Apologies for the long prose. The following problems are from a set of Topology Lecture Notes by Pete L. Clark that I found fascinating and am working through. Am a little unsure about my solutions here. Would be extremely grateful if someone could read through it (again sorry about the length) and give me some feedback.

$X$ and $Y$ are nonempty totally ordered sets. If $X \times Y$, equipped with the lexicographic order, is complete then so is $X$ and $Y$.

Proof: The completeness of $X$ is easier to establish and will be attempted first. Let $S$ be a subset of $X$. Since $Y$ is nonempty it contains some point $y$ and $S \times \{y\}$ is a subset of $X \times Y$. Now let $(\pi, \theta) = \sup (S \times \{y\})$.

  • Let $s \in S$. Then $(s, y) \in S \times \{y\}$. Then, $(s, y) \le (\pi, \theta)$. Then $s \le \pi$.

  • Now let $\eta$ be any other upper bound of $S$. Then $(\eta, y)$ is an upper bound of $S \times \{y\}$ which can be easily seen. Hence, $(\pi, \theta) \le (\eta, y)$, so, $\pi \le \eta$.

This shows that $\pi = \sup S$. So $X$ is complete.

The completeness of $Y$ is actually a bigger task. Now let $T$ be an arbitrary subset of $Y$. We have already proved the completeness of $X$. Hence $X$ admits a top element say $m$. Now consider the subset $\{m\} \times T$ of $X \times Y$. This set has a supremum, $(\gamma, \zeta)$. Now it can be easily argued that $\gamma = m$.

  • Let $t \in T$. Then $(m, t) \in \{m\} \times T \implies (m, t) \le (m, \zeta) \implies t \le \zeta$

  • Let $\psi$ be any other upper bound for the set $T$. Then $(m, \psi)$ is an upper bound for the set $\{m\} \times T$. Then, $(m, \zeta) \le (m, \psi) \implies \zeta \le \psi$

So, $\zeta = \sup T$, and therefore $T$ is complete.

Suppose $X \times Y$ is Dedekind complete. What can be said about $X$ and $Y$?

The proof above can be imitated for a nonempty subset $S$ of $X$ to show that $X$ is also Dedekind complete. But the second part of that proof cannot be imitated since there is no reason why $X$ should have a top element.

Counterexample: Let $X = \Bbb N$ and $Y = [0, 1)$ equipped with the usual orderings. Then we claim that $X \times Y$ is Dedekind complete. Let $S$ be a subset of $X \times Y$ that is bounded above. If $S = \emptyset$ then $(1,0)$, the bottom element of $X \times Y$ serves as the supremum of $S$. So suppose instead that $S$ is nonempty. Then let $S_X, S_Y$ be the projections of $S$ on $X$ and $Y$ respectively. Since $S$ is bounded so is $S_X$ and let $M$ be the top element of $S_X$. If $S_Y$ has a supremum $\lambda$ in $Y$ then $(M, \lambda ) = \sup S$ if not then $(M + 1,0) = \sup S$.

  • $\begingroup$ I have since realised that this counter-example won't work since $[0,1)$ actually is Dedekind complete under the usual order and am working on it. $\endgroup$ – Ishfaaq Dec 24 '15 at 11:42
  • $\begingroup$ You might want to formulate a few things clearer. A totally oredered set is complete if each non-empty set with an upper bound has a supremum. Hence you should not start with just any subset $S,T\subseteq X$ but rather demand that it be non-empty and have some upper bound. Indeed, if $S=\emptyset$ then $(\pi,\theta)$ may not exist. Also, $X$ need not have a top element, for example $\Bbb R$ is complete but has no top element. - Unless you work with a different definition of "complete" $\endgroup$ – justanotherhagman Dec 24 '15 at 12:36
  • $\begingroup$ Nah the definition in the notes is different. A totally ordered set is complete if any set has a supremum. And the definition you mention is called Dedekind-Completeness. The notes are here. math.uga.edu/~pete/pointset.pdf $\endgroup$ – Ishfaaq Dec 24 '15 at 12:40
  • 1
    $\begingroup$ Alright. One nitpick then: $\gamma =m$ follows only if $T\ne\emptyset$. To find $\sup T$ for $T=\emptyset$ consider the second component of the supremum of $\emptyset\subseteq X\times Y$ $\endgroup$ – justanotherhagman Dec 24 '15 at 14:09
  • $\begingroup$ @justanotherhagman: Yes, that's right. Thanks for that. $\endgroup$ – Ishfaaq Dec 24 '15 at 14:12

The completeness arguments are basically just fine, and justanotherhagman has already dealt with the one small nitpick. For Dedekind completeness of $Y$ you can argue as follows.

Suppose that $\varnothing\ne T\subseteq Y$, and $T$ is bounded above in $Y$, say by $y$. Let $x\in X$ be arbitrary, and let $T'=\{x\}\times T$. Then $T'$ is bounded in $X\times Y$ by $\langle x,y\rangle$, so it has a least upper bound $\langle x',y'\rangle$ in $X\times Y$. Clearly $\langle x',y'\rangle\le\langle x,y\rangle$, so $x'\le x$. On the other hand, $T\ne\varnothing$, so there is some $\langle x,t_0\rangle\in T'$, and since $\langle x,t_0\rangle\le\langle x',y'\rangle$, we must have $x\le x'$ and hence $x'=x$. It follows that $\langle x,t\rangle\le\langle x,y'\rangle$ for each $t\in T$ and hence that $t\le y'$ in $Y$ for each $t\in T$, i.e., that $y'$ is an upper bound for $T$ in $Y$. I’ll leave it to you to finish off the argument by showing that $y'$ actually the least upper bound of $T$.

Alternatively, you could suppose that $T$ is a non-empty subset of $Y$ that has an upper bound but no least upper bound, define $T'$ as above, and show that in $X\times Y$ the set $T'$ is bounded above but has no least upper bound; the argument is essentially the same.

| cite | improve this answer | |

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.