Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Chemical Engineering Science 61 (2006) 3917 – 3929

www.elsevier.com/locate/ces

Application of surface-renewal-stretch model for interface mass transfer


Babak Jajuee, Argyrios Margaritis ∗ , Dimitre Karamanev, Maurice A. Bergougnou
Department of Chemical and Biochemical Engineering, Faculty of Engineering, University of Western Ontario, London, Ont., Canada N6A 5B9

Received 4 April 2005; received in revised form 8 December 2005; accepted 20 January 2006
Available online 9 March 2006

Abstract
A new surface-renewal-stretch (SRS) model was developed to correlate experimental data for the time-average overall mass transfer coefficient,
KL,av , in liquid–liquid and gas–liquid mass transfer systems. The model is based on the equation of continuity, which includes both turbulent
and convective mass transfer at the liquid–liquid and gas–liquid interfaces. The model incorporates Dankwerts surface-renewal model with the
penetration theory for surface stretch proposed by Angelo et al. [Angelo, J.B., Lightfoot, E.N., Howard, D.W., 1996. Generalization of the
penetration theory for surface stretch: application to forming and oscillation drops. A.I.Ch.E. Journal 12 (4) 751–760]. We used our new SRS
mass transfer model to correlate successfully the existing interface mass transfer experimental data from published literature. As a result, the
experimental mass transfer coefficient data was predicted with a high degree of accuracy.
䉷 2006 Elsevier Ltd. All rights reserved.

Keywords: Mass transfer coefficient; Interface; Bubble column; Gaslift contactor; Airflit bioreactor; Absorption; Bubble; Drop

1. Introduction dispersed phase are both a Newtonian fluid and then it is


extended to cases in which the bulk phase would be also
Mass transfer theories are mainly developed for the process non-Newtonian.
of absorption of a gas into a liquid, even though their applica- The conventional and simplest picture of mass transfer be-
tion might be extended to the cases that mass transfer occurs tween two fluid phases, borrowed from a similar concept used
between any two immiscible fluid phases. Before turning to in the convective heat transfer, is that there exists a stagnant
the principles involved, the reader should be aware of certain film at the interface (Lewis and Whitman, 1924). In this case,
terminology which is basic to understanding the material pre- the surface concentration may be assumed to be the saturated
sented in this paper. For our purposes, a “fluid particle” is a ∗ , while the concentration of the bulk phase is
concentration, cA
self-contained body with maximum dimension between about kept uniform outside this film by turbulent mixing. The rate of
0.5 m and 10 cm, separated from the surrounding medium by mass transfer per unit area, also known as mass transfer flux,
a recognizable interface. The material forming the particle and will then be constant (steady state) and given by the following
the medium surrounding the particles will be termed the “dis- expression:
persed phase” and the “bulk phase”, respectively. If the dis-
persed phase is in the liquid state, the particle is called a “drop”, D
NA = (cA,i − cA0 ), (1)
and if it is a gas, the particle is referred to as a bubble. To- zF
gether, drops and bubbles comprise “fluid particles”. Hence-
where NA is the net flux defined as the rate of absorption per
forth, we use “continuous phase” to refer to the “bulk phase”
unit area, cA,i is the concentration at the interface, cA0 is the
with convective motion. Moreover, consideration is limited in
bulk concentration, D is the molecular diffusivity and zF is
the first place to the cases in which the bulk phase and the
the effective film thickness which depends upon the nature of
the flow conditions. Therefore, the net flux appears to conform
∗ Corresponding author. Tel.: +1 519 661 2146; fax: +1 519 661 4275. to the expression
E-mail address: amarg@uwo.ca (A. Margaritis)
URL: http://www.eng.uwo.ca/people/amargaritis/. NA = kL (cA,i − cA0 ), (2)

0009-2509/$ - see front matter 䉷 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2006.01.026
3918 B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929

where kL is the film mass transfer coefficient and is constant component before being submerged once again. To compensate
for a fluid–fluid system under given conditions. for the lack of turbulence at the interface, Dankwerts added a
Although it is preferable to refer to the conventional picture new constant parameter called the mean rate of production of
due to its simplicity as though it really exists, the fictitious na- fresh surface, s, to the penetration theory and assumed that the
ture of the film theory may lead to erroneous results in many chance of a surface element being replaced by another within
actual mechanisms of mass transfer. In particular, when fluid a given time is independent of how long it has been in the
particles rise or fall in infinite media with turbulent motion, surface; hence the fractional rate of replacement of the elements
the conditions required maintaining a stagnant film at the sur- belonging to different age groups is equal to s. Combination
face of moving fluid particles is lacking. In other words, it ap- of the continuity equation with the same conditions of Eq. (3)
pears unlikely that the surface layer on each particle keeps its and the concept of surface renewal gives a new expression
identity throughout all times. Higbie (1935) took a major step to the mean rate of mass transfer per unit area of turbulent
forward by introducing the “penetration theory”. This theory surface
emphasizes that in many situations the time of exposure of a √
fluid to mass transfer is too short to let the concentration gradi- NA,av = Ds(cA,i − cA0 ). (5)
ent of the film, characteristic of steady state, to develop. With
the assumption of constant but short time of exposure for all The penetration and surface-renewal theories tacitly assume
fluid particles, they are subject to unsteady state diffusion or that immediately underlying all freshly formed surfaces is fluid
penetration of solute. Taking the uniform initial concentration with the same concentration, cA0 , of dissolved component. It
of cA0 for the dissolved component and the surface concentra- is equal to consider the surface elements to be infinitely deep
tion of cA,i , as soon as the bulk phase is exposed to the dif- that diffusing component never reaches the region of constant
fusing component, which may be also taken as the equilibrium concentration below. Dobbins (1956) pointed out that this is
solubility of it in the bulk of the fluid, the initial and boundary not true for aeration systems where flowing streams continu-
conditions become ally absorb oxygen and the concentration of the bulk of the
fluid changes. Accordingly, Dobbins considered a finite depth

cA0 at  = 0 for all z, of surface elements and eddies by replacing the third boundary
c = cA0 for all  z = ∞, (3) condition in Eq. (3) with cA = cA0 for z = za and obtained
cA,i for all  z = 0, ⎡  ⎤
√ sz 2
in which  is the time of exposure to the diffusing component. NA,av = ⎣ Ds coth a⎦
(cA,i − cA0 ). (6)
Thus, the average flux over the time of exposure is D

4D It can be seen that the general dependence, KL ∝ DAB n , with n
NA,av () = (cA,i − cA0 ). (4)
 dependent upon circumstances, is satisfied in this expression.
Other researchers have made similar suggestions (King, 1966;
This theory pictures an unbroken film for component diffusion Kozinski and King, 1966; Toor and Marchello, 1958; Lamont
while the bulk phase is stagnant and has constant time of expo- and Scott, 1970). In summary, for rapid penetration (D), or for
sure for all eddies and fluid particles; nevertheless, the appli- thin surface elements, and small rate of surface renewal (s), the
cation of the penetration theory is probably extended to cases situation is close to steady state and the predominant character
where the bulk phase is also in turbulent motion. It can be seen of the mass transfer is described by the film theory; whereas for
that the rate of absorption becomes very slow after a time un- slow penetration, or for short time of exposure or rapid renewal
less the surface is renewed by stirring or by convection. This (unsteady state situation) it follows Eqs. (5) or (6). Remark-
was first clarified by Dankwerts (1951) with “surface-renewal ably, Eq. (5) may be applied for bulk phase moving parallel
theory” as he pointed out that Higbie theory with its constant to the surface with a velocity that varies with the depth while
time of exposure of eddies at the surface is a special case of the time of exposure is so short that the depth of penetration
what may be a more realistic picture of the processes occurring is much smaller than the depth at which the velocity is appre-
during absorption into an agitated fluid where in fact eddies ciably different from that at the surface. As a complimentary
are exposed for varying lengths of time. In the surface-renewal to this, Eq. (6) depicts a more realistic picture of the penetra-
theory, the interface is pictured then as a mosaic of different tion depth by assuming a finite depth of surface element for
exposure-time surface elements where eddies are continually fluid particles.
exposing fresh surfaces to the diffusing component and sweep- Molecular diffusivity is the most important attribute in con-
ing away and mixing into the bulk of the fluid. Dankwerts as- centration changes among unsteady state theories discussed
sumed that any portion of the bulk phase absorbs the component above. In addition to transport by molecular motion, however,
at a rate given by Eq. (4) in a case that the depth of penetration mass may also be transported by the bulk motion. Therefore, a
of the solute would be much smaller than the scale of turbu- comprehensive net flux of mass transfer in the liquid including
lence, so that the intensity of turbulence accentuates velocity bulk motion is the combined net flux vector, which consists of
gradient existing beneath the surface and at the same time de- the both molecular flux vector and convective flux vector
creases the depth of penetration by shortening the period for
any part of the bulk phase which is exposed to the diffusing N A = JA + cA v, (7)
B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929 3919

where JA and v are the molar molecular flux vector and the to the downcomer section. Aeration of one section exploits
molar average velocity vector, respectively. hydrostatic pressure differences by making lower hydrostatic
head compared to the non-aerated section at the same vertical
JA = −cD ∇x
 A, coordinate across the membrane wall (King, 1966). Unsteady
circulation with rocking motion of fluid particles increases the

N
intensity of turbulence, which in turn changes incessantly the
v = xA vA , (8)
conformation, and position of eddies and exposes fresh surfaces
A=1
to fluid particles. At the same time, bulk motion also drives ed-
where ∇ is the vector differential operator and is defined in dies and liquid particles in a continuous circulation. It is well
rectangular coordinates as known that the mass transfer processes occurring in fluid–fluid
contactors are too complicated to be explained by an implausi-
j j j ble theory. On the other hand, providing a thorough picture of
∇ = x + y + z (9)
jx jy jz mass transfer, for many purposes, is impossible without taking
the inevitable existence of both bulk and turbulent motions into
in which  is the unit vector. account.
Angelo and coworkers (Angelo et al., 1966; Angelo and
Lightfoot, 1968; Stewart et al., 1970) developed the penetration
concepts in the absence of turbulence when there is bulk mo- 3. Theoretical development
tion by introducing a promising theory called “surface-stretch
theory”. On the assumption that diffusion happens in the same 3.1. Surface-renewal-stretch (SRS) model
direction as convection, i.e., z, the surface-stretch theory ex-
presses the local velocity vz in time-dependent interface surface The main objective of this work is to extend the penetration
A, and gives the continuity equation a new form theory of mass transfer to fluid–fluid interacting systems, e.g.
gaslift contactors, in which both turbulent and bulk motions
j ln A jcA jcA j 2 cA exist. Consider a bulk phase with uniform rate of mass transfer
−z + =D 2 (10) through the interface that is maintained in turbulent and bulk
j x,y jz j jz
motion, as shown in Fig. 1. This case may be expounded from
whose solution leads to another expression for the net flux both macroscopic and microscopic standpoints. The former is
⎡ ⎤ pictured with eddies and fluid particles that have an explicit bulk
motion, while the latter with the surface of each element that is
⎢ (A(/0 )) D/0 ⎥ continuously replaced with fresh surface and has a distribution
NA,av = ⎣  ⎦ (cA,i − cA0 ), (11)
/0 2 of ages varying from zero to infinity at any particular instance.
0 A () d
Let the total area of the surface exposed to the gas would be
where A is the interfacial surface through which mass transfer equal to A∞ . On the assumption that the chance of any portion
takes place while it changes periodically with time, and 0 is a of surface element being replaced by another is quite indepen-
constant with the dimension of time defined for each system. dent of how long it has been in the surface, the average rate of
Eq. (10) will be justified in the discussion of surface-renewal- exchange of the film which produces fresh surfaces is taken to
stretch theory. In brief, the concepts of bulk motion and tur- be constant and equal to r(). The concept of surface–age dis-
bulent motion have been neglected in the surface-renewal and tribution function might be borrowed from the surface-renewal
surface-stretch theories, respectively. Presented here is a holis- theory, in the sense that at any moment the distribution of ages
tic approach to the treatment of mass transfer processes occur- of the film area is given by the function (), and the film area
ring in fluid–fluid contactors, while the symbiotic relationship comprising the elements having ages between  and  + d is
exist between turbulent and bulk motions synchronously. () d. This does not vary with time at steady state and is
equal to the film area in the age group between  − d and 
less the portion of this, which is replaced by fresh surface in a
2. Fluid–fluid contactors short time interval equal to d

More recently, fluid–fluid contactors are referred to a myr- r


iad of configurations developed for two or three-phase appli- () d = ( − d) d 1 − d . (12)
A∞
cations in which fluid particles, e.g. bubbles or drops, have
interactions. Fluid particles oscillate and wobble when rising Writing Taylor series for ( − d) gives
through a denser media and make surfaces wavy and rippled.
d() r
Gaslift systems considered as gas–liquid contactors, for exam- () = () − d − () d. (13)
ple, are widely used in many operations such as fermentation d A∞
and wastewater treatment especially where microorganisms are Thus
used as biocatalysts. In general, they are pictured as pneumati-
cally agitated devices characterized by circulation in a defined d() r
=− () (14)
cyclic pattern where slurry or substrate circulates from the riser d A∞
3920 B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929

vz

Distance, z
cA0
Finite depth of
element surface
za For all 

At z=0 for all  > 0

cA,i Concentration

Fluid
Particle

Fig. 1. Mass transfer mechanism in a fluid–fluid contactor when the scale of turbulence and convection are both taken into account.

and since where vz may be derived by making an overall momentum bal-


 ∞ ance over each contactor system; however, it is also attributable
() d = A∞ (15) to the corresponding interfacial area of the time-dependent sur-
0 face. In the case of constant molar density of the solution, con-
the distribution function can be obtained as sider a small flat element of surface xy moving with the
fluid bulk velocity. The direction of the z-axis is pointing into
() = re−s  , (16) the diffusing phase whose net rate of mass transfer is assumed
to be small. The area of the surface of this element changes
where s has the value of r/A∞ , defined as the rate of production with time
of fresh surface which is exposed to the gas (Dankwerts, 1951).
jA
The area of the mass transfer surface, () d, is a function of = [x(vy |y+y − vy |y ) + y(vx |x+x − vx |x )].
time and may be readily integrated to yield j ave
(20)
A() = −A∞ e−s  + C1 . (17) For any point in the interfacial surface, we may divide Eq. (20)
by xy and take the limit as x, y approach zero
Taking A = A0 at  = 0 gives
j ln A jvx jvy
A() = A∞ (1 − e−s  ) + A0 . (18) = + . (21)
j jx jy
This expression represents the area of the mass transfer surface When the density is constant this expression may be written
changes with time and the intensity of turbulence, s. On the with the aid of the equation of continuity as
other hand, this area may also change with time due to convec-
tion, which produces turbulent motion in turn. We limit consid- j ln A jvz
eration to situations in which diffusion happens in the direction =− . (22)
j jz
of convection, perpendicular to the interfacial surface area. We
shall also assume that at any given instant an interface, which Since the expression on the left-hand side is independent of z,
expands and shrinks with time, exists between the continuous we may write for negligible net transfer across the interface
fluid and fluid particles rising in the z direction with the bulk
motion, as shown in Fig. 1. Under these circumstances, diffu- j ln A
vz = −z , (23)
sion is negligible in directions parallel to the interface and is j
also considered one dimensional in the neighborhood of any
surface element. The continuity equation, therefore, takes the where the overall interfacial area A() is obtainable from
form Eq. (18) at any time. It is worth noting that z should be small
compared to the local radii of curvature for curved surfaces like
jcA jcA j 2 cA the interfacial surface of a gas bubble. This is consistent with
vz + =D 2 , (19)
jz j jz the limitations on the penetration theory. Eqs. (19) and (23) are
B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929 3921

then combined to give Eq. (10) Eq. (30) is a partial differential equation whose easiest way of
solving is considering a combination of variables in order to
j ln A jcA jcA j 2 cA convert it to a simple differential equation
−z + =D 2 . (10)
j x,y jz j jz
C = C(),  = Z × G(T ). (35)
For simplicity sake, we shall assume that D would be constant.
The boundary conditions over Eq. (10) then are Hence, Eq. (30) becomes
 
I.C. c = cA0 at  = 0 for all z, d2 C F G dC
(24) + − 3 = 0, (36)
d2 G2 G d
B.C.1 c = cA0 for all  at z = za , (25)
where G indicates the derivative of G with respect to T at
B.C.2 c = cA,i for all  at z = 0. (26) constant Z. The principle of “combination of variables” requires
 
F G
The initial condition and the first boundary condition above − = , (37)
follow the assumed uniform bulk concentration of gas in en- G2 G3
tire solution and finite depth of the surface elements or eddies. where  is an arbitrary constant. Eq. (36) is thus simplified to
Nevertheless, the rough assumption of infinitely deep surface
element is constantly referred to as though there are short con- d2 C dC
stant times. This may be regarded as a harmless and convenient 2
+  = 0. (38)
d d
usage for many systems but not for aerated flowing systems in
which bulk concentration continually changes with time. In air- In brief, Eq. (30) was reduced to Eq. (38), which is an ordinary
lift systems, the dissolving solute is able to reach the depth za differential equation and has the solution of
corresponding to the thickness of the eddy, so that from the so- √ √ 
lute point of view, za is essentially finite. The second boundary 2 
C() = C2 + C3 × erf . (39)
condition expresses the concentration in the continuous phase 2
at the interface, cA,i , which may be taken as the equilibrium
solubility of the fluid particles in the fluid. The concept of each If the assumed combination of variables is successful it should
concentration has been illustrated schematically in Fig 1. To be necessary that Eq. (39) satisfy all Eqs. (31)–(33). Application
solve Eq. (10) we may define the following dimensionless vari- of the initial condition and the two boundary conditions permits
ables: the evaluation of C2 and C3 which are constant

cA − cA,i C(0) = 0, (40)


C= , (27)
cA0 − cA,i
C(A ) = 1, (41)
z
Z= √ , (28) where A is the value of  when Z = ZA . Eq. (37) may be
D0
rearranged more simply as

T = , (29) F G − G = G3 . (42)
0
where 0 is a constant with units of time and is evaluated with If the chosen combination of variables is to be successful, Eqs.
time characteristic of each system; e.g. for a gaslift contactor (40) and (41) represent restrictions that should be met. For
0 might be bubble-formation time. The dimensionless form of convenience we take  = 2, thereby giving
our system becomes erf()
C() = . (43)
jC jC j2 C erf(A )
− F (T )Z = , (30)
jT jZ jZ 2 Eq. (42) is a Bernoulli equation and, therefore, by the substi-
tution
C = 1 at T = 0 for all Z, (31)
za G = h−1/2 (44)
C=1 for all T at Z = ZA = √ , (32)
D0 and the integrating factor
 
C=0 for all T at Z = 0, (33)
e2 F (T ) dT
= e2 (j ln A/jT ) dT = A(T )2 (45)
where
it takes the form
j ln A(T ) dh dA
F (T ) = . (34) A2 = −2hA + 4A2 . (46)
jT x,y dT dT
3922 B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929

We may readily solve Eq. (46) whereby it follows that where n is a constant equal to the ratio of A0 /A∞ which is
essentially less than unity and greater than zero, and  is a di-
A(T )
G(T ) = T , (47) mensionless variable taken as equal to s0 T . The price that
4 0 A2 () d + C4 is paid for a greater accuracy in analytical models is an in-
crease in complexity. The above analysis represents a compre-
where C4 is a constant that must be zero to satisfy Eq. (41). hensive expression for the average mass transfer coefficient in
Thus, any fluid–fluid system in which turbulent motion, bulk motion,
A(T ) and a finite depth of penetration exist. All previous mass trans-
G(T ) = T . (48) fer theories discussed earlier can be developed as special cases
2
4 0 A () d of the above treatment. It can be seen that the overall mass
transfer coefficient is a function of the diffusion coefficient, D,
To obtain the complete solution of Eq. (30) we should recall
the fractional rate of surface renewal, s, and the depth of pene-
that  = Z × G(T )
tration, za . The main drawback to such a complicated expres-

T sion for mass transfer coefficient is having an error function
erf Z × A(T ) 4 0 A2 () d in denominator which makes the integral of Eq. (54) too con-
C=  , (49) voluted to be employed for practical purposes; nevertheless, in
T
erf ZA × A(T ) 2
4 0 A () d many cases some assumptions help to circumvent the difficulty
of the problem. When the scale of turbulence is high, as in the
where ZA is a constant regarding to za in which C =1. This ex- case of many practical systems, the rate of production of fresh
pression satisfies Eqs. (32)–(33). It also automatically satisfies surface increases and produces infinitely deep liquid elements
Eq. (31) if one substitutes A(T ) from Eq. (18) into Eq. (49). corresponding to the thickness of eddies due to the short ex-
Returning to the concept of age–surface distribution function, posure times. Hence za approaches infinity and the error func-
the instantaneous rate of absorption for surface elements having tion becomes close to unity. Moreover, regarding the fact that
age  and film area () d is 0 < n < 1, the integrand in Eq. (54) finds virtually the value of
2 (Maple, 2003) when the error function is equal to unity, and
(T ) dT
dNA,z = jA,z  ∞ , (50) thus
0 (T ) dT 
4Ds
where KL,av = . (55)
  
jcA  jC  jZ
jA,z = −D = (c − c )D (51) This is a new model of time-average overall mass transfer co-
jz z=0 jZ Z=0
A,i A0
jz efficient coming out of the equation of continuity in the case
hence that the two immiscible fluid phases undergo turbulent motion
 with one-dimensional diffusion and convective motion in the
Ds 2 0 same direction. It is noteworthy to recall that Eq. (49) was
dNA,z = (cA,i − cA0 ) achieved by ascribing change of time-dependent exposed sur-

⎡  ⎤ face area and increase of intensity of turbulence to convection
T
2 motion, as expressed in Eq. (23), where vz is convective molar

A(T ) 0 A () d ⎥
×⎢⎣  ⎥

average velocity. As a result, s is a representative of the scale
T of turbulence vis-à-vis convective motion that renews the sur-
erf ZA A(T ) 4 0 A2 () d
face of the liquid by either stirring or convection, and hence
× exp((−s0 )T ) dT Eq. (55) embodies the influence of both turbulent and convec-
= (cA,i − cA0 )KL (T ) dT , (52) tive motions. In a special case that mass transfer takes place into
a stagnant fluid of infinite depth, Eq. (55) changes to the model
where KL is the local mass transfer coefficient at a given time on of penetration theory. The time-average overall mass transfer
the interface. Hence the average flux over the time of exposure flux becomes
per unit area of turbulent surface is 
 ∞ 4Ds
NA,av = (cA,i − cA0 ). (56)
NA,av = (cA,i − cA0 ) KL (T ) dT 
0
= KL,av (cA,i − cA0 ). (53) Provided no chemical reaction occurs, a different significance
is merely attached to the overall mass transfer coefficient while
To solve Eq. (53) through Eq. (52), A(T ) should be substituted all expressions derived for the rate of mass transfer are simi-
from Eq. (18) into Eq. (52) and integrated when T changes from lar to that based on the film theory. For systems of practical
zero to infinity.

  ∞ − + (1 + n))e−
Ds (−e −0.5e−2 + 2(1 + n)e− + (1 + n)2  − 1.5 − 2n
KL,av =    d, (54)
 0 erf (s × z2 )/4D(−e− + (1 + n)) −0.5e−2 + 2(1 + n)e− + (1 + n)2  − 1.5 − 2n
a
B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929 3923

importance, average mass transfer coefficients can be evaluated than 0.5, where H is the average distance to where eddies ap-
experimentally in such an easy way without being concerned proach the interface. In this case, the length and velocity scales
about parameters like D, s, za , if the average flux is obtained of the eddies for isotropic turbulence can be written as
from the general expression of Eq. (53). Having selected an 1/4
3
appropriate system, the equation of continuity for species A in l= , v = ( ˙)1/4 , (61)
a binary system is written over the volume of a shell in the form ˙
dNA,z dc where = / is kinematic viscosity and ˙ is viscous dissi-
+ (1 − p ) = 0. (57)
dz d pation of energy per unit mass of eddies. The connotation of
Kolmogoroffs’ time scale of eddies in viscous dissipation range
It should be noted that Eq. (57) is only valid when absorption
can be used in the jargon of surface-renewal theory, in the sense
takes place in the z direction with no chemical reaction. Assum-
that s is defined (Banerjee et al., 1968)
ing jz ≈ z = a −1 , it may be written at any given instant as
v
dc s≡ . (62)
−KL a(cA,i − c) + (1 − p ) = 0, (58) l
d
Substitution of Eq. (61) for v and l into Eq. (62) yields
where a is the specific surface area, i.e., the interfacial surface 
area divided by the bulk volume of fluid–fluid contactors. The ˙
s= . (63)
negative sign of the expression on the left-hand side indicates
that the net rate of mass transfer changes inversely with depth.
Energy dissipation rate is calculated as follows (Wang et al.,
We may readily solve Eq. (58) by rearranging and getting an
2005):
integral from zero to time 
 c  g
dc KL a 0 d ˙ = Usg − Usl g. (64a)
= (59) 1 − g
c0 (cA,i − c) (1 − p )
In bubble columns and gaslift systems with low superficial
then, liquid velocity (Usl ≈ 0) the energy dissipation rate can be
cA,i − c (−KL a) approximated as
ln = . (60)
cA,i − cA0 (1 − p ) ˙ = Usg g. (64b)
The experimental data might be obtained to evaluate KL a from Combination of Eqs. (63) and (55) provides an expression for
Eq. (60), while finding a good correlation for s as a function s, and hence the time-average overall mass transfer coefficient
of measurable parameters, e.g. fluid–fluid skip velocity, super- becomes
ficial velocities of two fluid phases, and diameter of fluid par-   1/2
ticles, provides a good picture based on the surface renewal- 4D Usg g
KL,av = . (65)
stretch model to predict the behavior of fluid–fluid contactors 
in different situations.
On the evidence of literature, Eq. (60) seems virtually am- In deriving this expression it has been assumed that the contact-
biguous by being directly derived from the film theory. This ing phases are Newtonian fluids with gas bubbles composing
happens because the net flux of mass transfer in the film theory the dispersed phase and that the rate of surface renewal and the
also follows the same general expression as in Eq. (53). The rate of energy dissipation follow the Eqs. (62) and (64b), re-
accuracy of Eq. (55) can be easily scrutinized by analyzing the spectively. The latter is almost certainly an oversimplification,
effects of the scales of turbulence and convection motion on but the approximations are probably no more drastic than that
the overall mass transfer coefficient. involved in real cases of mass transfer. Reviewing the literature
gives a general indication of the behavior which is expected
3.2. Mass transfer coefficient correlation for gas bubbles from this type of systems. Representing the non-Newtonian be-
havior of fluids by the power-law model,
Consideration is limited in the first place to the cases in
=K n
(66)
which the dispersed phase is in the gas state.
Due to its great acceptance, the determination of the mean leads to the expression
fractional rate of surface renewal, s, through physical and hy-  1/2
drodynamic arguments has been the subject of much research 4D Usg g 1/(1+n)
interest (Banerjee et al., 1968; Skelland and Lee, 1981). In this KL,av = . (67)

work, the value of the mean rate of production of fresh surface
has been determined without recourse to observations of mass Similarly the analogs of Eq. (61) are
transfer rates, as used in previous studies (Banerjee et al., 1968). 1/2(1+n)
3
Harriott (1962) showed that the concept of surface-renewal
√ the- l= , v = ( ˙n )1/2(1+n) (68)
ory could only be expected when the group H / D/s is less ˙2−n
3924 B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929

Table 1
A summary of investigator designs and operational variables for Fig. 2

Reference Liquid Dc × 103 × 102 C8 Ave. Abs. Dev.


(m) (Pa s) (kg/m3 ) (N/m) from Eq. (80) (%)

Kawase et al. (1987) Water 0.23 0.92 991 7.02 1.326 15.2
0.76 1.571 28.04

Nakanoh and Yoshida (1980) Water 0.1455 0.80 995 7.1 0.851 21.19
10% sucrose 1.0815 1030 7.1 1.044 1.65
30% sucrose 2.543 1130 7.1 8.13 26.1
50% sucrose 10.85 1230 7.1 0.512 50.44

Shamlou et al. (1995) Fermentation 0.317 0.894 997.1 7.26 0.52 22.83
Broth

El. Temtamy et al. (1984) Water 0.15 1 1000 7.26 0.687 23.1

Koide et al. (1983) Water 0.14 0.8941 997 7.196 0.682 57.1
100 mol/m3 BaCl2 0.9185 1016 7.242 0.846 23.73
50% glycerol 5.93 1141 6.688 1.496 30.09

Nakao et al. (1983) Water 0.6 1.73 998.5 7.39 1.838 37.09

Jackson and Shen (1978) Water 1.83 1.224 999.4 7.45 1.16 12.1
7.62 1.224 999.4 7.45 0.959 7.2

in which = K/ . Moreover, in both Newtonian and non- Similar correlations have been proposed by other researches in
Newtonian solutions the average shear rate varies about pro- which the range of the exponent  varies from 0.4 to 1 (Durst
portional to superficial gas velocity (Nishikawa et al., 1977) et al., 1979). Hence,

= 5000Usg . (69) ˙0.4 0.6


1− d
0.25
a = C7 0.6
p . (74)
This has been regarded, for many purposes, as a useful rela- c
tionship for derivation of mass transfer correlations (Nakanoh In the case of gas bubbles, Godbole et al. (1984) proposed
and Yoshida, 1980; Deckwer et al., 1982, Godbole et al., 1984). the following correlation for gas hold up in non-Newtonian
If unit volume of a two-immiscible-fluid mixture contains a solutions:
dispersed phase volume of p made up of n fluid particles of
diameter dp , then n = p /(dp3 /6). If the interfacial area in the −0.19
g = 0.255Usg
0.6
eff , (75)
unit volume is a, then n = a/dp2 . Equating the two expressions
for n provides the specific area where eff is defined as follows:

6p =K n−1
. (76)
a= (70) eff
dp
Substituting Eqs. (69) and (76) into Eq. (75) yields
and
6g g = (0.255 × 50000.19(1−n) )K −0.19 Usg
(0.79−0.19n)
. (77)
a= (71)
dBM This correlation agrees reasonably well with a vast body of lit-
when gas bubbles comprise the dispersed phase. For non- erature data dealing with bubbles (Kawase et al., 1987). Assum-
coalescing conditions, the size of gas bubbles produced in ing  = 21 in Eq. (74) and combining Eqs. (64), (74), and (77)
gas–liquid dispersion has been determined from a balance gives a new expression for specific gas–liquid interfacial area
between surface tension forces and those due turbulent fluctu-
g 0.4 0.6 U (8−n)/10 0.25
ations (Calderbank, 1958) a = (0.5 × 50001−n/10 )C7
sg d
. (78)
0.6 K 0.1
0.6 c
dBM = C5 . (72) Combining Eqs. (78) and (67) and rearranging into dimension-
˙0.4 0.6
less groups lead to a correlation for the volumetric mass trans-
As a result Calderbank (1958) proposed the following expres- fer coefficient
sion for fluid particle size, dF M :
C8 0.5 (4n+17.5)/10(n+2)
0.6 0.25 Sh = 5000(1−n)/10 √ Sc Re
 d 
dF M = C6 
0.6 p
. (73)
˙0.4 c × Bo F r
0.6 (5n−2)/15(n+1)
, (79)
B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929 3925

Table 2
A summary of investigator designs and operational variables for Fig. 3

Reference Liquid n Dc K × 102 C8 Ave. Abs. Dev.


(m) (Pa s) (kg/m3 ) (N/m) from Eq. (79) (%)

Kawase et al. (1987) Carbopol-1 0.82 0.23 2.78 × 10−2 992 6.92 0.522 3.16
CMC-2 0.476 2.32 992 7.11 0.109 17.37

Nakanoh and Yoshida (1980) 0.3% CMC 0.97 0.1455 1.15 1000 7.1 0.679 37.31
0.5% CMC 0.985 1.626 1000 7.1 0.696 40.31
1.0% CMC 0.835 12.11 1001 6.84 0.356 59.46

Godbole et al. (1984) CMC-1 0.697 0.305 0.095 1000 7.3 0.492 27.33
CMC-2 0.654 0.184 1000 7.25 0.349 20.65
CMC-3 0.668 0.256 1002 7.32 0.262 3.41
CMC-4 0.607 0.526 1003 6.84 0.192 3.47
CMC-5 0.492 2.816 1005 6.77 0.086 18.19

El. Temtamy et al. (1984) Yeast-B 0.82 0.15 9 × 10−3 1006 9.67 0.624 12.37
0.35% CMC 0.59 0.305 0.947 1002 7.10 0.189 4.62

Joseph et al. (1984) 3% CMC 0.5242 1.434 1004 7.12 0.127 3.07

108
Shamlou et al. (1995)
Kawase et al. (1987), Dc=0.76m
Kawase et al. (1987), Dc=0.23m 5%
El. Temtamy et al. (1984)
-2
Koide et al., (1983) 5%
10 7 Nakao et al. (1983) +2
Nakanoh & Yoshida (1980)
Jackson & Shen (1978), Dc=7.62m
Jackson & Shen (1978), Dc=1.83m
Shexp

106

105

5x104

5x104 105 106 107 108


Shcalc

Fig. 2. Experimental Sherwood number for Newtonian fluids vs. predicted Sherwood number from Eq. (80).

where function of a number of variables including the superficial gas


2−n
Dcn Usg velocity, the dispersed and continuous phase properties, and the
KL aD 2c K/ Dc1−n
Sh = , Sc = , Re = , bubble size distribution. To keep our scope sufficiently broad, a
D DU 1−n
sg K/ comprehensive review of literature abounded with mass transfer
2
gD 2c Usg data for both Newtonian and non-Newtonian fluids has been
Bo = and F r = . presented in Tables 1 and 2, respectively. The values of C8
Dc g
have been determined for each set of mass transfer data through
For Newtonian fluids where n = 1, Eq. (79) reduces to lack of reliable data for specific gas–liquid interfacial area.
C Accordingly, the constant C8 can be iteratively approximated as
Sh = √8 Sc0.5 Re0.72 Bo0.6 Fr0.1 , (80)

C8 = 1.022n3.3 . (81)
in which C8 is the value of C8 when n in Eq. (79) is equal to
unity. The exponential dependency of KL a on Usg is 0.95. It If the value of the integrand in Eq. (54) were to be exactly
can be seen that the volumetric mass transfer coefficient is a equal to 2 and no assumption was made to obtain Eq. (80),
3926 B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929

107
Kawase et al. (1987)
Nakanoh & Yoshida (1980)
9%
Godbole et al. (1984) -1
El. Temtamy et al. (1984)
Joseph et al. (1984) 9%
+1
106
Shexp

105

104
104 105 106 107
Shcalc

Fig. 3. Experimental Sherwood number for non-Newtonian fluids vs. predicted Sherwood number from Eq. (79).

i.e., the one in Eq. (64b), then the coefficient on the right- for which the surface renewal rate would be determined in terms
hand side of Eq. (81) would be expected to be 1 rather than of hydrodynamic parameters, is sparse. Skelland and Lee (1981)
1.022. The proposed correlation, Eq. (80), and the experi- suggested a periodically varying rate of surface renewal, rea-
mental results obtained by various workers for Newtonian soning that the condition required to maintain equal degree of
fluids are compared in Fig. 2. The upper and lower dashed turbulence at various locations in vessels for circulating drops
lines represent the negative and positive deviations from of the dispersed phase appears to be lacking. In other words,
Eq. (80). The results are in satisfactory agreement over the the degree of turbulence changes and finds its maximum value
wide range of 0.14mDc 7.62m with the average abso- near the impeller, minimum in the corners, and of intermedi-
lute deviation of only 25%. Absolute deviation is defined as ate strength in regions between these locations. The proposed
|(Calculated value − Experimental value)/Experimental value| relationship for the varying rate of surface renewal is
throughout this study. A very similar expression was introduced
2
by Kawase et al. (1987) s = sav + ϑ sin , (83)
c
0.8
Sh = √ Sc0.5 Re0.75 Bo0.6 Fr0.12 . (82) where c is the average circulation time of the droplets around

the vessel. This concept is not obviously in harmony with
Considering Higbie’s penetration theory as a particular form Dankwerts (1951) theory with its constant surface-renewal rate
of Eq. (55) and the theoretical basis for derivation of Eq. (82) and the one developed in this work. Substituting Eq. (83) into
is deemed to be the main reason for this close resemblance. Eq. (14) yields
Eq. (82) and other similar correlations (Nakanoh and Yoshida,
1980; Akita and Yoshida, 1973) approve the plausible picture ϑc 2
() = C9 exp −sav  + cos , (84)
of the model proposed in this work. Fig. 3 is also a plot of 2 c
the predicted values of Sherwood number for non-Newtonian
vs. those observed experimentally. Even though the exponential
Table 3
dependency of KL a on Usg , Dc , D, K/ , and is quite more The specifications of five systems studied for drops
complicated for non-Newtonian fluids, it provides much lower
average absolute deviation of about 19% compared to previous System Dispersed phase Continuous phase Solute
published correlations (Kawase et al., 1987). Figs. 2 and 3 show 1 Dimethyl siloxane Distilled water Heptanoic acid
how the points are evenly distributed around the diagonal. 2 Dimethyl siloxane Distilled water +20% Heptanoic acid
Colonial pure cane sugar
3.3. Mass transfer coefficient correlation for drops 3 Dimethyl siloxane Distilled water +30% Heptanoic acid
Colonial pure cane sugar
To date, we only considered the cases in which the dispersed 4 Ethyl acetate Distilled water Heptanoic acid
5 Benzaldehyde Distilled water Heptanoic acid
phase is a gas. The information on mass transfer between drops,
B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929 3927

10

0%
-2
System 1.
System 2.
0%
System 3. +2
System 4.
System 5.
KL,av /(ND) 0.5Exp

1 10
KL,av /(ND)0.5Calc
√ √
Fig. 4. Experimental KL,av / N D vs. predicted KL,av / N D from Eq. (88).

Table 4
A summary of investigator designs and operational variables for Fig. 4
3 3 × 102 D × 1010
System c × 10 d × 10 c d C10 Ave. Abs. Dev. from
(Pa s) (Pa s) (kg/m3 ) (kg/m3 ) (N/m) (m2 /s) Eq. (88) (%)

1 0.0010 0.0019 1000 873 0.039 6.01 4.857 × 10−7 16.72


2 0.0018 0.0019 1087 873 0.032 5.66 3.190 × 10−7 43.24
3 0.0029 0.0019 1131 873 0.033 4.02 4.399 × 10−7 12.50
4 0.0010 0.00046 1000 894 0.006 6.01 7.037 × 10−7 35.95
5 0.0010 0.0014 1000 1041 0.015 6.01 7.715 × 10−7 40.14

Internal diameters of vessels: 0.21 m and 0.246 m. Diameters of impellers: 0.076 m and 0.106 m.

where C9 is obtainable from Eq. (15). Thus, Taking s ≈ sav and substituting directly Eq. (86) into
 ∞ Eq. (55) yields
A∞ ϑc 2
= exp −sav  + cos d. (85)
C9 0 2 c KL,av 2C10 Di 0.592
√ = √ −0.502 Re1.328
N . (87)
This integral has been calculated and employed to evaluate the DN  d
DT
mass transfer coefficient for agitated vessels without bulk mo-
tion (Lee, 1978). In the case of agitation accompanied with bulk The experimental values of C10 have been tabulated in Table 4
motion, A() should be evaluated afresh from Eqs. (84), sub- over 120 runs for five different mass transfer systems analyzed
stituted in Eq. (52), and finally integrated from zero to infinity by Lee (1978). The five systems studied are listed in Table 3.
to evaluate the time-average overall mass transfer coefficient; By averaging out the value of C10 from mass transfer data Eq.
as presented in the case of constant surface-renewal rate in this (87) changes to
paper. For the case that H = DT , a general correlation for sav
obtained from statistical analysis method has been developed KL,av 8.942×−7 −0.502 Di 0.592
√ = √ d Re1.328
N . (88)
by Skelland and Lee (1981) as follows: DN  DT
0.592
√ Di This correlation contains the least number of dimensionless
sav = C10 −0.502
d Re1.328
N N 0.5 , (86) groups and has an average absolute deviation of only 19.46%.
DT √
The comparison of the experimental values of KL,av / DN
where ReN = c N D 2i / c , and C10 is constant that should be with the proposed correlation is shown in Fig. 4 (Tables 3
determined from mass transfer data. and 4).
3928 B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929

4. Conclusion NA overall flux of species A relative to a phase bound-


ary, mol/m2 s
A brief synopsis of the former mass transfer theories shows
that the mathematical model developed in this paper is pos- NA,av time-average overall flux of species A relative to a
tulated closer to the truth than the conventional picture of an phase boundary, mol/m2 s
undisturbed layer at the surface of the fluid and of those theo- r average rate of exchange of the film surface, m2 /s
ries lacking in turbulent or bulk motions. Nevertheless, it will Re Reynolds number, Dc Usg /
never seem entirely self-evident unless the time-average overall Re Reynolds number defined by Eq. (79)
mass transfer coefficient in mass transfer processes of practical ReN impeller Reynolds number, c N D 2i / c
interest approximates to Eq. (55) through qualitative observa- s fractional rate of surface renewal, 1/s
tions on stirred and flowing fluids. In this regard, a vast body of sav average rate of surface renewal, 1/s
literature dealing with bubbles and drops was reviewed to show Sc Schmidt number, D/
that it is possible to shed new light on systems in which turbu-
Sc Schmidt number defined by Eq. (79)
lent motion exists contemporaneously with bulk motion, while
Sh Sherwood number defined by Eq. (70)
intensity of turbulence is the sequel to intensity of convection.
However, the terminology of the stagnant film hypothesis would T dimensionless time variable defined by Eq. (29)
probably be retained as a matter of convenience in some cases. Usg superficial gas velocity, m/s
Usl superficial liquid velocity, m/s
v molar average velocity; velocity scale of eddies in
Notation viscous dissipation range, m/s
a specific interfacial surface, m2 /m3 x rectangular coordinate, m
A exposed area of time-dependent surface, m2 xA dimensionless mole fraction of species A
A0 initial area of the surface exposed to gas, m2 y rectangular coordinate, m
A∞ total area of the surface exposed to gas, m2 z rectangular coordinate, m
Bo Bond number defined by Eq. (79) za distance from the interface to the bulk of fluid, m
cA time-dependent molar concentration of species zF effective film thickness based of the film theory, m
A, mol/m3 Z dimensionless position variable defined by Eq. (28)
cA,i molar concentration of species A at the interface, ZA dimensionless position variable defined by Eq. (32)
mol/L3
cA0 molar concentration of species A at the bulk of Greek letters
fluid, mol/m3
c∗ saturation concentration, mol/m3 shear rate, 1/s
C dimensionless concentration profile defined by ∇ vector differential operator defined by Eq. (9)
Eq. (27)  unit vector, m
Ci constant  volume fraction, dimensionless
˙ energy dissipation rate per unit mass, W/kg
D molecular diffusivity, m2 /s
 time, s
Dc column diameter, m
c circulation time around vessel, s
Di impeller diameter, m
0 characteristic constant defined for each system; e.g.
DT tank diameter, m
for gas–fluid contactors 0 might be bubble forma-
F defined by Eq. (34)
tion time, s
Fr Froude number defined by Eq. (79)
ϑ amplitude of period, 1/s
g gravitational acceleration, m/s2
kinematic viscosity, m2 sn−2
G defined by Eq. (35)
viscosity, kg/m s
h defined by Eq. (44)
density, kg/m3
H̄ average distance of approach of eddies to interface,
m interfacial tension, N/m
JA molar flux of diffusion of species A relative to the  dimensionless variable used in Eq. (54)
molar average velocity, mol/m2 s  dimensionless variable defined by Eq. (35)
kL fluid-film mass transfer coefficient, m/s  surface–age distribution function, m2 /s
K consistency index in a power-law model, kg sn−2 /m Subscripts
KL,av time-average overall mass transfer coefficient, m/s
KL a overall volumetric mass transfer coefficient, 1/s c continuous phase
l length scale of eddies in viscous dissipation range, BM bubble mean (diameter)
m d dispersed phase
n dimensionless exposed surface area defined in Eq. FM fluid particle mean (diameter)
(54); flow index in a power-law model g gas
N impeller speed, rps p particles in general
B. Jajuee et al. / Chemical Engineering Science 61 (2006) 3917 – 3929 3929

Acknowledgments Higbie, R., 1935. The rate of absorption of a pure gas into a still liquid during
short periods of exposure. Transactions of the A.I.Ch.E. 31, 365–389.
Jackson, M.L., Shen, C., 1978. Aeration and mixing in deep tank fermentation
A. Margaritis, D. Karamanev, and M.A. Bergougnou ac- systems. A.I.Ch.E. Journal 24 (1), 63–71.
knowledge financial support from Imperial Oil Ltd., Sarnia, On- Joseph, S., Shah, Y.T., Carr, N.L., 1984. Two bubble class model for mass
tario, and Natural Sciences and Engineering Research Council transfer in a bubble column with internals. Institute of Chemical Engineers
(NSERC) of Canada. B. Jajuee acknowledges support by the Symposium Series 87, 223–230.
Kawase, Y., Halard, B., Moo-young, M., 1987. Theoretical prediction of
Western Engineering Scholarships.
volumetric mass transfer coefficients in bubble columns for Newtonian and
non-Newtonian fluids. Chemical Engineering Science 42 (7), 1609–1617.
King, C., 1966. Turbulent liquid phase mass transfer at a free gas–liquid
References interface. Industrial & Engineering Chemistry Fundamentals 5 (1), 1–8.
Koide, K., Sato, H., Iwamoto, S., 1983. Gas holdup and volumetric liquid-
Akita, K., Yoshida, F., 1973. Gas holdup and volumetric mass transfer phase mass transfer coefficient in bubble column with draught tube and
coefficient in bubble columns. Industrial & Engineering Chemistry Process with gas dispersion into annulus. Journal of Chemical Engineering of
Design and Development 12, 76–80. Japan 16, 407–413.
Angelo, J.B., Lightfoot, E.N., 1968. Mass transfer across mobile interfaces. Kozinski, A.A., King, C.J., 1966. The influence of diffusivity on liquid phase
A.I.Ch.E. Journal 14 (4), 531–540. mass transfer to the free interface in a stirred vessel. A.I.Ch.E. Journal 12
Angelo, J.B., Lightfoot, E.N., Howard, D.W., 1966. Generalization of the (1), 109–116.
penetration theory for surface stretch: application to forming and oscillation Lamont, J.C., Scott, D.S., 1970. An eddy cell model of mass transfer into
drops. A.I.Ch.E. Journal 12 (4), 751–760. the surface of a turbulent liquid. A.I.Ch.E. Journal 16 (4), 513–519.
Banerjee, S., Scott, D., Rhodes, E., 1968. Mass transfer to falling wavy liquid Lee, J.M., 1978. Mass transfer and liquid–liquid dispersion in agitated vessels.
films in turbulent flow. Industrial & Engineering Chemistry Fundamentals Ph.D. Dissertation, University of Kentucky, Lexington, KY.
7 (1), 22–27. Lewis, W.K., Whitman, W.G., 1924. Principles of gas absorption. Industrial
Calderbank, P.H., 1958. Physical rate processes in industrial fermentation, & Engineering Chemistry 16, 1215–1220.
Part I: the interfacial area in gas–liquid contacting with mechanical Nakanoh, M., Yoshida, F., 1980. Gas absorption by Newtonian and non-
agitation. Transactions of the Institution of Chemical Engineers 36, Newtonian liquids in a bubble column. Industrial & Engineering Chemistry
443–463. Process Design and Development 19, 190–195.
Dankwerts, P.V., 1951. Significance of liquid-film coefficients in gas Nakao, K., Takeuchi, H., Kataoka, H., Kaji, H., Otake, T., Mlyauchi, T.,
absorption. Industrial & Engineering Chemistry 43, 1460–1467. 1983. Mass transfer characteristics of bubble columns in recirculation
Deckwer, W.-D., Nguyen-Tien, K., Schumpe, A., Serpemen, Y., 1982. flow regime. Industrial & Engineering Chemistry Process Design and
Oxygen mass transfer into aerated CMC solutions in a bubble column. Development 22, 577–582.
Biotechnology and Bioengineering 24, 461–481. Nishikawa, M., Kato, H., Hashimoto, K., 1977. Heat transfer in aerated tower
Dobbins, W.E., 1956. The Nature of the Oxygen Transfer Coefficient in filled with non-Newtonian liquid. Industrial & Engineering Chemistry
Aeration Systems. Biological Treatment of Sewage and Industrial Wastes, Process Design and Development 16 (1), 133–137.
vol. 1, pt. 2-1. Reinhold, New York, pp. 141–148. Shamlou, P.A., Pollard, D.J., Ison, A.P., 1995. Volumetric mass transfer
Durst, F., Tsiklauri, G.V., Afgan, N.H., 1979. Two-Phase Momentum, Heat coefficient in concentric-tube airlift bioreactors. Chemical Engineering
and Mass Transfer in Chemical, Process, and Energy Engineering Systems, Science 50 (10), 1579–1590.
vol. 2, Washington, pp. 835–876 Skelland, A.H.P., Lee, L.M., 1981. Drop size and continuous-phase mass
El. Temtamy, S.A., Khalil, S.A., Nour-El-Din, A.A., Gaber, A., 1984. Oxygen transfer in agitated vessels. A.I.Ch.E. Journal 27 (1), 99–111.
mass transfer in a bubble column bioreactor containing lysed yeast Stewart, W.E., Angelo, J.B., Lightfoot, E.N., 1970. Forced convection in
suspensions. Applied Microbiology and Biotechnology 19, 376–381. three-dimensional flows: II. A.I.Ch.E. Journal 16 (5), 771–786.
Godbole, S.P., Schumpe, A., Shah, Y.T., Carr, N.L., 1984. Hydrodynamics and Toor, H.L., Marchello, J.M., 1958. Film-penetration model for mass and heat
mass transfer in non-Newtonian solutions in a bubble column. A.I.Ch.E. transfer. A.I.Ch.E. Journal 4 (1), 97–101.
Journal 30 (2), 213–220. Wang, T., Wang, J., Jin, Y., 2005. Theoretical prediction of flow regime
Harriott, P., 1962. A random eddy modification of the penetration theory. transition in bubble columns by the population balance mode. Chemical
Chemical Engineering Science 17 (3), 149–154. Engineering Science 60, 6199–6209.

You might also like