Show that $\operatorname{Lip}_\alpha $ is not an homogeneous Banach space

analysisbanach-spacesharmonic-analysis

This is an exercise of An Introduction to Harmonic Analysis of Yitzhak Katznelson:

Let $\operatorname{Lip}_\alpha(\mathbb{R}/2\pi\mathbb{Z}),\, \alpha \in (0,1)$ the Banach space of $2\pi$-periodic continuous functions such that
$$
r(f):=\sup_{t\in \mathbb{R}/2\pi \mathbb{Z},\,h\neq 0}\frac{|f(t+h)-f(t)|}{|h|^{\alpha }}<\infty
$$
with norm $\|f\|_{\operatorname{Lip}_\alpha }=\|f\|_\infty +r(f)$. Show that $\operatorname{Lip}_\alpha $ is not an homogeneous Banach space.

The statement to be proved is equivalent to show that there exists some $f\in \operatorname{Lip}_\alpha(\mathbb{R}/2\pi \mathbb{Z})$ such that
$$
\limsup_{\tau \to 0 }r(f_\tau -f)> 0
$$

for $f_\tau(x) :=f(x -\tau )$. I'm not sure how to continue with this exercise and I don't know some function that works as an example, some help will be appreciated.

Best Answer

I will use $a$ instead of $\alpha$. Let $T$ be the interval $[0,2\pi]$. Let $f$ be a function of the shape

  • $f(t)=t^a$ for $t\in[0,1]$,
  • $f$ is continuous and smooth on $(0,2)$ with $f'(2)=0$,
  • $f$ is zero on $[2,2\pi]$.

We identify $f:T\to \Bbb R$ (with $f(0)=f(2\pi)$) with a $2\pi$-periodic function on $\Bbb R$.

The quantity inside the supremum defining $r(f)$ is bounded, because (we may and do assume $h>0$, and then) the function $$ \begin{aligned} T^2&\to\Bbb R\ ,\\ (t,h)&\to\frac 1{h^a}(f(t+h)-f(t))&&\text{ for }t\in T\ ,\ h\in(0,2\pi]\ ,\\ (t,0)&\to\lim_{h\searrow0} \frac 1{h^a}(f(t+h)-f(t))=\delta_0(t)&&\text{ for }t\in T\ , \end{aligned} $$ is bounded. (It is continuous on a compact set obtained from $T^2$ obtained by removing a "small open (quarter)disc" with center in $(0,0)$. Then the control around $(0,0)$ comes from $(t+h)^a-t^a\le h^a$. This holds since for $t=0$ or $h=0$ this is clear, and else by $a$-homogenity it is enough to show $(1+h)^a\le 1+h^a$, i.e. $1+h\le 1+\frac 1ah^a\le (1+h^a)^{1/a}$.)


So $f$ is an element in the given Banach space. Let now $\tau>0$ be the number used for the shift of $f$. The shifted function $f_\tau$ is zero in a neighborhood of $0$, so $f-f_\tau=f$ in this small neighborhood. So for the special $t=0$ we have $$ \begin{aligned} r(f_\tau-f) &\ge \sup_{0< h}\frac{|(f_\tau-f)(0+h)-(f_\tau-f)(0)|}{h^a} \\ &= \limsup_{0< h}\frac{|f(0+h)-f(0)|}{h^a} \\ &= \sup_{0< h}\frac{h^a}{h^a} \\ &= 1\ . \end{aligned} $$

Related Question