American Mineralogist, Volume 88, pages 725–738, 2003

The reduction of aqueous Au3+ by sulfide minerals and green rust phases

D.M. HEASMAN,1,* D.M. SHERMAN,1 AND K.V. RAGNARSDOTTIR1

1
Department of Earth Sciences, University of Bristol, Bristol, BS8 1RJ, U.K.

ABSTRACT
The reactions of Au3+ with green rust, stibnite, pyrite, and chalcopyrite were investigated in the
laboratory and the size of the gold clusters formed was measured using extended X-ray absorption
fine structure (EXAFS) spectroscopy, scanning electron microscopy (SEM), and transmission elec-
tron microscopy (TEM). The individual clusters produced were between approximately 29 Å and 77
Å in diameter and may occur individually or as composite clusters from hundreds to thousands of
angstroms in size. This work shows that it is possible to form “invisible” gold through reduction of
Au3+ by green rust phases and the surfaces of sulfide minerals.

INTRODUCTION transport and subsequent concentration, are sulfide and chlo-

Understanding the mechanisms responsible for the accu- ride (Shenberger and Barnes 1989). Generally, in systems where
mulation of gold is essential for interpreting the history and the aqueous chloride concentration is low, the solutions are re-
genesis of known gold deposits. This information can then be ducing and are up to temperatures of approximately 350 ∞C,
used to prospect for new reserves. Within gold ores, gold is sulfide is considered to be the most important ligand respon-
often found as submicroscopic grains associated with mineral sible for gold transport (Benning and Seward 1996 and refer-
surfaces and as inclusions within other minerals and this form ences therein). In systems where sulfide concentration is low
is known as “invisible” gold. It is thought that “invisible” gold compared to chloride, and the solutions responsible for trans-
is formed by reduction of aqueous Au+ or Au3+ by sulfide min- port are acidic and oxidizing (Murphy et al. 2000) and the
erals and Fe2+ phases. temperature is greater than 350 ∞C (Henley 1973), chloride is
“Invisible” gold can be found in both primary and second- considered to be the most important ligand in transporting gold.
ary gold deposits. In primary deposits, formed by hydrother- Sulfide is chiefly responsible for gold transport in systems
mal activity, the “invisible” gold is usually associated with where the fluids are magmatic in origin (e.g., Matel et al. 2000)
sulfide minerals (Foster 1984 and references therein). In sec- and chloride complexes are responsible for gold transport in
ondary deposits, principally those formed by supergene enrich- systems where the fluids are meteoric (e.g., Mernagh et al. 1994)
ment in lateritic environments, “invisible” gold is primarily or oceanic in origin (e.g., Muller et al. 2001).
associated with iron (oxy)hydroxides (e.g., Darke et al. 1997). Experimental studies of gold solubility in high temperature
The origin and nature of this type of gold is of particular sulfide solutions have shown that gold concentrations can reach
interest because the form and distribution of gold within the between 10–5 and 10–4 M and that the gold is primarily trans-
ore affects the design and efficiency of the extraction process ported as AuHS0 and Au(HS)2– (Seward 1973; Shenberger and
(Cook and Chryssoulis 1990; Hong et al. 1999). Cyanidation Barnes 1989; Benning and Seward 1996).
is the principal process by which gold is leached from oxygen- Henley (1973) showed that high-temperature (£500 ∞C)
ated slurries of crushed ore (Fleming 1992) and “invisible” gold chloride solutions can contain between 10–3 and 10–2 M gold
is not as amenable to cyanidation as other forms of gold found and a Raman spectroscopy study of gold chloride solutions
in ores (Spry and Thieben 2000). The effect of the presence of showed that up to 300 ∞C the gold forms complexes with the
“invisible” gold on ore processing efficiency is demonstrated general formula Au3+ClxOH–(4–x), with the value of x dependant
at the Hillgrove gold-antimony deposit (Australia) where it is on pH (Murphy et al. 2000). It is likely that in chloride-domi-
estimated that up to 20% of the available gold is lost to tailings nated systems that some reduction of gold will occur and there-
due to being present in this “invisible” form (Ashley et al. 2000). fore some of the gold will be present as Au+. The use of Au3+ in
the work presented here is justified because firstly, some gold
The concentration and aqueous speciation of gold in in these systems will be present as Au3+ and secondly, since the
natural solutions reduction of Au3+ by mineral surfaces is a two stage process
The concentration of gold in natural solutions is typically (Mycroft et al. 1995; Eqs. 1 to 5 below) Au3+ represents a con-
10–12 to 10–9 M (McHugh 1988; Vlassopoulos and Wood 1990a). venient and stable precursor for Au+ (Maddox et al. 1994).
The aqueous speciation of gold depends on the physico-chemi- Aqueous speciation in secondary ore forming fluids. The
cal properties of the fluid. solubility, and therefore the transport, of gold in oxidizing sec-
Aqueous speciation in primary ore forming fluids. In hy- ondary ore forming environments has traditionally been attrib-
drothermal systems the two ligands that are sufficiently abun- uted to chloride (Krauskopf 1951; Cloke and Kelly 1964) but
dant to complex gold, increasing its solubility and allowing more recent work on the dissolution of metallic gold in sodium
hydroxide solutions by Vlassopoulos and Wood (1990a, 1990b)
* E-mail: [email protected] suggests that the Au+ species AuOH(H2O)0 is the dominant aque-
0003-004X/03/0506–725$05.00 725
726 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS

ous species in most circum-neutral, oxidized natural waters (Fig. “Invisible” gold associated with sulfide minerals
1). However, spectroscopic studies on the aqueous complex- Adsorption and/or reduction of gold in solution by mineral
ation of gold have failed to identify the AuOH(H2O)0 complex surfaces to form metallic gold have only recently been invoked
in solution. Raman and UV/Vis. (Peck et al. 1991), Raman as a mechanism for gold accumulation in hydrothermal sys-
(Murphy and Lagrange 1998), extended X-ray absorption fine tems. Observations from the active hydrothermal system at
structure (EXAFS), and X-ray absorption near edge spectro- Broadlands, New Zealand have shown that economic concen-
scopic (XANES) (Farges et al. 1993) studies have observed trations of gold (associated with arsenic and antimony sulfides)
the Au3+ species AuCl4– at low pH values and as the pH was can form from fluids containing gold concentrations almost as
increased the chloride ligands were successively replaced by low as seawater (0.04 mg/kg) (Seward 1984). Experimental
OH ligands with none or minor Au+ species observed. These work on the adsorption of aqueous gold to amorphous anti-
observations are in direct contradiction with the speciation pre- mony and arsenic sulfides has shown that the concentration of
dicted using the DG 0f 298 of AuOH(H 2 O) 0 calculated by gold found in the Broadlands system could have accumulated
Vlassopoulos and Wood (1990a, 1990b) (Fig. 1; plotted using due to the adsorption of aqueous gold alone (Renders and
the experimental composition of Murphy and Lagrange 1998). Seward 1989). 197Au Mössbauer spectroscopy of these experi-
This contradiction suggests that the oxidation of aqueous gold mental systems confirms that gold is present as complexes at
may be kinetically controlled and if the dissolution experiments the surface of amorphous antimony and arsenic sulfides and
of Vlassopoulos and Wood (1990a, 1990b) had been left for a not as metallic gold (Cardille et al. 1993). However, metallic
longer period of time then Au3+ would have been observed in gold is also found in association with stibnite in ancient hydro-
the experimental solutions. All of the spectroscopic studies of thermal systems that are now being mined (e.g., Michibayashi
aqueous gold speciation used Au3+ salts to produce their ex- 1995; Berzon et al. 1999; Mehrabi et al. 1999). Further work is
perimental solutions; therefore any potential kinetic effects on needed to decide whether this metallic gold is reduced by
the oxidation state of gold would not have affected the results. longer-term processes involving the mineral surface or by some
In the EXAFS study of aqueous gold speciation by Farges et external reducing agent. Adsorption alone could produce inter-
al. (1993) the same Au3+ solutions were analyzed again after 5 lattice “invisible” gold because adsorbed gold could be incor-
months and there was no distinguishable difference in the porated into the lattice of the adsorbent by subsequent
EXAFS data produced. In addition to the contradiction between precipitation of new mineral layers. To form the angstroms to
the results of the dissolution experiments and the spectroscopic micron-sized metallic gold grains observed in some natural
observations this suggests that the oxidation rate of aqueous samples (e.g., Mehrabi et al. 1999), however, reduction would
Au+ to aqueous Au3+ has been underestimated and that once the be required.
oxidation is complete the Au3+ produced is stable. Therefore, aque- The commonest sulfide phase with which “invisible” gold
ous gold may be present in oxidized natural solutions as Au3+. is associated is pyrite (e.g., Mao 1991; Santaguida and
Hannington 1996; Yang et al. 1998) and it is this mineral that
has attracted the most experimental attention with respect to
gold adsorption and reduction. Jean and Bancroft (1985) found
that Au3+ is rapidly adsorbed from solution and then reduced to
Au0 at the surface of pyrite via an Au+ intermediate phase. This
reaction is accompanied by oxidation of the pyrite to aqueous
Fe3+ and sulfate or the formation of polysulfides and elemental
sulfur on the pyrite surface (Hyland and Bancroft 1989). The

TABLE 1. Summary of thermodynamic data used to plot the Eh-pH

diagram in Figure 1
Species DGf298 Species DGf298
(KJ/mol) (KJ/mol)
H 2O –237.2* AuOH2+ 185.057†
OH– –157.3* Au(OH)2+ –63.639†
Cl– –131.3* Au(OH)4– –455§
Au0(aq) 53.591† Au(OH)3Cl– –404§
Au+ 163.6‡ Au(OH)2Cl2– –350§
Au3+ 440§ Au(OH)Cl3– –294§
AuOH(H2O)0 –345.6# AuCl2OH(H2O)0 –386§
AuOH0 –61.964† AuH2OCl30 –311§
Au(OH)2– –276.32† AuCl2+ 239.485†
Au(OH)Cl– –219.388† AuCl2+ 70.757†
FIGURE 1. Eh/pH diagram for the system Au-Cl-H2O. Composition AuCl0 5.024† AuCl30 –83.317†
is the same as Murphy and LaGrange (1998) used in their Raman AuCl2– –151.1432 AuCl4– –234.6§
experiments (0.02 M HAuCl4◊4H2O) and the DG0f 298 of AuOH(H2O)0 is * Stumm and Morgan (1981).
from Vlassopoulos and Wood (1990b). The diagram is calculated from † Baranova and Rhyzenko (1981).
‡ Renders and Seward (1989).
the thermodynamic data presented in Table 1 using Geochemists § Baes and Mesmer (1976).
Workbench (Bethke 1994). # Vlassopoulos and Wood (1990b).
 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS 727

rate of gold deposition is controlled almost entirely by the rate marized by the following reaction:
of reduction of Au3+ and the particles of Au0 formed on the
pyrite surface consist of composite particles approximately 3800 AuCl4– + 3 Fe2+ + 6 H2O ´ Au0 (s) + 3 FeOOH (s) + 9 H+
Å in diameter (Maddox et al. 1994). A study of the kinetics of (6)
this system (Mycroft et al. 1995) has confirmed the existence
of these steps, which are summarized below: However, this relatively straightforward mechanism may
be more complex due to the formation of intermediate iron
Step 1 AuCl4– (aq) ´ AuCl3 (ads) + Cl– (aq) (1) oxy(hydroxide) phases such as green rusts. Green rusts, [Fe2+ 4
Fe23+(OH)12Xy, where X is typically Cl– (y = 2) or SO42– (y = 1)],
Step 2 AuCl3 (ads) + 2 e– ´ AuCl (ads) + 2 Cl– (aq) were first identified as the products of the corrosion of steel in
(2) earth surface environments (Schwertmann and Fetcher 1994)
and due to the sensitivity to oxidation they have only recently
Step 3 2 (FeS2 + 2x H+ ´ x Fe3+ + Fe1–xH2xS2 + x e–) been identified in natural systems, such as ochre sludges in
(3) Denmark (Bender Koch and Mörup 1991). Green rusts contain
a mixture of Fe3+ and Fe2+ and thus only partial oxidation of the
Steps 2 and 3 combined (x = 1) aqueous Fe2+ is required for them to form. On further exposure
to an oxidizing agent they fully oxidize to goethite or
AuCl3 (ads) + 4FeS2 (s) + 4H+ (aq) ´ AuCl (ads) + 2Fe3+ lepidocrocite. It is likely that in oxidizing groundwater where
(aq) + 2FeH2S4 (s) + 2 Cl– (aq) (4) complexes of aqueous Au3+ are stable any aqueous Fe2+ present
will at least be partially oxidized because the standard poten-
Step 4 AuCl (ads) + 2FeH2S4 (s) ´ Au (0) + Cl– (aq) + tial for the Fe2+ to Fe3+ half reaction is higher than the standard
4H+ + oxidized FeS2 (5) potential for the half reaction Au0 to Au3+ (–0.77 V and –1.4 V,
respectively; Atkins 1994). Thus the AuCl–4 (aq) species in Equa-
tion 6 can probably be reduced by an intermediary solid phase
The electrons for the reduction are supplied by step 3, which such as green rust and not solely by aqueous Fe2+.
removes the Fe2+ from the pyrite lattice where it is oxidized. The aim of this study is to react aqueous Au3+ with minerals
SEM imaging of the reacted pyrites revealed that the metallic that could act as reductants. Then, using EXAFS spectroscopy,
gold clusters formed were of approximately equal size (500– SEM, and TEM, to determine if metallic gold particles are
1000 Å), “suggesting an instantaneous nucleation and growth formed and to constrain the size of the particles produced to
mechanism” (Mycroft et al. 1995). determine if the gold is indeed “invisible.”
“Invisible” gold is also found associated with chalcopyrite
(e.g., Hannington et al. 1991; Bowell et al. 1999), sphalerite MATERIALS AND METHODS
(e.g., Rubin and Kyle 1997), arsenopyrite (e.g., Fleet and
Mumin 1997), pyrrhotite (e.g., Oberthur et al. 1997), and mar- General
casite (e.g., Fleet and Mumin 1997). Jean and Bancroft (1985) Reagents used in this study were of analytical grade or higher
included chalcopyrite, sphalerite, and pyrrhotite in their study and were supplied by Fisher Scientific. All apparatus were acid
and showed that, like pyrite, these sulfide mineral surfaces can washed and MilliQ (18.2 MW) water was used for all stages of
rapidly adsorb aqueous gold and reduce it to metallic gold. the experiments. The pH measurements were made using
“Wilhelm” style electrodes (Sentek U.K. Ltd.) connected to an
“Invisible” gold associated with iron oxy(hydroxides) Orion model 720A pH meter or to a Metrohm auto-titrator
The “invisible” gold in supergene laterite deposits is thought (Metrohm U.K. Ltd. Titrino model 718). Electrodes were cali-
to form by two processes, either acting individually or in com- brated using constantly stirred certified buffers (4.007 ± 0.012,
bination with each other: adsorption to iron minerals and 6.869 ± 0.012, and 9.180 ± 0.012, Schott Gerate) stored in a
ferrolysis (Eq. 6). Adsorption studies have focused on the two water bath to ensure temperature equivalence with the experi-
main iron minerals found in lateritic deposits, hematite, and ments. Calibrations were only accepted if the R2 value of the
goethite (Valeton 1994 and references therein). Gold adsorbs calibration curve was better than 98%. Light was excluded from
to both minerals and whilst hematite is considered to be an all preparations to avoid any possible photochemical reactions.
effective scavenger of gold as aqueous complexes (Nechayev In all cases the samples were centrifuged for 20 min at 3300
1984; Karasyova et al. 1998) and as colloids (Enzweiler and rpm following which the supernatant was removed and filtered
Joekes 1991), goethite does not accumulate gold to ore grades using a 0.2 mm nitrocellulose filter mounted in a polypropy-
(Machesky et al. 1991; Schoonen et al. 1992). Whether the min- lene filter housing. The filtrate was then quantitatively diluted
eral is an effective scavenger of gold or not, adsorption could by 10 using 1% v/v HCl (aq) and these solutions were ana-
be responsible for the formation of inter-lattice “invisible” gold lyzed for dissolved species of interest by inductively coupled
since any gold adsorbed to either of these minerals could be plasma atomic-emission spectroscopy (ICP-AES). A Jobin Yvon
incorporated into the mineral lattice as new layers of iron JY 24 ICP-AES was calibrated using a series of standards pro-
(oxy)hydroxide are precipitated on pre-existing mineral duced by serial dilution of 1000 ppm stock solutions (Fisher
grains. Ferrolysis refers to the reduction of aqueous Au3+ by Scientific, ICP grade). All calibrations had R2 values better than
aqueous Fe2+ from the bedrock (Mann 1984) and can be sum- 99% (typically 99.99%). Small amounts of the wet pastes pro-
728 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS

duced were mounted in Teflon sample holders at room tem- sealed using Teflon tape and were shaken in a Grant SS30 wa-
perature using Sellotape. These were then frozen in liquid ni- ter bath maintained at 25 ∞C (±1 ∞C) for 12 hours. At the end of
trogen and stored in a freezer for subsequent analysis using this time the samples were separated as before so that the su-
EXAFS and XANES spectroscopies. pernatant could be analyzed by ICP-AES for aqueous gold and
The remaining wet pastes were analyzed using SEM and the wet pastes produced were analyzed using EXAFS, XANES,
TEM in an attempt to measure the size of any gold particles SEM, and TEM. Analysis of the supernatant for copper, anti-
present. X-ray diffraction (XRD) was used to identify the solid mony, and iron using ICP-AES showed that all of these cations
phases present. The wet pastes were mixed with a small amount were present in the supernatants below detection limits of the
of glycerol to prevent oxidation during analysis and were ICP-AES (9.7, 8.4, and 10 ppb, respectively) and therefore there
mounted on glass slides to be analyzed using CuKa radiation was no detectable mineral dissolution over the duration of the
on a Philips PW 1800 X-ray powder diffractometer. experiments.

Green rust sample preparation EXAFS sample preparation and analysis

Green rust was prepared using the method of Schwertmann The slides were transported to a synchrotron laboratory in a
and Fetcher (1994). All samples were taken via self-sealing plastic bottle floating in liquid nitrogen and were placed in a freezer
rubber septa using a 20 mL polypropylene syringe with a 50 upon arrival. In addition to the green rust and sulfide samples spec-
mm stainless steel needle attached. Samples were immediately tra were also collected from a 10 mm thick gold foil. EXAFS data
transferred to 50 mL polypropylene centrifuge tubes purged were collected at the Central Laboratory of the Research Councils
with argon gas and then sealed with Teflon® tape. Sample prepa- Synchrotron Radiation Source at Daresbury, U.K. Samples were
ration took place in a 500 mL polypropylene gas tight reaction mounted on a liquid nitrogen cryostat. EXAFS and X-ray absorp-
vessel in a water bath maintained at 25 ∞C (±0.01 C) by a Haake tion near edge spectroscopy (XANES) spectra were collected in
DC5 heater/circulator and a cooling circuit connected to a tap fluorescence mode at the Au L-III edge (11919 eV) on station 9.3,
water supply. Initially the vessel contained 300 mL of water which is equipped with a harmonic-rejecting double crystal Si (220)
purged with nitrogen gas for 12 hours. The purge gas was then monochromator. Each sample was scanned for 45 min 8 to 18
changed to argon gas prior to addition of a known weight of times. The storage ring energy was 2 GeV and the beam current
FeSO4◊7H2O (s) (approximately 20 g). The pH was then raised varied between 100 and 240 mA.
from an initial value of ~3.4 to 7.0 over 15 min by the addition of EXAFS data reduction was performed using Daresbury
carbon dioxide free 1 M sodium hydroxide by a Metrohm 718 pH- Laboratory software packages (EXCALIB, modified to allow
Stat titrator. Once pH 7 was reached the argon gas supply was for the non-linearity of the detector at high count rates, and
replaced by a laboratory air supply provided by a small pump. EXBACK; Dent and Mosselmans 1992). EXCALIB was used
Initial experiments had shown that three hours of oxidation were to calculate energy (eV) from the monochromator position (in
required for a complete transformation of the aqueous Fe2+ to green millidegrees) to and to average multiple spectra from individual
rust. At the end of the initial oxidation step a 20 mL sample of the samples. EXBACK was used to define the start of the EXAFS
suspension was removed for analysis by XRD, to confirm that the oscillations (determined from the inflection point on the L-III
dark blue green precipitate formed was green rust, and a filtered edge) and perform background subtraction. The pre-edge was
sample was analyzed by ICP-AES, to measure aqueous iron en- fitted to a linear function and the post-edge background to two
abling quantification of the amount of green rust produced. Im- second-order polynomial segments. The EXAFS oscillations
mediately a known volume of a 5543 ppm stock solution of HAuCl4 were simulated using the exact curved wave theory (Gurman
was added to give final gold concentrations of 0.1 wt% [(weight et al. 1984; Gurman 1988) as implemented in the EXAFS analy-
of gold/theoretical yield of green rust) ¥ 100]. The initial experi- sis program EXCURV98 (Binstead et al. 1998). The phase-
ments had shown that 100% of the aqueous gold was removed shift functions used in the curve fitting process were derived
from solution within 10 min of gold addition. At this point two by ab initio methods in EXCURV98 using Hedin-Lundqvist
samples were taken; one was frozen and used for EXAFS analysis potentials (Hedin and Lundqvist 1969) and von Barth ground
and the other was separated as above and was analyzed by XRD states. No Fourier filtering was performed during the data analy-
and ICP-AES. The preparation was repeated to produce a sample sis and all the shells were fitted without constraint.
with 0.5 wt% gold. Initially a value for the R-factor (the fitting parameter) for
the first co-ordination shell in each sample was generated by
Sulfide sample preparation optimizing co-ordination number (N), shell radius (R) and
All sulfide minerals (stibnite, pyrite, and chalcopyrite) used Debye-Waller factor (DWF) for each element present in the
in this study were obtained as powders from the Earth Sci- system. The element that gave the best fit (i.e., the lowest value
ences collection at the University of Bristol and were of un- of the R-factor) was then assigned to that shell. This was re-
known origin. A few grams of each mineral were washed using peated for successive shells and only shells that decrease the
1% v/v hydrochloric acid to remove any oxidized species from R-factor by 5% were included in the final fit.
the surface and dried at room temperature under argon gas.
Approximately 5 g of each mineral was then transferred to a Scanning electron microscope (SEM) imaging and energy
50 mL, argon gas purged, polypropylene centrifuge tube and dispersive X-ray (EDX) scanning
10 mL of each of the following solutions was added; 3 M NaCl, To confirm that the samples contain submicroscopic par-
18.2 MW water, and 5036 ppm HAuCl4. The tubes were then ticles of gold, as deduced from the EXAFS and XANES analy-
 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS 729

sis, SEM images of all of the samples were collected using a from Fletcher et al. (1996) (Table 2).These two calculations
Hitachi S/2300 scanning electron microscope fitted with an were then used in conjunction with the formula for Nt (Eq. 8)
EDX detector. Aliquots of the samples analyzed by EXAFS to produce expressions for the average Nx (2 < x < 5) for each
and XANES spectroscopies were dried overnight at room tem- shell as a function of edge length (m) (Table 3).
perature under argon gas, and subsequently were mounted di-
rectly onto carbon pads in a glove box under nitrogen gas, then RESULTS AND DISCUSSION
were coated with carbon and loaded directly into the SEM.
During the coating and loading stage it was not possible to Characterization of green rust and sulfide minerals
exclude oxygen. Initially low magnification images were Comparison of the reference XRD diffractograms with the
scanned with the EDX detector to locate gold within the sample. experimental data shows that all of the samples contain no other
Based on these data small areas where gold was known to be detectable minerals except that the chalcopyrite sample appears
present were selected for collection of a higher magnification to contain a minor amount of chalcocite (CuS). The BET sur-
image. face areas of the stibnite, pyrite, and chalcopyrite samples were
measured as 3.33, 3.3, and 6.59 m2/g, respectively.
Transmission electron microscope (TEM) imaging and
EDX analysis Qualitative analysis of XANES spectra
To further constrain the size of the gold particles produced The XANES regions of the EXAFS spectra collected for
in these experiments the samples were also imaged using a each sample are shown in Figure 2. Figure 2 also shows spec-
JEOL JSM 5600LV transmission electron microscope. Sepa- tra of reference compounds containing the two principal oxi-
rate aliquots of the wet pastes analyzed using EXAFS and dation states of aqueous gold found in nature: Au3+ (KAuCl4;
XANES spectroscopies were suspended in 18.2 MW water, Farges et al. 1993) and Au+ (sodium gold thiomalate, AuSTm;
were placed on a copper grid using a transfer pipette and then Watkins et al. 1987). The reference spectrum for metallic gold
loaded into the TEM. All stages of the sample preparation for was obtained from a 10 mm thick gold foil. Comparison of the
TEM analysis were performed without exclusion of oxygen. reference spectrum to the experimental data clearly shows that
To ensure that the images obtained represented areas of the gold is present as metallic gold in all of the samples analyzed.
sample where gold is present, an EDX detector was used to All experimental spectra show the same two absorption maxima
qualitatively measure the chemical composition of the area of as the gold foil spectrum (feature a, Fig. 2); none of the experi-
the sample represented by the image. mental spectra contain the absorption maxima characteristic of
Au3+ (feature b, Fig. 2) present in the KAuCl4 spectra and none
Calculation
of gold cluster sizes and co-ordination of the experimental spectra resemble the relatively featureless
environment
Benfield (1992) presented formulae that relate the co-ordi-
nation number in the first shell (N1) in metallic clusters to the TABLE 2. Population and co-ordination of the different sites in a
size of the cluster (valid for values of N1 < 12). The formulae cuboctahedral cluster in terms of edge length m (atoms)
allow N1 to be calculated for a cluster of edge length m (at- Site Population Co-ordinations
oms). For a cuboctahedron: N1 N2 N3 N4 N5
(2 m - 3)(5m - 15m + 13)
2
( m - 1)(10 m 2 - 14 m + 6) Interstitial 12 6 24 12 24
N1 (cuboct ) = 12 (7) 3
(2 m - 1)(5m 2 - 5m + 3) Vertex 12 5 2 6 5 5
Edge 24(m – 2) 7 3 10 7 10
The number of atoms in the cluster (Nt) can also be calcu- Square Face 6(m – 2)2 3 5 8 12 16
lated: Triangular Face 4(m – 3)(m – 2) 9 3 13 5 12

(2 m - 1)(5m 2 - 5m + 3)
Nt = (8) TABLE 3. Expressions for the average co-ordination environment
3
of each of the first five shells present in a cuboctahedral
Once m has been calculated it can be multiplied by the cluster
atomic diameter of gold (Å) to give the edge length of the clus- Shell Average Co-ordination
ter, and since the radius of a sphere circumscribing a 10 m 3 - 24 m 2 + 17m - 3
N2 6
cuboctahedron equals the edge length, the cluster diameter can (2 m - 1)(5m 2 - 5m + 3)
be easily calculated.
Calculation of m also allows predictions of the co-ordination 20 m 3 - 59m 2 + 65m - 30
number for the subsequent four shells to be made for compari- N3 12
(2 m - 1)(5m 2 - 5m + 3)
son with EXAFS results. The average co-ordination numbers
for each of the shells were calculated using the formulae for N4 5m 3 - 14 m 2 + 20 m - 15
24
the populations of the different sites (interstitial, vertex, edge, (2 m - 1)(5m 2 - 5m + 3)
square face, and triangular face) present within a cuboctahedral
cluster presented by Benfield (1992) (Table 2). The co-ordina- 5m 2 + 11m + 6
tion of each site for the first five shells was then measured N5 24
5m 2 - 5m + 3
using the structure for face centered cubic (FCC) bulk gold
730 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS

shown in Figures 4b (green rust samples) and 5b (sulfide

samples).
Metallic gold can form cuboctahedral and icosahedral clus-
ters (Benfield 1992; Fritsche and Benfield 1993) and these two
cluster types have different co-ordination numbers (N) at dif-
ferent distances R (Å) (Table 4).
Comparison of these two co-ordination environments with
the EXAFS data for all samples shows that the metallic gold is
crystallizing as cuboctahedral clusters because no indication
of the third, fifth, and sixth shells (4.18, 5.16, and 5.24 Å, re-
spectively) of the icosahedral geometry is observed in the
EXAFS data. In all of the samples the distances predicted us-
ing a cuboctahedral geometry (using values of Rx/R1 from
Benfield 1992) are in very good agreement with the measured
distances (Table 5). All measured distances agree with the pre-
dicted distances to within 1.5%.
The co-ordination numbers predicted by the cuboctahedral
geometry agree with the measured values within the ±20% er-
ror in N associated with EXAFS results, except for the third
FIGURE 2. XANES spectra of all of the samples analyzed by shell in the green rust sample with 0.1 wt% gold, which differs
EXAFS. Also included are reference XANES spectra for the three
principal oxidation states of gold (underlined). TABLE 4. Comparison of the co-ordination environments of
cuboctahedral and icosahedral gold clusters
Cuboctahedral Icosahedral
spectra of AuSTm. The XANES spectra also contain some in- Shell R Rx/ R1 N R Rx/ R1 N
formation about the size of the metallic gold clusters. Gener- 1 2.88 1 12 2.88 1 6
2 4.07 1.41 6 3.03 1.05 6
ally the intensity and width of the absorption maxima (feature 3 4.99 1.73 24 4.18 1.45 6
a, Fig. 2) present in the spectra for metallic gold clusters are 4 5.76 2.00 12 4.90 1.70 6
directly related to the number of neighbors and, more gener- 5 6.44 2.24 24 5.16 1.79 12
6 7.05 2.45 8 5.24 1.82 6
ally, to the size of the gold clusters (Benfield et al. 2001). Fig- 7 7.62 2.65 48 5.76 1.97 6
ure 2 shows that the sizes of these two maxima in the normalized 8 8.15 2.83 6 6.06 2.00 6
spectra of the samples are similar to that of gold foil. This sug- 9 8.64 3.00 36 6.51 2.10 12
10 9.11 3.16 24 2.26 2.26 12
gests the presence of at least five to six shells of gold atoms Note: Rx/R1 equals the ratio of the distance of shell x to the distance of the
(Benfield et al. 2001) and that the clusters are at least 36 to 43 Å in first shell.
diameter.
TABLE 5. Comparison of shell distances observed in the EXAFS
EXAFS spectra of gold foil data with those predicted for cuboctahedral gold clusters
The k3 weighted EXAFS spectrum of gold foil is shown in Sample Measured Predicted Measured Predicted Measured
Figure 3a and the Fourier transforms of these data and the in- R1 R2 R2 R3 R3
terpretation of the co-ordination environment in this foil is Gold Foil 2.87 4.06 4.06 4.97 5.00
Green rust 2.86 4.04 4.05 4.95 4.98
shown in Figure 3b. In these figures and all subsequent figures + 0.1 wt% Au
containing EXAFS data the thin line represents the experimental Green rust 2.87 4.06 4.06 4.97 5.00
+ 0.5 wt% Au
data and the bold line represents the calculated theoretical spec- stibnite 2.88 4.08 4.07 4.99 5.00
trum. The Debye Waller factors are also given (DWF). Calcu- + 1.07 wt% Au
lation of the theoretical co-ordination environment within bulk pyrite 2.87 4.06 4.07 4.97 4.98
+ 1.05 wt% Au
gold shows that it contains 12 atoms at 2.88 Å, 6 atoms at 4.08 chalcopyrite 2.87 4.05 4.07 4.96 4.98
Å, 24 atoms at 4.99 Å, 12 atoms at 5.77 Å, and 24 atoms at + 1.02 wt% Au
6.45 Å (Fletcher et al. 1996), which is in very good agreement
Sample Measured Predicted Measured Predicted Measured
with the distances observed in the EXAFS data presented here R1 R4 R4 R5 R5
(Fig. 3b). Gold Foil 2.87 5.73 5.67 6.41 6.44
Green rust 2.86 5.71 NP 6.38 NP
+ 0.1 wt% Au
EXAFS spectra of gold after reaction with green rust and Green rust 2.87 5.74 5.67 6.42 NP
sulfide minerals + 0.5 wt% Au
stibnite 2.88 5.77 5.68 6.45 6.42
The k3 weighted spectra for gold after reaction of aqueous + 1.07 wt% Au
Au3+ with green rust and after reaction with the sulfide miner- pyrite 2.87 5.74 5.67 6.41 6.42
als are shown in Figures 4a and 5a, respectively. The Fourier + 1.05 wt% Au
chalcopyrite 2.87 5.73 5.66 6.41 6.43
transforms of the k3 weighted EXAFS spectra and the interpre- + 1.02 wt% Au
tations of the co-ordination environments in these samples are Note: NP = shell not present.
 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS 731

FIGURE 3. (a) k3 weighted EXAFS spectrum for gold foil. (b) Fourier transform of the k3 weighted EXAFS spectrum shown in a.

TABLE 6. Comparison of co-ordination numbers observed in the proximately 20 to 250 Å. Further constraints can be put on
EXAFS data with those predicted for cuboctahedral gold these sizes because in all of these samples (except for the green
clusters rust samples) five shells, that have co-ordination numbers in
Sample Measured Predicted Measured Predicted Measured good agreement (i.e., within ±20%) with the predicted values, are
N1 N2 N2 N3 N3
present. In cuboctahedral clusters the edge length (m) is equiva-
Green rust 11.67 5.84 5.80 22.97 14.50
+ 0.1 wt% Au lent to the number of co-ordination shells present in the cluster.
Green rust 9.46 4.64 4.40 16.65 18.45 Therefore these data suggest that the value of m in these samples
+ 0.5 wt% Au
stibnite 9.13 4.45 3.98 15.77 18.45
is at least 5, which corresponds to a cluster diameter of 36 Å.
+ 1.07 wt% Au In the green rust sample with 0.1 wt% gold, only three shells
pyrite 10.94 5.45 5.32 20.74 18.29 are observed in the EXAFS data. We do not think, however,
+ 1.05 wt% Au
chalcopyrite 10.80 5.38 4.69 20.33 18.24 that this represents a cluster with a maximum value of m = 3
+ 1.02 wt% Au because the first shell contains 11.67 gold atoms, which corre-
sponds to a cluster diameter of 241 Å, suggesting there are
Sample Measured Predicted Measured Predicted Measured
N1 N4 N4 N5 N5 more than 3 shells present. It is more likely that the fourth and
Green rust 11.67 11.55 NP 23.13 NP fifth shells are not shown in the EXAFS spectrum because the
+ 0.1 wt% Au
Green rust 9.46 9.17 8.81 17.27 NP
concentration of gold in this sample is low, resulting in poorer
+ 0.5 wt% Au quality spectra (the R-factor is 42.32). However, the XANES
stibnite 9.13 8.88 5.02 16.38 15.51 spectrum of this sample qualitatively suggests that at least five
+ 1.07 wt% Au
pyrite 10.94 10.64 9.37 21.1 17.15 shells are present and therefore the clusters are at least 36 Å in
+ 1.05 wt% Au size (see above section “Qualitative Analysis of XANES spec-
chalcopyrite 10.80 10.48 10.83 20.80 20.56 tra”). In the green rust sample with 0.5 wt% gold only four
+ 1.02 wt% Au
Note: NP = shell not present.
shells are detectable in the EXAFS data. A cluster with m = 4
would have a diameter of 28.64 Å, which is in good agreement
with the measured size of the clusters in this sample. Given the
by 37%, and the fourth shell in the stibnite sample, which dif- good fit to the EXAFS data (R-factor of 25.65) it is thought
fers by 43% (Table 6). It seems, therefore, that using the that the clusters present in this sample contain between four
cuboctahedral geometry is a valid method to calculate the size and five co-ordination shells of gold atoms that represent an
of the gold clusters present in these samples. average cluster size of between 28.64 and 36 Å. Some qualita-
tive evidence for the presence of this number of shells is also
Calculation of gold cluster size contained in the XANES spectrum of this sample (see above
The results of the cluster size calculations are shown in Table section “Qualitative Analysis of XANES spectra”).
7. Cluster sizes have also been calculated including the error
of ±20% in N1. The results show that for all of the samples, SEM and EDX analysis
except for green rust with 0.5 wt% gold and stibnite, the clus- Figures 6 to 10 depict the SEM images and EDX scans of
ters produced are at the upper size limit that is resolvable with the samples analyzed in this work. Figures 6a and 7a show the
this technique in these types of systems because the error in N1 images of the green rust samples at low magnification (¥3000)
produces greater variations in diameter as N1 approaches 12. with the EDX data overlain and Figures 6b and 7b show the
However, these data can be used to provide a lower size higher magnification (¥10 000) images of the green rust samples
limit for the gold clusters produced in this experiment of ap- and the EDX scans taken of these views. In these and subse-
732 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS

FIGURE 4. (a) k3 weighted EXAFS spectra for gold after reaction of aqueous Au3+ with green rust. (b) Fourier transforms of the k3 weighted
EXAFS spectra.

TABLE 7. Results of cluster size calculations for all samples analyzed by EXAFS
Sample Diameter N1 Diameter N1 Diameter
N1 Å –20% Å +20% Å
Green rust + 0.1 wt% Au 11.67 241 9.34 32 14.01
Green rust + 0.5 wt% Au 9.46 34 7.57 20 11.36 123
Stibnite + 1.07 wt% Au 9.13 30 7.30 19 10.95 77
Pyrite + 1.05 wt% Au 10.94 76 8.75 27 13.13
Chalcopyrite + 1.02 wt% Au 10.80 68 8.64 26 12.96
Note: The blank fields represent results when N1 > 12.

quent images the presence of gold is indicated by the black spectra show that the areas of each sample imaged contain gold.
dots. Figures 8a, 9a, and 10a show the initial low magnifica- The large copper peaks in Figures 11, 12, 13, and 14 are due to
tion images of the sulfide samples and EDX scans. Figures 8b, the copper grids on to which the samples were loaded. The
9b, and 10b present the higher magnification images of the gold- morphology of the iron phases in the TEM images of the green
containing areas marked with a black rectangle in the lower rust samples are obviously different from the morphology ob-
magnification images. served in the SEM images of the same samples (Figs. 6 and 7).
The absence of any obvious features that can be identified The images obtained by SEM show the green rust as an amor-
as gold clusters in the high magnification images (Figs. 6b, 7b, phous precipitate whereas the TEM images show well formed
8b, 9b, and 10b) suggests that the metallic gold clusters seen in acicular crystals a few hundreds of angstroms wide and a few
the EXAFS data are smaller than the minimum size resolvable thousands of angstroms long.
on a SEM (0.5 mm, 5000 Å) and are indeed sub micrometer or The formation of these crystals was directly observed dur-
“invisible” gold particles (Yang et al. 1998). ing sample preparation for TEM analysis as a transformation
of the blue green green rust phase to an orange brown phase
TEM and EDX analysis upon suspension of the sample in water. Since green rust is an
Figure 11 shows the TEM images and EDX spectra obtained intermediate phase in the oxidation of Fe 2+ (aq) to Fe 3+
for the green rust samples analyzed in this work. The EDX (oxy)hydroxides and the crystals are acicular it seems very
 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS 733

FIGURE 5. (a) k3 weighted EXAFS spectra for gold after reaction of aqueous Au3+ with sulfide minerals. (b) Fourier transforms of the k3
weighted EXAFS spectra.

likely that this phase is goethite. This was confirmed by XRD are no obvious features that could be interpreted as metallic
analysis of an aliquot of the green rust with 0.1 wt% gold sample gold clusters. This could be because there is not sufficient dif-
used for TEM analysis. The difference is due to the fact that ference in contrast between the goethite and the gold clusters
the green rust samples prepared for SEM analysis were mounted or that they are too small to be resolved by the TEM (<5 nm or
and loaded directly into the SEM whereas the samples pre- <50 Å). A TEM study of gold associated with goethite shows
pared for TEM analysis had to be suspended in water before that metallic gold clusters are formed after hydrolysis of a
mounting and loading into the TEM and this step was suffi- ferrihydrite suspension containing Au3+ (Greffie et al. 1996).
cient to oxidize the green rust to goethite. The individual gold clusters are clearly seen in the TEM im-
The TEM images of the green rust samples show that there ages as very dark sub circular features of 30 to 600 Å in diam-
734 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS

FIGURE 6. SEM and EDX images of green rust +0.1 wt% gold. FIGURE 8. SEM and EDX images of stibnite +1.07 wt% gold.

FIGURE 7. SEM and EDX images of green rust +0.5 wt% gold. FIGURE 9. SEM and EDX images of pyrite +1.05 wt% gold.

eter or composite clusters representing multiple twinned par-

ticles up to 1500 Å in diameter growing on the edges of the
goethite lathes. This suggests that if the metallic gold clusters
present in the oxidized green rust samples are large enough to
be resolved by TEM then they would be clearly visible, there-
fore the gold clusters in this study are <50 Å in diameter. This
size is in good agreement with the size calculated for the green
rust sample with 0.5 wt% gold (28.64–36 Å).
The TEM images and EDX spectra of the stibnite samples
are shown in Figure 12. The TEM images show that there are
no features that could be interpreted as individual metallic gold
clusters. TEM images of natural samples of “invisible” gold
found in association with stibnite show that gold is clearly seen
as dark black spheres (400–1000 Å in diameter) located in
microfissures and in grain interstices of the lighter stibnite par-
ticles (Yang et al. 1998). The TEM images (Fig. 12) do contain
some darker areas of similar size to those observed in the natu-
ral samples (100–320 Å) associated with the finer mineral grains FIGURE 10. SEM and EDX images of chalcopyrite +1.02 wt% gold.
HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS 735

FIGURE 11. TEM images and EDX

spectra of the green rust samples with (a)
0.1 wt% gold and (b) 0.5 wt% gold.

FIGURE 12. (a) TEM images and

EDX spectrum of the stibnite sample
with 1.07 wt% gold. (b) Expanded image
of the highlighted area in a.
736 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS

F IGURE 13. TEM image and

EDX spectrum of the pyrite sample
with 1.06 wt% gold.

FIGURE 14. (a) TEM images

and EDX spectrum of the
chalcopyrite sample with 1.02 wt%
gold. (b) Expanded image of the
highlighted area in a.

(Fig. 12b) but we do not think that these features correspond to Maddox et al. (1994) it is possible that they are forming com-
individual metallic gold clusters. On closer examination some posite clusters a few hundreds of angstroms in diameter.
of these darker features appear to be where mineral grains over- Figure 13 shows a TEM image and the EDX spectra of the
lap, as none are as clearly defined as those observed in the sample of pyrite containing 1.05 wt% gold. The EDX spectra
study of Yang et al. (1998) and they do not appear to form the clearly shows that gold is present in this sample and the EXAFS
particles of regular spherical morphology (Yang et al. 1998) show that this gold is present as metallic gold; however, the
exhibited by metallic gold. Another possibility is that these fea- image contains no features that may be interpreted as metallic
tures in Figure 12b are irregular aggregates of smaller gold gold clusters. TEM images of natural samples of “invisible”
clusters (Maddox et al. 1994). The diameter of any individual gold found in association with pyrite show that gold is clearly
metallic gold cluster in this sample is therefore thought to be seen as dark black spheres (1000–2000 Å in diameter) either
below the resolution of the TEM used in this work (<5 nm or seen individually or forming chains on growth surfaces of the
<50 Å) though considering the EXAFS data and the study of much lighter pyrite mineral grains (Yang et al. 1998). In con-
 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS 737

trast to this, the pyrite grain shown in Figure 13 is very dark which Equation 7 is not valid (i.e., >12) and therefore these
and this may mean that the clusters may be present but there is data cannot be used to unambiguously constrain the upper size
an insufficient difference in contrast between the pyrite and limit of the individual clusters present in this sample. The
the gold for the clusters to be visible. This means that it is very XANES spectrum of this sample and the good agreement in
difficult, based on the TEM analysis, to apply any constraints predicted and measured co-ordination numbers for all five shells
on the size of the metallic gold clusters present in this sample. suggests that there are at least five co-ordination shells, corre-
Again the possibility that the any individual gold clusters are sponding to a minimum size of 36 Å. A best estimate of the
forming composite clusters should be considered and if they maximum size of the individual particles of 77 Å could be made
are present they have a maximum size of 5000 Å since they are on the basis that the good fit of the theoretical model to the
not resolved in the SEM images (Fig. 9). EXAFS data (R-factor is 27.94) implies reliability in the value
The darker features present in the TEM images of the chal- of N1, however, this should be considered tentative at best.
copyrite sample with 1.02 wt% gold (Fig. 14a) may represent The EXAFS data from the chalcopyrite sample with 1.02
metallic gold clusters 100 to 440 Å in diameter but on closer wt% gold implies that the individual metallic gold clusters
inspection (Fig. 14b) some resemble the darker areas of the present in this sample are between 26 Å and at least 68 Å and
stibnite sample in that some of them are caused by overlapping the upper size limit of individual clusters is further constrained
mineral grains and none represent the typical gold morpholo- by the TEM images to 50 Å. The good agreement between theo-
gies seen in TEM studies of gold reduction by minerals. As retical model and the EXAFS data (R-factor is 25.77) and the
with the stibnite sample it is also possible that these dark ir- theoretical co-ordination numbers and experimentally measured
regular features are composite clusters of smaller gold clus- values for all five shells is strong evidence that there are at
ters. It is thought that, like the stibnite sample, any individual least five shells present. The average size range for the indi-
clusters present in this sample are too small to be resolved by vidual metallic gold clusters in this sample is therefore 36 to
the TEM and therefore must have a maximum diameter of 50 Å. 50 Å and they may be forming composite clusters a few hun-
dreds of angstroms in diameter.
Assignment of average sizes to metallic gold clusters The average size ranges for the clusters present in all of the
The EXAFS data for the green rust sample with 0.1 wt% samples analyzed are summarized in Table 8.
gold suggests a metallic gold particle size of between 32 and
241 Å, however no gold particles are visible on the TEM im- Comparisons with previous studies
ages; therefore the upper limit of this range can be constrained In studies of gold reduction in laterites, Mann (1983, 1984)
to 50 Å. The XANES and EXAFS data suggest that these clus- proposed ferrolysis as the mechanism of metallic gold accu-
ters probably have five shells and are therefore at least 36 Å in mulation in these types of deposits. The last stage of ferrolysis,
diameter. The size range of the particles present in this sample the precipitation of metallic gold with oxidized iron minerals,
is therefore between 36 and 50 Å. The TEM images of the green was studied by Greffie et al. (1996) by co-precipitating metal-
rust with 0.5 wt% gold also constrain the maximum size of the lic gold with goethite and ferrihydrite. They observed individual
metallic gold clusters present in this sample to 50 Å. The good particles of “invisible” gold between 30 and 600 Å and com-
quality of fit of the EXAFS data strongly suggest that the clus- posite clusters consisting of multiple twinned particles up to
ters may be slightly smaller than this, with diameters between 1500 Å in diameter. The work presented here has shown that
29 and 36 Å. the overall process of ferrolysis is likely to be more complex in
The diameter of any individual gold clusters present in the that aqueous Au3+ will not only be reduced by aqueous Fe2+ to
stibnite sample suggested by the TEM work is in good agree- form metallic gold but will also be reduced by Fe2+ present in
ment with the cluster sizes implied by the EXAFS and XANES partially oxidized intermediate iron phases to produce the sub-
data (at least 36 Å). Although the EXAFS data gives a cluster microscopic “invisible” gold.
size of 19 to 77 Å, the TEM work constrains the size to below The individual clusters produced in the green rust samples
50 Å. This is in good agreement with the minimum size sug- are smaller than those produced by the work of Greffie et al.
gested by the normalized XANES data (five to six shells, 36 to (1996) and no composite clusters are observed. The difference
42.96 Å) but this conclusion should be considered wholly quali- between this work and the study of Greffie et al. (1996) prob-
tative. The value for N1 for this sample is 9.13 (diameter is 30 ably lies in the methods that were used to reduce the Au3+ and
Å), which seems a slight underestimation of the immediate co- the minerals that were used in the experiments. This suggests
ordination environment because the sample contains five shells that the oxidation potential and mineral phases present in the
(equivalent to N1 of 9.63), four of which have co-ordinations system may influence the size and morphology of the gold par-
in good agreement with the predicted values. Therefore the di-
ameter of any individual metallic gold clusters and those which
TABLE 8. Summary of approximate average sizes for the gold clus-
may be forming composite clusters in this sample is between ters formed in all the samples analyzed
36 and 50 Å. Sample Min. size Å Max. size Å Composite
The dark appearance of pyrite in the TEM means that this cluster size
data cannot be used to constrain the size of the individual gold Green Rust + 0.1 wt% Au 36 50
clusters present in this sample. The EXAFS data suggests a Green Rust + 0.5 wt% Au 29 36
stibnite + 1.07 wt% Au 36 50 £320 Å
size range of 27 Å to at least 77 Å but the high value of N1 pyrite + 1.05 wt% Au 36 77 £5000 Å
(10.94) means that +20% error in N1 gives a value for N1 for chalcopyrite + 1.02 wt% Au 36 50 £440 Å
738 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS

ticles found in nature. The gold particles produced in this work essential components of the laboratory equipment, and for building it. The
EXAFS work would not have been possible without C. Muskett and M. Collings
and that of Greffie et al. (1996) are much smaller than those who helped collect data and A. Dent, F. Mosselmans, I. Harvey, and R.
observed in nature. For example some of the “invisible” gold Billsborough of the Daresbury Laboratories, Cheshire, U.K. for their invalu-
found associated with iron oxide minerals at the Kori Kollo able help with EXAFS experiments. We are grateful to D. Robinson (Bristol)
and R. Wogelius (Manchester) for many useful comments and criticism. The
deposit (Bolivia) forms irregular shaped particles approximately authors are also very grateful to G. M. Bancroft and one other anonymous re-
1000 Å in size (Darke et al. 1997). A study of the physico- viewer for their comments that have greatly improved this paper. This work was
funded by a National Environmental Research Council (U.K.) Ph.D. Studentship.
chemical and experimental conditions that produce different
sizes and morphologies of gold particles during ferrolysis may REFERENCES CITED
yield information relating to the conditions under which later- Ashley, P.M., Creagh, C.J., and Ryan, C.G. (2000) Invisible gold in ore and mineral
itic “invisible” gold forms in nature. Another factor that may concentrates from the Hillgrove gold-antimony deposits, NSW, Australia.
Mineralium Deposita, 35, 285–301.
influence the mechanism of “invisible” gold formation is the Atkins, P.W. (1994) Physical Chemistry, 5th ed. Oxford University Press, U.K.
concentration of gold in solution. Due to technical constraints Baes, C.F. and Mesmer, R.E., Eds. (1976) Hydrolysis of Cations, 1st ed. Wiley,
it is necessary to use gold concentrations at least five to six New York.
Baranova, N.N. and Ryzhenko, B.N. (1981) Computer simulation of the Au-Cl-S-
orders of magnitude higher than those found in nature. This Na-H2O system in relation to the transport and deposition of gold in hydrother-
may lead to the formation of smaller clusters since at higher mal processes. Geokhimiyia, 7, 989–1001.
concentrations the nucleation rate will be higher. Bender Koch, C. and Mörup, S. (1991) Identification of green rust in an ochre sludge.
Clay Mineralogy, 26, 577–582.
Previous SEM studies of the reaction of gold with sulfide Benfield, R.E. (1992) Mean co-ordination numbers for Non-metal-Metal transition
minerals demonstrated that almost instantaneous reduction at in clusters. Journal of the Chemical Society, Faraday Transactions, 88, 1107–
1110.
the mineral surface does occur. The sizes of the gold clusters Benfield, R.E., Grandjean, D., Kröll, M., Pugin, R., Sawitowski, T., and Schmid, G.
produced in those experiments range from approximately 2000 (2001) Structure and bonding of gold metal clusters, colloids and nanowires
Å (Mycroft et al. 1995) to tens of microns (Scaini et al. 1997). studied by EXAFS, XANES and WAXS. Journal of Physical Chemistry B,
105, 1961–1970.
In addition to this the formation of composite clusters of Benning, L.G. and Seward, T.M. (1996) Hydrosulphide complexing of Au (I) in
smaller gold clusters was observed on the surface of arsenopy- hydrothermal solutions from 150–400 ∞C and 500–1500 bar. Geochimica et
rite (200–500 Å) and pyrite (3800 Å). The individual metallic Cosmochimica Acta, 60, 1849–1871.
Berzon, R.O., Bryzgalov, I.A., Konyshev, V.O., Meshcherina, T.V., and Nekrasova,
gold clusters produced by the experiments presented here are A.N. (1999) Geological setting, mineral composition and formation conditions
between 36 and 77 Å in diameter and probably form compos- of the Kyuchus gold-mercury deposit Sakha, Russia. Geology of Ore Deposits,
41, 440–459.
ite clusters of a few hundreds of angstroms in diameter on the Bethke, C.M., Ed. (1994) The Geochemists Workbench, v2.0: A users guide to Rxn,
surface of stibnite and chalcopyrite and up to 5000 Å on the Act2, Tact, React and Gtplot. University of Illinois, Champaign-Urbana.
surface of pyrite. Binstead, N., Campbell, J.W., Gurman, S.J., and Stephenson, P.C. (1998) SERC
Daresbury Laboratory EXCURV98 program, Daresbury, U.K.
Yang et al. (1998), using TEM, observed “invisible” gold Bowell, R.J., Baumann, M., Gingrich, M., Tretbar, D., Perkins, W.F., and Fisher,
particles with an average size of 1500 Å in natural samples of P.C. (1999) The occurrence of gold at the Getchell mine, Nevada. Journal of
pyrite and stibnite but due to the relatively low magnification Geochemical Exploration, 67, 127–143.
Cardille, C.M., Cashion, J.D., McGrath, A.C., Renders, P., and Seward, T.M. (1993)
of these images (¥10 000 to ¥28 000) it is not possible to de- 197
Au Mössbauer study of Au2S and gold adsorbed onto As2S3 and Sb2S3 sub-
termine whether these particles are individual or composite me- strates. Geochimica et Cosmochimica Acta, 57, 2481–2486.
Cloke, P.L. and Kelly, W.C. (1964) Solubility of gold under inorganic supergene
tallic gold clusters. If they are composite clusters then they are conditions. Economic Geology, 59, 259–270.
of a similar size to those synthesized by Maddox et al. (1994) Cook, N.J. and Chryssoulis, S.L. (1990) Concentrations of invisible gold in the
[gold (aq) is 10–5 to 10–4 M] and some of the features observed common sulphides. Canadian Mineralogist, 28, 1–16.
Darke, K.E., Boyce, A.J., Clapperton, C.M., Fallick, A.E., Redwood, S.D., and Rice,
in this work [gold (aq) is 10–3 M]. This suggests that the con- C.M. (1997) Supergene mineralisation at the Kori Kollo gold Mine, Bolivia.
centration of gold does not affect particle size when sulfide Exploration and Mining Geology, 6, 209–221.
minerals reduce aqueous gold. This is confirmed by the study Dent, A.J. and Mosselmans, J.F.W. (1992) A Guide to EXBACK, EXCALIB and
EXCURVE92. CLRC Daresbury Laboratory internal document, Daresbury, U.K.
of Maddox et al. (1994), which shows that the rate of deposi- Enzweiler, J. and Joekes, I. (1991) Adsorption of colloidal gold on colloidal iron
tion, and therefore the size of the gold clusters, is dependant on oxides. Journal of Geochemical Exploration, 40, 133–142.
Farges, F., Sharps, J.A., and Brown, G.E. (1993) Local environment around gold
the oxidation rate of the mineral. (III) in aqueous chloride solutions: An EXAFS spectroscopy study. Geochimica
The major problem with this and previous experimental stud- et Cosmochimica Acta, 57, 1243–1252.
ies of “invisible” gold formation is that they are performed un- Fleet, M.E. and Mumin, A.H. (1997) Gold-bearing arsenian pyrite and marcasite
and arsenopyrite from Carlin Trend gold deposits and laboratory synthesis.
der ambient laboratory conditions and therefore more elaborate American Mineralogist, 82, 182–193.
experiments, mimicking the physico-chemical conditions found Fleming, C.A. (1992) Hydrometallurgy of precious metals recovery.
in hydrothermal deposits, are needed to provide a better under- Hydrometallutgy, 30, 127–162.
Fletcher, D.A., McMeeken, R.F., and Parkin, D.J. (1996) The United Kingdom
standing of the conditions under which “invisible” gold forms Chemical Database Service. Journal of Chemical and Information Computing
in these systems. Science, 36, 746–749.
Foster, R.P., Ed. (1984) Gold ’82, 1st ed. Balkema, Neatherlands.
Fritsche, H.G. and Benfield, R.E. (1993) Exact analytical formulae for mean co-
ACKNOWLEDGMENTS ordination numbers in clusters. Atoms and Molecular Clusters, 26, 15–17.
Thanks go to all those in the Earth Sciences Department at Bristol who Greffie, C., Benedetti, M., Parron, C., and Amouric, M. (1996) Gold and iron oxide
made this work possible; including A. Kemp, K. Goodman, and D. Hawkins for associations under supergene conditions: An experimental approach. Geochimica
their assistance with some of the laboratory work, including the ICP-AES analy- et Cosmochimica Acta, 60, 1531–1542.
sis, and to S. Kearns for his help with the XRD work. Thanks must also go to J. Gurman, S.J. (1988) The small atom approximation theory. Journal of Physics C:
Wylde (Interface Analysis Center) for assistance with the SEM, to J. Dimmery Solid State Physics, 21, 3699–3717.
and K. Bean (Chemistry Department) for the surface area measurements, and in Gurman, S.J., Binstead, N., and Ross, I. (1984) Single scattering curved wave theory.
the same department, to S. Davies for his help with the TEM work. We also Journal of Physics C: Solid State Physics, 17, 143–151.
thank F. Wheeler and M. Dury for their technical help in designing some of the Hannington, M.D., Tivey, M.K., Larocque, A.C.L., Petersen, S., and Rona, P.A.
 HEASMAN ET AL.: REDUCTION OF AQUEOUS Au3+ BY MINERALS 739

(1991) The occurrence of gold in sulphide deposits of the TAG Hydrothermal chlorides, Part 1, A surface study. Geochimica et Cosmochimica Acta, 59, 3351–
Field, Mid-Atlantic Ridge. Canadian Mineralogist, 33, 1285–1310. 3365.
Hedin, L. and Lundqvist, S. (1969) Effects of electron-electron and electron-phonon Nechayev, Y.A. (1984) The effects of solution composition on the adsorption of
interactions on the one-electron states of solids. Solid State Physics, 23, 1–181. gold (III) complexes on hematite. Geokhimiyia, 4, 527–533.
Henley, R.W. (1973) Solubility of gold in hydrothermal chloride solutions. Chemi- Oberthur, T., Weiser, T., Amanor, J.A., and Chryssoulis, S.L. (1997) Mineralogical
cal Geology, 11, 73–87. siting and distribution of gold in quartz veins and sulphide ores of the Ashanti
Hong, H.L., Wang, Q.Y., Chang, J.P., Liu, S.R., and Hu, R.Z. (1999) Occurrence mine and other deposits in the Ashanti belt of Ghana: Genetic implications.
and distribution of invisible gold in the Shewushan supergene gold deposit, Mineralium Deposita, 32, 2–15.
southeastern Hubei, China. Canadian Mineralogist, 37, 1525–1531. Peck, J.A., Tait, C.D., Swanson, B.I., and Brown, G.E. (1991) Speciation of aque-
Hyland, M.M. and Bancroft, G.M. (1989) An XPS study of gold deposition at low ous gold (III) chlorides from ultraviolet/visible absorption and RAMAN/reso-
temperatures on sulphide minerals: Reducing agents. Geochimica et nance RAMAN spectroscopies. Geochimica et Cosmochimica Acta, 55, 671–
Cosmochimica Acta, 53, 367–372. 676.
Jean, G.E. and Bancroft, G.M. (1985) An XPS and SEM study of gold deposition at Renders, P.J. and Seward, T.M. (1989) The adsorption of thio gold(I) complexes by
low temperatures on sulphide mineral surfaces: Concentration of gold by ad- amorphous As2S3 and Sb2S3 at 25 ∞C and 90 ∞C. Geochimica et Cosmochimica
sorption/reduction. Geochimica et Cosmochimica Acta, 49, 979–987. Acta, 53, 255–267.
Karasyova, O.N., Ivanona, L.I., Lakshtanov, L.Z., Lövgren, L., and Sjöberg, S. (1998) Rubin, J.N. and Kyle, J.R. (1997) Precious metal mineralogy in porphyry, skarn,
Complexation of gold(III)-chloride at the surface of hematite. Aquatic Geochem- and replacement-type ore deposits of the Ertsberg (Gunung Bijih) district; Irian
istry, 4, 215–231. Jaya, Indonesia. Economic Geology, 92, 535–550.
Krauskopf, K.B. (1951) The solubility of gold. Economic Geology, 46, 61–84. Santaguida, F. and Hannington, M.D. (1996) Characteristics of gold mineralisation
Machesky, M.L., Andrade, W.O., and Rose, A. (1991) Adsorption of gold (III)- in volcanogenic massive sulphide deposits of the Notre Dame Bay Area, cen-
chloride and gold (I)-thiosulphate anions by goethite. Geochimica et tral Newfoundland. Canadian Journal of Earth Sciences, 33, 316–334.
Cosmochimica Acta, 55, 769–776. Scaini, M.J., Bancroft, G.M., and Knipe, S.W. (1997) An XPS, AES and SEM study
Maddox, L.M., Bancroft, G.M., Scaini, M.J., and Lorimer, J.W. (1994) Invisible of gold and silver chloride species with PbS and FeS2: Comparison to natural
gold: Comparison of Au deposition on pyrite and arsenopyrite. American Min- samples. Geochimica et Cosmochimica Acta, 61, 1223–1231.
eralogist, 83, 1240–1245. Schoonen, M.A.A., Fisher, N.S., and Wente, M. (1992) Gold sorption onto pyrite
Mann, A.W. (1983) Hydrogeochemistry and weathering of the Yilgarn Block, West- and goethite: A radiotracer study. Geochimica et Cosmochimica Acta, 56, 1801–
ern Australia—ferrolysis and heavy metals in continental brines. Geochimica 1814.
et Cosmochimica Acta, 47, 181–190. Schwertmann, U. and Fetcher, H. (1994) The formation of green rust and its trans-
——— (1984) Mobility of gold and silver in lateritic weathering profiles: Some formation to lepidocrocite. Clay Mineralogy, 29, 87–92.
observations from Western Australia. Economic Geology, 79, 38–49. Seward, T.M. (1973) Thio complexes of gold and the transport of gold in hydrother-
Mao, S.H. (1991) Occurrence and distribution of invisible gold in a Carlin-type mal ore solutions. Geochimica et Cosmochimica Acta, 37, 379–399.
gold deposit in China. American Mineralogist, 76, 1964–1972. ——— (1984) The transport and deposition of gold in hydrothermal systems. In
Matel, N.L., Kolpakova, N.N., and Il’ina, G.F. (2000) Genetic features of the Darasun R.P. Foster, Ed., Gold 82, 1st ed., 165–181. Balkema, Neatherlands.
gold-sulphide deposit (Eastern Transbaikal Region). Geology of Ore Deposits, Shenberger, D.M. and Barnes, H.L. (1989) Solubility of gold in aqueous sulphide
42, 474–495. solutions from 150 to 350 ∞C. Geochimica et Cosmochimica Acta, 53, 269–
McHugh, J.B. (1988) Concentration of gold in natural waters. Journal of Geochemical 278.
Exploration, 30, 85–94. Spry, P.G. and Thieben, S.E. (2000) The distribution and recovery of gold in the
Mehrabi, B., Yardley, B.W.D., and Cann, J.R. (1999) Sediment-hosted disseminated Golden Sunlight gold-silver telluride deposit, Montana, USA. Mineralogical
gold mineralisation at Zarshuran. Mineralium Deposita, 34, 673–696. Magazine, 64, 31–42.
Mernagh, T.P., Heinrich, C.A., Leckie, J.P., Carville, D.P., Gilbert, D.J., Valenta, Stumm, W. and Morgan, J.J., Eds. (1981) Aquatic Chemistry, 1st ed. Wiley, New
R.K., and Wyborn, L.A.I. (1994) Chemistry of low temperature hydrothermal York.
gold, platinum and palladium (± uranium) mineralisation at Coronation Hill, Valeton, I. (1994) Element concentration and formation of ore deposits by weather-
Northern-Territory, Australia. Economic Geology, 89, 1053–1073. ing. Catena, 21, 99–129.
Michibayashi, K. (1995) Two phase syntectonic gold mineralisation and barite Vlassopoulos, D. and Wood, S.A. (1990a) Gold speciation in natural waters: I. Solu-
remobilisation within the main ore body of the Golden Giant Mine, Hemlo, bility and hydrolysis reactions of gold in aqueous solution. Geochimica et
Ontario, Canada. Ore Geology Reviews, 10, 31–50. Cosmochima Acta, 54, 3–12.
Muller, D., Franz, L., Herzig, P.M., and Hunt, S. (2001) Potassic igneous rocks from ——— (1990b) Correction. Geochima et Cosmochima Acta, 54, 1849.
the vacinity of epithermal gold mineralisation, Lihir Island, Papua New Guinea. Watkins, J.W., Elder, R.C., Greene, B., and Darnell, D.W. (1987) Determination of
Lithos, 57, 163–186. gold binding in algal biomass using EXAFS and XANES spectroscopies. Inor-
Murphy, J.A. and LaGrange, M.S. (1998) Raman spectroscopy of gold chloro-hy- ganic Chemistry, 26, 1147–1151.
droxy speciation in fluids at ambient temperature and pressure: A re-evaluation Yang, S., Blum, N., Rahders, E., and Zhang, Z. (1998) The nature of invisible gold
of the effects of pH and chloride concentration. Geochimica et Cosmochimica in sulphides from the Xiangxi Au-Sb-W ore deposit in Northwestern Hunan,
Acta, 62, 3515–3526. Peoples Republic of China. Canadian Mineralogist, 36, 1361–1372.
Murphy, J.A., Stevens, G., and LaGrange, M.S. (2000) The effects of temperature
and pressure on gold-chloride speciation in hydrothermal fluids: A Raman spec-
troscopic study. Geochimica et Cosmochimica Acta, 64, 479–494. MANUSCRIPT RECEIVED MAY 5, 2002
Mycroft, J.R., Bancroft, G.M., McIntyre, N.S., and Lorimer, J.W. (1995) Spontane- MANUSCRIPT ACCEPTED NOVEMBER 21, 2002
ous deposition of gold on pyrite from solutions containing Au (III) and Au (I) MANUSCRIPT HANDLED BY LEE GROAT