Suppose I have a collection $U$ of transitive binary relations on an arbitrary set $A$; elements of $U$ are subsets $S$ of $A \times A$ such that if $(a,b) \in S$ and $(b,c) \in S$ then $(a,c) \in S$ also.
Does it make sense for me to define a set $T$ to be the minimal transitive set containing $ \cup_{S \in U} S$? I'm thinking of this as taking $\cup S$ and adding in all necessary pairs to make it transitive. But I'm worried there are problems with countability. How does the cardinality of $U$ (obviously this is restricted by that of $A$) affect my ability to do this?
Thanks,
Note that transitive set has a different meaning, it means that $x\in T$ then $x\subseteq T$ (recall that everything is a set in the universe of set theory).
Generally speaking now, $S$ is a relation on $A$ if $S\subseteq A\times A$. It is easy to see that the union of relations on $A$ is also a relation on $A$. Therefore in your definition we can automatically assume that $U$ is a single relation.
What you look for is given a relation $S$ we can define the transitive closure of $S$, $S^\ast$ as the minimal transitive relation containing $S$.
This can be defined internally and externally (like many other constructions in mathematics, such as topological closure, and such). The external definition is somewhat simpler: $$S^\ast = \bigcap\{T\subseteq A\times A\mid S\subseteq T\land T\text{ is transitive}\}$$
Since transitive relations intersect at a transitive relation, this is a good definition.
The second construction is internal, we start from $S$ and just add all the pairs needed for the transitivity. This is defined as $(a,b)\in S^\ast$ if and only if there is a finite sequence $a_0,\ldots, a_n$ such that $(a_i,a_{i+1})\in S$ for all $i<n$. If we allow $n=0$ then the result is reflexive as well.
Note that both the constructions did not care for the cardinality of $A$. The cardinality of $U$ is only limited by how many transitive relations there are to begin with, which can be fairly a lot for infinite sets.