head 1.1; access; symbols; locks http:1.1; strict; comment @# @; 1.1 date 2004.10.12.20.14.22; author MarkAllen; state Exp; branches; next ; desc @none @ 1.1 log @Details of testing ELAIS SEDs and PEGASE models @ text @ Tests of generating SEDs and plotting against Pegase Models ----------------------------------------------------------- Mark Allen, CDS, 12 October 2004 The Galaxies scenario for the Jan 2005 demo aims to compare SEDs of galaxies with spectra from stellar population models. Ideally this would include fitting the models to the data, or at least selecting a best match model to the data. The aim of the tests described here is to identify the scientific transformations of the data and models required in order to make sensible comparisons. In summary these tests involve 1. Extracting SEDs from the ELAIS final merged band catalog (Rowan-Robinson et al. 2004, MNRAS, 351, 1290), including the transformations of INT-WFS Ug'r'i'Z, and 2MASS JHK filter magnitudes into flux densities (Jy). Putting the SED information into Specview compatible format (specview text files), and plotting 2. Extracting spectra from the PEGASE output spectra files. Each model run generates a set of spectra for the different stellar population ages - so lots of spectra just for one model. 3. Putting PEGASE model spectra into Specview compatible format, and plotting models and data together -------------------------------------------------------------- 1. Extracting SEDs 1.1 ELAIS final merged band catalog. A VOTable version of the ELAIS final merged band catalog is available, and a fully meta-data marked up version is currently being made avaialble in Vizier. However, using VOTable it is not clear how to extract the data from the required columns and repackage them in any VO compatible SED or spectra format so I have simply used an ascii version of the table and used IDL to manipulate the data. There are some problems with the values in this table, namely there are some negative flux values, many fluxes with 0.0 listed as the error. The 15 micron flux error is not provided, but needs to be calculated from the signal-to-noise-ratio which is given. The transformation constants required to calculate fluxes from the Ug'r'i'Z band magnitudes are not provided anywhere. CASU lists the nightly zero points for all the data, but how this translates into values in the ELAIS catalog is unclear. Lists of 'standard' values of vega fluxes in these bands (like the one NED uses) are not complete and do not include the CCD quantum efficiency. To do this properly one need to integrate a Vega spectrum over the filter bandpass and convolve with the CCD response function. Thomas Babbedge (Imperial College) has done this for the required filters and has provided the accurate transformation constants. (Regenerated SEDs do however have some anomalies with respect to those plotted in (fig 28) of the Rowan-Robinson paper.) 1.2 IDL programs for applying transformations and writing specview files. The elais2specview.pro IDL procedure reads data and creates specview files. There are no arguments, simply compile and run the procedure with the data files elaiscatf_21a_id.dat and elaiscatf_21a_id.dat in the same directory. These data files are just the elaiscatf_21a.dat file split into string and numeric parts to ease reading with IDL. IDL> .run elais2specview % Compiled module: ELAIS2SPECVIEW. % Compiled module: READ_DATA. % Compiled module: READ_ID. IDL> elais2specview This generates 3523 files containing the SEDs of all objects in the catalog. e.g. ELAISC15_J163412.0+405652_SED.txt ELAISC15_J163412.0+405652 micron mJy 200000. 0.50600000 0.010000000 175.000 249.00000 54.000000 90.0000 83.000000 10.000000 15.0000 2.6570000 0.095096636 7.00000 1.0070000 0.11700000 1.25000 0.14680571 0.015008650 0.909870 0.071449729 0.0019742290 0.767141 0.055552059 0.0046048805 0.621577 0.030846118 0.0014205163 0.485727 0.016611136 0.00030598845 0.360900 0.0045414980 0.00016731498 Note however, that some objects ( 1434 out of 3523) have only one point in their SED. Specview does not input files with only one point, but I suggest it should. All the Specview compatible files generated with this method are contained in ELAIS_specview_files.tar.gz 1.3 Plotting the SEDs in Specview. Two examples of plotting these SEDs in specview: specview_ELAISC15_J163412.0+405652_SED.jpg - plotting problem: note the messed up values on the flux axis specview_ELAISC15_J163608.1+410507_SED.jpg Apart from the occasional hiccup with axis values, this is quite a good way for inspecting SEDs and changing units etc. 2. Extracting spectra from PEGASE outputs For each model star formation scenario, PEGASE generates a set of spectra representing the integrated luminosity of the population at different ages. All the spectra are in one big output file (e.g. spectra1.dat as described in the Pegase notes). One could imaging putting this into a hierarchical VOTable, but since there are no tools yet for handling such things there's not much point. pegase2specview.pro is an IDL proceudre for unpacking these output files. IDL> .run pegase2specview % Compiled module: PEGASE2SPECVIEW. IDL> pegase2specview, 'spectra1.dat',/w,/p spectra1.dat This generates a set of files (69 in the case of our spectra1.dat example), that are ready for plotting in Specview. These files are in PEGASE_model_1.tar.gz Using the multiple input file option in specview, this is what you get when you plot all models from spectra1.dat at once. all_models.jpg Note that the units of the PEGASE model spectra are in erg/s/Angstrom. Specview plots the correct units, but does not permit any transformations on it. This is because erg/s/Angstrom is not in specview's standard list - also because one cannot convert this to f-lambda or f-nu etc. without putting in a distance. These model spectra represent to the total luminosity generated by 1 solar mass of material, radiated into all directions. To really compare the data and the models, the model spectra need to be scaled, and redshifted to appropriate values. This involves transformations on the wavelength, and luminosity values. 3 Plotting SEDs and PEGASE models together In order to check the shapes of the models compared to the data, I've done a simple scaling (no redshift). model_data_eg.dat shows and example of a random model, plotted on a random SED. 4. Conclusions 4.1 SED info can be transformed into the correct units and file formats for plotting them, but getting accurate transformation constants is not easy. 4.2 Current VO tools are not flexible enough to generate the SED files from catalog, but it would not be difficult to write dedicated programs to apply the transformations detailed here. 4.3 Specview is fairly well suited for our purposes. Using its multiple input file capability, and it's combining functions it can almost act as a SED generator. This capability would be enhanced if specview could take files with only one point. That way, you could have a set of photometric point, each with its own units for wavelength and flux (not Vega magnitudes), and specview could put them together in a sensible way. Another VO style concept for this would be if a service that generated the sets of spectra as described here, made them available via SSA, so that Specview could load them via SSA (?) Specview also has a sopisticated set of fitting functions, but I have not been able to get sensible results using it yet. 4.4 VOSpec has not been tested 4.5 The PEGASE models can be put into specview format from their original output format, but work need to be done on the scaling and transformations of the models in order to compare with the SED data. @