Implementações sequencial e paralela de um novo algoritmo para a simulação de elementos e compostos magnéticos

Abstract

The magnetic phenomena are widely used in many devices, such as electronic, data storage and telecommunications devices. The understanding of this phenomenon is therefore of great interest to support the improvement and development of new technologies. To better understand the magnetic phenomena, it is essential to study interactions at nano scale. When magnetic atoms are brought together they interact magnetically, even with an external magnetic field, and can form structures at nanoscale. Special design computer programs can be developed to simulate this interaction. Such simulators can facilitate the study of magnetism in nanometer scale because they can provide detailed information about this phenomenon. Scientists may use a simulator to create and/or modify different physical properties of a magnetic system; visual and numerical data generated by the simulator can help to understand the physical processes associated with the magnetic phenomenon. However, there is a natural high complexity in the numerical solution of physical models. The interaction between spins occurs in a similar way to the classical n-body problem. The complexity of this problem is O(N2), where N is the number of spins or atoms in the system. In this work we propose a new algorithm that can substantially reduce the computational cost, and allows the simulation of a large number of spins. Besides, tools and environments for high-performance computing are used so that the costs of computation time may be further reduced.