Search NASA⌕ Search

SEARCH · Search NASA

Results for “experimental implementation”

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 865 records · Page 48

Sage: parallel semi-asymmtric graph algorithms for NVRAMs

Non-volatile main memory (NVRAM) technologies provide an attractive set of features for large-scale graph analytics, including byte-addressability, low idle power, and improved memory-density. NVRAM systems today have an order of magnitude more NVRAM than traditional memory (DRAM). NVRAM systems could therefore potentially allow very large graph problems to be solved on a single machine, at a modest cost. However, a significant challenge in achieving high performance is in accounting for the fact that NVRAM writes can be much more expensive than NVRAM reads. In this paper, we propose an approach to parallel graph analytics using the Parallel Semi-Asymmetric Model (PSAM), in which the graph is stored as a read-only data structure (in NVRAM), and the amount of mutable memory is kept proportional to the number of vertices. Similar to the popular semi-external and semi-streaming models for graph analytics, the PSAM approach assumes that the vertices of the graph fit in a fast read-write memory (DRAM), but the edges do not. In NVRAM systems, our approach eliminates writes to the NVRAM, among other benefits. To experimentally study this new setting, we develop Sage, a parallel semi-asymmetric graph engine with which we implement provably-efficient (and often work-optimal) PSAM algorithms for over a dozen fundamental graph problems. We experimentally study Sage using a 48--core machine on the largest publicly-available real-world graph (the Hyperlink Web graph with over 3.5 billion vertices and 128 billion edges) equipped with Optane DC Persistent Memory, and show that Sage outperforms the fastest prior systems designed for NVRAM. Importantly, we also show that Sage nearly matches the fastest prior systems running solely in DRAM, by effectively hiding the costs of repeatedly accessing NVRAM versus DRAM.

97 MATHEMATICS AND COMPUTING↗

Validation of CTF Droplet Entrainment and Annular/Mist Closure Models using Riso Steam/Water Experiments

This report summarizes the work done to validate the droplet entrainment and de-entrainment models as well as two-phase closure models in the CTF code by comparison with experimental data obtained at Riso National Laboratory. The Riso data included a series of over 250 steam/water experiments that were performed in both tube and annulus geometries over a range of various pressures and outlet qualities. Experimental conditions were set so that the majority of cases were in the annular/mist ow regime. Measurements included liquid lm ow rate, droplet ow rate, lm thickness, and two-phase pressure drop. CTF was used to model 180 of the tubular geometry cases, matching experimental geometry, outlet pressure, and outlet ow quality to experimental values. CTF results were compared to the experimental data at the outlet of the test section in terms of vapor and entrained liquid ow fractions, pressure drop per unit length, and liquid lm thickness. The entire process of generating CTF input decks, running cases, extracting data, and generating comparison plots was scripted using Python and Matplotlib for a completely automated validation process. All test cases and scripting tools have been committed to the COBRA-TF master repository and selected cases have been added to the continuous testing system to serve as regression tests. The dierences between the CTF- and experimentally-calculated ow fraction values were con- sistent with previous calculations by Wurtz, who applied the same entrainment correlation to the same data. It has been found that CTF's entrainment/de-entrainment predictive capability in the annular/mist ow regime for this particular facility is comparable to the licensed industry code, COBRAG. While lm and droplet predictions are generally good, it has been found that accuracy is diminished at lower ow qualities. This nding is consistent with the noted deciencies in the Wurtz entrainment model employed by CTF. The CTF predicted two-phase pressure drop in the annular/mist ow regime has been found to be highly inaccurate, exhibiting a clear bias with respect to the experimental data. This inaccuracy led to an investigation that revealed deciencies in the implementation of the annular/mist interfacial friction model, which should be investigated further in the future. Looking to published COBRAG results for this same facility reveal it exhibits no bias with regard to experimental pressure drop results. In addition to the problems with pressure drop prediction, the lm thickness was also signicantly under-predicted by CTF compared to both experimental data and Wurtz's analytical calculations. Film thickness is calculated using a simple geometric relationship and lm void fraction in CTF, which is dependent on slip ratio and interfacial friction. It is possible that the issues aecting the pressure drop and lm void prediction are related.

97 MATHEMATICS AND COMPUTING↗

Rate constants of chemical reactions from semiclassical transition state theory in full and one dimension

Semiclassical Transition State Theory (SCTST), a method for calculating rate constants of chemical reactions, offers gains in computational efficiency relative to more accurate quantum scattering methods. In full-dimensional (FD) SCTST, reaction probabilities are calculated from third and fourth potential derivatives along all vibrational degrees of freedom. However, the computational cost of FD SCTST scales unfavorably with system size, which prohibits its application to larger systems. In this study, the accuracy and efficiency of 1-D SCTST, in which only third and fourth derivatives along the reaction mode are used, are investigated in comparison to those of FD SCTST. Potential derivatives are obtained from numerical ab initio Hessian matrix calculations at the MP2/cc-pVTZ level of theory, and Richardson extrapolation is applied to improve the accuracy of these derivatives. Reaction barriers are calculated at the CCSD(T)/cc-pVTZ level. Results from FD SCTST agree with results from previous theoretical and experimental studies when Richardson extrapolation is applied. Results from our implementation of 1-D SCTST, which uses only 4 single-point MP2/cc-pVTZ energy calculations in addition to those for conventional TST, agree with FD results to within a factor of 5 at 250 K. This degree of agreement and the efficiency of the 1-D method suggest its potential as a means of approximating rate constants for systems too large for existing quantum scattering methods.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Mathematical nuances of Gaussian process-driven autonomous experimentation

Abstract The fields of machine learning (ML) and artificial intelligence (AI) have transformed almost every aspect of science and engineering. The excitement for AI/ML methods is in large part due to their perceived novelty, as compared to traditional methods of statistics, computation, and applied mathematics. But clearly, all methods in ML have their foundations in mathematical theories, such as function approximation, uncertainty quantification, and function optimization. Autonomous experimentation is no exception; it is often formulated as a chain of off-the-shelf tools, organized in a closed loop, without emphasis on the intricacies of each algorithm involved. The uncomfortable truth is that the success of any ML endeavor, and this includes autonomous experimentation, strongly depends on the sophistication of the underlying mathematical methods and software that have to allow for enough flexibility to consider functions that are in agreement with particular physical theories. We have observed that standard off-the-shelf tools, used by many in the applied ML community, often hide the underlying complexities and therefore perform poorly. In this paper, we want to give a perspective on the intricate connections between mathematics and ML, with a focus on Gaussian process-driven autonomous experimentation. Although the Gaussian process is a powerful mathematical concept, it has to be implemented and customized correctly for optimal performance. We present several simple toy problems to explore these nuances and highlight the importance of mathematical and statistical rigor in autonomous experimentation and ML. One key takeaway is that ML is not, as many had hoped, a set of agnostic plug-and-play solvers for everyday scientific problems, but instead needs expertise and mastery to be applied successfully. Graphical abstract

97 MATHEMATICS AND COMPUTING↗

Calculated fission-fragment yield systematics in the region 74 <=Z <= 94 and 90 <=N <= 150

Background: In the seminal experiment by Schmidt et al . [Nucl. Phys. A 665 , 221 (2000)] in which fission-fragment charge distributions were obtained for 70 nuclides, asymmetric distributions were seen above nucleon number A ≈ 226 and symmetric ones below. Because asymmetric fission had often loosely been explained as a preference for the nucleus to always exploit the extra binding of fragments near ¹³²Sn it was assumed that all systems below A ≈ 226 would fission symmetrically because available isotopes do not have a proton-to-neutron Z/N ratio that allows division into fragments near ¹³²Sn. But the finding by Andreyev et al . [Phys. Rev. Lett. 105 , 252502 (2010)] did not conform to this expectation because the compound system ¹⁸⁰Hg was shown to fission asymmetrically. It was suggested that this was a new type of asymmetric fission, because no strong shell effects occur for any possible fragment division. Purpose: We calculate a reference database for fission-fragment mass yields for a large region of the nuclear chart comprising 987 nuclides. A particular aim is to establish whether ¹⁸⁰Hg is part of a contiguous region of asymmetric fission, and if so, its extent, or if not, in contrast to the actinides, there are scattered smaller groups of nuclei that fission asymmetrically in this area of the nuclear chart. Methods: We use the by now well benchmarked Brownian shape-motion method and perform random walks on the previously calculated five-dimensional potential-energy surfaces. The calculated shell corrections are damped out with energy according to a prescription developed earlier. Results: We have obtained a theoretical reference database of fission-fragment mass yields for 987 nuclides. These results show an extended region of asymmetric fission with approximate extension 74 ≤ Z ≤ 85 and 100 ≤ N ≤ 120. The calculated yields are highly variable. We show 20 representative plots of these variable features and summarize the main aspects of our results in terms of “nuclear-chart” plots showing calculated degrees of asymmetry versus N and Z. Conclusions: Experimental data in this region are rare: only ten or so yield distributions have been measured, some with very limited statistics. We agree with several measurements with higher statistics. Regions where there might be differences between our calculated results and measurements lie near the calculated transition line between symmetric and asymmetric fission. To draw more definite conclusions about the accuracy of the present implementation of the Brownian shape-motion approach in this region experimental data, with reliable statistics, for a fair number of suitably located additional nuclides are clearly needed. Because the nuclear potential-energy structure is so different in this region compared to the actinide region, additional experimental data together with fission theory studies that incorporate additional, dynamical aspects should provide much new insight.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

The Impact of a Fault Tolerant MPI on Scalable Systems Services and Applications

Exascale targeted scientific applications must be prepared for a highly concurrent computing environment where failure will be a regular event during execution. Natural and algorithm-based fault tolerance (ABFT) techniques can often manage failures more efficiently than traditional checkpoint/restart techniques alone. Central to many petascale applications is an MPI standard that lacks support for ABFT. The Run-Through Stabilization (RTS) proposal, under consideration for MPI 3, allows an application to continue execution when processes fail. The requirements of scalable, fault tolerant MPI implementations and applications will stress the capabilities of many system services. System services must evolve to efficiently support such applications and libraries in the presence of system component failures. This paper discusses how the RTS proposal impacts system services, highlighting specific requirements. Early experimentation results from Cray systems at ORNL using prototype MPI and runtime implementations are presented. Additionally, this paper outlines fault tolerance techniques targeted at leadership class applications.

99 GENERAL AND MISCELLANEOUS↗

Attention-Augmented Parametric Kernel Graph Neural Network (APKGNN) for Node Classification

We present a new graph neural network, the Attention-based Parametric-Kernel augmented Graph Neural Network (APKGNN), developed for node classification tasks. Despite extensive work on modeling multi-faceted relationships between connected nodes of a graph, the effect of attention on edge features mapped to relationships has not yet been analyzed through learning representation. This study derives such an attention vector by first calculating node features corresponding to endpoints of an edge and then aggregating these with extracted local intrinsic patches of a given graph to generate augmented local patch vectors. This process uses a parametric kernel based on Gaussian mixture models (GMMs) to embed local neighborhoods of the graph in local patches. The patch vectors then convolve with the above node features to produce an updated node representation. We show that this new learning representation (APKGNN) achieves higher node classification accuracy on tasks - both standard benchmarks (Cora, PubMed, Citeseer) and new experimental short text corpora where nodes correspond to text documents and words. This implementation of the GNN convolution layer outperforms state-of-the-art (SOTA) algorithms, achieving higher training, validation, and test accuracy by a significant margin on three standard benchmark data sets under both SOTA experimental settings and those for new testbeds.

Bose, Avishek↗

Broken symmetry dielectric resonators for high quality factor Fano metasurfaces

We present a new approach to dielectric metasurface design that relies on a single resonator per unit cell and produces robust, high quality factor Fano resonances. Our approach utilizes symmetry breaking of highly symmetric resonator geometries, such as cubes, to induce couplings between the otherwise orthogonal resonator modes. In particular, we design perturbations that couple “bright” dipole modes to “dark” dipole modes whose radiative decay is suppressed by local field effects in the array. Our approach is widely scalable from the near-infrared to radio frequencies. We first unravel the Fano resonance behavior through numerical simulations of a germanium resonator-based metasurface that achieves a quality factor of ~1300 at ~10.8 μm. Then, we present two experimental demonstrations operating in the near-infrared (~1 μm): a silicon-based implementation that achieves a quality factor of ~350; and a gallium arsenide-based structure that achieves a quality factor of ~600, the highest near-infrared quality factor experimentally demonstrated to date with this kind of metasurface. Importantly, large electromagnetic field enhancements appear within the resonators at the Fano resonant frequencies. Here, we envision that combining high quality factor, high field enhancement resonances with nonlinear and active/gain materials such as gallium arsenide will lead to new classes of active optical devices.

36 MATERIALS SCIENCE↗

VTS Dynamic Model and Experimental Validation for Vibration Suppression

High Q-factor superconducting radio-frequency (SRF) cavities are extremely sensitive to mechanical vibrations, which can cause an uncontrolled shift of the electromagnetic resonance frequency, with detrimental effects on the cavity performance. To avoid this phenomenon (i.e. microphonics) the dynamic behavior of the assembly must be studied, to design vibration suppression systems. In this paper, the numerical model of the Vertical Test Stand (VTS) insert adopted at the Fermi National Accelerator Laboratory for cold tests was studied, in the framework of the experimental characterization of Dark SRF cavities conducted by the SQMS (Superconducting Quantum Materials and Systems) center. Due to the complexity of the assembly, experimental activity is of utmost importance to validate the model. Thus, vibration testing was performed on the actual system, and a good correlation between numerical and experimental results was found. Additionally, two different setups were compared, with and without implementing a preliminary passive vibration suppression system, to assess its effectiveness in reducing the microphonics phenomenon. The obtained results allowed for the design of an improved isolation system, which will be the subject of future experimental testing.

43 PARTICLE ACCELERATORS↗

A pulsed-power implementation of “Laser Gate” for increasing laser energy coupling and fusion yield in magnetized liner inertial fusion (MagLIF)

Magnetized Liner Inertial Fusion (MagLIF) at Sandia National Laboratories involves a laser preheating stage,where a few-ns laser pulse passes through a few-micron-thick plastic window to preheat gaseous fusion fuel contained within the MagLIF target. Interactions with this window reduce heating e ciency and mix window and target materials into the fuel. A recently proposed idea called \Laser Gate" involves removing the window well before the preheating laser is applied. In this article, we introduce experimental proof-of-principle results for a pulsed-power implementation of Laser Gate, where a thin current-carrying wire weakens the perimeter of the window, allowing the fuel pressure to push the window open and away from the preheating laser path. For this e ort, transparent targets were fabricated and a test facility capable of studying this version of Laser Gate was developed. A 12-frame bright- eld laser schlieren/shadowgraphy imaging system captured the window opening dynamics on microsecond timescales. The images exhibit that the window remains largely intact as it opens and detaches from the target. A column of escaping pressurized gas appears to prevent the detached window from inadvertently moving into the preheating laser path.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Protocols for x-ray transient grating pump/optical probe experiments at x-ray free electron lasers

Abstract Transient grating spectroscopy is a specialized application of the four-wave-mixing methodology and constitutes a versatile technique for investigating the dynamics of vibrational, magnetic and electronic degrees of freedom of matter in a background-free fashion. Recent developments in free-electron laser sources have enabled the extension of this technique into the extreme ultraviolet range. Ongoing efforts to expand transient grating spectroscopy into the x-ray regime promise numerous advantages: (1) substantial penetration depths that allow for probing bulk material properties, (2) element specificity via specific core-excited states, and (3) short wavelengths that allow for excitation gratings with higher momentum transfer and improved spatial resolution. In this study, we comprehensively outline the procedures for conducting x-ray transient grating pump/optical probe experiment. The process encompasses the design and alignment of the experimental setup, as well as the subsequent steps involved in data acquisition and analysis. This paper is intended as a comprehensive guide for researchers interested in implementing x-ray transient grating spectroscopy, providing valuable insights into the intricacies of the experimental workflow required for this novel technique. Furthermore, we discuss the potential for extending this methodology to an x-ray pump/x-ray probe scheme, envisioning a future direction that holds promise for enhancing the capabilities and scope of x-ray transient grating spectroscopy, opening new opportunities for studying ultrafast processes with unprecedented temporal and spatial resolutions.

Optics↗

Dose response of alanine detectors irradiated with carbon ion beams

Purpose: The dose response of the alanine detector shows a dependence on particle energy and type when irradiated with ion beams. The purpose of this study is to investigate the response behavior of the alanine detector in clinical carbon ion beams and compare the results to model predictions. Methods: Alanine detectors have been irradiated with carbon ions with an energy range of 89-400 MeV/u. The relative effectiveness of alanine has been measured in this regime. Pristine and spread out Bragg peak depth-dose curves have been measured with alanine dosimeters. The track structure based alanine response model developed by Hansen and Olsen has been implemented in the Monte Carlo code FLUKA and calculations were compared to experimental results. Results: Calculations of the relative effectiveness deviate less than 5% from the measured values for monoenergetic beams. Measured depth-dose curves deviate from predictions in the peak region, most pronounced at the distal edge of the peak. Conclusions: The used model and its implementation show a good overall agreement for quasimonoenergetic measurements. Deviations in depth-dose measurements are mainly attributed to uncertainties of the detector geometry implemented in the Monte Carlo simulations.

60 APPLIED LIFE SCIENCES↗

Comparison of plug flow and multi-node stratified tank modeling approaches regarding computational efficiency and accuracy

Residential water heaters contain water stratified by temperature-driven density differences. This implies that a water tank can reach a state in which the top and bottom sections have different temperatures, unless mixing happens. A high degree of thermal stratification can improve the efficiency of some water heaters, by saving the amount of energy required for the heat-up process. Studies of stratification became popular in the 1970s and it remains an active research topic today. The research has led to the development of different models and techniques to better predict and define a stratified tanks behavior. By comparing these models and techniques used previously to describe thermal stratification, the phenomenon could be better understood, exploited, and used to increase efficiency and thermal energy capacity in modern water tanks. From the existing models, we found the one-dimensional standard plug-flow and a multi node model to be appropriate for analyzing the processes of the heat up and cool-down in a water tank. These two models are based on energy balances. This work involved comparing the accuracy and computational effort needed to implement these models. To assess accuracy, we compared both types of existing models to experimental data (also collected in this work) which included a heat up process using an external heat pump. This external process included a layering process that has an eddy diffusivity at five times the rate of thermal diffusion. For this project, we implemented the models in MATLAB, the multi-paradigm numerical computing environment. We quantified model accuracy using the root mean squared error between modeled data and experimental data for six measured tank temperatures. Comparing the accuracy and the computational time taken to run the simulation provides a method to contrast the performance of each model and a way to rate it. The multi node model was run using from 6 to 96 spatial nodes; the plug flow model was run using 1 to 0.001 º C temperature bin sizes. Additionally, timesteps were varied from 4 to 236 s. The results quantify the tradeoff between accuracy and computational time, providing guidance for simulations to intelligently select the best model type and simulation parameters. This research can be used to validate the pre-existing models and possibly improve the modern water tank.

Bulnes, Fernando Karg↗

Most theoretical approaches used in nuclear astrophysics to model the nucleosynthesis of heavy elements incorporate the so-called statistical model in order to describe the excitation and decay properties of atomic nuclei. One of the basic assumptions of this model is the validity of the Brink–Axel hypothesis and the related concept of so-called photon strength functions to describe γ-ray transition probabilities. We present a novel experimental approach that allows for the first time to experimentally determine the photon strength function simultaneously in two independent ways by a unique combination of quasi-monochromatic photon beams and a newly implemented γ–γ coincidence setup. This technique does not assume a priori the validity of the Brink–Axel hypothesis and sets a benchmark in terms of the detection sensitivity for measuring decay properties of photo-excited states below the neutron separation energy. The data for the spherical off-shell nucleus 128 Te were obtained for γ-ray beam-energy settings between 3 MeV and 9 MeV in steps of 130 keV for the lower beam energies and in steps of up to 280 keV for the highest beam settings. We present a quantitative analysis on the consistency of the derived photon strength function with the Brink–Axel hypothesis. The data clearly demonstrate a discrepancy of up to a factor of two between the photon strength functions extracted from the photoabsorption and photon emission process, respectively. In addition, we observe that the photon strength functions are not independent of the excitation energy, as usually assumed. Thus, we conclude, that the Brink–Axel hypothesis is not strictly fulfilled in the excitation-energy region below the neutron separation threshold (S n = 8.78 MeV) for the studied case of 128 Te.

Not Available

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Improvements to CTF Closure Models for Modelling of Two-Phase Flow

This report documents the efforts to improve the CTF prediction of void fraction and two-phase pressure drop by improving two-phase closure models. Previous validation activities have revealed that CTF tends to overpredict void fraction and two-phase pressure drop. In response to this, two approaches were taken to improve CTF’s predictive capabilities. Based on findings that the interfacial drag and subcooled boiling models significantly impact void prediction, alternative closure models for these physical effects were found and implemented into the code. An extensive assessment was performed by using the existing and newly added two-phase experimental data, which showed that void prediction, wall temperature, and two-phase pressure-drop results are improved by using the newly implemented models. A second approach for improving CTF modeling accuracy involved the use of a Bayesian calibration process that uses CTF validation data to optimize selected modeling coefficients and closure model multipliers to achieve a more accurate prediction of experimental results. A similar assessment was performed with calibrated models that improved the void and pressure-drop prediction for test cases that were both included and not included in the calibration dataset. Results of this study will be used to change the models used in CTF to achieve more accurate BWR analyses moving forward. This study also identified opportunities for improving additional closure models and other opportunities to use calibration techniques to improve CTF.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Increasing of MERARG experimental performances: on-line fission gas release measurement by mass spectrometry

The MERARG device - implemented at the LECASTAR Hot Laboratory, at the CEA Cadarache - allows characterizing nuclear fuels with respect to the behaviour of fission gases during thermal transients representative of normal or off normal operating nuclear power plant conditions. The fuel is heated in order to extract a part or the total gas inventory it contains. Fission Gas Release (FGR) is actually recorded by mean of both on-line gamma spectrometry station and micro gas chromatography. These two devices monitor the quantity and kinetics of fission gas release rate. They only address {sup 85}Kr radioactive isotope and the elemental quantification of Kr, Xe and He (with a relatively low detection limit in the latter case, typically 5-10 ppm). In order to better estimate the basic mechanisms that promote fission gas release from irradiated nuclear fuels, the CEA fuel study department decided to improve its experimental facility by modifying MERARG to extend the studies of gamma emitter fission gases to all gases (including Helium) with a complete isotopic distribution capability. To match these specifications, a Residual Gas Analyser (RGA) has been chosen as mass spectrometer. This paper presents a review of the main aspects of the qualification/calibration phase of the RGA type analyser. In particular, results recorded over three mass ranges 1-10 u, 80-90 u and 120-140 u in the two classical modes of MERARG, i.e. on-line and off-line measurements are discussed. Results obtained from a standard gas bottle show that the quantitative analysis at a few ppm levels can be achieved for all isotopes of Kr and Xe, as well as masses 2 and 4 u. (authors)

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Considerations for a PDV diagnostic capability on the National Ignition Facility

Photonic Doppler Velocimetry (PDV) has become widely and routinely used in many high-velocity experimental applications due to its improved ease of use, cost, experimental flexibility, data return, and robustness compared to earlier velocimetric methods. However, these earlier methods have advantages in applications with requirements beyond PDV’s current capabilities. Various classes of experiments at the National Ignition Facility (NIF) that are characterized by extremely high velocity or acceleration, or diagnostic requirements for high precision in timing and/or velocity, have historically seen a VISAR (velocity interferometer system for any surface) diagnostic employed due to such advantages. VISAR, however, requires specific, and sometimes challenging, experimental features, including planar geometry and normal incidence, high-reflectivity surface treatment, and a relatively large and inflexible diagnostic footprint. Therefore, the potential for implementing a PDV diagnostic at NIF has been evaluated by researchers from National Security Technologies, LLC and Lawrence Livermore National Laboratory. We present the results of this study, weigh the relative merits of the two methodologies with consideration of experimental phenomena and requirements, and discuss possible implementations and future directions.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

DOE STTR Phase I Final Report Report: Machine-learning Based Prediction of Thermal Limits for Conjugated Organic Materials

The SCANN-DT Phase I STTR project led by NLM Photonics and partnering with the National Renewable Energy Laboratory (NREL) sought to apply machine learning techniques based on graph neural networks (GNNs) towards the prediction of decomposition temperatures (Td) of organic semiconductors, based on prior work on bond dissociation energy (BDE) prediction as implemented in NREL’s ALFABET prediction tool. Using a curated set of experimental decomposition energies, the project examined GNN-based, classical quantitative structure-property relationship (QSPR) based on DFT calculations, and combinations of both methods to predict Td. While the best MAEs in Td achieved were near 40°C, below project targets, the project led to improvements in the ALFABET model, improvements in cloud-based implementations of NWChem software, and a substantial dataset of calculations on medium-sized conjugated organic molecules.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗