Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This work introduces two benchmark datasets, SEFB-15 and CSFB-14, for single-electron and closed-shell fluorine bonding, respectively, providing high-level ab initio reference binding energies and bond distances. The study rigorously validates these references against coupled-cluster methods and assesses the performance of the M06-2X density functional. M06-2X is found to be reliable for closed-shell F-bonding but significantly overbinds and underestimates bond distances in single-electron F-bonding systems, an anomaly not resolved by basis set or BSSE corrections.
M06-2X density functional fails dramatically on single-electron fluorine bonding, overbinding by up to 75% and underestimating bond lengths by 9%, revealing a critical blind spot for this widely used functional.
Halogen bonding in which a covalently bound fluorine acts as halogen-bond donor (F-bonding) is rare, and its open-shell variant with radical electron donors remains unexplored. We introduce SEFB-15 dataset (15 open-shell DF$\cdots$CH$_3$ complexes) exemplifying single-electron F-bonding (SEFB) and CSFB-14 dataset (14 closed-shell DF$\cdots$NH$_3$ complexes) representing closed-shell F-bonding (CSFB). Reference binding energies (BEs) at DF-RMP2-F12/CBS//DF-RMP2-F12/VDZ-F12 are validated against CCSD(T), CCSD(T)-F12b, W1-F12 and W2-F12 for selected systems (DF = F$_2$, NCF, CNF, NH$_3$F$^+$ and NH$_2$F$_2^+$), with MA%D of 4.5% (DF$\cdots$CH$_3$) and 3.4% (DF$\cdots$NH$_3$) against CCSD(T)-F12b/CBS//CCSD(T)-F12b/VDZ-F12. BEs from W1-F12 and W2-F12 are identical. Reference F-bond distances at DF-RMP2-F12/VDZ-F12 are validated against CCSD(T)-F12b, CCSD(T), CCSD, and DCSD for the same systems, with MA%D of 0.9% (SEFB) and 0.4% (CSFB) against CCSD(T)-F12b/VDZ-F12. M06-2X is recommended for closed-shell halogen bonding with heavier halogens but has never been assessed for open-shell F-bonding. M06-2X/AVTZ systematically overbinds all SEFB-15 complexes and underestimates their F-bond distances, with MA%Ds of 75.8% on BEs and 8.9% on F-bond distances. M06-2X predicts comparable BEs for F$_2\cdots$CH$_3$ and F$_2\cdots$NH$_3$, contrary to all ab initio results, an anomaly persisting across all 16 basis sets. Neither basis set nor BSSE correction resolves the overbinding. BSSE correction worsens BEs for DF$\cdots$NH$_3$ complexes. For CSFB-14, M06-2X/AVTZ yields MA%Ds of 16.0% for BEs and 3.3% for F-bond distances. M06-2X is reliable for CSFB but unreliable for SEFB regardless of basis set. Companion paper II reports performance of other Minnesota functionals and ab initio methods for these datasets.