Equational logic of partial functions under Kleene equality: a complete and an incomplete set of rules

When equational logic for partial functions is interpreted using Kleene equality as the predicate, the relation of logical consequence may be said to express what identities of partial functions follow from a given set of identities. In the analogous situation for total functions, there is a complete set of inference rules consisting of reflexivity, symmetry, transitivity, replacement, and substitution; in the case of partial functions, unrestricted substitution fails to be a valid inference rule, and there remains the question of how to obtain a complete set of rules. The first part of the present paper shows that completeness cannot be obtained by a mere restriction of the substitution rule, for a counterexample shows that even a rule allowing substitution in all consequentially valid instances fails, in conjunction with the other four rules, to yield a complete set of rules. The second part of the paper defines a combined rule of transitivity-substitution which, in conjunction with reflexivity, symmetry, replacement, and substitution only of variables, yields a complete set of rules. The new rule is first stated in a form that allows an unbounded number of premises, and then is altered to a three-premise form. In both forms, the rule suffers from the shortcoming that in its formulation an auxiliary notion of conditional existence is involved, which is given by a recursive syntactic definition. As a result, the set of instantiations of the rule is recursively enumerable, but not (apparently) recursive (assuming a recursive set of premises).