Structure and Bonding in Au(I) Chloride Species: A Critical Examination of X-ray Absorption Spectroscopy (XAS) Data

Structure and Bonding in Au(I) Chloride Species: A Critical Examination of X-ray Absorption Spectroscopy (XAS) Au(I) chloride species are important reactants and intermediates in various processes across the chemical sciences and engineering. Structure and bonding in Au(I) species are often characterized by X-ray absorption spectroscopy (XAS), including measurements under reaction conditions. Previously reported XA spectra for Au(I) chloride species have varied significantly, likely as a result of radiation damage and/or partial disproportionation of [AuCl 2 ]  ions, which are metastable under ambient conditions. By monitoring the decomposition of tetrabutylammonium dichloroaurate(I), TBA[AuCl 2 ], in 1,2-dichlorobenzene we have obtained a reliable X-ray absorption spectrum of [AuCl 2 ] ‾ ions by combining the calculation of difference spectra with an extended X-ray absorption fine-structure (EXAFS) determination of the solution composition. The results show that the X-ray absorption near- edge structure (XANES) of [AuCl 2 ]  is characterized by a weak Au 2p 3/2 → 5d (‘white line’) transition, which agrees well with the spectrum predicted by electronic structure calculations using the FEFF8 code. Compared to [AuCl 4 ]  , the determined [AuCl 2 ] ‾ spectrum has several distinctive features of diagnostic analytical value. A more detailed densities of states (DOS) analysis of the electronic structure suggests that the weak white line arises from a hybrid Au 6s/5d DOS band that is partially occupied, up to the level of the highest occupied molecular orbital (HOMO). Correlation of Cl coordination numbers determined from the EXAFS with the intensity of the white line in the XANES indicates that the decomposition is a primarily radiation-induced oxidation to Au(III) species with an average formula of [AuCl 3 Au(I) chloride species are important reactants and intermediates in various processes across the chemical sciences and engineering. Structure and bonding in Au(I) species are often characterized by X-ray absorption spectroscopy (XAS), including measurements under reaction conditions. Previously reported XA spectra for Au(I) chloride species have varied significantly, likely as a result of radiation damage and/or partial disproportionation of [AuCl 2 ] ‾ ions, which are metastable under ambient conditions. By monitoring the decomposition of tetrabutylammonium dichloroaurate(I), TBA[AuCl 2 ], in 1,2-dichlorobenzene we have obtained a reliable X-ray absorption spectrum of [AuCl 2 ] ‾ ions by combining the calculation of difference spectra with an extended X-ray absorption fine-structure (EXAFS) determination of the solution composition. The results show that the X-ray absorption near-edge structure (XANES) of [AuCl 2 ] ‾ is characterized by a weak Au 2p 3/2 → 5d (‘white line’) transition, which agrees well with the spectrum predicted by electronic structure calculations using the FEFF8 code. Compared to [AuCl 4 ] ‾ , the determined [AuCl 2 ] ‾ spectrum has several distinctive features of diagnostic analytical value. A more detailed densities of states (DOS) analysis of the electronic structure suggests that the weak white line arises from a hybrid Au 6s/5d DOS band that is partially occupied, up to the level of the highest occupied molecular orbital (HOMO). Correlation of Cl coordination numbers determined from the EXAFS with the intensity of the white line in the XANES indicates that the decomposition is a primarily radiation-induced oxidation to Au(III) species with an average formula of [AuCl 3 OH] ‾ .


Introduction
Au(I) chloride species are important reactants and intermediates in heterogeneous and homogeneous Au catalysis, 1, 2 electrochemical processes, 3 nanoparticle synthesis, 4-6 geochemical Au speciation, [7][8][9] and gold mining. 10 X-ray absorption spectroscopy (XAS) measurements can provide incisive information on the chemical state of Au in solid Au(I) chloride and its associated solution species. 8 The X-ray absorption fine-structure (XAFS) in these spectra provides information on the electronic properties of the Au centers through the X-ray absorption near-edge structure (XANES) and on molecular structure, such as Au-Cl coordination numbers and bond lengths, through the extended X-ray absorption fine structure (EXAFS). 11,12 In the context of Au speciation analysis from L 3 -edge XANES, the spectral feature receiving most attention is the socalled 'white line' resonance at a photon energy of approximately 11921 eV ( fig. 1), which arises from the excitation of core Au 2p 3/2 electrons to unoccupied Au 5d or 6s states. Compared to metallic Au, Au(III) compounds tend to have lower 5d and 6s occupancies and therefore exhibit prominent white lines, [13][14][15][16][17][18] while Au(I) compounds tend to have weaker white line absorptions. 14,17 White line intensity analysis is therefore used to identify Au oxidation states 14 although there are cases in which such correlations need to be applied with care. 19,20 For example, in the context of heterogeneous catalysis, Au chloride species are common precursors in catalyst preparation 21 and reliable identification of Au species is important for generating the deep mechanistic understanding required for catalyst design. Certainty about the spectroscopic signature of Au(I) chloride is required to reach reliable conclusions. However, previously published Au L 3 -edge XANES spectra of Au(I) chloride species 6,8,[13][14][15][16] have varied considerably. Quite prominent white lines have been reported in some cases, 17,22 which suggest a superposition of absorption from significant concomitant Au(III) concentrations, either due to contamination or from reaction products formed spontaneously in solution. Even the published spectrum of [AuCl 2 ]‾ obtained under hydrothermal conditions, which strongly favor Au(I), has significant intensity in the white line region. 8 Generally, atmospheric pressure, non-acidic pH and absence of chloride ions favor Au(III), while hydrothermal and acidic chloride-rich aqueous electrolytes stabilize Au(I). 8 The variations between published spectra are likely due to (i) the sensitivity of cationic Au species to radiation-induced redox transformations, 23 and (ii) the metastability of Au(I) chloride under ambient conditions, especially in aqueous solution. 9,17 Examining the dichloroaurate(I) ion as its tetrabutylammonium salt, TBA[AuCl 2 ] in the non-aqueous solvents 1,2-dichlorobenzene (DCB), toluene and 1,2dichloroethane (DCE) we also observed evidence for Au(III) formation as a clearly enhanced white line feature ( fig. 1). This feature was weaker in toluene than in DCB solution, suggesting higher stability of [AuCl 2 ]‾ in toluene, but the low TBA[AuCl 2 ] solubility of approximately 1 mM in toluene did not allow us to collect EXAFS data with sufficient signal-to-noise quality for a more detailed analysis. For DCE, the spectrum of 5 mM TBA[AuCl 2 ] (not shown in fig. 1) indicated metallic Au only, suggesting rapid reduction from Au(I) to Au(0). However, in DCB solutions containing 6 mM TBA[AuCl 2 ], the spectral changes between each measurement were slow enough to enable monitoring of the reaction progress by XAS. In the following we will use this time-dependence of the Au L 3 -edge spectrum to advantage, to reliably isolate the [AuCl 2 ]‾ XANES through a combined EXAFS and XANES analysis. The analysis will show that TBA[AuCl 2 ] appears to be oxidized through the influence of radiation on the DCB solution, forming Au(III) species with an average composition of [AuCl 3 OH]‾. We will compare the determined [AuCl 2 ]‾ spectrum with the XANES predicted by electronic structure calculations and with the spectrum of aqueous [AuCl 2 ]‾.  Table 1) and 1 mM TBA[AuCl2] in toluene (Sample #2). Only the spectra arising from sample #1 will be analyzed in detail in this contribution.

Experimental
To obtain a spectrum of aqueous [AuCl 2 ]‾ (sample #4), it was extracted into water from a DCE solution through the DCE/water interface, similar to a previously described phase transfer process. 24 26 The data were first calibrated using gold foil and the edge-step heights normalized to 1. For the EXAFS analysis, a Hanning-type Fourier transform window was used. Fitting was done in R-space using k 1 -weighting. The EXAFS scattering paths were generated from the published crystal structure of AuCl. 27 The error values reported for fitted parameters are the diagonals of the covariance matrix multiplied by the square-root of the reduced chi-square. The EXAFS R-factor indicates the closeness-of-fit. 12 Only 3 variables out of the 15.1 independent points were used. The fitted k-and R-ranges were 3-11 Å -1 and 1.0-4.0 Å respectively.

1 EXAFS Analysis and the Au(III) Species
Without exposure to the X-ray beam, [AuCl 2 ]‾ in DCE is stable, as known from electrochemical experiments. 29 Using cyclic voltammetry we likewise established the stability of [AuCl 2 ]‾ in DCB solution ( fig. S1). We can therefore conclude that the observed changes in fig. 1 were indeed radiation-induced. We further established that radiation converted Au(I) to Au(III) in DCB by recording the spectrum of an unstirred solution, followed by a spectrum at the same position under stirring. As can be seen in fig, 1, the white line intensity decreased upon stirring, indicating that Au(III) concentrations accumulated in the region of the unstirred solution exposed to the X-ray beam. Aqueous [AuCl 2 ]‾ is known to undergo spontaneous disproportionation to [AuCl 4 ]‾ and Au metal. 9 However, for the radiation-induced oxidation observed in DCB, neither visual inspection of the sample cells nor EXAFS analysis (vide infra) provided any evidence for the presence of metallic Au. We return to the possible mechanism of the radiation-induced Au(III) formation in DCB further below, after establishing the likely composition of the Au(III) species from a combined EXAFS and XANES analysis.  2). Such linear MS paths are commonly used in EXAFS models for linear and square planar Au(III) compounds. 8,30 The parameters for the EXAFS model are summarized in Table 2. The amplitude factor S 0 2 for all the scattering paths was fixed at 0.82 and the Debye Waller (DW) factor for Cl, σ 2 Cl was fixed at 0.002 Å 2 . These values were obtained from the fitting of the TOA[AuCl 4 ] reference solution in DCB (sample #7, Table 1) and are in close agreement with values previously reported for [AuCl 4 ]‾. 8,30,31 We will in the following assume that identical DW factors apply to Au−Cl bonds in Au(I) and Au(III) complexes, which, strictly speaking, introduces uncertainty because the DW factor is dependent on the amplitude of the Au−Cl vibrations. However, given the similar Au−Cl bond ‾, it appears unlikely that the differences introduce a major error. No metallic Au-Au scattering was evident from the EXAFS analysis, suggesting the absence of dimers or other self-associated species. When carrying out the EXAFS fitting with k 1 -weighting we noted that the fit quality improved upon the inclusion of an Au-O scattering path, with σ 2 O and R O fixed at 0.002 Å 2 and 1.96 Å, which are values comparable to those previously found for hydrolyzed Au(III) species in water, 18,30,31 suggesting the presence of hydroxo ligands. Higher k-weightings were less sensitive to the contribution of this weak Au-O scattering, which is perhaps expected because Au-O scattering is strongest at low k-values. The initial E 0 value was chosen within the postwhite line and pre-EXAFS region at 11926 eV to minimize the shift in energy, ∆E in the EXAFS fitting. 32 As can be seen in Table 3 Table 3). Parameters S0 2 is the amplitude reduction factor; σ 2 is the Debye Waller factor; R is the Au-scatterer distance or the half-path length; x is the mole fraction; N is the coordination number or path degeneracy; ∆E is the edge energy shift; the initial choice for E0 was 11926 eV.   Table 3.

Determination of [AuCl 2 ]‾ XANES by Difference Spectrum Analysis
As a complementary and independent means for determining the mole fractions x of Au(I) and Au(III) in DCB solutions, a difference spectrum analysis of the XANES spectra was carried out. For this we assumed that the experimental spectra, ܵ .ୣ୶୮ can be expressed as a sum of the spectra from [AuCl 2 ]‾ , ܵ ሾ୳େ୪ మ ሿ‾ , and an arbitrary Au(III) compound, ܵ ୳ሺ୍୍୍ሻ : i is the scan number (scan 1, 2, or 3) of the experimental spectra ( fig. 1, Table 1). We can then determine the spectrum of pure [AuCl 2 ]‾, ܵ ሾ୳େ୪ మ ሿ‾, ୢୣ୲ from two experimental spectra (scan 1 and 2) by eliminating ܵ ୳ሺ୍୍୍ሻ from Equation 1, yielding One such spectrum is shown in fig. 4, labelled '[AuCl 2 ]‾ difference spectrum analysis'. With knowledge of ܵ ሾ୳େ୪ మ ሿ‾ ,ୢୣ୲ , we can now also determine the corresponding Au(III) spectrum,  We used a least-square method to calculate ‫ݔ‬ ୳ሺ୍୍୍ሻ,ଵ and‫ݔ‬ ୳ሺ୍୍୍ሻ,ଶ . The difference between the determined Au(III) spectrum, ܵ ୳ሺ୍୍୍ሻ,ୢୣ୲ and the experimentally obtained Au(III) spectrum, ܵ ୳ሺ୍୍୍ሻ,ୣ୶୮ was minimized by changing ‫ݔ‬ ୳ሺ୍୍୍ሻ,ଵ and ‫ݔ‬ ୳ሺ୍୍୍ሻ,ଶ using the Solver in Microsoft Excel. We  A FEFF8 simulation of the [AuCl 2 ]‾ L 3 -edge reproduces all salient features in the experimental XANES spectrum ( fig. 5), including the peaks at 11922 and 11930 eV. Analysis of the lprojected unoccupied density-of-states (DOS) plots shows that the weak absorption in the white line region arises from a transition to a hybrid partially unoccupied Au 6s/5d band intersected by what FEFF8 identifies as the 'Fermi energy', but should perhaps more correctly be referred to as the highest occupied molecular orbital (HOMO) for a molecular species. The well-known 6s/5d hybridization arises from the relativistic effect on the 6s state. 34,35 There is also significant p-DOS in this energy range, but its influence on the XANES spectrum is not evident because only transitions with ∆l = ±1, i.e. p 3/2 → d or p 3/2 → s, are dipole-allowed. Examination of the DOS in Fig.  5 further indicates that the resonance at 11930 eV, which is in the region of the onset of EXAFS backscattering, arises from a transition to unoccupied states with Cl 3d-Au 5d hybridization. 30 but clearly more work is required to unequivocally confirm this or any alternative mechanism.
To examine the effect of spontaneous disproportionation of aqueous [AuCl 2 ]‾ under ambient conditions we also measured its spectrum (sample #4 in Table 1, the group of spectra in the lower part of fig. 6). We obtained both the transmission and the fluorescence-yield XANES spectra. Comparison of these data with the spectrum obtained in DCB indicated that partial disproportionation to Au(0) and Au(III) had taken place. 9 This was evident from the fluorescence-yield XANES spectrum, which has a somewhat stronger white line than the transmission spectrum, as a result of metallic gold deposition on the walls of the sample cell during the measurement. Fluorescence-yield detection is much more sensitive to such thin wall deposits 37 than transmission detection, which probes the bulk composition of the solution. In line with this, subsequent scans of the same sample (not shown) yielded fluorescence-yield spectra that evolved to mainly metallic features; meanwhile the transmission spectra acquired Au(III) and Au(0) features, supporting the fact that disproportionation of Au(I) to Au(III) and Au(0) had taken place in the aqueous solution during the measurements. indicates that the internal electronic structure of the ion is not strongly influenced by the solvent and cation. Somewhat stronger white lines in other reported data 17,22 may be due to either residual Au(III) content or weak solvent interactions.
We note particularly that compared to Au foil ( fig. 6, upper set of spectra) the absorption in the white line region in the [AuCl 2 ]‾ spectrum is actually weaker. This indicates that the relationship between oxidation state and white line intensity may be more complex than sometimes assumed. 14 Indeed, the results of the FEFF8 DOS calculation (fig. 5) suggest a high sensitivity to the exact position of the 'Fermi energy' (highest occupied molecular orbital, HOMO), which determines Au 6s/5d band occupancy, relative to the contributions from Au 5d, 6s and 6p states. This reflects the well-known dependence of chemical bonding on ligand species, coordination number and geometric arrangement around the X-ray absorbing central atom. 39,40 More systematic studies are needed to elucidate these complex relationships further.  3 OH]‾ as shown in the correlation between the EXAFS-derived Cl coordination numbers and the white line intensities in the spectra. Difference spectrum analysis of the XANES arrives at the same quantitative conclusions and likewise allowed us to eliminate the influence of the Au(III) species on the spectrum. Compared with [AuCl 4 ]‾ the determined [AuCl 2 ]‾ spectrum has several distinctive features with diagnostic value: a small white line at 11922 eV, a second peak at 11930 eV and a broad shoulder at 11942 eV, which are also evident in some previously reported Au(I) chloride spectra. The energetic positions of these features are quantitatively reproduced by FEFF8 calculations of the [AuCl 2 ]‾ XANES. It appears that the radiation-induced oxidation and partial hydrolysis of [AuCl 2 ]‾ to [AuCl 3 OH]‾ most likely follows a mechanism involving radicals formed in the DCB solvent matrix as well as water dissolved in the solvent.

Introduction
Au(I) chloride species are important reactants and intermediates in heterogeneous and homogeneous Au catalysis, 1, 2 electrochemical processes, 3 nanoparticle synthesis, 4-6 geochemical Au speciation, 7-9 and gold mining. 10 X-ray absorption spectroscopy (XAS) measurements can provide incisive information on the chemical state of Au in solid Au(I) chloride and its associated solution species. 8 The X-ray absorption fine-structure (XAFS) in these spectra provides information on the electronic properties of the Au centers through the X-ray absorption near-edge structure (XANES) and on molecular structure, such as Au-Cl coordination numbers and bond lengths, through the extended X-ray absorption fine structure (EXAFS). 11,12 In the context of Au speciation analysis from L 3 -edge XANES, the spectral feature receiving most attention is the socalled 'white line' resonance at a photon energy of approximately 11921 eV ( fig. 1), which arises from the excitation of core Au 2p 3/2 electrons to unoccupied Au 5d or 6s states. Compared to metallic Au, Au(III) compounds tend to have lower 5d and 6s occupancies and therefore exhibit prominent white lines, [13][14][15][16][17][18] while Au(I) compounds tend to have weaker white line absorptions. 14,17 White line intensity analysis is therefore used to identify Au oxidation states 14 although there are cases in which such correlations need to be applied with care. 19,20 For example, in the context of heterogeneous catalysis, Au chloride species are common precursors in catalyst preparation 21 and reliable identification of Au species is important for generating the deep mechanistic understanding required for catalyst design. Certainty about the spectroscopic signature of Au(I) chloride is required to reach reliable conclusions. However, previously published Au L 3 -edge XANES spectra of Au(I) chloride species 6,8,[13][14][15][16] have varied considerably. Quite prominent white lines have been reported in some cases, 17,22 which suggest a superposition of absorption from significant concomitant Au(III) concentrations, either due to contamination or from reaction products formed spontaneously in solution. Even the published spectrum of [AuCl 2 ]‾ obtained under hydrothermal conditions, which strongly favor Au(I), has significant intensity in the white line region. 8 Generally, atmospheric pressure, non-acidic pH and absence of chloride ions favor Au(III), while hydrothermal and acidic chloride-rich aqueous electrolytes stabilize Au(I). 8 The variations between published spectra are likely due to (i) the sensitivity of cationic Au species to radiation-induced redox transformations, 23 and (ii) the metastability of Au(I) chloride under ambient conditions, especially in aqueous solution. 9,17 Examining the dichloroaurate(I) ion as its tetrabutylammonium salt, TBA[AuCl 2 ] in the non-aqueous solvents 1,2-dichlorobenzene (DCB), toluene and 1,2dichloroethane (DCE) we also observed evidence for Au(III) formation as a clearly enhanced white line feature ( fig. 1). This feature was weaker in toluene than in DCB solution, suggesting higher stability of [AuCl 2 ]‾ in toluene, but the low TBA[AuCl 2 ] solubility of approximately 1 mM in toluene did not allow us to collect EXAFS data with sufficient signal-to-noise quality for a more detailed analysis. For DCE, the spectrum of 5 mM TBA[AuCl 2 ] (not shown in fig. 1) indicated metallic Au only, suggesting rapid reduction from Au(I) to Au(0). However, in DCB solutions containing 6 mM TBA[AuCl 2 ], the spectral changes between each measurement were slow enough to enable monitoring of the reaction progress by XAS. In the following we will use this time-dependence of the Au L 3 -edge spectrum to advantage, to reliably isolate the [AuCl 2 ]‾ XANES through a combined EXAFS and XANES analysis. The analysis will show that TBA[AuCl 2 ] appears to be oxidized through the influence of radiation on the DCB solution, forming Au(III) species with an average composition of [AuCl 3 OH]‾. We will compare the determined [AuCl 2 ]‾ spectrum with the XANES predicted by electronic structure calculations and with the spectrum of aqueous [AuCl 2 ]‾.  Table 1) and 1 mM TBA[AuCl2] in toluene (Sample #2). Only the spectra arising from sample #1 will be analyzed in detail in this contribution.

Experimental
To obtain a spectrum of aqueous [AuCl 2 ]‾ (sample #4), it was extracted into water from a DCE solution through the DCE/water interface, similar to a previously described phase transfer process. 24 26 The data were first calibrated using gold foil and the edge-step heights normalized to 1. For the EXAFS analysis, a Hanning-type Fourier transform window was used. Fitting was done in R-space using k 1 -weighting. The EXAFS scattering paths were generated from the published crystal structure of AuCl. 27 The error values reported for fitted parameters are the diagonals of the covariance matrix multiplied by the square-root of the reduced chi-square. The EXAFS R-factor indicates the closeness-of-fit. 12 Only 3 variables out of the 15.1 independent points were used. The fitted k-and R-ranges were 3-11 Å -1 and 1.0-4.0 Å respectively.
FEFF8.2 28 was used to simulate the XANES of monomer [AuCl 2 ]‾, with Au-Cl distance at 2.27 Å. XANES, FMS and SCF cards were used -FMS for full-multiple scattering XANES calculation; SCF to enable self-consistent field iterations. The ION card was not used.

1 EXAFS Analysis and the Au(III) Species
Without exposure to the X-ray beam, [AuCl 2 ]‾ in DCE is stable, as known from electrochemical experiments. 29 Using cyclic voltammetry we likewise established the stability of [AuCl 2 ]‾ in DCB solution ( fig. S1). We can therefore conclude that the observed changes in fig. 1 were indeed radiation-induced. We further established that radiation converted Au(I) to Au(III) in DCB by recording the spectrum of an unstirred solution, followed by a spectrum at the same position under stirring. As can be seen in fig, 1, the white line intensity decreased upon stirring, indicating that Au(III) concentrations accumulated in the region of the unstirred solution exposed to the X-ray beam. Aqueous [AuCl 2 ]‾ is known to undergo spontaneous disproportionation to [AuCl 4 ]‾ and Au metal. 9 However, for the radiation-induced oxidation observed in DCB, neither visual inspection of the sample cells nor EXAFS analysis (vide infra) provided any evidence for the presence of metallic Au. We return to the possible mechanism of the radiation-induced Au(III) formation in DCB further below, after establishing the likely composition of the Au(III) species from a combined EXAFS and XANES analysis.  2). Such linear MS paths are commonly used in EXAFS models for linear and square planar Au(III) compounds. 8,30 The parameters for the EXAFS model are summarized in Table 2. The amplitude factor S 0 2 for all the scattering paths was fixed at 0.82 and the Debye Waller (DW) factor for Cl, σ 2 Cl was fixed at 0.002 Å 2 . These values were obtained from the fitting of the TOA[AuCl 4 ] reference solution in DCB (sample #7, Table 1) and are in close agreement with values previously reported for [AuCl 4 ]‾. 8,30,31 We will in the following assume that identical DW factors apply to Au−Cl bonds in Au(I) and Au(III) complexes, which, strictly speaking, introduces uncertainty because the DW factor is dependent on the amplitude of the Au−Cl vibrations. However, given the similar Au−Cl bond lengths in [AuCl 2 ]‾ and [AuCl 4 ]‾, it appears unlikely that the differences introduce a major error. No metallic Au-Au scattering was evident from the EXAFS analysis, suggesting the absence of dimers or other self-associated species. When carrying out the EXAFS fitting with k 1 -weighting we noted that the fit quality improved upon the inclusion of an Au-O scattering path, with σ 2 O and R O fixed at 0.002 Å 2 and 1.96 Å, which are values comparable to those previously found for hydrolyzed Au(III) species in water, 18,30,31 suggesting the presence of hydroxo ligands. Higher k-weightings were less sensitive to the contribution of this weak Au-O scattering, which is perhaps expected because Au-O scattering is strongest at low k-values. The initial E 0 value was chosen within the postwhite line and pre-EXAFS region at 11926 eV to minimize the shift in energy, ∆E in the EXAFS fitting. 32 As can be seen in Table 3 Table 3).    Table 3.

Determination of [AuCl 2 ]‾ XANES by Difference Spectrum Analysis
As a complementary and independent means for determining the mole fractions x of Au(I) and Au(III) in DCB solutions, a difference spectrum analysis of the XANES spectra was carried out. For this we assumed that the experimental spectra, ܵ .ୣ୶୮ can be expressed as a sum of the spectra from [AuCl 2 ]‾ , ܵ ሾ୳େ୪ మ ሿ‾ , and an arbitrary Au(III) compound, ܵ ୳ሺ୍୍୍ሻ : ܵ ୳ሺ୍୍୍ሻ,ୢୣ୲ , shown in fig. 4, labelled 'Au(III) difference spectrum analysis'. We used a least-square method to calculate ‫ݔ‬ ୳ሺ୍୍୍ሻ,ଵ and‫ݔ‬ ୳ሺ୍୍୍ሻ,ଶ . The difference between the determined Au(III) spectrum, ܵ ୳ሺ୍୍୍ሻ,ୢୣ୲ and the experimentally obtained Au(III) spectrum, ܵ ୳ሺ୍୍୍ሻ,ୣ୶୮ was minimized by changing ‫ݔ‬ ୳ሺ୍୍୍ሻ,ଵ and ‫ݔ‬ ୳ሺ୍୍୍ሻ,ଶ using the Solver in Microsoft Excel. We  A FEFF8 simulation of the [AuCl 2 ]‾ L 3 -edge reproduces all salient features in the experimental XANES spectrum ( fig. 5), including the peaks at 11922 and 11930 eV. Analysis of the lprojected unoccupied density-of-states (DOS) plots shows that the weak absorption in the white line region arises from a transition to a hybrid partially unoccupied Au 6s/5d band intersected by what FEFF8 identifies as the 'Fermi energy', but should perhaps more correctly be referred to as the highest occupied molecular orbital (HOMO) for a molecular species. The well-known 6s/5d hybridization arises from the relativistic effect on the 6s state. 34,35 There is also significant p-DOS in this energy range, but its influence on the XANES spectrum is not evident because only transitions with ∆l = ±1, i.e. p 3/2 → d or p 3/2 → s, are dipole-allowed. Examination of the DOS in Fig.  5 further indicates that the resonance at 11930 eV, which is in the region of the onset of EXAFS backscattering, arises from a transition to unoccupied states with Cl 3d-Au 5d hybridization. 30 but clearly more work is required to unequivocally confirm this or any alternative mechanism.
To examine the effect of spontaneous disproportionation of aqueous [AuCl 2 ]‾ under ambient conditions we also measured its spectrum (sample #4 in Table 1, the group of spectra in the lower part of fig. 6). We obtained both the transmission and the fluorescence-yield XANES spectra. Comparison of these data with the spectrum obtained in DCB indicated that partial disproportionation to Au(0) and Au(III) had taken place. 9 This was evident from the fluorescence-yield XANES spectrum, which has a somewhat stronger white line than the transmission spectrum, as a result of metallic gold deposition on the walls of the sample cell during the measurement. Fluorescence-yield detection is much more sensitive to such thin wall deposits 37 than transmission detection, which probes the bulk composition of the solution. In line with this, subsequent scans of the same sample (not shown) yielded fluorescence-yield spectra that evolved to mainly metallic features; meanwhile the transmission spectra acquired Au(III) and Au(0) features, supporting the fact that disproportionation of Au(I) to Au(III) and Au(0) had taken place in the aqueous solution during the measurements. There are several distinctive features in the determined [AuCl 2 ]‾ XANES spectrum that allow the identification of Au(I) chloride species. The upper group of spectra in fig. 6 compares the determined [AuCl 2 ]‾ spectrum with the spectrum of TOA[AuCl 4 ] and with the original spectrum of the TBA[AuCl 2 ] solution in DCB. It can be seen that the white line maximum for [AuCl 2 ]‾ is shifted to higher photon energy by approximately 1.5 eV relative to the value of 11920.5 eV observed for [AuCl 4 ]‾. A broad shoulder is evident in the photon energy range around ~11942 eV. However, the resonance at ~11930 eV is most distinctive for [AuCl 2 ]‾, as it occurs several eV below the corresponding feature in the [AuCl 4 ]‾ spectrum. We note that this transition is evident in a set of recently published time-resolved XANES data for the process of X-ray induced Au nanoparticle formation from [AuCl 4 ]‾ in ionic liquids, where it is most clearly visible after approximately 4 h. 38 Its presence suggests the formation of an [AuCl 2 ]‾ intermediate in the process, possibly supporting the AuCl 2 -dimer model put forward in the study, which was derived from an EXAFS analysis. Judging from the XANES obtained in our work it would appear that a model involving Au nanoparticles, [AuCl 2 ]‾ and [AuCl 4 ]‾ may provide an alternative explanation.
The [AuCl 2 ]‾ spectrum determined from XANES difference spectrum analysis has a weak white line, similar to that in some previously reported data for [AuCl 2 ]‾ species -including measurements under ambient and hydrothermal conditions (600 bar, 250°C) 8 as well as a spectrum of solid TBA[AuCl 2 ] ( fig. 6, lower set of spectra). The features at 11930 and 11942 eV are very similar in these Au(I) chloride XANES spectra. The agreement between these spectra in different environments indicates that the internal electronic structure of the ion is not strongly influenced by the solvent and cation. Somewhat stronger white lines in other reported data 17,22 may be due to either residual Au(III) content or weak solvent interactions.
We note particularly that compared to Au foil ( fig. 6, upper set of spectra) the absorption in the white line region in the [AuCl 2 ]‾ spectrum is actually weaker. This indicates that the relationship between oxidation state and white line intensity may be more complex than sometimes assumed. 14 Indeed, the results of the FEFF8 DOS calculation (fig. 5) suggest a high sensitivity to the exact position of the 'Fermi energy' (highest occupied molecular orbital, HOMO), which determines Au 6s/5d band occupancy, relative to the contributions from Au 5d, 6s and 6p states. This reflects the well-known dependence of chemical bonding on ligand species, coordination number and geometric arrangement around the X-ray absorbing central atom. 39,40 More systematic studies are needed to elucidate these complex relationships further.

Conclusions
By combined difference spectrum and EXAFS analysis we have determined the XANES spectrum of [AuCl 2 ]‾ from several consecutive XAS scans of TBA[AuCl 2 ] in DCB. The [AuCl 2 ]‾ spectrum obtained by our analysis can serve as a reliable reference for XANES studies requiring the identification and quantification of [AuCl 2 ]‾ content. Radiation-induced oxidation of TBA[AuCl 2 ] appears to result in [AuCl 3 OH]‾ as shown in the correlation between the EXAFS-derived Cl coordination numbers and the white line intensities in the spectra. Difference spectrum analysis of the XANES arrives at the same quantitative conclusions and likewise allowed us to eliminate the influence of the Au(III) species on the spectrum. Compared with [AuCl 4 ]‾ the determined [AuCl 2 ]‾ spectrum has several distinctive features with diagnostic value: a small white line at 11922 eV, a second peak at 11930 eV and a broad shoulder at 11942 eV, which are also evident in some previously reported Au(I) chloride spectra. The energetic positions of these features are quantitatively reproduced by FEFF8 calculations of the [AuCl 2 ]‾ XANES. It appears that the radiation-induced oxidation and partial hydrolysis of [AuCl 2 ]‾ to [AuCl 3 OH]‾ most likely follows a mechanism involving radicals formed in the DCB solvent matrix as well as water dissolved in the solvent.