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: AuFF Project title: Development of classical bond-order force field for gold nanoclusters from first principles Associated funding: LDRD (Materials for Energy Strategic Initiative) Other Systems: Carbon cluster at the Center for Nanoscale Materials; 600,000 core hours Science: Gold nano-clusters, owing to their exceptional chemical, optical, and electronic properties, hold promise in a wide range of applications, such as optoelectronics, bio-recognition, and catalysis [1]. Their enticing potential has fueled numerous theoretical and experimental investigations; however, a fundamental understanding of surface chemistry and atomic structure of Au nano-clusters is still elusive [2]. For instance, there is an open debate on the critical value of cluster size n (i.e., number of atoms) beyond which 3D isomers become energetically preferable over planar ones, with reported values ranging from 7 to 15 [2-4]. Furthermore, since many isomers at a given cluster size are energetically close to each other (~20 meV/atom) [4], it is possible that they may undergo structural transitions under the influence of external stimuli, e.g., temperature fluctuations; the knowledge of such transitions and the associated mechanisms is still in its i nfancy. Global optimization and molecular dynamics (MD) simulations based on classical potentials provide an ideal route to address these issues; however, the accuracy of these methods is severely limited by the force fields (FFs) currently available in the literature. To improve the predictive power of these available potentials, our group re-parameterized existing force fields with pre-defined functional forms by training them against thermodynamic and structural properties obtained from density functional theory (DFT) calculations. The training set was well represented by structures (clusters and bulk) that are near- and far- away from equilibrium. This work indicated that functional forms for pair wise potentials (e.g. Morse) and those that account for many body effects via a density term (e.g., Sutton-Chen) do not possess sufficient complexity to describe the whole range of structures exhibited by Au clusters such as planar, hollow cages/tubes, globular, space-filled icosahedral configurations. FFs based on bond-order formalism, e.g., Tersoff, are more likely to be capable of describing such a wide range of structures; for instance, in the case of carbon, Tersoff FF can describe graphene, nanotubes, graphite, as well as diamond [5]. Here, a FF for gold based on bond-order formalism will be developed using structural and thermodynamic properties obtained from DFT calculations. The initial efforts will be focused on a Tersoff type potential coupled with long-range attractive potentials (e.g. Lennard –Jones); we will also explore potentials with flexible functional forms determined using sparse regression techniques. To ensure robustness and accuracy of the FF, we will employ an extensive training set comprising of DFT calculated energies of numerous condensed phases, nano-clusters at various sizes, and low-/high- index surfaces. In addition, we will obtain metastable structures of Au nanoclusters at low sizes (n < 15) using evolutionary optimization techniques. Once the FF is developed for Au nano-clusters, it would (a) enable a fundamental understanding of structure of Au nano-clusters, (b) unravel the atomic-scale dynamical processes that underpin catalysis and other interesting phenomena (e.g., structural transitions) via MD simulations, as well as (c) 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] P. Pyykkö, “Theoretical chemistry of gold”, Angewandte Chemie 43, 4412 (2004). [2] S. Bulusu et al., “Evidence of hollow golden cages”, Proceedings of the National Academy of Sciences of the United States of America 103, 8326 (2006). [3] P. Gruene et al. "Structure of neutral Au7, Au19 and Au20 clusters in the gas phase" Science 321, 674 (5889). [4] X. Li, et al., “Structural study of gold clusters”, Journal of Chemical Physics 124, 114309 (2006). [5] L. Lindsay and D. Baido, “Optimized Tersoff and Brenner empirical potential parameters for lattice dynamics and phonon thermal transport in carbon nanotubes and graphene”, Physical Review B 81, 205441 (2010). Project description: For generating the training dataset of total energies and forces, we will employ the plane-wave DFT package VASP using the relativistic Perdew-Burke-Erzenhoff pseudopotentials in the framework of generalized gradient approximation. Our group is well experienced in these calculations in Fusion where the code has been tested for systems having up to 7000 electrons. For routine calculations, the code showed 70%-80% scaling efficiency for 64 to 128 cores compared to 8 cores. Given this performance, the VASP code has the efficiency to simulate large number configurations in reasonable amount of time. The calculations with classical FF will be performed using LAMMPS, which is well known to be a massively parallel molecular dynamics package, which scales excellently and has been routinely used for computational supercells containing unto million atoms. The proposed calculations along with their estimated compute time requirements are listed below: 1. Evolutionary algorithm based search for metastable clusters (planar and globular configurations) in the framework of DFT (PAW-PBE) at 2 different cluster sizes (n =12,14). Our preliminary calculations for the 13 atom cluster has shown such a search at each cluster size requires DFT relaxations of reasonable quality on ~300 random structures (~6 hours on 16 cores per DFT relaxation). So for three cluster sizes, we expect to need 2 cluster sizes * 2 (globular, planar) * 300 * 16 * 6 = 115,200 core hours. 2. At each of these cluster sizes (n = 12,14), we will perform high quality electronic structure calculations using DFT for 10 best structures to identify the electronic origins of energy differences between isomers that have identical number of dangling bonds. Such a calculation would require ~8 hours on 32 cores; so we need 2 * 10 * 8 * 32 = 5120 core hours. 3. For the fitting procedure, we would require ~1000 core hours per FF. We will fit (a) two existing bond order potentials (Tersoff, ReaxFF), (b) Tersoff potential coupled with long range attractive Lennard-Jones potential, and (c) a new reactive FF, wherein the potential is described as a linear superposition of various basis set functions. So we need 4 * 1000 = 4,000 core hours. 4. Gold nano-clusters have been found to exhibit interesting configurations, e.g., tubular Au24 and chiral Au34. To investigate whether such configurations possess exotic electronic, optical or magnetic properties, we will select 5 best structures at each size and perform high quality electronic structure calculations on them. Each of these calculations would require ~12 hours on 32 cores; so we need 45 * 10 * 6 * 32 = 172,800 core hours. 5. Finally, we will identify the atomistic origin of the shift in energetic preference from planar to globular isomers around cluster size of 13 atoms via MD simulations. Our preliminary computations have shown that a MD run of 50 ns needs ~1000 core hours. To explore the 2D-3D transition, we will study the trajectory of an ad-atom on planar Au13 cluster via MD simulations; 5 different candidate configurations for planar Au13 will be used and the simulations will be carried at 10 different temperatures; so we need 5 * 10 * 1000 = 50,000 core hours. The total time requested for this project is 350,000 core hours Dr. Badri Narayanan, Dr. Subramanian Sankaranarayanan, and Dr. Maria Chan will be responsible for carrying out the simulations. Industry partnership: Project URL: Requested allocation: 350000 Q1: 87500 Q2: 87500 Q3: 87500 Q4: 87500 Justification: Storage requirements: 1 TB 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