HPC Day 2016 at UMass Dartmouth
Titles and abstracts

Zlatan Aksamija (UMass Amherst) — Numerical Simulation of Thermal Transport in Semiconductor Nanostructures

In this talk, I will review my group’s recent efforts to study vibrational modes of semiconductor alloys (such as SiGe, and SiSn) and 2-dimensional materials (graphene and transition metal dichalcogenides). We use the code Quantum Espresso (QE) to perform Density Functional Perturbation theory calculations and obtain the vibrational modes from first principles. We use the output of QE in combination with an in-house Boltzmann transport solver code to study thermal transport in nanostructures made from aforementioned 3-dimensional alloys (predominantly group IV including SiGe and SiSn) and 2-dimensional materials. We are emphasizing in our work the role of extrinsic effects, such as boundaries of a nanostructure, grain boundaries in poly/nanocrystalline materials, and heterogeneous interfaces between dissimilar materials. Our calculations rely on the Massachusetts Green High Performance Computing Center (MGHPCC) in Holyoke, MA and our NSF-funded cluster installed recently at the MGHPCC.
Our results show that extrinsic effects are very significant in both 2D and 3D nanostructures and dominate the transfer of heat across grain boundaries and interfaces in both types of materials. This finding has potential impact on heat management in next-generation nanoelectronics but more immediately in designing more efficient solid-state thermo-electric (TE) energy converters for waste-heat scavenging from the many heat sources (car exhaust, industrial plants, etc.). The reduction of thermal conductivity we observe in group IV alloys points the way to novel TE converters which take advantage of the combination of alloy disorder and nanostructuring in the form of thin films to achieve larger TE conversion efficiency. 2-dimensional materials including graphene and TMDCs are grown by Chemical Vapor Deposition (CVD) and are hence poly/nanocrystalline. Our work shows that grain boundaries between single-crystalline domains in graphene and TMDCs significantly reduce thermal conductivity and impede heat removal from 2-d nanoelectronics.

Slides is available.

Valeri Barsegov (Umass Lowell) — Influence of Solvent-Induced Hydrodynamic Interactions on Dynamic Structural Transitions in Protein Assemblies

The hydrodynamic interactions (HI) are incorporated into the $C_\alpha$ based protein model by using Rotne-Prager-Yamakawa tensor and truncated expansion formalism. Computational performance of the model accelerated on a GPU demonstrates the model’s capability for describing protein systems of varying complexity ($10^2 - 10^5$ residues): from small polyprotein to large protein complex and to entire biological particles. We found that the HI couple unfolding of proteins, and speed up dissociation of non-covalent bonds and collapse of biological particles by facilitating cross-communication among protein domains, which result in more collective displacements of structure elements governed by more deterministic (less variable) dynamics. Hence, the biomolecular simulations without HI overestimate the role of tension/stress fluctuations. These findings establish the importance of including implicit water-mediated many-body effects into coarse-grained modeling of biological systems.

Vanni Bucci (Umass Dartmouth) — Engineering microbial systems for the prevention of enteric infections

Antibiotic treatment of bacterial infections has become one of society's biggest challenges as it results in the spreading of often deadly antibiotic-resistant bacteria. Alternative approaches that avoid the emergence of resistance are crucial for the advancement of modern medicine. Here I will present a summary of work from my laboratory where a combination of mathematical modeling, bioinformatics and experimentation is used to prototype therapies aimed at solving this problem. Specifically, I will illustrate the development and the application of novel methods for the inference of microbiome dynamical systems from DNA sequencing data that we used for the design of probiotic cocktails that provide protection against C. difficile infection and modulate host immunity.

Geoff Cowles (Umass Dartmouth) — A Multiscale Approach to Modeling the Tidal-Stream Energy Resource

Tidal in-stream energy conversion (TISEC) facilities provide a highly predictable and dependable source of energy. Given the economic and social incentives to migrate towards renewable energy sources there has been tremendous interest in the technology. Key challenges to the design process stem from the wide range of problem scales extending from device to array. In the present approach we apply a multi-model approach to bridge the scales of interest and select optimal device geometries to estimate the technical resource for several realistic sites in the coastal waters of Massachusetts, USA. The approach links two computational models. To establish flow conditions at site scales (~10m), a barotropic setup of the unstructured grid ocean model FVCOM is employed. The model is validated using shipboard and fixed ADCP as well as pressure data. For device scale, the structured multiblock flow solver SUmb is selected. A large ensemble of simulations of 2D cross-flow tidal turbines is used to construct a surrogate design model. The surrogate model is then queried using velocity profiles extracted from the tidal model to determine the optimal geometry for the conditions at each site.

Slides is available.

Hanchen Huang (Northeastern University) — Computation Enabled Discovery of Smallest Metallic Nanorods and Innovation of Metallic Glue in Ambient

Computation plays an important role in our daily life, as well as work. Focusing on a particular scientific topic – nanorod growth, this presentation serves to illustrate how computations have led to the discovery of smallest metallic nanorods, and thereby to the innovation of metallic glue in ambient. To start, this talk will first introduce the atomistic computational methods, in synergy with analytical formulations and experiments; the references at the end of this abstract may provide additional background information. Then, this talk turns to the discovery of new diffusion kinetics using a combination of classical molecular dynamics and quantum mechanics computations, continues on to identification of error in a classical growth theory using lattice kinetic Monte Carlo computations, and concludes with the formulation of an analytical theory for nanorod growth that is guided and verified by lattice kinetic Monte Carlo computations. Closing the loop, this talk will show how the computations have enabled the development of analytical theories, which in turn have enabled the design of nanorods and innovation of metallic glue in ambient – at room temperature, in air, and under a small mechanical pressure of less than 10 MPa.
Background references:

• X. B. Niu, S. P. Stagon, Hanchen Huang*, J. K. Baldwin, and A. Misra, “Smallest Metallic Nanorods Using Physical Vapor Deposition”, Physical Review Letters 110，136102 (2013).
• L. G. Zhou and Hanchen Huang*, “Response Embedded Atom Method of Interatomic Potentials”, Physical Review B 87, 45431 (2013).
• Hanchen Huang*, G. H. Gilmer, and T. Diaz de la Rubia, “An Atomistic Simulator for Thin Film Deposition in Three Dimensions”, Journal of Applied Physics 84, 3636-3649 (1998).

Kurt Keville (MIT) — RISC-V and the Path To Exascale

RISC-V is a completely open ISA that is freely available to academia and industry suitable for direct native hardware implementation. It is an ISA that avoids "over-architecting" for a particular microarchitecture style (e.g., microcoded, in-order, decoupled, out-of-order) or implementation technology (e.g., full-custom, ASIC, FPGA), but which allows efficient implementation in any of these and then can be separated into a small base integer ISA, usable by itself as a base for customized accelerators or for educational purposes, and optional standard extensions, to support general-purpose software development. It supports extensive user-level ISA extensions and specialized variants, the 2008 IEEE-754 floating-point standard, 32-bit, 64-bit, and 128-bit address space variants for applications, operating system kernels, and hardware implementations.