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 $N$ be a norm on $\mathbb R^2$, such that $N ( \mathbb Z^2) \subset \mathbb N $, where $\mathbb Z^2 =\{ (a,b)\mid a\mbox{ and }b \mbox{ are integers}\}$.

Help me to prove that for $u$, $v$ fixed vectors the following limit exists

$$\lim_{ t\to 0^+} \frac{N(u+tv)-N(u)}t. $$

share|cite|improve this question
Did you try to use the convexity of the norm? – Davide Giraudo Dec 18 '11 at 10:31
You can accept answers by clicking the tick symbol next to the answer you'd like to accept. People around here appreciate it if you accept one answer or the other from time to time. – Rudy the Reindeer Dec 18 '11 at 10:32
up vote 1 down vote accepted

I don't see where your additional assumption should be used. Are you trying to prove some stronger claim?

As pointed out in Davide Giraudo's comment, the key property to use here is convexity of the norm.

Let $X$ be any linear normed space.

For any $a\in[0,1]$ and $x,y\in X$ we get $$N(ax+(1-a)y) \le N(ax)+N((1-a)y)=aN(x)+(1-a)N(y),$$ i.e. norm is a convex function.

If we denote $g(t)=N(u+tv)$, then for any $a\in[0,1]$ and $x,y\in\mathbb R$ \begin{align*}g(ax+(1-a)y)&=N(u+axv+(1-a)ayv)\\ &=N(au+axv+(1-a)u+(1-a)ayv)\\ &=N(a(u+xv)+(1-a)(u+yv)) \\ &\le aN((u+xv))+(1-a)N(u+yv)\\ &=ag(x)+(1-a)g(y), \end{align*} i.e. $g$ is a convex real function.

The limit in the questions is precisely the right derivative of $g$ at the point $0$. It is known that every convex function has one-sided derivatives at each point, see e.g. Radulescu, Radulescu, Andreescu: Problems in real analysis, Problem 6.3.5, p.276 or Dudley: Real analysis and probability, Corollary 6.3.3., p.204. (And probably almost any textbook in real analysis, I've included simply the books which were among the first results when searching for convex derivative "one-sided" "real analysis").

Differentiability of norm (in the sense of directional, Frechet or Gateaux derivative) seems to be studied in many books. E.g. the following might be interesting for you: Pavel Drábek, Jaroslav Milota: Methods of nonlinear analysis, Example 3.2.9, p.120. The first part of that example answers your question, the argument is the same as I've given above; I've tried to include more details.

share|cite|improve this answer

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.