Sandia National Labs FY20 LDRD Annual Report

FY20 ANNUAL REPORT

Complex fluid partitioning revealed in surface and pore environments. Understanding the interaction of complex fluids in surface and pore environments is critical for technical advancements in energy-water applications such as resource recovery and water purification. This project combined experimental measurements and molecular modeling to investigate fundamental properties of these fluid interactions with mineral surfaces to provide a more complete picture of surfactant interactions. This research focused on an anionic surfactant and a well-characterized mineral phase representative of oil- and gas-producing shale deposits with the results demonstrating that the presence of surfactant significantly affects the mineral-fluid interfacial structure. Molecular modeling results also reveal details of the surfactant structure at the interface, and how this structure varies with surfactant coverage and fluid composition. This project will

enable a broader Sandia capability to solve technical challenges critical to the country’s energy-water future. (PI: Jeffery Greathouse) Snapshots from molecular dynamics simulations showing the effect of cation on surfactant micelle structure at a negatively charged mineral surface.

A flexible, highly scalable, configurable neuromorphic architecture. Neuromorphic architectures are proving to be very attractive options for computing since they have, in the past decade, provided potential 100X-1000X efficiency gains over conventional Von Neumann architectures. Both neuromorphic analog and digital technologies provide low-power and configurable acceleration of challenging artificial intelligence (AI) algorithms. This project explored placing deep artificial neural networks onto a hybrid analog-digital neuromorphic architecture that amplifies the advantages of both high-density analog memory and spike-based digital communication while mitigating their limitations. This approach provides a possible avenue for maximizing the benefits of these emerging

complementary approaches to neuromorphic computing. It can further U.S. leadership in advanced computing technology by forging a path to widespread applicability of brain-inspired computing from embedded domains to large-scale simulation tasks. (PI: James Bradley Aimone) This project focused on techniques allowing hybrid analog-spiking artificial neural networks to get into the “sweet spot” of low computational costs with high performance.

43

LABORATORY DIRECTED RESEARCH & DEVELOPMENT