separation of d‐psicose and d‐fructose using simulated moving bed chromatography

J. Sep. Sci. 2009, 32, 1987 –1995 N. Van Duc Long et al. 1987

Nguyen Van Duc Long1

Thai-Hoang Le1

Jin-Il Kim1

Ju Weon Lee2, 3

Yoon-Mo Koo1, 2

1Department of BiologicalEngineering, Inha University,Younghyundong, Namgu,Incheon, Republic of Korea

2ERC for Advanced BioseparationTechnology, Inha University,Younghyundong, Namgu,Incheon, Republic of Korea

3Department of ChemicalEngineering, Inha University,Younghyundong, Namgu,Incheon, Republic of Korea

Original Paper

Separation of D-psicose and D-fructose usingsimulated moving bed chromatography

Simulated moving bed (SMB) processes have been widely used in the sugar indus-tries with ion-exchange resin as a stationary phase. D-Psicose, a rare monosaccharideknown as a valuable pharmaceutical substrate, was synthesized by the enzymaticconversion from D-fructose. The SMB process was adopted to separate D-psicose fromD-fructose. Before the SMB experiment, the reaction mixture including D-psicoseand D-fructose was treated by a deashing process to remove contaminants, such asbuffers, proteins, and other organic materials. Four columns packed with Dowex50WX4-Ca2+ (200–400 mesh) ion-exchange resins were used in the four-zone SMB.Single-step frontal analysis was performed to estimate the isotherm parameters ofeach monosaccharide. The operating conditions of the SMB process were deter-mined based on the Equilibrium Theory. According to the simulation of the SMBprocess, the purity and yield of extract product (D-psicose) achieved were 99.04 and97.46%, respectively and those of raffinate product (D-fructose) were 99.06 and99.53%, respectively. Under the optimized operating condition, complete separation(extract purity = 99.36%, raffinate purity = 99.67%) was achieved experimentally.

Keywords: Chromatography / D-Psicose / D-Fructose / Separation / Simulated moving bed / SMB /

Received: December 23, 2008; revised: March 12, 2009; accepted: March 12, 2009

DOI 10.1002/jssc.200800753

1 Introduction

The simulated moving bed (SMB) technology, firstlypatented by universal oil products (UOP) as the Sorbexprocess [1], has emerged as a powerful tool for the contin-uous counter-current separation of binary mixtures.Compared with the batch chromatographic processes,the SMB process exhibits a number of advantages [2]. Inparticular, these are due to the continuous nature of theoperation and to an efficient use of the stationary andmobile phases, which allows the decrease of solvent con-sumption and the improvement of productivity. Solventsaving up to 90% and increases of productivity up to sev-eral times have been reported [2–4]. In recent years, SMBtechnology has found a wide range of applications [5], forinstance, the separation of lactic acid from acetic acid [6],separation of chiral drugs [7, 8], and it is now a key tech-nology for the separation of sugars using ion-exchangeresin.

The basic concept of a SMB process is to simulate thesolid phase movement of the corresponding true movingbed (TMB) where the fluid and solid phases move coun-

ter-currently [9]. The SMB process, schematically drawnin Fig. 1, consists of multiple identical packed columns,which are circularly interconnected to form a closedloop. By periodically switching the inlet and outlet portsto the same direction of the fluid flow, counter-currentcontact between the solid and liquid phases is simulated.As a result, not only do SMB processes benefit from thecounter-current contact between the solid and fluidphases, but also overcome the mechanical difficulties

Correspondence: Professor Yoon-Mo Koo, Department of Biolog-ical Engineering, Inha University, 253 Younghyundong, Namgu,Incheon 402-751, Republic of KoreaE-mail: [email protected]: +82-32-872-4046

i 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.jss-journal.com

Figure 1. Schematic illustration of four-zone SMB unit.

1988 N. Van Duc Long et al. J. Sep. Sci. 2009, 32, 1987 – 1995

related to the movement of the solid phase in a TMB proc-ess [5].

D-Psicose, a carbon-3 epimer of D-fructose, is a raresugar because D-psicose and its derivatives rarely exist innature. It is present in commercial carbohydrate or agri-cultural products in small quantities [10]. D-Psicose has awide range of applications, such as the source of D-alloseproduction which can be applied in the treatment ofmyeloid leukemia [11, 12], and as an alternative sweet-ener for weight reduction due to its zero calories [13, 14].D-Psicose is also used for repressing hepatic lipogenicenzyme activity [15]. The noncharacterized D-psicose 3-epimerase gene from Agrobacterium tumefaciens wascloned and expressed in Escherichia coli [10]. The expressedenzyme was purified by three-step chromatographybefore being used for D-psicose production. The goal ofthis work was to study the continuous separation of theD-psicose and D-fructose mixtures in a SMB unit.

2 Theory

With reference to the separation of binary mixtures, themore retained component (A) is collected in the extractstream, while the less retained component (B) is collectedin the raffinate stream. Each zone of the SMB processplays a specific role during the operation. Desorbent isused to desorb component A in zone 1 and to regenerateadsorbent. Component B is adsorbed in zone 4 and desor-bent is regenerated before recycling into zone 1.

Both batch and SMB chromatographic processes canbe modeled by the differential mass balance equation[16]

qci

qtþ F

qqi

qtþ u

qci

qx¼ DL;i

q2ci

qx2ð1Þ

where ci and qi are the concentration of component i inthe mobile and stationary phases, respectively, F is thephase ratio (= (1–e/e), e is the void fraction of the col-umn, u is the interstitial velocity, DL,i is the axial disper-sion coefficient of component i. For mass transferbetween the mobile and stationary phase, it is commonto use a linear lumped mass transfer model.

qqi

qt¼ apkeff ;iðc� c�Þ ¼ 6keff ;i

dpðc� c�Þ ð2Þ

where apkeff,i is the mass transfer coefficient of solute i, ap

the external surface area of the particles per unit volume(i. e., 6/dp for spherical particles), and c�i is the concentra-tion of solute i in the liquid film.

Lapidus and Amundson solved these equations for lin-ear isotherms and local equilibrium [17].

ciðz; tÞ ¼cF;i

21� erf

z� us;itffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

4Dap;ius;it=up

" #( )

ð3Þ

us;i ¼u

1þ 1� e

eKi

ð4Þ

where cF,i is the concentration of solute i in the feed, uS,i

the wave velocity of solute i, Dap,i the apparent axial dis-persion coefficient of solute i and Ki is the partition coef-ficient of solute i. The axial dispersion coefficient wasestimated by the Chung and Wen correlation [18].

Pe ¼ L2rpe

ð0:2þ 0:011Re0:48Þ ð5Þ

Re ¼ 2rpvlql

ellð6Þ

Pe ¼ vlLeDL

ð7Þ

where Pe is the Peclet number, Re the Reynolds number,and vl is the liquid superficial velocity. In chromatogra-phy, the Reynolds number is usually less than 0.1. Theapparent axial dispersion coefficient can be estimatedfrom either of the following relation [19].

Dap;i ¼ DL;i þu2Ki

apkeff ;i

ki

ð1þ kiÞ2ð8Þ

ki ¼1� e

eKi ð9Þ

where DL,i is the axial dispersion coefficient, ki is the reten-tion factor.

SMB design method based on the equilibrium theoryoffers simple design criteria for the operation of SMBunits [16]. In its framework, the performance of the SMBunit is characterized by the ratio of the net flow-rate ofthe mobile phase to that of the stationary phase, i. e. themj value which is defined as

mj ¼Q jt� � Ve

Vð1� eÞ ð10Þ

where Qj is the volumetric flow rate of the mobile phasein section j, t* the switching time, and V and e are the col-umn volume and the void fraction of the column, respec-tively. The adsorption mechanisms of carbohydrates onion-exchange resin are the combination of three exclu-sions called the size, ion, and ligand-exclusions. In theseexclusion mechanisms, the retention times of carbohy-drates are not changed according to the changing of thefeed concentration [20]. The adsorption equilibrium canbe represented by a linear isotherm

qi ¼ Hici ð11Þwhere Hi is the Henry constant of solute i. Under theseconditions, the necessary and sufficient operating condi-tions for complete separation in a SMB unit are deter-mined by the following inequalities [2]

HAa m1a 1 ð12Þ


J. Sep. Sci. 2009, 32, 1987 –1995 Other Techniques 1989

HBa m2a HA ð13Þ

HBa m3a HA ð14Þ

�ep

1� epa m4a HB ð15Þ

where A and B represent the more retained and lessretained solutes, respectively. The m1 and m4 values areeasily controlled by changing the flow rates of desorbent,raffinate, and extract streams. The m2 and m3 are relatedto the switching times of ports and flow rates of recycleand feed streams [20]. In the m2 –m3 plane, we can draw atriangular region called the separation region. One ofthe key issues in operating SMB process is the determina-tion of zone flow rates and switching time. Developed inthe frame of equilibrium theory, the triangle theory,which neglects the effect of axial mixing and mass trans-fer resistances, is one of the widely applied approachesfor SMB design [21].

3 Experimental

3.1 Materials

The reaction mixture including D-psicose and D-fructosewas obtained from Konkuk University (Korea). Calciumchloride 96% (powder), sodium hydroxide 98% (powder)and hydrochloric acid 37%, purchased from SamchunCo. (Korea) were used to treat the resin. Distilled anddeionized water (DDW) as the mobile phase was obtainedfrom a Milli-Q system (Millipore, U.S.A.). Dowex 50WX4H+ cation exchange resin (Sigma–Aldrich Co., U.S.A.) wasused to separate the binary sugar mixture which waspacked into the jacketed glass column (2.5 cm/ID619.5 cm bed height).

3.2 Apparatus

3.2.1 HPLC apparatus

The samples collected for the frontal analysis and SMBexperiment were analyzed using a 8 mm/ID6300 mmbed height SP0810 column filled with rigid styrene/DVBbased strong cation exchange resin (Showa Denko,Japan). The HPLC system was composed of a system con-troller (SCL-10Avp, Shimadzu, Japan), a pump (LC-6AD,Shimadzu, Japan), a RI detector (RID-10A, Shimadzu,Japan), an auto injector (SIL-10ADvp, Shimadzu, Japan).Temperature was controlled by a refrigerated circulatingbath (WCR-P22, Daihan, Korea).

3.2.2 SMB system

The laboratory-scale SMB unit was constructed at theSMB laboratory, Inha University. The scheme of the SMB-unit is shown in Fig. 1. It is a continuous chromato-graphic system constituted by four columns connected

in series. The jacketed glass columns (2.5 cm/ID625 cmheight) was purchased from Omnifit (U.S.A.) and werepacked with Dowex 50WX4 (Ca2+ form) by using a slurrymethod. Each column is connected with four lines (elu-ent, feed, extract, and raffinate lines) via a 4-way connec-tor. Two ECMT 16-port valves (Valco Instrument Inc. Co.,U.S.A.) were used inside the unit. The simulated counter-current was achieved by 16 port switching valves con-nected to the two inlet streams and two outlet steams inthe required manner. Four M50 pumps (Valco Instru-ment Inc. Co., U.S.A.) were positioned and installed infront of each column for the accurate control of zoneflow rate. The system temperature was controlledthrough a refrigerated circulating bath (WCR-P22, Dai-han, Korea). To collect the extract and raffinate fractions,a Foxy 200 fraction collector was used. The SMB systemwas controlled by a control software which was made bythe Inha SMB Lab.

3.3 Method

3.3.1 HPLC assay

HPLC was used to analyze the collected fractions fromboth batch and SMB experiments. In this assay, a sugarSP0810 column and DDW, at a flow rate of 1.5 mL/minand a temperature of 858C, were used. The sample injec-tion volume was 5 lL. Prior to analysis, all solvents weredegassed for approximately 45 min, and the column waspre-equilibrated with the mobile phase until a smoothbase line was obtained.

3.3.2 Deashing D-psicose and D-fructose mixture

Besides D-psicose and D-fructose, the solution obtainedfrom the enzymatic reaction contains many other com-ponents such as buffer, cations, anions, and other con-taminants which can adversely affect the chromato-graphic behaviors. Therefore, adsorption and ionexchange method were used to remove the contami-nants (Fig. 2) [22]. An �KTA FPLC system (GE Healthcare,U.S.A.) was used in the D-psicose and D-fructose mixturesolution treatment. This system consists of two pumps (P-920), an injection valve (INV-907), and monitor (UPC-900)which is a high precision on-line monitor for the com-bined measurement of UV absorption and conductivity.

A 3 L D-psicose and D-fructose mixture solution waspumped at a flow rate of 7 mL/min into three columns(2.5 cm/ID630.0 cm bed height) which contained DowexOptipore SD-2 resin, Dowex Monosphere 88 resin andDowex Monosphere 66 resin, respectively. The resinmust be regenerated before carrying out the deashingprocess. For the Dowex 88 strong acid cation resin, regen-eration involves forcing hydrogen ions back onto theexchange sites by pumping hydrochloric acid throughthe bed. In regenerating the Dowex 66 wk base anion



resin and Dowex Optipore SD-2 resin, sodium hydroxidewas used.

When D-psicose and D-fructose mixture solution waspumped into three columns, Dowex Optipore SD-2 resinadsorbed impurities that contribute to the formation ofcolor and off-flavors. Dowex 88 resin readily exchangedits hydrogen ions for other undesirable cations, such asmanganese added to the substrate to strongly enhance D-fructose epimerization as well as sodium, potassium,and other cations. The cation resin also removed somesoluble protein by providing an acidic environment inwhich amphoteric proteins take on cationic characteris-tics. The anions, e.g. chlorides, bromide and other nega-tively charged ions associated with the exchanged cati-ons, were passed on in the solution. Consequently, afterthe solution left the cation bed, Dowex 66 resin in thefree base form removed these anions. As a result, impur-ities picked up during isomerization and pH adjustmentwere removed. Therefore, the deashed solution was usedas a purified feed solution. After the pump was stopped,the columns were washed with distilled and deionizedwater. All the work was done at room temperature.

After carrying out the deashing process, protein con-centrations were determined by the Bradford methodusing BSA as a standard protein.

3.3.3 Determination of dead volume

Dowex 50WX4 Ca2+ ion exchange resin was packed intothe 195 mm long and 25 mm/ID column by the slurrypacking method. An �KTA FPLC pump (GE Healthcare)was used to deliver the mobile phase (DDW) at 3 mL/min.The conductivity detector was used to monitor the his-tory of the stream. The results were obtained in the elu-tion profiles of CaCl2 0.5 M which was injected in one ofthe SMB columns. CaCl2 is a molecule with molecularmass of 110.98 g/g mol and is not retained in Dowex50WX4 Ca2+ form. Thus, it is capable of diffusion throughthe interstitial spaces between particles, as well as

through the particle pores. Total column porosity wasmeasured as 0.470.

3.3.4 Frontal analysis

Among the various chromatographic methods availableto determine single-component isotherms, frontal anal-ysis (FA) is the most accurate [23]. For chromatographicsystems with low efficiency, FA is the most suitablemethod because of its accuracy and its relative simplicity[24]. The FA technique was used to determine the adsorp-tion equilibrium isotherm of D-fructose and D-psicose onthe ion-exchange resin Dowex 50WX4 Ca2+ form.

The adsorption equilibrium isotherms for D-fructoseand D-psicose were studied at 508C with a temperature-controlled water circulator. The eluent used was deion-ized water with a flow rate of 3.0 mL/min. The experi-ments were carried out using four different concentra-tions for each sugar which are described in Table 1. Thefeed concentration ranges of D-psicose and D-fructosewere 8.90 to 42.34 g/L and 23.23 to 116.13 g/L, respec-tively. Every 3.0 mL of elution mobile phase was frac-tioned and measured the concentration by HPLC anal-ysis.

3.3.5 SMB experiments

After degassing the feed and desorbent solution, bothwere continuously pumped into the columns. The feedwas a binary mixture of D-psicose and D-fructose of whichthe concentrations were 39.78 and 100.41 g/L, respec-tively, at a flow rate of 1 mL/min. Table 3 presents experi-mental conditions used in the SMB operation. The desor-bent was DDW. Fractions from the sampling port, the raf-finate outlet and extract outlet were collected with onefraction per one switching time after nine full cycles ofcontinuous operation. After nine cycles and a half ofswitching time, the remaining mobile phase in the tub-ing between columns was collected to measure the col-


Figure 2. Schematic illustration of deashing process.

Table 1. The feed conditions of frontal analysis.

Run 1 Run 2 Run 3 Run 4

D-Psicose (g/L) 42.34 27.91 19.81 8.90D-Fructose (g/L) 116.13 76.27 52.96 23.23Feed Volume (mL) 120 120 120 120

Table 2. Isotherm and kinetic parameters of D-fructose andD-psicose.

D-Psicose D-Fructose

Isotherm parameter (H) 1.337 0.503Axial dispersion coefficient (DL, cm2/min) 0.022 0.022Apparent axial dispersion coefficient(Dap, cm2/min)

0.046 0.050

Constant mass transfer coefficient(apkeff, 1/min)

8.316 19.513


umn profiles. All the fractions and the feed were ana-lyzed using the HPLC method. The experimental per-formance parameters were determined by analysis of theextract and raffinate samples collected during the lastfive cycles (after cyclic steady-state was achieved).

4 Results and discussions

4.1 Adsorption isotherms and mass transferparameters

On the ion exchange resin, Dowex 50WX4 Ca2+ form, sin-gle-step frontal analysis was carried out with four differ-ent concentrations of feed solution. All frontal analyseswere carried out with D-psicose and D-fructose mixturesolution. If the adsorption isotherms and mass-transferparameters are accurate and the assumptions of the ratemodel are valid, simulations would give reliable predic-tions of transient column profiles, effluent histories,product purity, and yield [25]. Figure 3 shows the elutionprofile of the frontal analysis for D-fructose and D-psicose,

respectively. The adsorption behaviors of D-psicose and D-fructose are observed as linear isotherm within the con-centration range to 42.34 g/L of D-psicose and 116.13 g/Lof D-fructose. The equilibrium isotherms, obtained forthe adsorption of D-fructose and D-psicose on Dowex50WX4 Ca2+ form, are presented in Fig. 4. The linearregression curves show good agreement with experi-ment data. According to the linear regression, theHenry's constants of D-psicose and D-fructose areobtained as 1.337 and 0.503, respectively. The axial dis-persion coefficients were estimated from the Chung andWen correlation, Eq. (5). The apparent axial dispersioncoefficients were estimated using frontal analyses of D-fructose and D-psicose with the concentrations of23.23 g/L and 8.90 g/L, respectively, and carrying out thefitting using Origin 7.5. After that, the mass transferparameters of D-psicose and D-fructose were estimated byEq. (8). Figure 3 shows the comparisons of the experimen-tal elution profiles and the simulated elution profiles ofthe frontal analyses of D-fructose and D-psicose. The simu-lated elution profiles followed the experimental resultsprecisely. The dispersion coefficients and mass transfercoefficients are shown in Table 2.

4.2 Results and analysis of the SMB experiments

The triangle theory was used to determine the operatingcondition of SMB process. The triangle region describesthe separation zone and the vertex point is the optimumoperating condition for the SMB operation in terms ofmaximum productivity (Pr)

Pr ¼ Amount of Product per TimeVolume of Adsorbent

ð16Þ

Although optimal operating conditions may be wellchosen by virtue of good measurement of isotherm


Figure 3. The comparisons of the experimental and simu-lated elution profile of D-fructose (a) and D-psicose (b).

Figure 4. The comparisons of the experimental data and cal-culated isotherm curves of D-psicose and D-fructose.


parameters, there still exists a difficult problem. TheSMB unit is rather sensitive to a number of disturbances,e. g., change in feed concentration and changes in iso-therm parameters, which may result from aging of pack-ing material or temperature changes. Consequently, it isa common strategy to keep the actual operating point ofa SMB process at a reasonable distance from these opti-mal operating conditions to guarantee a robust separa-tion within the specification [2].

The SMB productivity is proportional to the distancebetween the operating point and the diagonal of the m2 –m3 plane [2]. The purity and yield characterize the degreeof separation. Productivity, enrichment and desorbentconsumption describe the efficiency of the separation.To determine the operating conditions, Aspen Chroma-tography 2006m is used for the simulation of the SMBprocess. During the optimization of the SMB, the param-

eter, purity and yield were improved firstly. As soon asthey are acceptable (at about 99%), the quality of separa-tion was improved by increasing the productivity anddecreasing the desorbent consumption. Consequently,the enrichment of products also increased. Therefore, weapplied a 15.88% safety margin from the vertex pointtoward the diagonal line to decide the operating point.

According to the fixed feed flow rate of 1 mL/min, theseparation region was calculated in the m2 –m3 plane todecide the operating condition of SMB process. Figure 5shows the separation region and the operating point inthe m2 –m3 plane. The operating conditions are describedin Table 3. We assumed that all the columns have thesame properties, such as the dimensions and total voidfraction. The SMB experiment to separate D-psicose andD-fructose was operated during nine cycles. Figure 6shows the comparison of the experimental and simu-lated history profiles of extract and raffinate. History pro-files reached to the cyclic steady-state after four cycles. Inthe beginning of the switching time, the D-psicose con-centration at the extract port was high because the con-centration profile in Fig. 7 moves to the right and the


Table 3. Operating conditions and system parameters of theSMB unit.

Feed concentration (g/L) 39.79 (D-psicose)100.41 (D-fructose)

SMB configuration 1-1-1-1Temperature (8C) 508CLength of packed section (cm) 19.5Internal diameter of packed section (cm) 2.5Inter-particle/external voidage (m3 void/m3 bed) 0.47Particle radius of adsorbent (micron) 37.00Feed flow rate (mL/min) 1.00Desorbent flow rate (mL/min) 1.49Extract flow rate (mL/min) 1.21Raffinate flow rate (mL/min) 1.28Recycle flow rate (mL/min) 1.86Switching time (min) 36.00

Figure 5. Separation region on m2 –m3 plane. Square symbolis the operating point and circle symbol is simulation point.

Figure 6. The comparisons of experimental and simulatedextract (a) and raffinate (b) history profiles.


concentration of the extract slowly decreases to mini-mum just before the next switching time. In contrast tothis, the concentration of D-fructose sharply increasesduring every switching time. Figure 7 shows the compar-ison of the SMB experimental and simulated column pro-file. At the cyclic steady-state, 99.36 and 99.67% puritieswere obtained in extract and raffinate, respectively.Purity is defined as

PUX ¼ cD�psicoseExtract

cD�psicoseExtract þ cD�fructose

Extract

6100 ð17Þ

PUR ¼ cD�fructoseRaffinate

cD�psicoseRaffinate þ cD�fructose

Raffinate

6100 ð18Þ

where PUX and PUR are the purity of D-psicose and D-fruc-tose in the extract and raffinate streams, respectively,when the cyclic steady state has been reached. Table 4presents the simulation and experimental performanceparameters for the SMB operation. The possibility torecover 99.72 and 98.56% of D-psicose and D-fructose,respectively, from the post-reaction mixture was proved.The productivity obtained was 57 g of D-psicose per daywhen this system operates continuously.

The simulation runs were carried out with variousdesorbent flow rates. Eleven operating points have beenchosen which consisted of nine points were within theestimated complete separation region. Figure 5 showsthe separation region and eleven operating points. Theyhad the same distance between themselves and the diag-

onal. For the production purposes, D-psicose is of inter-est, two operating points located outside the triangleregion of complete separation where only the extractport is pure were investigated. Table 5 presents operationconditions and performance of the simulation runs. Fig-ure 8 shows the D-psicose purity increases from 96.50 to99.63% as desorbent flow rate increases. A simulationobservation is that, when the operating point is outsidebut close to the boundary of complete separation, thepurity of D-psicose increases, but there is no change ofthe productivity of D-psicose.


Figure 7. The comparisons of experimental and simulationelution profiles.

Table 4. Simulation and experimental results

Purity (%) Yield (%) Productivity(kg/L/h)

Desorbent(L/cycle)

Enrichment

Simulation D-fructose 99.06 99.53 2.94610 – 2 0.21 0.82D-psicose 99.04 97.46 1.14610 – 2 0.76

Experiment D-fructose 99.67 98.56 2.94610 – 2 0.21 0.81D-psicose 99.36 99.72 1.20610 – 2 0.78

Table 5. Operation conditions and performance of the simulation runs

Run Flow rates (mL/min) Tsw (min) Purity (%) Productivityof D-psicose

Desorbentconsumption

F D E R Re D-psicose D-fructose (kg/L/h) (mL/cycle)

Case 1 1.00 1.59 1.28 1.31 1.86 36.00 99.63 96.80 1.14610 – 2 228.96Case 2 1.00 1.57 1.28 1.29 1.86 36.00 99.59 97.31 1.14610 – 2 226.08Case 3 1.00 1.55 1.28 1.27 1.86 36.00 99.53 97.77 1.14610 – 2 223.20Case 4 1.00 1.53 1.28 1.25 1.86 36.00 99.45 98.17 1.14610 – 2 220.32Case 5 1.00 1.51 1.28 1.23 1.86 36.00 99.27 98.60 1.14610 – 2 217.44Case 6 1.00 1.49 1.28 1.21 1.86 36.00 99.04 99.06 1.14610 – 2 214.56Case 7 1.00 1.47 1.28 1.19 1.86 36.00 98.94 99.14 1.14610 – 2 211.68Case 8 1.00 1.45 1.28 1.17 1.86 36.00 98.61 99.22 1.14610 – 2 208.80Case 9 1.00 1.43 1.28 1.15 1.86 36.00 98.14 99.37 1.14610 – 2 205.92Case 10 1.00 1.41 1.28 1.13 1.86 36.00 97.47 99.45 1.14610 – 2 203.04Case 11 1.00 1.39 1.28 1.11 1.86 36.00 96.50 99.52 1.14610 – 2 200.16


5 Conclusions

After the enzymatic reaction, feed solution containingimpurities picked up during isomerization and pHadjustment was efficiently treated by adsorption and ionexchange resins. The isotherms under linear regimeswere measured on an analytical column using cationexchange resin as the stationary phase. All isothermparameters were measured by frontal analysis. Theseresults were used to identify regions of complete separa-tion for the SMB operation. Besides, unknown modelparameters – total void fraction, axial dispersion coeffi-cients and mass transfer coefficients – were estimated.The SMB process for separation of D-psicose and D-fruc-tose was designed and simulated based on the Triangletheory. The operating conditions were decided using a15.88% safety margin in the triangle separation region.Separation of D-psicose and D-fructose from binary mix-ture was successfully performed by using SMB chroma-tography. At the steady state, purity and yield of D-psicosewere 99.36 and 99.72%, respectively, and those of D-fruc-tose were 99.67 and 98.56%, respectively. The simulatedcolumn profiles and the history profiles of the extractand raffinate streams exhibited a good agreement withthe experimental results. It was further demonstratedthat the triangle theory is well suited for the design ofSMB units. It provides exact criteria to locate the regionin the operating parameters space, which allows achiev-ing the desired separation performance.

Nomenclature

apkeff mass transfer coefficient (min – 1)ci fluid phase concentration of solute i (g/L)c�i fluid phase concentration of solute i in the liquid

film (g/L)Dap,i apparent axial dispersion coefficient of solute i

(cm2/min)

dp diameter of the adsorption particle (cm)DL axial dispersion coefficient (cm2/min)H Henry constantk retention factorK partition coefficientL column length (cm)m flow rate ratio between the desorbent and adsorb-

ent flow ratePe Peclet numberPr productivity (kg/L/h)PUR raffinate purity (%)PUX extract purity (%)q solid phase concentration of solute (g/L)Q volumetric flow rate (mL/s)Re Reynold numbert* switching time (min)u interstitial velocity of the mobile phaseV column volume (mL)v the liquid superficial velocity (cm/min)

Greek letters

e the total void fraction of the column

Subscripts and superscripts

D desorbentE extractF feedi componentj section of SMBR raffinateRe recycle

This study was supported by the ERC for Advanced BioseparationTechnology, KOSEF. The authors are grateful to Prof. Deok-Kun Oh(Konkuk University, Korea) for kindly providing the D-psicose andD-fructose mixture and to Mr. Michael Hladik for his assistance inpreparing this paper.

The authors declare no conflict of interest

6 References[1] Broughton, D. B., Gerhold, C. G., U.S. Patent No. 2,985,589, 1961.

[2] Mazzotti, M., Storti, G., Morbidelli, M., J. Chromatogr. A 1997, 769,3 – 24.

[3] Nicoud, R. M., Majors, R. E., LCGC 2000, 18, 680 – 687.

[4] Negawa, M., Shoji, F., J. Chromatogr. A 1992, 590, 113 – 117.

[5] Ching, C. B., Ruthven, D. M., J. Chem. Eng. Sci. 1989, 44, 1011 –1038.

[6] Lee, H. J., Xie, Y., Koo, Y. M., Wang, N. H. L., J. Biotechnol. Prog. 2004,20, 179 – 192.

[7] Francotte, E., Richert, P., Mazzotti, M., Morbidelli, M., J. Chroma-togr. A 1998, 796, 239 – 248.

[8] Khattabi, S., Cherrak, D. E., Mihlbachler, K., Guiochon, G., J. Chro-matogr. A 2000, 893, 307 – 319.


Figure 8. Influence of desorbent flow rate on purity of D-psi-cose and desorbent consumption.


[9] Pais, L. S., Loureiro, J. M., Rodrigues, A. E., J. Chromatogr. A 1998,827, 215 – 233.

[10] Kim, H. J., Hyun, O. K., Kim, Y. S., Lee, Y. J., Oh, D. O. K., J. Appl. andEnviron. Microbiol. 2006, 72, 981 – 985.

[11] Takeshita, K., Shuga, A., Takada, G., Izumori, K., J. Biosci. Bioeng.2000, 90, 453 – 455.

[12] Itoh, H., Sato, T., Izumori, K., J. Ferment. Bioeng. 1995, 90, 101 –103.

[13] Matsuo, T., Suzuki, H., Hashiguchi, M., Izumori, K., J. Nutr. Sci.Vitaminol. (Tokyo) 2002, 48, 77 – 80.

[14] Matsuo, T., Tanaka, T., Hashiguchi, M., Izumori, K., Suzuki, H., J.Clin. Nutr. 2003, 12, 225 – 231.

[15] Matsuo, T., Baba, Y., Hashiguchi, M., Takeshita, K., Izumori, K.,Suzuki, H., J. Clin. Nutr. 2001, 10, 233 – 237.

[16] Storti, G., Mazzotti, M., Morbidelli, M., Carra, S., AIChE J. 1993, 39,471 – 492.

[17] Lapidus, L., Amundson, N. R., J. Phys. Chem. 1952, 56, 984 – 988.

[18] Chung, S. F., Wen, C. Y., AIChE J. 1968, 14, 857 – 866.

[19] Dunnebier, G., Weirich, I., Klatt, K. U., J. Chem. Eng. Sci. 1998, 53,2537 – 2546.

[20] Lee, J. W., Kim, J. I., Lee, C. H., Kim, J. K., Koo, Y. M., J. Liq. Chroma-togr. Relat. Technol. 2007, 30, 1 – 19.

[21] Migliorini, C., Mazzotti, M., Morbidelli, M., J. Chromatogr. A 1998,827, 161 – 73.

[22] Technical manual – Dow liquid separations, 177-01563-602QRP, June, 2002.

[23] Guiochon, G., Shirazi, S. G., Katti, A. M., Academic Press, Boston,M. A 1994, 80 – 96.

[24] Mihlbachler, K., Kaczmarski, K., Seidel-Morgenstern, A., Guio-chon, G., J. Chromatogr. A 2002, 955, 35 – 52.

[25] Wang, N. H. L., Xie, Y., Ho, N., Final Report, DE-FC36-01GO11071,A000, December, 2003.


separation of d‐psicose and d‐fructose using simulated moving bed chromatography

Documents