Let $f:\Bbb{R}^n\to \Bbb{R}$ be a differentiable function such that for some $p>1,$
\begin{align}f(tx)=t^pf(x),\;\forall \;t>0,\;\&\;x\in \Bbb{R}^n \qquad (1)\end{align}
$i.$ I want to show that \begin{align}f'(x)(x)=pf(x),\;\forall \;x\in \Bbb{R}^n \qquad (1)\end{align}
$ii.$ Is the converse also true?
I believe that we can take \begin{align}\varphi(t)=f(tx),\;\forall \;t>0\end{align} and show that \begin{align}\frac{d}{dt}\big(\frac{\varphi(t)}{t^p}\big)=0,\end{align} but I don't know how to tranform this to a proof. Any help please?
Answer
We can assume $n=1$ by restriction to an arbitrary line. Then
$$
f'(x) = \lim_{\epsilon\to 0}\frac{f((1+\epsilon)x)-f(x)}{\epsilon x} = \lim_{\epsilon\to 0}\frac{[(1+\epsilon)^p - 1]f(x)}{\epsilon x} = \frac{f(x)}{x}\frac{d}{d\epsilon}(1+\epsilon)^p|_{\epsilon=0} = p\frac{f(x)}{x}
$$
as desired. Note differentiability is guaranteed by this condition.
Similarly, if $\frac{df}{dx} = p\frac{f}{x}$ then $\frac{df}{f} = p\frac{dx}{x}$. Integrating both sides says
$$
\log f(x) - \log f(1) = p\log x = \log x^p
$$
so
$$
f(x) = C x^p,
$$
which provides the converse.
No comments:
Post a Comment