Previous soft (energy E < 10 keV) X-ray observations of NGC 1365 have consistently seen Compton-thin clouds that periodically cross the line of sight, absorbing the 3–5-keV flux, and changing the continuum on timescales of hours13,14. On timescales of a day, the spectrum has been seen to change from a state where a non-thermal continuum (associated with a tenuous coronal plasma) is viewed through a moderate, variably obscuring column to a state where the emission is completely dominated by Compton reflection off a dense material15. Based on the black-hole mass, and assuming velocities of a few thousand kilometres per second for the obscuring clouds, the inferred size of the X-ray source is a few gravitational radii, R g = GM/c2, where G is the gravitational constant, M is the black hole’s mass, and c is the speed of light)13.

The orbiting X-ray observatory XMM-Newton16 (X-ray Multi-Mirror Mission—Newton) and the space-based X-ray telescope NuSTAR17 (the Nuclear Spectroscopic Telescope Array) simultaneously observed the galaxy NGC 1365 with an exposure of around 130 ks from 25 to 27 July 2012 Universal time. These broadband X-ray data (0.5–79 keV) have sufficient statistics to study spectral variability on the relevant few-hour timescales. We find the low-energy (E < 3 keV) component to be constant, and dominated by thermal emission from an optically thin plasma (see Supplementary Information). Previous imaging observations from the space telescope Chandra X-ray Observatory show that this soft component is spatially extended, and physically distinct from the variable emission18. We restricted our analysis to the hard components observed above 3 keV.

The 3–79-keV spectrum shows strong emission features typical of relativistic disk reflection (Fig. 1). However, previous work has demonstrated that the 5–7 keV distortion can be also be explained by varying absorbers along the line of sight partially covering the source. These have been seen before in NGC 1365, and they strongly affect the spectral shape below 10 keV (refs 10 and 13).

Figure 1: The broadband 3–79-keV X-ray spectrum of NGC 1365. An absorbed power law has been fitted to the usually featureless continuum intervals 3–4 keV, 7–10 keV and 50–80 keV. The ratios relative to this fit are shown for NuSTAR modules A and B (black and red data in main panel), and for the XMM EPIC/PN instrument (inset). Three prominent features are apparent: an asymmetric excess between 5 and 7 keV, a broad prominent excess between 10 and 79 keV, and a series of absorption lines between 6.7 and 8.3 keV. These absorption lines are due to absorption by a highly ionized plasma26, which we include in all the spectral models discussed below. The two excesses are typical signatures of relativistically blurred emission. Quoted errors refer to a 90% confidence level. Full size image Download PowerPoint slide

In the 3–10-keV XMM band, the flux ratio F(3–5 keV)/F(6–10 keV) shows strong time variations, while the F(7–15 keV)/F(15–80 keV) ratio in the NuSTAR band is constant over the whole observation (Fig. 2), suggesting that the soft variation is due to variable absorption (see Supplementary Information). This enabled us to perform a time-resolved analysis and to decompose the different spectral components accurately. We broke the observation into four intervals of similar hardness and analysed each simultaneously. We considered two models to explain the spectrum and its variability, one containing a relativistic reflection component from the inner accretion disk and a variable, partially covering absorber, and one employing multiple variable absorbers instead. In addition, both models include neutral reflection from a constant, distant screen of gas to reproduce the unresolved, narrow iron emission line at 6.4 keV.

Figure 2: X-ray spectral variability of NGC 1365. a, NuSTAR F(7–15 keV)/F(15–80 keV) softness ratio light curve. b, XMM F(3–5 keV)/F(6–10 keV) softness ratio light curve. The vertical lines delimit four time intervals with significantly different average values of the softness ratio, from which four different spectra have been extracted and analysed simultaneously. These variations are due to absorption variability (see Supplementary Information). The changes of softness ratio within each interval reveal that the absorber has a complex structure; however, its detailed variability on timescales shorter than around 1,000 s is not relevant in the determination of the physical parameters of the other spectral components. Error bars are one standard deviation. Full size image Download PowerPoint slide

To illustrate the power of the broad energy range in determining model parameters, we first fitted the XMM data alone to both models. With XMM data alone, we could not distinguish between these scenarios, the broadened Fe-line being consistent with both. However, extrapolation of the two models to the 10–80 keV range broke the degeneracy: the absorption-only model failed to reproduce the new NuSTAR data by a large factor, while the model including the relativistic component reproduced the observed spectrum within around 5%, without even re-fitting the data (Fig. 3).

Figure 3: Comparison between the relativistic reflection model and the multiple absorber model. a, XMM and NuSTAR spectral data and models for one of the four time intervals in Fig. 2. The two models contain a relativistic reflection component plus variable partial covering (model 1; red dashed line), and a double partial covering (model 2; blue continuous line). Both models have been fitted to the data only below 10 keV, and reproduce the lower-energy observations equally well. However, the models strongly deviate at higher energies. b, Data-to-model ratio for the double partial covering (model 2; blue) and relativistic reflection plus variable absorber (model 1; red) models. The best-fitting partial-covering model using XMM data alone has two absorbers with N H values of 5 × 1022 cm−2 and 3 × 1023 cm−2; these have little influence on the spectrum above 10 keV. Instead, the relativistic reflection component reproducing the broad emission feature below 10 keV also includes a strong Compton reflection component above 10 keV. All errors are at a 90% confidence level. Full size image Download PowerPoint slide

We next considered the possibility that the shape of the hard spectrum arises not from disk reflection, but from an intervening screen of dense absorbing material that can scatter a fraction of the light, but primarily affects the high-energy continuum through absorption. A high equivalent hydrogen column density N H of this gas (far in excess of that found by the fits to XMM only) is required to affect energies above 10 keV. We fitted the broad 3–79-keV band allowing the absorber parameters in model 2 to vary relative to those found with XMM alone. The resulting fit drove the column of the second absorber to N H ≈ 5 × 1024 cm−2, which reproduces the shape of the hard continuum, but cannot reproduce the spectral variability below 10 keV (see Supplementary Information). We next added a third absorber to model 2, and allowed the parameters of all three to vary. By setting the third absorber to have N H = 4–6 × 1024 cm−2 and to cover at least 40% of the source, we achieved a formally acceptable fit (reduced χ2 of 1.02). However, this solution was rejected compared with the reflection model: an F-test performed on a ‘merged model’ that includes all the components of the two individual models showed that the probability that the model without relativistic reflection (model 2) is preferred is 8 × 10−5.

The three-zone model could also be ruled out on the basis of physical implications that are completely inconsistent with optical, infrared and X-ray observations. These arose when we considered the strong Compton scattering and reprocessing inevitable with the high-N H absorber. For low-N H (< 1024 cm−2) absorbers, the effect of continuum reprocessing can be ignored (and in fact is not included in our models). However, the high-density absorber required to fit the E > 10-keV data would strongly affect the predicted line and continuum emission. How strongly depends on the geometry.

We considered two extreme bounding possibilities. First, if the solid angle covered is small, the reprocessing into X-ray line emission or infrared flux will be negligible. But, in this case the effect of Compton scattering on the direct emission will strongly (by a factor exp(τ C ) ≈ 20–80, where τ C is the Compton optical depth) attenuate the intrinsic luminosity. Correcting for this, the implied total luminosity of 0.1–100 keV becomes implausibly high (X-ray luminosity of 1–3 × 1044 erg s−1, that is, 30–100% of the Eddington limit). This is completely incompatible with other luminosity indicators, such as the optical [O iii] 5,007 Å line19. Second, if the Compton-thick absorber covers a large fraction of the solid angle as seen from the source, there will be strong reprocessing, with two observable effects: a large fraction of the intrinsic luminosity is re-emitted in the infrared part of the spectrum, which is ruled out by high-resolution infrared observations20, and a strong narrow iron emission line21 with equivalent width >600 eV should be present, which is ruled out by our measurement: an observed equivalent width of approximately 60 eV (see Supplementary Information).

The observed source variability and unprecedented broad-band spectroscopy enable us to conclude unambiguously that reflection of a hard continuum off the inner accretion disk edge is producing the neutral Fe-line distortion and strong high-energy Compton reflection continuum. The flux of the relativistic reflection component is higher than the primary continuum at its peak around 30 keV (see Supplementary Information), implying a strong enhancement of the reflection efficiency due to relativistic distortion22 (the highest value with standard disk reflection is about 30%; ref. 23). This high efficiency is consistent with the disk and spin parameters presented below, implying that most of the reflection arises from the inner part of the accretion disk, close to the innermost stable circular orbit.

From the relativistic disk reflection, we can determine the black-hole spin parameter. Allowing all other model parameters (such as disk inclination, ionization state and emissivity profile) to vary, we found the minimum spin to be a* ≥ 0.84 at 90% confidence (Fig. 4), where a* is the dimensionless spin parameter, a* = Jc/GM2 (where J is the black hole’s angular momentum), equivalent to an innermost disk edge at ≤2.5 gravitational radii. These results are consistent with the ones from previous observations of NGC 1365 (ref. 10) from the X-ray astronomy satellite Suzaku. In that analysis, however, the relativistic model was assumed to be valid, and a complete comparison with the absorption-only scenario was not attempted, owing to the low statistics, especially at high energies (see Supplementary Information for further details).

Figure 4: Error contour for the spin parameter of the supermassive black hole in NGC 1365. The χ2 contour was obtained while allowing all model parameters to vary. The adopted best-fitting model consists of a power law with photon index Γ = , a neutral absorber with N H varying in the range 2.2–2.8 × 1023 cm−2, a ionized absorber with ionization parameter of about 3.5 and N H ≈ 1023 cm−2, a relativistically blurred reflection by a disk with inner radius less than 2.5R g and spin parameter a* as shown in the plot. The ionization state of the disk, its inclination, its iron abundance and the normalization of the reflection component are degenerate, and so are only loosely constrained; they also strongly affect the uncertainties on the spin parameter. To emphasize the importance of correctly considering systematic effects, if (based on previous work on Suzaku data27) we limit the disk inclination to 55–60°, the error on the spin measurement drops: a* = . Analogously, we would obtain an error of the order of ±0.01 if we fitted the whole data set with a model that does not allow for absorption variability. Full size image Download PowerPoint slide

Having tested this model (model 1) against the possible alternative (model 2, the multiple variable absorber), and having left all the model parameters unconstrained, the main potential source of residual systematic error is due to assuming truncation of the disk at the innermost stable circular orbit. However, recent magneto-hydro-dynamical simulations suggest that emission from within the innermost stable circular orbit is negligible24. Other errors could be introduced by assuming a constant, power-law emissivity profile, but the steepness of this profile suggests that its exact shape at radii larger than twice the innermost stable circular orbit is unimportant. The analysis therefore provides robust support for high spin values in active galactic nuclei2, constraining galaxy evolution and black-hole growth models25.