Annu. Rev. Astron. Astrophys. 2017. 55: 59-109 Copyright © 2017 by Annual Reviews. All rights reserved

4. CONCLUSION & OUTLOOK

Since the advent of self-consistent cosmological numerical simulations about 35 years ago significant progress in understanding galaxy formation has been made. Modern super-computers and numerical algorithms have become powerful enough to allow the simulation of individual galaxies at relatively high resolution as well as the evolution of galaxy populations in representative cosmological volumes. With different numerical techniques (smoothed particle hydrodynamics, meshless particle hydrodynamics, adaptive mesh refinement and moving mesh hydrodynamics, see review by Somerville & Davé 2015) it is now possible to simulate galaxy populations (at spatial resolutions of 0.5 - 3 kpc) with a realistic cosmological evolution of sizes, abundances, star formation rates, dark matter fractions, gas fractions as well as stellar and black hole masses, from well defined initial conditions, for a direct comparison to observational galaxy surveys (e.g. Davé et al. 2013, Hirschmann et al. 2014, Schaye et al. 2015, Sijacki et al. 2015, Khandai et al. 2015).

Zoom simulations of individual galaxies (at resolutions < 500 pc) allow a detailed investigation of the formation processes and the consequences for the internal galaxy structure for direct comparison to high resolution observations (e.g. Guedes et al. 2011, Stinson et al. 2013, Genel et al. 2014, Hopkins et al. 2014, Dubois et al. 2013b, Marinacci, Pakmor & Springel 2014, Aumer, White & Naab 2014, Naab et al. 2014, Wetzel et al. 2016). Simulations of this kind (see also Renaud, Bournaud & Duc 2015, Rosdahl et al. 2015, Hu et al. 2016, Forbes et al. 2016, Richings & Schaye 2016) with high enough spatial resolution to represent the multi-phase interstellar medium structure and stellar feedback more accurately accounting for major effects like stellar winds, radiation and supernovae have the potential to shed more light into the detailed physical processes governing galaxy formation.

The rapid recent progress can be considered a success in our quest for a better understanding of galaxy formation. It was mainly triggered by the realization that thermal energy input from supernovae is most likely insufficient to trigger outflows and 'galactic winds' with other mechanisms being explored as the relevant drivers. These outflows, however, play a major - if not the dominant - role for regulating the formation of galaxies at low and high masses. They are most likely driven by energy injection from newly formed stellar populations (cosmic rays, radiation and winds, in addition to supernovae) and accreting black holes. As the outflows are launched on parsec and sub-parsec scales, well below the resolution and physical complexity limit of any cosmological simulation this has triggered a wealth of sub-resolution models for stellar and AGN feedback (see Section 2.2.1 and 2.3). The fact that many models - even if conceptually very different - driving a 'reasonable galactic wind' can 'successfully' reproduce galaxy abundances and disk galaxy morphologies (see e.g. Figs. 4 and 2) indicates that the essential characteristics of the problem has been disclosed. However, the empirical nature of the subresolution models limit the predictive power of the simulations and the literature becomes enriched by parameter studies of particular implementations despite obvious shortcomings of the respective models which are compensated by adjusted parameters. 'Delayed cooling', 'stochastic thermal', and 'non-thermal' feedback models may significantly overestimate the energy and momentum input into the ISM and the 'delay' time-scales are uncertain. 'Decoupled wind' models might not capture (i.e. underestimate) the energy coupling to the local ISM and result in unrealistic wind structures. Empirical 'momentum driving' models rely on uncertain coupling efficiencies for infrared radiation. Similarly, almost all AGN feedback models - on cosmological scales - are of empirical nature with accretion and energy conversion efficiencies adjusted, in a plausible fashion, to match observed scaling relations.

It will be a major theoretical challenge in theoretical galaxy formation to understand stellar and AGN feedback in detail and identify physically correct sub-resolution models taking into account all relevant physical processes. First promising steps in this direction have been made from high resolution galaxy scale simulations as well as simulations on smaller scales. A full accounting for the energy input from stellar populations (e.g. Hopkins, Quataert & Murray 2012b, Agertz et al. 2013), the long range effect of low and high energy radiation from stars and AGN (e.g. Choi et al. 2012, Vogelsberger et al. 2013, Kannan et al. 2014, Rosdahl et al. 2015, Roos et al. 2015, Bieri et al. 2016) and the consideration of other significant non-thermal components of the ISM, namely magnetic fields and cosmic rays (e.g. Uhlig et al. 2012, Hanasz et al. 2013, Booth et al. 2013, Salem & Bryan 2014, Pakmor et al. 2016) are probably the most promising areas of galaxy formation research in the future. These research directions mainly refer to number (2) of our physics problem (Section 1.1): the knowledge of the physical processes primarily responsible for understanding each phase of galactic evolution. Starting with well defined initial conditions we can now roughly reproduce the scales and internal structures of common galaxy types and laboring with increasing physical precision to correctly model the detailed processes involved in feedback from stars and super-massive black holes.

Acknowledgements

The authors acknowledge valuable input on the manuscript from O. Agertz, R. Dave, A. Dekel, Y. Dubois, P. Girichidis, M. Hanasz, M. Hirschmann, P. Hopkins, N. Khandai, B. Moster, T. Peters, E. Puchwein, M. Rafieferantsoa, V. Springel, G. Stinson, R. Teyssier, H. Übler, M. Vogelsberger, and S. Zhukovska. The authors are particularly grateful for the many detailed and valuable comments from J. Kormendy, D. Nelson and J. Schaye.