Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Matrix Product State Representations

D. Perez-Garcia, F. Verstraete, M. M. Wolf, J. I. Cirac

TL;DR

Matrix Product State representations provide a precise, scalable framework for describing pure multipartite quantum states with entanglement concentrated along 1D structures. The paper develops open- and periodic-boundary canonical forms, clarifies the CP-map structure behind TI MPS, and analyzes when canonical decompositions are unique, including implications for the corresponding parent Hamiltonians and energy gaps. It also connects MPS to sequential generation schemes and to classical simulation, showing that ground states of 1D local Hamiltonians exhibit area laws and can be efficiently approximated and simulated using MPS-based variational methods. Overall, the work links structural MPS properties to practical generation, Hamiltonian construction, and scalable classical simulation, underpinning the effectiveness of DMRG and tensor-network techniques in 1D quantum systems.

Abstract

This work gives a detailed investigation of matrix product state (MPS) representations for pure multipartite quantum states. We determine the freedom in representations with and without translation symmetry, derive respective canonical forms and provide efficient methods for obtaining them. Results on frustration free Hamiltonians and the generation of MPS are extended, and the use of the MPS-representation for classical simulations of quantum systems is discussed.

Matrix Product State Representations

TL;DR

Matrix Product State representations provide a precise, scalable framework for describing pure multipartite quantum states with entanglement concentrated along 1D structures. The paper develops open- and periodic-boundary canonical forms, clarifies the CP-map structure behind TI MPS, and analyzes when canonical decompositions are unique, including implications for the corresponding parent Hamiltonians and energy gaps. It also connects MPS to sequential generation schemes and to classical simulation, showing that ground states of 1D local Hamiltonians exhibit area laws and can be efficiently approximated and simulated using MPS-based variational methods. Overall, the work links structural MPS properties to practical generation, Hamiltonian construction, and scalable classical simulation, underpinning the effectiveness of DMRG and tensor-network techniques in 1D quantum systems.

Abstract

This work gives a detailed investigation of matrix product state (MPS) representations for pure multipartite quantum states. We determine the freedom in representations with and without translation symmetry, derive respective canonical forms and provide efficient methods for obtaining them. Results on frustration free Hamiltonians and the generation of MPS are extended, and the use of the MPS-representation for classical simulations of quantum systems is discussed.

Paper Structure

This paper contains 32 sections, 25 theorems, 85 equations, 1 figure.

Table of Contents

  1. Introduction and Overview
  2. Definitions and Preliminaries
  3. MPS and the valence bond picture
  4. Finitely correlated states
  5. Frustration free Hamiltonians
  6. Examples
  7. The canonical form
  8. Open boundary conditions
  9. Periodic boundary conditions and translational invariance
  10. Site independent matrices
  11. MPS and CP maps
  12. The canonical representation
  13. Generic cases
  14. Uniqueness
  15. Obtaining the canonical form
  16. ...and 17 more sections

Key Result

Theorem 1

Any state $\psi\in\mathbb{C}^{d\otimes N}$ has an OBC-MPS representation of the form Eq.(eq.vidal) with bond dimension $D\leq d^{\lfloor N/2\rfloor}$ and

Figures (1)

  • Figure 1: Computing an expectation value of an MPS is equivalent to contract the tensor of the figure, where bonds represent indices that are contracted. The matrices associated to each spin are represented by the circles (the vertical bond of each matrix is its physical index) and observables are represented by squares. It is trivial to see that this contraction can be done efficiently.

Theorems & Definitions (25)

  • Theorem 1: Completeness and canonical form
  • Theorem 2: Freedom in the choice of the matrices
  • Theorem 3: Site-independent matrices
  • Theorem 4: TI canonical form
  • Theorem 5: Periodic decomposition
  • Theorem 6: Interpretation of $\Lambda$
  • Theorem 7: Uniqueness of the canonical form
  • Proposition 1
  • Proposition 2
  • Theorem 8: Obtaining the TI canonical form
  • ...and 15 more