In recent years, facial expression recognition (FER) has garnered significant attention within the realm of computer vision research. This paper presents an innovative network called the Dual-Direction Attention Mixed Feature Network (DDAMFN) specifically designed for FER, boasting both robustness and lightweight characteristics. The network architecture comprises two primary components: the Mixed Feature Network (MFN) serving as the backbone, and the Dual-Direction Attention Network (DDAN) functioning as the head. To enhance the network’s capability in the MFN, resilient features are extracted by utilizing mixed-size kernels. Additionally, a new Dual-Direction Attention (DDA) head that generates attention maps in two orientations is proposed, enabling the model to capture long-range dependencies effectively. To further improve the accuracy, a novel attention loss mechanism for the DDAN is introduced with different heads focusing on distinct areas of the input. Experimental evaluations on several widely used public datasets, including AffectNet, RAF-DB, and FERPlus, demonstrate the superiority of the DDAMFN compared to other existing models, which establishes that the DDAMFN as the state-of-the-art model in the field of FER.

A full-zone 30-band k · p model is developed as an efficient and reliable tool to compute electronic band structure in Ge_1−x Sn _x alloy. The model was first used to reproduce the electronic band structures in Ge and α-Sn obtained with empirical tight binding and ab initio methods. Input parameters for the 30-band k · p model are carefully calibrated against prior empirical predications and experimental data. Important material properties such as effective mass for electrons and holes, Luttinger parameters, and density of states are obtained for Ge_1−x Sn _x alloy with the composition range 0 < x < 0.3. The 30-band k · p model that requires far less computing resources is a necessary capability for optimization of sophisticated devices made from Ge_1−x Sn _x alloy with a large parameter space to explore.

Enhancing valley splitting in SiGe heterostructures is crucial for developing silicon spin qubits. Complex SiGe heterostructures, sharing a common feature of four-monolayer (4 ML) Ge layer next immediately to the silicon quantum well (QW), have been computationally designed in 2013 to have giant valley splitting approaching 9 meV and hence could be used to overcome the challenge of valley splitting towards the experimental realization of Si quantum computing. However, none of them has been successfully fabricated, perhaps due to their complexity. Here, we remarkably simplify the originally designed complex SiGe heterostructures by laying out the Si QW directly on the SiGe substrate followed by capping a ${\left({\mathrm{Ge}}_4{\mathrm{Si}}_4\right)}_n$ superlattice (SL) barrier with a sacrifice in valley splitting (VS), which is reduced from a maximum value of 8.7 to 5.2 meV. Even the smallest number of SL periods ($n=1$) will also give a sizable VS of 1.6 meV, which is large enough for developing stable spin qubits. We further develop an effective Hamiltonian model to reveal the physical mechanism underlying the enhanced valley splitting by the ${\left({\mathrm{Ge}}_4{\mathrm{Si}}_4\right)}_n$ SL barriers. We surprisingly find that the presence of the SL barrier will reduce rather than enhance the VS in most cases. The only exception is the ${\left({\mathrm{Ge}}_4{\mathrm{Si}}_4\right)}_n$ SL barriers, where their miniband states have such a strong coupling with Si QW valley states that they provide an even larger VS. These findings lay a solid theoretical foundation for overcoming the valley splitting issue of SiGe heterostructures in the experiment that is heading toward Si quantum computing.