First Detection of Outflowing Gas in the Outskirts of the Broad-line Region in 1H 0707−495^*

Narrow-line Seyfert 1 galaxies (NLS1s) are identified by their optical emission-line properties (Osterbrock & Pogge 1985; Goodrich 1989; Leighly 1999a; Mathur 2000; Sulentic et al. 2000; Véron-Cetty et al. 2001; Collin & Kawaguchi 2004; Zhou et al. 2006; Komossa & Xu 2007; Woo et al. 2015; Rakshit et al. 2017). Among their main characteristics, we highlight the presence of narrow permitted optical lines (FWHM of the broad component of Hβ < 2000 km s⁻¹) and weak forbidden lines ([O iii]/Hβ < 3; this distinguishes them from Seyfert 2 galaxies). Moreover, they frequently show strong Fe ii emission (R₄₅₇₀ > 1; ⁴ Osterbrock & Pogge 1985; Goodrich 1989). Another characteristic that makes these sources interesting is their black hole mass, which is systematically smaller (M_BH < 10⁸ M_⊙) than that in broad-line active galactic nuclei (AGNs). This implies a higher accretion rate relative to the Eddington value compared with Seyfert 1 galaxies with broad optical lines (see Komossa 2018, for a comprehensive review). A special subset of this class, known as extreme NLS1s or xA sources within the main-sequence classification (Marziani et al. 2018; Panda et al. 2019; Marinello et al. 2020), shows highly super-Eddington accretion sources (Jin et al. 2017; Panda & Marziani 2023). These xA objects exhibit spectral features such as extremely prominent Fe ii (i.e., R₄₅₇₀ > 1.5; Wang et al. 2014; Du et al. 2015; Marinello et al. 2020), very rapid X-rays, and UV flux variability (Gallo 2006). In some cases, they also harbor ultrafast outflow (UFO) signatures (Tombesi et al. 2010; Parker et al. 2017).

The current understanding of UFOs is that they are powerful winds from the AGNs with outflow velocities larger than 10,000 km s⁻¹. They are usually identified in the X-rays utilizing high-energy absorption features from Fe xxv/xxvi in the 7–10 keV energy band (e.g., Tombesi et al. 2010). It has been suggested that UFOs originate in winds magnetically or radiatively driven off the AGN accretion disk at high Eddington rates (Pounds et al. 2003; Reeves et al. 2003; Fukumura et al. 2015). If this scenario is correct, these winds are of great interest because they are very strong candidates for driving AGN feedback, as they couple with galactic gas much more efficiently than jets.

1H 0707−495 is a low-redshift (z = 0.04) NLS1 galaxy, well-known for its extreme variability and spectral shape (e.g., Leighly 1999b; Turner et al. 1999; Boller et al. 2002). Moreover, it displays a very strong soft X-ray excess and relativistic broad iron line, thought to arise from emission reprocessed by the accretion disk (Fabian et al. 2009). 1H 0707−495 was also the first AGN where an X-ray reverberation lag was detected (Fabian et al. 2009; Zoghbi et al. 2010). One characteristic that makes this AGN interesting is the presence of blueshifted absorption features from a UFO, detected by Dauser et al. (2012) and Hagino et al. (2016) using XMM-Newton spectra. Moreover, in its X-ray spectra emission lines from O viii and N vii are observed (Kosec et al. 2018), in particular when the continuum flux is low. Various authors have at times invoked low-ionization partial covering absorption in 1H 0707−495, either as a way of producing the spectral structure at 7 keV at the Fe K edge (e.g., Mizumoto et al. 2014) or as a way of producing soft X-ray variability (e.g., Boller et al. 2021).

In the ultraviolet (UV), Hubble Space Telescope (HST) Space Telescope Imaging Spectrograph (STIS) observations of 1H 0707−495 reported by Leighly & Moore (2004) show that at least part of the outflows detected in the X-rays are also observed in the UV emission lines. Overall, the STIS spectrum is characterized by a very blue continuum; broad, strongly blueshifted high-ionization lines (including C iv and N v); and narrow, symmetric intermediate-ionization (including C iii], Si iii], and Al iii) and low-ionization (e.g., Mg ii) lines. The last features are centered at the rest wavelength. The study of their emission-line profiles reveals that the high-ionization lines are associated with a wind while the intermediate- and low-ionization lines arise in low-velocity gas. The latter component is likely associated with the accretion disk or with the base of the wind.

Despite the wealth of information about 1H 0707−495 in the X-ray, UV, and optical, very little is known about the near-infrared (NIR) properties of this source. Indeed, to the best of our knowledge, only Durré & Mould (2022) employed this AGN as part of a southern sample of NLS1s to study the kinematics of the broad-line region (BLR) using the Paα line. The NIR wavelength range, though, is highly valuable because (i) the Fe ii emission lines, notably the Fe ii lines around the 1 μm region, are isolated or semi-isolated, unlike in the UV−optical region, allowing a better and more accurate determination of the line properties (Rudy et al. 2000; Rodríguez-Ardila et al. 2002; Riffel et al. 2006); (ii) the H i lines, particularly Paα and Paβ, are isolated, allowing the characterization of their emission-line profiles; and (iii) continuum emission due to hot dust starts to show up in this spectral region.

In this work, we report NIR spectroscopy carried out on 11H 0707−495 aimed at studying the outflow properties already detected in the high-ionization BLR gas but up to today elusive to detection in low-ionization BLR lines. This work is organized as follows: Section 2 provides an overview of the data and observations. Section 3 presents the results, and in Section 4 a discussion and interpretation of the findings are provided. Finally, in Section 5 we give the main conclusions found from the data analysis.

NIR spectroscopy of 1H 0707−495 was obtained using the Triplespec4 spectrograph (Schlawin et al. 2014) attached to the 4.1 m Southern Astrophysical Research Telescope (SOAR) on the night of 2022 February 23. The science detector employed is a 2048 × 2048 Hawaii-2RG Hg-Cd-Te array with a sampling of 041 pixel⁻¹. The slit assembly is 11 wide and 28'' long. The delivered spectral resolution R is ∼2850 across the different dispersion orders. Observations were done nodding in two positions along the slit. Right before the science target, the A1V star HIP 32913 (V = 7.57), close in air mass to the former (air mass = 1.06), was observed to remove telluric features and to perform the flux calibration. Wavelength calibration was carried out using skylines present across the NIR spectra.

The spectral reduction, extraction, and wavelength calibration procedures were performed using spextool v4.1, an IDL-based software developed and provided by the SpeX team (Cushing et al. 2004) with some modifications specifically designed for the data format and characteristics of ARCoIRIS, written by Dr. Katelyn Allers (private communication). Telluric feature removal and flux calibration were done using xtellcor (Vacca et al. 2003). The different orders, extracted using an aperture window of 2'' centered at the peak of the source light profile, were merged into a single 1D spectrum from 0.94 to 2.4 μm using the xmergeorders routine. The final reduced spectrum includes an error vector and appears as a third extension in the data file. It measures the uncertainty in flux calibration at every wavelength and takes into account errors propagated through the extraction process. That extension is employed to estimate the errors associated with the determination of the integrated flux of the lines and continuum fitting.

The final, merged spectrum was corrected for redshift, determined from the brightest lines detected. We employed Paα, Paβ, O i 1.128 μm, Fe ii 1.0501 μm and [S iii] 0.953 μm to this purpose. The average value obtained was z = 0.04126, in excellent agreement with the values reported in the literature (Leighly et al. 1997; Leighly & Moore 2004) and in the NASA/IPAC Extragalactic Database. Afterward, we corrected the spectrum for Galactic extinction using the Cardelli et al. law (Cardelli et al. 1989) and the extinction maps of Schlafly & Finkbeiner (2011). A value of E(B − V) = 0.084 was adopted.

Figure 1 displays the most relevant characteristics observed in the redshifted and Galactic-extinction-corrected NIR spectrum of 1H 0707−495. To the best of our knowledge, this is the first time in the literature that an NIR spectrum of this AGN covering simultaneously the 0.94–2.4 μm interval in rest wavelength is presented. The middle and upper insets display the region around the 1 μm Fe ii lines, Paβ, Paα, and the K-band region. The most important emission lines are identified.

Zoom In Zoom Out Reset image size

Overall, the NIR spectrum of 1H 0707−495 displays similar characteristics to that observed in other NLS1s (Rodríguez-Ardila et al. 2002; Riffel et al. 2006; Marinello et al. 2016, 2020). The most important difference is the lack of bright emission features from the narrow-line region (NLR), such as the forbidden [S iii] 0.953 μm, [Fe ii] 1.257 μm, molecular lines from H₂, and coronal lines, which are rather faint or not detected in 1H 0707−495. This result is consistent with previous observations (Paul et al. 2021). Moreover, the NLR component of the permitted H i lines seems absent, except likely in the Paschen lines.

The continuum emission is dominated by a strong featureless component, with the flux increasing steeply toward shorter wavelengths. No evidence of stellar absorption features was detected in the observed spectrum. Thus, we assume that the stellar contribution is below 10% of the observed continuum (Riffel et al. 2006).

To characterize the NIR continuum emission, we first fit a function described by a power law of the form F_λ ∝ λ^α , where λ is the wavelength and α the spectral index of the power law. Special care was taken to exclude emission lines in the spectral windows that were used in the fit. We found unsatisfactory results with that function and concluded that it alone cannot reproduce the observed NIR continuum. We then employed a composite function, consisting of an underlying power law plus a blackbody component. It has been employed successfully in the fitting of the NIR continuum in other AGNs, including NLS1s (Glikman et al. 2006; Riffel et al. 2009; Landt et al. 2011). Our results show that a power law with spectral index α = −1.95 ± 0.25 and a blackbody of temperature T_BB = 1299 ± 100 K reproduce satisfactorily the observed continuum. The errors quoted for the aforementioned parameters are within 2σ uncertainties. Figure 2 shows the fit (top panel) and the nebular spectrum after subtraction of that component (bottom panel).

Zoom In Zoom Out Reset image size

The value of α derived in this work in the NIR is, within errors, very close to that of −2.3 reported by Leighly & Moore (2004) for the UV/optical continuum. We note that the latter authors do not quote any uncertainty associated with the spectral index. Despite this, we conclude that the power law in the NIR likely represents the low-energy tail of the continuum produced by the central source. The blackbody component is associated with the hottest dust component, likely located in the inner face of the obscuring torus (Glikman et al. 2006; Landt et al. 2011).

After the continuum fitting, we subtracted the power law and the blackbody components from the observed spectrum. This procedure is carried out to study the pure emission-line spectrum of 1H 0707−495. However, the analysis that will be presented in Section 3.1 is not affected by this step (see below).

3.1. The Emission-line Spectrum of 1H 0707−495

The emission-line spectrum in 1H 0707−495 is dominated by permitted, broad features, most of them associated with the BLR. The only NLR lines detected are the [S iii] 0.953 μm in the blue edge and the H₂ line at 1.957 μm in the K band. Both features are intrinsically faint. It is important to notice that the forbidden emission line of [S iii] is usually the strongest forbidden line in the NIR (Riffel et al. 2006). We characterized the observed emission lines in terms of the integrated flux, FWHM, and centroid position of the line peaks. To this purpose, we fitted Gaussian or Lorentzian functions to individual lines or to sets of blended lines. This procedure was carried out using a set of custom scripts written in python by our team. The code uses nonlinear least squares, with scipy.optimize.curve_fit. Open-source software was employed, such as MATPLOTLIB (Hunter 2007), NUMPY (van der Walt et al. 2011), and SCIPY (Virtanen et al. 2020). For each emission line, up to three Gaussian components or two Gaussian components and one Lorentzian are allowed. Constraints for the lines were also added such that lines belonging to the same ion should have the same FWHM (in velocity space) and obey the theoretical wavelength separation.

Figure 3 shows a zoom-in around the region where the Paβ line is located. It is the strongest BLR line detected in the NIR spectrum. We first fit the observed profile with two components: a narrow one, from the NLR, and a broad profile, associated with the BLR. The results are presented in the top two panels of Figure 3. In the left panel the BLR is represented by a single Gaussian profile, whereas in the right panel it is assumed that the BLR contribution can be represented by a Lorentzian profile. A quick inspection of both fits evidences residuals as large as ∼10% of the peak intensity of the line. Moreover, the largest residuals are mostly concentrated toward the blue wing of the emission-line profile.

Zoom In Zoom Out Reset image size

We then tested a scenario of a BLR consisting of two components to account for the observed large residuals in the previous case. The results are shown in the bottom two panels of Figure 3. In both cases, we kept the former BLR component (Gaussian—left panel; Lorentzian—right panel) and included an additional blueshifted broad Gaussian component. All the parameters associated with this extra component were left free in the fit. It can be seen that after the addition of the latter component the residuals were considerably reduced in both cases, being limited now to ∼3% of the line peak.

Although the addition of a second BLR component offers a better description of the observed Paβ profile, it is well-known that the rms of the continuum residuals decreases with the increase of the number of components, regardless of their physical meaning. However, it is worth noticing at this point that previous works on 1H 0707−495 had already reported the detection of blue-asymmetric profiles in UV lines such as C iv, Si iv, and O iv (Leighly & Moore 2004). Leighly & Moore (2004) also point out that low-ionization lines such as Mg ii in the UV or Hβ in the optical were modeled using a Lorentzian profile, with no evidence for either blue or red asymmetries and centered at the systemic velocity. It is important to mention, though, that the last two lines are located in regions that are strongly contaminated by broad humps of Fe ii emission. This makes the full characterization of their line profiles very uncertain.

Even though our results on Paβ agree with previous observational findings using UV spectroscopy on this AGN, to have firm evidence of the need for a second component, we employed the Bayesian information criterion (BIC). It was first introduced by Schwarz (1978) to guide model selection, which is the problem of distinguishing competing models, sometimes featuring different numbers of parameters. Here we apply the BIC to test whether a second Gaussian component is indeed necessary. According to Schwarz (1978), BIC is defined by

$$\begin{eqnarray}&&\mathrm{BIC}=-2\ast \mathrm{ln}(L)+k\ast \mathrm{ln}(N),\end{eqnarray} \tag{ 1 }$$

where L is the likelihood of the model, k is the number of parameters, and N is the number of observations.

Assuming Gaussian errors and the boundary condition that the derivative of the log likelihood with respect to the true variance is zero, according to Liddle (2007), Equation (1) can be expressed in terms of the residual sum of squares (RSS):

$$\begin{eqnarray}&&\mathrm{BIC}=N\ast \mathrm{ln}(\mathrm{RSS}/N)+k\ast \mathrm{ln}(N).\end{eqnarray} \tag{ 2 }$$

The model giving the smallest BIC among the candidates is the one favored. If the difference in BIC values between two competing models is 0–2, this constitutes "weak" evidence in favor of the model with the smaller BIC; a difference in BIC values between 2 and 6 constitutes "positive" evidence; a difference in BIC values between 6 and 10 constitutes "strong" evidence; and a difference in BIC values greater than 10 constitutes "very strong" evidence in favor of the model with a smaller BIC.

It is important to mention that the BIC attempts to mitigate the risk of overfitting by introducing the penalty term $k\ast \mathrm{ln}(N)$, which grows with the number of parameters. This allows us to filter out unnecessarily complicated models, which have too many parameters to be estimated accurately on a given data set of size N.

Using Equation (2) and the values of N and k listed in Table 1, we determine the BIC for the profile modeling applied to Paβ and depicted in Figure 3. The results are in Columns (5)–(6) and (11)–(12) of that table. It is possible to see that the smallest BIC values (−6910 and −6938) are obtained when the BLR is modeled with two components: a broad profile (Gaussian or Lorentzian) with the line peak very close to the systemic velocity (the classical BLR) plus a blueshifted, broad Gaussian component. Moreover, the difference ΔBIC (Columns (7) and (13)) between the BLR represented by one or two components is >10 regardless of the profile employed to represent the classical BLR. Therefore, we conclude that a broad, blueshifted component is a requirement to properly model the observed Paβ profile. Here we suggest that this extra component represents the outflow already detected in UV lines by Leighly & Moore (2004) in this object.

Table 1. Parameters Employed in the Line Fitting Procedure and Resulting BIC Values

	Gaussian BLR						Lorentzian BLR
Line	N	Ka	Kb	BIC a	BIC b	Δ BIC	N	Ka	Kb	BIC a	BIC b	Δ BIC
Paβ	87	6	9	−6747	−6910	163	87	6	9	−6836	−6938	102
Fe ii	68	12	15	−5288	−5359	71	68	12	15	−5339	−5377	38
λ10501
O i	72	3	6	−5649	−5733	84	72	3	6	−5769	−5721	−48
Fe ii+	138	15	21	−10,690	−10,938	248	138	15	21	−10,848	−10,931	83
Paδ+	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
He ii+	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
Fe ii	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯
[S iii]+	53	15	21	−4104	−4170	66	53	15	21	−4090	−4161	71
Pa 8	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯	⋯

Notes. N is the number of data points, K is the number of parameters modeled (three for every Gaussian or Lorentzian function), and BIC is the Bayesian information criterion determined by Equation (1). Columns (2)–(7) list the results when a classical Gaussian BLR is assumed, and Columns (8)–(13) show the results for a Lorentzian BLR.

^a Values without a second Gaussian component. ^b Values when considering a second, blueshifted Gaussian component. ΔBIC is the difference of BIC for model i and the minimum BIC value, BIC${}_{\min }$.

Download table as:  ASCIITypeset image

Table 2 lists the centroid position, flux, FWHM, and shift of the line peak relative to the laboratory position found for each component of both the Gaussian and the Lorentzian approach. The errors in the fluxes are representative at the 2σ level across all cases.

Table 2. Line Parameters (Center, Flux, FWHM, and Shift from the Centroid Position) for the Gaussian (Columns (2)–(5)) and the Lorentzian Fits (Columns (6)–(9))

	Gaussian BLR				Lorentzian BLR
Line	Centre	Flux	FWHM	Δλ	Centre	Flux	FWHM	Δλ
	(Å)		(km s−1)	(km s−1)	(Å)		(km s−1)	(km s−1)
[S iii]NLR	9532	0.49 ± 0.04	122	41	9532	0.47 ± 0.04	119	41
[S iii]out	9526	0.83 ± 0.13	565	−170	9526	0.78 ± 0.14	583	−164
Pa 8NLR	9546	0.10 ± 0.05	118	0	9546	0.08 ± 0.04	118	0
Pa 8BLR	9547	1.18 ± 0.16	550	28	9547	2.33 ± 0.15	597	16
Pa 8out	9541	1.37 ± 0.44	1934	−157	9525	0.83 ± 0.64	2533	−669
Fe ii BLR	9997	2.11 ± 0.13	408	9	9997	4.02 ± 0.11	430	6
Fe ii out	9988	2.88 ± 0.54	1834	−279	9972	1.02 ± 0.53	1850	−765
PaδNLR	10049	0.17 ± 0.04	137	0	10049	0.10 ± 0.03	137	0
PaδBLR	10049	1.72 ± 0.14	537	0	10049	4.08 ± 0.14	598	0
Paδout	10044	3.18 ± 0.51	1954	−149	10026	1.87 ± 0.69	2615	−686
He ii BLR	10120	0.27 ± 1.87	771	−118	10120	0.28 ± 0.14	540	−118
Fe ii BLR	10172	0.39 ± 0.13	523	29	10172	0.62 ± 0.17	591	29
Fe ii BLR	10491	0.40 ± 0.16	415	0	10491	0.50 ± 0.14	427	0
Fe ii BLR	10503	1.63 ± 0.16	414	57	10502	3.04 ± 0.14	432	29
Fe ii out	10489	2.07 ± 0.62	1840	−341	10472	1.20 ± 0.55	1856	−826
O i BLR	11289	1.97 ± 0.15	440	53	11288	4.60 ± 0.47	546	27
O i out	11281	2.27 ± 0.58	1802	−159	⋯	⋯	⋯	⋯
PaβNLR	12818	0.36 ± 0.06	127	0	12819	0.23 ± 0.06	128	23
PaβBLR	12819	3.52 ± 0.20	543	23	12818	8.20 ± 0.21	602	0
Paβout	12809	5.38 ± 0.75	1945	−210	12787	2.70 ± 0.84	2545	−725

Note. Fluxes in units of 10⁻¹⁵ erg cm⁻² s⁻¹. The subscripts "NLR," "BLR," and "out" refer to the component emitted by the narrow-line region, the classical broad-line region, and the outflow component, respectively.

Download table as:  ASCIITypeset image

The BIC values from the Paβ fit also allow us to test whether the classical BLR is best described by a Gaussian or a Lorentzian profile. We found that the smallest BIC is obtained with the latter function. The ΔBIC between these two models (Columns (7) and (13)) is 28, that is, larger than 10. This agrees to previous results reported by Leighly & Moore (2004) to model the Hβ line.

When a Lorentzian profile is employed to represent the classical BLR, it has an FWHM of 602 km s⁻¹ and is centered at the systemic velocity. The outflow component, modeled using a Gaussian component, is found to be blueshifted by −725 km s⁻¹ and has an FWHM of 2545 km s⁻¹. We note that the maximum error associated with the peak position of the line is 30 km s⁻¹, while that associated with the FWHM is 35 km s⁻¹. These values were extracted from the maximum value of the rms of the wavelength calibration provided by the reduction pipeline and from the measurement of the FWHM around the Paβ line, assuming different continuum levels. They can be considered as standard for the analysis presented in the remainder of this work.

We notice that the Paα line in the K band is also detected in our data (see Figure 1). However, we refrain from using it in the analysis done here because the blue wing of the line is very close to the poor transmission region between the H and K bands. An inspection of Figure 1 allows us to confirm the presence of a strong blue-asymmetric profile in Paα, very similar to that of Paβ. However, because the former displays a lower signal-to-noise ratio (S/N) and is fainter than the latter (which contradicts the Paschen decrement), we opted to leave Paα out.

3.2. The NIR Fe ii Emission

In addition to the permitted H i lines, the NIR spectrum of 1H 0707−495 also shows the presence of prominent Fe ii emission features, in particular, the ones at 9997, 10501, 10863, and 11127 Å (Rudy et al. 2000; Rodríguez-Ardila et al. 2002; Marinello et al. 2020). Because of their proximity in wavelength, they are termed as the 1 μm Fe ii lines and are emitted after the decay of the common upper term b⁴G. Overall, Fe ii at 9997 and 10501 Å are very conspicuous and moderately isolated lines. For that reason, they will be employed here in the analysis of the Fe ii emission-line profiles. The former is ∼50 Å apart from Paδ, the nearest strongest emission feature, while Fe ii λ10501 is indeed a blend of Fe ii at 10493 and 10501 Å, rarely being resolved because of their proximity. Moreover, the expected flux ratio λ10501/λ10493 is ∼8 (Marinello et al. 2016). Thus, the flux of λ10501 dominates the blend. At the redshift of 1H 0707−495, the lines at 10863 and 11127 Å are located in regions of bad atmospheric transmission, preventing us from using them reliably.

In this section, we will focus on the results obtained for the Fe ii λ10501 blend. As with the Paβ line, We first tested the hypothesis of a line profile being represented solely by the classical BLR component. The results are shown in Figure A1. The BIC values are listed in Table 1. It is possible to see the presence of residuals in the blue wing in λ10501. For this reason, we included an outflow component to the most prominent line, that is, the one centered at 10501 Å. The left panel of Figure 4 illustrates the result after the addition of that component plus a Gaussian profile to account for the classical BLR contribution. In the right panel, a Lorentzian profile for that component is assumed. The lines at 10493, 10526, and 10546 Å are also emitted by Fe ii but are considerably fainter features than that at 10501 Å. For this reason, only the classical BLR component was used to fit them. The outflow component, if present, is at the S/N level. The BIC values found from these fits are in Table 1. Table 2 lists the best-fit parameters found for the Fe ii blend centered at 10501 Å.

Zoom In Zoom Out Reset image size

As in the Paβ fit, the presence of an outflow component is strongly favored. The fit with the classical BLR plus the outflow component (with the width, intensity, and centroid position left unconstrained in the fit) results in smaller BIC values, with the ΔBIC being >10 when compared to that without outflow. Moreover, the Lorentzian profile to represent the classical BLR is also favored, as the BIC in the latter case is smaller than when a Gaussian profile is assumed. The ΔBIC between these two cases amounts to 18.

The fits shown in Figure 4 confirm the results already gathered from Paβ. We found a classical BLR component, which within errors is characterized by a Lorentzian profile with an FWHM of 432 km s⁻¹ and coincident with the systemic velocity. This agrees with the results of Leighly & Moore (2004) for Mg ii and Fe ii. Moreover, there is a broad blueshifted component that we associate with outflowing gas from the BLR. It has an FWHM of ∼1856 km s⁻¹ and an outflow velocity between −341 and −826 km s⁻¹. To the best of our knowledge, this is the first time that such an outflow has been detected in Fe ii BLR emission in 1H 0707−495. It is important to notice here some additional findings. First, the Fe ii component emitted by the classical BLR displays an FWHM that is smaller than its counterpart in Paβ. This is in agreement with previous results found in the literature (Rodríguez-Ardila et al. 2002; Matsuoka et al. 2007; Barth et al. 2013; Marinello et al. 2016). The latter work reports that the BLR component of Fe ii is, on average, 30% narrower than that of Paβ. Second, the FWHM of the Fe ii lines is likely one of the smallest already reported in the literature, even for an NLS1 AGN. And third, the presence of a shoulder centered at ∼10492 Å, coinciding with Fe ii λ10493, suggests that we detected and resolved that line. The measured Fe ii λ10501/λ10493 flux ratio is 6.08 ± 1.73 using the Lorentzian BLR model, very close to the expected ratio of ∼8 (Marinello et al. 2016). Spectra with larger spectral resolution and S/N are necessary to confirm this result.

Finally, we have also considered the possibility of the Fe ii blend being fully dominated by the line at 10501 Å while the satellite lines are spurious features. This assumption comes from the fact that the calculated intensity of these lines relative to Fe ii λ10501 is at least a factor 2 × 10⁻² smaller. These values result by considering only the Gaunt factors and the Einstein coefficient A_ij of the lines. We then fit a Gaussian or a Lorentzian profile to the observed line plus the outflow component. The BIC values resulting in these two cases were −5297.3 and −5298.2, respectively, that is, larger than those derived when considering the satellite lines (see Table 1). We highlight that the line ratios between the different Fe ii lines may significantly depart from those expected from pure recombination because of collisional excitation, Lyα fluorescence, self-fluorescence, and other processes (see Sigut & Pradhan 2003) that contribute to the observed Fe ii spectrum in AGNs.

3.3. The O i Emission Line

We took advantage of the presence of the strong O i emission at 11287 Å in our spectrum and modeled that line using the two approaches already employed and described above. The fit with a single component to represent the BLR is in Figure A2, and the one including the outflow is shown in Figure 5. The corresponding BIC values are in Table 1.

Zoom In Zoom Out Reset image size

An inspection of Figure A2 reveals that a single Gaussian function does not reproduce satisfactorily the observed profile. Strong residuals are left in the blue and red wings of the emission line. However, when a single Lorentzian profile is employed, the residuals are considerably reduced. This is supported by the BIC, which is the smallest when the latter profile is employed.

We also tested the scenario where a blueshifted broad component is included in the fit. The results are displayed in Figure 5, and the corresponding BIC values are listed in Table 1. It can be seen that in a classical Gaussian BLR the outflow component produces a smaller BIC, with a difference >10 relative to the case of no outflow. In the case of a classical Lorentzian BLR, though, the inclusion of an outflow component does not improve the modeling. Indeed, the BIC increases considerably. Therefore, for O i λ11287 an outflow component is acceptable under the assumption that the classical BLR is represented by a Gaussian profile. Nonetheless, it is important to highlight that the smallest BIC is obtained when that latter region is modeled using a Lorentzian function, making the outflow scenario uncertain for that line. The outflow may be present, but it was not detected within our S/N because of the presence of a strong telluric absorption feature to the blue side of the line. Table 2 lists the parameters found for O i λ11287 using both the Gaussian and Lorentzian approaches.

3.4. Other Line Fittings

We next proceed to fit the blend formed by the Fe ii lines at 9997 and 10172 Å, the Paδ line, and the He ii line at 10124 Å. Figure A3 displays the fit when the BLR is modeled with a single component. The corresponding BIC is in Table 1. Figure 6 illustrates the scenario where a blueshifted Gaussian component is included in addition to a Gaussian profile (left panel) and a Lorentzian profile (right panel) for the classical BLR component. In this process, we tied the Paδ profile to have similar FWHM and centroid positions to those found for the Paβ line. BIC values are in Table 1, while Table 2 lists the best parameters of the different components found from the fit.

Zoom In Zoom Out Reset image size

The BIC values support the need for a blueshifted broad component in both Fe ii and Paδ, in agreement with the results already found using Paβ and Fe ii λ10501. Moreover, the BIC is smaller for the Lorentzian approach, indicating that the profile is the most suitable one to represent the classical BLR.

Finally, we fit the blend formed by the forbidden [S iii] λ9531 line, the Pa 8 line at 9547 Å, and the Fe ii lines at 9501 and 9571 Å. This fit is very important because it involves the only forbidden line detected in the NIR spectrum of 1H 0707−495. Moreover, [S iii] λ9531 is considered the NIR counterpart of [O iii] λ5007 in the optical (Fischer et al. 2017). Figure A4 shows the result when the BLR is modeled with a single component, while Figure 7 displays the profile fit after adding a blueshifted Gaussian component to both [S iii] and Pa 8. The corresponding BIC values are in Table 1.

Zoom In Zoom Out Reset image size

It can be seen that in addition to a narrow component due to the classical NLR, the [S iii] λ9531 line displays a strong blue-asymmetric wing, suitably fitting with a broad component. The velocity derived from the peak centroid of the blueshifted line is −170 km s⁻¹, with an FWHM of 570 km s⁻¹ (see Table 2). Whether the blueshifted component represents the propagation of the BLR outflow into the NLR cannot be confirmed or discarded from the data. However, it is rather tempting to assume that it is, considering that the BLR lines thought to be produced in the outskirts of that region do display an outflow component. The BIC listed in Table 1 shows that the smallest values are obtained when a blueshifted component is added to the blend, confirming the results discussed above. Moreover, the BIC${}_{\min }$ criterion favors the Lorenztian profile for the classical BLR.

UFOs are the most extreme subset of AGN winds, with velocities greater than 10,000 km s⁻¹. They are believed to originate from the inner accretion disk within a few hundred gravitational radii from the black hole (Tombesi et al. 2010; Nardini et al. 2015). The existence of a UFO has been confirmed in 1H 0707−495 (Dauser et al. 2012; Hagino et al. 2016; Kosec et al. 2018) at a velocity of ∼0.13c with an ionization parameter log(ξ/erg cm s⁻¹) ∼ 4.3. Blustin & Fabian (2009) detected Doppler-shifted emission lines, and Kosec et al. (2018) found that the velocities of the blueshifted emission increase in higher ionization species. This implies that the wind in 1H 0707−495 is likely stratified and is perhaps slowing down and cooling at larger distances from the SMBH.

The evidence gathered in this work using NIR spectroscopy and BIC analysis reveals the presence of a blueshifted emission in the low-ionization lines of H i (Paβ, Paδ, and Pa 8), Fe ii λ λ9997, 10501, and, to a lower extent, O i λ11287. Even the forbidden line of [S iii] λ9531 displays clear evidence of such outflow. In all cases, the outflow component is modeled using a Gaussian function with an FWHM between 1800 and 2600 km s⁻¹. The outflow velocity varies from −160 to −820 km s⁻¹ in the lines studied. The lowest velocity is derived when a Gaussian function is employed to model the classical BLR contribution, while the largest one is obtained when a Lorentzian profile is used instead.

We suggest that the outflow that is detected in the BLR high-ionization lines reported in Leighly & Moore (2004) is also present in the low-ionization lines of H i and Fe ii. Previous observations of low-ionization lines such as Mg ii, in the UV, and the Balmer lines, in the optical, made by the latter authors revealed that they appear to be rest-frame and of disk origin. Here we confirm the detection of such components using BLR NIR lines, at rest relative to the systemic velocity and with a turbulent velocity (FWHM) between 410 and 600 km s⁻¹. Moreover, the outflow component is unambiguous, regardless of the type of profile employed in the line fitting to represent the classical BLR, although our study favors the Lorentzian representation. The outflow velocity can be as high as ∼826 km s⁻¹.

Kosec et al. (2018) found a trend of increasing velocities with the increase of the ionization parameter of the ions. They also report possible similar kinetic powers of UV, soft X-ray emitters, and UFO absorbers, suggesting that we are witnessing the evolution of a stratified, kinetic-energy-conserving wind. Our results point out that the wind in 1H 0707−495 extends well into the outer boundary of the BLR, where the Fe ii is formed (Barth et al. 2013; Marinello et al. 2016). Indeed, adding the Fe ii measurements from this work to Figure 3 of Kosec et al. (2018), which relates the strength of the outflows with the ionization parameter, gives further support to our claims. They found a relationship of the form $V\,=b\times {(\xi /\mathrm{erg}\,\mathrm{cm}\,{{\rm{s}}}^{-1})}^{a}$, where V is the outflow velocity, ξ is the ionization parameter, and a and b are constants, a = 0.36 ± 0.04 and b = 1800 ± 300 km s⁻¹.

To derive the expected outflow velocity for Fe ii, we employed log U of ≈−3.25. This value is calculated using cloudy (Ferland et al. 2017) and the observed spectral energy distribution for I Zw 1, supersolar metal content (5–10 Z_⊙), and a microturbulence velocity of 20 km s⁻¹ within the BLR cloud. The model further assumes a cloud column density of 10²⁴ cm⁻² (see Panda 2021, for more details). From the above equation and employing the aforementioned value of U, an outflow velocity of −125 km s⁻¹ is predicted. We note that Kosec et al. (2018) choose arbitrary 500 km s⁻¹ error bars on UV ion velocities owing to a lack of uncertainties in Leighly & Moore (2004). Thus, assuming this uncertainty, the outflow velocity measured here in Fe ii and H i is less than a factor 2 than the one predicted. O i may still be taking part of the outflow, as the Gaussian approach for the classical BLR suggests. Using the same cloudy model as for Fe ii, a log U ≈ −3 is predicted. The corresponding outflow velocity is −222 km s⁻¹, while we measured a value of −159 ± 30 km s⁻¹. We remark that the blue wing of this line may be affected by telluric absorption, hindering its detection using the Lorentzian BLR scenario. The detection of a blueshifted component in the forbidden line of [S iii] indeed suggests that the outflow is coupled to the gas located in the NLR.

Another important result is that the disk component of the low-ionization lines displays a very low turbulent velocity of ∼546 km s⁻¹ in the Fe ii and the O i lines. These are likely some of the narrowest BLR lines already reported in the literature.

A full analysis of the NIR spectrum of 1H 0707−495 is beyond the scope of this work. However, upon inspection of our spectrum, it is evident that IFU observations of this object, both in the optical and in the NIR, are necessary to fully assess the size and energetic properties of the outflow.

We have analyzed the NIR spectrum of the xA NLS1 galaxy 1H 0707−495, widely known for displaying a UFO from X-ray emission lines. The UFO is likely propagated to the inner portion of the BLR because UV high- and mid-ionization lines produced in that region are strongly blue-asymmetric. In this work we provide convincing evidence that the outflow is observed farther out, reaching the outer portions of the BLR. Our results also suggest that at least part of the outflow is detected in the NLR because of the blue-asymmetric line associated with [S iii] λ9531.

The permitted emission lines studied in this work (i.e., H i, Fe ii, and O i) are thought to be formed in the middle and outer portions of the BLR. The analysis of the line profiles shows that, overall, the classical BLR component is best represented using a Lorentzian profile plus and outflow component of up to ∼−730 km s⁻¹ for H i and Fe ii. From our modeling, O i seems not to take part in the outflow. This result is supported by the use of the BIC to select the model that best represents each line. Previous observations have failed to detect the outflow component in low-ionization lines in this AGN. Our results are also consistent, within the uncertainties, to models predicting outflow velocities down to −300 km s⁻¹, for lines formed in gas with U < −3, as is the case here.

Finally, the analysis made on the NIR continuum of 1H 0707−495 shows that it is well represented by the low-energy tail of the optical power law, with a similar spectral index of α = 1.95. Moreover, we found strong evidence of the presence of hot dust, with a temperature of ∼1300 K. This dust very likely is located in the inner face of the torus.

The authors thank the anonymous referee for comments/suggestions on this manuscript. We thank the Brazilian Agencies: Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Agência de Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES).

Facility: 4.1 m Southern Astrophysical Research (SOAR) Telescope at Cerro Pachon - .

Software: Matplotlib (Hunter 2007), Numpy (van der Walt et al. 2011), Scipy (Virtanen et al. 2020), Cloudy (Ferland et al. 2017).