Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “GANs”

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 505 records · Page 28
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Physics-Informed Generative Adversarial Networks for Stochastic Differential Equations

We developed a new class of physics-informed generative adversarial networks (PI-GANs) to solve forward, inverse, and mixed stochastic problems in a unified manner based on a limited number of scattered measurements. Unlike standard GANs relying solely on data for training, here we encoded into the architecture of GANs the governing physical laws in the form of stochastic differential equations (SDEs) using automatic differentiation. In particular, we applied Wasserstein GANs with gradient penalty (WGAN-GP) for its enhanced stability compared to vanilla GANs. We first tested WGAN-GP in approximating Gaussian processes of different correlation lengths based on data realizations collected from simultaneous reads at sparsely placed sensors. We obtained good approximation of the generated stochastic processes to the target ones even if there is a mismatch between the input noise dimensionality and the effective dimensionality of the target stochastic processes. We also studied the overfitting issue for both the discriminator and the generator, and we found that overfitting occurs also in the generator in addition to the discriminator as previously reported. Subsequently, we considered the solution of elliptic SDEs requiring approximations of three stochastic processes, namely the solution, the forcing, and the diffusion coefficient. Here again, we assumed data collected from simultaneous reads at a limited number of sensors for the multiple stochastic processes. Furthermore, three generators were used for the PI-GANs: two of them were feed forward deep neural networks (DNNs), while the other one was the neural network induced by the SDE. For the case where we have one group of data, we employed one feed forward DNN as the discriminator, while for the case of multiple groups of data we employed multiple discriminators in PI-GANs. We solved forward, inverse, and mixed problems without changing the framework of PI-GANs, obtaining both the means and the standard deviations of the stochastic solution and the diffusion coefficient in good agreement with benchmarks. In this work, we have demonstrated the effectiveness of PI-GANs in solving SDEs for about 120 dimensions. In principle, PI-GANs could tackle very high dimensional problems given more sensor data with low-polynomial growth in computational cost.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Entangling Quantum Generative Adversarial Networks

Generative adversarial networks (GANs) are one of the most widely adopted machine learning methods for data generation. In this work, we propose a new type of architecture for quantum generative adversarial networks (an entangling quantum GAN, EQ-GAN) that overcomes limitations of previously proposed quantum GANs. Leveraging the entangling power of quantum circuits, the EQ-GAN converges to the Nash equilibrium by performing entangling operations between both the generator output and true quantum data. In the first multiqubit experimental demonstration of a fully quantum GAN with a provably optimal Nash equilibrium, we use the EQ-GAN on a Google Sycamore superconducting quantum processor to mitigate uncharacterized errors, and we numerically confirm successful error mitigation with simulations up to 18 qubits. Finally, we present an application of the EQ-GAN to prepare an approximate quantum random access memory and for the training of quantum neural networks via variational datasets.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2025

Temperature-Agnostic Pt/Au Ohmic Contacts on n -Type Gallium Nitride for Self-Aligned MOSFETs

Five alternative metals are investigated as ohmic contacts to n-GaN including Cr/Au, Mo/Au, Pt/Au, Pd/Au, and Ge/Au. Ti-based contacts are traditionally used for ohmic contacts on n-GaN. However, conventional Ti/Al/Ni/Au metallization is found to be incompatible with a self-aligned process for GaN trench MOSFETs due to wet etch restrictions. Therefore, an alternative metallization is needed that is unreactive to the etch chemistry used in the self-aligned process. Additionally, the contact should remain ohmic following anneal at 900 °C so that contact formation can precede the anneal required for p-dopant activation. Here, in the present work, an n-GaN bilayer, consisting of a thin heavily doped contact layer (n 0 = 1 × 10 20 cm −3 ) atop a thick lesser doped layer, is used to demonstrate ohmic contacts of alternative metals with low specific contact resistance and extended thermal budget. Cr/Au ohmic contacts are demonstrated up to anneal temperatures of 800 °C, an increase of 200 °C compared to the highest known reports for Cr/Au contacts on n-GaN. Pt/Au metallization is demonstrated as an ohmic contact to n-GaN for the first time and exhibits true temperature-agnostic behavior up to anneal temperatures of 900 °C with specific contact resistance that is near parity with Ti/Al/Ni/Au. The temperature-agnostic behavior of Pt/Au ohmic contacts on the n-GaN bilayer, in addition to chemical compatibility with the self-aligned process, positions Pt/Au contacts as a key enabling element for self-aligned trench MOSFETs on GaN.

Reilly, Kevin James [Sandia National Laboratories ↗
DOE OSTI2025

Effect of Thermal Oxidation on the Structure, Surface Texturing, and Microstructure Evolution in Nanocrystalline Ga-O-N Films

An extensive examination of the nanoscale, crystallographic growth dynamics of the system, which is impacted by the thermal energy given to the GaN, is carried out to derive a deeper understanding of the growth kinetics, morphology and microstructure evolution, chemical bonding, and optical properties of Ga-O-N films. Thermal annealing of GaN films is performed in the temperature range of 900–1200 °C. Crystal structure, phase formation, chemical composition, surface morphology, and microstructure evolution of Ga-O-N films are investigated as a function of temperature. Increasing temperature induces surface oxidation, which results in the formation of stable β-Ga2O3 phase in the GaN matrix, where the overall film composition evolves from nitride (GaN) to oxynitride (Ga-O-N). While GaN surfaces are smooth, planar, and featureless, oxidation induced granular-to-rod shaped morphology evolution is seen with increasing temperature to 1200 °C. The considerable texturing and stability of the nanocrystalline Ga-O-N on Si substrates can be attributed to the surface and interface driven modification because of thermal treatment. Corroborating with structure and chemical changes, Raman spectroscopic analyses also indicate that the chemical bonding evolution progresses from fully Ga-N bonds to Ga-O-N. While the GaN oxidation process starts with the formation of β-Ga 2 O 3 at an annealing temperature of 1000 °C, higher annealing temperatures induce structural distortion with the potential formation of Ga-O-N bonds. The structure-phase-chemical composition correlation, which will be useful for nanocrystalline materials for selective optoelectronic applications, is established in Ga-O-N films made by thermal treatment of GaN.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Utilization of Synthetic Near-Infrared Spectra via Generative Adversarial Network to Improve Wood Stiffness Prediction

Near-infrared (NIR) spectroscopy is widely used as a nondestructive evaluation (NDE) tool for predicting wood properties. When deploying NIR models, one faces challenges in ensuring representative training data, which large datasets can mitigate but often at a significant cost. Machine learning and deep learning NIR models are at an even greater disadvantage because they typically require higher sample sizes for training. In this study, NIR spectra were collected to predict the modulus of elasticity (MOE) of southern pine lumber (training set = 573 samples, testing set = 145 samples). To account for the limited size of the training data, this study employed a generative adversarial network (GAN) to generate synthetic NIR spectra. The training dataset was fed into a GAN to generate 313, 573, and 1000 synthetic spectra. The original and enhanced datasets were used to train artificial neural networks (ANNs), convolutional neural networks (CNNs), and light gradient boosting machines (LGBMs) for MOE prediction. Overall, results showed that data augmentation using GAN improved the coefficient of determination (R 2 ) by up to 7.02% and reduced the error of predictions by up to 4.29%. ANNs and CNNs benefited more from synthetic spectra than LGBMs, which only yielded slight improvement. All models showed optimal performance when 313 synthetic spectra were added to the original training data; further additions did not improve model performance because the quality of the datapoints generated by GAN beyond a certain threshold is poor, and one of the main reasons for this can be the size of the initial training data fed into the GAN. LGBMs showed superior performances than ANNs and CNNs on both the original and enhanced training datasets, which highlights the significance of selecting an appropriate machine learning or deep learning model for NIR spectral-data analysis. The results highlighted the positive impact of GAN on the predictive performance of models utilizing NIR spectroscopy as an NDE technique and monitoring tool for wood mechanical-property evaluation. Further studies should investigate the impact of the initial size of training data, the optimal number of generated synthetic spectra, and machine learning or deep learning models that could benefit more from data augmentation using GANs.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Selective incorporation of antimony into gallium nitride

Dilute concentrations of antimony (Sb) incorporation into GaN induce strong bandgap bowing and tunable room-temperature photoluminescence from the UV to the green spectral regions. However, the atomistic details of the incorporation of Sb into the GaN host remain unclear. In this work, we use first-principles calculations to understand the thermodynamics of Sb substitution into GaN and its effect on the optical and Raman spectra. Although it is empirically considered that Sb is preferentially incorporated as an anion (Sb3−) into the N sublattice, we demonstrate that Sb can also be incorporated as a cation (Sb3+, Sb5+) into the metal sublattice. Our thermodynamic analysis demonstrates that SbN0, SbGa2+, and SbGa0 can co-exist under Ga-rich conditions in n-type samples. We further confirm the dual incorporation of Sb by calculating the vibrational frequencies of different anionic and cationic substitutions to explain the origins of experimentally observed additional Raman peaks of Sb-doped GaN. Moreover, the calculated band structures of different Sb substitutions into GaN explain the experimental photoluminescence and optical absorption spectra. Overall, our analysis suggests that the coexistence of Sb3−, Sb3+, and Sb5+ substitutions into GaN explains the totality of experimental measurements. Our results demonstrate that the selective incorporation of Sb into GaN (and potentially other group-V elements such as As, P, or Bi) by tuning the growth conditions can drastically modify the electronic properties, for applications in visible light emitters and photocatalysis.

Physics↗
DOE OSTI2020

Conditional Generative Adversarial Networks for Solving Heat Transfer Problems

Generative Adversarial Networks (GANs) have been used as a deep learning approach to solving physics and engineering problems. Using deep learning for these problems is attractive in that reasonably accurate models can be inferred from only raw data, eliminating the need to define the exact physical equations governing a problem. We expand on previous work using GANs to generate steady-state solutions to the two-dimensional heat equation. Using a basic conditional GAN (cGAN), we generate accurate solutions for rectangular domains conditioned on four edge boundary conditions (MAE < 0.5%). For finding steady-state solutions over arbitrary two-dimensional domains (not constrained to rectangles), we use a cGAN designed for image-to-image translation. We train this GAN on various types of geometric domains (circles, squares, triangles, shapes with one circular or rectangular hole), achieving accurate results on test data made up of geometries similar to those in training (MAE < 1%). For both of these GANs, we experiment with different loss function terms, showing that a term using the gradients of solution images significantly improves the basic cGAN but not the image-to-image GAN. Lastly, we show that the image-to-image GAN performs poorly when applied to two-dimensional geometries that vary in structure from training data (MAE < 8% for shapes with multiple holes or different shaped holes). This demonstrates the cGAN's lack of generalizability. While the cGAN is an accurate and computationally efficient method when trained and tested on similarly structured data, it is a much less reliable method when applied to data that is slightly different in structure from the training data.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Improving luminescence response in ZnGeN 2 /GaN superlattices: defect reduction through composition control

Abstract Color-mixed (cm) light-emitting diodes (LEDs) are theoretically the most efficient white light emitters, projected to improve white light luminous efficacy by 34% compared to incumbent phosphor converted LEDs. Since white light technology is pervasive and essential, small improvements in LED technology can result in energy savings. However, cm-LEDs are not yet realized due to poor efficacy in green and amber emitting materials, a spectral region colloquially referred to as the Green Gap. ZnGeN 2 is nearly isostructural and closely lattice-matched to GaN and can be heteroepitaxially integrated with existing GaN devices; ZnGeN 2 /GaN hybrid structures are theorized to emit green (~530 nn) light with a spontaneous emission rate 4.6–4.9 times higher than traditional InGaN LEDs when incorporated into III-N LED structures. In this report we demonstrate the molecular beam epitaxy (MBE) growth of GaN and ZnGeN 2 superlattices, an important step towards realizing multiple quantum well structures required for efficient LEDs. Elemental analysis, including atom probe tomography, shows that Ga and Ge are observed in both ZnGeN 2 and GaN layers, degrading the structural uniformity. The lack of elemental abruptness also leads to increased defect luminescence and reabsorption of band edge luminescence. The source of unintentional Ga distributed throughout the ZnGeN 2 layers was identified as excess flux escaping from around the closed MBE shutter. The source of unintentional Ge, which tended to incorporate as a single delta-doped layer in GaN, was identified as Ge riding along the cyclical metal-rich Ga adlayer used for high quality GaN, incorporating during subsequent nitrogen-rich growth step. Modifying the growth strategy results in improved structural quality, elemental abruptness, and luminescence response. This realization of structurally and elementally abrupt interfaces demonstrates the potential of heteroepitaxially integrated binary and ternary nitrides for energy-relevant devices.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2022

External charge compensation in etched gallium nitride measured by x-ray photoelectron spectroscopy

Electronic states at GaN surfaces and at regrowth and heteroepitaxy interfaces inhibit electronic device performance. Understanding electronic state configuration at the GaN surface is, therefore, crucial for the development of GaN-based devices, which are currently of considerable interest in power electronic applications. GaN and other wurtzite III-nitrides possess large spontaneous polarization along the c-axis, producing a bound sheet charge at the surface, which affects the electronic state configuration through the formation of internal and external compensation charges. Defects induced by conventional plasma-based dry etching methods may inhibit the internal screening of this bound charge and thus increase the concentration of external charged states. The surface band bending of n-type Ga-face GaN (0001) was measured with x-ray photoelectron spectroscopy after inductively coupled plasma etching to investigate the impact of dry etching on external charge compensation. GaN samples were etched using inductively coupled plasma with varying rf power and a novel plasma-enhanced atomic layer etching method using an oxidation, fluorination, and ligand-exchange mechanism. The band bending varied from 0.0 to 0.8 ± 0.1 eV for the samples measured.

Hatch, Kevin A. (ORCID:0000000283118623)↗
DOE OSTI2020

Tuned and screened range-separated hybrid density functional theory for describing electronic and optical properties of defective gallium nitride

We apply a hybrid density functional theory approach, based on a tuned and screened range-separated hybrid (SRSH) exchange-correlation functional, to describe the optoelectronic properties of defective gallium nitride (GaN). SRSH and time-dependent SRSH (TDSRSH) are tuned to produce accurate energetics for the pristine material and applied to the study of a series of point defects in bulk GaN, a blue light-emitting material that degrades in the presence of defects. We first establish the accuracy of the method by comparing the predicted quasiparticle gap and low-energy excitation spectra of (TD)SRSH and many-body perturbation theory for both pristine GaN and GaN containing a single nitrogen vacancy. Aided by the reduced computational cost of (TD)SRSH, we then report on three additional technologically relevant point defects and defect complexes in GaN: the gallium vacancy, the carbon interstitial, and the carbon-silicon complex. We compute the low-energy optical absorption spectra for these defects and show the presence of defect-centered transitions. Furthermore, by estimating the Stokes shift, we predict, in agreement with previous studies, that the carbon substitutional defect is a candidate for the detrimental yellow luminescence in GaN. Furthermore, this study indicates that TDSRSH is a promising and computationally feasible approach for quantitatively accurate, first-principles modeling of defective semiconductors.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Interfacial oxide and other species in trimethylaluminum-pretreated atomic layer deposition-Al 2 O 3 /GaN characterized by sputter-assisted ToF-SIMS

Deposition techniques, like atomic layer deposition (ALD), are used to form high-quality dielectrics for GaN-based metal–oxide–semiconductor (MOS) gate structures due to the lack of a reliable thermal oxide in GaN. Moreover, interfacial GaO x from pre-existing native oxides is thought to adversely impact channel carrier dynamics and induce undesired threshold voltage shifts in GaN-based MOS gate structures. Exposure of the GaN surface to the trimethylaluminum (TMA) precursor prior to standard alumina ALD decreases the native oxide layer on GaN, but the extent of chemical modification has not been well studied in the context of interface composition in a MOS gate structure. Herein, we compare annealed 55 nm Al 2 O 3 dielectric films on GaN grown using either a water-first ALD process or a process including sequential pulses of TMA immediately before the initiation of Al 2 O 3 ALD. Time-of-Flight Secondary Ion Mass Spectrometry measures differences in the interfacial GaO x content between each ALD film. Here, it also detects surface contaminant species like Si, F, S, and C. Furthermore, we report the formation of an AlN species at the Al 2 O 3 /GaN interface, which is more prominent for the film grown using the TMA pre-pulse step. In general, this work demonstrates that the TMA pre-pulse step is an effective strategy for cleaning substrate surfaces prior to ALD.

Atomic layer deposition↗
DOE OSTI2020

MOCVD Growth and Characterization of ZnSnN 2

Since the first few reports of its synthesis in 2011 and 20121-3, interest in ZnSnN 2 for diverse applications, including photovoltaics1,3,4, visible light photocatalysis5, and optoelectronics, has been growing. Novel nitride-based LED device designs promise high gains in efficiency at emission wavelengths into the green, amber and even red.6-8 Their realization requires the precise insertion of an ultrathin layer of ZnGeN 2 or ZnSnN 2 into the active region of an InGaAs-GaAs LED structure. While thin films of ZnSnN 2 have been synthesized by RF sputtering1,9 and by MBE 2,10,11, and thin films of ZnGeN 2 , ZnSiN 2 and their alloys by MOCVD, the synthesis of ZnSnN 2 thin films by MOCVD has not yet been reported. It is a challenging goal, given the narrow temperature window between efficient activation of ammonia and the thermal decomposition of ZnSnN 2 at near-atmospheric pressures. We report here the results of studies of ZnSnN 2 films grown on r- and c-plane sapphire substrates, GaN templates on sapphire, and In 0.15 Ga 0.85 N templates on GaN-on-sapphire, at temperatures from 500 °C to 560 °C, using tetramethyltin (TMSn), diethylzinc (DEZn), germane and ammonia. Chamber pressures were varied between 200 and 500 torr. In addition, the ratios of the cation precursors and the total cation-to-anion precursor ratios were varied. Single crystal growth, as evidenced by 2θ-ω x-ray diffraction scans, was achieved on all of these substrates. Wurtzitic (0002) ZnSnN 2 diffraction were observed at 2θ ~ 32.90 for the samples grown on c-sapphire, GaN (0001) and InGaN (0001) templates, and wurtzitic (11-20) peak at 2θ ~ 54.80 for the films grown on r-sapphire. Higher pressures and higher temperatures promoted better crystallinity and more continuous films, as revealed by scanning electron microscopy (SEM) images and 2θ-ω x-ray diffraction peak widths and intensities. Crystal quality was observed to be highly dependent on growth temperature. For example, in one series a 10 °C change in growth temperature resulting in a factor of 15 increase in diffraction peak intensity. SEM images show that films grown on the GaN templates had better uniformity in nucleation than those grown on sapphire. The presence of satellite peaks around the Bragg peak in the XRD 2θ-ω scan for a film grown on a GaN template, and the absence of satellite peaks for the films grown simultaneously on c- and r-sapphire, also indicate superior quality for the film grown on GaN. HAADF-STEM images of films grown on these templates showed smooth interfaces at the film-substrate boundary and show growth rates as high as 167 nm/h, considerably higher than reported previously for MBE growth.9,10 Electron diffraction patterns show single crystallinity. Rocking curves show FWHM as low as 0.100. RMS surfaces roughnesses, measured by atomic force microscopy, are of the order of 3 nm for 1µm x 1µm areas. Current work is focusing on optimizing growth on InGaN templates. These templates are promising for their relatively lower lattice mismatch with ZnSnN 2 . Their use is a next step toward producing LED device structures.

Jayatunga, Benthara Hewage Dinushi↗
DOE OSTI2024

Simple self-consistent method for excited states in density functional theory to characterize defect-derived behavior in wide-band-gap-based microelectronic materials

This final report summarizes the results of the Laboratory Direct Research and Development (LDRD) Project Number 229740. Wide band gap semiconductors such as gallium nitride (GaN) have features highly desirable for multiple mission electronic applications. Realization of their potential requires atomic-scale understanding of electronic behavior. The principal experimental tools for electronically probing defects in GaN are chemically undifferentiating and lack a practical theoretical counterpart needed to identify and characterize specific defects. This project investigated whether a simple idea for modeling defect excited states and their associated photoluminescence (PL) energies is viable, as a path to accelerate the understanding of defect behavior and gain valuable insights into engineering new electronic materials and devices. The research implemented a non-self-consistent total-energy evaluation of a Koopmans-type estimation of an excited electronic state energy in density functional theory (DFT) calculations, and proceeded to design, implement, and assess a self-consistent method for computing excited states based upon an OCcupation-Constrained-DFT (occ-DFT). The occ-DFT was verified in test calculations of defect excited states and validated against well-characterized PL data for 3d transition metal defects in GaN. The method proved stable and robust in computing excited states and gave accurate predictions compared to experimental PL data. The combined ground state/excited-state capability proved capable of chemically differentiating defect species in GaN. In application to 3d dopants in GaN, we reinterpreted extensive experimental literature, proposed new defects as prospective candidates for use in quantum information applications, and outlined design strategies to create and exploit these potentially useful functional defects in GaN.

36 MATERIALS SCIENCE↗
NASA NTRS2001

Atomically Flat Surfaces Developed for Improved Semiconductor Devices

New wide bandgap semiconductor materials are being developed to meet the diverse high temperature, -power, and -frequency demands of the aerospace industry. Two of the most promising emerging materials are silicon carbide (SiC) for high-temperature and high power applications and gallium nitride (GaN) for high-frequency and optical (blue-light-emitting diodes and lasers) applications. This past year Glenn scientists implemented a NASA-patented crystal growth process for producing arrays of device-size mesas whose tops are atomically flat (i.e., step-free). It is expected that these mesas can be used for fabricating SiC and GaN devices with major improvements in performance and lifetime. The promising new SiC and GaN devices are fabricated in thin-crystal films (known as epi films) that are grown on commercial single-crystal SiC wafers. At this time, no commercial GaN wafers exist. Crystal defects, known as screw defects and micropipes, that are present in the commercial SiC wafers propagate into the epi films and degrade the performance and lifetime of subsequently fabricated devices. The new technology isolates the screw defects in a small percentage of small device-size mesas on the surface of commercial SiC wafers. This enables atomically flat surfaces to be grown on the remaining defect-free mesas. We believe that the atomically flat mesas can also be used to grow GaN epi films with a much lower defect density than in the GaN epi films currently being grown. Much improved devices are expected from these improved low-defect epi films. Surface-sensitive SiC devices such as Schottky diodes and field effect transistors should benefit from atomically flat substrates. Also, we believe that the atomically flat SiC surface will be an ideal surface on which to fabricate nanoscale sensors and devices. The process for achieving atomically flat surfaces is illustrated. The surface steps present on the "as-received" commercial SiC wafer is also illustrated. because of the small tilt angle between the crystal "basal" plane and the polished wafer surface. These steps are used in normal SiC epi film growth in a process known as stepflow growth to produce material for device fabrication. In the new process, the first step is to etch an array of mesas on the SiC wafer top surface. Then, epi film growth is carried out in the step flow fashion until all steps have grown themselves out of existence on each defect-free mesa. If the size of the mesas is sufficiently small (about 0.1 by 0.1 mm), then only a small percentage of the mesas will contain an undesired screw defect. Mesas with screw defects supply steps during the growth process, allowing a rough surface with unwanted hillocks to form on the mesa. The improvement in SiC epi surface morphology achievable with the new technology is shown. An atomic force microscope image of a typical SiC commercial epilayer surface is also shown. A similar image of an SiC atomically flat epi surface grown in a Glenn laboratory is given. With the current screw defect density of commercial wafers (about 5000 defects/cm2), the yield of atomically free 0.1 by 0.l mm mesas is expected to be about 90 percent. This is large enough for many types of electronic and optical devices. The implementation of this new technology was recently published in Applied Physics Letters. This work was initially carried out in-house under a Director's Discretionary Fund project and is currently being further developed under the Information Technology Base Program.

Powell, J. Anthony↗
DOE OSTI2026

Uranium Doped Gallium Nitride Epitaxial Thin Films

Gallium nitride (GaN) is near ubiquitous in modern day technologies, forming the backbone of solid-state lighting and high-power electronics. Engineering the physical properties of GaN has been investigated to some degree by the incorporation or doping of most of the elements of the periodic table, but the actinides remain unexplored. Molecular beam epitaxy is used to demonstrate uranium doping of GaN single crystals. High structural quality of the host matrix is maintained despite partial elemental segregation of the uranium dopant into 1D structures at the levels presented here. Electronic transport measurements reveal relatively high conductivity, which persists down to cryogenic temperature and is characterized by the formation of narrow gaps in the electronic band structures very close to the Fermi level. Photoluminescence measurements reveal that the U-doped GaN exhibits optical behavior similar to that of the GaN substrate. The addition of actinide materials to a non-centrosymmetric, optically active, radiation-hard, and electronically tunable host matrix opens a world of possibilities for investigating and leveraging elements with high electron correlations in the pursuit of novel devices.

36 - MATERIALS SCIENCE↗
DOE OSTI2025

Data efficiency assessment of generative adversarial networks in energy applications

This study investigates the data requirements of generative artificial intelligence (AI), particularly generative adversarial networks (GANs), for reliable data augmentation in energy applications. Generative AI, though seen as a solution to data limitations, requires substantial data to learn meaningful distributions—a challenge often overlooked. This study addresses the challenge through synthetic data generation for critical heat flux (CHF) and power grid demand, focusing on renewable and nuclear energy. Two variants of GAN employed are conditional GAN (cGAN) and Wasserstein GAN (wGAN). Our findings include the strong dependency of GAN on data size, with performance declining on smaller datasets and varying performance when generalizing to unseen experiments. Mass flux and heated length significantly influence CHF predictions. wGAN is more robust to feature exclusion, making it suitable for constrained synthetic data generation. In energy demand forecasting, wGAN performed well for solar, wind, and load predictions. Longer lookback hours and larger datasets improved predictions, especially for load power. Seasonal variations posed challenges, with wGAN achieving a relatively high error of Root Mean Squared Error (RMSE) of 0.32 for load power prediction, compared to RMSE of 0.07 under same-season conditions. Feature exclusions impacted cGAN the most, while wGAN showed greater robustness. This study concludes that, while generative AI is effective for data augmentation, it requires substantial data and careful training to generate realistic synthetic data and generalize to new experiments in engineering applications.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗