Recently, machine learning (ML) methods have been used to create powerful language models for a broad range of natural language processing tasks. An important subset of this field is that of generating code of programming languages for automatic software development. This review provides a broad and detailed overview of studies for code generation using ML. We selected 37 publications indexed in arXiv and IEEE Xplore databases that train ML models on programming language data to generate code. The three paradigms of code generation we identified in these studies are description-to-code, code-to-description, and code-to-code. The most popular applications that work in these paradigms were found to be code generation from natural language descriptions, documentation generation, and automatic program repair, respectively. The most frequently used ML models in these studies include recurrent neural networks, transformers, and convolutional neural networks. Other neural network architectures, as well as non-neural techniques, were also observed. In this review, we have summarized the applications, models, datasets, results, limitations, and future work of 37 publications. Additionally, we include discussions on topics general to the literature reviewed. This includes comparing different model types, comparing tokenizers, the volume and quality of data used, and methods for evaluating synthesized code. Furthermore, we provide three suggestions for future work for code generation using ML.

Titania is considered an interesting anode candidate for Li⁺-ion batteries, as it offers a high theoretical capacity (1280 mAh cm⁻³ or 336 mAh g⁻¹) and long term cycling stability. Unfortunately, the most commonly investigated anatase structure never reaches the theoretical capacity at practical charging rates (i.e. above 1 C). In this work, we compare amorphous (am-TiO₂) to anatase TiO₂ thin-films, and investigate the exceptional performance of am-TiO₂ as Li⁺-ion insertion electrode. An in-depth electrochemical characterization using cyclic voltammetry (CV), constant current lithiation and delithiation, and potentiostatic intermittent titration technique (PITT) is performed. From CV, the insertion and extraction kinetics of am-TiO₂ is found to be unrestricted by diffusion, contrary to anatase. Based on our combined electrochemical results, two different mechanisms are formulated for anatase and am-TiO₂. Whereas anatase is filled from the “top-down”, with a buildup of Li near the electrode/electrolyte interface, am-TiO₂ shows a “bottom-up” filling mechanism. This discrepancy is ascribed to the difference in diffusion coefficient measured for am-TiO₂ and anatase. This work highlights the differences of Li-ion insertion into amorphous TiO₂ compared to anatase, and gives guidance on material development for high capacity and fast charging electrodes.

One current key challenge in graphene research is to tune its charge carrier concentration, i.e., p- and n-type doping of graphene. An attractive approach in this respect is offered by controlled doping via well-ordered self-assembled networks physisorbed on the graphene surface. We report on tunable n-type doping of graphene using self-assembled networks of alkyl-amines that have varying chain lengths. The doping magnitude is modulated by controlling the density of the strong n-type doping amine groups on the surface. As revealed by scanning tunneling and atomic force microscopy, this density is governed by the length of the alkyl chain which acts as a spacer within the self-assembled network. The modulation of the doping magnitude depending on the chain length was demonstrated using Raman spectroscopy and electrical measurements on graphene field effect devices. This supramolecular functionalization approach offers new possibilities for controlling the properties of graphene and other two-dimensional materials at the nanoscale.

As machine learning continues to find applications, extensive research is currently underway across various domains. This study examines the current methodologies of machine learning being investigated to optimize semiconductor manufacturing processes. Our research involved searching the SPIE Digital Library, IEEE Xplore, and ArXiv databases, identifying 58 publications in the field of machine learning-based semiconductor process optimization. These investigations employ machine learning techniques such as feature extraction, feature selection, neural network architecture, and are analyzed using different algorithms. These models find applications in advanced process control, virtual metrology, and quality control, critical aspects in semiconductor manufacturing for enhancing throughput and reducing production costs. We categorize the papers based on the methods and applications employed, summarizing the primary findings. Furthermore, we discuss the general conclusions of several studies. Overall, the reviewed literature suggests that machine learning-based semiconductor manufacturing is rapidly gaining popularity and advancing at a swift pace.

The mechanism of graphene delamination from a Pt catalyst growth surface with electrochemical methods is studied. After a water intercalation step, an electrochemical graphene delamination process is done with a variety of different electrolytes. It is shown that (hydrogen or oxygen) bubble formation is not the main driving force to decouple graphene from its catalyst growth substrate. Ion intercalation is identified as the primary component for a fast graphene delamination process from its catalytic growth substrate. When the Pt/graphene sample is negatively charged, cations will intercalate, assuming they do not reduce within the electrochemical window of the solvent. This cation intercalation does result in graphene delamination. In the same way, anions intercalate in positively charged Pt/graphene samples when they do not react within the electrochemical window of the solvent. Furthermore, it is shown that applying a potential is sufficient (current is not needed) to induce ion intercalation and, as a result, graphene delamination. These findings open the door to avoid Na⁺ or K⁺ contamination introduced during currently described electrochemical graphene delamination. Alternative electrolytes (i.e. ammonium hydroxide and tetraethylammonium hydroxide) are proposed, due to the absence of alkali contaminants and rapid cation intercalation to delaminate graphene.

Positive charges in Hf-based gate stacks play an important role in the negative bias temperature instability of pMOSFETs, and their suppression is a pressing issue. The location of positive charges is not clear, and central to this letter is determining which layer of the stack dominates positive charging. The results clearly show that positive charges are dominated by the interfacial layer (IL) and that they do not pile up at the HfSiON/IL interface. The results support the assumption that positive charges are located close to the IL/substrate interface. Unlike electron trapping that reduces rapidly for thinner Hf dielectric layer, positive charges cannot be reduced by using a thinner HfSiON film.