Equisatisfiability

In mathematical logic (a subtopic within the field of formal logic), two formulae are equisatisfiable if the first formula is satisfiable whenever the second is and vice versa; in other words, either both formulae are satisfiable or both are not. Equisatisfiable formulae may disagree, however, for a particular choice of variables. As a result, equisatisfiability is different from logical equivalence, as two equivalent formulae always have the same models. Whereas within equisatisfiable formulae, only the primitive proposition the formula imposes is valued.

Equisatisfiability is generally used in the context of translating formulae, so that one can define a translation to be correct if the original and resulting formulae are equisatisfiable. Examples of translations that preserve equisatisfiability are Skolemization and some translations into conjunctive normal form such as the Tseytin transformation.

Examples
A translation from propositional logic into propositional logic in which every binary disjunction $$a \vee b$$ is replaced by $$((a \vee n) \wedge (\neg n \vee b))$$, where $$n$$ is a new variable (one for each replaced disjunction) is a transformation in which satisfiability is preserved: the original and resulting formulae are equisatisfiable. Note that these two formulae are not equivalent: the first formula has the model in which $$b$$ is true while $$a$$ and $$n$$ are false (the model's truth value for $$n$$ being irrelevant to the truth value of the formula), but this is not a model of the second formula, in which $$n$$ has to be true in this case.