Hello,
A yearly allocation for the LCRC cluster has been requested with the
following updated information:
Submitter/PI: Valerie Vallet
Project Name: MDcounterions
Division: CSE
Project title: Competing solvent and counter ion coordination and its effect on metal hydration properties: a polarizable molecular dynamics simulation
Associated funding: DOE, OBES, Chemical Sciences
Other Systems: We currently have access to the French GENCI CURIE computer center, with 200,000 core-hours for calendar year 2013.
Science: Understanding the speciation of ions at liquid interfaces is the first step toward building a predictive understanding of the partitioning and transport of metal ions under a wide range of biotic and abiotic processes. In FY2013 we initiated a joint experimental-theoretical study of molecular-scale interactions that play a fundamental role in the chemistry and physical properties of solute-solvent interactions in aqueous solutions. Our objective is to improve the basic understanding of ion interactions both in the bulk and at interfaces, and to explore how the chemistry occurring at an aqueous interface differs from that seen in the bulk. Amongst their relevance, these questions are key issues in the context of liquid-liquid separation processes of spent fuel materials, in particular as they pertain to outstanding issues of lanthanide/lanthanide and lanthanide/actinide separations germane to the international energy portfolio. From the experimental side, the ANL’s
Heavy Elements group has developed an expertise in using synchrotron-generated, high-energy (> 60 keV) X-rays in scattering (HEXS) experiments to probe atom-atom correlations in solutions out to distances of 1 nm or more. Recently they have built upon this effort (Luo et al. J. Phys. Chem. C, 2013, 117, 19082–19090), by demonstrating that X-ray scattering (reflectivity measurements) from a liquid surface of aqueous ErCl3 can reveal unprecedented detail of a non-monotonic electron density profile perpendicular to the interface. These new and unexpected experimental results call for the development of an accurate and predictive theoretical model to describe the metal-ion speciation at the interface, how it is different from the bulk, and the forces driving this difference. We propose to do this using classical force fields for molecular dynamics simulations. The classical force field we use to describe all interactions present in the solution (water-water, solute-water, wa
ter-counter-ion, solute-counterion) is more sophisticated than common polarizable force fields as it accounts for all the subtle many-body interactions taking place in a solute-solvent-anion system, including hydrogen bonding and covalent interactions. This is the level of sophistication needed to reach a unifying molecular picture of the lanthanide/actinide chemistry from the gas-phase, into the bulk solution and at its interface. Such a unified description offers a potentially break-through not just in separation chemistry but also in diverse areas as atmospheric chemistry and biophysics.
Project description: Work plan and methodology
In an attempt to explain, on a molecular level, the heuristic ordering of a solution in which metal cations, solvent molecules and anions are present, we will employ molecular dynamics modeling to compare to HEXS and X-Ray reflexivity measurements. This will complement our new experimental efforts to develop surface-derived pair-distribution functions (PDF)s from APS synchrotron data. The computational project proposed for FY2015 will focus on simulations of lanthanide and actinide solutions with chloride and bromide as counter-ions. We will build up upon the work achieved in FY2013, and FY2014. In particular, as described in our project report, we have devoted significant effort towards improving the description of water-water interactions (TCPEP2013 water model), and towards testing our force-field model for other types of cations, such as alkyl-ammonium cations [V. Vallet and M. Masella. Chem. Phys. Lett., submitted], for which reliable experimental data of hydration entha
lpies and free energies are available. Our co-workers [Houriez et al. J. Phys. Chem. B, 118 (2014), 6222–6233] have shown that it is possible to extrapolate behavior of these cations in the bulk phase from molecular dynamics simulations of droplets of growing size (50−1000 water molecules). The extrapolated values match, and confirm independently, the relative and absolute experiment-based ion solvation energies. Their results agree with experimental thermochemistry, that the relative solvation energies of alkylammonium ions by only four water molecules reproduce the relative bulk solvation energies, although the small clusters lack major bulk solvation factors. The droplet results also show a slow convergence of ion solvation properties toward their bulk limit. A key advantage of droplet simulations is that they are free from certain drawbacks present in periodic simulations, such as the absence of an explicit air/liquid water interface, which makes it compulsory to pos
t-process the data. However, the reliability of the correction schemes to be applied to periodic trajectories to obtain reliable thermodynamics quantities, is still subject to debate. Thus droplets simulations with varying droplet size offer a trustful way of probing behaviors from the gas-phase to the bulk, and when the droplet size is large enough (10,000 to 100,000 water molecules, which will be feasible this winter with the coming release of Polaris MD), the droplet bears a large interface with air making it possible to probe interfacial behaviors of a given ion. In addition to comparisons with ANL experiments, these droplet simulations pair on-going experimental data (using nanoelectrospray ionization) acquired by the group of Prof. E. R. Williams at UC Berkeley (see for instance J. Am. Chem. Soc. 2011, 133, 4810–4818)
1/ We have recently developed new sets of halide-water interaction parameters with the new TCPEP2013 water. The periodic molecular dynamics simulations we have performed in the past month indicate that this interaction model is able to accurately reproduce the bulk behavior of all halides (F-, Cl-, Br-, I-, At-) and thus can be used to also predict behaviors of the short-lifetime At-. Our objective is now to see whether the model is able to reproduce experimental hydration energies, and the behavior of these anions at an air-water interface. In this aim, we will perform batch of MD simulations for halides embedded in water droplets of increasing size, and scanning the position of the anion across the droplet. We expect to see if there is a correlation between polarizability and adsorption propensity at the surface, as previously suggested by molecular dynamics simulations and electrospray ionization mass spectrometry data.
2/ for a given Ln/An element, we will use the new TCPEP2013 water model to parametrize, Ln/An-water and Ln/An-halide interactions. We wish to investigate the effect of the counter ions on Ln/An bulk properties by comparing the results of simulations conducted without and with counter ions present in the simulation box. In the latter, the counter ion will be placed in the “second coordination shell”, as experiments predict, to start from a structural configuration close to what is observed in solution.
3/ for a given Ln/An-anion (bromide) system, we will explore the changes of free energy as a function of the Ln/An-bromide distance (potential of mean-force estimated from the results 10 trajectories at fixed Ln/An-bromide distances), to see if and where it reaches a minimum, which would correspond to the formation of contact-ion pairs.
The hydration of these elements will be both explored with droplet models and with periodic boundary condition simulations.
The MD simulations will be used to generate atomistic bulk pair-correlations or electron density profiles that can be directly compared with our newly developing experimental capabilities. Code to transform the MD simulations, obtained from our LCRC project in 2013, into HEXS PDF spectra has been written and is currently available (Y.J. Hu et al, manuscript in preparation). The fundamentally new aspect of this work is that these comparisons can be made, with significant accuracies, out to distances of about 0.8 nm in real space. Such comparisons will play a critical role in assessing and refining the force fields used in simulations.
Job characteristics and performance of the POLARIS(MD) code
Each trajectory described above should be run on a 5 ns time scales to reach system equilibrium. For droplet simulations, we have to scan about 40 positions through a droplet of given size, for 5 ns per point, each trajectory running on 8 cores, for a total cost of 80000 CPU hours for a droplet of 1000 water molecules). To scan the convergence of thermodynamics properties with respect to the droplet size, these simulations have to be repeated for several clusters (typically, Nw = 50, 100, 300, 600, 1000). For periodic simulations, for a simulation box size of a cation/anion system solvated by 1000 water molecules, the calculation requests 60 hours (walltime) on 32 processors, that is about 2000 core hours.
A new version of the POLARIS(MD) code will be released soon, with the feature of including fast-multiple moments techniques to speed up molecular dynamics simulations of large droplets. The program runs efficiently with hybrid MPI/OpenMP parallelization using 8 openMP processes across several hundred nodes. However, the MPI parallelization of the POLARIS(MD) code is being improved to enhance the scalability with more than 1000 CPUs. This development work is done with the help of Intel scientists and the French Exascale Research Laboratory (ERL). It consists in rewriting parts of the code in order to prepare for forthcoming exascale multicore computers. The new version of our code will be ready in the coming months.
Requested computer time.
Summing up the needed simulations for all five halide elements and Ln/An hydrated elements, a total of 1,000,000 hours is requested.
Industry partnership:
Project URL:
Current FY Hours Used: undetermined amount
New FY Requested allocation: 1200000
Q1: 300000
Q2: 300000
Q3: 300000
Q4: 300000
Justification: We justify the request of 1,500,000 from the large number of simultaneous trajectories that have to be explored to sample the phase-space of our systems and obtain thermodynamics data
Storage requirements: We request 15 Tb, to temporarily store trajectories for post-processing and thermodynamics integration.
Thank You,
The LCRC Accounts System
