Ni Enrichment and Stability of Al-Free Garnierite Solid-Solutions: A Thermodynamic Approach

Garnierites represent significant Ni ore minerals in the many Ni-laterite deposits worldwide. The occurrence of a variety of garnierite minerals with variable Ni content poses questions about the conditions of their formation. From an aqueous-solution equilibrium thermodynamic point of view, the present study examines the conditions that favor the precipitation of a particular garnierite phase and the mechanism of Ni-enrichment, and gives an explanation to the temporal and spatial succession of different garnierite minerals in Ni-laterite deposits. The chemical and structural characterization of garnierite minerals from many nickel laterite deposits around the world show that this group of minerals is formed essentially by an intimate intermixing of three Mg-Ni phyllosilicate solid solutions: serpentine-népouite, kerolite-pimelite, and sepiolite-falcondoite, without or with very small amounts of Al in their composition. The present study deals with garnierites which are essentially Al-free. The published experimental dissolution constants for Mg end-members of the above solid solutions and the calculated constants for pure Ni end-members were used to calculate Lippmann diagrams for the three solid solutions, on the assumption that they are ideal. With the help of these diagrams, congruent dissolution of Ni-poor primary minerals, followed by equilibrium precipitation of Ni-rich secondary phyllosilicates, is proposed as an efficient mechanism for Ni supergene enrichment in the laterite profile. The stability fields of the solid solutions were constructed using [{iexxxsbap19011700254003}] (predominance) diagrams. These, combined with Lippmann diagrams, give an almost complete chemical characterization of the solution and the precipitating phase(s) in equilibrium. The temporal and spatial succession of hydrous Mg- Ni phyllosilicates encountered in Ni-laterite deposits is explained by the small mobility of silica and the increase in its activity.


INTRODUCTION
Nickel laterites are regolith materials derived from ultramafic rocks (Trescases, 1975;Golightly, 1981;Brand et al., 1998). Weathering of the rocks results in enriched horizons, so that minor elements such as Ni, Co, and Mn contained in the unaltered parent rock become enriched in the laterite profiles (Brand et al., 1998;Freyssinet et al., 2005;Golightly, 2010). An iron cap (ferricrete or duricrust) is often found at the top of the weathering profile and the cap grades downward through a transitional zone of limonite to a saprolite zone transitional into bed rock. Saprolite is a zone of rapid change and of maximum supergene enrichment of nickel. The protolith or bedrock is mainly dunite, harzburgite, or lherzolite and their serpentinized equivalents.
Following Golightly (1981), Ni laterites are developed from: (1) unserpentinized peridotite (e.g. Poro, New Caledonia; Soroako West, Indonesia); (2) partially serpentinized peridotite (e.g. Falcondo Mine, Dominican Republic); or (3) totally serpentinized parent rock (e.g. Bonsora, Soroako East, Indonesia). As an example, selected analyses for the most important primary Ni-carrying minerals at Falcondo (Table 1) indicate that these minerals are mainly olivine and 'oceanic' serpentine. The fate of Ni can be tracked through the different values of the atomic ratio of Fe/Ni in these minerals. Initially, in olivine, Fe/Ni & 19. In oceanic serpentines this value is slightly smaller, indicating that some of the Fe is relocated in other phases like maghemite, which is readily hydrated to goethite. Goethite is able to incorporate or attach Ni (Manceau et al., 2003). The nickel contained in the limonite horizon and in primary serpentines is transferred to percolating meteoric solutions, and moves downward through the profile, being concentrated with Si and Mg within the underlying saprolite horizon to form secondary Ni-enriched phyllosilicates.
In many Ni lateritic deposits of the hydrous silicate type, the lower saprolite is the ore horizon, and the ore minerals are mainly nickeloan varieties of serpentine, talc, sepiolite, chlorite, or smectites, many of which are poorly defined and informally named ''garnierites'' (e.g. Brindley and Hang, 1973;Springer, 1974;Pelletier, 1983). Garnierite is generally used as a field term referring to a wide variety of Ni-bearing hydrous silicates with a characteristic green color (e.g. Gleeson et al., 2003). When only small amounts of Al are present in the protolith, garnierites can be formed by different combinations of the three main Mg-Ni solid solutions with serpentine, talc, and sepiolite-like structures; clays (chlorites and smectites) being absent (Wells et al., 2009;Tauler et al., 2009;Villanova-de-Benavent et al., 2011a).
Garnierites occur filling fractures, featureless voids, and cavities or as interstitial veins in the joints separating serpentine blocks, most commonly in the lower part of the saprolite horizon. Often, they have been removed from their original emplacement and have accumulated in cavities, forming breccias (Cluzel and Vigier, 2008). The most common and most studied occurrences of garnierites are found in the less serpentinized ophiolite-related peridotites of mantle origin as in New Caledonia (Trescases, 1975;Wells et al., 2009) and in the Caribbean (Gleeson et al., 2003;Lewis et al., 2006).
The progressive enrichment of Ni in the saprolite horizon occurs by means of two different processes. The dominant process is the direct replacement of preweathering minerals, such as olivine and serpentine, by NiFe-enriched lizardites and by Ni-rich smectites. In the minerals formed by replacement, Fe and Ni increase at a percentage range (see Table 1). The second process involves the precipitation of Mg-Ni phyllosilicates (Al-free garnierites) from aqueous solution. As described above, the garnierites form as fracture fillings, veins, and breccias. In contrast to the minerals formed by replacement, the garnierites are characterized by low Fe and are essentially free of Al. These garnierite minerals and the conditions of the precipitation process by which they form are the subject of the present study.
Here the authors explain, using equilibrium thermodynamics, three related, simultaneous phenomena: (1) supergene enrichment of Ni in a group of secondary phyllosilicates precipitating in Ni-laterite profiles À Lippmann's diagrams (Lippmann, 1980) were found to be very helpful for this; (2) simultaneous precipitation of Mg-Ni-phyllosilicate solid solutions (in theory, up to three different s.s. Ô quartz) À this can be represented by an activity diagram proposed below; and (3) the spatial and temporal evolution of garnierites towards more silica-rich members, from serpentine to talc and sepiolite-type minerals, as has been described previously (e.g. Golightly, 1981;Freyssinet et al., 2005). A simplified chemical system which is useful in the description of the stability relations among the Ni-containing phyllosilicate phases consists of four components: MgO, NiO, SiO 2 , and H 2 O. For aqueous solutions the basic species are: Mg 2+ , Ni 2+ , SiO 2 (aq), H + , and H 2 O. As garnierites are Fe-poor (<0.5 wt.% Fe) indicating a secondary origin (Pelletier, 1983;Proenza et al., 2008;Wells et al., 1999), Fe was not included in the system. Its inclusion would increase the complexity of the system without affording significant new information regarding the distribution of Ni among the main phases present. After weathering, residual Fe of the parent rock is concentrated in a highly insoluble limonitic horizon, formed mainly of goethite, hematite, and maghemite. This horizon is best defined by the complete disappearance of the primary silicates and is distinct from the heterogeneous saprolite below. The Mg-Ni-phyllosilicates are members of three solid-solution series extending from Mg and Ni endmembers (e.g. Springer, 1974Springer, , 1976Brindley et al., 1979;Gleeson et al., 2003;Tauler et al., 2009;Wells et al., 2009;Reddy et al., 2009;Villanova-de-Benavent et al., 2011b). The electron microprobe (EMP) analyses plotted in Figure 1 confirm this; a complete, stable solid solution between Mg and Ni end-members in the structures of these minerals must be possible at atmospheric temperatures and pressure.
Therefore, a complete solid solution between Mg and Ni end-members is a reasonable hypothesis. The terminology used here for the end-members, composition, and structure for the solid solutions are: serpentine ( (Brindley and Hang, 1973;Brindley and Maksimovic, 1974;Springer, 1974;Brindley and Wan, 1975;Brindley et al., 1979). Serpentine in saprolite can be present as a mixture of different structural varieties such as chrysotile, clinochrysotile, polygonal serpentine, and lizardite-1T. Indeed, in a single nanoscale particle, the transition between the different structural varieties can be observed using transmission electron microscopy (TEM) (Villanova-de-Benavent et al., 2011a). Kerolite and Ni-rich kerolite, a very fine-grained, hydrated variety of talc, is also present (Brindley et al., 1979;Soler et al., 2008).
Here, the published experimental dissolution constants for Mg end-members and the calculated constants for pure Ni end-members were used to calculate Lippmann diagrams for the serpentine-népouite, kerolite-pimelite, and sepiolite-falcondoite (Srp-Nep, Ker-Pim, and Sep-Fal) solid solutions, on the assumption that they are ideal. In addition a particular activity diagram for the system MgO-NiO-SiO 2 -H 2 O, delimiting the stability field of each Ni-bearing phyllosilicate and quartz, is proposed. Garnierites, as described below, can precipitate from aqueous solution in the form of one or two solid solutions, under near equilibrium conditions.

THEORY
The basic reactions involved in the formation or dissolution of Mg and Ni end-member silicates are given in Table 2. In the system MgO-NiO-SiO 2 -H 2 O, under constant temperature and pressure, the maximum number of phases present in equilibrium is four. In the presence of an aqueous phase, three solid phases are in equilibrium for zero degrees of freedom. Hence, in garnierites, a maximum of either three solid solutions or two solid solutions plus quartz is expected to precipitate from solution under near-equilibrium conditions. However, the most frequent observation is the existence of equilibrium between two solids (two solid solutions or one solid solution and quartz), or simply a single solid Figure 1. Projection of Caribbean garnierite ore compositions (Proenza et al., 2008;Tauler et al., 2009;Soler et al., 2010;Villanovade-Benavent et al., 2011a, 2011b and published data (Brindley and Hang, 1973;Brindley, and Wan, 1975;Springer, 1974Springer, , 1976 for garnierites from different Ni-laterite deposits, in a Si-Ni-Mg ternary diagram (at.%). The positions of the solid-solution series between Ni-sepiolite (Sep)-falcondoite (Fal), kerolite (Ker)-pimelite (Pim), and lizardite (Srp)-népouite (Nep) are also indicated. that is a solid solution or quartz (e.g. Vitovskaya et al., 1969;Brindley and Hang, 1973;Springer, 1974;Golightly, 1981;Soler et al., 2008, and examples in the present study).
With the given reactions (Table 2), all possible combinations of phases can be examined in light of thermodynamic considerations, provided that the corresponding equilibrium constants are known. A brief account of the equilibrium constants used was necessary, before the different combinations of phases in the system were reviewed. The most problematic choice was for the equilibrium constant of reaction (1) ( Table 2), (log K1 value). Several log K constants for the dissolution of serpentine are found in the literature (Table 3), and most are controversial because of the different crystal structures of serpentines and the close proximity of these structures in the same sample (Evans, 2004). The various equilibrium constants were calculated from the respective free energies of formation (Table 4). The value of log K1 (for reaction 1) was constrained by the experimental observation that at equilibrium Ni-kerolitie was more enriched in Ni than Ni-serpentine. The log K1 value that best fits the observed compositions of this equilibrium is the one derived from DG serpentine given by Bricker et al. (1973). Constant values for Mg endmembers of kerolite and sepiolite (log K3 and log K5)  (6) Quartz: SiO 2 (quartz) = SiO 2 (aq)  Nriagu (1975) * These constants correspond to the formula of sepiolite used by Stoessell, based on two octahedral cations. The underlined log K values were used in the present work.
were taken from dissolution experiments (Stoessell, 1988). The constants for the Ni end-members (népouite, pimelite, and falcondoite) were derived from DG f values calculated by the method proposed by Nriagu (1975), (log K2, log K4, and log K6). Log K values predicted by this model are in good agreement with available experimental data on Mg phases, encouraging the extension of the method to inaccessible phases such as pure Ni phyllosilicate end-members. The error estimated, <1 unit in log K, does not impede the evaluation of the relative stabilities of these phases.

Two-phase equilibria (one solid solution and one aqueous solution)
The simplest situation is when only one Mg-Niphyllosilicate solid solution precipitates from an aqueous solution. Equilibria between one solid solution and the aqueous solution were plotted in a Lippmann phase diagram (Lippmann, 1980;Glynn and Reardon, 1990;Prieto, 2009). In these diagrams, the aqueous solution in equilibrium with two (or more) end-member components of a solid solution is described by the ''total solubility product'' SPeq, (or logSPeq), represented on the ordinate, whereas on the abscissa two different variables are depicted: the mole fraction in the solid solution (represented here by X Mg ) and the activity fraction in the aqueous solution, X Mg 2+. SPeq as a function of X Mg and X Mg 2+ defines the 'solidus' and the 'solutus' curves, respectively. In equilibrium, every SPeq corresponds to an X Mg value and an X Mg 2+ value (both connected by a horizontal tie-line). An exception is made for pure endmembers or alyotropic points where the two values coincide, or when a miscibility gap exists in the solid solution. In general, the variable SP defines a state of supersaturation, equilibrium, or undersaturation. When SP lies above the solutus curve, the solution is supersaturated with respect to a series of solid compositions; below the solutus, the solution is undersaturated with respect to all possible compositions of the solid solution.
On the same diagrams, 'minimum stoichiometric saturation curves', MSS, can be depicted. These curves are useful for illustrating the relations between the composition of a given solid solution and the SP values of the solution when a congruent dissolution occurs. For mineral solid solutions with low solubility, dissolution that occurs congruently is expected, whereas precipitation is likely to occur in equilibrium or under supersaturation conditions. As discussed below, Lippmann diagrams can help to portray the supergene enrichment of Ni in phyllosilicates.
Lippmann diagrams (Figure 2a, 2b, 2c) for Mg-Ni solid solution in serpentine, kerolite, and sepiolite were calculated on the basis of the log K values given above and assuming ideal solid solutions. Note that these curves were calculated for the reduced formula unit of the minerals involved, so that only one octahedral atom enters the formula unit. This amounts to dividing the conventional formulae corresponding to serpentinenépouite and kerolite-pimelite solid solutions by three and dividing those corresponding to the sepiolitefalcondoite series by eight. Log K values for the dissolution reactions were also divided by the corresponding factors. The main reason for using reduced formulae was that the authors applied a 'mixing on sites' model to solid solutions rather than a molecular mixing model. With the reduced formulae, the equilibrium constants for the dissolution reactions and ionic activity product (IAP) for solutions are simpler and the exponents on the activities of substituting ions are unnecessary. The mole fractions in the solid, X Mg , and the activity fractions in solution, X Mg 2+, represented on the abscissa are given by: where N Mg and N Ni represent moles of Mg and Ni in the solid, and a Mg 2+ and a Ni 2+ are the activities of Mg 2+ and Ni 2+ , respectively, in solution.
The large contrast between the log K values corresponding to end-members of the solutions is reflected in the 'rectangular' rather than 'narrow loop' Table 4. Free energies of formation used in the approximations of DG f of the minerals listed in Table 3 and in the derivation of the corresponding log K of dissolution (kJ mol À1 ). À833.70 0.5 Phillips et al. (1988) aspect to the solidus and solutus curves (Figure 2a, 2b, 2c). The contrast also explains the efficiency of the supergene enrichment by the mechanism of congruent dissolution-equilibrium precipitation.
In order to use the calculated Lippmann diagrams in a semi-quantitative approach, the possible initial activity ratios of Mg 2+ and Ni 2+ in aqueous solution are compared with the resulting mole fractions of these metals in the enriched phases. The following example explains Ni enrichment using the Lippmann diagram ( Figure 2a). The values of the ratios Ni 2+ /(Mg 2+ + Ni 2+ ) in solution from congruent dissolution of primary serpentine ranges between 10 À3 and 10 À5 or less. The corresponding tie lines between the solutus and the solidus give the composition of the enriched precipitating solid solution with compositions in the range Mg 0.4À2.4 Ni 2.6À0.6 Si 2 O 5 (OH) 4 , which reflects much more Ni enrichment compared to the primary serpentine. These compositions are reasonable and representative of the amount of Ni incorporated into a neoformed phase in a single step of the dissolution-precipitation cycle. The repetition of several cycles of congruent dissolutionequilibrium precipitation depends on the hydrological regime in the soil (i.e. alternating dry and wet periods, fluctuations in the level of water table, etc.), and would produce more and more enriched phases. From one cycle to the next, the enriched precipitating phase type may change depending on the variation in silica activity that will determine the enriched precipitating phase type. For example, if the dissolving phase is a Ni-serpentine and the silica activity increases, the next precipitating phase may be a Ni-kerolite and so on.
For serpentine-népouite: For kerolite-pimelite: For sepiolite-falcondoite: where X Mg and (1ÀX Mg ) are the mole-fractions of Mg and Ni in the solids. Note that K 1 , K 2 , ...K 6 represent the equilibrium constants for the reduced formulae of the end-members, as indicated above. For every point in this diagram, three straight lines cross, corresponding to three possible solid solutions in equilibrium with the solution. However, apart from the boundaries between different phases, only one solid is in equilibrium with the aqueous solution, the other two candidates being undersaturated. Each solid solution (Srp-Nep, Ker-Pim, or Sep-Fal) is defined by a field in t h e ( p r e d o m i n a n c e a r e a ) a c t i v i t y d i a g r a m [log(a SiO 2 (aq) ), log((a Mg 2+ + a Ni 2+) /(a H + ) 2 )](see, for instance, Figure 3).

Three-phase equilibria (two solid-solutions and one aqueous solution)
Several authors have described garnierites as an intimate intermixing of two or more Ni-containing phases. The most common association encountered was serpentine-kerolite solid solutions (Vitovskaya et al., 1969;Brindley and Hang, 1973;Wells et al., 2009), but other associations such as chlorite-kerolite, with or without quartz, and a sample composed of sepiolite  plus minor serpentine and talc were described by Springer (1974). At a micrometric scale, all these combinations of mineral Ni-bearing phases in Caribbean Ni-laterites have been observed by the present authors. However, not all mixed particulate masses of garnierite are the result of an equilibrium precipitation, so it is important to distinguish the associations that occur as a result of mechanical mixing in breccias or accumulations in fractures, from those that are the result of co-precipitation at near-equilibrium conditions.
Simultaneous precipitation of serpentine-népouite and kerolite-pimelite solid solutions from aqueous solution When serpentine-népouite and kerolite-pimelite solid solutions precipitate from an aqueous solution in equilibrium, the equilibrium reactions 1, 2, 3, and 4 ( and its equilibrium constant is: The path of the solution in equilibrium with the two solid solutions is a curve (SK) that connects the intercept of serpentine-kerolite with that of népouite-pimelite ( Figure 3). One section of this curve presents a metastable extension (dashed line, Figure 3). The coordinates of this curve permit the calculation of the total solubility product (SP) of different minerals in the system MgO-NiO-SiO 2 -H 2 O, and this helps in the characterization of the solutions, in combination with the Lippmann diagrams. For example, the logSP values corresponding to sepiolite-falcondoite solid solution (equation 4) were plotted on the Lippmann diagram ( Figure 2c). All the terms of the Sep-Fal s.s. are undersaturated for all equilibria Srp-Nep+Ker-Pim solid solutions, with the exception of the end near the serpentine-kerolite intercept ( Figure 5).
An example of the use of the activity diagram (Figure 3), combined with Lippmann diagrams (Figures 2a and 2b), is as follows. Points falling to the left of the SK curve characterize solutions that will precipitate Srp-Nep s.s. For every point in this field, the logSP value for Srp-Nep and Ker+Pim can be calculated using equations 2 and 3. For a given value of logSP serpentine , the corresponding value of a Mg 2+/(a Mg 2+ + a Ni 2+) is given by the solutus in the corresponding Lippmann diagram. The same activity ratio applied to the Lippmann diagram for kerolite will plot under the solutus, indicating that kerolite s.s. composition is undersaturated. For example, for a point with coordinates (À5.5, 12) (Figure 3), a Srp-Nep crystal is in equilibrium with a solution with an activity ratio a Mg 2+/(a Mg 2+ + a Ni 2+) of 0.93. The corresponding point in the Lippmann diagram of Ker-Pim s.s. has coordinates (0.93, 4.66), falling under the solutus curve, and then Ker-Pim s.s. would be undersaturated.
Alternatively, the same statement can be inversely formulated: for given values of logSP and a Mg 2+/(a Mg 2+ + a Ni 2+), corresponding to an equilibrium Ker-Pim s.s. with solution, a series of (Srp-Nep) s.s. exists and they are supersaturated. This can be demonstrated graphically using the Lippmann diagrams (Figure 2a, 2b).
Similarly, points in the diagram to the right of the SK curve characterize the conditions for equilibrium precipitation of a Ker-Pim s.s.

Simultaneous precipitation of kerolite-pimelite and sepiolite-falcondoite solid solutions from aqueous solution
The equilibrium involved is: and its equilibrium constant is: Using this equilibrium constant for reaction 9, the equilibrium path for these two solid solutions (curve KSep, Figure 4) was constructed on the same basis as curve SK (Figure 3). This curve separates the stability field of Ker-Pim (left) from that of the Sep-Fal.
Along the equilibrium path of Ker-Pim with Sep-Fal, (curve KSep), Srp-Nep s.s. are undersaturated, except for high activity ratios of X Mg 2+ when they become supersaturated ( Figure 5). The corresponding logSP curve is depicted in Figure 2a.

Simultaneous precipitation of serpentine-népouite and sepiolite-falcondoite solid solutions from aqueous solution
Equilibrium between these solid solutions is given by the reaction: The equilibrium constant of this reaction was obtained by proper combinations of the reactions 1 À 2 À 5 + 6 (Table 2), as in the previous examples, yielding log K SSep = À0.76Ô0.42. The equilibrium path of the equilibrium Srp-Nep+Sep-Fal in the activity diagram is shown in Figure 5. Intimate mixing of these two solid solutions have not been observed. This is consistent with the fact that the curve for Srp-Nep+Sep-Fal equilibrium is metastable, except for the end near the intercept of serpentine-sepiolite ( Figure 5). The metastable extension falls in the field of Ker-Pim s.s.: the calculated logSP for Ker-Pim, translated to the Lippmann diagram lies above the solutus up to an activity ratio X Mg 2+ near 0.98 (Figure 2b).

Simultaneous precipitation of Ni-bearing sepiolite and quartz
One of the most commonly observed associations in garnierites is the solid solution Sep-Fal with quartz. The equilibrium path in this case is merely the vertical straight line log (a SiO 2 aq ) = À4.0 corresponding to the solubility of quartz. For greater activities of silica, Sep-Fal s.s. may be associated with amorphous silica.
Whether two solid solutions plus quartz can coexist in equilibrium with an aqueous solution is uncertain. The quartz boundary, represented by a vertical straight line at log(a SiO 2 ) = À4, intercepts the curve SSep and the metastable extension of KS, in the Mg-rich end of these curves ( Figure 6). Two solid solutions plus quartz might precipitate in near-equilibrium conditions. The position of the intersections of pure phase boundaries in the activity diagram and especially their intersections could change considerably due to uncertainties in the log K values, however.

MATERIALS AND METHODS
For the present study, garnierite samples from the Falcondo (Dominican Republic) and Loma de Níquel (Venezuela) Ni-laterite deposits were selected. The same samples were studied by Proenza et al. (2008), Tauler et al. (2009), Soler et al. (2010), and Villanova-de-Benavent et al. (2011a, 2011b. In addition, a list of garnierite minerals from Ni laterites worldwide was compiled in which accurate quantitative electron microprobe data are available (see Figure 1). Garnierites typically precipitate with strong compositional zoning, sometimes with oscillatory content in Ni, and with transitions from one structural type to another (e.g. from Ni-serpentine to Ni-kerolite). Thin veinlets of quartz cross the garnierite structures (Tauler et al., 2009;Villanova-de-Benavent et al., 2011b). As both the Lippmann and the activity diagrams were built on the basis of partially theoretically derived equilibrium constants (log K values for Ni end-members), having some experimental control over the reliability of the equilibrium constants used in the graphical representations was considered to be necessary. For this purpose,~100 polished sections of the material described above were examined to find garnierites consisting of two well characterized phases suitable for detailed chemical analyses, avoiding areas with pronounced zonation and oscillatory composition.

Determination of log K SK from biphasic particulate masses of serpentine-nepouite and kerolite-pimelite solid solutions
A garnierite sample from the Falcondo Mine was a clear example of different generations of co-precipitated Ni-enriched serpentine plus kerolite (Figure 7a). The garnierites consist of homogeneous light green masses cut in places by thin veinlets of quartz (1 mm thick or less) filling fractures or cavities in the saprolite horizon that appeared to have formed under near equilibrium conditions. Under the optical microscope, only quartz veinlets were distinguished from a homogeneous, poorly crystalline yellowish-green mass. Serpentine (mainly lizardite) and talc, in the form of the fine-grained hydrated variety known as kerolite (Brindley et al., 1977), was clearly distinguished by powder X-ray diffraction (XRD). The proportion of phases and the mean size of crystallites were estimated by Rietveld analysis of the XRD profile (Figure 8). Electron microprobe analyses of areas 1À2 mm in size revealed fractions of kerolite together with fractions of serpentine. The TEM studies also showed that the mineral particles were nm-sized (Soler et al., 2008;Villanovade-Benavent et al., 2011a).
A statistically sufficient number of points analyzed by EMP, in principle, allows the determination as to whether the precipitation of the mixture was produced under equilibrium conditions as well as the composition The fraction of serpentine (X serpentine ) probed was (Soctahedral atoms À 3)/3. If simultaneous precipitation of both phases proceeds in equilibrium, all particles of the same phase should have the same composition. For each point analyzed, a given increase in the serpentine fraction represents a constant increase in each octahedral cation (Mg, Ni, Fe 2+ ). Plotting these atomic contents against the serpentine fraction yielded straight lines for Mg, Ni, and Fe 2+ and the intercepts at X serpentine = 0 and X serpentine = 1 gave the compositions of (Ker-Pim) s.s. and (Srp-Nep) s.s. in the sample analyzed (Table 5). The reverse was also true: if the atomic contents of substitutable cations failed to lie on straight lines, the probed sample was inhomogeneous and consisted of particles of different compositions.
The log K SK 'analytical' value can be obtained from the composition of the equilibrium phases. For instance, the supposed ideal solid solutions in sample DR1 (Falcondo Mine) (Figure 7b) gave log K SK = À0.35, (with s(log K SK ) = Ô0.08) whereas that obtained from the analyses of the sample from Loma de Hierro Mine (Venezuela) (Figure 7c) gave log K SK = À1.19 (with s(log K SK ) = Ô0.55). This discrepancy provides a crude approach to the errors involved in the analytical estimation of log K SK . Nevertheless, both 'analytical equilibrium constants' for reaction 8 indicated that Ni atoms are better accommodated in the kerolite structure. These values fall around the value derived theoretically for log K SK (À0.73).
Determination of log K KSep from biphasic particulate masses of kerolite-pimelite and sepiolite-falcondoite solid solutions The same method was applied to estimate log K KSep . In this case, an homogeneous mass of greenish Ni-bearing minerals from Falcondo Mine were characterized unambiguously as a fine mixture of sepiolitefalcondoite and kerolite-pimelite solid solutions using powder X-ray micro-diffraction on an area of 500 mm ( Figure 9) and EMP analyses (Table 6). Diffraction was obtained using a Bruker D8 microdiffraction instrument equipped with a general area detector system (GADDS). As for the Ni-bearing serpentine and kerolite garnierites, the fraction of each phase in each individual analysis (the area illuminated by the electron beam) was obtained readily, expressing the unit formulae on a common basis (the same number of silicon atoms): The fraction of kerolite was then (Soctahedral cations À 8). The intercepts of the straight lines obtained by plotting Mg, Ni, and Fe 2+ contents against the fraction of kerolite gave the compositions of both solid solutions. From these, an estimated equilibrium constant for the next reaction (log K KSep = À0.16) was obtained. In contrast to the results for the mixtures of serpentine and kerolite, however, the points plotting Mg and Ni against the fraction of kerolite were more dispersed so that the estimation of the equilibrium constant (9) was less reliable. These data show that the sepiolite structure, compared with kerolite, is slightly preferred by Ni.
Analyses of polished sections of garnierites did not reveal biphasic mixtures of serpentine-nepouite and sepiolite-falcondoite solid solutions, so that the equilibrium distribution of Mg and Ni between these phases could not be ascertained analytically. This result is consistent with the narrow calculated stable boundary obtained theoretically for these solid solutions, as discussed below.

RESULTS AND DISCUSSION
By superposing SK, KSep, and SSep boundaries (Figures 3, 4, and 5) a mineral stability diagram for the system MgO-NiO-SiO 2 -H 2 O was obtained ( Figure 6). The six straight lines correspond to the solubilities of the six end-members (Srp, Nep, Ker, Pim, Sep, Fal). Three curves connecting the two intersecting points of pure end-members (curves SK, KSep, and SSep) depict the  The boundary between Srp-Nep and Ker-Pim fields (curve SK) displayed a pronounced change of slope. From the mid point to the Mg end, this boundary was almost parallel to the log(a SiO 2 (aq) ) axis, so that an increase in the activity of Ni in solution may favor the precipitation of Ker-Pim s.s., because the activity and activity ratio of Ni 2+ increased downward on the diagram. The mixtures of Srp-Nep+Ker-Pim, fell in a field where the variable log((a Mg 2+ + a Ni 2+)/(a H +) 2 ) was >13.75, with a Ni/Mg ratio in solution of 10 À5 À10 À4 and calculated pH of~9.2. The section of the SK boundary between the invariant point and the serpentine-kerolite intercept was metastable, running almost coincidental with, but below, the stable SSep boundary. For log((a Mg 2+ + a Ni 2+)/(a H +) 2 ) values of <13, in the Ni-rich portion, the boundary SK was parallel to the vertical axis, and only the activity of silica determined the type of solid solution that precipitates. This was also the case for the next boundary between Ker-Pim and Sep-Fal (curve KSep) which drew an almost straight vertical line at log(a SiO 2 (aq) ) = À4.5. The portion of the Srp-Nep field limiting with the Sep-Fal field was very narrow (boundary curve SSep), making the co-precipitation of these minerals less probable. In fact, coexistence of these two solid solutions was not observed in any of the samples.
Many processes leading to the formation of garnierites depend on factors not treated in this work (kinetic, climatic, hydrologic, tectonic, etc.). The stability diagram constructed (Figure 6) suggested, however, that two main temporal and spatial trends were related to the formation of Ni-enriched supergene phases. An increase in Ni 2+ in the solutions percolating through or contained within the altered parent rock may be explained by one or the other of these two mechanisms, e.g. the leaching of Ni from goethite in the limonitic horizon by meteoric waters, and the repeated cycle of congruent dissolution of Ni-poor phyllosilicates followed by equilibrium precipitation of a phyllosilicate, slightly richer in Ni. On the other hand, an increase in the silica activity may have occurred with the formation of a great variety of different forms of silicification (Golightly, 1981;Freyssinet et al., 2005;Trescases, 1975). Soler et al. (2008) observed in flow-through dissolution experiments of garnierites at low pH that the initial aqueous ratios of Mg/Si and (Mg+Ni)/Si were three times greater than in the solid. After 900 h, these ratios decreased to values only slightly above those of the dissolving garnierite. For higher pH values, the ratios in solution were always a little higher than in the dissolving phase. Preferential release of Mg over Si has been observed in Mgphyllosilicates previously (see Soler et al., 2008). Jurinski and Rimstidt (2001) observed the formation of a silica-rich surface on dissolving mineral grains. These experiments highlighted the fact that, for short dissolving events (i.e. a short period of rain), the dissolution of Ni-bearing phyllosilicates would not be stoichiometric with respect to the (Mg+Ni)/Si ratios but would be stoichiometric with respect to the Mg/Ni ratio. As a result, significant amounts of silica might accumulate and remain available until changes in pH allow its mobilization. In turn, the supply of silica to aqueous solutions explains the increase in the silica activity and the evolution of garnierites from Ni-serpentines to Nisepiolites. In summary, the evolution of garnierites, as shown in the stability diagram (Figure 6), may be depicted as a trend pointing downward (Ni enrichment) and to the right (increase in silica activity).

CONCLUSIONS
Garnierites are very often composed of an intimate mixture of Ni-enriched phyllosilicates which include serpentine-népouite, kerolite-pimelite, and sepiolitefalcondoite solid solutions. The variable and intricate