Monday, November 1, 2010

Session I

Accretion processes in massive galaxy formation

  • discrete vs continuous: mergers (small, large) vs cold flows
  • reservoirs/duty cycle
  • violent, extraordinary phases
Invited Talks
Dusan Keres
Gas Accretion in Galaxies
Jen Lotz
           Galaxy Mergers Through Cosmic Time
Avishai Dekel
           Feeding Massive Galaxies in their Most Active Phase of Assembly and Star Formation
Contributed Talks
Greg Rudnick
The Efficiency of Clusters at Stopping Gas Accretion in Galaxies
Mauro Giavalisco
Discovery of Massive Amounts of Cold HI Gas in a Galaxy Overdensity at Redshift z~1.6
discussion led by Frederic Bournaud


  1. Great that this blog is available for comments and discussion! It gives people like me who were invited but could not attend a chance to participate (as well presumably as other interested bystanders).

    The topic of this session is an interesting one, but its title is revealing. A better one would have been "Inflows and outflows during massive galaxy formation". In my view, work in this area has suffered substantially over the last five years from having too narrow a focus, and from being simultaneously too close and too far from the observations!

    Too narrow because it has focussed on gas inflows, which are not yet observed unambiguously, while neglecting gas outflows, which are observed and are predicted to involve similar mass fluxes but larger energy fluxes. Too narrow also because it has focussed on the structure of z=2 galaxies, while failing to check if it simultaneously matches their abundance, and the abundance and properties of lower redshift galaxies.

    No current simulation of the formation of an individual massive galaxy at z=2 has a star formation efficiency approaching that required to match the observed abundance of these objects. Most are too high by a factor of four or more, the best by a factor of two. This means that they almost certainly predict incorrect structure. Most of the baryonic mass in the regions being studied should, in fact, have been ejected. Proper treatment of this outflow would seem very likely to modify inflow properties.

    In addition to smooth accretion of pristine gas, it is important to consider growth through other channels: accretion of gas and stars in merging galaxies; reaccretion of ejected (and presumably enriched) gas. We currently have few observational or theoretical constraints on the relative importance of these channels.

    In my view there are two reasons why our understanding has advanced little in this area despite dramatic recent observational progress in measuring z=2 galaxy properties. I'll give them in the next post

  2. Simon White (cont.)November 6, 2010 at 10:33 AM

    My two reasons why our understanding has advanced little despite dramatic recent observational progress -- One is that we still have very little information about the gas components of the galaxies and even less about the diffuse gas flowing in and out of them. The other is the "cold flows" band-wagon, which has encouraged people to ignore or forget what had been learned in the previous quarter century of galaxy formation research.

    Despite statements in the introductions of most recent papers on high redshift galaxy formation, "Cold flows" are neither new nor a paradigm shift (though they are an interesting phenomenon!). Their critical aspects are: (i) that gas flowing into massive protogalaxies never builds a static atmosphere but falls directly onto the galaxy; (ii) that much of this infall is along filaments.

    (i) was first pointed out clearly by Rees & Ostriker (1977), and it was incorporated both in the DM+gas galaxy formation model of White & Rees (1978) and in the CDM galaxy formation model of Blumenthal et al (1984). All these papers identified the switch between the "cold flow" and "cooling atmosphere" regimes as causing the distinction between galaxies and clusters.

    This two regime model has featured in semi-analytic galaxy formation models since the early 1990s, and cold flows are also present, of course, in all CDM galaxy formation simulations, starting with those of Katz, Hernquist, Navarro and others, also in the early 1990s. Important lessons from this period which have been lost in the rebranding of the field since 2005 include, the need for strong outflows (often called "feedback") and the effects of metallicity on the boundary between the two regimes. Both are clear from the SA modelling, but the first was also discussed in some detail, for example, by Steinmetz & Navarro using their late 1990s Tucson-based simulations.

    (ii) was first pointed out by Zeldovich and his group in the early 1980's and was already clearly visible in the 1983 simulation of Klypin & Shandarin. It later became the basis for understanding the build-up of structures in the "cosmic web" (another rebranding) and indeed many of the techniques developed in the 1980s and 1990s for understanding the morphology of the web could be used to analyse the structure of gas flows onto protogalaxies.

    I think the main points for discussion are:
    (i) how to get direct observational evidence for nflows and improved observational information about outflows.
    (ii) how to make simulations which actually match the accumulation, ejection and star formation efficiencies required to fit the *population* properties of z=2 (and z=0!) galaxies. Only simulations which approximately match these requirements, can give realistic predictions for the structure and evolution of the galaxies and of their gaseous environment.

  3. Guinevere KauffmannNovember 6, 2010 at 6:25 PM

    >we still have very little information about the >gas components of the galaxies and even less about >the diffuse gas flowing in and out of them.

    I absolutely agree that this is the primary bottle-neck in this subject. I would like to see
    discussion about what we are leaning from absorption lines in the rest-frame far-UV spectra of galaxies. Do current simulations reproduce the results in the recent paper by Steidel et al,
    ApJ 717, 2010 ? Lots of new constraints must be coming from COS spectra of low redshift galaxies -- what have we learned from this new data?

  4. > absorption lines in the rest-frame far-UV spectra > of galaxies. Do current simulations reproduce the > results in the recent paper by Steidel et al,

    Good questions that are acutally already in my pre-discussion review draft, and on which I will try to insist.

  5. I fully agree with Simon that a theoretical and simulation study of inflows and outflows combined is one of our great open challenges. It should probably start with identifying and properly simulating a physically sound driving source for massive outflows in massive galaxies. I only wish to note that there may be conflicting constraints on the SFR efficiency. The observed high SFR~100 at z~2 in massive disks indicates that a large fraction of the inflowing gas must make stars. The known cosmological input rate, combined with the halo mass function at z~2, gives room to Mdot_in=SFR+Mdot_out, with Mdot_out~SFR. So the gas penetration into the galaxy should remain efficient also in the presence of outflows. The fact that the input is in narrow, relatively dense cold streams may help, but this is to be verified. Anyway, the abundances of SFR at z~2, as estimated or simulated, seem consistent with the observations to within a factor of 2. If this creates an apparent conflict with the abundances at z~0, the scenario will have to involve nontrivial dependence on mass and time. The high-res simulations are still struggling with connecting z~2 to z~0, but it will come soon.

    I think that the fact that the observed outflow signal overwhelms the observed inflow signal in stacked spectra (e.g. in Steidel et al) is actually consistent with inflow along narrow streams with a small sky coverage. I'll show some preliminary results on this.

  6. Guinevere KauffmannNovember 7, 2010 at 7:14 AM

    *In addition to smooth accretion of pristine gas, *it is important to consider growth through other *channels: accretion of gas and stars in merging *galaxies

    I would like to point out beautiful recent work on
    nearby galaxies in this regard:
    e.g. Mouhcine, Ibata, Rekjuba, 2010, ApJ, 714, L12
    A welath of information about the stellar accretion history is imprinted in the structure and kinematics of the diffuse outer light and streams. And, this galaxy has long been claimed to be on example of a "smoking gun" for recent
    accretion in the Universe, based on evidence for
    extra-planar gas.

    It will be a long time before this level of detail
    is available at z=2. Unfortunately, the simulators
    who focus on low redshift galaxies concentrate only on stellar accretion events, and neglect the

    In my opinion, galaxy evolution astronomers have tended to become far too narrowly focused on the
    same set of questions repeated over and over again at conferences involving the same set
    of people in a narrow subfield, and have forgotten that there are many ways to make progress.

  7. > galaxy evolution astronomers have tended to become far too narrowly focused

    Guinevere, are you referring to 'modelers' ? Because I'd agree with Simon, observational progress on distant massive galaxies has been substantial in the last few years.

    Even on their gaseous properties, one of the weak points that Simon mentions, we actually have made some progress observationally.

  8. Guinevere KauffmannNovember 7, 2010 at 9:55 AM

    >observational progress on distant massive galaxies has been substantial in the last few years.

    Agreed. But the definition of *evolution* is connecting progenitors to descendents. Please see Darwin.

  9. > Unfortunately, the simulators who
    > focus on low redshift galaxies
    > concentrate only on stellar
    > accretion events, and neglect the
    > gas.

    you can't really hold simulators responsible for this. facts are that reaching low redshift with gas and cosmological context is incredibly expansive, because of basic facts such as the expansion of the Universe and the 'current condition' of hydrodynamic models (the timesteps become shorter and shorter, making the calculations much much longer). Everyone in this field is doing its best to find solutions to reach low redshift with gaseous evolution, cold gas infall, and not just satelites. I agree with you that large samples of such simulations are still missing, but they are on their way. I will show examples tomorrow. the most recent simulations do combine gas infall/evolution and stellar accretion events, in line with Mouhcine et al's observations.

    The models you are asking (down to z=0 and not just with stellar accretion events) for are indeed the ones that are needed, I agree with your 'request', but is is unfair to say that simulators are neglecting these aspects.

  10. Good to see that this blog is being well used!

    I'm puzzled by Frederic's last comment. Until recently essentially all galaxy formation simulations continued down to z=0 because that is where all the observations used to be. This is still true for most simulations. Examples in the last year include papers led by Agertz, by Brooks, by Crain, by Piontek and by Scannapieco.
    It is true that some simulators have taken advantage of the new data at z=2 to shorten the timespan of their
    calculations by a factor of four, thus allowing better resolution, but it is unclear whether this improved resolution increases the realism of the results, given the poorer treatment of winds, enrichment and other feedback effects, and our necessarily poorer understanding of the z=2 galaxy population relative to that at z=0.

  11. Simon, my comment is less puzzling when put in the context of what Guinevere was asking for. Let me explain why.

    > Until recently essentially all
    > galaxy formation simulations
    > continued down to z=0

    Yes, this is basically what I meant: there are indeed *very recent* simulation samples or on-going ones that do reach z=0 with gas (the ones you cite are from a few months ago, no more, so my comment has to be read in a broader context than just exactly today). But they still often have less resolution, and hence a cruder treatement of the gas, than models stopped at z~2.

    To compare what a given code allows, and following your quote to Agertz, one can compare Agertz+09 (40pc down to z=2, star forming clumps captured) to Agertz+10 (done a year later, down to z=0 but 1D resolutions of 200-300pc, so 100 times lower in volume, no star forming clumps captured). SPH simulations down to z=0 would have a gravitational softening of ~200pc at very best (but often ~1kpc in examples you cite), the 'hydro resolution' (average smoothing length) can be worse. Then cooling below 10^4K could only be down at the risk of having artificial fragmentation, etc..

    So to come back to the context of what Guinevere was describing, modelling the stellar streams, the thick components, or the accreting gas disk require a resolution of ~100pc, a thin disk to be resolved, cold HI or molecular regimes to be captured in the gas. There are few simulations that can do that down to z=0. Some very recent ones (such as Agertz+10) are probably there or almost. On-going work by many teams will provide more samples of this kind within the next two years. This is the comment I made..

    By the way, reaching z=0 rather than z=2 with an hydro-code, at fixed physical resolution and fixed current condition (=sort of fixed hydrodynamic accuracy) does not require a four times longer computing time. The factor is much higher.

    Anyways, I personally don't need convincing that reaching z=0 is the right thing to do for modelers.

  12. I suppose Genel et al. 2010 (astro-ph) is a reasonable attempt (using a very simple model for radiation driven winds) to match the 'better' observed strong outflows at high redshift with high-resolution simulations. The ratio of SFR to outflow rate is about three and as a consequence most of the inflowing (gas)mass is not turned into stars but expelled from the galaxies. This also helps reducing the baryon conversion into stars at an epoch where a significant fraction of stars in the Universe is formed (and eventually at z-0). The stellar mass at z=2 might still be a factor of two too high but I suppose we are on the right track to at least get a reasonable baryon budget.

  13. Re Thorsten/Genel et al: Agreed. But what happens to the inflowing gas with this stronger outflow? Does the inflow rate change? If no, how is the SFR affected and/or gas fraction in the disk affected - is there still 50% of the gas in the disk?

  14. This relates to a question i had about Avishai's talk. Avishai argued that using predictions of accretion rates and halo number densities from theory, we can match the observed numbers of rapidly SF galaxies (SMGs, etc) at z~2 as long as the SFR is at least about *half* of the inflow rate. however his baryon accretion rate formula seemed to imply that a ~10^12 M_sun halo at z=0 should be accreting new baryons at about 80 M_sun/yr. clearly, unless the conversion of accreted baryons into stars is MUCH less efficient than mstardot ~ 1/2 mdot_inflow, this will greatly overpredict the SFR for ~MW sized galaxies.

    hydro people, does 80 M_sun /yr onto MW-sized halos at z=0 sound consistent with what you get in the absence of outflows?

    why should feedback be so much stronger at low z in MW-sized galaxies?

    or, if alternately the ratio of outflow rate to SFR is higher, then we will presumably underpredict the numbers of rapidly SF galaxies at high-z.


  15. Guinevere KauffmannNovember 8, 2010 at 3:03 PM

    Simon, Jerry Ostriker and Eliot Quataert were
    try to work out over lunch whether standard Zeldovich theory about collapse of sheets , filaments would predict 3 streams feeding each galaxy. I recall you worked this out, so maybe you
    would like to add some discussion on this matter to
    the blog?

  16. Rachel,

    The formula has 80 Msun/yr at z=2 (not at z=0), and it scales as (1+z)^2.4, so at z=0 it is 5.7 Msun/yr for 10^12 Msun halo. This is the average accretion rate of baryons into the virial radius of a halo, as deduced from the dark-matter accretion with a constant baryonic fraction 0.17, and the dark-matter part is confirmed in any cosmological N-body simulation. It is a robust result.

  17. Reply to Guinevere
    The generic number of filaments intersecting at a node of a Voronoi tesselation of a non-regular point distribution (for example a Poisson distribution) is four. The same is true for the skeleton of a Gaussian random field. The first case (the Voronoi tesselation) is easy to understand. The set of points equidistant from two adjacent centres is a plane, a face of the tesselation, or a sheet in Zeldovich's scheme. Where two such sheets intersect, one get a straight line (a filament) which is then equidistant from three adjacent centres. Such lines intersect at nodes (clusters) which are generically (i.e. in the absence of special symmetry) equidistant from four centres and so are linked to four filaments.
    Dynamical simulations of evolution from Gaussian IC's typically give more than four filaments coming out of each node, even though the skeleton of the ICs obeys the above argument. This is because dynamical evolution causes nodes to merge along their connecting filament. A cluster formed by one such merger would then typically be fed by six filaments, for example. Generally not all such filaments are equally strong, of course.

  18. Center -7 2 Kia Classic Leaderboard1. Starting from website
    designing, it covers a wide range of search engine metrics.

    my homepage: search engine placement

  19. Now here you might be very aware of your partner's addictive behavior, but very getting your ex back tips unaware of your own and can be expensive. High numbers of chap who are looking just for hooking are frightened to enter in a relationship.

    My website: get old boyfriend back


To post without an account, please select 'Name/URL' from the drop down box, enter your name and click 'continue'. Then write your comment and click 'Post Comment'. If you have any problems, please contact jeyhan [at] noao [dot] edu.