Our aim is to try to understand dark matter. Our approach is data driven. Galaxies are our laboratory. We begin with a study of spiral galaxy rotation curves [1] [2]. Next, we study the formation of dark matter halos starting from primordial linear density-velocity-gravitational perturbations [3]. In the present work we extend the previous dynamical studies to include baryons. In particular, we study the formation of the first generation of galaxies and stars.

The root-mean-square velocity of non-relativistic dark matter particles depends on the expansion parameter a of the universe as follows:

v h rms ( a ) = v h rms ( 1 ) a . (1)

The cold dark matter ΛCDM cosmology assumes that dark matter particles have negligible velocity dispresion, i.e. v h rms ( 1 ) ≈ 0 . The warm dark matter ΛWDM scenario adds the single parameter v h rms ( 1 ) to ΛCDM. The dark matter particle velocity dispersion results in a cut-off of the power spectrum of density fluctuations in the early universe due to free-streaming. The free-streaming cut-off wavenumber k fs ( t eq ) , at the time of equal densities of radiation and matter, can be calculated from v h rms ( 1 ) [4]. From the formation of first galaxies and reionization, we are able to set limits on v h rms ( 1 ) and k fs ( t eq ) .

In the following sections we discuss galaxy hydro-dynamical equations, their numerical solutions, the formation of first galaxies (that in ΛWDM depend on both the free-streaming power spectrum cut-off factor τ 2 ( k ) , and on the dark matter velocity dispersion), comment on the formation of first galaxies and stars in nodes, filaments and sheets, and estimate limits on the warm dark matter velocity dispersion based on the formation of first galaxies and reionization. We conclude with a brief comparison with previous measurements, and mention discrepancies with dark matter limits in the literature.

We use the standard notation in cosmology, and the cosmological parameters as in [5]. Throughout, the subscript h stands for the dark matter halo, and b stands for baryons.

The formation of the galactic halo can be illustrated by integrating numerically Newton’s equation

∇ ⋅ g = − 4 π G ( ρ h + ρ b ) , (2)

the continuity equations

∂ ρ h ∂ t = − ∇ ⋅ ( v h ρ h ) , (3)

∂ ρ b ∂ t = − ∇ ⋅ ( v b ρ b ) , (4)

and Euler’s equations

d v h d t = ∂ v h ∂ t + ( v h ⋅ ∇ ) v h = ( 1 − κ h ( t ) ) g − 1 ρ h ∇ ( 〈 v r h 2 〉 ρ h ) . (5)

d v b d t = ∂ v b ∂ t + ( v b ⋅ ∇ ) v b = ( 1 − κ b ( t ) ) g − 1 ρ b ∇ ( 〈 v r b 2 〉 ρ b ) . (6)

g ( x ) is the gravitation field, x is the proper coordinate vector, ρ h ( x ) and ρ b ( x ) are the mass densities, v h ( x ) and v b ( x ) are the velocities, and 〈 v r h 2 〉 and 〈 v r b 2 〉 are the radial (1-dimensional) velocity dispersions, of dark matter and baryons (hydrogen and helium), respectively. Equations (5) and (6) express the conservation of momentum. We use proper, not comoving, spatial coordinates. The functions κ h ( t ) and κ b ( t ) are included to describe spiral galaxy rotation [1] [2] [6]. We will discuss these functions below. For warm dark matter we supplement the preceding equations with the condition of adiabatic expansion in the core of the galaxy:

〈 v r h 2 〉 = v h rms ( 1 ) 3 ( ρ h ( r min , t ) Ω c ρ crit ) 1 / 3 . (7)

This condition is studied in [3]. Similarly, while the hydrogen and helium gas remains adiabatic, i.e. until excitations, radiation and shocks become significant, we require

〈 v r b 2 〉 = v b rms ( 1 ) 3 ( ρ b ( r min , t ) Ω b ρ crit ) 1 / 3 (8)

in the core of the galaxy. For baryons we take v b rms ( 1 ) = 21 m/s, corresponding to hydrogen decoupling from photons at z ≈ 150 [7]. We have made several data driven approximations [1] [3]: 〈 v r h 2 〉 , 〈 v r b 2 〉 , κ h ( t ) , and κ b ( t ) are taken to be independent of r, and hence, the adiabatic constraints (7) and (8) are applied, at each time step in the numerical integration, only at r min .

Static solutions of Equations (2), (5) and (6) are studied in [1] and [3]. The static solutions depend on the variables 〈 v ′ r h 2 〉 ≡ 〈 v r h 2 〉 / ( 1 − κ h ) , and 〈 v ′ r b 2 〉 ≡ 〈 v r b 2 〉 / ( 1 − κ b ) . For elliptical galaxies we may take κ h ≈ 0 and κ b ≈ 0 . For spiral galaxies we estimate κ h ≈ 0 and κ b ≈ 0.98 [1]. We solve the static equations by numerical integration, starting from the radius r min of the first measured rotation velocity, up to r max of the last one. To begin these integrations we need four boundary conditions: the radial velocity dispersions 〈 v r h 2 〉 and 〈 v r b 2 〉 , and the densities ρ h ( r min ) and ρ b ( r min ) , of dark matter and baryons, respectively. We vary these four parameters to minimize a χ 2 between the calculated and measured

rotation curves. Good fits to the data are obtained assuming 〈 v r h 2 〉 and 〈 v r b 2 〉 are independent of r. We predict that the adiabatic invariant 〈 v r h 2 〉 / ρ h ( r → 0 ) 1 / 3

is of cosmological origin, and hence has the same value for all relaxed spiral galaxies [3]. From measurements of approximately 60 spiral galaxies, we obtain the adiabatic invariant

v h rms ( 1 ) ≡ 3 〈 v r h 2 〉 ( Ω c ρ crit ρ h ( r → 0 ) ) 1 / 3 = 0.79 ± 0.33   ( tot )   km / s , (9)

at 68% confidence [8]. The conclusion is that dark matter is (arguably) warm. The velocity dispersion (9) corresponds to a warm dark matter power spectrum cut-off wavenumber, due to free-streaming, k fs ( t eq ) = 1.03 − 0.30 + 0.74   Mpc − 1 , at equality of the densities of radiation and matter [4] [8]. In summary, relaxed spiral galaxy density runs ρ h ( r ) and ρ b ( r ) are described by only three independent parameters. Note that v b rms ( 1 ) ≪ v h rms ( 1 ) , so baryons act like cold dark matter.

The static equations have been extended to degenerate fermion or boson dark matter [1] [6]. The onset of degeneracy brings disagreement with spiral galaxy rotation curves, so lower limits are set on the dark matter particle mass m h [1] [6].

Dynamic, spherically symmetric, solutions to Equations (2) to (8) are studied in [3] for the case of no baryons. In the cold dark matter scenario the dark matter particles fall to the origin, overshoot, splash-back, overshoot again, and relaxation phenomena are invoked to “virialize the galaxy” [9]. In the warm dark matter scenario, the dark matter halo may form adiabatically, due to the feedback from (7) that halts the halo collapse, reaching a relaxed density run ρ h ( r ) with the following asymptotes: the core density ρ c extends out to the core radius [3]

r c ≡ [ 〈 v ′ r h 2 〉 2 π G ρ c ] 1 / 2 , (10)

and thereafter ρ h ( r ) and M h ( < r ) approach

ρ h ( r ) = 〈 v ′ r h 2 〉 2 π G r 2 ,   M h ( < r ) = 2 〈 v ′ r h 2 〉 G r . (11)

The only independent parameter is 〈 v ′ r h 2 〉 , since ρ c can then be obtained from the adiabatic invariant (7). 〈 v ′ r h 2 〉 is obtained by numerical integration

starting from the initial linear density-velocity-gravitational perturbation. The radius of the warm dark matter halo keeps growing with constant speed 3 〈 v ′ r h 2 〉 , so the halo mass keeps increasing linearly with time, and no well defined static “virialized” halo mass can be defined. Note that the core is (arguably) evidence for warm dark matter. In the present work we extend these studies to include baryons.

As an example, let us consider the evolution of an initial spherically symmetric Gaussian perturbation. The initial densities, at expansion parameter a i , are ρ h i ( r ) = ρ i ( r ) Ω c / ( Ω c + Ω b ) , and ρ b i ( r ) = ρ i ( r ) Ω b / ( Ω c + Ω b ) , with

ρ i ( r ) = ρ ¯ i [ 1 + δ exp ( − r 2 r i 2 ) ] . (12)

The initial velocities are v r h i ( r ) = v r b i ( r ) = H ( a i ) r [ 1 − ( ρ ¯ < r − ρ 0 ) / ( 3 ρ 0 ) ] , corresponding to a growing mode. The average ρ ¯ < r is taken with weight r 2 . The initial parameters are z i = 135 , ρ ¯ i = 0.1 M ⊙   pc − 3 , δ = 0.3 , and r i = 4   kpc . The linear dark matter and baryon masses are M h = 2.3 × 10 10 M ⊙ and M b = 4.2 × 10 9 M ⊙ , respectively. Note that we have assumed that baryons track dark matter while perturbations are linear for the masses of interest (much greater than M J b W to be defined below). We assume cold dark matter. We assume no angular momentum, so set κ h ( t ) = κ b ( t ) = 0 . Numerical integration of Equations (2) to (8) with these initial conditions are shown in Figure 1. We note that ρ h ( r , t ) remains equal to Ω c ρ b ( r , t ) / Ω b , i.e. baryons act as cold dark matter. Our code crashes when ρ h ( r ) and ρ b ( r ) diverge at r → 0 . The equations are no longer valid beyond the collapse of dark matter or baryons

at the origin.

For comparison, in Figure 2 we present the formation of a similar galaxy, but with warm dark matter with v h rms ( 1 ) = 790   m / s , corresponding to k fs = 1   Mpc − 1 . The result is qualitatively different. The proportionality between ρ h ( r , t ) and ρ b ( r , t ) is lost. We note that the formation of the warm dark matter halo is delayed with respect to the corresponding cold dark matter case of Figure 1. The warm dark matter halo approaches a relaxed core, while baryons approach a “mini” core (not resolved by this simulation), and baryon radiation becomes important.

The collapse of baryons is halted if baryons have an angular momentum. To investigate this phenomenon, we set κ h ( t ) = 0 , and begin the numerical integration with κ b ( t i ) ≈ 0 , e.g. κ b ( t i ) = 0.0001 , and gradually increase κ b ( t ) to 1 so as to conserve the baryon angular momentum. Let v rot be the velocity of a

test particle in a circular orbit of radius r. Then g = − v rot 2 / r . Let v b rot be the rotation velocity of baryons at radius r. Then κ b ( t ) = v b rot 2 / v rot 2 . As a working approximation we take κ b ( t ) independent of r. The baryon angular momentum is

L b = ∫ r v b rot ρ b ( r ) d V . (13)

The rotating baryons approach a disk, so we approximate

L b ∝ ∫ 0 ∞ r 5 / 2 − κ b ( t ) g ( r ) ρ b ( r ) d r . (14)

This expression becomes well defined, i.e. finite, if ρ b ( r ) decreases faster than r − 3 . This becomes the case due to baryon energy loss by radiation. An example as in Figure 2, but with baryons with an initial angular momentum, is presented in Figure 3. We note that both warm dark matter and baryons reach relaxed

cores. In this way we are able to follow the formation of both relaxed elliptical and spiral galaxies starting from initial linear perturbations.

Keeping the spherical symmetry, we are also able to study the collapse of baryons in warm dark matter spherical sheets, see Figure 4. Note how baryons shock at collapse.

In Figure 5 we illustrate that no self-gravitating structure is formed if the linear mass M of the perturbation is below a threshold.

Let us define “first galaxy”. The definition needs to be of a probabilistic nature,

since, with a sufficiently large volume, it is possible to find a galaxy before the “first galaxy”. To this end, let us recall the variance of relative warm dark matter density perturbations δ ( x ) ≡ ( ρ ( x ) − ρ ¯ ) / ρ ¯ on the linear mass scale M, at redshift z: [7]

σ 2 ( M , z , k fs ) = f 2 ( 2 π ) 3 ( 1 + z ) 2 ∫ 0 ∞ 4 π k 2 d k P ( k ) τ 2 ( k ) W 2 ( k ) . (15)

f is a correction due to the accelerated expansion of the universe:

f = 1 , 1.252 , 1.269 for redshift z = 0 , 2 , 11 respectively. k is the comoving wavenumber. W ( k ) is a window function that defines the total (dark matter plus baryon) linear mass M. P ( k ) is the ΛCDM linear relative density power spectrum at z = 0 (we use the analytic expression in [7]). τ 2 ( k ) is a cut-off factor due to warm dark matter free-streaming. We take τ 2 ( k ) ≈ exp ( − k 2 / k fs 2 ) [4], where k fs is the warm dark matter free-streaming comoving cut-off wavenumber. Note our definition of k fs : τ 2 ( k fs ) ≡ e − 1 . P ′ ( k ) ≡ f 2 P ( k ) τ 2 ( k ) / ( 1 + z ) 2 is the

proper power spectrum at redshift z. For each k fs we adjust the amplitude of P ( k ) so that σ 8 = 0.811 [5], calculated with a top-hat window function of radius 8/h Mpc. For all other calculations we use the well behaved Gaussian window function (see [8] for details):

W ( k ) = exp ( − k 2 2 k 0 2 ) ,   M = 4 3 π ( 1.555 k 0 ) 3 ρ ¯ m . (16)

In the ΛCDM scenario, when δ ( x ) reaches 1.686 in the linear approximation, the true value of δ ( x ) diverges (for a spherically symmetric perturbation), and a galaxy forms. This is also true for the ΛWDM scenario for sufficiently large M so that velocity dispersion is negligible, i.e. above the “dots” in Figures 6-8 below. At z = 8 , galaxies with log 10 ( M ∗ / M ⊙ ) = 9.05 (9.55) (see distributions in [6]) correspond to fluctuations 1.686 / σ ( M ∗ ( Ω c + Ω b ) / Ω b , z , ∞ ) = 3.0 (3.4) standard deviations in the ΛCDM scenario if recently formed. At z = 6 , galaxies with log 10 ( M ∗ / M ⊙ ) = 9.05 (10.7) correspond to fluctuations of 2.3 (3.7) standard deviations. Note that the redshift at half reionization is z = 7.7 ± 0.7 [5]. Guided by these data, we define “first galaxy” (or “first star”, as the case may be) as a galaxy that forms from a 4 standard deviation fluctuation.

In the present article we consider the ΛCDM scenario, and several ΛWDM scenarios: ΛWDM₁ with v h rms ( 1 ) = 0.79   km / s , and the corresponding k fs ( t eq ) = 1.0   Mpc − 1 ; ΛWDM₂ with v h rms ( 1 ) = 0.395   km / s and k fs ( t eq ) = 2.0   Mpc − 1 ; and ΛWDM₄ with v h rms ( 1 ) = 0.2   km / s and k fs ( t eq ) = 4.0   Mpc − 1 .

Let us understand Figure 6. The horizontal axis is the redshift z. Time t ∝ ( 1 + z ) − 3 / 2 advances towards the left. The vertical axis is the linear baryon

mass M b (or the linear dark matter mass M h times Ω b / Ω c for the dark matter-only lines with short-dash style). Linear masses are well defined, since a linear, i.e. a relatively small, density perturbation has dimensions that scale as the expansion parameter a , and a density that scales as a − 3 , so the linear mass of a fluctuation is independent of a . The nearly vertical lines are defined by σ ( M b ( Ω c + Ω b ) / Ω b , z , k fs ) = 1.686 / 4 with k fs = 1,2,4, ∞   Mpc − 1 for the ΛWDM₁, ΛWDM₂, ΛWDM₄, and ΛCDM scenarios, respectively.

For completeness, we also present in Figure 6 several proper Jeans masses (with nearly horizontal lines). The warm dark matter Jeans mass M J h , for the case of no baryons, derived from the linear approximations of (2), (3) and (5), is

M J h = 4 3 π ( 1.555 k J h ) 3 ρ ¯ h , (17)

k J h ≡ 2 π λ J h ,   λ J h = 〈 v ′ r h 2 〉 π G ρ ¯ h , (18)

with all variables evaluated at redshift z. The short-dash lines correspond to the ΛWDM₁ and ΛWDM₂ scenarios. For the ΛCDM scenario, M J h ≈ 0 . Linear dark matter perturbations with mass greater than M J h grow due to self-gravitation, while perturbations with mass less than M J h are attenuated.

The curve marked M J b is the Jeans mass of baryons, for the case of no dark matter. It is calculated as above, but with the proper Jeans length

λ J b = v s 2 π G ρ ¯ b ,   v s = 5 k T b 3 μ m p . (19)

v s is the sound velocity in a gas of hydrogen and helium with mean molecular weight μ = 1.22 [7]. The baryon temperature T b ( a ) is calculated assuming baryons decouple from photons at z ≈ 150 [7], and thereafter T b ( a ) ∝ a − 2 . Thus, we take T b ( 1 ) = 2.7 / 151   K . Assuming no dark matter, linear baryon perturbations with mass greater than M J b grow due to self-gravitation, while perturbations with mass less than M J b propagate as sound waves.

The curve marked M J b W is a sort of Jeans mass for the ΛCDM scenario. In this scenario, linear perturbations of arbitrarily small dark matter mass can collapse, but with a reduced baryon fraction due to baryon pressure, unless the linear dark matter plus baryon perturbation mass is greater than [7]

M J b W = ( π G ) 3 / 2 v s 3 ρ ¯ m 1 / 2 . (20)

The line corresponding to k fs in Figure 6 is accurate above the dot, i.e. above the velocity dispersion cut-off. Below the dot, warm dark matter velocity dispersion becomes important, and lowers the redshift of baryon collapse, and finally prevents the formation of self-gravitating structures, see Figure 5. To obtain the “dot” we proceed as follows. We choose a δ for the simulations, e.g. δ = 0.3 . We choose a mass M b . Then, from Figure 6 we read off the redshift z C at which the galaxy forms in the ΛCDM scenario. Then, from δ / ( ( 1 + z C ) a i ) = 1.686 we obtain the initial expansion parameter a i , the initial mean matter density ρ ¯ i , and the initial radius r i , for the simulation. We then run the simulation with v h rms ( 1 ) = 0 for ΛCDM, and the corresponding v h rms ( 1 ) for the ΛWDM scenario, and obtain the difference in redshift Δ z of galaxy formation in these two simulations. We place the “dot” at the mass M b that obtains Δ z = 1 . Above the dot, the effect of the velocity dispersion can be neglected. Below the dot, velocity dispersion lowers the redshift of galaxy formation, if the galaxy forms at all (see Figure 5 above), and the curve in Figure 6 becomes invalid. The initial expansion parameter a i of the simulation is not critical.

Warm dark matter perturbations acquire two low mass cut-offs: the free-streaming cut-off due to the power spectrum free-streaming cut-off factor τ 2 ( k ) , and the velocity dispersion cut-off, i.e. the dots in Figures 6-8. The Press-Schechter linear galaxy mass distribution [10] includes the free-streaming cut-off, but does not include the velocity dispersion cut-off. Care should be taken not to apply the Press-Schechter relation to perturbations below the velocity dispersion cut-off mass.

We are now in a position to understand Figure 6. Considering only spherically symmetric perturbations, and neglecting non-linear regeneration of small scale structure, we obtain the following results. For the ΛWDM₁, ΛWDM₂ and ΛWDM₄ scenarios, “first galaxies” and stars form approximately at the red, blue and green dots, for k fs ( t eq ) = 1   Mpc − 1 ,2   Mpc − 1 and 4 Mpc⁻¹, at z ≈ 5.5,7.6 and 10.1, and M b ≈ 3 × 10 9 M ⊙ ,3 × 10 8 M ⊙ and 1 × 10 7 M ⊙ , respectively. For the ΛCDM scenario, “first halos” (with 4 σ fluctuations) with a full complement of baryons, i.e. first galaxies and stars, form approximately when the two black lines meet at z ≈ 18 and M b ≈ 5 × 10 3 M ⊙ . If we only consider spherically symmetric perturbations, and neglect non-linear regeneration of small scale structure, we conclude that the warmest dark matter consistent with reionization has v h rms ( 1 ) ≈ 270   m / s , k fs ≈ 3   Mpc − 1 , and first galaxies have masses of order 10 8 M ⊙ (these may break up).

These results are approximate because we have only considered spherically symmetric perturbations, and have not included non-linear regeneration of small scale structure. An ellipsoidal perturbation collapses first along its shortest axis (obtaining a Zeldovich pancake or sheet), then along the intermediate axis (obtaining a filament), and finally along the longest axis (obtaining a node). The large scale structure of the universe resembles a honeycomb with voids, surrounded by sheets, that meet at filaments, that meet at spheroidal nodes, that are punctuated by galaxies. Baryons collapse in nodes, filaments [11], and perhaps even in sheets (see Section 3), fragmenting into multiple galaxies and stars with a wide distribution of masses.

At the time t eq of equal radiation and matter densities, the warm dark matter cut-off function has the approximate form τ 2 ( k ) = exp ( − k 2 / k fs 2 ( t eq ) ) with k fs ( t eq ) given in [4]. After t eq the Jeans masses M J h and M J b decreas as a − 3 / 2 , and non-linear regeneration of small scale structure occurs as soon as relative density fluctuations approach unity. Warm dark matter only simulations obtain the cut-off factor τ 2 ( k ) corresponding to non-linear regeneration of small scale structure, see [12]. With the non-linear τ 2 ( k ) from [12], we obtain Figure 7. For k fs ( t eq ) ≈ 1   Mpc − 1 , 2 Mpc⁻¹, or 4 Mpc⁻¹ “first galaxies” and stars form approximately at the red, blue, or green dot, at z ≈ 8.0 , 8.6, or 9.3, and M b ≈ 1 × 10 9 M ⊙ , 3 × 10 8 M ⊙ , or 3 × 10 7 M ⊙ , respectively. Baryons collapse before warm dark matter, so including baryons in the simulations should obtain a larger non-linear regeneration effect.

Figure 8 is the same as Figure 7, but for 3 standard deviation fluctuations instead of 4. For k fs ( t eq ) ≈ 1   Mpc − 1 , 2 Mpc⁻¹, or 4 Mpc⁻¹, 3 σ fluctuations form galaxies approximately at the red, blue, or green dot, at z ≈ 6.0 , 6.6, or 7.1, and M b ≈ 1 × 10 9 M ⊙ , 2 × 10 8 M ⊙ , or 2 × 10 7 M ⊙ , respectively.

We conclude that the warmest dark matter consistent with reionization has v h rms ≈ 395   m / s , and k fs ( t eq ) ≈ 2   Mpc − 1 . This tentative result needs a correction from 3D simulations that include the formation of galaxies in filaments, and possibly in sheets, and both warm dark matter (free-streaming and velocity

dispersion) and baryons.

The conditions for the formation of a star are that the density becomes dominated by baryons, and that these baryons radiate effectively. In this case, as the baryons (hydrogen and helium) radiate, their potential energy drops, and their kinetic energy, i.e. their temperature, increases due to gravity. This increase in temperature increases the radiation rate, and the process is run-away, until halted by baryon angular momentum conservation, or by thermonuclear reactions in the core that burn hydrogen into helium at ≈1.5 × 10⁷ K (the p-p chain reaction) and a star is born on the main sequence. An example is presented in Figure 9.

For warm dark matter, baryons do dominate the core (if angular momentum is sufficiently small), see Figure 2. For cold dark matter, baryons do not initially dominate the core, see Figure 1, so a first stage of radiation results in cooling until baryons dominate the core. For the formation of first stars in the ΛCDM scenario I refer the reader to text books [13] [14] [15]. In the following, we focus our attention on warm dark matter.

Let us consider baryon energy loss by radiation. Channels considered for first generation (“Population III”) stars are the molecular hydrogen H₂ channels (the H⁻ channel, and the H 2 + channel), and the atomic hydrogen channels, i.e. Lyman-α transitions [16] [17]. The atomic hydrogen channels open up and dominate at T b ≳ 10 4     K , while H₂ channels dominate at lower temperatures. The

dominant channel of interest at low temperature is the H⁻ channel:

H + e − → H − + γ

H − + H → H 2 + e − . (21)

As an example, consider Figure 2. The resolution of our simulation allows us to follow the core down to r min = 0.7   kpc . At this radius the baryon density raises to 0.024 M ⊙ / pc 3 , corresponding to a number density n = 1   cm − 3 , T b = 447   K , and an energy 0.058 eV per hydrogen atom. The corresponding cooling function is ≈3 × 10⁻²⁶ erg cm³∙s⁻¹ [16], and the inverse of the cooling rate is 3 × 10¹² s. The age of the universe at the formation of this halo, is 7 × 10¹⁵ s. So even the H₂ cooling channel is effective. This calculation is conservative since a rough estimate of the baryon core temperature, using (8), and (11) at the pivot point, exceeds the atomic hydrogen channel threshold T b ≈ 10 4     K , even without including heating due to shocks at r ≈ 0 .

As another example, consider the baryon density peak in the sheet of Figure 4. This peak has a density ρ b = 0.0023 M ⊙ / pc 3 , corresponding to n = 0.1   cm − 3 , T b = 94   K , and a hydrogen atom kinetic energy 0.012 eV (these values are limited by the resolution of the simulation). The cooling function is ≈7 × 10⁻²⁹ erg cm³∙s⁻¹ [16]. The corresponding inverse of the cooling rate is 3 × 10¹⁴ s. The age of the universe at the formation of this sheet, is 3 × 10¹⁵ s. So, even for a sheet with hydrogen at this low temperature, the H₂ cooling channel may be marginally effective (a full detailed simulation is needed for confirmation). We have not included the shock observed in Figure 4: its velocity corresponds to Δ T b = 2.7 × 10 4     K .

A realistic, high resolution, and detailed simulation concludes that stars do indeed form in filaments [11].

The redshift at half reionization from the Planck mission is 7.7 ± 0.7 [5]. From the rest-frame ultra-violet luminosity distribution of galaxies, most of the reionization occurs in the redshift range 8 to 5.7, with a possible slow start at about z ≈ 10 , and with half-reionization at approximately 7.3 [18]. Star formation begins at about z ≈ 10 [18] [19]. So the onset of reionization lies in the approximate redshift interval from 10.0 to 8.2. This implies that 200   m / s ≲ v h rms ( 1 ) ≲ 395   m / s , corresponding to 4   Mpc − 1 ≳ k fs ( t eq ) ≳ 2   Mpc − 1 , see the green and blue dots in Figure 7 and Figure 8. These estimates include the non-linear regeneration of small scale structure obtained from dark matter only simulations [12], i.e. excluding the contribution from baryons that should be significant, and does not include galaxy formation in filaments and sheets.

By numerical integration of hydro-dynamical equations, we are able to obtain the density runs ρ h ( r ) and ρ b ( r ) of relaxed elliptical and spiral galaxies starting from primordial linear density-velocity-gravitational perturbations. The dark matter halo acquires a core due to the warm dark matter velocity dispersion. Baryons dominate the core density if angular momentum is below a threshold. These baryons radiate, and collapse, and their temperature increases, due to gravitation. The collapse is run-away until stopped by baryon angular momentum conservation, or by thermonuclear reactions that give birth to stars.

From Figures 6-8 we conclude that the onset of reionization in the ΛWDM scenarios is delayed with respect to ΛCDM. For spherically symmetric perturbations, and neglecting non-linear regeneration of small scale structure, we find that the warmest dark matter consistent with reionization has k fs ( t eq ) ≈ 3   Mpc − 1 , corresponding to v h rms ( 1 ) ≈ 270   m / s , with first galaxies with masses M b of order 10 8 M ⊙ (that may fragment), see Figure 6. However, taking into account the non-linear regeneration of small scale structure (obtained from warm dark matter only simulations [12]), we find that the warmest dark matter consistent with reionization has approximately k fs ( t eq ) ≈ 2   Mpc − 1 , corresponding to v h rms ( 1 ) ≈ 395   m / s , see Figure 7 and Figure 8.

From the measured redshift at half-reionization, and the rest frame ultra-violet luminosity distributions of galaxies, that determine the history of the star formation rate, we estimate 200   m / s ≲ v h rms ( 1 ) ≲ 395   m / s , and 4   Mpc − 1 ≳ k fs ( t eq ) ≳ 2   Mpc − 1 . The corresponding thermal relic mass m h depends on the temperature at which dark matter decouples from the standard model sector, and on the spin of dark matter particles, see Table 1 of [8].

This preliminary study needs to be extended to non-spherically symmetric perturbations, with high resolution 3D simulations that include dark matter (both free-streaming and velocity dispersion) and baryons, including the baryon physics that obtains the observable stellar luminosities, e.g. an extension of the studies in [11].

We have made three independent measurements of the warm dark matter adiabatic invariant v h rms ( 1 ) , and hence of the dark matter temperature-to-mass ratio T h ( a ) / m h : 1) From approximately 60 spiral galaxy rotation curves [1] [2]; 2) From galaxy stellar mass distributions at z = 3 , 4.5 , 6 , 7 , and 8 [6]; and 3) The present measurement from reionization. Each of these measurements has its issues. However, taken together, their consistency (within uncertainties that are still large, and a mild tension with the present measurement), and their agreement with the no freeze-in and no freeze-out warm dark matter scenario [8], is quite remarkable. Discrepancies with several limits that can be found in the literature are discussed in [8]. See also section 4, and [20].

The author declares no conflicts of interest regarding the publication of this paper.

Hoeneisen, B. (2022) Warm Dark Matter and the Formation of First Galaxies. Journal of Modern Physics, 13, 932-948. https://doi.org/10.4236/jmp.2022.136053