Session V6: The Impact of Large Scale Computing on Research in Physics

Challenges for Large Scale Simulations

With computational approaches becoming ubiquitous the growing impact of large scale computing on research influences both theoretical and experimental work. I will review a few examples in condensed matter physics and quantum optics, including the impact of computer simulations in the search for supersolidity, thermometry in ultracold quantum gases, and the challenging search for novel phases in strongly correlated electron systems. While only a decade ago such simulations needed the fastest supercomputers, many simulations can now be performed on small workstation clusters or even a laptop: what was previously restricted to a few experts can now potentially be used by many. Only part of the gain in computational capabilities is due to Moore's law and improvement in hardware. Equally impressive is the performance gain due to new algorithms - as I will illustrate using some recently developed algorithms. At the same time modern peta-scale supercomputers offer unprecedented computational power and allow us to tackle new problems and address questions that were impossible to solve numerically only a few years ago. While there is a roadmap for future hardware developments to exascale and beyond, the main challenges are on the algorithmic and software infrastructure side. Among the problems that face the computational physicist are: the development of new algorithms that scale to thousands of cores and beyond, a software infrastructure that lifts code development to a higher level and speeds up the development of new simulation programs for large scale computing machines, tools to analyze the large volume of data obtained from such simulations, and as an emerging field provenance-aware software that aims for reproducibility of the complete computational workflow from model parameters to the final figures. Interdisciplinary collaborations and collective efforts will be required, in contrast to the cottage-industry culture currently present in many areas of computational condensed matter physics.   

Unraveling the Supernova - Gamma-Ray Burst Mystery

Gamma-Ray Bursts are, simply put, the brightest explosions in the universe. Core-collapse Supernovae are the most energetic events in the modern universe. Observations show that both originate from massive stars, but the details of their central engines are essentially unknown. This remains one of the central challenges for astrophysics and cosmology today. Numerical calculations to address this challenge have been performed for decades, but have so far been unable to solve this problem. Comparison of current results indicates that large scale computing will be required to make further progress: Large scale (world-wide) collaborations to pool expertise from different fields of physics into a single, comprehensive code infrastructure, and large scale (at least petascale) calculations to allow accuracy and fidelity that are presently not yet reachable. I will first give brief overview over the (believed) necessary ingredients for a successful simulation of Gamma-Ray Bursts, and will then describe our approach towards addressing this challenge. I will outline the software infrastructure on which be base our research, and comment on the relation between the code infrastructure and the style of collaboration both within and outside of our group.   

Petascale Frontiers of Atomistic Materials Simulations

Large-scale classical molecular dynamics simulations with $10^6$ to $10^{12}$ atoms are providing unprecedented insight into material deformation processes under high strain-rate mechanical loading, and are increasingly being utilized to guide and interpret ultrafast {\it in situ} diffraction and microscopy experiments. I will describe our recent algorithm redesign motivated by the evolution towards hybrid multicore architechtures employing graphical processing unit (GPU) or Cell co-processors, specifically the heterogeneous petaflop Roadrunner platform at Los Alamos National Laboratory. I will then present the resulting performance achieved, as well as our approach to the challenges of terabyte data analysis and visualization on Roadrunner and Blue Gene systems. Finally, results from our initial scientific applications on Roadrunner to understand the response of copper single- and poly-crystals to shock compression and release, including material ejection and spall failure, will be described.   

Lattice QCD from algorithms to hardware

We shall discuss progress that has been made in studying the quantum field theory of the strong interaction (Quantum Chromodynamics, QCD). These range from improved discretization schemes that preserve chiral symmetry through advances in Monte Carlo algorithms to optimized hardware design.   

Ab initio calculations of correlated electron dynamics in ultrashort pulses

The availability of ultrashort and intense light pulses on the femtosecond and attosecond timescale promises to allow to directly probe and control electron dynamics on their natural timescale. A crucial ingredient to understanding the dynamics in many-electron systems is the influence of electron correlation, induced by the interelectronic repulsion. In order to study electron correlation in ultrafast processes, we have implemented an ab initio simulation of the two-electron dynamics in helium atoms. We solve the time-dependent Schr\"odinger equation in its full dimensionality, with one temporal and five spatial degrees of freedom in linearly polarized laser fields. In our computational approach, the wave function is represented through a combination of time-dependent close coupling with the finite element discrete variable representation, while time propagation is performed using an Arnoldi-Lanczos approximation with adaptive step size. This approach is optimized to allow for efficient parallelization of the program and has been shown to scale linearly using up to 1800 processor cores for typical problem sizes. This has allowed us to perform highly accurate and well- converged computations for the interaction of ultrashort laser pulses with He. I will present some recent results on using attosecond and femtosecond pulses to probe and control the temporal structure of the ionization process. This work was performed in collaboration with Stefan~Nagele, Renate~Pazourek, Andreas~Kaltenb\"ack, Emil~Persson, Barry~I.~Schneider, Lee~A.~Collins, and Joachim~Burgd\"orfer.