Sugar separations on a pilot scale by continuous ... - ACS Publications

Jan 1, 1990 - Charles H. Byers, Warren G. Sisson, Joseph P. DeCarli II, and Giorgio Carta. Biotechnol. Prog. , 1990, 6 (1), pp 13–20. Publication Da...
0 downloads 0 Views 907KB Size
Biotechnol. Prog. 1990, 6, 13-20

13

Sugar Separations on a Pilot Scale by Continuous Annular Chromatography Charles H. Byers* and Warren G. Sisson Chemical Technology Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831

Joseph P. DeCarli I1 and Giorgio Carta Department of Chemical Engineering, University of Virginia, Charlottesville, Virginia 22901

A pilot-scale continuous annular chromatograph has been developed for scale-up and optimization studies. The chromatographic apparatus consists of a slowly rotating annular bed of sorbent material. The feed mixture which is to be separated is continuously introduced a t a stationary point a t the top of the bed, while eluent flows over the remainder of the annulus. The rotation of the sorbent bed coupled with the elution development induced by the downflow of eluent causes the feed components to appear as individual helical bands, each of which has a characteristic, stationary exit point. The separation of sucrose, glucose, and fructose, using the calcium form of a cation-exchange resin, was investigated using water as the eluent. Experiments were carried out both with a synthetic mixture of the sugars as well as with an industrial feedstock. The separation was explored first in a bench-scale conventional fixedbed column and then in the pilot-scale continuous chromatograph. Factors such as column loading, feed-to-eluent ratio, and feed concentration were investigated both experimentally and theoretically. The results of bench-scale experiments were carried out in the limit of infinite dilution, almost perfectly scaled to the pilot unit when loading was low and feed mixtures were of low viscosity. Very significant deviations from ideality were, however, observed a t high feed loadings, when the feed viscosity exceeded that of the eluent by more than a factor of 2.

Introduction Liquid chromatography has been for many years the method of choice for a variety of analytical separations. More recently, chromatographic techniques have been extended to production-scale operations. An example in the food-processing industry, where significant industrial success of chromatography is evident, is in the glucose-fructose separation. The most common type of fructose syrup, usually called high-fructose corn syrup (HFCS), typically contains about 42% fructose, 52% glucose, and 6% oligosaccharides on a dry-weight basis. It is used primarily as a sucrose replacement in some foods and beverages, since fructose is considerably sweeter than sucrose.’ Recently, however, there has been an increasing demand for both purer fructose syrup and for crystalline fructose, and a variety of chromatographic processes have been devised for recovering enriched fractions from the same HFCS source. Several other potentially significant sugar separations are discussed by Sherman and Chao.2 A number of chromatographicmethods have been developed for sugar separations. The primary focus has been on alkali-metal-exchanged, strong-acid-type cationexchange resins like the calcium form of Dowex 50W-X8 used in the current study as a chromatographic More details are iven in our previous studys and by Welstein and Sauer.8 A significant number of industrial plants employing chromatography for sugar separations are now in existence. For instance, Finnish Sugar Company has designed

* Author to whom correspondence should be addressed.

and built several chromatographic systems for the separation of sucrose from beet molasses and to recover highpurity fructose from isomerized feedstock. American Xyrofin has a plant in Thomson, IL, which produces 27.5 million lb of crystalline fructose/year; the chromatographic separation system used in this plant consists of 17 separation columns and 8 crystallizers.” This system converts a high-dextrose syrup to fructose by using immobilized glucose isomerase; fructose is then isolated from other sugars in the chromatographic columns. Because of the large scale of these operations, a continuous or nearly continuous operation is desirable from the viewpoints of efficient sorbent utilization and simplicity of operation. A practical means of achieving nearly continuous operation is the UOP “SORBEX”p r o c e ~ s , l l - ~ ~ known as “SAREX” in its application to sugar fractionation. In this process, an effective countercurrent operation is provided by the sequential movement of feed and exit ports through a series of fixed beds in the direction opposite to that of fluid flow. The process has been successfully applied to the separation of glucosefructose mixtures both using zeolites and ion-exchange resins. The countercurrent contacting effect ensures efficient sorbent utilization and minimal dilution of the products. However, the process allows only two pure products, and multiple units are needed to fully resolve a multicomponent mixture. Further, the process is mechanically complex and may not be suitable for all scales of application. A simple, process-scale continuous chromatographicprocess, which is able to deal with multicomponent streams, may be desirable for some application in this industry.

8756-7938/90/3006-0013$02.50/0 0 1990 American Chemical Society and American Institute of Chemical Engineers

14

Biotechnol. Prog., 1990, Vol. 6,No. 1

With this goal in mind, some studies have been carried out to explore the separation of sugars, using continuous annular This process operates at steady state and potentially allows the full range of features that are characteristic of liquid chromatography; specifically, the process can be used to perform multicomponent separations on a continuous basis. The concept of continuous annular chromatography (CAC), which is discussed el~ewhere,~ involves an annular bed of sorbent particles, packed in the space between two concentric cylinders. The annular bed assembly is slowly rotated about its axis, while eluent and feed solutions are continuously supplied to the top of the bed. In isocratic operation, the eluent is supplied uniformly to the entire circumference, while the feed mixture is supplied only to a sector of the circumference which remains stationary in space. As time progresses, helical component bands develop from the feed point, with slopes dependent upon eluent velocity, rotational speed, and the distribution coefficient of the components. These bands are stationary in space and leads to fixed exit points at the bottom of the annular bed. Under constant conditions, the angular distance from the feed sector of the exit point of each species remains constant, and the separated components can be recovered continuously by means of stationary collection trays. Sugar separations by continuous annular chromatography have been demonstrated by Howard and cow o r k e r ~using ~ ~ a bench-scale CAC apparatus with an annular bed 29 cm in diameter and 60 cm in height. The continuous separation performance observed with the experimental CAC unit was essentially identical with that of a conventional fixed-bed chromatograph of the same height and operated at the same superficial velocity. These studies, however, were carried out exclusively with synthetic sugar mixtures and were limited to low feed concentrations and to modest column loadings. In scalingup the process, a number of factors need to be taken in consideration. Firstly, at increased column loadings, which in the CAC are obtained when a wide sector of the annulus receives feed, the resolution is reduced. This effect is, of course, analogous to the overloaded condition of a conventional chromatograph. Secondly, as the feed concentration is increased, flow distribution problems may arise as a result of a significant difference in viscosity between feed and eluent. In a conventional fixed-bed chromatograph, such flow nonidealities may appear in the radial direction of the column as a result of viscous fingering. In the CAC, additional band spreading may occur in the angular direction, and its extent needs to be ascertained. Thirdly, at increased concentrations, deviations from the limiting Henry’s law equilibrium line may arise, potentially causing considerable discrepancy with the commonly used linear models extrapolated from low concentration/low loading results. Finally, the feasibility of operating the CAC with industrial feedstocks should be demonstrated to elucidate the possible effects of impurities present in the feed mixture. These scale-up issues are addressed in the experimental and theoretical investigation reported in this paper.

Experimental Section An important objective of this investigation is a comparison of the performance of a conventional fixed-bed column with that of a pilot-scale CAC unit. Thus, both types of facilities were developed for this work and are described in this section. A bench-scale fixed-bed column was used, consisting of a straight stainless steel tube, 0.95-cm i.d. and 107 cm

h Figure 1. Construction detail of the continuous annular chromatograph. The unit has a 445-mm outside diameter, an annulus width of 31.7 mm,and a usable bed depth of 110 cm.

long. The column was equipped with a Reodyne sixport valve and a Milton Roy Constametric IIIG doublecam metering pump. Dowex 50W-X8 resin with a narrow particle size range (average 50 gm) was used for these experiments. The resin was first thoroughly washed, then converted to the calcium form, and finally slurry-packed in the stainless steel tube. The effluent from the column was continuously monitored with a refractive index detector (Waters Model 401), and samples were collected with a fraction collector for subsequent HPLC analyses. The CAC apparatus used in this study is shown in Figure 1. The annulus (44.5-cm 0.d.) is formed with two concentric open cylinders, both of which are flanged at the bottom and constructed of Carpenter-20 steel, pressure rated for 1.15 MPa. The inner cylinder, which is 10 cm shorter than the outer one, is closed at the top, allowing head room for stationary feed and eluent lines. The resin (Dowex 50W-X8,50-gmdiameter) was packed within the 3.17-cm-wide annulus to a depth of 107 cm. A layer of 0.018-cm-diameterglass beads was packed in the annulus to a depth of 8.5 cm on top of the resin. Eluent and feeds are introduced through a stationary inlet distributor that extends through the top flange and is sealed by two O-rings. All entering streams were delivered by metering pumps (Milroyal Model DC-1-175R)which provided a constant flow with a precision of f l % . The eluent was allowed to fill the head space above the glass bead

15

Biotechnol. Prog., 1990, Vol. 6, No. 1

layer, while stationary steel nozzles whose tips were located within the glass bead layer were used to distribute the feed. As the CAC unit rotates, the beads flow around the feed nozzles without significantly disturbing the resin bed below and preventing convective mixing of feed and eluent. In this manner, as long as the feed and eluent streams have similar viscosities, the fluid velocity remains constant throughout the annulus and the various streams are distributed along annular sectors whose widths are proportional to the corresponding inlet stream flow rate. A digitally controlled drive system was used to rotate the CAC unit. Porous, high-density polyethylene plugs located a t the top of each exit hole provide support for the sorbent. The effluent flows continuously out of the annular bed through 120 exit tubes. The feedstock used in this work was an industrial cornderived glucose-fructose mixture. The feedstock contained approximately 355 g/L of fructose, 325 g/L of glucose, and 26 g/L of oligosaccharides and other unknown dissolved solids of high molecular weight. Feed mixtures were prepared by diluting the feedstock with deionized distilled water without further purification. The eluent was also deionized distilled water. A preliminary monitoring of the separation performance of the CAC unit can be readily obtained by connecting an in-line, continuous detection instrument with a single eluate exit tube. As the CAC rotates, the eluate exit tubes also rotate, and the detector receives eluate from all circumferential points as the apparatus completes a 360" turn. Thus, if the concentration-detection instrument is attached to a timer, an apparently conventional (i.e., concentration vs time) chromatogram is recorded. Angular concentration profiles are immediately obtained, considering that the angular displacement from the feed point, 8, is related to the recorded time, t , and to the rotation rate, w , by 8 = ut. In practice, the connection between one of the eluate exit tubes and a refractive index detector (Waters Model R401) was provided by a digitally controlled Constametric pump to ensure constant flow, equal to 1/120 of the total flow rate into the top of the bed. If there is complete separation of the species, the calibrated refractive index detector allows an immediate analysis; otherwise, it gives the total sugar concentration as a function of 8. Where peaks overlapped, concentration profiles for the individual species were determined by collecting samples that were subsequently analyzed by HPLC using a Biorad Aminex HPX87C column at 85 "C. Samples were collected at 4-deg intervals along the annulus. All experiments reported in this paper were carried out at room temperature, 25 f 3 "C.

Theoretical Discussion As has been previously shown,18for a uniformly packed annular sorbent bed with void fraction e , the following steady-state continuity equation may be written for a solute with concentration c

a% €Do a% ac aq ac ED,-+ -- = ut + w ( I - C) +u az2 R: a 82

a8

a8

az

(I)

where D, and D, are the axial and angular dispersion coefficients, o is the rate of rotation, Ro is the radius of the annular chromatograph, u is the fluid superficial velocity, and z and 8 are the axial and angular coordinates, respectively. Transport to and from the sorbent parti-

cles for a linear system may be accurately described by17 ~ ( 1e )-aq = k,a(c - c*) a8 where koa is the overall mass-transfer coefficient and c* is the fluid concentration in equilibrium with the solid. If the angular dispersion term in eq 1is neglected and the transformation

8 = wt (3) is made, eq 1 and 2 are term-by-term equivalent to the continuity equations describing the propagation of a solute through a conventional chromatographic column. As a result, as pointed out by Wankat," under conditions of negligible angular dispersion, the unsteady-state, onedimensional chromatographic process is equivalent to the steady-state, two-dimensional CAC process. Thus, equations available to describe the transient performance of standard chromatographic units can be very simply modified to represent the steady-state operation of a CAC unit. For systems exhibiting a linear sorption isotherm, various analytical solutions are available to describe isocratic elution, while numerical techniques are in general required for the nonlinear system.lg Howard et al.15 have demonstrated that at the flow rates typically encountered in liquid chromatography applications, when fluid and intraparticle mass-transfer resistances are present, both angular and axial dispersions are only moderately important in the CAC. Thus, for example, if a linear system is considered, the analytic solution obtained by Sherwood et al.17 for an infinitesimal feed pulse applied to a conventional chromatographic bed can be used to describe the steady behavior of a CAC unit operated with an infinitesimal feed sector. The solute exit concentration profile is given by

where K is the distribution coefficient for the solute and

-

t = -8- - cz w

(5)

u

Equation 4 applies as an approximation when the number of transfer units available in the bed is greater than 5. The quantity Q, defined by Sherwood et al. as the amount of solute quickly introduced with the sample per unit cross-sectional area of the sorbent bed, may be calculated for the CAC from Q=-- CFUQF QT

360" 0

where cF is the solute concentration in the feed mixture, QF is the feed flow rate, and QT is the total flow rate of fluid through the annular bed. At high loadings that would be normally used in preparative applications, the feed occupies a significant sector of circumference and eq 4 fails. In this case, retaining the assumptions of negligible angular and axial dispersion and of linear equilibrium, the analytic solution obtained by Carta" for periodic applications of finite feed volumes to a chromatographic column may be used with the same transformation of eq 3. The CAC concentration profile for a component that enters the unit with

16

Biotechnol. Prog., 1990, Vol. 6, No. 1

Table I. Distribution and Mass-Transfer Coefficients Determined from Fixed-Bed Chromatographic Experimentss at Infinite Dilution (25 "C)

sugar K sucrose 0.0079 glucose 0.25 fructose 0.59 a Eluent was deionized distilled water.

where eF is the arc over which feed is applied to the column, OE is the elution arc, and r is given by koa(& + 8,) 2 n ( l - t)Ko

r=

(8)

This equation can be applied to the CAC when a finite feed sector is used or when multiple interfering feeds are applied. The average concentration of a product cut collected between any two angles, 81 and 82, may be computed by integrating eq 7 . The following result is obtained:20 E(z)

OF

-=CF

+

OF + BE

sin

[1"

OF

2(47 + 8 E ) ~ l l -exp ~ ( 6 2 - 8 1 ) j=1 j 2

sin

+ 8E

[

j d 0 2 - 0,) 8F

j d 4 + $2)

+ 8E

[

j2koaz -

(j2

] [--+ cos

- 2jxzwt

j.R8F

8F

-

1.

+ r2)u + $E

1)

OF + $E + 0,) ( jjrkoaz 2 + r2)u (9) The purity of the product cut can be easily determined by computing E for each solute from eq 9. Equations 4-9 are based on the assumption of ideal linear behavior for the various solutes. For the sugar system under investigation, the assumption of linear equilibrium is an excellent approximation for sugar concentrations up to at least 250 g/L.3 At higher concentrations, the uptake isotherms for glucose and fructose show a degree of upward curvature and of coupling between the two solutes.6924Nevertheless, such deviations from the limiting Henry's law limit remain small, and linear theories may be used with reasonable accuracy to gain insight in the separation process. Aside from these deviations from linear equilibrium, significant differences in performance between a fixed bed and the CAC may be observed in practice. Firstly, the observed axial dispersion could be somewhat different because of different wall and end effects in the two systems. Secondly, angular dispersion may, for some separation systems, be a very significant contribution to peak dispersion, while none is obviously possible in the fixedbed analogue. This is typically the case with the very small sorbent particles used in HPLC applications. Finally, there exists a possibility of flow nonuniformities both in the axial and in the angular directions that may result from packing irregularities and viscous fingering. Previous work on the CAC15-" has demonstrated that, for a variety of practical systems, the CAC behaves in a nearly ideal manner and these factors are not a concern. This has generally been observed a t low feed loadings and low feed concentrations. In this work, we test the limits of the CAC, investigating conditions of high loadings and feed concentrations for the sugar separation system. Such conditions are likely to be encountered in industrial applications. In the analysis, eq 4 and 7 are used to repre-

koa,s-l 0.0021 0.011 0.016

sent ideal conditions for the cases of low loading (infinitesimal feed sector) and high loading (finite feed sector), respectively. Experimentally observed deviations from concentration profiles calculated from these equations are attributed to the nonideal factors discussed above.

Results and Discussion Evaluation of Equilibrium and Mass-Transfer Parameters. Pure component injections to the fixedbed column were used for the determination of equilibrium and mass-transfer parameters, using the procedure illustrated by Howard et al.15 Sucrose was used to represent the oligosaccharides present in the industrial feedstock. Small sample loading and feed concentrations (125 g/L) were used in these experiments. The results are given in Table I. As previously discussed, glucose exists as two isomers that have slightly different K values. For the sake of simplicity, since only a very slight resolution of the two glucose forms is obtained in these experiments, we are assuming that glucose exists as a single component with an average K value. The corresponding koa value is, however, considerably smaller than the actual individual values to take into account the extra apparent peak spreading due to the partial resolution of the anomer forms. For all the sugars investigated, transport within the resin particles is the controlling mass-transfer resistance, and the koa values are essentially independent of the liquid flow rate. A 0.5-cm3 pulse of a mixture consisting of 10 g/L of sucrose, 25 g/L of glucose, and 25 g/L of fructose was then applied to the fixed-bed column and eluted a t a flow rate of 0.68 cm3/min. The exit concentration profile is shown in Figure 2 in comparison with calculations based on eq 4 and the parameters of Table I. Experimental and predicted curves are seen to be in good agreement, indicating the accuracy of the modeling procedure for these conditions. The parameter values given in Table I are used throughout the remainder of this investigation. It should be recognized, however, that, in HFCS, sucrose is accompanied by a variety of oligosaccharides, which can only be described by sucrose in an approximate manner. Furthermore, in the chromatography of actual HFCS, a small peak, presumably of rather high molecular weight, is found to elute very early, essentially in the column dead volume. The presence of this peak is ignored in this work. The same mixture of sugars used in the experiment whose results are shown in Figure 2 was separated in the pilot-scale CAC unit for equivalent conditions. Blue dextran, with molecular weight of 2 000 000, was added to the CAC feed mixture as a tracer and permitted determination of the bed void fraction ( 6 = 0.39). This component is virtually completely excluded from the resin matrix and is seen to elute in the bed dead volume. The feed sector in this case, has a width of 6 deg, which can be considered essentially infinitesimal. Comparison of experimental fixed-bed results of Figure 2 with the equivalent run on the pilot-scale CAC is shown in Figure 3.

17 O F F S E T FROM FEED (deg)

-

--

__

2.5

0

0

GLUCOSE (THEORY) EXPERIMENT SUCROSE (THEORY) FRUCTOSE (THEORY)

400

800

1200

3.5

-

3.0

-

40

80

120

160

200

240

280

I

I

I

I

I

I

I

320

-

FIXED BED (011988D)

1600

2000

2400

2800

3200

TIME (s)

Figure 2. Comparison of experimental fixed bed results with the prediction of the simple mass-transfer resistance theory (eq 4). Feed contains 25 g/L of glucose, 25 g/L of fructose, and 10 g/L of sucrose. Note that the time scale of the fixed-bed chromatogram and the angular distance scale of the CAC concentration profile are superimposed in this figure. The continuous separation performance of the CAC is clearly very similar to the batchwise performance of the fixed bed. In fact, for these conditions, the CAC peaks appear to be somewhat sharper than the correspondingfixed-bed peaks, which would indicate a better performance. In reality, some additional peak dispersion may have resulted in the fixed-bed run from a less than ideal injection of the feed sample. This favorable comparison between the two modes of operation confirms earlier work'* carried out with a bench-scale CAC unit.

High-Loading, Low-Concentration CAC Experiments The industrial HFCS feedstock, diluted 1O:l with deionized distilled water, was used for these runs. The viscosity of the diluted mixture was 1.12 CPand that of the eluent (deionized distilled water) was 0.97 cP both at 25 "C. Experiments were carried out with the pilot-scale CAC unit using two wide feed sectors, 30° and 55". In the latter case, the feed was distributed by means of three feed nozzles, spaced approximately 18.3 deg apart from each other. To ensure an even flow distribution, each feed nozzle was fed by an individual metering pump. A single feed nozzle was sufficient to evenly cover the 30' sector. A comparison of the experimental CAC concentration profiles, with theoretical predictions based on eq 7, is shown in Figures 4 and 5 for these two cases. The 30" feed sector comparison indicates good agreement between theory and experiment, although the experimental peaks appear to be again somewhat higher than expected from calculations based on parameters determined from the fixed-bed experiments. The wider feed sector experiment illustrated in Figure 5 is in good agreement with theoretical predictions and shows that the linear model can be used even for these overloaded conditions. As seen in Figure 5, these conditions lead to plateaus in the concentration profiles for values equal to the feed concentrations. High-Concentration Experiments. The total sugar concentration as well as the concentrations of individual sugar components has a rather complicated effect on the equilibrium distribution coefficient as recently reviewed

40

t

8

.

-1

DATA THEORY

OLIGO. GLUCOSE FRUCTOSE

---

A

': 0 0

48

96

144

192

240

288

336

OFFSET FROM FEED ( d e g )

Figure 4. Comparison of experimental 30° feed sector run with the theoretical predictions of eq 9. A diluted (101)industrial HFCS was used as the feed. The feed viscosity was 1.12 cP, while that of the eluent was 0.97 cP, both measured at 25 O C . The rotation rate was 144 deg/h. by Ganetsos.21 While the equilibrium is clearly nonlinear at high sugar concentrations both for resins and for zeolite sorbents, as pointed out earlier, these effects are usually rather small. An effect that is likely to be much more important in a comparison of performance between the CAC and a conventional fixed-bed chromatograph is that of viscosity differences between feed and eluent streams. The industrial feedstock used in this work contains approximately 724 g/L of sugar. At 25 OC, this mixture has a viscosity of 17.7 cSt, and it is much higher than that of the eluent. A series of three runs with different feed concentrations obtained by diluting the industrial feedstock with deionized distilled water is shown in Figures 6-8. The feed sector was 30" for all three experiments, while the feed concentrations were 181,362, and 724 g/L. The viscosities of these feeds at 25 "C were 2.20, 2.76, and 17.7 cSt, respectively. The concentration profiles shown in these figures clearly show a progressive deterioration of performance as the feed concentration is increased. The

ia

350

-2

I-

.

A

4

0 OLIGO.

A FRUCTOSE

300

J

D

250

51 c < 200 c a

50

150

z

s

100 50

0 80 OFFSET FROM FEED (deg)

HFCS was used as the feed. The rotation rate was 144 deg/h.

, , , , ,

00

t

120

.

-

p.

140

180

180

I

,

,

,

0 OLKIO.

I.

200

220

A FRUCTOSE

240

260

280

300

i

320

OFFSET FROM FEED (deg)

Figure 6. Comparison of experimental 30" feed sector run with the theoretical predictions of eq 7. A diluted (3:l) industrial HFCS was used as the feed. The feed viscosity was 2.2 cP, while that of the eluent was 0.97 cP, both measured a t 25 "C. The rotation rate was 144 deg/h. l

200

-2 Y

160

-

140

-

180

120

-

100

-

c

d z c 8 z

s

80

60

40

'

L

'

l

.{\

'

l

'

l

1

0 OLIGO.

GLUCOSE A FRUCTOSE

.*?

i

-

100

140

180

220

160

200

240

280

320

OFFSET FROM FEED (deg)

Figure 5. Comparison of experimental 55" feed sector run with the theoretical predictions of eq 9. A diluted (1O:l) industrial 100

120

260

300

OFFSET FROM FEED (de01

Figure 7. Comparison of experimental 30" feed sector run with the theoretical predictions of eq 9. A diluted (1:l)industrial HFCS was used as the feed. The feed viscosity was 2.76 cP, while that of the eluent was 0.97 cP, both measured a t 25 "C. The rotation rate was 144 deg/h. solid lines in these graphs are calculated from eq 7 using the parameters of Table I, which were determined from the low-concentration, low-loading, fixed-bed runs. As the feed concentration is increased, the experimental profiles show an increasing departure from the limiting theoretical concentration profiles. It should be pointed out

Figure 8. Comparison of experimental 30" feed sector run with

the theoretical predictions of eq 9. An undiluted industrial HFCS was used as the feed. The feed viscosity was 17.7 cP, while that of the eluent was 0.97 cP, both measured at 25 "C. The rotation rate was 144 deg/h.

that for these conditions, while the overall peak shape was very nearly the same, the finer details of the many peak irregularities were not exactly reproduced in repeated runs. This somewhat erratic behavior is indicative of the occurrence of viscous fingering. As discussed, for example, by Collins22and H ~ m s yviscous , ~ ~ fingering results from the hydrodynamic instability of viscous sample zones. The drop in viscosity at the rear boundary of the sample band causes the angular velocity profile to become unstable. In this zone, "fingers" of eluent break through the viscous layer and are projected downstream, while viscous sample fingers are delayed. The net result is a significant distortion of peaks, which become increasingly broader and skewed. The phenomenon is, of course, not unique for the CAC and finds its analogue in conventional fixed-bed operations as well. Similar peak broadening effects due to viscous fingering have, in fact, been reported for gel permeation chromatography separations carried out at high loadin s and feed concentrations in conventional fixed beds!4-27 In the industrial practice of sugar separations by chromatography, the problems caused by excessive differences in viscosity between feed and eluent are easily overcome by operating at temperatures well above 25 "C. At 60 "C, for instance, the viscosity of the undiluted feedstock is about 3.5 cP, only a factor of 9 higher than that of water at the same temperature. By operating at elevated temperature, the equilibrium distribution coefficient of fructose is significantly reduced, while that of fructose remains essentially constant." Nevertheless, the advantages of reducing the viscosity difference greatly offset the disadvantagesbrought about by a reduced intrinsic separation factor for the two sugars. In this work, the high feed viscosity is used as a test of the limit of operability of CAC. It appears that a viscosity difference of a factor of about 2 can exist without significant deterioration of separation performance. At higher viscosity differences, the deviations from ideal behavior rapidly reduce the separation quality to unacceptable values. A possible cause of performance changes in the highconcentration region is some nonlinearity in the isotherm at these concentrations. Component concentrations somewhat above the measured isotherms (350 vs 250 g/L)3 were used. The peak positions held constant, indicating linearity even at the highest concentrations. While it is possible that nonlinearity and interaction

19

Biotechnol. Prog., 1990, Vol. 6, No. 1 OFFSET FROM FEED (deg) 48

0 250

2 D

1

,

98

,

144

192

1

240

'

1

1

OLIGO. GLUCOSE FRUCTOSE

200 t

336

288

'

'

FIXED I

A

1

384

'

I

432

'

'

--CAC

---

480

l

j

I

1

1111 0

20

40

60

80

100

120

140

160

180

-

350 300

2

200

s

200

t

---

OLIGO.

I5OI

1

A AA A A

100

A

-1

TIME ( m i d

Figure 9. Comparison of experimental CAC 30' feed sector run with an equivalent fixed column run. An undiluted indus-

trial HFCS was used as the feed. The feed viscosity was 17.7 cP, while that of the eluent was 0.97 cP, both measured at 25 OC.

affected peak shapes, the observed deterioration a t the highest concentrations was in all likelihood driven by other effects. In order to determine to what extent the annular geometry of the CAC affects viscous fingering, a fixed-bed run was carried out with the undiluted feed mixture used in the experiment of Figure 8. The fixed-bed run conditions were chosen to be equivalent to those of the CAC run so that the two concentration profiles can be superimposed directly. A comparison of the two runs is shown in Figure 9. While the two concentration profiles differ somewhat, the deree of spreading appears to be of the same order of magnitude in the two apparatus. This leads to the conclusion that the viscous spreading effects are not significantly enhanced in the annular bed. Finally, a comparison of calculated and experimental CAC profiles for a concentrated feed and a wide feed sector of 72' is shown in Figure 10. The calculations are again based on eq 7 with the parameters of Table I. The calculations predict plateaus at the feed concentration values. The experimental peaks are, however, much broader than the calculated ones with a dramatically reduced separation performance. It is expected, however, that operation at higher temperature would lead to experimental results much closer to the limiting infinite dilution separation performance.

i O

The continuous chromatographic fractionation of industrial mixtures of glucose, fructose, and higher molecular weight sugars with the calcium form of Dowex 50W-X8 cation-exchange resin has been demonstrated by using a pilot-scale continuous annular chromatographic unit. Equilibrium and mass-transfer parameters evaluated in a fixedbed column in the limit of infinitesimal loadings and dilute feeds were used as the standard of "ideal" behavior for the CAC. Since earlier studies had determined the effects of feed rate, feed concentration, eluent velocity, and rotation rate on CAC performance, the primary purpose of the current study was to observe the effects of an increased scale of operation and the permissible loading and concentration of the feed solution on the CAC. At low loading and feed concentration, the observed continuous separation performance of the CAC unit was essentially the same as the batchwise performance of a conventional fixed-bed unit. Further, the separation performance of the CAC was not reduced relative to the limiting infinite dilution value by increased loading (a feed sector of up to 5 5 O was used), provided that the viscosity of the feed and that of the eluent did not differ by

96

144

192

240

288

336" 24

72

OFFSET FROM FEED (deg)

Figure 10. Comparison of experimental 72' feed sector run with the theoretical predictions of eq 9. An undiluted industrial HFCS was used as the feed. The feed viscosity was 17.7 cP, while that of the eluent was 0.97 cP, both measured at 25 "C.

more than a factor of about 2. For greater viscosity differences that result from high feed concentrations, viscous fingering effects dominated the scene, causing substantial amounts of peak spreading and distortion. In some systems, such as the sugar separation system investigated here, the viscosity effects can be mitigated or even completely eliminated by operating at elevated temperatures. For systems were this is not possible, care must be taken that the feed viscosity does not greatly exceed the viscosity of the eluent. Then, for these conditions, direct scale-up of fixed-bed experiments carried out in the limit of infinite dilution may be possible for the design of a continuous annular chromatographic system.

Acknowledgment This research was sponsored, in part, by the Office of Industrial Programs, US. Department of Energy, under Contract DE-AC05-840R21400 with Martin Marietta Energy Systems, Inc.

Notation particle-fluid interfacial area liquid-phase solute concentration CF feed solute concentration C* equilibrium concentration Dz axial dispersion coefficient D* angular dispersion coefficient ko global mass-transfer coefficient K equilibrium distribution coefficient 4 sorbent solute concentration Q column loading of solute QF feed flow rate QT total flow rate F O radius of annular bed t variable defined by eq 5 U superficial velocity 2 bed axial position Greek Symbols t bed void fraction 0 angular distance from feed point 0E elution angle OF feed angle W rotation rate a

Conclusions

48

C

Biotechnol. Prog., 1990, Vol. 6, No. 1

20

Literature Cited (1) Hashimoto, K.; Adachi, S.; Noujima, H.; Marayuma, H. J .

Chem. Eng. Jpn. 1983,16,400. (2) Sherman, J. D.; Chao, C. C. New Developments in Zeolite Science Technology. In Proceedings of the 7th Znternational Zeolite Conference; Murakami, Y., Iijima, A., Ward, J. W., Eds.; 1986. (3) Ho, C.; Ching, C. B.; Ruthven, D. M. Znd. Eng. Chem. Res. 1987,26, 1407. (4) Jones, J. K. N.; Wall, R. A. Can. J . Chem. 1960, 38, 2290. (5) Goulding, R. W. J . Chromatogr. 1975, 103, 229. (6) Barker, P. E.; Thawait, S. J . Chromatogr. 1984, 295, 479. (7) Barker, P. E.; Thawait, S. Chem. Eng. Res. Deu. 1986, 64, 302. (8) Byers, C. H.; Sisson, W. G.; DeCarli, J. P., 11; Carta, G. Appl. Biochem. Biotechnol. 1989,20/21, 635. (9) Welstein, H.; Sauer, C. In Ion Exchange Technology;Naden, D., Streat, M., Eds.; Ellis Horwood: Chichester, U.K., 1984. (10) Heikkila, H. Chem. Eng. 1983, Jan. 24, 50. (11)Broughton, D. B. US.Patent 3,291,726.13, 1966. (12) Broughton, D. B. Chem. Eng. Prog. 1977, 73,392. (13) Ching, C. B.; Ruthven, D. M. Can. J . Chem. Eng. 1984,62, 398. (14) Ching, C. B.; Ruthven, D. M.; Hidajat, K. Chem. Eng. Sei. 1985,40, 1411. (15) Howard, A. J.; Carta, G.; Byers, C. H. Ind. Eng. Chem. Res. 1988, 27, 1873.

(16) Carta, G.; Byers, C. H. In New Directions in Sorption Technology; Keller, G. E., 11, Yang, R. T., Eds.; Butterworth: Stoneham, MA, in press. (17) Sherwood, R. K.; Pigford, R. L.; Wilke, C. R. Mass Transfer;McGraw-Hill: New York, 1975. (18) Wankat, P. C. AIChE J. 1977,23,859. (19) Ruthven, D. M. Principles of Adsorption and Adsorption Processes;John Wiley and Sons: New York, 1984. (20) Carta, G. Chem. Eng. Sci. 1988, 43, 2877. (21) Ganetsos, G. J . Chromatogr. 1987, 411, 81. (22) Collins, R. E. Flow of Fluids Through Porous Materials; Reinhold, New York, 1960. (23) Homsy, G. M. In Disorder and Mixing; Guyon, E., Nadal, J. P., Pomeau, Y., Eds.; NATO AS1 Series E 152; Kluwer Academic Publishers: Boston, 1988; pp 237-251. (24) Moore, J. C. Sep. Sci. 1970,5, 723. (25) Altgelt, K. H. Sep. Sci. 1970, 5, 777. (26) Determan, H.; Brewer, J. E. In Chromatography: A Laboratory Handbook of Chromatographic and Electrophoretic Methods; Heftman, E., Ed.; Reinhold: New York, 1975. (27) Yamamoto, S.; Nomura, M.; Sano, Y. J . Chem. Eng. Jpn. 1986, 19, 227. Accepted April 14, 1989 Registry No. Dowex 5OW-X8,11119-67-8; sucrose, 57-50-1; glucose, 50-99-7; fructose, 57-48-7.