Exercise: Prove inequality using Riemann lower sum and integral

387 Views Asked by At

Let $\;F:\mathbb R^m \rightarrow \mathbb R\;$ be a continuous, non-negative function and $\;g:\mathbb R \rightarrow \mathbb R^m\;$ a continuous map. Consider $\;x_n \in \mathbb R\;$ a divergent sequence and $\;\varepsilon \gt 0\;$ such that:

$\;F(g(x))\ge \frac {\varepsilon}{2}\;\;\forall x\in I_n=(x_n-\frac{l^2}{2s},x_n-\frac{l^2}{2s})\;$ for some $\;l,s \gt 0\;$. Show that: For every $\;L\gt 0\;$

$\;\sum_{x_n \le L} {\frac{\varepsilon l^2}{s} \le \int_{0}^{x _n+\frac{l^2}{2s}} F(g(x)) \;dx }\;$.

My attempt:

By passing to a subsequence , I can assume that the intervals $\;I_n\;$ are disjoint. Then for every $\;I_n\;$ it holds, $\;\frac {\varepsilon}{2}(x_n+ \frac{l^2}{2s}-x_n+\frac{l^2}{2s}) \le \int_{x_n-\frac{l^2}{2s}}^{x_n+\frac{l^2}{2s}} F(g(x)) \;dx \Rightarrow \frac {\varepsilon l^2}{2s}\le \int_{x_n-\frac{l^2}{2s}}^{x_n+\frac{l^2}{2s}} F(g(x)) \;dx\;$.

At this point I've been stuck although I feel I'm pretty close to the solution.My questions are below:

  1. Can $\;I_n\;$ be considered as partition? I can't see how...
  2. Why does such $\;L\;$ exist?
  3. How can I conclude to $\;0\;$ instead of $\;x_n-\frac{l^2}{2s}\;$ in the integral?(I tried to change variable but then the whole integral changed)

I would appreciate any help. Hints are welcome too.

Thnaks in advance!

1

There are 1 best solutions below

0
On BEST ANSWER

I assume by divergent you mean $x_n \to +\infty$.

If you are, in fact, considering Riemann sums, the inequality probably should be

$$\sum_{x_n \leqslant L}\frac{\epsilon}{2} \frac{l^2}{s} \leqslant \int_0^{x_2 + \frac{l^2}{2s}}F(g(x)) \, dx $$

Consider $L$ such that $0 < x_1 \leqslant x_2 \leqslant L.$ If the intervals $I_1$ and $I_2$ do not overlap then

$$\int_0^{x_2 + \frac{l^2}{2s}}F(g(x)) \, dx \geqslant \int_{x_1 - \frac{l^2}{2s}}^{x_1 + \frac{l^2}{2s}}F(g(x)) \, dx +\int_{x_2 - \frac{l^2}{2s}}^{x_2 + \frac{l^2}{2s}}F(g(x)) \, dx \\ \geqslant \frac{\epsilon}{2}2\frac{l^2}{2s} + \frac{\epsilon}{2}2\frac{l^2}{2s} \\ = \epsilon \frac{l^2}{s} \\ = \sum_{x_n \leqslant L}\frac{\epsilon}{2} \frac{l^2}{s}.$$

This argument can be extended by induction to prove the inequality for $n > 2$.

If the intervals overlap then $x_2 - \frac{l^2}{2s} < x_1 + \frac{l^2}{2s}$ which implies $x_2 - x_1 < \frac{l^2}{s}$ and,

$$\int_0^{x_2 + \frac{l^2}{2s}}F(g(x)) \, dx = \int_{x_1 - \frac{l^2}{2s}}^{x_1 + \frac{l^2}{2s}}F(g(x)) \, dx +\int_{x_1 + \frac{l^2}{2s}}^{x_2 + \frac{l^2}{2s}}F(g(x)) \, dx \\ \geqslant \frac{\epsilon}{2}2\frac{l^2}{2s} + \frac{\epsilon}{2}\left(x_2 + \frac{l^2}{2s} - \left(x_1 + \frac{l^2}{2s}\right)\right) \\ = \epsilon\frac{l^2}{2s} + \frac{\epsilon}{2}(x_2 - x_1) \\ < \epsilon\frac{l^2}{s} \\ = \sum_{x_n \leqslant L}\frac{\epsilon}{2} \frac{l^2}{s}.$$

In this case the inequality could be violated. This case could be precluded if $l$ and $s$ were chosen such that intervals do not overlap.