Heterobimetallic Complexes That Point to When Bond Dissociation Energies Deviate from Computational Expectations

Measurement of the formal, gas-phase, d8-d10 bond dissociation energy across a series of structurally homologous heterobimetallic complexes of Pd(II) with Cu(I), Ag(I), Au(I), and Zn(II), themselves models for the transition states for transmetalation in Sonogashira and Negishi couplings, finds large discrepancies relative to predictions by a commonly used dispersion-corrected density-functional theory method, DFT-D3(BJ), but not in all cases. Control studies on the threshold collision-induced dissociation (T-CID) of electrosprayed molecular ions, as well as the deconvolution of the bond energy from the experimentally measured energy-resolved cross sections, indicate that the experimentally determined bond dissociation energies are most likely correct, which raises the question of why the computational methods, while sometimes agreeing acceptably with experiment, can also sometimes disagree egregiously. While initial attempts to characterize the discrepancy focused on the metal-metal interaction, the most likely origin of the discrepancy appears to be an uneven treatment of nonbonded interactions, among them medium-ranged correlation effects and London dispersion, between the ligands on the two metal centers. The contribution of these effects to the formal bond dissociation energy is large enough to be chemically significant, but it appears to depend on the nature of the interacting groups, specifically the hybridization at carbon, and, more importantly, their relative orientation. Whereas face-to-face aryl-aryl interactions seem to be modeled well by PBE-D3(BJ), a representative DFT-D3 method, alkyl-aryl, and edge-to-face aryl-aryl interactions appear to be overestimated. The consequences for structure and stability in organic and organometallic molecules are discussed, especially with regard to relative energies of conformers and interconverting valence isomers.