Assessment of local coupled cluster methods for excited states of BODIPY/Aza-BODIPY families

It was previously reported that Laplace transformed local CC2 (LCC2*) provided the best agreement (MAE = 0.145 eV) when comparing vertical excitation energies to experimental lambda(max) for a benchmark set of 17 BODIPY/Aza-BODIPY molecules. However, these energies did not agree with values obtained from canonical CC2. Here we report LCC2* computations of vertical excitation energies on the same benchmark set of molecules using a newly implemented treatment of the ground state. Comparison with resolution-of-identity approximate coupled cluster to second-order (RI-CC2) results demonstrate that the new LCC2* results agree quantitatively. Furthermore, these values can easily be corrected empirically to also provide excellent agreement with the experiment. We show that the local algebraic diagrammatic construction to second-order (LADC(2)) method exhibits the same differences between implementations as seen for LCC2. The source of the difference is traced to an improved treatment of the ground state in the local methods, which decreases agreement with the experiment (as attributed to a fortuitous cancellation of errors) but significantly improves agreement with RI-CC2. While the absolute vertical excitation energies now show larger deviations, there remains a strong linear correlation between the LCC2* results and the experiment. For the 17 BODIPY/Aza-BODIPY molecules vertical excitation energies are determined using DLPNO-STEOM-CCSD and shown to have excellent agreement with experimental lambda(max) (MAE = 0.145 eV), which is the best of all the single-reference methods. The vertical excitation energies are determined using LCC2*, empirically corrected LCC2*, and RI-CC2 for a series of eight large BODIPYs and Aza-BODIPYs.