MathOverflow is a question and answer site for professional mathematicians. 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

(I've edited this question)

I'm searching for a continuously differentiable function $f:\mathbb R^2\to\mathbb R$ such that $f(x)+f(x+u+v)\neq f(x+u)+f(x+v)$ for all $x$ and all linearly independent $u$ and $v$.

My original question was about the special case $x=0, f(x)=0$ for merely continuous functions, which turned out to be trivial.

(I was lead to this question when investigating whether one can always find the vertices of a parallelogram (or more specifically, a square) in the graph of a continuously differentiable function $f:\mathbb R^2\to\mathbb R$. The nonexistence of functions such as the above would imply that one cannot always find a parallelogram in the graph of a continuously differentiable function.)

share|cite|improve this question
I would suggest not introducing a name for that class of functions... – Mariano Suárez-Alvarez Feb 8 '10 at 18:55
I agree with Mariano - for example, it seems likely that the sum of two such maps will not necessarily be another such map, and so the collection of such functions will not form a vector space as we would usually want. – Zev Chonoles Feb 8 '10 at 19:09
Please do not edit the question so much that existing answers to it stop being answers. If you want to ask another question, ask another question! – Mariano Suárez-Alvarez Feb 8 '10 at 19:25
up vote 9 down vote accepted

For your new question: functions that satisfy your inequality don't exist


Suppose $f(x)+f(x+u+v)> f(x+u)+f(x+v)$ for all $x$ and all linearly independent $u$ and $v$. Let us get a contradiction from it. Take any square insribed in a circle, and rotate it leaving insrcibed. Rotating continuously for the angle 90 degrees you can exchange to pairs of opposite vertices.

Here is the previous counterexample:


share|cite|improve this answer
Sorry, I forgot the requirement f(0)=0. – Samuel Feb 8 '10 at 19:00
+10 if I could! +1 will have to do – HJRW Feb 8 '10 at 19:01
Samuel, idded 1 does not work anymore, but the other function works... – Dmitri Feb 8 '10 at 19:11

$F(v) = |v|$ works. To be more explicit, I mean $|(x,y)| = \sqrt{x^2 + y^2}$.

We have $F(u+v)^2 = F(u)^2 + F(v)^2 + 2 F(u) F(v) \cos \theta$, where $\theta$ is the angle between $u$ and $v$. When $u$ and $v$ are linearly independent, $\theta$ is nonzero, so $$F(u+v)^2 = F(u)^2 + F(v)^2 + 2 F(u) F(v) \cos \theta < F(u)^2 + F(v)^2 + 2 F(u) F(v)$$ and $$F(u+v) < F(u) + F(v).$$

share|cite|improve this answer
One can say that the triangle inequality happens only if the points are colinear. – Hailong Dao Feb 8 '10 at 19:15
Any strictly convex norm on $\mathbb{R}^2$ will do, because then the triangle inequality is strict $\|x+y\| \lt \|x\|+\|y\|$ whenever $x$ and $y$ are linearly independent. – Konrad Swanepoel Feb 8 '10 at 19:16

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.