UCI-HEP-TR-2017-17

On Profiles of Boson Stars with Self-Interactions

Felix Kling∗ and Arvind Rajaraman†

Department of Physics and Astronomy, University of California, Irvine, CA 92697, USA

Under the influence of gravity, light scalar fields can form bound compact objects called boson
stars. We use the semi-analytic approach of matching asymptotic expansions to obtain the profile
for boson stars where the constituent particles have self-interactions. We obtain parametric repre-
sentations of these profiles as a function of the self-interactions, including the case of very strong
self-interactions. We show that our methods agree with solutions obtained by purely numerical
methods. Significant distortions are found as compared to the noninteracting case.

CONTENTS which can distinguish it from backgrounds [20]. It is also

important to know how these profiles are distorted by
arXiv:1712.06539v2 [hep-ph] 15 Jun 2018

I. Introduction 1 the presence of self-interactions, and also by the gravita-

tional effects of matter. For all these reasons, it is timely
II. Gross-Pitaevskii-Poisson Equations 2 to have a precise description of the profiles of boson stars.
A. The real scalar field 2 In a previous communication [21], we analyzed the pro-
B. Limits and Validity 2 file of such objects in the special case when the bosons
C. Scaling Invariance 3 had no self-interactions. These objects are solutions to
a coupled set of equations called the Gross-Pitaevskii-
III. Weakly Coupled Systems and Series Expansion 4 Poisson equations. In that paper, we showed that one
A. Series Expansion 4 could find approximate solutions to these equations, by
B. Expansion Parameters 5 using a combination of analytical and numerical meth-
C. Fit for Weak Couplings: −1 < γ < 1 5 ods. We showed that our methods were numerically sta-
ble, and that they converged uniformly far away from the
IV. Strongly Coupled Systems 6 core of the star, and furthermore were much less compu-
A. Thomas Fermi Limit 6 tationally expensive compared to other purely numerical
B. Fit for Strong Couplings: γ > 1 7 methods [18, 22–35].
V. Applications 8 In this paper, we extend our previous methods to the
A. Using the Solution 8 case of interacting bosons. This is a more difficult situa-
B. Axion Stars 10 tion, because the self-interactions can be much stronger
than the gravitational binding. Nevertheless we show
VI. Conclusion 12 that our methods continue to provide excellent agree-
ment with a fully numerical solution.
Acknowledgments 12 Our results show that self-interactions can produce sig-
nificant distortions of the profile of the star, as compared
References 12 to the non-interacting star. We parametrize these de-
formations by finding a parametrization for the star in
terms of two matching asymptotic expansions, where the
I. INTRODUCTION expansions are taken far away from the core and close to
the core respectively. The parameters of the solution can
The increasingly strong constraints on weakly interact- be found systematically by matching the expansions in
ing massive particles have made axions more attractive an overlap region. We find these parameters as a func-
as a dark matter candidate (for reviews see [1, 2]). Many tion of the interaction strength, which also allows us to
searches are ongoing to find axion-like particles [3–15]. find any desired physical quantity (mass, central density
It has been noted by many authors that the axions etc.) in terms of the coupling. We do this both for weak
might bind into compact spatial structures (e.g. see the couplings, where we can perturb around the noninteract-
review [16]). These are more generally referred to as bo- ing star, as well as for strong couplings, where we can
son stars [17–19]). Such objects, if they exist, would pro- perturb around the Thomas-Fermi limit of the solution.
duce distinctive signatures in axion search experiments, We note that these expansions provide a solution to the
and understanding these signatures requires a descrip- axion stars which have an accuracy of at least 10−3 , and
tion of the profile of the boson stars, which could pro- can hence be used in lieu of complicated numerical cal-
duce unique time and spatial dependencies of the signal culations.
Our results improve on other semi-analytic approaches
(e.g. [33, 36, 37]), which have used variational and other
∗ fkling@uci.edu techniques to find approximations to the boson star pro-
† arajaram@uci.edu file. Our methods are particularly suited to accurate
 2

evaluations of the profile away from the center, where Using ∂t2 φ = −E 2 φ and E = m + e, we can rewrite the
the falloff is well described by a Whittaker function. the Klein-Gordon equation in the non-relativistic limit as
In the following sections we rederive the Gross-
Pitaevskii-Poisson equations satisfied by the boson star.
1 2 λ 3
We also describe various limiting cases, including the −eφ − ∇ φ + mΦφ + φ = 0. (4)
Thomas-Fermi limit of strong coupling. We then present 2m 6m
a series expansion in the two asymptotic regimes, and Inserting the explicit form of the field given in Eq. 2
find the solutions separately for weakly coupled systems and rephasing by eiEt , we obtain the Schrödinger-type
and for strongly coupled systems. We show how our so- equation
lutions apply to the special case of axion stars. We end
with conclusions and directions for future work. 1 2 Nλ 3
eψ = − ∇ ψ + mΦψ + ψ . (5)
2m 4m2

II. GROSS-PITAEVSKII-POISSON EQUATIONS Here we have dropped additional terms with rapidly os-
cillating phase factor e−inEt where n is a non-zero inte-
ger. For the non-relativistic approximation to be con-
A. The real scalar field
sistent, the last term should be sufficiently small, i.e.
Nλ
4m2  m.
In [21] we have derived the structure equations for the The Newtonian potential is related to the energy den-
ground state of a self-gravitating complex scalar field in sity via the Poisson equation ∇2 Φ = 4πGρ. The energy
the non-relativistic limit. Following the procedure de- density ρ of the real scalar field is
scribed in [38], we now show that the same set of equa-
tions also apply for a real scalar field. 1 1 1 λ
ρ= (∂t φ)2 + (∇φ)2 + m2 φ2 + φ4 ≈ N mψ 2 (6)
Let us consider the real scalar field φ(~r, t) described by 2 2 2 12
the Lagrangian
where we used the non-relativistic approximation in the
1 1 1 last step. Newton’s equation therefore takes the simple
L = g µν (∂µ φ)(∂ν φ) − m2 φ2 − λφ4 . (1)
form
2 2 12
In the presence of gravity, the scalar field can form grav- ∇2 Φ = 4πGN mψ 2 (7)
itational bound states, which are called boson stars. A
simple solution can be obtained assuming that only the Comparing with the results of [21], we find that both
ground state is populated. In this case the field can be ex- the ground state of a boson star for both a complex scalar
pressed in terms of a single real function ψ(r), sometimes field and a real scalar field are described by the same
called the wave function of the boson star, describing the set of equations given in Eq. 5 and Eq. 7. These are
radial profile of the boson star. We can write often referred to as Gross-Pitaevskii-Poisson equations.
r Note that the Gross-Pitaevskii-Poisson equations for a
N real scalar field have also been obtained by the authors
ψ(r) e−iEt + eiEt (2)


φ(t, r) =
2E of [29], which follow a semi-classical approach considering
a quantized scalar field φ.
where E is the ground state energy and N is number of
bosons in the ground state. Note that properly shifting
the time coordinate allows us to absorb a possible phase B. Limits and Validity
of the wave function and therefore to choose ψ toR be real.
We have chosen a wave function normalization ψ 2 dV =
The terms on the right-hand side of Eq. 5 represent the
1 which allows us to identify ψ 2 with as the probability
contribution to the energy of a scalar particle due to the
density.
quantum pressure, gravity and classical pressure respec-
The equation of motion of the scalar field is the Klein-
tively. The quantum pressure is a consequence of Heisen-
Gordon equation φ + m2 φ + λ3 φ3 = 0. Assuming that
berg’s uncertainty principle and is always repulsive, pre-
the field couples only weakly to gravity, we can use a
venting the star from gravitational collapse. Gravity on
Newtonian approximation. This allows us to introduce
the other hand is always attractive. The classical pres-
the Newtonian potential Φ in the metric gµν = diag(1 +
sure arises from he contact interaction term and can ei-
2Φ, −1, −1, −1). We can then rewrite the Klein-Gordon
ther be attractive or repulsive, depending on the sign of
equation as
the self coupling λ. It is illustrative to consider the limits
∂t2 φ λ in which one of the three contributions is negligible.
− ∇2 φ + m2 φ + φ3 = 0. (3) Using scaling relations between the coupling λ, the
1 + 2Φ 3
star’s mass M and the star’s radius R, we will quali-
Let us further assume that the ground state is non- tatively discuss both physical properties of the star and
relativistic. In this case we can write E = m + e with the validity of the solution. Without providing a for-
binding energy e  m. This implies eψ, Φψ, ∇ψ  mψ. mal definition of the star’s radius R, we know that the
 3

probability density inside the star scales like ψ 2 ∼ R13 . of particles and therefore the star’s mass. The binding
ψ energy of a scalar particle is
Similarly, a field derivative will scale as ∇ψ = R . For a
more rigorous discussion of the mass-radius relations of GM m M m3
boson stars see [36]. e ∼ mΦ ∼ ∼√ 3 (12)
R λ Mpl
The non-relativistic approximation requires e  m
Non-Interacting Limit λ = 0 max
√ M3 √ M max
which implies M  Mλ>0 = λ mpl 2 = λ mpl Mλ=0 .
Larger couplings increase the validity range of the non-
In the non-interacting case, λ = 0, the quantum relativistic approximation to higher masses M .
pressure balances gravity. We have obtained a semi- An analytic solution for the Thomas-Fermi limit has
analytical solution for this case in [21]. Note that for been obtained in [23] and is discussed in Sec. IV.
non-negligible couplings λ 6= 0, the boson star becomes
effectively non-interacting at large radius due to low den-
sities. Non-Gravitational Limit
We can rewrite Eq. 5 and 7 as
 2  For sufficiently strong attractive self-interactions, λ <
1 ∇ ψ
2
4πGM ψ = ∇ Φ =2
2
∇ 2
. (8) λ∗ , quantum pressure balances the attractive self-
2m ψ interaction while the effect of gravity becomes negligi-
Using the scaling behavior discussed above, we see that ble. Gravity and the classical pressure become equally
the radius of a non-self-interacting boson star scales as important at λ∗ when
R ∼ (GM m2 )−1 . This is an remarkable result: the star’s M |λ∗ | 2
mΦ = ψ (13)
radius decreases when its mass increases. The binding 4m3
energy of a scalar particle is
Using Φ ∼ GM
R and the radial size R ∼ (GM m2 )−1 this
GM m M2
e ∼ mΦ ∼ ∼ G2 M 2 m3 . (9) implies a critical coupling of λ∗ ∼ −Gm4 R2 ∼ − Mpl2 .
R For a given coupling λ, the non-gravitational limit ap-
The non-relativistic approximation requires e  m which plies for stars√with R < R∗ where the critical radius
2 M |λ|
max
implies M  Mλ=0 =
Mpl is R∗ = pl
with a corresponding critical mass
m . m2
1
M ∗ = (Gm2 R∗ )−1 = Mpl |λ|− 2 .
∗
For λ < λ we can rewrite Eq. 5 as
Thomas-Fermi Limit
1 2 ∇2 ψ
 
M |λ| 2 2
∇ = ∇ ψ . (14)
2m ψ 4m3
For strong repulsive self-interactions λ > λc , the quan-
tum pressure becomes negligible and the classical pres- Using the scaling behavior, we see that the radial size of
sure balances gravity. Here λc is the critical coupling at the star in the non-gravitational limit is R ∼ Mm|λ|
2 . The
which the quantum pressure and classical pressure are radius increases linearly with the mass of the star. Since
equally important: at larger radius R > R∗ we approach the non-interacting
limit in which the mass decreases for increasing radius,
1 2 M λc 3 we find that there is a maximum possible mass for bo-
∇ ψ= ψ . (10) 1
2m 4m3 son stars M max = M ∗ ∼ Mpl |λ|− 2 . However, it has be
We can use the scaling behavior introduced earlier to shown in [36], that the solutions for R < R∗ are unstable
2 with respect to perturbations. Therefore boson stars in
solve for the coupling and obtain λc ∼ Rm M . At the the non-gravitational limit cannot be realized in nature,
critical coupling, hydrostatic equilibrium requires R ∼
M2
at least for this simple class of interactions. For more
(GM m2 )−1 which implies λc ∼ Mpl2 . Already a very complicated interactions, such stars can exist, and fall
small coupling is sufficient for the Thomas-Fermi limit to under the general class of Q-balls [39–42].
apply.
For λ  λc we can rewrite Eq. 5 and 7 as
C. Scaling Invariance
M 2 2
4πGM ψ 2 = ∇2 Φ = ∇ ψ . (11)
4m4 Following the terminology of [21] we introduce the di-
Using that the gradient scales as ∇ψ ∼ ψ mensionless variables
R, we obtain that r
the radius of√a strongly self-interacting boson star scales e Φ πGN
M V = − , S= ψ,
as R ∼ mpl 2 λ. The radius is independent of the star’s 2m 2 2m (15)
mass. This is not surprising, since both gravity and the λ
repulsive self-interaction are proportional to the number x = 2rm, Λ= .
4πGm2
 4

We can then rewrite the Gross-Pitaevskii-Poisson equa- can be described via an (even) polynomial around the
tions given in Eq. 5 and 7 as center if the boson star z = 0
∞ ∞
∇2 V = −S 2 and ∇2 S = −V S + ΛS 3 . (16) X X
s= sn z n and v = vn z n . (21)
n=0 n=0
ψ 2 dV = 1
R
The wave function normalization condition
becomes Eq. 20 then leads to the recursion relations
Z ∞
n n X
m
x2 S 2 dx = GM m (17) X X
0 − sm vn−m + γ s` sm−` sn−m
m=0 m=0 `=0
where M = N m is the mass of the boson star. = (n + 2)(n + 3)sn+2 (22)
Let us note that Eq. 16 and 17 are invariant under the n
scaling
X
− sm sn−m = (n + 2)(n + 3)vn+2 .
m=0
x Λ
x→ , S → f 2 S, V → f 2 V, M → f M, Λ → 2 (18)
f f The smoothness of the profile at the origin implies
s1 = v1 = 0, and therefore, also all odd coefficients
where f is a scaling factor. This implies that we can s2n+1 , v2n+1 vanish. The profile at small radius z can
relate different solutions of the Gross-Pitaevskii-Poisson therefore be fully parametrized in terms of the wave func-
equations corresponding to different boson star masses tion and potential at the origin: s0 and v0 .
M and couplings Λ through rescaling. We will make At large radius, we can expand the profile using the
use of this scale invariance and solve for the Gross- series expansion
Pitaevskii-Poisson equations at a fixed reference scale k.
A particularly useful choice for our discussion is to set ∞,∞ n ∞,∞ n
e−z e−z
X  X 
e
−k 2 = V (∞) = 2m which transforms as k → f k. We s = snm z −m , v = vmn
z −m . (23)
can then introduce the scale invariant coordinate z, wave n,m=0,0
zσ n,m=0,0
zσ
function s, potential v, mass β and coupling γ via
By matching the coefficients, Eq. 20 we obtain the recur-
z = kx, S = k 2 s, V = k 2 v, sion relations
(19)
γ = k 2 Λ, GM m = 2kβ. n,m
X n−p
spq vm−q + n2 snm + 2n(nσ + m − 2) snm−1
Using the scale independent variables, we can write the p,q=0,0
Gross-Pitaevskii-Poisson equations as + (nσ + m − 2)(nσ + m − 3) snm−2 (24)
2 3 2 2 n,m p,q
∇ s = −sv + γs and ∇ v = −s . (20) X X p−r n−p
=γ srt sq−t sm−q
p,q=0,0 r,t=0,0
The scale choice implies the boundary condition v(∞) =
−1. The solution corresponding to a boson star with n,m
X
mass M can be obtained by performing the rescaling spq sn−p 2 n n
m−q + n vm + 2n(nσ + m − 2) vm−1
given in Eq. 19 with k = GM m
2β . In the following sec- p,q=0,0 (25)
tion, we will obtain an approximate analytical form for + (nσ + m − 2)(nσ + m − n
3) vm−2 = 0.
s, v.
Let us note the following properties of the solution: i)
Normalizability requires s00 = 0. Eq. 24 then implies that
III. WEAKLY COUPLED SYSTEMS AND all coefficients s0m vanish as well. This means that the
SERIES EXPANSION wave function decays at least exponentially. ii) Eq. 25
0
then implies that all vm = 0 for m > 1. This means
A. Series Expansion that at large radius, the potential is described by the
v0
Newtonian potential v (0) = v00 + z1 = −1 + 2β z . Here
We have seen in the discussion of the non-self- we both used the boundary condition imposed by our
interacting case, that we can describe the profile of the scale choice, v00 = −1, and used the notation introduced
boson star through an infinite series for the wave function in Eq. 19, v10 = 2β. All other terms in the expansion
and potential. of v are at least exponentially suppressed. iii) Setting
Following the same approach as for the non-self- n = m = 1, Eq. 24 can be written as 2β = v10 = 2(1 − σ).
interacting case [21], we will describe the profile at both This is a remarkable result: the exponent σ in the series
small and large radii through a series expansion of the expansion is related to the the total mass of the system
wave function and potential. At small radii, the profile σ = 1−β. iv) As derived in [21], the leading order, n = 1,

solution for the wave function at large radius is given by
the Whittaker function
α precision of the wave function needed for the numerical
s(1) = β Wβ,− 12 (2z). (26)
2 z
Here we have introduced the normalization parameter
α = s10 . v) The large radius solution can be fully
parametrized by the expansion parameters α = s10 and
β = v10 . The remaining coefficients can then be computed
using Eq. 24 and 25. Note however, that the series ex-
pansion in Eq. 23 does only converge for m < M , where
M is finite. vi) Eq. 24 and 25 further imply that the po-
n
tential contains only non-vanishing components vm for
even n while the wave function only has non-vanishing
components snm for odd n.
For practical purposes, we will truncate the infinite
series in Eq. 21 and 23 and only take into account the
leading terms with n ≤ N and m ≤ M . Let us define the
truncated series expansion at small and large radius z as
N N,M
e−z

(N )
X X
s(N ) = sn z n
and s(M ) = snm
n=0 n,m=0,0
zσ
B. Expansion Parameters
As we have seen in the previous section, the series
expansion for s and v at small and large radius can
be fully parametrized by the four expansion parameters
α, β, s0 , v0 . In the non-interacting case γ = 0 these were
just numbers, while in the general case they will be func-
tions of the coupling γ
α(γ), β(γ), s0 (γ), v0 (γ).
Following the strategy from [21], we first obtain the
expansion parameters from numerical simulations. In a
second step, we provide an analytic form for the expan-
sion parameter.
To obtain a numerical solution of the boson star’s pro-
file, it is convenient to solve the Gross-Pitaevskii-Poisson
equations as given in Eq. 16 using the boundary condition
V (0) = 1. The authors of [43] have shown that the solu-
tions of Eq. 16 can then be parametrized by the central
value of the wave function, S0 = S(0) and categorized
into three distinct classes: for S0 > S0∗ the wave function
diverges for at large radius towards positive infinity, for
S0 = S0∗ the wave function converges to zero, is posi-
tive definite and square integrable, while for S0 < S0∗ the
wave function diverges at large radius towards negative
infinity.
Using a Runge-Kutta 4 method with constant step size
∆x , we perform the numerical integration until the wave
5
function starts to diverge and iteratively optimize the
central value of the wave function S0 to find S0∗ . The
solution to stay finite until a large value of x, which is
needed to fit the large range solution, increases exponen-
tially with the radial coordinate x. The accuracy of the
numerical solution is limited by the step size ∆x . In this
study we use a precision of up to 150 significant figures
and ∆x = 10
−3 , providing an accuracy of the solution of
order O ∆4x ≈ 10−12 .
The Whittaker solution parametrization given in
Eq. 26 always describes the wave function profile at large
radius when the density is small enough that the self-
interaction becomes negligible. However, the central
mass parameter β and the normalization α depend on
the central profile of the star and therefore the coupling
parameter γ.
To obtain the expansion parameters α, β as well as the
scaling parameter k in Eq. 19, we fit the leading order
profiles, the Whittaker solution S(x) = 2kα
β x Wβ,− 1 (2kx)
2
n
and the Newtonian potential V (x) = −k 2 + 2kβ
x , to the
z −m . (27) numerical solution for V and S at large x. To avoid sys-
tematic effects due to the truncation of subleading terms
n > 1 of the series expansion in Eq. 23, we restrict the fit-
ting range to x > x∗ , where the fraction of mass outside
radius x∗ contributed less than 10−12 to the total mass
of the boson star. The expansion parameters at small
radius and the coupling are obtained via s0 = k −2 S0 ,
v0 = k −2 and γ = k 2 Λ.
In Fig. 1 we show the dependence of the expansion
parameters α (upper left), β (upper right), s0 (lower left)
and v0 (lower right) as function the coupling parameter
γ as red dots. Note that the horizontal axis switches
(28) from a linear scale to a logarithmic scale at γ = 1 as
indicated by the dashed black line. This indicated the
transition between the weakly coupled regime γ < 1 and
the strongly coupled regime γ > 1.
C. Fit for Weak Couplings: −1 < γ < 1
A vanishing self-coupling γ = 0 indicates the non-
interacting limit. The corresponding results for the ex-
pansion parameters are presented in [21]. If the self-
coupling is weak, |γ| < 1, we can treat the classical
pressure as perturbation. In this case we can write the
the expansion parameters as a series expansion in the
coupling γ around the non-interacting solution. To ob-
tain the coefficients of this expansion, we fit a 6th-degree
polynomial to the numerical solutions. We can write the
result in an analytic form as
 6

α(γ) = 3.495059 − 0.117682 γ − 0.391600 γ 2 + 0.191882 γ 3 − 0.041828 γ 4 − 0.041507 γ 5 + 0.033020 γ 6

β(γ) = 1.752717 + 0.703934 γ − 0.109101 γ 2 + 0.013436 γ 3 + 0.017778 γ 4 − 0.018281 γ 5 + 0.005129 γ 6
(29)
s0 (γ) = 1.021494 − 0.390946 γ + 0.171489 γ 2 − 0.064820 γ 3 + 0.004328 γ 4 + 0.028849 γ 5 − 0.017732 γ 6
v0 (γ) = 0.938204 + 0.102743 γ − 0.080310 γ 2 + 0.058708 γ 3 − 0.037703 γ 4 − 0.002557 γ 5 + 0.013512 γ 6 .

The result is shown in Fig. 1 as a green line. The lower
panels indicate the accuracy of the analytic form with
respect to the numerical solution. We can see that for
all four expansion parameters the solution from Eq. 29
reproduces the numerical results with accuracy better of
O(10−4 ) in the range −1 < γ < 1.
As mentioned before, boson stars with attractive self-
coupling and small radius R, or equivalently large neg-
ative coupling γ, become unstable with respect to per-
turbations. We show in Sec. V A, that this happens at
γmin = −0.722. The solutions for γ < γmin are unphysi-
cal.
star’s radius Z, the density drops until eventually the
self-interaction becomes negligible. The outer part of the
star can therefore be described by the non-interacting so-
lution, which at leading order can be written in terms of
the Whittaker function. In the following, we estimate
the remaining expansion parameter α by matching the
Thomas-Fermi solution for the interior of the star with
the Whittaker solution for the exterior part.
We define the matching point z ∗ as the radius where
the quantum and classical pressure terms become equal,
∇2 s(z ∗ ) = γs(z ∗ )3 . Using the Thomas-Fermi profile
from Eq. 31, we can write this condition as

z ∗2 z ∗2
 ∗
γs40 h
 
2z
4 3
1 + 3 − 1 + cos √
IV. STRONGLY COUPLED SYSTEMS 8z s γ γ γ
∗
 ∗i 3 32
 ∗ 3 (33)
2z 2z s0 γ z
A. Thomas Fermi Limit − √ sin √ = ∗3 sin √
γ γ z γ
In the previous section we have discussed the weakly √
We can write the matching point as z ∗ = (π − δ) γ and
coupled scalar field. Let us now consider the case of a expand the matching condition in δ. Keeping only the
large repulsive self-coupling γ  1. In this case the quan- 1
linear terms and using s0 = γ − 2 , we can solve for δ and
tum pressure becomes negligible and the classical pres- − 31
obtain δ = (4γ/π) . The wave function at z ∗ has a
sure balances gravity. This scenario is known as Thomas- value
Fermi limit and has been examined in [23, 44] We can
 δ  12
write the Gross-Pitaevskii-Poisson equations as 1 1
s(z ∗ ) = s0 sinc(π − δ) 2 ≈ s0 = (2πγ 2 )− 3 . (34)
π
γ∇2 v + v = 0 and v = γs2 . (30)
Following [45] (see Sec. 8, Eq. 18) we can write the
√
The (normalizable) solution for the profile is given by Whittaker function Wβ,− 21 (2z) for z = 2β = π γ as
√ √ 1
v = v0 sinc(z/ γ) and s = s0 [sinc(z/ γ)] 2 (31)  1  2z  61  β  
1
Wβ,− 21 (2z) ≈ Γ exp β log + (35)
3 6π 3 e 12β
where sinc(x) = sin(x)/x. The wave function becomes
√
zero at Z = π γ, implying that the boson star is com- Using the form of the exterior profile as given in Eq. 26
pact and has a radius Z. At z > Z the wave function and matching it to the Thomas-Fermi solution at z ∗ in
remains zero and the potential is described by the New- Eq. 34 allows us to extract the expansion parameter α.
tonian potential v(z) = −1 + 2β z . We find
Using that exterior and interior solution for the poten-

1 1
−1 − 1
3 6
tial of the star have to match at surface, v(Z) = 0, we 2β z ∗ s(z ∗ ) Γ 3 πγ 4
√ α= ≈ 4
can solve for the mass parameter β = 12 Z = π2 γ. We Wβ,− 21 (2z ) ∗
h 1 1 1 − 12
i (36)
exp π2 γ 2 log 4e
π 2


can further use the normalization condition (see Eq. 17) γ + 6π γ
√
π γ We have already seen that in the Thomas-Fermi limit the
√
Z Z
3
2 2
2β = s z dz = s20 sinc(z/ γ)z 2 dz = s20 πγ 2 (32) remaining expansion parameters are given by
0
π 1 1
1
to obtain the central density s0 = γ − 2 . Using Eq. 30, β= γ2, s0 = γ − 2 , v0 = 1 . (37)
2
this implies that the central value for the potential is
v0 = 1. We have therefore obtained a simple analytic form for all
expansion parameters in the limit of large self-couplings.
So far, we have ignored the effects of quantum pres- The results from Eq. 36 and 37 are shown in Fig. 1 as
sure to the boson star’s profile. However, close to the dashed cyan lines.
 7

50
3 TF-Limit
NI-Fit
2 20

(M)
β s(1)
TF-Fit
1
(M)
α s(1)
10
10-10
TF-Fit 5
10-20
-30 NI-Fit
10 TF-Limit
2
-40
10
Numerics 1 Numerics

10-1 10-1
(M)

(M)
(α-αnum)/αnum s(1)

(β-βnum)/βnum s(1)
-2
10 10-2
10-3 10-3
10-4 10-4
-5
10 10-5
10-6 10-6
-7
10 10-7
Linear Scale Logarithmic Scale Linear Scale Logarithmic Scale
1 1 1 1 2 3 1 1 1
-1 0 1 10 10 10 -1 0 1 101 102 103
γ γ
2
TF-Fit
1 1
TF-Limit
NI-Fit TF-Limit
0.5
(M)

(M)
s0 s(1)

v0 s(1)

NI-Fit
0.9
0.2

0.1
TF-Fit 0.8
0.05
Numerics Numerics

10-1 10-1
(M)

(M)
(s0-s0num)/s0num s(1)

(v0-v0num)/v0num s(1)

10-2 10-2
-3
10 10-3
10-4 10-4
-5
10 10-5
10-6 10-6
10-7 10-7
Linear Scale Logarithmic Scale Linear Scale Logarithmic Scale
1 1 1 1 2 3 1 1 1
-1 0 1 10 10 10 -1 0 1 101 102 103
γ γ

FIG. 1. The upper panels show the expansion parameters α (upper left), β (upper right), s0 (lower left) and v0 (lower right) as
function of the coupling parameter γ. The numerical solution is shown as red dots in the upper panels. The analytic fits to the
numerical results are shown in green for the Non-Interacting regime as discussed in Sec. III C and in blue for the Thomas-Fermi
regime as discussed in Sec. IV B. The results corresponding to the Thomas-Fermi limit discussed in Sec. IV A are shown in cyan
for comparison. The lower panels show the accuracy of the analytic solutions with respect to the numerical solution.

Note that for increasing self-coupling γ, the expansion B. Fit for Strong Couplings: γ > 1
parameter α exponentially decreases leading to a sharp
drop of the wave function profile at z = Z. The bosons
become more and more confined in the inner part while In the above discussion of the Thomas-Fermi limit, we
the large radius tails of the wave function vanish. have neglected the effect of the quantum pressure on the
structure of the star. For large but finite self-couplings γ,
 8

we can consider the quantum pressure as a perturbation coupling parameter. The coefficients of this expansion
to the Thomas-Fermi limit. We include this perturba- are obtained from a fit to the numerical solution. We
tion as a correction factor to the Thomas-Fermi solution, can write the result in an analytic form as
which we can express asseries expansion in the (inverse)

1 2 3 4 5
α(γ) = αT F 0.603380 + 0.485970γ − 6 − 4.422475γ − 6 + 8.719758γ − 6 − 8.363927γ − 6 + 4.397913γ − 6 − 1.001027γ −1
 
1 2 4 5
β(γ) = β T F 1 − 0.001478 γ − 3 + 0.045642 γ − 3 + 0.823049 γ −1 − 0.590994 γ − 3 + 0.347840 γ − 3 − 0.118132 γ −2
 
1 3 5
s0 (γ) = sT0 F 1 + 0.003712 γ − 2 − 0.067139 γ −1 − 0.436976 γ − 2 − 0.107433 γ −2 + 0.687868 γ − 2 − 0.327405 γ −3 (38)
 
1 3 5
v0 (γ) = v0T F 1 + 0.008062 γ − 2 + 0.388054 γ −1 − 1.245466 γ − 2 + 1.486280 γ −2 − 0.823199 γ − 2 + 0.178605 γ −3
 

with the leading order Thomas-Fermi solutions

3
16 1
−1 − 1 h π 1 π 1
1 −1 i π 1 1
αT F = Γ πγ 4 exp − γ 2 log γ2 − γ 2 , βT F = γ2, sT0 F = γ − 2 , v0T F = 1 . (39)
4 3 2 4e 6π 2

Note that the correction factor for the expansion param-
eter α contains a non-unity constant term. This should
not be surprising, considering that we obtained α in the
Thomas-Fermi limit by matching the Thomas-Fermi so-
lution and the Whittaker-solution at a matching point
z ∗ , even though both solutions describe the profile only
poorly at this point.
The results are shown in Fig. 1 as a blue line. We
can see that for all four expansion parameters the solu-
tion from Eq. 38 reproduces the numerical results with
accuracy of O(10−5 ) for self-couplings γ > 1.
Using the analytic expression for the expansion param-
eters, we can now obtain the profile of the boson star for
a given value of the self-coupling γ. For very large self-
couplings, γ & 10, we can directly use the Thomas-Fermi
and Whittaker solution to describe the central and outer
profile of the star. This is shown in Fig. 2 for γ = 10
(red), 100 (blue) and 1000 (green). The upper panel
shows the wave function profile s(z) normalized by its
central value s0 . The numerical solution is shown in gray,
the Thomas-Fermi solution for the central profile given
by Eq. 31 as dashed lines and the Whittaker solution
given by Eq. 26 as solid lines. The lower panel shows
the accuracy of the approximate solutions with respect
to the numerical solutions.
We can see that for large couplings, a combination of
the Thomas-Fermi and Whittaker solutions describe pro-
file with an accuracy of O(10−6 ) in the center and outside
the star, corresponding to the accuracy of the expansion
parameters. However, even for large couplings, there re-
mains a surface region around z ≈ Z that is poorly de-
scribed by both the Thomas-Fermi and Whittaker solu-
tions. Even though the wave function, and therefore the
mass density, is small in the surface region, the fraction
of the star’s mass contained in it can still be sizeable.
To obtain a better description, in particular for the
surface region, we can use the series expansion derived
in Sec. III A. This is shown in Fig. 3 for an intermediate
sized self-coupling γ = 10. The upper panel shows the
wave function profile s(z). Besides the numerical solu-
tion (gray), the Thomas-Fermi solution (magenta) and
the Whittaker solution (red), we show the truncated se-
ries expansion as given in Eq. 27 with N = 20 for the
inner solution (cyan) and N = 3, M = 8 for the outer
solution (blue). The corresponding accuracy of the solu-
tion is shown in the lower panel. We can see that includ-
ing higher order terms in the truncated series expansion
increases the accuracy of the solution by one order of
magnitude in the surface region and up to four orders
of magnitude in the center and outside star. For com-
parison we also show the truncated series expansion with
N = 50 for the inner part and N = 5, M = 10 in the
outer part in green. The corresponding accuracy in the
surface region is better than 10−3 .
V. APPLICATIONS
A. Using the Solution
In the above discussion, we have solved the Gross-
Pitaevskii-Poisson equations in Eq. 20, expressed in
terms of the dimensionless variable s and v. We have seen
that the solution of these equations is fully parametrized
by the dimensionless coupling parameter γ and then ob-
tained an analytic form for profile of boson stars. To use
the solution to describe a boson star with mass M made
of boson with mass m and self-coupling λ, we therefore
need to obtain the corresponding value of the dimension-
less coupling parameter γ.
Combining our scale choice k = GM 2β
m
with Eq. 15 and
19, we can see that
GM 2 λ
γ = k2 Λ = . (40)
16πβ 2
 9

1 Note that the value of the self-coupling parameter γ is

independent of the boson mass m and only depends on
0.8 the star’s mass M and the quartic self-coupling λ. We
γ=10 γ=100 γ=1000
can obtain the coupling parameter γ by solving
(M)
s(1)

0.6
s/s0

0.4
16πγβ(γ)2 = λGM 2 ≡ ξ (41)
0.2

0
for γ as a function of ξ where we have introduced the
short-hand notation ξ = λGM 2 . Alternatively, we can
follow our previous approach used to obtain the expan-
(M)
(s-snum)/snum s(1)

10-2 sion coefficients and obtain an expression for γ by finding

a suitable fit to the numerical solution. This is shown in
Fig. 4. The red dots correspond to the numerical solution
10-4 using the same data set as Fig. 1.

In the Thomas-Fermi limit, the expansion parameter

10-6 √
β is given by β = π2 γ. We can solve Eq. 41 and obtain

2 5 10 20 50 100 200
z
r
TF ξ (42)
FIG. 2. The upper panel shows the wave function profile s(z) γ =
4π 3
for the couplings γ = 10 (left), 100 (center) and 1000 (right).
We show the numerical solution (gray), the Thomas-Fermi
approximation (dashed) for the central profile (see Eq. 31)
and the Whittaker approximation (solid) for the outer pro- as shown in Fig. 4 as dashed cyan curve. To obtain a more
file (see Eq. 26). The lower panel shows the accuracy of the accurate result, we follow our approach used to obtain the
approximate solutions with respect to the numerical solution, expansion parameters and fit the numerical solution with
(s − snum )/snum , as indicated by the colored dashed lines. We a suitable series expansion. For large values of ξ > 100
use the expansion parameters according to Eq. 38. we expand around the Thomas-Fermi limit and obtain

1 1 3 5 3
γ(ξ) = γ T F 1 − 0.035941 ξ − 4 − 9.569558 ξ − 2 + 31.89268 ξ − 4 − 120.9668 ξ −1 + 316.0673 ξ − 4 − 308.2427 ξ − 2 (43)


For small ξ < 100 we find

  12    3
ξ − ξmin ξ − ξmin ξ − ξmin 2
γ(ξ) = −0.720960 + 1.157002 − 0.400828 + 0.420862
100 100 100
 2   52  3 (44)
ξ − ξmin ξ − ξmin ξ − ξmin
− 0.299337 + 0.125788 − 0.023160
100 100 100

with ξmin = −51.523602. Note that domain of ξ is re-
stricted to ξ > ξmin , and therefore, the coupling pa-
rameter γ is bound from below γ > γmin = −0.722.
Physical boson stars therefore need to fulfill the condi-
tion λGM 2 > ξmin . This implies that the maximal mass
of a boson star with attractive self-interaction, λ < 0, is
given by
r with accuracy better of O(10−4 ) for all physical values
λ<0 ξmin (45)
Mmax = Mpl .
λ
There exists a second branch of the solution correspond-
ing γ < γmin . However, the corresponding boson star
would have a higher total energy than the solution for
γ > γmin . Such configuration is unstable with respect to
perturbations and therefore unphysical [36].
The accuracy of the expressions given in Eq. 44
and Eq. 43 with respect to the numerical solution is
shown in the lower panel of Fig. 4. We can see that
the analytic expressions reproduce the numerical results
of ξ.
Knowing the value of γ for a given boson star, we can
obtain the profile for physical wave function ψ and grav-
 10

0.3 103
Thomas-Fermi γ=10
102
N=20
0.2

(M)
101

γ s(1)
(M) s(z)

Numerical
Solution 11
s(1)

0.1
Whittaker 01
N=3,M=8 Numerics
0 -11

mi W 10-1
Fer hi

(M)
as-
(M)

(γ-γnum)/γnum s(1)
tta 10-2
(s-snum)/snum s(1)

10-2 m
Tho ke
r 10-3
10-4
N=

10-4
N=
20

3,M

10-5
N=

5,M
N=50

=8
=1

10-6 10-6
0

10-7
Linear Scale Logarithmic Scale
1 1 1
0 5 10 15 20 -100 0 100 104 106 108
z ξ=16πγβ2=λGM2

FIG. 3. The upper panel shows the numerical solution (gray),
the Thomas-Fermi approximation as given in Eq. 31 (magenta
dashed), the Whittaker approximation as given in Eq. 26 (red
dashed), the truncated series expansion of the wave function
at small radius with N = 20 (cyan dotted) and with at large
radius with N = 3, M = 8 (blue dotted) as given in Eq. 27.
The lower panel shows the accuracy of the approximate solu-
tions with respect to the numerical solution, (s − snum )/snum .
We use the expansion parameters according to Eq. 38.
FIG. 4. The upper panels show the coupling parameter γ as
function of ξ = λGM 2 . The numerical solution is shown as
red dots in the upper panels. The analytic fit to the numerical
results is shown in green for the weak coupling regime (see
Eq. 44), in blue for the strong coupling regime (see Eq. 43)
and in cyan for the Thomas-Fermi limit. The lower panel
shows the accuracy of the analytic solutions with respect to
the numerical solution.

was initially introduced to solve the strong-CP problem.

itational potential Φ by rescaling the dimensionless solu- Furthermore it also provides a natural dark matter can-
tion s and v according to Eq. 15 and 19. We can write didate if its mass in the range between m = 10−5 eV
and 10−2 eV. The axion potential can heuristically be
described by the instanton potential
3 3
r
1 G 2 M 2 m3 GM m2
    
ψ(r) = s r a 1 λ
8π β2 β V (a) = m2π fπ2 1 − cos ≈ m2 a2 + a4 (49)
(46) f 2 12
2 2 2
GM m2
  
G M m
Φ(r) = − 1 + v r where a denotes the axion field, mπ = 135 MeV is the
2β 2 β pion mass, fπ = 92 MeV is the pion decay constant and
Furthermore, we can simply read off the binding energy f is the axion decay constant. After expanding the po-
tential and matching it to the form of Eq. 1, we find
G2 M 2 m3 that the axion mass and decay constant are related by
e = −2mk 2 = − (47) mf = mπ fπ . The quartic coupling is given by λ = − 2f m2
2β 2 2

where the negative sign indicates that the self-interaction

and the central density is attractive. Note that there exist higher terms in the
expansion. However, those terms will not contribute in
1
ρ0 = M ψ(0)2 = G3 M 4 m6 s20 . (48) the non-relativistic approximation and have been shown
8πβ 4 to be insignificant for dilute axions stars even in the rel-
ativistic limit [22].
In the previous section, we have seen that a boson star
B. Axion Stars
with attractive self-interaction has an upper mass. Using
the axion parameters this implies
In the following, we illustrate the use of the obtained  −5 2
solution on one particularly well-motivated scenario: ax- mπ fπ −12 10 eV
M < 10.14 1 = 2.74 · 10 M . (50)
ion stars. The axion is a real pseudo-scalar field, which G 2 m2 m
 11

10-9 5

(M)
ρ(r) [1026 GeV/cm3] s(1)
m=10-5 eV
it
Lim 4 M=2.5·10-12 M⊙
λ=0 TF 2
3 λ<0 |λ|= m 2 =3.2·10-53
2f
10-10
λ>0 2

1
e [m]

λ=0
10 -11 λ>0
λ<0 0

(M)
dM/dr [10-13 M⊙ /km] s(1)
λ<0
1.5
-12
10 λ=0
m=10-5 eV 1
2
|λ|= m 2 =3.2·10-53 λ>0
Mmax 2f
10-13 0.5
10-12 10-11 10-10
M [M⊙ ]
0
FIG. 5. Binding energy e as a function of the axion star’s 0 10 20 30 40
r [km]
mass M for an axion mass m = 10−5 eV. We show the results
for axion-star without self-interaction (green), with attrac-
tive self-interaction (red) and a wrong-sign axion-star with a FIG. 6. The upper panel shows the energy density profile ρ for
repulsive self-interaction (blue). The cyan line indicates the an axion star with mass M = 2.5 · 10−12 M and axion mass
Thomas-Fermi limit. The vertical line indicates the upper m = 10−5 eV. We show the results for an axion-star without
bound on the star’s mass Mmax in the case of an attractive self-interaction (green), with attractive self-interaction (red)
self-interaction. and a wrong-sign axion-star with a repulsive self-interaction
(blue). We use the truncated series expansion of the wave
function at small radius with N = 20 (dashed) and with at
We have seen in Sec. II B that the non-relativistic ap- large radius with N = 3, M = 2 (solid) as given in Eq. 27.
The lower panels shows the corresponding differential mass
proximation is valid for axion stars with mass
distributions dM/dr.
 −5 
1 10 eV
M = 5 · 10−7 M . (51)
Gm m
axions with mass m = 10−5 eV will form axion miniclus-
For axions in the dark matter axion window of masses ters with masses around M ∼ 10−12 M and therefore in
between m = 10−5 eV and 10−2 eV, the upper bound the regime in which self-interactions are important.
on the axion star’s mass is well below this limit, and For comparison, we also show the results for a wrong-
therefore, the axion star is well described by the non- sign axion star, in which the self-coupling has the same
m2
relativistic approximation. magnitude as the axion field but is repulsive, λ = + 2f 2 =

In Fig. 5 we show the binding energy e given in Eq. 47
as a function of the axion star’s mass M for a axion
mass m = 10−5 eV. The green line shows the result
for a non-interacting axion field, λ = 0. In this case
the binding energy is given by e = −0.1627 · G2 M 2 m3 ,
which increases quadratically with the star’s mass. The
red lines indicates the axion field including the attrac-
m2 −53
tive self-interaction, λ = − 2f 2 = −3.2 · 10 . For small
−12
axion star masses M . 10 M , it approaches the non-
interacting limit. In this case the classical pressure due
to the self-interaction is negligible since the axion den-
sity is small even in the center of the star. For masses
M > 10−12 M the self-interaction term becomes im-
portant and the binding energy increases relative to the
non-interacting case. The upper bound on the axion mass
star mass, Mmax = 2.74 · 10−12 M , is indicated by the
gray dashed line. The authors of [46] have shown that
+3.2 · 10−53 , as a blue line. At high masses, the binding
energy approaches the Thomas-Fermi limit, shown as a
dashed cyan line. In this case the binding energy is given
3
by e = − √4λπ G 2 M m3 .
In the upper panel of Fig. 6 we show the density pro-
file, ρ(r) = M ψ 2 (r). For illustration, we choose an
axion mass m = 10−5 eV and axion star mass M =
2.5 · 1012 M which is slightly below the maximal mass
Mmax . Following the previous discussion, we consider an
axion-star without self-interaction (green), including the
attractive self-interaction (red) and a wrong-sign axion-
star with a repulsive self-interaction (blue). To calculate
the density profile, we use the truncated series expansion
as given in Eq. 27 with N = 20 at small radius, as indi-
cated by the dashed lines, and N = 3, M = 2 at large
radius as indicated by the solid lines. We can see that the
two series expansions match well at intermediate values

of the radius. Note that the attractive self-interaction
of the axion field leads to a significant deformation of
the density profile with respect to the a non-interaction
axion star of same mass. This indicates the importance
of including the axion self-interaction when considering
axions stars close to their maximal mass M ∼ Mmax .
The lower panel of Fig. 6 shows the differential mass
distribution dM/dr = 4πr2 ρ for the three considered
stars. We can see that the mass distribution peaks at
intermediate radii and approximately half of the star’s
mass is described by both the small and large radius ex-
pansion of the wave function as given in Eq. 27. This
once again shows the importance of an accurate descrip-
tions of the tails of the wave function profile in order to
correctly describe the axion star.
VI. CONCLUSION
Light scalar fields can form gravitationally bound com-
pact objects, called boson stars. In the Newtonian limit,
the profiles of boson stars are described by the Gross-
Pitaevskii-Poisson equations. In a previous study [21],
we presented a semi-analytic solution to these equations
describing the profile of boson stars formed by a non-
interacting scalar field. The solution is based on a series
expansion which is parametrized by four expansion pa-
rameters that have been obtained from numerical simu-
lation at high accuracy.
In this study, we have generalized our semi-analytic
approach to boson stars where the constituent particles
have self-interactions. In this case, the expansion param-
[1] J. E. Kim and G. Carosi, “Axions and the Strong CP
Problem,” Rev. Mod. Phys. 82 (2010) 557–602,
arXiv:0807.3125 [hep-ph].
[2] H.-Y. Cheng, “The Strong CP Problem Revisited,”
Phys. Rept. 158 (1988) 1.
[3] P. Sikivie, “Experimental tests of the ’invisible’ axion,”
Physical Review Letters 51 (1983) 1415–1417.
[4] P. Sikivie, “Detection rates for “invisible”-axion
searches,” Phys. Rev. D 32 (1985) 2988–2991.
[5] S. J. Asztalos, G. Carosi, C. Hagmann, D. Kinion,
K. van Bibber, M. Hotz, L. J. Rosenberg, G. Rybka,
J. Hoskins, J. Hwang, P. Sikivie, D. B. Tanner,
R. Bradley, J. Clarke, and ADMX Collaboration,
“SQUID-Based Microwave Cavity Search for
Dark-Matter Axions,” Physical Review Letters 104
(2010) no. 4, 041301, arXiv:0910.5914 [astro-ph.CO].
[6] J. Hoskins, J. Hwang, C. Martin, P. Sikivie, N. S.
Sullivan, D. B. Tanner, M. Hotz, L. J. Rosenberg,
G. Rybka, A. Wagner, S. J. Asztalos, G. Carosi,
C. Hagmann, D. Kinion, K. van Bibber, R. Bradley,
and J. Clarke, “Search for nonvirialized axionic dark
matter,” Phys. Rev. D 84 (2011) no. 12, 121302,
arXiv:1109.4128.
12
eters are functions of the quartic self-coupling. Based
on results from numerical simulations, we found a cor-
responding analytic expression for all expansion param-
eters.
This allows to simply obtain profiles of boson stars in
an analytic form for arbitrary self-couplings at high pre-
cision directly from the series expansion. In particular,
no further time consuming and computational expansive
numerical integration is needed.
We have also applied our methods to axion stars, and
shown how the the mass and density profiles can be
obtained for both weak and strong interactions. The
profiles are significantly modified from the case of non-
interacting bosons. The methods developed in this paper
allows for systematic studies of the properties of boson
stars in an analytic way without further relying on nu-
merical simulations.
Finally, we note that there are several possible gener-
alizations of these results. In particular, we can extend
our results to rotating boson stars. It would also be
interesting to see how the profiles are modified in the
presence of other astrophysical objects like planets. We
leave these and other questions to future work.
ACKNOWLEDGMENTS
We thank Joshua Eby and William Shepherd for useful
discussions. This work is supported by NSF under Grant
No. PHY-1620638. F.K. performed part of this work at
the Aspen Center for Physics, which is supported by NSF
Grant No. PHY-1607611.
[7] G. Rybka, A. Wagner, A. Brill, K. Ramos, R. Percival,
and K. Patel, “Search for dark matter axions with the
Orpheus experiment,” Phys. Rev. D91 (2015) no. 1,
011701, arXiv:1403.3121 [physics.ins-det].
[8] P. Sikivie, N. Sullivan, and D. B. Tanner, “Proposal for
Axion Dark Matter Detection Using an LC Circuit,”
Phys. Rev. Lett. 112 (2014) no. 13, 131301,
arXiv:1310.8545 [hep-ph].
[9] O. K. Baker, M. Betz, F. Caspers, J. Jaeckel,
A. Lindner, A. Ringwald, Y. Semertzidis, P. Sikivie,
and K. Zioutas, “Prospects for searching axionlike
particle dark matter with dipole, toroidal, and wiggler
magnets,” Phys. Rev. D 85 (2012) no. 3, 035018,
arXiv:1110.2180 [physics.ins-det].
[10] D. Horns, J. Jaeckel, A. Lindner, A. Lobanov,
J. Redondo, and A. Ringwald, “Searching for WISPy
Cold Dark Matter with a Dish Antenna,” JCAP 1304
(2013) 016, arXiv:1212.2970 [hep-ph].
[11] D. Budker, P. W. Graham, M. Ledbetter, S. Rajendran,
and A. Sushkov, “Proposal for a Cosmic Axion Spin
Precession Experiment (CASPEr),” Phys. Rev. X4
(2014) no. 2, 021030, arXiv:1306.6089 [hep-ph].
[12] P. W. Graham and S. Rajendran, “New Observables for
Direct Detection of Axion Dark Matter,” Phys. Rev.

D88 (2013) 035023, arXiv:1306.6088 [hep-ph].
[13] K. Zioutas, C. E. Aalseth, D. Abriola, I. F. T. A., R. L.
Brodzinski, J. I. Collar, R. Creswick, D. E. D. Gregorio,
H. Farach, A. O. Gattone, C. K. Guérard,
F. Hasenbalg, M. Hasinoff, H. Huck, A. Liolios, H. S.
Miley, A. Morales, J. Morales, D. Nikas, S. Nussinov,
A. Ortiz, E. Savvidis, S. Scopel, P. Sievers, J. A. Villar,
and L. Walckiers, “A decommissioned LHC model
magnet as an axion telescope,” Nuclear Instruments
and Methods in Physics Research A 425 (1999)
480–487, astro-ph/9801176.
[14] Y. V. Stadnik and V. V. Flambaum, “Axion-induced
effects in atoms, molecules, and nuclei: Parity
nonconservation, anapole moments, electric dipole
moments, and spin-gravity and spin-axion momentum
couplings,” Phys. Rev. D89 (2014) no. 4, 043522,
arXiv:1312.6667 [hep-ph].
[15] J. L. Feng, I. Galon, F. Kling, and S. Trojanowski,
“Alps at faser: The lhc as a photon beam dump,”
1806.02348. https://arxiv.org/pdf/1806.02348.
[16] S. L. Liebling and C. Palenzuela, “Dynamical Boson
Stars,” Living Rev. Rel. 15 (2012) 6, arXiv:1202.5809
[gr-qc].
[17] D. J. Kaup, “Klein-Gordon Geon,” Physical Review
(1968) 1331–1342.
[18] R. Ruffini and S. Bonazzola, “Systems of selfgravitating
particles in general relativity and the concept of an
equation of state,” Phys. Rev. 187 (1969) 1767–1783.
[19] S. Carignano, L. Lepori, A. Mammarella,
M. Mannarelli, and G. Pagliaroli, “Scrutinizing the pion
condensed phase,” Eur. Phys. J. A53 (2017) no. 2, 35,
arXiv:1610.06097 [hep-ph].
[20] D. F. Jackson Kimball, D. Budker, J. Eby, M. Pospelov,
S. Pustelny, T. Scholtes, Y. V. Stadnik, A. Weis, and
A. Wickenbrock, “Searching for axion stars and Q-balls
with a terrestrial magnetometer network,”
arXiv:1710.04323 [physics.atom-ph].
[21] F. Kling and A. Rajaraman, “Towards an Analytic
Construction of the Wavefunction of Boson Stars,”
arXiv:1706.04272 [hep-th].
[22] J. Barranco and A. Bernal, “Self-gravitating system
made of axions,” Phys. Rev. D83 (2011) 043525,
arXiv:1001.1769 [astro-ph.CO].
[23] C. G. Böhmer and T. Harko, “Can dark matter be a
Bose Einstein condensate?,” JCAP 6 (2007) 025,
arXiv:0705.4158.
[24] M. Membrado, A. F. Pacheco, and J. Sañudo, “Hartree
solutions for the self-Yukawian boson sphere,” Phys.
Rev. A 39 (1989) 4207–4211.
[25] J.-W. Lee and I.-G. Koh, “Galactic halos as boson
stars,” Phys. Rev. D 53 (1996) 2236–2239,
hep-ph/9507385.
[26] J. Goodman, “Repulsive dark matter,” New Astron. 5
(2000) 103–107, astro-ph/0003018.
[27] A. Arbey, J. Lesgourgues, and P. Salati, “Galactic halos
of fluid dark matter,” Phys. Rev. D 68 (2003) no. 2,
023511, astro-ph/0301533.
[28] P. H. Chavanis and L. Delfini, “Mass-radius relation of
Newtonian self-gravitating Bose-Einstein condensates
with short-range interactions: II. Numerical results,”
Phys. Rev. D84 (2011) 043532, arXiv:1103.2054
[astro-ph.CO].
13
[29] J. Eby, P. Suranyi, C. Vaz, and L. C. R. Wijewardhana,
“Axion Stars in the Infrared Limit,” JHEP 03 (2015)
080, arXiv:1412.3430 [hep-th]. [Erratum:
JHEP11,134(2016)].
[30] J. Eby, C. Kouvaris, N. G. Nielsen, and L. C. R.
Wijewardhana, “Boson Stars from Self-Interacting Dark
Matter,” JHEP 02 (2016) 028, arXiv:1511.04474
[hep-ph].
[31] Q. Yang and H. Di, “Axion-like particle dark matter in
the linear regime of structure formation,” Int. J. Mod.
Phys. A32 (2017) no. 0, 1750051, arXiv:1507.04672
[hep-ph].
[32] N. Kan and K. Shiraishi, “A Newtonian Analysis of
Multi-scalar Boson Stars with Large Self-couplings,”
arXiv:1706.00547 [hep-th].
[33] J. Eby, P. Suranyi, and L. C. R. Wijewardhana,
“Expansion in Higher Harmonics of Boson Stars using a
Generalized Ruffini-Bonazzola Approach, Part 1:
Bound States,” arXiv:1712.04941 [hep-ph].
[34] L. Visinelli, S. Baum, J. Redondo, K. Freese, and
F. Wilczek, “Dilute and dense axion stars,” Phys. Lett.
B777 (2018) 64–72, arXiv:1710.08910 [astro-ph.CO].
[35] P.-H. Chavanis, “Phase transitions between dilute and
dense axion stars,” arXiv:1710.06268 [gr-qc].
[36] P.-H. Chavanis, “Mass-radius relation of Newtonian
self-gravitating Bose-Einstein condensates with
short-range interactions: I. Analytical results,” Phys.
Rev. D84 (2011) 043531, arXiv:1103.2050
[astro-ph.CO].
[37] E. D. Schiappacasse and M. P. Hertzberg, “Analysis of
Dark Matter Axion Clumps with Spherical Symmetry,”
arXiv:1710.04729 [hep-ph].
[38] E. Braaten, A. Mohapatra, and H. Zhang, “Dense
Axion Stars,” Phys. Rev. Lett. 117 (2016) no. 12,
121801, arXiv:1512.00108 [hep-ph].
[39] A. Kusenko, “Solitons in the supersymmetric extensions
of the standard model,” Phys. Lett. B405 (1997) 108,
arXiv:hep-ph/9704273 [hep-ph].
[40] A. Kusenko, “Q balls in the MSSM,” Nucl. Phys. Proc.
Suppl. 62A/C (1998) 248, arXiv:hep-ph/9707306
[hep-ph].
[41] B. Kleihaus, J. Kunz, and M. List, “Rotating boson
stars and Q-balls,” Phys. Rev. D72 (2005) 064002,
arXiv:gr-qc/0505143 [gr-qc].
[42] B. Hartmann and J. Riedel, “Supersymmetric Q-balls
and boson stars in (d+1) dimensions,” Phys. Rev. D87
(2013) no. 4, 044003, arXiv:1210.0096 [hep-th].
[43] P. Tod and I. M. Moroz, “An analytical approach to the
schrödinger-newton equations,” Nonlinearity 12 (1999)
no. 2, 201.
[44] Z. Slepian and J. Goodman, “Ruling Out Bosonic
Repulsive Dark Matter in Thermal Equilibrium,” Mon.
Not. Roy. Astron. Soc. 427 (2012) 839,
arXiv:1109.3844 [astro-ph.CO].
[45] H. Buchholz, The confluent hypergeometric function
with special emphasis on its applications. Springer
tracts in natural philosophy. Springer, 1969.
http://www.springer.com/us/book/9783642883989.
[46] J. Enander, A. Pargner, and T. Schwetz, “Axion
minicluster power spectrum and mass function,” JCAP
1712 (2017) no. 12, 038, arXiv:1708.04466
[astro-ph.CO].