Muroran Institute of Technology, Department of Electrical and Electronic Engineering, Muroran
In addition to standard high performance computing technologies such as supercomputers and grid computers, a method of dedicated computers have been attempted to construct portable high performance computing environments in the vicinity of office PC. The method of dedicated computers have also been adopted into electromagnetic field simulations, which are mainly in a linear algebra equation solver for general electromagnetic field analysis and the FDTD solver for microwave simulations. In this paper, attempts of FDTD/FIT dedicated computer (FDTD/FIT machine) are introduced*. The basic scheme of the FDTD/FIT method itself is very simple and suitable for implementation as hardware circuits. In addition, it is also essential to realize many other functions such as imposing of boundary conditions, treatment of non-uniform materials, power input, etc. Moreover, to fully bring out the advantage of the method of dedicated computer, the computer architecture should be designed to achieve efficient computing of all of FDTD/FIT scheme including the boundary condition setting, etc. Especially various efforts of minimization of memory access overhead are discussed in this paper.
UCLA, Los Angeles, California
New emerging multi-core technologies can achieve high performance, but algorithms often need to be redesigned to make effective use of these processors. We will describe a new approach to Particle-in-Cell (PIC) codes and discuss its application to Graphical Processing Units (GPUs). We will conclude with lessons learned that can be applied to other problems. Some of these lessons will be familiar to those who have programmed vector processors in the past, others will be new.
LLNL, Livermore, California
PPPL, Princeton, New Jersey
LBNL, Berkeley, California
The near-term mission of the Heavy Ion Fusion Science Virtual National Laboratory (a collaboration of LBNL, LLNL, and PPPL) is to study "warm dense matter" at ~1 eV heated by ion beams; a longer-term topic is ion-driven target physics for inertial fusion energy. Beam bunch compression factors exceeding 50x have been achieved on the Neutralized Drift Compression Experiment (NDCX) at LBNL, enabling rapid target heating; however, to meet our goals an improved platform, NDCX-II, is required. Using refurbished induction cells from the decommissioned Advanced Test Accelerator at LLNL, NDCX-II will compress a ~500 ns pulse of Li+ ions to ~1 ns while accelerating it to 3-4 MeV (a spatial compression of 100-150x) over ~15 m. Non-relativistic ions exhibit complex dynamics; the beam manipulations in NDCX-II are actually enabled by strong longitudinal space charge forces. We are using analysis, an interactive 1D PIC code (ASP) with optimizing capabilities and a centroid-offset model, and both (r,z) and 3D Warp-code simulations, to develop the NDCX-II accelerator. Both Warp and LSP are used for plasma neutralization studies. This talk describes the methods used and the resulting physics design.
LLNL, Livermore, California
The Finite Difference Time Domain (FDTD) method and the related Time Domain Finite Element Method (TDFEM) are routinely used for simulation of RF and microwave structures. In traditional FDTD and TDFEM algorithms the electric field E is associated with the mesh edges, and the magnetic flux density B is associated with mesh faces. It can be shown that when using this traditional discretization , projection of an arbitrary current density J(x,t) onto the computational mesh can be problematic. We developed and tested a new discretization that uses electric flux density D and magnetic field H as the fundamental quantities, with the D-field on mesh faces and the H-field on mesh edges. The electric current density J is associated with mesh faces, and charge is associated with mesh elements. When combined with the Particle In Cell (PIC) approach of representing J(x,t) by discrete macroparticles that transport through the mesh, the resulting algorithm conserves charge in the discrete sense, exactly, independent of the mesh resolution h. This new algorithm has been applied to unstructured mesh simulations of charged particle transport in laser target chambers with great success.
Northern Illinois University, DeKalb, Illinois
When faced with the challenge of the design optimization of a charged particle beam system involving beam-material interactions, a framework is needed that seamlessly integrate the following tasks: 1) high order accurate and efficient beam optics, 2) a suite of codes that model the atomic and nuclear interactions between the beam and matter, and 3) the option to run many different optimization strategies at the code language level with a variety of user-defined objectives. To this end, we developed a framework in COSY Infinity with these characteristics and which can be run in two modes: map mode and a hybrid map-Monte Carlo mode. The code, its applications to the FRIB, and plans involving large-scale computing will be presented.
Northern Illinois University, DeKalb, Illinois
ANL, Argonne
Most charged particle beams under realistic conditions have Gaussian density distributions in phase space, or can be easily made so. However, for several practical applications, beams with uniform distributions in physical space are advantageous or even required. Liouville’s theorem and the symplectic nature of beam’s dynamic evolution pose constraints on the feasible transformational properties of the density distribution functions. Differential Algebraic methods offer an elegant way to investigate the underlying freedom involving these beam manipulations. Here, we explore the theory, necessary and sufficient conditions, and practicality of the uniformization of Gaussian beams from a rather generic point of view.
ANL, Argonne
Northern Illinois University, DeKalb, Illinois
While COSY INFINITY provides powerful DA methods for the simulation of fragment separator beam dynamics, the master version of COSY does not currently take into account beam-material interactions. These interactions are key for accurately simulating the dynamics from heavy ion fragmentation and fission. In order to model the interaction with materials such as the target or absorber, much code development was needed. There were four auxiliary codes implemented in COSY for the simulation of beam-material interactions. These include EPAX for returning the cross sections of isotopes produced by fragmentation and MCNPX for the cross sections of isotopes produced by the fission and fragmentation of a 238U beam. ATIMA is implemented to calculate energy loss and energy and angular straggling. GLOBAL returns the charge state. The extended version can be run in map mode or hybrid map-Monte Carlo mode, providing an integrated beam dynamics-nuclear processes design optimization and simulation framework that is efficient and accurate. The code, its applications, and plans for large-scale computational runs for optimization of separation purity of rare isotopes at FRIB will be presented.