Enhanced characterization of the chemical structure of high-density polyethylene by size exclusion/thermal fractionation on poly(styrene-co-divinylbenzene) columns

Abstract Following the recent trends in polymer analysis, a novel fractionation method was developed using xylene as an alternative to trichlorobenzene (TCB) and offering a more exhaustive characterization of the chemical structure of high-density polyethylene (HDPE). The method is based on a linearized evaporative light scattering detector (LinELSD), which has a much better signal-to-noise for the solutions of HDPE in xylene, as compared with the traditional differential refractive index (DRI) and infrared (IR) detectors. The low viscosity of xylene opens the possibility to use the poly(styrene-co-divinylbenzene) (PS-DVB) columns not only for gel permeation chromatography (GPC) but also for temperature-rising elution fractionation (TREF). An immediate application is the simultaneous measurement of the wax fraction’s content and molecular weight (MW) in HDPE. Furthermore, there are strong indications that the thermogram profile corresponding to the isothermal extraction at 70 °C provides a fingerprint of the HDPE synthesis catalyst.


Introduction
From a chemical structure perspective, polyethylene (PE) is the simplest of all the polymer structures: just repeating units of methylene with an occasional comonomer inserted into the backbone of the macromolecule. Yet, this "simple" polymer provides a broad diversity of products, mostly obtained by changes to the molecular structure [1] . The following molecular characteristics have a substantial influence on the properties of PE: molecular weight (MW), MW distribution (MWD), long-chain branching (LCB), short-chain branching (SCB), and SCB distribution (SCBD). The technologies applied to manufacture PE, such as slurry, solution, and gas phase reactors, influence these five molecular characteristics to produce distinctive microstructure fingerprints [2] . In the absence of co-monomers, the high-density PE (HDPE) contains mainly linear chains, obtained by the synthesis in mild conditions by using chromium trioxide-based catalysts on silica support developed around 1953 by Phillips Petroleum Company. Alternative aluminum alkyl-activated titanium-based catalysts for HDPE synthesis were introduced almost at the same time by Professor Karl Ziegler and his team at Max Planck Institute in Mulheim. During the period from 1955 to 1975, chromium-based catalysts became the preferred production route to HDPE, because the manufactured resin grades were suited for stiff, crush-resistant containers due to a slightly higher density and a broader MWD, as compared with the competitive products.
The Ziegler products synthesized with a titanium tetrachloride-triethylaluminium (TEAL) complex exhibited slightly lower densities and narrower MWD. However, wide MWD and desired comonomer incorporation achieved in cascade slurry polymerization with Ziegler catalysts have been ensuring the high quality and performance of resins for blow molding and pipe applications.
During the last 50 years, the industrial production of PE by slurry process, in which the polymer is produced in an organic solvent in the presence of monomers and catalyst, became widespread throughout the world. Once formed, the insoluble PE is easily separated from the reaction medium. Therefore, one of the main problems of the slurry process is the solubility of the polymer in the solvent, especially when short molecules are formed, or lower polymer densities are targeted. Due to their "multi-site" characteristics and the conditions in which the high melt flow index (MFI) grades are prepared, i.e. excess of hydrogen, the current catalysts also generate, in addition to standard PE, very short PE chains that are soluble in organic media and particularly in the reaction solvent. We speak of waxes to designate this soluble PE fraction.
HDPE grades are developed for specific applications, and for a pertinent selection, among other properties, the HDPE datasheets stipulate the density and the MFI. The wax concentration tends to increase especially for low viscosity (high MFI) grades of PE. These waxes pose an operability problem: they accumulate in the solvent circuits, can be deposited on heat exchangers, and even lead to blockages and therefore costly production stoppages [3] . In the same conditions of polymerization, different catalysts can produce different amounts of waxes depending on the catalyst's chemical and morphological structure. The active centers can be more or less sensitive to the hydrogen and transfer reactions depending on their arrangement through the catalyst particle. Several models have been developed to describe the heterogeneous chemical composition distribution during Ziegler-catalyzed experiments [4] . For example, the "multi-site, isolated-site and selective poisoning" model for Ziegler PE catalysts consists of three different types of active titanium, located in clusters or isolated sites. The clusters are believed to produce low MW and hydrocarbon-soluble material, while the isolated titanium sites are producing high MW HDPE material. There is a clear need already at the catalyst developmental phase to design the catalyst in such a way that it would produce a minimum quantity of waxes at the industrial scale. This requires a better characterization of the obtained resins than the global information provided by density and MFI. For a more detailed description of the polymer chemical structure, the PE sample could be analyzed by differential scanning calorimetry (DSC) and by gel permeation chromatography (GPC), a liquid chromatography method introduced by Moore in 1964 [5] . DSC measures the polymer crystallinity, directly related to sample density, while the GPC elution profile can be converted into MWD if the appropriate polymer standards are available. Although these methods offer a better insight into the HDPE structure, the co-crystallization in DSC and the separation in function of the logarithm of MW in GPC impose severe difficulties in measuring the low concentrations of waxes in HDPE. Concerning DSC, instead of analyzing the sample in bulk, some improvements in the separation of the compounds can be attained by performing the crystallization and the melting of the PE in the presence of a solvent, a method called temperature rising elution fractionation (TREF) [6] .
Presently, the GPC and TREF methods for PE are routinely done in trichlorobenzene (TCB), a solvent banned as a hazardous substance by Annex XVII (Entry 49) of Registration, Evaluation, Authorization of Chemicals (REACH). On the other hand, xylene, which is known for a long time to be a possible substitute solvent to dissolve the polyolefins [7] , is registered under the REACH, being manufactured in and/or imported to the European Economic Area between 1000 and 10,000 tons per annum. However, the use of xylene for analytical methods was limited due to the low signal-to-noise of the infrared (IR) and differential refractive index (DRI) detectors for polyolefin solutions in xylene [8] .
Patented technology for evaporative light scattering detector (ELSD) [9] , recently upgraded to include a firmware performing the linearization of the output signal with an inverse power law (LinELSD) [10] , opened the possibility of using xylene for the analytical fractionation of HDPE. When compared with TCB, apart from lower toxicity according to REACH, the xylene has also a lower viscosity and a lower crystallization temperature, allowing the analyses with poly(styreneco-divinylbenzene) (PS-DVB) columns to be performed at room or even sub-ambient temperatures without damaging the crosslinked gel inside these columns.
In this work, a novel hybrid GPC/TREF method using xylene as solvent is described, and its advantages are demonstrated by investigating the residual waxes and the microstructures of three series of HDPE samples, synthesized with catalysts having different formulations.

Experimental part
Solvents 1,2,4-trichlorobenzene (Spectropure dry from Biosolve Chimie) is the traditional solvent for analytical GPC and TREF because it is IR transparent in the domain of CH 2 and has for polyolefin solutions a high refractive index increment (dn/dc¼ À0.1 mL/g), which is the key parameter defining the intensity of DRI detector signal.
Xylene (AR from Biosolve BV) is the preferred solvent for preparative techniques involving polyolefin fractionation. However, the common GPC and TREF detectors have a poor signal-tonoise when polyolefins are analyzed in this solvent, due to the presence of CH 3 groups, and due to the low value of dn/dc (À0.005 mL/g) [8] .

Samples
The certificate of analysis for the anionic PS standards (Agilent Technologies, Santa Clara, CA) containing the necessary information for GPC column calibration (Mp column) is shown in Figure S1 (Supplemental material).
All HDPE samples were produced and further selected for this study by Chemium: A first set of three HDPE samples, synthesized with a commercial catalyst in different conditions to obtain different MFI values, were considered as references and denoted R_23, R_56, and R_205. The number following the "R" gives the MFI values in g/10 min measured by ASTM D1238, at 190 C and 2.16 kg. The sample R_56 was used to investigate the concentrations and the MWs of the extracted soluble fractions as a function of the temperature of the isotherm step of the hybrid GPC/TREF method. Two additional sets of 3 HDPE samples each, obtained with proprietary catalysts A and B, and having different MFI values, were denoted A_21, A_49, A_191, and B_16, B_74, B_173. All synthesized HDPE samples were used to evaluate the Mark-Houwink parameters of HDPE in xylene, and to assess the hybrid GPC/TREF method.

Instruments
Traditional GPC analyses using TCB as solvent were performed on an Agilent PL-GPC 220 System equipped with a DRI detector. The HDPE samples were dissolved to obtain concentrations of 2 mg/mL in TCB at 160 C. The injection volume was 200 mL and the chromatographic separation was performed using three Agilent PLgel 10 mm Mixed-B (300 mm length and 7.5 mm ID) columns at a flow rate of 1 mL/min and a temperature of 160 C. The linear column calibration obtained with PS standards is given in Figure S2 (Supplemental material).
The hybrid GPC/TREF system was built by merging a high-temperature gel permeation chromatograph (HT-GPC), a gas chromatograph (GC) oven, and the LinELSD as shown in Figure 1. Details on the instrument design are provided in [11] . Apart from the main GPC functionality to measure the MWD of polymers, the injected samples can also be eluted in function of the chemical structure by employing different temperature programs on the column placed in the GC oven. The HDPE samples were dissolved to obtain concentrations of 1 mg/mL in xylene at 130 C. The injection volume was 200 mL and the chromatographic separation was performed using four Mixed-B columns at a flow rate of 1 mL/min. The main difference of the hybrid instrument as compared with the classical GPC is the possibility to apply a temperature program on the third Mixed B column in the column set, which was kept in the external GC oven. For example, isotherms at 40 C, 70 , and 130 C will give shifted GPC calibration curves as shown in Figure S3 (Supplemental material).
The ELSD linearization was done by injecting a mixture of four PS standards with narrow distributions as shown in Figure S4 (Supplemental material). The measured power coefficient was 1.74 in good agreement with the value of 1.75 previously measured for different polymers dissolved in good solvents [12] . By setting the value of the power coefficient in the ELSD firmware, the detector signal output becomes linear in function of the injected concentration.

Results part
Determination of MWD of synthesized HDPE samples by classical GPC in TCB Figure 2 presents the good correlation between the MFI values and the weight average molecular weights (M w ) of the nine HDPE samples measured with the classical high-temperature GPC instrument with a DRI detector and using TCB as solvent.

Evaluation of K and alpha for HDPE in xylene
By keeping the GC oven at the same temperature of 130 C as the GPC oven, the hybrid instrument becomes a high-temperature GPC allowing the analysis of HDPE samples in xylene. The Mark-Houwink parameters of HDPE in xylene at 130 C were evaluated by exporting the LinELSD chromatograms of HDPE samples to Excel. The K and alpha values for HDPE in xylene were chosen to minimize the differences between the average molecular weights (M n , M w , and M z ) measured in TCB@160 C and those measured in xylene@130 C. For HDPE in TCB@160 C, K was 0.039 mL/g and alpha was 0.725, as indicated by the ASTM D 6474. The alpha value of HDPE in xylene was found 0.745 in good agreement with literature [13] . The K value of HDPE in xylene was found 0.040 mL/g. The average MWs measured with the two methods are compared in Table 1.

Determination of the content of xylene solubles for different extraction temperatures
The sample R_56 was selected to demonstrate the capabilities of the hybrid instrument. Figure 3 shows the chromatograms obtained by applying different temperature programs on the column inside the GC oven. The baseline of each chromatogram was shifted to coincide in mV with the temperature in C of the initial isotherm, when a part of the sample is crystallized on the PS-DVB support inside the GPC column kept in the GC oven. The soluble part is passing through this column and is eluted in GPC mode, thus allowing the determination of its concentration and its MWD. For the tests at 10 C, the temperature was controlled using additional cooling with liquid nitrogen. In Figure 4, the temperature programs corresponding to different tests presented in Figure 3 are shown. Usually, the isotherm part in TREF programs has the same length of time, but for this work it was preferred to modify the initial isotherm step and to overlay the heating step for an easier comparison of the TREF fingerprints of the high crystalline fractions.
As expected, the concentration of the firstly eluted fraction (peak at 30 min) rises with the extraction temperature applied during the isothermal step. An exponential variation of this fraction concentration, calculated based on the peak areas, was observed for temperatures lower than 70 C, as shown in Figure 5.   Usually, the wax fraction in HDPE is measured by Soxhlet extraction in hexane, which has a boiling point of 69 C [14] . As with most preparative methods, it takes a lot of time and manpower and does not provide any information concerning the composition or the MW of the extracted wax. By applying the Soxhlet method to the sample HDPE-R_56, a wax content of 4.1% was  measured. Because, from the principle of TREF, fractions with the lowest crystallinity should be extracted during the isotherm step, the quantification of this very fraction could relate to waxes quantification as defined by the industry. By comparing the wax content measured by Soxhlet extraction with the relationship in Figure 5, a temperature of 40 C was selected for the initial isotherm to determine the wax concentrations in the investigated HDPE samples.
The chromatogram at 40 C, corresponding to a concentration of 4.4%, has a signal-to-noise for the wax peak (eluted at 32 min) of 40. Usually, the limit of quantification of chromatographic methods is given by a signal-to-noise of 10, so the hybrid GPC/TREF analytical method has a quantification limit of 1% wax in HDPE.
The reproducibility of the method was also studied by injecting 6 times the R_56 sample during two months. The obtained chromatograms are shown in Figure 6. The average value of the wax content is 4.55% ± 0.27%, indicating a relative error of 6%, usually encountered in analytical methods involving measuring the contaminant concentrations in polymers.
The method was applied to the nine HDPE samples produced with the 3 catalysts and the measured wax concentrations in function of the MFI values are shown in Figure 7. Differences are observed, the proprietary catalysts A and B generating a lower content of wax as compared with the commercial reference R catalyst.

Determination of the MW of xylene soluble fractions for different extraction temperatures
The analytical method using a hybrid GPC/TREF instrument has the following advantages over the traditional Soxhlet extraction method: automated; reproducible; requiring low manpower; and requiring a low quantity of about 1 mg of sample.
Moreover, the new analytical method provides the correlation between the initial isotherm temperature and the MWD of the extracted fraction during this step as shown in Figure 8. In Figure 8, the black curve corresponds to the MWD of the soluble fraction extracted at 40 C. It contains polyolefins with MWs of 100 g/mol up to 2000 g/mol, which is often the definition for wax in the industry [14] .
The scaled number average MW (M n /14) of the extracted fraction versus the extraction temperature, shown in Figure 9, follows a similar tendency as the plot of melting temperature against the number of carbon atoms for n-alkanes obtained by Mandelkern [15] by collecting the data from 10 independent references. Because in GPC/TREF method the extraction of the soluble fraction is governed not by melting temperature, but by the crystallization temperature, the curve in   It is important to mention that because waxes represent the part of MWD having MW lower than 2000 g/mol, their quantification is also possible by classical GPC. However, a quantification by GPC/TREF method offers more robustness: the target peak is better isolated, easier to integrate, and free from excessive baseline fluctuation and possible calibration shifts at low MW.  TREF fingerprint of the high crystalline fraction: a tool for PE microstructure characterization Up to this point, the study was focused on the information provided by the initial isotherm step, in which the separation of the soluble part is by size exclusion (GPC). However, additional crucial information is provided by the peak corresponding to the heating step (TREF), in which the elution is controlled by the melting temperature of the high crystalline fractions in the injected sample. For example, the thermograms during the heating step of R_56 and A_49 samples, corresponding to an isothermal step at 40 C, are shown in Figure 10. These samples were selected because they have similar MFI but were produced with different catalysts. For each sample, only a single peak is observed during this heating step, with a difference in elution time of Figure 10. Overlay of the chromatograms of the R_56 and A_49 samples (xylene, the initial isotherm of 40 C for 60 min was followed by heating up to 130 C at 1 C/min). Figure 11. Overlay of the chromatograms of the R_56 and A_49 samples (xylene, the initial isotherm of 70 C for 90 min was followed by heating up to 130 C at 1 C/min).
1 min, corresponding to 1 C for a heating rate of 1 C/min. This information on the melting behavior of HDPE samples is similar to the one provided by DSC analysis. However, by applying an initial isotherm at 70 C, it becomes possible to extract more information on the crystalline part, as presented in Figure 11. Two additional peaks at 118 and 122 min are visible in the thermograms of both samples, suggesting a possible fingerprint for the HDPE synthesis method.

Conclusions
In this work, a novel hybrid GPC/TREF method is presented, to determine the concentrations and the MW distributions of residual waxes, as well as the microstructures of the HDPE samples synthesized by the slurry process on a broad domain of MFI values.
The GPC analysis in xylene at 130 C of the HDPE samples, which is made possible by using a linearized evaporative LSD (LinELSD), gave similar results as the GPC analysis in TCB at 160 C, with the advantage that xylene is REACH registered, while TCB is flagged in Annex XVII to REACH as a restricted substance.
As compared with the classical GPC, the hybrid GPC/TREF method allowed us to isolate and to better quantify the low concentration of xylene soluble (wax) fraction by applying a fast cooling followed by an isothermal step at 40 C.
For applications requiring Food and Drug Administration (FDA) registration, by selecting an isothermal step at 10 C, it is possible to target the extraction of the fraction having MWs lower than 1000 g/mol, allowing a more robust quantification of this fraction.
Moreover, an important benefit of the hybrid GPC/TREF method seems to be the possibility to better investigate the microstructure of the HDPE samples, by applying an initial isotherm at 70 C. The thermograms obtained during the heating step for HDPE samples having similar MFI and densities suggest the possibility to identify the HDPE synthesis, and likely catalyst type, by a specific fingerprint, although more work is necessary to confirm this hypothesis.

Disclosure statement
No potential conflict of interest was reported by the author(s).