Idenification of AGN among the many thousands of galaxies that will be detected, in particular high redshift and/or very dusty objects, candidates for follow-up spectroscopic studies. Estimation of photometric redshifts.

• Use SEDs of various AGN and normal galaxy types, spanning the whole range (from X-rays to radio-wavelengths) that the GOODS data will cover. Combine these with GOODS flux limits and (anticipated) redshifts (luminosity functions, evolution, cosmology?). Create a multi-wavelength space.
  
• Study the n-colour space clustering of the various types to reveal the most efficient colour-criteria for finding AGN.

A set of IDL procedures was used to generate a simulated catalogue of various AGN types, a template of average AGN SEDs and to recover info as it would be observed with a combination of filters. At present, a limited set of filters and limited z and mag-ranges were used to test the use of the method. Later-on, it will be refined and improved to include a more extended range of galaxy properties and wavelengths.
The steps followed are described here:

• Compile (~60) individual AGN SEDs, of 7 types: radio-loud and radio-quiet (Elvis et al, 1994 AJSS 95,1), Seyferts of type 1, intermediate (Malkan, 2000 pc) and type 2 (Schmitt et al, 1997 AJ 114,592), and Starburst galaxies with high and low internal reddening (Schmitt et al., 1997 AJ 114,592).
  
• Choose representative (and in some cases two "extreme") SEDs for each type. These can be seen (with the 13 filter response curves, overplotted) in Figure 1.
  (I apologize for the .ps figures here, that all appear for some reason upside down. I will correct this, but in the mean time just use [orientation]-[swap_landscape] on the ps viewer)
  Randomize the magnitude (for m(3mum) lt 19 ), the redshift (for 0.01 lt z lt 2 or 4) and the SED type creating a fake catalog of 1000 "observed" galaxies.
  Use 13 filters (UBVRI (ESO 2.2m WFI), JHK (Kitt Peak IRIM), IRAC_1-4, MIPS_24) and convolve these with the above 1000 galaxies, thus obtaining a grid of (13,1000) fluxes/magnitudes.
  
• Construct average SEDs for each of the 7 AGN types (radio-loud, radio-quiet, Sy1, Sy2, SB, SB-high, SB-low) using the original data (~60 SEDs) of step 1, normalized at 7000 A. Repeat the steps described above. In Figure 2 you see the average SEDs with the 13 filter curves overplotted.
  
• Use the normal galaxy catalog of Fioc et al. to create a simulated catalog of 1000 objects, as done above for the AGN. I use the same magnitude and redshift ranges and 7 morphological types (E,S0, Sa, Sc, Sc, Sd, Irr)

• In Figure 3 are the (78) colour combinations as a function of AGN type (upper panels) and redshift (lower panels, with the same symbols for the SED type as the upper panels) and I use these to select which colours/combinations would be best diagnostic tools.
  
• In Figures 4,5, 6 and 7 I show various colour-colour plots, using the same symbols as before for the different SED types: RL: x, RQ: squares, Sy1: triangles, Sy2: diamonds, SB-low: +, SB-high: *. The different symbol sizes indicate the redshift of the galaxy in question, z ranging from 0-2 (smaller to larger symbols). The loci of (all) normal galaxies are shown as dots.
  
• In Figures 8, 9, 10 and 11 I show the same figures only for the normal galaxies now, with different symbols for each Hubble type: E:crossed squares, S0: + , Sa: * , Sb . , Sbc diamonds , Sc: triangles , Sd: squares and Irr: x
  The 24$\mu$m band seems to be essential for differentiating AGN from normal galaxies. Using a combination of colours, both the (nuclear for AGN, morphological for normal galaxies) type and redshift of an object can be derived almost solely on the basis of its position on those diagrams.
  
• Using the average AGN SEDs I constructed similar colour-colour plots (same symbols as above for different AGN types) for z=0.01-2 (Figures 12, 13, and 14) and with the z range expanded up to 4 (Figures 15, 16, and 17 )
  The colour ranges are shrunk and the tracks of different types are blurred. Both effects are expected and is clear that one shouldn't use average SEDs to do this type of work.

- Add more objects (ideally 30) for each AGN type.
- Add X-ray, UV, far-IR and radio-data.
- Define somehow the photometric segregation of different galaxy types in n-colour space.

last: Nov.6,2001