Large-Scale Quantum Dynamics with Matrix Product States

Alberto Baiardi, Markus ReiherJournal of Chemical Theory and Computation 2019 · DOI · arXiv

Abstract

Dynamical electronic- and vibrational-structure theories have received a growing interest in the past few years due to their ability to simulate spectra recorded with ultrafast experimental techniques. The exact time evolution of a molecular system can, in principle, be obtained from the time-dependent version of full configuration interaction. Such an approach is, however, limited to few-atom systems due to the exponential increase of its cost with the system dimension. In the present work, we overcome this unfavorable scaling by employing the time-dependent density matrix renormalization group (TD-DMRG) which parametrizes the time-dependent wave function as a matrix product state. The time-dependent Schrödinger equation is then integrated with a sweep-based algorithm, as in standard time-independent DMRG. Unlike other TD-DMRG approaches, the one presented here leads to a set of coupled equations that can be integrated exactly. The resulting theory enables us to study real- and imaginary-time evolutions of Hamiltonians comprising more than 20 degrees of freedom that are challenging for current state-of-the-art quantum dynamics algorithms. We apply our algorithm to the simulation of quantum dynamics of models of increasing complexity, ranging from simple excitonic Hamiltonians to more complex ab initio vibronic ones.

Introduction

In the field of quantum chemistry, the "Holy Grail" is the accurate solution of the time-dependent Schrödinger equation. Such solutions allow researchers to observe the intricate dynamics of electrons and nuclei during chemical reactions or under the influence of ultrafast laser pulses. However, traditional methods face the "curse of dimensionality": the complexity of a quantum wavefunction grows exponentially with the size of the system. Doubling the number of atoms does not merely double the computational cost; it causes it to explode, typically limiting exact simulations to very small, few-atom systems.

To address this, researchers have developed a powerful new approach that borrows the "tensor train" technique from condensed matter physics. By projecting quantum equations onto a specific mathematical manifold known as Matrix Product States (MPS), this method enables the simulation of complex molecular systems with over 20 degrees of freedom—a scale that typically exceeds the capabilities of standard computational frameworks.

Methodology

The core of this approach lies in the Density Matrix Renormalization Group (DMRG) framework, specifically utilizing Matrix Product States (MPS) to represent the time-dependent wavefunction. This choice transforms the problem from one of exponential scaling to one of polynomial scaling.

Tangent-Space Formulation

Rather than attempting to solve the full, infinite-dimensional quantum problem directly, the authors project the dynamics onto the "tangent space" of the MPS manifold. This mathematical innovation transforms the complex Schrödinger equation into a set of coupled differential equations. A key advantage of this specific formulation is that these equations can be integrated exactly using a sweep-based algorithm, ensuring high numerical stability.

Matrix Product Operators (MPO)

The molecular Hamiltonian (the operator representing the system's total energy) is represented as a Matrix Product Operator (MPO). This representation is crucial for handling long-range interactions, such as those between distant electrons, which are often difficult for older methods like Time-Evolving Block Decimation (TEBD) to process efficiently.

Adaptive Dynamics and Imaginary-Time Propagation

To maintain efficiency without sacrificing accuracy, the algorithm incorporates an adaptive bond dimension. The bond dimension is a measure of the quantum entanglement the model can carry; by adjusting it dynamically, the simulation avoids being bogged down by unnecessary data while ensuring it retains enough information to remain accurate. Furthermore, the method supports imaginary-time propagation, allowing for the efficient calculation of ground and excited states in addition to real-time dynamics.

Benchmark Applications

The effectiveness of the TD-DMRG approach was tested against several challenging molecular models:

  • Pyrazine: A classic benchmark in quantum dynamics. The researchers simulated the $S_2 \to S_1$ decay using a full-dimensional 24-mode model. Successfully capturing the dynamics of all 24 vibrational modes simultaneously is a significant milestone, as most state-of-the-art algorithms require simplifying hierarchical treatments to remain tractable.
  • Perylene Bisimide (PBI-I) Aggregates: These were used to test excitonic dynamics—how energy moves through dye molecules. The method was applied to monomers, dimers, and hexamers (up to 66 sites), demonstrating its scalability to large, repetitive structures.
  • Ethylene ($C_2H_4$): This model served to demonstrate the precision of imaginary-time propagation for calculating vibrational energy levels.

Key Insights and Results

Breaking the Exponential Wall

The transition to an MPS-based approach allows computational costs to grow polynomially. This shift makes "full-dimensional" quantum dynamics possible on standard hardware, even for systems that were previously considered too complex.

The Role of Entanglement and Ordering

The success of the simulation is intrinsically linked to "entanglement entropy." While entanglement grows during time evolution, the researchers found that for many spectroscopic problems, the most critical dynamics occur at short time scales, which can be captured with relatively small bond dimensions. Additionally, the physical ordering of atoms or modes within the MPS chain is vital; placing strongly interacting states close to one another significantly accelerates convergence.

Efficiency and Stability

The algorithm demonstrated remarkable stability, allowing for relatively large time steps (up to 1–2 femtoseconds). In the case of the PBI-I hexamer, the spectrum converged with a bond dimension of only $m=20$, highlighting the method's efficiency for large-scale molecular systems.

Conclusion and Future Directions

This TD-DMRG framework represents a significant step forward in simulating large-scale quantum dynamics. By overcoming the curse of dimensionality, it opens the door to studying realistic molecular processes with high precision.

Future work aims to further enhance the method by implementing an "on-the-fly" optimization of the local basis functions, which would make the description of each atom even more flexible. Additionally, the authors plan to incorporate spin symmetry to improve the efficiency of electronic-only dynamics and couple the method with real-time electronic structure calculations for long-term dynamics where molecular bonds may break.

Summary updated

Abstract

Dynamical electronic- and vibrational-structure theories have received a growing interest in the past few years due to their ability to simulate spectra recorded with ultrafast experimental techniques. The exact time evolution of a molecular system can, in principle, be obtained from the time-dependent version of full configuration interaction. Such an approach is, however, limited to few-atom systems due to the exponential increase of its cost with the system dimension. In the present work, we overcome this unfavorable scaling by employing the time-dependent density matrix renormalization group (TD-DMRG) which parametrizes the time-dependent wave function as a matrix product state. The time-dependent Schrödinger equation is then integrated with a sweep-based algorithm, as in standard time-independent DMRG. Unlike other TD-DMRG approaches, the one presented here leads to a set of coupled equations that can be integrated exactly. The resulting theory enables us to study real- and imaginary-time evolutions of Hamiltonians comprising more than 20 degrees of freedom that are challenging for current state-of-the-art quantum dynamics algorithms. We apply our algorithm to the simulation of quantum dynamics of models of increasing complexity, ranging from simple excitonic Hamiltonians to more complex ab initio vibronic ones.
BibTeX
@article{baiardi2019large,
  author = {Alberto Baiardi and Markus Reiher},
  title = {{Large-Scale Quantum Dynamics with Matrix Product States}},
  journal = {Journal of Chemical Theory and Computation},
  year = {2019},
  doi = {10.1021/acs.jctc.9b00301},
}
Figure