OPserver: opacities and radiative accelerations on demand C Mendoza (IVIC, Ven) LS Rodríguez (IVIC, Ven) MJ Seaton (UCL, UK) F Delahaye (OPM, France) P Buerger (OSC, USA) E Palacios (UC, Ven) A Bellorín (UCV,Ven) AK Pradhan (OSU, USA) M Meléndez (USB, Ven) CJ Zeippen (OPM, France) J González (UC/IVIC, Ven) High Accuracy Atomic Physics in Astronomy ITAMP 07/008/2006

e-Science is to do with the exploitation of the Internet as a powerful research environment Generate, store and analyze large volumes of scientific data LHC Virtual observatories Genomics/proteomics Perform large-scale modeling and simulations Nanoscience Establish and manage dynamic and distributed virtual organizations Grid

In GRID environments, computing is conceived as an efficient network service with unlimited capabilities GRID Portal

Scientific computing is mainly carried out under the traditional paradigm Source Compilation Input Disc Output Executable

Scientific computing in grid environments requires a more modern scheme: data-based centered Input Disc Output Executable DBMS

Scientific computing in grid environments is markedly distributed Disc CGI-scripts Executable 1/2 Web server Disc Executable 1/2 Supercomputer Disc HTML Java portlet Java application Web client

The standard model of the Sun interior has recently given rise to an intense polemic Helioseismology R(obs) =  R☼ R(theory) = R☼ (Basu & Antia 1997) Chemical abundance Z/H = (Grevesse & Sauval, 1998) Z/H = (Asplund et al. 2004) R(theory) = R ☼ Opacities OP: recently revised (Badnell et al. 2005) OPAL and OP in 2.5% accord The impact of the revised abundances has been extensively discussed by Bahcall & et al. (2005)

Detailed stellar models are now including effects due to microscopic diffusion Microscopic diffusion: Radiative levitation Gravitational Thermal diffusion It affects: Internal and thermal structures of star Convection zone depth Pulsations Anomalies in superficial abundances

Computing time of RMOs and RAs depends on the disk reading of monochromatic opacities (1Gb) For a chemical mixture with relative abundances f i, the Rosseland mean opacity (RMO) is given by  1/  R =    F(u)/  (u) du  where u=h /kT F(u) = [15/  4 ] u 4 exp(-u)/[1 – exp(-u)] 2 and the opacity cross section of the mixture  (u) =  f i  i (u) is the sum of the monochromatic opacities of each ion.

Computing time of RMOs and RAs depends on the disk reading of monochromatic opacities (1Gb) The radiative acceleration for the ith element is g rad =  R  i F/(c  i ) where F =  eff  R ☼ /r) 2 con B(T) = 2(  kT) 4 /(15c 2 h 3 ). The non-dimensional parameter   i =   i mta /  du  depends on the momentum transfer cross section  i mta =  i (u) [1- exp(-u)] – a i (u).

Stage 1 Stage 2 Mono 1 Gb mixv opfit end RMO mixv.xx 85 Kb mixv.in opfit.in opfit.xx start RA accv accfit end RA acc.xx 470 Kb acc.in accfit.in accfit.xx start RMO

OPserver exploits the client-server network architecture Opacity codes in OPCD (Seaton 2005) are restructured in a client-server network architecture User interaction is either through a web page or a linkable a subroutine library (OPlibrary) The RMOs and RAs are computed with the monochromatic opacities always loaded in RAM Remote calls are addressed through the HTTP protocol (URLs ) Codes have been parallelized (OpenMP)

OSC Web client Web server Supercomputer Modelling code OPlibrary Modelling code OPlibrary mono ABC