$$
This is essentially Theorem~5 of \shortciteN{Kelly1996closure}.
The only difference is that they only consider lexicographically
-forward relations, a special case of acyclic relation.
+forward relations, a special case of acyclic relations.
If, on the other hand, $R$ is cyclic, then we have to resort
to checking whether the approximation $K$ of the power is exact.
$$
However, when we intersect domain and range of this relation
with those of the input relation, then the result only contains
-the idenity mapping on the intersection of domain and range.
+the identity mapping on the intersection of domain and range.
\shortciteN{Kelly1996closure} propose to intersect domain
and range with then {\em union} of domain and range of the input
relation instead and call the result $R_i^?$.