Theoretical and Computational Chemistry

Active Learning Configuration Interaction for Excited States Calculations of Polycyclic Aromatic Hydrocarbons



We present the active learning configuration interaction (ALCI) method for multiconfigurational calculations based on large active spaces. ALCI leverages the use of an active learning procedure to find important electronic configurations among the full configuration space generated within an active space. We tested it for the calculation of singlet-singlet excited states of acenes and pyrene, by using different machine learning algorithms. The ALCI method yields excitation energies within 0.2–0.3 eV from those obtained by traditional complete active space configuration interaction (CASCI) calculations (affordable for active spaces up to 16 electrons in 16 orbitals), by including only a small fraction of the CASCI configuration space in the calculations. For larger active spaces (up to 26 electrons in 26 orbitals), not affordable with traditional CI methods, ALCI captures the trends of experimental excitation energies. Overall ALCI provides satisfactory approximations to large active-space wave functions with up to ten orders of magnitude fewer configurations. These ALCI wave functions are promising and affordable starting points for subsequent second order perturbation theory or pair-density functional theory calculations.


Thumbnail image of ALCI_ChemRxiv.pdf

Supplementary material

Thumbnail image of ALCI_SI.pdf
Supporting Information: Active Learning CI
Supporting Information. S1. Raw Data Extraction and Featurization, S2. Hyperparameter Tuning and ML Model Training, S3. Sensitivity to Iteration Parameters, S4. Generation of Excited Configurations, S5. Machine Learning Model Performance, S6. Active Learning CI Protocol Results, S7. Comparison of ALCI and CASCI wave functions, S8. Computational methods for the CASCI+PT2 calculations