University of Groningen Gas Absorption in an Agitated Gas-Liquid

University of Groningen

Gas absorption in an agitated gas-liquid-liquid system Cents, A.H.G.; Brilman, D.W.F.; Versteeg, Geert

Published in: Chemical Engineering Science

DOI: 10.1016/S0009-2509(00)00324-9

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record

Publication date: 2001

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA): Cents, A. H. G., Brilman, D. W. F., & Versteeg, G. F. (2001). Gas absorption in an agitated gas-liquid-liquid system. Chemical Engineering Science, 56(3), 1075-1083. DOI: 10.1016/S0009-2509(00)00324-9

Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

Download date: 11-02-2018 Chemical Engineering Science 56 (2001) 1075}1083

Gas absorption in an agitated gas}liquid}liquid system A. H. G. Cents*, D. W. F. Brilman, G. F. Versteeg Department of Chemical Engineering, University of Twente, P.O. Box 217, 7500 AE Enschede, Netherlands

Abstract

Gas}liquid}liquid systems have gained interest in the past decade and are encountered in several important industrial applications. In these systems an immiscible liquid phase may a!ect the gas absorption rate signi"cantly. This phenomenon, however, is not completely understood and underlying mechanisms need further study. In this work the well-known Danckwerts-plot technique is used to determine the liquid side mass transfer coe$cient k* and the gas}liquid interfacial area a simultaneously in this type of systems. As absorption/reaction system CO absorption in a 0.5 M KCO/0.5 M KHCO bu!er solution catalysed by sodium hypochlorite was chosen. Toluene, n-dodecane, n-heptane and 1-octanol were applied as dispersed liquid phases. The Danckwerts- plot could be well used in gas}liquid}liquid systems and from the results it appeared that two types of systems exist; systems that enhance mass transfer and systems that do not enhance mass transfer. E!ects at low dispersed phase hold-up were observed to be very strong and are thus important, but were not taken into account in further analysis of the e!ect of dispersed phase hold-up on mass transfer. In systems where dodecane and heptane were added to the bu!er solutions no enhancement of mass transfer was observed. However, the addition of toluene and 1-octanol caused an enhancement of mass transfer that could be well described using a homogeneous model of the shuttle mechanism. 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Mass transfer; Enhancement; Gas-liquid-liquid; Danckwerts-plot

1. Introduction (Falbe, 1980). Gas}liquid}liquid systems are also encountered in areas in which an additional inert liquid phase is Gas}liquid}liquid systems have gained interest in the added on purpose to a gas}liquid system to increase the past decade due to the introduction of homogeneous mass transfer rate. This latter concept is e.g. applied in biphasic catalysis in various reaction systems, e.g. hydro- a few biochemical applications (Rols, Condoret, Fonade formylation, carbonylation, hydrogenation and & Goma, 1990). However, the addition of a second liquid oligomerization (Cornils, 1999). The main advantage of phase can also retard the gas}liquid mass transfer these systems over catalysis in one phase is the easy (Yoshida, Yamane & Miyamoto, 1970). separation of the catalyst and the reactants or products. In this latter perspective, especially when considering In this way advantages of homogeneous catalysis (no that the gas absorption rate is frequently the capacity internal mass transfer limitations and often higher selec- determining factor, it is very important to understand the tivities) and of heterogeneous catalysis (easy catalyst phenomena occurring in these systems for design and separation) can be combined. Important industrial operation of these systems. applications are, e.g., the SHOP-process and hydro- The gas absorption rate in stirred multiphase reactors formylation of propene to butyraldehyde. is usually characterised by the volumetric mass transfer Gas}liquid}liquid systems are further encountered in coe$cient k*a. In this coe$cient the liquid side mass reaction systems which inherently consist of three phases transfer coe$cient k* as well as the interfacial area a are due to two (or more) immiscible reactants, reaction prod- discounted. ucts or catalyst. For example, in the Koch reaction sys- To study the underlying phenomena of the e!ect of tem all three reactants originate from di!erent phases adding an immiscible liquid phase on the gas absorption rate for a gas}liquid system, it is considered necessary to separate the e!ects on the interfacial area a from those on the liquid side mass transfer coe$cient k*. In this work * Corresponding author. the Danckwerts-plot technique to measure k* and a si- E-mail address: [email protected] (A. H. G. Cents). multaneously by chemical absorption is used to study the

0009-2509/01/$- see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 9 - 2 5 0 9 ( 0 0 ) 0 0 3 2 4 - 9 1076 A. H. G. Cents et al. / Chemical Engineering Science 56 (2001) 1075}1083 e!ects of the addition of di!erent immiscible organic Bruining, Joosten, Beenackers and Hofman (1986). They phases. In this work four di!erent gas}liquid}liquid sys- tested the model using oxygen absorption in aqueous tems will be studied. solutions using n-hexadecane as a dispersed phase. Brilman, Goldschmidt, Versteeg and van Swaaij (2000) recently developed 3-D heterogeneous mass transfer 2. Previous work models to study this grazing e!ect in more detail. From their overview of all models describing these e!ects, it can One of the "rst encounters in literature on be concluded that, although these heterogeneous models gas}liquid}liquid systems is the work of Yoshida et al. are to be preferred from a physical point of view, the (1970). In their work the oxygen absorption rate was homogeneous models describe the experimental data measured in aqueous dispersions of a.o. kerosene and equally well for simple cases and require less detailed toluene. The observed e!ects on the addition of these input parameters. dispersed phases were very di!erent. The addition of The second mechanism proposed to describe the mass kerosene resulted in a decrease of the volumetric mass transfer enhancement is the so-called coalescence}redis- transfer coe$cient k*a. On the other hand, toluene persion mechanism, which is based on direct contact caused, after a sharp initial decrease, a strong increase in between gas and the dispersed phase by formation of the mass transfer rate at higher hold-ups. gas}liquid complexes (Rols et al., 1990). The gas absorp- Linek and Benes\(1976) studied gas absorption in tion rate is then enhanced due to the introduction of this liquid}liquid systems with a known value for the inter- second transfer path. facial area to determine the in#uence on the liquid side Concluding from experiments and models reported in mass transfer coe$cient k*. They studied oxygen absorp- the literature, it appears that there is a lot of confusion tion in aqueous emulsions with a mixture of n-alkanes about the gas absorption mechanism and e!ects in and with oleic acid. The value of k* remained practically gas}liquid}liquid systems. By studying experimentally constant in the presence of n-alkanes before phase inver- the e!ects on k* and a separately, more insight on the sion, while the addition of oleic acid caused an initial mechanism may be obtained. decrease of k*. When more oleic acid was added, k* increased again. A few studies have been presented in literature on the 3. Experimental technique in#uence of a second immiscible liquid on the gas}liquid interfacial area Mehta and Sharma (1971) used a fast 3.1. Danckwerts plot reaction, CO}NaOH system, to study the in#uence of 2-ethyl-hexanol on the speci"c gas}liquid interfacial area. For a better analysis of the observed e!ects on the According to the authors the area increased due to a pre- mass transfer rate the simultaneous determination of the vention of bubble coalescence. Das, Bandopadhyay, liquid side mass transfer coe$cient k* and the interfacial Parthasarathy and Kumar (1985) studied the interfacial area a has been selected, using the well-known Danck- area in gas}liquid}liquid systems with both a physical werts-plot technique (Danckwerts, 1970). The major and a chemical method. As dispersed phases 2-ethyl- advantage of this chemical absorption method over a hexanol, toluene and methyl-isobutyl-ketone were used. physical absorption technique is that, when the reaction All systems showed initially an increase in the interfacial is fast enough, the bulk concentration of the dissolved gas area with higher dispersed phase fraction, but above is negligible, which implies that the driving force for mass approximately 10% hold-up the area started decreasing. transfer is well known. This allows for accurate deter- The increase was explained by prevention of bubble mination of the mass transfer parameters k* and a during coalescence, while a diminished level of turbulence in the stationary conditions. The determination of the inter- reactor caused the decrease at higher hold-ups, according facial area by physical measurement techniques in gas} to the authors. liquid}liquid systems is also di$cult to perform. The Several mechanisms have been proposed to explain the dispersions are often completely opaque, which implies experimental results, thereby focusing on the gas absorp- that laser- and photographic techniques are di$cult to tion enhancement e!ect. In the literature actually two use in these systems. Also the use of intrusive methods types of mechanisms were suggested. The "rst one is the like a capillary technique to determine bubble sizes is so-called shuttle- or grazing mechanism, analogous to not without severe complications in these systems. Non- the one in gas}liquid}solid three phase systems. The representative sampling, coalescence and break-up dur- mechanism is based on dispersed phase drops entering ing sampling and the disturbance of the #ow pattern can the mass transfer "lm at the gas}liquid interface to en- be disadvantageous in these methods. By using the hance mass transfer due to their higher solubility for the Danckwerts-plot technique a representative tank- gas-phase component to be transferred. A homogeneous averaged absorption rate is determined, circumventing model describing this mechanism was proposed by above-mentioned di$culties. A. H. G. Cents et al. / Chemical Engineering Science 56 (2001) 1075}1083 1077

According to the Danckwerts surface renewal model In this equation c represents the equilibrium concen- !- C for mass transfer the rate of absorption for a reaction tration of carbon dioxide, which depends a.o. on the with negligible bulk concentration can be described as carbonate and bicarbonate concentrations (Danckwerts, 1970). Under the experimental conditions the equilib- R "k (1#Hac a< . (1) rium concentration c is very small and can therefore * * * !- C be neglected. Eq. (6) can now be simpli"ed to: This equation holds when the gas side mass transfer resistance can be neglected. Ha is the Hatta number that r "k c with k "k #k c #k c \. is de"ned for a "rst- or pseudo-"rst-order reaction by !- * !- * A -& (7) k D Ha" * . (2) k Already at very low catalyst concentrations the * catalysed reaction is dominant: kAc k and kAc From these equations it can be seen that by changing the kc-&\. This means that the rate of reaction can be easily value of the apparent rate constant k* a straight line varied by changing the catalyst concentration. can be obtained by plotting (R /c *<*) versus k*D . The kinetics of the uncatalysed reaction (k + In this plot the slope is equal to a and (k*a) equals the 0.02 s\) were determined by Danckwerts and Sharma intercept with the vertical axis. In this way k and a can * (1966). The value of k is strongly dependent on ionic be determined simultaneously. strength and the nature of the cations. k was estimated to be 11.6 m mol\ s\ using the work of Pohorecki 3.2. System and Moniuk (1988). Sharma and Danckwerts (1963) and Benadda, Prost, Ismaily, Bressat and Otterbein (1994) As an absorption/reaction system carbon dioxide measured kinetic constants of the sodiumhypochlorite absorption in a 0.5 M potassium carbonate/0.5 M catalysed reaction in a 0.5 M NaCO/0.5 M NaHCO potassium bicarbonate bu!er solution was used. This bu!er and in a 0.4 M KCO/0.4 M KHCO bu!er, carbonate/bicarbonate bu!er solution is the continuous respectively. Although the value of kA depends on liquid in the experiments done. As dispersed liquid phase ionic strength and the counter ion of the bu!er the values several di!erent organic liquids have been applied; were not very di!erent at 293 K. A value for kA of toluene, n-dodecane, n-heptane and l-octanol. In the 1.8 m mol\ s\ (at 294 K) was used in the mass trans- continuous liquid phase the following reactions occur: fer calculations. Nevertheless, for studying the e!ect of an immiscible # &I # > liquid phase on the mass transfer coe$cient k and CO HO HCO H (3) * I\ a their absolute values are of minor importance, implying that the apparent reaction rate constants are known instantaneously followed by the equilibrium reaction: su$ciently accurate. In this work especially the relative # >& CO H HCO. (4) changes in k* and a will be studied. The solubility of carbon dioxide in the bu!er solution As a catalyst for reaction (3) sodiumhypochlorite is used. is estimated using the work of Weisenberger and This allows for a wide variation of the reaction rate on Schumpe (1996): addition of only small amounts of catalyst, so that the psycho-chemical parameters of the catalyst/bu!er system c log % "(h #h )c . (8) do not change signi"cantly. As catalytic species hypo- c G % G chlorite was chosen over arsenite because of its lower % toxicity. The value of (c% /c%) was determined to be 1.57. The Due to the pH (+10) also hydroxyl ions will be solubility of carbon dioxide and the di!usion coe$cient present in the bu!er solution. Carbon dioxide reacts with in water were determined using correlations proposed by OH\ according to the following reaction: Versteeg and van Swaaij (1988):

# &I 2044 CO OH HCO. (5) He "3.59;10 exp , (9) I\ !- & - ¹ The reaction rates for the reactions (3) and (5) are both !2119 "rst order with respect to CO . Danckwerts (1970) has D "2.35;10 exp . (10) !- & - shown that under certain conditions, which are all ful- ¹ "lled in this work, the reaction is pseudo-"rst order with The di!usion coe$cient of CO in the respect to CO and the rate equation can be de"ned by (KCO/KHCO)bu!er solution can be estimated by r "(k #k c #k c \ )(c !c ). (6) a correlation of Joosten and Danckwerts (1972) and was !- A -& !- !- C 1078 A. H. G. Cents et al. / Chemical Engineering Science 56 (2001) 1075}1083

; \ \ estimated to be 1.5 10 m s : state value of the CO concentration in the outgoing #ow. Depletion of CO in the gas phase was always less D !- " . (11) than 70%, implying that the assumption of an ideally D !- mixed gas phase does not in#uence the results. The gas hold-up was determined using the liquid The viscosity of the bu!er solution was estimated from height before the gassing and an estimation of the height the Handbook of Chemistry and Physics (1994). of the dispersions during the experiments. The Danckwerts-plot technique was initially applied 3.3. Set-up for the CO-bu!er two-phase system to check the validity of the method and operating conditions. Sub- The set-up consists of a gas premix section, a reactor sequently, the same technique was used in the presence of and an analysis section, which is schematically shown in toluene, n-heptane, n-dodecane and 1-octanol. Fig. 1. In the premix section the feed gas is composed All chemicals used in this research were of high purity ' using two mass #ow controllers (one for CO and one for ( 99%). Because sodiumhypochlorite solutions may N). The feed gas is introduced at the bottom of the decompose spontaneously, the hypochlorite concentra- stirred tank using a single ori"ce (1 mm). The unabsorbed tion was determined before every experiment by standard carbon dioxide and nitrogen leave the reactor at the top. iodometric titration. Condensable liquid components from the reactor are condensed in a cold trap. This cold trap consists of an ice bath at 273 K and a cryostat at 258 K. The CO-concen- 4. Results and discussion tration is then analysed by an infrared analyser and recorded versus time by a personal computer. 4.1. Danckwerts plots The reactor is jacketed and has a working volume of 2.5 l and an inner diameter of 149 mm. The dispersions Both for the CO}KCO/KHCO two-phase system inside the reactor are agitated by a 6-bladed Rushton- as well as for the systems in the presence of a dispersed turbine. The diameter of the turbine is 1/3 of the inner organic liquid phase the Danckwerts plots showed excel- diameter of the tank and is placed 50 mm above the lent linearity, implying that this method can be used for bottom of the reactor. Four ba%es with a width of 15 mm determining the mass transfer coe$cient and the inter- ensure adequate mixing within the reactor. The pressure facial area simultaneously. In Fig. 2 the Danckwerts plot control in the reactor consists of a 1.5 m water column, for the CO}KCO/KHCO two-phase system and an which also serves as a sparge for the gas that is not led example of a plot for a system with a dispersed organic through the CO-analyser. phase (20% dodecane) are presented. Experiments were carried out at a constant temperature of 294 K and a pressure of approximately 1.15 bars. 4.2. Mass transfer parameters Power input and total liquid volume were kept constant during all the experiments. The stirrer rate was approx- The Danckwerts plots have been measured for each imately 1100 rpm in all the experiments. The di!erent dispersed phase at di!erent hold-ups on the organic amounts of bu!er solution and organic liquid phase were liquids. In Fig. 3 the results are presented for the used batchwise in the reactor. volumetric mass transfer coe$cient, k*a, for every dis- The inlet #ow consisted of 1.70% carbon dioxide, persed phase used as measured by the Danckwerts-plot which could be accurately determined (within 0.3%) by technique. leading the gas directly to the CO-analyser. Gas absorption rate could be calculated by measuring the steady-

Fig. 1. Experimental set-up: 1: gas cylinders, 2: mass #ow controllers, Fig. 2. Danckwerts-plot without a dispersed phase and in the presence 3: reactor, 4: cold trap, 5: CO-analyser. of 20% n-dodecane. A. H. G. Cents et al. / Chemical Engineering Science 56 (2001) 1075}1083 1079

Fig. 3. Absolute k*a values for the four dispersed phases used. Fig. 4. Relative k*a values for the four dispersed phases used.

The initial decrease in k*a at low dispersed organic phase hold-ups is striking, as can be seen in Fig. 3. These initial e!ects are very strong in the case of toluene and 1-octanol and less pronounced with dodecane and heptane. From the results of the Danckwerts plots it was possible to determine both k* and a. It was found that the initial decrease in k*a on the addition of toluene was completely due to a decrease in the interfacial area (a part of this decrease could be contributed to a decrease in the gas hold-up). The initial decrease in k*a with toluene as a dispersed phase was also observed by Yoshida et al. (1970) in their work on oxygen absorption in aqueous dispersions. Fig. 5. Relative k* values for the four dispersed phases used. However, the initial decrease in k*a with the addition of 1-octanol was completely due to a decrease in the value of the mass transfer coe$cient k*, and are probably retards mass transfer. This e!ect was also observed by due to changes in the surface chemistry of the system Yoshida et al. (1970, see Section 2). In their work the (surface tension and the formation of semi-rigid bound- addition of kerosine, which is in fact comparable to ary layers). Analogous results were also observed by dodecane, retarded mass transfer and the addition of Eckenfelder and Barnhart (1961) in their work on the toluene-enhanced mass transfer (after a sharp initial in#uence of surfactants (to which 1-octanol can also be decrease). accounted for) on mass transfer. The e!ects on the relative values of the mass transfer As the scope of this work is to study the in#uence of coe$cient k* as determined from the Danckwerts-plots adding a dispersed phase on the mass transfer rate, the are presented in Fig. 5, showing that the relative value of initial e!ects are important. However, to be able to com- k* as calculated from the Danckwerts-plots increased on pare the results for the four dispersed phases used at the addition of toluene and 1-octanol. When dodecane higher organic phase hold-ups the initial e!ects should and heptane were added to the bu!er solutions the not be taken into account. In Fig. 4 the relative values for liquid side mass transfer coe$cient remained practically k*a are presented for the organic phases used. The rela- constant. tive value of k*a is de"ned by the ratio of the absolute In Fig. 6 the relative values of the interfacial area are value of k*a at higher hold-ups to the value obtained plotted versus the fraction dispersed organic phase in the from extrapolation to zero hold-up. reactor. The value decreased with the addition of 1-oc- From Fig. 4 it can be seen that the relative values of tanol and toluene. The relative value of the gas}liquid k*a increase on addition of toluene. The addition of interfacial area remained approximately constant up to 1-octanol caused a slight increase up to 20% organic 20% dodecane. Beyond this value, the area started also phase and at 40% 1-octanol the increment was some- decreasing. what lower. The addition of dodecane (and heptane) caused a decrease in the mass transfer rate. Despite the 4.3. Interpretation fact that all organic dispersed phases used have a higher physical solubility for carbon dioxide when compared to When a dispersed phase is present in the reactor it is the aqueous bu!er solution, two types of systems appear very important that the values for the mass transfer to exist; one that enhances mass transfer and one that parameters which are obtained from the Danckwerts 1080 A. H. G. Cents et al. / Chemical Engineering Science 56 (2001) 1075}1083

by Eq. (12), assuming a "rst-order reaction in the continuous phase: *c (x, t) D "k (1! )c (x, t) *x * " *c (x, t) #(1# (m !1)) (12) " 0 *t with boundary conditions " ' " t 0, x 0: c (x, t) c , (13) t'0, x"0: c (x, t)"c , (14) Fig. 6. Relative values of the interfacial area for the four dispersed * ' "R " phases used. t 0, x : c (x, t) c . (15) When the capacity of the liquid bulk is su$cient, the concentration of the gas in the bulk can be assumed zero; " plots are analysed correctly. In the absence of a second c 0. Equation (12) can now be solved analytically liquid phase the rate of absorption is described by Eq. (1), and the #ux through the gas}liquid interface, according as is con"rmed by the linearity of the Danckwerts plot to the Danckwerts surface renewal model, is given as (Fig. 2). Droplets with a higher solubility for the gas " ( # ! # ! < phase component to be transferred, however, may change R k* 1 "(m0 1) (1 ")Ha c *a *. (16) the concentration pro"le within the mass transfer zone at The Hatta number is de"ned in Eq. (2). Now it is possible the gas}liquid interface (if the droplets are actually pres- to de"ne an enhancement factor for the presence of ent in that zone). To describe the rate of absorption in a dispersed phase and a "rst-order reaction in the con- this latter situation a homogenous model of the shuttle tinuous phase according to a homogeneous model of the mechanism (Section 2) was used. This model was chosen shuttle-mechanism. The enhancement factor E" in Eq. to determine if the presence of immiscible dispersed (17) is the ratio of the #ux in Eq. (16) and the #ux in case phase droplets within the mass transfer zone could ex- " there is no dispersed phase present ( " 0): plain the observed e!ects on the gas}liquid absorption R ( O0) 1# (m !1)#(1! )Ha rates measured and on the corresponding mass transfer E " " " " 0 " . " " # parameters k* and a as determined by the Danckwerts- R ( " 0) 1 Ha plot technique. (17) When using a homogeneous model of the shuttle mechanism for the description of the absorption process, As is shown in Eq. (16) there is a di!erence in the implicitly a few assumptions are made: interpretation of the Danckwerts plot when there are no dispersed phase drops present in the mass transfer region 1. The dispersed phase is equally divided throughout the (case I) and when there are (case II). When the results of gas}liquid mass transfer zone. the gas absorption measurements are plotted as ex- 2. The concentration of carbon dioxide in the dispersed plained in Section 3.1, (R /c < ) versus k D , the phase is at any time and place within the mass transfer * * * intercept with the vertical axis and the slope of the zone at equilibrium with its concentration in the sur- Danckwerts plot can be interpreted as shown in Table 1. rounding continuous phase. Analysis of the mass transfer parameters for the di!er- 3. The presence of the microphase droplets does not ent organic liquid-phase fractions in the reactor showed in#uence the surface renewal frequency of the continu- two di!erent e!ects for the dispersions used in this re- ous liquid phase. search. The addition of toluene and 1-octanol caused an The second assumption implies that liquid}liquid mass increase in the value of the mass transfer coe$cient when transfer resistance is neglected, which may not necessar- case I is applied for analysis of the Danckwerts plot, like ily be true. The liquid}liquid mass transfer coe$cient is done in Fig. 5. However, this k* value remained prac- depends a.o. upon the di!usion coe$cients and the diameter of the droplets. Detailed information on these Table 1 parameters was not available for the system under Interpretation of the Danckwerts plot consideration. According to the Danckwerts surface renewal model Case I Case II the concentration pro"le of a gas-phase component A, ! that is being absorbed in a liquid containing microphase Slope a a (1 ") Intercept (k a) (k a).(1# (m !1)) droplets within the mass transfer zone, can be described * * " 0 A. H. G. Cents et al. / Chemical Engineering Science 56 (2001) 1075}1083 1081

Fig. 8. k versus dodecane hold-up when both case I and case II are Fig. 7. k* versus toluene hold-up when both case I and case II are * assumed. assumed. tically constant in case II, when assuming organic drops to be present within the mass transfer zone enhancing absorption rate. The value of k* was expected to remain practically constant, because this value mainly depends on power dissipation and hydrodynamic conditions, which could be assumed constant in all experiments. In Fig. 7 the dependence of k* on the toluene fraction is presented when cases I and II are assumed. From Fig. 7 it appears that the Danckwerts plot in the presence of toluene and 1-octanol is probably best interpreted using case II. Mass transfer in these systems is then enhanced by the addition of a dispersed phase and Fig. 9. k*/k* versus dispersed-phase hold-up when case II is assumed. the surface renewal frequency is constant and thus not in#uenced by the dispersed phase droplets. However, with dodecane and heptane as dispersed phases a completely di!erent e!ect was observed. When case I was chosen for the interpretation of the Danck- werts plot the value of k* remained at a constant level, while under case II the value decreased with increasing fraction of dispersed phase. In Fig. 8 an example of this e!ect is shown with dodecane as the dispersed phase. From Fig. 8, and the fact that mass transfer is not enhanced by dodecane or heptane (see Section 4.2), it can be concluded that for these systems no dispersed phase is likely to be present in the mass transfer region and case I is favored to interpret the mass transfer parameters. Fig. 10. a/a versus dispersed phase hold-up when case II is assumed. The mass transfer parameters k* and a as determined from the Danckwerts plots when case II is applied are presented in Figs. 9 and 10. In Fig. 11 the data of Fig. 5 are re-plotted together with The results indicate that probably two types exist the model predictions for the shuttle mechanism. When within the four of gas}liquid}liquid systems studied. In using a Case I Danckwerts' plot interpretation for a situ- this work an enhancement of mass transfer was observed ation in which droplets are present within the mass ' (when initial e!ects are discarded) with increasing transfer zone, the apparent k* coe$cient is described by amounts of toluene and 1-octanol added to the bu!er (see Table 1) solutions. The addition of dodecane and heptane, however, caused a decrease of mass transfer. k 1# (m !1) * " " 0 . (18) For the addition of toluene and 1-octanol the e!ects on ! k* 1 " k* can be explained best when assuming dispersed droplets to be present in the mass transfer zone (case II, see The value for the interfacial area, after accounting for '' Fig. 9). The value of k* /k* now remains practically initial e!ects, remains constant up to at least 20% or- constant as was expected for constant power dissipation. ganic phase hold-up (Fig. 10). 1082 A. H. G. Cents et al. / Chemical Engineering Science 56 (2001) 1075}1083

enhanced due to the addition of toluene and 1-octanol to aqueous bu!er solutions, which could be well described by a homogeneous model of the shuttle mechanism. Addition of n-dodecane and n-heptane, however, causes no enhancement of mass transfer. Initial e!ects (at low dispersed phase fraction) can be very strong and further study on these e!ects is required. To validate the conclusions drawn from the experimental results in this work for the e!ect on the mass transfer parameters k* and a, independent physical measuring techniques will be developed.

Fig. 11. Relative values for k* when case I is applied. The model lines represent the enhancement factor as determined from the model. Notation

a gas}liquid interfacial area, m/m liquid \ c concentration of component A, mol m c * concentration of component A at the gas}liquid interface, mol m\ \ D di!usion coe$cient of component A, m s E" enhancement factor due to the dispersed phase, dimensionless Ha Hatta number, de"ned in Eq. (2), dimensionless He Henry's law constant, mol m\ Pa\ \ hG, h% parameters in Eq. (8), m mol \ k* apparent "rst-order rate constant, s Fig. 12. Interpreted k* values. k\, k reverse and forward reaction rate constant of reaction (3), s\ k\, k reverse and forward reaction rate constant of The experimental results of dodecane and heptane reaction (5), m mol\ s\ dispersions, however, are best described assuming no kA rate constant of the catalysed reaction (3), dispersed phase in the mass transfer zone (case I). The m mol\ s\ \ value of the mass transfer coe$cient then remained prac- k* liquid side mass transfer coe$cient, m s tically constant (as expected) and in dodecane dispersions m0 ratio of the gas solubility in the dispersed and in the interfacial area was constant up to 20% dispersed the continuous phase, dimensionless \ \ phase fraction and then started decreasing (Fig. 6). In r reaction rate of component A, mol m s \ Fig. 11 the absence of mass transfer enhancement is R rate of absorption of component A, mol s clearly shown. ¹ temperature, K When applying the most probable situations for each < volume, m of the gas}liquid}liquid systems studied (thus: case I for dodecane and heptane and case II for toluene and 1- Greek letters octanol) the resulting k* values are plotted in Fig. 12. For a/a the value remained constant in all cases up to 20%. " dispersed phase fraction, dimensionless If this interpretation is valid (which needs further vali- viscosity, Pa s dation), these results show that the surface renewal frequency in these multiphase systems is not in#uenced by the presence of dispersed phase droplets. References

Benadda, B., Prost, M., Ismaily, S., Bressat, R., & Otterbein, M. (1994). 5. Conclusions Validation of the gas-lift capillary bubble column as a simulation device for a reactor by the study of CO absorption in NaCO/ The Danckwerts-plot technique was used for analysis NaHCO solutions. Chemical Engineering and Processing, 33, of mass transfer parameters in gas}liquid}liquid systems. 55}59. Brilman, D. W. F., Goldschmidt, M. J. V., Versteeg, G. F., & van Swaaij, From the experimental results and by interpretation of W. P. M. (2000). Heterogeneous mass transfer models for gas the Danckwerts plots the existence of two types of absorption in multiphase systems. Chemical Engineering Science, gas}liquid}liquid systems is indicated. Mass transfer is 55(12), 2793}2812. A. H. G. Cents et al. / Chemical Engineering Science 56 (2001) 1075}1083 1083

Bruining, W. J., Joosten, G. E. H., Beenackers, A. A. C. M., & Hofman, Linek, V., & Benes\, P. (1976). A study of the mechanism of gas absorp- H. (1986). Enhancement of gas}liquid mass transfer by a dispersed tion into oil}water emulsions. Chemical Engineering Science, 31, second liquid phase. Chemical Engineering Science, 41, 1873}1877. 1037}1046. Cornils, B. (1999). Bulk and "ne chemicals via aqueous biphasic Mehta, V. D., & Sharma, M. M. (1971). Mass transfer in mechanically catalysis. Journal of Molecular Catalysis A: Chemical, 143,1}10. agitated gas}liquid contactors. Chemical Engineering Science, 26, Danckwerts, P. V., & Sharma, M. M. (1966). Absorption of carbon 461}479. dioxide into solutions of alkalis and amines. Journal of Chemical Pohorecki, R., & Moniuk, W. (1988). Kinetics of reaction between Engineering Reviews Series No. 2, The Chemical Engineer,CE carbon dioxide and hydroxyl ions in aqueous electrolyte solutions. 244}280. Chemical Engineering Science, 43, 1677}1684. Danckwerts, P. V. (1970). Gas}liquid reactions. New York: McGraw- Rols, J. L., Condoret, J. S., Fonade, C., & Goma, G. (1990). Mechanism Hill. of enhanced oxygen transfer in fermentation using emulsi"ed Das, T. R., Bandopadhyay, A., Parthasarathy, R., & Kumar, R. (1985). oxygen-vectors. Biotechnology and Bioengineering, 35, 427}435. Gas}liquid interfacial area in stirred vessels: The e!ect of an immis- Sharma, M. M., & Danckwerts, P. V. (1963). Fast reactions of CO in cible dispersed phase. Chemical Engineering Science, 40, 209}214. alkaline solutions. (a) Carbonate bu!ers with arsenite, formalde- Eckenfelder, W. W., & Barnhart, E. L. (1961). The e!ect of organic hyde and hypochlorite as catalysts. Transactions of the Faraday substances on the transfer of oxygen from air bubbles in water. Society, 59, 386. A.I.Ch.E. Journal, 7, 631}634. Versteeg, G. F., & van Swaaij, W. P. M. (1988). Solubility and di!usivity Falbe, J. (1980). New synthesis with carbon monoxide: Chapter V: Koch of acid gases (CO,NO) in aqueous alkanolamine solutions. reactions (H. Bahrmann). Berlin: Springer. Journal of Chemical and Engineering Data, 32,29}34. Handbook of Chemistry and Physics (1994). (75th ed.). Boca Raton, FL: Weisenberger, S., & Schumpe, A. (1996). Estimation of gas solubilities in CRC. salt solutions at temperatures from 273 K to 363 K. A.I.Ch.E. Joosten, G. E., & Danckwerts, P. V. (1972). Solubility and di!usivity of Journal, 42, 298}300. nitrous oxide in equimolecular potassium carbonate-potassium Yoshida, F., Yamane, T., & Miyamoto, Y. (1970). Oxygen absorption bicarbonate solutions at 253C and 1 atm. Journal of Chemical and into oil-in-water emulsions. Industrial and Engineering Chemistry: Engineering Data, 17, 452}454. Process Design and Development, 9, 570}577.