An official website of the United States government
Machine learning potentials (MLPs) for atomistic simulations have an enormous prospective impact on materials modeling, offering orders of magnitude speedup over density functional theory (DFT) calculations without appreciably sacrificing accuracy in the prediction of material properties. However, the generation of large datasets needed for training MLPs is daunting. Herein, we show that MLP-based material property predictions converge faster with respect to precision for Brillouin zone integrations than DFT-based property predictions. We demonstrate that this phenomenon is robust across material properties for different metallic systems. Further, we provide statistical error metrics to accurately determine a priori the precision level required of DFT training datasets for MLPs to ensure accelerated convergence of material property predictions, thus significantly reducing the computational expense of MLP development.
more »
« less
Advancing the design of gold nanomaterials with machine-learned potentials
Gold nanoparticles (NPs), and their smaller (< 2 nm) counterpart, known as gold nanoclusters (NCs), have emerged in recent years as highly efficient catalysts. They exhibit unique properties, are highly tailorable, and are highly promising for applications in nanomedicine, sensing, and bioimaging. The design of nanomaterials with optimal properties hinges on our ability to understand and control their structure-function relationship, which has remained a challenge so far. The dual organic-metallic nature of ligand-protected Au NCs complicates the experimental characterization of their structure. Density Functional Theory (DFT) calculations are highly accurate but have a high computational cost, making such calculations on large NPs and over long simulation times beyond our reach. Classical simulations allow for a thorough exploration of the configuration space but the empirical force fields they rely on often lack accuracy. In this Topical Review, we discuss recent advances enabled by Machine-Learned Potentials (MLPs), which have the ability to predict energies and atomic forces with DFT-like accuracy for a fraction of the computational cost and can be readily used in molecular simulations. We further show how MLPs have led to the elucidation of the structure, stability, thermodynamics, and reactivity of nanomaterials, thereby paving the way for the accelerated computationally-guided design of Au nanomaterials.
more »
« less
- Award ID(s):
- 2247574
- PAR ID:
- 10591782
- Publisher / Repository:
- Institute of Physics
- Date Published:
- Journal Name:
- Nano Express
- Volume:
- 6
- Issue:
- 2
- ISSN:
- 2632-959X
- Page Range / eLocation ID:
- 022001
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Incorporation of metallic nanoparticles (NPs) in polymer matrix has been used to enhance and control dissolution and release of drugs, for targeted drug delivery, as antimicrobial agents, localized heat sources, and for unique optoelectronic applications. Gold NPs in particular exhibit a plasmonic response that has been utilized for photothermal energy conversion. Because plasmonic nanoparticles typically exhibit a plasmon resonance frequency similar to the visible light spectrum, they present as good candidates for direct photothermal conversion with enhanced solar thermal efficiency in these wavelengths. In our work, we have incorporated ∼3-nm-diameter colloidal gold (Au c ) NPs into electrospun polyethylene glycol (PEG) fibers to utilize the nanoparticle plasmonic response for localized heating and melting of the polymer to release medical treatment. Au c and Au c in PEG (PEG+Au c ) both exhibited a minimum reflectivity at 522 nm or approximately green wavelengths of light under ultraviolet-visible (UV-Vis) spectroscopy. PEG+Au c ES fibers revealed a blue shift in minimum reflectivity at 504 nm. UV-Vis spectra were used to calculate the theoretical efficiency enhancement of PEG+Au c versus PEG alone, finding an approximate increase of 10 % under broad spectrum white light interrogation, and ∼14 % when illuminated with green light. Au c enhanced polymers were ES directly onto resistance temperature detectors and interrogated with green laser light so that temperature change could be recorded. Results showed a maximum increase of 8.9 °C. To further understand how gold nanomaterials effect the complex optical properties of our materials, spectroscopic ellipsometry was used. Using spectroscopic ellipsometry and modeling with CompleteEASE® software, the complex optical constants of our materials were determined. The complex optical constant n (index of refraction) provided us with optical density properties related to light wavelength divided by velocity, and k (extinction coefficient) was used to show the absorptive properties of the materials.more » « less
-
Since their discovery, thiolate-protected gold nanoclusters (Au n (SR) m ) have garnered a lot of interest due to their fascinating properties and “magic-number” stability. However, models describing the thermodynamic stability and electronic properties of these nanostructures as a function of their size are missing in the literature. Herein, we employ first principles calculations to rationalize the stability of fifteen experimentally determined gold nanoclusters in conjunction with a recently developed thermodynamic stability theory on small Au nanoclusters (≤102 Au atoms). Our results demonstrate that the thermodynamic stability theory can capture the stability of large, atomically precise nanoclusters, Au 279 (SR) 84 , Au 246 (SR) 80 , and Au 146 (SR) 57 , suggesting its applicability over larger cluster size regimes than its original development. Importantly, we develop structure–property relationships on Au nanoclusters, connecting their ionization potential and electron affinity to the number of gold atoms within the nanocluster. Altogether, a computational scheme is described that can aid experimental efforts towards a property-specific, targeted synthesis of gold nanoclusters.more » « less
-
In this work, we propose a balanced multicomponent and multilayer neural network (MMNN) structure to accurately and efficiently approximate functions with complex features in terms of both degrees of freedom and computational cost. The main idea is inspired by a multicomponent approach in which each component can be effectively approximated by a single-layer network, combined with a multilayer decomposition strategy to capture the complexity of the target function. Although MMNNs can be viewed as a simple modification of fully connected neural networks (FCNNs) or multilayer perceptrons (MLPs) by introducing balanced multicomponent structures, they achieve a significant reduction in training parameters, a much more efficient training process, and improved accuracy compared to FCNNs or MLPs. Extensive numerical experiments demonstrate the effectiveness of MMNNs in approximating highly oscillatory functions and their ability to automatically adapt to localized features. Our code and implementations are available at GitHub.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript
- Journal Article:
- https://doi.org/10.1088/2632-959X/add6bc
