Search

We developed Electronic Structure Network (ESNet) to predict formation energies using density of states extracted from initial structures. ESNet outperformed previously reported models that used other input features and architectures.

The activity and selectivity for C2+ products from electrochemical CO2 reduction in a zero-gap membrane-electrode assembly (MEA) are improved using a synchronous KOH-activation and tailoring of Cu catalyst thickness.

In recent years, a development of appropriate crystal representations for accurate prediction of inorganic crystal properties has been considered as one of the essential tasks to accelerate materials discovery through high-throughput virtual screening (HTVS). However, many of them were developed aiming to predict properties of the given structures, although property predictions of ground state structures using unrelaxed structures as inputs are much more important in practical HTVS. To tackle this challenge, we develop a chemically inspired convolutional neural network based on convolution block attention modules using density of states of unrelaxed initial structures (IS-DOS) as inputs. Our model, Electronic Structure Network (ESNet), achieved the highest accuracy for predicting formation energy, proving that IS-DOS is appropriate input for the property prediction and the attention module is capable of properly featurizing DOS signals by capturing contributions of each spin and orbital state. In addition, we statistically evaluated a stability screening performance of ESNet, measuring computational cost and capability of materials discovery simultaneously. We found that ESNet outperformed previously reported models and various models with different types of input features and architectures. Indeed, ESNet successfully discovered 926 stable materials from 15,318 unrelaxed structures with 82 % reduced computational cost compared to the complete DFT validation.

Toluene diisocyanate (TDI) is an important raw material to produce a flexible polyurethane foam, and the demand for TDI is growing as the polyurethane market is driven by high demand. The cold phosgenation reactor plays a vital role in the production of TDI, in that the overall reaction selectivity is determined and the most of by-product urea, which is critical to the entire downstream process, is produced inside the reactor. Therefore, the optimal design of the cold phosgenation reactor is very important to improve the overall process efficiency and operability of the TDI production process. In this research, we develop a framework for designing and optimizing TDI reactors through Bayesian optimization methods with design parameters including the diameter of two inlet nozzles, the angle between the nozzles, the size of the mixing zone, and the ratio of the converging-diverging nozzle. A comprehensive 3-dimensional computational fluid dynamics (CFD) reactor model is incorporated into the Gaussian process (Kriging) to construct a surrogate model, whose posterior is subsequently updated with new sample points searched by the acquisition function evaluated within the Bayesian optimization algorithm. As a result, the optimal design is obtained, and the urea selectivity is reduced by 11.6% compared with the basic design scheme. Compared to the 6 × 107 simulations required for full grid search, only 61 function evaluations were performed to attain the optimum, demonstrating that the proposed framework will help efficiently achieve the optimal design of the expensive CFD reactor models that demand a high computational cost and time for evaluation.

Electrochemical reduction of O2 provides a clean and decentralized pathway to produce H2O2 compared to the current energy-intensive anthraquinone process. As the electrochemical reduction of O2 pro...

Electrochemical CO2 reduction is always accompanied by a competitive hydrogen evolution reaction as water is used as a hydrogen source. In addition to intrinsic activity control, geometrical factor...

Electroreduction systems to convert CO2 into CO via Ag electrodes have been intensely studied as a means of producing carbon-neutral fuels or chemical products. However, despite many efforts to maximize the performance of CO-producing systems, the performance of electrochemical cells that produce CO has not yet reached the level of economic viability. Moreover, compared with electrode development attempts, studies on the optimization of large-scale CO-producing systems are lacking, thus impeding the commercialization of electrochemical CO2 reduction systems. In this study, we present optimization results of a pilot-scale CO production system. Operating conditions such as pressure, temperature, and cell voltage were considered as the optimization variables to improve the CO partial current density. To facilitate experiment-based optimization of the pilot-scale operation, we adopted an efficient design of the experiment, for which data points were decided by input–output relations. As a result, the maximum CO partial current reached 2.56 A using a 50 cm2 electrode within 25 experiments. In addition, regression analysis results were provided for future studies on the systematic optimization of electrochemical systems. The operating temperature and CO2 solubility were more highly correlated with the current density and selectivity than was the applied cell voltage, and the CO current density could be predicted with high accuracy.

Various databases of density functional theory (DFT) calculations for materials and adsorption properties are currently available. Using the Materials Project and GASpy databases of material stability and binding energies (H* and CO*), respectively, we evaluate multiple aspects of catalysts to discover active, stable, CO-tolerant, and cost-effective hydrogen evolution and oxidation catalysts. Finally, we suggest a few candidate materials for future experimental validations. We highlight that the stability analysis is easily obtainable but provides invaluable information to assess thermodynamic and electrochemical stability, bridging the gap between simulations and experiments. Furthermore, it reduces the number of expensive DFT calculations required to predict catalytic activities of surfaces by filtering out unstable materials.

The electrochemical reduction of CO2 stores intermittent renewable energy in valuable raw materials, such as chemicals and transportation fuels, while minimizing carbon emissions and promoting carbon-neutral cycles. Recent technoeconomic reports suggested economically feasible target products of CO2 electroreduction and the relative influence of key performance parameters such as faradaic efficiency (FE), current density, and overpotential in the practical industrial-scale applications. Furthermore, fundamental factors, such as available reaction pathways, shared intermediates, competing hydrogen evolution reaction, scaling relations of the intermediate binding energies, and CO2 mass transport limitations, should be considered in relation to the electrochemical CO2 reduction performance. Intensive research efforts have been devoted to designing and developing advanced electrocatalysts and improving mechanistic understanding. More recently, the research focus was extended to the catalyst environment, because the interfacial region can delicately modulate the catalytic activity and provide effective solutions to challenges that were not fully addressed in the material development studies. Herein, we discuss the importance of catalyst-electrolyte interfaces in improving key operational parameters based on kinetic equations. Furthermore, we extensively review previous studies on controlling organic modulators, electrolyte ions, electrode structures, as well as the three-phase boundary at the catalyst-electrolyte interface. The interfacial region modulates the electrocatalytic properties via electronic modification, intermediate stabilization, proton delivery regulation, catalyst structure modification, reactant concentration control, and mass transport regulation. We discuss the current understanding of the catalyst-electrolyte interface and its effect on the CO2 electroreduction activity.

In this paper, we present a multicompartment model of an ethylene–vinyl acetate autoclave reactor including the mixing effects of the stirring device analyzed using computational fluid dynamics; th...

The development of new amine solvents without the major drawbacks of conventional amines is crucial to industrial applications of CO2 capture. This paper presents a water-lean CO2 capture solvent having a low regeneration energy and low degradation. The water-lean solvent, K2Sol, is a sterically hindered diamine; because of the hindered amine site, K2Sol easily forms bicarbonate, resulting in a high absorption capacity. The minimum solvent regeneration energy is obtained using Gaussian process Bayesian optimization (GPBO) and bench-scale pilot plant experiments. GPBO finds the optimal solution using the input and output relationship of experiments; thus, expensive first-principle model construction can be avoided. According to the pilot plant experiment, the optimal regeneration energies of monoethanolamine (MEA) and K2Sol are 4.3 and 2.8 GJ/t CO2, respectively, indicating that K2Sol requires only 65% of the regeneration energy of MEA. Fewer than 30 experiments are required to find the optimal pilot plant operation for both the MEA and K2Sol experiments. We also describe the superior properties of K2Sol in terms of the CO2 loading, cyclic capacity, regeneration temperature, and degradation.

We develop a rigorous mathematical model of aqueous mineral carbonation kinetics for carbon capture and utilization (CCU) and estimate the parameter posterior distribution using Bayesian parameter ...

The severe increase in the CO2 concentration is a causative factor of global warming, which accelerates the destruction of ecosystems. The massive utilization of CO2 for value-added chemical production is a key to commercialization to guarantee both economic feasibility and negative carbon emission. Although the electrochemical reduction of CO2 is one of the most promising technologies, there are remaining challenges for large-scale production. Herein, an overview of these limitations is provided in terms of devices, processes, and catalysts. Further, the economic feasibility of the technology is described in terms of individual processes such as reactions and separation. Additionally, for the practical implementation of the electrochemical CO2 conversion technology, stable electrocatalytic performances need to be addressed in terms of current density, Faradaic efficiency, and overpotential. Hence, the present review also covers the known degradation behaviors and mechanisms of electrocatalysts and electrodes during electrolysis. Furthermore, strategic approaches for overcoming the stability issues are introduced based on recent reports from various research areas involved in the electrocatalytic conversion.

As the scale and complexity of the chemical process increase, it is important to detect anomalies in the process at an early stage and respond in real-time. Currently, however, it is difficult for process operators to identify numerous alarms in the factory and to make a consistent and immediate abnormal diagnosis because each has different safety standards. To this end, this study proposed an adversarial autoencoder(AAE) based process monitoring model. AAE uses adversarial training to impose an arbitrary prior distribution on the latent vectors. In other words, the discriminator is trained to distinguish between the samples from the data distribution and the samples from the encoder, and the encoder is trained to match the latent vectors with a prior distribution. In the AAE-based process monitoring model, normal condition samples are used for train data and prior distribution is set up to be Gaussian distribution. T2 and SPE statistics are constructed in the feature space and residual space respectively to monitor the process. By employing AAE, the model learns a deep generative representation that maps the orignal data distribution. To demonstrate the performance of the proposed model, a case study using the Tennessee Eastman benchmark process is employed. False alarm rate (FAR) and false detection rate (FDR) are used as the assessment criteria to measure the monitoring performance.

Discovering new materials better suited to specific purposes is an important issue in improving the quality of human life. Here, a neural network that creates molecules that meet some desired multiple target conditions based on a deep understanding of chemical language is proposed (generative chemical Transformer, GCT). The attention mechanism in GCT allows a deeper understanding of molecular structures beyond the limitations of chemical language itself which cause semantic discontinuity by paying attention to characters sparsely. The significance of language models for inverse molecular design problems is investigated by quantitatively evaluating the quality of the generated molecules. GCT generates highly realistic chemical strings that satisfy both chemical and linguistic grammar rules. Molecules parsed from the generated strings simultaneously satisfy the multiple target properties and vary for a single condition set. These advances will contribute to improving the quality of human life by accelerating the process of desired material discovery.

Increasing wind and solar generation in power grids leads to more renewable power curtailments in some periods of time due to the fast and unpredictable variations of their outputs. The utilization of these sources for energy storage can unlock huge potential benefits. Therefore, aiming at minimizing the curtailments of renewable power from the viewpoint of an independent system operator (ISO), in this paper, we propose deep learning-driven optimal sizing and operation of alkaline water electrolyzers (AWE) and battery energy storage systems (BESS). For this purpose, a set of actual renewable power curtailment data of California ISO was fully investigated, and deep learning forecast methods were employed to determine the prediction error and its probability distribution function (PDF). Using the fitted PDF, a set of scenarios was generated and reduced to some accurate and probable ones. Consequently, a two-stage scenario-based stochastic model was proposed to determine the optimal planning of this system, and a penalty variable was defined in the second stage to maximize the utilization of curtailed renewable energy sources (RESs). The learning results showed that the prediction errors were minimized using the gated recurrent unit (GRU) method. It was also shown that 97% of curtailments were utilized using AWEs with annual costs of $233.55 million, which had 63.5% fewer costs than using BESSs. Furthermore, using AWEs reduced operational expenses by 89.1% compared with using BESSs, owing to their operational benefits.

Electrochemical reduction of O2 provides a clean and decentralized pathway to produce H2O2 compared to the current energy-intensive anthraquinone process. As the electrochemical reduction of O2 proceeds via either two-electron or four-electron path- way, it is thus essential to control the selectivity as well as to maximize the catalytic activity. Siahrostami et al. demonstrated a novel approach to control the reaction pathway by optimizing an adsorption ensemble to tune adsorption sites of reaction intermediates, and identified Pt-Hg catalysts from density functional theory (DFT) calculations and experimentally validated this catalyst (Nat. Mater. 2013, 12, 1137). Inspired by this concept, in this work, we apply a state-of-the-art high-throughput screening to develop O2 reduction catalyst for selective H2O2 production. Starting from Materials Project database, we evaluate activity, selectivity and electrochemical stability. To efficiently perform the screening, we introduce an active motif based approach which pre-screens unpromising materials and only performs DFT calculations for promising materials, which significantly reduce the number of the required calculations. We not only provide a list of promising candidates identified by DFT calculations, but also suggest element species to achieve high catalytic activity or H2O2 selectivity for future experimental attempts. Finally, we discuss a strategy for efficient future high-throughput screening using a machine learning pipeline consisting of a non-linear dimension reduction and a density-based clustering.

The electrochemical reduction of CO2 stores intermittent renewable energy in valuable raw materials, such as chemicals and transportation fuels, while minimizing carbon emissions and promoting carbon-neutral cycles. Recent technoeconomic reports suggested economically feasible target products of CO2 electroreduction and the relative influence of key performance parameters such as faradaic efficiency (FE), current density, and overpotential in the practical industrial-scale applications. Furthermore, fundamental factors, such as available reaction pathways, shared intermediates, competing hydrogen evolution reaction, scaling relations of the intermediate binding energies, and CO2 mass transport limitations, should be considered in relation to the electrochemical CO2 reduction performance. Intensive research efforts have been devoted to designing and developing advanced electrocatalysts and improving mechanistic understanding. More recently, the research focus was extended to the catalyst environment, because the interfacial region can delicately modulate the catalytic activity and provide effective solutions to challenges that were not fully addressed in the material development studies. Herein, we discuss the importance of catalyst-electrolyte interfaces in improving key operational parameters based on kinetic equations. Furthermore, we extensively review previous studies on controlling organic modulators, electrolyte ions, electrode structures, as well as the three-phase boundary at the catalyst-electrolyte interface. The interfacial region modulates the electrocatalytic properties via electronic modification, intermediate stabilization, proton delivery regulation, catalyst structure modification, reactant concentration control, and mass transport regulation. We discuss the current understanding of the catalyst-electrolyte interface and its effect on the CO2 electroreduction activity.

To reduce damage arising from accidents in chemical processing plants, detection of the incident must be rapid to mitigate the danger. In the case of the gas leaks, detectors are critical. To improve efficiency, leak detectors must be installed at locations after considering various factors like the characteristics of the workspace, processes involved, and potential consequences of the accidents. Thus, the consequences of potential accidents must be simulated. Among various approaches, computational fluid dynamics (CFD) is the most powerful tool to determine the consequences of gas leaks in industrial plants. However, the computational cost of CFD is large, making it prohibitively difficult and expensive to simulate many scenarios. Thus, a deep-neural-network-based surrogate model has been designed to mimic FLACS (FLame ACceleration Simulator), one of the most important programs in the modeling of gas leaks. Using the simulated results of a proposed surrogate model, a sensor allocation optimization problem was solved using mixed integer linear programming (MILP). The optimal solutions produced by the proposed surrogate model and FLACS were compared to verify the efficacy of the proposed surrogate model.

Design and operation of pilot-scale (1.0 BDP) compact GTL process comprising of reforming section, CO2 separating section, and Fischer-Tropsch (FT) synthesis section is presented. Detailed systematic computer-aided design procedure adopted to design a modular 0.5 BPD pilot-scale microchannel reactor used in the pilot plant operation is also presented. The modular microchannel FT reactor block design consists of 528 process channels and numerous coolant channels arranged in cross-cocurrent-cross configuration for adequate heat removal. On average 98.27% CH4 conversion to syngas in reforming section comprising of a pre-reformer unit and a tri-reformer unit, and CO2 separation rate of 36.75% along with CO/H2 reduction from 2.67 to 2.08 in CO2 membrane separation section were achieved from the entire pilot plant operation duration of 450 h. Parallel operation of FT microchannel reactor and multitubular fixed-bed type FT reactor for comparison showed that multitubular fixed-bed type reactor undergoes reaction runaway for the applied process conditions, while microchannel reactor showed adequate temperature control. Overall CO conversion of 83% and adequate temperature control at three different applied operating temperatures of 220 °C, 230 °C, and 240 °C subsequently during the 139 h FT reactor operation demonstrated the appreciable performance of the present microchannel FT reactor designed.

High-accuracy gas dispersion models are necessary for predicting toxic gas movement, and for reducing the damage caused by toxic gas release accidents in chemical processes. In urban areas, where obstacles are large and abundant, computational fluid dynamics (CFD) would be the best choice for simulating and analyzing scenarios of accidental release of toxic chemicals. However, owing to the large computation time required for CFD simulation, it is inappropriate in emergency situations and in real-time alarm systems. In this study, a non-linear surrogate model based on deep learning is proposed using a variational autoencoder with deep convolutional layers and a deep neural network with batch normalization (VAEDC-DNN) for real-time analysis of the probability of death (Pdeath). VAEDC can extract representation features of the Pdeath contour with complicated urban geometry in the latent space, and DNN maps the variable space into the latent space for the Pdeath image data. The chlorine gas leak accident in the Mipo complex (city of Ulsan, Republic of Korea) is used for verification of the model. The proposed model predicts the Pdeath image within a mean squared error of 0.00246, and compared with other models, it exhibits superior performance. Furthermore, through the smoothness of image transition in the variable space, it is confirmed that image generation is not overfitting by data memorization.