Sulphuric acid pressure leaching of laterites — a comprehensive model of a continuous autoclave

Hydrometallurgy 58 Ž2000. 89–101 www.elsevier.nlrlocaterhydromet

Sulphuric acid pressure leaching of laterites — a comprehensive model of a continuous autoclave D.H. Rubisov, V.G. Papangelakis ) Department of Chemical Engineering and Applied Chemistry, UniÕersity of Toronto, 200 College Street, Toronto, ON, Canada M5S 3E5 Received 27 April 1999; received in revised form 10 March 2000; accepted 29 March 2000

Abstract The high-temperature acid pressure leaching of nickeliferous laterites employs multi-compartment continuous autoclaves treating a variety of lateritic feeds. Previous work has correlated the kinetics of nickel dissolution, as well as the solubility of important metals in this process, with acid concentration levels calculated at the temperature of leaching. This was achieved by a speciation analysis resulting in a universal kinetic equation for Ni dissolution from limonitic feeds and from limoniticrsaprolitic blends. In the present work, these findings were combined to develop a comprehensive model for a continuous multi-compartment autoclave. To this end, the kinetics of Co, Fe, Al, and Mn was also estimated and included in the reactor model. The resulting model is capable of predicting Ni extraction and concentrations of major impurities during autoclave operation for a wide variety of process conditions and feed compositions. It was implemented in a computer simulator that can be used to investigate diverse process scenarios and to select acidity levels for different types of feed. Finally, the model is compared with continuous mini-plant pressure acid leaching data as provided by INCO Technical Services ŽITSL. during process development campaigns. q 2000 Elsevier Science B.V. All rights reserved. Keywords: Sulphuric acid; Laterites; Autoclave

1. Introduction High-temperature pressure acid leaching is currently the process of choice to recover nickel and cobalt from limonitic laterites. Previous experience demonstrated that a series of well-agitated continu-

)

Corresponding author. Tel.: q1-416-978-1093; fax: q1-416978-8605. E-mail address: [email protected] ŽV.G. Papangelakis..

ous autoclaves is the best choice for successful nickel and cobalt extraction by sulphuric acid leaching at 250–2708C w1,2x. An optimal mixture of limonites and saprolites may also form a high-grade feed that, at the same time, yields an acceptable acid consumption. Experimental studies of the direct sulphuric acid pressure leaching of laterites conducted at the University of Toronto over the last 6 years w1–6x have produced kinetic data for Ni, Co, Fe, and Al dissolution for a variety of conditions: temperatures from 2308C to 2708C, acidrore ratios from 0.15 to 0.5 and pulp densities from 10 to 30 wt.% solids. In addition,

0169-4332r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 3 8 6 X Ž 0 0 . 0 0 0 9 2 - X

90

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

a synthetic Al 2 ŽSO4 . 3 –H 2 SO4 system was studied experimentally and theoretically, allowing the identification of dissolutionrprecipitation kinetics and solubilities of aluminium w7–9x in this process. In the present work, a logical extension of previous work is made towards building a comprehensive continuous autoclave model capable of predicting the behaviour not only of Ni and Co, but also of Fe and Al, as well as Mg and Mn. The availability of data from continuous mini-plant tests with limonitic feeds and limoniticrsaprolitic blends that were provided by INCO Technical Services ŽITSL. facilitated the validation of the model. The solubility of metals that precipitate during the process was investigated first. A solution speciation approach was thus formulated for Al, Fe and Mg w10x. Furthermore, the use of the hydrogen ion concentration Aat temperatureB allowed us to develop a unified kinetic equation applicable to both limonitic feeds and limoniticrsaprolitic blends w11x. The successful development of Ni kinetics based on wHqx Aat temperatureB, as well the expression of metal solubilities as a function of wHqx Aat temperature,B constitutes the foundation for the autoclave model discussed in the present paper.

2. Review of model assumptions based on previous findings Equilibrium constants for the precipitation of Al, Fe and Mg and a kinetic equation for Ni dissolution were presented in the previous publications by the authors w10,11x. The major findings of our past work are used as assumptions to develop the model and are shown in Table 1. Additional assumptions are explained below. 2.1. Speciation The present model is entirely based on the process solution chemistry under autoclave conditions. Hence, a high-temperature speciation model w10x is used to evaluate wHqx T and wHSO4yx T . The subscript ATB hereafter denotes that the concentrations are Aat temperatureB. Solution volume expansion is taken into account w10x.

2.2. Ni kinetics Ni kinetic equations are given in Table 1. It should be noted that the ultimate extraction of nickel shown in Table 1 is achievable only if wHqx T exceeds 0.1 molrL w11x. The following polynomial regression with coefficients linearly dependent on the temperature uc Žin 8C. is used to correlate the ultimate Ni extraction with wHqx T if wHqx T - 0.1 molrL: 4

X` s y Ž 2.15uc y 529 . = 10 4 w Hq x T 3

q Ž 0.903uc y 223 . = 10 4 w Hq x T 2

y Ž 1.19uc y 286 . = 10 3 w Hq x T q Ž 0.514uc y 109 . = 10 2 w Hq x T .

Ž 1.

2.3. Mg kinetics The kinetics of Mg dissolution was found to be extremely fast to be properly identified via a kinetic equation w11x. Hence, it is assumed that in a continuous autoclave, 85% of AsolubleB Mg is dissolved during the first 25% of the total mean residence time, and the remaining 15% in the second 25%. The following considerations were taken into account in estimating the amount of AsolubleB Mg in the feed. Batch tests with a blend containing 2.47 wt.% Mg yielded Mg extraction values ranging from 73.2% to 76.7%. Hence, a mean value of X` s 75% is adopted for limoniticrsaprolitic blends. On the other hand, experimental data from limonitic laterites w10x suggested that a mean terminal Mg extraction of 25% is reasonable and this value was adopted. 2.4. Mg solubility The equilibrium constant for Mg given in Table 1 was derived for pure MgSO4 –H 2 SO4 systems. This equation was also used for the prediction of Mg solubility in real laterite leaching systems w10x, and it is adopted in the process model. Based on recent experimental findings w12x, it is further assumed that Ni, Co, and Mn also precipitate with Mg as a mixed sulphate compound when the solubility of Mg is exceeded. The molar ratio of the precipitated metals

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

91

Table 1 Model assumptions based on the previous findings Property

Assumption

Reference 87.75=103

Ž1. Ni kinetics Žlimonite. Ž2. Ultimate Ni extraction Žlimonite. Ž3. Ni kinetics Žlimoniticr saprolitic blend. Ž4. Ultimate Ni extraction Žlimoniticrsaprolitic blend. Ž5. Al solubility

7.806 = 10 7ey X` s 97.2%

wHq x T t s 1 y 3Ž1 y XrX` . 2r3 q 2Ž1 y XrX` .

w11x w11x

Same as in limonite for Ni associated with Fe Žgoethite.; instantaneous for Ni associated with Mg Žserpentine. X` s 97.2% for Ni associated with Fe, X` s 100% for Ni associated with Mg 3Al 2 ŽSO4 . 3 Žaq. q 14H 2 O

w lg b s lg Ž6. Fe solubility

RT

w

™ 2ŽH O.Al ŽSO . ŽOH. Žs. q 5H q 5HSO q

3

5r2 HSO4y T 3r2 Al 2 SO4 3 aq T 5r2 Hq T

x

Ž

w

x

. Ž .x

Fe 2 ŽSO4 . 3 Žaq. q 2H 2 O

3

4 2

y 4

6

' s y4464rTq 3.774 q 1 q 1.6'I 14.21 Ic

™ 2FeOHSO Žs. q HSO q H y 4

4

w11x w11x w10x

y 1.667Ic

c

q

w10x

9.905'Ic w Hq x T w HSO4y x T lg b s lg s y2740rTq 10.72 y q 0.0294 Ic w Fe 2 ŽSO4 . 3 Žaq. x T 1 q 1.6'Ic Ž7. Mg solubility

Mg 5 ŽSO4 . 2 ŽHSO4 . 6 Žaq. q 5H 2 O 3 Hq T

3 HSO4y T

™ 5MgSO P H OŽs. q 3HSO q 3 H 4

2

y 4

q

w10x

34.62'Ic w x w x lg b s lg s y1312rTy 9.369 q y 1.484 Ic w Mg 5 ŽSO4 . 2 ŽHSO4 . 6 Žaq. x T 1 q 1.6'Ic Ž8. Sulphate precipitation

Mixed sulphate of Ni, Co, Mn and Mg precipitates if the solubility of Mg is exceeded. The molar proportion of metals in the salt is: NirCorMnrMg s 0.22r0.03r0.1r1.

in the mixed salt is given in Table 1. Finally, when the solution becomes supersaturated, the model calculates a steady state magnesium concentration as the arithmetic mean of the solubility of Mg and its concentration assuming no precipitation.

3. Additional kinetic considerations 3.1. Fe kinetics The chemistry of iron dissolution–precipitation was reviewed in detail previously w10x. As also discussed in the past w3x, it is difficult to measure the rates of iron dissolution and precipitation separately because of the simultaneous occurrence of two reactions. However, it was assumed that nickel in limonites is associated with goethite only, although it is recognized that a small fraction of the Ni is associated with the asbolane phase. With this assumption, iron dissolution kinetics can be approximated using the nickel dissolution kinetics. Then,

w12x

iron precipitation kinetics may be recovered by subtracting the measured concentration of Fe in solution from the theoretical amount of Fe dissolved Žassuming no precipitation., which is calculated using the Ni dissolution kinetics. Further analysis showed that at any time after the first 2.5 min of leaching, the ratio of the measured concentration of dissolved Fe over the iron that was supposed to dissolve but had no precipitation, was always less than 0.02. This indicates a very rapid precipitation that closely follows the dissolution of goethite. It was also shown that the saturation ratio of iron linearly correlates with the relative Ni extraction, XrX` . A regression line was obtained from experiments with a limonitic feed at 22% solids and at various acidities and temperatures w3x: S s 6.725 y 5.735 XrX` ,

Ž 2.

where S is the saturation ratio of iron. Fig. 1a shows the fit of Eq. 2 to the experiments used to derive this equation w3x.

92

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

Fig. 1. Dependence of iron saturation ratio on Ni extraction. Data for the limonitic feed from Ref. w3x, for blends from Ref. w11x, aro is the acid-to-ore ratio by weight Ždry ore..

In the present model, it was also assumed that Eq. 2 is applicable to other feed compositions, including limoniticrsaprolitic blends. Fig. 1b and c show the fit of Eq. 2 to experimental data at higher slurry density w3x and to experiments with limoniticrsaprolitic blends w11x, respectively. As seen in Fig. 1c, Eq. 2 underpredicts the iron saturation ratio in the case of blends. This underestimation is not very important though, because ferric iron forms neutral sulphate complexes w10x and its concentration is low. As such, ferric iron concentrations do not affect the calculation of acidity Aat temperatureB. Thus, they do not affect the leaching kinetics and solubilities of other metals. By knowing the value of S for iron at any Ni extraction level, and the iron solubility at temperature, iron concentration in the solution can be calculated.

As stated above, iron precipitation as hematite is assumed to proceed through the formation of basic ferric sulphate. The latter quickly converts to hematite. For this process model, it is important to know the amount of basic ferric sulphate in the precipitate that is not converted to hematite. This allows us to calculate the total dissolved sulphate concentration as the amount of sulphate added with sulphuric acid, less that precipitated with Al and Fe. Equilibrating this total aqueous sulphate concentration with that calculated from the speciation analysis w10x makes it possible to calculate wHqx T and consequently all other metal concentrations. While the presence of unconverted basic ferric sulphate in solid residues will have to be confirmed by further experiments, the following observation may be considered as an indication that this conver-

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

Fig. 2. Sulphate losses with basic ferric sulphate as a function of bisulphate concentration.

sion is incomplete. Sulphate balance calculations in batch experiments showed that sulphate rejection from solution by means of hydronium alunite precipitation does not suffice. Hence, extra SO4 must report in the residue as basic ferric sulphate. In limonitic batch systems, calculations showed that the amount of iron in basic ferric sulphate ŽFeSO4 OH. is relatively small, ranging from 1% to 5% of the total iron in the feed. In limonoticrsaprolytic blends, it was found to range from 3% to 14%. These relatively small amounts greatly affect the calculation of the free acidity in the system. It was also found that the amount of FeSO4 OHŽs. depends on the wHSO4yx T . As seen in Fig. 2, there is a clear increase in the residual basic ferric sulphate at higher bisulphate concentrations. This trend is approximated with a straight line in Fig. 2 with a correlation coefficient of 0.81: FeSO4 OH Ž s . s 1.05 HSO4y

T y 0.05.

Ž 3.

It should also be noted that the calculated amount of FeSO4 OH was always 1.5–2 times higher at 2.5 min than at 10 min of leaching but stayed constant afterwards. This makes Eq. 3 acceptable for modelling steady state operations of continuous autoclaves with residence time above 10 min. 3.2. Kinetics of Al dissolution and precipitation Since the predominant Al phase in laterites is gibbsite, the kinetics of Al dissolution and precipitation was derived using data from a pure system of gibbsite and sulphuric acid w7x. In a departure from a

93

simultaneous search for dissolution and precipitation kinetic parameters w9x, the kinetics was separated using two speciation models; one at room temperature w8x, and another at high temperature w10x. The calculation of Al dissolved and Al precipitated was based on a total sulphate balance. Data for only five experiments at low acidity w8x Ž0.25 molrL of initial sulphuric acid, temperatures of 2308C, 2408C, 2508C, 2608C and 2708C. were used. At higher acidities Ž0.5 and 0.75 molrL., more than 90% of Al dissolves fast within 2.5 min, making it impossible to establish the dissolution kinetics. Similar to Ni, Al kinetics is based on the hydrogen ion concentration Aat temperatureB. The same kinetic equation as for Ni dissolution Žshrinking core model, spherical particles, diffusion through solid products control., albeit with different coefficients, was found to fit best the Al dissolution: 2.68 = 10 9 ey

101.9=103 RT

s 1 y 3 Ž 1 y XrX` .

w Hq x T t 2r3

q 2 Ž 1 y XrX` . ,

Ž 4.

where R is the ideal gas constant in Joules and T is the absolute temperature, t in min. An Arrhenius-type equation was used to express this dependence of the kinetic rate constant on temperature. The prediction given by Eq. 4 yielded a mean RMS error of 2.63; i.e., the percent extraction of Al is predicted with an accuracy of "2.63% at a confidence level of 70%. Eq. 4 was then checked against experiments with limonitic laterites w3x in which no precipitation of Al was observed. The experiments were conducted at 10% of solids, acidrore ratios of 0.25 and 0.35, and temperatures of 2308C and 2508C, respectively. The RMS error for X in percent evaluated from Eq. 4 was less than 4.8. Most probably, this higher error is due to gibbsite not being the only Al phase in laterites w1,2,10x. Nevertheless, Eq. 4 is used in the present model to estimate Al extraction. It is known w8x that the ultimate extraction Ži.e., dissolution. of gibbsite does not proceed to 100%. Our high-temperature speciation revealed that the ultimate extraction depends on the solution acidity if the latter is less than 0.3 molrL, but it is practically independent of temperature. A linear regression of the ultimate extraction X` of aluminium on hydro-

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

94

gen ion Aat temperatureB was established in percent units:

3.3. BehaÕiour of Mn and Co during laterite leaching

X` s 68.4 q 107 w Hq x T

Co in laterites is mainly associated with the manganese mineral, asbolane. Asbolane Žor asbolite., which is also known as earthy cobalt, mostly consists of MnO 2 and may contain up to 30% of Co and some Ni w17x. This mineral is usually amorphous, which makes it very difficult to detect using XRD analysis. However, a strong correlation between the concentrations of Mn and Co in ore samples has indeed been observed w18x and confirmed in our ore samples using TEM w15x. A comparison of Co and Mn kinetics also reveals their close association. Dissolution kinetics of Co and Mn in limonites for two typical experiments is given in Fig. 3. It is seen that the kinetics for both metals is practically identical. The ultimate extractions of cobalt and manganese showed different trends though, as shown in Table 2. The ultimate extraction of Co varied within a very narrow range, from 88% to 91.6% with an average of 90.4%. The variation of the ultimate extraction of Mn was much more significant, from 50.5% to 72.2%. Moreover, the terminal Mn concentration in limonitic feeds shown in Table 2 shows a dependence on the terminal free acidity. The following exponential plot fits the experimental data:

if w Hq x T - 0.3 molrL

Ž 5. X` s 100

if w Hq x T ) 0.3 molrL.

In establishing the precipitation kinetics of Al, the following considerations were taken into account. Precipitation is spontaneous and therefore, particles are growing by nucleus agglomeration rather than by crystal growth w13x. This is applicable to both iron and aluminium precipitates. Indeed, numerous photomicrographs of residues showed that hematite particles grow to approximately the same size w1–3,12x. This behaviour cannot be explained assuming a crystal growth mechanism. Instead, it is very common for agglomeration processes where particles eventually reach a certain cut-off size w14x. Transmission electron microscopy ŽTEM. microphotographs of aluminium precipitates showed that they are amorphous and form relatively big Ž3–10 mm. and very porous structures w15x. These properties are typical for agglomerates of extremely fine grains. The agglomeration of hydronium alunite particles does not affect the Al balance and, therefore, the agglomeration kinetics is not important for modelling purposes. To further simplify the kinetic model, it was also assumed that hydronium alunite precipitation starts when the saturation ratio is equal to 1. It was adopted that under conditions of non-controlled precipitation, a secondary nucleation mechanism dominates w13x. Results from 10 experiments w8x Ž0.25 and 0.5 molrL of initial sulphuric acid, temperatures of 2308C, 2408C, 2508C, 2608C and 2708C. were used to establish the kinetic equation. It was found that a first-order kinetic equation w16x with the rate constant proportional to wHqx T fits the experimental data: dM dt

s 0.408 w Hq x T Ž S y 1 . M.

Ž 6.

Here, M is the mass of precipitate in grams per liter, and S is the saturation ratio of aluminium.

w Mnx s 9.82 = 10y3 exp Ž 4.91 w H 2 SO4 x free . ,

Ž 7.

and is given in Fig. 4. This suggests that wMnx is limited by its solubility. However, Eq. 7 was tested against batch experimental data provided by ITSL and showed a significant deviation. A simple linear correlation of wMnx on total aqueous sulphate yielded:

w Mnx s 0.05 w SO4 x total ,

Ž 8.

and fitted experimental data for both limonitic feeds and limoniticrsaprolitic blends with a correlation coefficient of 0.75. Because Eq. 8 was fitted on data obtained by slurry sampling Žas opposed to filtered sampling., the concentrations in Eq. 8 are at room temperature.

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

95

Fig. 3. Comparison of Co and Mn leaching kinetics.

For limoniticrsaprolitic blends, both Eqs. 7 and 8 predict Mn extraction of 100%, whereas 90–95% of Mn extraction was in fact measured. This may be attributed to the fact that part of Mn is in a nonleachable phase that limits the applicability of Eq. 8 to approximately 90% of Mn extraction. Further investigation of the behaviour Mn was not undertaken because its value is affected by chromium. The dissolution of MnŽIV. proceeds through a reduction to MnŽII.. In limonitic feeds, CrŽIII., which exists in spinels, acts as a reducing agent. The reaction with pyrolosite leads to the formation of CrŽVI. w1x: 2Cr 3qq 3MnO 2 q 2H 2 O

™ 2HCrO q 3Mn y 4

2q

q 2Hq.

Ž I.

Since the solubilities of MnŽIV. and CrŽIII. in sulphuric acid at elevated temperatures are negligible w19x, Reaction I suggests that the molar concentrations of Mn and Cr in solution should have a ratio of 1.5. This conclusion was indeed confirmed in our batch tests. The wMnxrwCrx ratio in limonitic solutions Žexcept for the feed with Al ) 5%. had a mean value of 1.44. In a limoniticrsaprolitic blend, this

ratio was higher, having a mean value of 2.1. This is attributed to the presence of divalent iron in serpentine. Saprolites are rich in serpentine where part of Mg is replaced by other divalent metals, mostly Ni and FeŽII.. Ferrous iron in serpentine quickly dissolves w11x and takes over the reducing role of chromium in the Mn dissolution:

™ 2Fe

2Fe 2qq MnO 2 q 4Hq

3q

q Mn2qq 2H 2 O.

Ž II .

Finally, it is assumed that the relative extraction of Mn Ži.e., XrX` s wMnxrwMnx` . is the same as for Co. Since the ultimate extraction values of cobalt in limonitersaprolite blends were practically in the same range as for limonitic feeds, varying from 85.8% to 91.4%, the same value as for limonites Ži.e., 91.6%. was accepted. Ultimate Mn extraction

Table 2 Ultimate Co and Mn extractions Žexperiment number is from Table 2, Ref. w11x. Experiment 1

2

4

5

7

8

9

Mean Max

X` for Co 90.7 89.8 88.0 91.0 91.3 90.2 91.6 90.4 X` for Mn 61.8 61.4 72.2 50.5 55.7 54.5 68.3 60.6

91.6 72.2

Fig. 4. Terminal wMnx vs. free acid for limonitic feeds Žfeed analysis is from Table 1 in Ref. w11x..

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

96

was always estimated from Eq. 8 up to a maximum level of 90%. Finally, chromium is calculated from the Mn concentration. For limonitic feeds, the theoretical value of wMnxrwCrx s 1.5 is adopted, whereas a value of 2.1 is adopted for limoniticrsaprolitic blends. 3.4. Kinetics of Co dissolution The kinetics of Co dissolution was established in the same way as the Ni kinetics w11x. The following equation fits the data best of all: 3179

0.303ey

RT

2

w Hq x T t s Ž 1 y Ž 1 y XrX` . 1r3 . .

Ž 9.

Similar to the Ni kinetic equation, Eq. 9 is also a shrinking core model equation for spherical particles assuming a solids diffusion control. However, it was derived using the simplifying assumption for a constant curvature of the particles w20x. The RMS error for X in percent varied in the range from 1.1 to 7.5 with an average of 4.5. The value of the activation energy in Eq. 9 is low, 3.2 kJrmol K, suggesting a very weak dependence of the rate constant on temperature. The Co kinetic equation established for limonitic feeds ŽEq. 9. was also tested against limoniticr saprolitic blends. The RMS errors were close to those calculated for limonitic experiments, from 1.4 to 8.6 with an average of 4.5. This relatively high RMS error may be attributed to the influence of Cr and ferrous iron on the kinetics of Mn and Co. Additional experimental work with a pure synthetic Mn–Co–Cr system is required to elucidate this influence. 4. Continuous autoclave model 4.1. Model deÕelopment The speciation program, solubility formulae and kinetic equations were combined together to form a continuous steady state autoclave model. The computer code was written in AMathematica w B version 3.0 for Windows 95rNT w21x. The simulator takes into account the autoclave configuration Žnumber of compartments and distribution of acid injectors., operating conditions Žtemperature, residence time, acidrore ratio, solids percent in the slurry., and the

feed grade Žcontents of Ni, Co, Fe, Al, Mg and Mn.. In the case of a limoniticrsaprolitic blend, the assays of Ni, Mg and Fe in both limonitic and saprolitic parts of the blend are also required. The calculations are performed so that the results Žextractions and concentrations. for each continuous autoclave compartment form the input information for the next compartment. It is assumed that the size distribution of the feed is the same as in the batch experiments. This assumption should not introduce a serious error because all the leaching kinetics are diffusion-controlled, which means that the particle size refers to the size of crystals that form porous agglomerates w3x. The size distribution of the undissolved solids coming out of each compartment is evaluated w22x and used as the distribution of the feed to the following compartment. The simulator accounts for acid injection into any compartment. The simulator flowchart is given in Fig. 5. The equations used in the simulator are explained in Appendix A. The simulator calculates extractions of Ni and Co, as well as the concentrations of the impurities ŽMg, Mn, Al, Fe, Cr. at temperature. The concentrations of hydrogen and bisulphate ions at temperature are, of course, also calculated. These two parameters cannot be directly measured, and their prediction is extremely important in order to understand and correlate the chemical changes inside the autoclave under various operating conditions. Metal concentrations are then rendered back to room temperature for comparison with experimental data. 4.2. Model Õerification The model performance was tested against data supplied by ITSL from continuous mini-plant pressure acid leaching testwork. Continuous leaching tests were conducted both with limonitic feeds and with limoniticrsaprolitic blends of various compositions. The operating conditions were: approximately 30% solids Žbefore acid addition., 2508C and 2708C, 0.21–0.52 acidrore ratio, and 30 and 45 min mean residence time. Two slurry samples were taken. The first sampling point was after 25% or 50% of the total residence time. The second sampling point was at the end of the autoclave Žfrom the discharge.. The comparison of Ni extraction predicted and measured at the two sampling points is shown in

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

97

Fig. 5. Flowchart of the continuous autoclave simulator.

Fig. 6. As seen, the prediction of Ni extraction in the discharge is better than at the first sampling point. The prediction of Ni extraction was within "5% in five tests, whereas for the other three tests, the extraction values were underpredicted by 8–15%. Overall, the model underpredicted the extraction in six tests out of eight. This underprediction is attributed to the method of sampling employed. The pilot plant data were collected under finite cooling and filtration times. Moreover, these times were probably not the same for different pilot campaigns. At the same time, the developed model is based on variables Aat temperatureB. This in turn implies that

the predictions must be compared with the analysis of instantaneously filtered samples. During sample handling, nickel dissolution continues, albeit at a lower rate; hence, a higher extraction is obtained. The extraction of Co, given in the same figure, was also underestimated at the first sampling point. Since the activation energy of Co dissolution is much lower than that for Ni, the rate of Co dissolution should still be fast during solution handling. Thus, a greater underprediction is expected. On the other hand, given the relation of Co with Mn in asbolane, Co extraction may also be affected by the kinetics of Cr dissolution as discussed above.

98

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

Fig. 6. Comparison of Ni and Co extractions measured in ITSL continuous leaching testwork with model prediction.

Fig. 7 gives a comparison of Al and total sulphate concentrations. Contrary to Ni and Co, the model prediction is good at the first sampling point. However, the concentrations of Al in the discharge were up to 60% underestimated. Again, dissolution of hydronium alunite during sampling is the most probable cause. This dissolution was not important at the first sampling point because the amount of hydronium alunite precipitated within 25% of the total residence time was not significant. Dissolution of hydronium alunite also affects the prediction of total sulphate in the discharge. Similar to Al, it is expected that the predicted concentrations of Fe might be lower than those measured in the continuous tests because of hematite dissolution from the slurry samples during cooling.

The apparent predictive ability of the model can be improved if appropriate compensating corrections were introduced, or if the model were compared against instantaneously filtered samples. The former would require inclusion of dissolution kinetics and solubilities at lower temperatures, as well as exact timing of handling procedures, which renders the task impractical. The latter would require more sophisticated sampling techniques than currently employed in conventional piloting laboratories around the world. Finally, a more accurate solution chemical model is also expected to improve the predictive ability of the model. Nevertheless, for the given complexity of the chemical system, the present model Žbuilt on first principles., as tested on a variety of feeds and conditions, is considered acceptable.

Fig. 7. Comparison of aqueous Al and total sulphate concentrations measured in ITSL continuous leaching testwork with model prediction.

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

5. Conclusions A continuous autoclave model for sulphuric acid pressure leaching of laterites was built based on first principles, i.e., leaching and precipitation kinetics as well as metal solubilities. The model was developed on batch laboratory data and successfully tested against data generated during continuous mini-plant campaigns by ITSL. The model was found applicable to both limonitic feeds and limoniticrsaprolitic blends. This model universality is entirely due to its account for the process chemistry Aat temperatureB, i.e., under autoclave process conditions. In particular, the hydrogen ion concentration is used to correlate leaching kinetics. However, the selection of proper dominant metal ions and complexes is a prerequisite for the prediction of wHqx T . To this end, a speciation program was developed. The model was implemented in a computer simulator that can be used to investigate diverse process scenarios. The extra assumptions made to arrive at a model that correlated well with experimental data are also important findings that have their own merits. Hence, the findings reported in this paper are as follows:

v

v

v

v

v

Precipitation of hydronium alunite is not the only way for sulphate to be rejected from the system. Most probably, part of SO4 stays in the residue as basic ferric sulphate. Co and Mn are closely associated in laterites and follow similar dissolution kinetics. A solids diffusion-controlled shrinking core model equation that assumes a constant curvature of particles was found to fit the data. The ultimate Co extraction is approximately 91%. Mn dissolves up to its solubility, which depends on the sulphate concentration. This solubility may also be affected by the presence of chromium. Kinetics of Al dissolution is also described by a diffusion-controlled shrinking core model equation. A model of secondary nucleation with subsequent agglomeration of the nuclei was adopted for Al precipitation. Aluminium needs at least 0.3 molrL of Hq Aat temperatureB to dissolve completely from the ore and partially convert to alunite. Ni requires at least 0.1 molrL.

99

The kinetic equations used in the model were derived from batch leaching experiments with various feed compositions conducted at temperatures from 2308C to 2708C. The model considers the concentrations of seven metals ŽNi, Co, Mg, Mn, Al, Fe, Cr. correlated with wHqx T in sulphuric acid media. The model yields an underprediction of dissolution in the first compartment of a continuous autoclave and this is particularly true for cobalt. This is attributed to the sampling methods, and to the influence of CrŽIII. and FeŽII. on the dissolution of Mn with which cobalt is associated.

Acknowledgements This work was financially supported by ITSL and the Natural Science and Engineering Research Council of Canada. The authors wish to thank the personnel of ITSL for the continuous support, fruitful discussions over the course of the project, and for the permission to publish these results.

Appendix A. Equations for continuous reactor The development of the continuous autoclave model is based on the population balance equations derived by the authors w22x. Applying the population balance method on a compartment-by-compartment basis, one may obtain the population density function of the unleached particles coming out of the ith compartment as:

ci Ž D . 1 sy

Dm

H t D˙ D

ž

exp y

1

t D˙

/

Ž D y z . ciy1 Ž z . d z , Ž A1.

where D is the particle size, t is the mean residence time of the ith compartment, D˙ is the differential form of the batch kinetic equation. In the case of a solid diffusion control, shrinking core model Žkinet-

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

100

ics of Ni and Al dissolution., the differential form of the batch kinetics is as follows: D˙ s y

k w Hq x T D 02

,

6 D Ž 1 y DrD 0 .

Ž A2.

for the diffusion control, and: X 1rX` s

where k is the rate constant and D 0 is the initial particle size. For the Jander equation Žkinetics of Co dissolution., the differential form of the batch kinetics is as follows: D˙ s y

k w Hq x T D 0 2 Ž 1 y DrD 0 .

.

2

ž

q 1q

kt w H x

ž

=exp y

exp y

0

z2 kt w Hq x T

/ /

dz ,

for the Jander model. The population density functions evaluated from Eqs. A4 and A5 for the first compartment may be used in Eq. A1 to obtain the population density functions for the second compartment, etc. Finally, extraction for an i-compartment autoclave is given by: iy2

kt w Hq x T , j

i

X irX` s 1 y 6

Ý js1

i ,l/j

Ł Ž kt wHq x T , j y kt wHq x T ,l .

ls1

1

D

2

=

kt w H x T

1

H0

ž ž / 1y3

q

ž / //

D0

4

ž

z Ž 1 y z . exp y

z2 Ž3y2 z . kt w Hq x T , j

/

d z,

Ž A8.

,

Ž A4.

ž

kt w H x T

ž

1y

D D0

2

//

,

Ž A5.

for the Jander model. Eqs. A4 and A5 can be integrated over all sizes to calculate the corresponding extraction values for the first compartment:

=exp y

Ł Ž kt wHq x T , j y kt wHq x T ,l .

1

3

H0 z Ž 1 y z . exp

q

1 4

kt w Hq x T

i ,l/j ls1

=

1

6

Ý js1

kt w Hq x T D 0

=exp y

iy2

kt w Hq x T , j

i

X irX` s 1 y 2

2 n Ž 1 y DrD 0 .

ž

/H ž

y2

Ž A7.

for the diffusion control, and:

X 1rX` s 1 y

1

2 T D0

D0

c 1Ž D . s

2

/

3

D

q2

kt w Hq x T

kt w Hq x T

6 Dn Ž 1 y DrD 0 . q

1

ž ž

exp y

Ž A3.

Since goethite particles are agglomerates of needle-like grains and, therefore, a particle size analysis of these grains is hardly possible, it was assumed that the grains are monosized, i.e., the size distribution is given by d function. Then integrating Eq. A1 with Eq. A2 or Eq. A3 and replacing c 0 with ndŽ D 0 –D . Ž n is the number of particles per unit volume., one may obtain:

c 1Ž D . s

3kt w Hq x T

H0 z 1

kt w Hq x T

Ž1yz .

/

Ž1y3 z2 q2 z3. d z, Ž A6.

ž

z2 y

kt w Hq x T , j

/

d z,

Ž A9. for the diffusion control and the Jander models, respectively. If the extraction of Al obtained from Eq. A8 exceeds its solubility, Al precipitates as hydronium alunite. From a molar balance equation for hydronium alunite: Miy1rt y Mirt q 0.408 w Hq x T ,i Ž Si y 1 . Mi s 0, Ž A10. where Mi is the molar mass of hydronium alunite precipitated in the first i compartments per unit

D.H. RubisoÕ, V.G. Papangelakisr Hydrometallurgy 58 (2000) 89–101

volume, and Si s wAlxT,irwAlxT,S,i is the aluminium saturation ratio. The amount of the precipitate may be related to the change of aluminium concentration in the aqueous phase: Mi y Miy1 s Ž w Al x T ,0,i y w Al x T ,i . r3,

Ž A11.

where wAlx T,0,i is the concentration of Al Žassuming no precipitation in the ith compartment.. Resolving Eq. A10 with Eq. A11, one may obtain a secondorder polynomial equation for Al concentration: 0.408 w Hq x T ,it Ž w Al x T ,ir w Al x T ,S ,i y 1 .

Ž 1 q 3 Miy1r Ž wAlx T ,0,i y wAlx T ,i . . s 1.

Ž A12.

References w1x E. Krause, A. Singhal, B.C. Blakey, V.G. Papangelakis, D. Georgiou, Sulphuric acid leaching of nickeliferous laterites, in: W.C. Cooper, I. Mihaylov ŽEds.., Hydrometallurgy and Refining of Nickel and Cobalt, CIM, Montreal, PQ, 1997, pp. 441–458. w2x E. Krause, B.C. Blakey, V.G. Papangelakis, Pressure acid leaching of nickeliferous laterite ores, ALTA 1998 NickelrCobalt Pressure Leaching and Hydrometallurgy Forum, May 1998, ALTA Metallurgical Services, Perth, Australia, 1998. w3x V.G. Papangelakis, D. Georgiou, D.H. Rubisov, Control of iron during the sulphuric acid pressure leaching of limonitic laterites, in: J.E. Dutrizac, G.B. Harris ŽEds.., Iron Control in Hydrometallurgy, CIM, Montreal, PQ, 1996, pp. 263–274. w4x D. Georgiou, V.G. Papangelakis, Sulphuric acid pressure leaching of a limonitic laterite: chemistry and kinetics, Hydrometallurgy 49 Ž1998. 23–46. w5x P. Perdikis, V.G. Papangelakis, Scale formation in batch reactor under pressure acid leaching of a limonitic laterite, Canadian Metallurgical Quarterly 37 Ž5. Ž1999. 429–440. w6x J.M. Krowinkel, Kinetics of nickel dissolution during sulphuric acid pressure leaching of limonite–saprolite blends, Dipl. Eng. Thesis, University of Toronto, Canada, and Delft University of Technology, the Netherlands, 1997. w7x B.C. Blakey, V.G. Papangelakis, A study of solid–aqueous equilibria by the speciation approach in the hydronium alunite–sulphuric acid–water system at high temperatures, Metallurgical and Materials Transactions B 27 Ž1996. 555–566.

101

w8x B.C. Blakey, Synthesis and solubility of hydronium alunite under direct sulphuric acid pressure leaching conditions, MASc Thesis, University of Toronto, Canada, 1994. w9x X.H. Ye, Population balance modelling of a particulate dissolution–precipitation reaction sequence: hydronium alunite sequence, MASc Thesis, University of Toronto, Canada, 1995. w10x D.H. Rubisov, V.G Papangelakis, Sulphuric acid pressure leaching of laterites — speciation and prediction of metal solubilities ‘at temperature’, Hydrometallurgy Ž1999. submitted. w11x D.H. Rubisov, J.M. Krowinkel, V.G. Papangelakis, Sulphuric acid pressure leaching of laterites — universal kinetics of nickel dissolution for limonites and limoniticrsaprolitic blends, Hydrometallurgy Ž1999. submitted. w12x D.H. Rubisov, V.G. Papangelakis, The effect of acidity ‘at temperature’ on the morphology of precipitates during sulphuric acid pressure leaching of laterites, in: R. van Lier, G.B. Harris ŽEds.., Solid–Liquid Separation, CIM, Montreal, PQ, 1999, pp. 185–203. w13x E. Matijevic, Fundamentals of precipitations and uses of fine iron oxide particles, in: V.G. Papangelakis ŽEd.., Iron Control Short Course, in conjunction with the Second International Symposium on Iron Control, Ottawa, ON, October 19–23, 1996. w14x A.A. Adetayo, B.J Ennis, Unifying approach to modelling granule coalescence mechanisms, AIChE Journal 43 Ž1997. 927–934. w15x D. Georgiou, V.G. Papangelakis, Transmission electron microscope analysis of INCO’s lateritic samples, Unpublished report for ITSL, 1996. w16x J.A. Dirksen, T.A. Ring, Fundamentals of crystallization: kinetic effects on particle size distributions and morphology, Chemical Engineering Science 46 Ž1991. 2389–2427. w17x P.W. Trush, A Dictionary of Mining, Mineral and Related Terms, US Bureau of Mines, Washington, DC, 1968. w18x A. Taylor, Pressure acid leaching, NirCo Laterite Project Development Seminar, May 15, 1996, Alta Metallurgical Services, Perth, Australia, 1996. w19x A. Roine, HSC Chemistry for Windows 2.0., Outokumpu Research, Oy, Finland, 1994. w20x F. Habashi, Principles of Extractive Metallurgy vol. 1 Gordon & Breach, New York, 1969. w21x S. Wolfram, The Mathematica Book, 3rd edn., Wolfram Media, Champaign, ILrCambridge Univ. Press, Cambridge, England, 1996. w22x D.H. Rubisov, V.G. Papangelakis, Solution techniques for population balance equations as applied to heterogeneous aqueous processes in stirred tank reactors, Computers and Chemical Engineering 21 Ž1997. 1031–1042.