Influence of Microstructure on the Performance of Nitinol: A Computational Analysis

To investigate the effect of SIM on the fatigue behavior of a NiTi stent device, a 3D homogenous FEA model of a “v-strut” stent subcomponent was created. To accurately simulate the transformational behavior of Nitinol, an in-built material subroutine UMAT/Nitinol was employed. This material model incorporates the elastic behavior of the austentic and martensitic phases only; plasticity is neglected. To accurately characterize Nitinol’s asymmetric behavior for input into the UMAT/Nitinol, uniaxial tension-compression experimental testing was performed on linear support struts of custom manufactured specimens (Fig. 2a). The manufacturing and processing techniques used on the specimens are consistent with those performed on commercially available endovascular stents; austenite finish temperature (A _f) was found to be 11.2 ± 2 °C using bend-free recovery method. A load-unload procedure to 6% strain is first performed prior loading to failure; this is to identify upper and lower plateau stresses (Fig. 2b). Due to the highly temperature-dependent nature of Nitinol, procedures were performed at 37 °C using the EnduraTEC/ELF 3200 with fan-heated environmental chamber.

To create FEA models with explicit representation of the crystallographic microstructure of polycrystalline superelastic NiTi, microstructural geometries of suitable specimens were required. NiTi wire samples underwent a recrystallization anneal and transformational behavior was subsequently analyzed using the bend-free recovery test; A _f was found to be 1.5 ± 0.5 °C. Specimens were subsequently manually polished and chemically etched. Finally, specimens were viewed under an optical microscope. Grain orientation distribution and average grain size were specifically identified (Fig. 3a); average grain size in the examined material was 20 µm. Utilizing the obtained micrographs, realistic FEA models were created which explicitly incorporated the granular microstructural nature of the superelastic austenitic material (Fig. 3b); development of these models will be discussed further in “Effect of Texture on Fatigue Behavior” section.

The potential influence of SIMT on the observed macroscopic response of Nitinol “v-strut” stent subcomponents is evaluated through identification of the volume fraction of SIM at critical locations at various strain levels in a 3D homogenous “v-strut” FEA model. The volume fraction of SIM is an explicit output, entitled “SDV21,” of the UMAT/Nitinol. This value was extracted as a nodal from the node exhibiting the highest value of maximum principal tensile strain following loading to the specified strain; this node was located in the inner apex of the “v-strut” geometry. The model was constructed with 27,138 8-node hexahedral (C3D8) elements; this number of elements was deemed necessary following a convergence study. The initial boundary conditions constrained the bottom surface in all three directions (Ux = Uy = Uz = 0); where U is the displacement. In the first step, a boundary condition is employed to displace the “top” surface by the specified value. Simultaneously, constraining boundary conditions are placed on the nodes on the respective sides of the model to prevent out-of-plane motion (Uy = Uz = 0). Three strains spanning the experimentally established superelastic plateau of the specimen are targeted for analysis (namely, 2.5, 4, and 6%) to investigate various physiologically relevant conditions experienced by a stent (Fig. 4b). It is hypothesized, at sufficiently high mean strains, SIM is present and is stabilized, and thus resulting in increased fatigue performance.

To evaluate the effect of microstructural texture on Nitinol’s macroscopic response, a 2D rectangular model was created using Voronoi Tessellation based on the obtained micrographs (Fig. 3b). As the images had an aspect ratio close to 1, it was assumed that representation of the complex realistic geometries was not necessary; idealized hexagonal units were instead used (Fig. 3c). Models were constructed using 7600 2D 8-node biquadratic plane stress quadrilateral (CPS8) elements. As symmetry was assumed, movement was restricted along surfaces A and B in the Ux and Uy directions, respectively (Fig. 3c). Surfaces C and D were constrained to replicate the Uy and Ux movement, respectively, of the master node (as shown on top right corner of model in Fig. 3c). The constitutive equations for these micro-scale models were established using single crystal data adopted from the literature (Ref 9); material properties were extracted from uniaxial tension-compression stress-strain curves for each crystallographic orientation. The grains are assumed to have homogenous behavior and the properties vary according to their crystallographic orientation to the loading direction and grain boundaries were assumed to be perfectly bonded with no inter-granular slip. Employing FEA, the effect of individual grain orientation on SIMT in randomly generated microstructural textures was also successfully investigated in the complex “v-strut” geometry; this will be discussed further in “Effect of Texture on Fatigue Behavior” section of this paper.

The highly advantageous trait exhibited by Nitinol of increased fatigue performance with increasing mean strain remains unexplained in the literature. In this paper, it is hypothesized that the increased fatigue life is attributed to the phase transformation (SIMT) experienced by the superelastic material. In this study, the volume fraction of SIM is analysed at regions of peak stress during physiologically relevant compressive loading. When assuming homogenous NiTi material properties represented within the models as a continuum, the volume fraction of SIM at region of peak stress (inner apex) is shown to increase with increasing strain (Fig. 5). However, the most significant location for analysis is that of peak tensile stress, located on the outer apex of “v-strut” geometry; as the potential site of fatigue failure initiation. The volume fraction of SIM at this critical location is found to follow a similar trend to that of the experimentally established constant life diagram for Nitinol (Ref 1, 2) (Fig. 6). Furthermore, the observed on-set of increasing fatigue performance correlates with the on-set of SIM at approximately 1.5% mean strain level which presents further evidence for the proposed hypothesis (Fig. 6).

At these critically high strain levels, deformed martensite structures and defects are hindered leading to differing dominant de-twinning, reorientation modes, and dislocation densities in discrete areas of the microstructure, resulting in the presence of stabilized SIM (Ref 7, 8). This localized strengthening of Nitinol (by means of a proposed strain hardening effect (Ref 8)), in conjunction with the self-accommodating nature of the SIM itself, may provide an explanation for the observed improved fatigue performance of Nitinol.

In applying a load-unload procedure on the idealized hexagonal rectangular models described in “Effect of Texture on Fatigue Behavior” section, it is found grains orientated in (110) crystallographic direction inhibit, while grains in (111) orientations promote the SIMT (Ref 9). This becomes evident when comparing Fig. 7(a) and 7(b); a higher volume fraction of (111)-orientated grains in model (b) results in notably lower stress values along the loading plateau due to the ease at which SIMT occurs, as compared to model (a) with more (110) grains.

This variation in transformation behavior in the individual grains results in localized stress gradients over grain boundaries; these peak stresses become particularly important when discussing fatigue behavior in the “v-strut” stent-like specimen as potential fatigue failure initiation sites. To further explore this, the influence of texture was investigated in the “v-strut” stent subcomponent; a 2D idealized hexagonal model was created for analysis incorporating the idealised hexagonal grain geometry (Fig. 8).

In analysing the volume fraction of SIM at the critical location of maximum tensile stress (located at the inner apex of “v-strut”) with varying crystallographic textures, a similar trend to the experimentally established constant life diagram for Nitinol (Ref 1, 2) (Fig. 9) was observed. The three colored lines in Fig. 9a represents the stress-strain response of the “v-strut” FE model with either (100)-, (110)-, or (111)-orientated grain at the critical location (see Fig. 8). Variation in material response displayed is attributed to the different crystallographic textures generated in the “v-strut” models; in particular, varying grain orientations at critical location of maximum tensile stress impacts the macroscopic response. For example, due to the presence of (110)-orientated grains at critical location in the “v-strut” geometry (Fig. 8a), SIMT has been hindered and therefore localized peak strains have been induced. This may also have a role in explaining the significant scatter that is observed in experimental fatigue data.