I am defining a sequence of functions $f_n$ as follows:

$f_n(x) = nx$ if $x \leq \frac{1}{n}$, 1 if $x > \frac{1}{n}$.

I try computing a lower Darboux integral of $f_n(x)$ on [0,1] (like lower riemann integral but infimum of f on subinterval of partition rather than its value at midpoint). I use a partition of equally spaced points which define subintervals of length $\frac{1}{m}$ where $m < n$.

I arrive at an expression of the form: $(\sum_{i=1}^{i=k}\frac{1}{m}\frac{i-1}{m}n)$ + $\frac{n-k}{m}$. Where $k$ is the smallest integer such that $k \geq \frac{m}{n}$.

I tried to find an upperbound by setting $k = 1 + \frac{m}{n}$. And I arrive at an expression : $\frac{3n + 3m}{2nm} + \frac{n}{m}$. My conceptual problem is that my understanding is that as $m \rightarrow \infty$ the partition gets finer so the lowerbound approximation to the integral should get bigger. Yet I'm not sure about that from what I have. It seems as if $m\rightarrow \infty$ my expression tends to $\frac{3}{2n}$.

How would I use the definition of Riemann(darboux) integration to compute the integral of the functions $f_n$?

Thank you.

  • $\begingroup$ What interval are you considering doing integration on? $\endgroup$ – xbh Sep 14 '18 at 2:45
  • $\begingroup$ Sorry, [0,1], will edit my question. $\endgroup$ – trynalearn Sep 14 '18 at 2:47

First, I find it personally helpful to do some visualization, below are some examples of $f_n(x)$

enter image description here enter image description here enter image description here enter image description here

From these we can develop some useful intuition. First notice that $f_1(x) = x$ $(\textrm{for } x \in [0,1])$. Also, observe that as $n \rightarrow \infty$ we must have that $$\int_{[0,1]} f_n(x) \rightarrow 1$$ since the portion of the area under the curve taken up by when $x \leq n^{-1}$ goes to zero as $n$ tends to infinity and $$\int_{[0,1]}dx =1.$$ Define $x_i = \frac{i}{m}$, then I have $m-1$ subintervals of the form $[x_i, x_{i+1}]$, where $0 \leq i \leq m$. Observe that by the definition of $f_n(x)$ we have that $$\inf_{x \in [x_i,x_i+1]} f_n(x) = f_n(x_i),$$ as $f_n$ is either increasing or constant. so $$\int_{[0,1]}f_n(x)dx = \lim_{m\rightarrow \infty}\sum_{i=0}^{m-1}f_n(x_i)\Delta x$$ which is simply the Riemann Integral with left hand endpoints. For example, if I wanted to approximate $f_5$ with $m = 4$, then I would have the following situation: $$\int_{[0,1]}f_5(x) \approx f(0)\cdot\frac{1}{4} + f\left(\frac{1}{4}\right)\cdot\frac{1}{4} + f\left(\frac{1}{2}\right)\cdot\frac{1}{4} = 0 + 1\frac{1}{4} + \frac{1}{4} = \frac{1}{2}$$

What happens if I let $m = 10$? $m = 20$? I recommend doing this to develop intuition. In any case we can see that this is just the left hand sided Integral, and as $f_n$ is either increasing or constant this will be an underestimate, so yes the lower-bound estimate will increase as I make my partition finer.

As to explicitly calculating this integral. If you want to do it rigorously one can simply use the usual techniques for proving Riemann Integrals have a certain value while realizing that for particular $f_n$ $$\int_{[0,1]} f_n(x)dx = \frac{1}{2n} + 1 - \frac{1}{n} $$


Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.