Catálogo de publicaciones - libros

The KONWIHR project has developed a framework for the integration of simulation and visualization for large scale applications. This framework provides its own grid structure, the so called hierarchical hybrid grid, which is well suited for runtime efficient realization of multilevel algorithms. Furthermore, it offers flexible visualization functionality for both local and remote use on number crunchers and workstations. It is based on modern object-oriented software engineering techniques without compromising on performance issues.

The suitability of the Intel Pentium 4 Xeon, AMD Opteron, Intel Itanium2, IBM Power4 and Intel Pentium 4 architectures for application of the quantum chemical HF, B3LYP, MP2, CCSD and CCSD(T) methods is investigated, and conclusions on their relative performance are drawn.

We now allow ourselves the luxury of using basic complex analysis. In particular, we assume that the reader is familiar with contour integration and the residue theorem. We may view the residue theorem as yet another result that intimately connects the continuous and the discrete: it transforms a continuous integral into a discrete sum of residues.

Multi-beam laser welding is an advanced welding technique which can successfully prevent hot cracking, cf. [3], [4]. In order to guarantee that this technique prevents the initiation of hot cracks in the solid-liquid region, it is important to choose the positions, sizes, and powers of the additional heat sources suitably, e.g. optimally if an appropriate objective function can be established. In case of inappropriate choices for these parameters, hot cracking can even be enhanced. Until now these quantities are generally chosen by trial and error. This paper aims towards the simulation and optimization of multi-beam laser welding in order to demonstrate the potential of numerical optimization for the further improvement of this welding technique.

For this purpose a constrained nonlinear programming problem is formulated which provides a solution for the hot cracking problem by minimizing the accumulated transverse strain, i.e. the opening displacement, in the solid-liquid region. This approach is based on the so-called strip expansion technique, cf. [6]. For the objective function investigated in this paper it is sufficient to take into account a stationary temperature field in a moving reference frame. It is described by a partial differential equation for which it is possible to find a semi-analytical solution in terms of Bessel functions. Their computation is very time consuming and should be performed in parallel. If an optimization of the process is desired the amount of computation increases even more. This is due to the fact that, in addition to the solution of the partial differential equation, certain sensitivities must be computed in each loop of the optimization iteration, i.e., partial derivatives of the simulation output with respect to the optimization parameters.

We demonstrate the use of expression templates for the development of a numerical PDE software library for high performance computers. We discuss the library design and show that expression templates on performance tuned data structures achieve both a user-friendly interface and efficient runtime performance. Our performance results on various architectures including the Hitachi SR8000 illustrate that modern programming techniques like expression templates are well suited for large scale parallel computers.

Numerical simulations have become an important tool to predict the evolution of groundwater and subsurface contamination by organic compounds. Due to the observation of sharp interfaces on which biochemical degradation of the contaminants takes place, a reliable and accurate prognosis requires high spatial resolution of the considered subsurface domain and/or the application of higher order discretization methods. A successfull realization of these simulations can only be provided by parallel computer systems with high computing power and memory availability. In this article, our concepts and simulation results of “real world” scenarios of organic contamination are presented.

Performance analysis of parallel applications can be a time-consuming and daunting task, as it requires detailed a understanding of the interactions of the system’s components. We present the design and the prototypical implementation of a system for the automation of the performance analysis process developed within the Peridot project. Our system is based on the notion of cooperating agents that detect performance problems automatically at runtime and in a distributed fashion, avoiding several problems of classical performance analysis techniques such as overwhelmingly large trace files.

In this paper, we study numerically the renormalization of the electron-spin (el-sp) interaction or vertex due to Coulomb correlations in a two-dimensional one-band Hubbard model with spin-fluctuation momentum transfer = (π, π). Our simulations are based on a new numerically exact technique to extract the vertex, which is especially important for the physically relevant case, i.e., strong correlations, which cannot be controlled perturbatively. We find that the renormalized el-sp vertex decreases quite generally with increasing doping from the underdoped to the over-doped region. In the underdoped region, the corresponding effective pairing interaction increases strongly with lowering temperature in the weak- to intermediate-correlation regime. In contrast to this, it depends weakly on temperature in the strong-correlation regime. This behavior in the physically relevant strong-correlation case is due to a near cancellation between the temperature-driven enhancement of the spin susceptibility χ and the reduction of the el-sp interaction vertex. Thus, the spin-mediated d-wave attraction, which is peaked in weak coupling due to χ, is strongly reduced due to the el-sp vertex corrections for strong correlations.

Enzymes play a key role in research of the pharmaceutical industry because they represent targets for the design of new drugs. Therefore, the determination of the mode of action of enzymes is one of the great challenges of modern chemistry and an important task in . The situation is aggravated by the fact that the number of enzymes with known three-dimensional structure is small compared to the number of pharmaceutically relevant enzymes. Therefore, approaches for searching for a new depend on the information available about the protein structure and the ligands binding to a particular target. In this article we present a methodology based on a ligand-based approach. It can also be employed if the three-dimensional structure of the target of interest is not known. The structures of a set of molecules are superimposed based on a of a (GA) to evaluate their . This is an important step in the identification of a for molecules that bind to the same receptor. With this method it is possible to determine a complementary map of the receptor binding pocket.

Metal foams are interesting as lightweight materials that have an excellent combination of mechanical, thermal, and acoustic properties. However, the production process is currently not fully understood. Therefore, the goal of the FreeWiHR project is the development and high performance implementation of a model for simulating the formation process of metal foams based on the lattice Boltzmann method.