1H NMR is not a proof of hydrogen bonds in transition metal complexes.

M. A. Bakar et al. Nature Commun. 10.1038/s41467-017-00720-3 (2017)

class="Chemical">Hydrogen boclass="Chemical">ndiclass="Chemical">ng to gold(I) aclass="Chemical">nd its effect oclass="Chemical">n the structure aclass="Chemical">nd dyclass="Chemical">namics of molecules have beeclass="Chemical">n a matter of loclass="Chemical">ng debate.[1] A class="Chemical">number of X-ray studies have reported gold compouclass="Chemical">nds with short class="Chemical">n class="Chemical">AuI···H contacts, but solid spectroscopic evidence for AuI···H bonding has been missing.[1] Notably, during the revision of this work, Bourissou et al.[2] and Straka et al.[3] have provided evidence of true intramolecular AuI···H hydrogen bonds in [Cl–Au–L]+ complexes, where L is a protonated N-heterocyclic carbene. The studied compounds feature intramolecular AuI···H+–N bonds detected by means of NMR[2] and infrared spectroscopies.[2,3]

Previously in this Journal, Bakar et al.[4] reported compound 1 (Fig. 1a) with four short class="Chemical">Au···H coclass="Chemical">ntacts (2.61–2.66 Å X-ray determiclass="Chemical">ned). Assumiclass="Chemical">ng the ceclass="Chemical">ntral cluster iclass="Chemical">n 1 to be [class="Chemical">n class="Chemical">Au6]2+ and observing the 1H (13C) NMR resonances at respective H(C) nuclei in 1 highly deshielded with respect to precursor 2 (Fig. 1b), the authors concluded that “the present Au···H–C interaction is a kind of hydrogen bond”, where the [Au6]2+ serves as an acceptor”.

Fig. 1

Schematic structures of 1 and 2, and selected ETS-NOCV channels in 1. a The schematic structure of 1 with indicated differences in calculated charges and experimental 1H NMR shifts[4] (in brackets) between 1 and 2 for selected hydrogens. b Precursor 2. ETS-NOCV channels in 1′ corresponding to c Au2···H2–C2 interaction, and d side-on Au3···H2–C2 interaction. Large P(Ph)2 sidechains are omitted for clarity. Cutoff of 0.0002 is used in c and d

Schematic structures of 1 and 2, and selected ETS-class="Chemical">NOCV chaclass="Chemical">nclass="Chemical">nels iclass="Chemical">n 1. a The schematic structure of 1 with iclass="Chemical">ndicated differeclass="Chemical">nces iclass="Chemical">n calculated charges aclass="Chemical">nd experimeclass="Chemical">ntal class="Chemical">n class="Chemical">1H NMR shifts[4] (in brackets) between 1 and 2 for selected hydrogens. b Precursor 2. ETS-NOCV channels in 1′ corresponding to c Au2···H2–C2 interaction, and d side-on Au3···H2–C2 interaction. Large P(Ph)2 sidechains are omitted for clarity. Cutoff of 0.0002 is used in c and d

Here, we show that the class="Chemical">Au6 cluster iclass="Chemical">n 1 bears class="Chemical">negative charge aclass="Chemical">nd the class="Chemical">n class="Chemical">Au···H contacts lead to only a rather weak (~1 kcal mol−1) auride-like···hydrogen bonding interaction. In addition, computational analysis of NMR chemical shifts reveals that the deshielding effects at respective hydrogen nuclei are not directly related to Au···H–C hydrogen bonding in 1. It is well known that interactions of hydrogen with transition metals compounds may influence the 1H NMR shifts in unexpected ways.[5]

In the following, we analyze class="Chemical">Au2···class="Chemical">n class="Chemical">H2–C2 contact in 1, which is one of the four Au···H–C contacts in the molecule (Fig. 1a). Computational methodology is described in Supplementary Information. The calculated C–H distances (1.08 Å) in 1 are about 0.15 Å longer than those derived from the X-ray structure (0.95 Å) as proposed in ref. [4]. The calculated minimum Au···H distances (2.61–2.62 Å) are in excellent agreement with the reported ones (2.61–2.65 Å). To afford computational analysis of NMR chemical shifts[7-9] (δ), the P(Ph)2 groups in 1 and 2 were replaced by P(CH3)2 groups in model systems 1′ and 2′ (Fig. 1). Such changes are known to have minimum impact on δ(1H).[8] Notably, the absence of the bulky P(Ph)2 ligands causes rotation of the central phenyl groups away from the Au6 cluster. The Au···H distances increase from 2.6 Å to 2.77 Å and the Au–H2–C2 angles bend from 167° to 144° (Supplementary Fig. 1). The short Au···H–C contacts in 1 are thus likely enforced indirectly by sterically demanding P(Ph)2 ligands that potentially stabilize the whole cluster via dispersion interactions among themselves.[10] To avoid these undesirable changes in calculations, we fixed the core of 1 and 2 in optimization of 1′ and 2′ and only methyl groups were optimized.

Quantum theory of atoms in molecules (QTAIM) analysis of 1′ shows a low ED (0.016 e.bohr−3) with positive Laplacian (0.037 e.bohr−5) at the line critical point (LCP) of class="Chemical">Au2···class="Chemical">n class="Chemical">H2 interaction. These values are less than a half of those for reported Au···H+–N bonds.[2,3,6] Small electron exchange between Au2 and H2 of 0.07 e (e = electron) is consistent with a dispersive interaction.[11,12] The direction of the charge transfer in Au2···H2 interaction is from Au2 to H2 (0.04 e, Supplementary Table 2) similar to auride···hydrogen weak interaction.[1] All four Au atoms in contact with phenylene H2 atoms (Au2–Au5) in 1′ have negative charge, about −0.15 e (using QTAIM, Supplementary Table 3). Thus, the Au6 cluster, although formally a di-cation, strongly pulls the electron density from the ligands in 1.

Extended transition state-natural orbitals for chemical valence (ETS-class="Chemical">NOCV)[13] aclass="Chemical">nalysis of 1′ reveals a weak class="Chemical">n class="Chemical">Au2···H2–C2 interaction channel (0.9 kcal mol–1, Fig. 1c), which is about 10 times less than for recently reported Au···H+–N bonds.[2,3] Smaller, weakly stabilizing “side-on” interaction (0.4 kcal mol–1) is found between H2–C2 and Au3 6p orbitals (Fig. 1d). Notably, the Au2···H2–C2 channel (Fig. 1c) is also found in the fully relaxed structure of 1′ (0.8 kcal mol–1). This further points to a minimal stabilization effect of Au···H–C bonding in 1.

The calculated differences in class="Chemical">1H chemical shifts betweeclass="Chemical">n 1 aclass="Chemical">nd 2 (1′ aclass="Chemical">nd 2′) are iclass="Chemical">n excelleclass="Chemical">nt agreemeclass="Chemical">nt with the experimeclass="Chemical">ntal oclass="Chemical">nes, for class="Chemical">n class="Chemical">H2 Δ = 4.4 ppm, Δ = 3.6 ppm, and Δexp = 4.4 ppm (absolute values are shown in Supplementary Table 1). The analysis of the NMR chemical shifts affordable only for 1’ and 2’ reveals that 1.6 of 3.6 ppm of calculated Δ(H2) arises from the diamagnetic part of the NMR chemical shift, Δδdia, which can be rationalized only by a depletion of electron density (ED) at the H2 nuclei. Molecular orbital (MO) analysis of Δδdia identifies that main part Δδdia (1.3 of 1.6 ppm) originates from four Au–P π-back-bonding MOs (Supplementary Fig. 2). No through-space interactions between Au2 and H2 can be seen in these MOs. Quite to the contrary, the HOMO and HOMO-1 have Au2–H2 antibonding character (Supplementary Fig. 2). Despite the fact that Au2 shares some ED with H2 (see above), the overall ED at the H2 decreases by 0.010 e bohr−3 from 2′ to 1′. Notably, ED increases at H4 and H6. Changes in ED are reflected in NMR chemical shifts, as deshielding is observed at H2 and shielding is observed at H4 and H6. This interpretation is supported by the calculated differences of atomic charges (Supplementary Tables 3, 4), which correlate well with the reported experimental differences of δ(1H) at H2, H4, and H6 nuclei[4] given in brackets in Fig. 1a.

The paramagnetic part of the Δ deshielding difference at class="Chemical">H2 class="Chemical">nuclei (2 ppm) is domiclass="Chemical">nated by local Ramsey-type paramagclass="Chemical">netic coupliclass="Chemical">ngs[9,14,15] betweeclass="Chemical">n class="Chemical">n class="Chemical">H2–C2 σ-bond (HOMO-1 of 1′ in Fig. 2) and vacant MOs* formed by π* C2 2py orbitals (e.g., LUMO + 5 of 1′ in Fig. 2). Mixing of Au3 6p* and 5dz2* atomic orbitals (AOs) with C2 2py* in 1′, which is not possible in 2′, increases the MO ↔ MO* overlap in orbital magnetic couplings.[9,15,16] This leads to the ~0.5 ppm larger paramagnetic deshielding at H2 in 1′ in this particular coupling (Fig. 2). Notably, this coupling strongly resembles the dispersive side-on Au3···H2–C2 NOCV interaction channel discussed above (Fig. 1d). Overall, HOMO-1 in 1′ is responsible for ~2 ppm of paramagnetic deshielding at H2, while similar Au–P π-back-bonding orbital (HOMO-1) in 2′ contributes only by 0.2 ppm. An analogous mechanism is likely to be responsible also for the deshielding resonance at C2.

Fig. 2

An example of Ramsey-type MO ↔ MO* coupling in 1′. Orbitals are cut along Au2–H2–Au3 plane

An example of Ramsey-type MO ↔ MO* coupling in 1′. Orbitals are cut along class="Chemical">Au2–class="Chemical">n class="Chemical">H2–Au3 plane

We conclude that the short class="Chemical">Au···H coclass="Chemical">ntacts iclass="Chemical">n 1 are aclass="Chemical">n example of a weak (~1 kcal/mol per coclass="Chemical">ntact) class="Chemical">n class="Chemical">auride-like···hydrogen interaction, with small overall (~4 kcal/mol) stabilizing effect on the cluster structure. Instead, the stabilizing effect can be attributed to the dispersion interactions among the P(Ph)2 groups, as documented previously.[10] Distinct δ(1H) NMR deshielding of C–H groups in contact with Au6 cluster in 1 as compared with the precursor 2 is due to (a) the differential ED at the H2 atom in 1 as compared to precursor 2, and (b) side-on orbital interactions between nearby Au3 atom and H–C MOs that increase the efficiency of the local Ramsey-type deshielding paramagnetic couplings in molecule 1 as compared with corresponding couplings in precursor 2.