7. THE MILKY WAY BULGE IN THE CONTEXT OF EXTERNAL GALAXIES

7.1. The X-shaped Bulge of the Milky Way: how rare is this?

It can be argued that while in the context of Galactic research box/peanuts are somewhat a recent discussion, in Extragalactic studies such structures are known for about twice as much the time. So it came as no surprise to the Extragalactic community when evidence suggested that the Milky Way has a box/peanut, especially because there is also evidence that it has a bar. Perhaps the first mention in the literature about these deceptively unusually-looking structures is from Burbidge and Burbidge (1959), referring to the prototypical example that is NGC 128 (see also Sandage 1961). More detailed investigation came with de Vaucouleurs (1974), Jarvis (1986) and Shaw (1987). But it was in the pioneering studies of de Souza and Dos Anjos (1987) that the major step of connecting these structures with bars was made for the first time from an observational viewpoint, using a statistical argument. Basically, they argued, the frequency of box/peanuts in edge-on lenticulars is similar to that of bars in face-on lenticulars, consistent with the idea that box/peanuts are bars seen at a different projection. This conclusion was corroborated years later by Lütticke et al (2000), who reported a fraction >40 per cent of box/peanuts in disc galaxies covering most of the Hubble sequence (from S0 to Sd classes).

The complicating factor here is of course the fact that bars are difficult to be seen when the inclination angle of the galaxy is too large. Therefore, simulations of barred galaxies played a major role here. In fact, the starting point for this observational connection between bars and box/peanuts was the work published in Combes and Sanders (1981). These authors have shown, using collisionless simulations, that bars seen at a given edge-on projection show a very characteristic peanut-like morphology.

A number of studies came thereafter dedicated to extend this connection into a dynamical context. Kuijken and Merrifield (1995) came up with an ingenious diagnostic to test in this context whether box/peanuts are just bars seen at a different projection. This consisted in producing diagrams in which the line of sight velocity is plotted against the galactocentric radius for highly inclined or edge-on systems. By producing such diagrams corresponding to orbits in a purely axisymmetric potential, and orbits in a barred potential, they showed that the presence of a bar produces a clear distinctive signature. Because at a region around the bar corotation radius (the radius at which the pattern speed of the bar matches the local circular speed) there are no close, non-self-intersecting orbits available, clear gaps appear in this diagram, producing a figure-of-eight pattern. The matter became then just to produce such diagrams for galaxies presenting box/peanuts in order to test for the presence of a bar. Kuijken and Merrifield (1995) did that for NGC 5746 and NGC 5965, providing observational evidence that box/peanuts and bars are related phenomena.

More evidence was produced in Merrifield and Kuijken (1999) and Bureau and Freeman (1999). To this point, almost all galaxies with box/peanuts studied showed evidence of a bar. In only a few extreme cases the box/peanut could have formed through accretion of external material. In addition, none of the galaxies without box/peanuts showed signatures of a bar. Further development also happened in the theoretical background. Bureau and Athanassoula (1999) refined and corroborated the orbital study of Kuijken and Merrifield (1995), while Athanassoula and Bureau (1999) provided strong support to the bar detection diagnostic with hydrodynamical simulations.

It must be noted that the detailed morphology of box/peanuts, i.e. if they either have a boxy shape, a peanut shape or an X shape, depends on projection effects, as well as the strength of the box/peanut. More about X shape bulges can be found in Laurikainen & Salo (this volume), but essentially the X shape is more clear in the strongest peanuts. Techniques such as unsharp masking are able to reveal X shapes more clearly when the peanut is not so pronounced. Therefore, all these morphologies come actually from the same physical process, i.e. the increase in the vertical extent of stellar orbits in the inner parts of the bar. Both simulations and observations point out that box/peanuts extend to galactocentric distances which are about a third to a half of the bar semi-major axis (see Erwin and Debattista 2013). What causes this change in the orbits vertical extent is reviewed in detail by Athanassoula (this volume). In the Milky Way, we see the extension of the B/P to galactocentric distances of ∼ 1.5 kpc, which is nearly two-thirds the length of the semi-major axis of the long bar of 2-2.3 kpc (Hammersley et al 2000, Benjamin et al 2005, López-Corredoira et al 2007).

A review on the observed properties of box/peanuts in external galaxies is given by Laurikainen & Salo (this volume). They also discuss a structure called barlens, which is interpreted as the projection of box/peanuts when seen face-on. This structure was noticed by Laurikainen et al (2005), who included a model to fit barlenses in their image decompositions. Later, based on Fourier analysis, Laurikainen et al (2007) suggested that barlenses are part of the bar, while Gadotti (2008) also noticed their existence in a sample of local barred galaxies. However, only in Laurikainen et al (2011) the term 'barlens' is introduced as a new morphological feature in galaxies. Very recently, more detailed studies have made a robust connection between barlenses and box/peanuts (Laurikainen et al 2014, Athanassoula et al 2014). This connection implies that also on the plane of the disc, the stellar orbits in the inner part of the bar become wider. Figure 8 describes schematically the connection between bars, box/peanuts and barlenses.

Figure 8. Connection between bars, box/peanuts and barlenses. The diagram on the left shows a schematic representation of a barred galaxy seen a) face-on, and b) edge-on. Box/peanuts can be seen at edge-on projections, but the flat, more extended part of bar is difficult to realize, as its vertical extent is similar to that of the disc. The same galaxy seen face-on would reveal the bar and barlens. The barlens and the box/peanut appear to be the same structure seen at different projections [see Laurikainen & Salo (this volume) and Athanassoula (this volume)]. The panels on the right show an R-band image of NGC 4608 (top), where the bar and barlens can be clearly seen (see also Laurikainen et al 2005, Laurikainen et al 2007, Laurikainen et al 2011). The bottom panel shows a residual image, after the subtraction of a 2D model of the bulge, bar and disc of this galaxy. In the residual image the barlens stands out even more clearly. The red circle points out the barlens. This circle was not in the original Gadotti (2008) paper, but added now that we understand that the structure is a barlens. The red arrows point out empty regions in the disc within the bar radius, where stars from the disc were captured by the bar. [Right panels adapted from Gadotti 2008.]

7.2. Bulge formation scenarios

The formation of bulges in general, and of the Milky Way bulge in particular, have been discussed many times elsewhere (e.g. Aguerri et al 2001, Kormendy and Kennicutt 2004, Athanassoula 2005, Laurikainen et al 2007, Hopkins et al 2010, Fisher and Drory 2010, Gadotti 2012 see also Bournaud, and Brooks & Christensen, this volume). Here we will assess each bulge formation scenario in the light of evidence obtained from data on the Milky Way bulge, as presented above. There is mounting evidence that the Milky Way has a bar and a box/peanut, and thus secular evolution processes induced by bars in disc galaxies must have played a non-negligible role in the evolution of the Galaxy. On the other hand, a scenario in which the Galactic bulge was formed in violent processes such as mergers has weak support from data. In fact, an important question that observers must focus now is whether the Galaxy has a merger-built central stellar component at all.

7.2.1. Bulges formed via disc instabilities

Dynamical disc instabilities can originate bars, spiral arms and ovals, the latter being just a distortion in the disc stellar orbits that make them acquired less circular orbits, but not as eccentric as those in bars. All these structures, being non-axisymmetric, produce perturbations in the galaxy potential, with the result that material (gas and stars) within the corotation resonance radius loses angular momentum, whereas material outside the corotation radius absorb this angular momentum. The effect is particularly important for the collisional gas component, which thus falls towards the centre. At some point, the in-falling gas gets compressed and form stars, contributing to a rejuvenation of the stellar population in the central regions (e.g. Gadotti and dos Anjos 2001, Fisher 2006, Coelho and Gadotti 2011, 502011llison et al 2011).

Because most disc galaxies with bars should have one or two Inner Lindblad Resonances (ILRs) near the centre ( ∼ a hundred to a few hundred parsecs from it), the in-falling gas cannot reach the galaxy centre immediately (see Fig. 9). Instead, it usually forms an inner disc, decoupled from the large-scale disc. These inner discs may form nuclear bars and spiral arms, as often observed, and are often called disc-like bulges (see e.g. Athanassoula 2005, Gadotti 2012) or discy pseudobulges, to contrast with the fact that box/peanuts are often called as well pseudobulges (see Kormendy and Kennicutt 2004). At the ILRs gas can get accumulated and compressed, often forming a star-bursting nuclear ring.

Figure 9. Angular speed of stars in circular orbits in a potential reproducing a disc galaxy with a bar, as a function of the galactocentric distance (in arbitrary units). The bar pattern speed is represented by the solid horizontal line (ΩB). The epicyclic frequency of the stellar orbits is denoted by κ. Barred galaxies present several dynamical resonances. This figure shows schematically how four of the main resonances come to be. Whenever the bar pattern speed is equal to Ω or Ω ± κ/2 this is the position of a dynamical resonance. From the centre outwards, the resonances depicted here are: the Inner Inner Lindblad Resonance, the Outer Inner Lindblad Resonance, Corotation, and the Outer Lindblad Resonance. In this case, the main families of stellar orbits change their orientation by 90 degrees at each resonance. This effect is at the origin of a number of dynamical effects in barred galaxies, in particular the transfer of angular momentum from material inside the corotation radius to material outside this radius, and the resulting formation of disc-like bulges.

However, it must be noted that simulations indicate that, initially, the infall of gas within a bar corotation radius occurs rapidly after the formation of the bar, with a time scale of the order of 10⁸ years, i.e. a dynamical time (Athanassoula 1992, Emsellem et al 2014). ² This means that disc-like bulges not necessarily present ongoing star formation or a very young stellar population, if the bar has formed long ago and have been able to push most of the gas to the centre quickly, and if the gas content in the disc is not being replenished. Sheth et al (2008, 2012) present results that suggest that the first long-standing bars ³ formed after redshift 1. These bars are still the ones seen at redshift zero, since bars are difficult to destroy, unless the disc is extremely gas-rich (see Athanassoula et al 2005, Bournaud and Combes 2002, Bournaud et al 2005, Kraljic et al 2012). This means that they first induced star formation at the centres of their host galaxies about 8 Gyr ago. Thus, disc-like bulges with stars as old as 8 Gyr are perfectly possible.

For the disc to be replenished with gas so that the bar can push this gas to the centre and produce a new central burst of star formation and rebuilding of the disc-like bulge, it has to fall into the disc plane from a direction not parallel to the galaxy disc, and inside the bar corotation radius. Otherwise, this gas will be pushed outwards by the bar or accumulate at the corotation (see Bournaud and Combes 2002). Evidence for gas infall from directions not parallel to the galaxy disc has recently been presented by Bouché et al (2013), but how often it occurs is still unknown, as is whether the gas reaches the disc within corotation.

Thus, although Ellison et al (2011) find that, statistically, barred galaxies present ongoing, central star formation more often than unbarred galaxies, there is still a significant fraction of barred galaxies with star formation rates comparable to those in unbarred galaxies. In addition, Coelho and Gadotti (2011) find that the younger bulges found in barred galaxies have a mean stellar age of a few Gyr. This is in contrast to unbarred galaxies, which show on average older mean stellar ages (see Fig. 10). This means that replenishing the disc with gas inside the corotation radius is a phenomenon that does not occur very often. Otherwise, very young stellar populations should be more conspicuous at the centres of barred galaxies. As discussed above, Bensby et al (2013b) find a stellar component in the Milky Way bulge with ages less than 5 Gyr. Although this means that the mean stellar age for the Milky Way bulge as a whole is above this value, this younger component has a mean stellar age thus in very good agreement with the mean stellar ages of the young bulges in other barred galaxies. The bottom line is that stars originating from gas infall to the centre through disc instabilities do not necessarily have to be extremely young now. A fraction of these younger stars could be elevated out of the disc plane and populate the box/peanut, but most of these stars are expected to be at or near the large scale disc plane.

Figure 10. Relative distributions of bulge mean stellar ages for barred and unbarred galaxies in bins of bulge stellar mass, as indicated. For massive bulges, the distribution is bimodal only for barred galaxies, consistent with the picture in which secular evolution processes build disc-like bulges. However, note that the mean stellar ages of such bulges can be as high as a few Gyr. [Adapted from Coelho and Gadotti 2011.]

Some studies using both simulations and observations have suggested that some star formation may occur along the bar as soon as the bar forms and induces shocks in the gas content in the interface between the bar and the disc (see Athanassoula 1992, Phillips 1996, Sheth et al 2002). However, they also indicate that, due to shearing in the gas clouds when they start falling towards the centre, this is interrupted shortly afterwards, and from then on, star formation is limited to components at the bar ends and the disc centre. Discs stars trapped by the bar mostly do not leave the bar, keeping their elongated orbits. As it evolves, however, the bar can capture stars from the disc formed more recently (see an example of such capture in the right panels of Fig. 8), and thus a fraction of young stars can be present in the bar, as long as there is ongoing star formation in the disc within the radius at which the bar ends, and the bar can grow stronger and keep capturing disc stars. Stars in bars are thus predominantly old, and therefore box/peanuts, being just part of bars, should as well be populated mostly by old stars, with some younger component.

For a galaxy as massive as our own, the bar is expected to form at redshifts close to 1 (see Sheth et al 2012). The ages of the stellar populations seen at the Galactic bulge are thus consistent with the picture of it being built purely from the bar instability in the disc, i.e. the Galactic bulge can well be just a box/peanut plus a disc-like bulge, as far as the ages of its stellar populations is concerned.

In trying to assess how the bulge of the Galaxy has formed, the chemical content of its stellar populations are better consider closely with their kinematical properties. Bars and their box/peanuts are expected to rotate cylindrically, i.e. the mean stellar velocity is independent of the height above the plane of the disc – as a rigid body. Variations in this pattern are usually attributed to the presence of other structural components with different kinematical properties (see e.g. Williams et al 2011 and references therein). As discussed above, one stellar population in the Bulge can be described as having high metallicity, low content of alpha-elements, and kinematics consistent with the eccentric stellar orbits in bars in cylindrical rotation. The alpha-element content thus indicates that the box/peanut formed after the thick disc. In external galaxies, recent evidence suggests that thick discs form early, during the short initial formation stages of the galaxy, which qualitatively agrees with the picture for the Milky Way (Comerón et al 2011, Comerón et al 2014). Nevertheless, although most of the thick disc stars thus form in situ, a significant fraction of stars in the thick disc may come from the accretion of satellites, and another fraction (likely smaller), may consist of heated up stars from the thin disc. Evidently, this complicates substantially the interpretation of observations.

Likewise, the observations from the ARGUS survey indicating variations in metallicity within the box/peanut (see Sect. 4.1), are suggestive of a complex stellar population content in the box/peanut itself. This can be a result of similar processes that complicate just as well the stellar population content of the thick disc, as described in the previous paragraph. As discussed in Sect. 4, different populations or population gradients in the disc from which the box/peanut forms can produce a similar result. However, it can also be the result of more than one buckling event forming the box/peanut. In the simulations of Martinez-Valpuesta et al (2006) a bar goes through a first buckling event about 2 Gyr after the beginning of the simulation, and this event is fast (< 1 Gyr). However, a second, powerful buckling event occurs around 5 Gyr later and lasts for about 3 Gyr. How this would affect the chemical content of stars seen today in the box/peanut depends on how the stellar population content and kinematical properties of the bar and box/peanut vary during these periods. But it is clear that if the bar of the Milky Way has gone through such recurrent buckling, the presence of multiple populations in the box/peanut is not surprising. Add this to the complex composition of the thick disc and one sees how complicated the stellar populations can be away from the disc plane.

7.2.2. Bulges formed via violent processes

The classical picture of bulges in disc galaxies is that of mini-ellipticals: massive, smooth and extended spheroids with dense centres, with old stellar populations showing alpha-enhanced chemical composition and relatively low rotational support (as compared to discs). A natural formation scenario for these structural components would be that of a monolithic collapse (Eggen et al 1962), in which a single gas cloud collapses in short time scales (< 10⁸ yr), producing a violent burst of star formation that originates the stellar halo and the bulge. While this scenario might explain the formation of the first spheroids, it faces many difficulties. It does not reproduce for instance the heterogenous distributions of stellar ages and metallicities observed in bulges (e.g. McWilliam and Rich 1994, Wyse et al 1997, a monolithic collapse implies a more homogeneous population) and regions of ongoing star formation (e.g. Carollo et al 1997).

It should be noted, however, that the monolithic collapse scenario was formulated within a perspective that does not include box/peanuts, and thus should not be compared against the properties of such observed structures. On the other hand, modern dissipative collapse models similar to the pioneer model of Eggen et al (1962), that however include as well cosmological ingredients, generate bulges with more realistic properties. Such key ingredients include in particular a long time scale history of accretion of dark matter haloes into the central halo, with the associated evolution of angular momentum and star formation episodes (see e.g. Samland and Gerhard 2003, Obreja et al 2013).

A natural picture within ΛCDM cosmology is that of merger-built bulges. Brooks & Christensen (this volume) review this picture. Another scenario to explain the formation of classical bulges is the coalescence of clumps in discs at high redshifts. This is also reviewed in this volume by Bournaud.

The evidence from observations of the Galactic bulge as reviewed above, however, give little support to the presence of a massive classical bulge. The observed box/peanut and its cylindrical rotation cannot be originated in violent scenarios. On the other hand, the possibility of a small classical bulge embedded within the box/peanut is not yet ruled out. Such composite bulges are discussed in the next section.

Nevertheless, we have seen above that there is a component in the Galactic bulge with low metallicity and an alpha-element content consistent with it being formed concomitantly with the thick disc, i.e. before the box/peanut. In addition, the morphological properties of this component seem to point out a more spherically distributed structure (see Fig. 1). This is revealed by RR Lyrae stars and are properties that are shared by classical bulges. This component has a spatial extent similar to that of the box/peanut, but it is not revealed by images such as those from COBE. However, a number of early-type disc galaxies with massive classical bulges shows bars, which probably went through a buckling process that originated a box/peanut. So it is perfectly plausible to have a classical bulge and a box/peanut coexisting in the same galaxy. A possible example is our own massive neighbor, M31. Athanassoula and Beaton (2006) have shown that this galaxy has a bar and a box/peanut, and there is evidence that it also hosts a classical bulge (e.g. Courteau et al 2011 and references therein). 2D decompositions in Gadotti & Erwin (in preparation) show that the classical bulge has a similar extent as the box/peanut. The morphology of the M31 bulge (see Fig. 2 in Athanassoula and Beaton 2006) is similar to that seen in the COBE/DIRBE image for the Milky Way, although the vertices of the box/peanut are more clearly recognized in the Galaxy (perhaps due to projection effects in M31). Nevertheless the same 2D decompositions of Gadotti & Erwin reveal the X shape outstandingly (see Gadotti 2012). Another critical issue is the understanding of how bright/massive is the component revealed by RR Lyrae stars in the Galaxy, and how does this compare to other classical bulges. It clearly cannot be large enough as to mask the vertices of the box/peanut revealed by COBE.

7.2.3. Composite bulges

In the previous subsections, we explored the possibility that the Milky Way has a disc-like bulge, apart from its box/peanut. We also remarked about the possibility of a classical bulge. Here we will briefly summarise recent work on the presence of such composite bulges in external galaxies. It is not hard to contemplate the possibility of such composite systems. A pure disc galaxy can form at high redshifts, say z ∼ 3, and acquire a classical bulge, be it through minor mergers, accretion events or the coalescence of clumps of stars and gas in the disc. At z ∼ 1 the same disc – now hosting a classical bulge at its centre – might become unstable to the formation of a bar, and develop one. Quickly this bar pushes gas to the central regions of the disc, originating a disc-like bulge. Give it a couple of Gyr and the box/peanut is formed. We thus end up with a galaxy containing a classical bulge, a disc-like bulge, and a box/peanut.

Gadotti (2009) has shown evidence that 34 per cent of his disc galaxies hosting classical bulges are galaxies possessing bulges with structural properties typical of classical bulges, but with an intensity of star formation activity characteristic of disc-like bulges. While one cannot rule out the possibility that some classical bulges may present ongoing star formation, it is also plausible that many of these galaxies actually host composite bulges. Those would be composed by an extended classical bulge with an embedded disc-like bulge. Because the classical bulge dominates the disc-like bulge in terms of mass, the composite bulge shows structural properties of classical bulges. However, using the D_n(4000) spectral index, Gadotti (2009) was able to realize the intense star formation in the bulge region. Such star-forming activity, according to this interpretation, occurs at the embedded disc-like bulge. Further evidence for this and other types of composite bulges is presented in Méndez-Abreu et al (2014).

Kormendy and Barentine (2010) report the existence of a small disc-like bulge embedded in the box/peanut of NGC 4565. Nowak et al (2010) argue that NGC 3368 and NGC 3489 actually have an embedded classical bulge within component(s) built via disc instabilities. Finally, very recently, Erwin et al (2014) have shown further evidence of such components that appear to be embedded classical bulges.

It will thus be no surprise if the Galactic bulge is a composite bulge.

² Gas outside corotation receives angular momentum from the bar, and other factors govern the gas infall rate at these outer radii, such as dynamical effects induced by spiral arms and the dissipative nature of the ISM. At these distances the infall of gas no longer occurs in a dynamical time scale. Back.

³ Some simulations (Kraljic et al, 2012) have reported the early formation of bars, at redshifts above 1. However, these bars are short-lived. In these simulations, bars formed at ≤ 1 generally persist down to z = 0. Back.