Hydrometallurgy 61Ž. 2001 105–119 www.elsevier.nlrlocaterhydromet

The dependence of sorbed copper and speciation on ion exchange resin type Versiane A. Leao˜ a,1, Grant C. Lukey a, Jannie S.J. van Deventer a,), Virginia S.T. Ciminelli b a Department of Chemical Engineering, The UniÕersity of Melbourne, ParkÕille, Victoria 3010, Australia b Department of Metallurgical and Materials Engineering, UniÕersidade Federal de Minas Gerais. Rua Espõrito´ Santo, 35. Belo Horizonte, MG, 30160.030, Brazil Received 17 October 2000; accepted 20 February 2001

Abstract

The present study investigates the influence of functional group structure and resin matrix on the speciation of copper and nickel sorbed onto two commercially available ion exchange resins. Batch experiments were performed using synthetic copper and nickel solutions containing 50 and 200 mgrL free cyanide, respectively. Despite the presence of 2y 3y CuŽ. CN34 and CuŽ. CN in solution, it has been found using Raman spectroscopy that the Imac HP555s resin, which has a 2y polystyrene–divinylbenzene matrix, loads predominantly the CuŽ. CN3 complex. In contrast, the polyacrylic resin, 2y 3y Amberlite IRA958, sorbed significant amounts of both CuŽ. CN34 and CuŽ. CN . It has been found that the speciation of nickel cyanide sorbed onto each resin was the same. A recently developed mathematical model based on statistical thermodynamic principles has been used as a tool to understand further the equilibrium sorption of copper and nickel cyanide complexes onto each resin studied. A higher sorption energy for nickel compared to copper has been observed for the sorption onto Imac HP555s. In contrast, the sorption energy for copper was found to be higher than for nickel for the polyacrylic resin, Amberlite IRA958. The values of the model parameters obtained were correlated with the chemical features of each complex in solution as well as sorbed onto the resins. q 2001 Elsevier Science B.V. All rights reserved.

Keywords: Copper cyanide complexes; Speciation; Raman spectroscopy; Ion exchange

1. Introduction The widespread presence of ores with low gold content as well as high levels of base metals is The processing of gold–copper ores is at present challenging the research community to identify and a major focus of attention in the minerals industry. develop new methods andror technologies for the cost-effective processing of these types of ores. The processing of gold–copper ores is currently under ) Corresponding author. Tel.: q61-3-8344-6620; fax: q61-3- investigation and some alternatives are being pre- 8344-4153. sented. Since the majority of copper minerals are E-mail address: [email protected]Ž. J.S.J. van Deventer . 1 soluble in cyanide solutions, chrysocolla and chal- Permanent address: Department of Metallurgical and Materi- wx als Engineering, Escola de Minas, Universidade Federal de Ouro copyrite being the main exceptions 1 , the main Preto, Minas Gerais, Brazil. challenge is to decrease the cyanide consumption

0304-386Xr01r$ - see front matter q 2001 Elsevier Science B.V. All rights reserved. PII: S0304-386XŽ. 01 00159-1 106 V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119 during leaching. Some alternatives have been pro- is important to identify the metal cyanide speciation posed such as the use of selective flotation, leaching because this information is used to estimate parame- with ammonia–cyanide solutions and also cyanide ter values in the mathematical model that has been recycling. developed to describe the equilibrium sorption be- In the case of cyanide recycling, the proposed haviour of each system studied. methods usually involve the acidification of cyanide As the sorption step is dependent on the physico- solutions. In the AVR process, the cyanide solution chemical features of both the solution and the resin it is acidified Ž.Acidification , then contacted with air in is necessary to assess the loading for different con- a counter-current tower to volatilise the soluble centrations of the metals in solution. This is particu- cyanide Ž.Volatilisation . Finally, the HCN–air mix- larly important in the case of the gold–copper ores ture is neutralised Ž.Reneutralisation in a caustic located on a transition zone between the sulphides at wx solutionŽŽ.. NaOH or Ca OH2 . Riveros et al. 2 the bottom of the mine and the upper part where the showed the feasibility of applying the AVR process oxide ore is found. This transition zone usually to solutions containing high amounts of copper presents a rather irregular copper contentwx 8 . In this cyanides. The standard AVR practice requires the regard, the modelling of these conditions allows the use of clarified solutions or barren bleeds. The appli- prediction of sorption equilibria when the concentra- cation of the AVR process to pulps must be done at tions of these metals in solution change significantly. higher pHŽ. usually 5.5 to 7.5 pH range . However, Many mathematical models have been proposed copper cyano-complexes start to decompose at a pH for the equilibrium sorption of ions onto soils, acti- lower than 3 and so the acidification in the above- vated charcoal and also ion exchange resins. These mentioned pH range does not allow the copper-bound models are effective in predicting the loading of the cyanide to be releasedwx 3,4 . sorbent in relatively simple systemsŽ no complex Recently developed resin technology can be used species in solution, homogeneous solid phase and for the recycling of cyanide from pulps as well as low ionic strength. . However, they usually fail in from clarified solutions. In this process, ion ex- predicting sorption equilibria in more complex sys- change resins are used to sorb and concentrate the tems such as those found in hydrometallurgical oper- cyano-complexes that have been leached from the ations—high ionic strength, different complexes in ore. The resins are then screened from the pulp and solution. eluted. Current examples of this technology are the In the present work the mathematical model origi- Augment and the Vitrokele processeswx 5,6 . nally formulated by De Kockwx 9 has been used as a Leao˜ and Ciminelliwx 7 have studied previously tool to understand further the equilibrium sorption of the sorption of metal cyano-complexes onto strong copper and nickel cyanide onto two commercially base resins from solutions containing high concentra- available ion exchange resins. This model has re- tions of copper and minor amounts of iron and cently been used to describe the equilibrium sorption nickel. The objective of the work was to identify a of gold and copper cyanide onto a variety of ion suitable resin for cyanide recycling in the processing exchange resins in non-saline and highly saline solu- of gold–copper ores. It was observed by Leao˜ and tionswx 10,11 . It is based on the principles of statisti- Ciminelliwx 7 that resins having a polyacrylic matrix cal thermodynamics as well as the Metropolis Monte possess a greater affinity for copper cyanide com- CarloŽ. MMC numerical method. An important as- pared to nickel cyanide, while for polystyrene-based pect of this modelling approach is that it does not resins the converse is observed. It was proposed that use any solely empirical parameters. It is therefore this selectivity was due to the physico-chemical fea- possible to include fundamental information of the tures of each resinŽ. matrix and exchange groups and system in order to facilitate easier estimation of also the solution, however, no direct measurement of parameters. It is important to note that the emphasis metal speciation was undertaken. In the present work, of the current work is to compare the equilibrium Raman spectroscopy is used to determine the specia- sorption of copper and nickel cyanide onto ion ex- tion of copper and nickel cyano-complexes in solu- change resins for the purpose of recycling the bound tion and also sorbed onto each ion exchange resin. It cyanide. The two resins investigated have a different V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119 107 functional group structure and also resin matrix. Amberlite IRA958Ž. 1 mL and Imac HP555s Ž 1 Therefore, the present work investigates the sorption mL. were loaded by contacting each resin with a of copper and nickel cyano-complexes on resins with solution containing 0.002 M copper and nickel different affinities for the metals. The effect of speci- cyanide. The amount of resin used was chosen in ation of each metal cyanide complex in solution as order to obtain significant sorption of each metal on well as sorbed onto the resin is discussed. The both resins. Potassium cyanideŽ. 98%, Aldrich was present study combines the results obtained from added to these solutions in order to increase the Raman spectroscopy and also the mathematical model concentration of free cyanide to approximately 50 or to develop for the first time a complete understand- 200 mgrL. Consequently, the CNrCu and CNrNi ing of the equilibrium sorption of copper and nickel molar ratio in these solutions was the same as the cyanide onto different ion exchange resins. solutions described previously. A Renishaw Raman MicroscopeŽ Ramascope 2000. was used to obtain the Raman spectra. A 2. Experimental procedures linearly polarised HeNe laserŽ Spectra Physics model a127, ls632.8 nm, 60 mW output. was used as 2.1. Ion exchange resins the excitation source. The Raman spectra of all Two commercially available ion exchange resins solutions were obtained by placing the cyanide solu- were selected for the sorption experiments. Both tions directly under the objective of the microscope. resins are macroporous and contain strong base qua- Optimum Raman spectra of each fluid were obtained = ternary ammonium functional groups. Imac HP555s by using a 5 objective to focus the laser source m is a polystyrene-based resin containing the trieth- approximately 10–20 m below the surface of the ylammonium as the exchange group. Amberlite solution. Directly focussing the laser source onto the IRA958 is a polyacrylic resin that contains the surface of the fluid was not possible as the height of trimethylammonium functionality. Both resins were the fluid decreased with time due to evaporation of obtained from Rohm and Haas, France. Prior to each the solution. The spectra of the resins were obtained experiment, the resin was allowed to swell in dis- by placing one resin bead directly under the objec- tilled water. tive of the microscope. In order to minimise the effects of the conversion of free cyanide to 2.2. Preparation and Raman analysis of metal on the spectra, large numbers of short scans were cyanide solutions and resins accumulated to yield the 10-min integration time. The spectra presented are the raw data with no Copper and nickel cyanide solutionsŽ. 0.01 M for post-accumulation processing. Raman analysis were prepared by dissolving either CuCNŽ. 99%, Aldrich in a Ž 98%, Aldrich.Ž solution CNrCus3.5 . or potassium tetra- 2.3. Determination of cyanide in solution = cyanonickelateŽ. II hydrate Ž K442 Ni Ž CN . H O, F.W. 240.99, Aldrich. in distilled water. An additional The concentration of cyanide in solution was quantity of potassium cyanideŽ. 98%, Aldrich was determined by using the silver nitrate titrimetric added to each solution in order to increase the methodwx 12 . This method is based on the formation w xy concentration of free cyanide to approximately 250 of AgŽ. CN2 as silver nitrate is added to a cyanide or 1000 mgrL. This was necessary to ensure that solution. The end point of the titration is reached these solutions had the same CNrCu and CNrNi when a silver sensitive indicator detects an excess of molar ratios as the solutions used during the sorption silver. Unless otherwise stated, rhodinineŽ Rowe Sci- experiments. It is important to note that a high entific, AR Grade. was used as an indicator. The end concentrationŽ. 0.01 M of each metal in solution was point was taken to be the final colour change from necessary in order to obtain well-defined Raman canary yellow to salmon hue. The presence of labile spectra. No attempt was made to alter the pH of copper cyanide interferes with the end point by the either the copper or nickel cyanide solution. loss of cyanide associated with the copper complex. 108 V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119

As a result, an increase in the value of ‘free’ cyanide ment. The metal content in the aqueous phase was is obtained. This effect was minimised by the dilu- determined by Inductively Coupled Plasma Optical tion of sampleswx 12 . In the present work the loss of Emission SpectroscopyŽ. ICP-OES . This enabled the cyanide during each sorption experiment was defined metal content on the resin to be determined by mass as the change in the concentration of free cyanide, balance. The pH of each solution varied between determined at the beginning and at the end of each 10.7 and 10.2 for the experiments using 200 mgrL experiment. The loss of cyanide for solutions con- free cyanide and 10.4–9.7 for the experiments using taining an initial cyanide concentration of 200 mgrL 50 mgrL free cyanide. No attempt was made to was between 15 and 60 mgrL. For solutions having maintain the pH or the free cyanide concentration in an initial cyanide concentration of 50 mgrL, the loss each container during the sorption run. of cyanide was between 10 and 40 mgrL. This indicates that the loss of cyanide is linked to the initial copper concentration in the solution. 3. Description of the mathematical model 2.4. Equilibrium sorption of metal cyanides onto ion A complete derivation of the mathematical model exchange resins used in this study has been presented by De Kock Synthetic solutions of copperŽ. I cyanide were pre- wx9 . The model has been reformulated and applied pared by mixing CuCNŽ. 99%, Aldrich and potas- recently to the equilibrium sorption of gold and sium cyanideŽ. 98%, Aldrich at a CNrCu molar copper on a series of polystyrene-based ion exchange ratio of 3.5. Nickel cyanide solutions were prepared resinswx 10,11 . In the present work the model is used by dissolving potassium tetracyanonickelateŽ. II hy- as a tool to gain further insight into the sorption = drateŽŽ. K442 Ni CN H O, F.W. 240.99, Aldrich . in behaviour of copper and nickel on two ion exchange distilled water. These solutions were placed in PVC resins which have different functionalities and matri- bottles and kept refrigerated at 48C until used. ces. The model is used to provide further support to To assess the loading of copper and nickel the observations made using Raman spectroscopy. cyano-complexes on both resins the concentration of The mathematical model used is based on the each metal in aqueous solution was varied in a way minimisation of the Gibbs free energy during sorp- that allowed the molar fraction of metals on the resin tion. Using the Boltzmann criterionwx 10,11 it is to vary between 0 and 1. Sorption tests were per- possible to formulate a system of equations that formed in baffled 1 L PVC containers that were describe sorption equilibria for systems that exhibit attached to a simultaneous overhead stirrer bench ideal properties. However, to successfully describe that had the capacity to stir 28 containers. Each the equilibrium sorption in real systems, modifica- container was stirred at 200 miny1 for 24 h. This tions are made to the model to incorporate the time was sufficient for the attainment of a pseudoe- non-ideal sorption behaviour given below. quilibrium. Each container was immersed in a water Ž.i Heterogeneous surfaces: Heterogeneous sur- bath and the temperature kept constant at 308C. faces are modelled by dividing the surface of the Before agitation, copper and nickel solutions of dif- sorbent into a series of equally sized homogeneous fering metal concentrations were prepared by mixing patches. The energy of each homogeneous patch is specific amounts of the reserved copper and nickel determined from a statistical distribution. In the pre- solutions. Using potassium cyanide solutionŽ KCN sent work, it has been found that the Normal cumula- 98%, Aldrich. , the concentration of free cyanide in tive distribution function adequately describes the each container was adjusted to either 50 or 200 heterogeneous surface of each resin studied. mgrL approximately. These systems are referred to Ž.ii Lateral interactions: The Coulombic interac- hereafter as 50 and 200 mgrL cyanide solutions. tion between sorbed species such as copper and One milliliter of resin was used throughout the ex- nickel cyanide on the resin may be significant and periments. therefore these possible interactions must be incorpo- A sample of the solution in each container was rated into the model. Intuitively, higher interaction taken at the beginning and at the end of each experi- energies would be expected for systems where the V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119 109 sorbed molecules are in close proximity. The interac- Table 1 tion energies are considered constant for each system Raman assignments for cyano-complexes in aqueous solution and if their values are positive, the species on the Metal n, polarised n, depolarised Reference y1 y1 surface repel or destabilise each other due to speciesŽ. cm Ž. cm y wx Coulombic forces. This type of interaction corre- CuŽ. CN2 2137 13 2y wx CuŽ. CN3 2108 2094 13 sponds to an increase in the total system energy. 3y wx CuŽ. CN4 2094 2078 13 Ž.iii SelectiÕe sorption: Selective sorption is asso- 2y wx NiŽ. CN4 2146 2138 14 ciated with particular features of the sorption sites on 3y wx NiŽ. CN5 2128 2114 14 the resin, for example steric effects or the nature of CNy 2077wx 13 the matrix. To incorporate selective sorption into the model, each patch is divided into separate areas according to the number of species being modelled. The equilibrium loading of each species in each area the peak at 2108 cmy1 is assigned to the symmetric 2y wx can then be determined. This allows an estimate of Ž.strong vibrational stretching of Cu Ž. CN3 13 . the loading for each patch to be made using the The Raman spectra obtained for copper cyanide relative size of each area to perform a weighted solution containing 250 and 1000 mgrL free cyanide summation. are shown in Fig. 1. It is important to note the Ž.iv IrreÕersible sorption: The possibility of the change in the relative intensities of the peaks at 2108 y1 2y y1 3y irreversible sorption of species onto the sorbent is cmŽŽ Cu CN .34. and 2078 cmŽŽ Cu CN .. as also incorporated into the mathematical model. The the concentration of free cyanide in solution is in- irreversibility parameter can have values greater than creased from 250 to 1000 mgrL. For the copper 0 and less than or equal to 1. The value of unity for cyanide solution containing 250 mgrL free cyanide, the irreversibility parameter corresponds to com- the existence of peaks at 2108 and 2078 cmy1 2y pletely reversible sorption. indicates the presence of both CuŽ. CN3 and 3y In the present study the fit of the model to the CuŽ. CN4 in solution. The greatest relative intensity experimental data was determined as follows: is observed for the peak at 2094 cmy1. As discussed previously, this peak is attributed to the asymmetric 100 m x yx i,Exp i,Pr vibrational stretch of CuŽ. CN2y and also the strong %Errors = Ý Ž.1 3 mx 3y 1 i,Pr symmetric stretch of CuŽ. CN4 . As shown by Fig. 1, the relative intensities of the above-mentioned peaks where x sthe experimental equivalent fraction i,Exp are changed as the free cyanide concentration is of species i. x sthe predicted equivalent fraction i,Pr increased from 250 to 1000 mgrL. For a free cyanide of species i. msthe number of equilibrium data concentration of 1000 mgrL, the peak at 2078 cmy1 points. 3yy ŽŽ.assigned to Cu CN4 and CN. has a greater relative intensity compared to the peak at 2108 cmy1 2y y1 ŽŽCu CN .3 . . Furthermore, the peak at 2094 cm 4. Results and discussion 2y 3y ŽŽ.assigned to Cu CN34 and CuŽ. CN. has greater relative intensity than the peak at 2108 cmy1. These Table 1 presents the assignments of Raman peaks changes in peak height imply that the equilibrium observed in solution for metal cyano-complexes. The wx15 depicted in Eq.Ž. 2 is shifted to the right as the 3y CuŽ. CN4 species exhibits peaks at 2078 and 2094 concentration of free cyanide is increased from 250 cmy1 which are assigned to the weak asymmetric to 1000 mgrL. and the strong symmetric vibrational stretching of y y 2 q yz 3 s the complex respectively. The peak at 2078 cmy1 CuŽ. CN 34CN Cu Ž. CN log K 1.5 Ž. 2 overlaps with the vibrational stretch of free cyanide, The Raman spectra obtained for the Ni–CN–H2 O which occurs at a similar wavelengthwx 13 . The peak system are presented in Fig. 2. Although the forma- y1 3y at 2094 cm also corresponds to the weak asym- tion of NiŽ. CN5 in solutions containing excess 2y wx metric vibrational stretching of CuŽ. CN3 . Finally, cyanide has been suggested previously 15 , the peaks 110 V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119

Fig. 1. Raman spectra of aqueous cuprous cyanide as a function of the free cyanide concentration.

at 2146 and 2138 cmy1 in Fig. 2 are assigned to the conditions usedŽ 250 mgrL and 1000 mgrL free 2y NiŽ. CN4 complex only. According to Terzis et al. cyanide. . wx14 the pentacyanonickelateŽ. II complex is observed The Raman spectra obtained for the Imac HP555s only in concentrated cyanide solutionsŽ. 7.8 M . resin loaded with copper and nickel cyano-com- However, even in the extreme conditions used by plexes are presented in Fig. 3. The Imac HP555s wx 2y Terzis et al. 14 the total conversion of NiŽ. CN4 to resin contains triethylammonium functional groups 3y NiŽ. CN5 was not observed. Consequently, the ab- attached to a polystyrene–divinylbenzene matrix. It 3y sence of peaks attributed to NiŽ. CN5 was expected has been established previously that this resin has a in the current work because of the experimental high affinity for cyano-complexes with low charge

Fig. 2. Raman spectra of aqueous nickel cyanide containing 250 and 1000 mgrL free cyanide, respectively. Raman spectrum of cyanide solutionŽ. as KCN showed peaks in the 2070–2090 cmy1 range. V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119 111

Fig. 3. Raman spectra for copper and nickel cyanides sorbed on the Imac HP555s resin. The peaks at 2094 and 2108 cmy1 are related to copper and the peak at 2137 cmy1 corresponds to the sorbed nickel cyano-complex. Resin loadings: 12.3 mgCurmL resin and 16.3 mgNirmL resin for 50 mgrLCNy solution; 11.4 mgCurmL resin and 15.3 mgNirmL resin for 200 mgrLCNy solution.

y wx 2y 2y such as AuŽ. CN244 16 , ZnŽ. CN and NiŽ. CN effect of metal cyanide speciation on the selectivity compared to the cyano-complexes of copper and iron of ion exchange resins. wx7,17,18 . This sorption phenomenon was attributed The Raman spectra presented in Fig. 3 clearly to the steric hindrance of the triethylammonium elucidate the speciation of copper and nickel cyano- functional groupwx 7,18 . complexes sorbed on the Imac HP555s resin. As A previous investigation by Leao˜ et al.wx 7,17 shown previously in Fig. 1, a copper cyanide solu- used both batch and column experiments to study the tion containing a concentration of either 250 or 1000 sorption behaviour of Imac HP555s. It was observed mgrL cyanideŽ this corresponds to the same CNrCu by Leao˜ et al.wx 7,17 that for a solution containing a molar ratio used during the sorption of the metals on high concentration of copper and minor amounts of the resin. exhibits Raman peaks that are characteris- 2y 3y nickel and iron cyano-complexes, the Imac HP555s tic of both CuŽ. CN34 and CuŽ. CN complexes. resin exhibited a high affinity for the nickel cyano- Furthermore, the distribution of the copper cyanide complex while iron was rejected by the resin. Copper complexes in solution was dependent upon the con- presented an intermediate behaviour. In particular for centration of free cyanide. Fig. 2 also shows that 2y the batch experiments, which involved contacting 1 g nickel forms only the NiŽ. CN4 complex in solu- of resin with 100 mL of a solution containing 500 tions containing a concentration of 250 or 1000 mgrL Cu, 50 mgrL Ni and 15 mgrL Fe, Leao˜ and mgrL free cyanide, respectively. Ciminelliwx 7 observed the loading of nickel and iron Comparing Fig. 3 with Fig. 1 it can be seen that to be 99% and 6%, respectively. Copper exhibited an the weak asymmetric vibrational stretch at 2078 y1 3y intermediate level of sorption with 63% being ex- cm , assigned previously to the CuŽ. CN4 , com- tracted from the solution. The behaviour of copper plex is not observed in the Raman spectra of the during sorption was credited to the formation of both loaded resin. This result suggests that the Imac 2y 3y 2y CuŽ. CN34 and CuŽ. CN in solution. Conse- HP555s resin has a strong affinity for the CuŽ. CN 3 quently, for the experimental conditions used in the complex. The peaks at 2108 and 2094 cmy1 in Fig. study by Leao˜ and Ciminelliwx 7 copper was not 3 are assigned to the strong and weak stretching 2y completely rejected nor did the complex have a modes of vibration of the CuŽ. CN3 complex, re- 3y strong affinity for the resin. In the present work, spectively. As the polarised peak of CuŽ. CN4 oc- Raman spectroscopy is used to investigate further the curs at the same Raman shift of the depolarised 112 V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119

2yy1 Ž.weak peak of Cu Ž. CN3 , the peak at 2094 cm occurs without change in its coordination. The only alone cannot be used to identify the tetra- nickel species observed in solution and also on the 2y cyanocuprite complex. Fig. 3 shows clearly that the resin is the NiŽ. CN4 complex. It is noteworthy that Imac HP555s resin has a strong affinity for the the slight shift in the position of the peak assigned to 2y 2y CuŽ. CN3 complex, despite a distribution of both NiŽ. CN4 in Fig. 3 compared with Fig. 1 can be 2y 3y CuŽ. CN34 and CuŽ. CN in solutionŽ. Fig. 1 . The attributed to the interaction of the nickel complex polystyrene–divinylbenzene matrix of Imac HP555s with the resin. Jones and Pennemanwx 19 observed leads to the resin having a predominantly hydropho- that the loading of the aurocyanide species onto bic character. This hydrophobic character, combined resins occurs without any change of the gold–cyanide with the steric hindrance associated with the trieth- speciation. In aqueous solutions, gold forms only the y ylammonium functional group explains the strong AuŽ. CN2 complex regardless of the amount of 2y affinity of the resin for CuŽ. CN3 . cyanide present in the solution or in the resin. Fur- Jones and Pennemanwx 19 have observed this sorp- thermore, Vernon et al.wx 20 and Fawell et al. wx 21 tion phenomenon previously for the observed no speciation changes during the loading of system. Using infrared spectroscopy, Jones and Pen- copper, nickel and gold cyano-complexes on the nemanwx 19 observed the formation of three different IRA400 resin. Riveros and Cooperwx 22 also ob- silver cyanide complexes in solution as the concen- served no change in the speciation of iron during y tration of free cyanide was increased, viz. AgŽ. CN2 , sorption. In summary, the above-mentioned work has 2y 3y AgŽ. CN34 and AgŽ. CN . However, regardless of shown that the nature of the sorbed species is depen- the concentration of free cyanide in solution, it was dent upon the chemical structure of the resin as well y observed using infrared that only AgŽ. CN2 loaded as the chemistry of the sorption solution. onto the Dowex A-1 resin. Other peaks were ob- As shown by Figs. 4 and 5, the equilibrium served in the infrared spectrum, however, these were sorption of copper and nickel on the Imac HP555s y2 y3 unable to be assigned to AgŽ. CN34 or AgŽ. CN . resin is predicted satisfactorily using the mathemati- Although the peak associated with the formation cal model described. As indicated by Table 2, a y of CuŽ. CN2 in solution is not observed in Fig. 1, it unimodal Normal distribution was used to describe is possible that this complex may be sorbed onto the heterogeneity of the resin surface experienced by Imac HP555s due to the high affinity that this resin copper and also nickel. A recent study by Lukey et has for cyano-complexes with low charge. From al.wx 10 modelled the loading of copper and gold on a Table 1 it can be seen that the Raman peak for the series of polystyrene-based resins using a bimodal yy1 CuŽ. CN2 complex is located at 2137 cm . This Normal distribution for copper and unimodal Normal y means that the possible sorption of CuŽ. CN2 on the distribution for gold. It was reasoned by Lukey et al. Imac HP555s resin cannot be determined from Fig. 3 wx10 that a bimodal distribution for copper was re- y because the CuŽ. CN2 peak overlaps with the peak quired due to the significant sorption of both 2y 2y 3y assigned to the vibrational stretch of NiŽ. CN4 Ž Ta- CuŽ. CN34 and CuŽ. CN on the resins studied. As ble 1. . In an attempt to determine whether the stated previously, the Imac HP555s resin used in the y CuŽ. CN2 complex does load onto the Imac HP555s current work is highly selective for complexes with a resin, a separate sorption experiment was performed low charge. In particular, it was established using which loaded only copper cyanide species onto the Raman spectroscopy that the resin sorbs predomi- 2y resin. Although the Raman spectrum is not shown, nantly the CuŽ. CN3 complexŽ. Fig. 3 despite the y 2y 3y the characteristic peak for CuŽ. CN2 was not ob- fact that both CuŽ. CN34 and CuŽ. CN exist in served on the Imac HP555s resin loaded only with solution for the experimental conditions used in this copper cyanide. Therefore, provided that nickel workŽ. Fig. 1 . It was noted that due to the overlap- cyanide does not affect the speciation of copper ping Raman peaks of the weak asymmetric stretch of y 2y sorbed onto the resin, it can be inferred that CuŽ. CN 2 CuŽ. CN3 with the strong symmetric vibrational 3y does not load onto the Imac HP555s resin. stretch of CuŽ. CN4 it was not possible to conclude 2y The Raman peaks assigned to nickel cyanide indi- beyond doubt that only the CuŽ. CN3 sorbed onto cate that the loading of this metal on Imac HP555s the resin. However, the fact that a unimodal Normal V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119 113

Fig. 4. A comparison between predicted and experimental copper and nickel loading on the Imac HP555s resin. Initial cyanide concentration: 50 mgrL. distribution more accurately describes the sorption of sites are available for both complexes. In this regard copper on Imac HP555s supports the contention that the most important factor is the electrical charge of 2y 2y only the CuŽ. CN3 sorbs onto the resin. the complex. Due to the double charges of CuŽ. CN 3 2y As an approximation, the equivalent fraction of and NiŽ. CN4 at least two functional groups must copper in solution was calculated using a charge of be in close enough proximity for sorption to take 3y 2.5. This was necessary because it is not possible to place. This is a further reason why CuŽ. CN4 does determine the ratio of both copper cyanide species in not load onto the Imac HP555s resin, because this solution. Although the nickel cyano-complex is would require at least three triethyl groups to be in 2y square planar and the CuŽ. CN3 complex is trigonal close proximity for sorption to occur. This is made planar, no improvement in the fit of the model was difficult due to the steric effects associated with the 3y observed by limiting the surface area available for triethyl functional group. In addition, the CuŽ. CN 4 the sorption of copper or nickelŽ. Table 2 . This complex is labile and so readily converts to the more implies that although the copper and nickel com- stable tricyanocuprite complexŽŽ.. Eq. 2 . plexes have different geometries, no selective sorp- It can also be seen from Table 2 that the mean tion of either complex takes place on the Imac sorption energies of copper and nickel are only HP555s resin. This means that the same sorption slightly different and these values are not signifi-

Fig. 5. A comparison between predicted and experimental copper and nickel loading on the Imac HP555s resin. Initial cyanide concentration: 200 mgrL. 114 V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119

Table 2 affinity of the resin for complexes of a low charge. Parameters used for the modelling of copper and nickel sorption This corresponded to a decrease in the mean sorption on Imac HP555s resin containing triethylammonium functional groups energy for copper compared with the other resins studied. Lukey et al.wx 10 also observed a better Surface Solvent 50 mgrL 200 mgrL model fit when the area available for the sorption of Nickel Copper Nickel Copper copper was decreased to 92%. This was reasoned to Energy – Normal Normal Normal Normal be due to steric hindrance caused by the long alkyl distribution chain of the tripropylammonium functional group. Mean – 1 1 1 1 Standard – 0.05 0.1 0.03 0.1 As discussed previously, a restriction in the area deviation available for the sorption of copper did not improve Selective 100 100 100 100 the fit of the model in the present work. However, it sorptionŽ. % is important to note that the Imac HP555s resin Species contains triethylammonium groups. Therefore, due to Mean sorption 0 y17.97 y15.83 y18.24 y15.99 the shorter alkyl chain length and the higher strong energy base capacity of the Imac HP555s resin compared Interaction 0 0 0 0 0 with the tripropylammonium resin, the selective energies sorption of copper is not expected. In addition, in the current work the copper and nickel cyanide system is investigated. As shown by Fig. 3, the predominant cantly changed whether the concentration of free complexes sorbed onto the Imac HP555s resin are 2y 2y cyanide used during the sorption experiments was 50 CuŽ. CN34 and NiŽ. CN . Both of these complexes or 200 mgrL. As expected, a less negative value for are planar in geometry and have the same charge. the mean sorption energy for copper is observed. Consequently, it would be expected that the resin This is due to the physico-chemical features of the surface available for sorption of both complexes sorbent i.e. the highly hydrophobic character due to would be the same. the polystyrene–divinylbenzene matrix and also steric As mentioned previously, the Amberlite IRA958 effects attributed to the triethylammonium functional resin has trimethylammonium functional groups at- groups. The substantially negative values of the sorp- tached to a polyacrylic matrix. The presence of C5O tion energies also represent the high affinity of the groups in the matrix allows the formation of hydro- resin for both species relative to the solvent. This is gen bridges with water molecules. This leads to a to be expected due to the highly polarised metal-to- higher hydration of the resin, represented by its cyanide bonds associated with metal cyanide com- water contentwx 7,18 . This results in polyacrylic-based plexes. This also means that a high proportion of the resins having a more hydrophilic character compared surface area available for sorption is covered by both with polystyrene resins. Consequently, the Amberlite copper and nickel complexes. IRA958 resin has a high affinity for highly charged 4y 3y Upon modelling the equilibrium sorption of gold complexes such as FeŽ. CN64 and CuŽ. CN com- and copper onto an experimental resin containing pared with polystyrene-based resins. Due to the high tripropylammonium functional groups, Lukey et al. charges of these complexes, they have a higher wx10 observed that the resin had a higher affinity for hydration level than cyano-complexes with low va- the tricyanocuprite complex than for the tetracyan- lence. As a result, they exhibit a high affinity for cuprite species. This phenomenon was attributed to hydrophilic resins such as Amberlite IRA958. the presence of the long alkyl chain of the tripropy- The Raman spectra obtained for the Amberlite lammonium functional group. This also resulted in IRA958 resin loaded with copper and nickel are the resin having a lower strong base capacity com- presented in Fig. 6. The presence of three substantial pared with a resin containing trimethylammonium peaks in the 2070–2110 cmy1 region indicates that 2y 3y groups. It was proposed in the study by Lukey et al. CuŽ. CN34 and CuŽ. CN sorb onto the Amberlite wx10 that the tripropylammonium groups increase the IRA958 resin. It should be noted that there is a shift 3y hydrophobicity of the resin and thereby enhance the in the position of the peaks for the sorbed CuŽ. CN 4 V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119 115

Fig. 6. Raman spectra of copper and nickel cyanides loading on the Amberlite IRA958 resin. The peaks at 2071, 2090, and 2108 cmy1 are related to copper and the peak at 2146 cmy1 corresponds to nickel cyano-complex. Resin loadings: 7.3 mgCurmL resin and 10.0 mgNirmL resin for 50 mgrLCNy solution; 7.5 mgCurmL resin and 9.3 mgNirmL resin for 200 mgrLCNy solution.

3y complex. In aqueous solution the CuŽ. CN4 com- sorption solution had a free cyanide concentration of y1 r 2y plex exhibits peaks at 2078 and 2094 cmŽ. Fig. 1 . approximately 50 mg L. The loading of CuŽ. CN 3 The peaks for the sorbed complex are located at is prominent in view of the intense peak at 2108 2071 and 2091 cmy1 Ž. Fig. 6 . The shift in the cmy1. The overlapping of shifted peaks at 2091 position of these peaks is attributed to the interaction cmy1 makes it difficult to quantify the loading of 3y 3yy1 between the labile CuŽ. CN4 complex and the poly- CuŽ. CN4 , but the shifted peak at 2071 cm indi- 3y acrylic resin matrix. The slight shift in these peaks to cates that some CuŽ. CN4 is indeed present. At a lower Raman values was also observed when Am- free cyanide concentration of 200 mgrL the shifted y1 3y berlite IRA958 was loaded with copper cyanide only. peak at 2071 cmŽŽ Cu CN .4 . is more dominant Consequently, this phenomenon is not due to the than at a lower cyanide concentration, and the shifted sorption of nickel cyanide on the resinŽ. Fig. 6 . It is overlapping peak at 2091 cmy1 becomes more sig- noteworthy that Coleman et al.wx 23 also observed a nificant relative to the peak at 2108 cmy1 Ž 2108 y1 3y shift in the position of the absorption peaks for cm.Ž. . This suggests that Cu CN4 is present in 2y copper cyano-complexes sorbed on the quaternary relatively higher concentrations than CuŽ. CN3 at ammonium resin, Dowex 1-X10. these higher cyanide concentrations. It is important 3y It was discussed previously that the CuŽ. CN 4 to note that the ion exchange resins studied do not complex is the predominant species in solution when possess a high affinity for free cyanide. These results the free cyanide concentration is 1000 mgrLŽ Fig. are expected because it is shown by Fig. 1 that for 1. . This is because the relative intensity of the peak the experiments at 1000 mgrLCNy the concentra- y1 3y 2y observed at 2078 cmŽŽ. characteristic of Cu CN 4 . tion of CuŽ. CN3 in solution is lower than at 250 in Fig. 1 is greater than the relative intensity of the mgrL—the relative intensity of the peak at 2108 y1 2y y1 2y peak at 2108 cmŽŽ. characteristic of Cu CN3 . . cmŽŽŽ. characteristic of Cu CN3 . is lower. Con- The relative distribution of copper cyanide species versely, at 1000 mgrL cyanide the concentration of 3y sorbed on Amberlite IRA958 can be obtained by CuŽ. CN4 species in solution is higher than at the comparing the relative intensities of Raman peaks lower cyanide concentration. assigned to each copper complexŽ. Fig. 6 . It is The parameters used to model the equilibrium shown by Fig. 6 that the Amberlite IRA958 resin sorption of copper and nickel cyano-complexes on 2y 3y loaded both CuŽ. CN34 and CuŽ. CN when the the polyacrylic resin, Amberlite IRA958, are pre- 116 V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119

Table 3 Parameters used to model the equilibrium sorption of copper and nickel onto the Amberlite IRA958 resin, which contains trimethyl- ammonium functional groups Surface Solvent 50 mgrL 200 mgrL Nickel Copper Nickel Copper Energy distribution – Normal Normal Normal Normal Normal Normal Distribution fraction – – 0.42 0.58 – 0.35 0.65 Mean – 1 0.8 1 1 0.8 1 Standard deviation – 0.05 0.05 0.2 0.05 0.05 0.2 Selective sorptionŽ. % 100 100 100 100

Species Mean sorption energy 0 y19.40 y21.28 y19.87 y22.31 Interaction energy 0 0 0 0 0 sented in Table 3. As shown by Figs. 7 and 8, the Normal distribution has been used to describe the sorption of both metals on the resin has been mod- surface heterogeneity experienced by the copper elled successfully. It can be seen from Table 3 that cyanide complexes. As shown by Table 3, a distribu- the mean sorption energy for copper is slightly more tion fraction of 0.42 and 0.35Ž for 50 and 200 mgrL negative than that for nickel. In addition, the values cyanide, respectively. is used to specify the bimodal of these parameters did not change significantly with Normal distribution. The latter value corresponds to 3y an increase in the free cyanide concentration from 50 a higher proportion of CuŽ. CN4 on the resinŽ Fig. to 200 mgrL. 6. . According to Lukey et al.wx 10,11 the mean 2y As indicated by Figs. 1 and 6, the NiŽ. CN 4 sorption energy parameter is related to the electrical complex is the only nickel species present in solution interaction between the ion exchange resin and the and also sorbed onto the resin. For this reason, a sorbed metal cyanide complex. Therefore, from clas- unimodal Normal distribution is used to describe the sical ion exchange theory this means that a more surface heterogeneity of the resin experienced by the highly charged species would have a greater affinity nickel cyanide complex. It can also be seen from for the resin. In the model, this increased affinity for Fig. 6 that both copper complexes sorb onto the the resin corresponds to a more negative mean sorp- Amberlite IRA958 resin. Consequently, a bimodal tion energy.

Fig. 7. A comparison between predicted and experimental copper and nickel loading on the Amberlite IRA958 resin. Initial cyanide concentration: 50 mgrL. V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119 117

Fig. 8. A comparison between predicted and experimental copper and nickel loading on the Amberlite IRA958 resin. Initial cyanide concentration: 200 mgrL.

From Table 3 it can be seen that the mean sorp- stronger sorption. than for Imac HP555s. This is tion energy of copper is slightly more negative than because Amberlite IRA958 has a greater hydrophilic- that of nickel. This is attributed to the presence of ity due to its polyacrylic matrix compared with the 2y 3y both CuŽ. CN34 and CuŽ. CN on the resin. As the polystyrene-based Imac HP555s. The difference in 3y CuŽ. CN4 complex has a more negative charge than the resin functional groups has also been shown to 2y the NiŽ. CN4 complex, a more negative sorption affect the mean sorption energies. energy is expected. Limiting the surface area avail- The results of the present work show that both able for the sorption of either copper or nickel did resins can be used for the sorption of the cyano-com- not improve the fit of the model to the sorption data. plex of nickel, while the Imac HP555s resin is more However, selective sorption of either copper or nickel suitable for the loading of copper in solutions where 2y is not expected to take place in the Amberlite IRA958 CuŽ. CN3 is the predominant species. However, the system because this resin is hydrophilic and contains successful elution of metals from each of these resins trimethylammonium groups. As a consequence, the is an important consideration. Most of the proposed affinity of the resin for copper or nickel cyanide elution procedures are based on the acidification of complexes is not affected by their geometry or extent the resin in an AVR-like procedure. While nickel of co-ordination. cyanide can be directly decomposed to Ni2q and Comparing the mean sorption energies of copper free cyanide, the acidification of resins that contain a and nickel for the two resins studiedŽ Tables 2 and high concentration of copper cannot be carried out 3. , it can be seen that these values are more negative unless sufficient acid or a complexing agent is ap- for the Amberlite IRA958 resin than the Imac HP555s plied. This is necessary because in the case of direct resin. Although the concentration of exchange groups acidification, copper decomposes to the very stable Ž.the strong base capacity is higher for the Amberlite CuCN precipitate. The decomposition reaction is IRA958 resin compared with Imac HP555s, this accompanied by the liberation of HCN: alone does not account for the differences in sorption 1yn q CuŽ.Ž. CN q ny1HzCuCNq Ž.ny1 HCN energy. Of more importance is the difference in resin n 3 matrix and also functional group on each of the two Ž. resins. As expected, the model indicates that the The CuCN precipitate could lead to the poisoning energy associated with the sorption of copper and of the ion exchange resin. Thiourea is capable of nickel onto Amberlite IRA958 is more negativeŽ i.e. replacing the cyanide ligand in the complex, thereby 118 V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119 producing a cationic species that does not load onto sorbed onto two different ion exchange resins. It has the resin. It has been shown previously that an been established that the Imac HP555s resin loads 2y 2y acidifiedŽ. H24 SO thiourea solution is able to elute NiŽ. CN43 and CuŽ. CN . However, the sorption of 3y more than 90% of copper and nickel loaded onto CuŽ. CN4 on the Imac HP555s resin was not ob- Imac HP555s as well as a resin with the same served. It is proposed that the predominant sorption wx 2y polyacrylic matrix as Amberlite IRA958 17 . of CuŽ. CN3 onto the Imac HP555s resin is due to As discussed above, the sorption of cyano-com- the sterically bulky triethylammonium group and plexes onto ion exchange resins relies on the chemi- also the hydrophobic polystyrene–divinylbenzene cal features of both the complex and the resin. If the matrix. Conversely, the Amberlite IRA958 resin 2y 2y metal can form more than one complex the resin loads NiŽ. CN43 as well as both CuŽ. CN and 3y may have a higher affinity for one of these com- CuŽ. CN4 complexes. It is proposed that the sorp- plexes. The present study has shown that the factors tion of the cuprous tetracyano complex is due to the which affect the affinity of the resin for metal cyanide hydrophilic nature of the polyacrylic matrix and also complexes include: the charge of complex, hy- the greater concentration of exchange groups on drophobicrhydrophilic character of the resin, type of Amberlite IRA958 compared to the Imac HP555s matrix, and also steric effects such as the alkyl chain resin. length of the functional group. The equilibrium sorption behaviour of copper and The affinity of the Dowex A1 resin for the nickel cyano-complexes on the Imac HP555s resin y wx AgŽ. CN2 complex in the Ag–CN system 19 , the and the Amberlite IRA958 resin has been success- precipitation of AuCN in a polydiallylaminewx 20,21 fully described using a mathematical model based on and the results of the present work show a change in statistical thermodynamic principles. The advantage the nature of the species upon sorption onto a resin. of this modelling approach over models proposed On the other hand, it was also shown that there is no previously is that it uses parameters that have a change in the speciation of copper or nickel during fundamental meaning. In particular, the Raman spec- sorption onto the polydiallylamine resinwx 20,21 . It troscopic results regarding metal cyanide speciation was also shown in the present study that the specia- have been used to provide estimates of parameter tion of nickel was the same in solution and on the values in the model. It has been found for the Imac sorbent. These results indicate that the mechanisms HP555s resinŽ. polystyrene–divinylbenzene matrix of sorption are specific for each system under con- that nickel has a slightly more negative sorption sideration. The generalisation of sorption studies to energy than copper. In contrast, it has been found for broader environments should be carried out consider- Amberlite IRA 958Ž. polyacrylic matrix that copper ing the differences in the nature of the solutions and has a slightly more negative sorption energy than the resins used during these studies. The sorption nickel. behaviour of copper and nickel on the Imac HP555s It was found for both resins that the concentration and IRA 958 resins is dependent upon the hydropho- of free cyanideŽ. either 50 or 200 mgrL during bic nature of the ions and the resins themselves. The sorption did not affect the model parameters signifi- type of functional group on the resin is also an cantly. The results of the present study show that the important consideration, because it is has been shown Amberlite IRA958 resin can be used for the sorption in the current work that the triethylammonium func- of copper and nickel cyanide complexes, while the tional group increases the steric hindrance in the Imac HP555s resin is not recommended for the system. recycling of cyanide from solutions containing a high CNrCu ratio. For the first time, the equilibrium sorption behaviour of copper and nickel on both resins studied has been successfully described. By 5. Conclusions combining Raman spectroscopy and a probabilistic model a greater understanding of the equilibrium Raman spectroscopy has been used to determine sorption of both metals onto ion exchange resins has the metal cyanide speciation in solution and also been achieved. V.A. Leao˜ et al.rHydrometallurgy 61() 2001 105Ð119 119

Acknowledgements wx11 G.C. Lukey, J.S.J. Van Deventer, D.C. Shallcross, Equilib- rium model for the sorption of gold cyanide and copper cyanide on trimethylamine resin in saline solutions. Hy- This work was supported by Conselho Nacional drometallurgy 59Ž.Ž 1 2000 . 101–113. de Pesquisas, CNPq, Brazil and by RECOPE and wx12 W.C. Lenahan, R.D.L. Murray-Smith, Assay and Analytical PRONEX grants. The assistance of Dr. Shane T. Practice in South African Mining Industry, Johannesburg. Huntington from the Particulate Fluids Processing SAIMM, Johannesburg, South Africa, 1986. wx CentreŽ. PFPC , The University of Melbourne is also 13 G.C. Lukey, J.S.J. Van Deventer, S.T. Huntington, R.L. Chowdhury, D.C. Shallcross, Raman study on the speciation appreciated. of copper cyanide complexes in highly saline solutions. Hydrometallurgy 53Ž.Ž 9 1999 . 233–244. wx14 A. Terzis, K.N. Raymond, T.G. Spiro, On the structure of 3y NiŽ. CN5 . Raman Infrared and X-Ray Crystallographic Evi- References dence. Inorganic Chemistry 9Ž.Ž 11 1970 . 2415–2420. wx15 A.G. Sharpe, The Chemistry of Cyano Complexes of the wx1 N. Headley, H. Tabachnick, Chemistry of cyanidation. Min- Transition Metals. Academic Press, London, 1976 302 pp. eral Dressing Notes, vol. 17, American Cyanamid, 1950, 54 wx16 R.S. Costa, V.A. Leao,˜ V.S.T. Ciminelli, Copper and cyanide pp. recovery using ion exchange resins. ALTA Copper 1999— wx2 P.A. Riveros, D. Koren, V.M. McNamara, J. Binvignat, Copper Sulphides Symposium and Copper Hydrometallurgy Cyanide recovery from a gold mill barren solution containing Forum September 6–8, Gold Coast, Queensland, Australia. high levels of copper. CIM Bull. 91Ž.Ž. 1025 1998 73–81. ALTA Metallurgical Services Ltd., Victoria, Australia, 1999. wx3 P.A. Riveros, R. Molmar, F. Basa, Treatment of a high- wx17 V.A. Leao,˜ R.S. Costa, V.S.T. Ciminelli, Cyanide recycling cyanide waste solution for cyanide and metal recovery. CIM using ion exchange resins–application to the treatment of Bull. 89Ž.Ž. 998 1996 153–156. gold–copper ores. in: MassaciŽ. Ed. , XXI International Min- wx4 V.A. Leao,˜ R.S. Costa, V.S.T. Ciminelli, Cyanide recycling eral Processing Congress, Rome, Italy, vol. A, 2000, pp. using strong base ion exchange resins. J. Met. 50Ž.Ž 10 1998 . A6:1–A6:9. 71–74. wx18 P.A. Riveros, Selectivity aspects of the extraction of gold wx5 L. Whittle, in: H. von MichaelisŽ. Ed. , The Piloting of from cyanide solutions with ion exchange resins. Hydromet- Vitrokele for the Cyanide Recovery and Waste Management allurgy 33Ž. 1993 43–58. at Two Canadian Gold Mines, Randol Gold Forum ’92, wx19 L.H. Jones, R.A. Penneman, Infrared absorption studies of Vancouver, Canada. Randol International Ltd., Golden, CO, aqueous complex ions: I. Cyanide complexes of AgŽ. I and USA, 1992, pp. 379–384. AuŽ. I in aqueous solution and adsorbed on anion resin. J. wx6 C.A. Fleming, in: B. MishraŽ. Ed. , The potential role of Chem. Phys. 22Ž.Ž 6 1954 . 965–970. anion exchange resins in the gold industry. EPD Congress wx20 C.F. Vernon, P.D. Fawell, C. Klauber, XPS investigation of 1998, TMS, San Antonio, TX, USA, 1998, pp. 95–117. the states of sorption of aurocyanide onto crosslinked poly- wx7 V.A. Leao,˜ V.S.T. Ciminelli, Application of ion exchange dallylamine and commercial anion exchange resins. React. resins in gold hydrometallurgy. A tool for cyanide recycling. Polym. 18Ž.Ž 1 1992 . 35–45. Solvent Extr. Ion Exch. 18Ž.Ž 3 2000 . 567–582. wx21 P.D. Fawell, C.F. Vernon, C. Klauber, H.G. Linge, The wx8 S.R. La Brooy, H.G. Linge, G.S. Walker, Review of gold extraction of metal cyanides by crosslinked polydiallylamine. extraction from ores. Miner. Eng. 7Ž.Ž 10 1994 . 1213–1241. React. Polym. 18Ž.Ž 1 1992 . 57–65. wx9 F.P. De Kock, A Generalised Competitive Adsorption The- wx22 P.A. Riveros, W.C. Cooper, Speciation study on the ion ory of Non-Ideal Substances, PhD Thesis, The University of exchange extraction of silver, copper, iron and zinc cyano Stellenbosch, South Africa, 1995, 218 pp. complexes with an iterative slope analysis technique. Solvent wx10 G.C. Lukey, J.S.J. Van Deventer, D.C. Shallcross, Equilib- Extract. Ion Exch. 3Ž.Ž 6 1993 . 909–930. rium model for the selective sorption of gold cyanide on wx23 J.S. Coleman, R. George, L. Allaman, L.H. Jones, Stepwise different ion-exchange functional groups. Miner. Eng. 13 formation of cyanide complexes of CopperŽ. I in anion ex- Ž.Ž12 2000 . 1243–1261. changers. J. Phys. Chem. 72Ž.Ž 7 1968 . 135–139.