Microstructure and texture development during high-strain torsion of NiAl

In this study polycrystalline NiAl has been subjected to torsion deformation. Torsion has been used because of its characteristics. By this deformation mode high shear strains (gamma = 18 in this study) can be imposed on the sample. The deformation conditions are well-defined because of the local deformation mode, which is simple shear. Due to the monoclinic sample symmetry one half of the pole figure is needed in order to obtain the complete texture information, which is more than is needed e.g. by extrusion or rolling. Therefore, texture analysis might be more sensitive with respect to texture components. Furthermore, torsion deformation is characterized by being inhomogeneous in terms of the amount of shear strain and shear strain rate along the sample radius. The shear strain gradient makes the analysis of different stages of deformation on the same sample (i.e. under the same deformation conditions) possible. Another characteristic being special for torsion is that samples change their length, although no axial stress is applied. This effect is known as Swift effect and will be analyzed in detail. The deformation, microstructure and texture development subject to the shear strain are studied by different techniques (Electron Back-Scatter and High Energy Synchrotron Radiation). Beside the development of microstructure and texture with shear strain, the effect of an initial texture as well as the deformation temperature on the development of texture and microstructure constitute an important part of this study. Therefore, samples with three different initial textures were deformed in the temperature range T = 700K – 1300K. The development of the microstructure is characterized by two different regimes depending on the deformation temperature T. For T up to 1000K, continuous dynamic recrystallization (CDRX) takes place. This mechanism leads to the deformation-induced dislocations forming low angle grain boundaries (LAGBs) or being incorporated into them and the successive transformation of these boundaries into high angle grain boundaries (HAGBs) by a further increase of their misorientation. The predictions of this model were compared with the experimental results. The shear stress – shear strain curves are characterized by a peak at low strains, which is followed by softening and a steady state at high strains. This condition is fulfilled for a number of samples, but especially <111> oriented samples do not show a softening stage at low temperatures. Grain refinement takes place for all samples and the average grain size decreases with temperature. The predicted LAGB decrease is in best agreement with the experiments at the lowest temperatures (T = 700K and 800K). Deviations from the model can be explained by the temperature dependence of the grain boundary mobility. For temperatures T > 1000K, discontinuous dynamic recrystallization (DDRX) occurs, by which new grains form by nucleation and subsequent growth. The texture is characterized by two components, {100}<100> (cube, C) and {110}<100> (Goss, G). The intensity of G increases with temperature, while that of C decreases independent of the initial orientation. Both components have their maximum deviated about the 1 axis. The deviation is larger for grains containing the C component and decreases with temperature. Grains containing the G component have the smaller deviation, which decreases with temperature and strain. Texture simulations based on the full constraint Taylor model under the assumption of {110}<100> and {110}<110> slip were done with the experimental <110> and <111> fibres as well as a theoretical <100> fibre and a {100}<100> single orientation (ideal as well as rotated about the torsion axis). The G component is predicted by the simulations and is therefore a deformation texture. However the C component does not appear in the simulation. It therefore must originate by different mechanisms. For the non-<100> oriented samples, possibly nucleation is responsible for the formation of C oriented nuclei. Simulations with single orientations lead to the conclusion, that the ideal C orientation rotates about the 1 axis, while other C orientations, which are rotated about the torsion axis, increasingly converge towards the G component with strain. A single G orientation on the other hand is stable against such a rotation and is therefore the most likely steady state texture. Based on these results it is proposed, that ideally C oriented nuclei rotate until an orientation is reached into which they grow. These new grains are further rotated up to a critical angle, at which a part of them disappears either by adjacent grains or new C oriented nuclei. The recrystallization texture for T > 1000K is most likely the C component as well. Torsional creep of NiAl is characterized by a stress exponent, which depends on temperature and an activation energy, which is stress dependent. A model incorporating both dependencies is proposed and applied to the creep data. It is shown that these equations are able to describe the experimental findings. Thus creep of NiAl based on this model is dominated by non-diffusional processes such as cross slip of <100> screw dislocations for T 1000K. For T > 1000K the stress exponent and the activation energy are in a region, which according to previous reports is rather dominated by dislocation-climb controlled creep. The Swift effect, due to which samples change their axial dimension during torsion without applied axial stress, is observed for NiAl. It is strongly related to the texture development and in the case of NiAl the C component is identified as being responsible for shortening, whereas the G component leads to lengthening as long as it is not aligned with the shear system. Both tendencies can be explained based on the active slip systems. Simulations fail to predict the experimental observation, because the C component is not present. HESR and EBSD were compared with respect to local texture measurements. It was concluded depending on the average grain size HESR has an advantage in terms of grain statistics. For DDRX samples however, both methods are limited. Local texture inhomogeneities can be better detected using EBSD, whereas for an overall local texture information HESR is better suited