The density matrix renormalization group in chemistry and molecular physics: Recent developments and new challenges

Alberto Baiardi, Markus ReiherThe Journal of Chemical Physics 2020 · DOI

TL;DR

The Density Matrix Renormalization Group (DMRG) has revolutionized quantum chemistry by providing a powerful, yet computationally efficient, method for tackling complex electronic structure problems. This review highlights DMRG's evolution, its strengths in handling strongly correlated systems, and its expansion into new frontiers like time-dependent dynamics and vibrational spectroscopy.

What Problem Does This Paper Address?

Traditional quantum chemistry methods struggle with "strongly correlated" systems – molecules where electrons interact so intensely that simple approximations fail. This leads to an exponential increase in computational cost with system size, limiting our ability to accurately model complex molecules like those found in catalysis or materials science. The paper addresses this by reviewing the development and application of DMRG, a method that offers a more scalable approach to these challenging problems. It also explores how DMRG can be extended to capture dynamic electron correlation, which is crucial for accurately describing phenomena like spectroscopy.

Methods

The paper delves into the core of DMRG, explaining its foundation in Matrix Product States (MPSs). Unlike traditional methods that diagonalize large matrices, DMRG iteratively optimizes a wave function represented as an MPS. This approach leverages the "area law" of entanglement, which allows for a compact representation of the wave function for systems with short-range interactions. The review covers:

  • The MPS/MPO (Matrix Product Operator) formulation: A more flexible and powerful framework for applying DMRG to complex Hamiltonians, including those with long-range interactions.
  • Targeting excited states: Techniques to go beyond ground-state calculations, essential for simulating spectroscopic properties.
  • Multidimensional generalizations: Extending DMRG beyond one-dimensional systems using Tensor Network States (TNSs).
  • Hybrid approaches: Combining DMRG with perturbation theory, coupled-cluster methods, and density functional theory (DFT) to capture both static and dynamic correlation effects.
  • Time-dependent DMRG (TD-DMRG): Applying DMRG to simulate quantum dynamics, crucial for understanding processes like electron transfer and light-matter interactions.
  • Vibrational DMRG (vDMRG): Adapting DMRG to solve the vibrational Schrödinger equation, enabling the study of molecular vibrations in larger systems.

Datasets

Not discussed in the paper. The paper focuses on theoretical methods and their application to general quantum chemical problems, rather than specific datasets.

Key Insights and Ideas

  • DMRG's "area law" advantage: For systems with short-range interactions, the entanglement entropy is limited, allowing MPSs to represent the wave function compactly, thus taming the exponential scaling problem.
  • MPS/MPO formulation is key: This generalized framework is crucial for handling the complex, long-range interactions present in realistic quantum chemical Hamiltonians.
  • DMRG as a "gold standard": When combined with perturbation theory, DMRG has become a reliable method for obtaining accurate reference data for complex molecules, aiding in the development of new theoretical models.
  • TD-DMRG opens new doors: Simulating quantum dynamics without the need for diagonalization offers a powerful alternative for studying phenomena like X-ray spectroscopy and electron transfer.
  • Vibrational DMRG tackles molecular dynamics: Extending DMRG to vibrational problems allows for more accurate modeling of molecular vibrations, especially in larger systems where traditional methods falter.

Key Results

The paper highlights DMRG's success in:

  • Enabling large-scale CI calculations: Making calculations feasible for systems with up to 100 orbitals, which were previously intractable.
  • Accurate energy calculations: Providing reliable ground-state energies for complex molecules, including transition metal complexes and clusters.
  • Improved efficiency: Demonstrating significant speedups and reduced computational cost compared to traditional methods for strongly correlated systems.
  • Successful application to new fields: Showing promising results in time-dependent dynamics and vibrational spectroscopy, expanding the scope of DMRG applications.

Limitations and Future Work

The authors acknowledge several limitations and suggest future directions:

  • Long-range interactions: While MPS/MPO helps, the efficiency of DMRG can still be impacted by strong long-range interactions, requiring higher bond dimensions.
  • Dynamical correlation: Recovering dynamical correlation efficiently remains a challenge, often requiring hybrid approaches with perturbation theory. The evaluation of high-order density matrices can be a bottleneck.
  • TD-DMRG convergence: Ensuring convergence for long-time propagations in TD-DMRG can be difficult, as the MPS bond dimension can grow.
  • Multidimensional TNS optimization: Generalizing ALS (alternating least squares) optimization to arbitrary TNSs is still an open problem, often necessitating less efficient Monte Carlo methods.
  • Automated active space selection: While progress has been made (e.g., AutoCAS), developing truly general and automated methods for selecting the active space remains an active area of research.
  • Vibro-rotational coupling: Incorporating vibro-rotational effects in vDMRG is an area that needs further exploration.
  • Spectroscopy applications: While promising, the application of DMRG to molecular spectroscopy is still in its early stages and requires further development, particularly for higher-order properties.
  • Embedding schemes for DMRG: Extending advanced embedding schemes to DMRG is crucial for tackling even larger systems.

Summary updated

Abstract

In the past two decades, the density matrix renormalization group (DMRG) has emerged as an innovative new method in quantum chemistry relying on a theoretical framework very different from that of traditional electronic structure approaches. The development of the quantum chemical DMRG has been remarkably fast: it has already become one of the reference approaches for large-scale multiconfigurational calculations. This perspective discusses the major features of DMRG, highlighting its strengths and weaknesses also in comparison with other novel approaches. The method is presented following its historical development, starting from its original formulation up to its most recent applications. Possible routes to recover dynamical correlation are discussed in detail. Emerging new fields of applications of DMRG are explored, such as its time-dependent formulation and the application to vibrational spectroscopy.
BibTeX
@article{baiardi2020density,
  author = {Alberto Baiardi and Markus Reiher},
  title = {{The density matrix renormalization group in chemistry and molecular physics: Recent developments and new challenges}},
  journal = {The Journal of Chemical Physics},
  year = {2020},
  doi = {10.1063/1.5129672},
}
Figure