Jeremy A. Templeton

Model based design of a microfluidic mixer driven by induced charge electroosmosis

Mixing chemical or biological samples with reagents for chemical analysis is one of the most time consuming operations on microfluidic platforms. This is primarily due to the low rate of diffusive transport in liquid systems. Additionally, much research has focused on detection, rather than sample preparation. In response, we describe a mixer for microfluidic sample preparation based on the electrokinetic phenomenon of induced-charge-electroosmosis (ICEO). ICEO creates microvortices within a fluidic channel by application of alternating current (AC) electric fields. The microvortices are driven by electrostatic forces acting on the ionic charge induced by the field near polarizable materials. By enabling mixing to be turned on or off within a channel of fixed volume, these electronically controlled mixers prevent sample dilution-a common problem with other strategies. A three-dimensional model based on the finite volume method was developed to calculate the electric field, fluid flow, and mass transport in a multi-species liquid. After preliminary experiments, the model was used to rapidly prototype a wide range of designs. A new microfabrication process was developed for devices with vertical sidewalls having conductive metal coatings and embedded electrodes. Mixing experiments were carried out in the devices and the results were compared to the model.

A material frame approach for evaluating continuum variables in atomistic simulations

We present a material frame formulation analogous to the spatial frame formulation developed by Hardy, whereby expressions for continuum mechanical variables such as stress and heat flux are derived from atomic-scale quantities intrinsic to molecular simulation. This formulation is ideally suited for developing an atomistic-to-continuum correspondence for solid mechanics problems. We derive expressions for the first Piola-Kirchhoff (P-K) stress tensor and the material frame heat flux vector directly from the momentum and energy balances using localization functions in a reference configuration. The resulting P-K stress tensor, unlike the Cauchy expression, has no explicit kinetic contribution. The referential heat flux vector likewise lacks the kinetic contribution appearing in its spatial frame counterpart. Using a proof for a special case and molecular dynamics simulations, we show that our P-K stress expression nonetheless represents a full measure of stress that is consistent with both the system virial and the Cauchy stress expression developed by Hardy. We also present an expanded formulation to define continuum variables from micromorphic continuum theory, which is suitable for the analysis of materials represented by directional bonding at the atomic scale.

Machine learning strategies for systems with invariance properties

In many scientific fields, empirical models are employed to facilitate computational simulations of engineering systems. For example, in fluid mechanics, empirical Reynolds stress closures enable computationally-efficient Reynolds Averaged Navier Stokes simulations. Likewise, in solid mechanics, constitutive relations between the stress and strain in a material are required in deformation analysis. Traditional methods for developing and tuning empirical models usually combine physical intuition with simple regression techniques on limited data sets. The rise of high performance computing has led to a growing availability of high fidelity simulation data. These data open up the possibility of using machine learning algorithms, such as random forests or neural networks, to develop more accurate and general empirical models. A key question when using data-driven algorithms to develop these empirical models is how domain knowledge should be incorporated into the machine learning process. This paper will specifically address physical systems that possess symmetry or invariance properties. Two different methods for teaching a machine learning model an invariance property are compared. In the first method, a basis of invariant inputs is constructed, and the machine learning model is trained upon this basis, thereby embedding the invariance into the model. In the second method, the algorithm is trained on multiple transformations of the raw input data until the model learns invariance to that transformation. Results are discussed for two case studies: one in turbulence modeling and one in crystal elasticity. It is shown that in both cases embedding the invariance property into the input features yields higher performance at significantly reduced computational training costs.

Comparison of Molecular Dynamics with Classical Density Functional and Poisson–Boltzmann Theories of the Electric Double Layer in Nanochannels

Comparisons are made among Molecular Dynamics (MD), Classical Density Functional Theory (c-DFT), and Poisson–Boltzmann (PB) modeling of the electric double layer (EDL) for the nonprimitive three component model (3CM) in which the two ion species and solvent molecules are all of finite size. Unlike previous comparisons between c-DFT and Monte Carlo (MC), the present 3CM incorporates Lennard-Jones interactions rather than hard-sphere and hard-wall repulsions. c-DFT and MD results are compared over normalized surface charges ranging from 0.2 to 1.75 and bulk ion concentrations from 10 mM to 1 M. Agreement between the two, assessed by electric surface potential and ion density profiles, is found to be quite good. Wall potentials predicted by PB begin to depart significantly from c-DFT and MD for charge densities exceeding 0.3. Successive layers are observed to charge in a sequential manner such that the solvent becomes fully excluded from each layer before the onset of the next layer. Ultimately, this layer filling phenomenon results in fluid structures, Debye lengths, and electric surface potentials vastly different from the classical PB predictions.

Application of a field-based method to spatially varying thermal transport problems in molecular dynamics

This paper derives a methodology to enable spatial and temporal control of thermally inhomogeneous molecular dynamics (MD) simulations. The primary goal is to perform non-equilibrium MD of thermal transport analogous to continuum solutions of heat flow which have complex initial and boundary conditions, moving MD beyond quasi-equilibrium simulations using periodic boundary conditions. In our paradigm, the entire spatial domain is filled with atoms and overlaid with a finite element (FE) mesh. The representation of continuous variables on this mesh allows fixed temperature and fixed heat flux boundary conditions to be applied, non-equilibrium initial conditions to be imposed and source terms to be added to the atomistic system. In effect, the FE mesh defines a large length scale over which atomic quantities can be locally averaged to derive continuous fields. Unlike coupling methods which require a surrogate model of thermal transport like Fouriers law, in this work the FE grid is only employed for its projection, averaging and interpolation properties. Inherent in this approach is the assumption that MD observables of interest, e.g. temperature, can be mapped to a continuous representation in a non-equilibrium setting. This assumption is taken advantage of to derive a single, unified set of control forces based on Gaussian isokinetic thermostats to regulate the temperature and heat flux locally in the MD. Example problems are used to illustrate potential applications. In addition to the physical results, data relevant to understanding the numerical effects of the method on these systems are also presented.

Atomistic and molecular effects in electric double layers at high surface charges

The Poisson-Boltzmann theory for electrolytes near a charged surface is known to be invalid due to unaccounted physics associated with high ion concentration regimes. To investigate this regime, fluids density functional theory (f-DFT) and molecular dynamics (MD) simulations were used to determine electric surface potential as a function of surface charge. Based on these detailed computations, for electrolytes with nonpolar solvent, the surface potential is shown to depend quadratically on the surface charge in the high charge limit. We demonstrate that modified Poisson-Boltzmann theories can model this limit if they are augmented with atomic packing densities provided by MD. However, when the solvent is a highly polar molecule, water in this case, an intermediate regime is identified in which a constant capacitance is realized. Simulation results demonstrate the mechanism underlying this regime, and for the salt water system studied here, it persists throughout the range of physically realistic surface charge densities so the potentials quadratic surface charge dependence is not obtained.

A Long-Range Electric Field Solver for Molecular Dynamics Based on Atomistic-to-Continuum Modeling

Understanding charge transport processes at a molecular level is currently hindered by a lack of appropriate models for incorporating nonperiodic, anisotropic electric fields in molecular dynamics (MD) simulations. In this work, we develop a model for including electric fields in MD using an atomistic-to-continuum framework. This framework provides the mathematical and the algorithmic infrastructure to couple finite element (FE) representations of continuous data with atomic data. Our model represents the electric potential on a FE mesh satisfying a Poisson equation with source terms determined by the distribution of the atomic charges. Boundary conditions can be imposed naturally using the FE description of the potential, which then propagate to each atom through modified forces. The method is verified using simulations where analytical solutions are known or comparisons can be made to existing techniques. In addition, a calculation of a salt water solution in a silicon nanochannel is performed to demonstrate the method in a target scientific application in which ions are attracted to charged surfaces in the presence of electric fields and interfering media.

Polarization as a field variable from molecular dynamics simulations

A theoretical and computational framework for systematically calculating the macroscopic polarization density as a field variable from molecular dynamics simulations is presented. This is done by extending the celebrated Irving and Kirkwood [J. Chem. Phys. 18, 817 (1950)] procedure, which expresses macroscopic stresses and heat fluxes in terms of the atomic variables, to the case of electrostatics. The resultant macroscopic polarization density contains molecular dipole, quadrupole, and higher-order moments, and can be calculated to a desired accuracy depending on the degree of the coarse-graining function used to connect the molecular and continuum scales. The theoretical and computational framework is verified by recovering the dielectric constant of bulk water. Finally, the theory is applied to calculate the spatial variation of the polarization vector in the electrical double layer of a 1:1 electrolyte solution. Here, an intermediate asymptotic length scale is revealed in a specific region, which validates the application of mean field Poisson-Boltzmann theory to describe this region. Also, using the existence of this asymptotic length scale, the lengths of the diffuse and condensed/Stern layers are identified accurately, demonstrating that this framework may be used to characterize electrical double layers over a wide range of concentrations of solutions and surface charges.

Thermal Transport Mechanisms in Carbon Nanotube-Nanofluids Identified From Molecular Dynamics Simulations

Atomistic simulations of carbon nanotubes (CNTs) in a liquid environment are performed to better understand thermal transport in CNT-based nanofluids. Thermal conductivity is studied using nonequilibrium molecular dynamics (MD) methods to understand the effective conductivity of a solvated CNT combined with a novel application of Hamilton‐Crosser (HC) theory to estimate the conductivity of a fluid suspension of CNTs. Simulation results show how the presence of the fluid affects the CNTs ability to transport heat by disrupting the low-frequency acoustic phonons of the CNT. A spatially dependent use of the Irving‐Kirkwood relations reveals the localized heat flux, illuminating the heat transfer pathways in the composite material. Model results can be consistently incorporated into HC theory by considering ensembles of CNTs and their surrounding fluid as being present in the liquid. The simulation-informed theory is shown to be consistent with existing experimental results. [DOI: 10.1115/1.4029913]

Optimal Compressed Sensing and Reconstruction of Unstructured Mesh Datasets

Exascale computing promises quantities of data too large to efficiently store and transfer across networks in order to be able to analyze and visualize the results. We investigate compressed sensing (CS) as an in situ method to reduce the size of the data as it is being generated during a large-scale simulation. CS works by sampling the data on the computational cluster within an alternative function space such as wavelet bases and then reconstructing back to the original space on visualization platforms. While much work has gone into exploring CS on structured datasets, such as image data, we investigate its usefulness for point clouds such as unstructured mesh datasets often found in finite element simulations. We sample using a technique that exhibits low coherence with tree wavelets found to be suitable for point clouds. We reconstruct using the stagewise orthogonal matching pursuit algorithm that we improved to facilitate automated use in batch jobs. We analyze the achievable compression ratios and the quality and accuracy of reconstructed results at each compression ratio. In the considered case studies, we are able to achieve compression ratios up to two orders of magnitude with reasonable reconstruction accuracy and minimal visual deterioration in the data. Our results suggest that, compared to other compression techniques, CS is attractive in cases where the compression overhead has to be minimized and where the reconstruction cost is not a significant concern.