Você está na página 1de 12

Chemical Engineering Science 91 (2013) 90–101

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Science


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

Particle-fluid mass transfer in multiparticle systems at low


Reynolds numbers
Fabrizio Scala n
Istituto di Ricerche sulla Combustione—CNR, Piazzale Tecchio 80, 80125 Napoli, Italy

H I G H L I G H T S

c Mass transfer in multiparticle systems is theoretically examined.


c Different Sh can be defined depending on the choice of the reference driving force.
c Effective and local Sh are derived for the general case and for asymptotic limits.
c Allowance is given for variation of voidage and volume fraction of active particles.
c The use of the local Sh is strongly suggested.

a r t i c l e i n f o a b s t r a c t

Article history: The problem of mass transfer between active solid particles and a fluid in multiparticle systems is
Received 9 August 2012 examined with a focus on the stagnant and the low Reynolds number cases. This problem has attracted
Received in revised form significant attention with regard to operation of fixed and fluidized beds. It is recognized that different
19 December 2012
Sherwood numbers can be defined depending on the choice of the reference concentration difference
Accepted 8 January 2013
Available online 16 January 2013
(driving force). An effective Sh has often been used to analyze experimental data based on an overall
concentration difference across the bed. A local Sh can also be introduced based on a concentration
Keywords: difference close to the active particle. However, the use of these two different Sherwood numbers
Mass transfer implies a different implementation of the mass balance equations.
Sherwood number
The mass balance equations are here analytically solved both under stagnant and non-stagnant
Multiphase flow
conditions in a multiparticle system under suitable simplifying assumptions. Equations for the effective
Diffusion
Packed bed and local Sherwood numbers are derived for the general case and for the asymptotic limits. Allowance
Fluidized bed is given for the variation of bed voidage and volume fraction of active particles in the bed. It is shown
that the local Sh only depends on geometrical and fluid-dynamics considerations and accordingly has a
general validity. On the contrary, the effective Sh also depends on the assumptions made in deriving the
mass balance equations across the bed (e.g., fluid plug flow). The use of the local Sh is therefore
suggested.
Results show that for Re-0 the limiting value of the local Sh is always a finite number, while the
effective Sh tends to zero linearly with Re. It is shown that this result is simply a consequence of the
plug flow assumption made in the bed mass balance. The general expressions derived here compare
very well to experimental trends for the cases of large Reynolds numbers and of few isolated active
spheres immersed in a bed of inert particles, where most of the reported experimental data gather.
Unfortunately, for the most controversial case of very low Re in beds made entirely of active particles no
reliable data appears to be available to check the accuracy of the expressions.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction often expressed as a function of the Reynolds (Re) and Schmidt


(Sc) numbers using a Frössling-type equation:
Mass transfer data in fixed and fluidized beds are usually
Sh ¼ kþ b  Rec  Scd ð1Þ
correlated in terms of the Sherwood number (Sh), which is most
where k, b, c, d are parameters to be estimated in each specific
case. This expression was derived by dimensional analysis and
n
Tel.: þ39 81 7682 969; fax: þ 39 81 5936 936. boundary layer theory (Frössling, 1938; Ranz and Marshall, 1952),
E-mail address: scala@irc.cnr.it and consists of the sum of two contributions. The first term on the

0009-2509/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ces.2013.01.012
F. Scala / Chemical Engineering Science 91 (2013) 90–101 91

right hand side (k) represents mass transfer in stagnant condi- of the model was that at low Reynolds numbers the predicted
tions (diffusive term), while the second one accounts for the values of Sh were much lower than two and approached zero at
enhancement of mass transfer caused by the fluid flow around the Re-0. This model was however questioned by Schlünder (1977),
particle (convective term). In the limit Re-0, the second term Sohn (1977) and Glicksman and Joos (1980), who pointed out that
vanishes and the diffusive term prevails. Representative values of the penetration theory is not appropriate al low Reynolds num-
the parameters (k, b, c, d) will be discussed later on. bers, and that the boundary conditions used were incorrect
The minimum possible rate of mass transfer from an isolated (in fact, setting the concentration derivative equal to zero at the
active spherical particle in a stagnant fluid is well known to shell surface is in contradiction to having a sink in this position).
correspond to a Sherwood number of two (Bird et al., 2002). With The latter authors claimed that accounting for channeling and
active particle we mean a particle that is exchanging mass with axial diffusion would explain the low apparent Sherwood num-
the fluid phase, because either a chemical reaction or a physical bers found experimentally (Glicksman and Joos, 1980). They also
process (phase change) is taking place. When the single active showed that with an appropriate definition of the mass transfer
particle is surrounded by a bed of inert particles this limiting driving force, based on the local concentration gradient close to
value should be corrected for the effect on diffusion of the bed the particle, the Sherwood number value would become always
tortuosity and of the decrease of the available fluid volume larger than two. A similar approach was also used by Sohn (1977).
(Scala, 2007). This situation corresponds to many cases of prac- These models will be discussed in more detail later on.
tical interest, e.g., the combustion of char particles in a fluidized The above literature review shows that the controversy on the
bed boiler. Another interesting practical situation is when the bed theoretical lowest possible value of Sh is still not solved, as
is composed entirely of active particles (e.g., in fluid catalytic witnessed by recent literature reviews (Kunii and Levenspiel,
cracking (Kunii and Levenspiel, 1991), or in chemical looping 1991; Ho, 2003), and a generalized reliable expression for Sh in
combustion (Chuang et al., 2009)). Many experimental works multiparticle systems is not available. In addition, the definition
conducted in such a situation reported limiting Sherwood num- of the correct mass transfer driving force to be used for the
bers at Re-0 much lower than two (Dwivedi and Upadhyay, Sherwood number calculation appears not to be generally realized.
1977; Kunii and Levenspiel, 1991). This result has triggered a Finally, it is noted that all the works reported in the literature
dispute on the explanation of this result and on the theoretical examine the two limiting cases (either a bed composed entirely of
lowest possible value of Sh in the low Reynolds number regime. active particles, or one or few active particles surrounded by a bed of
Some authors argued that these very low Sh values were inert particles), while intermediate situations where the active and
the result of inaccuracies or artifacts in the experiments inert bed particles are mixed together in variable amounts has not
(Van Heerden, 1952; Schlünder, 1977; Glicksman and Joos, 1980). been treated so far. This paper tries to answer all these questions, by
Bed channeling, gas bypass in bubbles, backmixing, particle agglom- analyzing in detail the mass transfer problem in multiparticle
eration, inaccuracy in the concentration measurement, too simple systems under stagnant and low Reynolds number conditions, with
reactor model, and erroneous driving force definition were some of a particular focus to gas–solid systems. On the basis of the following
the possible explanations proposed. These arguments were also analysis a generalized analytical expression for the local and the
supported by the fact that it is very difficult to carry out reliable effective Sherwood numbers is proposed, which should be more
experiments in such systems, since the diffusing species concentra- useful than available numerical procedures (e.g., Sørensen and
tion gradient extends along bed lengths of only few particle Stewart, 1974; Glicksman and Joos, 1980).
diameters in gas–solid systems. A dilution method was in fact
proposed to overcome this difficulty, consisting in mixing of the
active particles with inert ones (Van Heerden, 1952). However, 2. Mass transfer in stagnant conditions (Re¼0)
increasing the bed dilution by mixing with inert particles simply
shifts the problem from the second case (bed composed entirely by 2.1. Theory
active particles) to the first case (one or few active particles
surrounded by a bed of inert particles). The first point to be discussed is the definition of the correct
Cornish (1965) first pointed out that, from a theoretical point mass transfer driving force to be used for the calculation of the
of view, the presence of other active particles around the active Sherwood number. The typical choice is the difference between
particle under consideration affects the concentration profiles and the concentration at the particle surface and a reference ‘‘bulk’’
may account for the apparent low Sh values found experimen- fluid concentration (the sink), whose location most often coin-
tally. His reasoning was based on the analogy between heat/mass cides with the outer edge of the concentration boundary layer
transfer and the field of electrostatics, where a number of workers around the particle. But what about the situation of a stagnant
presented solutions to the problem of charges on two or more fluid? In this case the most reasonable choice is to consider this
spherical condensers whose surfaces are at the same potential. point to be infinitely far from the particle, where the fluid sink is
Cornish did not derive any equation to predict Sh, but using the located. This is also consistent with the well known result of the
above analogy he argued that when the active sphere is entirely boundary layer theory that for Re-0 the boundary layer becomes
surrounded by other active spheres in stagnant conditions Sh infinitely thick. This choice corresponds to calculating an overall
should approach zero. Nelson and Galloway (1975) started from effective bed Sherwood number (Sheff). A second possible choice is
this concept and developed a mass transfer model based on the to use the difference between the concentration at the particle
penetration theory where, instead of applying the boundary surface and the concentration at a fixed distance from the surface.
condition for the ‘‘sink’’ infinitely far from the sphere (relevant In this case the most reasonable choice is to select this point as
for an isolated active sphere), they forced a stationary solution by half the average distance between two adjacent active particles.
imposing the boundary condition for the sink at a position where This choice corresponds to calculating a local particle Sherwood
the concentration gradient vanishes symmetrically between the number (Shn). It is noted that in this second case the concentra-
particles (on a surface assumed to be a concentric spherical shell). tion difference is evaluated along a distance which is not
This approach requires that an external flow is present in order to dependent on the boundary layer thickness, but only on the
provide the sink between the active particles, and clearly is not geometrical position of the active particles. Clearly, the value of
applicable under stagnant conditions where no sink exists the Sherwood number strongly depends on this choice of the
between the active particles. However, an interesting outcome driving force.
92 F. Scala / Chemical Engineering Science 91 (2013) 90–101

Let us first consider the second choice (Shn). A spherical active In each cell the local stationary mass balance on the diffusing
particle (AP) with radius RAP is surrounded by an isothermal species can be written:
infinite bed of particles. The bed is made of a mixture of inert and  
d dC n
active particles. All the active bed particles are spheres with r2 ¼0 ð4Þ
dr dr
radius RAP, while all the inert bed particles are spheres with
radius RIP. We use a reference system in radial coordinates with with the following boundary conditions:
the origin in the center of AP. The volume fraction of active bed
r ¼ RAP ) C n ¼ C AP
particles (with respect to the total volume occupied by inert and
active solids) is XAB. It is assumed that the active particles are r ¼ RC ) C n ¼ C ð5Þ
evenly distributed in the bed volume. A species which is either n
The second conditions imposes the local concentration (C ) at
released or captured by the active particles diffuses through a the cell boundary to be equal to the interstitial average bulk
stagnant fluid in the interstices between the bed particles with an
concentration (C).
effective diffusivity Def f ¼ ðebed =tÞDm , where Dm is the molecular Eq. (4) with the boundary conditions (5) can be easily solved to
diffusivity of the diffusing species in the fluid mixture, ebed the
yield:
bed voidage and t the bed tortuosity. The concentration of the  
diffusing species is assumed to be very low so that the Stefan flow C n C 1RC =r
¼  ð6Þ
can be neglected. The number of active bed particles per unit bed C AP C 1RC =RAP
volume is:
The flux of the diffusing species at the active bed particle
surface is:
3 ð1ebed Þ  X AB 
N¼  ð2Þ dC n  Def f ðC AP C Þ
4 p  R3AP Def f ¼ 2   ð7Þ
dr AP RAP 1=RAP 1=RC

It is supposed that the unit bed volume can be divided into N Given the definition of the local active bed particle mass
non-overlapping cells (assumed spherical) with an active bed transfer coefficient (Kn), we can write:

particle at their center (Fig. 1). This is a simplifying assumption Def f dC n  Def f 1
Kn ¼  ¼ 2   ð8Þ
implying an approximation in the division of the whole bed ðC AP C Þ dr AP RAP 1=R AP 1=RC
domain into a number of non-overlapping spherical cells. Clearly,
from a geometrical point of view the bed cannot be exactly Finally, the local bed particle Sherwood number is given by:
divided into a number of non-overlapping spherical cells. It is 2  RAP  K nAP 2  ebed =t
n
here assumed that this approximation is not far from reality. Sh ¼ ¼  ð9Þ
Dm 1RAP =RC
A similar approach has been adopted by several other authors in
the past (e.g., Pfeffer, 1964; Nelson and Galloway, 1975). Each Substituting Eq. (3) into Eq. (9) gives:
active bed particle has a volume of bed equal to 1/N in its cell. The 2  ebed =t
n
radius of each cell is: Sh ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð10Þ
1 3 ð1ebed Þ  X AB
RAP The term:
RC ¼ p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3Þ
3
ð1ebed Þ  X AB 1
yn ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð11Þ
1 3 ð1ebed Þ  X AB
represents the enhancement factor of the local Sherwood number
INERT PARTICLES
due to the presence of other active particles in the surroundings.
n
ACTIVE PARTICLES Obviously, for XAB ¼0 or ebed ¼1 it is y ¼ 1.
It is here underlined that this approach is approximate, since it
assumes that each spherical cell does not overlap with the
surrounding ones and with the interstitial diffusing flux. This is
obviously not true, since in a stagnant fluid the diffusing fluxes
generated by the different active particles will all interact among
each other determining a complex concentration field in the
domain. The simplifying assumption underlying the present
approach is that the influence of the surrounding cells on the
cell under consideration can be entirely concentrated at the cell
RAP outer boundary (RC) where an average diffusing species concen-
RC tration C is considered. In general, this average concentration can
be calculated once a flow model for the fluid in the bed is
assumed. For the stagnant case examined here the derivation is
detailed immediately below.
Let us now consider the first choice (Sheff). We divide the bed
into two zones: a first (finite) zone close to AP made of a mixture
of inert and active particles; a second (infinite) zone made of inert
particles. For simplicity we consider a system which can be
treated in radial coordinates (Fig. 2) with the origin in the
center of AP. The first (active bed) zone extends in the range
RAP or oRAB, and the volume fraction of active bed particles in
this zone is XAB. The second (inert bed) zone extends in the range
Fig. 1. Bed schematization for local mass transfer around one active particle. RAB or oN.
F. Scala / Chemical Engineering Science 91 (2013) 90–101 93

INERT PARTICLES The boundary conditions are:


ACTIVE PARTICLES
r ¼ RAP ) C ¼ C AP

dC  ðC AB C 1 Þ
r ¼ RAB )   ¼ ð17Þ
dr R
AB AB

The second boundary condition was written using Eq. (15) and
imposes the equality of the inlet and outlet fluxes of the diffusing
species across the boundary between the two bed zones.
This diffusion problem can be solved by operating the follow-
RBP ing change of variable: u ¼ ðC AP C Þ  r. In this way Eq. (16)
becomes:
RAP
2
RAB d u a  Kn
¼ u ð18Þ
dr 2 Def f

The new boundary conditions read:


r ¼ RAP ) u ¼ 0

du
r ¼ RAB )  ¼ ðC AP C 1 Þ ð19Þ
dr AB

Solution of Eq. (18) with boundary conditions (19) gives the


following result (after changing the variable again to the original
one):
  
C AP C 1 sinh a r=RAP 1
¼     ð20Þ
Fig. 2. Bed schematization for mass transfer under stagnant conditions. C AP C 1 a  r=RAP cosh a RAB =RAP 1
qffiffiffiffiffiffiffiffinffi
where a ¼ RAP aK Def f .
In the external inert bed zone (RAB or oN) the stationary
The flux of the diffusing species at the AP surface is:
mass balance on the diffusing species can be written: 
  dC  ðC C 1 Þ 1
d dC Def f  ¼ Def f AP     ð21Þ
r2 ¼0 ð12Þ dr AP RAP cosh a RAB =RAP 1
dr dr
Given the definition of the effective mass transfer coefficient
with the following boundary conditions:
(similar to that applicable to the case of an isolated sphere), we
r ¼ RAB ) C ¼ C AB can write:
r ¼ 1 ) C ¼ C1 ð13Þ 
Def f dC  Def f 1
K ef f ¼  ¼     ð22Þ
It is worth to note that a stationary solution to this equation ðC AP C 1 Þ dr AP RAP cosh a RAB =RAP 1
exists only because the bed extends to infinite, and so acts as an Finally, the effective particle Sherwood number is given by:
infinite capacity sink for the diffusing gaseous species.
Eq. (12) with the boundary conditions (13) can be easily solved 2  RAP  K ef f 2  ebed 1
Shef f ¼ ¼     ð23Þ
to yield: Dm t cosh a RAB =RAP 1
n
CC 1 RAB Noting that K n ¼ Sh2R Dm
AP
, we can express the quantity a in
¼ ð14Þ
C AB C 1 r Eq. (23) as a function of the local particle Sherwood number Shn:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
The flux of the diffusing species at the boundary between the RAP  a  Sh
n

two bed zones (Fig. 2) is: a¼ ð24Þ


2  ebed =t

dC  ðC C 1 Þ
Def f ¼ Def f AB ð15Þ In a bed of uniformly sized active spheres with radius RAP the
dr AB RAB specific active surface of the bed is:
Let us now consider the internal active bed zone (RAP or - 3  ð1ebed Þ  X AB
oRAB). In this zone the stationary mass balance on the diffusing a¼ ð25Þ
RAP
species can be written:
  We have previously found that the local particle Sherwood
1 d dC number (in stagnant conditions) can be expressed by Eq. (10).
Def f 2 r2 ¼ a  K n ðC AP C Þ ð16Þ
r dr dr Combining Eqs. (10) and (25) with Eq. (24) we obtain:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where a is the specific surface of the active bed particles, and Kn is 3  ð1ebed Þ  X AB
the local mass transfer coefficient between the active bed parti- a¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð26Þ
1 3 ð1ebed Þ  X AB
cles and the interstitial fluid. It is here assumed that a fixed
concentration CAP of the gaseous species establishes on the sur- Examination of Eqs. (23) and (26) shows that the effective
face of all the active particles (e.g., the equilibrium concentration Sherwood number (apart from the bed voidage and tortuosity) is
for a phase change process or for a mass transfer limited chemical a function of the geometrical quantity RAB =RAP , and of the active
reaction). In addition, it must be underlined that in this situation bed particles volume fraction XAB. Interestingly, Sheff is not a direct
C represents an average concentration of the diffusing species on function of the inert bed particle size.
a spherical shell of radius r. This is because in reality the presence It is worth now to examine the two limiting cases where
of other active particles produces concentration gradients on each RAB =RAP ¼ 1 and RAB =RAP -1, respectively. The first case repre-
spherical shell, which overlap with (and of course affect) the sents the case of an isolated active particle dispersed in a bed
concentration gradient along the radial direction. made entirely of inert particles. Equivalently, this case also occurs
94 F. Scala / Chemical Engineering Science 91 (2013) 90–101

when XAB ¼0. It is easy to show that in this situation Eq. (23)
reduces to:
2  ebed
Shef f ¼ ð27Þ
t
This value represents the well known solution of the mass
transfer problem in a stagnant fluid around an isolated sphere
(giving Sh ¼2), corrected for the influence of the inert bed
particles. The inert bed is responsible for two different effects
(Scala, 2007): first, the bed particles decrease the fluid volume
available for mass transfer; second, diffusion is affected by the
hindering effect of the bed tortuosity.
The second limiting case (RAB =RAP -1) represents the case of an
active particle dispersed in an infinite active bed, with the particular
situation XAB ¼1 where the bed is entirely composed of other active
particles. In this case (whatever the value of XAB 40) the only
stationary solution to Eq. (23) is the trivial one Sheff ¼0. This means
that in the case of an infinite active bed there is not a theoretical
lower limiting value for the effective Sherwood number.
In contrast, it can be noted that if the Sherwood number is Fig. 4. Local Sherwood number under stagnant conditions as a function of the
defined on a local basis, according to Eq. (10) its value does not active particle volume fraction for different values of the bed voidage.

depend on RAB =RAP . Whatever the active bed thickness, Shn in


stagnant conditions assumes a finite value. Again, for XAB ¼0, it is: more than three active particle radii is sufficient to determine
n 2  ebed Sheff E0. This means that in all situations of practical interest the
Sh ¼ ð28Þ limiting effective Sherwood number can be assumed to be zero.
t
As regards the intermediate situations (0 oXAB o1), Fig. 3
shows that XAB exerts a strong influence on the limiting effective
2.2. Results Sherwood number. Values of XAB 40.01 are necessary to induce a
significant decrease of Sheff. Active bed particles volume fractions
Since in practical situations the bed is obviously not infinite, it XAB 40.1 are always associated to Sheff E0 for beds deeper than
is interesting to examine the effect of the ratio between the active 15–20 active particle radii.
bed depth and the active particle radius (RAB =RAP ) on the limiting Fig. 4 reports the local particle Sherwood number calculated
effective Sherwood number. Eq. (23) was solved using the with Eq. (10) as a function of XAB at different values of the bed
following typical values for the bed parameters: ebed ¼0.45; voidage. For all parameter values Shn assumes always finite
t ¼1.2. It is here noted that the range of variability of these two values. The figure shows that Shn increases with the bed voidage
parameters in fixed beds and dense fluidized beds is rather at low XAB values, while the opposite occurs at high XAB values.
limited (0.4 o ebed o0.5; 1.1 o t o1.6). For ebed ¼1 a constant value of 2 is obviously found. Interestingly,
Fig. 3 shows the calculated effective Sherwood number as a values of Shn at XAB ¼ 1 for dense beds (ebed ¼0.45) are very close to
function of RAB =RAP for different values of XAB. As explained before, those numerically evaluated by Sørensen and Stewart (1974) for a
the horizontal line corresponding to XAB ¼0 represents the case of cubic array of spheres ( E4).
an isolated active particle dispersed in a bed made entirely of
inert particles, giving in this case Sheff ¼0.75. This is also the
limiting value for all the curves at RAB =RAP ¼ 1.
Let us examine the case of a bed made entirely of active 3. Mass transfer at low Reynolds numbers
particles (XAB ¼1). It is easy to see that in this case a bed length of
3.1. Statement of the problem and critical analysis of available
models

Let us examine the case in which an external fluid flow is


imposed across the bed. When an isolated active sphere is
immersed in a bed of inert particles the situation is similar to
that of an isolated sphere in a free fluid flow. As Re increases, the
thickness of the boundary layer decreases, and it is always
possible to find a point outside the boundary layer where a
constant bulk concentration establishes (at each distance along
the direction of the fluid flow). So in this case it is easy to consider
as the driving force the difference between the concentration at
the particle surface and the bulk concentration outside the
boundary layer.
When the active particle is immersed in a bed containing other
active particles the problem gets more intriguing. In this situation
we can identify two different characteristic lengths. The first is
the theoretical concentration boundary layer thickness (to be
calculated as if no other active particle was present). The other is
Fig. 3. Effective Sherwood number under stagnant conditions as a function of the half the average distance between two adjacent active particles.
active bed radius for different active particle volume fractions. ebed ¼0.45; t ¼ 1.2. The first quantity can be calculated from the boundary layer
F. Scala / Chemical Engineering Science 91 (2013) 90–101 95

theory to be (Hayhurst, 2000): the concentration gradient is assumed to establish in a fixed space
shell between the particle radius and dID . As a consequence,
2  RAP
dBL ¼  c ð29Þ thinner concentration boundary layers (relevant for larger Re)
b  Re=ebed  Scd are not allowed. This paper, however, has the merit to highlight
  the importance of the selection of the reference concentration
where Sc ¼ m= r  Dm and Re ¼ 2  RAP  U  r=m. In Eq. (29) the
Reynolds number was divided by the bed voidage, since the gradient on the Sherwood number value.
boundary layer thickness depends on the real gas velocity around Sørensen and Stewart (1974) first recognized that the Sherwood
the particle rather than on the superficial velocity. The second (or Nusselt) number in a packed bed can be calculated in two
quantity can be calculated from Eq. (3) to be: different ways, depending on the definition of the fluid bulk
" # concentration (temperature). The difference in the two definitions
1 depends on if axial diffusion (conduction) is considered or not in the
dID ¼ RAP  p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 ð30Þ mass (heat) balance along the bed. These authors solved numerically
3
ð1ebed Þ  X AB
the three dimensional mass (heat) transfer problem when a slow
Once the properties of the different solid and gaseous phases fluid flow passes through a cubic array of spheres. Noteworthy,
have been defined the first quantity is solely a function of the analytical expressions of the asymptote for Re-0 showed that
fluid velocity, while the second quantity is fixed. Now we can depending on the definition, Sh (Nu) tended either to a constant
conceptually imagine two different situations. For relatively large finite value or to zero (linearly with Re).
values of the Reynolds number it is dBL o dID . This means that the Glicksman and Joos (1980) pursued further the numerical
concentration boundary around the active particle is not affected analysis by considering the two possible approaches for mass
by the neighboring active particles, and that the mass transfer transfer in a packed bed: one consistent with an effective Sh based
problem around the active particle can be treated as if the particle on the overall bed behavior, and the other consistent with a
were an isolated active sphere. A bulk interstitial fluid flow can be particle Sh based on local mass transfer rates and concentration
individuated with a constant concentration at each distance along differences. The authors noted that in practice most of the
the direction of the fluid flow. So again the driving force to be previous researchers did calculate the mass transfer coefficient
used is the difference between the concentration at the particle from experimental data using a simple plug flow model for the
surface and the bulk concentration outside the boundary layer. bed. A second possible approach would be that of calculating the
The difference between this case and the previous one obviously mass transfer coefficient using a more detailed model that
relies in the different specific active surface available in the bed. accounts for axial diffusion along the bed. By comparing the
When the Reynolds number is so low that dBL Z dID the two different approaches, the authors found that the effective bed
problem gets much more complicated. In this case the concentra- Sh was well below the particle local Sh, and this result was mostly
tion boundary layers of the neighboring active particles overlap, attributed to axial diffusion. However, the authors made an error
and a bulk interstitial fluid flow with a constant concentration in the application of the model. They considered that the particle
does not exist anymore. This is exactly the situation that has local Sh can be estimated by using Ranz and Marshall (1952)
attracted the attention of previous researchers. However, the Frössling-type correlation. This is obviously not correct, since this
problem has not always been faced correctly. Let us critically correlation already accounts for diffusion in stagnant conditions,
examine the available theoretical work. and inserting such a correlation in a bed model with axial
Pfeffer (1964) developed a model for mass transfer in multi- diffusion would mean to account for diffusion two times. In other
particle systems valid for low Reynolds numbers and high Peclet words, at Re-0 the Ranz and Marshall correlation predicts a
numbers. This model was based on the assumption that a boundary layer thickness dBL -1 (whatever the value of XAB)
concentration thin boundary layer solution exists. This assump- which is not consistent with the definition of the local Sh.
tion may be reasonable for some liquid–solid systems at not too From the above literature review it is clear that the Sherwood
low Re, but it is certainly not applicable to gas–solid systems number value depends on the selection of the reference concen-
where Peclet numbers are much lower. As detailed before, for tration driving force. So one could decide to use either an overall
Re-0 the boundary layer thickness extends beyond dID and or a local Sherwood number, but the correlations to be used are
cannot be considered to be thin. So this model cannot be used different. In our opinion, however, the use of a local Sherwood
to study the situation of our interest. number is a more reasonable choice, as will be discussed later on.
As written in Section 1, Nelson and Galloway (1975) developed In the following we will examine this problem theoretically in
a mass transfer model based on the penetration theory where detail.
they forced a stationary solution by imposing the boundary
condition on a surface assumed to be a concentric spherical shell 3.2. Theory
around the spherical particle. This approach can be questioned for
two different aspects. First, the penetration theory appears not to Let us consider a spherical active particle (AP) with radius RAP
be an appropriate choice for stagnant or low Re conditions, since which is surrounded by a bed of particles through which a fluid
it requires a continuous fluid renewal by a constant concentration flows in the direction z with a superficial velocity U. The system is
bulk stream, which is obviously absent in this case. Second, the isothermal and all the bed particles are spheres with radius RIP.
boundary condition at the shell surface (zero concentration Since the fluid flow occurs along one single direction, at each z the
gradient) is not correct, since it is in contrast with the presence concentration of the diffusing species as well as the bed and fluid
of a sink in this position, which is postulated by the model. properties are assumed to be constant along the orthogonal
Sohn (1977) stated that for estimating the correct value of Sh directions.
at low Re the local concentration gradient from the particle We can schematize this system considering a semi-infinite
surface to a shell with radius dID must be used. Setting up a active bed starting from z¼0 (Fig. 5a). The bed of particles is
simple diffusion model in radial coordinates and accounting for divided into two zones: a first zone (  Noz o0) made of inert
an approximate reasonable fluid velocity profile, the author found particles, a second zone (0oz oN) made of a mixture of inert
Sh 42 for all experimental conditions. Sh was solely a function of and active particles. The volume fraction of active bed particles in
the bed voidage. This approach (which is somewhat similar to this zone is XAB. If we assume plug flow through the bed, in the
that used to obtain Eq. (10)) cannot be extended to larger Re, since first zone the mass balance gives the simple final result C (z)¼CIN,
96 F. Scala / Chemical Engineering Science 91 (2013) 90–101

INERT PARTICLES
b local mass transfer coefficient):
ACTIVE PARTICLES
a
2
d C dC
Def f þU ¼ a  K n ðC AP C Þ ð36Þ
dz2 dz
with boundary conditions:
z= L
z ¼ 0 ) C ¼ C IN
RAP 
dC 
z-1 ) ¼0 ð37Þ
dz 1

Solution gives:
0 0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 1
n
CC AP U 4  a  K  D ef f
¼ exp@  @1 1 þ A  zA
C IN C AP 2  Def f U2
ð38Þ
z=0 z= 0 OUT
If we measure the concentration C at a certain bed length L
we can write (after some manipulations):
0 0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 1
z C OUT C AP Pe 24  ð1ebed Þ  X AB  ebed =t  Sh A
n
L A
IN
¼ exp@  @1 1þ 
C C AP 4  ebed =t Pe 2 RAP

ð39Þ
By substituting Eq. (39) in Eq. (35) we can calculate the
effective bed Sherwood number as:
Fig. 5. Bed schematization for mass transfer under non-stagnant conditions: (a)
Pe2
semi-infinite active bed; (b) finite length active bed. Shef f ¼
12  ð1ebed Þ  X AB  ebed =t
0sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
n
where CIN is the inlet concentration. In the second zone the 24  ð1ebed Þ  X AB  ebed =t  Sh
@ 1 þ 2
1A ð40Þ
stationary mass balance on the diffusing species reads (introdu- Pe
cing the effective bed mass transfer coefficient Keff, i.e., a mass
transfer coefficient based on the mean concentration driving force One possible objection to this approach regards the first
in the bed): boundary condition (37). At low Re the diffusing component can
back-diffuse towards the fluid inlet. So in principle a concentra-
dC tion gradient exists also in the first inert bed zone and the
U ¼ a  K ef f ðC AP C Þ ð31Þ
dz concentration at z ¼0 is not equal anymore to CIN. In order to
with the boundary condition: solve this situation, axial diffusion should be accounted for in the
mass balance in the first zone. This problem is solved in Appendix 1,
z ¼ 0 ) C ¼ C IN ð32Þ and the resulting solution is Eq. (A1.5). As demonstrated in
Appendix 1, for all practical situations this effect can be neglected.
Solution gives: A second possible objection is that real fixed and fluidized beds
  have a finite length, so that considering a semi-infinite active bed
CC AP a  K ef f
¼ exp   z ð33Þ may lead to deviations from the correct results. In Appendix 2 we
C IN C AP U
solve the problem of mass transfer in a bed of length L, as
If we can measure the concentration COUT at a certain bed schematized in Fig. 5b. Again, in Appendix 2 we demonstrate
length L (e.g., using an on-line analyzer or a chromatographic that results for the finite length active bed model are practically
technique), the effective bed Sherwood number can be calculated coincident with those for the semi-infinite active bed model, so
as: that Eq. (40) appears to be a reasonably simple and accurate
relation.
!
2  RAP  K ef f 2  RAP  U C OUT C AP In order to use Eq. (40), the local particle Sherwood number
Shef f ¼ ¼ ln ð34Þ Shn must be expressed. The value estimated with Eq. (10) is
Dm Dm  a  L C IN C AP
strictly valid only in stagnant conditions. On the other hand, we
If we now introduce the Peclet number Pe ¼ Re  Sc ¼ 2RAP U would like to find a general expression valid in the whole range of
Dm
Eq. (34) becomes: Reynolds numbers of interest. Now, on the basis of the discussion
! given in Section 3.1, it is reasonable to assume that the effect of
Pe R C OUT C AP the fluid velocity on the local Sherwood number is the same as
Shef f ¼   AP ln ð35Þ
3  ð1ebed Þ  X AB L C IN C AP that predicted by Frössling for an isolated sphere using the
boundary layer theory. This means that we can combine
Most of the researchers have used this equation to estimate Eqs. (1) and (10) to yield:
the Sherwood number in fixed and fluidized beds from 2  ebed =t  c
þ b  Re=ebed  Scd
n
experimental data. Sh ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð41Þ
1 3 ð1ebed Þ  X AB
A more detailed approach should include axial diffusion in the
mass balance (Glicksman and Joos, 1980), since this contribution Again, the Reynolds number was divided by the bed voidage,
becomes important as Re-0. Let us, however, for the moment since the boundary layer thickness depends on the real gas
assume that the first inert bed can still be considered to be in plug velocity around the particle rather than on the superficial velo-
flow. In the second active zone the stationary mass balance on the city. The values of the parameters b, c, and d should be found
diffusing species reads (where we now use the previously defined experimentally. Inserting Eq. (41) in Eq. (40) gives the general
F. Scala / Chemical Engineering Science 91 (2013) 90–101 97

correlation for the effective bed Sherwood number: For Re-0 (and XAB a0, ebed a1) Eq. (42) has the following
asymptotic expression:
Re2  Sc2
Shef f ¼
12  ð1ebed Þ  X AB  ebed =t Re  Sc
Shef f ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð46Þ
0v
u
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
! 1 3  ð1ebed Þ  X AB  ð1 3 ð1ebed Þ  X AB Þ
u 24  ð1ebed Þ  X AB  ebed =t 2  ebed =t  c
@t1þ 2 2
 p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi þ b  Re=ebed  Scd 1A
Re  Sc 1 ð1ebed Þ  X AB
3
indicating that the effective Sherwood number goes to zero for
ð42Þ Re-0 and varies linearly with the Reynolds number at very low
superficial velocities. In this case obviously the behavior of the
where the equality Pe ¼ Re  Sc was used. Now we will examine effective Sherwood number is different from that of the local
the asymptotic behavior of this expression with the different particle Sherwood number which tends to a finite value. Unfortu-
quantities of interest. First, for ebed-1 Eq. (42) reduces to: nately, this is the case where it is most difficult to obtain
experimental data, especially in fluidized beds. Some evidence
Shef f ¼ Sh ¼ 2 þb  Rec  Scd
n
ð43Þ of a linear behavior of Shef f with Re at low fluid velocities has been
reported in the literature (Richardson and Szekely, 1961; Koloini
that is the well-known Frössling-type correlation for mass trans- et al., 1977), but the reliability of the published data obtained
fer around an isolated sphere in a convective flow. This correlation under these conditions is questionable. More reliable data are
has been shown to accurately predict mass transfer data around an needed to test the accuracy of the present model. Interestingly,
isolated sphere in a fluid flow (in the range 20oReo2000). The the same asymptotic dependency of Shef f with Re was obtained in
following parameter values were suggested for gas–solid operation: the models by Nelson and Galloway (1975) and by Sørensen and
b¼0.69; c¼1/2; d¼1/3 (Rowe et al., 1965). Stewart (1974), and was also shown by numerical results of the
For XAB-0 Eq. (42) reduces to: model by Glicksman and Joos (1980).
 c One final consideration regards the comparison between the
Shef f ¼ Sh ¼ 2  ðebed =tÞ þ b  Re=ebed  Scd
n
ð44Þ model results obtained with the stagnant model (Eq. (23)) and those
reported here for Re-0. Apart from the limiting case XAB ¼0
This expression is relevant for the case of one or few isolated (or ebed ¼1), the present model always predicts Shef f ¼ 0 for Re¼0.
active spheres immersed in a bed of inert particles. This case was Results of the stagnant model, instead, predict finite values of the
recently reviewed by Scala (2007) referring to bubbling gas overall particle Sherwood number for finite active bed lengths.
fluidized beds, where most of the experimental data are available. This inconsistency depends on the different geometrical domain of
This author showed that the two-phase behavior of bubbling integration for the two models. In fact, the one-dimensional approach
fluidized beds should be taken into account in analyzing the used here is based on the assumption that gas concentration can be
experimental data. In particular, evidence was given that the assumed to be constant in the directions orthogonal to the main gas
active particles only reside in the dense phase and never enter the flow. This assumption clearly fails at Re¼0 when no imposed gas
bubble phase. As a consequence, Eq. (44) can still be used to flow exists anymore. In fact, the very definition of the effective
correlate the experimental data, but U and ebed should be Sherwood number in Eq. (34) implies that for U-0 it is Shef f ¼ 0,
substituted with Umf and emf, which represent the relevant whatever the value of the other parameters (except for XAB ¼0 or
quantities in the dense phase. With the above substitution Eq. ebed ¼ 1). However, since the value for Re¼0 is only a limiting
(44) was shown to excellently fit the experimental data using the asymptote and has no practical interest, the present approach is
same parameter values given before for Eq. (43) in the range considered to be applicable to all multiparticle systems of interest.
0.5 oRemf ¼ 2  RAP  U mf  r=m o50 (in reality a slightly larger
value of b¼0.7 was found to provide the best fit to the experi-
3.3. Results
mental data, but using b ¼0.69 would give practically the same
results).
The local particle Sh and the effective bed Sh, given by Eqs. (41)
For Re-N Eq. (42) reduces to:
and (42), respectively, have been calculated as a function of the
 c Reynolds number with the following base case parameter values:
Shef f ¼ Sh ¼ b  Re=ebed  Scd
n
ð45Þ

i.e., the presence of other active particles does not affect the mass
transfer coefficient anymore, since the concentration boundary
layer is much thinner than half the average distance between two
adjacent active particles. A large number of experimental data
exist for this case both in fixed and fluidized beds and both for
liquid and gas operation. A number of correlations of the type of
Eq. (45) have been successfully used to fit the experimental data
with different parameter values for the different systems and
conditions considered (Dwivedi and Upadhyay, 1977; Bird et al.,
2002). For example, the two most widely used correlations of
experimental data for mass transfer coefficients in packed beds at
large Reynolds numbers predict Sh¼ 1.8 Re0.5 Sc1/3 (Ranz, 1952)
and Sh¼1.1 Re0.6 Sc1/3 (Wakao and Funazkri, 1978). The correla-
tion developed by Dwivedi and Upadhyay (1977) at large Rey-
nolds numbers gives practically the same coefficients than the
one proposed by Wakao and Funazkri (1978).
Note that in all these three cases examined the effective bed Fig. 6. Local and effective Sherwood numbers as a function of Reynolds number
Sherwood number and the local particle Sherwood number for different active particle volume fractions. ebed ¼0.45; t ¼ 1.2; Sc ¼2.5; b ¼0.7;
coincide. c¼ 1/2; d ¼ 1/3.
98 F. Scala / Chemical Engineering Science 91 (2013) 90–101

ebed ¼0.45; t ¼1.2; XAB ¼1; Sc ¼2.5; b¼ 0.7; c¼ 1/2; d ¼1/3. operate at Re 41, and it is clearly visible in Fig. 6 that in this range
The value of the Schmidt number was selected to represent a of Re whatever the value of XAB there is not so much difference
typical gas–solid system (Kunii and Levenspiel, 1991). The values between Shn and Sheff (the largest error is about 40% at Re¼1 and
of b, c and d are those found to give the best fit to mass transfer XAB ¼1).
data in gas–solid fluidized bed when the active particles are very Fig. 7 reports Shn and Sheff, as a function of Re, for different
diluted in inert particles (Scala, 2007). Other values for these values of ebed. The curve for ebed ¼ 1 (valid for both Shn and Sheff)
parameters might be more appropriate for fixed beds and/or was calculated with Eq. (43). Again, for large values of Re the
liquid–solid processes. curves representing Shn converge with those representing Sheff,
Fig. 6 reports Shn and Sheff, as a function of Re, for different while at low Re values the two series of curves diverge. When the
values of XAB. The curve for XAB ¼0 (valid for both Shn and Sheff) bed voidage increases the value of Shn decreases. The same is true
was calculated with Eq. (44). As expected, for large values of Re, for Sheff for large Re, while the opposite effect is found at low Re. In
all the curves converge into one representing Eq. (45). At low Re general, the effect of ebed is more significant at large bed voidages.
values, the two series of curves diverge: Shn tends towards a finite Fig. 8 shows the effect of the Schmidt number on Shn and Sheff.
value, while Sheff tends to zero linearly with Re. The effect of an This effect is clearly very significant. When we pass from gas–
increase of XAB on Shn is to enhance the local mass transfer solid systems (0.1 oSc o10) to liquid-solid ones (100oSc o
coefficient (see e.g., Eq. (11)), except for very large values of Re. 100,000) the behavior of the Sherwood number changes drama-
A similar effect is visible on Sheff for relatively large Re, while the tically. For example, for Sc ¼10,000 at Reynolds numbers as low
opposite effect is noted for low Re. From a practical point of view as Re¼0.001 the two Sherwood numbers coincide. Again,
it is seen that for XAB o0.1 (i.e., if the active particles occupy less most liquid–solid fixed and fluidized bed processes operate at
than 10% of the solids volume) for Re41 the two Sherwood Re40.001.
numbers coincide and can be substituted with a reasonable
accuracy by the simpler Eq. (44). This is for example the case of
solid fuel combustion in fluidized bed boilers (Scala, 2007). It is 4. Conclusions
also noted that most gas–solid fixed and fluidized bed processes
The problem of mass transfer in multiparticle systems was
examined in detail with a particular attention to the stagnant and
low Reynolds number cases. The following conclusions can be
drawn:

1. Different Sherwood numbers can be calculated depending on


the choice of the reference concentration difference (driving
force). In particular, a local Sh based on a concentration
difference close to the active particle and an effective Sh based
on a macroscopic concentration difference can be used.
The use of these different Sherwood numbers implies a
different implementation of the mass balance equations.
2. The local Sherwood number depends only on geometrical and
fluid-dynamics considerations and therefore has a general
validity. On the other hand, the effective Sherwood number
also depends on the assumptions made in deriving the mass
balance equations across the bed (e.g., typically a fluid plug
flow assumption). As a consequence, the effective Sherwood
Fig. 7. Local and effective Sherwood numbers as a function of Reynolds number number can be used only if the same assumptions are made in
for different values of the bed voidage. XAB ¼ 1; Sc¼ 2.5; b ¼0.7; c ¼1/2; d ¼1/3. the bed model as those used in deriving it. For example, if we
want to schematize a fluidized bed reactor as a perfectly mixed
reactor, it is not correct to use an effective Sh derived under
the assumption of a plug flow reactor. The same consideration
applies if we want to include in the bed model phenomena
like bed channeling, gas bypass in bubbles, or backmixing.
The local Sherwood number should be used in all these cases.
3. For low volume fractions of active particles (roughly XAB o0.1)
a very simple correlation for the Sherwood number can be
used in all situations, given by Eq. (44). If the system under
consideration is a gas–solid bubbling fluidized bed, care should
be taken to use the relevant dense phase quantities Umf and emf
in Eq. (44).
4. For large Reynolds numbers the diffusive term can be
neglected and the Sherwood number can be estimated in all
situations with Eq. (45). The values of the parameters in this
relation depend on the different systems and conditions
considered.
5. For all the other situations the local Sherwood number can be
calculated with the general expression Eq. (41), while the
Fig. 8. Local and effective Sherwood numbers as a function of Reynolds number effective Sherwood number (only under the assumption of
for different Schmidt numbers. ebed ¼0.45; t ¼1.2; XAB ¼ 1; b¼ 0.7; c¼ 1/2; d ¼ 1/3. plug flow) with Eq. (42). The accuracy of these expressions in
F. Scala / Chemical Engineering Science 91 (2013) 90–101 99

the very low Reynolds number limit, however, has not been eff effective
proved due to the lack of reliable experimental data. Never- ID half average inter-particle distance
theless, most fixed and fluidized bed real processes operate at IP inert particle
larger Reynolds numbers where the two expressions give m molecular
approximately the same results. The use of the simpler local mf relative to minimum fluidization conditions
Sherwood number (Eq. (41)) is therefore suggested. N infinitely far
6. The fact that the effective Sherwood number tends to zero
linearly with Re when Re-0, is simply a consequence of the Superscripts
plug flow assumption in the bed mass balance (as suggested
by previous researchers (Richardson and Szekely, 1961; IN inlet
Glicksman and Joos, 1980)). The limiting value of the local Sh OUT outlet
when Re-0 instead is always a finite number. n local
7. Finally, it must be recognized that for large volume fractions of I first bed zone
active particles equilibrium conditions are easily reached close II second bed zone
to the fluid inlet and that the mass transfer potential in the
upper part of the bed is very small. As a consequence, in most
practical applications the detailed description of fluid-bed
mass transfer is not really important, and a simple mass
balance across the bed is sufficient for design purposes. Appendix 1

With reference to the system schematized in Fig. 5a, the mass


balance on the diffusing species in the first bed zone
(  Nozo0) taking into account axial diffusion reads:
Nomenclature
2
d C dC
a specific surface area (m ) 1 Def f þU ¼0 ðA1:1Þ
dz2 dz
b constant in Eq. (1) (  )
c constant in Eq. (1) (–) This equation can be coupled with Eq. (36) in the second bed
C concentration (kmol/m3) zone (0 ozoN) with the following four boundary conditions:
d constant in Eq. (1) (–)
z-1 ) C ¼ C IN
D diffusion coefficient (m2/s)
k constant in Eq. (1) (–) z ¼ 0 ) C I0 ¼ C II0
K mass transfer coefficient (m/s) I II
L active bed length (m) dC  dC 
z¼0 )  ¼ 
N number of active particles per unit volume (m  3) dz 0 dz 0
Pe Peclet number (–) 
dC 
r radial coordinate (m) z-1 ) ¼0 ðA1:2Þ
dz 1
R radius (m)
Re Reynolds number (–) The second and third conditions impose the equality of the
Sc Schmidt number (–) concentrations and of the concentration gradients calculated in
Sh Sherwood number (–) the two bed zones at z¼0.
u auxiliary variable in Eq. (18) (kmol/m2) Solution of Eqs. (A1.1) and (36) with the boundary conditions
U superficial velocity (m/s) (A1.2) gives in the second bed zone:
X volume fraction (–)
0 0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 1
z length coordinate (m) 4  a  K n  Def f A
CC AP @ U @
¼ exp  1 1 þ  zA
C IN C AP 2  Def f U2
Greek letters
2 0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi13

4  a  K n  Def f
a quantity introduced in Eq. (20) (–) 41  @1 þ 1þ A5 ðA1:3Þ
g1 quantity introduced in Eq. (A2.2) (–)
2 U2
g2 quantity introduced in Eq. (A2.2) (–)
g3 quantity introduced in Eq. (A2.6) (–) This expression is similar to that in Eq. (38), with the right
d thickness (m) hand side divided by a term Z1. For Re-N Eq. A1.3 reduces to
y enhancement factor (–) Eq. (38). If we measure the concentration COUT at a certain bed
e voidage (–) length L we can write (after some manipulations):
r density (kg/m3)   qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n

24ð1ebed ÞX AB ebed =tSh
t bed tortuosity (–) OUT exp 4ePe
bed =t
 1 1 þ Pe 2  L
RAP
C C AP
m viscosity (kg/m s) ¼  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffin 
C IN C AP 1 24ð1ebed ÞX AB ebed =tSh
2  1þ 1þ Pe2

Subscripts ðA1:4Þ

AB active bed Finally by inserting Eq. (A1.4) into Eq. (35) we can calculate
AP active particle the effective bed Sherwood number:
bed relative to the bed
BL boundary layer Pe
Shef f ¼ 
C cell 3  ð1ebed Þ  X AB
100 F. Scala / Chemical Engineering Science 91 (2013) 90–101

10 Solution gives in the active bed zone:


   
g1  exp g1  RLAP  exp g2  RzAP g2  exp g2  RLAP  exp g1  RzAP
Effective bed Sherwood number

CC AP
¼  
C IN C AP g1  exp g1  RLAP g2  exp g2  RLAP
1
ðA2:2Þ

where
0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
0.1 n
Pe 24  ð1ebed Þ  X AB  ebed =t  Sh A
g1 ¼  @1 1þ
4  ebed =t Pe2

0.01 Eq. (40)


and
Eq. (A1.5) - L/RAP = 2
Eq. (A1.5) - L/RAP = 10 0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1
n
Eq. (A1.5) - L/RAP = 20 Pe 24  ð1ebed Þ  X AB  ebed =t  Sh A
g2 ¼  @1þ 1þ :
4  ebed =t Pe 2
0.001
0.001 0.01 0.1 1 10 100 1000
Peclet number At the bed outlet it is:
OUT
C C AP
Fig. A1. Comparison of effective Sherwood numbers calculated with Eqs. (40) and
(A1.5) as a function of Peclet number for different bed lengths. ebed ¼0.45; t ¼ 1.2; C IN C AP
   
XAB ¼ 1; Sh* ¼4.15. g1  exp g1  RLAP  exp g2  RLAP g2  exp g2  RLAP  exp g1  RLAP
¼  
g1  exp g1  RLAP g2  exp g2  RLAP

8   qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 9 ðA2:3Þ
>
>exp 4ePe  1 1 þ 24ð1ebed ÞX AB ebed =tSh
n
 L >
>
R < bed =t Pe 2 RAP =
 AP ln  By inserting Eq. (A2.3) into Eq. (35) we can calculate the
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffin 
L >
> >
> effective bed Sherwood number:
: 1
2  1þ 1 þ 24ð1ebed ÞX Pe2
AB ebed =tSh ;
Pe
ðA1:5Þ Shef f ¼ 
3  ð1ebed Þ  X AB
8    9
L L L L
We note that contrary to Eq. (40), in this case Sheff is a function RAP <g1  exp g1  RAP  exp g2  RAP g2  exp g2  RAP  exp g1  RAP =
 ln  
of L=RAP . Fig. A1 reports Sheff as a function of Pe calculated with L : g1  exp g1  RLAP g2  exp g2  RLAP ;
Eq. (A1.5) at different L=RAP values, and that calculated with
ðA2:4Þ
Eq. (40), for comparison. The following parameter values were
used: ebed ¼0.45; t ¼1.2; XAB ¼1; Shn ¼4.15. In particular, this last If we remove the plug flow assumption in the first and third
value is that resulting from Eq. (10) relevant for stagnant condi- inert zones, the mass balance in these zones is now given by
tions. For very low Reynolds numbers the local Sherwood number Eq. (A1.1). It is first noted that in the third zone (Loz oN) the
would not be very far from this value. only possible solution that gives a finite concentration infinitely
Examination of Fig. A1 indicates that a not negligible deviation far from the active bed is C (z)¼COUT. So we only have to solve the
between the two models is found for L=RAP ¼2, representing the two coupled Eqs. (A1.1) and (36) in the first and second zone,
lowest possible value (a monolayer of active particles). On the respectively, with boundary conditions:
other hand, for L=RAP 410, representing a bed height larger than
5 active particle diameters, the difference between results of the z-1 ) C ¼ C IN
models is very small. This means that for all situations of practical z ¼ 0 ) C I0 ¼ C II0
interest the bed entrance effect due to axial diffusion can be I II
neglected. Only for purposely carried out lab-scale experiments at dC  dC 
z¼0 ) ¼
Re-0 using shallow beds consisting of few active particle layers dz 0 dz 0

this effect should be taken into account. dC 
z¼L ) ¼0 ðA2:5Þ
dz L

where the second and third conditions impose the equality of the
Appendix 2 concentrations and of the concentration derivatives calculated in
the two bed zones at z ¼0. Solution gives in the active bed zone:
Let us consider the system schematized in Fig. 5b, representing
CC AP
a bed of active particles of length L through which a fluid flows in
C IN C AP
the direction z with a superficial velocity U. Two semi-infinite    
beds of inert particles are placed below and above the active bed. g1  exp g1  RLAP  exp g2  RzAP g2  exp g2  RLAP  exp g1  RzAP
¼      
If we assume plug flow through the bed, in the first zone g1 þ g3  exp g1  RLAP  g2 þ g3  exp g2  RLAP
(  Noz o0) the result is C (z) ¼CIN, in the second zone (0 ozoL)
ðA2:6Þ
the result is still given by Eq. (35), in the third zone (Loz oN)
the result is C (z)¼COUT. Now let us remove the plug flow where g3 ¼ 3  ð1ebed Þ  X AB  Sh =Pe. At the bed outlet it is:
n

condition only in the active bed zone. The mass balance in this
C OUT C AP
zone is given by Eq. (36) with boundary conditions:
C IN C AP
IN    
z¼0 ) C¼C g1  exp g1  RLAP  exp g2  RLAP g2  exp g2  RLAP  exp g1  RLAP
 ¼      
dC  g1 þ g3  exp g1  RLAP  g2 þ g3  exp g2  RLAP
z¼L ) ¼0 ðA2:1Þ
dz L ðA2:7Þ
F. Scala / Chemical Engineering Science 91 (2013) 90–101 101

10 References
Effective bed Sherwood number

1 Bird, R.B., Stewart, W.E., Lightfoot, E.N., 2002. Transport Phenomena, second ed.
Wiley, New York.
0.1 Eq. (40) Chuang, S.Y., Dennis, J.S., Hayhurst, A.N., Scott, S.A., 2009. Kinetics of the chemical
Eq. (A2.4) - L/RAP = 2 looping oxidation of CO by a co-precipitated mixture of CuO and Al2O3. Proc.
0.01 Eq. (A2.4) - L/RAP = 10 Combust. Inst. 32, 2633–2640.
Eq. (A2.4) - L/RAP = 20 Cornish, A.R.H., 1965. Note on minimum possible rate of heat transfer from a
0.001 sphere when other spheres are adjacent to it. Trans. Inst. Chem. Eng. 43,
T332–T333.
1 Eq. (A1.5) - L/RAP = 2
Dwivedi, P.N., Upadhyay, S.N., 1977. Particle-fluid mass transfer in fixed and
fluidized beds. Ind. Eng. Chem. Process Des. Dev. 16, 157–165.
Eq. (A1.5) - L/RAP = 10 Frössling, N., 1938. The evaporation of falling drops (in German). Gerlands Beitr.
0.1 Eq. (A1.5) - L/RAP = 20 Geophys. 52, 170–216.
Eq. (A2.8) - L/RAP = 2 Glicksman, L.R., Joos, F.M., 1980. Heat and mass transfer in fixed beds at low
0.01 Eq. (A2.8) - L/RAP = 10 Reynolds numbers. J. Heat Transfer 102, 736–741.
Eq. (A2.8) - L/RAP = 20 Hayhurst, A.N., 2000. The mass transfer coefficient for oxygen reacting with a
0.001 carbon particle in a fluidized or packed bed. Combust. Flame 121, 679–688.
0.001 0.01 0.1 1 10 100 1000 Ho, T.C., 2003. Mass transfer (Chapter 11). In: Yang, W.-C. (Ed.), Handbook of
Peclet number Fluidization and Fluid-Particle Systems. Dekker, New York.
Koloini, T., Sopčič, M., Žumer, M., 1977. Mass transfer in liquid-fluidized beds al
Fig. A2. Comparison of effective Sherwood numbers calculated with Eqs. (40), low Reynolds numbers. Chem. Eng. Sci. 32, 637–641.
Kunii, D., Levenspiel, O., 1991. Fluidization Engineering, second ed. Butterworth-
(A2.4), (A1.5) and (A2.8) as a function of Peclet number for different bed lengths.
Heinemann, Newton, USA.
ebed ¼ 0.45; t ¼1.2; XAB ¼1; Sh* ¼ 4.15.
Nelson, P.A., Galloway, T.R., 1975. Particle-to-fluid heat and mass transfer in dense
systems of fine particles. Chem. Eng. Sci. 30, 1–6.
By inserting Eq. (A2.7) into Eq. (35) we can calculate the Pfeffer, R., 1964. Heat and mass transport in multiparticle systems. Ind. Eng. Chem.
effective bed Sherwood number: Fundam. 3, 380–383.
Ranz, W.E., Marshall Jr., W.R. 1952. Evaporation from drops. Chem. Eng. Prog.
Pe 48,141–146 (part I), 173–180, and (patr II).
Shef f ¼ 
3  ð1ebed Þ  X AB Ranz, W.E., 1952. Friction and transfer coefficients for single particles and packed
8    9
L L L L beds. Chem. Eng. Prog. 48, 247–253.
RAP <g1  exp g1  RAP  exp g2  RAP g2  exp g2  RAP  exp g1  RAP =
 ln       Richardson, J.F., Szekely, J., 1961. Mass transfer in a fluidised bed. Trans. Inst.
L : g1 þ g3  exp g1  RLAP  g2 þ g3  exp g2  RLAP ;
Chem. Eng. 39, 212–222.
Rowe, P.N., Claxton, K.T., Lewis, J.B., 1965. Heat and mass transfer from a single
ðA2:8Þ
sphere in an extensive flowing fluid. Trans. Inst. Chem. Eng. 43, T14–T31.
Fig. A2 reports Sheff as a function of Pe calculated with Scala, F., 2007. Mass transfer around freely moving active particles in the dense
phase of a gas fluidized bed of inert particles. Chem. Eng. Sci. 62, 4159–4176.
Eqs. (A2.4) and (A2.8) at different L=RAP values, compared with Schlünder, E.U., 1977. On the mechanism of mass transfer in heterogeneous
those calculated with Eqs. (40) and (A1.5), for comparison. The systems—in particular in fixed beds, fluidized beds and on bubble trays. Chem.
following parameter values were used: ebed ¼0.45; t ¼1.2; XAB ¼ 1; Eng. Sci. 32, 845–851.
Shn ¼4.15. As noted in Appendix 1, this value of the local Sher- Sohn, H.Y., 1977. The limiting Sherwood and Nusselt numbers in sphere-packed
beds. Lett. Heat Mass Transfer 4, 403–415.
wood number is that resulting from Eq. (10) relevant for stagnant Sørensen, J.P., Stewart, W.E., 1974. Computation of forced convection in slow flow
conditions. For very low Reynolds numbers Shn should not be through ducts and packed beds—III. Heat and mass transfer in a simple cubic
much different from this value. array of spheres. Chem. Eng. Sci. 29, 827–832.
Van Heerden, C., 1952. Some fundamental characteristics of the fluidized state.
Examination of results shown in Fig. A2 indicates that the
J. Appl. Chem. 2, S7–S17.
more detailed model reported here for a finite bed length gives Wakao, N., Funazkri, T., 1978. Effect of fluid dispersion coeffcients on particle-
practically the same results as the semi-infinite bed model, both to-fuid mass transfer coeffcients in packed beds. Correlation of Sherwood
with and without the inclusion of the axial diffusion bed entrance numbers. Chem. Eng. Sci. 33, 1375–1384.
effect.

Você também pode gostar