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.
The main goal of our project is to maintain the level of performance of the suite, extending the pure message passing paradigms with mixed parallelism using MPI and OpenMP (where possible). In particular, we will start working on one of the most relevant algorithms of the suite: an ad-hoc 3D Fast Fourier Transform. We try 2 approaches to ibridize the code: an implicit approach linking native multi-thread library provided by third-part vendors and an explicit one using OpenMP directly into the code. If the experimental hybrid/mixed paradigm with the FFT is successful, we will try to extend it to other relevant and heavy computational cores of the suite, particularly in the iterative diagonalization of Hamiltonian, which is the second algorithm into suite in order of importance.
We implemented two test programs for a detailed performance analysis of the parallel ad-hoc hybrid 3D FTT. Both programs share the same main modules of Quantum ESPRESSO and are able to initialize the FFT grid like a real simulation using a real input parameters. We devoted a large part of the time to design a test program that shares the same structure of the main modules of Quantum ESPRESSO in order to simplify the procedure of integrating the modifications into the main package. This point is very important because Quantum ESPRESSO has fixed data structures that cannot be changed. Both master-only and funneled approach have been developed. We have test the modified algorithm using different combination of MPI and threads number.
After the testing phase, we have transferred all the modifications into the main code of Quantum ESPRESSO and our improvements have been committed into the official repository. We have had the possibility to work on CRAY XT4 (HECToR) and IBM POWER5 (HPCx) clusters hosted by Edinburgh Parallel Computing Centre and, also, IBM BladeCenter LS21 (BCX) in CINECA. Now we are able to run Quantum ESPRESSO using MPI and OpenMP together.
The whole research project is focused on designing new ceramic materials for protonic conduction in the range of temperature 350-750 °C. To accomplish this, it was planned to deepen the informations on the systems obtained by doping the octahedral tetravalent cation site of the BaZrO3 perovskite matrices with a trivalent atom. In these structures, zirconium atom has been substituted by an Y one, in order to create oxygen vacancies that could be filled by hydroxyls groups. In this way, proton defects are inserted into the structure.
Once protons are introduced into the host matrix, their diffusion is mainly driven by phonon-assisted dynamics. Accordingly, the mechanism of protonic conduction is strongly affected by local distortions involving the doped Zr sites. For this reason, in order to know the details of the protonic conduction mechanism, hence to improve the performance of the related materials, it is of fundamental importance to study the local environment surrounding the doped sites.
Following this, the final tasks of this computational project are:
the development of structural and kinetic models to study the barium zirconate perovskite;
the validation, by mimicking experimental findings, of the structural and kinetic computing methods and models employed to study ceramic conductors, and the involved proton transport phenomena.
Integrated with the experimental findings, these computer simulations come out to be a further contribute to get a deep understanding of the physical and chemical characteristics of conducting materials and the related conduction phenomena.
Due to the characteristics of the studied materials, DFT approaches implemented into the SIESTA package have been investigated. The computational methods and the choice of the models and protocols have been tuned according to the machines capabilities, balancing the reliability of the results against the computational time. Geometry calculations have been performed on 3x3x3 and 4x3x3 supercells of pure and Y-doped BaZrO3 systems, by applying PBC.
Singly and doubly Y substitutions on Zr site have been considered, with and without an oxygen vacancy, together with different charge states. In this way, we were able to investigate the details of the dopant geometrical environment and its electronic features.
The systems above were used as a starting point to study the electronic and structural changes induced by inserting a proton into the BaZrO3 perovskite matrix. In order to do this, one proton was added to the pure and doped structures discussed above and the systems geometry were relaxed.
The results obtained for the undoped systems allowed us to validate and tune the computational methods, models and protocols suitable for this kind of compound, balancing the amount of the obtainable informations and the computational cost needed to get them at the required precision. Moreover, matching the structural and the electronic data so calculated, we were able to rationalize some experimental findings, giving a further contribution to the study of the proton transport phenomena in perovskite compounds.
Light-harvesting (LH) complexes are used by photosynthetic organisms to increase the overall efficiency of photosynthesis. This is accomplished by harvesting light energy and funneling it to the reaction center, where it is converted into electrochemical potential. Dinoflagellates, unicellular algae constituting one of the most important classes of phytoplankton, use a water-soluble LH complex called peridinin-chlorophyll-a-protein (PCP) with a 4:1 peridinin/chlorophyll ratio. The presence of peridinin molecules in the PCPs enables the organism to collect light in the visible spectral region where chlorophyll poorly absorbs. The peridinins of PCPs are also able to play a photoprotective role by efficient quenching of the chlorophyll triplet states, which may occasionally be populated, thus preventing the formation of the highly toxic singlet oxygen.
Infrared spectroscopy, and in particular time-resolved IR difference spectroscopy is a well-established technique which has been successfully used to investigate photophysical phenomena and photochemical reactions taking place in photosynthetic reaction center and LH complexes. This technique allows reaction-induced changes in both the protein and the cofactors to be monitored. In both static and time-resolved IR band assignment remains a difficult task. Vibrational frequencies can be obtained from theoretical calculations performed at molecular level such that the band assignment can be subsequently performed. A detailed and complete characterization of peridinin vibrational modes appears highly desirable also as a general purpose. In fact, it has been demonstrated that with a microspectroscopic Resonance Raman (RR) approach peridinin can be visualized directly in vivo.
In studying IR signal, for both singlet and triplet states, a key signature is provided by the carbonyl function of the lactone ring. In particular the effect of different protein environments of the four peridinins present in PCP complex, was evocated to understand IR spectroscopy.
Based on this motivation, a set of IR and RR experiments is going to be performed by Dr. A.Mezzetti of the University of Lille (France) to understand environmental effects on vibrational properties of peridinin. At this end three prototypical solvents were used in simulations: 1) an apolar/aprotic solvent, like cyclohexane; 2) a polar/aprotic solvent, like deuterated acetonitrile; 3) a polar/protic solvent like ethanol.
Ab-initio molecular dynamics should help for the band assignment. Using a mixed QM/MM approach, we studied the vibrational spectroscopy of a peridinin molecule, treated at DFT level, immersed in different classical solvents. This is achieved by using the CPMD package coupled with Gromos96. The active collaboration with prof. L. Guidoni of University of L’Aquila (former at University of Rome) the HPC fellowship was fundamental for the realization of the project.
From the resulting dynamics we obtained the vibrational signatures of the system, calculating the vibrational density of states by Fourier Transform (FT) of the velocity-velocity autocorrelation function. The IR absorption spectrum and the Raman spectrum are also calculated by FT of the dipole-dipole correlation function and the polarizabiliy-polarizability correlation function, respectively. All the resulting spectra will directly take into account anharmonic effects that should play a big role in such flexible molecule, in particular in solution. Also temperature effects are directly accounted for since molecular dynamics are performed at the same finite temperature of corresponding experiments. The assignment of vibrational bands will be done using the recently developed public license code based on the work of M.-P. Gaigeot and co-workers to analyze finite temperature equilibrium dynamics in terms of effective normal modes.
 Mezzetti A. and Spezia, R. 2008. Spectroscopy: Int. J., 22, 235-250.
Within this first HPC-Europa2 fellowship, we have firstly set-up QM/MM molecular dynamics of peridinin in three solvents: cyclohexane, deuterated acetonitrile and ethanol. This first part was also very useful to find the correct parameters to be used in QM/MM simulations. Actually, a Born-Oppenheimer dynamics was found to be the best solution, as the best compromise between accuracy and computational time.
We have thus run simulations within QM/MM scheme. First, we have performed cyclohexane and ethanol as solvents for which classical parameters were available in the standard classical force field. Two trajectories of 15 ps each were performed and peridinin dipoles calculated on the fly via the electron density. With these information we have produced first vibrational density of states (VDOS) and InfraRed (IR) spectra that show some expected features (C=O, allyle and skeletal positions) and solvation effect in particular on the C=O group. This reflects the expected H-bond effect.
The 15 ps simulation time-length seems to be enough to correctly reproduce vibrational features by inspecting the convergence of spectra as a function of simulation time for high energy bands. A final decision on that will come anyway from a further investigation that will take into account all the three simulations and experimental results in different solvents. Experiments in solution are actually work in progress.
The band assignment was done using the standard approach of VDOS decomposition initially. Three expected bands emerged : a high frequency band corresponding to C=C=C (at about 1960 cm-1), a typical C=O intense band at about 1720 cm-1 (depending strongly on the solvent : 1719 in gas phase, 1748 in cyclohexane, 1695 in ethanol) and a band at about 15000 cm-1 that corresponds to skeletal motion.
The quantitative contribution of each internal mode to the vibrational signals will be done with the MolSimu code that was modified to take into account a complex molecule like peridinin that is characterized by non-standard groups (this code developing was done during the HPC-Europa stay of Daniele Bovi in Evry).
The complete picture of solvation effects will be obtained when the simulations in acetonitrile (a polar/aprotic solvent), currently running by means of other HPC-Europa grants, will be achieved.