Earth Sciences & Env.
Engineering & Tech.
Information & Comm. Tech.
Life Sciences & Biotech
Use this search facility to find out more about the profile of our HPC-Europa2 visitors, the type of work they have been doing, and their project achievements.
According to a recent expert review on Density Functional Theory (DFT) [[i]], one of the current challenges is "the need to improve the description of ... dispersion/van der Waals interactions". In recent years a number of approaches addressing this issue have appeared. One of promising methods (BH-DFT-D) have been developed by my host, Dr. Alisa Krishtal and coworkers [[ii]]. This approach is based on distributed atomic polarizabilities obtained from perturbed densities. First, molecular polarizabilities are calculated within the finite-field approach. Subsequently, atomic components of the polarizability tensors are obtained using atomic weighting functions defined according to Hirshfeld [[iii]] or iterative Hirshfeld partitioning [[iv]]. From the atomic polarizabilities, intermolecular interaction energies are obtained.
There are indications in the literature [[v]] that most common density functionals overestimate molecular polarizabilities and this deficiency can be improved by applying a density functional with a correct asymptotic behavior. This suggests that the atomic polarizabilities and, eventually, intermolecular interaction energies can be also improved if an asymptotically corrected functional is used. Therefore, the central idea of the current project is to implement and to test asymptotically corrected functionals within the BH-DFT-D method.
[i]. Cohen A. J., Mori-Sánchez P., Yang W., Chem. Rev. 2012, 112, 289–320
[ii]. Krishtal A., Vanommeslaeghe K., Olasz A., Veszprémi T., van Alsenoy C., Geerlings P., J. Chem. Phys. 2009, 130, 174101
[iii]. Hirshfeld F. L. Theoret. Chim. Acta, 1977, 44, 129--139.
[iv]. Bultinck P., van Alsenoy C., Ayers P. W., Carbó-Dorca R., J. Chem. Phys. 2007, 126, 144111.
[v]. Vasiliev I., Chelikowski J. R., Phys. Rev. A, 2010, 82, 012502.
We implemented two different asymptotically corrected functionals. The first one is the van Leeuwen–Baerends correlation functional (LB94)
[[i]]. However, it has a problematic performance in the high-density region. Therefore, Casida and Salahub proposed a method that combines a shifted LDA (or GGA) potential in the core region with the LB94 potential in the low-density region [[ii], [iii]].
The implementation of LB94 in the BRABO code was relatively straightforward. BRABO already has subroutines necessary to calculate the electron density and gradient. We wrote a subroutine implementing the LB94 functional in a close analogy to existing subroutines in BRABO implementing other exchange-correlation functionals. The subroutine was tested by comparison with MOLPRO results.
The implementation of the Casida–Salahub functional is a little more tricky, since it depends, in addition, on the ionization potential IP. In our implementation, the ionization potential was obtained from a Hartree–Fock calculation according to the Koopmans theorem. The values of IP and are input for a Casida–Salahub calculation. Therefore, a Casida–Salahub calculation consists essentially of three different ones: a Hartree–Fock run (to get the IP), an LDA run (to obtain ), and the proper Casida–Salahub run. We implemented the above equation in the BRABO program with an energy shift as an input. The preceding Hartree–Fock and LDA calculation are performed in separate BRABO runs, but the IP and were extracted automatically from the output by a shell script. The same script also creates an input for the final Casida–Salahub calculation.
In order to obtain all the necessary molecular polarizabilities within the finite-field approach, a large series of BRABO runs is necessary, since about 40 various perturbations (±x, ±y, ±z, ±xx, ±xy, ±xz, ..., ±xxx, ±xxy, ±xyz, ..., ±zzz) must be run in order to obtain the perturbed density matrices. These calculations are also governed by a script. We refrained from parallelization of BRABO, since about 40 independent calculations, which can be run simultaneously, are needed anyway.
The atomic polarizabilities are compute from the perturbed density matrices by the STOCK program. An MPI-parallelized version developed by Dr. Krishtal already existed. Upon a slight modification it was successfully compiled on the Nehalem computer.
For a number of reasons, calculations took much more time and effort, and I eventually managed to calculate only a part of what had been planned. The calculation were performed for a selection of monomers from the S22 set [[iv]] with the 6-311+G*. The BRABO jobs (perturbed-density evaluation) were carried out using Hartree–Fock, B3LYP, and Casida–Salahub functionals, while the final STOCK jobs (polarizability evaluation ) have been so far computed only for B3LYP.
By now, the polarizabilities were computed for the following molecules: C2H2 (several geometries), C2H4 (several geometries), C6H6 (several geometries), (CH3)2C(H)(C2H5), N(CH2)4. For other monomers from the S22 set, a part of calculations is finished.
[i]. van Leeuwen R., Baerends E. J., Phys. Rev. A, 1994, 49, 2421--2431
[ii]. Casida M. E., Salahub D. R., J. Chem. Phys., 2000, 113, 8918.
[iii]. Casida M. E., Casida K. C., Salahub D. R., Int. J. Quantum Chem., 1998, 70, 933
[iv]. Jurečka P., Šponer J., Černý J., Hobza P. Phys.Chem.Chem.Phys 2006, 8, 1985
In this Project we aim to analyze the aromaticity and the different distortive nature of s and p electrons employing the energy decomposition analysis (EDA) technique as used in previous works by Prof. Bickelhaupt’s method for the study of the aromaticity of classical organic molecules. This is a very promising molecular orbital (MO) model of aromaticity that allows understanding the reasons behind the fact that aromatic molecules such as benzene have a regular structure with delocalized bonds, while antiaromatic molecules like cyclobutadiene prefer distorted geometries with localized bonds. In none of the systems analyzed by Prof. Bickelhaupt and coworkers does the p-electron system favour a symmetric, delocalized ring. It is the sigma-electron system that forces the system to acquire its regular structure.
This Project is carried out in cooperation with Prof. Dr. F. Matthias Bickelhaupt of the Vrije Universiteit Amsterdam. During the realization of the project the computational resources of SARA were extensively used. All calculations were performed on Huygens installation.
In this Project we were interested in the analysis of s and p electron distortive character in the case of all-metal aromatic clusters, and, in particular, in the quintessential Al42- aromatic species. Just for comparison we added the isoelectronic B42- and Ga42- species. In addition, we compared the results obtained with those derived for the C4H4, C4H42+, and C4H42- species. Finally, the Al44- species having ambiguous aromatic character was also included in the study. The ADF software package based on Density Functional Theory (DFT) has been used for calculations. Calculations were performed using BP86 functional in conjunction with a triple-zeta quality Slater-type basis set augmented with two sets of polarization functions (TZ2P). For the EDA analysis we have considered two diagonal X2 fragments of the X4 clusters analyzed and we have changed the angle between the two fragments from 90 to 100º. In this way we change from D4h to D2h molecular symmetries and we can discuss which energy terms increase or decrease during this process. Preliminary results of our calculations for Al42- show that the orbital interaction term of the EDA is clearly distortive while the Pauli repulsion prefers the symmetric structure. Since there is only Pauli repulsion among the s electrons it is possible to sum the s orbital interaction term and the Pauli repulsion term to arrive at the conclusion that the s electrons favour the symmetric structure in Al42- while the p electrons are slightly distortive but much less than in C4H4 for instance. Similar results were obtained for the rest of the systems.
The objectives of my visit were the following:
- Testing the parallel 2D Multisubband Monte Carlo simulator in the supercomputing resources of the EPCC.- Comparing the execution time results of the EPCC resources for the nanoelectronic simulator with the previous results obtained in Spain: in our local cluster, in the resources of the Supercomputing Center of Galicia (CESGA) and in the Spanish National Grid Initiative (NGI-es).- Developing a job submission and monitoring system based on Rapid.- Consulting the current techniques employed in data analysis and visualisation in grid computing environments.- Establishing academic links that enable us to exchange the obtained knowledge and experience in the field of grid computing.
The achivements of my visit were as follows:- I have parallelised a 2D Multisubband Monte Carlo simulator in Spain for my PhD research. During my visit I could evaluate the parallel performance of this code in the AMD processors of the Ness parallel machine.- These execution time results of the parallel 2D Multisubband simulator run on Ness were compared with the execution results obtained in the Supercomputing Centre of Galicia (CESGA) using a Itanium2 Processor and with other execution times obtained with a Xeon Processor that belongs to a cluster of our department. The executed code was statically compiled and we obtained speed up differences up to 15% using 8 cores among the different processors. These results provide an important information in order to submit this parallel application to heterogeneous resources such as the Grid infrastructure.- I have designed a database to manage the simulation results. This was an important objective of my visit because my research group does not have a data management system and we are working on the study of nanoelectronic fluctuations in MOSFET transistors which require hundreds of thousands of simulations. During my visit my host introduces me to a person with experience on scientific databases who helps me with the design. Therefore, I had to learn some new concepts about database design and entity relationship (ER) diagrams. Finally, the design is ready to be implemented in my group.- I have also started the design of a job submission portal using Rapid, thinking about how to connect it with the database. I have installed Rapid on an Ubuntu virtual Machine and the LifeRay Web portal to deploy the Rapid portlets. When I tested several examples, I stopped working on the job submission portal to prioritize the database design due to time limitations.- This visit enabled me to meet people working for the data intensive research group and some of their research interests. Although, we work on different research areas, I think my research group has some software requirements related to data management and other common interests, such as Cloud and Grid architectures, that could be common points to stablish academic links in future.