## Introduction

Carbon monoxide is amongst the most important ligands in transition metal chemistry. Thus, carbonyl complexes have been continuously and extensively studied over the past 130 years after the discovery of the first homoleptic carbonyl complex Ni(CO)4 by Mond in 18901. Their beauty lies in their simplicity and their rich substitution and redox chemistry, which leads to a wide range of applications, spanning from catalysis to biochemistry and medicine2,3,4,5,6,7. Most condensed phase homoleptic transition metal carbonyl complexes and all neutral mononuclear homoleptic transition metal carbonyl complexes in particular, obey the 18-electron rule7; the only exception is V(CO)6 as a 17 valence electron (VE) species. Neutral complexes can be reduced to gain anionic carbonyl metallates, such as [V(CO)6], [Fe(CO)4]2− (discovered as early as 1930) or even clustered radical anions such as [Fe3(CO)11]•−8,9. Transition metal carbonyl cations (TMCCs), however, could not be accessed until about 1960, when the octahedral carbonyl cation [Mn(CO)6]+ was discovered10. Since the 1990s, mainly superacidic media enabled the synthesis, isolation, and characterization of several homoleptic TMCCs, such as [Au(CO)2]+11,12,13, [Fe(CO)6]2+14, [Co(CO)5]+15 or even superelectrophilic [Pd(CO)4]2+16, or [Ir(CO)6]3+17. Despite the fact that all these complexes were prepared by an oxidative approach, all of them feature an even number of electrons. No open-shell TMCC with mononuclear metalloradical configuration18 has been synthesized and structurally characterized to date (cf. Table 1)19,20,21,22,23,24,25,26,27,28,29. Typically, those would evade the open-shell situation by forming metal–metal-bonded dimers, like seen for the reliably on spectroscopic data assigned dimeric platinum carbonyl cation [(OC)3Pt–Pt(CO)3]2+23.

However, the chromium hexacarbonyl cation, as a prototype example for such open-shell TMCC systems, was the subject of several electrochemical investigations30,31,32,33, as well as gas phase and theoretical studies34,35,36. In principle, these rather weakly bound TMCCs should be ideal starting materials for further substitution chemistry. Yet, the established routes to TMCCs via superacidic media require specialized equipment/laboratories and superacids are often incompatible with commonly used ligands. This combination limited hitherto the possibilities for follow-up chemistry. Quite contrary, the approach delineated here uses the weakly coordinating anions (WCAs) [Al(ORF)4] and [F-{Al(ORF)3}2] (RF = C(CF3)3)) that allow the use of regular organic solvents, as well as standard laboratory equipment. In combination, this approach facilitates synthesis of the target TMCC salts—if desired, also in multigram-scale—and enables a rich follow-up chemistry.

## Results and Discussion

### Synthesis

A suitable oxidant (strong and innocent) is required to prepare the target salts [Cr(CO)6]•+[WCA]. Whilst Ag+ (E° = +0.65 V vs. Fc/Fc+ in CH2Cl2)37 deelectronates [Co2(CO)8] in CO atmosphere generating [Co(CO)5]+ at room temperature (r.t.) in ortho-difluorobenzene (oDFB)38, its redox potential is too low to oxidize Fe(CO)5. This leads to the formation of the peculiar metal-only Lewis pair [Ag{Fe(CO)5}2]+39,40. [NO]+ might be a strong enough oxidant (cf.: E° = +1.00 V vs. Fc/Fc+ in CH2Cl2)37, but it is prone to an inner-sphere electron transfer, which then often leads to a coordination of the resulting NO to the metal center37,41. This is also reflected in the reaction of [Co2(CO)8] with [NO]+ that yielded an inseparable mixture of [Co(CO)5]+ and [Co(CO)2(NO)2]+42. However, the octahedral configuration of Cr(CO)6 should be kinetically hindered towards substitution due to an energetically disfavored mechanism43. This feature enabled using [NO]+ as oxidant in combination with the WCAs [Al(ORF)4] and [F-{Al(ORF)3}2]. In order to minimize the inevitable NO/CO-ligand exchange, the 1:1 deelectronation reaction of Cr(CO)6 with [NO]+[WCA] (WCA: [Al(ORF)4] or [F-{Al(ORF)3}2]) was carried out at low temperatures in CH2Cl2 (−78 °C to r.t.). Application of a dynamic vacuum over 1 h in order to immediately remove the released NO, furnished near quantitative yields of the 17 VE species [Cr(CO)6]•+[Al(ORF)4] (1) or [Cr(CO)6]•+[F-{Al(ORF)3}2] (2) as off-white to pale-yellow solids (Eq. 1). By allowing the reaction to proceed for 7–14 days in a closed vessel at r.t. (Eq. 2), the heteroleptic 18 VE complexes [Cr(CO)5(NO)]+[Al(ORF)4] (3) and [Cr(CO)5(NO)]+[F-{Al(ORF)3}2] (4) were obtained in near quantitative yield as orange solids. Here, the liberated NO displaced one CO ligand and acts as a 3e donor (Fig. 1).

The success of the reaction was monitored by gas-phase IR spectroscopy. The loss of either CO or NO was observed in the reactions providing 3 and 4 or 1 and 2, respectively (see S.I. section 8, Supplementary Figures 35 and 36). The purity and phase homogeneity of 1–4 were confirmed by IR and Raman spectroscopy, as well as by phase analysis with powder X-ray diffraction measurements (pXRD, see S.I. section 13, Supplementary Figures 5158). All complexes are stable at room temperature and can be stored for months under argon atmosphere. In addition, magnetic measurements using the Evans method44 in oDFB/CH2Cl2 solutions furnished the expected magnetic moments of 2.04μB for 1 and 2.06μB for 2, respectively (see S.I. section 6, Supplementary Table 1, Supplementary Figures 2732). Hence, apart from being a powerful method for the synthesis of TMCCs, Eq. 2 also opened the door to the family of heteroleptic chromium carbonyl/nitrosyl cations. To put this into perspective, [Cr(CO)5(NO)]+ has only been subject to a few theoretical studies yet45,46 and structurally authenticated examples are only known for anionic ([Cr2(CO)9(NO)], [Cr(CO)4(NO)])47,48,49 or neutral (Cr2(CO)8(NO)2)47 representatives.

### Kinetic vs. thermodynamic product

From a series of reactions performed at intermediate temperatures and monitored at various times, it appeared that the oxidation and formation of 1 and 2 is the kinetically preferred reaction pathway, whereas the nitrosyl complexes 3 and 4 are the thermodynamic final products. Full ab initio calculations at the CCSD(T)/TZ→QZ level (see S.I. section 16, Supplementary Figure 75) support the claim that [Cr(CO)6]•+ is indeed the kinetic product in CH2Cl2 as solvent (ΔrH°gasrG°solv = −83/−30 kJ mol−1 for [Cr(CO)6]•+ (Eq. 1) vs. −232/−173 kJ mol−1 for [Cr(CO)5NO]+, Eq. 2). The high reactivity of [NO]+ and rather poor solubility of 1 and 2 in CH2Cl2 seem to be crucial for the success of the reaction (Eq. 1), since it leads to a quick formation of the relatively stable [Cr(CO)6]•+ intermediate that shortly after precipitates at low temperatures. By contrast, carrying out the reaction in oDFB, where all the products are well dissolved, apparently led to a less reactive [NO]+ and therefore only the isolation of [Cr(CO)5(NO)]+. We note that the reaction of the obviously formed charge-transfer π−complex [NO+-oDFB] with Cr(CO)6 must be faster than the formation of a Wheland-intermediate and nitrosation of the solvent. However, once pure 3 and 4 were prepared, their crystals dissolve unchanged in oDFB.

### Vibrational spectroscopy

The solid state ATR IR spectra (Fig. 2 and Table 2; ATR = attenuated total reflection) of 1 and 2 show a single broadened CO band at 2094 cm−1 (small 13CO shoulder at about 2070 cm−1), which verifies the assumptions made of a possible [Cr(CO)6]•+ in a previous matrix study50. The respective Raman spectra show a sharp band at about 2174 cm−1 for the all-symmetric stretch vibration, as well as a broad band centered around 2127 cm−1. The position and shape of these bands are in agreement with the Raman spectrum of the isoelectronic but neutral D3d-symmetric V(CO)6 (cf. Table 2 and Supplementary Figure 34)51.

A Jahn–Teller-induced fluxionality at room temperature leads to the broad band at about 2127 cm−1 not affecting the all-symmetric stretching mode. This fluxionality freezes out in the vanadium case only at temperatures below 16 K. At higher temperatures, a very low-lying transition state probably allows for equilibration—even on the fast time scale of vibrational spectroscopy, leading to the discrepancy between experimental and the D3d-simulated spectrum. This assignment is in agreement with the EPR spectra as well as DFT and ab initio calculations for the cation (vide infra, Table 3). The presence of a band of very low intensity at about 1840 cm−1 is indicative for the formation of very small amounts of the (by-)products [Cr(CO)5(NO)]+ 3 and 4. Furthermore, as opposed to 1 and 2, some Cr–N and Cr–C vibrations at 658/640 cm−1 (IR) and 640/487 cm−1 (Raman) are not obscured by anion bands and can be assigned unambiguously (cf. Supplementary Figure 33). Their entire vibrational spectra are in good agreement with the simulation (Fig. 2), as well as the known bands of the isoelectronic neutral V(CO)5(NO)52 (Table 2).

### Molecular structures

Quite unsurprisingly and due to their very similar molecular structure, the pale yellow complexes 1, 2 and the orange colored complexes 3, 4 are isostructural. The high symmetry of the crystal structures of 1 and 3 (tetragonal, P4/n) leads to only three symmetry-independent ligands, rendering a differentiation between the NO or CO positions impossible. Even more so for 2 and 4, where the cubic structures ($$Pa\bar 3$$) include only one symmetry-independent ligand (Supplementary Figure 68 and Supplementary Figure 71). Yet, spectroscopically clean pale yellow colored (orange colored) crystals suitable for single crystal XRD (scXRD), in agreement with an assignment as 1 and 2 (3 and 4), were isolated from batches of reactions according to Eq. 1 (Eq. 2) (Fig. 3).

The observed average Cr–ligand bond lengths are about 3 pm shorter for 3 and 4. This is consistent with the shorter (calculated) Cr–N bond length (Table 3) and is in addition reflected in smaller unit cell volumes of 3 and 4, if compared to 1 and 2. In addition, this also holds for the bulk microcrystalline powders of all compounds 14, as shown by full Rietveld-refinement of powder XRD data collected at 100 K (see S.I. section 13, Supplementary Table 7). Overall, the determined experimental parameters and the results of DFT calculations collected in Table 3 are in good agreement and in accordance with the reported bond lengths for Cr(CO)653 and V(CO)654.

### EPR spectroscopy

To verify the identity of the [Cr(CO)6]•+ radical cation, a diluted frozen solution of 1 in a 1:1 (v:v) mixture of oDFB and toluene (pure oDFB solutions led to very noisy spectra) was investigated by means of X-band EPR spectroscopy at 100 K (Fig. 2c). The paramagnetic 17 VE d5 species gives rise to one single EPR-line corresponding to an isotropic g-value of g = 2.003 being close to the g-factor of the free electron (gel = 2.0023)55. Similar g-values were reported for tetra-, tri-, and dicarbonyl chromium(I) phosphine and phosphite complexes56. In contrast to prior EPR investigations from electrochemical oxidations33, we did not observe any signal splitting due to hyperfine coupling to the 53Cr nucleus (I = 3/2, natural abundance: 9.55%). Probably any coupling pattern is hidden by the line broadening in the EPR spectra, since our calculations of hyperfine couplings (HFCs) reveal isotropic coupling constants always lower than 30 MHz (Supplementary Table 5). Note, however, that a larger hyperfine coupling of 61 MHz was reported for a species assigned as [Cr(CO)6]•+ and generated by in situ electrolysis of Cr(CO)6 in saturated solutions of [NBu4]+[ClO4] in CH2Cl2 at −80 °C33. This 53Cr hyperfine coupling is about three times larger than the coupling constants of chromium(I) carbonyl phosphine and phosphite complexes56. Furthermore, the g-value of g = 2.003 we identified for 1 is larger than the one previously reported (g = 1.982)33. We note, however, that the presence of the—compared to our WCAs—much stronger coordinating perchlorate anion in the experimental setup may have led to ion-pair formation, thus shifting the g-value with respect to the naked [Cr(CO)6]•+ cation. Our isotropic g-value obtained at 100 K would be consistent with an octahedral symmetry Oh of the [Cr(CO)6]•+ cation. However, the d5 complex should be subject to a dynamical Jahn–Teller effect, lowering its symmetry as it was found in the crystal structure of 1 and 2 and that of the isoelectronic V(CO)6, which has D3d symmetry in its electronic ground state51 (see S.I. section 9 for an in-depth discussion). Consequently, the pseudo-isotropic signal at 100 K may result from a thermal fluctuation of different Jahn–Teller distorted [Cr(CO)6]•+ structures (e.g., the D3dD2hD3d sequence shown in Fig. 2f and also suggested for V(CO)651). To verify this assumption, we conducted EPR measurements of 1 at 4 K. At this temperature, the spectrum indicates axial anisotropy with two principle g-values of $$g_\parallel = 1.953$$ and $$g_ \bot = 2.156$$, being consistent with an axial D3d-symmetry of the [Cr(CO)6]•+ cation (Fig. 2d).

### Ab initio and DFT investigations

To complete the answer to the question of the correct point group of the [Cr(CO)6]•+ cation, we performed symmetry restricted structure optimizations at the TPSSh/def2-TZVPP level of theory. In a comparative study, we found the TPSSh functional to yield reasonable structural parameters (Supplementary Table 4). The so-obtained structures were used for DLPNO-CCSD(T) single-point electronic energy calculations. In addition, we calculated g-tensors for minimum structures by means of the NEVPT2-SA-CAS-SCF method. The results are summarized in Table 4.

In accordance with an extensive study51 of the isoelectronic V(CO)6 complex, we found the D3d structure to be the lowest in energy. The D2h structure is a transition state connecting two equivalent D3d minima. Alike, the Oh symmetric structure is not a stationary point and convergence in the SCF procedure can only be achieved by allowing (unphysical) broken occupation numbers in the unrestricted treatment. Similar to V(CO)657, we identified a C2h symmetric structure, which shows only a very small structural and energetic deviation from the D3d one. We note that the calculated minimal relative energy of the D2h transition state (+0.85 kJ mol−1) is in perfect agreement with the EPR observation of a fluxional structure at 100 K. Thermal energy corresponds to 0.83 kJ mol−1 at 100 K. Only at considerably lower temperature (4 K) the structure is static. The D4h symmetric structure is a local minimum notably higher in energy and seems only to be accessible on the potential energy surface by passing the D2h transition state to higher energies. Furthermore, the CAS-SCF calculations yield a much more anisotropic g-tensor for the D4h structure than for the D3d one. Hence, even though there is some deviation, the comparison of calculated and experimental g-values clearly confirms the [Cr(CO)6]•+ cation to have a D3d symmetric ground state structure. Thus, the symmetry taken by the [Cr(CO)6]•+ cation in the crystal structure—best realized in the cubic structure 2 with an isolated Cr–CO moiety residing on a $$\bar 3$$-position of local D3d site symmetry—is really rooted in its electronic structure and not caused by interactions with the anion or crystal packing effects. This is in agreement with our claim that the WCA [F-{Al(ORF)3}2] is one of the least coordinating anions known58 and as such truly introduces pseudo-gas phase conditions in the solid state59.

## Discussion

The right combination of [NO]+ as oxidant with the WCAs [Al(ORF)4] and [F–{Al(ORF)3}2] and the unique reactivity of Cr(CO)6 gives facile access to the unexplored field of homoleptic 17 VE radical carbonyl cations, as well as heteroleptic 18 VE chromium carbonyl/nitrosyl cations. [Cr(CO)6]•+[WCA] shows surprising stability and is closely related to the isoelectronic and isostructural V(CO)6 with the same D3d symmetric ground state, which is supported by experimental IR, Raman, and EPR spectroscopic investigations, as well as magnetic measurements and calculated DFT/full ab initio data. Apart from its textbook importance, [Cr(CO)6]•+ might be a valuable 17 VE metalloradical precursor for substitution chemistry, which is not part of this report. Especially, follow-up chemistry and take up by others is facilitated by the access to 1 and 2 using standard glassware, organic solvents, and commercially available starting materials. We expect transfer of this methodology to other transition metal carbonyls. Of those, the group 6 metal carbonyls molybdenum and tungsten are currently in our focus.

## Methods

### General

Full details of the employed methods and additional information are given in Supplementary Information: pictures of the glassware used and the compounds themselves (Supplementary Figures 14); synthesis and spectra of NO[F-{Al(ORF)3}2] (Supplementary Figures 58); EPR spectroscopy (Supplementary Figure 37, Supplementary Tables 4 and 5, Supplementary Data 1); calculated geometries (Supplementary Figures 76 and 77, Supplementary Data 811), simulated spectra (Supplementary Figures 3841) and calculated energies (Supplementary Figure 75, Supplementary Tables 10 and 11); UV/Vis spectra (Supplementary Figure 50); crystal structures (Supplementary Figures 6772, Supplementary Data 27, Supplementary Tables 8 and 9) and Hirshfeld plots (Supplementary Figures 73 and 74). For space restrictions, here we only report the data of the [Al(ORF)4] salts. However, all experimental details for the [F-{Al(ORF)3}2] salts and their spectra are given as well (Supplementary Figures 1316, 22–26 (NMR), Supplementary Figures 44, 45, 48, 49 (IR/Raman)). For all compounds, pXRD measurements including Rietveld refinements strongly support their phase pure nature (Supplementary Figures 5166). Elemental analyses proved to be unreliable for salts with these kinds of anions; a more thorough discussion is given in the Supplementary Information (S.I. section 12 and Supplementary Table 6).

### Synthesis of [Cr(CO)6][Al(ORF)4] (1)

A double-Schlenk flask was equipped inside the glove box with Cr(CO)6 (51 mg, 0.232 mmol, 1 eq.) and NO[Al(ORF)4] (230 mg, 0.231 mmol, 1 eq.). Then, CH2Cl2 (~4 mL) was condensed at −196 °C onto the reaction mixture and it was allowed to thaw to −78 °C under dynamic vacuum. Upon dissolution, the color turned initially blue and a gas evolution was visible. The mixture was stirred under dynamic vacuum (so that the solvent was very mildly boiling) until it reached r.t. over the period of an hour. The CH2Cl2 (~2 mL) was then filtered off and the remaining solid was dried in vacuo in order to yield 1 as an off-white to beige solid (260 mg, 0.219 mmol, 94%). Single crystals were obtained by slow vapor diffusion of n-pentane in a solution of 1 in 1,2-F2C6H4 (oDFB) at r.t. FTIR (ZnSe, ATR): $$\tilde v$$/cm−1 (intensity) = 2094 (s), 1843 (vvw, trace of [Cr(CO)5(NO)]+), 1508 (vvw, trace of residual oDFB), 1352 (vw), 1298 (mw), 1272 (ms), 1239 (s), 1208 (vvs), 1161 (ms), 969 (vvs), 831 (vw), 756 (vvw), 726 (vvs), 571 (vvw), 560 (vw) (Supplementary Figure 42). FT Raman (1000 scans, 100 mW, 4 cm−1): $$\tilde v$$/cm−1 (intensity) = 2175 (vvs), 2128 (mw), 2062 (vw), 1304 (vvw), 1272 (vvw), 1235 (vvw), 1163 (vvw), 977 (vvw), 798 (w), 747 (w), 572 (vvw), 563 (vvw), 538 (vw), 368 (vvw), 332 (m), 288 (w), 234 (vvw), 173 (vvw) (Supplementary Figure 43). 1H NMR (400.17 MHz, oDFB, 298 K): only solvent signals; 13C{1H} NMR (100.62 MHz, oDFB, 298 K): δ/ppm = 121.7 (q, 1J(C,F) = 293 Hz, 12C, CF3, in part overlapped by solvent signals); 19F NMR (376.54 MHz, oDFB, 298 K): δ/ppm = −75.3 (s, 36F, 4 × C(CF3)3); 27Al NMR (104.27 MHz, oDFB, 298 K): δ/ppm = 35.0 (s, 1Al, Al(ORF)4) (Supplementary Figures 912).

### Synthesis of [Cr(CO)5(NO)][Al(ORF)4] (3)

A double-Schlenk flask was equipped inside the glove box with Cr(CO)6 (51 mg, 0.232 mmol, 1 eq.) and NO[Al(ORF)4] (228 mg, 0.229 mmol, 1 eq.). Then, CH2Cl2 (~3 mL) was added at r.t. to the reaction mixture and the closed vessel was stirred for 14 days at r.t. After drying in vacuo, 3 yielded as an orange solid (250 mg, 0.210 mmol, 92%). Orange single crystals were obtained by slow vapor diffusion of n-pentane in a solution of 3 in oDFB at r.t. FTIR (ZnSe, ATR): $$\tilde v$$/cm−1 (intensity) = 2184 (vw), 2164 (vvw), 2127 (vvw), 2108 (ms), 2074 (vvw), 1841 (mw), 1821 (vvw, trace of unknown impurity), 1353 (vw),1299 (mw), 1273 (ms), 1251 (ms),1239 (ms), 1210 (vvs), 1162 (mw), 1140 (vw), 971 (vvs), 866 (vvw), 832 (vw), 756 (vvw), 727 (vs), 657 (w), 640 (vw), 571 (vvw), 561 (vw) (Supplementary Figure 46). FT Raman (100 scans, 500 mW, 4 cm−1): $$\tilde v$$/cm–1 (intensity) = 2185 (m), 2175 (vw), 2164 (m), 2127 (vvs), 2112 (vw), 2076 (vvw), 1843 (vvw), 1304 (vvw), 1275 (vvw), 977 (vvw), 797 (vw), 747 (vw), 640 (vw), 562 (vvw), 539 (vvw), 487 (vvw), 369 (vvw), 332 (vw), 290 (vvw), 234 (vvw) (Supplementary Figure 47). 1H NMR (300.18 MHz, oDFB, 298 K): only solvent signals; 13C{1H} NMR (100.62 MHz, oDFB, 298 K): δ/ppm = 201.9 (s, 4Cequatorial, Cr(CO)5(NO)), 187.3 (s, 1Caxial, Cr(CO)5(NO)), 121.7 (q, 1J(C,F) = 293 Hz, 12C, CF3, in part overlapped by solvent signals), 79.5 (m, 4C, 4 × C(CF3)3); 14N NMR (21.69 MHz, oDFB, 298 K): δ/ppm = 17 (br. s, 1N, Cr(CO)5(NO)); 19F NMR (376.54 MHz, oDFB, 298 K): δ/ppm = −75.3 (s, 36F, 4 × C(CF3)3; 27Al NMR (78.22 MHz, oDFB, 298 K): δ/ppm = 35.0 (s, 1Al, Al(ORF)4) (Supplementary Figures 1721).