Large-Scale Liquid Chromatography

8/10/2019 Large-Scale Liquid Chromatography

1/6

Large-Scale Liquid Chromatography

H. Colin and G. B. Cox , Prochrom R&D,Champigneulles, France

Copyright ^ 2000 Academic Press

Introduction

The 1990s have seen the birth of high performancepreparative liquid chromatography (HPPLC) and itsrecognition by the pharmaceutical and Rne chemicalindustries as a very powerful and versatile puri Rca-tion technique. In contrast to traditional low perfor-amance / low pressure preparative chromatography,HPPLC is based on the use of columns of medium tohigh ef Rciency, operated at high mobile phase vel-ocities, and consequently at moderate to high pres-sure } conditions quite similar to those met in analyti-cal chromatography.

The very early work carried out in liquid chrom-atography at the beginning of the twentieth century(at a time when liquid chromatography was used forpreparative purposes only) made use of particles1}15 m in diameter. For several reasons, preparativeliquid chromatography (PLC) was later carried outwith large particles ( ' 100 m) packed in columns of very low ef Rciencies operated at pressures close toatmospheric. Under such conditions, the kinetic prop-erties of the columns are obviously very poor; thepreparative separations require large quantities of mobile phase, long separation times and are expen-sive. For this reason, for a long time PLC was con-sidered to be inappropriate for industrial puri Rca-tions and was the choice of last resort. There werethree reasons for this poor image: Rrst, the misunder-standing of the effect of the column ef Rciency and thelack of a theoretical framework to describe the behav-iour of columns under highly overloaded conditions;second, the use of low-quality packing materials(large particle size and large size distribution); andthird, the lack of proper equipment in general, andadequate column technology in particular. Now thatthese problems have been addressed, PLC has manyadvantages.

Particular Effects Related toColumn Overloading

For the sake of simplicity, the following discussion islimited to a binary mixture, but the phenomena de-scribed are the same for more complex mixtures. It isalso assumed that the adsorption isotherms are Lang-

muirian. This is the case for most practical situations

in chromatography. The total band broadening oc-curring in the column is the combination of twocontributions: that of the intrinsic column ef Rciency(kinetic term } the column ef Rciency under an in R-nitely small injected quantity) and that of the Rnitemass of product injected (thermodynamic term } re-lated to the amount injected). The equipment itself,besides the column, can also contribute to bandbroadening, but this is not considered in the presentdiscussion.

When a suf Rciently large quantity of product isinjected, the thermodynamic term (which is propor-tional to the injected quantity, to a Rrst approxima-tion) becomes larger than the kinetic term and thetotal band broadening is primarily controlled by thequantity injected. It has been concluded from thisobservation that, under high mass overload, there isno reason to use ef Rcient columns in PLC since thecontribution of the intrinsic column ef Rciency is mas-ked by the thermodynamic contribution. This is trueas long as pure products are injected, which is obvi-ously not the case when PLC is used.

When a mixture of two products is injected, theelution bands occupy the same section of the columnduring a certain fraction of their migration throughthe bed. When this happens, the molecules of eachcomponent compete for the same retention sites.When the injected quantity becomes suf Rcientlylarge, the molecules that have more af Rnity for thestationary phase prevent the other ones from interac-ting with the adsorbent. As a result, the weakly re-tained product is eluted faster than expected. This isknown as a displacement effect. This effect is parti-cularly strong when the weakly retained product ispresent at a smaller concentration than the stronglyretained one. Under such conditions, the peak for theweakly retained component is not only eluted morequickly than if it were injected alone, but it is alsonarrower and has the typical shape shown in Fig-

ure 1 A. Better separation is achieved than would befound if the displacement effect were not effective.When the more strongly retained compound is atlower concentration than the weakly retained one,the peak for the second compound is broader than itwould be if injected alone. It also starts to be elutedearlier, an effect known as the tag-along effect (seeFigure 1B).

It is clearly preferable to optimize the choice of thechromatographic conditions so that the Rrst productis the one at the smaller concentration (typically, theimpurity to be removed). Under these conditions, the

region where the two products interfere is strongly

II / CHROMATOGRAPHY: LIQUID / Large-Scale Liquid Chromatography 685


2/6

Figure 1

Figure 2

controlled by the column ef Rciency (see Figure 2 ) andincreasing the column ef Rciency results in the possi-bility of injecting (much) more product or recoveringmore product for a given injected quantity. The possi-bility of using displacement effects and injectinglarge quantities is thus associated with columns of suf Rcient ef Rciency. Such columns enable higher

production rates, reduce solvent consumption andpuri Rcation times, and thus decrease puri Rcationcosts.

The displacement and tag-along effects, as well asthe role of the various operating parameters, havebeen modelled by several groups. Unfortunately, it isnot yet possible (and it is questionable if it will be ina near future) to make accurate predictions of theproperties of a given chromatographic system undermoderate and severe overloading conditions, andthus optimize the operating conditions to minimizethe puri Rcation cost, for instance. The rigorous treat-ment of Guiochon and co-workers, based on thesimultaneous resolution of the mass balance equa-tions for the various species involved in the chro-matographic process, involves numerical integrationof a set of differential equations. This treatment isbased on the model of competitive Langmuir iso-therms and the approach is not simple since itrequires knowledge of the composite adsorption iso-therms. Major dif Rculties lie in experiments tomeasure these composite isotherms and the lack of suitable models to describe them with suf Rcientaccuracy.

Column Technology

In order to get ef Rcient columns of large diameter (i.e.' 5 cm) for preparative purposes, it is necessary tohave a packing material of adequate quality and alsothe appropriate column technology. Column techno-logy is a very critical issue, since it is known thatabove a certain diameter columns tend to be unstablebecause of bed settling and the resulting formation of voids in the packing. These voids are very detrimentalto the column ef Rciency and are responsible for banddistortion. In addition, large columns are potentiallymore dif Rcult to pack than small ones, particularlywhen small particles ( ( 50 m) are used. Indeed, thehigh pressure slurry-packing technique used to pack

686 II / CHROMATOGRAPHY: LIQUID / Large-Scale Liquid Chromatography


3/6

Figure 3 Methods of dynamic compression. (A) Radial com-pression; (B) axial compression.

Figure 4 Comparison of chromatograms from columns.

(A) 25 cm 4.6 mm and (B) 25 cm 150 mm.

analytical columns cannot be used for large diametercolumns for both practical and technical reasons.

Various solutions have been proposed to solve theproblem of bed stability. It seems that the best ap-proach is to compress the bed of particles, eitherradially (using columns with Sexible walls immersedin a pressurized Suid) or, preferably, axially usinga piston (see Figure 3 ). This bed compression must bedynamic so that voids are eliminated as soon as theyform. Using dynamic axial compression, it is possibleto pack columns of large diameter with small par-ticles offering the same ef Rciency as analytical col-umns (see Figure 4 ). This technology is particularlyversatile because it gives the operator the possibilityof packing the column very easily and quickly with

any type of material. This is clearly important forequipment used in production where downtime hasto be minimized. In addition, the length of the bedcan be easily adjusted to optimize the column ef Rcien-cy. Finally, whatever the compression technique used,it is very important to design the column ends proper-ly. Indeed, in order to generate piston S ow and avoidband distortion due to unequal S ow distribution, it isnecessary to spread the Sow of liquid evenly acrossthe cross-section of the column.

Packing Materials

The best column technology is useless without theproper packing material. There are two main aspectsto consider when selecting the packing material:chemical and physical. The chemical aspect is relatedto the chemistry of the puri Rcation problem and is notdiscussed here since it is very much product depen-dent. A material well suited for a given applicationmay be inadequate for another. The physical aspects,however, can be discussed in general terms. The vari-ous models of HPPLC have shown that there is anoptimum value of the ratio d 2p/ L for any separationproblem ( d p is the particle size and L the columnlength). In typical conditions, for columns operated ata pressure of 30 to 60 bar, an optimum particle sizeis about 10 }20 m with a column length between 20and 50 cm. Accordingly, rather small particles haveto be used for PLC.

It is also very important to keep the particle sizedistribution as narrow as possible, since small par-ticles in the distribution have a very strong effect onthe column permeability and large particles stronglyinS uence the column ef Rciency. In other words, a ma-terial with a large size distribution tends to producelow-ef Rciency columns with low permeability. It isusually considered that the ratio d p10 / d p90 should beless than 1.5 for a good material ( d p10 and d p90 beingthe particle sizes corresponding to 10 % and 90 % of the cumulated volume-based size distribution).

Besides the average particle size and the size distri-

bution, other physical properties of the packing ma-terial are important, particularly the speci Rc surfacearea and the pore size. It is desirable to maximize thesurface area of the material in order to increase itsadsorption capacity. Increasing the surface area,however, usually means decreasing the pore size andthus the access of the molecules to the internal poresof the particles. There exists an optimum pore sizeand speci Rc surface for each compound to be puri Red.

Among the other important physical parameters,one should also mention the mechanical strength of the particles, since they must resist the stress of the

compression technique used.



4/6

Optimization of Separation Conditions

The optimization of PLC is not a simple task sincemany parameters have to be taken into account:kinetics (column ef Rciency and column design interms of d p and L), thermodynamics (choice of thechromatographic system, temperature, degree of col-umn overloading, etc.) and economics. The necessityto take into account economic aspects is importantsince the most favourable conditions in terms of puri-Rcation cost do not necessarily correspond to the bestchromatographic conditions in terms of quality of separation. Indeed, a better chromatographic mobilephase providing more selectivity may turn out to bemore expensive to buy or to recycle (mobile phaserecycling is often essential to keep puri Rcation costslow). This means that such parameters as the heat of vaporization of the mobile phase and its cost have tobe considered. The mobile phase viscosity is also veryimportant since it controls not only the velocity of theeluent at a given operatingpressure (and thus the timerequired for the separation), but also the values of the diffusion coef Rcients of the products in the mobilephase, and accordingly the column ef Rciency. Veryviscous eluents (such as propanol }water mixtures,for example) should be avoided in preparativechromatography.

A general economic analysis of a puri Rcation pro-cess by PLC indicates that there are typically twosituations to be considered. When the quantity to bepuri Red is low, the optimization should be aimed atmaximizing the production rate because time (labour)is the largest contribution to the puri Rcation cost.When large quantities have to be processed, the op-timization should be aimed at reducing the solventcost (solvent consumption and regeneration). Manystrategies can be envisaged to reduce the cost of PLC,depending on the conditions of the puri Rcation interms of expected production, required purity andrecovery ratio, etc. For instance, a puri Rcation can bemade in several steps. This is often advisable whenthe concentration of the product of interest in the

crude feedstock is low (i.e. less than 50 % ). In thisrespect, it is important to clean up the mixture to bepuri Red as much as possible before injecting it (usinga preliminary treatment such as Sash chromatogra-phy, for example).

Various techniques can be used to reduce time andsolvent consumption, and thus puri Rcation cost.Among them are shave-recycling and S ip-S op. Shave-recycling consists of injecting (very) large quantitiesof feedstock } too large to provide satisfactory recov-ery of the product of interest at the expected purity.The fraction of the product peak where the purity is

acceptable is collected and returned to the column

inlet (typically by using the solvent pump } adequatevalving is required). During the second passagethrough the column, the separation between theproduct of interest and the impurities is further im-proved. The pure fraction of the product peak is thencollected and, if necessary, the contaminated fractionof the peak is returned to the column for a third cycle.Typically, three or four cycles are made before a newlarge injection is performed. This shave-recycling ap-proach saves on solvent (no fresh solvent is consumedwhen the column ef S uent is returned to the columninlet), reduces product dilution and decreases puri R-cation costs.

The S ip-S op technique is an optimized back Sushoperation. It is very common for several impurities tobe eluted after the product of interest. If these impu-rities are not removed from the column before thenext injection is made, they can contaminate theproduct of interest in subsequent puri Rcation runs.BackS ushing the column is often done to remove thestrongly retained impurities. Flip- S op consists of per-forming the next injection a certain time after theback S ush process has been initiated. During this timethe weakly retained impurities of the next injectionare eluted. Accordingly, during each elution run theS ow direction in the column is reversed (hence thename S ip-S op). This technique requires very goodcolumn stability because of these changes of Sowdirection.

Another powerful approach employed to reducepuri Rcation costs and solvent consumption is the re-cently rediscovered technique of simulated movingbed (SMB), which is of particular interest for binaryseparation.

Simulated Moving Bed

When working on a large production scale, there aremany arguments in favour of continuous processes.The majority of chromatographic processes are, how-ever, batch processes that are incompatible with con-tinuous production. Some years ago, continuous

chromatographic processes were developed for large-scale separations in the food and petroleum indus-tries. These processes solved some of the problems of implementation of large-scale chromatography. Theyare based upon the technique that is now known assimulated moving bed chromatography. The originalapplication areas for this process were in the puri Rca-tion of p-xylene (a precursor of polyester plastics) andthe separation of glucose from fructose to make high-fructose corn syrups, which are used in the soft drinksindustry as a sucrose replacement. A few other ap-plications also based upon this technology exist.

There are perhaps 150 of these large-scale, low



5/6

Figure 5 TMB unit.

pressure SMB systems in the world, responsible forthe production of many thousands of tonnes of prod-uct each year. More recently, the combination of thenew regulatory atmosphere in the pharmaceuticalsindustry, where pure enantiomers, of racemic drugsare becoming more and more in demand, togetherwith the development of new media for the chrom-atographic resolution of enantiomers, has resulted innew areas of opportunity for SMB applications.These applications, by virtue of the high value of theproducts and the requirements for high purity in theindustry, are for the most part operated using smallparticles in high pressure SMB equipment; this paral-lels the extensive use of HPLC puri Rcation methodsin production for the pharmaceutical industry.

The basic principle of a true moving bed (TMB)unit is shown in Figure 5 . The basis of TMB lies in theidea that in a chromatographic system the stationaryphase may be moved as well as the mobile phase. Ina binary separation, the two components move atdifferent velocities through the column. If one were tomove the packing material in a direction opposite tothat of the mobile phase in the column, one couldchoose a velocity of the packing material relative tothat of the mobile phase such that one of the compo-nents moved in the original direction, carried by themobile phase, while the other component, whichspends a greater proportion of its time in the station-ary phase, would be transported by the packing ma-terial in the opposite direction. By injecting in themiddle of such a column, one product would emergefrom each end of the column (the more retainedproduct is often called the extract and the less retainedproduct the raf Rnate, see Figure 5). Thus one couldinject and collect product in a continuous fashion.

In practice, many attempts have been made torealize such a countercurrent S ow of packing materialand mobile phase. None has been really successful,due mainly to the incompatibility of the needs of chromatography (tightly packed beds) and those of the transport of the solid particles (loose, S uid beds).The solution to the problem is to move the entirecolumn, but in this case a column cut into shortsections. In this way, rather than moving the bed ina continuous fashion, the individual sections of thecolumn are moved periodically, such that the averagebed speed remains that required for continuous separ-ation. The feed inlet and the product outlets arearranged at the interfaces between the sections of thecolumn, facilitating the entry and removal of mater-ials. This principle operation simulates a true movingbed system and is thus called a simulated moving bed,SMB. It has been shown that when the number of columns exceeds about eight (this number depends onthe separation problem considered), the performanceof SMB is very close to that of TMB.

The removal of the products requires some specialarrangements of S ow rates in the system. The slowermoving product (extract) moves during the separ-ation phase in the direction of the packing material.If, in the section between the solvent inlet and theproduct outlet, the solvent Sow is increased, a velo-city will eventually be attained at which the productslows and becomes stationary. At a higher velocity,the product will reverse direction in the system andmove with the solvent. This extra Sow is removedfrom the product port; downstream of this port thevelocities in the system remain the same as before.This means that in the Rrst part of the column theproduct is moving toward the exit port in the samedirection as the solvent. In the second section, theproduct is still moving with the packing material,also in the direction of the exit port. The onlypossible route for this product to take is out of theport, along with the extra solvent used to increase theS ow rate beyond that critical for the change in direc-tion of the product. In certain cases, this extra S ow is

not very large, and can be close in value to that of thefeed.At the other end of the system, the raf Rnate is

removed by bleeding off some of the Sow rate at theexit port. This reduces the solvent velocity in the Rnalsection of the column set such that the product moveswith the packing material rather than the solvent.This mans that, in this Rnal section, the product ismoved toward the exit port. In the penultimate sec-tion, the Sow continues as before and the productagain is moved in the direction of the outlet port.Thus, this product, too, is eluted from the system at

the appropriate product port.



6/6

This arrangement for the removal of the productsallows the solvent to be recycled. The solvent thatelutes from the end of the column set does not (if theS ow rates, etc., are controlled correctly) contain anyproduct. It can then be redirected back to the solventpump with the addition of a make-up Sow to com-pensate for the excess Sow removed in the raf Rnateand extract streams in comparison with that added bythe feed. By the same token, the columns that pass thesolvent inlet contain no product and so these can beadded back at the other end of the column set forreuse. In this way a limited number of columns } usu-ally between 8 and 16 } are used for the separationsand the solvent use is also reduced in comparisonwith the batch operation. It should be noted that, inthe majority of SMB system, the columns themselvesremain stationary. Their movement is simulated bythe use of switching valves, which are used to movethe entry and exit ports around the column set.

In comparison with batch chromatography, SMBseparations can be performed using smaller columnsand less solvent than the equivalent batch separation.This is the result of the reduction in solvent required,the increased ef Rciency of use of the column and alsothe reduced need for column ef Rciency in terms of thenumber of theoretical plates in the system. Thus,although one requires a greater number of columns,the column set does not have to be as long as fora batch separation using the same column packing.This can lead to some bizarre solutions, in that batchHPLC separations are usually optimized for columnlengths of 25 to 50 cm, using particle sizes of around10 to 25 m. The equivalent SMB separation, usingthe same particle size, would require column lengthsof only 1 or 2 cm. This is not dif Rcult to realize fora small-scale system, but becomes dif Rcult if largediameter columns } of 50 to 100 cm internal dia-meter, for example } are envisaged. In this case, largerparticle sizes are required in order that the columnscan have reasonable product distribution to eliminateband deformation. The optimum size of the particles,columns, etc., for the separation then becomes a ques-

tion of economics rather than of science.

Conclusions

HPPLC is certainly one of the most powerful puri Rca-tion techniques available today, and as such must be

considered by process and production engineers withthe same attention as distillation, crystallization andother more traditional techniques. Provided it is proper-ly used, HPPLC can be cheaper than traditional tech-niques, provide higher yields, higher degrees of purityand save on puri Rcation times, a particularly criticalaspect in the highly competitive world of the modernpharmaceutical industry. Properly using HPPLC meansunderstanding the behaviour of columns under highlyoverloaded conditions and using such bene Rcial effectsas displacement effects. It also means using the rightpacking material, the right column ef Rciency and theappropriate column technology, as well as consideringalternative strategies such as shave-recycling, back Sush,SMB, etc. SMB is of particular interest for binaryseparations (but not only) and in the near future isexpected to become the technique of choice for suchseparations at large and very large scales.

See Colour Plate 23.See also: II/Chromatography: Liquid: Column Techno-logy. III/Chiral Separations: Liquid Chromatography.Flash Chromatography. Medium-pressure LiquidChromatography.

Further Reading

Bidlingmeyer BA (ed.) (1987) Preparative Liquid Chromatography Elsevier: Amsterdam.

Golshan-Shirazi S and Guiochon G (1988) Analytical Chemistry 60: 2364.

Gviochon G, Shirazi SG and Katti AM (1994) Funda-mentals of Preparative and Nonlinear ChromatographyBoston: Academic Press.

Knox JH and Pyper HM (1986) Journal of Chromatogra- phy 363: 1.

Miller L, Orihuela C, Fronek R, Honda D and DapremontO (1999) Chromatographic resolution of the enantio-mers of a pharmaceutical intermediate from the milli-gram to the kilogram scale. Journal of ChromatographyA 849: 309 }317.

Nicoud RM (1992) LC-GC INTL , 5: 43.Pynnonen B (1998) Simulated moving bed processing: es-

cape from the high-cost box. Journal of Chromatogra- phy A 827: 143 }160.

Snyder LR, Dolan JW, Antle DE and Cox GB (1987)Chromatographia 24: 82.

Zhang GM and Guiochon G (1998) Fundamentals of simulated moving bed chromatography under linearconditions. Advances in Chromatography 39: 351 }400.

Mass Spectrometry Detection in LiquidChromatography

See II / CHROMATOGRAPHY: LIQUID / Detectors: Mass Spectrometry


Documents

Large-Scale Liquid Chromatography