Hello, A new project on the LCRC cluster has been requested. Please forward the information on to the LCRC Allocation sub-committee. Applicant's name: Badri Narayanan Applicant's institution: ANL Applicant's division: NST Project Name: CoC_tersoff Project title: Tersoff potential for cobalt-carbon systems to model organic-inorganic heterostructures Tersoff potential for cobalt-carbon systems to model organic-inorganic heterostructures Tersoff potential for cobalt-carbon systems to model organic-inorganic heterostructures Tersoff potential for cobalt-carbon systems to model organic-inorganic heterostructures Associated funding: LDRD Other Systems: Carbon (NST; Argonne) : 600,000 Science: A fundamental understanding of the inter-relationships between structure, morphology, chemistry, and physical properties of mixed metallic-covalent systems is essential to design novel functional materials for applications in flexible nano-electronics, energy storage and catalysis. The dynamic atomic-scale processes occurring at the interfaces hold the key to such knowledge; to explore these atomistic mechanisms, it is imperative to describe the atomic interactions in a robust, accurate, and a computationally efficient manner. One of the 3d transition metal of exceptional importance in technology is cobalt, especially at nano-meter length scales. It has a wide range of applications encompassing catalysis, magnetic devices, shape memory alloys, and molecular electronics [1-4]. In particular, the size- and shape-dependent reactivity of Co nanoparticles (NPs) has been leveraged to catalyze various chemical processes, such as ammonia synthesis from N2, Fischer-Tropsch process for producing hydrocarbons from H2 and CO, and carbon nanotube formation (CNT) [2,3]. Recently, hybrid hierarchical architectures made up of porous network of metal (e.g., Co) NPs and organic material (mainly carbon binder) has attracted a lot of attention owing to their tunable elastic, electronic, and optical properties. As such, these nanoscale structures hold tremendous promise in various specific applications, e.g., energy storage, flexible opto-electronics, and magnetic devices [4,5]. Despite significant advances in the synthesis protocols for organic-inorganic hybrid hetero-structures and metal catalyzed CNTs, several questions pertaining to the chemistry, atomic structure, and dynamical processes at the metal-carbon interface still remain unanswered. For instance, a few open questions include probing (1) factors that govern chirality of a grown CNT, (2) the principal atomic forces that govern the growth of organic-inorganic hybrid structures a nd the final architecture of the assembly, (3) stability of the assembled structure, (4) dependence of the pore morphology on the size/shape of metal NPs, and carbon content, and (5) relationship between morphology of the aggregate porous structure and its physical properties. Molecular dynamics (MD) simulations based on classical interatomic potentials provide an ideal route to address these issues owing to the associated length (tens of nm), and timescales (several ns). The success of global optimization and MD techniques hinges on the ability of the employed empirical force field (EFF) to accurately describe interatomic interactions. It is a common practice in the literature to employ embedded atom method (EAM) to describe interactions between metal atoms, in which an effective local electron density term is added to pairwise interactions [1]. Such a spherically symmetric potential works well for bulk systems; however, the lack of 3-body terms leads to poor description of clusters, surfaces, and interfaces wherein bonding directionality has been found to be important [6,7]. For cobalt, these angular effects have been considered in the ReaxFF framework [2], and in an atomic coordination dependent potential form by Shibuta et al [1]. These potentials are, however, either computationally expensive for large systems (ReaxFF), or have been trained against limited data set (only upto 4-atom clusters for Shibuta potential). On the other hand, carb on, a covalently bonded material, is conventionally treated using Tersoff type bond order potential (BOP) to describe directional bonds in a computationally robust and efficient way [8]. Brenner demonstrated that the Tersoff BOP is equivalent to EAM [9]; indeed, the Tersoff BOP has been successfully employed to a variety of metals (e.g, Au, Fe) [6]. Furthermore, this provides a unique opportunity to describe Co-Co, Co-C, and C-C interactions within a single formalism, which is well equipped to probe atomic scale processes at the metal-carbon interface, as well as in bulk metal and carbon. In a broader context, this approach can be extended to any system that exhibits mixed metallic-covalent character. In this project, we propose to parameterize a Tersoff BOP for the binary Co-C system, which can accurately describe both the pure components (i.e, Co and C) as well as the compound (Co-C). To achieve a new generation of robust, accurate and yet computationally efficient FF for studying dynamics and structure of Co-C hetero-structures, it is crucial to employ a extensive training set comprising of forces and energies computed using density functional theory (DFT) calculations on numerous structures (containing only Co, only C, and Co-C) that are both near and far-away from equilibrium. Once the Tersoff BOP is developed for Co-C, it would (a) enable a fundamental understanding of the factors governing growth of hybrid organic-inorganic porous networks, and their final architecture, (b) unravel the atomic-scale dynamical processes that occur at Co-C interfaces and their impact on observable characteristics, e.g., chirality of grown CNT, (c) identify structure-property relations hips in hybrid organic-inorganic architectures properties via MD simulations, as well as (d) establish a novel methodology for developing force fields that can be easily extended to other relevant systems (e.g., Li-Si). In a broader context, the development of such a generic methodology for robust force-field generation is imperative to supply necessary computational tools to access the length and time scales necessary for studying atomic-scale processes at reactive interfaces. Such a fundamental understanding is urgently required to make truly pathbreaking advances in the design of functional materials at the nano-scale for energy applications. [1] G. P. Pun, and Y. Mishin, Phys. Rev. B 86, 134116 (2012). [2] X-Q. Zhang, E. Iype, S. V. Nedea, A.P.J. Jansen, B.M. Szyja, E.J.M. Hensen, and R. A. van Santen, J. Phys. Chem. C. 118, 6882 (2014). [3] Y. Shibuta, and S. Maruyama, Comp. Mat. Sci 39, 842 (2007). [4] J.T. Bahns, S.K.R.S. Sankaranarayanan, S.K. Gray, and L. Chen, Phys. Rev. Lett. 106, 095501 (2011). [5] J.T. Bahns, S.K.R.S. Sankaranarayanan, N.C. Giebink, H. Xiong, and S.K. Gray, Adv. Mater. 24, OP242 (2012). [6] M. Backman, N. Juslin, and K. Nordlund, Eur. Phys. J. B 85, 317 (2012). [7] B. Narayanan, A. Kinaci, F.G. Sen, M.J. Davis, S.K. Gray, M.K.Y. Chan, and S. K. R. S. Sankaranarayanan, Describing the diverse geometries of gold nanoclusters – a first principles based bond order potential, In review at Journal of Chemical Theory and Computation [8] J. Tersoff, Phys Rev B, 39, 5566 (1989) [9] D.W. Brenner, Phys. Rev. Lett. 63, 1022 (1989) Project description: The proposed work involves generating an extensive training dataset computed using DFT calculations that would be employed to parameterize Tersoff BOP for Co-C system. For these DFT calculations, we will employ the plane-wave DFT package VASP using the relativistic Perdew-Burke-Erzenhoff pseudopotentials in the framework of generalized gradient approximation. The kinetic energy cut-off for the plane waves will be chosen appropriately to ensure convergence and sufficient accuracy of energy and force values. The training set would contain DFT computed (1) lattice parameters, cohesive energies, and elastic properties for 5 bulk polymorphs of Co, namely hexagonal close-packed, face-centered cubic, body-centered cubic, simple cubic, and diamond cubic to consider various possible atomic coordinations, (2) energy values for each of these bulk phases will also be computed at 11 different volumetric strains (20% tensile to 20% compressive) to account for far-from equilibrium configurations. (3) energies of 6 low/high index Co surfaces, (4) adsorption energies of Co on graphene, and C on hexagonal close packed Co surfaces at three different sites (for each), (5) structural properties, cohesive energies, and equation of state (11 volumetric strains) of two metastable Co carbides (Co2C, Co3C) , and (6) energy as a function of separation distance for Co-C dimer at 11 different Co-C separations. In all, we need to perform 110 DFT calculations each of which requires ~3 hours on 32 cores; so, we expect to need 110 * 32 * 5 = 17,600 core hours. For parameterizing Tersoff BOP, genetic algorithms (GA) based approach will be employed, wherein, an initial population of possible solutions (here, sets of parameters) is randomly generated; these solutions are then ranked by their fitness according to an objective function. Several genetic operators, e.g., crossover, mutation etc. are then applied to the solutions of high fitness to result in a new generation of solutions (offsprings); this process is iteratively performed until the parameter-set that results in global minimum of the objective function (usually weighted mean-squared differences between FF and DFT values). The parameterization will be performed in two stages: 1) first, the parameters for Co-Co interactions will be obtained by fitting against the training set for pure Co configurations, 2) then the Co-C cross interactions will be fitted to the adsorption energies (Co on graphene/ C on Co surface), Co-C bond stretching curve, as well as structure/energetics of metastable Co carbides. The C-C interactions will be taken from Ref. 9. Using GA, the parameter space will be exhaustively searched for a global minimum in the objective function; this would require performing about 20 searches over different domains in the parameter space. Each GA search takes ~144 hours on 8 cores; so we expect 20 * 144 * 8 = 23,040 core hours. Once the GA runs converge, we will perform local optimization using Simplex around 5 different parameter sets from the final population of each run to obtain the nearest local minimum. Each of these take ~ 10 hours on 8 cores; so we expect 5 * 20 * 10 * 8 = 8,000 core hours Using the Tersoff BOP for Co-C system optimized in this study, we will perform two sets of MD simulations to investigate 1) the atomic forces that govern the assembly of Co clusters and C60 fullerenes into hybrid organic-inorganic porous hetero-structures, and 2) atomic scale mechanisms underlying chirality of CNT grown on a Co cluster. For studying the assembly of Co clusters and C60 fullerenes, we will perform large-scale MD simulations, wherein Co and C clusters will initially be placed at random locations in the computational supercell. We will then monitor the morphological evolution of the aggregate by employing canonical (isothermal) MD simulations in LAMMPS for 1 ns. Two different temperatures will be employed, namely 600 K, and 1000 K. 3 different sizes of Co clusters (100, 200, 500 atoms) will be employed; the number of Co clusters will be kept constant at 100 for all runs. At each of 3 Co cluster sizes, we will perform runs at 5 different Co/C ratios (100/1, 50/1, 1/1, 1/50, 1/100) while keeping the number density constant. Each of these MD runs will take ~50 hours on 256 cores; so we expect 2* 3* 5 * 50 * 256 = 384,000 core hours. For studying the growth of CNT on Co clusters, isolated C atoms will be deposited on the surface of Co cluster placed in the center of a computational supercell. 2 different sizes of Co cluster (100, 500) will be employed. For each of these sizes, C atoms will be deposited on Co cluster at 800 K at 10 different deposition rates over a time period of 50 ns. Each of these calculations will take ~96 hours on 32 cores; so we expect 2 * 10 * 32 * 96 = 61440 hours Total core hours requested = 475,000 Industry partnership: Project URL: Requested allocation: 475000 Q1: 0 Q2: 0 Q3: 0 Q4: 475000 Justification: Storage requirements: The requester has used undetermined amount hours of their initial startup project. In addition to approving an initial amount, please specify a Category and Subcategory for this project. For a list of the current selection of approved categories, please see: https://wiki.lcrc.anl.gov/wiki/Processes/Categories Once the Allocation committee has approved the project, please go to the Project Management page to create it: https://accounts.lcrc.anl.gov/projects.php Thank You, The LCRC Accounts System