| t-DMRG | |
|---|---|
| Name | Time-dependent density matrix renormalization group |
| Acronym | t-DMRG |
| Developer | Steven R. White; developments by Ulrich Schollwöck, G. Vidal and others |
| Introduced | 1992 (DMRG); time-dependent extensions mid-1990s–2000s |
| Field | Quantum many-body physics; Condensed matter physics |
| Related | Density matrix renormalization group, Matrix product state, Tensor network |
t-DMRG
t-DMRG is a set of numerical methods for simulating the real-time (and imaginary-time) evolution of low-dimensional quantum many-body systems using tensor network representations, notably matrix product states. It matters because it enables quantitative study of non-equilibrium dynamics, quench phenomena, transport, and thermalization in paradigmatic models of condensed matter physics and quantum information where exact diagonalization is infeasible.
t-DMRG grew from the success of the Density matrix renormalization group (DMRG) for ground states of one-dimensional systems and was adapted to dynamics to address problems in quantum quenches, transport in Hubbard model and Heisenberg model chains, and dynamics relevant to ultracold atoms experiments at institutions such as Institut für Quantenoptik groups and groups led by experimentalists like Immanuel Bloch. t-DMRG sits at the intersection of computational many-body physics, quantum statistical mechanics, and quantum information theory by leveraging entanglement structure to compress many-body wavefunctions efficiently.
The method represents a quantum state as a matrix product state (MPS) and evolves it under a Hamiltonian using decompositions such as the Trotter–Suzuki decomposition or variational principles. Early time-dependent schemes include the time-evolving block decimation (TEBD) introduced by Guifre Vidal and time-adaptive DMRG refinements by Steven R. White and Adrian E. Feiguin. Key theoretical ingredients are Schmidt decompositions, reduced density matrix truncation, and control of discarded weight. Algorithms implement operator exponentials via Suzuki–Trotter steps or use the time-dependent variational principle (TDVP) over the MPS manifold pioneered in later works by Haegeman et al., connecting to concepts from differential geometry in variational optimization.
Practical t-DMRG requires careful management of bond dimension, local basis truncation, and time-step errors. Implementations appear in software packages such as ALPS project, ITensor, and specialist codes developed in research groups at Riken and Max Planck Institute for the Physics of Complex Systems. Computational cost scales with system size and maximum bond dimension χ; memory and CPU demands rise rapidly as entanglement grows. Techniques like adaptive time-step control, Krylov subspace methods, and MPO (matrix product operator) compression are commonly employed. Parallelization strategies target shared-memory and distributed systems used at national supercomputing centers like Oak Ridge National Laboratory and CINECA.
t-DMRG has been applied across a range of problems: real-time dynamics after quantum quenchs in the Bose–Hubbard model relevant to optical lattice experiments; spin transport and spintronics questions in Heisenberg chains; computation of spectral functions and response in the Hubbard model and Kondo model; non-equilibrium steady states and driven open systems when combined with Lindblad formalisms in collaboration with work on open quantum systems. Studies have informed understanding of thermalization versus many-body localization (MBL), and comparisons with experiments at Harvard University and MIT have validated predictions about light-cone spreading and correlation propagation.
A central limitation of t-DMRG arises from linear (or faster) growth of bipartite entanglement entropy after quenches, which forces bond dimensions to increase exponentially to maintain accuracy; this constrains accessible time scales. Other error sources include Trotterization errors, truncation errors from discarded Schmidt weights, and finite-size effects. Methods such as error extrapolation, monitoring discarded weight, and using TDVP can mitigate but not eliminate these limits. The challenge has equity implications: accurate long-time simulations require large computational resources available primarily to well-funded institutions, raising issues about access and reproducibility in computational physics.
Numerous extensions address t-DMRG weaknesses: the time-dependent variational principle (TDVP) on MPS, matrix product operator (MPO) representations for mixed states, and infinite-size algorithms like iTEBD for translationally invariant systems. Related tensor network families include projected entangled pair states (PEPS) for two dimensions, multiscale entanglement renormalization ansatz (MERA) for critical systems, and tensor network approaches used in quantum chemistry and high-energy contexts. Hybrid strategies combine t-DMRG with quantum Monte Carlo, dynamical mean-field theory (DMFT), or variational quantum algorithms on near-term quantum computer prototypes to extend reach. Research communities coordinating workshops and conferences—such as the Gordon Research Conferences and specialized sessions at the American Physical Society meetings—continue to drive method development with attention to open science and inclusive collaboration.
Category:Numerical methods in quantum mechanics Category:Tensor network states Category:Many-body physics