Mikhlin multiplier theorem

1.5k Views Asked by At

The Mikhlin multiplier states the following:

Let $m : \mathbb{R}^{n} \backslash \{0\} \rightarrow \mathbb{C}$ satisfy the following:

$$|\partial^{\alpha}m(\xi)|\leq C_{0}|\xi|^{-|\alpha|}, \forall \alpha \in \mathbb{N}_{0}^{n} \text{ i.e. alpha is a multi-index with } |\alpha| \leq n+2.$$ $$\text{ Then, for all } 1 < p < \infty, \exists B=B(m,n,p) > 0 \text{ such that } ||T_{m}f||_{L^{p}} \leq B||f||_{L^{p}}, \forall f \in \mathcal{S}(\mathbb{R}^{n}).$$

In proving this one considers a partition of unity as follows $\Psi \in C_{c}^{\infty}(\mathbb{R}^{n}),$ with support of $\Psi \subseteq \mathbb{R}^{n} \backslash \{0\}$ and $ \sum_{j=- \infty}^{\infty} \Psi(\frac{x}{2^{j}}) =1 \,\forall x \neq 0$. This can be choosen radial and non-negative.

Then defining $\Psi_{j}(x):= \Psi(\frac{x}{2^{j}})$ for $\Psi$ as above and observing that each $\Psi_{j}$ is supported in a dyadic annulus of size $2^{j}$ and given $j \in \mathbb{Z}, m_{j}(\xi):=\Psi_{j}(\xi)m(\xi)$

How the The Fourier multiplier of the above $m_{j}(\xi)$ is $m_{j}(D)$ How do we obtain $m(D)f = \sum_{j}m_{j}(D)f$ with convergence in $L^{2}$ (I know we have to use Plancherel and the Dominated Convergence Theorem somehow)

1

There are 1 best solutions below

5
On

As you mentioned, we can first apply Plancherel's theorem to work in the frequency variable. We now need to show that $$m(\xi) \hat{f}(\xi) = \sum_j m_j(\xi) \hat{f}(\xi).$$ Since $\sum_j m_j(\xi)$ converges to $m(\xi)$ pointwise a.e., it suffices to show that $\sum_j m_j(\xi) \hat{f}(\xi)$ is bounded pointwise by some $L^2$ function. Now, by the decay assumption on $m$, $m \in L^\infty$, so $m(\xi) \hat{f} \in L^2$. Since the $m_j$ are nonnegative and sum to $m$ pointwise, it follows that $$\sum_{j = -N}^N m_j \hat{f} \leq m\hat{f}\qquad \text{a.e.}$$ so since $m\hat{f} \in L^2$, we can conclude.