arXiv:2609.23295
Non-Markovian Quantum Dynamics of Exciton-Polaritons
Rajanya Sarkar, Pritha Ghosh, Arshath Manjalingal, David R. Reichman, Arkajit Mandal
Abstract
Exciton-polaritons, hybrid light-matter quasiparticles formed when a material interacts with a confined electric field, have experimentally been shown to exhibit mesoscale coherent quantum propagation that remains robust at room temperature. However, an accurate and direct quantum dynamical simulation of this phenomenon that does not resort to semi-classical approximations is prohibitively expensive computationally, limiting the microscopic understanding of the rich dynamical interplay among phonons, photons, and electrons under collective light-matter coupling. To address this fundamental challenge, we develop a non-Markovian master equation approach which enables the fully quantum mechanical simulation of non-equilibrium exciton-polariton dynamics and captures phonon-induced decoherence and dissipation beyond the conventional Markovian limit. To carry out this task, a procedure is developed in which the wave vector space is coarse-grained and each diagonal element of the density matrix is evolved in parallel. To demonstrate the utility of this approach, we simulate exciton-polariton transport in TIPS-pentacene. We find that our approach reasonably captures the experimentally observed renormalization of the polariton group velocity, which originates from the phonon-induced non-Markovian Lamb shift. We further show that this renormalization cannot be reproduced within conventional Markovian theories.
AI Summary
The paper addresses whether phonon memory effects are needed to describe room-temperature exciton-polariton transport, particularly the experimentally observed reduction of propagation velocity. It develops a non-Markovian reduced-density-matrix method for a multimode Holstein–Tavis–Cummings model projected onto the lower-polariton manifold, with a memory kernel evaluated to second order in polariton–phonon coupling. The computational scheme combines an exponential Prony decomposition of bath correlations, auxiliary-variable time-local propagation, independent evolution of density-matrix diagonals at fixed momentum difference, and nonuniform momentum-space coarse-graining. It is applied to a two-dimensional TIPS-pentacene cavity model representing $12001\times12001\approx1.4\times10^8$ emitters. Phonon spectral densities are parametrized using DFTB/TD-DFTB molecular simulations. The simulations show ballistic propagation over hundreds of femtoseconds with a phonon-induced reduction in wave-front velocity, in reasonable agreement with existing experiments. Comparisons with Markovian calculations attribute the reduction to a phonon-induced Lamb shift: omitting this shift misses the effect, while retaining its Markovian form recovers it only partially and remains quantitatively discrepant. Purity calculations with varied refractive index support a connection between the low polariton density of states, suppressed phonon scattering, and longer-lived coherence. The treatment is perturbative rather than numerically exact and excludes the upper-polariton manifold. Residual disagreement with experiment is attributed to possible Hamiltonian-parametrization errors and the second-order memory-kernel approximation. Higher-order or numerically evaluated kernels are proposed as extensions, but are not demonstrated.
Central claim may not hold: The printed auxiliary-tensor updates and density-matrix equation produce a positive second derivative of the total population instead of conserving trace, so they cannot constitute a valid reduced quantum dynamics as written. The stated interaction-picture transformation has the opposite phase sign from that required to remove the free evolution in the Nakajima–Zwanzig equation. The purity approximation uses squared sampling multiplicities where the stated point-sampling quadrature requires a single multiplicity, leaving the reported coherence diagnostic unsupported by that formula.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.