Machine learning was applied to classify the device characteristics of indium gallium zinc oxide (IGZO) thin-film transistors (TFTs). A K-means approach was employed for initial clustering of IGZO transfer curves into three of four grades (high, medium-high, medium, and low) of TFT performance according to qualitative features. A 2-layered artificial neural network (ANN) and 4-layered deep neural network (DNN) were used to extract mobility, threshold voltage, on/off current ratio, and sub-threshold slope device parameters from high-grade and medium-high-grade oxide TFTs. Ground-truth device parameters were calculated using in-house codes based on a rules-based approach consistent with the definitions employed to train the ANN and DNN. The DNN-predicted parameters were in closer agreement with manual and macro-based calculations than were those obtained from the ANN. Synergistic integration of K-means clustering and DNN effectively extracted TFT device parameters encountered in processing high volumes of data in industrial and academic domains of the microelectronics field.

Zinc oxide (ZnO) is a transparent wide band gap semiconductor material with various possible applications in form of thin films. Most previous studies on atomic layer deposition (ALD) of ZnO thin films utilized a few well-known processes with diethylzinc (DEZ) and counter-reactants such as H2O and O3. However, O3 and H2O reactants have relatively strong reactivity, so that they are not suitable for substrates sensitive to oxidation. Therefore, development of milder non-aqueous alternative ALD process for ZnO is highly desired. In this study, we introduce ALD of ZnO using alcohols with theoretically optimized molecular structure. To discover suitable alcohol reactants for ZnO ALD, reaction pathways and reactivity between various types of alcohol reactants with surface-ethyl groups were evaluated through density functional theory calculations. It was found that unsaturated allylic alcohols would have the lowest activation energy for ZnO ALD via an allylic rearrangement mechanism. Experimental, novel ALD processes for ZnO using DEZ with alcohol as oxygen sources are set up. Ethanol as a typical simple alcohol is compared to 2-methyl-3-buten-2-ol (MBO) as an alcohol with optimal molecular structure setup. ALD ZnO films using MBO showed processes and material properties comparable to those of H2O-ALD ZnO. ZnO thin films as transparent conducting oxide could be obtained, and device performances of thin-film transistors based on alcohol-ALD processes are evaluated.

A high-performance machine learning-assisted gas sensor strategy based on the integration of supervised and unsupervised learning with a gas-sensitive semiconductor metal oxide (SMO) gas sensor array is introduced. A 4-SMO sensor array was chosen as a test sensor system for detecting carbon monoxide (CO) and ethyl alcohol (C2H5OH) mixtures using 15 different combinations. Gas sensing detection/classification was performed with different numbers of gas sensor and machine learning algorithms. K-Means clustering was successfully employed to rationally identify the similarity features of targeted gases among 4 different groups, i.e., matrix gas, two single-component gases, and one two-gas mixture, based on only unlabeled voltage-based gas sensing information. Detailed classification was performed through a multitude of supervised algorithms, i.e., 2-layer artificial neural networks (ANNs), 4-layer deep neural networks (DNNs), 1-dimensional convolutional neural networks (1D CNNs), and 2-dimensional CNNs (2D CNNs). The numerical-based DNNs and image-based CNNs are shown to be excellent approaches for gas detection and classification, as indicated by the highest accuracy and lowest loss indicators. Through the analysis of the influence of the number of sensors on the arrayed gas sensor system, the application of machine learning methodology to an arrayed gas sensor system demonstrates four unique features, i.e., a data augmentation methodology, machine learning approach of combining K-means clustering and neural networks, and a systematic approach to optimized sensor combinations, potentially leading to the practical sensor networks based on chemical sensors. Even two SMO sensor combinations are shown to be highly effective in gas discrimination against diverse gas environments assisted through numeric-based DNNs and image-based 1D CNNs, overcoming the simple clustering proposed through the unsupervised K-means clustering.