Advanced Algorithm and Systems (Japan)

BaAl$_2$O$_4$ is a ferroelectric material that exhibits structural quantum criticality through chemical composition tuning. Although theoretical calculations and several diffraction experiments have suggested the involvement of a soft mode in its ferroelectric structural phase transition, direct experimental verification is still lacking. In this study, we successfully observed two soft modes of BaAl$_2$O$_4$ using x-ray inelastic scattering, providing direct experimental evidence for their role in the structural phase transition. Furthermore, we reveal that the soft modes at the M and K points are nearly degenerate in energy, indicating a delicate balance in which either mode could potentially freeze. The K-point mode simultaneously softens toward the transition temperature ($T_{\rm C}$) in a manner nearly identical to the M-point mode. However, the phase transition condenses only at the M point, with the M-point mode stabilizing as an acoustic mode in the low-temperature structure and the K-point mode hardening as temperature decreases.

Terahertz (THz) waves are valuable for evaluating non-metallic materials due to their strong penetrative capabilities. Conventional THz time-domain spectroscopy (THz-TDS) systems based on asynchronous optical sampling (ASOPS) use ultrashort pulse lasers however require a mechanical delay stage, resulting in a trade-off between spectral resolution and acquisition time. Dual-comb spectroscopy (DCS) eliminates the need for delay stage, enabling broadband, high-resolution, and rapid measurements [1], and has been applied to ASOPS THz-TDS systems [2]. However, conventional DCS systems are typically large and complex, as they require precise stabilization of two independent lasers.

Speaker verification based on deep neural network is one of the most prospective biometric authentication methods. But massive computation and parameters make it difficult to deploy speaker verification systems on memory and power constrained embedded devices. Binary neural networks with weights and activations compressed to 1 bit can reduce computational and memory cost significantly. However, speaker verification system based on binary neural networks suffers performance degradation problem. In this paper, to mitigate this problem and facilitate information flow across layers, we optimize binary neural network structure by adding a shortcut to connect layers. We propose an energy and memory efficient speaker system based on the optimized binary networks. Experimental results on text-dependent and text-independent datasets show that performance of proposed speaker verification system outperforms conventional binary neural networks significantly. Compared with full precision networks, the proposed system can largely maintain the performance while reducing memory and computational costs.

Worldwide energy infrastructures are shifting to renewable, but intermittent sources, requiring new solutions for energy storage and conversion. Solid oxide fuel cells are attractive for their high energetic efficiencies, but poor performance in oxygen electrodes contributes to high material costs preventing their widespread adoption. Improving the performance of these technologies is often hindered by inadequate knowledge of localized kinetic and transport behavior. Spatial averaging across the electrode structure makes these properties inaccessible to typical bulk electrochemical characterization techniques like DC polarization and AC impedance spectroscopy. Conversely, chemically sensitive operando and in situ techniques spatially resolve local behavior but struggle to separate physics by timescale. We aim to bridge this gap by extending on previous work using synchrotron-based operando micro X-ray absorption spectroscopy (µ-XAS)1–3 with steady periodic voltage perturbations to isolate and spatiotemporally resolve physics.We introduce this technique – called frequency-resolved X-ray absorption spectroscopy (fr-XAS)– by showing one-dimensional oxygen vacancy distribution images in a patterned thin film La0.6Sr0.4CoO3-δ electrode. Oxygen vacancies in the electrode are compensated by Co oxidation state, resulting in Co K-edge absorption shifts. At a fixed incident X-ray energy, the edge shifts appear as vertical changes, providing a direct measure of electrode oxidation state in the irradiated volume. Scanning the X-ray beam across the electrode and measuring the absorption shift at each point over time (with Fourier analysis to accurately distinguish shifts caused by the voltage perturbation from noise in the signal) allows us to construct the oxygen vacancy distribution images. The oxygen chemical potential is equilibrated at the electrode/electrolyte interface controlled by the voltage with gradients extending away from the interface determined by surface exchange and ionic transport processes. Changing the perturbation frequency separates observed physics by timescale and allows for directly extracting kinetic and transport parameters.Initial results show good quantitative agreement with a one-dimensional Gerischer model, and extracted vacancy diffusion coefficients and surface exchange rates are within a factor of 10 from literature values on porous electrodes.4,5 Impedance results from the same sample are obfuscated by extraneous processes past the point of reasonable interpretation, thereby highlighting the power of fr-XAS to extract meaningful parameters despite extraneous factors. These promising results demonstrate proof-of-concept for fr-XAS and provide motivation for the community to adapt this framework for other material systems relevant to energy storage and conversion technologies.(1) Fujimaki, Y.; Watanabe, H.; Terada, Y.; Nakamura, T.; Yashiro, K.; Hashimoto, S.; Kawada, T.; Amezawa, K. Direct Evaluation of Oxygen Chemical Potential Distribution in an SOFC Cathode by In Situ X-Ray Absorption Spectroscopy. ECS Trans. 2013, 57 (1), 1925–1932. https://doi.org/10.1149/05701.1925ecst.(2) Amezawa, K.; Fujimaki, Y.; Nakamura, T.; Bagarinao, K. D.-; Yamaji, K.; Nitta, K.; Terada, Y.; Iguchi, F.; Yashiro, K.; Yugami, H.; Kawada, T. (Invited) Determination of Effective Reaction Area in a Mixed-Conducting SOFC Cathode. ECS Trans. 2015, 66 (2), 129–135. https://doi.org/10.1149/06602.0129ecst.(3) Amezawa, K.; Fujimaki, Y.; Mizuno, K.; Kimura, Y.; Nakamura, T.; Nitta, K.; Terada, Y.; Iguchi, F.; Yugami, H.; Yashiro, K.; Kawada, T. (Invited) Triple Phase Boundary Reaction in a Mixed-Conducting SOFC Cathode. ECS Trans. 2017, 77 (10), 41–47. https://doi.org/10.1149/07710.0041ecst.(4) Lu, Y.; Kreller, C.; Adler, S. B. Measurement and Modeling of the Impedance Characteristics of Porous La1 − x Sr x CoO3 − δ Electrodes. J. Electrochem. Soc. 2009, 156 (4), B513–B525. https://doi.org/10.1149/1.3079337.(5) Søgaard, M.; Hendriksen, P. V.; Mogensen, M.; Poulsen, F. W.; Skou, E. Oxygen Nonstoichiometry and Transport Properties of Strontium Substituted Lanthanum Cobaltite. Solid State Ionics 2006, 177 (37), 3285–3296. https://doi.org/10.1016/j.ssi.2006.09.005.

1. Purpose Solid-state lithium-ion batteries (SSLIBs) have attracted extensive attentions because of their potentials to improve safety and to achieve higher power and energy densities compared with the conventional lithium-ion batteries. In composite electrodes for SSLIBs, particles of active material (AM) and solid electrolyte (SE) are three-dimensionally distributed and form complicated Li-ion and electron pathways. Then, especially during high rate charging/discharging, the inhomogeneous reaction distribution may be formed in the electrode. The reaction distribution can substantially deteriorate the battery performances of SSLIBs such as capacity, power output, rate capability, and cyclability. Therefore, understanding how the reaction distribution affects the battery performances is crucial for the development of high-performance SSLIBs. In this study, we performed operando three-dimensional (3D) observation of reaction distributions in composite electrodes for SSLIBs using CT-XAFS1 to elucidate the influence of reaction distributions on the battery performances of SSLIBs. 2. Experimental An SSLIB cell with a configuration of Li x CoO2 (LCO)-Li2.2C0.8B0.2O3 (LCBO) composite electrode|LCBO electrolyte|poly (ethylene oxide) (PEO)-based polymer electrolyte|Li metal electrode was fabricated based on the literature2. The composite electrode comprised 8 mg of LCO and 2 mg of LCBO. The SSLIB cell was charged to 4.5V and discharged to 2.0 V with a current of 240 μA (0.3 C) for 3 cycles at 100 °C. The reaction distribution in the composite electrode was observed every 40 minutes during charging and discharging using operando CT-XAFS. The CT-XAFS measurements were carried out in the energy range between 7725.5-7730 eV with an energy step of 0.1 eV under an exposure time of 20 ms per energy and projection angle from -65 to 65° with an angle step of 0.1°. The observation area was 517 × 517 × 49 μm. The spatial and time resolutions were 3.1 μm and 40 min., respectively. 3. Results and Discussion Figure 1(a) shows the charge-discharge curves of the SSLIB cell. The SSLIB cell exhibited the initial charge and discharge capacities of 84 and 46 mAh·g−1, respectively. Those degraded to 24 and 13 mAh·g−1 in the 2nd cycle, and 12 and 12 mAh·g−1 in the 3rd cycle, respectively. Figure 1(b) shows the 3D charging state (Li content) maps of the composite electrode and the representative 2D cross-sectional charging state maps in the in-plane direction at the end of 1st, 2nd, and 3rd charge. Colored/uncolored areas stand for the regions where the AMs exist or not, respectively. The red/blue areas stand for the charged (x = 0.6) /discharged (x = 1.0) AMs, respectively. After the 1st charge, the AMs generally showed a high charging state (red). However, the regions with high charging state significantly decreased after the 2nd charge, and most of the AMs showed a charging state corresponding to x = 0.7 ~ 0.8 (green). The regions with high charging state further decreased after the 3rd charge, and AMs with a low charging state were increased (blue). To more precisely investigate where the less-charged AMs existed after the 2nd and 3rd charge, we subtracted the charging state map after the 1st discharge from that after the 2nd charge, and also the charging state map after the 2nd discharge from that after the 3rd charge (Fig. 1(c)). The 2D maps in Fig. 1(c) thus indicate to what extent the charge reaction progressed during the 2nd or 3rd charge compared with the end of the 1st or 2nd discharge, respectively. The red/blue areas in Fig. 1(c) represent the AM regions where the reaction progressed (Δx = 0.4) or not (Δx = 0) from the end of the 1st or 2nd discharge, respectively. As shown in the left map in Fig. 1(c), the inner parts of the aggregated AMs were significantly less charged compared to the regions near AM/SE interfaces during the 2nd charge. Similarly to the 2nd charge, the charge reaction preferentially progressed near the AM/SE interfaces, while the reaction insufficiently progressed at the inner parts of the aggregated AMs during the 3rd charge. Such a variation in the charging state distribution during cycling could be explained by considering the microstructural change due to the expansion/shrinkage of AM particles. The above results suggest that the expansion/shrinkage of AM particles reduced the connections between AM particles but did not reduce the AM/SE interfaces. We further discuss the mechanism of the capacity degradation during cycling in the presentation.

In autonomous driving, detecting vehicles together with their parts, such as a license plate is important. Many methods with using deep learning detect the license plate based on number recognition. However, there is an idea that the method using deep learning is difficult to use for autonomous driving because of the complexity in realizing deterministic verification. Therefore, development of a method that does not use deep learning(DL) has become important again. Although the authors have made the world’s best performance in 2018 for Caltech data with using DL, this concept has now turned to another research without using DL. The CT5L method is the latest type, that includes techniques of the continuity of vertical and horizontal black-and-white pixel values inside the plate, unique Hough transform, only vertical and horizontal lines are detected, the top five in the order of the number of votes to ensure good performance. In this paper, a method to determine the threshold value for binarizing input by machine learning is proposed, and good results are obtained. The detection rate is improved by about 20 points in percent as compared to the fixed case. It achieves the best performance among the conventional fixed threshold method, Otsu’s method, and the conventional method of JavaANPR.

To investigate the initiation and propagation processes of internal fatigue crack in high strength metallic materials, high resolution and non-destructive observation technique was proposed. The opening and closing behaviors of fatigue cracks in Ti-6Al-4V, and the non-metallic inclusions in SNCM439, which are regarded as a potential crack initiation site, were observed by synchrotron radiation CT (SR-CT) at SPring-8. To ensure a sufficient X ray transmission, fatigue specimens with a small diameter were designed, and a tensile loading jig was newly developed to measure the CTOD during SR-CT. In Ti-6Al-4V, micro CT and nano CT with a resolution around 1 μm and 200 nm, respectively, were used to observe internal cracks. As a result, it was clarified that surface cracks and internal cracks showed different behaviors. The crack tip of surface cracks had a sharp shape and the CTOD was changed by the tensile load. On the other hand, the shape of crack tips of internal cracks was blunted, and the tensile load did not affect CTOD. This crack blunting was explained by the vacuum-like environment around the internal crack. In SNCM439, small non-metallic inclusions with the size around 8 μm in diameter were clearly observed. Accordingly, it is expected that we can observe the initiation and propagation of internal cracks in high strength steel by repeating fatigue tests and SR-CT alternately.

We investigate the Leggett-Garg inequalities and the relative entropy of coherence in the Bixon-Jortner model. First, we analytically derive the general solution of the Bixon-Jortner model by a technique of the Laplace transform. So far, only a special solution has been known for this model. The model has a single state coupled to equally spaced quasi-continuum states. These couplings cause discontinuities in the time evolution of the occupation probability of each state. Second, using the analytical solution, we show that the probability distribution of the quasi-continuum states approaches the Lorentzian function in a period of time between the initial time and the first discontinuity. Third, we examine violation of the Leggett-Garg inequalities and temporal variation of the relative entropy of coherence in the model. We prove that both the inequalities and the relative entropy are invariant under transformations of the energy-level detuning of the single state.