Yanyan Liu,Chaoqun Yao*,Guangwen Chen*

1 Dalian National Laboratory for Clean Energy,Dalian Institute of Chemical Physics,Chinese Academy of Sciences,Dalian 116023,China

2 University of Chinese Academy of Sciences,Beijing 100049,China

Keywords:Mass transfer Multiphase flow Slug flow Microreactor Colorimetric method

ABSTRACT Gas-liquid-liquid three-phase slug flow was generated in both hydrophilic and hydrophobic microreactors with double T-junctions.The bubble-droplet relative movement and the local mass transfer within the continuous slug and the dispersed droplet were investigated.It was found that bubbles moved faster than droplets under low capillary number(Ca),while droplets moved faster upon the increase of Ca due to the increased inertia.For the first time,we observed that the increased viscosity of droplets fastened the droplet movement.The mass transfer in the continuous slug was dominated by convection,leading to nearly constant global mass transfer coefficient(kLa);while that in the dispersed droplet was dominated by diffusion,resulting in kL decreasing along the channel.Such features are analogical to the corresponding gas-liquid or liquid-liquid two-phase slug flow,but the formation of bubble-droplet clusters caused by relative movement lowered the absolute mass transfer coefficient.These results provide insights for the precise manipulation of gas-liquid-liquid slug flow in microreactors towards process optimization.?2022 The Chemical Industry and Engineering Society of China,and Chemical Industry Press Co.,Ltd.All rights reserved.

1.Introduction

Gas-liquid-liquid (G-L-L) systems play an important role in reaction-extraction coupling processes,biochemical and chemical synthesis [1,2].The current research in this area is mainly limited in conventional reactors,because of the lack of tools/knowledge for a precise process control and understanding of the process.For instance,usually poor three-phase dispersion is involved and thus limits the process efficiency;the reaction/extraction performance is complicated by the presence of complex mechanisms of mass transfer crossing both gas-liquid and liquid-liquid interfaces.

Continuous flow microreactors hold great promises for carrying out G-L-L operations towards realizing an efficient chemical process [1,3-5].With the lateral dimension of the flow channel reduced to belowca.1 mm,microreactors provide a short diffusion distance and large interfacial area in multiphase systems,leading to a substantial mass transfer enhancement.Moreover,a precise control over multiphase flow,especially slug flow,significantly reduced back-mixing in microreactors and narrowed the residence time distribution.Together with the large interfacial area,microreactors function as an efficient operating platform to carry out various (chemical) processes where the system performance can be precisely elucidated [1].Lastly,microreactors provide a viable scale-up method by numbering up,while maintaining the same flow/transfer characteristics as those in a single channel.These features make microreactors ideal for process intensification.

In fact,G-L-L slug flow microreactors have received more and more research attention,with promising results in the targeted applications [1].As an example,?nalet al.[6] applied G-L-L slug flow to investigate the kinetics of the selective hydrogenation of α,β-unsaturated aldehydes with aqueous catalyst (i.e.,Ru(II)-TPPTS)in a microreactor.Due to the uniformly dispersed gas bubbles and aqueous droplets,the microreactor enabled an accurate mediation of the interfacial area and mass transfer,which is crucial for the accurate determination of reaction kinetics.The flow regime can also significantly intensify G-L-L reactions,as shown by Yapet al.[7]in the hydrogenation of several olefins,which presented higher conversions and yields within a few minutes compared with its batch counterpart.However,due to the complexity of mass transfer scenarios in the G-L-L microreactors,these investigations have so far been confined to the characterization of global mass transfer coefficient (kLa),with little consideration of the local transfer mechanism elucidation.As shown in Fig.1,the involved mass transfer scenarios in G-L-L slug flow include various routines,such as G-L1(from gas to liquid phase 1),L1-L2(from liquid phase 1 to liquid phase 2) and G-L1-L2(from gas to liquid phase 1 to liquid phase 2,see compound A in Fig.1(b)).Such parallel and/or successive mass transfer steps are very likely to limit the overall extraction/reaction performance,especially when the mass transfer rate was not properly tuned according to the reaction kinetics.

To achieve the target performance and address the controlling mass transfer step,an in-depth understanding is needed.Due to the similarity of G-L-L slug flow to typical two-phase slug flow,here we refer to the mechanism in two-phase flow for inspiration.Take a gas-liquid slug flow as an example,assume that the mass transfer resistance lies in the continuous phase where a perfect mixing is achieved,van Batenet al.[9] proposed that the overall mass transfer coefficientkLain a unit cell could be assessed by the separate contributions from the bubble cap (kL,cap) and the bubble body surrounded by liquid film(kL,film),i.e.,kLa=kL,capacap+kL,filmafilm.Based on the penetration model,the mass transfer coefficientskL,capandkL,filmwas respectively estimated by the contact time of a liquid element passing by half bubble cap and through the liquid film,at the bubble velocity.Thus,under given flow condition(i.e.,bubble/droplet size,velocity),the predictedkLa viathis model remains constant along the channel.However,such perfect mixing assumption is usually not fulfilled in reality,a significant decrease of measuredkLawas observed [10-14],revised models have therefore been proposed to describe this effect [15-17].On the other hand,under the dominance of diffusion(e.g.,limited convection in liquid-liquid slug flow),Susantiet al.[11]considered the entire liquid film surrounding the dispersed droplet as a curved stationary plane.Thus,the mass transfer coefficients in the continuous and dispersed phases were determined by the residence time,askL,C=2·(DC/πτ)0.5andkL,D=2·(DD/πτ)0.5(DCandDDare corresponding diffusivities of the solute).Then,the reciprocal of overall mass transfer coefficient 1/kLbecomes the sum of the reciprocal of that in the continuous and dispersed phases(i.e.,1/kL=1/kL,C+1/kL,D).However,the validity of the above-mentioned mass transfer models in G-L-L system remains unknown.

Moreover,the influence of hydrodynamics on three-phase mass transfer is also of great importance.In G-L-L slug flow,the different moving velocities between bubbles and droplets lead to close attachment between the adjacent bubble and droplet,resulting in additional mass transfer resistance [18].Whereas,before the attachment occurs,Yaoet al.[19]showed that the inert water droplet could intensify gas-to-oil mass transfer.These results indicate that it is critical to get insights into the mass transfer mechanism for proper choice over models and reliable process predictions in G-L-L slug flow operations.

The present work has focused on the hydrodynamics and mass transfer of G-L-L slug flow in both hydrophilic glass and hydrophobic PTFE microchannels,to create mass transfer routines of G-L1and G-L1-L2,respectively.An online colorimetric method based on the oxidation-reduction reaction of resazurin was used [20].The relative motion between bubbles and droplets and mass transfer characteristics were investigated in both microchannels.These contents would pave road for the design of G-L-L slug flow microreactors towards the optimization of target processes.

2.Experimental

2.1.Materials and fluid properties

Resazurin sodium salt (CAS 62758-13-8,molecular mass:251.17 g·mol-1) was purchased from Sigma Aldrich,glucose (CAS 14431-43-7),sodium hydroxide (CAS 1310-73-2) andn-octane(CAS 111-65-9) were purchased from Aladdin (China),all chemicals were used as received.

In this work,the resazurin-based colorimetric method was used to characterize local mass transfer,as shown in Fig.2.The purple resazurin solution was first reduced to the colorless dihydroresorufin by glucose and sodium hydroxide under an oxygen-free environment,then the colorless dihydroresorufin was delivered to the microchannel and oxidized back to pink resorufin when contacting oxygen therein.The gray value of pink resorufin was employed to quantify the amount of transferred oxygen (refer to[8] for details).Based on the literature [8,15,21] and our preliminary experiments,0.3 g·L-1resazurin solution with 20 g·L-1glucose and 20 g·L-1sodium hydroxide was chosen as the aqueous phase for sufficient optical intensity and reaction rate.The aqueous feed solution was freshly made and deoxygenized by N2flush in each run and used only for maximum 40 min.Under this condition,the mass transfer resistance is located at the aqueous phase and the enhancement factor of oxidazition is very close to 1,due to which the physical mass transfer coefficients were directly measured [21].

The surface tension and viscosity of the aqueous solution were constantly monitored and found to be stable at 73 mN·m-1(measured by DataPhysics OCA 15EC,Germany) and 1.29 mPa·s (measured by a viscometer DV-II+Pro,Brookfield,USA).Air andnoctane served respectively as the gas phase and the inert organic phase in the microreactor (vide infra).Physical properties of employed fluids are summarized in Table 1.

Table 1 Physical properties of the used working fluids (20 °C,0.1 MPa)

2.2.Experimental setup and procedure

Fig.1.Schematics of mass transfer in a gas-liquid-liquid three-phase system in the microreactor(adapted from[8]).A,B are the reactants and P the product.The subscripts 1 and 2 represent the continuous phase and the dispersed phase,respectively.Catalysts reside in either of the continuous phase and the dispersed droplet phase.

Fig.2.Reactions invloved in the colorimetric method used for mass transfer characterization (reproduced from Ref.[21]).

To study the mass transfer from bubbles to continuous aqueous slugs(i.e.,G-L1mass transfer),the experiments were conducted in a glass microreactor as shown in Fig.3.The meandering microchannel after the second T-junction (denoted as T2) is 440 mm long in total,composed of 20 mm-long straight channel segments (the first and the last ones are 12 mm long) jointed by half circles (radiusrc=1 mm).All channels were of 600 μm wide and 300 μm deep (i.e.,w=600 μm,h=300 μm).The exact geometry was fabricated on PMMA plates (i.e.,PMMA microreactor) to study the effect of hydrophobicity on fluid dispersion.In the experiments,bothn-octane and the aqueous solution with resazurin were first flushed by N2and then respectively injected through inlets 1 and 2 to the first T-junction (denoted as T1) by syringe pumps (LSP02-1B,LongerPump,China).Air from a pressurized gas bottle was delivered to inlet 3 through a pre-calibrated mass flow meter (SC200,Sevenstar,China).Thereafter,the gas-aqu eous-organic three-phase slug flow was generated at T2.The flow regime and color change in the aqueous phase was recorded by a CMOS camera (Phantom M310,Vision Research,USA,working at 100-500 frames per second)with the aid of an optical microscope(SZX 16,Olympus,USA).To see the overall flow regime,a low magnification(×0.8) was adopted,as shown by the blue dotted box in Fig.3 (resolution ofca.0.0176 millimeters per pixel).A higher magnification (×2.0;ca.0.0071 millimeters per pixel) was employed for mass transfer analysis,as shown by the red solid box in Fig.3.The imaging processing method and determination of mass transfer coefficients are referred to Refs.[13-15].

To investigate the mass transfer from bubbles to the dispersed droplets through the continuous slugs(i.e.,the G-L1-L2mass transfer),a PTFE microreactor was employed.As shown in Fig.4,the PTFE tube of 0.5 mm inner diameter was straightenedviaan immersion pool,in order to improve image quality (see inserted images inside and outside the immersion pool in Fig.4(b))and thus the accuracy of mass transfer analysis.The three phases were injected in the same way as shown in Fig.3,i.e.,organic and aqueous phases were injected into T1 and air to T2.Volumetric flow rates of the three phases were also comparable to those in glass microreactor.

For simplicity,the flow rates in an experimental run is expressed using the notation as gas flow rate (QG)-aqueous flow rate (QW)-organic flow rate (QO) with units being ml·min-1.For example,the condition‘G0.1-W0.6-O0.2’refers to the gas,aqueous and organic flow rates being 0.1,0.6 and 0.2 ml·min-1,respectively.The ranges of volumetric flow rate for the gas,aqueous and organic phases were 0.207-0.995,0.30-1.00 and 0.10-1.00 m l·min-1,respectively.Under such conditions,the oxygen (in air)was in excess to resazurin,and mass transfer resistance located in the aqueous phase.The employed fluids in above mentioned glass,PMMA and PTFE microreactors,as well as the corresponding contact angles are presented in Tables S1 and S2 in Supplementary Material.All experiments were conducted at room temperature((20 ± 2) °C) and ambient pressure (0.1 MPa).

3.Results and Discussion

3.1.Relative movement between bubbles and droplets

In the hydrophilic glass reactor,the droplets and slugs were generated in T1 (similar to typical two-phase slug flow),of which the lengths were respectively correlated asLd/w=1.88QO/QW+1.48 andLslug/w=1.05QW/QO+1.36.While the generating bubbles at T2 were simultaneously cut/squeezed by both droplets and slugs,therefore the corresponding length is related to bothQG/QWandQG/QO,and correlated asLb/w=3.06QG/QW+1.52QG/QO(errors within ±15 %,see Fig.S1 in Supplementary Material).

Fig.3.Structure of the glass microreactor (horizontally placed) for the G-L1 mass transfer study in three-phase flow.The viewing sections for flow visualization and mass transfer characterization are indicated by the blue dotted box and the red solid box,respectively.

It has been reported that bubbles and droplets move at different velocities in microchannels [23,24].In the hydrophilic microchannel,the bubbles moved faster than the droplets,leading to the formation of bubble-droplet cluster.As shown in Fig.5,two different bubble-droplet structures were observed.Fig.5(a) shows that under low flow rate,the bubble would gradually catch up with the droplet,forming the ‘bubble-pushing-droplet’ cluster.Under higher flow-rate (Fig.5(b)),the bubble can not only catch up with the droplet,but also intrude into and flow through the droplet with certain distortions,ending up with a ‘bubble-intruding-droplet’cluster.This phenomenon was caused by the interfacial tension[25].According to Chenet al.[25],only when the interfacial tension between the continuous phase and the gas phase (γW-G) is higher than the sum of the interfacial tension between the continuous phase and the droplet phase (γO-W) and that between two dispersed phases(γO-G),the bubble would intrude into the oil droplet.In this work,γO-W=49.7 mN·m-1,resulting in γW-G≈γO-W+γO-G(γW-Gand γO-Gshown in Table 1).Such approximate equality led to the formation of two different bubble-droplet structures,depending on the bubble inertia.At low flow rates,the inertia of the bubble could not overcome the interfacial tension,thus only the bubble-pushing-droplet cluster formed.While at higher flow rates,the bubble with larger inertia intruded into the droplet,hence formed the bubble-intruding-droplet cluster.While in the PTFE capillary,the oil phase acted as the continuous phase and only bubble-pushing-droplet cluster was formed (γO-G＜γO-W+γW-G).

Fig.4.Structure of the PTFE microreactor to investigate the G-L1-L2 mass transfer in three-phase flow: detailed dimensions of immersion pool refers to [22].

Fig.5.Two different bubble-droplet dispersion structures in the glass microreactor: (a) bubble-pushing-droplet cluster (operating condition: G0.2-W0.6-O0.1) and(b) bubble-intruding-droplet cluster (G0.8-W0.8-O0.3).

Up to now,it has been reported that bubbles can move either faster or slower than droplets in microchannels [26-28],which was simply ascribed to the wetting properties of the continuous phase[28],while the convincing reason has not been revealed yet.The liquid film between the bubble/droplet and the wall can increase the bubble/droplet velocity [29].However,according to the film thickness law [23,24,29,30],the droplet with lower interfacial tension,leading to thicker film,moves faster than the bubble,which is opposite to what we observed in experiments.One possible reason is that bubble interface was more rigid than droplet interface,resulting in smaller drag from the film.Accordingly,an increase in droplet viscosity will also lead to an increase in droplet velocity,due to the decreased interface rigidity.

In order to verify the above speculation,a PMMA microreactor with the same microchannel geometry as the glass microreactor(Fig.3) was fabricated.Octane with 2.5 % (mass) Span 80 and air respectively served as the continuous phase and gas phase,while the aqueous phase was 0 %,30 %,50 % and 65 % (mass) glycerol solutions.This system ensured that bubbles and droplets are discrete in the main channel (γO-G＜γO-W+γW-G,21.94 ＜4.77+73)[25],as shown in Fig.6.When increasing either the total flow rate or the droplet viscosity,we observed a transition from ‘bubble-p ushing-droplet’ cluster (bubbles moved faster than droplets) to‘droplet-pushing-bubble’cluster(bubbles moved slower than droplets).Fig.6(a) shows a droplet initially located at the middle of two bubbles was gradually caught up by its following bubble,forming a ‘bubble-pushing-droplet’ cluster at the end.Fig.6(b)shows a typical ‘droplet-pushing-bubble’ cluster at increased flow rate for the same fluid system.Similarly,in Fig.6(c)-(d),when the mass concentrations of glycerol increased from 0 % to 65 % under the fixed flow rates,the ‘bubble-pushing-droplet’ cluster shifted to‘droplet-pushing-bubble’cluster.These results not only showed that wetting condition was not the cause of the phenomenon because the oil phase with Span 80 always wet the channel perfectly,but also confirmed the speculation described above.At small flow rate,the drag from the liquid film on the droplet body was the dominating factor,leading to smaller droplet velocity;while at higher flow rate,the distortion of droplet dominated,the film thickness becomes thicker and the droplet velocity was increased.

Fig.6.Dispersion structures of bubbles and droplets in gas-liquid-liquid slug flow in PMMA microchannel.Left panel: (a) ‘bubble-pushing-droplet’ clusters under G0.35-W0.2-O0.3;(b)‘droplet-pushing-bubble’clusters under G0.97-W0.2-O0.3;the aqueous phase was 30%(mass)glycerol solutions in both conditions.Right panel:(c)‘bubblepushing-droplet’ clusters under G0.35-W0.2-O0.3 with deionized water;(d) ‘droplet-pushing-bubble’ clusters under G0.35-W0.2-O0.3 with 65% (mass) glycerol solution.Liquids flow from left to right in all images.

Since both film thickness and interface rigidity are closely related to interfacial tension and capillary number,we proposed an integrated capillary number for the three-phase system and plotted a figure with the viscosity ratio of the dispersed liquid phase to continuous liquid phase λ (=μW/μO) to map the distribution of the above two different bubble-droplet structures.

where γDrepresents the interfacial tension between dispersed phase and the continuous phase.In current gas-liquid-liquid system,both gas and water phases serve as dispersed phases.Therefore,γDwas estimated by the averaged interfacial tension of both dispersed phases,according to the volume fractions of gas and water phases (?G=QG/(QG+QW) and ?W=QW/(QG+QW)),i.e.,γD=γO-G?G+γO-W?W.As shown in Fig.7,this map successfully distinguishes the ‘bubble-pushing-droplet’ clusters and ‘droplet-push ing-bubble’ clusters under experimental conditions.The overlapping area at low capillary number and low viscosity ratio indicates a comparable effect from both factors.In this area,the final structure of bubble-droplet clusters could be vulnerable to external factors,such as disturbance from inlet and bends of the capillaries.

3.2.Mass transfer

3.2.1.Bubble-to-slug mass transfer in hydrophillic glass microchannel

In the hydrophilic glass microchannel,bubbles contacted directly with the aqueous slugs,thus the oxygen transferred directly to slugs.Accordingly,the bubble-to-slug (G-L1) mass transfer process was locally characerized.A Typical raw image was presented in Fig.8(a),which shows the colorimetric evolution of oxidated resazurin.It was found that mass transfer in different types of continuous slugs (DSB and BSB in Fig.8) deviated from each other.Similar to two-phase systems,for each operation condition there were repeating units that contained the same flow structures,e.g.,a droplet followed by two bubbles as shown in Fig.8(b).Therefore,to calculate the overall mass transfer coefficient,the concentration should be averaged in the whole repeating unit.For the shooting zone under high magnification (red box in Fig.3),three locations(or residence times)were selected for mass transfer calculation before the bubble-droplet clusters formed,considering the relatively long repeating units.

Fig.7.Distribution map of bubble-droplet clusters in the main channel.

Fig.9 shows the values ofkLaunder various conditions.It can be seen thatkLavaried little with residence time τ(=Lch/Utot),which is of high analogy with the results from gas-liquid slug flow systems.The influence of flow rates of different phases was also not obvious.This may be attributed to the complex flow behavior in G-L-L systems.For example,the addition of the oil phase resulted in faster convection rate,but also produced more oil droplets that may hinder fluid mixing in the continuous slugs.Despite this,the overallkLakept in a high range of values for all experimental conditions.In our previous work on the mass transfer of gas-liquid and liquid-liquid two phase flows,we found that there was strong convective mixing in the continuous slug,which lead to rapid mass transfer rate [15].In addition,there was also strong leakage flow[29,31] that promoted mixing between the film and bulk slug,as well as mixing between adjacent slugs.Due to such strong convective mixing in the continuous phase,kLaof two-phase slug flow kept nearly unchanged along the channel length.The agreement on this phenomenon between two-phase and G-L-L slug flow suggests that convective mixing dominated in the present study.The strong convective mixing resulted from either the increased flow velocity with the introduction of a third droplet phase,or the relative movement phenomenon described in the prior section,which has been observed in our previous studies [8,19].

Based on the strong mixing in the bulk slug and between the film and bulk slug,we developed a revised van Baten model,which applied excellently to both gas-liquid and liquid-liquid two-phase slug flow [15-17].The original van Baten model [9] divides the overall mass transfer into two parts: from the bubble cap to the bulk slug (kL,capacap) and from the lateral bubble body to the film(kL,filmafilm).van Baten derived the model on the basis of penetration model,by estimating the contact time around the bubble cap as the time a circulating liquid element sweeping over half of the bubble cap at the bubble velocity and that in the liquid film was approximated by the time a liquid element passing through(Fig.10).The final formulas were described as.

The van Baten model directly sums the two parts as the overall mass transfer rates,which in fact implies an assumption of ideal mixing in the continuous phase.However,this is not true,though the convective mixing is strong within the continuous slugs.Realizing this problem,we proposed an additional term to describe the mixing efficiency (i.e.,mixing in bulk slug by circulation and mixing between bulk slug and the film) [15-17].

In the present work,the validity of the new model for G-L-L slug flow was examined.As shown in Fig.11,without refitting the parameter in Eq.(5),the predictions agreed reasonably with the experimental results from current three-phase slug flow system,and the accuracy is only slightly larger than its two-phase counterpart (±30 %vs±20 % [15]).Such discrepancy was caused by the above-mentioned relative motion between bubbles and droplets and the maldistribution of internal circulation (especially at the corner of the channel [15]).

Fig.8.(a) Raw image of mass transfer process and (b) normalized equivalent concentration of the oxidized resorufin at three locations/residence times.

Fig.9.Variation of kLa with the residence time under gas-liquid-liquid slug flow with alternate bubbles and droplets in glass microchannel.

3.2.2.Bubble-to-droplet mass transfer in hydrophobic PTFE capillary

Fig.10.The unit cell mass transfer model for the continuous phase.

Fig.11.Comparison of experimental and predicted kLa with Eq.(5).

In the hydrophobic PTFE capillary(Fig.4),the oxygen in bubbles must transfer through the continuous organic slug to the aqueous droplet (i.e.,bubble-to-droplet,or G-L1-L2mass transfer).In this section,the mass transfer process was characterized and discussed,at five locations.To exclude the effect of dissolved oxygen in the oil phase,degassing as described in the experimental section was performed prior to experiments.As shown in Fig.12(a),the dissolved oxygen in the oil phase could be completely removed.Another note should be made is that due to the limitation of T-junction and observation setup,bubble-droplet clusters occurred in the observation zone (Fig.12(b) and (c)).Because of the higher diffusivity/solubility of oxygen in organic phase compared to the aqueous phase,the excess supply of oxygen [15] and the close contact between the bubble and the droplet,the mass transfer resistance mainly located in the droplet phase.From Fig.12(c),it can be seen that there were isolated vortices within the droplet.The transferred oxygen concentrated near the two caps of the droplet while little in the middle of droplet body.This suggests that convective mixing was strong inside the vortices,but the mixing between each vortex was dominated by diffusion.

Fig.13 shows the results of overallkLaandkLin the hydrophobic PTFE capillary.kLwas calculated bykLa/a,where the specific interfacial areaawas calculated as indicated in Eq.(6).It can be seen that bothkLaandkLdecreased rapidly along the channel length (or with the increase of residence time),which is in agreement with the mass transfer inside droplets in two-phase slug flow[15].This confirms that diffusion plays an important role in the overall mass transfer.Due to the limitation of diffusion,the solute accumulated near the interface(Fig.12(c)),which reduces the driving force and mass transfer coefficient.

Fig.13 also shows that bothkLaandkLincreased first with the increase ofQG(=0.2-0.4 ml·min-1)and then decreased(QG=0.4-1.2 ml·min-1),given fixed flow rates of aqueous and organic phases.The same trend ofkLaandkLindicates that the interfacial area was not the dominant factor forkLa,becauseaincreased withQG.The effect ofQGon the overall mass transfer was closely related to the detailed flow pattern,which can be identified from Fig.12(c).When the gas flow rate was too low(i.e.,G0.2-W0.6-O0.6),oxygen could hardly be transferred to the front of the following droplet,due to the relatively long slug.Therefore,the mass transfer mainly located at the droplet rear.Under medium gas flow rates (e.g.,G0.4-W0.6-O0.6 and G0.6-W0.6-O0.6),the length of slug decreased,the mass transfer at the droplet front was intensified.With the further increasing of gas flow rate,the gas phase at the second T-junction was able to cut off the aqueous droplet from the first T-junction(i.e.,droplet breakup scenario),forming a smaller daughter droplet,which presented much weaker inner circulation.Therefore,thekLavalues decreased.Overall,the mass transfer rate reached the highest value when the flow rates of the involved three phases approaching 1:1:1 [8].

Fig.12.(a) Degassing effect in liquid-liquid flow at the T1 (b) raw image of mass transfer process (c) normalized equivalent concentration of the oxidized resorufin under various conditions.

Fig.13.Effect of QG respectively on(a)overall mass transfer coefficient kLa and(b)mass transfer coefficient kL.

Similar to the case of mass transfer in the glass microchannel,a model developed for liquid-liquid two-phase slug flow in our previous work [15] was applied to the current three-phase system in the hydrophobic microchannel.The empirical model based on the penetration theory was proposed as following.

kdiffis the mass transfer coefficient for complete diffusion in the absence of convection inside the droplet.Then the contact time is the residence time τ of the droplet flowing in the microchannel.As there were vortices in the droplet,half of the fitting parameter(i.e.,5.01/2 ≈2.5) represents the intensification of the vortices on the mixing and the term(QW/QO)0.176describes the effect of droplet size on the size of the vortices.The predictions of Eq.(7) overestimatedkLof the three-phase system.This is reasonable because the G-L1-L2mass transfer is more complicated and the G-L1mass transfer may not be fast enough to supply the L1-L2mass transfer.Accordingly,we refitted the parameter and obtained a more accurate model for the three-phase mass transfer,as shown in Eq.(9).As shown in Fig.14,for both cases of droplet non-breakup (low to medium gas flow rates)and breakup(high gas flow rate,formation of small daughter droplets,using average droplet length during calculation),the model excellently predicted the mass transfer coefficients,at all the five observation locations.

Fig.14.The comparison of experimental and predicted kL under G-L-L slug flow in PTFE capillary.

So far,we have established the mass transfer mechanisms and theory-based models for G-L1and G-L1-L2mass transfer in G-L-L three-phase slug flow in microreactors.

4.Conclusions

Gas-liquid-liquid microflow has gained enormous attention from both academic and industrial communities towards process intensification.However,the hydrodynamics and mass transfer mechanism have not been sufficiently studied yet,due to their intrinsic complexity.In this work,the relative movement of bubbles and droplets and mass transfer characteristics therein were investigated via a colorimetric method.Specifically,the local mass transfer in both continuous slugs and dispersed droplets were demonstrated and modeled in both hydrophilic and hydrophobic microreactors.

The relative movement between bubbles and droplets was found to be subjected to the interfacial tension among three phases,Canumber and the viscosity ratio of the dispersed liquid phase to the continuous phase.Under low flow rates and low viscosity ratio of the dispersed liquid phase to the continuous liquid phase,bubbles moved faster than droplets,leading to the formation of ‘bubble-pushing-droplet’ clusters;under high flow rates and high viscosity ratio,droplets moved faster than bubbles,forming ‘droplet-pushing-bubble’ clusters.

The mass transfer in the continuous slugs and the dispersed droplets under three-phase slug flow showed distinct characteristics.Within the glass microreactor,kLavalues were nearly constant along the main channel under fixed flow rates,due to the dominated convection resulting from the strong internal circulation and/or the leakage flow.By applying a mass transfer model that considers the above two factors,good prediction performance was obtained,without refitting the parameters.For three phase flow within the PTFE capillary,bothkLaandkLvalues decreased monotonically along the main channel (i.e.,with the increase of residence time),because of the significant role of diffusion against convection.In this case,penetration theory was directly applied with a pre-factor parameter to quantify the contribution of convection against diffusion.These contents will ultimately contribute to the smart design of gas-liquid-liquid three-phase microreactors for chemical reactions and material synthesis.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

We gratefully acknowledge the financial support for this work from National Natural Science Foundation of China (21991103,92034303,91634204).Yanyan Liu would like to acknowledge Dr.Lixia Yang for the kind help on the preparation of resazurin solution.

Supplementary Material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.cjche.2022.07.023.

Nomenclature

Cequivalent oxygen concentration in a unit cell,mol·m-3

C* saturated concentration of oxygen in the aqueous phase,mol·m-3

Cacapillary number

Ddiffusion coefficient,m·s-2

ddiameter,m

hheight of the microchannel,m

kLliquid-side mass transfer coefficient,m·s-1

Llength,m

Qvolumetric flow rate,ml·min-1

wwidth of the microchannel,m

μ viscosity,mPa·s

γ interfacial tension,mN·m-1

τ residence time along the microchannel,s

Subscripts

C continuous phase

ch main channel

D dispersed phase

f liquid film around the bubble/droplet body

G gas phase

h hydraulic diameter

O organic phase

uc unit cell

W aqueous phase