Superluminous Transients at AGN Centers from Interaction between Black Hole Disk Winds and Broad-line Region Clouds

The basic idea of our BH disk-wind model is to make STACs bright through the interaction between the BH disk wind and the clouds in the BLR surrounding the central BH. Our model is mainly composed of two phenomena: (i) the mass ejection from the BH disk and (ii) the interaction between the ejected mass and the BLR clouds.

At first, at stage (i), a strong wind from the central BH accretion disk with the mass ${M}_{\mathrm{ej}}^{\mathrm{BH}}$ and the velocity ${v}_{\mathrm{ej}}^{\mathrm{BH}}$ is launched. Our model is independent of the wind launching mechanism, but we suggest that the BH disk wind is triggered by the limit–cycle oscillation that is induced by the thermal-viscous instability when the mass accretes onto the BH with the near-Eddington rate (Section 4.1). This is because the observed luminosities of STACs are near- or super-Eddington. In addition to the sudden increase of the disk luminosity, the limit–cycle oscillation can make the mass ejection rate temporarily more than 10 times larger than the Eddington rate (e.g., Ohsuga 2006). A typical velocity of the BH wind is ${v}_{\mathrm{ej}}\simeq 0.1c$, where c is the speed of light. The timescale of the activity is ∼10–100 s for stellar-mass BHs and is considered to be ∼10–1000 days for M_BH ∼ 10⁶–10⁸ M_⊙ (Section 4.1). For the central BH mass of 10⁷ M_⊙, if the mass ejection with the Eddington rate (L_Edd ≃ 10⁴⁵ erg s⁻¹) occurs for 100 days, the total kinetic energy output is E_BH ∼ 10⁵² erg. This is comparable to the typical total radiated energy of STACs. In our model, this powerful outflow powers STACs. For the velocity of ${v}_{\mathrm{ej}}^{\mathrm{BH}}\simeq 0.1c$, the ejected mass needs to be ${M}_{\mathrm{ej}}^{\mathrm{BH}}\simeq 1\,{M}_{\odot }$ (${E}_{\mathrm{kin}}^{\mathrm{BH}}={M}_{\mathrm{ej}}^{\mathrm{BH}}{v}_{\mathrm{ej}}^{\mathrm{BH}\ 2}/2$).

Although the disk luminosity can exceed the Eddington luminosity during the limit–cycle oscillation, the disk wind will immediately obscure the luminous disk. In the above example of M_BH = 10⁷ M_⊙, ∼10⁻² M_⊙ is ejected in a day. With the velocity of ≃0.1c, it reaches ∼3 × 10¹⁴ cm. If we assume that the spherically symmetric fully ionizing gas flows as the BH disk wind, we can apply the same approximation to estimate its opacities as in Roth et al. (2016). Assuming that the inner radius of the ejecta is ∼10¹³ cm, the electron-scattering optical depth of the ejecta at one day is τ_es ≃ 180. Because the absorption optical depth (τ_ab) is ∼10⁻⁴ times smaller than the scattering optical depth (Roth et al. 2016), the effective optical depth in the BH disk wind is ${\tau }_{\mathrm{eff}}=\sqrt{{\tau }_{\mathrm{sc}}{\tau }_{\mathrm{ab}}}\sim 2$. Therefore, at one day after the mass ejection, the BH disk wind can already obscure the central X-ray and ultraviolet source. As long as the mass ejection continues, the central source is kept obscured via the wind with τ_eff of the order of unity, and the accretion disk does not directly contribute to the luminosities of STACs in our model.

Now we consider the second phenomenon, i.e., (ii) the interaction between the BH disk wind and the BRL clouds. We illustrate this phase in Figure 1. When n_e ≃ 10⁹ cm⁻³ in the BLR clouds, the corresponding cloud density is ≃2 × 10⁻¹⁵ g cm⁻³. This density is as high as those found in SNe IIn (e.g., Moriya et al. 2014). Therefore, similar physical processes to those in SNe IIn are presumed to occur in these clouds when BH disk wind collides with them. SN ejecta are replaced with BH disk winds, and dense circumstellar media are replaced with BLR clouds. Strong radiative shock waves are created between the BH disk wind and the BLR clouds. High-energy photons from the shock are immediately absorbed by the clouds due to their high density and heat up the clouds. Most of radiation is emitted in near-ultraviolet to optical wavelengths and the shocks also make narrow lines as in SNe IIn. The efficient conversion from the kinetic energy to radiation at the shocks makes the BH disk wind bright and results in STACs.

The emission timescale from the interaction (t_em) is determined by the timescale for the BH wind to lose its kinetic energy. The deceleration timescale corresponds to the timescale for the shock to sweep the comparable mass to the BH disk wind. Keeping in mind that only f_c M_ej interacts with the BLR clouds, we obtain the following emission timescale:

$$\begin{eqnarray}&&{t}_{\mathrm{em}}\simeq 240{\left(\displaystyle \frac{{f}_{c}{M}_{\mathrm{ej}}^{\mathrm{BH}}}{{M}_{\odot }}\right)}^{\tfrac{1}{3}}{\left(\displaystyle \frac{{v}_{\mathrm{ej}}^{\mathrm{BH}}}{0.1c}\right)}^{-1}{\left(\displaystyle \frac{\varepsilon }{{10}^{-3}}\right)}^{-\tfrac{1}{3}}{\left(\displaystyle \frac{{n}_{e}}{{10}^{9}{\mathrm{cm}}^{-3}}\right)}^{-\tfrac{1}{3}}\,\mathrm{days}.\end{eqnarray} \tag{ 8 }$$

The light-crossing time of the shock is ∼10 times smaller than t_em, and it does not affect STAC timescales.

We briefly summarize the general picture. First, the BH disk wind is launched by a disk instability like limit–cycle oscillations. Although the disk will be bright due to the instability, the BH disk wind becomes optically thick, and the central disk is obscured as long as the mass ejection continues. The mass ejection typically continues for about 100 days or less. The BH disk wind becomes bright enough to be observed as STACs due to its collision with the surrounding clouds in the BLRs, which efficiently converts kinetic energy to radiation. The emission timescale is determined by the deceleration timescale, which is typically more than 100 days.

We apply our BH disk-wind model to CSS100217. We start by looking into the AGN properties hosting CSS100217.

The spectrum of the galaxy hosting CSS100217 obtained by Sloan Digital Sky Survey suggests that it is a Seyfert and therefore an AGN (Abazajian et al. 2009). The AGN has the g-band magnitude of 17.9 mag, and its redshift is 0.147 (Drake et al. 2011). Thus, λL₅₁₀₀ ≃ 9 × 10⁴³ erg s⁻¹. Its bolometric luminosity L_bol is ≃7 × 10⁴⁴ erg s⁻¹, if we apply the bolometric correction of L_bol ≃ 8.1λL₅₁₀₀ (Runnoe et al. 2012). The broad component of its Hβ emission has 2900 km s⁻¹ (FWHM) with the luminosity of L(Hβ) ≃ 4 × 10⁴¹ erg s⁻¹ (Drake et al. 2011). Adopting the equations in Section 2, we estimate R_BLR ≃ 0.03 pc, M_BLR ≃ 4 M_⊙, ε ≃ 10⁻³, and M_BH ≃ 5 × 10⁷ M_⊙. Therefore, the Eddington luminosity is ≃6 × 10⁴⁵ erg s⁻¹. Thus, the Eddington ratio Γ = L_bol/L_Edd is about 0.1 at the central BH.

The integrated radiation energy emitted by CSS100217 is ≃1.3 × 10⁵² erg (Drake et al. 2011). Assuming that this kinetic energy originates from the ejecta from the BH disk wind and only f_c M_ej (f_c ≃ 0.4) interacts with the BLR clouds, ${E}_{\mathrm{kin}}^{\mathrm{BH}}\simeq 3\times {10}^{52}\,\mathrm{erg}$ is required. With the typical BH wind velocity of ${v}_{\mathrm{ej}}^{\mathrm{BH}}\simeq 0.1c$ (e.g., Hashizume et al. 2015), the required total ejecta mass is estimated as ${M}_{\mathrm{ej}}^{\mathrm{BH}}\simeq 4\,{M}_{\odot }$.

We find that ${f}_{c}{M}_{\mathrm{ej}}^{\mathrm{BH}}\simeq 2\,{M}_{\odot }$ interacts with the BLR clouds. Because ${M}_{\mathrm{BLR}}\simeq 4\,{M}_{\odot }\gtrsim {f}_{c}{M}_{\mathrm{ej}}^{\mathrm{BH}}\simeq 2\,{M}_{\odot }$, the BLR region contains enough amount of matter to decelerate the BH disk wind. The emission timescale t_em is t_em ≃ 280 days. This timescale roughly matches that observed in CSS100217. Overall, the properties of CSS100217 can be explained by our BH disk-wind model, and CSS100217 can be powered by the interaction between the BH disk wind and the BLR clouds.

The host AGN of PS16dtm has λL₅₁₀₀ ≃ 10⁴³ erg s⁻¹, L(Hβ) ≃ 7 × 10⁴⁰ erg s⁻¹, and FWHM(Hβ) ≃ 1200 km s⁻¹ (Blanchard et al. 2017). These properties result in R_BLR ≃ 0.009 pc, M_BLR ≃ 0.6 M_⊙, ε ≃ 7 × 10⁻³, and M_BH ≃ 3 × 10⁶ M_⊙. Because of the small BH mass, the AGN continuum could be affected by stellar light, and M_BH may actually be slightly lower (≃10⁶ M_⊙; e.g., Xiao et al. 2011). Adopting the same as in CSS100217, we obtain L_bol ≃ 9 × 10⁴³ erg s⁻¹ and Γ ≃ 0.3.

The g-band peak magnitude of PS16dtm is −22 mag (Blanchard et al. 2017). Because it has not been faded away, total radiated energy in PS16dtm exceeds ≃3 × 10⁵¹ erg. To acquire this amount of radiation energy, ${f}_{c}{M}_{\mathrm{ej}}^{\mathrm{BH}}$ needs to be larger than 0.3 M_⊙. If we assume that ${f}_{c}{M}_{\mathrm{ej}}^{\mathrm{BH}}$ is as much as M_BLR ≃ 0.6 M_⊙, t_em can be as long as t_em ≃ 110 days. PS16dtm is likely to start declining soon if it is powered by the BH disk wind as its duration now matches t_em ≃ 110 days. The observed broad (∼10,000 km s⁻¹) Mg ii absorption could be related to the fast BH disk wind.

Blanchard et al. (2017) argue that PS16dtm is not due to an AGN variability because (i) they have a luminosity increase much larger than those observed in other AGNs and (ii) viscous timescales in disks are much longer than timescales observed in STACs. However, both difficulties are overcome if we consider the interaction between the BH disk wind and the BLR clouds. An important characteristic of PS16dtm is the X-ray fading during the optical brightening. Blanchard et al. (2017) suggest that TDE debris surrounding the X-ray emitting region obscures the X-ray emission from the center. As discussed in Section 3.1, the BH disk wind can obscure the central region and the X-ray fading is naturally expected in our BH disk-wind model. In the case of PS16dtm, assuming that the ejecta mass is ∼0.5 M_⊙ and the inner and outer radii of the ejecta are ∼10¹³ cm and 10¹⁶ cm (a typical outer shock radius at around 100 days), respectively, the effective optical depth in the BH disk wind becomes τ_eff ∼ 3. Therefore, the BH disk wind can actually obscure the central X-ray source. Because τ_eff is close to unity, the degree of the obscuration in each STAC can easily change depending on, e.g., the mass and asphericity of the BH disk wind.

A possible physical mechanism to initiate the transient disk winds from the BH accretion disks is limit–cycle oscillations (e.g., Honma et al. 1991; Szuszkiewicz & Miller 1997; Watarai & Mineshige 2003). When the mass accretion rate of the disk is near-Eddington rate, the disk cycles between the gas-pressure-dominated standard disk state and the slim disk state via the disk instability. When it transforms from a gas-pressure-dominated state to a slim disk state, its luminosity can increase significantly and the luminosity can exceed the Eddington luminosity for a while. During this time, strong radiation-driven disk winds can be launched. By the two-dimensional radiation hydrodynamics simulations, Ohsuga (2006) has successfully reproduced the limit–cycle behavior and the transient radiation-driven winds, of which the kinetic energy flux is above the Eddington luminosity (see also Teresi et al. 2004a, 2004b; Ohsuga 2007).

The most famous candidate that exhibits this kind of limit–cycle oscillation is microquasar GRS 1915+105. The luminosity of GRS 1915+105 is comparable to the Eddington luminosity, and this object clearly shows the quasi-periodic luminosity variation with the timescale of several tens of seconds (Belloni et al. 1997a, 1997b). Interestingly, the possible double-peak shape in the LC of PS16dtm is sometimes found in microquasar GRS1915+105 (e.g., Janiuk & Czerny 2005). Janiuk & Czerny (2011) have reported that the disk instability in the radiation-pressure-dominated region occurs in several stellar-mass accreting BHs (M_BH ∼ 10 M_⊙) with Γ ≳ 0.1. Indeed, the AGNs hosting CSS100217 and PS16dtm have an Eddington ratio of around 0.1 and are likely to be able to trigger the limit–cycle oscillation.³ The timescales for stellar-mass BH disks to have super-Eddington accretion are ∼10–100 s. As the timescale is expected to be roughly proportional to M_BH, the timescales for the BH disk wind from M_BH ∼ 10⁶–10⁸ M_⊙ are thought to be ∼10–1000 days. These are comparable to or less than the emission timescale t_em (Equation (8)). Thus, STACs usually radiate with the emission timescale in our model.

If the BH disk wind caused by the limit–cycle oscillation is responsible for the STAC activities, STACs are expected to be brightened intermittently. The prediction for the duration of the quiescent phase strongly depends on the viscosity adopted in the disk model (Watarai & Mineshige 2003). The quiescent phase can be more than 10 times longer than the luminous phase if the disk viscosity is sensitive to the radiation pressure. If we simply assume that the properties of the viscosity remain unchanged, the duration of the quiescent phases is roughly proportional to the viscous timescale and therefore the central BH mass. In the case of M_BH ∼ 10⁶–10⁸ M_⊙, the quiescent phase can last for several months to decades.

We have proposed that STACs like CSS100217 and PS16dtm are related to AGN activities. We argue that STACs are preferentially observed in NLS1 galaxies because of their relatively small BH masses (∼10⁶–10⁸ M_⊙; Zhou et al. 2006). Here, we consider diversities of STACs due to the diversities in BHs in AGNs in the context of our BH disk-wind model.

The BLR density does not vary much due to their similar electron densities (n_e ∼ 10⁹ cm⁻³) and filling factors (ε ∼ 10⁻³; Equation (5)). Also, the BH disk-wind velocity is presumed to be similar (${v}_{\mathrm{ej}}^{\mathrm{BH}}\simeq 0.1c$). Therefore, the diversities in STACs in our BH disk-wind model mainly come from the diversities in the total amount of the mass ejected from the central BH accretion disks.

${M}_{\mathrm{ej}}^{\mathrm{BH}}\simeq 4\,{M}_{\odot }$ is required to explain the observational properties of CSS100217. Because ${M}_{\mathrm{ej}}^{\mathrm{BH}}$ can be proportional to L_Edd and L_Edd ∝ M_BH, ${M}_{\mathrm{ej}}^{\mathrm{BH}}$ can be proportional to M_BH. Then, the STAC luminosity $({E}_{\mathrm{kin}}^{\mathrm{BH}}/{t}_{\mathrm{em}})$ is proportional to ${M}_{\mathrm{BH}}^{2/3}$ as ${t}_{\mathrm{em}}\propto {M}_{\mathrm{ej}}^{\mathrm{BH}1/3}$. On the other hand, the AGN luminosities hosting STACs are presumed to be proportional to M_BH because Γ ≃ 0.1 is required to activate the limit–cycle oscillations. Therefore, when a BH mass increases by a factor of 10, a STAC luminosity increases only by a factor of 5 but an AGN luminosity increases by a factor of 10. In Figure 2, we show expected STAC and host AGN luminosity evolutions scaled with the properties of CSS100217. STACs by the BH disk wind start to be "buried" in AGNs with the increasing AGN BH mass. On the contrary, they become easier to detect in low-mass AGNs, even though the STAC luminosities are also expected to decrease with smaller BH masses. PS16dtm is observed in an AGN whose host g-band magnitude is −19.1 mag and peak g-band magnitude is −22 mag. These properties roughly matches the above scaling based on CSS100217.

Zoom In Zoom Out Reset image size

CSS100217 has been monitored even after the major burst observed in 2010, and its light curve is available in the Catalina Real-Time Transient Survey website.⁴ Interestingly, the post-burst brightness of the CSS100217 host is about 0.5 mag fainter than the pre-burst brightness. The brightness difference in the host is likely associated with changes in the AGN accretion disk. If CSS100217 was a phenomenon that is not related to the central AGN activities, we do not naturally expect changes in the accretion disk shortly after the burst. Thus, the observed luminosity difference in the AGN hosting CSS100217 suggests that the transient is related to the AGN activity. In addition, the post-burst brightness is gradually recovering to the pre-burst brightness that matches a prediction of the limit–cycle oscillation (Watarai & Mineshige 2003). If the limit–cycle oscillation is actually triggering STACs, CSS100217 may become bright again in coming years or decades.