We propose a new principle of modularity for product design in cases where functions and constraints contribute to modularity. A modularity matrix is defined as a set of permissible ranges for both functions and constraints. We prove that a design structure matrix (DSM) of physical components for functions is equivalent to a modularity matrix of axiomatic design (AD). Next, from statistical mechanics, we introduce entropy for selected modules with constrantis after regularizing a modularity matrix by a new parameter. This concept of entropy highlights the number of possible options to realize the optimal value of a product. We show that the two limits of the entropy are logically equivalent to modularization of real option theory and information axiom in DSM and AD. From the principle of entropy maximization, we clarify the manner in which the matrix elements of modularity matrix should change in order to realize the highest value of products.