研究目的
To improve the predictibility power of the standard GW/BSE approach for triplet states by optimal use of random phase approximation (RPA) screening in order to obtain best estimate of quasi-particle (QP) gap and exciton binding energy which are key for an accurate prediction of excitation energies (EEs).
研究成果
The screening mixing ansatz reduces the mean absolute error of the traditional HF-based GW/BSE by 56% to 0.26 eV, providing a parameter-free alternative for accurate triplet state predictions. It offers computational efficiency by replacing eigenvalue self-consistency with a one-shot procedure. However, it has limitations for certain molecules and systems with high degeneracy.
研究不足
The screening mixing ansatz may not perform optimally for all molecules, such as pyridine, s-tetrazine, and p-benzoquinone, due to less effective compensation effects in RPA screening. It is not suitable for systems with significant static correlation, like transition metal complexes, where it may fail similarly to other methods.
1：Experimental Design and Method Selection:
The study employs a screening mixing many-body ansatz (Scheme B) to optimize the GW and BSE calculations. It involves injecting RPA screening from different starting points (e.g., W0-RPA@LDA into GW and W0-RPA@HF into BSE) to achieve optimal QP gap and exciton binding energy. Theoretical models include GW self-energy and BSE eigenvalue equations.
2：Sample Selection and Data Sources:
The Thiel set of 20 organic molecules with 63 triplet excitation energies is used, with theoretical best estimates (TBE-2) from Silva-Junior et al. as references.
3：List of Experimental Equipment and Materials:
Computational methods are used; no physical equipment is mentioned. Software: MOLGW package for ab initio many-body calculations. Basis sets: aug-cc-pVQZ Dunning basis set. Techniques: Resolution-of-the-identity for computational efficiency.
4：Experimental Procedures and Operational Workflow:
Calculations start with HF orbitals to avoid DFT biases. QP energies are obtained by solving the GW equation graphically. BSE is solved for the full matrix without Tamm–Dancoff approximation. Screening mixing involves generating RPA screenings from LDA and HF and injecting them into GW and BSE steps.
5：Data Analysis Methods:
Excitation energies are compared to TBE-2 references. Errors are quantified using mean absolute error, mean signed error, mean squared error, and root mean squared error.
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
