Final Report Summary - OMSQC (Orthogonalization Models in Semiempirical Quantum Chemistry)
The objective of the OMSQC project is to develop fast, practical, and sufficiently accurate electronic structure methods that cover the gap between high-level quantum chemistry and classical force fields, and thus allow realistic treatments of electronic effects in large (bio)chemical systems. The project targets semiempirical quantum-chemistry (SQC) methods that go beyond standard SQC approaches by including explicit orthogonalization as well as dispersion corrections and that are therefore capable of providing a balanced description of ground and excited states in large molecules. The major advances are as follows.
Method development: Novel SQC methods with integrated orthogonalization and dispersion corrections were devised and implemented (ODM2 and ODM3). A comprehensive analysis of the NDDO (neglect of diatomic differential overlap) integral approximation provided further justification for the chosen theoretical framework and allowed us to propose an integral scheme that goes beyond NDDO. The configuration interaction (CI) module was supplemented by a faster general CI algorithm and by very efficient CI-singles variants for large-scale applications. The parametrization technology was further developed both with regard to established optimization algorithms and machine learning. The surface hopping module was improved by adding a variable time step algorithm and the option to simulate internal conversion and intersystem crossing simultaneously.
Parametrization and validation: Extended reference databases were generated covering large sets of reference molecules and reference properties, both from experiment and high-level quantum calculations. These databases were used in the parametrization of our novel SQC methods and in thorough benchmark studies of our orthogonalization-corrected methods (OMx), their variants with a posteriori dispersion corrections (OMx-Dn), and the most recent methods with integrated dispersion corrections (ODMx). These extensive benchmarks demonstrate that our methods outperform all other available NDDO-based semiempirical methods, and they document their accuracy both for ground-state and excited-state properties.
Applications: OMx-CI methods were used in trajectory surface hopping simulations of ultrafast excited-state dynamics in organic chromophores, which provided detailed insight into the photoinduced processes in a large variety of interesting systems, including photoswitches, light-driven molecular motors, and green fluorescent proteins. Further applications involved multiscale QM/MM studies using molecular dynamics simulations and free energy calculations, which addressed reaction mechanisms and photoinduced processes in complex biomolecular systems as well as solvent effects.
Code development: Our in-house SQC program was significantly enhanced by incorporating the methodological advances achieved during the project. It is highly efficient and exploits the benefits from vectorization, shared-memory and distributed-memory parallelization, and acceleration through the use of GPUs. Concerning functionality, the code is unique with regard to the implementation of our novel SQC methods (OMx, ODMx) and the excited-state technology (general CI module, nonadiabatic surface hopping dynamics). It is made available to other groups upon individual request.
In summary, the project has achieved the major objectives of the OMSQC project.
Method development: Novel SQC methods with integrated orthogonalization and dispersion corrections were devised and implemented (ODM2 and ODM3). A comprehensive analysis of the NDDO (neglect of diatomic differential overlap) integral approximation provided further justification for the chosen theoretical framework and allowed us to propose an integral scheme that goes beyond NDDO. The configuration interaction (CI) module was supplemented by a faster general CI algorithm and by very efficient CI-singles variants for large-scale applications. The parametrization technology was further developed both with regard to established optimization algorithms and machine learning. The surface hopping module was improved by adding a variable time step algorithm and the option to simulate internal conversion and intersystem crossing simultaneously.
Parametrization and validation: Extended reference databases were generated covering large sets of reference molecules and reference properties, both from experiment and high-level quantum calculations. These databases were used in the parametrization of our novel SQC methods and in thorough benchmark studies of our orthogonalization-corrected methods (OMx), their variants with a posteriori dispersion corrections (OMx-Dn), and the most recent methods with integrated dispersion corrections (ODMx). These extensive benchmarks demonstrate that our methods outperform all other available NDDO-based semiempirical methods, and they document their accuracy both for ground-state and excited-state properties.
Applications: OMx-CI methods were used in trajectory surface hopping simulations of ultrafast excited-state dynamics in organic chromophores, which provided detailed insight into the photoinduced processes in a large variety of interesting systems, including photoswitches, light-driven molecular motors, and green fluorescent proteins. Further applications involved multiscale QM/MM studies using molecular dynamics simulations and free energy calculations, which addressed reaction mechanisms and photoinduced processes in complex biomolecular systems as well as solvent effects.
Code development: Our in-house SQC program was significantly enhanced by incorporating the methodological advances achieved during the project. It is highly efficient and exploits the benefits from vectorization, shared-memory and distributed-memory parallelization, and acceleration through the use of GPUs. Concerning functionality, the code is unique with regard to the implementation of our novel SQC methods (OMx, ODMx) and the excited-state technology (general CI module, nonadiabatic surface hopping dynamics). It is made available to other groups upon individual request.
In summary, the project has achieved the major objectives of the OMSQC project.