Structure, processing and performance of ultra-high molecular weight polyethylene (IUPAC Technical Report). Part 2: crystallinity and supra molecular structure

2.1 Materials

In support of this project, a leading manufacturer provided three large batches of UHMWPE powder, which had nominal weight-average relative molar masses, M ¯ w , of approximately 0.6 × 10⁶, 5 × 10⁶, and 9 × 10⁶, respectively. Accordingly, they were given the code names PE06, PE5, and PE9. Unless otherwise stated, all results presented in this report were obtained from tests on compression mouldings made by the polymer manufacturer under standardized processing conditions.

2.2 Morphology of powder particles – scanning electron microscopy

The first step in characterizing structure in PE06, PE5, and PE9 was to determine the dimensions and morphology of powder particles obtained from the polymerization reactor. Scanning electron microscopy undertaken by Michler (code ML)^[1] showed very little difference between the three grades at this stage. Typical morphology is illustrated in Fig. 1. At a low magnification, PE9 powder is seen to consist of irregularly shaped grains with diameters between 50 and 250 μm. However, higher magnification revealed that each grain is an aggregate of much smaller, roughly spherical particles, with diameters between 0.4 and 0.8 μm. Some of these small particles had fused together. Others were connected by long, straight fibrils between 50 and 75 nm in diameter. Similar features have been reported in previous publications, most notably in an encyclopedia article by Kurtz [2]. Their formation can be attributed to continuous swelling of the spherical microparticles as the polymer chains increased in length. That swelling eventually led to the formation of macroparticles when the microparticles came into contact, and at some stage chain entanglement probably occurred at the contact points. During subsequent polymerization, both the microparticles and the macroparticles continued to expand to accommodate freshly-formed polymer. Also, in some less densely populated regions, the contacting microparticles were pulled apart as a result of this expansion, thereby forming the extended fibrils seen at high magnification in Fig. 1 [4], [5].

Fig. 1:

Reactor powder particles of PE9 viewed at two magnifications – transmission electron micrographs provided by Michler (laboratory code ML).

2.3 Lamellar morphology – TEM and SEM

Michler (code ML) used transmission electron microscopy (TEM) to characterize lamellar morphology in PE06, PE5, and PE9. Samples cut from compression mouldings were embedded in epoxy resin, repeatedly exposed to RuO₄ vapour, and sectioned using a Leica Ultracut UCT ultramicrotome fitted with a Diatome diamond knife, to produce sections about 80 nm thick. A typical section is shown in Fig. 2. Each section revealed linear features, which were previously identified as lamellae seen edge-on [6]. The lamellae have a broad distribution of thicknesses, ranging approximately from 7 to 28 nm, with a peak at (16 ± 3) nm.

Fig. 2:

Transmission electron micrograph of a section through embedded PE9 particles at a high magnification. Michler (ML).

A group led by Slouf used scanning electron microscopy (SEM) to study lamellae in PE06, PE5, and PE9. Specimen blocks were cut from the centres of compression mouldings. Smooth surfaces were then prepared by microtoming at room temperature and etched with a permanganic mixture, as described in an earlier article [7]. The etched surfaces were covered with a 4 nm thick platinum layer using a Balzers SCD 050 vacuum sputter coater and observed using a secondary electron detector at an accelerating applied potential of 15 kV. The results are presented in Fig. 3, where projecting crystalline lamellae are seen as bright features against a dark background.

Fig. 3:

Scanning electron micrographs showing lamellar structure in PE06, PE5, and PE9, revealed by microtoming followed by etching with permanganic acid. Slouf (code CZ).

2.4 Crystallinity and signs of degradation – infrared spectroscopy

Slouf (CZ) used infra-red spectroscopy to measure crystallinity and check for signs of degradation in moulded plaques. This technique is helpful in identifying chemical changes in UHMWPE bearing surfaces, especially in total joint replacements [8]. All three grades of PE feature notable peaks in the IR spectrum at 1303 and 1897 cm⁻¹, which are characteristic of the amorphous and crystalline components, respectively. Crystallinity was calculated from the areas of these two peaks. The values obtained corresponded to 70.6 % by weight in PE06, 57.8 % in PE5, and 55.8 % in PE9, with combined standard uncertainty, u_c, of ±2 % of the measured value.

Published studies have shown that degradation of PE samples results in peaks in the infra-red spectrum that indicate oxidation (>C=O at 1715 cm⁻¹) and crosslinking (trans>C=C< at 960 cm⁻¹). In the present study, measurements were made at three locations in a single specimen of each polymer. Both oxidation and trans-vinylene crosslinking peaks were negligibly small, with an integral intensity close to zero in all three grades of PE.

2.5 Crystallinity, crystallization and melting – SAXS, WAXS, DSC

Galeski (PA) used small-angle X-ray scattering (SAXS) to study lamellar thicknesses in compression moulded discs that were 25 mm thick and 136 mm in diameter. Two slices, 0.6 mm thick, were machined from a single disc of each material, one on a plane containing the central compression axis and the other on a plane parallel to the tangential surface of the disc. Lamellar structure was characterized using two-dimensional small angle X-ray scattering (2-D SAXS). A Kissing-type camera was placed at a distance of 1.2 m from the sample and coupled to a low-divergence CuK_α X-ray micro-source operating at 50 kV and 1 mA (GeniX Cu-LD from Xenics, France). The sealed-tube micro-source was integrated with multilayer collimation optics to produce a highly collimated beam with a divergence of 0.8 × 0.8 mrad². The collimation optics were combined with two additional hybrid scatter-free slit systems to produce a beam with a square cross-section. The two slit assemblies were placed 1.2 m apart. Scattering produced by the sample was recorded by a Pilatus 100 K solid-state detector with a resolution of 172 × 172 μm² (Dectris, Switzerland). For unoriented samples, the long period (P_L) was determined by applying Bragg’s law to 1D sections through 2-D scattering patterns. Background and Lorentz corrections were applied.

Figure 4 compares SAXS patterns from PE06, PE5, and PE9. They show that the lamellae in PE5 and PE9 are distributed preferentially on planes perpendicular to the compression axis of the disc, but there is no orientation in PE06, where the directions of lamellar axes are randomly distributed. Fig. 5 shows the relationship for all three grades of polyethylene between the scattering vector q and Iq², where I is the scattering intensity at angle θ, q is given by:

and λ is the X-ray wavelength. Long period (P_L) values determined from peaks in the SAXS profiles according to Bragg’s Law were 34.5, 41.5, and 44.5 nm for PE06, PE5, and PE9 respectively. Lamellar thicknesses, calculated from density correlation functions, were 9.5, 11.8, and 11.3 nm for PE06, PE5, and PE9 respectively, with a u_c of ±3 %. All three grades of UHMWPE have similar scattering profiles, except for a shift along the q axis due to reduced lamellar thickness in PE06, which can also be seen in Fig. 2. For this polymer, there is no sign of an extended tail at higher scattering angles, which would indicate the presence of thinner, more easily deformed lamellae.

Fig. 4:

SAXS patterns for PE06 (top), PE5 (middle), and PE9 (bottom). Galeski (code PA).

Fig. 5:

SAXS profiles for compression moulded PE06 (black), PE5 (red), and PE9 (blue). Galeski (PA).

Slouf (CZ) carried out small-angle and wide-angle X-ray scattering (SAXS and WAXS) tests on sections cut in three mutually perpendicular directions from compression moulded PE06, PE5, and PE9 discs. Cutting was performed under intense cooling in order to minimize changes in supramolecular structure. The first pair of samples (XZ and YZ) were cut parallel to the compression direction (Z axis). The second pair of samples, designated XYi and XYe, were cut on planes lying normal to the compression direction, where letters i and e denote interior and edge locations on the disc, respectively. This procedure made it possible to study the effects of orientation and sub-surface depth on supra-molecular structure. A Freiberger Präzisionsmechanik HZG/4A powder diffractometer was used to obtain a complete set of one-dimensional WAXS (1D-WAXS) patterns for PE06, PE5, and PE9. Samples XZ, YZ, XYi, and XYe were examined in two orientations, parallel (||) and perpendicular (⊥) to the long edges of the machined sections. Two-dimensional WAXS (2D-WAXS) and SAXS (2D-SAXS) diffraction patterns were recorded using a pinhole camera (Molecular Metrology SWAXS System) attached to a micro-focused X-ray beam generator (Osmic MicroMax 002) operating at 45 kV and 0.66 mA (30 W). Crystallinity, χ_c, was calculated from the ratio of I_cr, the integral intensity of diffraction from the crystalline phase, to I_tot, the total diffraction intensity. The 2D-SAXS patterns were converted to one-dimensional (1D) curves, with the Lorentz correction applied. Peak positions (q_m) were used to obtain long periods (P_L) according to Bragg’s law.

The results are shown in Table 1. Apart from two anomalously high readings in the 1D⊥ column, the data are consistent with the other data recorded in this report. Crystallinity was substantially higher in PE06 than in the two UHMWPE grades and long periods increased with molecular weight.

Slouf also used differential scanning calorimetry to measure crystallinity in compression moulded discs. Tests were carried out, under nitrogen at a heating rate of 10 K min⁻¹, using a Perkin-Elmer Pyris1 DSC. For each material, measurements were made on three separate samples, each taken from the centre of a single moulded disc. Average crystallinity values obtained were 56.4 % in PE06, 35.8 % in PE5, and 37.8 % in PE9, with uncertainty u_c =±3 %.

Handge (code HG) and Piorkowska (code PA) measured crystallinity using TA Instruments differential scanning calorimeters, models Q1000 and Q20. Samples between 4 and 6 mg were cut from the centres of compression-moulded plaques and heated under nitrogen at 10 K min⁻¹. They were cooled to room temperature and then heated again at 10 K min⁻¹. Percentage crystallinity was calculated assuming a specific enthalpy of fusion Δ_fusH⁰ of 293 kJ kg⁻¹ [9]. The same value of Δ_fusH⁰ was used for all calculations of crystallinity in polyethylene presented in this report.

The data shown in Table 2 are averages from two separate samples of each polymer. They demonstrate that crystallinity was much higher in PE06 than in PE5 and PE9 and that the same relationship was maintained through both heating stages, although levels of crystallinity increased after the first heating stage. These effects are in line with expectations. Hoffman and Miller have explained the reduced crystallinity of UHMWPE as the result of many macromolecular chains being involved in two or more adjacent lamellae, or at two widely-separated locations in a single lamella, so that chain segments between these anchor points have less freedom to participate in crystallization than those with unrestricted chain ends [10].

At any given temperature, the relaxation times in the non-crystalline phase are shorter in PE06 than they are in the two UHMWPE grades. Consequently, as the polymer cools in the mould, PE06 is able to come closest to the levels of crystallinity that are found, for example, in extrusion grades of linear polyethylene.

Piorkowska (PA) also used DSC to measure the melting point temperatures, θ_m, of powders. Her results are presented in Table 3. They show a substantial decrease in θ_m between nascent powder and samples that had been melted, cooled, and reheated. It should be noted that the powders have higher crystallinities than the compression moulded samples.

2.6 High pressure crystallization

Figure 6 is a schematic phase diagram for UHMWPE. The triple point was found at about 360 MPa and 215 to 220 °C. We note that Hikosaka et al. concluded that the triple point pressure and temperature were higher by at least 50 MPa and 20 degrees, respectively, based on experiments on a high-density polyethylene with a relatively low molecular weight [11].

Fig. 6:

Schematic phase diagram for polyethylene, showing TP and PT routes to the region in which hexagonal crystals are stable. On the TP route, the temperature is raised from 23 to 235 °C or 245 °C at atmospheric pressure, and the pressure is then raised to 480 MPa. On the PT route, the pressure is raised to 480 MPa, and the samples are then heated to 235 or 245 °C. Black arrows mark return paths for both routes [2], [11].

As indicated by the arrows and black points, the region in which hexagonal crystals are stable can be reached by raising the temperature to 235 or 245 °C and applying pressure to the melt (TP route) or by applying a pressure greater than 360 MPa to the sample and heating it to 235 or 245 °C to induce a solid state phase transformation (PT route). These transformations are particularly interesting, because experiments on standard grades of linear polyethylene show that entanglement densities decrease substantially under these conditions and thick pseudo-hexagonal crystals are formed [12], [13], [14]. In the hexagonal phase, lamellae also grow thicker than they do in the orthorhombic phase, so that X-ray reflections are difficult (or in some cases impossible) to detect. Furthermore, the structural changes induced in the hexagonal phase are not reversed when the sample is cooled to room temperature [15], [16].

In the present project, a group led by Piorkowska (PA) studied the effects of crystallization under high pressure on melting, crystallization, and phase transformation in PE06, PE5, and PE9. Following procedures used in earlier studies on medium molecular weight grades of linear polyethylene, they used an Instron tensometer fitted with a specially-designed high-pressure cell to compress samples of reactor powder between tightly-fitting 9.5 mm diameter plungers. A full description of the equipment used for this work is given by Masirek and co-workers [16]. Pressures and temperatures were controlled to an accuracy of ±0.5 MPa and ±1 K and applied sequentially, as illustrated in Fig. 6. On the TP route, a pressure of 1.5 MPa was first applied for 5 min to compact the powder slightly. The temperature was then raised to 235 or 245 °C, and the sample was held for 5 min before the pressure was increased to 480 MPa. When the PT route was followed, the pressure was first increased to 480 MPa, and the temperature was then raised to 235 or 245 °C. After annealing for 1 or 2 h, all samples were cooled to room temperature under full pressure, thereby inducing a solid-state transformation from pseudo-hexagonal to orthorhombic crystals.

The melting temperatures, θ_m, and enthalpies of fusion, Δ_fusH⁰ of both nascent powder and annealed samples were measured using a TA Instruments DSC 2920 calorimeter. Samples were heated to 200 °C at 10 K min^-1, cooled to room temperature, and then heated again to 200 °C.

The results of these tests are compared in Tables 4 and 5. They show that both θ_m and Δ_fusH⁰ increase during annealing at elevated pressure and temperature, the extent of the increase depending on the molecular weight of the polymer, the duration of annealing, and the sequence in which pressure and temperature are raised. In the case of PE06, χ_c increased from 76 % in the powder to about 90 % in annealed samples. High pressure annealing led to thickening of existing crystals (PT route) or to the formation of thick crystals from the melt (TP route). It can also reduce the number of chain folds and hence the free energy of basal surfaces of crystals.

The observed increase in melting temperatures is directly related to an increase in lamellar thickness L*, in accordance with the Gibbs-Thomson equation:

where G_e is the lamellar basal surface free energy, Δ_fusH⁰ is the heat of fusion, and T m 0 is the extrapolated equilibrium melting point. For PE, G_e is 9 μJ cm^-2 and T m 0 is 418.2 K [16]. In the limit, the chains are fully extended and there are no lamellae left. It is possible to reach this state in standard grades of HDPE, in which M ¯ w is less than 10⁵, but not in UHMWPE, because relaxation rates are too low at all temperatures that do not cause degradation.

It should be noted that θ_m values for UHMWPE samples crystallized under high pressures exceeded θ m 0 in some cases. This indicates that the crystals were not only very thick but were also severely constrained. It is reasonable to conclude that extremely long crystallization times would be required to reach a state in which the chains are fully extended, and that a significant amount of thermal degradation would take place during this period. It is interesting to note that the PE5 and PE9 samples had higher melting points than PE06 samples that were crystallized under the same conditions. The crystallinity reached at high pressures was also higher in the PE06 samples than in the PE5 and PE9 samples, as in samples crystallized under atmospheric pressure.

2.7 Fusion defects

To study the effects of processing conditions on the integrity of UHMWPE mouldings, Buckley (code OU) compression moulded three sets of plaques, at 150, 165, and 180 °C, respectively. Those dwell temperatures were selected on the basis of preliminary trials, which showed that mouldings made below 150 °C were too brittle to be machined and that processing above 180 °C produced no obvious improvements in strength. Dwell times at maximum temperature and full pressure were limited to 15 min. Applied pressure was kept relatively low during initial heating, then raised to 9 MPa when the mould reached 145 °C. After consolidation at the selected dwell temperature, the 3 mm thick mouldings were cooled rapidly under full pressure. Measurements based on Archimedes’ principle then gave densities of 954, 935, and 932 kg m⁻³ for PE06, PE5, and PE9, respectively.

Specimens cut from these mouldings were subjected to two types of tensile test at room temperature. In standard tests, specimens were stretched at constant strain rate, either to failure or to the limit of the extensometer, which in these experiments was reached when the specimens reached 600 % extension. In cyclic tests, specimens were repeatedly taken to a fixed strain, ε_max, unloaded to zero stress, and returned to ε_max. For the first four cycles, ε_max was set at 40 %. During the fifth cycle, ε_max was raised to 60 %. The specimen was subjected to a further four cycles with ε_max at 60 %. For the final five cycles, ε_max was increased to 80 %. The stress-strain behaviour observed during these tests is discussed in Part 3.

Optical microscopy (OM) and environmental scanning electron microscopy (ESEM) were used to study structure in the compression moulded specimens before and after tensile testing. Previous work showed that the PTFE spray used as a release agent in this program leaves a fine coating of PTFE particles on the moulded surfaces. Most of these were removed by washing with an acetone solution; subsequent inspection using ESEM found only a small scattering of PTFE particles. Sections approximately 50 μm thick were prepared using a sledge microtome and sandwiched between cover slips in a fluid with a refractive index matching the sample.

In optical microscopy, sections from untested PE06 showed no visible evidence of grain-boundary fusion defects. After load cycling, the tensile bars became significantly whiter, but it was not possible to see individual voids, which suggests that they were all below the limit of resolution for light microscopy (about 0.2 μm). Controlled cavitation of this type usually has the effect of enhancing the fracture resistance of the polymer. It reduces constraints close to the notch tip while leaving the load-bearing network of highly-stressed chains essentially intact [17]. The same toughening mechanism is responsible for the fracture resistance of all commercial grades of rubber-modified thermoplastics, including ABS, high-impact polystyrene, toughened PVC, and super-tough polyamides [18], [19]. From that perspective, polyethylene can be regarded as a rubber-toughened thermoplastic, in which PE chains form both the rigid, continuous crystalline phase and the particulate elastomeric amorphous phase in which cavitation is initiated.

Figure 7 shows a pattern of discrete domains in an untested PE9 specimen, with domain sizes corresponding approximately to the dimensions of reactor powder particles (50 to 250 μm). Similar patterns are seen in Fig. 8, where the section was taken from a cyclically loaded PE5 specimen. The faint parallel lines running horizontally across the micrograph are due to small irregularities in the cutting edge of the microtome blade. It is possible that forces imposed by the blade were responsible for exposing weak inter-particle boundaries in both specimens. The effects of cyclic loading are not apparent in Fig. 8.

Fig. 7:

Microtomed section from untested PE9 moulding, viewed in non-polarized transmitted light. Illuminated diameter of 0.8 mm. Buckley (OU).

Fig. 8:

Microtomed section from cyclically loaded PE5 moulding, viewed in non-polarized transmitted light. Loading direction horizontal. Illuminated diameter of 1.6 mm. Buckley (OU).

Evidence for inter-particle shear was obtained by concentrating on areas where the microtome blade cut through the outer surface of a tensile bar. Fig. 9 compares two sections viewed under polarized light. Before testing, the surfaces of test specimens were flat and featureless. After cyclic loading, they were roughened to a depth of about 15 μm, owing to unhindered shear at inter-particle boundaries adjacent to the surface.

Fig. 9:

Microtomed sections from PE5 tensile specimen, cut at 90 ° to the outer surface and viewed under polarized light. Top: untested specimen. Bottom: cyclically-loaded specimen, showing roughened surface. Tensile direction horizontal. Illuminated diameter 0.3 mm. Buckley (OU).

An alternative method for observing fusion defects, dark field imaging, is illustrated in Fig. 10. Dark field imaging is a special technique for observing discontinuities in translucent specimens using light microscopy. The images are formed only by scattered light, so that tiny holes and intergranular impurities produce bright spots, while larger holes and featureless solid regions appear black in the micrographs. Other types of discontinuity, including ‘grain’ boundaries between powder particles, generate intermediate shades of grey. All three levels of scattering can be seen in Fig. 10. In particular, the granular structure of the mouldings is clearly defined, indicating that bonding between powder particles is less than ideal in the mouldings prepared by the polymer manufacturer for this project. That does not necessarily mean that they were too weak to make satisfactory orthopaedic implants. To determine the extent to which visible grain boundaries constitute active fusion defects, which could cause wear or fatigue problems in the more demanding prosthetic applications, it is necessary to measure resistance to mechanical failure and fracture, and to fatigue crack propagation in particular (see Part 4 in this series).

Fig. 10:

Sections microtomed from UHMWPE mouldings, viewed in a light microscope using dark-field imaging: (a) PE06; (b) PE5; (c) PE9. Section thickness about 10 μm. Slouf (CZ).

Conventionally, fusion defects are divided into two categories. Type 1 defects are defined as crack-like voids, where there is no contact between adjacent reactor powder particles. They are caused by inadequate compaction of the powder and are easy to detect. Type 2 defects are regions in which contact has been made during melt processing, but the quality of the bond is less than perfect; in these regions the polymer is not fully consolidated. This description covers a very broad range of imperfections, ranging from extremely weak regions, where little reptation has taken place across grain boundaries, to moderately strong bonds that are quite well established but have not reached the optimum level of entanglement density. Because reptation rates are low in UHMWPE, it is extremely difficult (and probably impossible) to achieve perfect bonding, where all traces of inter-particle boundaries have been eliminated and entanglement levels are comparable with those found, for example, in cast PMMA.