You are on page 1of 11

Chemical Engineering Science 62 (2007) 2375 2385 www.elsevier.

com/locate/ces

Non-isothermal lipase-catalyzed kinetic resolution in a packed bed reactor: Modeling, simulation and miniplant studies
W.R. Berendsen , A. Lapin, M. Reuss
Institute of Biochemical Engineering (IBVT), University of Stuttgart, Allmandring 31, 70569 Stuttgart, Germany Received 27 November 2006; received in revised form 4 January 2007; accepted 14 January 2007 Available online 26 January 2007

Abstract A rigorous model is developed for the exothermic kinetic resolution of 1-methoxy-2-propanol with vinyl acetate catalyzed by immobilized Candida antarctica lipase B in a packed bed reactor. The non-isothermal two-dimensional heterogeneous model takes into account the coupled mass and energy balances, the uneven ow distribution and irreversible ping-pong bibi kinetics with competitive substrate inhibition by both enantiomers. This model is based on kinetic parameters, which were estimated in earlier work. The model simulation is validated with experimental results obtained in a fully automated modular miniplant and is shown to be capable of predicting the key parameters needed for process design of a kinetic resolution, the enantiomeric excess and the extent of conversion at a given supercial velocity. 2007 Elsevier Ltd. All rights reserved.
Keywords: Non-isothermal; Mathematical modeling; Packed bed; Enzyme; Kinetic resolution; Simulation

1. Introduction Due to recent advancements in enzyme production, purication and immobilization technology, packed bed reactors are becoming more and more the reactor of choice for biochemical reactions. Fixed bed reactors offer higher enantioselectivity in case of kinetic resolutions (Indlekofer et al., 1993), higher substrate conversion (Lin, 1991), highest catalyst to liquid ratio and easier catalyst retainment on cost of lower heat exchange rates compared to continuous stirred tank reactors (CSTRs). Hence, they provide much opportunity for industrial application of biocatalytic conversions. The transport mechanisms in such a wall-cooled xed bed reactor consist of convective and dispersive mass transport as well as convective and conductive energy transport in the bulk phase, mass and heat transfer between the solid and uid phase, heat transfer between the reaction mixture and the cooling agent,
Corresponding author. Present address: Danone GmbH, R&D, Gervais Str. 4, 97199 Ochsenfurt, Germany. Tel.: +49 711 6854573; fax: +49 711 6855164. E-mail addresses: WRBerendsen@gmx.net (W.R. Berendsen), Reuss@ibvt.uni-stuttgart.de (M. Reuss).

dispersive mass and conductive energy transport and biocatalytic conversion in the solid phase (Fig. 1). Models of enzymatic reactions in packed bed reactors have been built by considering one or more of these phenomena in various combinations and for various reactions. An overview of activities in this eld is presented in Table 1. As is evident from this table, most studies apply simple models based on one-dimensional pseudo-homogeneous or heterogeneous models. While all give insight through modeling and/or simulation, only a few studies actually provide experimental data for verication of these results. The majority of them apply a simple kinetic model, such as the irreversible or a reversible MichaelisMenten mechanism. In general, isothermal conditions are assumed, thus energy balances are not needed (e.g. Indlekofer et al., 1996; Jung and Bauer, 1992; Xiu et al., 2001), neglecting the heat of reaction and its potential consequences on reactor performance. This is justied for most enzymatic reactions, as their heat of reaction and productivity are usually low. Some examples, however, exist of biocatalytic conversions with a considerable heat of reaction and productivity, such as the decomposition of hydrogen peroxide catalyzed by catalase from beef liver ( Hm(25 C,1 atm) = 93 3 kJ mol1 ;

0009-2509/$ - see front matter doi:10.1016/j.ces.2007.01.006

2007 Elsevier Ltd. All rights reserved.

2376

W.R. Berendsen et al. / Chemical Engineering Science 62 (2007) 2375 2385

In the liquid: Material transport Mass & heat transfer liquid / particle by convection (axial) and dispersion (axial & radial) Energy transport by convection (axial) and effective conduction (mainly radial) Reactor T Liquid Particles Wall Cooling Liquid

In the particle: Diffusion Heat conduction Biocatalytic reaction

x Uneven flow distribution


Fig. 1. Transport mechanisms in a xed bed reactor packed with immobilized enzymes.
Table 1 Non-exclusive overview of scientic publications focusing on modeling of immobilized enzyme reactions in packed bed reactors
Authors 1D 2D Mass and energy balances Intraparticle Experimental verication x x M.-M.a kinetics x x x

Carrara et al. (2003) Ching and Chu (1988) Faqir and Attarakih (2002) Faqir (2004) Gonzlez et al. (1989) Hassan et al. (1995) Hassan et al. (1996) Indlekofer et al. (1996) Jung and Bauer (1992) Kim et al. (1982) Kobayashi and Moo-Young (1971) Lin (1972) Lin (1991) Lortie (1994) Lortie and Thomas (1986) Marrazzo et al. (1975) Marsh and Tsao (1976a) Marsh and Tsao (1976b) Moynihan et al. (1989) Patwardhan and Karanth (1982) Stadler (2005) Xiu et al. (2001)

x x x x x x x x x x x x x x x x x x x x x

x x x x x x x x x x x x x x x x x x x x x x

x x

x
b b

x x x x x

x x

1D, 2D = one-, two-dimensional model. a MichaelisMenten kinetics. b Through effectiveness factor correlation.

Liang et al., 1997; Tran Minh et al., 1976). Lipase-catalyzed transesterication with the acyl donor vinyl acetate is another example, where the major contribution to the heat of reaction originates from tautomerization of vinyl alcohol to acetaldehyde (42 13 kJ mol1 ; Holmes and Lossing, 1982; Toullec, 1990). The use of vinyl esters, such as vinyl acetate, in Candida antarctica lipase B catalyzed transesterications is popular for kinetic resolution of secondary alcohols, since the tautomerization causes the reaction to be irreversible (Degueil-Castaing

et al., 1987), which results in high conversion and enantioselectivity values (e.g. Kazlauskas and Bornscheuer, 1998; Resnick et al., 2003; Rotticci, 2000). In this paper, the enantioselective conversion of (R/S)1-methoxy-2-propanol with vinyl acetate catalyzed by immobilized C. antarctica lipase B is investigated as a model system for the application of exothermic kinetic resolution reactions in packed bed reactors. A non-isothermal two-dimensional heterogeneous model is developed and used for simulations, which are validated with experimental results obtained in a fully

W.R. Berendsen et al. / Chemical Engineering Science 62 (2007) 2375 2385

2377

automated modular miniplant. In addition to intraparticle mass transport, which plays a role in this system (Berendsen et al., 2006c), uneven ow distribution is taken into account, as this effect signicantly improved modeling of a wall-cooled reactor (Hein and Vortmeyer, 1995). As kinetic mechanism, an irreversible ping-pong bibi mechanism with alternative substrate inhibition by both enantiomers is applied. The coupled mathematical model is solved numerically using the control volume technique (Lapin et al., 2004, 2006; Patankar, 1982). 2. Mathematical model development The following assumptions are made in model development for the xed bed reactor system involving immobilized C. antarctica lipase B (CAL-B) catalyzed kinetic resolution of (R/S)-1-methoxy-2-propanol (MP) with vinyl acetate (VA): all physical and transport properties are independent of temperature, except rate constants; catalyst particles are spherical (Eigtved, 1992) and randomly packed inside the reactor; enzyme molecules are uniformly distributed inside evensized particles. Of course, this is a simplication of reality, as the particle radius ranges from 0.15 to 0.45 mm (Eigtved, 1992) and the enzyme is highly unevenly distributed in an external shell of 0.1 mm (Mei et al., 2003). While the distribution of the enzymes may affect substrate concentration and temperature proles in the enzyme particle (Hossain et al., 1986; Hossain and Do, 1989; Scharer et al., 1992), this assumption is necessary here, however, as determination of the exact enzyme location in each applied particle and at each spot in the reactor is beyond the scope of this paper; diffusion of substrates and products inside the particles can be described by Ficks law; adsorption is assumed to be negligible, as the model simulations presented here describe the xed bed reactor measurements accurately, using prior estimated kinetic parameters (Berendsen et al., 2006c) and parameters using empirical relationships. This would not be the case if adsorption would be signicant, since the kinetic parameters were estimated on the basis of dynamic experiments; enzyme deactivation is not taken into account. CAL-B is known to exhibit good stability at high temperatures (compared to other enzymes). For example, it did not lose any activity at all during 250300 h of operation at 65 C in organic solvents (Bousquet et al., 2000; Fishman et al., 2001) and it showed prolonged activity at 130 C (Turner and Vulfson, 2000).

can be written as l jcl,i j2 cl,i jcl,i 1 j Der,l rr + = Dax,l jt rr jrr jrr jx 2 jcl,i + kl ap (cs,i cl,i ), u0,l jx jTl = jt j2 Tl jTl 1 j + er,l rr rr jrr jrr jx 2 jTl + lp ap (Ts Tl ). u0,l l cp,l jx
ax,l

(1)

l l cp,l

(2)

The boundary conditions at the entry (x = 0), exit (x = X), center (rr = 0) and wall (rr = Rr ) of the reactor are given by x = 0: u0,l cl,i Dax,l u0,l x = X: rr = 0: rr = Rr : jcl,i = u0,l cl,inlet,i , jx jTl = u0,l l cp,l Tl,inlet , l cp,l Tl ax,l jx jcl,i = 0, jx jcl,i = 0, jrr jcl,i = 0, jrr jTl = 0, jx jTl = 0, jrr
er,l

(3) (4) (5) (6)

jTl = kh (Tc Tl ). jrr

The initial conditions pertaining to Eqs. (1) and (2) are as follows: t = 0: cl,inlet,i = cl,inlet,i,0 , Tl,inlet = Tl,inlet,0 , (7)

where ap is the specic external particle surface area per unit reactor volume (m1 ), cl,i , cs,i the concentration in liquid and solid phase, respectively (mol L1 ), cp,l the specic heat capacity of liquid (J kg1 K 1 ), Dax,l , Der,l the effective axial/radial dispersion coefcient in liquid phase (m2 s1 ), kL the mass transfer coefcient (m s1 ), kh the lumped heat transfer coefcient (W m2 K 1 ), rr the distance from reactor center (m), Rr inner reactor radius (m), t the time (s), Tl , Tc the temperature of liquid and cooling medium, respectively (K), u0,l the supercial velocity (m s1 ), x the distance from reactor inlet (m), X the reactor height (m), lp the heat transfer coefcient between liquid and particle (W m2 K 1 ), l the reactor porosity (), ax,l , er,l the effective axial and radial heat conduction in liquid phase (W m1 K 1 ), l the liquid density (kg m3 ). The internal mass and energy balance for a representative particle must be simultaneously simulated at every position in the reactor. The mass (i = R-MP, S-MP, VA) and energy balances for the immobilized enzyme particle are represented by the following equations: 1 j jcs,i = Dep,l 2 jt rp jrp
p cp,p 2 rp

2.1. Mass and energy balances The mass balance equations for the substrates (i = R-MP, S-MP, VA) and energy balance equations in the bulk liquid phase

jcs,i jrp

vi ,

(8)

jTs = jt

1 j 2 jTs rp 2 jr jrp rp p Hm (vR -MP + vS -MP ),


ep,l

(9)

2378

W.R. Berendsen et al. / Chemical Engineering Science 62 (2007) 2375 2385

which are subject to the following initial and boundary conditions at the center (rp =0) and surface (rp =Rp ) of the particle: jcs,i = 0, rp = 0: jrp rp = Rp : Dep,l
ep,l

where vmax,a = kcat,a and


T vmax,a =vmax,a e Hact,a /R(1/T 1/301.15)

jTs = 0, jrp

(10)

1 1 l

with a = R, S

(20)

jTs = kh (Tl Ts ), jrp Ts = Ts,0 ,


1

jcs,i = kl (cl,i cs,i ), jrp (11) (12)

with a=R, S,

(21)

with kcat,a is the maximum specic conversion rate (mol g1 s1 ), Km,i the afnity constant (mol L1 ), p the particle density (kg m3 ), Hact,a the activation energy (J mol1 ). 2.3. Calculation of radial porosity and supercial velocity prole

t = 0: cs,i = cs,i,0 , where kh = 1


wl

w w

1
wc

(13)

with Dep,l is the effective diffusion coefcient in particle (m2 s1 ), rp the distance from particle center (m), Rp the particle radius (m), vi the volumetric reaction rate (mol L1 s1 ), wc the heat transfer coefcient between wall and cooling medium (W m2 K 1 ), wl the heat transfer coefcient between wall and liquid (W m2 K 1 ), w the thickness of glass wall between cooling medium and mixture (m), Hm the enthalpy of reaction (J mol1 ), w the thermal conductivity of the glass reactor wall (W m1 K 1 ), ep,l the effective heat conduction in particle (W m1 K 1 ). For the cooling liquid, the energy balance may be written as jTc = u0,c c cp,c jt kh jTc (Tc Tl ), c cp,c jx c (14)

The porosity in a xed bed reactor may be considered constant, assuming it is an unlimited wide bed ( ), however, in doing so, the increase in porosity close to the reactor wall is neglected. This porosity prole results in a radially uneven ow distribution. By taking this effect into account, a signicantly improved model prediction of a wall-cooled reactor was obtained by Hein and Vortmeyer (1995). The radial porosity prole is calculated with the exponential function (Tsotsas, 2002): l = 1 + a1 exp b1 Rr r r Rp , (22)

a1 = 1.36, b1 = 2.5, = 0.4.

with the following initial and boundary conditions: x = 0: Tc = Tc,inlet , t = 0: Tc = Tc,inlet,0 , (15) (16)

The dependency of the supercial velocity in radial direction can be obtained by applying iterative methods for numerically solving the enhanced Brinkmann equation (Tsotsas, 2002): jp j = f1 u0 f2 u2 + eff 0 jx rr jrr r ju0 jr , (23)

where cp,c is the specic heat capacity of cooling medium (J kg1 K 1 ), c the cooling jacket width (m), c the density of cooling medium (kg m3 ). 2.2. Kinetics The biocatalytic reaction, the enantioselective conversion of 1-methoxy-2-propanol (MP) with vinylacetate (VA) into R- and S-1-methoxy-2-propyl-acetate (R- and S-MPA) and acetaldehyde (Ac), may be modeled using an irreversible ping-pong bibi mechanism taking into account competitive substrate inhibition by both enantiomers (Berendsen et al., 2006c). The corresponding equations for the volumetric rate of both enantiomers of MP and VA, vi , are vR -MP = vmax,R cs,R -MP cs,V A Km,R -MP cs,V A 1 +
cs,S -MP Km,S -MP

2 with f1 = a2 ((1 l )2 /l3 ) l /Rp , f2 = b2 ((1 l )/l3 ) l /Rp , = c2 l exp(d2 Rep ), a2 = 37.5 (), b2 = 0.875 (Pa m), eff c2 = 2 (), d2 = 2.0 103 (), with the boundary conditions: ju0 = 0, (24) rr = 0: jrr

rr = Rr : u0 = 0,

(25)

where p is the pressure (Pa), Rep the particle Reynolds (), l the viscosity of liquid phase (Pa s). In order to determine the pressure gradient (jp/jx), the following condition is used: the cross-section integrated uid velocity equals the given ow rate.

1+

Km,V A cs,V A

+ Km,V A cs,R -MP + cs,R -MP cs,V A ,

(17)

vS -MP =

vmax,S cs,S -MP cs,V A Km,S -MP cs,V A 1 +


cs,R -MP Km,R -MP

1+

Km,V A cs,V A

+ Km,V A cs,S -MP + cs,S -MP cs,V A (19)

(18)

vV A = vR -MP + vS -MP

W.R. Berendsen et al. / Chemical Engineering Science 62 (2007) 2375 2385

2379

3. Materials and methods 3.1. Chemicals and immobilized enzyme Vinyl acetate and 1-methoxy-2-propanol (99%) were kindly provided by The Dow Chemical Company, Stade, Germany.

Acetonitrile and pentanone (99%) used for GC-analysis originated from Sigma-Aldrich, Hannover, Germany. C. antarctica lipase B immobilized on an acrylic resin, Chirazyme L-2, c.-f., C2 with a wet particle density of 200 g L1 and a particle diameter of 0.30.9 mm originated from Roche Diagnostics, Mannheim, Germany (Eigtved, 1992).

Fig. 2. Reaction part of fully automated miniplant (1: PC for process control, 2: balances, 3: MP substrate vessel, 4: VA substrate vessel, 5: substrate pumps, 6: mixing vessels, 7: pump, 8: xed bed reactor, 9: intermediate vessel between reaction and distillation part of miniplant, 10: cryostats for temperature control).

S4

WT2 0-50 C

Z5
TIRC 14

PI 406

30 cm
TIRC 104X

20 cm 20 cm 80 cm

Z4 Z3

50 cm

20 cm 20 cm

XWT2

Z2

TIRC 103

Fixed bed reactor R1

Z1 S3 1.4 cm 3.4 cm
Fig. 3. Detailed process information diagram of the miniplant reactor used for verication of the reactor model (PI: pressure sensor, S: stream, TIRC: temperature sensor, WT: cryostat, XWT: automatic valve, Z: sampling valve).

2380

W.R. Berendsen et al. / Chemical Engineering Science 62 (2007) 2375 2385

3.2. Miniplant design, construction, start-up and operation The design and construction of the miniplant was successfully completed through the combination of in-house expertise, outsourcing and technology transfer with Dow Chemical, Midland, MI, USA. In Figs. 2 and 3 a photo of the plant and a detailed process information diagram (PID) of the xed bed reactor are displayed. The miniplant was built within a BoschRexroth frame (UTZ Ratio Technik, Korb). Process control equipment and software were supplied by National Instruments (Labview , Austin, TX, USA). 3.3. Typical miniplant operation and measurements Substrates were added separately to 25 L stainless steel vessels (V1, V2) after which they were automatically pumped (gamma/L ProMinent Dosiertechnik GmbH, Heidelberg) in the correct ratio and amount to 20 L glass vessels (V3, V4, HWS Labortechnik, Mainz) and mixed by air-driven stirrers (Gebr. Buddeberg GmbH, Mannheim). The temperature of this mix was adjusted prior to being pumped by an Ismatec MCP-CPF process pump with QP-Q1 pump head (Wertheim, Germany) into the bottom of the jacketed xed bed reactor (Glasgertebau Ochs GmbH, Bovenden Lenglern). The temperature of the reactor was controlled using a PT100 sensor (RS Components GmbH, Mrfelden-Walldorf) at x = 0.63 as setpoint for a cryostat (F33, Julabo, Seelbach), which employed a 5050 mixture of water and ethanol as cooling medium. Flow rates were calibrated and measured accurately using Sartorius Combics 3 balances (HBH Ebinger, Fichtenberg). Samples were taken at 25%, 50%, 75% and 100% relative reactor height at various conditions. These samples were further processed, by pipetting in triplicate 50 L of each sample into

950 L acetonitrile, containing 10 mM Pentanone as internal standard. These samples were further analyzed by gas chromatography to determine the local concentrations of R-MPA and S-MPA in the reactor. In addition to these concentration measurements, the temperature of the reaction medium was measured continuously at 19%, 63% and 95% relative reactor height and the mass ow was determined by analyzing the loss of weight over time of the tanks containing the substrate mixture (V3, V4). The temperature in the reactor was controlled using the sensor at 63% relative reactor height. 3.4. Gas chromatography analysis method A Varian, CP3800 GC, equipped with a Chirasil -Dex chiral capillary column (25 m 0.25 mm ID0.25 m, CP7502, Varian, Darmstadt, Germany) was used with the following temperature programming: 1 min at 40 C, 5 C min1 to 75 C, 10 C min1 to 100 C, 80 C min1 to 200 C and a 2 min hold at 200 C. Helium was used as carrier gas (1.3 mL min1 ) and the compounds were detected using a ame ionization detector (FID) at 250 C. Injector was set to 200 C and the split ratio was 1:20. Response times for S-MPA and R-MPA were 9.6 and 10.2 min, respectively. 4. Results and discussion While investigating the kinetic resolution of 1-methoxy2-propanol with vinyl acetate catalyzed by immobilized C. antarctica lipase B in a xed bed reactor, a signicant temperature gradient is observed in axial direction (Fig. 4). While a temperature gradient is common in chemical engineering, it is rare in biocatalysis, as the heat of the reaction is usually low and/or the applied biocatalysts exhibit low productivity. Since

50 40 30 T(C) 20 Tx=0.19 10 Tx=0.97 0 -10 0 1 2 t/ (-) 3 4 Tc,inlet Tx=0.63

Tx=0.97 Tx=0.19

Tx=0.63

Fig. 4. Dynamic temperature prole in a xed bed reactor during kinetic resolution of 1-methoxy-2-propanol by immobilized Candida antarctica lipase B (u0,l = 322 cm h1 , Tsetpoint,x=0.63 = 10 C; Tl,inlet = 10 C; RR = 0.015 m; X = 0.3 m; cl,inlet,MP = 4.4 M, cl,inlet,V A = 6.2 M).

W.R. Berendsen et al. / Chemical Engineering Science 62 (2007) 2375 2385

2381

0.5 , 0.4 cR-MPA / cMP,0 (-) , , 0.3 ,

u0,l(cm h-1) 305 682 1022 3190

0.020 0.018 0.016 cS-MPA / cMP,0 (-) MP (%) 0.014 0.012 0.010

0.2

0.008 0.006

0.1

0.004 0.002

0.0 0.0 0.2 0.4 x/X (-) 0.6 0.8 1.0

0.000

Fig. 5. Measured and simulated axial concentration prole of R- and S-enantiomer in the packed bed reactor at several average supercial velocities, u0,l (lines: model, dots: experiments; solid lines and closed symbols: R-MPA; dotted lines and open symbols: S-MPA; RR = 0.007 m; X = 0.8 m; cl,inlet,MP = 4.4 M, cl,inlet,V A = 6.2 M, Tsetpoint,x=0.63 = 30 C; Tc,inlet = 24 C; Tl,inlet = 23 C).

100 90 80 70 e.e.h (%) 60 50 40 30 20 10 0 0 500 1000 1500 2000 u0,l (cm h-1) 2500 3000 e.e.MP e.e.MPA x/X (-) , , , , , 1 0.75 0.5 0.25

70 60 50 40 30 20 10 0

Fig. 6. Verication of non-isothermal heterogeneous model by experiments in a miniplant reactor: enantiomeric excess (e.e.h ) versus average supercial velocity (u0,l ) at different dimensionless axial positions. The prole of the total enantiomer conversion ( MP ) versus supercial velocity is shown exemplary for x = 1 (lines: model, dots: experiments, RR = 0.007 m; X = 0.8 m, cl,inlet,MP = 4.4 M, cl,inlet,V A = 6.2 M, Tsetpoint,x=0.63 = 30 C; Tc,inlet = 24 C; Tl,inlet = 23 C).

transesterication with vinyl esters is an attractive means for production of chiral intermediates for pharmaceutical, agricultural and ne-chemical industry, the kinetic resolution of MP in a xed bed reactor is investigated here and the performance of the developed model discussed. Although the miniplant set-up used for obtaining the results displayed in Fig. 4 clearly shows the axial temperature gradient in the reactor, it is not practical for additional experimental investigations, as the time needed to reach steady state is very long. Due to the limited cooling capacity available in the miniplant, a reactor with smaller diameter is used from now on (1.4 cm instead of 3 cm).

In this investigation, the radial porosity and supercial velocity proles are determined for each mass ow, Eqs. (22)(25), before the coupled mass and energy balances, Eqs. (1)(16), and kinetic equations, Eqs. (17)(21), are solved simultaneously using implicit numerical methods with iterations, as described by Patankar (1982). The development of kinetic equations and estimation of its parameters is described in another publication (Berendsen et al., 2006c). The mathematical model is then executed at various conditions (e.g. at various average supercial velocities, u0,l in Fig. 5) to yield modeling simulations. These simulations are veried with experimental results obtained with the miniplant and shown in Figs. 58.

2382

W.R. Berendsen et al. / Chemical Engineering Science 62 (2007) 2375 2385

100 90 80 70 e.e.h (%) 60 50 40 30 20 10 0 0 500 1000 1500 u0,l (cm h-1) 2000 2500 3000 e.e.MP , , , , e.e.MPA x/X (-) 1 0.75 0.5 0.25

Fig. 7. Verication of non-isothermal heterogeneous model by experiments in a miniplant reactor: enantiomeric excess (e.e.h ) versus average supercial velocity (u0,l ) at different dimensionless axial positions (lines: model, dots: experiments, RR = 0.007 m; X = 0.8 m, cl,inlet,MP = 4.4 M, cl,inlet,V A = 6.2 M, Tsetpoint,x=0.63 = 10 C; Tc,inlet = 4 C; Tl,inlet = 6 C).

100 90 80 70 60 T (C) 50 40 30 20 10 0 0.0 0.2 0.4 x/X (-)


Fig. 8. Measured and simulated axial temperature prole in the packed bed reactor at several average supercial velocities and set point temperatures (lines: model, dots: experiments; RR = 0.007 m; X = 0.8 m; cl,inlet,MP = 4.4 M, cl,inlet,V A = 6.2 M; closed symbols and solid lines: Tsetpoint,x=0.63 = 30 C, Tc,inlet = 24 C, Tl,inlet = 23 C; open symbols and dotted lines: Tsetpoint,x=0.63 = 10 C, Tc,inlet = 4 C, Tl,inlet = 6 C).

u0,l (cm h-1) 138 138 214 214 658 658 682 682 TrR=RR

TrR=RR

TrR=0

TrR=0

0.6

0.8

1.0

The rst gure, Fig. 5, shows the steady state axial product concentration prole of both enantiomers in the reactor. As is typical for a kinetic resolution reaction, the conversion of one of the enantiomers (here the R-enantiomer) progresses faster than the other one. In biocatalysis, this characteristic property is usually analyzed using the enantiomeric excess, e.e.h , and total enantiomer conversion, MP , dened as (Chen et al., 1987): e.e.h = = cRh cSh , cRh + cSh h = MP, MPA, (26) (27)

MP

cR -MP ,0 cR -MP + cS -MP ,0 cS -MP , cR -MP ,0 + cS -MP ,0

whereas an enantiopure compound has an e.e. of 100%.

The dependency of the enantiomeric excess on supercial velocity and axial position in the reactor is illustrated in Figs. 6 and 7 for two set point temperatures at dimensionless axial position x/X = 0.5 and steady state conditions (10 C, 30 C). These setpoint temperatures correspond to the temperature, which was used as input for temperature control of the cryostat. As is evident from both graphs, enantiomeric values close to 9598% can be achieved depending on the conditions applied. Finding the right supercial velocity is a trade-off between having a high endproduct concentration of the respective enantiomer and obtaining a high enantiopurity. At high supercial velocity, the conversion of both enantiomers, MP , is lower (shown exemplary for x/X = 1 in Fig. 6) than at low supercial velocity, resulting in good e.e.-values for the

W.R. Berendsen et al. / Chemical Engineering Science 62 (2007) 2375 2385


Table 2 Applied parameters in model simulations shown in Figs. 58
Parameters cp,c cp,l cp,p Dax,l Der,l Dep,l kl Rep Rp Rr u0,c X
lp wc wl c w

2383

Table 3 Previously estimated kinetic parameters (Berendsen et al., 2006c)


Unit J kg J kg1 K 1 J kg1 K 1 m2 s1 m2 s1 m2 s1 m s1 m m m s1 m W m2 K 1 W m2 K 1 W m2 K 1 m m J mol1 Pa s W m1 K 1 W m1 K 1 W m1 K 1 W m1 K 1 kg m3 kg m3 kg m3
1

Value 2440 2130 1460 1.1 105 2.5 108 1.66 1010 1 104 0.56 3.0 104 7.0 103 9.9 102 0.8 2300 300 970 8 103 1 103 68 700 1 103 11.5 0.186 0.2 1.05 789 900 200

Parameters K 1 kcat,R kcat,S Km,R -MP Km,S -MP Km,V A Hact,R Hact,S

Value 1.4 10 9.4 109 0.28 0.72 1.40 37 000 45 000


6

Unit mol g1 s1 mol g1 s1 mol L1 mol L1 mol L1 J mol1 J mol1

Hm
l ax,l ep,l er,l w c l p

reactor center (rr = 0) and wall (rr = Rr ). Unfortunately, it is unclear which spot in radial direction the temperature sensors in the miniplant reactor actually measure. Therefore an accurate comparison between predicted and measured temperatures is not possible. However, the measured temperatures are well within the simulated temperatures at rr = 0 and Rr . Since the mass and energy balances are coupled, the prediction of the temperature appears to be correct also, as otherwise the concentration proles would not be correctly described either. 5. Conclusions The presented comprehensive mathematical model is capable of predicting accurately the critical parameters needed for xed bed reactor operation of an exothermic kinetic resolution reaction without additional estimation of parameters. The C. antarctica lipase B catalyzed transesterication of 1-methoxy-2-propanol with vinyl acetate in a xed bed reactor is shown to be inuenced by axial and radial temperature gradients. If the temperature in the reactor is appropriately maintained, however, the reaction yields high enantioselectivity values at nearly complete conversion of the faster enantiomer. This demonstrates the technological feasibility of this reaction and indicates the importance of iterating reactor design improvements though the combination of modeling, simulation and miniplant experiments to optimize selectivity and productivity. Appropriate control of reactor temperature is key, especially in relation to scale-up of xed bed reactors, which is usually more pronounced in the radial than in the axial direction. In chemical industry, the application of miniplant technology, modeling and simulation plays a major role in fast development of new products and processes (Behr et al., 2004; Heimann, 2003). This technology is especially suited for mimicking technical operations at the smallest scale. While stateof-the-art in chemical industry, its application in biocatalysis is rare. Only one case has been reported previously (Stadler, 2005). This technology offers also much potential for the eld of biocatalysis. The veried model and the learnings presented in this paper are an example for this. By application of such miniplant validated models early in development, the gained knowledge and experience may be utilized for sustainable decision making or rationally controlled improvement of the complete process and biocatalyst (Berendsen and Samorski, 2006).

The empirical calculation methods summarized by Koning (2002) were used for determination of heat and mass transport parameters.

product (e.e.MP A ). By lowering the supercial velocity, the conversion increases, just like the e.e. of the substrate (e.e.MP ). At one point, however, when the R-enantiomer is nearly fully converted, its reaction rate drops in comparison to the rate of the S-enantiomer causing a decrease in product e.e. As soon as the faster enantiomer is converted, the process may yield the S-enantiomer of the reaction substrate in high enantiopurity. Hence, it is crucial to adjust the supercial velocity accurately in xed bed reactor operated kinetic resolutions. While this behavior of enantioselectivity and conversion is typical for kinetic resolutions and is usually shown in plots of e.e. versus conversion (e.g., Berendsen et al., 2006a,b), such plots do not give information about the relationship of these parameters with the reaction velocity, i.e., the way the reactor needs to be operated, and thus are not suited for complete validation of the reactor model. Figs. 6 and 7 offer this possibility, while giving insight into process behavior. Overall, the model applied in Figs. 58 describes the experimental ndings relatively well considering the fact that all parameters needed for model simulations are calculated using correlations available in literature (Table 2) and, in case of kinetic parameters, are previously estimated using a set of dynamic experiments in a stirred tank reactor (Berendsen et al., 2006c, Table 3). In particular, the dependency of enantiomeric excess on supercial velocity, which is crucial for kinetic resolution reactions, is nicely predicted by the model. The simulated axial temperature prole is presented in Fig. 8 for the

2384

W.R. Berendsen et al. / Chemical Engineering Science 62 (2007) 2375 2385


er,l

Notation ap cp,c cl,i cp,l cp,p cl,i cs,i Dax,l Dep,l Der,l e.e.h Hact,a Hm kcat,a kh kL Km,i p rp rr Rep Rp Rr t Tc Tl u0,l u0 vi x X specic external surface area of the particle per unit reactor volume =3 (1 l )/Rp , m1 specic heat capacity of cooling medium, J kg1 K 1 concentration in liquid phase, mol L1 specic heat capacity of liquid, J kg1 K 1 specic heat capacity of particle, J kg1 K 1 concentration in liquid phase, mol L1 concentration in solid phase, mol L1 effective axial dispersion coefcient in liquid, m2 s1 effective diffusion coefcient in particle m2 s1 effective radial diffusion coefcient in liquid phase, m2 s1 enantiomeric excess, dimensionless activation energy, J mol1 enthalpy of reaction, J mol1 maximum specic conversion rate, mol g1 s1 lumped heat transfer coefcient, W m2 K 1 mass transfer coefcient, m s1 afnity constant, mol L1 pressure, Pa distance from particle center, m distance from reactor center, m particle Reynolds =u0 Rp l /2 l , dimensionless particle radius, m inner reactor radius, m time, s temperature of cooling medium, K temperature of liquid, K average supercial velocity, m s1 supercial velocity, m s1 volumetric reaction rate, mol L1 s1 distance from reactor inlet, m reactor height, m

MP c l p

effective radial heat conduction in liquid phase, W m1 K 1 total enantiomer conversion, dimensionless density of cooling medium, kg m3 liquid density, kg m3 particle density, kg m3

Abbreviations and Subscripts 0 Ac a i inlet l MP MPA p, s VA initial conditions acetaldehyde R, S (R- or S-enantiomer) R-MP, S-MP, VA reactor feed liquid phase 1-methoxy-2-propanol 1-methoxy-2-propyl-acetate particle or solid phase vinyl acetate

Acknowledgments This work has been supported by the Federal Ministry of Education and Research (BMBF) as part of the program sustainable bioproduction and The Dow Chemical Company, Stade, Germany. The authors would like to specially acknowledge Andreas Freund, Wilhelm Leffers and Sol Resnick for technological support and discussions. References
Behr, A., Brehme, V.A., Ewers, C.L.J., Groen, H., Kimmel, T., Kueppers, S., Symietz, I., 2004. New developments in chemical engineering for the production of drug substances. Engineering in Life Sciences 4, 1524. Berendsen, W.R., Samorski, M., 2006. Einsatz und Potenzial der integrierten Miniplant-Technologie fr die Enzymkatalyse. Chemie Ingenieur Technik 78, 10131021. Berendsen, W.R., Gendrot, G., Freund, A., Reuss, M., 2006a. A kinetic study of lipase-catalyzed reversible kinetic resolution involving verication at miniplant-scale. Biotechnology and Bioengineering 95 (5), 883892. Berendsen, W.R., Gendrot, G., Resnick, S., Reuss, M., 2006b. Kinetic modeling of lipase catalyzed hydrolysis of (R/S)-1-methoxy-2-propylacetate as a model reaction for production of chiral secondary alcohols. Journal of Biotechnology 121, 213226. Berendsen, W.R., Lapin, A., Reuss, M., 2006c. Investigations of reaction kinetics for immobilized enzymesIdentication of parameters in the presence of diffusion limitation. Biotechnology Progress 22, 13051312. Bousquet, M.P., Willemot, R.M., Monsan, P., Boures, E., 2000. Enzymatic synthesis of alpha-butylglucoside linoleate in a packed bed reactor for future pilot scale-up. Biotechnology Progress 16, 589594. Carrara, C.R., Mamarella, E.J., Rubiolo, A.C., 2003. Prediction of the xedbed reactor behaviour using dispersion and plug-ow models with different kinetics for immobilized enzyme. Chemical Engineering Journal 92, 123129. Chen, C.-S., Wu, S.-H., Girdaukas, G., Sih, G.J., 1987. Quantitative analysis of biochemical kinetic resolution of enantiomers. 2. Enzyme-catalyzed esterications in water-organic biphasic systems. Journal of American Chemical Society 109, 28122817. Ching, C.B., Chu, K.H., 1988. Modelling of a xed bed and a uidized bed immobilized enzyme reactor. Applied Microbiology and Biotechnology 29, 316322. Degueil-Castaing, M., De Jeso, B., Drouillard, S., Maillard, B., 1987. Enzymatic reactions in organic synthesis: 2-ester interchange of vinyl esters. Tetrahedron Letters 28, 953954.

Greek letters
lp wc wl c w

l
l w ax,l ep,l

particle-to-uid heat transfer coefcient, W m2 K 1 heat transfer coefcient between wall and cooling medium, W m2 K 1 heat transfer coefcient between wall and liquid, W m2 K 1 cooling jacket width, m thickness of glass wall between cooling medium and mixture, m reactor porosity, dimensionless viscosity of liquid phase, Pa s thermal conductivity of the glass reactor wall, W m1 K 1 effective axial heat conduction in liquid phase, W m1 K 1 effective heat conduction in particle, W m1 K 1

W.R. Berendsen et al. / Chemical Engineering Science 62 (2007) 2375 2385 Eigtved, P., 1992. Immobilization of lipase by adsorption on a particulate macroporous resin. US patent 5.156.963. Faqir, N.M., 2004. Optimization of operating temperature for an continuous immobilized glucose isomerase reactor with pseudo linear kinetics. Engineering in Life Science 4, 450459. Faqir, N.M., Attarakih, M.M., 2002. Optimal temperature policy for immobilized enzyme packed bed reactor performing reversible MichaelisMenten kinetics using the disjoint policy. Biotechnology and Bioengineering 77, 163173. Fishman, A., Eroshov, M., Dee-Noor, S.S., van Mil, J., Cogan, U., Effenberger, R., 2001. A two-step enzymatic resolution process for largescale production of (S)- and (R)-ethyl-3-hydroxybutyrate. Biotechnology and Bioengineering 74, 256263. Gonzlez, G., Caminal, G., de Mas, C., Lopez-Santin, J., 1989. An approach in mathematical modeling of an upow packed-bed reactor for enzymatic hydrolysis of wheat straw. Biotechnology and Bioengineering 34, 242251. Hassan, M.M., Atiqullah, M., Beg, S.A., Chowdhury, M.H.M., 1995. Analysis of non-isothermal tubular reactor packed with immobilized enzyme systems. Chemical Engineering Journal 58, 275283. Hassan, M.M., Atiqullah, M., Beg, S.A., Chowdhury, M.H.M., 1996. Effects of enzyme microcapsule shape on the performance of a nonisothermal packed-bed tubular packed bed reactor. Journal of Chemical Technology and Biotechnology 66, 4155. Heimann, F., 2003. Process intensication through the combined use of process simulation and miniplant technology. Computer Aided Chemical Engineering 14, 155160. Hein, S., Vortmeyer, D., 1995. Wandgekhlte chemische Festbettreaktoren und deren Modellierung mit Ein- und Zweiphasenmodellen. Zeitschrift fr Naturforschung 50a, 568576. Holmes, J.L., Lossing, F.P., 1982. Heats of formation of the ionic and neutral enols of acetaldehyde and acetone. Journal of American Chemical Society 104, 26482652. Hossain, Md.M., Do, D.D., 1989. General theory of determining intraparticle active immobilized enzyme distribution and rate parameters. Biotechnology and Bioengineering 33, 963975. Hossain, Md.M., Do, D.D., Bailey, J.E., 1986. Immobilization of enzymes in porous solids under restricted diffusion conditions. A.I.Ch.E. Journal 32, 10881098. Indlekofer, M., Brotz, F., Bauer, A., Reuss, M., 1996. Stereoselective bioconversions in continuously operated xed bed reactors: modeling and process optimization. Biotechnology and Bioengineering 52 (4), 459471. Indlekofer, M., Reuss, M., Barth, S., Effenberger, F., 1993. Kinetic analysis and simulation studies for lipase-catalysed resolution of racemic 2-methyl1-pentanol. Biocatalysis 7, 249266. Jung, H.J., Bauer, W., 1992. Determination of process parameters and modelling of lipase-catalyzed transesterication in a xed bed reactor. Chemical Engineering Science 15, 341348. Kazlauskas, R.J., Bornscheuer, U.T., 1998. Biotransformations with Lipases. In: Rehm, H.J., Reed, G., Phler, A., Stadler, P.J.W., Kelly, D.R. (Eds.), Biotechnology, Biotransformations I, second ed., vol. 8a, VCH, Weinheim, pp. 37191. Kim, C., Kim, H.S., Ryu, D.D.Y., 1982. Optimization of operating temperature for continuous glucose isomerase reactor system. Biotechnology and Bioengineering 24, 18891896. Kobayashi, T., Moo-Young, M., 1971. Backmixing and mass-transfer in the design of immobilized enzyme reactors. Biotechnology and Bioengineering 13, 893910. Koning, G.W., 2002. Heat and mass transport in tubular packed bed reactors at reacting and non-reacting conditions. Ph.D. Thesis, University of Twente, Twente University Press, ISBN 903651813x. Lapin, A., Mller, D., Reuss, M., 2004. Dynamic behavior of microbial populations in stirred bioreactors simulated with EulerLagrange methodstraveling along the lifelines of single cells. Industrial and Engineering Chemistry Research 43, 46474656. Lapin, A., Schmid, J., Reuss, M., 2006. Modeling the dynamics of E. coli populations in the three-dimensional turbulent eld of a stirredtank bioreactorA structured-segregated approach. Chemical Engineering Science 61, 47834797.

2385

Liang, Y., Wu, Y., Dinghuo, L., Wang, C., Liu, Y., Songsheng, Q., Zou, G., 1997. Thermokinetic models of enzyme-catalyzed reactions in batch and plug-ow reactors. Thermochimica Acta 307, 149153. Lin, S.H., 1972. Nonisothermal immobilized enzyme reaction in a packed bed reactor. Biophysik 8, 302309. Lin, S.H., 1991. Optimal feed temperature for an immobilized enzyme packedbed reactor. Journal of Chemical Technology and Biotechnology 50, 1726. Lortie, R., 1994. Evaluation of the performance of immobilized enzyme reactors with MichaelisMenten kinetics. Journal of Chemical Technology and Biotechnology 60, 189193. Lortie, R., Thomas, D., 1986. Heterogeneous one-dimensional model for xed bed enzyme reactors. Biotechnology and Bioengineering 28, 12561260. Marrazzo, W.N., Merson, R.L., McCoy, B.J., 1975. Enzyme immobilized in a packed-bed reactor: kinetic parameters and mass transfer effects. Biotechnology and Bioengineering 17, 15151528. Marsh, D.R., Tsao, G.T., 1976a. A heat transfer study on packed bed reactors of immobilized glucoamylase. Biotechnology and Bioengineering 18, 349362. Marsh, D.R., Tsao, G.T., 1976b. A theoretical and computational study of the heat effects on packed-bed reactors of immobilized enzymes. Journal of Solid-Phase Biochemistry 1, 6779. Mei, Y., Miller, L., Gao, W., Gross, R.A., 2003. Imaging the distribution and secondary structure of immobilized enzymes using infrared microspectroscopy. Biomacromolecules 4, 7074. Moynihan, H.J., Lee, C.K., Clark, W., Wang, N.H.L., 1989. Urea hydrolysis by immobilized urease in a xed bed reactor: analysis and kinetic parameter estimation. Biotechnology and Bioengineering 34, 951963. Patankar, S.V., 1982. Numerical methods in heat transfer. Proceedings of the Seventh International Heat Transfer Conference, Mnchen, vol. 1. pp. 8390. Patwardhan, V.S., Karanth, N.G., 1982. Film diffusional inuences on the kinetic parameters in packed-bed immobilized enzyme reactors. Biotechnology and Bioengineering 24, 763780. Resnick, S.M., Donate, F.A., Frank, T.C., Foley, P., 2003. Enzymatic resolution of propylene glycol alkyl (or aryl) ethers and ether acetates. World patent, WO 03/083126 A2. Rotticci, D., 2000. Understanding and engineering the enantioselectivity of Candida antarctica lipase B towards sec-Alcohols. Ph.D. Thesis, Stockholm, ISSN 110-7974. Scharer, R., Hossain, Md.M., Do, D.D., 1992. Determination of total and active immobilized enzyme distribution in porous solid supports. Biotechnology and Bioengineering 39, 679687. Stadler, H.-G., 2005. Entwicklung eines kontinuierlichen Verfahrens zur enzymkatalysierten Synthese eines strukturierten Triglycerides. Ph.D. Thesis, University of Stuttgart. Toullec, J., 1990. Keto-enol equilibrium constants. In: Rappaport, Z. (Ed.), The Chemistry of Enols. Wiley, New York, pp. 323398. Tran Minh, C., Bougnol-Bilvou, V., Lancelot, F., 1976. Enzymatic reactor using high activity immobilized enzyme during exothermic reaction. In: Proceedings of the fourth International Symposium in Chemical Reaction Engineering, DECHEMA, Frankfurt/Main.pp. 125134. Tsotsas, E., 2002. Wrmeleitung und Dispersion in durchstrmten Schttungen. In: VDI-Wrmeatlas, ninth ed., Chapter Mh1-14. Springer, Berlin. Turner, N.A., Vulfson, E.N., 2000. At what temperature can enzymes maintain their catalytic activity? Enzyme Microbial Technology 27, 108113. Xiu, G.-H., Jiang, L., Li, P., 2001. Mass-transfer limitations for immobilized enzyme-catalyzed kinetic resolution of racemate in a xed bed reactor. Biotechnology and Bioengineering 74 (1), 3139.

You might also like