dark matter halos from the inside out.pdf

7/23/2019 Dark Matter Halos from the Inside Out.pdf

1/18

Hindawi Publishing CorporationAdvances in AstronomyVolume 2011, Article ID 604898,17pagesdoi:10.1155/2011/604898

Review ArticleDark Matter Halos from the Inside Out

JamesE. Taylor

Department of Physics and Astronomy, University of Waterloo, 200 University Avenue West, Waterloo, ON, Canada N2L 3G1

Correspondence should be addressed to James E. Taylor,[email protected]

Received 2 July 2010; Accepted 16 August 2010

Academic Editor: Rob Ivison

Copyright 2011 James E. Taylor. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The balance of evidence indicates that individual galaxies and groups or clusters of galaxies are embedded in enormousdistributions of cold, weakly interacting dark matter. These dark matter halos provide the scaffolding for all luminous structuresin the universe, and their properties comprise an essential part of the current cosmological model. I review the internal propertiesof dark matter halos, focussing on the simple universal trends predicted by numerical simulations of structure formation.Simulations indicate that halos should all have roughly the same spherically averaged density profile and kinematic structureand predict simple distributions of shape, formation history, and substructure in density and kinematics, over an enormous rangeof halo mass and for all common variants of the concordance cosmology. I describe observational progress towards testing thesepredictions by measuring masses, shapes, profiles, and substructure in real halos using baryonic tracers or gravitational lensing.An important property of simulated halos (possibly the most important property) is their dynamical age, or degree of internalrelaxation. I review recent gravitational lensing studies of galaxy clusters which will measure substructure and relaxation in a largesample of individual cluster halos, producing quantitative measures of age that are well matched to theoretical predictions.

1. Introduction

Everywhere in the universe, on scales comparable to thesize of galaxies or larger, the effects of gravity appear to beanomalously strong. The earliest of these observations datesto 1933, in work by Zwicky [1] on the Coma cluster ofgalaxies. Using redshifts derived from optical spectroscopy,Zwicky measured a velocity dispersion of 1600 km/s forthe galaxies in Coma, indicating that the cluster had

an enormous amount of internal kinetic energy. For thestructure to be dynamically stable, the gravitational potentialenergy required was 20 times what would have been inferredfrom the distribution of visible stars and gas alone. Zwickysuggested that some sort of dark matter, cold enough tobe effectively invisible in his optical observations, mightaccount for the deficit. The term dark matter has beenwith us since, although the idea remained dormant for nearly40 years following Zwickys discovery. A generation later,the idea of dark matter was rediscovered by Rubin andFord. Starting with a study of the Andromeda galaxy, M31[2], they gradually established that rotation speeds in theoutskirts of spiral galaxies were far higher than would be

possible for stable systems bound together by their visiblegas and stars alone. Once again, the mass needed to explainthe observations was enormous, and the concept of a darkhalo around every galaxy was eventually introduced. It wasunclear at the time whether this matter was normal, baryonicmatter (that is matter whose mass was comprised of baryons,protons, and neutrons) in the form of cold gas or compactobjects, or whether it was some novel form of nonbaryonicmatter.

Although individual galaxies like M31 or clusters ofgalaxies like the Coma cluster are still powerful probes ofthe nature and distribution of dark matter, the evidence fordark matter (and/or nonbaryonic matter) is now much morebroadly based than it was in the 1930s or the 1970s. Thespec-trum of fluctuations in the cosmic microwave background(CMB), for instance, strongly constrains the composition ofthe universe when it was less than a million years old. At thistime the radiation density is high enough that interactionswith photons have a substantial effect in smoothing outvariations in the density of normal matter (i.e., matter withan electromagnetic coupling). Weakly interacting matter isfree to go about its business, however, and fluctuations in
http://-/?-mailto:[email protected]://-/?-http://-/?-http://-/?-http://-/?-http://-/?-mailto:[email protected]://-/?-


2/18

2 Advances in Astronomy

this component can grow once the corresponding particlehas cooled sufficiently to become nonrelativistic. The relativeheights of the second and third acoustic peaks in the CMBangular power spectrum constrain the ratio of the twocomponents; together with the first peak they indicate thatthe total matter density in the universe, as a fraction of the

critical density, is

m = 0.267 0.029, while the fractionin baryonic matter coupling to photons is only bar =0.045 0.003 [3]. The difference between the two figuresmust come from a component with interactions of the weakscale or less. In principle this dark component could havedecayed since that time, but measurements of large-scalestructure (see, e.g., [4]) and cosmic expansion (see, e.g.,[5]) find a consistent value for the total matter densityat low redshift. Meanwhile, abundances of light elementsprovide an independent estimate of the baryon densitythat is roughly consistent with CMB measurements andmuch lower than the density needed to explain large-scalestructure or cosmic expansion [6]. Thus the best evidencefor cold dark matteror more specifically for a nonbaryonic,pressureless component that dominates the matter densityand is nonrelativistic at early timesnow comes fromthe largest scales in the universe. This point is sometimesmissed by alternative theories that focus exclusively on galaxyrotation curves or similar tests.

The idea of a weakly interacting particle massive enoughto be cold (or nonrelativistic) at the time of the CMB is notunwelcome in proposed extensions to the Standard Modelof particle physics. To explain the many cases of fine tuningor the strange discrepancies in scale in the Standard Model,these extensions introduce new high-energy symmetries andnew families of massive particles, partnered to the familiarlow-energy states. Supersymmetry, with its family of massivesuperpartners, is the most commonly cited example, butthe Kaluza-Klein modes from higher-dimensional theoriesare similar in many respects, and other well-motivatedmodels exist[7]. There is great hope that the Large HadronCollider will produce one of these candidates in the nearfuture, that direct detection experiments will detect it inthe lab, or that indirect detection experiments will detect itsannihilation or decay products. Not all candidates are easilydetectable, however; some may elude most or all attemptsto identify them [8]. Furthermore, determining the fullproperties of a candidate will probably require astrophysicalinput as well as laboratory or accelerator measurements[8]. Thus astrophysics will remain an important source of

information about dark matter for some time to come. Thecosmic abundance, mass, decay channels, self-couplings, orcouplings to known particles and possibly even excited statesof the dark matter particle(s) may all eventually be derivedfrom astrophysical measurements [9].

As a central component in the current picture of cos-mological structure formation, cold dark matter (CDM) hasbeen extremely successful. The standard cosmological modelof structure formation (which I will refer to loosely as theCDM model although it contains many other ingredients)posits that the universe contains known particlesbaryons,photons, a small contribution from warm or hot neutrinosbut also two dominant dark components, weakly interacting

cold dark matter and a cosmological constant or some similarform of dark energy. Given these ingredients and startingfrom an inflationary power spectrum at early times, theCDM model predicts the subsequent growth of fluctuationsin the matter distribution, the 3D power spectrum ofthese fluctuations after radiation-matter equality, the angular

power spectrum of temperature fluctuations in the CMBat the time of last scattering, and the properties of largescale structure post CMB. The predictions are in such goodagreement with large-scale measurements of these quantitiesthat it is hard to construct sensible alternatives to thestandard picture.

On somewhat smaller scales, the CDM model does notspecifyhowgalaxies form or evolve, but it does suggest wherethey may form. Analytic theory and numerical simulationsof structure formation show how fluctuations grow intovirialized halos and predict the abundance and clusteringof dark matter halos as a function of mass and redshift.Empirical models can then be used to place galaxies in thesehalos in a way that is consistent with the galaxy abundancesand clustering measured in surveys. Finally, measurements ofthe average gravitational potential associated with galaxies,through ravitational lensing, for example, close the loop,the abundance and clustering pattern of galaxies and theirassociation with excess gravitational potential are now thor-oughly established, so much so that any alternative model ofstructure formation is constrained to make predictions verysimilar to the standard CDM theory at the redshifts we haveobserved to date.

In short, CDM has passed all observational tests onscales of individual halos or larger. In the process, theobservational evidence has narrowed the field of competitionfor alternative models considerably. Any working theoryof structure formation, whatever its physical basis, is nowconstrained by observations to look a lot like the CDMmodel. An unfortunate corollary is that large-scale testsof structure formation may no longer be enough to makeprogress in this field. Future observations of large-scalestructure, for example, from forthcoming experiments tomeasure baryon acoustic oscillations [1012], may teach ussomething about the nature of dark energy, but they willprobably not teach us much about the nature of dark matter.For this, we will need more detailed studies of the individualhalos detected by these spectroscopic surveys and by the nextgeneration of wide-field imaging surveys [1316].

Indeed, while models of structure formation are highly

constrained by large-scale observations, the constraints onsmall scales are considerably looser. If the standard CDMmodel is correct, luminous matter does not trace most darkmatter structure below the scale of bright galaxies (halos ofmass 1011 1012M), and does not traceanyof it below thescale of the smallest dwarf galaxies (halo masses of 106 107M). In the default model, the dark matter candidate ismassive, cold, weakly interacting, and stable. In this case,dark matter structure extends down to scales of 0.1 A.U.or less (halo masses of 106M or less) and is almost scaleinvariant all the way down [17]. To test this incredibleprediction of 10 decades or more of invisible structure fillingour universe, we need to study dark matter in the highly
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


3/18

Advances in Astronomy 3

nonlinear regime, deep in the heart of halos. This is wherethe smallest and oldest dark matter structures end up, andit is also where dark matter reaches its highest density andwhere new physicsscattering, annihilation, or decay intoother particlesshould be most evident. In fact, since theMilky Way is embedded in a dark matter halo and we reside

relatively close to its centre (within the central 3% in radius),any local study of dark matter must come to grips with thehighly nonlinear regime of CDM structure formation. Inthat sense, our view of dark matter halos is necessarily aninsiders view.

In this paper, I review some of what we know and whatwe can learn about the internal structure of dark matterhalos. The literature in this field is extensive, so rather thanproviding an exhaustive survey I have focussed on a few keyconcepts and simple results. Since reviews on this subjectare unfortunately rare, I have tried to include enough basicexplanatory material to make the main results accessible toa broader audience of nonspecialists. In Section 2, I firstintroduce some basic elements from the standard theoryof structure formation, and clarify why halos correspondto nonlinear structure.Section 3 reviews the surprisinguniversality of halo properties that has emerged fromnumerical simulations, explains how halos grow and evolvethrough mergers and accretion, and introduces the conceptof halo age as a description of the degree of internalrelaxation. Section 4 considers methods for determininghalo ages observationally, using semianalytic models ofhalo substructure to evaluate their effectiveness. Finally inSection 5I discuss the prospects for these methods in currentand future observations. I do not discuss specific dark mattercandidates or their properties in detail in this paper, sincethese have been extensively reviewed elsewhere (see, e.g.,[9]), but I summarize in Section 5 how measurements ofhalo properties can help constrain these candidates and otheraspects of fundamental physics.

2. From Linear toNonlinear Structure Formation

It is worth clarifying how dark matter halos and nonlinearstructure relate to the linear physics of fluctuations in theearly universe. Reviews of basic cosmology and the linearperturbation theory can be found in almost any textbook oncosmology (see, e.g., [19] for an elementary introduction, or

[20] or[21] for more advanced treatments). A more specificdiscussion of large-scale structure can be found in [22]. Anexcellent review of Press-Schechter theory is given in [23].

If we perturb the smooth matter distribution in a regionof the early universe by a small amount= ( )/aroundthe average density, then the perturbation will obey

+ 2H 32mH

2= 0, (1)

where a = 1/(1 + z) is the scale factor, H = a/a is theHubble constant, andm=/cis the matter density relativeto the critical densityc. Solving this perturbation equationin a flat, matter-dominated universe with m = 1 and

H= 2/(3t), we find solutions that grow asD(t) t2/3 andsolutions that decay asD(t) t1. Sincea t2/3, this meansthat the growing mode will increase in amplitude linearlywith scale factor and independent of amplitude. For moregeneral cosmologies the relative amount of growth is nolonger proportional toa (see[24] for a general discussion),

but it will still be independent of amplitudeprovided thatthe amplitude is small. If we decompose a given patternof fluctuations into distinct Fourier modes kof (spatial)wavevectork, linear growth will preserve the relative phasesand amplitudes of the different modes since it is independentof. Thus the spatial pattern of fluctuations and the shapeof the corresponding power spectrum will be preserved aslong as the amplitude of fluctuations remains small. Thisfeature of linear growth makes it particularly easy to comparefluctuations at different epochs and has led to the highlydeveloped statistical machinery used to study the CMB andlarge-scale structure.

At early times, the matter density field resulting frominflation should consist of Gaussian fluctuations with uncor-related phases around a nearly critical mean density. Sincethe phases and relative amplitudes of fluctuations in thisfield are unchanged during subsequent linear growth, wecan describe them independent of redshift by dividing theiramplitude by a linear growth factorD(z) that measures theamount by which linear growth has amplified fluctuationsby redshiftzrelative to some reference epoch. (Note that thelinear growth factor is sometimes defined asg= D(a)/a, andD(a) is referred to as the amplitude of the growing mode,e.g., in [24].) Normalizing the growth factor to unity at z= 0,this is equivalent to considering the initial field as it would beif evolved linearly to the present day. Smoothing this linearlyevolved field on a scaleRwe obtain a new Gaussian randomfield of variance 2(R), and this variance alone is enoughto characterize the matter distribution statistically on thesescales. The function(R) summarizes all information aboutthe (linear) power spectrum, while all information aboutlinear growth is contained inD(z).

As fluctuations grow to amplitudes 1, it is cleartheir growth must accelerate beyond the linear rate. Sincethe early universe is very close to the critical density, positivedensity fluctuations exceed this limit and represent regionswith net positive curvature. These regions will eventuallystop their expansion, turn around, and recollapse, at whichpoint their density diverges formally. In 1972, Gunn and Gottconsidered the behavior of a spherical region of uniform

(over-)density and showed that it would recollapse at a well-defined time [25]. The collapse time can be expressed interms of the initial conditions at some early time when theamplitude of the fluctuation is small; it corresponds to thetime by which the initial fluctuation, had it grown at thelinear rate, would have reached a critical thresholdc. Thiscritical threshold has the valuec= 3/5(3/2)2/3 1.69 fora wide range of cosmologies (see, e.g., [22,26]).

This simple result leads to a very clever analytic estimateof the abundance of collapsed halos, derived by Press andSchechter [27]. Consider a region which has a sphericallyaveraged density contrast 0< (z)< cat redshiftz(or scalefactor a = 1/(1+ z)). The region will recollapse by the present
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


4/18


day if its linearly evolved density contrast exceedscatz= 0.We can write this condition:

(0) = (z)D(z)

> c, (2)

whereD(z) is the linear growth factor normalized to unity at

the present day and(0) is the amplitude of the fluctuationevolved linearly to the present day. The fraction of thelinearly evolved matter density fieldFcontained in collapsedobjects at redshiftz= 0 is thus simply the fraction of regionswith density contrast(0)> c; that is,

F(0, R) = 1

c

exp

222(R)

= 1

2erfc

(0)

2

, (3)

while the fraction contained in collapsed objects at a higherredshiftzis

F(z, R) = 1

c/D(z)exp

222(R)

= 1

2erfc

2, (4)

where (0) = (0)/(R) and = (z)/(R) = c/(D(z)(R))are the normal variates of the smoothed Gaussian randomfield, and erfc is the complementary error function. The twointegrals differ only in their lower limit; since less growth hasoccurred by redshiftza smaller faction of the field will havecollapsed, so the threshold for collapse must be higher.

The insight of Press and Schechter was that if a fractionF(R) of the matter distribution met the collapse criterionwhen smoothed on a scale R(M) = (3M/4)1/3, and thefraction was reduced to F < Fwhen the smoothing scaleincreased toR= R(M+ dM), then the difference could beassumed to have collapsed to form objects in the mass range

[M,M + dM]. Thus differentiating the previous equationwith respect to filtering scale (or equivalently enclosed mass)leads to the Press-Schechter expression for the halo massfunction:

dn

dMdM= 2

M

dF

dM(z, R(M))dM

= M

2

exp

222(R)

d

dMdM,

(5)

where the factor/Mconverts from mass density to numberdensity. An extra factor of two has crept in here to correct forthe under-counting of underdense regionsa subtle point in

Press-Schechter theory (see [28] for a rigorous deviation).The beauty of the Press-Schechter approach is that it

relates the abundance of nonlinear halos to fundamentalquantities from linear physicsthe growth factor D(z)and the spectrum of fluctuations, represented by (M).The model can also be extended to calculate conditionalprobabilities, for example, the probability that a point willbe contained in a collapsed region of massM1 at redshiftz1and then in a region of mass M2 at redshiftz2. I will discussextended Press-Schechter (EPS) theory and these conditionalstatistics further inSection 4. In its simplest form, however,Press-Schechter theory has many inconsistencies and onlyapproximately describes the behavior seen in simulations

(see, e.g., [29]). I will discuss more recent numerical workon halo evolution in the next section.

What do analytic arguments tell us about the internalstructure of dark matter halos? The collapse of a completelycold spherical shell, for example, in the model of Gunn &Gott discussed previously, would produce a spike of infinite

density at the centre of the perturbation. Real perturbationswill only reach finite density, however, because even colddark matter has some residual random motion relative tothe Hubble expansion, and because the region surroundinga perturbation will not be perfectly spherically symmetric.Both these effects add angular momentum to the orbits ofinfalling particles, keeping them away from the point at r=0. In general, any deviation from symmetry will be amplifiedby the collapse, leading to a mixing of orbits and the rapidvirialization of the system, through which its central regionreaches the virial equilibrium:

W= 2K (6)between potential energy

Wand kinetic energy

K. One of the

main results of the spherical collapse model of Gunn & Gottis an estimate of the final density a virialized region achieves.This can be derived from energy conservation, which showsthat the final size of a spherical region after collapse will behalf its size at turn-around (the moment at which its radialvelocity is instantaneously zero). Meanwhile the surroundinguniverse will continue to expand, increasing the contrastbetween the collapsed region and the mean density of theuniverse. A numerical calculation (see, e.g., [21]) shows thatthe density contrast relative to the background densitywillbe cvir/ = 182 178 in an Einstein-deSitter universe,or a slightly larger value in LCDM cosmologies.

Having reached this equilibrium, the virialized region nolonger feels the universal expansion around it, and in theabsence of subsequent accretion, its physical size will remainconstant with time. This idealized situation is never achievedfor any real halo, however. Since a halo represents a regionwith > 0, one can always find a larger region aroundit with smaller density contrast such that > > 0.Since this larger region also has net positive curvature, it willrecollapse in turn, adding new material to the virialized halowithin it. Thus halos never remain isolated, but continue toaccrete material from the universe around them. While theircentral regions are in approximate equilibrium, some newmaterial is being mixed into these regions at all times, andtheir outer regions are constantly growing to include more

and more mass. In a system like the Milky Way, for instance,the virial radius (i.e., the radius within which W 2K)is estimated to be around 300 kpc (see, e.g., [30, 31] andreferences therein), but matter is accreting onto the halofrom a much larger region extending out to1 Mpc. Sincevirialization produces a constant density contrast relative tothe background, all halos at a given redshift are predicted tohave the same mean density interior to their virial radius,and this virial density decreases with time as the backgrounddensity decreases.

Averaging over large scales, the density field of theuniverse can be decomposed into several distinct compo-nents (see, e.g., [32]): (1) the linear regime of large-scale,
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


5/18


low-amplitude fluctuations with fixed comoving size anda linearly growing amplitude; (2) the collapsed, virializedregions corresponding to individual halos, whose sizes anddensities evolve slowly through mergers and accretion; (3)a quasilinear regime which interpolates between these two.The division into different regimes is made formal in halo

models, which decompose the clustering of galaxies intoclustering within a single halo on small scales (the one-halo term) and clustering of multiple halos due to linear orquasilinear fluctuations in the large-scale matter distribution(the two-halo term) [33]. In the present day universe, thedivision between the two regimes occurs on scales of roughly1 Mpc. In what follows I will consider the matter distributionwithin halos, which determines the nonlinear, or one-halo,contribution to the matter distribution.

3. The Universality of Halo Properties

3.1. A Universal Density Profile. Neither the spherical col-

lapse model nor Press-Schechter theory alone specifies whatthe internal structure of dark matter halos should be.One analytic approach to estimate this is to generalize thespherical collapse model, considering the one-dimensionalcollapse of concentric shells of different radii and densities.A sensible choice for the initial run of density with radiusin such a model is the matter (auto-)correlation function,which describes clustering statistically in terms of the averageexcess density of matter around a point, and can be derivedfrom the power spectrum. Several early analytic models ofthis kind, notably [25,34,35], predicted that matter wouldcluster around a point to produce a radial density profile thatwas a steep power law with a constant slope.

Despite these analytic insights, real progress on thematter distribution inside halos did not occur until numer-ical simulations of structure formation started to resolveindividual halos. Building on initial work which made it clearthat halo density profiles were not simple power laws [3639], Navarro et al. [40,41] determined that the halo densityprofile was well fit by a single functional form:

(r) = sr3s

r(rs+r)2 . (7)

Not only was this form not a simple power law, with suitablechoices ofrs andsit seemed to fitallthe simulation results,independent of halo mass, power spectrum, and cosmology.

The profile, since named the Navarro-Frenk-White (NFW)profile or universal density profile (UDP), was the firstof many indications of the surprising simplicity of haloproperties.

The radial profiles of halos have a number of interestingfeatures. First, rather than being scale invariant they containa characteristic length rs, the NFW scale radius, at whichthe logarithmic slope of the profile is d ln /d lnr= 2.This scale is often defined with respect to the virial radiusrvir via the concentration parameterc rvir/rs. As discussedbelow, the scale radius seems to mark the division betweentwo phases in the assembly of the halo, a rapid phase wherethe central part of the profile builds up with (r) r1, and

a slower phase where an outer envelope forms around thehalo with a steeper profile going as(r) r3. The velocitydistribution inside halos is roughly isotropic, that is,v,rv, v,, although the outer regions have a slight radialbias, probably reflecting the continued infall of materialon radial orbits [42]. Perhaps the greatest surprise is the

pseudo-phase space density (r)/3

(r) one can constructfrom the density and velocity dispersion profiles. In therelaxed part of the halo this quantity is a featureless powerlaw, suggesting the possibility of a fundamental connectionbetween the inner and outer parts of the density profile [43].

3.2. Universal Patterns in Halo Growth. As suggested above,the features of the UDP seem to be connected to theevolutionary history of halos. The growth of an individualhalo can be described by specifying its total mass (say themass within the virial radius, defined as the region with meandensityctimes the critical density) as a function of redshiftor scale factor. In what follows I will refer to this function

M(z) as the mass accretion history (MAH) of a halo andoften consider it normalized by the valueM(z= 0). Studiesof halo growth in N-body simulations show that MAHs havea characteristic shape, consisting of rapid growth at earlytimes and slower growth at late times, with the break pointbetween the two varying from halo to halo. They have be

fitted to the functional formM(z)/M(0)= (1 + z) [44] aswell as to the formM(z)/M(0)= exp(z)[45], but morerecent work [18,46] makes it clear both terms are requiredto fit the full range of MAHs seen in simulations. In [ 18],McBride et al. fit a general 2-parameter form for the MAH:

M(z)

M(0)=(1 +z) exp

z, (8)but find that the fitted values of and they derive arestrongly correlated in their ensemble of MAHs, suggestingan even better parameterization may exist.

These analytic approximations to the typical MAH areparticularly interesting because of the strong connectionbetween the MAH of a halo and its density profile. Crudelyspeaking, when halos grow rapidly their density profiles stayrelatively shallow. From the work of [47], the scale radiusrsappears to increase in sync with the virial radius during thesegrowth spurts, such that the concentration parameter staysat a roughly constant value c 4. During slower phases ofevolution, while the virial radius continues to grow roughly

as the scale factor of the universe (since the virial densitycontrast is approximately constant), the scale radius staysconstant. Thus concentration increases as scale factor duringthese quiescent periods. Based on these patterns, Wechsleret al. [45] and Zhao et al. [47] provide similar algorithmsfor predicting a halos concentration at any time, given itsMAH. (See also Lu et al. [48], which ties the precise shapeof the density profile to the MAH.) In the particularly simplemodel of Wechsler et al., c = c0(a/ac), wherec0 4 is theconcentration of a halo undergoing rapid accretion and theterma/acis the ratio of the present scale factorato the scalefactorac at the time the halo last stopped accreting rapidly.Present day galaxy halos have typical concentrations of12,
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


6/18


indicating that they stopped growing rapidly at ac = 1/3or z= 2, while galaxy clusters have concentrations of 46,indicating that they have just formed recently (ac= 1 2/3,orz= 0 0.5).

3.3. Halo Shape and Spin. Just as simulated halos have a

well-defined distribution of MAHs, correlated with theirconcentration parameter, so too they show a regular anduniversal distribution of shapes. Halos are generally triaxial,with axis ratios ofb/a c/b 0.6 0.8 (see, e.g., [4951]) (although their potentials may have slightly differentshapescf. [52]). Typical examples are more often prolate(cigar-shaped) than oblate (disk shaped). Shape is correlatedwith age or merger history (see, e.g., [51]), although thiscorrelation has been established only in an average sense. Aphysical mechanism which may account for the correlationwas outlined in [53]. A system of two halos merging on aradial orbit can be described in terms of a tensorial versionof the virial theorem, in which the contributions of kineticand potential energy must eventually reach an equilibriumcomponent by component. Dissipationless mergers betweendark matter halos should roughly conserve the individualtensor components, such that the final merger remnant willremember the original orientation of the infalling pair andwill remain more extended along that axis. From this pointof view, a halos shape may provide an interesting clue to theorientation of the merger that formed it. Tests of this ideaare possible using gravitational lensing (see, e.g., [54]) or theorbits of satellite galaxies (see, e.g., [55]) to measure haloellipticity with respect to local structure.

The internal velocity distribution of halos also showsregular patterns. In particular, halos seem to have a fairlyuniversal distribution of internal angular momentum [56]and of net spin (see, e.g., [51] and earlier references therein).Overall the spin is small; defining a dimensionless spinparameter from the size, mass and energy of a system, thetypical value for halos is = 0.030.05. Thus spin contributeslittle to the support of the system against gravity. The angularmomentum distribution within halos presumably relates tothe MAH and to the features of the density profile (see, e.g.,[51] shows that rounder halos have less net spin on average,for instance), although the connection is still somewhatunclear.

3.4. Halo Substructure. The standard picture of halo for-mation suggests halos should contain dense substructure,

corresponding to the visible baryonic structure in groupsand clusters. Halos accrete matter continuously from theirsurroundings, and this matter may include other virializedhalos. In fact, in CDM cosmologies the shape of the powerspectrum is such that the variance2(M) increases at smallmasses, so small regions have a greater range of densitycontrast than large ones and are more likely to cross thethreshold for collapse at early times. As a result, at anytime there are many more small halos in the universe thanlarge ones. The smallest halos collapse and virialize at theearliest times on average, reaching high virial densities. Theymay then be incorporated into larger halos where theircores can survive as dense substructures, or subhalos. The

most massive subhalos (i.e., those closest to their parenthalo in mass) evolve quickly, but otherwise the evolutionis relatively independent of subhalo mass. Thus the massspectrum of subhalos within a halo at any given time reflectsthe average cosmic mass function of halos present in theaccreted material, although the normalization of the subhalo

mass function decreases slowly as individual subhalos losemass through tidal heating and stripping.

From this picture, substructure should correlate with theage of the halo, in the sense of the mean time since itsmass was assembled into a single object. For a given initialspectrum of subhalos, this prediction can be made quanti-tative by determining the mean number of orbits subhaloshave spent in the parent halo and assuming a certain amountof mass loss occurs once per orbit at pericentric passage.Slightly more elaborate analytic or semianalytic models ofthe evolution of the subhalo mass function were developedin[5759]. Beyond the systematic variation with halo age,the shape of the subhalo mass function is approximately

universal. It consists of a power law with an exponentialcutoff at Msat/Mmain 0.1, as discussed below (see also[60,61]).

While the relationship between simulated halos andobserved field galaxies seems reasonably straightforward,the connection between subhalos and group or clustermembers has historically been much harder to understand.The earliest simulations found no substructure at alltheovermerging problem, which turned out to be due to alack of mass and force resolution [62]. After finally resolvinghalo substructure in clusters [6365], simulations quicklystarted producing too much of it in group or galaxy-sizedhalos [66, 67]far more than was needed to host visible

dwarf galaxies in the Local Group, for instance. This issuecontinues to generate controversy, but is beyond the scope ofthis paper (see [68] for a recent review). Suffice itto say thatifthe LCDM model is correct, the current generation of high-resolution simulations of individual halos [69,70] indicatethe existence of a huge amount of substructure around theMilky Way that is not traced by baryons.

3.5. The Dynamical Evolution of Halos. So far I have focussedon the static properties of halos, but halos are dynamicalsystems. As mentioned above, they are constantly accretingnew material, both relatively smooth matter and a spectrumof other halos. Mergers of one halo with another have long

interested simulators, as they may give rise to galaxy mergersand drive some of the more spectacular evolution in galaxymorphology. Thus there has been extensive analytic theoryand numerical work on this subject. I will summarize somesalient points here.

In the spherically-symmetric limit of halo growth, sub-halo orbits are expected to be purely radial and start theirinfall with a fixed energy corresponding to their potentialenergy at turn around. In real halos, departures fromsymmetry in the immediate surroundings of a system willscatter the initial energy and angular momentum of satellites.The angular momentum distribution of subhalo orbits atinfall (and to a lesser extent the energy distribution) have
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


7/18


been studied in detail by many authors (see, e.g., [63, 64,71, 72]); see also recent work by [73, 74]. The angularmomentum of an orbit can be characterized by its circularityparameter = L/Lc, where L is the angular momentum of theorbit and Lcis the angular momentum of a circular orbit withthe same energy. In terms of, the initial orbital distribution

is almost uniform, with a slight peak around= 0.5. Giventhe steep density profile of dark matter halos, this angularmomentum distribution produces mainly radial orbits, witha mean axis ratio of 6 : 1[64,75].

Subsequent to the merger, subhalo properties evolveunder the influence of dynamical friction, tidal mass loss, andtidal heating. Dynamical friction drags in the most massivesatellites, but becomes negligible belowMsat/Mmain 0.001(see, e.g., [58, 76, 77]). Tidal mass loss peaks stronglyonce per orbit at passage through pericentre. This episodicmass loss should produce coherent streams of (dark matter)debris like those seen in disrupting stellar systems. Afterdetaching from the progenitor subhalo, tidal streams evolvequite simply with time, and their physics is well understood(see, e.g., [78,79]). Because mass loss, heating, and streamformation occur once per orbit at the time of pericentricpassage, the characteristic timescale for evolution in haloproperties is the timescale for radial oscillations in the orbit,Prad = 2/, where is the epicyclic frequency. At anyredshift, this timescale is comparable to the instantaneousHubble time H(z)1 [58]. Beyond the first few orbits, thelonger-term evolution of subhalos is an unresolved problem.It is not clear when (if ever) subhalos are disrupted byrepeated mass loss [80]. This issue is particularly relevant tothe fate of the smallest subhalos, which may or may not havesurvived for hundreds of orbits in the halo of the Milky Way[17,8183].

In summary, halos show many simple trends andproperties that are universal, in the sense that they applyindependent of mass or cosmology. The universal densityprofile is the best known of these properties, but theregular trends in mass accretion history (MAH) and thecorrelations of other quantities with the MAH deservefurther consideration. They suggest that age is possibly themost important property of a halo. Recently assembledhalos have massive substructure, nonspherical shapes, lowconcentration parameters and possibly more spin and/orvelocity anisotropy. Older halos are generally smoother, morespherical, and more concentrated. In next section I considerone specific example of a trend with age, the correlation

between age and substructure, in more detail.

4. Defining and Measuring Halo Age

If age is the organizing principle in halo properties, thereremains the question of how to define and measure it. Apragmatic approach is simply to construct model halos,study them with mock observations, and see what observa-tions produce the clearest determination of age or formationhistory. A large number of model halos are required to fullyexplore the multivariate distributions in halo properties, sosemianalytic halo models are a convenient tool to use forthis task. In this section I will use the model developed in

[57, 58, 84, 85] to study halo substructure and ask howwell lensing or X-ray studies of cluster substructure candetermine the cluster age distribution.

4.1. Models. The semianalytic model introduced in[57,58]consists of two components: a Monte-Carlo algorithm for

generating random merger histories, or merger trees, forindividual halos, and an analytic description of subhaloevolution which is applied to each satellite subhalo as itmerges with the main system. I summarize each componentbriefly below; for a full description see [57,58,84,85].

4.1.1. Merger Trees. In the extension to Press-Schechtermentioned in Section 2, the growth history of a halo isdetermined by the distribution of mass around it, averagedspherically on successively lager scales. Ignoring higher-order correlations, this distribution is Gaussian with variance2(M) when averaging on a scale R = (3M/4)1/3, where

is the mean density of the universe. The entire history of

a halo can be generated by taking random walks in densityfluctuation , using a Gaussian normal variate scaled by(M(R)), as the scaleRdecreases from infinity (where= 0,since the density must equal the mean value by definition) tozero, wheremay diverge. The resulting trajectory(R) or(M) is then mapped onto evolution with time or redshiftusing the condition for collapse > c(z) = c(0)/D(z).Since can increase or decrease with scale, the trajectorymust be filtered such that (R) increases monotonically asRdecreases; this corresponds to finding thelargestvalue ofRin the trajectory with a givenand assuming that that scalecollapses when > c(z)[26].

This process produces a single, randomly generated but

representative mass accretion historyM(z). Because of thefiltering operation, this trajectory may have discontinuousjumps in mass. These are interpreted as mergers, in which themass of the halo instantaneously increases by a finite amount.If each merger involves a single other halo, we can associatethe mass change with this new halo and follow its evolutionalong with the main branch. By following every merged haloin every branch iteratively down to some mass resolutionlimit, we can generate a merger tree describing the entireformation history of our original object.

There are many subtleties to generating accurate mergertrees, and no one method is ideal. A recent comparison ofmethods is given in [86]. An analytic expression for the

merger probability in a given redshift step is available onlyfor binary mergers in which two halos merge, but theseevents do not normally account for the whole mass of thefinal system, leading to additional branches in the merger

tree and/or accreted mass below the mass resolution of thetree. The trees used in this section were generated using themethod of Somerville and Kolatt [87], which picks time stepsshort enough that binary mergers are very likely in the tree.In practice, this method breaks the main progenitor downinto small progenitors or accreted mass slightly too quickly,leading to mass accretion histories that are slightly too young;this is a common problem for many merger tree algorithms[44,86,88].
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


8/18


5 0 5 10

0

2

4

6

8

10

(a)

5 0 5 10

0

2

4

6

8

10

(b)

Figure 1: Distribution of parameters describing the mass accretion histories of 1000 randomly generated merger trees. (a)Colors correspondto the classes defined by McBride et al. in [18]: blue: class I, yellow: class II, orange: class III, red: class IV. (b) colors correspond to theformation redshiftz50: magenta:z50 < 0.3, blue:z50= 0.30.5, green:z50= 0.50.8, yellow:z50= 0.81.1, red:z50 > 1.1. The dotted lineindicates the locus of trajectories with = 0. For 1, formation redshift is roughly constant along this lines parallel to this one.

Part of the inaccuracy may come from the conditionalprobability used in the method. In standard EPS theory[26], the probability of going from a halo of mass M1 atredshift z1 to a halo of mass M2 at redshift z2 can bewritten as a simple function of a single variable

=(c(z))/(2(M)), where (c(z)) = c(z1) c(z2) =c,0(1/D(z1) 1/D(z2)) and (2(M))= 2(M1) 2(M2).This function is simply the unconditional probability fromthe mass function, shifted to a new origin in :

P()dM= PPS()dM=

2

exp

2

2

d()

dM dM.

(9)

This conditional probability assumes a collapse thresholdindependent of mass: c(M, z) = c(z), such that (c) =c(z1) c(z2). This would be the case for spherical halos;for nonspherical halos the collapse threshold is different, andsystematic trends in halo shape with mass introduce a net

mass dependence in the collapse threshold, which can befitted by

c(M, z) =

asc

1 +b

a2

, (10)

with a 0.707, b = 0.485 and 0.615, where sc isthe normal spherical collapse threshold [89, 90]. Now thechange in barrier height (c) = c(M1, z1) c(M2, z2)will depend on both masses and both redshifts separately,complicating the merger tree calculations. We can simplifythe calculations, however, by taking = 0.5 and using anapproximation to the change in collapse threshold:

(c)

a(sc(z))+b. (11)

For a fixed jump in redshift z1 z2, finding the exactsolution to this equation still requires iteration since itself depends on c, but using the spherical value as anapproximation to , the approximate change in thresholdcan be calculated in one pass. Comparison with simula-

tions (Taylor, in preparation) shows that this approximateellipsoidal barrier provides a good match to measured halomerger probabilities at high values of, while the sphericalbarrier works well at low values, so in practice the code usedhere switches between the two forms at 0.15. Usingthese adjustments to the merger probability, the correct massdependence can be included in the merger trees, improvingtheir age distribution and MAHs.

Figure 1shows the distribution ofandvalues fitted tothe semianalytic merger trees using2-fitting of their MAH.The distribution is very similar to the one found by McBrideet al. for halos in the Millenium simulation [18], althoughwith a slight shift to higher values of and/or lower values

of. This may indicate that the semianalytic trees are stillslightly too young. (Note that the distribution plotted infigure A1 of McBride et al. includes halos of much lowermass, so this explains some of the difference.) Mcbride et al.classify MAH shapes into 4 classes, as indicated by the colorsin the left-hand panel. Compared to their statistics, 47%,36%, 10%, and 8% of the semianalytic trees are in classesI, II, III, and IV, respectively, versus to 18%, 57%, 17%, and8% of the most massive halos of the Millenium simulation.Thus, while the two distributions overlap significantly, thedifference between them suggests that absolute numbers orage distributions derived from semianalytic merger trees willneed to be calibrated using simulated halos. The merger trees
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


9/18


0.0001

z= 0z= 0.2

0.001 0.01 0.1

Msat/Mmain

0.1

1

10dN/dlnM

100

1000

(a)

0.0001 0.001 0.01 0.1

Msat/Mmain

0.1

1

10dN/dlnM

100

1000

(b)

Figure2: (a) The average differential mass function of subhalos within the virial radius for halos at redshift z= 0 (solid blue line) andz= 0.2 (dashed red line). (b) The average differential mass function for halos at radii less than 1.0, 0.5, 0.25, 0.125, and 0.0625 times thevirial radius (different histograms, from top to bottom). The excess of massive systems at small radii is clearly visible.

should capture relative trends in age and substructure withmass, redshift, and cosmology, however.

It is also worth noting one other complicating factorin the fit to MAHs. Both in the Millenium runs

and in the semianalytic models, measured parameter valuescluster along a sequence roughly defined at high values of

by = (dotted line in Figure 1(b)). Different pointsalong lines parallel to this sequence actually represent almostindistinguishable fits to the numerical MAH, so it is notclear that and represent the best parameterization of anaccretion history. The difference correlates much moreclosely with physical quantities like the formation redshift z50by which a halo had assembled 50% of its present-day (z= 0)mass, as shown by the colours inFigure 1(b).

4.1.2. Subhalo Evolution. Given a sequence of mergers froma merger tree, the next question is, how do these affect halo

structure? This can be divided in two distinct questions: howdo the overall properties (mass, concentration, shape, etc.)of a halo change during a merger, and how much densesubstructure survives the merger?

The model introduced in [58] assumes the halo mass(really the gravitational potential) increases instantaneouslyat the moment of the merger and ignores changes in shape,treating the halo potential as spherical. For concentration,it uses a running fit of the MAH, together with theconcentration model of [45] (chosen for its simplicity) ateach redshift. These choices give only a crude approximationto full 3-dimensional behavior of halos seen in simulations,but they do capture some of its essential features.

The evolution of merging subhalos is complex, and is dis-cussed in detail in [57,58]. Briefly, halo orbits are calculatedin the (evolving) potential of the main system; dynamicalfriction gradually reduces the energy and angular momen-

tum of subhalos, particularly massive ones; tidal strippingremoves mass from the outer parts of satellites; tidal heatingmodifies their internal structure and accelerates mass-loss;and finally encounters and collisions also contribute to massloss and scatter subhalo velocities. This combination ofphysics is required to produce realistic subhalo distributionsthat match high-resolution simulations [84,85].

4.2. The Subhalo Mass Function and Radial Distribution.Halos retain traces of successive mergers, both in thepositions and in the velocities of dark matter particles.Velocity substructure, in the form of coherent streams ofparticles on correlated orbits, is expected to be common

but is probably only detectable in the halo of the MilkyWay, where it can modify count rates in direct detectionexperiments (e.g., [9194]). Real-space substructure consistsof the dense cores of subhalos surviving from previousmergers. Since CDM structure formation proceeds fromsmaller mass halos to larger mass halos as the universeevolves from high density to low density, massive low-density halos at low redshift will contain the undigestedcores of many low-mass, high-density halos, which are stableagainst tidal stripping provided they have a few times themean density of the background (e.g., [80]).

The mass spectrum and spatial distribution of thesecores, or subhalos, are particularly simple, as illustrated
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


10/18


0 2 4 6 8 10 12Look-back time (gyr)

1015

1

1014

10Nsats

100

M(M)

0 0.2 0.4 0.6

z

0.8 1 2 3

Figure3: The evolution of a single cluster halo versus time. Thebottom panel shows how the total mass (upper black line) and themass within the central 390 kpc (lower red line) build up with time.The upper panel shows how the number of satellites changes withtime. The dashed curves are for satellites with masses in excess of1011M, 1012M,and1013M, respectively, from topto bottom. Thesolid curves are for satellites with masses of more than 0.1% or 1%of the main systems mass, respectively, from top to bottom.

in Figure 2 which shows average mass functions for an

ensemble of 3000 semianalytic halo models. Relative to themass of the main halo, the distribution of subhalo masses isa power law with an exponential cutoffaroundMsat/Mmain0.1. Normalized in this way, the mass function is almostinvariant with redshift and varies only slowly with halo mass(see also [60,61,84,85,95]). The spatial distribution withhalo-centric radius (Figure 2(b)) is also very simple untilone reaches 10%20% of the virial radius from the centre. Asmall population of massive subhalos (1 or less per halo onaverage, with masses greater than 1% of the main halo mass)is predicted in this region. These are systems which havebeen dragged in by dynamical friction and are in the processof disruption. The semianalytic models ignore the effects

of baryons, however, which should be strong in the centralregion. Thus it is not clear how many of the central mergingsystems exist in real cluster or galaxy haloshydrodynamicsimulations would help clarify the situation here.

4.3. Evolution with Time. The subhalo mass function isnot completely static, but changes systematically with time.While low-mass halos reach an equilibrium in mass, energy,and angular momentum loss after a few orbits in the mainhalo, massive subhalos evolve quickly due to dynamicalfriction. As a result, massive substructure is a key indicatorof recent growth and merging.Figure 3shows the evolutionof a single semianalytic model halo over time. The lower

z50= 0 0.25z50= 0.25 0.45

z50= 0.45 0.65z50> 0.65

0.001 0.01 0.1

Msat/Mmain

1

10

dN/dlnM

100

1000

Figure 4: The differential mass function of subhalos for systemswith different ranges of formation epoch:z50= 00.25 (blue),z50=0.250.45 (green),z50= 0.450.65 (yellow), andz50> 0.65 (red).

panel shows how the total mass of the main halo (upper,black line) and the mass within the central 390 kpc buildup with time. The upper panel shows how the numberof satellites changes with time. The dashed curves are forsatellites with masses in excess of 1011M, 1012M, and

1013M, respectively, from top to bottom. The solid curvesare for satellites with masses of more than 0.1% or 1% ofthe main systems mass, respectively, from top to bottom. Asthe halo grows, its total number of satellites increases, but themass ratio of each individual subhalo decreases, both becauseit loses mass through tidal stripping and because its parenthalo increases in mass. Thus the amplitude of the halo massfunction expressed in terms of the ratio Msat/Mmain decayswith time, particularly for large values ofMsat/Mmain (e.g.,the green solid curve in the upper panel).

As a result of these trends, the slope of the subhalomass function changes with time. This suggests a firstobservational test of halo age, based on substructure. If one

can measure subhalo masses for individual halos, particularlyat the high mass end, then the amplitude of the mass functionwill correlate with evolutionary history. The correlation isstrongest with recent tracers of halo growth such asz90, theredshift by which a halo built up 90% of its present-day(z= 0) mass [84], but is evident even for longer-term tracerslikez50, as shown inFigure 4.(Note that this is a differentialmass function; integrating the contribution from all haloswould produce an even stronger signal.) I explore this ideafurther below.

4.4. Inferring Age from Observation of Substructure. Thedifferences shown in Figure 4 suggest there should be
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


11/18


10.90.80.70.60.50.40.30.20.1

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1x

1

y 00.10.2

0.30.40.50.60.70.80.9

(a)

10.90.80.70.60.50.40.30.20.1

1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1x

1

y 00.10.2

0.30.40.50.60.70.80.9

(b)

Figure5: Two cluster halos extracted from an n-body simulation atz= 0. One is isolated and relaxed (a), while the other is in the processof forming through a major merger (b). In each case, only particles within 1 Mpc of the halo are plotted.

a strong correlation between the total amount of sub-structure in a halo and its formation history. Numericalstudies find similar results, albeit with large halo-to-haloscatter [60, 96]. At a broader level, it has long beenappreciated that cluster substructure indicates recent growth[97].Figure 5shows two regions extracted from a recent n-body simulation run with WMAP-7 parameters. For clarity,only the particles within 1 Mpc of the centre of each region

are plotted. Figure 5(a) contains relatively isolated halo,while Figure 5(b)catches an ongoing merger between twocomponents. Even with an imperfect reconstruction of thematter density, for example, via lensing measurements ofthe convergence field or X-ray measurements of the gaseousemission in each region, observations could clearly tell onefrom the other. The challenge is to detect more subtle signsof recent growth, and to do so for a large sample of halos.

Galaxy clusters are a tempting target for this sort of study,for several reasons. Luminous baryons can be used to traceat least part of the substructure in cluster halos down tovery small values ofMsat/Mmain, since individual galaxies havemass ratios of 105 or less relative to the cluster. Hot gas

provides an independent measure of the shape of the mainhalo potential, as does lensing. Lensing is the more appealingof these tracers, since it avoids possible offsets or biasesbetween the baryonic matter distribution and the dark mat-ter distribution, but it requires deep observations of massiveclusters at moderate redshifts. Many groups have used lens-ing to reconstruct the matter distributions in famous systemslike Abell 1689 (e.g., [98, 99]) or the Bullet Cluster [100].The LoCuSS survey (http://www.sr.bham.ac.uk/locuss/) is amore general attempt to measure the shape of the potentialin a large sample of cluster halos and to relate the massdistribution in clusters to their dynamical state and galaxypopulations.

These studies combine large scale weak lensing measure-ments (measurements of weak distortions in galaxy shapeonly apparent in averages over hundreds of galaxies), toget the overall mass distribution in the cluster, with stronglensing observations, which can reveal small-scale featuresin the central part of the cluster. The relative contributionof structure to the projected mass distribution in the clustercan be estimated by comparing a smooth ellipsoidal model

to the full mass distribution inferred from the observations.For a particular choice of aperture (roughly 390 kpc in thecase of LoCuSS, based on the field of view of the camera andthe redshifts of the clusters), the difference between the two,normalized to the total mass, gives the substructure fractionfsub. It is straightforward to calculate the distribution offsub in the semianalytic models, assuming a perfect set ofobservations which reconstruct the full mass distribution tohigh precision over the entire aperture. It is less clear (andthe subject of ongoing work) how gaps in the lensing mapaffect measurements of fsub in real clusters. Nonetheless,preliminary comparisons of the models with the data showgood general agreement[101].

The predicted distribution of fsubis roughly log normal,as discussed below. The substructure fraction correlatesstrongly with the formation redshiftsz90andz50defined pre-viously, although in both cases the relationship is not quitemonotonic (Figure 6). As discussed in [101], for systems thathave formed very recently, merging material will not havereached the centre of the halo yet. Thus systems withz90< 0.1sometimes have relatively low central substructure fractions.The dotted lines in the two panels indicate the formationredshift such that merging substructure has reached thecentre of the halo and is on its first pericentric passage.Beyond this point the correlation between formation epochand substructure fraction is clear and monotonic. Only the
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://www.sr.bham.ac.uk/locuss/http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://www.sr.bham.ac.uk/locuss/http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


12/18


0 0.2 0.4 0.6

z90

1

fsub

0.1

(a)

0 0.5 1 1.5

z50

1

fsub

0.1

(b)

Figure6: Lensing substructure fraction versus the formation reshifts z90(a) andz50(b). The dotted line indicates the epoch approximatelyone infall time before present.

vertical axis of these figures corresponds to an observablequantity, fsub, but clearly by binning in fsub one can selectsamples of clusters with very different age distributions. Thusmeasurements of fsub represent one realistic approach todetermining the underlying distribution of halo ages.

Finally, I note that similar work constraining substruc-ture fractions with lensing observations may also be possible

on much smaller mass scales, in the halos of individualgalaxies. While individual galaxies are not massive enoughto be detected through weak lensing, their central regionsreach surface densities high enough to produce stronglensing of background galaxies or quasars. Substructure inthe halo of a lensing system can produce variations inbrightness (e.g., [102104]), image position (e.g., [105]),or the time delays between multiple images of variablebackground sources (see, e.g., [106])see [107] for a recentreview. Separating the signals from dark and luminoussubstructure is more complicated on this scale, since thereis possible contamination from microlensing by stars in thelensing galaxy, and since many subhalos large enough to

produce detectable variations in the lensing potential donot host galaxies bright enough to detect at cosmologicalredshifts. Nonetheless, with forthcoming large samples ofstrong lenses, this may be promising avenue for measuringdark substructure on very small scales (e.g., [108]).

5. Prospects

The simple, universal distributions of halo parametersconcentration, shape, spin, substructurethat emerge fromnumerical simulations, and the strong correlations foundbetween these quantities, suggest the one or two mainproperties determine much of the internal structure of

dark matter halos. Section 3 reviewed the evidence thata halos concentration parameter is determined directlyby its mass accretion history, and in Section 4 I shownhow substructure should also reflect formation history. Thesemianalytic models used to study this relationship aresomewhat approximate and need confirmation from a largesuit of numerical simulations (Wong et al. in preparation),

but the predicted trends are clear and match those seenin earlier numerical work. Thus in these two examples,the fundamental parameter which accounts for most of thescatter in halo properties seems to be age, in the broad senseencapsulated by the mass accretion history.

Mass accretion histories can in principle contain anarbitrary number of independent degrees of freedom. Basedon the numerical fits discussed inSection 3, real examplesseem to form a much more limited set with only one(or possibly two) main degree(s) of freedom. Likewise,while the shape of the substructure mass function can inprinciple have many degrees of freedom, it seems to followa single-parameter sequence in practice, and of course scalar

properties like concentration necessarily form 1-dimensionaldistributions. All this suggests there may be a single bestmeasure of halo age, best in the sense that it capturesmost of the variety in halo properties. This exact identityof this parameter, its physical interpretation, and the bestobservational estimators for it remain to be determined.

Assuming a single-parameter sequence of halo ageor degree of relaxation exists, why bother measuring it?Understanding how halos achieve their particular internalstates is intrinsically interesting, of course, but informa-tion about halo structure and dynamics also has severalimmediate practical applications. As discussed in Section 2,halo occupation (HOD) models predict or interpret galaxy
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


13/18


m= 0.27m= 0.23

0.1 1

fsub

0

0.05

P(fsub)

0.1

0.15

(a)

m= 0.27m= 0.23

0.1 1

fsub

0

0.05

P(fsub) 0.1

z= 0.20.15

(b)

Figure7: Distributions of lensing substructure fraction at z= 0 (a) andz= 0.2 (b). The solid histogram is for 1000 merger trees in acosmology with m= 0.27, 8= 0.80. The dashed histogram is for 1000 merger trees in a cosmology with m= 0.23, 8= 0.85. Thesetwo cosmologies are chosen to produce the same abundance of massive clusters at low redshift, so cluster number counts alone would notdistinguish between them.

clustering by matching galaxies to halos, and this approachis also useful in producing mock catalogues of galaxies

from dark-matter-only simulations. These models normallyassume that mass is the most important halo property tomatch, despite evidence that halo age is correlated withlarge-scale clustering and environment [109115]. Betterdefinitions and metrics for age, together with simulations ofgalaxy formation in halos with different formation histories,would clarify the connection between the age of the host haloand its resident galaxys properties.

A second practical application is testing standard struc-ture formation and breaking parameter degeneracies. Onesimple example is the age distribution of galaxy clusters.Clusters form through the amplification of peaks in theearly density field by the growth factorD(z). Measuring the

abundance of clusters constrains the product of the powerspectrum and the growth factor, leading to a degeneracybetween m and 8 (8 = (M(R)) for R = 8h1 Mpc,where h = (H0/100kms1 Mpc1)) or for more generalcosmologies, between late-time growth and initial power.At fixed cluster abundance, there is a physical differencebetween cosmologies with low m and high 8 and thosewith highmand low8, however; in the former case clustersform earlier, and should be more relaxed, rounder, moreconcentrated and have less substructure.Figure 7shows howthe substructure fraction distributions reflect the differencein cluster ages for two very similar cosmologies locatedalong the degeneracy in the m 8 plane. The shift in

the distribution is subtle (partly because substructure isbeing measured in projectionseeSection 4), but should be

measurable with a sample ofO(100) clusters, provided theeffect can be calibrated carefully in simulations.

In the longer term, a more fundamental goal of the studyof nonlinear structure is to understand dark matter itself.Nonlinear structureor halo properties more specificallyare connected to the fundamental physics of dark matterin several ways. Throughout this paper I have assumedthat all dark matter is a single particle of a plain vanillasort, that is cold, collisionless, and stable over cosmologicaltimescales and on all the length scales we can probe. Infact a much greater range of possibilities existsee [9]for a comprehensive review. Real candidates could behavequite differently on small scales or at high densities, their

properties could vary over cosmological time, or they couldstart offwith a quite different initial distribution in the earlyuniverse. Detailed studies of dark matter halos will helpconstrain all three of these possibilities.

There are a number of particle properties which wouldonly become evident at high densities or on small spatialscales. Supersymmetric dark matter particles such as neu-tralinos are their own antiparticle and can annihilate withone another, producing gamma rays and other secondaryparticles with energies in the GeV-TeV range. This emissionis strongly weighted to the densest part of halos and thusdepends sensitively on the amount of substructure within ahalo (see, e.g., [116] for a recent review). In principle, dark
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


14/18


matter particles could also have important collisional cross-sections with themselves and/or normal matter, or theircoupling to gravity might even be different. All these possibil-ities are already quite strongly constrained by current obser-vations, however. Observations by the Fermi Gamma-raytelescope constrain the annihilation cross-section for dark

matter [117], observations of the bullet cluster constrainthe elastic scattering cross-section [100] and observationsof the disruption of the Sagittarius dwarf galaxy constrain amodified gravitational coupling [118]. As our measurementsof halo structure improve, probing denser substructure andsmaller mass scales, these constraints will grow stronger.The time evolution of dark matter properties is slightly lesswell constrained, but since most of the evidence for densesubstructure in dark matter halos comes from low redshift,this suggests that dark matter particles have remained stableover the age of the universe. Better observational constraintson dense halo substructure and the overall distribution ofhalo ages could strengthen this argument considerably.

The initial conditions for structure formation are amuch more open question. Modifications to dark mattercouplings, or variations in the equation of state of theuniverse at early times, could leave their imprint in thepower spectrum of fluctuations from which halos formed.Warm dark matter, strongly annihilating or rapidly decayingdark matter, collisional dark matter, and other variationson the plain vanilla model all produce a truncated powerspectrum and a minimum scale to structure formation. Thiswould eliminate halo substructure below that scale, andproduce rounder and more relaxed halos on slightly largerscales. There have been many appeals to modified physicsof this kind to explain apparent discrepancies between thedistributions of dark and luminous matter in galaxy cores(as reviewed in [119]) or in the halo of the Milky Way (asreviewed in [68]). These arguments have the problem thatthe baryonic tracers themselves are expected to disappear onsmall scales, due to strong negative feedback effects in starformation and galaxy formation on these scales. When we failto observe dense, small-scale dark structure in the universe,it is unclear whether this is telling us something aboutthe behavior of dark matter on these scales, or somethingabout the behavior of baryons. Current and future lensingexperiments will be crucial for resolving this problem andestablishing or disproving definitively the existence of darkstructure on subgalactic scales. This in turn should lead tostrong constraints on warm or collisional dark matter, or

similar models.A generation of numerical experiments and analytic

work have made remarkable progress in understandingnonlinear structure formation and the deeply nonlinearinternal properties of dark matter halos. While the fieldlacks the analytical simplicity of linear theory and may seemless suited to precision cosmology as a result, some simplepatterns have emerged from simulations. Dark matterhalo properties correlate strongly with their formationhistory and may even form a single-parameter sequencein age, provided this parameter can be defined clearly.With powerful new observational measures of halo structureand substructure becoming available, small-scale nonlinear

structure formation may be the next great source of tests ofcosmology, dark matter, and fundamental physics.

Acknowledgments

This work was supported by a Discovery grant from the

Natural Sciences and Engineering Research Council ofCanada. The auther thanks Anson Wong for providing thedata used to make Figure 5. The auther also is happy toacknowledge many useful conversations about dark mattermodels, astrophysical constraints on dark matter, and theprospects for dark matter identification, with the partic-ipants of the Keck Institute for Space Studies workshopShedding Light on the Nature of Dark Matter, which wasfunded by the W. M. Keck Foundation. Finally, I thank P.Salucci for clarifying the original evidence for dark matterin rotation curves.

References

[1] F. Zwicky, Die rotverschiebung von extragalaktischenNebeln,Helvetica Physica Acta, vol. 6, pp. 110127, 1933.

[2] V. C. Rubin and W. K. Ford Jr., Rotation of the andromedaNebula from a spectroscopic survey of emission regions, The

Astrophysical Journal, vol. 159, p. 379, 1970.

[3] D. Larson, J. Dunkley, G. Hinshaw et al., Seven-yearWilkinson Microwave Anisotropy Probe (WMAP) obser-vations: power spectra and WMAP-derived parameters,http://arxiv.org/abs/1001.4635.

[4] B. A. Reid, W. J. Percival, D. J. Eisenstein et al., Cosmologicalconstraints from the clustering of the Sloan Digital SkySurvey DR7 luminous red galaxies, Monthly Notices of theRoyal Astronomical Society, vol. 404, no. 1, pp. 6085, 2010.

[5] P. Astier, J. Guy, N. Regnault et al., The supernova legacysurvey: measurement of M,L and w from the first yeardata set,Astronomy and Astrophysics, vol. 447, no. 1, pp. 3148, 2006.

[6] R. H. Cyburt, B. D. Fields, K. A. Olive, and E. Skillman, NewBBNlimits on physics beyondthe standard model from 4He,

Astroparticle Physics, vol. 23, no. 3, pp. 313323, 2005.

[7] J. L. Feng, Dark matter candidates from particle physicsand methods of detection,Annual Review of Astronomy and

Astrophysics, vol. 48, pp. 495545, 2010.

[8] D. Hooper and E. A. Baltz, Strategies for determining thenature of dark matter,Annual Review of Nuclear and ParticleScience, vol. 58, pp. 293314, 2008.

[9] G. Bertone, D. Hooper, and J. Silk, Particle dark matter:evidence, candidates and constraints, Physics Reports, vol.405, no. 5-6, pp. 279390, 2005.

[10] M. J. Drinkwater, R. J. Jurek, C. Blake et al., The WiggleZdark energy survey: survey design and first data release,

Monthly Notices of the Royal Astronomical Society, vol. 401,no. 3, pp. 14291452, 2010.

[11] D. Schlegel, M. White, and D. Eisenstein, The Baryonoscillation spectroscopic survey: precision measurement ofthe absolute cosmic distance scale, http://arxiv.org/abs/0902.4680.

[12] D. J. Schlegel, C. Bebek, H. Heetderks et al., BigBOSS: theground-based stage IV dark energy experiment,http://arxiv.org/abs/0904.0468.
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://arxiv.org/abs/1001.4635http://arxiv.org/abs/1001.4635http://arxiv.org/abs/0902.4680http://arxiv.org/abs/0902.4680http://arxiv.org/abs/0902.4680http://arxiv.org/abs/0904.0468http://arxiv.org/abs/0904.0468http://arxiv.org/abs/0904.0468http://arxiv.org/abs/0904.0468http://arxiv.org/abs/0904.0468http://arxiv.org/abs/0902.4680http://arxiv.org/abs/0902.4680http://arxiv.org/abs/1001.4635http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


15/18


[13] Y. Komiyama, S. Miyazaki, H. Nakaya, H. Furusawa, andK. Takeshi, Subaru next-generation wide-field camera:HyperSuprime, in Society of Photo-Optical InstrumentationEngineers, A. F. M. Moorwood and M. Iye, Eds., vol. 5492 ofProceedings of SPIE, pp. 525532, September 2004.

[14] https://www.darkenergysurvey.org/.

[15] http://pan-starrs.ifa.hawaii.edu/public/.

[16] http://www.lsst.org/.

[17] J. Diemand, B. Moore, and J. Stadel, Earth-mass dark-matter haloes as the first structures in the early Universe,

Nature, vol. 433, no. 7024, pp. 389391, 2005.

[18] J. McBride, O. Fakhouri, and C.-P. Ma, Mass accretion ratesand histories of dark matter haloes, Monthly Notices of theRoyal Astronomical Society, vol. 398, no. 4, pp. 18581868,2009.

[19] B. Ryden, Introduction to Cosmology, Addison Wesley, SanFrancisco, Calif, USA, 2003.

[20] J. A. Peacock, Cosmological Physics, Cambridge UniversityPress, Cambridge, UK, 1999.

[21] T. Padmanabhan, Theoretical Astrophysics: Volume 3, Galaxies

and Cosmology, Cambridge University Press, Cambridge, UK,2002.

[22] P. J. E. Peebles, The Large-Scale Structure of the Universe,Princeton University Press, Princeton, NJ, USA, 1980.

[23] A. R. Zentner, The excursion set theory of halo massfunctions, halo clustering, and halo growth, International

Journal of Modern Physics D, vol. 16, no. 5, pp. 763815, 2007.

[24] A. J. S. Hamilton, Formulae for growth factors in expandinguniverses containing matter and a cosmological constant,

Monthly Notices of the Royal Astronomical Society, vol. 322,no. 2, pp. 419425, 2001.

[25] J. E. Gunn and J. R. Gott III, On the infall of matter intoclusters of galaxies and some effects on their evolution,The

Astrophysical Journal, vol. 176, p. 1, 1972.

[26] C. Lacey and S. Cole, Merger rates in hierarchical models ofgalaxy formation,Monthly Notices of the Royal AstronomicalSociety, vol. 262, pp. 627649, 1993.

[27] W. H. Press andP. Schechter, Formation of galaxies andclus-ters of galaxies by self-similar gravitational condensation,The Astrophysical Journal, vol. 187, pp. 425438, 1974.

[28] J. R. Bond, S. Cole, G. Efstathiou, and N. Kaiser, Excursionset mass functions for hierarchical Gaussian fluctuations,

Astrophysical Journal, vol. 379, no. 2, pp. 440460, 1991.

[29] B. E. Robertson, A. V. Kravtsov, J. Tinker, and A. R. Zentner,Collapse barriers and halo abundance: testingthe excursionset ansatz,The Astrophysical Journal, vol. 696, pp. 636652,2009.

[30] L. L. Watkins, N. W. Evans, and J. H. An, The masses of the

Milky Way and Andromeda galaxies,Monthly Notices of theRoyal Astronomical Society, vol. 406, no. 1, pp. 264278, 2010.

[31] Y.-S. Li and S. D. M. White, Masses for the local group andthe Milky Way, Monthly Notices of the Royal AstronomicalSociety, vol. 384, no. 4, pp. 14591468, 2008.

[32] E. Hayashi and S. D. M. White, Understanding the halo-mass and galaxy-mass cross-correlation functions, Monthly

Notices of the Royal Astronomical Society, vol. 388, pp. 214,2008.

[33] A. Cooray and R. Sheth, Halo models of large scalestructure,Physics Report, vol. 372, no. 1, pp. 1129, 2002.

[34] J. A. Fillmore and P. Goldreich, Self-similar gravitationalcollapse in an expanding universe,The Astrophysical Journal,vol. 281, pp. 18, 1984.

[35] E. Bertschinger, Self-similar secondary infall and accretionin an Einstein-de Sitter universe, The Astrophysical JournalSupplement, vol. 58, pp. 3965, 1985.

[36] P. J. Quini, J. K. Salmon, and W. H. Zurek, Primordialdensity fluctuations and the structure of galactic haloes,

Nature, vol. 322, no. 6077, pp. 329335, 1986.

[37] C. S. Frenk, S. D. M. White, M. Davis, and G.Efstathiou, Theformation of dark halos in a universe dominated by colddark matter, The Astrophysical Journal, vol. 327, pp. 507525,1988.

[38] J. Dubinski and R. G. Carlberg, The structure of cold darkmatter halos,Astrophysical Journal, vol. 378, no. 2, pp. 496503, 1991.

[39] M. M. Crone, A. E. Evrard, andD. O.Richstone, The cosmo-logical dependence of cluster density profiles, Astrophysical

Journal, vol. 434, no. 2, pp. 402416, 1994.

[40] J. F. Navarro, C. S. Frenk, and S. D. M. White, The structureof cold dark matter halos,Astrophysical Journal, vol. 462, no.2, pp. 563575, 1996.

[41] J. F. Navarro, C. S. Frenk, and S. D. M. White, A universal

density profile from hierarchical clustering, AstrophysicalJournal, vol. 490, no. 2, pp. 493508, 1997.

[42] J. F. Navarro, A. Ludlow, V. Springel et al., The diversityand similarity of simulated cold dark matter haloes,Monthly

Notices of the Royal Astronomical Society, vol. 402, no. 1, pp.2134, 2010.

[43] J. E. Taylor and J. F. Navarro, The phase-space densityprofiles of cold dark matter halos,Astrophysical Journal, vol.563, no. 2, pp. 483488, 2001.

[44] F. C. van den Bosch, The universal mass accretion historyof cold dark matter haloes, Monthly Notices of the Royal

Astronomical Society, vol. 331, no. 1, pp. 98110, 2002.

[45] R. H. Wechsler, J. S. Bullock, J. R. Primack, A. V. Kravtsov,and A. Dekel, Concentrations of dark halos from their

assembly histories,Astrophysical Journal, vol. 568, no. 1, pp.5270, 2002.

[46] A. Tasitsiomi, A. V. Kravtsov, S. Gottlober, and A. A. Klypin,Density profiles ofCDM clusters,Astrophysical Journal,vol. 607, no. 1, pp. 125139, 2004.

[47] D. H. Zhao, H. J. Mo, Y. P. Jing, and G. Borner, The growthand structure of dark matter haloes,Monthly Notices of theRoyal Astronomical Society, vol. 339, no. 1, pp. 1224, 2003.

[48] Y. Lu, H. J. Mo, N. Katz, and M. D. Weinberg, On the originof cold dark matter halo density profiles,Monthly Notices ofthe Royal Astronomical Society, vol. 368, no. 4, pp. 19311940,2006.

[49] Y. P. Jing and Y. Suto, Triaxial modeling of halo density pro-files with high-resolution N-body simulations,Astrophysical

Journal, vol. 574, no. 2, pp. 538553, 2002.[50] B. Allgood, R. A. Flores, J. R. Primack et al., The shape of

dark matter haloes: dependence on mass, redshift, radius andformation,Monthly Notices of the Royal Astronomical Society,vol. 367, no. 4, pp. 17811796, 2006.

[51] P. Bett, V. Eke, C. S. Frenk, A. Jenkins, J. Helly, and J. Navarro,The spin and shape of dark matter haloes in the Millenniumsimulation of a cold dark matter universe,Monthly Noticesof the Royal Astronomical Society, vol. 376, no. 1, pp. 215232,2007.

[52] E. Hayashi, J. F. Navarro, and V. Springel, The shape of thegravitational potential in cold dark matter haloes, Monthly

https://www.darkenergysurvey.org/https://www.darkenergysurvey.org/http://pan-starrs.ifa.hawaii.edu/public/http://pan-starrs.ifa.hawaii.edu/public/http://www.lsst.org/http://www.lsst.org/http://www.lsst.org/http://pan-starrs.ifa.hawaii.edu/public/https://www.darkenergysurvey.org/


16/18


[53] B. Moore, S. Kazantzidis, J. Diemand, and J. Stadel, Theorigin and tidal evolution of cuspy triaxial haloes, Monthly


[54] R. Mandelbaum, C. M. Hirata, T. Broderick, U. Seljak,and J. Brinkmann, Ellipticity of dark matter haloes withgalaxy-galaxy weak lensing, Monthly Notices of the Royal

Astronomical Society, vol. 370, no. 2, pp. 10081024, 2006.[55] T. G. Brainerd, Anisotropic distribution of SDSS satellite

galaxies: planar (not polar) alignment,Astrophysical Journal,vol. 628, no. 2, pp. L101L104, 2005.

[56] J. S. Bullock, A. Dekel, T. S. Kolatt et al., A universal angularmomentum profile for galactic halos,Astrophysical Journal,vol. 555, no. 1, pp. 240257, 2001.

[57] J. E. Taylor and A. Babul, The dynamics of sinking satellitesaround disk galaxies: a poor mans alternative to high-resolution numerical simulations,Astrophysical Journal, vol.559, no. 2, pp. 716735, 2001.

[58] J. E. Taylor and A. Babul, The evolution of substructurein galaxy, group and cluster haloesI. Basic dynamics,

Monthly Notices of the Royal Astronomical Society, vol. 348,

no. 3, pp. 811830, 2004.[59] F. C. van den Bosch, G. Tormen, and C. Giocoli, The

mass function and average mass-loss rate of dark mattersubhaloes,Monthly Notices of the Royal Astronomical Society,vol. 359, no. 3, pp. 10291040, 2005.

[60] L. Gao, C. S. Frenk, M. Boylan-Kolchin, A. Jenkins, V.Springel, and S. D. M. White, The statistics of the sub-halo abundance of dark matter haloes,http://arxiv.org/abs/1006.2882.

[61] C. Giocoli, G. Tormen, R. K. Sheth, and F. C. van den Bosch,The substructure hierarchy in dark matter haloes, Monthly


[62] B. Moore, Simulating clusters of galaxies: a brief historyof N and overmerging, in Constructing the Universe withClusters of Galaxies, 2000.

[63] G. Tormen, A. Diaferio, andD. Syer, Survival of substructurewithin dark matter haloes, Monthly Notices of the Royal

Astronomical Society, vol. 299, no. 3, pp. 728742, 1998.[64] S. Ghigna, B. Moore, F. Governato, G. Lake, T. Quinn, and J.

Stadel, Dark matter haloes within clusters,Monthly Noticesof the Royal Astronomical Society, vol. 300, no. 1, pp. 146162,1998.

[65] A. Klypin, S. Gottlober, A. V. Kravtsov, and A. M. Khokhlov,Galaxies in N-body simulations: overcoming the overmerg-ing problem,Astrophysical Journal, vol. 516, no. 2, pp. 530551, 1999.

[66] A. Klypin, A. V. Kravtsov, O. Valenzuela, and F. Prada,Where are the missing galactic

dark matter halos from the inside out.pdf

Documents