Bioresource Technology 143 (2013) 360–368

Contents lists available at SciVerse ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Modeling of breakthrough curves of single and quaternary mixturesof ethanol, glucose, glycerol and acetic acid adsorption onto amicroporous hyper-cross-linked resin

0960-8524/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.biortech.2013.06.009

⇑ Corresponding author at: College of Biotechnology and Pharmaceutical Engi-neering, Nanjing University of Technology, Nanjing 210009, PR China. Tel.: +86 02586990001; fax: +86 025 86990001.

E-mail address: yinghanjie@njut.edu.cn (H. Ying).

Jingwei Zhou a,b,c, Jinglan Wu a,b,c, Yanan Liu b,c, Fengxia Zou b,c, Jian Wu b,c, Kechun Li b,c, Yong Chen a,b,c,Jingjing Xie a,b,c, Hanjie Ying a,b,c,⇑a State Key Laboratory of Materials-Oriented Chemical Engineering, Nanjing 210009, PR Chinab College of Biotechnology and Pharmaceutical Engineering, Nanjing University of Technology, Nanjing 210009, PR Chinac National Engineering Technique Research Center for Biotechnology, Nanjing 211816, PR China

h i g h l i g h t s

� Adsorption of ethanol from actualfermentation broth in packed bedswas conducted.� HD-01 resin shows large adsorption

capacity and high selectivity forethanol.� Effect of pH on the adsorption of

quaternary mixtures was investigatedin detail.� Multicomponent mass transport

model was proposed and validated byexperiment data.� Satisfactory fit of breakthrough

curves in single/multi-componentsystems.

g r a p h i c a l a b s t r a c t

a r t i c l e i n f o

Article history:Received 5 April 2013Received in revised form 31 May 2013Accepted 3 June 2013Available online 13 June 2013

Keywords:EthanolMulti-componentMicroporous resinCompetitive adsorptionModeling

a b s t r a c t

The adsorption of quaternary mixtures of ethanol/glycerol/glucose/acetic acid onto a microporous hyper-cross-linked resin HD-01 was studied in fixed beds. A mass transport model based on film solid lineardriving force and the competitive Langmuir isotherm equation for the equilibrium relationship was usedto develop theoretical fixed bed breakthrough curves. It was observed that the outlet concentration ofglucose and glycerol exceeded the inlet concentration (c/c0 > 1), which is an evidence of competitiveadsorption. This phenomenon can be explained by the displacement of glucose and glycerol by ethanolmolecules, owing to more intensive interactions with the resin surface. The model proposed was vali-dated using experimental data and can be capable of foresee reasonably the breakthrough curve of spe-cific component under different operating conditions. The results show that HD-01 is a promisingadsorbent for recovery of ethanol from the fermentation broth due to its large capacity, high selectivity,and rapid adsorption rate.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Ethanol produced from lingo-cellulosic biomass has been recog-nized as one of the most important second-generation liquid bio-fuels (Nigam and Singh, 2011). However, ethanol fermentation isa product inhibition process, which leads to low concentrations

Table 1Physicochemical properties of the microporous resin HD-01.

Property

Matrix Poly(styrene-co-divinylbenzene)Appearance Brown–red, translucentPolarity Weak polarBET surface area (m2/g) 1645.5Particle size (mm) 0.42–0.56Moisture capacity (%) 43.0HK desorption average pore diameter (nm) 1.92Total pore volume (cm3 g�1) 0.732

J. Zhou et al. / Bioresource Technology 143 (2013) 360–368 361

of ethanol in the final fermentation broth. The effect of feedbackinhibition can be circumvented through integrated bioreactor de-signs employing in situ product recovery (ISPR) methods, e.g., fer-mentation coupled with gas stripping (Hashi et al., 2010),adsorption (Jones et al., 2011) or pervaporation (Chen et al.,2012) and so on. Among proposed technologies, ethanol fermenta-tion coupled with adsorption is one of the most promising meth-ods (Vane, 2008). It was found that direct addition ofhydrophobic adsorbents to the fermentation flask promoted theglucose consumption and ethanol production rates of strains(Einicke et al., 1991; Ikegamai et al., 2000; Jones et al., 2011). How-ever, the spent sorbents in fermentation broth are difficult to re-cover. Moreover, the fermentation is hard to operate in thecontinuous mode. Accordingly, Yang and Tsao (1995) proposed aprocess to integrate the repeated fed-batch fermentation with con-tinuous product removal by adsorption with two externally lo-cated adsorption columns. The separation process can be madecyclical by alternating between adsorption and desorption withthe two columns.

Adsorption in a fixed bed is more preferable in industrial pro-duction because of its ability to process large quantities of feedin the continuous mode. To date, many studies have been con-ducted to concentrate ethanol from aqueous solution on a fixedbed (Delgado et al., 2013; Jones et al., 2010; Kawabata et al.,1988; Pitt et al., 1983). However, most of them used ethanol–waterbinary aqueous to simulate fermentation broth, while the effect ofother main components co-existing in the fermentation broth areneglected. In addition to ethanol, by-products (e.g., glycerol, andorganic acids) and substrate sugars co-exist in fermentation broth.These impurities may have negative effect on the adsorption ofethanol. Bui et al. (1985) found that glucose could compete withethanol for the same adsorption sites on the surface of activatedcarbon thus reduce the ethanol adsorption capacity. Bowen andVane (2006) observed that the presence of small amounts of aceticacid (<0.6 wt%) significantly reduced the ethanol adsorption onzeolite ZSM-5. These studies were conducted with the binary sys-tem (glucose/ethanol, acetic acid/ethanol) in a batch adsorber. Re-ports on the adsorption of ethanol from the quaternary systems(ethanol/glycerol/glucose/acetic acid) and actual fermentationbroth in a packed bed are few and fragmentary.

Therefore, this work is focused on the adsorption of ethanolfrom quaternary systems in a fixed bed. In our previous study, amicroporous hyper-cross-linked resin HD-01 with large adsorptioncapacity and rapid kinetics for ethanol from aqueous solution wasscreened. As a continuation of our previous works, the main aimsof the present study are:

(1) Evaluate the adsorption selectivity for ethanol from actualfermentation broth on HD-01 resin.

(2) Establish a mass transport model to predict the competitivebreakthrough curves of ethanol, glucose, glycerol and aceticacid in a HD-01 resin packed bed.

(3) Investigate the effects of operating conditions, includingflow rate, bed length, initial concentration of the solutesand pH on the performance of HD-01 bed.

2. Methods

2.1. Materials

2.1.1. ChemicalsEthanol, acetic acid, glucose and glycerol were supplied by Sin-

opharm Chemical Reagent Co., Ltd. (Shanghai, China). All the chem-icals used were analytical grade reagents, and distilled de-ionizedwater was used in the preparation of all solutions.

2.1.2. ResinThe microporous hyper-cross-linked polymeric resin HD-01

was kindly provided by the National Engineering Technique Re-search Center for Biotechnology (Nanjing, China). The main phys-ico-chemical properties of HD-01 were listed in Table 1.

2.1.3. Fermentation broth preparationThe detailed ethanol fermentation process could be found in S1

(Supplementary materials). Prior to use, the fermentation brothwas centrifuged at 8000 rpm and 15 min to remove the yeast cell.After about 40 h of culture, the fermentation broth contained90.0 ± 5.0 g/L ethanol and 9.0 ± 2 g/L glucose. In addition, the fer-mentation broth contained 9.0 ± 1.0 g/L glycerol and 1.5 ± 0.5 g/Lacetic acid as by-products.

2.2. Experimental methods

2.2.1. The static equilibrium adsorption experimentsBatch uptake experiments were performed to study the single/

multi-component equilibrium adsorption isotherms of ethanol (orother solutes of interest) from single/multi-component systemsonto resins at the temperature of 298 K. The mass ratios of etha-nol:glycerol:glucose:acetic acid in quaternary mixture solutionwas equal to 100:10:10:5, which was close to the ratio of thesecomponents in actual final fermentation broth. The solution withdifferent initial concentrations (50 mL) were added to erlenmeyerflasks (100 mL) containing 5 g wet resin. After that, the flasks weresealed and shaken at 200 rpm in a shaker for 8 h at 298 K to attainthe equilibrium. Then a sample was withdrawn from the superna-tant fluid with a syringe and the concentration of solute of interestwas measured by HPLC.

The equilibrium adsorption capacity of the individual soluteonto resin was determined by the following relationship:

qi;e ¼ðci;0 � ci;eÞV

mð1Þ

where ci,0 and ci,e represent the initial and equilibrium aqueous con-centration of ethanol (or other solutes of interest) (mmol/L), respec-tively. V is the volume of the aqueous solution (L); m is the mass ofthe resin used (kg).

2.2.2. Dynamic column adsorption experimentsDynamic adsorption experiments were carried out in a water

jacketed glass column with a diameter of 2.05 cm and a length of30 cm. The column was packed with wet resin of given mass (20,30, or 40 g). The void faction of the bed (eb) was 0.27, which wascalculated according to (Kleinübing et al., 2012). By doing this, abed height around 7.6–15.2 cm was obtained. A peristaltic pump(BT300, longer pump, China) was used to pump the solution down-wards through the column at a given flow rate (0.2–1.2 mL/min).By using an automatic fraction collector (BSZ-100, Qingpu, Shang-hai), samples of approximately 5 mL were collected at differenttimes. The concentration of ethanol (or other components of inter-

362 J. Zhou et al. / Bioresource Technology 143 (2013) 360–368

est) was measured via HPLC thus the breakthrough curves was ob-tained. The adsorption capacities in dynamic experiments weredetermined from the mass balance as follows:

qs ¼ðcf ðVf � ebVbÞ �

PciðtÞViðtÞÞ

mð2Þ

where qs represents the dynamic adsorption capacity (mmol/kg), cf

and ci(t) represent the inlet and outlet concentrations of ethanol (orother solutes of interest), respectively. Vf is the volume of the feed,Vb is the volume of the packed bed, Vi(t) is the outlet volume of col-lection in the cuvette. The breakthrough point was used to compareand evaluate the adsorption performance of the HD-01 bed undervarious conditions. The breakthrough point (tb) was defined as thetime when the effluent concentration, ci, is about 5% of the influentconcentration, cf.

2.3. Analytical method

The concentrations of ethanol, glucose, glycerol, and acetic acidin the aqueous phase were measured by HPLC (1200 series, AgilentTechnologies, USA) equipped with a refractive index detector usingan Aminex HPX-87H ion exclusion column (Bio-Rad Laboratories,USA). The mobile phase was 5 mM H2SO4 with a flow rate of0.6 mL/min at 298 K. External standards were used for calibration.

2.4. The model calculation

2.4.1. Adsorption equilibriumIn the single-component system, the experimental equilibrium

data of each individual solute obtained were fitted by the Langmuirisotherm model which is represented as follows:

qi;e ¼qi;mKici;e

1þ Kici;eð3Þ

where qm is the Langmuir isotherm constant (mmol/kg), and K isequilibrium coefficient (L/mmol).

In the multi-component systems, the adsorption capacity ofeach solute at equilibrium depends on the concentration of othercomponents present locally. The competitive Langmuir isothermmodel was used to describe the adsorption equilibrium of each sol-ute as follows (Guiochon et al., 2006):

qi ¼qm;i

Kici

1þPN

j Kjcj

ð4Þ

2.4.2. Fixed bed modelThe film solid linear driving force (LDF) model was used in this

work to simulate the adsorption behavior of specific component inthe fixed bed due to its simplicity. The model comprised a massbalance equation, adsorption isotherm, and a linear rate equation.The assumptions of this model are as follows:(1) Fluid density andviscosity are constant; (2) isothermal operation; (3) constant fluidflow rate; (4) radial distributions are negligible; (5) mass transferrate in the adsorption process is described by linear driving forcerate equation.

The differential mass balance in a volume element for species iin the liquid phase can be expressed by Schmidt-Traub (2005):

v @ci

@zþ @ci

@tþ q

1� eb

eb

@qi

@t¼ DL

@2ci

@z2 ð5Þ

where v is the interstitial velocity (m/s), v ¼ Q=ðAebÞ, A is the cross-sectional area of column (m2), Q is the volumetric flow rate (m3/s), zis the axial position (m), t is the time (s), q is the density of adsor-bent (kg/m3), DL (m2/s) is the axial dispersion coefficient and can be

calculated from the following empirical correlation (Suzuki andSmith, 1972):

DL ¼ 0:44Dm þ 0:83udp ð6Þ

where dp is the particle diameter (m), u is the superficial velocity(m/s), Dm (m2/s) is the molecular coefficient and could be calculatedby the Wilke–Chang equation (Wilke and Chang, 1955):

Dm ¼7:4� 10�8ð/MBÞ1=2T

gBV0:6A

ð7Þ

The first term in Eq. (6) represents the molecular diffusion in theinterparticle voids, and the second term in Eq. (6) represents theeddy diffusion. Both of them contribute to axial dispersion.Theboundary conditions are written as was suggested by Danckwerts(1953):

z ¼ 0; DL@ci

@zjz¼0 ¼ vðcijz¼0 � cfe;iÞ ð8Þ

z ¼ L;@ci

@zjz¼L ¼ 0 ð9Þ

Initial conditions are:

t ¼ 0; qi ¼ 0 ð10Þ

t ¼ 0; ci ¼ 0 ð11Þ

The mass transfer rate is represented by a linear driving force (LDF)equation according to Glueckauf (1955):

@qi

@t¼ keff ðqi;e � qiÞ ð12Þ

where keff is the effective mass transfer coefficient (s�1).Adsorption equilibrium isotherm of species i in single/multi-

component systems can be described by the Langmuir isotherm(Eq. (3)) or the competitive Langmuir isotherm model (Eq. (4)).

This mathematical model was solved by the commercial soft-ware MATLAB 2010a. The partial differential equation was discret-ized with respect to the space coordinate by the method oforthogonal collocation method on finite elements (OCFE) with 20finite elements and three interior collocation points in each bedvolume element. The parameters of the model (keff) were deter-mined by minimizing the objective function given as follows:

FðpÞ ¼Xn

i

XN

i

ci;preðtj; LÞ � ci;expðtjÞci;expðtjÞ

� �2

ð13Þ

where n is the number of solutes and N is the number of experimen-tal data; L is the length of the bed (m); ci,pre is the concentration attime tj in the outlet flow calculated by the model and ci,exp is theexperimental concentration at time tj.

3. Results and discussion

3.1. Adsorption equilibrium studies

For the sake of comparison, the single-component adsorptionisotherms of ethanol, glycerol, glucose and acetic acid on HD-01 re-sin at 298 K along with the experimental equilibrium data in qua-ternary systems were shown in Fig. 1. It is clear that the relativeadsorption affinity of four compounds with resin HD-01 followsthe order: acetic acid > ethanol > glycerol > glucose, which matchesthe reversed order of polarity strength of solute molecule: glucose(11 ± 3 D) (Alfimova et al., 1975) > glycerol (2.56 D) (Lide and Bru-no, 2002) > acetic acid (1.70 ± 0.03 D)(Lide and Bruno, 2002) > eth-anol (1.69 ± 0.03 D) (Lide and Bruno, 2002). It should be noticedthat acetic acid shows stronger adsorption affinity with HD-01 in

Fig. 1. The adsorption isotherms of ethanol, acetic acid, glycerol and glucose ontoHD-01 resin in single- and multi-component systems at the temperature of 298 K.The dash lines represent the single-component isotherms fitted with Langmuirequation, the solid lines represent the multi-component isotherms fitted withcompetitive Langmuir isotherm equation.

Fig. 2. (a) The breakthrough curves of ethanol, acetic acid, glycerol and glucose insingle-component system; (b) the breakthrough curves of ethanol, acetic acid,glycerol and glucose in multi-component systems at solution pH = 3.0.

J. Zhou et al. / Bioresource Technology 143 (2013) 360–368 363

contrast to ethanol due to the extra p�p the carbonyl of acetic acidand the phenyl ring of the polymer matrix (Zhai et al., 2003). Theresults indicate that the adsorption mechanism of these solutesonto non-ionic polymeric resin HD-01 is mainly by hydrophobicinteractions (Nielsen et al., 2010).

To determine the effect of the presence of one adsorbate on theadsorption characteristics of the other, the adsorption equilibriumof ethanol, glycerol, glucose and acetic acid in multi-componentsystems were also conducted. It can be seen from Fig. 1, theadsorption of ethanol in quaternary systems is slight lower thanthat in the single-component system, which indicates that compe-tition adsorption by other solutes occurs. At the same time, thepresence of ethanol in the multi-component solutions greatly re-duced the adsorption of acetic acid, glucose and glycerol on HD-01 due to the competitive effect. This result is consistent withthe report described by Bui et al. (1985), who investigated thecompetitive adsorption behavior of ethanol and glucose onto acti-vated carbon. The experimental data is predicted by the competi-tive Langmuir isotherm equation (Eq. (4)) well (the inset showsthe data at the low concentration range). The corresponding iso-therm parameters are listed in Table 2.

3.2. The column adsorption

3.2.1. The single component column dynamic adsorption experimentsBreakthrough curves for each component were obtained by the

frontal analysis method on preparative column packed with HD-01resin. For the sake of comparison, the breakthrough curve was rep-resented as the effluent/influent concentration ratio (c/c0) versustime/volume. Fig. 2(a) shows the experimental and the simulatedconcentration histories for ethanol, glycerol, glucose and aceticacid in the single-component system at 298 K. It can be observed

Table 2The isotherm parameters for solutes adsorption on resin HD-01.

Solute Single system

qm (mmol/kg) K R2

Ethanol 6945.5 0.00068 0.9Glycerol 2692.7 0.00043 0.9Glucose 1279.5 0.00041 0.9Acetic acid 6851.7 0.00124 0.9

a D %, the average absolute percent deviations. %D ¼ 1N

PN1

qexp�qpredqexp

��� ���� 100%;

that the column model fitted the experimental data very well(Fig. 2a). It is clear that the effluent orders are as follows: glu-cose > glycerol > ethanol > acetic acid, which is contributed to boththe initial concentration and the adsorption isotherm (adsorptionaffinity). The adsorption affinity of glucose and glycerol on HD-01 is weak. Thus the packed bed was soon saturated with these sol-utes. Acetic acid shows the strongest adsorption affinity with HD-01, which is consistent with the results obtained from equilibriumdata (Fig. 1). The experimental conditions used in the model forbreakthrough curves prediction are depicted in Table 3. The axialdispersion coefficient DL of the four solutes was found to be inthe range 1.09 � 10�8 to 1.11 � 10�8 m2/s (Eq. (6)). Mass transferkinetics is described by parameter keff (Eq. (12)), which is estimatedby the best fit procedure of the sets of breakthrough data in whichthe objective function F (p) has been minimized (Eq. (13)). The

Quaternary system

qm (mmol/kg) K D%a

986 6705.6 0.00075 1.60996 2680.2 0.00039 9.15956 1021.5 0.00058 7.65949 6827.0 0.00110 8.16

Table 3Summary of operation conditions for the adsorption experiments in packed column and the corresponding results.

Exp. No. Solute c0 (mmol/L) tb (min) qs (mmol/kg) Keff (min�1) DL (�108 m2/s) F (P)

Single-component system1 Glucose 56.1 11.9 29.1 0.25 1.09 0.592 Glycerol 110.8 28.6 122.4 0.50 1.10 0.383 Ethanol 108.5 128.0 466.7 0.62 1.11 0.214 Ethanol 542.7 112.1 1929.7 0.62 0.315 Ethanol 1085.3 93.5 2999.8 0.62 0.256 Ethanol 2177.1 66.7 4074.2 0.62 0.377 Acetic acid 81.6 203.6 575.5 0.56 1.11 0.67

Multi-component systems8 Glucose 56.6 12.3 11.8 0.20 1.09 1.21

Glycerol 108.6 22.6 37.7 0.37 1.11Ethanol 2155.4 63.3 3718.0 0.60 1.11Acetic acid 83.4 72 215.0 0.39 1.11

15a Glucose 59.4 11.6 8.9 0.20 1.09 2.10Glycerol 93.0 22.8 43.4 0.37 1.10Ethanol 1886.5 68.8 3631.4 0.60 1.11Acetic acid 16.7 11.5 3.3 0.22 1.11

⁄ The ratio of height over diameter of HD-01 packed bed (H/D) was 3.7, the flow rate was 1.5 BV/h.a The adsorption of ethanol and other solutes with actual fermentation broth.

364 J. Zhou et al. / Bioresource Technology 143 (2013) 360–368

mass transfer rate for the four solutes on HD-01 resin is quite fastas one can conclude from the estimated values of the mass transfercoefficient (keff). The mass transfer rate of ethanol is almost thesame as acetic acid but higher than glycerol and glucose (Table 3).

To evaluate the performance of HD-01 packed column for sep-aration of ethanol from other solutes and to investigate the effectof initial concentrations on the breakthrough curves, the dynamicadsorption of ethanol at different initial concentrations were con-ducted. It can be observed that the initial concentrations have asignificant impact on the ethanol breakthrough curves (Fig. 2a).First, the shape of ethanol breakthrough curve becomes sharperwhen the initial concentration increasing. This is because the con-centration gradient between the surface of HD-01 resin and thesolution is larger at higher initial concentration, which improvesthe driving force for mass transfer (Kleinubing et al., 2011). Themass transfer zone is short within the bed and results in a sharperbreakthrough profile. Second, the breakthrough point (tb) of etha-nol decreased with increasing the initial concentrations. Accordingto Guiochon et al. (2006), the retention time can be given by thebasic equation of chromatography:

tR;iðcþi Þ ¼ t0 1þ 1� eb

eb

@qi

@ci

����Cþi

!ð14Þ

where t0 and tR represent dead time of the column and retentiontime, respectively.

The differential term in Eq. (14) represents the slope of the iso-therm at the concentration ci, which is varied with concentrationfor a non-linear isotherm. The isotherm of ethanol is convex iso-therm, and thus the retention time decreased with increasing con-centration (Fig. 2a). It can be observed that the separation betweenethanol and other compounds on HD-01 is fairly good in the lowconcentration range (108.5–1085.3 mM). But the separation be-tween ethanol and glucose, glycerol becomes lower at high initialethanol concentration due to the decreasing breakthrough point(tb).

3.2.2. The dynamic adsorption experiments in multi-componentsystems

The breakthrough curves for ethanol and other solutes adsorp-tion from quaternary mixtures are plotted in Fig. 2b. In contrastwith the single-component breakthrough curves, it was observedthat the outlet concentration of glucose and glycerol exceed the in-let concentration (c/c0 > 1), which is an evidence of competitive

adsorption. The plots in Fig. 2b reveal: (i) glycerol exhibits thehighest overshoot value in the mixtures; (ii) the shape of thebreakthrough curve of ethanol is almost the same as the single-component system; (iii) acetic acid do not overshoot in thequaternary mixtures, but the breakthrough point (tb) of acetic acidis greatly decreased from 203.6 min to 72 min, which is close tothat of ethanol. The observed overshoot is in agreement with theadsorption affinity for the solutes on HD-01 discussed in Section3.1. The adsorbate with the lower adsorption affinity (glucoseand glycerol) is displaced by the adsorbate with higher adsorptionaffinity (ethanol and acetic acid) (Escudero et al., 2013). Acetic acidexhibits the highest affinity for the adsorbent and is not overshootby other compounds.

The overshoot of breakthrough curves could be predicted wellby the column model incorporating the competitive Langmuir iso-therm equation. The obtained model parameters are presented inTable 3. The effective mass transfer coefficient keff for the four sol-utes were slightly lower in the quaternary system. But it is ob-served that the shape of the breakthrough curves of glycerol andglucose were shaper in the multi-component systems due to thedisplacement effect by ethanol and acetic acid (Fig. 2b). The shar-per breakthrough curves of glycerol and glucose are favor to sepa-rate ethanol from these two solutes. But the breakthrough files ofacetic acid are even smoother in the multi-component systemsthan in single component system. The drag breakthrough curvesof acetic acid and the close tb with ethanol indicate that it is diffi-cult to separate ethanol from acetic acid from quaternary mixture.

3.3. Model validation and effect of operating conditions on thebreakthrough curves in multi-component systems

In order to check the model capability, model predictions havebeen compared with the experimental data for quaternary mix-tures adsorption on HD-01 resin at various conditions using theoptimum effective diffusion coefficients keff (Table 3).

3.3.1. Effect of influent flow rate on the breakthrough curvesThe breakthrough curves as a function of effluent volume are

shown in Fig. 3. The detailed operating conditions and the perfor-mance parameters of packed column for the adsorption of ethanoland other solutes are listed in Table A.2 (Supplementary materials).It can be seen from Fig. 3, (i) the higher the flow rate, the lower thebreakthrough point (tb) is, while the amount of solutes adsorbed onHD-01 resin are not obviously affected by the flow rate. This result

Fig. 3. Effect of influent flow rate on the breakthrough curves of glucose (a), glycerol (b), ethanol (c) and acetic acid (d) on HD-01 in multi-component systems as a function ofvolume.

J. Zhou et al. / Bioresource Technology 143 (2013) 360–368 365

is reasonable because the adsorption is determined by the equilib-rium isotherm. (ii) The breakthrough of ethanol and acetic acid be-came more smooth at higher flow rate, thus the overshoot ofglycerol and glucose are slight lower (Fig. 3a and b). These phe-nomenons could be explained by the increased axial diffusion coef-ficient (DL) when the flow rate increasing (Table A.2). Both ofmolecular diffusion in the interparticle voids and eddy diffusioncontribute to axial dispersion (Eq. (6)). The effect of eddy diffusionbecomes significant at high flow rate.

3.3.2. Effect of bed length on the breakthrough curvesThe effect of ratio of the height over the inner diameter (H/D) on

the breakthrough performance of HD-01 was studied at a flow rateof 0.6 mL/min and a constant feed concentration. The break-through curves are shown in Fig. A1 (Supplementary materials).The experimental data at three different H/D ratios were fittedby the mass transport model with the same effective mass transfercoefficient (keff) fairly well. It is observed that the retention time in-creased with H/D ratio, whereas the loading capacities were almostidentical (Table 4). The separation degree (resolution) betweenethanol and other solutes was increased by increasing H/D ratio.But the breakthrough points of ethanol and acetic acid were closeto each other at three different values of H/D, which indicates thatit is difficult to separate them by increasing H/D ratio.

3.3.3. Effect of initial concentration on the breakthrough curvesThe breakthrough curves as a function of contact time at differ-

ent initial concentrations are depicted in Fig. 4. An interesting phe-

nomenon can be observed that the retention time of the ethanoland the acetic acid concentration profiles decrease with increasingthe initial feed concentrations, while keeps almost the same forglucose and glycerol. This phenomenon is influenced by thedifferent types of isotherm for the solutes at the experimental con-ditions. The retention time can be given by the basic equation (Eq.(14)).

In the special case of linear isotherms, Eq. (14) reduces to:

tR;iðcþi Þ ¼ t0 1þ 1� eb

ebKi

� �ð15Þ

where Ki is the linear isotherm coefficient. The isotherms of glucoseand glycerol are near linear isotherm in the concentration range, sothe retention time is constant. But the isotherm of ethanol and ace-tic acid are convex isotherms, and thus the retention time decreasedwith increasing concentration (Table A.3, Supplementarymaterials).

As the initial concentrations increasing, the competition forbinding sites on the surface of HD-01 resin also increase, whichescalates the displacement effect. It can be observed that the over-shoot of glycerol and glucose increase with increasing the initialfeed concentrations.

3.3.4. Effect of pH on the breakthrough curvesAs discussed above, acetic acid shows the highest affinity with

HD-01 and is difficult to be separated from ethanol in multi-com-ponent system. Thang and Novalin (2008) pointed out that thenon-ionic polymeric resin selectively adsorb non-ionic species

Table 4Effect of H/D on the adsorption experiments in packed column and the corresponding results in quaternary system.

Exp. no. H/D L (cm) Solute tb (min) qs (mmol/kg) F (p)

8 3.7 7.6 Glucose 12.3 11.8 1.21Glycerol 22.6 37.7Ethanol 63.3 3718.0Acetic acid 72 215.0

11 5.7 11.5 Glucose 17.2 11.8 1.01Glycerol 38.5 37.6Ethanol 99.2 3735.6Acetic acid 108.5 226.0

12 7.4 15.2 Glucose 34.1 12.2 1.05Glycerol 54.2 38.1Ethanol 133.1 3709.6Acetic acid 146.2 227.0

⁄ The flow rate was set at 0.6 mL/min; the solution pH was 3.0. The initial concentration of ethanol, glycerol, glucose and acetic acid in quaternary solution was 2151.1, 108.6,56.1 and 83.4 mmol/L (equivalent to 99.1, 10, 10.1 and 5.01 g/L), respectively.

Fig. 4. Effect on initial concentrations on the breakthrough curves of glucose (a), glycerol (b), ethanol (c), and acetic acid (d) in multi-solute systems. The solid lines representthe simulation results with the mass transport model.

366 J. Zhou et al. / Bioresource Technology 143 (2013) 360–368

compared to ionic species. This is may be that the main adsorptionmechanism of undissociated organic acids onto the non-ionic poly-meric resin is due to the hydrophobic interactions (Nielsen et al.,2010), which is prominent only at a pH that is well below itspKa. The effect of pH on the equilibrium adsorption of acetic acidand other solutes on HD-01 could be found in S5 (Supplementalmaterials).

The dynamic adsorption experiments were conducted atpH = 3.0 and 7.0, respectively. The experimental data along withthe simulation results are depicted in Fig. 5. It was evident that

the adsorption of acetic acid on HD-01 was significantly decreasedwhen the solution pH increased to 7.0 (Fig. 5d). The breakthroughpoint of acetic acid is decreased from 72 to 11.1 min, indicating theadsorption of acetic acid at pH = 7.0 is negligible. Thus the separa-tion of ethanol from acetic acid could be achieved by adjust thesolution pH to 7.0. Further more, due to the weaken affinity withHD-01 resin at pH = 7.0, acetic acid could be displaced by ethanolduring the dynamic adsorption process. Therefore, the concentra-tion history profiles of acetic acid (c/c0) exhibit overshoot in mul-ti-component system at solution pH = 7.0 and could be predicted

Fig. 5. Effect of pH on the breakthrough curves of glucose (a), glycerol (b), ethanol (c) and acetic acid (d), the solid line represent the results simulated with the column model.

J. Zhou et al. / Bioresource Technology 143 (2013) 360–368 367

by the column model incorporating the competitive Langmuir iso-therm at pH = 7.0 (not shown) very well.

Besides, the overshoot of glycerol is slightly lower at pH = 7.0 inmulti-component systems (Fig. 5b) due to the weaken competitionby acetic acid. The breakthrough point (tb) of ethanol is increasedfrom 63.3 to 65.3 at pH = 7.0, which is close to that in single-solutesystem (66.7 min).

Fig. 6. The breakthrough curves of ethanol, glucose, glycerol and acetic acid withactual fermentation broth at pH = 7.0.

3.3.5. Column dynamic adsorption process with actual ethanolfermentation broth

In addition to ethanol and substrate (glucose), a variety by-product also existed in fermentation broth, e.g., glycerol, organicacids, inorganic salt, proteins, pigment and so on, which could hin-der the purification. In this section, HD-01 resin was evaluated torecovery ethanol from actual ethanol fermentation broth. In orderto separate ethanol from acetic acid, the pH of fermentation brothwas adjust to 7.0 by NaOH solution prior the column adsorption.The composition of fermentation broth and the model parametersare listed in Table 3. The breakthrough curves of ethanol and othersolutes are shown in Fig. 6. It can be seen that the column modelincorporating the competitive Langmuir isotherm at pH = 7.0 fittedthe experimental data very well, which indicate that the effect ofimpurities, such as proteins, pigment and inorganic salt, on theadsorption of ethanol was negligible. HD-01 resin shows highadsorption for ethanol from actual fermentation broth(qe = 3631.4 mmol/kg wet resin, ce = 1886.5 mmol/L), although thiswas a slight (7.0%) lower than that from single-component solution(qe = 3902.8 mmol/kg wet resin, ce = 1886.5 mmol/L, calculatedwith Langmuir isotherm equation). The adsorption for glucose

and glycerol on HD-01 resin from fermentation broth was negligi-ble due to the competitive adsorption and their low adsorptionaffinity with HD-01. It also can be seen from Fig. 6, the separationbetween ethanol and glucose (glycerol or acetic acid) at pH = 7.0 isvery high. The results indicate that HD-01 resin is a promisingadsorbent to recovery ethanol from fermentation broth due to itshigh adsorption capacity and selectivity.

368 J. Zhou et al. / Bioresource Technology 143 (2013) 360–368

4. Conclusion

A microporous hyper-cross-linked resin HD-01 was evaluatedto recover ethanol from actual fermentation broth. It shows excel-lent adsorption properties for ethanol with large capacity, highselectivity, and rapid adsorption rate. The competitive adsorptionof ethanol, glycerol, glucose and acetic acid on HD-01 in a fixedbed at various operating conditions could be predicted by the masstransport model with the constant mass transfer coefficients (Keff)well. Acetic acid has a negative effect on the adsorption of ethanoland is difficult to be separated from ethanol in multi-componentsystems, which could be prevented by increasing the pH to 7.0.

Acknowledgements

This work was supported by ‘‘National Outstanding YouthFoundation of China’’ (Grant No.: 21025625), ‘‘PCSIRT’’, ‘‘PAPD’’,and ‘‘National High-Tech Research and Development Plan of China’’(863 Program, 2012AA021202).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.biortech.2013.06.009.

References

Alfimova, R., Andreeva, N., Khoroshikh, N., Shishkin, Y.L., 1975. Rotation ofmolecular fragments about bonds and the dipole moment of glucose and theAmadori rearrangement products of glucose. J. Struct. Chem. 15, 830–831.

Bowen, T.C., Vane, L.M., 2006. Ethanol, acetic acid, and water adsorption from binaryand ternary liquid mixtures on high-silica zeolites. Langmuir 22, 3721–3727.

Bui, S., Verykios, X., Mutharasan, R., 1985. In situ removal of ethanol fromfermentation broths. 1. Selective adsorption characteristics. Ind. Eng. Chem.Process Des. Dev. 24, 1209–1213.

Chen, C., Tang, X., Xiao, Z., Zhou, Y., Jiang, Y., Fu, S., 2012. Ethanol fermentationkinetics in a continuous and closed-circulating fermentation system with apervaporation membrane bioreactor. Bioresour. Technol. 114, 707–710.

Danckwerts, P., 1953. Continuous flow systems: distribution of residence times.Chem. Eng. Sci. 2, 1–13.

Delgado, J.A., Águeda, V.I., Uguina, M.A., Sotelo, J.L., García, A., Brea, P., García-Sanz,A., 2013. Separation of ethanol–water liquid mixtures by adsorption on apolymeric resin Sepabeads 207�. Chem. Eng. J. 220, 89–97.

Einicke, W.D., Gläser, B., Schöoullner, R., 1991. In situ recovery of ethanol fromfermentation broth by hydrophobic adsorbents. Acta Biotechnol. 11, 353–358.

Escudero, C., Poch, J., Villaescusa, I., 2013. Modelling of breakthrough curves ofsingle and binary mixtures of Cu(II), Cd(II), Ni(II) and Pb(II) sorption onto grapestalks waste. Chem. Eng. J. 217, 129–138.

Glueckauf, E., 1955. Theory of chromatography. Part 10. Formulae for diffusion intospheres and their application to chromatography. Trans. Faraday Soc. 51, 1540–1551.

Guiochon, G., Felinger, A., Shirazi, D.G., 2006. Fundamentals of preparative andnonlinear chromatography. Academic Press, Amsterdam.

Hashi, M., Thibault, J., Tezel, F.H., 2010. Recovery of ethanol from carbon dioxidestripped vapor mixture: adsorption prediction and modeling. Ind. Eng. Chem.Res. 49, 8733–8740.

Ikegamai, T., Yanagishita, H., Kitamoto, D., Haraya, K., 2000. Accelerated ethanolfermentation by Saccharomyces cerevisiae with addition of activated carbon.Biotechnol. Lett. 22, 1661–1665.

Jones, R., Gandier, J., Thibault, J., Tezel, F., 2011. Enhanced ethanol productionthrough selective adsorption in bacterial fermentation. Biotechnol. BioprocessEng. 16, 531–541.

Jones, R.A., Thibault, J., Tezel, F.H., 2010. Simulation and validation of ethanolremoval from water in an adsorption packed bed: isotherm and mass transferparameter determination in batch studies. Can. J. Chem. Eng. 9999, 1–10.

Kawabata, N., Sumiyama, Y., Matsuura, N., 1988. Separation of alcohols from waterby adsorption on crosslinked polymethacrylic ester containing a pyridiniumgroup. Ind. Eng. Chem. Res. 27, 1882–1886.

Kleinübing, S.J., Guibal, E., da Silva, E.A., da Silva, M.G.C., 2012. Copper and nickelcompetitive biosorption simulation from single and binary systems bySargassum filipendula. Chem. Eng. J. 184, 16–22.

Kleinubing, S.J., da Silva, E.A., da Silva, M.G., Guibal, E., 2011. Equilibrium of Cu(II)and Ni(II) biosorption by marine alga Sargassum filipendula in a dynamicsystem: competitiveness and selectivity. Bioresour. Technol. 102, 4610–4617.

Lide, D.R., Bruno, T.J., 2002. CRC Handbook of Chemistry and Physics, 82nd ed. CRCPress Llc, Florida.

Nielsen, D.R., Amarasiriwardena, G.S., Prather, K.L.J., 2010. Predicting the adsorptionof second generation biofuels by polymeric resins with applications for in situproduct recovery (ISPR). Bioresour. Technol. 101, 2762–2769.

Nigam, P.S., Singh, A., 2011. Production of liquid biofuels from renewable resources.Prog. Energ. Combust 37, 52–68.

Pitt Jr., W.W., Haag, G.L., Lee, D.D., 1983. Recovery of ethanol from fermentationbroths using selective sorption–desorption. Biotechnol. Bioeng. 25, 123–131.

Schmidt-Traub, H., 2005. Preparative Chromatography of Fine Chemicals andPharmaceutical Agents. WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

Suzuki, M., Smith, J.M., 1972. Axial dispersion in beds of small particles. Chem. Eng.J. 3, 256–264.

Thang, V.H., Novalin, S., 2008. Green biorefinery: separation of lactic acid from grasssilage juice by chromatography using neutral polymeric resin. Bioresour.Technol. 99, 4368–4379.

Vane, L.M., 2008. Separation technologies for the recovery and dehydration ofalcohols from fermentation broths. Biofuels. Bioprod. Biorefin. 2, 553–588.

Wilke, C.R., Chang, P., 1955. Correlation of diffusion coefficients in dilute solutions.AlChE J. 1, 264–270.

Yang, X., Tsao, G.T., 1995. Enhanced acetone-butanol fermentation using repeatedfed-batch operation coupled with cell recycle by membrane and simultaneousremoval of inhibitory products by adsorption. Biotechnol. Bioeng. 47, 444–450.

Zhai, Z., Chen, J., Fei, Z., Wang, H., Li, A., Zhang, Q., 2003. Adsorption ofphenylhydrazine derivatives on hypercrosslinked polymeric adsorbents.React. Funct. Polym. 57, 93–102.