Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $\mathcal{C}^1([a,b],\mathbb{R})$ be the vectorial space of functions of class $C^1$ (i.e. functions $f:[a,b]\to\mathbb{R}$ such that $f'$ is continuous) with a norm given by


Let $H\subset\mathcal{C}^1([a,b],\mathbb{R})$ the subset of injective functions prove that if $h\in H$ and there exists $x\in[a,b]$ such that $h'(x)=0$ $h\notin \operatorname{int}(H)$. In particular $H$ is not open.


If $h\in H$ then is a continuous injection therefore $h$ is a homeomorphism over some close interval. Given $\epsilon>0$ exist $f\in B(h,\epsilon)$ such that $f'=0$.

share|cite|improve this question
up vote 1 down vote accepted

Suppose $h \in H$ such that $\exists x_0$ with $h'(x_0) = 0$. Since $h$ is injective, it must be monotonic. Hence $h'$ must either be $\geq 0$, or $\leq 0$. I will assume that it is decreasing, the other case is handled similarly. Note that since $h$ is decreasing, we have $h(a) > h(b)$.

Let $\epsilon >0$, and define $h_{\epsilon}$ as $h_{\epsilon}(x) = h(x) + \epsilon (x-x_0)$. Clearly $h_{\epsilon}'(x) = h'(x) + \epsilon$, and we have $||h-h_{\epsilon}||_1 \leq \epsilon \max(|a-x_0|,|b-x_0|)+\epsilon$, so $h_{\epsilon}$ can be chosen arbitrarily close to $h$.

Suppose $x_0 \in (a,b)$. Then we have $h(a) > h(x_0) > h(b)$. Now choose $\epsilon$ small enough so that $h_{\epsilon}(a) > h_{\epsilon}(x_0) > h_{\epsilon}(b)$. Since $h_{\epsilon}'(x_0) = \epsilon >0$, we can conclude that there must be a point $x_1 \in (a,x_0)$ such that $h_{\epsilon}(x_1) = h_{\epsilon}(x_0)$, which means that $h_{\epsilon} \notin H$. Hence $h\notin H^{\circ}$.

If $x_0 = a$, choose $\epsilon$ small enough so that $h_{\epsilon}(a) = h_{\epsilon}(x_0) > h_{\epsilon}(b)$. By similar reasoning, we see that there must be a point $x_1 \in (x_0,b)$ such that $h_{\epsilon}(x_1) = h_{\epsilon}(x_0)$, with the same consequences.

The case $x_0=b$ is handled similarly.

share|cite|improve this answer
Nice construction! Thanks!! – Gastón Burrull May 26 '12 at 21:14

Suppose that given $h\in H$, there exists some $x\in [a,b]$ such that $h'(x)=0$.

Because $h$ is injective, we cannot have $h'\equiv 0$ on any open interval, so there is some open interval $(c,d)\subset [a,b]$ containing $x$ such that $|h'(t)|>0$ for all $t\in(c,d)$, $t\neq x$. Because $h'$ is continuous, $h'(t)$ has the same sign for all $t\in(c,x)$.

Because $h'$ is continuous and $h'(x)=0$, for any $\epsilon>0$ we can choose our interval $(c,d)$ containing $x$ to also have the property that $|h'(t)|<\frac{\epsilon}{4}$ for all $t\in (c,d)$.

For any $r,s$ with $c<r<x<s<d$, we can create a non-zero $C^\infty$ function $w:[a,b]\to\mathbb{R}$ supported on $[r,s]\subset (c,d)$ such that

  • $\|w\|<\epsilon$
  • $|w'(z)|>\frac{\epsilon}{2}$ for some $z\in (r,x)$
  • $w'$ has the opposite sign of $h'$ on $(r,x)$

Thus $$\|(h+w)-h\|=\|w\|<\epsilon,$$ and $h+w\in C^1([a,b],\mathbb{R})$, but the sign of $(h+w)'$ changes between $r$, where $$(h+w)'(r)=h'(r)+w'(r)=h'(r)+0=h'(r)$$ has the same sign as $h'$, and $z$, where $$(h+w)'(z)=h'(z)+w'(z)$$ has the opposite sign of $h'$ because $w'$ has the opposite sign of $h'$ on $(r,x)$ and $|w'(z)|>\frac{\epsilon}{2}>\frac{\epsilon}{4}>|h'(z)|$.

Thus $h+w$ is not injective, and thus $h+w\notin H$. Because, for any $\epsilon>0$, we can find some $h+w$ such that $\|(h+w)-h\|<\epsilon$ and $h+w\notin H$, the set $H$ is not closed in $C^1([a,b],\mathbb{R})$.

share|cite|improve this answer
Thanks for an interesting and simple proof. – Gastón Burrull May 26 '12 at 21:14

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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