Introduction Chemical composition and molecular weights are the two most important properties of polymers.

While the chemical composition can usually be determined relatively easily by a variety of techniques, the determination of polymer molecular weights may not be as straightforward. This is further complicated by the fact that, with polymers, we are usually confronted by a distribution of molecular weights. A proper understanding of a polymer comes only from considering the molecular weight distributions (MWD) as a whole. A number of techniques have been developed to determine the molecular weight distribution of polymeric materials. These methods include osmometry, light scattering, mass spectrometry, gel permeation chromatography, high performance liquid chromatography (HPLC) and many others. This essay is aimed at reviewing the use of gel permeation chromatography in determining the molecular weight distributions and average molecular weights of polymers [1]. Gel permeation chromatography (GPC) is a type of size exclusion chromatography (SEC), that separate analytes on the basis of size. This is a chromatographic technique in which the separation obtained is a function of molecular size. GPC has revolutionised polymer characterization. Using this technique it is now possible to obtain the molecular weight distributions of a very small sample of polymer in two or three hours. More traditional methods, which involve fractionation followed by characterization of the resulting fractions, need much more materials and may take weeks rather than hours. This being the case, GPC has opened up entirely new fields of research which were previously practically not possible. With the introduction of commercially-manufactured gel chromatograph, this useful technique has rapidly become established and is now available to many small laboratories with no previous experience in chromatographic techniques [2]. This technique is the most recently used technique for fractionation of polymers. It has two major advantages, good resolution in short time periods, and it is susceptible to automation. Therefore, it is being used increasingly widely.

Origin of GPC The term gel permeation chromatography was first applied by J.C. Moore [3] in 1964. Although the basic mechanism had been used for biopolymers [4]. It was the development of macrorecticular (rigid) polymeric packings, by Moore, and their application to characterising synthetic polymers, which led to the rapid advancement of the technique. By the end of the 1960s, suitable instrumentation had been developed and GPC was established as an important technique in characterising synthetic polymers. At that time, the instrumentation used in GPC was well advance of other forms of liquid chromatography [1]. In the 1970s, gas chromatography workers applied their methods and experiences to working with a liquid mobile phase and high performance liquid chromatography (HPLC) emerged. These general advances in liquid

chromatography have now been applied to GPC and this has led to a new generation of benchtop instruments for the fast molecular weight characterization of polymers.

The Development of GPC The separation process which defines GPC was first observed in the elution of low molecular weight non-ionic substances with water from a column packed with ion-exchange resin particles [5, 6]. Swollen starch gels and agar gel [7], which are non-ionic natural products, were also used successfully for the separation of polysaccharides and proteins by this process. The technique did not become widespread, however, until a series of chemically-produced hydrophilic gels became available [8]. Application to hydrophobic polymers was initially less encouraging. Gels were produced by the copolymerisation of styrene with divinyl benzene [9, 10]. However, these produced only small separation in the high molecular weight region, despite being very highly swollen. Much more successful results were achieved by copolymerising in the presence of a diluent which was a solvent for the monomers, but not for the polymer [5]. The resulting gels have a rugged internal structure and high permeability which can be controlled by varying the amount and nature of the diluent present at the time of crosslinking. These gels are the ones most widely used for GPC at the present time. Much recent work has been directed at producing a porous packing material more stable at the high temperatures necessary for the analysis of many polymers such as polyethylene. Promising results have so far been obtained using porous silica and glass beads [7] and such column packings are now commercially available. The disadvantage of these column packings is that they have polar groups on their surfaces and there is a tendency for the solutes to be adsorbed on the column packings causing a decrease in separation efficiency. Such effects can, however, be eliminated by a surface reaction in which the hydroxyl groups are replaced by methyl or phenyl groups before column packing. Other current research is directed towards the production of gels with greater separation efficiency. Polyvinyl acetate gels have been produced by copolymerisation of vinyl acetate and divinyl esters of dicarboxylic acids [7]. These gave very high resolution and separation of the low molecular weight oligomers of a disperse polymer sample. Such gels are now commercially available.

The Separation Process It is generally assumed that the separation process relies on the preferential penetration of smaller molecules into the pores of the column packing material. This may be due to either the inability of large molecules to enter certain regions of the gel due to their size (steric exclusion) or the larger diffusion coefficients of smaller molecules. Steric exclusion theories [11] which assume simple geometrical models for the gel have been presented, but implicit in these theories is the assumption that the time of molecular diffusion into the gel is very small compared to the
2

time the solution zone takes to pass the gel particle, i.e. there is an equilibrium distribution between the mobile and stationary phases. They relate the total available pore volume per molecule to the mean molecular diameter in solution. The predicted dependence of elution volume on molecular weight is in broad agreement with experiment in each case. Diffusion theories [12] have also been presented and these too, in certain cases, successfully predict the molecular weight dependence of the elution volume. In one theory [12], the equation of Renkin [7] is used to relate the effective pore radius to the Stokes radius of the solute. The other theory [13] uses the solution of Fick's diffusion equation for the simplified model of diffusion along a single coordinate that extends to infinity. Using the known polymer diffusion coefficients, it has been calculated that, with normal GPC conditions, sufficient time is available for complete diffusion into the gel to set up an equilibrium distribution. Also, at very slow flow rates the elution volume of high molecular weight polymers does not approach that of low molecular weight material [1]. Thus, the basic separation mechanism cannot be incomplete diffusion. Incomplete diffusion may modify the basic exclusion process however, particularly for high molecular weight material and high flow rates. A more general theory [7] incorporating both finite diffusion rate and steric exclusion shows how the non-equilibrium effects decrease the separation efficiency. In the theory it is assumed that diffusion occurs with radial symmetry into a spherical stationary phase. The theory cannot be compared with experiment as easily as the other theories since, in general, an analytical solution of the equations is not possible [2]. A theory has been presented [14] which suggests that separation can be achieved simply by flow of a polymer solution through fine capillaries or through a column of non-porous beads. The separation is caused by the quadratic velocity profile of fluid in a capillary; the centres of the mass of the larger molecules cannot approach as close to the capillary wall as the smaller molecules and so the larger molecules have larger average velocities. Such a mechanism could explain the separation which was observed when proteins were eluted from a column packed with impermeable glass beads. Up to the present time the relative importance of steric exclusion and flow separation in normal gel permeation chromatography has not been evaluated [7].

GPC Apparatus Gel permeation chromatography has been carried out for several years using very simple equipment. A single glass column packed with Sephadex gel and fed by a constant head solvent reservoir has proved adequate for many applications, particularly in the biochemistry field [15]. Following the production of gels compatible with organic solvents, a commercial gel permeation chromatograph was developed by Waters Associates Inc. [15]. This liquid chromatography apparatus has enabled laboratories with no previous chromatographic experience to establish polymer characterisation on a routine basis. It is with this type of apparatus that the phenomenal growth of GPC has occurred and with which the majority of this review is concerned. The apparatus consists essentially of a pumping system, a series of sample separation columns with a
3

sample injection system, a series of reference columns, a detector for measuring the weight of polymer in the eluant at a given time, and a system for recording the sample elution volume, as shown in figure 1. The pump pumps the polymer in solution through the system. Different polymers produce solutions of different viscosities. To compare data from one analysis to the next, the pump must deliver the same flow rates independent of viscosity differences. In addition, some detectors are very sensitive to the solvent flow rate precision. Such constant flow must be a critical feature of the instrument. The injector introduces the polymer solution into the mobile phase. It must be capable of small volume injections (for molecular weight determinations) and large volume injections (if fraction collecting is desirable). The injector should not disturb the continuous mobile phase flow. It should also be capable of automatic multiple sample injection when the sample volume is large. The column set efficiently separates sample components from one another. High efficiency columns give maximum separating capability and rapid analyses. Every column must provide reproducible information over extended periods for both analytical and fraction collecting purposes. The detector monitors the separation and responds to components as they elute from the column. Detectors must be nondestructive to eluting components if they are to be collected for further analysis. In addition, the detectors must be sensitive and have a wide linear range in order to respond to both trace amounts and large quantities of material if necessary. For most applications of GPC, the differential refractive index (DRI) detector is that most commonly employed. However, other detectors are used either in place of, or in conjunction with the DRI to overcome specific problems or to obtain additional information. These other detectors are specific detectors which only respond to specific chemical groups or are sensitive to difference in molecular weights. In outline, the method employs a column, often 0.5-3 cm in diameter with lengths varying from perhaps 1 to 10 m. The column is packed with particles, frequently a crosslinked polymer, which contain pores of various diameters. The column is filled with a solvent which permeates both the pores and the column void spaces. A small amount of the polymer, 1-20 rag, dissolved in the same solvent is introduced on to the head of the column. Solvent is continually passed through the column with a typical flow rate being 1 ml/min. During the separation process, small polymer molecules can go freely into the pores while large molecules can be completely excluded from all pores. Intermediate-size molecules have access to part of the available pore volume. Thus the large molecules move through the column more rapidly than the smaller ones. Separation is due to the differing volumes of solvent in the gel that are available to the different sizes of molecules [2].

Figure 1: Schematic diagram of a gel permeation chromatograph [2].

The Chromatogram Figure 2 is a typical illustration of a chromatogram. With the usual differential refractometer detector, the ordinate records the difference between the refractive indices of the polymer solution and the solvent. For the dilute solutions analysed, this is proportional to the weight of polymer in an increment of elution volume. Although a comparison of the chromatograms of two or more samples is all that is necessary in many applications, for complete utilisation of the data, it is often desirable to make the GPC measurements quantitative. This requires correction of the chromatogram for instrumental broadening and calibration of the sample elution volume in terms of molecular weight of the polymer analysed. After calibration, the more usual differential molecular weight distribution Z(M) can be determined from the sample chromatogram y(v). Z(M) is defined such that Z(M)dM is the weight of polymer which has molecular weight between M and M + dM. y(v) is the weight of polymer which is eluted from the chromatograph between elution volumes v and v - dr. It can be easily shown that Z(M) and y(v) are related by the equation [18] Z(M) = [K/(-dM/dv)]y(y) (1)

In other words, after replacement of the elution volume scale by a molecular weight scale using the calibration curve, it is incorrect to simply expand the molecular weight scale to make it linear. On expansion, it is also necessary to divide the chromatogram deflection by a factor proportional to the slope of the M/v calibration curve (i.e. proportional to the product of M and the slope of the log M/v calibration curve). Average molecular weights can be calculated from
5

the chromatogram directly. For example using equation (1) and the definition of number- and weight-average molecular weights, Mn and Mw, it is easily shown that [16]

Mn =

[ ]

(2)

Mw =

(3)

The burden of calculation of average molecular weights and of production of a differential molecular weight distribution from chromatograms can be considerably reduced by use of a computer programme [7].

Figure 2: Chromatogram of a typical commercially available linear pelyethylene-Marlex 600g. The elution volume abscissa has been related to molecular weight using the GPC calibration curve [7].

GPC Calibration Calibration is necessary in order that the elution volume abscissa of chromatograms can be related to polymer molecular weight, as shown in Figure 2. It is performed most simply by using very narrow distribution fractions of the polymer under investigation. In this case, the peak elution volume of the narrow chromatogram is assumed to correspond to the known polymer molecular weight. A calibration curve can be constructed if a number of samples of different molecular weights are available. Using narrow fractions and low molecular weight monodisperse
6

samples, it has been shown experimentally that, over the useful separation range of a GPC column, the functional relationship between the molecular weight M of a molecular species, and its elution volume v is of the approximate form [5]: log M = a bv where a and b are constants. For many polymers, the fractions available are not always sufficiently narrow. It is then necessary to use samples with appreciable dispersion of molecular weight for which either 21~rn or 21~w, or both, are known. With such samples there are two possible methods for the determination of the calibration curve. The first [7] assumes that the calibration curve has the form of equation (4). A "seek and find" computer programme is used whereby trial values of a and b are selected and for each pair of values the molecular weight averages are calculated from the chromatograms of the samples and assumed calibration curve. The calibration curve is defined as the pair of values which gives the best fit between the calculated and experimental molecular weight averages. The second approach [16] assumes that the calibration curve can be taken to have the form of equation (4) for the range of molecular weight in each sample. It does, however, allow for the curve to have different values of a and b in different regions. In this procedure, a single molecular weight average together with the chromatogram of the sample is used to calculate the envelope curve, formed on the log M/V diagram by all possible calibration lines which are consistent with the chromatogram and average molecular weight. The calibration curve is tangential to this computed curve. Thus, from a series of samples, the best calibration curve (which need not be linear) can be drawn through the series of computed curves. As an alternative to using molecular weight, it was initially suggested [15] that extended chain length be used in calibration. It was then hoped that, since GPC separation appears to be due to differences in molecular size, such a curve would be independent of polymer type. This has been shown to be incorrect [7]. Several attempts have been made to determine a parameter which will give a common calibration curve for different polymers. It has been shown that many different types of polymers, including branched polymers, all fall on a single curve when a plot of log []M against elution volume is made, where [] is the intrinsic viscosity of the narrow distribution polymer fraction of molecular weight M [17]. Now the radius of the sphere which in suspension would have the same effect on the macroscopically observed viscosity as the polymer molecule, i.e. the hydrodynamic radius R, may be defined from [] using the equation [18], [] = R3/M (5) (4)

The universal calibration obtained using []M therefore, suggests that it is the hydrodynamic radius of the polymer in solution which governs GPC separation.
7

The above results are of great practical significance since they enable a calibration curve, applicable to one polymer, to be calculated from the data of a different polymer. Thus narrow distribution polystyrene fractions, which are commercially available, can be used to predict directly the log M/elution volume calibration curve of a different polymer for which characterised fractions may not be available. For example, if the elution volume of a molecular species is determined by the product []M then, by introducing the dependence of [] on M (i.e. the Mark-Houwink equation ([] = K Ma), the log M/v calibration curves can be directly related by the Mark-Houwink constants, which are often available in the literature [18]. This approach has been successfully applied to the important practical case of polystyrene and polyethylene [19]. Some of the factors that influence the calibration curve include concentration of the solution, injection time, temperature and flow rate. The details by which these factors affect the calibration curve are found elsewhere [7].

What Information can be obtained from GPC? Gel permeation chromatography is widely used for determination of polymer molecular weight and molecular weight distribution (MWD). For polymer characterization, molecular weight and molecular distribution are important because they can have significant impact on physical properties of the polymer. Because GPC is useful for determination of molecular weight and MWD, it is an important technique for aiding in the establishment of structure/property relationships for polymers. Once molecular weight/property relationships are established, GPC can be used in quality control fashion to ensure the production of polymers having desirable physical properties [17]. The term molecular weight as it is applied to polymeric materials implies something different from what is generally meant for small molecules. This is due to the fact that a polymer sample does not have a single molecular weight but rather a range of values. In a given sample there may be polymer chains which contain widely different numbers of repeat units. In fact, most polymers contain some residual monomer (Degree of Polymerization DP = 1) and some chains which contain several hundred repeat units (DP > 100). These molecules will differ in relative molecular mass by several orders of magnitude. GPC can also be used to aid in optimization of process conditions for production of materials with desirable molecular weight. GPC may also be used for comparing the MWD of samples that may perform differently in end-use applications. Synthetic polymers differ from small molecules in that they cannot be characterized by a single molecular weight. In a synthetic polymerization, a distribution of chain lengths (i.e., molecular weights) is produced. This distribution can be described by any number of molecular weight averages. For this reason, the molecular weight of a polymer is reported using averages. These averages are intended to describe the distribution of molecular weight values for the polymer chains.
8

Three different molecular weight averages are commonly used to provide information about polymers. These are the number average molecular weight (Mn), weight average molecular weight (Mw) and viscosity average molecular weight. In addition, several other averages are used to lesser extents including the Z average molecular weight (Mz) and the Z+1 average molecular weight (Mz+1). The averages are defined mathematically using the following formulas [19 - 21]: Number Average Molecular Weight : Weight Average Molecular Weight: Z Average Molecular Weight: Viscosity Average Molecular Weight:

(5) (6) (7) (8)

Here Ni and Wi are the number and weight, respectively, of molecules having molecular weight Mi. The subscript i is an index representing all molecular weights present in the ensemble of chains. It is noted that each of the above equations contains three representations of the particular molecular weight average. The viscosity average molecular weight also makes use of an additional term a. This parameter is an empirical constant that is dependent upon the hydrodynamic volume or the effective volume of the solvated polymer molecule in solution, and varies with polymer, solvent, and temperature. [21]. It should be noted that Mv and Mw are the same when a has a value of 1. The third representation in each case (farthest right) defines how one obtains these averages from GPC chromatograms. hi is the height (from baseline) of the GPC curve at the ith elution increment and Mi is the molecular weight of species eluting at this increment. Mi is obtained via calibration with appropriate standards, as described above [22]. From the definitions of the molecular weight averages it can be observed that for a monodisperse polymer (polymers with chains of only one mass) the four averages are equal Mn = Mw = Mv = Mz. If however the polymer has many different sized chains (this is typically true) then the relationship between the averages is Mn Mv Mw Mz. The presence of the additional Mi term in each succeeding average results in greater emphasis being placed on the highest molecular weight molecules. For step growth polymers which have a most probable distribution of molecular sizes the ratio of Mz:Mw:Mn is 3:2:1 [21]. To provide a reasonable estimate of the molecular weight, it is necessary to provide more than one average (unless the polymer is monodisperse). To fully describe the molecular weight distribution it is necessary to prepare a plot showing the weight fraction of the polymer chains which have a given molecular weight [9, 10].

Figure 3: Plot of the distribution of molecular weights in a typical polymer sample [21]. Figure 3 shows an example of a plot of the distribution of molecular weights in a typical polymer and the approximate location of the various averages on the curve. Since most polymers are polydisperse, the four molecular weight averages are often used in conjunction with one another to define more thoroughly the nature of the molecular weight distribution. As can be seen from Figure 3, Mn provides information about the molecular weight which is present at the greatest frequency in the sample. Mw is a weighted average which favours higher molecular weight molecules and appears on the high side of the molecular weight distribution. The Mz value has the term Mi to the third power and so it is even more skewed toward the highest molecular weights. The value for Mv ranges depending upon the constant a such that when a = 1 this average is equal to Mw. When a is less than 1, Mv is found to be between Mn and Mw. A typical value for a is between 0.5 - 0.9. [23] When polymers are provided for sale, the molecular weight reported is generally Mn or Mw. Mv is reported less frequently. experimental experience, has shown that the Z-average molecular weight has never been listed apart from another average. This average is skewed too far out into the distribution to be a viable number to list individually [21]. Another parameter which is frequently used when describing polymers is the polydispersity index or PDI. This parameter gives an indication of how broad a range of molecular weights are in the sample. The PDI is defined as [7]: PDI = (10)

The PDI can best be explained by looking at two polymers of the same molecular weight but with different PDI values. Figure 4 shows two theoretical polymer systems which have the same

10

number average molecular weight (Mn). The two polymers differ from one another in that one polymer has a PDI of 1.1 and the second polymer has a PDI of 2.2.

Figure 4: Comparison of two theoretical polymer distributions. One polymer has PDI 1.1 (black) and a polymer of the same molecular weight with PDI 2.2 (red) [7].

Summary GPC is entering its "second generation" of sophistication with the development of more advanced chromatographic theory and improved instrumentation and column packings. Already it is the most widely employed analytical technique for the laboratory MWD characterization of synthetic polymers and analysis and purification of many biological materials. The future can only bring proliferation of commercial and scientific uses of GPC in the fields of quality control, preparative-scale operation, and analytical separations. Complete understanding of the separation mechanism and definition of differences between packing material behaviour remain elusive. In this respect, GPC is one of the most fertile research fields in chromatography.

11

References 1. S. R. Holding, Gel permeation chromatography, Endeavour, New Series, Volume 8. No. 1, 1994. 2. Gel Permeation Chromatography, J. Polymer Sci, 168 (1970) 3931. 3. Moore J. C., J. Poly. Sci., Part A, 2 835, 1964. 4. Porath J. and Flodin P., Nature, Lond. 183 1657, 1959. 5. R. M. Wheaton and W. C. Baumann, Ann. N.Y. Acad. Sci. 57 (1953) 159. 6. R. T. Clack, Analyst. Chem. 30 (1958) 1676. 7. T. Williams, Gel Permeation Chromatography: A Review, Journal of material Science 5 (1970) 118-820. 8. J. Porath and P.Hlodin, Nature 183 (1959) 1657. 9. Yau W. W. and Kirkland J. J., J. Chromaton. 218. 217. 1981. 10. Tijssen R., Bleumer J. P. A., and Van Kreveld M. E., J. Chromatogr. 260. 297. 1983. 11. E.F. Casassa, J.polymer Sci. B5 (1967) 773. 12. G.K. Ackers, Biochemistry 3 (1964) 723. 13. W. W. Yau and C. P. Malone, J of polymer Sci. B5 (1967) 663. 14. W. F. Busse, Phys. Today 17 (1964) 32. 15. L.E. Maley, J. Polymer Sci. C8 (1965) 253. 16. F.C Frank, I. M. Ward, and T. Williams, J. of polymer Sci. A-2 6 (1968) 1357. 17. C. Tanford Physical Chemistry of Macromolecules (J. Wiley and Sons, New York, 1961. 18. H. Coll and L.R. Prusinowski, J. Polymer Sci. B5 (1967) 1153. 19. Barker P. E., Hatt B. W., and Holding S. R.. .I. Chromatoer. 206. 27. 1981. 20. Dawkins J. V. and Yeadon G., High Performance Gel Permeation Chromatography, In Developments in Polymer Characterisation-1. (J. V. Dawkins, ed.) Applied Science Publishers, London, 1978. 21. G. G. Odian, Principles of Polymerization, Wiley and Sons, New York, New Edition, 2004. 22. Dumelow T., Holding S. R., and Maisey L. J., Polymer Communications, 24, 307, 1983. 23. High Temperature Gel Permeation Chromatography-Polymer Solubilities, Solvents, Temperatures Waters Application Note 1981. 24. Cooper A. R., Analysis of Polymers by Gel Permeation Chromatography In Analvsis of Polvmer Svstems (L. S. Bark and N. S. Allen, eds) Applied Science Publishers, London, 1982.

12