## Introduction

Transition metal complexes bearing terminal oxido (O2−, oxo) ligands are quite common. Nevertheless, those of the late transition metals are very rare1,2,3,4,5,6,7,8, and in particular group 11 terminal oxo complexes (M = Cu, Ag, Au) are to our knowledge still elusive1,9,10,11,12. On the other side molecular binary group 11 metal oxygen complexes, MOx, have extensively been studied in gas phase and matrix-isolation studies13,14. Given the well-known activity of silver15 and of molecular gold complexes9,16 as catalysts in oxygen transfer reactions, there is interest in such molecular group 11 terminal oxo complexes as prospective catalysts in oxidation reactions17,18,19,20. Copper complexes bearing a terminal oxo ligand were frequently considered as highly reactive intermediates in biological systems to catalyze hydrogen atom transfer and homogeneous oxygenation reactions11,12. However, despite experimental and theoretical efforts such proposed group 11 terminal oxo species have never been observed experimentally.

Moreover, computational studies and gas phase experiments revealed an intriguing electronic structure of group 11 metal-oxo complexes11,17,18,19,20,21. The diatomic group 11 monoxide cations [MO]+ showed a diradical character with two unpaired electrons to reside in antibonding π* orbitals of predominant oxygen 2p character. For [CuO]+ this results in a spin density of 1.68 on O and 0.38 on Cu20,21. For [AuO]+ a higher spin density of 1.92 at oxygen and of only 0.08 at gold indicates a smaller π back-donation from the d orbitals of Au19. Triplet ground states have also been predicted quantum mechanically for ligated [LMO]+ species in various ligand environments (M = Cu11,17,18,19 and Ag10). They are described as M(II) oxido species bearing oxygen centered radicals (O•−)10,17,18,19. Consistent with their biradical character these species are found to be highly potent oxidants in conjunction with a very low M-O bond energy. Both, the low M–O bond energy and the high reactivity may have as yet prevented such complexes to be studied in condensed phases.

Experimental data about molecular oxo complexes of the group 11 metals are extremely rare and simple molecular oxo compounds such as LnMO (M = Cu, Ag, Au, L = any ligand except oxygen) are completely unknown.

Here we describe for the first time the synthesis of molecular oxygen fluorides of the coinage metals (L = F). They are characterized by IR matrix-isolation spectroscopy at cryogenic conditions supported by state-of-the-art quantum-chemical calculations. This study gives insight into the nature of the unusual group 11 metal-oxygen bond and proves the radical-like character of the oxo ligands in these complexes. Because of the non-innocent character of the terminal oxo ligand, assigning metal oxidation states in these complexes becomes ambiguous22,23,24.

## Results

### Investigation of hypofluorites

The novel group 11 metal oxygen fluorides have been prepared by laser ablation of the metals in the presence of OF2. Laser-ablated metal atoms were allowed to react with either 16OF2 or 18OF2 seeded in an excess of argon or neon and the mixture was co-deposited onto a rhodium-plated mirror cooled to 5–15 K (Supplementary Methods). It is expected that the metal atoms insert into the O−F bond of OF2 to yield the primary product FMOF, Equation (1).

$${\mathrm{M}} ^\ast + {\mathrm{OF}}_2 \to {\mathrm{FMOF}}\,\left( {{\mathrm{M}} = {\mathrm{Cu}},\,{\mathrm{Ag}},\,{\mathrm{Au}}} \right)$$
(1)

The hypofluorites FMOF were indeed detected for all coinage metals, see Figs. 1 and 2 (Supplementary Figures 111). Based on their characteristic 16/18O isotope shift we unambiguously assigned several stretching bands of these novel species. A compilation of the experimentally observed band positions of oxygen metal fluorides is given in Table 1 together with computed values at the CCSD(T)/aug-cc-pVTZ(-PP) level of theory. For the Cu and Ag species we also observed an additional 63Cu/65Cu (69:31) and 107Ag/109Ag (52:48), respectively, isotope splitting according to the natural isotope abundances of these elements. Gold is a neat element and the IR spectra do not show any Au isotope splitting.

As can be judged from the data in Table 1 the prediction of the band positions of the novel species and even of their 16/18O isotope shift is a challenge even for the high CCSD(T) level used therein. In agreement with the CCSD(T) predictions only the Cu−F stretching band of FCuOF has sufficiently high intensity to be observed experimentally. On the contrary, for FMOF, M = Ag and Au, the O−F stretch is predicted to be stronger than the F−M stretches. For both hypofluorites these two stretching bands are clearly assigned. While the O−F stretches show a characteristic Δν(16/18O) shift of about 23.5 cm–1 (Table 1), the band positions as well as the associated 16/18O shift of the F−Ag stretch in FAg16/18OF revealed an unusual matrix dependence, as these values were found to be larger in solid Ar than in solid Ne (Table 1). This observation indicates a rather strong interaction of this species with the solid hosts. Furthermore, for the FAg16/18OF isotopologues also a strong M−O stretch is computed around 300 cm–1 at the CCSD(T) level (Table 1). However, our search for these additional bands in the FIR region failed (Supplementary Figures 6 and 8).

### Investigation of OMF2

We noticed a strong dependence of the intensity of the hypofluorite bands on experimental conditions, such as the choice of the matrix gas, its OF2 concentration, and the amount of the laser energy used for metal ablation. The FMOF species are found to be highly photo-sensitive. By irradiation of the deposits using visible light of λ > 500 nm they rapidly photo-rearrange to yield the monooxygen fluorides OMF2, Equation (2) (Supplementary Figure 25, 7, 9, 11).

$${\mathrm{FMOF}}\mathop{\longrightarrow}\limits^{{{{h\nu}}}}_{{\lambda > 500\,{\mathrm{nm}}}}{\mathrm{OMF}}_2\quad \left( {{\mathrm{M = Cu,}}\,{\mathrm{Ag,}}\,{\mathrm{Au}}} \right)$$
(2)

### Investigation of binary metal fluorides

Bands associated with the known binary metal fluorides MFn (M = Cu, Ag, Au)27,28 are safely assigned in oxygen-free experiments, in which OF2 was replaced by gaseous fluorine. In these experiments none of the bands assigned to an oxygen-containing species appeared. Our band positions for the difluorides MF2, AuF3 and of the rare gas Ng-MF (Ng = Ne, Ar) adducts are listed in Supplementary Tables 13. They agree well with previously reported experimental and calculated ones27,28. We note that the Ne−CuF complex was previously not detected28. Our present results show a weak band at 625.3/623.4 cm–1 with a corresponding isotope splitting for a 63Cu/65Cu species, which we tentatively assign to the elusive Ne−CuF complex (Supplementary Figure 1, 10). The gas-phase value for the 63CuF IR band was reported at 614.8 cm–129, which is close to a computed value of 613.2 cm–1 at the CCSD(T)/aug-cc-pVTZ level28.

### Quantum-chemical calculations

Both, the hypofluorites FOMF (Cs – X 2A″) and the oxygen fluorides OMF2 (C2v – X 2B2, Fig. 3) have planar structures, and also for the monomolecular FOMF to OMF2 rearrangement a planar low-energy transition state has been localized at the UB3LYP-D3BJ/def2-TZVPP level of theory (Supplementary Table 5). The observed fluorine migration from FOMF to OMF2 is slightly exothermic for M = Au and Cu by –11.4 and –25.0 kJ mol−1, respectively, at the CCSD(T)/augcc-pVTZ level (Supplementary Table 4). For the silver compound FOAgF we computed an endothermic rearrangement to OAgF2 by +15.7 kJ mol−1, which, however, was photochemically initialized in our experiments.

## Discussion

In accordance with the planar T-shape structures of the OMF2 species, featuring an equatorial mono-oxygen ligand and a near linear F−M−F unit (Fig. 3), only the asymmetric M−F stretching frequency was observed for these species (see Figs. 1 and 2 and Table 1). This mode revealed no observable 16/18O isotope shift in experiments using mixtures of 16OF2 and 18OF2. The close vicinity of these M−F frequencies to the known bands of the corresponding MF2 species27,28 (Supplementary Tables 13) indicates a similar M−F bonding situation. In agreement with the absence of an absorption for the O−M stretches in the recorded spectra a very low intensity is predicted for this mode (Table 1). Both observations indicate an unexpected low polarity of the O−M bond in the OMF2 species.

The IR spectra obtained for the gold species in solid Ne (Fig. 1a, b) show another pronounced absorption in the Au−F stretching region at 629.4 cm–1 directly after the deposition. This band exhibits a small but significant 16/18O shift of 1.4 cm–1 in experiments using isotopically enriched 18OF2 (Table 1), which points to the formation of another so far unknown oxygen containing gold fluoride. Surprisingly, the carrier of this band has not been detected in solid argon, but it is formed in addition to OAuF2 in solid neon matrices, for instance by the photo-decomposition of FAuOF using LED light of λ = 375 nm (Supplementary Figure 2c). On the other side, this band was destroyed by visible LED light of λ = 405–455 nm, while simultaneously some OAuF2 was formed (Supplementary Figure 1c). These observations suggest that the band at 629.4 cm–1 is due to a simple oxygen gold fluoride such as linear OAuF.

Our tentative assignment of the OAuF species is supported by coupled-cluster calculations. We have performed high-level CCSD(T) calculation for the 3Σ ground state of OAuF (C∞v) using large basis sets (cc-pVQZ, aug-cc-pVQZ)30,31. Anharmonicity corrected vibrational frequency calculations via Dunham analysis32,  or by the use of a vibrational self-consistent field procedure (VSCF)33 (Supplementary Methods). While these calculations predict only one pronounced absorption for 16OAuF associated with a strong Au–F stretching mode and a harmonic frequency around 642 cm–1, they consistently overestimate its 16/18O isotopic shift to about 5 cm–1 compared with the experimentally recorded value (1.4 cm–1, Table 1). Nevertheless, even this computed 16/18O isotopic shift for the Au–F stretching mode of linear OAuF is unusually small, compared to e.g., the linear OAuF unit in FOAuF. The weak coupling between the Au–F and the O–Au stretching vibrations in OAuF indicates an unusual Au–O bond character.

We finally have recorded FIR spectra in the range <500 cm–1 using a liquid He cooled bolometer. In the reaction of laser ablated silver atoms with a mixture of 16OF2 and 18OF2 (4:6) we observed two additional weak bands at 491.6 and 472.9 cm–1 (solid Ne, Supplementary Figure 6; solid Ar: 481.5 and 463.3 cm–1, Supplementary Figure 8) showing an unresolved 107/109Ag isotope splitting and a significant Δν(16/18O) shift of 18.7 (solid Ar: 18.2) cm–1. Both, positions and 16/18O isotope shift of these bands match very well the CCSD(T) predicted values for the antisymmetric stretching vibrations of the linear triplet 16/18OAgF isotopologues (Table 1). Our CCSD(T) calculations also predict a rather strong absorption for the lighter OCuF homolog at 837 cm–1 (Δν(16/18O) = 22 cm–1), which, however, has not been detected in the experimental spectra.

To get more insight into the nature of the M-O bond and the proposed radical-like character of the terminal oxygen ligand in the group 11 complexes we had a closer look on the molecular orbitals and the spin densities for the linear OAuF species (Supplementary Figures 12, 13 and Supplementary Note 1). The O–Au bond length in OAuF of 181.1 pm (Fig. 3), computed at the CCSD(T)/aug-cc-pVTZ(-PP) level of theory, is just in between the reported standard bond length for a single (187 pm)34 and double (178 pm)35 O–Au bond, while the computed Au–F bond length of 188.2 pm is close to the value expected for an Au–F single bond length of 188 pm34. The MO scheme of OAuF bears similarities with that of the valence isoelectronic and well studied [OAuO] anion14. Like the [OAuO] anion also OAuF shows an open-shell configuration (7σ2424422) with two unpaired electrons in the antibonding 4π* molecular orbitals (MO’s) (Supplementary Table 6 and Figures 12, 13). The key message arising from this scheme is that the oxygen (2px,2py) orbitals, which mainly contribute to the antibonding 4π* MO’s, are higher in energy than the Au 5d orbitals. This is a characteristic feature of a so called inverted ligand field36,37,38, for which the binary group 11 oxides such as the [OAuO] anion14 may serve as an additional example. The computed spin density of 1.34 at oxygen and of 0.53 at gold for triplet OAuF (Supplementary Figure 4) corresponds well to the singly occupied antibonding 4π* MO (Supplementary Figure 13). This bonding pattern is similar to that previously described for the diatomic triplet 3[MO]+ cations (M = Cu, Au), in which also two biradicaloid π* electrons are present17,18,19,20,21. However, while the Au–O bond in the triplet 3[AuO]+ cation has recently been described by a donor-acceptor model19, it is obvious that the Au-O bonding in triplet OAuF is strongly influenced by the bonding properties of the Au 5d orbitals (Supplementary Figures 12, and Supplementary Note 1).

Another difference between the Au–O bonds in triplet OAuF and the related 3[AuO]+ cation is an enhanced spin density at the gold atom (0.53 and 0.0819, respectively). There is a significant π back-donation from the Au 5d orbitals in OAuF (Fig. 4) but not for the 3[AuO]+ cation19. This likely can be attributed to the fully occupied F(2px, 2py) orbitals, which in an antibonding fashion mix into the Au(5dxz, 5dyz) orbitals (Supplementary Figure 13).

In the above described experiments the elusive OMF complexes (M = Au, Ag) are likely formed from free O atoms and MF molecules during deposition or within the solid matrices. These species were always present in the deposits. As it was shown by us, AuF forms considerable strong Ar complexes (ΔH = −47.3 kJ mol−1), but only weak complexes with neon (ΔH = −8.8 kJ mol−1)27,28. It is therefore reasonable that this reaction was observed using neon matrices by substitution of a weakly bound Ne atom in a NeAuF complex, but not in solid argon where the ArAuF complex was found to be one of the main reaction products (Fig. 2). The elusive OMF (M = Cu, Ag, Au) compounds bear also some similarities to the weakly-bound NeMF adducts, also recently studied by us28. While the NeCuF adduct (ΔH = −11.2 kJ mol−1)28 is predicted by CCSD(T) calculations to be even more stable than NeAuF, attempts to observe this elusive species failed until it could now tentatively be assigned in this work. On the other side, the NgAgF (Ng = Ne, Ar) show significantly smaller bonding energies ((ΔH = −3.7 (Ne) and −21.1 (Ar) kJ mol−1)28, but both, the neon and the oxygen complexes of AgF are detected experimentally. Like FOAgF, but unlike OAuF, the elusive OAgF complex is highly photosensitive and easily destroyed by irradiation of visible light (Supplementary Figures 6, 8).

Also in the 2B2 ground state of the OMF2 species the unpaired electron occupies an oxygen 2p orbital, and the computed spin-density is solely located at the oxygen atom (Supplementary Figure 14). There is no O−M π bond, which is also evident from the computed CCSD(T) O−M bond length indicated in Fig. 3. As pointed out by a referee we noticed that such a breakdown of the conventional Lewis electron–pair model is not a unique phenomenon of group 11 oxygen compounds. One closely related example is the recently investigated oxygen-centered OHgF radical39, but also other transition metal terminal oxo complexes bearing terminal oxygen-centered radicals have been predicted22,23,24,40. Terminal oxygen radicals have also been found to occur in metal oxide clusters of early transition metals, when these metals are in their highest oxidation state. Such oxide clusters were found to be crucial intermediates in catalytic oxidation reactions41,42.

We have shown that excited coinage metal atoms M (M = Au, Ag, Cu) react with OF2 to yield the hypofluorites FOMF, as well as the elusive group 11 oxygen fluorides OAgF, OAuF, and OMF2. The isolation of these novel species in rare-gas matrices allowed a direct spectroscopic investigation of group 11 compounds bearing a terminal oxygen ligand in a condensed phase. The new oxygen fluorides are likely formed by photolysis of the initially prepared hypofluorites. The linear molecules OAgF and OAuF have a 3Σ ground state with a biradical character. Two unpaired electrons are located mainly at the oxygen ligand in antibonding O–M π* orbitals. For the 2B2 ground state of the OMF2 compounds only an O–M single bond arises and a significant spin-density contribution was found at the oxygen atom as well. The radical character of the terminal oxygen ligand in these group 11 fluorides can be explained with an inverted ligand field, which previously has also been suggested for the related binary group 11 oxides14,19,21. It turns out that the investigation of these novel oxygen fluorides is a challenge for prospective experimental and high-level quantum chemical studies.

## Methods

### Experimental details

The experimental set-up used a closed-cycle helium cryostat (Sumitomo Heavy Industries, RDK-205D) inside a vacuum chamber (Supplementary Methods). The Nd:YAG laser fundamental with a pulse energy of up to 50 mJ/cm² was focused onto the metal targets. The OF2 and ablated metal atoms and their reaction products have been trapped in rare-gas matrices on a cold rhodium-plated mirror. 16/18OF2 was synthesized by a known procedure using elemental fluorine and 16/18OH2 dispersed in solid KF. FTIR spectra were recorded on a Bruker Vertex 80 v spectrometer at a 0.5 cm−1 resolution and with a 0.5 cm−1 accuracy using an MCTB detector (Supplementary Methods). For quantum-chemical details see also Supplementary Methods, Supplementary Table 7 and Supplementary Data 1.

### Data availability

Data will be available on request from the authors.