DIAGNOSTICS 83

6450 6500 6550 6600 6650 6700 6750 6800 Wavelenght

0 100 200 300 400 500 600 700

Flux

CIG: 214

(a) CIG 214 shows a clear strong wide component in Hα.

6500 6550 6600 6650 6700 6750

Wavelenght 0

100 200 300 400 500 600

Flux

CIG: 336

(b) CIG 336 has a weak wide component in Hα

Figure 7.11: Examples of the spectra of galaxies classified as Seyfert 1. The wave- length units are Å and the flux units are 10⁻²⁰W/m². The stellar component has been already subtracted from the spectra. The theoretical location of the Hαline is marked with a solid black line, the locations of [Nii] and [Sii] lines are marked as blue dashed lines and as yellow dashed lines respectively.

Table 7.7: Wavelengths of the emission lines.

Line Wavelength in vacuum (Å) Wavelength in air (Å)

Hβ(4861) 4861.3 4862.72

[Oiii] 5007 5006.9 5008.24

[Oi] 6364 6363.8 6365.54

[Nii] 6548 6548.0 6549.86

Hα(6563) 6562.8 6564.61

[Nii] 6583 6583.4 6585.27

[Sii] 6716 6716.5 6718.29

[Sii] 6730 6730.8 6732.68

The line ratios should be between an Hydrogen line of the Balmer series and a forbidden line, less sensitive to abundances.

The lines should be easily accessible by the instrumentation, hence they choose optical lines and rejected UV ones.

The chosen forbidden lines were [Oiii] at 5007Å, [Oi] at 6364Å, [Nii] at 6583Å, [Sii] at 6716Å and 6730Å and the Balmer series lines of Hydrogen Hαand Hβ(see Table 7.7). Based on these lines they proposed 4 line ratios: a) log([Oiii]/Hβ);

b) log([Nii]/Hα); c) log([Sii]/Hα) and d) log([Oi]/Hα). The [Oiii]/Hβratio is mainly an indicator of the mean level of ionization and temperature, the [Sii]/Hα and [Oi]/Hαratios are indicators of the relative importance of a large partially ion- ized zone produced by high-energy photoionization. The significance of the ratio [Sii]/Hαis not immediately obvious but provides a good separation between star forming nuclei and AGN (Osterbrock & Ferland 2006).

Finally, these ratios were grouped to form the 3 following diagnostic diagrams:

log([Oiii]/Hβ) - log([Nii]/Hα) log([Oiii]/Hβ) - log([Sii]/Hα) log([Oiii]/Hβ) - log([Oi]/Hα)

Veilleux & Osterbrock(1987) proposed several boundaries separating the theoret- ical starburst region from objects of other types of excitation like an AGN in the diag- nostic diagrams. These boundary lines are semi-empirical.Ho et al.(1997a) proposed different empirical selection lines for these three diagrams that also take into account the difference between Seyfert and LINERs.Kewley et al.(2001) proposed a theoret- ical line to divide the regions. They used PEGASE v2.0 (Fioc & Rocca-Volmerange 1997) and STARBURST99 (Leitherer et al. 1999) codes to generate the spectral en- ergy distribution of the star forming galaxies and used their code MAPPINGS III (Dopita et al. 2000) to compute the photoionization models for these galaxies. E.g. in the figure 4 ofKewley et al.(2001) (Figures7.12aand7.12bhere) are plotted, with a grid, the different locations of a theoretical starburst with different metallicities (z) and ionization parameters (q), over this grid they plot the location of a sample of in- frared starburst galaxies. The location for some galaxies can be explained directly by a starburst but for some galaxies it is clear that there is a systematic separation from the pure starburst models. To explain the location of these galaxies it is needed to take

7.6. DIAGNOSTICS 85 into account an additional ionization source like an AGN.Kauffmann et al.(2003a) provided an empirical separation line between pure star forming galaxies and galaxies that they assumed powered by an AGN. This study made use of a sample of 122 808 galaxies from the SDSS, one of the larger if not the largest sample used until that moment. The separation line is below the one ofKewley et al. because it takes into account the empirical shape of the distribution of galaxies, which is well below the theoretical “extreme starburst” line ofKewley et al.. Stasi´nska et al.(2006) provided a new line of separation used their own photoionization code (see Figure7.13a). This theoretical line was even below the one ofKauffmann et al..

Some authors use different lines from the above mentioned to perform their own classificationDecarli et al.(2007);Miller et al.(2003). The difference in the defini- tions must be taken into account to compare between different samples.

Stasi´nska et al.(2008) studied the possibility that some of the galaxies in the region of AGN could actually be galaxies powered by stellar processes. The photoionization of some of these galaxies could be produced by hot post-asymptotic giant branch stars and withe dwarfs (Binette et al. 1994). In Figure7.13bare shown the new grids and extension of the photoionization models ofStasi´nska et al.(2008) in the diagnostic diagram.

LINERs were defined as an own class of galactic nuclei byHeckman(1980). The classification was based on the relative strengths of [Oi], [Oii] and [Oiii] emission.

[Oii] at 3727Å has to be at least as strong as [Oiii] at 5007Å and [Oi] at 6300Å has to be at least as strong as one third of [Oiii] at 5007Å. AsHo et al.(1997a) explain in their paper, although a criterion using the log([Oiii]/Hβ) - log([Oi]/Hα) diagnostic diagram is different from the original criterion ofHeckman(1980), in practice the two sets of classification criteria identify the same objects (Ho et al. 1997a). This comes from the inverse correlation between [Oiii]/Hβand [Oii]/[Oiii] for conditions of ow excitation. Kewley et al.(2006) gave an empirical classification scheme to separate Seyferts and LINERs. It is based in the empirical separation of Seyferts and LINERs in two clearly separated branches in the diagnostic diagrams.

Classification criteria

The lines of separation between different types of galactic nuclei are usually clearly and unambiguously defined in the literature. This is not the case for the final classi- fication of a nucleus. The first source of ambiguity occurs when a condition for the classification of a galaxy is not hold simultaneously in the three different diagnostic diagrams. In some cases the weight of the classification accuracy is assigned depend- ing on the diagnostic diagram. Some diagrams are considered as more reliables than others. E.g.Ho et al.(1997a) consider the log([Oiii]/Hβ) - log([Oi]/Hα) diagnostic diagram as the most sensitive to the shape of the ionizing spectrum, hence, a higher weight is given to the classification of this diagram. In other cases galaxies classi- fied as different types are directly considered as “ambiguous galaxies” (Kewley et al.

2006).

Another source of confussion appears when a line is not measured due a low signal to noise ratio. This means that one or two of the ratios involved in a diagnostic diagram is an upper or lower limit or is even not defined when the two lines involved are non- detections. These cases are usually discarded from the classification or are studied in a case by case basis.

Some studies only use one diagnostic diagram or even one line ratio to classify the galaxies. The usual diagnostic diagram used is log([Oiii]/Hβ) - log([Nii]/Hα) and

(a) Starburst models ofKewley et al.(2001) computed with PEGASE

(b) Starburst models ofKewley et al.(2001) computed with STARBURST99

Figure 7.12: Theoretical models in the diagnostic plot of log([Oiii]/Hβ) and log([Nii]/Hα). In both the subplots the grid corresponds to the theoretical model for different metallicities (z) and ionization parameters (q) fromKewley et al.(2001).

A sample of infrared starburst galaxies is plotted over the theoretical models. Source:

Kewley et al.(2001).