Astronomical Data Analysis Software and Systems XVII P8.8 ASP Conference Series, Vol. XXX, 2008 J. Lewis, R. Argyle, P. Bunclarck, D. Evans, and E. Gonzales-Solares, eds. GOSSIP, a new VO compliant tool for SED fitting P. Franzetti, M. Scodeggio, B. Garilli, M. Fumana, L. Paioro INAF - IASF Milano, Milan, Italy 8 0 0 Abstract. We present GOSSIP (Galaxy Observed-Simulated SED Interac- 2 tive Program), a new tool developed to perform SED fitting in a simple, user n friendlyandefficientway. GOSSIPautomaticallybuilds-uptheobservedSEDof a anobject(oralargesampleofobjects)combiningmagnitudesindifferentbands J andeventuallyaspectrum;thenitperformsaχ2 minimizationfittingprocedure 6 versusaset ofsynthetic models. The fitting resultsareusedto estimate a num- 1 berofphysicalparametersliketheStarFormationHistory,absolutemagnitudes, stellar mass and their Probability Distribution Functions. User defined models ] can be used, but GOSSIP is also able to load models produced by the most h commonly used synthesis population codes. GOSSIP can be used interactively p withother visualizationtoolsusing the PLASTIC protocolforcommunications. - o Moreover,sinceithas beendevelopedwithlargedatasets applicationsinmind, r itwillbe extendedtooperatewithin the VirtualObservatoryframework. GOS- t s SIP is distributed to the astronomical community from the PANDORA group a web site (http://cosmos.iasf-milano.inaf.it/pandora/gossip.html). [ 1 v 8 1 5 2 1. 1. Introduction 0 8 0 GOSSIP is a tool created to fit the electro-magnetic emission of an object (the : SED, Spectral Energy Distribution) against synthetic models, to find the sim- v ulated one that best reproduces the observed data. It has been developed to i X perform this task in a simple, user friendly and efficient way. r GOSSIP was born within the frameworks of the VVDS (Le F´evre et al. 2005) a and the zCOSMOS (Lilly et al. 2007) surveys; therefore it has been optimized to work on huge amounts of data as the ones provided by modern photometric and spectroscopic surveys. 1 GOSSIP has been written by the PANDORA Group at INAF IASF-Milano, 2 using the Python language for the graphical part and the C language for the high performance computational tasks. 1http://cosmos.iasf-milano.inaf.it/pandora 2http://www.python.org 1 2 Franzetti et al. Figure 1. The GOSSIP SED visualization window. In this example it is shown a SED composed by an optical spectrum (the rapidly varying thick solidline)andmagnitudepoints(thedots). ThiszCOSMOSSEDiscomposed by datafromVLT-VIMOS,CFHT,SUBARU andSPITZERdata. The best- fitting PEGASE model (thin smooth line) is over-imposed to the data while its main parameters are summarized in the right panel. 2. GOSSIP flow chart 2.1. SED building GOSSIP builds-up the observed SED of an object combining magnitudes in differentbandsfromdifferentinstrumentsand,eventually, alsoaspectrum. Data can be read both from ASCII files or from a MySQL database table. Figure 1 shows an example of a SED built by GOSSIP using data from the zCOSMOS survey. 2.2. Synthetic models GOSSIP fits sets of synthetic models against observational data. It can use user-defined models, or “standard” ones computed by the most commonly used synthesis population codes like PEGASE (Fioc & Rocca-Volmerange 1997) or Bruzual & Charlot (Bruzual & Charlot 2003) by directly reading the output files produced by these codes. It can also compute a set of parameters on the loaded models (both user-defined and “standard”) like a photometric color or a spectral index. In figure 2 the Bruzual & Charlot import panel is shown. 2.3. The fitting procedure 2 GOSSIP performs a χ minimization to find the model which best represents observational data in interactive or in batch mode. The minimization results for each object are stored into a file, so that different GOSSIP, a new VO compliant tool for SED fitting 3 Figure2. TheBruzual&Charlotmodelsimportpanel. Theusercanchoose originalmodels parametersto be loadedandalsoparametersto be computed by GOSSIP itself. filters can be applied to them to select the best-fitting model. Model ages can be left unconstrained, or bounded by the age of the Universe at the object’s redshift, or the user can select a fixed galaxy formation epoch for the whole sample. Moreover the input data set can be limited to the use of only the pho- tometric or the spectroscopic data. Fitting results can be plotted in the main SED visualization window (see figure 1) where the user can switch very rapidly between the various filters. The fitting procedure can run on a single CPU or on a BEOWULF cluster; porting to the computational GRID is currently on-going. Once the fitting pro- cedure has been performed, a number of “post-fitting” operations are carried out, which includethe computation of physical parameters, like absolute magni- tudesandstellar masses,fromthebest-fittingmodelandoftheirfullProbability Distribution Functions from the full set of synthetic models. 2.4. Interoperability 3 GOSSIP can connect to a running PLASTIC hub to send the fitting results to other specialized tools, like TOPCAT or VISIVO for visualization. 3http://plastic.sourceforge.net/ 4 Franzetti et al. 3. GOSSIP and the Virtual Observatory A standard VO interface within GOSSIP is currently under development. Using thisinterfaceGOSSIPwillbeabletodownloadsinglespectrafromaSSAservice to be fitted against synthetic models. However, since a relatively narrow wave- lengthcoverage resultsintolargeuncertaintiesintheSEDfittingresults,asingle spectrum very poorly constrains a GOSSIP fit. The results can be significantly improved using also magnitudes to extend the wavelength coverage. Individual aperture magnitude measurements in several optical and NIR filters can be cur- rently gathered using VO protocols, but they are not suited for a correct SED fitting, as detailed information about the aperture at which each data point has been obtained is required to build up a reliable SED. Aperture characterization is already provided in the Spectrum Data Model (McDowell, 2007), but it is not yet much used by data providers. An alternative approach could be to down- load a ready-made observed SED provided by some VO service; in this case the SED would be built by the data provider after having properly normalized the magnitude measurements to a single common aperture. Unfortunately the SED Data Model is still under definition and it is not yet possible to obtain ready made SED through the VO. Another limitation we are facing, irrespective of the approach used to build a SED, is the lack of a detailed characterization of the magnitudes. The SED fitting procedure requires the knowledge of the exact shape of the filter transmission curve associated to each magnitude in order to 2 computethesyntheticmagnitudesfromthemodelsthatareusedintheχ mini- mization. However a specification of the filter responsecurve is not yet included in the Spectrum Data Model. Despite the fact that these limitations prevent SED fitting within the current VO implementation, GOSSIP is being developed paying close attention to the development of the VO technology. When these issues will be solved, it will be ready to exploit all the VO potentialities. 4. Conclusions We have presented GOSSIP, a new tool dedicated to SED fitting. It has been developed in order to perform this task in a simple, user friendly and efficient 2 way. GOSSIP builds-up observed SEDs, loads synthetic models, performs χ minimization fitting between them and estimates physical parameters for huge samples of objects. The features implemented within this program make it a very useful tool for the analysis of the large data samples currently available to the astronomical community. References Bruzual, G. & Charlot, S. 2003, MNRAS, 334,1000 Fioc, M. & Rocca-Volmerange,B. 1997,A&A, 326, 950 Le F´evre, O. et al. 2005, A&A, 439,845 Lilly, S. et al. 2007, ApJS, 172, 70 McDowell, J. et al. 2007, IVOA Documents, Spectrum Data Model Version 1.02, available at: http://www.ivoa.net/Documents/latest/SpectrumDM.html