Skip to main content
. 2021 Nov 17;17(12):7518–7530. doi: 10.1021/acs.jctc.1c00769

Table 3. ALCI Protocol Results with the Optimized Input Parameters for Tetracene, Pentacene, and Hexacenea.

system ML algorithm threshold for CI coeff. average number of iterations number of important configurations number of important SDs excitation energy (eV)
tetracene (18e, 18o) KRC 0.01 15.0 1759 53 213 3.86
  0.005 22.7 4788 251 320 3.74
ANN 0.01 10.3 1491 32 663 3.91
  0.005 18.0 3944 179 952 3.73
XGBoost 0.01 15.0 1741 53 125 3.88
  0.005 24.3 4642 234 622 3.75
CASCI (18e, 18o) 44 152 809b 2 363 904 400b N/A
pentacene (22e, 22o) KRC 0.01 12.7 1793 31 491 3.50
  0.005 25.0 4780 216 678 3.46
ANN 0.01 10.7 1713 22 622 3.44
  0.005 16.0 4195 175 336 3.48
XGBoost 0.01 17.0 1979 47 645 3.47
  0.005 25.0 4941 231 838 3.46
CASCI (22e, 22o) 3 241 135 527b 497 634 306 624b N/A
hexacene (26e, 26o) KRC 0.01 17.3 2430 27 468 2.89
  0.005 18.7 4061 58 237 3.02
ANN 0.01 12.3 2366 24 760 2.89
  0.005 15.0 4574 101 015 3.03
XGBoost 0.01 24.0 2670 51 247 2.93
  0.005 36.3 5952 245 534 3.03
CASCI (26e, 26o) 241 813 226 151b 108 172 480 360 000b N/A
a

Results are average values of three separate calculations for each model that are performed to obtain better statistics.

b

Total number of configurations or determinants in the active space.