3 Emission-line properties of MG 2016+112 B and C

We now describe the observed properties of the emission lines of image B and C. The velocity width (FWHM) corrected for instrumental resolution and the relative flux values measured by Gaussian fitting procedure are listed in Tables 1 and 2, respectively. The flux values are normalized to CIV lines. We corrected for the reddening by the Galaxy using the value of the extinction in this region, A_V=0.67 and A_I=0.36, derived by Benítez et al. (1999). No correction for the internal reddening was applied since the rest-frame wavelength of the emission lines concerned here are rather close. We treat the unresolved doublet lines as a single line.

   

Table 1: Line width$^{\rm a}$

	MG 2016+112 A$^{\rm b}$	MG 2016+112 B	MG 2016+112 C	F10214$^{\rm c}$	SMM 02399$^{\rm d}$
Ly$\alpha$	<1000	460	630	900	1850
N V	<1600	880	840	1700	1790
C IV	<1100	580	580	1200	1560
He II	<1200	460	440	1150	2800
C III]	--	750	1100		7700
C III] narrow	--	360	380	1000
C III] broad	--	1750	2550	3700

(a) FWHM in km s^-1.
(b) From Lawrence et al. (1984).
(c) From Searjeant et al. (1998).
(d) From Ivison et al. (1998).

   

Table 2: Relative flux of the emission lines$^{\rm a}$

	MG 2016+112 B	MG 2016+112 C	F10214$^{\rm b}$	SMM 02399$^{\rm c}$	HzPRG$^{\rm d}$	Quasar BLR$^{\rm d}$	Seyfert 2$^{\rm d}$
Ly$\alpha$	2.73 (0.09)	3.81 (0.43)	0.53	3.39	2.17	8.52	4.52
N V	0.66 (0.07)	1.19 (0.48)	0.74	2.20	0.57	0.42
C IV	1.00	1.00	1.00	1.00	1.00	1.00	1.00
He II	0.28 (0.03)	0.55 (0.10)	0.53	0.36	0.12	0.87	0.17
C III]	0.40 (0.04)	1.32 (0.27)		2.70	0.45	0.49	0.46
C III] narrow	0.17 (0.02)	0.51 (0.10)	0.34
C III] broad	0.29 (0.06)	1.25 (0.51)	0.28

(a) Normalized to CIV line. Numbers in parenthesis show the error values.
(b) From Searjeant et al. (1998).
(c) Evaluated by us from the values of equivalent width in Ivison et al. (1998).
(d) From McCarthy et al. (1993).

The velocity width of the lines are between 450-900 kms^-1. These lines are much narrower than quasar broad lines which have typical width of 5000-10000 kms^-1 but as narrow as those of HzPRG, 500-1000 kms^-1 (e.g., McCarthy 1993). In Table 1, we compare the emission-line width to the two infrared-selected gravitationally-lensed type-2 AGN at high redshift, namely 10214 (e.g., Rowan-Robinson et al. 1991) and SMM 02399-0136 (Ivison et al. 1998).

CIII] line of both B and C seem to have broad wings. Figures 3a,b shows the results of the two-component Gaussian fitting of the CIII] lines for B and C and 3c and 3d shows those with single-component for comparison. They cannot be perfectly fitted with a single component but can be better fitted with a narrow plus broad components as shown in the figures. It seems strange that the CIII] line has a broad component while the CIV line, which comes from the gas at higher ionization stage and may be closer to the central engine, shows only a narrow one. It may be due to the extinction of the broad component of the CIV line. Another possibility is that the wings are the lines of other ions contaminating this spectral range.

The broad CIII] feature is not a unique characteristics of MG 2016+112 lensed object. 10214 (Searjeant et al. 1997) and SMM 02399-0136 (Ivison et al. 1998) also show similar property. Searjeant et al. (1997) considered the possibility of contamination by SiIII] lines to the blue-wing feature of the CIII] line in the spectrum of 10214 but the model cannot fully explain the observed feature (Searjeant et al. 1997). For the case of MG 2016+112, the broad component seems to extend both side of the narrow component (Figs. 3a,b) and thus it is also difficult to be explained by the contamination of SiIII] lines. It is interesting that those two gravitationally-lensed type-2 AGN as well as MG 2016+112 show evidence of a broad CIII] line.

Next, we investigate the line flux ratios. There are significant differences between line ratios of component B and C. While the HeII/CIV ratio of image B is 0.28, the ratio of C is 0.55. Also, the CIII]/CIV ratio of B and C is 0.17 and 0.51, respectively if we consider the narrow components. If we fit the lines with single Gaussian component, then the ratio is 0.40 and 1.32. The emission-line gas in image B seems to be at higher ionization stage than in C. In Fig 4, we compare these line ratios with those predicted by photoionization models (CLOUDY90, Ferland 1988). We examined the cases with hydrogen density $n_{\rm H}=100$ and 1000 cm^-3, power-low ionization continuum with $\alpha=1$ and 1.4, and a range of ionization parameter, $\Gamma$=10^-2.5-10^-1 and solar-abundance. The obtained line ratios are fairly consistent with the typical photo-ionization models considered for narrow-line region of AGN. The differences in line ratios between B and C may be interpreted as the difference in ionization parameters.

Figure 3: Results of the two-components Gaussian fitting for the C III] lines of component B (panel a)) and C (panel b)) and of the one-component ones (panels c) and d)). The LDECONV task developed by Michitoshi Yoshida was used for the fitting

There may be an effect of the reddening by dust in the object on the observed line ratio. According to Calzetti's reddening formula for starburst galaxies (Sawicki & Yee 1998), the reddening of E(B-V) = 1 shifts the observed values with $\Delta$log(CIV/HeII) = -0.12 and $\Delta$log(CIII]/HeII) = +0.30, respectively, which is shown by the arrow in Fig. 4. Even if there is a fairly large amount of reddening, $E(B-V) \sim 1$, the discussion we give here is not so much affected.

There are weak but fairly significant NV lines in the spectra of B and C. In Fig. 5 we plotted the observed NV line ratios as well as those predicted by photoionization models. Clearly, photoionizatin models that can explain the line ratios at lower ionization state are not consistent with the observed NV flux. The observed NV lines are more than several times stronger than the predicted value. Reddening correction moves the points further away from the models. The origin of NV line may be different from those of other lines since the ionization potential of N⁺⁺⁺⁺ is 77.5 eV which is significantly higher than those of CIV (47.9 eV), CIII] (24.4 eV), and HeII (24.6 eV). Indeed, the observed line width of the NV line is larger than others (see Table 1). The enhancement of the NV line may also be due to the nitrogen over abundance. The excess of the NV line intensity is not special for MG 2016+112 B; Hamman & Ferland (1993) also compare their photoionization models with the observed lines of broad-line region (BLR) of quasars and found a large nitrogen over abundance. There is also similar nitrogen problem for the optical NII lines observed in the spectra of local AGN (e.g., Osterbrook 1986).

The fast shock models (Mouri & Taniguchi, private communication) can provide the NV/CIII] and NV/HeII ratios that match the observed values but then predict too strong CIV lines. The line ratios observed in image B and C are not perfectly understood with simple photoionization or shock models.

In Fig. 6, we also plot typical line ratios of various types of high-redshift AGN taken from McCarthy (1993), ultra-steep spectrum HzPRGs compiled by Röettgering et al. (1997), and the infrared-selected type-2 quasars, 10214 (Serjeant et al. 1997). Many of the HzPRGs in Röettgering et al. (1997) are distributed at around log(CIV/HeII) $\sim 0.2$ and log(CIII]/HeII) $\sim -0.2$, which is consistent with the average flux ratio given by McCarthy (1993). The line ratios of image B is very different from typical HzPRGs. The line ratios of C and 10214 lies near the edge of the distribution of line ratios of HzPRGs.

We also plotted the line ratio of the broad lines observed in quasar spectra compiled by Baldwin et al. (1979). They are very different from both HzPRGs and image B and C and can be explained by the photoionization models with high density such as 10^9-10 cm^-3.