $1 < p < \infty$, states that there exists $A_{n,p} > 0$ depending only on $n$ and $p$ such that for every $f \in L^p(\mathbb{R}^n)$, \begin{align*} A_{n,p}^{-1}\,\|f\|_{L^p(\mathbb{R}^n)} \le \Bigl\|\Bigl(\sum_{\jmath} |\Delta_{\jmath} f|^2\Bigr)^{1/2}\Bigr\|_{L^p(\mathbb{R}^n)} \le A_{n,p}\,\|f\|_{L^p(\mathbb{R}^n)}. \end{align*} This inequality is obtained by iterating the one-dimensional Littlewood--Paley inequality $n$ times — once along each coordinate axis. The one-dimensional Littlewood--Paley inequality is itself a consequence of the [Marcinkiewicz Multiplier Theorem in Dimension One](/theorems/3190) applied to the symbol $\mathbb{1}_{I_j^\pm}$, or equivalently of the vector-valued Mihlin multiplier theorem applied with values in $\ell^2$. The iteration combines the one-dimensional inequality with vector-valued Mihlin multipliers acting in the remaining variables; since $L^p(\mathbb{R}^{n-1};\,\ell^2)$ is a UMD lattice for $1 < p < \infty$, the iteration goes through with constants depending only on $n$ and $p$. The full verification — together with the pointwise iterated-Stieltjes form of the inequality used below — is given in Stein, *Singular Integrals and Differentiability Properties of Functions*, Chapter IV §§5--6, and we take it as a cited input. [/step]

[step:Pointwise summation-by-parts bound on each dyadic rectangle] Fix $\jmath = (\jmath_1, \ldots, \jmath_n)$ and let $R_{\jmath} = I_{\jmath_1} \times \cdots \times I_{\jmath_n}$. Write $\xi_k^L < \xi_k^R$ for the left and right endpoints of $I_{\jmath_k}$, so $I_{\jmath_k} = [\xi_k^L, \xi_k^R)$ (with the obvious modification for negative-sign intervals). The fundamental theorem of calculus, applied iteratively in each coordinate $k = 1, \ldots, n$ to the restriction $m|_{R_{\jmath}}$, gives the **iterated Stieltjes representation**: \begin{align*} m(\xi)\,\mathbb{1}_{R_{\jmath}}(\xi) = \sum_{S \subseteq \{1, \ldots, n\}}\,(-1)^{|S^c|}\, \int_{\prod_{k \in S^c}[\xi_k, \xi_k^R]}\,\partial^{S^c}\!m(\eta)\,\Bigl|_{\eta_k = \xi_k\,(k \in S)}\,d\mathcal{L}^{|S^c|}(\eta_{S^c})\,\mathbb{1}_{R_{\jmath}}(\xi), \end{align*} where $\partial^{S^c}\!m$ is the mixed partial derivative in the variables indexed by $S^c$, evaluated as a Radon measure (the Marcinkiewicz hypothesis exactly guarantees this regularity). The decomposition is the standard $n$-fold telescoping identity for a function on the rectangle $R_{\jmath}$, expressing $m\,\mathbb{1}_{R_{\jmath}}$ as a sum over $2^n$ "boundary--interior" terms indexed by which coordinates remain free ($S$) versus which are integrated against the corresponding partial derivative ($S^c$); see Stein, Singular Integrals, Ch IV §6, Lemma 1. For each subset $S \subseteq \{1, \ldots, n\}$, write $T_{\jmath}^{(S)}$ for the Fourier multiplier with symbol given by the $S$-term of the above decomposition. Each $T_{\jmath}^{(S)}$ has the form \begin{align*} T_{\jmath}^{(S)} g(x) = \int_{\mathbb{R}^n}\,\Bigl[\,\int_{\prod_{k\in S^c}[\xi_k, \xi_k^R]} \partial^{S^c}\!m(\eta)\,d\mathcal{L}^{|S^c|}(\eta_{S^c})\Bigr]\,\mathbb{1}_{R_{\jmath}}(\xi)\, e^{i\xi \cdot x}\,\hat g(\xi)\,d\mathcal{L}^n(\xi). \end{align*} Integrating in $\xi_S$ first and applying the elementary $L^\infty$ bound $\|\mathbb{1}_{R_{\jmath}} \cdot (\text{interior integrand})\|_{L^\infty} \le A$ from the $|S^c|$-dimensional Marcinkiewicz hypothesis (uniform in the frozen $\xi_S$-coordinates), \begin{align*} |T_{\jmath}^{(S)} g(x)| \le A\,\bigl|\Pi_{R_{\jmath}} g(x)\bigr|, \end{align*} where $\Pi_{R_{\jmath}}$ is the Fourier projector onto $R_{\jmath}$ (i.e., $\Pi_{R_{\jmath}} = \Delta_{\jmath}$). Summing the $2^n$ terms, \begin{align*} |T_m \Delta_{\jmath} f(x)| \le 2^n\,A\,|\Delta_{\jmath} f(x)|. \end{align*} This is the load-bearing pointwise estimate: the multiplier $m\,\mathbb{1}_{R_{\jmath}}$, applied to $f$, is dominated pointwise (up to the dimensional constant $2^n\,A$) by the unweighted dyadic projection $\Delta_{\jmath} f$. [/step]

[step:Combine the pointwise bound with the product Littlewood--Paley square function] By Step 1, $T_m f = \sum_{\jmath} T_m \Delta_{\jmath} f$ in $L^p$ for $f \in \mathcal{S}(\mathbb{R}^n)$ (with the partial sums converging in $L^p$ once we have the bound). Each summand $T_m \Delta_{\jmath} f$ has Fourier support in $R_{\jmath}$, and the rectangles $\{R_{\jmath}\}$ are pairwise disjoint, so by the **upper** product Littlewood--Paley inequality applied to the function $T_m f = \sum_{\jmath} T_m \Delta_{\jmath} f$ (whose dyadic-rectangle pieces are precisely $T_m \Delta_{\jmath} f$), \begin{align*} \|T_m f\|_{L^p(\mathbb{R}^n)} \le A_{n,p}\,\Bigl\|\Bigl(\sum_{\jmath}|T_m\Delta_{\jmath}f|^2\Bigr)^{1/2}\Bigr\|_{L^p(\mathbb{R}^n)}. \end{align*} Inserting the pointwise bound $|T_m\Delta_{\jmath}f| \le 2^n A\,|\Delta_{\jmath}f|$ from Step 3, \begin{align*} \Bigl\|\Bigl(\sum_{\jmath}|T_m\Delta_{\jmath}f|^2\Bigr)^{1/2}\Bigr\|_{L^p} \le 2^n A\,\Bigl\|\Bigl(\sum_{\jmath}|\Delta_{\jmath}f|^2\Bigr)^{1/2}\Bigr\|_{L^p} \le 2^n A\,A_{n,p}\,\|f\|_{L^p(\mathbb{R}^n)}, \end{align*} using the **upper** product Littlewood--Paley inequality once more (this time on $f$ itself). Combining the two steps, \begin{align*} \|T_m f\|_{L^p(\mathbb{R}^n)} \le 2^n A\,A_{n,p}^2\,\|f\|_{L^p(\mathbb{R}^n)}. \end{align*} Setting $C_{n,p} := 2^n\,A_{n,p}^2$ — which depends only on $n$ and $p$ via the product Littlewood--Paley constant — we obtain \begin{align*} \|T_m f\|_{L^p(\mathbb{R}^n)} \le C_{n,p}\,A\,\|f\|_{L^p(\mathbb{R}^n)}, \qquad f \in \mathcal{S}(\mathbb{R}^n). \end{align*} [/step]

[step:Extend to $L^p$ by density and conclude] By density of $\mathcal{S}(\mathbb{R}^n)$ in $L^p(\mathbb{R}^n)$ for $1 < p < \infty$, $T_m$ extends to a bounded linear operator on $L^p(\mathbb{R}^n)$ with $\|T_m\|_{\mathcal{L}(L^p(\mathbb{R}^n))} \le C_{n,p}\,A$. The constant $C_{n,p}$ depends only on $n$ and $p$, completing the proof. [/step]