On the theory of superconductivity