Talanta 148 (2016) 419–426

Contents lists available at ScienceDirect

Talanta
journal homepage: www.elsevier.com/locate/talanta

Determination of gold and silver in geological samples by focused

infrared digestion: A re-investigation of aqua regia digestion
Yong Wang, Laura A. Baker, Ian D. Brindle n
Department of Chemistry, Brock University, St. Catharines, Ontario, Canada L2S 3A1

art ic l e i nf o a b s t r a c t

Article history: Focused infrared radiation-based digestions, for the determination of gold and silver, can be achieved
Received 10 August 2015 in a timeframe as short as 10–15 min, making it an attractive candidate technology for the mining
Received in revised form industry, where very large numbers of samples are analyzed on a daily basis. An investigation was
4 November 2015
carried out into gold and silver dissolution chemistry from geological samples using this novel di-
Accepted 5 November 2015
Available online 6 November 2015
gestion technique. This study investigated in-depth the issue of low recoveries of gold from aqua
regia (AR) digestions, reported by a number of researchers. Conventional AR digestions consistently
Keywords: delivered gold recoveries in a range of 69–80% of the certified values for the four certified reference
Gold materials (CRM) employed (CCU-1d, SN26, OREAS 62c, and AMiS 0274), while silver recoveries were
Silver
satisfactory. By gradually shifting the HCl:HNO3 ratio (v/v) from 3:1 to a reversed 1:3 ratio, recoveries
Infrared
of gold and silver exhibited inverse trends. At a HCl:HNO3 ratio of 1:3, complete recovery of gold was
Sample digestion
Aqua regia achieved with excellent reproducibility in all CRMs. Meanwhile, silver recoveries plunged sig-
Reverse aqua regia nificantly at this ratio in samples with higher silver concentrations. Silver values were recovered,
ICP-MS however, when the silver was re-solubilized by adding a small volume of concentrated HCl to the
cooled reverse aqua regia digests. Recoveries of base metals, such as Fe and Cu, were satisfactory
throughout and were much less sensitive to changes in the digestion medium. Using four CRMs and
five real-world gold/silver containing samples, the utility of the proposed reverse aqua regia was
systematically studied. The uncomplicated nature of the digestion methods reported here, that are
fast, effective and inexpensive, may be useful to analysts developing/optimizing their methods for
the rapid determination of Au and Ag in a variety of mineral phases, particularly where rapid results
are desirable, such as in prospecting and mine development.
& 2015 Elsevier B.V. All rights reserved.

1. Introduction digestion. Aqua regia, in particular, has been employed as the

preferred sample preparation method, capable of digesting
Accurate quantification of gold and silver in geological various sample types since medieval times [1]. The perfor-
samples provides vital information to exploration geologists mance attributes of the acid mixture fit well with the criteria
in the decision-making process of whether a mine should be of elemental analysis of geological materials, such as wide
developed, maintained, or closed. The analytical procedure commercial availability, strong oxidizing ability to bring me-
demands near-total recovery of both elements for an accurate tals into solution, and relatively low cost. One of the many
estimation of the economic potential of the mine. Laboratory analytical applications developed historically, and still a
determination of the two elements, however, largely depends prevalent and effective method in modern day operations, is
its use in extracting precious metals, such as silver and gold,
on both the efficiency of extraction methods, and on the mi-
and/or strategic metals from rocks and ores in a wet dis-
neralogy of the rock samples. In current practice, both AR and
solution process [2].
a more aggressive acid attack, with the addition of hydro-
Amongst popular gold and silver extraction and determination
fluoric acid and perchloric acid, are routinely employed to
methodologies for geological materials, AR competes primarily
serve general sample preparation purposes, with the latter with fire assay (FA) [3–5]. The main reason FA has been a trusted
being termed as “total digestion (TD)”, or a “four-acid” gold extraction method within the geological community is two-
fold: (a) it is assumed to be a ‘total’ recovery method, and (b) past
n
Corresponding author. experience suggests that using a large sample size, that can be
E-mail address: ibrindle@brocku.ca (I.D. Brindle). accommodated readily by the FA process. Sample size typically

http://dx.doi.org/10.1016/j.talanta.2015.11.019
0039-9140/& 2015 Elsevier B.V. All rights reserved.
420 Y. Wang et al. / Talanta 148 (2016) 419–426

ranges from 30–50 g, and effectively reduces the risk of the Table 1
‘nugget effect’*. It should be noted that, despite being the preferred Comparison of recovered Ag (μg g
1) by different sample preparation techniquesa.
method for gold and precious metals extraction, the performance
CRM ARb TDb FUb Maximum particle Description
of FA remains laboratory- and technician-dependent. Its ability to size (μm)
extract total gold from geological materials has also been found less
satisfactory, as gold recovery by FA can sometimes be 14–26% lower OREAS 37 5.19 – 5.0 30 Sulfidic
OREAS 131a 29.5 30.9 30.2 30 Siliceous Zn–Pb–Ag
than the values obtained by instrumental neutron activation analysis
ore
(INAA) [6]. INAA is an essentially non-destructive and matrix-in- OREAS 134b 204 209 206 30 Siliceous and
dependent technique. It offers reporting of total metal concentrations sulphidic
in contrast to AR, which is a leaching process. The use of INAA is OREAS 926 2.97 2.92 o3c 75 Siliceous copper ore
OGGeo08 20.2 20.4 19.8 30 Granite-based ore
limited, particularly where time of analysis is an important factor,
blend
since other activated elements, notably sodium, must be allowed to
undergo nuclear decay for approximately seven days before the a
Data taken from Ore Research and Exploration data sheets.
b
measurement of gold can take place [7]. Aside from the need to gain AR: aqua regia, TD: total digestion, FU: fusion, sodium peroxide or lithium
borate.
access to the beam lines of a nuclear reactor, the turnaround time for c
Reported as lower than method reporting limit.
analysis fails to justify its use in high-sample throughput mining
operations. Acid digestions for gold determination serve the purpose
case-by-case analytical scrutiny. For instance, in 2007 Barrick Gold
where a cheaper but relatively reliable method is sought. This ap-
offered a $10 million “Unlock the Value” program to release silver
proach, however, requires samples to be sufficiently homogenized so
from a silicate-encapsulated silver deposit in their Veladero mine,
that a smaller sample size can be used. In the case of commercial gold
in which more than 90% of the silver was encapsulated [13].† It
reference materials, a maximum particle size that passes through a
appears that the silver mineralogy of this deposit may be un-
sieve of 200-mesh or finer is recommended to mitigate the extent of
common as the total leachable silver by a regular AR attack ac-
gold encapsulation and the nugget effect. Gold results reported pre-
counts for only a small portion of total silver.
viously by various acid treatments and sample types, however, sug-
Although AR is clearly effective for silver extraction from geo-
gest that the determination of gold in geological materials can be
logical materials [14–16], the interpretation of gold values gener-
more complex than other metals and must be dealt with on a case-
ated by AR leaching should be interpreted with caution, since gold
by-case basis. For instance, one study employed as little as 0.1–0.2 g of
recoveries can be broadly lower than the values determined by FA
relatively coarsely ground samples (∼170 μm, or 90-mesh), and
claimed excellent precision for gold and silver recoveries [8]. Where [17]. This observation has been replicated in our laboratory, re-
gold is measured in soils and alluvial deposits, Ward and coworkers gardless of the sample types and choice of digestion method
suggested that the gold nugget effect can be greatly reduced by (hotblock or ColdBlock). Hall and coworkers reported that AR ty-
performing a clay-size separation of soil and till prior to the AR di- pically produces negative reporting biases in a range of 24–42%,
gestion [9]. On the other hand, an investigation using 0.5–1 g of re- compared with INAA [6]. Wang and Brindle discussed this lim-
ference materials (SARM-7, SU-1a) in closed-vessel low-pressure itation of AR and suggested that the reduced gold recoveries with
microwave digestion yielded unsatisfactory gold recoveries and re- the AR method, using a conventional HCl/HNO3 ratio of 3:1, are
producibilities [10]. Other reports suggest that encapsulation of gold likely associated with the volatility of the gold chloride species,
in silicates samples, another source of variation in the determination Au2Cl6, at elevated temperatures [18]. This notion was originally
of gold in mining samples, may not be as common a phenomenon as reported by the US Bureau of Mines in 1971 [19].
previously thought, as little difference was found in gold concentra- It occurred to the authors that the ratio of HCl:HNO3 in AR,
tions between AR and TD methods [11,12]. In the past, there was also although sometimes modified in in-house methods for particular
a lack of consensus on the minimum number of replicates (at a given extraction purposes, has not been investigated systematically, and
set of conditions) required, which inevitably contributes to the overall that it may have a profound impact on extraction efficiency for Au
uncertainty of gold determination by acid digestion. and Ag. General trends in silver recoveries, as a function of HCl:
The notion of encapsulation of silver, though rarely reported in HNO3 ratio (v/v), were discussed in environmental samples by
the literature, may also be case-dependent, as is the case with Link et al. [20]. They proposed that silver recoveries may increase
silver in the Veladero mine, described below. Ore Research and proportionately to the concentration of free chloride ions, which
Exploration (ORE) has released a series of silver-containing certi- react with insoluble AgCl to form soluble AgCl−2 . When the acid
fied reference materials, representing a range of frequently en- ratio is reversed, however, the recovery quickly plummets as a
countered ores, and compared data from various extraction consequence of the chloride being exhausted by its oxidation to
methods such as AR, TD, and fusion (FU). The certificates reported volatile chlorine and nitrosyl chloride. In this paper, we present
no statistically significant differences in silver values between our investigations into the determination of Au, Ag, and major
these methods (Table 1), suggesting that encapsulation of Ag in base metals by AR dissolution in various acid ratios, with the ad-
those materials is not likely a common mineralogy, or the amount dition of HF where necessary, using four commercial CRMs and
of encapsulated silver appears analytically negligible at the given five real-world mining samples that represent significantly dif-
particle sizes. Results obtained from some real-world mining ferent mineralogies and ore types. Together with our continuing
samples, however, suggest that there are exceptions that require effort to develop fast and accurate digestion methods using the
infrared-based ColdBlock digester, we present data that will be
*
The nugget effect is usually described as being due to the occurrence of useful to analysts in developing or optimizing methods for rapid
discrete particles, often of pure gold in the sample. The number of these particles and accurate determination of Au and Ag in mining samples.
required to make up the concentration of gold is relatively small. Thus, from a
probabilistic perspective, a relatively large sample size is needed to provide an
†
appropriate number of these particles to give a subsample that is representative of Typically, mines grind ores to a much coarser mesh than analytical facilities
the whole and therefore generate a reasonably accurate concentration of gold. Gold do, so the potential for the silver to be associated with the groundmass of the ore,
is also difficult to grind in conventional comminution equipment, since it is ductile rather than encapsulation into a silica matrix, must also be considered. The authors
and malleable and is likely to smear on the grinding equipment, leading to losses thank Mr. Russ Calow, Vice President of Global Analytical Services, SGS, Lakefield,
from the sample. Ontario, Canada, for this valuable insight.
 Y. Wang et al. / Talanta 148 (2016) 419–426 421

2. Experimental (Table 3). CCU-1d from CANMET (Ottawa, ON, Canada) is a highly
sulfidic ore mixture containing chiefly chalcopyrite (CuFeS2,
2.1. Sample digestion apparatus – focused infrared-based digester 66.8%). SN26 came from Rocklabs (Auckland, New Zealand) and is
composed of feldspar and pyrites. OREAS 62c from ORE (Bayswater
The current study uses the prototype of the ColdBlock digester North, Australia) is a siliceous epithermal vein ore. AMiS 0274 was
described previously, to perform all sample digestions [18,21]. This obtained from African Mineral Standards (Johannesburg, South
device is capable of dramatically accelerating the digestion process Africa) and is also a siliceous epithermal vein ore. Four pulverized
by transmitting focused short-wavelength infrared radiation di- mining samples were donated by SGS and one by Barrick Gold.
rectly to the surface of sample particles, rather than heating the They were treated following their respective sample preparation
acid-sample mixture as in conventional hot-block and microwave procedures.
systems. Regular AR digestions of the four CRMs were also per-
formed using a 9-channel programmable hot-block digester (SCP 2.5. Method of sample dissolution and preparation
Science, Baie d’Urfé, QC, Canada) for gold and silver data com-
parison purpose. Reference materials were used as received from the supplier.
No additional grinding was performed. Since different miner-
2.2. Analytical instrumentation alogies may require differing energy treatment for digestion, the
proposed heating profile for each sample was based on a complete
A PerkinElmer-SCIEX (Woodbridge, ON, Canada) ELANs DRCs recovery of base metals. The reference materials and acids were
II ICP-MS was used for determining Au and Ag in all samples and mixed in a quartz tube and the individual heating profile for the
standards. Base metals were determined with a PerkinElmer chosen CRM is presented in Table 4.
(Sheldon, CT, USA) OPTIMAs DV3300 ICP-OES, following the same The total initial acid volume used was kept constant at 8 mL
analytical procedure described previously [18]. The sample in- throughout. The amount of sample weight was held at approxi-
troduction system of the DRC II ICP-MS consists of a Meinhards A3 mately 1 g. No significant nugget effects for the chosen CRMs were
type glass concentric nebulizer, a baffled quartz cyclonic spray observed at this weight and particle size levels, due to fine
chamber and a 2.0 mm (I.D.) quartz injector. Daily performance grinding (200 mesh and below). The digestion time of each ma-
checks were carried out to ensure that the baseline elemental terial was held constant for all acid ratios for comparison purposes.
sensitivities recommended by the manufacturer were met. Other Each CRM was digested with five acid ratios where the volume of
important operating parameters of the instrument are listed in HNO3 increases and HCl decreases.‡ Details of acid treatment for
Table 2. the five geological samples are described in Section 3. After the
digestates had cooled to room temperatures, they were diluted
2.3. Reagents with de-ionized water to 25 mL.

Analytical-grade (single distilled) nitric acid (68–70% m/v) and 2.6. Matrix selection, calibration, detection limit, and quality control
hydrochloric acid (36.5–38% m/v) were purchased from Caledon
(Georgetown, ON, Canada). All elemental stock solutions The strong memory effect of gold is a well-known phenomenon
(1000 mg mL
1) were purchased from High-Purity Standards in silica-based sample introduction and nebulization systems, due
(Charleston, SC, USA) to prepare calibration standards. A second to adsorption and desorption during wash-in and wash-out times
gold standard (also 1000 mg mL
1) and a multi-element QC solu-
and was described in a previous publication [22]. This problem has
tion containing Ag (QC STD4) were purchased from SCP Science
been solved by using dilute L-cysteine in 1% (v/v) HCl [18,23]. It is
(Baie d’Urfé, QC, Canada) to serve as QC solutions. L-Cysteine was
worth noting that this L-cysteine-based matrix solution shows
manufactured by Amresco (Solon, OH, USA). All sample and
superb Au stabilizing ability, as standards prepared in this matrix
working solutions were prepared with ultra-pure de-ionized wa-
show no sign of Au signal decay over a period of 180 days. The
ter (18.2 MΩ-cm resistivity) from an Elgastat-Maxima purification
wash-in and wash-out times of gold-containing solutions are as
system (High Wycombe, UK). Research-grade argon (99.998%) and
short as 3–5 s, indicating that concerns with wash-in and wash-
nitrogen (99.999%) gases were supplied by Praxair (Hamilton, ON,
out times and gold carryovers are essentially eliminated. Silver is
Canada).
stable in dilute (1% v/v) nitric acid and its wash-in and wash-out
times are not a concern. In this work, matrix matching was per-
2.4. Reference materials and Au/Ag containing samples
formed by ensuring that both gold working standards and pre-
pared samples were made up in a 0.5% (m/v) L-cysteine, 1% (v/v)
Certified gold and silver reference materials of differing mi-
HCl solution, as described below.
neralogies were obtained for evaluation of the proposed study
Since silver does not exhibit mid-term stability in an L-cysteine
solution, two individual sets of calibration standards were prepared
Table 2
Instrumental settings for ELAN ICP-MS.
at 1.0, 5.0, 10.0, 20.0 μg L
1 (ppb) levels for determination of Au and
Ag. Reduction by L-cysteine from Ag þ to Ag0 may occur, as pointed
Plasma RF power (W) 1250 out by Chen et al. [23]. Stabilization of gold by L-cysteine, both in
Plasma gas flow (L min
1) 15.0 solution and during introduction into silica or glass-based sample
Auxiliary gas flow (L min
1) 1.35
introduction systems, as described by Wang and Brindle for the ICP-
Nebulizer gas flow (L min
1) 0.82–0.84 (optimized frequently)
Operating mode Standard MS determination [18], was deemed necessary in this study, where
Sample uptake (mL min
1)  1.0 gold and silver were to be determined from the same digest. The
Interface Pt sampler and skimmer cones ICP-MS instrument was calibrated every 8–10 measurements to
Auto lens calibration On (Be 6.5 V, Co 7.5 V, In 8.5 V)
ensure that results are protected from any potential plasma drift.
Detector Dual
Scanning mode Peak hopping Calibration curves were blank-subtracted (linear through zero) and
Dwell time (ms) 1000
Sweeps 1
‡
Replicate 3 Five HCl:HNO3 acid mixtures used were 6:2 (mL), 5:3, 4:4, 3:5, and 2:6,
respectively.
422 Y. Wang et al. / Talanta 148 (2016) 419–426

Table 3
Au and Ag containing CRMs investigated.

CRM Au (μg g
1) Ag (μg g
1) Maximum particle size (μm) Description

CCU-1d 14.017 0.18 120.7 7 1.3 75 Sulfidic copper concentrate, 32.8% S

SN26a 8.543 7 0.072 19.63 7 0.37 – Siliceous matrix, 62.5% SiO2
OREAS 62c 8.79 7 0.10 8.767 0.28 20 Siliceous matrix, 60.9% SiO2
AMiS 0274 3.317 0.44b 323 7 24 54 Siliceous matrix, 67.3% SiO2, 2.0% S

a
Mesh size was not disclosed in the certificate, but confirmed by the manufacturer that the particle sizes are no larger than 75 micron.
b
Provisional data.

Table 4 3. Results and discussion

ColdBlock digestion program for four CRMs containing Au and Ag.
3.1. Evaluation of gold and silver recoveries by regular hot-block
Power (%)a and time (min) CCU-1D SN-26 OREAS 62c AMiS 0274
digestions
30% – – 1 1
35% 1 1 1 1 Regular AR (6 ml HCl þ2 ml HNO3) leaching was performed for
40% 1 9 13 13 the four CRMs (1 g per digestion) in a hot-block regulated at 95 C
45% 13 – – –
for one hour. No replenishment of the acids was needed as the
Total acid (mL) 8 8 8 8
Total time (min) 15 10 15 15 remaining acid was about three quarters of the initial acid volume.
The gold recoveries were 20–24% lower than the certified values,
a
Based on actual percent usage of maximum power of a double IR-ring di- whereas silver showed excellent recoveries and reproducibilities
gester (500 W).
(Table 5). Further trials indicated that longer digestion times re-
sulted in worse gold recoveries, which is suspected to be due to
presented a minimum correlation coefficient of 0.9999. Method the loss of gold via gold chloride volatilization. To confirm this
detection limits, based on 8 continuous measurements of the blank, hypothesis, gold spike recovery experiments, using clean silica
were determined at 6.8 ng L
1 (ppt) for gold and 2.0 ng L
1 (ppt) blank, were conducted in both hot-block and ColdBlock under
for silver. All measurements were frequently (every 6–8 samples) regular AR attack. The recoveries were in line with those obtained
checked with a 10 μg L
1 (ppb) QC solution and only data meeting from the CRMs.
75% of QC guideline were reported.
Attention should be paid for possible polyatomic interferences 3.2. Au trend in response to changing HCl:HNO3 ratio in the Cold-
that can bias gold and silver signals in ICP-MS. Polyatomic inter- Block digestion
ferences that could interfere with the gold signal include
157
Gd40Ar þ , 181Ta16O þ and 180Hf16OH þ and were scanned using Although the ability of aqua regia to extract gold from geolo-
100 ppb solutions of Gd, Ta and Hf in the matrix solution of 0.5% gical materials was well-characterized, little to nothing has been
(m/v) L-cysteine in 1% (v/v) HCl. The Gd solution returned baseline published on significantly varying ratios of nitric acid and hydro-
intensity when analyzed using ICP-MS. Ta and Hf solutions chloric acids, and specifically the use of reverse AR for this pur-
(100 ppb) did form measurable peaks at the mass to charge ratio pose. Ratios, varying from 3:1 to 1:3 HCl:HNO3 were tested and
of 197, necessitating a preliminary scan for these two elements in the recoveries of silver and gold were determined, together with
prepared sample solutions before quantification. If Ta or Hf were iron (and copper, in the case of the sulfidic copper concentrate,
present in appreciable amount in samples, a dynamic reaction cell CCU-1d) to help provide an independent measure of the re-
(DRC) or collision cell (CC)-based method would need to be de- coveries. As expected, iron (and copper) values were relatively
veloped to eliminate the interference. In the current study, Ta and consistent for all ratios of the two acids. Reverse AR yielded
Hf exist in ultra-trace amounts (sub-ppb to ppb levels) in all di- complete recovery of gold for all four reference materials. Figs. 1–4
luted samples and their respective polyatomic interferences on show that, as the relative concentration of nitric acid increases,
197
Au do not produce positive bias in any significant way. Polya- and the concentration of hydrochloric acid diminishes, the re-
tomic interferences on the silver isotopes, 107Ag and 109Ag, pri- covery of gold increases. Complex and competitive redox reactions
marily derive from 91Zr16O and 93Nb16O. Therefore, Zr and Nb take place and work to convert elemental gold to stable tetra-
should be pre-scanned in sample solutions to evaluate whether or chloroaurate(III) species (reaction (1)). Chloride is critically needed
not either or both silver isotopes suffer interference. For quality to create chlorine and nitrosyl chloride to oxidize gold to its higher
control purpose, it is strongly recommended that 107Ag and 109Ag oxidation state, Au(III), which is thermodynamically more stable
to be determined simultaneously so that interference on a silver than Au (I), since recovery suffers when nitric acid is used alone
isotope can be readily identified. All four CRMs and five samples [20,24]. When the relative concentration of nitric acid increases
investigated were scrutinized for the abovementioned polyatomic further,/chloride in solution becomes exhausted as a result of re-
interferences and the determination was able to be carried out in action (2), and this is evidenced by the pronounced decline in
standard mode ICP-MS, since no significant polyatomic inter-
ference was detected for the analytes. Table 5
No internal standard was used for the determination of either Au and Ag results (μg g
1): regular aqua regia by hot-block digestiona.
gold or silver due in part to the fact that a suitable internal stan-
CRM Au certified Au determined Ag certified Ag determined
dard for these analytes is difficult to choose. To attenuate the risk
of ICP-MS signal drift due to changes in the plasma conditions that CCU-1d 14.01 70.18 117 1 120.7 7 1.3 1187 5
could cause run-to-run variability, regular calibration and quality SN26 8.543 70.072 6.5 7 0.4 19.63 7 0.37 19.5 7 1.0
control procedures were performed. If the QC results did not fall OREAS 62c 8.79 70.10 6.3 7 0.4 8.767 0.28 8.9 7 0.2
AMiS 0274 3.31 70.44 2.6 7 0.3 323 7 24 320 7 20
within a controlled fluctuation range (75%) the system was re-
calibrated and the samples were analyzed again, followed by an- a
Data points shown in Table 5 reflect triplicate analysis and 95% confidence
other QC check. interval.
 Y. Wang et al. / Talanta 148 (2016) 419–426 423

silver recoveries. The predominant nitrate ions in solution replace

chloride ligands in the previous AuCL−4 ion to form a new tetra-
nitratoaurates (III) species, Au(NO3)−4 (reaction (3)), which is ther-
modynamically stable under highly oxidizing conditions [25,26].
Au(s) + 3HNO3 + 4HCl → AuCl−4(aq) + H3O+(aq) + 3HNO2 + 2H2 O (1)

HNO3 + 3HCl → NOCl(g) + Cl2(g) + 2H2 O (2)

Au(3aq
+ − −
) + 4NO3(aq) → Au(NO3)4(aq) (3)

Some relatively large variances from triplicate analyses were

observed, particularly in those digestions where sample volumes
of the two acids used were close (e.g. 3:5, 4:4, and 5:3). The gold
variance becomes small when the acid ratio reaches conventional
Fig. 1. Recoveries vs. AR ratios for CCU-1d. Data points in Figs. 1–4 are based on the
mean of triplicate determinations and their standard deviations.
and reverse AR ratios. This interesting improvement in the ob-
served gold reproducibility suggests that, at intermediate ratios,
the gold dissolution reactions may be more susceptible to subtle
changes in reaction conditions.
The theory that gold can be encapsulated within silicate lattices
in significant amounts has been a topic of debate [27,28]. Some of
the reference materials that were used in this investigation were
rich in silicate, making them possible candidates for this phe-
nomenon. The complete recovery of gold and the relatively small
standard deviation of the results suggest that silicate-gold en-
capsulation is not of concern for any of these reference materials.
The complete recovery on the other hand suggests that the par-
ticle size of all materials was sufficiently fine for complete ex-
traction or that the particles of that size ( o75 μm) were suffi-
ciently permeable to the solution components to allow the dis-
solution and extraction of gold [29]. Encapsulation may occur
through a different mechanism during the digestion itself. For-
Fig. 2. Recoveries vs. AR ratios for SN26.
mation of insoluble materials, specifically the formation of sulfur
beads in CRM, CCU-1d, occurred by the oxidation of sulfide by
nitric acid during the initial phase of the digestion (CCU-1d con-
tains 33% sulfur). The formation of elemental sulfur is common
and expected when an AR leach is performed on a sulfur-rich
material. These beads can be prevented by various processes such
as roasting, prior to digestion; the use of HBr instead of HCl [18];
or use of perchloric acid. This avenue was not pursued, however,
since the objective was to observe the AR trends for ores of varying
mineralogies. Although it is possible that some small portion of
gold may have been incorporated into the sulfur beads, the
achievement of essentially complete recovery of gold from CCU-1d
suggests that potential gold encapsulation in sulfur beads is
negligible.
Liberation of volatile species could potentially be a reason for
loss of gold during conventional AR digestion. Although the Peltier
Fig. 3. Recoveries vs. AR ratios for OREAS 62c. cooling block should condense some of the vapors as they are
being emitted, since the vessel is semi-open, some volatile species
may be lost via evaporation and/or carried in nitrogen dioxide gas
that is released during the digestion. As the concentration of
chloride in the solution increases so might the formation of vo-
latile gold–chloride complexes (AuCl3 and Au2Cl6). This chemistry
is summarized at length by Puddephatt [30]. In brief, the halides of
gold in two oxidation states, 1 and 3, undergo redox processes in
response to the presence of free chlorine. This chemistry has been
used to prepare AuCl3 (Au2Cl6) and AuCl (Au2Cl2). The presence of
free chlorine is important in this reaction but, as the level of
chlorine falls to zero, another reaction takes over in which AuCl
disproportionates to form Au and AuCl3 even at room temperature.
Thus there are two modes by which gold may be removed from
solution; by volatilization and/or by precipitation. The large excess
of nitric acid used in the reverse aqua regia digestion would give
Fig. 4. Recoveries vs. AR ratios for AMiS 0274. rise to a stable complex ion of Au(NO3)−4 in the absence of chloride
ions [25]. Also, the US Bureau of Mines noted that gold losses
424 Y. Wang et al. / Talanta 148 (2016) 419–426

through heating of gold silicates in the presence of pyrite and Table 6

chlorine might be attributed to the volatilization of FeAuCl6 near Au and Ag results (μg g
1): reverse aqua regia followed by HCl additiona.
250 C [19]. This species could be synthesized under the conditions
CRM Au certified Au determined Ag certified Ag determined
used in this investigation and would consume little iron, since
there is a preponderance of it, relative to gold, in all of the samples CCU-1d 14.01 70.18 14.4 7 0.6 120.7 7 1.3 1257 6
investigated. SN26 8.543 70.072 8.3 7 0.5 19.63 7 0.37 207 1
OREAS 62c 8.79 70.10 8.9 7 0.6 8.767 0.28 9.0 7 0.6
Although it is assumed that these trends are inherent for all AR
AMiS 0274 3.31 70.44 3.40 7 0.05 323 7 24 320 7 20
digestion methods, if evaporation plays a significant role in the
loss of gold, then closed-vessel methods may yield different re- a
Data points shown in Table 6 reflect the means of triplicate analyses and their
sults, but if both reactions are likely at play, then lower recoveries 95% confidence interval.
would be anticipated. It should be noted, however, that, if Teflons
liners are used, there is the potential for volatile species to be re-solubilize previously precipitated AgCl by forming AgCl−2 ions,
trapped in the pores of that medium that are not readily released while having no analytically significant effect upon Au(NO3)−4 ions
back into the digestion solutions. This phenomenon was observed in solution. This revised reverse AR method was tested on the four
in the pressurized closed vessel microwave digestions of geologi- reference materials and yielded satisfactory results for both gold
cal materials for gold and mercury determinations [10,31,32]. In and silver (Table 6).
the case of the determination of mercury in coal [31], quartz
sleeves prevented this migration, and full recoveries of mercury 3.6. Statistical analysis of the analytical data of gold
were observed.
Method repeatability and accuracy in the current study were
3.3. Ag trend in response to changing HCl:HNO3 ratio satisfactory following the procedures in “Assessment of Laboratory
Performance” recommended by CANMET [35]. However, the ratio
Figs. 1–4 show that good silver recovery was only achieved in of the experimental standard deviations to the published be-
all four CRMs when regular AR was used. This trend supports the tween-laboratories standard deviations, which is used to assess
need for a complexing agent for the extraction of silver, as dis- the minimum replication required for validity of accuracy test,
cussed in the introduction. In the presence of higher chloride varies significantly among the CRMs. This variation is mainly due
concentrations, silver will pass into solution as AgCl−2 [33,34]. For to the uncertainty associated with sample mean, which is a sam-
this reason, the dissolution of silver does not solely depend on the ple-specific characteristic. Overall we suggest that if one-gram
AR ratio but also specifically on the ratio between silver and sample size for gold analysis is to be used, a minimum 10–12 re-
chloride ions at the end of the digestion. On the other hand when plicates would be required to adequately present the sample mean
the chloride concentration is reduced, insoluble AgCl remains according to CANMET's guideline (Table 7).
uncomplexed. The dramatic drop of silver recoveries for CCU-1d,
SN 26 and AMiS 0274 (Figs. 2–4) suggests that there is a critical 3.7. Evaluation of the ‘nugget effect’ and encapsulation of Au and Ag
chloride concentration below which silver is not solubilized, and in real-world samples
which would depend both on the AR ratio and the silver to
chloride ratio. The less dramatic descent shown in Fig. 3 for OREAS Aside from the interesting digestion chemistry that reverse
62c can be explained by its relatively low silver concentration aqua regia and its derivative methods revealed using standard
(8.76 μg g
1) for which there is insufficient silver, and chloride reference materials, the effectiveness and applicability of these
concentrations to exceed the solubility product (Ksp methods would be better understood from those samples which

10
(AgCl) ¼ 1.8  10 ), hence the silver is not precipitated. feature real-world sample diversity and complexity. Four gold-
containing ore samples and one gold and silver bearing sample
3.4. Base metals recoveries were received from two independent mining laboratories for
evaluation. Data generated by fire assay, INAA, as well as various
The purpose of monitoring base metal recoveries in this study derivatives of AR methods used in the current study were tabu-
was to provide a reference point to evaluate the degree that the lated for comparison. In addition, after being leached by reverse
mineral structure has been destroyed by the acid attack. Although AR, three residual samples were also collected, dried, and well
iron was completely recovered for SN26, OREAS 62c and AMiS0274 mixed before submission to Activation Laboratories (Ancaster,
at all acid ratios, CCU-1d shows a slightly increasing trend in re- Ontario, Canada) for INAA analysis.
coveries of both copper and iron, similar to that of gold (Fig. 2), as In Table 8 it can be seen that the mean gold values agree well
the HCl:HNO3 ratio gradually changes from 3:1 to 1:3. This trend between different extraction methods. However, about 10–20%
can largely be attributed to the different oxidizing power of the relative standard deviations were reported for the reverse AR
acid medium, that has a decisive impact on how efficient the method, exhibiting markedly wider gold ranges than those seen
sulfide can be removed and metals be exposed for acid attack.
Table 7
Recommended minimum replication for validity of
3.5. Simultaneous recovery of Au and Ag by a variant of reverse aqua accuracy test.
regia
a
SwL/rLm Minimum n
Based on the gold and silver trends summarized above, a fast
r 0.33 1
and convenient remedy to reverse aqua regia (HCl:HNO3 1:3) 0.5 3
method was proposed, aiming to generate satisfactory results for 0.67 5
both gold and silver from the same digestion run. The new method 1 10
proceeded with a two-step procedure, an initial 10 minutes re- 1.5 22

verse AR at moderate heating power, followed by addition of 2 mL a

SwL: standard deviation of the replicates, sLm:
of concentrated HCl after the digestate has cooled down. The the between-lab standard deviation of the certified
purpose of this method was to provide sufficient free chloride to value.
 Y. Wang et al. / Talanta 148 (2016) 419–426 425

Table 8 ores. Reverse AR significantly outperformed regular AR in re-

Determination of Au in mine samples by selected extraction methods. covering gold from geological materials, thanks to the formation of
stable tetranitratoaurate ions in solution and attenuation of vola-
Sample Reverse ARa Reverse FAb INAAc INAA
(μg g
1) AR þ HCla (residual)c tile gold chloride species. Full silver recovery, on the other hand,
can be readily achieved by re-solubilizing precipitated silver ha-
SGS A 8.0 7 0.7 – 8.256 10.2 – lides by addition of sufficient chloride, except when the silver is
SGS B 87 2 – 7.258 7.97 –
encapsulated. Aside from the proper chemistry treatment of gold
SGS C 3.2 7 0.6 – 3.411 3.205 0.065
SGS D 4.0 7 0.9 – 4.285 3.46 0.034 and silver containing ores, the overall quality of simultaneous gold
BGV 4.6 7 0.6 4.8 7 0.3 4.715 3.84 0.092 and silver determination, however, is also linked to sample weight,
particle size, sample homogeneity, and mineralogy. To minimize
a
Based on 6 replicate analyses, 95% confidence interval at 1-gram sample size the ‘nugget effect’ of gold in particular, a minimum degree of
and a 10-min digestion.
b
Results from a single 30-gram fire assay run.
sample comminution and pulverization (200-mesh) should be
c
Results from a single 10-gram INAA run. maintained. A reasonable number of replicates for gold analysis
should also be studied should a one-gram sample size to be
standardized in the method. An increasingly significant ‘nugget
Table 9 effect’, analytically manifested as higher relative standard devia-
Determination of Au and Ag in mining ore samples BGV. tions, was also observed in real-world gold samples due to the
uneven particle sizes (typically, mining samples are reported as
Method (n ¼6) Au (μg g
1) Ag (μg g
1)
80–85% o200 mesh) and inhomogeneity of the sample. This in-
Reverse AR (HCl:HNO3 1:3) 4.6 7 0.6 4.7 7 0.6 creased uncertainty in gold values may be improved, however, by
Reverse AR/HCla 4.8 7 0.3 307 2 using a larger quantity of sample in acid digestion, if better com-
Regular AR (HCl:HNO3 3:1) 3.6 7 0.2 317 1 minution and improved homogenization are not possible, or in-
Reverse AR þHF/HClb 4.7 7 0.4 51 74
creasing the number of sample digestions to better delineate the
Fire assayc 4.72 50.0
INAAd 3.84 46 gold range in an unknown sample.

a
Method was the same as described in Section 3.5. Total digestion time 10 min.
b
Method modified from 3.5 by adding 2 ml of HF in the reverse AR stage.
c
Acknowledgments
Results from a single 30-gram FA run.
d
Results from a single 10-gram INAA run.
Financial support for this research project, provided by the
Ontario Centres of Excellence (OCE), the Natural Sciences and
from standard reference materials, which typically range within Engineering Research Council of Canada (NSERC), the Centre of
5%. This significant decrease in gold reproducibility suggests that Excellence for Mining Innovation (CEMI), Barrick Gold, and
the nugget effect could pose an underlying risk when the sample Chemquant Technologies (operating as ColdBlock Technologies)
size for gold determination is relatively small. The nugget effect under grants numbered OCE 11537 and NSERC CRDPJ 445997-12,
may also speak directly to samples with particle sizes less uni- is gratefully acknowledged. SGS Canada and Barrick Gold kindly
formly distributed. In the current study, the five mining samples contributed gold and silver samples for data comparison. The Ca-
were known to have about 85% mass passing through a 200-mesh nada Foundation for Innovation and the Ontario Research Fund
sieve. provided funds for the purchase of the SCIEX Elan ICP-MS (DRC II).
Encapsulation of Au and Ag in pulverized ore samples was also The OPTIMA 3300DV ICP-OES was donated by PerkinElmer. The
assessed using limited data. Table 8 provides some INAA data from authors are especially grateful to an anonymous reviewer who
analyses of gold ore residuals after reverse AR leach. The mean contributed many valuable and constructive suggestions to im-
values obtained by reverse AR are comparable with the FA and prove the manuscript.
INAA results, and there was an insignificant amount of gold in
digestion residuals as measured by INAA. Table 9 also compares
gold and silver concentrations detected by reverse AR and total References
digestion methods for sample BGV. Amongst three residual sam-
ples examined, no evidence was found to signal gold encapsula- [1] J.O. Marsden, C.I. House, The Chemistry of Gold Extraction, Society for Mining,
Metallurgy, and Exploration, Inc., Littleton, CO, USA (2006), p. 14.
tion in a statistically significant amount. Silver encapsulation,
[2] T.T. Chao, R.F.J. Sanzolone, Decomposition techniques, J. Geochem. Explor. 44
however, was discovered to a significant extent in one sample. (1–3) (1992) 65–106.
From extensive investigation on the one gold- and silver-bearing [3] C.H. Branch, D. Hutchison, Comparison between isobutyl methyl ketone and
diisobutyl ketone for the solvent extraction of gold and its determination in
samples, BGV, total leachable silver was determined to be around
geological materials using atomic absorption spectrometry, Analyst 111 (2)
60% of total recoverable silver. When a TD method is employed, (1986) 231–233.
100% silver can be recovered, indicating roughly 40% of en- [4] A. Frimpong, B.J. Fryer, H.P. Longerich, Z. Chen, S.E. Jackson, Recovery of pre-
capsulated silver even at the current state of grinding cious metals using nickel sulfide fire assay collection: problems at nanogram
per gram concentrations, Analyst 120 (6) (1995) 1675–1680.
(Tables 8 and 9). [5] R. Juvonen, E. Kallio, T. Lakomaa, Determination of precious metals in rocks by
inductively coupled plasma mass spectrometry using nickel sulfide con-
centration. Comparison with other pre-treatment methods, Analyst 119 (4)
(1994) 617–621.
4. Conclusion [6] G.E.M. Hall, J.E. Vaive, J.A. Coope, E.F. Weiland, Bias in the analysis of geological
materials for gold using current methods, J. Geochem. Explor. 34 (2) (1989)
157–171.
The effects of conventional AR and modified AR digestions on
[7] E.L. Hoffman, J.R. Clark, J.R. Yeager, Gold analysis- fire assaying and alternative
the determination of gold and silver in geological materials were methods, Explor. Min. Geol. 7 (1–2) (1998) 155–160.
carefully examined. Reverse AR and its derivatives, together with [8] J. Alvarado, A. Petrola, Determination of cadmium, chromium, lead, silver and
infrared-based ColdBlock digestion technology, provide fast, con- gold in Venezuelan red mud by atomic absorption spectrometry, J. Anal.
Spectrom. 4 (5) (1989) 411–414.
venient, and inexpensive alternatives to the existing methods for [9] B.C. Ward, M.I. Leybourne, D.N. Sacco, Drift Prospecting Within the QUEST
accurate assessment of gold and silver concentrations in rocks and Project Area, Central British Columbia (NTS 093J): Potential for Porphyry
426 Y. Wang et al. / Talanta 148 (2016) 419–426

Copper-gold, Volcanic Massive Sulfide Mineralization and Gold-copper Veins, [22] C.A. Williams, F.R. Abou-Shakra, N.I. Ward, Investigations into the use of in-
Geoscience BC Summary of Activities 2010, Geoscience BC, Report 2011-1, ductively coupled plasma mass spectrometry for the determination of gold in
2011, pp. 73–96. plant materials, Analyst 120 (2) (1995) 341–346.
[10] M.M. Totland, I. Jarvis, K.E. Jarvis, Microwave digestion and alkali fusion pro- [23] W.-C. Chen, P. Wee, I.D. Brindle, Elimination of the memory effects of gold,
cedures for the determination of the platinum-group elements and gold in mercury and silver in inductively coupled plasma atomic emission spectro-
geological materials by ICP-MS, Chem. Geol. 124 (1–2) (1995) 21–36. scopy, J. Anal. Spectrom. 15 (2000) 409–413.
[11] A. Chow, F.E. Beamish, An experimental evaluation of neutron activation, wet [24] I. De La Calle, N. Cabeleiro, M. Costas, F. Pena, S. Gil, I. Lavilla, C. Bendicho,
assay and fire assay methods of determining gold in ores, Talanta 14 (2) (1967) Ultrasound-assisted extraction of gold and silver from environmental samples
219–231. using different extractants followed by electrothermal-atomic absorption
[12] V. Balaram, K.V. Anjaiah, Direct estimation of gold in geological samples by spectrometry, Microchem. J. 97 (2) (2011) 93–100.
inductively coupled plasma mass spectrometry, J. Indian Chem. Soc. 74 (1997) [25] C.C. Addison, G.S. Brownlee, N. Logan, Tetranitratoaurates(III) : preparation,
581–583. spectra, and properties, J. Chem. Soc. Dalton Trans. 14 (1972) 1440–1445.
[13] Barrick Annual Review, 2007. 〈http://www.barrick.com/files/doc_financials/ [26] T.R. Dulski, A Manual for the Chemical Analysis of Metals, American Society for
2007/Annual%20Reports/2007%20-%20Innovation_v001_c45p31.pdf〉 (ac- Testing and Materials, West Conshohocken, PA, USA (1996), p. 75.
cessed 04.08.15). [27] G. Senanayake, Gold leaching in non-cyanide lixiviant systems: critical issues
[14] L. Hämäläinen, K. Smolander, P. Karttunen, Direct determination of silver in on fundamentals and applications, Miner. Eng. 17 (6) (2004) 785–801.
geological reference samples after aqua regia leach by graphite furnace atomic [28] J.T. Nash, W.C. Utterback, J.A. Saunders, Geology and Geochemistry of the
absorption spectrometry with matrix modification, Fresenius Z. Anal. Chem. Sleeper Gold-Silver deposit, Humboldt County, Nevada – an Interim Report, US
330 (2) (1988) 107–110. Department of Interior, USA, 1989.
[15] J.S. Barros, Determination of antimony, arsenic, bismuth, cadmium, cobalt and [29] T.A. Delaney, W.K. Fletcher, Efficiency of cyanidation in gold exploration using
silver in complex sulphide minerals by flame atomic absorption spectrometry, soils, J. Geochem. Explor. 66 (1–2) (1999) 229–239.
Analyst 114 (3) (1989) 369–373. [30] R.J. Puddephatt, The Chemistry of Gold, Monograph 16: Topics in Inorganic
[16] B. Ebarvia, E. Macalalad, N. Rogue, I. Rubeska, Determination of silver, cad- and General Chemistry, Elsevier, Amsterdam (1980), p. 33–34.
mium, selenium, tellurium and thallium in geochemical exploration samples [31] J. Xiao, Sample Preparation and Heavy Metal Determination by Atomic Spec-
by atomic absorption spectrometry, J. Anal. Spectrom. 3 (1) (1988) 199–203. trometry (M.Sc. Thesis), Brock University, St. Catharines, ON, Canada, 2004.
[17] M. Leybourne, S. Rice, Determination of gold in soils and sediments by fire [32] J. Liu, W. Goessler, K.J. Irgolic, Mercury determination with ICP-MS: signal
assay or aqua regia digestion: choosing the optimal method, Explore 158 suppression by acids, Fresenius J. Anal. Chem. 366 (1) (2000) 48–53.
(2013) 1–10. [33] EPA method 3051A, Microwave Assisted Acid Digestion of Sediments, Sludges,
[18] Y. Wang, I.D. Brindle, Rapid high-performance sample digestion for ICP de- Soils, and Oils. 〈http://www.epa.gov/osw/hazard/testmethods/sw846/pdfs/
termination by ColdBlock™ digestion: part 2: gold determination in geological 3051a.pdf〉 (accessed 20.05.15).
samples with memory effect elimination, J. Anal. Spectrom. 29 (2014) 1904–1911. [34] R.R. Brooks, J. Holzbecher, D.E. Ryan, H.F. Zhang, A rapid method for the de-
[19] Gold Transport by Complex Metal Chloride Vapors, Bureau of Mines, Report of termination of Au and Ag in sulphide ores and rocks, At. Spectrosc. 2 (1981)
Investigation 7489, US Department of the Interior, 1971. 151–154.
[20] D.D. Link, P.J. Walter, H.M. Kingston, Development and validation of the new [35] Assessment of Laboratory Performance with CCRMP Certified Reference Ma-
EPA microwave-assisted leach Method, Environ. Sci. Technol. 3051A (32)
terials. 〈http://www.nrcan.gc.ca/sites/www.nrcan.gc.ca/files/mineralsmetals/
(1998) 3628–3632.
pdf/mms-smm/tect-tech/pdf/asscrm-eng.pdf〉 (accessed 28.10.15).
[21] Y. Wang, R. Kanipayor, I.D. Brindle, Rapid high-performance sample digestion
for ICP determination by ColdBlock™ digestion: part 1:environmental sam-
ples, J. Anal. Spectrom. 29 (2014) 162–168.