To obtain high photovoltaic performances for the emerging copper zinc tin sulfide/selenide (CZTSSe) thin film solar cells, much effort has deservedly been placed on CZTSSe phase purification and CZTSSe grain size enhancement. Another highly crucial but less explored factor for device performance is the elemental constitution of CZTSSe surface, which is at the heart of p-n junction where major photogenerated carriers generate and separate. In this work we demonstrate that, despite the well-built phase and large grained films are observed by common phases and morphology characterization (XRD, Raman and SEM), prominent device efficiency variations from short circuited to 6.4% are obtained. Insight study highlights that the surface (0–250 nm) compositions variation results in different bulk defect depths and doping densities in the depletion zone. We propose that suitable sulfurization (at ~10 kPa sulfur pressure) drives optimization of surface constitution by managing the Cu, Zn and Sn diffusion and surface reaction. Therefore, our study reveals that the balance of elemental diffusion and interface reactions is the key to tuning the surface quality CZTSSe film and thus the performance of as resulted devices.

Cross-subject electroencephalogram (EEG) based seizure subtype classification is very important in precise epilepsy diagnostics. Deep learning is a promising solution, due to its ability to automatically extract latent patterns. However, it usually requires a large amount of training data, which may not always be available in clinical practice. This paper proposes Multi-Branch Mutual-Distillation (MBMD) Transformer for cross-subject EEG-based seizure subtype classification, which can be effectively trained from small labeled data. MBMD Transformer replaces all even-numbered encoder blocks of the vanilla Vision Transformer by our designed multi-branch encoder blocks. A mutual-distillation strategy is proposed to transfer knowledge between the raw EEG data and its wavelets of different frequency bands. Experiments on two public EEG datasets demonstrated that our proposed MBMD Transformer outperformed several traditional machine learning and state-of-the-art deep learning approaches. To our knowledge, this is the first work on knowledge distillation for EEG-based seizure subtype classification.

Epilepsy is a pervasive neurological disorder affecting approximately 50 million individuals worldwide. Electroencephalogram (EEG) based seizure subtype classification plays a crucial role in epilepsy diagnosis and treatment. However, automatic seizure subtype classification faces at least two challenges: 1) class imbalance, i.e., certain seizure types are considerably less common than others; and, 2) no a priori knowledge integration, so that a large number of labeled EEG samples are needed to train a machine learning model, particularly, deep learning. This paper proposes two novel Mixture of Experts (MoE) models, Seizure-MoE and Mix-MoE, for EEG-based seizure subtype classification. Particularly, Mix-MoE adequately addresses the above two challenges: 1) It introduces a novel imbalanced sampler to address significant class imbalance; and, 2) It incorporates a priori knowledge of manual EEG features into the deep neural network to improve the classification performance. Experiments on two public datasets demonstrated that the proposed Seizure-MoE and Mix-MoE outperformed multiple existing approaches in cross-subject EEG-based seizure subtype classification. Our proposed MoE models may also be easily extended to other EEG classification problems with severe class imbalance, e.g., sleep stage classification.

The magnetic transparency of biological tissue allows the magnetic nanoparticle (MNP) to be a promising functional sensor and contrast agent. The complex susceptibility of MNPs, strongly influenced by particle concentration, excitation magnetic field and their surrounding microenvironment, provides significant implications for biomedical applications. Therefore, magnetic susceptibility imaging of high spatial resolution will give more detailed information during the process of MNP-aided diagnosis and therapy. In this study, we present a novel spatial magnetic susceptibility extraction method for MNPs under a gradient magnetic field, a low-frequency drive magnetic field, and a weak strength high-frequency magnetic field. Based on this novel method, a magnetic particle susceptibility imaging (MPSI) of millimeter-level spatial resolution (<3 mm) was achieved using our homemade imaging system. Corroborated by the experimental results, the MPSI shows real-time (1 s per frame acquisition) and quantitative abilities, and isotropic high resolution.

Magnetic iron oxide nanoparticles are relatively advanced nanomaterials, and are widely used in biology, physics and medicine, especially as contrast agents for magnetic resonance imaging. Characterization of the properties of magnetic nanoparticles plays an important role in the application of magnetic particles. As a contrast agent, the relaxation rate directly affects image enhancement. We characterized a series of monodispersed magnetic nanoparticles using different methods and measured their relaxation rates using a 0.47 T low-field Nuclear Magnetic Resonance instrument. Generally speaking, the properties of magnetic nanoparticles are closely related to their particle sizes; however, neither longitudinal relaxation rate $r_1$ nor transverse relaxation rate $r_2$ changes monotonously with the particle size $d$ . Therefore, size can affect the magnetism of magnetic nanoparticles, but it is not the only factor. Then, we defined the relaxation rates $r_i^{\prime }$ (i = 1 or 2) using the induced magnetization of magnetic nanoparticles, and found that the correlation relationship between $r_1^{\prime }$ relaxation rate and $r_1$ relaxation rate is slightly worse, with a correlation coefficient of $R^2$ = 0.8939, while the correlation relationship between $r_2^{\prime }$ relaxation rate and $r_2$ relaxation rate is very obvious, with a correlation coefficient of $R^2$ = 0.9983. The main reason is that $r_2$ relaxation rate is related to the magnetic field inhomogeneity, produced by magnetic nanoparticles; however $r_1$ relaxation rate is mainly a result of the direct interaction of hydrogen nucleus in water molecules and the metal ions in magnetic nanoparticles to shorten the $T_1$ relaxation time, so it is not directly related to magnetic field inhomogeneity.

The effect of the nonlinear time-drift gain in the power amplifier on the instability in a magnetic nanothermometer was studied to allow a very stable magnetic nanotermometer to be designed, which was then used to measure the temperature of a light-emitting diode pn-junction. Experiments showed that the system response fluctuated by less than 0.04% and the temperature measurement error was less than 0.1 K when real-time measurements were made at 1 s intervals. The temperature of a pn-junction of a light-emitting diode cooled using a heat sink or circulating water was measured successfully using the improved magnetic nanothermometer.