Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “strong correlation”

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 1,531 records · Page 85

Effects of Respiratory Motion on Passively Scattered Proton Therapy Versus Intensity Modulated Photon Therapy for Stage III Lung Cancer: Are Proton Plans More Sensitive to Breathing Motion?

Purpose: To quantify and compare the effects of respiratory motion on paired passively scattered proton therapy (PSPT) and intensity modulated photon therapy (IMRT) plans; and to establish the relationship between the magnitude of tumor motion and the respiratory-induced dose difference for both modalities. Methods and Materials: In a randomized clinical trial comparing PSPT and IMRT, radiation therapy plans have been designed according to common planning protocols. Four-dimensional (4D) dose was computed for PSPT and IMRT plans for a patient cohort with respiratory motion ranging from 3 to 17 mm. Image registration and dose accumulation were performed using grayscale-based deformable image registration algorithms. The dose–volume histogram (DVH) differences (4D-3D [3D = 3-dimensional]) were compared for PSPT and IMRT. Changes in 4D-3D dose were correlated to the magnitude of tumor respiratory motion. Results: The average 4D-3D dose to 95% of the internal target volume was close to zero, with 19 of 20 patients within 1% of prescribed dose for both modalities. The mean 4D-3D between the 2 modalities was not statistically significant (P<.05) for all dose–volume histogram indices (mean ± SD) except the lung V5 (PSPT: +1.1% ± 0.9%; IMRT: +0.4% ± 1.2%) and maximum cord dose (PSPT: +1.5 ± 2.9 Gy; IMRT: 0.0 ± 0.2 Gy). Changes in 4D-3D dose were correlated to tumor motion for only 2 indices: dose to 95% planning target volume, and heterogeneity index. Conclusions: With our current margin formalisms, target coverage was maintained in the presence of respiratory motion up to 17 mm for both PSPT and IMRT. Only 2 of 11 4D-3D indices (lung V5 and spinal cord maximum) were statistically distinguishable between PSPT and IMRT, contrary to the notion that proton therapy will be more susceptible to respiratory motion. Because of the lack of strong correlations with 4D-3D dose differences in PSPT and IMRT, the extent of tumor motion was not an adequate predictor of potential dosimetric error caused by breathing motion.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Transverse momentum fluctuations and their correlation with elliptic flow in nuclear collision

In this article, we propose observables v 0 and v 0 (p T ) which quantify the relative fluctuations in the total transverse momentum at fixed multiplicity. We first study the factorization of the fixed multiplicity momentum dependent two particle correlation function into a product of v 0 ($p^{a}_{T}$) and v 0 ($p^{b}_{T}$) within realistic hydrodynamic simulations. Then we present computations of v 0 (p T ) for different particle types. We determine the relation between the integrated v 0 and previously measured observables, and compare results from a hybrid hydrodynamics based model to experimental data. The effects of bulk viscosity and an initial pre-equilibrium stage on the results are quantified. We find that v 0 is strongly correlated with the initial state entropy per elliptic area, S/A. Using this result, we explain how the observed correlations between the elliptic flow and the transverse momentum (both in simulations and experiment) reflect the initial state correlations between 1/A and ellipticity ε 2 at fixed multiplicity. Furthermore, we argue that the systematic experimental study of v 0 , with the same sophistication as used for the other v n , can contribute significantly to our understanding of quark gluon plasma properties.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Investigating the correlation of biomass recalcitrance with pyrolysis oil using poplar as the feedstock

Pyrolysis of five poplar samples with differing degrees of recalcitrance was performed; the correlations between the poplar enzymatic hydrolysis glucose yields and the physicochemical properties of pyrolysis product were studied in this work. Sugar release of five poplar samples varied from 48.1 to 112.3 mg/g for glucose, and 12.0 to 32.4 mg/g for xylose. The yield of pyrolysis products was calculated and the molecular weight distribution of pyrolysis oils was measured by GPC, ranging from 268 to 289 g/mol for its weight-average molecular weight. GC–MS analysis of the bio-oil exhibited a strong correlation between biomass recalcitrance and guaiacyl-type structures in bio-oils. The correlation between biomass recalcitrance and the ratio of syringyl-to-guaiacyl-type-related structures was also assessed. The results from quantitative 31 P NMR indicated some correlation between biomass recalcitrance and the guaiacyl hydroxyl groups in bio-oils. These conclusions illustrate correlations and differences between converting biomass to biofuels via the biological and thermal platform.

09 BIOMASS FUELS↗

A NEAR-INFRARED SPECTROSCOPIC SURVEY OF CLASS I PROTOSTARS

We present the results of a near-IR spectroscopic survey of 110 Class I protostars observed from 0.80 {mu}m to 2.43 {mu}m at a spectroscopic resolution of R = 1200. This survey is unique in its selection of targets from the whole sky, its sample size, wavelength coverage, depth, and sample selection. We find that Class I objects exhibit a wide range of lines and the continuum spectroscopic features. Eighty-five percent of Class I protostars exhibit features indicative of mass accretion, and we found that the veiling excess, CO emission, and Br {gamma} emission are closely related. We modeled the spectra to estimate the veiling excess (r{sub k} ) and extinction to each target. We also used near-IR colors and emission line ratios, when available, to also estimate extinction. In the course of this survey, we observed the spectra of 10 FU Orionis-like objects, including 2 new ones, as well as 3 Herbig Ae-type stars among our Class I young stellar objects. We used photospheric absorption lines, when available, to estimate the spectral type of each target. Although most targets are late-type stars, there are several A- and F-type stars in our sample. Notably, we found no A or F class stars in the Taurus-Auriga or Perseus star-forming regions. There are several cases where the observed CO and/or water absorption bands are deeper than expected from the photospheric spectral type. We find a correlation between the appearance of the reflection nebula, which traces the distribution of material on very large scales, and the near-IR spectrum, which probes smaller scales. All of the FU Orionis-like objects are associated with reflection nebulae. The spectra of the components of spatially resolved protostellar binaries tend to be very similar. In particular both components tend to have similar veiling and H{sub 2} emission, inconsistent with random selection from the sample as a whole. There is a strong correlation between [Fe II] and H{sub 2} emission, supporting previous results showing that H{sub 2} emission in the spectra of young stars is usually shock excited by stellar winds.

79 ASTRONOMY AND ASTROPHYSICS↗

HST imaging of fading AGN candidates. I. Host-galaxy properties and origin of the extended gas

We present narrow- and medium-band Hubble Space Telescope imaging, with additional supporting ground-based imaging, spectrophotometry, and Fabry–Perot interferometric data, for eight galaxies identified as hosting a fading active galactic nucleus (AGN). These are selected to have AGN-ionized gas projected >10 kpc from the nucleus and energy budgets with a significant shortfall of ionizing radiation between the requirement to ionize the distant gas and the AGN as observed directly, indicating fading of the AGN on ≈50,000 yr timescales. This paper focuses on the host-galaxy properties and origin of the gas. In every galaxy, we identify evidence of ongoing or past interactions, including tidal tails, shells, and warped or chaotic dust structures; a similarly selected sample of obscured AGNs with extended ionized clouds shares this high incidence of disturbed morphologies. Several systems show multiple dust lanes in different orientations, broadly fit by differentially precessing disks of accreted material viewed ∼1.5 Gyr after its initial arrival. The host systems are of early Hubble type; most show nearly pure de Vaucouleurs surface brightness profiles and Sérsic indices appropriate for classical bulges, with one S0 and one SB0 galaxy. The gas has a systematically lower metallicity than the nuclei; three systems have abundances uniformly well below solar, consistent with an origin in tidally disrupted low-luminosity galaxies, while some systems have more nearly solar abundances (accompanied by such signatures as multiple Doppler components), which may suggest redistribution of gas by outflows within the host galaxies themselves. These aspects are consistent with a tidal origin for the extended gas in most systems, although the ionized gas and stellar tidal features do not always match closely. Unlike extended emission regions around many radio-loud AGNs, these clouds are kinematically dominated by rotation, in some cases in warped disks. Outflows can play important kinematic roles only in localized regions near some of the AGNs. We find only a few sets of young star clusters potentially triggered by AGN outflows. In UGC 7342 and UGC 11185, multiple luminous star clusters are seen just within the projected ionization cones, potentially marking star formation triggered by outflows. As in the discovery example, Hanny’s Voorwerp/IC 2497, there are regions in these clouds where the lack of a strong correlation between Hα surface brightness and ionization parameter indicates that there is unresolved fine structure in the clouds. Together with thin coherent filaments spanning several kpc, persistence of these structures over their orbital lifetimes may require a role for magnetic confinement. Overall, we find that the sample of fading AGNs occur in interacting and merging systems, that the very extended ionized gas is composed of tidal debris rather than galactic winds, and that these host systems are bulge-dominated and show no strong evidence of triggered star formation in luminous clusters.

79 ASTRONOMY AND ASTROPHYSICS↗

CLaSPS: A NEW METHODOLOGY FOR KNOWLEDGE EXTRACTION FROM COMPLEX ASTRONOMICAL DATA SETS

In this paper, we present the Clustering-Labels-Score Patterns Spotter (CLaSPS), a new methodology for the determination of correlations among astronomical observables in complex data sets, based on the application of distinct unsupervised clustering techniques. The novelty in CLaSPS is the criterion used for the selection of the optimal clusterings, based on a quantitative measure of the degree of correlation between the cluster memberships and the distribution of a set of observables, the labels, not employed for the clustering. CLaSPS has been primarily developed as a tool to tackle the challenging complexity of the multi-wavelength complex and massive astronomical data sets produced by the federation of the data from modern automated astronomical facilities. In this paper, we discuss the applications of CLaSPS to two simple astronomical data sets, both composed of extragalactic sources with photometric observations at different wavelengths from large area surveys. The first data set, CSC+, is composed of optical quasars spectroscopically selected in the Sloan Digital Sky Survey data, observed in the x-rays by Chandra and with multi-wavelength observations in the near-infrared, optical, and ultraviolet spectral intervals. One of the results of the application of CLaSPS to the CSC+ is the re-identification of a well-known correlation between the {alpha}{sub OX} parameter and the near-ultraviolet color, in a subset of CSC+ sources with relatively small values of the near-ultraviolet colors. The other data set consists of a sample of blazars for which photometric observations in the optical, mid-, and near-infrared are available, complemented for a subset of the sources, by Fermi {gamma}-ray data. The main results of the application of CLaSPS to such data sets have been the discovery of a strong correlation between the multi-wavelength color distribution of blazars and their optical spectral classification in BL Lac objects and flat-spectrum radio quasars, and a peculiar pattern followed by blazars in the WISE mid-infrared colors space. This pattern and its physical interpretation have been discussed in detail in other papers by one of the authors.

79 ASTRONOMY AND ASTROPHYSICS↗

VARIATIONS IN SOLAR WIND FRACTIONATION AS SEEN BY ACE/SWICS AND THE IMPLICATIONS FOR GENESIS MISSION RESULTS

We use Advanced Composition Explorer (ACE)/Solar Wind Ion Composition Spectrometer (SWICS) elemental composition data to compare the variations in solar wind (SW) fractionation as measured by SWICS during the last solar maximum (1999–2001), the solar minimum (2006–2009), and the period in which the Genesis spacecraft was collecting SW (late 2001—early 2004). We differentiate our analysis in terms of SW regimes (i.e., originating from interstream or coronal hole flows, or coronal mass ejecta). Abundances are normalized to the low-first ionization potential (low-FIP) ion magnesium to uncover correlations that are not apparent when normalizing to high-FIP ions. We find that relative to magnesium, the other low-FIP elements are measurably fractionated, but the degree of fractionation does not vary significantly over the solar cycle. For the high-FIP ions, variation in fractionation over the solar cycle is significant: greatest for Ne/Mg and C/Mg, less so for O/Mg, and the least for He/Mg. When abundance ratios are examined as a function of SW speed, we find a strong correlation, with the remarkable observation that the degree of fractionation follows a mass-dependent trend. We discuss the implications for correcting the Genesis sample return results to photospheric abundances.

79 ASTRONOMY AND ASTROPHYSICS↗

DISSECTING THE STELLAR-MASS-SFR CORRELATION IN z = 1 STAR-FORMING DISK GALAXIES

Using a mass-limited sample of 24 {mu}m detected, star-forming galaxies at 0.5 < z < 1.3, we study the mass-star formation rate (SFR) correlation and its tightness. The correlation is well defined ({sigma} = 0.28 dex) for disk galaxies (n{sub Sersic} < 1.5), while more bulge-dominated objects often have lower specific SFRs (sSFRs). For disk galaxies, a much tighter correlation ({sigma} = 0.19 dex) is obtained if the rest-frame H-band luminosity is used instead of stellar mass derived from multi-color photometry. The sSFR correlates strongly with rest-frame optical colors (hence luminosity-weighted stellar age) and also with clumpiness (which likely reflects the molecular gas fraction). This implies that most of the observed scatter is real, despite its low level, and not dominated by random measurement errors. After correcting for these differential effects a remarkably small dispersion remains ({sigma} = 0.14 dex), suggesting that measurement errors in mass or SFR are {approx}< 0.10 dex, excluding systematic uncertainties. Measurement errors in stellar masses, the thickening of the correlation due to real sSFR variations, and varying completeness with stellar mass, can spuriously bias the derived slope to lower values due to the finite range over which observables (mass and SFR) are available. When accounting for these effects, the intrinsic slope for the main sequence for disk galaxies gets closer to unity.

79 ASTRONOMY AND ASTROPHYSICS↗

Dissecting molecular descriptors into atomic contributions in density functional reactivity theory

Density functional reactivity theory (DFRT) employs the electron density of a molecule and its related quantities such as gradient and Laplacian to describe its structure and reactivity properties. Proper descriptions at both molecular (global) and atomic (local) levels are equally important and illuminating. In this work, we make use of Bader's zero-flux partition scheme and consider atomic contributions for a few global reactivity descriptors in DFRT, including the density-based quantification of steric effect and related indices. Earlier, we proved that these quantities are intrinsically correlated for atomic and molecular systems [S. B. Liu, J. Chem. Phys. 126, 191107 (2007); ibid. 126, 244103 (2007)]. In this work, a new basin-based integration algorithm has been implemented, whose reliability and effectiveness have been extensively examined. We also investigated a list of simple hydrocarbon systems and different scenarios of bonding processes, including stretching, bending, and rotating. Interesting changing patterns for the atomic and molecular values of these quantities have been revealed for different systems. This work not only confirms the strong correlation between these global reactivity descriptors for molecular systems, as theoretically proven earlier by us, it also provides new and unexpected changing patterns for their atomic values, which can be employed to understand the origin and nature of chemical phenomena.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Persistence of Gamma-H2AX Foci in Bronchial Cells Correlates with Susceptibility to Radiation Associated Lung Cancer in Mice

The risk of developing radiation-induced lung cancer differs between different strains of mice, but the underlying cause of the strain differences is unknown. Strains of mice also differ in how quickly they repair radiation-induced DNA double-strand breaks (DSBs). We assayed mouse strains from the CcS/Dem recombinant congenic strain set for their efficacy in repairing DNA DSBs during protracted irradiation. We measured unrepaired γ-H2AX radiation-induced foci (RIF), which persisted after chronic 24-h gamma irradiation, as a surrogate marker for repair efficiency in bronchial epithelial cells for 17 of the CcS/Dem strains and the BALB/c founder strain. We observed a very strong correlation (R2 = 79.18%, P < 0.001) between the level of unrepaired RIF and radiogenic lung cancer incidence measured in the same strains. Interestingly, spontaneous levels of foci in nonirradiated mice also showed good correlation with lung cancer incidence when incidence data from male and female mice were combined. These results suggest that genetic differences in DNA repair capacity largely account for differing susceptibilities to radiationinduced lung cancer among CcS/Dem mouse strains, and that high levels of spontaneous DNA damage are also a relatively good marker of cancer predisposition. In a smaller pilot study, we found that the repair capacity measured in peripheral blood leucocytes also correlated well with radiogenic lung cancer susceptibility, raising the possibility that the assay could be used to detect radiogenic lung cancer susceptibility in humans.

60 APPLIED LIFE SCIENCES↗

The X-ray AGN content of the Molonglo 408 MHz Survey: Bulk properties of previously optically identified sources

The cross-correlation of a source list from the ROSAT All Sky Survey with the Molonglo 408 MHz Survey (MRC) yields a calalogue of 549 sources. The majority of the objects are quasars and radio galaxies with known optical counterparts, most of them detected in X-rays for the first time. More than one third of the objects are optically unidentified. In this paper we will present a discussion of the representative properties of the 337 previously optically identified active galactic nuclei (AGN). We find strong correlations between luminosities in the radio, optical, and X-ray bands which differ for quasars and radio galaxies. Several of the statistical source characteristics seem not to support simple unification schemes. However, for many objects some of their key properties like redshift, radio morphology and environmental conditions are not known yet. The fraction of X-ray detected quasars in the MRC is around 30%, higher than that of radio galaxies (less tha 8%). It is not currently known which intrinsic properties are responsible for the X-ray loudness of these radio objects. We further have evidence for a considerable number of objects being radio- and X-ray loud, but optically very faint.

Brinkmann, W.↗

An experimental search for near-wall boundary conditions for large eddy simulation

Instantaneous wall shear stress and streamwise velocities have been measured simultaneously in a flat plate, turbulent boundary layer at moderate Reynolds number in an effort to provide experimental support for large eddy simulations. Data were obtained by using a buried-wire wall shear gage and a hot-wire rake positioned in the log region of the flow. All data processing was accomplished with digital data analysis techniques on a minicomputer. Fluctuations of the instantaneous U plus versus Y plus profiles about a mean law of the wall are shown to be significant and complex. Peak cross-correlation values between wall shear stress and the velocities are high and reflect the passage of a large structure inclined at a small angle to the wall. Estimates of this angle are consistent with those made by other investigators. Conditional sampling techniques were used to detect the passage of various sizes and types of flow disturbances (events) and to estimate their mean frequency of occurrence. Events characterized by large and sudden streamwise accelerations were found to be highly coherent throughout the log region and were strongly correlated with large fluctuations in wall shear-stress. Phase randomness between the near-wall quantities and the outer velocities was small. The results suggest that the flow events detected by conditional sampling applied to velocities in the log region may be related to the bursting process.

Robinson, S. K.↗

Instantaneous velocity profile measurements in a turbulent boundary layer

Instantaneous wall shear stress and streamwise velocities have been measured simultaneously in a flat-plate, turbulent boundary layer at moderate Reynolds number in an effort to provide experimental support for large eddy simulations. Data were obtained using a buried-wire, wall shear gage and a hot-wire rake positioned in the log region of the flow. Fluctuations of the instantaneous U(+) versus Y(+) profiles about a mean law of the wall are shown to be significant and complex. Peak cross-correlation values between wall shear stress and the velocities are high, and reflect the passage of a large structure inclined at a small angle to the wall. Estimates of this angle are consistent with those made by other investigators. Conditional sampling techniques were used to detect the passage of various sizes and types of flow disturbances (events), and to estimate their mean frequency of occurrence. Events characterized by large aand sudden streamwise accelerations were found to be highly coherent throughout the log region and were strongly correlated with large fluctuations in wall shear stress. Phase randomness between the near-wall quantities and the outer velocities was small. The results suggest that the flow events detected by conditional sampling applied to velocities in the log region may be related to the bursting process.

Robinson, S. K.↗

The Growth of Instabilities in Annular Liquid Sheets

An annular liquid sheet surrounded by parallel co-flowing gas is an effective atomiser. However, the initial instabilities which determine the primary break-up of the liquid sheet are not well understood. Lack of agreement on the influence of the boundary conditions and the non-dimension scaling of the initial instability persists between theoretical stability analyses and experiments. To address this matter, we have undertaken an experimental parametric study of an aerodynamically-driven, non-swirling annular water sheet. The effects of sheet thickness, inner and outer gas-liquid momentum ratio were investigated over an order of magnitude variation in Reynolds and Weber number. From high-speed image correlation measurements in the near-nozzle region, we propose new empirical correlations for the frequency of the instability as a function of the total gas-liquid momentum ratio, with good non-dimensional collapse. From analysis of the instability velocity probability densities, we find two persistent and distinct superimposed instabilities with different growth rates. The first is a short-lived, rapidly saturating sawtooth-like instability. The second is a slower-growing stochastic instability which persists through the break-up of the sheet. The presence of multiple instabilities whose growth rates do not strongly correlate with the shear velocities may explain some of the discrepancies between experiments and stability analyses.

annular↗

A resource adequacy assessment of correlated wide-area outages in the power grid

As the power grid is undergoing rapid transformations, numerous questions are emerging about its vulnerability to wide-area extreme events (WAEE), which could influence its operations. Relatively few analyses have been conducted to date regarding the impact of correlated outages during WAEEs on the grid’s ability to balance resources with demand. This study addresses this gap by conducting a resource adequacy analysis for a hurricane-inspired WAEE on a 2035 synthetic power grid system for the United States. A sensitivity analysis was also conducted to characterize the relative impact of weather on unserved energy. Our results indicate that although the magnitude and duration of the shortfalls vary depending on weather conditions, persistent shortfalls are observed in some regions. Initial explorations indicate a strong correlation between transmission-constrained regions and regions with persistent shortfalls. Future work could generate empirically-grounded representations for generator outages as well as conduct causal analyses of these shortfalls to improve understanding of drivers as well as possible mitigation strategies. Continued exploration of extreme weather impacts on the grid is important to develop more robust understanding of the reliability and resilience of our power systems, especially as they undergo rapid transformations.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Hard x ray highlights of AR 5395

Active Region 5395 produced an exceptional series of hard x ray bursts notable for their frequency, intensity, and impulsivity. Over the two weeks from March 6 to 19, 447 hard x ray flares were observed by the Hard X Ray Burst Spectrometer on Solar Maximum Mission (HXRBS/SMM), a rate of approx. 35 per day which exceeded the previous high by more than 50 percent. During one 5 day stretch, more than 250 flares were detected, also a new high. The three largest GOES X-flares were observed by HXRBS and had hard x ray rates over 100,000 s(exp -1) compared with only ten flares above 100,000(exp -1) during the previous nine years of the mission. An ongoing effort for the HXRBS group has been the correlated analysis of hard x ray data with flare data at other wavelengths with the most recent emphasis on those measurements with spatial information. During a series of bursts from AR 5395 at 1644 to 1648 UT on 12 March 1989, simultaneous observations were made by HXRBS and UVSP (Ultra Violet Spectrometer Polarimeter) on SMM, the two-element Owens Valley Radio Observatory (OVRO) interferometric array, and R. Canfield's H-alpha Echelle spectrograph at the National Solar Observatory at Sacramento Peak. The data show strong correlations in the hard x ray, microwave, and UV lightcurves. This event will be the subject of a combined analysis.

Schwartz, R. A.↗

Persistence of Gamma-H2AX Foci in Irradiated Bronchial Cells Correlates with Susceptibility to Radiation Associated Lung Cancer in Mice

The risk of developing radiation-induced lung cancer differs between different strains of mice, but the underlying cause of the strain differences is unknown. Strains of mice also differ in their ability to efficiently repair DNA double strand breaks resulting from radiation exposure. We phenotyped mouse strains from the CcS/Dem recombinant congenic strain set for their efficacy in repairing DNA double strand breaks during protracted radiation exposures. We monitored persistent gamma-H2AX radiation induced foci (RIF) 24 hours after exposure to chronic gamma-rays as a surrogate marker for repair deficiency in bronchial epithelial cells for 17 of the CcS/Dem strains and the BALB/cHeN founder strain. We observed a very strong correlation R2 = 79.18%, P < 0.001) between the level of persistent RIF and radiogenic lung cancer percent incidence measured in the same strains. Interestingly, spontaneous levels of foci in non-irradiated strains also showed good correlation with lung cancer incidence (R2=32.74%, P =0.013). These results suggest that genetic differences in DNA repair capacity largely account for differing susceptibilities to radiation-induced lung cancer among CcS/Dem mouse strains and that high levels of spontaneous DNA damage is also a relatively good marker of cancer predisposition. In a smaller pilot study, we found that the repair capacity measured in peripheral blood leucocytes also correlated well with radiogenic lung cancer susceptibility, raising the possibility that such phenotyping assay could be used to detect radiogenic lung cancer susceptibility in humans.

Ochola, Donasian O.↗

Machine Learning Prediction of the Experimental Transition Temperature of Fe(II) Spin-Crossover Complexes

Spin-crossover (SCO) complexes are materials that exhibit changes in the spin state in response to external stimuli, with potential applications in molecular electronics. It is challenging to know a priori how to design ligands to achieve the delicate balance of entropic and enthalpic contributions needed to tailor a transition temperature close to room temperature. Here, we leverage the SCO complexes from the previously curated SCO-95 data set [Vennelakanti et al. J. Chem. Phys. 159, 024120 (2023)] to train three machine learning (ML) models for transition temperature (T 1/2 ) prediction using graph-based revised autocorrelations as features. We perform feature selection using random forest-ranked recursive feature addition (RF-RFA) to identify the features essential to model transferability. Of the ML models considered, the full feature set RF and recursive feature addition RF models perform best, achieving moderate correlation to experimental T 1/2 values. We then compare ML T 1/2 predictions to those from three previously identified best-performing density functional approximations (DFAs) which accurately predict SCO behavior across SCO-95, finding that the ML models predict T 1/2 more accurately than the best-performing DFAs. In addition, we study ML model predictions for a set of 18 SCO complexes for which only estimated T 1/2 values are available. Upon excluding outliers from this set, the RF-RFA RF model shows a strong correlation to estimated T 1/2 values with a Pearson’s r of 0.82. In contrast, DFA-predicted T 1/2 values have large errors and show no correlation to estimated T 1/2 values over the same set of complexes. Overall, our study demonstrates slightly superior performance of ML models in comparison with some of the best-performing DFAs, and we expect ML models to improve further as larger data sets of SCO complexes are curated and become available for model training.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗