NMR Spectroscopic Evidence for the Intermediacy of XeF 3– in XeF 2 /F – Exchange, Attempted Syntheses and Thermochemistry of XeF 3– Salts, and Theoretical Studies of the XeF 3– Anion

The existence of the trifluoroxenate(II) anion, XeF 3– , had been postulated in a prior NMR study of the exchange between fluoride ion and XeF 2 in CH 3 CN solution. The enthalpy of activation for this exchange,  H ‡ , has now been determined by use of single selective inversion 19 F NMR spectroscopy to be 74.1  5.0 kJ mol –1 (0.18M) and 56.9  6.7 kJ mol –1 (0.36 M) for equimolar amounts of [N(CH 3 ) 4 ][F] and XeF 2 in CH 3 CN solvent. Although the XeF 3– anion has been observed in the gas phase, attempts to prepared the Cs + and N(CH 3 ) 4+ salts of XeF 3– have been unsuccessful, and are attributed to the low fluoride ion affinity of XeF 2 and fluoride ion solvation in CH 3 CN solution. The XeF 3– anion would represent the first example of an AX 3 E 3 VSEPR arrangement of electron lone pair and bond pair domains. Fluoride-ion exchange reactions between XeF 2 and the F – anion have been probed computationally using CCSD and DFT (PBE1PBE) methods. The energy-minimized geometry of the ground state shows that the F – anion is only weakly coordinated to XeF 2 (F 2 Xe---F – ; distorted Y-shape possessing C s symmetry), while the XeF 3– anion exists as a first-order transition state in the fluoride-ion exchange mechanism, and is planar and Y-shaped ( C 2 v symmetry). The molecular geometry and bonding for the XeF 3– anion has been described and rationalized in terms of electron localization function (ELF) calculations, as well as the VSEPR model of molecular geometry. Quantum-chemical calculations at the CCSD/aVTZ level of theory, using a continuum solvent model (CH 3 CN), accurately reproduced the transition-state enthalpy observed


Introduction
Of the binary xenon fluorides, only XeF4 and XeF6 are known to form anionic salts with fluoride ion donors. Xenon tetrafluoride behaves as a weak fluoride ion acceptor (calculated gasphase fluoride ion affinity (FIA), 247.3 kJ mol -1 ) 1 towards alkali metal fluorides and the naked fluoride ion source [N(CH3)4][F] to give salts of the pentagonal planar XeF5anion. 2 Xenon hexafluoride is a considerably stronger fluoride ion acceptor (FIA, 313.8 kJ mol -1 ), 1 forming the XeF7and XeF8 2anions with alkali metal (Na, K, Rb, Cs) fluorides. 3,4 In addition to the alkali metal fluoride salts, the NO + and NO2 + salts have been prepared by direct reaction of XeF6 with NOF or NO2F, namely, [NO2][XeF7], 5  salt was shown, by single-crystal X-ray diffraction, to contain the slightly distorted square antiprismatic XeF8 2anion. 6,7 Xenon hexafluoride also reacts with [NF4][HF2] in anhydrous hydrogen fluoride (aHF) to give [NF4] [XeF7], which was converted to [NF4]2[XeF8] by selective laser photolysis and characterized by vibrational spectroscopy. 8 The XeF7and Xe2F13anions have also been characterized by X-ray crystallography as their Cs + and NO2 + salts. 5 The XeF7anion has a monocapped octahedral structure, while the Xe2F13anion may be described as an XeF6 molecule bridged by two long Xe-F bonds to an XeF7anion such that the bridging fluorines avoid the axial nonbonding electron pair of the XeF6 molecule.
In contrast, XeF2 has not been conclusively shown by experiment to exhibit fluoride-ion acceptor properties in solution or in the solid state (FIA, 83.3 kJ mol -1 ). 1 The trifluoroxenate(II) anion, XeF3 -, was first proposed to be a plausible anionic noble-gas species based on well-known diagonal trends within the Periodic Table. 9 The XeF3  anion was later proposed as an intermediate in the "base-catalyzed" fluorination of SO2 by XeF2, 10 however, reasonable alternative mechanisms were subsequently proposed for this reaction which did not involve this anion. 11 Experimental evidence for the XeF3anion has been obtained in the gas phase from the negative ion mass spectra of XeF2 12 and XeOF4, 13 and from energy-resolved collision-induced dissociation studies of XeF2. 14 The related XeF3 • radical has recently been stabilized in an argon matrix and characterized by Fourier-transform infrared spectroscopy and explored by kinetic measurements. 15 Radiochemical experiments using 18 F (half life = 109.7 min) have failed to confirm fluoride-ion exchange with XeF2 in water, 16 CH2Cl2, 17 or CH3CN solvents. 17 However, 18 Fexchanges between [ 18 F]-HF, [ 18 F]-SiF4, and [ 18 F]-AsF5 and XeF2 have been successfully used for the preparation of [ 18 F]-XeF2 (through XeF + and Xe2F3 + as intermediates), 18 that was, in turn, used for the preparation of [ 18 F]-2-fluoro-2-deoxy-D-glucose 19 and [ 18 F]-6-fluoro-L-3,4dihydroxyphenylalanine. 20 Computational chemistry has more recently been employed to study the nature of the XeF3anion in the gas phase, 1,14 as well as the thermodynamics of its formation (for details see Computational Results). 14 In an earlier study, the authors observed exchange between Fand XeF2 in CH3CN solvent under rigorously anhydrous conditions by use of 2-D 19 F-19 F EXSY experiments. 17 Fluorine-19 exchange occurred between the "naked" fluoride-ion source, [N(CH3)4][F], and XeF2 in CH3CN solvent at 15 ºC, providing the first conclusive evidence for XeF2/Fexchange on the NMR timescale. 17 The 19 F exchange was postulated to proceed through the formation of XeF3 -(eq 1).
The objectives of the present study are to better define the nature of the XeF2/Fexchange, using single selective inversion NMR and to establish the fluoride ion acceptor properties of XeF2 by attempting the syntheses of representative salts containing the XeF3anion.
Computational methods (CCSD and DFT) have also been used to explore the nature of the XeF2/Fexchange. The current study also clarifies inconsistencies that have arisen in recent computational studies relating to the structure of the XeF3anion. 1 and by reaction of XeF2 with CH3CN. 23 The HF2anion was not found to exchange with XeF2 by single selective inversion NMR spectroscopy (   24 and was attributed to the high solvation energies of the fluoride ion in these polar solvents. The syntheses of the Cs + and N(CH3)4 + salts containing the XeF3anion were also Thermochemistry. To account for failed attempts to prepare either the Cs + or N(CH3)4 + salt of the XeF3  anion, quantum-chemical calculations and established semi-empirical methods [25][26][27][28][29] were used in conjunction with known thermodynamic quantities to estimate Hºrxn, Sºrxn, and Gºrxn for eq 1 in the absence of a solvent. The standard enthalpies for the reactions were determined by analyzing their Born-Fajans-Haber cycles (eq 2). The enthalpy change for the gas-phase reaction (eq 3) corresponds to the negative of the fluoride ion affinity (FIA) of XeF2. A prior reported fluoride ion affinity value of 83.3 kJ mol 1 has been used. 1 The experimental value for the enthalpy of sublimation (Hº(XeF2)sub) for XeF2 (55.71 kJ mol -1 ) was used. 30 The salt. For the salts under investigation, which are singly charged and non-linear, the following values were used: I = 1, α = 117.3 mm kJ mol -1 , β = 51.9 kJ mol -1 , and p = 2. In this formalism, ∆HL o is the lattice enthalpy and is defined as the energy required to break the crystal lattice, and therefore has a positive value. The volume-based approach is particularly useful because the formula unit volume (Vm) of an unknown salt can be estimated with reasonable accuracy using several methods. 27 A method for estimating the absolute standard entropy of a salt from its unit volume has been reported by Jenkins and Glasser (eq 5) where k = 1360 JK -1 mol -1 nm -3 (formula unit) -1 and c = 15 J mol -1 K -1 . 29 When coupled with the experimental standard entropies of XeF2(s) (115.09 J mol -1 K -1 ), 30 this method allows Sºrxn (eq 6) and Gºrxn (eq 7) to be calculated for the reactions of interest.
The Sºrxn and Gºrxn values obtained for these reactions are: The reaction enthalpies and Gibbs free energies are positive indicating that both reactions are endothermic and non-spontaneous under standard conditions. As expected, the reaction employing the larger N(CH3)4 + cation is far less endothermic and closer to spontaneity.
Computational Results. It was first suggested that the XeF3anion would have an octahedral To fully appreciate the complexity of this seemingly simple system, it is important to distinguish the two, discrete forms of the anion that have stationary points on the potential energy surface (PES), as computed in the present work and in previous studies: 1,14 a planar, distorted Y-shaped anion (Cs symmetry, Figure 3a) possessing two short Xe-F bonds and one long Xe-F bond (which is still within the sum of the van der Waals radii of xenon and fluorine, 3.63 Å 32 ), and a planar, Y-shaped anion (C2v symmetry, Figure 3b) possessing three similar Xe-F bond lengths. The former will be referred to as the F2Xe---Fadduct, while the latter will be referred to as the XeF3anion. A third, T-shaped adduct of C2v symmetry was also computed as a stationary point on the PES, 1,14 but it was found that a slight, symmetry-lowering distortion to Cs symmetry gave the lower energy, ground-state structure, which is, in fact, the mer-structure (albeit with one long Xe---F interaction). Finally, a Y-shaped anion (D3h) was also computed as a diradical species, in accord with the structure predicted by MO theory, 14 but was found to be 59.8 to 110.0 kJ mol -1 higher in energy than the ground-state structure. the results, along with the present work, are summarized in Table 1. In the former report, DFT calculations with the B3LYP functional indicated that the XeF3anion (C2v) is the energyminimized structure (all frequencies real, Table S1 in the Supporting Information) when either an all-electron Maroulis basis set or an ECP basis set (SDB-cc-pVTZ) is used for xenon. 14 The F2Xe---Fadduct (C2v) was higher in energy relative to the XeF3anion by ca.  (Table 1). Furthermore, the existence of the XeF3anion as a transition state was not explored. 1 Calculations performed in the present work (DFT using the PBE1 functional, and CCSD; basis sets: aVTZ (for F) and aVTZ(-PP) (for Xe) were done without symmetry constraints, and reveal that the global-energy minimum of the anion is the F2Xe---Fadduct (Cs symmetry), in In order to elucidate the structure of the transition state, a scan involving the step-wise reduction in the long Xe-F bond length of the F2Xe---Fadduct was computed. In both cases (PBE1 and CCSD), the transition-state structure was that of the Y-shaped XeF3anion ( Figure   3b). The structures thus obtained were further optimized as transition states at their respective levels of theory (Table 2). Calculations were also performed to determine if the T-shaped, C2vsymmetric transition state was a potential, competing transition state. The C2v-symmetric transition state was 14.6 (PBE1) and 16.7 (CCSD) kJ mol -1 higher in energy than the Y-shaped XeF3anion, indicating that the latter is the more favored transition state. Thermodynamic calculations for the intermolecular exchange between XeF2 and Fwere also reinvestigated in the present work (see Computational Assessment of Intermolecular Exchange).
The transition state (XeF3 -) appears to be far more covalently bound than the groundstate F2Xe---Fadduct; using DFT methods, the XeF3anion possesses three similar Xe-F bond lengths (2.115 Å and 2.182 Å (×2)). The two large (145.2º) angles and one small (69.6º) angle allow for the retention of the lone-pair torus around the xenon atom, as shown by ELF analyses (see Electron Localization Function (ELF) Analyses). The structure computed at the CCSD/aVTZ level of theory was found to possess similar geometric parameters to that of the DFT structure (Table 1). It is unclear why the XeF3anion was computed to be an energy minimum in the previous study 14 since the structure remains a transition state even when the calculations are performed using the same basis set-density functional combination as employed in the original work.
(ii) In CH3CN Solution. Because of the thermodynamics for the enthalpy of activation of Fion exchange with XeF2 in CH3CN solvent were found to be in good agreement with the experimental findings (see Computational Assessment of Intermolecular Exchange), the structures for F2Xe---Fand XeF3were also determined in CH3CN (using the CPCM model, 33 Table 1). For the F2Xe---Fadduct, the bond between XeF2 and F  loses some of its covalent character. The optimized structure contains a near-linear XeF2 fragment (PBE1, 178.2º; CCSD, 178.3º) with a long Xe-F bond equal to 3.112 Å (PBE1) and 3.035 Å (CCSD) that is still within the sum of the Xe and F van der Waals radii. Unlike the gas-phase structures, the geometric parameters in CH3CN solution do not change significantly between PBE1 and CCSD (Table 1), and show that the interaction between XeF2 and Fis primarily electrostatic in nature, precluding the observation of an F2Xe---Fadduct by 19 F NMR spectroscopy in this solvent or by lowtemperature Raman spectroscopy in frozen CH3CN solution. In contrast to the adduct, the geometry for the XeF3anion modeled in CH3CN solution remains essentially unchanged from that determined in the gas phase (Table 1). solution. Because the enthalpy of formation is small for the F2Xe---Fadduct, the enthalpy for the formation of XeF3 -(i.e., the enthalpy of activation) could be determined relative to free XeF2 and F -, and gave energy barriers for the fluoride ion exchange reaction that were much higher than those in the gas phase: 24.7 (PBE1) and 50.2 kJ mol -1 (CCSD). While the value using the PBE1 functional was underestimated, the value obtained using the CCSD method fell within the range of those experimentally determined for the fluoride-ion exchange (65.5 ± 8.4 kJ mol -1 ).

(b) Computational Assessment of Intermolecular Exchange between Fluoride
Taken as whole, the calculations confirm that the direct reaction between XeF2 and F  in solution leads to fluoride ion exchange via a transition state corresponding to the Y-shaped XeF3  anion.
Finally, in order to assess the validity of averaging the two activation enthalpies obtained by 19 F NMR spectroscopy (i.e., that the two energies did not correspond to two different

Conclusions
The standard enthalpy of formation for the intermolecular 19   Bruker RFS 100 FT Raman spectrometer using 1064-nm excitation as previously described. 36 Spectra were recorded in glass or ¼-in. FEP sample tubes using a laser power of 200 mW and a total of 300 scans.
Computational Methods. Molecular geometries were optimized with CCSD and DFT (using the PBE1PBE exchange-correlation functional) with aug-cc-pVTZ (F) and aug-cc-pVTZ-PP (Xe) basis sets (basis sets obtained from https://bse.pnl.gov/bse/portal). The nature of all stationary points found was assessed by performing frequency analyses. Calculations involving solvent effects were performed using the CPCM model 33 and employing the default parameters for CH3CN. Natural atomic orbital and natural bond orbital analyses were conducted using the NBO 5.0 code; 37 electron localization function analyses were performed with the program package TopMod. 38 Visualizations of molecular structures and ELF isosurfaces were done with the gOpenMol program. 39,40 Quantum-chemical calculations were carried out using the program Gaussian 09 41 for geometry optimizations, vibrational frequencies, and their intensities and the program Gaussian 03 42 for NBO analysis.