Figure 3.4: UV/optical emission lines in starburst spectra from 1500 ˚A to 10000 ˚A from
Kinney et al. [1996]. The figure is based on the central regions of 35 starburst galaxies (described in Calzetti et al. [1994]) with low extinction (E(B-V)< 0.1). Line identifica- tions have been added: for identification of other emission lines, see Figure 3.3.
Metallicity
Line intensity ratios can be used to estimate metallicity. A well-known for- mula (referred to as R23) uses oxygen abundance (12+log(O/H)) as a tracer for metallicity (Pagel et al. [1979]; Kobulnicky and Kewley [2004]; Geller et al. [2014]):
R23 = [OII]λ3727 + [OIII]λ4959 + [OIII]λ5007
Hβ (3.1)
although this may need to be corrected for the ionisation level, to which it is also sensitive. Better estimates start with the electron temperature, which can be measured by [OIII] λ5007/λ4363; however the λ4363 was usually too weak to be observed in the data analysed in this thesis.
3.2.2
Separating AGNs from star forming galaxies
A key question in examining mid- and far-infrared selected galaxies is to separate the luminosity from AGN accretion and the luminosity from star formation. One
3.2.2 Separating AGNs from star forming galaxies
method is to use line widths: the accretion disc around an AGN (and outflows in the narrow-line region) will cause Doppler broadening of the emission lines observed, and a Hα line width over ∼500 km s−1 is likely to indicate an AGN.
Another commonly used method is to use line ratios: the relative strengths of the lines of highly ionised elements can be used as a proxy for the hardness of the radiation. These ratios are most reliable when the effect of relative dust obscuration is minimised (by choosing lines which are close in wavelength in the UV/optical, or FIR lines which are not affected by dust obscuration) and when the effect of differing abundances is eliminated (by using the same element). The first effect is achieved in the BPT diagram, which is described next.
BPT diagram
The Baldwin Phillips Terlevich (BPT) diagram (Baldwin et al. [1981]) uses a plot of [NII]/Hα vs. [OIII]/Hβ emission line intensity ratios31 to distinguish
AGNs from star-forming galaxies. AGNs have much harder radiation than hot stars, causing more ionisation of the nitrogen and oxygen atoms, increasing the intensities of these two forbidden lines. Two demarcation lines are often used with this plot: firstly, the Kewley line (Kewley et al. [2001a]; Kewley et al. [2001b]) to identify starburst galaxies:
log10([OIII]/Hβ) < 0.61/[log10([NII]/Hα)− 0.47] + 1.19 (3.2)
and secondly, the Kauffmann line (Kauffmann et al. [2003]) to identify AGNs: log10([OIII]/Hβ) > 0.61/[log10([NII]/Hα)− 0.05] + 1.3 (3.3)
The region between the two lines is considered a composite region showing evi- dence of emission from both AGNs and star-forming galaxies. Earlier demarcation lines were empirical, but the Kewley starburst line was modified by the results from population synthesis models and photoionisation models to produce more
31Baldwin et al. [1981] also proposed various other pairs of line ratios; the advantage of the
ones which have become popular is that, as well as being relatively strong, the lines in each ratio are close in wavelength, eliminating the need to estimate extinction, which varies with wavelength.
3.2.2 Separating AGNs from star forming galaxies
Figure 3.5: Left: A BPT diagram from Kewley et al. [2006] showing the Kewley (red) and Kauffmann (black dashed) demarcation lines discussed in the text, plotted over a background of ∼ 85, 000 SDSS DR4 emission-line galaxies. Starburst galaxies form a sequence on the left; the region between the two lines identifies sources with both star-formation and an AGN. Right: an example of a colour-colour diagram separating AGNs (blue) from star-forming galaxies (green); figure from Kirkpatrick et al. [2013].
reliable results (Kewley et al. [2001b]), in particular by placing galaxies with ev- idence of weak AGNs (‘low-ionisation nuclear emission regions’, LINERs) below the starburst line. The demarcation line is based on both the hardness of the radi- ation field and the metallicity of the line-emitting gases, resulting (non-trivially) in a curved boundary. The Kauffmann line is empirical. The two lines are shown in relation to ∼ 85, 000 SDSS emission line galaxies in Figure 3.5 (left). The distinction between star-forming sequence (labelled HII in the figure) and AGNs is shown to be clearly separated by the lines.
The BPT diagram diagnostic was recently confirmed out to z ∼ 1.5 (Juneau et al. [2014]). It has recently been modified to a formula (though still using the same ratios) for higher-redshift galaxies (Kewley et al. [2013]), who also found that the usual diagram is not appropriate at z ≥ 1.5. However, another recent study found that the BPT diagram was effective in distinguishing AGNs from starbursts out to z = 2.3 (Coil et al. [2015]). A variation to the BPT diagram proposed using Hβ, [OIII] and [OII] was proposed in Davies et al. [2014]. Also, a recent paper has suggested a 3-D extension of these ratios (Vogt et al. [2014]).
3.2.2 Separating AGNs from star forming galaxies
Figure 3.6: Left: The Spitzer-IRAC mid-infrared bands wedge diagnostic for AGNs of Donley et al. [2012] compared to classification of 154 galaxies from the 5MUSES sample into AGNs vs. star-forming galaxies using 6.2 µm PAH equivalent widths (Magdis et al. [2013]). Red circles show sources with EW6.2≤ 0.2 µm (suggesting destruction of PAHs by an AGN) and black circles show sources with EW6.2 > 0.2 µm. Grey outer circles show sources with a further test: their SEDs suggest >50% LFIRis due to AGN activity. Right: Line ratio diagnostics for sample of 53 ULIRGs (Farrah et al. [2007]): [NeIII] λ15.56 µm / [NeII] λ12.81 µm vs. [SIV] λ10.51 µm / [SIII] λ18.71 µm. Green symbols represent starbursts and red symbols represent AGNs.
An analysis to find the split between SBG- and AGN- powered galaxies in Spitzer 70 µm data in the Extended Groth Strip field found that although ∼ 13% hosted an AGN, virtually all (60 out of a sample of 61) were powered predomi- nantly by starburst activity (Symeonidis et al. [2010]). In this chapter, ∼ 13% of the 192 sources for which the BPT diagram could be used (i.e., at relatively low redshifts) was found to be dominated by AGNs (see Figure 3.15). Mid-infrared selection is expected to find a higher proportion of AGNs: Toba et al. [2013], using AKARI IRC data, found that about 26% of their 9 µm sources and 41% of their 18 µm sources were AGNs. They found most local AGNs to be obscured, but with obscuration decreasing with increasing mid-infrared luminosity.
Evidence from mid-infrared spectroscopy
Mid-infrared spectroscopy has helped to distinguish between AGNs and star- bursts. This region includes PAH features (discussed in Section 1.5) and a variety
3.2.2 Separating AGNs from star forming galaxies
Figure 3.7: A diagnostic from Genzel et al. [1998], showing the [OIV] λ25.9 µm /
[NeII] λ12.8 µm ratio plotted against the relative strength of the 7.7 µm PAH feature. Known starbursts are shown as triangles, known AGNs are squares with crosses, and ULIRGs are filled circles.
of emission lines from ionised atoms, in particular neon and sulphur, in addition to some molecules. The ratios between emission lines of atoms at different ion- isation levels indicate the hardness of the radiation (i.e. the likelihood of AGN activity), in particular ratios involving [SIV] λ10.51 µm, [NeII] λ12.81 µm, [NeV] λ14.32 µm and [NeIII] λ15.56 µm (e.g. Madden et al. [2006]; Farrah et al. [2007]; Magdis et al. [2013]; see Figure 3.6, right).
The first extensive work in this region used spectroscopy from the ISO (e.g. Genzel et al. [1998]; Lutz et al. [1998]; Sturm et al. [2000]; see Figure 3.7). Many features were found, some previously unobserved; however, these emission features were largely absent in the hard radiation field close to an AGN, and also in intensely star-forming regions. Using an alternative diagnostic diagram for mid-infrared spectra and with combined ISO and Spitzer data, Spoon et al. [2007] has shown that most ULIRGs have simultaneous AGN and starburst activity in their nuclei.
The Great Observatories All-Sky LIRG Survey (GOALS; Armus et al. [2009]) obtained multi-wavelength data to study 202 local (z < 0.088) LIRGs (1011L
3.2.2 Separating AGNs from star forming galaxies
Figure 3.8: Spitzer-IRS mid-infrared spectra of representative starburst-dominated
(top) and AGN-dominated (bottom) GOALS sources (figure from Inami et al. [2013]). Note the higher [NeV]/[NeII] line flux ratio and the absence of the PAH features (in particular at 11.3 µm) in the AGN-dominated sources.
LIR < 1012L⊙) which included both AGNs and starbursts identified by optical
line-ratio diagrams. It was found that the energy from LIRGs was dominated by starbursts. Petric et al. [2011], using Spitzer mid-infrared spectroscopy, found that ∼12% of the total energy emitted by the LIRGs came from AGNs; that 18% contained an AGN, and that in 10% AGN emission was the dominant energy source. Inami et al. [2013], also using the Spitzer mid-infrared spectroscopy, detected a variety of emission lines as well as PAH features (see Figure 3.8). They found that more than 75% of the galaxies were dominated by starbursts based on the [NeV]/[NeII] line flux ratios and the width of the PAH features. Far- infrared line ratios have been used with observations by Herschel (e.g. Rigopoulou et al. [2014]; Farrah et al. [2013]).
Other colour diagrams
Other colour-colour plots, using photometric data rather than relative emis- sion line intensities, may be used to select or segregate types of galaxy popu-
3.2.2 Separating AGNs from star forming galaxies
lations. The success of this technique rests on the fact that different types of galaxies have differently shaped SEDs. Using IRAS fluxes, Helou [1986] found that for star-forming galaxies, the flux ratio 12 µm / 25 µm decreased as the flux ratio 60 µm / 100 µm increased. Soifer et al. [1987a] used a plot of 25 µm / 60 µm against 60 µm / 100 µm to separate starburst galaxies from quasars, BL Lacs and Seyferts. Star formation was characterised by warm 100 µm / 60 µm colours and AGNs by high 25 µm / 60µm colours (as would be expected from their SEDs).
In the mid-infrared the existence, equivalent widths and relative strengths of the various PAH bands can help to distinguish between AGNs and star- forming galaxies. Since these are bands rather than narrow lines, photometry from Spitzer-IRAC can be used since its four bands coincide to varying degrees with the PAH features (e.g. Laurent et al. [2000]; Peeters et al. [2004]; Donley et al. [2012]; Magdis et al. [2013]). For example, plotting 8.0 µm / 4.5 µm against 5.8 µm / 3.6 µm provides a successful diagnostic (see Figure 3.6, left).
Kirkpatrick et al. [2013] used a sample of 151 galaxies at 0.5 < z < 4 with Spitzer spectroscopy and Herschel PACS and SPIRE data to evaluate the reliabil- ity of various colour plots in separating AGNs from star-forming galaxies. They found the best plots, with only ∼10% contamination rate, were (a) 250 µm / 24 µm against 8 µm / 3.6 µm (see Figure 3.5, right) and (b) 100 µm / 24 µm against 8 µm / 3.6 µm. Applying these diagnostics to the entire GOODS-Herschel survey, they found ∼10% of the galaxies have a significant AGN, rising to �50% for the brightest sources. Huang et al. [2007] using an 8 µm selected local sample of Spitzer sources found that dust emission (mostly PAH) accounted for 80% of 8 µm luminosity, with AGN emission accounting for only 1%.
The IR8 ratio is also used to provide a diagnostic for AGN / SFG differ- entiation since PAHs are destroyed by hard radiation (this ratio was discussed in Section 2.7). Other diagnostic diagrams including colour-excitation diagrams (CEx; Yan et al. [2011]) and mass-excitation diagrams (MEx; Juneau et al. [2011]) have been proposed for cases where Hα and [NII] emission lines used in the BPT diagram are shifted out of the optical observing range (at about z > 0.35) or where all four lines are outside the optical range (z � 1). These have not been used in this work.
3.3.1 AAOmega choice of targets