Deformation twinning in AZ31: Influence on strain hardening and texture evolution

Abstract

This paper describes the main results from an experimental investigation into the consequences of deformation twinning in AZ31 on various aspects of plastic deformation, including the anisotropic strain-hardening rates, the tension/compression yield asymmetry, and the evolution of crystallographic texture. It was seen that AZ31 exhibited unusually high normalized strain-hardening rates compared to α-Ti that occurred beyond the strain levels where extension twins have completely altered the underlying texture. This observation challenges the validity of the generally accepted notion in the current literature that the high strain-hardening rates in AZ31 are directly caused by extension twins. It is postulated here that the thin contraction twins are very effective in strain hardening of the alloy by restricting the slip length associated with pyramidal 〈c + a〉 slip. This new hypothesis is able to explain the major experimental observations made in this study and in the prior literature. We have also presented a new hypothesis for the physical origin of the observed differences in the thicknesses of the extension and contraction twins. The stress fields in selected matrix–twin configurations were modeled using crystal plasticity finite element models. The contraction twin $(0\overline{1}11)$$[0\overline{1}1\overline{2}]$ was predicted to form an internal extension twin $(01\overline{1}2)$ $[0\overline{1}11]$, resulting in the commonly observed “double twin” sequence. The extension twin is suggested to inhibit thickening of this double twin by loss of twin–matrix coherency. Extension twins were predicted to retain their coherency and thus thicken.

Introduction

Magnesium alloys are being increasingly evaluated for applications in automotive components due to their outstanding properties such as low density, high specific strength, castability, machinability at high speeds, and weldability under controlled atmospheres. The automobile industry is already utilizing cast magnesium alloys [1]. Although wrought magnesium alloys are expected to possess better mechanical properties [2], the main hurdle has been their limited room temperature formability. Enhancing the formability of Mg alloys requires a better understanding of the deformation mechanisms and their influence on the overall mechanical behavior of these alloys.

Pure magnesium has a hexagonal close-packed (hcp) structure with a c/a ratio of 1.624 [3], which is slightly lower than the ideal ratio of 1.633. Basal $〈a〉$ slip systems are the easy glide systems in Mg [4]. Although prism $〈a〉$ and pyramidal $〈a〉$ slip systems also operate in Mg, they require higher driving forces or elevated temperatures [5], [6], [7]. All of the slip systems mentioned above provide only a total of four independent slip systems. In particular, none of these is capable of accommodating any plastic strain along the crystallographic c direction. The activation of second pyramidal $〈c+a〉$ slip systems¹ is generally assumed to provide the additional degree of freedom needed for an arbitrary isochoric plastic strain, and has been confirmed experimentally [6], [8], [9], [10], [11], [12], [13]. Alternatively, deformation twinning can also provide the needed c-axis strain component for generalized deformation. Two common twin modes have been observed in magnesium and its alloys. These are classified as $\{10\overline{1}2\}〈\overline{1}011〉$ extension twins and $\{10\overline{1}1\}〈10\overline{1}\overline{2}〉$ contraction twins [14], because they result in the extension and the contraction of the crystal along the crystallographic c direction, respectively. The extension and contraction twins produce roughly similar shear strains of 0.1289 and 0.1377, respectively. However, they re-orient the crystal lattice differently. The extension twins re-orient the crystal lattice by 86.3° about $〈11\overline{2}0〉$ directions, while the contraction twins re-orient the crystal lattice by 56.2° about the same $〈11\overline{2}0〉$ directions. Twin thickness is expected to be influenced by the twinning shear [15]. However, in spite of the close similarity in their twinning shears, the extension twins in Mg alloys are generally thick, abundant, and are readily observed [16], [17], while the contraction twins are very thin, scarce, and hard to detect [17], [18]. It is also interesting to note that extension twins are often found within the contraction twin lamella, leading to a “double-twinning” phenomenon [12], [13], [16], [18], [19], [20], [21], [22], [23], [24], [25].

There are three main mechanisms by which deformation twinning contributes to strain hardening of the sample: (i) a Hall–Petch-like effect that arises from sub-division of grains (e.g. [26], [27], [28], [29], [30], [31]), (ii) a glissile-to-sessile transformation of dislocations already present in the region experiencing the twinning shear transformation (referred to as the Basinksi mechanism) [29], [31], [32], and (iii) changes in crystal lattice orientation to either a harder or a softer orientation (referred to as texture hardening or texture softening, respectively). It is generally thought that the high strain-hardening rates in AZ31 result from the production of extension twins (producing a Hall–Petch-like effect) and from texture hardening because the extension twins typically re-orient the grains into hard orientations [7], [33], [34], [35]. The first prevailing notion is unlikely to be accurate for extension twins in AZ31 because they grow quickly and consume the entire grain. Therefore, there is no significant grain-size reduction with the production of extension twins in AZ31. The contributions from texture hardening and the Basinski mechanism are also unlikely to be high enough to explain convincingly the unusually high strain-hardening rates measured in this alloy. This will be demonstrated later in this paper.

There have been several reports of asymmetry in tension/compression yield in AZ31 (e.g. [7]), and this has been customarily attributed to the differences in the active modes of deformation twinning in these tests. However, the current theories [36], [37] for the activation of deformation twinning rely on the availability of a critical level of dislocation density in the sample produced by initial dislocation generation and motion. Consequently, the asymmetry in the tension/compression yield values in an annealed sample should not arise at the initiation of plastic deformation in the sample. Instead, the observed differences in the reported off-set yield points are likely to have evolved between the initiation of plastic deformation and the off-set yield strain (typically 0.002). This slip before twinning hypothesis needs be to be verified so that we can better understand the physical origin of the much observed tension/compression asymmetry in AZ31.

In this paper, we have conducted several experiments and numerical simulations in an effort to develop new physical insights into the phenomena described above. We acknowledge that the basic experiments described here have already been conducted and reported in prior literature (e.g. [35]). However, our analysis of the experimental observations and conclusions are substantially different from what has been presented in prior literature. In particular, we have measured the anisotropic stress–strain response in simple compression and simple tension tests in different sample directions and documented the evolution of the strain-hardening rates. We have also carefully mapped these strain-hardening rates with the concomitant evolution of microstructure in the deformed samples using electron backscatter diffraction (EBSD) based orientation image microscopy (OIM) [38]. From a thorough analysis of all of the available experimental evidence, we have postulated new hypotheses for the physical mechanisms responsible for both the unusually high strain-hardening rates and the dramatically different morphologies (and volume fractions) of the extension and contraction twins in AZ31.