Journal Archive

Johnson Matthey Technol. Rev., 2015, 59, (4), 322
doi: 10.1595/205651315X689487

Selective Removal of Mercury from Gold Bearing Streams

Exploring the use of solid adsorbents to avoid the undesirable loss of gold

    • By James G. Stevens
    • Johnson Matthey Technology Centre,
    • Blounts Court, Sonning Common, Reading, RG4 9NH, UK


Article Synopsis

Many gold ore bodies contain high levels of mercury which are co-extracted with the gold. This mercury then travels through the process circuit to pose health, environmental and technical issues. This article highlights a method to selectively remove the mercury whilst leaving the gold to be processed as normal. The removal of mercury from the circuit mitigates the need for retorting of the produced gold, reduces the potential environmental impact of any waste solutions and decreases any potential mercury exposure to plant workers.

The solid bound thiol species used have been shown by inductively coupled plasma optical emission spectrometry (ICP-OES) to reduce the mercury to undetectable levels whilst having no measurable effect on the gold concentration. The control of the cyanide concentration at the adsorption step has been shown to be key to ensuring that the mercury removal is achieved selectively. This in turn ensures that no precious metal value is lost in the mercury removal process. The process has been shown to be applicable to both batch and continuous operation which will allow the technology to be applied to a variety of flow rates and applications.



In modern gold mining processes, typically it is necessary to extract gold from complex ores which comprise gold in addition to other metals, including mercury. A common technique for extracting gold from its ores is the cyanide process, wherein leaching of gold is achieved by the addition of cyanide at alkaline pH following the Elsner equation (Equation (i)) (1). Cyanide is a strong lixiviant for gold and so leaches the gold out of the ore into solution (2). The gold is typically present in the leaching solution as a gold cyanide complex. Silver can also be extracted from its ores using a similar cyanide leaching process.

4Au + 8CN + O2 + 2H2O → 4[Au(CN)2] + 4OH (i)

A problem with this process is that cyanide is an equally strong lixiviant for many other metals, including mercury. Accordingly mercury, which is typically present in the ore along with gold or silver, is also leached into the solution. The mercury may be present in the leach solution as a variety of complex anions with the general formula [Hg(CN)2+x]x– (where x is 0, 1 or 2) depending on the ratio of mercury to cyanide ions. However, typically it is present as [Hg(CN)4]2–.

The removal of mercury from mining waters is very important, both on health and safety grounds and on environmental grounds. In particular, mercury volatilisation during extraction processes can be a threat to the health of plant workers and the presence of mercury in waste waters from mining is of significant environmental concern. Environmental legislation limits the concentration of mercury permitted in waste waters to very low levels in many countries. Accordingly, effective removal of mercury from mining waters is of significant interest to the industry. However, it is important that mercury removal technologies do not remove significant quantities of the gold or silver being mined, to avoid undesirable loss of these products during processing.

A range of different methods have been employed for mercury removal in this field. Miller et al. reviewed different technologies for the removal of mercury, including precipitation with inorganic sulfides or sulfur-based organic compounds; adsorption with activated carbon or crumb rubber; solvent extraction by alkyl phosphorus esters or thiol extractants; ion exchange with isothiouronium groups or polystyrene-supported phosphinic acid; and electrochemical cementation all with varying degrees of selectivity. They deem further work on resins with thiol functionality necessary in order to achieve the desired selectivity (3).

Dithiocarbamates form stable mercury precipitates which have been used to selectively precipitate mercury from gold bearing solutions (4), this can be carried out more efficiently by using colloidal hydroxides to cause coagulation which can then be removed by dissolved air flotation (5). Hutchison and Atwood further reviewed mercury remediation methods highlighting dimethyldithiocarbamate as an effective precipitation reagent. However the long term stability has been questioned with suggestion that mercury leaches from the precipitate over time; this combined with its degradation into toxic byproducts means that it is only applicable under certain circumstances (6). Alkyl thiols such as 1,3-benzenediamidoethanethiol can also be used to precipitate mercury (7).

This paper explores the use of solid adsorbents and how they can be applied to the selective adsorption of mercury from gold cyanide bearing process streams such as those found within the gold mining circuit.

Experimental Procedures

Test solutions were made by dissolving the appropriate salts in deionised water adjusted to the correct pH using sodium hydroxide solution to prepare a stock solution. This stock solution was diluted by pipette to provide the required concentration solution.

Batch adsorption experiments were carried out by weighing 0.5 dry wt% of the adsorbent into a 60 ml glass tube, dry weight was determined using a Sartorius infrared balance. 15 ml of the metal solution was then added and allowed to stir for the required time period on a Radleys 12 position stirrer. After stirring the solution was filtered using a 0.45 μm syringe filter, the solution was analysed by ICP-OES and compared against the initial solution. The percentage metal removed is calculated from the difference of the initial concentration to the final concentration divided by the initial concentration.

Kinetic data was obtained by running multiple tubes in parallel with each tube being filtered after an appropriate time. For the cyanide addition experiments an aliquot of cyanide solution was added to the appropriate tubes after an appropriate initial period, the tubes were then filtered as required.

Two samples of real mining process solutions were analysed by inductively coupled plasma mass spectrometry (ICP-MS). From this a model solution concentration of 4.0 ppm gold and 1.0 ppm mercury was set; this initial concentration was used unless stated otherwise.

The pH of the electrowinning (EW) pregnant and barren was pH 12.6 and 12.0 respectively; samples of heap leach were pH 10.0. The pKa of cyanide is reported as 9.2 (8); therefore, pH 9.2 is considered the minimum safe operating pH as below this the conversion to hydrogen cyanide occurs and loss of cyanide as HCN(g) causes both experimental and safety concerns. For these experiments a pH range of 10 to 13 was adopted.

Column experiments were carried out by loading the adsorbent material into a glass column, rinsing with pH adjusted deionised water and then pumping the test solution through the column using a peristaltic pump. The flow rate is controlled proportional to the bed volume (BV) with a standard flow rate of 6 BV h–1 being utilised. Outlet samples were collected and analysed by ICP-OES and compared against the inlet solution.

Materials and Reagents

Johnson Matthey produce a range of metal adsorbents for metal removal applications under the brands Smopex® (9), QuadraPure® and QuadraSil® which are based on polymer fibres, polymer beads and silica spheres respectively with an additional industrial silica based Functional Silica (FS series) range. Further details of the adsorbent materials are detailed in Table I. From internal knowledge and experience and preliminary screening several materials were identified to explore the selective adsorption of mercury. Initial screening showed both thiol and quaternary amine materials to remove mercury; however quaternary amines strongly adsorbed gold therefore this selectivity study concentrated on the thiol based materials.

Table I

Details of the Tested Materials

Name Support Material properties Functional group
FS1 Granular silica 250–710 μm, 90 Å pore size Alkyl thiol
Smopex®-111 Polymer fibre 300 × 50 μm (length × diameter) Benzyl thiol
Smopex®-112 Polymer fibre 300 × 50 μm (length × diameter) Alkyl thiol

The following reagents were used as received: mercury cyanide (Sigma-Aldrich), potassium gold cyanide (Alfa Aesar), potassium cyanide (Sigma-Aldrich), sodium cyanide (Sigma-Aldrich), sodium cyanate (Sigma-Aldrich), sodium thiocyanate (Sigma-Aldrich), sodium sulfate (Alfa Aesar), sodium thiosulfate (Fisher Chemicals), sodium hydroxide (Fisher Chemicals). Deionised water was used from an Elga Purelab DV35 at 15 M Ω.

Results and Discussion


Metal cyanide coordination complexes are well known and Nakamoto discussed how the easily identifiable CN stretching band at 2200–2000 cm–1 can provide information on the structure of the complex, as coordination of cyanide to the metal centre causes the CN band to be shifted to a higher frequency. This shift relates to both the coordination number and the metal oxidation state (10). Several mercury cyanide species are known to exist with varying coordination numbers which depends on the cyanide concentration of the solution. Gold cyanide does not vary its coordination but can exist as either gold(I) or gold(III); however, the Elsner equation (Equation (i)) predicts the gold to exist only as gold(I) in cyanidation process solutions.

Model solutions of gold and mercury cyanide were made and used to explore the likely species in a cyanidation circuit. Varying the cyanide ratio with mercury showed that the tetracoordinate [Hg(CN)4]2– is formed easily once the required ratio of cyanide is added (Figure 1). Above this ratio additional free cyanide is observed. Comparison of dissolved gold with known standards of potassium gold(I) cyanide (KAu(CN)2) and potassium gold(III) cyanide (KAu(CN)4) demonstrates that no gold(III) is present under standard conditions (Figure 2). This is in accordance with the Elsner equation. Model adsorption testing was carried out using KAu(CN)2 and mercuric potassium cyanide (K2Hg(CN)4), made in situ (Equation (ii)):

2KCN + Hg(CN)2 → K2Hg(CN)4 (ii)

Fig. 1.

Speciation in a cyanidation circuit using model solutions of gold and mercury cyanide with varying ratios HgCl2:KCN. Formation of the 4 coordinate species is observed even at 1:3 ratios indicating that its formation is favoured. At higher ratios the free cyanide peak is observed at 2079 cm–1. Mercury concentration 0.1 M

Fig. 2.

Comparison of dissolved gold with known standards. The gold powder dissolved in KCN shows formation of KOCN from the oxidation under air of KCN. Gold concentration 0.1 M

Model Adsorption

An initial comparison of the three materials (Figure 3) at pH 11 showed that Smopex®-111 gave low removal rates of mercury compared against the other materials. Both Smopex®-112 and FS1 gave excellent removal of mercury although in both cases some gold was removed. The lower performance of Smopex®-111 is probably due to the hydrophobic nature of the material. When it was tested at a higher pH or when using the sodium thiolate form of the material then the performance was improved. Following this result, further testing focused on the FS1 and the Smopex®-112 which are both alkyl thiols in a more hydrophilic environment (Figure 4).

Fig. 3.

Comparison of the three materials for removing 4 ppm gold and 1 ppm mercury cyanide at pH 11. FS1 and Smopex®-112 show good mercury removal whilst Smopex®-111 had poor wetting and showed low mercury removal

Fig. 4.

Effect of pH on removal of 4 ppm gold and 1 ppm mercury cyanides using: (a) FS1; (b) Smopex®-112. As the pH is increased the amount of gold adsorbed increases. This presumably relates to the pKa of the thiol functional group. The higher adsorption with Smopex®-112 at pH 10 has not been explained but may relate to additional hydroxyl functionality in the material

The amount of gold adsorbed by the materials generally increases with increasing pH whilst the adsorption of mercury is unaffected. This is likely due to the pKa of thiols occurring at around 12 to 13 (11); therefore, over the pH range 10 to 13 the thiols will become increasingly deprotonated changing the behaviour of their adsorption. Mercury forms a strong bond to sulfur and the adsorption is not affected by this change from free thiol to thiolate.

Various salts can be present in mining solutions depending on the source of the ore. These could potentially include sulfur containing salts such as thiosulfate or sulfate from oxidation of sulfide minerals, excess cyanide from the heap leach solution or products of cyanide decomposition including thiocyanate or cyanate. Cyanide is often used in the region of 100 to 500 ppm (12), this decreases during the process as the cyanide becomes bound to both the desired metal (gold) and undesired metals such as mercury, nickel and iron; it is also oxidised to the cyanate ion or reacts with sulfides to form thiocyanate. Infrared spectroscopy of the EW samples (Supplementary Information) shows no free cyanide but does show peaks indicative of cyanate and nickel cyanide. All of these species could potentially cause a change in the adsorption behaviour; particularly chelating species such as thiocyanide, cyanide and cyanate.

To explore the effect of these species, 100 ppm of each was added to the model solution and the adsorption was retested. Figure 5 shows that in all cases the addition of 100 ppm of the anions had little effect on the adsorption of mercury when using FS1 with 99% removal being achieved in most cases and 97% removal with cyanide addition. For gold adsorption some significant differences were observed. The addition of sulfate, thiocyanate or thiosulfate had no effect; however both cyanate and cyanide caused a reduction in the removal of gold from 87% removal with no added anion to 65% and <1% removal with cyanate and cyanide respectively. Lewis and Shaw demonstrated that gold thiolate and gold cyanide are in equilibrium (as in Equation (iii)) (13). At low cyanide concentrations the gold forms an insoluble gold thiolate with the surface bound thiol and is removed from solution; at higher cyanide concentrations the equilibrium is shifted to the soluble gold cyanide and the gold stays in solution. Mercury more easily forms an insoluble complex with the thiol and the mercury cyanide equilibrium is not strong enough to maintain the mercury in solution.

Fig. 5.

Effect on metal removal using FS1 when adding various sodium salts (100 ppm) to a solution of 4 ppm gold and 1 ppm mercury at pH 12. The commonly encountered ions show little change in the amount of gold removed from the system. Adding additional sodium cyanide meant that the gold removal was completely prevented whilst mercury removal was unaffected


Adsorption Kinetics

Simple adsorption kinetics (Figure 6) show that mercury is rapidly adsorbed with both FS1 and Smopex®-112. From the initial mercury concentration of 1 ppm both materials reduced the solution concentration to 0.01 ppm (instrument detection limit) within 5 minutes, less than 1 minute for Smopex®-112. Once adsorbed the mercury was tightly bound to the material and is not eluted under standard conditions, moderate pH or temperature.

Fig. 6.

Adsorption kinetics of 4 ppm gold and 1 ppm mercury cyanides at pH 12 using: (a) FS1; (b) Smopex®-112. Both materials give rapid removal of the mercury with only 0.04 ppm mercury detectable after 1 minute and 0.01 ppm after 5 minutes. Gold adsorption is moderate with FS1 being reduced to 0.57 ppm (85% removed) after 15 minutes. Smopex®-112 had not yet reached equilibrium with gold after 1 h indicating gold adsorption is slow with that material

Gold is adsorbed more slowly, particularly with Smopex®-112. This is not expected as Smopex® usually exhibits rapid kinetics due to its small particle size; therefore, the poor gold adsorption kinetics maybe indicative that the adsorption process is strongly equilibrium limited.

Further kinetic measurements were carried out whereby an aliquot of cyanide was added at 30 minutes, these are shown in Figure 7. This shows how upon cyanide addition the gold is rapidly desorbed from the resin. This provides further evidence for the gold thiolate ⇌ cyanide equilibrium and its effect on the gold adsorption. The addition of cyanide has little effect on the mercury which remains adsorbed onto the resin, confirming that the insoluble mercury thiolate is strongly favoured over the soluble mercury cyanide species.

Fig. 7.

Adsorption kinetics of 4 ppm gold and 1 ppm mercury cyanides at pH 12 using FS1. After 30 minutes an aliquot of sodium cyanide is added equivalent to 100 ppm. The full lines/symbols show the concentration for the standard adsorption; the dashed line/open symbols show the solution concentration with cyanide addition. This clearly shows that upon addition of cyanide the gold is rapidly desorbed from the resin whilst mercury remains adsorbed

Adsorption Isotherms

The mercury-only adsorption isotherms were measured for both FS1 and Smopex®-112 at an initial mercury concentration of 100 ppm (Figure 8); with mercury in the form of the K2Hg(CN)4 salt. The data were fitted to a Langmuir isotherm, shown in Equation (iv), where: qA = mg adsorbate per g adsorbent (mg g–1), b = adsorption parameter (l mg–1), Ce = equilibrium concentration (mg l–1) and Q0 = maximum capacity (mg g–1), which is deemed to be more suitable due to the expected chelation mechanism.

Fig. 8.

Graph showing adsorption isotherm of mercury as K2Hg(CN)4; initial mercury concentration 100 ppm, with the dashed lines showing the Langmuir isotherm fitting (anomalous data point with open symbol ignored for Langmuir fitting)


Both materials show a similar maximum capacity (Q0) from the Langmuir fitting at 28.8 mg g–1 and 31.2 mg g–1 for FS1 and Smopex®-112 respectively. However, the adsorption with FS1 is much more favourable with an adsorption parameter (b) of 5.1 l mg–1 compared to 0.50 l mg–1 for Smopex®-112.

Column Adsorption Trials

Due to the scale of liquid flow in a mining application the final process must run continuously; therefore columns are preferred. To test the feasibility column trials were run with a model solution containing 5 ppm gold and 5 ppm mercury at pH 10 with no added cyanide. Figure 9 shows that the outlet mercury concentration was below the instrument detection limit (0.01 ppm) during the entire test period; gold was initially adsorbed but then partially displaced as indicated by the outlet concentration going above the inlet concentration ([M]/[M]0>1). Figure 10 shows the effect of adding 100 ppm cyanide to the test solution. In this case no gold was adsorbed by the material whilst mercury adsorption was maintained; with the outlet mercury concentration being below the detection limit of the analysis (0.01 ppm). The low gold concentration of the first bed volume is due to the dilution effect of the column pre-rinse.

Fig. 9.

Flow test of adsorption of 5 ppm gold and 5 ppm mercury by FS1 at a flow rate of 6 BV h–1. Mercury concentration was below the instrument detection limit (0.01 ppm) during the entire test period; gold was initially adsorbed but then partially displaced (outlet concentration above inlet concentration)

Fig. 10.

Flow test of adsorption of 5 ppm gold, 5 ppm mercury and 100 ppm cyanide by FS1 at a flow rate of 6 BV h–1. Mercury concentration was below the instrument detection limit (0.01 ppm) during the entire test period; gold outlet concentration was at the inlet concentration immediately (ignoring any initial sample dilution effect)

The material loaded in Figure 9 was subjected to washing with an increasing concentration of cyanide solution at pH 10, shown in Figure 11. Initially, when washed with dilute sodium hydroxide the gold concentration rapidly decreased as the load solution was rinsed from the packed bed (BV 1 to 6); when 100 ppm cyanide was added to the wash solution (BV 7 to 12) then the gold was rapidly eluted. The mercury was retained by the solid even when the cyanide concentration was increased to 1000 ppm. This provides further evidence for the equilibrium nature of the gold-thiolate interaction which can easily be perturbed by increasing the free cyanide concentration. Meanwhile the mercury-thiolate interaction is considerably more stable.

Fig. 11.

Elution of loaded resin from Figure 9. The material was washed with deionised water at pH 10 with increasing cyanide concentration. The initial gold concentration is a dilution effect of the load solution. As 100 ppm cyanide is added to the wash water the gold is rapidly eluted. No mercury is detected (<0.01 ppm) even at a cyanide concentration of 1000 ppm

Real Stream Adsorption

Due to the complexity of the real solution streams the batch adsorption was repeated using a real sample of heap leach pregnant solution containing 1.22 ppm gold and 0.31 ppm mercury at pH 10.0; both FS1 and Smopex®-112 were tested as adsorbents. The solution was tested both as received and with cyanide added at 100 ppm in order to assess whether the addition of cyanide would improve the selectivity of adsorption in a real feed. Gold adsorption was observed from the as received test solution with both materials (Figure 12), with 7% and 20% gold removal with FS1 and Smopex®-112 respectively. In any real world application non-selective adsorption would represent loss of product and is therefore unacceptable. When the cyanide was added to the received material the gold removal was reduced to less than 1% with Smopex®-112 and to undetectable levels with FS1. Both materials reduced the mercury concentration from 0.31 ppm to below the detection limit of the instrument (<10 ppb), this was regardless of whether cyanide was added to the solution. This represents the ideal result for the desired application whereby a minor modification or monitoring of the stream allows highly selective adsorption of mercury from a gold bearing stream. The process is currently being piloted in Nevada, USA (shown in Figure 13).

Fig. 12.

Batch adsorption from real feed (heap leach pregnant solution) containing 1.22 ppm gold and 0.31 ppm mercury at pH 10.0 using: (a) FS1; (b) Smopex®-112. Both adsorbents show unacceptable levels of gold adsorption with the as received samples whilst adding 100 ppm of cyanide led to less than 1% of the gold being adsorbed with Smopex®-112 and undetectable gold adsorption with FS1

Fig. 13.

70 l pilot unit in Nevada, USA, operating at flow rates in the order of tens of litres per minute



In this study infrared spectroscopy was used to show that [Hg(CN)4]2– and [Au(CN)2] are the most likely species to occur within a mercury containing gold cyanide stream, as would be found in a gold mining circuit where cyanide is used as the lixiviant. These species were then used to conduct model adsorption tests in order to identify a method of selective mercury adsorption.

By adding free cyanide to the system (or ensuring that free cyanide exists in the solution) it was found that the selective adsorption of mercury in the presence of gold could be achieved when using thiol based adsorbents. The adsorbents were effective regardless of whether the thiol was bound to a silica or polymer based support. The theory was then applied to a real sample from a mining circuit, with the real stream it was found that when the adsorbents were applied to the solution as received then an unacceptable level of gold adsorption was obtained in conjunction to the mercury adsorption. When cyanide was added to the received solution then gold adsorption was completely prevented whilst mercury adsorption was maintained, with the adsorbent reducing the mercury concentration to below the detection limit (0.01 ppm) of the analytical equipment.

The adsorbed mercury is strongly bound to the resin as an insoluble complex without leaching under mild conditions; therefore the material can be more easily handled or stored than many mercury complexes. Alternatively, it is envisaged that the material could be regenerated with strong concentrated acids (14).

The concept was also tested using a flow system which is likely to be the final application method. Here it was also found that the addition of cyanide to the solution led to the prevention of gold adsorption whilst maintaining a mercury outlet concentration below the detection limit of the equipment (0.01 ppm) during the test period. Additionally, the solution was run with no free cyanide and a dilute cyanide solution was then used to rinse any adsorbed gold from the material. This shows that alternative process solutions can be used depending on the individual process economics with both methods being based on the same scientific concept.


  1.  L. Elsner, J. Prakt. Chem., 1846, 37, (1), 441 LINK
  2.  J. S. MacArthur, R. W. Forrest and W. Forrest, ‘Improvements in Obtaining Gold and Silver From Ores and Other Compounds’, British Patent 14,174; 1888
  3.  J. D. Miller, E. Alfaro, M. Misra and J. Lorengo, ‘Mercury Control in the Cyanidation of Gold Ores’, in “Pollution Prevention for Process Engineers”, eds. P. E. Richardson, Frank Lanzetta Jr and B. J. Scheiner, Engineering Foundation, New York, USA, 1996, pp. 151–64
  4.  M. Misra and J. A. Lorengo, Board of Regents of the University and Community College System of Nevada, ‘Method of Removing Mercury From Solution’, US Patent 5,599,515; 1997
  5.  F. Tassell, J. Rubio, M. Misra and B. C. Jena, Min. Eng., 1997, 10, (8), 803 LINK
  6.  A. R. Hutchison and D. A. Atwood, J. Chem. Cryst., 2003, 33, (8), 631 LINK
  7.  M. M. Matlock, B. S. Howerton, M. A. Van Aelstyn, F. L. Nordstrom and D. A. Atwood, Environ. Sci. Technol., 2002, 36, (7), 1636 LINK
  8.  R. M. Smith and A. E. Martell, “Critical Stability Constants: Volume 4: Inorganic Complexes”, 3rd Edn., Springer Science+Business Media, New York, USA, 1976
  9.  S. Phillips and P. Kauppinen, Platinum Metals Rev., 2010, 54, (1), 69 LINK
  10.  K. Nakamoto, “Infrared and Raman Spectra of Inorganic and Coordination Compounds”, 3rd Edn., Wiley and Sons Inc, New York, USA, 1978
  11.  M. M. Kreevoy, E. T. Harper, R. E. Duvall, H. S. Wilgus III and L. T. Ditsch, J. Am. Chem. Soc., 1960, 82, (18), 4899 LINK
  12.  “Fact Sheet - Cyanide and its Use by the Minerals Industry”, Minerals Council of Australia, Australian Capital Territory, Australia, 2005 LINK
  13.  G. Lewis and C. F. Shaw III, Inorg. Chem., 1986, 25, (1), 58 LINK
  14.  A. Arencibia, J. Aguado and J. M. Arsuaga, Appl. Surf. Sci., 2010, 256, (17), 5453 LINK

Supplementary Information

Johnson Matthey supplies advanced ion exchange materials and services for selective metal removal and recovery processes. Find out more by visiting the website:


Advanced Ion Exchange

Johnson Matthey Ion Exchange for mining industry

Johnson Matthey Ion Exchange for mining industry

The Author

James Stevens is a Senior Scientist at the Johnson Matthey Technology Centre, Sonning Common, UK. His work is focused on the development and application of solid adsorbents to the removal of metals from process water. Previously he gained his PhD at the University of Nottingham, UK, under the supervision of Martyn Poliakoff. His thesis focused on the hydrogenation of biorenewables in supercritical carbon dioxide.

Related articles

“Practical Guidelines for the Chemical Industry: Operation, Processes, and Sustainability in Modern Facilities”

Microbubble Intensification of Bioprocessing

Find an article