Crackless high-aspect-ratio processing of a silica glass with a temporally shaped ultrafast laser.

In this Letter, we propose a crackless high-aspect-ratio processing method based on a temporally shaped ultrafast laser. The laser pulse is temporally split into two sub pulses: one with smaller energy is used to excite electrons but without ablation so that the applied pressure to the sample is weak, and the other one is used to heat the electrons and achieve material removal after it is temporally stretched by a chirped volume Bragg grating (CVBG). Compared with the conventional ultrafast laser processing, the crack generation is almost suppressed by using this proposed method. The hole depth increases more than 3.3 times, and the aspect ratio is improved at least 2.2 times. Moreover, processing dynamics and parameter dependence are further experimentally studied. It shows that the processing highly depends on the density of electrons excited by the first pulse (P1) and the energy of the second pulse (P2). This novel, to the best of our knowledge, method provides a new route for the precise processing of wide-bandgap materials.

Splitting tensile strength prediction of Metakaolin concrete using machine learning techniques.

Splitting tensile strength (STS) is an important mechanical property of concrete. Modeling and predicting the STS of concrete containing Metakaolin is an important method for analyzing the mechanical properties. In this paper, four machine learning models, namely, Artificial Neural Network (ANN), support vector regression (SVR), random forest (RF), and Gradient Boosting Decision Tree (GBDT) were employed to predict the STS. The comprehensive comparison of predictive performance was conducted using evaluation metrics. The results indicate that, compared to other models, the GBDT model exhibits the best test performance with an R2 of 0.967, surpassing the values for ANN at 0.949, SVR at 0.963, and RF at 0.947. The other four error metrics are also the smallest among the models, with MSE = 0.041, RMSE = 0.204, MAE = 0.146, and MAPE = 4.856%. This model can serve as a prediction tool for STS in concrete containing Metakaolin, assisting or partially replacing laboratory compression tests, thereby saving costs and time. Moreover, the feature importance of input variables was investigated.

Temporal-spatial characteristics of filament induced by a femtosecond laser pulse in transparent dielectrics.

The evolution mechanism of femtosecond laser-induced filaments has been widely investigated owing to its application prospects in microprocessing. However, the material dependence of the excitation, stability, and decay of filaments is not well understood despite the importance of their precise utilization. In this study, the spatiotemporal evolution of filaments induced by a single femtosecond laser pulse in sapphire and silica glass was investigated using time-resolved pump-probe shadowgraphy on femtosecond and picosecond timescales. The results revealed that the evolution was significantly different in the two typically transparent dielectrics in terms of the electronic plasma dynamics and filament lifetimes. This difference can be attributed to the self-trapped excitons (STEs) in silica glass. Furthermore, the filament dependence on pump energy and focal position was experimentally analyzed. Divergent filaments were observed when the focal position was near the surface because of the effect of the excited plasma on beam propagation. Moreover, the evolution of filament length in the two materials was discussed. This study contributes to the applications of filaments in precise processing.

All-in-One Deposition to Synergistically Manipulate Perovskite Growth for High-Performance Solar Cell.

Nonradiative recombination losses originating from crystallographic distortions and issues occurring upon interface formation are detrimental for the photovoltaic performance of perovskite solar cells. Herein, we incorporated a series of carbamide molecules (urea, biuret, or triuret) consisting of both Lewis base (-NH2) and Lewis acid (-C=O) groups into the perovskite precursor to simultaneously eliminate the bulk and interface defects. Depending on the different coordination ability with perovskite component, the incorporated molecules can either modify crystallization dynamics allowing for large crystal growth at low temperature (60°C), associate with antisite or undercoordinated ions for defect passivation, or accumulate at the surface as an energy cascade layer to enhance charge transfer, respectively. Synergistic benefits of the above functions can be obtained by rationally optimizing additive combinations in an all-in-one deposition method. As a result, a champion efficiency of 21.6% with prolonged operational stability was achieved in an inverted MAPbI3 perovskite solar cell by combining biuret and triuret additives. The simplified all-in-one fabrication procedure, adaptable to different types of perovskites in terms of pure MAPbI3, mixed perovskite, and all-inorganic perovskite, provides a cost-efficient and reproducible way to obtain high-performance inverted perovskite solar cells.

Supervised-actor-critic reinforcement learning for intelligent mechanical ventilation and sedative dosing in intensive care units.

BACKGROUND: Reinforcement learning (RL) provides a promising technique to solve complex sequential decision making problems in healthcare domains. Recent years have seen a great progress of applying RL in addressing decision-making problems in Intensive Care Units (ICUs). However, since the goal of traditional RL algorithms is to maximize a long-term reward function, exploration in the learning process may have a fatal impact on the patient. As such, a short-term goal should also be considered to keep the patient stable during the treating process. METHODS: We use a Supervised-Actor-Critic (SAC) RL algorithm to address this problem by combining the long-term goal-oriented characteristics of RL with the short-term goal of supervised learning. We evaluate the differences between SAC and traditional Actor-Critic (AC) algorithms in addressing the decision making problems of ventilation and sedative dosing in ICUs. RESULTS: Results show that SAC is much more efficient than the traditional AC algorithm in terms of convergence rate and data utilization. CONCLUSIONS: The SAC algorithm not only aims to cure patients in the long term, but also reduces the degree of deviation from the strategy applied by clinical doctors and thus improves the therapeutic effect.

Incorporating causal factors into reinforcement learning for dynamic treatment regimes in HIV.

BACKGROUND: Reinforcement learning (RL) provides a promising technique to solve complex sequential decision making problems in health care domains. However, existing studies simply apply naive RL algorithms in discovering optimal treatment strategies for a targeted problem. This kind of direct applications ignores the abundant causal relationships between treatment options and the associated outcomes that are inherent in medical domains. METHODS: This paper investigates how to integrate causal factors into an RL process in order to facilitate the final learning performance and increase explanations of learned strategies. A causal policy gradient algorithm is proposed and evaluated in dynamic treatment regimes (DTRs) for HIV based on a simulated computational model. RESULTS: Simulations prove the effectiveness of the proposed algorithm for designing more efficient treatment protocols in HIV, and different definitions of the causal factors could have significant influence on the final learning performance, indicating the necessity of human prior knowledge on defining a suitable causal relationships for a given problem. CONCLUSIONS: More efficient and robust DTRs for HIV can be derived through incorporation of causal factors between options of anti-HIV drugs and the associated treatment outcomes.

Low-Temperature Atomic Layer Deposition of Metal Oxide Layers for Perovskite Solar Cells with High Efficiency and Stability under Harsh Environmental Conditions.

Rapid progress achieved on perovskite solar cells raises the expectation for their further development toward practical applications. Moisture sensitivity of perovskite materials is one of the major obstacles which limits the long-term durability of the perovskite solar cells, especially in outdoor operation where rainfall and water accumulation on the solar panels often occur. Micro/nanopinholes within the functional layers of the devices usually lead to water vapor penetration, thus subsequent decomposition of perovskites, and finally poor device performance and shortened operational lifetime. In this work, low-temperature atomic layer deposition (ALD) technique was utilized to incorporate pinhole-free metal oxide layers (TiO2 and Al2O3) into an inverted perovskite solar cell consisting of indium tin oxide/NiO/perovskite/PC61BM/TiO2/Ag. The interface properties between the inserted TiO2 layer and the perovskite layer were investigated by X-ray photoelectron spectroscopy. The results showed that TiO2 ALD fabrication process had made negligible degradation to the perovskite layer. The TiO2 layer can significantly reduce interfacial charge recombination loss, improve interfacial contact, and enhance water resistance. A maximum power conversion efficiency (PCE) of 18.3% was achieved for devices with TiO2 interface layers. A stacked Al2O3 encapsulation layer was designed and deposited on top of the devices to further improve device stability under harsh environmental conditions. The encapsulated devices with the best performance retained 97% of the initial PCE after being stored in ambient condition for a thousand hours. They also showed great water resistance, and no significant degradation in terms of PCE and photocurrent of the devices was observed after they were immersed in deionized water for as long as 2 h. Our approach offers a promising way of developing highly efficient and stable perovskite solar cells under real-world operational conditions.