Effect of Prestrain on the Fatigue Life of Superelastic Nitinol

Testing was conducted on a 50.8 atomic percent nickel straightened wire with a diameter of 0.28 mm. The mechanical properties were adjusted through thermal aging to achieve a UTS of 1330 MPa, and upper and lower plateaus of 545 and 340 MPa at 37 °C, respectively. Also, the transformation temperatures, measured by differential scanning calorimeter (DSC) per ASTM F2004 (Ref 14), were M_p = − 80 °C, M _p ^*  = − 9 °C, and A_f = + 8 °C (following the notation convention set forth in Ref 15 representing martensite formation peak, martensite reversion peak and austenite finish temperatures, respectively). The thermal aging process and transformation temperatures are representative of typical medical grade wires used for medical device applications. Prior to testing, wires were prestrained in tension to various strain levels at room temperature using an Instron equipped with a video extensometer. High-cycle fatigue tests were conducted to a maximum of 10⁶ (1MM) cycles¹ at a frequency of 20 Hz in a temperature-controlled water bath at 37 °C with a minimum of three wires per strain level. Readers are referred to Ref 16 for detailed description of the test setup used for this study. Strain amplitudes were estimated \(\left( {\varepsilon_{\text{a}} } \right)\) based on the wire diameter (\(d\)) and the radius of curvature of the mandrels (\(\rho\)) through the expression \(\varepsilon_{\text{a}} = d/2\rho\).

Figure 2 shows the strain amplitude vs. number of cycles fatigue (S-N curve) comparison for the specimens subjected to fully reversed bending without prestraining and after various tensile prestrains. The data demonstrate that fatigue behavior is strongly influenced by prestraining: with no prestrain, runouts are consistently observed at 0.6% strain amplitude, while the 8 and 10% prestrain conditions consistently ran out at 1.0 and 1.1% strain amplitude, respectively, corresponding to a 67 and 83% increase in tolerable strain amplitude. The effect of prestrain in improving fatigue life can be seen even at lower prestrains of 4 and 6%, even though macroscopic plasticity is not observed. Such improvements are consistent with prior published results (Ref 9-12), but seemingly inconsistent with others (Ref 13). One must take care, however, in that the latter testing (Ref 13) was prestrained in bending, not tension. The importance of this will be discussed later.

The tests were performed on the same wire as mentioned in the previous section. The midsection of wires was reduced from 0.28 mm to 0.22 mm in diameter through electropolishing to create dog-bone specimens. Prior to testing, specimens were prestrained in tension to various strains (4, 6, 9, 10 and 11%) using an extensometer to confirm the prestrain level in the gauge section. Prestraining to 11% and beyond was challenging as the samples began to neck. Hence, a minimal number of samples were tested at the 11% prestrain condition. After prestraining, the specimens were released to a 2% tensile strain and cycled at various strain amplitudes holding the mean strain at 2%. Testing was performed in water at 37 °C with a minimum of five samples at each strain amplitude. Cycling was conducted to 10⁷ (10MM) cycles at 24 Hz, at which time it was defined as a runout.

Tension–tension fatigue results (Fig. 3) show increased lifetimes with prestrain that are generally similar to that found in rotary bending. Runout strains are substantially lower in tension–tension loading compared with rotary bending because the former subjects the entire test specimen volume to the maximum strain amplitude rather than just the surface, increasing the probability of there being a critical defect. The magnitude of the improvement due to prestraining, however, is substantially greater than in rotating bending. Also, prestrain plays a prominent role in improving the low-cycle fatigue life of Nitinol when subjected to tensile fatigue (Ref 12) due to transformation-induced plasticity that produces dislocation-induced local residual stresses which lead to increase in fatigue lifetime.

In this tension–tension specimen, strain localization (so-called Lüders bands) expands to encompass the gauge volume as tensile strains increase to about 6%. Increasing strain beyond 6%, martensite elasticity is experienced up to about 8% strain, followed by plasticity with increasing strain. At strains exceeding 8%, the effects of plasticity apply uniformly throughout the gauge volume, which may explain the durability benefit observed at these elevated prestrain levels.

The diamond-shaped specimen used in this study is shown in Fig. 4. The intrados and extrados marked in that figure experience inverse stress states when the diamond is opened or closed. Figure 5, for example, shows the stress states that arise when the diamond is closed. (Details of the finite-element modeling are discussed in the subsequent section.) This action (closing the diamond) represents the most common scenario encountered in medical implants: a stent crimped into a catheter, for example. In the case of Fig. 5, the diamond is crimped until the maximum principal strain reaches 9% (a displacement of 4.7 mm). At the end of crimping, the maximum tensile strain and stress are found in the region of the extrados, while the maximum compressive strains and stresses are observed at the intrados (see inset at lower right). Importantly, if the crimping is sufficient to cause plastic deformation at the extrados, then releasing the load on the diamond results in residual stresses of the opposite sense: continuing the example above, a stent that is deployed freely from a constraining catheter results in residual compressive stresses at the extrados and residual tensile stresses at the intrados (inset at lower left). Had the diamond instead been stretched open for prestraining, the stresses in Fig. 5 would be of a reverse sense: tension at the intrados during deformation leading to compressive residual stresses upon release.

Returning to the above example of a stent compressed into a delivery catheter, in practice the stent is not freely deployed, but rather deployed into a vessel that constrains it at some diameter smaller than its free diameter: from there, a duty cycle is imposed, defined by a mean strain state and a cyclic strain. Stents are constrained in a radially compressed state; thus, the sense of stresses and strains at the extrados is of the same sense as the initial prestrain. From a testing perspective, however, one can also apply a duty cycle of the opposite sense: prestraining by closing the diamond (creating a tensile stress state at the extrados), but then stretching the diamond to impose tension at the intrados. This scenario is shown in Fig. 6. This ability to apply a duty cycle of the same or opposite global displacement as the prestrain allows one to reverse the residual stress state which will help to determine whether residual stresses are dominant in modifying the fatigue lifetime of superelastic Nitinol.

There are thus four possible loading scenarios as outlined in Table 1. Two of the four create compressive residual stresses at the locations that will be exposed to tension during cycling (conditions O-O and C-C), and two create tensile residual stresses at those locations (C-O and O-C). If residual stresses are controlling fatigue life, we expect the former situation to extend fatigue life and the latter to accelerate fracture.

The material used in this study was ultra-high-purity Ni_50.8Ti_49.2, also known as ELI (extra low interstitial), with a total C + O + N content of below 100 wppm. The diamond specimens shown in Fig. 4 were laser-cut from an 8.0 mm OD × 7.0 mm ID tube and expanded in multiple deformation and thermal aging steps to achieve a UTS of 1210 MPa, and upper and lower plateaus of 404 and 134 MPa at 37 °C, respectively. Also, the transformation temperatures measured by DSC per ASTM F2004 were M_p = − 53 °C, M _p ^*  = + 12 °C, and A_f = + 20 °C. The thermal aging process and transformation temperatures are representative of typical laser-cut medical devices. After electropolishing, the samples were measured and dimensionally characterized to assure accurate determinations of strain from displacements using finite-element analysis (FEA).

The finite-element model of the diamond specimens was meshed with C3D8R elements. The geometry was discretized with a mesh density of 12 × 12 elements across the strut width and wall thickness to determine the resulting stresses and strains. The 12 × 12 mesh density is representative of the mesh discretization used in modeling (including submodels) of typical medical devices that provides optimal solution. The final ABAQUS User Material (UMAT) properties of the aged material were determined from tensile testing of coupon samples subjected to heat treating representative of the diamond specimens. The key ABAQUS UMAT parameters are shown in Table 2, and for tension–compression asymmetry, a ratio of 1.5 (start of transformation stress in compression/start of transformation stress in tension) was assumed representative of typical Nitinol tubes (Ref 17). Sensitivity to material parameters was not analyzed as part of this study, as the primary intent is to understand the interaction of prestresses on fatigue life and not on the absolute value of the stresses/strains.

Testing was performed on a 12-station Instron E-3000 machine with load cells on each station. Six diamond samples were tested for each of the four combinations. Since fixation is rigid (Fig. 7), a fracture on one side of the diamond does not influence the conditions experienced by the other. Thus, each diamond can be considered to be two “V”-shaped test specimens. Hence, a total of 12 “V” specimens were tested per condition.

All samples were prestrained in a water bath controlled at 37 °C to a displacement corresponding to a 9% maximum strain (samples C-O and C-C by closing the diamond and samples O-O and O-C by opening). “Baseline” samples that were not prestrained were also included in the study. FEA was used to determine the required deformation and strains based on the actual part geometry. The fatigue strains (mean strain and strain amplitude) were computed using the tensor method by computing the average and differences of the principal values of the strain tensors at the extremes of the loading. It has been shown that both normal tensor method and modified tensor method consistently predict fracture or survival under different loading conditions for stents (Ref 18). The actual load–displacement curves for the four prestrain scenarios are shown in Fig. 8 with negative displacements corresponding to closing of diamonds and vice versa. Following the prestrain, all samples (including the reference samples) were stretched or closed to impart a 3.5% maximum mean strain and then cycled at 24 Hz with a strain amplitude of 0.75%. If no fracture was observed after 10⁶ (1MM) cycles, the test was considered a “runout” and the cyclic strain amplitude was increased while maintaining the same mean strain level. This was continued until a substantial number of fractures were observed. The baseline condition was only tested with a closing duty cycle (tension at the extrados), the most common scenario experienced by Nitinol medical devices. The experimental load–displacement of condition C-O in Fig. 8 corresponds to the stress strain history for the diamond specimen shown in Fig. 6.

As one would expect, scanning electron microscopy images (Fig. 9) verified that fractures initiated at the areas of highest tensile duty stress: the intrados of conditions C-O and O-O and the extrados of conditions C-C, O-C and the reference sample. Table 3 summarizes the results of the testing with number of “V” specimen fractures specified—the numbers listed indicate the cumulative number of fractures from the total of twelve “V” specimens. As with previous testing, prestraining has a significant effect on lifetime, but here we see that the effect can be an enhancement or a degradation depending on the sense of the prestrain with respect to the duty cycle: conditions O-O and C-C performed better than the reference samples, while conditions C-O and O-C performed worse. This is consistent with the expectation that compressive residual stresses are protective when superimposed onto areas of tensile stress, and tensile residual stresses accelerate the fatigue process, regardless of whether the intrados or extrados is critical (tension bearing). These same results are plotted based on probability of survival in Fig. 10, with the baseline condition denoted by dashed green line.

Lifetimes in this test are substantially longer than those resulting from the previous two testing modes. This can be attributed to the fact that this study employed ultrapure material which is known to have a significantly longer lifetime (Ref 19), because the peak strains and strain amplitudes are experienced by a very small volume of material.