High Glass Transition Temperature Fluorinated Polymers Based on Transfer Learning with Small Experimental Data.

Machine learning can be used to predict the properties of polymers and explore vast chemical spaces. However, the limited number of available experimental datasets hinders the enhancement of the predictive performance of a model. This study proposes a machine learning approach that leverages transfer learning and ensemble modeling to efficiently predict the glass transition temperature (Tg) of fluorinated polymers and guide the design of high Tg copolymers. Initially, the QM9 dataset is employed for model pre-training, thus providing robust molecular representations for the subsequent fine-tuning of a specialized copolymer dataset. Ensemble modeling is used to further enhance prediction robustness and reliability, effectively addressing the problems owing to the limited and unevenly distributed nature of the copolymer dataset. Finally, a fine-tuned ensemble model is used to navigate a vast chemical space comprising 61 monomers and identify promising candidates for high Tg fluorinated polymers. The model predicts 247 entries capable of achieving a Tg over 390 K, of which 14 are experimentally validated. This study demonstrates the potential of machine learning in material design and discovery, highlighting the effectiveness of transfer learning and ensemble modeling strategies for overcoming the challenges posed by small datasets in complex copolymer systems. This article is protected by copyright. All rights reserved.

Room Temperature Reduction of Nitrogen Oxide on Iron Metal-Organic Frameworks.

Nitrogen oxides represent one of the main threats for the environment. Despite decades of intensive research efforts, a sustainable solution for NOx removal under environmental conditions is still undefined. Using theoretical modelling, material design, state-of-the-art investigation methods and mimicking enzymes, it is found that selected porous hybrid iron(II/III) based MOF material are able to decompose NOx, at room temperature, in the presence of water and oxygen, into N2 and O2 and without reducing agents. This paves the way to the development of new highly sustainable heterogeneous catalysts to improve air quality.

Optical property dataset of inorganic phosphor.

Developing inorganic phosphor with desired properties for light-emitting diode application has traditionally relied on time-consuming and labor-intensive material development processes. Moreover, the results of material development research depend significantly on individual researchers' intuition and experience. Thus, to improve the efficiency and reliability of materials discovery, machine learning has been widely applied to various materials science applications in recent years. However, the prediction capabilities of machine learning methods fundamentally depend on the quality of the training datasets. In this work, we constructed a high-quality and reliable dataset that contains experimentally validated inorganic phosphors and their optical properties, sourced from the literature on inorganic phosphors. Our dataset includes 3952 combinations of 21 dopant elements in 2238 host materials from 553 articles. The dataset provides material information, optical properties, measurement conditions for inorganic phosphors, and meta-information. Among the preliminary machine learning results, the essential properties of inorganic phosphors, such as maximum Photoluminescence (PL) emission wavelength and PL decay time, show overall satisfactory prediction performance with coefficient of determination ( R 2 ) scores of 0.7 or more. We also confirmed that the measurement conditions significantly improved prediction performance.

Investigation of Laser-Induced Graphene (LIG) on a Flexible Substrate and Its Functionalization by Metal Doping for Gas-Sensing Applications.

Graphene materials synthesized using direct laser writing (laser-induced graphene; LIG) make favorable sensor materials because of their large surface area, ease of fabrication, and cost-effectiveness. In particular, LIG decorated with metal nanoparticles (NPs) has been used in various sensors, including chemical sensors and electronic and electrochemical biosensors. However, the effect of metal decoration on LIG sensors remains controversial; hypotheses based on computational simulations do not always match the experimental results, and even the experimental results reported by different researchers have not been consistent. In the present study, we explored the effects of metal decorations on LIG gas sensors, with NO2 and NH3 gases as the representative oxidizing and reducing agents, respectively. To eliminate the unwanted side effects arising from metal salt residues, metal NPs were directly deposited via vacuum evaporation. Although the gas sensitivities of the sensors deteriorate upon metal decoration irrespective of the metal work function, in the case of NO2 gas, they improve upon metal decoration in the case of NH3 exposure. A careful investigation of the chemical structure and morphology of the metal NPs in the LIG sensors shows that the spontaneous oxidation of metal NPs with a low work function changes the behavior of the LIG gas sensors and that the sensors' behaviors under NO2 and NH3 gases follow different principles.

Data-Driven Prediction of Configurational Stability of Molecule-Adsorbed Heterogeneous Catalysts.

The design of new heterogeneous catalysts that convert small molecules into valuable chemicals is a key challenge for constructing sustainable energy systems. Density functional theory (DFT)-based design frameworks based on the understanding of molecular adsorption on the catalytic surface have been widely proposed to accelerate experimental approaches to develop novel catalysts. In addition, a machine learning (ML)-combined design framework was recently proposed to further reduce the inherent time cost of DFT-based frameworks. However, because of the lack of prior information on chemical interactions between arbitrary surfaces and adsorbates, the efficacy of the computational screening approaches would be reduced by obtaining unexpected structural anomalies (i.e., abnormally converged surface-adsorbate geometries after the DFT calculations) during an exhaustive exploration of chemical space. To overcome this challenge, we propose an ML framework that directly predicts the configurational stability of a given initial surface-adsorbate geometry. Our benchmark experiments with the Open Catalysts 20 (OC20) dataset show promising performance on classifying stable geometry (i.e., F1-score of 0.922, the area under the receiver operating characteristics (AUROC) of 0.906, and Matthews correlation coefficient (MCC) of 0.633) with a high precision of 0.921 by utilizing an ensemble approach. We further interpret the generalizability and domain applicability of the trained model in terms of the chemical space of the OC20 dataset. Furthermore, from an experiment on the training set size dependence of model performance, we found that our ML model could be practically applicable to classify stable configurations even with a relatively small number of training data.

Data-Driven Enhancement of ZT in SnSe-Based Thermoelectric Systems.

Doping and alloying are fundamental strategies to improve the thermoelectric performance of bare materials. However, identifying outstanding elements and compositions for the development of high-performance thermoelectric materials is challenging. In this study, we present a data-driven approach to improve the thermoelectric performance of SnSe compounds with various doping. Based on the newly generated experimental and computational dataset, we built highly accurate predictive models of thermoelectric properties of doped SnSe compounds. A well-designed feature vector consisting of the chemical properties of a single atom and the electronic structures of a solid plays a key role in achieving accurate predictions for unknown doping elements. Using the machine learning predictive models and calculated map of the solubility limit for each dopant, we rapidly screened high-dimensional material spaces of doped SnSe and evaluated their thermoelectric properties. This data-driven search provided overall strategies to optimize and improve the thermoelectric properties of doped SnSe compounds. In particular, we identified five dopant candidate elements (Ge, Pb, Y, Cd, and As) that provided a high ZT exceeding 2.0 and proposed a design principle for improving the ZT by Sn vacancies depending on the doping elements. Based on the search, we proposed yttrium as a new high-ZT dopant for SnSe with experimental confirmations. Our research is expected to lead to novel high-ZT thermoelectric material candidates and provide cutting-edge research strategies for materials design and extraction of design principles through data-driven research.

Electrochemical Transparency of Graphene.

In the present study, we used the electrochemical transparency of graphene to show that the direct intercalation of alkali-metal cations is not a prerequisite for the redox reaction of Prussian blue (PB). PB thin films passivated with monolayer graphene still underwent electrochemical redox reactions in the presence of alkali-metal ions (K+ or Na+) despite the inability of the cations to penetrate the graphene and be incorporated into the PB. Graphene passivation not only preserved the electrochemical activity of the PB but also substantially enhanced the stability of the PB. As a proof of concept, we showed that a transparent graphene electrode covering PB can be used as an excellent hydrogen peroxide transducer, thereby demonstrating the possibility of realizing an electrochemical sensor capable of long-term measurements.

Nonlinearity encoding to improve extrapolation capabilities for unobserved physical states.

The fundamental goal of machine learning (ML) in physical science is to predict the physical properties of unobserved states. However, an accurate prediction for input data outside of training distributions is a challenging problem in ML due to the nonlinearities in input and target dynamics. For an accurate extrapolation of ML algorithms, we propose a new data-driven method that encodes the nonlinearities of physical systems into input representations. Based on the proposed encoder, a given physical system is described as linear-like functions that are easy to extrapolate. By applying the proposed encoder, the extrapolation errors were significantly reduced by 48.39% and 40.04% in n-body problem and materials property prediction, respectively.

Electrode-Induced Self-Healed Monolayer MoS2 for High Performance Transistors and Phototransistors.

Contact engineering for monolayered transition metal dichalcogenides (TMDCs) is considered to be of fundamental challenge for realizing high-performance TMDCs-based (opto) electronic devices. Here, an innovative concept is established for a device configuration with metallic copper monosulfide (CuS) electrodes that induces sulfur vacancy healing in the monolayer molybdenum disulfide (MoS2 ) channel. Excess sulfur adatoms from the metallic CuS electrodes are donated to heal sulfur vacancy defects in MoS2 that surprisingly improve the overall performance of its devices. The electrode-induced self-healing mechanism is demonstrated and analyzed systematically using various spectroscopic analyses, density functional theory (DFT) calculations, and electrical measurements. Without any passivation layers, the self-healed MoS2 (photo)transistor with the CuS contact electrodes show outstanding room temperature field effect mobility of 97.6 cm2 (Vs)-1 , On/Off ratio > 108 , low subthreshold swing of 120 mV per decade, high photoresponsivity of 1 × 104  A W-1 , and detectivity of 1013 jones, which are the best among back-gated transistors that employ 1L MoS2 . Using ultrathin and flexible 2D CuS and MoS2 , mechanically flexible photosensor is also demonstrated, which shows excellent durability under mechanical strain. These findings demonstrate a promising strategy in TMDCs or other 2D material for the development of high performance and functional devices including self-healable sulfide electrodes.

Polycrystalline SnSe with a thermoelectric figure of merit greater than the single crystal.

Thermoelectric materials generate electric energy from waste heat, with conversion efficiency governed by the dimensionless figure of merit, ZT. Single-crystal tin selenide (SnSe) was discovered to exhibit a high ZT of roughly 2.2-2.6 at 913 K, but more practical and deployable polycrystal versions of the same compound suffer from much poorer overall ZT, thereby thwarting prospects for cost-effective lead-free thermoelectrics. The poor polycrystal bulk performance is attributed to traces of tin oxides covering the surface of SnSe powders, which increases thermal conductivity, reduces electrical conductivity and thereby reduces ZT. Here, we report that hole-doped SnSe polycrystalline samples with reagents carefully purified and tin oxides removed exhibit an ZT of roughly 3.1 at 783 K. Its lattice thermal conductivity is ultralow at roughly 0.07 W m-1 K-1 at 783 K, lower than the single crystals. The path to ultrahigh thermoelectric performance in polycrystalline samples is the proper removal of the deleterious thermally conductive oxides from the surface of SnSe grains. These results could open an era of high-performance practical thermoelectrics from this high-performance material.

Tuplewise Material Representation Based Machine Learning for Accurate Band Gap Prediction.

The open-access material databases allowed us to approach scientific questions from a completely new perspective with machine learning methods. Here, on the basis of open-access databases, we focus on the classical band gap problem for predicting accurately the band gap of a crystalline compound using a machine learning approach with newly developed tuplewise graph neural networks (TGNN), which is devised to automatically generate input representation of crystal structures in tuple types and to exploit crystal-level properties as one of the input features. Our method brings about a highly accurate prediction of the band gaps at hybrid functionals and GW approximation levels for multiple material data sets without heavy computational cost. Furthermore, to demonstrate the applicability of our prediction model, we provide a data set of GW band gaps for 45835 materials predicted by TGNN posing higher accuracy than standard density functional theory calculations.

Machine-guided representation for accurate graph-based molecular machine learning.

In chemistry-related fields, graph-based machine learning has received significant attention as atoms and their chemical bonds in a molecule can be represented as a mathematical graph. However, many molecular properties are sensitive to changes in the molecular structure. For this reason, molecules have a mixed distribution for their molecular properties in molecular space, and it consequently makes molecular machine learning difficult. However, this problem has not been investigated in either chemistry or computer science. To tackle this problem, we propose a robust and machine-guided molecular representation based on deep metric learning (DML), which automatically generates an optimal representation for a given dataset. To this end, we first adopt DML for molecular machine learning by integrating it with graph neural networks (GNNs) and devising a new objective function for representation learning. In experimental evaluations, machine learning algorithms with the proposed method achieved better prediction accuracy than state-of-the-art GNNs. Furthermore, the proposed method was also effective on extremely small datasets, and this result is impressive because many real world applications suffer from a lack of training data.

Costless Performance Improvement in Machine Learning for Graph-Based Molecular Analysis.

Graph neural networks (GNNs) have attracted significant attention from the chemical science community because molecules can be represented as a featured graph. In particular, graph convolutional network (GCN) and its variants have been widely used and have shown a state-of-the-art performance in analyzing molecules, such as molecular label classification, drug discovery, and molecular property prediction. However, in molecular analysis, existing GCNs have two fundamental limitations: (1) information of the molecular scale is distorted and (2) global structures in a molecule are ignored. These limitations can seriously degrade the performance in the machine learning-based molecular analysis because the scale and global structure information of a molecule occasionally have a significant effect on the molecular properties. To overcome the limitations of existing GCNs, we comprehensively analyzed the structure of GCNs and developed a costless solution for the limitations of GCNs. To demonstrate the effectiveness of our solution, extensive experiments were conducted on various benchmark datasets.

Tuning graphene transistors through ad hoc electrostatics induced by a nanometer-thick molecular underlayer.

We report on the modulation of the electrical properties of graphene-based transistors that mirror the properties of a few nanometers thick layer made of dipolar molecules sandwiched in between the 2D material and the SiO2 dielectric substrate. The chemical composition of the films of quinonemonoimine zwitterion molecules adsorbed onto SiO2 has been explored by means of X-ray photoemission and mass spectroscopy. Graphene-based devices are then fabricated by transferring the 2D material onto the molecular film, followed by the deposition of top source-drain electrodes. The degree of supramolecular order in disordered films of dipolar molecules was found to be partially improved as a result of the electric field at low temperatures, as revealed by the emergence of hysteresis in the transfer curves of the transistors. The use of molecules from the same family, which are suitably designed to interact with the dielectric surface, results in the disappearance of the hysteresis. DFT calculations confirm that the dressing of the molecules by an external electric field exhibits multiple minimal energy landscapes that explain the thermally stabilized capacitive coupling observed. This study demonstrates that the design and exploitation of ad hoc molecules as an interlayer between a dielectric substrate and graphene represents a powerful tool for tuning the electrical properties of the 2D material. Conversely, graphene can be used as an indicator of the stability of molecular layers, by providing insight into the energetics of ordering of dipolar molecules under the effect of electrical gating.

Scratch to sensitize: scratch-induced sensitivity enhancement in semiconductor thin-film sensors.

Semiconductor gas sensors are advantageous in miniaturization and can be used in a wide range of applications, yet consume large power due to high operating temperature. Here we demonstrated the ability of nanoscale scratches produced with mechanical abrasion to enhance the chemical sensitivity of thin-film-type semiconductor sensors. Well-aligned arrays of scratches parallel to the electrical current direction between the source and drain electrodes were made, using typical polishing machines with diamond suspensions, on semiconductor thin films produced with various deposition methods such as atomic layer deposition (ALD), sputtering, and the sol-gel technique. Processing with sharp diamond microparticles left nano-grooves on the surface, together with changes in chemical composition. For all of the tested metal oxide thin films, the introduction of scratches yielded increased quantities of oxygen vacancies and metallic components. Scratched ZnO devices exhibited superior performance even at room temperature, as predicted by a computational simulation that showed increased binding energy of gas molecules on defects. The scratch technique shown in the present study may be used to produce dense arrays of nanometer-scale, chemically functionalized line patterns on substrates larger than a few tens of centimeters with minimum cost, which in turn may be used in a variety of applications including massive arrays of sensors displaying high sensitivity.

Atomic-Level Customization of 4 in. Transition Metal Dichalcogenide Multilayer Alloys for Industrial Applications.

Despite many encouraging properties of transition metal dichalcogenides (TMDs), a central challenge in the realm of industrial applications based on TMD materials is to connect the large-scale synthesis and reproducible production of highly crystalline TMD materials. Here, the primary aim is to resolve simultaneously the two inversely related issues through the synthesis of MoS2(1- x ) Se2 x ternary alloys with customizable bichalcogen atomic (S and Se) ratio via atomic-level substitution combined with a solution-based large-area compatible approach. The relative concentration of bichalcogen atoms in the 2D alloy can be effectively modulated by altering the selenization temperature, resulting in 4 in. scale production of MoS1.62 Se0.38 , MoS1.37 Se0.63 , MoS1.15 Se0.85 , and MoS0.46 Se1.54 alloys, as well as MoS2 and MoSe2 . Comprehensive spectroscopic evaluations for vertical and lateral homogeneity in terms of heteroatom distribution in the large-scale 2D TMD alloys are implemented. Se-stimulated strain effects and a detailed mechanism for the Se substitution in the MoS2 crystal are further explored. Finally, the capability of the 2D alloy for industrial application in nanophotonic devices and hydrogen evolution reaction (HER) catalysts is validated. Substantial enhancements in the optoelectronic and HER performances of the 2D ternary alloy compared with those of its binary counterparts, including pure-phase MoS2 and MoSe2 , are unambiguously achieved.

New red-emitting phosphor Rb x K3-x SiF7:Mn4+ (x = 0, 1, 2, 3): DFT predictions and synthesis.

Finding new phosphors through an efficient method is important in terms of saving time and cost related to the development of phosphor materials. The ability to identify new phosphors through preliminary simulations by calculations prior to the actual synthesis of the materials can maximize the efficiency of novel phosphor development. In this paper, we demonstrate the use of density functional theory (DFT) calculations to guide the development of a new red phosphor. We performed first-principles calculations based on DFT for pristine and Mn-doped Rb x K3-x SiF7 (x = 0, 1, 2, 3) and predicted their stability, electronic structure, and luminescence properties. On the basis of the results, we then synthesized the stable Rb2KSiF7:Mn4+ red conversion phosphor and investigated its luminescence, structure, and stability. As a result, we confirmed that Rb2KSiF7:Mn4+ emitted red light with a longer wavelength than that emitted by K3SiF7:Mn4+ and a wavelength similar to that of K2SiF6:Mn4+. These results show that DFT calculations can provide rational insights into the design of a phosphor material before it is synthesized, thereby reducing the time and cost required to develop new red conversion phosphors.

Ultrathin Metal Crystals: Growth on Supported Graphene Surfaces and Applications.

Controlled nucleation and growth of metal clusters in metal deposition processes is a long-standing issue for thin-film-based electronic devices. When metal atoms are deposited on solid surfaces, unintended defects sites always lead to a heterogeneous nucleation, resulting in a spatially nonuniform nucleation with irregular growth rates for individual nuclei, resulting in a rough film that requires a thicker film to be deposited to reach the percolation threshold. In the present study, it is shown that substrate-supported graphene promotes the lateral 2D growth of metal atoms on the graphene. Transmission electron microscopy reveals that 2D metallic single crystals are grown epitaxially on supported graphene surfaces while a pristine graphene layer hardly yields any metal nucleation. A surface energy barrier calculation based on density functional theory predicts a suppression of diffusion of metal atoms on electronically perturbed graphene (supported graphene). 2D single Au crystals grown on supported graphene surfaces exhibit unusual near-infrared plasmonic resonance, and the unique 2D growth of metal crystals and self-healing nature of graphene lead to the formation of ultrathin, semitransparent, and biodegradable metallic thin films that could be utilized in various biomedical applications.

Synthesis of silver sulfide nanoparticles and their photodetector applications.

Silver sulfide nanoparticles (Ag2S NPs) are currently being explored as infrared active nanomaterials that can provide environmentally stable alternatives to heavy metals such as lead. In this paper, we describe the novel synthesis of Ag2S NPs by using a sonochemistry method and the fabrication of photodetector devices through the integration of Ag2S NPs atop a graphene sheet. We have also synthesized Li-doped Ag2S NPs that exhibited a significantly enhanced photodetector sensitivity via their enhanced absorption ability in the UV-NIR region. First-principles calculations based on a density functional theory formalism indicated that Li-doping produced a dramatic enhancement of NIR photoluminescence of the Ag2S NPs. Finally, high-performance photodetectors based on CVD graphene and Ag2S NPs were demonstrated and investigated; the hybrid photodetectors based on Ag2S NPs and Li-doped Ag2S NPs exhibited a photoresponse of 2723.2 and 4146.0 A W-1 respectively under a light exposure of 0.89 mW cm-2 at 550 nm. Our novel approach represents a promising and effective method for the synthesis of eco-friendly semiconducting NPs for photoelectric devices.

Graphene as a thin-film catalyst booster: graphene-catalyst interface plays a critical role.

Due to its extreme thinness, graphene can transmit some surface properties of its underlying substrate, a phenomenon referred to as graphene transparency. Here we demonstrate the application of the transparency of graphene as a protector of thin-film catalysts and a booster of their catalytic efficiency. The photocatalytic degradation of dye molecules by ZnO thin films was chosen as a model system. A ZnO thin film coated with monolayer graphene showed greater catalytic efficiency and long-term stability than did bare ZnO. Interestingly, we found the catalytic efficiency of the graphene-coated ZnO thin film to depend critically on the nature of the bottom ZnO layer; graphene transferred to a relatively rough, sputter-coated ZnO thin film showed rather poor catalytic degradation of the dye molecules while a smooth sol-gel-synthesized ZnO covered with monolayer graphene showed enhanced catalytic degradation. Based on a systematic investigation of the interface between graphene and ZnO thin films, we concluded the transparency of graphene to be critically dependent on its interface with a supporting substrate. Graphene supported on an atomically flat substrate was found to efficiently transmit the properties of the substrate, but graphene suspended on a substrate with a rough nanoscale topography was completely opaque to the substrate properties. Our experimental observations revealed the morphology of the substrate to be a key factor affecting the transparency of graphene, and should be taken into account in order to optimally apply graphene as a protector of catalytic thin films and a booster of their catalysis.