Redox and Ligand Exchange during the Reaction of Tetrachloroaurate with Hexacyanoferrate(II) at a Liquid-Liquid Interface: Voltammetry and X-ray Absorption Fine-Structure Studies

Voltammetry for charge (ion and electron) transfer at two immiscible electrolyte solutions (VCTIES) has been used to provide insight into the ligand exchange and redox processes taking place during the interfacial reaction of aqueous hexacyanoferrate(II) with tetrachloroaurate ([AuCl 4 ] (cid:1) ) in 1,2-dichloro-ethane (DCE). VCTIES permitted the detection of the reactants, intermediates and products at the liquid/ liquid interface. A model for the sequence of interfacial processes was established with the support of speciation analysis of the key elementary reactions by X-ray absorption spectroscopy (XAS). The potential-driven transfer of [AuCl 4 ] (cid:1) from the organic into the aqueous phase is followed by reduction and ligand exchange


Introduction
Au(III) and Au(I) complexes in aqueous solution can undergo simultaneous transformations such as ligand substitution and redox reactions, which are of particular interest due to their importance in the synthesis of metallic Au nanoparticles, and the extraction/purification of the metal.The redox transformations between Au(III), Au(I) and Au(0) species are influenced by reducing agents, coexisting ligands, electrolytes, and pH, resulting in Au complexes with varying standard redox potentials in water [1][2][3][4][5].
The reaction between tetrachloroaurate, [AuCl 4 ] À and hexacyanoferrate(II) (ferrocyanide, [Fe(CN) 6 ] 4À ) has previously been examined [6][7][8][9][10].It was proposed that hexacyanoferrate(II), used as a reducing agent, produced dicyanoaurate, [Au(CN) 2 ] À , as well as complexes of Fe 3+ and [AuCl 2 ] À [6][7][8].On the other hand, hexacyanoferrate(III) (ferricyanide, [Fe(CN) 6 ] 3À ) was employed to decrease the size of Au nanoparticles previously prepared by the oxidation of Au(0) [6].The pH and electrolyte dependence of aqueous Au(III) and Au(I) species need to be taken into consideration when modelling the reduction process though the thermodynamic properties of species such as [AuCl 3 (OH)] -and [AuCl 2 (OH) 2 ] - [11,12].Au(I) chloro-complexes, which are plausible reduction intermediates, also readily decompose by disproportionation to Au(III) and metallic Au [13].Voltammetry for charge (ion and electron) transfer between two immiscible electrolyte solutions (VCTIES) has been used to investigate the reduction of Au ions where the Au species are initially located in an organic phase in contact but immiscible with the aqueous phase containing the reducing reagent.Au deposition at the liquid/liquid interface is well known having been the subject of a number of previous studies including [9,10,[14][15][16][17][18][19][20].In the case of hexacyanoferrate, the charge transfer current corresponding to the reduction of [AuCl 4 ] À and the concomitant oxidation of hexacyanoferrate(II) has been observed [9,10].This reaction relies on the initial transfer of [AuCl 4 ] À from the organic to the aqueous phase which is then followed by a homogeneous reaction with hexacyanoferrate(II) to form Au(I), [AuCl 2 ] À .VCTIES can be applied to identify the ionic species based on their ion transfer potential between the aqueous and organic solutions, which is proportional to the difference in solvation energy of ions between the two solutions [21].The electrochemical transfers of [AuCl 4 ] À [9,10,14,[17][18][19][20], [AuBr 4 ] À [10] or [AuCl 2 ] À [19,20] between water and organic solutions have been reported, however, the transfers of other Au complexes are, to the best of our knowledge, yet to be reported.
Our VCTIES have been combined with X-ray absorption fine structure (XAFS) measurements which can provide detailed information about the local structure around the atoms of a specific element.The coordination structure of Au complexes have been examined in detail, including the identification of valency from X-ray absorption near-edge structure (XANES) [20,22], and characterization of AuÀ ÀCl [23][24][25][26], AuÀ ÀOH [23][24][25] and AuÀ ÀCN [26,27] bonds through the analysis of extended X-ray absorption fine-structure (EXAFS).As XAFS is an elementally specific technique, in-situ XAFS measurements are capable of identifying Au(III), Au(I) and Au(0) species formed in the presence of reducing agents [25,26,28,29] or at high temperature [30].As such, XAFS analysis of the homogenous solutions provides quantitative understanding of the ligand exchange reactions of the Au complexes and their redox reactions.
In the present study, we combine VCTIES measurements and XAFS as complimentary techniques to understand the interfacial reaction between Au ions and hexacyanoferrate(II).The ionic species [AuCl 4 ] À and [AuCl 2 ] À formed in aqueous solutions were found to be dependent on the concentrations of H + , electrolyte and hexacyanoferrate(III) and (II) in the system.From these observations we discuss a possible mechanism for the redox process between hexacyanoferrate(II) in aqueous solution and [AuCl 4 ] À in organic solution.

Measurement of the voltammogram for charge transfer at the macro and micro water | DCE interface
Two electrochemical cells were employed for the VCTIES: one with a macroscale and one with a micro-scale contact between the two phases.In a conventional macro-interface cell [10], cyclic voltammetry experiments were performed using a four electrode configuration with an IVIUM "Compactstat" potentiostat (IVIUM Technologies, the Netherlands).No iR compensation was applied for the electrochemical measurements.Homemade Ag|AgCl and platinum gauze were used as the reference electrodes (RE) and counter electrodes (CE) respectively.The counter electrode in DCE was coated with glass to avoid the contact of platinum with water.The water | DCE interface had a cross-sectional area of 0.64 cm 2 and a volume of 2 cm 3 .Further details are described elsewhere [10].The micro-interface cell consists of a water chamber and a DCE chamber separated by a 16 mm thick polyester film with a micro hole 30 mm in diameter.The water | DCE interface was formed at the micro hole [31][32][33].
In both cells, the potential difference at the water | DCE interface, E, was measured.The potential of an Ag|AgCl electrode in water, was referred to the potential of a BTPPA + Where E ref is the potential of the reference electrodes employed.In the calculations of the Gibbs energy of ion transfer between water and DCE, D W DCE G ð¼ ÀzFD W DCE f ; z and F are the charge and the Faraday constantÞ, the measured E was converted using the extra-thermodynamic assumption of Parker [34].

XAFS measurements
XAFS spectra were acquired at the spectroscopy beamline I18 of DIAMOND Light Source (Harwell Science and Innovation Campus, UK) [35].All measurements were collected in fluorescence-yield mode using an Ortec multi-element solid-state Ge detector to measure the Au L 3 edge.The electron storage ring runs at 3 GeV with a current of 300 mA.A double crystal Si(111) monochromator with an intrinsic resolution of 1.4 Â 10 À4 DE/E was used.When using the full beam the flux at the gold L 3 -edge is 8 Â 10 11 photons.
Each measurement took $30 min.The beam size was $420 mm Â 280 mm.The strong XANES resonance visible in the spectra at about 11 918 eV reflects an intra-atomic electronic transition of Au 2p 3/2 core electrons to unoccupied valence states.Spectral features in the Au L 3 -edge XANES beyond this photon energy range are additionally influenced by back scattering of photoelectrons.The high sensitivity of XAFS to unoccupied valence 5d and 6s-states of Au allows the identification of the oxidation state, electronic configuration and coordination geometry.
For each measurement, a 1 mL sample solution was placed in a 1.5 mL Eppendorf microcentrifuge tube positioned vertically on a magnetic stirrer plate.XAFS analysis were carried out using the Demeter software package [36].Energy scales were calibrated according to a standard procedure where the first inflection point of gold foil is known to be 11 919 eV.EXAFS for the AuÀ ÀClÀ ÀOH system (Table 2) were fitted in FT-space using simultaneous k 1 , k 2 and k 3 weightings where k is the photoelectron wave vector.A Hanning-type window with dk = 1 was used for Fourier Transformation from k-space into FT-space.A k-range of 3-12 Å À1 and Rrange of 1-4 Å were used for the fittings, except for the pH 11.77 sample where the window ranges are 3-10 Å À1 and 1.25-4 Å.One energy shift parameter (DE 0 ) was used for all of the shells.To account for the changes in relative Au(III) and Au(0) compositions, x Au(III) S 0 2 and x Au(0) S 0 2 were assigned to each fit where x denotes composition and S 0 2 denotes amplitude reduction factor.Note that x Au(III) S 0 2 and x Au(0) S 0 2 almost always sum to $0.9.Any instances in the text referring to FT-peak before phase shift will be indicated.Linear combination fitting for the AuÀ ÀClÀ ÀCN system (Fig. 8) was carried out over the range of 11 899 to 11 949 eV.The samples for the Au-Cl-CN system were measured as soon as possible after sample preparation; $30 min gap between sample preparation and XAFS measurements.

Results and discussion
The stability of [AuCl 4 ] À and [AuCl 2 ] À ions as a function of H + concentration and electrolytes (in the absence of hexacyanoferrate) was first studied (sections 3.1 and 3.2 respectively) with the aim to prevent decomposition of these Au species.Once the solution conditions that stabilize these gold chloride ions were found, the reactions between [AuCl 4 ] À or [AuCl 2 ] À with hexacyanoferrate (II) or (III) were investigated by VCTIES and XAS (sections 3.3 and 3.4).

Stability of [AuCl 4 ] À as a function of pH and Cl -concentration
The dependence of [AuCl 4 ] À stability on Cl -concentration and pH has been investigated previously by UV-visible spectroscopy [11].VCTIES is introduced here as an alternative method to determine the stability of [AuCl 4 ] À in water at various pH and Cl -concentrations.The stability of [AuCl 4 ] À in an acidic aqueous solution was examined by cyclic voltammogram (CV) recorded at the interface between an aqueous phase containing Na + [AuCl 4 ] -(0.2 mM) in 10 mM HCl (pH = 2) and DCE (10 mM TOA + TFPB -).Positive and negative current features at 0.148 and 0.082 V corresponding to the transfer of [AuCl 4 ] À between water and DCE were observed.The mid-point potential was calculated to be 0.115 V (peak A) which is close to the previously reported values [9,10].
In view of this evidence, a previous assignment of the peak to [AuCl 2 ] -appears less likely [18].Assuming that peak B corresponds to the transfer of [AuCl 3 (OH)] À , its ion transfer potential can be estimated from D W DCE G of Cl -(51.1 kJ mol -1 [37] or 46.2 kJ mol -1 [38]) and OH -(67.6 kJ mol -1 [37] or 63.3 kJ mol -1 [39]) as shown in Table 1.The difference between D W DCE G of Cl -and that of OH -is thus 16.8 kJ mol -1 on average, close to the difference between the ion transfer potentials of peaks A and B (17.3 kJ mol -1 , Fig. 1(a) and  (b)).We therefore propose that peak B corresponds to the transfer EXAFS structural parameters for the Au(III) chloride-hydroxide systems shown in Fig. 2. [Cl -À ÀOH -], and the difference between ion transfer potentials of peaks A and B suggests that the latter is dependent on the difference between the solvation energies of Cl -and OH -in water and DCE, rather than the difference between AuÀ ÀCl and AuÀ ÀOH bond strengths.When 5 mM Li 2 SO 4 was used instead of LiCl (Fig. 1(c)), another pair of negative and positive peaks with a mid-point potential at -0.254 V (peak C) was observed in addition to peaks A and B. SO 4 2À is more hydrophilic than Cl -therefore extending the negative region of the potential window [38].Peak C is assigned to the transfer of [AuCl 2 (OH) 2 ] À because the difference of the midpoint potentials of peaks B and C is close to the difference between the D W DCE G of Cl -and OH -, in a similar manner to the assignment of peak B to [AuCl 3 (OH)] À described above.
The dependence of the Au(III) hydrolysis on H + concentration was investigated at a constant Cl À concentration of 10 mM (Fig. S2).Peak A attributed to the transfer of [AuCl 4 ] À was seen to be more pronounced as H + concentration was increased whereas peak B assigned to [AuCl 3 (OH)] À was only observed at lower H + concentrations.Based on the data obtained, we suggest that the pre-peak observed in the literature [9] when 10 mM KCl was employed as a supporting electrolyte in water corresponds to the transfer of [AuCl 3 (OH)] À .In addition, the dependence of Au(III) hydrolysis on Cl -concentration was independently examined at a constant H + concentration of 10 mM (Fig. S3).In this case peak A corresponding to the transfer of [AuCl 4 ] À increases with an increase in Cl -concentration.Peak B assigned to [AuCl 3 (OH)] À was observed at lower concentrations of Cl -.When no Cl -was added, both peaks A and B are less pronounced because of the preferential formation of [AuCl 2 (OH) 2 ] À through further hydrolysis of Au(III) as observed as peak C in Fig. 1.
[AuCl 4 ] À hydrolysis as a function of pH was followed by XAS (Fig. 2).Solution pH was varied by the addition of OH À while maintaining a constant Cl -concentration at 0.5 M. [AuCl 4 ] -in an acidic solution (pH 2.40) was also measured to serve as a reference.The strong white line at 11 920 eV showed constant intensity, except for the most basic case at 50 mM OH -(pH = 11.77,Fig. 2(a)).The presence of the intense white line for all the solutions indicates that the gold oxidation state remained at +III independent of pH.The FT peak at 1.87 Å (before phase shift, Fig. 2(c)) corresponding to the AuÀ ÀCl scattering path, decreased in intensity with increasing pH.As Cl À was gradually exchanged with OH À with an increase in pH, the somewhat shorter FT peak at 1.55 Å became more intense.The weaker intensity of the AuÀ ÀOH scattering path arises from the fact that oxygen is a lighter element than chlorine, with fewer core electrons and hence a smaller back-scattering amplitude.The coordination numbers of Cl -and OH -(N Cl,exp and N OH,exp ) and the bond distances of AuÀ ÀCl and AuÀ ÀOH (R Cl,exp and R OH,exp ) were quantified from the EXAFS data and plotted as a function of pH (Table 2 and Fig. 3).The decrease in N Cl,exp with pH is accompanied by an increase in N OH,exp .R Cl,exp decreased with increasing pH, whereas R OH,exp was constant.The average coordination numbers of Cl -and OH -were calculated from the stability constants of Au(III)À ÀClÀ ÀOH complexes (N Cl,cal and N OH,cal ) [11] at each pH (Fig. 3).It was found that N Cl,exp and N OH,exp at pH 2.40, 5.75, and 7.01 were very close to N Cl,cal and N OH,cal .For instance at pH 5.75, a solution composition of 34% [AuCl 4 ] -, 58% [AuCl 3 (OH)] À and 8% [AuCl 2 (OH) 2 ] À was predicted; thus N Cl,cal = 3.32 and N OH,cal = 0.68 which are in very good agreement with N Cl,cal = 3.2 (2) and N OH,cal = 0.77 (9).However, N Cl,exp and N OH,exp at pH = 10.25 and 11.77 deviated from the N Cl,cal and N OH,cal values expected for homogeneous molecular hydroxo complexes [11] because of metallic Au formation.The EXAFS analysis revealed AuÀ ÀAu scattering, i.e. the presence of metallic Au.The AuÀ ÀAu scattering parameters have also been included in Table 2.

Stability of [AuCl 2 ] À in water as a function of pH and Cl - concentration
The chemical stability of [AuCl 2 ] À in water was investigated in the presence of either HCl, LiCl or Li 2 SO 4 (Fig. 4).There are few reliable papers on the stability of [AuCl 2 ] À in aqueous solution because of the spontaneous disproportionation of [AuCl 2 ] À at ambient conditions [13,30].Since [AuCl 2 ] À is stable in DCE [22], voltammetry for the transfer of [AuCl 2 ] À at the water | DCE interface is used here to shed more light on the stability of this complex.The stability of [AuCl 2 ] À in acidic aqueous solutions was examined in the same way as that of [AuCl 4 ] À above.
A voltammogram at the interface between an aqueous solution with 10 mM HCl (pH = 2) and a DCE phase containing 0.2 mM TOA + [AuCl 2 ] À and 10 mM TOA + TFPB -was recorded (Fig. 4-a).Positive and negative currents were observed at 0.058 and -0.004 V, which were controlled by the diffusion of [AuCl 2 ] À from DCE to water.The mid-point potential was 0.026 V (peak D).To examine the stability of [AuCl 2 ] À in acidic aqueous solutions a controlled potential electrolysis experiment was carried out.Under the same reactant conditions as above, a potential of 0.4 V was applied for 30 min as a means to extract AuCl 2 -into the aqueous phase electrochemically.The voltammogram measured after the transfer (data not shown) was very similar to the original sample indicating that there was no electrolysis.If [AuCl 2 ] À was unstable in aqueous solution we would anticipate the development of peaks at the transfer potential of [AuCl 4 ] À (peak A) generated by disproportionation of [AuCl 2 ] À , or a peak at the negative end of the potential window indicative of hydrolysis of the chloride complex.As no such peaks were visible in Fig. 4(a) we therefore suggest that [AuCl 2 ] À was stable in the presence of 10 mM HCl. Surprisingly, 100 mM LiCl produced a very similar response where the diffusion controlled [AuCl 2 ] À ion transfer was the only peak observed (peak D, Fig. 4(b).By analogy with Au(III) hydrolysis, when 100 mM LiCl was used instead of 10 mM HCl, currents corresponding to the transfer of hydrolyzed species such as [AuCl(OH)] À were expected at the negative end of the potential window.As with HCl, a fixed potential of 0.4 V was applied for 30 min to transfer [AuCl 2 ] À into the aqueous phase to detect the possible hydrolysis products.However, the initial and final voltammogram was similar suggesting that [AuCl 2 ] À is stable in a neutral aqueous solution for at least 30 min, possibly because [AuCl 2 ] À was additionally stabilized by the presence of Cl - electrolyte.
In line with this, an additional test was carried out to check the stability of [AuCl 2 ] À in water.An aqueous solution of [AuCl 2 ] À was prepared by exchange with [AuCl 4 ] À as described in the experimental section.[AuCl 2 ] À in water was stable in the short term, but metallic gold was found at the interface after a few hours, indicating the well-known disproportionation of [AuCl 2 ] À in water to metallic Au and [AuCl 4 ] À [13], Eq. ( 5).  2. N Cl,cal and N OH,cal were calculated based on the stability constants reported in Ref. [11].
When Li 2 SO 4 was used as the supporting electrolyte therefore lowering the Cl À and H + concentrations in the solution, the peak current for the transfer of [AuCl 2 ] À decreased with each scan.In this case the application of 0.4 V for 1800 s to drive transfer into the aqueous phase did induce electrolysis as shown by the difference in the initial state (solid line) and final state (dotted line) CVs in Fig. 4(c).Also, a broad positive and negative current corresponding to the transfer of Cl -formed by the disproportionation of [AuCl 2 ] À was observed at -0.25 V, which is close to the transfer of SO 4 2À at -0.32 V (corresponding to the negative end of the potential window).
The negative current corresponding to the transfer of [Au (CN) 2 ] À was indeed observed (Fig. 6).The half wave potential for the transfer of [Au(CN) 2 ] À agreed with that obtained in r = 10 in Fig. 5(a-1).These results indicate the concomitant reduction of Au (III) to Au(I) and ligand exchange.Although [AuCl 2 ] À was not observed we cannot rule out its possible involvement as a fast lived intermediate as per Eq. ( 6).The [AuCl 2 ] À generated in water may then form a complex with hexacyanoferrate(II) or hexacyanoferrate(III) to yield [Au(CN) 2 ] À as Eqs.( 7) and/or (8) [8]. And/or XAS characterization of this system of coupled equilibria was conducted under conditions similar to those of the voltammetric measurements again examining HAuCl 4 and hexacyanoferrate(II) solutions at different values of r.Fig. 7(a) shows XANES spectra for Au complexes as a function of hexacyanoferrate(II) concentration.The spectra of [AuCl 4 ] À and [Au(CN) 2 ] À in the absence of hexacyanoferrate (II) were also measured as references (Fig. 7(a)).Fig. 7(b) and (c) show the k 3 -weighted EXAFS spectra and their corresponding Fourier transforms.As expected with increasing r, the FT peak at $1.8 Å (before phase shift, Fig. 7(c)) corresponding to AuÀ ÀCl decreased and the $1.5 Å and 2.8 Å peaks corresponding to AuÀ ÀCN became more intense.We performed linear combination fitting at each value of r (0.5, 1, 2 and 5) assuming only [Au(CN) 2 ] -and [AuCl 4 ] -in the solution (Fig. 8(a) and (b), and Fig. S4).Good fits can be obtained for r = 2 and 5 where hexacyanoferrate(II) is in excess resulting in a predominantly [Au(CN) 2 ] -solution.In contrast, data for r = 0.5 and 1 suggest the presence of an unknown intermediate in the region of 11 924 eV where the fits could not be described by a combination of [AuCl 4 ] - and [Au(CN) 2 ] -.We examined the possible presence of [Au (CN) 4 ] -or [AuCl 2 ] À [22] (fits not shown) but neither offered a good fit and principal component analysis did not suggest their presence in the sample.As such further analysis is required to determine the structure of the unknown species.The standards mentioned here are provided for reference in Fig. S5.At r = 2 and higher, the AuÀ ÀCN species is dominant based on both EXAFS observations and XANES linear combination analysis which is in line with the anticipated stoichiometry of the reaction where two moles of hexacyanoferrate(II) are required to reduce one mole of [AuCl 4 ] -.
It can therefore be concluded that [Au(CN) 2 ] À must be the species responsible for peak E in the presence of excess Table 3 Standard potential of each redox couple in water [43].
Alongside the reaction between [AuCl 4 ] À and hexacyanoferrate (II), hexacyanoferrate(II) could decompose into hexacyanoferrate (III) which would in turn form Prussian blue with hexacyanoferrate (II) according to Eq. ( 9) [40].We confirmed the presence of Prussian blue by UV-vis spectroscopy in this study (data not shown).We note that no reaction was observed between aqueous [AuCl 4 ] À and hexacyanoferrate(III) not hexacyanoferrate(II) described above.
Recall that in the VCTIES experiment described earlier in this section (Figs. 1 and 5(a-1)), the first negative current (peak A) was assigned to the transfer of [AuCl 4 ] À from water to DCE.Peak A remains similar even after the sample was aged for one day.The relative compositions of each species can be quantified from the limiting currents obtained in the VCTIES.The cyclic voltammetry for the fresh samples are shown in Fig. 5(a-1) with the relative compositions for the samples shown in Fig. 5(b-1); the samples aged for one day are shown in Fig. 5(a-2) and (b-2).When a twofold excess of hexacyanoferrate(II) to [AuCl 4 ] À was added, i.e. = 2, a second negative current (peak E) corresponding to the transfer of [Au(CN) 2 ] À was also observed with 21.3% for the fresh sample and gradually changed to 75.0% after ageing for one day; the trend is expected because [Au(CN) 2 ] À is the reaction product.Interestingly, an additional ion transfer peak was observed for r = 0.5, 1, 2 and 4 for the fresh samples at À0.25 V (peak F), which is close to the negative current limit (transfer of Cl -as a supporting electrolyte).The limiting current of peak F was found to decrease after 1 day.Although initially present in all samples apart from r = 0 and r = 10 after 24 h, the signal only remained present for r = 1 and 2, having decreased from 42.7% to 16.7% for the r = 1 sample and from 49.7% to 25.0% for the r = 2 sample.Thus, we suggest that peak F corresponds to a transient intermediate species in the reaction between [AuCl 4 ] À and hexacyanoferrate(II).
Cheng et al. [9] reported a peak similar to peak F. They assigned the peak to the electron transfer at the interface corresponding to the reduction of [AuCl 4 ] À to metallic Au in DCE, accompanied by the oxidation of hexacyanoferrate(II) to hexacyanoferrate(III) as per Eq (10).
If the electron transfer reaction described in Eq. ( 10) were to occur, Cl -would form in the organic solution instead of the aqueous.However, no current for the transfer of Cl -from the organic to the aqueous phase was observed within the potential window of the voltammogram (Fig. 5(a)).

Stoichiometric analysis of the reduction of [AuCl 2 ] À
To further investigate the identity of the unknown peak F we examined the possibility of reactions between hexacyanoferrate (II) and [AuCl 2 ] À .Even though the presence of [AuCl 2 ] À was not detected by VCTIES or XAFS, considering that (i) [Au(CN) 2 ] À is formed, and that (ii) [AuCl 2 ] À is itself able to react with hexacyanoferrate (II), [AuCl 2 ] À could possibly be an intermediate for the reaction between [AuCl 4 ] À and hexacyanoferrate (II).Fig. 9(a) shows the voltammogram recorded at the macro interface between an aqueous phase containing hexacyanoferrate(II) and a DCE phase containing TOA + [AuCl 2 ] À .Here, [AuCl 2 ] À was directly dissolved in DCE to speed up the reaction with hexacyanoferrate (II).In the first scan, a positive current for the transfer of [AuCl 2 ] À from DCE to water was observed (peak D).The negative current pair for the return transfer of [AuCl 2 ] À from water to DCE was however not observed as the [AuCl 2 ] À was consumed by the reaction.Instead, when the potential cycle was repeated, the positive current seemingly shifted to a more negative potential (from 0.045 V to 0.005 V) because of the formation of [Au (CN) 2 ] -through ligand exchange between [AuCl 2 ] À and hexacyanoferrate (II) in the aqueous phase as per Eq. ( 7).
This was then compared to a voltammogram for the presence of hexacyanoferrate(III) instead of hexacyanoferrate(II) and DCE containing [AuCl 2 ] À is shown in Fig. 9(b).Hexacyanoferrate(III) had failed to react with [AuCl 4 ] À .The positive current for the transfer of [AuCl 2 ] À from DCE to water (peak D) decreased with successive scans and the corresponding negative current was not observed.However, positive and negative currents at -0.2 V were observed.Even though the ion transfer current of [AuCl 2 ] À decreased with scan number, the transfer of [Au(CN) 2 ] À which could be formed by the ligand exchange reaction with [AuCl 2 ] À and hexacyanoferrate(III) was not observed.Therefore, assuming that the reactions proceed via the substitution pathway shown above, the mechanism was via Eq.( 7) and not Eq.( 8), during the experiment.These results indicate that the positive and negative currents at -0.2 V correspond to an ion transfer reaction (instead of an electron transfer reaction) between water and DCE as there is no redox reaction in this example, and that the ionic species formed is more hydrophilic than [AuCl 2 ] À and [Au(CN) 2 ] À based on the ion transfer potentials.
To confirm that the formation of the unknown species is occurring in the aqueous phase, an aqueous [AuCl 2 ] À solution was prepared.Even though ion transfer of [AuCl 2 ] À was observed in the absence of hexacyanoferrate(III) (dotted line in Fig. 10), a negative and positive current at À0.2 V was observed in the presence of hexacyanoferrate(III) (solid line in Fig. 10), verifying that the unknown ionic species is formed.

Interfacial nanoparticle synthesis
We have been able to make observations with regard to the stability of Au species in solution.The voltammetry data suggests that the hydrolysis of Au species is dependent on the Cl À ion concentration as well as the strong influence of pH.A high pH or low Cl À concentration results in the formation of Au hydrolysis products ([AuCl 3 (OH)] À or [AuCl 2 (OH) 2 ] À ).A mixture of Au species has a significant influence on nanoparticle formation because of the different reduction potentials of the Au complexes.For instance, the hydrolysis species of Au(III) halides as well as cyanide species of Au(III) and Au(I) do not easily react with reducing agents.Mixed speciation can therefore result in size polydispersity or a reduction in the concentration of nanoparticles formed [41].
Examination of the reduction of [AuCl 4 ] À by hexacyanoferrate (II) was motivated by the previous by the work of Cheng and Schiffrin on the interfacial reduction of organic TOA + [AuCl 4 ] À by hexacyanoferrate(II) [9].In their report electron transfer at the liquid/liquid interface results in the formation of gold nanoparticles which are stabilized at the interface.However our own observations for the system in aqueous solution have instead pointed to the preferential formation of [Au(CN) 2 ] À which appears to be too stable to undergo further reduction by hexacyanoferrate (II).Previous reports have indicated that nanoparticles can be 0 0.5  formed by the reduction of [Au(CN) 2 ] À by the stronger reducing agent sodium borohydride [42].

Conclusions
The chemical stability of [AuCl 4 ] À and [AuCl 2 ] À have been investigated as a function of electrolyte concentration and pH.Using VCTIES, [AuCl 4 ] À was found to undergo hydrolysis at high pH and low Cl À concentration as has been shown in previous studies.In the case of measurements at the liquid/liquid interface it has been shown that the difference between D W DCE G of [AuCl 4 ] À and [AuCl 3 (OH)] -was close to the difference between D W DCE G of Cl -and OH -.The XAFS data shows a clear pH dependence on Au hydrolysis which agrees well with calculations based on stability constants in acidic conditions, however at high pH the data suggests that there is a higher concentration of chlorinated Au(III) species than would be anticipated by stability constants possibly due to the formation of metallic Au species.The N Cl,exp and N OH,exp at neutral pH calculated using the EXAFS agreed with N Cl,cal and N OH,cal calculated from the stability constants of Au-Cl-OH complexes.The decomposition of [AuCl 2 ] À in aqueous phase was observed at neutral pH and in the absence of Cl À .
Using the combination of VCTIES and XAFS experiments we have been able to examine the reaction between [AuCl 4 ] À and hexacyanoferrate(II) in water.It was found that [AuCl 4 ] À readily underwent reduction by hexacyanoferrate(II) to form [Au(CN) 2 ] À .The formation of [Au(CN 2 )] À was confirmed calculations of the D W DCE G and XAFS combination analysis.The reaction is complicated by the presence of an as yet undetermined intermediate species which was consumed within 24 h at most values of r investigated, however it was stable enough to be measured by both VCTIES and XAFS.The species is shown by VCTIES to be more hydrophilic than [Au(CN) 2 ] À , [AuCl 2 ] À and [AuCl 4 ] À .Although not directly detected by VCTIES or XAFS, [AuCl 2 ] À could be a reaction intermediate as it readily reacts with hexacyanoferrate(II) forming [Au(CN) 2 ] À and a peak in the voltammogram that resembles that of the undetermined intermediate species in the reaction of hexacyanoferrate(II) with [AuCl 4 ] À (peak F).The lack of metallic Au formation by hexacyanoferrate(II) is related to the rapid formation of [Au(CN) 2 ] À and the stability of [Au(CN) 2 ] À with respect to further reduction.

Fig. 2 .Fig. 3 .
Fig. 2. XAFS of the AuÀ ÀClÀ ÀOH solution systems as a function of pH.All samples contain 5 mM Na + [AuCl 4 ] À and 500 mM NaCl.The pH 2.40 sample contains 5 mM HCl.The pH 5.75, 7.01, 10.25 and 11.77 samples contain 5, 10, 15 and 50 mM NaOH respectively.(a) The normalized XANES spectra.(b) The k 3 -weighted EXAFS spectra and (c) the corresponding FT-space spectra converted from k-space using a window range of 3.0-11 Å À1 .The measured spectra are indicated by solid lines and the fitted spectra by dashed lines.

Fig. 8 .Fig. 9 .
Fig. 8. Linear combination fitting of XANES spectra collected in a single aqueous phase solution at different r.The solution contains 5 mM H + [AuCl 4 ] À and either 2.5 mM hexacyanoferrate(II) (r = 0.5, a-1) or 25 mM hexacyanoferrate(II) (r = 5, a-2).For both (a-1) and (a-2) the original spectra are plotted alongside the linear combination fitting result (dotted line) and the relative ratios of the fitting components AuCl 4 À and AuCN 2 À .(c) The relative ratios of the fitting components AuCl 4 À and AuCN 2 À as a function of r.

Fig. 10 .
Fig.10.Cyclic voltammogram at a water|DCE interface.The aqueous phase contains 0.5 mM [AuCl 2 ] À and 10 mM HCl either in the absence (dotted line) or presence (solid line) of 2.5 mM hexacyanoferrate(II). The DCE phase contains 10 mM TOA + TFPB À .The scan rate was 10 mV s À1 .Voltammetric measurements were carried out directly after the preparation of aqueous phase [AuCl 2 ] À through ion exchange with [AuCl 4 ] À as described in section 2.1.

Table 1
Standard ion transfer potential between water and DCE.