0

Let $B$ be a set of positive real numbers with the property that adding together any finite subset of elements from $B$ always gives a sum of $2$ or less. Show $B$ must be finite or countable.

Can this proof be criticized?

Let $m_1,m_2,m_3...$ be $B$'s (respectively) greatest element, second greatest element, third greatest element etc(the list could also be finite, or even empty if $B$ does not have a greatest element). I claim that the set $B\setminus\left\{m_1,m_2,...\right\}$ is empty, i.e. $B$ is either countable or finite:

Let $s$ be the supremum of $C=B\setminus\left\{m_1,m_2,...\right\}\neq\left\{\ \right\}$ because although it has no greatest element, it is bounded (all elements of $B\supseteq C$ are less than $2$ in particular).

For an arbitrary $\epsilon_1>0$ there is a $c_1$ in $C$ such that $s-\epsilon_1<c_1$; and there is a $\epsilon_2$ such that $0<\epsilon_2<s-c_1$(because $s$ is strictly greater than every element of $C$), and for such $\epsilon_2$ there is a $c_2>s-\epsilon_2>c_1$ in $C$ etc.:

$s-\epsilon_1<c_1<s-\epsilon_2<c_2<s-\epsilon_3<c_3<...<s\ \ $ i.e. there are $c$'s in $C$ such that $c_1<c_2<...$

Let $n$ be an integer greater than $\frac{2}{c_1}$: the finite sum $c_1+...+c_n>nc_1$ is greater than $2$. Contradiction in $C=B\setminus\left\{m_1,m_2,...\right\}\neq\left\{\ \right\}$.

ludo1337
  • 169
  • 1
    Are you sure $m_1$ exists? There can be real sets having no greatest element. – 311411 Feb 24 '22 at 16:39
  • The set {m1,m2,...} can be empty or not, it doesn't matter for the proof – ludo1337 Feb 24 '22 at 16:50
  • Could you elaborate? I don't think I assumed B can be written as a list, I just listed its greatest elements (for example the set [0,1]U{2,3,4} has greatest element 4, second greatest element 3, third greatest element 2, fourth greatest element 1, and it has no fifth greater element; yet [0,1]U{2,3,4} can not be written as a list(U stands for union)) – ludo1337 Feb 24 '22 at 18:06

2 Answers2

2

I can elaborate and criticize. I see now that your idea is to eliminate all the large elements of the set provided they are isolated. So, yes, I agree that you are not assuming all the elements of $B$ are isolated. Even so the proof is flawed, but not completely misguided. (The error is, perhaps, the example that you give in your comment, $B= [0,1]\cup \{2,3,4\}$, mislead your intuition about this.)

You start: "Let $m_1,m_2,m_3,\dots$ be $B$'s (respectively) greatest element, second greatest element, third greatest element etc(the list could also be finite, or even empty if B does not have a greatest element). I claim that the set $B\setminus \{m_1,m_2,.\dots\}$ is empty."

Suppose that the set $B$ consists of all these numbers:

$$1, \dots, 1+\frac{1}{n}, 1+\frac{1}{n-1}, \dots, 1+\frac{1}{3}, 1+\frac{1}{2}$$ together with all of these

$$\frac12, \dots, \frac12+\frac{1}{n}, \frac12+\frac{1}{n-1}, \dots, \frac12+\frac{1}{3} $$

etc. if you want more.

Your argument does not work because the point $s$ that you claim exists is not an accumulation point of the new set $C$, so you cannot find that many values of $c$ in your set $C$ close to $s$.

Specifically: your list from largest down would be $$1+\frac{1}{2}, 1+\frac{1}{3}, 1+\frac{1}{4}, \dots $$ and would not end. So you would be left with $C$ (which is $B$ minus your list) and that contains the number $1$ so $s=1$. There does not exist a number $c_1$ with $s -\epsilon_1 <c_1< \dots$, etc.

There are simpler ideas to prove this, but let's go with your idea. You just want a positive number that is an accumulation point of $B$. But any set $B$ has a countable number of isolated points only. If there is no positive accumulation point then the set is countable. I think that is your proof. Is it?

For a nontopological, simply arithmetic proof, write the sequence of sets $$B_n=\left\{b\in B: b>\frac1n\right\}$$ and show each is finite.

  • Thanks for answering, I appreciate! Can you explain why the $C$ set of this $B$ you exemplified is not empty? The definiton of $C$, removing the greatest element of the set successively says that $C=B\setminus \left{1+\frac{1}{2},1+\frac{1}{3}\ ...\ ,1\right}$ which is just empty: $B \setminus B=\left{\ \right}$, and therefore there is no supremum of $C$ (if $C$ wasn't empty, it being also bounded would imply existence of supremum). It doesn't work because removing the greatest element successively would never end because 1, 1+1/2 , 1+ 1/3, ... is infinite? – ludo1337 Feb 24 '22 at 21:12
  • (Also the main idea of the proof was to show that given an uncountable set of reals $B$, there is a $b$ element in $B$ such that there are infinitely many other elements of $B$ that are greater than $b$. I am not sure about your suggestion with accumulation points, as I am not familiar with this definition just yet) – ludo1337 Feb 24 '22 at 21:12
  • 1
    @ludo1337 I added a bit. Your argument was to produce a positive number $s$ so that there are infinitely members of $B$ greater than (say) $s/2$. That is a good approach, but the details you gave don't quite work in general. In fact in this example there are indeed infinitely many numbers larger than $1$, but they are not between $1$ and $1-\epsilon$ as you claim. You are ready to learn about isolated points and accumulation points and then your argument will make more sense. – B. S. Thomson Feb 24 '22 at 21:42
0

Suppose $B$ has $n$ elements greater than $2/n$

$$x_1>\frac{2}{n}$$ $$x_2>\frac{2}{n}$$ $$...$$

$$x_n>\frac{2}{n}$$

$$x_1+x_2+...+x_n>n\frac{2}{n}=2$$

Contradiction.

So $\forall \ n>0$, there are fewer than $n$ elements greater than $2/n$.

$\forall n>0,$ define $A_n=\{x\in B|x>2/n\}$.

We know every $A_n$ is finite. The infinum of $2/n$ is $0$, so we have every positive number in B in some $A_n$. So B is the countable union of finite subsets. From here we know the countable union of countable subsets is countable, and so is $N \times N$. For any element of an $A_n$ we can associate an natural number $k$ (e.g. by counting in ascending order) and therefor an ordered pair $(n,k)$. In this way we map $B$ into a subset of $N\times N$. A subset of a countable subset must be countable or finite else we have a contradiction, so $B$ must be countable or finite.

TurlocTheRed
  • 5,683