Search

Since 2020, synchrotron radiation facilities in several Asia-Pacific countries have been collaborating in a major project called "SYNAPSE"(Synchrotrons for Neuroscience: an Asia-Pacific Scientific Enterprise). They use x-ray imaging to attack in a coordinated fashion one of the major issues in modern science: the structure of animal and human brains, including neurons and connections. The objective is to develop Google-like maps also including detailed structural and functional information for selected regions of interest. The sheer mass of data needed for the objective poses huge problems for the acquisition, processing, storage and use of images. In order to complete the task within a reasonable time, the key element of the SYNAPSE strategy is the parallel and coordinated work of several facilities on the same specimens. This article reviews different aspects of the enterprise, including the foundations of synchrotron radiation, coherence and of its role in advanced imaging, electron accelerators, x-ray optics and detectors. This will provide the foundation for an extensive presentation of the different components of SYNAPSE, with an overview of results already obtained within the consortium.

In this work, the methodology for the reserve design in GaN power MIS-HEMTs based on hand-drawn characteristics is demonstrated for the first-time using Graph Attention Network. The hand-drawn ID-VG characteristic is constructed by Ramer-Douglas-Peucker algorithm. Then, the extracted information is sent to the Graph Attention Network (GAT) to receive the corresponding device design variables, including tAlGaN, recessed depth, Al%, Lg, Lgd, and Lgs. Less than 8% of the differences in the key extracted parameters, such as Vth, Ion, and subthreshold slope (SS), can be obtained and less than 30 seconds is consumed to generate the design variables. Therefore, the developed GAT approach is promising for the reverse design of GaN power MIS-HEMTs, which can provide users with efficient and valuable design suggestions to optimize the devices toward the targeting performance.

Engineering simulation accelerates the development of reliable and repeatable design processes in various domains. However, the computing resource consumption is dramatically raised in the whole development processes. Making the most of these simulation data becomes more and more important in modern industrial product design. In the present study, we proposed a workflow comprised of a series of machine learning algorithms (mainly deep neuron networks) to be an alternative to the numerical simulation. We have applied the workflow to the field of dental implant design process. The process is based on a complex, time-dependent, multi-physical biomechanical theory, known as mechano-regulatory method. It has been used to evaluate the performance of dental implants and to assess the tissue recovery after the oral surgery procedures. We provided a deep learning network (DLN) with calibrated simulation data that came from different simulation conditions with experimental verification. The DLN achieves nearly exact result of simulated bone healing history around implants. The correlation of the predicted essential physical properties of surrounding bones (e.g. strain and fluid velocity) and performance indexes of implants (e.g. bone area and bone-implant contact) were greater than 0.980 and 0.947, respectively. The testing AUC values for the classification of each tissue phenotype were ranging from 0.90 to 0.99. The DLN reduced hours of simulation time to seconds. Moreover, our DLN is explainable via Deep Taylor decomposition, suggesting that the transverse fluid velocity, upper and lower parts of dental implants are the keys that influence bone healing and the distribution of tissue phenotypes the most. Many examples of commercial dental implants with designs which follow these design strategies can be found. This work demonstrates that DLN with proper network design is capable to replace complex, time-dependent, multi-physical models/theories, as well as to reveal the underlying features without prior professional knowledge.

This work investigates the correlation between dipole formation and surface energy at a nonplanar dielectric interface on a Si nanowire (NW) by means of molecular dynamics simulations. According to the atom and charge distribution obtained by simulations, the built-in electric field and potential are calculated at the dielectric interface based on the Poisson equation. It is found that the built-in potential is dependent on the diameter of the Si NW, which is attributed to the dependence of the surface energy difference on the surface curvature of the dielectric heterointerface, driving atom diffusion and leading to a dipole layer at the interface. The impact of the built-in potential on the threshold voltage of a transistor with a multigate configuration is discussed as well.

According to quantum transport simulations, source-to-drain tunneling (SDT) has been recognized as the main cause leading to subthreshold swing ( SS ) saturation and degradation of short-channel MOSFETs at cryogenic temperatures. Generally, at a given low temperature, the steeper constant SS of thermionic currents may be overwhelmed by less-steep SDT currents at lower gate bias, degrading the average SS . Our simulations show that the SS of SDT currents is insensitive to temperatures for a MOSFET with a given channel length, accounting for SS saturation as lowering temperature. This work reveals the key points differentiating the possible reasons (interface traps, band tail and SDT) of SS saturation at cryogenic temperatures. We also study the impacts of carrier effective masses, gate-SD underlapping and a tunneling barrier at the source junction on SS saturation. SDT may pose a potential challenge and limit for scaling cryogenic CMOS of a quantum processor.

Mapping the large neural networks of animal and human brains is a fundamental but so far elusive task, because of the massive amount of data and the consequent prohibitively long image taking and processing times. We developed an effective strategy called “AXON” (Accelerated X-ray Observation of Neurons) to solve this problem. AXON can achieve comprehensive whole-brain mapping within a reasonable time by combining fast image taking and processing, plus two other critical performances: three-dimensional (3D) imaging with high and isotropic spatial resolution, and multi-scale resolution. We successfully tested this strategy with coordinated experiments at four synchrotron facilities in Japan, Taiwan, Singapore and Korea on two animal models, Drosophila and mouse. Its performances notably allowed full 3D mapping of the Drosophila brain in a few days. With reasonable improvements, AXON can deliver full mapping of large animal and human brains on a realistic time scale of a few years.

The dielectric screening property of a semiconductor is very crucial for the electrical characteristics of a MOSFET, and which can be described mathematically by Poisson equation via the permittivity. While the theory and experiments have corroborated the permittivity reduction of nanoscale Si, this paper studies the electrical characteristics of MOSFETs considering the reduced channel permittivity by quantum transport simulations. It is found that the channel permittivity reduction may mitigate the short-channel effects, showing subthreshold swing improvement and threshold voltage shift of MOSFETs in nanoscale. Compared to quantization effects, the positive and negative impacts of the channel permittivity reduction on the devices in particularly nanoscale have been investigated. This paper elucidates the necessity of considering semiconductor permittivity reduction for nanoscale device design and simulations.

Detailed microstructure evolution in shape memory alloys (SMAs) is typically studied by molecular dynamics (MD) simulations. However, the conventional post-processing tools for atomistic calculations, such as CNA and PTM, fail to identify distinct crystal variants and to reveal twin alignments in SMAs. In the current work, a powerful and efficient post-processing tool based on GraphSAGE neural network is developed, which can identify multiple phases in martensitic transformation, including the orthorhombic, monoclinic and R phases. Where the network was trained by the results of sets of temperatureand stress-induced martensitic transformation MD calculations. The accuracy and generality were also verified by the application to the cases which did not appear in the training dataset, such as unrecoverable nanoindentation process. The proposed method is rapid, accurate, and is ready to be integrated with any visualization tool for MD simulations. The outcome of the current work is expected to accelerate the pace of atomistic studies on SMAs and related materials.

This paper reports on the demonstration of the characteristics modeling and prediction in GaN-on-Si power devices (MIS-HEMTs and p-GaN HEMTs) using the artificial neural network (ANN)-based approach. A multi-layer ANN is developed to model the electrical characteristics, e.g., V$_{TH}, {I}_{D} V_{G}$, hysteresis, breakdown characteristics, and time-dependent dielectric breakdown (TDDB), etc. Furthermore, an autoencoder with two ANNs is also developed to reconstruct the device designs based on the specific characteristics. We show that the ANN-based approach is promising for modeling and prediction with multidimensional parameters, further assisting in the optimization for GaN-based devices towards the targeted performance.

In this article we explore Ni/HfO 2 /TiN resistive random access memory (RRAM) as ultralow power synaptic element in deep neural networks (DNN) for artificial intelligence applications. Low power RRAM devices are fabricated and measured, with very low RESET current and 2–3 orders of resistance window. The SET-RESET current-voltage characteristics, high- and low-resistance state statistical distribution, and analog programming characteristics are calibrated to analytical models. Training and inference for the MNIST handwritten digits dataset using a multilayer perceptron was simulated based on the calibrated model using CIMulator, a novel neuromorphic simulation platform for compute-in-memory circuitry to predict DNN inference accuracy and energy consumption. Despite larger inherent device-to-device variability due to low on current, 97% inference accuracy is achieved with only 2-bit for the weights using weight update accumulation technique.

Rechargeable batteries, such as LiFePO4/graphite cells, age differently by variability in manufacturing, charging (energy inflow) policy, temperature, discharging conditions, etc. Great economic and environmental value can be extracted if we can predict how a battery ages and ascertain its current state of health and residual useful life, based on just a few cycles of testing. Here, by developing novel-architecture deep neural networks with a special convolutional training strategy and taking advantage of recently published battery cycling data, we show that one can predict the residual life of a battery to a mean absolute percentage error of 6.46%, using only one cycle of testing. The cycle-by-cycle profiles, such as discharge voltage, capacity, and power curves of any given cycle, of used batteries with unknown age can also be accurately predicted for the first time. Moreover, our models can extract data-driven features from the data which were much more influential on the predicted properties than human-picked features. This work has shown that single cycle data contains a sufficient amount of information to predict essential battery properties with high accuracy. It is expected to provide tremendous economic and environmental benefits since reuse and recycling of batteries can be better planned and less lithium-ion batteries end up in landfills.

Effective 3D visualization is essential for connectomics analysis, where the number of neural images easily reaches over tens of thousands. A formidable challenge is to simultaneously visualize a large number of distinguishable single-neuron images, with reasonable processing time and memory for file management and 3D rendering. In the present study, we proposed an algorithm named “Kaleido” that can visualize up to at least ten thousand single neurons from the Drosophila brain using only a fraction of the memory traditionally required, without increasing computing time. Adding more brain neurons increases memory only nominally. Importantly, Kaleido maximizes color contrast between neighboring neurons so that individual neurons can be easily distinguished. Colors can also be assigned to neurons based on biological relevance, such as gene expression, neurotransmitters, and/or development history. For cross-lab examination, the identity of every neuron is retrievable from the displayed image. To demonstrate the effectiveness and tractability of the method, we applied Kaleido to visualize the 10,000 Drosophila brain neurons obtained from the FlyCircuit database ( http://www.flycircuit.tw/modules.php?name=kaleido ). Thus, Kaleido visualization requires only sensible computer memory for manual examination of big connectomics data.

A device design technique using tunneling barriers (TBs) for reducing the short-channel effects (SCEs) is proposed. By introducing TBs at the source and drain junctions of a Si FET, the threshold voltage ( ${V}_{\text {th}}$ ) roll-off can be significantly suppressed. This is because the TBs weaken the electrical coupling between drain bias and transmission/current spectrum in energy. Specifically, as compared with a conventional FET, the ${V}_{\text {th}}$ roll-off for channel length reduction from 20 to 5 nm is mitigated by more than 40% when a thin TB is embedded at the source junction. This paper further reveals that the TB at the source junction dominates the physical mechanism minimizing the SCEs of the TBFET, and thus the device performance can be improved appreciably by removing the TB at the drain side and by decreasing the TB height at the source side.

High resolution synchrotron microtomography capable of revealing microvessels in three dimensional (3D) establishes distinct imaging markers of mouse kidney disease strongly associated to renal tubulointerstitial (TI) lesions and glomerulopathy. Two complementary mouse models of chronic kidney disease (CKD), unilateral ureteral obstruction (UUO) and focal segmental glomerulosclerosis (FSGS), were used and five candidates of unique 3D imaging markers were identified. Our characterization to differentially reflect the altered microvasculature of renal TI lesions and/or glomerulopathy demonstrated these image features can be used to differentiate the disease status and the possible cause therefore qualified as image markers. These 3D imaging markers were further correlated with the histopathology and renal microvessel-based molecular study using antibodies against vascular endothelial cells (CD31), the connective tissue growth factor or the vascular endothelial growth factor. We also found that these 3D imaging markers individually characterize the development of renal TI lesions or glomerulopathy, quantitative and integrated use of all of them provide more information for differentiating the two renal conditions. Our findings thus establish a practical strategy to characterize the CKD-associated renal injuries by the microangiography-based 3D imaging and highlight the impact of dysfunctional microvasculature as a whole on the pathogenesis of the renal lesions.

article i nfo Complete profiling would substantially facilitate the fundamental understanding of tumor angiogenesis and of possible anti-angiogenesis cancer treatments. We developed an integrated synchrotron-based methodology withexcellent performances: detection of very small vessels by high spatialresolution (~1 μm)and nanoparticle contrast enhancement, in vivo dynamics investigations with high temporal resolution (~1 ms), and three- dimensional quantitative morphology parametrization by computer tracing. The smallest (3-10 μm) microves- sels were found to constitute >80% of the tumor vasculature and exhibit many structural anomalies. Practical applications are presented, including vessel microanalysis in xenografted tumors, monitoring the effects of anti-angiogenetic agents and in vivo detection of tumor vascular rheological properties.

Dynamics of the formation of self-assembled sub-monolayer networks of colloidal nanoparticles is studied in the two-dimensional lattice gas model by Monte Carlo simulation. The site-percolation threshold ( p c ) is studied for various temperatures and chemical potentials. Our numerical results show that p c is greatly reduced due to the evaporation-driving self-organization. The results are qualitatively consistent with our experimental observation.

Quantitative analysis of neurons is a very important issue in neural science especially after numerous three-dimensional neural images in Drosophila brains were taken from confocal laser scanning microscope [1]. However, analyzing these messy data is mostly by human being with some semi-automatic software packages so far. Not only the task is very labor intensive but also the result is susceptible to errors and usually lacks objectivity. Therefore, fast and accurate analyzing tools are crucial and very desirable. Recently, we developed a computational algorithm, FAST (Fast Automatically Structural Tracing algorithm), which can trace neurons and get characteristic quantities of neuron fibers from their morphology in a very efficient way. These characteristic quantities (called SIs, Structural Indexes) are, for example, number of branch points, number of end points, cross section area of fibers, branch angle of fibers, distribution of fiber length, curvature of fibers, and innervation in neuropils, etc. After structural indexes of neuron fibers were obtained, isomap [2] and modularity [3] methods are applied to classify neurons without depending on human intervention. The isomap method can defined the similarity between neurons by geodesic paths in a high-dimensional manifold as well as the modularity method can find the best community structure of classification by optimization, i.e., to maximize the intra module connections as many as possible and to minimize the inter module connections as few as possible. With these tools, large-scale neural morphological structures, their annotations as well as quantified characteristics, and neural classifications can be facilely and reliably retrieved as useful data for computational neuroscience. Figure 1 A schematic diagram for innervation table and classification results of local neurons in olfactory system of Drosophila.

We explore the possibility for the native state of a protein being inherently a multi-conformation state in an ab initio coarse-grained model. Based on the Wang-Landau algorithm, the complete free energy landscape for the designed sequence 2D4X: INYWLAHAKAGYIVHWTA is constructed. It is shown that 2DX4 possesses two nearly degenerate native states: one has a helix structure, while the other has a hairpin structure and their energy difference is less than 2% of that of local minimums. Two degenerate native states are stabilized by an energy barrier of the order 10kcal/mol. Furthermore, the hydrogen-bond and dipole-dipole interactions are found to be two major competing interactions in transforming one conformation into the other. Our results indicate that degenerate native states are stabilized by subtle balance between different interactions in proteins; furthermore, for small proteins, degeneracy only happens for proteins of sizes being around 18 amino acids or 40 amino acids. These results provide important clues to the study of native structures of proteins., 15 pages, 7 figures, submitted to Biophysical Journal

Refractive-index (phase-contrast) radiology was able to detect lung tumors less than 1 mm in live mice. Significant micromorphology differences were observed in the microradiographs between normal, inflamed, and lung cancer tissues. This was made possible by the high phase contrast and by the fast image taking that reduces the motion blur. The detection of cancer and inflammation areas by phase contrast microradiology and microtomography was validated by bioluminescence and histopathological analysis. The smallest tumor detected is less than 1 mm3 with accuracy better than 1 × 10−3 mm3. This level of performance is currently suitable for animal studies, while further developments are required for clinical application. Refereed/Peer-reviewed

A coarse-grained off-lattice model that is not biased in any way to the native state is proposed to fold proteins. To predict the native structure in a reasonable time, the model has included the essential effects of water in an effective potential. Two new ingredients, the dipole-dipole interaction and the local hydrophobic interaction, are introduced and are shown to be as crucial as the hydrogen bonding. The model allows successful folding of the wild-type sequence of protein G and may have provided important hints to the study of protein folding., 4 pages, 4 figures, to appear in Physical Review Letters

Recently, numerous three-dimensional neural images in Drosophila brains were taken from confocal laser scanning microscope [1]. However, how to obtain useful neuronal information from these messy raw data is very challenging. In order to conquer this problem, two critical issues need to be addressed: the first is to segment the image of single neuron from image data; the second is to trace the neuron fibers for quantitative analysis. Therefore, a robust segmentation process and an efficient tracing algorithm for single neuron are crucial and very desirable. At present there are methods and commercial software packages for these functions. But it requires a viewer to use his/her vision and judgment to segment and trace the neurons. Not only the task is very labor intensive but also the result is susceptible to errors and is usually lack of objectivity. Here we proposed an automatic procedure to segment and trace neural image data on a large scale. We first developed a new algorithm, Fast Automatically Structural Tracing Algorithm (FASTA), which encodes all image voxels on the idea of source field method and traces whole neuron via these codelets with some stopping criteria. Then, all image data were segmented into several single-neuron images with reasonable intensity threshold by using of FASTA iteratively. With this automatic procedure, single-neuron images and their annotations as well as quantified characteristics can be facilely and reliably retrieved as useful data for computational neuroscience. Figure 1