Section: New Results
The study of hybrid materials based on a coupling between molecular dynamics (MD) and quantum mechanism (QM) simulation has been conducted in collaboration with IPREM (Pau) within the ANR CIS 2007 NOSSI (ended December 2011). These simulations are complex and costly and may involve several length scales, quantum effects, components of different kinds (mineral-organic, hydro-philic and -phobic parts). Our goal was to compute dynamical properties of hybrid materials like optical spectra. The computation of optical spectra of molecules and solids is the most consuming time in such coupling. This requires new methods designed for predicting excited states and new algorithms for implementing them. Several tracks have been investigated in the project and new results obtained as described bellow.
Some new improvements in our TD-DFT code have been introduced. Our method is based on the LCAO method for densities and excited states that computes electronic excitation spectra. We have worked in two directions:
As the method introduces a regularization parameter to obtain regularized spectra we have used it to build better algorithms. In particular, we have developed a new hierarchical algorithm that builds a well adapted frequency distribution to better capture the biggest peaks (strongest oscillator strengths) in the spectrum. Moreover, a nonlinear fit method was added and used to compute the transitions and the oscillator strengths of the spectrum.
In our algorithm, we used a coarse grain paradigm to parallelize the spectrum computation. This approach leads to a memory bottleneck for large systems. In that respect, we have explored a new parallel approach based on a fine grain paradigm (matrix-vector parallelization) to better exploit the manycore achitecture of the emerging computers.
Finally, we have improved the packaging of the code to prepare a public release of the code. Our TD-DFT code will be soon available on request.
QM/MM algorithm. For structure studies or dynamical properties, we have coupled QM model based on pseudo-potentials (SIESTA code) with dynamic molecular (DL-POLY code). Therefore we have developed a new algorithm to avoid accounting twice for the forces and the quantum electric field in the molecular model. All algorithms involved in the coupling have been introduced both in SIESTA and in DL-POLY codes. The following new developpements needed by the coupling have been introduced in the SIESTA code:
We have implemented a fast evaluation of the molecular electrostatic field on the quantum grid.
We have introduced a non periodic Poisson solver based on the parallel linear Hypre solver. This solver allows us to use computation domains as small as possible.
We have implemented the ElectroStatic Potential (ESP) fit method to obtain more physical point charges than those given by SIESTA with the Mulliken method. These point charges are used by the MM codes to compute electrostatic forces.
Thanks to all our develpements introduced in SIESTA a collaboration with the SIESTA research team has started. This enables us to have acces to their private svn like repository. Preliminary resuls on a water dimer and a water box systems show good agreement with other methods developed in SIESTA and DL-POLY teams.
All these results were presented in the final international NOSSI workshop in Biarritz on december.
We have started in the context of the OPTIDIS ANR to work on dislocation simulations. The main characteristic of these simulations is that they are highly dynamical. This year, we have started the study of the state of the art on this topic in two directions. The first direction concerns the study of the algorithms used in such simulations and how we can efficiently parallize them on manycore clusters. In the second one for isotropic materials, we are investigating how to adapt our fast multipole method to compute constraints and then forces in this kind of simulations.