On the Extraction of HCl and H2PtCl6 by Tributyl Phosphate: A Mode of Action Study

ABSTRACT Combining computational modeling with experimental measurements has revealed the self-assembly of nano-aggregate structures in the transfer of HCl and PtCl62– from an aqueous phase into toluene by the common industrial extractant tributyl phosphate (TBP). Molecular dynamics simulations have been coupled to analytical measurements to provide an atomistic interpretation of the mode of action of TBP under 6 M and 10 M HCl conditions. The structures conform to reverse micelles, where the Cl– or PtCl62– core is encapsulated by a hydration shell that acts as a mediating bridge to the electronegative oxygen atom in the TBP phosphate groups. For the 6 M HCl extraction model, the data support stable aggregates forming from 2–3 TBP molecules around one chloride anion if the number of water molecules encapsulating the chloride anion is no more than five; increasing the water content to 10 molecules allows a fourth TBP molecule to coordinate. For the 10 M HCl extraction model, stable structures are obtained that conform to the empirical formula (TBP.HCl.H2O)3–5. At 6 M HCl, extraction of PtCl62– is achieved by encapsulation by four TBP molecules; the data for extraction at 10 M HCl indicate larger aggregates containing multiple PtCl62– anions are likely to be forming. In all cases, the hydrated core regions of the reverse micelles are considerably exposed. The diameters of the self-assembled structures around chloride ions agree well with available literature data from small-angle neutron-scattering experiments.


Introduction
Tri-n-butyl phosphate (TBP, Fig. 1) is a well-known compound employed in the recovery of inorganic acids (notably HCl, HClO 4 , and HNO 3 ) and metals by solvent extraction from acidic media. [1][2][3][4] It is used extensively in the nuclear industry for the reprocessing of uranium, plutonium, and thorium in the PUREX process [5] and finds application as a modifier and solvent in many processes. [6,7] This article concerns the mode of action of TBP in the recovery of PtCl 6 2from aqueous HCl solutions of the type obtained by oxidative leaching of platinum group metal (PGM) concentrates. [8] In such a process, PtCl 6 2is present in an acidic solution that contains other chloridometalate anions and a large excess of Cl −. [9] In many cases, the extraction of chloridometalates from acidic solutions can be represented by a simple ion pairing, to create charge-neutral assemblies that contain little or no water. [10][11][12][13][14][15][16] TBP can function as a "solvating extractant," [17] bonding in either the inner or the outer coordination sphere (or both) of a neutral metal complex, as in extraction of uranyl nitrate. [18] However, it can also operate by a different mechanism that involves formation of reverse micelle aggregates in which TBP behaves in a similar manner to a surfactant. [19] The reverse micelle structure is achieved by the polar head group of TBP associating with a water pool of the extracted anion(s) that is subsequently encapsulated by the aliphatic hydrophobic chains. These chains protrude into the organic solvent, essentially providing a layer between the water and the hydrophobic solvent. The formation of reverse micelles and, more generally third phases, by TBP has been studied extensively for nitrate-containing systems because of its importance in the separation and recovery of radionuclides. [20][21][22] The concept of a reverse micelle mode of action to transfer metal ions through water-oil interfaces by a range of surfactant ligands is now well established. [17,[23][24][25][26][27][28] Given that the extraction of any chloridometalate anion will compete with a massive excess of chloride, it is important that the extraction behavior of TBP with respect to hydrochloric acid is understood. Early work on the uptake of water and HCl by TBP (neat and in hydrocarbon solvents) is the subject of an excellent review by Osseo-Asare. [7] Chiarizia et al. [21,29] report small angle neutron scattering (SANS) data on HCl extractions by TBP into n-octane that provides a mechanistic explanation for third phase formation. Their data suggest the presence of reverse micelles that exist in solution as simple dimers that grow from aggregates containing two TBP molecules (when no HCl is present) to ({HCl} 3 {H 2 O} 3 · 7TBP) org , until the organic layer splits at 7.6 M HCl. This picture is somewhat at odds, however, with thermodynamic modeling of the TBP/HCl/water system by Lum et al., [30] which predicted that the trisolvate species (HCl{H 2 O} 7 3TBP) org will dominate at low acid concentrations, while the di-and mono-solvates (HCl{H 2 O} 6 2TBP) org and (HCl{H 2 O} 3 TBP) org become more prevalent as the acid concentration is increased. At HCl concentrations greater than 8 M, water is lost from the organic phase, and only a slight excess of water over HCl is observed, with a typical stoichiometry being ({HCl} 1.6 {H 2 O} 1.8 TBP) org .
While the extraction of metals by TBP has been the subject of many papers, there are gaps in knowledge that require attention. Its behavior as an extractant in toluene appears not to have been reported, and an in-depth investigation beyond the stoichiometries of components postulated to aggregate in the organic phase is missing. Information on the molecular structure, bonding, and the fundamental factors guiding the formation of the postulated macromolecular reverse micelles can be provided through molecular dynamics simulations, where the impact on the self-assembly of structures arising through systematic variation of the different components involved can be studied in depth. The focus of this article is therefore to probe more deeply the behavior of TBP for the extraction of HCl and PtCl 6 2into toluene. This study began with an analytical investigation to determine, as far as possible, the relative proportions of TBP, H 2 O, HCl, and PtCl 6 2that comprise the reverse micelles as a function of HCl concentration. These results were then used to guide computational modeling work, where self-assembly of reverse micelle structures was observed from components placed randomly in a toluene solvent box. Two HCl concentrations formed the basis for the computational investigation, 6 M HCl, since this is the acid concentration typically used for the commercial extraction of platinum group metals (PGMs) from HCl leach streams, [9] and 10 M HCl because the analytical measurements indicated a significant change in extractant behavior at very high acid concentration.

Extraction setup
All solvents and reagents were used as received from Sigma-Aldrich, Fisher Scientific UK, Alfa Aesar, Acros Organics or VWR International. Deionized water was obtained from a Milli-Q purification system. Tri-n-butyl phosphate (TBP) was dissolved in toluene to create stock solutions of 0.5 and 1 M. The HCl concentration was varied from 2 M to 12 M. As HCl is known to promote hydrolysis of TBP, [31] contact and mixing of the equi-volume organic and aqueous phases were kept to a minimum (15 minutes, stirrer plate) before being physically separated. Any entrained water in the organic phase was subsequently removed either by phase separation paper (Whatman 1PS), centrifuge or sonicator.

Water and acid content analysis
Because of the high excess of chloride (2-12 M HCl) present in the (0.01 M) PtCl 6 2extractions, and as TBP will co-extract both these species, it was not possible to determine the levels of H 2 O and H + associated with transport of just PtCl 6 2into the organic phase because the values are swamped by those involved in the extraction of HCl. Consequently, water and acid content analyses are only reported in the main text for the HCl extractions; data obtained in the presence of 0.01 M Na 2 PtCl 6 can be found in the electronic supplement S1.2.
Water content of the organic phase before and after extraction was measured by Karl Fischer titration, using a Metrohm 831 KF Coulometer charged with HYDRANAL ® Coulomat AG reagent. Background levels of H 2 O present in toluene following contact with 2-12 M HCl were also obtained to apply as a correction to the data when TBP was present.
The concentration of H + in the organic phase after extraction was determined by making up 1 ml aliquots to 5 ml in propanol, ensuring availability of sufficient volume to cover the Radiometer red rod REF201 universal electrode and Radiometer pHG 201/8 glass pH electrodes, against either standardized 0.1 M or 1 M NaOH solutions. The pH titrations were carried out using a Metrohm Titrando equipped with 800 Dosino dispensers.

Metal-content analysis
The concentration of Pt in the organic phase was determined using a Perkin Elmer Optima 5300DV inductively coupled plasma optical emission spectrometer. Sample solutions in 1-methoxy-2-propanol were taken up into a Gem Tip cross flow nebulizer and a Glass Cyclonic spray chamber at a rate of 2.0 mL min -1 and analyzed using a RF forward power of 1500 W with argon gas flows of 20, 1.4, and 0.45 L min -1 for plasma, auxiliary, and nebulizer flows, respectively. Resulting data were processed using WinLab32 for ICP-OES, version 3.0.0.0103. ICP-OES calibration standards for platinum in 10% hydrochloric acid were obtained from VWR International Ltd (UK).

Classical molecular dynamics simulations
All MD simulations were performed using the software package LAMMPS, [32] with the OPLS-AA force field [33] for all atom types with the exceptions of water (for which the TIP3P parameters were employed) [34] and PtCl 6 2-(for which custom parameters were derived for the partial charge (q), Van der Waals well depth (ε) and collision diameter (σ), see electronic supplement S2.4). Partial charges (Mulliken) for TBP, TBPH + , and H 3 O + were calculated at the B3LYP/6-31G(d) level of theory using Gaussian09, [35] in a similar fashion to that reported by Cui [36] and Wipff. [37] Initial models were constructed using Packmol [38] to fit specific numbers of desired molecules randomly inside a cubic box of length 40 Å. Models comprised 3 to 6 TBP molecules, 0 to 10 H 2 O molecules, 1 to 5 Cl -anions, along with the appropriate number of charge balancing protons located on either TBP or H 2 O molecules, and a sufficient number of toluene molecules to match the experimental solvent density at 298 K. Note the apparent low concentrations of components in the simulation box prevent multiple aggregates forming in the simulation, which would require a substantially larger periodic boundary unit cell. The xyz file was then converted into a LAMMPS data file using the VMD Topo tools. [39] Initial minimization was carried out on all models by iteratively adjusting the atomic coordinates to reach a local potential energy minimum. Simulations were then run under NVT ensemble conditions to achieve equilibrium for approximately 0.05 ns (in integration time steps of 0.1 fs using the standard Velocity-Verlet algorithm), followed by simulations using the NPT ensemble for a minimum of 9 ns (integration time steps of either 0.5 or 1 fs), thermostated at room temperature and pressure using the Nosé-Hoover thermonstat/barostat system. [40,41] Further details can be found in the electronic supplement.

Results and discussion
Extraction of HCL by TBP into toluene

Water content
As observed for other organic solvents, [31,42] varying the HCl concentration in an aqueous phase in contact with toluene affects the amount of water that is transferred into the organic phase.  Data points above 10 M HCl in Fig. 2 were adjusted to account for changes in the relative volumes of the two phases by using a metal tracer in the aqueous phase, and the amount of water transferred to the organic phase was corrected for background levels present in toluene (Electronic Supplement S1.1). Similarly shaped curves were obtained by Kertes [43] and Hardy [44] Fig. 2(a). [42] Acid content Results for H + extraction are displayed in Fig. 2(a), where minimal H + was detected in the organic phase upon contact with HCl below 6 M. Above 6 M HCl, [H + ] rises in a linear fashion. Similar behavior has been reported in other solvents. [31,[42][43][44] Water:acid ratios in loaded organic phase The ratio of H 2 O to HCl measured upon extraction with varying initial HCl concentration in the aqueous phase is shown in Fig. 2(b). The ratio falls rapidly, to levels of around 1:1 at 9-10 M HCl, dropping further thereafter. Two key concentrations of HCl are the focus of discussion below: 6 M HCl, which is the acid concentration typically used in the extraction of PGMs from HCl-leach streams, [9] where our analysis suggests HCl{H 2 O} 2.5 [ Fig. 2(b) red point], and 10 M HCl, where the concentration of H 2 O detected in the organic phase reached a maximum, and our analysis suggests HCl{H 2 O}. Our data match very closely the report by Lum et a.l for high acid extraction, [30] but our lower acid extraction model suggests a notably less hydrated core, which may arise due to the different solvents used.
Extraction of PtCl 6 2by TBP into toluene PtCl 6 2extraction The ability of TBP solutions in toluene to extract PtCl 6 2as a function of TBP and HCl concentration is summarized in Fig.3(a). Extractant performance of TBP is highly dependent on HCl concentration. Data reported here are comparable with those reported by Sun et al. [45] For extractant solutions containing more than 1 M TBP, the platinum loading decreases significantly as the initial concentration of HCl in the aqueous feed solution is increased. This behavior is typical for extraction of chloridometalates, which are competitively extracted alongside chloride anions. [9] Slope analysis TBP:PtCl 6 2ratios pertaining to the extractions at 6 M and 10 M HCl were obtained from analysis of the distribution ratio, D, where D = [Pt] org /[Pt] aq , and the gradient of a plot of log D against log [TBP] gives the ratio of the two components in the extracted species. The results of these analyses are given in Fig. 3 (b,c). At 10 M HCl, the straight line fit (R 2 = 0.992) suggests that two TBP molecules are associated with each PtCl 6 2ion in the organic phase. The curved form of the plot at 6 M HCl indicates that the extraction process involves the formation of more than one type of assembly in the organic phase. The curved form is not associated with experimental errors, as the Pt content of each phase was estimated to be subject to ≤3% error and no third phase was detected. While, as might be expected for a complex system in which aggregation of species in the organic phase is possible, the log D plots did not provide evidence for the formation of a single species under all conditions, they proved useful in providing guidelines for the range of stoichiometries (i.e., 2-5 TBP to PtCl 6 2-) that should be considered in modeling the formation of reverse-micelle type assemblies.

Molecular dynamics simulations
Models for extraction of Clfrom 6 M HCLl MD simulations were performed to establish the maximum number of TBP molecules that can assemble around one Clanion in the presence of a small amount of water. The number of water molecules was varied (0, 5, or 10) in order to directly assess the impact of water in the assemblies thus formed. Note that care needs to be taken not to over interpret the observations from the modeling work as, unlike the solvent extraction experiment, the periodic boundary condition model used does not have an interface to an aqueous layer. As such, any water molecules present in the simulation cell will naturally gravitate to the anion, whereas in the real extraction, the polar environment of the aqueous phase may act as a stronger attractant than the organic phase. To establish charge neutrality, an H + counter-ion must be present, which can be attached to either a TBP molecule or a H 2 O molecule. While it is more likely that the proton resides on a water molecule (and in all likelihood hops between water molecules), [46] we pursued both options in this work, as geometry optimization calculations [M06/6-31G(d)] on the system TBP. . ..H 3 O + resulted in the proton "hopping" to the TBP, leaving a hydrogen-bound water molecule, indicating that, in the gas phase in any case, TBPH + is a stable species. Also, within a small reverse micelle containing only a few "structured" water molecules H-bonded to an anion (which will reduce their basicity), the option of a proton residing on a TBP molecule might be more favorable than in a bulk phase. In all simulations, a single aggregate was observed to form spontaneously, where the Clanion was surrounded by a water shell, which in turn was surrounded by a TBP shell (see Figs. 4 and 5). The structure therefore conforms to what could be regarded as a reverse micelle. Regardless of whether the countercation is H 3 O + or TBPH + , it quickly became apparent that, while the maximum number of TBP molecules that could assemble around the chloride anion was four, assemblies containing two or three TBP molecules were far more common [see electronic supplement S2.1]. This finding is in line with the thermodynamic modeling, which suggested a ratio of HCl:3TBP at low acid concentration. [30] The simulations also appear to support the levels of micelle hydration identified by the Karl-Fischer titrations: in the complete absence of water, the number of associated TBP molecules most often found was just two, while increasing the number of water molecules (to ten) usually permitted four TBP molecules to aggregate. This suggests that one function of the water molecules is to provide a larger volume around the chloride anion with which extractant molecules can interact.
For the MD simulations comprising TBPH + as the counter-ion, simple electrostatics would predict this ion to lie closer to the Clanion than neutral TBP molecules, and this indeed was always found to be the case [ Fig. 4(b) and electronic supplement S2.1]. The Cl -. . . O(TBPH + ) distance, 3.05 Å, is recorded as a sharp peak on the g(r) plot, which indicates that this is a strong interaction that varies little throughout the simulation. In contrast, the Cl -. . . O(TBP) distances form a broader continuum, centered around 3.6 Å. Optimization of the aggregate (M06/6-31G(d) level, gas phase model) gave similar distances: Cl -. . . O(TBPH + ) = 2.8 Å, Cl -. . . O (TBP) = 3.0 Å, which lends credence to the force-field parameters used. The water molecules are most commonly located between the chloride ion and the TBP molecules but never disrupt the Cl -. . . O(TBPH + ) ion pair. The model that emerges is thus one where the water molecules act as mediating bridges between the negatively charged chloride ion and electronegative oxygen atoms in the P = O group of the TBP molecules, while the positively charged (BuO) 3 P = OH + unit forms the ion pair contact, as summarized in Fig. 4(c).
Information on the size of the reverse micelle formed can be obtained from atomic density plots [ Fig. 4(d)], where the coordinates of all atoms bar the toluene solvent molecules from each frame of the MD simulation are shifted to ensure that the center of mass lies at the center of the box, and the positions of atom types are binned to generate probability distribution plots. The chloride anion does not sit at the center of mass, implying mobility of the anion within the aggregate toward the TBPH + cation and/or a shifting center of mass due to flexibility of the hydrocarbon chains on the TBP molecules. The probability density plots allow the sizes of each of the approximately Gaussian concentric shells making up the aggregate to be determined. This gives a maximum radius of 5 Å for the chloride core, which rises to 8-10 Å when the water and TBP layers are included. This closely matches the diameter (ca. 18 Å) reported by small-angle neutron and X-ray scattering (SANS and SAXS) for aggregates formed on extraction of HCl by TBP into n-dodecane [20] and n-octane. [47] Switching the countercation to H 3 O + has a marked effect on the structures obtained. While the most commonly observed aggregates again contain two or three TBP molecules (Electronic Supplement S2.1), the shape of the aggregate changes from an approximately spherical arrangement to a more prolate structure [ Fig. 5(a)   simulation box. The schematic diagram in Fig. 5(c) shows that the short contact distance relates to binding of the charged hydronium ion, which in turn "pins" two TBP molecules to the hydrated core, while the longer distance relates to the neutral H 2 O molecules that also surround the chloride anion. The atomic density plot [ Fig. 5(d)] appears to suggest a lack of coherent structure, with each component equally dispersed within the aggregate, in marked contrast to that described from the simulation where the counter-ion is TBPH + . However, as this analysis is essentially a histogram of atomic positions as a function of distance from the centre of mass, averaged over many trajectory snapshots, it will not perform well for randomly orientated structures that deviate substantially from a sphere.
To establish how well TBP molecules encapsulate a hydrated chloride anion, 50 random frames were selected from the production run trajectories that contained three TBP molecules, five H 2 O molecules, and one Clion, for both the TBPH + and H 3 O + counter-ion simulations. The amount of core region exposed to the surrounding solvent (the percentage core exposure) can be calculated via a Monte Carlo script that shoots a probe point from 10,000 random positions on the surface of a sphere (which is centered on the chloride ion and which completely encapsulates the aggregate) toward the chloride ion and counts what atom type it hits first on its path; the number of times the probe hits an atom in a water molecule or a Clanion (represented as spheres of appropriate van der Waal radii) is then recorded as a percentage of the total probe runs. Low core-exposure values therefore indicate that the water molecules are less exposed. The analysis was repeated for each of the 50 random frames, which allows an average and standard deviation to be reported (see Table 1) and reveals core exposure values of 53(2)% when the counter-ion in the simulation was TBPH + , compared to 57(5)% for counter-ion H 3 O + . In both cases, the core is therefore relatively exposed.
The picture that emerges from the computational modeling work is one where stable assemblies readily form that comprise three TBP molecules to one chloride anion if the number of water molecules included in the aggregate is below ten, which finds good agreement with the conclusions reached by Lum et al. [30] The size distribution obtained for the more spherical aggregates formed when the counter-ion is TBPH + is also in excellent agreement with SANS and SAXS measurements.- [20,47] Turning to the details that the atomistic simulations can provide, the new insight gained relates to direct observation of water molecules acting as a mediating shell between (nBuO) 3 P δ+ = O δand Cl -, which would otherwise be a repulsive interaction. The observed structures also have relatively exposed core regions, indicating that rather "leaky" reverse micelles have formed. If the TBP carries the proton, the structure formed is essentially a reasonably spherical hydrated ion-pair, with a welldefined water shell; if the proton shifts to a water molecule, the structure becomes markedly more prolate.
Models for extraction of Clfrom 10 M HCL The analytical measurements reported in Fig. 2 suggest a substantial change in mode of action for extraction of HCl with increasing acid concentration, with likely ratios of core components fitting the empirical formula (H 2 O.HCl) y at 10 M HCl, in line with the thermodynamic modeling work of Lum et al. [30] MD simulations were explored with equimolar quantities (3, 4, and 5 molecules) of TBP, H 2 O, and Cl -, with counter-ions of either TBPH + or H 3 O + , as before. In all simulations, a single aggregate was observed to form spontaneously, containing all of the TBP present (Electronic Supplement S2.2). The nature of the counter-ion was again observed to have a marked effect on the structures formed, as demonstrated by the atom probability plots in Fig. 6, which show a noticeably more diffuse water structure when protonated. The simulations with TBPH + counter-ions return aggregates of very similar size for structures containing 3-5 Clions, with encapsulated hydrated chloride core diameters of about 15 Å, which rises to around 20 Å including the TBP shell. Aggregates appear to be larger (diameters ca. 25 Å) when the counter-ion is H 3 O + , and the atom probability plots show evidence of layered structures for cores comprising 4 or 5 Clions, which suggests a return to a more spherical structure. Chiarizia et al. have performed SANS measurements for TBP extraction of HCl into n-octane and concluded that, upon increasing acid concentration, the aggregates swell by about 4.5 Å and have polar core diameters of around 10-15 Å. [48] This compares favorably with the results obtained in our study.  Fig. 6. In essence, this suggests a model comprising a hydrated chloride core where the Clions protrude when surrounded by a charged TBPH + shell, or the H 3 O + cations protrude when attracted by the P = O δgroup from a shell of charge-neutral TBP molecules. These structural characteristics are displayed schematically alongside the corresponding g(r) plots in Figs. 7 and 8. Other noteworthy features of the g(r) plots are that Clbinds tightly to both H 3 O + and TBPH + (sharp peaks), but while H 3 O + also binds tightly to TBP (2.55 Å), H 2 O appears to cluster more loosely around TBPH + ions.   4 aggregates, with counter-ions TBPH + , obtained over the production run trajectories, with general schematic diagram highlighting the main structural details shown alongside. Integrating the g(r) plots (red lines) permit the number of interactions, as a function of distance between the denoted atom pair, to be counted. The tight-binding of H 3 O + is mirrored in RMSD plots, which suggest that the (H 3 O + .Cl -) 3-5 core is always more rigid than a (H 2 O.Cl -) 3-5 core (Electronic supplement, S2.2). Interestingly, a superconcentrated HCl water pool has been reported in the extraction of PtCl 6 2by a tertiary amine ligand in a reverse micelle mechanism. [49] The core exposure measurements (Table 1), averaged over 50 snapshots selected at random from each MD production run trajectory, suggest the hydrated chloride core is shielded more effectively as the number of bound TBP surfactant molecules rises. It is tempting to attribute the relatively exposed hydrated cores to the cause of third-phase formation reported for TBP in solvent extraction processes under high acid or high metal loading conditions, [29] as aggregates may merge to combine exposed water pools, which would lead to larger assemblies carrying increasing amounts of water into the organic phase, resulting in the solvent interface breaking down. SANS measurements for extractant aggregation have typically been interpreted using the Baxter model of sticky hard spheres,- [50] suggesting the existence of a short-range attractive potential between aggregates. Recent measurements by diffusion NMR spectroscopy on the extraction of Zr(NO 3 ) 4 by TBP [22] have suggested that long-range repulsive interactions can also arise if the negatively charged core region is exposed; the modeling work performed here clearly supports a fluxional aggregate with significantly exposed core region.
Finally, we explored the effect of increasing the number of TBP molecules (up to eight) available to aggregate around a core of four Clions and four water molecules. Regardless of the nature of the counter cation, results indicate that it is unusual for more than four TBP molecules to interact with the core (electronic supplement S2.3), which is at odds with Chiarizia's high acid concentrations model (which has seven TBP molecules around a (H 3 O + .Cl -) 3 core), [29] but does support the thermodynamically postulated structures by Lum et al, which propose a lower chloride to TBP ratio at high acid concentration. [30] Models for extraction of PtCl 6 2from 6 M HCL Given that the Karl Fischer titrations could not be used to ascertain the quantity of water involved in the extraction species (due to the simultaneous extraction of Cl -, which pulls substantial quantities of water into the organic phase, and is present in vast excess compared to PtCl 6 2-, see electronic supplement S1.2), in this work, the impact of any water associated with the aggregates has instead been probed via computational modeling. The number of TBP molecules present in the toluene simulation box was varied from four to six, and the number of water molecules present was either 0, 2, 5, or 10. With the anion acting as the nucleation point now carrying a 2-charge, two cations are now needed to ensure neutrality. In this study, the impact of assigning the counter-ions as TBPH + or H 3 O + was assessed as these mark the two extreme cases. Note also that a reverse micelle could contain both PtCl 6 2and Cl -, but we have not explored this possibility here. With the counter-ions set to TBPH + , modeling indicates that four TBP molecules can be comfortably accommodated around a single PtCl 6 2anion in the absence of any water molecules (see electronic supplement S2.5), which is in line with the slope-analysis measurements reported above. Increasing the number of TBP molecules to six indicated that a fifth molecule can be incorporated into an aggregate, but only transiently. Introducing five H 2 O molecules into the simulation box gives a similar result: the most common number of coordinating TBP molecules is still four. Increasing still further (to ten H 2 O) increases the frequency of a fifth interacting TBP molecule, but this is still in a transient fashion. In contrast, if both counter-ions are set to H 3 O + , it is only when a total of ten water molecules are present that association of four neutral TBP molecules can be maintained (electronic supplement S2.5). Core exposure analysis indicates that around 20% of the water pool is exposed in these reverse micelles ( Table 1).
As  Fig. 9(a) and 9(b)]. The atomic density plots [ Fig. 9(c)] show that in the presence of water, a spherical structure is seen comprising distinct shells with the PtCl 6 2unit found close to the center of mass of the aggregate. The diameter of the aggregate including the TBP shell is approximately 25 Å [ Fig. 9(c)], which decreases slightly on removal of the water molecules (electronic supplement S2.5). Thus, it is clear that these assemblies are notably larger than those formed on extraction of single chloride ions. As with the chloride extraction models discussed earlier, if water molecules are present, they effectively negate repulsive interactions between the negatively charged anion and the negative dipoles on the oxygen atoms of the surfactant TBP molecules through hydrogen bonding. One notable difference compared to the chloride reverse micelle model is that water molecules are more likely to disrupt the TBPH +. . . PtCl 6 2-"ion-pair" interactions, which presumably relates to the more diffuse nature of the larger anion. The loose-binding of the water molecules to the anion is neatly captured in the g(r) plots [contrast PtCl 6 2- Fig. 9(a) against the corresponding more tightly bound Cl -. . .H 2 O in Fig. 5(b)].
The analogous data obtained for the PtCl 6 2extraction model when the counter-ions were set to H 3 O + are presented in Fig. 10. Both H 3 O + counter-ions coordinate firmly to opposing octahedral faces of the PtCl 6 2anion and act to "pin" two TBP molecules in place, with the remaining eight water molecules forming a hydrogen-bonded cluster close to the PtCl 6 2anion [see Fig. 10(b)]. This assembly forms early in the production run trajectory, with the only variant being the migration of one TBP molecule along one or two positions of the water cluster. The atom probability distribution Figure 9. (a) Plots of g(r) for the aggregate formed by 4 TBP molecules around a core of (H 2 O) 10 (PtCl 6 2-), obtained from the production run trajectory, (b) representative structure (water molecules are contained within blue shell), and (c) atom probability plot, where the counter-ions are TBPH + . Integrating the g(r) plots (red lines) permit the number of interactions, as a function of distance between the denoted atom pair, to be counted. plot shown in Fig. 10(c) indicates a highly structured reverse micelle structure has formed, which is a little larger than the structure formed by the TBPH + counter-ions.
Models for extraction of PtCl 6 2from 10 M HCL Evidence from the slope-analysis measurements suggest a change in the mode of action from (PtCl 6 {H 3 O} 2 {H 2 O} x TBP 4 ) org at 6 M HCl to (PtCl 6 {H 3 O} 2 {H 2 O} x TBP 2 ) org at 10 M HCl. From the modeling work reported in the previous section, it is apparent that four TBP molecules can be readily accommodated around a solvated PtCl 6 2core, and thus the only way to drop the PtCl 6 2-: TBP ratio is to explore higher order aggregates containing multiple PtCl 6 2anions, or that the mode of action is changed to a simple ion pairing. This latter option seems unlikely from the modeling work, as in the presence of excess TBP, it is rare for the coordination number to drop below four (electronic supplement, S2.5). For the former option, the models quickly become large, however, and in the absence of any further experimental data to establish the size of aggregates formed, a preliminary calculation was performed on a simulation box containing two PtCl 6 2anions only with four TBPH + and two TBP molecules (i.e., two TBP molecules potentially in excess) and six water molecules (arbitrary choice). While an aggregate quickly formed, it did not conform to the sought-after PtCl 6 2-:2TBP ratio, as two of the TBPH + ligands were observed to straddle the two PtCl 6 2anions, which in turn were each bound by one TBPH + and one TBP unit. Thus, the number of interacting TBP ligands per PtCl 6 2remained four, suggesting that higher-order aggregates, containing three or more PtCl 6 2anions, need to be explored. The possibility that two PtCl 6 2anions can be extracted by two TBP with two H 3 O + counter-ions cannot be ruled out, nor the possibility that PtCl 6 2and Clare co-extracted in the same assembly. It is therefore apparent that further experimental investigation is required to guide the computational study any further. Figure 10. (a) Plots of g(r) for the aggregate formed by 4 TBP molecules around a core of (H 2 O) 10 (PtCl 6 2-), obtained from the production run trajectory, (b) general schematic structure (water molecules are contained within blue shell), and (c) atom probability plot. Counter-ions are H 3 O + . Integrating the g(r) plots (red lines) permit the number of interactions, as a function of distance between the denoted atom pair, to be counted.

Conclusions
This article has probed TBP mode of action in the extraction of Cland PtCl 6 2from acidic chloride solutions. To the best of our knowledge, the experimental results presented are the first report of TBP performance in toluene and suggest the absence of any third-phase formation, which has hampered studies in other hydrocarbon solvents. Classical molecular dynamics simulations performed on models constructed from experimentally derived ratios of HCl and H 2 O resulted in the spontaneous formation of assemblies that can be described as reverse micelles. The diameter of chloride-containing micelles (16-18 Å for the 6 M HCl extraction model, rising to ca. 20-22 Å for the models explored for the 10 M HCl extraction model) matches well with data reported from SANS experiments, giving validity to the structures described. Modeling suggests that the experimental component ratios expressed at 6 M HCl can be best accommodated by a structure derived from two to three TBP molecules interacting with one chloride anion, which is surrounded by no more than ten water molecules. For the 10 M HCl extraction data, modeling indicates that reverse micelles containing multiple chloride anions (we explored up to five), with matching ratios of TBP and water molecules, all gave stable assemblies. For PtCl 6 2extraction, slope analysis indicated that more than one type of assembly can form in the organic phase and that four or more TBP molecules can be associated with each PtCl 6 2at 6 M HCl, which was matched by the computational modeling. At 10 M HCl, the experimental data indicated that the number of coordinating TBP molecules to each PtCl 6 2falls to two, suggesting the formation of a larger aggregate that the modeling work suggests will contain more than two PtCl 6 2ions. The role of water in the extraction of PtCl 6 2is unclear, but it appears to be less important than for the extraction of Cl -, with simulations devoid of any water molecules still capable of attracting the expected four TBP molecules as observed experimentally under 6 M HCl extraction conditions. The chloridoplatinate-containing micelles are significantly larger than the chloride-containing structures.
The MD simulations applied in this work prevent the spontaneous "hopping" of H + ions between molecular units, and so by pursuing both H 3 O + or TBPH + as counter-ions, the two extremes of behavior have been explored. Results indicate that while both options result in stable assembly formation, those constructed from H 3 O + counter-ions give more rigid hydration core structures. This supports the concept of a highly concentrated acidic core, as suggested by earlier literature reports. [49] The modeling work also reveals that the observed levels of TBP aggregation result in structures with relatively exposed hydrated core regions, which provides insight into why thirdphase formation occurs in solvent extraction processes.