Suppose $R$ is a total order on $A$ and $S$ is a total order on $B$. Define a relation $L$ on $A \times B$ as follows: $L = \{((a, b), (a', b')) \in (A \times B) \times (A \times B) | aRa' \land (a = a' \implies bSb') \}$. Is $L$ a total order? (Assume that we have also proven that L is a partial order).
I would appreciate a suggestion when looking for a counterexample (I suspect that the $L$ is not a total order).
Why I think it is not a total order:
To show that $L$ is a total order let $(a,b), (c,d) \in A\times B$ and we will show that $(a,b)L(c,d) \lor (c,d)L(a,b)$. Since $R$ and $S$ are total orders we have:
- $aRc \lor cRa$
- $bSd \lor dSb$
We can split the proof into cases:
- $aRc \land bSd$. Then it doesn't matter whether $a =c$, we have $(a,b)L(c,d)$
- $aRc \land dSb$.
- If $a\ne c$ we can have $(a,b)L(c,d)$.
- But if $a=c$ to have this result we need to have $bSd$. If $b=d$ then $bSd$ so $(a,b)L(c,d)$. However if $b \ne d$, suppose $bSd$. Since $S$ is antisymmetric it would follow that $b=d$ which is a contradiction, so $(b,d)\notin S$.
It seems to me that I am stuck at point 2.2. $(a,b)L(c,d)$ cannot be true so the only way the proof would work out is if $(c,d)L(a,b)$ which I am unlikely to prove in this case.