IN
DEGREE PROJECT TECHNOLOGY, FIRST CYCLE, 15 CREDITS
STOCKHOLM SWEDEN 2017,
X-ray Spectroscopy of
Distant Active Galactic Nuclei
CHARLIE BÖRJESON JULIA COHEN
KTH ROYAL INSTITUTE OF TECHNOLOGY
Abstract
The physical structure and behaviour of stars have been studied extensively throughout centuries.
In contrast, the existence of objects now known as Active Galaxies has only been established since a couple of decades. The nucleus of such an object radiates an extraordinary amount of light across the electromagnetic spectrum, sometimes outshining the rest of the galaxy.
However, since many Active Galactic Nuclei (AGNs) live in the distant universe, they are difficult to observe even with the advanced instruments of the 21st century and plenty of effort is being put into this field by astronomers wishing to understand the inner structure of these nuclei. One survey was started in 1999 by the 2002 Nobel Prize winner Riccardo Giacconi, where the Chandra telescope made several observations in the sky direction called Chandra Deep Field South. A separate pencil-beam survey of ∼ 700 arcmin2 in this area has been performed by the telescope XMM-Newton in the ranges 2−10 keV and 5−10 keV, where the number of X-ray sources classified as AGNs is currently estimated to be 339 and 137 for each energy range, respectively.
We studied two sources amongst the most brilliant in the surveys to explore their X-ray spectra and their X-ray variability and found that one is strongly absorbed in the line of sight while the other does not display any absorption. Their variability properties are also different, since the non-absorbed source displays variability behaviour similar to well known nearby AGNs in the literature.
The study is limited to two intrinsically different AGNs finding that they differ not only in the spectral domain but also in the time domain.
Contents
1 Introduction 4
1.1 Physical structure . . . 4
1.2 X-ray spectrum components . . . 5
1.3 Variability . . . 8
1.4 Chandra Deep Field South . . . 9
2 Method and instruments 10 2.1 XMM-Newton . . . 10
2.2 Observations and data reduction . . . 11
2.3 Procedure . . . 13
3 Results 15 3.1 Source 337 . . . 15
3.2 Source 358 . . . 15
4 Discussion 20
5 Summary and Conclusions 22
Chapter 1
Introduction
Active Galactic Nuclei (AGNs) are extremely luminous cores in the centre of some galaxies. They are the most luminous stable sources of electromagnetic radiation in the universe and exhibit quite complex continuum spectra. AGNs typically radiate 1042− 1046ergs/s. For comparison, the luminosity of our sun is L = 4 · 1033ergs/s (Gandhi 2005). These properties make AGNs an interesting topic to study since radiation originating from vast (even in astronomical terms) distances may contain information about the early Universe.
In 1943 Carl Seyfert published a paper where he showed that some galaxies have strong, broadened emission lines and that these galaxies are especially bright, particularly in their nuclear regions. These galaxies are known today as ‘Seyfert Galaxies’.
Almost two decades later, in 1963, Maarten Schmidt reported that the quasar 3C 273 had a remarkably high (or so it was considered at the time) redshift at z = 0.158, thereby demonstrating that it lay far beyond the Milky Way (several billion light-years), possessing a tremendous lumin- osity. This caught on the attention of many astronomers. In 1964 Yakov Zel’dovich and Edwin Ernest Salpeter independently suggested that the origin of the vast amount of energy could be black holes (BHs), three years before the term ‘black hole’ was taken into general use.
In 1968 Donald Lynden-Bell reasoned that each massive galaxy should contain a super massive black hole (SMBH). After this, the subject took off. There has been an influx of people investigating this topic and it is still an area of active research.
1.1 Physical structure
The complete physical structure of AGNs is still not completely understood and there exists several models, however in this text we will consider the currently favoured orientation-based unification scheme depicted in Figure 1.1a.
It is today generally believed that the strong emission from AGNs is the result of accretion onto black holes with masses ranging from 106M to 109M . The accretion disk is a few light-days wide, and emits photons which scatter on a hot corona above the black hole, see the red arrows in Figure 1.1b. The photons gain energy in the corona through inverse Compton-scattering. Some lines of sight to the black hole are blocked by a dusty torus that absorbs radiation, particularly in the UV-optical range (Gandhi 2005).
Jets occur in about 10% of AGNs (Beckmann et al. 2012) and emit radiation in almost the entire electromagnetic spectrum, where the narrower part of the jet closer to the SMBH emits X- rays. The patterns follow as the distance from the SMBH increases, resulting in longer wavelength radiation all the way to radio waves. Highly energetic collimated plasma is believed to be launched off the accretion disk forming such a jet. They often retain a extraordinary degree of collimation and can extend up to megaparsecs in length.
Surrounding the accretion disk lies the Broad Line Region (BLR). The naming of this region is due to the Doppler-broadened emission lines that have been concluded to originate from the end of the accretion disk up to a radius of RBLR ' 1017±1 cm (Beckmann et al. 2012). This is measured from the time delay between the continuum variability and the BLR variability, which is τ = RBLRc . Since such time lags have been observed for several AGNs, it is possible to get a
(a) Unification model. The green arrows indicate how different AGN are categorised due to various viewing angles. Image credit: Pierre Auger Obser- vatory.
(b) Schematic structure of the surroundings of the black hole. Soft accretion disk photons are up- scattered on the optically-thin hot corona plasma.
This process results in a power-law element in the hard spectrum.
Figure 1.1: Physical models of the current understanding of AGNs (components not to scale).
rough estimate of the BLR radius through reverberation mapping. Moreover, using reverberation mapping and invoking the virial theorem is a key method when estimating the mass of the SMBH.
In fact, this is currently the most accurate and common practice when performing this estimation.
Using BLR reverberation data from 35 AGNs B. M. Peterson et al. (2004) show that the black hole masses range between 106− 109M , given a scaling factor f = 5.5. The scaling factor f for the virial theorem depends on SMBH geometry and kinematics.
The Narrow Line Region (NLR) is named by the same convention as the BLR; observed narrow lines originate from what is believed to be a much larger region (RN LR' 1018±2cm) (Beckmann et al. 2012).
There are many different classes of AGNs. For a long time these were thought to have entirely different physical properties, but today most astronomers acknowledge the unification scheme.
The unification scheme hypothesises that all types of AGNs have the same basic structure. The difference in observed emission is explained to be the result of the obscuring torus. Moreover, the observed emission from the AGN depends on the angle of observation so if the torus is in the line of sight, the BLR will not be visible. If instead the line of sight is face on both BLR and NLR emission will be observed.
1.2 X-ray spectrum components
It is customary to name X-rays at energies 0.1 keV < E < 2 keV ‘soft X-rays’ and 2 keV <
E < 100 keV ‘hard X-rays’. Energies above 100 keV are γ-rays. These are somewhat arbitrary distinctions in X-ray astronomy imposed mainly by the technology used in observations.
When attempting to understand AGNs, high-energy radiation (γ- and X-rays) are import- ant since they provide a probe of the inner region of the nucleus, displaying rapid variability.
Roughly 10% of the emitted bolometric luminosity originates from such energy intervals (Brad- ley M. Peterson 1997). Approximately 80% of all cosmic X-ray radiation are accounted to AGNs (Singh 2013).
The main X-ray component from AGNs is a general power-law following the relation N (E) = KE−Γ, where N is the photon density, Γ the photon index and K a constant (usually called the norm of the power-law). The power-law feature comes from the up-scattering of photons from the accretion disk. The average value of the photon index Γ for AGNs in the local universe is 1.9 (Gandhi 2005).
There is a peak in the X-ray spectrum at 6.4 keV (rest-frame), believed to be the result of
Figure 1.2: Simulated absorption (rest-frame) by Gandhi (2005). The photon index used for the simulation was Γ = 1.9. A relexion component was also included in the model, as well as a few other parameters (see Gandhi (2005)). The grey area is the energy interval observed by Chandra and XMM-Newton. Absorption is indicated by the exponent in NH = 10xcm−2.
Figure 1.3: Spectrum components of an average type 1 AGN by Guido Risaliti et al. (2004).
(a) The different steps of the iron line broadening.
A Newtonian Doppler split result in the upper spec- trum. Adding that the blue side will be brighter we get the second panel. Also adjusting for redshift gives us the third panel, and the final line profile as observed from earth can be seen in the last panel.
Produced by Fabian, Iwasawa et al. (2000).
(b) The different broadening effects on the iron line, as calculated and illustrated by Fabian, Rees et al.
(1989). ri is the inner radius, ro the outer radius, i the inclination angle and q the line emissivity vari- ability (assumed to be rq). For each step all other parameters were fixed to ri = 10rs, ro = 100rs, i = 30◦and q = −2.
Figure 1.4: The principles behind iron line broadening.
fluorescence in cold iron in the inner parts of the accretion disk (Singh 2013) as well as the outer parts of the disk and the torus (Yang et al. 2016). However, the observed line is not a narrow line as one would initially expect from fluorescence. It can be rather broad, up to a few keV (Fabian 2005), but more commonly a few hundred eV (Falocco, S. et al. 2013). Different models predict the broadening effects on the iron line differently, and the best prediction of the broadening is obtained by taking both special and general relativity into account (see Figure 1.4a). Various parameters of the accretion disk conditions also affect the final line profile. Because the broadening is affected by the iron’s proximity to the BH, the inner and outer radii of the accretion disk are relevant to the line profile as well as the iron distribution in the accretion disk. The viewing angle also affects the perceived Doppler broadening. The effects of each parameter was thouroghly calculated by Fabian, Rees et al. (1989) and can be seen in Figure 1.4b.
The column density NH of the obscuring torus affects the spectrum shape if it is in our line of sight (see Figure 1.2). The higher the column density, the more soft photons are absorbed, resulting in a hardened spectrum. Moreover, the iron line becomes more prominent, as it originates partly from the torus.
Some photons from the corona are reflected on the accretion disk. High energy photons are Compton scattered on the disk, resulting in a high energy excess bump mainly in the 20 − 40 keV energy interval and fluorescence lines at lower energies (Reynolds 1998; Singh 2013).
Even when accounting for all the above factors, there is an excess of soft X-rays. The source
of this soft excess is debated.
All different components can be viewed in Figure 1.3.
1.3 Variability
AGNs display some erratic variability over time (down to a few hundred seconds (Sobolewska et al. 2009)) in luminosity over large domains in the electromagnetic spectrum (Beckmann et al.
2012). This attribute is in fact one of the defining characteristics of an AGN, however the exact interpretation of the underlying physical conditions for the variability is still an active area of research. Furthermore, variability from the optical and near-IR bands to the X-ray band is often correlated, where the X-ray emission displays distinguished variability compared to the optical and UV wavelengths, having the largest amplitude variations on timescales of days to weeks (Lira et al.
2009). However, the optical wavelengths show larger variability on longer timescales (months or years) for which reason one single cause of variability is unlikely. The multi-wavelength study by Lira et al. (2009) found that the near-IR emission from MR 2251-178 correlated with the B- and V-bands (using photometric system nomenclature) which in turn originate from the accretion disk.
In addition, they observed very small time lags between the J- and B-bands and concluded that this emission too emerges from the accretion disk (more precisely colder territories on the disk exposed to X-ray illumination). In contrast, for the source NGC 3783 the near-IR lag was found to be ∼ 85 days. The natural consequence of this finding is that one single cause of variability fails to explain the different observed lag timescales between distinct emitting regions. In short, one cannot hope to explain all variability phenomena by simply studying the X-ray spectrum yet it is an important informant simply due to the earlier mentioned fact that X-rays emanate from the inner region in AGNs.
Three of the types of observed variability in the X-ray spectrum are (i) variable column density NH, (ii) variations in the spectral slope and (iii) flux variability:
(i) As stated earlier, the column density is suggested to be a result of a torus obscuring the central engine or absorption from BLR clouds. Indeed, one study by G. Risaliti, Elvis and Nicastro (2002) concludes that a satisfactory explanation for the observed NH variability is a clumpy gas absorber that is not spherically symmetric. However, in a more recent investigation by G. Risaliti, Elvis, Bianchi et al. (2010), the fast variability (hours to days) suggests absorbtion originating from distances and sizes related to the BLR. The observed source UGC 4203 is known to be obscured by a farther absorber in the line of sight, unrelated to the BLR since this region lies inside the dust sublimation radius. This absorber may be the torus, but it is not certain since different absorption components could not disentangled. Risaliti also mentions that the evidence is based on a small sample of sources without homogeneous selection criteria wherefore short-term NH variability should be studied in a more representative sample in the future.
(ii) There is a significant variability in the slope of the power-law in many AGNs, as shown in Sobolewska et al. (2009). They also found that intrinsic variation in the spectral slope can be explained through variations in mass accretion rate (Sobolewska et al. 2009). A positive correlation between the accretion rate and the observed photon index Γ implies that higher accretion rate sources have steeper average spectra. They found the correlation to be Γ ' 2.7 ˙m0.08, where ˙m is the accretion rate.
(iii) Sobolewska et al. (2009) also noted a high variability in flux. Ponti et al. (2012) have found that the flux variability properties between AGNs can be accounted for entirely by the black hole mass MBH and ˙m. The study included more than 100 mostly local (z ≤ 0.2) AGNs.
In many AGNs the photon index is also directly correlated with flux (Lamer et al. 2003;
Sobolewska et al. 2009). When Γ is high, the AGN is also brighter. This is known as the ‘steeper when brighter’ trend. It is thought to be a result of a truncation of the accretion disk (Matt, G.
et al. 2005). When the accretion disk is truncated it radiates less energy and the corona stays hot, resulting in a hard spectrum. When the accretion disk extends in to the BH the corona undergoes Compton cooling, and we see a softer spectrum (Done et al. 2004).
1.4 Chandra Deep Field South
The Chandra Deep Field South (CDF-S) is an image captured by the NASA Chandra satellite during the survey that was launched by Riccardo Giacconi in 1999. He was awarded the Nobel Prize in Physics ‘for pioneering contributions to astrophysics, which have led to the discovery of cosmic X-ray sources’ (”The Nobel Prize in Physics” 2002).
The ESA satellite XMM-Newton has also studied the X-rays from CDF-S. With a nominal exposure of 3.45 Ms covering an area around 700 arcmin2, this survey of CDF-S is the deepest X-ray survey performed by XMM-Newton and its estimated number of point sources is 339 in the 2 − 10 keV range and 137 in the 5 − 10 keV range (Ranalli et al. 2013). Deep X-ray surveys is the primary tool for understanding AGNs in the earlier stages of the universe.
In this paper we focus on two AGNs in CDF-S: sources 337 and 358 (IDs from Ranalli et al.
(2013)). They are the third and fourth most luminous sources in the XMM-Newton observations of CDF-S. We analyse these two because the two most luminous sources (319 and 203) have already been extensively studied in Iwasawa, K. et al. (2015). Spectral extraction and source detection were performed by Ranalli et al. (2013). Source 337 has coordinates 03h32m38s, -27◦3904500, and a redshift of z = 0.837. Source 358 is located at 03h32m08s, -27◦3703200, and has a redshift of z = 0.976 (Ranalli et al. 2013). The redshifts have been confirmed by Luo, Brandt et al. (2017) in their 7 Ms catalogue, with source 337 corresponding to source 716 in their catalogue, and source 358 corresponding to their source 168 (as indicated by column 73 in the Chandra 7 Ms catalogue).
The observations used in our analysis are from XMM-Newtons observations in 2001–2002 and 2008–2010 with a total exposure of 3.45 Ms. The X-ray sources were identified at 4σ-confidence level and the surrounding background radiation was mapped. The exact procedure can be found in Ranalli et al. (2013).
Chapter 2
Method and instruments
2.1 XMM-Newton
XMM-Newton is a European Space Agency (ESA) X-ray satellite launched in 1999. It features three different scientific instruments including a very sensitive telescope for the optical/UV range,
‘The European Photon Imaging Camera’ (EPIC) imaging cameras and the Reflection Grating Spectrometer (RGS). There are three telescopes on XMM-Newton of Wolter type I and two of them are equipped with reflective grating assemblies, dispersing X-rays in much the same as a prism does with visible light. In this way an improved spectral resolution is achieved at secondary focus and more information about specific elements such as oxygen or iron can be extracted by the RGS detector (NASA’s HEASARC n.d.). In these two telescopes 40% of the X-rays hit the RGS’
own Charge Couple Device (CCD) camera while about 44% of the radiation strikes the EPIC, leaving the remainder absorbed by the support structures of the reflexion grating arrays.
Since the atmosphere of the earth does not let cosmic X-rays through, one must send a satellite into space in order to observe them. In addition to this difficulty, X-rays do not reflect when striking an ordinary mirror surface, they are so energetic they pass right through. In order to redirect an X-ray a special mirror is needed, one that is made of ceramic or metal foil coated in gold or iridium, and the incident angle has to be very small. In XMM-Newton the material used is gold-plated Silicon carbide (SiC). Furthermore, the mirror needs to be constructed in such a way that the X-rays hit the surface in a very small so-called ‘grazing’ angle.
XMM-Newton has an orbital period of 48 hours, enabling it to make long undisturbed obser- vations (ESA 2015). It has a moderate spectral resolution of E/∆E = 20 − 50 (ESA 2016b), an angular resolution of 60 Full Width at Half Maximum (FWHM) and an impressive mechanical stability resulting in a pointing accuracy of 0.250when observing targets over long periods of time.
Another strength of this satellite is its large mirror surface, with an effective area of 1900 cm2 up to energies at 150 eV, 1500 cm2 at 2 keV and 350 cm2 at 10 keV (Beckmann et al. 2012). The relevant effective area in this paper is somewhat smaller and is illustrated by the 2-module MOS curve in Figure 2.1. The mirrors are nested concentrically about the telescope axis, where the thinnest mirror is 0.47 mm thin and the thickest 1.07 mm (ESA 2015). The straylight effective area of the EPIC depends on the off-axis angle to the focal plane, it is negligible for angles > 1.4◦. For angles between 200 and 1.4◦ it extends to about 3 cm2.
The main focal plane cameras on this satellite are the EPIC cameras of which two are EPIC- MOS (‘Metal Oxide Semiconductor’), associated with the RGS detectors, carrying 7 CCDs (each 10.90× 10.90) each and one EPIC-pn with 12 CCDs (each 13.60× 4.40) in the focal plane of the unobstructed X-ray path (ESA 2016b). The CCDs are sensitive in the energy range 0.15 − 12 keV, resulting in a field of view around 300. See Figure 2.2 for a visual explanation of their geometry and setup.
Understanding the different components in the variable EPIC-MOS background is crucial before performing any kind of data analysis. According to ESA there are three categories of components contributing to the background: photons, particles and electronic noise (ESA 2017). One can also view the background components as either internal, originating from interference with the materials in the instruments, or externally emerging from the Galactic and cosmic background. For instance,
Figure 2.1: Effective area of the instruments of XMM-Newton plotted against energy. Image from ESA (2016b).
around 1 keV there is a temporally variable ‘quiescent’ component from highly energetic particles interacting with the detectors and their surrounding structures (Kuntz et al. 2008). Moreover, Ranalli et al. (2013) observed a doubling of background radiation between 2001–2002 and 2008–
2010. This increase was due to the external component, and it is thought to relate to the Sun’s cycle (Ranalli et al. 2013).
(a) The X-ray view of CDF-S as observed by the two MOS-cameras. The brightest sources are labelled by their ID:s from Ranalli et al. (2013). Image from Falocco (2017).
(b) At the focal point of the optical axis the central CCD is placed. The other CCDs are 4.55 mm closer to the mirror the in order to mimic the curvature of the focal plane. Image from ESA (2017).
Figure 2.2: Two pictures of an EPIC-MOS camera. In (a) the image produced by the camera is displayed and (b) shows the physical CCD arrangement.
2.2 Observations and data reduction
The spectra obtained by XMM-Newton from sources 337 and 358 were extracted by Ranalli et al.
(2013) and then further processed in order to be properly analysed. The background radiation was calculated from nearby regions (for exact procedure, see Georgantopoulos, I. et al. (2013)).
Counts and exposure corresponding to the same source, observation and filter were summed from the two MOS cameras into one spectrum. The data from the PN detector were excluded from the summation, because of the unstable background radiation detected in the PN camera. The weighted average of the response matrices and ancillary files were computed with Ftools task
(a) The observations of source 337.
(b) The observations of source 358.
Figure 2.3: Flux levels and hardness ratios are calculated in the 0.5 − 8 keV observed frame band.
Cyan diamonds and green triangles correspond to MOS1 and MOS2 respectively. Circles indicate observations with more than 10% of the flux outside the MOS chip or in-between chips, or where source and background fall on different chips. Squares indicate observations where more than 30%
of the background is outside the chip or in-between chips. Black crosses indicate the summed spectra from MOS1 and MOS2, excluding the bad epochs mentioned above. The average fluxes are shown as dotted lines. The solid black line is the zero level. The observations are grouped into 6 epochs, marked by the 6 quadrants in the figure. The hardness ratio is defined by HR = H−SH+S where S is the soft flux and H is the hard flux. Image courtesy of Serena Falocco (Falocco 2017).
addrmf. The observations were then grouped into epochs, defined as adjacent observations. In total there are 6 epochs, visualised by the 6 quadrants in Figure 2.3 (Falocco 2017). We see from the lightcurves (Figure 2.3 that there is great variability in flux for both of our sources, but the hardness ratio is quite stable. The variation in flux is 19% in source 337 and 17% in source 358 (Falocco 2017).
Epochs 2, 3, 4 and 6 for source 337 and epochs 2, 4, 6 for source 358 were chosen for our study. The other epochs had observations with photometric problems, making them unsuitable for analysis. The observations in each used epoch were merged to single epoch spectra before the spectral analysis. One summed spectrum of the used epochs was also produced for each source in order to increase the precision of the fits.
2.3 Procedure
Xspec was used for analysis of the X-ray spectra. It is a program developed by NASA/GSFC specifically for X-ray spectral fitting from satellite data. It is command-driven and can be run on a variety of platforms (NASA 2017). We have used Xspec version 12.9.1 for our analysis. Errors were calculated in the 90% confidence interval, the default in Xspec.
The spectrum to be analysed was loaded with the data command. Channels outside the 0.5 − 10 keV range (rest-frame) were ignored with the ignore command. Bad channels were also ignored. Energies were converted to the observed frame using the equation Eobs= Eem/(1 + z).
In source 358 the softest part of the spectra were dominated by background radiation, wherefore channels below 0.8 keV rest-frame (1.6 keV observed frame) were ignored in all spectra from source 358. The spectra of the available epochs can be seen in Figure 2.4. From Figure 2.4a we see that there is an apparent variability in flux in source 337, especially in the softer X-rays. The flux variability is not apparent in source 358 (Figure 2.4b), because the epochs average out the variability seen within the epochs for source 358 (Figure 2.3b). However, we notice that the epochs in source 358 are reduced in the soft X-ray band, suggesting that the source is absorbed.
The used individual epochs and the summed spectra for each source were fitted to a simple power-law model with absorption factors. The primary Xspec model used was wabs*zwabs*pow.
The galaxy absorption was fixed to 8 · 10−19cm−2, and the redshift was fixed to 0.837 and 0.976 for sources 337 and 358 respectively. Parameters NH, Γ, and Γ-norm were free to vary. We call this model P (power-law).
The P model was modified using the command editmodel to include a Gaussian line for the iron peak. The Xspec model used for this was wabs*zwabs*(pow+gauss). Using the editmodel command instead of model allows Xspec to add a new component while keeping the values of the parameters from the previous fit. This should result in an improved goodness-of-fit. σ was fixed to 0.1 keV and the energy of the line to 6.4 keV (converted to observed frame). The norm of the Gaussian peak was free to vary, along with the free parameters from model P. We call this model PG (power-law gauss). Both individual epochs and summed spectra were fitted the PG model.
The PG model was further modified to wabs*zwabs*(pow+gauss+pow) in order to see if the spectrum could be interpreted with an additional power-law representing soft excess. The two power-laws were set to have the same Γ but different norms. Both photon index and norms were allowed to vary. This was performed only on the summed spectra. We call this model PGP (power-law gauss power-law).
The modified model wabs*zwabs*(pow+gauss+pexrav) which accounts for the reflexion, was also fitted to the summed spectra only. The reflexion scaling factor was set to vary between -2 and 0 in order to add a reflective component only (the primary power-law is accounted for in the power-law component). Γ was set to be equal to Γ of the power-law, the redshift was fixed and the norm was allowed to vary. All other parameters in the reflexion component were fixed to the Xspec default values. We call this model PGR (power-law gauss reflexion).
The 90% confidence interval for the values of the parameters were extracted with the error command in Xspec. The flux was calculated with the flux command for the specified energy intervals. The error of the flux was calculated from the relative error of the power-law norm, because the relative error in the flux is equal to the relative error of the power-law norm.
The calculated total fluxes and photon indexes for the individual epochs were used for analysis
1
0.5 2 5
10−510−410−3
normalized counts s−1 keV−1
Energy (keV) Source: 337
(a) The counts for the used epochs in 337 plot- ted together. Black: epoch 2. Red: epoch 3.
Blue: epoch 4. Magenta: epoch 6. We see that epoch 3 has higher flux than the others, just like observed in Figure 2.3a.
1 2 5
10−510−42×10−55×10−52×10−45×10−4
normalized counts s−1 keV−1
Energy (keV) Source: 358
charlie 25−Mar−2017 09:39
(b) The counts for the used epochs in 358 plot- ted together. Black: epoch 2. Red: epoch 4.
Blue: epoch 6.
Figure 2.4: The unfitted spectra for each source with epochs plotted together.
of the ‘steeper when brighter’ trend. We used the values from the PG fits, with total flux calculated in the range 0.5 − 10 keV and 1.6 keV for sources 337 and 358 respectively. Linear regression was performed in Python, with the function linregress from the package SciPy (The Scipy community 2014).
Chapter 3
Results
The values of the fitted parameters for each model and source can be found in Table 3.1. The calculated fluxes can be found in Table 3.2.
3.1 Source 337
All models give consistent results on the absorption of source 337. It has very little to no absorption in our direct line of sight. This is true for both the different epochs and the summed spectrum.
The photon index is ∼ 2.19 for all models except model PGR, where we got a photon index of
∼ 2.39. This difference is expected, because reflexion has the effect to flatten the spectrum, thus if it is not taken into account the Xspec fit will find a lower photon index.
The iron line was found to have an EW of 0.18 ± 0.08 keV in the PG and PGP model, consistent with the values in Falocco, S. et al. (2013) where they calculated the average EW from the spectra of the whole CDF-S survey. The PG model improved the χ2-value significantly from the P model.
The PG fit of the spectrum can be seen in Figure 3.1a.
Model PGP resulted in one power-law with a high norm and one with an almost non-existent norm. The χ2-value did not improve from the previous model. This means that we do not find a second power-law to account for the soft excess.
In the fit of the PGR model, the photon index increased, and the iron line almost disappeared.
The χ2-value also dropped. However, the introduction of the reflexion factor does not affect the χ2-value. We checked this with the steppar command, and the χ2is 209.07 for values of R in the range [-2, 0).
Source 337 appears to have a correlation between photon index Γ and flux (see Figure 3.2a).
The linear regression has a p-value of p = 0.03 and the correlation factor is r = 0.97, indicating that we have 97% certainty of having found a correlation.
There is also an inverse correlation between flux and EW (see Figure 3.2c). The p-value for the linear regression is p = 0.04 and the correlation factor is r = −0.96, indicating a 96% certainty of an anti-correlation.
3.2 Source 358
Unlike source 337, source 358 has a high absorption with an average of (8.3 ± 0.7) · 1022cm−2 for the different models. The three epochs have variable absorptions, but due to the high uncertainty it is not significant.
The spectrum is relatively hard, with a photon index of Γ = 1.56. This is lower than typical for AGNs, because the source is heavily absorbed in the soft X-rays.
The iron line is almost non-detectable. The calculated EW from the summed spectrum was 0.03 keV with an uncertainty of ±0.04 keV for the PG model. We also notice that the χ2-value hardly improves, but remains essentially the same as for model P. The iron line might be very faint, or there is too much noise in the data, but only future studies can conclude which. The fit to the spectrum can be seen in Figure 3.1b.
10−6 10−5 10−4 10−3
normalized counts s−1 keV−1
Source: 337. Model: Powerlaw + Gaussian, with residual.
1
0.5 2 5
−2×10−4 0 2×10−4 4×10−4
normalized counts s−1 keV−1
Energy (keV)
juliacohen 25−Apr−2017 15:11
(a) Summed spectra of all used epochs of 337, fitted with model PG. Residual is plotted be- neath. The iron peak lies at 3.48 keV observed frame.
10−7 10−6 10−5 10−4
normalized counts s−1 keV−1
358 total. (wabs*zwabs+gauss).
1 2 5
−10−4 0 10−4 2×10−4
normalized counts s−1 keV−1
Energy (keV)
juliacohen 24−Mar−2017 08:28
(b) Summed spectra of all used epochs of 358, fitted with model PG. Residual is plotted be- neath. The iron peak lies at 3.24 keV observed frame.
Figure 3.1: The fits of the PG model to the summed spectra of 337 and 358.
The double power-law model resulted in one high norm and one zero-norm, and the χ2did not improve, meaning we do not find a second power-law to account for the soft excess.
In the fit with the PGR model, Γ and NH both increased like in source 337. The χ2-value decreased, but not very much (see Table 3.1). However, just like for source 337 the χ2-value was as good as non-dependent on the reflexion factor, with a constant χ2 of 172.06.
Source 358 has no apparent correlation between photon index and flux (see Figure 3.2b). A linear regression was performed (though not displayed), giving a p-value of p = 0.952 and a correlation factor of r = −0.075.
3.5 4 4.5 5 5.5 6 6.5 7 7.5 Total flux [ergs/cm 2/s] ×10-14 1.9
1.95 2 2.05 2.1 2.15 2.2 2.25 2.3
Photon Index Γ
Source: 337
(a) The relation between photon index and flux in the 0.5 − 10 keV (rest-frame) range for source 337. The photon indexes and fluxes were calcu- lated for the used epochs with the PG model.
The p-value for the linear regression is p = 0.03, meaning there is a 97% certainty that there is a correlation.
1 2 3 4 5 6 7
Total flux [ergs/cm 2/s] ×10-14 1
1.2 1.4 1.6 1.8 2 2.2
Photon Index Γ
Source: 358
(b) The relation between photon index and flux in the 1.6 − 10 keV (rest-frame) range for source 358. The softest X-rays were cut from the spec- trum and the analysis due to high background radiation (see Section 2.3). The photon indexes and fluxes were calculated for the used epochs with the PG model. There is no apparent cor- relation between Γ and flux in source 358.
3.5 4 4.5 5 5.5 6 6.5 7 7.5
Total flux [ergs/cm 2/s] ×10-14 0
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
EW (keV)
(c) The relation between EW and flux in the 0.5 − 10 keV range for source 337. The EWs and fluxes were calculated from the used epochs with the PG model. The p-value for the linear regression is p = 0.04, meaning there is a 97%
chance that there is an anti-correlation.
Figure 3.2: Various parameters for the epochs were extracted from the PG models, analysed with linear regression and plotted together. Black crosses indicate epochs with error bars, and red lines indicate the calculated linear correlations. Error bars were plotted with the user-made by Zoergiebel (2009).
Table 3.1: The calculated values of the relevant parameters for every spectrum and model. (1):
The model used for the fits displayed on the corresponding rows. (2): The source and epoch used for the fit. (3): The column density in units of 1022cm−2. (4) The photon index Γ. (5) The χ2-value of the fit / degrees of freedom. (6) The calculated EW of the iron line (if present in model). (7) The norms of the two power-laws in the PGP model, in units of 10−5. (8) The reflexion component from the PGR model.
(1) (2) (3) (4) (5) (6) (7) (8)
Model Source NH Γ χ2/ deg EW [keV] Γ-norms Refl
P 337 : 2 < 0.003 2.06 ± 0.10 36.95 / 51 - - -
337 : 3 < 0.013 2.24 ± 0.05 129.79 / 124 - - -
337 : 4 < 0.008 2.15 ± 0.06 124.20 / 106 - - -
337 : 6 < 0.004 2.21 ± 0.05 133.94 / 130 - - -
337 : sum < 0.002 2.18 ± 0.03 251.29 / 222 - - -
358 : 2 7.84 ± 2.03 1.79 ± 0.41 19.48 / 26 - - -
358 : 4 9.66 ± 1.75 1.72 ± 0.28 46.65 / 56 - - -
358 : 6 8.41 ± 1.77 1.41 ± 0.27 64.33 / 61 - - -
358 : sum 8.27 ± 0.15 1.55 ± 0.15 174.30 / 185 - - - PG 337 : 2 < 0.057 2.07 ± 0.12 33.67 / 49 0.25 ± 0.25 - - 337 : 3 < 0.015 2.25 ± 0.05 126.13 / 122 0.14 ± 0.12 - - 337 : 4 < 0.028 2.16 ± 0.08 112.97 / 104 0.23 ± 0.17 - - 337 : 6 < 0.011 2.22 ± 0.05 127.15 / 128 0.21 ± 0.14 - - 337 : sum < 0.06 2.19 ± 0.03 229.91 / 220 0.18 ± 0.08 - -
358 : 2 8.19 ± 2.08 1.77 ± 0.40 19.60 / 25 0 - -
358 : 4 9.45 ± 1.73 1.73 ± 0.29 42.92 / 55 0.10 ± 0.09 - - 358 : 6 8.32 ± 1.77 1.41 ± 0.28 63.89 / 60 0.04 ± 0.05 - - 358 : sum 8.21 ± 0.91 1.56 ± 0.15 173.19 / 184 0.03 ± 0.04 - - PGP 337 : sum < 0.006 2.19 ± 0.03 229.91 / 219 0.18 ± 0.08 1.24 / 0 - 358 : sum 8.21 ± 0.91 1.56 ± 0.15 173.19 / 188 0.03 ± 0.04 1.85 / 0 - PGR 337 : sum 0.02 ± 0.03 2.39 ± 0.14 209.07 / 218 0.06 ± 0.06 - -2
358 : sum 8.35 ± 0.92 1.76 ± 0.29 172.06 / 188 0.01 ± 0.03 - -2
Table 3.2: Calculated flux from each model. (1) The model used for flux calculations. (2) The spectrum used for the fit. (3) The total flux of the soft and hard X-rays in columns (4) and (5).
(4) The soft flux, calculated in the range 0.5 − 2 keV (rest-frame) for source 337 and in the range 1.6 − 2 keV (rest-frame) for source 358. The softest X-rays from source 358 had to be cut from the spectrum because of high background radiation in that band (see Section 2.3). (5) Calculated hard flux in the range 2 − 10 keV (rest-frame).
(1) (2) (3) (4) (5)
Model Source Flux [ergs/cm2/s] Soft flux Hard flux P 337 : 2 (1.94 ± 0.11) · 10−14 (1.74 ± 0.01) · 10−14 (2.01 ± 0.11) · 10−14
337 : 3 (7.04 ± 0.20) · 10−14 (3.74 ± 0.10) · 10−14 (3.29 ± 0.09) · 10−14 337 : 4 (4.91 ± 0.18) · 10−14 (2.46 ± 0.09) · 10−14 (2.45 ± 0.09) · 10−14 337 : 6 (5.74 ± 0.18) · 10−14 (2.98 ± 0.09) · 10−14 (2.76 ± 0.08) · 10−14 337 : sum (5.59 ± 0.09) · 10−14 (2.86 ± 0.05) · 10−14 (2.73 ± 0.04) · 10−14 358 : 2 (3.87 ± 2.07) · 10−14 (9.03 ± 4.83) · 10−17 (3.87 ± 2.07) · 10−14 358 : 4 (4.75 ± 1.74) · 10−14 (5.12 ± 1.87) · 10−17 (4.75 ± 1.74) · 10−14 358 : 6 (4.31 ± 1.59) · 10−14 (5.84 ± 2.16) · 10−17 (4.30 ± 1.59) · 10−14 358 : sum (4.47 ± 0.88) · 10−14 (5.84 ± 2.16) · 10−17 (4.30 ± 1.59) · 10−14 PG 337 : 2 (3.78 ± 0.26) · 10−14 (1.73 ± 0.12) · 10−14 (2.05 ± 0.14) · 10−14 337 : 3 (7.06 ± 0.20) · 10−14 (3.74 ± 0.11) · 10−14 (3.32 ± 0.10) · 10−14 337 : 4 (4.94 ± 0.21) · 10−14 (2.42 ± 0.10) · 10−14 (2.51 ± 0.10) · 10−14 337 : 6 (5.77 ± 0.17) · 10−14 (2.97 ± 0.09) · 10−14 (2.80 ± 0.08) · 10−14 337 : sum (5.62 ± 0.09) · 10−14 (2.84 ± 0.05) · 10−14 (2.77 ± 0.05) · 10−14 358 : 2 (3.89 ± 2.05) · 10−14 (8.76 ± 4.61) · 10−17 (3.88 ± 2.04) · 10−14 358 : 4 (4.77 ± 1.77) · 10−14 (5.55 ± 2.06) · 10−17 (4.76 ± 1.77) · 10−14 358 : 6 (4.31 ± 1.61) · 10−14 (6.03 ± 2.26) · 10−17 (4.30 ± 1.61) · 10−14 358 : sum (4.47 ± 0.90) · 10−14 (7.80 ± 1.55) · 10−17 (4.46 ± 0.89) · 10−14 PGP 337 tot (5.62 ± 2.85) · 10−14 (2.84 ± 1.44) · 10−14 (2.77 ± 1.41) · 10−14 358 tot (4.47 ± 2.73) · 10−14 (7.80 ± 4.76) · 10−17 (4.46 ± 2.72) · 10−14 PGR 337 tot (4.34 ± 0.14) · 10−14 (1.44 ± 0.04) · 10−14 (2.90 ± 0.09) · 10−14 358 tot (4.46 ± 0.74) · 10−14 (7.94 ± 1.32) · 10−17 (4.45 ± 0.74) · 10−14
Chapter 4
Discussion
Our values for absorption NH agree with those of Luo, Brandt et al. (2017). Their observations were performed by NASA satellite Chandra, from observations in 1999–2000 and 2007 (2 Ms) (Luo, Bauer et al. 2011), 2010 (2 Ms) (Xue et al. 2011) and 2014–2016 (3 Ms). In their main-catalogue the absorption for 337 is 0, and the absorption for 358 is 7.90 · 1022cm−2. These are similar to our own results, even though the times of observation are different. We can draw the conclusion that source 337 has no absorption in our line of sight, and that source 358 has absorption in our line of sight, with an average column density of ∼ 8 · 1022cm−2. Source 337 has also previously been classified as a broad-line AGN (non-absorbed) from its optical spectrum (Szokoly et al. 2004), indicating that there is no variability in the absorption of source 337.
We evaluated the correlation between Γ and flux in our AGNs. From Figure 3.2a it appears that there is a positive correlation in source 337. The linear regression gave a p-value of 0.03, meaning that we have a significant correlation at the p < 0.05 level. However, we have to reject the theory for source 358, which is apparent from both Figure 3.2b and the calculated p > 0.05. Judging both by the large errors and the proximity of the data points, it is not an unexpected result. According to Paolillo, M. et al. (2004), it is probable that the ‘steeper when brighter’ trend is generally not observed in absorbed AGNs, because the highly variable central core is obscured by the torus.
The variable power-law will then be diluted with the more constant reflexion component, which is non-variable and emitted from material far away from the central engine. It is also generally hard to distinguish the power-law from the absorption, making it even harder to analyse the variability in absorbed sources.
The study of variability in AGNs by Sobolewska et al. (2009) clearly identified the ‘steeper when brighter’ trend in almost all of their studied AGNs. Their AGNs had a redshift of z < 0.05 and they had between 200 and 1200 observations for each source, each observation having a typical exposure of 1 − 2 ks. When summing observations into different epochs, they had ∼ 10 epochs each containing more than 20 observations. This resulted in low errors for the properties of each epoch, and the ‘steeper when brighter’ trend was clear and apparent in almost all of their sources.
Our study essentially does the same, with fewer, but longer, observations (∼ 100 ks) summed into epochs used for variability analysis.
One interesting feature of variability in AGNs is that it seems to be anti-correlated with lumin- osity (Almaini et al. 2000). The AGNs analysed in Sobolewska et al. (2009) have lower luminosity than ours (∼ 1043ergs/s compared to 1044ergs/s in source 337), and so we can expect their sources to show higher variability than ours (which they do). Because distant AGNs generally are more luminous (they have to be in order for us to observe them), it is currently unclear whether this anti-correlation is an effect of the luminosity alone or if there is also an effect from the higher redshift.
We found that the variations in NH in source 358 are consistent within the uncertainties. It is not unexpected, because significant variation in column density is rare: in a study of the CDF-S by Yang et al. (2016) they only found one out of 68 AGNs that had a variable absorption (with a time-scale of ∼ a year). Faster variability in absorbed AGNs have been found in e.g. G. Risaliti, Elvis, Bianchi et al. (2010), but our epochs are far spaced and variations on those time-scales (a few days) can not be analysed with our samples.
There is an inverse correlation between the EW and flux in source 337. This relation was noted in Sobolewska et al. (2009), but not found in all of their studied AGNs. Bianchi, S. et al. (2007) also performed an extensive study of this relation. These studies identified an inverse correlation between different AGNs, not as variability within the same AGN. However, Bianchi, S. et al. (2007) proposed that the anti-correlation is due to changes in the conditions of the central engine. The anti-correlation is then also expected as variability in a individual AGNs. However, it is the case that our linear regression, though showing a p-value of significance, is probably not significant.
This is because of the big errors of the EW, which the linear regression does not consider.
More general conclusions about the variability in these two sources, especially in source 358, must be deferred to future X-ray surveys with increasingly sensitive X-ray instruments, such as the ESA satellite Athena which is planned to be launched in 2028 (ESA 2016a). Athena will be able to detect three times as many sources in a single pointing as Chandra would have with identical exposure time, and the limiting flux will be a fifth of that of XMM-Newton for the same exposure time (Barcons et al. 2012). This along with greater sensitivity will enable spectral analysis of faint, distant X-ray sources and will facilitate the study of variability in highly absorbed AGNs and the anti-correlation between luminosity and variability.
Chapter 5
Summary and Conclusions
We have analysed two distant AGNs in Chandra Deep Field South: source 337 with a redshift of z = 0.837 and source 358 with a redshift of z = 0.976 (IDs from Ranalli et al. (2013)). They were observed by European Space Agency satellite XMM-Newton in 2001–2002 and 2008–2010 with a total exposure of 3.45 Ms. The analysis was performed with NASA developed software Xspec.
We fitted the individual epoch spectra with a simple power-law and absorption, and also with an expanded model including an iron line. This was in order to analyse variability in the photon index, flux and also in the iron line. The summed spectra were fitted to these models to get better precision in the fits. The summed spectra were also fitted to more complicated models, including an extra power-law for soft excess or a reflexion component.
Our main conclusions are:
• The flux is highly variable between observations in both sources.
• Source 337 has no absorption in our line of sight and a steep spectrum with a photon index of Γ = 2.2. We detect an iron line with an EW of 0.18 keV, a value consistent with the average in CDF-S (Georgantopoulos, I. et al. 2013).
• Source 358 has an absorption of NH = 8 · 1022cm−2 and a flat spectrum with a photon index of Γ = 1.56. The iron line is very faint. Most probably there is an iron line, but the data is too scarce to properly detect it, which often is the case with distant, absorbed AGNs.
• The ‘steeper when brighter’ trend is confirmed in source 337, but not in source 358. It has been argued that this is not commonly seen in absorbed sources, because the highly variable central source is blocked from view and the power-law is diluted by a constant reflexion component (Paolillo, M. et al. 2004).
• There is an inverse correlation between flux and EW of the iron line in source 337. Although this phenomenon has mainly been studied as a trend between different AGNs, the anti- correlation is most probably due to changing conditions in the central engine.
• Variability in absorption of source 358 could not be established due to high uncertainty in the absorption.
In summary, the non-absorbed source 337 follows the ‘steeper when brighter’ trend, previously observed in non-absorbed sources at lower redshift and luminosities by Sobolewska et al. (2009).
The absorbed source 358 does not show any variability, possibly due to the low count rate and high uncertainties. The results of this thesis were part of the oral contribution to the conference
”Big Questions in Astrophysics” in Lund, April 2017, presented by our supervisor Serena Falocco.
Acknowledgements
We would like to thank our supervisor Serena Falocco, for her extensive support and guidance.
Abbreviations and Units
Abbreviations
AGN Active Galactic Nucleus BLR Broad Line Region CCD Charge Couple Device CDF-S Chandra Deep Field South
EPIC European Photon Imaging Camera ESA European Space Agency
EW Equivalent Width
FWHM Full Width at Half Maximum
GSFC Goddard Space Flight Center (part of NASA) MOS Metal Oxide Semi-conductor
NASA National Aeronautics and Space Administration NH Column density
NLR Narrow line region
RGS Reflection Grating Spectrometer SMBH Super massive black hole XMM X-ray Multimirror Mission
z Redshift
Γ Photon index
Units
1 erg = 10−7J
1 arcmin = 10= 601 ◦≈ 0.0167◦ 1 parsec = 3.0857 · 1016m
1 solar luminosity = 1 L = 3.839 · 1033erg s−1 1 solar mass = 1 M = 1.989 · 1033g
Bibliography
Almaini, O. et al. (2000). “X-ray variability in a deep, flux-limited sample of QSOs”. In: Monthly Notices of the Royal Astronomical Society 315, pp. 325–336. doi: 10.1046/j.1365- 8711.
2000.03385.x. eprint: astro-ph/9912068.
Barcons, X. et al. (2012). “Athena (Advanced Telescope for High ENergy Astrophysics) Assess- ment Study Report for ESA Cosmic Vision 2015-2025”. In: ArXiv e-prints. arXiv: 1207.2745 [astro-ph.HE].
Beckmann, Volker and Chris Schrader (2012). Active Galactic Nuclei. Wiley-VCH. isbn: 9783527410781.
Bianchi, S. et al. (2007). “On the Iwasawa-Taniguchi effect of radio-quiet AGN”. In: Astronomy
& Astrophysics 467.1, pp. L19–L22. doi: 10.1051/0004-6361:20077331. url: https://doi.
org/10.1051/0004-6361:20077331.
Done, Chris and Marek Gierli´nski (2004). “Spectral Transitions in X-Ray Binaries”. In: Progress of Theoretical Physics Supplement 155, p. 9. doi: 10.1143/PTPS.155.9. eprint: /oup/backfile/
content_public/journal/ptps/155/10.1143/ptps.155.9/2/155-9.pdf. url: +%20http:
//dx.doi.org/10.1143/PTPS.155.9.
European Space Agency (2015). XMM-Newton. url: http://sci.esa.int/xmm-newton/.
— (2016a). Athena mission summary. url: http : / / sci . esa . int / cosmic - vision / 54517 - athena/.
— (2016b). XMM-Newton Users Handbook. url: https : / / xmm - tools . cosmos . esa . int / external/xmm_user_support/documentation/uhb/epic.html.
— (2017). Background Analysis. url: https : / / www . cosmos . esa . int / web / xmm - newton / background#EPIC.
Fabian, A. C. (2005). “Broad Iron Lines in AGN and X-Ray Binaries”. In: Astrophysics and Space Science 300.1, pp. 97–105. doi: 10.1007/s10509-005-1203-x.
Fabian, A. C., K. Iwasawa et al. (2000). “Broad Iron Lines in Active Galactic Nuclei”. In: Public- ations of the Astronomical Society of the Pacific 112.775, p. 1145. url: http://stacks.iop.
org/1538-3873/112/i=775/a=1145.
Fabian, A. C., M. J. Rees et al. (1989). “X-ray fluorescence from the inner disc in Cygnus X-1”. In:
Monthly Notices of the Royal Astronomical Society 238.3, p. 729. doi: 10.1093/mnras/238.3.
729. eprint: /oup/backfile/Content_public/Journal/mnras/238/3/10.1093/mnras/238.
3.729/2/mnras238-0729.pdf. url: +%20http://dx.doi.org/10.1093/mnras/238.3.729.
Falocco (2017). “The XMM deep survey in the CDF-S”. In: Astronomy & Astrophysics.
Falocco, S. et al. (2013). “The XMM Deep survey in the CDF-S - V. Iron K lines from active galactic nuclei in the distant Universe”. In: A&A 555, A79. doi: 10.1051/0004-6361/201321083. url:
https://doi.org/10.1051/0004-6361/201321083.
Gandhi, P. (2005). “X-ray studies of active galactic nuclei”. In: Asian Journal of Physics 13, pp. 90–107.
Georgantopoulos, I. et al. (2013). “The XMM deep survey in the CDF-S - IV. Compton-thick AGN candidates”. In: Astronomy & Astrophysics 555, A43. doi: 10.1051/0004-6361/201220828.
url: https://doi.org/10.1051/0004-6361/201220828.
Iwasawa, K. et al. (2015). “The XMM deep survey in the CDF-S - VIII. X-ray properties of the two brightest sources”. In: A&A 574, A144. doi: 10.1051/0004- 6361/201425086. url:
https://doi.org/10.1051/0004-6361/201425086.
Kuntz, K. D. and S. L. Snowden (2008). “The EPIC-MOS particle-induced background spectra”.
In: Astronomy & Astrophysics 478, pp. 575–596. doi: 10.1051/0004-6361:20077912.
Lamer, G. et al. (2003). “X-ray spectral variability of the Seyfert galaxy NGC 4051”. In: Monthly Notices of the Royal Astronomical Society 338.2, p. 323. doi: 10.1046/j.1365-8711.2003.
06005.x. eprint: /oup/backfile/Content_public/Journal/mnras/338/2/10.1046/j.1365- 8711.2003.06005.x/2/338- 2- 323.pdf. url: +%20http://dx.doi.org/10.1046/j.1365- 8711.2003.06005.x.
Lira, Paulina et al. (2009). “X-Ray, Optical, and Near-IR Long-Term Monitoring of AGN”. In: Pro-
ceedings of the International Astronomical Union 5.S267, pp. 90–95. doi: 10.1017/S1743921310005624.
Luo, B., F. E. Bauer et al. (2011). “The Chandra Deep Field-South Survey: 2 Ms Source Catalogs”.
In: The Astrophysical Journal 179.1, p. 19.
Luo, B., W. N. Brandt et al. (2017). “The Chandra Deep Field-South Survey: 7 Ms Source Cata- logs”. In: The Astrophysical Journal Supplement Series 228.1, p. 2. url: http://stacks.iop.
org/0067-0049/228/i=1/a=2.
Matt, G. et al. (2005). “A changing inner radius in the accretion disc of Q0056–363?” In: Astronomy
& Astrophysics 435.3, pp. 857–861. doi: 10.1051/0004-6361:20042581. url: https://doi.
org/10.1051/0004-6361:20042581.
NASA’s HEASARC. Heasarc. url: https : / / heasarc . gsfc . nasa . gov / docs / xmm / about _ satellite.html.
National Aeronautics and Space Administration (2017). Xspec Manual. url: https://heasarc.
gsfc.nasa.gov/docs/xanadu/xspec/manual/manual.html.
Paolillo, M. et al. (2004). “Prevalence of X-Ray Variability in the Chandra Deep Field-South”. In:
The Astrophysical Journal 611.1, p. 93.
Peterson, B. M. et al. (2004). “Central Masses and Broad-Line Region Sizes of Active Galactic Nuclei. II. A Homogeneous Analysis of a Large Reverberation-Mapping Database”. In: The Astrophysical Journal 613.2, p. 682.
Peterson, Bradley M. (1997). An introduction to active galactic nuclei. Cambridge University Press.
isbn: 0521479118.
Ponti, G. et al. (2012). “CAIXA: a catalogue of AGN in the XMM-Newton archive. III. Excess variance analysis”. In: Astronomy and Astrophysics 542, A83, A83. doi: 10 . 1051 / 0004 - 6361/201118326. arXiv: 1112.2744 [astro-ph.HE].
Ranalli, P. et al. (2013). “The XMM deep survey in the CDF-S. III. Point source catalogue and number counts in the hard X-rays”. In: Astronomy & Astrophysics 555, A42, A42. doi: 10.
1051/0004-6361/201321211. arXiv: 1304.5717 [astro-ph.HE].
Reynolds, C.S. (1998). “Compton reflection and iron fluorescence in AGN and GBHCs”. In: url:
https://arxiv.org/abs/astro-ph/9810018.
Risaliti, G., M. Elvis, S. Bianchi et al. (2010). “Chandra monitoring of UGC 4203: the structure of the X-ray absorber”. In: Monthly Notices of the Royal Astronomical Society: Letters 406.1, p. L20. doi: 10.1111/j.1745-3933.2010.00873.x. eprint: /oup/backfile/content_public/
journal/mnrasl/406/1/10.1111/j.1745- 3933.2010.00873.x/2/406- 1- l20.pdf. url:
+%20http://dx.doi.org/10.1111/j.1745-3933.2010.00873.x.
Risaliti, G., M. Elvis and F. Nicastro (2002). “Ubiquitous Variability of X-Ray-absorbing Column Densities in Seyfert 2 Galaxies”. In: The Astrophysical Journal 571.1, p. 234. url: http : //stacks.iop.org/0004-637X/571/i=1/a=234.
Risaliti, Guido and Martin Elvis (2004). “A Panchromatic View of AGN”. In: Supermassive Black Holes in the Distant Universe. Ed. by Amy J. Barger. Dordrecht: Springer Netherlands, pp. 187–
224. isbn: 978-1-4020-2471-9. doi: 10.1007/978-1-4020-2471-9_6. url: http://dx.doi.
org/10.1007/978-1-4020-2471-9_6.
Singh, K.P. (2013). “An X-ray view of quasars”. In: Bulletin of the Astronomical Society of India 41.2, p. 137.
Sobolewska, M. A. and I. E. Papadakis (2009). “The long-term X-ray spectral variability of AGN”.
In: Monthly Notices of the Royal Astronomical Society 399.3, p. 1597. doi: 10.1111/j.1365- 2966.2009.15382.x. eprint: /oup/backfile/Content_public/Journal/mnras/399/3/10.
1111/j.1365-2966.2009.15382.x/3/mnras0399-1597.pdf. url: +%20http://dx.doi.org/
10.1111/j.1365-2966.2009.15382.x.
Szokoly, G. P. et al. (2004). “The Chandra Deep Field-South: Optical Spectroscopy. I.” In: The Astrophysical Journal 155, pp. 271–349. doi: 10.1086/424707. eprint: astro-ph/0312324.
”The Nobel Prize in Physics” (2002). url: http : / / www . nobelprize . org / nobel _ prizes / physics/laureates/2002/.
The Scipy community (2014). scipy.stats.linregress. url: https://docs.scipy.org/doc/scipy- 0.14.0/reference/generated/scipy.stats.linregress.html.
Xue, Y. Q. et al. (2011). “The Chandra Deep Field-South Survey: 4 Ms Source Catalogs”. In: The Astrophysical Journal 195, p. 10.
Yang, G. et al. (2016). “Long-term X-Ray Variability of Typical Active Galactic Nuclei in the Distant Universe”. In: The Astrophysical Journal 831.2, p. 145.
Zoergiebel, Felix (2009). ploterr. url: https://se.mathworks.com/matlabcentral/fileexchange/
22216-ploterr?.
