0

I am going through my analysis-notes from when I was a student. I did not solve the following exercise back then and wanted to give it a try:

Determine minimum, maximum, infimum/supremum of the set $C = \{\frac{n-m}{n+m} \vert n,m \in \mathbb{N}_0\}$. Here $\mathbb{N}_0$ denotes the positive naturals

First of all, I rewrote the fractions as $1 - \frac{2m}{n + m}$. Hence, $1$ is an upper bound. Also, $\frac{-2m}{n+m} > -2$, so $-1$ is a lower bound.

By the supremum property of $\mathbb{R}$, I know that both supremum and infimum exist. I think they are 1 and -1. I have trouble prooving this.

Also, these bounds can not be attained (otherwise either $n$ or $m$ has to be zero), so there is no minimum/maximum.

Any hints on how to show that sup(C) = 1 and inf(C) = -1?

Based on Omnomnomnom's hint is my proof that the supremum equals 1:

Suppose that $\alpha < 1$ is an upper bound for $C$, then $\alpha \geq 1 - \frac{2m}{n+m}$ for all $n,m \in \mathbb{N}_0$. However, there exists $n_1 \in \mathbb{N}_0$ such that $$\frac{1}{n_1 + 1} < \frac{1}{n_1} < \frac{1 - \alpha}{2}.$$ Hence we have that $$\frac{2}{n_1+1} < 1 - \alpha \Rightarrow 1 - \frac{2}{n_1+1} > \alpha.$$ This is a contradiction ($n = n_1, m = 1$). Therefore, $\operatorname{Sup}(C) = 1$.

Proof that the infimum equals -1:

Suppose $\alpha > -1$ is a lower bound. Then $1 - \alpha < 2$. Therefore $\frac{1 - \alpha}{2} < 1$. There exists $m_1 \in \mathbb{N}_0$ such that $\frac{1}{m_1 +1} + \frac{1 - \alpha}{2} < 1$. This however implies that $$\frac{1 - \alpha}{2} < 1 - \frac{1}{m_1 + 1} = \frac{m_1}{m_1 + 1}.$$ Therefore, $$1 - \alpha < \frac{2m_1}{m_1 + 1}$$ from which it follows that $1 - \frac{2m_1}{m_1 + 1} < \alpha$, which is a contradiction (take $m = m_1, n = 1$).

Student
  • 4,438
  • Student: $m,n \in \mathbb{Z^+}$?, the positive naturals? – Peter Szilas Apr 14 '20 at 16:06
  • @PeterSzilas yes. I'm currently working on the infimum-case. I solved the supremum-case based on Omnomnomnom's answer. – Student Apr 14 '20 at 16:07
  • Student. Then better remove the index 0 in \mathbb{N_0}. You can post your solutions, maybe feedback. – Peter Szilas Apr 14 '20 at 16:16
  • @PeterSzilas Belgian over here, in Belgium the notation $\mathbb{N}_0$ is used to denote the naturals without zero :) I'll adjust it, thanks for reminding me. – Student Apr 14 '20 at 16:18

2 Answers2

1

Hint: Suppose for the purpose of contradiction that $\alpha < 1$ is an upper bound for $C$. Show that there exists an $m,n$ such that $1 - \frac{2m}{n+m} > \alpha$.

Once you have proven this, then you have shown that every upper bound $\alpha$ of $C$ must satisfy $\alpha \geq 1$. In other words, $1$ is the least possible upper bound, which is to say that it is the supremum.

The approach for the infimum is similar.

Ben Grossmann
  • 225,327
  • I think a was able to prove the supremum-case, however I struggling with adjusting the argument to find the infimum. The direction of the inequalties should become reversed... I guess I don't really have to prove this, since switching the terms $n$ and $m$ just switches the sign of the fraction. Any $\alpha > -1$ which is a lower bound, would also be an upper bound, hence contradiction. But how to make that formal? – Student Apr 14 '20 at 16:30
  • 1
    @Student: "We note that $s \in S \implies -s \in S$. So, if $\alpha > -1$ is a lower bound to $S$, then $-\alpha$ would be an upper bound to $S$ and we would have $-\alpha < 1$. This contradicts our earlier result regarding the supremum". – Ben Grossmann Apr 14 '20 at 16:41
  • Neat! I suppose it can be done using inequalities too, but I don't find the right estimations. – Student Apr 14 '20 at 16:57
  • Hi, I updated my answer. I think I found estimations to prove the infimum equals -1. Was this somewhat how you would have done it? Or am I overcomplicating this? – Student Apr 15 '20 at 22:41
1

Your answer is fine.

An option.

1) Supremun.

$1$ is an upper bound .

Consider $a_n:= 1-\frac{2}{n+1}$.

As suggested by Omnomnomnom assume there is a bound $b$ s.t. $b <1$.

$\lim_{n \rightarrow \infty} a_n =1$.

Given $\epsilon$ there is a $n_0$ s.t. for $n\ge n_0$

$|a_n-1| <\epsilon$,

$-\epsilon+1<a_n < \epsilon +1$.

Choose $\epsilon =1-b (>0)$, then

$1-\epsilon =b <a_n$, a contradiction.

Peter Szilas
  • 20,344
  • 2
  • 17
  • 28
  • In my coursenotes, this exercise is in the first chapter, so I'm trying to prove this without using limits (although I used limits to find the upper and lower bounds). Thanks for the suggestion and the feedback :) – Student Apr 14 '20 at 16:31
  • Fine. When you write : There is $a n_1$ s.t. $1/n_1<(1-a)/2$, you use the Archimedean principle. Greetings to Belgium. – Peter Szilas Apr 14 '20 at 16:39