Materials Chemistry and Physics, 14 (1986) 259-261 259

A NOVtiL ELtiCTROREPINING METHOD FOR GOLD ALLOYS

B. CHATTERJEE

Lucas CAV Limited, Materials Laboratory, P.O.Box 36, Warple Way,

London W3 7SS (U.K.)

Received 31 July 1985; accepted 16 September 1985

SUMMARY

The separation of gold-silver alloys into component metals by

electrorefining was investigated using packed-bed electrodes in
fused sodium chloride at 1170°C in an inert atmosphere. The
electrolytic cell, a closed system, performed reliably and
reproducibly with-no evolution of chlorine or any other gas during
the electrolysis of Au-20An alloy. A nolarisation effect at the
cathode due 50 sodium vapoir formation was investigated. The
current efficiency of the process is estimated to be low (68%)
which could primarily be due to a single-orifice metal entry
mechanism and also the loose packing of graphite granules making
up the anode bed.

INTRODUCTION

The conventional process of refining gold alloys (i.e. the

Miller process) involves chlorination of the molten alloys whereby
the chlorides of the reactive alloying constituents are formed in
preference to the thermodynamically unstable gold chloride. The
chlorides thus formed will either vapourise during chlorination
or will float on the surface of the refined gold. The reaction
rates with chlorine are fast owing to good gas-liquid contact and
stirring of the bath by the rising bubbles. Although the
efficiency of the process is high at a large concentration of the
reactive element, it decreases considerably at low concentrations.
In addition, there are obvious pollution and corrosion problems
with the escape of free chlorine and chloride products.
In contrast to chlorination, fused-salt electrolysis is
generally pollution-free and can separate an alloy into its

0254-0584/86/$3.50 0 Elsevier Sequoia/Printed in The Netherlands

260

constituent elements quite effectively. The major advantages of

Using fused salts are their high conductivities and diffusivities
compared to aqueous electrolytes (Wohlwill process), coupled with
low viscosities. This allows very high currents to be passed
through the electrolyte with only a modest drop in voltage across
the electrolyte leading to low energy consumption. Present day
cells using fused-salts have long anode to cathode paths, and at
high current densities depletion of the most reactive element at
the anode/electrolyte interface occurs causing contamination of the
cathode metal. However, use of a packed bed of graphite granules
in a fused-salt cell has overcome these and other difficulties [I,
21. In such a cell, as the molten metal flows through the bed, it
is dispersed into fine droplets and streams ensuring excellent
mixing, good mass transfer and high energy efficiency. The
technique has so far been applied to low-melting alloys (<7OO"C)
such as Pb-Sn, Pb-Zn, Al-Mg and Bi-Zn-Pb systems p-31 . This is
presumably due to both construction and operation difficulties at
high temperatures. In the present paper, the technical feasibility
of a packed-bed fused-salt electrolysis of a gold-silver alloy of
Au-2OAg composition is reported at 117O'C. For simplicity, the Cell

was operated on a batch basis only.

EXPERIMENTAL
Cell Design

The cell (Big.1) basically consisted of a packed bed of

conducting granules of graphite with a vertical ceramic diaphragm
separating the anode and cathode chambers. From a survey of
variables that could affect the mass transfer rate withinthin
films of liquid flowing over a bed of packed solids and of the
factors that may affect electrochemical transfer in such systems,
it was decided that the cell should be tall and narrow to minimize
inter-electrode distance, to maximize residence time of the molten
anode metal, to provide an adequate pre-mixing length and to
maximize granule surface area. An anode chamber of 0.6 m high and
0.45 m internal diameter was used which contained a graphite anode
and a diaphragm of 8.8 X 10 -3 ,2 and 0.0283 m2 surface areas
respectively.

Cell Material

The materials for the cell (Table I) were chosen from a

preliminary work on their stability in fused chlorides at high
 261

Impure gold

r-
L+

w
Cathode
Anode chamber/cathode
chamber
(

Fused
salt

Graphi ite anode

t
Spreader
plate

Ref rat tory base

Metallic
impurities
Yig. 1. Schematic diagram of a packed-bed electrorefining cell.

temperatures. An argon atmosphere was used during electrolysis in

order to avoid oxidation of the cell components. This was particu-
larly important to minimize erosion of the cathode chamber of
carbon-bonded silicon carbide which also served as a container for
the electrolyte. The current collector for the cathode was an
Inconel plate placed at the base of the cathode chamber (not shown
in Fig. 1). A porous silicon carbide crucible with a Refrasil sleeve
262

Table I . Average Analyses ($) of i"iatesials

Used in the Cell.

Composition Cathode/cathode Anode chamber Diaphragm

chamber material
(Suprex EC 323) (Tsisngle 84E) (Refrasil)

sic 45 84
c 30
A1203 4 5
SiO2 32 9 99.3
Fe203 3 0.1 0.07
Ti02 0.1 0.17
CaO/MgO 0.3 0.18
Na20/K20 0.9 0.01
Unknown 6* 0.6 0.02**
* **
mainly Cu, MO and alkali metals '2'3

(TableI) acted as a diaphragm and contained a graphite rod for use

as the current feeder (anode).

Electrolyte
Generally the fused salt in the cell is expected to be a salt
of the metal to be refined. But it was considered that the refining
kinetics at a high temperature such as 1170°C would be fast enough
not to require any salt of the precious metals. A most commonly
available inexpensive salt such as sodium chloride was used which
is highly conducting and has a decomposition potential not low
enough to interfere with the refining process.

PROCEDU~
The cell was heated under vacuum to 800°C, when the salt was
added under an argon blanket. Beating was continued under argon
until the anode chamber reached 1170°C. A rod (0.003 m diameter)
of Au-2OAg alloy was pushed into the cell at lOg/min througha
neoprene-sealed single-orifice entry mechanism. Electrolysis Was
carried out at a constant current for 15 minutes using a Farnell
rectifier (H30/100 model). The reason for keeping the current and
not the voltage cnstant ,is that the latter is a function of variab-
les such as anode-cathode spacing which may alter from cell to
cell. At the end of the test, the furnace was switched off to cool
down the cell, in order to recover the metals for analysis.
 263

KKXJLTS

All results (TableII) were obtained after one pass only.

TableII. Results on a packed;bed fused-salt electrorefining of

Au-20Ag alloy at 1170°C in an argon atmosphere (data are
for one pass only)

Run UC supply % of metal Estimated Current Alloy

Current Average retention by residence efficiency composi-
voltage the granules time of metal in terms tion
in the anode of silver after
chamber removal electro-
refining
(A) (v) (min)

1 10 1.6 48 7.1 69 Au-l 6Ag

2 10 1.6 46 7.2 68 Au-l6Ag
3 20 3.1 45 6.1 68 Au-l 2Ag
4 40 5.3 14 2.2 39 Au-IOAg
5 40 5.3 50 6.2 69 Au-8Ag

Reproducibility of results at any current is typically shown in

runs 1 and 2 at IOA. The average current density on the graphite
anode was 4500 A/m2 at 40A. From an analysis of the silver content
of the metal in the anode chamber, the rate of silver removal is
estimated to be 1.84 g/min compared to the theoretical rate of
2.68 g/min at WA. The residence time of the molten alloy in the
graphite granules averaged around 7 minutes. A longer residence
time would obviously prolong the electrolysis leading to efficient
refining. The residence time is affected by the density of packing
of the granules, as shown in runs 4 and 5 where under identical
conditions, a looser packing hence lower residence time in the
anode chamber produced inferior current efficiency and refinement.

DISCUSSION

Theoretical Consideration

The gold compounds are fairly unstable. The thermodynamic data

show that gold chlorides are relatively unstable even above 5OO*C
compared to the free energy of formation of about 70 kJ/mol for
silver chloride at 12OO*C [41 . This difference is considered
264

adequate to permit separation of silver from gold by electrolysis

in which the more reactive silver will electrotransport to the
cathode leaving gold behind in the anode chamber.
Based on eqn.(l) of appendix 1, the electrode potential of
silver in the alloy increases as silver is gradually electrolysed
out from the alloy. This theory is supported by Grube and Hantlem-
an D] who demonstrated that in a thermally equilibrated system,
the voltage needed to maintain a constant current density increased
during the removal of a more reactive base metal. Eventually a
condition will be reached when the ionisation potential of the
more noble metal is exceeded with the result that ions of both
metals will start to liberate. It is estimated (appendix 2) that
gold is likely to ionise along with silver when the silver content
in Au-Ag alloy is reduced down to about 0.0%

Polarisation Effect
An initial experiment was performed using graphite granules in
the anode chamber only. It was noticed that the applied voltage
increased faster during electrolysis than expected, with a
consequent joule heating and a rapid rise in temperature from
117O'C to l25O*C in IO min of electrolysis. The sharp rise in cell
voltage could originate from poor electrical contacts and/or
electrochemical effects due to electrode polarisation. Since the
electrical connections were checked, the voltage rise is considere-
d to be due to polarisation. Any anodic polarisation due to
chlorine gas evolution at the anode is considered negligible,
because based on thermodynamics, chlorine gas cannot liberate
without any anodic dissolution of gold, and there was no detection
of gold loss or chlorine gas from the anode chamber. The cathodic
polarisation, on the other hand, can originate from a formation of
a sodium vapour blanket at the cathode. In the absence of any other
salt in the molten electrolyte of sodium chloride, Naf ions are
formed in the cathode chamber at the start of eleCtrOlYSiS and are
discharged at the cathode before my Ag’ ions could arrive from the
anode. The time lag between the arrival of silver ions and the
discharge of sodium ions at the cathode is believed to be long
enough to allow some deposition of sodium on the cathode. Since
the operating temperature of 1170°C is too high for sodium to stay
in liquid state, a vapour blanket of sodium is formed on the cath-
ode resulting in cathodic polarisation, hence an increase in voltage.
This effect would predominate as the applied current is raised.
Such a phenomenon was observed in the present work with a sharp
 265

rise in voltage at a higher current when more Na+ ions would be

expected to be discharged at the cathode. The behaviour of the cell
was thus investigated at various temperatures without any metal
addition. A sharp rise in voltage at a high current was observed
only when the experiment was performed above 883OC which is the
boiling point of sodium. It is therefore obvious that the effect is
due to polarisation resulting from a layer of sodium vapour at the
cathode. Experiments were next carried out to eliminate the polari-
sation effect by 'l)increasing the cathode area, 2) applying a
periodic reverse current and 3) changing the salt formulation. An
increase in cathode area by using graphite granules in the cathode
compartment proved to be the most satisfactory solution to the
polarisation problem and was subsequently adopted for all tests
reported here. Thus both chambers of the cell were filled up with
graphite granules for different basic reasons, namely to reduce
polarisation in the cathode and to improve both mass transfer and
mixing in the anode chamber.
It is evident from the results in TableII that refining of
Au-2OAg alloy improved with increasing applied current, namely
Au-16Ag at 'l0Aand Au-8Ag at WA. However, the current efficiency
of the present work which was of course conducted for one pass
only, remained more or less constant at 68% The main reasons for
the low efficiency besides any leakage current were Sought. An
interesting feature of the cell is that the alloy was introduced
through single and not multi-orifice feeding mechanism around the
graphite anode (Pig.1). The mechanics of current conduction from
the graphite feeder-anode to the granules where the molten alloy is
likely to be held back for electrolysis,was therefore such as
not to be able to utilise the total active surface area along the
anode length, resulting in poor current efficiency. Also the degree
of packing of granules can affect electrolysis;a denser packing
than the present work would be likely to result in better electrol-
ysis with improved efficiency. This is supported by the results on
runs 4 and 5 of TableII. Unfortunately, the present cell dimensions
and sizes of granules available did not physically allow any better
packing; thus there is definite room for improvement along this
line.
Results from the present feasibility study of a batch process
SUggest that the attraction of the new method compared to the
Miller process lies in its potential to act as a fast, pollution
free process. It is estimated that a packed-bed cell of 0.45 m
266

diameter and 0.2 m height would refine 30 tonnes of Au-Ag alloy

per year containing as high as 60"'
,O Ag at only about 6kW energy with
a current density of about 66,000 A/m2 . Any comparative study of
the present method with the Wohlwill process has not been carried
out because such investigation would be beyond the scope of the
present project and it would also involve a detailed study requiri-
ng a lot of gold alloy, for which resources are not available.

REFERENCES
1 J.H.Clelend and D.J.Fray, in M.Jones (ed.) Advances in Extracti-
ve Metallurgy, Inst.Min.Met., London, 1977, p.141.
2 D.J.Fray and J.H.Cleland, British Patent 1,515,216 (1978).
3 J.H.Cleland and D.J.Fray, Trans.Inst.Min.Met., 88C (1979) Cl911
4 W.J.Hamer, M.S.Malmberg end B.Rubin, J.Electrochem.Soc., 103
(1956) 8.
5 G.Grube and P.Hentelmann, Z.Elektrochem., 56 (1952) 1.
6 G.J.Janz, Molten Salts Handbook, Academic Press, London, 1967.
7 R.Hultgren, p.D,Desai, D.T.Hawkins, M.Gleiser, K.K.Kelley and
D.D.Wagman, Selected Values of the Thermodynamic Properties of
Binary Alloys, Am.Soc.Met.,Metals Park, Ohio, 1973 op

Appendix 1 Electrorefining principles for AU-AK alloys

Based on thermodynamics, silver will ionise in preference to the

more noble gold metal. The present cell system can be expressed as
Au~z~z;;oy IAg+l\ A@;+\ A~~m,,emz;,)

The half-cell reactions with the standard sign conventions would be

RT In aAg (alloy)
(1) Ag++ e-= Au-Ag(alloy) with El= E' +
Ag /Au-A@; - 3 aAg+
and
a
(2) Ag++ e-= Ag with E 2 = Elg+,Ag - g In $$

where z the charge carried is unity in the present case. At equili-

brium E2 - El = 0. Assume E*= E + when at equilib-
Ag /Ag - EAg+/Au-Ag
rium E* becomes equal to RT In 1 , since aAg = 1,
r
aAg(alloy)
or E* = - RT In (1)
F- aAg(alloy)
Thus electrorefining will start when the difference in electrode
potentials between the anode and cathode exceeds E*. However, in
order to pass a current it is necessary to overcome the resistance
of the electrolyte and the polarisation at both electrodes, so that
the applied voltage will always exceed E*.
Appendix 2 Estimation of minimum silver content prior to gold
dissolution during electrorefining of AU-AK alloy
The electrode potentials of both silver and gold must be equal
when both metals will simultaneously start to dissolve from the
anode, i.e.
aA + 0
'Ii;/Ag+ - F lnB=E
aArr -plnz
Ag/Ag + = %u/Au+ = %/Au+
0
No data are available in the literature on Eiu,Au+ at a high tempera-
ture such as 1170°C. Iiowever, it appears from data on fused salts
at a lower temperature [4,6] that the theoretical difference between
";g,*g+ and R*;,*u + is about IV. The above equation can therefore
be re-written as RT In (a + . a,)/(a . aAu+) = 1 which would
Ag Ag
yield (aAc+ . aAuFcaAg 0 aAu + )~3100 at T = 1443*K. It is diffic-
ult to estimate the activity of the ions of gold and silver at the
electrode surface. Since the ionic activity at the surface must be
very high, It’ ’ assumed that a +S aAu+ which would mean a W
1s
Ag Ag
aAu/300. With the progress of electrorefining, the level of metall-
ic silver decreases in the anode chamber. At a very low atomic
fraction (2) of metallic silver, the activity coefficients (f) of
gold and silver in Au-Ag alloy are given as 1 and 0.5 respectively
at a high temperature [7]. Therefore, one obtains

2Ag = aAg/ ‘rAg= aA,/(O.T X 3100) = %Au. $&550 = (I -')(11g)/1550*

Hence,%Ag is approximately equal to 6 X 10m4 which corresponds to
0.03% by weight [7]. It is at this low level of silver when gold
is expected to start to ionise i.e. dissolve along with silver
from the anode chamber during electrorefining.