Optical singly-ionized iron emission in radio-quiet and relativistically jetted active galactic nuclei

The issue of the difference between optical and UV properties of radio-quiet and radio-loud (relativistically"jetted") active galactic nuclei (AGN) is a long standing one, related to the fundamental question of why a minority of powerful AGN possess strong radio emission due to relativistic ejections. This paper examines a particular aspect: the singly-ionized iron emission in the spectral range 4400 -- 5600 A, where the prominent HI H$\beta$ and [OIII] 4959, 5007 lines are also observed. We present a detailed comparison of the relative intensity of Fe II multiplets in the spectral types of the quasar main sequence where most jetted sources are found, and afterwards discuss radio-loud narrow-line Seyfert 1 (NLSy1) nuclei with $\gamma$-ray detection and with prominent Fe II emission. An Fe II template based on I Zw 1 provides an accurate representation of the optical Fe II emission for RQ and, with some caveats, also for RL sources. CLOUDY photoionization simulations indicate that the observed spectral energy distribution can account for the modest Fe II emission observed in composite radio-loud spectra. However, spectral energy differences alone cannot account for the stronger Fe II emission observed in radio-quiet sources, for similar physical parameters. As for RL NLSy1s, they do not seem to behave like other RL sources, likely because of their different physical properties that could be ultimately associated with a higher Eddington ratio.


Introduction
The wide majority of active galactic nuclei (AGN) are characterised by the presence of broad and narrow optical and UV lines emitted by ionic species over a wide range of ionization potential χ. 1 Restricting the attention to broad lines, type-1 AGN 2 spectra invariably show the same low ionization * ESO fellow † CNPq fellow 1 For an introduction to AGN spectra and the interpretation of their spectra in physical terms, see e.g., [1][2][3], and references therein. 2 Type-2 AGN do not show broad permitted lines in natural light; they are believed to be mostly obscured type-1 and will not be further considered here because they lack the diagnostics offered by the broad lines measurements. rather to establish whether the optical FeII emission might be different in radio-quiet (RQ) and RL AGN. We then present an overview of the sample used for the analysis of the differences between RQ and RL sources, keeping the distinction between CD and FR-II among jetted sources. The comparison is carried out within the MS context, namely comparing sources that are RQ and RL belonging to the same spectral type ( §3), considering a "solid" FeII template (i.e, a template with fixed relative line intensities), and a template in which the relative multiplet strengths are free to vary ( §4). The results ( §5) confirm a substantial equality for the FeII emission, within the non-trivial constraints imposed by signal-to-noise ratio (S/N) and resolution. Possible implications for the line emitting region structure in RQ and RL are briefly discussed, also with the help of photoionization computations ( §6).

The FeII emission lines
The main FeII optical lines in the spectral region between 4400 Å and 5600 Å are associated with 5 main multiplets. It is expedient to distinguish 3 main features: • the blend on the blue side of Hβ, made up of lines from the m37 and m38, and usually referred to as FeIIλ4570, whose intensity is the FeII measure that enters in the definition of R FeII and applied in most papers following Boroson and Green [4]. • m42, with two lines appearing as satellite lines of [OIII]λλ4959,5007. The line at λ5018 might be affected by contamination from He Iλ5016, and is strongly affected also by the red wing of [OIII]λ5007. Given these difficulties the m42 lines were included in the fits but no result concerning this multiplet is considered; • the lines of multiplets m48 and m49 that provide the bulk of the emission of the FeII blend on the red side of Hβ (referred to as FeIIλ5270 or as the FeII red blend).
The transitions giving rise to these lines are schematically shown in the highly-simplified Grotrian diagram of Fig. 2 [3,50,51]. The observed Fe II opt emission is produced by a combination of collisional and resonance-fluorescence processes [51][52][53][54][55]. Collisional excitation, resonance fluorescence, in which a line or continuum photon is absorbed in a line transition, both play an important role. Lyα fluorescence can populate the upper levels of Fig. 2 via population of much higher energy level at ≈ 10 eV, producing a cascade with two branches leading to the z 6 D, 6 F, 6 P, 4 D, 4 F, and 4 P terms between 4.8 and 5.6 eV responsible for the optical emission between 4500 and 5600 Å. Collisional excitation, albeit significant [56], alone cannot explain the emission of the stronger FeII emitters.

Sample
This paper considers two samples of broad-line, type-1 sources: • a sample of bright quasars from the Sloan Digital Sky Survey (SDSS) database in the redshift range 0.4 z 0.7 [28,57]. Composite median spectra were extracted for the RQ, CD and FR-II radio classes. A significant number of RL sources ( out of the 680 quasars of the original sample) is found only in the spectral bins A1, B1, B1 + : 8, 16, 9 CDs and 10, 23, 11 FR-IIs in each bin, respectively. All sources (save a borderline one) in the sample that are RL with extended emission satisfy the power criterion at 1.5 GHz log Pν 31.7 erg s −1 Hz −1 that separate FR-II sources from the lower luminosity FR-I [58,59]. 4 In the last bin, however, the FeII intensity is consistent with 0, so that only sources in bin A1 and B1 are considered. In addition, in A1 FR-II the Fe II opt emission is too weak for a meaningful analysis.In conclusion, our bright quasar sample belongs in the following composite spectra: A1 RQ, A1 CD, B1 RQ, B1 CD, B1 FR-II; • three individual relativistically jetted NLSy1s with γ-ray detections, that show more prominent FeII emission than the composites. The γ-ray detection supports the presence of a relativistic jet as the origin of the radio power for these sources [61]. They were selected because high-quality optical spectra were available for the three of them.
The three γ-ray sources are 1H 0323+342 (≡ B2 0321+33 ≡ 1H 0323+342 ≡ J03246+341), 3C 286 (≡J13311+305), and PKS 2004-447 (≡J20079-445). J03246 is the closest γ-ray NLSy1 (z = 0.063), hosted either by a spiral or an interacting late-type galaxy [62,63], and it is the only known γ-NLSy1 showing a Fe Kα emission line in X-rays [64]. This object was detected by the Fermi Gamma-ray Space Telescope 4 Both FR-I and FR-II sources are characterized by extended radio emission. FR-Is however show lower radio surface brightness toward the outer extremities of the lobes, at variance with FR-II sources that are often described as "edge brightened." According to the unification schemes [60], both classes are observed at relatively high inclination. Their optical spectra can be of both type 1 and type 2 AGN. soon after its launch [65]. J13311 is a rather high redshift (z ≈ 0.850) object, identified as a γ-ray source in all Fermi catalogs (e.g., [66]). Originally it was classified as a compact steep-spectrum (CSS) source based on its radio morphology [67], and later as a NLSy1 from optical spectroscopy [68,69]. Also J20079 (z ≈ 0.240) is classified as both a CSS and a NLSy1 [70][71][72] and, as J03246, it belongs to the first batch of NLSy1s detected by Fermi [65]. The spectra of these three sources were obtained using the Asiago 1.22m Telescope for J03246 (full description of the data reduction in Foschini et al. [73]), the SDSS archive for J13311, and the FORS2 instrument on the Very Large Telescope for J20079-445 [72].

Analysis
The analysis is mainly empirical and, as mentioned, there is no pretension to solve the problem of the optical FeII emission in AGN. Two approaches are followed in the measurement of the relative intensity of the multiplets: • the modeling of the spectrum using a "solid" FeII template, scaled and broadened to minimize the χ 2 in a multi-component fit. It is basically the one of Boroson and Green [4] actualized with a higher resolution spectrum and a model of the FeII emission underlying Hβ [20]. In this case, a multi-component fit was carried out including all known emission components, as detailed in several recent works [74][75][76]. Specifically, the redward asymmetric Hβ has been modeled by the use of three components -a narrow, a broad and a very broad component [77, and references therein]. After verification that the host galaxy spectral emission is not contributing significantly, the local continuum was fit with a power law; • the use of a "liquid" template that permits to change the relative intensity of the multiplets in the optical spectral range [78,79] 5 . In this case, the continuum subtracted spectrum was fit to a set of variable multiplets and to Hβ and [OIII]λ5007 profiles that were approximated as Gaussians.

FeII emission comparison between RQ and RL
The results of the application of the multi-component fit with the solid FeII template are shown in Figs. 3 and 4 for the spectral types A1 and B1, respectively. In Fig. 3, the only comparison we show is between A1 RQ and A1 CD sources, as there is no A1 FR-II composite. In Fig. 4, the left panel shows the comparison between B1 RQ and B1 CD composites, while on the right the B1 RQ composites are compared to B1 FR-II. The immediate result is that RL (CD and FR-II) composites show weaker FeII emission [44]. The second immediate result [already reported in 19] is the similar appearance of the Hβ profiles. The Hβ profiles of Figs. 3 and 4 confirm that the prominent redward asymmetry is not a prerogative of RL sources, but is a common feature in both RQ and RL sources, provided that they belong to Population B [11,81,82].
The bottom panels of Figs. 3 and 4 provide constraints on the nature of the difference between FeII emissions in the RQ and RL composite spectra. The brown line traces the difference in the templates, which is positive since FeII is stronger in RQ than in CD and FR-II as well. The black filled lines traces the difference between the observed FeII emission after rescaling of the RL FeII emission by a factor k to obtain an average of 0. The difference is computed between the observed FeII spectra, as follows: where the index i refers to CD or FR-II, the f λ to the observed flux as a function of wavelength, and the calligraphic symbols refer to model spectrum (M λ ) and model FeII (F e) via the scaled and broadened template. Model parameters were obtained with a multi-component, non-linear minimum χ 2 technique, as implemented in IRAF [83].
The basic results of the spectral fits are reported in Table 1. The Table columns list, in the following order, the spectrum identification, the rest-frame equivalent width in Å for FeIIλ4570, R FeII , FWHM Hβ, the spectral type defined on the basis of the reported R FeII and FWHM Hβ, and the rescaling factor k. Table 2 lists the average of the absolute difference between the observed iron blends and the solid FeII template. The averages are computed over the full ranges 4434 Å -4684 Å (hereafter indicated as B blend or FeIIλ4570), and 5100 Å -5600 Å (hereafter R blend) also used for the CLOUDY photoionization simulations (Sect. 6.1). The next columns yield the average and relative absolute differences between RL and RQ after rescaling, for the restricted spectral regionB between 4500 and 4590 Å (roughly corresponding to multiplets 38 and 39) andR (between 5200 and 5330 Å, with the strongest features of m48 and m49). The reason of this restriction is to avoid the contamination by HeII on the blue side of Hβ. The HeIIλ4686 line can have a significant effect and its estimation is based only from fit.
In the case of spectral type A1, the comparison between RQ and CD relative intensities of the multiplets is shown in the panels of Fig. 3. On the blue side, the δFeII λ,i is consistent with 0 over the range where most of the emission due to multiplets 37 and 38 is expected. Similar considerations apply to the red side emission; the main fluctuations according to Eq. 2 occur when FeII emission is weak.
The comparison for the composite spectra of bin B1 is shown in Fig. 4. On the blue side of Hβ in correspondence of the m37 and m38 emission there could be a slight excess for the CD composite, accompanied by a deficit on the red side of Hβ where emission is associated with multiplets m48 and    m49. The average relative difference is a few positive and negative percents for the B and R regions, respectively ( Table 2). Accepted as real the absolute difference is constrained within 3 %, and the relative difference is contained within ± 20 %. The ≈ 3% average excess translates into an error in terms of δR FeII ≈ 0.03, and in W FeIIλ4570 the difference is also ≈ 3 Å implying that there is no effect on the placement in the MS optical plane. To estimate the measurement errors on R FeII for the same S/N, equivalent width and FWHM of the lines, we used the quality parameter defined in an unrelated investigation, and derived from Monte Carlo simulations estimates of the errors for Hβ and R FeII . The errors on FeII are the dominant one, and are δR FeII 0.05 for the RQ composites and δR FeII ≈ 0.10 for the radio-loud ones.

γ-ray detected RL NLSy1s
We apply the same fitting techniques to three relativistically jetted NLSy1s with γ-ray detection. Fig. 5 shows that the solid template used for the fitting of the composites is providing a good agreement, in a case with higher values of R FeII and much narrower lines that is better posed to appreciate possible differences. The fits are successful, and the residuals consistent with 0, with average systematic differences are 1 %, and the rms scatter is 5% of the normalized continuum value at 5100 Å. a : B/R ratio measured on the observed spectrum; b : same as in the previous column, butB/R in the restricted range; c : ratio between the sum of m37+m38 and m48+m49 intensity obtained with the fits following Kovačević et al. [78]. Note that the ratio includes only the sums of these four multiplets, while both the standard and the restricted range include additional FeII emission as well as a possible residual contribution of HeI and HeII lines. d : ratio between the FeII emission of the B and R restricted range computed with the simulations of Kovačević et al. [78]. e : absolute average difference for the restricted B and R ranges, respectively, using the total FeII emission following Kovačević et al. [78].

Models with relative intensity of the multiplets free to vary
We considered first the measurements based on the continuum-subtracted spectra in the full and restricted ranges (second and third columns of Table 3), and we utilized the web-based fitting tool 6 for implementing the scheme of Kovačević et al. [78] for the optical FeII emission. The measurement of the ratio between the sum of the m37+m38, and m48+m49 blends is reported in the fourth column of Table  3. The fifth column reports the ratioB k /R k measured on the full spectral model of Kovačević et al. [78]. The last two columns report the average differences between the observed and the Kovačević et al. [78] modelling for the restricted rangesB k andR k , respectively, and are meant to provide an estimate of how well the model FeII reproduces the actual observations. The agreement is very good with all cases having the average difference less than 0.05, and most less than 0.01. The same procedure was also applied to the solid FeII template used in the previous analysis (Fig. 6). The values reported in Table 3 explain why the solid FeII template provides satisfactory results in all cases, and especially in the RQ cases.
We stress three main results: • There is a significant difference (by a factor ≈2) between RQ and RL sources in the same spectral types (meaning similar mass, Eddington ratio 0.1, and luminosity). • The B/R ratio for the RQ A1 and B1 spectral types is ≈ 0.9 − 1.1 ( Fig. 7 and Table 3).
• The various measurements of the blue and red blend ratios suggest a somewhat higher values for the A1 and B1 RQ than for the RL sources (≈ 0.9 − 1 vs 0.7). The effect is not strong: yet, it is apparent especially in Fig. 7.
It is advisable to consider that there could be effects not related to the estimation of the FeII intensity. The main reason could be a difference in the intensity of HeIIλ4686, even if HeIIλ4686 is included in the fit. Other possibilities involve internal extinction [84,85] affecting the FeII emitting region, limb-darkening [6], incorrect continuum placement, contamination by lines other than HeII, and blueshifted Hβ emission. To overcome at least the problem of the HeIIλ4686 uncertain contribution, and of the contribution of the Hβ wings, we defined a restricted range for B and R,B andR. The effect is however not significantly reduced ( Table 2). If we apply the same technique to the γ−detected sources, the intensity ratio between B and R is closer to the value of the template ( Fig. 8 and Table 3) but not for 1H 0323+342. The values reported in Table 3 confirm that the template is providing a good approximation for the RQ sources in bin A1 and B1. This is likely true also for spectral bins with higher R FeII (A2 and A3), as the template is based on the spectrum of I Zw 1, a NLSy1 accreting at a high rate and especially emitting strong FeII, with R FeII slightly larger than 1, and so of spectral type A3 7 . At the other end of the MS, FeII emission is faint and systematic differences might be not appreciable. So the issue of a possible, significant disagreement concerns only ≈ 3% of all type 1 AGN.
In summary, the use of a solid FeII template appears fully justified in this work. There could be a genuine effect associated with radio-loudness, in the sense of a less prominent blue blend with respect to the red one.

Discussion
The composites that we have considered in the present work all show modest FeII emission and satisfy the condition that R FeII 0.5. In principle, on the context of photoionization, they do not pose a serious challenge to the conventional view of the of the BLR as a system of emitting clouds characterized by typical densities n H ∼ 10 9.5 cm −3 , and column density N c = 10 23 [cm −2 ] [87]. Indeed, physical parameters similar to these n H ∼ 10 10−10.5 cm −3 , along with metallicity solar or slightly sub-solar, and a flattish spectral energy distribution (SED) [88,89], can account for the emission in the R FeII range of spectral bins A1 and B1, at BLR distances from the continuum sources consistent with their luminosity [31].
Thanks especially to the results of reverberation mapping and [27], the emitting region radius was found to be a factor of 10 smaller than previously thought [90,91]. The conventional view outlined in the previous paragraph may be only partially valid in Population A sources, as it is unable to account for the strong FeII emission in quasars [92,93]. Different physical conditions are required for sources 7 I Zw 1 is however not a very extreme object: the most extreme accretors show spectra similar to the one of PHL 1092 [86].

Photoionization computations
To ascertain the role of the SED in the strength of FeII we performed exploratory calculations of the RQ and RL differences, as well as for the γ-detected NLSy1s. As a first attempt, we assumed a column density N c = 10 23 cm −2 , solar metallicity, and 0 micro-turbulence. We focused on the exploration of the parameter ranges of the density n H (∼ 10 9−11 cm −3 ), and of BLR radius within the range expected from the Bentz et al. [97] scaling law, corrected for high-accretion rates ( [98,99]; see also [100]). We utilized the SEDs from Laor et al. [88] and Korista et al. [89], which are most appropriate for bins A1 and B1, for RQ and RL separately (top panel of Fig. 9). The two RQ SEDs most likely bracket the distribution of SEDs in the two bins, as also assumed in previous work [31]. For the composites, we consider a fixed luminosity λL λ (5100 Å) ≈ 10 45 erg s −1 , consistent with the values derived for the original sample [28]. For RL NLSy1s, we built a specific SED for each one of them by using the multi-frequency data of the SED builder available at the Space Science Data Center of the Italian Space Agency 8 . The results are shown in the second, third, and fourth panel of Fig. 9.  Fig. 10 shows the expected R FeII , the ratio between the red blend R and Hβ, and the ratio B/R for the RQ and RL SEDs, as a function of r BLR and n H . As mentioned above, the range in r BLR has been centered on the value expected from the correlation r BLR -L by Bentz et al. [97]. The density range includes the most likely values for the BLR along the MS. Typical values are lower for Population B, and much higher for Population A. [6,101].

Composites RL and RQ
The most important result is that the SED for RL sources is able to account for the modest R FeII and W(FeIIλ4570) reported in Table 1. The observed equivalent width ∼ 10 − 20 Å and R FeII ∼ 0.15 are explained at the r BLR expected from the r BLR -L relation, and by the moderate density n H ∼ 10 11 cm −3 . The result seems to be especially robust against changes in n H and r BLR (Fig. 10, middle column). Restricting the ranges toB andR confirms this conclusion ( Fig. A1; the result of the analysis for the restricted range are shown in Appendix A).
Regarding the interpretation of the composite spectra, a second important issue is whether the difference in RQ and RL SED can explain the differences in their FeII strength. We stress that the comparison is being carried out between sources that are in the same spectral types, meaning that they have similar M BH , L, and Eddington ratio. The R FeII computed over a large area of the parameter plane n H -r BLR shows differences that are at most 0.05, i.e., less than 15% (Fig. 10). This has the important implication that the parameter k ≈ 2, i.e., R FeII twice as strong in RQ than in RL AGNs, cannot be explained on the basis of the SED only. However, the RQ SED that we adopt is not necessarily equivalent to the one of the sources used to construct the Population A and B. Using the SED from Korista et al. [89] that has a stronger big blue bump (implying for fixed r BLR and n H an increase in ionization parameter by a factor 3) cannot lead to a higher R FeII (Fig. 11 and Fig. A2).
The second issue investigated in the present paper is the possibility of a systematic difference in the B/R FeII ratio. This result should be viewed with care, and confirmatory data are needed, also 16 considering that we are dealing mainly with weak FeII emitters. The agreement between the observed and photoionization-predicted B/R is good, and in most cases the B/R ∼ 1, with little dependence on the r BLR and n H parameters. Fig. 12 (Fig. A3 for the restricted ranges) shows that there is a possible lowering of the B/R ratio in the B1 RL (both CD and FR-II) with respect to the RQ. Accepted at face value, it might imply a lower density, and a higher ionization parameter for the RL sources (Fig. 12), although Figs. 10 and 11 show that this implies fainter FeII emission than what is expected for the typical values of r BLR .
Sulentic et al. [102] speculated that the cocoon associated with the shock wave due to the expansion of the jet could push outward the line emitting regions. It is known that the CIV high-ionization line reaches lower blueshift amplitudes in RL than in RQ sources [14,103]. The present paper confirms that an effect on the low-ionization lines is minor. Considering the layout of the atomic levels in the Grotrian diagram of Fig. 2, the lower B in radio loud sources could be due a lower electron T in RL than in RQ, which could be, according to Fig. 12, in turn associated with a lower U and a larger distance.

γ-detected NLSy1s
The photoionization calculations performed with the parameters appropriate to model 1H 0323+342 (Fig. 13, repeated in Appendix A for the restricted range as Fig. A4) reveal that is not possible to achieve R FeII ∼ 1 with the SED shown in Fig. 9. This raises an important issue about the FeII emission in these jetted sources. A change in SED within the limit of Population B, such as the one provided by Korista et al. [89] is likely not enough to explain the increase in FeII. This leaves several other possibilities to account for this intrinsic difference between jetted and non-jetted sources. They can be loosely grouped in "evolutionary" (chemical evolution of the host and of the circumnuclear regions, in turn related to the host morphology), and "intrinsic" (density, column density, ionization degree, covering fraction of the line emitting gas). The host morphologies of RL and RQ are systematically different, and recent work has validated the long-held paradigm that RL are hosted in earlier morphological types than RQ [105,106]. This result might not be applicable to NLSy1s, and especially to the ones that are moderate-to-strong FeII emitters. Basically all of the recent studies on the host of jetted RL NLSy1s [107][108][109] showed that they are in late-type hosts. Merging is instead rather common, and it may constitute the biggest difference with respect to RQ NLSy1. In particular, Olguín-Iglesias et al. [109] clearly pointed out on a rather large sample that disk galaxies dominate among jetted NLSy1s. The visual inspection of almost all known RL NLSy1s suggests that they never reach the extreme R FeII value obtained by RQ sources. This impression -even if consistent with the difference found for weak FeII emitters -should be validated by a systematic study that goes beyond the scope of the present paper. Enrichment in chemical composition appears a likely possibility, as the strongest FeII emitters are also believed to have the highest metal content [96,[110][111][112]. Photoionization analyses indicate that significant FeII emission is associated with super-solar metallicity, in addition to high density and high column density [30,31], although a complete observational analysis with matching R FeII and UV line diagnostics for metallicity and physical parameters estimates along the main sequence is still lacking.

Summary and conclusion
This work presented an analysis of the FeII emission in the spectral bins along the main sequence where both RQ and RL coexist (A1, B1) with measurable FeII strength, and a comparison between the emission of RQ and RL composite template, along with the analysis of 3 RL sources found to have a stronger FeII emission. The main results can be summarized as follows: • the template based on the I Zw 1 spectrum works equally well for RQ and RL objects. Deviation in R FeII and equivalent width measurements due to differences between the template and the observed spectra have been found to be δR FeII 0.03, and δW a few Å. In particular, no correction is needed for the placement of RL in the MS built from RQ-dominated samples.
• The RL SED of Laor et al. [88] can account for the modest emission in RL sources, for conditions that are very likely in Population B objects [31]. • Somewhat surprisingly, SED differences between RQ and RL cannot account for the stronger FeII emission in RQ 9 . We suggest that other factors related to the evolutionary pattern of the circumnuclear regions of the active nucleus should be investigated. • Last, the SED shape does not seem so important in determining the optical FeII prominence with respect to Hβ for such sources with intermediate Eddington ratios (∼0.1), as the change from the Laor et al. [88] to the Korista et al. [89] SED produces no significant change in R FeII , even if the number of ionizing photons is increased by a factor ≈ 3. We remark that, for sources at higher Eddington ratios, this might be different, as has been tested and shown in Panda et al. [31]. Those authors found that the SED shape matters for pushing the R FeII higher, especially when the contribution from a soft-X-ray excess is accounted for. • The ratio between the B and the R blend is found to be ≈ 0.9 -1.0 and to be consistent with the predictions of the photoionization computation. A slightly lower value (≈ 0.7) might be possible for the B1 RL composites. However, the measurement is excruciatingly difficult, considering the uncertain influence of the HeIIλ4686 emission that is known to be very strong in cases where FeII is negligible [e.g., 113], and of several other factors that should play a lesser role. More importantly, the B/R value around ≈ 1 is found also in higher Eddington ratio sources which are stronger FeII emitters such as the 3 γ-detected NLSy1s considered in this study, all with Eddington ratio ≈ 0.5, typical of Population A quasars. Since R FeII is correlated with Eddington ratio [114,115], and the B/R is apparently independent from R FeII , a lower B/R might be a genuine radio loudness effect or, perhaps more likely, an effect dependent on other RQ/RL sample differences.
In summary, we find that the FeII emission is stable, and we confirm that a solid template based on I Zw 1 provides an accurate representation of the optical FeII emission in the spectral region of Hβ within the limit of precision allowed by moderate dispersion spectroscopy.
Author Contributions: All authors contributed equally to this paper.

Abbreviations
The The Appendix shows the results of the CLOUDY simulations in the case of the restricted ranges 4500 -4590 Å and 5200 -5300 Å (B andR through the paper).
Appendix Restricted range 16 Figure A3. Results of CLOUDY simulations: the three panels yield the ratio R FeII B to R FeII R (in practice B/R) as a function of density and BLR radius, for the RQ Korista et al. [89] (leftmost), the RQ Laor et al. [104] (middle) and RL SED (rightmost).  ratio (blue/red) Figure A4. Results of CLOUDY simulations for the 1H0323+342 SED: the two top panel yields the equivalent width of FeIIλ4570 and the two middle panels R FeII for the blue and red blends, as a function of density and BLR radius. The bottom panel shows the ratio of the blue and red FeII emission.