5
$\begingroup$

Evaluate $\int ^1 _0 x^2 \, dx$ without using the Fundamental Theorem of Calculus.

I can find the Riemann sum of some partition and the intermediate points, what should I do next?

$\endgroup$
  • $\begingroup$ If you use a uniform partition of size $1/n$ and take Darboux upper and lower sums, you'll end up with a known sum of squares, and ultimately you'll get a limit of $1/3$. You do need to argue that the integral exists, too. $\endgroup$ – Pedro Tamaroff Dec 2 '13 at 14:44
  • $\begingroup$ Riemann sums are far from the most elegant way to do this problem, but I suspect that most people who teach calculus don't even suspect that. $\endgroup$ – Michael Hardy Dec 2 '13 at 16:28
10
$\begingroup$

$$\text{As }\lim_{n \to \infty} \frac1n\sum_{r=1}^n f\left(\frac rn\right)=\int_0^1f(x)dx$$

$$\int_0^1x^2dx=\lim_{n\to\infty} \frac1n\sum_{r=1}^n \frac{r^2}{n^2}=\lim_{n\to\infty} \frac1{n^3}\sum_{r=1}^nr^2$$

$$\text{Now }\sum_{r=1}^nr^2=\frac{n(n+1)(2n+1)}6$$

So, $\displaystyle\lim_{n\to\infty}\frac{n(n+1)(2n+1)}{6n^3}=\lim_{n\to\infty}\left(\frac16\right)\left(1+\frac1n\right)\left(2+\frac1n\right)=\frac13$

$\endgroup$
  • $\begingroup$ I wonder if you can explain the third line for me please? How do you got $∑_{r=1} ^n r^2=(n(n+1)(2n+1))/6$ $\endgroup$ – Diane Vanderwaif Dec 2 '13 at 15:05
  • 1
    $\begingroup$ @DianeVanderwaif, math.stackexchange.com/questions/48080/… $\endgroup$ – lab bhattacharjee Dec 2 '13 at 15:06
  • $\begingroup$ why there is no 6 there on last line from previous line? $\endgroup$ – Aleksey Dec 2 '13 at 17:22
  • $\begingroup$ @Aleksey, I willingly kept out $\frac16$ from the limit calculation which is legal as it is independent of $n$ $\endgroup$ – lab bhattacharjee Dec 3 '13 at 3:19
10
$\begingroup$

A cute way to do it while avoiding sums of squares is to use a geometric progression. (The same method can easily be adapted to $\int_0^1 x^n\,dx$ for other choices of $n$ without having to cook up formulas for $\sum k^n$.)

By general theory, $x \mapsto x^2$ is Riemann integrable, so we can compute the integral as a limit of any Riemann sum, as long as the "size of the partition" tends to $0$.

Fix an integer $n$ and choose points $x_j = \dfrac{r^j}{n}$ ($0 \le j \le n$) where $r > 0$ is chosen so that $r^n = n$.

Then a corresponding Riemann sum is \begin{align} \sum_{j=0}^{n-1} f(x_j)(x_{j+1}-x_{j}) &= \sum_{j=0}^{n-1} \frac{r^{2j}}{n^2}\left( \dfrac{r^{j+1}}{n} - \dfrac{r^{j}}{n} \right) \\ &= \frac{1}{n^3}\sum_{j=0}^{n-1} (r^{3j+1}-r^{3j}) \\ &= \frac{1}{n^3} \cdot \frac{r^{3n+1}-r^{3n}-r+1}{r^3-1} \\ &= \frac{1}{n^3} \cdot \frac{(n^3-1)(n^{1/n}-1)}{n^{3/n}-1}. \end{align}

I'll leave it as an exercise to check that this tends to $\dfrac13$ as $n\to\infty$.

$\endgroup$
  • $\begingroup$ I have liked this method since I first learned of it. It seems much underappreciated. $\endgroup$ – Harald Hanche-Olsen Dec 2 '13 at 15:10
  • $\begingroup$ @HaraldHanche-Olsen I agree. The idea is apparantly due to Fermat, so even predating Newton and Leibniz (not to mention Riemann). $\endgroup$ – mrf Dec 2 '13 at 19:29

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.