THE MULTI-LAYER VARIABLE ABSORBERS IN NGC 1365 REVEALED BY XMM-NEWTON AND NuSTAR 

We reduced the XMM-Newton data for all four observations using the Science Analysis System (v13.0.0) following the procedure detailed in the online guide and Walton et al. (2014). We processed the data using EPPROC and EMPROC for the EPIC-pn (Strüder et al. 2001) and EPIC-MOS (Turner et al. 2001) data, respectively. Spectra and light curves were extracted from circular source and background regions with 40'' and 50'' radius for the pn and MOS, respectively. Response and ancillary response matrices were generated using the FTOOLs RMFGEN and ARFGEN.

During the last portion of observation 3 the source reached a peak in the flux level sufficient for pile-up to be a concern. We found some evidence for pileup in the MOS below 10 keV, and therefore excluded the central 8'' for the pn and 10'' for the MOS for observation 3 and interval P12 (the last interval of observation 3). Note that we did not include data from observation 3 past 110 ks where a background flare occurred, possibly contaminating the data (see Figure 1).

We reduced data from both NuSTAR modules, FPMA and FPMB (Harrison et al. 2013), using the standard pipeline in the NuSTAR Data Analysis Software v1.1.1. Instrumental responses were taken from the NuSTAR CALDB v20130315. The unfiltered event files were cleaned with the standard depth correction, which significantly reduces the internal background at high energies, and SAA passages were excluded from our analysis. For both modules we extracted spectra and light curves from a 100'' circular source region and a 100'' background region on the same chip as the source. We grouped the NuSTAR spectra with a minimum of 25 counts per bin.

Initially we fit each of the four observations separately in the 3–70 keV range with an absorbed power-law. This fit was universally poor $({{\chi }^{2}}$/degrees of freedom (dof) = 2832/1047, 6926/1145, 4551/1137, and 3968/1113 for observations 1, 2, 3, and 4, respectively) and showed strong residuals in the Fe K bandpass (∼5–9 keV) as well as broad residuals in the 20–30 keV range (see Figure 1 of Risaliti et al. 2013; Walton et al. 2014 for which the continuum and reflection modeling of these data has been done in great detail). In order to model the apparent neutral Fe emission lines we added a neutral reflection component. We used the xillver xspec model, which includes Fe K emission lines and a Compton reflection hump from a disk geometry (García & Kallman 2010). There was significant improvement in the fit statistic for each observation (${{\chi }^{2}}/$dof = 2106/1044, 5727/1142, 3630/1134, and 3059/1110). Note that this model choice was not a significant improvement over a phenomenological modeling using multiple Gaussian components and a neutral Compton hump such as from pexrav, but we prefer a self-consistent physical model whenever possible.

Strong negative residuals in the Fe K band remained, necessitating the addition of four absorption lines from highly ionized species of Fe in a high velocity outflow (Risaliti et al. 2005; Brenneman et al. 2013). We tied the velocities and line widths of the four lines, assuming intrinsic line energies of 6.70, 6.97, 7.88, and 8.27 keV. This greatly improved the fit statistics (${{\chi }^{2}}/$dof = 1341/1039, 2325/1137, 2197/1129, and 1784.1/1105), though broad residuals still remained near the Fe Kα line and above 10 keV. Though there are several models available which might be used at this juncture, we have elected to use relconv ×xillver to model relativistic reflection from the inner parts of an ionized accretion disk. Inclination angle, ionization state, Fe abundance, black hole spin, and normalization were left as free parameters in our initial fitting. This model provided a good fit to the data in all four observations (${{\chi }^{2}}/$dof = 1087/1035, 1445/1133, 1283/1125, and 1300/1101, for the four observations, respectively). Parameter values, detailed justification of the model, and physical interpretations for this analysis can be found in Walton et al. (2014).

Those parameters that we expect to remain constant over the timescale of our observations were frozen at their average values: black hole spin (0.98), disk inclination (63°), Fe abundance (4.7). Note that modeling the distant reflection with a torus model can lead to a much lower measurement of the Fe abundance for the distant reflector; however, this does not change our primary results so for simplicity we adhere to the model presented in Walton et al. (2014), tying the Fe abundance between the two reflectors. We also froze the following parameters that showed no evidence for variability over the four observations: ionization of the inner disk (logξ = 1.8), and disk emissivity index (6.75). Note that we do not see evidence for a high energy cutoff in this source and it was therefore not included in our final model.

In order to fully characterize the partial-covering absorption in this source, we must extend our spectral analysis down to lower energies. It was immediately evident that observations 2 and 3, and possibly observation 4, were partially unobscured in the 1–3 keV range. We therefore replaced the full-covering absorber in our model with a partial-covering absorber, though for observation 1 the covering fraction remained pegged at 1. Additionally, we chose to analyze the spectrum down to 0.4 keV in order to model the soft emission from diffuse plasma in the region so that we could be confident that our measurements of the absorption were not influenced by this component.

The extended plasma was previously studied in detail by Wang et al. (2009) using Chandra gratings data and by Guainazzi et al. (2009) using XMM-Newton RGS data. These analyses determined that the plasma was likely a combination of thermal and photoionized plasma. When the AGN is in an absorbed state the extended plasma dominates the soft X-ray flux and it is expected to remain essentially constant due to its spatial extent. For our modeling we used a phenomenological double apec component (i.e., two-temperature collisionally ionized gas) with temperatures of 0.3 and ∼0.8 keV, and five additional Gaussian components to model emission line complexes from photoionized gas at 0.50, 0.85, 1.03, 1.24, and 2.74 keV, similar to Brenneman et al. (2013). These components are shown in red in Figure 2. We determined the normalizations of these components using data from observation 1 only because it is the only observation that is fully covered and therefore allowed for the best determination of the plasma parameters. Since we do not expect the extended plasma to undergo any changes over the months between our observations we froze these parameters to those measured in observation 1. We included an additional soft power-law component in the modeling to account for differences in flux level below 2 keV for the four observations. The photon index was tied to that of the primary continuum power-law and in all fits was found to have a normalization around ∼2% that of the primary power-law, as expected for a scattered continuum component (e.g., Turner et al. 1997; Guainazzi et al. 2005; Eguechi et al. 2009). Compared to the flux of the combined plasma components, ${{F}_{0.5-2}}=3.6\times {{10}^{-13}}$ erg cm⁻² s⁻¹, the scattered power-law 0.5–2 keV flux values were $2.7\times {{10}^{-13}}$ erg cm⁻² s⁻¹, $4.4\times {{10}^{-13}}$ erg cm⁻² s⁻¹, $7.8\times {{10}^{-13}}$ erg cm⁻² s⁻¹, and $3.5\times {{10}^{-13}}$ erg cm⁻² s⁻¹for the four observations, respectively.

Zoom In Zoom Out Reset image size

We applied this model to the other three observations and discovered the need for an additional layer of full-covering absorption, with an improvement in fit of ${\Delta }{{\chi }^{2}}$/dof = 6000/1 for observation 2, ${\Delta }{{\chi }^{2}}$/dof = 1500/1 for observation 3, and ${\Delta }{{\chi }^{2}}$/dof = 1050/1 for observation 4, with a column density of ${{N}_{{\rm H}}}$ $\;\sim \;$ 1 $\;\times \;{{10}^{22}}$ cm⁻² for all three observations. We will refer to this absorbing layer as the "full-covering low column" absorber for the remainder of the paper. Given the consistency of this component over the last three observations it seems reasonable to assume that it is also present in observation 1, but it is completely degenerate with the higher column density absorber which is fully covering during that observation. We have included the component in our modeling of observation 1 with a fixed column density of 1.0 $\;\times \;{{10}^{22}}$ cm⁻²in the hopes that this gives us a more accurate measurement of the column density of the partial-covering absorber layer during that observation.

We also found unexpected variability below 1 keV which was not fit by the changing absorption or the scattered power-law. The unidentified excess appeared in the less absorbed observations, strongest in observations 2 and 3, while completely absent in observation 1. None of the usual components that might account for this increase in flux below 1 keV were able to fit the data (i.e., a blackbody or power-law soft excess, changes in the fully covering absorber, or reflection from ionized material). We therefore fit the excess with an additional phenomenological Gaussian at 0.68 keV with a width of $\sigma \sim$ 200 eV. Including this component resulted in an improvement in the fit of ${\Delta }{{\chi }^{2}}$/dof = 200/3 for observation 3 and ${\Delta }{{\chi }^{2}}$/dof = 30/3 for observation 2, with a null hypothesis probability of $4\times {{10}^{-4}}$. It was not significant to include the component in observations 1 or 4, both of which result in a normalization of the component consistent with 0. We investigate the source of this feature in Section 3.6. Residuals to models excluding the full-covering low column absorber and the phenomenological Gaussian are shown for each observation in Figure 2, panels (c) and (d), respectively.

Our final model consisted of two collisional plasma components plus five Gaussian emission lines to model the extended plasma, a scattered power-law, the phenomenological Gaussian at 0.68 keV, a full-covering neutral absorber (${{N}_{{\rm H}}}$ ∼1 $\;\times \;{{10}^{22}}$ cm⁻²), a partial covering neutral absorber, four absorption lines from highly ionized species of Fe in a high velocity outflow, a continuum power-law, relativistic disk reflection, and cold distant reflection. The final form of the model in xspec is: apec[×2] + zGauss[×5] + scattered power-law + Gauss + zphabs × zpcfabs × gauabs[×4] × (power-law + relconv × xillver) + xillver.

For consistency checks, we fit all four observations simultaneously, exploring which parameters were consistent across the observations. We find that the full-covering absorber was very steady over all three observations where it was measurable. In observation 1 this layer is completely degenerate with the high column density layer of absorption. We therefore included this layer in all fits to observation 1 with a column density frozen at 1.0 $\;\times \;{{10}^{22}}$ cm⁻². Figure 2 shows each of the four observations with final best-fit model components in panel (a) and residuals to the best-fit model in panel (e). Parameters are listed in Table 2.

Table 2.  Broadband Model Parameters

Interval	Unabsorbed	Photon	PC Column	Covering	FC Column	Scattered	Soft Gaussian	Relativistic	Distant	${{\chi }^{2}}$/dof
	Continuum	Index	Density (${{N}_{{\rm H}}}$)	Fraction	Density	Power-Law	at 0.68 keV	Reflection	Reflection
	F2-10a	(Γ)	(1022 cm−2)	(f)	(1022 cm−2)	F$_{0.5-2}$ a/F2–10a	Norm (10−5)	Norm (10−6)	Norm (10−6)
Obs. 1	16.8 ± 0.3	1.84 ± 0.02	22.4 ± 0.2	1.00$^{{\rm b}}$ ($\gt 0.99$)	1.0$^{{\rm c}}$	0.27 / 0.38 ± 0.01	⋯	1.0 ± 0.0	3.0 ± 0.2	1627/1243
Obs. 2	27.3 ± 0.6	2.01 ± 0.01	8.4 ± 0.2	0.76 ± 0.01	1.4 ± 0.1	0.44 / 0.52 ± 0.03	4.3 ± 1.3	2.0 ± 0.1	3.4 ± 0.4	2131/1353
Obs. 3	23.0 ± 0.5	2.04 ± 0.01	5.8 ± 0.4	0.49 ± 0.01	1.1 ± 0.1	0.78 / 0.75 ± 0.05	20.5 ± 2.4	2.1 ± 0.1	3.2 ± 0.5	1877/1347
Obs. 4	21.6 ± 0.5	1.92 ± 0.01	11.3 ± 0.2	0.97 ± 0.01	1.0 ± 0.2	0.35 / 0.46 ± 0.02	0.2$^{{\rm b}}$ ($\lt 1.0$)	1.5 ± 0.1	3.3 ± 0.3	1888/1323
P1	16.0 ± 0.7	1.87 ± 0.02	24.0 ± 0.5	1.00$^{{\rm b}}$ ($\gt 0.99$)	1.0$^{{\rm c}}$	0.26 / 0.37 ± 0.01	⋯	1.1 ± 0.1	2.4 ± 0.4	953/804
P2	21.2 ± 1.1	1.90 ± 0.03	21.3 ± 0.5	1.00$^{{\rm b}}$ ($\gt 0.99$)	1.0$^{{\rm c}}$	0.34 / 0.46 ± 0.02	⋯	1.3 ± 0.2	3.5 ± 0.6	918/708
P3	17.4 ± 1.2	1.89 ± 0.03	25.6 ± 0.8	1.00$^{{\rm b}}$ ($\gt 0.99$)	1.0$^{{\rm c}}$	0.26 / 0.36 ± 0.02	⋯	1.2 ± 0.2	2.5 ± 0.6	641/544
P4	17.5 ± 1.3	1.91 ± 0.04	22.6 ± 0.7	1.00$^{{\rm b}}$ ($\gt 0.99$)	1.0$^{{\rm c}}$	0.26 / 0.35 ± 0.02	⋯	1.2 ± 0.2	2.6 ± 0.7	569/515
P5	23.9 ± 1.2	2.05 ± 0.03	8.8 ± 0.4	0.82 ± 0.01	1.4$^{{\rm c}}$	0.39 / 0.48 ± 0.04	1.5 ± 0.6	2.1 ± 0.2	4.7 ± 0.8	806/690
P6	31.2 ± 1.1	2.03 ± 0.02	8.2 ± 0.3	0.80 ± 0.01	1.4$^{{\rm c}}$	0.36 / 0.51 ± 0.03	3.1 ± 0.5	2.6 ± 0.2	5.7 ± 0.8	1180/906
P7	23.1 ± 0.9	1.99 ± 0.02	7.8 ± 0.3	0.78 ± 0.01	1.4$^{{\rm c}}$	0.39 / 0.54 ± 0.03	1.2 ± 0.5	1.9 ± 0.2	4.5 ± 0.7	1100/883
P8	27.5 ± 1.1	1.97 ± 0.02	7.8 ± 0.4	0.70 ± 0.01	1.4$^{{\rm c}}$	0.45 / 0.62 ± 0.04	2.8 ± 0.6	1.7 ± 0.2	3.7 ± 0.8	1067/873
P9	16.4 ± 1.3	1.93 ± 0.04	9.1 ± 0.6	0.90 ± 0.01	1.1$^{{\rm c}}$	0.49 / 0.69 ± 0.08	1.3 ± 1.7	1.8 ± 0.2	4.9 ± 1.0	529/536
P10	23.3 ± 1.1	2.00 ± 0.03	5.7 ± 0.4	0.75 ± 0.01	1.1$^{{\rm c}}$	0.54 / 0.61 ± 0.06	10.7 ± 2.0	2.3 ± 0.3	3.8 ± 0.9	1102/774
P11	24.0 ± 1.1	2.04 ± 0.03	4.2 ± 0.4	0.59 ± 0.01	1.1$^{{\rm c}}$	0.58 / 0.56 ± 0.06	15.1 ± 2.2	2.2 ± 0.2	6.6 ± 0.9	1102/768
P12	38.9 ± 1.3	2.07 ± 0.03	4.3 ± 0.6	0.29 ± 0.03	1.1$^{{\rm c}}$	1.00 / 1.08 ± 0.10	31.5 ± 3.7	2.6 ± 0.3	4.3 ± 1.3	1011/874
P13	40.7 ± 1.3	2.07 ± 0.03	7.2 ± 0.2	1.00$^{{\rm b}}$ ($\gt 0.99$)	1.0$^{{\rm c}}$	0.37 / 0.39 ± 0.03	0.2$^{{\rm b}}$ ($\lt 1.0$)	2.0 ± 0.3	2.5 ± 1.2	908/719
P14	18.9 ± 1.1	1.93 ± 0.04	11.7 ± 0.5	0.97 ± 0.01	1.0$^{{\rm c}}$	0.33 / 0.43 ± 0.05	0.3$^{{\rm b}}$ ($\lt 1.0$)	1.6 ± 0.2	3.9 ± 0.9	556/592
P15	20.9 ± 0.8	1.90 ± 0.03	12.9 ± 0.3	0.98 ± 0.01	1.0$^{{\rm c}}$	0.31 / 0.42 ± 0.03	2.0 ± 0.6	1.3 ± 0.1	3.4 ± 0.6	917/816
P16	20.8 ± 0.8	1.93 ± 0.02	16.9 ± 0.4	0.99 ± 0.01	1.0$^{{\rm c}}$	0.34 / 0.44 ± 0.03	1.0 ± 0.6	1.4 ± 0.1	2.3 ± 0.5	1089/950

Notes. Best-fit parameters for the four observations and the 16 intervals, four per observation. PC and FC stand for partial-covering and full-covering, respectively. The intrinsic continuum fluxes and soft scattered power-law fluxes were determined using the PEGPWRLW model in xspec.

^aFlux is in units of 10⁻¹² erg cm⁻² s⁻¹. ^bIndicates a frozen parameter. ^cIndicates a pegged parameter.

Download table as:  ASCIITypeset image

We then fit the 16 intervals independently in the 0.3–70 keV range using our final model. In order to reduce degeneracy we froze the column density of the full-covering constant absorber to the time-averaged value for each observation and to 1.0 $\;\times \;{{10}^{22}}$ cm⁻² for all intervals of observation 1. Best-fit parameters for all 16 intervals are listed in Table 2 and the evolution of the parameters with the most interesting variability (${{N}_{{\rm H}}}$, f, Γ, and intrinsic flux) is shown in Figure 3. Values for the Fe K ionized wind absorption were also left free, but did not vary significantly over the observations (i.e., they varied by less than the measured error bars).

Zoom In Zoom Out Reset image size

In observation 3 we saw a dramatic drop in the absorption by the partial-covering absorber. The line-of-sight covering fraction (f) fell from 0.9 to 0.3 over the course of the observation, almost completely uncovering the source. The column density also dropped significantly (nearly a factor of 2) to 5 $\;\times \;{{10}^{22}}$ cm⁻². Figure 3 shows the rapid decrease in both parameters with a corresponding increase in the unabsorbed power-law flux. In order to get a clearer picture of the evolution of these parameters we performed additional time-resolved analysis on observation 3, breaking each interval in half and fitting the 8 subintervals. We used the final model as described in Section 3.3 with the full-covering absorber column density held fixed at the time-averaged value of 1.1 $\;\times \;{{10}^{22}}$ cm⁻².

Figure 4 shows the evolution of the column density, covering fraction, photon index, and intrinsic power-law flux for observation 3. The rapid decrease in both ${{N}_{{\rm H}}}$ and f could plausibly be due to parameter degeneracy; however, the error bars seem too small for this to be the case. A contour plot of ${{N}_{{\rm H}}}$versus f for the eight subintervals shown in Figure 5 verifies that little to no parameter degeneracy is apparent between ${{N}_{{\rm H}}}$and f. While there is an expected degeneracy between the column density and flux of the source shown in Figure 6, it is generally small compared to the changes in the parameters and is in the opposite sense to the observed evolution.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The decreasing column density and covering fraction both seem to have an inverse relationship with the increasing intrinsic flux. The photon index is relatively stable over the same period. This is unusual behavior for absorption variability caused by clumps moving into and out of the line-of-sight. Column density and covering fraction do not typically vary in lock step with one another and we would not expect either parameter to correlate with intrinsic flux. This may be indicative of an evolution in the absorbing material with luminosity, such as in the wind absorber scenario proposed by Connolly et al. (2014; discussed in Section 4.1). We will refer to this layer of absorption (seen primarily in observation 3) as the "patchy partial-covering" absorber to distinguish it from the sometimes partial-covering "high column density" absorber, which dominates in observations 1 and 4.

Observation 4 shows a rapid increase in column density during the first 32 ks (intervals P13 and P14), with little to no change in the covering fraction. We broke the first 32 ks down into eight 4 ks subintervals and fit just the PN+FPMA data in the 1–40 keV range for computational brevity, using the final model as described in Section 3.3. Parameters for the full-covering absorber, scattered power-law, highly ionized wind absorber, and reflection were also held constant. We find an evolution of the continuum and absorption parameters shown in Figure 7. The column density peaked around 20–24 ks, with a seemingly symmetrical profile strongly indicative of a clump of material passing through the line-of-sight. Though degeneracy between the column density and intrinsic flux is again present in these fits, the lack of change in the covering fraction makes this behavior very different from that seen in observation 3. Again, the level of degeneracy is less than the observed changes (see Figure 8).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

If this is indeed an occultation event then it is consistent in duration with events seen previously in this source and thought to be due to BLR clouds passing through the line-of-sight (Maiolino et al. 2010). Using their estimate for ${{R}_{{\rm BLR}}}$ of ∼10¹⁶ cm, and assuming a black hole mass of $2\times {{10}^{6}}$ ${{M}_{\odot }}$ we estimate the velocity of the cloud in a circular orbit to be ∼1600 km s⁻¹. Following the analysis of Rivers et al. (2011) for a smooth, symmetrical occultation event assuming a circular orbit around the black hole, we fit the absorption profile with a solid sphere, a beta profile ($\rho \propto {{r}^{\beta }}$) and a sphere of linear density profile ($\rho \propto {{r}^{-1}}$), shown in Figure 9. We find a cloud size of ∼4 × 10¹² cm (∼10 R_g) with a central density of ∼3 × 10¹⁰ cm⁻³ for the linearly decreasing density sphere. These numbers are consistent with the physical properties inferred by Maiolino et al. (2010) for the cloud cores, although we do not see evidence of the same cometary structure inferred in that work (${{N}_{{\rm H}}}$ rises again after the first 32 ks of observation 4 and the covering factor is relatively stable over the course of the event).

Zoom In Zoom Out Reset image size

From Table 2 we see that the phenomenological Gaussian component at 0.68 keV peaks strongly in observation 3. The strength of this feature seems to be anti-correlated with the covering fraction, not showing up at all in observation 1, which is fully covered, and only very weakly in observation 4 where the covering fraction is 0.97–1.0. This component is therefore clearly not associated with the extended plasma and likely arises within the radius of the variable high column density partial-covering absorber.

Braito et al. (2014) have analyzed the RGS spectrum of observation 3 and found evidence for an increasing ionization in their low column absorber over the course of the observation. What we are seeing in the MOS/PN may be due to further leakage of the power-law below 1 keV due to increasing ionization of the low column full-covering absorber as the source increases in luminosity. The material would have to be quite close to the central source since it is directly correlated with the observed increase in intrinsic power-law flux. With time-resolved fitting on timescales of around 20 ks this would mean the material is at a distance of ≲10¹⁶ cm, which is roughly the radius of the BLR and is consistent with being inside the radius of the high column density partial-covering absorber.

Another possibility is the uncovering of actual emission lines such as from O viii Lyα at 0.654 keV. This could be from highly ionized material that is very close to the central source and only visible when the full-covering low column absorber becomes ionized enough to be semi-transparent at energies below 1 keV. Braito et al. (2014) also noticed that some broader soft X-ray lines started to emerge during the last part of observation 3, for instance from Mg xi and Mg xii, which were much broader than the usual distant narrow line emission, and which seemed to have P Cygni like profiles. These lines could be associated with a disk wind very close to the central source. It is worth noting that a visual inspection of the RGS residuals in Figure 4 of Braito et al. (2014) reveals systematically high residuals around 0.65–0.7 keV.

The uncovering of the continuum by the high column partial covering absorber seen during 2012 December (observation 2) and 2013 January (observation 3) is an unusual event, particularly the extreme uncovering witnessed in the January observation, where both the column density and covering fraction dropped dramatically. This is in contrast to earlier observations of fast absorption variability due to "comet-like" BLR clouds, which show an increase in the covering fraction as the column density declines, indicative of a denser core leading a diffuse tail (Maiolino et al. 2010). The simultaneous drop in both the column density and the covering fraction in our third observation is inconsistent with this kind of event.

In order to discover the origin of the variability in the absorber we looked for relationships between the parameters that showed the greatest variations. Figure 10 shows the relationships between ${{N}_{{\rm H}}}$ and intrinsic flux, and between ${{N}_{{\rm H}}}$ and f. The flux plotted here is the unabsorbed power-law 2–10 keV flux which corresponds directly to the intrinsic luminosity of the source. For the 16 intervals, ${{N}_{{\rm H}}}$ appears to be correlated with flux at the 99.5% level and with the covering fraction at the 99.9% level by the Pearson correlation test. These trends are also seen for the subintervals at around the ∼95% confidence level (although note the small sample sizes). There is some indication that on longer timescales the photon index softens as the source brightens (correlated at the 99.9% level for the 16 intervals). However, this correlation is not seen on shorter timescales (for the subintervals of observations 3 and 4).

Zoom In Zoom Out Reset image size

The relationship between ${{N}_{{\rm H}}}$ and flux in observation 3 is consistent with the anti-correlation noticed by Connolly et al. (2014) from analyzing Swift monitoring data of NGC 1365. Our data do not show a simple linear correlation between these parameters. It seems clear, however, that when the X-ray source is brighter ($\gtrsim 2\times {{10}^{-11}}$ erg cm⁻² s⁻¹) the column density is much lower ($\lesssim 1\times {{10}^{23}}$ cm⁻²). For lower values of flux ($\lesssim 2\times {{10}^{-11}}$ erg cm⁻² s⁻¹), the column density is much higher ($\gtrsim 1\times {{10}^{23}}$ cm⁻²). Connolly et al. (2014) suggested that this could be due to a wind absorber which launches from further out when the source flux goes up.

This patchy wind absorber is only clearly seen in observation 3. In the other three observations a high column, sometimes partial-covering absorber dominates. This high column absorber is likely associated with the BLR (as evidence by the occultation event in observation 4 and those seen previously) as well as with the torus. The timescale of the uncovering between observations 2, 3, and 4 is weeks to months rather than hours, indicating that it is either due to a gap in the torus clouds or to a global attenuation of material. Since we do not see any other evidence of a temporary drop in overall accretion rate, we favor the former scenario.

One other possibility is that the drop in covering fraction in observation 3 could be explained by a cloud moving out of the line-of-sight. If the cloud were decreasing in size, such as with a drawn out filamentary tail, then instead of the comet-like increase in covering fraction we would see a shrinking covering fraction as less and less of the tail covered the source. This would match our observed absorption parameter evolution, though it does not explain the correlation between the absorber parameters and intrinsic flux seen in observation 3. We therefore reject this hypothesis.

In this data set we see changes in the absorbers on multiple timescales. We see an initial drop in total column density over the course of ∼5 months from 22 to 8$\;\times \;{{10}^{22}}$ cm⁻². Then in observation 3 ${{N}_{{\rm H}}}$ goes from 11 to 4 $\;\times \;{{10}^{22}}$ cm⁻² in ∼130 ks while the covering fraction drops to 0.45. And in the first 20–24 ks of observation 4 ${{N}_{{\rm H}}}$ rises from 7 to 14$\;\times \;{{10}^{22}}$ cm⁻². This last value is the most rapid change and can be used to place constraints on the size of the X-ray emitting region. Since the occultation in observation 4 is nearly fully covering, we can infer that the X-ray emitting region is no larger than the size of the occulting cloud: $\lesssim 10\;{{R}_{{\rm g}}}$.

This agrees with previous estimates of the compactness of the X-ray corona from the measured relativistic reflection, observed reverberation and BLR occultation events (Maiolino et al. 2010; Kara et al. 2015; Walton et al. 2014). However, it presents a conundrum when we consider the extremely low covering fraction seen in observation 3. To see such a clear drop in covering fraction would require that the absorber be either very close to the source or made up of clumps/filaments that are smaller than the size of the X-ray emitting region. Since the drop in covering fraction occurs slowly over the entire observation and taking into account the low ionization state of the absorbers, the latter scenario seems more plausible, possibly due to a patchy disk wind.

Between 2012 July and 2013 February, NuSTAR and XMM-Newton performed four long-look joint observations of NGC 1365. We have analyzed the variable absorption seen in these observations in order to characterize the geometry of the absorbing material. Fortuitously, two of the observations caught the source in an unusually low absorption state, revealing additional complexity that had previously been hidden. This "peak between the clouds" allowed us to see past the typical torus/BLR clouds, which tend to have column densities of around ∼10²³ cm⁻², uncovering a patchy absorber with a variable column around ∼10²² cm⁻² and a measured covering fraction of f = 0.3–0.9. Additionally, we found that this patchy absorber seems to respond to the intrinsic source flux, with the column density and covering fraction dropping as the source grows brighter. This could be due to a high luminosity event pushing an absorbing wind out to a large radius where the covering fraction and effective ${{N}_{{\rm H}}}$ both drop dramatically. This latter theory is espoused by Connolly et al. (2014), who noticed an anti-correlation between ${{N}_{{\rm H}}}$ and luminosity in NGC 1365, a trend which our data confirms.

We also find evidence of an additional constant absorber with a low column density of 1 × 10²² cm⁻², the geometrical location of which is still unclear. The ionized wind absorbers seen in this source (Risaliti et al. 2005; Braito et al. 2014) most likely reside closer to the central source than the three layers of neutral absorbers we have characterized in this work.

A short occultation event in 2013 February was observed, likely due to a BLR clump passing through the line-of-sight. We estimate a clump size of ∼4 × 10¹² cm with a central density of ∼3 × 10¹⁰ cm⁻³. From this we also infer a small X-ray corona with a linear dimension of only a few R$_{{\rm g}}$.