Separation and Purification Technology 136 (2014) 94–104

Contents lists available at ScienceDirect

Separation and Purification Technology

journal homepage: www.elsevier .com/locate /seppur

Poly-thiosemicarbazide membrane for gold recovery

http://dx.doi.org/10.1016/j.seppur.2014.08.0271383-5866/� 2014 Elsevier B.V. All rights reserved.

⇑ Corresponding author. Tel.: +966 2 8082769.E-mail address: klausviktor.peinemann@kaust.edu.sa (K.-V. Peinemann).

Luis Francisco Villalobos a, Tahir Yapici b, Klaus-Viktor Peinemann a,⇑a Advanced Membrane and Porous Materials Center, King Abdullah University of Science and Technology (KAUST), 23955-6900 Thuwal, Saudi Arabiab Analytical Core Laboratory, King Abdullah University of Science and Technology (KAUST), 23955-6900 Thuwal, Saudi Arabia

a r t i c l e i n f o

Article history:Received 25 March 2014Received in revised form 14 August 2014Accepted 14 August 2014Available online 6 September 2014

Keywords:Gold recoveryMembrane adsorberPoly-thiosemicarbazidePolymeric membraneMetal-ion capture

a b s t r a c t

A novel polymeric membrane adsorber with a high density of adsorption sites that can selectively captureAu(III) ions, is proposed as an efficient alternative to recover gold from dilute solutions. Poly-thiosemicar-bazide (PTSC), a polymer that contains one chelate site per monomeric unit, was used to fabricate themembranes. This polymer can be easily processed into membranes by a phase inversion technique,resulting in an open and interconnected porous structure suitable for high flux liquid phase applications.This method overcomes the usual low capacities of membrane adsorbents by selecting a starting materialthat contains the adsorption sites within it, therefore avoiding the necessity to add an external agent intothe membrane matrix.

The resulting mechanically stable PTSC membranes can operate in a pressure driven permeation pro-cess, which eliminates the diffusion limitations commonly present in packed column adsorption pro-cesses. This process can selectively recover 97% of the gold present in a solution containing a 9-foldhigher copper concentration, while operating at a flux as high as 1868 L/m2 h. The maximum gold uptakemeasured without sacrificing the mechanical stability of the membrane was 5.4 mmol Au/g. Furthermorethe gold can be easily eluted from the membrane with a 0.1 M thiourea solution and the membrane canbe reused for at least three cycles without any decrease in its performance. Finally, the ability of thismembrane for recovering metals from real-life samples, like seawater and tap water, was tested withpromising results.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction

Growing demand for gold and high market prices motivate thedevelopment of technologies to recover the precious metal fromsecondary sources, like scrap (electronic, jewelry, dental), industrywaste (spent catalyst, alloy scraps, copper anode slime, mine tail-ing), and seawater [1]. To put into perspective how realistic it isto consider gold recovery and recycling as a future source of gold,Ogata et al. reported that cellular phone scrap contain around200 g of gold per ton of scrap, while the gold present in the goldores used in mining contains between 5 and 30 g of gold per tonof ore [2].

A hydrometallurgical process, the preferred method for recover-ing gold from secondary sources, consists of three main steps: (i)leaching to extract the gold and bring it into solution, (ii) selectiverecovery of the precious metal by precipitation, ion exchange, sol-vent extraction or adsorption, and (iii) refining it to improve thepurity. The second step, the bridge between the leaching processand the final metal-winning process, is the key step to make

hydrometallurgy a more efficient process [3]. Concentrating thegold from dilute solutions to appropriate concentrations for therefining process is a challenging endeavor. The leaching solutioncontaining the dissolved metals is quite toxic if cyanide is usedas the leaching agent, so technologies are moving away from thistechnique. Another alternative is the use of halides at low pH’sas leaching agents. The most common halide solution for this pur-pose is called ‘‘aqua regia’’, which is a hydrochloric acid/nitric acidmixture with a ratio of 3:1. Usually the gold concentration in theleaching solution is low and several other unwanted metal ionsare present. Especially copper can be found in large excess.

Out of all the possible processes for concentrating the gold, spe-cial attention is directed to the use of solid adsorbents, becausethey are easy to handle and efficient and they produce lessamounts of secondary waste compared to other processes. Theideal adsorbent is one that can selectively adsorb gold very fast,with a high capacity and is easy to regenerate.

Several adsorbents with impressive gold uptake capacitiesranging from 4 mmol Au/g to more than 40 mmol Au/g can befound in literature [2,4–9]; the time they take to reach their max-imum uptake capacity varies from 5 h up to more than 300 h.Capacity is an important characteristic for an adsorbent; however,

L.F. Villalobos et al. / Separation and Purification Technology 136 (2014) 94–104 95

high adsorption rates are crucial for making the process economi-cally viable. To achieve this, the film and pore diffusion limitationsmust be eliminated or reduced to a great extent [10]. Lam et al.grafted organic amine groups to a mesoporous silica support, theeasy access to the adsorption sites through the large cylindricalpores present in the silica allowed this adsorbent to reach equilib-rium in just 30 min with a loading of 1.4 mmol Au/g. SimilarlyHuang et al. immobilized bayberry tannin into a mesoporous silicasupport resulting in fast adsorption rates with a maximum uptakeof 2.21 mmol Au/g. Both of these configurations achieve highadsorption rates, but at the cost of reducing the specific active areaavailable for adsorption.

Tuning the morphology of the adsorbent plays a key role inreducing the pore diffusion limitation. However, this must be donein such a way as to retain the available specific active surface area,so the capacity is not diminished. Membrane technology can beused to reduce or even eliminate the pore diffusion limitation;when a solution is forced to pass through a porous membranethe molecules travel by convective flow through the membranepores, eliminating the usually present long diffusion time thatthe molecules or ions take to reach the adsorption sites insidethe pores [11]. Adsorption sites can be incorporated inside themembrane pores by different methods: (i) dispersion of adsorbentparticles in the membrane matrix, known as mixed matrix mem-branes (MMMs) [12,13]; (ii) incorporation of a liquid carrier inthe membrane matrix by either filling the pores of the membraneswith the carrier, known as supported liquid membranes (SLMs), orby mixing the liquid carrier with the polymer solution beforemembrane casting, know as polymer inclusion membranes (PIMs)[14–18]; (iii) functionalization of polymeric membranes to incor-porate chelate-forming groups, ion-exchange groups or affinityligands [19–23]; (iv) fabrication of membranes from blendsbetween membrane-forming polymers and polymers containingthe appropriate functional groups for the adsorption process[24–28]; (v) fabrication of membranes from metal-complexingpolymers [29–31]. The latter enables the highest concentration ofadsorption sites per membrane volume, because each monomericunit has at least one adsorption site. It is also the easiest methodto fabricate a membrane adsorber and the most scalable one,because it involves the least amount of production steps and themembranes can be fabricated with current industrial manufactur-ing facilities. In spite of this, this method is not common because itis challenging to find a metal-complexing polymer that has both:good film-forming properties to fabricate a mechanically stablemembrane that can withstand the pressure-driven permeationprocess; and monomeric units with adsorption sites that can formstrong chemical complexes with the desired adsorbate.

In this work we propose the use of poly-thiosemicarbazide(PTSC) for the fabrication of asymmetric membranes to recovergold(III) from a solution by permeating the solution through themembrane. Each monomeric unit of this polymer contains two thi-osemicarbazide groups that form a chelate site. This site containsseveral groups that can contribute to the formation of complexeswith metal ions; the structure of PTSC is illustrated in Fig. 1. Camp-bell and Tomic first synthesized poly-thiosemicarbazide in 1962, intheir work they demonstrated that this polymer forms chemicalcomplexes with iron, nickel, silver, copper, mercury and lead

Fig. 1. Poly-thiosemicarbazide structure.

[32,33]. Further tests from our group demonstrated its ability toform strong complexes with gold ions, even at extreme acidic con-ditions. We manufactured this polymer for the first time into a por-ous asymmetric membrane providing the possibility of performingpermeation experiments, in which contact of the gold ions with thechelating group of the polymer is promoted, and all the diffusionallimitations of the adsorption process are reduced or eliminated.

2. Materials and methods

2.1. Membrane fabrication and characterization

Poly-thiosemicarbazide was synthesized according to literatureby the reaction of N,N0-diaminopiperazine and methylenebis(4-phenylisothiocyanate) in dimethyl sulfoxide (DMSO). PTSC mem-branes were prepared by non-solvent induced phase separation(Fig. 2). First a solution of 15 wt% PTSC, 10 wt% 1,4-dioxane and75 wt% DMSO was stirred for 5 h and allowed to rest for another5 h until all the air bubbles had escaped. Then a 250-lm-thickpolymer solution film was cast on a glass plate. The glass platewas then immersed in a non-solvent bath (water) at room temper-ature for at least 12 h. Finally, the membranes were cut in circleswith diameter of 2.5 cm and an average dry weight of 14.3 mgfor the permeation experiments and in squares of 1 cm2 for thebatch experiments. The membranes were stored in a water/ethanolsolution until use.

Scanning electron microscope (SEM) images of the membranesurface and cross-section were taken with a FEI Nova NanoSEM.Cross-section samples were prepared by fracturing the membranesright after having been exposed to liquid nitrogen for about 2 min.Surface and cross-section samples were attached to an aluminumsupport with a conductive aluminum tape and coated with a thinfilm of Iridium to reduce charging during imaging.

2.2. Batch adsorption experiments

Three gold solutions (100, 500 and 1000 ppm Au) were pre-pared by diluting a mother solution of 1000 mg of gold per literof 10% HCl solution obtained from Perkin Elmer atomic spectros-copy standards with a 10% HCl aqueous solution. This solutionwas chosen for all the experiments reported in this work. The con-centration of HCl was kept at 10% in all the experiments.

PTSC squares of 1 cm2 were dried at 45 �C for 24 h and weightedafterwards. Each of the membrane squares was immersed in one ofthe three previously prepared gold solutions. The solutions were inconstant agitation via a shaker and aliquots of 0.2 ml were taken atcertain time intervals to measure the progression of gold adsorp-tion. The gold solution aliquots were analyzed by inductively cou-pled plasma optical emission spectrometry (ICP-OES) using aVarian Inc. 720-ES Inductively Coupled Plasma-Optical EmissionSpectrometer. All the experiments were done at room temperatureunless other temperature is specified.

2.3. Dynamic adsorption experiments

For all these experiments PTSC membrane circles with a diam-eter of 2.5 cm were used. Three circular membranes were driedunder vacuum at 45 �C for 24 h and weighted to obtain an averageweight value. The permeation experiments were done in a Milli-pore Stirred Cell model 8010 with an active membrane area of4.1 cm2, a maximum process volume of 10 ml and a hold-up vol-ume of 0.2 ml. Pressurized air was used to force the solutionsthrough the membrane. A non-woven polyester circle of the samesize of the membrane was used below the membrane to providemechanical support during the permeation experiments; the

(b) (c)(a)

Fig. 2. Membrane casting process. (a) PTSC solution poured in a glass plate and casting knife; (b) PTSC solution film with an approximate thickness of 250 lm; (c) membraneformed after immersion in non-solvent (water) bath.

96 L.F. Villalobos et al. / Separation and Purification Technology 136 (2014) 94–104

complete set up is illustrated in Fig. 3. A difference in pressure assmall as 0.5 bar was sufficient to permeate water through themembranes with an average flux of 101 L/m2 h; increasing the airpressure will result in higher fluxes.

2.3.1. Permeation of gold solutionsFour gold solutions were prepared by diluting the Perkin Elmer

1000 ppm Au atomic spectroscopy standard solution with 10% HCl.The concentrations were 100, 200, 500, and 1000 ppm Au respec-tively. To evaluate the effect of residence time and initial gold con-centration in the adsorption process, 10 ml of each solution werepermeated through a PTSC membrane at two different fluxes:101 L/m2 h using a DP of 0.5 bar and 1868 L/m2 h using the highestpossible pressure recommended by the cell manufacturer as safe(DP of 4 bar). Feed and permeate of each experiment were ana-lyzed using ICP–OES to determine the gold concentration.

2.3.2. Breakthrough curvesA pressurized container (RC800 Mini-Reservoir, Millipore Co.)

was added to the set-up illustrated in Fig. 3 between the pressur-ized air and the stirred cell allowing the continuous permeation ofup to one liter of solution. Three 10% HCl solutions with 10 ppmAu, 100 ppm Au, and 300 ppm Au concentrations respectivelywhere continuously permeated through a PTSC membrane usingthe described set-up. Permeate samples of 5 ml were taken everycertain time throughout the entire experiment to monitor the fluxand the gold concentration. The flux was determined by dividingthe mass flow by the water density and the active membrane area;aliquots were analyzed using ICP–OES to determine the gold con-centration. After these experiments the membranes were driedand analyzed using a thermogravimetric analysis (TGA) (TA Instru-ments Q5000IR) to measure the amount of gold captured by the

Fig. 3. Permeat

membrane. The samples were equilibrated at 100 �C under a nitro-gen flow, afterwards the gas was changed to air and the tempera-ture was raised at a rate of 10 �C/min up to 800 �C, finally in the laststep the temperature was maintained constant for 10 min. Thegold uptake was calculated using the weight after all the water isevaporated from the sample and the weight at the end of the runwhere the organic material was decomposed leaving only the gold.Three measurements for each membrane were done and the aver-age is reported. One blank experiment of a membrane without goldwas measured to warranty that all the organic material is decom-posed under the conditions of the analysis.

2.3.3. Selectivity experimentsCopper is usually the most abundant ion present in the gold

recovery process. For this reason it is important to evaluate theselectivity of poly-thiosemicarbazide between gold and copper.The following solutions were permeated through the PTSC mem-branes in order to evaluate its selectivity for Au ions: (i) 10 ml ofsolution containing 100 ppm Cu, (ii) 10 ml of solution containing100 ppm Cu and 100 ppm Au, (iii) 10 ml of solution containing500 ppm Cu and 100 ppm Au, (iv) 10 ml of solution containing500 ppm Cu and 500 ppm Au, (v) 10 ml of solution containing900 ppm Cu and 100 ppm Au. Feed and permeate of each experi-ment were analyzed using ICP–OES to determine the gold andcopper concentrations.

2.3.4. Regeneration of the membranesRecovery of the metal ions for a possible re-use of the mem-

brane was also investigated. Thiourea forms strong complexes withgold [34]; for this reason its ability to strip gold from the mem-brane was further investigated. The regeneration and reusabilityof the PTSC membrane was studied for up to three consecutive

ion set-up.

L.F. Villalobos et al. / Separation and Purification Technology 136 (2014) 94–104 97

cycles. Each cycle consisted of the permeation of 10 ml of a100 ppm Au in 10% HCl solution through the membrane followedby permeation of 10 ml of a 0.1 M thiourea solution and finally per-meation of 10 ml of water. The gold concentration in the permeateof each step was analyzed using ICP–OES. The membrane was driedand analyzed in the TGA under the same conditions as in thebreakthrough experiments to verify that all the gold was extracted.

2.3.5. Recovery of metals from Red Sea water and tap waterThe PTSC membranes were able to capture gold ions even from

very diluted gold solutions (10 ppm Au). These results suggesttrace metal recovery as a tentative application for these mem-branes. Two experiments were done in order to see which metalscould the membrane recover from tap water and from seawater.In the first experimental set-up a PTSC membrane was mountedin the small Millipore cell described before and connected directlyto the tap water (no pump was needed, the pressure from thewater supply system was enough to run the permeation experi-ment). The tap water available in our laboratory comes from areverse osmosis plant and the distribution system consists mainlyof copper pipes. The experiment ran continuously for 100 days;during this time constant stirring was maintained and the fluxwas monitored sporadically.

For the second experimental set-up a PTSC membrane with adiameter of 5 cm was mounted in a Sterlitech HP4750 stirred cell.The Sterlitech cell was connected to a Millipore pressure filtrationvessel with a capacity of 20 L filled with Red Sea water collectedfrom the shores of King Abdullah University of Science and Tech-nology. The whole system was pressurized to 4 bar, allowing waterto permeate continuously through the membrane for 10 days. Dur-ing this time constant stirring was maintained and the flux wasmonitored every day.

After the permeation experiments the membranes were driedunder vacuum at 60 �C and analyzed qualitatively with an X-rayfluorescence equipment (XGT 7000, Horiba Jobin Yvon) to lookfor the most abundant metals in the sample. A small piece of themembrane used in the tap water permeation experiment was frac-tured, after having been exposed to liquid nitrogen for about 2 min.The cross-section was imaged in the SEM to analyze the structureand the fouling layer. An energy-dispersive X-ray spectrometer(EDX) coupled to the SEM was used to observe the distributionof copper (the most abundant metal captured by the membrane)throughout the membrane’s cross-section. Afterwards the mem-branes were digested in 8 ml of aqua regia (6 ml of concentratedhydrochloric acid and 2 ml of concentrated nitric acid) using amicrowave oven (ETHOS One, Milestone Microwave Systems) at230 �C for 20 min. The resulting solutions were diluted to at least20 ml with MilliQ water and analyzed with an ICP–MS (Elan DRCII, PerkinElmer) to quantify the concentration of the most abun-dant metals (although gold was not shown in the X-ray fluores-cence as one of the most abundant metals, it was also includedin the ICP–MS analysis). Two blanks were prepared for the ICP–MS analysis: the first one consisted in digesting a PTSC membranethat had never been used for any permeation experiment; the sec-ond one consisted in permeating 10 ml of a 100 ppm Au solution,through a PTSC membrane and digesting the membrane after ithad captured all the gold present in the solution.

Fig. 4. SEM images of PTSC membrane. (a) High magnification image of mem-brane’s surface; (b) low magnification image of membrane’s cross-section; (c) highmagnification image of top layer in membrane’s cross-section.

3. Results and discussion

3.1. Membrane characterization

The non-solvent induced phase separation process resulted inthe formation of an asymmetric membrane with a thin dense layerover a porous support. SEM images of the membrane’s surface and

cross-section (Fig. 4) revealed the asymmetric nature of the mem-brane and the presence of macrovoids. The high porosity of themembrane and the presence of macrovoids in the support resultedin high flux membranes with an average flux of 535 L/m2 h using aDP of 1 bar. Macrovoids are often related with low mechanical sta-bility, however this is not the case for this membrane, it presentedgood mechanical stability withstanding pressures up to 20 bar andcontinuous operation over three months. After measuring five SEM

Table 1Pseudo first-order model kinetic parameters for the batch adsorption of gold(III) onPTSC membranes at different initial concentrations (C0).

C0 (ppm) r2 kp1 (min�1) qe (mmol Au/g) qe exp (mmol Au/g)

1000 0.964 1.75E�03 4.93 5.39500 0.959 1.77E�03 3.49 3.70100 0.980 3.61E�03 0.88 0.97

Table 2Pseudo second-order model kinetic parameters for the batch adsorption of gold(III) onPTSC membranes at different initial concentrations (C0).

C0

(ppm)r2 kp2 (g/mmol Au

min)qe (mmol Au/g)

qe exp (mmol Au/g)

1000 0.991 5.05E�04 5.02 5.39500 0.985 7.22E�04 3.55 3.70100 0.997 9.05E�03 1.01 0.97

98 L.F. Villalobos et al. / Separation and Purification Technology 136 (2014) 94–104

images of the membrane’s cross section the resulting averagethickness was 125 lm. The average dry weight of the circularmembranes with a diameter of 2.5 cm was 14.3 mg.

3.2. Batch experiments

An important parameter to compare different adsorbents istheir maximum uptake capacity; the uptake increases with theincrease of equilibrium concentration until the adsorbent is satu-rated. The maximum uptake capacity measured in this work was5.4 mmol Au/g; it was measured at room temperature and in a10% HCl solution. Although we knew that the polymer couldachieve higher uptakes if exposed to higher equilibrium concentra-tions, no experiments were done to push it to its limit, because themembrane’s mechanical stability is compromised when the uptakeof gold is near the value of 5.4 mmol/g. PTSC membranes withuptakes of more than 5.4 mmol Au/g became brittle and even alight stress broke them.

The results of the three batch experiments using different initialconcentrations of gold are presented in Fig. 5.

Pseudo first-order, pseudo second-order, and intra particle dif-fusion models were tested for the experimental data obtained.Comparing the values of Tables 1 and 2 it is clear that the adsorp-tion process followed pseudo second-order kinetics; the coeffi-cients of determination (r2) are higher and the calculatedequilibrium capacities (qe) are closer to the experimental ones(qe exp).

The pseudo first-order and pseudo second-order kinetic modelsdo not account for the possibility of the adsorption being limitedby diffusion, and are unable to determine a precise mechanism;hence the need to test for diffusion limitations. Webber and Morrisnoticed that in several adsorption cases the solute uptake variesproportionally with t1/2 when diffusion is a rate-limiting step[35]. Moreover if the plot passes through the origin, diffusion isthe sole limiting step. It is observed in Fig. 6 that the intraparticlediffusion model matches the experimental data, and that the linearregressions pass through the origin. This is evidence enough tostate that the adsorption process rate-controlling step in this batchexperiment was the diffusion of the gold ions through the pores ofthe membrane. The advantage of membrane adsorbers is that theycan overcome pore-diffusion limitations by operating in a

Fig. 5. Batch adsorption of gold by PTSC for three different initial concentrations:C0 = 1000 ppm Au (j), C0 = 500 ppm Au ðiÞ, and C0 = 100 ppm Au (h).

permeation set-up similar to the one illustrated in Fig. 3 in whichthe solution containing the adsorbate is forced to pass through themembranes pores through convective flow. The permeation set-upanalyzed in the next sections will provide an enhanced contactbetween the gold ions and the chelate sites of the polymereliminating the pore-diffusion limitations present in the batchexperiment, achieving faster rates of adsorption.

3.3. Dynamic experiments

3.3.1. Permeation of gold solutionsThe permeation set-up achieves much faster adsorption rates

than the batch experiment, it recovers more than 99% of the goldpresent in 10 ml of a 100 ppm Au solution operating at fluxes ashigh as 500 L/m2 h.

Fig. 7a shows the effect of flux in the recovery of gold; for a100 ppm Au solution it is observed that after 500 L/m2 h the recov-ery starts to diminish slightly, however even with a flux close to2000 L/m2 h more than 93% of the gold is recovered. The effect ofgold concentration on the recovery at low and high fluxes is pre-sented in Fig. 7b and c respectively. With a flux of 101 L/m2 h arecovery of 90% or higher is achieved for solutions of up to500 ppm Au. On the other hand, when using a high flux (1868 L/m2 h), the recovery is diminished, but high levels are still main-tained: 93%, 78%, and 44% for 100 ppm Au, 200 ppm Au and500 ppm Au respectively. The gold uptake of the membrane afterthe permeation of 1000 ppm Au solutions at low and high fluxesis 3.25 mmol Au/g and 1.13 mmol Au/g respectively, and wasachieved in less than 15 min for the low flux case and in less than1 min for the high flux case.

3.3.2. Breakthrough curvesIt is important to know how much gold the membrane can

adsorb before it saturates and its performance begins to decrease.For this reason continuous permeation experiments were done forthree different initial concentrations, and the resulting curves ofrelative gold concentration on the permeate (C/C0) versus the efflu-ent volume (permeate volume) are shown in Fig. 8. A relevantpoint in these curves is the breakthrough point that is definedwhen a preset target concentration emerges in the effluent. For thisanalysis the breakthrough point is set to C/C0 = 0.1 correspondingto the point where the permeate concentration of gold is 10% ofthe feed concentration.

When a solution of 10 ppm Au was permeated continuouslythrough a PTSC membrane of 2.5 cm in diameter, almost all thegold was recovered for the first 550 ml that passed through the

(a)

(b)

L.F. Villalobos et al. / Separation and Purification Technology 136 (2014) 94–104 99

membrane (Fig. 8a). At this point (breakthrough point) the mem-brane gold uptake, also referred to as breakthrough capacity, was2.6 mmol Au/g and the membrane had captured 6 mg of gold.Almost complete exhaustion of the membrane occurred afteraround 1000 ml of solution had been permeated, at this point theexperiment was stopped. The uptake of the membrane after thisexperiment was 3.8 mmol Au/g.

As expected the PTSC membranes saturate faster when moreconcentrated gold solutions were permeated through them. Suchcases are depicted in Fig. 8b and c; for an initial concentration of100 ppm Au at a low flux (170 L/m2 h) the breakthrough point islocated at 50 ml; while at a flux more than four times greater(800 L/m2 h), the breakthrough point is half this value. Lowerfluxes will result in greater amounts of gold recovered per volumeof solution permeated, however higher fluxes will increase theoverall throughput. In the case of an initial concentration of300 ppm Au, the PTSC membrane saturated very fast. The break-through point was reached just after 5 ml of solution had passedthrough the membrane, however it still recovers more than 50%of the gold present in the solution for up to 20 ml. This experimentwas stopped after 235 ml had been permeated; at this point themembrane had an uptake of 3.38 mmol Au/g.

High permeate fluxes resulted in the gold ions spending a lowerresidence time inside the membrane, allowing less time for themto react with the adsorption sites. After analyzing Fig. 9a it canbe concluded that although the membrane configuration elimi-nates the pore diffusion limitations, the adsorption of gold ions isstill mass transfer limited; the rate of gold capture increases athigher fluxes, lower residence times, which is the case for a masstransport limited process. If the adsorption reaction were the lim-iting step, no matter how high the flux, the adsorption rate wouldbe constant. In this case, the convective transport of gold ions tothe adsorption sites is the limiting step. Moreover, higher fluxeswill produce a steeper decrease on the available adsorption sitesallowing more gold ions to pass through the membrane withoutbeing adsorbed, deviating the process from the ideal case whereall the gold is trapped. This can be observed in Fig. 9b where thedeviation from the ideal case (straight line) is larger for the highflux (low residence time) experiment.

Fig. 6. Linear fitting of qt versus t1/2 (intraparticle diffusion model) to theexperimental data of the adsorption of gold on PTSC at different initial con-centrations: C0 = 1000 ppm Au (j), C0 = 500 ppm Au (ðiÞ), and C0 = 100 ppm Au(h).

(c)

Fig. 7. Influence of flux and gold concentration on the recovery; gold recoveredfrom solution (j), gold uptake by the membrane (h). (a) Effect of flux for a 100 ppmAu solution; (b) effect of gold concentration for a flux of 101 L/m2 h; (c) effect ofgold concentration for a flux of 1868 L/m2 h.

(a) (b) (c)

Fig. 8. Breakthrough curves at different gold initial concentrations (C0). (a) C0 = 10 ppm Au, flux of 520 L/m2 h; (b) C0 = 100 ppm Au, flux of 800 L/m2 h (j) and 170 L/m2 h (h);(c) C0 = 300 ppm Au, flux of 540 L/m2 h.

(a)

(b)

Fig. 9. (a) Gold captured as a function of time for an initial concentration of100 ppm Au and two different fluxes: 800 L/m2 h (j) and 170 L/m2 h (h); (b) effectof residence time on the saturation of the PTSC membrane (800 L/m2 h (j) and170 L/m2 h (h)). The solid line represents the ideal case in which all the gold fed iscaptured by the membrane.

Table 3Selectivity of PTSC membrane towards gold ions. Experiments made by permeating10 ml solutions at 1868 L/m2 h.

Initial goldconcentration(ppm Au)

Initial copperconcentration(ppm Cu)

Goldrecovered(%)

Copperrecovered(%)

Golduptake(mmol Au/g)

100 0 93 – 0.400 100 – 24 –

100 100 99 15 0.42100 500 91 11 0.39100 900 97 16 0.41500 0 44 – 0.95500 500 46 3 0.98

100 L.F. Villalobos et al. / Separation and Purification Technology 136 (2014) 94–104

Depending on the gold concentration of the feed, the mem-brane’s area and flux can be adjusted to meet the desired recoveryand throughput. The gold concentration in real feeds differs widelydepending on their source; however in the vast majority of cases itwill be in a range where the membrane presented in this workoperates optimally. In the case of gold recovery from electronicscrap, the minimum volume of aqua regia needed per gram ofscrap to achieve a complete immersion is approximately two mil-liliters [36]. An optimized process reported a volume of 20 ml ofaqua regia per gram of scrap [37]. Taking into account this and thatthe gold content in mobile phones range from 300 to 350 g/t and incomputer circuit boards from 200 to 250 g/t [38], the gold concen-tration in electronic scrap real feeds can be easily estimated. It willcommonly range from 14 to 25 ppm Au and have an upper limit of175 ppm Au. For all these concentrations, the PTSC membrane pre-sented in this work can achieve high recoveries at very high fluxes.

3.3.3. Selectivity experimentsDue to the fast kinetics of adsorption the selectivity experi-

ments were done at high fluxes and the results are presented inTable 3. Comparing the runs made with 100 ppm of single compo-nents (gold and copper), we can see that 93% of gold was recoveredwhile in the case of copper only 24%. On the other hand, a run con-taining 100 ppm of both metals in the same solution resulted in99% recovery of gold and only 15% recovery of copper. The higherselectivity of PTSC towards gold and the fact that higher selectivityis obtained in mixed component experiments (Au/Cu = 6.7) com-pared to single components experiments (Au/Cu = 3.9) can beexplained with the hard-soft acid-base (HSAB) concept [39]. Bothsulfur and nitrogen atoms in the chelate site have an easily polar-izable ion pairs of electrons (soft base), which could interact stron-ger and faster with soft acids. Cu2+ is classified as a borderline acidwhile Au3+ is a soft acid [40], for this reason Au3+ will react faster

Table 4Membrane regeneration using 0.1 M thiourea solution. Each cycle consisted onpermeating first 10 ml of a 100 ppm Au solution, then 10 ml of 0.1 M thiourea andfinally 10 ml of H2O.

Cycle(#)

Gold recovered by PTSCmembrane (%)

Gold eluted from the membrane by0.1 M thiourea (%)

1 100 972 100 973 100 99

(a)

(b)

Fig. 10. Flux decrease during the experiments. (a) Tap water; (b) Red Sea water.

L.F. Villalobos et al. / Separation and Purification Technology 136 (2014) 94–104 101

and form more stable complexes with the chelate sites present inpoly-thiosemicarbazide. This is even more evident in the run con-taining 500 ppm of both metals in the test solution, which showedlittle adsorption of copper due to the presence of a high concentra-tion of the faster adsorbing gold ions (Au/Cu = 14.6). Furthermore,the recovery of gold was not affected at all by the presence of veryhigh concentrations of copper ions; even a 9-fold excess amount ofcopper did not decrease the gold recovery.

As can be observed in Table 3 the recovery of gold for a 100 ppmAu feed solution with concentrations of copper ions ranging fromzero to 900 ppm Cu was maintained above 90%. No trend can beobserved form the results but taking into consideration theexperimental error associated with the high flux used in theseexperiments, it is adequate to conclude that the presence of copperdid not affected the gold recovery.

3.3.4. Regeneration of the membranesWe can observe in Table 4 that the membrane did not loose any

capacity after the three cycles and that a diluted solution of thio-urea was enough to regenerate the membrane and elute all thegold from it.

There are two objectives for the third step in which we perme-ated pure water: (i) take out of the system the remaining goldeluted by the previous step present in the hold-up volume of thecell; (ii) clean the system before the next cycle. The flux remainedconstant throughout the three cycles with an average value of490 L/m2 h. The white PTSC membrane changed color to an intensebrown after the gold adsorption step and after the elution step itgoes back to its original color but maintains a slightly gray shade.After the three cycles the membrane was dried at 60 �C under vac-uum and three small pieces of it were analyzed in the TGA along ablank sample (PTSC membrane that has never been in contact witha gold solution). The TGA analysis of the three samples of the PTSCmembrane that was used for the three adsorption cycles showedan average residue weight percentage of 1.4 wt% while the blankmeasurement showed a residue weight percentage of 0.7 wt%.These results confirms that the thiourea solution is able to eludealmost all the gold captured by the membrane. The amount of goldthat remained in the membrane after the three cycles was0.069 mg according to the ICP-OES measurements and 0.072 mgaccording to the TGA measurements. This means that only 2 wt%of the total gold captured by the membrane during the three cycleswas not stripped by the 0.1 M thiourea elution steps. Finally, anelectrowinning process can process the gold concentrated in thethiourea solution into the final product [41].

3.3.5. Recovery of metals from Red Sea water and tap waterThe flux decreased in both cases throughout the experiments.

However, for the tap water the decrease was gradual, while inthe case of seawater the flux suffered a sharp decrease the firstdays and then stabilized. The behavior mentioned before anddepicted in Fig. 10a and b is, to some extent, due to fouling ofthe membrane and also to the adsorption of metals by the mem-brane [42]. A plausible explanation for the sharp decrease in theflux during the first few days, in the case of seawater, could bethe higher fouling potential of the water, collected from the shoresof the Red Sea and used without any pretreatment, in comparisonto tap water.

As shown in Fig. 11, the membranes change their color fromwhite to brown after the permeation experiment. An evidentchange of color occurred on both sides of the membranes: the sidefacing the feed presented an homogeneous dark brown while theother side presented a majority of brown color with some whitespots.

The fact that the color changed in both sides on the membraneis evidence enough to state that the change in color was primarily

due to the adsorption of metals and to a lesser extent due to theformation of a fouling layer. Qualitative analysis of the surface ofthe membrane using XRF revealed that the metal recovered in lar-ger quantities was copper, followed by zinc and nickel. SEM imagesof the membrane used for the permeation of tap water for100 days, show the formation of a fouling layer on the surfaceand no changes on the general structure of the membrane. It canbe observed in Fig. 12 how the structure endured operation for

Fig. 11. Photos of the poly-thiosemicarbazide membrane before and after the continuous permeation of tap water for 100 days.

Fig. 12. SEM images of the membrane’s cross-section after the continuouspermeation of tap water for 100 days.

102 L.F. Villalobos et al. / Separation and Purification Technology 136 (2014) 94–104

more than three months without collapsing, and how a foulinglayer of approximately 2.7 lm was formed. The EDX elementalmapping shown in Fig. 13 revealed that copper, the metal recov-ered in the largest quantity, was adsorbed homogenously through-out the membrane and did not concentrate only on the surface.

Fig. 13. Cross-seetion scanning electron microscope image and X-ray element map of tSecondary electron image, 10 keV (b) C map; (c) Cu map.

Table 5Amount of nickel, copper, zinc, and gold recovered by PTSC membranes.

Sample Nickel cont

Unused PTSC membrane 0PTSC membrane after permeating 10 ml of 100 ppm Au solution 0PTSC membrane after permeating 20 L of seawaterb 17.5PTSC membrane after continuously permeate tap water for 100 days 0.2a

a Approximate values.b Membrane with a diameter of 5 cm was used for this experiment.

The metal content of the membranes after the permeationexperiments is presented in Table 5, along with the results fromtwo blanks: an unused PTSC membrane and a PTSC membrane thatcaptured all the gold present in 10 ml of a 1000 ppm Au solution.Both blanks presented no nickel, and negligible amounts of copperand zinc. The measured gold content of the PTSC membrane usedto recover the gold from a 1000 ppm Au solution, matches theamount of metal present in the original 10 ml of solution thatwas permeated through the membrane. The unused PTSC mem-brane, as expected, did not contain any gold.

Poly-thiosemicarbazide membranes were able to recover met-als from solutions containing very low concentrations, like tapwater and seawater. The concentration of metals captured in largerquantities by the membrane used with tap water was 10.9 ppb forcopper and 2.7 ppb for zinc. An approximate permeate volume of67 L was calculated by fitting an exponential function to the fluxdecay data. Assuming the metal content in tap water was relativelyconstant throughout the duration of the experiment, and using theapproximate permeate volume calculated, the PTSC membranerecovered 16 wt% of the copper and 13 wt% of the zinc present inthe tap water.

After permeating 20 L of seawater, the PTSC membrane recov-ered some of the metals present in the water. Interestingly, theamount of nickel, zinc, and copper captured by the membranewas orders of magnitude more than the expected content of these

he PTSC membrane after the continuous permeation of tap water for 100 days. (a)

ent (lg) Copper content (lg) Zinc content (lg) Gold content (lg)

0.5a 1.3a 00.3a 0.7a 1062.0

1336.0 195.8 1.0a

115.4 23.8 0.9a

L.F. Villalobos et al. / Separation and Purification Technology 136 (2014) 94–104 103

metals in 20 L of Red Sea water [43]. These results suggest that theconcentration of these metals in the area where the sample wastaken were much higher than the average reported. Higher concen-trations are often linked to site-specific variables like intenseanthropogenic activity, pseudo stagnant conditions, shallow water,and the potential role of ecosystems to serve as sinks for thesemetals as is the case for mangrove ecosystems [44]. The samplingsite of the experiment presented in this work met all of the afore-mentioned conditions, which is the most likely explanation for thehigh metal content found in the membrane. Although the amountof gold captured by the membrane after permeating 20 L of seawa-ter could not be precisely determined, a well-defined peakcorresponding to an approximate concentration of 5.2 ppb wasobserved in the ICP–MS; while in the case of the blank membraneno peak was detected at all. The amount of gold captured by thePTSC membrane corresponds to a concentration in solution in therange of parts per trillion, which is in concordance with thereported range for concentration of gold in the oceans [45–47].Even though the PTSC membranes are able to recover metals fromseawater, the limitation is the large amount of energy needed topermeate enough water to recover significant amounts of thesemetals. The results presented in this section have gone some waytowards unveiling the potential applications of poly-thiosemicar-bazide membranes for the recovery of metals from very dilutedsolutions; nevertheless further research is needed to test theperformance of these membranes in specific scenarios.

4. Conclusions

Using chelating polymers to manufacture membranes providesa simple approach to obtain membrane adsorbers with a highcapacity and fast adsorption rates. Poly-thiosemicarbazide mem-branes presented in this work have proven to be a promising alter-native for the selective extraction of gold(III) ions from solutions.They can be easily manufactured in current industrial set ups,installed in a very compact module at any point along a process,and can recover gold as fast as needed by controlling the mem-brane area and flux. PTSC membranes reached gold uptakes com-parable to the best adsorbents available, and much higheradsorption rates due to convective mass transport inside the pores,that overcomes diffusion limitations. Additionally, the goldtrapped in the membrane can be easily stripped, allowing for themembrane to be used again without any decrease in its perfor-mance. The versatility, scalability and easy manufacturing of theprocess proposed in this work, along with its efficiency and lowenergy consumption, makes it a real solution to the sustainabilityissues usually present in the gold industry. Poly-thiosemicarbazidemembrane adsorbers applications are not limited to gold recovery,as it can be seen in the initial findings presented in Section 3.3.5.These results suggest potential applications for the membrane inthe treatment of metal-contaminated water.

Acknowledgments

The authors acknowledge Dr. Pradeep Neelakanda and Dr.Madhavan Karunakaran for helping with the synthesis of PTSC.The authors also acknowledge Rodrigo Valladares for valuable dis-cussions. Research reported in this publication was supported bythe King Abdullah University of Science and Technology (KAUST).

References

[1] S. Syed, Recovery of gold from secondary sources – a review, Hydrometallurgy115 (2012) 30–51.

[2] T. Ogata, Y. Nakano, Mechanisms of gold recovery from aqueous solutionsusing a novel tannin gel adsorbent synthesized from natural condensed tannin,Water Res. 39 (2005) 4281–4286.

[3] D.S. Flett, Solution purification, Hydrometallurgy 30 (1992) 327–344.[4] C.R. Adhikari, D. Parajuli, K. Inoue, K. Ohto, H. Kawakita, H. Harada, Recovery of

precious metals by using chemically modified waste paper, New J. Chem. 32(2008) 1634–1641.

[5] C.R. Adhikari, D. Parajuli, H. Kawakita, K. Inoue, K. Ohto, H. Harada,Dimethylamine-modified waste paper for the recovery of precious metals,Environ. Sci. Technol. 42 (2008) 5486–5491.

[6] B. Pangeni, H. Paudyal, M. Abe, K. Inoue, H. Kawakita, K. Ohto, B.B. Adhikari, S.Alam, Selective recovery of gold using some cross-linked polysaccharide gels,Green Chem. 14 (2012) 1917–1927.

[7] B. Pangeni, H. Paudyal, K. Inoue, H. Kawakita, K. Ohto, S. Alam, An assessmentof gold recovery processes using cross-linked paper gel, J. Chem. Eng. Data 57(2012) 796–804.

[8] Y. Xiong, C.R. Adhikari, H. Kawakita, K. Ohto, K. Inoue, H. Harada, Selectiverecovery of precious metals by persimmon waste chemically modified withdimethylamine, Bioresource Technol. 100 (2009) 4083–4089.

[9] X. Li, C. Zhang, R. Zhao, X. Lu, X. Xu, X. Jia, C. Wang, L. Li, Efficient adsorption ofgold ions from aqueous systems with thioamide-group chelating nanofibermembranes, Chem. Eng. J. (2013).

[10] V. Vadivelan, K.V. Kumar, Equilibrium, kinetics, mechanism, and processdesign for the sorption of methylene blue onto rice husk, J. Colloid InterfaceSci. 286 (2005) 90–100.

[11] E. Klein, Affinity Membranes: Their Chemistry and Performance in AdsorptiveSeparation Processes, Wiley, New York, 1991.

[12] A. Ladhe, P. Frailie, D. Hua, M. Darsillo, D. Bhattacharyya, Thiol-functionalizedsilica-mixed matrix membranes for silver capture from aqueous solutions:experimental results and modeling, J. Membr. Sci. 326 (2009) 460–471.

[13] K.K.R. Tetala, D.F. Stamatialis, Mixed matrix membranes for efficientadsorption of copper ions from aqueous solutions, Sep. Purif. Technol. 104(2013) 214–220.

[14] G. Argiropoulos, R.W. Cattrall, I.C. Hamilton, S.D. Kolev, R. Paimin, The study ofa membrane for extracting gold(III) from hydrochloric acid solutions, J.Membr. Sci. 138 (1998) 279–285.

[15] C. Fontas, R. Tayeb, M. Dhahbi, E. Gaudichet, F. Thominette, P. Roy, K.Steenkeste, M.P. Fontaine-Aupart, S. Tingry, E. Tronel-Peyroz, P. Seta, Polymerinclusion membranes: the concept of fixed sites membrane revised, J. Membr.Sci. 290 (2007) 62–72.

[16] C.A. Kozlowski, W. Walkowiak, Removal of chromium(VI) from aqueoussolutions by polymer inclusion membranes, Water Res. 36 (2002) 4870–4876.

[17] I. Van de Voorde, L. Pinoy, R.F. De Ketelaere, Recovery of nickel ions bysupported liquid membrane (SLM) extraction, J. Membr. Sci. 234 (2004) 11–21.

[18] J.Y. Xu, W. Shen, R. Paimin, X.G. Wang, Improved membranes for the extractionof heavy metals, Fiber Polym. 5 (2004) 68–74.

[19] D. Bhattacharyya, J.A. Hestekin, P. Brushaber, L. Cullen, L.G. Bachas, S.K. Sikdar,Novel poly-glutamic acid functionalized microfiltration membranes forsorption of heavy metals at high capacity, J. Membr. Sci. 141 (1998) 121–135.

[20] S. Konishi, K. Saito, S. Furusaki, T. Sugo, Sorption kinetics of cobalt in chelatingporous membrane, Ind. Eng. Chem. Res. 31 (1992) 2722–2727.

[21] S.M.C. Ritchie, L.G. Bachas, T. Olin, S.K. Sikdar, D. Bhattacharyya, Surfacemodification of silica- and cellulose-based microfiltration membranes withfunctional polyamino acids for heavy metal sorption, Langmuir 15 (1999)6346–6357.

[22] E. Salehi, S.S. Madaeni, F. Heidary, Dynamic adsorption of Ni(II) and Cd(II) ionsfrom water using 8-hydroxyquinoline ligand immobilized PVDF membrane:isotherms, thermodynamics and kinetics, Sep. Purif. Technol. 94 (2012) 1–8.

[23] S. Tsuneda, K. Saito, S. Furusaki, T. Sugo, J. Okamoto, Metal collection usingchelating hollow fiber membrane, J. Membr. Sci. 58 (1991) 221–234.

[24] H. Bessbousse, T. Rhlalou, J.F. Verchere, L. Lebrun, Removal of heavy metal ionsfrom aqueous solutions by filtration with a novel complexing membranecontaining poly(ethyleneimine) in a poly(vinyl alcohol) matrix, J. Membr. Sci.307 (2008) 249–259.

[25] Z.A. Chen, M.C. Deng, C. Yong, G.H. He, W. Ming, J.D. Wang, Preparation andperformance of cellulose acetate/polyethyleneimine blend microfiltrationmembranes and their applications, J. Membr. Sci. 235 (2004) 73–86.

[26] L. Lebrun, F. Vallee, B. Alexandre, Q.T. Nguyen, Preparation of chelatingmembranes to remove metal cations from aqueous solutions, Desalination 207(2007) 9–23.

[27] C.X. Liu, R.B. Bai, Adsorptive removal of copper ions with highly porouschitosan/cellulose acetate blend hollow fiber membranes, J. Membr. Sci. 284(2006) 313–322.

[28] E. Salehi, S.S. Madaeni, L. Rajabi, V. Vatanpour, A.A. Derakhshan, S. Zinadini, S.Ghorabi, H.A. Monfared, Novel chitosan/poly(vinyl) alcohol thin adsorptivemembranes modified with amino functionalized multi-walled carbonnanotubes for Cu(II) removal from water: preparation, characterization,adsorption kinetics and thermodynamics, Sep. Purif. Technol. 89 (2012)309–319.

[29] A. Denizli, S. Bektas, Y. Arica, O. Genc, Metal-chelating properties of poly(2-hydroxyethyl methacrylate-methacryloylamidohistidine) membranes,J. Appl. Polym. Sci. 97 (2005) 1213–1219.

[30] W. Kaminski, Z. Modrzejewska, Application of chitosan membranes inseparation of heavy metal ions, Separ. Sci. Technol. 32 (1997) 2659–2668.

[31] Z. Modrzejewska, W. Kaminski, Separation of Cr(VI) on chitosan membranes,Ind. Eng. Chem. Res. 38 (1999) 4946–4950.

[32] T.W. Campbell, E.A. Tomic, Polythiosemicarbazides. 1. Preparation andproperties of polymers and some simple metallic chelates, J. Polym. Sci. 62(1962) 379–386.

104 L.F. Villalobos et al. / Separation and Purification Technology 136 (2014) 94–104

[33] E.A. Tomic, V.S. Foldi, T.W. Campbell, Polythiosemicarbazides. 2. Effect ofchemical structure on chelating properties – structure of chelates, J. Polym. Sci.62 (1962) 387–399.

[34] M. Nicol, C. Fleming, R. Paul, The chemistry of the extraction of gold, ExtractionMetallurgy of Gold in South Africa, SAIMM (1987).

[35] W. Weber, J. Morris, Kinetics of adsorption on carbon from solution, J. Sanit.Eng. Div. 89 (1963) 31–60.

[36] P.P. Sheng, T.H. Etsell, Recovery of gold from computer circuit board scrapusing aqua regia, Waste Manage. Res. 25 (2007) 380–383.

[37] Y.J. Park, D.J. Fray, Recovery of high purity precious metals from printed circuitboards, J. Hazard. Mater. 164 (2009) 1152–1158.

[38] C. Hagelüken, C.W. Corti, Recycling of gold from electronics: cost-effective usethrough ‘Design for Recycling’, Gold Bull. 43 (2010) 209–220.

[39] R.G. Pearson, Hard and soft acids and bases, J. Am. Chem. Soc. 85 (1963) 3533–3539.

[40] W. Stumm, J.J. Morgan, Aquatic Chemistry: Chemical Equilibria and Rates inNatural Waters, John Wiley & Sons, 2012.

[41] C. Juarez, A. Dutra, Gold electrowinning from thiourea solutions, Miner. Eng.13 (2000) 1083–1096.

[42] L.F. Villalobos, P. Neelakanda, M. Karunakaran, D. Cha, K.V. Peinemann, Poly-thiosemicarbazide/gold nanoparticles catalytic membrane: in-situ growth ofwell-dispersed, uniform and stable gold nanoparticles in a polymericmembrane, Catalysis Today, <http://dx.doi.org/10.1016/j.cattod.2013.10.067>.

[43] L.G. Danielsson, Cadmium, cobalt, copper, iron, lead, nickel and zinc in Indian-ocean water, Mar. Chem. 8 (1980) 199–215.

[44] W. Machado, E.V. Silva-Filho, R.R. Oliveira, L.D. Lacerda, Trace metal retentionin mangrove ecosystems in Guanabara Bay, SE Brazil, Mar. Pollut. Bull. 44(2002) 1277–1280.

[45] K.K. Falkner, J.M. Edmond, Gold in seawater, Earth Planet Sci. Lett. 98 (1990)208–221.

[46] M. Koide, V. Hodge, E.D. Goldberg, K. Bertine, Gold in seawater: a conservativeview, Appl. Geochem. 3 (1988) 237–241.

[47] J.B. Mchugh, Concentration of gold in natural-waters, J. Geochem. Explor. 30(1988) 85–94.