Convergence of smoothing operators, Schwarz seminorms and mistakes in the book Modern Fourier Analysis

59 Views Asked by At

The question is about exercise 8.3.2 part (c) of Grafakos's book Moder Fourier Analysis.

The hint seems to be wrong and as there has been a couple of situations like this with the book I want to make sure that I do not miss anything. I have tried to find a list of mistakes in Grafakos but there does not seem to be one.

The Problem (Exercise 8.3.2 part c): Let $\Phi$ be smooth, radial, positive function on $\mathbb{R}^n$ with support in $B(0,1)$ and integral $1$. Denote $\Phi_t (x) = t^{-n} \Phi (x/t)$ and $P_t (f) = \Phi _t * f$. Now let $T$ be a continuous linear operator from the space of Schwarz functions $\mathcal{S} (\mathbb{R}^n)$ to its dual $\mathcal{S} ' (\mathbb{R}^n)$. We want to show that $P_t T P_t (f) \to 0$ in $\mathcal{S} ' (\mathbb{R}^n)$ as $t\to \infty$, where $f$ is Schwarz with $\hat f$ vanishing near the origin.

Hint(from the book): Pair with a Schwarz function $g$ and use part (a)[$P_t(f) \to 0$ in $\mathcal{S}(R^{n})$ as $t\to \infty$] and the fact that all Schwarz seminorms of $P_t(g)$ are bounded uniformly in $t>0$.

My approach: First, let us see that the hint is wrong. Let $n=1$. It is not difficult to see that the second part of the hint is true. So show $\sup_{\xi } | ( \hat{ P_t (g) } )' (\xi) |$ is not bounded in $t$ which contradicts that the Schwarz seminorms of $P_t (g)$ are bounded uniformly in $t>0$.

Standard computations give $$\mathcal{F} (P_t (g)) ' (\xi) = (\hat \Phi (t\xi) \hat g(\xi) )' = t\hat \Phi '(t\xi) \hat g (\xi) + \hat \Phi (t\xi) \hat g ' (\xi).$$ Putting $\xi = \frac{1}{4} t^{-1}$ we get that the last expression is equal to $t\hat \Phi '(1/4) \hat g (t^{-1}/4) + \hat \Phi (1/4) \hat g ' (t^{-1}/4).$ Obviously, for certain admissible choices of $g$ and $\Phi$ this expression is not bounded.

On the other hand, we can show that not only $P_t (f) \to 0$ in $\mathcal {S} (\mathbb{R} ^n)$ as $t\to 0$ but the convergence happens $o(t^M)$ for any $M>0$. Moreover, for arbitrary Schwarz $g$ we have $\rho_{\alpha, \beta} (P_t g) \leq C(g, \Phi) t^{|\alpha|}$, where $\rho$ denotes the Schwarz seminorm. In this case, we get something of the following form \begin{align*} | \langle P_t T P_t f, g \rangle | &= | \langle T P_t f, P_t g \rangle | \\ & \leq \left[ \sum_{|\alpha|, |\beta| \leq N} c_{\alpha, \beta} \rho_{\alpha, \beta} (P_t f)\right] \cdot \left[ \sum_{|\alpha|, |\beta| \leq N} c_{\alpha, \beta} \rho_{\alpha, \beta} (P_t g)\right] \\ &\leq C(f, \Phi, N) t^{-(N + 1)} \cdot C(g, \Phi, N) t^{N}. \end{align*} Thus, we get the desired result.

I'd appreciate if anyone would check this and maybe could provide any information on the existence of the list of mistakes in this book.