Determining the temperature structure of the hot plasma halo around M87 with XMM-Newton

Contents

1 Introduction
1.1 Clusters of galaxies
1.2 The cooling flow problem
1.2.1 Cooling flows before XMM-Newton and Chandra
1.2.2 The fall of the cooling flow model
1.2.3 Emerging new models for cooling core clusters
1.3 Previous observations of M
1.4 Scientific goals of this work

2 Observational details
2.1 XMM-Newton
2.2 The present observation of M

3 Data analysis methods
3.1 X-ray brightness profile
3.2 Temperature profile
3.2.1 Determining the temperature from color maps
3.2.2 Determining the temperature from spectral fitting

4 Results and interpretation

5 Summary and Outlook

1.Introduction

The launch of the advanced X-ray observatories Chandra and XMM-Newton revolutionized our view of the X-ray sky providing valuable new and more detailed data which answered some scientific questions and raised many more, shaking some of the main physical models representing our understanding of X-ray astrophysics. Among the most disputed problems posed by the new observation capabilities of these two satellites is the so-called ”cooling-flow problem” in galaxy clusters. XMM-Newton and Chandra data proved that the hot gas at the center of many galaxy clusters does not cool below a certain threshold temperature although its cooling time is significantly lower than the age of the Universe, thus challenging scientists to find a source of heating which would be so well-tuned as to give the gas just the right amount of energy to prevent cooling below the observed limit. Known as the central dominant galaxy of the closest cooling-core cluster, as well as the first and one of the brightest extragalactic X-ray sources ever discovered, M87 is a perfect object for detailed studies which may indicate the origin and properties of the inferred heat input halting the cooling flow. The aim of the current work is to analyse the most recent, deepest observation of M87 with XMM-Newton to date and produce detailed, accurate and objective temperature maps which may enable us to better understand the heating mechanism as well as the physics of the intracluster medium in the vicinity of the central galaxy in general.

1.1 Clusters of galaxies

Contrary to what the name suggests, clusters of galaxies are not merely collections of galaxies grouped close together; they are well-defined, connected structural entities. In fact, only roughly 2% of the total mass of such a cluster is contained in galaxies while 11% consists of hot intracluster gas and 87% of the mass is in dark matter. The hot, highly ionized intracluster medium (ICM) has temperatures of the order of tens of millions of degrees, radiating therefore primarily in the X-ray domain. Images of galaxy clusters in this wavelength regime provide a very different picture from that seen in the optical range, since the intracluster gas fills the whole cluster volume and therefore the diffuse X-ray emission from the ICM traces the cluster structure contiguously. X-ray imaging information and X-ray spectra are thus crucial elements in understanding the astrophysics of galaxy clusters.

1.2 The cooling flow problem

By emitting X-rays, the intracluster gas loses energy and therefore should cool if no other source of heating is available. The intensity radiated in the process of thermal bremsstrahlung can be derived to be approximately proportional to the square of the gas density and the square root of the gas temperature. Therefore, the cooling is largest in the central parts of clusters where the density is high. Observations of clusters with the Uhuru satellite first showed that the mean cooling time of the gas in some cluster cores which show a very peaked density profile is close to the Hubble time (Lea et al 1973 [30]), leading scientists to develop the cooling flow model in an attempt to describe the effects of significant cooling of the central gas (Fabian and Nulsen 1977 [19], Mathews and Bregman 1978 [33]). Later observations with Einstein and EXOSAT set the estimated percentage of clusters which host a cooling core to a significant 70-80% [15], which emphasizes the importance of understanding cooling flows in cluster astrophysics.

1.2.1 Cooling flows before XMM-Newton and Chandra

The cooling flow model, based on early observations mainly with the Uhuru and Einstein observatories, states that upon cooling (which takes place within the so-called cooling radius where the cooling time is equal to or less than the Hubble time) the gas condenses to cold blobs which sink or ”flow” through the remaining hotter atmosphere to the very center of the cluster which is often associated with a central dominant galaxy. The cooling radius is small compared to the extent of the entire cluster such that the gas loss through cooling is negligible compared to the total amount of gas in the ICM, however the cooling radius can be as large as some hundred kiloparsecs, allowing the cooling region to be resolved by X-ray satellites. It was inferred that

through the cooling flow, the central dominant (cD) galaxy is still growing at the present time. However, no significant signatures of star-formation were observed in cD’s, therefore it was assumed that either the gas from the ICM cools into optically thick objects [16] or that mostly small stars are produced which show much less observational evidence of star formation, which would imply a modified stellar initial mass function in the central galaxy [20].

The cooling flow model implies that each radial zone in the cooling flow region comprises different plasma phases (the cooling blobs), covering a wide range of temperatures, and that there is no energy exchange between these different phases, otherwise the blobs would be mixed with the hotter ICM before condensing to the center of the cluster. Consequently, heat conduction must be suppressed, which was presumed to be due to cluster-scale magnetic fields.

According to the cooling flow model, the mass deposition rate - that is, the mass of cool gas which sinks to the core per unit time - can be calculated from the relation

illustration not visible in this excerpt

whereLcool is the luminosity associated with the cooling region, usually of the order of 10% of the total luminosity, and T is the temperature of the gas at the cooling radius. In deriving this formula, it was assumed that the luminosity associated with the cooling region is due to the thermal emission from the cooling gas and the PdV work done on the compressed cooling gas. Typical mass deposition rates calculated from this formula are in the range of hundreds of solar masses per year [16].

Observational evidence for the cooling flow model came mostly from the X-ray domain. Strongly peaked surface brightness profiles confirmed by ever higher resolution detectors with the launch of ROSAT indicated very short cooling times. Low temperature components were detected with the Solid State Spectrometer onboard the Einstein satellite in the Perseus and Virgo clusters (Canizares et al 1979 [12], Lea et al 1982 [29]). Further evidence from spectral data was the correlation between the X-ray luminosity, X-ray emission-weighed temperature, metallicity and inferred mass deposition rate, such that strong cooling flows were shown to have a systematically higher luminosity and metallicity at a given temperature compared to weaker ones (Fabian et al 1994 [17]).

However convincing the X-ray data, no other wavelength regime showed strong evidence in favor of the cooling flow model. The cD’s in cooling cores were found to show anomalies in the optical band, such as the emission of diffuse blue light (McNamara and O’Connell 1989 [36]), and in the radio band, such as high Faraday rotation or depolarization (Owen et al 1990 [42]), but the strength of these signatures implied mass deposition rates one to two orders of magnitude smaller than those calculated from the cooling flow model. Many scientists were thus still skeptical of the cooling core model and it was the launch of Chandra and XMM-Newton that proved them entitled to their skepticism.

1.2.2 The fall of the cooling flow model

High-resolution spectra of cooling core clusters obtained with the Reflecting Grating Spectrometer (RGS) onboard XMM-Newton showed a different picture from what the cooling flow model had predicted. The temperature of the ICM did show a decrease at the center of cooling core clusters, but no evidence of gas cooling below a certain threshold, usually one half to one third of the virial temperature, was found. The work of Peterson et al [44],[45] clearly showed that the cooling flow model also overpredicted the line emission from low-temperature ions, thus giving a very different spectrum to what the XMM RGS measured. For example, in clusters with a virial temperature of 7-10 keV the predicted Fe XVII-XXII lines were not observed; for virial temperatures of 4-7 keV, Fe XVII-XXI were also not observed while for clusters with temperatures of 2-4 keV the Fe XVII-XVIII lines were much weaker than the predictions of the cooling flow model. A visual example of this is given in Figure 1.1.

Some solutions were proposed to reconcile these findings with the cooling flow model, among which the presence of a cool absorber along the line of sight which would account for the deficiency in soft X-ray detection, or an inhomogeneous distribution of the metal abundances in the ICM resulting in the suppression of low-temperature line emission (Fabian et al 2001 [18]). All of these explanations however were refuted by the available data.

1.2.3 Emerging new models for cooling core clusters

Since the cooling flow model could not explain the spectra obtained with the most recent X-ray satellites, new perspectives came to the attention of astrophysicists. Based on detailed spectra, especially those obtained from M87 at the center of the nearby Virgo cluster, it was inferred that the mass deposition rates must in fact be at least an order of magnitude less than the value predicted by the cooling flow model. For this to be possible, an extra

illustration not visible in this excerpt

Figure 1.1: RGS spectra of the cooling region in Abell 1835 from Peterson et al 2001 [44]. The classical cooling flow model in blue clearly predicts line emission not seen in the data. A cooling flow model with a forced cutoff at 2.7 keV in green and an isothermal model with a temperature of 8.2 keV shown in red are much better fits to the observation

source of heating must exist at the cluster center which was not previously included in the energy balance in the calculation ofLcool.

Several forms of possible heat input have been discussed, among which heat conduction from the ICM outside the cooling radius, reconnection of magnetic field lines and energy output from the active galactic nucleus (AGN) of the central galaxy. The constraints to the heating source in order to reproduce observations are that the energy input must be high enough to balance the cooling flow and very fine-tuned for each cluster such as to maintain the peaked density profile and prevent overheating which would cause the gas to flow outwards from the central cluster region [8]. Moreover, the heating must be ”gentle” enough to maintain the observed steep gradients of heavy elements such as iron undisturbed.

The most promising heating source under the given constraints is the central AGN, where the required fine-tuning can be accomplished through a feedback mechanism between the gas accreted onto the central AGN massive black hole and the AGN output power. One possibility of such mechanism discussed by Churazov et al (2002) [13] involves classical Bondi accretion theory which provides a link between the accretion rate and the gas parameters in the ICM, notably the gas entropy. As discussed by Kaiser and Binney (2003) [28], an AGN outburst produces bubbles of high-entropy gas near the bottom of the gravitational potential well, disturbing the stable hydrostatic equilibrium characterized by a stratified specific entropy profile. Following this disturbance, the accretion rate is altered and the AGN goes into a quiescent stage, during which the specific entropy is redistributed through the ICM by the buoyant bubbles and the entropy excess is radiated away. When the excess entropy has been lost, the density of the gas at the bottom of the well becomes again extremely large and a further outburst of the AGN is stimulated when some of this dense gas accretes onto the central massive black hole, starting the next cycle of the system.

The influence of an AGN on the intracluster medium was considered and modeled as early as 1988 by Böhringer and Morfill [9] who looked into the effects of associate cosmic ray production and 1993 and 1995 by Binney and Tabor [5] who discussed mechanical energy input via the AGN jet. Apart from the achievable fine-tuning necessary, AGNs are attractive candidates for heat sources also because cooling flow clusters were shown to be statistically more likely to contain radio emission indicating recent black hole activity (Burns 1990 [11]), emission which often coincides with cavities in the X-ray images suggesting interaction between the AGN and the ICM. Based on the sizes of these X-ray cavities and the work needed thus to inflate them, it was estimated by Birzan et al (2004) [6] that each AGN outburst introduces between 1058 and 1061 ergs into the ICM, sufficient energy to stop the cooling flows provided that an AGN outburst occurs roughly every 108 years.

A further reason which makes AGNs attractive as cooling flow quenchers is that, according to recent simulation by Springel et al (2005) [48], the formation and properties of the massive cD galaxies such as M87 cannot be modeled in agreement with observations unless AGN feedback is included. In the absence of a ”cure” for the cooling flow problem, the most massive galaxies resulting from the simulations are too bright, too blue and diskdominated. With cooling flows suppressed by radio AGNs, this discrepancy to the observations is solved and the massive galaxies formed become less massive, red and elliptical.

It was observed that entropy profiles flatten towards the inner parts of the cluster, reaching a plateau at about 10 keV cm−2 independently of the individual cluster properties [50]. This can be used as an important tool for constraining the properties of the heating source. The duration of a heating outburst, independently of what produces it, is estimated to 107 years, the entropy input of an outburst to 10-20 keV cm−2, the total energy deposited to 1060 ergs, and the time between two events to 108 years. The latter two quantities agree surprisingly well with the values computed from the AGN-related X-ray cavities, mentioned above. The shape of entropy profiles clearly favors heating by a central AGN also because an AGN outburst with a constant kinetic power output is a natural way to produce a constant entropy boost throughout the cluster, which combined with the pure cooling model reproduces very well the observed radial entropy trend. Detailed entropy maps are likely to contain information which would allow further constraints on the outburst history and nature and properties of the heat source.

A detailed model based on an observational framework presented recently by Voit and Donahue [50] employs a three-stage heating process. In the innermost parts of the core, outflow-driven shock heating is the main mechanism. The authors argue that, since the shocks are outflow-driven, the iron gradient should remain undisturbed. If the flow is still supersonic when the kinetic outflow shuts off, further heating will occur beyond the outflow-shock region up to the point where the flow becomes subsonic. The third stage of heating effective for even larger distances is through rising bubbles inflated at the supersonic to subsonic transition radius. These bubbles are believed to correspond to the cavities observed in X-ray images. As the bubbles which are less dense than the surrounding medium rise, the gas they displace sinks and thermalizes the gained gravitational energy.

Clearly, as the arguments aligned above show, cooling core clusters present some the most intricate phenomenology found in modern astrophysics. While there is significant observational evidence arguing for the scenario of AGN heating of the ICM, the exact feedback and heat transport mechanisms are still grounds for scientific debate. Due to the complexity of these systems, progress in understanding the physics of cooling-core clusters relies in great part on the results of detailed observations with the help of which we may hope to unveil episodes of the ICM history. With its proximity and brightness, M87 is an ideal candidate for such detailed observations, on which we will elaborate in the following section.

1.3 Previous observations of M87

The giant elliptical galaxy M87 is the central dominant galaxy in the nearby Virgo cluster, the distance to it having been estimated at 17-20 kpc [23]. As it is among the closest and most luminous extragalactic objects, M87 has been intensely studied in all wavebands and with all the available observatories, becoming one of the main drivers for the development of extragalactic astrophysics. M87 was among the first known extragalactic radio sources and was the first X-ray source identified outside the Milky Way. Observations in the X-ray domain with the Einstein satellite showed a peaked brightness profile of the extended emission from the halo of intracluster gas surrounding M87 indicative of a cooling flow which, using the cooling flow model, gave an estimated mass deposition rate of about 10 M /year [49]. A subsequent observation of M87 with XMM-Newton was among the most important data sets which lead to contradictions with the cooling flow model and motivated the need for a heat input into the ICM. As seen in the previous section, AGNs are considered an attractive candidate for providing such heat input, thus it comes as no surprise that M87 does host an active galactic nucleus, whose one-sided jet has been observed in detail at several wavelengths, namely in the optical [47], radio [43] and X-ray bands [46] [40]. The jet and the unseen counterjet are believed to be the cause for the observed complex system of radio lobes, on the basis of which Owen, Eilek and Kassim (2000) [41] showed that the mechanical power input from the AGN is more than sufficient to compensate for the energy loss through X-ray radiation. The proximity of M87 furthermore allowed the disk surrounding the AGN central black hole to be resolved with the Hubble Space Telescope (HST) providing an estimate of its mass as 3.2 × 10 9 M [25]. Yet another feature which makes M87 a very attractive candidate particularly for X-ray studies is that the mean temperature of the gas surrounding it is about 2-2.5 keV, rather low compared to most ICMs, therefore M87 offers a very rich line spectrum from which distributions of various heavy elements in the ICM can be determined to good accuracy and which also provides the basis of good temperature diagnostics using temperature-sensitive emission lines.

Having hopefully convinced the reader of the importance of detailed studies of M87 in resolving the cooling flow puzzle, we will further in this section focus on the results of the first observation of M87 with XMM-Newton (complemented also by Chandra data in some aspects), which will serve as a guide to our analysis of the second, deeper observation to be discussed in this paper. M87 was observed with XMM-Newton on June 19th, 2000 for 75.6 ksec, of which effective unflared exposure times were about 39 and 25.9 ksec for the MOS and PN detectors respectively (please refer to the next chapter for more technical details about the XMM-Newton detectors). The spectra of the nucleus and X-ray knot of the jet were found to be well fitted by powerlaws with slopes of 2.2 ± 0.2 and 2.5 ± 0.4 respectively [7]. The extended emission from the gas halo of M87 was roughly spherically symmetric with the exception of two localized surface brightness enhancements to the SW and E of the nucleus, referred to as the southwestern and eastern arms. Excluding these regions from the analysis, Böhringer et al (2001) [7] divided the halo into concentric rings and determined their temperature and abundance in heavy elements by spectral fitting. Outside a radius of 1 arcminute, it was found that a one-temperature model fits the data very well therefore it was deduced that the ICM is single-phase locally, in contradiction with the multiphase cooling flow model. An upper limit to the mass deposition rate was calculated to below about 1 M /year, at least an order of magnitude below the expectations from the cooling flow model. Outside 1’, the abundances of most elements (Fe, Si, S) were found to decrease with radius, with the exception of oxygen which remained constant. This was interpreted as a variable rate of supernovae Ia and II (which produce these elements in different proportions) with distance from the center of M87 (Finoguenov et al 2002 [21]). Belsole et al (2001) [4] in turn focused on the structure of the E- and SW-arms and found that a two temperature model is needed here in order to explain the data. The lower temperature in the fits was about 0.9 keV and the higher temperature around 1.5-1.6 keV. Adding a third temperature however did not improve the fit further. Spectral analysis revealed that the abundances of Si, S and Fe are larger in the arms than in the surroundings, although for the E arm the effect was much more pronounced than in the SW arm. The two arms can be well seen by creating a mean energy map in the 0.9-1.1 keV energy range corresponding to the Fe L line complex. The arm structures show a lower mean energy than the ambient medium and ”patchy” substructures of very low mean energy can be seen. The arms so-seen in the X-ray image correspond to, but are not exactly coincident with the lobes seen in the radio map (see Figure 1.2), the mechanism generating the slight shift between the two being still unexplained.

illustration not visible in this excerpt

Figure 1.2: The radio map at 90 cm with superimposed contours of the MOS1 mean energy map [4]

A more in-depth study by Matsushita et al (2002) [34] confirmed that, apart from the arm regions, the gas halo around M87 has a single-phase temperature structure locally (except the very central 1’-2’) and that no signatures of components below 0.8 keV can be found in any of the analyzed spectra. In this analysis, a spectral deprojection method based on the assumption of spherical symmetry was employed. Matsushita et al [34] used several different tests to check the reliability of the results from spectral fitting, including fitting the spectrum only above 1.6 keV to eliminate uncertainties in the calculation of the Fe L line complex, studying the temperature-sensitive ratios of He-like and H-like Kαlines, as well as comparing different fitting codes (APEC and vmekal). All these methods gave a consistent deprojected single temperature profile (apart from the arm regions) which shows a positive gradient from 1.7 keV at a radius of 1’ to 2.5 keV at 15’, as can be seen in the graph in Figure 1.3. Including the arms and performing a two-temperature fit to the data, it was found that the low-temperature component has an almost constant temperature of about 1 keV at all radii and that the ratio of its emission measure to the emission measure of the hot component decreases strongly with radius, becoming zero to within the calculated uncertainty levels outside of 2’ (further analysis leading to a similar conclusion is provided also by Molendi (2002) [38]). Regarding elemental distribution in deprojected spectra, a second paper by Matsushita et al [35] found that a twotemperature fit needed in the inner 2’ region provided much higher element abundances than a one-temperature fit (this influence of a two-temperature fit in increasing the predicted abundances is discussed also by Molendi and Gastaldello (2001) [39]). A strong gradient was found for Si, S, Ar, Ca and Fe outside of 2’ (where a one-temperature model was used), while the deprojected profiles of these elements flattened inside this radius (where the two-temperature model was used). The oxygen profile however was much flatter, in agreement with Böhringer at al (2001). In contrast, Gastaldello and Molendi (2002) [24] find a significant gradient also for the light elementabundances, in particular oxygen. Matsushita et al [35] found the abundance pattern of O, Mg, Ca, Si and Fe to be typical of galactic stars, therefore it was suggested that precise measurements of elemental abundances in the ICM can be used to constrain the products of supernovae Ia and II, whose yields are not yet very accurately determined by the present models.

The detailed X-ray morphology of M87 as seen with XMM-Newton but also with Chandra and ROSAT is given in a paper by Forman et al (2005) [22].

illustration not visible in this excerpt

Figure 1.3: Deprojected radial profile of the temperature obtained with EMOS, by fitting the spectra of the whole energy band (bold black crosses), 0.7-1.3 keV (black diamonds), above 1.6 keV (black crosses), 1.8-2.1 keV (gray crosses), and 2.3-2.7 keV with and APEC model (gray diamonds). The solid line corresponds to the best fit regression line for the MEKAL model using the whole energy band of the EMOS [34]

The authors describe a rich variety of substructure on all the different scales covered by the three detectors. With the good resolution of Chandra, the inner part of M87 is revealed in great detail (Figure 1.4). Among the observed features are the X-ray jet extending to about 20” from the nucleus and showing a substructure consisting of at least 7 bright knots; several elliptical cavities surrounding the jet and its unseen counterpart, cavities which were coincident with radio emission at 6 cm; the X-ray bright core of M87 with a sharp edge at around 50”. Moreover, by combining Chandra, XMM-Newton and ROSAT data from which a model intensity had been subtracted to account for the spherically symmetric gas halo, further substructure could be identified in the form of the already discussed E and SW arms, a NW-SE asymmetry consisting of X-ray brightness enhancements in these directions, as well as several arcs (at 3.6’ and 8’) and a ring of enhanced emission (with a radius of 3’), interpreted as weak shocks (all these features are labeled in Figure 1.5). By modelling the parameters of the burst which could have caused the emission enhancement at 3’ (corresponding to 14 kpc for an as-sumed distance to M87 of 16 Mpc), Forman et al find that an energy deposit of 8×1057 ergs should have happened 107 years ago. Assuming the expansion of the jet and counterjet associated cavities seen in the Chandra image to have been subsonic, the authors also calculate the enthalpy of these cavities to be 1.1 × 1056 ergs, therefore certainly below the energy input into the shock. Therefore, it is argued that shocks are the most important channel of energy input into the ICM. The substructure observed in the arms seems to also be associated with the 14 kpc ring. Both arms brighten significantly at the position of the ring (the E arm shows a brightening as large as 25%), and beyond this ring the SW arm bifurcates and the radio intensity is increased between the two resulting filaments. Deeper observations are thus important in confirming the existance and further establishing the properties of shocks in the gas halo of M87.

illustration not visible in this excerpt

Figure 1.4: X-ray morphology of M87 seen with Chandra, as identified by Forman et al [22]

1.4 Scientific goals of this work

We aim to analyze the most recent and deepest observation of M87 with XMM-Newton, in order to further understand the substructure and heavy element abundance profiles in the gas halo. The first and most important

illustration not visible in this excerpt

Figure 1.5: X-ray morphology of M87 from the combined Chandra, XMMNewton and ROSAT data, processed to remove the spherically symmetric emission. The observed substructure is labeled according to [22] step in this direction is to understand the temperature distribution, from which entropy and pressure maps can be generated which are the best indications of existing substructure and especially shocks and cavities. A good understanding of the temperature maps is also needed for detailed analysis of the distribution of heavier elements in the ICM, since most emission lines are strongly temperature-dependent. We will thus present and compare in this work two independent methods for determining the temperature pro le in M87 and their results.

[...]