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

Fundamental lemma in calculation of variation states that $\int_a^b f(x)g(x) dx =0$ for $f$ continuous in $(a, b) $ and and $g(x)$ which has compact support and is continuous in $(a,b)$ , then $f(x)=0$ $\forall x\in (a,b)$ . I found out the proof of this theorem is easy . But i have some doubts related to it if i want to extend this concept in heigher dimensions and to different class of functions .

a) Instead of interval $(a,b) $ if i take a set $\Omega \subset R^n$ , then what do i have to think about here, to prove the above theorem ?

b) Is it necessary that $g(x)$ should be infinitely differentiable and compactly supported ?

c) If $\int_a^b f(x) g'(x) =0$ with $g(x) , f(x) \in C^1[a,b]$ and $g(a)=g(b)=0$ . Then $f(x)$ is constant but does it still remain the same if i take $f(x)\in C^0[a,b]$ and $g(x) \in C_c^1[a,b]$ ?

A bit vague question : If i want to extend this idea to large class of functions what are the important things that i need to go about ? Thank you for your help .

What i am thinking about $c$ is to use mollification but the problem is it will be valid only almost everywhere, which i am not interested . is it possible after all ?

share|cite|improve this question
I can't quite follow what you're asking in (c). But see my response about the other two. – coffeemath Oct 18 '12 at 2:25

It needs to be included in the theorem statement that the statement $\int_a^b f(x)g(x) dx =0$ is to hold for $every$ allowable $g$. The idea of it is to fix a particular point $x$ in the interior of the domain of $f$, and based on that, and how $f$ behaves near $x$, you then select your $g$ to have a bump nearly 1 at $x$, and tapering off quickly enough that the behavior of $f$ away from $x$ puts only a negligible amount into the integration of $f(x)g(x)$ over the region.

When I've seen it applied, $g$ is taken to be actually zero outside of a ball about the point $x$. Then the size of this ball is shrunk to zero, at the same time keeping the $g$'s each individually continuous. After all this, if $f(x)$ were not zero we'd get a contradiction in the limit.

I gather you understand what I just wrote already. If so then it seems for your question (a) it should be clear it works in $n$ dimensions, especially if you use boxes instead of balls around the $x$ for ease of notation/calculation.

For $b$ the answer is: absolutely not! We can select our $g$ ourselves to be anything we want, as long as the relevant properties of integration of the product work. Certainly continuous is enough. And I guess the compact support hypothesis is there so the product integrals exist. If the given funcion $f$ has an unbounded domain, then integrating it against another function of unbounded domain might give no value. But if $g$ has compact domain the integration of the product $f(x)g(x)$ can be done over that domain, and so will converge provided $g$ is also continuous. [It's already assumed that $f$ is continuous for this application]

The only place I ever heard talk of integrating against functions which had to be $C^\infty$ with compact support was in a quite abstract course about distribution theory, in which things like the Dirac delta function had rigorous meaning. That course was over my head at the time...

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.