Bioprocess Engineering 13 (I995) 1 11 t~' Springer-Verlag 1995

Description and evaluation of reciprocating plate bioreactors M. Lounes, J. Audet, J. Thibault, A. LeDuy

Abstract The environment in which live microorganisms has a major impact on their productivity. One important factor is the mechanical mixing that is used to promote good heat and mass transfer in bioreactors. In this paper, the performance of reciprocating plate bioreactors is first evaluated for their ability to produce high oxygen transfer coefficients. Pure water and a glycerol water (50:50 wt%) solution are used for this evaluation. Then, the performance of reciprocating plate bioreactors for the production of an exocellular polysaccharide (pullulan) by yeast Aureobasidium pullulans is analyzed in terms of quantity and quality of the polysaccharide. Results clearly show that a more efficient substrate utilisation is achieved with reciprocating plate bioreactors.

List of symbols A m amplitude C - constant in Eq. (1) D m diameter HL m height of liquid in the column KLa 1/s overall oxygen mass transfer coefficient P g/1 pullulan concentration PG W gassed power input Us m/s superficial gas velocity VL m 3 liquid volume

Greek letters ~,fi - constant in Eq. (1)

- gas holdup ~) s- 1 shear rate t

~1 Pa s viscosity q5 - fractional free area

1 Introduction The rapid development of the bioprocessing industry and the aggressive market competition are forcing biotechnological industries to continually strive to obtain

Received 22 February 1994

M. Lounes, J. Audet, I. Thibault, A. LeDuy Department of Chemical Engineering, LavaI University, Sainte-Foy (Quebec), Canada G1K 7P4

Correspondence to: J. Thibault

higher yields and to improve the quality and uniformity of production. These objectives are accomplished by critically examining each aspect of the production: selection and improvement of microorganisms, choice of substrate and nutrients, fermenter geometry, operating conditions, supervision and control, downstream processes, etc. This paper is mainly concerned with the choice of the mixing device, for aerobic fermentations. The advantages and limitations of a relatively new type of bioreactor, the reciprocating plate bioreactor (RPB), are discussed with respect to the other more conventional bioreactors and some experimental results are presented.

Rheological properties of culture media may change drastically during the course of a fermentation and, as a result, strongly influence bioreactor performance. For example, in typical fermentations where exocellular polysaccharides are produced, the culture medium initially exhibits Newtonian behaviour and then the rheology of the broth progressively changes to a highly viscous and non-Newtonian behaviour. For low viscosity fermentation broths and for microorganisms and product with low shear sensitivity, almost any type of bioreactors can be used, provided it leads to adequate homogeneity and good oxygen mass transfer. However, when high density cell slurries and/or high biopolymer concentration are produced, it becomes very difficult to mix the fermentation broth, resulting in poor distribution of oxygen and nutrients in the bioreactor. Non-Newtonian rheology, tendency to aggregate and shear sensitivity add to the problems [1]. In the selection of a mixing device for a particular application, it is therefore necessary to consider the homogeneity of the fermentation broth, the shear stress produced by the mixing element, the oxygen mass transfer efficiency and the ease of scale-up. These factors will be discussed in turn by focusing the discussion on mechanically agitated bioreactors and, more particularly, on three types of mixing devices: impellers, helical-ribbon screw and reciprocating perforated plates.

1.1 Homogeneity of the fermentation broth A large number of agitator geometries are used in the industry. An excellent summary of numerous types of bioreactors with their associated mixing devices has been presented in a review article by Schiigerl [2]. Some stirrer types (turbine, paddle, MIG, etc.) are more frequently used than others for highly viscous aerated reactors [3]. However, for highly viscous fermentation broths, remote clearance impellers such as

Bioprocess Engineering 13 (1995)

Rushton turbines or marine impellers do not lead to good mixing. Under highly viscous and non-Newtonian conditions, the mixing state within a bioreactor using Rushton impellers can be divided into three zones: a micromixing zone in the vicinity of the impeller where mixing is vivid, a macromixing zone where the broth circulates very slowly with very little intimate mixing and a stagnant zone in the periphery of the bioreactor where the broth is motionless. Outside the central well mixed zone, the metabolic activities are significantly reduced due to a poor distribution of nutrients and low oxygen mass transfer. To circumvent this problem, it has been recommended to use close clearance or proximity impellers [4-6]. However, the use of proximity impellers does not necessarily guarantee to obtain an uniformly mixed system. Indeed, it has been observed that helical-screw ribbon impellers lead to a significant reduction of the shear mixing zone in the case of the mixing of shear thinning fluids [7]. If gas dispersion is required, agitators specially designed for mixing highly viscous fermentation broths (helical, screw and anchor stirrers) are not recommended [3]. Reciprocating Plate Bioreactors (RPB), with its particular geometry, possess desired characteristics to produce very efficient mixing with the result that microorganisms, nutrients and air bubbles are properly distributed [8]. For mammalian cells which have a high shear sensitivity, the axial type of flow is preferred rather than radial type flow because of the more uniform local energy distribution [9]. In RPB, due to the upward and downward motions of the perforated plates, the fermentation broth in one elementary volume is exchanged with neighbouring elementary volumes, resulting in a good mixing in the vertical direction. In the horizontal direction, good mixing is produced by the uniform distribution of the perforated plates, and the formation, destruction and reformation of the ring vortex. By increasing the frequency of reciprocation, very intense mixing can be achieved.

In poorly mixed bioreactors, characterized by long mixing times, as cells move around the bioreactor and encounter spatially inhomogeneous conditions, the metabolism of the cell may always be in a transient state and never achieve steady state conditions. Several experimental investigations have been reported on the influence of transient conditions on metabolism. For instance, variations in dissolved oxygen levels produce, most often, a significant decrease in biomass to substrate yields and a decrease in growth rates [10].

1.2 Shear stress In the previous section, it was mentioned that inhomogeneity in the fermentation broth can have a major influence on the metabolic functions of the cells. On a smaller scale, turbulent velocity fluctuations may have a profound impact on the morphology and the metabolic state of certain microorganisms [10]. Shear stress may influence growth rate, cellular volume, metabolite production rate and distribution, and membrane permeability [11]. For plant cells, in addition to mechanical disruption and loss of viability, it has been reported that, to react to shear stress, cells use available substrate to build a heavier cell wall [1]. Large velocity fluctuations, resulting in

shear stress, may also have an influence on the molecular weight of biopolymers. On one hand, shear stress is necessary to promote good heat and mass transfer and, on the other hand, excessive shear stress is detrimental to the microorganism functions. It is therefore highly desirable to use a mixing device that will create the adequate shear stress, distributed as uniformly as possible within the bioreactor. Toma et al. [12] have introduced a neologism to describe the inhibition of shear stress on microbial growth and productivity. They use the term turbohypobiosis. It is the shear stress rather than the shear rate that appears to be the key variable in assessing the shear sensitivity of cells. The shear rate is equal to the velocity gradient between adjacent fluid layers, and the shear stress is the product of the shear rate and the viscosity. An excellent review of shear effects on suspended cells has been presented by Merchuk [11].

According to the statistical theory of turbulence, when the mixing device transfers its energy to the fluid, large vortex elements are produced. Large vortices or eddies are disintegrated into smaller vortices and yet into smaller ones until ultimately the energy is dissipated as internal energy. What is most important is the local energy dissipated per unit volume. In the design or selection of a mixing element, it is therefore highly desirable to have a mixing system that distributes as uniformly as possible its energy to the fluid to create uniform mixing and homogeneous conditions within the fermenters without unduly adding stress on the product and microorganisms. With flat-blade turbines, propellers or paddies in stirred reactors, the energy given to the fluid is highly concentrated and very large eddies are created at the tip of the mixing element. High velocities and high velocity gradients are thereby created in the vicinity of the impeller where the shear rates can be more than 100 times larger than those in other areas [11]. The use of two or more sets of impellers mounted on the same shaft, turning at a lower rotational speed, is an attempt to more adequately distribute the energy within the fermenter [13]. However, the problem of a static outer zone still remains for viscous fermentation broths. With a helical-ribbon screw mixer, the energy is distributed more uniformly and it has been shown that less energy is required to reach a given degree of uniformity than turbine-type impellers [14]. However, inadequately mixed zones with a HRS mixer may be created in highly viscous systems. In a reciprocating plate bioreactor, there is a large number of perforated plates, occcupying the whole cross- sectional area of the bioreactor. The axial reciprocating motion of the plates imparts energy more gently and uniformly to the fermentation broth than radial mixing devices. In RPB, the shear stress depends on the speed of reciprocation, the diameter of the plate perforations and the open area fraction of the plates. For a constant power input per unit volume, an increase in the number of plates would have for consequence to distribute more uniformly the energy to the liquid thereby reducing the shear on the fermentation broth. On possible disadvantage of a RPB is the sinusoidal motion of the plate that creates time-varying shear stress within the bioreactor.

It is clear that bioreactor design should consider the fluid dynamic behaviour and should have for objective to have a homogeneous shear rate field throughout the bioreactor.

M. Lounes et al.: Reciprocating plate bioreactors

1.3 Oxygen mass transfer rate In all aerobic fermentations, oxygen supply to growing cells is often critical. Of all the nutrients involved in an aerobic fermentation, oxygen is often the most difficult to provide in non-limiting quantities [1]. It is evident that the cells will only be able to utilize nutrients only if oxygen can be maintained at a high enough concentration in the vicinity of the cells. Due to the relatively low solubility of oxygen in aqueous systems, it is important that the oxygen mass transfer between the dispersed gas phase and the liquid be as efficient as possible. Even temporary depletion of dissolved oxygen could mean irreversible cell damage [15]. The evaluation and comparison of the performance of oxygen mass transfer within a mixing vessel is usually quantified by the volumetric mass transfer coefficient KLa. It has been clearly shown that KLa is not uniform throughout most mixing vessels. It is usually large in the vicinity of the mixing element and lower away from the region of vivid mixing. This difference is accentuated as the rheology of the fermentation broth evolves to a highly viscous system. The rheological properties of the fermentation broth will therefore greatly influence the oxygen mass transfer rate. Higher KLa values are normally reported for radial flow impellers than for HRS. This is undoubtedly valid for a system of low viscosity but not necessarily true for non-Newtonian viscous systems since KLa determinations have been performed under conditions of nonuniform mixing. If the dissolved oxygen probe is placed near the impeller tip, higher KLa values would be observed and lower values would be obtained away from the impeller. In a reciprocating plate bioreactor, KLa depends on the number of plates, i.e. the spacing between the plates. When other operating conditions are identical, higher KLa values are obtained with a larger number of plates due to higher power consumption [16].

1.4 Scale-up In the selection of a mixing device, it is important to be able to transpose the optimum process conditions, found at the laboratory scale, to a larger scale. This is not a simple task for most types of bioreactors and definitely more difficult than other processes in the chemical industry because biological principles must also be considered [17]. The two most common approaches are to try to obtain the same KLa values or an equal agitation power per unit volume of liquid for the smaller and greater vessels [15]. Biochemical engineers are in general agreement that to maintain the volumetric oxygen transfer coefficient at the same value in different fermenters is one of the best methods to scale up aerobic fermentation systems [18]. In systems, where KLa greatly varies with position, this is not a simple task. In a reciprocating plate bioreactor, it has been shown that, apart from possible pressure effects, KLa was not a function of position [19]. Godfrey et al. [20] argue that the use of rotating agitators may be cheaper but it is considered that scalling-up of such column is more difficult than for a reciprocating plate column RPC. In RPC, the energy dissipation rate is more uniform. In addition, once constructed, radial agitated columns offer one degree of variation of agitation intensity, whereas reciprocating plate

columns offer the possibility to vary the frequency and the amplitude of agitation.

One of the most important parameters in agitated systems is the shear stress of the agitated liquid and its maintenance at a constant value is an important criterion in scale-up [18]. An important limitation, with radial mixing devices, is that, for equal power input per unit volume, the velocity of the impeller speed, and as a result the shear stress, increases with the diameter of the impeller. The RPB offers some potential advantages over conventional stirred tanks and other bioreactors where radial mixing is predominant. Because, for axial mixing devices, flow patterns of the gas and liquid phases are more easily characterized, scale-up can be done with greater confidence [15].

2 Materials and methods

2.1 RPB systems Two reciprocating plate bioreactors have been constructed in our laboratory. Schematic representations of the two bioreactors are given respectively in Figs 1 and 2. Their respective diameter to height ratios are 1 : 12 and 1:2. The column geometrical parameters are presented in Table 1.

The first bioreactor has an 8-1 working volume. It is made of two concentric plexiglass tubes 1.26 m high (Fig. 1). The annular section, in which water at 30 ~ continuously circulates, is used as a heat exchanger to maintain a constant temperature inside the column where biological reactions occur. The central tube has an internal diameter of 101.6 ram. The plate stack consists of 18 perforated stainless steel discs, 96 mm in diameter and 1.25 mm thick, mounted on a central shaft and uniformly distributed with a distance of 50 mm between adjacent plates. The perforations, 6.35 mm in diameter, are distributed on an equilateral triangular pitch. The fractional free area is 0.28.

The second reciprocating plate bioreactor has a working volume of 13 litres. This bioreactor is made of a central stainless steel tube of 430 mm high and 206 mm in diameter (Fig. 2). A water jacket, in which water at a desired temperature circulates, surrounds the central cylindrical section and serves as a heat exchanger to maintain a constant temperature inside the column. Two viewing windows have been inserted to visualize the evolution of the fermentation. The height and the width of the front window (F1) are respectively 390 mm and 75.7 mm whereas the dimensions of the rear window (F2) are 240.5 mm and 75.7 mm. The plate stack consists of 6 perforated stainless steel plates, 194 mm in diameter and 1.25 mm thick, mounted on central shaft and uniformly distributed with a distance of 50 mm between adjacent plates. The perforations, 6.35 mm in diameter, are distributed on an equilateral triangular pitch. The fractional free area is 0.33. The top and bottom lids are made of stainless steel. The top lid has appropriate openings for the reciprocating axis of the plate stack, for gas-outlet, for introduction of antifoam agent, inoculum and substrate. It also has openings for a sampling port, a differential pressure transducer and receptacles for the

Bioprocess Engineering 13 (1995)

~ FT

C Air compressor D Gas distributor DP Differential pressure transducer E i ELectrodes F Fi[tre FCD Frequency control disk FT Force transducer IL Interface [eve[ N Motor drive P Pressure transducer

Rotameter

Fig. 1. Schematic diagram of the first reciprocating plate bioreactor (RPB-1)

dissolved oxygen probe and the pH probe. The outlet gas must go through a small water-cooled heat exchanger, for which the main function is to reduce water loss from the fermentation broth. The four openings in the bot tom lid serve to introduce the air, to empty the column and to connect the pressure and differential pressure transducers.

A schematic diagram of the plate stack is presented in Fig. 3. The reciprocation of the plate stack is generated by a mechanical system composed of a variable speed motor, controlled by a microcomputer, a tenfold reducing speed transmission and a connecting rod. An aluminium disc, containing 100 uniformly distributed perforations and mounted on the output shaft of the reducing transmission, is used in conjunction with an infrared optical switch (HOA- 2001, Honeywell) to measure and control with a microcomputer the frequency of the reciprocation by manipulating the power to the motor.

The plate stack is attached directly to a force transducer (Intertechnology Inc., Don Mills, Ontario, Model 363-D3-300- 20P3) which gives the instantaneous measurement of the force necessary to reciprocate the whole mixing mechanism. In addition, the instantaneous pressure at the bot tom of the column as well as the pressure drop inside the column were measured respectively with a pressure transducer P (Ashcroft, Stratford, Connecticut, Model ASHK1 G100 D7 MO201) and

FCD

~ Z

-2 'i

. . . . .

PJ

pH = ~ - - ~ Gas outlet

~R ~

C

Fig. 2. Schematic diagram of the second reciprocating plate bioreactor (RPB-2)

Table 1. Column geometrical parameters and ranges of operating conditions

Parameters RPB-1 RPB-2

Column diameter, cm 10.16 20.6 Column height, cm 126 42 Plate diameter, cm 9.60 19.4 Piate spacing, cm 5 5 Plate thickness, cm 0.125 0.125 Size of perforations, cm 0.635 0.635 Number of plates (stainless steel) 18 6 Fractional free area, % 0.28 0.33 Amplitude of plate vibration, cm 5 5 Frequency of plate vibration, Hz 0-2.0 0-2.0 Superficial gas velocity, cm/s 0-1.86 0-0.4

with a differential pressure transducer DP (Celesco, Canoga Park, California, Model DP31-0030 111). The ranges of pressure of the P and DP transducers were respectively 0 to 413 kPa and -103 to 103 kPa. The average power consumption per unit volume of the RPB can be calculated from either the pressure P or the pressure drop DP data [21].

Compressed air for aeration is fed at the base of the column after passing through a mass flow meter (Matheson, Model 8272-0414), a sterile air filter and a rotameter. A stainless steel

M. Lounes et al.: Reciprocating plate bioreactors

Fig. 3. Schematic view of the plate stack

plate containing a great number of 1 mm diameter holes is used to uniformly distribute small bubbles at the base of the column. The pressure and the temperature of the exit gas were monitored with a pressure transmitter (Model 3-8400, Signet Scientific) and a temperature transmitter (Model 3-8300, Signet Scientific).

The experimental data (reciprocating frequency, force, pressure at the base of the column, pressure drop) were sampled via a multiplexer, with an IBM PS/2 Model 30 microcomputer equipped with an IBM Data Acquisition and Control Card (Number 6451502). The experimental data were stored on a hard disc for further processing.

The data acquisition system is also comprised of a New Brunswick Scientific (NBS) Model DO-40 Dissolved Oxygen Analyzer. The dissolved oxygen probe used in this investigation was a galvanometric oxygen electrode (NBS Model 600) having a membrane thickness of 25 gm. In the first bioreactor, the dissolved oxygen probe was installed inside an ex situ continuous measuring cell where the culture medium in the bioreactor is continuously circulated via a peristaltic pump. In the second bioreactor, it was installed inside the column and the first two plates have been perforated to accommodate its passage.

2.2 Microorganisms and inoculum Aureobasidium pullulans (strain 2552) was obtained from our collection. It was adapted and maintained by a multiple- transfer technique in culture medium. The organism was cultured in a liquid medium composed of the following constituents in g/L: (NH4)SO 4, 0.6; KiHPO 4, 5.0; MgSO4.7H20, 0.41; NaC1, 0.1; yeast extract, 0.4 and sucrose (from supermarket), 50.0. The initial pH was adjusted to 5.3-5.5 using concentrated HC1.

To prepare the inoculum for the first reciprocating plate bioreactor (RPB-1), 900 ml of medium in 21 Erlenmeyer flask were autoclaved for a period of 10 min at 121 ~ and 0.2 MPa

(15 psig). The sterile medium was then inoculated with 45 ml of liquid culture of A. pullulans and incubated at 25 ~ for 4 days on a rotating shaker (150 rpm).

RPB-2 inoculum was prepared with 1.5 1 medium shared (2 x 750 ml) into two 2 1 Erlenmeyer flasks. These Erlenmeyers were autoclaved for a period of 15 min at 121 ~ and 0.2 MPa (15 psig). Each of them were inoculated with 37 ml of liquid culture and incubated at 25 ~ for 4 days on a rotating shaker (150 rpm).

2.3 Sterilisation of RPB systems RPB-1 was too tall to be sterilized in the autoclave. Instead, it was chemically sterilized using the following procedure. The bioreactor and the dissolved oxygen measuring cell were first cleaned with tap water, before assembling it to the mechanical mixing and data acquisition system. A solution of 0.2% sodium hypochlorite was introduced inside the column using a centrifugal pump (Model PP-1, Pump Co.). The mechanical mixing system and the air supply were then turned on for a period of 10 min, in order to sterilize the system. The same procedure was repeated twice with a solution of 0.2% sodium hypochlorite and three more times with sterile distilled water.

RPB-2 was designed to be sterilized in the autoclave. The bioreactor was first cleaned with tap water and detergent before filling it with 12 1 of non-sterile culture medium (pH 5.5). The bioreactor and its content were autoclaved for a period of 45 rain at 121 ~ and 0.2 MPa (15 psig).

2.4 Experimental procedure With the first bioreactor, the culture medium for the fermentation experiments was of the same composition as the one used for inoculum preparation. Nine liters of medium in a 20 1 Pyrex carboy, adjusted to pH 5.3-5.5, were sterilized in an autoclave at 121 ~ and 0.2 MPa (15 psig) for 45 minutes, cooled and then inoculated with 900 ml of inoculum. Eight liters of the content of the carboy were aseptically transferred into the previously sterilized RPB system. With the second bioreactor, the sterilized bioreactor and its culture medium (11-12 i) were cooled and then inoculated with 1.5 1 of inoculum. The total volume of the liquid was adjusted to 12.3 1.

With both bioreactors, the experiments were performed with a constant agitation frequency of 0.25 Hz and a gas flow rate of 0.5 win. Fifteen milliliters of culture medium were sampled immediately after inoculation of the bioreactors and subsequent samples of the same size were taken twice or once daily during the progress of the fermentation for biomass concentration, pullulan concentration and theological property measurements. Other variables were monitored by the data acquisition system described above.

2.5 Analytical procedures The dry weight of the cells in the broth was determined as follows. A sample of 10 ml of the liquid broth, diluted twice with distilled water to reduce its viscosity, was centrifuged at 11300 rpm (15000 g) for 15 minutes. The supernatant~ conserved. The cell pellets were washed with 1 volume of

Bioprocess Engineering 13 (1995)

distilled water and centrifuged again. The second supernatant was combined with the first one. The washed cells, transferred to pre-weighted aluminium dish, were then dried to a constant weight in an oven at 105 ~

Pullulan was recovered by precipitation using two volumes of ethanol for one volume of supernatant. The precipitated polysaccharide was filtered through a previously dried and weighed aluminium dish containing a glass fibre filter (Gelman type A, 45 gin) and dried to a constant weight at 105 ~

The rheological property of the whole fermentation broth samples was measured using a Contraves Model Low Shear 30 rheometer at a constant temperature of 30 ~

3 Hydrodynamic characteristics The main purpose of using a RPC is to obtain a high and uniform oxygen mass transfer coefficient throughout the bioreactor. The efficiency of a bioreactor is often measured in terms of its ability to transfer as much oxygen as possible to the liquid phase with the minimum energy requirement. This efficiency criterion is justified by the fact that in most fermentations, the dissolved oxygen concentration is very low and often a limiting factor for greater productivity. Most KLa data, reported in the literature, were obtained with model fluids and rheologically simple broths so that fluid properties were better characterized and controlled.

The results of the oxygen mass transfer KLa, obtained in this investigation by the gassing out method with pure water and an aqueous glycerol solution must be compared respectively with the results obtained with other mixing devices to properly evaluate the efficiency of a reciprocating plate column for oxygen mass transfer. Although the physical properties of the fluids and operating conditions, used by many investigators, are hardly comparable over a wide range, their results have a common trend of increasing the oxygen mass transfer coefficient values with increasing agitation and gas flow rate. In order to compare the efficiency of mixing of different contacting devices, the latter two variables are usually expressed in terms of power input per unit volume Pc~ VL, and superficial gas velocity Ua. The effects of the two parameters have been combined by many researchers to propose correlations having the following form:

KLa=C -~L (Ua)e" (1)

Despite the fact that the comparison can only be made over a limited range of power input, KLa results obtained in the present investigation were compared with data calculated from correlations proposed by other researchers using both similar and different mixing devices. Table 2 summarizes the various correlations obtained in the present investigation as well as the correlations of Baird and Rama Rao [22] for a reciprocating plate column, Linek et al. [23] for Rushton

Table 2. K~a correlations for different mixing devices

Authors Equations Range of Validity Column Parameters (PG/VL in W/m3; U G in cm/s)

Water

/p ~0.59 This work: RPB-1 Kza=0.00723~7 j U~ "44 290<~PJVL<~7200 HL=1058 mm

kVL / 0.68 ~< U c ~< 1.86 D = 101.6 mm H L/n = 10.41 /p \0.632

This work: RPB-2 KLa =0.281 ( - e l U~ .... 35<~PjVL<~6600 HL=385

\ VL / 0.05 <~ U c ~ 0.40 D = 206

H L / D = 1.87 //9 \0.364 Baird and Rama Rao (1988) KLa=4.86 ( - e l U~SS 23 <~PJV~.<~3620 H L ~ 2000

(Reciprocating column) \ /VL " 0.49 ~ U a ~< 0.99 D = 50.8 HL/D ~ 40 /v \0..3

Linek et al. (1987) Kfa=O.O0495 {-G } UOG. 4 lO0<~PG/VL<~2600 HL=290

(Rushton turbines) kVL / 0.212 ~< U G ~< 0.424 D = 290 Hz /D= 1.0

Tecante (1991) (HRS) KLa = 0.0622 U ~ 23 ~< PJ V L <~ 375 H L = 323

0.386 ~< Uo ~< 0.749 D = 210 HLID=l.54

Water-Glycerol(50:50)

This work: RPB-2

Nocentini et al. (1993)

(Multiple Rushton turbines)

"PG ~0.735 0.668 KLa=0'0039 \ ~ UG 35<~PJVL<~6600 HL=385

0.05 ~< U c ~< 0.40 D = 206 H L/D = 1.87

[ ' G ] U0.4 300<~PjVL<~9000 HL=920 KLa =0'05 \VLj G , 0.464 <~ UGh< 1,035 D=230

HL/D=4.0

M. Lounes et ai.: Reciprocating plate bioreactors

10 -1

10 -2

d

10 -3

10-4 1

10 4

10-2

10 -3 @ 5

U~ = 0.15 cm.s -t

UG = 0.40 cm.s-' I I

101 10 2 10 3 10 4

PG/VL, W.m -a

Water

O RPB-I~ This work �9 RPB-2 J

-I- Tecante (1991) �9 Baird et Rama Rao (1988) �9 Linek et al. (1987)

Water-Glycerol solution �9 RPB-2 Thiswork

Nocentini et al. (1993)

Fig. 4. Comparison between KLa values in RPB, and those predicted by empirical correlations in HRS and Rushton agitated columns as a function of the gassed power input per unit volume

turbines and Tecante [14] for a Helical Screw Ribbon mixer, all obtained for pure water. The results obtained for a glycerol- water solution in the RPB-2 were compared with the correlation of Nocentini et al. [24]. The correlations for RPB-1 and RPB-2 were obtained respectively for two and four different gas superficial velocities. Table 2 also provides information on the type of mixing device, the range of validity of each correlation and the height to diameter ratio of the experimental system. All the experimental and predicted data of the volumetric oxygen transfer coefficients are plotted on Fig. 4 as a function of the power input per unit volume for two different gas superficial velocities.

For the first reciprocating plate bioreactor (RPB-1), the correlation obtained for pure water, in the range of superficial velocity varying from 0.68 to 1.86, is nearly identical to the correlation of Linek et al. [23] for a Rushton turbine. These correlations and all the correlations of Table 2 were then

Table 3. KLa correlations for the two regimes in RPB-2

Fluid First Regime Second Regime

Water KLa = 1.6769 U} m

Glycerol KL a = 0.775 U~ ~ -water

KLa = 0.0196 ( ~ ) '~ U~ 195

KLa =0.336 x 10 -4 U~ 624

evaluated for smaller values of the gas superficial velocity and compared with the results obtained with RPB-2 at these superficial velocities. For RPB-1, KLa values were very close to the values that are normally encountered with other mixing devices and follow very well the correlation expressed by Equation (1) over a wide range of power input per unit volume. For RPB-2, there are two distinct regimes of operation. At lower power input per unit volume, the increase of KLa with the power input per unit volume is lower than what is commonly reported in the literature whereas it is quite steep at higher values of the power input. The correlations for water and the glycerol-water solution, given in Table 2, were obtained using the data for the two regimes of operation combined. It is interesting to note that the exponent of the power input per unit volume of the correlation, for water in RPB-2, is very similar to the one in RPB-1. However, for a proper evaluation, it is important to report the correlations obtained in both regimes. These results are presented in Table 31 This two-regime phenomenon was previously reported by Perez and Sandall [25] with a Rushton turbine for pure water and carbopol solutions where KLa was independent of agitation until the speed of rotation of the turbine reached 200 rpm. It can be postulated that, at low agitation speeds, the energy given to the fluid is not intense enough to impart its energy to the gas phase and disrupt the gas bubble flow. The gas bubbles are simply deviated without being significantly affected by the mixing element. Increasing the intensity of agitation results in the production of smaller bubbles that have a greater probability to be retained for a longer period of time in the bioreactor and thereby leading to an increase in the gas holdup. Fig. 5 presents a plot of the gas holdup in RPB-2 as a function of the power input per unit volume. It is evident that there are also two distinct regimes for the gas holdup. As the power input per unit volume is increased, the increase of the gas holdup and the decrease of the average bubble diameter result in a large increase of KLa. Indeed, in the first regime, the exponents on PG/VL of the correlations of Table 3 for RPB-2 are respectively 0.296 and 0.159 for water and the glycerol-water solution, in the range of gas superficial velocities varying from 0.05 to 0.40 whereas they are 1.017 and 1.308 for the second regime.

The comparison with the results of Nocentini et al. [24] for glycerol-water solution must be done with caution since they were obtained in a range of superficial gas velocities different than the range of the results obtained with RPB-2 (Table 2). The range of superficial gas velocity, chosen in this investigation, was limited by the capacity of the gas flow rate controller and restricted to the volumetric gas flow rates (0-0.65 vvm) normally used in our fermentations. Nocentini

Bioprocess Engineering 13 (1995)

0.12

0.09

%

# 0.06 13 O I

0.03

a 0 0.15

0.12

% 0.09 0,. :J

-~ 0.06

0,03

UG=0.15 cm.s -1

lk " ' � 9 A ~ ~ ~

, ' 1

0""

UG= 0.40 cm-s -1

�9 Water-gtycero[ sotution

�9 Water

g l '

0" , i

.0

,A

0 I E I I I I l l l t I I E I l l l l t I I I I l l [ !

10 ~ 10 2 10 3 10 ~

b PdV, W m -3

Fig. 3a, b. Variation of the gas holdup in RPB-2 as a function of the gassed power input per unit volume

et al. [24] found slightly higher KLa values at low to medium glycerol concentrations (5-50 wt%), than the values that they obtained for pure water. They argued that this increase is due to the effect of glycerol to modify the liquid behaviour from coalescing to noncoalescing and thereby increasing the gas-liquid interfacial area. This is clearly shown in Fig. 4 if their results are compared with those obtained for water in the RPB-1 and by Linek et al. [23]. With RPB-2, KLa values obtained for the water-glycerol solution was significantly lower than those obtained in water and the phenomenon reported by Nocentini et al. [24] was not observed over the range of power input per unit volume and gas superficial velocities considered in this investigation.

It is interesting to note that many of the KLa correlations, published in the literature, are usually given for ranges of superficial gas velocities that exceed the gas flow rate that would normally be used in fermenters so that most correlations are not of great use for inferring KLa values for a fermentation broth. If there is indeed a presence of two regimes, it is important to report it.

The variation of KLa with the power input per unit volume, obtained by Tecante [14] with a FIRS mixer, shows approximately the same dependency than the variation observed in the present investigation with RPB-2, in the first regime of operation. This may suggest that a HRS may also have two distinct regimes of operation. Tecante [14] was not able to achieve high values of the power input per unit volume in water because of the occurrence of a large central vortex. He concluded that bioreactors using a HRS mixer

would not be very efficient in the early stage of a polysaccharide fermentation, where the viscosity of the fermentation broth would be very low.

In view of the results for RPB-1 and RPB-2 and the correlation of Baird and Rama Rao [22], it is clear that a correlation like Eq. (1) is useful to characterize the variation of KLa with the power input per unit volume and the gas superficial velocity for a given column. However, to be general, this equation should take into consideration geometrical factors of the column such as the height of the column. Much work needs to be done to derive such a general equation if it is to be useful for scaleup operations.

4 Comparative fermentation This section has not for objective to be a comprehensive fermentation study, but mainly to determine if the reciprocating plate bioreactor can produce a product of good quality. The main purpose was to test the performance of the RPB, designed and built in our laboratory [19, 21] for non-Newtonian viscous fermentation using the yeast Aureobasidium pullulans Pullulan fermentation was chosen for this study because it represents very well all other exocellular polysaccharide fermentations, where oxygen transfer is severely limited due to the high viscous and non-Newtonian character of the fermentation broth [26, 27]. Reciprocating plate bioreactors have previously been used with success for three different fermentations. Cyathus striatus for the production of antibiotics, Aspergillus niger for the production of citric acid and Zymomonas mobilis for the production of ethanol [8].

Fig. 6 presents the evolution of the biomass and pullulan concentrations as a function of the fermentation time for the two reciprocating plate bioreactors, used in this investigation, and their comparison with the data obtained by Lacroix [28] with a Rushton turbine and in an Erlenmeyer. No reliable data for pullulan fermentation with a HRS mixer were available to compare with the other results. Fig. 6 also presents the variation of the gassed power input per unit volume as a function of time. Table 4 presents the concentrations of biomass and pullulan at the seventh day of fermentation along with the maximum apparent viscosity, evaluated at a shear rate ~) of 0.01 s -1, and the day at which this maximum occurred.

The fermentation performed in RPB-2 led to the highest pullulan production. Results obtained in RPB-1 and the bioreactor using Rushton impellers [28] showed a similar pullulan production. As it was expected, lower pullulan production, for fermentations conducted in Erlenmeyers, was obtained. For biomass production, concentrations obtained in RPB-1 were significantly higher than the biomass concentrations achieved in the other three cases (Fig. 6). It appears that the geometry of RPB-1 favours biomass production and the geometry of RPB-2, pullulan synthesis. This difference in geometry (HL/D more than five times larger for RPB-1 and slightly larger free open area for RPB-2) resulted in a higher gassed power input per unit volume in RPB-1 as shown in the bottom curves of Fig. 6. In addition, as it will be discussed below, the viscosity in RPB-2 is significantly higher

M. Lounes et al.: Reciprocating plate bioreactors

7 ....I o3

~3 E

,0

t-

14

12~

&

2

0 25 - -

20

15

10

5

0

~00

"?, 30O E

~. 200 J

100

O

�9 RPB-1 �9 RPB-2 �9 Rushton A Erlenme

0 1 2 3 Z, 5 6 7 8 9 10 11 Time,d

Fig. 6. Evolution of the biomass and pullulan concentrations and the gassed power input per unit volume as a function of fermentation time

Table 4. Performances of pullulan fermentation under different agitation modes. Concentration data taken on the seventh day

Concentration (g/L) RPB-1 RPB-2 Rush ton Erlenmeyer

Biomass 12.3 6.9 3.0 7.3 Pullulan 15.8 23.8 13.3 11.5 Maximum viscosity 4 28 18 8 (Pas) Day of occurrence 3 5 9 9

than in RPB-1 so that it should have required more energy to mix the fermentation broth. With these two effects combined, it is clear that significantly more energy per unit volume was dissipated in RPB-I, resulting in a higher shear stress. More fermentations are now being performed to determine the contribution of the various factors (shear stress and geometrical factors) on the respective quantities of biomass and pullulan formed. It is important to note that the total anabolic conversion was nearly identical for the two reciprocating plate bioreactors and much higher than the conversion achieved in the other two cases.

The objective of these fermentations is to have as much product as possible with the longest polymeric chains. A high molecular weight of polysaccharide is a product quality indicator and the associated apparent viscosity is therefore a measure of quality. Thus, the level of viscosity, achieved

100

~- 10-~

10 -2

10 -~

5.6

6.6

RPB-2 f=0.25Hz T=30~

1 . 0 ~ _ 1.6

0.6 -''- = = ~

10 -2 I ~ I I I I I L I I I I t l t l H I I i I l t l j l I I I I I I I H

10 -1 100 101 102 ) j S 1

Fig. 7. Viscosity rheograms of the fermentation broth samples for different fermentation times in RPB-2

during fermentation, was chosen as the main comparative parameter to evaluate different mixing devices. Fig. 7 presents the rheograms of a typical fermentation performed in RPB-2 as function of time. There is a drastic increase of viscosity up to the sixth day and thereafter, it decreased slightly. It is hypothesised that pullulanolysis may play a role in this viscosity reduction [28]. In RPB-1, the maximum viscosity was reached on the third day of fermentation. On subsequent days, the fermentation broth started rapidly to show a viscosity decrease until the end of the fermentation. In RPB-2, the maximum viscosity (20 Pa.s, ~ = 0.01 s -1) is attained on the fifth day and it is approximately 7 times higher than in RPB-1 (~4 Pa.s, ,)=0.01 s-l). Also, the degradation of the broth consistency is slower and less apparent in RPB-2 where the viscosity is still near its maximum value on the sixth day of fermentation.

To evaluate the quality of the fermentation broth, in a way that is independent of the quantity of polysaccharides produced, it is more significant to use the viscosity to pullulan concentration ratio 0lIP). This ratio is similar for RPB-2, Rushton impellers and Erlenmeyers. The ratio seems to reach a plateau as the fermentation progresses. However, in RPB-1, the r//P ratio reached a maximum value at the end of the first day and then, decreases progressively. This diminution could be interpreted as a reduction in the product quality.

Thus, it can be concluded that fermentations performed using RPB-2 lead to an efficient polysaccharide production in terms of quantity as well as quality. In addition, higher production level are achieved in significantly less time than in. bioreactor using a Rushton turbine. Moreover, it is important to note that the experiments and the rheological analyses for fermentations in the bioreactor using Rushton turbines and in Erlenmeyers were conducted with a temperature of 25 ~ Since the temperature used for RPB-1 and RPB-2 was 30 ~ it is obvious that the results must be interpreted with care. Indeed, in terms of viscosity, a 5 ~ difference does have a great impact

Bioprocess Engineering 13 (1995)

lO

10 ~

l~176 I , " "

..a q t & O"

/

/ " RPB-1 t/�9 4 �9 Rushton

10 3 ~ I A=~r[enmeyer

[L a.-

0 1 2 3 4 5 6 7 8 9 10 11 Time,d

Fig. 8. Apparent viscosity to pullulan concentration ratio as a function of fermentation time

on rheograms. It was est imated that at 25 ~ the viscosity is approximately 50% higher than at 30 ~ Also, the pullulan

product ion could be very much affected by this temperature

variat ion [29].

5 Conclusion This paper was concerned with the evaluat ion of reciprocat ing

plate bioreactors for its use in viscous fermentat ions. First, the oxygen mass transfer coefficient in these relatively new

bioreactors was evaluated with pure water and a glycerol-water

solution. The mass transfer coefficients in RPB-1 was found to be nearly identical to the correlat ion of Linek et al. [23],

obtained with Rushton turbines. For RPB-2, at the lower superficial gas velocities used in this investigation, there was

the presence of two regimes of operat ion which arise f rom the ability of the mixing element to break the gas into small

bubbles and to retain a high gas volume. The performance of reciprocat ing plate bioreactors

was also evaluated for the product ion of pullulan, a polysaccharide synthesised by the yeast Aureobas id ium pullulans. It was shown that the substrate util isation efficiency

was far greater for reciprocat ing plate bioreactors than for the other two modes of mixing. RPB-2 was part icularly efficient to produce a large concentra t ion a pullulan of a good quality.

It is envisaged to pursue the investigation to per form a comparat ive study, for at least three mixing mechanisms (Rushton turbines, Helical Ribbon Screw mixers and Reciprocating perforated plates) for their oxygen mass transfer

efficiency and their efficiency to produce a high quality product in pullulan fermenta t ion under an identical power input per unit vo lume and, more important ly, with the identical microorganism strain. The replacement of some perforated plates by static mixers in the reciprocat ing plate bioreactors is also considered.

References 1. Doran, P. M.: Design of reactors for plant cells and organs. In:

Advances in Biochemical Engineering Biotechnology (Ed.: A. Fiechter), Springer-Verlag 48 (1993) 113 168

2. Schiigerl, K.: Oxygen transfer into highly viscous media. In: Advances in Biochemical Engineering Biotechnology (Ed.: A. Fiechter), Springer-Verlag 19 (1981) 71 174

3. Schiigerl, K.: New bioreactors for aerobic processes. Int. Chem. Eng. 22, 4 (1982) 591-610

4. Nagata, S.: Mixing principles and appIications, John Wiley & Sons, New York, 1975

5. Skelland, A. H. P.: Mixing and agitation of non-Newtonian fluids. In: Handbook of Fludis in Motion, Chap. 7 (Eds: N. P. Cheremisionoff and R. Gupta), Ann Arbor Science, New York (1993) 179-209

6. Ulbrecht, ]. ]4 Carreau, P.: Mixing of viscous non-Newtonian liquids. In: Mixing of Liquids by Mechanical Agitation (Eds.: I. ]. Ulbrecht and G. R. Patterson), Gordon and Breach Science Pub., New York 1 (1985) 93 137

7. Tanguy, P. A.; Lacroix, R.; Bertrand, F.; Choplin, L.; Brito De la Fuente, E.: Finite element analysis of viscous mixing with helical ribbon-screw impeller. A1ChE ]. 38, 6 (1991) 939 944

8. Brauer, H.: Growth of fungi and bacteria in the reciprocating jet bioreactor. Bioprocess Engineering 6 (1991) 1-15

9. Prokop, A.; Rosenberg, M. Z.: Bioreactor for mammalian cell culture. In: Advances in Biochemical Engineering Biotechnology (Ed.: A. Fiechter), Springer-Verlag 39 (1989) 29-71 Bailey, I- E.; Ollis, D. F.: Biochemical Engineering Fundamentals, 2 nd Ed., McGraw-Hill Book Company, New York 1986 Merchuk, ]. C.: Shear Effects on suspended cells. In Advances in Biochemical Engineering Biotechnology (Ed.: A. Fiechter), Springer- Verlag 44 (1991) 65-95 Toma, M. K.; Ruldisha, M. P.; Vanags, ]. ].; Zeltina, M. 04 Leite, M. P.; Galinina, N. I.; Viesturs, V. E.; Tengerdy, R. P.: Inhibition of microbial growth and metabolism by excess turbulence. Biotechnol. Bioeng. 38 (199i) 552-556 Nocentini, M.; Magelli, F.; Pasquali, G.; Fajner, D.: A fluid-dynamic study of a gas-liquid, non-standard vessel stirred by multiple impellers. The Chem. Eng. ]ournal 37 (1988) 53-56 Tecante, A. C." Mass transfer in rheologically complex fluids in helical ribbon screw-agitated and aerated tank. Ph.D. Thesis, Univ. Laval, Quebec 1991 Moo-Young, M.; Blanch, H. W.: Design of biochemical reactors - Mass transfer criteria for simple and complex systems. In: Advances in Biochemical Engineering Biotechnology (Ed.: A. Fiechter), Springer- Veriag 19 (1981) 1-69 Veljkovic, V.; Skala, D.: Mass transfer characteristics in a gas-liquid reciprocating plate column. II. Interfacial area. Can. ]. Chem. Eng. 66 (1988) 200-210 Aiba, S.; Humphrey, A. E.; Millis, M." Biochemical Engineering. New York, Academic Press 1973 larai, M,. Factors affecting the scale-up of aerated ferementation processes. Int. Chem. Eng. 19, 4 (1979) 710 707 Lounes, M.; Thibault, ].: Mass transfer in a reciprocating plate bioreactor. Chem. Eng. Comm. 127 (1994) 169-189 Godfrey, ]. C.; Houlton, D. A.; Marley, S. T.; Marrocchelli, A.; Slater, M. ].: Continuous phase axial mixing in pulsed sieve plate liquid-liquid extraction columns. Chem. Eng. Res. Des. 66 (1988) 445-457 Lounes, M.; Thibault, l.: Hydrodynamics and power consumption of a reciprocating plate gas-liquid column. Can. ]. Chem. Eng. 71 (1993) 497 506 Baird, M. H. I.; Rama Rao, N. V.: Characteristics of countercurrent reciprocating plate bubble column. IL Axial mixing and mass transfer. Can. J. Chem. Eng. 66 (1988) 222-231 Linek, V.; Vacek, V.; Benes, P.: A critical review and experimental verification of the correct use of the dynamic method for the determination of oxygen transfer in aerated agitated vessels to water, Chem. Eng. ]. 34 (1987) 11-34

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

M. Lounes et al.: Reciprocating plate bioreactors

24. Nocentini, M.; Fajner, D.; Pasquali, G.; Magelli, F.; Gas-liquid mass transfer and holdup in vessels stirred with multiple Rushton turbines: Wter and water-glycerol solutions. Ind. Eng. Chem. Res. 32 (1993) 19 26

25. Perez, I. F.; Sandall, O. C.: Gas absorption by non-Newtonian fluids in agitated vessels. A1ChE J. 20, 4 (1974) 770-775

26. LeDuy, A.; Marsan, A. A.; Coupal, B.: A study of the rheological properties of a non-Newtonian fermentation broth, Biotechnol. Bioeng. 16 (1974) 6i 76

27. Rho, D.; Mulchandani, A.; Luong, J. H. T.; LeDuy, A.: Oxygen requirement in pullulan fermentation. Appl. Microbiol. Biotechnol. 28 (1988) 361-366

28. Lacroix, C.: Effet du pH sur la production en discontinu du pullulane par fermentation du sucrose. Master Thesis, Universit~ Laval, Quebec, Canada 1985

29. McNeil, B.; Kristiansen, B.: Temperature effects on polysaccharide formation by Aureobasidium pullulans in stirred tanks. Enzyme Microb. Technol. 12 (1990) 521 526

11