Let $A$ be an infinite set and $R$ be a transitive relation on $A$ such that sets $\{a:(a,y)\in R\}$ and $\{a:(x,a)\in R\}$ are finite for any $a\in A$. What does $R$ look like?
-
2$\begingroup$ What do you mean? What kind of answer are you looking for? $\endgroup$– Gerry MyersonCommented Jul 12, 2022 at 9:29
-
$\begingroup$ I am interested in the description of such relations $\endgroup$– Markiian KhylynskyiCommented Jul 13, 2022 at 5:57
-
$\begingroup$ And what does that mean? What counts as the description of a relation? $\endgroup$– Gerry MyersonCommented Jul 13, 2022 at 6:19
-
$\begingroup$ Perhaps one can say more specifically about these relations than that these intersections are finite $\endgroup$– Markiian KhylynskyiCommented Jul 13, 2022 at 6:38
1 Answer
Interesting question, I'm assuming you arrived at it by thinking about some possible relations, and realising that transitivity seems to force those sets to be infinite, for example the "less than" operator on $\mathbb N$
Since $1<2$ and $2<3$, and so on, for transitivity to hold every number must be less than all of its successors, and so those sets you described are all infinite
But there are ways around this (2 that I can think of)
Let $$ R := \{ (a,b) \in \mathbb N × \mathbb N | a = b\}$$
Indeed, $|[R]_a| = 1$ for all $a \in \mathbb N$
And it is trivially transitive
A more exciting example could be
$R^k := \{(a,b) \in \mathbb N × \mathbb N| a//k = b//k\}$
For some $k\in \mathbb N^+$ Where "//" denotes floor division
This works because it "groups" together subsets of size $k$ in $\mathbb N$
So if k = 5, then the numbers 0-4 all appear together in all there permutations, and with nothing else, 5-9 same thing, etc.
A subset of this relationship could contain only cycles within these groupings
For uncountably infinite sets its a bit harder, since grouping intervals in the reals analgous to our second example won't work because each point will appear in a k-ball with infinitely many others, I'f you don't require that every point appears at least once in the relation, then the above examples still work since $\mathbb N \subset \mathbb R$, if you do require that, then I'm not sure, I don't think $\mathbb R$ can be covered by infinitely many finite discrete cycles for example
-
$\begingroup$ Thanks for the examples $\endgroup$ Commented Jul 13, 2022 at 6:00
-
$\begingroup$ I don't require that every point appears at least once in the relation. But in effect it doesn't give any new examples. $\endgroup$ Commented Jul 13, 2022 at 6:12
-
$\begingroup$ These examples can be generalized as follows. Let $\{A_i\}_{i\in I}$ be a family of finite disjoint sets and $R_i$ be a transitive relation on a set $A_i$. Consider the relation $R=\bigcup\limits_{i\in I}R_i$ on the set $\bigsqcup\limits_{i\in I}A_i$. This relation is transitive and has finite intersections. $\endgroup$ Commented Jul 14, 2022 at 12:15
-
$\begingroup$ Ah very nice! What does the squared off union sign mean? Or is it a typo $\endgroup$– CarlyleCommented Jul 15, 2022 at 22:02
-
$\begingroup$ It is used when the sets are disjoint $\endgroup$ Commented Jul 16, 2022 at 4:30