I am trying to find a function $f: \mathbb{R}^+ \to \mathbb{R}^+$ that fullfils the following conditions

  • $$f \in \mathcal{C}^1(\mathbb{R}^+,\mathbb{R}^+)$$

  • $$\int_{\mathbb{R}^+} f \in \mathbb{R}^+$$

  • $$\int_{\mathbb{R}^+} \mid f' \mid \in \mathbb{R}$$

  • $$f \text{is not $\frac{1}{2}$-Hölder}$$

I've tried functions with smooth spikes but I am unable to express this function as combinations of usual functions.

Moreover, It's worth noticing that if we have the assumption : $f'^2$ is integrable then $f$ is necessarily $\frac{1}{2}-$Hölder :

We have (using CS) : $$ \mid f(x) -f(y) \mid \leq \int_x^y 1 \times f' \leq \sqrt{\int_x^y f'^2}\sqrt{y-x} $$ Hence it follows that $f$ is $\frac{1}{2}-$Hölder continous since that $\sqrt{\int_x^y f'^2}$ is bounded.


Indeed, disjoint tiny smooth spikes, of small heights and even much-much smaller widths, will do.

Let $K\in C^\infty(\mathbb R)$ be such that $K\ge0$, $K(x)=0$ if $|x|>1/2$, and $a:=K(1/3)-K(0)\ne0$. For instance, we may take $K(x)=\exp\frac1{4x^2-1}$ if $|x|<1/2$ and $K(x)=0$ if $|x|\ge1/2$.

Let $c:=\int_{\mathbb R} K<\infty$ and $b:=\int_{\mathbb R}|K'|<\infty$. For $j=2,3,\dots$, let $h_j:=1/j^2$, $d_j:=1/j^6$, and \begin{equation} f_j(x):=h_j K\Big(\frac{x-j}{d_j}\Big). \end{equation} Let \begin{equation} f:=\sum_2^\infty f_j. \end{equation} Then $f\in C^\infty(\mathbb R)$, $f\ge0$, \begin{equation} \int_{\mathbb R}f=\sum_2^\infty h_j d_j c<\infty, \end{equation} and \begin{equation} \int_{\mathbb R}|f'|=\sum_2^\infty h_j b<\infty. \end{equation} Yet, for $j=2,3,\dots$ \begin{equation} f(j+d_j/3)-f(j)=f_j(j+d_j/3)-f_j(j)=h_j a=a/j^2, \end{equation} which latter is much greater in absolute value than $\sqrt d_j=1/j^3$ as $j\to\infty$. So, $f$ is not $1/2$-Hölder-continuous.

  • $\begingroup$ Great answer. It's quite impressive that adding the assumption : $f'^2$ is integrable prevent these kind of counter-examples. Do you have any ideas why (on an intuitive level) the integrability of $f'^2$ prevent the existence of a function with smooth spikes, of small heights and even much-much smaller widths which is not $1/2$-Hölder ? Thank you. (+1) $\endgroup$ – Thinking Nov 15 '18 at 12:00
  • 1
    $\begingroup$ @Thinking : Thank you for your comment. Concerning your question: This is a matter of scale. Indeed, in the example presented in my answer, given the height $h_j$, the derivative $f'_j\asymp h_j/d_j$ of the $j$th peak $f_j$ is inversely proportional to its "width" $d_j$ and hence large if $d_j$ is much smaller than $h_j$, but $|f'_j|$ is integrated over a small interval of length $d_j$, so that the integral $\int|f_j'|=h_j b$ is small (and does not depend on $d_j$). However, $(f'_j)^2$ is much greater than $|f_j|$, so that $\int(f_j')^2$ is on the order of $(h_j/d_j)^2d_j=j^2\to\infty$. $\endgroup$ – Iosif Pinelis Nov 15 '18 at 14:04
  • $\begingroup$ Indeed this make a lot of sense. Thank you for the detailed comment ! $\endgroup$ – Thinking Nov 15 '18 at 22:31

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.