1

$f$ is defined on $[0,1]$ and $f(x) = 0$ for $x$ irrational and $f(x) = 1/q$ for $x = p/q$ in lowest terms, then $\int_0^1 f = 0$

I have found this Is Thomae's function Riemann integrable? and this $f(x)=1/q$ for $x=p/q$ is integrable but my proof is different, so I don't think this question is a duplicate.

My proof goes like this.

We already have that the lower integral is $0$, so it's enough to show that there exists partitions $P$ with $U(f,P) < \epsilon$ for any $\epsilon > 0$

For $t \in \mathbb{N}$ we consider the partition $$P_t = \bigg\{\bigg[\frac{i-1}{2^{2t}}, \frac{i}{2^{2t}}\bigg] \bigg| i \text{ is an integer between 1 and } 2^{2t}\bigg\}$$

What I'll do is to show that $U(f.P_t) < \frac{1}{t}$

If $M_i$ is the supremum of $f$ on the $ith$ interval, then either $M_i = f\bigg(\frac{i-1}{2^{2t}}\bigg)$ or $M_i = f\bigg(\frac{i}{2^{2t}}\bigg)$ because for any rational in that interval, expressed in lowest terms, $q > 2^{2t}$

Now we have the following:

$$ \sum_{i=1}^{2^{2t}} M_i = 1 + 2t$$ The proof is with induction.

For $t = 1$

$$ P_t = \{[0, 1/4], [1/4, 2/4], [2/4,3/4],[3/4,4/4] \}$$

Thus $ \sum_{i=1}^4 M_i = 1 + 1/2 + 1/2 +1 = 3 =1 +2t$

Suppose it holds for any $t$.

The partition $P_{t+1} = \bigg\{\bigg[\frac{i-1}{2^{2t + 2}}, \frac{i}{2^{2t+2}}\bigg] \bigg| j \text{ is an integer between 1 and } 2^{2t+2}\bigg\}$ is obtained in the following way.

For any interval $[\frac{i-1}{2^{2t}}, \frac{i}{2^{2t}}]$ of $P_t$, by multiplying by $4$ we get

$$\bigg[\frac{i-1}{2^{2t}}, \frac{i}{2^{2t}}\bigg] = \bigg[\frac{4(i-1)}{2^{2t+2}},\frac{4(i-1) + 1}{2^{2t+2}}\bigg] \bigcup \bigg[\frac{4(i-1)+1}{2^{2t+2}},\frac{4(i-1) + 2}{2^{2t+2}}\bigg] \bigcup \bigg[\frac{4(i-1)+2}{2^{2t+2}},\frac{4(i-1) + 3}{2^{2t+2}}\bigg] \bigcup \bigg[\frac{4(i-1)+3}{2^{2t+2}},\frac{4(i-1) + 4}{2^{2t+2}}\bigg] $$

$M_j'$ is the sup of $f$ on the $jth$ interval and without loss of generality, we'll assume that $M_i = f\bigg(\frac{1}{2^{2t}}\bigg)$

$$ \sum_{j = 4(i-1)}^{4i} M_j' = M_i + \frac{1}{2^{2t+1}} + \frac{1}{2^{2t+1}} + \frac{1}{2^{2t}} $$

The term $\frac{1}{2^{2t+1}}$ comes from the fact that if the numerator of one of the endpoint is even, then we can factor it out.

Therefore, the intervals of $P_{t+1}$ are between intervals of $P_t$ and we have

$$ \sum_{j = 1}^{2^{2t+2}} M_j' = \sum_{i = 1}^{2^{2t}} M_i + \frac{2}{2^{2t}} = \sum_{i = 1}^{2^{2t}} M_i + 2\sum_{i = 1}^{2^{2t}} \frac{1}{2^{2t}} = 1 + 2t + 2 = 1+ 2(t+1) $$

So the induction is closed.

Finally we get $$U(f,P_t) = \sum_{i=1}^{2^{2t}}\frac{M_i}{2^{2t}} = \frac{1}{2^{2t}} \sum_{i=1}^{2^{2t}} M_i = \frac{1+2t}{2^{2t}} < \frac{1}{t}$$

The last inequality can easily be proven with derivatives.

Is my proof correct?, is it possible to make it simpler? and if it's not, how can I fix it using this approach?

  • @ user807138 I don't understand what you are saying, the integral is equal to the infimum of all upper sums, the proof shows that the upper sums can be made as small as desired. So the infimum has to be equal to 0. it follows that the function is riemann/darboux integrable – Donlans Donlans Aug 03 '20 at 23:28

1 Answers1

1

I thinks it is simple. Let $\mathbb{X}$ is set of rational numbers in [0,1] and $\mathbb{Y}$ is set of irrational numbers in [0,1]. $\mu (A) $ is measure of set $A$. We know $\mu (\mathbb{X})=0$ and $\mu (\mathbb{Y})=1$. Lebesgue integral: $\int_{0}^{1} f(x) dx= \int_{\mathbb{X}}f(x)d\mu + \int_{\mathbb{Y}}f(x)d\mu =0$ Because

$\int_{\mathbb{X}}^{}f(x)d\mu =0$

ratatuy
  • 151