Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “KNOWLEDGE BASE”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 919 records · Page 51
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Real-time dynamic wavelength tuning and intensity modulation of metal-clad nanolasers

To realize ubiquitously used photonic integrated circuits, on-chip nanoscale sources are essential components. Subwavelength nanolasers, especially those based on a metal-clad design, already possess many desirable attributes for an on-chip source such as low thresholds, room-temperature operation and ultra-small footprints accompanied by electromagnetic isolation at pitch sizes down to ∼50 nm. Another valuable characteristic for a source would be control over its emission wavelength and intensity in real-time. Most efforts on tuning/modulation thus far report static changes based on irreversible techniques not suited for high-speed operation. In this study, we demonstrate in-situ dynamical tuning of the emission wavelength of a metallo-dielectric nanolaser at room temperature by applying an external DC electric field. Using an AC electric field, we show that it is also possible to modulate the output intensity of the nanolaser at high speeds. The nanolaser’s emission wavelength in the telecom band can be altered by as much as 8.35 nm with a tuning sensitivity of ∼1.01 nm/V. Additionally, the output intensity can be attenuated by up to 89%, a contrast sufficient for digital data communication purposes. Finally, we achieve an intensity modulation speed up to 400 MHz, limited only by the photodetector bandwidth used in this study, which underlines the capability of high-speed operation via this method. This is the first demonstration of a telecom band nanolaser source with dynamic spectral tuning and intensity modulation based on an external E-field to the best of our knowledge.

Deka, Suruj S. (ORCID:0000000297812317)↗
DOE OSTI2021

Assessing Teleconnections-Induced Predictability of Regional Water Cycle on Seasonal to Decadal Timescales Using Machine Learning Approaches

Focal Area: (3) Insight gleaned from complex data (both observed and simulated) using AI, big data analytics, and other advanced methods, including explainable AI and physics- or knowledge-guided AI. Science Challenge: Predictive understanding of the regional water cycle based on teleconnections of climate modes of variability using AI methods.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Summer 2021 Internship Report

The internship was a 12 week program which started on 7th of June 2021. During this internship program, I worked as a Research Intern on a project which focused on feature extraction from scientific PDF documents using open-source, machine-learning-based tools.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
DOE OSTI2021

Predictive Data-driven Platform for Subsurface Energy Production

Subsurface energy activities such as unconventional resource recovery, enhanced geothermal energy systems, and geologic carbon storage require fast and reliable methods to account for complex, multiphysical processes in heterogeneous fractured and porous media. Although reservoir simulation is considered the industry standard for simulating these subsurface systems with injection and/or extraction operations, reservoir simulation requires spatio-temporal “Big Data” into the simulation model, which is typically a major challenge during model development and computational phase. In this work, we developed and applied various deep neural network-based approaches to (1) process multiscale image segmentation, (2) generate ensemble members of drainage networks, flow channels, and porous media using deep convolutional generative adversarial network, (3) construct multiple hybrid neural networks such as convolutional LSTM and convolutional neural network-LSTM to develop fast and accurate reduced order models for shale gas extraction, and (4) physics-informed neural network and deep Q-learning for flow and energy production. We hypothesized that physicsbased machine learning/deep learning can overcome the shortcomings of traditional machine learning methods where data-driven models have faltered beyond the data and physical conditions used for training and validation. We improved and developed novel approaches to demonstrate that physics-based ML can allow us to incorporate physical constraints (e.g., scientific domain knowledge) into ML framework. Outcomes of this project will be readily applicable for many energy and national security problems that are particularly defined by multiscale features and network systems.

58 GEOSCIENCES↗
DOE OSTI2022

Interplay of Magnetism and Superconductivity in van der Waals Heterostructures (Final report)

This project has investigated the magnetic and electric properties of 2D materials, with emphasis on the (anti-)ferromagnetic properties of 2D crystalline magnets and a new concept for stacking engineering of 2D ferroelectricity. The new basic physics fundamental knowledge obtained through this research may have important implications for future technologies based on magnetism and ferroelectricity, for example spintronics, quantum information science, or neuromorphic computing.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2021

Machine Learning-Based Classification of Lignocellulosic Biomass from Pyrolysis-Molecular Beam Mass Spectrometry Data

High-throughput analysis of biomass is necessary to ensure consistent and uniform feedstocks for agricultural and bioenergy applications and is needed to inform genomics and systems biology models. Pyrolysis followed by mass spectrometry such as molecular beam mass spectrometry (py-MBMS) analyses are becoming increasingly popular for the rapid analysis of biomass cell wall composition and typically require the use of different data analysis tools depending on the need and application. Here, the authors report the py-MBMS analysis of several types of lignocellulosic biomass to gain an understanding of spectral patterns and variation with associated biomass composition and use machine learning approaches to classify, differentiate, and predict biomass types on the basis of py-MBMS spectra. Py-MBMS spectra were also corrected for instrumental variance using generalized linear modeling (GLM) based on the use of select ions relative abundances as spike-in controls. Machine learning classification algorithms e.g., random forest, k-nearest neighbor, decision tree, Gaussian Naïve Bayes, gradient boosting, and multilayer perceptron classifiers were used. The k-nearest neighbors (k-NN) classifier generally performed the best for classifications using raw spectral data, and the decision tree classifier performed the worst. After normalization of spectra to account for instrumental variance, all the classifiers had comparable and generally acceptable performance for predicting the biomass types, although the k-NN and decision tree classifiers were not as accurate for prediction of specific sample types. Gaussian Naïve Bayes (GNB) and extreme gradient boosting (XGB) classifiers performed better than the k-NN and the decision tree classifiers for the prediction of biomass mixtures. The data analysis workflow reported here could be applied and extended for comparison of biomass samples of varying types, species, phenotypes, and/or genotypes or subjected to different treatments, environments, etc. to further elucidate the sources of spectral variance, patterns, and to infer compositional information based on spectral analysis, particularly for analysis of data without a priori knowledge of the feedstock composition or identity.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2022

DERMS Online: A New Voltage Sensitivity-Enabled Feedback Optimization Framework: Preprint

This paper proposes a distributed energy resource management system (DERMS) solution by developing a new voltage sensitivity enabled feedback optimization framework. The key idea is to adopt a measurement feedback scheme to reformulate the original nonlinear optimization into a linear programming (LP) problem via perturb and observe-based voltage sensitivity analysis. The proposed solution eliminates the dependence on load knowledge and can be implemented online thanks to an efficient open-source solver for LP problems. The proposed DERMS online platform is generalizable to deal with various types of distributed energy resources (DERs), including distributed photovoltaics (PVs), energy storage, electric vehicles, demand response, etc. Comparison results with other control methods on a realistic distribution feeder in Southern California highlight the feasibility as well as benefits for the proposed framework.

distributed energy resources management↗
DOE OSTI2019

Aluminum Clad SNF: Technical Considerations and Challenges for Extended (>50 years) Dry Storage - 1900y

The U.S. Department of Energy (DOE) manages a significant inventory (>13 MTHM) of aluminum clad spent nuclear fuel (ASNF) using a combination of wet and dry storage at three primary sites; Hanford, Idaho National Laboratory (INL), and Savannah River Site (SRS). The majority of this ASNF is generated from U.S. and international research test reactors (RTR), examples include Advanced Test Reactor (ATR) at INL and High-Flux Isotope Reactor (HFIR) at Oak Ridge National Laboratory (ORNL). The Obama Administration suspension of the Yucca Mountain repository, issuance of the January 2013 DOE 'Strategy for the Management and Disposal of Used Nuclear Fuel and High-Level Radioactive Waste,' delays in implementing the 'regionalization' SNF approach recommended in the 1995 'Department of Energy Programmatic Spent Nuclear Fuel Management and Idaho National Engineering Laboratory Environmental Restoration and Waste Management Programs Final Environmental Impact Statement,' and limited ASNF processing capability at SRS H-Canyon means that interim storage has become the de-facto strategy for managing ASNF. The 1995 Idaho Settlement Agreement requires the DOE to transfer all SNF stored at INL to interim dry storage prior to December 31, 2023 and wet storage capacity at the SRS is nearing capacity. This situation was recognized by the DOE Spent Nuclear Fuel Working Group (SNFWG) and at its June 2016 meeting formed a ASNF sub-working group comprised subject matter experts to evaluate this area and provide a report of findings. The ASNF sub-working group met many times over several month period to review the engineering and technical basis of existing and planned DOE ASNF dry storage facilities and additional studies on the behavior of ASNF under different conditions, including during reactor service. The ASNF sub-working group focused its review on RTR SNF. Based on these reviews, the ASNF sub-working group identified five knowledge gaps and technical data needs that should be addressed to ensure viability and safety of extended, greater than 50 years, sealed or vented interim dry storage of ASNF. These five areas include; behavior/chemistry of aluminum oxyhydroxide layer in dry storage, radiolytic gas generation properties of aluminum oxyhydroxide layers, effect of episodic breathing and radiolytic generation of corrosive gases during storage, performance of ASNF in existing dry storage systems, and effects of high temperature (>1000 C) drying on chemistry and behavior of aluminum oxyhydroxide layers. The ASNF sub-working group also recommended development of an action plan to identify technical and engineering activities needed to address the knowledge gaps and technical data needs. Results of the ASNF sub-working group are documented in 2017 report 'Aluminum Clad Spent Nuclear Fuel: Technical Considerations and Challenges for Extended (>50 years) Dry Storage,' DOE-ID/RPT-1575. The INL assembled a team comprised of twenty INL and Savannah River National Laboratory SNF management and subject matter experts to develop this action plan. Over the course of 2 months the team worked to develop the action plan. The plan identified and recommended six technical and engineering tasks needed to address the previously identified knowledge gaps and technical data needs. DOE Environmental Management Office of Technology Development is providing funding to the INL to execute these tasks. This paper will present and discuss the results of the ASNF sub-working group, tasks identified in the action plan, and experimental activities underway at INL and SRNL to address the action plan. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2026

Active space selection with self-healing diffusion Monte Carlo algorithms for periodic solids

Multideterminant Diffusion Monte Carlo (DMC) displays improved accuracy over single determinant DMC. Self-Healing Diffusion Monte Carlo (SHDMC) is a DMC based method that iteratively improves a multideterminant trial wavefunction. Although configuration interaction or complete active space (CAS) methods are very accurate and computationally feasible for many systems, they are not optimal for application to solids. SHDMC is accurate and designed for application to solids, so developing SHDMC based active space selection algorithms is a worthy endeavor. Here, we present and compare active space selection algorithms that are designed for use in conjunction with SHDMC, without relying on external approaches. For benchmarking, we calculated the ground state energy of a small unit cell of graphene and compared the results with a complete basis set extrapolated selected CI and a reference SHDMC trajectory. We found that systematically expanding the active space using an “auto-branching” algorithm optimally balances accuracy with computational practicality. To the best of our knowledge, this is the first work that demonstrates completely self-contained DMC-based active space selection algorithms that do not depend on external methods for determinant selection.

Spanedda, Nicole [ORNL]↗
DOE OSTI2022

A behavior and physiology-based decision support tool to predict thermal comfort and stress in non-pregnant, mid-gestation, and late-gestation sows

Background: Although thermal indices have been proposed for swine, none to our knowledge differentiate by reproductive stage or predict thermal comfort using behavioral and physiological data. The study objective was to develop a behavior and physiology-based decision support tool to predict thermal comfort and stress in multiparous (3.28 ± 0.81) non-pregnant (n = 11), mid-gestation (n = 13), and late-gestation (n = 12) sows. Results: Regression analyses were performed using PROC MIXED in SAS 9.4 to determine the optimal environmental indicator [dry bulb temperature (T DB ) and dew point] of heat stress (HS) in non-pregnant, mid-gestation, and late-gestation sows with respiration rate (RR) and body temperature (T B ) successively used as the dependent variable in a cubic function. A linear relationship was observed for skin temperature (T S ) indicating that TDB rather than the sow HS response impacted TS and so TS was excluded from further analyses. Reproductive stage was significant for all analyses (P < 0.05). Heat stress thresholds for each reproductive stage were calculated using the inflections points of RR for mild HS and T B for moderate and severe HS. Mild HS inflection points differed for non-pregnant, mid-gestation, and late gestation sows and occurred at 25.5, 25.1, and 24.0 °C, respectively. Moderate HS inflection points differed for non-pregnant, mid-gestation, and late gestation sows and occurred at 28.1, 27.8, and 25.5 °C, respectively. Severe HS inflection points were similar for non-pregnant and mid-gestation sows (32.9 °C) but differed for late-gestation sows (30.8 °C). These data were integrated with previously collected behavioral thermal preference data to estimate the T DB that non-pregnant, mid-gestation, and late-gestation sows found to be cool (T DB < T DB preference range), comfortable (T DB = T DB preference range), and warm (T DB preference range < T DB < mild HS). Conclusions: The results of this study provide valuable information about thermal comfort and thermal stress thresholds in sows at three reproductive stages. The development of a behavior and physiology-based decision support tool to predict thermal comfort and stress in non-pregnant, mid-gestation, and late-gestation sows is expected to provide swine producers with a more accurate means of managing sow environments.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

A Generation-Storage Coordination Dispatch Strategy for Power System Based on Causal Reinforcement Learning

In the backdrop of global energy transformation, power systems integrating high proportions of renewable energy sources are facing unprecedented challenges in operational stability and dispatch efficiency. To address these challenges, this study introduces a generation-storage coordination real-time dispatch strategy based on Causal Power System Dynamic Reinforcement Learning (CPSDRL). Diverging from traditional reinforcement learning approaches, CPSDRL innovatively incorporates causal inference within the state prediction model - the crux of model-based reinforcement learning - thereby establishing the Power Causal Dynamic Model (PCDM). Assisted by the prior knowledge of power systems, the model significantly enhances prediction accuracy and reliability through a two-stage training process. Utilizing PCDM, this study further applies a direct policy search algorithm to optimize the real-time dispatch strategy. Experimental results indicate that the proposed method improves the stability of generation-storage coordination real-time dispatch and exhibits competitive advantages in sample efficiency and computational speed, compared to traditional model-based and model-free reinforcement learning algorithms. This method is expected to enhance the practicality and adaptability of causal reinforcement learning techniques in power system scheduling and control.

causal reinforcement learning↗
DOE OSTI2023

Ultrafiltration separation of Am(VI)-polyoxometalate from lanthanides

Partitioning of americium from lanthanides (Ln) present in used nuclear fuel plays a key role in the sustainable development of nuclear energy. This task is extremely challenging because thermodynamically stable Am(III) and Ln(III) ions have nearly identical ionic radii and coordination chemistry. Oxidization of Am(III) to Am(VI) produces AmO 2 2+ ions distinct with Ln(III) ions, which has the potential to facilitate separations in principle. However, the rapid reduction of Am(VI) back to Am(III) by radiolysis products and organic reagents required for the traditional separation protocols including solvent and solid extractions hampers practical redox-based separations. Herein, we report a nanoscale polyoxometalate (POM) cluster with a vacancy site compatible with the selective coordination of hexavalent actinides ( 238 U, 237 Np, 242 Pu and 243 Am) over trivalent lanthanides in nitric acid media. To our knowledge, this cluster is the most stable Am(VI) species in aqueous media observed so far. Ultrafiltration-based separation of nanoscale Am(VI)-POM clusters from hydrated lanthanide ions by commercially available, fine-pored membranes enables the development of a once-through americium/lanthanide separation strategy that is highly efficient and rapid, does not involve any organic components and requires minimal energy input.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2025

Ground-Active Arthropod Diversity Under Energycane and Biomass Sorghum Production

Energycane and biomass sorghum are two of the most promising cellulosic energy crops in the southeastern US. Research on these two energy crops has focused mainly on biomass production, and there is a lack of knowledge on their ability to promote biodiversity and ecosystem services. This paper presents results from a comprehensive study on ground-active arthropod diversity in seven sites across five states in the southeastern US (Florida, Georgia, Louisiana, Mississippi, and Texas). Pitfall traps were deployed four times during each crop season for energycane, biomass sorghum, and a local reference conventional crop from 2020 to 2022. Arthropod abundance (individuals/(trap × day)) values were 4.9 ± 0.46, 3.7 ± 0.18, and 2.6 ± 0.16 (mean ± stderr) for conventional crops, biomass sorghum, and energycane, respectively, with a significant difference found only between conventional crops and energycane. Individuals were identified to arthropod orders, and Hill’s diversity indices were calculated based on the number of individuals in each arthropod order instead of the number of individuals in each arthropod species. Order-based arthropod richness values were 5.3, 5.2, and 4.8 for biomass sorghum, conventional crops, and energycane, with significant difference found only between biomass sorghum and energycane. There was no significant difference in the order-based Shannon diversity and Simpson diversity between the three crop types. The effective number of arthropod orders for the two energy crops decreased from 5.0 to 3.4 to 2.9 with increasing order of diversity from arthropod richness to Shannon diversity to Simpson diversity. The explained variability by environmental factors also decreased with increasing Hill’s order of diversity. The results from this study indicate no significant advantage in order-based arthropod diversity in growing biomass sorghum and energycane. This research fills a critical knowledge gap in understanding the impacts of cellulosic energy crop production on biodiversity and ecosystem services.

Yang, Yubin↗
DOE OSTI2019

Interfacial Engineering for High Efficiency Nanorod Array Structured Perovskite Solar Cells

TiO 2 nanorod (NR) array for perovskite solar cells (PSCs) has attached great importance due to superb power conversion efficiency (PCE) compared to traditional mesoporous TiO 2 film. A TiO 2 compact layer for the growth of TiO 2 NR array via spin-coating cannot meet the requirements for efficient NR based PSCs. Herein we have developed and demonstrated inserting an bifunctional extra-thin TiO 2 interlayer (5 nm) by atomic layer deposition (ALD) at the interface of FTO/TiO 2 compact layer to achieve an alleviated electron exchange and a reduced energetic barrier. Thus, an accelerated extraction of electrons from TiO 2 NR arrays via the compact layer to FTO substrate is ready for improving the PSC efficiency. The thickness of the spin-coated TiO 2 compact layer on the ALD-deposited TiO 2 layer is spontaneously optimized. Finally, an outstanding efficiency of 20.28% has been achieved from a champion PSC with negligible hysteresis and high reliability. To the best of our knowledge, this is the first time to demonstrate the superiority of TiO 2 -NR based PSCs withstanding in dry heat and thermal cycling tests. The results are of great importance for the preparation of efficient and durable PSCs for our real-world applications.

25 ENERGY STORAGE↗
DOE OSTI2024

Robust wind farm layout optimization

Wake interactions in wind farms cause losses in annual energy production (AEP) on the order of 10%. Wind farm designers optimize the layout of the farm to mitigate wake losses, especially in the dominant site-specific wind directions. As wind turbines and wind farms grow in scale, optimization becomes more complex. Offshore wind farms regularly comprise more than 100 wind turbines and are characterized by complex boundaries due to shipping lanes, neighboring wind farms, and other constraints. Layout optimization methods are broadly split between gradient-based and gradient-free approaches. Gradient-based approaches can converge quickly and perform well for smaller, academic problems but are often sensitive to initial conditions and tuning parameters and require expert knowledge to use. On the other hand, gradient-free approaches can be more robust to problem complexities. We present a robust layout optimization approach based on a random search algorithm. The algorithm is intended for those who are not optimization experts and has few tuning parameters that need specification to achieve satisfactory results. Unlike off-the-shelf methods, which use generally available, non-domain-specific optimization routines that accept as inputs an optimization function and constraint definitions, this approach takes advantage of the relative computational costs of the different evaluations by evaluating cheaper computations first (boundary and minimum distance constraints) and running expensive AEP evaluations only if all other checks pass. Moreover, an outer genetic algorithm allows multiple solutions to evolve in parallel, enabling rapid solution development on high-performance computers. We discuss the relative ease of selecting necessary tuning parameters and demonstrate the efficacy of the genetic random search on a complex layout problem consisting of placing 70 turbines in a nonconvex and unconnected boundary region.

17 WIND ENERGY↗
DOE OSTI2023

Degradation of Insulating Glass Units: Thermal Performance, Measurements and Energy Impacts

Insulating glass unit (IGU) degradation has been studied extensively. However, there is limited understanding of how present durability evaluation standards relate to product lifetime. Furthermore, there is debate on how to quantify performance of installed windows over time to better understand degradation processes. More knowledge on these topics is required to link durability evaluation to product lifetime projections based on energy performance. Energy models provide helpful estimations of total annual building energy consumption. However, most models are based on “as installed” performance of envelope components and fail to account for performance degradation. This can lead to an underestimation of building lifetime energy consumption. A better understanding of the relationship between durability and energy performance can inform integration of degradation dynamics into energy modeling software. This will improve lifetime building energy consumption estimations as well as inform appropriate retrofit strategies and timing. This paper reviews current durability literature, various standards for window performance ratings and weathering methods, existing in situ IGU energy performance measurement techniques, and whole-building energy effects. The challenges and disparities among various studies are analyzed and discussed. The authors hope that further work in this area will lead to the development of improved in situ test methods to assess IGU degradation in the field and link this knowledge to improved energy performance modeling approaches.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2022

The pursuit of net-positive sustainability for industrial decarbonization with hybrid energy systems

We report signatories of the Paris Agreement are set to miss their climate targets. The net-zero pledges announced to date across many countries and private industries are insufficient to achieve carbon neutrality, which requires implementation of far-reaching and significantly scaled-up climate-positive actions. All technology options that pertain to deep decarbonization and carbon removal must be part of the mitigation portfolio. Transformative action plans must be established in which every individual/organization around the globe is an actor of changes in relation to net-zero goals. Such plans must involve governmental policy support but also encourage voluntary efforts to boost innovations, investments, initiatives, and behavioral changes. Recognition of these efforts is made quantifiable with the concepts of carbon handprint and net positivity. This paper presents a carbon handprint perspective on characterizing the environmental benefits of hybrid energy systems (HESs)—a widely applicable solution to cleaner production leveraging the capabilities of a portfolio of low-carbon energy sources—that provide heat and electricity to industrial processes. First, the carbon handprint and net positivity concepts and their calculation approach are introduced. The state of the art of HES-enabled industrial cogeneration is then surveyed, and the greenhouse gas emission intensities of different energy sources that power HESs are compared. Next, drawing on a case study about a US chemical facility's voluntary initiative to explore replacing its fossil fuel–based cogeneration infrastructure with a clean energy–generated HES, several technically viable scenarios are evaluated—especially those with small modular nuclear reactors—to illustrate how the positive-thinking handprint approach helps encourage and inform the search for widespread influence pathways in pursuit of net-positive sustainability. Finally, current knowledge gaps in the case study are identified, and opportunities to scale up the proposed handprint-based analysis are outlined with consideration of an expanded role of HESs in fulfilling climate objectives.It is envisioned that like-minded decision makers in the industry sector and beyond will adopt this perspective and act synergistically to enhance their environmental stewardship through voluntary actions and make greater contributions to the planet's climate future.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2020

Generative adversarial networks (GAN) based efficient sampling of chemical composition space for inverse design of inorganic materials

A major challenge in materials design is how to efficiently search the vast chemical design space to find the materials with desired properties. One effective strategy is to develop sampling algorithms that can exploit both explicit chemical knowledge and implicit composition rules embodied in the large materials database. Here, we propose a generative machine learning model (MatGAN) based on a generative adversarial network (GAN) for efficient generation of new hypothetical inorganic materials. Trained with materials from the ICSD database, our GAN model can generate hypothetical materials not existing in the training dataset, reaching a novelty of 92.53% when generating 2 million samples. The percentage of chemically valid (charge-neutral and electronegativity-balanced) samples out of all generated ones reaches 84.5% when generated by our GAN trained with such samples screened from ICSD, even though no such chemical rules are explicitly enforced in our GAN model, indicating its capability to learn implicit chemical composition rules to form compounds. Our algorithm is expected to be used to greatly expand the range of the design space for inverse design and large-scale computational screening of inorganic materials.

36 MATERIALS SCIENCE↗