8. A PARTIAL SUMMARY OF OUTSTANDING PROBLEMS

I conclude with a summary of the most important outstanding problems. I restrict myself to big-picture issues and do not address the myriad engineering details that are unsolved by our present state of the art. They are, of course, vitally important. But a comprehensive list would require a paper of its own. I therefore refer readers to earlier chapters of this book, which discuss many of these problems in detail.

• (1) I emphasized in Section 4.1.3 that, to me, the most important goal is to produce realistic classical bulges+ellipticals and realistic disks that overlap over a factor of > 1000 in mass but that differ from each other in ways that we observe over the whole of this range. They can combine with any B / T from 0 to 1, but the differences between bulges and disks depend very little on B / T.
• (2) Four decades of work on z ≃ 0 galaxies showed convincingly that major mergers convert disks into classical bulges and ellipticals with the observed properties, including Sérsic index, fundamental plane parameter correlations, intrinsic shape and velocity distributions, both as functions of mass, the presence of cores or central extra light, and isophote shape. This work also suggested that merger rates were higher in the past, and modern observations confirm this prediction. By the mid-1990s, we had converged on a picture in which classical bulges and ellipticals were made in major mergers. Enthusiasm for mergers was probably overdone, but now, the community is overreacting in the opposite direction. The successes of the 1970s–1990s are being forgotten, and – I believe – we have come to believe too strongly that minor mergers control galaxy evolution. Reality probably lies between these extremes. For today's audience, the important comment is this: The observations that led to our picture of E formation via major mergers have not been invalidated. I suggest that the profitable way forward is to use what we learn from z ≃ 0 mergers-in-progress to explore how mergers make bulges and ellipticals at higher z, including (of course) differences caused (for example) by large gas fractions and including new ideas, such as violent disk instabilities that make clumps that make bulges. For this still-elusive true picture, it is OK that mergers are rare, because ellipticals are rare, too, and classical bulges are rarer than we thought. And it is OK that most star formation does not happen in mergers, because ellipticals are rare anyway, and because their main bodies are made up of the scrambled-up remnants of already-stellar progenitor disks.
• (3) The most important unsolved problem is this: How did hierarchical clustering produce so many giant galaxies (say, those with V_circ ≳ 150 km s⁻¹) with no sign of a classical bulge? This problem is a very strong function of environment – in field environments such as the Local Group, most giant galaxies are bulgeless, whereas in the Virgo cluster, most stars live in classical bulges and elliptical galaxies. The clue therefore is that the solution involves differences in accretion (gentle versus violent) and not largely internal physics such as star-formation or AGN feedback.
• (4) Calculating galaxy evolution ab inito, starting with ΛCDM density fluctuations, constructing giant n-body simulations of halo hierarchical clustering, and then adding baryonic physics is the industry standard today and the way of the future. It is immensely difficult and immensely rewarding. It is not my specialty, and I have only one point to add to the excellent review by Brooks & Christensen: Observations hint very strongly that we put too much reliance on feedback to solve our engineering problems in producing realistic galaxies. Observations of supermassive BH demographics tell us that AGN feedback does not much affect galaxy structure or star formation until mergers start to make classical bulges. And point (3) emphasizes that environment and not gravitational potential well depth is the key to solving the problem of giant, pure-disk galaxies.
• (5) We need to fully integrate our picture of disk secular evolution into our paradigm of galaxy evolution. As observed at z ≃ 0, this picture is now quite detailed and successful. Essentially all of the commonly occurring morphological features of galaxies – bars, (nuclear, inner, and outer) rings, nuclear bars, and pseudobulges – are at least qualitatively explained within this picture. Some of these details are beyond the “targets” of present galaxy-formation simulations. But pseudobulges are immediately relevant, because our recognition of them has transformed our opinions about classical bulges. They are much rarer than we thought. In particular, small classsical bulges are very rare. And although some galaxies have structure that is completely determined by the physics of hierarchical clustering, others – and they dominate in the field – appear to have been structured almost exclusively by secular processes. Incorporating these processes is a challenge, because slow processes are much more difficult to calculate than rapid processes. But secular evolution is an ideas whose time has come (Sellwood 2014), and we need to include it in our paradigm.
• (6) At the same time, our quantitative understanding of secular evolution needs more work. For example, we need a study similar to Dressler's (1980) work on the morphology-density relation: We need to measure the luminosity and mass functions of disks, pseudobulges, and classical bulges+ellipticals, all as functions of environmental density. At present, we have essentially only two “data points” – the extreme field (Kormendy et al. 2010; Fisher & Drory 2011) and the Virgo cluster (see Kormendy et al. 2010). This is already enough to lead to point (3) in this list. We need corresponding studies in more environments that span the density range from the field to the richest clusters. This will not be easy, first because we need high spatial resolution whereas observing more environments drives us to larger distances, and second because of point (7).
• (7) Our picture of disk secular evolution predicts that many galaxies should contain both a classical and a pseudo bulge. Work on the subject has concentrated on extremes – on galaxies that are dominated by one kind of bulge or the other. Samples of large numbers of galaxies will inevitably have to face the challenge of separating at least three components (bulge, pseudobulge, and disk) and in many cases more (bar, lens, …). We also need to be able to find pseudobulges in face-on barred galaxies (see Section 2.1). But it is easy to overinterpret details in the photometry. The best way to approach this problem is probably to begin with infrared observations of nearly-edge-on galaxies (e.g., Salo et al. 2015).
• (8) Are classical bulges really indistinguishable from ellipticals? The structural parameter scaling relations shown in Figure 4 (based on many authors's work) show that they are closely similar. I use this result thoughout the present paper. It is central to Renzini's (1999) paraphrase of the classical morphological definition: “A bulge is nothing more nor less than an elliptical galaxy that happens to live in the middle of a disk.” But not everybody agrees. Based on multi-component decompositions, different fundamental plane correlations for classical bulges and ellipticals have been found by Gadotti (2008, 2009, 2012) and by Laurikainen et al. (2010). We need to resolve these differences. At stake is an understanding of whether classical bulges and ellipticals form–as I suggest– by essentially the same major merger process or whether important variations in that process produce recognizably different results. In particular, it is not impossible that we can learn to distinguish ellipticals and perhaps some bulges that form via mergers of distinct galaxies from other bulges that form via the mergers of mass clumps that form in unstable disks. Both processes drive additional gas toward the center, but it is possible that bulge formation via disk instabilities is intrinsically more drawn out in time with the result (for example) that “extra light components” such as those studied in Kormendy (1999), KFCB, and Hopkins et al. (2009a) are smoothed away and unrecognizable in the resulting classical bulges but not in disky-coreless-rotating ellipticals.
• (9) Returning to elliptical galaxies: KFCB present a detailed observational picture and ARA&A-style review of the two kinds of ellipticals in large part as seen in the Virgo cluster. Hopkins et al. (2009a, b) present modeling analyses of wet and dry mergers, respectively. We need to know how this very clean picture as seen in the nearest rich cluster translates into other environments. Much of the work published by Lauer et al. (1995, 2005, 2007a, b), by Faber et al. (1997), by Kormendy & Bender (1996, 2013), and by Bender et al. (1989) applies to broader ranges of environments. It suggests that the picture summarized here in Section 4.1.1 is basically valid but that the distinction between coreless-disky-rotating and core-boxy-nonrotating galaxies is somewhat “blurred” in a broader range of environments. For example, M_V = −21.6 cleanly separates the two kinds in Virgo, with only one partial exception (NGC 4621 at M_V = −21.54 has n = 5.36_−0.28^+0.30 characteristic of core galaxies, but it has a small amount of extra light near the center). However, the above papers and others show that the two galaxy types overlap over a range of absolute magnitudes from about M_V = −20.5 to about M_V = −23. In the overlap range and occasionally outside it, some classification criteria in Section 4.1.1 conflict with the majority. We should not be surprised that heterogeneous formation histories can have variable outcomes; on the contrary, it is encouraging to see as much uniformity as we see. Still, a study of how the systematics depend on environment should be profitable.
• (10) Still on ellipticals and classical bulges: The SAURON and ATLAS^3D teams have carried out an enormous amount of truly excellent work on nearly all aspects of bulge+E structure and evolution. A review is in preparation by Cappellari (2015). It is natural to ask how the picture of bulges and ellipticals developed by the SAURON and ATLAS^3D papers compares with the one outlined in Section 4 here. The answer is that they agree exceedingly well. There are differences in emphasis, and the large SAURON + ATLAS^3D teams address many subjects that are beyond the scope of studies by our team or by the Nuker team. There is also one difference in analysis that makes me uncomfortable – in their work, they generally do not decompose galaxies into bulge and disk parts. It is therefore all the more remarkable that careful work without using component decomposition and our work that always is based on component decomposition converge on pictures that are so similar. E.g., the separate parameter correlations for bulges and disks that are shown here in Figure 4 are visible as pure-bulge and pure-disk boundaries of parameter correlation regions shown in Cappellari et al. (2013b). In their diagrams, the parameter space between our bulge and disk correlations is filled in with intermediate-Hubble-type galaxies that have 0 < B / T < 1. Similarly, Cappellari et al. (2011) and Kormendy & Bender (2012) both revive the “parallel sequence” galaxy classification of van den Bergh (1976), as do Laurikainen et al. (2011). Kormendy and Bender (2012) also add Sph galaxies (as distinct from ellipticals) to the classification.

  What may appear as a difference between Section 4 and the SAURON + ATLAS^3D work is our emphasis on many E – E dichotomy classification criteria versus their distinction based only on fast versus slow rotation. However, Lauer (2012) shows that the SAURON + ATLAS^3D division into fast and slow rotators is essentially equivalent to the division between coreless and core galaxies. The equivalence is not exact based in the rotation amplitude parameter λ_re/2 (within 1/2 of the effective radius r_e) chosen by the SAURON and ATLAS^3D teams. But it becomes much more nearly exact if slow and fast rotators are divided at a slighly higher rotation rate, λ_re/2 = 0.25. In unpublished work, I found an essentially equivalent result for the original SAURON kinematic classification, in which slow rotators have λ_R < 0.1 and fast rotators have λ_R > 0.1 as defined in Emsellem et al. (2007). If the division is instead made at λ_R = 0.175, then core and coreless ellipticals are separated essentially perfectly. (The only exception in KFCB is NGC 4458, which is slowly rotating but coreless. But it is almost exactly round, and rotating galaxies that are seen face-on will naturally look like slow rotators.) The more nuanced ATLAS^3D look at elliptical galaxy dynamics leads to a revised suggestion that fast and slow rotators should be separated at λ_re/2 = (0.265 ± 0.01) × √є_e/2 (Emsellem et al. 2011, Equation 4). A typical є = 0.2 for core-boxy galaxies and є = 0.35 for coreless-disky galaxies (from Tremblay & Merritt 1996) then implies a division at λ_re/2 = 0.16 and 0.12, respectively. The typical intrinsic ellipticaly of 0.4 found by Sandage, Freeman, & Stokes (1970) for all ellipticals implies λ_re/2 = 0.17. These values are closer to the rotation parameters 0.25 and 0.175 that divide core and coreless galaxies as found by Lauer (2012) and by my work, respectively.

  I suggest that the best way to divide slow rotators from fast rotators is not to pick some arbitrary value of the rotation parameter but rather to ask the galaxies what value of the rotation parameter produces the cleanest distinction into two kinds of galaxies as summarized in Section 4.1.1. When this is done, the E – E dichotomy as discussed in this paper and the large body of work done by the SAURON and ATLAS^3D teams are remarkably consistent.

• (2+3 redux) A partial exception to the above conclusion is some of the n-body simulation work, e.g., by Naab et al. (2014). They acknowledge the importance of major mergers in some ways that are consistent with the story advocated in this paper. But their conclusion that “The galaxies most consistent with the class of non-rotating round early-type galaxies grow by gas-poor minor mergers alone” (emphasis added) is at best uncomfortable within the picture presented here. The core-boxy-nonrotating galaxies have a large range of mostly homogeneous properties with respect to which the round ones do not stand out as different (e.g., KFCB). In particular, our understanding of cores – especially the tight correlations between core properties and BH masses – depends on our picture that cores are scoured by black hole binaries that are formed in major mergers (see KFCB and Kormendy & Bender 2009 for both the data and a review). At best, it remains to be demonstrated that minor mergers – which necessarily involve many small galaxies with (from Figure 7) undermassive BHs – can produce the very large BH masses and cores that are seen in giant core ellipticals. Dry minor mergers cannot do better than to preserve the M_• / M_host mass ratio. Also, if many minor mergers are necessary–and these galaxies are so massive that very many minor mergers are necessary to grow them–then there is a danger of producing a central cluster of low-mass BHs that is never observed as a cluster of compact radio sources and that is inherently unstable to the ejection of objects in small-n n-body systems (see KH13, p. 634).
• (11) I conclude with two sociological points: It is worth emphasizing that galaxy evolution work did not start in the 2000s. Many results that were derived in the 1960s – 1990s remain valid today. We should not forget them. We should integrate them into our current picture of galaxy evolution.
• (12) And finally: Galaxy evolution work has changed profoundly in the SDSS and HST eras. Before the early 1990s, our goal was to understand the evolution of galaxy structure. Now, most emphasis on galaxy structure has disappeared. Now, our goal is to understand the history of star formation in the universe. The main reason for this change is the common ground found between SDSS studies of many thousands of galaxies and HST studies of very distant galaxies. Necessarily, both kinds of studies concentrate on galaxies whose images are a few arcsec across. We do not resolve structural details. Mainly, we measure colors and magnitudes. So galaxy evolution has evolved into the study of the red sequence and blue cloud in the color-magnitude relation. Star formation and its quenching are, of course, important. But it would be enormously healthy if we could improve the dialog between SDSS+HST people and those – such as this author – who work on nearby galaxies whose star formation histories and structures can be studied in great detail. Conselice (2014) is an example of a paper that tries to bridge the gap. We would benefit greatly if we could completely connect the two approaches to galaxy evolution.

Many of my ideas about galaxy evolution were forged in intense and enjoyable collaborations with Ralf Bender, Luis Ho, and the Nuker team. It is a pleasure to thank all these people and many more who I do not have room to list for fruitful conversations over many years. I am especially grateful to Reinhard Genzel for stimulating and insightful discussions and to Ralf Bender, Dimitri Gadotti, and Eija Laurikainen for very helpful comments on this paper. Any errors of interpretation that remain are of course my responsibility. I thank Steve Allen, Xinyu Dai, Ying-Jie Peng, and Simon Lilly for permission to copy figures. My work on this paper was supported by the Curtis T. Vaughan, Jr. Centennial Chair in Astronomy at the University of Texas.