Figure 9.

Errors in pairwise solvation (charge–charge interaction) cross terms compared between the canonical GB and the proposed model based on Eqs. 5, 9, 11. The error in $\Delta G_{ij}^{\mathrm{el}}$ is computed relative to numerical PE as error$=\Delta G_{ij}^{\mathrm{el}}\left(\text{GB}\right)-\Delta G_{ij}^{\mathrm{el}}\left(\text{PE}\right)$. The two red dashed lines at ±2k_BT indicate the gross error threshold.