2

The question is from Mathematical Analysis, Douglass pg.286 Exercise 18.

Define a function $f(x)$ on $[0, 1]$ such that $f(x)=x$ if $x$ is rational, and $f(x)=0$ if $x$ is irrational. Find $L(f)$ and $U(f)$, and determine whether $f$ is in $R[a, b]$.

My attempt:

For any partition $\pi=\{0=x_0, x_1, ...x_p=1\} \in \Pi[0, 1]$, there exists an irrational between every interval $[x_{j-1}, x_j]$, so the minimum of each interval would be $0$, so we could conclude that $L(f)=0$. $U(f)$ is the part that I am having some hard time with.

Take the same partition as above, but in this case, take such that each $x_i$'s are rational. Then, for each interval $[x_{j-1}, x_j]$, the supremum of $f$ in this interval would be the largest rational, which is $x_j$. Then, $$U(f, \pi)=\sum_{j=1}^{p}M_j\Delta x_j=\sum_{j=1}^{p}x_j(x_j-x_{j-1})$$ At this stage, I struggled to get an inequality that would make me conclude that $L(f)\neq U(f)$, but I failed to do it. I saw that $U(f)=\frac{1}{2}$ from another article, and I'm not sure whether this specific partition would do me the job. I plugged in about 5 to 6 small partitions of $[0, 1]$ and it seems that it would do the job, but I'm not sure how to prove it. Can someone help me out with this proof?

Jose Avilez
  • 13,432
Joshua Woo
  • 1,373
  • 1
    Wait....I think I got it, can I just assume that $U(f, \pi)=\sum_{j=1}^{p}x_j(x_j-x_{j-1})$ is a riemann sum for $y=x$ in the interval $[0, 1]$, and use the fact that $\int_{0}^{1}xdx=\frac{1}{2}$? Is this a valid way? – Joshua Woo Jun 09 '21 at 14:07
  • https://math.stackexchange.com/questions/437711/is-dirichlet-function-riemann-integrable/2388783 – Dhanvi Sreenivasan Jun 09 '21 at 14:07
  • @DhanviSreenivasan I haven't learnt the fact that $f$ is Riemann integrable if and only if the function is bounded and continuous almost everywhere, though. I think I need to prove this just by definition and Riemann condition. – Joshua Woo Jun 09 '21 at 14:11
  • If you are familiar with the notion of sets of measure zero, there is a well known result by Lebesgue that may be used to settled these type of questions. The result states that $f$ is Riemann integrable in a closed interval iff $f$ is bounded and continuous almost everywhere on $[a, b]$. The function in your OP is discontinuous everywhere (except at $0$) – Mittens Jun 09 '21 at 14:12
  • @OliverDiaz Thanks for your comment. Sorry but I haven't learned about Lebesgue measure. – Joshua Woo Jun 09 '21 at 14:16

2 Answers2

4

Given any partition $P$ with points $0=x_0< x_1 < \ldots < x_n = 1$ we have $\sup_{x\in [x_{j-1},x_j]} f(x) = x_j$ for $j=1,2,\ldots, n$. This holds whether the endpoints are rational or irrational, as a consequence of the density of the rationals. Consequently, it follows that

$$U(f,P) = \sum_{j=1}^n x_j(x_j - x_{j-1})$$

As you observed, this is an upper Darboux sum for the function $x \mapsto x$ on $[0,1]$ and for any partition $P$ we have

$$U(f,P) \geqslant \inf_{P'} U(f,P') = \int_0^1 x \, dx = \frac{1}{2}$$

It is also not difficult to show that $U(f,P) > \frac{1}{2}$ directly. Note that $x_j > \frac{1}{2}(x_j + x_{j-1})$ and

$$U(f,P) = \sum_{j=1}^n x_j(x_j - x_{j-1})> \sum_{j=1}^n \frac{1}{2}(x_j+x_{j-1})(x_j - x_{j-1})\\ = \frac{1}{2}\sum_{j=1}^n (x_j^2 - x_{j-1}^2) = \frac{1}{2}(x_n^2 - x_0^2)= \frac{1}{2} (1-0) = \frac{1}{2}$$

As you showed, $L(f,P) = 0$ for any partition. Since $U(f,P - L(f,P) > \frac{1}{2}$ it follows that $f$ is not Riemann integrable.

RRL
  • 92,835
  • 7
  • 70
  • 142
1

Let $P$ be a partition of $[0,1].$ Then

$$U(f,P)\ge U(f,P\cup \{1/2\})$$ $$\ge U(f,(P\cup \{1/2\})\cap [1/2,1]))\ge \frac{1}{2}\cdot \frac{1}{2} =\frac{1}{4}.$$

This is true for any partition $P.$ Thus $U(f)\ge 1/4 >0 = L(f).$

zhw.
  • 107,943