Huisheng Lü,Guoqing Wang ,Minhua Zhang ,*,Zhongfeng Geng ,Miao Yang ,Yanpeng Sun

1 Key Laboratory for Green Chemical Technology of the Ministry of Education,Tianjin University R&D Center for Petrochemical Technology,Tianjin 300072,China

2 Petro China Northeast Re fining&Chemical Engineering Co.Ltd.,Jilin Design Institute,Jilin 132000,China

Keywords:Adsorption isotherm Citric acid Cyano column Isotherm model

ABSTRACT Supercritical adsorption equilibrium has a significant role in defining supercritical adsorption behavior.In this paper,the adsorption equilibrium of citric acid from supercritical CO2/ethanol on a cyano column was systematically investigated with the elution by characteristic point method.Equilibrium loading was obtained at313.15 K and 321.15 K with supercriticalCO2/ethanoldensities varying from 0.7068 g·cm-3 to 0.8019 g·cm-3.The experimental results showed that the adsorption capacity of citric acid decreased with increasing temperature and increasing density of the supercritical CO2/ethanol mobile phase.The adsorption equilibrium data were fitted well by the Quadratic Hill isotherm model and the isotherms showed anti-Langmuir behavior.The monolayer saturation adsorption capacity of citric acid is in the range of44.54 mg·cm-3 to 64.66 mg·cm-3 with an average value of 56.86 mg·cm-3.

1.Introduction

Citric acid(CA,2-hydroxy-1,2,3-propanetricarboxylic acid,C6H8O7)is one of the most important organic acids used massively worldwide because of its low toxicity compared with other acidilants[1].It can be used as an additive to add an acidic taste to food and soft drinks,and also as an ingredient in detergents and cleaning products[2].However,there is very limited natural supply of citric acid and the high demand for it can only be met by biotechnological fermentation processes[3].Carbohydrates,including sucrose and molasses,are usually adopted as rawmaterials to produce citric acid.The fermentation process is relatively easy to achieve a high yield,whereas it is difficult to separate citric acid from the fermentation broth.The conventional method of precipitation by calcium salt suffers from disadvantages of high cost and considerable amount of environmentally harmful waste.The other possible methods such as solvent extraction,electrodialysis,membrane separation,supercritical CO2(sc-CO2)extraction and ion-exchange[2]also have more or less inherent drawbacks.

In recent years,the incentive of employing green,sustainable technologies in industrial processes is increasing[4].Supercritical fluid chromatography(SFC),with its mobile phases generally using environmental friendly and less expensive solvent,presents strong economical and ecological advantages and gains more and more attention in re fining processes.Applications range from micro-scale analysis of complex mixtures to macro-scale purification of chiral enantiomers in a variety of industries,such as pharmaceuticals,foods,cosmetics,agrochemicals,petrochemical and natural products.The inherentspeed,efficiency,and versatility of SFC have transformed the perceptions of the technology from novelty to integral tool for the modern labs,especially for those expecting to maximize throughput[5].Patel et al.have successfully adopted packed column SFC forthe separation oftwo-pairsofwatersoluble peptides of identical mass,composition and charge that differ only in amino acid sequence[6].Desmortreux et al.have studied the separation of furocoumarins of essential oils by SFC with a green mobile phase of CO2/ethanol.The results obtained showed that SFC was a perfectly suitable method to investigate the essential oil composition because of the great number of compounds separated in a reduced analysis time(about10 min)with a very short time for re-equilibration of the system atthe end ofthe gradientanalysis[7].Based on the advantages and wide application of SFC,it is thought to be an attractive method for the separation of citric acid from the fermentation broth.

During the SFC separation process,the optimization of separating conditions in terms of production rate,yield,and costs is of much importance.In general,the knowledge of the shape of adsorption isotherms of components is necessary to understand the retention mechanism,to establish optimal conditions and to design the whole chromatographic separation process[8].Adsorption isotherms are considered like “working curves”whose knowledge is required to maximize the yield of product purification[9].In spite of the increasing importance and necessity,the knowledge of adsorption behavior under high pressures is still scarce[10].Only a few researches have been reported on adsorption isotherms from supercritical fluid.The adsorption equilibrium of some solutes,such as benzoic acid and acetylsalicylic acid[11],salicylic acid[12],ethyl benzene[13],toluene and ethyl acetate[14,15], α-tocopherol and δ-tocopherol[16,17],terpene[18,19],furfural[20],and eicosane and 1,2-hexanediol[21],from sc-CO2has been studied.However,because of the complex nature of adsorbent–adsorbate interactions,it is hard to theoretically predict adsorption isotherms.Therefore,isotherms of the supercritical fluid adsorption process have to be determined experimentally.For the operation and development of SFC adsorption processes,especially for adsorption systems with sc-CO2,the measurement of thermodynamic data and the determination of adsorption equilibrium are important.

The present work was focused on the adsorption equilibrium of citric acid from sc-CO2/ethanol with a cyano column as the stationary phase.Firstly,citric acid was separated from simulated fermentation broth(citric acid,glucose and aconitic acid)and the effect of temperature on RC–A(the Resolution of citric acid to aconitic acid)was studied.The molecular dynamic simulation(MD)method was used to predict the density of the sc-CO2/ethanol binary mobile phase at different temperatures and pressures.The in fluence of temperature,pressure and density of sc-CO2/ethanol on adsorption capacity was studied.The method of elution by characteristic point(ECP)was used to analyze the adsorption isotherm of citric acid from sc-CO2/ethanol on a ZorBaxSB-CN column.The data were modeled with the Quadratic Hill isotherm model.In our work,the preparative SFC(pre-SFC)was firstintroduced to the separation of citric acid from simulated fermentation broth and for adsorption isotherm measurements.Data from this work can be applied for engineering design and optimization of separation of citric acid with SFC.

2.Theoretical Fundamentals

2.1.Elution by characteristic points(ECP method)

Determination of adsorption isotherms can be done by the ECP method.This method is simple and fastand only a smallamountofsample is required.Itis based on the use ofa simple equation giving the rear diffusive part of an overloaded elution band.When a large amount of sample is injected into a chromatographic column packed with an adsorbent,an unsymmetrical band with a steep front and a diffuse rear profile or a steep rear and a diffuse front profile is obtained by elution.Assuming that the column efficiency is in finite and the instant adsorption equilibrium is reached between the adsorption phase and the mobile phase,the retention time(tR)at the concentration C can be expressed by the equation:

The equation is solved for d q/d C,giving the relationship[22]:

2.2.Adsorption isotherms

Adsorption is a process in which molecules from the mobile phase attach themselveson the surface ofthe stationary phase.The adsorption equilibrium can be reached during the process and the adsorption isotherm is always applied to describe this process[23].Adsorption equilibrium models including Langmuir,Freundlich and Brunner–Emmet–Teller(BET)are usually used.They are limited to model isotherms with a single curvature,either convex or concave.Isotherms exhibiting both curvatures cannot be described.For describing this type of isotherms,a moreflexible model of Hill isotherms can be used[24].

In the equation,for N=1,the isotherm is identical to the Langmuir model.For N≥2,isotherms of different curvatures with a point of inflection can be described.In this work the experimentaldata were fitted well by the Quadratic Hill model(N=2).

3.Experimental Section

3.1.Experimental set-up and materials

The experimentalset-up(Thar Company,USA)isillustrated in Fig.1.The cyano column(ZorBaxSB-CN,250 × 9.4 mm,i.d.,5 μm,purchased from Agilent,USA)was adopted for the separation of citric acid from simulated fermentation broth and also for the measurements ofadsorption isotherms underspecified conditions.The physicalproperties ofthe ZorBaxSB-CN column are listed in Table 1.The wavelength of UV detection was set to 215 nm.The sample of citric acid was filtered by a filter membrane of 0.45 μm before being injected into the chromatographic system.

Carbon dioxide with purity＞99.9%(by mass)was purchased from Tianjin Liufang Ind.,Tianjin China.Ethanol(analysis grade)and citric acid(analysis grade)were provided by Tianjin Jiangtian Chemical Technology Co.

3.2.Experimental method and procedure

3.2.1.Separation of citric acid from simulated fermentation broth

Referring to the component of the fermentation broth produced in the plant,the simulated fermentation broth whose composition is citric acid,glucose and aconitic acid was prepared.Supercritical fluid chromatography(SFC)with ZorBaxSB-CN(cyano column)as the stationary phase was adopted for the purification of citric acid from simulated fermentation broth.According to West's classification of the stationary phases,three major groups have been defined:non-polar,moderately polar and very polar[25].The cyanopropyl-bonded stationary phase(CN)is in the very polar group and is suitable for the separation of polar and medium-or small-sized molecules.The cyanopropylbonded stationary phase(CN)is able to obtain symmetric peak for acid and basic substances,and the selectivity for separation of double bond compound is satis fied.Considering the polarity and acidity ofcitric acid as wellas aconitic acid and the property ofthe cyanopropyl-bonded stationary phase,ZorBaxSB-CN(cyano column)was chosen as the stationary phase in the experiment.

3.2.2.Measurement of hold-up volume and column porosity

In preparative chromatography,the column porosity of the stationary phase is an importantparameter.Itcan be calculated from the retention volume of an unretained substance,the so-called hold-up volume.1,3,5 Tri-tert-butyl-benzene was adopted as the marker substance to quantify the hold-up volume of the cyano column.The density of the mobile phase changed from 0.7 to 0.8 g·cm-3.At these conditions,it was viewed that there is no dependency between hold-up volume and density of the mobile phase.The measured total porosity(εt)of the cyano column was about0.6.The particle porosity(εp)was 0.38.According to the relationshipεt=εb+(1-εb)εp,itcan be calculated that the bed void porosity(εb)was 0.35.

3.2.3.Calculation of the number of theoretical plates

High column efficiencies were required while the ECP method was used to determine the adsorption isotherm.The number of theoretical plates under experimental conditions has been measured.10 μl citric acid aqueous solutions of different concentrations were injected into the chromatographic system.The calculation was done according to theformula n=

Fig.1.Experimental set-up for determination of adsorption isotherms.1—CO2 cylinder,2—stop valve,3—CO2 condenser,4— flow meter,5—CO2 pump,6—modifier,7—modifier pump,8—preheater,9—sample,10—sample pump,11—injection loop,12—heat exchanger,13—column,14—auto-pressure regulator,15—pressure regulator,16—collections CS1,CS2,CS3,and CS4.

Table 1 Physical properties of ZorBaxSB-CN column

3.2.4.Calibration of C vs.S

Before data processing,the relationship of S–t(S:the detectorsignal,t:the retention time)required to be converted to that of C–t(C:the mobile phase concentration,t:the retention time).Calibration of the mobile phase concentration(C)in sc-CO2vs.the detector signal(S)was carried out by the static method[14].It was found that in the range of experimental concentration,the detector signal(S)was linear with respectto the mobile phase concentration(C),i.e.,S=k×C.k was the response factor and its value depended on temperature and density of the mobile phase.However,when temperature and density of the mobile phase changed in a small range,the value of k could be treated as keeping constant.Thus,k was measured at 313.15 K and 0.7142 g·cm-1by injecting different amounts of citric acid aqueous solution.Its value was fitted by the method of least squares.As a result,k=25.1 was calculated with an R2of 0.9990,and thus the relationship of S=25.1×C could be obtained.

3.2.5.Measurement of adsorption capacity at different conditions

The amount of ethanol added into sc-CO2was 9.7%in mass fraction.The flow rate of CO2was set to 5 g·min-1.Experiments were carried out at a temperature of 313.15 K with pressures of 13 MPa,14 MPa and 15.1 MPa,and a temperature of 321.15 K with pressures of 15.05 MPa,16.2 MPa and 18 MPa,respectively(The pressures are outlet column pressure).A 50 μl citric acid aqueous solution with a concentration of 50 mg·ml-1was injected into the chromatographic system atdifferenttemperatures and pressures.The adsorption capacity q was then calculated according to Eqs.(1)and(2).

The value of q in supercritical fluid is notonly related to temperature,but also in fluenced by the density of the mobile phase.In our experiment,the COMPASS force field of the MDmethod was adopted to calculate the density of the sc-CO2/ethanol binary mobile phase.Then,the effectoftemperature and density of the mobile phase on the adsorption capacity was studied.

4.Results and Discussion

4.1.Separation of citric acid from simulated fermentation broth

Separation of 10%(by mass)citric acid aqueous solution was firstly done with pure sc-CO2as the mobile phase.The results showed that there was not elution peak of citric acid in the chromatogram.This was mainly due to the fact that the polar citric acid molecules were nearly non-soluble in non-polar CO2molecules so that the citric acid molecules were adsorbed to the cyano column and could not be eluted by sc-CO2.The mobile phase combining sc-CO2with modifiers can extend the utility of SFC to polar and ionic compounds.The addition of modifiers to sc-CO2acted on the retention and separation of solutes was mainly by modifications of the polarity of the fluid,and therefore of the eluting strength and deactivation of the stationary phase.Thus,methanol and ethanol were added to sc-CO2respectively to separate citric acid from simulated fermentation broth.The chromatograms are shown in Fig.2(a)and(b).Since the solubility and eluting strength of the mobile phase are dramatically enhanced,citric acid could be eluted and separated with glucose and aconitic acid from simulated fermentation broth successfully.

It can be calculated from Fig.2(a)and(b)that the resolution of citric acid to aconitic acid(RCA)is 1.39 and 2.02,respectively.It demonstrates that the separating effect with ethanol as modifier is better than methanol.Moreover,considering that citric acid is mainly used in the food and medicine industries,whereas methanol is toxic and harmful to human health,sc-CO2/ethanol is chosen as the mobile phase in all the following studies.

Fig.2.Chromatogram of simulated fermentation broth in SFC:(a)methanol content of 10%in mobile phase；and(b)ethanol content of 9.7%in mobile phase.

The effectoftemperature on RCAhas also been studied.Experimental data are illustrated in Fig.3.The results indicate that RCAdecreases with increasing temperature.When the temperature≥321.15 K and the pressure ≤ 16 MPa,the value of RC–A≤ 1.5.(In the chromatographic separation process,R=1.5 is viewed as the symbolof complete separation for the neighboring chromatographic peak.)In order to guarantee complete separation of citric acid,it is better to selectan operating temperature lowerthan 321.15 K on the condition ofrelative low pressures.

Fig.3.Effect of temperature on R C–A.

4.2.Determination of adsorption isotherms by the ECP method

The adsorption process is related to the temperature and density of the mobile phase.In our study,density of the sc-CO2/ethanol mobile phase was predicted by the COMPASS force field of Material studio(MS)software in different temperatures and pressures before isotherm analysis[26].Atypicalsystem contained 282 molecules with 26 ethanol and 256 CO2molecules.The molecular dynamic equation was solved by the fifth-order Gear prediction calibration algorithm.The simulation results are summarized in Table 2.Densities of each group are similar.These three groups of similar densities could be viewed as constant when studying the effect of temperature on adsorption capacity.

Table 2 Calculated value ρ of sc-CO2/ethanol for different temperatures and pressures

The column efficiency was calculated under experimental conditions.When a sample of 10 μl of citric acid solution was injected into the chromatographic system,a peak of Gaussian distribution was obtained.According to the formula in Subsection 3.2.3,the number of theoretical stages was calculated.The results indicated that the number of theoretical stages of the column for all experiments was higher than 3000,partly higher than 5000.This ensured thatthe instantequilibrium modelfor the ECP method was valid and thatthe ECP method was accurate to calculate the diffuse profiles[27,28].

Typical diffuse profiles of the chromatogram at 313.15 K and 321.15 K are shown in Fig.4.The big peak is assigned to be a citric acid peak because the absorbance of citric acid is maximal at its maximumabsorption wavelength(λmax=215 nm).The smallpeak between 2 and 3 min may be a pressure peak due to pressure impulse during sample injection.The citric acid elution peaks indicate that the rear of the chromatogram is almost vertical to the abscissa,which is a typical result of favorable adsorption isotherm under the instant equilibrium condition.The adsorption characters do not conform to Langmuir adsorption type.By analyzing the chromatogram,the equilibrium adsorption capacity q at different temperatures and densities of the mobile phase is calculated according to Eqs.(1)and(2).

The effect of temperature at constantdensity on adsorption capacity is plotted in Fig.5 and the effect of densities is plotted Fig.6.

Experimental data in Fig.5 demonstrates that the adsorption capacity q of citric acid decreases with increasing temperature at similar densities of the mobile phase,suggesting that the adsorption of citric acid on cyano column has an exothermic nature.This trend is conformed to the basic principle of thermodynamics that low temperature facilitates adsorption.In the case of citric acid adsorption on a cyano column,the molecular interaction of citric acid with the propyl cyano silica stationary phase includes the electron donor–acceptor interaction,the π–π interaction and the hydrogen bond interaction.The electron donor–acceptor interaction exists between lone pair electrons of nitrogen in CN-and electropositive carbon atomofcarbonylin citric acid.The π–π interaction exists between the π electron of sp hybridized orbital in CN-andπelectron ofsp2hybridized orbitalin carbonylofcitric acid.The hydrogen bond interaction exists as O–H…O between citric acid and the hydroxyl group of silica matrix.The increased temperature causes an enhanced molecular motion,which results in the decrease in the interaction of these groups.As a result,the adsorption capacity of citric acid decreases.In addition,with increasing temperature,the interaction between molecules of citric acid and the mobile phase increases.Thus,the enhanced eluting power of the mobile phase leads to the reduction of adsorption capacity.

Fig.4.Chromatograms ofoverloading ofcitric acid atdifferent temperatures and densities of mobile phase:(a)313.15 K；and(b)321.15 K.

In comparison to adsorption from ordinary liquid solutions,the adsorption capacity from supercritical fluid depends on not only the temperature but also the density of the mobile phase.Fig.6 illustrates the adsorption data at fixed temperatures.It can be seen from Fig.6 that the adsorption capacity q ofcitric acid decreases with increasing density of the sc-CO2/ethanol mobile phase.Ethanol added to the mobile phase as modifier can block active sites of hydroxyl group on the stationary phase by the formation of hydrogen bond.Since a significant portion of hydroxyl sites is occupied by ethanol molecules,the adsorption of citric acid has been inhibited.Meanwhile,the addition of ethanol as modifier can promote the dissolution of citric acid in the mobile phase and enhance the eluting powerofthe mobile phase.In addition,ethanol may also suppress the ionization of citric acid because the solvating power of ethanol is weaker in comparison to water.When adding ethanol to the mobile phase,the solvating power of the solution may be weakened.The increased density of the mobile phase at fixed temperatures results from the increase of pressure.Parcher had concluded that CO2and polar modifiers,such as methanol and ethanol,adsorbed to a considerable extent onto the SFC stationary phase,and the amounts of CO2and modifier adsorbed decreased with increasing pressure at the supercritical region[29,30].With increasing pressure,the adsorbed ethanol on the stationary phase decreased while the ethanol in the mobile phase increased.Therefore,the deactivation of the stationary phase is weakened and meanwhile the eluting power of the mobile phase is intensified.As a result,the adsorption capacity q of citric acid decreases significantly with increasing density.

Fig.5.Adsorption isotherm ofcitric acid atdifferent temperatures fitted by Quadratic Hillisotherm:(a)0.71 g·cm-3；(b)0.77 g·cm-3；and(c)0.80 g·cm-3.

Fig.6.Adsorption isotherm of citric acid at different densities of mobile phase fitted by Quadratic Hill-isotherm:(a)313.15 K；and(b)321.15 K.

Both Figs.5 and 6 indicate that the adsorption capacity q increases with increasing concentration of citric acid.The increase is less at low concentration than athigh one,suggesting thatthe in fluence oftemperature and density of the mobile phase on the adsorption process becomes more obvious with increasing concentration of citric acid in the sc-CO2/ethanol binary mobile phase.

4.3.Modeling of the adsorption data

The diffuse profiles of the chromatogram in Fig.4 show an obvious diffusive part in the front of the elution band.It is the typical character of an anti-Langmuir behavior.The Hill isotherm model can be applied to represent this type of adsorption data.This model is obtained from statistical thermodynamics and the most usable models are the Quadratic-Hill isotherm and Cubic-Hill isotherm for the fitting of adsorption data.In this paper,as the Quadratic-Hill isotherm model can give significantly better results for citric acid,it is chosen to describe the adsorption equilibrium data.The adsorption data are fitted by the method of least squares.Isotherm parameters and correlation coef ficients are shown in Table 3.The fitted curves of Quadratic-Hill isotherm are illustrated in Figs.5 and 6.

From the fitting curves in Figs.5 and 6,it is apparent that the Quadratic-Hill isotherm model achieves a suitable fit of the adsorption data.However,it has to be pointed out that these fitting curves arenot in good agreement with the experimental data at low concentrations.When the concentration of citric acid(CCA)is lower than 0.2 mg·ml-1,the data points give considerable deviation from the fitting curves.The lower the concentration is,the larger the deviation willbe.The reason maybe thatthe data points close to the top ofthe elution profile are more easily affected by the source ofband broadening so that these points are not proper for the determination of the isotherm.However,it is not possible to reject these data points close to the baseline since the integration of the profile has to be made from concentration 0 to C.The deviation is the systematic error and the error is also a drawback of the ECP method due to the use of inaccurate lowconcentration data for the determination of high-concentration points of the isotherm[31].

Table 3 Quadratic Hill-isotherm parameters for citric acid at different conditions

The fitting parameters in Table 3 show that the correlation coef ficients R2are＞0.9993 under all conditions.The monolayer adsorption capacity(qS)of citric acid is in the range of 44.54 to 64.66 mg·cm-3with an average value of 56.86 mg·cm-3.The values of qSfluctuate in a smallrange.Butto some extent,the adsorption capacity is almostconstant under different temperatures and densities of the mobile phase.This is because the monolayer adsorption capacity depends mainly on the character of the adsorbent and the way the adsorbed molecules are packed on the adsorbentsurface.Thus,itwould notchange with adsorption conditions.

5.Conclusions

Citric acid was separated from simulated fermentation broth by SFC successfully and the effectof temperature on RCAwas studied.Then,the adsorption isotherms of citric acid on a cyano column from supercritical carbon dioxide(sc-CO2)modified with 9.7%(by mass)ethanol were measured using the elution by characteristic point(ECP)method.Molecular dynamic simulation(MD)was carried out to predict the density of the sc-CO2/ethanol binary mobile phase with the COMPASS force field.Citric acid adsorbing experiments were performed at two temperatures(313.15 K and 321.15 K)and six sc-CO2/ethanol densities(0.7142 g·cm-3,0.7068 g·cm-3,0.7665 g·cm-3,0.7721 g·cm-3,0.8019 g·cm-3and 0.7958 g·cm-3).Adsorption capacity q decreased with increasing temperature,which conformed to the basic principle of thermodynamics that low temperature was favorable for adsorption.In addition,adsorption capacity q also decreased with increasing density of the sc-CO2/ethanol mobile phase,which was related to the enhanced eluting power of the mobile phase.

The equilibrium isotherm data were fitted well by the Quadratic Hill adsorption isotherm model with correlation coefficients of R2＞0.9993.The obtained model showed an anti-Langmuir behavior with an increasing isothermslope athigherconcentrations.The average monolayer saturation adsorption capacity qSof citric acid was 56.86 mg·cm-3.Data from this work have been described by an appropriate mathematical model.The obtained model can be applied for the simulation of chromatographic processes to ease the choices of suitable process parameter as well as for engineering design and optimization of recovery and purification of citric acid with SFC.

Nomenclature

bathe numerical coefficient

bbthe numerical coefficient

C the mobile phase concentration,mg·ml-1

n the number of theoretical plates

q(C) the equilibrium capacity when the fluid concentration is C,mg·cm-3

qSthe monolayer saturation adsorption capacity of the adsorbent,mg·cm-3

RCAthe Resolution of citric acid to aconitic acid

S the detector signal

t0the hold-up time,s

tPthe injected time,s

tRthe retention time,s

V the retention volume ofthe characteristic pointofdiffuse profile at concentration C,cm3

V0the hold-up volume of the column,cm3

Vathe volume of adsorbent in the column,cm3

Y1/2the peak width at half-height,s

ε the porosity

εbthe bed void porosity

εpthe particle porosity

εtthe measured total porosity of the cyano column