Jia-Min Shieh

Outdoor air monitoring for possible air pollution is always one of the important tasks in managing our living environment. A module of five gas sensors and a new learning model is developed by this study to provide precision, real-time multiple gas concentrations in public spaces. The module suits well an Internet of Things (IoT) node of smart gas sensing that integrates a custom-designed sensor array, driving circuits, and wireless transmission via Bluetooth/Wi-Fi. The sensed gas data are uploaded to a cloud platform for remote storage and then retrieved by an edge device for on-site processing. A cascaded multilayer perceptron (MLP) and long short-term memory (LSTM) model, deployed at the edge using TensorFlow Lite for Microcontrollers (TFLM), compensates successfully for sensor cross-sensitivity, temperature and humidity variations, and other environmental interferences, leading to precise multiple-to-multiple gas concentration estimation. This edge AI deployment reduces cloud computation load and makes it possible to an inference in resource-constrained environments. Experimental validation demonstrates that the proposed MLP–LSTM approach achieves accuracies with mean absolute errors (MAEs) within 113 ppb, 0.108 ppm, 0.110 ppm, 0.124 ppm, and 0.13 ppb for NO_x, H₂S, TVOC, CO, and SO₂, respectively, superior to those by traditional analysis methods. The system architecture also shows great capability for future integration into IoT applications and an alert system for environmental air monitoring.

Model-based optical proximity correction (OPC) is a widely adopted resolution enhancement technique that compensates for lithography proximity effects by adjusting and correcting mask pattern shapes. However, its application to large-area, high-density patterns such as metasurfaces is constrained by the computational demands of image calculation in the lithographic model. This paper introduces an intelligent proximity correction (IPC) system that leverages a deep-learning U-net-based lithographic model to optimize mask designs through iterative algorithms. By eliminating the need for optical calculations while maintaining accurate predictions of photoresist contours directly from mask patterns, the intelligent lithography model demonstrated a 3-fold reduction in simulation time compared to conventional optical models. Overall the IPC system with parallel computation achieves a remarkable 35-fold speed increase compared to conventional OPC approaches without compromising correction accuracy. Experimental results demonstrate that IPC is effective in fabricating high-efficiency, large-area metalenses and metasurface holograms for the visible wavelength using KrF photolithography, opening possibilities to produce complex, large-scale metasurfaces for various applications in optics and photonics.

In this study, we report a low ohmic contact resistance process on a 650 V E-mode p-GaN gate HEMT structure. An amorphous silicon (a-Si) assisted layer was inserted in between the ohmic contact and GaN. The fabricated device exhibits a lower contact resistance of about 0.6 Ω-mm after annealing at 550 °C. In addition, the threshold voltage shifting of the device was reduced from −0.85 V to −0.74 V after applying a high gate bias stress at 150 °C for 10⁻² s. The measured time to failure (TTF) of the device shows that a low thermal budget process can improve the device’s reliability. A 100-fold improvement in HTGB TTF was clearly demonstrated. The study shows a viable method for CMOS-compatible GaN power device fabrication.

Heterogeneous integration of monolayers is an emergent route of spatially combining materials with available platforms for unprecedented properties. A long-standing challenge along this route is to manipulate interfacial configurations of each unit in stacking architecture. A monolayer of transition metal dichalcogenides (TMDs) offers an embodiment of studying interface engineering of integrated systems because optoelectronic performances generally trade off with each other due to interfacial trap states. While ultrahigh photoresponsivity of TMDs phototransistors has been realized, a long response time commonly appears and hinders applications. Here, fundamental processes in excitation and relaxation of the photoresponse are studied and correlated with interfacial traps of the monolayer MoS₂. A mechanism for the onset of saturation photocurrent and the reset behavior in the monolayer photodetector is illustrated based on device performances. Electrostatic passivation of interfacial traps is achieved with the bipolar gate pulse and significantly reduces the response time for photocurrent to reach saturated states. This work paves the way toward fast-speed and ultrahigh-gain devices of stacked two-dimensional monolayers.

While applying machine learning (ML) to semiconductor manufacturing is prevalent, an efficient way to sample the search space has not been explored much in key processes such as lithography, annealing, deposition, and etching. The aim is to use the fewest experimental trials to construct an accurate predictive model. Here, we proposed a technology computer added design (TCAD)-assisted meta-learned sampling approach. The meta-learner adjusts the way of sampling in terms of how to hybridize the TCAD with ML when selecting the next sampling point. While an advanced semiconductor process is expensive, efficient sampling is indispensable. Using laser annealing as an example, we demonstrate the effectiveness of the proposed algorithm where the mean square error (MSE) at the first 100 sampling steps using TCAD-assisted meta-learned sampling is significantly lower than the pure ML approach. Besides, with reference to the pure TCAD approach, the TCAD-assisted sampling prevents the MSE degradation at 200-400 sampling steps. The proposed approach can be used in other manufacturing or even any applied machine intelligence fields.

The emerging applications of silicon photonics in free space, such as LiDARs, free‐space optical communications, and quantum photonics, urge versatile emission shaping beyond the capabilities of conventional grating couplers. In these applications, silicon photonic chips deliver free‐space emission to detect or manipulate external objects. Light needs to emit from a silicon photonic chip to the free space with specific spatial modes, which produce focusing, collimation, orbital angular momentum, or even holographic projection. A platform that offers versatile shaping of free‐space emission, while maintaining the CMOS compatibility and monolithic integration of silicon photonics is in pressing need. Here we demonstrate a platform that integrates metasurfaces monolithically on silicon photonic integrated circuits. The metasurfaces consist of amorphous silicon nanopillars evanescently coupled to silicon waveguides. We demonstrate experimentally diffraction‐limited beam focusing with a Strehl ratio of 0.82. The focused spot can be switched between two positions by controlling the excitation direction. We also realize a meta‐hologram experimentally that projects an image above the silicon photonic chip. This platform can add a highly versatile interface to the existing silicon photonic ecosystems for precise delivery of free‐space emission.

Back-end-of-line compatible 400°C thermally robust perpendicular spin-orbit torque (p-SOT) cells with reduced MgO short fails are demonstrated by the etch-stop-on-MgO process. The stop-on-MgO cell features the SOT channel continuity and no metal redeposition at MgO sidewall after ion beam etching. To the best of our knowledge, the endurance as high as 1010 cycles using the field-free spin-transfer torque (STT) assisted SOT writing is achieved for the first time. The SOT switching current density can be reduced by increasing the STT current density to save write energy. The stop-on-MgO cell does not degrade the cell switching speed, since the switching always starts from the inner free layer and the domain propagation at the extended free layer does not affect junction resistance, as shown by micromagnetic simulation. The simulation also reveals that the thermal stability factor of stop-on-MgO cells is enhanced by the extended free layer, which suffers less from the interference of pinned layer edge stray field.