日本原子力研究開発機構 (JAEA) / 量子ビーム応用研究部門
放射光科学研究ユニット / 放射光量子シミュレーショングループ
|Date and time:||17th July (Thu.) 16:15〜|
|Place:||SPring-8, "HOUKOUKAN" seminar room|
|Title:||Large-scale DFT calculations using a linear-scaling code CONQUEST|
|講演者:||宮崎 剛 氏 (物材機構)|
|Speaker:||Dr. Tsuyoshi Miyazaki (NIMS)|
We report our recent progress and applications of our linear-scaling (or O(N)) DFT code CONQUEST. The code is efficient on massively parallel computers and has an ability to treat systems containing more than ten thousands of atoms. CONQUEST minimises DFT total energy with respect to the Kohn-Sham density matrix and it can be run at different levels of precision, ranging from empirical tight-binding, through ab initio tight-binding, to full plane-wave precision. In the seminar, the theory behind the code and the way in which this is achieved will be outlined. As one of the research examples by CONQUEST, we show our theoretical study on the nanostructured Ge 3D islands on Si(001). We demonstrate that it is now possible to perform DFT calculations on such large and scientifically important systems. We have succeeded in employing DFT structure optimization on the systems containing more than 20,000 atoms. Recently, we also use the code for the study on biological systems. To check the accuracy of our theoretical methods such as localised orbital or density matrix minimisation methods, we have performed DFT calculations on single DNA bases, DNA base pairs and hydrated DNA systems. We will show the errors from the approximation in the O(N) method are extremely small.
CONQUEST is a collaborative code, developed by teams in University College London (UCL) and NIMS. This work has been done with D. R. Bowler (UCL), M. J. Gillan (UCL), T. Otsuka (NIMS), T. Ohno(NIMS), M. Todorovic (NIMS), A. S. Torralba (NIMS) and V. Bradzova (UCL).