Narrow-line Seyfert 1 galaxies in Sloan Digital Sky Survey: a new optical spectroscopic catalogue

Narrow-line Seyfert 1 (NLSy1) galaxies are an enigmatic class of active galactic nuclei (AGN) that exhibit peculiar multiwave-length properties across the electromagnetic spectrum. For example, these sources have allowed us to explore the innermost regions of the central engine of AGN using X-ray observations and have also provided clues about the origin of relativistic jets considering radio and 𝛾 -ray bands. Keeping in mind the ongoing and upcoming wide-field, multi-frequency sky surveys, we present a new catalogue of NLSy1 galaxies. This was done by carrying out a detailed decomposition of > 2 million optical spectra of quasars and galaxies from the Sloan Digital Sky Survey Data Release 17 (SDSS-DR17) using the publicly available software “Bayesian AGN Decomposition Analysis for SDSS Spectra". The catalogue contains 22656 NLSy1 galaxies which is more than twice the size of the previously identified NLSy1s based on SDSS-DR12. As a corollary, we also release a new catalogue of 52273 broad-line Seyfert 1 (BLSy1) galaxies. The estimated optical spectral parameters and derived quantities confirm the previously known finding of NLSy1 galaxies being AGN powered by highly accreting, low-mass black holes. We conclude that this enlarged sample of NLSy1 and BLSy1 galaxies will enable us to explore the low-luminosity end of the AGN population by effectively utilizing the sensitive, high-quality observations delivered by ongoing/upcoming wide-field sky surveys. The catalogue has been made public at https://www.ucm.es/blazars/seyfert .


INTRODUCTION
Narrow-line Seyfert 1 (NLSy1) galaxies were, initially defined as, low-luminosity active galactic nuclei (AGN, absolute -band magnitude  B > −23, Schmidt & Green 1983) which were identified purely based on their optical spectroscopic properties.Osterbrock & Pogge (1985) originally proposed to classify them based on the presence of narrow 'broad permitted lines' and the strength of the forbidden [O III] emission line compared to H line with flux ratio of [O III] 5007/H<3.The spectropolarimetric study of NLSy1 galaxies was later carried out by Goodrich (1989) who quantified the first selection criterion of Osterbrock & Pogge (1985) with the full width at half maximum (FWHM) of the broad H line to be <2000 km s −1 .These objects exhibit several peculiar observational features such as the strong permitted FeII complexes, steep soft X-ray spectra, rapid X-ray flux variations (cf.Boller et al. 1996;Leighly 1999a,b) and strong outflows (e.g., Boroson 2002;Komossa et al. 2008;Grupe et al. 2010;Xu et al. 2012).These observations have indicated the existence of rapidly accreting, low mass black hole systems ( BH ∼ 10 6−8  ⊙ ) powering these enigmatic AGN (e.g., Peterson et al. 2000;Grupe & Mathur 2004;Xu et al. 2012).How-★ E-mail: vaidehi.s.paliya@gmail.com(VSP) ever, alternative theoretical models have also been put forward by attributing the observed characteristics of NLSy1 galaxies to geometrical parameters, e.g., covering factor, leading to the proposition that NLSy1 source population has been preferentially viewed at small angles compared to their broad line counterparts (cf.Decarli et al. 2008).
NLSy1 galaxies have been used to study a variety of AGN physics problems.For example, the first fundamental correlation vector or the eigenvector 1 of Type 1 AGN (EV1) which represents the correlations of various observables such as the steep X-ray spectrum or strength of the [O III] or optical FeII emission with the H line width, has been argued to be an important Type 1 AGN unification scheme found so far (Boroson & Green 1992;Sulentic et al. 2000).These objects have been found to lie at the extreme negative end of EV1 thereby providing us clues about the central engine parameters, e.g., black hole mass or accretion rate, and/or geometrical aspects such as the viewing angle (cf.Boller et al. 1996;Sulentic et al. 2000;Marziani et al. 2001).Furthermore, the X-ray spectra of NLSy1 galaxies often show a strong soft X-ray excess and reflection dominated hard Xray emission (e.g., Fabian et al. 2009).Deep X-ray observations of these sources have permitted us to study the behaviour of matter and energy and their possible interaction in the immediate vicinity of the central supermassive black hole (Parker et al. 2014;Kara et al. 2017).Though most of the NLSy1 galaxies are radio-quiet, ∼5% of them are found to be radio-loud indicating the presence of jets (Komossa et al. 2006;Yuan et al. 2008).Some of the very radio-loud NLSy1s have also been detected in the all-sky -ray survey being conducted with the Fermi-Large Area Telescope (Abdo et al. 2009;Paliya et al. 2018).This has led to the idea of them being the nascent blazars (e.g., Paliya et al. 2020).Indeed, the general NLSy1 population has been considered as rapidly accreting, low-luminosity AGN in the early stage of their evolution (cf.Mathur 2000;Berton et al. 2018).Moreover, these sources have also been reported to exhibit rapidly rising and long-lasting optical flaring activity thereby making them crucial targets in the era of time-domain astronomy (Frederick et al. 2021).
The above-mentioned research problems highlight the pivotal role that NLSy1 galaxies can play in advancing our current understanding of AGN science.Also considering the latest and upcoming widefield, multi-frequency sky surveys and missions, e.g., Very Large Array Sky Survey (VLASS, Lacy et al. 2020), eROSITA (Predehl et al. 2021), and Rubin observatory (Ivezić et al. 2019), it is imperative to increase the sample size of the known NLSy1 galaxies.The latest catalogue of this class of AGN was prepared using the Sloan Digital Sky Survey data release 12 (SDSS-DR12) and contains 11,101 NLSy1s (Rakshit et al. 2017) which superseded the earlier catalogues containing 150 and 2011 sources using SDSS early data release and SDSS-DR3, respectively (Williams et al. 2002;Zhou et al. 2006).Since then, there have been a number of major updates, both in data collection and analysis software, that motivated us to prepare a new sample of these enigmatic AGN.We highlight a few such updates below: (i) The number of spectroscopically observed sources has considerably increased in the most recent DR17 which is the final survey from the fourth phase of SDSS (Blanton et al. 2017).For example, SDSS-DR16 contains 750,414 quasars 1 (Lyke et al. 2020) which is more than two times larger than the 297,301 objects published in DR12 (Pâris et al. 2017).
(ii) The work of Rakshit et al. (2017) only considered objects classified as quasars (SDSS pipeline keyword QSO).In addition to quasars, we have searched for NLSy1s among ∼1.9 million sources identified as galaxies by the SDSS data reduction pipeline (keyword GALAXIES).
(iii) In Rakshit et al. (2017), the analysis of SDSS spectra was done using a custom-built private software.In this work, we have instead adopted the publicly available optical spectroscopic data analysis software 'Bayesian AGN Decomposition Analysis for SDSS Spectra 2 ' (BADASS, Sexton et al. 2021).This package is an open-source spectral analysis tool designed for detailed decomposition of SDSS spectra.
(iv) Analyzing SDSS spectra with the motivation to identify NLSy1 galaxies will naturally lead to finding new broad-line Seyfert 1 (BLSy1) galaxies.While earlier studies have only reported the NLSy1 galaxy sample (Williams et al. 2002;Zhou et al. 2006;Rakshit et al. 2017), we release the catalogues of both NLSy1 and BLSy1 sources.
In Section 2, we briefly describe the sample selection criteria, while details of the BADASS spectroscopic data analysis steps are elaborated in Section 3. The new catalog of NLSy1 galaxies and their multi-wavelength properties are discussed in Sections 4 and 5, respectively.We summarize our findings in Section 6. Throughout, a flat cosmology with  0 = 70 km s −1 Mpc −1 and Ω M = 0.3 was adopted.

SAMPLE SELECTION
We considered all DR-17 sources that were classified either as QSO (1370779 objects) or GALAXIES (3237535 sources) by the SDSS pipeline.Next, we applied the following filters to retain sources for spectroscopic analysis: (i) reject all sources with  >0.8 and >0.9 for SDSS and Baryon Oscillation Spectroscopic Survey (BOSS) spectrographs, respectively.This is because the optical spectra taken with the BOSS spectrographs have larger wavelength coverage compared to the SDSS spectrographs.This redshift cut ensured that both H and [O III] emission lines, needed to characterize a NLSy1 galaxy, are present in the optical spectrum.
(ii) reject all objects with the keyword zWarning >0 and zWarning ≠ 16.The latter condition, i.e., zWarning = 16, usually indicates a high S/N spectrum or broad emission lines in a galaxy3 .
(iii) reject all sources with the keyword specprimary= 0 since the best observations of all unique objects have specprimary >0.
This exercise left us with 111506 QSO and 1918262 GALAXIES.Furthermore, we also included 11754 quasars from the SDSS-DR16 quasar catalog that were left out possibly due to one/more filters mentioned above.Recently, Wu & Shen (2022) reported the results of the optical spectroscopic analysis of SDSS-DR16 quasars and also provided improved redshifts measurements.We used their redshift measurements for sources common in both samples and adopted the SDSS pipeline redshift for the rest of them.Overall, our final sample contains 123260 quasars and 1918262 galaxies.The analyses of these >2 million spectra were carried out using BADASS software as described in the next section.

Bayesian AGN Decomposition Analysis for SDSS Spectra
BADASS is an open source optical spectroscopic data analysis package designed to automate the deconvolution of AGN and host galaxy spectra, simultaneously fitting both emission line and continuum features, and estimating robust parameter uncertainties using Markov Chain Monte Carlo (MCMC) approach.The full description of the tool can be found in Sexton et al. (2021) and Sexton et al. (2022) and here we summarize its salient features.
In the first step, the optical spectrum is brought to the rest-frame and corrected for the Galactic extinction using the extinction map of Schlafly & Finkbeiner (2011) and extinction law of Cardelli et al. (1989) considering  V = 3.1.Next, the software simultaneously fits all components, e.g., emission lines, to constrain their relative contribution and covariances.
The BADASS software provides several line profile shapes, e.g., Gaussian or Lorentzian, to fit the broad and narrow emission lines.The measured line widths are corrected for the SDSS instrumental resolution.The flux ratios of [O III] and [N II] doublets are kept fixed to 3 during the fit (e.g., Shen et al. 2011).Optionally, the widths of narrow lines can also be tied, e.g., all narrow line width components in the H region are tied to the [O III] line width.
The model fitting in BADASS is done using the Bayesian affineinvariant MCMC sampler emcee (Foreman-Mackey et al. 2013).To obtain the initial parameter values, a maximum-likelihood fit is performed adopting a Gaussian likelihood distribution.The full parameter space is then scanned using MCMC to derive robust parameter and uncertainty estimates.Additionally, the software also provides an option of performing multiple iterations of maximum-likelihood fitting by applying a Monte Carlo bootstrapping technique.The spectra are perturbed by adding a random normally distributed noise at every pixel using the spectral flux uncertainties and re-fitting the spectra.The median (50th percentile) of the distribution is adopted as the best-fit values of the spectral parameters and the semi amplitude of the range covering the 16th-84th percentiles of the distributions is considered as 1 uncertainties on each parameter.

Spectral Analysis of Quasars
The spectral fitting of 123260 SDSS quasars was done in the wavelength range 3500−7000 Å.The host galaxy emission was modeled with three single stellar population templates of ages 0.1, 1, and 10 Gyr from EMILES library (Vazdekis et al. 2016).We did not attempt applying more sophisticated penalized Pixel-Fitting software (pPXF, Cappellari 2017) to model the host galaxy emission since the overall spectrum is expected to be dominated by the emission lines.To reproduce the continuum, we also considered a power-law and optical FeII complex template from Véron-Cetty et al. (2004).Since Balmer lines in NLSy1 galaxies are usually better represented by a Lorentzian function (cf.Sulentic et al. 2002;Goad et al. 2012;Cracco et al. 2016), we fitted the broad components of H and H emission lines with a Lorentzian function.The narrow emission line profiles, on the other hand, were modeled with a Gaussian function.The [O III] doublet was fitted with a single or double Gaussian functions, one each for the core and wing, depending on the line shapes and signal-to-noise (S/N) ratio of the data.We also applied the condition that the width of the broad components must be larger than that of the narrow emission lines.A maximum-likelihood fitting technique using a normal likelihood distribution was adopted to carry out the initial optimization of the spectral parameters which was followed by MCMC fitting for the robust parameter and uncertainty estimations.
A maximum of 5000 iterations of MCMC sampling were performed with 100 walkers per parameter and we considered the final 1000 iterations (4000 burn-in) for the posterior distributions.The median of the distribution was taken as the best-fit value and the 16th and 84th percentiles of the distributions were used as the lower-and upper-bound 1 uncertainties on each measured quantity.Figure 1 shows the fitted optical spectrum of one of the analyzed quasars.

Spectral Analysis of Galaxies
The optical spectral analysis of ∼1.9 million SDSS galaxies was carried out following a strategy similar to that adopted to model the quasar spectra.However, given the large number of sources and to optimize the computational resources, we divided the analysis in two parts.First, we modeled the continuum with a power law and optical FeII template and also applied the pPXF software to reproduce the strong host galaxy emission.The emission lines were modeled with Gaussian functions and the S/N ratio of the broad H emission line was computed.To speed up the process, results obtained from a single maximum-likelihood fitting were considered.At this stage, we rejected all sources in which the S/N ratio of the broad H emission line was found to be <2.This exercise led to the selection of 1143 galaxies.Then, we repeated the full spectral fitting, similar to that adopted for quasars, on these objects and estimated the parameters and uncertainties by applying the MCMC fitting technique.An example of the fitting is demonstrated in Figure 1.

Reliability of Spectral Modeling
The BADASS software carries out the spectral data analysis in fully automatic mode.As usual for any modeling technique, the reliability of the fitting results strongly depends on the S/N ratio of the spectrum.In Figure 2, we show the distribution of the S/N ratio of the H and H emission line regions.For H line, majority of sources have S/N ratio ∼5 possibly due to the fact that DR17 quasars probe fainter flux limits compared to previous data releases given the improved sensitivity of BOSS spectrographs (see, e.g., Shen et al. 2011;Wu & Shen 2022).In this regard, the application of MCMC fitting that has allowed us to derive the robust parameter uncertainties by properly taking into account the data quality, ensures that the measured parameters are reliable.Unlike the conventionally adopted least-square fitting method which may get stuck to a local minima, MCMC fitting scans the full parameter space so that the global minima is achieved.Furthermore, we also performed a number of checks and cross-matched our results with other published works to verify the robustness of the obtained results as described in the next section.We also provide the S/N ratio measured for all of the broad and narrow emission lines in our NLSy1 and BLSy1 catalogues so that a user can customize the selection of objects for a particular research problem.

RESULTS
We derived the emission line parameters for 123260 quasars and 1143 galaxies using the BADASS software.All fitted spectra were visually inspected to filter out contaminating objects whose spectral parameters could not be well constrained.If required, the fit was repeated, e.g., with modified wavelength coverage to avoid noisy edges of the spectrum, and parameters were calculated again.This led to the rejection of 49428 optical spectra.A major fraction of these objects turned out to be either Type 2 AGN with no broad lines detected in their optical spectra or those with extremely poor quality data.The spectral parameters of the remaining 74975 AGN, including 222 galaxies, were further analyzed to identify NLSy1 and BLSy1 galaxies as discussed in the next section.
In Figure 3, we plot some of the measured spectral parameters, namely FWHM and flux of the broad H and H components for the full sample.We also show the variation of the FWHM of the broad H line as a function of the optical FeII strength ( 4570 ) which is defined as the ratio of the FeII flux in the wavelength range 4434 Å−4684 Å to the flux of the broad component of the H emission line (e.g., Boroson & Green 1992).The FWHM of the broad components of the Balmer lines are found to be correlated as    H ∝ (0.82 ± 0.01)   H confirming the results found in earlier studies (cf.Zhou et al. 2006).Similarly, the flux values of the broad H and H lines are also correlated with  H ∝ (3.41 ± 0.01) H (e.g., Domínguez et al. 2013).
In Appendix, Table A1 reports the measured spectral parameters for the sample which is published in the form of online NLSy1 and BLSy1 catalogues4 .Along with the measured quantities, we also derived -band absolute magnitude ( B ), bolometric luminosity ( bol ), single epoch virial black hole mass ( SE ), and Eddington ratio ( Edd ).
We used the SDSS  and -filter magnitudes and adopted the following transformation equation to estimate the Bessel -band magnitude (Jordi et al. 2006): which was then used to estimate  B .The objective of calculating this parameter was to consider the fact that most of the studies on NLSy1 galaxies do not differentiate them with narrow-line quasars.Though we also do not apply any selection filter using the absolute magnitude based quasar/Seyfert classification ( B ≶ −23, Schmidt & Green 1983), this piece of information may enable the user to select genuine NLSy1 objects.For the sake of consistency, we throughout mention all broad/narrow line objects as BLSy1/NLSy1 galaxies.
We considered the commonly adopted single epoch virial technique to estimate the mass of the central black hole that assumes the broad line region to be virialized (e.g., Vestergaard & Peterson 2006).In particular, the following equation was used to calculate  SE (e.g., McLure & Dunlop 2004): where  BLR is the Keplerian velocity of the line emitting BLR clouds,  BLR is the BLR radius, and  is the gravitational constant.The parameter  refers to the scale factor considering the kinematics and geometry of the BLR and taken as  = 3/4 assuming spherical distribution of clouds (Rakshit et al. 2017).The  BLR was estimated following Du & Wang (2019) who updated the scaling relation between  BLR and 5100 Å continuum luminosity including the relative strength of FeII emission, which is important for highly accreting AGN such as NLSy1 galaxies.In particular, the following equation was adopted: where coefficients , , and  are estimated as 1.65, 0.45, and −0.35 (Du & Wang 2019).The reported uncertainty in  SE refers to measurement errors and does not take into account any systematics which can be as large as 0.4 dex (e.g., Shen 2013).Furthermore, we computed  bol from 5100 Å continuum luminosity adopting the bolometric correction factor of 9.26 (Richards et al. 2006).The Eddington ratio,  Edd , was estimated from the derived  SE and  bol .

The NLSy1 Catalogue
To identify the genuine NLSy1 galaxies, we used the following two conditions: (i) the FWHM of the broad H emission line within measured uncertainties is smaller than 2000 km s −1 , i.The first filter follows the classic definition of NLSy1 galaxies (Goodrich 1989).It is slightly different from that used in previous NLSy1 catalogues (cf.Zhou et al. 2006;Rakshit et al. 2017).These works used the H FWHM threshold of 2200 km s −1 without considering the uncertainties in the measured quantity which we have accounted for.The second condition was proposed to separate NLSy1s from Seyfert 2 galaxies.To calculate the ratio, the fluxes of those [O III] 5007 and narrow H emission lines were considered that have S/N ratio >1 and non-zero flux uncertainties, otherwise flux values were assumed zero since in such cases the line detection is marginal at best.The application of above two filters led to the final sample of 22656 NLSy1 galaxies present in SDSS-DR17.There are 46 sources that qualified the first selection filter but not the second, i.e., the flux ratio of [O III] 5007 and H emission lines was found to be >3.If the threshold of FWHM of the broad H component is relaxed to 2200 km s −1 , the NLSy1 sample size grows to 27298.Nevertheless, we stick to the original threshold of 2000 km s −1 and list all other objects in the BLSy1 catalogue so that if a user wishes to use the relaxed criterion, they can find all of the information in this catalogue.Furthermore, 17206 sources have  B > −23 indicating a large fraction of the NLSy1 sample to be genuine Seyfert galaxies (Schmidt & Green 1983).

Comparison with previous works
For a consistency check, we compared our catalogue of NLSy1 galaxies with earlier published works.Cross-matching with the SDSS-DR12 NLSy1 catalogue (Rakshit et al. 2017) by using a maximum search radius of 3 arcsec, we found that 884 quasars classified as NLSy1 galaxies in SDSS-DR12 were rejected during the visual inspection.Among the remaining 10217 objects, 8904 i.e., ∼87.1%, sources are common in both the catalogues.If we adopt the relaxed threshold of    H = 2200 km s −1 to select NLSy1s, the match percentage increases to 94.5% with 9650 sources present in both the works.A comparison of the redshift distributions indicates that a majority of the newly identified NLSy1 galaxies in our work are at higher redshifts (Figure 4).
In Figure 5, we have shown comparison of some of the parameters derived in the two papers.The FWHM values of the broad H emission line were found to be similar to that measured for SDSS-DR12 NLSy1 objects.The logarithmic ratio of the measured quantities is −0.04 ± 0.10.A small fraction of the sources appear to be systematically shifted with DR12 measurements having higher values (Figure 5, top left panel).To investigate the differences, we crossmatched the SDSS-DR12 NLSy1 catalogue with SDSS-DR3 NLSy1 catalogue (Zhou et al. 2006) and compared their measurements of the broad H FWHM values.The top middle panel of Figure 5 reveals the comparison where the same pattern can be seen.This observation indicates a possible issue in estimating the broad H FWHM values for some of the SDSS-DR12 NLSy1 objects.Furthermore, the broad H flux measurements done by us reasonably matches with that published for SDSS-DR12 NLSy1s (average logarithmic ratio 0.06 ± 0.09, Figure 5, top right panel).In this diagram also, a small fraction of objects have systematically lower broad H flux values as reported in SDSS-DR12 NLSy1 catalogue.These objects turned out to be the same that have systematically offset FWHM of the broad H line seen in the top left panel of Figure 5.
We made an attempt to understand such outliers with larger FWHM and smaller flux values of the broad H component reported in the SDSS-DR12 NLSy1 catalogue with respect to our measurements.The fitting of the H line region is complex due to the presence of the [N II] 6549,6585 doublet overlapping with the broad and narrow H components.Therefore, one possibility could be due to issues in properly decomposing these emission lines, which can affect the broad H measurements.In Figure 6, we show the spectral fitting results for one of the outliers.The top panel shows the final result derived after running 5000 iterations of the MCMC fitting.We obtained the broad H FWHM of 758 km s −1 and the flux of 2.9 × 10 −14 erg cm −2 s −1 .The bottom panel shows the fitting result for one of the MCMC iterations where the H line was mainly fitted with the narrow H component, thereby suppressing the broad H component.This led to broader FWHM (1856 km s −1 ) and lower flux (1.7×10 −14 erg cm −2 s −1 ) values for the broad H component.Indeed, outliers with larger FWHM have lower flux values in the SDSS-DR12 NLSy1 catalogue with respect to our measurements (Figure 5, top left and right panels).Therefore, it is possible that the H line region fitting done by Rakshit et al. (2017) might have found local minima (with fitting solutions similar to the bottom panel of Figure 6) for such outliers.
We found close matches between broad H flux and FWHM measurements with the logarithmic ratio distribution having an average of 0.06±0.09and −0.01±0.10,respectively (Figure 5,middle left and central panels).The average logarithmic ratio of the [O III] 5007 line flux measured in this work to that in SDSS-DR12 is 0.04 ± 0.19.The large dispersion is possibly due to larger scatter seen in fainter sources where two measurements appears to deviate though the overall trend remains the same (Figure 5, middle right panel).We have also compared the AGN continuum luminosity at 5100Å and found the results to be similar (logarithmic ratio −0.01 ± 0.21).Finally, a comparison of the FeII strength, i.e.,  4570 parameter, reveals our measurements to be slightly higher than that found by Rakshit et al. (2017) for SDSS-DR12 NLSy1 galaxies.The logarithmic ratio of the two measurements is 0.08 ± 0.18.Altogether, though the spectral modeling strategies of both works are different, the overall results appears similar and we were able to retrieve >80% of SDSS-DR12 NLSy1 galaxies thereby hinting at the robustness of the fitting procedure.
We have also compared our results with the SDSS-DR3 NLSy1 catalogue of 2011 sources (Zhou et al. 2006).There are 23 SDSS-DR3 NLSy1 galaxies which were rejected in our analysis.Among the remaining 1988 objects, 1849, i.e., 93%, were found common in both the works.Adopting the relaxed threshold of    H =2200 km s −1 , the number of common AGN increases to 1928 i.e., 97% of the SDSS-DR3 NLSy1 galaxies.We show the comparison of various spectral parameters derived in this work and that obtained by Zhou et al. (2006) in Figure 7.The logarithmic ratio of the FWHM of the broad H emission line has an average of 0.01 ± 0.05 and unlike SDSS-DR12 NLSy1s, there is no significant offset of sources (Figure 7, top left panel).The flux values of the broad H and H emission lines were also similar with the average logarithmic ratio of 0.05 ± 0.07 and 0.01 ± 0.08, respectively (Figure 7, top middle and right panels).The FWHM of the broad H component measured in both the works has an average logarithmic ratio of 0.01 ± 0.07 indicating a close match (Figure 7, bottom left panel).Similarly, a comparison of the [O III] 5007 line flux and FeII complex strengths reveal the measured values to be very similar with average logarithmic ratio of −0.004 ± 0.160 and 0.05 ± 0.14, respectively.All these results highlight the accuracy of the adopted fitting technique and the robustness of the derived spectral parameters.

The BLSy1 Catalogue
The BLSy1 catalogue was prepared from the parent sample after removing the NLSy1 galaxies and 46 narrow line objects that have the flux ratio of [O III] 5007 and H emission lines >3.The total number of BLSy1 galaxies is 52273.We caution that it may not be appropriate to consider all of them as genuine Seyfert galaxies since luminous broad line quasars are also present in the sample.Indeed, dividing the objects based on their  B values, we found 24401 broad line sources to be quasars with  B < −23 and 27761 broad line Seyferts with absolute -band magnitude > −23 (Schmidt & Green 1983).

MULTI-WAVELENGTH PROPERTIES
The NLSy1 galaxies exhibit peculiar observational characteristics across the electromagnetic spectrum as outlined in Section 1.Though a detailed multiwavelength study of these sources is beyond the scope of the current work, we briefly describe interesting observational features by cross-matching our catalogues with several multi-frequency catalogues.

Radio observations
The NLSy1 objects are generally faint radio emitters with only a small fraction (∼5%) detected in radio surveys (cf.Komossa et al. 2006;Singh & Chand 2018).We cross-matched our NLSy1 sample with the Faint Images of the Radio Sky at Twenty-Centimeters (FIRST, White et al. 1997) with 5 arcsec search radius and found 730 radio emitting NLSy1 galaxies.Interestingly, though our catalogue is more than two times larger than SDSS-DR12 NLSy1 catalogue, the number of radio detected sources has increased only by a factor of ∼1.3.The fraction of radio detected NLSy1 galaxies in our sample is ∼3% which is smaller than ∼5% reported for SDSS-DR12 NLSy1 catalogue (Rakshit et al. 2017).This can be understood by considering the fact that the enhancement in the number of NLSy1s is largely at higher redshifts (Figure 4).If we assume the radio luminosity of the sources to be similar, a higher redshift implies a fainter flux which may remain below the sensitivity of the FIRST survey.
The difference in the adopted methods to compute the  parameter in both works could explain this observation.Indeed, Rakshit et al.
(2017) calculated it as the ratio of the flux densities at 1.4 GHz and optical -band flux, whereas, we have used the conventional definition (Kellermann et al. 1989).
The cross-matching of the BLSy1 catalogue with the FIRST survey led to the identification of 2568 radio detected AGN among which 1975, i.e., ∼77% are found to have  > 10.On comparing the radio-loudness with that estimated for NLSy1 sample, the overall distributions appear similar (logarithmic dispersion ∼0.9, Figure 8) though the median average  value for BLSy1s (∼51) is higher than that estimated for NLSy1 galaxies (∼17).Similarly, the distributions of the 1.4 GHz radio power for both populations also have a similar dispersion (∼1.1, on logarithmic scale) though BLSy1 galaxies are more luminous (median average ∼ 5.7 × 10 40 erg s −1 ) compared to NLSy1 sources (∼ 1.6 × 10 40 erg s −1 ).The distributions are plotted in the right panel of Figure 8.

Optical observations
A strong correlation between continuum luminosity and emission line luminosities has been reported both for quasar and NLSy1 populations covering a wide range of redshift and bolometric luminosities (Greene & Ho 2005;Zhou et al. 2006;Jun et al. 2015;Rakshit et al. 2017).We show the variations of the H, H, and [O III] 5007 emission line luminosities as a function of the 5100Å continuum luminosity for NLSy1 sources in Figure 10.The following relations were found by applying a linear least-square fit on the data: log( H ) = (1.13 ± 0.12) + (0.938 ± 0.002) log  5100 .
All correlations were found >5 significant as also reported for SDSS-DR12 NLSy1 catalogue (Rakshit et al. 2017).These correlations between the continuum and emission line luminosities enable us to adopt the latter while estimating virial black hole masses for the cases where measuring the former is tedious, e.g., in blazars, galaxydominated low-luminosity sources, and Type 2 AGN (Zakamska et al. 2003;Greene & Ho 2005).
We show some of the parameters and physical properties of NLSy1 and BLSy1 sources estimated from the optical spectroscopic analysis in Figure 11.The redshift distribution suggests an increase in number for both populations with redshift.At lower redshifts ( < 0.5), NLSy1s are more common whereas BLSy1 galaxies dominate at higher redshifts.Comparing the absolute -band magnitudes, NLSy1s tend to appear a bit fainter (⟨ B ⟩ = −22.1)with respect to BLSy1 objects (⟨ B ⟩ = −22.9)though their dispersions are similar, ∼1.2 magnitudes.The black hole masses of NLSy1s (⟨log  SE ⟩ = 6.98 ± 0.37,  ⊙ ) were also found to be lower than that of BLSy1 galaxies (⟨log  SE ⟩ = 7.85 ± 0.47,  ⊙ ).This is likely due to the adopted FWHM of the broad H line threshold which is used to compute  SE .Furthermore, since the differences in the bolometric luminosity distributions for both types of sources are not significant (⟨log  bol ⟩ = 45.01,44.74 erg s −1 , for BLSy1 and NLSy1, respectively), a lower  SE also implies a higher Eddington ratio or  Edd for NLSy1 galaxies (Figure 11, bottom middle panel).These results are consistent with those reported for SDSS-DR12 NLSy1 catalogue (Rakshit et al. 2017).
NLSy1 sources typically exhibit strong FeII emission which is quantified using  4570 ( e.g., Vanden Berk et al. 2001).We found the median  4570 parameter value to be 0.71 and 0.38 and 1 dispersion to be 0.41 and 0.24 for NLSy1 and BLSy1 galaxies, respectively.These quantities are similar to those reported by Rakshit et al. (2017) and follow the same trend observed in earlier studies (cf.Bergeron & Kunth 1984;Vanden Berk et al. 2001;Zhou et al. 2006).

X-ray observations
We cross-matched our NLSy1 catalogue with the Chandra source catalogue (CSC v2.0, Evans et al. 2010), the eROSITA Final Equitorial Depth Survey AGN catalogue (eFEDS, Liu et al. 2022), XMM-Newton serenditipious source catalogue (4XMM-DR12, Webb et al. 2020), and live Swift-X-ray Telescope Point Source Catalogue (LSXPS, Evans et al. 2023), in sequential order to identify X-ray emitting sources.Using a search radius of 5 arcsec, we found 567, 195, 979, and 819 matches, respectively.Some of these catalogues, e.g., eFEDS, provide spectral index measurements which we have used to determine its possible correlation with the accretion rate in Eddington units (cf.Risaliti et al. 2009).No assumptions were made while considering spectral indices even though these X-ray surveys cover overlapping yet different energy bands.In Figure 12, we show the variation of the photon index with Eddington ratio for sources detected with Chandra and eROSITA satellites.A positive correlation is evident with the null hypothesis of no correlation is rejected at >5 confidence level.The best-fitted line slope ( = 0.91 ± 0.02) is steeper than that reported in other studies (Risaliti et al. 2009;Laurenti et al. 2022) which can be understood since no detailed X-ray data reduction was performed by us.Indeed, the X-ray spectra of NLSy1 sources exhibit several interesting features, e.g., soft excess below 2 keV (cf.Fabian et al. 2009), which should be properly taken into account while estimating spectral parameters.

Gamma-ray observations
The detection of variable -ray emission from some of the radioloud NLSy1 galaxies have provided strong evidence that even low black hole mass systems can launch powerful relativistic jets (e.g., Abdo et al. 2009).To identify -ray emitting NLSy1 galaxies in our sample, we considered the fourth data release of the fourth catalogue The plotted isodensity contours refer to WISE thermal sources and locations of various source classes are also highlighted.The acronyms QSOs, ULIRGs, LIRGs, and LINERs refer to quasars, ultraluminous infrared galaxies, luminous infrared galaxies, and low-ionization nuclear emission region galaxies, respectively.The contour data are adopted from Massaro et al. (2011).
of -ray sources detected by the Fermi-Large Area Telescope (4FGL-DR4, Ajello et al. 2022;Ballet et al. 2023).Adopting the counterpart positions given in this catalogue and using a search radius of 5 arcsec, we found 21 -ray detected NLSy1 galaxies.The list is provided in Table 1.All sources previously known in the -ray band except one are present in the list (Paliya 2019).The only missing object is the flat spectrum radio quasar GB6 J0937+5008 ( = 0.28) for which our spectral fitting analysis resulted in the broad H line FWHM of 2350.2 ± 231.6 km s −1 thus classifying it as a broad line AGN.Among the newly identified NLSy1s, 9 sources are found to be -ray emitters.
Among the broad line sources, 45 are identified as -ray emitters by cross-matching with the 4FGL-DR4 catalogue (Table 1).Interestingly, there are 10 objects that have absolute -band magnitude  B > −23, i.e., they are genuine Seyfert galaxies.This observation strengthens the idea that low-luminosity AGN can also host relativistic jets.

SUMMARY
In this work, we have carried out a detailed analysis of >2 million SDSS spectra using the publicly available software BADASS to identify NLSy1 galaxies in the latest SDSS-DR17.By reproducing the Balmer emission lines with the Lorentzian profile and also adopting the multi-component continuum fitting which includes contributions from the optical FeII emission, host galaxy, and nuclear power-law AGN radiation, we identified 22656 NLSy1 galaxies with broad H emission line FWHM <2000 km s −1 within uncertainties.Furthermore, >75% of these objects are found to be low-luminosity AGN with  B > −23 (Schmidt & Green 1983).This exercise also led to a new catalogue of 52273 BLSy1 galaxies.Comparing with previous works (Zhou et al. 2006;Rakshit et al. 2017), we found >80% of their NLSy1s to be present in our catalogue and the corresponding optical spectroscopic parameters were also found to be comparable, thereby supporting the robustness of our analysis procedure and results.Based on the estimated parameters and derived quantities, e.g.,  SE , we confirm earlier findings of NLSy1s being AGN powered by rapidly accreting low-mass black holes and the steepening of the X-ray spectrum with increasing Eddington rate (cf.Boller et al. 1996).We conclude that with the advent of the ongoing and upcoming wide-field, multiwavelength sky surveys, e.g., VLASS, this new catalogue of NLSy1 galaxies will enable us to explore the physics of this enigmatic class of AGN in an unprecedented detail hence sowing the seeds for their future observations with the next-generation of giant telescopes.The catalogue is made public at https://www.ucm.es/blazars/seyfert and also provided as a supplementary material of this article.

Figure 2 .
Figure 2.This plot shows the distributions of the measured S/N ratio for the H and H regions.The areas of the plotted histograms are normalized to unity.

Figure 3 .
Figure 3.The left and middle panels show the variation of the broad H emission line FWHM and flux, respectively, with that measured for the broad H line.The red line refers to the best-fitted correlation for the plotted quantities.The distribution of the sources on the optical FeII strength versus H line FWHM plane is shown in the right panel.In all plots, the number density of the data points is colour coded with lighter colours representing larger density of sources.
e.,    − Δ   ≤ 2000 km s −1 .(ii) the flux ratio of [O III] 5007 and H emission lines within estimated uncertainties is <3, where we propagated the uncertainties in the H and [O III] 5007 flux values while computing the ratio.

Figure 4 .
Figure 4.The redshift histograms of NLSy1 galaxies present in our sample and that included in the SDSS-DR12 NLSy1 catalogue.

Figure 5 .
Figure 5.This plot shows the comparison of various spectral parameters obtained in this work and that published for SDSS-DR12 NLSy1 catalogue (R17, Rakshit et al. 2017).The colour coding is done based on the number density of sources.The red line refers to the one-to-one correlation.

Figure 6 .
Figure 6.Results of the spectral fitting done on one of the outliers found in Figure 5, top left panel.The top panel shows the fitting result obtained after 5000 iterations of the MCMC fitting, whereas, the bottom panel refers to that obtained for one of the MCMC iterations.The measured FWHM and flux values for the broad H component is mentioned.See the text for details.

Figure 7 .Figure 8 .
Figure 7.This plot shows the comparison of various spectral parameters obtained in this work and that published for SDSS-DR3 NLSy1 catalogue (Z06, Zhou et al. 2006).Other information are same as in Figure 5.

Figure 9 .
Figure 9. WISE colour-colour diagram for objects studied in this work.The plotted isodensity contours refer to WISE thermal sources and locations of various source classes are also highlighted.The acronyms QSOs, ULIRGs, LIRGs, and LINERs refer to quasars, ultraluminous infrared galaxies, luminous infrared galaxies, and low-ionization nuclear emission region galaxies, respectively.The contour data are adopted fromMassaro et al. (2011).

Figure 10 .
Figure 10.A comparison of the luminosity of the emission lines and 5100Å continuum luminosity .The colour coding is done based on the number density of sources.The red line refers to the best-fitted correlation.See the text for details.

Figure 11 .
Figure 11.Comparisons of various optical spectroscopic parameters, measured/derived, for NLSy1 and BLSy1 sources.See the text for details.

Figure 12 .
Figure 12.The variation of the X-ray photon index as a function of the Eddington ratio for NLSy1 galaxies detected with Chandra and eROSITA satellites.The red line corresponds to the best-fitted correlation.