Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “transformer”

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 847 records · Page 47
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Deep Learning for Subsurface Flow: A Comparative Study of U‐Net, Fourier Neural Operators, and Transformers in Underground Hydrogen Storage

Subsurface flow research is essential for the sustainable management of natural resources and the environment. Deep learning (DL) has significantly advanced this field by developing efficient and accurate surrogate models to replace computationally expensive physics‐based simulations. These surrogate models are commonly used to predict the spatiotemporal evolution of state variables, such as gas saturation and reservoir pressure, in heterogeneous geological formations. Despite the various DL models applied to this task, there is a lack of studies systematically comparing their performance. This absence of comparative analysis leads to somewhat arbitrary DL model selection in subsurface flow research, resulting in suboptimal performance and potentially inaccurate predictions. To bridge this gap, we conduct a systematic comparison study of three popular DL architectures—U‐Net, Fourier Neural Operators (FNO), and Segmentation Transformer (SETR)—in surrogate modeling of underground hydrogen storage (UHS). We focus on UHS due to its promise of enhancing clean energy resilience and its cyclic operational conditions that represent common scenarios in various subsurface applications. We evaluate the models based on accuracy, training cost, and inference speed. The comparison shows that U‐Net achieves the highest accuracy, followed by SETR and FNO. Despite its lower accuracy, FNO has the highest inference speed. SETR offers competitive accuracy with the least training memory usage, demonstrating the potential of transformers in learning subsurface flow. Our results provide guidance for selecting DL models for surrogate modeling in a wide range of subsurface flow problems.

42 ENGINEERING↗
DOE OSTI2021

CRL4A DTL degrades DNA-PKcs to modulate NHEJ repair and induce genomic instability and subsequent malignant transformation

Genomic instability induced by DNA damage and improper DNA damage repair is one of the main causes of malignant transformation and tumorigenesis. DNA double strand breaks (DSBs) are the most detrimental form of DNA damage, and nonhomologous end-joining (NHEJ) mechanisms play dominant and priority roles in initiating DSB repair. A well-studied oncogene, the ubiquitin ligase Cullin 4A (CUL4A), is reported to be recruited to DSB sites in genomic DNA, but whether it regulates NHEJ mechanisms of DSB repair is unclear. Here, we discovered that the CUL4A-DTL ligase complex targeted the DNA-PKcs protein in the NHEJ repair pathway for nuclear degradation. Overexpression of either CUL4A or DTL reduced NHEJ repair efficiency and subsequently increased the accumulation of DSBs. Moreover, we demonstrated that overexpression of either CUL4A or DTL in normal cells led to genomic instability and malignant proliferation. Consistent with the in vitro findings, in human precancerous lesions, CUL4A expression gradually increased with increasing malignant tendency and was negatively correlated with DNA-PKcs and positively correlated with γ-H2AX expression. Collectively, this study provided strong evidence that the CUL4A-DTL axis increases genomic instability and enhances the subsequent malignant transformation of normal cells by inhibiting NHEJ repair. These results also suggested that CUL4A may be a prognostic marker of precancerous lesions and a potential therapeutic target in cancer.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2023

Calcium promotes persistent soil organic matter by altering microbial transformation of plant litter

Abstract Calcium (Ca) can contribute to soil organic carbon (SOC) persistence by mediating physico-chemical interactions between organic compounds and minerals. Yet, Ca is also crucial for microbial adhesion, potentially affecting colonization of plant and mineral surfaces. The importance of Ca as a mediator of microbe-mineral-organic matter interactions and resulting SOC transformation has been largely overlooked. We incubated 44 Ca labeled soils with 13 C 15 N labeled leaf litter to study how Ca affects microbial transformation of litter and formation of mineral associated organic matter. Here we show that Ca additions promote hyphae-forming bacteria, which often specialize in colonizing surfaces, and increase incorporation of litter into microbial biomass and carbon use efficiency by approximately 45% each. Ca additions reduce cumulative CO 2 production by 4%, while promoting associations between minerals and microbial byproducts of plant litter. These findings expand the role of Ca in SOC persistence from solely a driver of physico-chemical reactions to a mediator of coupled abiotic-biotic cycling of SOC.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

TransPolymer: a Transformer-based language model for polymer property predictions

Accurate and efficient prediction of polymer properties is of great significance in polymer design. Conventionally, expensive and time-consuming experiments or simulations are required to evaluate polymer functions. Recently, Transformer models, equipped with self-attention mechanisms, have exhibited superior performance in natural language processing. However, such methods have not been investigated in polymer sciences. Herein, we report TransPolymer, a Transformer-based language model for polymer property prediction. Our proposed polymer tokenizer with chemical awareness enables learning representations from polymer sequences. Rigorous experiments on ten polymer property prediction benchmarks demonstrate the superior performance of TransPolymer. Moreover, we show that TransPolymer benefits from pretraining on large unlabeled dataset via Masked Language Modeling. Experimental results further manifest the important role of self-attention in modeling polymer sequences. We highlight this model as a promising computational tool for promoting rational polymer design and understanding structure-property relationships from a data science view.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Feature engineering descriptors, transforms, and machine learning for grain boundaries and variable-sized atom clusters

Abstract Obtaining microscopic structure-property relationships for grain boundaries is challenging due to their complex atomic structures. Recent efforts use machine learning to derive these relationships, but the way the atomic grain boundary structure is represented can have a significant impact on the predictions. Key steps for property prediction common to grain boundaries and other variable-sized atom clustered structures include: (1) describing the atomic structure as a feature matrix, (2) transforming the variable-sized feature matrix to a fixed length common to all structures, and (3) applying a machine learning algorithm to predict properties from the transformed matrices. We examine how these steps and different combinations of engineered features impact the accuracy of grain boundary energy predictions using a database of over 7000 grain boundaries. Additionally, we assess how different engineered features support interpretability, offering insights into the physics of the structure-property relationships.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Fourier-based three-dimensional multistage transformer for aberration correction in multicellular specimens

High-resolution tissue imaging is often compromised by sample-induced optical aberrations that degrade resolution and contrast. Although wavefront sensor-based adaptive optics (AO) can measure these aberrations, such hardware solutions are typically complex, expensive to implement and slow when serially mapping spatially varying aberrations across large fields of view. Here we introduce AOViFT (adaptive optical vision Fourier transformer)—a machine learning-based aberration sensing framework built around a three-dimensional multistage vision transformer that operates on Fourier domain embeddings. AOViFT infers aberrations and restores diffraction-limited performance in puncta-labeled specimens with substantially reduced computational cost, training time and memory footprint compared to conventional architectures or real-space networks. We validated AOViFT on live gene-edited zebrafish embryos, demonstrating its ability to correct spatially varying aberrations using either a deformable mirror or postacquisition deconvolution. By eliminating the need for the guide star and wavefront sensing hardware and simplifying the experimental workflow, AOViFT lowers technical barriers for high-resolution volumetric microscopy across diverse biological samples.

Alshaabi, Thayer [Howard Hughes Medical Institute,↗
DOE OSTI2020

RIFTA: A Robust Iterative Fourier Transform-based dwell time Algorithm for ultra-precision ion beam figuring of synchrotron mirrors

With the rapid evolution of synchrotron X-ray sources, the demand for high-precision X-ray mirrors has greatly increased. Single nanometer profile error is required to keep imaging capability at the diffraction limit. Ion Beam Figuring (IBF), as a highly deterministic surfacing technique, has been used for ultra-precision finishing of mirrors. One crucial step that guides the IBF process is dwell time calculation. A valid dwell time solution should be non-negative and duplicate the shape of the desired removal map. Another important aspect is to minimize the total dwell time. In this study, we propose a Robust Iterative Fourier Transform-based dwell time Algorithm (RIFTA) that automatically fulfills these requirements. First, the thresholded inverse filtering in Fourier transform-based deconvolution is stabilized and automated by optimizing the threshold value using the Nelder-Mead simplex algorithm. Second, a novel two-level iterative scheme is proposed to guarantee the minimized total dwell time with its non-negativity at each dwell point. Third, a bicubic resampling is employed to flexibly adapt the calculated dwell time map to any IBF process intervals. The performance of RIFTA is first studied with simulation, followed by a comparison with the other state-of-the-art dwell time algorithms. We then demonstrate with an experiment that, using the dwell time calculated by the RIFTA, the total dwell time is shortened by a factor of two and the RMS in a 5 × 50 mm clear aperture was reduced from 3.4 nm to 1.1 nm after one IBF run, which proves the effectiveness and the efficiency of the proposed algorithm.

42 ENGINEERING↗
DOE OSTI2020

Correlating work hardening with co-activation of stacking fault strengthening and transformation in a high entropy alloy using in-situ neutron diffraction

Transformation induced plasticity (TRIP) leads to enhancements in ductility in low stacking fault energy (SFE) alloys, however to achieve an unconventional increase in strength simultaneously, there must be barriers to dislocation motion. While stacking faults (SFs) contribute to strengthening by impeding dislocation motion, the contribution of SF strengthening to work hardening during deformation is not well understood; as compared to dislocation slip, twinning induced plasticity (TWIP) and TRIP. Thus, we used in-situ neutron diffraction to correlate SF strengthening to work hardening behavior in a low SFE Fe 40 Mn 20 Cr 15 Co 20 Si 5 (at%) high entropy alloy, SFE ~ 6.31 mJ m –2 . Cooperative activation of multiple mechanisms was indicated by increases in SF strengthening and γ-f.c.c. → ε-h.c.p. transformation leading to a simultaneous increase in strength and ductility. The present study demonstrates the application of in-situ, neutron or X-ray, diffraction techniques to correlating SF strengthening to work hardening.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Physics-constrained machine learning for electrodynamics without gauge ambiguity based on Fourier transformed Maxwell’s equations

We utilize a Fourier transformation-based representation of Maxwell’s equations to develop physics-constrained neural networks for electrodynamics without gauge ambiguity, which we label the Fourier–Helmholtz–Maxwell neural operator method. In this approach, both of Gauss’s laws and Faraday’s law are built in as hard constraints, as well as the longitudinal component of Ampère–Maxwell in Fourier space, assuming the continuity equation. An encoder–decoder network acts as a solution operator for the transverse components of the Fourier transformed vector potential, $\hat{A}_⟂(k,t)$, whose two degrees of freedom are used to predict the electromagnetic fields. This method was tested on two electron beam simulations. Among the models investigated, it was found that a U-Net architecture exhibited the best performance as it trained quicker, was more accurate and generalized better than the other architectures examined. We demonstrate that our approach is useful for solving Maxwell’s equations for the electromagnetic fields generated by intense relativistic charged particle beams and that it generalizes well to unseen test data, while being orders of magnitude quicker than conventional simulations. We show that the model can be re-trained to make highly accurate predictions in as few as 20 epochs on a previously unseen data set.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Dynamic and reversible transformations of subnanometre-sized palladium on ceria for efficient methane removal

Reversibly adjusting the active structures of supported metal catalysts in response to dynamic working conditions has long been pursued. Here we report the reaction-environment-modulated transformations of subnanometre-sized Pd on CeO 2 for efficient methane removal, leveraging the reaction environments at different stages of automotive exhaust aftertreatment. During the cold start of vehicles, inactive Pd 1 single atoms are readily transformed into PdO $x$ subnanometre clusters by CO even at room temperature with excess O 2 , resulting in boosted low-temperature CH 4 oxidation. At elevated temperatures, dispersion of PdO $x$ cluster into Pd 1 against metal sintering renders outstanding hydrothermal stability to the catalyst, to be activated during the next vehicle start. Combined experimental and computational studies elucidate the dynamically evolved Pd speciation on CeO 2 at an atomic level. In conclusion, modulating the reversible nature of supported metals helps overcome the long-existing trade-off between low-temperature activity and high-temperature stability, also providing a new paradigm for designing intelligent catalysts that brings single-atom/cluster catalysts closer to real applications.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

SwinCell: a 3D transformer and flow-based framework for improved cell segmentation

Segmentation of three-dimensional (3D) cellular images is fundamental for studying and understanding cell structure and function. However, 3D cellular segmentation is challenging, particularly for dense cells and tissues. This challenge arises mainly from the complex contextual information within 3D images, anisotropic properties, and the sensitivity to internal cellular structures, which often lead to incorrect segmentation. In this work, we introduce SwinCell, a 3D transformer-based framework that leverages Swin-transformer to predict flow and differentiate individual cell instances. We demonstrate SwinCell’s utility in the segmentation of nuclei, colon tissue cells, and densely cultured cells. SwinCell strikes a balance between maintaining detailed local feature recognition and understanding broader contextual information. Through extensive testing with both public and in-house 3D cell imaging datasets, SwinCell shows utility in segmenting dense cells, making it a valuable tool for 3D segmentation in cellular analysis that could expedite research in cell biology and tissue engineering.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Crack mitigation in additively manufactured AlCrFe 2 Ni 2 high-entropy alloys through engineering phase transformation pathway

The far-from-equilibrium solidification during additive manufacturing often creates large residual stresses that induce solid-state cracking. Here we present a strategy to suppress solid-state cracking in an additively manufactured AlCrFe 2 Ni 2 high-entropy alloy via engineering phase transformation pathway. We investigate the solidification microstructures formed during laser powder-bed fusion and directed energy deposition, encompassing a broad range of cooling rates. At high cooling rates (10 4 –10 6 K/s), we observe a single-phase BCC/B2 microstructure that is susceptible to solid-state cracking. At low cooling rates (10 2 –10 4 K/s), FCC phase precipitates out from the BCC/B2 matrix, resulting in enhanced ductility (~10 %) and resistance to solid-state cracking. Site-specific residual stress/strain analysis reveals that the ductile FCC phase can largely accommodate residual stresses, a feature which helps relieve residual strains within the BCC/B2 phase to prevent cracking. Our work underscores the value of exploiting the toolbox of phase transformation pathway engineering for material design during additive manufacturing.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Stress-dependent χ phase transformation in a Ni-based superalloy

The ongoing push to elevate operating temperatures in aerospace gas turbine engines – driven by goals of enhanced fuel efficiency and reduced CO 2 emissions – mandates advancements in the creep resistance of Ni- and Co-based superalloys, which are integral for critical engine components. This study elucidates the role of stress assisted localized phase transformations in the creep properties of these alloys. By leveraging chemo-mechanical coupling, self-healing γ′ precipitates are designed to immobilize planar defects, thereby increasing creep resistance. Employing advanced characterization techniques such as high-resolution Scanning Transmission Electron Microscopy (HR-STEM), in conjunction with atomistic simulations and thermodynamic calculations, novel deformation pathways facilitated by χ local phase transformation (LPT) strengthening have been uncovered; notably, the formation of χ nano-laths through microtwinning and superlattice intrinsic stacking fault (SISF) shearing. This study highlights critical insights into the compositional boundaries necessary for optimizing LPT strengthening while avoiding deleterious bulk formation of η/χ phases. These advancements will guide the design of new alloys maximizing high-temperature creep strength for advanced aerospace applications.

Egan, Ashton J. [Friedrich-Alexander University Er↗
DOE OSTI2025

Atomic-scale insights into topotactic transformations in an extra-large-pore zeolite using time-resolved 3D electron diffraction

Understanding the atomic-scale structural dynamics of phase transformations is crucial for developing materials and tailoring their properties. However, many materials are obtained as polycrystalline powders with large unit cells and/or complex structures, making it challenging to investigate detailed structural changes using conventional X-ray diffraction techniques. Here we employ time-resolved three-dimensional electron diffraction to reveal the topotactic reactions and transformations that convert the extra-large-pore silicate zeolite ECNU-45 into ECNU-46. ECNU-45 features three-dimensional interconnecting 24 × 10 × 10-ring channels, while ECNU-46 consists of one-dimensional 24-ring channels connected to 10-ring pockets. ECNU-45 and ECNU-46 are both examples of pure silicate zeolites with pore openings larger than 22-ring. Our findings indicate changes at six distinct tetrahedral silicon sites, involving atom displacement, addition and removal of framework atoms through bond breakage and formation. This work presents the synthesis of zeolites and also provides atomic-level insights into the dynamic processes of topotactic reactions. Our results have implications for advancing materials engineering and understanding complex solid-state reactions at an atomic scale.

Inorganic chemistry↗
DOE OSTI2026

Kinetic-controlled transformations of group-III arsenide nanocubes

Tracking the structural evolution of colloidal nanocrystals (NCs) facilitates the mechanistic studies of their materials chemistry. NC engineering via phase transformation reveals the chemical and physical determinants that drive lattice-scale dynamic processes such as cation exchange. Here, in this study, we employed NCs to demonstrate the cation exchange process from Cu 3 As to InAs and GaAs within nanocubes. The symmetry conversion in unit cells from cubic Cu 3 As to hexagonal InAs and GaAs can be described using a schematic cellular automaton model, which suggests a simplified cube-to-sphere transition. The strong covalent characteristics of III–V materials highlight the kinetic control that navigates the tailorable transformation through either an isotropic trajectory, leading to hollow structures, or a topotaxial trajectory, with abundant stacking faults. The reconstruction of complex covalent bonds is envisioned as the foundation for the synthesis of NCs.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2020

Transforming energy using quantum dots

Transforming and directing the flow of energy from one form (e.g., sunlight, electricity, etc.) to other useful forms of energy (e.g., electricity, chemical bonds, light, etc.) in an efficient and controllable manner is critical to meet the increasing energy demands and build a sustainable society. In search of such energy mediators, colloidal semiconductor nanocrystals, or quantum dots (QDs) are promising building blocks for building and designing systems that can efficiently capture light and convert and direct that energy into other useful forms of energy. In this work, we summarize recent advances using QDs in energy conversion architectures with the express goal of converting optical energy to other forms of energy, including electricity (i.e., photovoltaics) , photons with different energies (i.e., photo up- or downconversion), and chemical bonds (i.e., photocatalysis). The advantages of employing QDs over molecular chromophores in absorbing and then directing and converting optical energy are highlighted. Finally, we discuss ongoing challenges as well as unique opportunities associated with the use of QDs for transforming energy.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2022

Achieving a blue-excitable yellow-emitting Ca-LMOF phosphor via water induced phase transformation

Luminescent metal-organic frameworks (LMOFs) with diverse structural features and promising fluorescence-based applications have attracted wide attention in the past two decades. In this work, a LMOF with the formula [Ca 4 (tcbpe-F) 2 (H 2 O) 3 ] (1, LMOF-411) has been constructed from calcium (Ca) and 1,1,2,2-tetrakis(4-(4-carboxyphenyl)phenyl)ethene (H 4 tcbpe-F). Compound 1 features a three-dimensional framework with a 10-nodal net topology. Due to the relatively high hydration energy of Ca 2+ , compound 1 readily transforms into a new phase formulated as [Ca(H 2 tcbpe-F)(H 2 O) 2 ] (1') upon exposure to water. Combining experimental characterization and theoretical calculations, we elucidated the mechanism of H 2 O-induced phase transition from 1 to 1'. Notably, the water induced phase transformation can be detected visibly from the change in luminescence, which originates from the fluorescent linker. Compound 1 emits green light ( λ em = 490 nm) under UV excitation, while compound 1' emits bright yellow light ( λ em = 550 nm) under blue excitation (450 nm). Compound 1' represents the first Ca based LMOF yellow phosphor and its luminescence quantum yield reaches 68%. It can be coated directly onto a commercial blue light-emitting-diode (LED) chip to fabricate a white LED (WLED).

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Mechanochemistry-driven phase transformation of crystalline covalent triazine frameworks assisted by alkaline molten salts

We report covalent triazine frameworks (CTFs) have shown wide applications in the fields of separation, catalysis, energy storage, and beyond. However, it is a long-term challenging subject to fabricate high-quality CTF materials via facile procedures. Herein, a mechanochemistry-driven procedure was developed to achieve phase transformation of crystalline CTFs assisted by alkaline molten salts. The transformation of CTF-1 from staggered AB to eclipsed AA stacking mode was achieved by short time (30 min) mechanochemical treatment in the presence of molten salts composed of LiOH/KOH, generating high-quality CTF-1 material with high crystallinity, high surface area (625 m 2 g -1 ), and permanent/ordered porosity without carbonization under ambient conditions. This facile procedure could be extended to provide nanoporous three-dimensional CTF materials.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗