Why do we sometimes have to multiply a half-reaction by a coefficient before adding it to another one?