The Search for Diffuse Sources

Clusters of Galaxies

In the 1920's Edwin Hubble discovered that the Milky Way, far from being the entire universe was in fact an isolated island, one of many self-contained systems of stars. He went on to observe and classify these extragalactic sources, establishing the basic properties of what are now known as galaxies and providing the first accurate insight into the large scale structure of the universe. Hubble developed a classification scheme which is still used to separate the main classes of galaxies; spiral, elliptical and irregular, demonstrated by Figure 1.1.

  

Figure: Galaxies vary greatly in size, composition and morphology; Left: Spiral galaxy NGC4535, known as the `Southern Pinwheel'. Centre: Giant elliptical galaxy NGC4486 in the Virgo cluster. Right: Irregular Spiral galaxy IC4182. Images from http://www.noao.edu/

Almost everything we know about the large scale structure of the universe comes from studying the distribution of galaxies. Galaxies are not distributed evenly throughout the universe, rather they are located in gravitationally bound aggregations, called clusters. There are two general categories of clusters, rich and poor, which depend on the number of galaxies they contain. A typical poor cluster such as the Local Group (which contains the Milky Way) contains $\sim 10 - 100$ galaxies, whereas a much richer cluster such as Coma can contain $\sim 1000$.

Rich Clusters

In rich clusters, the number density of galaxies is low at the edges of the cluster, and increases smoothly towards a central peak. Usually galaxies are isolated systems; however, the large number of galaxies orbiting around the central region in a rich cluster means that there are frequent gravitational interactions between the galaxies in this region.

During these interactions there is dynamical friction between the gas of the galaxies. The stars rarely undergo collisions because the average separation between stars is about 10⁷ times their diameter. This dynamical friction distorts the shape of the galaxies and can strip away their outer regions. As a result, much of the interstellar matter is removed and star formation in the galaxies decreases. On a larger scale, the intracluster medium (ICM) is heated. These processes may result in spiral galaxies being converted into elliptical galaxies, which could explain the population bias in the type of galaxies $(\sim 90\%$ elliptical and S0) found in the dense central region of rich clusters [Snow1991]. An alternative explanation for this bias is that since disks take longer to form than the elliptical components of galaxies, the local density of the region may play a role. For example, interactions between galaxies could halt the disk formation for galaxies in high density regions [Dressler1980].

Another effect of collisions between galaxies is the formation of a giant elliptical galaxy, the central dominant (cD), in the centre of the cluster. Observations show that the hot gas and galaxies are in hydrostatic equilibrium within a common cluster potential. So both are moving under the influence of the deep gravitational potential well at the centre of the cluster. Large galaxies orbiting the centre gradually move towards this potential well, and may eventually merge with other galaxies, creating a single giant elliptical galaxy which then attracts more galaxies and hence becomes more massive. This model of `galaxy cannibalism' is one of several which explain the observed properties of rich clusters. A comprehensive review can be found in [Dressler1984].

Multiwavelength Observations of Clusters

Multiwavelength studies have been crucial in understanding the nature of clusters of galaxies. The combined information from radio, X-ray and optical studies is necessary to get an overall picture of clusters of galaxies and is an important part of this project. One technique used is to overlay images observed at different wavelengths. For example, an optical image showing individual galaxies which appear as point sources may be overlaid with radio contours to show the full extent of the sources. Alternatively, a radio image of a cluster could be overlaid with smoothed X-ray contours to show how the radio sources are aligned with respect to the density of the hot ICM. The following sections outline the respective benefits that observations at radio, X-ray and optical wavelengths have to offer in the context of studying clusters of galaxies.

Radio Observations

Radio emission from clusters is predominantly synchrotron emission from individual galaxies within the cluster. Synchrotron radiation is the result of highly relativistic electrons being accelerated as they spiral around magnetic field lines. The radiation is strongly beamed into a forward cone in the direction the electrons are travelling. This is the main process of emission from non-thermal plasmas. Most galaxies in a given cluster are radio quiet. In other words, the radio power contributes less than $1\%$ of their total output and so they are not detectable at radio wavelengths [Rowan-Robinson1996]. Cluster galaxies detected at radio wavelengths tend to be either the most massive ones, which have a higher overall power output, or those with dramatically enhanced non-thermal emission, such as radio galaxies with large synchrotron jets. The term radio galaxy is used to describe a wide range of enhanced radio emission associated with, for example, cD galaxies, Seyfert galaxies, and quasars. As well as strong, discrete emission, it is sometimes possible to observe much fainter regions of diffuse emission. These regions, which are the principal subject of this project, are known as halo or relic sources. They will be discussed further in section 1.2.

X-Ray Observations

X-ray emission comes predominantly from Bremsstrahlung, or free-free, emission in the ICM. Bremsstrahlung occurs when electrons travel through a hot, ionized gas such as the ICM. The electrons are accelerated in the Coulomb fields of the ions, converting kinetic energy into radiation. This is the main emission process from thermal plasmas.

  

Figure: X-ray image of the Coma cluster

The ROSAT All-Sky-Survey [Voges et al.1999], with energy range 0.1-2.4 keV, is the main source of X-ray data for research into clusters of galaxies, as it includes low luminosity clusters, and those at high redshifts. X-ray data are important because the X-ray emission is directly linked to the physical parameters of the clusters. It provides information about dynamics of the ICM and also the structure and evolution of clusters on a large scale. X-ray surveys are less susceptible to observational bias than optical surveys and the signal to noise ratio of detections is generally higher [Giacconi1993]. The main cluster parameters that can be directly observed or determined from X-ray observations are luminosity, temperature, morphology of intracluster gas and the mass within a given radius.

Optical Observations

Traditionally most observations have been carried out at optical wavelengths and it was on the basis of optical information (galaxy density within a given radius) that[Abell1958] defined clusters. Optical studies are still the main source of information about the type and distribution of galaxies within a cluster. Radiation at optical wavelengths comes predominantly from stars in the galaxies. Light from star formation regions is dominated by young, hot stars and so appears blue whereas light from regions of older stars appears at the red end of the spectrum.

  

Figure: Optical CCD image of Coma

The importance of multiwavelength observations is highlighted in its application to one of the central concerns of modern astrophysics, the `missing matter' problem. From optical observations we obtain the distribution and mass of galaxies within the clusters. From X-ray observations, the distribution and mass of the hot ICM gas can be found. Recent studies show that the intracluster hot gas constitutes $\sim15 \%$ of the total mass of clusters and visible galaxies constitute $\sim3 \%$ [Einasto and Einasto1999]. Hence $\sim 80\%$ of the mass of a cluster must be some form of baryonic (eg. planets, cold gas) or non-baryonic (eg. axions, neutralinos) dark matter [Bothun1998].

   
Cooling Flows and Cluster Mergers

Observations of the cosmic background radiation show a high degree of isotropy, which implies the density distribution in the early universe was essentially uniform. However, small density perturbations would grow, forming clouds of gas. The clouds of gas condense, getting hotter as they collapse under their own gravity. The gas then cools, forming the stars, galaxies and clusters that we see today (the order that these form in depends on whether a `top-down' or `bottom-up' model of cosmological evolution is correct). In the case of clusters, the gas cools very slowly and, after the cluster has formed, continues to lose energy by emission of X-ray radiation [Fabian1994]. The cooling of the gas at the centre of the cluster means the gas density must rise to maintain pressure equilibrium. The only way this can occur is for gas to flow inwards, establishing what is called a cooling flow. Cooling flows are quite common and have been found to occur in about $70\%$ of clusters [Edge et al.1992].

The gas in a cooling flow cluster is cooling out of the hot phase $(\sim 10^6 - 10^8$ K) at a rate of hundreds of solar masses per year, over several billion years. The hot phase is detectable due to the X-ray emission. However, once the gas cools below $\sim 10^4$ K it can no longer be detected by its thermal emission (although there could still be spectral line emission) and eventually becomes baryonic dark matter [Jaffe1991]. X-ray emission from the hot phase causes the characteristic observational evidence for a cooling flow, an X-ray surface brightness distribution sharply peaked at the cD galaxy [Edge et al.1992].

The disruption of the cooling flow structure is an indication of a cluster merger. Mergers are massive dynamic events, evidence for which is seen in $20 \% - 30 \%$ of clusters [Edge et al.1992,Jones and Forman1990]. Evidence for cluster mergers comes from studying cluster substructure and morphology, using X-ray or optical data. The disruption of the intracluster medium (ICM) and the presence of a high temperature gas during a cluster merger has been predicted by simulations [Schindler and Müller1993,Burns et al.1994,Roettiger et al.1997] and confirmed by X-ray observations [Honda et al.1996,Knopp et al.1996,Markevitch et al.1998]. While not all mergers disrupt cooling flows, the clusters without cooling flows detected so far all appear to have undergone recent mergers.

   
Diffuse Radio Sources

Relic and halo sources are diffuse radio sources which are not clearly associated with a particular optical host. They are characterised by a low surface brightness, large extent $\sim 1$ Mpc and a steep spectral index up to $(\alpha \le -1)$. The lifetime of diffuse sources is estimated to be of order 10⁸ years.

In the literature, a source with these properties that is located near the centre of a cluster is referred to as a halo, whereas a source near the cluster boundary is called a relic. Projection effects limit the usefulness of this distinction and are a problem whenever the three dimensional nature of a object is inferred from its projection onto a two dimensional surface. For example, if, on a two dimensional image of a cluster, the diffuse source appears near the outer boundary it must also be near the `real' boundary of the cluster in three dimensions. The problem arises when the diffuse source appears near the centre of the two dimensional projection. In that case, the source could be contained anywhere within the cluster, including near the boundary. In this report the two types will not be distinguished and the term `diffuse source' used to refer to both.

Properties of Clusters That Contain Diffuse Sources

Since diffuse sources appear to occur only in clusters of galaxies, the properties of the clusters in which they have been found may give clues to their origin and evolution. The clusters which contain diffuse sources are characterised by the following properties:

• high X-ray luminosity $(L_X \sim 10^{44} - 10^{45}$ erg s^-1) [Röttgering et al.1997]
• high temperature
• no cooling flow
• high velocity dispersion of the galaxies
• widespread distribution of hot intracluster gas

These are empirical results, so it is quite possible that diffuse sources could occur in clusters with different properties.

There are several other properties of the clusters that contain diffuse sources which are not as well established. These include a higher than expected number of radio sources and the presence of head-tail sources. Head-tail, or narrow-angle-tail (NAT), sources consist of a higher luminosity `head' and radio jets bent back at extreme angles giving the impression of a wake. The standard explanation for this morphology has been that it is due to the motion of the galaxy through the ICM. However, the observation by [Bliton et al.1998] that NAT sources generally occur in dynamically evolving clusters with X-ray substructure suggests that they may be the result of ICM turbulence, or `cluster weather', that accompanies cluster-subcluster mergers. It is probable that both of these mechanisms apply.

   
Previously Studied Diffuse Sources

Until recently there were only $\sim15$ known diffuse sources (reviewed in[Feretti and Giovannini1995]). However, new research by [Giovannini et al.1999] has increased the number to $\sim30$. Some of the sources which have been classed as diffuse sources, are shown in Figure 1.4.

  

Figure: Radio/optical overlays of six of the previously studied diffuse sources. Clockwise, from top left; A85, A520, A773, A2744, A2254, A2163. A85 was first reported by [Slee and Reynolds1984]. A2163 was reported by [Herbig and Birkinshaw1994]. A520, A773, A2254 and A2744 have all been found recently by [Giovannini et al.1999]. Contours in each image are at 0.9, 1.35, 2, 4, 8, 16, 32, 64, 128, 256 mJy/beam

The Origin of Diffuse Sources

Early Models for the Origin of Diffuse Sources

The origin of diffuse sources has remained a mystery since the discovery of the first and now canonical example in the Coma cluster [Large et al.1959], mainly due to the small number of relics that have been observed. The main models put forward to explain the origin and evolution of diffuse sources are summarized below, along with some physical arguments for or against them. The model which currently seems most plausible is discussed in the next section.

The first model for diffuse sources was proposed by [Jaffe1977] who suggested that diffuse sources result from synchrotron emission from relativistic electrons originating from a single host galaxy. The electrons are emitted from the radio galaxy (a strong radio source, often with large radio lobes and jets) and disperse over time until they occupy the large volumes now observed. The most significant problem with this model is that by the time the electrons have travelled the distance required from the source galaxy, they would have radiated away most of their energy. [Dennison1980] proposed the secondary electron model to overcome the problem of energy loss. In this model, collisions between relativistic protons from radio galaxies and the protons in the ICM produce secondary electrons. These electrons then emit synchrotron radiation which forms the diffuse source. The secondary electron model is an improvement because protons lose less energy to synchrotron radiation than electrons do, and so could travel much further from their host galaxy than electrons could. The major argument against this model is that we would expect to observe more diffuse sources, since protons have a long lifetime [Hanisch1982].

The third model is the galactic wake model put forward by [Roland1981]. Galactic wakes are caused as a galaxy moves through the ICM. Bow and tail shocks are created by the movement and a turbulent wake forms. This results in an intracluster magnetic field being generated and relativistic electrons being accelerated. The aim of this model is to overcome the problem of the electrons having to travel such a large distance in a relatively short time. It opens up the possibility that the electrons are already spread over a large area and then are reaccelerated by the galactic wake. Although this model resolves the problems of the previous two, it appears that galactic wakes do not have enough energy to power the diffuse source [DeYoung1992] and so this is not a completely satisfactory explanation.

The Merger Hypothesis and the Role of Cooling Flows

It is the merger hypothesis that provides the most plausible explanation for the presence of diffuse sources, although it has not been extensively tested. The merger hypothesis suggests that interactions between the ICM of two clusters during a merger, may produce shocks which can reaccelerate relativistic electrons and amplify the magnetic field within the clusters. The physical size of diffuse sources (> 0.5 Mpc) relative to their short lifetimes may be explained as a result of the reacceleration of electrons which are already dispersed over a large region [Roettiger et al.1999]. For example, the dispersed electrons could be the low energy remnants of a radio galaxy. Cluster mergers, which can have collisional kinetic energy of $\sim 10^{57}$ J [Markevitch et al.1998], are a likely source of the energy required to create and power diffuse sources.

Evidence supporting the merger hypothesis is that there has been no cooling flow detected in any of the clusters observed to contain diffuse sources. As discussed above (in section 1.1.2), cooling flows are disrupted during a cluster merger, so the absence of a cooling flow suggests a recent cluster merger. However, [Feretti and Giovannini1995] note that mergers can not be the only contributing factor in the formation of diffuse sources, since cluster mergers are quite common, whereas diffuse sources are quite rare. The role of cooling flows (or the absence of them) in creating diffuse sources is discussed by [Burns et al.1992] who compare Coma, which has a large diffuse source and no cooling flow, with the Perseus cluster, which has a mini-halo and does have a cooling flow. Mini-halos are much smaller than other diffuse sources (about $1\% - 20\%$ of the size [Feretti and Giovannini1995]) and are confined to the cluster core, usually centred on the dominant galaxy. The magnetic field strength and the cluster kinetic energy density are highest at the cluster centre. In the absence of a cooling flow, the magnetic field will diffuse outwards and possibly produce a large scale diffuse source as in Coma. However, if a cooling flow is present, it will impede the diffusion of the magnetic field, resulting in a mini-halo confined to the cluster core as in Perseus.

A New Model for the Origin of Diffuse Sources

Although the merger hypothesis is currently accepted as the most plausible model, [Liang1999] points out that the correlation between the presence of a diffuse source and a cluster merger does not conclusively show that mergers cause diffuse sources. Instead it is proposed that this correlation could be a selection effect. Mergers disrupt the cooling flows in clusters, and clusters without a cooling flow are less likely to have a central dominant radio source. This means that a low surface brightness, diffuse source would be easier to detect in such a cluster than in a cluster with higher levels of radio emission. Hence diffuse sources could actually be present in most clusters but are not observed due to the difficulties in achieving the high dynamic range required and the fact that they have a relatively short lifetime.

[Liang1999] proposes a new model for the origin of diffuse sources in which merger activity and thermal electrons both contribute to the formation of diffuse sources. In this model, the relativistic particle necessary for synchrotron emission originate from the high energy tail of the ICM thermal electron Maxwellian distribution. [Liang1999] suggests that recent X-ray observations show that rich clusters are dense enough for the acceleration of thermal electrons to relativistic speeds to take place, a process that has previously been ruled out as a possible mechanism.

The Scope of this Project

Research in this field has been limited primarily by the difficulty of observing diffuse, low surface brightness sources. The small number of these sources detected so far has made it hard to characterize them, even in a general way. The field has evolved by looking for patterns and trends that may have a common physical origin.

Trying to model how diffuse sources are created and sustained is even more challenging. A general problem of any astrophysics research is that we can not observe the time development of most sources, as we only see them at a snapshot in time. For many types of sources, such as spiral galaxies, this is partially compensated for by the large numbers accessible to observation, each at different stages in their evolution. From these observations, a picture of the time evolution of the galaxies can be developed. Clearly that is not the case in this field, and any sense of time development must be obtained from estimates of the lifetime of the diffuse sources, and the characteristics of the phenomena that might cause them. Hence, the models are based on empirical evidence from what is still a relatively small sample.

The primary objective of this project is to further define what characterizes diffuse sources. We have radio observations of $\sim 130$ southern clusters which have not been extensively searched for diffuse sources. Due to the small number of diffuse cluster sources currently known, any new discoveries will be significant to the field. To help understand the nature of diffuse sources, and their environment, we will study both the statistical properties of a distance-limited cluster sample, and the detailed characteristics of a particular cluster which is known to contain a diffuse source (A3266).

In summary, our objectives are to;

• Create a distance limited catalogue of Abell southern clusters
• Identify any diffuse sources in this sample and study their properties
• Analyze the properties of clusters that contain diffuse sources
• Carry out a detailed study of a particular cluster, A3266
• Find which, if any of the current models best describes our sample

Throughout the project it was necessary to convert angular sizes (the projected separation in radians) to linear sizes (the `real' distance in the cluster.) To do this we had to assume a particular cosmology and a value for the Hubble constant H₀. For this project we assumed an Einstein-de Sitter universe ( $q_0 = \frac{1}{2}$) with $H_0 = 70 \mbox{ kms}^{-1} \mbox{Mpc}^{-1}$. It was then possible to convert between linear size, l_kpc, and angular size, $\theta_{rad}$, using:

$$\theta_{rad} = \frac{l_{kpc} H_0}{2c} \frac{1+z}{1-(1+z)^{-\frac{1}{2}}}$$ (1.1)

where z is the redshift of the cluster, $c = 2.998 \times 10^8 \mbox{ms}^{-1}$ and 1 pc = 3.086 $\times 10^{16}$ m.