betaFeedback

WorksPhysics and Astronomy

The Galaxy Activity, Torus, and Outflow Survey: N. Little Red Dots are not dust deficient: constraints from Seyfert 1 JWST spectra

Abstract

Little Red Dots (LRDs) are compact, red sources discovered by JWST. Most show broad permitted emission lines characteristic of AGN, but other properties differ from local AGN, such as X-ray and radio weakness and an apparent scarcity of dust. If LRDs at z>4 are indeed dust deficient, the bolometric luminosities and black hole (BH) masses measured from SED fitting would be overestimated, and their red optical continuum would be intrinsic rather than reddened. Here we use rest-frame 3-27 micron JWST spectra of a sample of nearby (D<60 Mpc) Seyfert 1 (Sy1) nuclei to place constraints on the dust emission of LRDs. We normalised the individual Sy1 spectra by their broad Hαα and 5100 A continuum luminosities and stacked them, obtaining scalings that reproduce, within a factor of three, the rest-frame near-infrared emission of a NIRSpec+MIRI stack of 249 LRDs at 2.3<z<9.3 and the near- and mid-infrared emission of a local LRD at z=0.1, with the 5100 A continuum giving the more reliable prediction when modest extinction is considered (Av<1 mag). However, simultaneously reproducing the optical and infrared SEDs of LRDs requires a more reddened accretion disc (Av ~ 2 mag) and a torus with a lower reprocessing efficiency (Ct ~ 0.1-0.2) than inferred for local Sy1 galaxies. This combination offers an alternative to BH-star scenarios that explain the optical continuum as an unreddened blackbody of T ~ 5000 K. Our model, based on local Sy1s, yields a more plausible infrared-to-optical energy balance, suggesting that LRDs are not dust deficient. From the torus model that reproduces the optical and near-infrared SED of the high-redshift LRD stack, we estimate a dust mass of Mdust≃104 M⊙M_{\rm dust}\simeq10^4\,M_\odot, consistent with the dust masses typical of AGN tori.

Introduction

The study of active galactic nuclei (AGN) has become increasingly compelling with the advent of the James Webb Space Telescope (JWST). This is particularly the case for AGN at redshift z≳4z \gtrsim4 [54] (Kovács et al. 2023; Amdam et al. 2025), which are no longer restricted to bright quasars (Lbol≳1046L_{\mathrm{bol}} \gtrsim10^{46} erg s−1^{-1}). JWST/NIRSpec spectroscopy has revealed a large population of broad-line AGN with Lbol≃1044−45L_{\mathrm{bol}} \simeq10^{44-45} erg s−1^{-1} and black hole (BH) masses of MBH≃106−8M⊙M_{\mathrm{BH}} \simeq10^{6-8} M_\odot (e.g. [91, 74, 130, 41, 70, 104]). These AGN luminosities and BH masses are now closer to the average values of local active galaxies (see Table 1), which now, thanks to JWST/MIRI, can be studied in the mid-infrared with unprecedented spectral resolution and coverage [140, 43, 108, 56, 32, 47] (García-Bernete et al. 2024).

IDSeyfertDist (Mpc)Scale (pc arcsec−1^{-1})ii (deg)fcf_{\mathrm{c}}log⁡NH\log N_{\mathrm{H}} (cm−2^{-2})log⁡L2−10 keV\log L_{2-10\ \mathrm{keV}} (erg s−1^{-1})log⁡Lbol\log L_{\mathrm{bol}} (erg s−1^{-1})log⁡MBH\log M_{\mathrm{BH}} (M⊙M_{\odot})log⁡LbolLEdd\log\frac{L_{\mathrm{bol}}}{L_{\mathrm{Edd}}}
NGC1097114.5 (a)70500.1520.6040.9041.908.10-4.30
NGC32271.523.7 (b)115560.4020.9542.1043.406.58-1.28
NGC40511.516.6 (c§)80350.8020.0041.3342.635.95-1.42
NGC43951.54.30 (d)21400.7421.0440.5041.804.43-0.73
NGC68141.221.7 (e)105170.5120.9742.3143.616.89-1.38
NGC72131.522.0 (a)107200.1220.0041.9543.256.41-1.26
NGC73141.8/1.9*18.3 (a)89640.8421.6042.3343.636.70-1.17
NGC7469158.0 (f)281450.4020.5343.1944.497.48-1.09

Table 1. Properties of the Seyfert 1 galaxies.

Interestingly, the population of AGN at z≳4z \gtrsim4 discovered by JWST show some properties that are substantially different from those of nearby AGN. Despite the broad Balmer lines detected in a significant fraction of them, indicating that they are type-1 AGN, they are extremely weak in the X-rays [5, 139, 134, 42] (Maiolino et al. 2025). They are also faint at radio frequencies [95, 96], and they show weak or no significant ultraviolet (UV) to optical variability [131] (Kokorev & Harkane 2025), although this aspect has been debated [35, 82, 121] (Ji et al. 2025; Lambriades et al. 2026).

While many of the broad line AGN found by JWST have blue colors (so-called Little Red Dots, [53, 15, 42]), a fraction of them (which depends on luminosity and selection band) is characterized by red optical slope and blue UV slope, giving them a characteristic “V-shaped” continuum – these have been named “Little Red Dots” [93, 50, 74]. In addition to the several peculiar properties discussed above, LRDs are often characterized by Balmer absorption lines [93, 68, 67, 29, 94] and also Balmer breaks. The latter were initially interpreted as tracing evolved stellar populations, resulting in extremely large stellar masses (possibly cosmology-breaking; [62]) and extremely high stellar densities [9]. However, subsequent studies have revealed that such features are tracing dense gas along the line of sight [63, 98, 27] (Ji et al. 2025). The nature of such absorbing gas along the line of sight is matter of debate, with scenarios ranging from Broad Line Region (BLR) clouds to a low-dust-content torus [68 Madau & Maiolino] 2026; Zhong et al. 2026), to a partially ionized cocoon [114, 126], and to an embedding pseudostellar atmosphere [98, 86].

Another important question is the origin of the red optical shape of LRDs. Many authors suggest that the optical continuum redward of the Balmer break resembles a black body (BB) with temperature around 5000 K [26, 83], and this is interpreted as the signature of a pseudostellar photosphere [86, 72]. Sneppen et al. [125] ascribe the optical spectral shape to nebular emission, based on the detection of a Paschen jump in a few LRDs. Other authors interpret the optical shape in terms of dust reddening, possibly associated with the classical torus [e.g. 87], but with lower dust-to-gas ratio than in typical AGN, owing the lower metallicity typical of LRDs (∼0.1\sim0.1 Z⊙Z_\odot, consistent with the average metallicity normal galaxies at the same cosmic epoch).

The scenario of dust absorption was initially discarded because of early claims that LRDs show no or little evidence for dust emission [136, 2]. The non-detections with the Atacama Large Millimeter/submillimeter Array (ALMA), which provide upper limits on the dust mass down to ∼106\sim10^6 M⊙M_\odot (Casey et al. 2025; Greene et al. 2026; Akins et al. 2025), are often taken as evidence of lack of dust. However, one should take into account that the dust mass of the nuclear torus is typically very low: ∼6×103\sim6 \times10^3 M⊙M_\odot as measured from high-angular resolution ALMA observations of local Seyfert galaxies [40], and between ∼102\sim10^2 and 10410^4 M⊙M_\odot from SED fitting of SDSS quasars at z≤0.5z\leq0.5 [55]. For what concerns the host galaxy, the vast majority of distant, low metallicity galaxies are undetected with ALMA - therefore the ALMA non-detection of LRDs are not surprising, but simply consistent with the fact that LRDs are hosted in low-metallicity and low-mass galaxies. Additionally, sub-mm non-detections at high-redshift should be interpreted cautiously, because the increasing temperature of the Cosmic Microwave Background (CMB) reduces the contrast between dust and the CMB emission [24].

Interestingly, more recent works have found evidence for hot dust emission (T∼1000T\sim1000-15001500 K), typical of AGN, in the rest-frame near-infrared spectrum of many LRDs, both individually and in stacks [68, 28, 84, 15, 104]. Obviously, the detection of this hot dust emission has been more difficult at high-zz, because the thermal bands are shifted outside JWST’s sensitive bands, which required the use of stacking, but in LRDs at z≲3z\lesssim3 hot dust is clearly detected (Juodžbalis et al. 2024; Ji et al. 2026; Lin et al. 2026b,c; Brazzini et al. 2026). In the case of local LRDs (z≲0.5z\lesssim0.5), for which there are mid-infrared SEDs available (Ji et al. 2026; Lin et al. 2026b), dust at temperatures of T∼100T\sim100-300 K, and up to T∼1000T\sim1000 K, have been reported. The question is whether the dust emission of LRDs is drastically different from that of normal AGN or not. Analysing the Red Rosetta Stone, an LRD at z=2.6z=2.6, Brazzini et al. [15] find a deficiency of mid-infrared emission by a factor ranging from 2 to 6 relative to the continuum at 1 μm1\ \mu\mathrm{m} and using the local AGN templates from García-Bernete et al. [37]. Madau & Maiolino [87] show that the LRD stack by Delvecchio et al. [28], as well as the spectrum and photometry of The Cliff LRD [27], can be modelled in the framework of the simple AGN unified model (though with a high accretion rate), with a dusty torus covering factor of CT∼0.2C_T\sim0.2.

These various results indicate that LRDs might not be as dust-deficient as initially claimed. More generally, the understanding of dust emission in LRDs is far from settled, and this can have profound implications for the scenarios invoked to explain their nature.

In this paper, we explore more thoroughly the dust emission of LRDs. We start by re-assessing the nuclear mid-infrared emission of local type-1 AGN. While this was already explored in the past based on ground-based and low resolutions space facilities, JWST now offers detailed mid-infrared information on the dust and nebular emission of local AGN by isolating the nuclear component from circumnuclear emission associated with stars and star formation. Additionally, we derive the scaling between the infrared and optical emission in type-1 AGN. This provides a more careful comparison to infer the amount of dust emission expected in AGN at high-zz, especially with luminosities not as extreme as those of powerful quasars. We use these local templates to revisit the dust emission in distant and local LRDs. As we will illustrate, this analysis indicates that LRDs are not dust deficient, although their torus properties differ from those of local AGN. In the following, we assume a cosmology with H0=70H_0=70 km s−1^{-1} Mpc−1^{-1}, Ωm=0.3\Omega_m=0.3, and ΩΛ=0.7\Omega_\Lambda=0.7.

Targets and JWST observations

We selected the eight Seyfert 1 (Sy1) galaxies that had publicly available spectra from the JWST instruments MIRI and NIRSpec by the end of 2025. Most of the archival NIRSpec observations of nearby galaxies only cover the ∼\sim3-5 μ\mum range because it includes features such as the 3.3 and 3.4 μ\mum polycyclic aromatic hydrocarbon bands (PAHs) and because it is difficult to observe it with ground-based observatories due to atmospheric transmission. For the sake of homogeneity, we downloaded NIRSpec and MIRI integral field unit observations covering the range ∼\sim3-28 μ\mum. This includes the whole wavelength range covered by MIRI/MRS (∼\sim5-28 μ\mum; R∼R\sim1300-3700; σinst≈35\sigma_{\rm inst}\approx35-100 km s−1^{-1}; Labiano et al. 2021; Argiroffi et al. 2023) and the 2.87-5.27 μ\mum range provided by NIRSpec using the grating and filter combination G395H/F290LP (R∼R\sim2700; σinst≈50\sigma_{\rm inst}\approx50 km s−1^{-1}; Jakobs(en) et al. 2022; Böker et al. 2023). Most of the data come from the Galaxy Activity, Torus, and Outflow Survey (GATOS) JWST General Observer (GO) programs 3535 (PI: I. García-Bernete), 4972 (PI: L. Zhang), and 7195 (PI: A. Alonso-Herrero), except in the case of NGC 4395 (GO program 2016; PI: A. Seth) and NGC 7469 (Early Release Science Program 1328; PI: L. Armus).

The data were reduced using the same version of the JWST Science Calibration Pipeline as in García-Bernete et al. [38] and Pereira-Santaella et al. [102]. Additional correction steps to correct hot and cold pixels were done, outlined in García-Bernete et al. [39] for the case of MIRI/MRS data and in Pereira-Santaella et al. [103] for NIRSpec.

The MIRI nuclear spectra were extracted as point source from all the MRS sub-channels, using aperture sizes ranging from 0.3 to 0.8″ with increasing wavelength. The spectra of the 12 bands were then stitched together using small scaling factors to make them coincide in the overlapping regions [see 108 for further details]. We then stitched the NIRSpec nuclear spectra, extracted in an aperture size of 0.1″, to the nuclear MIRI spectra, taking advantage of their overlapping region at ∼\sim5 μ\mum. The multiplicative factors that we applied to the NIRSpec spectra range from 0.94 to 1.07. The resulting nuclear infrared spectra of the eight Sy1 galaxies are shown in Figure 1.

Figure 1. Left panel: Infrared nuclear spectra of the eight Seyfert 1 (Sy1) galaxies, obtained from NIRSpec and MIRI observations. Right panel: Individual Sy1 spectra normalized at 15 μ\mum (grey lines), resulting mean stack (blue line), and standard deviation (light blue area).

The basic properties of the sample, including line-of-sight column density (NHN_{\rm H}), X-ray and bolometric luminosity, BH mass, and Eddington ratio are shown in Table 1. They have

Notes. Galaxy inclination (ii) are from NED outer diameters and geometrical torus covering factor (fcf_{\mathrm{c}}) derived by Ramos Almeida et al. [109] and García-Bernete et al. [37] from infrared SED fitting using CLUMPY torus models (Nenkova et al. [99]). The column density and intrinsic 2-10 keV luminosities are from Ricci et al. [112] and MBHM_{\mathrm{BH}} from the Hα\alpha measurements reported in Mejía-Restrepo et al. [97], assuming a virial factor of f=1f=1. The only exceptions are NGC 1097, whose values come from Cisternas et al. [23] and Lewis & Eracleous [81] respectively, and NGC 7314, whose BH mass is from Schulz et al. [120]. LbolL_{\mathrm{bol}} is calculated as 20×L2−10 keV20 \times L_{2-10\ \mathrm{keV}} except for NGC 1097, for which we used 10×L2−10 keV10 \times L_{2-10\ \mathrm{keV}} (Ho [57]), and LEdd=1.26×1038×MBHL_{\mathrm{Edd}}=1.26 \times10^{38} \times M_{\mathrm{BH}}. * Usually classified as Seyfert 1.8/1.9 but broad Hα\alpha and Hβ\beta components reported by da Silva et al. [25] and broad Pfund lines detected in the NIRSpec spectra presented here.

References. (a) Tully [135]; (b) Tonry et al. [133]; (c) Yuan et al. [138]; (d) Thim et al. [132]; (e) Bentz et al. [11]; (f) Koshida et al. [77]. § With adjustments from Blakeslee et al. (2001).

luminosity distances below 60 Mpc and relatively low galaxy inclinations (see Table 1) that allow us to isolate the central, AGN-dominated regions, minimizing the host galaxy contribution. Considering the angular resolution of NIRSpec (∼0.1′′\sim0.1^{\prime\prime} at 3 μ\mum) and MIRI (∼0.3\sim0.3-0.8′′0.8^{\prime\prime} across the whole range of 5-28 μ\mum), the nuclear spectra probe the central ∼2\sim2-17 pc of the closest target, NGC 4395, and ∼28\sim28-225 pc for the most distant one, NGC 7469 (see Table 1).

Results

NGC 1097 and NGC 4395 are the dimmest nuclei in the sample, and NGC 7469 is the brightest (see Figure 1). The spectra show tens of emission lines and several PAH bands in the case of galaxies with nuclear star formation such as NGC 1097 and NGC 7469. The 9.7 μ\mum silicate feature varies from weak absorption (NGC 7314) to strong emission (NGC 7213), although in the majority of the galaxies is either in weak emission or flat. This was already seen from ground-based mid-infrared spectra of Sy1 galaxies (Alonso-Herrero et al. [3]), and it is consistent with clumpy (i.e. non-smooth) and two-phase torus model predictions (Ramos Almeida et al. [110, 109]; Hönig et al. [60]; Alonso-Herrero et al. [4]; García-Bernete et al. [37, 38]; González-Martín et al. [44, 46]).

All the targets were previously observed with the InfraRed Spectrograph (IRS) of the Spitzer Space Telescope, whose spectra cover a much larger fraction of the galaxies than JWST (the central ∼3.6\sim3.6-10.7″ in the case of IRS versus ∼0.1\sim0.1-0.8″ for NIRSpec+MIRI). In Fig. A1 we show a comparison between the JWST and Spitzer spectra of the eight Sy1 galaxies. From this comparison we can see that the targets with AGN-dominated infrared spectra, which have little or no PAH emission (NGC 4395, NGC 6814, and NGC 7213), show similar JWST and Spitzer spectra. On the other hand, NGC 1097, NGC 3227, NGC 4051, and NGC 7469 show strong PAH bands in their Spitzer/IRS spectra that are either not present or much weaker in the JWST spectra. Finally, the Spitzer/IRS spectrum of NGC 7314 shows deeper silicate absorption than the JWST/MIRI spectrum, indicating a host-galaxy contribution of cold dust that the JWST spectrum does not include. This qualitative comparison shows that even in the case of Sy1 galaxies, the nuclear JWST/NIRSpec+MIRI spectra is significantly different from those observed with Spitzer/IRS, except in the case of targets with AGN-dominated infrared spectra.

JWST average stacked spectrum of Seyfert 1 galaxies

To produce the Sy1 stack, we first interpolated the rest-frame individual spectra shown in the left panel of Figure 1 to a common wavelength grid. We adopted the wavelength sampling of the closest target, NGC 4395, which sets the starting wavelength of the new grid (x1=2.87 μmx_1=2.87\ \mu\mathrm{m}) up to an end wavelength of x2=28.2 μmx_2=28.2\ \mu\mathrm{m}, defined by the most distant galaxy, NGC 7469.

Prior to stacking, the regrid spectra were normalized to their corresponding mean flux density within a narrow window centered at 15 μm15\ \mu\mathrm{m} (14.9814.98–15.01 μm15.01\ \mu\mathrm{m}). These flux densities at 15 μm15\ \mu\mathrm{m} range from 17 to 891 mJy, depending on the target, with a mean value of 348±317348 \pm317 mJy. The normalization ensures that all spectra contribute equally at the reference wavelength while preserving relative differences in spectral shape (see right panel of Figure 1). We chose to normalize at 15 μm15\ \mu\mathrm{m} because it provides a relatively clean measure of the mid-infrared continuum while minimizing contamination from strong spectral features and variations in spectral slope. At shorter wavelengths (e.g. ∼12 μm\sim12\ \mu\mathrm{m}), the continuum can be affected by the wings of the 9.7 μm9.7\ \mu\mathrm{m} silicate feature and/or 11.3 μm11.3\ \mu\mathrm{m} PAH band, introducing object-dependent biases when scaling the spectra. At longer wavelengths (e.g. ∼20 μm\sim20\ \mu\mathrm{m}), the emission becomes increasingly sensitive to variations in the warm-to-cool dust ratio, leading to a larger dispersion in spectral slopes among sources. In contrast, the region around 15 μm15\ \mu\mathrm{m} is featureless and retains high signal-to-noise, providing a more stable and representative anchor for the intrinsic AGN continuum. In any case, we verified that the variations introduced in the stack by normalizing at 12 and 20 μm20\ \mu\mathrm{m} are smaller than the standard deviation of the stack normalized at 15 μm15\ \mu\mathrm{m} (see the right panel of Figure 1).

The resulting set of uniformly sampled and normalized spectra was then used to construct the Sy1 stack, computed as the mean of the individual spectra at each wavelength. The associated dispersion was quantified via the standard deviation. This procedure yields an average spectrum, shown in the right panel of Figure 1, while providing a measure of the object-to-object variability. The stack is representative of the eight Sy1 galaxies here considered, but not necessarily of the local Sy1 population. Nevertheless, these Sy1s cover a relatively broad range of Eddington ratios, BH masses, X-ray and bolometric luminosities (see Table 1). Neither the stack nor the individual spectra were corrected from extinction. For Sy1 galaxies and intermediate types, typical AVA_V values measured from ratios of Balmer and Paschen narrow emission lines and infrared-to-optical colors are ≤1\leq1–2 mag (Landt et al. [80]; Schnorr-Müller et al. [118]; Prieto et al. [106]). This would correspond to an A15μm∼0.02A_{15\mu\mathrm{m}} \sim0.02–0.04 mag using the extinction-law of Chiar & Tielens [22].

Torus models

To facilitate the comparison with different datasets including bluer and redder wavelengths than those covered by our Sy1 stack, we fitted it with BayesClumpy2 (Ramos Almeida et al. [107]). This new version of the Bayesian tool BayesClumpy (Asensio Ramos & Ramos Almeida [7, 8]) allows the user to fit photometry or spectroscopy data with the CLUMPY torus models of Nenkova et al. [99] and the CAT3D-WIND models of Hönig & Kishimoto [59]. We used an emission-line free, smoothed, and interpolated version of the Sy1 stack. We smoothed the spectrum using a Lee filter algorithm with a box size of 100, and then interpolated it to a grid of 120 wavelengths that successfully captures the spectral shape, including the silicate features (see blue solid line in Figure 2 and B.1).

Figure 2. Median CAT3D-wind torus model (black dot-dashed line) fitted to the Sy1 stack shown in Figure 1 (blue solid line). The four-component power-law used as input spectrum to the CAT3D-wind models is the red dashed line. The orange lines are the sum of the median torus model and the accretion disc, scaling the disc to have energetic covering factors from CT=0.7C_{\mathrm{T}}=0.7 (solid line) to CT=0.3C_{\mathrm{T}}=0.3, 0.2, 0.1, 0.09, 0.08, and 0.07 (dot-dashed lines). The blue dot is the average flux at 5100 Å of the eight Sy1s, estimated from the FHα,5100F_{\mathrm{H}\alpha,5100} values reported in Table 3, normalized by the average flux at 15 μm15\ \mu\mathrm{m} (Fν‾=347.5\overline{F_\nu}=347.5 mJy).

We ran the fits using nested sampling, uniform priors for all the parameters, and no foreground extinction. As flux error, we used the standard deviation of the stack (shaded area in the right panel of Figure 1) plus 10% of calibration uncertainty (Ramos Almeida et al. [107]), summed in quadrature.

Figure B.1 shows the fitted CLUMPY and CAT3D-wind models (those between the 16 and 84 percentiles of the posterior distribution) in orange. The black lines are the median model and corresponding residuals, which in the case of the CAT3D-wind model are smaller than 15% across the whole wavelength range considered in the fit. The CAT3D-wind models better reproduce the shape of the mean Sy1 stack than the CLUMPY torus models, which produce larger residuals in the near-infrared (up to ∼50%\sim50\% in the case of the median model). However, it is noteworthy that the residuals from the fit with the CLUMPY model at λ>20 μm\lambda>20\ \mu\mathrm{m} are smaller than those from the fit with the CAT3D-wind model. Previous work on SED fitting of nearby AGN reported evidence for torus models including polar emission better accounting for the infrared emission of the majority of type-1 AGN than those without (González-Martín et al. [45, 43]; García-Bernete et al. [38]).

Since we are working with type-1 AGN, we added the AGN input spectrum of the CAT3D-wind models (the four-component power-law described in Appendix A of Hönig & Kishimoto [58]), shown as a red dashed line in Figure 2, to the median torus model (black dot-dashed line in Figure 2). The resulting model SED, which accounts for both torus and accretion disc direct emission, is the orange solid line in Figure 2. If we use Eq. 2 in Maliino et al. (2007) to derive the energetic covering factor (CTC_{\mathrm{T}}), but considering a factor of 9 instead of 7 to convert from LbolL_{\mathrm{bol}} to L5100L_{5100} (Stern & Laor [128]),

CT=Ldust/Lbol≈0.3⋅λLλ(6.7 μm)/λLλ(5100 A˚)(1)C_{\mathrm{T}} = L_{\mathrm{dust}}/L_{\mathrm{bol}} \approx0.3 \cdot\lambda L_{\lambda}(6.7\ \mu\mathrm{m})/\lambda L_{\lambda}(5100\ \text{Å}) \tag*{(1)}

Notes. The second column is the FWHM of the Hα\alpha broad component from Mejía-Restrepo et al. [97], except in the case of NGC 1097, that is from Schimoia et al. [116]. The mean, minimum, and maximum broad line fluxes were calculated from the values reported in the works indicated in column seven. Column five lists the variation in flux from minimum to maximum. Columns six and seven list the broad line Hα\alpha luminosities calculated from the fluxes reported in columns 3 and 4 and the distances from Table 1. Columns 8 and 9 indicate the Balmer decrement measured from the broad Hα\alpha and Hβ\beta components and corresponding AVA_{\mathrm{V}} assuming and intrinsic Hα\alpha/Hβ\beta=3.06 (Dong et al. [31]) and adopting the extinction law of Cardelli et al. [17] with RVR_{\mathrm{V}}=3.1. The AVbA_{\mathrm{V}}^{b} values of NGC 6814 and NGC 7213 were taken from Schnorr-Müller et al. [118] and Caglar et al. [16], where several line ratios were used to compute the extinction towards the BLR. *NGC 7213 also has broad double-peaked lines with a velocity separation of ∼9000 km s−1^{-1}, whose fluxes, from Schimoia et al. [117], are included in the broad Hα\alpha flux.

References. (a) Schimoia et al. [116]; (b) Schimoia et al. [115]; (c) Landt et al. [80]; (d) Mejía-Restrepo et al. [97]; (e) Balmaverde & Capetti [10]; (f) Gu et al. (2024); (g) Schimoia et al. [117]; (h) da Silva et al. [25]; (i) Storchi-Bergmann et al. [129]; (j) Devereux [30]; (k) Schnorr-Müller et al. [118]; (l) Caglar et al. [16].

we get CTC_{\mathrm{T}} = 0.7. This energetic covering factor is often referred to as reprocessing efficiency (RE) of the torus or reprocessed fraction of the accretion disc emission (Ramos Almeida et al. [109]; Stalevski et al. [127]), and it shall not be mistaken with the geometrical covering factor derived from the torus model parameters (fCf_{\mathrm{C}} or f2f_{2}; see Ramos Almeida et al. [111]).

Figure 2 also shows the model SEDs resulting from scaling up the accretion disc emission to get CTC_{\mathrm{T}}=0.3, 0.2, 0.1, 0.09, 0.08, and 0.07 from Eq. (1). These model SEDs are used in § 4.1. Although the accretion disc and the torus SEDs should not be scaled independently if we seek for consistency between them, it can be done as an exercise to simulate AGN with different properties than those in the CAT3D-wind database, which were developed to reproduce local AGN. In any case, we checked that there are torus+disc models in the BayesClumpy2 database having the covering factors used here.

The model SED corresponding to CTC_{\mathrm{T}}=0.3 (orange solid line in Figure 2) matches the average 5100 Å flux of the Sy1 galaxies (blue dot in Figure 2). This average 5100 Å flux and its standard deviation was calculated from the λFHαavg\lambda F_{\mathrm{H}\alpha}^{\mathrm{avg}} fluxes reported in Table 3 and then divided by the average flux of the Sy1 stack at 15 μm, the wavelength of normalization (FνF_{\nu}=347.5 mJy).

IDλF5100Hα,avg×1012\lambda F^{\mathrm{H}\alpha,\mathrm{avg}}_{5100} \times10^{12} (erg s−1^{-1} cm−2^{-2})log⁡λL5100Hα,avg\log\lambda L^{\mathrm{H}\alpha,\mathrm{avg}}_{5100} (erg s−1^{-1})log⁡λL5100Hα\log\lambda L^{\mathrm{H}\alpha}_{5100} (erg s−1^{-1})λF5100×1012\lambda F_{5100} \times10^{12} (erg s−1^{-1} cm−2^{-2})log⁡λL5100\log\lambda L_{5100} (erg s−1^{-1})λF6200×1012\lambda F_{6200} \times10^{12} (erg s−1^{-1} cm−2^{-2})log⁡λL6200\log\lambda L_{6200} (erg s−1^{-1})
NGC10975.59741.15[40.69, 41.49]............
NGC322759.3042.60[42.48, 42.69]63.1042.6348.9842.52
NGC405143.6842.16[42.14, 42.18]53.7042.2585.1142.45
NGC43958.85540.29[40.29, 40.30]2.75439.783.09039.83
NGC681462.5542.54[42.39, 42.70]19.5042.0432.3642.26
NGC721346.7342.43[42.39, 42.51]8.31841.6885.1142.69
NGC731425.0442.00[41.98, 42.02]......1.12240.65
NGC746971.8943.46[43.40, 43.51]107.243.63104.743.62

Table 3. Continuum properties derived from Hα\alpha broad fluxes and from single-epoch optical spectra.

Connecting the BLR and optical continuum emission of Sy1 galaxies to their infrared spectra

Here we attempt to assess whether the rest-frame infrared emission of LRDs is consistent with predictions derived from our local Sy1 galaxies. In the literature, expectations on the infrared emission are set by the chosen normalization assumed from the optical. The issue with using the rest-frame optical continuum to predict the infrared at high-zz is the uncertainty on whether that continuum is dominated either by the AGN or by the stellar component. For this reason, in principle it would be better to use the broad Hα\alpha luminosity (LHαbL_{\mathrm{H}\alpha}^{b}) as normalizing quantity. The caveat then is that the large EWs measured for Hα\alpha in distant AGN (see Greene et al. [51] and references therein) could imply that the covering factor of the BLR is higher than in local AGN (Maiolino et al. 2025; Yanagisawa et al. [137]) and/or that the Hα\alpha emission is boosted by processes such as collisional excitation (Brazzini et al. [15]). As a consequence, our local conversion from LHαbL_{\mathrm{H}\alpha}^{b} to infrared luminosities might need to be re-scaled by a certain factor (∼3-10; Brazzini et al. [15]).

An extra source of complication affecting optical normalizations for local type-1 AGN, both using the continuum or emission line luminosities, is variability. In the case of Sy1 galaxies, this variability usually ranges from weeks to months and it can be significant (Peterson et al. [105]; Schimoia et al. [116]; Tao et al. 2015). To account for this, we compiled broad Hα\alpha fluxes for the eight Sy1 galaxies from the literature and calculated average values and ranges, which we report in Table 2. In some cases, these flux measurements span several years (e.g., NGC 3227 and NGC 4051). All the galaxies but NGC 1097 have broad Hα\alpha fluxes reported by Mejía-Restrepo et al. [97], together with continuum measurements at 5100 and 6200 Å (see Table 3). According to the broad Hα\alpha fluxes shown in Table 2, NGC 1097 is the most variable source, with a 8.5 variation between maximum and minimum flux. The rest of the sources show variations ranging from 1.03 to 1.81. For NGC 6814 we found only one measurement of the broad Hα\alpha flux, and thus we allowed a 1.3 factor of variation, based on the average of all the Sy1 galaxies except the extreme case of NGC 1097. The range of broad Hα\alpha luminosities that we are probing with our Sy1 galaxies, uncorrected from extinction, is 38.4–42.2 erg s−1^{-1}, with a mean value of 40.4±1.2 erg s−1^{-1}.

IDFWHM (km s−1^{-1})FHαavg×1013F_{\mathrm{H}\alpha}^{\mathrm{avg}} \times10^{13} (erg s−1^{-1} cm−2^{-2})FHα×1013F_{\mathrm{H}\alpha} \times10^{13} (erg s−1^{-1} cm−2^{-2})FHαmax⁡/FHαmin⁡F_{\mathrm{H}\alpha}^{\max}/F_{\mathrm{H}\alpha}^{\min}log⁡LHαavg\log L_{\mathrm{H}\alpha}^{\mathrm{avg}} (erg s−1^{-1})log⁡LHα\log L_{\mathrm{H}\alpha} (erg s−1^{-1})Refs.Hαb\alpha^{b}/Hβb\beta^{b}AVbA_{\mathrm{V}}^{b} (mag)Refs.
NGC1097100001.045[0.306, 2.585]8.4539.42[38.89, 39.81]a,b2.8±0.10.0i
NGC3227290018.74[13.49, 24.10]1.7941.10[40.96, 41.21]c,d,e3.1±0.70.0c,j
NGC4051200011.76[11.22, 12.30]1.1040.59[40.57, 40.61]d,e<2.80.0d
NGC439510001.212[1.193, 1.230]1.0338.43[38.42, 38.43]d,f<2.80.0d
NGC6814330014.45. . .. . .40.91[40.85, 40.97]d. . .0.4±0.4k
NGC72132300*13.90[12.39, 17.23]1.3940.91[40.86, 41.00]g. . .0.0±0.3l
NGC731413000.706[0.670, 0.741]1.1139.45[39.43, 39.47]d,h9.3±0.63.0±0.2h
NGC7469270031.04[26.60, 35.48]1.3342.10[42.03, 42.15]c,d<2.80.0c

Table 2. Properties of Hα\alpha broad compiled from the literature.

In Table 2 we report the extinction measured from broad recombination line ratios (AVbA_{\mathrm{V}}^{b}) reported in the literature. For all of the galaxies in the sample, AVbA_{\mathrm{V}}^{b}<0.5 mag, in good agreement with the median values reported for BASS broad-line AGN (Mejía-Restrepo et al. [97]), of 0.2 mag for galaxies with log⁡NH≤20\log N_{\mathrm{H}} \leq20 cm−2^{-2} and 0.6 mag for those with log⁡NH>20\log N_{\mathrm{H}} > 20 cm−2^{-2}. The exception is NGC 7314, for which da Silva et al. [25] reported AVbA_{\mathrm{V}}^{b} values of 3.0 mag from the Balmer decrement of the broad components. This is related with the Seyfert 1.8/1.9 classificaNotes. Columns 2-4 are the continuum flux, luminosity, and luminosity range at 5100 Å derived from the extinction-corrected broad Hα\alpha fluxes reported in Table 2 and using Eq. (1) in Greene & Ho [49]. Columns 5-8 are the single-epoch continuum fluxes and luminosities at 5100 and 6200 Å from Mejía-Restrepo et al. [97]. Luminosities were calculated using the distances reported in Table 1.

tion of the galaxy, which has changed to type 1/1.5 depending on when the observations were taken (see da Silva et al. [25] and references therein). Therefore, we corrected its broad Hα\alpha flux from extinction using the extinction law of Cardelli et al. [17] and RV=3.1R_{\mathrm{V}}=3.1. We did the same for NGC 6814, for which Schnorr-Müller et al. [118] reported an AVb=0.4±0.4A_{\mathrm{V}}^{\mathrm{b}}=0.4\pm0.4 mag. Using the extinction-corrected LHαavgL_{\mathrm{H}\alpha}^{\mathrm{avg}} values we get the same range of broad Hα\alpha luminosities, and a mean value of 40.5±1.140.5\pm1.1 erg s−1^{-1}.

We then normalized the JWST spectra of the individual Sy1 galaxies, shown in Figure 1, by their corresponding broad Hα\alpha luminosities (using LHαavgL_{\mathrm{H}\alpha}^{\mathrm{avg}} from Table 2, corrected from AVbA_{\mathrm{V}}^{\mathrm{b}}). The resulting spectra are shown in the left panel of Figure 3 and C.1. In logarithmic scale, the normalized spectra vary between ∼1.1\sim1.1 and 2, with NGC 4051 showing the highest values and NGC 6814 the lowest. We also used the continuum luminosities at 5100 Å to normalize the spectra, which are shown in Figure 3 and C.1. These continuum luminosities were calculated from LHαavgL_{\mathrm{H}\alpha}^{\mathrm{avg}}, corrected from extinction in the cases of NGC 6814 and NGC 7314, using Eq. (1) in Greene & Ho [49]. They are reported in Table 3 and they range between 40.3 and 43.5 erg s−1^{-1} (mean value of 42.1±1.042.1\pm1.0 erg s−1^{-1}). From the right panel of Figure 3 and C.1 it can be seen that the normalized spectra range from -0.7 to 0.6 in logarithmic scale, with NGC 7469 showing the highest values and NGC 4395 the lowest. The difference between the two normalizations shown in Figure 3 is of ∼1.6\sim1.6 dex (λL5100≈40×LHα\lambda L_{5100}\approx40\times L_{\mathrm{H}\alpha}), which comes from Eq. (1) in Greene & Ho [49].

Figure 3. Rest-frame spectra of the Sy1 galaxies (dark grey lines), normalized to their corresponding extinction-corrected broad Hα\alpha luminosities (left panel) and 5100 Å luminosities derived from them (right panel). Their corresponding mean stacks are shown in pink and green, respectively, with the grey shaded region indicating the standard deviation around the mean. Squares correspond to the average fluxes of the stacked spectrum measured in the wide MIRI filters (orange) and in the two NIRCAM wide filters F356W and F444W (yellow).

We note that although the conversion from LHαL_{\mathrm{H}\alpha} to L5100L_{5100} might not be adequate for high-redshift AGN due to high Hα\alpha EWs (Greene et al. [51]), we can safely use it for our local Sy1s. Indeed, for all the galaxies but NGC 1097 and NGC 7314 there are single-epoch 5100 Å continuum luminosities reported by Mejía-Restrepo et al. [97], shown in Table 3, and they are in reasonable good agreement with the LHαavgL_{\mathrm{H}\alpha}^{\mathrm{avg}}-derived measurements (mean scatter of ∼0.3\sim0.3 dex). The largest discrepancy is found for NGC 7213 (0.75 dex). We prefer to use log⁡L5100\log L_{5100} instead of the single-epoch values because they account for optical variability, which in the case of some of the targets is important (see Table 2), and there are measurements for all the targets (see Table 3).

Using the normalized spectra shown in the left and right panels of Figure 3 and C.1, we produced the mean stacks shown in Figure 3, using the same method as in Section 3.1. In the case of the L5100L_{5100} normalization, NGC 4395 is below the standard deviation of the mean stack, but including it or not does not make any significant difference to the latter. The mean normalized stacks can be easily multiplied by the Hα\alpha and 5100 Å continuum luminosities measured for high-redshift AGN in order to predict their rest-frame ∼3\sim3-27 μ\mum emission. In Table C.1 we also report average values measured in the JWST MIRI and NIRCAM wide filters from the normalized stacks shown in Figure 3. In the case of the broad Hα\alpha normalization, λLλ/LHαcorr≃[30,60]\lambda L_{\lambda}/L_{\mathrm{H}\alpha}^{\mathrm{corr}} \simeq[30,60], from 3 to 27 μ\mum, and in the case of the 5100 Å normalization, λLλ/λL5100≃[0.8,1.8]\lambda L_{\lambda}/\lambda L_{5100} \simeq[0.8,1.8].

The bolometric luminosities of high-redshift AGN, including LRDs (log LbolL_{\mathrm{bol}} (erg s−1^{-1}) ∼44.1\sim44.1-44.4 for the LRD stack of Pérez-González et al. [104] and see also Maiolino et al. [91]; Geris et al. [41]; Scholtz et al. [119]) are not that different anymore from those of local Sy1 galaxies. The normalizations provided by NGC 7469, the most luminous Sy1 in our sample (log LbolL_{\mathrm{bol}} (erg s−1^{-1}) ∼44.5\sim44.5), are fully consistent with those of the stacks shown in Figure 3. Therefore, despite the relatively low luminosities of the Sy1 galaxies used here, of log LbolL_{\mathrm{bol}} (erg s−1^{-1}) =41.8-44.5, they can be used to predict the infrared emission of LRDs with luminosities of at least up to log LbolL_{\mathrm{bol}} (erg s−1^{-1}) ∼\sim44.5.

Predicting the rest-frame infrared emission of LRDs from Hα\alpha and optical continuum measurements.

We now use the Sy1 stacks shown in Figure 3 to predict the infrared emission of the z>2z>2 LRD stack from Pérez-González et al. [104] and the SED of the local LRD known as LORD/The Egg at z=0.1z=0.1 (Ji et al. 2026; Lin et al. [84]) from their broad Hα\alpha and 5100 Å luminosities.

The infrared emission of LRDs at 2.3<z<9.3

Pérez-González et al. [104] stacked NIRSpec spectra of 249 LRDs at 2.3<z<9.32.3<z<9.3 obtained with the low resolution, R∼30R\sim30-330, double-pass prism (PRISM) and MIRI photometry of between 5 and 17 LRDs, depending on the filter (9, 12, 8, 10, and 5 LRDs for the five MIRI points at λrest≳2 μ\lambda_{\mathrm{rest}} \gtrsim2\ \mum). They reported a flattening of the MIRI SED at λrest<2 μ\lambda_{\mathrm{rest}}<2\ \mum, in agreement with the NIRSpec Prism continuum, and a rising continuum at λrest≳2 μ\lambda_{\mathrm{rest}} \gtrsim2\ \mum, with a peak at 2-3 μ\mum (see Figure 4). Here we consider the MIRI stack at 4.7 μ\mum rest-frame as an upper limit, as it includes only five infrared-bright LRDs. Pérez-González et al. [104] claim that the four MIRI stacks with rest-frame wavelengths between 2 and 3.8 μ\mum are consistent with the presence of hot dust at ∼\sim1500 K. From this NIRSpec LRD stack we measured a rest-frame 5100 Å flux of 5.70×10−175.70 \times10^{-17} erg s−1^{-1} cm−2^{-2} (see Table D.1) in a narrow window between 5080 and 5120 Å. We used two Gaussian components of FWHM=3400 (broad) and 1050 km s−1^{-1} (intermediate) to reproduce the Hα\alpha profile (see left panel of Fig. D.1 and Table D.1), however, for our goals, the detailed functional form for fitting the Hα\alpha profile is not critical.

Figure 4. Normalized Sy1 stacks from Figure 3, shown as pink (purple) and green lines, multiplied by the broad (total) Hα\alpha and 5100 Å fluxes measured from the LRD stack of Pérez-González et al. [104]. The gray shaded area corresponds to the standard deviation around the average for the broad Hα\alpha normalization (left panel) and for the extinction-corrected 5100 Å normalization (right panel). The lighter purple line is the total Hα\alpha normalization divided by three. The brown solid line in the right panel is the normalized Sy1 stack, but multiplied by the 5100 Å flux corrected by an extinction of AV_V=0.7 mag to match the MIRI 3-4 μ\mum LRD stacks. The NIRSpec Prism LRD stacked spectrum from Pérez-González et al. [104], in solid grey, and MIRI stacks (blue diamonds), in λFλ\lambda F_\lambda units, are shown in both panels.

In the left panel of Figure 4 we show the Sy1 stack from the left panel of Figure 3, multiplied by the broad and total (broad+intermediate) Hα\alpha flux of the LRD stack from Pérez-González et al. [104]. The prediction from broad Hα\alpha (pink line and shaded grey area in the left panel of Figure 4) fully agrees with the ∼\sim3-4 μ\mum MIRI LRD stacks. It is noteworthy that the Hα\alpha fluxes that we measure from the NIRSpec LRD stack do not include any extinction correction. If we correct the broad Hα\alpha flux by different extinctions (e.g., the ones considered in Pérez-González et al. [104], of AVA_{\mathrm{V}}=0.4, 0.7, and 1.6 mag), the predictions on the infrared emission of LRDs would be a factor 1.3-3.2 higher (see left panel of Fig. E.1). Using the flux of Hα\alpha/Hβ\beta from Table D.1 and the Small Magellanic Cloud (SMC) bar extinction curve of Gordon et al. [48], we get AVA_{\mathrm{V}}=1.1±\pm0.4 mag, which would correspond to an infrared emission a factor of 2.2 higher than with an AVA_{\mathrm{V}}=0 mag. We note that there is debate on the interpretation of the Balmer decrement of the broad components of LRDs, with models ranging from those entirely invoking density and radiative effects within the BLR (or surrounding cocoon) to those claiming significant dust reddening effects (e.g. D’Eugenio et al. [29]; Nikopoulos et al. [101]; Kiyota et al. [73]; Madau & Maiolino [87]; Madau et al. [88]). Here we do not enter into this debate and note that the inferred reddening is similar to the range reported by Pérez-González et al. [104]. We will discuss more thoroughly the role of dust extinction in § 4.1.

In the left panel of Figure 4 we also show the Sy1 stack, but multiplied by the total (broad + intermediate) Hα\alpha flux from Table D.1 (purple line), which is a factor two higher than the broad Hα\alpha flux. Given that the LRD stack was built from low resolution PRISM spectra, it is not clear whether the intermediate component traces a narrow (NLR/SF) component or it also contains part of the BLR emission. In the extreme case in which the total flux traces the full BLR emission, it would over-predict their infrared emission by a factor of ∼\sim2. This would be consistent with a higher covering factor of the BLR in these systems (Maiolino et al. 2025; Brazzini et al. [14]; Yanagisawa et al. [137]). However, in the latter scenario of BLR with high covering factor in LRDs, the Hα\alpha scaling given by the Sy1 stack should be corrected by about a factor of 3 (Brazzini et al. [15]). This is shown by the light purple line in the left panel of Figure 4, which is also consistent with the observed data.

In the right panel of Figure 4 we show the normalized Sy1 stack from the right panel of Figure 3, multiplied by the 5100 Å flux measured from the LRD stack (see Table D.1). The predicted infrared emission, shown as a green solid line, is below the MIRI LRD stacks, indicating that the 5100 Å flux might not be as good as a good predicting quantity as the broad Hα\alpha flux. However, considering that Hα\alpha (and particularly, Hα\alpha broad) is less affected by extinction than the 5100 Å flux, we repeated the exercise of correcting the continuum flux by different extinction values. The results are shown in the right panel of Fig. E.1. In the case of the 5100 Å flux, the predictions on the infrared emission of LRDs would be a factor 1.5-5 higher if we consider AVA_{\mathrm{V}} values of 0.4, 0.7, and 1.6 mag. As can be seen from the right panel of Figure 4 and E.1, an extinction-correction of AV∼0.7A_{\mathrm{V}} \sim0.7 mag (a factor 2 higher in flux) nicely matches the MIRI 3-4 μ\mum LRD stacks.

The infrared emission of an LRD at z∼0.1

In the case of the z≳4z\gtrsim4 LRD stacks, the longest wavelengths for which there are detections are ∼\sim3-4 μ\mum. For this reason, in Figure 5 we also show a comparison between the Sy1 stacks and the photometric SED of the local LRD (z=0.1z=0.1) known as “LORD” (Ji et al. 2026) or “The Egg” (Lin et al. [84]). This SED includes WISE fluxes from 3.4 to 22 μ\mum that can be directly compared with the mid-infrared continuum of the Sy1 stack. Here we consider the broad (BB in Ji et al. 2026) and total (NB+BB, but excluding the narrow component of 90 km s−1^{-1}) Hα\alpha fluxes measured from the Gemini spectrum and reported in Table 1 from Ji et al. (2026), and a 5100 Å flux of 2.722×10−132.722 \times10^{-13} erg s−1^{-1} cm−2^{-2}, measured from the same spectrum.

Normalized Sy1 stacks and LORD SED

Figure 5. Same as in Figure 4, but for the local LRD (z=0.1) LORD (Ji et al. 2026; Lin et al. [84]). The brown solid line in the right panel is the normalized Sy1 stack, but multiplied by the 5100 Å flux of LORD corrected by the extinction of AV_V=0.4 mag reported by Ji et al. (2026) for the ISM/NLR of LORD. The LORD SED (yellow triangles) is shown in both panels.

In the left panel of Figure 5 we show the Sy1 stack from the left panel of Figure 3, multiplied by the broad and total Hα\alpha fluxes of LORD. The broad and total Hα\alpha fluxes overestimate the WISE W4 flux (22 μ\mum) by factors of 1.6 and 2.7. The template’s scatter around the broad Hα\alpha normalization is consistent with the WISE W3 and W4 fluxes, although the W3 flux could be boosted by emission spectral features, such as the 9.7 μ\mum silicate feature, the 11.3 and 12.5 μ\mum PAH features, and/or emission lines such as [SIV]λ\lambda10.51 μ\mum and [NeII]λ\lambda12.81 μ\mum. Upcoming JWST/MIRI observations of LORD (GO proposal 12316; PI: F. Bauer) will reveal its mid-infrared spectrum.

Similarly to the analysis of the high-z stack, in the left panel of Figure 5 we also show the normalization using the total Hα\alpha flux divided by three, which matches the W1, W2, and W4 fluxes. As in the case of the high-z LRD stack of Pérez-González et al. [104], we are not considering extinction correction for the Hα\alpha fluxes in the left panel of Figure 5. Ji et al. (2026) estimated AV_V < 5.1 mag from the Balmer decrement, but they acknowledged that it can be enhanced by collisional excitation (Ilić et al. [61]), resembling dust attenuation.

In the right panel of Figure 5 we show the Sy1 stack from the right panel of Figure 3, multiplied by the 5100 Å flux of LORD (green solid line). In this case, the continuum normalization matches the W1 and W2 fluxes and it underpredicts the W4 flux by a factor of 0.7. However, if we correct the 5100 Å flux from extinction using the value reported by Ji et al. (2026) for the ISM/NLR of LORD, of AV_V = 0.4 mag, the continuum normalization accommodates all the WISE fluxes (brown solid line and gray area in Figure 5).

ISM/NLR of LORD, of AV_V = 0.4 mag, the continuum normalization accommodates all the WISE fluxes (brown solid line and gray area in Figure 5).

Considering the comparisons presented in Sections 3.4.1 and 3.4.2, we find that the infrared emission of LRDs inferred from their Hα\alpha luminosities (uncorrected from extinction) exceeds the actual infrared emission by a factor of ∼\sim1–3, placing our estimates at the lower end of the factor of ∼\sim3–10 predicted by Brazzini et al. [15] for a high covering factor BLR. However, this 1–3 factor would increase if AV_V(BLR) > 0 mag. Overall, the interpretation of the results when normalizing by Hα\alpha are not trivial, as the deviations can be interpreted either in terms of higher covering factor of the BLR, excess production of Hα\alpha due to radiative and density effects in the BLR, or Hα\alpha extinction – the effects are degenerate at this level.

When we use the 5100 Å flux, the predicted infrared emission is consistent with the actual values using a modest amount of extinction, of AV_V ≲\lesssim 1 mag (i.e. within a factor of ∼\sim2). Despite the large galaxy areas probed by the LRD stack of Pérez-González et al. [104] and, to a lesser extent, by the SED of LORD, our comparisons are consistent with an AGN-dominated optical continuum of LRDs. Therefore, here we favor the use of the 51005100 Å continuum as a predicting quantity. We discuss the AGN-origin of the optical emission of LRDs in § 4.

Discussion

In § 4.1 we compare the Sy1 stack and torus model fits shown in Figure 2 with different stacks of LRDs at z≳2z\gtrsim2 (Delvecchio et al. [28]; Akins et al. [2]; Pérez-González et al. [104]). In § 4.2 we briefly discuss the optical-infrared energy balance aspects of our model. Finally, in § 4.3 we estimate the dust mass of the torus model resembling the optical and near-infrared SED of these LRD stacks and we compare it with different estimates published in the literature.

A reddened accretion disc and a low-reprocessing efficiency torus to explain the optical-to-infrared SED of LRDs

In Section 3.4 we showed that the infrared emission of LRDs can be predicted from their 51005100 Å continuum luminosity using the Sy1 stacks shown in Figure 3. This approach has the advantage of being model-independent. However, the accretion disc+torus model that fits both the Sy1 stack and optical point (see Figure 2) does not resemble the optical and infrared SED of LRDs, suggesting that their accretion disc and torus properties differ from those of Sy1s. Therefore we attempt to reproduce the optical-to-infrared continuum emission of high-redshift LRDs by varying the torus covering factor (i.e., its reprocessing efficiency, which is the scaling of the accretion disc relative to the torus model) and the foreground extinction.

We scaled our Sy1 stack and fitted torus model (blue solid line and black dot-dashed line in Figure 2) to the rest-frame 3 μm3\ \mu\mathrm{m} MIRI stack from Pérez-González et al. [104]. We then considered different scalings of the accretion disc component relative to the torus model (see § 3.2 for details), which correspond to covering factors ranging from CT=0.7C_{\mathrm{T}}=0.7 to 0.070.07 (orange lines in Figure 2). To account for additional foreground extinction not included in the CAT3D-WIND models, we applied an extinction law, with AVA_{\mathrm{V}} varying from 0.50.5 to 33 mag, to the Sy1 stack and model SEDs. We adopted the SMC bar extinction curve of Gordon et al. [48], which may be more appropriate for high-redshift LRDs (Madau & Maiolino [87]), but also for normal AGN and quasars at high-zz (Richards et al. [113]). We scaled it to match the Chiar & Tielens [22] local ISM extinction curve at the K-band (1.25 μm1.25\ \mu\mathrm{m}), which we used up to 27 μm27\ \mu\mathrm{m}, and then extrapolated it beyond this wavelength with a power law (Aλ∝λ−2A_{\lambda}\propto\lambda^{-2}).

We find that the torus model with CT=0.08±0.01C_{\mathrm{T}}=0.08\pm0.01 and an additional, foreground extinction of AV=1.9±0.1A_{\mathrm{V}}=1.9\pm0.1 mag, is the one that best captures the shape of the LRD stack for rest-frame λ≳0.3\lambda\gtrsim0.3-0.4 μm0.4\ \mu\mathrm{m}, as illustrated by the orange solid line in Figure 6 (with the orange dot-dashed lines corresponding to the torus models with CT=0.07C_{\mathrm{T}}=0.07 and 0.090.09). This exercise shows that a reddened accretion disc and a small covering factor torus can explain the shape of the optical and near-infrared continuum of high-redshift LRDs. The low torus covering factor or reprocessing efficiency (RE) indicates that the obscuring structures of LRDs differ from those of local type-1 AGNs, which typically exhibit intermediate values (Ldust/Lbol≃0.4L_{\mathrm{dust}}/L_{\mathrm{bol}} \simeq0.4–0.70.7; Ramos Almeida et al. [109]).

Figure 6. Sy1 stack and standard deviation (blue solid line and cyan shaded area) and median CAT3D-wind torus models (orange lines) with energetic covering factors of CT=0.07C_{\mathrm{T}}=0.07, 0.080.08, and 0.090.09 (see Figure 2), reddened by AV=1.9A_{\mathrm{V}}=1.9 mag using the SMC bar extinction law. The accretion disc and torus components of the model with CT=0.08C_{\mathrm{T}}=0.08 are shown as red dashed and black dot-dashed lines, respectively, and the ratio between their integrated luminosities is Ldust/Ldiscreddened=2.8L_{\mathrm{dust}}/L_{\mathrm{disc}}^{\mathrm{reddened}}=2.8 (see Table 4). The Sy1 stack without extinction (see Figure 2) was scaled to the rest-frame 3 μm3\ \mu\mathrm{m} MIRI stack of Pérez-González et al. [104]. The NIRSpec Prism median stacked spectrum is shown in solid grey, and the MIRI stacks as blue diamonds. We also show the photometric LRD stacks from Delvecchio et al. [28] and Akins et al. [2] as green dots and a solid light purple line. The stacked photometry from Delvecchio et al. [28] was multiplied by a factor of 1.41.4 to match the other two stacks. The yellow triangles correspond to the SED of the local LRD LORD from Ji et al. (2026), divided by 35003500 to match the high-zz LRD stacks.

It is worth mentioning that the foreground extinction that we apply to the CT=0.08C_{\mathrm{T}}=0.08 model to match the optical-to-infrared shape of the LRD stack of Pérez-González et al. [104], of AV=1.9A_{\mathrm{V}}=1.9 mag, is higher than the value of 0.7 mag used in the right panel of Figure 4. This is because the CT=0.08C_{\mathrm{T}}=0.08 model has an intrinsically brighter optical continuum than the 5100 Å average flux of the Sy1s (see Figure 2), and thus requires a higher extinction to have the same optical-to-infrared ratio.

Figure 6 also shows two photometric LRD stacks for comparison. The green dots correspond to the stack of 300 LRDs with redshifts of 2<z<112<z<11 from Delvecchio et al. [28], and the light purple solid line, to the COSMOS-Web Maximal SED from Akins et al. [2]. The latter is the maximal SED allowed by the photometry upper limits, which at λrest>3\lambda_{\mathrm{rest}}>3 µm are from Spitzer, Herschel, SCUBA-2, and ALMA, and they include 434 LRDs at 5<z<95<z<9 (except in the case of the ALMA stacks, which include 30 and 128 LRDs at 1.2 and 2 mm; see Fig. 13 in Akins et al. [2] for further details). The LRD stack from Delvecchio et al. [28], which we multiplied by 1.4 to match the optical and near-infrared emission of the Pérez-González et al. [104] stack, is similar in shape to the other two LRD stacks, also showing a rise in the near-infrared continuum. Therefore, our reddened accretion disc + torus model also accommodates them.

In Figure 6 we show the SED of the local LRD LORD/The Egg ([84] (Ji et al. 2026)). Scaling its W4 flux (22 µm) to the Sy1 stack, the W1 (3.4 µm) and W2 (4.6 µm) fluxes fall within its standard deviation, while the W3 flux (12 µm) is above it, possibly due to a contribution from emission features (see § 3.4.2). Although in FλF_{\lambda} units the SED of LORD does not appear different from the Akins et al. [2] LRD stack (see Fig. 1 in Ji et al. (2026)), from Figure 6 it becomes clear that the V-shape is more pronounced and the fluxes between ∼0.7\sim0.7 and 1 µm are above the high-zz LRD stacks, partly due to contamination from emission lines. In any case, it is normal for individual sources to differ from stacks of hundreds of galaxies, as the stacking process washes out source peculiarities. We refer the reader to Madau et al. [88] for an example of an alternative accretion disc model and a clumpy dusty screen successfully replicating the LORD SED.

Energy balance arguments

Lin et al. [84] used three BBs of T∼\sim1000, 300, and 90 K to reproduce the infrared SED of LORD/Egg, and a fourth BB of T∼\sim5000 K to account for the optical continuum. As can be seen from their Table D1, the integrated luminosity of the three infrared BB components is almost twice the luminosity of the 5000 K BB. Therefore, in the scenario in which the optical emission is intrinsically red and not dust-reddened/extincted (i.e. the BH-star/cocoon scenarios), then the dust covering factor or RE (Ldust/Lbol∼Ldust/LoptL_{\mathrm{dust}}/L_{\mathrm{bol}}\sim L_{\mathrm{dust}}/L_{\mathrm{opt}} in the BH-star case) would be larger than unity. This model-independent infrared versus optical energy balance issue represents a challenge for the previously mentioned scenarios.

In the case of the accretion disc+torus model shown in Figure 6 (reddened model in Table 4), the energy of the dust component is almost three times higher than that of the apparent optical component (disc), but the key difference with the BH-star model is that in our case the optical corresponds to a reddened (by AV≃2A_{\mathrm{V}}\simeq2 mag) accretion disc component – the intrinsic bolometric luminosity is significantly higher. This means that if we consider the

Model CT=0.08C_{\mathrm{T}}=0.08LdiscL_{\mathrm{disc}} (104410^{44} erg s−1^{-1})LdustL_{\mathrm{dust}} (104410^{44} erg s−1^{-1})LtotalL_{\mathrm{total}}Ldust/LdiscL_{\mathrm{dust}}/L_{\mathrm{disc}}
Reddened (AV=1.9A_{\mathrm{V}}=1.9 mag)0.872.473.342.8
Unreddened (AV=0A_{\mathrm{V}}=0)17.42.9520.40.17

Table 4. Integrated luminosities of the accretion disc and torus components of the model SED in Figure 6, with and without extinction.

Notes. Luminosities are integrated over 0.01–36000 µm (z=7z=7, dL=69d_{\mathrm{L}}=69 Gpc), using the SMC bar extinction law for the reddened model. The proxy of Maiolino et al. [92], i.e. Eq. (1), gives CT=0.08C_{\mathrm{T}}=0.08 for the unreddened model, whereas directly integrating the disc and torus components gives 0.17. This mismatch comes from the conversion factors used in Eq. (1).

unreddened model (see Table 4), the luminosity of the dust component is lower than that of the disc (Ldust/Ldisc=0.17L_{\mathrm{dust}}/L_{\mathrm{disc}}=0.17), resulting in a reasonable dust covering factor.

Our proposed scenario of reddened accretion disc and low RE or covering factor torus to explain the optical-to-infrared SED of LRDs is supported by the recent work of Chandna et al. [20]. They used the SKIRT radiative transfer code to post-process zoom-in cosmological simulations and compared their models to LRD observations, including the stack of Pérez-González et al. [104]. They reported that the steep red optical slope of LRDs arises from a dust-attenuated AGN, whereas a less obscured stellar component dominates the UV emission. A similar approach, which includes a more detailed geometrical radiative transfer model, was proposed by Madau & Maiolino [87] and, for the specific case of the Egg/LORD, by Madau et al. [88], reaching similar results as our phenomenological approach.

Dust masses

If the torus model shown in Figure 6 (black dot-dashed line) indeed resembles the rest-frame mid- and far-infrared emission of high-redshift LRDs, the 5σ5\sigma upper limits from Akins et al. [2] would be largely consistent with our analysis. Their Herschel/PACS upper limits at 70 and 160 µm (∼\sim10 and 23 µm rest-frame) are ∼\sim90 and 60 times above the flux of the scaled Sy1 stack and torus model, respectively. The upper limit at λrest>100\lambda_{\mathrm{rest}}>100 µm from Delvecchio et al. [28] is more restrictive than those from Akins et al. [2], but still well above the torus emission at those wavelengths (see Figure 6). This comparison shows that the rest-frame infrared non-detections obtained for high-redshift LRDs should not be interpreted as a lack of dust in these targets. The increasing temperature of the CMB reduces the contrast between dust and the CMB emission (da Cunha et al. [24]), making it difficult to obtain constraining upper limits even at ALMA sensitivities ([19] (Casey et al. 2025)).

We fitted the dust torus model in Figure 6 (dot-dashed black line) with a sum of four modified BBs (Lν=4πMdustκνBν(T)L_{\nu}=4\pi M_{\mathrm{dust}}\kappa_{\nu}B_{\nu}(T), with κν=κ100(100 μm/λ)β\kappa_{\nu}=\kappa_{100}(100\,\mu\mathrm{m}/\lambda)^{\beta}; see Fig. F.1) of T≃\simeq100, 220, 570, and 1175 K. We adopted β=1.7\beta=1.7, set by the slope of the long-wavelength tail (λFλ∝λ−1.5\lambda F_{\lambda}\propto\lambda^{-1.5}), and κ100=25\kappa_{100}=25 cm2^{2} g−1^{-1} (Draine [33]). The fit reproduces the torus SED at λ≥1\lambda\geq1 µm with an rms of 0.05 dex. More than 99% of the dust mass is in the coolest component (T≃100T\simeq100 K), giving Mdust≃1.4×104 M⊙M_{\mathrm{dust}}\simeq1.4\times10^{4}\,M_{\odot}. Varying β\beta between 1.5 and 2 and κ100\kappa_{100} between 10 and 60 cm2^{2} g−1^{-1} changes this value between ∼4×103\sim4\times10^{3} and ∼6×104 M⊙\sim6\times10^{4}\,M_{\odot}. Since the cold component is constrained only by the far-infrared tail of the torus model, which is an extrapolation beyond the data (λ≲4 μm\lambda\lesssim4\,\mu\mathrm{m}), we regard this value as an order-of-magnitude estimate. Regardless, our dust mass estimation is in excellent agreement with the average dust mass predicted by Casey et al. [19], of Mdust=1.6−0.9+4.8×104 M⊙M_{\mathrm{dust}} = 1.6^{+4.8}_{-0.9} \times10^{4}\ M_{\odot}, for 675 LRDs at z≳4z \gtrsim4. They used the available measurements of AVA_{\mathrm{V}}, MUVM_{\mathrm{UV}}, and galaxy effective radius (ReffR_{\mathrm{eff}}) to derive dust mass and temperature, the latter being Tdust≃110T_{\mathrm{dust}} \simeq110 K. Most importantly, the dust masses inferred by our modelling are consistent with the range of dust masses typical of AGN dusty tori (102−104 M⊙10^{2} - 10^{4}\ M_{\odot}; García-Burillo et al. [40]; Hatziminaoglou et al. [55]).

The high-redshift LRD stack from Delvecchio et al. [28], shown as green dots in Figure 6, was reproduced by Madau & Maiolino [87] using a dust covering factor of 0.2 and a reddened super-Eddington AGN continuum with AV≃3A_{\mathrm{V}} \simeq3 mag. They reported a dust mass of ∼60\sim60-120120 solar masses for that torus, assuming a gas-to-dust ratio of 1000, which is probably more appropriate for metal-poor hosts such as those of LRDs (Madau & Maiolino [87]). The dust mass that we derived from the torus component in Figure 6, of Mdust≃104 M⊙M_{\mathrm{dust}} \simeq10^{4}\ M_{\odot}, is two orders of magnitude above the ∼102 M⊙\sim10^{2}\ M_{\odot} of Madau & Maiolino [87]. The difference is most likely due to the different modelling of the cold component of the dust emission, which is poorly constrained due to the lack of tight constraints at longer wavelengths, especially taking into account the steep dependence of the dust emission on the temperature. The torus modelling is also significantly different between the two approaches (especially in terms of density profile). Our estimated dust mass is two orders of magnitude below the current upper limits for high-redshift LRDs (Mdust≲106 M⊙M_{\mathrm{dust}} \lesssim10^{6}\ M_{\odot}; Casey et al. 2025; Chen et al. [21]; Setton et al. [122]).

Conclusions

We used rest-frame 3–27 μ\mum JWST spectra of eight local Sy1 galaxies to make predictions on the dust emission of LRDs. We normalised the individual Sy1 spectra by their broad Hα\alpha and 5100 Å continuum luminosities and stacked them, obtaining λLλ/LHα≃30\lambda L_{\lambda}/L_{\mathrm{H}\alpha} \simeq30–60 and λLλ/λL5100≃0.8\lambda L_{\lambda}/\lambda L_{5100} \simeq0.8–1.8 across 3–27 μ\mum.

We tested these scalings with the NIRSpec+MIRI stack of 249 LRDs at $2.3 < z < 9.3 from Pérez-González et al. [104] and with the SED of a local LRD at z = 0.1 ({citet}Ji et al. 2026; {citet}Lin et al. 2026b`). The 5100 Å continuum gives the more reliable prediction on the infrared emission of LRDs (rest-frame λ≲4 μ\lambda\lesssim4\ \mum for the high-zz stack and λ≲22 μ\lambda\lesssim22\ \mum for the local LRD) when modest extinction is considered (AV≲1A_{\mathrm{V}} \lesssim1 mag). Although the high-zz stack and, to a lesser extent, the local LRD SED probe large areas that may include host-galaxy emission, this agreement is consistent with an AGN-dominated optical continuum.

Although our results indicate that the infrared emission of LRDs can be predicted, within a factor of ∼1.5\sim1.5-2, from their 5100 Å continuum, the torus+accretion disc model that fits the local Sy1 stack does not simultaneously reproduce the optical and infrared SED of LRDs. Reproducing both requires a more reddened accretion disc (AV≃2A_{\mathrm{V}} \simeq2 mag) and a torus with a lower covering factor or RF (CT≃0.1C_{\mathrm{T}} \simeq0.1–0.2) than inferred for local Sy1 galaxies. This combination offers an alternative to BH-star scenarios (which ascribed the optical continuum to an unreddened blackbody of T∼5000T \sim5000 K), while yielding a more plausible infrared-to-optical energy balance.

From the fit of the torus model that reproduces the SED of the high-redshift LRD stack of Pérez-González et al. [104] with a sum of four optically thin modified BBs (T≃100T \simeq100, 220, 570, and 1175 K), we derive a dust mass of ∼104 M⊙\sim10^{4}\ M_{\odot} (between ∼4×103\sim4 \times10^{3} and ∼6×104 M⊙\sim6 \times10^{4}\ M_{\odot} considering different values of β\beta and κ100\kappa_{100}).

This dust mass is in good agreement with the prediction from Casey et al. [19], Mdust=1.6−0.9+4.8×104 M⊙M_{\mathrm{dust}} = 1.6^{+4.8}_{-0.9} \times10^{4}\ M_{\odot}, obtained from measured values of AVA_{\mathrm{V}}, MUVM_{\mathrm{UV}}, and ReffR_{\mathrm{eff}} for 675 LRDs at z≳4z \gtrsim4. This dust masses are two orders of magnitude below the current upper limits for high-redshift LRDs, based on their ALMA non detections (Mdust≲106 M⊙M_{\mathrm{dust}} \lesssim10^{6}\ M_{\odot}; Casey et al. 2025; Chen et al. [21]; Setton et al. [122]). According to the results presented here, LRDs are not dust deficient, but their accretion disc and torus properties would be different from those of local AGN.

Acknowledgements

This work is based on observations made with the NASA/ESA/CSA James Webb Space Telescope. The data were obtained from the Mikulski Archive for Space Telescopes at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-03127 for JWST and from the European JWST archive (eJWST) operated by the ESAC Science Data Centre (ESDC) of the European Space Agency. These observations are associated with programs GO 2016, 3535, 4972, 7195, and the Early Release Science Program 1328. CRA thanks Thaisa Storchi-Bergmann and Allan Schnorr-Müller for their help with the BLR extinction values. CRA and AA acknowledge support from the Agencia Estatal de Investigación of the Ministerio de Ciencia, Innovación y Universidades (MCIU/AEI) under grants “Tracking active galactic nuclei feedback from parsec to kiloparsec scales”, with reference PID2022–141105NB–I00, and “Local windows into the most extreme phases of black hole growth”, with reference PID2025–176305NB–I00, and the European Regional Development Fund (ERDF). This work is part of grant CEX2025-001609-S, awarded to the IAC under the Severo Ochoa Centre of Excellence program and funded by MICIU/AEI/10.13039/501100011033. AAH and LHM acknowledge support from grant PID2021-124665NB-I00 funded by MCIN/AEI/10.13039/501100011033 and by ERDF A way of making Europe. NAL acknowledges support from the Fundación Occident and the Instituto de Astrofísica de Canarias under the Visiting Researcher Programme 2025-2028 agreed between both institutions. PGP-G acknowledges support from grants PID2022-139567NB-I00 and PID2025-169195NB-I00 funded by Spanish Ministerio de Ciencia, Innovación y Universidades MICIU/AEI/10.13039/501100011033, and the European Union F EDER program Una manera de hacer Europa. AA is funded by the European Union (Widening Participation, ExGal-Twin, GA 101158446). EB acknowledges support from the Spanish grant PID2022-138621NB-I00 and PID2021-1234701B-I00, funded by MCIN/AEI/10.13039/501100011033/FEDER, EU. MS acknowledges support by the Ministry of Science, Technological Development and Innovation of the Republic of Serbia (MSTDIRS) through contract no. 451-03-2026-03/200002 with the Astronomical Observatory (Belgrade).

{sup}5` Department of Physics and Astronomy, University College London, Gower Street, London WC1E 6BT, UK

{sup}6` Centro de Astrobiología (CAB), CSIC-INTA, Camino Bajo del Castillo s/n, E-28692, Villanueva de la Cañada, Madrid, Spain

{sup}7` Department of Physics and Astronomy, The University of Texas at San Antonio (UTSA), 1 UTSA Circle, San Antonio, TX 78249-0600, USA

{sup}8` Instituto de Física Fundamental, CSIC, Calle Serrano 123, 28006 Madrid, Spain

{sup}9` School of Physics & Astronomy, University of Southampton, High-field, Southampton SO171BJ, UK

{sup}10` Space Telescope Science Institute, Baltimore, MD 21218, USA

{sup}11` Instituto de Radioastronomía and Astrofísica (IRyA-UNAM), 3-72 (Xangari), 8701, Morelia, Mexico

{sup}12` School of Mathematics, Statistics and Physics, Newcastle University, Newcastle upon Tyne NE1 7RU, UK

{sup}13` Max Planck Institute for Extraterrestrial Physics (MPE), Giessenbachstr. 1, D-85748 Garching, Germany

{sup}14` Department of Physics & Astronomy, University of South Carolina, Columbia, SC 29208, USA

{sup}15` LUX, Observatoire de Paris, Collège de France, PSL University, CNRS, Sorbonne University, Paris, 75014, France

{sup}16` Depto. de Física de la Tierra y Astrofísica, Facultad de CC. Físicas, Universidad Complutense de Madrid, E-28040 Madrid, Spain

{sup}17` Instituto de Física de Partículas y del Cosmos IPARCOS, Facultad CC. Físicas, Universidad Complutense de Madrid, E-28040 Madrid, Spain

{sup}18` Department of Physics, University of Oxford, Oxford OX1 3RH, UK

{sup}19` Institute of Astrophysics, Foundation for Research and Technology-Hellas (FORTH), Heraklion, GR-70013, Greece

{sup}20` Department of Astrophysics, University of California San Diego, 9500 Gilman Drive, San Diego, CA 92093, USA

{sup}21` Observatorio Astronómico Nacional (OAN-IGN)-Observatorio de Madrid, Alfonso XII, 3, 28014 Madrid, Spain

{sup}22` Departamento de Física, Universidad de Oviedo, Campus de Llamaquique, C/ Calvo Sotelo s/n, 33007 Oviedo, Spain

{sup}23` Department of Physics & Astronomy, University of Alaska Anchorage, Anchorage, AK 99508-4664, USA

{sup}24` National Astronomical Observatory of Japan, National Institutes of Natural Sciences (NINS), 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan

{sup}25` Department of Astronomy, School of Science, Graduate University for Advanced Studies (SOKENDAI), Mitaka, Tokyo 181-8588, Japan

{sup}26` Telespazio UK for the European Space Agency, ESAC, Camino Bajo del Castillo s/n, 28692 Villanueva de la Cañada, Spain

{sup}27` Department of Astronomy, University of Geneva, ch. d’Ecogia 16, 1290, Versoix, Switzerland

{sup}28` Instituto de Estudios Astrofísicos, Facultad de Ingeniería y Ciencias, Universidad Diego Portales, Avenida Ejército Libertador 441, Santiago, Chile

{sup}29` Kavli Institute for Astronomy and Astrophysics, Peking University, Beijing 100871, China

{sup}30` Astronomical Observatory, Volgina 7, 11060 Belgrade, Serbia

{sup}31` Department of Physics and Astronomy, Universiteit Gent, Proeftuinstraat 86 N3, B-9000 Ghent, Belgium

Comparison between JWST NIRSpec+MIRI and Spitzer/IRS spectra of the targets

Figure A.1 shows a comparison between the nuclear JWST NIRSpec+MIRI used in this work and their Spitzer/IRS spectra. The IRS spectra was dowloaded from the Combined Atlas of Sources with Spitzer IRS Spectra (CASSIS). The observations were done using the Short-Low (SL) and Long-Low (LL) spectrograph modules, which cover the wavelength range 5.3–38 µm with a spectral resolution of R=λ/Δλ∼60R = \lambda/\Delta\lambda\sim60–130. The SL1 and SL2 slits have widths of 3.6 and 3.7″, and the LL slits of 10.5 and 10.7″. Since the targets are type-1 AGN, optimal extraction was chosen.

Torus models

Figure B.1 shows the Sy1 stack (see right panel of Figure 1), in blue, fitted with the CAT3D-wind and CLUMPY torus models (orange and black lines).

Normalizations of individual galaxies and average normalized fluxes in JWST filters

Figure C.1 shows the rest-frame infrared spectra of the Sy1 galaxies, normalized to their corresponding Hα (left panel) and 5100 Å luminosities (right panel). The 5100 Å luminosities were calculated from LHαavgL_{\mathrm{H}\alpha}^{\mathrm{avg}} in Table 2 and Eq. (1) in Greene & Ho [49]. The Hα luminosities of NGC 6814 and NGC 7314 were extinction-corrected using the values reported in Table 2, and then used to calculated the corresponding L5100L_{5100}.

Filterλ\lambda (µm)λb\lambda_b (µm)λr\lambda_r (µm)log⁡(λLλLHα)\log(\frac{\lambda L_\lambda}{L_{\mathrm{H}\alpha}})log⁡(λLλL5100)\log(\frac{\lambda L_\lambda}{L_{5100}})
F356W3.5653.1363.9811.46842-0.09200
F444W4.4023.8814.9821.46868-0.08839
F560W5.65.0546.1711.47254-0.08317
F770W7.76.5818.6871.49753-0.05972
F1000W10.09.023010.8911.584330.03371
F1130W11.310.95311.6671.652420.10138
F1280W12.811.58814.1151.657170.10856
F1500W15.013.52716.6401.724380.17832
F1800W18.016.51919.5021.789630.24330
F2100W21.018.47723.1591.779760.23429
F2550W25.523.30126.7331.801680.25419

Table C.1. Average normalized fluxes in NIRCAM and MIRI wide filters.

Notes. Columns 1-4 are the name, pivot, blue, and red wavelengths of the JWST NIRCAM and MIRI wide filters. We refer the reader to the instruments webpages for details on their definition. Columns 5 and 6 are the average fluxes measured between the blue and red wavelengths of each filter in the normalized stack spectra shown in Figure 3.

Bolometric corrections

We measured the broad Hα flux from the fit shown in the left panel of Fig. D.1. We used two Gaussian components, of FWHM=3400 (broad) and 1050 km s−1^{-1} (intermediate), typical of high-redshift LRDs (Matthee et al. [93]; Greene et al. [50]; Brazzini et al. [15]), both resolved at the PRISM resolution (see Table D.1). More complex profiles have been tested with higher spectral resolution data (see Geris et al. [41] and references therein), but at the PRISM resolution we rather keep the fit as simple as possible. We fitted two additional Gaussians to account for the [NII] emission, having the same FWHM as the “narrow” (intermediate, hereafter) Hα component (see left panel of Fig. D.1). HeIλ6678 Å is also detected in the LRD stack, overlapping with the red wing of Hα. The fluxes and luminosities of the broad and intermediate Hα components are reported in Table D.1.

Using the integrated Hα fluxes and the fitted continuum (orange dashed line in Fig. D.1 and Table D.1), we measure EWs of 282 and 285 Å for the broad and intermediate components, which are within the range of nearby AGN (EW=[15,500] Å for log⁡LHαb=[40,44]\log L_{\mathrm{H}\alpha}^{\mathrm{b}}=[40,44] erg s−1^{-1}; Stern & Laor [128]). If we sum the two Hα components (red and blue lines in the left panel of Fig. D.1) we measure an EW=567 Å, and if we also include the [NII] line fluxes, EW=625 Å. The values considering the whole Hα and Hα+[NII] profile are typical of high-redshift LRDs (~400–1000 Å; Lin et al. [85]; Setton et al. [122]; Brazzini et al. [15]; Geris et al. [42]). Instead, the EW of the broad Hα component alone would imply that the covering factor of the BLR might not be that different from that of local AGN (see also Geris et al. [41]). We refer the reader to theoretical and observational studies of type-1 AGN at z≳3z \gtrsim3 using high spectral resolution data (e.g. Brazzini et al. [14]; Jouzdablis et al. 2026b; Geris et al. [41]; Matthee et al. [94]; Madau et al. [89]; Scholtz et al. [119]) for further discussion on the evolution of the structure and properties of the BLR with cosmic time.

BroadInterm.
λHα\lambda_{\mathrm{H}\alpha} (Å)6564.5±0.46564.5\pm0.46563.8±0.26563.8\pm0.2
FWHMHα\mathrm{FWHM}_{\mathrm{H}\alpha} (km s−1^{-1})3380±1103380\pm1101050±301050\pm30
EWHα\mathrm{EW}_{\mathrm{H}\alpha} (Å)282285
FHαF_{\mathrm{H}\alpha} (erg cm−2^{-2} s−1^{-1})(3.68±0.19)×10−18(3.68\pm0.19)\times10^{-18}(3.72±0.17)×10−18(3.72\pm0.17)\times10^{-18}
LHαL_{\mathrm{H}\alpha} (erg s−1^{-1})(2.10±0.11)×1042(2.10\pm0.11)\times10^{42}(2.12±0.09)×1042(2.12\pm0.09)\times10^{42}
λHβ\lambda_{\mathrm{H}\beta} (Å)4864.9±1.54864.9\pm1.54862.3±0.94862.3\pm0.9
FWHMHβ\mathrm{FWHM}_{\mathrm{H}\beta} (km s−1^{-1})3400±4303400\pm430…\ldots
EWHβ\mathrm{EW}_{\mathrm{H}\beta} (Å)7730
FHβF_{\mathrm{H}\beta} (erg cm−2^{-2} s−1^{-1})(7.98±1.17)×10−19(7.98\pm1.17)\times10^{-19}(3.10±1.27)×10−19(3.10\pm1.27)\times10^{-19}
LHβL_{\mathrm{H}\beta} (erg s−1^{-1})(4.55±0.67)×1041(4.55\pm0.67)\times10^{41}(1.77±0.72)×1041(1.77\pm0.72)\times10^{41}
λF5100\lambda F_{5100} (erg cm−2^{-2} s−1^{-1})5.69×10−175.69\times10^{-17}
λL5100\lambda L_{5100} (erg s−1^{-1})3.23×10433.23\times10^{43}

Table D.1. Optical spectral properties of NIRSpec LRD stack.

Notes. Central wavelength, FWHM corrected from instrumental broadening (~1100 km s−1^{-1} for Hα and ~1750 km s−1^{-1} for Hβ, using the median resolution values from Pérez-González et al. [104], corrected by a factor of 1/0.70), EW, integrated flux, and luminosity of the Hα and Hβ broad and intermediate components (blue and red lines in Fig. D.1). The line fluxes were multiplied by a factor of (1+z)(1+z), with zz=7, because they were measured in the rest-frame spectrum, in flux density units (FλF_\lambda). The 5100 Å flux was measured as the average of the λFλ\lambda F_\lambda rest-frame spectrum between 5080 and 5120 Å, so no (1+z)(1+z) factor was applied.

If we use the broad Hα flux measured from the LRD stack to estimate the continuum flux at 5100 Å using Eq. (1) in Greene & Ho [49], as we did for our local Sy1 galaxies in Section 3.3 (see Table 2 and Table 3), we derive a value of λF5100=7.95×10−17\lambda F_{5100}=7.95\times10^{-17} erg s−1^{-1} cm−2^{-2}. This flux is only a factor of 1.4 higher than the flux at 5100 Å that we measured from the LRD stack (see Table D.1). Indeed, if we use Eq. 5 in Stern & Laor [128], more commonly used for high-redshift and more luminous AGN, we get λF5100=5.15×10−17\lambda F_{5100}=5.15\times10^{-17} erg s−1^{-1} cm−2^{-2}, which is 0.91 times the flux measured directly from the continuum. This would indicate that broad Hα\alpha is a good proxy of the optical continuum luminosity in high-redshift LRDs and therefore of the AGN bolometric luminosity, contrary to what it has been recently claimed (Greene

Figure A.1. Rest-frame JWST/NIRspec+MIRI spectra of the eight Seyfert 1 galaxies (black line). The CASSIS optimal extraction Spitzer/IRS spectra is shown in green, and the purple line corresponds to the same spectra but scaled to the JWST/NIRspec+MIRI (at 20 μ\mum). Vertical dotted lines indicate the wavelength of the Pfund β\beta, γ\gamma, δ\delta, and ϵ\epsilon lines. Vertical dashed lines indicate the location of the 6.2, 7.7, 6.8, 11.3, and 17 μ\mum PAH bands.

Smoothed and interpolated mean Sy1 stack and model fits

Figure B.1. Smoothed and interpolated mean Sy1 stack (solid blue line) and fits the with the CAT3D-wind and CLUMPY torus models (shaded orange area: model SEDs between the 16 and 84 percentiles of the posterior. Solid black line: 50 percentile of the posterior). Vertical dotted lines indicate the peak wavelengths of the silicate features. The bottom panel shows the residuals as a percentage of the continuum flux, with the horizontal dotted lines indicating 0 and 15% error for guidance.

Figure C.1. Rest-frame infrared spectra of the Sy1 galaxies, with same colors as in the left panel of Figure 1, normalized to their corresponding Hα\alpha (left panel) and 5100 Å luminosities (right panel).

et al. 2026; [15]). If instead we use the total Hα\alpha flux (broad+intermediate components), we get a λF5100\lambda F_{5100} a factor of 2.6 and 1.8 higher than that measured directly from the continuum, using the Greene & Ho [49] and Stern & Laor [128] relations, respectively. If this total Hα\alpha flux is truly representative of the BLR emission, it might be indicating that the BLR covering factor is indeed somehow higher than that of local AGN.

Using the 5100 Å luminosity reported in Table D.1, which is measured directly from the LRD stack, and Lbol=9×λL5100L_{\mathrm{bol}} = 9 \times\lambda L_{5100} (Kaspi et al. [71]; Shen et al. [124]), we get log⁡Lbol=44.5\log L_{\mathrm{bol}} = 44.5 erg s−1^{-1}. This value is in agreement with the bolometric luminosity derived from SED fitting of the LRD stack (log⁡Lbol=44.1\log L_{\mathrm{bol}}=44.1–44.444.4 erg s−1^{-1}; Pérez-González et al. [104]). If we use the broad (total) Hα\alpha luminosity reported in Table D.1 and Lbol=(130−2.4+2.4)×LHαbL_{\mathrm{bol}} = (130^{+2.4}_{-2.4}) \times L_{\mathrm{H}\alpha}^{\mathrm{b}} from Stern & Laor [128], we measure log⁡Lbol=44.4\log L_{\mathrm{bol}}=44.4 (44.7) erg s−1^{-1}. If instead we use the bolometric corrections of Lbol=5.4×λL5100L_{\mathrm{bol}} = 5.4 \times\lambda L_{5100} and Lbol=19×LHαL_{\mathrm{bol}} = 19 \times L_{\mathrm{H}\alpha} derived for two luminous LRDs by Greene et al. [51], we get log⁡Lbol=44.2\log L_{\mathrm{bol}} = 44.2 erg s−1^{-1} from the 5100 Å luminosity of the LRD stack and 43.6 (43.9) erg s−1^{-1} from the broad (total) Hα\alpha luminosity reported in Table D.1. Therefore, we find that the Hα\alpha bolometric corrections derived from local samples of type-1 AGN reproduce the bolometric luminosity derived from SED fitting of the LRD stack by Pérez-González et al. [104]. More specifically, they require bolometric corrections of 60–130 (30–60) if we use the broad (total) Hα\alpha luminosity, consistent with Stern & Laor [128].

In the case of LORD (z=0.1z=0.1), Ji et al. (2026) fitted the Hα\alpha profile detected in their Gemini/GMOS spectra (R≃3400R \simeq3400) with two broad Gaussians of FWHM = 730 (NB) and 2050 (BB) km s−1^{-1} plus a narrow Gaussian of 90 km s−1^{-1} (see Figure 2 and Table 1 in Ji et al. (2026)). Adding the flux of the NB and BB components, they obtained a broad Hα\alpha luminosity of 41.90 erg s−1^{-1}. Using Stern & Laor [128] and log⁡MBH/M⊙=6.49\log M_{\mathrm{BH}}/M_{\odot}=6.49, they reported an Eddington ratio of 0.26±0.010.26 \pm0.01, similar to that of NGC 4395 (see Table 1). The EW reported by Ji et al. (2026) for NB+BB Hα is 450 Å, similar to the LRD stack from Pérez-González et al. [104] but still within the range of typical EWs of local AGN of similar bolometric luminosities (Stern & Laor [128]).

Figure D.1. Rest-frame NIRSpec LRD stack from Pérez-González et al. [104] around the Hα (left panel) and Hβ and [OIII] emission lines (right panel). The stacked Hα and Hβ profiles were reproduced with two Gaussian components (red and blue lines), both of them resolved at the spectral resolution of the NIRSpec/PRISM in the case of Hα (see Table D.1). The [NII] and HeIλ6678 Å lines are shown in pink, and their FWHMs have been fixed to that of the intermediate component of Hα (∼1050\sim1050 km s−1^{-1}). The [OIII] lines were fitted with two components, shown in pink and cyan. In both panels, the continuum is shown as a dashed orange line, the fit as a green solid line, and the residuals in black.

As in the case of the high-zz LRD stack, the 5100 Å fluxes of LORD estimated from its broad (BB) and total (NB+BB) Hα fluxes using eq:5 in Stern & Laor [128] are 0.9 and 1.5 times the 5100 Å flux directly measured from the optical continuum. If confirmed for more examples of local and high-redshift LRDs, the optical continuum of LRDs can be estimated either from the broadest component fitted to their Hα profiles or by the total Hα flux (broad+intermediate components) divided by a factor of ∼1.5\sim1.5-2.

Finally, considering the 3σ3\sigma upper limit on the hard X-ray luminosity of LORD, obtained by Ji et al. (2026) from NuSTAR and Chandra data (log⁡L2−10keV<41.9\log L_{2-10\mathrm{keV}} < 41.9 erg s−1^{-1}), Lbol/L2−10keV>137L_{\mathrm{bol}}/L_{2-10\mathrm{keV}}>137 if we estimate LbolL_{\mathrm{bol}} as 130λLHαb130\lambda L_{\mathrm{H}\alpha}^{\mathrm{b}}. If we calculate the bolometric luminosity from the flux of the BB Hα component only, reported in Table 1 in Ji et al. (2026), we get Lbol/L2−10keV>80L_{\mathrm{bol}}/L_{2-10\mathrm{keV}}>80, and with the 5100 Å luminosity measured from the Gemini spectrum, of 42.8742.87 erg s−1^{-1}, Lbol/L2−10keV>88L_{\mathrm{bol}}/L_{2-10\mathrm{keV}}>88 using Lbol=9×λL5100L_{\mathrm{bol}}=9\times\lambda L_{5100} (Kaspi et al. [71]; Shen et al. [124]). This factor is closer to the limit derived from the BB Hα component, but they are still much larger than the local Lbol/L2−10keVL_{\mathrm{bol}}/L_{2-10\mathrm{keV}} values of ∼8\sim8-25 (Ho [57]; Duras et al. [34]), due to the X-ray faintness of LRDs. In the case of the Sy1 galaxies considered here, if we use the broad Hα luminosities reported in Table 2 and the X-ray luminosities reported in Table 1, we find Lbol/L2−10keV<24L_{\mathrm{bol}}/L_{2-10\mathrm{keV}}<24, with a median value of 10.5.

Dust masses

We fitted the dust torus model in Figure 6 with a sum of four modified BBs (Lν=4πMdust,kκνBν(T)L_{\nu}=4\pi M_{\mathrm{dust},k}\kappa_{\nu}B_{\nu}(T)), with κν=κ100(100 μm/λ)β\kappa_{\nu}=\kappa_{100}\left(100\ \mu\mathrm{m}/\lambda\right)^{\beta}; see Fig. F.1). We adopted β=1.7\beta=1.7, as constrained from the slope of the long-wavelength tail, from ∼200\sim200 to 5000 μ\mum (λFλ∝λ−4.65\lambda F_{\lambda}\propto\lambda^{-4.65}), and κ100=25\kappa_{100}=25 cm2^{2} g−1^{-1} (Draine [33]). The fit reproduces the torus SED at λ≥1 μ\lambda\geq1\ \mum with an rms of 0.05 dex. Using three BBs instead of four produces an rms of 0.07 dex. More than 99% of the dust mass is in the coolest component (T∼100T\sim100 K), giving Mdust≃1.4×104 M⊙M_{\mathrm{dust}}\simeq1.4\times10^{4}\ M_{\odot} (1.3×104 M⊙1.3\times10^{4}\ M_{\odot} using three BBs instead of four).

Extinction corrections

Here we show the same plots as in Figure 4, but correcting the broad Hα and 5100 Å fluxes measured from the LRD stack of Pérez-González et al. [104] by different AVA_{\mathrm{V}} values using the SMC bar extinction curve of Gordon et al. [48].

Figure E.1. Same as in Figure 4, but correcting the Hα\alpha and 5100 Å fluxes measured from the LRD stack of Pérez-González et al. [104] by different AVA_{\mathrm{V}} values using the SMC bar extinction curve of Gordon et al. [48]. The Sy1 stacks have also been corrected from the same values of the extinction, but using the local ISM extinction law of Chiar & Tielens [22].

Torus model and sum of four optically thin modified blackbodies

Figure F.1. Torus model shown in Figure 6 (black dots) and best-fit sum of four optically thin modified blackbodies (β=1.7\beta= 1.7, κ100=25 cm2 g−1\kappa_{100} = 25\ \mathrm{cm}^{2}\,\mathrm{g}^{-1}; red solid line). The dashed lines show the individual components. The bottom panel shows the residuals in dex.

References

  1. [2]Akins, H. B., Casey, C. M., Lambrides, E., et al. 2025, ApJ, 991, 37
  2. [3]Alonso-Herrero, A., Esquej, P., Roche, P. F., et al. 2016, MNRAS, 455, 563
  3. [4]Alonso-Herrero, A., Ramos Almeida, C., Mason, R., et al. 2011, ApJ, 736, 82
  4. [5]Ananna, T. T., Bogdán, Á., Kovács, O. E., Natarajan, P., & Hickox, R. C. 2024, ApJ, 969, L18
  5. [7]Asensio Ramos, A. & Ramos Almeida, C. 2009, ApJ, 696, 2075
  6. [8]Asensio Ramos, A. & Ramos Almeida, C. 2013, MNRAS, 428, 195
  7. [9]Baggen, J. F. W., van Dokkum, P., Brammer, G., et al. 2024, ApJ, 977, L13
  8. [10]Balmaverde, B. & Capetti, A. 2014, A&A, 563, A119
  9. [11]Bentz, M. C., Ferrarese, L., Onken, C. A., Peterson, B. M., & Valluri, M. 2019, ApJ, 885, 161
  10. [14]Brazzini, M., D’Eugenio, F., Maiolino, R., et al. 2025, MNRAS, 544, L167
  11. [15]Brazzini, M., D’Eugenio, F., Maiolino, R., et al. 2026, A&A, 713, A332
  12. [16]Caglar, T., Burchtser, L., Brandl, B., et al. 2020, A&A, 634, A114
  13. [17]Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245
  14. [19]Casey, C. M., Akins, H. B., Kokorev, V., et al. 2024, ApJ, 975, L4
  15. [20]Chandna, S., Di Mascia, F., Gallerani, S., et al. 2026, arXiv e-prints, arXiv:2609.34903arxiv.org/abs/2609.34903
  16. [21]Chen, K., Li, Z., Inayoshi, K., & Ho, L. C. 2025, ApJ, 994, L42
  17. [22]Chiar, J. E. & Tielens, A. G. G. M. 2006, ApJ, 637, 774
  18. [23]Cisternas, M., Gadotti, D. A., Knapen, J. H., et al. 2013, ApJ, 776, 50
  19. [24]da Cunha, E., Groves, B., Walter, F., et al. 2013, ApJ, 766, 13
  20. [25]da Silva, P., Menezes, R. B., Díaz, Y., et al. 2023, MNRAS, 519, 1293
  21. [26]de Graaff, A., Hviding, R. E., Naidu, R. P., et al. 2026, MNRAS, 551, stag1567
  22. [27]de Graaff, A., Rix, H.-W., Naidu, R. P., et al. 2025, A&A, 701, A168
  23. [28]Delvecchio, I., Daddi, E., Magnelli, B., et al. 2025, A&A, 704, A313
  24. [29]D’Eugenio, F., Nelson, E., Ji, X., et al. 2025, arXiv e-prints, arXiv:2510.00101arxiv.org/abs/2510.00101
  25. [30]Devereux, N. 2021, MNRAS, 500, 786
  26. [31]Dong, X., Wang, T., Wang, J., et al. 2008, MNRAS, 383, 581
  27. [32]Donnan, F. R., García-Bernete, I., Rigopoulou, D., et al. 2026, arXiv e-prints, arXiv:2603.12200arxiv.org/abs/2603.12200
  28. [33]Draine, B. T. 2003, ARA&A, 41, 241
  29. [34]Duras, F., Bongiorno, A., Ricci, F., et al. 2020, A&A, 636, A73
  30. [35]Furtak, L. J., Secunda, A. R., Greene, J. E., et al. 2025, A&A, 698, A227
  31. [37]García-Bernete, I., Ramos Almeida, C., Alonso-Herrero, A., et al. 2019, MNRAS, 486, 4917
  32. [38]García-Bernete, I., Rigopoulou, D., Alonso-Herrero, A., et al. 2022, A&A, 666, L5
  33. [39]García-Bernete, I., Rigopoulou, D., Donnan, F. R., et al. 2024b, arXiv e-prints, arXiv:2409.05686
  34. [40]García-Burillo, S., Alonso-Herrero, A., Ramos Almeida, C., et al. 2021, A&A, 652, A98
  35. [41]Geris, S., Maiolino, R., Isobe, Y., et al. 2026a, MNRAS, 545, staf1979
  36. [42]Geris, S., Maiolino, R., Ji, X., et al. 2026b, arXiv e-prints, arXiv:2606.21614arxiv.org/abs/2606.21614
  37. [43]González-Martín, O., Díaz-González, D. J., Martínez-Paredes, M., et al. 2025, MNRAS, 539, 2158
  38. [44]González-Martín, O., Masegosa, J., García-Bernete, I., et al. 2019a, ApJ, 884, 11
  39. [45]González-Martín, O., Masegosa, J., García-Bernete, I., et al. 2019b, ApJ, 884, 11
  40. [46]González-Martín, O., Ramos Almeida, C., Fritz, J., et al. 2023, A&A, 676, A73
  41. [47]Goold, K., Seth, A., Molina, M., et al. 2026, arXiv e-prints, arXiv:2601.16977DOI
  42. [48]Gordon, K. D., Clayton, G. C., Misselt, K. A., Landolt, A. U., & Wolff, M. J. 2003, ApJ, 594, 279
  43. [49]Greene, J. E. & Ho, L. C. 2005, ApJ, 630, 122
  44. [50]Greene, J. E., Labbe, I., Goulding, A. D., et al. 2024, ApJ, 964, 39
  45. [51]Greene, J. E., Setton, D. J., Furtak, L. J., et al. 2026, ApJ, 996, 129
  46. [53]Hainline, K. N., Maiolino, R., Juodžbalis, I., et al. 2025, ApJ, 979, 138
  47. [54]Harikane, Y., Zhang, Y., Nakajima, K., et al. 2023, ApJ, 959, 39
  48. [55]Hatziminaoglou, E., Fritz, J., Franceschini, A., et al. 2008, MNRAS, 386, 1252
  49. [56]Hermosa Muñoz, L., González Fernández, J. R., Alonso-Herrero, A., et al. 2025, arXiv e-prints, arXiv:2509.14019
  50. [57]Ho, L. C. 2009, ApJ, 699, 626
  51. [58]Hönig, S. F. & Kishimoto, M. 2010, A&A, 523, A27
  52. [59]Hönig, S. F. & Kishimoto, M. 2017, ApJ, 838, L20
  53. [60]Hönig, S. F., Kishimoto, M., Gandhi, P., et al. 2010, A&A, 515, A23
  54. [61]Ilić, D., Popović, L. Č., La Mura, G., Ciroi, S., & Rafanelli, P. 2012, A&A, 543, A142
  55. [62]Inayoshi, K. & Ichikawa, K. 2024, ApJ, 973, L49
  56. [63]Inayoshi, K. & Maiolino, R. 2025, ApJ, 980, L27
  57. [67]Juodžbalis, I., Ji, X., D’Eugenio, F., et al. 2026a, arXiv e-prints, arXiv:2607.19057arxiv.org/abs/2607.19057
  58. [68]Juodžbalis, I., Ji, X., Maiolino, R., et al. 2024, MNRAS, 535, 853
  59. [70]Juodžbalis, I., Maiolino, R., Baker, W. M., et al. 2026c, MNRAS, 546, stag086
  60. [71]Kaspi, S., Smith, P. S., Netzer, H., et al. 2000, ApJ, 533, 631
  61. [72]Kido, D., Ioka, K., Hotokezaka, K., Inayoshi, K., & Irwin, C. M. 2025, MNRAS, 544, 3407
  62. [73]Kiyota, T., Ouchi, M., Yanagisawa, H., et al. 2026, arXiv e-prints, arXiv:2608.10832DOI
  63. [74]Kocevski, D. D., Finkelstein, S. L., Barro, G., et al. 2025, ApJ, 986, 126
  64. [77]Koshida, S., Yoshii, Y., Kobayashi, Y., et al. 2017, ApJ, 842, L13
  65. [80]Landt, H., Bentz, M. C., Ward, M. J., et al. 2008, ApJS, 174, 282
  66. [81]Lewis, K. T. & Eracleous, M. 2006, ApJ, 642, 711
  67. [82]Lin, R., Zheng, Z.-Y., Wang, J., et al. 2026a, ApJ, 1005, 150
  68. [83]Lin, X., Fan, X., Cai, Z., et al. 2026b, ApJ, 997, 364
  69. [84]Lin, X., Fan, X., Cai, Z., et al. 2026c, arXiv e-prints, arXiv:2605.21574arxiv.org/abs/2605.21574
  70. [85]Lin, X., Wang, F., Fan, X., et al. 2024, ApJ, 974, 147
  71. [86]Liu, H., Jiang, Y.-F., Quataert, E., Greene, J. E., & Ma, Y. 2025, ApJ, 994, 113
  72. [87]Madau, P. & Maiolino, R. 2026, arXiv e-prints, arXiv:2602.22386arxiv.org/abs/2602.22386
  73. [88]Madau, P., Maiolino, R., Ji, X., Lin, X., & Fan, X. 2026a, arXiv e-prints, arXiv:2611.22316arxiv.org/abs/2611.22316
  74. [89]Madau, P., Maiolino, R., Scholtz, J., & D’Eugenio, F. 2026b, arXiv e-prints, arXiv:2604.04216arxiv.org/abs/2604.04216
  75. [91]Maiolino, R., Scholtz, J., Curtis-Lake, E., et al. 2024, A&A, 691, A145
  76. [92]Maiolino, R., Shemmer, O., Imanishi, M., et al. 2007, A&A, 468, 979
  77. [93]Matthee, J., Naidu, R. P., Brammer, G., et al. 2024, ApJ, 963, 129
  78. [94]Matthee, J., Torralba, A., Pezzulli, G., et al. 2026, arXiv e-prints, arXiv:2603.17667arxiv.org/abs/2603.17667
  79. [95]Mazzolari, G., Gilli, R., Brusa, M., et al. 2024, A&A, 687, A120
  80. [96]Mazzolari, G., Scholtz, J., Maiolino, R., et al. 2025, A&A, 700, A12
  81. [97]Mejía-Restrepo, J. E., Trakhtenbrot, B., Koss, M. J., et al. 2022, ApJS, 261, 5
  82. [98]Naidu, R. P., Matthee, J., Katz, H., et al. 2026, Nature, 656, 329
  83. [99]Nenkova, M., Sirocky, M. M., Ivezić, Ž., & Elitzur, M. 2008a, ApJ, 685, 147
  84. [101]Nikopoulos, G. P., Watson, D., Sneppen, A., et al. 2026, A&A, 710, A136
  85. [102]Pereira-Santaella, M., Álvarez-Márquez, J., García-Bernete, I., et al. 2022, A&A, 665, L11
  86. [103]Pereira-Santaella, M., González-Alfonso, E., García-Bernete, I., García-Burillo, S., & Rigopoulou, D. 2024, A&A, 681, A117
  87. [104]Pérez-González, P. G., Barro, G., Carniani, S., et al. 2026, arXiv e-prints, arXiv:2602.20247arxiv.org/abs/2602.20247
  88. [105]Peterson, B. M., Wanders, I., Bertram, R., et al. 1998, ApJ, 501, 82
  89. [106]Prieto, M. A., Nadolny, J., Fernández-Ontiveros, J. A., & Mezcua, M. 2021, MNRAS, 506, 562
  90. [107]Ramos Almeida, C., Asensio Ramos, A., Westerdorp Plaza, C., et al. 2026, A&A, 706, A100
  91. [108]Ramos Almeida, C., García-Bernete, I., Pereira-Santaella, M., et al. 2025, A&A, 698, A194
  92. [109]Ramos Almeida, C., Levenson, N. A., Alonso-Herrero, A., et al. 2011, ApJ, 731, 92
  93. [110]Ramos Almeida, C., Levenson, N. A., Rodríguez Espinosa, J. M., et al. 2009, ApJ, 702, 1127
  94. [111]Ramos Almeida, C., Piqueras López, J., Villar-Martín, M., & Bessiere, P. S. 2017, MNRAS, 470, 964
  95. [112]Ricci, C., Trakhtenbrot, B., Koss, M. J., et al. 2017, ApJS, 233, 17
  96. [113]Richards, G. T., Hall, P. B., Vanden Berk, D. E., et al. 2003, AJ, 126, 1131
  97. [114]Rusakov, V., Watson, D., Nikopoulos, G. P., et al. 2026, Nature, 649, 574
  98. [115]Schimoia, J. S., Storchi-Bergmann, T., Grupe, D., et al. 2015, ApJ, 800, 63
  99. [116]Schimoia, J. S., Storchi-Bergmann, T., Nemmen, R. S., Winge, C., & Eracleous, M. 2012, ApJ, 748, 145
  100. [117]Schimoia, J. S., Storchi-Bergmann, T., Winge, C., Nemmen, R. S., & Eracleous, M. 2017, MNRAS, 472, 2170
  101. [118]Schnorr-Müller, A., Davies, R. I., Korista, K. T., et al. 2016, MNRAS, 462, 3570
  102. [119]Scholtz, J., D’Eugenio, F., Maiolino, R., et al. 2026, arXiv e-prints, arXiv:2603.22277arxiv.org/abs/2603.22277
  103. [120]Schulz, H., Knake, A., & Schmidt-Kaler, T. 1994, A&A, 288, 425
  104. [121]Secunda, A., Somerville, R. S., Jiang, Y.-F., et al. 2026, ApJ, 996, 6
  105. [122]Setton, D. J., Greene, J. E., Spilker, J. S., et al. 2025, ApJ, 991, L10
  106. [123]Shapovalova, A. I., Popović, L. Č., Chavushyan, V. H., et al. 2017, MNRAS, 466, 4759
  107. [124]Shen, Y., Richards, G. T., Strauss, M. A., et al. 2011, ApJS, 194, 45
  108. [125]Sneppen, A., Matthews, J. H., Watson, D., et al. 2026a, arXiv e-prints, arXiv:2604.09399arxiv.org/abs/2604.09399
  109. [126]Sneppen, A., Watson, D., Matthews, J. H., & Sim, S. A. 2026b, A&A, 712, A61
  110. [127]Stalevski, M., Ricci, C., Ueda, Y., et al. 2016, MNRAS, 458, 2288
  111. [128]Stern, J. & Laor, A. 2012, MNRAS, 423, 600
  112. [129]Storchi-Bergmann, T., Eracleous, M., Teresа Ruiz, M., et al. 1997, ApJ, 489, 87
  113. [130]Taylor, A. J., Finkelstein, S. L., Kocevski, D. D., et al. 2025, ApJ, 986, 165
  114. [131]Tee, W. L., Fan, X., Wang, F., & Yang, J. 2025, ApJ, 983, L26
  115. [132]Thim, F., Hoessel, J. G., Saha, A., et al. 2004, AJ, 127, 2322
  116. [133]Tonry, J. L., Dressler, A., Blakeslee, J. P., et al. 2001, ApJ, 546, 681
  117. [134]Tortosa, A., Ricci, C., Du, P., et al. 2026, A&A, 708, A293
  118. [135]Tully, R. B. 1988, Nearby galaxies catalog
  119. [136]Williams, C. C., Alberts, S., Ji, Z., et al. 2024, ApJ, 968, 34
  120. [137]Yanagisawa, H., Ouchi, M., Kiyota, T., et al. 2026, arXiv e-prints, arXiv:2606.09726arxiv.org/abs/2606.09726
  121. [138]Yuan, W., Macri, L. M., Peterson, B. M., et al. 2021, ApJ, 913, 3
  122. [139]Yue, M., Eilers, A.-C., Ananna, T. T., et al. 2024, ApJ, 974, L26
  123. [140]Zhang, L., Packham, C., Hicks, E. K. S., et al. 2024, ApJ, 974, 195

Paper details

vlnc:W_6e7d9cp3
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents