- Email: [email protected]
Evaluation of residence time distribution and mineralogical characterization of the biooxidation of sulfide minerals in a continuous stirred tank reactor
Minerals Engineering 46–47 (2013) 128–135
Contents lists available at SciVerse ScienceDirect
Minerals Engineering journal homepage: www.elsevier.com/locate/mineng
Evaluation of residence time distribution and mineralogical characterization of the biooxidation of sulfide minerals in a continuous stirred tank reactor Diana M. Arroyave G. a,⇑, Darío Gallego S. b, Marco A. Márquez G. a a b
Applied Mineralogy and Bioprocess Research Group, School of Materials, National University of Colombia, Medellin, Colombia Sanitary Engineering Laboratory, School of Processes and Energy, National University of Colombia, Medellin, Colombia
a r t i c l e
i n f o
Article history: Received 10 April 2012 Accepted 13 March 2013 Available online 3 May 2013 Keywords: RTD Biooxidation Tracer Jarosite XRD CSTR
a b s t r a c t The residence time distribution (RTD) of the liquid phase and the mineralogical characterization of the biooxidation of refractory gold mineral was studied in a continuous stirred tank reactor. The latter was achieved using a native mixed culture of acidophilic mesophiles. The RTD was assessed using a mathematical model of stirred tanks in parallel. The oxidation of sulfide and the phases generated through the biooxidation process were evaluated via X-ray diffraction (XRD).The results indicated that the experimental RTD fit to the model. The reactor has a high tendency to behave as a completely mixed reactor. However, the mixed flow inside the reactor has disturbances such as by-pass and dead zones. The estimated mean residence time for the model was approximately 36% greater than the theoretical residence time. It was probably caused by the delay in the outflow of the tracer due to gas hold-up, foaming at the top and the design of the reactor outlet structure. The XRD outcomes showed that the oxidation of arsenopyrite was greater than that of pyrite. Similarly, the formation of jarosite and brushite was observed. It was concluded that the dead zones could increase the probability of jarosite precipitations. Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction The biooxidation of sulfide minerals in continuous stirred tank reactors (CSTR) has become an important method for the pretreatment of refractory gold ore (Chandraprabha et al., 2002; Akcil, 2004; Watling, 2008). Industrial-scale biooxidation was first carried out in South Africa (Fairview) in 1986 as pretreatment for refractory gold concentrates. Currently, CSTR biooxidation is being used successfully in commercial operation in countries such Brazil, Peru, Australia, Ghana, South Africa, India and China (Gonzalez et al., 2003; Akcil, 2004; Sand and Gehrke, 2006; Watling, 2008). The growth of this technology is promising, and new projects are currently in the research and engineering stages (Van Niekerk, 2009). Biooxidation is catalyzed by bacteria that oxidize reduced iron and sulfur compounds. In industrial applications of mining and metallurgy, cultures of mesophiles or moderate thermophiles are widely used for the oxidation of sulfide ores (Rossi, 1990; Rawlings et al., 2003; Akcil, 2004; Sand and Gehrke, 2006). Such microorganisms include the iron- and sulfur-oxidizing Acidithiobacillus ferrooxidans, the sulfur-oxidizing Acidithiobacillus thiooxidans and Acidithiobacillus caldus, and the iron-oxidizing Leptospirillum ferrooxidans and Leptospirillum ferriphilum. (Coram and Rawlings, 2002; Rawlings et al., 2003; Bryan et al., 2011). The operations tend to be ⇑ Corresponding author. Tel.: +57 (4) 4255341/4255342. E-mail address: [email protected] (D.M. Arroyave G.). 0892-6875/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.mineng.2013.03.012
dominated by L. ferriphilum and A. caldus at 40–45 °C (Rawlings, 2007). The success of bacterial sulfide oxidation in a CSTR requires a mixing that reaches a degree of uniformity that maintains suitable conditions for bacterial growth and solids suspension, thus minimizing the development of concentration, temperature and pH gradients within the reactor (Rossi, 1990; Hayward et al., 1997; González et al., 2003; Deveci, 2004). Currently, the research and development areas related to biooxidation technology include the development of an improved agitation system, and the reduction of retention time (Van Niekerk, 2009). The scientific literature provides a number of studies on different impellers and their effects on bacterial activity in CSTR biooxidation (d’Hugues et al., 1997; Deveci, 2002, 2004; Liu et al., 2007; Sun et al., 2012). It has been demonstrated that within in the biooxidation process the best alternative is the use of axial flow impellers (Dew et al., 1997; Deveci, 2002, 2004; Gonzalez et al., 2003; Van Aswegen et al., 2007). Indeed, nowadays all commercial stirred-tank biooxidation plants use axial flow hydrofoil impellers (Dew et al., 1997; Arrascue and Van Niekerk, 2006; Batty and Rorke, 2006; Van Aswegen et al., 2007; Van Niekerk, 2009). Hydrodynamic behavior studies in biooxidation reactors are limited in the literature. The mathematical description of residence time distribution (RTD) is usually expressed either through the dispersion model or through the tanks-in-series model. Romero et al. (1998) determined the RTD for two stirred tank reactors in series at the Rio Tinto plant in Spain. The RTD curve that best fit the experimental data was the two completely stirred tanks-in-series model.
129
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
Nomenclature Ci E(t) E(h) Em(h) Ee(h) G(s) ti (h)tm s
tracer concentration in the effluent at time ti (mg/L) Residence time distribution function Dimensionless residence time distribution function residence time distribution of the model experimental residence time distribution transfer function of the stirred tanks time at which the sample is taken from the reactor effluent during the tracer test experimental mean residence time (h) complex variable
Mazuelos et al. (2002) reported the assessment of RTD at various bed heights of a packed bed reactor using the tanks-in-series model. The results indicated that for the maximum bed height of 84 cm, the hydrodynamic behavior fits two completely mixed tanks in series. A hydrodynamic assessment of a reactor conducted by analyzing the residence time distribution (RTD) is crucial to determine possible flow abnormalities such as by-pass and dead zones which may affect the process operation and performance (Levenspiel, 1981; Andrade and Hodouin, 2005). Additionally, RTD can be used together with biooxidation kinetic models to simulate these processes (Brochot et al., 2004). The previous statement shows that characterizing the hydrodynamic behavior in biooxidation reactors is relevant. Mixing conditions in stirred tank reactors for the biooxidation of gold concentrates depend on many factors including pulp density, aeration and agitation rate, reactor design features such as tank geometry, impeller type and diameter, and air injection type and location (Gonzalez et al., 2003; Deveci, 2004). It is therefore necessary to determine the RTD for each individual case and thus guarantee a good process performance. In a reactor, hydrodynamic evaluation is performed using a tracer test: a tracer is injected into the reactor’s feed and the response to this stimulus is recorded. The response analysis provides information about the mixing properties of the system and the performance of non-ideal chemical reactors (Levenspiel, 1981). The aim of this paper was to determine the RTD of the liquid phase of a continuous stirred tank reactor used for biooxidizing sulfides at laboratory scale. This was done in order to characterize the mixing conditions established by the operating conditions and reactor design. The RTD was analyzed through a developed mathematical model of completely stirred tanks in parallel. After determining the RTD the bacterial oxidation of sulfides was evaluated and a mineralogical characterization through X-ray diffraction (XRD) was performed to determine the level of sulfide oxidation and the phases generated in the biooxidation process, respectively.
s1 s2
mean residence time in the large stirred tank (h) mean residence time in each small stirred tank in series (h) mean residence time of the three agitated tanks in series (h) flow fraction of the upper path flow fraction of the lower path dimensionless time (ti/tm)
s3 fq f1-q h
GðsÞ ¼
Z
1
EðtÞ est dt
ð1Þ
0
where s is a complex variable. The response of any array of ideal reactors can be derived from the standard complex algebraic domain rules, and then converted into its time domain expression (Levenspiel, 1981). The solution of Eq. (1) for each ideal continuous stirred tank in Fig. 1 is represented by a transfer function in the Laplace domain in terms of mean residence time of the liquid inside the reactor (s) by Eq. (2) (Yianatos et al., 2005; Andrade and Hodouin, 2005):
GðsÞ ¼
1
ð2Þ
ss þ 1
The overall transfer function for the arrangement of tanks in parallel in Fig. 1 is given by Eq. (3):
GðsÞ ¼ ð1 fq Þ G1 ðsÞ G2 ðsÞ2 þ fq G3 ðsÞ3
ð3Þ
where fq is the flow fraction circulating through the upper branch, f1q is the fraction of the flow circulating through the lower branch, G1(s) is the transfer function of the large stirred tank, G2(s) is the transfer function of each ideal small stirred tank in series with the large tank, and G3(s) is the transfer function of each of the three stirred tanks in series in the upper branch of the array in parallel. By substituting Eq. (2) in each stirred tank in Eq. (3), the transfer function in terms of the parameters of the tanks-in-parallel is obtained as follows (Eq. (4)):
GðsÞ ¼
ð1 fq Þ ðs1 s þ 1Þ ðs2 s þ 1Þ
2
þ
fq ðs3 s þ 1Þ3
ð4Þ
If Eq. (4) is decomposed into partial fractions, Eq. (5) is obtained.
GðsÞ ¼
ð1 fq Þ s1
"
ðs1 s2 Þ2 " # 1 1
ðs þ s2 Þ
2
# 1 1 ð1 fq Þ ðs1 s2 Þ s2 ðs þ s11 Þ ðs þ s12 Þ
þ
fq
s ðs þ s13 Þ3 3 3
2. Stirred tanks in parallel model This model consists of three small perfectly mixed reactors in parallel with a combination of a large perfectly mixed reactor in series with two smaller reactors (Fig. 1). A very useful concept for modeling the RTD of real reactors is the transformation of RTD into the Laplace domain (Pinheiro and Oliveirah, 1998; Yianatos et al., 2005), because the Laplace transform is used to solve problems where time is the independent variable. The Laplace transform of the residence time distribution, or the E(t)-curve, that is, the transfer function of the mixing system, G(s), is given by Eq. (1): Fig. 1. Graphical representation of the stirred tanks in parallel model.
ð5Þ
130
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
By applying the inverse Laplace transform to both sides of Eq. (5) and rearranging the terms, the residence time distribution of the tanks-in-parallel model (time dependent) was obtained. The RTD is described by Eq. (6):
3.3. Evaluation of RTD
3. Materials and methods
The hydrodynamic evaluation of the reactor was conducted without using microorganisms. To avoid changes in fluid density, the reactor was operated under the same conditions established in the biooxidation process. The pH of the reactor was controlled at 1.7 ± 0.1, with the addition of concentrated sulfuric acid or 9.0 N sodium hydroxide. The temperature in the reactor was maintained at 35 °C. The solids and the 9 K medium proposed by Silverman and Lundgren (1959) – g/L: (NH4)2SO4, 3.0; MgSO47H2O, 0.5; KH2PO4, 0.5; KCl, 0.1; Ca(NO3)2, 0.01 – were fed to maintain a pulp density of 10% w/v inside the reactor (Arroyave, 2008). Lithium chloride (LiCl) was selected as the tracer. Lithium chloride (LiCl) is frequently used as a tracer in mineral processing and hydrometallurgical operations because it is usually not present in the slurry and does not interact significantly with the ore or solution (Andrade and Hodouin, 2005; Yianatos et al., 2005). This was proved by preliminary experiments (Arroyave, 2008). Tracer was added as a liquid pulse over the liquid surface in a central point located between the impeller and the baffle (Bujalski et al., 1997). Samples in the effluent were collected immediately after tracer injection and every 12 h. The sampling frequency was always the same for calculation purposes (Levenspiel, 1981). The samples were filtered and the liquid analyzed for lithium using atomic absorption spectrometry (AAS).
3.1. Initial characterization of the mineral
3.4. Fitting of experimental RTD and the tanks-in-parallel model
A refractory gold ore from the ‘‘El Zancudo’’ mine (Titiribí – Colombia) was used in this study. The mineralogical characterization of the mineral was determined via optical microscopy using plane-polarized light and X-ray diffraction (XRD). The mineral composition was 33.23% pyrite, 26.19% arsenopyrite, 8.3% sphalerite, 6.07% galena, 4.15% chalcopyrite and 22.04% gangue. The gangue minerals present were quartz, muscovite, kaolinite and carbonates (dolomite and aragonite). The mineral’s chemical composition was determined using Atomic Absorption Spectrometry (AAS). The results were 23.3% Fe, 2.3% As, 2.3% Pb, 1.05% Zn, 0.22% Sb and 0.085% Cu. The mineral was crushed and passed through a 325 mesh screen (Arroyave et al., 2010).
The fitting of experimental results to the model was performed using the least squares method. This method allows calculating the parameter values of the model. Mathematically, this is expressed by Eq. (8):
EðTÞ ¼
ð1 fq Þ s1 ðti =s1 Þ ð1 fq Þ ti e eðti =s2 Þ eðti =s2 Þ ðs1 s2 Þ s2 ðs1 s2 Þ2 þ
fq t 2i ðti =s3 Þ e 2 s33
ð6Þ
where ti is the time at which the sample is taken from the reactor effluent during the tracer test. Eq. (6) is used to determine the distribution function of the reactor’s effluent age of the tracer’s input pulse. The age distribution curve E(t) can be expressed as a function of the dimensionless time (h = ti/tm), as given by Eq. (7) (Levenspiel, 1981; Jafari and Soltan Mohammadzadeh, 2005).
EðhÞ ¼ t m EðtÞ
ð7Þ
where tm and h are the experimental mean residence time and the dimensionless time, respectively. Eq. (7) is useful for comparing the design characteristics of two different reactors or the operating conditions of processes (Jafari and Soltan Mohammadzadeh, 2005).
3.2. Equipment and operating conditions of the reactor The study was conducted in an 8 L acrylic reactor with a working volume of 5 L, using an up-pumping axial turbine impeller with six-blades pitched at 45°. The tank was equipped with four equally spaced baffles. Impeller and baffle sizes were based on the reactor’s diameter (T = 19 cm) in order to comply with the standard design of a stirred tank reactor. The dimensions of the reactor and the impeller were H = T, B = 0.1T, C = T/3, D = 0.42T, W = D/5. Air was supplied using a sparger at the bottom of the tank. The sparger contained a small-pore-size inert membrane within it which is capable of splitting the air into small air bubbles to enhance the oxygen mass transfer in the biooxidation process (Arroyave, 2008). The agitation and aeration conditions were previously established for the biooxidation process using a 22 augmented factorial design with four center points (Arroyave, 2008). The reactor was operated at an agitation rate of 800 rpm, an aeration rate of 1.8 vvm, 10% pulp density, and a dilution rate of 0.25 d1 corresponding to a theoretical residence time of 96 h. The continuous operation of the reactor was performed using two separate inlets: the mineral through a hopper with a rake and the liquid medium (9 K without ferrous sulfate) by a peristaltic pump. The solid and liquid feed rates were controlled in order to ensure the presence of 10% w/v of solids in the feed. The slurry was withdrawn from the reactor via overflow (Arroyave, 2008).
/¼
n X ðEm ðhÞ Ee ðhÞÞ2
ð8Þ
J¼1
where Em(h) is the RTD predicted by the model and Ee(h) is the experimental RTD. 3.5. Evaluation of bacterial oxidation 3.5.1. Preparation of the inoculum for the experiments The mixture of acidophilic microorganisms used in this study was previously isolated at the ‘‘El Zancudo’’ Mine (Titiribí – Colombia). This culture was identified and showed to be compatible with A. ferrooxidans and A. thiooxidans (Ossa, 2004). The adaptation of the microorganism mixture was carried out in flasks containing 200 mL of the 9 K medium. In the culture, ferrous sulfate was replaced with mineral until a pulp density of 10% w/v was reached. The inoculum size in the biooxidation test was 10% v/v. The culture was kept within a pH of 1.5–1.9, and at a temperature of 35 °C. Agitation was set to 230 rpm using an orbital agitator. 3.5.2. Reactor operation in continuous mode The inoculum added to the biooxidation reactor was 500 mL (10% v/v). The inoculum cell concentration was 109 cells per mL, as determined by the Neubauer chamber. The process was initially operated in batch mode until the maximum concentration of ferric iron in solution was reached. Then, the reactor was switched to continuous operation. The time a continuous system takes to stabilize is called transitory state. Steady-state operation was considered to be reached when the ferric iron concentration in solution changed less than 5% during a period of time equal to the real retention time obtained with the tracer test. The pH of the reactor was controlled at 1.7 ± 0.1 and temperature was kept at 35 °C (Arroyave, 2008). Ferric iron was estimated as the difference
131
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
between total iron and ferrous iron that was determined by the ophenanthroline method (APHA, 2000). 3.5.3. Mineralogical characterization The mineralogical characterization of the biooxidized mineral was performed using X-ray diffraction (XRD). This made it possible to obtain information regarding the sulfides oxidation level, the pre-existing mineral phases and the mineral phases generated during the biooxidation process, before and after initiating the continuous mode operation. Samples were taken after 240 and 494 h of operation in the batch mode and after 622 and 766 h of operation in the continuous mode. For the XRD analyzes, a PANalytical X’pert PRO MPD diffractometer was used. In order to have a standard of comparison in the mineralogical characterization, a control experiment was also conducted without inoculating the microbial culture. This was done without changing any of the operating conditions of the biooxidation reactor.
4. Results and discussion 4.1. RTD analysis Fig. 2 shows that the maximum tracer concentration in the reactor effluent is obtained after 12 h and has a value of 2.1 mg/ L. After this value, an exponential decrease in tracer concentration can be seen. This is typically observed in completely mixed reactors. The peak at the beginning of the curve may be characterized as the presence of by-pass zones (Levenspiel, 1981; Andrade and Hodouin, 2005). The substantial decrease of the slope at the end of the curve indicates the possible presence of dead zones in the reactor. As a consequence of such a disturbance in the flow, the outflow of the tracer is slow and extends the descending branch of the concentration curve (Szekely and Themelis, 1971; Levenspiel, 1981). The mass balance of the tracer indicates that after 504 h, 98.6% of tracer had flowed out from the reactor. (These results are considered significant). The experimental mean residence time obtained from the distribution curve of tracer concentration in the reactor effluent was 130.7 h. The experimental mean residence time determined by the tracer test was 36% greater than the theoretical residence time of the liquid in the reactor. This behavior can be explained by the presence of stagnant zones that may delay the outflow of a certain amount of tracer, lengthening its residence time inside the system (Szekely and Themelis, 1971; Levenspiel, 1981; Andrade and Hodouin, 2005). Eq. (9) is the function of the dimensionless residence time distribution found for the biooxidation reactor by using the tanks-in-parallel model. This function is derived from Eq. (6) and
Fig. 2. Distribution of tracer concentration in the liquid flowing in the reactor.
Table 1 Parameters of the tanks in parallel model. fq 1fq
s1 s2 s3
0.05 0.95 0.9158tm 0.0171tm 0.0167tm
– – 119.66 h 2.23 h 2.18 h
Eq. (7). Table 1 shows the parameter values of the tanks-in-parallel model, i.e., the mean residence time for each stirred tank of Fig. 1.
EðhÞ ¼ 1:077 ½eð1:092hÞ eð58:479hÞ 61:818 h eð58:479hÞ E1 ðhÞ |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} 2 þ 5400 e60h |fflfflfflfflfflfflfflfflfflfflfflfflfflh{zfflfflfflfflfflfflfflfflfflfflfflffl ffl}E2 ðhÞ
ð9Þ Fig. 3 shows the RTD fitting from the experiments and the tanks-in-parallel model (Eq. (9)). In this model, 91.58% of the reactor volume behaves as a completely mixed tank. To gain a better understanding of the system, the two dimensionless RTD functions that compose the tanks-in-parallel model, E1(h) and E2(h) (Fig. 4) were analyzed separately. Function E1(h) is the major component of the system. It gives a completely mixed behavior composed of a large, perfectly mixed tank (about 91.58%) followed by two smaller tanks in series that correspond to 3.42%. As can be seen in Fig. 4a, this 3.42% does not contribute appreciably to the distortion of a perfectly mixed tank. Function E2(h) represents the outflow of a tracer pulse indicating the presence of a plug flow behavior within a short time during the early hours of the test. E2(h) in Fig. 4b can be interpreted physically as a by-pass zone indicating that a small fraction of lithium may have gone through the reactor much faster than the rest of the tracer. Since the fit of the early experimental data (Fig. 3) arose from the distribution function E2(h), the percentage of by-pass zones can be considered to be approximately equal to the flow fraction (5%) that goes through the upper branch of the array in parallel. The by-pass fraction in the liquid probably appears early in the process while the homogenization of the reactor is taking place (Yianatos et al., 2005). The results indicate that the model of completely mixed tanks in parallel had a good fit to experimental data. The estimated mean residence time for the model is approximately 36% greater than the theoretical residence time. This behavior can be explained because the tracer is expected to delay its outflow probably due to: gas hold up, foam formation at the top and the design of the reactor’s outlet structure (Vardar Sukan, 1998; Andrade and Hodouin, 2005). Foaming delays the outflow of the tracer because the foam can be regarded as a stagnant zone where there is not a good mixing
Fig. 3. RTD experimental and the tanks in parallel model.
132
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
Fig. 4. Functions that compose the residence time distribution of the tanks in parallel model.
with the bulk liquid that leads to an increase in the effective reactor volume and enhances the gas hold-up (Vardar Sukan, 1998). Aeration also modifies the flow patterns established by the impeller (Nienow and Bujalski, 1997; Kasat and Pandit, 2005) which may contribute to the formation of foam (Vardar Sukan, 1998). As the gas flow rate rises the height of the foam layer increases because more bubbles reach the surface being converted into foam (Vardar Sukan, 1998). The liquid inside the reactor flowed out of it through a plastic hose when there was overflow. Slurry drops fall off the hose when the surface tension forces can no longer counterweight the slurry outside the hose. This causes the liquid to flow out when the pressure inside is greater than the surface tension in the form of pulses. It is important to note that the outflow due to overflow depends on the random motion and the liquid level inside the reactor. 4.2. Characterization of the biooxidation process through XRD Fig. 5 shows the variation in the concentration of ferric and ferrous iron in solution and the diffractograms obtained during the biooxidation process in batch and continuous mode. The concentration of ferrous iron in solution decreased while the ferric ion concentration increased over time within the batch mode (Fig. 5a). This behavior was interpreted as evidence that the main catalytic mechanism of bacteria involves the oxidation of Fe2+ to Fe+3 in accordance with the findings of Williamson and Rimstidt (1994), Rawlings et al. (2003), Rodríguez et al. (2003); Sand and Gehrke (2006). During the time lapse between hours 336 and 477, the concentration of ferric iron in solution had a tendency to decrease possibly due to jarosite precipitation. The XRD analysis showed that the greatest formation of jarosite occurred approximately after 477 h in the batch mode (Fig. 5c). The total iron in solution after 477 h was equivalent to 70% of the oxidation (theoretical value estimated stoichiometrically). The steady-state operation was considered to be
reached after 634.6 h since the difference in ferric iron concentration in solution was approximately 5%. During the steady-state, an average concentration of ferric iron in solution of 8.3 g/L was reached corresponding to an oxidation of 33% of total iron (this is an underestimated value because the precipitated iron in the oxidized mineral was not measured). The XRD results show that the oxidation rate of the arsenopyrite was greater than that of pyrite in both of the biooxidation reactor’s operating modes (Fig. 5). Advanced and nearly complete oxidation of the arsenopyrite and pyrite occurred after 477 h of operation in the batch mode. This could be due to fact that the mineral was in contact with the oxidizing medium for a longer time. Furthermore, high ratios of Fe3+/Fe2+ were found in solution compared to the samples obtained after 240 h of operation in batch mode, and 622 h and 722 h in continuous mode (steady state). In both modes of operation, arsenopyrite oxidation was higher than pyrite oxidation. This is possibly due to the fact that arsenopyrite oxidation is produced by the combined attack of ferric iron and protons (polysulfide mechanism). In contrast, pyrite oxidation is the result of the attack of ferric iron (thiosulfate mechanism) (Tributsch, 2001; Schippers and Sand, 1999; Fowler et al., 2001; Rawlings et al., 2003). The pH in the reactor was kept at 1.7 ± 0.1, thus maintaining the supply of protons. This is likely to have favored arsenopyrite oxidation in the continuous and batch modes. In the case of pyrite, a longer contact time with ferric iron in batch mode possibly caused higher oxidation after 477 h. It has been reported that arsenopyrite oxidizes faster than pyrite (Márquez et al., 2006; Ciftci and Akcil, 2010 and Ossa and Márquez, 2010). The cause could be its lower oxidation potential (Rimstidt et al., 1994). The XRD analyzes showed the formation of jarosite in both operation modes. The highest formation of jarosite occurred after 477 h of operation in the batch mode. This could be due to the high concentration of ferric iron in solution (Fig. 5a and c) according to what was reported by Ciftci and Akcil (2010). The decrease in the output of jarosite within the continuous mode is caused by the outflow of ferric iron and some jarosite until the system reaches a steady-state with a lower average concentration of ferric iron in solution (Fig. 5a). The formation rate of jarosite in the process was not as high as the rate observed by Zapata et al. (2004); Ossa (2004); Ossa and Márquez (2010). It might be a result of the applied control over pH (1.7 ± 0.1) during the biooxidation process. Daoud and Karamanev (2006) found that the principal parameter affecting jarosite formation is pH. A low precipitation of jarosite was obtained in a culture of A. ferrooxidans with a pH of 1.6–1.7 under a temperature of 35 °C. When the pH values are above 2.5, ferric iron has low solubility that causes the formation of basic Fe(III) hydroxyl sulfates (Gomez and Cantero, 2005; Daoud and Karamanev, 2006). There was no formation of jarosite in the control experiment because there was neither oxidation of sulfides nor increased ferric iron in solution. Jarosite is a common byproduct of the biooxidation of sulfides containing iron (García et al., 1995; Márquez et al., 2006). The final byproduct of the reaction of pyrite and ferric iron is sulfate (Schippers and Sand, 1999). In arsenopyrite oxidation, it is possible that polysulfide and elemental sulfur are generated as intermediate byproducts (Schippers and Sand, 1999). This elemental sulfur is relatively stable because it can be oxidized into sulfate through the action of sulfur-oxidizing microorganisms (Suzuki, 2001); in this case, A. thiooxidans was one of the organisms present in the culture. Furthermore, it has been proposed that jarosite is an abundant and common mineral phase in the biooxidation process of arsenopyrite (Márquez et al., 2006; Ossa and Márquez, 2010; Jiang et al., 2008; Márquez et al., 2012). According to Ciftci and Akcil (2010) the increased sulfate in the residual matter resulting from the biooxidation of a refractory gold concentrate is probably due to the precipitation of jarosite.
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
133
Fig. 5. Variation of ferric and ferrous iron concentration in solution and diffractograms of the biooxidation process in the batch and continuous modes (p: pyrite, a: arsenopyrite. j: jarosite, b: brushite, q: quarz, c: clinozoisite).
The set of data from the XRD show a high presence of silicates in the biooxidation process as a result of their exhaustive dissolution resulting from the controlled pH conditions (1.7 ± 0.1) (Fig. 5). The predominant type of silicate was quartz, and its higher proportion was observed after 477 h of operation in batch mode. This was probably due to the increased contact time with the oxidizing medium and to the fact that the biooxidation of concentrates containing no sulfides (such as oxides, carbonates and silicates) occurs simply through an acid attack (Sand and Gehrke, 2006). Fig. 5 shows the presence of brushite (CaHPO42(H2O)) in the oxidized samples. There was formation of brushite in the control sample (data is not shown in Fig. 5f) The formation of brushite could have been due to the reaction of the HPO 4 in the 9 K medium and the carbonates in the ore. The highest formation of
brushite was observed in the continuous mode. This might be explained by the permanent inflow and outflow of mineral and 9 K medium. RTD results from the liquid phase of the reactor has indicated there is a 36% of dead zones meaning that in the continuous operation mode of the reactor a 36% of a delayed outflow of ferric iron in solution is probable. This outcome could increase the possibility of jarosite formation. XRD analyses show the presence of jarosite in the continuous mode for concentrations of ferric iron in solution during the steady-state approximately 8.3 g/L and pH 1.7 ± 0.1 on average. On the other hand, the delayed outflow of the tracer decreases both the dilution rate and the specific growth rate of the microorganisms, thus affecting sulfide oxidation in the continuous operation mode (Gomez and Cantero, 2003).
134
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
It may be deduced from the outcome of this study that jarosite precipitation could be related to dead zones since the delay in the outflow of the tracer indicates that the concentration of ferric iron in solution is likely to delay its outflow by approximately 36%, hence increasing the probability of iron precipitation inside the reactor. Further research is required in order to identify the effects of hydrodynamic behavior in the formation of precipitates in biooxidation processes, since these are carried out using large commercial-scale continuous stirred tank reactors that usually exhibit high concentrations of ferric iron during the steady state (Rawlings et al., 2003; Arrascue and Van Niekerk, 2006; Van Niekerk, 2009). Moreover, during the operation they may exhibit dead zones and by-passes that affect process performance. In this regard, jarosite precipitation in the presence of different types of microorganisms and sulfide concentrates should be assessed. Likewise, the assessment must be done for several design and operating conditions (aeration, agitation, pulp density, temperature and different impellers) that allow the variation of the by-passes and dead zones percentages inside the reactor. The hydrodynamic evaluation of the biooxidation reactor of refractory gold concentrate was carried out without the use of microorganisms. Cell growth could have affected foaming, since the dissolution of sulfides may affect the physicochemical properties of the medium and therefore the surface phenomena. According to Vardar (1998) foaming is affected by gas flow, agitation, temperature, pH, nature and composition of the medium, presence of cell growth and by the nature of the metabolites and substances produced. Foaming affects the effective volume of the liquid inside the reactor and consequently its RTD (Vardar, 1998). Likewise, sulfide concentrates other than refractory gold may affect the physicochemical properties of the medium and therefore of foaming. Considering that microorganisms are responsible for providing sulfuric acid for a proton attack and for maintaining high concentrations of ferric iron for an oxidative attack on the mineral (Rawlings et al., 2003; Sand and Gehrke, 2006), future studies should evaluate jarosite precipitation in the presence of mesophiles, moderate thermophiles and extreme thermophiles in CSTR. Ciftci and Akcil (2010) evaluated the oxidation ability of a mixed microbial culture of mesophiles, moderate thermophiles and extreme thermophiles during the biooxidation of a refractory gold concentrate at different pulp densities in flasks. These authors also evaluated the effect of biooxidation residues on sodium cyanide consumption. They found that sodium cyanide consumption was lower in cultures with extreme thermophiles compared to the cultures with mesophiles and moderate thermophiles due to the lower amount of jarosite precipitation. The hydrodynamic behavior of the reactor depends directly on the impeller. The formation of precipitates should therefore be assessed with different impellers such as axial flow hydrofoil impellers, e.g. Lightnin A315 impellers, which are used in biooxidation plants (Dew et al., 1997; Arrascue and Van Niekerk, 2006; Batty and Rorke, 2006; Van Niekerk, 2009). RTD and XRD analyzes from a laboratory-scale biooxidation reactor make it possible to obtain observations that can be used for the design and scale-up of reactors in order to improve their operation and performance.
5. Conclusions Hydrodynamic evaluation showed that the reactor had a high tendency to behave as a perfectly mixed reactor. However, the mixed flow inside the reactor had disturbances such as by-pass and dead zones that make the process far from ideal. It is highly probable that the outflow of the tracer was delayed. This was caused mainly by gas hold-up, foaming at the top and the design
of the reactor output structure itself. The tanks-in-parallel model fit well the experimental residence time distribution. This indicated that the estimated mean residence time for the model was approximately 36% greater than the theoretical residence time this was due to the delay in the outflow of the tracer. Additionally, 91.58% of the reactor’s volume behaved as a completely mixed tank and 5% of it had by-pass zones. The XRD outcomes showed a greater oxidation of arsenopyrite than pyrite besides the formation of jarosite and brushite. Although the presence of jarosite in continuous mode is lower than in batch mode, the dead zones in the continuous operation of the reactor may increase the possibility of jarosite precipitation.
Acknowledgements The authors would like to thank the Biotechnology National Program of COLCIENCIAS, CDI S.A. Mines Company and the Biomineralogy Laboratory from the National University of Colombia (Medellin campus)
References Akcil, A., 2004. Potential bioleaching developments towards commercial reality: Turkish metal mining’s future. Minerals Engineering 17 (3), 477–480. Andrade, L., Hodouin, D., 2005. Residence time distribution of an industrial mechanically agitated cyanidation tank. Minerals Engineering 18 (6), 613–621. APHA, 2000. American public health association. standard methods for the examination of water and wastewater. In: Clescerl, L.S., Greenberg, A.E., Eaton, A.D. (Eds.), American Water Works Association & Water Environment Federation, 20th ed. American Public Health Association, Washington, DC. Arrascue, M.E.L., Van Niekerk, J., 2006. Biooxidation of arsenopyrite concentrate using BIOXÒ process: industrial experience in Tamboraque, Peru. Hydrometallurgy 83 (1–4), 90–96. Arroyave, D. M., 2008. Evaluation of the biooxidation process to laboratory scale of the auriferous mineral of the ‘‘El Zancudo’’ Mine, Titiribi, Antioquia. M.Sc. Thesis, School of Materials, National University of Colombia (Medellin). Arroyave, D., Marquez, M., Gallego, D., Pacheco, G., 2010. Evaluation and mineralogical characterization of biooxidation process in a continuous stirred tank reactor. Dyna 164, 18–29. Batty, J.D., Rorke, G.V., 2006. Development and commercial demonstration of the BioCOP™ thermophile process. Hydrometallurgy 83 (1–4), 83–89. Brochot, S., Durance, M.V., Villeneuve, J., d’Hugues, P., Mugabi, M., 2004. Modelling of the bioleaching of sulfide ores: application for the simulation of the bioleaching/gravity section of the Kasese Cobalt Company Ltd. process plant. Minerals Engineering 17 (2), 253–260. Bryan, C.G., Joulian, C., Spolaore El Achbouni, P.H., Challan-Belval, S., Morin, D., d’Hugues, P., 2011. The efficiency of indigenous and designed consortia in bioleaching stirred tank reactors. Minerals Engineering 24 (11), 1149–1156. Bujalski, J.M, Jaworski, Z, Bujalski, W., Nienow, A.W., 1997. The influence of the addition position of a tracer on CFD simulated mixing times in a vessel agitated by a Rushton turbine. In: Engineering Foundation Conference on Biochemical Engineering X.
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135 García Jr., O., Bigham, J.M., Touvinen, O.H., 1995. Sphalerite oxidation by Thiobacillus ferrooxidans and Thiobacillus thiooxidans. Canadian Journal of Microbiology 41, 578–584. Gomez, J.M., Cantero, D., 2003. Kinetic study of biological ferrous sulphate oxidation by iron – oxidising bacteria in continues stirred tank and packed bed bioreactors. Process Biochemistry 38 (6), 867–875. Gómez, J. M., y Cantero, D., 2005. Biooxidación del ión ferroso. In: Fernando Acevedo y Juan Carlos Gentina (Ed.). Fundamentos y Perspectivas de las Tecnologías Biomineras. pp. 25–43. Gonzalez, R., Gentina, J., Acevedo, F., 2003. Optimisation of the solids suspension conditions in a continuous stirred tank reactor for the biooxidation of refractory gold concentrates. Electronic Journal of Biotechnology 6 (3), 233–240. Hayward, T., Satalic, D.M., Spencer, P.A., 1997. Engineering, equipment and materials: developments in the design of a bacterial oxidation reactor. Minerals Engineering 10 (10), 1047–1055. Jafari, M., Soltan Mohammadzadeh, J.S., 2005. Mixing time, homogenization energy and residence time distribution in a gas-induced contactor. Trans IChemE, Part A, Chemical Engineering Research and Design 83 (A5), 452–459. Jiang, T., Li, Q., Yang, Y.B., Li, G.H., Qiu, G.Z., 2008. Bio-oxidation of arsenopyrite. Transactions of Nonferrous Metals Society of China 18, 1433–1438. Kasat, G.R., Pandit, A.B., 2005. Review on mixing characteristics in solid–liquid and solid–liquid–gas reactor vessels. The Canadian Journal of Chemical Engineering 83 (4), 618–643. Levenspiel, O., 1981. Chemical Reaction Engineering, 2 ed. Reverté S.A, Barcelona. Liu, G., Yin, J., Cong, W., 2007. Effect of fluid shear and particles collision on the oxidation of ferrous iron by Acidithiobacillus ferrooxidans. Minerals Engineering 20 (13), 1227–1231. Márquez, M.G., Gaspar, J., Bessler, K., Magela, G., 2006. Process mineralogy of bacterial oxidized gold ore in Sao Bento mine (Brazil). Hydrometallurgy 83 (1–4), 114–123. Márquez, M.A., Ospina, J.D., Morales, A.L., 2012. New insights about the bacterial oxidation of arsenopyrite: a mineralogical scope. Minerals Engineering 39, 248– 254. Mazuelos, A., Romero, R., Palencia, I., Carranza, F., Borjas, F.J., 2002. Oxygen transfer in ferric iron biological production in a packed-bed reactor. Hydrometallurgy 65 (1), 15–22. Nienow, A. W., Bujalski, W., 1997. Recent studies on agitated three phase (gas– solid–liquid) systems in the turbulent regime. In: Engineering Foundation Conference on Biochemical Engineering X.
135
Rawlings, D.E., Dew, D., du Plessis, C., 2003. Biomineralization of metal containing ores and concentrates. TRENDS in Biotechnology 21 (1), 38–44. Rimstidt, J., Chermak, J.A., Cagen, P.M., 1994. Rates of reaction of galena, sphalerite, chalcopyrite, and arsenopyrite with Fe(III) in acidic solutions. In: Alpers, C.N., Blowes, D.W. (Eds.), Environmental Geochemistry of Sulfide Oxidation, vol. 550.. ACS Symposium Series. American Chemical Society, Washington, D.C, pp. 1–13. Rodríguez, Y., Ballester, A., Blázquez, M.L., González, F., Muñoz, J.A., 2003. New information on the pyrite bioleaching mechanism at low and high temperature. Hydrometallurgy 71 (1–2), 37–46. Romero, R., Palencia, I., Carranza, F., 1998. Silver catalyzed IBES process: application to a Spanish copper–zinc sulfide concentrate Part 3. Selection of the operational parameters for a continuous pilot plant. Hydrometallurgy 49 (1–2), 75–86. Rossi G. 1990. Biohydrometallurgy. McGraw – Hill Book Company Gmbh. Sand, W., Gehrke, T., 2006. Extracellular polymeric substances mediate bioleaching/ biocorrosion via interfacial processes involving iron(III) ions and acidophilic bacteria. Research in Microbiology 157 (1), 49–56. Schippers, A., Sand, W., 1999. Bacterial leaching of metal sulfides proceeds by two indirect mechanisms via thiosulfate or via polysulfides and sulfur. Applied and Environmental Microbiology 65 (1), 319–321. Silverman, M.P., Lundgren, D.G., 1959. Studies on the chemolithotrophic iron bacterium Ferrobacillus ferrooxidans an improved medium and a harvesting procedure for securing high cell yields. Journal of Bacteriology 77, 642–647. Sun, Li-Xin, Zhang, Xu, Tan, Wen-Song, Zhu, Ming-Long, 2012. Effect of agitation intensity on the biooxidation process of refractory gold ores by Acidithiobacillus ferrooxidans. Hydrometallurgy 127–128, 99–103. Suzuki, Isamu, 2001. Microbial leaching of metals from sulfide minerals. Biotechnology Advances 19 (2), 119–132. Szekely, J., Themelis, N., 1971. Rate Phenomena in Process Metallurgy. WileyInterscience, a Division of John Wiley & Sons, Inc.. Tributsch, Helmut, 2001. Direct versus indirect bioleaching. Hydrometallurgy 59 (2–3), 177–185. Van Aswegen, Pieter C., Van Niekerk, Jan, 2007. The BIOXTM process for the treatment of refractory gold concentrates. In: Rawlings, Douglas E., Johnson, B.D. (Eds.), Biomining. Springer Verlag, Berlin, Germany, pp. 1–32. Van Niekerk, J., 2009. Recent advances in BIOXÒ technology. In: Hydrometallurgy Conference 2009. The Southern African Institute of Mining and Metallurgy. Vardar, Sukan Fazilet, 1998. Foaming: consequences, prevention and destruction. Biotechnology Advances 16 (5–6), 913–948. Watling, H.R., 2008. The bioleaching of nickel–copper sulfides. Hydrometallurgy 91, 70–88. Williamson, M.A., Rimstidt, J.D., 1994. The kinetics and electrochemical ratedetermining step of aqueous pyrite oxidation. Geochimica Cosmochimica Acta 58 (24), 5443–5454. Yianatos, J.B., Bergha, L.G., Díaz, F., Rodríguez, J., 2005. Mixing characteristics of industrial flotation equipment. Chem Engineering Science 60 (8–9), 2273–2282. Zapata, D., Gallego, A., Marquez M. A., 2004. Definition of the type of refractoriness in El Zancudo Golg Mine, Antioquia, Colombia. Appl. In: Miner., Pecchio et al. (Eds.).
Contents lists available at SciVerse ScienceDirect
Minerals Engineering journal homepage: www.elsevier.com/locate/mineng
Evaluation of residence time distribution and mineralogical characterization of the biooxidation of sulfide minerals in a continuous stirred tank reactor Diana M. Arroyave G. a,⇑, Darío Gallego S. b, Marco A. Márquez G. a a b
Applied Mineralogy and Bioprocess Research Group, School of Materials, National University of Colombia, Medellin, Colombia Sanitary Engineering Laboratory, School of Processes and Energy, National University of Colombia, Medellin, Colombia
a r t i c l e
i n f o
Article history: Received 10 April 2012 Accepted 13 March 2013 Available online 3 May 2013 Keywords: RTD Biooxidation Tracer Jarosite XRD CSTR
a b s t r a c t The residence time distribution (RTD) of the liquid phase and the mineralogical characterization of the biooxidation of refractory gold mineral was studied in a continuous stirred tank reactor. The latter was achieved using a native mixed culture of acidophilic mesophiles. The RTD was assessed using a mathematical model of stirred tanks in parallel. The oxidation of sulfide and the phases generated through the biooxidation process were evaluated via X-ray diffraction (XRD).The results indicated that the experimental RTD fit to the model. The reactor has a high tendency to behave as a completely mixed reactor. However, the mixed flow inside the reactor has disturbances such as by-pass and dead zones. The estimated mean residence time for the model was approximately 36% greater than the theoretical residence time. It was probably caused by the delay in the outflow of the tracer due to gas hold-up, foaming at the top and the design of the reactor outlet structure. The XRD outcomes showed that the oxidation of arsenopyrite was greater than that of pyrite. Similarly, the formation of jarosite and brushite was observed. It was concluded that the dead zones could increase the probability of jarosite precipitations. Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction The biooxidation of sulfide minerals in continuous stirred tank reactors (CSTR) has become an important method for the pretreatment of refractory gold ore (Chandraprabha et al., 2002; Akcil, 2004; Watling, 2008). Industrial-scale biooxidation was first carried out in South Africa (Fairview) in 1986 as pretreatment for refractory gold concentrates. Currently, CSTR biooxidation is being used successfully in commercial operation in countries such Brazil, Peru, Australia, Ghana, South Africa, India and China (Gonzalez et al., 2003; Akcil, 2004; Sand and Gehrke, 2006; Watling, 2008). The growth of this technology is promising, and new projects are currently in the research and engineering stages (Van Niekerk, 2009). Biooxidation is catalyzed by bacteria that oxidize reduced iron and sulfur compounds. In industrial applications of mining and metallurgy, cultures of mesophiles or moderate thermophiles are widely used for the oxidation of sulfide ores (Rossi, 1990; Rawlings et al., 2003; Akcil, 2004; Sand and Gehrke, 2006). Such microorganisms include the iron- and sulfur-oxidizing Acidithiobacillus ferrooxidans, the sulfur-oxidizing Acidithiobacillus thiooxidans and Acidithiobacillus caldus, and the iron-oxidizing Leptospirillum ferrooxidans and Leptospirillum ferriphilum. (Coram and Rawlings, 2002; Rawlings et al., 2003; Bryan et al., 2011). The operations tend to be ⇑ Corresponding author. Tel.: +57 (4) 4255341/4255342. E-mail address: [email protected] (D.M. Arroyave G.). 0892-6875/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.mineng.2013.03.012
dominated by L. ferriphilum and A. caldus at 40–45 °C (Rawlings, 2007). The success of bacterial sulfide oxidation in a CSTR requires a mixing that reaches a degree of uniformity that maintains suitable conditions for bacterial growth and solids suspension, thus minimizing the development of concentration, temperature and pH gradients within the reactor (Rossi, 1990; Hayward et al., 1997; González et al., 2003; Deveci, 2004). Currently, the research and development areas related to biooxidation technology include the development of an improved agitation system, and the reduction of retention time (Van Niekerk, 2009). The scientific literature provides a number of studies on different impellers and their effects on bacterial activity in CSTR biooxidation (d’Hugues et al., 1997; Deveci, 2002, 2004; Liu et al., 2007; Sun et al., 2012). It has been demonstrated that within in the biooxidation process the best alternative is the use of axial flow impellers (Dew et al., 1997; Deveci, 2002, 2004; Gonzalez et al., 2003; Van Aswegen et al., 2007). Indeed, nowadays all commercial stirred-tank biooxidation plants use axial flow hydrofoil impellers (Dew et al., 1997; Arrascue and Van Niekerk, 2006; Batty and Rorke, 2006; Van Aswegen et al., 2007; Van Niekerk, 2009). Hydrodynamic behavior studies in biooxidation reactors are limited in the literature. The mathematical description of residence time distribution (RTD) is usually expressed either through the dispersion model or through the tanks-in-series model. Romero et al. (1998) determined the RTD for two stirred tank reactors in series at the Rio Tinto plant in Spain. The RTD curve that best fit the experimental data was the two completely stirred tanks-in-series model.
129
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
Nomenclature Ci E(t) E(h) Em(h) Ee(h) G(s) ti (h)tm s
tracer concentration in the effluent at time ti (mg/L) Residence time distribution function Dimensionless residence time distribution function residence time distribution of the model experimental residence time distribution transfer function of the stirred tanks time at which the sample is taken from the reactor effluent during the tracer test experimental mean residence time (h) complex variable
Mazuelos et al. (2002) reported the assessment of RTD at various bed heights of a packed bed reactor using the tanks-in-series model. The results indicated that for the maximum bed height of 84 cm, the hydrodynamic behavior fits two completely mixed tanks in series. A hydrodynamic assessment of a reactor conducted by analyzing the residence time distribution (RTD) is crucial to determine possible flow abnormalities such as by-pass and dead zones which may affect the process operation and performance (Levenspiel, 1981; Andrade and Hodouin, 2005). Additionally, RTD can be used together with biooxidation kinetic models to simulate these processes (Brochot et al., 2004). The previous statement shows that characterizing the hydrodynamic behavior in biooxidation reactors is relevant. Mixing conditions in stirred tank reactors for the biooxidation of gold concentrates depend on many factors including pulp density, aeration and agitation rate, reactor design features such as tank geometry, impeller type and diameter, and air injection type and location (Gonzalez et al., 2003; Deveci, 2004). It is therefore necessary to determine the RTD for each individual case and thus guarantee a good process performance. In a reactor, hydrodynamic evaluation is performed using a tracer test: a tracer is injected into the reactor’s feed and the response to this stimulus is recorded. The response analysis provides information about the mixing properties of the system and the performance of non-ideal chemical reactors (Levenspiel, 1981). The aim of this paper was to determine the RTD of the liquid phase of a continuous stirred tank reactor used for biooxidizing sulfides at laboratory scale. This was done in order to characterize the mixing conditions established by the operating conditions and reactor design. The RTD was analyzed through a developed mathematical model of completely stirred tanks in parallel. After determining the RTD the bacterial oxidation of sulfides was evaluated and a mineralogical characterization through X-ray diffraction (XRD) was performed to determine the level of sulfide oxidation and the phases generated in the biooxidation process, respectively.
s1 s2
mean residence time in the large stirred tank (h) mean residence time in each small stirred tank in series (h) mean residence time of the three agitated tanks in series (h) flow fraction of the upper path flow fraction of the lower path dimensionless time (ti/tm)
s3 fq f1-q h
GðsÞ ¼
Z
1
EðtÞ est dt
ð1Þ
0
where s is a complex variable. The response of any array of ideal reactors can be derived from the standard complex algebraic domain rules, and then converted into its time domain expression (Levenspiel, 1981). The solution of Eq. (1) for each ideal continuous stirred tank in Fig. 1 is represented by a transfer function in the Laplace domain in terms of mean residence time of the liquid inside the reactor (s) by Eq. (2) (Yianatos et al., 2005; Andrade and Hodouin, 2005):
GðsÞ ¼
1
ð2Þ
ss þ 1
The overall transfer function for the arrangement of tanks in parallel in Fig. 1 is given by Eq. (3):
GðsÞ ¼ ð1 fq Þ G1 ðsÞ G2 ðsÞ2 þ fq G3 ðsÞ3
ð3Þ
where fq is the flow fraction circulating through the upper branch, f1q is the fraction of the flow circulating through the lower branch, G1(s) is the transfer function of the large stirred tank, G2(s) is the transfer function of each ideal small stirred tank in series with the large tank, and G3(s) is the transfer function of each of the three stirred tanks in series in the upper branch of the array in parallel. By substituting Eq. (2) in each stirred tank in Eq. (3), the transfer function in terms of the parameters of the tanks-in-parallel is obtained as follows (Eq. (4)):
GðsÞ ¼
ð1 fq Þ ðs1 s þ 1Þ ðs2 s þ 1Þ
2
þ
fq ðs3 s þ 1Þ3
ð4Þ
If Eq. (4) is decomposed into partial fractions, Eq. (5) is obtained.
GðsÞ ¼
ð1 fq Þ s1
"
ðs1 s2 Þ2 " # 1 1
ðs þ s2 Þ
2
# 1 1 ð1 fq Þ ðs1 s2 Þ s2 ðs þ s11 Þ ðs þ s12 Þ
þ
fq
s ðs þ s13 Þ3 3 3
2. Stirred tanks in parallel model This model consists of three small perfectly mixed reactors in parallel with a combination of a large perfectly mixed reactor in series with two smaller reactors (Fig. 1). A very useful concept for modeling the RTD of real reactors is the transformation of RTD into the Laplace domain (Pinheiro and Oliveirah, 1998; Yianatos et al., 2005), because the Laplace transform is used to solve problems where time is the independent variable. The Laplace transform of the residence time distribution, or the E(t)-curve, that is, the transfer function of the mixing system, G(s), is given by Eq. (1): Fig. 1. Graphical representation of the stirred tanks in parallel model.
ð5Þ
130
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
By applying the inverse Laplace transform to both sides of Eq. (5) and rearranging the terms, the residence time distribution of the tanks-in-parallel model (time dependent) was obtained. The RTD is described by Eq. (6):
3.3. Evaluation of RTD
3. Materials and methods
The hydrodynamic evaluation of the reactor was conducted without using microorganisms. To avoid changes in fluid density, the reactor was operated under the same conditions established in the biooxidation process. The pH of the reactor was controlled at 1.7 ± 0.1, with the addition of concentrated sulfuric acid or 9.0 N sodium hydroxide. The temperature in the reactor was maintained at 35 °C. The solids and the 9 K medium proposed by Silverman and Lundgren (1959) – g/L: (NH4)2SO4, 3.0; MgSO47H2O, 0.5; KH2PO4, 0.5; KCl, 0.1; Ca(NO3)2, 0.01 – were fed to maintain a pulp density of 10% w/v inside the reactor (Arroyave, 2008). Lithium chloride (LiCl) was selected as the tracer. Lithium chloride (LiCl) is frequently used as a tracer in mineral processing and hydrometallurgical operations because it is usually not present in the slurry and does not interact significantly with the ore or solution (Andrade and Hodouin, 2005; Yianatos et al., 2005). This was proved by preliminary experiments (Arroyave, 2008). Tracer was added as a liquid pulse over the liquid surface in a central point located between the impeller and the baffle (Bujalski et al., 1997). Samples in the effluent were collected immediately after tracer injection and every 12 h. The sampling frequency was always the same for calculation purposes (Levenspiel, 1981). The samples were filtered and the liquid analyzed for lithium using atomic absorption spectrometry (AAS).
3.1. Initial characterization of the mineral
3.4. Fitting of experimental RTD and the tanks-in-parallel model
A refractory gold ore from the ‘‘El Zancudo’’ mine (Titiribí – Colombia) was used in this study. The mineralogical characterization of the mineral was determined via optical microscopy using plane-polarized light and X-ray diffraction (XRD). The mineral composition was 33.23% pyrite, 26.19% arsenopyrite, 8.3% sphalerite, 6.07% galena, 4.15% chalcopyrite and 22.04% gangue. The gangue minerals present were quartz, muscovite, kaolinite and carbonates (dolomite and aragonite). The mineral’s chemical composition was determined using Atomic Absorption Spectrometry (AAS). The results were 23.3% Fe, 2.3% As, 2.3% Pb, 1.05% Zn, 0.22% Sb and 0.085% Cu. The mineral was crushed and passed through a 325 mesh screen (Arroyave et al., 2010).
The fitting of experimental results to the model was performed using the least squares method. This method allows calculating the parameter values of the model. Mathematically, this is expressed by Eq. (8):
EðTÞ ¼
ð1 fq Þ s1 ðti =s1 Þ ð1 fq Þ ti e eðti =s2 Þ eðti =s2 Þ ðs1 s2 Þ s2 ðs1 s2 Þ2 þ
fq t 2i ðti =s3 Þ e 2 s33
ð6Þ
where ti is the time at which the sample is taken from the reactor effluent during the tracer test. Eq. (6) is used to determine the distribution function of the reactor’s effluent age of the tracer’s input pulse. The age distribution curve E(t) can be expressed as a function of the dimensionless time (h = ti/tm), as given by Eq. (7) (Levenspiel, 1981; Jafari and Soltan Mohammadzadeh, 2005).
EðhÞ ¼ t m EðtÞ
ð7Þ
where tm and h are the experimental mean residence time and the dimensionless time, respectively. Eq. (7) is useful for comparing the design characteristics of two different reactors or the operating conditions of processes (Jafari and Soltan Mohammadzadeh, 2005).
3.2. Equipment and operating conditions of the reactor The study was conducted in an 8 L acrylic reactor with a working volume of 5 L, using an up-pumping axial turbine impeller with six-blades pitched at 45°. The tank was equipped with four equally spaced baffles. Impeller and baffle sizes were based on the reactor’s diameter (T = 19 cm) in order to comply with the standard design of a stirred tank reactor. The dimensions of the reactor and the impeller were H = T, B = 0.1T, C = T/3, D = 0.42T, W = D/5. Air was supplied using a sparger at the bottom of the tank. The sparger contained a small-pore-size inert membrane within it which is capable of splitting the air into small air bubbles to enhance the oxygen mass transfer in the biooxidation process (Arroyave, 2008). The agitation and aeration conditions were previously established for the biooxidation process using a 22 augmented factorial design with four center points (Arroyave, 2008). The reactor was operated at an agitation rate of 800 rpm, an aeration rate of 1.8 vvm, 10% pulp density, and a dilution rate of 0.25 d1 corresponding to a theoretical residence time of 96 h. The continuous operation of the reactor was performed using two separate inlets: the mineral through a hopper with a rake and the liquid medium (9 K without ferrous sulfate) by a peristaltic pump. The solid and liquid feed rates were controlled in order to ensure the presence of 10% w/v of solids in the feed. The slurry was withdrawn from the reactor via overflow (Arroyave, 2008).
/¼
n X ðEm ðhÞ Ee ðhÞÞ2
ð8Þ
J¼1
where Em(h) is the RTD predicted by the model and Ee(h) is the experimental RTD. 3.5. Evaluation of bacterial oxidation 3.5.1. Preparation of the inoculum for the experiments The mixture of acidophilic microorganisms used in this study was previously isolated at the ‘‘El Zancudo’’ Mine (Titiribí – Colombia). This culture was identified and showed to be compatible with A. ferrooxidans and A. thiooxidans (Ossa, 2004). The adaptation of the microorganism mixture was carried out in flasks containing 200 mL of the 9 K medium. In the culture, ferrous sulfate was replaced with mineral until a pulp density of 10% w/v was reached. The inoculum size in the biooxidation test was 10% v/v. The culture was kept within a pH of 1.5–1.9, and at a temperature of 35 °C. Agitation was set to 230 rpm using an orbital agitator. 3.5.2. Reactor operation in continuous mode The inoculum added to the biooxidation reactor was 500 mL (10% v/v). The inoculum cell concentration was 109 cells per mL, as determined by the Neubauer chamber. The process was initially operated in batch mode until the maximum concentration of ferric iron in solution was reached. Then, the reactor was switched to continuous operation. The time a continuous system takes to stabilize is called transitory state. Steady-state operation was considered to be reached when the ferric iron concentration in solution changed less than 5% during a period of time equal to the real retention time obtained with the tracer test. The pH of the reactor was controlled at 1.7 ± 0.1 and temperature was kept at 35 °C (Arroyave, 2008). Ferric iron was estimated as the difference
131
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
between total iron and ferrous iron that was determined by the ophenanthroline method (APHA, 2000). 3.5.3. Mineralogical characterization The mineralogical characterization of the biooxidized mineral was performed using X-ray diffraction (XRD). This made it possible to obtain information regarding the sulfides oxidation level, the pre-existing mineral phases and the mineral phases generated during the biooxidation process, before and after initiating the continuous mode operation. Samples were taken after 240 and 494 h of operation in the batch mode and after 622 and 766 h of operation in the continuous mode. For the XRD analyzes, a PANalytical X’pert PRO MPD diffractometer was used. In order to have a standard of comparison in the mineralogical characterization, a control experiment was also conducted without inoculating the microbial culture. This was done without changing any of the operating conditions of the biooxidation reactor.
4. Results and discussion 4.1. RTD analysis Fig. 2 shows that the maximum tracer concentration in the reactor effluent is obtained after 12 h and has a value of 2.1 mg/ L. After this value, an exponential decrease in tracer concentration can be seen. This is typically observed in completely mixed reactors. The peak at the beginning of the curve may be characterized as the presence of by-pass zones (Levenspiel, 1981; Andrade and Hodouin, 2005). The substantial decrease of the slope at the end of the curve indicates the possible presence of dead zones in the reactor. As a consequence of such a disturbance in the flow, the outflow of the tracer is slow and extends the descending branch of the concentration curve (Szekely and Themelis, 1971; Levenspiel, 1981). The mass balance of the tracer indicates that after 504 h, 98.6% of tracer had flowed out from the reactor. (These results are considered significant). The experimental mean residence time obtained from the distribution curve of tracer concentration in the reactor effluent was 130.7 h. The experimental mean residence time determined by the tracer test was 36% greater than the theoretical residence time of the liquid in the reactor. This behavior can be explained by the presence of stagnant zones that may delay the outflow of a certain amount of tracer, lengthening its residence time inside the system (Szekely and Themelis, 1971; Levenspiel, 1981; Andrade and Hodouin, 2005). Eq. (9) is the function of the dimensionless residence time distribution found for the biooxidation reactor by using the tanks-in-parallel model. This function is derived from Eq. (6) and
Fig. 2. Distribution of tracer concentration in the liquid flowing in the reactor.
Table 1 Parameters of the tanks in parallel model. fq 1fq
s1 s2 s3
0.05 0.95 0.9158tm 0.0171tm 0.0167tm
– – 119.66 h 2.23 h 2.18 h
Eq. (7). Table 1 shows the parameter values of the tanks-in-parallel model, i.e., the mean residence time for each stirred tank of Fig. 1.
EðhÞ ¼ 1:077 ½eð1:092hÞ eð58:479hÞ 61:818 h eð58:479hÞ E1 ðhÞ |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} 2 þ 5400 e60h |fflfflfflfflfflfflfflfflfflfflfflfflfflh{zfflfflfflfflfflfflfflfflfflfflfflffl ffl}E2 ðhÞ
ð9Þ Fig. 3 shows the RTD fitting from the experiments and the tanks-in-parallel model (Eq. (9)). In this model, 91.58% of the reactor volume behaves as a completely mixed tank. To gain a better understanding of the system, the two dimensionless RTD functions that compose the tanks-in-parallel model, E1(h) and E2(h) (Fig. 4) were analyzed separately. Function E1(h) is the major component of the system. It gives a completely mixed behavior composed of a large, perfectly mixed tank (about 91.58%) followed by two smaller tanks in series that correspond to 3.42%. As can be seen in Fig. 4a, this 3.42% does not contribute appreciably to the distortion of a perfectly mixed tank. Function E2(h) represents the outflow of a tracer pulse indicating the presence of a plug flow behavior within a short time during the early hours of the test. E2(h) in Fig. 4b can be interpreted physically as a by-pass zone indicating that a small fraction of lithium may have gone through the reactor much faster than the rest of the tracer. Since the fit of the early experimental data (Fig. 3) arose from the distribution function E2(h), the percentage of by-pass zones can be considered to be approximately equal to the flow fraction (5%) that goes through the upper branch of the array in parallel. The by-pass fraction in the liquid probably appears early in the process while the homogenization of the reactor is taking place (Yianatos et al., 2005). The results indicate that the model of completely mixed tanks in parallel had a good fit to experimental data. The estimated mean residence time for the model is approximately 36% greater than the theoretical residence time. This behavior can be explained because the tracer is expected to delay its outflow probably due to: gas hold up, foam formation at the top and the design of the reactor’s outlet structure (Vardar Sukan, 1998; Andrade and Hodouin, 2005). Foaming delays the outflow of the tracer because the foam can be regarded as a stagnant zone where there is not a good mixing
Fig. 3. RTD experimental and the tanks in parallel model.
132
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
Fig. 4. Functions that compose the residence time distribution of the tanks in parallel model.
with the bulk liquid that leads to an increase in the effective reactor volume and enhances the gas hold-up (Vardar Sukan, 1998). Aeration also modifies the flow patterns established by the impeller (Nienow and Bujalski, 1997; Kasat and Pandit, 2005) which may contribute to the formation of foam (Vardar Sukan, 1998). As the gas flow rate rises the height of the foam layer increases because more bubbles reach the surface being converted into foam (Vardar Sukan, 1998). The liquid inside the reactor flowed out of it through a plastic hose when there was overflow. Slurry drops fall off the hose when the surface tension forces can no longer counterweight the slurry outside the hose. This causes the liquid to flow out when the pressure inside is greater than the surface tension in the form of pulses. It is important to note that the outflow due to overflow depends on the random motion and the liquid level inside the reactor. 4.2. Characterization of the biooxidation process through XRD Fig. 5 shows the variation in the concentration of ferric and ferrous iron in solution and the diffractograms obtained during the biooxidation process in batch and continuous mode. The concentration of ferrous iron in solution decreased while the ferric ion concentration increased over time within the batch mode (Fig. 5a). This behavior was interpreted as evidence that the main catalytic mechanism of bacteria involves the oxidation of Fe2+ to Fe+3 in accordance with the findings of Williamson and Rimstidt (1994), Rawlings et al. (2003), Rodríguez et al. (2003); Sand and Gehrke (2006). During the time lapse between hours 336 and 477, the concentration of ferric iron in solution had a tendency to decrease possibly due to jarosite precipitation. The XRD analysis showed that the greatest formation of jarosite occurred approximately after 477 h in the batch mode (Fig. 5c). The total iron in solution after 477 h was equivalent to 70% of the oxidation (theoretical value estimated stoichiometrically). The steady-state operation was considered to be
reached after 634.6 h since the difference in ferric iron concentration in solution was approximately 5%. During the steady-state, an average concentration of ferric iron in solution of 8.3 g/L was reached corresponding to an oxidation of 33% of total iron (this is an underestimated value because the precipitated iron in the oxidized mineral was not measured). The XRD results show that the oxidation rate of the arsenopyrite was greater than that of pyrite in both of the biooxidation reactor’s operating modes (Fig. 5). Advanced and nearly complete oxidation of the arsenopyrite and pyrite occurred after 477 h of operation in the batch mode. This could be due to fact that the mineral was in contact with the oxidizing medium for a longer time. Furthermore, high ratios of Fe3+/Fe2+ were found in solution compared to the samples obtained after 240 h of operation in batch mode, and 622 h and 722 h in continuous mode (steady state). In both modes of operation, arsenopyrite oxidation was higher than pyrite oxidation. This is possibly due to the fact that arsenopyrite oxidation is produced by the combined attack of ferric iron and protons (polysulfide mechanism). In contrast, pyrite oxidation is the result of the attack of ferric iron (thiosulfate mechanism) (Tributsch, 2001; Schippers and Sand, 1999; Fowler et al., 2001; Rawlings et al., 2003). The pH in the reactor was kept at 1.7 ± 0.1, thus maintaining the supply of protons. This is likely to have favored arsenopyrite oxidation in the continuous and batch modes. In the case of pyrite, a longer contact time with ferric iron in batch mode possibly caused higher oxidation after 477 h. It has been reported that arsenopyrite oxidizes faster than pyrite (Márquez et al., 2006; Ciftci and Akcil, 2010 and Ossa and Márquez, 2010). The cause could be its lower oxidation potential (Rimstidt et al., 1994). The XRD analyzes showed the formation of jarosite in both operation modes. The highest formation of jarosite occurred after 477 h of operation in the batch mode. This could be due to the high concentration of ferric iron in solution (Fig. 5a and c) according to what was reported by Ciftci and Akcil (2010). The decrease in the output of jarosite within the continuous mode is caused by the outflow of ferric iron and some jarosite until the system reaches a steady-state with a lower average concentration of ferric iron in solution (Fig. 5a). The formation rate of jarosite in the process was not as high as the rate observed by Zapata et al. (2004); Ossa (2004); Ossa and Márquez (2010). It might be a result of the applied control over pH (1.7 ± 0.1) during the biooxidation process. Daoud and Karamanev (2006) found that the principal parameter affecting jarosite formation is pH. A low precipitation of jarosite was obtained in a culture of A. ferrooxidans with a pH of 1.6–1.7 under a temperature of 35 °C. When the pH values are above 2.5, ferric iron has low solubility that causes the formation of basic Fe(III) hydroxyl sulfates (Gomez and Cantero, 2005; Daoud and Karamanev, 2006). There was no formation of jarosite in the control experiment because there was neither oxidation of sulfides nor increased ferric iron in solution. Jarosite is a common byproduct of the biooxidation of sulfides containing iron (García et al., 1995; Márquez et al., 2006). The final byproduct of the reaction of pyrite and ferric iron is sulfate (Schippers and Sand, 1999). In arsenopyrite oxidation, it is possible that polysulfide and elemental sulfur are generated as intermediate byproducts (Schippers and Sand, 1999). This elemental sulfur is relatively stable because it can be oxidized into sulfate through the action of sulfur-oxidizing microorganisms (Suzuki, 2001); in this case, A. thiooxidans was one of the organisms present in the culture. Furthermore, it has been proposed that jarosite is an abundant and common mineral phase in the biooxidation process of arsenopyrite (Márquez et al., 2006; Ossa and Márquez, 2010; Jiang et al., 2008; Márquez et al., 2012). According to Ciftci and Akcil (2010) the increased sulfate in the residual matter resulting from the biooxidation of a refractory gold concentrate is probably due to the precipitation of jarosite.
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
133
Fig. 5. Variation of ferric and ferrous iron concentration in solution and diffractograms of the biooxidation process in the batch and continuous modes (p: pyrite, a: arsenopyrite. j: jarosite, b: brushite, q: quarz, c: clinozoisite).
The set of data from the XRD show a high presence of silicates in the biooxidation process as a result of their exhaustive dissolution resulting from the controlled pH conditions (1.7 ± 0.1) (Fig. 5). The predominant type of silicate was quartz, and its higher proportion was observed after 477 h of operation in batch mode. This was probably due to the increased contact time with the oxidizing medium and to the fact that the biooxidation of concentrates containing no sulfides (such as oxides, carbonates and silicates) occurs simply through an acid attack (Sand and Gehrke, 2006). Fig. 5 shows the presence of brushite (CaHPO42(H2O)) in the oxidized samples. There was formation of brushite in the control sample (data is not shown in Fig. 5f) The formation of brushite could have been due to the reaction of the HPO 4 in the 9 K medium and the carbonates in the ore. The highest formation of
brushite was observed in the continuous mode. This might be explained by the permanent inflow and outflow of mineral and 9 K medium. RTD results from the liquid phase of the reactor has indicated there is a 36% of dead zones meaning that in the continuous operation mode of the reactor a 36% of a delayed outflow of ferric iron in solution is probable. This outcome could increase the possibility of jarosite formation. XRD analyses show the presence of jarosite in the continuous mode for concentrations of ferric iron in solution during the steady-state approximately 8.3 g/L and pH 1.7 ± 0.1 on average. On the other hand, the delayed outflow of the tracer decreases both the dilution rate and the specific growth rate of the microorganisms, thus affecting sulfide oxidation in the continuous operation mode (Gomez and Cantero, 2003).
134
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135
It may be deduced from the outcome of this study that jarosite precipitation could be related to dead zones since the delay in the outflow of the tracer indicates that the concentration of ferric iron in solution is likely to delay its outflow by approximately 36%, hence increasing the probability of iron precipitation inside the reactor. Further research is required in order to identify the effects of hydrodynamic behavior in the formation of precipitates in biooxidation processes, since these are carried out using large commercial-scale continuous stirred tank reactors that usually exhibit high concentrations of ferric iron during the steady state (Rawlings et al., 2003; Arrascue and Van Niekerk, 2006; Van Niekerk, 2009). Moreover, during the operation they may exhibit dead zones and by-passes that affect process performance. In this regard, jarosite precipitation in the presence of different types of microorganisms and sulfide concentrates should be assessed. Likewise, the assessment must be done for several design and operating conditions (aeration, agitation, pulp density, temperature and different impellers) that allow the variation of the by-passes and dead zones percentages inside the reactor. The hydrodynamic evaluation of the biooxidation reactor of refractory gold concentrate was carried out without the use of microorganisms. Cell growth could have affected foaming, since the dissolution of sulfides may affect the physicochemical properties of the medium and therefore the surface phenomena. According to Vardar (1998) foaming is affected by gas flow, agitation, temperature, pH, nature and composition of the medium, presence of cell growth and by the nature of the metabolites and substances produced. Foaming affects the effective volume of the liquid inside the reactor and consequently its RTD (Vardar, 1998). Likewise, sulfide concentrates other than refractory gold may affect the physicochemical properties of the medium and therefore of foaming. Considering that microorganisms are responsible for providing sulfuric acid for a proton attack and for maintaining high concentrations of ferric iron for an oxidative attack on the mineral (Rawlings et al., 2003; Sand and Gehrke, 2006), future studies should evaluate jarosite precipitation in the presence of mesophiles, moderate thermophiles and extreme thermophiles in CSTR. Ciftci and Akcil (2010) evaluated the oxidation ability of a mixed microbial culture of mesophiles, moderate thermophiles and extreme thermophiles during the biooxidation of a refractory gold concentrate at different pulp densities in flasks. These authors also evaluated the effect of biooxidation residues on sodium cyanide consumption. They found that sodium cyanide consumption was lower in cultures with extreme thermophiles compared to the cultures with mesophiles and moderate thermophiles due to the lower amount of jarosite precipitation. The hydrodynamic behavior of the reactor depends directly on the impeller. The formation of precipitates should therefore be assessed with different impellers such as axial flow hydrofoil impellers, e.g. Lightnin A315 impellers, which are used in biooxidation plants (Dew et al., 1997; Arrascue and Van Niekerk, 2006; Batty and Rorke, 2006; Van Niekerk, 2009). RTD and XRD analyzes from a laboratory-scale biooxidation reactor make it possible to obtain observations that can be used for the design and scale-up of reactors in order to improve their operation and performance.
5. Conclusions Hydrodynamic evaluation showed that the reactor had a high tendency to behave as a perfectly mixed reactor. However, the mixed flow inside the reactor had disturbances such as by-pass and dead zones that make the process far from ideal. It is highly probable that the outflow of the tracer was delayed. This was caused mainly by gas hold-up, foaming at the top and the design
of the reactor output structure itself. The tanks-in-parallel model fit well the experimental residence time distribution. This indicated that the estimated mean residence time for the model was approximately 36% greater than the theoretical residence time this was due to the delay in the outflow of the tracer. Additionally, 91.58% of the reactor’s volume behaved as a completely mixed tank and 5% of it had by-pass zones. The XRD outcomes showed a greater oxidation of arsenopyrite than pyrite besides the formation of jarosite and brushite. Although the presence of jarosite in continuous mode is lower than in batch mode, the dead zones in the continuous operation of the reactor may increase the possibility of jarosite precipitation.
Acknowledgements The authors would like to thank the Biotechnology National Program of COLCIENCIAS, CDI S.A. Mines Company and the Biomineralogy Laboratory from the National University of Colombia (Medellin campus)
References Akcil, A., 2004. Potential bioleaching developments towards commercial reality: Turkish metal mining’s future. Minerals Engineering 17 (3), 477–480. Andrade, L., Hodouin, D., 2005. Residence time distribution of an industrial mechanically agitated cyanidation tank. Minerals Engineering 18 (6), 613–621. APHA, 2000. American public health association. standard methods for the examination of water and wastewater. In: Clescerl, L.S., Greenberg, A.E., Eaton, A.D. (Eds.), American Water Works Association & Water Environment Federation, 20th ed. American Public Health Association, Washington, DC. Arrascue, M.E.L., Van Niekerk, J., 2006. Biooxidation of arsenopyrite concentrate using BIOXÒ process: industrial experience in Tamboraque, Peru. Hydrometallurgy 83 (1–4), 90–96. Arroyave, D. M., 2008. Evaluation of the biooxidation process to laboratory scale of the auriferous mineral of the ‘‘El Zancudo’’ Mine, Titiribi, Antioquia. M.Sc. Thesis, School of Materials, National University of Colombia (Medellin). Arroyave, D., Marquez, M., Gallego, D., Pacheco, G., 2010. Evaluation and mineralogical characterization of biooxidation process in a continuous stirred tank reactor. Dyna 164, 18–29. Batty, J.D., Rorke, G.V., 2006. Development and commercial demonstration of the BioCOP™ thermophile process. Hydrometallurgy 83 (1–4), 83–89. Brochot, S., Durance, M.V., Villeneuve, J., d’Hugues, P., Mugabi, M., 2004. Modelling of the bioleaching of sulfide ores: application for the simulation of the bioleaching/gravity section of the Kasese Cobalt Company Ltd. process plant. Minerals Engineering 17 (2), 253–260. Bryan, C.G., Joulian, C., Spolaore El Achbouni, P.H., Challan-Belval, S., Morin, D., d’Hugues, P., 2011. The efficiency of indigenous and designed consortia in bioleaching stirred tank reactors. Minerals Engineering 24 (11), 1149–1156. Bujalski, J.M, Jaworski, Z, Bujalski, W., Nienow, A.W., 1997. The influence of the addition position of a tracer on CFD simulated mixing times in a vessel agitated by a Rushton turbine. In: Engineering Foundation Conference on Biochemical Engineering X.
D.M. Arroyave G. et al. / Minerals Engineering 46–47 (2013) 128–135 García Jr., O., Bigham, J.M., Touvinen, O.H., 1995. Sphalerite oxidation by Thiobacillus ferrooxidans and Thiobacillus thiooxidans. Canadian Journal of Microbiology 41, 578–584. Gomez, J.M., Cantero, D., 2003. Kinetic study of biological ferrous sulphate oxidation by iron – oxidising bacteria in continues stirred tank and packed bed bioreactors. Process Biochemistry 38 (6), 867–875. Gómez, J. M., y Cantero, D., 2005. Biooxidación del ión ferroso. In: Fernando Acevedo y Juan Carlos Gentina (Ed.). Fundamentos y Perspectivas de las Tecnologías Biomineras. pp. 25–43. Gonzalez, R., Gentina, J., Acevedo, F., 2003. Optimisation of the solids suspension conditions in a continuous stirred tank reactor for the biooxidation of refractory gold concentrates. Electronic Journal of Biotechnology 6 (3), 233–240. Hayward, T., Satalic, D.M., Spencer, P.A., 1997. Engineering, equipment and materials: developments in the design of a bacterial oxidation reactor. Minerals Engineering 10 (10), 1047–1055. Jafari, M., Soltan Mohammadzadeh, J.S., 2005. Mixing time, homogenization energy and residence time distribution in a gas-induced contactor. Trans IChemE, Part A, Chemical Engineering Research and Design 83 (A5), 452–459. Jiang, T., Li, Q., Yang, Y.B., Li, G.H., Qiu, G.Z., 2008. Bio-oxidation of arsenopyrite. Transactions of Nonferrous Metals Society of China 18, 1433–1438. Kasat, G.R., Pandit, A.B., 2005. Review on mixing characteristics in solid–liquid and solid–liquid–gas reactor vessels. The Canadian Journal of Chemical Engineering 83 (4), 618–643. Levenspiel, O., 1981. Chemical Reaction Engineering, 2 ed. Reverté S.A, Barcelona. Liu, G., Yin, J., Cong, W., 2007. Effect of fluid shear and particles collision on the oxidation of ferrous iron by Acidithiobacillus ferrooxidans. Minerals Engineering 20 (13), 1227–1231. Márquez, M.G., Gaspar, J., Bessler, K., Magela, G., 2006. Process mineralogy of bacterial oxidized gold ore in Sao Bento mine (Brazil). Hydrometallurgy 83 (1–4), 114–123. Márquez, M.A., Ospina, J.D., Morales, A.L., 2012. New insights about the bacterial oxidation of arsenopyrite: a mineralogical scope. Minerals Engineering 39, 248– 254. Mazuelos, A., Romero, R., Palencia, I., Carranza, F., Borjas, F.J., 2002. Oxygen transfer in ferric iron biological production in a packed-bed reactor. Hydrometallurgy 65 (1), 15–22. Nienow, A. W., Bujalski, W., 1997. Recent studies on agitated three phase (gas– solid–liquid) systems in the turbulent regime. In: Engineering Foundation Conference on Biochemical Engineering X.
135
Rawlings, D.E., Dew, D., du Plessis, C., 2003. Biomineralization of metal containing ores and concentrates. TRENDS in Biotechnology 21 (1), 38–44. Rimstidt, J., Chermak, J.A., Cagen, P.M., 1994. Rates of reaction of galena, sphalerite, chalcopyrite, and arsenopyrite with Fe(III) in acidic solutions. In: Alpers, C.N., Blowes, D.W. (Eds.), Environmental Geochemistry of Sulfide Oxidation, vol. 550.. ACS Symposium Series. American Chemical Society, Washington, D.C, pp. 1–13. Rodríguez, Y., Ballester, A., Blázquez, M.L., González, F., Muñoz, J.A., 2003. New information on the pyrite bioleaching mechanism at low and high temperature. Hydrometallurgy 71 (1–2), 37–46. Romero, R., Palencia, I., Carranza, F., 1998. Silver catalyzed IBES process: application to a Spanish copper–zinc sulfide concentrate Part 3. Selection of the operational parameters for a continuous pilot plant. Hydrometallurgy 49 (1–2), 75–86. Rossi G. 1990. Biohydrometallurgy. McGraw – Hill Book Company Gmbh. Sand, W., Gehrke, T., 2006. Extracellular polymeric substances mediate bioleaching/ biocorrosion via interfacial processes involving iron(III) ions and acidophilic bacteria. Research in Microbiology 157 (1), 49–56. Schippers, A., Sand, W., 1999. Bacterial leaching of metal sulfides proceeds by two indirect mechanisms via thiosulfate or via polysulfides and sulfur. Applied and Environmental Microbiology 65 (1), 319–321. Silverman, M.P., Lundgren, D.G., 1959. Studies on the chemolithotrophic iron bacterium Ferrobacillus ferrooxidans an improved medium and a harvesting procedure for securing high cell yields. Journal of Bacteriology 77, 642–647. Sun, Li-Xin, Zhang, Xu, Tan, Wen-Song, Zhu, Ming-Long, 2012. Effect of agitation intensity on the biooxidation process of refractory gold ores by Acidithiobacillus ferrooxidans. Hydrometallurgy 127–128, 99–103. Suzuki, Isamu, 2001. Microbial leaching of metals from sulfide minerals. Biotechnology Advances 19 (2), 119–132. Szekely, J., Themelis, N., 1971. Rate Phenomena in Process Metallurgy. WileyInterscience, a Division of John Wiley & Sons, Inc.. Tributsch, Helmut, 2001. Direct versus indirect bioleaching. Hydrometallurgy 59 (2–3), 177–185. Van Aswegen, Pieter C., Van Niekerk, Jan, 2007. The BIOXTM process for the treatment of refractory gold concentrates. In: Rawlings, Douglas E., Johnson, B.D. (Eds.), Biomining. Springer Verlag, Berlin, Germany, pp. 1–32. Van Niekerk, J., 2009. Recent advances in BIOXÒ technology. In: Hydrometallurgy Conference 2009. The Southern African Institute of Mining and Metallurgy. Vardar, Sukan Fazilet, 1998. Foaming: consequences, prevention and destruction. Biotechnology Advances 16 (5–6), 913–948. Watling, H.R., 2008. The bioleaching of nickel–copper sulfides. Hydrometallurgy 91, 70–88. Williamson, M.A., Rimstidt, J.D., 1994. The kinetics and electrochemical ratedetermining step of aqueous pyrite oxidation. Geochimica Cosmochimica Acta 58 (24), 5443–5454. Yianatos, J.B., Bergha, L.G., Díaz, F., Rodríguez, J., 2005. Mixing characteristics of industrial flotation equipment. Chem Engineering Science 60 (8–9), 2273–2282. Zapata, D., Gallego, A., Marquez M. A., 2004. Definition of the type of refractoriness in El Zancudo Golg Mine, Antioquia, Colombia. Appl. In: Miner., Pecchio et al. (Eds.).