Carbon nanotube fibers have been considered as a reinforcement in lightweight composites. Despite recent improvements in carbon nanotube fiber production rates and tenacities, mechanical properties of carbon nanotube fiber composites and corresponding processing methods have been limited to tensile property studies. Consideration of carbon nanotube fiber reinforced polymer composites in structural applications requires a broad characterization of their performance under a variety of loading conditions. To that end, short beam strength, axial flexure, transverse flexure, compression, and tension tests are conducted on carbon nanotube fiber reinforced epoxy composites. For each test, composites made from three different formats of carbon nanotube fiber reinforcement with a high weight (or volume) percentage of reinforcement up to 70 wt.% (68 vol.%) are compared. Each composite shows unique mechanical behaviors dependent on the loading environment and carbon nanotube fiber format. This dataset may serve as guidance for the further improvement of carbon nanotube fiber reinforced composites' mechanical properties and processing optimization.

From batteries to biology, many important technologies and physical phenomena operate as out-of-equilibrium reactive systems. Accurately modeling the nanoscale dynamics of non-equilibrium reactive systems and how they respond to external stimuli is challenging, especially if both atomistic resolution and large scales (>105 atoms) are required. REACTER is a protocol for modeling chemical reactions during classical molecular dynamics (MD) simulations. Coupling traditional fixed-valence force fields with heuristic reactive MD is advantageous for large-scale simulations of dynamic systems that can include the complex reaction mechanisms common in organic chemistry. This paper details the current features of the LAMMPS implementation of REACTER, known as fix bond/react, and surveys recent applications of the protocol in a variety of fields, including photopolymers, high-performance composites, and membranes. Conceived as a tool for modeling polymerization processes, the scope of REACTER is expanding as it is applied to new materials and supporting features are implemented. Three new case studies are presented that highlight the capabilities of REACTER, including modeling hierarchical materials, the mechanics of molecular machines, and large-scale dynamics of heterogeneous catalysis.

Despite improvements in carbon nanotube fiber tenacities, their shear strength continues to limit their performance in composite applications. Shear strength can be improved through resin penetration into the nanotube fiber hierarchical microstructure to enhance internal interfaces. In this work, a manufacturing process uses an ionic liquid and epoxy to simultaneously stretch and infiltrate nanotube roving resulting in a nanotube-epoxy-ionic liquid prepreg. The ionic liquid also serves as a latent curing agent/hardener for the prepreg. After curing, the resulting composite fiber has comparable tenacity and specific modulus to fibers made without epoxy but with improved shear properties. The dry core shear failure mode typically observed in CNT fibers was eliminated and the apparent interfacial shear strength was improved by over an order of magnitude.

Carbon-Carbon composites (C/C composites) are an excellent thermal protection system for aerospace vehicles. The matrix material of most C/C composites, glassy carbon (GC), demonstrates excellent thermal stability and mechanical response with low mass densities. Although GC materials have been in development and use for decades, molecular simulation protocols need to be developed to accelerate the optimization of processing cycles and to drive the development of the next generation of C/C composites. This research aims to establish molecular dynamics (MD) simulation protocols to accurately predict the evolution of the molecular structure and properties of furan resin during the pyrolysis processes that convert the polymer into a GC material. MD workflows are developed using a reactive force field to simulate the evolution of the molecular structure, mass density, and elastic properties of furan resin-derived GC. MD simulation parameters are optimized, and the predicted structures and properties are shown to agree with experimental measurements from the literature. The modeling methodology established in this work can provide guidance in driving the development of the next-generation C/C composite precursor chemistries.

A simulation technique has been developed for predicting the char yield of organic resins during high-temperature processing. In silica methods can aid in the screening of new advanced materials for a number of important properties, but no chemistry-sensitive protocol currently exists for predicting the important experimental value of char yield. The proposed method utilizes a reactive force field (ReaxFF) to model the chemical transformation of precursor monomers into carbonized structures during three processing stages: ramp-up to pyrolysis temperatures (similar to 3000 K), pyrolysis, and quenching. Achieving good agreement with experimental char yields requires continuous removal of small by-product molecules to mimic outgassing, and the application of high pressure to encourage the formation of a dense, glassy network. Six different resin chemistries were investigated: an ethynyl, a phenylethynyl, a cyanate ester, a phthalonitrile, acrylonitrile, and adamantane. These candidates represent a diverse group of precursors with respect to initial cyclic content, presence of heteroatoms, and types of reactive groups. The protocol developed accurately predicts the relative char yield between the investigated chemistries and provides quantitative agreement with experimental values, especially for high char yield resins. Several simulated properties of the carbonized structures are compared with experimental results, including outgassing products, morphology of the final chemical configurations, cyclic content, and mechanical properties.

Glassy carbon (GC) material derived from pyrolyzed furan resin was modeled by using reactive molecular dynamics (MD) simulations. The MD polymerization simulation protocols to cure the furan resin precursor material are validated via comparison of the predicted density and Young's modulus with experimental values. The MD pyrolysis simulations protocols to pyrolyze the furan resin precursor is validated by comparison of calculated density, Young's modulus, carbon content, sp(2) carbon content, the in-plane crystallite size, out-of-plane crystallite stacking height, and interplanar crystallite spacing with experimental results from the literature for furan resin derived GC. The modeling methodology established in this work can provide a powerful tool for the modeling-driven design of next-generation carbon-carbon composite precursor chemistries for thermal protection systems and other high-temperature applications.

Carbon nanotube assemblies such as fibers and sheets are an emerging lightweight material class with potential to enable aerospace structures beyond what is achievable with existing materials. Load transfer within these materials can be attributed to a combination of cohesion, static friction, covalent cross-links, and entanglements. Of these mechanisms, entanglements are the least studied or understood and are not well defined when applied to nanotube materials. In this work, an entanglement is defined with sufficient detail for molecular models to be built and tested. Non-reactive models where the covalent bond topology does not change and reactive models where covalent bonds can form and break were developed. In both model types, entanglement load transfer was observed and can be attributed to buckles (i.e., wrinkles) that form under bending compression. In non-reactive models, there are energy barriers to restructure the shape of the buckles, while reactive models formed covalent bonds at the high-curvature edges of the buckles. Reactive models produced an average load transfer approximately 14 times greater than non-reactive models due to these covalent bonds.

Achieving simultaneous enhancement of multiple mechanical properties in fiber reinforced polymer composites is a difficult challenge. In this work, unidirectional carbon nanotube (CNT) fiber reinforced polymer composites have been fabricated that exhibit improvement in not only axial and transverse strength, but also Mode I fracture toughness through multi-scale hierarchical structures. The multi-scale hierarchical microstructures in CNT composites are the result of pre-infiltrating polymer resin into the loosely connected CNT networks of the roving material to strengthen the bundle networks at the nano- and micro-scale. At the macro-scale, the polymer pre-infiltrated CNT fibers were laid up in a ‘brick-and-mortar’ pattern to produce an additional level of structural interlocking in the composite. The resulting multi-scale load transfer pathways combine to increase the deflection of crack paths, the distribution of damage ahead of the crack, and the dissipation of energy during deformation. Crack-bridging toughening mechanisms operating at multiple structural scales are responsible for the improved properties in the CNT composites. The measured specific axial tensile strength of the unidirectional CNT fiber composite was 0.85 GPa/(g/cm3), which was approximately 120% of the starting CNT material {0.71 N/tex [numerically equivalent to 0.71 GPa/(g/cm3)]}. The transverse tensile strength and Mode I fracture toughness, measured at a constant crack propagation, were 81.3 MPa/(g/cm3) and 0.670 kJ/m2, which are promising values for structural composite applications.

We present a computational framework for developing physics-based, high-fidelity structure-property relationships with atomic systems. In this framework, atomic structure is quantified by directionally resolved two point spatial correlations of the charge density field, projected to a salient low-dimensional feature space via principal component analysis (PCA), and correlated to physical properties by Gaussian process regression (GPR). The charge density field provides a complete, purely physics-based definition of the atomic structure that is independent of chemical species information and does not require additional feature engineering or idealizations beyond those of first-principles computations. The two-point spatial correlations capture the salient spatial features underlying the atomic structure that dictate the physics underlying the material response. Since the feature engineering approach explored in this work is universally applicable to all atomic structures independent of the chemical species present in the structure, it offers new avenues for efficiently exploring the space of atomic structures for desired property combinations. A further contribution of this work comes from utilizing the uncertainty quantification inherently provided by GPR to deploy a Bayesian experiment design strategy to minimize the number of computationally expensive physics simulations required to achieve the desired accuracy. In this work, we demonstrate the proposed framework to elucidate the relationship between the chemical composition and bulk modulus in AlNbTiZr high entropy alloys. It is shown that a highly accurate structure-property relationship with less than 2% average error can be established using a small training dataset of less than 30 samples.

Phenolic resin is a thermosetting polymer that has historically been used as a carbon matrix precursor for carbon-carbon composite manufacturing due to its relatively high char yield. However, the complex structural and chemical changes occurring during pyrolysis are difficult to characterize in situ. This work presents a novel method for modeling the pyrolysis processes for a polymerized phenolic resin using reactive molecular dynamics. The characteristics of the pyrolyzed model structures agree with experimental X-ray diffraction studies on glassy carbon matrices, with interplanar spacings of 3.80 ± 0.06 Å and crystallite heights of 10.98 ± 0.35 Å. The resulting structures are free of defects, and the mass densities of 2.01 ± 0.03 g/cm3 and Young’s moduli of 123.29 ± 22 GPa are found to be in reasonable agreement when compared to skeletal mass densities of glassy carbon and Young’s moduli of nanoscale glassy carbon thin films, respectively10.12783/asc38/36544

The complex structural and chemical changes that occur during polymerization and pyrolysis critically affect material properties but are difficult to characterize in situ. This work presents a novel, experimentally validated methodology for modeling the complete polymerization and pyrolysis processes for phenolic resin using reactive molecular dynamics. The polymerization simulations produced polymerized structures with mass densities of 1.24 ± 0.01 g/cm3 and Young’s moduli of 3.50 ± 0.64 GPa, which are in good agreement with experimental values. The structural properties of the subsequently pyrolyzed structures were also found to be in good agreement with experimental X-ray data for the phenolic-derived carbon matrices, with interplanar spacings of 3.81 ± 0.06 Å and crystallite heights of 10.94 ± 0.37 Å. The mass densities of the pyrolyzed models, 2.01 ± 0.03 g/cm3, correspond to skeletal density values, where the volume of pores is excluded in density calculations for the phenolic resin-based pyrolyzed samples. Young’s moduli are underpredicted at 122.36 ± 16.48 GPa relative to experimental values of 146 – 256 GPa for nanoscale amorphous carbon samples.

A framework using peridynamic theory is developed and demonstrated for deformation and failure analysis of carbon nanotube (CNT) yarn-based structural composites. Experimental work involved tension testing of a CNT yarn/polymer composite resulting in stress–strain response up to and including failure. The as-prepared specimen was characterized using x-ray micro computed tomography (CT), which was then converted into voxel-based data with CNT yarn, polymer, and void phases as well as surface undulation. A Density-Based Spatial Clustering of Applications with Noise algorithm was applied to detect and quantify the clusters of voids, of CNT-rich, and of resin-rich regions. The voxel data, with all microstructural details, were used in peridynamic simulations. These demonstrate the critical roles of resin and void clusters and surface undulation in fracture initiation and propagation. Additional analysis was performed to construct probability density functions (PDFs) of different phases (yarn, resin, and void) with the goal of constructing synthetic virtual composite specimens. The synthetically reconstructed peridynamic models correctly captured the experimental stress–strain response. The similarities and differences between the failure (initiation and propagation) behaviors predicted by x-ray CT-based and PDF-based peridynamic model simulations are presented in detail and discussed.

Carbon-based composites have become indispensable materials in aerospace and other high-performance applications, yet obtaining a detailed, nanoscale understanding of their morphology and failure mechanisms using only experimental methods remains a difficult challenge. REACTER is a versatile computational modeling tool for atomistic molecular dynamics simulations designed to model chemical reactions at the speed and length scales of classical force fields. In this work, several recent features of REACTER were applied to the creation and subsequent mechanical testing of two classes of carbon composites: carbon nanotube (CNT) composites and carbon fiber reinforced polymers (CFRP). A network of CNTs was grown dynamically using the new ‘create atoms’ feature of REACTER. The CNT filler was embedded into a polyarylacetylene (PAA) matrix by simulated in situ polymerization to obtain the final composite model. To generate the second class of carbon composite, fully carbonized (graphitic) carbon fiber morphologies were created by the method of Desai et al. [1], but using the advanced reaction constraints framework of REACTER. Two fiber models were created, representing a circular carbon fiber core and a flat surface, and similarly infiltrated with resin to obtain the final CFRP structure. Failure mechanisms were elucidated by simulating mechanically induced bond breaking, as characterized by third order DFT-based tight-binding simulations, via a reaction constraint on the total potential energy of the involved atoms.

Large-diameter carbon nanotubes (CNTs) synthesized by floating-catalyst chemical vapor deposition (FC-CVD) assemble into bundles and subsequently into aerogel networks from which yarns and sheets are mechanically drawn. The CNT bundles exhibit unique microstructures with collapsed CNT packing, not found in other types of CNT yarns. At the same time, the bundle structure is not homogeneous and the wide variability of CNT cross-sectional shapes reflects the bundling process. Transmission electron microscopy (TEM) images allow detailed quantification of CNT diameters, shapes, and number of walls. Molecular dynamics (MD) simulations of CNT assemblies are subsequently built to match the observed CNT shape characteristics and mass densities. Contrary to an established notion of an applied "buckling" pressure requirement for radial collapse, MD demonstrated that interacting large-diameter CNTs can collapse spontaneously, suggesting that collapse can occur during bundling. Computational explorations of this mechanism yield conditions for large double-walled CNTs (with a mean diameter around 7.8 nm) to assemble into phases with significantly increased packing efficiencies and Young's moduli approaching 1 TPa. The results may play a guiding role in advancing FC-CVD aerogel synthesis and processing methods to yield yarn and sheets with larger densities and superior mechanical properties.

Achieving high strength in fiber reinforced structural composites requires effective load transfer between the high-performance fiber reinforcement, e.g., carbon nanotube (CNT) yarn, and the matrix. Various processing approaches to enhance the interaction between CNT fiber and the matrix were investigated. The apparent interfacial shear strengths (IFSS) of pristine CNT yarns, post-treated CNT yarns (cross-linked, functionalized, and polymer incorporation by resistive heating), and pre-infiltrated polymer/CNT composite fibers measured using single fiber pull-out tests were used to screen the efficacy of the processing methods. Pristine CNT yarns had a low apparent IFSS (<5 MPa) due to shear failure within their dry cores. In post-treated CNT yarns which did not exhibit good IFSS, the failure surface consisted of a resin-infiltrated sheath near the surface of the yarn and a dry section within the yarn core; failure occurred in the dry core. This failure mode is unlike those observed in traditional carbon fiber reinforced composites which fail at the fiber/matrix interface. In contrast to the sword-in-sheath failure modes of post-treated CNT yarns, pre-infiltrated polymer/CNT composite fibers displayed high apparent IFSS (>20 MPa). Improved wet-out of the fiber eliminated the dry-core shear failure mode.

This paper introduces voxelized atomic structure (VASt) potentials as a machine learning (ML) framework for developing interatomic potentials. The VASt framework utilizes a voxelized representation of the atomic structure directly as the input to a convolutional neural network (CNN). This allows for high-fidelity representations of highly complex and diverse spatial arrangements of the atomic environments of interest. The CNN implicitly establishes the low-dimensional features needed to correlate each atomic neighborhood to its net atomic force. The selection of the salient features of the atomic structure (i.e., feature engineering) in the VASt framework is implicit, comprehensive, automated, scalable, and highly efficient. The calibrated convolutional layers learn the complex spatial relationships and multibody interactions that govern the physics of atomic systems with remarkable fidelity. We show that VASt potentials predict highly accurate forces on two phases of silicon carbide and the thermal conductivity of silicon over a range of isotropic strain.

REACTER is a heuristic protocol that allows complex, predefined reactions to be modeled in atomistic, fixed-valence molecular dynamics (MD) simulations. The method is applicable to a broad range of chemical reactions and permits much larger and longer reactive simulations than existing approaches. One or more competing multistep reactions or series of reactions can be invoked simultaneously. Special treatment can be applied to neighboring atoms to relax high-energy configurations while the simulation progresses. The original implementation of REACTER, which was included in the open-source LAMMPS simulation package as fix bond/react, was only available for serial simulations. This work describes the expansion of the REACTER protocol for use in parallel simulations, as well as the addition of various new options, including deletion of reaction byproducts, reversible reactions, and custom reaction constraints. The capability of the parallel implementation is demonstrated through large-scale simulations (200000+ atoms) of the polymerization of polystyrene and nylon-6,6. The morphologies of both polymers are analyzed after reaching >99% extent of polymerization. Finally, the newly added reversible reactions feature is demonstrated by rupturing these highly entangled systems under uniaxial strain by defining a chain scission reaction.

To better understand the molecular level mechanisms behind the self-healing of polymer systems, this work developed and validated a molecular dynamics model for a self-healing polymer, Surlyn® 8940. The polymer chains were built using a free-radical polymerization between ethylene monomers (~95 mol %) and methacrylic acid monomers. Predicted thermal and structural properties including thermal conductivity, heat capacity, transition temperatures, thermal expansion coefficient, density, and Young’s modulus were shown to be in good agreement with experimental values. Furthermore, the effects of cluster formation among the acid groups on these properties were explored. † Corresponding author email: kevin.hadley@sdsmt.edu https://ntrs.nasa.gov/search.jsp?R=20190029225 2019-11-21T20:56:25+00:00Z

Quality control and repeatability of 3D printing must be enhanced to fully unlock its utility beyond prototyping and noncritical applications. Machine learning is a potential solution to improving 3D printing performance and is explored for areas including flaw identification and property prediction. However, critical problems must be resolved before machine learning can truly enable 3D printing to reach its potential, including the very large data sets required for training and the inherently local nature of 3D printing where the optimum parameter settings vary throughout the part. This work outlines an end-to-end tool for integrating machine learning into the 3D printing process. The tool selects the ideal parameter settings at each location, taking into consideration factors such as geometry, hardware and material response times, and operator priorities. The tool demonstrates its usefulness by correcting for visual flaws common in fused filament fabrication parts. An image recognition neural network classifies local flaws in parts to create training data. A gradient boosting classifier then predicts the local flaws in future parts, based on location, geometry, and parameter settings. The tool selects optimum parameter settings based on the aforementioned factors. The resulting prints show increased quality over prints that use global parameters only.