Kai Zhu,Chaoqun Yao,Yanyan Liu,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:Microchannel Microreactor Separation Purification High pressure

ABSTRACT The usage of capillary tubes for CO2 absorption suffers from small residence time,which leads to reduced performance for large throughput.This work presents a method of connecting expansion units to capillary tubes to serve as a residence time delayer.The effect of the expansion unit on gas–liquid hydrodynamics,pressure drop and mass transfer coefficient(kL a)are investigated under various operating pressures up to 4.0 MPa,for both physical and chemical absorption.A novel periodic jetting flow is found in the expansion unit,which can intensify the CO2 absorption.Experimental results show that the strategy can significantly decrease the pressure drop while maintaining the absorption performance to a large extent.The overall kL a for physical and chemical absorption are correlated to pressure drop,respectively.Besides,CO2 loading in rich absorbents increases dramatically compared to literature studies with only micromixers or capillary tubes,which is beneficial to regenerate solvent.The study verifies the concept that pre-treatment with water can largely reduce the usage of amines,and can also provide a guide for process design in natural gas purification such as biogas recovery.

1.Introduction

Natural gas,as an important energy source,covers about 20%–30%of the worldwide energy consumption every year[1].However,acid gas impurities such as CO2are commonly found in natural gas reserves at high levels(i.e.,5%–80%by volume)[2].To meet specifications for pipeline transportation and also improve heating value of the natural gas,CO2must be removed.Chemical absorption with amine solutions is the most mature technology for natural gas purification.However,the conventional column technology suffers from high-energy consumption and operating costs due to significant limitations in mass transfer and huge equipment volumes.Additionally,the large equipment size makes these processes not suitable for purification of biogas,which shows a huge prospect in the worldwide energy supply[3].Collecting the biogas sources,such as farm-scale biogas and landfill gas from anaerobic digesters,can not only serve as renewable energy,but also slow down the greenhouse effect since methane is about 21 times larger than CO2with respect to the greenhouse gas(GHC)warming effect[4].Hence,for both economic and environmental reasons it is imperative to develop more efficient and versatile technologies to effectively remove CO2from natural gas.

The utilization of microreactors in CO2separation is one promising technology,which has attracted plenty of interest during the last decade[5–9].Compared with conventional column equipment,microreactor possesses many distinguished advantages,including small volume,efficient mixing performance,large heat/mass transfer rate,increased safety and easy scaling up[10–15].Hence,for gas–liquid absorption,substantial process intensification can be obtained using the microreactors.Particularly,the small size of microreactors provides a huge advantage in on-site purification of small-scale biogas,as well as in applications limited by space such as floating liquefied natural gas(FLNG)production.Up to now,a series of microreactors with various structures have been designed and attempted for CO2separation[16–19].These studies all show that mass transfer performance are substantially enhanced,indicating great potential of microreactors.In addition,a decrease in channel diameter or the introduction of meandering/zigzag structures results in increased mass transfer rate[20].Chen et al.[17,21,22]investigated the effect of absorption temperature in microreactors for both MEA (Monoethanolamine) and DEA(Diethanolamine)systems.The optimized temperature was found to be around 40°C.A further increase in temperature will lead to an important impact of the reduced CO2solubility and CO2release from the reverse reaction.All these works revealed plenty of understanding on the absorption processes.

Although there are a lot of research studies in literature,there are two main limitations considering the microreaction technology.Firstly,current studies rarely aimed at the applications for natural gas purification.They were mostly conducted under atmospheric pressure,which usually required large amount of absorbents [17].It is economically unattractive especially when CO2content is high.Considering this drawback,we previously proposed a method to first pre-absorb CO2with water and then treat with DEA solutions under elevated pressures[17].The usage of DEA was effectively reduced.Elevating the system pressure will not bring additional cost since the feed gas source usually ranges from 2 MPa to 7 MPa.The other limitation is that due to the small reactor volume and low residence time,the absorption efficiency will rapidly decrease when the throughput is increased to a certain extent[21,23–25].If methods such as zig-zag design and long microchannels are induced to increase the performance/throughput,the pressure drop will increase largely[17].In fact,the pressure drop is also a serious issue for the applications of microreactors.Therefore,it can be concluded that there is still a lot of efforts needed concerning the CO2removal from natural gas streams.

In the present work,a simple strategy is proposed to decrease the pressure drop while maintaining good absorption efficiency.The concept is that a microreactor is used as an efficient gas–liquid mixer while capillary tubes with larger diameters serve as the delayer to increase residence time.Expansion units are additionally adopted to increase the residence time.The aim is to test the validity of the concept and study the absorption characteristics.The hydrodynamics in the expansion unit is studied.The absorption processes with water and DEA solution as the absorbents are investigated under different system pressures,in terms of pressure drop and mass transfer performance.The effects of expansion unit number,system pressure,gas fraction and molar ratiowill be presented and discussed.

2.Experimental

2.1.Experimental apparatus

Fig.1 shows a diagrammatic sketch of the micromixer and residence time delayers.In the micromixer which was placed horizontally,the gas that was dispersed by a metal foam with cavity of>90%and mean pore size of 5 μm,contacted liquid flowing in the parallel channels in the bottom plate.The aperture in the up plate has a size of 15×15 mm2.There are four parallel channels with a cross section of 2.0 mm×0.6 mm and a length of 25 mm on the bottom plate.The whole size of the micromixer is 74 mm×64 mm×32 mm.Due to the relatively short distance and large cavity of the metal foam,the gas–liquid two phase pressure drop in the micromixer was very small,which will be illustrated in the discussion section.After exiting from micromixer,the gas–liquid two phase flow entered the delayer composed of capillary tubes and expansion units.The capillary tubes had the same size(2 mm in Din).The expansion unit was a chamber designed with a cylindrical shape with a Dinof 9 mm and a volume of 1.2 ml.The expansion units and capillary tubes are made of stainless steel.Four different combinations were used for study,as displayed in Table 1.For the Case 1 without expansion unit,only one capillary tube with a length of 2.0 m was used and the total volume was 6.28 ml.For other cases,the total length of the capillary tubes was mediated to keep the whole volume of the delayer to be 6.28 ml,and was divided equally by the expansion units.For example,for the Case 4,3 expansion units connected 4 capillary tubes with the same length.Additionally,we also designed one cylindrical chamber with the same diameter of 9 mm and volume of 6.28 ml,in order to test the absorption characteristics without the capillary tubes.It needs to be noted that the design parameters are not optimized.Hence the present work was mainly to prove the validity of the concept.

The experimental apparatus is schematically displayed in Fig.2.The detailed setup can be found in our previous study[17].Some important details are provided in the present work.Gas phase was supplied from a gas cylinder and gas flow rate was controlled by a MFC(D07-7B,0–10 L·min?1,Beijing Sevenstar Electronics Co.,Ltd.,China).The MFC could supply constant mass flow rates(or volumetric flow rate under atmospheric pressure),which is expressed as QGAhereafter in this work.Comparatively,the gas flow rate under operating system pressure was expressed as QGS.The liquid phase was delivered by a piston pump(Beijing Xingda Technology Co.,Ltd.,China).A buffer tank was installed to eliminate the pulsation at the outlet of the piston pump.The gas–liquid mixture leaving the delayer was separated by a separation tank(ID:50 mm).The operating pressure was adjusted by a back pressure valve.The content of exhaust gas from the outlet was detected by an infrared sensor(Foshan Analytical Instrument Co.,Ltd.,China).The capillary tube was directly inserted into the separator to suppress the absorption from end effect[16].The level of liquid was maintained high in the separation tank(shown in Fig.2)with the liquid simultaneously being delivered out from the separator by a piston pump.These designs were arranged so that quite short gas–liquid two-phase contact time was obtained in the separation tank.The experiments were repeated by three times for each condition.

Fig.1.Diagrammatic sketch of(a)micromixer and(b)residence time delayers.

Table 1 The characteristics of the delayers

2.2.Fluid properties and parameter settings

10 vol%carbon dioxide(CO2)balanced with nitrogen(N2)with purity of>99.995%(Dalian Special Gas Co.,Ltd)and 30 wt%aqueous DEA solution were used for the majority of the absorption tests,while 30%and 50 vol%CO2were used to investigate the effect of the initial fraction.The DEA(analytical reagent)was purchased from Guoyao Co.,Ltd.,and was used without purity.If the CO2concentration is not specified in this paper,it then refers to the 10 vol%case.The gas phase flow rate was set in the range of 0.5–6 L·min?1.The system pressure varied in the range of 0.1–4.0 MPa for physical absorption and of 0.1–2.0 MPa for chemical absorption,respectively.The absorption temperature was controlled at 293 K.

The absorption performance was characterized by the absorption fraction η and mass transfer coefficient kLa.η was calculated according to Eq.(1).

where N represents the molar flow rate.kLa was determined according to Eq.(2),

where Vmrrepresents the total volume for absorption.The log-mean interfacial concentrationwas defined as Eq.(3),

where H represents the Henry constant of carbon dioxide.in the pure water at 293 K was acquired as 2610.78 kPa·L·mol?1from Cents et al.[26].

DEA is generally considered to be a secondary amine with a molecular formula expression of R2NH,and R refers to CH2CH2OH.According to the zwitterion mechanism[27–29],the reaction between DEA and CO2first forms a complex as a transient intermediate(Eq.(4)),which is sequentially by the deprotonation of zwitterions by bases,

where B represents the bases,which can be DEA,H2O,OH?,or.kLa for chemical absorption was also calculated from Eq.(2).in the 30 wt% DEA solution at 293 K was acquired as 3178.89 kPa·L·mol?1from Yao et al[17].

3.Result and Discussion

3.1.Multiphase flow in the expansion unit

In order to investigate the effect of the expansion unit on the gas–liquid two phase flow,a transparent microreactor was fabricated with polymethal methacrylate plate (i.e.,PMMA) using milling method(FANUC KPC-30a).The microreactor contained an expansion unit(with a length of 18.9 mm and cross section of 9 mm × 9 mm)connecting two microchannels(with a length of 50 and cross section of 2 mm×2 mm).With this transparent microreactor,the hydrodynamic of gas–liquid two phase flow entering the expansion unit could be observed by a high speed camera.For hydrodynamic study,air was used instead of the CO2/N2mixture.

Typical flow images are shown in Fig.3.Under experimental conditions,the two phase flow in the microchannel was mainly annular flow or slug-annular flow,in accordance with the flow regimes in literature[30,31].Such flow pattern is very common in tubes,especially in micro/milli tubes where the effect of gravity can be neglected.When the two-phase flow entered the expansion unit with relatively large diameter,the gravity played an important role,leading to the flow pattern with liquid phase at the bottom layer and large bubbles at the upper layer.The large bubbles were generated as the gas core of annular flow penetrated into the liquid layer.They would coalesce fast with the largest bubble at the top of the expansion unit for small flow rates,as shown in Fig.3.At larger flow rates,the formed bubbles may move in the liquid layer for certain distances,or smaller bubbles may be generated and flowed out directly without coalescence with the top bubble.Fig.3 also shows the difference between the water-air system and the DEA solution-air system.The difference under small flow rates was small while under high flow rates it formed plenty of fine bubbles for the DEA solution-air system due to the larger liquid viscosity.In addition,there were also much bubbles crowded at the periphery of the expansion unit.

Fig.2.Schematic of the experimental apparatus.

If the flow in the expansion unit was always kept as described above,the expansion could only serve as the residence time delayer as the mixing/mass transfer was poor.However,an interesting jetting flow was observed,which could dramatically intensify the two phase mixing/mass transfer.As can be seen in Fig.4,various jetting flows were observed,depending on the flow conditions.For each condition,the jetting flow occurred frequently.We found that the formation frequency of the jetting flow was nearly constant for small flow rates.For example,the formation frequency was about 6 Hz for water-air system under QGAof 100 ml·min?1and QLof 10 ml·min?1.But under larger flow rates,the formation frequency was not stable and varied largely.For water-air system under QGAof 100 ml·min?1and QLof 20 ml·min?1,the formation frequency ranged from 7 to 16 Hz while under QGAof 100 ml·min?1and QLof 30 ml·min?1the formation frequency ranged from 8 to 40 Hz.The jetting intensity also varied from time to time under fixed flow condition.This suggested that the jetting flow was caused by a chaotic flow disturbance.Generally,the occurrence of the jetting flow became more frequently when either gas or liquid flow rate increased.The jetting intensity also increased,as shown in Fig.4.In this study,we measured the maximum jetting velocity observed,and compared to the superficial velocity in the microchannel.The result is displayed in Fig.5.It indicated that for small flow velocity(i.e.,around 0.5 m·s?1),the maximum jetting velocity was close to or slightly smaller than the superficial velocity.This is reasonable that the jetting was likely caused by the inertia of the two-phase flow from the microchannel under these conditions.Whereas,the maximum jetting velocity increased drastically as the superficial velocity increased.The maximum jetting velocity could be several times larger than the superficial velocity.The jetting could be so strong that it could be directly sprayed into the outlet microchannel and induce intensive chaotic vortices in the expansion unit.Besides,the results suggest that the jetting was more intensive in the DEA/H2O-air system than in the H2O-air system since the maximum jetting velocity was larger.This again indicated that the jetting flow was caused by a chaotic flow disturbance,instead of just the inertia effect.

Fig.6 displays the formation process of a jetting flow for DEA/H2Oair system.As can be seen,the jetting flow was formed after the generation of a liquid slug in the upstream annular flow in the microchannel.Usually,it is common for annular flow to generate wavy gas core[30,31],similar to t=7.5 ms in Fig.6.However,it is very difficult to rupture the wavy neck and form liquid slugs in annular flow.Even for slugannular flow,the moving velocity of the slug should be approximately the total superficial velocity.Therefore,the formation of the slug was not the intrinsic reason for the jetting flow.While the exact reason is unknown yet,the jetting may be caused by the“spring effect”of the expansion unit.It is clear that the two-phase flow in the expansion unit was a flexible system(i.e.,bubbles,unconfined fluids)that could store and release energy,and influence the upstream flow(i.e.,pressure fluctuation).Such spring effect could lead to the formation of liquid slugs and the jetting of the slugs subsequently.This can explain why increased liquid flow rate could promote the jetting as the formation of liquid slugs was easier.

3.2.Frictional pressure drop

To investigate the frictional pressure drop,the acceleration pressure drop resulting from the gas expansion effect,needs to be first evaluated.It is given by Eq.(6)[18,31],

where x represents the gas mass fraction and G represents the total mass flux.α represents the void fraction and was calculated by the correlation as follow[32,33].

The frictional pressure drop was calculated by excluding the acceleration pressure drop.

Fig.3.Typical images of gas–liquid flow hydrodynamics(top view)in the expansion unit.The flow direction is from left to right.

Fig.4.The jetting flows in the expansion unit.

Fig.5.The maximum jetting velocity for different flow conditions.

It needs to be noted that these parameters for analysis were based on capillary size as it contributed the majority of the pressure drop,which will be illustrated in this section.The results indicated that for physical absorption ΔPTP,Aonly accounts less than 0.03%of ΔPTP.The reason is that x was quite low and large amount of gas absorption also reduced the gas expansion.Similar results have been observed in our previous studies[17,31].For the same reason,the acceleration pressure drop was also negligibly small in chemical absorption.

The influence of operating pressure is shown in Fig.7.Under fixed QGA,the actual flow velocity decreases as the system pressure increases,thus leading to obviously declined pressure drop.In the case of fixed QGS,the effect of system pressure is not apparent.In our previous study[17],we found that the pressure drop first rapidly increased as the operating pressure increased from the 0.1 MPa to 1.0 MPa,and then varied only a little when the system pressure further increased to 4.0 MPa.A conclusion was made that both gas density and superficial velocity play important roles in the pressure drop.Since ΔPTPin the present study is several times smaller than that study,the negligible effect of system pressure for fixed QGSis reasonable.

Fig.6.The formation process of the jetting flow.DEA/H2O-air system,QGA =400 ml·min?1,QL =20 ml·min?1.

The effect of the expansion unit number on frictional pressure drop is displayed in Fig.8.Clearly,the pressure drop decreases obviously with increase of the expansion unit number.In order to evaluate the contribution of the expansion unit,as well as the micromixer to the pressure drop,we implemented blank experiments only with the micromixer and a single expansion unit.As we have stated before,for all the designs combining the capillary tube and the expansion unit,the total volume was kept the same.The expansion unit in the blank experiment was separately fabricated with the same diameter as the standard expansion unit.The pressure drop with this setup is far smaller than the others,indicating that the pressure drop in the micromixer and the expansion unit is negligible.Further,the pressure drop was plotted against the total length of the capillary tubes for each setup during experiments,as shown in Fig.8(b).A linear relationship is observed for each operating condition.This again confirms that the two-phase pressure drop mainly originates from capillary tube.Therefore,it is reasonable to analyze the pressure drop based on the flow conditions in the capillary tubes.

Since the flow in the capillary tubes was mainly annular flow the Lockhart-Martinelli model(L-M model)[34]was suitable for analyzing and modeling the pressure drop.This model is a typical separate-flow model which correlates the gas–liquid pressure drop with pressure drop of single phase flows.The L-M model is described as follows,

where C is an empirical parameter that depends on a lot of factors,including mass flux[31],channel cross-sectional area[36],and fluid characteristics[37].For laminar flow,a value of 5 was given[38]as the upper limit of laminar flow.

Discussion above indicates that ΔPTP,Fin the capillary equipped with expansion units mainly results from the capillary tube.The major influencing parameters include the phase flow rates and system pressure.Therefore,for the pressure drop prediction,an empirical correlation of C is proposed in Eq.(11).Its validity was also verified by the chemical absorption system.As shown in Fig.10,both the pressure drop in physical and chemical absorption systems is well predicted.

Fig.7.The frictional pressure drop under different system pressures at fixed QGS .

Fig.8.The effect of(a)expansion unit number(b)capillary tube length on ΔPTP ,F .

3.3.Physical absorption

It is very appealing to pre-absorb the CO2by water from the gas stream with high concentration of CO2.First,such pre-treatment can significantly reduce the usage of amines,which can save their regeneration cost to a great extent.Second,water itself is cheap and environmentally benign.Third,only pressure reduction can realize water regeneration without excessive heat inputs.For the purification of natural gas,which either is of high pressure or needs to be pressurized for the grid-pipeline usage,it will not increase the operation cost additionally.Although there are many studies concerning CO2absorption with water in microchannels[31,39–41],they mainly focused on the fundamental study of gas–liquid mass transfer,as well as absorption with pure carbon dioxide.In the present study,the absorption performance of gas mixture with 10%CO2is tested so as to evaluate the influence of the expansion unit,as well as other parameters such as flow rates and system pressure.

Fig.9.Measured friction multiplier at different operating pressures for physical absorption.

The influences of operating pressure and the expansion units are displayed in Figs.11 and 12,respectively.Under experimental conditions,the concentration in water at the outlet accounted 36%–97%of the saturation concentration,suggesting that the capacity of water was highly utilized in the present setup.The overall kLa varied in the range of 0.015–0.15 s?1,which agrees well with literature with capillary tube of similar diameters[42,43].As shown in Fig.11,for fixed QGA,the outlet CO2content decreases with increasing the system pressure,whereas kLa presents the opposite tendency.This is because under higher system pressure,the solubility of CO2in water is larger,and the residence time is longer due to the reduced superficial velocity of gas phase.Both factors are beneficial for the absorption.However,as QGSis decreased,due to the reduced renewal and convection rate,as well as the reduced interfacial area (i.e.,transition from annular flow to slug-annular flow),it eventually leads to smaller kLa.

For fixed QGS,both the absorption fraction and kLa present a slight decrease with increasing the operating pressure from 1.0 MPa to 4.0 MPa.In our previous studies[17,44],increasing system pressure in a similar range was shown to have little influence on or slightly increase the mass transfer coefficient.However,from 0.1 MPa to 1.0 MPa,a relatively large increase in kLa was observed[17],which was explained by an obvious change in the film thickness of the annular flow[32,45].It suggests that the effect of system pressure may behave differently,depending on the detailed flow conditions and reactor structures.This is similar to conditions in conventional reactors,in which different phenomena are also observed [46,47].Anyway,in micro/milli channels,due to the similar superficial velocities and flow patterns for fixed QGS,the mass transfer coefficient is generally influenced by system pressure to a small extent.However,the absorption flux can be substantially increased by elevating the pressure.

Fig.10.Predicting performance of Eq.(11).

With respect to the influence of expansion unit number on the absorption process,the absorption nearly shows no decline when the expansion units are used together with the capillary tube,as shown in Fig.12.When the expansion unit with the same volume is used solely,only slight decline at high flow rates can be observed.Such amazing performance may be attributed to the jetting phenomena in the expansion units as discussed previously.Though the interfacial area is smaller than in the micromixer and capillary tubes,the frequent jetting can enhance the two phase mixing/mass transfer to a great extent by creating strong turbulence and fine bubbles.Hence,with the expansion units,the pressure drop can be significantly reduced while the high mass transfer performance is maintained,which is very beneficial to reduce the operating cost.This result also indicates the success of the strategy to utilize the efficient gas–liquid mixing in the micromixer,and increase the residence time with combination of the long capillary tubes and expansion units.

3.4.Chemical absorption

Fig.11.Influence of operating pressure on kL a for(a)fixed QGA and(b)fixed QGS .

Fig.12.Effect of expansion unit number on physical absorption.

The chemical absorption in these reactors was investigated by varying the molar ratiosystem pressure and gas inlet fraction.Compared with physical absorption,mass transfer is substantially enhanced due to chemical enhancement,as has been widely shown by literature studies.Fig.13 indicates the influences of DEA/CO2and operating system pressure on the absorption for Case 2 with one expansion unit.The former parameter is less paid attention to in literatures.For study with fixed DEA/CO2,it can reveal its influence on the loading of CO2in rich solvents,which is an important issue for regeneration processes[48].It can be seen that for fixed DEA/CO2,η rapidly reduces as the flow rates increase.This is a main drawback of microreactors,which retards the large scale application[25,49].The absorption fraction,as well as kLa,can be significantly elevated by increasing DEA/CO2at fixed gas flow rate.The reason is that both the total flow velocity and the reaction rate are increased.Whereas,it also leads to a rapid decline in the CO2loading in rich solvents,which is not desired in regeneration.Overall,the CO2loading ranges from 0.15 to 0.41 under experimental conditions.It is quite higher than literature studies[18,19,49].Specifically,in the study of Ganapathy et al.[49]with the same inlet CO2fraction,to obtain a similar absorption fraction in a single microreactor,much larger amount of DEA needs to be used,leading to the small CO2loading ranging from 0.01 to 0.12.In our previous study[17],a large value of 25 of DEA/CO2was required and a low CO2loading of 0.03 was obtained.This again proves the effectiveness of the current strategy in the present study.When the system pressure is elevated,the CO2loading can be further increased.

The influence of operating pressure on chemical absorption is also studied,as displayed in Fig.13.Since it is impossible to keep the same QGSwhile not changing the DEA/CO2,only the same QGAwas fixed for comparison.Clearly,the absorption performance is improved as operating pressure is elevated,especially from 0.1 MPa to 1.0 MPa.Similar to the physical absorption,kLa reduces with the increase in operating pressure due to the reduced QGS.But since both the residence time and CO2partial pressure in the gas phase were increased,the absorption performance was strongly intensified.Under elevated pressure,CO2loading in the rich solvent could be as high as 0.41 mol·(mol DEA)?1.

CO2fraction in natural gas sources can vary largely from 2%–50%,which may lead to unexpected phenomenon[25].Hence,it is necessary to investigate the absorption performance with different CO2fractions.In this study,three fractions,namely 10%,30%and 50%,are involved.Typical results are shown in Fig.14.It can be seen that higher absorption fraction as well as higher CO2loading in the rich solvent,is achieved for larger initial CO2fraction.The same tendency was observed by Ganapathy et al.[49]with initial CO2fraction of 2.5%–7.5%.Meanwhile,kLa also increases as the CO2fraction increases.Under 0.1 MPa,the outlet CO2fraction can be reduced from about 20%to 7.7%,13%to 5.4%,and 5.6%to 1.5%for the 50%,30%,and 10%concentration cases,respectively.From the results,it can be inferred that it is easier to reduce the CO2concentration to below 2% (i.e.,maximum concentration in commercial natural gas product)with smaller initial fraction.Hence,for the pipeline fuel application,to reduce the CO2concentration by pre-treatment with water is essentially important.

Fig.13.The influence of operating pressure on(a)η(b)kL a and(c)CO2 loading in rich solvents.

With respect to the effect of the expansion units,it presents different behavior with the case for physical absorption.η and kLa decrease as more expansion units are used,as displayed in Fig.15.For the case with only the expansion unit,the absorption performance is far worse than the other cases with both expansion units and capillary tubes.Only when the throughput is very small,the absorption performance of these cases is close,and approaches complete absorption.A possible reason can be inferred from the results.Due to larger liquid viscosity of DEA solution,the intensification effect of the jetting phenomenon is less significant compared to physical absorption despite the jetting is stronger.However,though gas–liquid mixing and absorption in the expansion unit are declined,its combination with capillary tubes can still achieve reasonably good absorption.As shown in Fig.15,even though three standard expansion units are used,with remaining only 1/3 of the capillary length(Fig.8),the absorption performance still pertains at a relatively high level.

3.5.Mass transfer correlation and comparison with conventional reactors

Fig.14.The influence of CO2 fraction on(a)η and(b)kL a.

Transport theory implies that convective mass transfer is highly dependent on the power consumption of the gas–liquid systems[50].High kLa is obtained when the energy is effectively transformed to mass transfer,instead of heat generation.Therefore,kLa can be correlated to the total power consumption[51].As the capillary tubes are the main cause of the pressure drop in this study,the overall kLa can be correlated to the flow conditions therein.Since the flow was pressure-driven,the two phase pressure drop was used to calculate the power consumptions(kW·m?3),which are calculated with the following expressions,respectively[17,52].As the expansion unit is shown to affect the absorption performance,the number of the expansion unit is involved to estimate kLa for physical absorption(Eq.(14)),with parameter range:10

Fig.15.The effect of the expansion number on(a)η and(b)kL a.

The mass transfer coefficients are compared to other kinds of reactors,with the typical ranges of mass transfer coefficients obtained from Yue et al.[31],as shown in Fig.17.The microreactor used in our previous work[17]is also induced as the operation conditions are similar.Obviously,kLa in the present work is much higher than in conventional large-scale reactors,showing the great advantage of microreactors[31,54].The microreactor in our previous work is better than the present study with respect to mass transfer coefficient.However,the pressure drop is reduced by several times while achieving better absorption efficiency,especially for chemical absorption.The results show that strong process intensification is achieved by the simple strategy introduced in this work,for the purpose to capture CO2and sweeten natural gas.

4.Conclusions

This work presents a strategy to intensify the CO2absorption process with a microreactor by implementing capillary tubes and expansion units,for the purpose to increase the residence time and reduce pressure drop.The absorption characteristics are investigated by varying several important parameters,such as the system pressure,number of expansion unit and gas fraction.DI water and DEA solution of 30 wt%are used as the absorbents.The gas–liquid pressure drop,as an important design parameter,is studied with the Lockhart-Martinelli model.It is shown that by introducing the expansion units,the pressure drop can be significantly reduced.An empirical correlation is developed for the prediction of two-phase pressure drop,which gives very good predicting performance for both the DI water and DEA solution.

Fig.16.The prediction performance over kL a for(a)physical and(b)chemical absorption.

Fig.17.Comparison of kL a from different reactors.

A periodic jetting flow is observed in the expansion units,which may be caused by the“spring effect”of the expansion unit.The jetting velocity can be several times larger than the flow velocity in the capillary tubes,leading to significant intensification in the gas–liquid mixing/mass transfer in the expansion units.For physical absorption,an amazing finding is that the absorption performance is rarely affected by the expansion units.While the pressure drop is largely reduced,it presents a significant advance in reducing the operation cost.For chemical absorption,the absorption performance decreases as the number of expansion unit increases.But,the decline is restricted to a limited value and the overall performance is still excellent.Therefore,strong process intensification is also achieved.Additionally,a highlight of the present study is that very high CO2loading in rich solvents after absorption,which is comparable to conventional column,can be achieved in the present setup.It is of great importance since amine regeneration is the main concern in energy consumption.The study indicates great potential of microreaction technology in the removal of CO2,as well as other acidic gases,such as biogas purification,landfill gas and FLNG(floating liquefied natural gas).We also proposed correlations for predicting the mass transfer for both physical and chemical absorptions,by correlating to the pressure drop.

Nomenclature

A cross-sectional area,m2

C parameter in the Lockhart-Martinelli model

DHhydrodynamic diameter,μm

G total mass flux,kg·m?2·s?1

H Henry constant,kPa·L·mol?1

j superficial velocity,m·s?1

kLa overall mass transfer coefficient,s?1

N molar flow rate,mol·s?1

P pressure,Pa

Q flow rates,ml·min?1

Re Reynolds number

T temperature,K

U velocity,m·s?1

V volume,m3

x mass fraction of gas

α void fraction

η absorption efficiency

ε energy consumption,kW·m?3

ρ density,kg·m?3

Subscripts

A acceleration

F friction

G gas

GA gas under atmospheric pressure

GS gas under system pressure

L liquid

mr microreactor

TP two phase

Acknowledgements

We acknowledge gratefully the financial supports for this project from National Natural Science Foundation of China (No.21676263,U1608221),the CAS supports of the Youth Innovation Promotion Association CAS(No.2017229),DICP(DICP I201925).