The near-infrared broad emission line region of active galactic nuclei. II. The one-micron continuum

We use quasi-simultaneous near-infrared (near-IR) and optical spectroscopy from four observing runs to study the continuum around 1 micron in 23 well-known broad-emission line active galactic nuclei (AGN). We show that, after correcting the optical spectra for host galaxy light, the AGN continuum around this wavelength can be approximated by the sum of mainly two emission components, a hot dust blackbody and an accretion disc. The accretion disc spectrum appears to dominate the flux at ~1 micron, which allows us to derive a relation for estimating AGN black hole masses based on the near-IR virial product. This result also means that a near-IR reverberation programme can determine the AGN state independent of simultaneous optical spectroscopy. On average we derive hot dust blackbody temperatures of ~1400 K, a value close to the sublimation temperature of silicate dust grains, and relatively low hot dust covering factors of ~7%. Our preliminary variability studies indicate that in most sources the hot dust emission responds to changes in the accretion disc flux with the expected time lag, however, a few sources show a behaviour that can be attributed to dust destruction.

we outlined the rationale of our programme, presented the observations of the first three epochs and addressed briefly some of the important issues regarding the physics of the most prominent broad emission lines. Here, we present the fourth epoch of observation and investigate the continuum around the 1-μm inflection point.
The AGN spectral continuum region around the rest-frame wavelength of ∼1 μm is believed to sample simultaneously two important emission components, namely the accretion disc (e.g. Malkan & Sargent 1982;Malkan 1983) and the hottest part of the putative dusty torus (e.g. Barvainis 1987;Neugebauer et al. 1987). However, although it is assumed to be understood, it has not yet been sampled spectroscopically in its entirety. By probing the long-wavelength end of the accretion disc spectrum (Kishimoto, Antonucci & Blaes 2005;Kishimoto et al. 2008), such an investigation has the potential to solve the discrepancy often found between theoretical models and observations (see e.g. review by Koratkar & Blaes 1999). Furthermore, since the accretion disc most likely illuminates directly the inner parts of the dust structure, monitoring the change in spectral flux and slope of these two emission components relative to each other can constrain the location and geometry of the obscurer. So far, only few sources have been observed in dust reverberation programmes (e.g. Glass 1992Glass , 2004Nelson 1996;Oknyanskij et al. 1999;Minezaki et al. 2004;Suganuma et al. 2006;Koshida et al. 2009).
The paper is organized as follows. In Section 2, we briefly introduce the sample and discuss the observations. In Section 3, we derive pure AGN continuum spectral energy distributions (SEDs), based on which we constrain the individual continuum components (Section 4). The variability of these components is discussed in Section 5. Finally, in Section 6 we summarize our main results and present our conclusions. Throughout this paper, we have assumed cosmological parameters H 0 = 70 km s −1 Mpc −1 , M = 0.3 and = 0.7.

O B S E RVAT I O N S
The target selection, observational strategy and data reduction procedures have been described in detail in Paper I. In short, we obtained for a sample of 23 well-known relatively nearby (z 0.3) and bright (J 14 mag) broad emission line AGN during four observing runs contemporaneous (within 2 months) near-IR and optical spectroscopy. The observations were carried out between 2004 May and 2007 January with a single object being typically observed twice within this period.
In the near-IR we used the SpeX spectrograph (Rayner et al. 2003) at the NASA Infrared Telescope Facility (IRTF), a 3-m telescope on Mauna Kea, Hawaii. We chose the short cross-dispersed mode (SXD, 0.8-2.4 μm) and a short slit of 0.8 × 15 arcsec. The optical spectra were obtained with the FAST spectrograph (Fabricant et al. 1998) at the Tillinghast 1.5-m telescope on Mt. Hopkins, Arizona, using the 300 l mm −1 grating and a 3-arcsec long-slit, resulting in a wavelength coverage of ∼3720-7515 Å.
Paper I presented details for the first three epochs May, 2006January and 2006. Similarly, we give in Tables 1 and 2 the details for the fourth IRTF epoch carried out on 2007 January 24-26. These nights were mostly photometric with a seeing in the range of ∼0.7-1 arcsec. Note. The columns are as follows: (1) object name; (2) observation date; (3) exposure time; (4) average airmass; S/N in the continuum over ∼100 Å measured at the central wavelength of the (5) J, (6) H and (7) K bands; for the star used to correct for telluric absorption (8) name, (9) distance from the source and (10) average airmass.

AG N S P E C T R A L E N E R G Y D I S T R I B U T I O N S
Our aim is to study the spectral shape of the AGN continuum over the entire observed frequency range and how it changes as the AGN flux varies. For this purpose, we need to isolate the spectrum of the pure AGN and ensure an acceptable relative photometry between epochs. In practice, this means that we need to subtract the flux of the underlying host galaxy and correct the individual spectra for flux calibration errors (e.g. varying weather conditions, aperture and seeing effects). In the absence of short-term variability, these steps should yield an alignment between the AGN optical and near-IR spectral parts. In principle, accurate (within a few per cent) relative photometry of the spectra can be achieved a posteriori by using the fact that the narrow emission line fluxes of AGN are constant over several years (Peterson 1988(Peterson , 1993. However, this requires that the observations were performed under similar conditions, i.e. using the same aperture and slit position angle and having the same seeing. In general, these requirements were not met for our optical observations, which were obtained in service mode under widely varying weather conditions. On the other hand, the near-IR spectra were obtained uniformly during four separate epochs, with one of them (2006 June) having photometric weather conditions throughout. In addition, the much better seeing achieved in the near-IR relative to the optical band allowed us to use a relatively small slit for these spectra, which minimized the contamination by host galaxy light.
Given these considerations, we chose the following approach to obtain pure AGN continuum SEDs with acceptable relative photometry. First, we performed relative photometry of the near-IR spectra using the flux of the strong narrow emission line [S III] λ9531 with the epoch observed under the best weather conditions (usually 2006 June or 2007 January) as a reference. We have measured the emission line fluxes as described in Paper I and list them in Table 3. We note that [S III] λ9531 is not observed for the source HE 1228+013, and therefore no adjustment of the near-IR spectra was made in this case. Secondly, we subtracted the host galaxy flux from the (adjusted) near-IR spectra. Finally, we corrected the flux scale of the optical spectra by requiring that after the subtraction of the (optical) host galaxy contribution the optical and near-IR AGN spectral parts align. In Fig. A1, we show the original data (left-hand panels) and the data after host galaxy subtraction and spectral alignment were applied (right-hand panels).
We have estimated the host galaxy contribution in the apertures of both the near-IR and optical spectra using the Hubble Space Telescope (HST) images of Bentz et al. (2006aBentz et al. ( , 2009) and following their approach. The observed HST fluxes were transformed to a rest-frame wavelength of 5100 Å by applying a colour correction factor based on the model bulge galaxy template of Kinney et al. (1996) and were corrected for Galactic extinction using A λ values derived from the hydrogen column densities of Dickey & Lockman (1990). The unabsorbed rest-frame 5100-Å fluxes were then used to scale the galaxy template from Mannucci et al. (2001) of the appropriate Hubble type, which was subsequently subtracted from the (rest-frame) spectrum. Details of our host galaxy flux estimates are given in Tables 3 and 4 for the near-IR and optical spectra, respectively. As expected, the relatively small aperture used in the near-IR includes significantly less host galaxy flux than the optical aperture, on average a factor of ∼3.
For 8/23 objects, we did not have suitable HST images. In these cases, we have estimated the host galaxy contribution based on the linear correlation present between the logarithms of the aperture and the enclosed host galaxy luminosity for the sources with HST images (Fig. 1). The observed correlation is log L host = 0.73 (±0.04) × log (aperture) + 39.05(±0.04), where L host is the host galaxy luminosity at rest-frame 5100 Å (in erg s −1 Å −1 ) and aperture is the spectral aperture (in kpc 2 ). Additionally, we could not find information on the host galaxy type for four sources. In these cases, we have assumed elliptical hosts for the two high-redshift sources IRAS 1750+508 and PDS 456 and spiral (S0) hosts for the two low-redshift sources H 2106−099 and H 1934−063.
The frequency gap between our near-IR and optical spectra is large enough (log ν ∼ 0.05) that we need to assume an overall spectral shape in order to judge the two parts as being 'aligned'. Therefore, since the AGN continuum bluewards of ∼1 μm is generally thought to be emitted by the accretion disc, we have considered this component as the model and calculated its spectrum as detailed in Section 4.1. Assuming an accretion disc spectrum, excess host galaxy contribution to the optical spectrum (but not to the near-IR spectrum) was apparent in the majority of our low-redshift (z 0.1) sources, requiring also a host galaxy flux correction factor. However, as Fig. 2 shows, in most cases the host correction factors are similar to the correction factors applied to the total optical spectra, and in only a few sources they indicate a real underestimation (by factors of 3).
The relative photometry in the near-IR required flux correction factors of <2, with the exception of Mrk 590 (factor of ∼3), and most spectra had flux differences relative to the best-weather epoch of no more than ∼20-30 per cent (Fig. 3, top panel). The correction factors for the optical spectra spanned a larger range of ∼0.6-3.8. The narrow emission line [O III] λ5007 is usually used for relative photometry in the optical, and we list in Table 4 also its flux. As Fig. 3 shows, whereas our approach did not succeed in bringing also the [O III] λ5007 line fluxes of the different epochs into alignment (bottom panel), it did reduce the scatter of the original flux ratio distribution (middle panel). Finally, for a subsample of 12 sources we compare in Fig. 4 the optical correction factors with the ratio between the [O III] λ5007 line fluxes observed by reverberation programmes (Korista et al. 1995;Peterson et al. 1998;Bentz et al. 2006b;Denney et al. 2010) and us. Fig. 4 indicates that the optical correction factors are mostly overestimates relative to photometric conditions. This result could be partly due to our near-IR flux calibration, which was based on the telluric standard stars. Since the target integration times were considerably larger than those of the stars, the seeing would have varied substantially for the former, leading overall to an overestimate of the extraction aperture and so of the flux scale.

AG N C O N T I N U U M C O M P O N E N T S
Our observations cover the continuum rest-frame frequency range of ν ∼ 10 14 -10 15 Hz, which is believed to sample two important emission components, namely the accretion disc (e.g. Malkan & Sargent 1982;Malkan 1983) and the hottest part of the putative dusty torus (e.g. Barvainis 1987;Neugebauer et al. 1987). These two components are predicted to have opposite spectral behaviours, namely the accretion disc and the hot dust emission rising and falling towards longer frequencies (in a logarithmic νf ν versus ν plot), respectively, resulting in an inflection point at the location where they meet (at ∼1 μm; e.g. Carleton et al. 1987;Ward et al. 1987;Elvis et al. 1994;Glikman, Helfand & White 2006;Riffel, Rodríguez-Ardila & Pastoriza 2006).
Our present knowledge of this special continuum region is based mainly on optical spectroscopy combined with near-IR photometry, with the two frequency ranges usually not observed  Note. The columns are as follows: (1) object name; (2) redshift from NED; (3) Galactic extinction at rest-frame 5100 Å; (4) Hubble type of the host galaxy; contemporaneously. Our observations now give us for the first time the opportunity to define the separate continuum components at spectroscopic rather than photometric precision. This permits us to investigate their properties in greater detail than was possible before. In particular, we are interested to understand if a near-IR reverberation mapping campaign generally requires simultaneous optical spectroscopy to determine the state of the ionizing flux or if the latter can also be derived from the continuum behaviour at longer wavelengths.
For the purpose of the following analysis, we have normalized all SEDs at ∼1 μm (Fig. 5)   Note. The columns are as follows: (1)  in both their near-IR and optical spectral parts (Fig. 6) and are not considered further.

Accretion disc
The AGN continuum bluewards of ∼1 μm is believed to be emitted by the accretion disc, which is considered the main source of ionizing flux producing the broad emission lines. In order to test if this component is indeed seen in our data, we have calculated accretion disc spectra. We have assumed a steady geometrically thin, optically thick accretion disc, in which case the emitted flux is independent of viscosity and each element of the disc face radiates roughly as a blackbody with a characteristic temperature depending only on the mass of the black hole, M BH , the accretion rate,Ṁ, and   the radius of the innermost stable orbit (e.g. Peterson 1997;Frank, King & Raine 2002). We have adopted the Schwarzschild geometry (non-rotating black hole) and for this the innermost stable orbit is at r in = 6 r g , where r g is the gravitational radius defined as r g = GM BH /c 2 , with G being the gravitational constant and c the speed of light. Furthermore, we have assumed that the disc is viewed face-on.
The accretion disc spectrum is fully constrained by the two quantities, the mass and accretion rate of the black hole. Twothirds of our sample have black hole masses derived from reverberation mapping campaigns. For the remainder (seven sources), we have estimated this quantity by applying the virial theorem M BH ∝ v 2 r/G, where v and r are the velocity and radial dis-  Table 4 applied, respectively.  disc spectrum to the data. We show our results in Fig. 5 (left-hand panels, dotted lines) and list the relevant values in Table 5.
The calculated accretion disc spectrum approximates well the AGN continuum slope bluewards of ∼1 μm in all high-redshift (z 0.1) sources, where the host galaxy contribution to the total flux is negligible at all wavelengths. In particular the near-IR spectral part (∼0.8-1 μm, covered by the IRTF spectrum) is well reproduced, which is free from major contaminating components such as e.g. strong Fe II emission and the 'small blue bump' (Grandi 1982;Wills, Netzer & Wills 1985) sometimes found in the optical spectral part. However, most important for our future near-IR reverberation programme is the result that the accretion disc emission can still dominate at the large wavelength of ∼1 μm. This means that we observe the flux of the ionizing component directly in the near-IR, which will allow us to determine the AGN state without optical spectroscopy.  In the lower redshift sources, the host galaxy contribution is expected to be minimal in the near-IR spectrum, since these observations were taken through a relatively narrow slit, and in the bluer part of the optical spectrum, since at these frequencies the AGN emission lies well above that of the host galaxy. Indeed, in both these frequency regimes the calculated accretion disc spectrum is always a satisfactory approximation. However, excess host emission is often observed at the red end of the optical spectrum, which has the effect of flattening the spectrum (in a logarithmic νf ν versus ν plot). Nevertheless, subtraction of sufficient host galaxy flux leads also in this spectral range to an alignment of the data with the expected accretion disc emission.

Hot dust
The AGN continuum redwards of ∼1 μm is believed to be produced by the hottest part of a dusty torus, which surrounds the central ionizing source and obscures the BELR for lines of sight close to the plane of the accretion disc. Assuming that this component is indeed seen in our data, we have subtracted from the total spectrum that of the accretion disc and have fitted to the result a blackbody spectrum. For this purpose we have used the C routine MPFIT (ver-   Table 5. A hot blackbody spectrum appears to approximate well the near-IR AGN continuum in all our sources. In particular, the typical curvature of such a spectrum is evident in our spectra due to their relatively large wavelength coverage. This behaviour was also noted by Rodríguez-Ardila & Mazzalay (2006), whose near-IR spectrum of one broad-line AGN (Mrk 1239) covered an even larger wavelength range of 0.8-4.5 μm. In this respect, we note that the blackbody curvature is not evident in the original spectra (without the accretion disc component subtracted), which resemble rather a single power law. The resulting temperatures for the hot blackbody component are in the range of T hot ∼ 1100-1700 K, which are typical values of the dust sublimation temperature for most astrophysical grain compositions (≈1000-2000Salpeter 1977). The overall temperature distribution is relatively narrow and has a well-defined mean of T hot = 1365 ± 18 K (Fig. 8). Similarly narrow temperature distributions were obtained previously by AGN studies using low-and medium-resolution near-IR spectra (Kobayashi et al. 1993;Riffel et al. 2009) and more recently by studies based on high-resolution near-IR photometric and interferometric observations (Kishimoto et al. 2007(Kishimoto et al. , 2009(Kishimoto et al. , 2011. This well-defined peak of the blackbody temperatures is close to the sublimation temperature of silicate dust grains (Kimura et al. 2002). More refractory dust types are known, notably carbonaceous dust, e.g. graphite, which can survive up to ∼2000 K. The absence of such hot dust, except in one object (NGC 5548), suggests that either carbonaceous dust is rare in AGN or some other mechanism than dust sublimation sets the maximum dust temperature. Dust formation is critically dependent on the numerical ratio of carbon to oxygen atoms in the parent gas; most pair up to form CO and do not form dust (Whittet 2003). An imbalance in C/O atom numbers biases dust formation strongly towards either carbon-rich or oxygen-rich (including silicates). The ∼1400-K peak in AGN suggests an oxygen-rich environment from which the dust formed.
The strength of our data set is that it allows us to observe simultaneously the accretion disc and the hot dust emission. Therefore, we can derive for the first time meaningful covering factors for the dusty obscurer in AGN. If the UV radiation from the accretion disc emitted into the solid angle, , defined by the dust distribution is completely absorbed and re-emitted in the IR, the dust covering factor is C = /4π = hot L ν dν/ acc L ν dν ≈ 0.4 × (νL hot /νL acc ), where νL hot and νL acc are the peak luminosities of the hot blackbody and accretion disc spectrum, respectively (e.g. Barvainis 1987;  Granato & Danese 1994). Note that whereas ν L hot lies only slightly outside the observed spectral range and is therefore well constrained by the data, the accretion disc peak luminosity is strongly modeldependent. In Fig. 9, we show the distribution of hot dust covering factors. We obtain values in the range of C ∼ 0.01-0.6 and a mean of C = 0.07 ± 0.02. Our average value is a factor of ∼6 lower than the average total dust covering factor obtained by Sanders et al. (1989) ( C = 0.40 ± 0.01), who considered the ratio of the entire integrated IR luminosity (in the frequency range ν ∼ 10 12 -10 14.5 Hz) to the integrated accretion disc luminosity for a large (∼100 sources) sample of bright quasars.
The relatively low hot dust covering factors could be a result introduced by the correction factors that we applied to the optical spectra (Section 3), i.e. we have artificially raised the accretion disc flux relative to the hot dust emission. In order to address this concern, we have plotted in Fig. 10 the covering factors versus the optical correction factors. Whereas we see an envelope in the sense that the higher the optical correction factor, the lower the highest covering factor, overall we observe large hot dust covering factors (C 0.1) in only a few cases (mostly in Ark 120 and NGC 4151) or, put differently, AGN states cluster around relatively low C values independent of the optical correction factor.

Additional continuum components
The observed AGN SEDs can be explained mainly by the sum of an accretion disc spectrum and emission from hot dust. However, two additional continuum components are expected to be present, which albeit weaker could alter the AGN continuum slope. The first is the diffuse continuum (thermal emission and scattering) from the BELR, which can be strong due to the relatively high gas densities found in these regions. The Balmer jump and continuum (λ < 3646 Å) were identified in AGN spectra long ago (e.g. Malkan & Sargent 1982) and predicted in early photoionization models as well (e.g. Wills et al. 1985). If the BELR is emitting a Balmer continuum, it is also emitting a Paschen continuum (λ < 8205 Å), the strength of which could be important throughout the optical and near-IR spectrum. Other, weaker, recombination continua (Brackett, etc.) will also be present within the IR. Korista & Goad (2001) used the photoionization code CLOUDY ) and calculated the full diffuse BELR continuum assuming clouds with simple distributions in gas density and incident ionizing flux. They found that after correcting for host galaxy light ∼20 per cent of the flux at 5100 Å may be due to this component, which is dominated by the thermal continua (nearly 40 per cent at the Balmer jump; see Korista & Goad for details). In a representation of the SED as in Fig. 5 (log νf ν versus log ν)  component is expected to strengthen with increasing wavelength as the accretion disc spectrum falls, resulting in a flattening of the near-UV to near-IR AGN continuum spectrum as compared to an accretion disc alone.
A second additional component is due to the finite albedo of the same dusty clouds emitting the thermal IR continuum.  investigated the effective albedo of a variety of gas clouds potentially present within the BELR. These contributions are included in the diffuse continuum just described. They also presented the effective albedo of a high column density, dusty cloud with parameters grossly approximating the clouds producing most of the hot dust emission. They found an ∼20 per cent albedo spanning the wavelength range of λ = 2500 Å-1 μm that declines significantly for wavelengths outside. While the detailed wavelength-dependent albedo depends on the grain size distributions and composition, their overall amplitudes should not differ greatly. Although the spatially integrated scattered nuclear light will depend on factors such as the geometry of the scattering region and the observer's viewing angle, contributions of ∼10 per cent to the AGN continuum are possible. Due to the finite grain albedo, the dusty torus will contribute additional light to the near-UV to the near-IR spectrum that otherwise would not be expected from its thermal emission alone. As with the contribution from the diffuse continuum light from the BELR, this   contribution will act to flatten (in log ν f ν versus log ν) the AGN continuum.
Due to the uncertainties in the absolute flux calibration and the relative normalizations of our near-IR and optical spectra, we have not attempted to isolate the contributions of the diffuse BELR continuum and the dust scattering of nuclear light to the AGN continuum. However, doing so should provide additional physical and geometrical constraints on the BELR and inner dusty torus. Here we only point out that the strongest contributor to the diffuse BELR continuum, namely the Balmer jump and continuum, is clearly visible in most of our optical spectra, forming together with Fe II the so-called small blue bump (Grandi 1982;Wills et al. 1985).

AG N C O N T I N U U M VA R I A B I L I T Y
Both AGN continuum components sampled by our spectra are expected to be variable. The main cause of variability will be a change in the accretion rate, which will lead to a change in the dust illumination. However, although the variability behaviour of the accretion disc has been studied extensively by optical reverberation campaigns, only few sources have been observed in dust reverberation programmes so far (e.g. Glass 1992Glass , 2004Nelson 1996;Oknyanskij et al. 1999;Minezaki et al. 2004;Suganuma et al. 2006;Koshida et al. 2009). Therefore, the exact location of the dusty obscurer as well as its geometry remain uncertain. In this section, we discuss the first results of our coordinated optical and near-IR reverberation campaign. Based on these results, we will select the final sample for monitoring and develop a suitable observational strategy.
We have multiple observation epochs available for 15/21 sources listed in Table 5, of which five sources have more than two epochs. Defining the variability factor as the flux change at 1 μm relative to the lowest flux state in a fixed period of 1 yr, we obtain the distribution shown in Fig. 11. Note that this variability factor is different from that listed in Table 5 (column 6), which refers to the total period between two epochs, and was calculated assuming uniform variability between the lowest flux epoch and the epoch with a higher flux. The most variable of our sources are 3C 273, HE 1228+013 and NGC 4151, which have at least one epoch with a 1-yr variability factor of 2, followed by PG 0844+349, Ark 120, NGC 5548 and 4593, with variability factors of ∼1.5-2. The source Mrk 817 shows no variability and our three least variable sources are IRAS 1750+508, Mrk 509 and H 2106−099.
With our future near-IR reverberation programme, we will be able to study also the variability behaviour of the hot dust and its dependence on the state of the ionizing source. A first look gives Fig. 12 Peterson et al. (2004) and est: estimated based on the correlation in Fig. 7; (4) IRTF run; (5) total AGN (host galaxy-subtracted) luminosity at 1 µm; (6) variability factor relative to the lowest flux epoch; (7) peak luminosity of the accretion disc; for the hot dust component; (8) its blackbody temperature; (9) peak luminosity; (10) covering factor and (11) average radius (in light-years).
Two important results become evident from Fig. 12. First, for most sources we observe a response of the hot dust to the change in accretion disc luminosity, which is always lower than the latter, as expected for a response lag. Most interestingly, however, is that a dust response is observed even in those sources with estimated hot dust radii well above 1 light-year, i.e. for which we would expect no response. The only exception in this group is the source IRAS 1750+508. Secondly, for a few sources we observe a deficit of hot dust in the high state. This effect is most pronounced in the source NGC 4151, for which it was observed also by Koshida et al. (2009) using long-term optical and near-IR imaging and attributed to dust destruction.

S U M M A RY A N D C O N C L U S I O N S
We have used four epochs of quasi-simultaneous (within 2 months) near-IR and optical spectroscopy of 23 broad-line AGN to study the continuum spectral shape around 1 μm. Our main results can be summarized as follows.
(i) The accretion disc spectrum appears to dominate the flux at ∼1 μm, which allows us to derive a new relation that can be used to estimate AGN black hole masses. It is based on the near-IR virial product, defined here as the product between the width of the Paβ broad emission line and the integrated 1-μm continuum luminosity. The dominance of the accretion disc spectrum at such  . Histogram of the hot dust covering factor C ≈ 0.4 × (νL hot / νL acc ), where νL hot and νL acc are the peak luminosities of the hot blackbody and accretion disc spectrum, respectively. long wavelengths means that the AGN state can be determined directly from the near-IR spectrum, making simultaneous optical spectroscopy for a reverberation programme unnecessary.
(ii) An adequate subtraction of (in particular optical) host galaxy light reveals that the AGN continuum in the rest-frame frequency range of ν ∼ 10 14 -10 15 Hz can be approximated by the sum of mainly two emission components, a hot dust blackbody and an accretion disc spectrum.
(iii) For the hot dust component, we derive temperatures in the range of T hot ∼ 1100-1700 K, which are typical values of the dust sublimation temperature, with a mean of T hot = 1365 ± 18 K. This mean value is close to the sublimation temperature of silicate dust grains, indicating that either carbonaceous dust is rare in AGN or some other mechanism than dust sublimation sets the maximum dust temperature. The resulting hot dust covering factors are relatively Figure 10. The hot dust covering factor versus the correction factor applied to the optical spectrum. Figure 11. Histogram of the variability factor measured at 1 µm and calculated over a fixed period of 1 yr. low and in the range of C ∼ 0.01-0.6, with a mean of C = 0.07 ± 0.02.
(iv) Our preliminary variability studies have identified promising candidates for a future near-IR reverberation programme. Our three most variable sources in the near-IR are 3C 273, HE 1228+013 and NGC 4151. Furthermore, we have studied the response of the hot dust emission to changes in the accretion disc flux. Most sources show the expected time lag, but a few sources have a deficit of hot dust in the high state, which indicates dust destruction.
In our future work, we will study the variability of the near-IR broad emission lines and constrain their physical conditions using detailed photoionization models. In the longer term, we plan to image our sample with current and future near-IR interferometers (Elvis & Karovska 2002).