Optical spectroscopy of blazars for the Cherenkov Telescope Array - III

Blazars are a type of active galactic nuclei (AGNs) that display exceptional observational properties. Their unique characteristics include beamed non-thermal emission from the radio to high-energy $\gamma$ rays, as well as strongly polarized optical and radio emission ($\geq$ 3%) (Angel & Stockman 1980; Angelakis et al. 2016; Lister et al. 2011, see e.g.), variability from a few percent up to a few orders of magnitude on different timescales at all wavelengths (Wagner & Witzel 1995; Falomo et al. 2014, see e.g.), and (for some) the ejection of superluminal radio blobs (Vermeulen & Cohen 1994, see e.g.). These characteristics are generally explained by Doppler boosting of the jet emission with Lorentz factors of up to $\sim$40 (Jorstad et al. 2017, e.g.).

Recent decades have seen the emergence of a new observational window on the Universe: very-high-energy (VHE; $E$ $>$ 100 GeV) $\gamma$-ray astronomy. Thanks to three major imaging atmospheric Cherenkov telescopes (Aharonian et al. 2006; Aleksić et al. 2016b; Aleksić et al. 2016a; Holder et al. 2006, H.E.S.S.¹¹ 1 https://www.mpi-hd.mpg.de/HESS/, MAGIC²² 2 https://magic.mpp.mpg.de/, and VERITAS³³ 3 https://veritas.sao.arizona.edu/; see e.g.), 252 sources have been detected so far at VHE (Wakely & Horan 2008, TeVCAT⁴⁴ 4 http://tevcat.uchicago.edu/;), and AGNs make up approximately one-third of these sources. Since 2008, this field has also benefitted from the detection of $\gamma$ rays at high-energy (HE; 100 MeV $<$ $E$ $<$ 100 GeV) with the Large Area Telescope (LAT) on board the Fermi Gamma Ray Space Telescope⁵⁵ 5 https://fermi.gsfc.nasa.gov/ (Atwood et al. 2009). In the coming years, the future world-wide Cherenkov Telescope Array Observatory (CTAO⁶⁶ 6 https://www.cta-observatory.org/) will start its operations with a lower energy threshold for VHE $\gamma$-ray detections, down to a few tens of GeV, and a flux sensitivity improved by a factor of $\sim$10 compared to the current facilities.

Blazars, which include BL Lacs and flat-spectrum radio quasars (FSRQs), are by far the most numerous type of AGNs detected at VHE. The study of their population and evolution is key to high-energy astrophysics (Cherenkov Telescope Array Consortium et al. 2019). However, one of the biggest difficulties in investigating VHE $\gamma$-ray emission from AGNs resides in their limited number, since only 84 of them are known to date. In the future, CTAO is expected to detect hundreds of AGNs of various types and make it possible to carry out population studies. The broad energy range covered by CTAO includes energies where $\gamma$ rays are unaffected by absorption while propagating in the poorly characterized extragalactic background light (Gould & Schréder 1966; Hauser & Dwek 2001; Dwek & Krennrich 2013, EBL; e.g.) and extends to the VHE regime, where the spectra are strongly distorted by the EBL. This will help reduce systematic effects combining spectra from different instruments, leading to a more reliable EBL determination, thus making it possible to constrain blazar models up to the highest energies with less ambiguity. One challenge in the study of the cosmological evolution of blazars is the difficulty in obtaining redshifts, particularly from the nearly featureless, continuum-dominated spectra of BL Lacs. Indeed, many of the early studies using X-ray or radio-selected samples had highly incomplete redshift measurements, even though the samples were confined to relatively bright sources (Massaro et al. 2015a, BZCAT⁷⁷ 7 https://www.ssdc.asi.it/bzcat/;). Uncertainties related to extrapolating unknown redshifts from the measured set of redshifts have further complicated population studies.

The difficulty in measuring BL Lacs redshifts has not yet been solved and it also plagues present-day samples of blazars (Shaw et al. 2013; Paiano et al. 2017a; Massaro et al. 2015b; Massaro et al. 2015c; Peña-Herazo et al. 2020). Indeed, while the redshift distribution of $\gamma$-ray FSRQ ($\geq$ 90 % of them with known redshifts) displays a maximum at $z$ $\sim$ 1 (Ajello et al. 2012), the redshift distribution of BL Lacs peaks at $z$ $<$ 0.5 (Ajello et al. 2014). This is reflected in the luminosity functions of the two classes. The luminosity function of FSRQ has a positive, luminosity-dependent evolution, while BL Lacs have a negative evolution at low luminosities, suggesting a very different evolution among the two classes. However, the results of BL Lacs are very likely biased by the lack of redshifts. For example, considering redshift lower limits from absorbing systems, a relevant population of BL Lacs with redshifts greater than 0.3–0.5 appears plausible. It is thus clear that samples with poor redshift coverage cannot be used to obtain reliable results on the cosmic evolution of $\gamma$-ray BL Lacs (Ajello et al. 2014). The problem is more severe for VHE BL Lacs, where the number of sources detected at higher redshift becomes sparser (less than ten BL Lacs at $z$ $>$ 0.3, of which two are at $z$ $>$ 0.4).

The VHE $\gamma$-ray detection of many BL Lacs in different redshift bins at $z$ $>$ 0.3 will allow for a more complete determination of the EBL density (see CTA Consortium et al. 2019). The redshift evolution of diffuse light in the Universe provides a powerful constraint on the galaxy evolution, the formation of stars and galaxies, and cosmology. In this context, the EBL is of fundamental importance. BL Lacs have so far been successfully used to investigate the EBL density (Ajello et al. 2014; Biteau & Williams 2015; Acciari et al. 2019, e.g.). However, a major ingredient for its success is a robust redshift determination of the sources used for estimating it. It is therefore of great importance to make an effort to measure the redshift of a large fraction of AGNs detected with Fermi-LAT and that are likely to be detected with CTAO based on the extrapolation of their LAT spectra.

This is the third paper in a series that aims to determine the redshift of a sample of blazars that are promising targets for CTAO. In the previous two papers, we determined 25 spectroscopic redshifts with values between 0.0838 and 0.8125, 2 tentative redshifts, and 5 lower limits. In this paper, we present detailed results for 24 new targets. The results of the Lick spectra for 17 additional targets in our sample (see Section 2) are reported in Appendix A.

In this work, we use the sample introduced in Goldoni et al. 2021 and Kasai et al. 2023, which we briefly describe here. The sample has been selected to comprise the best candidates for future observations with the CTAO, extrapolating the spectral information of the hardest sources detected by Fermi-LAT at $E$ $>$ 10 GeV to the VHE.

We started from the sample of BL Lacs and blazar candidates of uncertain type (BCUs) in the Third Fermi LAT Catalog of High energy Sources (Ajello et al. 2017, 3FHL;), where the majority of the 1040 sources (64%) have no redshift measurement. To estimate their emission in the CTAO band, we extrapolated their average 3FHL energy spectrum, adding a 3 TeV exponential cutoff to simulate the expected spectral curvature in the VHE band. We then performed a Monte Carlo simulation with the Gammapy⁸⁸ 8 https://gammapy.org software (Deil et al. 2017; Nigro et al. 2019; Donath et al. 2023), using the publicly accessible CTAO performance files⁹⁹ 9 https://www.cta-observatory.org/wp-content/uploads/2019/04/CTA-Performance-prod3b-v2-FITS.tar.gz^,1010 10 https://zenodo.org/record/5163273#.Yg9-yPVBzPZ. To account for the energy- and redshift-dependent $\gamma$-ray opacity of the EBL, we used the model of Domínguez et al. 2011. For 3FHL sources with no redshift available, we assumed a value of $z=0.3$, similar to that of $z_{\rm med}=0.33$ reported by Shaw et al. 2013 and $z_{\rm med}$ = 0.285 by Peña-Herazo et al. 2020. Incidentally, the median redshift of the sources reported in our two previous papers is $z_{\rm med}=0.255$, which excludes the lower limits with a mean value of $z_{\rm med\,ll}=0.618$. This value is consistent with this choice. After a first simulation run, we selected a smaller sample on which we performed a literature search that allowed us to correct the redshift of 32 sources (see Paper I) and rerun the simulations. As a result of these simulations, we estimated the minimum observing time required to get a 5$\sigma$ detection with CTAO and the 3FHL sources with no redshift, which require less than 30 hours to be detected, are included in our sample. The final sample contains 165 targets.

As in the first two papers in the series (Paper I and Paper II), our main goal is to determine the spectroscopic redshifts or at least a spectroscopic lower limit for the BL Lacs in our sample. This is achieved by looking at stellar absorption features found in the luminous elliptical galaxies that usually host BL Lacs (Urry et al. 2000), such as Ca ii HK doublet, Mg_b, and Na i D. Despite the scarcity of their detection in BL Lac spectra, we also searched for emission lines such as [O ii], [O iii], H$\alpha,$ and [N ii]. A spectral resolution, $\lambda$/$\Delta\lambda,$ of a few hundred (ideally up to 1000), along a signal-to-noise ratio (S/N) per pixel of $\sim$100 is required to detect features with EWs $\lesssim$ 5 Å expected in these sources. These two conditions have demonstrated good results in the past, as shown in Pita et al. 2014 and Papers I and II in our series. In difficult cases, the observations were intended to achieve at least one of these criteria. When possible, observations were carried out during a photometric minimum of the target to benefit from the higher S/N of the host galaxy features in the spectrum due to a reduced non-thermal emission of the central AGN.

Observations were performed for 41 blazars between 2020-09-20 and 2022-03-26 for a total of 68 hours using the Keck/ESI¹¹¹¹ 11 Echellette Spectrograph and Imager (ESI) on the Keck II telescope, https://www.keckobservatory.org/about/telescopes-instrumentation (Sheinis et al. 2002), Lick/KAST¹²¹² 12 KAST Double Spectrograph on the Shane telescope at the Lick Observatory, https://mthamilton.ucolick.org/techdocs/instruments/kast/, SALT/RSS¹³¹³ 13 Robert Stobie Spectrograph (RSS) on the Southern African Large Telescope (SALT), www.salt.ac.za/telescope (Burgh et al. 2003), VLT/FORS¹⁴¹⁴ 14 FOcal Reducer and low dispersion Spectrograph (FORS) on the Very Large Telescope (VLT), https://www.eso.org/sci/facilities/paranal/instruments/fors.html (Appenzeller et al. 1998), and GTC/OSIRIS ¹⁵¹⁵ 15 Optical System for Imaging and low-Intermediate-Resolution Integrated Spectroscopy (OSIRIS) on the Gran Canarias Telescope (GTC) http://www.gtc.iac.es/instruments/osiris/. We note that Keck II and GTC have primary mirror diameters of 10 m and SALT’s is 11 m, whereas the primary mirror diameter for the VLT is 8.2 m and the mirror diameter of the Shane telescope is 3 m.

The observations and data analysis follow similar procedures as described in Papers I and II, particularly with respect to the observational configuration, data reduction, flux calibration, telluric correction, and spectral dereddening procedures. After this treatment, the spectral shape is that generally expected for BL Lacs, showing a power-law continuum superposed with the emission of a host galaxy and (at times) weak emission lines. For some objects with a power law spectrum, namely, WISE J053629.06$-$334302.5, PKS 1424+240, and RBS 1457, as well as (to a smaller extent) 1RXS J203650.9$-$332817 and PMN J2221$-$5224, a flattening is displayed in the red part of the spectrum, which may be due to problems with flat-field frames. In this paper we give only a brief description of the VLT/FORS and GTC/OSIRIS instruments and of their data reduction as data from these instruments were not included in the first two papers in the series.

The visual and near-UV FOcal Reducer and low dispersion Spectrograph (FORS), part of the VLT of the European Southern Observatory (ESO), is a multi mode (imaging, polarimetry, long slit and multi-object spectroscopy) optical instrument mounted on the VLT UT1 Cassegrain focus (Appenzeller et al. 1998). To achieve the desired coverage, sensitivity, and spectral resolution on most targets, we selected GRISM 600RI+19, given its sensitivity in the range of 5000–8500 Å, with a spectroscopic resolution of $\lambda$/$\Delta\lambda\sim 800$ and a 40-80% efficiency. For RBS 1751, we selected GRISM 1200B+97, which is sensitive in the range of 3600–5100 Å, with a spectroscopic resolution of $\lambda$/$\Delta\lambda\sim 1200$ and a 60-80% efficiency. In all cases, we used a 1.3 arcsec slit, whereas a 5 arcsec slit was used for standard stars. Data reduction was performed using the FORS pipeline (Izzo et al. 2010) up to spectral extraction, which was then performed using Image Reduction and Analysis Facility (Tody 1986, IRAF,).

OSIRIS is the work-horse imaging and spectroscopic instrument for the GTC (Cepa et al. 2003) installed at the Nasmyth focus. We carried out observations using the R1000B and R1000R gratings, with an overall spectral coverage from 3650 Å to 10000 Å at a resolution of $\lambda$/$\Delta\lambda$ $\sim$ 1000. The 1.0 arcsec slit was used for the target and for the standard star. We used standard IRAF (Tody 1986) procedures for long-slit observations to reduce tha data. Table 1 lists the sources and observational parameters used for the spectroscopic measurements of the 24 targets with spectral features or high S/N featureless spectra. Table 2 gives the parameters used in the observations at each telescope.

All sources, except for 1RXS J035000.4+064053 and SUMMS J224017$-$524111, have been included not only in the 3FHL, but also in the Second Fermi-LAT Catalog of High-Energy Sources (Ackermann et al. 2016, 2FHL;); therefore, they has been detected by Fermi-LAT at energies higher than 50 GeV. 1RXS J035000.4+064053 and SUMMS J224017$-$524111 are the only sources presented in Table 1 not classified as a BL Lac object in the 4FGL-DR4 catalogue (Ballet et al. 2023). However, those two sources are classified as high-energy BL Lac objects (HBLs) in the 3HSP catalogue (Chang et al. 2019).

Table 10 in Appendix A presents 26 spectra of 18 blazars observed with Lick/KAST, one of which is also reported in Table 1 (1RXS J035000.4$+$064053). The 17 remaining are not discussed in detail in this paper because of their low S/N and featureless spectra.

To determine the redshift, we searched for absorption and emission features in the spectra, requiring a minimum of two independent detections corresponding to the same redshift for a robust determination. Details on the lines we searched can be found in Paper II. For each feature, we measured the EW by fitting the continuum with cubic splines and integrating the flux for each pixel. The uncertainties in the EW are determined by the quadrature sum of the normalized flux errors and the continuum placement contribution (Sembach & Savage 1992, see).

The redshift uncertainty is estimated by considering the uncertainty in the position of the detected features. The feature position and its uncertainty were determined by a Gaussian fit. To this uncertainty, we added in quadrature the wavelength calibration uncertainty. The results are presented in Tables 3, 4, 5, and 6. In the case of SUMMS J224017$-$524111, by considering the large number of emission lines detected, the results are reported in a dedicated table.

We modelled the SED with the sum of a power law for describing the jet emission and an elliptical galaxy template for describing the host (Mannucci et al. 2001; Bruzual & Charlot 2003). When needed, Gaussian emission features are added in the modelling (Pita et al. 2014) and only one template was used per spectrum for simplicity. Only two free parameters are needed: the power -law slope and jet-to-galaxy ratio. Table 6 shows the results of the fits.

We also estimated the absolute magnitude of the detected host galaxies. Slit losses are estimated by assuming the host galaxy effective radius, $r_{\rm e}$, to be 10 kpc for a de Vaucouleurs profile. From the template spectra, we computed the K-corrections and did not apply evolutionary corrections. In the case of non-detection of a host, the spectra are fitted with a power law and normalised at the band centre. Due to the high S/N of the spectra, the relative errors on the fitted parameters are very small (10^-3 or so). However, there is residual curvature seen in the spectra that maybe intrinsic (i.e. slope change) or due to calibration issues (fluxing and/or extinction correction). To take this into account, we fit the red and blue halves of the spectra separately and take the difference between the blue and the red parameters as 1 sigma errors on the parameters of the total spectrum. The uncertainties quoted in Table 6 are three times (i.e. three sigma) these values.

For the 24 sources in Table 1, twelve spectroscopic redshifts were determined, ranging from 0.2223 to 0.7018, one tentative redshift and two redshift lower limits ($z$ $>$ 0.6185 and $z$ $>$ 0.6347) were obtained. In the following we discuss the results of our observations for each of the sources.

Spectroscopic observations triggered during a photometric minimum can improve the redshift measurement efficiency (see Section 6.12 of this paper and Paper I). This technique takes advantage of the weakening of the non-thermal emission during the photometric minimum of the AGN, which allows for the emergence of the host galaxy features that are used to determine the redshifts. For testing the importance of observing objects during a low-activity state, we investigated the light curves of the sources in Table 1. The light curves were collected from the Zwicky Transient Facility (ZTF) (Masci et al. 2019)²⁰²⁰ 20 http://www.ztf.caltech.edu and the All Sky Automated Survey for SuperNovae (ASAS-SN) (Shappee et al. 2014; Kochanek et al. 2017)²¹²¹ 21 https://asas-sn.osu.edu// in the period between January 2020 – March 2022 to check if our spectroscopic measurements are performed during an high-, intermediate-, or low-activity state and the effect on the redshift measurement efficiency. We briefly review the two survey characteristics. ZTF, which uses the 48-inch Oschin Schmidt telescope at Palomar Observatory, is sensitive down to r$\sim$20.6 (5$\sigma$ in 30 sec) and has a 1″ pixel scale, but its coverage is limited mostly to the northern hemisphere. ASAS-SN is sensitive to g$\sim$ 18 (5$\sigma$ in 5 minutes) with a 8″ pixel scale and, thanks to several sites around the globe, it covers the entire sky. Whenever available, we report ZTF measurements that have smaller uncertainties.

We checked the light curves for consistency with archival photometric measurements and with our spectrophotometric magnitudes. This check led to the rejection of the ASAS-SN light curves of 1RXS J184230.6$-$584202 and SUMSS J113032$-$780105, which are strongly contaminated by nearby stars. Furthermore, the good-quality light curves of B2 0557+38, 1RXS J085802.6$-$313043, and TXS 1742$-$078 are not available due to their weakness. In Tables 7 and 8, the time separation between the spectroscopic measurement and the nearest observing data collected by ZTF and ASAS-SN, respectively, are reported together with the observed magnitude, the median, maximum, and minimum magnitude calculated over the entire period considered, and the redshift or lower limit measured in this paper.

Taking a look at the ZTF light curves (Table 7), at the time of our spectroscopic observations, 7 out of 11 sources have been observed at a magnitude comparable or higher (thus, a lower brightness) with respect to the median magnitude (hereafter, ‘low state’) during 2020–2022. In particular, in the case of 1RXS J165655.0$-$201049 and RBS 1888, the sources have been observed almost at the minimum magnitude. For five of these seven sources, a redshift has been determined with our new spectroscopic observations, the only exceptions being RBS 1457 and RBS 1751 (for which a high lower limit has been estimated). The other four sources have been observed at a lower magnitude (thus higher brightness) with respect to the median magnitude (‘high state’) and no redshift has been determined.

In case of the ASAS-SN light curves, five out of eight sources have been observed in a low or intermediate (i.e. at a magnitude comparable to the median value) state, resulting in 2 redshift determinations. On the other hand, three sources have been observed in an high state with respect of the median value resulting in one redshift determination. The only source for which the redshift has been measured in high state is SUMSS J224017$-$524111, which has a low jet-to-galaxy ratio (2.6 $\pm$ 0.2) and several emission lines, thereby simplifying the detection.

In total, 12 sources in a low state (or intermediate in case of PMN J1539$-$1128 and SUMSS J082627$-$640414) yield 7 redshifts, while 7 sources in high state yield 1 redshift. As expected, ‘low state’ observations are preferable for redshift determination. We suggest that observers, when possible, should take into account this parameter when programming their spectroscopic observations.

In this work, 41 BL Lacs, which have been included in the 3FHL catalogue and selected as potential candidates for future observations with CTAO, were observed with Keck II, Lick, SALT, VLT, and GTC telescopes. The 41 targets are divided in 24 taken with 8-m class telescopes and whose spectra have been described in detail and 17, which only have low S/N, featureless spectra taken with the Lick telescope and are not discussed in detail here. Of the 24 sources, 12 spectroscopic redshift (values between 0.2223 and 0.7018) and 1 tentative redshift (0.6622) have been determined. We also measured two redshift lower limits, $z$ $>$ 0.6185 and $z$ $>$ 0.6347. Taking a look at the quality of the spectra collected, for 15 out 24 BL Lacs, the S/N of the spectrum has been higher than 100; but it is only for 6 of them that a redshift measurement has been obtained. For the remaining 9 objects, the S/N values range between 23 and 95, from which we measured 6 firm redshifts, 1 tentative redshift, and 2 lower limits. As already noted in Paper II, an high S/N of the spectrum is an important but not sufficient condition for obtaining a redshift measurement. In fact, the contribution of the non-thermal jet emission can play an even more important role for the detection. We checked the activity level of the AGN in our sample investigating the ZTF and ASAS-SN light curves at the time of our spectroscopic observations and we found a high level of efficiency in the redshift detection in case of a low- or intermediate-activity state of the AGN; and, thus, a lower contribution of the non-thermal jet emission to the continuum spectrum. The case of PKS 1424+240, in which the emission lines observed by GTC during a low activity state are three times more intense than the lines previously observed, allowing us to improve the redshift determination of the source and confirming the importance of the optical state of the AGN. Following this indication, we are monitoring the optical state of the sources in our sample with the Telescopi Joan Oro (Colomé et al. 2010, TJO;) and the Rapid Eye Mount (Zerbi et al. 2001; Covino et al. 2004, REM;) in order to perform follow-up spectroscopic observations of the targets in a low state.

The 11 firmly detected host galaxies in this paper have an average magnitude of $M_{R}$ = $-$22.9, brighter than the values obtained in Paper I and II. All the 11 sources can be adequately fitted with a local elliptical template (Mannucci et al. 2001). Emission lines are detected for five objects, all of them with an equivalent width smaller that the 5 Å, limit historically used to separate FSRQ and BL Lacs.

The redshift measurement efficiency for the 24 sources taken with 8-m class telescopes is about 54 per cent, comparable to the results in Paper I and Paper II. The median redshift obtained with the new observations $z_{\rm\,med}^{\text{III}}$ = 0.39 is comparable to the value reported in Paper II ($z_{\rm\,med}^{\text{II}}$ = 0.37) and higher than the value reported in Paper I ($z_{\rm\,med}^{\text{I}}$ = 0.21). We also computed the combined results of our programme to date (see Table 9). To do this, we took into account repeated observations. In this paper, we re-observed PKS 1424+240, PMN J1539$-$1128, 1RXS J184230.6$-$584202 (confirming its tentative redshift reported in Paper I), 1RXS J203650.9$-$332817, and RBS 1751 (confirming its tentative lower limit reported in Paper I). Thus, we have two tentative redshifts remaining (B2 0557+38 from this paper and RX J0819.2+0756 from Paper II) and no tentative lower limits remaining. Up to now, we observed 63 independent targets and obtained 37 solid redshifts with an efficiency of about 59% and a median redshift $z_{med}^{\text{tot}}$ = 0.30.

By considering the high efficiency of redshift determinations in cases of low activity of the AGN, re-observations of sources for which a featureless spectrum has been obtained, despite a S/N $>$ 100 spectrum, will be performed during a low activity state. For a better characterization of our sample of BL Lac objects, we are also currently performing deep near-IR imaging of these sources (Nilsson et al., in prep.). The most distant host galaxy we detected is that of RX J2156.0+1818, for which we report a spectroscopic lower limit of 0.6347 here.

With the new spectroscopic observations performed here, we obtained 9 redshifts and 1 tentative redshift higher than 0.3, similar to the numbers obtained in paper II (i.e. 8 redshift and 1 tentative redshift with $z$ $>$ 0.3), but significantly higher than the numbers obtained in paper I (i.e. 2 redshifts and 1 tentative redshift with $z$ $>$ 0.3). In total, 19 redshifts and 1 tentative redshift with $z$ $>$ 0.3 have been determined in our programme. As a comparison, 9 BL Lacs with redshifts of $z$ $>$ 0.3 are reported in TeVCat presently. Our sources thus represent an important pool of possible TeV sources at $z$ $>$ 0.3 that can be targets of VHE ground-based observatories such as CTAO. These objects are important for determining the luminosity function of BL Lacs and for EBL studies. For this reason, this redshift measurement campaign is recognized as necessary to support for future CTAO observations of AGN, in particular, for CTAO Key Science Projects on AGNs (Cherenkov Telescope Array Consortium 2020; Abdalla et al. 2021, e.g.).