I have a continuous function $f$ from $[0,1]\to\mathcal{L}(\Bbb{R}^n,\Bbb{R}^p)$ such that $f(t)=L_t.$

Assume that $L_0$ is one ton one, I would like to prove that $L_t$ is one to one for $t$ sufficiently small.

Set $\varepsilon>0$ there exists $\gamma>0$ such that for any $t\in[0,\gamma)$ we have $$\Vert L_t-L_0\Vert<\varepsilon$$

Now for any $x\in \Bbb{R}^n$ I have $\Vert (L_t-L_0)(x)\Vert\le \Vert L_t-L_0\Vert \Vert x\Vert$

In fact, I am not sure how can I use the linearity here.

If $x\ne0$ I can choose $\varepsilon=\frac{1}{\Vert x\Vert}$, I will have $\Vert L_t(x)-L_0(x) \Vert<1$ for $t\in [0,\gamma)$, not sure that help.

Geometrically I cannot see why $L_t$ will be one to one, I can imagine that $L_t(x)$ is equal to $L_0(t)+\vec{v}$ and the norm of linear map can be visualize as the taking the biggest vector on the unit sphere.

Question: Can someone give a nice view of why is going to be one to one ?


Notice first that that there exists $c$ such that $\|L_0(x)\| > c > 0$ for all vectors $x$ such that $\|x\| =1$. This holds since $L_0$ is linear (which implies $L_0(0) = 0$), $L_0$ is one-to-one, and the set $\{ x : \|x\| = 0\}$ is closed.

Thus linearity implies that $\|L_0(x)\| \geq c \|x\|$ for any vector $x$.

Continuity says that there exists $\delta>0$ such that if $t < \delta$, then $\|L_t - L_0 \| \leq c/2$.

Then for $t < \delta$, we have \begin{align*} \| L_t(x) - L_t(y) \| &= \|L_t(x-y)\| \\ &\geq \| L_0(x-y)\| - \|[L_t -L_0](x-y) \| \\ &\geq c\|x-y\| - c/2(\|x - y\|) \\ &= c/2\|x-y\|\\ & > 0 \qquad \qquad \qquad \text{ if } x \neq y, \end{align*} so $L_t$ is one-to-one.

  • $\begingroup$ Your solution is a bit confusing, my problem is that I don't see where you are using the fact that $L_t$ is linear, it's seems that your proof will works for any continuous maps. $\endgroup$ – Alex Sep 23 '16 at 16:56
  • $\begingroup$ Apologies -- the proof I gave earlier was wrong. I believe it's correct now. And it now uses linearity extensively. $\endgroup$ – ec92 Sep 25 '16 at 3:37

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.