9 months, starting as soon as possible considering the time needed by the person retained to get all the administrative papers necessary to work in France.
Financial conditions:
The credits obtained allow us to warrant a gross salary of 1839 € per month, which in reality means 1508 € per month transferred on the bank account.
Requested profile:
The applicant should preferentially have with experience in numerical analysis concerning implementation of the fast wavelet transform in fortran. Experience of parallelization and/or knowledge of Green’s function techniques and/or knowledge of the scattering matrix formalism in optics would be a plus, but are not required. In any case, the person retained must have completed his/her PHD since less than 2 years and NOT be of French nationality!
Therefore, the applicants should enclose in their application, considerations allowing assessing the time they would need to get the authorization to work in France.
Summary of the project:
Since more than 10 years, we have been carrying theoretical studies concerning carbon nanotubes, first in the framework of the French CNRS research group GDR n°1752, then in the European research group GDR-E n°2756). In parallel, we got interested in applying the fast wavelet transform to speed up our simulations which require resolution of integral equations by Green’s function techniques. The limiting step is most of the time the resolution of large and dense systems of linear equations.
The idea was to consider the matrix of the system just as an image and to use the fast wavelet transform and a thresholding process to sparsify it. The resulting sparse system is solved by means of specialized techniques and the solution inverse transformed. This was already part of the subject of Rachel Langlet’s PHD thesis co-supervised by Martin Meyer (Mathematics lab of Besançon) and Michel Devel (UTINAM institute of Besançon), with a great help from Yiwei LI (Xidian University, China) during and after his one-month visit in Besançon in July 2004.
Thanks to these efforts we now have several pieces of fortran 90 codes, which implement this algorithm for real matrices (of arbitrary size, not necessarily a power of 2), with various levels of refinement. First experiments indicate that we can indeed use far less memory than with the traditional LAPACK based code and also gain in computation speed. The goal of this post-doc would be to apply the same technique with complex coefficient matrices and possibly parallelize it. This would allow us to simulate interaction with light of carbon nanotubes or carbonaceous soot, with sizes closer to the actual experimental size of these objects than what we can presently do with the LAPACK based code. Cooperation with experimentalists in these fields is already under way.
Contact person :
Please send a CV and e-mail and phone numbers of some past collaborators to:
Dr. Michel DEVEL
Institut UTINAM,
UMR CNRS 6213 16 route de GRAY,
25030 Besançon CEDEX, FRANCE
E-mail : Michel.devel@univ-fcomte.fr
Website : http://www.utinam.cnrs.fr
0 komentar:
Posting Komentar