High-throughput Calculations Research Articles

Leveraging multiplexed metasurfaces for multi-task learning with all-optical diffractive processors

Abstract Diffractive Neural Networks (DNNs) leverage the power of light to enhance computational performance in machine learning, offering a pathway to high-speed, low-energy, and large-scale neural information processing. However, most existing DNN architectures are optimized for single tasks and thus lack the flexibility required for the simultaneous execution of multiple tasks within a unified artificial intelligence platform. In this work, we utilize the polarization and wavelength degrees of freedom of light to achieve optical multi-task identification using the MNIST, FMNIST, and KMNIST datasets. Employing bilayer cascaded metasurfaces, we construct dual-channel DNNs capable of simultaneously classifying two tasks, using polarization and wavelength multiplexing schemes through a meta-atom library. Numerical evaluations demonstrate performance accuracies comparable to those of individually trained single-channel, single-task DNNs. Extending this approach to three-task parallel recognition reveals an expected performance decline yet maintains satisfactory classification accuracies of greater than 80 % for all tasks. We further introduce a novel end-to-end joint optimization framework to redesign the three-task classifier, demonstrating substantial improvements over the meta-atom library design and offering the potential for future multi-channel DNN designs. Our study could pave the way for the development of ultrathin, high-speed, and high-throughput optical neural computing systems.

Chemical space-informed machine learning models for rapid predictions of x-ray photoelectron spectra of organic molecules

Abstract We present machine learning models based on kernel-ridge regression for predicting X-ray photoelectron spec- tra of organic molecules originating from the ionization energies of 1s electrons in carbon (C), nitrogen (N), oxygen (O), and fluorine (F) atoms. We constructed training dataset through high-throughput calculations of core-electron binding energies (CEBEs) for 12,880 small organic molecules in the bigQM7ω dataset, employing the ∆-SCF formalism coupled with meta-GGA-DFT and a variationally converged basis set. The models are cost-effective, as they require the atomic coordinates of a molecule generated using universal force fields while estimating the target-level CEBEs corresponding to DFT-level equilibrium geometry. We explore transfer learning by utilizing the atomic environment feature vectors learned using a graph neural network framework in kernel-ridge regression. Additionally, we enhance accuracy using the ∆-machine learning framework by leveraging inexpensive baseline spectra derived from Kohn–Sham eigenvalues. Upon application to 208 com- binatorially substituted uracil molecules, larger than those in the training set, our analyses reveal that while the models may not yield quantitatively accurate predictions of CEBEs on a molecule-by-molecule basis, they do exhibit a strong linear correlation, which proves valuable for virtual high-throughput screening purposes. We present the dataset and models as the Python module, cebeconf, to facilitate further explorations.

Maize Phenotypic Parameters Based on the Constrained Region Point Cloud Phenotyping Algorithm as a Developed Method

As one of the world’s most crucial food crops, maize plays a pivotal role in ensuring food security and driving economic growth. The diversification of maize variety breeding is significantly enhancing the cumulative benefits in these areas. Precise measurement of phenotypic data is pivotal for the selection and breeding of maize varieties in cultivation and production. However, in outdoor environments, conventional phenotyping methods, including point cloud processing techniques based on region growing algorithms and clustering segmentation, encounter significant challenges due to the low density and frequent loss of point cloud data. These issues substantially compromise measurement accuracy and computational efficiency. Consequently, this paper introduces a Constrained Region Point Cloud Phenotyping (CRPCP) algorithm that proficiently detects the phenotypic traits of multiple maize plants in sparse outdoor point cloud data. The CRPCP algorithm consists primarily of three core components: (1) a constrained region growth algorithm for effective segmentation of maize stem point clouds in complex backgrounds; (2) a radial basis interpolation technique to bridge gaps in point cloud data caused by environmental factors; and (3) a multi-level parallel decomposition strategy leveraging scene blocking and plant instances to enable high-throughput real-time computation. The results demonstrate that the CRPCP algorithm achieves a segmentation accuracy of 96.2%. When assessing maize plant height, the algorithm demonstrated a strong correlation with manual measurements, evidenced by a coefficient of determination R2 of 0.9534, a root mean square error (RMSE) of 0.4835 cm, and a mean absolute error (MAE) of 0.383 cm. In evaluating the diameter at breast height (DBH) of the plants, the algorithm yielded an R2 of 0.9407, an RMSE of 0.0368 cm, and an MAE of 0.031 cm. Compared to the PointNet point cloud segmentation method, the CRPCP algorithm reduced segmentation time by more than 44.7%. The CRPCP algorithm proposed in this paper enables efficient segmentation and precise phenotypic measurement of low-density maize multi-plant point cloud data in outdoor environments. This algorithm offers an automated, high-precision, and highly efficient solution for large-scale field phenotypic analysis, with broad applicability in precision breeding, agronomic management, and yield prediction.

Developing High-Performance Ceramics from Multicomponent Grain Boundary Entropy Design.

Grain boundary (GB) glassy phase often results in poor ceramic performances. Here, a Multicomponent Grain Boundary Entropy (MGBE) descriptor extracted from high-throughput first-principle calculations is proposed to capture the nature of high-entropy GB phases in ceramics. In a Si3N4 ceramic model system, MGBE is found to have a direct correlation with GB phase crystallinity, element segregation, and formation of pores. The predicted highest MGBE sintering additive combination (MgO-Y2O3-Er2O3-Yb2O3) leads to high-performance ceramics of homogenous microstructure and pure GB (YErYb)2Si3O3N4 phase without observable glassy film. Conversely, low MGBE additives result in a substantial amount of GB glassy phase, element segregation, and pore clusters. The MGBE descriptor can make a rapid screening of multicomponent sintering additives, offering a novel approach for rational designing of ceramics with targeted microstructure and performances.

Low-pressure superconductivity of Th-B-H ternary compounds: An insightful study of stable and metastable phases

The low-enthalpy structures of Th-B-H ternary compounds below 50 GPa are investigated by using an efficient framework that combines high-throughput screening, genetic algorithms, and first-principles calculations. Four thermodynamically stable hydrides (Th2BH11, Th2BH10, ThB2H10, and ThBH) and seven thermodynamically metastable hydrides (ThBH2, ThBH3, ThBH4, ThBH5, ThBH7, ThB2H4, and ThB3H3) are revealed. There are no stable thorium boron hydrides at ambient pressure. R3m-ThB2H10 and R3m-Th2BH11 are indirect bandgap semiconductors, while R3m-Th2BH10 and P6/mmm-ThBH show metallic nature. The relatively low λ values of R3m-Th2BH10 and P6/mmm-ThBH result in their Tc around 5.8 K and 0.1 K at 20 GPa, respectively. Two metastable phases, F4¯3m-ThBH5 and P63/mmc-ThBH7, exhibit excellent superconducting properties at low pressures. The values of Tc of F4¯3m-ThBH5 and P63/mmc-ThBH7 reach 46.9 K at 20 GPa and 40.0 K at 50 GPa, respectively. Our findings offer valuable insights for discovering and designing superconducting rare earth hydrides under low and even ambient pressures.

Physics-Informed Machine Learning with Data-Driven Equations for Predicting Organic Solar Cell Performance.

Organic solar cells (OSCs) have emerged as a promising solution in pursuing sustainable energy. This study presents a comprehensive approach to advancing OSC development by integrating data-driven equations from quantum mechanical (QM) descriptors with physics-informed machine learning (PIML) models. We circumvent traditional experimental limitations through high-throughput QM calculations, prioritizing transparent and interpretable models. Using the SISSO++ method, we identified key descriptors that effectively map the relationships between input variables and photovoltaic performance metrics. Our innovative predictive models, derived from SISSO outputs, excel in forecasting critical OSC parameters such as short-circuit current (JSC), open-circuit voltage (VOC), fill factor (FF), and power conversion efficiency (PCEmax), achieving high accuracy even with limited data sets. To validate our models' practical utility, we applied the PIML framework to a newly compiled data set of OSC devices, demonstrating their versatility and capability in pinpointing high-performance materials. This research underscores the strong predictive power of our models, bridging the gap between experimental results and theoretical predictions and making significant contributions to the advancement of sustainable energy technologies.

Low lattice thermal conductivity induced by rattling-like vibration in Zintl phase Na2CaCdSb2 compound with high ZT from two-channel model

Zintl phase compounds, characterized by the traits of a phonon-glass and electron-crystal, stand out as promising candidates for thermoelectric applications. This study delves into the thermoelectric properties of Zintl phase Na2CaCdSb2 compound through a synergy of first-principles calculations, machine learning interatomic potential approach based on the two-channel model, and high-throughput screening calculations. With an indirect bandgap of 1.02 eV, Na2CaCdSb2 compoundpossesses an high carrier mobility. The high valence band degeneracy and anisotropic band (light-heavy band) lead to a high power factor for the p-type Na2CaCdSb2 compound. The strong fourth-order anharmonicity of the rattling-like Na atom manifests the low lattice thermal conductivity on the basis of two-channel model. The loosely bonding Na atom displays random rotations and dynamically disordered orientations, leading to the weakly disordered nature of Na2CaCdSb2 compound. The unique crystal structure of Na2CaCdSb2 compound coupled with the abundance of diffusion-like phonons play a pivotal role in the phonon transport. Considering the low lattice thermal conductivity and high power factor with the framework of two-channel model, the n-type and p-type Na2CaCdSb2 compounds exhibit optimal figure-of-merits (ZT) of 1.0 and 2.0 at 600 K, showcasing excellent thermoelectric performance. Meanwhile, an optimal ZT of 2.3 is achieved for p-type Na2CaCdSb2 compound at the same temperature along the x-direction. Our present study not only provides fundamental insights into the thermal and electronic transport properties of Na2CaCdSb2 compound under the two-channel model in combination with the four-phonon scattering and multiple carrier scattering rates, but also rationalizes the high-order phonon interactions and thermal transport properties under two-channel model.

Machine learning assisted screening of two dimensional chalcogenide ferromagnetic materials with Dzyaloshinskii Moriya interaction

Magnetic skyrmions are potential candidates for high-density storage and logic devices because of their inherent topological stability and nanoscale size. Two-dimensional (2D) Janus transition metal chalcogenides (TMDs) are widely used to induce skyrmions due to the breaking of inversion symmetry. However, the experimental synthesis of Janus TMDs is rare, which indicates that the Janus configuration might not be the most stable MXY structure. Here, through machine-learning-assisted high-throughput first-principles calculations, we demonstrate that not all MXY compounds can be stabilized in Janus layered structure and a large proportion prefer to form other configurations with lower energy than the Janus configuration. Interestingly, these new configurations exhibit a strong Dzyaloshinskii–Moriya interaction (DMI), which can generate and stabilize skyrmions even under a strong magnetic field. This work provides not only an efficient method for obtaining ferromagnetic materials with strong DMI but also a theoretical guidance for the synthesis of TMDs via experiments.

High-Throughput Computational Study and Machine Learning Prediction of Electronic Properties in Transition Metal Dichalcogenide/Two-Dimensional Layered Halide Perovskite Heterostructures.

Heterostructures formed by transition metal dichalcogenides (TMDs) and two-dimensional layered halide perovskites (2D-LHPs) have attracted significant attention due to their unique optoelectronic properties. However, theoretical studies face challenges due to the large number of atoms and the need for lattice matching. With the discovery of more 2D-LHPs, there is an urgent need for methods to rapidly predict and screen TMDs/2D-LHPs heterostructures. This study employs first-principles calculations to perform high-throughput computations on 602 TMDs/2D-LHPs heterostructures. Results show that different combinations exhibit diverse band alignments, with MoS2 and WS2 more likely to form type-II heterostructures with 2D-LHPs. The highest photoelectric conversion efficiency of type-II structures reaches 23.26%, demonstrating potential applications in solar cells. Notably, some MoS2/2D-LHPs form type-S structures, showing promise in photocatalysis. Furthermore, we found that TMDs can significantly affect the conformation of organic molecules in 2D-LHPs, thus modulating the electronic properties of the heterostructures. To overcome computational cost limitations, we constructed a crystal graph convolutional neural network model based on the calculated data to predict the electronic properties of TMDs/2D-LHPs heterostructures. Using this model, we predicted the bandgaps and band alignment types of 9,360 TMDs/2D-LHPs heterostructures, providing a comprehensive theoretical reference for research in this field.

Predictive Modeling of High-Entropy Alloys and Amorphous Metallic Alloys Using Machine Learning.

High entropy alloys and amorphous metallic alloys represent two distinct classes of advanced alloy materials, each with unique structural characteristics. Their emergence has garnered considerable interest across the materials science and engineering communities, driven by their promising properties, including exceptional strength. However, their extensive compositional diversity poses substantial challenges for systematic exploration, as traditional experimental approaches and high-throughput calculations struggle to efficiently navigate this vast space. While the recent development in data-driven materials discovery could potentially help, such efforts are hindered by the scarcity of comprehensive data and the lack of robust predictive tools that can effectively link alloy composition with specific properties. To address these challenges, we have deployed a machine-learning-based workflow for feature selection and statistical analysis to afford predictive models that accelerate the data-driven discovery and optimization of these advanced materials. Our methodology is validated through two case studies: (i) a regression analysis of the bulk modulus, and (ii) a classification analysis based on glass-forming ability. The Bayesian-optimized regression model trained for the prediction of bulk modulus achieved an R2 of 0.969, an mean absolute error (MAE) of 3.958 GPa, and an root mean square error (RMSE) of 5.411 GPa, while our classification model for predicting glass-forming ability achieved an F1-score of 0.91, an area-under-the-curve of the receiver-operating-characteristic curve of 0.98, and an accuracy of 0.91. Furthermore, by leveraging a wide array of chemical data from diverse literature sources, we have successfully predicted a broad range of properties. This success underscores the efficacy of our modeling approach and emphasizes the importance of a comprehensive feature analysis and judicious feature selection strategy over a mere reliance on complex modeling techniques.

A high-throughput screening of catalyst for efficient nitrogen fixation: Transition metal single-atom anchored on an emerging synthesized biphenyl network

Synthesizing efficient and selective green ideal catalysts has been an increasingly critical, yet unsolved, issue for the ammonia synthesis industry, hindering the growing global demand for environmental protection and energy efficiency. Electrocatalytic nitrogen reduction reaction (eNRR) is a promising technology for low-energy ammonia synthesis. However, designing efficient electrocatalysts for NRR remains challenging. The emergence of graphene-like substrates offers exciting prospects for addressing this challenge and facilitating single-atom catalysts in eNRR. Here, we report the innovative selection of a recently synthesized two-dimensional biphenyl network (2D BPN) compound as a substrate. Its excellent conductivity and porosity enable stable transition metal atoms (TMs) support for constructing eNRR electrocatalysts. we evaluated the feasibility of 23 TMs anchored on BPN for eNRR by high-throughput first-principles calculations. Through a systematic five-step strategy, we identified several single-atom catalysts (SACs) with potential for eNRR, including Mo@BPN, V@BPN, W@BPN, and Re@BPN. Among them, Mo@BPN exhibited the best balance in the adsorption of key reaction intermediates (e.g., N2H and NH3) and demonstrated a low limiting potential (-0.37 V). In addition, the underlying mechanism of NRR activity was elucidated by analyzing the extrinsic patterns revealed through the screened catalysts. A triangular volcano diagram, incorporating the initial protonation step, adsorption free energy, and final protonation step, revealed the NRR activity trend. Overall, this study provides a solid theoretical foundation and valuable guidance for future experimental exploration of efficient electrocatalysts for ammonia synthesis on BPN. Crucial insights into the theoretical design of efficient electrocatalysts are also offered.

Striving for clarity in language about gene expression.

What do we mean when we say 'gene expression'? In the decades following Crick's 1958 central dogma of molecular biology, whereby genetic information flows from DNA (genes) to RNA (transcripts) to protein (products), we have learned a great deal about DNA, RNA, proteins, and the ensuing phenotypic changes. With the advent of high-throughput technologies (1990s), molecular biologists and computer scientists forged critical collaborations to understand the vast amount of data being generated, rapidly escalating gene expression research to the 'omics' level: entire sets of genes (genomes), transcribed RNAs (transcriptomes), and synthesized proteins (proteomes). However, some concessions came to be made for molecular biologists and computer scientists to understand each other-one of the most prevalent being the increasingly widespread use of 'gene' to mean 'RNAs originating from a DNA segment'. This loosening of terminology, we will argue, creates ambiguity and confusion. We propose guidelines to increase precision and clarity when communicating about gene expression, most notably to reserve 'gene' for the DNA template and 'transcript' for the RNA transcribed from that gene. Striving to use perspicuous terminology will promote rigorous gene expression science and accelerate discovery in this highly promising area of biology.

Electrochemical stability of biodegradable Zn-Cu alloys through machine-learning accelerated high-throughput discovery.

Zn-Cu alloys have attracted great attention as biodegradable alloys owing to their excellent mechanical properties and biocompatibility, with corrosion characteristics being crucial for their suitability for biomedical applications. However, the unresolved identification of intermetallic compounds in Zn-Cu alloys affecting corrosion and the complexity of the application environment hamper the understanding of their electrochemical behavior. Utilizing high-throughput first-principles calculations and machine-learning accelerated evolutionary algorithms for screening the most stable compounds in Zn-Cu systems, a dataset encompassing the formation energy of 2033 compounds is generated. It reveals that most of the experimentally reported Zn-Cu compounds can be replicated, especially the structure of R32 CuZn5 is first discovered which possesses the lowest formation energy of -0.050 eV per atom. Furthermore, the simulated X-ray diffraction pattern matches perfectly with the experimental ones. By formulating 342 potential electrochemical reactions based on the binary compounds, the Pourbaix diagrams for Zn-Cu alloys are constructed to clarify the fundamental competition between different phases and ions. The calculated equilibrium potential of CuZn5 is higher than that of Zn through the forward reaction Zn + CuZn5 ⇌ CuZn5 + Zn2+ + 2e-, resulting in microcell formation owing to the stronger charge density localization in Zn compared to CuZn5. The presence of chlorine accelerates the corrosion of Zn through the reaction Zn + CuZn5 + 6Cl- + 6H2O ⇌ Cu + 6ZnOHCl + 6H+ + 12e-, where the formation of ZnOHCl disrupts the ZnO passive film and expands the corrosion pH range from 9.2 to 8.8. Our findings reveal an accurate quantitative corrosion mechanism for Zn-Cu alloys, providing an effective pathway to investigate the corrosion resistance of biodegradable alloys.

Unidirectional and hierarchical on-chip interconnected architecture for large-scale hardware spiking neural networks

Spiking Neural Networks (SNNs) exhibit the strong capability to address spatiotemporal dynamic problems. Recent research has explored the hardware SNN systems to solve the spatiotemporal problems in real-time. The Network-on-Chip (NoC) is an effective scheme for building large-scale hardware SNNs. However, for the existing NoC-based hardware SNNs, large area overhead and hardware power are consumed by their interconnections, because of complex topologies and router structures. Therefore, in this work a novel Unidirectional and Hierarchical on-Chip Interconnected Architecture (UHCIA) is proposed to address this problem. The proposed UHCIA mainly combines the novel hybrid topology of unidirectional multiple loops and rings, and uses a deflection router technique. Experimental results show that compared to other works, the UHCIA achieves ∼23.6X of area reduction and ∼6.4X of power reduction, with high system throughput and biological real-time computations.

Discovery of promising Cu-based catalysts for selective propane dehydrogenation to propylene assisted by high-throughput computations

The development of efficient catalysts with high selectivity is a major challenge for propane dehydrogenation (PDH), which is an on-purpose route to produce propylene. We have developed a high-throughput computational strategy to screen out the promising catalysts for propane dehydrogenation to propylene on Cu-based metal alloys. A total of 1500 candidates of Cu3M1 alloys are constructed, and four efficient descriptors are applied in our high-throughput screening to evaluate the activity and selectivity of catalysts. Four Cu-based alloys of Cu3Cr1, Cu3Pd1, Cu3Pt1, and Cu3Ag1 are identified as promising PDH catalysts, and their superior selectivity towards propylene could be explained by the synergistic effect of Cu and M atoms. Our theoretical screening method could assist the new catalyst design and accelerate the discovery of environment-friendly alloy catalysts for PDH.

Sustainable H2O2 production via solution plasma catalysis

Clean production of hydrogen peroxide (H2O2) with water, oxygen, and renewable energy is considered an important green synthesis route, offering a valuable substitute for the traditional anthraquinone method. Currently, renewable energy-driven production of H2O2 mostly relies on soluble additives, such as electrolytes and sacrificial agents, inevitably compromising the purity and sustainability of H2O2. Herein, we develop a solution plasma catalysis technique that eliminates the need for soluble additives, enabling eco-friendly production of concentrated H2O2 directly from water and O2. Screening over 40 catalysts demonstrates the superior catalytic performance of carbon nitride interacting with discharge plasma in water. High-throughput density functional theory calculations for 68 models, along with machine learning using 29 descriptors, identify cyano carbon nitride (CCN) as the most efficient catalyst. Solution plasma catalysis with the CCN achieves concentrated H2O2 of 20 mmol L-1, two orders of magnitude higher than photocatalysis by the same catalyst. Plasma diagnostics, isotope labeling, and COMSOL simulations collectively validate that the interplay of solution plasma and the CCN accounts for the significantly increased production of singlet oxygen and H2O2 thereafter. Our findings offer an efficient and sustainable pathway for H2O2 production, promising wide-ranging applications across the chemical industry, public health, and environmental remediation.

Machine learning enables the discovery of 2D Invar and anti-Invar monolayers

Materials demonstrating positive thermal expansion (PTE) or negative thermal expansion (NTE) are quite common, whereas those exhibiting zero thermal expansion (ZTE) are notably scarce. In this work, we identify the mechanical descriptors, namely in-plane tensile stiffness and out-of-plane bending stiffness, that can effectively classify PTE and NTE 2D crystals. By utilizing high throughput calculations and the state-of-the-art symbolic regression method, these descriptors aid in the discovery of ZTE or 2D Invar monolayers with the linear thermal expansion coefficient (LTEC) within ±2 × 10−6 K−1 in the middle range of temperatures. Additionally, the descriptors assist the discovery of large PTE and NTE 2D monolayers with the LTEC larger than ±15 × 10−6 K−1, which are so-called 2D anti-Invar monolayers. Advancing our understanding of materials with exceptionally low or high thermal expansion is of substantial scientific and technological interest, particularly in the development of next-generation electronics at the nanometer or even Ångstrom scale.

Dxtb-An efficient and fully differentiable framework for extended tight-binding.

Automatic differentiation (AD) emerged as an integral part of machine learning, accelerating model development by enabling gradient-based optimization without explicit analytical derivatives. Recently, the benefits of AD and computing arbitrary-order derivatives with respect to any variable were also recognized in the field of quantum chemistry. In this work, we present dxtb-an open-source, fully differentiable framework for semiempirical extended tight-binding (xTB) methods. Developed entirely in Python and leveraging PyTorch for array operations, dxtb facilitates extensibility and rapid prototyping while maintaining computational efficiency. Through comprehensive code vectorization and optimization, we essentially reach the speed of compiled xTB programs for high-throughput calculations of small molecules. The excellent performance also scales to large systems, and batch operability yields additional benefits for execution on parallel hardware. In particular, energy evaluations are on par with existing programs, whereas the speed of automatically differentiated nuclear derivatives is only 2 to 5 times slower compared to their analytical counterparts. We showcase the utility of AD in dxtb by calculating various molecular and spectroscopic properties, highlighting its capacity to enhance and simplify such evaluations. Furthermore, the framework streamlines optimization tasks and offers seamless integration of semiempirical quantum chemistry in machine learning, paving the way for physics-inspired end-to-end differentiable models. Ultimately, dxtb aims to further advance the capabilities of semiempirical methods, providing an extensible foundation for future developments and hybrid machine learning applications. The framework is accessible at https://github.com/grimme-lab/dxtb.

Computational Modeling and Machine Learning Approaches for Accelerated Materials Design

A substantial part of research in energy storage and conversion systems concerns the discovery and characterization of novel materials with improved performance, but most advancements are still generally attributed to costly and time-consuming trial-and-error experimentation. Computational chemistry methods combined with machine learning techniques offer paradigm shift in how materials are fundamentally understood and designed. Namely, high-throughput calculations combined with machine learning can help evaluate different properties of complex materials and efficiently screen millions of candidates to identify the ones with improved targeted properties, such as higher activity and selectivity and/or enhanced durability.In this talk we will first illustrate the use of density functional theory (DFT) in the design of fuel cell and electrolyzer ionomers. More specifically, we will discuss how simple descriptors such as DFT-calculated adsorption energies and adsorption modes of polymer fragments can be used to guide the rational design of ionomers for anion exchange membrane fuel cells and alkaline membrane water electrolyzers. Our design principle has led to the synthesis of new polymers, e.g., poly(fluorene) and polynorbornene-based ionomers, with weak adsorption on the electrocatalyst surface and improved hydrogen oxidation reaction activity. We will further introduce the computational workflow that integrates high-throughput DFT calculations with machine learning with a goal of designing new amine-based CO2 adsorbents, as well as platinum group metal-free electrocatalysts for conversion of CO2 to value added products. In our workflow we use descriptors together with DFT-calculated binding energetics to train artificial neural networks for activity prediction. Our trained machine learning models are then used to screen millions of candidate chemistries and identify the ones with optimal activity. Descriptor importance evaluation provides additional design principles based on the structure to property relationships for the synthesis of targeted materials for CO2 capture and electro-conversion.

(Invited) Tuning Selective CO2 Electrohydrogenation Under Mid Temperature and Pressure

To address rising environmental concerns and energy challenges, it is highly desirable to develop green technologies for a sustainable future. Hydrogenation reactions are essential processes in the chemical industry, giving access to a variety of valuable compounds. Electrochemical CO2 hydrogenation using renewable electricity is considered one of the most promising pathways to reach the environmental economics. Among various electrochemical devices, a solid state electrolyzer working at intermediate temperatures has the advantages of high reaction rates and low overpotentials. As CO2 molecules is quite stable, it is critical to develop electrocatalysts with high activity to reduce electricity engagement. Also, CO2 hydrogenation has multiple pathways which normally results in low selectivity for a target product. A tunable electrocatalyst with high selectivity to different products is desirable for CO2 conversion reactions. Heterostructured nanomaterials attracts great attentions in electrochemical systems. With careful design, they can show very high activity and selectivity towards an electrochemical reaction pathway. In this work, we combine theoretical simulations, including density functional theory (DFT) and ab initio molecular dynamics (AIMD) calculations, with experimental methods (characterizations and electrochemical measurements) for rational design of highly efficient heterostructured nanomaterials for CO2 hydrogenations. We studied the strong metal-support interaction (SMSI) in a doped CeO2 supported noble metal nanoparticles (NPs) system using theoretical calculations. The results indicate that SMSI largely depends on chemical conditions of the support and particle size of dispersed metal NPs. By tuning the SMSI, we can successfully change the catalytic activity and selectivity towards CO2 hydrogenation reactions. Based on these findings, we designed tunable heterostructured nanomaterials for efficient hydrogenation reactions at intermediate temperatures. These predictions were further confirmed by experimental method. By using a combination of high-throughput theoretical calculations and electrochemical measurements, we successfully developed highly active catalysts for electrochemical CO2 hydrogenation reactions. This framework is also applicable to other electrochemical systems using heterostructured materials.