查看更多>>摘要：The modified Becke-Johnson exchange potential combined with local-density approximation correlation (mBJLDA) has recently attracted interest because it provides highly improved band gaps at a very low computational cost. In this work we performed an extensive investigation of the performance of the mBJLDA potential employing a norm-conserving pseudopotential plane-waves approach (mBJLDA@PP), as implemented in the AbiNit code, using a test set of 83 solids representing a wide range of semiconductors and insulators. Our results confirm the conclusion of our previous study that the number of electrons treated as valence in the pseudopotentials of the cations can have a significant impact on the calculated mBJLDA@PP band gaps. More specifically, while the use of typical pseudopotentials leads to accurate band gaps of certain systems, it yields significantly underestimated band gaps for other systems compared to experiment and to those of the all-electron mBJLDA (mBJLDA@AE) approach. The classes of the latter systems are identified, and this problem is resolved by including some outer core states as valence. The resulting mean absolute error in the calculated band gaps (compared to experiment) is of 0.46 eV, which is comparable to that of the mBJLDA@AE band gaps, reflecting the accuracy and reliability of the mBJLDA@PP approach for the band gap calculations.
查看更多>>摘要：We study and compare four coarse-grained models of cis-1,4-polyisoprene distinguished by mapping schemes for locating superatoms. First, coarse-grained potentials are obtained by iterative Boltzmann inversion method for the polymer melt. For all the coarse-grained models considered, time-scale factors based on translational and rotational motion were found to be different. However, coarse-grained potentials were unable to reproduce stress-strain behaviour of the underlying detailed model due to weak attractive nature of the nonbonded part of the coarse-grained potential. Consequently, the nonbonded potentials were optimized using particle swarm optimization to match the tensile behaviour of a detailed model of the polymer below glass transition temperature. While the modified potentials seemed to better predict the mechanical behaviour, the ability to accurately predict simultaneously the structural distributions also, depends on the mapping scheme.
查看更多>>摘要：Classical molecular dynamics (MD) has been widely used to study atomistic mechanisms and emergent behavior in materials at length and time scales beyond the capabilities of first-principles approaches. The success of classical MD simulations relies on the ability of classical interatomic potentials to accurately map complex many-body interacting systems of electrons and nuclei into effective few-body interacting systems of atoms. In practice, the development of interatomic potentials is a nontrivial process and requires considerable amount of effort. Recently, machine learning has become a promising approach to accelerate interatomic potential development. However, these machine learning approaches are often computation and data intense, as they require a large amount of training data from first-principles calculations, such as total energies, atomic forces, and stress tensors of many atomistic structures. Here we propose an active learning approach combined with first-principles theory calculations to expedite the development of machine learning interatomic potentials. In particular, we develop a batch active learning method which combines both energy uncertainty and structure similarity metrics to efficiently sample the highly uncertain structures that are difficult to predict. This active sampling approach maximizes the utility of the dataset in each batch and generates interatomic potential with highly accurate and robust model coefficients which are difficult to achieve with conventional sampling approaches. To demonstrate this batch active learning method, we develop an active learning potential for monolayer GeSe, a two-dimensional ferroelectric-ferroelastic material, and compare the quality and robustness of the active learning potential with the potential obtained from random sampling. Batch active learning method opens up avenues for accelerating the development of robust and accurate machine learning potential using a small set of atomistic structures which will be valuable for computational materials, physics, and chemistry community.
查看更多>>摘要：Structural materials properties are highly dependent on their microstructure. Their microstructure is in turn affected by multiple fabrication and thermo-mechanical treatment parameters, all of which conform a highly-dimensional parametric space with often hidden correlations that are difficult to extract by experimentation alone. This is particularly true for alloys of the dual-phase Ti-6Al-4V family, with their greatly complex and rich microstructures, which combine several intrinsic lengthscales associated with multiple grain and subgrain structures, grains with different crystal lattices (a and 6 phases), and complex chemistry. Here we use a comprehensive set of machine learning techniques to develop predictive tools relating the yield strength and hardening rate of these alloys to a set of input parameters covering extensive ranges. The data generator is a finite-element crystal plasticity model for polycrystal deformation that takes into account slip anisotropy and employs standard dislocation evolution models for the a and 6 phases of Ti-based alloys. Our dataset includes over two thousand independent simulations and is used to train the machine learning models, which are then used to establish correlations between microstructural parameters and the alloys' mechanical response. Our results point to the most influential parameters affecting yield strength and hardening rate, information that can then be used to guide experimental synthesis and characterization efforts to save time and resources.
查看更多>>摘要：This paper presents an implementation of the elasto-visco-plastic fast Fourier transform (EVPFFT) crystal plasticity model in the implicit finite element (FE) method of Abaqus standard through a user material (UMAT) subroutine to provide a constitutive relationship between stress and strain at FE integration points. To facilitate the implicit coupling ensuring fast convergence rates, an analytical Jacobian is derived. The constitutive response at every integration point is obtained by the full-field homogenization over an explicit microstructural cell. The implementation is a parallel computing approach involving multi-core central processing units (CPUs) and graphics processing units (GPUs) for computationally efficient simulations of large plastic deformation of metallic components with arbitrary geometry and loading boundary conditions. To this end, the EVPFFT solver takes advantages of GPU acceleration utilizing Nvidia's high performance computing software development kit (SDK) compiler and compute unified device architecture (CUDA) FFT libraries, while the FE solver leverages the message passing interface (MPI) for parallelism across CPUs. The high-performance hybrid CPU-GPU multi-level framework is referred to as FE-GPU-EVPCUFFT. Simulations of simple compression of Cu and large strain cyclic reversals of dual phase (DP) 590 have been used to benchmark the accuracy of the implementation in predicting the mechanical response and texture evolution. Subsequently, two applications are presented to illustrate the potential and utility of the multi-level simulation strategy: 4-point bending of textured Zr bars, in which the model captures the shape variations as a consequence of texture with respect to the bending plane and another bending of DP1180, in which the model reveals details of spatial micromechanical fields.
查看更多>>摘要：ABSTR A C T Nanomaterials or structures usually exhibit characteristic performance under complex stress states. The tensile and compressive behaviors of [001]-oriented single-crystalline nanopillars were studied, by performing mo-lecular dynamic simulations on several typical FCC metals. For all those metallic nanopillars, their yield strengths for nucleating the initial dislocation show strong loading direction dependence, i.e., the strength under tension is higher than that under compression, showing the typical T/C asymmetry. The origins of the T/C asymmetry were investigated from the new aspects of the surface energy difference under tension and compression, the large ultimate elastic deformation, and the non-Schmid stress, in detail. The results indicate that both the Schmid factor and the non-Schmid factors change considerably due to the large elastic deformation under tension or compression, which contribute negatively to the T/C asymmetry. The difference in surface energy reduction due to the large elastic deformation is one but not the only factor that results in the T/C asymmetry. Although the non-Schmid factor contributes negatively to the T/C asymmetry, the non-Schmid stress can increase the dif-ference of unstable stacking fault energies under tension and compression, which has a significant positive in-fluence on the T/C asymmetry by changing the ideal shear strength of the slip plane.
NSTL
Elsevier