1.4 Where ould axions and ALPs hide?
2.1.1 Realignment mehanism
In the realignment mehanism a eld, whih in the early universe an take a
random initial state, rolls down towards the minimum of the potential. One it
has reahed the bottom,itovershootsthe minimumand startstoosillatearound
it. Ifthe quantaof the eld are osmologiallystable, theseosillationsbehaveas
a old darkmatter uid. Their energy density infat is dilutedby the expansion
of the universe as
ρ ∝ R −3
.The simplest exampleis that of a salar eld
φ
of massm φ
with LagrangianL = 1
2 ∂ µ φ∂ µ φ − 1
2 m 2 φ φ 2 + L I ,
(2.12)where
L I
enodes interationsofthesalareldwith itselfandtherestofpartiles inthe primordialbath. In general, the mass reeivesthermal orretions fromL I
whih might be ruial,thus
m φ = m φ (t)
should be understood. In eah ausally onnetedpathofthe universe thesalar eldhas aninitialvalue,φ i
. Ifination0 0.5 1 5 10 -0.5
0.0 0.5 1.0
Log 10 Htt 1 L Log 10 Hm Φ HL
ΦΦ i
Figure 2.2: Numerial solution of equation (2.13) (solid) and theevolution of
log 10 (m φ /H)
(dashed) for the valuesused inthe equation.alreadyhappened,
φ
isuniformlyequaltoφ i
inthewholeobservableuniverse. Theequation of motion in the expanding universe for the homogeneousomponent of
φ
alled the zero mode isobtained negleting the gradient eets,φ ¨ + 3H φ ˙ + m 2 φ φ = 0 .
(2.13)Its solutionanbeseparatedintotworegimes. Inarst epoh,
3H ≫ m φ
,soφ
isanoverdamped osillatorand getsfrozen,
φ ˙ = 0
. Atalatertime,t 1
,haraterized by3H(t 1 ) = m φ (t 1 ) ≡ m 1
, the damping beomes underritial and the eld an roll down the potential and starts to osillate. During this epoh, the mass termis the leading sale in the equation and the solution an be found in the WKB
approximation,
φ ≃ φ i
m 1 R 3 1 m φ R 3
1/2
cos Z t
t 1
m φ dt
,
(2.14)where
φ(t 1 ) ∼ φ i
, sine up tot 1
the eld evolution is frozen. Notethat toobtainthis solution only the denition of
H = ˙ R/R
has been taken into aount, notits atual time dependene, and so it is valid for radiation, matter, and vauum
energy dominatedphases ofthe universe and theirtransitions. Figure2.2shows a
The approximate solution (2.14) orresponds to fast osillations with a slow
amplitude deay. Dening this amplitude
A (t) = φ i (m 1 R 3 1 /m φ R 3 ) 1/2
and thephase
α(t) = R t
m φ (t)dt
, the energy density of the salar eld isρ φ = 1
2 φ ˙ 2 + 1
2 m 2 φ φ 2 = 1
2 m 2 φ A 2 + ... ,
(2.15)where the dots stand for terms involving derivatives of
A
, whih by assumptionare muh smallerthan
m φ
, beausem φ ≫ H
in this regime. The pressure isthenp φ = 1
2 φ ˙ 2 − 1
2 m 2 φ φ 2 = − 1
2 m 2 φ A 2 cos (2α) − A A ˙ m φ sin (2α) + ˙ A 2 cos 2 (α) .
(2.16)When the eld just starts to osillate the equation of state is a non-trivial and
strongly time dependent funtion. However, at muh later times,
t ≫ t 1
, theosillationsin the pressure our at time sales
1/m φ
, muh faster than theos-mologialevolution. Wean thereforetakeanaverage overthese osillations,and
the pressure is then
h p φ i = h A ˙ 2 cos 2 (α) i = 1
2 A ˙ 2 .
(2.17)At leading orderin
A ˙ /( A m)
, the equation of state is justw = h p i / h ρ i ≃ 0 ,
(2.18)whih is exatlythat of non-relativisti matter.
It follows fromequation (2.14) that the energy density ina omovingvolume,
ρ φ R 3
,is not onserved if the salar mass hanges intime. However, the quantityN φ = ρ φ R 3
m φ
= 1
2 m 1 R 1 3 φ 2 i ,
(2.19)isonstant,andanbeinterpretedasaomovingnumberofnon-relativistiquanta
ofmass
m φ
. Weanuse this onservationtoompute the energydensitytodayto obtainρ φ (t 0 ) = m 0
N φ
R 3 0 ≃ 1
2 m 0 m 1 φ 2 i R 1
R 0
3
,
(2.20)wherequantities with a
0
-subsript are evaluated atpresent time.Through the onservation of omovingentropy
S
, the dilution fator(R 1 /R) 3
inequation (2.20) an be expressed as
R 1
R 3
= g ∗S (T )T 3
g ∗S (T 1 )T 1 3 .
(2.21)10 -5 0.001 0.1 10 1000 0.2
0.4 0.6 0.8 1.0
T @GeVD
F H T L
Figure 2.3: Thefuntion
F (T )
,usedinequations (2.22) .Using the expression for the Hubble onstant in the radiation dominated era,
equation (2.6), and the denition of
T 1
,3H(T 1 ) = m 1
,T 1
an bewritten intermsof
m 1
andthePlankmass. Inthisway,equation(2.20)forthematterdensityfromthe misalignementmehanismtodayan thenbeexpressed inamore quantitative
way as
ρ φ,0 ≃
0.17 keV cm 3
r m 0
eV r m 0
m 1
φ i
10 11 GeV 2
F (T 1 ) ,
(2.22a)Ω φ ≃ 0.016 h −2
r m 0
eV r m 0
m 1
φ i
10 11 GeV 2
F (T 1 ) ,
(2.22b)where
F (T 1 ) ≡ (g ∗ (T 1 )/3.36) 3 4 (g ∗S (T 1 )/3.91) −1
isa smooth funtion ranging from1
to∼ 0.3
in the intervalT 1 ∈ (T 0 , 200GeV)
, and it is plotted in gure 2.3. Thequantity
Ω φ = ρ φ,0 /ρ c
is the fration of energy density of the universe in ALPs.The abundane ismost sensitiveto the initialamplitude of the osillations, being
proportionalto
φ 2 i
,andtoalesserdegreetothepresent massvaluem 0
. Thefator1/ √ m 1
reetsthedampingoftheosillationsintheexpandinguniverse: thelater the osillationsstart,i.e.thesmallerT 1
and thereforeH 1
andm 1
,the lessdampedthey are for agiven
m 0
.and other large sale struture probes[86℄,
ρ CDM = 1.17 keV
cm 3 , Ω CDM = 0.11 h −2 ,
(2.23)it seems that only very large values of
φ i
an provide aρ φ,0
that aounts for allthe dark matter. However, a relativelysmall
φ i
ouldbe ompensated by a smallm 1 ≪ m 0
.Itisnowneessary toprovideamodelforthethermalmassofthepseudosalar
partiletoompare the estimate(2.22a)with theobserved abundane (2.23). For
thepotential(1.25),theALP satisestheequationofmotion(2.13)aslong asthe
angle
ϕ ≡ φ/f φ
is small. The inauray of the quadrati approximation an be anyway ured by an additional orretion fator to (2.22a). This is normally anO (1)
fator exept if we ne tune the initialondition toϕ = π
.Figure 2.4represents the region of the parameter spae of ALPs in whih the
realignamentmehanismouldprovideasuientprodutionofDM.Theallowed
regionsof ALPdarkmatterinthe
m φ
g φ
planeingure2.4anbeobtained usingφ i = ϕ i αC γ
2πg φ
(2.24)
with
ϕ i = φ i /f φ
,theinitialmisalignmentangle,whoserangeisrestritedtovalues between− π
and+π
. The modeldependent fatorC γ
willfrom now on be takento be unity for presentation purposes, but the reader should keep in mind that
in priniple it an assume a very dierent value aording to the model it omes
from. To tune the rightDM abundanewe have used the a priori unknown value
of
ϕ i
. The upper bound ong φ
reets the fat thatϕ i
annot be larger thanπ
, and thus assumesϕ i ∼ π
. Moving to lower values ofg φ
requires inationhappening after SSB in order to have a homogeneous small value of
ϕ i
whih isinreasingly ne-tuned to zero to avoid over-abundant DM prodution. In this
sense the values losest to the boundary, orresponding to the largest values of
the photon oupling, an be onsidered the most natural ones. We will further
ommenton the ne-tuning issue later on.
In the simplest realization of an ALP model, the mass reeives no thermal
orretions, thus
m φ
is onstant throughout the universe expansion and the DMyieldan be inferred diretly from equation (2.22a) using
m 1 = m 0
. In gure 2.4itis the dark red region labelled
m 0 = m 1
.HB excluded
Haloscope excluded
Τ<10 17 s
m 1 >3HHT eq L
m 1 =m 0 m 1 m 0 =HLTL Β
Β= 0 Β= 1 Β= 3 Β= 5 Β= 7 Β= 9
Axion models
-8 -6 -4 -2 0 2 4 6
-16 -14 -12 -10 -8 -6
Log 10 m Φ @eVD Log 10 g Φ @ GeV - 1 D
Figure 2.4: Axion and ALP DM from therealignment mehanism. The axion
modelband is plotted asin gure1.4. The regions ofthe ALP parameter spae
in whih the realignment mehanism provides suient or too large DM
abundaneareplotted indierentredgradationsfordierentthermalmass
mod-els. The pink region fulls theminimumrequirement for non-thermal produed
ALP to behave like CDM. The greyregion over the part ofparameter spaein
whihtheALPlifetime, whihisprovidedbyequation(2.36) ,isshorterthanthe
age of the universe. The turquoise regions have been tested by halosopes and
no ALPs have been found. As a referene, theHB upperbound is plotted as a
If the global symmetry assoiated to the ALP is anomalous, then the mass is
providedby theinstantonipotentialjustasthe
η ′
orthe axionaquiretheirmassthankstoQCDinstantons. TheALP needsinprinipleanotherunbroken
SU(N )
group,whihondenses atasale
Λ
. IfT
isthetemperatureoftheSU(N )
setor,then the thermalALP mass is
m φ ≃
Λ ′2
f φ ≡ m 0
forT ≪ Λ , m 0 Λ ′′
T
β
for
T ≫ Λ .
(2.25)
It makes sense to naively assume
Λ ′ ∼ Λ ′′ ∼ Λ
, while the preise relation amongthese sales depends on the details of the model. At temperatures larger than
Λ
, eletri-sreening damps long range orrelations in the plasma and thus the instantoni ongurations, resulting in a derease of the ALP mass. In speimodels the exponent
β
an be obtained for instane from instanton alulations, but here it is left as a free parameter. Assuming the onset of ALP oherentosillations to happen in the mass suppression regime, it is easy to obtain an
expression for
m 0 /m 1
whih is the expeted enhanement inthe DM abundane,r m 0
m 1
=
√ m 0 m Pl
Λ ′′
β+2 β
(3 × 1.66 √
g ∗1 ) 2β+4 −β
(2.26)and the fator that ontrols the enhanement is
√ m 0 m Pl
Λ ′′ ∼ Λ ′ Λ ′′
r m Pl
f φ
.
(2.27)These models an provideonly a moderate enhanement of the DM density with
respettotheonstant
m φ
ase. ThegainedregionsfortheALPDMaseforvaluesof
β = 1, 3, 5, 7, 9
an be seen in gure 2.4 from bottom to top the lowermostregion
m 1 = m 0
orresponds, of ourse, toβ = 0
. Atually, even onsideringunrealistially huge values of
β
does not help muh, as an be seen from theasymptotiapproahof the highest
β
ases. This isreeted by the nite limitofequation (2.26) when
β → ∞
, but it follows from its denition, equation (2.25).In the
β → ∞
limit,m φ
is a step funtion of temperature,m φ ∝ Θ(T − Λ)
, andthe relation
m 0 = Λ 2 /f φ
determinesΛ
fromm 0
andf φ
. Thus eah point in them φ
g φ
parameter spae has an impliit maximum DM abundane, independent ofβ
. The ruial assumption that leads to these onlusions is thatΛ ′ ∼ Λ
,beauseitdoesnot allowtoonsider arbitrarysmallvaluesfor
Λ
foragiven mass.Therefore, models in whih
Λ ′ ≫ Λ
imply generially higher DM abundane andrequire smaller initial amplitudes
φ i
. Unfortunately, at the moment we an not provide afully motivated example.Finally,theminimalrequirementforanALPtobehaveasCDMisthatatlatest
at matter-radiation equality, at a temperature
T eq ∼ 1.3
eV, the mass attains itsurrent value
m 0
and therefore the DM density starts to sale truly as1/R 3
. Inpartiular, at this point the eld should already have started to osillate. This
orresponds toalowerlimiton
m 1
,m 1 > 3H(T eq ) = 1.8 × 10 −27
eV,whihimpliesan upper bound on
ρ φ,0
,ρ φ,0 < 1.17 keV cm 3
m 0
eV
φ i
54 TeV 2
.
(2.28)In other words, if we want these partilesto be the DM, we need
(m 0 /eV)(φ i /54 TeV) 2 > 1 ,
(2.29)givingaonstraintontherequiredinitialeldvalueasafuntionofthemasstoday,
whih is plotted in gure 2.4 in pink. Again, we are not able to provide a well
motivatedmodel for anALP that ould perform this late realignmentprodution
and this limithas to be onsidered just asthe broadest region of ALP parameter
spae that in prinipleould provide the right amountof CDM.
Thedarkmattergeneratedbytherealignmentmehanismhasinteresting
prop-erties beyond thoseof olddarkmatter. Atthe timeof theirprodution, partiles
from the realignmentmehanism are semi-relativisti. Their momenta are of the
order of the Hubble onstant
p ∼ H 1 ≪ T 1
. Aordingly we have todayout-side of gravitational wells a veloity distribution with a very narrow width of
roughly,
δv(t) ∼ H 1
m 1 R 1
R 0
≪ 1 .
(2.30)Combined with the high number density of partiles,
n φ,0 = N φ /R 3 0 = ρ CDM /m 0
,quantum state,
N occupation ∼ (2π) 3 4π/3
n φ,0
m 3 0 δv 3 ∼ 10 42 m 1
m 0
3/2 eV m 0
5/2
,
(2.31)where we used
R 0 /R 1 ∼ T 1 /T 0 ∼ p
m 1 m Pl /T 0
. If the ALP self-interations are strong enough to ahieve thermalisation and ALP-number-onserving, as arguedin referenes [87,88℄ for the ase of axions, this high oupation number leads
to the formation of a Bose-Einstein ondensate. This ould imprint interesting
signatures inosmologial observations, likegalatiausti rings [8790℄.
The axion realignment prodution is a partiular ase of what is desribed
above. The
Λ
parameter in this ase isΛ QCD
, and the phase transition thatturns on the axion mass is the QCD phase transition. The axion thermal mass
was rst alulated from the dilute instanton gas approximation [91℄ and more
reently usingthe instanton liquidone [92℄. The theoretial unertainties in these
alulationsmaketheparametersforthethermalaxionmassslightlydierent. For
example,asimpleand reentapproximationinthedilutegas approximation,that
alsoagrees very well athigh temperaturewith the liquid instanton one, is [92℄
m a (T ) = 4.1 × 10 −4 Λ 2 f a
Λ T
3.34
,
(2.32)for
Λ = 400
MeV. This approximation is valid aboveT ∼ 0.1
GeV, sine belowthis value the axion mass approahes the zero-temperature value
m a
given byequation (1.6). The exat shape of the thermal axionmass determines
t 1
, thus itaets the