DETECTION OF THE INTERMEDIATE-WIDTH EMISSION LINE REGION IN QUASAR OI 287 WITH THE BROAD EMISSION LINE REGION OBSCURED BY THE DUSTY TORUS

The observed spectra are displayed in Figure 1. It is striking that the UV emission lines of Lyα and C iv are much narrower compared with the optical and NIR broad lines. To make a clear comparison, we show the profiles of Lyα and Hα in their common velocity space as an example in the insert panel. In order to further study the unusual emission lines, we first subtract the underling continuum from the observed spectra by fitting the spectra in three spectral ranges. (1) We fit the continuum of the HST spectrum using a single power law in continuum windows free from strong emission lines. The derived continuum, with a spectral index α ≈ −1.75 (f_ν ∝ ν^α), is redder compared with the quasar composite spectrum (α = −0.46; Vanden Berk et al. 2001). (2) The continuum of the SDSS+DoubleSpec spectrum is modeled with the combination of a power law, a Balmer continuum (supplemented with blended high-order Balmer emission lines), and Fe ii multiplets, as described in detail in Dong et al. (2008). This continuum, with a spectral index of α ≈ −2.04, is also redder than the quasar composite spectrum. (3) The continuum of the TripleSpec spectrum is fit with a second-order polynomial. We overplot these continuum models in Figure 1.

Zoom In Zoom Out Reset image size

After subtracting the continuum model, we obtain the emission line spectrum. Figure 2 displays the strong permitted emission lines, including Lyα, C iv, Mg ii, Hγ, Hβ, Hα, and He i, in their common velocity space. Lyα is dominated by an intermediate-width component with FWHM ≈ 2000 km s⁻¹, while its broad component is almost completely absent. Similarly, the adjacent line C iv also presents a prominent intermediate-width component, but has a more apparent broad component. At longer wavelengths, the emission lines such as Mg ii, Hγ, Hβ, Hα, and He i are dominated by their broad components. The trend that broad components gradually become weaker toward shorter wavelengths indicates that the BELR of OI 287 may be reddened.

Zoom In Zoom Out Reset image size

To quantify the different components of emission lines, we decompose these emission lines into three components: broad, narrow, and intermediate-width. Each component is modeled with a single Gaussian. The same components in different lines are assumed to have the same redshift and line width. For the doublets of C iv and Mg ii, each doublet component is modeled separately with their relative intensity ratios fixed at 1:1, assuming that the emission is optically thick. Hβ shows a red wing extending underneath the [O iii] double lines (the "red shelf," Meyers & Peterson 1985; Véron et al. 2002), which might be attributed to (a) broad Hβ, (b) broad [O iii] λλ4959, 5007, (c) He i λλ4922, 5016, or (d) Fe ii. This feature unlikely originates from Hβ emission since it is absent in Hγ and Hα. Detailed study of this feature is beyond the scope of this paper and we use an additional broad Gaussian to eliminate its influence. The red wing of He i is blended with Paγ, which is also modeled by the three components with the similar profiles as the corresponding components of Balmer lines. The UV lines of N v λ1240 and Si iv λ1397, usually obvious in the spectrum of AGNs, are very weak in this spectrum. These weak lines may also provide important information and we also fit them using the three components to detect their upper limits. The forbidden lines including [O iii] λλ4363, 4959, 5007, [N ii] λλ6548, 6583, and [S ii] λλ6716, 6731 are also fitted. Most of these lines are fitted using one narrow component, except for the three [O iii] lines, which are fitted with two components, one narrow component for the line core and one free Gaussian component for the blue wing. The relative intensity ratios of [O iii] λλ4959, 5007 (for both core and wing components) and [N ii] λλ6548, 6583 are fixed to their theoretical value of 1:3. Then we simultaneously fit all of these emission lines using an Interactive Data Language code based on MPFIT (Markwardt 2009), which performs χ²—minimization by the Levenberg Marquardt technique. During the fitting process, absorption lines in C iv, Mg ii, and He i are carefully masked. The best-fit results are shown in Figure 2 and the emission-line parameters are summarized in Table 3 (Model 1).

In order to check whether the fitting results are model dependent, we decompose these lines also with a different method. Because the three components dominate different lines (for example, NELs are dominant in the forbidden lines, IELs are dominant in the UV permitted lines, and BELs are dominant in the optical permitted lines), their shifts and profiles can be reliably obtained in the corresponding lines. We first obtain the redshift and line width of the NELs from the core of the [O iii] double lines, the BELs from the Hα emission line, and the IELs from the Lyα emission line. Then, we use the three-component model to fit all of the emission lines. These best-fit results, also shown in Table 3 (Model 2), are similar to those from the method above, indicating the line decompositions are independent of models we used.

Due to the weakness of the UV BELs in the quasar OI 287, the UV IELs become prominent and thus can be reliably measured. We made a simulation by measuring the Lyα IEL as an example to investigate the dependence of measurement accuracy over the IEL/BEL flux ratios. The simulation result shows that the 1σ measurement flux errors for OI 287 are greatly reduced from 18.7% to 0.8%. Meanwhile, the profile decomposition uncertainties for the optical/NIR IELs can also be reduced, because their redshifts and profiles can be fixed to those of the prominent UV IELs. We also carried out a simulation by measuring the Hα IEL as an example to inspect this fitting strategy. This simulation shows that by fixing the redshift and profile of the Hα IEL to those of Lyα IEL, the 1σ measurement flux errors are reduced from 37.3% to 15.4%. (See the Appendix for how we perform the simulations and calculate the measurement errors.)

With the measurements of emission lines, we first investigate the extinction using the Balmer decrement (Hα/Hβ). The intrinsic Balmer decrement under the CASE B condition is Hα/Hβ = 2.76–3.30 (Osterbrock & Ferland 2006). The measured Balmer decrement for NELs of OI 287, Hα/Hβ = 3.06 ± 0.19, is within the theoretical range. Similarly, the Balmer decrement for IELs, Hα/Hβ = 2.74 ± 0.90, is also consistent with the theoretical value. These results indicate that the NELR and IELR are not reddened. On the other hand, the Balmer decrement for BELs, Hα/Hβ = 5.79 ± 0.07, is much larger than the theoretical value, indicating that the BELR is obviously reddened. We further investigate the BELR extinction through intensity ratios of BELs in OI 287 to BELs in the quasar composite. As a rough approximation, we evaluate the BELs intensities of the composite quasar spectrum by simply employing the measurements for whole emission lines (Lyα, C iv, Mg ii, Hγ, Hβ, and Hα from Vanden Berk et al. 2001; He i and Paγ from Glikman et al. 2006), since these emission lines in the composite spectrum are dominated by their broad component. He i is blended with Paγ, therefore we use the summed intensity (He i + Paγ) for both OI 287 and the composite spectrum. Figure 3 shows the derived intensity ratios of BELs in OI 287 to BELs in the quasar composite, which is normalized to unity at He i + Paγ. The intensity ratios gradually decrease toward shorter wavelengths, suggesting that the BELs of OI 287 are reddened by dust. We fit the intensity ratios using three commonly used extinction curves of the Small Magellanic Cloud (SMC), Large Magellanic Cloud, and Milky Way. The parametrizations of these extinction curves are taken from Pei (1992). All of these three extinction curves can interpret the BEL intensity ratios, but the SMC-like extinction, which yields an E(B − V) of 0.29 ± 0.03, is relatively more likely. Anyway, all of these extinction curves indicate that the BELR of OI 287 is obscured by dust.

Zoom In Zoom Out Reset image size

Since dust extinction is stronger toward shorter wavelengths, the broad components of Lyα and C iv are heavily suppressed, which can naturally explain why the intermediate-width component in these two lines are prominent. On the other hand, the prominent IELs in the UV imply that they should arise from a distinct region that is not obviously obscured by dust. The location of the IELR will be further discussed in Section 4.1.

Since the BELR of OI 287 is obscured, the central accretion disk is also likely to be obscured because the optical-emitting part of the accretion disk is smaller than the BELR. Therefore, we investigate the SED, which can reveal the properties of the accretion disk, to confirm the extinction.

With the multi-wavelength spectroscopic and photometric data, we construct the broadband SED of OI 287 spanning from 1100 Å to 15 μm in the rest frame. Figure 4 displays the broadband SED. From the infrared to optical band, photometric and spectroscopic data are well consistent with each other, implying that the variation is not significant among different observation epochs. The long-term intensive monitoring observations by the Catalina Sky Survey⁸ for nearly 10 years (2005 April 10–2014 January 23) demonstrate that the optical V-band variation amplitudes of OI 287 are within 0.1 mag. The only exception is that the GALEX photometry is larger than the HST spectrum for about 2 times. It is not clear whether this difference is caused by the calibration uncertainties or intrinsic variabilities. As the HST spectrum provides more information on the continuum, we use the HST data in the following analysis.

Zoom In Zoom Out Reset image size

For comparison, a rescaled quasar composite spectrum (Zhou et al. 2010), which is obtained by combining the SDSS composite (λ < 3000 Å; Vanden Berk et al. 2001), the NIR template (3000 Å < λ < 2 μm; Glikman et al. 2006), and the mid-infrared template (λ > 2 μm; Netzer et al. 2007), is overplotted in Figure 4 and normalized to the SED of OI 287 at the W3 band. The observed SED of OI 287 is nearly identical to the composite spectrum in the longer wavelength range (λ > 7000 Å), while gradually deviates from the composite spectrum toward the shorter wavelength range (λ < 7000 Å). This indicates that the accretion disk may be reddened.

To further analyze the reddening of the accretion disk emission, we model the broadband SED using the following four components. (1) Reddened power law: we use a power law with a spectral index α reddened by the SMC extinction curve with an E(B − V) to represent the continuum radiation from the obscured accretion disk. (2) Scattered power law: spectropolarimetrice observations showed that OI 287 has significant polarization of up to 8% caused by scattering (Moore & Stockman 1981; Goodrich & Miller 1988; Rudy & Schmidt 1988). In the observer-frame wavelength range of λ ≈ 4600–8000 Å (3200–5500 Å in the quasar rest frame), the polarization modestly increases toward shorter wavelength (Rudy & Schmidt 1988), inferring that the scattering may be more important in the UV. The scattered light, arising from an extensive region, should not be reddened as indicated by the unreddened NELR. Based on these considerations, we add an unreddened power law with the same spectral index α to represent the scattered component. (3) Hot black body: the near infrared bump is caused by the thermal radiation of hot dust in the inner region of the dusty torus. We use a black body with temperature T_HD in the range of 500–1500 K to model the hot dust radiation. (4) Warm black body: from middle to far infrared, the SED is dominated by the radiation from warm dust in the outer region of the dusty torus. A black body with temperature T_CD < 500 K is used to model the warm dust radiation. We do not consider the contribution of the quasar host galaxy. There is no significant feature of starlight found in the spectrum. In addition, from the SDSS r-band photometry data, we do not find significant difference between the point-spread function magnitude (17.71 mag) and the model magnitude (17.66 mag), which also implies that the contribution of starlight is not important.

The best-fit results are displayed in Figure 4. The four-component model provides a good fit to the observational data and the derived parameters are also reasonable. The derived spectral index of the power law is α = −0.56 ± 0.13, similar to that of the quasar composite spectrum −0.46. The E(B − V) derived from the SED fitting is 0.31 ± 0.03, which agrees well with that derived from the BELs (0.29 ± 0.03) in Section 3.1. The fraction of scattered component increases from 5.3% at 5500 Å to 10.1% at 3200 Å, roughly consistent with the observed degree of polarization from p ≈ 6% to 8% across the wavelength range (Rudy & Schmidt 1988). Our fitting suggests that the hot dust has a temperature of T_HD = 1250 ± 60 K, similar to the value found by Glikman et al. (2006) by modeling the composite quasar spectrum. By combining the measured hot dust luminosity (L_HD) with the bolometric luminosity L_bol evaluated using the bolometric correction L_bol = 9 λ L_λ (5100 Å) (Kaspi et al. 2000), we derive an estimation for the covering factor f_C of the hot dust as L_HD/L_bol ≈ 7%. We will use this estimation to study the properties of the IELR in Section 4.1.

The results above indicate that the accretion disk and BELR of OI 287 are obscured by dust located somewhere along the line of sight. Since the dust is always mixed with gas, we can trace the location of the dust via the gas absorption lines. In the spectrum of OI 287, we found one absorption-line system, involving C iv λλ1548, 1551, Mg ii λλ2796, 2803, and He i* λλ3189, 3889, 10830. We show the local spectrum around these absorption lines in the left column of Figure 5.

Zoom In Zoom Out Reset image size

We normalize the absorption lines using the best-fit models above of emission lines (NEL, IEL, and BEL) and SED (reddened power law, scattered power law, hot black body, and warm black body). First, we subtract the models of NEL, IEL, black body, and the scattered power law from the observed spectrum, assuming that these emission regions are not covered by the absorption gas. This is based on the following considerations. (a) These regions are located in extensive regions far away from the compact central source, which are usually hard to obscure. (b) As shown in Section 3.1, NELs and IELs are not significantly dust reddened, which implies that neither one is covered by the obscuring gas. (c) As shown in Figure 5, after subtracting the emission models above, there is almost no residual flux at the bottom of the C iv and He i absorption lines. It can be readily interpreted as that the absorption gas fully covers the BELR and accretion disk (similar to the dust mentioned above) and does not obscure the light from the other regions, i.e., IELR, hot and warm dust, and the scattering region. We then normalize the observed spectrum by the sum of the power law and BELs. These two emission regions are assumed to be covered by gas, as indicated by their reddening by dust.

The normalized absorption lines are shown in the right column of Figure 5. Most of these absorption lines have similar shifts and line widths, except those of the C iv doublet, which have larger blueshifts, broader line widths, and multiple absorption troughs. Therefore, we cannot acquire reliable results from the C iv absorption lines and focus on the other absorption lines in the following analysis. We simultaneously fit all of the absorption lines using a single Gaussian for each line, assuming that they have the same shifts and line widths. The best-fit results are presented in the right column of Figure 5. The equivalent width ratio of the Mg ii doublets, EW(Mg ii λ2796)/EW(Mg ii λ2803) = 1.84 ± 0.21, is consistent with the theoretical ratio of 2.00 for unsaturated lines with a full coverage. The He i* λ3189 absorption line is marginally detected with a 3σ upper limit of EW(He i* λ3189) ≲ 0.21 Å. The ratio of EW(He i* λ3889)/EW(He i* λ3189) ≳ 3.04 is also consistent with the theoretical ratio of 3.08. These results suggest that the absorption gas fully covers the accretion disk and BELR, and that the absorption lines are not saturated. Under this scenario, the column density of corresponding ions can be derived from their equivalent widths using the equation from Jenkins (1986), $N=({m}_{e}{c}^{2}/\pi {e}^{2}f{\lambda }^{2})\mathrm{EW},$ where f is the oscillator strength, m_e is the electron mass and e is the electron charge. We derive the column densities for Mg⁺ from the Mg ii λ2796 line with ${N}_{{\mathrm{Mg}}^{+}}=(1.91\pm 0.54)\times {10}^{13}{\mathrm{cm}}^{-2},$ and for He i* from the He i* λ3889 line with ${N}_{\mathrm{He}\;{{\rm{I}}}^{*}}=(1.93\pm 0.22)\times {10}^{14}{\mathrm{cm}}^{-2}.$ In addition, with the same redshift and line width, we also measure the 3σ upper limit of the Hα absorption line, EW(Hα) ≲ 0.09 Å, and acquire the maximum column density of hydrogen at the n = 2 level, N_{H i,2} ≲ 3.6 × 10¹¹ cm⁻². These measurements will be used to study the properties of the absorption gas in Section 4.2.

With the measurements of the Hβ broad line width and extinction corrected continuum luminosity at 5100 Å, the central black hole mass (M_BH) is estimated to be ${1.46}_{-0.82}^{+1.89}\times {10}^{9}$ M_⊙ by⁹ employing the empirical formula of Wang et al. (2009). By combining M_BH with the measurement of FWHM(IELs), and assuming that the clouds in the IELR are virialized, the distance of the IELR to the central black hole (R_IELR) can be derived as R_IELR = G M_BH/(f FWHM(IELs))², where G is the gravitational constant and f is a scaling factor. With a simple approximation of an isotropic IELR and Gaussian-profile IELs, $f=\sqrt{3}/2.354$.¹⁰ With these assumptions, we derive an estimate of ${R}_{{\rm{IELR}}}={2.9}_{-2.1}^{+4.3}\;\mathrm{pc}$. This is similar to the dust sublimation radius (R_sub) of 1.3 pc, estimated using the formula in Barvainis 1987, ${R}_{{\rm{sub}}}=1.3\;{L}_{\mathrm{uv},46}^{0.5}\;{T}_{1500}^{-2.8}\;\mathrm{pc},$ where L_uv,46 is the UV luminosity of the central source in units of 10⁴⁶ erg s⁻¹, and T₁₅₀₀ is the grain sublimation temperature in units of 1500 K. The coincidence of these two values implies that the IELR may be located in the inner part of the dusty torus. According to the widely accepted unified model of AGNs (e.g., Antonucci 1993), gas on the inner surface of the dusty torus is exposed to the central ionizing source. As a result, it can be inferred that illuminated gas in this region may produce emission lines through photo-ionization processes.

If the IELs are produced through photo-ionization processes, we can constrain the IELR physical conditions by comparing the observed IELs with those of the photo-ionization models. We perform a simulation using CLOUDY (Version 13.03, Ferland et al. 1998) by considering a gas slab, which is illuminated by a quasar with an SED defined by Mathews & Ferland (1987, hereafter MF87). The quasar monochromatic luminosity at 1450 Å is scaled to that of OI 287 after extinction correction, λ L_λ(1450 Å) ≈ 5.4 × 10⁴⁵ erg s⁻¹, and the distance of the gas to the quasar (d) is fixed to the R_IELR of OI 287 (i.e., the hydrogen ionizing photon volume density is fixed to be Φ(H) ≈ 10^17.2 photons s⁻¹ cm⁻²). The covering factor is set to that of the hot dust (∼7% in Section 3.2), assuming that the IELR is located in the inner part of the dusty torus. The effect of dust grains is taken into account in the calculation. As the extinction of BELs in OI 287 can be described using the SMC-like extinction law, we model the dust with the same composition, consisting of graphite and silicate, and the same size distribution as SMC suggested in Weingartner & Draine (2001). The total abundance of the gas and dust is assumed solar and the dust-to-gas ratio is set to that of the dusty torus of OI 287 estimated by the absorption lines (Section 4.2). We calculate a grid of models by varying the hydrogen density (n_H) from 10³ to 10¹¹ cm⁻³ and hydrogen column density (N_H) from 10¹⁹ to 10²³ cm⁻².

The calculated results are shown in Figure 6, where we plot the contours of EW(Lyα), EW(C iv) and EW(Hα) as functions of n_H and N_H. In each panel, the dashed lines denote the basic models and the filled areas represent the observed¹¹ range with 1σ confidence level. Within the overlapping region, the parameters are constrained in a narrow range of n_H ∼ 10^8.8–10^9.4 cm⁻³ and N_H ∼ 10^19.6–10^20.2 cm⁻². The derived n_H is consistent with the gas density near the sublimation radius suggested by the recent observation by Kishimoto et al. (2013) and modeling by Stern et al. (2014).

Zoom In Zoom Out Reset image size

Adopting the derived n_H and N_H, we predict the equivalent widths of all permitted IELs. The results are shown in Figure 7. The corresponding observed values are also plotted for comparison. Since the observed N v and Si iv are marginally detected, we show their 3σ upper limit. The equivalent widths predicted by the photo-ionization model agree with their observed values within the uncertainties, which supports that IELs may be produced through photo-ionization processes. More importantly, adopting the covering factor estimated from hot dust emission gives results which are consistent with observations. Such results strongly favor that the IELR is closely associated with the hot dust region, the inner part of the dusty torus.

Zoom In Zoom Out Reset image size

With the derived parameters above, the mass of emitting clouds in the IELR is estimated to be ${M}_{\mathrm{IELR}}$ = m_p $4\pi {R}_{\mathrm{IELR}}^{2}$ f_C N_H ≈ 4.5 M_⊙, where m_p is the mass of a proton. Compared with the typical mass range of the dusty torus ∼10⁵–10⁷ M_⊙ (Mor et al. 2009), the mass of the IELR is negligible. This is generally consistent with the scenario that the IELR located in the inner part of the dusty torus. In addition, the thickness of the IELR is estimated to be l ∼ N_H/n_H = 2 × 10⁻⁸ pc, which is also much smaller than the typical size of a dusty torus. However, clouds in the IELR may be not smoothly distributed. The dusty torus is suggested to be very clumpy or filamentary (e.g., Krolik & Begelman 1988), which requires a small volume filling factor ${f}_{{\rm{F}}}\ll 1$ (Nenkova et al. 2002). If so, the IELR may consist of a large number of ionized clouds that are distributed throughout much larger region of the torus.

As shown previously, the central accretion disk and BELR of OI 287 are obscured by dust, while the IELR and NELR are not. This difference implies that the obscuring material is likely associated with the dusty torus. More precisely, the properties of the obscuring material can be estimated using the gas absorption lines, since dust is always mixed with gas and only one set of gas absorption-line system is found in the spectrum of OI 287. To constrain the obscuring material, we also perform a simulation with CLOUDY by considering a dusty gas with SMC-type grains, which is illuminated by a quasar with an MF87 SED. The total abundance of the gas and dust is assumed to be solar and the dust-to-gas ratio of A_v/N_H increases with a grid of 1.0, 1.5, 2.0, 2.5, 3.0 × 10⁻²² cm².¹² For each dust-to-gas ratio, we calculate a two-dimensional grid with variable U of 10^−1.5–10^−0.5 and n_H of 10³–10⁸ cm⁻³ and stop the calculation when A_v reaches the observed value of 0.83. (With R_V = 2.87 for SMC (Gordon & Clayton 1998) and E(B − V) = 0.29, we get ${A}_{V}\equiv {R}_{V}E(B-V)=0.83.$) Figure 8 shows the calculated results. We use the observed ranges of ${N}_{{\mathrm{Mg}}^{+}},$ ${N}_{\mathrm{He}\;{{\rm{I}}}^{*}}$, and N_{He i,2} to constrain the parameters. When A_V/N_H = 3 × 10⁻²² cm², ${N}_{{\mathrm{Mg}}^{+}},$ ${N}_{\mathrm{He}\;{{\rm{I}}}^{*}}$ and N_{He i,2} form an overlap of U ∼ 10^−0.9–10^−0.8 and n_H ∼ 10^4.0–10^6.5 cm⁻³. With U and n_H, the distance of absorber to central ionizing source is derived as ${R}_{{\rm{absorber}}}={(Q({\rm{H}})/4\pi {{cUn}}_{{\rm{H}}})}^{0.5},$ where Q(H) is the number of ionizing photons, $Q({\rm{H}})={\displaystyle \int }_{\nu }^{\infty }{L}_{\nu }/h\nu d\nu \approx 2.2\times {10}^{56}\mathrm{photons}{{\rm{s}}}^{-1}.$ In each panel of Figure 8, we also show the contour of R_absorber as functions of U and n_H. As shown in Panel (e) of Figure 8, in the overlapping region, R_absorber is constrained in the range of ∼10–200 pc. This indicates that the accretion disk and BELR of OI 287 is obscured by dust located in an outer region of the dusty torus.

Zoom In Zoom Out Reset image size

In brief, the derived properties of the obscuring material and IELR suggest that both of them are part of the dusty torus. The IELR, located in the inner part of the dusty torus, is exposed to the central ionizing source and produces IELs through photo-ionization processes; while the obscuring material, in an outer region of the dusty torus, obscures the central accretion disk and BELR like a "coronagraph." Figure 9 displays the cartoon for detecting the IELR in which the dusty torus can be treated as a "coronagraph." The central accretion disk and BELR are obscured by the boundary of the dusty torus, which can account for the observed facts of a reddened SED and BELs. However, the IELR is not fully obscured by the dusty torus, which yields the observed IELs.

Zoom In Zoom Out Reset image size

In addition, as shown in the cartoon, a fraction of the IELR may also be obscured, especially the near side of the IELR (marked with "B" in Figure 9). Therefore, the observed intensities of the IELs might be smaller than their intrinsic ones. As a result, some parameters (such as n_H, N_H, and M_IELR) derived from the strength of IELs could be moderately underestimated, but not change dramatically. The qualitative properties of the IELR analyzed previously remain the same. Although the "coronagraph" may obscure a fraction of the IELR, it makes the detection of IELs much more robust.

With the reddening quantities derived above, we recover the emission lines of OI 287 before dust extinction. Figure 10 displays the extinction corrected emission-line profiles of Lyα and C iv, both normalized to the continuum. In both Lyα and C iv, the unreddened BELs strongly outshines the IELs. The intensity ratio of IEL/BEL before extinction is only 2.8% in Lyα and 3.2% in C iv, respectively. Such low ratios hereby elude detection of the IELs if BELs are not suppressed, which demonstrates the importance of obscuring the BELs for detecting the weak IELs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As a comparison, we overplot the Lyα and C iv emission-line profile of the composite spectrum (Vanden Berk et al. 2001) in Figure 10. The extinction corrected emission-line profiles of OI 287 are similar to those of the composite spectrum, indicating that the IELs found in OI 287 are common and in normal quasars the IELs are usually "hidden" in the spectrum.

Since there is nothing special about the IELs of OI 287, we infer that more objects similar to OI 287 could be found. Taking OI 287 as a prototype, we have found a few tens of analogues of OI 287 from the Baryon Oscillation Spectroscopic Survey (Dawson et al. 2013). In addition, Alexandroff et al. (2013) recently presented a sample of candidate type II quasars selected to have strong lines of Lyα and C iv with average FWHM ∼ 1500 km s⁻¹. In follow-up observations, Greene et al. (2014) reported that some of these objects have a broad Hα component with FWHM up to 7500 km s⁻¹. Their broadband SED and Balmer decrement indicate that these objects are moderately obscured. These characteristics are very similar to those of OI 287, implying that they are likely analogues of OI 287. The findings of these objects indicate that the IELR is common in AGNs and OI 287 may represent a population of AGNs. It is possible and meaningful to compile a large sample of objects similar to OI 287, which can be used to further study the properties of IELRs.