Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “model structural error”

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 613 records · Page 34
All sourcesNASA NTRSDOE OSTI
DOE OSTI2018

A biologically-inspired approach for adaptive control of advanced energy systems

Here in this article, a novel approach is proposed for integrating a Biologically-Inspired Optimal Control Strategy (BIOsingle bondCS) with an Artificial Neural Network (ANN)-based adaptive component for advanced energy systems applications. Specifically, BIOsingle bondCS employs gradient-based optimal control solvers in a biologically-inspired manner, following the rule of pursuit for ants, to simultaneously control multiple process outputs at their desired setpoints. Also, the ANN component captures the mismatch between the controller and the plant models by using a single-hidden-layer technique with online learning capabilities to augment the baseline BIOsingle bondCS control laws. The resulting approach is a unique combination of biomimetic control and data-driven methods that provides optimal solutions for dynamic systems. The applicability of the proposed framework is illustrated via an Integrated Gasification Combined Cycle (IGCC) process with carbon capture as an advanced energy system example. In particular, a multivariable control structure associated with a subsystem of the IGCC plant simulation in DYNSIM® is addressed. The proposed control laws are derived in MATLAB®, while the plant models are built in DYNSIM®, and a previously developed MATLAB®-DYNSIM® link is employed for implementation purposes. The proposed integrated approach improves the overall performance of the process up to 85% in terms of reducing the output tracking error when compared to stand-alone BIOsingle bondCS and Proportional-Integral (PI) controller implementations, resulting in faster setpoint tracking. The proposed framework thus provides a promising alternative for advanced control of energy systems of the future.

42 ENGINEERING↗
DOE OSTI2024

Transformer Neural Networks with Spatiotemporal Attention for Predictive Control and Optimization of Industrial Processes

In the context of real-time optimization and model predictive control of industrial systems, machine learning, and neural networks represent cutting-edge tools that hold promise for enhancing dynamic modeling. This work presents a novel transformer neural network architecture for real-time optimization and model predictive control. This network design includes a modified attention mechanism inspired by positional embedding attention from vision transformers and task-specific modifications to the input-output structure of the transformer’s decoder stack. Experiments were conducted using data from a 450 MW coal-fired power plant to evaluate this approach's effectiveness. The transformer neural network was compared with conventional recurrent models, including GRU and LSTM. The transformer exhibited a 6% increase in the R-squared (R2) value of predictions and an 83% reduction in mean squared error (MSE). Computation time was also reduced by 84% compared to conventional recurrent models.

Gallup, Ethan R.↗
DOE OSTI2012

XCOM intrinsic dimensionality for low-Z elements at diagnostic energies

Purpose: To determine the intrinsic dimensionality of linear attenuation coefficients (LACs) from XCOM for elements with low atomic number (Z = 1-20) at diagnostic x-ray energies (25-120 keV). H{sub 0}{sup q}, the hypothesis that the space of LACs is spanned by q bases, is tested for various q-values. Methods: Principal component analysis is first applied and the LACs are projected onto the first q principal component bases. The residuals of the model values vs XCOM data are determined for all energies and atomic numbers. Heteroscedasticity invalidates the prerequisite of i.i.d. errors necessary for bootstrapping residuals. Instead wild bootstrap is applied, which, by not mixing residuals, allows the effect of the non-i.i.d residuals to be reflected in the result. Credible regions for the eigenvalues of the correlation matrix for the bootstrapped LAC data are determined. If subsequent credible regions for the eigenvalues overlap, the corresponding principal component is not considered to represent true data structure but noise. If this happens for eigenvalues l and l + 1, for any l{<=}q, H{sub 0}{sup q} is rejected. Results: The largest value of q for which H{sub 0}{sup q} is nonrejectable at the 5%-level is q = 4. This indicates that the statistically significant intrinsic dimensionality of low-Z XCOM data at diagnostic energies is four. Conclusions: The method presented allows determination of the statistically significant dimensionality of any noisy linear subspace. Knowledge of such significant dimensionality is of interest for any method making assumptions on intrinsic dimensionality and evaluating results on noisy reference data. For LACs, knowledge of the low-Z dimensionality might be relevant when parametrization schemes are tuned to XCOM data. For x-ray imaging techniques based on the basis decomposition method (Alvarez and Macovski, Phys. Med. Biol. 21, 733-744, 1976), an underlying dimensionality of two is commonly assigned to the LAC of human tissue at diagnostic energies. The finding of a higher statistically significant dimensionality thus raises the question whether a higher assumed model dimensionality (now feasible with the advent of multibin x-ray systems) might also be practically relevant, i.e., if better tissue characterization results can be obtained.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2012

Constraining primordial magnetic fields with future cosmic shear surveys

The origin of astrophysical magnetic fields observed in galaxies and clusters of galaxies is still unclear. One possibility is that primordial magnetic fields generated in the early Universe provide seeds that grow through compression and turbulence during structure formation. A cosmological magnetic field present prior to recombination would produce substantial matter clustering at intermediate/small scales, on top of the standard inflationary power spectrum. In this work we study the effect of this alteration on one particular cosmological observable, cosmic shear. We adopt the semi-analytic halo model in order to describe the non-linear clustering of matter, and feed it with the altered mass variance induced by primordial magnetic fields. We find that the convergence power spectrum is, as expected, substantially enhanced at intermediate/small angular scales, with the exact amplitude of the enhancement depending on the magnitude and power-law index of the magnetic field power spectrum. Specifically, for a fixed amplitude, the effect of magnetic fields is larger for larger spectral indices. We use the predicted statistical errors for a future wide-field cosmic shear survey, on the model of the ESA Cosmic Vision mission Euclid, in order to forecast constraints on the amplitude of primordial magnetic fields as a function of the spectral index. We find that the amplitude will be constrained at the level of ∼ 0.1 nG for n{sub B} ∼ −3, and at the level of ∼ 10{sup −7} nG for n{sub B} ∼ 3. The latter is at the same level of lower bounds coming from the secondary emission of gamma-ray sources, implying that for high spectral indices Euclid will certainly be able to detect primordial magnetic fields, if they exist. The present study shows how large-scale structure surveys can be used for both understanding the origins of astrophysical magnetic fields and shedding new light on the physics of the pre-recombination Universe.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Predictive links between microbial communities and biological oxygen utilization in the Arctic Ocean

Microbial metabolism influences rates of net community production (NCP), exerting a direct biological control on marine oxygen and carbon fluxes. In the Arctic, it is increasingly important to understand and quantify this process, as ecological and oceanographic conditions shift due to changing climate. Here, we describe potential ecological links between pelagic microbial diversity and an NCP precursor, biological oxygen utilization, using machine learning and paired observations of community structure and metabolic activity from a seasonally and spatially variable transect of the Arctic Ocean (2019–2020 MOSAiC Expedition). Community structure was determined using 16S (prokaryotic) and 18S (eukaryotic) rRNA gene amplicon sequencing, and metabolic activity was derived from ΔO 2 /Ar. Using self-organizing maps, we identified clear successional patterns in observed microbial community structure that were seasonally driven in the upper ocean and vertically stratified with depth. Metabolic activity was also stratified, with a primarily net heterotrophic water column (median −1.5% biological oxygen saturation), excepting periodic oxygen supersaturation (maximum: 13.6%) within the mixed layer. Using DNA sequences as predictor variables, we then constructed a random forest regression model that reliably reconstructed biological oxygen concentrations (root mean squared error = 4.14 μmol kg −1 ). Top predictors from this model were from heterotrophic (bacteria) or potentially mixotrophic (dinoflagellate) taxa. These analyses highlight biologically driven diagnostic tools that can be used to expand biogeochemical datasets and improve the microbial perspectives and metabolisms represented in ecological models of net productivity and carbon flux in a changing Arctic Ocean.

Chamberlain, Emelia J. [Univ. of San Diego, San Di↗
DOE OSTI2017

Life cycle assessment based environmental impact estimation model for pre-stressed concrete beam bridge in the early design phase

The late rise in global concern for environmental issues such as global warming and air pollution is accentuating the need for environmental assessments in the construction industry. Promptly evaluating the environmental loads of the various design alternatives during the early stages of a construction project and adopting the most environmentally sustainable candidate is therefore of large importance. Yet, research on the early evaluation of a construction project's environmental load in order to aid the decision making process is hitherto lacking. In light of this dilemma, this study proposes a model for estimating the environmental load by employing only the most basic information accessible during the early design phases of a project for the pre-stressed concrete (PSC) beam bridge, the most common bridge structure. Firstly, a life cycle assessment (LCA) was conducted on the data from 99 bridges by integrating the bills of quantities (BOQ) with a life cycle inventory (LCI) database. The processed data was then utilized to construct a case based reasoning (CBR) model for estimating the environmental load. The accuracy of the estimation model was then validated using five test cases; the model's mean absolute error rates (MAER) for the total environmental load was calculated as 7.09%. Such test results were shown to be superior compared to those obtained from a multiple-regression based model and a slab area base-unit analysis model. Henceforth application of this model during the early stages of a project is expected to highly complement environmentally friendly designs and construction by facilitating the swift evaluation of the environmental load from multiple standpoints. - Highlights: • This study is to develop the model of assessing the environmental impacts on LCA. • Bills of quantity from completed designs of PSC Beam were linked with the LCI DB. • Previous cases were used to estimate the environmental load of new case by CBR model. • CBR model produces more accurate estimations (7.09%) than other conventional models. • This study supports decision making process in the early stage of a new construction case.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

Redshift-space streaming velocity effects on the Lyman- α forest baryon acoustic oscillation scale

The baryon acoustic oscillation (BAO) scale acts as a standard ruler for measuring cosmological distances and has therefore emerged as a leading probe of cosmic expansion history. However, any physical effect that alters the length of the ruler can lead to a bias in our determination of distance and expansion rate. One of these physical effects is the streaming velocity, the relative velocity between baryons and dark matter in the early Universe, which couples to the BAO scale due to their common origin in acoustic waves at recombination. In this work, we investigate the impact of streaming velocity on the BAO feature of the Lyman- α forest auto-power spectrum, one of the main tracers being used by the recently commissioned Dark Energy Spectroscopic Instrument (DESI). To do this, we develop a perturbative model for Lyman- α flux fluctuations which is complete to second order for a certain set of fields, and applicable to any redshift-space tracer of structure since it is based only on symmetry considerations. We find that there are 8 biasing coefficients through second order. We find streaming velocity-induced shifts in the BAO scale of 0.081%–0.149% (transverse direction) and 0.053%–0.058% (radial direction), depending on the model for the biasing coefficients used. These are smaller than, but not negligible compared to, the DESI Lyman- α BAO error budget, which is 0.46% on the overall scale. Overall, the sensitivity of these results to our choice of bias parameters underscores the need for future work to measure the higher-order biasing coefficients from simulations, especially for future experiments beyond DESI.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Glass formulation and composition optimization with property models: A review

Abstract Glass is a versatile material with a remarkable history and many practical applications. It plays a critical role in our everyday lives, the advancement of science, and the development of many technologies. The Edisonian type trial‐and‐error method was commonly used for conventional design of glass compositions, which was time‐consuming and costly. With the urgent need to develop new glass compositions for technology applications rapidly, it has become necessary to develop precise property models with predictive powers using large databases and efficient formulation approaches. This paper reviews the design of glass compositions using these analytical and numerical models of composition–structure–property relations of glasses, some based on large databases and machine learning approaches. Aspects of data collection, model fitting, feature extraction, model evaluation, and uncertainty quantification will be covered. Furthermore, advances in the glass optimization framework and available tools are summarized with examples. The outlook and perspective for further glass property model development and formulation approaches are discussed.

Lu, Xiaonan↗
DOE OSTI2019

WSBP/NWBP small-world neural network and its performance for wind power forecasting

A small-world neural network has stronger generalization ability with high transfer efficiency than that of the regular neural networks. This paper presents two novel small-world neural networks, the Watts-Strogatz small-world based on a BP neural network (WSBP) and a Newman-Watts small-world neural network based on a BP neural network (NWBP), related to previous research of complex networks. The algorithms are developed separately by adopting WS and NW small-world networks as their topological structures, and their derivation and convergence criterion are progressively discussed. After that, the proposed models are subsequently tested by two typical nonlinear functions which confirm their significant improvement over the regular BP networks and other algorithms. Finally, a wind power prediction system is advanced to verify their generalization abilities, and show that the models are practically feasible and effective with improved accuracy and acceptable forecasting errors caused by wind fluctuation and randomness with a time scale up to 24 h.

17 WIND ENERGY↗
DOE OSTI2019

Subgrid Treatment of Spacer Grids in the 2D/1D Subplane Approach

Recent efforts have focused on improving the performance of the 2D/1D subplane implementation in MPACT. One of the developments recently added, presented here, focuses on treating spacer grids using a subgrid approach within each subplane, avoiding the extra computation and potentially improving stability issues that can be present by explicitly modeling the spacer grid with a separate MOC plane. From the cases analyzed, it seems that the approach does a sufficient job of maintaining the accuracy of explicitly meshing the spacer grid (<10 pcm and <1% max pin power error) while offering relatively modest performance gains. Overall, the success of the approach with spacer grids opens up possibilities to apply similar approaches to other structural components, particularly those in the axial reflectors to reduce the number of MOC planes and overall computational burden.

Stimpson, Shane G.↗
DOE OSTI2024

Bridging the time scale in exascale computing of chemical systems (Final Technical Report)

This report summarizes the work carried out with support of the United States Department of Energy under Award DE-SC0019441. The theme of this project was to develop and apply methods that allowed for the acceleration of atomistic calculations, particularly in challenging areas such as multiphase systems, electrified interfaces, uncertainty estimation, and applications requiring chemical accuracy, which tend to be applications where simulation time is severely bottlenecked by the computational time requirements. Much of the focus was on the application of emerging machine-learning methodologies, although a wide range of methodologies were employed. This report has two major sections. The first focuses on the methodological advances themselves. Within this part, we report a number of major advances, a few examples of which are described here. We report the first machine-learning scheme for the acceleration of electronically grand-canonical calculations (that is, those applicable to electrochemistry). We report new methods of performing transfer learning, in which physics-based priors can be used to provide predictions, often with uncertainty estimates, of images well outside of training sets; we also offer ways to fine-tune these transfer-learning models. We provide a new systematic means to generate and apply minimal training data sets to very large (10,000’s of atoms) systems, with only small training sets appropriate for electronic structure. We developed new methodologies to integrate surface vibrations into surface adsorption calculations. We made advances to the applicability of diffusion Monte Carlo methods to allow (learned) force prediction, finite-size error correction, and force-free means of searching for transition states. We integrated machine-learned atomistic predictions into mechanism generation codes. Additionally, we released new software including AmpTorch, a modernized version of our original atomistic machine-learning code Amp. The second part of this report focuses on the scientific applications that accompanied, and were often enabled by, the methodological advances described earlier. A few examples follow, but full details are in the individual chapters of the report. For example, we developed a general theory of phonon-induced friction on molecular adsorbates. We showed fundamentally how solvent influences the adsorption and desorption process and how it differs from the processes typically involved at the solid–gas interface, making aqueous-phase and electrocatalysis different from traditional thermocatalysis. We examined how metal–insulator and magnetic transitions can be probed, and accelerated exciton dynamics via Frenkel Hamiltonian parameters. We showed that the nearsighted force-training approach, developed within this project, can predict both the stability and reactivity of large nanoparticles, and can also lead to insights on catalyst coverage on binding energies and entropies. These applied studies, which generally integrated with our method development, allowed us to push forward the theoretical understanding of several reaction classes.

08 HYDROGEN↗
DOE OSTI2025

Data-driven reduced-order models for port-Hamiltonian systems with operator inference

Hamiltonian operator inference has been developed in Sharma et al. (2022) to learn structure-preserving reduced-order models (ROMs) for Hamiltonian systems. The method constructs a low-dimensional model using only data and knowledge of the functional form of the Hamiltonian. The resulting ROMs preserve the intrinsic structure of the system, ensuring that the mechanical and physical properties of the system are maintained. In this work, we extend this approach to port-Hamiltonian systems, which generalize Hamiltonian systems by including energy dissipation, external input, and output. Based on snapshots of the system’s state and output, together with the information about the functional form of the Hamiltonian, reduced operators are inferred through optimization and are then used to construct data-driven ROMs. To further alleviate the complexity of evaluating nonlinear terms in the ROMs, a hyper-reduction method via discrete empirical interpolation is applied. Accordingly, we derive error estimates for the ROM approximations of the state and output. Lastly, we demonstrate the structure preservation, as well as the accuracy of the proposed port-Hamiltonian operator inference framework, through numerical experiments on a linear mass–spring-damper problem and a nonlinear Toda lattice problem.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2017

A density functional theory based approach for predicting melting points of ionic liquids

Accurate prediction of melting points of ILs is important both from the fundamental point of view and from the practical perspective for screening ILs with low melting points and broadening their utilization in a wider temperature range. In this work, we present an ab initio approach to calculating melting points of ILs with known crystal structures and illustrate its application for a series of 11 ILs containing imidazolium/pyrrolidinium cations and halide/polyatomic fluoro-containing anions. The melting point is determined as a temperature at which the Gibbs free energy of fusion is zero. The Gibbs free energy of fusion can be expressed through the use of the Born-Fajans-Haber cycle via the lattice free energy of forming a solid IL from gaseous phase ions and the sum of the solvation free energies of ions comprising IL. Dispersion-corrected density functional theory (DFT) involving (semi)local (PBE-D3) and hybrid exchange-correlation (HSE06-D3) functionals is applied to estimate the lattice enthalpy, entropy, and free energy. The ions solvation free energies are calculated with the SMD-generic-IL solvation model at the M06-2X/6-31+G(d) level of theory under standard conditions. The melting points of ILs computed with the HSE06-D3 functional are in good agreement with the experimental data, with a mean absolute error of 30.5 K and a mean relative error of 8.5%. The model is capable of accurately reproducing the trends in melting points upon variation of alkyl substituents in organic cations and replacement one anion by another. The results verify that the lattice energies of ILs containing polyatomic fluoro-containing anions can be approximated reasonably well using the volume-based thermodynamic approach. However, there is no correlation of the computed lattice energies with molecular volume for ILs containing halide anions. Moreover, entropies of solid ILs follow two different linear relationships with molecular volume for halides and polyatomic fluoro-containing anions. As a result, continuous progress in predicting crystal structures of organic salts with halide anions will be a key factor for successful prediction of melting points with no prior knowledge of the crystal structure.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2015

Reliable four-point flexion test and model for die-to-wafer direct bonding

For many years, wafer-to-wafer (W2W) direct bonding has been very developed particularly in terms of bonding energy measurement and bonding mechanism comprehension. Nowadays, die-to-wafer (D2W) direct bonding has gained significant attention, for instance, in photonics and microelectro-mechanics, which supposes controlled and reliable fabrication processes. So, whatever the stuck materials may be, it is not obvious whether bonded D2W structures have the same bonding strength as bonded W2W ones, because of possible edge effects of dies. For that reason, it has been strongly required to develop a bonding energy measurement technique which is suitable for D2W structures. In this paper, both D2W- and W2W-type standard SiO{sub 2}-to-SiO{sub 2} direct bonding samples are fabricated from the same full-wafer bonding. Modifications of the four-point flexion test (4PT) technique and applications for measuring D2W direct bonding energies are reported. Thus, the comparison between the modified 4PT and the double-cantilever beam techniques is drawn, also considering possible impacts of the conditions of measures such as the water stress corrosion at the debonding interface and the friction error at the loading contact points. Finally, reliability of a modified technique and a new model established for measuring D2W direct bonding energies is demonstrated.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2026

Model-observation discrepancies in Arctic moisture intrusions: causes and pathways for improved simulation

Arctic moisture intrusions (MIs), narrow filaments of strong moisture transport, are key drivers of poleward moisture flux and Arctic weather extremes, yet their representation in climate models is poorly understood. Using a new Arctic MI detection algorithm, we document persistent biases across three CMIP generations (CMIP3–CMIP6): models overestimate MI occurrence over the Pacific sector and underestimate it over the Atlantic sector. These errors stem from misrepresented midlatitude westerly jets, with an equatorward North Atlantic jet associated with too few Atlantic MIs, and a poleward, weakened North Pacific jet linked to too many Pacific MIs. Experiments that correct sea surface temperature and sea ice concentration biases and increase atmospheric resolution improve jet structure and MI statistics, while a cloud-locking simulation indicates that better high-frequency cloud–radiation–circulation interactions can yield further gains. Our results clarify pathways to reducing long-standing MI and jet biases, providing guidance for improving simulations of Arctic and midlatitude climate.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2015

Comment on “Towards high efficiency thin-film crystalline silicon solar cells: The roles of light trapping and non-radiative recombinations” [J. Appl. Phys. 115, 094501 (2014)]

In the above paper, an analytical approach including a new solution to the differential diffusion equation in illuminated quasi-neutral regions (QNR) is exploited to calculate the short-circuit current density (J{sub sc}), open-circuit voltage (V{sub oc}), fill factor (FF), and efficiency (η) of light-trapping (LT) c-Si solar cells with a given structure. Comparisons with numerical results calculated by the Silvaco ATLAS device simulator in the same LT cells show that the analytical results are systematically overestimated. According to the authors, the inaccuracies in J{sub sc}, V{sub oc}, and η are due to the fact that assuming ideal collection from space-charge region (SCR) and using the superposition approximation introduce systematic errors into analytical models. In this comment, an analytical approach using reported solutions to the transport equations in QNR and SCR, where ideal collection from SCR is assumed and the superposition approximation is used, is shown to agree with both the Silvaco and PC1d numerical approaches in calculating J{sub sc}, V{sub oc}, and η, in the same LT devices as considered in the commented paper. Reasons for the inaccuracies detected in the commented paper are suggested.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2015

Optical constants of Cu(In, Ga)Se{sub 2} for arbitrary Cu and Ga compositions

The optical constants of Cu(In, Ga)Se{sub 2} (CIGS)-based polycrystalline layers with different Cu and Ga compositions are parameterized completely up to a photon energy of 6.5 eV assuming several Tauc-Lorentz transition peaks. Based on the modeled optical constants, we establish the calculation procedure for the CIGS optical constants in a two-dimensional compositional space of (Cu, Ga) by taking the composition-induced shift of the critical point energies into account. In particular, we find that the variation of the CIGS optical constants with the Cu composition can be modeled quite simply by a spectral-averaging method in which the dielectric function of the target Cu composition is estimated as a weighted average of the dielectric functions with higher and lower Cu compositions. To express the effect of the Ga composition, on the other hand, an energy shift model reported earlier is adopted. Our model is appropriate for a wide variety of CIGS-based materials having different Cu and Ga compositions, although the modeling error increases slightly at lower Cu compositions [Cu/(In + Ga) < 0.69]. From our model, the dielectric function, refractive index, extinction coefficient, and absorption coefficient for the arbitrary CIGS composition can readily be obtained. The optical database developed in this study is applied further for spectroscopic ellipsometry analyses of CIGS layers fabricated by single and multi-stage coevaporation processes. We demonstrate that the compositional and structural characterizations of the CIGS-based layers can be performed from established analysis methods.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2025

Many-body expansion based machine learning models for octahedral transition metal complexes

Abstract Graph-based machine learning (ML) models for material properties show great potential to accelerate virtual high-throughput screening of large chemical spaces. However, in their simplest forms, graph-based models do not include any 3D information and are unable to distinguish stereoisomers such as those arising from different orderings of ligands around a metal center in coordination complexes. In this work we present a modification to revised autocorrelation descriptors, a molecular graph featurization method, for predicting spin state dependent properties of octahedral transition metal complexes (TMCs). Inspired by analytical semi-empirical models for TMCs, the new modeling strategy is based on the many-body expansion (MBE) and allows one to tune the captured stereoisomer information by changing the truncation order of the MBE. We present the necessary modifications to include this approach in two commonly used ML methods, kernel ridge regression and feed-forward neural networks. On a test set composed of all possible isomers of binary TMCs, the best MBE models achieve mean absolute errors (MAEs) of 2.75 kcal mol −1 on spin-splitting energies and 0.26 eV on frontier orbital energy gaps, a 30%–40% reduction in error compared to models based on our previous approach. We also observe improved generalization to previously unseen ligands where the best-performing models exhibit MAEs of 4.00 kcal mol −1 (i.e. a 0.73 kcal mol −1 reduction) on the spin-splitting energies and 0.53 eV (i.e. a 0.10 eV reduction) on the frontier orbital energy gaps. Because the new approach incorporates insights from electronic structure theory, such as ligand additivity relationships, these models exhibit systematic generalization from homoleptic to heteroleptic complexes, allowing for efficient screening of TMC search spaces.

Meyer, Ralf (ORCID:0000000322360261)↗