Journal of Membrane Science 473 (2015) 189200

Contents lists available at ScienceDirect

Journal of Membrane Science

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

Simulation and analysis of the performance of tubular enzymatic

membrane reactors under different congurations, kinetics
and mass transport conditions
R. Abejn, C.L. Gijiu, M.P. Belleville, D. Paolucci-Jeanjean, J. Sanchez-Marcano n
Institut Europen des Membranes, UMR 5635 ENSCM, UM2, CNRS Universit de Montpellier 2, CC 047, Place Eugne Bataillon, 34095 Montpellier, France

art ic l e i nf o a b s t r a c t

Article history: A model was developed to simulate tubular enzymatic membrane reactors under three different
Received 11 June 2014 congurations: dead-end, tangential ow with a porous enzymatic membrane and a non-permeable
Received in revised form enzymatic wall. The simulations were applied to analyze the inuence of reactor conguration, kinetics
25 August 2014
and mass transport conditions over the reactor performance in order to identify the main aspects to be
Accepted 10 September 2014
Available online 19 September 2014
taken into consideration for attaining optimal designs. The non-permeable enzymatic wall conguration
under the evaluated conditions seems to be more valuable than the dead-end case in terms of substrate
Keywords: conversion and the tangential conguration looked more favorable to promote the best conversion in the
Enzymatic membrane reactor permeate but not in the retentate. It was demonstrated that for a similar value, the Damkhler number
Modeling
can result in very dissimilar performances. The simulated results demonstrated that the most signicant
Simulation
variable of the global performance of the enzymatic membrane reactors is the reaction kinetics: fast
Sensitivity analysis
reactions attained very considerable conversion values under very different conditions.
& 2014 Elsevier B.V. All rights reserved.

1. Introduction
Enzymes are gaining relevance as biocatalysts in industrial
processes, as their capacities have been adapted during the last
50 years for successful production of some commodities, ne
chemicals and pharmaceuticals in an environmentally friendly
way (high reaction rates at room temperature and pressure with-
out auxiliary chemicals). Enzymes are specically useful to facil-
itate the synthesis and modication of organic molecules by
chemo-, regio-, and stereoselective bond-forming and -breaking
reactions [1,2]. But enzymes have to compete economically with
inexpensive traditional chemical processes which are nowadays
industrially established. Therefore, the recovery and reuse of
enzymes, which are generally expensive, becomes really essential
in most cases in order to assure the economic viability of the
processes [3,4]. Indeed, immobilization of enzymes is generally
required in industry.
In addition, immobilization can also improve enzyme perfor-
mance by increasing their activity, stability and selectivity and
allowing optimal process reaction conditions, such as acidity,
alkalinity, organic solvents or elevated temperatures [5,6]. Many
immobilization methods and protocols have been put into practice
n
Corresponding author. Tel.: 33 467149149; fax: 33 467149119.
E-mail address: Jose.Sanchez-Marcano@univ-montp2.fr (J. Sanchez-Marcano).
and are compiled in bibliography [7], but in a broad framework, all
different immobilization strategies can be categorized in one of
the following groups: adsorption, covalent linking, entrapment
encapsulation, cross-linking or afnity [8]. Membranes are con-
sidered as very attractive supports for immobilizing enzymes since
they can be placed in multifunctional membrane reactors, which
integrate biocatalytic reaction, product separation and enzyme
recovery [911]. Enzymatic membrane reactors (EMRs) can over-
come some of the negative aspects that are related to immobiliza-
tion and show some advantages over other alternatives, but some
disadvantages are still apparent as summarized in Table 1 [10].
EMRs contribute to the implementation of continuous operation
mode and avoid the intrinsic discontinuous nature of batch processes
and their problems. Nevertheless, although several continuous
enzymatic processes have been fully developed, improvement of
the entire processes is still required to maximize the benets derived
from their use [912]. Reaction engineering must provide solutions
to improve enzymatic processes at the commercial level [13].
Decisions made in reactor design have a very signicant impact on
the enzymatic process performance, but there are no simple and
direct procedures available to determine the optimal conditions [14].
These reaction conditions also determine the downstream processing
and, under the consideration of the high importance of the costs due
to downstreaming, have a deep impact in the economic viability of
the whole process [15].

http://dx.doi.org/10.1016/j.memsci.2014.09.020
0376-7388/& 2014 Elsevier B.V. All rights reserved.
190 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200

Table 1
Advantages and drawbacks of EMRs. Adapted from [10].
Advantages
Effective retention and reuse of catalyst
Continuous mode operation
Reduction in substrate/product inhibition
Final product free of enzyme
Enhanced enzyme stability when grafted
Decrease of diffusional limitations
Process intensication (single-step reaction
and separation)
which are more usual at industrial scale like the tangential cong-
uration. Moreover, a detailed theoretical analysis was carried out in
order to study the main process variables controlling the global
Drawbacks
performance and conversion of EMRs. For this purpose we based
Enzyme activity decay our simulations on the different enzymatic kinetics reported in the
Heterogeneous reaction literature. This analysis was carried out considering initially
conditions the Damkhler number, and then carrying out a careful analysis
Polarization layers
of the inuence of the different parameters on the conversion of
Membrane fouling
Possible less enzyme activity enzymatic membrane reactors in tangential conguration.
when grafted
2. Process modeling and case study resolution

2.1. Process modeling and presentation of the case study

In a preliminary 3D model for the simulation of enzymatic

Modeling and simulation are among the most useful tools to be
membrane reactors working in dead-end mode (feed stream
applied to reaction and process engineering. They complement
leaves as permeate stream) [29], we were able to represent, in a
purely empirical procedures, accelerate and simplify process
successful way, the hydrodynamics inside the membrane reactors
development and help to have a better process understanding
and the resulting ux proles, as well as the reactive regions
[16], so these tools should be applied to the complete set of
through membranes length and the corresponding concentration
chemical and biochemical processes [15,17], including some reluc-
proles. In this previous work, [29] the model was validated with
tant sectors such as the pharmaceutical or biotechnological ones
experimental results of the hydrolysis of butyl acetate with a
[18,19]. As modeling allows experimental effort reduction by
lipase immobilized on the internal surface of tubular ceramic
avoiding time, money and resources consuming experiments,
membranes. The model presented here is inspired in the previous
it has to be recommended for the investigation and analysis of
one but it has been modied and completed in order to simulate
the inuence of the different factors that have to be taken into
the behavior of other different congurations of the EMR, like an
account in the design of reactors based on enzymatic activity [20].
enzymatic non-permeable wall reactor (feed stream leaves only as
The objective of this work is the employment of computer
retentate stream) and in tangential mode (feed stream is split
aided process engineering tools such as modeling and simulation
between permeate and retentate streams). The congurations
to advance in the knowledge of the performance of enzymatic
considered are shown in Fig. 1.
membrane reactors under different design and operation condi-
Although the conguration under tangential mode could be
tions by building a three dimensional (3D) model of the system.
considered as the most relevant for industrial purposes, the other
The obtained results for the different analyzed congurations
congurations can also have promising applications. For instance,
provided a very valuable support to identify the most relevant
the dead-end conguration can be employed for situations with
aspects that have to be taken into account for attaining optimal
quite clean streams (no ltration is needed) where the permeation
conditions for each application. Steady state one dimensional
through the membrane involves a signicant improvement
models (1D) of membrane bioreactors utilizing free enzymes or
through the ow through membrane reactor concept [9]. The
conned in a volume have been developed by different groups
opposite case can be applied to the non-permeable wall: the
[2123]. Among these models, one of the most interesting model
permeation of the uid should be avoided because of, for instance,
is the model of Salzman et al. [23], who make a ne analysis of the
very high solids content which will drive to membrane fouling, in
effect of diffusional limitations that may prevail in this type of
such case the membrane acts only as a catalyst support, this case
reactor with the aid of two dimensionless groups, an effectiveness
can be approached to the honeycomb supports.
factor (related with reaction rates) and a second dimensionless
The 3D developed models were built with COMSOL Multi-
number related with the mass transport through the membrane.
physics. The modeling work was focused on mono-channel
In the literature, have been reported more recent 1D or 2D models
membranes because these membranes do not require long calcu-
analyzing the inuence of the main operation variables on the
lation times. In addition to the classical model considerations
performance of hollow-bers membrane reactors with entrapped
which integrate the ux through the inner compartments, like it
enzymes inside the porosity of the membranes [2428]. All of them
has been reported in the previous work named above; we studied
are based on interesting analyses of the balance between the reaction
here the relative inuence of the permeation rate (transmembrane
rates and mass transport with explicit equations and through various
ux through the different membrane layers in radial direction)
dimensionless numbers (Thiele, Peclet, Bodenstein).
and the reaction kinetics. These both parameters are essential in
In a rst published part of this work [29], we have reported a
order to determine theoretically the cases when the use of EMRs is
preliminary 3D model which was developed for a dead-end
interesting in terms of reaction conversion.
conguration and the enzymatic hydrolysis of butyl acetate by a
As illustrating case study, the butyl acetate hydrolysis reaction
lipase as model reaction. The enzymes were considered to be
to acetic acid and butanol catalyzed by immobilized Candida
grafted on the internal surface of the membranes. The model was
antarctica lipase B (CALB) was selected:
based on the coupling of mass balances, hydrodynamics of the
tubular membranes and enzymatic kinetics. The results in this C6H12O2 H2O-C2H4O2 C4H10O (1)
preliminary work were reported in terms of reaction productivity
which is not an interesting magnitude for chemical engineering This reaction was selected because of its very well-known
and design purposes. In addition any analysis with respect to other MichaelisMenten kinetics, which has facilitated the appearance
congurations or the controlling phenomena (enzymatic kinetics of previous studies about its implementation in EMRs.
or mass transfers) has not been carried out. In this second part and The mono-channel tubular membranes considered have 13 cm
for the same hydrolysis reaction, we complete the previous of effective length, and external and internal diameters of 1 and
reported model in order to consider other EMR congurations 0.7 cm respectively. They present an asymmetric structure with
 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200 191

Fig. 1. Reactor congurations: (a) Enzymatic non-permeable wall, (b) Tangential and (c) Dead end.

layer and enzymes grafted modied the original hydraulic perme-

ability of the ceramic membranes.

2.2. Model formulation

The main assumptions taken for this model are:

Steady-state and isothermal operation.

Enzymes are considered grafted uniformly and only on the
surface of the gelatin layer.
The enzymatic activity and permeate ow rate are constant.
The retentate and permeate ow rate are always in laminar
ow conditions (Relower than 1200).
The physical properties (viscosity, density) are constant under
different operating conditions.
Diffusion coefcients were estimated independently for the
product and substrate and then considered constant under
different operating conditions.

The 3D model was built in Comsol Multiphysics and taking in

Fig. 2. Schematic draw of the EMR and the symmetric view represented on
simulations. A: inner retentate compartment, B: gel layer, C: porous support.
consideration simultaneously:

1) radial and axial uxes in the inner side of the tubular mem-
branes and in laminar conditions;
2) reaction takes place only at the surface of the separative layer;
Table 2 3) radial ux occurs through the separative layer and then
Membrane layer characteristics [29]. through the membrane porosity;
4) the permeate ux is immediately depleted out of the reactor.
Filtration active gel layer Support

Thickness (m) 1.4  10  5 1.5  10  3 The experimental Lp reported in Table 2 allowed to calculate
Porous diameter (m) 8  10  7 1  10  5 the derived membrane permeability coefcients by using the
Porosity (  ) 0.37 0.37 HagenPoiseuille equation:
Tortuosity (  ) 1.0 1.0
Hydraulic permeability JP
3.2  10  10 5.6  10  7 LP 2
(m s  1 Pa  1) P
Material permeability (m2) 4.48  10  18
8.34  10  13

dP
2
LP U 3
32 U U

LP U U 4
a thin ltration layer of 8.0  10  7 m of mean pore size supported
on a thick porous ceramic support with a mean pore size of where JP is the permeate ux, P is the transmembrane pressure,
1.0  10  6 m. A very thin (less than one micron of thickness) is the porosity, is the tortuosity, dP is the pore diameter, is the
but continuous gelatin layer was deposited on the internal surface membrane thickness, and m is the dynamic viscosity.
of the ceramic membranes, the enzymes were immobilized on The liquid velocities in channels were calculated from station-
the surface of this very thin cross-linked gelatin layer. The applied ary incompressible NavierStokes equations for a laminar regime
immobilization protocol, was originally developed by this as follows:
research group for the immobilization of -chymotrypsin [30] U u Uu P U u U g 5
but successfully adapted to other enzymes like CALB [31] and
laccase [32,33]. In our preliminary work [29], we considered Uu 0 6
that the structural characteristics of the support were not where is the uid density, u is the velocity and P is the pressure.
affected by the gelatin deposition and that enzymes were largely In the porous matrix (ceramic and gel layers of the membrane)
located on the membrane surface. In this work we took the same Brinkman equations describe in an adequate way the mass
assumption. transfer [34,35]:
To simplify the geometry, thickness and pore diameter of each    
1 2U
layer were considered constant and equal to the average values Uu P U2 u   U U u U g 7
shown in Table 2, which includes also the hydraulic permeability 3
coefcients LP. The permeability coefcients were obtained from
Uu 0 8
experimental permeate uxes because the cross-linked gelatin
192 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200

The concentration eld C for each specie was calculated from 2.3. Models resolution
the mass balance and reaction rate taking into account the
reactant (butyl acetate) consumption (  vR) or product (acetic The explained mathematical model was implemented in COM-
acid) appearance (vR). As far as the reaction extend was based on SOL Multiphysics 4.4, by coupling free (Eqs. (5) and (6)) and
the reactant consumption for the modeling, the following equation porous media (Eqs. (7) and (8)) ows, transport of diluted species
can be employed: (Eq. (9)), and reaction kinetics (Eq. (12)) considering that the
enzymatic reaction takes place at the inner surface of retentate
u U C DX U 2 C  vR 9 channel. The resolution was carried out for steady state conditions.
The mono-channel simulated membrane reactors can be repre-
where DX is the diffusion coefcient. On the contrary of some
sented by an axisymmetric 2D scheme, the enzymes were con-
previous literature works, the diffusion coefcients in the liquid
sidered to be grafted on the internal surface of the membrane.
media were not considered identical but they were estimated
In order to optimize the calculation time, the simulations were
independently for each compound with the Wilke and Chang
done taking into account this symmetry and the results were given
correlation [36]: as the evolution of concentrations inside the tubular membrane
but representing only the 2D prole of the inner compartment and
M W 0:5 T
DX 7:4 U10  13 10 the membrane (retentate, gelatin thin layer and porous support)
V 0:6
M following the schematic draw shown in Fig. 2. The membrane
supports are 1.3  10  1 m long (L) and their internal and external
where is the association factor (2.6 for water), MW is the
molecular weight of solvent (water in this example), T is the
temperature, is the viscosity, and VM is the molecular volume of
each specie, this last parameters was calculated for each compo-
nent according to Prausnitz et al. [37]. The values of diffusion
coefcients calculated are presented in Table 3. Inside the pores of
the membrane layers, the diffusion coefcient values were cor-
rected with the porosity and tortuosity values (0.37 and 1 for the
proposed case, see for instance Table 2):

DXP DX 11

where DXP is the corrected diffusion coefcient. For further details
about the mathematical model, specically about the mass and
momentum balances related to Brinkman equations, bibliography
can be consulted [38].
Previously obtained experimental results about butyl acetate
hydrolysis were successfully tted to the MichaelisMenten kinetics
reaction model:

CS
vMM vMAX 12
CS K M

where CS is the substrate (butyl acetate) concentration, vMM is the

modeled reaction rate, vMAX is the apparent maximum reaction rate,
and KM is the apparent MichaelisMenten constant. The values for
these kinetic constants, determined experimentally in a previous
work [29], are equal to 1.572  10  4 mol s  1 m  2 and 20.19 mol m  3
respectively.
As the reaction rate for the immobilized enzyme was expressed
as a surface reaction rate, it was implemented in the program as a
ux discontinuity on the inner surface of the channel. If NXin is the
inlet ux of X species entering the surface, NXout is the outer ux
leaving the surface and NXR is the ux discontinuity due to
reaction, the following relationship can be dened among the
three terms:

N Xin N Xout 7 N XR 13

The reaction term is either negative or positive, depending if

the species are reactants or products of the reaction.

Table 3
Diffusion coefcient values.

Specie VM (m3 mol  1) DX (m2 s  1)

Butyl acetate (A) 1.562  10  4 1.17  10  9

Acetic acid (B) 6.84  10  5 1.92  10  9 Fig. 3. Axial ow velocity (m/s) prole for the different reactor congurations:
n-Butanol (C) 1.072  10  4 1.47  10  9 (left) Enzymatic non-permeable wall; (center) Tangential conguration; and (right)
Dead-end conguration.
 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200 193

Table 4
Initial and boundary conditions.

Reactor conguration Dead- Enzymatic Tangential

end wall

Feed concentration Cs0 (mol m  3) 50 50 50

Outer permeate radial velocity (m s  1) 2.2  10  5 2.2  10  5
3 1
Feed ow (m s ) 9  10  8 9  10  8 18  10  8

diameters are 7.0  10  3 and 1.0  10  2 m respectively. The thick-

ness of the gel layer over the internal face of the membrane
support is 1.4  10  5 m (C).
All the enzymatic membrane reactors were integrated with the
corresponding initial and boundary conditions. The transmembrane
pressure at the channel input is 1  105 Pa. Feed concentration, outer
permeate radial velocity, and feed ow depend on the conguration
and the values are compiled in Table 4.
The velocity and concentration gradients near the membrane
surface must be calculated on a very ne mesh because of the high
hydraulic and concentration gradients. As a consequence, models
with around one million degrees of freedom have to be solved.
The non-linear differential equations built in Comsol Multi-
physics are presented in the annex; here below we present only
the general equations. The model was solved in nite elements,
and taking into consideration the 3D geometry of tubular mem-
branes; however, taking into account the axysimetry only are
represented 2D in simulations views of the half of each type of
enzymatic reactor studied. The global conversions (Eqs. (15) and
(16)) presented in Section 3, were calculated after integration of
the concentrations and ows in the different volumes of the
reactors. Simulations curves which concern the Damkhler num-
bers were built with a homemade software but using the inte-
grated values of ows and concentrations obtained with Comsol
and combined with the kinetics parameters.

3. Results and discussion

3.1. Comparison of different reactor congurations

The model had been validated in a previous work for the dead-
end conguration [29]. The experimentally obtained results were
compared to the simulated ones and, as the found differences
were not signicant, the model was concluded as valid. Therefore,
the same modeling approach was followed for the other two
congurations investigated in this work (enzymatic wall and
tangential ltration conguration). In all simulation gures shown
below, the enzymatic layer and membrane support are placed at
the right side of the illustrations as shown in the schematic
representation in Fig. 2.
The hydrodynamics proles, expressed as axial and radial ow
velocities, are shown in Figs. 3 and 4 respectively. The ow
limitations that are imposed under enzymatic wall (no permeate
stream) and dead-end (no retentate stream) congurations have
clear impact over the hydrodynamics. The enzymatic wall mode is
characterized by null radial velocity as uid is not allowed to
permeate through the support, so a totally developed laminar
ow along the axial direction can be observed. By contrast, the
dead-end mode does not allow retentate outlet and all the feed
stream has to permeate. As a consequence, the axial velocity
becomes negligible near the end of the inner channel whereas
the radial velocity is almost constant along the membrane length
(except for the both ends where some boundary effects appear).
Nonetheless, the radial velocity magnitude is lower than the
Fig. 4. Radial ow velocity (m/s) prole for the different reactor congurations:
(left) Enzymatic non permeable wall; (center) Tangential conguration; and (right)
Dead-end conguration.
axial velocity and only at the end of the inner channel the radial
component dominates over the axial one.
The tangential mode can be analyzed as an intermediate
situation between the two extreme cases. The feed ow of the
tangential mode is the sum of the one of the other two cases and,
consequently, the corresponding ows of permeate and retentate
streams are similar to the resulted ones in the extreme cases.
The ow prole inside the membrane is analogous to that
obtained for the dead-end conguration whereas the ow prole
in the inner channel looks like the enzymatic wall case. Never-
theless, some differences are visible. The axial velocity prole
shows how the higher feed ow implies higher initial velocity.
However, the velocity is reduced along the reactor length as a
consequence of the mass transfer by permeation. Otherwise, the
radial velocity prole differs only in the boundary effect on the
inner channel outlet when compared to the dead-end case.
The concentration proles of the substrate A (butyl acetate) are
shown in Figs. 5 and 6. To have a better understanding of the
results, three different conversion expressions are dened: global
194 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200

Fig. 6. Detail of the substrate (A) concentration (mol m  3) prole for the outlet of
the enzymatic non-permeable wall conguration.

Fig. 5. Substrate (A) concentration (mol/m3) prole for the different reactor
congurations: (left) Enzymatic non-permeable wall; (center) Tangential cong-
uration; and (right) Dead-end conguration.

conversion (Xglobal), permeate conversion (Xperm) and retentate

conversion (Xret): Fig. 7. Conversion of butyl acetate in function of the Qperm/Qret ratio.

 
F 0 U C 0  F perm U C perm F ret U C ret
X global 100 14 the substrate is attained close to the wall, with a minimal
F 0 U C 0 concentration of 21.4 mol m  3.
C 0  C ret The global conversion in the tangential conguration is 3.5%,
X ret 100 15
C0 but it has to be kept in mind that the treated ow is just the
double of the feed ow of the dead-end conguration. The
C 0  C perm conversion for the permeate stream is 6.9%, very close to the value
X perm 100 16 obtained for the dead-end case. However, the conversion for the
C0
retentate stream is very low, only 0.1%. Indeed, the major part of
Global conversions were obtained by integrating the concen- the uid that gets in contact with the enzymatic layer placed at
tration proles and ows through the different reactors volumes. the membrane internal surface, crosses the membrane as soon as
The enzymatic wall conguration is slightly more effective than it had reacted and leaves the system as permeate, not as retentate,
the dead-end one (global conversions of 7.2% and 7.0% are reached so that the uid has not as much contact time as in the case of the
respectively). In the enzymatic wall reactor, a high conversion of non-permeable enzyme support.
 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200
Fig. 8. Inuence of the transport rate in terms of axial and radial Damkhler number over the conversion of butyl acetate in EMR in a non-permeable enzymatic wall
conguration (a) and dead-end conguration (b).
In order to analyze more in detail for the tangential congura-
tion the inuence of the relative ows of retentate and permeate,
we carried out simulations considering different ratios of perme-
ate/retentate (Qperm/Qret) and calculated the hydrolysis conver-
sions reached in each compartment. As it was explained
previously for these simulations, global conversions and ows
were integrated through the different volumes of the reactors.
The results presented in Fig. 7 show that at very low Qperm/Qret
ratio the axial and global conversion are very low, in such case the
axial ow rate is very high and the contact time is very short,
whereas the permeate ow rate is low and then even if its
conversion is not negligible it has a very low inuence on the
global conversion because its contribution to the total volume is
very small. On the contrary, at very low Qperm/Qret ratio we
approach the conditions and conversion of the dead-end cong-
uration. Indeed, the conversion of the permeate ow is always
higher than the retentate ow because all the permeate has been
in contact with the biocatalyst, whereas only a small fraction of
the retentate (near the enzymatic membrane) is really in contact
with the grafted enzymes. These results are in good agreement
with the concept of ow-through membrane reactor which
considers that the physical approaching of substrates and catalyst
or biocatalyst by the permeation is fundamental for the effective-
ness of the membrane reactors.
After the analysis of these initial simulated results for the
hydrolysis of butyl acetate, two ideas can be outlined. On the one
hand, the enzymatic wall conguration under these particular
conditions seems to be more advantageous than the dead-end
case in terms of substrate conversion, but under this situation
ltration is not carried out. On the other hand, in terms of
global conversion for porous membranes, both tangential and
dead-end congurations can reach similar results but tangential
conguration allows a continuous ltration while giving the
possibility of avoiding the formation gel layers and a continuous
permeate ow.
The Da number has been frequently used to evaluate and study
the sensitivity of enzymatic or other types of reactors performance
by investigation of the effect of variation of Da number [39,40].
Thus, the analysis carried out above can also be made considering
the Damkhler number (Da) which is a classical dimensionless
number used to relate the chemical reaction rate to the transport
phenomena rate occurring in a reactor. In the case of membrane
reactors, the Da has also been studied to determine if the process
is controlled by the biological or chemical kinetics or by the mass
transport through the membrane [9]. For this purpose we con-
sidered only two congurations: the non-permeable enzymatic
wall and dead-end. Both congurations present two different
uxes: radial and axial which can be described by two different
Da numbers. The following expressions have been dened for the
195
axial and radial Damkhler numbers:
Reaction rate v UC S =C S K M
DaZ MAX 17
Axial mass rate Q RET UC S =ARET
Reaction rate vMAX UC S =C S K M
DaR 18
Radial mass rate Q PERM UC S =APERM
where QRET and QPERM are the ows of the retentate and permeate
streams respectively, and ARET and APERM are the outlet areas of the
retentate and permeate streams, which correspond to the surfaces
perpendicular to the ow direction, i.e. the cross section of the
inner channel for ARET and the external surface of the membrane
support for APERM.
For the simulations we calculated DaZ and DaR and conversions
considering the actual measured kinetics of butyl acetate hydrolysis
(Eq. (12), where vMAX 1.572  10  4 mol s  1 m  2 and KM
20.19 mol m  3 [29]) and varying theoretically the ow rates (axial
and radial) from 3.2  10  3 m s  1 to 3.2  10  7 m s  1. The simulation
results are shown in Fig. 8.
The obtained results show the same trend: high Da numbers
are required to achieve high conversion. These high Da numbers
represent situations with small feed ows, and then high resi-
dence time, which afford these membrane reactors to attain even
complete conversion. If we consider the actual experimental
conditions for the dead-end conguration with an experimental
ow rate of 3.2  10  2 m s  1 it corresponds to a Da of 7.2  10  2,
and a conversion of 7%. This result indicates that in the actual
experimental conditions [29], the process is controlled by the
kinetics (very low Da).
The sigmoidal shape of the obtained graphs has some clear
implications in the sensitivity of the reactors: changes in the
values that represent extreme conditions (both low and high Da
numbers values) are much less signicant than the ones in the
intermediate range, and therefore, very small improvement of the
conversion can be attained even by important ow modications
in these extreme situations.
Nonetheless, the investigated reactors are mainly in the regime
of reaction-controlled process and only under extreme conditions
out of the proposed framework the mass transfer becomes the
controlling agent of the process [41]. The DaZ of the non-
permeable enzymatic wall reactor is lower than 1 for all the
proposed feed ow range, and, for the case of the dead-end
conguration, none of the values of the DaR is higher than 10.
3.2. A general analysis of the inuence of kinetics and mass transport
conditions in EMRs
In order to generalize the model, we will present in the next
sections simulations considering different kinetics and mass
196 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200

Fig. 9. Simulations of an enzymatic non-permeable wall reactor. Inuence of the reaction kinetics and transport terms over: (a) the axial Damkhler number and (b) the
performance expressed as conversion.

transport properties to extend the model to all types of EMRs with
enzymes grafted at the membrane surface.
The analyses with different kinetics showed that the results
obtained by modifying the reaction rate or the mass transfer
conditions led to different results in spite of an identical subse-
quent Da number. As illustrative example, Fig. 9 compares the
inuence of the reaction and mass terms over the axial Da number
and the conversion obtained in an enzymatic wall reactor. As it can
be observed, diagonal lines in Fig. 9a represent different operation
conditions with identical Da number values, but this pattern is
very dissimilar in Fig. 9b and very different reactor performance is
expected for points with the same DaZ value. For instance, Fig. 10
shows the substrate concentration prole of three different situations
in an enzymatic non-permeable wall reactor characterized by a same
DaZ value (0.096). However, as this dimensionless number is the
quotient between the reaction and mass transfer rates, all the
situations that maintain the ratio between both terms do not change
the resulting value of the DaZ number, but the operation conditions
can correspond to high reaction rate high mass transfer or low
reaction rate low mass transfer situations, which entail very
different reactor performances.
In Fig. 10 we can notice that the lower substrate concentrations
are reached in the case of low reaction rate and low mass transfer.
Nevertheless, it is clear that the mass transfer has greater impor-
tance than kinetics for the investigated situations. For the case of
equal DaZ number values in enzymatic non-permeable wall
reactor, the lower the treated ow, the more effective the reaction
is (never mind the reduction in the kinetics term). High feed ows
 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200
Fig. 10. Substrate (A) concentration (mol m  3) prole for three different enzymatic
non-permeable wall reactor operation conditions with a same DaZ value: Left: high
reaction rate (1.12  10  2) and high mass transfer (1.17  10  1); Center: moderate
reaction rate (1.12  10  3) and moderate mass transfer (1.17  10  2); Right: low
reaction rate (1.12  10  4) and low mass transfer (1.17  10  3).
are problematic even for very fast kinetics, as the uid in the
center of the inner channel under high velocity conditions has
very short residence time in the reactor.
As a consequence of the weakness of the Da numbers as
suitable parameters to continue the sensitivity analyses, it was
decided to complete the analysis of the tangential mode cong-
uration, which can be considered as the most polyvalent one,
on the basis of the inuence of three design and operation
conditions: reaction kinetics, membrane permeability and feed
variables.
197
Fig. 11. Inuence of the apparent maximum reaction rate over the performance of
the EMR on tangential conguration expressed as permeate, retentate and global
conversions for a xed apparent MichaelisMenten constant value (KM 20 mol/m3).
The respective feed and permeate ow rates are 18  10  8 and 9  10  8 m3/s.
Fig. 12. Inuence of the membrane permeability over the performance of the
tangential mode reactor expressed as permeate, retentate and global conversions for
intermediate reaction kinetics (vMAX 1.57  10  4 mol/s m2 and KM 20.2 mol/m3).
To study the inuence of the reaction kinetics, we reported
some other examples of enzymatic reactions over membrane
surfaces which can be tted by MichaelisMenten kinetics. The
works which published kinetics parameters of surface enzymatic
reactions over membrane support are few [4246], and sometimes,
the found data were not suitable to be converted to the desired
units [47]. Nevertheless, the intervals of values for both kinetics
parameters can be dened according to the found data: the
apparent MichaelisMenten constant KM ranges from 0.5 to
250 mol m  3 and the apparent maximum reaction rate vMAX from
10  1 to 10  7 mol s  1 m  2.
In Fig. 11, the evolution of the conversion in function of the
apparent vMAX is reported. For these simulations, the values of
the apparent vMAX were varied within the dened range while
apparent MichaelisMenten constant was not changed (KM
20 mol m  3). The value of 1.57 10  4 mol s  1 m  2, which corre-
sponds to the butyl acetate hydrolysis, is placed in the central
zone of the studied interval. It is clear that this variable has
a great inuence: while almost total conversion in the permeate
stream (99%) can be achieved for a value of 5  10  2 mol s  1 m  2,
lower kinetics parameter implies very poor conversion of the
same stream. Nevertheless, even for very fast reaction rates,
the conversion in the retentate stream is not highly improved
and the corresponding conversion value is not higher than
1%. Similar gures (not shown) were obtained for different
values of the MichaelisMenten constant. In these cases, only
198 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200
slight displacements of the curves to the right or the left resulted
from the changes to higher or lower values of the Michaelis
Menten constant, respectively.
The inuence of varying the permeability of the enzymatic
membrane and the resulting permeate stream over the reactor
performance was also investigated. Three different situations,
characterized by different reaction kinetics parameters, were
analyzed: a slow reaction (vMAX 1.53  10  6 mol s  1 m  2 and
KM 11.5 mol m  3), fast reaction kinetics (vMAX 3.64  10  2
mol s  1 m  2 and KM 0.5 mol m  3) and an intermediate kinetics
reaction (vMAX 1.57  10  4 mol s  1 m  2 and KM 20.2 mol m  3).
The results of this last case correspond to the butyl acetate
hydrolysis. The simulation curves for this last case are shown in
Fig. 12. The upper bound of the analyzed permeability range was
limited to avoid the ow of the permeate stream attaining
the value of the feed stream. Under these high-permeate circum-
stances, the results are very analogous to the dead-end congura-
tion (global conversion close to 7% for permeate stream). On the
other side, the lower bound was xed to a value 10 times smaller
than the baseline of the membrane permeability, and such a value
implies a situation comparable to the enzymatic non-permeable
wall reactor (global conversion for the retentate stream close to
7%). If the full interval is observed, the constant value of the global
conversion has to be mentioned. Therefore, the membrane perme-
ability is a variable that has little inuence over the global
performance of the reactor under these kinetics circumstances,
but, however, it can be very useful to design different types of
Fig. 13. Inuence of the membrane permeability over the performance of the
tangential conguration EMR expressed as permeate, retentate and global conver-
sions for a fast reaction kinetics (vMAX 3.64  10  2 mol/s m and KM 0.5 mol/m3).
Fig. 14. Inuence of the feed ow over the performance of the tangential
conguration EMR reactor expressed as permeate, retentate and global conversions
for a reaction with similar kinetics than the butyl acetate hydrolysis.
Fig. 15. Inuence of the feed concentration over the performance of the tangential
mode reactor expressed as permeate, retentate and global conversions for a
reaction with similar kinetics than the butyl acetate hydrolysis.
reactors depending if the outlet stream is preferred as retentate
stream (low membrane permeability) or permeate stream (high
membrane permeability).
The inuence of the membrane permeability over the perfor-
mance of a fast kinetics reactor can be observed in Fig. 13. In this
case, the high reaction rate allows a complete conversion of the
permeate stream within the full proposed permeability. Contrary
to intermediate reaction rate, the global performance of this
reactor is strongly inuenced by the membrane permeability:
while very permeable membranes lead to very high total conver-
sion, less permeable membranes improve the retentate conversion
at the expense of the moderate global conversion.
The results for the slow kinetics reaction (not shown) are quite
disappointing: not higher than 1% conversion can be obtained for
the permeate stream and lower values resulted for retentate
stream. Therefore, the evaluated conditions are totally inefcient
for slow enzymatic reactions.
Both feed ow and feed concentration were subjected to a
sensitivity analysis for the tangential EMR under acetate hydrolysis
kinetics situation. The rst investigated variable was the feed ow
and the corresponding results are shown in Fig. 14. It was decided to
limit the lower bound of the studied range to 9.4  10  8 m3 s  1 to
ensure a feed ow higher than the permeate ow (9.0  10  8 m3 s  1).
Under these conditions, the tangential conguration behaves in a
very similar way to a dead-end one (global conversion of around 7%,
with permeate as main outlet stream). As the feed ow increases, the
retentate stream becomes more relevant and, due to the poor
conversion obtained for the retentate, the global performance
decreases very fast to reach the retentate value for sufciently high
feed ows (around 0.1% of global conversion). The other feed
condition studied was the concentration of the solution (Fig. 15).
Despite of the slow reaction rate consequence of low substrate
concentration, the results show how the maximum conversion
corresponds to the minimum feed concentration. This improvement
of the reactor efciency is especially remarkable for the permeate
stream. On the other extreme, too concentrated feed streams (which
has to be under the conditions were the substrates are soluble)
cannot be highly converted by only EMR and more complex reaction
systems (for instance, recirculation or multiple reactors in series)
may be needed in order to avoid the substrate waste.
4. Conclusions
In this work we have built a 3D model able to successfully
represent the hydrodynamics and concentration proles in EMRs
under three different congurations: dead-end, non-permeable
enzymatic wall reactor and tangential ow mode. Two results can
 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200 199

be highlighted from the study of the different congurations. CS substrate concentration (mol m  3)
On the one hand, the non-permeable enzymatic wall conguration C0 feed concentration (mol m  3)
seems to be more advantageous than the dead-end case in terms DaR radial Damkhler number
of substrate conversion for the studied conditions. On the other DaZ axial Damkhler number
hand, the tangential conguration looks more favorable to pro- dP pore diameter (m)
mote best conversions in permeate streams when compared with DX diffusion coefcient (m2 s  1)
retentate streams. DXP corrected diffusion coefcient in pores (m2 s  1)
Kinetics and mass transport parameters were incorporated in Fperm permeate ow (m3 s  1)
Damkhler dimensionless numbers to investigate their inuence Fret retentate ow (m3 s  1)
over the performance of EMRs. Nevertheless, in spite the interest F0 feed ow (m3 s  1)
of Damkhler numbers to determine the controlling step of the JP permeate ux (m s  1)
process in some situations, similar Damkhler values result in very Km MichaelisMenten constant (mol m  3)
dissimilar performances of the EMRs. LP hydraulic permeability coefcient (m s  1 Pa  1)
A detailed analysis of the relative inuence of mass transport MW molecular weight of water (kg mol  1)
and enzymatic kinetics showed that this last process controls NXin inside ux (mol s  1 m  2)
generally the global process. Almost total conversion in the NXout outer ux (mol s  1 m  2)
permeate stream can be attained for the fastest reactions, while NXR reaction ux (mol s  1 m  2)
slow reactions suffer from irrelevant conversions. Nevertheless, P pressure (Pa)
even for very fast reaction rates, the conversion in the retentate T temperature (K)
stream is not signicant, with conversion values not higher than u velocity (m s  1)
1%. The membrane permeability is an important factor to dene VM molecular volume (m3 mol  1)
different types of reactors depending if the outlet stream is VMAX maximal reaction rate (mol s  1 m  2)
preferred as retentate stream (low membrane permeability) or VMM modeled reaction rate (mol s  1 m  2)
permeate stream (high membrane permeability). High feed ows VR reaction rate (mol s  1 m  3)
imply an irrelevant performance and conversion decreases drasti- association factor
cally. Despite of the slower reaction rate consequence of lower membrane thickness
substrate concentration, the obtained results demonstrate how P applied pressure (Pa)
the maximum conversion corresponds to the minimum feed porosity
concentration. material permeability (m2)
Therefore, the concurrence of fast reaction kinetics and high m viscosity (Pa s)
permeability appears as a very promising target because of its great tortuosity
potential to obtain very high conversion for the permeated stream,
so it is clear that the efforts should be oriented to progress in the
achievement of more efcient reaction kinetics. These enzymatic
kinetics improvements can be attained by different ways, including
the search and production of new enzymes showing higher activity, References
specicity and stability; or the application of better innovative
immobilization techniques for higher enzymatic load in the mem- [1] O. Kirk, T.V. Borchert, C.C. Fuglsang, Industrial enzyme applications, Curr. Opin.
brane supports, for example, the extension from only membrane Biotechnol. 13 (2002) 345351.
[2] B.M. Nestl, S.C. Hammer, B.A. Nebel, B Hauer, New generation of biocatalysts
surface immobilized enzyme to enzyme inside membrane pores for organic synthesis, Angew. Chem. Int. Ed. 53 (2014) 30703095.
too. Nevertheless, meanwhile the improved kinetics are achieved, [3] J.R. Cherry, A.L. Fidantsef, Directed evolution of industrial enzymes: an update,
the implementation of more advanced enzymatic reaction systems, Curr. Opin. Biotechnol. 14 (2003) 438443.
[4] C. Mateo, J.M. Palomo, G. Fernandez-Lorente, J.M. Guisan, R. Fernandez-
like recirculated streams and multiple EMRs in parallel or in series
Lafuente, Improvement of enzyme activity, stability and selectivity via
or even under cascade congurations, can be considered to get over immobilization techniques, Enzyme Microb. Technol. 40 (2007) 14511463.
the current limitations. [5] D. Brady, J. Jordaan, Advances in enzyme immobilization, Biotechnol. Lett. 31
(2009) 16391650.
[6] D.N. Tran, K.J. Balkus, Perspective of recent progress in immobilization of
enzymes, ACS Catal. 1 (2011) 956968.
Acknowledgments [7] S.D. Minteer, Enzyme stabilization and immobilization: methods and proto-
cols, Methods in Molecular Biology, Humana Press, New York, 2011.
This project has received funding from the European Union's [8] A. Sassolas, L.J. Blum, B.D. Leca-Bouvier, Immobilization strategies to develop
enzymatic biosensors, Biotechnol. Adv. 30 (2012) 489511.
Seventh Framework Programme for research, technological devel- [9] J. Sanchez Marcano, T.T Tsotsis, Catalytic Membranes and Membrane Reactors,
opment and demonstration under Grant agreement no. 282818. Wiley VCH, Weinheim, 2002.
[10] G.M. Rios, M.P. Belleville, D. Paolucci, J. Sanchez, Progress in enzymatic
membrane reactors - a review, J. Memb. Sci. 242 (2004) 189196.
[11] G. Rios, M.P. Belleville, D. Paolucci, Enzymatic membrane reactors in applica-
Appendix A. Supporting information
tions of membrane separations technology, in: K. Mohanty, M.K. Purkait (Eds.),
Membranes Technologies and Applications, CRC Press Taylor & Francis Group,
Supplementary data associated with this article can be found in Boca Raton, FL, USA, 2012, pp. 445461 (Chap. 24).
[12] E. Magnan, I. Catarino, D. Paolucci-Jeanjean, L. Preziosi-Belloy, M.P. Belleville,
the online version at http://dx.doi.org/10.1016/j.memsci.2014.09.020.
Immobilization of lipase on a ceramic membrane: activity and stability,
J. Membr. Sci. 241 (2004) 161166.
[13] M. Miyazaki, H. Maeda, Microchannel enzyme reactors and their applications
for processing, Trends Biotechnol. 24 (2006) 463470.
[14] P.M. Doran, Bioprocess Engineering Principles, Academic Press, San Diego,
Nomenclature 1995.
[15] J.M. Woodley, M. Breuer, D. Mink, A future perspective on the role of industrial
C concentration (mol m  3) biotechnology for chemicals production, Chem. Eng. Res. Des. 91 (2013)
20292036.
Cperm permeate concentration (mol m  3) [16] J.M. Brass, F.W.J.M.M. Hoeks, M. Rohner, Application of modelling techniques
Cret retentate concentration (mol m  3) for the improvement of industrial bioprocesses, J. Biotechnol. 59 (1997) 6372.
200 R. Abejn et al. / Journal of Membrane Science 473 (2015) 189200
[17] C. Jimnez-Gonzlez, J.M. Woodley, Bioprocesses: modeling needs for process
evaluation and sustainability assessment, Comput. Chem. Eng. 34 (2010)
10091017.
[18] D. Vasic-Racki, Z Findrik, A. Vrsalovic Presecki, Modelling as a tool of enzyme
reaction engineering for enzyme reactor development, Appl. Microbiol.
Biotechnol. 91 (2011) 845856.
[19] R. Abejn, A. Garea, A. Irabien, Analysis and optimization of continuous
organic solvent nanoltration by membrane cascade for pharmaceutical
separation, AIChE J. 60 (2014) 931948.
[20] S. Leuchs, J. Lima-Ramos, L. Greiner, N. Al-Haque, P Tufvesson, J.M. Woodley,
Reaction engineering of biocatalytic enantioselective reduction: a case study
for aliphatic ketones, Org. Process Res. Dev. 17 (2013) 10271035.
[21] P.R. Salmon, C.R. Robertson, A theoretical analysis of a hollow-ber reactor
with two substrate, Chem. Eng. J. 35 (1987) B1B8.
[22] T. Fuji, O. Miyawaki, T. Yano, Modelling of hollow-ber cappilary reactor for
the production of L-alanine with co-enzyme regeneration, Biotechnol. Bioeng.
38 (1991) 11661172.
[23] G. Salzman, R. Tadmor, S. Guzy, S. Sideman, N. Lotan, Hollow ber enzymatic
reactors for a two substrate process: analytical modeling and numerical
simulations, Chem. Eng. Process. 38 (1999) 289299.
[24] V. Calabro, S. Curcio, G. Iorio, A theoretical analysis of transport phenomena in
a hollow ber membrane bioreactor with immobilized biocatalyst, J. Membr.
Sci. 206 (2002) 217241.
[25] W.S. Long, S. Bhatia, A.H. Kamaruddin, Modeling and simulation of enzymatic
membrane reactor for kinetic resolution of ibuprofen ester, J. Membr. Sci. 219
(2003) 6988.
[26] S. Bhatia, W.S. Long, A.H. Kamaruddin, Enzymatic membrane reactor for the
kinetic resolution of racemic ibuprofen ester: modeling and experimental
studies, Chem. Eng. Sci. 59 (2004) 50615068.
[27] S.H. Choi, F. Scura, G. Barbieri, R. Mazzei, L. Giorno, E. Drioli, J.H. Kim,
Bio-degration of phenol in wastewater by enzyme-loaded membrane reactor:
numerical approach, Membr. J. 19 (2009) 7282.
[28] S.Y. Lau, M.H. Uzir, A.H. Kamaruddin, S. Bhatia, Lipase-catalyzed dynamic
kinetic resolution of racemic ibuprofen ester via hollow ber membrane
reactor: modeling and simulation, J. Membr. Sci. 357 (2010) 109121.
[29] S. Ben Ameur, C.L. Gijiu, M.P. Belleville, J. Sanchez, D. Paolucci-Jeanjean,
Development of a multichannel monolith large-scale enzymatic membrane
and application in an immobilized enzymatic membrane reactor, J. Membr.
Sci. 455 (2014) 330340.
[30] M.P. Belleville, P. Lozano, J.L. Iborra, G.M. Rios, Preparation of hybrid mem-
branes for enzymatic reaction, Sep. Purif. Technol. 25 (2001) 229233.
[31] P. Lozano, A.B. Prez-Marn, T. De Diego, D. Gomez, D. Paolucci-Jeanjean,
M.P. Belleville, G.M. Rios, J.L. Iborra, Active membranes coated with immobi-
lized Candida antarctica lipase B: preparation and application for continuous
butyl butyrate synthesis in organic media, J. Membr. Sci. 201 (2002) 5564.
[32] V. Chea, D. Paolucci-Jeanjean, M.P. Belleville, J. Sanchez, Optimization and
characterization of an enzymatic membrane for the degradation of phenolic
compounds, Catal. Today 193 (2012) 4956.
[33] M. de Cazes, M.P. Belleville, E. Petit, M. Llorca, S. Rodrguez-Mozaz, J. de
Gunzburg, D. Barcelo, J. Sanchez-Marcano, Design and optimization of an
enzymatic membrane reactor for tetracycline degradation, Catal. Today 236
(2014) 146152.
[34] S.I. Vasin, A.N. Filippov, Permeability of complex porous media, Colloid J. 71
(2009) 3145.
[35] W. Yao, Y.B. Li, N. Chen, Analytic solutions of the interstitial uid ow models,
J. Hydrodyn. 25 (2013) 683694.
[36] C.R. Wilke, P. Chang, Correlation of diffusion coefcients in dilute solutions,
AIChE J. 1 (2) (1955) 264270.
[37] R.C. Reid, J.M. Prausnitz, B.E. Poling, The Properties of Gases & Liquids, 4th
Edition, McGraw-Hill, New York (1987) 53 (table 3-8).
[38] M. Le Bars, M.G. Worster, Interfacial conditions between a pure uid and a
porous medium: implications for binary alloy solidication, J. Fluid Mech. 550
(2006) 149173.
[39] R. Lortie, Evaluation of the performance of immobilized enzyme reactors with
Michaelis-Menten kinetics, J. Chem. Technol. Biotechnol. 60 (1994) 189193.
[40] E. Abbaszadeh Molaei, S.M.H. Hashemi Amrei, A. Molaei Dehkordi, M. Haghi,
Transient and steady-state analysis of heat, mass, and momentum transfer in
developing and fully-developed regions of homogeneous tubular reactors
with non-Newtonian uid ow, Energy Convers. Manag. 65 (2013) 308321.
[41] I.M. Abu-Reesh, Predicting the performance of immobilized enzyme reactors
using reversible Michaelis-Menten kinetics, Bioprocess Eng. 17 (1997)
131137.
[42] J. Ceynowa, M. Raucheisz, High enantioselective resolution of racemic
2-arylpropionic acids in an enzyme membrane reactor, J. Mol. Catal. B: Enzym.
23 (2003) 4351.
[43] N. Vasileva, T. Godjevargova, V. Konsulov, A. Simeonova, S. Turmanova,
Behavior of immobilized glucose oxidase on membranes from polyacryloni-
trile and copolymer of methylmethacrylate dichlorophenylmaleimide,
J. Appl. Polym. Sci. 101 (2006) 43344340.
[44] J. Xu, Y Wang, Y. Hu, G. Luo, Y. Dai, Immobilization of lipase by ltration into a
specially designed microstructure in the CA/PTFE composite membrane,
J. Mol. Catal. B: Enzym. 42 (2006) 5563.
[45] S.F. Li, W.T. Wu, Lipase-immobilized electrospun PAN nanobrous membranes
for soybean oil hydrolysis, Biochem. Eng. J. 45 (2009) 4853.
[46] Vasileva, V. Iotov, Y. Ivanov, N. Kotia Godjevargova, Immobilization of
-galactosidase on modied polypropilene membranes, Int. J. Biol. Macromol.
51 (2012) 710719.
[47] A. Trusek-Holownia, A. Noworyta, Peptides removing in enzymatic membrane
bioreactor, Desalination 221 (2008) 543551.