Accelerated diradical character assessment in large datasets of polybenzenoid hydrocarbons using xTB fractional occupation

Polybenzenoid hydrocarbons (PBHs) have garnered significant attention in the field of organic electronics due to their unique electronic properties. To facilitate the design and discovery of new functional organic materials based on these compounds, it is necessary to assess their diradical character. However, this usually requires expensive multireference calculations. In this study, we demonstrate rapid identification and quantification of open-shell character in PBHs using the fractional occupation number weighted electron density metric (<i>N</i>_FOD) calculated with the semiempirical GFN2-xTB (xTB) method. We apply this approach to the entire chemical space of PBHs containing up to 10 rings, a total of over 19k molecules, and find that approximately 7% of the molecules are identified as having diradical character. Our findings reveal a strong correlation between xTB-calculated <i>N</i>_FOD and the more computationally expensive Yamaguchi <i>y</i> and DFT-calculated <i>N</i>_FOD, validating the use of this efficient method for large-scale screening. Additionally, we identify a linear relationship between size and <i>N</i>_FOD value and implement a size-dependent threshold for open-shell character, which significantly improves the accuracy of diradical identification across the chemical space of PBHs. This size-aware approach reduces false positive identifications from 6.97% to 0.38% compared to using a single threshold value. Overall, this work demonstrates that xTB-calculated <i>N</i>_FOD provides a rapid and cost-effective alternative for large-scale screening of open-shell character in PBHs.