Skip to main content

Full text of "A Near-Infrared Template Derived from I Zw 1 for the FeII Emission in Active Galaxies"

See other formats


(N 



m 



O 



> 



A Near-Infrared Template Derived from I Zw 1 
for the Fell Emission in Active Galaxies 

A. Garcia-Rissmann and A. Rodriguez-Ardila^ 



O '. Laboratorio Nacional de Astrofisica, 

(N 
;-( I Rua Estados Unidos 154, 37504-364, Itajuba, MG, Brazil 



T.A.A. Siguti 
The University of Western Ontario, 



O '. London, Ontario N6A 3K7 Canada 

u 



and 

A.K. Pradhan 

4055 McPherson Laboratory, The Ohio State University, 

140 W. 18th Ave., Columbus, OH 43210-1173, USA 

arissmannSlna . br 



O ■ Received ; accepted 



K^ ' to appear in ApJ 



^Visiting Astronomer at the Infrared Telescope Facility, which is operated by the Univer- 
sity of Hawaii under Cooperative Agreement no. NNX-08AE38A with the National Aeronau- 
tics and Space Administration, Science Mission Directorate, Planetary Astronomy Program. 



-2- 



ABSTRACT 



In AGN spectra, a series of Fe ll multiplets form a pseudo-continuum that ex- 
tends from the ultraviolet to the near- infrared (NIR). This emission is believed to 
originate in the Broad Line Region (BLR), and it has been known for a long time 
that pure photoionization fails to reproduce it in the most extr eme cases, as does 



the co Uisional-excitation alone. The most recent models by 



Sigut fc Pradhan 



( 120031 ) include details of the Fell ion microphysics and cover a wide range in 
ionization parameter logf/ion= (-3.0 — )■ -1.3) and density logrin = (9.6 -^ 12.6). 
With the aid of such models and a spectral synthesis approach, we study for the 
first time in detail the NIR emission of IZwl. The main goals are to confirm 
the role played by Lya fluorescence mechanisms in the production of the Fe ll 
spectrum and to construct the first semi-empirical NIR Fe ll template that best 
represents this emission and can be used to subtract it in other sources. A good 
overall match between the observed Fe ll-|-Mg ll features with those predicted by 
the best fitted model is obtained, corroborating the Lya fluorescence as a key 
process to understand the Fell spectrum. The best model is then adjusted by 
applying a deconvolution method on the observed Fe ll+Mg ll spectrum. The de- 
rived semi-empirical template is then fitted to the spectrum of Ark 564, suitably 
reproducing its observed Fe ll-|-Mg ll emission. Our approach extends the current 
set of available Fe ll templates into the NIR region. 

Subject headings: galaxies: active - galaxies: individual (IZwl, Ark 564) - infrared: 
general - quasars: emission lines - techniques: spectroscopic 



-3- 



Introduction 



The broad-line region (BLR) of active galactic nuclei (AGN) is thought to consist of 
a roughly spherical mist of cloudlets with characteristic densities in the range 10^-10^^ 
cm"^ and column densities of ^10^^ cm~^, surroud i ng a central s ource emitting ionizing 



radiation roughly isotropically ( ISulentic et al. 



2000 : 



Gaskell 



20091 ) ■ Despite the success of 



this traditional picture, in order to explain the strengths of the BLR lines, the need for a 
high covering factor and the lack of Lyman continuum absorption, a BLR having a flattened 



distribution at least for the low-ionization gas has been proposed (JGaskell 



20091 ) 



The BLR has been extensively stud ied from the X-rays to the NIR r egime during 



the last two decades (see the reviews of 



Sulentic et al. 



2000 



Gaskell 



20091 ) ■ One of the 



most puzzling aspects of the line spectra emitted by the BLR is the Fe ll emission, whose 
numerous multiplets form a pseudo-continuum which extends from the UV to the optical 
region due to the blending of approximately 10^ lines. This emission constitutes one of the 
most important contributors to the cooling of the BLR. 

Indeed, the blending of several BLR emission lines, including the large number of Fe ll 
emission multiplets, prevents a reliable study of individual line profiles and the identification 
and measurement of weaker lines. As blending is minimized in the class of objects known 
as Narrow-Line Seyfert 1 galaxies (NLSls), their study can lead to a significantly more 
accurat e study of the properties of the emission line region that is closer to the central 



source. 



Boroson fc GreenI (jl992[ ). for instance, derived an optical template for the Fell 



multiplets from IZwl, which has served since then to adjust the Fell strength of several 
other objects, after a proper scaling and convolution to match the BLR velocity dispersions 
(estimated from strong emiss ion lines). The advent of HST UV quasar spectral data allowed 
Vestergaard fc Wilkesl (120011 ) to extend the template method into the UV regime. They 
presented the first high S/N, high-resolution, quasar empirical UV iron template spectrum 



-4- 



ranging from rest frame 1250 A to 3090 A, which is apphcable to quasar data. The template 
was based on HST (archival) data of IZw 1. 

The iron emission templates have importance not only for our ability to measure and 
subtract the iron emission in quasar spectra, but also as a tool through which we can 



study the iron emission stren gth itself. Iron is a key coo 



energy output from the BLR (jWills. Netzer fc Wills 



ant emitting ~25% of the total 



19851), emphasizing the importance of 



including the iron emission in studies of the BLR. 

However, despite the wide use of such templates, most of the physical mechanisms that 
produce such lines remain under debate. There have been a number of pioneering theoretical 
investigations about the Fe ll emission spectra in active galactic nuclei (AGNs). For example, 
Phillips (1978a, 1978b) discussed conti nuum pumping as one of th e excitation mechanisrn s 



that a re responsible for that emission. 



Netzer fc Willsl fll983[ ) and 



Wills. Netzer fc Wills 



( 119851 ) calculated the strengths of 3407 Fell emission lines assuming coUisional excitation 
and continuum fluorescence of Fe ll, with radiative transfer in the spectral lines treated in 
the first-order escape probability approximation. They found a good fit to the overall shape 
of Fe II features in the AGN UV and optical spectra but recognized that the total strength 
of the Fe ll emission is l arger than the one p r edicte d by the models by a factor of ~4. 



Other attemps made by 



CoUin-Souffrin et al 



( 119861 ) to solve the apparent weakness of the 



Fe II emission using multi-component photoionization models were also unsuccessful. These 
failures point out that not all the excitation mechanisms have been taken into account or 
even that non-radiatively heated material with possibly even greater density exists within 
the BLR. 



In order to solve the Fell problem. 



PenstonI ( 119871 ) suggested Lya fiourescence as 



the main physical process involved in the production of Fe ll lines. It takes advantage 
of the various near-coincidences between the wavelength of Lya and the wavelengths 



-5- 



corresponding to transitions between the levels of the a^ D term and the 5p odd parity 
levels in Fell, as described by I Johansson &: JordanI (Il984l ). The largest calculated transition 



probabilities from the odd 5p levels are those to e^D and e^D, and cascades from these 
levels to odd parity levels at 5 eV and then to a^D and a'^D would produce the bulk of the 
Fell spectrum located between 2000 A and 3000 A. 

Model calculations including L yg fluorescence as the excitation mechanism for the Fell 



lines (ISigut &: Pradhan 



1998 



20031 ) showed that this process is of fundamental importance 



in determining the strength of the Fe ll emission. Previously, Fe ll features in the intervals 
2400-2560 A and 2830-2900 A that origi nate from high excitation levels (~10 eV) had bee n 



LaoretaL 



ident ified in the spectra of some AGNs (iGraham. Clowes fc Campusand 11996 : 

19971 ). favou ring the Lya fluor e scenc e scenario. The key feature to test this process, as 



predicted by lSigut &: PradhanI (1l998l ). is signiflcant Fell emission in the wavelength range 
8500-9500A, where the primary cascade lines from the upper 5p levels to e^D and e^D are 
located. Up to a few years ago, that emission had been elusive to ob serve in AGNs althou gh 
they are common features in the spe ctra of some Galactic sources ( jHamman fc Persson 



1989 



Kelly. Rieke. fc Campbell 



19941 ). 



Fortunately, sensitive NIR spectroscopy carried out on AGN samples during the last 
decade at moderate spectral resolution (R~800) revealed a wealt h of emission lines froni 
Fell i n a previously unexplored wavelength region. For instance 



Rodriguez- Ardila et al. 



f l2002h . hereafter RA02, identifled for the flrst time in four NLSl galaxies (Ark 564, Mrk335, 



IH 1934-063 A and Mrkl044) the str ongest primary cascade lines of Lya fluorescence 



predicted by ISigut fc PradhanI ( 1l998l ). In addition, the secondary UV lines resulting from 
the decay of the e^D and e^D levels are also present in these objects. Those results provided 
strong observational support to the hypothesis that Lya fluorescence is, indeed, contributing 
to the emitted Fe ll spectrum. Furthermore, RA02 reports the presence of the so-called 



-6- 



l/iin Fell lines (Fell A9997, A10501, A10862 and A11126). These are the most prominent 
Fell features observed in the rest wavelength interval 0.8-2.4 fim. The importance of that 
finding comes from the fact that nearly ~50% of the optical Fe ll emission results from 
decays of the z'^D^ and z'^F'^ levels, which are populated either by the transitions leading to 
the emission of the l/xm lines and the second ary UV lines r nentio ned above, or by collisions 



from lower levels and direct photoionization. jLandt et al. 



(120081 ) reported similar findings 



in a s ample of 23 well-k i iown broad-emission line AGNs. They also confronted, for the first 



time. 



Si gut &: PradhanI (119981 ) theoretical predictions of the NIR iron emission spectrum 



with observations. However, the prototypical I Zw 1 was not included in their sample. 



Rudy et al. 



mm, 



The only observation to date of I Zw 1 in the NIR was reported by 
hereafter RMPH. Their work clearly reveals Fell A9997, A10501, A10862, A11126. Based on 
the absence of the crucial cascade lines that feed the common upper state where the 1 /im 
Fe II lines originate (assuming Lya fluorescence as the dominant mechanism) as well as the 
relatively low energy of that state, RMPH suggest that the observed lines are coUisionally 
excited. Note, however, that no individual identification of the Fe ll lines in the 8500-9300 A 
interval has yet been done in I Zw 1 . 

For all said above, I Zw 1 is a particularly good choice as a test target to construct 
a semi-empirical NIR Fell template as it is so well studied, especially in terms of its 
UV-optical iron emission. The strong and narrow Fe ll emission lines in I Zw l|j, a 
conspicuous characteristic of NLSl galaxies, make it an ideal object for the cons truction of 
such a template and compleraents the ones previously publishe d in both the UV ( Laor et al. 



1997 



Vestergaard fc Wilkes 



200ll ) and in the optical region (JBoroson fc Green 



1992 



^Although I Zw 1 is classified as a Seyfert 1 galaxy, its absolute luminosity {My = - 
23.8 for Ho = 50 km s~^ M pc~^, Q'o = 0) actually qualifies it as a low-luminosity quasar 



( IVeron-Cetty fc Veron 



19911 ). 



-7- 



Veron-Cettv. Jolv fc Veronll2004h . 



and 



Primary cas c ading lines following Lya fluorescence, previously confirmed by RA02 



LandtetaL 



( I2OO8I ) in other NLSls, can also be studied and characterized in this 
source. The main 1 /im diagnostic lines have the advantage of not being heavily blended 
with other multiplets, as normally happens with the Fell optical lines. The observation 
of such features in I Zw 1 can serve as a useful benchmark for photoionization models, 
in particular, for models predicting the complex Fell emission spectrum. Moreover, the 
proposed semi-empirical Fe ll would allow the subtraction of this emission in other AGNs. 
This is important for at least two reasons: to decontaminate other BLR features and to 
evaluate the amount of Fe ll emission present in the NIR region, along with its relationship 
to that of the UV and optical region. 

In this paper we describe the first detailed work to study the Fe ll lines emitted by the 
BLR in IZwl. The aim is twofold: (i) provide tight observational constraints to model the 
Fe II emission in this source and (ii) construct the first semi-empirical template in the NIR 
region. The structure of this paper is as follows: Section 2 describes the observations and 
data reduction. Section 3 discusses the characteristics of the theoretical models used in this 
work. In Section 4 we perform a template fitting to the observed spectrum of IZwl, using 
the theoretical models described in the former section. Section 5 describes the construction 
of the semi-empirical Fe ll template and tests it on the NIR spectrum of Ark 564. A general 
discussion and conclusions are given in Sections 6 and 7, respectively. 

2. Observations and Data Reduction 

Near-infrared spectra of I Zw 1 were obtained at the NASA 3m Infrared Telescope 
Facility (IRTF) on the night of 23 October 2003. The SpeX spectrograph (Rayner et al. 



2003) was used in the short cross- dispersed mode (SXD, 0.8 - 2.4 /im). The detector 
employed consisted of a 1024x1024 ALADDIN 3 InSb array with a spatial scale of 
0.15"/pixel. A 0.8" x 15" slit oriented at the paralactic angle to minimize differential 
refraction was used, providing a spectral resolution of 360 kms~^. This value was 
determined both from the arc lamp and the sky line spectra and was found to be constant 
with wavelength along the observed spectra. 

During the observations the seeing was ~1" in J. Observations were done nodding in an 
ABBA pattern with integration time of 120 s per frame and total on-source integration time 
of 28 minutes. After the galaxy, the AOV star SAO 92128 (V=7.38) was observed as telluric 
standard and flux calibrator. The spectral redu ction, extraction an d wavelength calibration 



2004 



the in-house software 



procedures were performed using SPEXTOOL (1 Gushing et all 
developed and provided by the SpeX team for the IRTF community. An aperture window 
3" wide was employed to integrate all the signal from the galaxy nucleus along the spatial 
direction. Extended emission is likely to be present but it is outside the 3" region. Indeed, 
the FWHM of the I Zw 1 light proflle matches, within the natural seeing fluctuations during 
the observations, that of the telluric standard (0.91" for the former and 0.89" for the latter 
in the i^-band). The root-mean-square (RMS) of the dispersion solution for the wavelength 
calibration was 0.17A. 

The 1-D I Zw 1 spectrum was then correcte d for telluric absorption and flux calibrated 



using Xtellcor (jVacca. Gushing fc Rayner 



20031 ). another in-house software developed by 



the IRTF team. Finally, the different orders of the galaxy spectrum were merged to form 
a single 1-D frame. It was later corrected for the redshift of 2;=0. 061105, determined from 
the average z measured from the positions of Pa(5, He I 1.083/im, Pa/3 and Br7. A Galactic 



^SPEXTOOL is available from the IRTF web site at 



http://irtf.ifa.hawaii.edu/Facility/spex/spex.htnil 



-9- 



extinction correction of E(B-V)=0.065 (jSchlegel. Finkbeiner fc Davislll998l ) was applied. 



Figure [T] shows the final ID spectrum of I Zw 1 already calibrated by wavelength and 
fiux. A visual comparison of our SpeX data with that of RMPH reveals an improvement in 
the spectral resolution along the 0.8-2.2 /im, allowing us to better constrain most spectral 
features, in particular, those that are heavily blended. Moreover, the higher S/N (>150) 
of our spectrum eases the identification of new emission features not detected before. It 
can also be seen that the continuum fiux scale is about 30% lower in the 2;, J andif band 
and 50% lower in the K band than that of RMPH. This discrepancy is very likely due 
to the differences in apertures between the two observations, as our slit width is about 
2.5 X narrower, and therefore encompassing a much smaller contribution of the host galaxy. 
The minimum in the continuum emission, at ~ 13000 A is characteristic of AGNs and is 
interpreted as due to a shift from a nonthermal continuum to the thermal dust emission 
that dominates at longer wavelengths (RMPH; RA02; Riffel, Rodriguez- Ardila & Pastoriza 
2006; Landt et al. 2011). Note, however, that most line fluxes measured in our observation 
(see Table [2]) agree within erros to that of RMPH. 

Line identifications for the most conspicuous lines detected in the NIR spectrum of 
I Zw 1 are indicated in Figure [H 

It is easy to see from Figure [1] that the NIR spectrum of I Zw 1 is rich in Fe ll emission 
features. The so-called l/xm Fe ll lines, for instance, are particularly strong. Significant Fe ll 
emission in t he 8500-9500 A w a veleng th range, very likely produced by Lya fiuorescence as 



predicted by lSigut &: PradhanI (Il998l ). is also observed. Other prominent lines detected in 
IZwl include Hi, He I, Ol and Call. Moreover, forbidden lines of [Fell], and [Slll] and 
molecular H2 were also detected. 

In the following sections we will discuss the method employed for the construction of 
the first semi-empirical NIR Fe ll template published in the literature, suitable to remove 



-10- 

this emission in AGNs after a proper subtraction of the continuum emission and scaling/ 
line broadening of the template. 

3. The Fell models 

Theoretical models of Fe ll including the NIR region are rare in the literature. Up to 
our knowledge, all works published before the year 1998 (Wills, Netzer & Wills 1985 and 
references therein) predicted emission line intensities for the UV and optical regions only. 
The main reason for ignoring the NIR is likely due to the fact that most model predictions 
pointed out to very weak Fe ll emission in that region. Moreover, the lack of good S/N 
observations of AGNs redwards of 8000 A by that time prevented a confrontation between 
models and observations. 



Sigut fc PradhanI ( 119981 ) proposed that the inclusion of Lya fluorescence excitation 
process results in significant NIR Fell emission in the region 8500-9500 A. In these 
calculations, a limited, non-LTE atomic model with 262 fine structure levels, sufficiently 
large for Lya fluorescent excitation, was included. They showed that Lya excitation can be 
of fundamental importance in enhancing the UV and optical Fe ll fluxes. 



Later, ISigut fc PradhanI ( l2003l ) presented improved theoretical non-LTE Fe ll emission 
line strengths for physical conditions typical of active galactic nuclei with broad-line 
regions. In these new set of models, updated to also include the Mgll ion (Sigut, 2004, 
private communication), the Fell line strengths were computed with a precise treatment 
of radiative transfer using exte nsive and accurate atomic data from the Iron Projecc 



( ISigut. Nahar fc PradhanI I2004J ). Excitation mechanisms for the Fell emission included 



continuum fluorescence, collisional excitation, self-fluorescence among the Fe ll transitions. 



^NORAD database at www.astronomy.ohio-state.edu/~nahar 



- 11 - 



and fluorescent excitation by Lya and Ly/3. A larger Fell atomic model consisting 
of 827 fine structure levels (including states to E~15 eV) was used to predict fiuxes 
for approximately 23,000 Fell transitions, covering most of the UV, optical, and NIR 
wavelengths of astrophysical interest. Detailed radiative transfer in the lines, including 
self-fluorescence overlap , was perfomed with an approximate A operator scheme - see 



Sieut fc PradhanI fl2003h for details. 



Currently, owing to the complexity of the observed iron emission in AGNs, this 



emission is typical 



(JBoroson &: Green 



modeled using empirical tera plates derived from speciflc AGN spectra 



1992 



Corbin fc Boroson 



19961 ). A more recent example of application 



of this method consisted in deriving a Fe ll- ill 



the NLSl galaxy IZwl ( IVestergaard fc Wilkes 



template from high-quality UV spectra of 



20011 ). Such templates play a critical role in 



extracting a measure of the total iron emission from heavily blended and broadened AGN 
emission line spectra. Following this approach, and taking advantage of the availability of 
the Fell templates and the NIR spectrum of 1 Zw 1, we will construct the flrst semi-empirical 
Fe II template of that galaxy, complementing the ones existing in the UV and optical 
regions. The approach that we will employ consists of comparing the NIR spectrum of I Zw 1 
with a grid of Fell+Mgll models developed by Sigut & Pradhan (2003), with these latter 
covering a wide range of ionization parameters (f/ion=-l-3, -2 and -3 dex) and densities (wh 
= 9.6, 10.6, 11.6 and 12.6 dex cm~^). The internal cloud turbulent velocity, in all cases, 
was V|;ur=10 kms~^. Fell-I-Mgll spectra were computed for BLR cloud models with typical 
conditions thought to exist in the Fe ll emitting clouds. The calculations have been made 
for traditional clouds of a single specifled density and ionizat ion parameter, as opp osed to 



Baldwin et al. 



(11995[ l. as the 



the more realistic locally optimally emitting cloud models of 

main interest is to study the interplay of the various iron emission excitation mechanisms 

and not the detailed structure of the BLR. 



- 12- 

Figure [2] shows the peak-to- valley intensity variability among the Sigut & Pradhan 
models versus wavelength. It shows a rather large variation in strength in some emission 
lines, such as those contained in the 8300-8500 and 9000-9400 A intervals (including a 
contribution from the Mg emission), Fell at 8927, 9997, 10502, 10862 and 11126 A, as well 
as Mgll at 9218 and 9244 A. Indeed, the amplitude of the variation of these lines are about 
two orders of magnitude or more larger than the median peak-to- valley values. 

4. Analysis Procedure 

In order to estimate a NIR template for I Zw 1 a comparison of its observed emission 
line spectrum with that predicted by the available models was made. The best matching 
model can also indicate the most probable physical conditions of the BLR. As a by-product 
of the template fitting, we also determined the emission line positions/intensities of other 
BLR lines in order to minimize the residual RMS. The whole fitting procedure is decribed 
below. 

The flux calibrated spectrum of I Zw 1 was first continuum-subtracted in order to leave 
us with a pure emission line spectrum. For this purpose, a spline function was fitted to 
the continuum, choosing regions free of emission lines. The fitld task of IRAF was used 
for this purpose, and chosen for simplicity. A more elaborate approach, consisting of a 
simultaneous fitting of the intrinsic AGN continuum, stellar population template and dust 
is out from the scope of this paper. Since the NIR spectrum is very rich in emission lines 
other than Fell, it is also necessary to model them and then perform a multi-parametric fit. 

The Paschen series was modeled using, as a reference, the observed Paa (1.8751 fim) 
profile in velocity space. In a first approach, we set the scaling factors of individual lines 
as free parameters in the fit. Pa/3 is located in a region of strong telluric contamination 



-13- 



and, for that reason, was not considered. Preliminary tests produced results without 
physical significance (i.e. not justifiable by internal reddening and/or deviations of case B 
recombination rates). For that reason, we decided to set limits on the relative intensities 
of such lines consistent with decreasing values as moving towards the bluer part of the 
spectrum, starting from Paj. This behavior was forced through the adoption of reasonable 
boundary conditions in the their scaling parameters. Moreover, Paj/Paj+i line ratios in 
the 9000-12000 A range were allowed to have a 20-30% error, given the uncertainties in 
subtracting the continuum. 

The permitted lines of Call A8498, A8542, A8662, HelA10830, HellA10124 and 
O I A8446, A11287 are also generally blended with the iron multiplets. As the bulk of all 
these lines is produced by the BLR, they were also modeled through the Paa profile, 
allowing some broadening (by up to 80 kms~^ using Gaussian filtering) to improve the 
results. The centroid position of He I and He ll were found to be blueshifted by -332 and 
-622 kms~^, respectively. This may indicate that the lar gest contribution of these two line s 



is produced in the NLR, in agreement with the results of IVeron- Getty. Joly fc Veron 



(12004] ). 



They report two regions emitting broad and blueshifted [O ill] lines in the optical region. 
One of the systems has V=-500 kms~^, compatible to the line shift of the helium lines 
found here. The other system has V=-1450 kms~^. This latter is not detected here, very 
likely due to the lower spectral resolution of our data and its relative weakness compared 
to the strength of the other system. The fact that the iron line profiles do not appear 
asymmetric nor are significantly blueshifted indicates that this emi ssion does not originat e 



i n the outfiowing gas itself. This agrees with the results found by iVestergaard fc Wilkes 



fl200lh . 



Forbidden lines present in this spectral region such as [Gal] A9850 {vj. ~213 kms ^), 
the highly excited [S vill] A9913 {v,. ~-785 kms^^), [Slll] AA9069,9532 (both with Vr ~-300 



-14- 



kms ^) and [Sll] A10280, A10320 have all been assumed to have Lorentzian profiles with 
widths of 600 kms"^. The blue shift found in m ost of these lines are compatible with the 



shift found by 



Veron-Cetty. Joly &: Veron 



( 12004 ) in the optical spectrum, confirming the 
presence of a complex NLR in this object. We have fixed the FWHlVJj of the forbidden 
lines because we verified that by progressively increasing their FWHM the residual RMS 
improved by clearly fitting through them also features of the Fe ll models, a not desirable 
effect. Also, the [Slll] doublet at 9069 and 9532 A were assumed to have their peak 
intensities constrained to the value of 2.4 ([Slll] A9532/[Slll] A9069), as determined by 
atomic physics. The list of all modeled emission lines is shown in table [T] 

Our 12 available Fell+Mgll (or simply the Fell, when not considering the Mgll lines) 
theoretical templates had to be convolved with a line profile representative of the BLR. For 
that purpose, it is useful to examine Figure El where the FellA11126, OlA11287 and Paa 
emission line profiles are plotted in the velocity space. It is easy to see that all three profiles 
can be well represented by a Lorentzian function of FWHM = 875 kms~^ (dashed curve). 
For comparison, the figure also shows a Gaussian profile (dotted) with the same FWHM 
as the Lorentzian. Clearly, the Gaussian fails at reproducing the extense wings observed 
in both lines. Because of the strong similarity of the BLR profiles with the Lorentzian 
curve, we adopted this theoretical profile to convolve ou r models. Notice that this system 



Veron-Cettv. Jolv fc Veronl f l2004h . 



would be equivalent to the relatively broad Ll system of 
associated to the BLR. 

Although it may sound appealing to use the form and width of Fe ll A9997 or 
A10491+10502 to broaden the templates, note that the former is heavily blended with Pa(5 
(narrow and broad components), [S vill] and [Cl], and the latter is a blend of two lines 



^AU FWHM listed in this work refer to the instrumental width, i.e. are not corrected 
from the intrinsic line width of 360 kms^^. 



-15- 



very close in wavelength. Therefore, the characterization of these profiles are more subject 
to uncertainties. As shown above, the form and width of either FellA11126 or OlA11287 
to broaden the Fell +Mgll template provides a similar result. This is c onsistent also 



with previous works (JRodriguez-Ardila et al. 



2002 



Matsuoka et al. 



20071 ) that presented 



consistent observational and theoretical evidence, respectively, that both Fe ll and O I are 
originated in the same parcel of gas. 

Let each line intensity of a given Fe ll+Mg ll (or only Fe ll) template to be denoted by 
Ij (a (5-function at Xj) and the Lorentzian velocity profile to be given by P = P{vij). The 
flux fi at a certain wavelenght Aj of the convolved template is then: 

all j 

with 

Vij = C [{\i - \j)/\j] 

where c is the speed of light. The proportionality constant on equation [1] is a parameter of 
the fit. 

Once defined all the emission line profiles, we proceeded with the least-squares fit of all 
scaling parameters (19 in total for each model, correponding to the lines listed in Tabled] 
plus the convolved template scaling factor jfl. This was carried out using the Fortran routines 
of the Minuit v. 94.1 software, available at the CERN librar}o. The boundary conditions 
provided Hi line ratios (Pa7/Pa5, Pa5/Pa8, Pa8/Pa9, Pa9/Pal0) with an average of ~1.7, 
consistent with the theoretical Paschen decrements within 20—30% error. Note that the 
reduced-x^ provided by the routine does not take into account possible sources of errors 
such as residuals from telluric absorption corrections and continuum subtraction. For this 



^Notice that the S ill lines count as one, since they are constrained in intensity ratio, 
^wwwasdoc. web. cern.ch/wwwasdoc/minuit/minmain. html 



-16- 



reason, we used the RMS of the residuals as a measure for the quahty of the fit. The residual 
RMS obtained from the fit of each of the 12 models, calculated in the the main region of 
interest (8300—11600 A), is shown in Figure IH These results are used as a discriminant of 
the best suitable model. 

Models with medium/low ionization parameters and large densities such as (log f/ion=- 
2.0, log nH=12.6) and (log [/ion=-l-3, log raH=12.6) are the most successful ones, having 
residual RMS between 15-20% smaller with respect to the average (around 30% smaller 
considering the worst models). For consistency, we decided to check the effect of the Mgll 
emission in the fits. For this purpose, we performed the same fitting procedure using only 
the Fe ll multiplets in the models. This approach is justified by the large uncertainties 
concerning the Mgll strength in the models, specially in the spectral region of 9200 A. 
The excitation mechanism for these Mgll lines is Ly/3 fluorescence, and the uncertainties 
come from the small transition probabilities between the levels 3s ^S — )► 5p ^P°. Moreover, 
the pumping source function for Ly/3 is more uncertain than for Lya due to the fact that 
photons in Ly/3 can also cycle through Ha, and this "cross- redistribution" can be important. 

Considering that the number of Mg ll lines included in the Fe ll-|-Mg ll models is less 
than 1% of that of Fell, the exclusion of Mgll does not affect significantly the fit, as 
expected. The RMS values of the fit residuals considering only the Fe ll multiplets are also 
shown in Figure H] as solid symbols. The results are very similar to those obtained using 
the composite Fell-I-Mgll models, favoring a high-density and moderate/low ionization 
parameter. 

Another consistency check in the results was made considering only the 8300- 
11600 A region in the fit, where the main Fell diagnostic lines are located. In the above 
interval only Pa7 to PalO were included in the fit, reducing the number of free parameters 
to 18. It turned out that either by fitting the larger 8300-20000 A or just the 8300-11600 A 



-17- 

interval one obtains similar results. For this reason, and also because of the lack of 
significant Fell+Mgll emission redwards of 11600 A, we will concentrate in the remainder 
of this paper to this smaller spectral region. 

Figures |5] and |6] show the best fits for all the models. Table [2] lists the emission line 
intensities derived from such fits. 



5. Semi-empirical NIR Fell+Mgll Template 

In the previous section we found that the model with log U=-2 and log nH=12.6 is 
the one that best represents the observed Fell+Mgll emission in IZwl. Although not 
crucial at this point, it is important to question if that model is consistent with the physical 
conditions of a Fe ll emission region believed to exist in AGNs. 



Jolyi (jl99ll ) computed purely collisional models showing that low temperature 
(T <8000 K), high density {nu > 10^^ cm"^) and high column density (A^(H)> 10^^ cm~^) 
clouds provide Fell opt /H/3 in good agreement with observations of Seyfert 1 galaxies. 
Detailed modeling of the Fell us ing Cloudy and inclu ding Lya fluorescence as one of the 



excitation mechanisms made by iBaldwin et al. 



( I2OO4I ) points out to similar conditions. 



Indeed, densities between 9 < ran < 13 and log f/ion=-l-4 are consistent with observations. 
Additional observational evidence that strong i ron (Fell and Fel ll ) emi s sion may be 



connected wit 



1 high densities was provided 



Baldwin et al. 



(1l996h 



Lawrence et al. 



by 



Hartig fc BaldwinI ( 



1986|li 



Kuraszkiewicz et al. 



Jolvl (119911 ): 



(120001 ). From that 



(119971) and 

we can see that the physical conditions of the best matching model for I Zw 1 are, in general 
terms, representative of the Fe ll emitting region in AGNs. 

In the remainder of this section we analyze in detail the best Fe ll+Mg ll template and 
propose modifications to it, based on the mismatches between the observed spectrum and 



-18- 



the convolved model. We then present a semi-empirical template capable of minimizing the 
residual RMS in the NIR window of 8300-11600 A for I Zw 1. 

The equation [1] for the convolution of a given model can be succinctly rewritten in a 
matricial form: 

f = pe (2) 

where / denotes a spectrum, the N- length vector of Fell-I-Mgll fluxes for a given 
wavelength range, P the NxM convolution matrix and £ the M- length vector with the set 
of Fell and Mgll line intensities which contribute to the wavelength range comprised by /. 
Neglecting any measurement errors, we can estimate / from the residuals of the observed 
spectrum after the subtraction of all emission lines (except Fell and Mgll), which we call 
hereafter f^bs- The convolution matrix can be easily built from the Lorentzian velocity 
profile (FWHM=875 km s~^) and the Vij differences between each given multiplet line Ij 
and the wavelength which corresponds to /j. Adjusting the template line intensities to 
match as close as possible the observed residuals then constitutes a deconvolution problem, 
for which several solving approaches exist (see, for instance, a discussion on deconvolution 
techniques in Thiebaut 2005). 

The number of Fell+Mgll lines comprised between 8300 and 11600 A are 1529, 
out of which only a few contribute significantly to the emission spectrum. The Wiener 
deconvolution was chosen for our problem for its simplicity. However, it can lead to an 
unrealistic overdensity of line contributions, along with some unphysical results, if we 
consider all of the 1529 line intensities as free parameters. For this reason, we have decided 
to build a £* vector containing only the contribution of multiplets with the ~3% largest 
intensity contributions (selected from the best fitted model), plus some others whose 
intensities are clearly underestimated in the models (e.g. the unusual two-peaked bump 
around 11400 A; see discussion below). In practice, this means zeroing the elements in the 



-19- 



t vector (to be multiplied by P) which fall outside the selection criteria. The intensity 
threshold for building t* lies at about 4% of the peak intensity in this wavelength range, 
and it is at least twice higher than the template median intensity. This criterium produces 
a t with 51 elements. 

In order to account for the spectral contribution of the Fe ll+Mg ll multiplets not 
present in the € vector, which we call hereafter the "background" spectrum /^, we first 
subtract their emissioiij from the input spectrum. This allows us to rewrite equation ([2]) as: 

f obs — fobs ~ fb = ^ ^ (3) 

Note that /^ is scaled accordingly to the fitted parameter of Fell+Mgll found in the 
previous section. The line intensities in the vector £* found from the deconvolution of /*^g 
plus the "background" subset of line intensities provide what we call the semi-empirical 
Fell+Mgll template. 

A quick test used as f^f^^ a spectrum produced by convolving our best model, which 
we called f model- The deconvolution technique described above was then applied to check 
whether we could reliably recover the original £* , and consequently £. Convolving this 
"deconvolved" vector then produced f deconv The difference of this latter with f model 
showed a RMS of less than 2% (relative to the intensity at 9997 A). The reliability of this 
result is also shown in Figure [3, where the initial normalized model intensities {£*modei) ^^^ 
plotted against the ones found from the deconvolution {£*iieconv)- Despite of some scattering, 
specially for the less intense lines, the agreement is very good. 

When tackling real data, we decided to make a small tuning to the £* vector. In order 
to conserve the physical interpretation given by the best model (-2.0,12.6) with predictions 



^The "background" spectrum /f, is obtained by convolving the low intensity lines selected 
from the best model with the Lorentzian profile. 



-20- 



about the intensities of the most prominent Fe ll hnes, we decided to discard them from 
the t vector. In practice this means that Fell A9997, A10501, A10862, A11126 (and the 
close neighbors Fell A10491, A10871) had their original values preserved, and have been 
included in the /^ contribution. We should also make a brief remark on the relatively strong 
two-peaked bump of Fell observed at ~11400 A: this feature is underpredicted by any of 
the models, and because of its strength, we ruled out the possibility of being part of the 
spectrum noise or a feature introduced by the OlA11287 profile fitting. Telluric absorption 
corrections are not critical in this region and, besides that, in the model templates one 
can notice two small peaks around this region, not exactly coincident in wavelength with 
the observed ones. These arguments give us support to believe that these features, located 
around 11381 and 11402 A, are real. Notice that these wavelengths, estimated by visual 
inspection, are close to 11380.32 and 11403. 54A, which refer to features that appear in 
absorption in the models. We decided to include these two hypothetical lines in the £* 
vector, as well as those around 10686 A, clearly underestimated in the models but visible in 
f*obs- This left us with a final 50-element C vector. 

We then applied the deconvolution to the observed spectrum (after the /f, subtraction). 
The intensities obtained from the deconvolution method applied to the observed Fe ll+Mg ll 
I Zw 1 spectrum are shown in Table [3] together with those given by the best model. These 
values are all normalized with respect to the intensity of the Fe ll A9997 line. For building 
the semi-empirical Fell-I-Mgll spectrum we convolved this newly derived intensity vector 
with the P matrix and added back the background contribution. 

The total semi-empirical template (corresponding to £, according to our notation) in 
the region of interest (0.83-1.16 /im) is shown in Figure El on top. The bottom panel shows 
an expanded view (x25) of the template, in order to highlight the contribution of the less 
intense lines. 



-21 - 



Figure [9] shows the resuhing semi-empirical spectrum (bottom), as well as the one 
derived from best model (top), for comparison, superposed on the observed Fell+Mgll 
spectrum of I Zw 1. The semi-empirical template reduces by ~31% the RMS of the observed 
spectrum with respect to the best model. The bottom figure also indicates the lines whose 
estimated intensities varied more significantly with respect to those given by the model. 
Notice that negative intensities obtained from the deconvolution process have been zeroed 
in the computation of the estimated spectrum, but are shown in table |3] for completion. 
Such negative line intensities are consistent with zero within 2cr of the spectral residuals. 



Landt et al. 



model of 



(|2008[) [se e their Section 5.4.1.] have compared the predicted Fell 



Si gut fc PradhanI (120031 ) to observations and found a discrepancy in the the 



Fell 1.0491+1.0502 /im and Fell 1.0174 /im emission lines, in the sense that the former were 
overpredicted whereas the latter were underpredicted by th eory, and this by a similar factor 



Sigut fc PradhanI fcoosh with 



of ~2. Note that they used the predictions of model A of 
logf/ion=-2 and lognH=9.6. Our best matching model has the same ionization parameter 
but a density that is three orders of magnitude larger (lognH=12.6). As can be seen in 
Table |2] and Figures [5] and IH our best model leads to a better fit of the most conspicuous 
lines including Fell 1.0491+1.0502 /im. Indeed, the discrepancy is removed in this pair of 
emission lines. Fell 1.0174 /xm, on the other hand, continues to be underpredicted even in 
the best model. Actually, this was one of the iron features that needed a fine tuning: its 
strength increased in the semi-empirical template by a factor of almost 3 (see Figure [TT| 
when a comparison between the model and the semi-empirical template is done). Very 



likely, atomic parameters for that line would need to be reviewed in future modeling. 



-22- 



5.1. Application to Ark 564 



Ark 564 is another well-known NLSl with strong Fell emission, suitable to test our 
semi-empirical template. The spectrum of this AGN was observed and reduced in a similar 



way t o that of of I Zw 1. Detai l s of t 
are in 



Rodriguez- Ardila et al. 



le extraction, reduction, wavelength and flux calibration 
( l2004l ). As previously, we fitted and subtracted the continuum 
emission through a spline function. For modeling the line profiles, we adopted Paa as 
representative of the emission line profile to be convolved with the Fe ll+Mg ll intensity 
vector derived from IZwl, and also as a template for other permitted emission lines. 
Forbidden lines were assumed to have Gaussiaio profiles with widths of 500-550 km s^^. A 
quick fit over all the theoretical models also indicates a very high density and low/moderate 
ionization parameter for the Fe ll emitting region, as shown in Table H] with the estimated 
RMS over the 8300-11600 A region. The sharpness of the emission lines makes small 
uncertainties in emission line positioning and scaling to give rise to high amplitude residuals 
in the fit, what explains the larger observed residual RMS. Neverthless, the general trend is 
similar to that observed for IZwl, favoring similar physical conditions for the Fell emitting 
region. 

Figure [10] is similar to figure [H but now showing the suitability of our derived 
semi-empirical template to the observed Fe ll-|-Mg ll emission of Ark 564. A smaller RMS 
in the residuals with respect to those obtained from the theoretical models confirms the 
usefulness of the semi-empirical template to represent the Fe ll and Mg ll emission in AGNs. 
In general, lines that needed a fine tuning in IZwl, like Fell 10174 A, properly reproduce 
the observations. Note, however, that the bump observed at 11400 A in IZwl and that we 
deliberately introduced in the template seems to be unusually large with respect to the one 



^Ark 564 shows sharper forbidden emission features in its spectrum, so we decided to use 
a Gaussian instead of a Lorentzian to model them. 



-23- 

in Ark 564. Clearly, the transitions leading to these particular set of lines should strongly 
depend on local physical parameters that vary from object to object. 

6. Discussion 

The analysis carried out in the previous sections confirms several pieces of evidence 
already suggested by other works, which are: (i) Lya fluorescence is indeed a process that 
should be taken into account in any systematic study of the Fe ll emission in AGNs as it 
produces a considerable amount of emission lines that otherwise would be absent. This is 
particularly evident for the 9200 A feature, composed of numerous Fell multiples as well 
as some contribution from Mgll lines, (ii) Similarly to the optical region, the NIR Fell 
emission also produces a subtle pseudo-continuum, particularly in the region between 8600 
and 10000 A (see Figure |8]). Without a proper modeling and subtraction of this emission, 
fluxes of other BLR and NLR features can be severely overestimated. (3) Unlike in the 
optical region, individual Fell emission lines can be isolated in the NIR (i.e., the lines at 
10501 A, 10862 A and 11126 A). This is particularly useful to characterize, for instance, the 
line profiles and emission line fluxes, in an already complex emission region. This individual 
Fe II line charaterization can be possible because other Fe ll lines very close in wavelength 
to the above three are at least 25 x weaker. 

The above thoughts can better be visualized in Figure [HI which shows the spectrum of 
1 Zw 1 with (top) and without (bottom) the Fe II+Mg II contribution. The spectrum "clean" 
of iron and magnesium emission highlights the remaining emission lines, properly identified 
in the figure. Small spikes, coincident with the position of the strongest Fe ll lines can still 
be seen as, for example, in the blue part of Pa6, between He I and Pa7 and at the position 
of the FellA11126 line . We interpret these small r esidua ls as a due to emission from the 



NLR, as suggested by 



Veron- Getty. Joly fc VeronI ( 12004J ). It might indicate that a more 



-24- 



complex modeling of the convolving profile might be required, for example, by including a 
possible contribution from the NLR. However, because we are primarily interested in the 
construction of a semi-empirical template for the Fe ll+Mg ll emission which originates at 
the BLR, these small residuals are out of concern. 

Figure [T^ shows the same as figure [TT] but for Ark 564. It can be seen that the spectrum 
at the bottom is nicely clean of Fell and Mgll, as evidenced by the small residuals left in 
the region between 10200-10600 A. It exemplifies the use of our NIR IZwl template to 
remove that emission in other AGNs. 

At this point we call the attention to an apparent absorption feature blueward of 
HelA10818 that appeared after the subtraction of the Fell+Mgll semi-empirical template 
in Ark 564. It is possible that this feature is artificial and due to a bad subtraction of 
the Fell blends with peak at 10750 A. Yet another possibility is that it c ould be a real 



feature, similar to the one reported by 



Leighly. Dietrich. &: Barberl ( 120111 ) in the quasar 



FBQ,SJ1151 +3822, which thev attribute to a 



that source. 



Droad absorption line (BAL) system in 



Leighly. Dietrich, fc Barberl (1201 if ) discussed the prospects of finding other 



HelA10830 BALQSOs on six additional objects and pointed out that several well-known, 
bright low-redshift BALQSOs have no HelA10830 absorption, a fact that can place upper 
limits on the column densities in those objects. Observations with higher spectral resolution 
are needed to confirm if the absorption in Ark 564 is indeed real. Although it is out of 
the scope of this paper the study of such a system of absorbers in the BLR of Ark 564 or 
in other sources, our results strengthen the need of an adequate Fe ll subtraction around 
Hel A10830 to further constrain the inner physical properties of such AGNs. 

Note also that the peak observed around 10740 A is due to [Fexill] emission. In 
addition, the 11400 A bump is overestimated in the semi-empirical template, meaning that 
some particular lines may need a fine tuning for a better match. It implies also that not 



-25- 

all NIR Fe ll lines may scale up by the same factor, as would be expected. However, as in 
the optical and UV regions, the semi-empirical template suitably reproduces most of the 
observed Fe ll+Mg ll emission features in the NIR. Clearly, testing the template in a large 
number of objects is necessary to verify its suitability in more general terms. 

At this point is important to draw our attention to how significant is the NIR Fell 
emission in I Zw 1 compared to that of the optical and UV. For this purpose we measured 
the integrated flux in the 8300—11600 A region using the Fell semi-empirical template 
derived for that object. We found that Fpeii = 2.79 ± 0.20 x ip- is ergcm~^s~^ . Note 



Tsuzuki et al 



(120061 ) 



that the Mgll was not taken into account in the computed value, 
measured the integrated flux of Fe ll for I Zw 1 using HST and ground-based observations 
of this galaxy in five different wavelength bands: f/1 [2200-2660 A], U2 [2660-3000 A], U3 
[3000-3500 A], 01 [4400-4700 A] and 02 [5100-5600 A]. The values they found, relative to 
H/3, are shown in Table |5l 

It can be seen that nearly half the amount of II/3 flux in I Zw 1 is emitted by Fe ll in the 
NIR. Moreover, the integrated NIR iron emission carries a flux that is equivalent to ~10% 
of the Fell emission in the optical region (sum of the fluxes in the 01 and 02 intervals of 
Table [5D, to ~10% of the Fell in the near-UV region (3000-3500 A) and to ~3% of the Fell 
emission in the UV (2200-3000 A). 

At first sight the above numbers may indicate that the role of Lya pumping, which is 
behind most of the Fe ll NIR emission, is negligible. However, we should take into account 
that after the cascading transitions that result in the NIR emission the z ^D and z ^F levels 
are populated. These latter are responsible for part of the transitions leading to the Fe ll 
emission in the 01 and 02 optical regions. Therefore, a 10% in flux means that up to 20% 
of the optical Fe ll photons can be attributed to Lya fluorescence. 



-26- 

7. Conclusions 

We have compared theoretical models for the Fell+Mgll NIR emission in active 
galaxies, which have the same turbulent velocities in the medium, but differ in physical 
conditions such as density and degree of ionization of the emitting region. For that purpose, 
we chose to model the NLSl galaxy IZwl, which has traditionally provided the template 
for the Fe ll emission in the optical and UV. 

The best match among all models was obtained by comparing the results of a 
multi-parametric fit comprising the main emission line features to the observed spectrum 
in the region 0.83-1.16 yum. Low/moderate ionization parameters and high gas densities of 
]^gi2.6 Qjji"^) are favored, as they reduce the residuals of the fitted spectrum. However, since 
some Fe ll lines are clearly underestimated even by the best models, we decided to derive 
a semi-empirical template to improve the fitting, by adjusting the intensities of the most 
prominent Fell and Mgll lines, taken from the best fitted model. 

The newly derived template reduces the fit residuals by about 31% with respect to what 
is obtained using the best theoretical model. We performed a quick check on the spectrum 
of another NLSl with conspicuous and very narrow emission lines. Ark 564. Despite some 
small differences, this test corroborated the reliability of this new semi-empirical template 
to reproduce AGN Fe ll and Mg ll emission lines in the NIR. 

We also highlight that the Fell bump around 11400 A which we introduced to match 
the observed spectrum of 1 Zw 1 does not reproduce well the same observed feature of 
Ark 564. This lead us to the conclusion that this emission is abnormal in IZwl, given 
that none of the models too could predict such a large emission. Also, I Zw 1 might 
contain a narrow contribut i on to the Fell spectrum, as suggested in an optical study of 



Veron-Cetty. Joly fc VeronI ( 120041 ). Further tests will be futurely carried out on a larger 



sample, in order to fine tune our derived template. 



-27- 

AGR acknowledges Institute Nacional de Ciencia e Tecnologia de Astrofisica (INCT-A) 
for funding support under process CPNq 573648/2008-5. ARA acknowledges CNPq 
for partial support to this research through grant 308877/2009-8. AKP would like to 
acknowledge partial support from the U.S. National Science Foundation, and Sultana 
Nahar for the Iron Project data. We thank to an anonymous Referee for useful corrections 
suggested to the manuscript. 



-28- 

REFERENCES 

Baldwin, J.A., Ferland, G., Korista, K., Verner, D., 1995, ApJ, 455, L119 

Baldwin, J. A., Ferland, G.J., Korista, K.T., Carswell, R.F., Hamann, F., Phillips, M.M., 
Verner, D., Wilkes, B.J., & Williams, R.E., 1996, ApJ, 461, 664 

Baldwin, J. A., Ferland, G. J., Korista, K. T., Hamann, F., LaCluyz, A., 2004, ApJ, 615, 
610 

Boroson, T.A., & Green, R.F., 1992, ApJS, 80, 109 

Collin-SouflFrin, S., Joly, M., Pequignot, D., & Dumont, S., 1986, A&A, 166, 27 

Corbin, M.R., Boroson, T.A., 1996, ApJS, 107, 69 

Gushing, M., Vacca, W.D., Rayner, J.T., 2004, PASP, 116, 362 

Gaskell, G. M. 2009, New Astronomy Reviews, 53, 7 

Graham, M.J., Glowes, R.G., & Gampusano, I.E. 1996, MNRAS, 279, 1349 

Johansson, S., & Jordan, G., 1984, MNRAS, 210, 239 

Joly, M. 1991, A&A, 242, 49 

Hamann, F., & Persson, S.E., 1989, ApJS, 71, 931 

Hartig, G. F., & Baldwin, J. A. 1986, ApJ, 302, 64 

Kelly, D.M., Rieke, G.H., & Gampbell, B., 1994, ApJ, 425, 231 

Kuraszkiewicz, J., Wilkes, B. J., Gzerny, B., & Mathur, S. 2000, ApJ, 542, 692 

Landt, H., Bentz, M.G., Ward, M.J., Elvis, M., Peterson, B.M., Korista, K.T., & Karovska, 
M., 2008, ApJS, 174, 282 



-29- 

Landt, H., Elvis, M., Ward, M. J., Bentz, M. C, Korista, K. T., & Karovska, M. 2011, 
MNRAS, 414, 218 

Laor, A., Jannuzi, B.T., Green, R.F., & Boroson, T.A. 1997, ApJ, 489, 656 

Lawrence, A., Elvis, M., Wilkes, B. J., McHardy, I., & Brandt, W. N. 1997, MNRAS, 285, 
879 

Leighly, K.M., Dietrich, M., Barber, S., 2011, ApJ, 728, 94 

Lipari, S., 1994, ApJ, 436, 102 

Matsuoka, Y., Oyabu, S., Tsuzuki, Y., & Kawara, K., 2007, ApJ, 663, 781 

Netzer, H., & Wills, B.J., 1983, ApJ, 275, 445 

Penston, M.V., 1987, MNRAS, 229, IP 

Phillips, M. M., 1977, ApJ, 215, 746 

Phillips, M. M., 1978, ApJS, 38, 187 

Riffel, R., Rodriguez- Ardila, A., Pastoriza, M., 2006, A&A, 457, 61 

Rodriguez-Ardila, A., Viegas, S.M., Pastoriza, M.G., & Prato, L., 2002, ApJ, 565, 140 

Rodriguez-Ardila, A., Pastoriza, M.G., Viegas, S.M., Sigut, T.A.A., & Prahan, A.K., 2004, 

A&A, 425, 457 

Rudy, R.J., Mazuk, S., Puetter, C., & Hamann, P., 2000, ApJ, 539, 166 

Sigut, T.A.A., & Pradhan, A. K., 1998, ApJ, 499, L139 

Sigut, T.A.A., & Pradhan, A. K., 2003, ApJS, 145, 15 

Sigut, T.A.A., Nahar, S.N., & Pradhan, A.K., 2004, ApJ, 611, 81 



-30- 

Schlegel, D.J., Finkbeiner, D.P., Davis, M., 1998, ApJ, 500, 525 

Sulentic, J. W., Marziani, P., & Dultzin-Hacyan, D. 2000, ARA&A, 2000, 38, 521 

Thiebaut, E., 2005, Optics in astrophysics: Proceedings of the NATO Advanced Study 

Institute on Optics in Astrophysics, Foy, R., & Foy, F.-C, Dordrecht: Springer, 198, 
397 

Tsuzuki, Y., Kawara, K., Yoshii, Y., Oyabu, S., Tanabe, T., & Matsuoka, Y. 2006, ApJ, 
650, 57 

Vacca, W.D, Gushing, M.C., Rayner, J.T., 2003, PASP, 115, 389 

Veron-Cetty, M.-R, Joly, M., & Veron, P., 2004, A&A, 417, 515 

Veron-Cetty, M.-R, & Veron, P., ESO Scientific Report, 1991, 5th edition 

Vestergaard, M., & Wilkes, B.J., 2001, ApJS, 134, 1 

Wills, B.J., Netzer, H., & Wills, D., 1985, ApJ, 288, 94 



This manuscript was prepared with the A AS lATfrjX macros v5.2. 



-31 - 



Table 1. Template lines that were added to the Fell+Mgll fitting process. 



Line 


Arest (A) 


■^meas (A) 


Profile 


Pact 


18750 


18750 


itself 


Pa7 


10937 


10937 


Pao 


Pa<5 


10049 


10049 


Paa 


Pa8 


9544 


9544 


Paa 


Pa9 


9230 


9230 


Paa 


PalO 


9014 


9014 


Pao 


Can 


8498 


8498 


Pao 


Can 


8542 


8542 


Pact 


Can 


8662 


8662 


Pact 


Oi 


8446 


8446 


OIA11287 


Oi 


11287 


11287 


OIA11287 


He I 


10830 


10818 


Pao 


Hen 


10124 


10103 


Pao 


[Cal] 


9850 


9857 


Lorcntzian 


[Sii] 


10280 


10280 


Lorentzian 


[Sii] 


10320 


10320 


Lorcntzian 


[8111]=' 


9069 


9060 


Lorcntzian 


[8111]=' 


9532 


9521 


Lorcntzian 


[8vin] 


9913 


9888 


Lorcntzian 



Line peak intensity ratio fixed in 2.4. 



Table 2. Fluxes of IZw 1 emission lines obtained from the fitting of Fell+Mgll (in units of 10 erg s cm 







logUi 


on — "J 






logC/i 


on — -^ 






log t/ion=-1.3 




lognn 


9.6 


10.6 


11.6 


12.6 


9.6 


10.6 


11.6 


12.6 


9.6 


10.6 


11.6 


12.6 


Pa7 (10937A) 


9.18 


9.16 


9.01 


8.64 


8.81 


8.96 


8.64 


8.64 


8.82 


8.64 


8.64 


8.64 


Pa5 (10049A) 


4.41 


4.41 


4.41 


4.41 


4.41 


4.41 


4.41 


4.41 


4.41 


4.41 


4.41 


4.41 


Pa8 (9544A) 


2.78 


2.78 


2.78 


2.78 


2.78 


2.78 


2.78 


2.78 


2.78 


2.78 


2.78 


2.78 


Pa9 (9230A) 


2.22 


2.22 


2.22 


2.22 


2.22 


2.22 


2.22 


2.22 


2.22 


2.22 


2.17 


2.17 


PalO (9014A) 


1.43 


1.43 


1.39 


1.26 


1.38 


1.40 


1.26 


1.24 


1.38 


1.32 


1.17 


1.12 


He I (10830A) 


18.9 


18.9 


18.7 


17.8 


18.6 


18.8 


18.0 


17.3 


18.7 


18.4 


17.3 


17.4 


Hen (10124A) 


1.73 


1.73 


1.75 


1.59 


1.61 


1.73 


1.63 


1.39 


1.63 


1.67 


1.49 


1.40 


O I (8446A) 


7.21 


7.21 


7.15 


7.03 


7.14 


7.17 


7.04 


7.30 


7.14 


7.15 


7.28 


7.30 


Oi (11287A) 


4.62 


4.63 


4.62 


4.52 


4.60 


4.62 


4.55 


4.47 


4.62 


4.58 


4.50 


4.44 


Can (8498A) 


3.63 


3.62 


3.54 


3.06 


3.57 


3.57 


3.21 


3.09 


3.56 


3.45 


3.19 


3.00 


Can (8542A) 


5.93 


5.93 


5.92 


5.86 


5.92 


5.92 


5.88 


5.79 


5.92 


5.90 


5.83 


5.75 


Can (8662A) 


5.19 


5.19 


5.18 


5.10 


5.17 


5.18 


5.12 


4.95 


5.17 


5.16 


5.04 


4.83 


[Cai] (9850A) 


0.25 


0.25 


0.25 


0.24 


0.24 


0.25 


0.24 


0.22 


0.24 


0.24 


0.23 


0.23 


[Sii] (10280+10320A) 


1.00 


1.00 


0.99 


0.87 


0.98 


1.00 


0.91 


0.79 


0.99 


0.96 


0.84 


0.84 


[Sin] (9530+9068A) 


3.03 


3.03 


2.99 


2.85 


2.96 


2.99 


2.86 


2.89 


2.96 


2.93 


2.88 


2.83 


[Sviii] {9913A) 


0.51 


0.51 


0.50 


0.43 


0.49 


0.50 


0.45 


0.35 


0.49 


0.48 


0.39 


0.35 


Fc 11 (9997A) 


0.34 


0.34 


0.45 


1.99 


0.69 


0.45 


1.49 


3.78 


0.68 


0.84 


2.77 


3.73 


Fell (10501A) 


0.35 


0.36 


0.50 


2.06 


0.74 


0.49 


1.68 


3.16 


0.74 


0.99 


2.37 


1.94 


Fen (10863A) 


0.28 


0.31 


0.43 


1.69 


0.65 


0.42 


1.43 


2.41 


0.67 


0.85 


2.33 


1.86 


Fen (11126A) 


0.14 


0.16 


0.25 


1.11 


0.33 


0.23 


0.89 


1.70 


0.34 


0.49 


1.51 


1.37 


Fe n (9000-9400A)« 


6.37 


6.38 


6.81 


8.93 


7.14 


6.70 


8.53 


7.80 


7.12 


7.23 


7.82 


8.44 


Mgn (8300-11600A)^ 


0.93 


0.97 


1.32 


2.47 


1.74 


1.51 


2.46 


2.75 


1.75 


3.25 


3.51 


3.77 



00 
to 



''Integrated fluxes of all lines in the interval. 



00 
00 



-34- 



Table 3. Relative intensities of Fell and Mgll lines in the i* vector as given by the 

original best model (-2.0, 12.6) and by the one computed from the observed spectrum. 

Values have been normalized with respect to the intensity of the 9997 A line. 



A (A) 


^model 


^deconv 


Ion 


A (A) 


^model 


deconv 


Ion 


8213.99 


0.055 


0.032 


Mgii 


9204.05 


0.093 


0.132 


Fen 


8228.93 


0.057 


0.043 


Fell 


9218.25 


0.304 


0.274 


Mgn 


8234.64 


0.111 


0.049 


Mgii 


9244.26 


0.186 


0.151 


Mgn 


8287.85 


0.103 


0.350 


Fen 


9251.72 


0.080 


0.146 


Fell 


8357.18 


0.062 


0.212 


Fen 


9272.16 


0.062 


0.006 


Fell 


8423.87 


0.076 


0.102 


Fen 


9296.85 


0.046 


0.028 


Fen 


8450.99 


0.160 


0.127 


Fell 


9297.23 


0.102 


0.027 


Fen 


8469.22 


0.066 


-0.050 


Fen 


9303.59 


0.071 


0.004 


Fen 


8490.05 


0.124 


0.085 


Fen 


9326.93 


0.048 


-0.089 


Fen 


8499.56 


0.091 


-0.034 


Fen 


9406.67 


0.041 


0.190 


Fen 


8508.61 


0.051 


0.220 


Fen 


9572.62 


0.061 


0.140 


Fen 


8926.64 


0.158 


0.040 


Fen 


9661.15 


0.041 


-0.107 


Fen 


9075.50 


0.102 


0.074 


Fen 


9956.25 


0.103 


0.088 


Fen 


9077.40 


0.080 


0.088 


Fen 


10173.51 


0.081 


0.223 


Fen 


9095.07 


0.087 


-0.007 


Fen 


10402.83 


0.042 


0.260 


Fen 


9122.94 


0.202 


0.103 


Fen 


10546.38 


0.096 


0.062 


Fen 


9132.36 


0.168 


-0.007 


Fen 


10685.17 


<0.001 


0.091 


Fen^ 


9155.77 


0.071 


-0.022 


Fen 


10686.80 


<0.001 


0.069 


Fen'' 


9171.62 


0.040 


0.045 


Fen 


10686.92 


<0.001 


0.067 


Fen^ 


9175.87 


0.169 


0.038 


Fen 


10749.72 


0.042 


0.308 


Fen 


9178.09 


0.092 


0.034 


Fen 


10826.50 


0.060 


-0.159 


Fen 


9179.47 


0.103 


0.032 


Fen 


10914.24 


0.110 


-0.014 


Mgn 


9187.16 


0.086 


0.038 


Fen 


10951.78 


0.078 


-0.021 


Mgn 


9196.90 


0.055 


0.057 


Fen 


11381.00 


<0.001 


0.059 


Fen'' 


9203.12 


0.075 


0.120 


Fen 


11402.00 


<0.001 


0.127 


Fen'' 



^ Lines included by visual inspection of the residual spectrum, after applying the thresh- 
old criterium for line selection. 



-35- 



Table 4. Residual RMS (in 10 ^^ erg s ^ cm ^ A ^) obtained from the fit of Ark 564, 
using the models and the template derived from I Zw 1. 



log nH=9.6 
log nH=10.6 
log nH=11.6 
log nH=12.6 



log Uion 

-3.0 -2.0 -1.3 



0.466 0.447 0.447 

0.459 0.462 0.436 

0.455 0.421 0.396 

0.409 0.392 0.392 



Semi-empirical 



0.377 



Table 5. Integrated emission line fluxes for I Zw 1 in the UV, optical and near-infrared 

regions. 



Spectral Region^ Integrated Flux Ratio'' 



Reference 



(71 (2200-2660) 
(72 (2660-3000) 
US (3000-3500) 

01 (4400-4700) 

02 (5100-5600) 
NIR (8300-11600) 



8.06±0.39 
5.45±0.27 
4.50±0.23 
2.65±0.18 
2.68±0.16 
0.46±0.02 



Tsuzuki et al. f2006^ 
Tsuzuki et al. (2006) 
Tsuzuki et al. (2006^ 
Tsuzuki et al. (2006) 
Tsuzuki et al. (2006) 
this work 



^(71, (72, (73, Ol, 02 and NIR denote the integrated Fell emission in 
the intervals (in A) shown between brackets, relative to H/3 flux. 

Values ar e relative to the integrated H/3 flux of 6.08it0.24 
ergcm~^s~^ I Tsuzuki et al. 



20061'! 



»< 



OB 




-36- 



[Fell] 



I I 1 1 

[SIII] + HI i^„ Fell 



_L 



_L 



_L 



_L 



IZwl 




X 



_L 



X 



8000 10000 12000 14000 16000 18000 20000 22000 

Wavelength [A] 



Fig. 1. — NIR SpeX spectrum of I Zw 1, from 8000 A to 22100A rest wavelength. Prominent 
emission lines are identified. 



-37- 



> 

Q_ 

> 

a. 



1.0 


-n 






1 1 1 ■ 1 


' 1 ' 

+ 




' 1 ' 


■ Ill 


0.5 


_ 










+ 


+ 


_ 


0.0 


ik 


^ 


Ufebr 


+ o 




+ 


+ 


■ 


-J 






• 1 • ■ • 


. 1 . 




. 1 . 





9000 



10000 



11000 



Wavelength [A] 



Fig. 2. — Peak-to-valley variability of the Fell+Mgll models at each wavelength, normalized 
by its value at 9997 A, for the 8300-11600 A region. Open circles refer to the Mgll emission. 



-38- 



X 

ZJ 



■D 
CD 

N 

"cB 



1.0 - 



-| — I — I — I — I — I — I — r 



I I I 



I I I I 




AV[kms-^] 



Fig. 3.— The normalized OlA11286 A (solid), Paa (solid-bold) and FellA11126 A (dot- 
dashed) profiles. Gaussian (dotted) and Lorentzian (dashed) functions with the same FWHM 
of 875 kms^^ are also shown for comparison. 



-39- 







— 1 




1 1 1 






1— 


7 0.16 


— 












— 


o< 


■ 






f 


■ 




■ 


7 
if) 


♦ 






* 


D 




- 


CM 

1 


C 






• 






. 


£ ... 








o 








o 0.14 


~ 












— 


O) 


. 








+ 




. 


1 


- 








* 


■ 


- 


O 


■ 










D 


" 


1 — 
















" 0.12 


- 












- 


C/) 










• 






^ 


■ 








o 




" 


DC 
















3 


o log 


u = 


-1.3 




$ 


_ 


■D 

w 0.10 


- * log 


u = 


-2 






* 


- 


CD 


□ log 


u = 


-3 










DC 


1 1 1 L 


_l 


1 L 


, , 1 


—I 1 1 1 L 


—I 1 1 L- 


J_ 



10 



11 



12 



13 



log hh 



Fig. 4. — RMS of the residual spectrum of IZwl after subtraction of fitted Fell+Mgll 
models and emission lines, in the region between 8300 and 11600 A. Solid symbols denote 
the results obtained using only the Fell multiplets (Mgll neglected). 



-40- 



E 



(-3.0,9.6) 

J:...l: I I I I I I 



rms = 0.154 
I I I .1 ;.| I ij 




"I I I I I I I I I I I I I I I I 



-3.0,10.6) rms = 0.154 
j;.!;. I I I I I I 




9000 10000 11000 



9000 10000 11000 



(-2.0,9.6) 

j;.!:; I I I I I I 



a3 



rms = 0.147 
I I I I :.| I IJ 




"I I I I 11 I I I I I I I I I I I 



9000 10000 11000 



(-1.3,9.6) 

J:..l: I I M I I 



E 
a3 



rms = 0.147 
I I I .1:1 I. IJ 




"I I I I '1 I I I I I I I I I I I 



-2.0,10.6) rms = 0.151 





i:;!:: 1 1 






1 i:;i: 


0.5 


H J 


i r 




i ■': - 


0.0 


|W 


^J 


fSm^ 


0.5 


" 1 1 1 '1 1 1 1 1 1 1 1 1 1 1 1 H 




9000 10000 11000 




(-1.3,10.6) rms = 0.141 




J.-l: 1 1 


1 1 1 


; 1 II 1 ; 


1 i;i- 


0.5 


H J 


.[: 


i. J il 


~ 


0.0 


|«^^ 


0.5 


1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 r 



9000 10000 11000 



9000 10000 11000 



Wavelength [A] 



Wavelength [A] 



Fig. 5. — Results of the fitting procedure for the templates of Fell+Mgll, for lognn of 9.6 
and 10.6. Dotted lines are the continuum-subtracted spectrum of IZwl, solid-bold lines 
show the fitted (convolved) template, and the lighter solid lines are the fit residuals. 



-41- 



E 



(-3.0,11.6) 

J:...l: I I I I I I 



rms = 0.151 




"I I I I 'i I I I I I I I I I I I 



(-3.0,12.6) rms = 0.124 
j;.!. I I I I I I j. I I I r.| I I 




I I I I ' I I I I I I I I I I I I 



9000 10000 11000 



9000 10000 11000 



(-2.0,11.6) 



rms = 0.132 



j: I: I I II I I |: I I I 



a3 




"I I I I '' I I I I I I I I I I I 



9000 10000 11000 



(-2.0,12.6) rms = 0.102 

j; I: I I I; I I I I: I I I I I I I 




I I I I I I I I I I I I I I I I 



9000 10000 11000 



(-1.3,11.6) 

J:..l: I I M I I |: I I I 



rms = 0.114 



E 
a3 




"I I I I I I I I I I I I I I I I 



9000 10000 11000 



(-1.3,12.6) rms = 0.107 

J.-l;. I I I. I I III I I r| l;l_ 




I I I I I I I I I I I I I I I I 



9000 10000 11000 



Wavelength [A] 



Wavelength [A] 



Fig. 6. — (Cont.): results of the fitting procedure for the templates of Fell+Mgll, for lognn 
of 11.6 and 12.6. Dotted lines are the continuum-subtracted spectrum of IZwl, solid-bold 
lines show the fitted (convolved) template, and the lighter solid lines are the fit residuals. 



-42- 



s 

o 

O 







-1 - 



-3 



/ 


/ 



/ 

' / ■ 







log^ 



model 



Fig. 7. — Comparison between the 51 elements of the input intensity vector i'^odei (=niodel) 
with those of the output one ideconv (=deconvolved), normahzed by the intensities at 9997A. 
Filled and open circles denote Fe ll and Mg ll lines, respectively. The Wiener deconvolution 
gives a consistent result, specially for the strongest contributions. The dashed line represents 
the locus of the 1:1 relation. 



CO 




-43- 



Fell+Mgll 



.L.I.I 11,1 ,.ll.l.UuliL.I,,.i 111 iliJ<L.j,[l Aa. I I 



Il_L 



^.aj. 



^Ji 




bilillJljllllliLIMlM^ 



± 



± 



9000 



10000 



11000 



Wavelength [A] 



Fig. 8. — Top: Semi-empirical template derived from the best fitted model and the decon- 
volution procedure applied to the observed spectrum of IZwl. Bottom: a zoomed view of 
the template. Intensities are plotted in arbitrary units. 



-44- 



15 




9000 



10000 



11000 



Wavelength [A] 



Fig. 9. — Top: Comparison of the observed Fell+Mgll spectrum with the one derived from 
the best model (-2.0,12.9). Bottom: again the observed Fell+Mgll spectrum, and the semi- 
empirical spectrum, computed through the template derived in this work. Dotted black line: 
total emission line spectrum of I Zw 1; solid grey and black lines: observed and model/semi- 
empirical Fell-|-Mgll spectra, respectively. Dashed lines indicate the zero level intensity. 
Line intensities which varied significantly with respect to the model are also marked in the 
lower plot. 



-45- 



"c 
>^ 

CO 

15 




9000 



10000 



11000 



Wavelength [A] 



Fig. 10. — Top: Comparison of the observed Fell+Mgll spectrum of Ark 564 with the one 
derived from the best model (-2.0,12.9). Bottom: again the observed Fell+Mgll spectrum, 
and the semi-empirical spectrum, computed through the template derived from IZwl. Dot- 
ted black line: total emission line spectrum of Ark 564; solid grey and black lines: observed 
and model/semi-empirical Fell+Mgll spectra, respectively. Dashed lines: zero intensity 
level. 



-46- 



"c 

Z5 
CO 

JO 




Zw 1 Spectrum 



w/o (Fell+Mgll) 




9000 



10000 



11000 



Wavelength [A] 



Fig. 11. — Top: Continuum-subtracted spectrum of IZw 1, with the spectrum of Fell+Mgll 
calculated from the semi-empirical template (in bold) superposed, highlighting its main NIR 
emission lines. Bottom: Spectrum of I Zw 1 without this contribution. All the lines (except 
Paa and those of the template) used in the fit are marked. 



-47- 



"c 

Z5 
CO 

JO 




9000 



10000 



11000 



Wavelength [A] 



Fig. 12. — Top: Continuum-subtracted spectrum of Ark 564, with the spectrum of 
Fell+Mgll calculated from the semi-empirical template (in bold) superposed, highlight- 
ing its main NIR emission lines. Bottom: Spectrum of Ark 564 without this contribution. 
All the lines (except Pa« and those of the template) used in the fit are marked.