Вы находитесь на странице: 1из 7

Hydrometallurgy 79 (2005) 62 – 68

www.elsevier.com/locate/hydromet

Shrinking core models in hydrometallurgy: What students are not


being told about the pseudo-steady approximation
KNona C. Liddell
Chemical Engineering Department, Washington State University, Pullman, WA 99164-2710, USA
Received 19 August 2002; received in revised form 22 April 2003; accepted 7 July 2003
Available online 16 August 2005

Abstract

There is a significant amount of early work that has been largely forgotten that deals with the validity of the pseudo-steady
approximation for liquid–solid reactions. Much of this was originally published in the chemical engineering literature but is
highly relevant to hydrometallurgical processes. The purpose of this paper is to summarize some criteria that have been
developed by various authors for determining the error associated with use of shrinking core models for liquid–solid systems
and to provide an illustrative example using data for a mineral leaching reaction.
D 2005 Elsevier B.V. All rights reserved.

Keywords: Shrinking core models; Pseudo-steady state approximation

1. Introduction easily assimilated by beginning students and has


been emulated in newer texts (e.g., Fogler, 1992).
Studies of the fundamental aspects of shrinking Two curious facts about the early development of
core models have undergone a renaissance in the last shrinking core models have generally been over-
10–15 years, with Crundwell and Godorr (1997), looked. First, Levenspiel specifically restricts his dis-
Lapidus (1992), and Pritzker (1996) among those cussion to gas–solid systems. He briefly comments on
making important recent contributions. Their work some limitations on the validity of the familiar shrink-
adapts and extends earlier model equations to situa- ing core equations (Levenspiel, 1972, p. 373) but does
tions more complex than those that were first consid- not mention liquid–solid cases at all. The assumptions
ered. Although these authors emphasized mineral Levenspiel points out are that there must be a sharp
leaching reactions, the simple models in Levenspiel’s interface between the product layer and fresh reactant
(1972) classic text remain much more widely used by and that the system must be isothermal. The latter
hydrometallurgists. Levenspiel’s approach is also condition would virtually always be met for a leaching
process, but it is less clear that the former would be.
Microscopic examination of sections of partially
E-mail address: liddell@che.wsu.edu. reacted particles is necessary to determine whether
0304-386X/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.hydromet.2003.07.011
K.C. Liddell / Hydrometallurgy 79 (2005) 62–68 63

there is a definite product–reactant interface or a has an analytical solution that involves the error func-
diffuse reaction zone. Second, from the early 1960s tion. In making the PSS assumption, the time deriva-
through the early 1980s, a number of research papers tive is neglected, so that the partial differential
and review articles examined the validity of shrinking equation is converted into an ordinary one.
core models for liquid–solid systems, using both ana- In Levenspiel’s development of the equation relat-
lytical and numerical methods. The pseudo-steady ing particle conversion (fraction reacted) and reaction
state (PSS) approximation was the focus of this time, the movement of the interface between the
work. Results from several research groups appeared unreacted and product solids is initially assumed to
in leading chemical engineering journals but are be very slow relative to the rate species A diffuses
almost unknown today in the hydrometallurgical com- through the porous shell. This allows the concentra-
munity. One consequence is that the shrinking core tion profile across the shell to be determined from a
models have been presented to several generations of steady-state diffusion equation. This is the physical
extractive metallurgy students with relatively little meaning of the pseudo-steady state assumption. Then
attention to the range of applicability of the key PSS the motion of the interface is dealt with separately by
assumption. From an instructional viewpoint, this is equating the mass flux to the rate of disappearance of
clearly a lost opportunity to develop students’ ability the reacting solid. Elementary calculus is all that is
to make engineering approximations. required to obtain the familiar result:
The purpose of this paper is to summarize the
 2  3
relevant early papers on the PSS approximation and t Rc Rc
its use in liquid–solid shrinking core models. It is ¼13 þ2 ð2Þ
s Rp Rp
hoped that information from these publications will
become an integral part of the course materials taught Here the time required for the particle to react com-
to future hydrometallurgists and metallurgical engi- pletely is s; R c is time-dependent.
neers. The scope here is limited to shrinking core In 1968, Wen (1968) presented a tutorial review of
models of monosized particles in a large volume of shrinking core and shrinking particle models for non-
solution, i.e., effects of particle size distribution, mix- catalytic reactions. Based on a series expansion of the
ing, and reactor type are not covered. The relevant analytical solution for unsteady diffusion through a
literature citations are from 1963 to 1983. porous solid slab and comparison with predictions
from the corresponding shrinking core equation,
Wen concluded that the shrinking core model for
2. The pseudo-steady state assumption liquid–solid reactions was likely to give erroneous
results unless the fluid reactant’s concentration is
We first review the pseudo-steady assumption for very low. For accurate results, the ratio
systems controlled by product layer diffusion, then
consider the literature about it. For a spherical particle eCA; 0
surrounded by a large volume of fluid with which it aCs; 0
reacts, the differential mass balance on the particle is
as follows: would need to be small (b 0.1), but a detailed ana-
  lysis of the error for particular circumstances was not
BCA DA B 2 BCA given. In this expression, a is the stoichiometric
e ¼ 2 r Rc brbRp ð1Þ
Bt r Br Br coefficient for the liquid phase reactant, C A,0 is the
bulk concentration of the solution phase reactant, and
R c and R p are the unreacted core and particle radius, C s,0 is the initial concentration of the solid reactant.
respectively; e is the porosity of the product layer; and This method of analysis can be applied to other
C A and D A are the concentration and effective diffu- geometries but the details become somewhat more
sivity of the fluid phase reacting species A. The complicated.
porosity and effective diffusivity are assumed con- Also in 1968, Luss (1968) found upper and lower
stant. This is a well-known diffusion problem and bounds on the exact solution for Eq. (1) and compared
64 K.C. Liddell / Hydrometallurgy 79 (2005) 62–68

those with the PSS, shrinking core result. Although and the dimensionless radial position (r) and dimen-
the final conclusions are based on physical property sionless core radius (R c) expressed as
values typical of gas–solid reactions, the limits given
r ¼1x Rc ¼ 1  d ð6Þ
in Luss’ Eq. (19) are more general. The criterion in
that equation is the first-order approximation (PSS approximation) to
"  # / was shown to be
1  s ð1  sÞ2 1 Ca ðsÞ
0bdb þ 1 Max : x
Nu 2 Nu qs /1 ¼ 1  : ð7Þ
d
ð3Þ
The second-order approximation is
The error d is 0 when the shrinking core model is
accurate and in other cases is a function of the C0 n3
 An þ
dimensionless radius of the unreacted core s ¼ RRc ðpsÞ , 6aCs
/2 ¼ 1 þ : ð8Þ
the Nusselt number for mass transfer (also known as B
k R
the Sherwood number, Sh) Nu ¼ gD p , the ambient
Here C 0 and C s are fluid concentration at the surface
concentration C a (which need not be constant), and
of the particle and the solid concentration, respec-
the reactant solid density q s. Other quantities are the
tively; a is the stoichiometric coefficient; and
core and particle radii R c and R p, the mass transfer
coefficient k g, the (effective) diffusivity D, and the x C0 C0
n¼ A¼1dþ B¼1dþ :
dimensionless time s ¼ RDt2 . A noteworthy feature of d 2aCs 3aCs
p
this treatment is that the boundary conditions chosen ð9Þ
do not require any assumption about the kinetics of
the reaction. This criterion is in a form readily applic- The third-order approximation is also given in full.
able to liquid–solid reactions but is not widely Comparison with previous work, including that of
known. It is also worth noting that Luss’ criterion Bischoff (1965) (see below) indicated that the var-
can be simplified slightly if the ambient concentration ious methods were in good agreement, that there
is kept constant. Also, the ratio s approaches zero as would rarely be a need to use the third-order approx-
the reaction nears completion, leading to the further imation for a gas–solid system, and that numerical
simplification results might be subject to significant round off
  error. However, liquid–solid cases were not explicitly
1 1 Ca
0bdb þ1 : ð4Þ considered.
2 Nu qs
Some 10 years later, Lindman and Simonsson
Consistent with the common observation that shrink- (1979) noted that Luss’ criterion indicates a large
ing core models tend to fail as the conversion exceeds upper bound on the error if the reaction involves a
0.3 or 0.4, the error d increases as leaching progresses liquid reactant at high concentration. They used
at constant C a. numerical methods to solve a generalized form of
An approximate solution to unsteady transport pro- Eq. (1) in which a term was added to account for
blems in confined spatial domains was developed by convection associated with the creation of additional
Theofanous and Lim (1971). Their treatment is quite porosity during a first-order liquid–solid reaction.
general in that it applies to a rather wide class of heat Results of the computations were used to prepare a
transfer and mass transfer problems, including shrink- series of graphs of the expected error in various
ing core models. The method involves a transforma- situations. The general conclusion from this work
tion of variables and leads to integrodifferential was that the shrinking core model gives acceptable
equations from which successive approximations are accuracy even when there is a concentrated solution
obtained. With the fluid reactant’s concentration C phase reactant and high conversions are reached; the
and the new dependent variable defined by error is smaller at lower concentrations. The convec-
tive term was reported to have a significant effect at
/ ¼ Cr ð5Þ high concentrations.
K.C. Liddell / Hydrometallurgy 79 (2005) 62–68 65

Bischoff revisited the question of the validity of the Unfortunately neither the original nor the transformed
PSS assumption several times over a period of some 20 problem lends itself to an analytical solution except in
years (Bischoff, 1963, 1965; Del Borghi et al., 1976; special cases. Numerical integration would usually be
Stakgold et al., 1983). The second paper in the series needed even if the PSS approximation was invoked,
(Bischoff, 1965) is a response to criticisms of the first and the governing differential balances must first be
(Bischoff, 1963) and refines the general approach, set up in a form consistent with those of Stakgold et al.
introducing an expansion in the perturbation parameter (1983).
d ¼ CqA . For small values of the parameter, the reaction In another paper from 1983, Taylor et al. (1983)
s
interface would be expected to move slowly and the explicitly considered liquid–solid reactions. This
PSS shrinking core solution is generated. The first- work is subject to a number of assumptions: the
order correction for the position of the moving bound- reaction itself is first order and irreversible; the
ary (radius of the unreacted core) is given by solution phase reactant concentration is constant;
  and the effective diffusivity in the product layer is
d 1  Rc;0
Rc ¼ Rc;0 1 þ þN ð10Þ constant. The governing equations were put in
3 Rc;0 dimensionless form, allowing dimensionless groups
where R c,0 is the 0th order shrinking core approxima- to be identified, and then solved numerically under
tion. The error associated with the PSS approximation conditions of porous medium diffusion control. The
is small for d bb 1. dimensionless group
The most mathematically rigorous examination of
the validity of the PSS approximation was given by aCB;0
c1 ¼ ð14Þ
Stakgold et al. (1983). Their derivation considers a bqs
generalized form of Eq. (1), which includes a term for
the rate of reaction of the fluid-phase reactant; trans- was shown to determine the validity of the PSS
formation of the coupled differential mass balances approximation. Here C B,0 is the bulk concentration
for the fluid and solid leads to a single reaction- of the fluid reactant, q s is the molar density of the
diffusion equation. It was shown that application of reacting solid, and a and b are the stoichiometric
the PSS assumption leads to an error Z, which is coefficients for the solid and fluid reactants, respec-
bounded by
e  2 
0bZb Rp þ 2aRp : ð11Þ Table 1
2n Data requirements for selected criteria for the validity of the pseudo-
steady state approximation
Here e is the porosity of the product layer, n is the
Reference Data Needed
dimension (typically 3), R p is the radius of the parti-
cle, and a is a non-negative constant in the boundary Luss (1968) Density of B
Bulk concentration of A
condition of the untransformed problem that indicates
Radius of unreacted core
whether there is a film diffusion contribution. With a Particle radiusa
flux (Neumann) boundary condition, it can be shown Mass transfer coefficienta
that Effective diffusivitya
Bischoff (1965) Density of B
Deff Bulk concentration of A
a¼ ð12Þ
km Rp Stakgold et al. (1983) Porosity
Particle radius
in the expression for the upper bound on Z. Sometimes Diffusivityb
a Dirichlet condition may be encountered instead, i.e., Mass transfer coefficientb
Taylor et al. (1983) Stoichiometric coefficient
the surface concentration is a known constant. In this
Density of B
case Bulk concentration of A
a
Sherwood number is Sh = (k mr p/D eff).
a ¼ 0: ð13Þ b
Needed with flux boundary condition at particle surface.
66 K.C. Liddell / Hydrometallurgy 79 (2005) 62–68

tively. Several graphs were constructed to permit Ti Leached from FeTiO3


quick evaluation of the approximation’s validity, 0.6
and it was concluded that pseudo-steady state Expt'l Data
could be safely assumed if 0.5

Conversion (X)
0.4
c1 V0:1: ð15Þ
0.3

This criterion is easily applied and fairly general, but


0.2
despite these advantages it appears to be rarely used
by hydrometallurgists. 0.1
Table 1 summarizes the data requirements of sev-
eral of the above criteria. Following Levenspiel 0.0
0 10 20 30 40
(1972), the reaction is represented Time (h)

Fig. 1. Fraction Ti leached from +35 to 45 Am FeTiO3 (trace impu-


AðfluidÞ þ bBðsolidÞYproducts: ð16Þ rities) in 3 M HCl, at 87 8C in a batch reactor stirred at 700 rpm.

Spherical particles are assumed.


The criteria of Bischoff (1965) and Taylor et al. btighterQ of the two, with d b 1, while the other is
(1983) are very simple and easily applied. That of c 1 b 0.1. One might therefore expect that a shrinking
Stakgold et al. (1983) will not be useful unless the core model for spherical particles would fit the data,
porosity of the particles has been determined or can but the R 2 value might be rather low.
be estimated; the surface concentration of the fluid Four shrinking core models were fit to this set of
reactant may also be needed, depending on the data:
boundary condition assumed. Similarly, Luss’s
(1968) error criterion requires knowledge of the Model I, spherical particles under reaction control:
effective diffusivity (or Nu), a quantity that may be
unknown for a particle in a concentrated electrolyte t~1  ð1  X Þ1=3 : ð18Þ
solution.
Model II, spherical particles under product layer
diffusion control:
3. Examples t~1  3ð1  X Þ2=3 þ 2ð1  X Þ: ð19Þ

For illustrative purposes, we now consider the Model III, cylindrical particles under reaction
leaching of ilmenite (FeTiO3) by HCl (Orth, 1985): control:

t~1  ð1  X Þ1=2 : ð20Þ


2+ 2+ 
FeTiO3+4HClYFe +TiO +2H2O+4Cl (17) Model IV, cylindrical particles under product
layer diffusion control:
Ti solution concentration data from a typical t~X þ ð1  X Þlnð1  X Þ: ð21Þ
experiment (3 M HCl, 87 8C, 700 rpm, +35 to 45
Am FeTiO3 with trace Fe3O4 and Fe2O3) in a stirred Here X is conversion and t is time. The results are
batch reactor are shown in Fig. 1, presented as con- shown in Figs. 2–5. All four models fit the data very
version (X) versus time. Taking an ilmenite specific well, and standard statistical methods would not
gravity of 4.7, Bischoff’s criterion is d = 0.0965, sug- cause any of them to be ruled out. Indeed, not a
gesting that the conditions for the PSS approximation single data point lies outside the 95% prediction
are not satisfied; Taylor’s criterion is c 1 = 0.024, con- interval on any of the graphs, even for models that
sistent with validity. The Bischoff criterion is the assume cylindrical particles. (Models III and IV
K.C. Liddell / Hydrometallurgy 79 (2005) 62–68 67

Shrinking Core or Particle, Sphere, Reaction Control Shrinking Core, Cylinder, Reaction Control
0.30 0.5
Expt'l Data
0.25 R 2 = 0.9921 Expt'l Data
95% Prediction Interval 0.4 R2 = 0.9905
95% Prediction Interval
0.20
1 - (1-X)1/3

1 - (1-X)1/2
0.3
0.15

0.10 0.2

0.05
0.1

0.00
0 10 20 30 40 0.0
Time (h) 0 10 20 30 40
Time (h)
Fig. 2. Ti leach data fitted to Model I, shrinking core in spherical
particles under reaction control. Fig. 4. Ti leach data fitted to Model III, shrinking core in cylindrical
particles under reaction control.
neglect the ends of the cylinder, so that it is effec- transport within a reacting particle. As a result, in some
tively infinitely long; diffusion and reaction occur in instances topochemical models fail to fit experimental
the radial direction only.) For the ilmenite–HCl reac- data even when criteria for PSS validity are satisfied.
tion, these commonly used shrinking core equations One such case is chrysocolla leached by NH4Cl/NH3
are insensitive to both leaching mechanism and par- (Raghavan and Gajam, 1986); the fibrous structure and
ticle shape! The ability of so many different models high microporosity of the mineral were believed to be
to give excellent fits to these data results from the responsible, and a structurally based, three-parameter
fact that values of X cover a limited range (0 V X V 1) model was more successful in representing the leaching
to begin with, and appear in functions (variously behavior.
ln(1  X), (1  X)1/2, (1  X)1/3, and (1  X)2/3) that Examples of these kinds have a variety of uses in an
flatten the range of values even further. instructional setting. Perhaps most importantly, stu-
It is also important to note that besides explicitly dents can see for themselves that there is no uniquely
neglecting the time derivative, shrinking core models right answer in fitting leaching data, and no uniquely
implicitly assume that a single lumped parameter, the correct mechanism. Additional complications that
effective diffusivity, adequately describes the mass might be covered with an appropriate reaction include

Shrinking Core, Sphere, "Ash" Diffusion Control Shrinking Core, Cylinder, "Ash" Diffusion Control
0.16 0.25
0.14 Expt'l Data Expt'l Data
R2 = 0.9626 0.20 R 2 = 0.9656
1 - 3(1-X)2/3 + 2(1-X)

0.12 95% Prediction Interval


X + (1-X) ln(1-X)

95% Prediction Interval

0.10
0.15
0.08

0.06 0.10

0.04
0.05
0.02

0.00 0.00
0 10 20 30 40 0 10 20 30 40
Time (h) Time (h)

Fig. 3. Ti leach data fitted to Model II, shrinking core in spherical Fig. 5. Ti leach data fitted to Model IV, shrinking core in cylindrical
particles under product layer (bashQ layer) diffusion control. particles under product layer (bashQ layer) diffusion control.
68 K.C. Liddell / Hydrometallurgy 79 (2005) 62–68

extended leaching (X Y 1), and wide or bimodal parti- Del Borghi, M., Dunn, J.C., Bischoff, K.B., 1976. A technique for
cle size distributions. The use (and misuse) of statistical solution of the equations for fluid–solid reactions with diffusion.
Chem. Eng. Sci. 31, 1065 – 1069.
methods can be illustrated with a varying amount of Fogler, H. Scott, 1992. Elements of Chemical Reaction Engi-
detail, depending on the objectives and level of the neering, Second ed. Prentice-Hall, Englewood Cliffs, NJ,
course. The derivation of the shrinking particle- pp. 582 – 591.
shrinking core family of models can be covered at Lapidus, Gretchen, 1992. Mathematical modelling of metal leaching
in nonporous minerals. Chem. Eng. Sci. 47, 1933 – 1941.
the undergraduate level using a treatment similar to
Levenspiel, Octave, 1972. Chemical Reaction Engineering, Second
Levenspiel’s (1972), and then revisited more rigor- ed. John Wiley & Sons, Inc., New York, pp. 361 – 373.
ously in graduate transport and reactor design Lindman, Nils, Simonsson, Daniel, 1979. On the application of the
courses. In both cases, the consequences of the shrinking core model to liquid–solid reactions. Chem. Eng. Sci.
pseudo-steady state assumption can be explored 34, 31 – 35.
using experimental data on mineral leaching, some- Luss, Dan, 1968. On the pseudo steady state approximation for gas
solid reactions. Can. J. Chem. Eng. 46, 154 – 156.
times with surprising results. R.J. Orth, M.S. Thesis, Washington State University (1985).
In summary, there is a fairly extensive body of Pritzker, Mark D., 1996. Shrinking-core model for systems with
literature, largely unknown to hydrometallurgists, on facile heterogeneous and homogeneous reactions. Chem. Eng.
the validity of shrinking core models for liquid–solid Sci. 51, 3631 – 3645.
reactions. These papers provide some simple criteria Raghavan, S., Gajam, S.Y., 1986. Application of an enlarging pore
model for the ammoniacal leaching of chrysocolla. Hydrome-
that could be taught to beginning students, and the tallurgy 16, 271 – 281.
more mathematically sophisticated approaches should Stakgold, I., Bischoff, K.B., Gokhale, V.V., 1983. Validity of
be of interest to those teaching advanced courses. the pseudo-steady-state approximation. Int. J. Eng. Sci. 21,
537 – 542.
Taylor, P.R., de Matos, M., Martins, G.P., 1983. Modeling of
noncatalytic fluid–solid reactions: the quasi-steady state
References assumption. Metall. Trans., B, Process Metall. 14B, 49 – 53.
Theofanous, T.G., Lim, H.C., 1971. An approximate analytical
Bischoff, K.B., 1963. Accuracy of the pseudo steady state approx- solution for non-planar moving boundary problems. Chem.
imation for moving boundary diffusion problems. Chem. Eng. Eng. Sci. 26, 1297 – 1300.
Sci. 18, 711 – 713. Wen, C.Y., 1968. Noncatalytic heterogeneous solid fluid reaction
Bischoff, K.B., 1965. Further comments on the pseudo steady state models. Ind. Eng. Chem. 60 (9), 34 – 54.
approximation for moving boundary diffusion problems. Chem.
Eng. Sci. 20, 783 – 784.
Crundwell, Frank K., Godorr, Sven A., 1997. A mathematical
model of the leaching of gold in cyanide solutions. Hydrome-
tallurgy 44, 147 – 162.

Вам также может понравиться