THIS WEEK: 24/25 Seminars on HPC

HG E 42, ETH Main Building

Thursday,  April 24
13:15 h  Prof. Giulia Galli, Department of Chemistry, UC Davis
Talk: Predictive computational science: successes, challenges and future
goals (Abstract below)

15:45 h  Prof. Romain Teyssier, Head Computational Astrophysics Group,
CEA Saclay
Talk: Recent advance in Computational Cosmology (Abstract below)


12:15 h  Prof. Thomas Schulthess, Oak Ridge National Laboratory
Talk: Computational studies of superconductivity in the cuprates:
anticipating the role of peta/exascale computing (Abstract below)

14:45 h  Prof. Edward Seidel, Director Center for Computation and
Technolgoy, Louisiana State University
Talk: From Black Holes to Black Oil:  Cyberinfrastructure-Enabled
Complex Problem Solving (Abstract below)


Abstract of Prof. Giulia Galli's talk:
Predictive computational science: successes, challenges and future goals 
 

In the past forty years, the use of scientific computing has become pervasive in all
disciplines: collection and interpretation of most experimental data is carried out using
computers, and numerical solutions of physical models with various degrees of
complexity and sophistication, are utilized in all fields of science. However, discovery
and scientific design by computation are revolutions still in the making. In particular, the
prediction of fundamental properties of materials and molecular systems from the
numerical solutions of the basic laws of quantum mechanics is in its infancy; some
formidable theoretical and computational challenges lay ahead of us. We describe recent
progress and successes obtained in predicting properties of matter by quantum
simulations, and discuss algorithmic challenges and their connection with the use of
evolving high-performance computing architectures.  We also discuss open issues related
to the validation of the approximate, first principles theories used in large scale
simulations, and the resulting complex interplay between computation and experiment. In
our discussion of applications, we focus on materials for energy applications, high
pressure physics and physical chemistry, in particular aqueous environments.



Abstract of Prof. Romain Teyssier's talk:
Recent advance in Computational Cosmology

Structure formation in the universe is considered to be a
well-established
theory based on gravitational instability and,  as initial conditions,
primordial
fluctuations observed with unprecedented accuracy on the cosmic
microwave
background. High performance computing has played a
fundamental role in our understanding of (non linear) gravitational
clustering.
Using large supercomputers and sophisticated N body algorithms, it is
now possible
to generate virtual universe as large as the actual observable universe.
Although dark matter dynamics dominates at large scale, baryons physics
is fundamental
at small scale. Galaxy formation is the current frontier in our
understanding of cosmic structure
formation. Recent state-of-the-art simulations of galaxy formation will
be presented, with
some new interesting ideas on how complex gas physics (MHD, turbulence)
may play a
fundamental role in the formation of our own galaxy, the Milky Way.


Abstract of Prof. Thomas Schulthess' talk:
Computational studies of superconductivity in the cuprates: anticipating
the role of peta/exascale computing

With their discovery of superconductivity in the cuprates some twenty
years ago, Bednorz and Muller set in motion new research directions in
transition metals oxide materials with many promising applications. But
the most promising of all, high temperature superconductivity in the
cuprates, has still fallen far short of expectations. The development of
new superconducting materials has been hampered by a lack of theoretical
guidance, and despite two decades of intensive research there is still
no agreement on the basic mechanisms of superconductivity in the
cuprates. The recent staggering advances in high-performance computing,
however, open new possibilities to investigate this problem. In this
talk, I will review recent progress made with quantum cluster
simulations of superconductivity in the Hubbard model. I will report on
insights we have gained from attempts to link these simulations with
first principles electronic structure calculations, and I will discuss
possible paths fo rward towards performing materials specific
simulations of superconductivity in the cuprates. Furthermore, I will
discuss recent experiences with full system runs on the newly upgraded
250 TFlop/s Cray XT4 at Oak Ridge National Laboratory, as well as
production runs that we have executed on clusters using Graphic
Processing Units (GPU) compute accelerators. These preliminary runs
allow us to realistically estimate the type of simulations that will be
possible on future peta- and exascale computer systems.


Abstract of Prof. Edward Seidel's talk:
From Black Holes to Black Oil:  Cyberinfrastructure-Enabled Complex
Problem Solving

Computing is revolutionizing our ability to solve complex problems in
all areas of research.  For example, detailed, multiscale  theoretical
simulations of the inspiral, merger, and ringdown of binary black hole
systems, impossible for decades until just a few year ago, are now
routinely performed and used in experimental searches for gravitational
waves on a daily basis.  Using recent breakthroughs in numerical
relativity as a leading example, I will show how modern
cyberinfrastructure---the comprehensive set of hardware, software, and
algorithmic tools and approaches to scientific enquiry---are
accelerating and transforming research, education, and increasingly
collaboration across and among all disciplines, whether they are
experimental, data-driven, theoretical, and/or computational.