Progress in the understanding of strongly correlated materials depends crucially on powerful many-body methods. The major methods currently available are various mean field theories, Monte Carlo and numerical renormalization group. Over the recent years, we have been developing a novel many-body method, entanglement perturbation theory (EPT), for calculating partition functions, ground states and elementary excitations. The idea of EPT is divide and conquer: no Hilbert space truncation, no finite-size problem, no fermion sign problem, just SVD from operators to eigenstates, leading to tensor product representations, and reducing the problem to a key concept, iterative solution of the generalized eigenvalue problems. Physically, it pursues a conviction that the systematic inclusion of correlations ~ entanglements, will be the only perturbation theory which converges. Also important conviction is, the fermion sign problem arising from the infinite-range anticommutation algebra, is inherently incompatible with any methods which include as a core step an update of the wave function based on the application of local operators, not the entire Hamiltonian, a well-known example being the real-space renormalization group. In this talk, I will give an overview of this new theoretical endeavor as well as a recent result for the long-range correlation functions, long enough to be comparable with field theories.
