0
$\begingroup$

Let $f:\Bbb{R}^n\to\Bbb{R}^n$ be a function of class $C^{1}$. We suppose that there exists $k\in ]0,1[$ such that $$\Vert g'(x)\Vert\leq k,\forall\;x\in\Bbb{R}^n$$ where $f(x)=g(x)+x,\forall\;x\in\Bbb{R}^n$

Prove that $(i)$ \begin{align}\langle f'(x)h,h\rangle\geq (1-k)\ \Vert h\Vert^2\end{align} and $(ii)$ \begin{align}\lim\limits_{\Vert x\Vert\to \infty}f(x)=\infty\end{align}

For the first:

I think we can use directional derivatives \begin{align}\langle \lim\limits_{t\to 0} \frac{f(x+th)-f(x)}t,h\rangle=\sum^{n}_{i=1}\frac{\partial f_{i}(x)}{\partial x_i}h^2_i\end{align} To me, directional derivatives may not be the only way to show this but I don't even seem to be finding my way through. Please, could someone help? Thanks for your time and efforts!

$\endgroup$
1
$\begingroup$

I do not know, if there is any way to use directional derivatives, but the following should work: We have $g = f - \def\id{\mathrm{id}}\id$ and hence for each $x \in \mathbb R^n$ that $$ g'(x) = f'(x) - \id \iff f'(x) = g'(x) + \id $$ Hence $$ \def\<#1>{\left<#1\right>} \< f'(x)h,h> = \<g'(x)h,h> + \|h\|^2 $$ By Cauchy-Schwarz $$ \<g'(x)h,h> \ge -\|g'(x)h\| \|h\| \ge -k\|h\|^2 $$ and hence $$ \<f'(x)h,h> = \<g'(x)h,h> +\|h\|^2 \ge (1-k)\|h\|^2 $$ For (ii), use (i). Note that for $x \in \mathbb R^n$ we have \begin{align*} \<f(x),x> &= \<f(0),x> + \int_0^1 \frac{d}{dt} \<f(tx),x>\, dt\\ &= \<f(0),x> + \int_0^1 \<f'(tx)x,x> \, dt\\ &\ge \<f(0),x> + \int_0^1 (1-k)\|x\|^2\, dt\\ &\ge -\|f(0)\|\|x\| + (1-k)\|x\|^2 \end{align*} By Cauchy-Schwarz \begin{align*} \|f(x)\| &\ge \frac{\<f(x),x>}{\|x\|}\\ &\ge -\|f(0)\| + (1-k)\|x\| \end{align*} and the result follows.

$\endgroup$
  • $\begingroup$ Thanks a lot! I'll go through the proof, ask questions before thicking! $\endgroup$ – Omojola Micheal Aug 23 '18 at 6:02
  • $\begingroup$ Please, can you explain how you got \begin{align*} \<f(x),x> &= \<f(0),x> + \int_0^1 \frac{d}{dt} \<f(tx),x>\, dt\\ \end{align*} and how did you arrive at this point? \begin{align*} &= \<f(0),x> + \int_0^1 \<f'(tx)x,x> \, dt \end{align*} $\endgroup$ – Omojola Micheal Aug 23 '18 at 6:51
  • $\begingroup$ The first is just the fundamental theorem of calculus. Consider the function $h(t) = \left<f(tx), x\right>$, then we have $$ h(1) - h(0) = \int_0^1 h'(t) \, dt $$ which is just the first statement. The second statement is derived from the chain rule, note that $$ h'(t) = \left< \frac{d}{dt} f(tx), x\right> = \left< f'(tx) x, x\right> $$ by the chain rule, as $f'(tx)$ is the outer and $x$ the inner derivative (note that $\frac{d}{dt}(tx) = x$). $\endgroup$ – martini Aug 23 '18 at 10:05
  • $\begingroup$ Thanks for the explanation! Everything is clear! $\endgroup$ – Omojola Micheal Aug 23 '18 at 11:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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