We introduce a fingerprint representation of molecules based on a Fourier series of atomic radial distribution functions. This fingerprint is unique (except for chirality), continuous, and differentiable with respect to atomic coordinates and nuclear charges. It is invariant with respect to translation, rotation, and nuclear permutation, and requires no preconceived knowledge about chemical bonding, topology, or electronic orbitals. As such, it meets many important criteria for a good molecular representation, suggesting its usefulness for machine learning models of molecular properties trained across chemical compound space. To assess the performance of this new descriptor, we have trained machine learning models of molecular enthalpies of atomization for training sets with up to 10 k organic molecules, drawn at random from a published set of 134 k organic molecules with an average atomization enthalpy of over 1770 kcal/mol. We validate the descriptor on all remaining molecules of the 134 k set. For a training set of 10 k molecules, the fingerprint descriptor achieves a mean absolute error of 8.0 kcal/mol. This is slightly worse than the performance attained using the Coulomb matrix, another popular alternative, reaching 6.2 kcal/mol for the same training and test sets.Such approaches have already delivered convincing results for highly relevant applications, such as enhanced sampling, [6] screening of heterogeneous catalyst candidates based on Sabatier's principle, [7] and devising simple materials design rules leading to topological insulators, semiconductors, and others. [8] With increasingly available simulation data stemming from routine applications of first principles methods, such as Born-Oppenheimer or Car-Parrinello molecular dynamics, [9] statistical ML methods can be applied, in the hope of detecting trends and relationships that hitherto were difficult, if not impossible, to spot for the human expert. Applications of such approaches include data-mining for crystal structure discovery, [10] regression for reorganization energies that enter Marcus charge transfer rates, [11,12] learning of potential energy surfaces [a] O.
Eighty years after its experimental discovery, a description of induced nuclear fission based solely on the interactions between neutrons and protons and quantum many-body methods still poses formidable challenges. The goal of this paper is to contribute to the development of a predictive microscopic framework for the accurate calculation of static properties of fission fragments for hot fission and thermal or slow neutrons. To this end, we focus on the 239 Pu(n,f) reaction and employ nuclear density functional theory with Skyrme energy densities. Potential energy surfaces are computed at the Hartree-Fock-Bogoliubov approximation with up to five collective variables. We find that the triaxial degree of freedom plays an important role, both near the fission barrier and at scission. The impact of the parameterization of the Skyrme energy density and the role of pairing correlations on deformation properties from the ground-state up to scission are also quantified. We introduce a general template for the quantitative description of fission fragment properties. It is based on the careful analysis of scission configurations, using both advanced topological methods and recently proposed quantum many-body techniques. We conclude that an accurate prediction of fission fragment properties at low incident neutron energies, although technologically demanding, should be within the reach of current nuclear density functional theory.
Scientific data is continually increasing in complexity, variety and size, making efficient visualization and specifically rendering an ongoing challenge. Traditional rasterization-based visualization approaches encounter performance and quality limitations, particularly in HPC environments without dedicated rendering hardware. In this paper, we present OSPRay, a turn-key CPU ray tracing framework oriented towards production-use scientific visualization which can utilize varying SIMD widths and multiple device backends found across diverse HPC resources. This framework provides a high-quality, efficient CPU-based solution for typical visualization workloads, which has already been integrated into several prevalent visualization packages. We show that this system delivers the performance, high-level API simplicity, and modular device support needed to provide a compelling new rendering framework for implementing efficient scientific visualization workflows.
Figure 1: Several animated models ray traced using our coherent grid traversal: a) A gesturing hand of 16K triangles. b) An animated "Poser" model (78K triangles). c) Animated wind-up toys (11K triangles) walking and jumping incoherently around each other. d) A rigid-body dynamics simulation of marbles (8.8K triangles). e) A complex scene of 174K animated triangles, where a fairy and a dragonfly dance through an animated forest. Scenes are rebuilt from scratch every frame, allowing fully dynamic animation. Including shading, texturing, and hard shadows, as used in the above images, we can render these scenes at 1024 × 1024 pixels with 15.3, 7.8, 10.2, 26.2, and 1.4 frames per second on a dual 3.2 GHz Xeon. Excluding shading, texturing, and shadows, we achieve 34.5, 15.8, 29.3, 57.1, and 3.4 frames per second. AbstractWe present a new approach to interactive ray tracing of moderatesized animated scenes based on traversing frustum-bounded packets of coherent rays through uniform grids. By incrementally computing the overlap of the frustum with a slice of grid cells, we accelerate grid traversal by more than a factor of 10, and achieve ray tracing performance competitive with the fastest known packet-based kd-tree ray tracers. The ability to efficiently rebuild the grid on every frame enables this performance even for fully dynamic scenes that typically challenge interactive ray tracing systems.
Figure 1: Fiber Surfaces of electron density and reduced gradient in an ethane-diol molecule: (a) While an isosurface of electron density identifies regions of influence of atoms (grey), it does not distinguish atomic type. An isosurface of reduced gradient identifies bonding interaction sites (blue) but does not distinguish non-covalent (top) from covalent bonds (others). (b) Continuous scatter plot (log scale) of electron density and reduced gradient. Isosurfaces and fiber surfaces are shown as dashed lines and polygons respectively. (c) Fiber surfaces distinguish atom types (oxygen in red, carbons in grey) as well as bond types (non-covalent in green, covalent in blue). AbstractScientific visualization has many effective methods for examining and exploring scalar and vector fields, but rather fewer for multi-variate fields. We report the first general purpose approach for the interactive extraction of geometric separating surfaces in bivariate fields. This method is based on fiber surfaces: surfaces constructed from sets of fibers, the multivariate analogues of isolines. We show simple methods for fiber surface definition and extraction. In particular, we show a simple and efficient fiber surface extraction algorithm based on Marching Cubes. We also show how to construct fiber surfaces interactively with geometric primitives in the range of the function. We then extend this to build user interfaces that generate parameterized families of fiber surfaces with respect to arbitrary polylines and polygons. In the special case of isovalue-gradient plots, fiber surfaces capture features geometrically for quantitative analysis that have previously only been analysed visually and qualitatively using multi-dimensional transfer functions in volume rendering. We also demonstrate fiber surface extraction on a variety of bivariate data
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.