Nonadiabatic Dynamics Simulations for Photoinduced Processes in Molecules and Semiconductors: Methodologies and Applications

Generalization of Quantum-Trajectory Surface Hopping to Multiple Quantum States

In this work, we present a generalization of the quantum trajectory surface hopping (QTSH) to multiple states and its implementation in the Libra package for nonadiabatic dynamics. In lieu of the ad hoc velocity rescaling used in many trajectory-based surface hopping approaches, QTSH utilizes quantum forces to evolve nuclear degrees of freedom continuously. It also lifts the unphysical constraint of enforcing the total energy conservation at the individual trajectory level and rather conserves the total energy at the trajectory ensemble level. Leveraging our new implementation of the multistate QTSH, we perform a comparative analysis of this method with the conventional fewest switches surface hopping approach. We combine the QTSH and decoherence corrections based on the simplified decay of mixing (SDM) and exact factorization (XF), leading to the QTSH-SDM and QTSH-XF schemes. Using the Holstein, superexchange, and phenol model Hamiltonians, we assess the relative accuracy of the resulting combined schemes in reproducing branching ratios, population, and coherence dynamics for a broad range of initial conditions. We observe that the decoherence correction in QTSH is crucial to improve energy conservation as well as the internal consistency between the population from the quantum probability and active state.

Machine learning accelerated nonadiabatic dynamics simulations of materials with excitonic effects

This study presents an efficient methodology for simulating nonadiabatic dynamics of complex materials with excitonic effects by integrating machine learning (ML) models with simplified Tamm-Dancoff approximation (sTDA) calculations. By leveraging ML models, we accurately predict ground-state wavefunctions using unconverged Kohn-Sham (KS) Hamiltonians. These ML-predicted KS Hamiltonians are then employed for sTDA-based excited-state calculations (sTDA/ML). The results demonstrate that excited-state energies, time-derivative nonadiabatic couplings, and absorption spectra from sTDA/ML calculations are accurate enough compared with those from conventional density functional theory based sTDA (sTDA/DFT) calculations. Furthermore, sTDA/ML-based nonadiabatic molecular dynamics simulations on two different materials systems, namely chloro-substituted silicon quantum dot and monolayer black phosphorus, achieve more than 100 times speedup than the conventional linear response time-dependent DFT simulations. This work highlights the potential of ML-accelerated nonadiabatic dynamics simulations for studying the complicated photoinduced dynamics of large materials systems, offering significant computational savings without compromising accuracy.

Conformational and Solvent Effects on the Photoinduced Electron Transfer Dynamics of a Zinc Phthalocyanine-Benzoperylenetriimide Conjugate: A Nonadiabatic Dynamics Simulation

Herein, we employed a combination of static electronic structure calculations and nonadiabatic dynamics simulations at linear-response time dependent density functional theory (LR-TDDFT) level with the optimally tuned range-separated hybrid (OT-RSH) functional to explore the ultrafast photoinduced dynamics of a zinc phthalocyanine-benzoperylenetriimide (ZnPc-BPTI) conjugate. Due to the flexibility of the linker, we identified two major conformations: the stacked conformation (ZnPc-BPTI-1) and the extended conformation (ZnPc-BPTI-2). Since the charge transfer states are much lower than the lowest local excitation in ZnPc-BPTI-1, which is contrary to ZnPc-BPTI-2, the ultrafast electron transfer (~3.6 ps) is only observed in the nonadiabatic simulations of ZnPc-BPTI-1 upon local excitation around the absorption maximum of ZnPc. However, when considering the solvent effects in benzonitrile: the lowest S1 states are both charge transfer states from ZnPc to BPTI for different conformers. Subsequent nonadiabatic dynamics simulations indicate that both conformers experience ultrafast electron transfer in benzonitrile with two time constants of 90 [100] fs and 1.40 [1.43] ps. Our present work not only agrees well with previous experimental study, but also points out the important role of conformational changes and solvent effects in regulating the photodynamics of organic donor-acceptor conjugates.

How Sophisticated Are Neural Networks Needed to Predict Long-Term Nonadiabatic Dynamics?

Nonadiabatic dynamics is key for understanding solar energy conversion and photochemical processes in condensed phases. This often involves the non-Markovian dynamics of the reduced density matrix in open quantum systems, where knowledge of the system's prior states is necessary to predict its future behavior. In this study, we explore time-series machine learning methods for predicting long-time nonadiabatic dynamics based on short-time input data, comparing these methods with the physics-based transfer tensor method (TTM). To understand the impact of memory time on these approaches, we demonstrate that non-Markovian dynamics can be represented as a linear map within the Nakajima-Zwanzig generalized quantum master equation framework. We further propose a practical method to estimate the effective memory time, within a given tolerance, for reduced density matrix propagation. Our predictive models are applied to various physical systems, including spin-boson models, multistate harmonic (MSH) models with Ohmic spectral densities and for a realistic organic photovoltaic system composed of a carotenoid-porphyrin-fullerene triad dissolved in tetrahydrofuran. Results indicate that the simple linear-mapping fully connected neural network (FCN) outperforms the more complicated nonlinear-mapping networks including the gated recurrent unit (GRU) and the convolutional neural network/long short-term memory (CNN-LSTM) in systems with short memory times, such as spin-boson and MSH models. Conversely, the nonlinear CNN-LSTM and GRU models yield higher accuracy in the triad MSH systems characterized by long memory times. These findings offer valuable insights into the role of effective memory time in non-Markovian quantum dynamics, providing practical guidance for the application of time-series machine learning models to complex chemical systems.

Photoinduced Nonadiabatic Dynamics of a Single-Walled Carbon Nanotube-Porphyrin Complex

Single-walled carbon nanotubes (SWCNTs) have gained a lot of attention in the past few decades due to their promising optoelectronic properties. In addition, SWCNTs can form complexes that have good chemical stability and transport properties with other optical functional materials through noncovalent interactions. Elucidating the detailed mechanism of these complexes is of great significance for improving their optoelectronic properties. Nevertheless, simulating the photoinduced dynamics of these complexes accurately is rather challenging since they usually contain hundreds of atoms. To save computational efforts, most of the previous works have ignored the excitonic effects by employing nonadiabatic carrier (electron and hole) dynamics simulations. To properly consider the influence of excitonic effects on the photoinduced ultrafast processes of the SWCNT-tetraphenyl porphyrin (H2TPP) complex and to further improve the computational efficiency, we developed the nonadiabatic molecular dynamics (NAMD) method based on the extended tight binding-based simplified Tamm-Dancoff approximation (sTDA-xTB), which is applied to study the ultrafast photoinduced dynamics of the noncovalent SWCNT-porphyrin complex. In combination with statically electronic structure calculations, the present work successfully reveals the detailed microscopic mechanism of the ultrafast excitation energy transfer process of the complex. Upon local excitation on the H2TPP molecule, an ultrafast energy transfer process occurs from H2TPP (SWCNT-H2TPP*) to SWCNT (SWCNT*-H2TPP) within 10 fs. Then, two slower processes corresponding to the energy transfer from H2TPP to SWCNT and hole transfer from H2TPP to SWCNT take place in the 1 ps time scale. The sTDA-xTB-based electronic structure calculation and NAMD simulation results not only match the previous experimental observations from static and transient spectra but also provide more insights into the detailed information on the complex's photoinduced dynamics. Therefore, the sTDA-xTB-based NAMD method is a powerful theoretical tool for studying the ultrafast photoinduced dynamics in large extended systems with a large number of electronically excited states, which could be helpful for the subsequent design of SWCNT-based functional materials.

Multiconfigurational Surface Hopping: a Time-Dependent Variational Approach with Momentum-Jump Trajectories

The Ehrenfest mean field dynamics and trajectory surface hopping have been widely used in nonadiabatic dynamics simulations. Based on the time-dependent variational principle (TDVP), the multiconfigurational Ehrenfest (MCE) method has also been developed and can be regarded as a multiconfigurational extension of the traditional Ehrenfest dynamics. However, it is not straightforward to apply the TDVP to surface hopping trajectories because there exists momentum jump during surface hops. To solve this problem, we here propose a multiconfigurational surface hopping (MCSH) method, where continuous momenta are obtained by linear interpolation and the interpolated trajectories are used to construct the basis functions for TDVP in a postprocessing manner. As demonstrated in a series of representative spin-boson models, MCSH achieves high accuracy with only several hundred trajectory bases and can uniformly improve the performance of surface hopping. In principle, MCSH can be combined with all kinds of mixed quantum-classical trajectories and thus has the potential to properly describe general nonadiabatic dynamics.

"On-the-Fly" Nonadiabatic Dynamics Simulation on the Ultrafast Photoisomerization of a Molecular Photoswitch Iminothioindoxyl: An RMS-CASPT2 Investigation

Iminothioindoxyl (ITI) is a new class of photoswitch that exhibits many excellent properties including well-separated absorption bands in the visible region for both conformers, ultrafast Z to E photoisomerization as well as the millisecond reisomerization at room temperature for the E isomer, and switchable ability in both solids and various solvents. However, the underlying ultrafast photoisomerization mechanism at the atomic level remains unclear. In this work, we have employed a combination of high-level RMS-CASPT2-based static electronic structure calculations and nonadiabatic dynamics simulations to investigate the ultrafast photoisomerization dynamics of ITI. Based on the minimum-energy structures, minimum-energy conical intersections, linear interpolation internal coordinate paths, and nonadiabatic dynamics simulations, the overall photoisomerization scenario of ITI upon excitation is established. Upon excitation around 416 nm, the molecule will be excited to the S2 state considering its close energy to the experimentally measured absorption maximum and larger oscillator strength, from which ultrafast decay of S2 to S1 state can take place efficiently with a time constant of 62 fs. However, the photoisomerization is not likely to complete in the S2 state since the dihedral associated with the Z to E isomerization changes little during the relaxation. Upon relaxing to the S1 state, the molecule will decay to the S0 state ultrafast with a time constant of 232 fs. In contrast, the decay of the S1 state is important for the isomerization considering that the dihedral related to the isomerization of the hopping structures is close to 90°. Therefore, the S1/S0 intersection region should be important for the isomerization of ITI. Arriving at the S0 state, the molecule can either go back to the original Z reactant or isomerize to the E products. At the end of the 500 fs simulation time, the E configuration accounts for nearly 37% of the final structures. Moreover, the photoisomerization mechanism is different from the isomerization mechanism in the ground state; i.e., instead of the inversion mechanism in the ground state, the photoisomerization prefers the rotation mechanism. Our results not only agree well with previous experimental studies but also provide some novel insights that could be helpful for future improvements in the performance of the ITI photoswitches.

Nonadiabatic Molecular Dynamics in Momentum Space Beyond Harmonic Approximation: Hot Electron Relaxation in Photoexcited Black Phosphorus

We simulated hot-electron relaxation in black phosphorus using the nonadiabatic molecular dynamics (NA-MD) approach with a non-Condon effect in momentum space beyond the harmonic approximation. By comparing simulations at the Γ point in a large supercell with those using a few k-points in a smaller supercell─while maintaining the same number of electronic states within the same energy range, we demonstrate that both setups yield remarkably consistent energy relaxation times, regardless of the initial state energy. This consistency arises from the complementary effects of supercell size in real space and the number of k-points in the reciprocal space. This finding confirms that simulations at a single k-point in large size supercells are an effective approximation for NA-MD with a non-Condon effect. This approach offers significant advantages for complex photophysics, such as intervalley scattering and indirect bandgap charge recombination, and is particularly suitable for large systems without the need for a harmonic approximation.

On-the-fly simulation of time-resolved fluorescence spectra and anisotropy

We combine on-the-fly trajectory surface hopping simulations and the doorway-window representation of nonlinear optical response functions to create an efficient protocol for the evaluation of time- and frequency-resolved fluorescence (TFRF) spectra and anisotropies of the realistic polyatomic systems. This approach gives the effective description of the proper (e.g., experimental) pulse envelopes, laser field polarizations, and the proper orientational averaging of TFRF signals directly from the well-established on-the-fly nonadiabatic dynamic simulations without extra computational cost. To discuss the implementation details of the developed protocol, we chose cis-azobenzene as a prototype to simulate the time evolution of the TFRF spectra governed by its nonadiabatic dynamics. The results show that the TFRF is determined by the interplay of several key factors, i.e., decays of excited-state populations, evolution of the transition dipole moments along with the dynamic propagation, and scaling factor of the TFRF signals associated with the cube of emission frequency. This work not only provides an efficient and effective approach to simulate the TFRF and anisotropies of realistic polyatomic systems but also discusses the important relationship between the TFRF signals and the underlining nonadiabatic dynamics.