Umberto Ravaioli

1
Nanoscale Simulation and High-Performance
Computing

• Umberto Ravaioli
• Dept. of Electrical and Computer Engineering
• and National Center for Supercomputing
  Applications
• University of Illinois at Urban-Champaign

2
Objectives

• Starting from traditional MOS systems, develop
  physical simulation
• tools for nanoscale structures at the limit of
  scalability.
• Develop new device concepts based on 3-D
  topology, looking
• at interaction and cooperation between adjacent
  elements.
• Integrate hierarchically semi-classical and
  quantum simulation in
• nano-structure application and look forward at
  links with molecular
• electronics for hybrid solid-state/molecular
  structures.
• Leverage NCSA infrastructure to establish
  efficient applications on
• large scale cluster computing environments.

3
Hierarchical Simulation Approach
Provide links between different available levels
of simulation for flexibility and efficiency.

• Quantum-level simulation
• (Ballistic and scattering regimes)
• Schrödinger/Poisson
• Greens function approaches

Quantum corrected semi-classical simulation
Semi-classical simulation Particle Monte Carlo
methods
4
From Standard MOS to Quantum Nanostructures
Evolutionary approach component integrated in the
overall program on nano-structure development
Hybrid structures Docking for molecular and
bio-molecular electronics
Coupled 3-D structures Multi-functional Multi-term
inal
3-D nanoscale devices based on MOS technology
Ultra-scaled MOS transistors
5
Building from Existing Expertise
Strong program in Monte Carlo simulation

• 3-D parallel particle Monte Carlo simulation
  has been developed
• for deep sub-micron MOSFET structures.
• Approaches for quantum correction and detailed
  charge-charge
• interaction are actively investigated.

Quantum level simulation
Numerically advanced quantum solvers for
coupled Schrödinger- Poisson equation,
tight-binding Greens function methods,
transient simulation and absorbing boundary
conditions.
Supercomputing
Strong working experience over 15 years over
virtually all major supercomputing platforms,
with present focus on clusters.
Other advanced simulation interests
Detailed simulation of transport in ionic
channels micro- and nano-electro-mechanical
systems.
6
An Example of Ongoing Work 1
3-D Multiple-Gate MOS structure
Directional coupling properties of two adjacent
channels are studied as a function of gate
voltage. Properties of the structure are also
controlled by varying the length and (LI) and
width (WI) of the quantum coupling region. One
goal is to build the simulation tools necessary
to understand and design novel functionality for
complete 3-D structures.
7
An Example of Ongoing Work 2
A result example
8
Large-Scale Computing The NCSA Advantage
Prof. Ravaioli is serving as the leader of the
Application Technology Team on
Nanomaterials/Devices of the National
Computational Science Alliance (NCSA). He has
also served with NCSA, for the last 5 years, as
Senior Academic lead for Computational
Electronics and Nanoelectronics in the PET
program of the DoD Major Shared Resource Centers.
Super-Cluster Platforms
NCSA is pursuing an aggressive upgrade path to
realize large scale cluster computational
environments, leading the way in developing
public domain system software and acting as
national resource in this area.
Multidisciplinary Environment
NCSA application groups in many computational
disciplines interact closely and interface with
Enabling Technology Teams (from compiler issues
to advanced visualization and remote
collaboration).
9
Large-Scale Clusters The Future of
Supercomputing
The trend towards large commodity clusters for
computation and the coming of age of Linux is
shaping the present and the future of
super- computing. The Illinois group has been
pursuing vigorously the development of
cluster-based parallel computational
applications. The expertise will be directly
applied to the projects involved in this
effort. Emphasis will be on standardized
approaches, so that other groups involved may
more readily benefit from collaboration and
transfer technology. MPI parallelization
procedures will be approach of choice, due to
the wide availability and maturity of the related
system software. We believe that this collective
effort will demonstrate how a broad research
program in nanoelectronics can be integrated in
high performance supercomputing cluster
environments.
10
Opportunities for Collaboration

• There are unique opportunities for collaboration
  that the Illinois group intends to pursue.
• Existing ties and commonality of interests in
  both nano-science and high-performance computing
  make the relationship with the Purdue group
  particularly strong. Due to the rich range of
  activities existing at the University of
  Illinois, there are also interests in pursuing
  interaction with all the groups involved with
  molecular electronics, to pursue possible MEMS
  and NEMS sensing and computing applications.
• Remote collaboration of a distributed group
  presents also challenges. Our group has been
  working on advanced collaborative visualization
  systems together with NCSA. Compatibly with
  available resources, we will make an effort to
  prototype and deploy collaborative technology to
  facilitate sharing of complex data visualization
  interactively among collaborating groups.