Fourier Transform Lemma 2.3 from Stein Singular Integrals

85 Views Asked by At

I'm working through Stein Singular Integrals, and can't figure out how to prove this lemma (and find Stein's proof either wrong or un-readable). It is the following:

Lemma 3.3

Suppose $K:\mathbb{R}^n \rightarrow \mathbb{R}$ satisfies the following three properties: $$(a)\ \ \ |K(x)| \leq B|x|^{-n} $$ $$(b)\ \ \ \int_{|x|>2|y|}{|K(x-y) - K(x)|dx} \leq B\ \ \ \forall y\in\mathbb{R}^n $$ $$(c)\ \ \ \int_{R_1 < |x| < R_2}{K(x)dx} = 0\ \ \ \forall 0 < R_1 < R_2 < \infty $$

For $\varepsilon > 0$ define $K_\varepsilon(x) = K(x)\chi_{B_\varepsilon}(x)$ where $\chi_{B_\varepsilon}$ is the characteristic function of the ball $B_\varepsilon$ of radius $\varepsilon$ centered at the origin. Then we have

$$ \sup_{y\in\mathbb{R}^n}{|\mathcal{F}(K_\varepsilon)(y)|} \leq CB $$

where the $\mathcal{F}$ is the Fourier Transform, and $C$ is a dimensional constant not depending on $K$ or $\varepsilon$.

(end of statement of Lemma 3.3)

First of all, I don't even think this is true unless we interpret the supremum as an essential supremum, but I'm not even concerned about that. We are to first prove it for the case $\varepsilon = 1$ and then use a scaling argument. This is fine with me, but I do not understand the case $\varepsilon = 1$. The main maneuver suggested by Stein is the following:

$$ \int_{\mathbb{R}^n}{e^{2\pi i y \cdot x} K_1(x) dx} = \cfrac{1}{2} \int_{\mathbb{R}^n}{e^{2\pi i y \cdot x} [K_1(x) - K_1(x-z)]dx} $$

where $z = \cfrac{y}{2|y|^2}$. Then, how I understand it is, we should split this into an integral over $\{ |x| > 2|z| \}$ and one over $\{ |x| \geq 2|z| \}$. The first can be estimated with the help of assumption (b), but I do not seem to agree with Stein that the second can be estimated, or I mis-understood what he wrote. If anyone can give a reference or a hint in the right direction, I'll appreciate it. I've been tinkering with this for far too long, so if I can't get help, I'm going to have to leave this theorem alone...