이학석사학위논문 The Broad Line Region Gas Metallicity of the Palomar Green Quasars Palomar Green 퀘이사의 넓은방출선폭지역의가스금속성 2013 년 2 월 서울대학교대학원 물리 천문학부천문학전공 신재진

Transcription

1 저작자표시 - 비영리 - 변경금지 2.0 대한민국 이용자는아래의조건을따르는경우에한하여자유롭게 이저작물을복제, 배포, 전송, 전시, 공연및방송할수있습니다. 다음과같은조건을따라야합니다 : 저작자표시. 귀하는원저작자를표시하여야합니다. 비영리. 귀하는이저작물을영리목적으로이용할수없습니다. 변경금지. 귀하는이저작물을개작, 변형또는가공할수없습니다. 귀하는, 이저작물의재이용이나배포의경우, 이저작물에적용된이용허락조건을명확하게나타내어야합니다. 저작권자로부터별도의허가를받으면이러한조건들은적용되지않습니다. 저작권법에따른이용자의권리는위의내용에의하여영향을받지않습니다. 이것은이용허락규약 (Legal Code) 을이해하기쉽게요약한것입니다. Disclaimer

2 이학석사학위논문 The Broad Line Region Gas Metallicity of the Palomar Green Quasars Palomar Green 퀘이사의 넓은방출선폭지역의가스금속성 2013 년 2 월 서울대학교대학원 물리 천문학부천문학전공 신재진

3 The Broad Line Region Gas Metallicity of the Palomar Green Quasars A thesis submitted in partial fulfillment of the final requirement for the degree of Master of Science Astronomy Program Department of Physics and Astronomy Seoul National University Candidate: Supervisor: Jaejin Shin Jong-Hak Woo December 2012

4 This thesis was originally published in The Astrophysical Journal in January 2013 (Shin et al. ApJ 2013, 763, 58) and is revised here.

5 Abstract We investigate the chemical properties of low-z QSOs, using archival UV spectra obtained with the HST and IUE for a sample of 70 Palomar-Green QSOs at z < 0.5. By utilizing the flux ratio of UV emission lines (i.e., N v /C iv, (Si iv+o iv])/c iv, and N v/he ii) as metallicity indicators, we compare broad-line region (BLR) gas metallicity with AGN properties, i.e., black hole mass, luminosity, and Eddington ratio. We find that BLR metallicity correlates with Eddington ratio while the dependency on black hole mass is much weaker. Although these trends of low-z AGNs appear to be different from those of high-z QSOs, the difference between low-z and high-z samples is partly caused by the limited dynamical range of the samples. We find that metal enrichment at the center of galaxies is closely connected to the accretion activity of black holes and that the scatter of metallicity correlations with black hole mass increases over cosmic time. Keywords: galaxies: active galaxies: nuclei galaxies: abundances galaxies: evolution quasars: emission lines Student Number: i

6 Contents Abstract i 1 Introduction 1 2 Sample Selection and the Data Sample selection Data AGN properties Analysis Multi-component fitting procedure Fitting result Result Comparison of emission-line fluxes Comparison among metallicity indicators Comparison between metallicity and AGN properties Discussion 37 6 Summary & Conclusion 42 ii

7 Chapter 1 Introduction Measuring chemical properties of galaxies and their redshift evolution is a crucial step in understanding galaxy evolution since the metallicity of galaxies is closely related to the history of star-formation, gas inflow, and outflow. A number of observational studies have been devoted to measure the metallicity of galaxies, revealing metallicity correlations with various galaxy properties. In the local universe, it has been shown that metallicity scales with galaxy luminosity and mass (the luminosity-metallicity relation and mass-metallicity relation, respectively) based on the metallicity measured from gas emission lines (Tremonti et al. 2004, and references therein) or stellar absorption lines (e.g., Gallazzi et al. 2005; Panter et al. 2008). These scaling relations indicate that metal enrichment is closely connected to the galaxy mass assembly. Recently the redshift evolution of these scaling relations in star-forming galaxies has been extensively investigated. At z < 3, many studies have suggested apparent metallicity evolution as a function of redshift. The evolution is significant especially for low stellar-mass galaxies (e.q., Savaglio et al. 2005; Erb et al. 2006; Maiolino et al. 2008; Mannucci et al. 2009; Yabe et al. 2012), though it could be due to observational selection effects (see Mannucci et al. 2010). At z > 3, exploring the metallicity of galaxies is extremely challenging, because typical galaxies at such high redshifts are 1

8 Chapter 1. Introduction 2 very faint and the classical metallicity indicators in the rest-frame optical spectra shift out of the atmospheric windows (but see also, e.g., Laskar et al. 2011; Nagao et al. 2012). To extend the metallicity measurement toward higher redshifts, one possible approach is to focus on active galactic nuclei (AGNs). Thanks to their high luminosity (L AGN ) and a various metallic emission lines in their rest-frame ultraviolet spectra, it is possible to infer the metallicity of broad-line regions (BLRs) even for QSOs at z 6 7 with ground-based telescopes (Kurk et al. 2007; Juarez et al. 2009; see also Mortlock et al. 2011). Previously a positive relation between the metallicity of BLRs (Z BLR ) and the redshift of AGNs has been reported (e.g., Hamann & Ferland 1992, 1993); however it turned out that the apparent relation was caused by a selection bias and that the correlation between AGN luminosity and BLR metallicity (the L AGN Z BLR relation) was fundamental (see, e.g., Nagao et al. 2006b). Note that the luminosity-metallicity relation of AGNs has been also reported based on the emission lines in the narrow-line region (NLR), which is much more extended than BLR and traces the chemical properties in the spatial scale of AGN host galaxies (Nagao et al. 2006a; Matsuoka et al. 2009). Interestingly, the luminosity-metallicity relation of high-luminosity QSOs shows no strong redshift evolution in the redshift range of 2 < z < 6 (Nagao et al. 2006a,b; Juarez et al. 2009; Matsuoka et al. 2011b), implying that the chemical evolution at the center of host galaxies is mostly completed at a very high redshift (see also Matsuoka et al. 2011a). Although the observed L AGN Z BLR relation in AGNs and its redshift dependence are crucial to constrain evolutionary scenarios of the supermassive black hole (BH), the host galaxy, and the interplay between these two (i.e., the galaxy-bh coevolution), there are two main drawbacks that should be resolved. First, while the L AGN Z BLR relation is well established, its origin is still controversial. For example, Warner et al. (2004) reported that the metallicity of BLR showed a correlation with the mass of

9 Chapter 1. Introduction 3 BH (M BH ), but no correlation with the Eddington ratio (L/L Edd ). Their result was recently confirmed by a larger sample of QSOs (Matsuoka et al. 2011b). On the other hand, Shemmer et al. (2004) claimed that the observed L AGN Z BLR relation was caused by the dependence of Z BLR on the Eddington ratio, not on M BH (see also Dietrich et al. 2009). At lower redshifts, it has been reported that narrow-line Seyfert 1 galaxies (NLS1s, whose Eddington ratios are believed to be high; see, e.g., Boroson 2002; Grupe 2004) show higher metallicity than typical broad-line AGNs (e.g., Wills et al. 1999; Nagao et al. 2002; Shemmer & Netzer 2002). The higher metallicity of NLS1s in these studies is qualitatively consistent with the result reported by Shemmer et al. (2004). The other drawback comes from observational limitations. The AGN metallicity based on both BLRs and NLRs has been examined predominantly for AGNs at z > 2, since the AGN metallicity studies generally utilizes emission lines in the rest-frame ultraviolet spectra. Consequently, the observational studies with ground-based telescopes are limited to AGNs at z > 2. This prevents us from studying the difference in the chemical properties of BLRs between high-z and low-z QSOs. For instance, Nagao et al. (2006b) studied the L AGN Z BLR relation of QSOs at 2.0 z 4.5. The possible physical origin of this L AGN Z BLR relation has been examined by Shemmer et al. (2004) for 2.0 < z < 3.5, and by Matsuoka et al. (2011b) for 2.3 < z < 3.0. Although Warner et al. (2004) investigate this issue for AGNs in a wide redshift range of 0 < z < 5, they did not examine the redshift dependence of BLR chemical properties. Although there are a few attempts to infer the NLR metallicity based on the rest-frame optical spectra (e.q., Storchi-Bergmann et al. 1998; Nagao et al. 2002; Groves et al. 2006), those methods are in turn difficult to apply for high-z AGNs since infrared spectroscopy is required to obtain rest-frame optical emission lines, thus inconvenient for the comparative study between high and low redshifts. Using space observations with the IUE and the HST, Shemmer & Netzer (2002) investigated UV spectra of low-z AGNs and showed that a

10 Chapter 1. Introduction 4 significant L AGN Z BLR relation was present also in low-z AGNs. However, they did not examine the physical origin of the L AGN Z BLR relation and thus it is not clear whether BH mass or the accretion rate derives the observed L AGN Z BLR relation, and whether the low-z and high-z AGNs show the same correlations. Motivated by these considerations, in this paper we investigate the correlation between BLR metallicity with various AGN properties, including mass, luminosity, and Eddington ratio, for a sample of 70 low-z Palomar-Green (PG) QSO at z < 0.5, by utilizing the archival UV spectra. We describe the sample selection and the data in 2, the data analysis and the fitting procedure in 3. The main results are presented in 4, followed by discussion in 5 and summary and conclusions in 6. We adopt a cosmology of H 0 = 70 km s 1 Mpc 1, Ω Λ = 0.7 and Ω m = 0.3.

11 Chapter 2 Sample Selection and the Data 2.1 Sample selection To investigate metallicity of high-z QSOs, both permitted and weak semi-forbidden lines have been used. The weak semi-forbidden emission lines in the rest-frame UV spectra, i.e., N iv]λ1486, O iii]λ1663, and N iii]λ1749 are good metallicity indicators (e.g., Shields 1976; Baldwin & Netzer 1978; Osmer 1980; Uomoto 1984; Warner et al. 2002), since the flux ratio among these lines do not show strong dependences on physical properties of gas clouds (such as the density and ionization parameter). However, the actual application to observational data is generally difficult since those semi-forbidden emission lines are too faint to be measured accurately. Thus, stronger emission lines are preferred in the studies of Z BLR. For example, the flux ratios of N vλ1240 to C ivλ1549 and N vλ1240 to He iiλ1640 have been utilized to infer Z BLR by comparing them with the prediction of photoionization models (e.q., Hamann & Ferland 1992, 1993; Ferland et al. 1996; Korista et al. 1998; Dietrich et al. 1999; Dietrich & Wilhelm-Erkens 2000; Hamann et al. 2002; Dietrich et al. 2003). Nagao et al. (2006b) showed that the flux ratios of (Si ivλ1397+o iv]λ1402)/c ivλ1549 and Al iiiλ1857/c ivλ1549 are also useful to infer Z BLR through photoionization model runs, that are used for inferring Z BLR in 5

12 Chapter 2. Sample Selection and the Data 6 high-z QSOs (e.g., Juarez et al. 2009; Matsuoka et al. 2011b). Since most previous studies used the rest-frame UV spectra to infer BLR metallicity of high-redshift Type-1 QSOs, we therefore focus on high-luminosity QSOs at low-redshfit, for which UV spectra are available, in order to investigate the chemical properties of low-z QSOs compared to high-z QSOs. The Palomar-Green (PG) QSOs (Schmidt & Green 1983) are well-studied low-z luminous Type-1 AGNs, and the UV spectra of many PG QSOs have been previously obtained with space facilities, thus suitable for our BLR metallicity study. We selected all PG QSOs at z < 0.5 (89 objects), for which reliable black hole masses are available from either reverberation mapping results (Peterson et al. 2004; Denney et al. 2010) or single-epoch estimates (Vestergaard & Peterson 2006). To investigate BLR metallicity, we will use the flux ratios of N vλ1240/c ivλ1549, N vλ1240/he iiλ1640, and (Si ivλ1397+oiv]λ1402)/c ivλ1549, since these emission lines have relatively large equivalent widths. Thus, we searched for available UV spectra previously obtained from space facilities, using the Mikulski Archive for Space Telescopes (MAST). Among 89 PG QSOs at z < 0.5, the archival UV spectra covering the required emission lines were available for 86 objects. Among them, we excluded 7 broad absorption line (BAL) QSOs, since the strong absorption features prevent us from measuring the emission-line fluxes accurately. We also excluded 9 additional objects, for which the spectral quality is too low to identify the aforementioned emission lines. Thus, we finalized a sample of 70 PG QSOs for this work as listed in Table

13 Chapter 2. Sample Selection and the Data 7 Table 2.1: Log of archival UV data & AGN properties Object Redshift Observation Date Telescope/Instrument log[mbh/m ] Ref. log[lbol/ergs 1 ] S/N (1) (2) (3) (4) (5) (6) (7) (8) PG Nov 05,07 HST/FOS 9.25 ± ± PG Feb 08 HST/COS 7.13 ± ± PG Jun 08 IUE/SWP 8.71 ± ± PG Nov 27 HST/FOS 8.57 ± ± PG Jul 31 IUE/SWP 8.33 ± ± PG Dec 22,23 IUE/SWP 7.42 ± ± PG Jun 29 IUE/SWP 8.55 ± ± PG Aug 09 IUE/SWP 8.15 ± ± PG Jun 12 HST/COS 8.82 ± ± PG Sep 24 HST/COS 8.13 ± ± PG Nov 30;Dec 01 IUE/SWP 7.95 ± ± PG Feb 28,29 IUE/SWP 7.38 ± ± PG May 01 IUE/SWP 8.58 ± ± PG May 06 HST/FOS 8.66 ± ± PG Mar 26 HST/COS 7.30 ± ± PG Apr 10, IUE/SWP 8.23 ± ± PG May 31;Jun 01 IUE/SWP 6.32 ± ± PG Nov 13 IUE/SWP 8.35 ± ± PG Apr 01, HST/FOS 9.16 ± ± PG Dec 29 HST/FOS 9.30 ± ± Col. (1): Target ID. Col. (2): Redshift. Col. (3): Observed date. Col. (4): Telescope and Instrument. Col. (5): Black hole mass from Peterson et al. (2004); Vestergaard & Peterson (2006); Denney et al. (2010) with a new virial factor (Woo et al. 2010). Col. (6): References for redshift and black hole mass. 1 - Reverberation-mapped AGNs (Peterson et al. 2004), 1* - Reverberation-mapped AGNs (Denney et al. 2010), 2 - AGNs with single-epoch black hole mass (Vestergaard & Peterson 2006). Col. (7): AGN bolometric luminosity calculated from the monochromatic luminosity at 1350Å by multiplying a bolometric correction factor, Col. (8): Signal-to-noise ratio per resolution element at 1350Å in the rest-frame.

14 Chapter 2. Sample Selection and the Data 8 Table 2.2: Log of archival UV data & AGN properties Object Redshift Observation Date Telescope/Instrument log[mbh/m ] Ref. log[lbol/ergs 1 ] S/N (1) (2) (3) (4) (5) (6) (7) (8) PG May 19 HST/FOS 7.65 ± ± PG Feb 19,20 HST/FOS 8.51 ± ± PG Nov 21,26 IUE/SWP 7.45 ± ± PG Jan 08 IUE/SWP 8.01 ± ± PG Dec 29 IUE/SWP 7.90 ± ± PG Jan 29,30 IUE/SWP 8.53 ± ± PG Jul 21 HST/FOS 8.59 ± ± PG Feb 04,07 HST/STIS 7.94 ± ± PG Mar 15 HST/FOS 9.18 ± ± PG Jul 9 HST/FOS 8.93 ± ± PG May 01,02 IUE/SWP 7.84 ± ± PG Feb 08 IUE/SWP 6.50 ± ± PG Dec 27 HST/FOS 8.90 ± ± PG Jul 25,26 IUE/SWP, LWP 8.86 ± ± PG May 04 IUE/SWP 8.62 ± ± PG Feb 11 IUE/SWP 7.86 ± ± PG Jan 19 HST/FOS 8.26 ± ± PG Mar 22 IUE/SWP 8.02 ± ± PG Jun 27 HST/COS 7.52 ± ± PG May26 HST/FOS 8.40 ± ± PG Aug 25 HST/FOS 7.92 ± ± PG Feb 23 HST/FOS 6.87 ± ± PG Jan 02 HST/FOS 7.99 ± ± PG Mar 03 IUE/SWP 9.02 ± ± PG Jun29 HST/FOS 9.71 ± ± Table 2.1 continued

15 Chapter 2. Sample Selection and the Data 9 Table 2.3: Log of archival UV data & AGN properties Object Redshift Observation Date Telescope/Instrument log[mbh/m ] Ref. log[lbol/ergs 1 ] S/N (1) (2) (3) (4) (5) (6) (7) (8) PG Jul 27, 28, 29 HST/STIS 9.09 ± ± PG Jan 07 HST/FOS 8.07 ± ± PG Aug 04 HST/COS 7.77 ± * ± PG Jun 12 IUE/SWP 8.34 ± ± PG Dec 05 HST/FOS 7.45 ± ± PG May 23 HST/FOS 8.27 ± ± PG Jun 18 HST/COS 6.95 ± ± PG Jun 30; Jul 02 IUE/SWP 8.50 ± ± PG Jan 26 HST/FOS 9.35 ± ± PG Jun 11 IUE/SWP 7.92 ± ± PG Oct 28 HST/COS 7.37 ± * ± PG Mar 14 HST/FOS 7.98 ± ± PG Apr 08,10 HST/FOS 9.29 ± ± PG Apr 28 IUE/SWP 7.52 ± ± PG Sep 10 IUE/SWP 8.04 ± ± PG Apr 08, 09, 10 HST/COS 8.43 ± ± PG May 13 IUE/SWP 8.75 ± ± PG Nov 19 HST/FOS 8.48 ± ± PG Sep 19 HST/FOS 8.98 ± ± PG Oct 28, HST/COS 8.64 ± ± PG Jun 03 IUE/SWP 8.53 ± ± PG May 13 HST/STIS 8.02 ± ± PG May 01 HST/STIS 8.97 ± ± PG Dec 29 IUE/SWP 8.54 ± ± PG Oct 12 HST/FOS 9.57 ± ± Table 2.1 continued

16 Chapter 2. Sample Selection and the Data Data We obtained the UV spectra taken with International Ultraviolet Explorer (IU E) or Hubble Space Telescope (HST ) through the MAST database. We collected all available spectra of our targets and used the best quality spectrum with an order of Cosmic Origins Spectrograph (COS), Space Telescope Imaging Spectrograph (ST IS), Faint Object Spectrograph (F OS), and IUE when multiple instruments have been used. In summary, we utilized 10 COS spectra, 4 STIS spectra, 26 for FOS and 30 IUEspectra. Specifically, we used the SWP ( Å) data of IUE, G130H ( Å) and G160H ( Å) data of HST /FOS, and G140M ( Å) data of HST /STIS, for lower-z QSOs. For relatively higher-z QSOs, we used the LWP ( Å) data of IUE and G270H ( Å) data of HST /FOS data. Finally, for the HST /COS data, we used the combined data in two spectral ranges, i.e., G130M ( Å) and G160M ( Å), in order to cover the N v and C iv lines at the same time. We combined the spectra of each exposures by calculating the error-weighted mean. For the STIS data, we combined the spectra using the exposure-time as a weight, because we could not eliminate artificial spark features effectively when we adopted the error-weighted mean. However, the spectra are qualitatively consistent. In the case of COS spectra, we used IDL routines developed by the COS GTO team (Danforth et al. 2010). We smoothed the spectra in the wavelength direction by adopting 7 pixel smoothing for the COS data and 2 pixel smoothing for the STIS data. If a target has been observed at multiple-epochs, we chose only one epoch with the best data quality to avoid any time-variation effects. Table lists the observation data and the instrument for each target. In this table we also list the signal-to-noise ratio per resolution element calculated at the rest-frame 1350Å continuum.

17 Chapter 2. Sample Selection and the Data AGN properties To compare with BLR metallicity, we measure and collect other AGN properties, i.e., black hole mass, bolometric luminosity, and Eddington ratio. We collected black hole mass of the sample QSOs, which has been previously determined by the reverberation mapping studies for 18 objects (Peterson et al. 2004; Denney et al. 2010) or by the single-epoch method for 52 objects (Vestergaard & Peterson 2006). A black hole mass measurement based on reverberation-mapping results is availble for PG (Peterson et al. 2004), however it has large uncertainty due to the low data quality, and it has been excluded in other reverberation sample studies. Thus, we will use single-epoch mass for PG We re-calculated black hole mass of the sample by adopting the updated virial factor of 5.2 (Woo et al. 2010), which is slightly smaller than the previous virial factor (5.5; Onken et al. 2004, see also Park et al. 2012) As the uncertainty of black holes masses, we adopted the values given by Peterson et al. (2004); Vestergaard & Peterson (2006); Denney et al. (2010). For AGN bolometric luminosity (L bol ), we used the obtained UV spectra to measure monochromatic luminosity at 1350Å, which is presumably not heavily contaminated by the host galaxy stellar light. To measure the flux at 1350Å, we fitted the AGN continuum between 1210Å and 1700Å with a power-law function. The measured 1350Å monochromatic continuum luminosity is then used for calculating AGN bolometric luminosity by multiplying a bolometric correction factor, 3.81 (Shen et al. 2008). Note that this bolometric correction factor is the same as adopted by Matsuoka et al. (2011b) for high-z QSOs. The measurement uncertainty of AGN luminosities was calculated based on the signal-to-noise ratio of the spectra. Finally, we determined Eddington ratio, L bol /L Edd (hearafter, L/L Edd ), with following equation.

18 Chapter 2. Sample Selection and the Data 12 L L Edd = L bol M BH (2.1) In Figure 2.1, we present the distribution of the sample properties; the redshift, black hole mass, bolometric luminosity, and Eddington ratio. The black hole mass ranges over 3 orders of magnitude (from 6.32 to 9.71) with an average of 8.26 ± 0.71 M. The bolometric luminosity also ranges over a large range from to erg s 1 while some fraction of the sample has relatively low luminosity and belong to Seyfert class rather than QSOs. The mean Eddington ratio of the sample is 10% with 0.66 dex dispersion, indicating that there is a large range of accretion activity (Woo & Urry 2002).

19 Chapter 2. Sample Selection and the Data 13 Figure 2.1: The distributions of the redshift, BH mass, AGN luminosity, and the Eddington ratio of the sample.

20 Chapter 3 Analysis 3.1 Multi-component fitting procedure The flux measurement of BLR emission lines is an important step for investigating Z BLR. It is known that BLR emission lines sometimes show significant asymmetric velocity profiles (e.q., Corbin 1997; Vanden Berk et al. 2001; Baskin & Laor 2005), hence a single-gaussian model does not generate a reliable fit for such cases. To fit asymmetric velocity profiles of QSO UV emission lines, various models have been adopted. Here, we examined 4 models, namely, double-gaussian, Gauss-Hermitian, modified-lorentzian, and 2 power-law functions to determine the best line profile to use. In Figure 3.1, we present the C iv line of PG as well as 4 different model fits, which show slightly different results, particularly at the wing of the line. Through our visual inspection, we decided to adopt the double-gaussian function as an emission line profile model. Note that our results do not significantly depend on the choice of the model since the difference in flux measurements is 5%. In the rest-frame UV spectra, many AGN emission lines are blended; Lyα λ1216+ N vλ1240, Si ivλ1397+o iv]λ1402, and He iiλ1640+o iii]λ1663+al iiλ1671. Thus, it is necessary to perform a multi-component fitting analysis for secure flux measurements. 14

21 Chapter 3. Analysis 15 Double Gaussian Gauss Hermitian Arbitrary flux Modified Lorentzian Power law Rest frame wavelength (Å) Figure 3.1: Comparison of different fitting functions for the C ivλ1549 emission line of PG Double-Gaussian (upper left), Gauss-Hermitian (upper right), modified- Lorentzian (lower left) and 2 power-law (lower right) functions are examined. The red lines denote the fitting results, and the blue line in the upper left panel represents each Gaussian component. Residual spectrum is shown at the bottom in each panel. Masked regions are indicated with gray hatches.

22 Chapter 3. Analysis 16 We simultaneously fitted all 10 emission lines, that were used as BLR metallicity indicators. First, we divided these emission lines into two groups based on their ionization degree, and assumed that the emission lines in each group have the same velocity profile (see Nagao et al. 2006b). Specifically, we categorized N vλ1240, O iv]λ1402, N iv]λ1486, C ivλ1549 and He iiλ1640) in the high-ionization group, while Si iiλ1263, Si ivλ1397, O iii]λ1663 and Al iiλ1671 in the low-ionization group, following Nagao et al. (2006b). We adopted the same velocity width and the velocity shift for each group. We excluded the spectral range between 1570Å and 1631Å from the fit, since an unidentified emission feature is reported in this range (see, e.g., Wilkes 1984; Boyle 1990; Laor et al. 1994; Nagao et al. 2006b). In the case of Lyα, the absorption by the intergalactic matter (IGM) affects the line profile significantly, particularly below 1210Å. Thus, we treated Lyα separately, by allowing the velocity dispersion and velocity shift to be free. We excluded the spectral range below 1210Å from the fitting procedure, because of the IGM absorption. For the continuum fitting, we used a power-law and determined the slope by using three spectral windows (1345Å 1355Å, 1445Å 1455Å, and 1687Å 1697Å), where no strong emission lines are present. Figure 3.2 shows an example of multi-component fitting for four different spectral regions of PG From upper left to bottom right, 1) Lyα λ1216+n vλ1240, 2) Si ivλ1397+o iv]λ1402, 3) N iv]λ1486+c ivλ1549, and 4) He iiλ1640+o iii]λ1663+al iiλ1671. The line center is denoted with dashed lines. 3.2 Fitting result Based on the multi-component fitting using double-gaussian profiles, we measured the flux of 10 broad emission lines in the rest-frame UV spectra for the sample. In Figure , we present the fitting results of our all targets. In some cases, spectral quality is too low to fit the weak emission lines, i.e., N vλ1240, Si ivλ1397, O iv]λ1402, C ivλ1549, and He iiλ1640, thus we only measure the flux of N v and C iv. The weak

23 Chapter 3. Analysis 17 Lyα N V Si II Si IV O IV Arbitrary flux N IV C IV He II O III Al II Rest frame wavelength (Å) Figure 3.2: An example of Multi-component fitting analysis with PG Each panel shows Lyα λ1216+n vλ1240 (upper left), Si ivλ1397+o iv]λ1402 (upper right), N iv]λ1486+c ivλ1549 (bottom left), and He iiλ1640+o iii]λ1663+al iiλ1671 (bottom right), respectively. The color of lines are the same as in Figure 3.1. The dashed lines indicate the center of each emission line.

24 Chapter 3. Analysis 18 lines (i.e., Si iv+o iv] and He ii+o iii]+al ii) are shown in the inset panels only when these lines were successfully fitted. In summary, we measured the N v and C iv fluxes for the entire sample (70 objects) while we measured the flux of Si iv, O iv], He ii for a subsample of 34 objects. Table lists the measured fluxes and the inferred uncertainties, which were estimated by averaging the signal-to-noise ratio of each pixel within the spectral range of each lines. We list the sum of Si ivλ1397 and O iv]λ1402 fluxes instead of individual flux measurements, since the sum of two lines will be used to compare with the flux of the C iv line.

25 Chapter 3. Analysis 19 PG PG PG PG Arbitrary flux PG PG PG PG Rest frame wavelength (Å) Figure 3.3: Fitting results of all targets. The inset panels show the fit for weak lines, namely, Si iv+o iv] (left) and He ii+o iii]+al ii (right). No inset panel means that fluxes of these weak lines were not measured due to the low signal-to-noise ratio. The dashed lines in the inset panels represent the center of each line as shown in Figure 3.2. The fitting residual is shown in the lower panel for each object. The color of lines and masked regions are the same as in Figure 3.1.

26 Chapter 3. Analysis 20 PG PG PG Arbitrary flux PG PG PG PG PG Rest frame wavelength (Å) Figure 3.4: Continued 1600

27 Chapter 3. Analysis 21 PG PG PG PG Arbitrary flux PG PG PG PG Rest frame wavelength (Å) Figure 3.5: Continued

28 Chapter 3. Analysis 22 PG PG PG PG Arbitrary flux PG PG PG PG Rest frame wavelength (Å) Figure 3.6: Continued

29 Chapter 3. Analysis 23 PG PG PG PG Arbitrary flux PG PG PG PG Rest frame wavelength (Å) Figure 3.7: Continued

30 Chapter 3. Analysis 24 PG PG PG Arbitrary flux PG PG PG PG PG Rest frame wavelength (Å) Figure 3.8: Continued 1600

31 Chapter 3. Analysis 25 PG PG PG Arbitrary flux PG PG PG PG PG Rest frame wavelength (Å) Figure 3.9: Continued 1600

32 Chapter 3. Analysis 26 PG PG PG PG Arbitrary flux PG PG PG PG Rest frame wavelength (Å) Figure 3.10: Continued

33 Chapter 3. Analysis 27 PG PG PG Arbitrary flux PG PG PG Rest frame wavelength (Å) Figure 3.11: Continued

34 Chapter 3. Analysis 28 Table 3.1: Measurements of emission Line fluxes Object N V Si IV+O IV] C IV He II (1) (2) (3) (4) (5) PG ± ± ± ± 1.2 PG ± ± ± ± 1.9 PG ± ± 6.0 PG ± ± ± 0.3 PG ± ± 30.8 PG ± ± ± 2.6 PG ± ± ± 0.2 PG ± ± ± 0.2 PG ± ± ± 5.8 PG ± ± ± 2.4 PG ± ± 6.6 PG ± ± ± ± 0.5 PG ± ± ± 1.3 PG ± ± ± ± 0.3 PG ± ± ± ± 0.4 PG ± ± ± 1.6 PG ± ± 4.5 PG ± ± 3.4 PG ± ± ± ± 0.7 PG ± ± ± 0.1 PG ± ± 2.0 PG ± ± ± ± 2.2 PG ± ± ± ± 0.4 PG ± ± 3.5 PG ± ± 7.4 PG ± ± ± 0.5 PG ± ± 5.1 PG ± ± ± 12.1 PG ± ± ± ± 0.9 PG ± ± ± ± 1.4 PG ± ± ± ± 0.3 PG ± ± 1.2 PG ± ± ± ± 0.3 PG ± ± 0.6 Col. (1): Target ID. Col. (2): line flux and error of N v. Col. (3): line flux and error of Si iv+o iv. Col. (4): line flux and error of C iv. Col. (5): line flux and error of He ii.

35 Chapter 3. Analysis 29 Table 3.2: Measurements of emission Line fluxes Object N V Si IV+O IV] C IV He II (1) (2) (3) (4) (5) PG ± ± 4.6 PG ± ± ± 2.4 PG ± ± ± 0.3 PG ± ± 5.3 PG ± ± ± ± 0.9 PG ± ± 2.8 PG ± ± 3.4 PG ± ± 1.6 PG ± ± ± ± 0.3 PG ± ± 4.3 PG ± ± 1.5 PG ± ± 36.8 PG ± ± ± ± 0.3 PG ± ± ± 16.2 PG ± ± 4.1 PG ± ± ± ± 0.7 PG ± ± 2.8 PG ± ± ± ± 0.5 PG ± ± ± 0.2 PG ± ± ± ± 0.8 PG ± ± ± 5.9 PG ± ± ± ± 0.6 PG ± ± ± 0.5 PG ± ± ± 0.4 PG ± ± 4.4 PG ± ± ± 0.1 PG ± ± ± 4.5 PG ± ± 2.1 PG ± ± ± ± 0.3 PG ± ± ± 2.0 PG ± ± ± 0.9 PG ± ± ± 1.5 PG ± ± ± 0.3 PG ± ± ± 1.2 PG ± ± 8.8 PG ± ± ± 4.8 Table 3.1 continued

36 Chapter 4 Result 4.1 Comparison of emission-line fluxes In this study, we use three line flux ratios as metallicity indicators among various diagnostics previously proposed, due to the limited data quality; namely, N vλ1240/c ivλ1549, (Si ivλ1397+o iv]λ1402)/c ivλ1549, and N vλ1240/he iiλ1640 (hereafter N v/c iv, (Si iv+o iv])/c iv, and N v/he ii, respectively). Before using them as metallicity indicators in comparison with AGN properties, here we first examine the correlation among the emission line fluxes. Figure 4.1 (top panel) presents the comparison between the C ivλ1549 and He iiλ1640 fluxes, which are used as a denominator of metallicity diagnostics. As photoionization models predict that the flux ratio of these two lines does not depend on Z BLR (see, e.g., Figure 29 in Nagao et al. 2006b), a clear linear relation with a small scatter (0.20 dex) is present, showing that the flux ratio of these two lines is nearly constant. The constant flux ratio between C iv and He ii suggests that our flux measurements of the weak He ii line are reasonable although the He ii is difficult to fit due to blending with Al ii. In Figure 4.1 (bottom panel) we compare the N vλ1240 flux and the sum of the 30

37 Chapter 4. Result Slope : 0.98 Scatter : 0.20 Log (He II) Log (C IV) 12.0 Slope : 0.94 Scatter : 0.24 Log (Si IV+O IV]) Log (N V) Figure 4.1: Comparison of emission-line fluxes. Top: Comparison between the He iiλ1640 and C ivλ1549 fluxes. Bottom: Comparison between the sum of the Si ivλ1397 and O iv]λ1402 fluxes and N vλ1240 flux. The derived Spearman s rank order correlation coefficients and their statistical significance are , (top) and , (bottom).

38 Chapter 4. Result 32 Si ivλ1397 and O iv]λ1402 fluxes, which are used as a numerator of metallicity indicators. Again a clear positive correlation is presented between them, with a somewhat larger scatter than that shown in the comparison between the C iv and He ii fluxes. This larger scatter is partly caused by the fact that the N v flux depends on Z BLR as well as the relative abundance of N v (see, e.g., Matsuoka et al. 2011b; Araki et al. 2012), while the sum of the Si iv and O iv] fluxes mainly depends on Z BLR. 4.2 Comparison among metallicity indicators We compare 3 line flux ratios, namely, N v/c iv, (Si iv+o iv])/c iv, and N v/he ii, as metallicity indicators adopted in this study. As shown in Figure 4.2, the flux ratios of N v/c iv and N v/he ii present a clear correlation with a relatively small scatter, reflecting the constant flux ratio between C iv and He ii (see Figure 4.1). The comparison between (Si iv+o iv])/c iv and N v/c iv also shows a relatively good correlation although a few outliers dominate the scatter. In the case of (Si iv+o iv])/c iv and N v/he ii, the comparison shows a less clear correlation, probably due to the larger combined uncertainties on the flux measurements of weak lines (Si iv+o iv]] and He ii).

39 Chapter 4. Result Slope : 0.73 Scatter : Log (N V/He II) 1.0 Slope : 0.71 Scatter : Log ((Si IV+O IV])/C IV) 1.0 Slope : 0.44 Scatter : 0.19 Log ((Si IV+O IV])/C IV) Log (N V/C IV) Log (N V/C IV) Log (N V/He II) Figure 4.2: The relation among three metallicity indicators. Red line represents error weighted linear fit to the data. The slopes of the linear fit and the data dispersion are shown at the upper-left corner in each panel. The derived Spearman s rank order correlation coefficients and their statistical significance are , (top), , (middle) and , (bottom).

40 Chapter 4. Result Comparison between metallicity and AGN properties In this section, we investigate the correlation between BLR metallicity inferred from the emission line ratios and AGN properties, i.e., black hole mass, luminosity, and Eddington ratio, using the selected low-z QSOs. Figure 4.3 compares 3 metallicity indicators (N v/c iv, (Si iv+o iv])/c iv, and N v/he ii) with AGN properties (M BH, L bol, and L/L Edd ), respectively. A positive correlation is presented between N v/c iv or N v/he ii with AGN luminosity, indicating a luminosity dependence of the BLR metallicity, while it is less certain in the case of (Si iv+o iv])/c iv, probably due to the larger measurement uncertainty in Si iv+o iv] flux. The apparent luminosity - BLR metallicity correlation of low redshift QSOs is similar to the results from previous studies based on high redshift AGNs (Shemmer et al. 2004; Warner et al. 2004; Nagao et al. 2006b). We investigate which parameter between M BH and L bol /L Edd is more fundamental in driving the observed L AGN Z BLR relation. As for M BH, there is no obvious M BH dependence on N v/c iv and (Si iv+o iv])/c iv while there is a possible positive correlation between N v/he ii and M BH. These behaviors are in contrast to previous results obtained at high redshifts, where significant positive correlations were reported between the metallicity indicators and M BH (Warner et al. 2004; Matsuoka et al. 2011b). On the contrary, clear positive dependences on L bol /L Edd are present in N v/c iv and N v/he ii, while it is less certain in the case of (Si iv+o iv])/c iv, due to the lack of objects in the range of L bol /L Edd < 1.5. For these low Eddington ratio objects, Si iv+o iv] lines are very weak and no secure measurements are available. In order to examine the statistical significance of these possible correlations, we applied the Spearman s rank-order test to the data. The derived Spearman s rankorder correlation coefficients (r S ) and their statistical significance (p), which is the probability of the data being consistent with the null hypothesis that the flux ratio is not correlated with an AGN parameter, are given in Table 4.1. The rank-order tests

41 Chapter 4. Result 35 Figure 4.3: The relation between metallicity indicators and AGN properties (M BH, L bol, and L/L Edd ). The symbols and colors represent the Eddington ratio or mass bin as indicated in the top panels.

42 Chapter 4. Result 36 suggest that there are statistically significant positive correlations of N v/c iv with L bol and L bol /L Edd, and a positive correlation of N v/he ii with L bol. In contrast, all three diagnostic flux ratios show no statistically significant correlation with M BH. Table 4.1: Results of Spearman s rank-order correlation test Flux ratio M BH L bol L/L Edd (1) (2) (3) (4) N v/c iv r S = r S = r S = p = p = p = (Si iv+o iv])/c iv r S = r S = r S = p = p = p = N v/he ii r S = r S = r S = p = p = p =

43 Chapter 5 Discussion We compare our results obtained for AGNs at z < 0.5 with the previous results obtained for high-z QSOs (z 2.5; Matsuoka et al. 2011b), in order to investigate possible redshift evolution and the origin of the metallicity scaling relations in AGNs. In Figure 5.1, we overplot the metallicity indicators as functions of M BH and L/L Edd for high-z QSOs adopted from (Matsuoka et al. 2011b), along with the measurements of low-z QSOs. The low-z objects are more dispersed than high-z objects, partly because of the larger measurement uncertainties of the low-z objects. Note that the emission line fluxes were measured for individual objects in the low-z sample, while the composite spectra of high-z QSOs in each black hole mass and Eddington ratio bin were used for the emission line flux measurements (see Matsuoka et al. 2011b). Interestingly, the low-z and high-z samples appear to show different trends in Figure 5.1 (top panels). As mentioned in 4.3, the low-z sample shows only marginal correlations between metallicity indicators and M BH, while the high-z sample clearly shows positive correlations of metallicity indicators with M BH. On the other hand, the metallicity indicators of the low-z sample show much stronger positive correlations with L/L Edd than with M BH, while the correlations between metallicity indicators and L/L Edd are less evident in the high-z sample. 37

44 Chapter 5. Discussion 38 Figure 5.1: Top: Metallicity indicator (N v/c iv flux ratio) as functions of M BH (left) and L/L Edd (right). Low-z QSOs are represented with various symbols while high-z QSOs are denoted by solid lines (Matsuoka et al. 2011b). The colors represent different mass and Eddington ratios as indicated in the upper panels. Bottom: Metallicity indicator vs. M BH (left) and L/L Edd (right), after excluding low M BH and low Eddington ratio AGNs from low-z sample for proper comparison.

45 Chapter 5. Discussion 39 The difference in the trend with Eddington ratios between low-z and high-z samples is partly caused by the much wider range of L/L Edd ( 3 < log[l/l Edd ] < 0) covered by the low-z sample than that covered by the high-z sample ( 1 < log[l/l Edd ] < 0). As shown in Figure 5.1, low-z AGNs with log[l/l Edd ] < 1.5 (that is not covered in the high-z sample) show systematically lower N v/c iv flux ratios, leading to the more evident dependence on L/L Edd. Therefore, our results do not necessarily suggest that the L/L Edd dependence on the metallicity indicators is systematically different between low-z and high-z QSOs. For proper comparison, low Eddington ratio AGNs (log[l/l Edd ] < 1.5) are required at high-z. These results imply that the accretion activity of black holes are closely related with metal enrichment at the central part of host galaxies. The dependence of metallicity indicators on M BH appears to be different between the low-z and high-z samples. At high-z, more massive AGNs have higher metallicity although the M BH range is small (8.5 < log M BH < 10). In contrast, the low-z sample shows much larger scatter without clear trend between M BH and metallicity indicators. However, the large scatter in the low-z sample is partly caused by the low Eddington ratio AGNs, which have systematically lower N v/c iv flux ratios, as described as the metallicity-eddington ratio relation. However, if AGNs with similar Eddington ratios are selected (e.g., same color objects in Figure 5.1), then there appears to be a weak trend of metallicity with M BH, suggesting that at low-z, Z BLR has weak dependency on M BH at fixed Eddington ratios. For high-z QSOs, the tightness of the observed Z BLR M BH correlation is probably caused by the limited range of the Eddington ratios since only high Eddington QSOs were included in the high-z sample. However, inclusion of low Eddington ratio AGNs will presumably weaken the correlation as in the case of the low-z sample. As a consistency check, we match the ranges of M BH and Eddington ratio between high-z and low-z samples, by excluding low M BH and low Eddington ratio objects from the low-z

46 Chapter 5. Discussion 40 sample as presented in the bottom panels of Figure 5.1. As expected, both high-z and low-z samples in the matched dynamical range show similar metallicity dependency on both M BH and Eddington ratio although the M BH -Z BLR relation is much weaker in low-z than at in high-z. Our results imply that the BLR metallicity of low-z AGNs mainly depends on the Eddington ratio, and weakly depends on M BH. Currently, it is unknown how Z BLR scales with average gas metallicity of host galaxies. Nevertheless, the observed metallicity dependency on M BH may imply that there is connection among BH growth, gas enrichment, and galaxy evolution. Assuming that Z BLR correlates with gas metallicity of host galaxies, we try to understand the observed relations with several scenarios. The correlation between M BH and Z BLR at high-z can be interpreted as a consequence of the combination of the galaxy mass-metallicity relation and the M BH M bulge relation (Warner et al. 2003; Matsuoka et al. 2011b). For QSOs at z 2 3 that corresponds to the peak of the quasar activity in the cosmological timescale, it has been often claimed that the major merger triggers the AGN activity through the efficient mass fueling onto black holes (Hasinger 2008; Li et al. 2010). Here the major merger event reduces the angular momentum of gas clouds at the nuclear regions of the quasar host galaxies, providing efficient mass fueling. In this case, the metallicity of accreting gas onto the nucleus may be characterized by the mass-metallicity relation of the host galaxy. Regarding the black hole to host galaxy connection, the M BH M bulge relation has not been observationally defined at highz although several studies indicated that M BH -to-m bulge ratio increases with redshift (e.q., Woo et al. 2006, 2008; Bennert et al. 2010, 2011; Merloni et al. 2010). Thus, if we assume a scaling relation between M BH and galaxy mass, presumably with higher normalization than the local M BH -M bulge relation, then black hole mass scales with both stellar mass and gas metallicity, hence the M BH -Z BLR relation is naturally expected at high-z.

47 Chapter 5. Discussion 41 The weaker correlation of M BH with Z BLR at low-z can be interpreted as combination of two additional effects. First, gas metallicity increased from high-z to low-z by further star formation after major BH growth. In other words, for the same galaxy mass (or black hole mass), metallicity has been increased, particularly at the nuclear region, leading to higher metallicity at fixed M BH compared to high-z objects. For example, nuclear star formation induced by secular process (e.g., bar instability) and galaxy interaction may sufficiently increase metallicity (e.q., Maiolino et al. 2008). Another difference between low-z and high-z is the gas fraction, especially for massive galaxies like quasar host galaxies. The minor merger process is more important as an AGN triggering mechanism at low-z (e.q., Taniguchi 1999; Cisternas et al. 2011). Thus, as a consequence of minor mergers between a gas-poor massive galaxy (i.e., a quasar host galaxy) and a relatively gas-rich less-massive galaxy, the metallicity of accreting gas onto the nucleus can be largely affected by the chemical property of the merging, less-massive galaxy. In this case, the metallicity of accreting gas does not simply reflect the mass-metallicity relation of the host galaxy. Second, the M BH -to-m bulge ratio may change over cosmic time. If black hole mass was higher at fixed stellar mass at high-z, then by combining galaxy mass-metallicity relation and M BH M bulge relation, we may expect similar M BH -metallicity relation in low-z, but with a different normalization. In other words, at fixed metallicity, M BH is lower in low-z than in high-z. Although it is beyond the scope of the current work to quantify and compare these two effects, it is reasonable to conclude that coupling between black hole mass with gas metallicity becomes much weaker with decreasing redshift.

48 Chapter 6 Summary & Conclusion To investigate the chemical properties of low-z QSOs, we measured the flux ratios of the rest-frame UV emission lines as metallicity indicators using a sample of 70 low-z PG QSOs at z < 0.5. By comparing BLR gas metallicity with black hole mass, luminosity and Eddington ratio, we find that Z BLR correlates with Eddington ratio while Z BLR shows much weaker correlation with M BH, indicating that the metal enrichment at the central part of host galaxies is closely connected to the accretion activity of AGN. These trends are different from high-z QSOs, which shows a tighter correlation between Z BLR and M BH and a weaker correlation between Z BLR and Eddington ratio. The apparent difference between low-z and high-z samples seems to be caused by the limited dynamical range in the high-z sample. Various star formation mechanism can increase BLR gas metallicity the cosmic time, increasing the scatter in the metallicity correlation with properties of AGN in low-z. 42