CFD simulation of homogeneous reaction characteristics of dehydration of fructose to HMF in micro-channel reactors

Xin Zhang ,Hui Liu ,*,Amar Samb ,Guofeng Wang

Keywords:Fructose dehydration Micro-channel reactor Simulation Optimization

A B S T R A C T In this work the applicability of the micro-channel reactor technique to the production of promising platform chemical 5-hydroxymethyl furan(HMF)from fructose in aqueous solution is systemically investigated by performing CFD simulations.Influential factors including solvents,residence time distribution of reaction mixtures,heat transfer conditions and micro-channel configurations are evaluated in terms of the reaction performance indices,i.e.,conversion of fructose,HMF selectivity and yield.A scale-up method from a single channel to a multiple channel reactor is also proposed.It is demonstrated that:1)at the single channel scale,controlling residence times and temperature distribution of the reaction mixture within the channel is crucial for enhancing the reaction performance,while different channel configurations lead to marginal improvements;2)for the scaling-up of the reaction process,a reactor module containing 15 circular parallel channels could be used as module blocks,which can be stacked one by one to meet the required reactor performance and production capacity.The present results show that micro-reactors are quite suitable for HMF production.

1.Introduction

Nowadays,with the gradual depletion of non-renew able fossil fuels,the development of new energy technologies receives much attention world-widely in order for sustainable utilization of available energies and proper solution to environmental problems.As a result,plant based biomass resources are considered an ideal candidate of future energy sources[1].Lignocellulose,the main constituent of plant biomass,is the most abundant renew able resource on the earth,and could be obtained from agricultural residues,forestry wastes and herbaceous plants.In lignocellulose,cellulose is the primary substance and can be converted into glucose and/or fructose when subjected to a catalyst,from both of which 5-hydroxymethyl furan(HMF)can be produced in the presence of an acid catalyst.Along this synthesis route,Dumesic and colleagues[2]have done a ground-breaking work in which the biologically derived fructose,which is proved to be superior to glucose as a feedstock[2],was used to produce HMF by multi-step dehydration via hybrid catalytic techniques.The HMF thus obtained constitutes a platform compound for further processing and conversion into a wide spectrum of chemicals and liquid fuels[3]through oxidation,hydrogenation and polymerization reactions,etc.

At present,there exist mainly two types of method for producing HMF,depending on the different reaction solvent systems adopted[4].In the first method,HMF is formed in a single aqueous[5,6]or organic phase[7–10],using either homogeneous[11–13]or heterogeneous[14–17]catalysts.The second method is the two-phase method where two immiscible liquid phases are used,with HMF produced in the aqueous solution and extracted into the organic phase immediately[17,18]such that the decomposition of HMF to levulinic acid and formic acid could be prevented effectively.Although there are many such investigations focusing on the reaction kinetics of the dehydration systems,researches on the reactor techniques for this process have received little attention,which is quite necessary for the industrialization of this promising synthesis route.In this work,we consider the applicability of the micro-reactor technique to the above mentioned single phase reaction system.To the best of our know ledge,there is no such study reported openly in the micro-reactor literature.

In a micro-channel reactor,the reduction of the characteristic reactor dimension,typical to tens to hundreds of microns[19],presents a variety of advantages over conventional reactor configurations,such as enhanced driving forces of mass transfer and heat transfer[20–22],enlarged specific surface areas up to 10000–50000 m2·m?3in comparison with less than 100 m2·m?3in conventional units[23],rapid transfer and efficient mixing of fluids which facilitate avoiding local hot spots and uneven concentration distributions[24],and controlling the side-reactions by precisely controlling the residence time in the reaction zone[25].

Based on the above considerations,the objective of the present work is to present a systematic conceptual design of micro-reactors for the single phase production of HMF from fructose.To this end,a reactor model is developed to simulate the reaction performance in microchannel reactors by investigating the effects of solvents,residence times of reaction mixtures,heat transfer conditions and micro-channel configurations;and then scaling-up from the single micro-channel to a multi-channel reactor is made.The present results could provide useful design and operational information for the further development of micro-channel reactors of HMF production.

2.Model Development

The single phase dehydration of HMF from fructose is considered to take place in a micro-tubular reactor.Reactant mixture is fed in the channel and a homogeneous reaction proceeds.The flow is axisymmetric due to the cylindrical nature of the channel,and laminar due to the small scale of the flowing cross-section.

2.1.Governing equations

A set of 2-D governing equations is set up for the bulk liquid phase including momentum,heat and mass balance equations.For axisymmetric steady state flow in a circular tube,the radial and axial component governing equations in their conservative forms is given by

w here ?jrepresents generic transport properties(see Table 1),i.e.,u(j=2)and v(j=3)are velocity components in the axial and radialdirections;T(j=4)are temperature,and CAj(j=5 to11)are concentrations of j-th reactive species;ajare parameters;Sjare source terms,and Γjare the transport coefficients,each of which corresponds to an individual generic transport property.Unless otherwise noted,in what follow s,we report simulating results of 2-D axisymmetric channels with the governing Eq.(1).

Table 1Parameters and source terms for Eq.(1)

2.2.Geometry and boundary conditions

The straight reaction channel has a diameter of 1 mm and a length of 20 mm.The boundary conditions used are the following:

(1)At the inlet of the channel,

(2)At the axis of symmetry of the channel,

(3)At the outlet of the channel,

(4)At the impermeable wall

2.3.Additional relations

In the production of HMF from fructose,the dehydration of fructose could be carried out in different solvents.In this work,three solvents are considered,i.e.,water,methanol and acetic acid;the corresponding reactions and rate expressions are listed in Table 2[26,27],where all the relevant reactions are in the first order.

The physical properties of the reaction mixtures including density,viscosity,and heat conductivity were calculated by using the mixing rulesbuilt in the software Aspen 8.8[28],when the corresponding properties of the pure components were known as tabulated in Table 3.

All the coupled differential governing equations were solved by the finite volume difference method using the commercial FLUENTsoftware.Structured meshes were generated using software Gambit to build up the geometry model.The flow was calculated by the SIMPLE scheme for pressure–velocity coupling,the standard method for pressure and the second order upwind scheme for momentum,species and energy equations were adopted.Preliminary computations were made to ensure that the converged results are grid independent,see Fig.1.It is noted that when the grid spacing is less than 0.1 mm,the grid fineness has no effect on the results.

Table 2Reactions and rates expressions in different solvents[26,27]

Table 3Physical properties of relevant species in fructose dehydration

Fig.1.Grid independent in a micro-tubular reactor of ?1 mm × 20 mm(inlet velocity=0.005 m·s?1(Re=29.1),inlet temperature=483 K,wall temperature=483 K,operating pressure=4 MPa,water as solvent and hydrochloric acid(p H=1.8)as the catalyst).

In order to verify the accuracy of the present simulation,the experimental data of the reaction in a similar micro-reactor[29]w ere compared with the simulated relationship between the residence time of the mixture and the outlet concentrations as shown in Fig.2,and good agreement is noted.

3.Results and Discussion

3.1.Basic features of the temperature and concentration distribution

Typically,the temperature and concentration distributions from the present axisymmetric simulations are presented in Fig.3.The axial temperature distribution is seen to decrease first and then rise(Fig.3a),in accordance with the endothermic dehydration reaction and the exothermic hydrolysis reaction of the intermediate product HMF,respectively.The radial temperature distribution is nearly uniform with a very thin boundary layer of temperature formed over the walls,owing to the small scale of the microchannel.The maximal radial temperature difference was only 0.4 K,at z=10 mm.Fig.4 shows the typical variation in HMF concentration;along the radial direction the concentration varies negligibly(at most the variation is 0.05%that is not shown here),while in the axial direction,the concentration variation exhibits roughly a spectrum of three zones,the entrance zone where the HMF concentration increases,the middle zone where the concentration attains a peak value,and the outlet zone where the concentration is decreasing due to HMF hydrolysis and reaches a steady value at local thermodynamic equilibrium.

Fig.2.Effect of the residence time on the concentration of fructose and HMF in a microtubular reactor of ?1 mm × 20 mm(inlet velocity=0.05 mm·s?1,inlet temperature=463 K,operating pressure=10 MPa,inlet fructose=10 wt%,water as solvent,catalyst C HCl=25 mol·m?3).

3.2.Simulation of fructose dehydration in different solvents

Fig.5a and b show the simulated selectivity and yield along the axial direction for the three solvent systems,respectively.From Fig.5a,it can be seen that the local selectivity decreases continuously in the axial direction;at the same axial distance,the selectivity in the water solvent is the highest.Fig.5b shows that the yield of 5-HMF increases first and then decreases from certain axial distances because 5-HMF is an intermediate product in the three cases,while the yield in the water solvent is highest within the three solvents adopted.In addition,we notice that the reaction pressure in water is 4 MPa,which is much lower than the 20 MPa in the other two solvents.Therefore,in what follow s we will focus our investigation solely on the aqueous reaction system with hydrochloric acid as the catalyst.

3.3.Effect of residence time on reaction performance

Fig.3.Axial(a)and radial(b)distributions of temperature in a micro-tubular reactor of ?1 mm × 20 mm(inlet velocity=0.005 m·s?1,inlet temperature=483 K,inlet fructose concentration=30 wt%,wall temperature=483 K,operating pressure=4 MPa,HCl as catalyst(p H=1.8)).

Fig.4.HMF concentration distribution in along the axial distance of the microreactor(The conditions are the same as that in Fig.3).

Fig.6.The mass fraction distribution of HMF in the axial direction at r=0 mm in a microtubular reactor of ?1 mm × 20 mm(inlet temperature=483 K,inlet velocity=0.005 m·s?1,wall temperature=487 K,operating pressure=4 MPa,inlet fructose=30 wt%,water as solvent and hydrochloric acid as the catalyst,p H=1.8).

Fig.5.HMF selectivity(a)and yield(b)in the axial direction at r=0 mm in different solvents(methanol and acetic acid at 483 K,20 MPa,sulfuric acid as the catalyst;water at 483 K,4 MPa,HCl as the catalyst.Process conditions:inlet velocity u=0.005 m·s?1,inlet fructose=30 wt%).

In Fig.6,a typical concentration distribution of HMF in the water solvent is show n.We see that the mass fraction of 5-HMF is the highest at a distance of 3.7 mm,which suggests that for this continuous reaction process with two competing reactions,i.e.,dehydration and hydrolysis in series,and the residence time is an important parameter in determining the reactor performance.In general,by properly adjusting the ratio of reactor volume to the flow rate of the feed to the reactor[31],the residence time of the reactants can be controlled and hence the conversion and the selectivity of the product be optimized[32].In the present case,the effect of residence time on the intermediate product of 5-HMF can be seen from Fig.7,in which the conversion and selectivity,as well as the yield are shown as a function of the residence time.We notice that an increase in conversion with the residence time is associated with a decrease in selectivity(Fig.7a),which gives rise to the peak value of the yield(Fig.7b),namely,a yield up to 41%at the residence time 4 s,with the corresponding conversion and selectivity being 75%and 55%respectively.Compared with the performance in a conventional batch reactor where the reaction temperature and initial concentration were the same as those adopted in the present simulation(see Fig.6),but the reaction time was3 min[11],the present conversion is increased by 25%and the yield is increased by 15%with the present residence time of the reactive mixture being merely 4 s.This indicates that the reaction performance could be significantly improved by controlling the residence time in micro-channel reactors.

3.4.Thermal effect on reaction performance

In the process of fructose dehydration to HMF in water as solvent and hydrochloric acid as the catalyst,the first dehydration step and the second step(see Table 1)are endothermic reactions,and the third step is exothermic;the overall process is highly exothermic[30].Therefore,temperature control is an important factor in determining the reactor performance.Here we consider two ways of temperature controlling,i.e.,by imposing constant wall temperatures and wall heat fluxes for removal of the reaction heat.

Fig.7.Effect of the residence time on the selectivity(a)and yield(b)of HMF in a micro-tubular reactor of ?1 mm × 20 mm(area weighted averaged concentration at the outlet)in water solvent at 483 K,inlet velocity=0.0025 m·s?1,inlet fructose=30 wt%and 4 MPa with hydrochloric acid(p H=1.8)as the catalyst.

Fig.8.Effect of the wall temperature on the selectivity(a)and yield(b)of HMF in a micro-tubular reactor of ?1 mm × 20 mm(inlet velocity=0.005 m·s?1,inlet temperature=483 K,inlet fructose=30 wt%,operating pressure=4 MPa,water as solvent and hydrochloric acid(p H=1.8)as the catalyst).

The influence of the wall temperature on the reaction is shown in Fig.8;obviously,for the exothermic reaction,an increase in wall temperature is favorable for the conversion but not for the HMF selectivity(Fig.8a),and consequently a maximum HMF yield of 36.61%is observed at the wall temperature of 487 K(Fig.8b),with the corresponding conversion being 77%and selectivity 48%(Fig.8a),respectively.In Fig.9,a typical temperature distribution in the channel is show n.We see that at the axial distance of 3.7 mm,the temperature is the low est,which is consistent with the highest mass fraction of HMF observed in Fig.6.The reason for this fact is that initially the endothermic dehydration is dominant and after a certain distance the exothermic hydrolysis outperforms.

Fig.9.Temperature distribution in the axial direction at r=0 mm in a micro-tubular reactor of ?1 mm × 20 mm(inlet temperature=483 K,inlet velocity=0.01 m·s?1,inlet fructose=30 wt%,wall temperature=487 K,operating pressure=4 MPa,water as solvent and hydrochloric acid(p H=1.8)as the catalyst).

The other way of controlling temperature distribution is to adjust the wall heat flux.As shown in Fig.10,the conversion rate increases with an increase in the heating flux,and the yield decreases with the increase of heat flux.The yield increases first and then decreases with the increase of heat flux.It is noted that when the heat flux is 150 W·m?2,the maximum yield is39.86%.In this case,the temperature distribution in the micro-channel is shown in Fig.11;we see that the temperature is monotone increasing without a minimum as observed in Fig.9,and hence,the yield is increased by around 3%from a comparison of the yields in the two cases.From this we conclude that the control of wall heat flux is more favorable than that of wall temperature in the range of reaction conditions investigated.

Fig.11.Temperature distribution in the axial direction at r=0 mm in a micro-tubular reactor of ?1 mm × 20 mm(inlet temperature=483 K,inlet velocity=0.005 m·s?1,inlet fructose=30 wt%,operating pressure=4 MPa,heat flux=150 W·m?2,water as solvent and hydrochloric acid(p H=1.8)as the catalyst).

3.5.Effect of channel configuration on reaction performance

In the micro-channel literature,two types of configuration received particular attention for enhancing transport processes therein,e.g.,by adopting curved channels other than the straight one such that compact structure and efficient heat transfer could be achieved[33,34],or channels with rough internal surfaces to enhance heat transfer and modify the flow structure[35].In the present work,in order to improve the reaction performance further,we consider two types of channel configuration including a bent 3D tubular channel with the same dimension(Fig.12a),and a 2D channel with internal surface roughness(Fig.12b),both of which have the same mean residence time as before,i.e.,the channel length divided by the inlet velocity as that of the straight channel micro-reactor investigated before(Fig.5).In this case,preliminary simulations were performed to make sure that the simulation results obtained are grid-independent;in Fig.13,the results are shown for the bent 3D channel,indicating that the pressure drop is basically the same when the grid spacing is less than 0.1 mm.In order to verify the present simulation,the species mixing performance of the bent reported in Chen et al.[36]was simulated by using the same methodology including the governing equations and numerical method used therein,see Fig.14;a comparison of the two sets of results shows that the agreement is good.

Fig.10.Effect of the heat flux on the selectivity(a)and yield(b)of HMF in a micro-tubular reactor of ?1 mm × 20 mm(inlet temperature=483 K,inlet velocity=0.005 m·s?1,inlet fructose=30 wt%,operating pressure=4 MPa,water as solvent and hydrochloric acid(p H=1.8)as the catalyst).

Fig.12.Geometric structure and boundary conditions:curved tube(a);w ide channel=1 mm,narrow channel=0.6 mm(b).

Fig.13.Mesh independent of the curved tube(inlet velocity=0.01 m·s?1,inlet temperature=483 K,operating pressure=4 MPa,wall temperature=483 K,diameter=1 mm,inlet fructose=30 wt%,water as solvent and hydrochloric acid(p H=1.8)as the catalyst).

Fig.14.Experimental[36]and simulated mixing efficiencies in a curved tube of ?1 mm at various Reynolds numbers(inlet temperature=378 K,wall temperature=378 K,operating pressure=0.6 MPa).

In the case of the curved tube,the yield of 5-HMF is shown in Fig.15.It is seen that when the residence time is 4 s,the yield is 36.5%,with the corresponding conversion being 70%and the selectivity 52.2%;these values are less than that in the case of the previous straight channel under the same operating conditions(Fig.5).The reason for this could be explained with the help of the following additional numerical evidences:

(1)Performing CFD simulation in the way where a Delta like injection or impulse of a tracer spices is introduced at the inlet of the channel at a certain time and then the concentration of the tracer at the outlet of the channel is monitored and recorded as the residence time distribution(RTD)curve[37].In Fig.16,the resulting RTD curves are shown for the straight tube and the curved tube.

(2)Using the continuous stirred tank reactors(CSTRs)model in the

Fig.15.Effect of the residence time on the yield of HMF in curved tube(inlet temperature=483 K,inlet velocity=0.01 m·s?1,inlet fructose=30 wt%,wall temperature=483 K,operating pressure=4 MPa,water as solvent and hydrochloric acid(p H=1.8)as the catalyst).

Fig.16.Residence time distribution in curved tube and straight channel(inlet temperature=483 K,inlet velocity=0.01 m·s?1,inlet fructose=30 wt%,wall temperature=483 K,operating pressure=4 MPa,water as solvent and hydrochloric acid(p H=1.8)as the catalyst).

RTD theory[38],model parameters which represent the degree of back-mixing can be calculated from the RTD curves.In the case of the straight channel micro-reactor investigated before(Fig.5),the square of variance σ2t=1.84,and the number of CSTRs N=8.23;in the case of the curved tube,σ2t=2.78,and N=5.98.The comparison of these parameters suggests that the back-mixing in the latter case is increased due to the curved conifguration of the reactor,which is not beneficial to the formation of the intermediate product HMF.The increased back-mixing is further evidenced by the flow structure within the curved tube,as shown in Fig.17,where the streamlines indicate that recirculation or back flowing occurs in some parts of the field.

In the case of the tube with roughness,an increased yield of 5-HMF up to 44.6%w hen the residence time is 4 s,is observed compared to the previous two cases,as shown in Fig.18;the corresponding conversion is 80%,and the selectivity is 55.9%,respectively,which are better than those in the previous straight tube.Following the same procedure of carrying out a numerical RTD experiment,we obtained the two back mixing parameters in this case,which are σ2t=1.31 and N=12.21.Obviously,a reduction of back-mixings is achieved due to the rough internal surface of the channel,which increases the HMF yield.

Fig.17.Pathlines colored by velocity magnitude in the bend of the first section of curved tube(inlet temperature=483 K,inlet velocity=0.01 m·s?1,inlet fructose=30 wt%,wall temperature=483 K,operating pressure=4 MPa,water as solvent and hydrochloric acid as the catalyst).

Fig.18.Effect of the residence time on the yield of HMFin roughness(inlet temperature=483 K,inlet velocity=0.01 m·s?1,inlet fructose=30 wt%,wall temperature=483 K,operating pressure=4 MPa,water as solvent and hydrochloric acid(p H=1.8)as the catalyst).

3.6.Scaling up from single channel to multiple channels

In order to make a scaled-up practically applicable process based on the above single channel information,we consider to carry out the reaction in a reactor composed of an array of uniformly structured parallel channels of circular geometry with the same diameter of 1 mm,the geometrical details of which are shown in Fig.19.The reactor consists of three parts:the fluid distribution part,parallel channels,and exit part,which is quite similar to the one studied by Luo et al.[39].To investigate the flow distribution characteristics of the reactor,we carried out three dimensional CFD simulations using a similar single phase CFD flow model as reported in Table 2.Structured and full- field uniform meshes without local encryption were generated using Gambit to build up the geometry,and because of the symmetry of the geometry,half of the field was calculated to reduce the load of computation.When the size of the square elemental units is less than 0.1 mm the simulation results become grid independent.The volume flow rate at the entrance of each channel is shown in Fig.20;it can be seen that the volume flow is almost equal to each channel,and the extracted data from the simulation show that the maximum volume flow rate is 0.0217 cm3·s?1and the minimum is 0.0208 cm3·s?1,corresponding to a relative error of 4.1%.Comparing the actual flow rate and its target value in a channel,if the actual flow rate is higher than the target,the size of the orifices becomes smaller to reduce the flow rate passing through the channel.Otherwise,if the actual flow rate is lower than the target,the orifices should larger to increase the downstream fluid flux,and the geometric design can satisfy the uniform distribution of the fluid in the multi-channel configuration.In Fig.21,the concentration distribution of species fructose is show n;it is evident that the reaction proceeds nearly the same in all channels due to the uniform flow distribution established.In this case,the exit yield of 5-HMF is 42.1%,the selectivity of 5-HMF is 58%,and the conversion rate of fructose is 73%,in contrast to the yield of 41%in the single channel.This leads us to conclude that the present reactor configuration could be used as a module in the scaling-up of the reaction process,in which case many such modules could be stacked one by one in order to meet the required production capacity.

Fig.19.(a)Multiple channel geometric configuration(inlet rectangle tube:with 5 mm×5 mm×10 mm;distribution chamber:rectangle with 45 mm×10 mm×5 mm,the thickness of perforated baffle:3 mm;15 circular parallel channels:diameter=1 mm,40 mm in length;exit part:rectangle with 45 mm×10 mm×7.5 mm;outlet tube:5 mm×5 mm×10 mm);(b)Two-dimensional structure(mm).

Fig.20.Volume flow rate distribution among the entrance of multichannels(singlechannel average velocity=0.01 m·s?1,inlet temperature=483 K,wall temperature=483 K,operating pressure=4 MPa,inlet fructose=30 wt%,water as solvent and hydrochloric acid as the catalyst,p H=1.8).

4.Conclusions

Fig.21.Concentration distribution of species fructose in y=0(water as solvent and hydrochloric acid as the catalyst,inlet velocity=0.01 m·s?1,inlet temperature=483 K,wall temperature=483 K,operating pressure=4 MPa).

In this work,the applicability of the micro-channel reactor technique to the production of the promising platform chemical 5-HMF from fructose in water solvent is systemically investigated by performing CFD simulations.Influential factors including solvents,residence time and its distribution of reaction mixtures,heat transfer conditions and micro-channel configurations are evaluated in terms of the reaction performance in dices,i.e.,conversion of fructose,HMF selectivity and yield;and a scale-up method from a single channel to a multiple channel reactor is proposed.Based on the work done and discussions presented,the following major conclusions can be draw n.

Simulations at the single channel scale demonstrate that for the dehydration of fructose to HMF,controlling residence times and temperature distribution of the reaction mixture within the channel is crucial for enhancing the reaction performance,while different channel configurations lead to marginal improvements.In contrast to the conventional batch reactor,the adoption of the micro-channel reactor gives rise to the typical yield of 5-HMF up to 41%,with the selectivity of 5-HMF being 55%and the conversion of fructose being 75%,under properly selected operational conditions.

For the scaling-up of the reaction process,a reactor module which contains 15 circular parallel channels is designed and proposed to be used as module blocks stacked one by one in order to meet there quired production capacity.It is demonstrated that the multiplechannel geometry provides uniform distribution of the flow to the reactor and hence good reaction performance comparable to that in the single channel.

Nomenclature

a parameter defined in Eq.(1)

CAconcentration,mol·m?3

Cpthermal capacity,J·kg?1·K?1

D diffusivity,m2·s?1

Dhhydraulic diameter,m

f friction factor

k thermal conductivity,W·m?1·K?1

L channel length,m

P pressure,Pa

r radial coordinate,m

rACreaction rate,mol·s?1·m?3

S source terms defined in Eq.(1)

T temperature,K

u axial velocity component,m·s?1

v radial velocity component,m·s?1

z axial coordinate,m

Γ transport coefficient defined in Eq.(1)

ε absolute roughness

μ viscosity,Pa·s

ρ density,kg·m?3? generic transport property defined in Eq.(1)

Subscripts

i variable(i=1,axial velocity;i=2,radial velocity;i=3,temperature;i=4,concentration)

j phase

0 inlet