研究目的
Investigating the use of deep learning to accurately predict various electronic properties of oligothiophenes (OTs), including their HOMO and LUMO energies, excited-state energies, and associated transition dipole moments, to address the computational challenge in modeling optoelectronic properties of organic semiconductors.
研究成果
Deep learning methods, particularly SchNet, are capable of predicting electronic properties of OSCs with an accuracy comparable to TDDFT. SchNet retains its superior performance even as molecule size increases and outperforms shallow feed-forward neural networks, especially in difficult cases with large molecules or limited training data. The study demonstrates the promise of using SchNet in modeling exciton dynamics in OSCs and opens the door to routine calculation of electronic spectroscopy.
研究不足
The accuracy of transition dipole prediction could be affected by the charge-transfer character of the excited states. The computational cost of training DNNs is high, requiring expensive hardware and long training time.
1：Experimental Design and Method Selection:
High-temperature gas-phase MD simulations were used to generate non-equilibrium molecular configurations. Quantum chemistry calculations were then performed to obtain the quantum properties of these configurations. The datasets of quantum properties were used to train DNN models.
2：Sample Selection and Data Sources:
Non-equilibrium configurations of OTs were harvested from classical MD simulations using the OPLS/2005 force field. Quantum chemical calculations were performed on these configurations to generate datasets of the quantum properties of interest.
3：List of Experimental Equipment and Materials:
Desmond package 3.6 for MD simulations, PySCF program for quantum chemical calculations.
4：6 for MD simulations, PySCF program for quantum chemical calculations.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: MD simulations were performed in the NVT ensemble at 1000K. DFT with CAM-B3LYP functional and 6-31+G(d) basis set was employed to compute HOMO and LUMO energies, and TDDFT calculations were performed within the Tamm-Dancoff approximation to obtain the singlet excited-state energies and associated transition dipoles.
5：0K. DFT with CAM-B3LYP functional and 6-31+G(d) basis set was employed to compute HOMO and LUMO energies, and TDDFT calculations were performed within the Tamm-Dancoff approximation to obtain the singlet excited-state energies and associated transition dipoles.
Data Analysis Methods:
5. Data Analysis Methods: The performances of the selected DNNs were evaluated on HOMO energy, LUMO energy, HOMO-LUMO gap, and the first excited-state energy of OTs.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
