Cube oriented grains in face centered cubic materials are metastable under plane strain. Once loading begins, the lattice rotates, and the axis of that rotation, the normal direction ND, the rolling direction RD, or the transverse direction TD, has not been settled by experiment alone. Measurements on single crystals and on cube oriented specimens since the 1930s show that the micro texture that builds up inside a single cube grain depends sensitively on how the specimen is loaded and on how much initial misorientation the crystal carries before deformation. Early models attributed the splitting to rotation about ND and RD, while later analyses favored rotation about TD. A consistent picture requires a model that responds sharply to two small inputs: a slight initial orientation spread and a change in friction at the contact surfaces.
The study reported here provides such a picture. Raabe, Zhao, and Roters simulate the plane strain compression of a cube single crystal with a crystal plasticity finite element method in which the starting texture is not a single orientation but a quantitative distribution in orientation space. All simulations run to a 50 percent engineering thickness reduction, the regime of a typical channel die experiment.
The sample is a block of 500 three dimensional linear elements, each with eight integration points, bounded by four rigid surfaces. Linear Coulomb friction is applied separately to the top and bottom contact surfaces and, in a separate series, to all four side surfaces. The constitutive equations are integrated implicitly with the viscoplastic formulation of Kalidindi, Bronkhorst, and Anand, using {111}<110> slip systems and rate dependent hardening. The specific part of the model is the way the starting texture is fed into the mesh.
The cube orientation is described as one spherical Gaussian component in orientation space, defined by its height and its full width at half maximum. In a first step the ideal cube orientation, Euler triple 0°, 0°, 0° in Bunge notation, is assigned as a single rotation matrix to every integration point. In a second step the individual orientation portions are re oriented so that their ensemble reproduces the prescribed Gaussian, here with a full width at half maximum of 2.5°, distributed in random lateral order across the 500 elements. After this mapping the component concept is dropped: every portion evolves independently during the deformation. The procedure converts a distribution that originally exists only in orientation space into a distribution that exists simultaneously in real space and orientation space, which is the situation of an as grown single crystal that already contains small misorientations.
Three series of calculations differ only in the initial texture and in the friction conditions.
Ideal cube, no initial scatter. With an exactly cube oriented, perfectly homogeneous crystal, friction coefficient μ = 0.1 on the two contact surfaces and zero friction on the side surfaces, the elements keep a regular brick like shape after 50 percent thickness reduction and the von Mises stress is nearly homogeneous. The {111} pole figure shows orientational splitting about the normal direction only, and the split is symmetric. In this limit the cube orientation is stable apart from the well known normal direction component.
Cube with 2.5 degree Gaussian scatter. Introducing the 2.5° orientation scatter under the same friction conditions changes the result. The pole figure now shows splitting about both the normal direction and the rolling direction. A section of the orientation distribution function at f(g) = 10 shows that the spread about the rolling direction, approximately 22°, is clearly larger than the spread about the normal direction, approximately 14°. Most elements exhibit slight shear distortion, indicating that the laterally random scatter locally violates the plane strain state.
Higher friction. Raising the friction coefficient to μ = 0.3 adds a third rotation mode. Orientation scatter about the transverse direction appears, while the spread about the normal direction becomes smaller. The maximum of the orientation density at the ideal cube component is strongly reduced, so that the micro texture is smeared out, and both element distortion and stress inhomogeneity increase. Applying friction on all four side surfaces at μ = 0.1 produces a strong transverse direction component together with weaker normal and rolling direction components; the rolling direction scatter in this case is smaller than in the two surface case.
The results admit two complementary readings. The first is kinematical. In homogenization and crystal plasticity analyses that treat the lattice reorientation rate as a vector field in orientation space, the divergence of that field measures whether an orientation is a source or a sink of reorientation. Orientations with positive divergence amplify arbitrarily small initial misorientations, so that orientation gradients develop even under gradient free external loads. Cube in face centered cubic metals belongs to this weakly divergent class, which is why it responds so strongly to small changes in starting texture. The finite element results confirm this directly: zero scatter gives pure normal direction rotation, while a scatter of only 2.5° switches on a strong rolling direction component that exceeds the normal direction spread.
The second reading concerns local symmetry. The lateral distribution of the initial misorientations breaks the symmetry of the plane strain state inside the grain, and the resulting local shear drives the rolling direction rotation. Friction acts in the same spirit but by a different route: it introduces longitudinal shear strains at the contact surfaces, and these preferentially promote transverse direction rotation. Both effects, initial texture and friction, destroy the ideal plane strain state, but they load different rotation modes. The cube orientation is therefore best understood as a weakly unstable state whose final micro texture is set by small, experimentally controllable imperfections rather than by the loading path alone.