For an ordered set, the least upper bound property means that every nonempty subset that is bounded above has a least upper bound, that is, a supremum, in the set. Thus, the completeness axiom says that \(\R\text{,}\) with its usual order, has the least upper bound property. Dually, this is equivalent to saying that every nonempty subset of \(\R\) that is bounded below has an infimum in \(\R\text{;}\) see Exerciseย 1.3.9.
Suppose instead that some \(x \in \R\) were an upper bound for \(\N\text{.}\) Then \(\N\) would be non-empty and bounded above, so by the least upper bound property, \(u:=\sup\N\) would exist. Since \(u-1<u\text{,}\) the number \(u-1\) cannot be an upper bound of \(\N\text{.}\) So there is \(n\in\N\) with \(u-1<n\text{.}\) Then \(u<n+1\text{,}\) and \(n+1\in\N\text{,}\) contradicting that \(u\) is an upper bound of \(\N\text{.}\)
Since \(\varepsilon > 0\text{,}\) the number \(x/\varepsilon\) is real. By Propositionย 1.2.2, there exists \(N \in \N\) such that \(N > x/\varepsilon\text{.}\) Multiplying by \(\varepsilon\) gives \(N\varepsilon > x\text{.}\)
Apply Propositionย 1.2.3 with \(x=1\text{.}\) Then there exists \(n \in \N\) such that \(n\varepsilon > 1\text{.}\) Since \(n>0\text{,}\) dividing by \(n\varepsilon\) yields \(1/n < \varepsilon\text{.}\)
Then \(n \le x\text{.}\) Also \(n \ne N\text{,}\) because \(x < N\text{.}\) Hence \(n+1 \in S\text{.}\) By maximality of \(n\text{,}\) the integer \(n+1\) cannot satisfy \(n+1 \le x\text{.}\) Therefore \(x < n+1\text{,}\) and so \(n \le x < n+1\text{.}\)
Since \(N>0\text{,}\) dividing by \(N\) gives \(x < m/N < y\text{.}\) Because \(m/N \in \Q\text{,}\) this proves that \(\Q\) is dense in \(\R\text{.}\)
For existence, define \(S=\{x\in\R : x>0,\ x^2\le a\}\text{.}\) The set is non-empty: if \(a\ge 1\text{,}\) then \(1\in S\text{;}\) if \(0<a<1\text{,}\) then \(a\in S\text{.}\) It is bounded above by \(M=\max\{a,1\}\text{,}\) since \(x>M\) implies \(x^2>a\text{.}\) By completeness, \(u:=\sup S\) exists.
If \(u^2<a\text{,}\) then for sufficiently small \(\varepsilon>0\text{,}\) we still have \((u+\varepsilon)^2\le a\text{,}\) so \(u+\varepsilon\in S\text{,}\) contradicting that \(u\) is an upper bound of \(S\text{.}\) If \(u^2>a\text{,}\) then for sufficiently small \(\varepsilon>0\text{,}\)\((u-\varepsilon)^2>a\text{;}\) then every \(x\in S\) satisfies \(x\le u-\varepsilon\text{,}\) so \(u-\varepsilon\) is an upper bound of \(S\text{,}\) contradicting leastness of \(u\text{.}\) Hence \(u^2=a\text{.}\)
A standard comparison example is the ordered field \(\Q\text{:}\) the set \(\{q\in\Q:q>0,\ q^2<2\}\) is non-empty and bounded above in \(\Q\text{,}\) but it has no supremum in \(\Q\text{.}\) Therefore \(\Q\) does not satisfy the least upper bound property.