Efficient numerical method for predicting nonlinear optical spectroscopies of open systems

Abstract

Nonlinear optical spectroscopies are powerful tools for probing quantum dynamics in molecular and nanoscale systems. While intuition about ultrafast spectroscopies is often built by considering impulsive optical pulses, actual experiments have finite-duration pulses, which can be important for interpreting and predicting experimental results. We present a new freely available open source method for spectroscopic modeling, called Ultrafast Ultrafast (UF$^2$) Spectroscopy, which enables computationally efficient and convenient prediction of nonlinear spectra, including treatment of arbitrary finite duration pulse shapes. UF$^2$ is a Fourier-based method that requires diagonalization of the Liouvillian propagator of the system density matrix. We also present a Runge-Kutta Euler (RKE) direct propagation method. We include open-systems dynamics in the secular Redfield, full Redfield, and Lindblad formalisms with Markovian baths. For non-Markovian systems, the degrees of freedom corresponding to memory effects are brought into the system and treated nonperturbatively. We analyze the computational complexity of the algorithms and demonstrate numerically that, including the cost of diagonalizing the propagator, UF$^2$ is 20-200 times faster than the direct propagation method for secular Redfield models with arbitrary Hilbert space dimension; that it is similarly faster for full Redfield models at least up to system dimensions where the propagator requires more than 20 GB to store; and that for Lindblad models it is faster up to dimension near 100, with speedups for small systems by factors of over 500. UF$^2$ and RKE are part of a larger open source Ultrafast Software Suite, which includes tools for automatic generation and calculation of Feynman diagrams.