Adiabatic extraction of nonlinear optical properties from real-time time-dependent electronic-structure theory

TL;DR

This paper introduces improved methods for extracting nonlinear optical properties from real-time electronic-structure simulations, highlighting the importance of nonadiabatic effects and demonstrating that quadratic ramping enhances accuracy and efficiency.

Contribution

The study identifies nonadiabatic effects as the cause of deviations in property extraction and proposes quadratic ramping as a more accurate and cost-effective approach.

Findings

Quadratic ramping yields highly accurate polarizabilities and hyperpolarizabilities.

Nonadiabatic effects from finite-time ramping cause deviations in response extraction.

The methods reduce computational cost while maintaining accuracy.

Abstract

Real-time simulations of laser-driven electron dynamics contain information about molecular optical properties through all orders in response theory. These properties can be extracted by assuming convergence of the power series expansion of induced electric and magnetic multipole moments. However, the accuracy relative to analytical results from response theory quickly deteriorates for higher-order responses due to the presence of high-frequency oscillations in the induced multipole moment in the time domain. This problem has been ascribed to missing higher-order corrections. We here demonstrate that the deviations are caused by nonadiabatic effects arising from the finite-time ramping from zero to full strength of the external laser field. Three different approaches, two using a ramped wave and one using a pulsed wave, for extracting electrical properties from real-time time-dependent electronic-structure simulations are investigated. The standard linear ramp is compared to a quadratic ramp, which is found to yield highly accurate results for polarizabilities, and first and second hyperpolarizabilities, at roughly half the computational cost. Results for the third hyperpolarizability are presented along with a simple, computable measure of reliability.