Machine Learning Predictions of Simulated Self-Diffusion Coefficients for Bulk and Confined Pure Liquids.

Diffusion properties of bulk fluids have been predicted using empirical expressions and machine learning (ML) models, suggesting that predictions of diffusion also should be possible for fluids in confined environments. The ability to quickly and accurately predict diffusion in porous materials would enable new discoveries and spur development in relevant technologies such as separations, catalysis, batteries, and subsurface applications. In this work, we apply artificial neural network (ANN) models to predict the simulated self-diffusion coefficients of real liquids in both bulk and pore environments. The training data sets were generated from molecular dynamics (MD) simulations of Lennard-Jones particles representing a diverse set of 14 molecules ranging from ammonia to dodecane over a range of liquid pressures and temperatures. Planar, cylindrical, and hexagonal pore models consisted of walls composed of carbon atoms. Our simple model for these liquids was primarily used to generate ANN training data, but the simulated self-diffusion coefficients of bulk liquids show excellent agreement with experimental diffusion coefficients. ANN models based on simple descriptors accurately reproduced the MD diffusion data for both bulk and confined liquids, including the trend of increased mobility in large pores relative to the corresponding bulk liquid.

Adsorption at Nanoconfined Solid-Water Interfaces.

Reactions at solid-water interfaces play a foundational role in water treatment systems, catalysis, and chemical separations, and in predicting chemical fate and transport in the environment. Over the last century, experimental measurements and computational models have made tremendous progress in capturing reactions at solid surfaces. The interfacial reactivity of a solid surface, however, can change dramatically and unexpectedly when it is confined to the nanoscale. Nanoconfinement can arise in different geometries such as pores/cages (3D confinement), channels (2D confinement), and slits (1D confinement). Therefore, measurements on unconfined surfaces, and molecular models parameterized based on these measurements, fail to capture chemical behaviors under nanoconfinement. This review evaluates recent experimental and theoretical advances, with a focus on adsorption at solid-water interfaces. We review how nanoconfinement alters the physico-chemical properties of water, and how the structure and dynamics of nanoconfined water dictate energetics, pathways, and products of adsorption in nanopores. Finally, the implications of these findings and future research directions are discussed.

Probing electrolyte-silica interactions through simulations of the infrared spectroscopy of nanoscale pores.

The structural and dynamical properties of nanoconfined solutions can differ dramatically from those of the corresponding bulk systems. Understanding the changes induced by confinement is central to controlling the behavior of synthetic nanostructured materials and predicting the characteristics of biological and geochemical systems. A key outstanding issue is how the molecular-level behavior of nanoconfined electrolyte solutions is reflected in different experimental, particularly spectroscopic, measurements. This is addressed here through molecular dynamics simulations of the OH stretching infrared (IR) spectroscopy of NaCl, NaBr, and NaI solutions in isotopically dilute HOD/D2O confined in hydroxylated amorphous silica slit pores of width 1-6 nm and pH ∼2. In addition, the water reorientation dynamics and spectral diffusion, accessible by pump-probe anisotropy and two-dimensional IR measurements, are investigated. The aim is to elucidate the effect of salt identity, confinement, and salt concentration on the vibrational spectra. It is found that the IR spectra of the electrolyte solutions are only modestly blue-shifted upon confinement in amorphous silica slit pores, with both the size of the shift and linewidth increasing with the halide size, but these effects are suppressed as the salt concentration is increased. This indicates the limitations of linear IR spectroscopy as a probe of confined water. However, the OH reorientational and spectral diffusion dynamics are significantly slowed by confinement even at the lowest concentrations. The retardation of the dynamics eases with increasing salt concentration and pore width, but it exhibits a more complex behavior as a function of halide.

Ab initio molecular dynamics free energy study of enhanced copper (II) dimerization on mineral surfaces.

Understanding the adsorption of isolated metal cations from water on to mineral surfaces is critical for toxic waste retention and cleanup in the environment. Heterogeneous nucleation of metal oxyhydroxides and other minerals on material surfaces is key to crystal growth and dissolution. The link connecting these two areas, namely cation dimerization and polymerization, is far less understood. In this work we apply ab initio molecular dynamics calculations to examine the coordination structure of hydroxide-bridged Cu(II) dimers, and the free energy changes associated with Cu(II) dimerization on silica surfaces. The dimer dissociation pathway involves sequential breaking of two Cu2+-OH- bonds, yielding three local minima in the free energy profiles associated with 0-2 OH- bridges between the metal cations, and requires the design of a (to our knowledge) novel reaction coordinate for the simulations. Cu(II) adsorbed on silica surfaces are found to exhibit stronger tendency towards dimerization than when residing in water. Cluster-plus-implicit-solvent methods yield incorrect trends if OH- hydration is not correctly depicted. The predicted free energy landscapes are consistent with fast equilibrium times (seconds) among adsorbed structures, and favor Cu2+ dimer formation on silica surfaces over monomer adsorption.

Using Computationally-Determined Properties for Machine Learning Prediction of Self-Diffusion Coefficients in Pure Liquids.

The ability to predict transport properties of liquids quickly and accurately will greatly improve our understanding of fluid properties both in bulk and complex mixtures, as well as in confined environments. Such information could then be used in the design of materials and processes for applications ranging from energy production and storage to manufacturing processes. As a first step, we consider the use of machine learning (ML) methods to predict the diffusion properties of pure liquids. Recent results have shown that Artificial Neural Networks (ANNs) can effectively predict the diffusion of pure compounds based on the use of experimental properties as the model inputs. In the current study, a similar ANN approach is applied to modeling diffusion of pure liquids using fluid properties obtained exclusively from molecular simulations. A diverse set of 102 pure liquids is considered, ranging from small polar molecules (e.g., water) to large nonpolar molecules (e.g., octane). Self-diffusion coefficients were obtained from classical molecular dynamics (MD) simulations. Since nearly all the molecules are organic compounds, a general set of force field parameters for organic molecules was used. The MD methods are validated by comparing physical and thermodynamic properties with experiment. Computational input features for the ANN include physical properties obtained from the MD simulations as well as molecular properties from quantum calculations of individual molecules. Fluid properties describing the local liquid structure were obtained from center of mass radial distribution functions (COM-RDFs). Feature sensitivity analysis revealed that isothermal compressibility, heat of vaporization, and the thermal expansion coefficient were the most impactful properties used as input for the ANN model to predict the MD simulated self-diffusion coefficients. The MD-based ANN successfully predicts the MD self-diffusion coefficients with only a subset (2 to 3) of the available computationally determined input features required. A separate ANN model was developed using literature experimental self-diffusion coefficients as model targets. Although this second ML model was not as successful due to a limited number of data points, a good correlation is still observed between experimental and ML predicted self-diffusion coefficients.

Quantum Calculations of VX Ammonolysis and Hydrolysis Pathways via Hydrated Lithium Nitride.

Recently, lithium nitride (Li3N) has been proposed as a chemical warfare agent (CWA) neutralization reagent for its ability to produce nucleophilic ammonia molecules and hydroxide ions in aqueous solution. Quantum chemical calculations can provide insight into the Li3N neutralization process that has been studied experimentally. Here, we calculate reaction-free energies associated with the Li3N-based neutralization of the CWA VX using quantum chemical density functional theory and ab initio methods. We find that alkaline hydrolysis is more favorable to either ammonolysis or neutral hydrolysis for initial P-S and P-O bond cleavages. Reaction-free energies of subsequent reactions are calculated to determine the full reaction pathway. Notably, products predicted from favorable reactions have been identified in previous experiments.

Simulations of the IR and Raman spectra of water confined in amorphous silica slit pores.

Water in nano-scale confining environments is a key element in many biological, material, and geological systems. The structure and dynamics of the liquid can be dramatically modified under these conditions. Probing these changes can be challenging, but vibrational spectroscopy has emerged as a powerful tool for investigating their behavior. A critical, evolving component of this approach is a detailed understanding of the connection between spectroscopic features and molecular-level details. In this paper, this issue is addressed by using molecular dynamics simulations to simulate the linear infrared (IR) and Raman spectra for isotopically dilute HOD in D2O confined in hydroxylated amorphous silica slit pores. The effect of slit-pore width and hydroxyl density on the silica surface on the vibrational spectra is also investigated. The primary effect of confinement is a blueshift in the frequency of OH groups donating a hydrogen bond to the silica surface. This appears as a slight shift in the total (measurable) spectra but is clearly seen in the distance-based IR and Raman spectra. Analysis indicates that these changes upon confinement are associated with the weaker hydrogen-bond accepting properties of silica oxygens compared to water molecules.

Ab initio and force field molecular dynamics study of bulk organophosphorus and organochlorine liquid structures.

We performed ab initio molecular dynamics (AIMD) simulations to benchmark bulk liquid structures and to evaluate results from all-atom force field molecular dynamics (FFMD) simulations with the generalized Amber force field (GAFF) for organophosphorus (OP) and organochlorine (OC) compounds. Our work also addresses the current and important topic of force field validation, applied here to a set of nonaqueous organic liquids. Our approach differs from standard treatments, which validate force fields based on thermodynamic data. Utilizing radial distribution functions (RDFs), our results show that GAFF reproduces the AIMD-predicted asymmetric liquid structures moderately well for OP compounds that contain bulky alkyl groups. Among the OCs, RDFs obtained from FFMD overlap well with AIMD results, with some offsets in position and peak structuring. However, re-parameterization of GAFF for some OCs is needed to reproduce fully the liquid structures predicted by AIMD. The offsets between AIMD and FFMD peak positions suggest inconsistencies in the developed force fields, but, in general, GAFF is able to capture short-ranged and long-ranged interactions of OPs and OCs observed in AIMD. Along with the local coordination structure, we also compared enthalpies of vaporization. Overall, calculated bulk properties from FFMD compared reasonably well with experimental values, suggesting that small improvements within the FF should focus on parameters that adjust the bulk liquid structures of these compounds.

Impact of intrinsic framework flexibility for selective adsorption of sarin in non-aqueous solvents using metal-organic frameworks.

Molecular modeling of mixture adsorption in nanoporous materials can provide insight into the molecular-level details that underlie adsorptive separations. Modeling of adsorption often employs a rigid framework approximation for computational convenience. All real materials, however, have intrinsic flexibility due to thermal vibrations of their atoms. In this article, we examine quantitative predictions of the adsorption selectivity for a dilute concentration of a chemical warfare agent, sarin, from bulk mixtures with aqueous and non-aqueous (methanol, isopropyl alcohol) solvents using metal-organic frameworks (MOFs). These predictions were made in MOFs approximated as rigid and also in MOFs allowed to have intrinsic flexibility. Including framework flexibility appears to have important consequences for quantitative predictions of adsorption selectivity, particularly for sarin/water mixtures. Our observations suggest the intrinsic flexibility of MOFs can have a nontrivial impact on adsorption modeling of molecular mixtures, especially for mixtures containing polar species and molecules of different sizes.

Molecular dynamics simulation of zirconium tungstate amorphization and the amorphous-crystalline interface.

Classical molecular dynamics (MD) simulations were performed to provide a conceptual understanding of the amorphous-crystalline interface for a candidate negative thermal expansion (NTE) material, ZrW2O8. Simulations of pressure-induced amorphization at 300 K indicate that an amorphous phase forms at pressures of 10 GPa and greater, and this phase persists when the pressure is subsequently decreased to 1 bar. However, the crystalline phase is recovered when the slightly distorted 5 GPa phase is relaxed to 1 bar. Simulations were also performed on a two-phase model consisting of the high-pressure amorphous phase in direct contact with the crystalline phase. Upon equilibration at 300 K and 1 bar, the crystalline phase remains unchanged beyond a thin layer of disrupted structure at the crystalline-amorphous interface. Differences in local atomic structure at the interface are quantified from the simulation trajectories.

Determining Diffusion Coefficients of Chemical Warfare Agents in Metal-Organic Frameworks.

Metal-organic frameworks (MOFs) have shown potential for selective capture of chemical warfare agents (CWAs). To determine characteristic adsorption times, the kinetics of CWA uptake in MOFs must be known. Here, we calculate diffusion coefficients of the CWA sarin and simulants in prototypical MOFs using classical molecular simulations. Sarin can diffuse throughout a one micrometer crystal in less than a second in MIL-47 and Cu-BTC, but this process takes more than 3 h in ZIF-8 and UiO-66. A simple estimate based on Knudsen diffusion is able to describe diffusion of sarin in MIL-47 but fails to do so in other MOFs. This work has implications in designing devices to detect and capture CWAs.

Spectroscopically Resolved Binding Sites for the Adsorption of Sarin Gas in a Metal-Organic Framework: Insights beyond Lewis Acidity.

Here we report molecular level details regarding the adsorption of sarin (GB) gas in a prototypical zirconium-based metal-organic framework (MOF, UiO-66). By combining predictive modeling and experimental spectroscopic techniques, we unambiguously identify several unique bindings sites within the MOF, using the P═O stretch frequency of GB as a probe. Remarkable agreement between predicted and experimental IR spectrum is demonstrated. As previously hypothesized, the undercoordinated Lewis acid metal site is the most favorable binding site. Yet multiple sites participate in the adsorption process; specifically, the Zr-chelated hydroxyl groups form hydrogen bonds with the GB molecule, and GB weakly interacts with fully coordinated metals. Importantly, this work highlights that subtle orientational effects of bound GB are observable via shifts in characteristic vibrational modes; this finding has large implications for degradation rates and opens a new route for future materials design.

Revealing Transition States during the Hydration of Clay Minerals.

A molecular-scale understanding of the transition between hydration states in clay minerals remains a challenging problem because of the very fast stepwise swelling process observed from X-ray diffraction (XRD) experiments. XRD profile modeling assumes the coexistence of multiple hydration states in a clay sample to fit the experimental XRD pattern obtained under humid conditions. While XRD profile modeling provides a macroscopic understanding of the heterogeneous hydration structure of clay minerals, a microscopic model of the transition between hydration states is still missing. Here, for the first time, we use molecular dynamics simulation to investigate the transition states between a dry interlayer, one-layer hydrate, and two-layer hydrate. We find that the hydrogen bonds that form across the interlayer at the clay particle edge make an important contribution to the energy barrier to interlayer hydration, especially for initial hydration.

Distinguishing between bulk and edge hydroxyl vibrational properties of 2 : 1 phyllosilicates via deuteration.

Observation of vibrational properties of phyllosilicate edges via a combined molecular modeling and experimental approach was performed. Deuterium exchange was utilized to isolate edge vibrational modes from their internal counterparts. The appearance of a specific peak within the broader D2O band indicates the presence of deuteration on the edge surface, and this peak is confirmed with the simulated spectra. These results are the first to unambiguously identify spectroscopic features of phyllosilicate edge sites.

Concerted Metal Cation Desorption and Proton Transfer on Deprotonated Silica Surfaces.

The adsorption equilibrium constants of monovalent and divalent cations to material surfaces in aqueous media are central to many technological, natural, and geochemical processes. Cation adsorption-desorption is often proposed to occur in concert with proton transfer on hydroxyl-covered mineral surfaces, but to date this cooperative effect has been inferred indirectly. This work applies density functional theory-based molecular dynamics simulations of explicit liquid water/mineral interfaces to calculate metal ion desorption free energies. Monodentate adsorption of Na+, Mg2+, and Cu2+ on partially deprotonated silica surfaces are considered. Na+ is predicted to be unbound, while Cu2+ exhibits binding free energies to surface SiO- groups that are larger than those of Mg2+. The predicted trends agree with competitive adsorption measurements on fumed silica surfaces. As desorption proceeds, Cu2+ dissociates one of the H2O molecules in its first solvation shell, turning into Cu2+(OH-)(H2O)3, while Mg remains Mg2+(H2O)6. The protonation state of the SiO- group at the initial binding site does not vary monotonically with cation desorption.

Enhanced Ion Adsorption on Mineral Nanoparticles.

Classical molecular dynamics simulation was used to study the adsorption of Na+, Ca2+, Ba2+, and Cl- ions on gibbsite edge (1 0 0), basal (0 0 1), and nanoparticle (NP) surfaces. The gibbsite NP consists of both basal and edge surfaces. Simulation results indicate that Na+ and Cl- ions adsorb on both (1 0 0) and (0 0 1) surfaces as inner-sphere species (i.e., no water molecules between an ion and the surface). Outer-sphere Cl- ions (i.e., one water molecule between an ion and the surface) were also found on these surfaces. On the (1 0 0) edge, Ca2+ ions adsorb as inner-sphere and outer-sphere complexes, whereas on the (0 0 1) surface, outer-sphere Ca2+ ions are the dominant species. Ba2+ ions were found as inner-sphere and outer-sphere complexes on both surfaces. Calculated ion surface coverages indicate that, for all ions, surface coverages are always higher on the basal surface compared to those on the edge surface. More importantly, surface coverages for cations on the gibbsite NP are always higher than those calculated for the (1 0 0) and (0 0 1) surfaces. This enhanced ion adsorption behavior for the NP is due to the significant number of inner-sphere cations found at NP corners. Outer-sphere cations do not contribute to the enhanced surface coverage. In addition, there is no ion adsorption enhancement observed for the Cl- ion. Our work provides a molecular-scale understanding of the relative significance of ion adsorption onto gibbsite basal versus edge surfaces and demonstrates the corner effect on ion adsorption on NPs.

Synthesis and characterization of thallium-salen derivatives for use as underground fluid flow tracers.

A pair of thallium salen derivatives was synthesized and characterized for potential use as monitors (or taggants) or as models for Group 13 complexes for subterranean fluid flows. These precursors were isolated from the reaction of thallium ethoxide with N,N'-bis(3,5-di-tert-butylsalicylidene)-ethylenediamine (H2-salo-But), or N,N'-bis(3,5-di-tert-butylsalicylidene)-1,2-phenylenediamine (H2-saloPh-But). The products were identified by single crystal X-ray diffraction as: [((µ-O)2,κ1-(N)(N')salo-But)Tl2] (1) and {[((µ-O)2saloPh-But)Tl2][((µ-O)2,κ1-(N)(N')saloPh-But)Tl2]} (2). Both structures are similar, wherein each O atom of the salo moiety bridges the two Tl atoms, leading to a TlTl interaction, which is further stabilized by an intramolecular π-bond with neighboring phenyl rings. For 1, an additional TlN interaction was solved for each metal center; whereas, for 2, one of the two molecules in the matrix has a weak TlN interaction but no bonding noted in the other molecule. Both Density Functional Theory (DFT) calculations and variable temperature solution 205Tl NMR studies of 1 and 2 further confirmed the TlTl interaction. The UV-vis absorbance spectra of these compounds had an absorbance peak at 392 nm for 1 and a broad absorbance peak centered at 469 nm for 2, which were found to be in good agreement with the DFT calculated spectra that were dominated by the singlet state. Fluorescence emission and excitation studies reveal absorptions at 360 and 380 nm for 1 and 2, respectively, which are attributed to the TlTl metal centers. To demonstrate practicality, fluorescence spectra of 1 and 2 were obtained using a handheld 405 nm cw laser pointer and portable spectrometer where compound 1 was found to glow 15 times brighter than compound 2. Only compound 1 was found to survive the simulated deep-well conditions explored, which was attributed to the TlN interaction noted for 1 but not for 2.

Synthesis and Characterization of Structurally Diverse Alkaline-Earth Salen Compounds for Subterranean Fluid Flow Tracking.

A family of magnesium and calcium salen-derivatives was synthesized and characterized for use as subterranean fluid flow monitors. For the Mg complexes, di- n-butyl magnesium ([Mg(Bu n)2]) was reacted with N, N'-ethylene bis(salicylideneimine) (H2-salen), N, N'-bis(salicylidene)-1,2-phenylenediamine (H2-saloPh), N, N'-bis(3,5-di- t-butylsalicylidene)-ethylenediamine (H2-salo-Bu t), or N, N'-bis(3,5-di- t-butylsalicylidene)-1,2-phenylenediamine (H2-saloPh-Bu t), and the products were identified by single-crystal X-ray diffraction as [(κ3-(O,N,N'),µ-(O')saloPh)(µ-(O),(κ2-(N,N'),µ-(O')saloPh)2(µ-(O),κ3-(N,N',O')saloPh')Mg4]·2tol (1·2tol; saloPh' = an alkyl-modified saloPh derivative generated in situ), [(κ4-(O,N,N',O')saloPh)Mg(py)2]·py (2·py), [(κ4-(O,N,N',O')salo-Bu t)Mg(py)2] (3), [(κ4-(O,N,N',O')saloPh-Bu t)Mg(py)2]·tol (4·tol), and [(κ3-(O,N,N'),µ-(O')saloPh-Bu t)Mg]2 (5), where tol = toluene; py = pyridine. For the Ca species, a calcium amide was independently reacted with H2-salo-Bu t and H2-saloPh-Bu t to generate the crystallographcially characterized compounds: [(κ4-(O,N,N',O')salo-Bu t)Ca(py)3] (6), [(κ4-(O,N,N',O')saloPh-Bu t)Ca(py)3]·py (7·py). The bulk powders of these compounds were further characterized by a number of analytical tools, where 2-7 were found to be distinguishable by Fourier transform infrared and resonance Raman spectroscopies. Structural properties obtained from quantum calculations of gas-phase analogues are in good agreement with the single-crystal results. The potential utility of these compounds as taggants for monitoring subterranean fluid flows was demonstrated through a series of experiments to evaluate their stability to high temperature and pressure, interaction with mineral surfaces, and elution behavior from a loaded proppant pack.

First-Principles Structural, Mechanical, and Thermodynamic Calculations of the Negative Thermal Expansion Compound Zr2(WO4)(PO4)2.

The negative thermal expansion (NTE) material Zr2(WO4)(PO4)2 has been investigated for the first time within the framework of the density functional perturbation theory (DFPT). The structural, mechanical, and thermodynamic properties of this material have been predicted using the Perdew, Burke and Ernzerhof for solid (PBEsol) exchange-correlation functional, which showed superior accuracy over standard functionals in previous computational studies of the NTE material α-ZrW2O8. The bulk modulus calculated for Zr2(WO4)(PO4)2 using the Vinet equation of state at room temperature is K 0 = 63.6 GPa, which is in close agreement with the experimental estimate of 61.3(8) at T = 296 K. The computed mean linear coefficient of thermal expansion is -3.1 × 10-6 K-1 in the temperature range ∼0-70 K, in line with the X-ray diffraction measurements. The mean Grüneisen parameter controlling the thermal expansion of Zr2(WO4)(PO4)2 is negative below 205 K, with a minimum of -2.1 at 10 K. The calculated standard molar heat capacity and entropy are C P 0 = 287.6 and S 0 = 321.9 J·mol-1·K-1, respectively. The results reported in this study demonstrate the accuracy of DFPT/PBEsol for assessing or predicting the relationship between structural and thermomechanical properties of NTE materials.

Atomistic Structure of Mineral Nano-aggregates from Simulated Compaction and Dewatering.

The porosity of clay aggregates is an important property governing chemical reactions and fluid flow in low-permeability geologic formations and clay-based engineered barrier systems. Pore spaces in clays include interlayer and interparticle pores. Under compaction and dewatering, the size and geometry of such pore spaces may vary significantly (sub-nanometer to microns) depending on ambient physical and chemical conditions. Here we report a molecular dynamics simulation method to construct a complex and realistic clay-like nanoparticle aggregate with interparticle pores and grain boundaries. The model structure is then used to investigate the effect of dewatering and water content on micro-porosity of the aggregates. The results suggest that slow dewatering would create more compact aggregates compared to fast dewatering. Furthermore, the amount of water present in the aggregates strongly affects the particle-particle interactions and hence the aggregate structure. Detailed analyses of particle-particle and water-particle interactions provide a molecular-scale view of porosity and texture development of the aggregates. The simulation method developed here may also aid in modeling the synthesis of nanostructured materials through self-assembly of nanoparticles.