EFFECTS OF SURFACE-ACTIVE SUBSTANCES ON GAS HOLDUP AND GAS-LIQUID MASS TRANSFER IN BUBBLE COLUMN

Transcription

1 EFFECTS OF SURFACE-ACTIVE SUBSTANCES ON GAS HOLDUP AND GAS-LIQUID MASS TRANSFER IN BUBBLE COLUMN Kozo KOIDE, Seiji YAMAZOEand Shigehiko HARADA Department of Chemical Engineering, Shizuoka University, Hamamatsu 432 Key Words: Mass Transfer, Gas Holdup, Bubble Column, Surfactant, Antifoam Agent The effects of surfactants and antifoam agent on the gas holdup s, the liquid-phase mass transfer coefficient kl and the volumetric liquid-phase mass transfer coefficient kla in a bubble column were studied experimentally. An addition of surfactants such as //-alcohols to water increases s and decreases kl, and an addition of antifoam agent to water decreases e, kl and kla. However, the degree of reduction of kl value by an addition of surfactant to water is lower for bubble swarms in a bubble column than that for a single bubble in stagnant liquid. Based on these observations, the previous model for estimating kl of a single bubble in aqueous solutions of surface-active substance is modified so as to be applicable to the estimation of kl for bubble swarms in a bubble column. Introduction It is knownthat surface-active substances in liquids retard mass transfer from a single bubble.4'10'12~15) Raymond et al.10) and Koide et al.4) concluded that the decrease in the liquid-phase mass transfer coefficient kl on the addition of surfactants is primarily due to changes in hydrodynamic characteristics of the system. That is, surfactants adsorbed at the gas-liquid interface retard surface flow by the surface tension gradient at the interface and hence decrease kl value of a single bubble. Along this line of thought, Koide et al.4) proposed a model to estimate kl of a single bubble in aqueous solutions of surfactants. Recently, bubble columns are receiving attention as bioreactors. Liquids in bioreactors contain surfactants, antifoam agents or both. Yagi et al.15) studied experimentally the effects of surfactants, antifoam agents and sterilized cells on the gas holdup s and the volumetric liquid-phase mass transfer coefficient kla in a bubble column, and showed that both and kla were decreased by the addition of surfactants or antifoam agents to water. On the other hand, Oels et al.8) and Kelkar et al.3) showed that e was increased by the addition of normal alcohols to water and that the degree of increase in e was higher in aqueous solutions of alcohols with longer carbon chain length. Further, Oels et al.8) showed that the gas-liquid inter facial area was increased remarkably by the addition of alcohols to water. However,the effects of surfactants and antifoam agent on kl in a bubble column have not been sufficiently studied, and only a few data of kl in Received September 8, Correspondence concerning this article should be addressed to K. Koide. Dept. of Chem. Eng., Tokyo Institute of Technology, Tokyo 152. VOL 18 NO % methanol solution have been obtained by Oels et The purpose of this work was to study experimentally the effects of surfactants and antifoam agent on 8, kl and kla in a bubble column. 1. Experimental The experimental apparatus used in this work is shown in Fig. 1. Dimensions of the plexiglass column with a rectangular cross section were 0.2m in width, 0.15m in length and 1.5m in height. Two kinds of perforated plates with holes oriented in triangular pitches were used as gas distributors: one with 5= 0.5mm and n=66 and the other with (5=lmm and /2=32. The gas velocity UG of air was adjusted to cm-s"1. The liquids used in this work were distilled water and aqueous solutions of n-hexanol, rc-heptanol, n- octanol and an antifoam agent (Nissan Disfoam CC- 1 18). The clear-liquid height HL was 1 m. During each run, liquid was neither fed nor discharged, and liquid temperature was kept at 298.2±0.5K. Table 1 shows the properties of liquid at operating temperature. The average gas holdup e in the column was evaluated by using experimental data of clear-liquid height and aerated-liquid height. The axial distribution of gas holdup was obtained from the axial gradient of static pressure. Physical absorption of oxygen in the air by the liquid was employed to determine kla. The details of this method are similar to those in the previous 5) paper. Photographs of bubbles in the column were taken in three regions (the region j, 7= 1-3) whose levels from the gas distributor were respectively m, 287

2 The liquid-phase mass transfer coefficient kl was evaluated by Eq. (6). kl = kla/a (6) Fig. 1. Experimental apparatus. Table 1. Properties ofliquids at 298.2K B <7x lo3 c' q [ppm] >Bx [N-m-1] l010[m^s"1] Water Aq. soln of «-hexanol * Aq. soln of w-heptanol * Aq. soln of Tz-octanol * Aq. soln of Nissan ** Disfoam CC pc=997.0 kg-m~3,,uc=8.937x10~4 Pa s, Z)c=2.42x10"9 m2-s-\ (fic/pcdc)=313, (Pd/pc)=1.25x \0~3 and <jidlnc)= 2.05 x 10~4 for water and aqueous solutions. * Defay et al2) ** Polson.8) m and m bubbles were sampled for each region, and the volume mean diameter dvj and the volume-surface mean diameter dvsj in each region were respectively evaluated by the following equations. dvj={6(sv^inj*\113> V=1~Nj) (1) dvsj=6(lv^l(l,s^> (*=1"nj) (2) Further, the average values of dvj and dvsj in the column were respectively evaluated by the following equations. ^=[(l^)/{l(v^3,)}]1/3> 0=1-3) (3) *v*=(l^)/ Z(V^)}' (7=1-3) (4) Specific gas-liquid inter facial area a was obtained from Eq. (5). 288 a = 6s/dvs (5) 2. Results and Discussion Figure 2 shows that an addition of alcohol to water increases e. This increase might be due to the fact that bubble coalescence is hindered in the aqueous solution of alcohol. A comparison of s in aqueous heptanol solution shows that e decreases with increasing hole diameter 3 of the perforated plate. Meanwhile, an addition of antifoam agent to water decreases since bubble coalescence occurs easily. Values of s in aqueous butanol solution observed by Kelkar et al.3) are also shown in Fig. 2 for comparison. The equation of Marrucci7) for s in the homogeneous flow regime and that of Akita et al.x) for s in the heterogeneous flow of an air-water twophase system are shown in Fig. 2. Two curves based on these equations are nearly applicable to other aqueous solutions used in this work. The region between the two curves corresponds approximately to that where UGvs. e curves for the transition regime lie. The observed values for e shown in Fig. 2 indicate that most of the experiments in this work were carried out in the transition regime. Figure 3 shows the effect of c 'B, the concentration of surface-active substance, on e/e0, the ratio of s in aqueous solution to that in water. s/s0 seems to increase at lower concentration in an aqueous solution of alcohol with longer carbon chain length. dvj and dvsj were almost constant in the axial direction since UGin this work was small, and bubble coalescence was not so frequent. Sj was almost constant in the axial direction in the range of UG<\ cm- s"1 but increased slightly with increasing height from the gas distributor in the range of UG^l cm-s"1. The shapes of babbles were spherical or oblate-spheroidal, and the values for the ratios of longer axis to shorter axis of bubbles were Figure 4 shows that dvs in aqueous alcohol solution is smaller than that in water and increases with increasing <5, and that the effect of UG on dvs is small in the experimental range of UGin this work. In the aqueous solution of the antifoam agent, bubble sizes were less uniform than those in water and other aqueous solutions and there existed many fine bubbles of dvs~1mmor smaller and mediumsize bubbles of 4-5mmor larger. Therefore, dvs in this solution was small comparedto that in water. Figure 5 shows that specific gas-liquid inter facial area a increases with increasing UG,and the values of a in aqueous alcohol solutions are larger than those in water for the same gas distributor. Figure 6 shows that kla increases with increasing JOURNAL OF CHEMICAL ENGINEERING OF JAPAN

3 Fig. 4. Effects of UGand additives on bubble size. Fig. 2. Fig. 5. Effects of UG and additives on specific gas-liquid inter facial area. Fig. 3. Effects of additives and their concentrations on s/e0. UGwhile kla in aqueous solution of alcohol decreases with increasing carbon chain length of the alcohol. Figure 7 shows that at UG=2cmà"s~1the values ofkla in aqueous solutions of alcohol and antifoam agent are lower than those in water except for kla in aqueous hexanol solution of cb below 100 ppm. Surfactants, such as alcohols, adsorbed at the gasliquid interface retard surface flow by the surface tension gradient at the interface and hence reduce the value of &L,4) whereas they hinder bubble coalescence^ and then increase e and a. In aqueous hexanol solution, where the curve of kla/(kla)0 vs. cb has a maximumin Fig. 7, the effect ofhexanol on increasing VOL. 18 NO Fig. 6. Effects of UG and additives on kla. s surpasses that on decreasing kl in the range of cb below 100 ppm, and vice versa in aqueous solutions of other alcohols. Therefore, kla/(kla)0 values for hexanol aqueous solution of cb below 100 ppm are 289

4 Fig. 7. Effects of additives and their concentrations on kla/ (kla)0. Fig. 8. Effects of bubble size and additive concentration on kl. larger than unity and those for other alcohol aqueous solutions are smaller than unity. In the range of UG below l cm-s"1, kla/(kla)0 for the alcohol aqueous solutions are smaller than unity, as can be seen from Fig. 7. This might be caused by the fact that bubble coalescence in water is not so frequent in the range of UGbelow lcm-s"1 and so the effect of alcohol on increasing s doesn't contribute enough to increase kla/(kla)0. In aqueous solution of the antifoam agent, both s/e0 and kla/(kla)0, are smaller than unity, since the agent accelerates bubble coalescence and reduces kl value. Figures 3 and 7, respectively, show that both s/s0 and kla/(kla)0 are smaller than unity in aqueous solutions of Tween 85 (as a surfactant) and Antifoam AF emulsion (Dow Corning, as an antifoam agent).15) Though Yagi et al.15) defined Tween 85 as a surfactant, it accelerated bubble coalescence like an antifoam agent. Figure 8 shows that kl increases slightly with dv, and the increase of kl with dv becomes smaller in aqueous solution of alcohols with longer carbon chain length. The values of kl estimated by the equation of Shirotsuka et al}l) for bubble and solid sphere are also shown in Fig. 8. kl values for aqueous alcohol solutions approach those for a solid sphere as carbon chain length increases. kl values observed by Oels et al.8) for 1 % methanol solution in the column with S=0.5mm are also shown in Fig. 8. Figure 9 shows that the decrease in kl with increasing alcohol concentration is more drastic for aqueous solutions of alcohols with longer carbon chain length. These phenomena are similar to those of single bubbles observed by Koide et al.4) and Raymond et al.10) Koide et al^ assumed the retarding effect on surface flow by surface-active substances adsorbed on 290 Fig. 9. Effects of additives and their concentrations on kl. the gas-liquid interface to be expressed by the apparent increase in the viscosity of the dispersed phase, Va=Vd+Vb (7) where \ia is the apparent viscosity and \ih the retardation coefficient. Further, they assumed the flow around a single bubble in the aqueous surfactant solution to be similar to the flow of pure liquid around a dispersed phase with the viscosity of \ia. fia was determined so that the value of kl calculated by the equation of Shirotsuka et al.11] coincided with that of kl experimentally obtained, where k in the equation of Shirotsuka et al. was defined in the model JOURNAL OF CHEMICAL ENGINEERING OF JAPAN

5 Fig. 10. Empirical correlation of fib/fic vs. (knm/u/ic)/yjre. ofkoide et al. by Eq. (8) instead of K=fijfic: K = Val^c (8) Using the \ia value thus determined, they calculated \xh by Eq. (7), and correlated \ih with the flow condition and the properties of aqueous surfactant solutions by Eq. (9). ^=173 p^\ure\ (9) In this work, the same model as stated above was assumed to apply to the bubble swarms in aqueous solutions of surface-active substances, and \ih was calculated by Eq. (7). In the calculation, the bubblerising velocity u was assumed to be equal to the terminal velocity ut ofa single bubble in water.*1 \ih\\ic in this work was correlated empirically with (knjuii^/yjre in Fig. 10. The empirical equation for \xh\\xc was obtained by the least squares method with data shown in Fig. 10: Vc IVWeJlV J in the experimental ranges of 667^iterg1354 and ^(knju/ic)/yfre^ i6a. The average error for estimating \xh by Eq. (10) was within 44% for 63 data. As shown in Fig. 10, \xh in this work is smaller than that for a single bubble in stagnant liquid. The liquid where bubble swarms rise is more turbulent than that where only a single bubble rises, and this might reduce the retarding effect of the surface-active substance on surface flow. Figure 10 also shows that the values of *x ut of single bubbles in aqueous solutions was not measured. As ut values of CO2bubble in alcohol aqueous solutions were at most 7% lower than those in water,4) ut values observed in contaminated water by Kubota et al.6) were used. VOL 18 NO Fig. ll. Comparison of kl calculated from Eqs. (7), (8), (10) and the equations of Shirotsuka et al.ll} with values obtained in this work. \xh\[ic calculated with kl data of Oels et al.8) shown in Fig. 8 roughly agree with Eq. (10). (kl)cal, the mass transfer coefficient of bubble swarms calculated from Eqs. (7), (8), (10) and the equations of Shirotsuka et al.,n) was compared with kl experimentally obtained in Fig. ll. The average error of estimating kl was within 12% for 63 data. kl values observed by Oels et al.8) for a perforated plate of S=0.5mm are also compared with those calculated in Fig. ll. The kl value observed at UG= 2.5cm-s"1 is much larger than the value calculated. This might be due to underestimation of the gasliquid inter facial area which was used to evaluate the kl value from kla data. IfEq. (9) is used instead ofeq. (10), kl is underestimated as shown in Fig

6 dvs = volume-surface mean diameter of bubbles [m] HL = clear-liquid height [m] k = - (da/dcb)(3url2db)1/2(rrt) [-] kl = liquid-phase mass transfer coefficient [m - s"1] kla = volumetic liquid-phase coefficient mass transfer [s ~ x] Nj = number of bubbles in regionj [-] n = number of holes in perforated plate [-] R = gas constant [J-K^-moF1] Re = dvupc\\ic, Reynolds number [-] r = dv/2, bubble radius [m] Sb T = surface area = temperature of bubble [m2] [K] UG = superficial gas velocity [m - s"1] u = bubble rising velocity [m-s"1] ut = terminal velocity of single bubble [m-s"1] Vb = bubble volume [m3] Fig. 12. Comparison ofkl calculated from Eqs. (7), (8), (9) and the equations of Shirotsuka et al.11} with values obtained in this work. Conclusions The addition of surfactants such as ^-alcohols to water increases the gas holdup e but decreases the liquid-phase mass transfer coefficient kl. Therefore, the properties and concentration of the surfactant determine whether the volumetric liquid-phase mass transfer coefficient kla increases or decreases by the addition of the surfactant to water. The addition of antifoam agent reduces a, kl and kla. The effect of surface-active substances such as surfactants and antifoam agent on reducing kl value is smaller for bubble swarms than for a single bubble in stagnant liquid. The previous model for estimating kl of a single bubble in aqueous solutions of a surface-active substance can be used to estimate kl of bubble swarms, if the empirical equation of the retardation coefficient which indicates the degree of retardation of surface flow at gas-liquid interfaces of bubble swarms is used. Acknowledgment The authors are very grateful to Nihon Yushi Co., Ltd. for offering antifoam agents and data regarding their properties. Nomenclature a = specific gas-liquid inter facial area based on aerated liquid [m"1] cb = concentration of surface-active substance [mol-m"3] c 'B = concentration of surface-active substance [ppm] DB = diffusivity of surface-active substance in liquid [m2 ^ " 1] Dc = diffusivity of dissolved oxygen in liquid K-s-1] dv = volume mean diameter of bubbles [m] 292 S = hole diameter of perforated plate [m] s = gas holdup [-] k = vjvc [-] \ia = ll(1+hi apparent viscosity of dispersed phase [Pa-s] /ib = retardation coefficient [Pa à"s] \ic = viscosity of continuous phase [Pa à" s] \xd = viscosity of dispersed phase [Pa à" s] pc = density of continuous phase [kg-m~3] pd = density of dispersed phase [kg-m~3] 77^ = - cb(d<j/dcb), surface pressure [N -m"1] o = surface tension of liquid [N-rrT1] (Subscripts) cal = calculated value ex = experimental value / = i-th bubble j = in regionj 0 = in water Literature Cited 1) Akita, K. and F. Yoshida: Ind. Eng. Chem., Process Des. Develop., 16, 76 (1973). 2) Defay, R. andj. R. Hommelen: /. ColloidSci., 14, 411 (1959). 3) Kelkar, B. G., S. P. Godbole, M. F. Honath, Y. T. Shah, N. L. Carr and W.-D. Deckwer: AIChE /., 29, 361 (1983). 4) Koide, K., T. Hayashi, K. Sumino and S. Iwamoto: Chem. Eng. Sci., 31, 963 (1976). 5) Koide, K., 407 (1983). H. Sato arid S. Iwamoto: /. Chem. Eng. Japan, 16, 6) Kubota, M., 1074 (1967). T. Akehata and T. Shirai: Kagaku Kogaku, 31, (7) Marrucci, G.: Ind. Eng. Chem., Fundam., 4, 224 (1965). 8) Oels, U., J. Lucke, R. Buchholz and K. Schiigerl: Ger. Chem. Eng., 1, 115 (1978). 9) Poison, A.: /. Phys. Colloid Chem., 54, 649 (1950). 10) Raymond, D. R. and S. A. Zieminski: AIChE J., 17, 57 (1971). ll) Shirotsuka, T. and A. Hirata: Kagaku Kogaku, 35, 123 (1971). 12) Vogtlander, J. G. and F. W. Meijboom: Chem. Eng. Sci., 29, 799 (1974). 13) Vogtlander, J. G. and F. W. Meijboom: Chem. Eng. Sci., 29, 949 (1974). 14) Weber, M. E.: Chem. Eng. Sci., 30, 1507 (1975). 15) Yagi, H. and F. Yoshida: /. Ferment. Technol., 52, 905 (1974). (Presented in part at the 18th Autumn Meeting of Society of Chemical Engineers, Japan at Fukuoka, October 18, 1984.) JOURNAL OF CHEMICAL ENGINEERING OF JAPAN