betaFeedback

WorksPhysics and Astronomy

Cosmological structure formation in ultralight dark matter

Abstract

An important challenge in cosmology is to describe the formation of galaxies and structures, starting from small inhomogeneities in the early Universe, which later grew and formed the so- called dark matter halos. We consider the dark matter to be composed of ultralight bosons and simplify the general complex problem of halo formation by using the time-dependent variational analysis approach. This yields, for the first time, a fully analytical equation for the time-evolution of perturbations towards isolated haloes. We follow the evolution of the halo size and density during different stages of its formation, revealing the role of the background density, Universe expansion, and quantum pressure.

Introduction and model

While the existence of dark matter (DM) was first predicted based on astrophysical observations [1, 2], its composition and properties play an important role across various fields of physics, astrophysics, and cosmology. The widely used Λ\LambdaCDM model of cosmology assumes that the DM is a non-relativistic collisionless fluid, referred to as cold dark matter (CDM) [3]. The CDM explains many of the observed astrophysical phenomena, however encounters a few challenges, such as cusp-core tension [4, 5], which could be resolved by accounting for the microscopic nature of DM. This characterizes the ultralight dark matter (ULDM) class of models, which assume that the DM is composed of ultralight bosons in the mass range m∈[10−18,10−24]m \in[10^{-18}, 10^{-24}] eV/c2c^2 [6, 7, 8, 9]. The question of viability of the ULDM can be addressed from an astrophysical perspective based on such observations as rotation curves [10, 11], velocity dispersion [12, 13], and from a cosmological perspective by analyzing the evolution of the DM in an expanding Universe [14, 15, 16].

According to the standard Λ\LambdaCDM model of cosmology, dark matter comprises about 85% of the matter in the Universe and governs its expansion during the matter-dominated epoch, between redshifts zi≃3100z_i \simeq3100 and zf≃0.5z_f \simeq0.5 (tf=9.8×109t_f = 9.8 \times10^9 yr.) [14, 17]. During this epoch, galaxy formation takes place in the flat Einstein-de Sitter universe [14]; the scale factor evolves as a(t)=(t/tf)2/3a(t) = (t/t_f)^{2/3} to its final value, which, for convenience, we hereby set to a(tf)=1a(t_f) = 1. Both CDM and ULDM rely on the cosmological principle that assumes the Universe, and hence the matter is homogeneous and isotropic at very large scale. The spacetime metric of the homogeneous expanding Universe is given by the Friedmann-Lemaître-Robertson-Walker (FLRW) metric [14, 18], in which the bosonic ULDM obeys the Klein-Gordon equation of motion

ψ¨0+3Hψ˙0+(mc2ℏ)2ψ0=0,(1)\ddot{\psi}_0 + 3H\dot{\psi}_0 + \left(\frac{mc^2}{\hbar}\right)^2\psi_0 = 0, \tag*{(1)}

where mm denotes the DM particle mass and H=a˙/aH = \dot{a}/a is a Hubble parameter. The exact solution to Eq. (1) exhibits fast oscillations starting at t∼H−1t \sim H^{-1} with period ℏ/(mc2)\hbar/(mc^2) [15, 16]. For a typical boson mass range m=10−22−10−20m = 10^{-22} - 10^{-20} eV/c2c^2, fast oscillations start well before the matter-radiation equality and have a very small period compared to H−1H^{-1} [16]. Therefore, for the considered matter-dominated epoch, one can exclude these fast oscillations by taking the nonrelativistic limit ψ=eimc2t/ℏψ0\psi= e^{imc^2t/\hbar}\psi_0 in Eq. (1), which gives

ψ˙0+32Hψ0=0,(2)\dot{\psi}_0 + \frac{3}{2}H\psi_0 = 0, \tag*{(2)}

and agrees with the standard result in cosmology for non-relativistic matter [14, 18], implying that

∣ψ0∣2=ρ0=38πG(a˙a)2=16πGt2.(3)|\psi_0|^2 = \rho_0 = \frac{3}{8\pi G}\left(\frac{\dot{a}}{a}\right)^2 = \frac{1}{6\pi Gt^2}. \tag*{(3)}

Hence, ULDM in a non-relativistic limit and the cosmological setting has the same behavior as Λ\LambdaCDM [16] reproducing the observed distribution of matter at large scales [19].

While the cosmological principle holds on large scales, on the smaller scales the Universe shows inhomogeneities such as galaxies and clusters of galaxies [14]. To describe the formation of these structures, one considers small density perturbations on the uniform density (3), see Ref. [20], which generate small metric perturbations described by the gravitational potential Φ(r,t)\Phi(r,t). The metric for a scalar perturbation can be written in the linearized pseudo-Newtonian form [21]

ds2=(1+2Φc2)c2dt2+a2(t)(−1+2Φc2)δijdxidxj,(4)ds^2 = \left(1+\frac{2\Phi}{c^2}\right)c^2dt^2 + a^2(t)\left(-1+\frac{2\Phi}{c^2}\right)\delta_{ij}dx^i dx^j, \tag*{(4)}

where Φ/c2≪1\Phi/c^2 \ll1, see Refs. [22, 23] for further discussion of the post-Newtonian corrections. The bosonic field ψ\psi of ULDM in this spacetime obeys the system of Schrödinger-Poisson equations (SPE) in the expanding Universe

iℏ(∂∂t+32H)ψ=−ℏ22ma2∇2ψ+mΦψ,(5)i\hbar\left(\frac{\partial}{\partial t}+\frac{3}{2}H\right)\psi=-\frac{\hbar^2}{2ma^2}\nabla^2\psi+m\Phi\psi, \tag*{(5)}
ΔΦ=4πGa2(∣ψ∣2−ρ0),(6)\Delta\Phi=4\pi Ga^2\left(\lvert\psi\rvert^2-\rho_0\right), \tag*{(6)}

as has been derived and extensively analyzed in Refs. [24, 25]. These equations can be further reformulated in a fluid description using the Madelung transformation [26] of the complex field ψ=ρeiθ\psi=\sqrt{\rho}e^{i\theta} in terms of its absolute value ρ(r,t)\sqrt{\rho(\mathbf{r},t)} and phase θ(r,t)\theta(\mathbf{r},t). This leads to the hydrodynamic equations for density ρ(r)\rho(\mathbf{r}) and velocity v=ℏ/(ma)⋅∇θ\mathbf{v}=\hbar/(ma)\cdot\nabla\theta

ρ˙+3Hρ+1a∇(ρv)=0,(7)\dot{\rho}+3H\rho+\frac{1}{a}\nabla(\rho\mathbf{v})=0, \tag*{(7)}
v˙+Hv+1a(v∇)v=−1a∇Φ−1ma3∇Q,(8)\dot{\mathbf{v}}+H\mathbf{v}+\frac{1}{a}(\mathbf{v}\nabla)\mathbf{v}=-\frac{1}{a}\nabla\Phi-\frac{1}{ma^3}\nabla Q, \tag*{(8)}
ΔΦ=4πGa2(ρ−ρ0).(9)\Delta\Phi=4\pi Ga^2(\rho-\rho_0). \tag*{(9)}

Jeans instability and ansatz– Without the second term on the right-hand side of Eq. (6b), those equations are the equivalent of the continuity equation for a collisionless (pressureless) fluid, corresponding to CDM. The additional term describes the quantum pressure potential Q=−ℏ2/(2m)⋅(Δρ)/ρQ=-\hbar^2/(2m)\cdot(\Delta\sqrt{\rho})/\sqrt{\rho} and is a unique feature of ULDM compared to CDM, producing repulsion to counteract attractive gravity [27]. The balance between them is reached for ULDM overdensities with momentum

kJ=(16πGρ0m2a4ℏ2)1/4,(10)k_J=\left(\frac{16\pi G\rho_0m^2a^4}{\hbar^2}\right)^{1/4}, \tag*{(10)}

known as the critical Jeans wavenumber in the expanding Universe [28] (see also Refs. [29, 30, 31] and Supplemental Material [32]). Due to the Jeans instability [25, 29, 30, 31, 33], density fluctuations with k<kJk<k_J grow and undergo gravitational collapse. Thus, in terms of the Jeans wavelength λJ=2π/kJ\lambda_J=2\pi/k_J, perturbations with length scales larger than λJ\lambda_J cause instability and may lead to structure formation [25] with the corresponding Jeans mass [29, 30, 33]

MJ=4a33πρ0(λJ2)3=π46ℏ2Gm2λJ.(11)M_J=\frac{4a^3}{3}\pi\rho_0\left(\frac{\lambda_J}{2}\right)^3=\frac{\pi^4}{6}\frac{\hbar^2}{Gm^2\lambda_J}. \tag*{(11)}

Notably, the length and mass scales λJ\lambda_J and MJM_J are finite and determine the lower limit of the size and mass of ULDM halos formed due to the gravitational collapse of a homogeneous distribution of non-interacting bosons [34]. It is worthwhile to remind here that in CDM (pressureless fluid) corresponding to ℏ=0\hbar=0 and MJ=0M_J=0, density inhomogeneities are unstable on all scales inside the Hubble radius since CDM has no quantum pressure.

While from the discussion above, we can get some qualitative insight into structure formation in ULDM, the complete description of this process in the expanding Universe is computationally demanding and requires large-scale numerical simulations of the Schrödinger-Poisson equations [8, 19, 35, 36, 37, 38, 39]. While these simulations provide a final state of an ULDM halo for a specific given set of system parameters, there have been many efforts to understand its structure while covering the whole parameter space with analytical approximations and variational methods [27, 34, 40, 41]. In this Letter, we extend these methods to the cosmological problem of structure formation, providing analytical insights that complement and help validate previous numerical studies.

In what follows, we consider an overdensity δ\delta with respect to the background density

ρ(r,t)=ρ0(t)(1+δ(r,t)),(12)\rho(\mathbf{r},t)=\rho_0(t)(1+\delta(\mathbf{r},t)), \tag*{(12)}
ψ(r,t)=ρ(r,t)eiθ(r,t),(13)\psi(\mathbf{r},t)=\sqrt{\rho(\mathbf{r},t)}e^{i\theta(\mathbf{r},t)}, \tag*{(13)}

where ρ0\rho_0 is the background density in the matter-dominated FLRW spacetime given by Eq. (3), c.f., Ref. [18]. A simple model to describe the formation of a galaxy would be to consider a spherically symmetric overdensity of a shape

δ(r,t)=R03R(t)3e−r2/R(t)2,(14)\delta(r,t)=\frac{R_0^3}{R(t)^3}e^{-r^2/R(t)^2}, \tag*{(14)}

where R(t)R(t) is the size of the halo. Here, the distance scale R0R_0 is related to the total mass of the ULDM inhomogeneity M=∫∣ψ∣2dr=π3/2R03ρfM=\int\lvert\psi\rvert^2dr=\pi^{3/2}R_0^3\rho_f. We note that this type of ansatz was first introduced to describe nonlinear pulse propagation in optical fibers [42], an atom BEC in a harmonic trap [43], and then further developed in the context of ULDM in the absence or weak ULDM self-interaction [34]. Armed with ansatz (9), one maps the problem of solving the SPE system (5) for field ψ(r,t)\psi(\mathbf{r},t) to a simpler problem with time-dependent parameter R(t)R(t). In the discussed cosmological setting, we expect that the perturbation δ\delta is small in the linear regime realized initially and becomes locally much more dense than the background later in the nonlinear regime. The density inhomogeneity ρ(r)−ρ0\rho(\mathbf{r})-\rho_0 generates the gravitational potential Φ\Phi according to Eq. (5b)

Φ=−Ga2ρ0π3/2R03Erf⁡[r/R]r,(15)\Phi=-Ga^2\rho_0\pi^{3/2}R_0^3\frac{\operatorname{Erf}[r/R]}{r}, \tag*{(15)}

where the contribution from the background density ρ0\rho_0 is given by Φ0=Ga2ρ02πr2\Phi_0=Ga^2\rho_0^2\pi r^2 and is subtracted from the total gravitational potential to obtain the gravitational potential Φ\Phi of the perturbation.

The corresponding ansatz for the phase θ(r,t)\theta(r,t) of the wavefunction (9b) has to satisfy the continuity equation (6a). For the considered density ansatz (9a) this equation reads

δ˙+1a∇[(1+δ)v]=0.(16)\dot{\delta}+\frac{1}{a}\nabla[(1+\delta)\mathbf{v}]=0. \tag*{(16)}

In accord with the spherical symmetry of our variational ansatz we assume that v=v(r)er\mathbf{v}=v(r)\mathbf{e}_r is directed along ∇δ\nabla\delta. In addition, we expect that the physically meaningful solution should be finite at r=0r=0. Thus, we find

v(r)=ra(t)1+er2/R2(R/R0)3R˙R=raR˙1+δ1R(17)v(r)=\frac{r a(t)}{1+e^{r^2/R^2}(R/R_0)^3}\frac{\dot{R}}{R}=\frac{r a\dot{R}}{1+\delta}\frac{1}{R} \tag*{(17)}
j(r)=ρ(r)v(r)er=ρ0aR03R3e−r2/R2R˙Rrer(18)\mathbf{j}(r)=\rho(r)v(r)\mathbf{e}_r=\rho_0a\frac{R_0^3}{R^3}e^{-r^2/R^2}\frac{\dot{R}}{R}r\mathbf{e}_r \tag*{(18)}

so that v=0\mathbf{v}=0 in equilibrium when R˙=0\dot{R}=0. The result (13) implies that the current j\mathbf{j} and the velocity v(r)v(r) decay exponentially at large distances from the fluctuation. At small distances, in contrast, when the fluctuation density dominates over the background, i.e., R≪R0R\ll R_0, we recover the velocity profile v=a(t)R˙Rr\mathbf{v}=a(t)\frac{\dot{R}}{R}\mathbf{r} and the corresponding phase θ∼r2\theta\sim r^2 as is well known for the variational analysis of the Gaussian density in the absence of background, c.f. Refs. [44] and further details in Supplemental Material [32]. In the general case, phase θ(r,t)\theta(r,t) can be derived from Eq. (13a) and reads

θ(r,t)=−βR2ln⁡(1+(R0R)3e−r2/R2),(19)\theta(r,t)=-\beta R^2\ln\left(1+\left(\frac{R_0}{R}\right)^3e^{-r^2/R^2}\right), \tag*{(19)}

where β(t)=ma2/(2ℏ)⋅R˙/R\beta(t)=ma^2/(2\hbar)\cdot\dot{R}/R. Following the ideas of Refs. [43, 45], we introduce β(t)\beta(t) as a canonically conjugate variational parameter to the halo size R(t)R(t). Here, this particular parametrization is chosen to be analogous to the previous studies with a Gaussian ansatz in the absence of background [43], when θ(r,t)=β(t)r2\theta(r,t)=\beta(t)r^2.

Equations of motion— Let us now determine the dynamics of the ULDM following the approach of Refs. [34, 43] using the SP action

SSP=∫dt L=a3m∫dt∫dr ψ∗[iℏ∂∂t+3iℏ2H+ℏ22ma2∇2−m2Φ]ψ.(20)S_{\mathrm{SP}}=\int dt\,\mathcal{L}=\frac{a^3}{m}\int dt\int d\mathbf{r}\,\psi^*\left[i\hbar\frac{\partial}{\partial t}+\frac{3i\hbar}{2}H+\frac{\hbar^2}{2ma^2}\nabla^2-\frac{m}{2}\Phi\right]\psi. \tag*{(20)}

Here we plug our ansatz (9) for the wavefunction ψ\psi using the Gaussian ansatz (10) for overdensity δ\delta, Eq. (19) for phase θ\theta, and Eq. (15) for gravitational potential Φ\Phi. With this, we find

L=2πℏρ0a3m(2R(β˙R+2βR˙)∫dr r2(1+δ)log⁡[1+δ]−ℏma2[4β2+1R4∫dr r4δ21+δ+π3/2Gma2ρ0R0623/2ℏR])(21)\mathcal{L}=\frac{2\pi\hbar\rho_0a^3}{m}\left(2R\left(\dot{\beta}R+2\beta\dot{R}\right)\int dr\,r^2(1+\delta)\log[1+\delta]-\frac{\hbar}{ma^2}\left[4\beta^2+\frac{1}{R^4}\int dr\,r^4\frac{\delta^2}{1+\delta}+\frac{\pi^{3/2}Gma^2\rho_0R_0^6}{2^{3/2}\hbar R}\right]\right) \tag*{(21)}

see details on the derivation in Supplemental Material [32]. Here the first line describes the time dependence of the wavefunction parameterized by the time derivative of the phase θ\theta. The first term in the second line is the kinetic energy - both classical ∼(θ′)2∼∣v∣2\sim(\theta')^2\sim|\mathbf{v}|^2 and quantum, associated with the quantum pressure, c.f. Refs. [27, 46].

The last term is the gravitational energy due to the potential Φ\Phi generated by the fluctuation. The common multiplier of the Lagrangian (16) is constant since a3∼t2a^3\sim t^2 and ρ0∼1/t2\rho_0\sim1/t^2 (see Eq. (3)).

For the given Lagrangian (16) the evolution of two variational parameters R(t)R(t) and β(t)\beta(t) can be predicted from the Lagrange equations of motion. The equation for β(t)\beta(t) simplifies to the solution of the continuity equation β(t)=ma2/(2ℏ)⋅R˙/R\beta(t)=ma^2/(2\hbar)\cdot\dot{R}/R, as shown in Supplemental Material [32]. The Lagrange equation of motion for R(t)R(t) reads

R¨+2a˙aR˙+R˙2R(I22I1−1)+3π322I1(a˙a)2R+ℏ2m2a4R3(I22I1−2)=0,(22)\ddot{R}+2\frac{\dot{a}}{a}\dot{R}+\frac{\dot{R}^2}{R}\left(\frac{I_2}{2I_1}-1\right)+\frac{3\sqrt{\pi}}{32\sqrt{2}I_1}\left(\frac{\dot{a}}{a}\right)^2R+\frac{\hbar^2}{m^2a^4R^3}\left(\frac{I_2}{2I_1}-2\right)=0, \tag*{(22)}
I1(R)=∫0+∞dx x4e−2x21+(R0/R)3e−x2,(23)I_1(R)=\int_0^{+\infty}dx\,\frac{x^4e^{-2x^2}}{1+(R_0/R)^3e^{-x^2}}, \tag*{(23)}
I2(R)=∫0+∞dx x4(2x2−3)e−2x21+(R0/R)3e−x2,(24)I_2(R)=\int_0^{+\infty}dx\,\frac{x^4(2x^2-3)e^{-2x^2}}{1+(R_0/R)^3e^{-x^2}}, \tag*{(24)}

where we introduced a dimensionless integration variable x=r/Rx=r/R and used ρ0=3/(8πG)⋅(a˙/a)2\rho_0=3/(8\pi G)\cdot(\dot{a}/a)^2 from Eq. (3). Integrals I1/2(R)I_{1/2}(R), which quantify different contributions in Eq. (17a), are illustrated and further discussed in Supplemental Material [32].

Linear and nonlinear regimes— Having found the equation of motion for the halo size in the expanding Universe, we in what follows investigate the two limits, frequently examined in the literature, where the halo density is either small (linear stage) or locally dominant (nonlinear stage) compared with the background density. During the linear stage (R0/R)3≪1(R_0/R)^3\ll1, we have I1=3π/(322)I_1=3\sqrt{\pi}/(32\sqrt{2}) and I2=−I1/2I_2=-I_1/2. Hence, the equation for R(t)R(t) simplifies to

R¨+2a˙aR˙−5R˙24R+(a˙a)2R−9ℏ24m2a4R3=0,(25)\ddot{R}+2\frac{\dot{a}}{a}\dot{R}-\frac{5\dot{R}^2}{4R}+\left(\frac{\dot{a}}{a}\right)^2R-\frac{9\hbar^2}{4m^2a^4R^3}=0, \tag*{(25)}

where the fourth term is the attractive gravity and the last term is the repulsive quantum pressure. We find that these two effects compensate each other at the critical fluctuation size

RJ=(aa˙)1/2(9ℏ24m2a4)1/4,(26)R_J=\left(\frac{a}{\dot{a}}\right)^{1/2}\left(\frac{9\hbar^2}{4m^2a^4}\right)^{1/4}, \tag*{(26)}

which, up to a prefactor RJ/λJ≈0.3R_J/\lambda_J\approx0.3, coincides with the Jeans wavelength λJ=2π/kJ\lambda_J=2\pi/k_J, where kJk_J is given by Eq. (10) and relation (3) is used. Thus, in accord with the Jeans instability mechanism, for an initial condition R≫RJR\gg R_J and R˙=0\dot{R}=0 the linear gravitational attraction ∼R\sim R dominates and drives the collapse of the halo. In the ultimate case of negligible quantum pressure, Eq. (25) can be solved assuming R=AtαR=At^\alpha, which leads to the two solutions α=2(1±5)/3\alpha=2(1\pm\sqrt{5})/3, with the negative branch describing contraction of the overdensity. The corresponding central contrast evolves as δ(0,t)∼t2(5−1)\delta(0,t) \sim t^{2(\sqrt{5}-1)}, rather than the standard pressureless linear growing mode δ∼t2/3\delta\sim t^{2/3} [28]. This discrepancy reflects the imposed relation between the amplitude and width of the Gaussian profile (10).

In the nonlinear regime (R0/R)3≫1(R_0/R)^3 \gg1, we have I1=3π/8×(R/R0)3I_1 = 3\sqrt{\pi}/8 \times(R/R_0)^3 and I2=2I1I_2 = 2I_1 that gives

R¨+2a˙aR˙−ℏ2m2a4R3+142(a˙a)2R03R2=0.(27)\ddot{R}+2\frac{\dot{a}}{a}\dot{R}-\frac{\hbar^2}{m^2a^4R^3}+\frac{1}{4\sqrt{2}}\left(\frac{\dot{a}}{a}\right)^2\frac{R_0^3}{R^2}=0. \tag*{(27)}

The latter equation unlike Eq. (25), where the gravitational term is ∼R\sim R, includes gravity as a nonlinear term ∼1/R3\sim1/R^3. This behavior allows the gravitational attraction to play an important role throughout the entire evolution of a ULDM halo. We also note that Eq. (27) can be obtained as an exact Lagrange equation for the parameter R(t)R(t) if, instead of wavefunction (9), one would consider

ψ=ρ0(eiα+δeiθ)(28)\psi=\sqrt{\rho_0}\left(e^{i\alpha}+\sqrt{\delta}e^{i\theta}\right) \tag*{(28)}

in the action of SP equations given by Eq. (20). Here phase α(r)\alpha(r) is a coordinate-dependent random phase, which makes the background and the fluctuation an incoherent mixture, similarly to the description of gravitating solitons and finite-temperature BECs in Refs. [47, 48]. This reflects halo-background decoupling rather than dephasing because, as soon as the background amplitude becomes negligible, the relative phase becomes dynamically irrelevant. At this point an isolated halo is formed, see further details in Supplemental Material [32].

Numerical analysis— To describe the complete evolution of the halo size, we solve Eq. (17) numerically. For that we account for the time dependency of scale factor a(t)=(t/tf)2/3a(t)=(t/t_f)^{2/3} and introduce convenient dimensionless variables y=R/R0y=R/R_0 and τ=t/tf\tau=t/t_f, in terms of which the equation reads

d2ydτ2+43τdydτ+1y(dydτ)2(I22I1−1)+2πy48I1τ2+qτ8/3y3(I22I1−2)=0,(29)\frac{d^2y}{d\tau^2}+\frac{4}{3\tau}\frac{dy}{d\tau}+\frac{1}{y}\left(\frac{dy}{d\tau}\right)^2\left(\frac{I_2}{2I_1}-1\right)+\frac{\sqrt{2\pi}y}{48I_1\tau^2}+\frac{q}{\tau^{8/3}y^3}\left(\frac{I_2}{2I_1}-2\right)=0, \tag*{(29)}
I1(y)=∫0+∞dxx4e−2x21+e−x2/y3,(30)I_1(y)=\int_0^{+\infty}dx\frac{x^4e^{-2x^2}}{1+e^{-x^2/y^3}}, \tag*{(30)}
I2(y)=∫0+∞dxx4(2x2−3)e−2x21+e−x2/y3,(31)I_2(y)=\int_0^{+\infty}dx\frac{x^4(2x^2-3)e^{-2x^2}}{1+e^{-x^2/y^3}}, \tag*{(31)}

where we used q=ℏ2tf2/(m2R04)q=\hbar^2t_f^2/(m^2R_0^4) as a dimensionless parameter characterizing quantum pressure. Here the length scale R0=(M/(π3/2ρf))1/3R_0=(M/(\pi^{3/2}\rho_f))^{1/3} is related to the total mass MM of overdensity (10).

To fully formulate the problem, we need to fix our initial conditions for the density contrast δ(ti)\delta(t_i) of the primordial fluctuations that will lead to a halo of given mass. For that, we take input from the cosmic microwave background (CMB) that was released at the epoch of recombination, approximately ti=3.8×105t_i=3.8\times10^5 years (z=1100z=1100) after the Big Bang. Measurements of the CMB temperature and polarization anisotropies show that the Universe contained small primordial fluctuations, with a characteristic amplitude of order ΔT/T∼10−5\Delta T/T\sim10^{-5} [17]. These temperature fluctuations allow reconstructing a CDM total matter power spectrum at z=1100z=1100 using the CAMB code [49], assuming a [17] cosmology. For a given ULDM particle mass, the power spectrum is then modified using the corresponding ULDM transfer function [50]. To relate the final halo mass and the density contrast, we first compute the value of the perturbation wavenumber at the origin of collapse. This is done by considering the mass enclosed in a sphere of radius equal to half its wavelength [51 e.g.]. Having determined kk, the density contrast is obtained from the power spectrum at this wavenumber.

The primordial fluctuations provide the initial conditions for the subsequent growth of cosmic structure through gravitational instability, as shown in Figure 1. A CDM halo experiences singular collapse due to the absence of quantum pressure to compensate gravity, resulting in zero Jeans scale. On the other hand, a ULDM halo collapses due to Jeans instability y(τi)=RJ/R0y(\tau_i)=R_J/R_0, see Eq. (26), and further enters the nonlinear stage only if the initial overdensity δ(τi)=y−3(τi)\delta(\tau_i)=y^{-3}(\tau_i) fulfills δ(τi)>1.5×10−8\delta(\tau_i)>1.5\times10^{-8}. This limits the boson mass mm from below, since for small boson masses with δ(τi)<1.5×10−8\delta(\tau_i)<1.5\times10^{-8} the collapse within the matter-dominated epoch is not possible, see m=10−22m=10^{-22} eV/c2c^2 in Figure 1.

Numerical solution to Eq. (22)

Figure 1. Numerical solution to Eq. (22) assuming the initial condition dy(τi)/dτ=R˙(ti)=0dy(\tau_i)/d\tau=\dot{R}(t_i)=0, y(τi)=R(ti)/R0=δ−3(τi)y(\tau_i)=R(t_i)/R_0=\delta^{-3}(\tau_i) at the initial point of time τi=3.88×10−5\tau_i=3.88\times10^{-5} (ti=3.8×105t_i=3.8\times10^5 yr). Here δ(τi)\delta(\tau_i) is evaluated with the CAMB code and equals 7.372×10−117.372\times10^{-11}, 2.864×10−62.864\times10^{-6}, and 3.944×10−33.944\times10^{-3} for the boson masses 10−2210^{-22} eV/c2c^2, 10−2110^{-21} eV/c2c^2, and 10−2010^{-20} eV/c2c^2 correspondingly. The solution R(t)R(t) for these different values of the boson mass mm is shown with solid blue lines. For the CDM δ(τi)=6.335×10−3\delta(\tau_i)=6.335\times10^{-3} and q=0q=0, see orange dashed line. For all lines, the halo mass M=108M⊙M=10^8M_\odot is fixed, giving R0=52.7R_0=52.7 kpc. Note that while the time-scale of the oscillations for τ>10−2\tau>10^{-2} appears to decrease on the log-scale, it remains constant on a linear scale, c.f., Figure 2.

We also find that after a ULDM halo decouples from the background density, its size R(t)R(t) oscillates. To further investigate this dynamics, presented in Figure 1, we analyze the energy of the system. The kinetic energy Θk\Theta_k and gravitational energy WW, which enter Lagrangian (16), read

Θk=Θkq+Θkc(32)\Theta_k=\Theta_k^q+\Theta_k^c \tag*{(32)}
Θkq=R053Gtf4I1(y)y3qq4/3y2(33)\Theta_k^q=\frac{R_0^5}{3Gt_f^4}\frac{I_1(y)}{y^3}\frac{q}{q^{4/3}y^2} \tag*{(33)}
Θkc=R053Gtf4I1(y)y3y4/3(dydτ)2(34)\Theta_k^c=\frac{R_0^5}{3Gt_f^4}\frac{I_1(y)}{y^3}y^{4/3}\left(\frac{dy}{d\tau}\right)^2 \tag*{(34)}
W=−π362R05Gtf41τ2/3y(35)W=-\frac{\sqrt{\pi}}{36\sqrt{2}}\frac{R_0^5}{Gt_f^4}\frac{1}{\tau^{2/3}y} \tag*{(35)}

where the dimensionless units τ=t/tf\tau=t/t_f and y=R/R0y=R/R_0 are used, see details of the derivation in Supplemental Material [32]. Here we explicitly separated the classical kinetic energy Θkc\Theta_k^c due to the velocity dy/dτdy/d\tau of the change of the halo size and the quantum kinetic energy Θkq∼q\Theta_k^q\sim q, which is induced by the quantum pressure. The latter is a unique feature of ULDM models, c.f. Refs. [27, 46] and is sometimes referred to as thermal energy [52]. The three energy components are illustrated in Figure 2, from where we also see that the total energy E=Θk+WE=\Theta_k+W of the ULDM halo becomes time-independent during the nonlinear stage, when the halo decouples from the background.

Numerical solution for the quantum kinetic energy, classical kinetic energy, gravitational energy and total energy

Figure 2. Numerical solution for the quantum kinetic energy Θkq\Theta_k^q, classical kinetic energy Θkc\Theta_k^c, gravitational energy WW and total energy E=Θkq+Θkc+WE=\Theta_k^q+\Theta_k^c+W depending on τ=t/tf\tau=t/t_f in the interval τ∈[0.75,1]\tau\in[0.75,1] as given by Eq. (23) for R(t)R(t) presented in Figure 1. The line EeqE_{\rm eq} stands for the stationary solution at yeq=92qy_{\rm eq}=9\sqrt{2}q in the case of a ULDM halo decoupled from the background. Here all energies are given in the units of ϵ=R05/(Gtf4)\epsilon=R_0^5/(Gt_f^4) and the boson and halo masses are fixed to m=10−20 eV/c2m=10^{-20}\,\mathrm{eV}/c^2 and M=108M⊙M=10^8M_\odot. The y-axis is scaled on a symmetric arcsinh-scale.

At the minima of R(t)R(t) both the kinetic and the gravitational energy have sharp extrema, and further gravitational collapse is prevented mainly by quantum pressure, which dominates the kinetic energy. Thus, our findings show how the quantum pressure stabilizes a ULDM halo and sets a finite minimum length scale of the structure formation. At plateaus between the minima of R(t)R(t), all energy components approach their absolute minimum values corresponding to the big halo size R(t)R(t) and small velocity R˙(t)\dot{R}(t). At these points, the static equilibrium is, however, not established due to the presence of an excess energy in the system compared with the energy

Eeq=−π1296qR05Gtf4(36)E_{\rm eq}=-\frac{\sqrt{\pi}}{1296q}\frac{R_0^5}{Gt_f^4} \tag*{(36)}

of the static solution of the same mass

Req=32πGρfR03m2ℏ2=0.99(10−22 eV/c2m)2/3kpc,(37)R_{\rm eq}=\frac{3}{\sqrt{2\pi}G\rho_fR_0^3m^2}\hbar^2=0.99\left(\frac{10^{-22}\,\mathrm{eV}/c^2}{m}\right)^{2/3}\mathrm{kpc}, \tag*{(37)}

decoupled from the background and the Universe expansion, see Supplemental Material [32] for details. In the given dimensionless units, it reads yeq=Req/R0=92qy_{\rm eq}=R_{\rm eq}/R_0=9\sqrt{2}q. Since the total energy of the halo EE is larger than the energy EeqE_{\rm eq} of an equilibrium Gaussian soliton of the same mass, the isolated halo cannot evolve into such a soliton but instead evolves into a higher energy state.

Core and envelope– The density profile of this higher energy state appears more complex than the one introduced by (14) due to nonspherical evolution, and hence cannot be described by a spherically-symmetric analytical model, presented in this Letter. This state is formed via the complex virialization dynamics that leads to the formation of a halo composed of a soliton and a Navarro-Frenk-White envelope, which emerges in numerical simulations of ULDM structure formation [46, 53, 54]. There, non-spherical gravitational collapse converts potential energy into kinetic (wave) energy. Violent relaxation and wave interference redistribute this energy, while gravitational cooling can transfer energy out of the bound region through the emission of waves, allowing the remaining system to approach virial equilibrium [55, 56]. In the inner region, wave dynamics subsequently leads to the formation of a dense, approximately stationary solitonic core surrounded by an extended virialized envelope [56, 57, 58].

While the spherically-symmetric analytical ansatz (10) does not allow for a full description of this virialization dynamics, we can still infer the properties of the final core-envelope configuration from our calculations, using the core-halo relations investigated in Refs. [10, 59]. They relate the core energy EcE_c and mass McM_c with the total energy EE and mass MM of the halo, consisting of the core and an envelope. In accord with them, the fraction of the mass in the soliton core lies between the limit of energy equipartition [10]

EcMc=EM,(38)\frac{E_c}{M_c}=\frac{E}{M}, \tag*{(38)}

describing a halo directly after virialization, and the limit of all energy being located in the central core [59]

Ec=E,(39)E_c=E, \tag*{(39)}

describing a halo at later times after gravitational cooling. We obtain the total energy E=Θk+WE=\Theta_k+W for given halo mass MM and boson mass mm by substituting the numerical solution of Eqs. Eqs. (22) into Eqs. Eqs. (23) at a nonlinear stage, when the energy approaches a constant, as illustrated in Figure 2. The density of the virialized core can be approximated with a Gaussian soliton, see Supplemental Material [32], whose energy at equilibrium reads

Ec=−G2m2Mc36πℏ2,(40)E_c=-\frac{G^2m^2M_c^3}{6\pi\hbar^2}, \tag*{(40)}

and hence the core mass McM_c remains the only unknown parameter to be determined from the relations (26) and (27). The core-halo mass ratio Mc/MM_c/M for these two relations is illustrated in Figure 3. We find that, with an increase of the boson mass mm, the core in the center becomes smaller and the halo tends to the CDM limit with no core. This behavior emerges due to the quantum pressure, whose impact is most pronounced for small mm. It has a two-fold effect in our calculations: it suppresses initial overdensities, see CAMB code [49] and Figure 1, and impacts further evolution of the halo size, entering Eqs. Eqs. (22) with the parameter q∼1/m2q\sim1/m^2. In the ultimate case of m>10−19 eVm>10^{-19}\,\mathrm{eV}, the core mass constitutes only ≲2%\lesssim2\% of the total halo mass, and hence the ULDM tends to the CDM limit with q=0q=0, i.e., no quantum pressure.

Ratio of core and total halo masses and initial overdensity depending on particle mass

Figure 3. Ratio Mc/MM_c/M of the core and total halo masses (left y-axis, linear scale) and the initial overdensity δ(τi)=y−3(τi)\delta(\tau_i)=y^{-3}(\tau_i) used for numerical solution of Eq. (22) (right y-axis, log-scale) depending on the particle mass mm for a fixed halo mass M=108M⊙M=10^8M_\odot. The yellow line is calculated assuming energy equipartition between Ec/Mc=E/ME_c/M_c=E/M, while the green line corresponds to Ec=EE_c=E, i.e., all energy located in the central core. The dark blue line shows the initial overdensity δ(τi)\delta(\tau_i). The shaded region between the two lines covers all possible intermediate virialization stages.

Let us discuss the decrease of the core mass McM_c as a function of mm, predicted by our model, in the context of astrophysical observations. The total halo mass of the order of 108M⊙10^8M_\odot is typical for the dwarf spheroidal galaxies, orbiting the Milky Way [60]. The stellar dynamics in these galaxies was fitted assuming CDM with a cuspy NFW density distribution [60], ULDM with a boson mass ∼10−20 eV/c2\sim10^{-20}\,\mathrm{eV}/c^2 [12], and ∼10−22 eV/c2\sim10^{-22}\,\mathrm{eV}/c^2 [13]. It was found that for a bigger boson mass, the central core constitutes only a small fraction of the total halo mass and size, while for the boson mass as small as ∼10−22 eV/c2\sim10^{-22}\,\mathrm{eV}/c^2 the stellar dynamics can be completely described by a Gaussian soliton, hence with only a smaller fraction of halo mass located in the envelope. This example of astrophysical observations shows the same tendency as our results in Figure 3, however for a more general conclusion further studies are required.

We also note that the prediction of mass distribution between the core and envelope in a ULDM halo, illustrated in Figure 3, could be used to further test the underlying assumptions of the model. For instance, deviations from the results presented in Figure 3 may indicate that the collapse of a particular halo was non-spherical or significantly influenced by collisions with other haloes during nonlinear stage.

Summary and discussion– To summarize, in this Letter the formation of a ULDM halo in the expanding matter-dominated Universe was investigated by using the approach of variational ansatz, which complements the previous research on structure formation with numerical simulations [8, 19, 35, 36]. This approach allowed us to formulate an analytical expression for the ULDM wavefunction consisting of a uniform background and an inhomogeneous spherically symmetric halo. Making no further assumptions about the relative densities of the halo and the background, we formulated an analytical model, which addresses both linear and nonlinear stages of structure formation. In particular, we derived an effective Lagrangian as a function of two variational parameters characterizing the size and phase of the ULDM halo; in view of the Universe expansion, the Lagrangian is explicitly time-dependent. With this, we isolated different contributions to the equation of motion for the halo size R(t)R(t), i.e., gravitational attraction, quantum pressure, classical kinetic energy, and Universe expansion.

The initial conditions for the structure formation are set by the power spectrum of primordial fluctuations, which is sensitive to the quantum pressure. While for a large boson mass m≳10−19 eV/c2m\gtrsim10^{-19}\,\mathrm{eV}/c^2 the initial overdensity is the same as for CDM, for smaller boson masses the quantum pressure becomes significant and suppresses density fluctuations, providing much smaller initial overdensities. For the subsequent evolution, during the linear stage when the density fluctuations are small, the coherence between them and the uniform background plays a crucial role in the structure formation. According to our results, this leads to a gravitational attraction linear in R(t)R(t) and drives the gravitational collapse towards a ∼103\sim10^3 times more compact halo compared with the initial small overdensity.

The presence of a uniform background plays a significant role only during the linear stage of structure formation, while it can be neglected during the nonlinear stage when the halo completely decouples from the background. Moreover, in contrast to the linear stage, the gravitational term in the equation for the halo size evolution is nonlinear ∼1/R2\sim1/R^{2}, as also evident from the variational analysis of a Gaussian halo density in a static Universe [34]. The self-gravity is counteracted by the quantum pressure, which is a unique feature of ULDM and suppresses structure formation on small scales. The interplay of these two effects leads to oscillations of the halo size while the total energy tends to a constant. With this, we analyzed the possible scenarios of how this energy is distributed between a core and an envelope of the halo after virialization. This allowed us to formulate the predictions of our model for the final core-envelope structure, which are in agreement with the ULDM haloes, fitted to describe astrophysical observations [12, 13, 60].

Acknowledgments– The work of E.V.G. and Y.R. was support by the Swiss National Science Foundation through the Ukrainian-Swiss Joint research project "Cosmic waltz of baryonic and ultralight dark matter: interaction and dynamical interplay" (grant No. IZURZ2_224972). K.K. acknowledges funding by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy—EXC 2123 QuantumFrontiers—390837967. The authors are grateful to Bohdan Hnatyk, Yurii Bidasyuk, Andrii Chaika, and Srikanth Nagesh for useful discussions.

References

  1. [1]F. Zwicky, Helv. Phys. Acta 6, 110 (1933).
  2. [2]H. Andernach and F. Zwicky, arXiv preprint arXiv:1711.01693 (2017).
  3. [3]G. R. Blumenthal, S. Faber, J. R. Primack, and M. J. Rees, Nature 311, 517 (1984).
  4. [4]D. H. Weinberg, J. S. Bullock, F. Governato, R. Kuzio de Naray, and A. H. Peter, Proceedings of the National Academy of Sciences 112, 12249 (2015).
  5. [5]W. J. G. de Blok, Advances in Astronomy 2010, 10.1155/2010/789293 (2009).
  6. [6]E. G. M. Ferreira, The Astronomy and Astrophysics Review 29, 10.1007/s00159-021-00135-6 (2021).
  7. [7]D. F. D. Marsh and K. V. Bubbier, The search for ultralight bosonic dark matter (Springer Nature, 2023).
  8. [8]H.-Y. Schive, Living Reviews in Computational Astrophysics 12, 10.1007/s41115-026-00027-5 (2026).
  9. [9]P.-H. Chavanis, Frontiers in Astronomy and Space Sciences 12, 1538434 (2025).
  10. [10]N. Bar, D. Blas, K. Blum, and S. Sibiryakov, Physical Review D 98, 083027 (2018).
  11. [11]K. Korshynska, Y. M. Bidasyuk, E. V. Gorbar, J. Jia, and A. I. Yakimenko, The European Physical Journal C 83, 10.1140/epjc/s10052-023-11548-1 (2023).
  12. [12]I. S. Goldstein, S. M. Koushiappas, and M. G. Walker, Physical Review D 106, 10.1103/physrevd.106.063010 (2022).
  13. [13]K. Korshynska, E. V. Gorbar, Y. M. Bidasyuk, A. I. Yakimenko, and Y. Revaz, Phys. Rev. D 113, 063006 (2026).
  14. [14]J. Binney and S. Tremaine, Galactic dynamics (Princeton university press, 2011).
  15. [15]D. J. Marsh, Physics Reports 643, 1–79 (2016).
  16. [16]D. J. E. Marsh and P. G. Ferreira, Phys. Rev. D 82, 103528 (2010).
  17. [17]N. Aghanim, Y. Akrami, M. Ashdown, J. Aumont, C. Baccigalupi, M. Ballardini, A. Banday, R. Barreiro, N. Bartolo, S. Basak, R. Battye, K. Benabed, J.-P. Bernard, M. Bersanelli, P. Bielewicz, J. Bock, J. Bond, J. Borrill, F. Bouchet, and A. Zonca, Astronomy & Astrophysics 641, A6 (2020).
  18. [18]V. A. Rubakov and D. S. Gorbunov, Introduction to the Theory of the Early Universe: Hot big bang theory
  19. [19]H.-Y. Schive, T. Chiueh, and T. Broadhurst, Nature Physics 10, 496 (2014).
  20. [20]T. Harko, Monthly Notices of the Royal Astronomical Society 413, 3095 (2011).
  21. [21]M. W. Jacobs, E. V. Linder, and R. V. Wagoner, Phys. Rev. D 48, 4623 (1993).
  22. [22]J.-c. Hwang, H. Noh, and D. Puetzfeld, Journal of Cosmology and Astroparticle Physics 2008 (03), 010.
  23. [23]J.-c. Hwang and H. Noh, Phys. Rev. D 107, 083508 (2023).
  24. [24]L. M. Widrow and N. Kaiser, Astrophys. J. Lett. 416, L71 (1993).
  25. [25]W. Hu, R. Barkana, and A. Gruzinov, Physical Review Letters 85, 1158–1161 (2000).
  26. [26]E. Madelung, Zeitschrift für Physik 40, 332 (1927).
  27. [27]P.-H. Chavanis, Physical Review D 100, 083022 (2019).
  28. [28]P.-H. Chavanis, Astronomy & Astrophysics 537, A127 (2012).
  29. [29]K. Ourabah, The European Physical Journal C 82, 565 (2022).
  30. [30]A. Suárez and P.-H. Chavanis, Physical Review D 98, 083529 (2018).
  31. [31]P.-H. Chavanis, Physical Review D 103, 10.1103/physrevd.103.123551 (2021).
  32. [32]See supplemental material at [url will be inserted by publisher] for intermediate derivations and supplemental illustrations.
  33. [33]T. Harko, The European Physical Journal C 79, 10.1140/epjc/s10052-019-7285-3 (2019).
  34. [34]P.-H. Chavanis, Physical Review D—Particles, Fields, Gravitation, and Cosmology 84, 043531 (2011).
  35. [35]S. May and V. Springel, Monthly Notices of the Royal Astronomical Society 506, 2603 (2021), https://academic.oup.com/mnras/article-pdf/506/2/2603/39207201/stab1764.pdf.
  36. [36]M. Mina, D. F. Mota, and H. A. Winther, Astronomy & Astrophysics 662, A29 (2022).
  37. [37]J. Veltmaat and J. C. Niemeyer, Phys. Rev. D 94, 123523 (2016), arXiv:1608.00802 [astro-ph.CO].DOI
  38. [38]M. Nori and M. Baldi, MNRAS 478, 3935 (2018).
  39. [39]P. Mocz, A. Fialkov, M. Vogelsberger, F. Becerra, M. A. Amin, S. Bose, M. Boylan-Kolchin, P.-H. Chavanis, L. Hernquist, L. Lancaster, F. Marinacci, V. H. Robles, and J. Zavala, Phys. Rev. Lett. 123, 141301 (2019).
  40. [40]J. Goodman, New Astronomy 5, 103–107 (2000).
  41. [41]H.-Y. Schive, M.-H. Liao, T.-P. Woo, S.-K. Wong, T. Chiueh, T. Broadhurst, and W. P. Hwang, Physical review letters 113, 261302 (2014).
  42. [42]D. Anderson, Phys. Rev. A 27, 3135 (1983).
  43. [43]L. Salasnich, International Journal of Modern Physics B 14, 1–11 (2000).
  44. [44]L. Salasnich, International Journal of Modern Physics B 14, 1 (2000).
  45. [45]P.-H. Chavanis, Phys. Rev. D 94, 083007 (2016).
  46. [46]I.-K. Liu, N. P. Proukakis, and G. Rigopoulos, Monthly Notices of the Royal Astronomical Society 521, 3625–3647 (2023).
  47. [47]J. Garnier, K. Baudin, A. Fusaro, and A. Picozzi, Phys. Rev. Lett. 127, 014101 (2021).
  48. [48]A. Griffin, T. Nikuni, and E. Zaremba, Bose-Condensed Gases at Finite Temperatures (Cambridge University Press, 2009).DOI
  49. [49]A. Lewis, Journal of Cosmology and Astroparticle Physics 2026 (10), 013.
  50. [50]W. Hu, R. Barkana, and A. Gruzinov, Phys. Rev. Lett. 85, 1158 (2000), arXiv:astro-ph/0003365 [astro-ph].DOI
  51. [51]J. S. Bullock and M. Boylan-Kolchin, Annual Review of Astronomy and Astrophysics 55, 343 (2017), arXiv:1707.04256 [astro-ph.CO].DOI
  52. [52]P. R. Holland, The quantum theory of motion: an account of the de Broglie-Bohm causal interpretation of quantum mechanics (Cambridge university press, 1995).
  53. [53]H.-Y. Schive, M.-H. Liao, T.-P. Woo, S.-K. Wong, T. Chiueh, T. Broadhurst, and W.-Y. P. Hwang, Physical Review Letters 113, 10.1103/physrevlett.113.261302 (2014).
  54. [54]S. May, A. J. Bond, and V. Springel, MNRAS 506, 2603 (2021).
  55. [55]E. Seidel and W.-M. Suen, Physical Review Letters 72, 2516–2519 (1994).
  56. [56]J. Veltmaat, J. C. Niemeyer, and B. Schwabe, Physical Review D 98, 043509 (2018), arXiv:1804.09647 [astro-ph.CO].DOI
  57. [57]H.-Y. Schive, T. Chiueh, and T. Broadhurst, Nature Physics 10, 496 (2014), arXiv:1406.6586 [astro-ph.GA].DOI
  58. [58]B. Schwabe, J. C. Niemeyer, and J. F. Engels, Physical Review D 94, 043513 (2016), arXiv:1606.05151 [astro-ph.CO].DOI
  59. [59]P. Mocz, M. Vogelsberger, V. H. Robles, J. Zavala, M. Boylan-Kolchin, A. Fialkov, and L. Hernquist, Monthly Notices of the Royal Astronomical Society 471, 4559 (2017).
  60. [60]M. G. Walker, M. Mateo, E. W. Olszewski, O. Y. Gnedin, X. Wang, B. Sen, and M. Woodroofe, The Astrophysical Journal 667, L53 (2007).

Paper details

vlnc:W_qybvt33z
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents