I've been asked to prove the following:

Assume $f(x)$ is continuous on $[-L,\ L]$ and $\int_{-a}^af = 0$ for all $a\in[0,\ L]$. Show that $f(x)$ is odd.

The catch? I'm only allowed to use Riemann sums.

Note: what makes this question different from most other ones I've read, such as this one, is that here I'm asked to prove $f(x)$ is odd, where in the others you are asked to prove the integral is $0$.

What I have so far is fairly straightforward: $$ \begin{aligned} \int_{-a}^a f(x)\text{d} x & = \int_0^af(x)\text{d} x + \int_{-a}^0 f(x)\text{d} x \\ & = \lim_{N\to\infty}\sum_{i=0}^{N-1}f\left(\frac{ia}N\right)\frac aN + \lim_{N\to\infty}\sum_{i=0}^{N-1}f\left(-\frac{ia}N\right)\frac aN \\ & = \lim_{N\to\infty}\left[\sum_{i=0}^{N-1}f\left(\frac{ia}N\right)\frac aN + \sum_{i=0}^{N-1}f\left(-\frac{ia}N\right)\frac aN\right] \\ & = \lim_{N\to\infty}\frac aN\sum_{i=0}^{N-1}\left[f\left(\frac{ia}N\right)+f\left(-\frac{ia}N\right)\right] \\ & = \lim_{N\to\infty}\frac aN\sum_{i=0}^{N-1}g\left(\frac{ia}N\right) \\ & = \int_0^ag(x)\text{d} x = 0\qquad\forall a\in[0,\ L] \end{aligned} $$

I was thinking that it would be easier to show that if $\int_0^af(x)\text{d}x = 0$ for all $a$ then $f(x)=0$, but I'm still unsure where to go from here (if this is the right route to take).

To be honest, I'm skeptical there's a solution involving just Riemann sums, since if you're given $\lim_{N\to\infty}\frac aN\sum_{i=0}^{N-1}f(ia/N) = 0$ for all $a$, then you'll only be able to conclude that $f(x)$ is zero except at a countable number of points. I'm not sure how you'd use the continuity of $f(x)$.

Any help would be appreciated.

  • $\begingroup$ Presumably you mean for all $a \in [0,L]$? $\endgroup$ – copper.hat Jan 25 '17 at 19:40
  • $\begingroup$ Yes indeed, thank you. $\endgroup$ – user3002473 Jan 25 '17 at 19:42

One possible approach:

Since $-\int_{-a}^0 f(x) dx = \int_0^a f(x) dx$, we have $\int_0^a (f(x)+f(-x)) dx = 0$ for all $a\in [0,L]$.

In particular, $\int_b^c g(x) dx = 0$ for all $0 \le b \le c \le L$, where $g(x) = f(x)+f(-x)$.

Now suppose $g(x) \neq 0$ and use continuity along with the Riemann sums to get a contradiction.

  • $\begingroup$ I considered that but couldn't find a contradiction specifically involving Riemann sums (of course, otherwise, finding a contradiction isn't difficult). $\endgroup$ – user3002473 Jan 25 '17 at 19:54
  • 1
    $\begingroup$ If $g(x_0) >0$ for example, then $g(x) > {1 \over 2} g(x_0)$ in some $b < x_0 < c$. Then $L(g,P) \ge {1 \over 2} g(x_0)(c-b)$ for all partitions and hence $\int_b^c g \ge {1 \over 2} g(x_0)(c-b) >0 $, a contradiction. $\endgroup$ – copper.hat Jan 25 '17 at 19:56
  • $\begingroup$ Huh, not what I had in mind but I suppose that works. Thanks! $\endgroup$ – user3002473 Jan 25 '17 at 19:57

Recall that a continuous function can't be zero except at a countable number of points. If a function $h$ is nonzero only on a countable number of points, then the set of points at which it is zero is dense. Fix $\epsilon > 0$. Let $|h(x)| > \epsilon$; there are $y$ arbitrarily close to $x$ so that $h(y) = 0$, so $|h(x) - h(y)| > \epsilon$, contradicting continuity. So $|h(x)| \leq \epsilon$. But $\epsilon$ was arbitrary, so $h(x) = 0$.

Now, you know $\int_0^ag(x)dx = 0$ for all $a$, so $g$ is zero almost everywhere. Since $g$ is also continuous, $g$ is zero everywhere.

  • $\begingroup$ Sorry, I don't understand: a continuous function can't be zero except at a countable number of points? What about a piecewise continuous function? Also, where in this do you use Riemann sums? $\endgroup$ – user3002473 Jan 25 '17 at 19:53
  • $\begingroup$ The first statement is incorrect. A continuous function can be zero on an interval. Take $f(x) = \max(0,x)$, this is zero on $(-\infty,0]$. $\endgroup$ – copper.hat Jan 25 '17 at 19:59
  • $\begingroup$ @copper.hat Not incorrect, just poorly phrased. I should have said "it can't be that a continuous function is only nonzero on a countable number of points." $\endgroup$ – Reese Jan 25 '17 at 20:26
  • $\begingroup$ @user3002473 This builds off of what you've already done. You used Riemann sums to show $\int_0^ag (x)dx = 0$; what I did was show that this means that $g $ must be zero. $\endgroup$ – Reese Jan 25 '17 at 20:35

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.