0

The following is a very simple exericse in Rudin.

Let $E$ be a nonempty subset of an ordered set; suppose $\alpha$ is a lower bound of $E$ and $\beta$ is an upper bound of $E$. Prove that $\alpha \leq \beta$.

The proof goes something like this.

Let $x \in E$ (since $E$ is nonempty). Then $x \geq \alpha$ (definition of lower bound), $x \leq \beta$ (definition of upper bound). Then by transitivity, $\alpha \leq \beta$.

I am trying to understand where the assumption that $E$ is nonempty comes into play in this proof, because this fact surely does not hold when $E$ is empty. For example, I could take $\alpha = -1$, an upper bound, and $\beta = 3$, a lower bound. Could I not reword the above proof like this, for example:

By the definition of lower bound, $\forall x \in E, x \geq \alpha$, and by the definition of upper bound, $\forall x \in E, x \leq \beta$. Hence, $\forall x \in E, \alpha \leq x \leq \beta$, so by transitivity, $\alpha \leq \beta$.

I am trying to figure out why the above proof is wrong, though it very obviously is. Specifically, the first two assertions in the first sentence sound fine even if $E$ is empty. The next assertion sounds incorrect, but I suppose it could be vacuously true because I cannot provide a counterexample to it: there's nothing in $E$. Is the only incorrect step that I cannot invoke transitivity when there is no actual $x$ validating the fact that $\alpha \leq x \leq \beta$?

John P.
  • 2,136
  • Yes, you've put your finger on it. – saulspatz Jun 05 '20 at 18:16
  • I think you are right, if you have no element in $E$ then you cannot apply transitivity, hence you cannot state that $\alpha\leq\beta$ – Maryam Jun 05 '20 at 18:17
  • For the empty set, actually all reals are upper bounds as well as lower bounds. One defines $\inf\emptyset = +\infty$ and $\sup\emptyset = -\infty.$ This is the only case when $\inf E > \sup E.$ – md2perpe Jun 05 '20 at 19:11

1 Answers1

2

The last sentence of your reworded proof might be better reworded as

$\exists x\in E$ such that $\alpha\leq x \leq \beta,$ so by transitivity $\alpha\leq x\leq \beta.$

It is not sufficient that the statement holds for all $x,$ as you claim in your proof, but it is sufficient that it holds for at least one $x.$

Dasherman
  • 4,206