recovery of acetoin from the ethanol–acetoin–acetic acid ternary mixture based on adsorption...

Recovery of Acetoin from the Ethanol−Acetoin−Acetic Acid TernaryMixture Based on Adsorption Methodology Using a Hyper-Cross-Linked ResinJinglan Wu,∇,†,‡ Xu Ke,∇,†,‡ Lili Wang,†,‡ Renjie Li,†,‡ Xudong Zhang,†,‡ Pengfei Jiao,†,‡ Wei Zhuang,†,‡

Yong Chen,†,‡ and Hanjie Ying*,†,‡,§

†College of Biotechnology and Pharmaceutical Engineering, Nanjing University of Technology, Xin mofan Road 5, Nanjing 210009,China‡National Engineering Technique Research Center for Biotechnology, Nanjing University of Technology, Nanjing 211816, China§State Key Laboratory of Materials-Oriented Chemical Engineering, Nanjing University of Technology, Nanjing 210009, China

ABSTRACT: The separation performances, in terms of adsorption selectivity, desorption, and regeneration of an innovativehyper-cross-linked resin (HD-02) for recovery of acetoin from the ethanol−acetoin−acetic acid ternary mixture were explored inthis work. The competitive adsorption behaviors of the ternary mixture were determined experimentally. The results showed theHD-02 resin had a good adsorption selectivity toward acetoin over acetic acid and ethanol. Subsequently, the desorptionbehaviors in terms of desorption isotherms and kinetics were systematically investigated. Using ethanol as a desorbent, therecovery could be achieved as high as 98% in the batch desorption experiments. The Fick model was adopted to simulate thedesorption process. The simulation results revealed that the intraparticle diffusion was the rate-limiting step and the obtainedeffective diffusivities (1.530 × 10−9 m2/min) were independent of the acetoin concentrations. In the end of this work, threecycles of adsorption−desorption−regeneration operations confirmed good reproducibility of the resin for the attainment ofacetoin.

1. INTRODUCTION

Acetoin (3-hydroxy-2-butanone or acetyl methyl carbinol),which is a naturally occurring chemical in the dairy and certainfruits, is widely used to add flavor to food and also serves as aprecursor in the synthesis of many important compounds.1,2

This bio-based chemical is defined as one of the high value-added platform compounds and selected by the U.S. Depart-ment of Energy as one of the potential top 30 chemical buildingblocks from sugars.3 Currently, the industrial production ofacetoin via microbial fermentation is a leading method, becauseof its cost efficiency, compared with the chemical method andenzymatic conversion.4,5 Although some specific microorgan-isms have the ability to produce acetoin, the long fermentationperiod that is generally needed hinders its large-scaleproduction.6−8 For many strains, acetoin is only generated asa byproduct.3,9 For instance, the strain C. acetobutylicumB3,

10−14 which is widely known as a good producer ofacetone−butanol−ethanol (ABE), is able to accumulate acetoinas well. Chen,10 Lin,12 and Liu13 have used this type of strain toperform batch ABE fermentation experiments. Besides the mainproducts, i.e., butanol (∼12 g/L), acetone (∼6 g/L), andethanol (∼2 g/L), the concentrations of byproducts, i.e., aceticacid, acetoin, and butyric acid in the final fermentation brothare ∼1, 2, and 0.6 g/L, respectively. Traditionally, only ABEwere extracted in the separation process, resulting in a loss oftotal product yield. Since acetoin is a high value-addedcompound with the price of $41.0/kg, much higher thanbutanol ($1.75/kg), co-production of ABE and acetoin couldbe one of the novel alternatives to the ABE production.

In the past few years, we have dedicated ourselves to thedevelopment of adsorption methodology to recover butanolfrom the ABE solution. So far, a weak-polar adsorption resinKA-I12,15 has been selected and proved to have the ability toadsorb butanol, butyric acid, and acetone efficiently, while theresiduals (i.e., ethanol, acetic acid, and acetoin, existing in theABE solution are unrestrained on this type of resin, because ofthe competitive adsorption effects and remained in the effluent.Hence, after adsorption, the effluent contains almost nobutanol, butyric acid, and acetone. In order to realize the co-production of ABE and acetoin, a new resin must be developedto separate acetoin from ethanol and acetic acid in the effluent.In our previous work, a hyper-cross-linked resin (HD-02) wasapplied to adsorb pure acetoin. The thermodynamic equili-brium as well as kinetic adsorption of acetoin on the HD-02resin has been investigated systematically.16

As a continuation of our previous work, the main objective ofthe present study is to further explore the potential applicationof the HD-02 resin for recovery of acetoin from the ethanol−acetoin−acetic acid ternary mixture. For this purpose, inaddition to the resin adsorption capacity, the adsorptionselectivity, desorption and resin regeneration play an essentialrole in the acetoin recovery process as well. To the best of ourknowledge, no previous studies have applied adsorptionmethodology to separate acetoin from the ABE solution.

Received: December 26, 2013Revised: July 9, 2014Accepted: July 14, 2014Published: July 14, 2014

Article

pubs.acs.org/IECR

© 2014 American Chemical Society 12411 dx.doi.org/10.1021/ie502105q | Ind. Eng. Chem. Res. 2014, 53, 12411−12419

pubs.acs.org/IECR

Accordingly, the following work was carried out in the presentstudy:

(i) Screening of six adsorption resins with various polarityand specific surface areas to explore the adsorptionmechanism of resins to acetoin. The HD-02 resinshowed a great advantage in adsorption capacity andadsorption rate.

(ii) Investigation of the competitive adsorption behaviors ofthe ternary mixture, i.e., ethanol, acetoin, and acetic acidto evaluate the adsorption selectivity of the resin. Thefeasibility of successful separation of acetoin from themixture was validated by the experimentally obtainedbreakthrough curves.

(iii) Establishment of a mathematical column model, whichtakes the axial dispersion and mass transfer into account,for the accurate design of the operating conditions in thesole/multicomponent separation systems. The synthe-sized ethanol−acetoin−acetic acid solution was used as afeed solution.

(iv) Systematic investigation of the acetoin desorptionbehaviors from the HD-02 resin. The desorptionisotherms of acetoin were measured experimentally invarious molar ratios of ethanol to water solutions. Thekinetics of desorption acetoin on the resin weresimulated and the effective diffusivity was obtained.

(v) In the end of this work, the adsorption/desorption/regeneration cycles were carried out to assess thereproducibility of the HD-02 resin.

2. MATERIALS AND METHODS2.1. Materials. Pure 99% acetoin was purchased from

Sigma−Aldrich. Ethanol and acetic acid (>99.5 wt %) wereobtained from Sinopharm Chemical Reagent Co., Ltd. The sixadsorbents used in this research were kindly provided byNational Engineering Technique Research Center for Bio-technology (Nanjing, China). The chemical structures andphysical properties of the resins are listed in Table 1. Prior to

the adsorption experiments, the resins were pretreated first bysoaking it into the ethanol solution for 24 h and then by 1 MHCl and 1 M NaOH for 8 h to remove preservative agents andpolymerization residuals. Finally, the resins were washed toneutral pH with deionized water.2.2. Methods. 2.2.1. Preliminary Selection of the Proper

Resin for Adsorption of Acetoin. The batch static method wasused for determination of the adsorption capacities of differentresins. Equilibrium adsorption experiments were carried out inclosed 100-mL Erlenmeyer flasks containing 50 mL of aqueous

solution and 2.0 g of resin. Solutions were prepared indeionized water with acetoin at a known initial concentration(10 g/L). The mixtures were then equilibrated for 24 h at 293K, with agitation at 120 rpm. After measuring soluteconcentrations both before and after equilibration, the amountsof adsorbate (i) adsorbed onto resin were calculated by themass balance relation, as shown in eq 1:17

=−

qC C V

m

( )i,e

i,0 i,e 0

(1)

where Ci,0 and Ci,e (g/L) are the initial and the equilibriumconcentrations of the component, m (g) and V0 (L) representthe weight of the adsorbent and the volume of the solution, qi,e(mg/g wet resin) is the mass of acetoin adsorbed per unit massof adsorbent at equilibrium.Determination of the adsorption rate was performed using a

500-mL Erlenmeyer flask containing 300 mL of a 10 g/Lacetoin solution and 15 g of different resins. The flask wasplaced in a thermostated water bath with agitation at 120 rpmto maintain the desired temperature (293 K). The sampleswere withdrawn from the vessel at regular intervals and thenanalyzed by gas chromatography (GC). Adsorption rate curve(ln(Ct/C0) − t) was applied to investigate the adsorption rateof the six resins.

2.2.2. Batch Adsorption Experiments. Batch adsorptionexperiments were performed to study the single/multi-component equilibrium adsorption isotherms onto HD-02resin at the temperature of 293 K. The concentration ratio ofethanol:acetoin:acetic acid in ternary mixture solution wasequal to 2:2:1, which was close to the ratio of thesecomponents in the actual final fermentation broth.10,13 Theexperiments with the same laboratory equipment andprocedures as the method described in section 2.2.1 wererepeated. The amount of single/multicomponent adsorbed perunit mass of resin (qi,e) and distribution coefficient (Kd)

18,19

were calculated by eqs 1 and 2, respectively:

=−

KV C C

mC

1000 ( )d

0 i,0 i,e

i,e (2)

2.2.3. Fixed-Bed Experiments. Fixed-bed adsorption runswere carried out in glass columns with an inside diameter of1.90 cm and a length of 30 cm. Every column had a water jacketto maintain the desired constant operating temperature (293K). Solutions with a known concentration were fed to the topof the column at different flow rates regulated by a constant-speed pump. The effluent samples were collected at intervalsand analyzed via GC. In fixed-bed studies, the adsorptionperformance of a column is mostly evaluated by analyzing theform of normalized concentration, Cout/CF of the effluent vstime curves (i.e., breakthrough curves).

2.2.4. Batch Desorption Experiments. The desorptionequilibrium experiments were performed in 100-mL Erlen-meyer flasks that contain 2 g of resin and 50 mL of acetoin−ethanol−water solution with different molar ratios of ethanol towater (i.e., 0:1 to 1:0). The procedures described in section2.2.1 then were repeated and the adsorption capacities (mg/g)of acetoin in the mixture solutions were calculated using eq 1.The desorption kinetics experiments were conducted

immediately after the adsorption experiments using acetoin-loaded adsorbent, which was previously exposed to differentconcentrations (initial concentration from 4.675 g/L to 24.82g/L) of acetoin solution. After the attainment of adsorption

Table 1. Chemical Structures and Physical Properties of theSix Resins

type ofresin

matrixstructure polarity

pore sizedistribution

surface area(m2/g)

q(mg/g)

KA-I PS-PVBa weak polarity macropore 850−900 61.81KA-II PS-PVBa strong

polaritymacropore 810 8.572

KA-III PS-PVBa nonpolarity macropore 845 22.07HD-01 PS-PVBa weak polarity micropore 1646 73.44HD-02 PS-PVBa weak polarity micropore/

mesopore1294 66.55

HD-03 PS-PVBa weak polarity mesopore 1129 72.66aPolystyrene−poly(vinyl)-benzene.

Industrial & Engineering Chemistry Research Article

dx.doi.org/10.1021/ie502105q | Ind. Eng. Chem. Res. 2014, 53, 12411−1241912412

equilibrium, the adsorbed resin was washed with distilled waterand then dried by filter paper. The following operation wascarried out in a 250-mL Erlenmeyer flask that contained 200mL of absolute ethanol and 5 g of saturated resin. The flask wasplaced in a thermostated water bath to maintain the desiredtemperature (293 K). The agitation was fixed to 120 rpm toensure that the film mass-transfer effects caused by the lowagitation speed were neglected. Samples were withdrawn atdifferent times with a syringe, and then measured by GC. Thepercentage desorption of acetoin (D%) from acetoin-loadedadsorbent was calculated using eq 3:

=−C V

C C VD%

( )x x

e0 0 (3)

where Cx (g/L) is the concentration of acetoin in the elutionand Vx (L) is the volume of elution.2.2.5. Regeneration Experiments. After fixed-bed adsorp-

tion experiments, the acetoin was eluted from the loadedadsorbent by the desorbent. All effluent samples were collectedby a fraction collector at definite intervals and measured by GC.Then, the HD-02 resin was washed with deionized water andused again in the next cycle. Acetoin was used to saturate theloaded adsorbent with a flow rate of 3.0 mL/min previously, theelution and regeneration rate flow were 1.5 mL/min and 3.0mL/min, respectively. The recycling studies of adsorption/desorption/regeneration were repeated for three times in orderto examine the resin reproducibility.2.2.6. Analytical Method. The ethanol, acetoin, and acetic

acid concentration were determined via gas chromatography(GC) (Model 7890A, Agilent, USA) equipped with a 60 m ×0.25 mm × 0.25 μm Agilent HP-INNOWAX 19091N-236column. The oven temperature program used was as follows:70 °C for 0.5 min and then increased to 190 °C at 20 °C/min(6 min) for 1 min. The carrier gas was nitrogen, which had aflow rate of 2 mL/min. The injector temperature was 180 °C,and the FID detector temperature was 220 °C. The split ratiowas set to be 1/90.

3. THEORY3.1. Single/Multicomponent Adsorption Equilibrium.

In the single-component system, the experimental equilibriumdata of each individual solute were fitted by the Langmuirisotherm model,20 which is represented as follows:

=+

qq K C

K C1i,ei,m i i,e

i i,e (4)

where qm is the Langmuir isotherm constant (mg/g wet resin)and K is the equilibrium coefficient (L/g).In the multicomponent systems, the adsorption capacity of

each solute at equilibrium depends on the concentration ofother components present locally.21 As a result, the competitiveLangmuir isotherm model was used to describe the adsorptionequilibrium of each solute as follows:22

=+ ∑ =

qa C

b C1i ji i j

jn

j i j,

,

1 , (5)

where a and b are the equilibrium coefficients (L/g), theconstants a (a = qmax × KL) and b (b = KL) in the competitiveLangmuir isotherm model are dependent on the equilibriumconcentrations of other components and are not independent; jand n are the number of components.

3.2. Adsorption Selectivity. The separation factor23−25

(∂i/i′), which is an index of selectivity, was defined as the ratioof distribution coefficient (Kd) of component i to component i′.The separation factor values of ∂Ac

A , ∂EA, ∂Ac

E were used toevaluate the resin selectivity of acetoin/acetic acid, acetoin/ethanol, and ethanol/acetic acid, respectively.

3.3. Mathematical Modeling of a Single Chromato-graphic Column. The fixed-bed column model is the core fordesign and optimization of the operating conditions in thesole/multicomponent separation systems. The transportdispersive model (TDM),26−28 which takes into account theaxial dispersion as well as the mass-transfer resistance in thesolid phase, was considered to simulate the experimentallyattained breakthrough curves of the single component(acetoin) and the ternary mixture (acetoin, ethanol, and aceticacid).The overall mass balance in the mobile phase is29

ρ εε

∂∂

+∂∂

+ − ∂∂

=∂∂

ct

vcz

q

tD

cz

1i i i ii,L

2

2 (6)

wherev (mL/min) is the interstitial velocity, z (cm) and ρ (g/mL) are the axial coordinate of the column and the resin beddensity, t (min) is the time, ε (ε = 0.45) is the bed porosity andci (g/L) represents the concentration of the component i in thebulk liquid phase. DL (m2/min) denotes the axial dispersioncoefficient and can be obtained using following correlationsuggested by Suzuki and Smith:30

= +D D ud0.44 0.83i,L i,m p (7)

where dp (cm) is the particle diameter, u (cm/min) thesuperficial velocity, and Dm (m2/min) the molecular coefficient,which can be calculated as31

ϕη

= × −DM T

V7.4 10i,m

8 B

B A0.6

(8)

where VA (cm3) is the molar volume of the liquid solute at itsnormal boiling point, MB (g) is the molecular weight of thesolvent, ηB(N S m−2) its viscosity, and φ is a constant whichaccounts for solute−solvent interactions and has a recom-mended value of 2.6 for water.The differential mass balance equation in the solid phase32 is

given by

∂∂

= −q

tk q q( )i

eff i,e i (9)

where keff (m/min) is the mass-transfer coefficient, which isobtained by the best fitting of the simulation results to theexperimental data.The initial conditions are written as follows,

= = =t c t z q t zat 0: ( , ) 0, ( , ) 0i i (10)

The boundary conditions33 are written from the followingexpressions:

=∂ =

∂= − =

z

DC t z

zvC t vC t z

at 0:( , 0)

( ) ( , 0)ii iL ,0 (11)

=∂ =

∂=z L

C t z Lz

at :( , )

0i(12)



where L (cm) is the length of the fixed bed.Adsorption equilibrium isotherms of species i in the single/

multicomponent systems (qi,e = f(ci)) can be described by theLangmuir isotherm (eq 4) or the competitive Langmuirisotherm model (eq 5). The partial differential equationswere solved efficiently in MATLAB 2010a. After thediscretization step, the time integration is performed by theordinary differential equation solver ODE23. An absolute andrelative tolerance of 10−5 was used.3.4. Desorption Kinetics Model. The Fick model34,35 was

applied to predict the acetoin desorption process on the HD-02resin. The model was simplified by making the followingassumptions: (i) the resin was treated as a quasi-homogeneousphase; (ii) diffusion was restricted to radial diffusion; and (iii)intraparticle diffusion was the rate-limiting step throughout theprocess. Accordingly, the conservation equations for acetoin inthe bulk fluid and in the particles include

ρ= −

∂∂ =

VCt

mR

Dqr

dd

3e

r R (13)

∂∂

=∂∂

+∂∂

⎛⎝⎜

⎞⎠⎟

qt

Dq

r rqr

2e

2

2(14)

where V (cm3) is the volume of adsorbent, R (cm) is the radialdistance from the center of the pellet, and De (m2/min)represents the effective diffusion coefficient, which wasestimated by matching the experimental concentration decaydata (cexp, g/L), with the concentration decay predicted with anumerical solution of the diffusion model (ccal, g/L). The bestvalue for De was obtained when the diffusion model best-fit theexperimental data, considering that the optimum fit wasachieved by minimizing the following objective function:36

∑=−

=

⎛⎝⎜

⎞⎠⎟

c cc

minimumi

N

1

e cal

e

2

(15)

Meanwhile, the average relative deviation (ARD%, as definedby eq 16) is used to evaluate the model fitness as well:

∑=−

×=N

c c

cARD%

1100

i

N

1

exp cal

exp (16)

where the subscripts “exp” and “cal” denote the experimentaland calculated values, respectively, and N is the number ofexperimental data points.

4. RESULTS AND DISCUSSION4.1. Preliminary Selection of the Proper Resin for

Adsorption of Acetoin from Aqueous Solution. Sixdifferent types of resins were screened for their ability totake-up acetoin from the aqueous solution with an initialconcentration of 10 g/L, in which the series of KA resins werethe macropore adsorption polymers with various polarities,whereas the series of HD resins were hyper-cross-linkedpolymers with the specific surface areas all above 1000 m2/g(see our previous work in ref 16). The adsorption capacities ofthese resins to acetoin are compared in Table 1. It can beobserved that HD series resins generally display higher affinityto acetoin, because of the large specific surface areas and theuniform micropore/mesopore structures that supply bettercompatibility with the adsorbate acetoin. The uptake of weakpolarity resin KA-I is higher than that of the nonpolarity resin

KA-II, while the strong polarity resin KA-III shows practicallyno adsorption to acetoin. It indicates that, irrespective of theirnature, the polymer resins that possess hydrophobic surfacechemistries (to support stronger surface interactions throughvan der Waals forces) as well as high specific surface areasmight serve as effective adsorbents for the separation of acetoinfrom the aqueous solutions. Meanwhile, the adsorption ratesfor the six adsorbents have been determined experimentally andthe results were presented with adsorption rate curves (ln(Ct/C0) − t) in Figure 1. As can be seen, all of the acetoin

adsorption attain equilibrium within 30 min; the six adsorbentspossess almost the same adsorption rate. All the HD series resincan be used for the recovery of acetoin. As shown in Table 1,we can see that the HD-01 resin possesses a uniform microporestructure while the HD-02 resin has both micropores andmesopores, which means the pore distribution is much widerthan that of HD-01. Therefore, it is more complicated tosynthesize the HD-01 resin, resulting in the high cost ofproducing the HD-01 resin. Proceeding from the economicpoint of view, the HD-01 resin would be inappropriately usedin the industrial application. Since the HD-02 resin presentsrelatively high adsorption capacity, as well as a high adsorptionrate to acetoin, and the resin has been used to adsorb acetoin inour previous work, this resin is selected preliminarily as theproper resin for adsorption of acetoin and used in the followingwork.

4.2. Single/Multicomponent Adsorption Equilibrium.For the sake of comparison, the single-component andmulticomponent adsorption isotherms, in terms of ethanol,acetoin, and acetic acid, on the HD-02 resin at a temperature of298.15 K are presented in Figure 2. The acetoin concentrationsranged from 0 to 30.0 g/L. According to the composition in thereal ABE fermentation broth (i.e., ethanol:acetoin:acetic acid =2:2:1), in the multicomponent system, the ethanol concen-trations were set to be 0 to 30.0 g/L and those of acetic acid tobe 0 to 15.0 g/L. The Langmuir isotherm (eq 4) and thecompetitive Langmuir isotherm models (eq 5), which describethe single-component and multicomponent adsorption behav-iors, were used to fit the adsorption isotherm data. The fittingresults also are shown in Figure 2. Good agreements betweenthe model predictions and the experimental data can beobserved. The model parameters are listed in Table 2.

Figure 1. Adsorption rate curves of the six different resins to acetoin atthe temperature of 293 K (C0 = 10 g/L).



It is clear that, in the single-component system, acetoin,acetic acid, and ethanol can be all retained on the HD-02 resin.The relative adsorption affinity of these three compounds tothe HD-02 resin follows the order: acetoin > acetic acid >ethanol. However, in the multicomponent system, theadsorption capacity is reduced significantly, in comparisonwith its individual pure component. For instance, the uptake ofacetoin is reduced from 228.31 mg/g to 178.43 mg/g. At thesame time, with the coexistence of acetoin, the uptakes ofethanol and acetic acid on the HD-02 resin are also reducedgreatly, indicating that the competition adsorption occursamong these components. This phenomenon would be furtherverified in the multicomponent competitive breakthroughcurves in section 4.4. Nevertheless, no matter either in thesingle-component or in the multicomponent system, acetoinalways displays stronger affinity to the resin than acetic acid andethanol, which implies the selected HD-02 resin possesses goodadsorption selectivity toward acetoin. Hence, the separationfactors of acetoin/acetic acid, acetoin/ethanol, and acetic acid/ethanol would be calculated to prove the resin selectivityquantitatively.4.3. Adsorption Selectivity of HD-02 Resin. The

separation factor was calculated according to section 3.2, toevaluate whether the HD-02 resin could be applicable for theselective adsorption toward acetoin from a multicomponentsystem. The obtained separation factors of acetoin/acetic acid,acetoin/ethanol, and acetic acid/ethanol are presented inFigure 3. At a fixed initial concentration ratio of ethanol,acetoin, and acetic acid (2:2:1), the separation factors varied inthe range of 2.122 < ∂E

A < 2.670, 1.387 < ∂AcA < 1.906, 0.7987 <

∂AcE < 1.1719. Since acetoin is the strongest retained

component, while ethanol is the weakest retained one, theseparation factor ∂E

A shows the maximum value of above 2,

demonstrating that these two substances can be easily separatedby the HD-02 resin. The average value of ∂Ac

A is ∼1.5, indicatingthat these two substances can also achieve baseline separation.However, the separation of acetic acid from ethanol is not easy,because of the small value of ∂Ac

E , i.e., ∼1.0. Anyway, the mainobjective of this study is to recover acetoin, according to thevalues of ∂E

A and ∂AcA , we can conclude that the selected novel

HD-02 resin possesses a good adsorption selectivity towardacetoin over acetic acid and ethanol.

4.4. Mathematical Modeling of the Column Adsorp-tion Process for the Sole and Ternary Mixture Systems.In principle, purification and recovery of acetoin from the ABEfermentation broth would be realized on the chromatographiccolumn in the end. Consequently, investigation of thebreakthrough curves, especially the competitive breakthroughcurves that are used to evaluate the simultaneous adsorptionbehavior and species interactions, are of great importance.In this section, the breakthrough curves of the single

component, with regard to acetoin, acetic acid, and ethanol, aswell as their competitive breakthrough curves were exper-imentally recorded and simulated by means of the TDM model.The corresponding results are presented in Figure 4. Themodel predictions fit the experimental data fairly well. Since theadsorption affinities of ethanol and acetic acid on the HD-02resin are weak, the packed-column is soon saturated with thesesolutes. Acetoin shows the strongest adsorption affinity to theresin, which is consistent with the results obtained from thesingle-component equilibrium data (Figure 2). For the ternarysystem, the breakthrough time of each component is ahead oftheir individual pure component, indicating that the adsorptioncapacities are reduced. It also corresponds to the resultsobtained from the ternary-component equilibrium data (Figure2). Moreover, the overshoots for ethanol and acetic acidconcentrations (Cout/CF) can be observed in Figure 4. This isbecause the least-adsorbed substance ethanol was displaced anddesorbed simultaneously by the middle adsorbed one (aceticacid), whereas the most adsorbed component (acetoin)replaced acetic acid as well. According to Moreira andFerreira,37 the overshoot phenomena in adsorption multi-component systems as a result of the competition for activesites on the resin can be predicted by the equilibrium theoryand considering the column divided into zones of unique

Figure 2. Adsorption isotherms of ethanol, acetoin, and acetic acidonto HD-02 resin in single-component and ternary systems at atemperature of 293 K.

Table 2. Isotherm Parameters for Solutes Adsorption onHD-02 Resin

single component multicomponent

solute qm(mg/g) KL(L/g) R2 a(L/g) b(L/g)

ethanol 209.5 0.0192 0.9942 4.551 0.0037acetoin 228.3 0.0748 0.9983 10.36 0.0581acetic acid 192.3 0.0347 0.9943 6.516 0.0010

Figure 3. Separation factors of the ternary system, in terms of theamounts of ethanol, acetion, and acetic acid, at a temperature of 293 K.



compositions, called plateaus, separated from each other byzones of varying compositions, called transitions. The peaks inthe concentrations during the breakthrough of ethanol andacetic acid occur at times of 96 and 107 min, respectively. Thestoichiometric time for acetoin is 130 min. The time intervalbetween the overshoots of ethanol and acetic acid and thestoichiometric point of acetoin indicates a good separation canbe achieved. Because of the good agreement between theexperimentally obtained breakthrough curves and the modelpredictions, the adsorption equilibriums, as well as the TDMmodel proposed in this work, can be considered correct.4.5. Product Recovery and Resin Regeneration. The

operating steps, in terms of desorption and regeneration, playsan important role in a complete acetoin recovery process.However, most of the literature has focused on the adsorptionprocess.16,38,39 Relatively limited information is available ondesorption behavior and resin regeneration. Accordingly, in thissection, the desorption equilibrium isotherms and thedesorption kinetics of acetoin on the HD-02 resin weredetermined experimentally and simulated. The adsorption−desorption-regeneration cycles were carried out to access thereproducibility of the resin.4.5.1. Desorption Equilibrium Isotherm. Selection of a

proper desorbent is of primary importance in the desorptionprocess. Generally, the following factors would be taken intoaccount for choosing a suitable desorbent:12,40 (i) highsolubility of the adsorbate(s) in the desorbent, (ii) easyseparation of the adsorbate(s) from the desorbent, and (iii)enrichment of the adsorbate(s) in the eluent, if possible. In thecase of acetoin recovery, pure ethanol was chosen to be adesorbent. Besides fulfillment of the above-mentioned require-ments, another important reason to select ethanol as adesorbent is that it is already existed in the butanolfermentation broth. Therefore, taking advantage of ethanolcan avoid the introduction of additional impurities into thecomplete separation process.The desorption isotherms of acetoin at different molar ratios

of ethanol and water solutions are shown in Figure 5. TheLangmuir isotherm model was used to fit the experimental data.The resulting parameters (i.e., the maximum adsorptioncapacity (qmax) and the Langmuir constant (k)) are listed inTable 3. As can be seen, qmax and k decrease as the ethanol

concentration increases. It seems that the solution compositionhas a strong influence on the retention behavior of acetoin onthe HD-02 resin. As a matter of fact, this phenomenon is well-known in analytical chromatography, in that changing themobile phase composition (e.g., either changing the ratiobetween a polar and an apolar solvent, or changing the pH ofthe solution) significantly modifies the solute adsorptivity onthe stationary phase.41 As competitive adsorption presentsbetween acetoin and ethanol, it is reasonable to assume that thedesorption of acetoin from the resin could be a result ofcompeting interactions between the intermolecular forces ofadsorption on the resin and dissolution in the solvent. Whenintermolecular forces are recessive, acetoin desorbs from theresin into the solvent. Therefore, a high ethanol concentrationwould decrease the intermolecular forces of acetoin and theresin, resulting in a reduction of adsorption capacity.

4.5.2. Modeling the Desorption Kinetics. Batch desorptionexperiments of loaded acetoin on the HD-02 resin were carriedout to study the desorption kinetics. The concentration decaycurves of acetoin with various initial concentrations wererecorded experimentally and are presented in Figure 6. Thefitting lines were calculated by the Fick model. Good agreementbetween the experimental data and the predictions can beobserved. The desorption process is found to be faster thanadsorption, because the desorption rate is rapid for the first 10min and thereafter it proceeds at a slower rate and finally attainsthe desorption equilibrium within ∼16 min, while in theadsorption step, the equilibrium time was over 30 min (see ourprevious work in ref 16). Since 200 mL of pure ethanol wasused to desorb 5 g of saturated resin in the desorption

Figure 4. Breakthrough curves of ethanol, acetoin, and acetic acid ontoHD-02 resin in single-component and ternary systems at atemperature of 293 K.

Figure 5. Desorption isotherms of acetoin in different molar ratios ofethanol and water at T = 293 K.

Table 3. Coefficients of the Desorption Isotherms atDifferent Molar Ratios of Ethanol and Water, T = 293 K

nethanol:nwater K qmax R2

1:0 0.0102 58.71 0.99371:7 0.0210 83.49 0.99681:11 0.0213 96.84 0.99891:19 0.0265 110.5 0.99301:50 0.0336 125.5 0.99421:150 0.0551 154.2 0.99821:250 0.0616 168.0 0.99870:1 0.0748 228.3 0.9949



experiments, the desorption recovery could be calculated to beas high as 98%. It confirms that ethanol could be a suitabledesorbent to elute acetoin from the resin. The effectivediffusivities De in the Fick model obtained by the best-fitting ofthe experimental data are given in Table 4. It can be seen that

the values are independent of the acetoin concentrations. Sinceacetoin is almost not retained on the resin, the diffusion isrestricted to the radial diffusion and is the rate-limiting stepthroughout the desorption process.42

4.5.3. Resin Regeneration. The adsorption−desorption-regeneration steps were carried out in the chromatographiccolumn to assess the reproducibility of the HD-02 resin. Theexperimentally attained adsorption breakthrough curves, as wellas the elution curves of acetoin, are depicted in Figure 7. Thefix-column bed was saturated with 5.011 g/L acetoin solution of∼8.11 BV (bed volume) and then the acetoin recovery attainedvalues as high as 98.7% when eluted by 2.45 BV pure ethanol.Consequently, the acetoin concentration would be enriched inthe elution. As can be seen in Figure 7, the concentration ofacetoin at the peak of the elution curve could achieved values ashigh as 77.67 g/L, ∼15.5 times higher than the feedconcentration. However, comparison of the total bed volumeused for adsorption with that used for desorption, acetoin wasonly concentrated 3.31 times after the adsorption anddesorption process. The HD-02 resin was immersed in theethanol solution after the elution process. Since ethanol andwater are miscible with each other, the best regenerant wouldbe water. The optimal amount of water to completeregeneration of the resin was obtained experimentally of 3.2BV. After three adsorption−desorption−regeneration cycles,the adsorption capacity of resin to acetoin remains unchange-able. The results confirm that the HD-02 resin is a potential

good adsorbent that can be used successfully in the acetoinrecovery process.After regeneration steps, the residual components in the

effluent are ethanol, acetoin, and water. Since an azeotrope canbe formed between ethanol and water, the difference betweenthe boiling points of acetoin and the azeotrope is thereforeenhanced. The commonly used distillation method can then beapplied to obtain acetoin from ethanol−water solution and itwill be discussed and detailed in the future work.

5. CONCLUSIONS

Six adsorption resins with various polarity and specific surfaceareas were screened to explore the adsorption mechanism toacetoin. The weak-polarity HD-02 resin showed excellentadsorption properties for acetoin. The adsorption selectivity,desorption equilibrium isotherms and kinetics, as well asregeneration studies were investigated systematically to evaluatethe feasibility of successful separation of acetoin from theternary mixture, in terms of ethanol, acetoin, and acetic acid.The results revealed that the affinity of these three compoundsto the HD-02 resin follows the order: acetoin > acetic acid >ethanol. The obtained effective diffusivity De (1.530 × 10−9 m2/min) was independent of the acetoin concentrations, indicatingthat the radial diffusion was the rate-limiting step throughoutthe desorption process. These preliminary experimentssuggested that the adsorption of acetoin by the HD-02 resincould be an attractive method to separate acetoin from theethanol−acetoin−acetic acid ternary system, and could beparticularly useful for realizing the co-production of acetone−butanol−ethanol (ABE) and acetoin in the future.

■ AUTHOR INFORMATION

Corresponding Author*Tel.: +86 25 86990001. Fax: +86 25 58139389. E-mail:[email protected].

Author Contributions∇These authors contributed equally to this work.

NotesThe authors declare no competing financial interest.

Figure 6. Desorption kinetics of acetoin at different initialconcentrations when T = 293 K.

Table 4. Coefficients of the Desorption Kinetic Model atDifferent Initial Concentrations, T = 293 K

C0 (g/L) Ce (g/L) De (× 10−9 m2 min−1) desorption rate (%) ARD%

4.675 0.9987 98.26 7.36810.22 1.443 1.530 98.55 6.50714.44 2.240 98.40 4.07124.82 2.506 98.74 7.297

Figure 7. Three adsorption/desorption/regenaration cycles of acetoinonto the HD-02 resin (T = 293 K).



mailto:[email protected]

■ ACKNOWLEDGMENTSThis work was partly supported by a grant from the NationalOutstanding Youth Foundation of China (Grant No.21025625), the PCSIRT, and the PAPD. We would also liketo acknowledge the financial support provided by the NationalHigh-Tech Research and Development Plan of China (863Program, No. 2012AA021202) and by the State KeyLaboratory of Motor Vehicle Biofuel Technology.

■ NOMENCLATUREa, b = multicomponent competitive Langmuir isothermconstants, L/gC = concentration of liquid phase, g/Lcal = calculated valuedp = particle diameter, cmD% = percentage desorption of acetoin from an acetoin-loaded adsorbentDe = effective diffusion coefficient, m2/minDL = axial dispersion coefficient, m2/minDM = molecular coefficient, m2/mine = equilibrium conditionexp = experimental valuei = sorbate species ij = all sorbate speciesk = single-component Langmuir isotherm constant related tothe rate of adsorption, L/gkeff = mass-transfer coefficient, m/minkd = distribution coefficient, L/gL = length of fixed bed, cmm = mass of wet resin, gMB = molecular weight of the solvent, gn = number of components in the systemN = number of experimental data pointsout = effluent solutionq = adsorption capacity, mg/gr = radius of a resin particle, cmR = radial distance from the center of the pellet, cmt = time, minT = temperature, Kv = interstitial velocity, mL/minV = volume of adsorbent, cm3

V0 = volume of aqueous solution, LVA = molar volume of the liquid solute at its normal boilingpoint, cm3

x = acetoin in the elutionz = axial coordinate of the column, cm0 = initial conditionα = separation factorρ = resin bed density, g/mLε = bed porosity (ε = 0.45)u = superficial velocity, cm/minφ = constant that accounts for solute−solvent interactions;the recommended value is 2.6 for waterηB = solvent viscosity, N S m−2

■ REFERENCES(1) Zhang, X.; Zhang, R.; Bao, T.; Yang, T.; Xu, M.; Li, H.; Xu, Z.;Rao, Z. Moderate expression of the transcriptional regulator ALsRenhances acetoin production by Bacillus subtilis. J. Ind. Microbiol.Biotechnol. 2013, 40 (9), 1067−1076.(2) Effendi, C.; Shanty, M.; Ju, Y.-H.; Kurniawan, A.; Wang, M.-J.;Indraswati, N.; Ismadji, S. Measurement and mathematical modelingof solubility of buttery-odor substance (acetoin) in supercritical CO2 at

several pressures and temperatures. Fluid Phase Equilib. 2013, 356,102−108.(3) Ji, X.-J.; Xia, Z.-F.; Fu, N.-H.; Nie, Z.-K.; Shen, M.-Q.; Tian, Q.-Q.; Huang, H. Cofactor engineering through heterologous expressionof an NADH oxidase and its impact on metabolic flux redistribution inKlebsiella pneumoniae. Biotechnol. Biofuels 2013, 6 (1), 7.(4) Zhang, Y.; Li, S.; Liu, L.; Wu, J. Acetoin production enhanced bymanipulating carbon flux in a newly isolated Bacillus amyloliquefaciens.Bioresour. Technol. 2013, 130, 256−260.(5) Cho, S.; Kim, K. D.; Ahn, J.-H.; Lee, J.; Kim, S.-W.; Um, Y.Selective Production of 2,3-Butanediol and Acetoin by a NewlyIsolated Bacterium Klebsiella oxytoca M1. Appl. Biochem. Biotechnol.2013, 1−12.(6) Fan, Y.; Tian, Y.; Zhao, X.; Zhang, J.; Liu, J. Isolation of Acetoin-Producing Bacillus Strains from Japanese Traditional FoodNatto.Preparative Biochem. Biotechnol. 2013, 43 (6), 551−564.(7) Xu, H.; Jia, S.; Liu, J. Development of a mutant strain of Bacillussubtilis showing enhanced production of acetoin. Afr. J. Biotechnol2011, 10 (5), 779−788.(8) Liu, Y.; Zhang, S.; Yong, Y.-C.; Ji, Z.; Ma, X.; Xu, Z.; Chen, S.Efficient production of acetoin by the newly isolated Bacilluslicheniformis strain MEL09. Process Biochem. 2011, 46 (1), 390−394.(9) Wu, Z.; Wang, Z.; Wang, G.; Tan, T. Improved 1,3-propanediolproduction by engineering the 2,3-butanediol and formic acidpathways in integrative recombinant Klebsiella pneumoniae. J.Biotechnol. 2013, 168 (2), 194−200.(10) Chen, Y.; Zhou, T.; Liu, D.; Li, A.; Xu, S.; Liu, Q.; Li, B.; Ying,H. Production of butanol from glucose and xylose with immobilizedcells of Clostridium acetobutylicum. Biotechnol. Bioprocess. Eng. 2013, 18(2), 234−241.(11) Doremus, M. G.; Linden, J. C.; Moreira, A. R. Agitation andpressure effects on acetone−butanol fermentation. Biotechnol. Bioeng.1985, 27 (6), 852−860.(12) Lin, X.; Wu, J.; Jin, X.; Fan, J.; Li, R.; Wen, Q.; Qian, W.; Liu,D.; Chen, X.; Chen, Y. Selective separation of biobutanol fromacetone−butanol−ethanol fermentation broth by means of sorptionmethodology based on a novel macroporous resin. Biotechnol. Progress2012, 28 (4), 962−972.(13) Liu, D.; Chen, Y.; Li, A.; Ding, F.; Zhou, T.; He, Y.; Li, B.; Niu,H.; Lin, X.; Xie, J.; Chen, X.; Wu, J.; Ying, H. Enhanced butanolproduction by modulation of electron flow in Clostridiumacetobutylicum B3 immobilized by surface adsorption. Bioresour.Technol. 2013, 129, 321−328.(14) Siemerink, M. A.; Kuit, W.; Contreras, A. M. L.; Eggink, G.; vander Oost, J.; Kengen, S. W. D-2,3-butanediol production due toheterologous expression of an acetoin reductase in Clostridiumacetobutylicum. Appl. Environ. Microbiol. 2011, 77 (8), 2582−2588.(15) Lin, X.; Li, R.; Wen, Q.; Wu, J.; Fan, J.; Jin, X.; Qian, W.; Liu,D.; Chen, X.; Chen, Y. Experimental and modeling studies on thesorption breakthrough behaviors of butanol from aqueous solution in afixed-bed of KA-I resin. Biotechnol. Bioprocess Eng. 2013, 18 (2), 223−233.(16) Wu, J.; Wang, L.; Zhou, J.; Zhang, X.; Liu, Y.; Zhao, X.; Wu, J.;Zhuang, W.; Xie, J.; He, X.; Ying, H. Recovery of acetoin from theaqueous solution by means of a novel hyper-cross-linked resin:Equilibrium and kinetics. J. Food Eng. 2013, 119 (4), 714−723.(17) Su, H.; Wang, Z.; Tan, T. Adsorption of Ni2+ on the surface ofmolecularly imprinted adsorbent from Penicillium chysogenummycelium. Biotechnol. Lett. 2003, 25 (12), 949−953.(18) Muslu, N.; Gulfen, M. Selective separation and concentration ofPd(II) from Fe(III), Co(II), Ni(II), and Cu(II) ions using thiourea−formaldehyde resin. J. Appl. Polym. Sci. 2011, 120 (6), 3316−3324.(19) Long, K. M.; Goff, G. S.; Ware, S. D.; Jarvinen, G. D.; Runde, W.H. Anion Exchange Resins for the Selective Separation of Technetiumfrom Uranium in Carbonate Solutions. Ind. Eng. Chem. Res. 2012, 51(31), 10445−10450.(20) Langmuir, I. The Constitution and Fundamental Properties ofSolids and Liquids. Part I. Solids. J. Am. Chem. Soc. 1916, 38 (11),2221−2295.



(21) Zhou, J.; Wu, J.; Liu, Y.; Zou, F.; Wu, J.; Li, K.; Chen, Y.; Xie, J.;Ying, H. Modeling of breakthrough curves of single and quaternarymixtures of ethanol, glucose, glycerol and acetic acid adsorption onto amicroporous hyper-cross-linked resin. Bioresour. Technol. 2013, 143,360−368.(22) Shirazi, D. G.; Felinger, A.; Katti, A. M. Fundamentals ofPreparative and Nonlinear Chromatography; Academic Press: NewYork, 2006.(23) Li, L.; Liu, F.; Jing, X.; Ling, P.; Li, A. Displacement mechanismof binary competitive adsorption for aqueous divalent metal ions ontoa novel IDA-chelating resin: Isotherm and kinetic modeling.Water Res.2011, 45 (3), 1177−1188.(24) Li, B.; Liu, F.; Wang, J.; Ling, C.; Li, L.; Hou, P.; Li, A.; Bai, Z.Efficient separation and high selectivity for nickel from cobalt-solutionby a novel chelating resin: Batch, column and competitioninvestigation. Chem. Eng. J. 2012, 195, 31−39.(25) Wu, X.; Arellano-Garcia, H.; Hong, W.; Wozny, G. n. Improvingthe Operating Conditions of Gradient Ion-Exchange SimulatedMoving Bed for Protein Separation. Ind. Eng. Chem. Res. 2013, 52(15), 5407−5417.(26) Lv, L.; Zhang, Y.; Wang, K.; Ray, A. K.; Zhao, X. S. Modeling ofthe adsorption breakthrough behaviors of Pb2+ in a fixed bed of ETS-10 adsorbent. J. Colloid Interface Sci. 2008, 325 (1), 57−63.(27) Strohlein, G.; Aumann, L.; Mazzotti, M.; Morbidelli, M. Acontinuous, counter-current multi-column chromatographic processincorporating modifier gradients for ternary separations. J. Chromatogr.A 2006, 1126 (1), 338−346.(28) Schmidt-Traub, H. In Preparative Chromatography of FineChemicals and Pharmaceutical Agents; Wiley−VCH Verlag: Weinheim,Germany, 2005.(29) Schmidt-Traub, H.; Schulte, M.; Seidel-Morgenstern, A.Preparative Chromatography; John Wiley & Sons: New York, 2012.(30) Suzuki, M.; Smith, J. Axial dispersion in beds of small particles.Chem. Eng. J. 1972, 3, 256−264.(31) Wilke, C.; Chang, P. Correlation of diffusion coefficients indilute solutions. AIChE J. 1955, 1 (2), 264−270.(32) Glueckauf, E. Theory of chromatography. Part 10.Formulaefor diffusion into spheres and their application to chromatography.Trans. Faraday Soc. 1955, 51, 1540−1551.(33) Danckwerts, P. Continuous flow systems: distribution ofresidence times. Chem. Eng. Sci. 1953, 2 (1), 1−13.(34) Li, X.; Zhang, L.; Chang, Y.; Shen, S.; Ying, H.; Ouyang, P.Kinetics of adsorption of thymopentin on a gel-type strong cation-exchange resin. Chromatographia 2007, 66 (3−4), 231−235.(35) Cayan, F. N.; Pakalapati, S. R.; Elizalde-Blancas, F.; Celik, I. Onmodeling multi-component diffusion inside the porous anode of solidoxide fuel cells using Fick’s model. J. Power Sources 2009, 192 (2),467−474.(36) Zhou, X.; Fan, J.; Li, N.; Qian, W.; Lin, X.; Wu, J.; Xiong, J.; Bai,J.; Ying, H. Adsorption Thermodynamics and Kinetics of Uridine 5′-Monophosphate on a Gel-Type Anion Exchange Resin. Ind. Eng.Chem. Res. 2011, 50 (15), 9270−9279.(37) Moreira, M. J. A.; Gando-Ferreira, L. M. Separation ofphenylalanine and tyrosine by ion-exchange using a strong-baseanionic resin. I. Breakthrough curves analysis. Biochem. Eng. J. 2012,67, 231−240.(38) Forlani, G.; Mantelli, M.; Nielsen, E. Biochemical evidence formultiple acetoin-forming enzymes in cultured plant cells. Phytochem-istry 1999, 50 (2), 255−262.(39) Speckman, R.; Collins, E. Separation of diacetyl, acetoin, and2,3-butylene glycol by salting-out chromatography. Anal. Biochem.1968, 22 (1), 154−160.(40) Faria, R. P.; Pereira, C. S.; Silva, V. M.; Loureiro, J. M.;Rodrigues, A. E. Glycerol valorisation as biofuels: Selection of asuitable solvent for an innovative process for the synthesis of GEA.Chem. Eng. J. 2013, 233, 159−167.(41) Abel, S.; Mazzotti, M.; Morbidelli, M. Solvent gradientoperation of simulated moving beds: I. Linear isotherms. J.Chromatogr. A 2002, 944 (1), 23−39.

(42) Wang, F. Y.; Zhu, Z. H.; Rudolph, V. Diffusion through orderedforce fields in nanopores represented by Smoluchowski equation.AIChE J. 2009, 55 (6), 1325−1337.



recovery of acetoin from the ethanol–acetoin–acetic acid ternary mixture based on adsorption...

Documents