Universidad de Valencia - Tirant Node

From AnnualReport

Jump to: navigation, search


The TIRANT Supercomputer

The RES node located at the University of Valencia was inaugurated in January 2008. The supercomputer's name Tirant refers to the main character in the "Tirant lo Blanc" novel written by the Catalan author Joanot Martorell in 1490. The node is installed in a specially designed data centre and is managed by technicians of the Servei d'Informàtica de la Universitat de València (SIUV). The SIUV has broad experience in managing supercomputers and has been key to the university's efforts in the field of scientific computing since 1978. The SIUV also hosts the RedIris PoP of the Valencian region, the network infrastructure of the university, the central database, the university web page, e-mail services, application services and other computer services.

In December 2012, Tirant was upgraded as a consequence of the installation of MareNostrum3, by virtue of the agreement between RES nodes. Tirant now has 2048 processors and 4TB of distributed memory.

Organisational Structure

From the left: Alejandro Soriano - System Analyst, Jose María González - Operator and Josep Vicent Sala - System Analyst

Tirant is managed by technicians from SIUV who report to the systems group director and who are responsible for system management (dealing with hardware problems, installation and configuration of system software) and user support (compiling scientific programs, managing the system queue, solving user problems). The RES access committee is responsible for assigning the Tirant's CPU hours to users by evaluating new incoming projects. In addition, a local access committee (CARS) receives and evaluates scientific projects from all Valencia and is responsible for the distribution of local hours among them.

Technical and Scientific Highlights 2014

In 2014, Tirant offered more than 7,7 million CPU hours, of which 4,8 million (~62%) hours were used by the scientific community of Valencia, including researchers at the host University. The rest was consumed by RES users.

Key Publications 2014

  • Aranda, J., Zinovjev, K., Roca, M., and Tuñón, I. Dynamics and Reactivity in Thermus Aquaticus N6-Adenine Methyltransferase, Journal of the American Chemical Society 136, 16227-16239 (2014)
  • Zinovjev, K., and Tuñón, I. Exploring Chemical Reactivity of Complex Systems with Path-Based Coordinates: Role of the Distance Metric, Journal of Computational Chemistry 35, 1672-1681 (2014)
  • Luk, L. Y. P., Ruiz-Pernía, J. J., Dawson, W. M., Loveridge, E. J., Tuñón, I., Moliner, V., and Allemann, R. K. Protein Isotope Effects in Dihydrofolate Reductase from Geobacillus Stearothermophilus Show Entropic–Enthalpic Compensatory Effects on the Rate Constant, Journal of the American Chemical Society 136, 17317-17323 (2014)
  • M. Boronat, A. Leyva-Pérez, A. Corma. Theoretical and Experimental Insights into the Origin of the Catalytic Activity of Subnanometric Gold Clusters: Attempts to Predict Reactivity with Clusters and Nanoparticles of Gold. Accounts of Chemical Research 47, 834 – 844 (2014)
  • M. Boronat, T. López-Ausens, A. Corma. Making C-C bonds with gold catalysts: a theoretical study of the influence of gold particle size on the dissociation of the C-X bond in aryl halides. Journal of Physical Chemistry C 118, 9018 – 9029 (2014)
  • M. Boronat, S. Laursen, A. Leyva-Perez , J. Oliver-Meseguer, D. Combita, A. Corma. Partially oxidized gold nanoparticles: A catalytic base-free system for the aerobic homocoupling of alkynes. Journal of Catalysis 315, 6 – 14 (2014)
  • M. Boronat, A. Pulido, P. Concepción, A. Corma. Propene Epoxidation with O2 or H2/O2 Mixtures over Silver Catalysts: Theoretical Insights into the Role of Particle Size. Phys. Chem. Chem. Phys. 16, 26600 – 26612 (2014)

Contributed talks and posters 2014

  • Oral contribution at 15th International Conference on Theoretical Aspects of Catalysis ICTAC 15, London. “Theoretical investigation of the exceptional oxidation activity of gold clusters of low atomicity” M. Boronat, P. Concepción, A. Corma
  • Conference: Simulation of atomization process in diesel. F.J. Salvador, D.Jaramillo-Ciscar, J.-V. Romero and M. D. Rosello, Modelling engineering and human behaviour, September 2014, ISBN 978-84-606-5746-0.

Key Projects 2014

Development of novel treatments for myotonic dystrophy: in vivo drug discovery

MBNL1 (to the left) is sequestered by binding to CUG RNA (right in the image)

Myotonic dystrophy type 1 (DM1) is caused by the expansion of noncoding CTG repeats in the DMPK gene. Mutant transcripts form CUG hairpins that sequester RNA-binding factors into nuclear foci, including MBNL1 protein, which regulate alternative splicing and gene expression. A previous study in our lab proved that D-amino acid hexapeptides can bind to CUG hairpins and improve the phenotype of a DM1 Drosophila melanogaster model, but the activity of these peptides was weak. Using the Tirant supercomputer we developed a computational model of the interaction between the hexapeptides and the CUG RNA and are using this structural information to develop peptides with stronger affinity for CUG repeat expansions.

Thermal conductivity of GaN nanowires

In this work the group led by Heruy Mengistu investigates the thermal conductivity of GaN nanowires with wurtzite structure by using Green-Kubo approach. The molecular dynamics calculations are performed by using LAMMPS code. The nanowire is aligned along [0001] crystallographic direction and has hexagonal cross sectional geometry. Periodic boundary conditions are applied in the axial direction. Nanowires of diameters 0.781, 1.86 and 2.61 nm are investigated by using a periodic computational cell of length 15.51 nm. The interatomic interactions are described by means of the Stillinger-Weber potential. The heat fluxes are obtained over a simulation time of 9ns in the NVE ensemble for a temperature range of T=300-1200 K. To accurately compute the thermal conductivity, first the normalized heat current autocorrelation function integral is adjusted to decay to zero with time. Calculations show that the thermal conductivity of GaN nanowires decreases with decrease of diameter from 2.6 to 0.781 nm. It was also observed that for all the diameters considered, the thermal conductivity becomes less affected by higher temperatures values. This shows that the contribution of the surface scattering to the conductivity of the wires is dominant but decreases with temperature.

Acid base reactivity of zirconia polymorphs

ZrO2 has high thermal and chemical stability, as well as structural versatility. It has weak acid and basic centres, and therefore shows interesting acid-base bifunctional catalytic properties. The strength of acid and basic sites depends on the coordination of the exposed metal cations and oxygen anions, respectively, and therefore on the morphology of the surface exposed. ZrO2 presents different polymorphs that can be obtained by controlling the synthesis parameters and that might show different acid-base properties. This project, led by Mercedes Boronat, theoretically investigated the acid-base catalytic properties of the most stable facets of monoclinic and tetragonal ZrO2 polymorphs, and their influence on selected acid-base catalysed reactions: ketonic decarboxylation, Knoevenagel and Claisen-Schmid condensations, and Meerwein-Ponndorf-Verley reduction.

Energy profile and structures involved in the condensation and decarboxylation steps of the mechanism of ketonic decarboxylation of carboxylic acids over monoclinic zirconia

Monte Carlo study of the SU(2) Higgs Model

One of the goals of the project, led by Pilar Hernandez, is to study the existence of a scaling window in the four-dimensional gauged non-linear sigma model, which at low energy could describe the Higgs-gauge sector of the Standard Model. That would be an alternative to the Higgs mechanism, however it could only be seen as an effective theory as the model can not be extended to arbitrarily large scales. In JHEP 1311 (2013) 213 the group found strong indications for the existence of such a scaling window up to the TeV scale. However the results for the observables used for this scaling study had rather large statistical errors, and a large share of the CPU time allocated on TIRANT in 2014 was devoted to reduce those. Also, the group studied the effective tree-level vector-bosons coupling at finer lattice resolutions and confirmed that such coupling is a factor 2 – 3 larger than in the Standard Model. For the model to be viable, the inclusion of fermions, and higher order terms in the matching between the fundamental and the effective theory, must account for the discrepancy. Finally the group optimized the setup for the search of excited states and non Standard-Model states, which would represent striking signatures of the model and could be searched at LHC. 8 lectures were given in Europe on the subject during 2014 (2 in Spain, 3 in Denmark, one in the UK, one in Germany, one in Ireland).

Molecular simulations of enzymatic reactions

The research group lead by Ignacio Tuñón focuses on the simulation of enzymatic reactions using hybrid techniques, combining mechano quantum description (QM) of the active site while using molecular mechanics (MM) for the rest of the system (protein and solvent).

Structure of the N6-DNA-Methyltransferase M.TaqI enzyme in complex with a decamer of DNA used in our simulations, (PDB ID: 1G38). An inset of the active site is shown

A detailed analysis of enzymatic reactions requires an appropriate representation of the environment and then the consideration of a huge number of atoms. Moreover, in order to describe the breaking and forming of bonds that take place in the enzyme’s active site we have to describe the atoms involved in the process with the Quantum Mechanics (QM) methodology. As the full quantum mechanical treatment of the enzymatic system is unaffordable nowadays, we restrict this description to those atoms directly involved in the chemical reaction (bond breaking and forming processes). The effect of the rest of the system can be included by means of classical force fields (or molecular mechanics, MM) that influence the QM subsystem. We use these hybrid QM/MM methods to simulate reaction mechanisms in complex biomolecular systems, testing them against accurate experimental data and reproducing subtle effects such as the change in the mass of some protein atoms. In addition, methodological development of QM/MM methods leading to more efficient and faster algorithms were carried out in order to expand the chemical information that can be obtained in our simulations. One of the latest applications of our QM/MM methodology was carried out for the study of the reaction mechanism catalysed by the N6-DNA-Methyltransferase, an enzyme involved in the epigenetic methylation process of DNA in bacteria. The most feasible reaction mechanism in terms of free energy barriers was characterized. The knowledge about the characteristics comprising the transition state for the rate limiting step is a perfect mold for the development of new drugs and antibiotics against undesirable pathogens.

Simulations of relativistic flows in active galaxies and binary stars

Snapshot of rest-mass density extracted from a 2D axisymmetric simulation of a relativistic flow interacting with the bubble formed a by a stellar wind. This scenario is relevant to explain mass-load of extragalactic jets and possibly high-eenrgy emission from these sources. The simulation includes energy losses via thermal cooling and a relativistic gas equation of state. Perucho et al., in preparation

During 2014, this group led by Manel Perucho published results of simulations performed during previous years at Tirant and Mare Nostrum. Results were published on long-term jet evolution and their influence on the intergalactic medium (Perucho, Martí, Quilis, Ricciardelli, MNRAS) and on mass-load of jets by stellar winds (Perucho, Martí, Laing, Hardee, MNRAS). Some of the simulations presented in those papers were performed at Tirant. The image shows maps of rest-mass density and temperature of the simulation of a slow-massive jet evolving through a galaxy and the intergalactic medium.

In addition, the group continued the study of wind/wind interaction in a binary system formed by a non-accreting pulsar and a massive star. The aim was to study the locations of high-energy emission produced by the interaction of the dilute, relativistic pulsar wind, with the slow, dense and inhomogeneous wind from a massive star. The objective is to achieve a quasi-steady-state scenario and perturb it with typical stellar wind inhomogeneities - a two-dimensional study of this scenario has already been published (Paredes-Fortuny et al. 2015).

Simulation of atomization process in diesel

Incompressible solver

In 2014, the group led by Francisco Javier Salvador continued the study on the internal flow and cavitation phenomenon on Diesel injectors. The study focused on geometrical influences of nozzle parameters and on the differences between turbulence models to proper simulate diesel behaviour in the injector. The study focused on an incompressible monophasic solver. It showed similar values in terms of mass flow error, around 2% for all the turbulence models studied except for the standard k-epsilon that showed an average error of 6%. This study was also extended to a non-simplified solver, considering the compressibility of the liquid under a non-cavitating geometry. Together with standard k-epsilon and RNG k-epsilon some others k-epsilon models were tested: 2 equations models as Launder-Sharma (apply a near wall treatment, considering some damping functions over the standard k-epsilon model coefficients) and realizable k-epsilon model (this model provides better performance under strong streamline curvature, vortices and boundary layers under strong pressure gradients). As a non-linear eddy viscosity model, Lien cubic k-epsilon model was tested.

Personal tools