32 research outputs found
Materials Design With Molecular Dynamics: Novel Properties in Martensitic Alloys Through Free Energy Landscape Engineering
Get PDFNovel properties in martensitic alloys are predicted using molecular dynamics (MD) simulations, the uncertainty from input models in MD is explored and quantified, and new data infrastructure for MD is developed. The possible regimes of material properties are extended through free energy landscape engineering (FELE), where coherent, epitaxial integration of two materials at the nanoscale can result in metamaterials with fundamentally different behavior. Free energy as a function of strain is calculated using MD simulations and analytically combined to predict potential new properties for a Ni 63% - Al 37% martensitic alloy combined with B2 NiAl, shown to be an ideal candidate for modifying the martensitic landscape. Direct MD simulations show the landscape predictions hold, producing large reduction in thermal hysteresis while retaining transformation strain, second order martensitic transformations, and both tunable and ultra-low stiffness. These properties are shown for structures including nanolaminates, nanowires, and nanoprecipitates; the final case provides an example of a more easily accessible structure through standard metallurgical processing routes. The uncertainty in these results is described, including atomic level variability, cycle variability, and the interatomic model; the model uncertainty is most significant, but only generally approximated through similar results with a second, independent interatomic potential. Beginning from simpler models, functional uncertainty quantification (FunUQ) is developed to directly address the errors between multiple interatomic models. Functional derivatives describe the sensitivity to local changes to the input function and a computationally feasible calculation method for MD simulations is derived. Together with the discrepancy between two models, the functional error between models can be calculated. These capabilities are shown for relatively simple MD models, structures, and properties, with possible extension to more complex systems and properties. Challenges are addressed, with primary attention to cases where the discrepancy between input functions is large. Improvements for community use of MD simulations is finally presented through use of the nanoHUB online collaboration and cloud computing platform to improve a growing materials data infrastructure (MDI). Tools which introduce unfamiliar users to MD simulations are shown first, without coding, downloads, or installation. Next, connections with existing atomistic MDI are described including databases for interatomic models, structures, tests, and properties. Tools to document and improve MD workflows with Jupyter notebooks are then demonstrated, further connecting these tools and resources. To conclude, future possibilities and challenges with FELE are considered within martensitic materials and beyond, extension of FunUQ to complex interatomic models and applicability to other materials modeling is examined, and new directions in atomistic MDI are discussed
Harnessing mechanical instabilities at the nanoscale to achieve ultra-low stiffness metals
Get PDFAlloy and microstructure optimization have led to impressive improvements in the strength of engineering metals, while the range of Young’s moduli achievable has remained essentially unchanged. This is because stiffness is insensitive to microstructure and bounded by individual components in composites. Here we design ultra-low stiffness in fully dense, nanostructured metals via the stabilization of a mechanically unstable, negative stiffness state of a martensitic alloy by its coherent integration with a compatible, stable second component. Explicit large-scale molecular dynamics simulations of the metamaterials with state of the art potentials confirm the expected ultra-low stiffness while maintaining full strength. We find moduli as low as 2 GPa, a value typical of soft materials and over one order of magnitude lower than either constituent, defying long-standing composite bounds. Such properties are attractive for flexible electronics and implantable devices. Our concept is generally applicable and could significantly enhance materials science design space
Fast and Accurate Predictions of Total Energy for Solid Solution Alloys with Graph Convolutional Neural Networks
No full textWe use graph convolutional neural networks (GCNNs) to produce fast and accurate predictions of the total energy of solid solution binary alloys. GCNNs allow us to abstract the lattice structure of a solid material as a graph, whereby atoms are modeled as nodes and metallic bonds as edges. This representation naturally incorporates information about the structure of the material, thereby eliminating the need for computationally expensive data pre-processing which would be required with standard neural network (NN) approaches. We train GCNNs on ab-initio density functional theory (DFT) for copper-gold (CuAu) and iron-platinum (FePt) data that has been generated by running the LSMS-3 code, which implements a locally self-consistent multiple scattering method, on OLCF supercomputers Titan and Summit. GCNN outperforms the ab-initio DFT simulation by orders of magnitude in terms of computational time to produce the estimate of the total energy for a given atomic configuration of the lattice structure. We compare the predictive performance of GCNN models against a standard NN such as dense feedforward multi-layer perceptron (MLP) by using the root-mean-squared errors to quantify the predictive quality of the deep learning (DL) models. We find that the attainable accuracy of GCNNs is at least an order of magnitude better than that of the MLP
Role of electronic thermal transport in amorphous metal recrystallization: A molecular dynamics study
No full textRole of electronic thermal transport in amorphous metal recrystallization: a molecular dynamics study
Get PDFRecrystallization of glasses is important in a wide range of applications
including electronics and reactive materials. Molecular dynamics (MD) has been
used to provide an atomic picture of this process, but prior work has neglected
the thermal transport role of electrons, the dominant thermal carrier in
metallic systems. We characterize the role of electronic thermal conductivity
on the velocity of recrystallization in Ni using MD coupled to a continuum
description of electronic thermal transport via a two-temperature model. Our
simulations show that for strong enough coupling between electrons and ions,
the increased thermal conductivity removes the heat from the exothermic
recrystallization process more efficiently, leading to a lower effective
temperature at the recrystallization front and, consequently, lower propagation
velocity. We characterize how electron-phonon coupling strength and system size
affects front propagation velocity. Interestingly, we find that initial
recrystallization velocity increases with decreasing in system size due to
higher overall temperatures. Overall, we show that a more accurate description
of thermal transport due to the incorporation of electrons results in better
agreement with experiments
Tuning martensitic transformations via coherent second phases in nanolaminates using free energy landscape engineering
No full textUncharacteristic second order martensitic transformation in metals via epitaxial stress fields
No full textGovernment and elementary education in Britain in the mid-nineteenth century
Get PDFThis thesis attempts to describe the growth of the central government’s involvement in elementary education, and the corresponding growth of the staffing and expenditure of the Education Department in Whitehall, in terms that have explanatory force. It goes from 1833 to the early 1860s, covering the 1840s and 1850s in most detail. The first chapter establishes a theoretical framework within which education can take its place beside other examples of government intervention. It reasserts the relevance of A.V. Dicey's analysis of the movements of opinion and the corresponding legislative trends, and concludes that in the mid-nineteenth century a description as far as possible in terms of demand factors is the appropriate one. The next two chapters describe the structure and growth of the systems of building grants and pupil-teacher grants; and the consequences for the staffing and expenditure of the Education Department. These are traced in detail, allowing an assessment of the Department's efficiency and the adequacy of the staff to the work, and how these changed over the period. Chapter 4 examines the evidence for Treasury restrictiveness of the Education Department's activities, and finds little, contrary to the assumptions of many accounts of the period. Chapter 5 traces the development of the views of the Newcastle Commission, and of Gladstone's interventions, and relates them to the Revised Code. These are together interpreted as a reassertion, ultimately unsuccessful, of an individualist approach to government intervention against the increasingly collectivist tendency of the system as it had become
Multi-task graph neural networks for simultaneous prediction of global and atomic properties in ferromagnetic systems
No full textWe introduce a multi-tasking graph convolutional neural network, HydraGNN, to
simultaneously predict both global and atomic physical properties and
demonstrate with ferromagnetic materials. We train HydraGNN on an open-source
ab initio density functional theory (DFT) dataset for iron-platinum (FePt) with
a fixed body centered tetragonal (BCT) lattice structure and fixed volume to
simultaneously predict the mixing enthalpy (a global feature of the system),
the atomic charge transfer, and the atomic magnetic moment across
configurations that span the entire compositional range. By taking advantage of
underlying physical correlations between material properties, multi-task
learning (MTL) with HydraGNN provides effective training even with modest
amounts of data. Moreover, this is achieved with just one architecture instead
of three, as required by single-task learning (STL). The first convolutional
layers of the HydraGNN architecture are shared by all learning tasks and
extract features common to all material properties. The following layers
discriminate the features of the different properties, the results of which are
fed to the separate heads of the final layer to produce predictions. Numerical
results show that HydraGNN effectively captures the relation between the
configurational entropy and the material properties over the entire
compositional range. Overall, the accuracy of simultaneous MTL predictions is
comparable to the accuracy of the STL predictions. In addition, the
computational cost of training HydraGNN for MTL is much lower than the original
DFT calculations and also lower than training separate STL models for each
property.Comment: 13 pages, 6 figure
Tunability of martensitic behavior through coherent nanoprecipitates and other nanostructures
No full text
