2
$\begingroup$

May be a simple question, but I still don't know what the definition of the supremum or infimum of a set in the extended real line is. For example when we define the Lebesgue measure, we define the Lebesgue outer measure as the infimum of a set in the extended real line. My questions are:

  1. What is the definition of the supremum/infimum of a set in the extended real line?

  2. FORMALLY, when do we say that the infimum of a set in the extended real line is equal to positive infinity?

$\endgroup$
  • 1
    $\begingroup$ Infimum of empty set is positive infinity and supremum is negative infinity. $\endgroup$ – jdods Jul 15 '16 at 18:06
  • $\begingroup$ @jdods how is this relevant? $\endgroup$ – Topological cat Jul 15 '16 at 18:09
  • $\begingroup$ Is it not an example of #2? $\endgroup$ – jdods Jul 15 '16 at 18:11
  • $\begingroup$ @jdods in 2. I ask: "when do we say that the infimum of a set in the extended real is positive infinity?". You gave me an example of a set having infimum positive infinity, this isn't exactly helpful $\endgroup$ – Topological cat Jul 15 '16 at 18:14
  • $\begingroup$ Wouldn't the infimum be $+\infty$ when the set is either empty or $\{\infty\}$? $\endgroup$ – jdods Jul 15 '16 at 18:19
2
$\begingroup$

What jdods wrote in the comments essentially covdrs it, but here I make things very explicit.

The supremum of a set is the smallest upper bound and the infimum is the largest lower bound. This is the formal definition that holds for every ordering.

Let's decode this. There is a usual ordering $\leq$ on $\mathbb{R}$, that one extends to the extended real line by letting $-\infty\leq r\leq$ for each real number $r$. If $A$ is a set of extended real numbers (in particular, a set of real numbers), then $u$ is an upper bound of $A$ if $r\leq u$ for all $r\in A$ and $l$ is a lower bound of $A$ if $l\leq r$ for all $r\in A$. Note that $-\infty$ is a lower bound of every set and $\infty$ is an upper bound of every set. Moreover, every extended real number is both an upper and lower bound of the empty set, since the condition "for all $r\in\emptyset$" holds vacuosly.

A smallest element in a set of extended real numbers is a lower bound of the set that is also an element of the set. If a set has a smallest element, it is unique. The set $(0,1]$ has no smallest element. Similarly, a largest element in a set is an upper bound of the set that is actually in the set.

Now let $A$ be a set of extended real numbers with infimum $\infty$. Then $\infty$ is a lower bound of $A$ and is the largest such lower bound. If $r\in A$, then we must have $\infty\leq r$, since $\infty$ is a lower bound of $A,$ and this is only possible if $r=\infty$. So the only element $A$ can have is $\infty$. There are then two possibilities, either $A$ is nonempty, then we must have $A=\{\infty\}$, in which case $\infty$ is indeed a lower bound of $A$, and, as the only lower bound, is also the largest lower bound. The other possibility is that $A=\emptyset$. Now every extended real number is a lower bound of the empty set, so the largest lower bound is the largest extended real number, namely $\infty$. So these are all possibilities for $\infty$ to be the infimum of a set of extended real numbers.

$\endgroup$
  • $\begingroup$ +1. I don't think my answer adds too much to this, but I'll leave it there in case it helps someone. $\endgroup$ – 6005 Jul 15 '16 at 19:41
2
$\begingroup$

You are probably overthinking it a little. Infimum and supremum can be defined for any totally ordered set; the definition of $\inf$ is the "least upper bound", and the definition of $\sup$ is the "greatest lower bound" (explicit definitions given below). Both infimum and supremum are unique if they exist; on the other hand, not all totally ordered sets have the existence of supremum and infimum for every subset. However, crucially, the extended real line $\overline{\mathbb{R}} = \mathbb{R} \cup \{-\infty,\infty\}$ does have a supremum and infimum for every subset.

Formal details

We start out by ignoring addition, multiplication, and anything else about $\overline{\mathbb{R}}$ in particular, and focusing only on the structure $\overline{\mathbb{R}}$ under the order $\le$. Here are the definitions, which may be familiar:

Definition 1. Given a totally ordered set $T$ (under the order $\le$), and a subset $S \subseteq T$, an upper bound of $S$ is an element $a \in T$, such that $s \le a$ for all $s \in S$ (lower bound likewise);

Definition 2. Given a totally ordered set $T$ (under the order $\le$), and a subset $S \subseteq T$, a supremum of $S$ is a least upper bound of $S$, i.e. an upper bound that is less than or equal to all other upper bounds (infimum likewise).

As I already mentioned, the sup and inf are unique when they exist. In the totally ordered set $\mathbb{R}$, every bounded subset has a least upper bound. In $\overline{\mathbb{R}}$, every subset has a least upper bound (and you should prove this). This justifies using the notation $\sup S$ for any subset $S$ of the extended real line.

I believe that answers (1). To answer (2) and some related questions you may have, then, here are some specifics for $\overline{\mathbb{R}}$:

  • For $S \subseteq \overline{\mathbb{R}}$, $\inf S = \infty$ exactly when $S \subseteq \{\infty\}$. That is, $\inf \varnothing = \infty$, $\inf \{\infty\} = \infty$, and there are no other examples.
  • Likewise $\sup S = -\infty$ occurs exactly when $S = \varnothing$ or $S = \{-\infty\}$.

  • For $S \subseteq \overline{\mathbb{R}}$, $\sup S = \infty$ exactly when $S$ is not bounded above by a finite real number; in particular this occurs if $\infty \in S$.

  • Similarly $\inf S = -\infty$ exactly when $S$ is not bounded below by a finite real number.

  • If $S$ is bounded both above and below by finite real numbers, then $\inf S$ and $\sup S$ agree with the supremum and infimum in $\mathbb{R}$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.