Naphthalenediimide (NDI) derivatives without or with one or two vinyl groups were synthesized to prepare thin films by the vapor deposition. NDI with larger number of vinyl substitution tends to form thin films with poorer crystallinity and lower electrical conductivity. On the other hand, the vinyl modification is effective in improving film morphology and temporal stability. Electron-assisted deposition of the vinyl-substituted NDIs produces smooth and stable amorphous polymer films with a lower conductivity. The vinyl substitution brings about contradictory effects of improving film morphology and stability at a cost of reducing crystallinity and conductivity. The vinyl modification also enables controlling the film/substrate interface by covalent tethering via a self-assembled monolayer.

The I-V measurement results of poly(vinyl alcohol)/ITO triboelectric generators are reported and discussed with a focus on the orientational ordering of molecular dipoles as electrical power sources induced by mechanical rubbing. The results showed that the rubbing induces dc current flowing through the ammeter in the direction from the poly(vinyl alcohol) to the cotton rubbing cloth. The polarity of induced current is opposite to what is expected from triboelectric series table, where poly(vinyl alcohol) is listed in the negative side relative to cotton. Assuming that the OH molecular groups are orientationally ordered in the direction away from the rubbing surface, and the current is induced through depolarization from the ordered state, the current direction is well explained. The results suggest that the molecular orientational ordering induced by rubbing is key for the operation of triboelectric generators using polar materials.

Organic multifunction diodes with light-emitting, photovoltaic, and photo sensing functions were demonstrated using a stack of N, N’-ditridecylperylene-3,4,9,10-tetracarboxylic diimide and 2,7-di(9H-fluoren-2-yl)benzo[lmn][3,8]-phenanthroline-1,3,6,8(2H,7H)-tetraone as an electron transporter/acceptor and rubrene as an emitter/donor. Polyethylenimine ethoxylated as an electron injection layer was introduced in an inverted-structured device. By using the inverted structure and introducing HFl-NDI to reduce the absorption in the acceptor layer, an increase in the emission and an increase in the short-circuit current density in the organic solar cell were obtained.

S-shaped current (I)-voltage (V) characteristics are routinely observed in emerging photovoltaic cell research. The opposed two-diode equivalent circuit model can reproduce such characteristics. The present study demonstrates Bayesian estimation of equivalent circuit parameters of a photovoltaic cell from S-shaped I-V characteristics with affordable computational cost below 15 min. The demonstration codes have been made publicly available on GitHub to cultivate transparency and facilitate reproducibility. This initiative aims not only to advance the understanding of photovoltaic cell behaviors but also to provide a practical, accessible tool for researchers in the field.

We developed the stable, ultra-fast vertical polyimide capacitive-type humidity sensor with a response time of 10 ms order using fluorinated polyimide with a multi-walled carbon nanotube (CNT) gas permeable upper electrode. On the basis of the diffusion model, the response time is reduced to less than 0.1s by reducing the hydrophobic partially fluorinated polyimide dielectric layer to ca. 100 nm. Although the sensitivity increases with the reduction of polyimide (inversely proportional) thickness, the recovery time increases and non-linear sensitivity are observed after the exposure to a high-humid air (> 80 %RH). The slow recovery and instability were improved by the insertion of ultra-thin perfluoro polymer (Cytop).

We focus on modifying Au surface with UiO-66-NH₂ for heavy metal detection via surface plasmon resonance (SPR) technique. Two conditions were tested for the deposition of UiO-66-NH₂, i.e., spin-coating immediately after dropping the UiO-66-NH₂ dispersion solution on the surface (0-minute) and spin-coating after incubating for 15 min. (15-minute). The 0-minute condition resulted in a more uniform UiO-66-NH₂ distribution, thus selected for further investigation of heavy metal ion adsorption. The SPR signal showed a response to As(V) ion adsorption on UiO-66-NH₂ layer in aqueous solution, demonstrating a novel and effective method for As(V) detection in the 10 to 50 mg mL^-1 range. This modified sensor exhibited improved sensitivity for detecting As(V) ions.

An enzymatic biofuel cell (EBFC) that uses lactic acid as fuel to generate electricity is an attractive power source for wearable devices. Power density of an EBFC was improved by using two paired electrodes: LOD (anode)-BOD (cathode) and LDH+NAD (anode)-BOD (cathode). The purpose was to reduce the pyruvic acid produced by LOD at anode. The enzymes that generate power output were LOD at the anode and BOD at the cathode, and the additional electrode pair modified with LDH and NAD at the anode and BOD at the cathode was placed near the power output electrodes. The combination of LDH and NAD can perform both oxidation and reduction. Therefore, the pyruvic acid produced at the anode can be reduced. The output of the single pair electrodes (LOD-BOD) was increased from 25 μW/cm² to approximately 43 μW/cm² (1.7 times) with the additional paired electrodes.

Many studies have been conducted on the application of microneedles (MNs) for drug administration in the medical field. In medical applications, large amounts of drugs need to be administered. We investigated the method of melting and diffusing a solidified regent on the punctured MN inside human skin by ultrasonic heating. This was because when the MN was punctured into human skin, the liquid did not migrate along the needle groove into the skin. As a countermeasure, the following method was adopted: after transporting liquid to the needle tip owing to capillary force, the liquid was solidified. Then the MN was punctured pseudo human skin (gel), and the solidified liquid was melted and diffused by ultrasonic heating. Temperature characteristics of the MN was investigated changing the needle material, needle diameter, and ultrasonic wave direction. In addition, the behavior of the liquid in the gel was observed under ultrasonic heating.

With the growth of deep learning and machine learning applications, an efficient processing element array (PEA) has become increasingly important. To address this need, this paper introduces a quantized bit-serial PEA, which improves data reusability by integrating a weight ring (WR) dataflow mechanism and increases operation frequency through the use of bit-serial circuits. This design substantially reduces the number of feature map accesses, thereby optimizing data processing efficiency. A key aspect of our approach is the use of quantization techniques. By converting floating-point values to signed 8-bit fixed-point numbers, we reduce computational complexity and ease memory bandwidth pressure. We briefly discuss that ignoring bias terms may not impact model inference accuracy when the appropriate neural network type and dataset are chosen. Our proposed WR dataflow, inspired by the weight stationary (WS) dataflow, only updates the outdated row with a new row. This not only boosts data reuse rates but also diminishes costly data access operations. Notably, the 3×3 WR PEA requires 38.54% of the off-chip accesses per second as compared to the 3×3 WS PEA and merely 11.25% compared to its no local reuse (NLR) PEA counterpart. Empirical results show its excellent trade-off between area, power, and speed, ensuring robust data reuse efficiency. By combining quantization and WR dataflow, our high-reuse, quantized bit-serial PEA offers a fresh perspective on deep learning hardware design.

Wired-logic processor architecture is a promising technology for energy-efficient computing. They have achieved several orders of magnitude higher energy efficiency than conventional FPGA-based deep neural network (DNN) processors by eliminating DRAM/BRAM access. The technical challenge of the wired-logic architecture is a huge amount of hardware resources to implement all weights and processing elements as wired-logic circuits. While the non-linear neural network (NNN) was proposed which can save hardware resources by a ternary weight, highly sparse neural network, the area overhead is still large for non-linear function implementation of NNN. Here we developed a coarse- and fine-grained lookup table (LUT) segmentation technique for resource-efficient FPGA-based NNN wired-logic processors. Two techniques are designed and analyzed: (1) an LUT segmentation technique based on coarse and fine granularity, and (2) accuracy optimization through the incorporation of redundant bits. The application of these proposed techniques to state-of-the-art wired-logic processors markedly enhances the scalability achievable with a single FPGA, thereby facilitating the implementation of larger-scale neural networks across various tasks, including CIFAR-10 classification and keyword spotting. The hardware resource requirements for non-linear functions in processing elements decreased by 94.4%, and 95.4%, respectively. Notably, the recognition accuracy for both CIFAR-10 and the keyword spotting task decreased by less than 0.2%, a negligibly small degradation.

In order to achieve a compact robot vision system that can perform object identification in real environments exposed to large changes in illumination, it is essential to consider both the algorithm and the hardware architecture. In this study, we have developed a compact and low-power image recognition system that is robust to illumination changes, consisting of a CMOS image sensor, field-programmable gate array (FPGA), and graphics processing unit (GPU) for embedded devices. To minimize the effects of changes in the illumination, our system uses the center/surround (C/S) retinex model, which is a color constancy model of the visual nervous system. Since the C/S retinex model involves large-scale spatial filtering with high computational costs, we compared the processing speeds of different hardware implementations of this model. This comparison showed that the FPGA implementation of the C/S retinex model used in this study is more than 10 times faster than the GPU implementation. Using the output of this efficiently processed C/S retinex model, a relatively small convolutional neural network (CNN), which runs on a GPU for embedded devices, performs object classification. We also investigated the impact of the spatial parameters of the C/S retinex model on the classification accuracy of CNNs using a dataset of images acquired under various lighting conditions. This investigation revealed parameters that provide better classification accuracy, and these parameters were largely independent of the CNN architecture used. This system performed object classification under various illumination colors at 52.1 frames per second with a power consumption of approximately 10.9 W.

This paper proposes an optoelectronic architecture of Convolutional Recurrent Neural Network (C-RNN in short). It employs RNN layers that replace area-consuming fully connected (FC) layers in typical convolutional neural network (CNN) architectures. The convolution and RNN layers in this architecture process input data in a pipelined manner that improves the throughput of the inference processing. It takes advantage of both the high input compression capabilities of CNNs and the compact and power-efficient nature of RNNs. The proposed optoelectronic C-RNN architecture achieves over 97.8% accuracy on the MNIST dataset while maintaining the advantages of power-efficient and high-speed characteristics of photonics. Our proposed optoelectronic C-RNN architecture can reach 240 TOPs/W, which is ten times more efficient than CMOS-based dedicated CNN accelerators.

We report on low-power operation of the half-flux-quantum (HFQ) shift-register circuit. We used a superconducting quantum interference device composed of two conventional switching junctions (0-junctions) and one π-shifted non-switching magnetic Josephson junction (π-junction) as a switching element, called 0-0-π SQUID. Because the π-shift assists switching by inducing a spontaneous current, the 0-0-π SQUID shows a nominal small critical current value and is easily switched by a weak driving force. This feature allows us to significantly reduce static and dynamic power consumption at junctions and bias-feeding resistors. We carefully designed the HFQ circuit elements using a circuit parameter optimization tool and introducing additional non-switching 0-0-π SQUIDs to compensate for superconductor phase shifts. We fabricated the test circuit of the 4-bit shift-register by forming Nb/PdNi/Nb π-junctions on the Nb four-layer, 10-kA/cm² device. We successfully obtained correct operation with measured power consumption of 0.12-0.18 μW/bit, which was about 1/10 of the conventional single-flux-quantum shift-register designed with the 0-junctions of the same critical currents and the standard bias voltage.

We demonstrated a coincidence circuit based on single flux quantum (SFQ) circuits with high-time accuracy for a two-photon interferometer using superconducting nanostrip photon detectors (SNSPDs). The coincidence circuit was designed to have 3 different time windows of 30 ps, 100 ps, and 200 ps, which can be selectable on demand. Furthermore, each time window can be fine-tuned by changing the applied bias currents. The circuit was fabricated by the AIST high-speed standard process based on Nb/AlO_x/Nb junctions with critical current density of 10 kA/cm². We experimentally evaluated the time window of the coincidence circuit at temperatures below 2.4 K in a 0.1-W Gifford-McMahon (GM) cryocooler.

We report a hybrid fabrication process for superconductor large-scale integrated circuits (ICs) with ferromagnetic π-shifted Josephson junctions (π junctions) and conventional 0 junctions. π junctions have a Nb/PdNi/Nb sandwich structure and are formed on the reliable Nb four-layer structure used for Nb/AlO_x/Nb 0-junction-based ICs. The additional process for making π junctions causes little damage to the 0 junctions. The π phase shift for the superconducting macroscopic wave function is confirmed by forming 0-0-π SQUIDs that have two 0 junctions and one π junction in a superconducting loop. The π junction serves as the π phase shifter because the critical currents of the π junctions are much larger than those of the 0 junctions. The modulation patterns of the 0-0-π SQUIDs show a clear shift by the magnetic field corresponding to the half flux quantum (HFQ). The critical currents of the 0-0-π SQUIDs without fields, which are referred to as the nominal critical currents, are reduced depending on the product of the loop inductance and the critical current of the 0 junction. The nominal critical current reaches 1/5 of the critical current of the single 0 junction. The small values of the nominal critical currents contribute to the reduction of the power consumption in the HFQ circuit directly.

We describe the operation of a rapid-single-flux-quantum (RSFQ) hardware random number generator comprising two Josephson oscillators, an XOR gate, and a non-destructive read-out (NDRO) cell. Single-flux-quantum (SFQ) pulse trains fed to the input terminals of the XOR gate contain timing jitters due to thermal noises. Then, the output of the XOR gate becomes random, which is stored in the NDRO cell and fetched by the trigger signal. We experimentally evaluated the output randomness of a test circuit fabricated using a 10 kA/cm² Nb/AlO_x/Nb integration process. For evaluation, the NIST FIPS 140-2 test suite was used. 24 random number sequences of 20 kb length were acquired with the trigger signal of 1 MHz, and evaluated for each operation condition. Relationships between the output randomness and operation conditions were obtained. In addition, numerical simulation demonstrated true random number generation at 20 GHz.