Search

Rapid growth in data, computational methods, and computing power is driving a remarkable revolution in what variously is termed machine learning (ML), statistical learning, computational learning, and artificial intelligence. In addition to highly visible successes in machine-based natural language translation, playing the game Go, and self-driving cars, these new technologies also have profound implications for computational and experimental science and engineering, as well as for the exascale computing systems that the Department of Energy (DOE) is developing to support those disciplines. Not only do these learning technologies open up exciting opportunities for scientific discovery on exascale systems, they also appear poised to have important implications for the design and use of exascale computers themselves, including high-performance computing (HPC) for ML and ML for HPC. The overarching goal of the ExaLearn co-design project is to provide exascale ML software for use by Exascale Computing Project (ECP) applications, other ECP co-design centers, and DOE experimental facilities and leadership class computing facilities.

Atomic-scale disorder plays an important role in the chemical and physical properties of oxide materials. The structural flexibility of pyrochlore-type oxides allows for crystal-chemical engineerin...

A fast and exact algorithm to calculate the powder pair distribution function (PDF) for the case of periodic structures is presented. The algorithm especially improves X-ray and electron PDF calculations, and the handling of instrumental resolution functions., A fast and exact algorithm to calculate the powder pair distribution function (PDF) for the case of periodic structures is presented. The new algorithm calculates the PDF by a detour via reciprocal space. The calculated normalized total powder diffraction pattern is transferred into the PDF via the sine Fourier transform. The calculation of the PDF via the powder pattern avoids the conventional simplification of X-ray and electron atomic form factors. It is thus exact for these types of radiation, as is the conventional calculation for the case of neutron diffraction. The new algorithm further improves the calculation speed. Additional advantages are the improved detection of errors in the primary data, the handling of preferred orientation, the ease of treatment of magnetic scattering and a large improvement to accommodate more complex instrumental resolution functions.

One of the main goals of neutron data analysis is to determine the internal structure of materials from their neutron scattering profiles. These structures are defined by a crystallographic class label and a set of real-valued parameters specific to that class. Existing structure analysis approaches use computationally expensive loop refinements methods that routinely take days, and even weeks, to complete. Additionally, the outcomes often rely on the fidelity of physical models that are computed during the refinement process.

Large-scale simulations and data analysis are often required to explain neutron scattering experiments to establish a connection between the fundamental physics at the nanoscale and data probed by neutrons. However, to perform simulations at experimental conditions it is critical to use correct force-field (FF) parameters which are unfortunately not available for most complex experimental systems. In this work, we have developed a workflow optimization technique to provide optimized FF parameters by comparing molecular dynamics (MD) to neutron scattering data. We describe the workflow in detail by using an example system consisting of tRNA and hydrophilic nanodiamonds in a deuterated water (D2O) environment. Quasi-elastic neutron scattering (QENS) data show a faster motion of the tRNA in the presence of nanodiamond than without the ND. To compare the QENS and MD results quantitatively, a proper choice of FF parameters is necessary. We use an efficient workflow to optimize the FF parameters between the hydrophilic nanodiamond and water by comparing to the QENS data. Our results show that we can obtain accurate FF parameters by using this technique. The workflow can be generalized to other types of neutron data for FF optimization, such as vibrational spectroscopy and spin echo.

Computational science is well established as the third pillar of scientific discovery and is on par with experimentation and theory. However, as we move closer toward the ability to execute exascale calculations and process the ensuing extreme-scale amounts of data produced by both experiments and computations alike, the complexity of managing the compute and data analysis tasks has grown beyond the capabilities of domain scientists. Thus, workflow management systems are absolutely necessary to ensure current and future scientific discoveries. A key research question for these workflow management systems concerns the performance optimization of complex calculation and data analysis tasks. The central contribution of this article is a description of the PANORAMA approach for modeling and diagnosing the run-time performance of complex scientific workflows. This approach integrates extreme-scale systems testbed experimentation, structured analytical modeling, and parallel systems simulation into a comprehensive workflow framework called Pegasus for understanding and improving the overall performance of complex scientific workflows.

A combination of Rietveld refinement and pair distribution function analysis of total neutron scattering data are used to provide insight into the negative thermal expansion mechanism of siliceous faujasite. The negative thermal expansion mechanism of siliceous faujasite is attributed to the transverse vibrations of bridging oxygen atoms resulting in the coupled librations of the SiO4 tetrahedra. The constituent SiO4 tetrahedra are revealed to expand in size with temperature and they are also shown to undergo some distortion as temperature is increased. However, these distortions are not distinct enough in any geometric manner for the average behaviour of the SiO4 tetrahedra not to be considered as that of a rigid units. The work displays the benefits of using total scattering experiments to unveil the finer details of dynamic thermomechanical processes within crystalline materials.

Understanding the role of retarder on the chemical nature and molecular architecture of hydrating cement paste is essential for engineering oil well cements with additives. Here, synchrotron X-ray and total neutron scattering with pair distribution function (PDF) analysis were performed in combination with calorimetry and nuclear magnetic resonance (NMR) to examine the retarder effect in hydrating tri-calcium silicate (C3S) and Class G oil well cement paste. Primarily, the retarder, Diethylenetriamine pentamethylene phosphonic acid (DTPMP) influenced the hydration by affecting the Ca-O and Ca-Si pair correlation providing evidence of calcium playing a predominant role in the retardation process. Secondary effects related to Calcium-Silicate-Hydrate (C-S-H) nuclei poisoning influencing the suppression of calcium hydroxide precipitation were observed. These findings provide insights into the retardation mechanism of hydrating cement paste influenced by calcium depletion when subjected to phosphonate retarders.

Scientific research data provides unique challenges that are distinct from classic “Big Data” sources. One common element in research data is that the experiment, observations, or simulation were designed, and data were specifically acquired, to shed light on an open scientific question. The data and methods are usually “owned” by the researcher(s) and the data itself might not be viewed to have long-term scientific significance after the results have been published. Often, the data volume was relatively low, with data sometimes easier to reproduce than to catalog and store. Some data and meta-data were not collected in a digital form, or were stored on antiquated or obsolete media. Generally speaking, policies, tools, and management of digital research data have reflected an ad hoc approach that varies domain by domain and research group by research group. This model, which treats research data as disposable, is proving to be a serious limitation as the volume and complexity of research data explodes. Ch...

We use a combination of neutron and X-ray total scattering measurements together with pair distribution function (PDF) analysis to characterise the variation in local structure across the orbital order--disorder transition in LaMnO$_3$. Our experimental data are inconsistent with a conventional order--disorder description of the transition, and reflect instead the existence of a discontinuous change in local structure between ordered and disordered states. Within the orbital-ordered regime, the neutron and X-ray PDFs are best described by a local structure model with the same local orbital arrangements as those observed in the average (long-range) crystal structure. We show that a variety of meaningfully-different local orbital arrangement models can give fits of comparable quality to the experimental PDFs collected within the disordered regime; nevertheless, our data show a subtle but consistent preference for the anisotropic Potts model proposed in \emph{Phys Rev.\ B} {\bf 79}, 174106 (2009). The key implications of this model are electronic and magnetic isotropy together with the loss of local inversion symmetry at the Mn site. We conclude with a critical assessment of the interpretation of PDF measurements when characterising local symmetry breaking in functional materials., 14 pages, 8 figures, 3 tables

Total scattering and pair distribution function (PDF) methods allow for detailed study of local atomic order and disorder, including materials for which Rietveld refinements are not traditionally possible (amorphous materials, liquids, glasses and nanoparticles). With the advent of modern neutron time-of-flight (TOF) instrumentation, total scattering studies are capable of producing PDFs with ranges upwards of 100–200 Å, covering the correlation length scales of interest for many materials under study. Despite this, the refinement and subsequent analysis of data are often limited by confounding factors that are not rigorously accounted for in conventional analysis programs. While many of these artifacts are known and recognized by experts in the field, their effects and any associated mitigation strategies largely exist as passed-down `tribal' knowledge in the community, and have not been concisely demonstrated and compared in a unified presentation. This article aims to explicitly demonstrate, through reviews of previous literature, simulated analysis and real-world case studies, the effects of resolution, binning, bounds, peak shape, peak asymmetry, inconsistent conversion of TOF to d spacing and merging of multiple banks in neutron TOF data as they directly relate to real-space PDF analysis. Suggestions for best practice in analysis of data from modern neutron TOF total scattering instruments when using conventional analysis programs are made, as well as recommendations for improved analysis methods and future instrument design.

Even though the importance of sharing data is frequently discussed, data sharing appears to be limited to a few fields, and practices within those fields are not well understood. This study examines perspectives on sharing neutron data collected at Oak Ridge National Laboratory’s neutron sources. Operation at user facilities has traditionally focused on making data accessible to those who create them. The recent emphasis on open data is shifting the focus to ensure that the data produced are reusable by others. This mixed methods research study included a series of surveys and focus group interviews in which 13 data consumers, data managers, and data producers answered questions about their perspectives on sharing neutron data. Data consumers reported interest in reusing neutron data for comparison/verification of results against their own measurements and testing new theories using existing data. They also stressed the importance of establishing context for data, including how data are produced, how samples are prepared, units of measurement, and how temperatures are determined. Data managers expressed reservations about reusing others’ data because they were not always sure if they could trust whether the people responsible for interpreting data did so correctly. Data producers described concerns about their data being misused, competing with other users, and over-reliance on data producers to understand data. We present the Consumers Managers Producers (CMP) Model for understanding the interplay of each group regarding data sharing. We conclude with policy and system recommendations and discuss directions for future research.

The development of novel lignin-based carbon composite anodes consisting of nanocrystalline and amorphous domains motivates the understanding of the relationship of the structural properties characterizing these materials, such as crystallite size, intracrystallitedspacing, crystalline volume fraction and composite density, with their pair distribution functions (PDFs), obtained from both molecular dynamics simulation and neutron scattering. A model for these composite materials is developed as a function of experimentally measurable parameters and realized in 15 composite systems, three of which directly match all parameters of their experimental counterparts. The accurate reproduction of the experimental PDFs using the model systems validates the model. The decomposition of the simulated PDFs provides an understanding of each feature in the PDF and allows for the development of a mapping between the defining characteristics of the PDF and the material properties of interest.

Reduction in reversible hydrogen storage capacity with increasing hydrogenation and dehydrogenation cycle number is observed in numerous hydrogen storage materials, but the mechanism behind this unfavorable change has not been elucidated yet. In this study, we have investigated the development of structural defects or disorders in V1–xTixH2, x = 0, 0.2, and 0.5, during the first 15 hydrogen absorption and desorption cycles using the atomic pair distribution function (PDF) analysis of synchrotron X-ray total scattering data to find out the possible structural origin of the poor cyclic stability of V1–xTix alloys. While pure vanadium shows no significant change in the PDF, alloy samples subject to several hydrogenation and dehydrogenation cycles display fast decaying of the PDF profile due to a progressive increase in the PDF peak width with increasing r. This r-dependent PDF peak broadening effect becomes stronger with cycle number. Molecular dynamics (MD) simulations demonstrated that dislocation defects ...

The crystal structure and defect chemistry of hydrothermally synthesized LiFe1-xMnxPO4 (x = 0, 0.25, and 0.50) particles have been characterized by simultaneous neutron and X-ray Rietveld refinement as well as X-ray and neutron pair distribution function (PDF) analysis, crystallinity determination, Mossbauer spectroscopy, ion coupled plasma (ICP) studies, and scanning electron microscopy (SEM). The very detailed structural refinements show that fast hydrothermal synthesis causes partial Fe-occupancy and vacancies on the Li (M1) site, while the Fe (M2) site is always fully occupied by iron. Thus, the defect is not merely a Li/Fe antisite defect, and excessive amounts of Fe are the origin of the disorder in the structure. Neutron and X-ray total scattering with PDF analysis show that after fast hydrothermal synthesis, the crystalline, defective LixFeyPO4 coexists with amorphous Li/Fe-PO4 structures having just short-range order. Iron excess is only seen in the crystalline part of the particles, and as the c...

Pair distribution function (PDF) analysis is a long-established technique for studying the local structure of amorphous and disordered crystalline materials. In today's increasingly complex materials landscape, the coexistence of amorphous and crystalline phases within single samples is not uncommon. Though a couple of reports have been published studying samples with amorphous and crystalline phases utilizing PDF analysis, to date little has been done to determine the sensitivity that the method currently has in resolving such contributions. This article reports a series of experiments that have been conducted on samples with known ratios of crystalline quartz and amorphous glassy silica to examine this question in detail. Systematic methods are proposed to obtain the best possible resolution in samples with unknown phase ratios and some problems that one might encounter during analysis are discussed.

This paper presents the outcomes of a series of beamline-based studies, the results of which are combined to provide a more detailed multiscale understanding of the structure and chemistry of geopolymer binders. The range of beamline-based characterization techniques which have been applied to the study of geopolymer binders is increasing rapidly; although no single technique can provide a holistic view of binder structure across all the length scales which are of importance in determining strength development and durability, the synergy achievable through the combination of multiple beamline techniques is leading to rapid advances in knowledge in this area. Studies based around beamline infrared and X-ray fluorescence microscopy, in situ and ex situ neutron pair distribution function analysis, and nano- and micro-tomography, are combined to provide an understanding of geopolymer gel chemistry, nano- and microstructure in two and three dimensions, and the influences of seeded nucleation and precursor chemistry in these key areas. The application of advanced characterization methods in recent years has brought the understanding of geopolymer chemistry from a point, not more than a decade ago, when the analysis of the detailed chemistry of the aluminosilicate binder gel was considered intractable due to its disordered (“X-ray amorphous”) nature, to the present day where the influence of key compositional parameters on nanostructure is well understood, and both gel structure and reaction kinetics can be manipulated through methods including seeding, temperature variation, and careful mix design. This paper therefore provides a review outlining the value of nanotechnology – and particularly nanostructural characterization – in the development and optimization of a new class of environmentally beneficial cements and concretes. Key engineering parameters, in particularly strength development and permeability, are determined at a nanostructural level, and so it is essential that gel structures can be analyzed and manipulated at this level; beamline-based characterization techniques are critical in providing the ability to achieve this goal.

The ORNL Spallation Neutron Source (SNS) provides the most intense pulsed neutron beams in the world for scientific research and development across a broad range of disciplines. SNS experiments produce large volumes of complex data that are analyzed by scientists with varying degrees of experience using 3D visualization and analysis systems. However, it is notoriously difficult to achieve proficiency with 3D visualizations. Because 3D representations are key to understanding the neutron scattering data, scientists are unable to analyze their data in a timely fashion resulting in inefficient use of the limited and expensive SNS beam time. We believe a more intuitive interface for exploring neutron scattering data can be created by combining immersive virtual reality technology with high performance data analytics and human interaction. In this paper, we present our initial investigations of immersive visualization concepts as well as our vision for an immersive visual analytics framework that could lower the barriers to 3D exploratory data analysis of neutron scattering data at the SNS.

Modeling the pair distribution function (PDF) of molecular compounds is a challenging task because intra- and intermolecular interactions lead to very different features in the PDF. This article discusses the different peak shapes in PDFs of molecular compounds in detail. Moreover, the common methods to calculate PDFs from structural models are summarized and evaluated with respect to molecular systems and an approach to calculate PDFs from molecular crystals more accurately is introduced.p-Terphenyl was chosen as a test compound. It adopts a crystal structure with disordered features and short-range order. The short-range order was previously investigated by analyzing single-crystal diffuse scattering and it was also extracted from experimental PDFs during this study.

The local structure of mechanically alloyed Mg50Co50 and its deuteride, Mg50Co50D75, was investigated using the atomic pair distribution function (PDF) analysis of neutron and synchrotron X-ray total scattering data. The purpose of our study was to obtain the structural information of these amorphous alloys and understand their hydrogen storage properties. We found that the body centered cubic (bcc) structural model with the uniform distribution of Mg and Co atoms proposed by earlier studies was not able to explain neutron and X-ray PDFs at the same time. Our data suggest that the local environment of Mg is different from that of Co. A two-phase model composed of MgCo2 and Mg2CoD5 (Mg2Co for the alloy sample) qualitatively reproduced the main features of experimental PDFs and explained the amount of hydrogen absorbed by MgxCo100-x.

To date, the fundamental details of the molecular structural changes and associated mechanisms, which take place during the formation of aluminosilicate geopolymer gels, have remained largely elusive. Here, density functional theory-based coarse-grained Monte Carlo modeling, a multiscale simulation technique, is used to simulate the geopolymerization reaction and to determine the molecular mechanisms controlling this process. Silica supplied by the alkaline solution plays a significant role in enhancing the dissolution of the solid aluminosilicate precursor (metakaolin, in this case) and the polymerization of the gel. In the reaction between NaOH and metakaolin, in the absence of initially dissolved silica, the solid precursor completely dissolves and the aluminosilicate gel forms via the percolation of small aluminosilicate clusters. On the other hand, in the presence of dissolved silicate, the metakaolin only partially dissolves, as the aluminosilicate gel precipitates on the surfaces of the metakaolin particle after a period of time. © 2011 American Institute of Chemical Engineers AIChE J, 2012

The decomposition of hexacarbonyltungsten, W(CO)6, has been studied. The decomposition was induced by heating W(CO)6in an autoclave at 523 K and pressures up to 1.8 MPa, and by laser heating in a diamond anvil cell at pressures between 5 and 18 GPa. The products have been characterized using synchrotron X-ray diffraction, pair distribution function analysis, Raman spectroscopy and scanning electron microscopy. Decomposition in the autoclave at the lower pressures resulted in the formation of a metastable tungsten carbide, W2C, with an average particle size of 1–2 nm, and an unidentified nanocrystalline tungsten oxide and nanocrystalline graphite with average particle sizes of 1–2 and 11 nm, respectively. The existence of nanocrystalline graphite was deduced from micro-Raman spectra and the graphite particle size was extracted from the intensities of the Raman modes. The high-pressure decomposition products obtained in the diamond anvil cell are the monoclinic tungsten oxide phase WO2and the high-pressure phase W3O8(I). The approximate average size of the graphite particles formed here was 6–8 nm. The bulk modulus of W(CO)6isB0≃ 13 GPa.

Nuclear incoherent neutron scattering contributions present a challenge in the structural characterization of many classes of materials. This article introduces methods for the correction of nanoparticle, bulk crystalline and amorphous powder neutron scattering data with significant incoherent contributions from hydrogen, and describes the effects the corrections have on the resulting atomic pair distribution function data sets. The approach is presented in the context of thePDFgetNdata-reduction program [Peterson, Gutmann, Proffen & Billinge (2000).J. Appl. Cryst.33, 1192].

The mechanistic details of the structural changes occurring during the initial stages of silica polymerization to form gels or zeolites remain largely unknown due mainly to the complexity of sol–gel synthesis processes. Previous simulation studies have applied simple lattice models to qualitatively replicate spontaneous silica nanoparticle formation using fitted interaction energy parameters to replicate the behavior of real systems. This study moves for the first time to the use of quantum chemical-based interaction (dimerization) energies, determined through density functional theory computations, in a coarse-grained Monte Carlo simulation of the initial stages of gel/cluster formation in sodium silicate systems across a range of concentrations. The use of accurate dimerization energies as model inputs enables semiquantitatively accurate results to be obtained, as determined by comparisons with 29Si nuclear magnetic resonance data. The most concentrated system simulated (9.36 m silica concentration) und...

Fluctuations in the microstructure with chemical alloying of Fe multicomponent bulk metallic glasses were characterized via the pair density function analysis of neutron diffraction data. In the two systems investigated, (Fe/Mo)76P12C10B2 and Fe49Cr15Mo14Er1(C/B)21, it is shown that the atomic topology is sensitive to even small chemical fluctuations. In phosphorous containing glasses, (Fe/Mo)76P12C10B2, increasing the atomic percent of Mo brings about changes of the microstructure, particularly in the short-range environment involving Mo-C/B and Fe-P pair correlations. In Fe49Cr15Mo14Er1(C/B)21, the systematic increase of B results in a shift of the center of mass of the first coordination shell and an increase in the coordination of the second shell, reflecting a local rearrangement brought upon the substitution of B for C.

Geopolymer cement is fast becoming a technologically important alternative to ceramics and traditional cement. However, the amorphous nature of the phases which participate in the molecular processes occurring during evolution of geopolymer gel has made nanoscale research challenging. Here, for the first time, the local structural correlations of metakaolin-based geopolymer gel have been elucidated using in situ neutron pair distribution function analysis, following the structural changes occurring due to dissolution and repolymerization molecular processes. Over the initial 17 h of reaction, the subtle structural changes observed predominantly relate to dissolution of the initial metakaolin precursor before formation of the gel. After 90 days the gel has formed and has transitioned from the initially formed geopolymer structure (gel 1) to a more stable and more ordered state (gel 2), via an increase in cross-linking within the geopolymer gel. Through analysis of precursor dissolution behavior in different activator solutions, the impact of morphology on the rate of dissolution has been postulated, with layered precursors (metakaolin) shown to behave differently than spherical precursors (fly ash) depending on the type of activator solution used. Hence, this investigation reveals the important structural changes occurring during synthesis of this new class of low-temperature ceramics.

Milling-time-dependent local structural evolution of mechanically alloyed Mg50Co50 was investigated by the atomic pair distribution function analysis using both neutron and synchrotron X-ray powder...

High-energy X-ray and spallation neutron total scattering data provide information about each pair of atoms in a nanoparticle sample, allowing for quantitative whole-particle structural modeling based on pair distribution function analysis. The realization of this capability has been hindered by a lack of versatile tools for describing complex finite structures. Here, the implementation of whole-particle refinement for complete nanoparticle systems is described within two programs,DISCUSandDIFFEV, and the diverse capabilities they present are demonstrated. The build-up of internal atomic structure (including defects, chemical ordering and other types of disorder), and nanoparticle size, shape and architecture (including core–shell structures, surface relaxation and ligand capping), are demonstrated using the programDISCUS. The structure refinement of a complete nanoparticle system (4 nm Au particles with organic capping ligands at the surface), based on neutron pair distribution function data, is demonstrated usingDIFFEV, a program using a differential evolutionary algorithm to generate parameter values. These methods are a valuable addition to other probes appropriate for nanomaterials, adaptable to a diverse and complex set of materials systems, and extendable to additional data-set types.

Kaolinite intercalated with potassium acetate is of great interest in the areas of environmental remediation and industrial application; however, its exact atomic structure and the changes which occur when heated have remained largely elusive. Here, neutron pair distribution function analysis is used to investigate the local structural characteristics of this complex material, revealing that hydrated potassium acetate exists as a single layer in the interlamellar spacing of kaolinite. Furthermore, the potassium ions within the intercalated complex are most likely associated with the resonance structure of the acetate molecules, and upon heating (and decomposition of the carbon containing molecules), these ions become strongly associated with the negative charge located on the oxygen atoms in the alumina layers of dehydroxylated kaolinite. Several possible orientations of hydrated potassium acetate within the interlamellar spacing of kaolinite have been proposed and investigated using density functional mo...

A Hot Dipole In a ferroelectric material, there is an alignment of local electric dipole moments that produces a net overall electric polarization. This state is accompanied by a decrease in symmetry, which can be restored by heating above a critical temperature. In contrast, through a combination of theory and experiments, Božin et al. (p. 1660 ) now show that with increasing temperature, rock-salt–structured lead telluride and lead sulfide go through a phase transition from a high symmetry phase to a low symmetry phase with an associated dipole moment. Paradoxically, the dipoles are stabilized in the disordered state at high temperature, even though the undistorted structure has lower internal energy.