Oxygen Transfer into Xanthan Solutions in Partitioned Bubble Columns Koichi TERASAKA, Junya NITTA, Hisashi FUKUDA 3, Misa IWAHASHI, Takashi GOSHIMA and Satoko FUJIOKA Department of Applied Chemistry, Faculty of Science and Technology, Keio University, 3-4-, Hiyoshi, Kohoku-ku, Yokohama-shi, Kanagawa 3-85, Japan School for Open and Environmental Systems, Graduate School of Science and Technology, Keio University, 3-4-, Hiyoshi, Kohoku-ku, Yokohama-shi, Kanagawa 3-85, Japan 3 Enzyme Engineering Division, National Research Institute of Brewing, 3-7-, Kagamiyama, Higashi-hiroshima, Hiroshima 739-0046, Japan keywords; Yield stress, Bubble column, Oxygen transfer, Xanthan, Non-Newtonian liquid The escapable gas holdups and volumetric oxygen transfer coefficients were measured in several concentrations of xanthan aqueous solutions in four standard bubble columns, in which slug bubble flow was observed. The experimental escapable gas holdups were well estimated by the semi-theoretical Nicklin's equation modified to apply for non-newtonian xanthan aqueous solutions having yield stress. Although the escapable gas holdup increased with decreasing column diameter and increasing superficial gas velocity, it was hardly influenced by the concentration or apparent viscosity of xanthan aqueous solutions. The volumetric oxygen transfer coefficient increased with decreasing yield stress of the liquid, increasing diffusion coefficient, decreasing column diameter and increasing gas holdup. By correlation of all experimental results, an empirical equation was proposed. To design a more efficient bioreactor, the partitioning perforated plate which is applicable for larger column than those of Terasaka and Shibata (003a) was inserted into the standard bubble column. The gas holdup and volumetric oxygen transfer coefficient were measured in the partitioned bubble columns and compared with those in the standard bubble columns. The volumetric oxygen transfer coefficients in the partitioned bubble columns became larger than those in the standard bubble columns at a fixed superficial gas velocity even for large column diameter. Therefore, the partitioned bubble column bioreactors were developed for more suitable production of xanthan gum in this study. Introduction Xanthan gum, which is an exocellular heteropolysaccharide produced aerobically by Xanthomonas campestris, is a useful thickening additive in foods, cosmetics and pharmaceuticals so that its demand increases year by year. In industry, xanthan has been usually produced aerobically from glucose and so on using a stirred tank bioreactor because it is customarily believed that the stirred tank has higher oxygen transfer rate and better liquid mixing in highly viscous liquid than other types of bioreactors. When the xanthan aqueous solution has a yield stress, however, the liquid mixing comes to be drastically bad with increasing distance from the impeller. On final stage of xanthan fermentation, large stagnant zone was observed in a stirred tank (Suh et al., 99), where the oxygen feed to fungi was limited so that xanthan could not be produced. On the other hand, bubble column bioreactors presented well-mixing performance under the slug flow operation for high concentration of xanthan aqueous solutions (Suh et al., 99). Terasaka and Shibata (003a) also presented that the stagnant zone in non-newtonian liquids having a yield stress is smaller in a bubble column than in a stirred tank using the visualization technique. Schumpe et al. (987) and Eickenbusch et al. (995) studied the oxygen transfer into xanthan aqueous solutions in bubble columns and proposed correlation equations,
whereas the yield stress was neglected. Nicklin et al. (96) had presented a semi-empirical equation to predict the gas holdup for slug bubble flow in Newtonian liquids. Terasaka et al. (003b) modified their equation to predict the gas holdup in viscous liquids having a yield stress in bubble columns. The equation estimated well gas holdup for slug bubble flow through viscous media having a yield stress. Utilizing the presented equation for gas holdup, Terasaka and Shibata (003a) proposed a correlation equation for volumetric oxygen transfer coefficient k L a in viscous non-newtonian liquids having yield stress in standard bubble columns, which predicted within the accuracy of ±30%. Iwahashi et al. (004) investigated the xanthan production from Xanthomonas campestris using two typical bioreactors such as a stirred tank bioreactor and a standard bubble column bioreactor on laboratory scale. The xanthan was not suddenly produced in both bioreactors when the oxygen supply for fermentation came short at a specific concentration of xanthan solution. That is, those bioreactors may not have efficient oxygen transfer rate for demand higher productivity of xanthan in culture media. A novel bubble column bioreactor is, therefore, necessary to improve the xanthan productivity (Terasaka, 003). In this study, the partitioned bubble column was proposed to dissolve oxygen faster into xanthan aqueous solution than conventional bioreactors. At first, to clarify the behavior of oxygen transfer and bubble behavior, the volumetric oxygen transfer coefficient k L a, escapable gas holdup G and stagnant gas holdup GVS were investigated for some standard bubble columns. By correlating those experimental results, an improved empirical equation was obtained to estimate a volumetric oxygen transfer coefficient k L a. Secondly, the partitioned bubble column bioreactors, which consist of standard bubble columns inserted some partitioning perforated plates, were developed. The gas holdup and volumetric oxygen transfer coefficient were measured in the partitioned bubble columns and compared with those in the standard bubble columns.. Experimental. Characteristics of xanthan solutions used Table shows the physical properties of eight kinds of xanthan gum (Sigma-Aldrich Inc.) aqueous solutions, which were prepared between.5 and.3 wt% for in this study. To prevent decomposition, 0. wt% of formaldehyde was added into the xanthan solutions. The rheological characteristics of the xanthan solutions used were approximated using Herschel and Bulkley model (96) as follows: 0 () where, 0,, and are shear stress, yield stress, H-B model viscosity coefficient, H-B model viscosity index and shear rate, respectively. Figure shows the rheological behaviors of all solutions measured using a co-axial revolving cylindrical viscometer (MR-3: Rheology, Co., Ltd.). The yield stress was measured with the stress remaining after stopping revolution of the viscometer. All solutions had the yield stress and the ratio of shear stress to shear rate, / decreased with increasing shear rate Shear stress τ [Pa] Shear stress [Pa] 80 70 60 50 40 30 0 0 0 X- X-5 X- X-6 X-3 X-7 X-4 X-8 0 0 40 60 80 00 0 40 Shear rate [s - ] Shear rate γ [s - ] Fig.. Rheological characteristics of xanthan solutions used On Table, the density of xanthan aqueous solution was measured using a pycnometer. The interfacial tensionwas measured using a surface tension meter (CBVP A-: Kyowa Kagaku Co., Ltd.). Henry s constant of xanthan solutions H was experimentally evaluated by the saturated oxygen concentration divided by the partial pressure of oxygen in air. The diffusivity of oxygen in xanthan solution D L was evaluated by the time course of oxygen concentration in very small volume sample measured using a fluorescence oxygen sensor (FO-960: ASR Co., Ltd.) and the theoretical linear diffusion model.. Equipment and operation Figure shows a schematic diagram of the experimental apparatus. The superficial gas velocity U G was changed from to 0.0 m/s for all experimental conditions. The gas holdup is an important parameter for designing of bubble columns. When the solution has a yield stress, the.
very small bubbles were not able to escape from the solution due to less buoyancy than the yield stress. Although the escapable bubbles are oxygen source, the stagnant bubbles have lost the driving force due to gas-liquid equilibrium state. Therefore, the gas holdup for escapable large bubbles G and the gas holdup for stagnant small bubbles GVS should be measured separately. The escapable gas holdup was measured using an ultrasonic liquid level sensor (UD-500: Keyence Co., Ltd.). The xanthan solution is not suitable to use a popular liquid-manometers-method because the solution cannot ascend nor descend through narrow manometer tube due to the yield stress. The technique for ultrasonic liquid level sensor has been proposed by Terasaka et al. (003b). To measure the dissolved oxygen concentration in solutions, the fluorescence oxygen sensor (FO-960: ASR Co., Ltd.) was used. Although galvanic electrode oxygen sensor is more popular, the fresh liquid is needed on the probe s tip. The yield stress of xanthan solution let the liquid stagnant, so that the fluorescence oxygen sensor was more suitable than galvanic electrode oxygen sensor to measure the dissolved oxygen concentration as presented by Terasaka et al. (998). All those experimental data such as escapable gas holdup G, oxygen concentration c L and temperature T were recorded every second using a computer. The stagnant gas holdup GVS was determined from the apparent density of the dispersion measured using a pyknometer (Terasaka et al., 998). column. In this study, the bubble columns and partitioned plates were made of transparent acrylic resin. was a standard bubble column whose diameter is changed 0.03, 0.06, 0.0 or 0.0 m and height is. m. Those standard bubble columns were used to correlate the experimental data for gas holdups and volumetric oxygen transfer coefficients in each xanthan aqueous solution. Column B and Column C are partitioned bubble columns. Column B whose diameter and height were fixed at 0.0 m and. m was a bubble column quarterly partitioned by perforated plates whose hole diameter, opening ratio and height were fixed 0 mm, 30% and 0.95 m, respectively. Column C was vertically partitioned a ϕ0.0 m H.4 m bubble column to the 6 sections with a partitioned plate whose height was.0 m and layout was shown in Figure 3 together with Column B. Liquid was filled up h = 0.65, 0.88, 0.86 and.6 m in height for D C = 0.03, 0.06, 0.0 and 0.0 m of, respectively. For Column B, the liquid height h was 0.89 m. For Column C, the liquid height h was.6 m. Gas was fed into the center of each partitioned section. The equivalent diameters D C of each cross section for Column B and Column C, which was defined as four times hydrodynamic radius of each partitioned section, was evaluated 0.044 m. 5 9 0 4 Air compressor N gas cylinder 3 Mass flow controller 4 Pressure gauge 5 Three-way valve 6 Nozzle 7 Replacable bubble column 8 Thermo-regulator 9 Slug bubble 0 Ultrasonic level sensor Fluorescence O sensor Fluorescence O meter 3 Thermo-sensor 4 A/D converter 5 Personal computer 8 3 4 7 5 6 3 Fig.. Experimental apparatus Figure 3 shows the illustration of the bubble columns used in this study. On previous works, i.e., Terasaka et al. (003b) and Terasaka and Shibata (003a), used a standard bubble column, an airlift bubble column or small four-partitioned bubble Column B Column C Fig. 3. Bubble columns used. (A) Standard bubble column, (B) and (C) Partitioned bubble columns 3
. Theoretical. Escapable gas holdup The escapable bubbles G was estimated by the equation of Nicklin et al. (96) modified by Terasaka and Shibata (003a) as follows: 4 UG UL 0.35 G C () UG Fr where Fr and U G are Froude number (=U G /(D C g) 0.5 ) and superficial gas velocity. The superficial liquid velocity U L was zero due to the semi-batchwise operation in this study. The modification factor C is given for non-newtonian liquids having a yield stress. For Herschel and Bulkley model fluids, C R r R r C 3 0 RC 0.0 h / R C 3.4e C 0 r0 r0 (3) where R C, r 0 and h are a radius of bubble column, yield point in bubble column (=τ 0 /ρg) and static liquid height from gas distributor.. Volumetric oxygen transfer coefficient The volumetric oxygen transfer coefficient based on aerated liquid phase, k L a, was measured by the dynamic method (Schumpe et al., 987). After the feed gas into the bubble column was switched from air to nitrogen, the time course of dissolved oxygen concentration, c L, was measured. When a number of small bubbles cannot disengage from the liquid phase due to the yield stress, Schumpe et al. (987) removed the effect of stagnant bubbles as an unavailable oxygen sink on k L a by the relation of oxygen-liquid equilibrium, so that the oxygen balance in liquid phase is described as follows: ln c L Journal of Chemical Engineering of Japan, Advance Publication kla G L GVS H RT t (4) where L and R are liquid holdup and gas constant, respectively. 3. Results and Discussion 3. Standard bubble column Escapable gas holdup The slug bubbles were observed over the entire operating range in this study. Those slug bubbles smoothly rose without coalescence or breakage after they had been accelerated upward just above a gas distributor. Figure 4 shows the escapable gas holdup, G, in a typical standard bubble column (). G increased with increasing U G. Moreover, G did not depend on the rheological characteristics of xanthan solutions when the diameter of bubble columns D C was identical. Escapable gas holdup G [-] 0. Exp. Cal. X- X-5 X-6 X-7 X-8 D C =0.0m 0. Figure 5 shows effect of D C on G in Columns A. G increased with decreasing D C. That was caused by the reduction of the slug bubble velocity due to the increase of the wall friction. The escapable gas holdups were estimated by using the Nicklin s equation modified by Terasaka et al. (003b), i.e., Eqs. () and (3). The calculated results as lines in Figures 4 and 5 agreed fairly with the experimental values. Escapable gas holdup G [-] 0. Fig. 4. Escapable gas holdup D C [m] Exp.Cal. 0.03 0.06 0.0 0.0 Liquid: X-8 0. Fig. 5. Effect of D C on ε G Stagnant gas holdup The small bubbles could not escape from the liquid by their insufficient buoyancy due to the yield stress of the liquid. Figure 6 shows the stagnant small bubbles gas holdup, GVS, in steady state. Terasaka et al.(998) presented that very small bubbles accumulated in xanthan solutions due to yield stress. Stagnant small bubbles were often generated when large bubbles were crushed by shearing with wall and/or neighbor large bubbles. On the other hand, when the large bubbles rise rapidly through the aerated liquid phase, the stagnant small bubbles were often absorbed and wiped out from the liquid. The generation and disappearance of stagnant small bubbles balanced independently from U G and D C. Therefore, the
stagnant gas holdup GVS hardly depend on U G and D C. Liquid:X-8 Stagnant gas holdup GVS [-] 0. X- X-5 X-6 X-7 X-8 D C =0.0m 0. Fig. 6. Stagnant gas holdup in xanthan solutions Volumetric oxygen transfer coefficient Figure 7 shows the volumetric oxygen transfer coefficients based on aerated liquid phase k L a in a typical. Although k L a was increased with increasing U G, it hardly depended on the concentration of xanthan in solutions. Figure 8 shows the effect of column diameter D C on k L a in Columns A. For the solutions used in this study, k L a increased with decreasing D C in the same way as escapable gas holdup G in Figure 5. It was remarkable when the highest yield stress of the solutions, i.e., X-8. The oxygen transfer rate would decrease for the bubble column would be designed simply with larger diameter for scale-up. Terasaka et al. (003a) presented that rising slug bubbles let the liquid replace well in a bubble column, even if the liquid has a yield stress, so that liquid-mixing became better with developing slug flow. For slug flow condition, the shape of the slug bubbles was usually similar so that the gas-liquid interfacial area, a, increase with increasing escapable gas holdup G. k La [s - ] 0.00 Liquid Exp.Cal. X- X-5 X-6 X-7 X-8 D C =0.0m 0.000 0. Fig. 7. k L a in xanthan solutions in standard bubble column k L a[s - ] 0.00 0.000 0. To estimate k L a in viscous liquids such as xanthan solutions and other non-newtonian liquids having yield stress, Terasaka and Shibata (003a) proposed an empirical equation, where k L a was proportional to G 0.73 and (D L /τ 0 ) 0.5. According to the penetration theory, the oxygen transfer in liquid phase was related to a square root of diffusivity of oxygen in liquid phase. Moreover, the larger the yield stress, 0, the larger the volume of stagnant liquid zone. The effect on k L a of other important parameters such as U G, D C and the physical properties were included in the value of G. In this study, for the wider range of xanthan concentrations than in Terasaka and Shibata (003a), the correlation equation was reconstructed as follows: 3. DL k La 3.89 0 (5) G which are applicable for 3 < 0 < 6 Pa and < U G < 0. m/s. It is commonly understood that the slug bubbles have less specific surface area so that k L a must become small under slug flow operation. In this case, however, very small bubbles whose specific surface area is very large did not have available oxygen source in liquid phase. On the other hand, the large slug bubbles can mix well liquid phase so that they remove stagnant zone and then the oxygen transfer rate must be higher. The estimated k L a values calculated by Eq. (5) were shown by lines in Figures 7 and 8. For all experimental results on k L a in standard bubble columns, the parity plots were shown in Figure 9. By the correlation, Eq. (5), the volumetric oxygen transfer coefficients, k L a, were predicted within an accuracy of ±40%. 0 D C [m] Exp. Cal. 0.03 0.06 0.0 0.0 Fig. 8. Effect of column diameter on k L a 5
0. +40% Experimetal k L a [s - ] 0.00-40% 0.000 0.000 0.00 0. Calculated k L a [s - ] Fig. 9. Parity plots of k L a in standard bubble columns 3. Partitioned bubble column When the diameter of, D C, was reduced, the value of k L a increased as shown in Figure 8. Therefore, the bubble column was improved so as to have a decreased diameter in large diameter of columns in this study. The partitioned bubble columns, Column B and C, were developed by vertically dividing the standard bubble columns, Columns A, so that each divided cross section was made the same. Feed gas was also separately dispersed into each section and homogeneous slug bubbles were generated just at the gas distributor. The slug bubbles were smoothly rising along the vertical partitioned paths with neither breakage nor passing through holes of the plates. Moreover, the slug bubbles rose with scratching the wall and plates because of very narrow paths compared with the bubble size. On the other hand, the liquid was often exchanged over the partitioning plates with holes due to squeezing by slug bubbles. Therefore, the stagnant zone was minimized in these bubble columns. Figure 0 shows the comparison of escapable gas holdup between standard bubble columns and partitioned bubble columns. When the column diameter D C was 0.0 m, the quarter partitioned bubble column (Column B) had slightly larger G than the standard bubble column (). For 0.0 m ID bubble columns, the /6 partitioned bubble column (Column C) had larger G than the standard bubble column (). The equivalent diameter of those partitioned bubble columns D C was 0.044 m so that G was calculated with D C instead of the column diameter D C in Eq. (). The estimated G in the partitioned bubble columns drawn as broken lines in Figure 0 agreed fairly with experimental values. Fig. 0. Comparison of G between standard bubble columns and partitioned bubble columns Figure shows the comparison of k L a between standard bubble columns and partitioned bubble columns. By inserting the partitioning plates, the stagnant zone was reduced and then the liquid-mixing was improved so that the partitioned bubble columns, which are Column B for 0.0 m ID column and Column C for 0.0 m ID column, presented larger k L a than the standard bubble columns (). The observed k L a in the partitioned bubble columns were a little bit larger than the prediction as shown in broken lines in Figure. Although the fluid dynamics in the most viscous xanthan solution having the highest yield stress, X-8, was very complicated even in a standard bubble column, the increment of k L a must be caused by liquid mixing derived from the liquid exchange through a number of holes on the perforated partitioning plates. Fig.. Comparison of k L a between standard bubble columns and partitioned bubble columns Figure shows the relation between k L a and equivalent diameter of bubble columns D C considering D C = D C for. In xanthan aqueous solutions, k L a increased with decreasing D C '. Therefore, when the large capacity is required as xanthan bioreactors, the partitioned bubble 6
column is available because it can keep high oxygen transfer rate by dividing the column cross section into small sections. k La [s - ] 4. Conclusions To design the more efficiently aerobic bioreactors for xanthan production, the escapable gas holdup, stagnant gas holdup and volumetric oxygen transfer coefficient were investigated for a wide range of xanthan concentration in solutions having a yield stress in standard bubble columns. Due to the yield stress, the small bubbles could not be readily disengaged from liquid phase and were apparently stagnant. The gas holdup of escapable bubbles increased with increasing superficial gas velocity and decreasing equivalent column diameter. By using Nicklin s equation modified by Terasaka et al (003b)., the gas holdups of escapable bubble were estimated well. The liquid phase volumetric oxygen transfer coefficient based on aerated liquid phase was investigated for the same range of xanthan concentrations as above so that the correlation equation was proposed as a function of escapable gas holdup, yield stress and diffusivity. The values of k L a in xanthan aqueous, which were estimated fairly with the proposed equation, increased with increasing superficial gas velocity and decreasing equivalent column diameter at a fixed superficial gas velocity. To design the large capacity bioreactors, the novel partitioned bubble columns were proposed in this study. The partitioned bubble columns presented the larger oxygen transfer rates than the standard bubble columns for identical column diameter. Acknowledgements This research was supported by Japan Society for the Promotion of Science under a Grant-in-Aid for Scientific Research (C) in 004 and the General 7 Journal of Chemical Engineering of Japan, Advance Publication U G =0.03 m/s Liquid: X-8 Key Column A B C 0.00 0. Equivalent column diameter D C '[m] Fig.. Effect of equivalent diameter of bubble columns D C on k L a Research Grant of Nestlé Science Promotion Committee in 00. Nomenclature c L = dissolved oxygen concentration [mol m -3 ] D C = column diameter [m] D C = equivalent column diameter [m] D L = liquid phase diffusivity [m s - ] Fr = Froude number [ ] g = gravitational acceleration [m s - ] H = Henry s constant [Pa m 3 mol - ] h = static liquid height from gas distributor [m] k L a = volumetric mass transfer coefficient [s - ] R = gas constant [J mol - K] R C = column radius [m] r 0 = yield point in bubble column [m] T = temperature [K] t = aeration time [s] U G = superficial gas velocity [m s - ] U L = superficial liquid velocity [m s - ] = shear rate [s - ] ε G = escapable gas holdup [-] ε GVS = stagnant gas holdup [-] ε L = liquid holdup [-] η = H-B model coefficient [Pa s ν ] ρ = liquid density [kg m -3 ] ν = H-B model index [-] σ = interfacial tension [N m - ] τ = shear stress [Pa] τ 0 = yield stress [Pa] Literature cited Eickenbusch, H., P.-O. Brunn, A. Schumpe, Mass Transfer into Viscous Pseudoplastic Liquid in Large-diameter Bubble Columns, Chem. Eng. Processing, 34, 479-485 (995) Herschel, V. W. H. and R. Bulkley, Konsistenzmessungen von Gummi- Benzollösungen, Kolloid-Zeit., 39, 9-300 (96) Iwahashi, M., H. Fukuda and K. Terasaka, Xanthan gum production using a bubble column bioreactor and stirred tank bioreactor, Proc. 69th Annual Meeting of SCEJ., B309 (004) Nicklin, D. J., J. O. Wilkes and J. F. Davidson, Two-phase flow in vertical tube, Trans. Inst. Chem. Eng., 40, 6-68 (96) Schumpe, A. and W. -D. Deckwer, Viscous media in tower bioreactors: Hydrodynamic characteristics and mass transfer properties, Bioprocess Eng.,, 79-94 (987)
Suh, I.-S., A. Schumpe and W. -D.Deckwer, Gas-liquid mass transfer in the bubble column with viscoelastic liquid, Can. J. Chem. Eng., 69, 506-5 (99) Suh, I.-S., A. Schumpe and W. -D. Deckwer, Xanthan production in bubble column and air-lift reactors, Biotech. Bioeng., 39, 85-94 (99) Terasaka, K, Research and development of bubble column bioreactors for production of highly concentrated xanthan gum, Nestlé Science Prom. Committee Annual Rep., 7-84 (003) Terasaka, K., D. Hullmann and A. Schumpe, Mass transfer in bubble columns studied with an oxygen optode, Chem. Eng. Sci., 53, 338-384 (998) Terasaka, K. and H. Shibata, Oxygen transfer in viscous non-newtonian liquids having yield stress in bubble columns, Chem. Eng. Sci., 58, 533-5337 (003a) Terasaka, K. and H. Tsuge, Gas holdup for slug bubble flow of viscous liquids having yield stress in bubble columns, Chem. Eng. Sci., 58, 53-57 (003b) Table. Physical properties of xanthan solutions used (93 K) No. [kg/m 3 ] [mn/m] 0 [Pa] [Pa s ] [-] D L [m /s] H [Pa m 3 /mol] X- 007.9 58.4.9 5.5 0.49 4.80 0-0 7.50 0 4 X- 00. 50.4 3.3 6. 0.58 3.90 0-0 7.90 0 4 X-3 999.3 59.5 3.55.8 0.98 3.87 0-0 7.50 0 4 X-4 00.5 6.6 3.59 0.0 0.68 3.47 0-0 8.76 0 4 X-5 000.5 56.8 3.83 9.8 0.9 3.85 0-0 7.80 0 4 X-6 008.3 5.6 4.58 6.4 0.0908 3.70 0-0 6.8 0 4 X-7 993.6 56.6 4.6 7.6 0.76.90 0-0 8.66 0 4 X-8 004.0 49.8 5.85 3. 0.83 3.00 0-0 8.53 0 4 8