Ultrafast two-step process in the non-adiabatic relaxation of the CH2 molecule

Ehrenfest dynamics with localized atomic-orbital basis sets within the projector augmented-wave method

Density functional theory with linear combination of atomic orbitals (LCAO) basis sets is useful for studying large atomic systems, especially when it comes to computationally highly demanding time-dependent dynamics. We have implemented the Ehrenfest molecular dynamics (ED) method with the approximate approach of Tomfohr and Sankey within the projector augmented-wave code GPAW. We apply this method to small molecules as well as larger periodic systems and elucidate its limits, advantages, and disadvantages in comparison with the existing implementation of Ehrenfest dynamics with a real-space grid representation. For modest atomic velocities, LCAO-ED shows satisfactory accuracy at a much reduced computational cost. This method will be particularly useful for modeling ion irradiation processes that require large amounts of vacuum in the simulation cell.

Non-adiabatic Couplings in Surface Hopping with Tight Binding Density Functional Theory: The Case of Molecular Motors

Nonadiabatic molecular dynamics (NAMD) has become an essential computational technique for studying the photophysical relaxation of molecular systems after light absorption. These phenomena require approximations that go beyond the Born-Oppenheimer approximation, and the accuracy of the results heavily depends on the electronic structure theory employed. Sophisticated electronic methods, however, make these techniques computationally expensive, even for medium size systems. Consequently, simulations are often performed on simplified models to interpret the experimental results. In this context, a variety of techniques have been developed to perform NAMD using approximate methods, particularly density functional tight binding (DFTB). Despite the use of these techniques on large systems, where ab initio methods are computationally prohibitive, a comprehensive validation has been lacking. In this work, we present a new implementation of trajectory surface hopping combined with DFTB, utilizing nonadiabatic coupling vectors. We selected the methaniminium cation and furan systems for validation, providing an exhaustive comparison with the higher-level electronic structure methods. As a case study, we simulated a system from the class of molecular motors, which has been extensively studied experimentally but remains challenging to simulate with ab initio methods due to its inherent complexity. Our approach effectively captures the key photophysical mechanism of dihedral rotation after the absorption of light. Additionally, we successfully reproduced the transition from the bright to dark states observed in the time-dependent fluorescence experiments, providing valuable insights into this critical part of the photophysical behavior in molecular motors.

SpaiNN: equivariant message passing for excited-state nonadiabatic molecular dynamics

Excited-state molecular dynamics simulations are crucial for understanding processes like photosynthesis, vision, and radiation damage. However, the computational complexity of quantum chemical calculations restricts their scope. Machine learning offers a solution by delivering high-accuracy properties at lower computational costs. We present SpaiNN, an open-source Python software for ML-driven surface hopping nonadiabatic molecular dynamics simulations. SpaiNN combines the invariant and equivariant neural network architectures of SchNetPack with SHARC for surface hopping dynamics. Its modular design allows users to implement and adapt modules easily. We compare rotationally-invariant and equivariant representations in fitting potential energy surfaces of multiple electronic states and properties arising from the interaction of two electronic states. Simulations of the methyleneimmonium cation and various alkenes demonstrate the superior performance of equivariant SpaiNN models, improving accuracy, generalization, and efficiency in both training and inference.

Optimizing Conical Intersections without Explicit Use of Non-Adiabatic Couplings

We present two alternative methods for optimizing minimum energy conical intersection (MECI) molecular geometries without knowledge of the derivative coupling (DC). These methods are based on the utilization of Lagrange multipliers: (i) one method uses an approximate calculation of the DC, while the other (ii) do not require the DC. Both methods use the fact that information on the DC is contained in the Hessian of the squared energy difference. Tests done on a set of small molecular systems, in comparison with other methods, show the ability of the proposed methods to optimize MECIs. Finally, we apply the methods to the furimamide molecule, to optimize and characterize its S1/S2 MECI, and to optimizing the S0/S1 MECI of the silver trimer.

Parametrically Managed Activation Functions for Improved Global Potential Energy Surfaces for Six Coupled 5A' States and Fourteen Coupled 3A' States of O + O2

We report new potential energy surfaces for six coupled 5A' states and 14 coupled 3A' states of O3. The new surfaces are created by parametrically managed diabatization by deep neural network (PM-DDNN). The PM-DDNN method uses calculated adiabatic potential energy surfaces to discover and fit an underlying adiabatic-equivalent set of diabatic surfaces and their couplings and obtains the fit to the adiabatic surfaces by diagonalization of the diabatic potential energy matrix (DPEM). The procedure yields the adiabatic surfaces and their gradients, as well as the DPEM and its gradient. If desired one can also compute the nonadiabatic coupling due to the transformation. The present work improves on previous work by using a new coordinate to guide the decay of the neural network contribution to the many-body fit to the whole DPEM. The main objective was to obtain smoother potentials than the previous ones with better suitability for dynamics calculations, and this was achieved. Furthermore, we obtained suitably small deviations from the input reference data. For the six coupled 5A' surfaces, the 60,366 data below 10 eV are fit with a mean unsigned error (MUE) of 49 meV, and for the 14 coupled 3A' surfaces, the 76,733 data below 10 eV are fit with an MUE of 28 meV. The data below 5 eV fit even more accurately with MUEs of 37 meV (5A') and 20 meV (3A').

On the description of conical intersections between excited electronic states with LR-TDDFT and ADC(2)

Conical intersections constitute the conceptual bedrock of our working understanding of ultrafast, nonadiabatic processes within photochemistry (and photophysics). Accurate calculation of potential energy surfaces within the vicinity of conical intersections, however, still poses a serious challenge to many popular electronic structure methods. Multiple works have reported on the deficiency of methods like linear-response time-dependent density functional theory within the adiabatic approximation (AA LR-TDDFT) or algebraic diagrammatic construction to second-order [ADC(2)]-approaches often used in excited-state molecular dynamics simulations-to describe conical intersections between the ground and excited electronic states. In the present study, we focus our attention on conical intersections between excited electronic states and probe the ability of AA LR-TDDFT and ADC(2) to describe their topology and topography, using protonated formaldimine and pyrazine as two exemplar molecules. We also take the opportunity to revisit the performance of these methods in describing conical intersections involving the ground electronic state in protonated formaldimine-highlighting in particular how the intersection ring exhibited by AA LR-TDDFT can be perceived either as a (near-to-linear) seam of intersection or two interpenetrating cones, depending on the magnitude of molecular distortions within the branching space.

Machine Learning for Electronically Excited States of Molecules

Electronically excited states of molecules are at the heart of photochemistry, photophysics, as well as photobiology and also play a role in material science. Their theoretical description requires highly accurate quantum chemical calculations, which are computationally expensive. In this review, we focus on not only how machine learning is employed to speed up such excited-state simulations but also how this branch of artificial intelligence can be used to advance this exciting research field in all its aspects. Discussed applications of machine learning for excited states include excited-state dynamics simulations, static calculations of absorption spectra, as well as many others. In order to put these studies into context, we discuss the promises and pitfalls of the involved machine learning techniques. Since the latter are mostly based on quantum chemistry calculations, we also provide a short introduction into excited-state electronic structure methods and approaches for nonadiabatic dynamics simulations and describe tricks and problems when using them in machine learning for excited states of molecules.

Machine Learning for Electronically Excited States of Molecules

Electronically excited states of molecules are at the heart of photochemistry, photophysics, as well as photobiology and also play a role in material science. Their theoretical description requires highly accurate quantum chemical calculations, which are computationally expensive. In this review, we focus on not only how machine learning is employed to speed up such excited-state simulations but also how this branch of artificial intelligence can be used to advance this exciting research field in all its aspects. Discussed applications of machine learning for excited states include excited-state dynamics simulations, static calculations of absorption spectra, as well as many others. In order to put these studies into context, we discuss the promises and pitfalls of the involved machine learning techniques. Since the latter are mostly based on quantum chemistry calculations, we also provide a short introduction into excited-state electronic structure methods and approaches for nonadiabatic dynamics simulations and describe tricks and problems when using them in machine learning for excited states of molecules.

On-the-Fly Symmetrical Quasi-Classical Dynamics with Meyer-Miller Mapping Hamiltonian for the Treatment of Nonadiabatic Dynamics at Conical Intersections

The on-the-fly version of the symmetrical quasi-classical dynamics method based on the Meyer-Miller mapping Hamiltonian (SQC/MM) is implemented to study the nonadiabatic dynamics at conical intersections of polyatomic systems. The current on-the-fly implementation of the SQC/MM method is based on the adiabatic representation and the dressed momentum. To include the zero-point energy (ZPE) correction of the electronic mapping variables, we employ both the γ-adjusted and γ-fixed approaches. Nonadiabatic dynamics of the methaniminium cation (CH₂NH₂⁺) and azomethane are simulated using the on-the-fly SQC/MM method. For CH₂NH₂⁺, both ZPE correction approaches give reasonable and consistent results. However, for azomethane, the γ-adjusted version of the SQC/MM dynamics behaves much better than the γ-fixed version. Further analysis indicates that it is always recommended to use the γ-adjusted SQC/MM dynamics in the on-the-fly simulation of photoinduced dynamics of polyatomic systems, particularly when the excited state is well separated from the ground state in the Franck-Condon region. This work indicates that the on-the-fly SQC/MM method is a powerful simulation protocol to deal with the nonadiabatic dynamics of realistic polyatomic systems.

Molecular Photochemistry: Recent Developments in Theory

Photochemistry is a fascinating branch of chemistry that is concerned with molecules and light. However, the importance of simulating light-induced processes is reflected also in fields as diverse as biology, material science, and medicine. This Minireview highlights recent progress achieved in theoretical chemistry to calculate electronically excited states of molecules and simulate their photoinduced dynamics, with the aim of reaching experimental accuracy. We focus on emergent methods and give selected examples that illustrate the progress in recent years towards predicting complex electronic structures with strong correlation, calculations on large molecules, describing multichromophoric systems, and simulating non-adiabatic molecular dynamics over long time scales, for molecules in the gas phase or in complex biological environments.

SSAIMS-Stochastic-Selection Ab Initio Multiple Spawning for Efficient Nonadiabatic Molecular Dynamics

Ab initio multiple spawning provides a powerful and accurate way of describing the excited-state dynamics of molecular systems, whose strength resides in the proper description of coherence effects during nonadiabatic processes thanks to the coupling of trajectory basis functions. However, the simultaneous propagation of a large number of trajectory basis functions can be numerically inconvenient. We propose here an elegant and simple solution to this issue, which consists of (i) detecting uncoupled groups of coupled trajectory basis functions and (ii) selecting stochastically one of these groups to continue the ab initio multiple spawning dynamics. We show that this procedure can reproduce the results of full ab initio multiple spawning dynamics in cases where the uncoupled groups of trajectory basis functions stay uncoupled throughout the dynamics (which is often the case in high-dimensional problems). We present and discuss the aforementioned idea in detail and provide simple numerical applications on indole, ethylene, and protonated formaldimine, highlighting the potential of stochastic-selection ab initio multiple spawning.

Combining SchNet and SHARC: The SchNarc Machine Learning Approach for Excited-State Dynamics

In recent years, deep learning has become a part of our everyday life and is revolutionizing quantum chemistry as well. In this work, we show how deep learning can be used to advance the research field of photochemistry by learning all important properties-multiple energies, forces, and different couplings-for photodynamics simulations. We simplify such simulations substantially by (i) a phase-free training skipping costly preprocessing of raw quantum chemistry data; (ii) rotationally covariant nonadiabatic couplings, which can either be trained or (iii) alternatively be approximated from only ML potentials, their gradients, and Hessians; and (iv) incorporating spin-orbit couplings. As the deep-learning method, we employ SchNet with its automatically determined representation of molecular structures and extend it for multiple electronic states. In combination with the molecular dynamics program SHARC, our approach termed SchNarc is tested on two polyatomic molecules and paves the way toward efficient photodynamics simulations of complex systems.

A molecular perspective on Tully models for nonadiabatic dynamics

We present a series of standardized molecular tests for nonadiabatic dynamics, reminiscent of the one-dimensional Tully models proposed in 1990.

Machine learning enables long time scale molecular photodynamics simulations

Photo-induced processes are fundamental in nature but accurate simulations of their dynamics are seriously limited by the cost of the underlying quantum chemical calculations, hampering their application for long time scales. Here we introduce a method based on machine learning to overcome this bottleneck and enable accurate photodynamics on nanosecond time scales, which are otherwise out of reach with contemporary approaches. Instead of expensive quantum chemistry during molecular dynamics simulations, we use deep neural networks to learn the relationship between a molecular geometry and its high-dimensional electronic properties. As an example, the time evolution of the methylenimmonium cation for one nanosecond is used to demonstrate that machine learning algorithms can outperform standard excited-state molecular dynamics approaches in their computational efficiency while delivering the same accuracy.

Surface Hopping Dynamics beyond Nonadiabatic Couplings for Quantum Coherence

Description of correct electron-nuclear couplings is crucial in modeling of nonadiabatic dynamics. Within traditional semiclassical or mixed quantum-classical dynamics, the coupling between quantum electronic states and classical nuclear trajectories is governed by nonadiabatic coupling vectors coupled to classical nuclear momenta. This enables us to develop a very powerful nonadiabatic dynamics algorithm, namely, surface hopping dynamics, which can describe the splitting of nuclear wave packets and detailed balance. Despite its efficiency and practicality, it suffers from the lack of quantum decoherence due to incorrect accounts for the electron-nuclear coupling. Here we present a new surface hopping algorithm based on the exact electron-nuclear correlation from the exact factorization of molecular wave functions. This algorithm demands comparable computational costs to existing surface hopping methods. Numerical simulations with two-state models and a multidimensional multistate realistic molecule show that the electron-nuclear coupling beyond the nonadiabatic coupling terms can describe the quantum coherence properly.

An On-the-Fly Surface-Hopping Program JADE for Nonadiabatic Molecular Dynamics of Polyatomic Systems: Implementation and Applications

Nonadiabatic dynamics simulations have rapidly become an indispensable tool for understanding ultrafast photochemical processes in complex systems. Here, we present our recently developed on-the-fly nonadiabatic dynamics package, JADE, which allows researchers to perform nonadiabatic excited-state dynamics simulations of polyatomic systems at an all-atomic level. The nonadiabatic dynamics is based on Tully's surface-hopping approach. Currently, several electronic structure methods (CIS, TDHF, TDDFT(RPA/TDA), and ADC(2)) are supported, especially TDDFT, aiming at performing nonadiabatic dynamics on medium- to large-sized molecules. The JADE package has been interfaced with several quantum chemistry codes, including Turbomole, Gaussian, and Gamess (US). To consider environmental effects, the Langevin dynamics was introduced as an easy-to-use scheme into the standard surface-hopping dynamics. The JADE package is mainly written in Fortran for greater numerical performance and Python for flexible interface construction, with the intent of providing open-source, easy-to-use, well-modularized, and intuitive software in the field of simulations of photochemical and photophysical processes. To illustrate the possible applications of the JADE package, we present a few applications of excited-state dynamics for various polyatomic systems, such as the methaniminium cation, fullerene (C20), p-dimethylaminobenzonitrile (DMABN) and its primary amino derivative aminobenzonitrile (ABN), and 10-hydroxybenzo[h]quinoline (10-HBQ).

Evaluation of the importance of spin-orbit couplings in the nonadiabatic quantum dynamics with quantum fidelity and with its efficient "on-the-fly" ab initio semiclassical approximation

We propose to measure the importance of spin-orbit couplings (SOCs) in the nonadiabatic molecular quantum dynamics rigorously with quantum fidelity. To make the criterion practical, quantum fidelity is estimated efficiently with the multiple-surface dephasing representation (MSDR). The MSDR is a semiclassical method that includes nuclear quantum effects through interference of mixed quantum-classical trajectories without the need for the Hessian of potential energy surfaces. Two variants of the MSDR are studied, in which the nuclei are propagated either with the fewest-switches surface hopping or with the locally mean field dynamics. The fidelity criterion and MSDR are first tested on one-dimensional model systems amenable to numerically exact quantum dynamics. Then, the MSDR is combined with "on-the-fly" computed electronic structure to measure the importance of SOCs and nonadiabatic couplings in the photoisomerization dynamics of CH(2)NH(2)(+) considering 20 electronic states and in the collision of F + H(2) considering six electronic states.

Ultrafast internal conversion pathway and mechanism in 2-(2'-hydroxyphenyl)benzothiazole: a case study for excited-state intramolecular proton transfer systems

We study the ultrafast electronic relaxation of the proton transfer compound 2-(2'-hydroxyphenyl)benzothiazole (HBT) in a joint approach of femtosecond pump-probe experiments and dynamics simulations. The measurements show a lifetime of 2.6 ps for the isolated molecule in the gas phase in contrast to approximately 100 ps for cyclohexane solution. This unexpected decrease by a factor of 40 for the gas phase is explained by ultrafast internal conversion to the ground state promoted by an inter-ring torsional mode. The quantum chemical calculations based on multireference configuration interaction clearly demonstrate that a S(0)/S(1) conical intersection at a 90 degrees twisted structure exists and is responsible for the ultrafast decay. The reaction path leading from the keto form of HBT to this intersection is practically barrierless on the S(1) surface. The on-the-fly dynamics simulations using time-dependent density functional theory show that after electronic excitation to the S(1) state and after fast excited-state proton transfer (30-50 fs), HBT reaches the region of the S(1)/S(0) crossing within about 500 fs, which will lead to the observed 2.6 ps deactivation to the ground state. After the internal conversion, HBT branches in two populations, one that rapidly closes the proton transfer cycle and another (trans-keto) that takes approximately 100 ps for that step.

Photophysics and deactivation pathways of thymine

Combined complete active space perturbation theory (CASPT2) and multireference configuration interaction calculations with single and double excitations (MR-CISD) were performed in order to explore possible deactivation pathways of thymine after photoexcitation. Equilibrium geometries are reported together with a total of eight extremes (minima or maxima) on the crossing seam (MXS), through which such radiationless transitions may occur. Furthermore, conformational analysis allows grouping these conical intersections in five distinct types. Reaction paths were calculated connecting the S1 (1)n pi* minimum with the lowest-energy MXS of each group. Two distinct types of paths were observed, both with features that should delay the internal conversion to the ground state. This is shown to provide a possible explanation for the relatively long excited-state lifetime of thymine.

Simulation of the photodeactivation of formamide in the nO-pi* and pi-pi* states: an ab initio on-the-fly surface-hopping dynamics study

The short-time photodynamics (1 ps) of formamide in its low-lying singlet excited n(O)-pi(*) and pi-pi(*) states have been investigated by the direct trajectory surface-hopping method based on multiconfigurational ab initio calculations. The simulations showed that in both states, the primary deactivation process is C-N bond dissociation. In the ground state, the energy is transferred to (a) translational motion of the HCO and NH(2) fragments, (b) additional C-H dissociation from the vibrationally hot HCO fragment, or (c) formation of NH(3) and CO. In addition to the C-N dissociation pathway, C-O bond fission is found to be an additional primary deactivation path in the pi-pi(*) dynamics. From fractional occupations of trajectories, lifetimes of formamide were estimated: tau(S(1))=441 fs and tau(S(2))=66 fs.