In this paper, the method how to further study the representational and the representational transformation is introduced by geometry coordinates, linear algebra and physical concept. According to this representational theory which can be simply compared with coordinate theory in geometry, two simple memory methods have been provided. Firstly, the transformation formula can be written as b =S+a from representation A to representation B, in which the matrix element of the transformation matrix is <ψ|φ>, which can be regarded as the inner product between the basis vector of representation A and representation B. At the same time, this formula can be thought of as the expression of the representation B of the state vector equals the transformation matrix acting on the expression of the representation A. Secondly, the transformation formula can be written as F' =S+FS for the mechanical operator from representation A to representation B, in which the operator F and operator F' are the expression of representations A and representations B, respectively, which this formula can be thought of as the expression of the representation B of the mechanical operator equals the conjugate matrix of transformation matrix and the expression of the representation A of the mechanical operator acting on transformation matrix. Finally, an example is given to prove the superiority of representation transformation. Meanwhile, the intrinsic function under two representations is deduced and the transformation between them is realized by the Fourier transform, thus deepening the understanding of representation transformation.
A method is given for obtaining the density matrix or matrices, expressible in terms of a given set of spin geminals, which are as nearly
N representable as possible. A measure of
N representability is introduced as the norm of the antisymmetric component of a normalized wave function leading to the density matrix. An error estimate is given for the maximum deviation of an energy calculated with the density matrix from that calculated with an antisymmetric wave function. Linear variational parameters may occur in the density matrix, and their number can be increased at the expense of
N representability. Implementation of the method requires diagonalization of a large matrix, and the consequences of truncation are considered. The method is also related to the problem of the exact
N representability of a given density matrix.