Structure, processing and performance of ultra-high molecular weight polyethylene (IUPAC Technical Report). Part 3: deformation, wear and fracture

2.1 Materials

Three large batches of polyethylene, with weight average relative molar masses of about 0.6 × 10⁶, 5 × 10⁶, and 9 × 10⁶ were supplied by a major manufacturer of UHMWPE. The batches are identified in this report by the codenames PE06, PE5, and PE9, respectively. For convenience, the term UHMWPE is applied to all three materials, although PE06 should strictly be classified as HMWPE, because its M ‾ W is below 10⁶. Except for one set of plaques moulded at Oxford University (see Section 2.2 below), all mechanical tests were carried out on compression mouldings made by the polymer manufacturer purely from as-polymerized reactor powder, without additives, as the company’s contribution to the IUPAC project. Consequently, the materials tested during this study do not necessarily have the same properties as commercial grades of UHMWPE.

2.2 Standard tensile testing

Three laboratories carried out conventional uniaxial tensile tests at 23 °C, in accordance with ASTM and ISO standards [1], [2]. Following standard practice, most of the resulting data are expressed in this section in terms of nominal stress, σ ( = F / A 0 ) , and nominal strain, ε ( = Δ L / L 0 ) , where F is the applied tensile force, A₀ is the initial cross-sectional area, and L₀ is the initial gauge length. However, in some cases it is more convenient to discuss relationships between true stress, σ_t, and true strain, ε_t, which are defined as follows:

Results from one set of tensile tests are presented in Fig. 1 and in Table 1, where the combined standard uncertainty, u_c, for yield stress is ±2 %. They show that PE06 has the highest yield stress and the largest elongation at break. There are only small differences between the stress–strain curves of the two UHMWPE grades. Both have relatively low yield stresses, but subsequently strain-harden quite rapidly before breaking at an elongation of about 250 %.

Fig. 1:

Nominal stress–strain curves for PE06, PE5, and PE9. Nominal strain rate 0.833 mm/s. Data of Handge (HG).^[1]

Handge also measured secant modulus E at 23 °C, over the strain interval 0.05 to 0.25 %, at a strain rate of 3.3 × 10⁻⁴ s⁻¹. The resulting values of E were 1020 MPa for PE06, 624 MPa for PE5, and 610 MPa for PE9, with u_c = ±1.5 %.

Instead of testing mouldings made by the UHMWPE manufacturer, Buckley (OU) carried out a separate program of work to investigate the effects of molecular weight and consolidation temperature on mechanical properties by deliberately using below-optimum processing conditions to make some of the mouldings. Reactor powders from batches PE06, PE5, and PE9 were compression moulded at three dwell temperatures: 150, 165, and 180 °C, to produce sets of 3 mm thick sheets measuring 140 × 200 mm. The mould was heated gradually over about 30 min, with pressure applied when the temperature reached 145 °C. After reaching the selected dwell temperature, each melt was held at full pressure for 15 min; the mould was then cooled to 100 °C within 3 min. Previous compression moulding experiments at Oxford have shown poor consolidation of UHMWPE at temperatures below 150 °C, while raising the dwell temperature above 180 °C not only has little effect on consolidation, but also can cause significant oxidative degradation, leading to losses in both strength and ductility. A CNC milling machine was used without cutting lubricant to prepare dog-bone test specimens with 20 mm gauge length and 105 mm grip separation.

Tensile tests were carried out at 20 °C, in accordance with the ISO standard [2]. Results from these experiments are set out in Fig. 2 and Table 2. The one factor that clearly affects initial yielding is crystallinity. Both elastic modulus and yield stress are higher in PE06 than in PE5 or PE9, because both of these true UHMWPE grades exhibit retarded relaxation while they are cooling in the mould, as discussed in Part 2, so that crystallization is restricted. During subsequent stretching, drawing stresses in PE06 were not only lower than in PE5 and PE9, but they also showed less variation with moulding temperature. All three PE06 specimens extended to nominal strains of over 600 %. Tests were then terminated because grip separation in the Instron tensometer had reached its limit. At that point, the nominal strain was 198 % and nominal stresses were over 140 MPa. In PE5 and PE9, the curves for mouldings made at different temperatures began to diverge after reaching a nominal strain of 40 % and a nominal stress of 28 MPa. Coincidentally, the curves for all nine specimens meet at that same point on the loading curve. Fracture occurred at nominal strains between 140 and 155 %. The similarity in behaviour between PE5 and PE9 shows that raising M ‾ W from 5 × 10⁶ to 9 × 10⁶ did not produce the expected effects of increased entanglement density in PE9. The simplest explanation for this result is that entanglement densities at ‘grain’ boundaries were lower than in the bulk of the moulding and made a disproportionate contribution to yielding and tensile drawing.

Fig. 2:

Stress–strain curves for PE06, PE5, and PE9 at 20 °C, showing the effects of moulding temperature. Data from Buckley (OU). Laboratory record in which symbols σ_t and ε_t represent tensile stress and strain.

2.3 Cyclic tensile testing

Tensile specimens cut from plaques moulded at 150, 165, and 180 °C were subjected to cyclic loading at 20 °C, with a nominal strain rate dε/dt of 5 × 10⁻³ s⁻¹. Each specimen was first stretched under uniaxial tension to a maximum nominal strain, ε_max, of 40 %. It was then subjected to five loading cycles between a tensile stress, σ, of zero and ε_max of 40 %. At the end of the fifth cycle, the maximum strain was raised to 60 % and the specimen was subjected to a further five cycles between zero stress and the new ε_max. For a final five cycles, ε_max was increased to 80 %. Results from these tests are presented in Fig. 3 as nominal stress–strain curves. All specimens produced large hysteresis loops. The effects of varying moulding temperature were found to be negligible and the differences between PE5 and PE9 were again small. As in the standard tensile tests, PE06 was distinguished by its high yield and drawing stresses during the early stages of each test.

Fig. 3:

Nominal stress–strain curves for PE06 and PE5 obtained in cyclic loading tensile tests at 19 °C and Hencky strain rate of 0.05 s⁻¹. Data from Buckley.

To clarify the discussion of plastic deformation under cyclic loading, it is helpful to convert nominal stresses and strains into true stresses and true strains. A quantity of particular interest in the present context is the relative residual true strain, ε r ∗ , which measures the residual strain in the specimen after unloading relative to the strain reached at peak load during that most recent cycle. It is defined as follows:

where ε_t,max is the maximum true strain at peak load for each cycle and ε_t,r is the residual true strain after subsequent unloading. Table 3 shows that ε r ∗ is unaffected by moulding temperature over the range 150–180 °C but increases with maximum true strain. Residual strains were higher in PE06 than in PE5 and PE9, indicating that plastic flow processes played a greater part in the deformation of PE06 than they did in PE5 or PE9, where the exceptionally long chains of these polymers hindered relaxation and promoted elastic behaviour.

In cyclic loading tests, the specific hysteresis energy ΔW is defined as the energy dissipated per unit volume during a single cycle, and the specific work input W is the total work input per unit volume during the loading stage of that cycle. For each hysteresis loop, values of ΔW and W were calculated from true stress–strain data. Subscripts identify specific loops within a set of five that were all taken to a particular maximum strain. Thus ΔW₄ measures the area enclosed by the fourth hysteresis loop in a given set, while W₄ measures all of the area under the fourth loading curve, between zero stress at the beginning of that loading-unloading cycle, and σ_t,max. The fourth cycle in each set is of particular interest because at that stage hysteresis has almost reached a steady state; the fifth cycle is less suitable as a basis for comparison because maximum strains of 40 and 60 % are followed by large increases in ε_t,max.

The results of this analysis are shown in Fig. 4. To a good approximation, all of the points lie on a single straight line through the origin, thereby demonstrating that the same deformation mechanisms govern energy dissipation in all three polymers. To that extent, PE06, PE5 and PE9 exhibit the same behaviour. Fig. 4 shows that the main difference between them is that maximum values of W and ΔW are relatively high in PE5 and PE9, while they are relatively low in PE06. Two factors contribute to these differences: (a) at every stage of the test, residual strains at zero stress are lower after unloading in PE5 and PE9 than in PE06, so that larger strain increases are required to reach ε_max; (b) strain hardening effects are much more prominent in PE5 and PE9, so that peak stresses rise more rapidly in these materials. Despite the significantly higher initial yield stresses exhibited by PE06 because of its high crystallinity, it reaches slightly lower peak stresses than PE5 or PE9 during the final five cycles of the test, where ε_max = 80 %.

Fig. 4:

Energy dissipated through hysteresis plotted against work input per unit volume in PE06 (open black symbols), PE5 (red symbols), and PE9 (blue symbols). Data obtained from fourth cycles in Fig. 3, at ε_max = 40 % (☐), 60 % (○), and 80 % (Δ). Data from Buckley (OU).

Optical microscopy was used to study the damage caused by cyclic loading. Small blocks of UHMWPE were cut for the microscope stage and the original moulded surfaces were carefully cleaned using an acetone solution. Nevertheless, the samples showed evidence of embedded debris, in addition to many scratches. To detect fusion defects and characterize anisotropy, sections were therefore prepared from the interiors of both freshly made and previously tested specimens. For examination under polarized light, each section was cut to a thickness of 50 μm, on planes perpendicular to the tensile bar’s cross-section, then mounted between a glass slide and a cover slip in a liquid of matching refractive index.

In addition, environmental scanning electron microscopy (ESEM) was used to examine specimens both before and after testing. Prior to testing, there was little detectable evidence of surface void formation in any of the specimens. Following cyclic loading, both PE5 and PE9 appeared to be undamaged, but the PE06 specimens displayed a high density of surface voiding, as illustrated in Fig. 5.

Fig. 5:

Surface voids formed in a PE06 bar subjected to cyclic tensile loading in the horizontal direction, parallel to the voids. Data from Buckley (OU).

2.4 Small punch testing

Small punch tests (SPT) were carried out by Slouf at (23 ± 1) °C in accordance with the ASTM standard [3]. This test is designed specifically for testing UHMWPE, using specimens taken either from pristine manufactured implants or from implants retrieved from the human body [4]. The specimens are 0.5 mm thick and 6.4 mm in diameter. They are clamped around their circumference against the rim of a cylindrical die with an inner diameter of 3.81 mm. The punch is a cylindrical rod with a diameter of 2.54 mm and a hemispherical head, which is driven into the die at a velocity of 0.5 mm/min. In this study, at least five specimens of each material were tested. The standard requires that average and standard deviations of peak load, ultimate load, ultimate displacement, and work to failure be recorded.

The graphical results of small punch tests are presented in Fig. 6. They are broadly similar to those obtained in standard and cyclic tensile tests. The force-deflection curves for PE06 rise steeply to a peak, but the polymer then strain-softens quite rapidly. Curves for PE5 and PE9 have lower initial slopes and slightly lower peak forces. Beyond maximum load, the curves for both UHMWPE grades diverge by a small amount; strain softening takes place more slowly than in PE06 and is followed by strain hardening, which is more prominent in PE5 and PE9 because of their higher entanglement densities. The curves reach a minimum at a deflection of about 4 mm, which is a convenient point at which to compare the force data. This comparison demonstrates that the defining force F_{4 mm} at this deflection is only moderately sensitive to variations in the molecular weight of the selected polymers.

Fig. 6:

Force-deflection curves from small punch tests on PE06, PE5, and PE9, each represented by five specimens. Data from Slouf (CZ).

2.5 Channel die compression

Galeski (PA) carried out channel die compression tests at a true strain rate of 0.067 s⁻¹ in a 6 mm wide, 49.3 mm long rectangular channel. The compression tool was a closely fitting rectangular block. Specimens were 6.0 mm wide, 23.4 mm high, and 49.3 mm long. A full description of the test method is given in an article by Bartczak et al. [5]. Basically, the specimen extends along the length of the channel as it is compressed vertically. As illustrated in Fig. 7, the onset of plastic deformation is marked in all three grades of UHMWPE by a change in slope rather than a clear peak. It is therefore difficult to obtain accurate data on yield stresses from channel die compression tests. Best estimates of yield stresses in compression at 22 °C are 47 MPa for PE06 and 31 MPa for PE5 and PE9, with combined standard uncertainty u_c = ±5 %. These relatively high values are due to the pressure-dependence of plastic deformation, which is a characteristic feature of yielding in thermoplastics. As in the tensile tests, raising M ‾ w from 5 × 10⁶ to 9 × 10⁶ had little effect on deformation behaviour, again indicating that increasing the entanglement density in the central regions of the powder particle ‘grains’ during compression moulding was not translated into higher resistance to deformation at room temperature. This observation supports the view that deformation in lightly entangled ‘grain’ boundaries made a disproportionately large contribution to yielding in compression.

Fig. 7:

Channel die compression curves showing compressive stress σ as a function of strain ε for PE06, PE5, and PE9 at 22 °C. Data from Galeski (PA).

2.6 Microhardness

Slouf carried out Vickers microhardness tests at 23 °C using a standard indenter [6]. Samples were first spattered with a 4 nm thick layer of platinum. Loads of 5, 10, and 50 g were then applied for 6 s, and the diagonal lengths d_i of the indentations were measured using an optical microscope. Plots of applied force F against d² were linear. In these tests, Vickers microhardness, H_Vμ, is given by:

where the unit for microhardness is MPa, F is the force in N (Newton) and d_i is the indentation diameter in mm. Quoted results for H_Vμ are based on the average of 10 separate measurements. The hardest material was PE06, with H_Vμ = 57.6 MPa. Hardness values for PE5 and PE9 were substantially lower, at 41.5 MPa, and 40.3 MPa, respectively, with u_c = ±2 %. Once again, the increase in M ‾ W from 5 × 10⁶ to 9 × 10⁶ had little effect.

2.7 Wear testing

Slouf used an OrthoPOD multidirectional pin-on-disc (POD) machine to measure wear rates, following the ASTM standard [7]. The metal discs were polished to a surface roughness Ra of 0.1, and the load on the UHMWPE pins was 4.0 MPa, which is approximately the contact stress in human joints.^[2] The average pin-to-disc velocity was 6 cm/s. The total multidirectional path for each pin corresponded to 1 million cycles. Wear rate was calculated as the ratio of the mass lost by the sample to the mass lost by a control, which was a standard non-crosslinked Chirulen 1020 grade of UHMWPE. At least three repeat measurements were made on each material. The measured wear rates for PE06, PE5, and PE9 were 6.57, 0.86, and 1.52 respectively.

Wu used the Durham multidirectional pin-on-plate machine to perform bio-tribological wear tests on PE06, PE5, and PE9, following an ASTM standard with respect to patterns of motion, sliding speed, lubricant, and temperature [7]. The tests were carried out at 60 rpm, with 25 mm strokes at a sliding speed of 50 mm s⁻¹. A load of 40 N was applied vertically to each pin to give a contact stress of 2 MPa. The pins and plates were lubricated with 25 % (v/v) diluted bovine serum (Harlan Laboratories Limited, UK), which gave a protein content of approximately 17.5 g L⁻¹, with 0.2 % sodium azide (NaN₃) added to inhibit bacterial growth, along with 20 mM ethylene-diamine-tetra-acetic acid (EDTA) to prevent calcium deposition. The wear tests were conducted for up to 3 million cycles, stopping approximately every 0.25 million cycles to enable the pins and plates to be cleaned in accordance with the ISO cleaning/drying protocol [8], prepared for surface examination, and weighed using a Mettler AE 200 balance with an accuracy of 0.01 mg. Finally, before resuming the next 0.25 million cycles, fresh lubricant was applied to the clean surface.

The results are presented in Fig. 8. They show that PE06 had the highest rate of wear, at 0.153 μg m⁻¹, while the wear rates of PE5 and PE9 were lower and almost identical, at 0.08 μg m⁻¹. Under an optical microscope, a grain boundary pattern is clearly visible on the worn surface of the PE9 specimen shown in Fig. 9.

Fig. 8:

Curves showing loss of mass m_wr with increasing length of travel L_wr in wear tests on PE06, PE5, and PE9. Wu (DU).

Fig. 9:

Surface of PE9 plate after 3 million cycles of bio-tribological wear testing at 37 °C. Wu (DU).

2.8 Fatigue crack growth testing

Wu studied fatigue crack propagation (FCP) in compact tension (CT) specimens under sinusoidal loading at 5 Hz, at 23 °C and 50 % relative humidity, in accordance with the ASTM standard [10]. Specimens 35 mm high and 36.5 mm wide were cut from 3 mm thick compression moulded plates provided by the UHMWPE manufacturer. Starter-crack slots were cut from one edge of each specimen to a length, a₀, of 13.5 mm along the central plane, with faces 1 mm apart. Their tips were sharpened with a razor blade. Crack-opening cyclic tensile stresses were applied through pneumatic clamps attached across the upper and lower edges of the test specimen. The crack-tip stress intensity factor, K, was varied between a minimum, K_min, and a maximum, K_max, with R (the ratio K_min/K_max) set at 0.1. The magnitude of K in fracture mechanics is determined by the applied stress, σ, the crack length, a, and the geometry of the specimen, in accordance with the following equation:

where Y is a numerical factor determined by the geometry of the test specimen [10].

Preliminary trials indicated that an initial crack-tip stress intensity amplitude ΔK of 1.75 MPa m^0.5 would initiate stable crack growth at a measurable rate. From an initial crack length a_o of 13.5 mm, a preferred nominal stress amplitude Δσ of 4.47 MPa was identified. For each cycle, nominal applied stresses therefore ranged from a minimum of 0.497 MPa to a maximum of 4.97 MPa. The nominal stress is defined as the applied force, F, divided by the nominal cross-sectional area, BW, where B is thickness and W is width measured from the loading point. Earlier trials showed that temperature rises were negligible during fatigue crack propagation at 5 Hz in the 3 mm thick CT specimens.

With Δσ and R held constant, ΔK increased with increasing crack length a. Specimens were marked with a felt pen at increments of 2 mm along the anticipated route of crack propagation, and tests were filmed using a Casio EXILIM EXZ720 digital camera with a resolution of 7.2 megapixels. Crack length a was recorded as a function of time t. The film was used to determine the time intervals required for visible increments in stable crack growth. These data were used to calculate crack propagation rate, da/dN, as a function of ΔK, where N is the number of loading cycles.

Results from these tests are presented in Fig. 10, which covers the regions of stable crack growth in all three materials. In this region, propagation rates typically follow the Paris equation [11]:

where C and m are coefficients determined by the material’s properties. Fatigue crack propagation can generally be divided into three stages. During the initial stage, there is little observable crack growth, but load cycling finally leads to a transition into the second stage, where stable crack propagation takes place at an increasing rate. In the third, terminal stage of the test, da/dN is large enough to make propagation unstable. Since stable crack growth behaviour is of the greatest interest in the present context, data from the initial and final stages of propagation are not included in Fig. 10. Both PE5 and PE9 clearly follow Equation (6), with m equal to 4.8 and 4.9 respectively. There are not enough data on PE06 to determine whether this grade of UHMWPE obeys Equation (5) in the stable growth region.

Fig. 10:

Relationship between stress intensity amplitude ΔK and fatigue crack growth rate da/dN in 3 mm thick CT specimens. Data of Wu (DU).

The most significant and unexpected result recorded in Fig. 10 is that resistance to fatigue crack propagation decreased with increasing molecular weight, so that the highest crack growth rates were exhibited by PE9. The obvious conclusions to be drawn from this observation are that (a) bonding between reactor powder particles was relatively weak, so that there was no gain in fracture resistance when M ‾ w was raised from 5 × 10⁶ to 9 × 10⁶; and (b) fatigue tests are particularly effective at exposing this kind of weakness. The levels of entanglement at grain boundaries between adjacent powder particles appear to be relatively low in both UHMWPE grades, especially PE9. Environmental scanning electron microscopy showed that growth steps at intervals of approximately 2 μm were formed on the fracture surfaces of both PE5 and PE9 specimens towards the end of the fatigue test, but the fracture surface of PE06 was almost featureless – see Fig. 11.

A more extensive study led by Altstädt (UB) found that fatigue crack propagation behaviour in PE06, PE5, and PE9 was strongly dependent on the thickness of compact tension specimens. There was a certain amount of scattering in the data from 4 mm thick specimens, but da/dN generally increased with increasing ΔK, approximately in accordance with Equation (6). By contrast, crack propagation in almost all of the 10 mm thick specimens tested was extremely erratic, and in some cases impossible to initiate. Because of the significance and complexity of these results, they are presented and discussed separately in Part 4, the final report in this series).

Fig. 11:

Environmental scanning electron micrographs of fracture surfaces from compact tension specimens subjected to fatigue tests by Wu (UD); left PE06; right PE5.