Massive gravitons as FIMP dark matter candidates

Haiying Cai1 ,∗ Giacomo Cacciapaglia2,3 ,† and Seung J. Lee1‡

1
Department of Physics, Korea University, Seoul 136-713, Korea
2
University of Lyon, Université Claude Bernard Lyon 1, F-69001 Lyon, France
3
Institut de Physique des 2 Infinis de Lyon (IP2I),
UMR5822, CNRS/IN2P3, F-69622 Villeurbanne Cedex, France

We discovered a chiral enhancement in the production cross-sections of massive spin-2 gravitons,

below the electroweak symmetry breaking scale, that makes them ideal dark matter candidates for
the freeze-in mechanism. The result is independent on the physics at high scales, and points towards
masses in the MeV range. The graviton is, therefore, a warm dark matter particle, as favoured by
the small scale galaxy structures. We apply the novel calculation to a Randall-Sundrum model with
three branes, showing a significant parameter space where the first two massive gravitons saturate
the dark matter relic density.
arXiv:2107.14548v1 [hep-ph] 30 Jul 2021

Despite the overwhelming evidence for the presence of
a Dark Matter (DM) component in our Universe, also
indirectly observed in galaxies and galaxy clusters, the
nature of this matter component remains a mystery. In
the Standard Model (SM), no known particle can play
the role of DM: the only candidates, neutrinos, have a
relic density many orders of magnitude below the re-
quired one, which is roughly 5 times the relic density
of ordinary baryons. Extended objects, like primordial
black holes, remain a possibility, alas still requiring new
physics to explain their presence [1].
A particle DM candidate can only emerge from new
physics beyond the SM. The most popular and time
honored possibility has been the WIMP (weakly inter-
active massive particle), which requires substantial in-
teractions with the SM particles, with an annihilation
cross-section of the order of the electroweak ones σv '
2.0×10−26 cm3 /s independently of the DM mass in order
to achieve the observed relic density ΩDM h2 ∼ 0.12 [2].
The non-observation of new physics signals at colliders
(the LHC) and at DM direct and indirect detection ex-
periments has, however, put this scenario under stress [3].
Hence, this ‘WIMP crisis’, has prompted the exploration
of alternative possibilities. Here we will be interested
in the freeze-in mechanism of feebly interacting massive
particles (FIMPs), which never attain thermal equilib-
rium with the SM bath [4]. This is usually obtained by
tuning a coupling to very small values, typically of the
order of 10−8 − 10−10 provided the DM is stable (pro-
tected by a parity). Another possibility, called ultravi-
olet (UV) freeze-in [5], relies on higher dimensional op-
erators suppressed by a large scale. In the latter case,
the naive expectation is that unitarity violation renders
the DM relic density highly sensitive to the unknown re-
heat temperature at the end of the inflationary phase.
Here, we will present a novel scenario based on a spin-
2 DM candidate where, albeit the freeze-in is induced
by a non-renormalizable operator, the UV sensitivity of
the predictions is curbed by the electroweak scale. The
mechanism behind is a chiral enhancement of some scat-
tering amplitudes involving massive SM fermions, which
is only activated below the electroweak symmetry break-
ing (EWSB) scale, TEW ≈ 160 GeV.
The couplings of the massive graviton Gµν to the
SM particles can be parametrized by the following 4-
dimensional effective Lagrangian:
X 
δLi

Lef f = Ci Gµν 2 µν − ηµν Li |ĝ=η , (1)
δĝ
i=spin
where η is the Minkowski metric, and the factors within
i
parenthesis are the stress-energy tensors Tµν for the SM
1
particles of different spins (i = 0, 2 , 1). As such, Li is the
SM Lagrangian for the particles of spin i and, since the
spin sum of the graviton polarization tensor is traceless,
the terms proportional to Li in Eq. (1) do not contribute.
Explicit expressions for the various spins can be found in
Ref. [6]. Note that universal couplings between the gravi-
ton and all the SM particles will ensure the unitarity in
the high energy limit before the chiral symmetry break-
ing. Hence, we will stick to this assumption and denote
the universal coupling as CH .
The freeze-in generally proceeds via decays of heavier
particles or pair annihilations of the SM particles in the
thermal bath. The FIMP belongs to a hidden sector that
communicates to the SM sector via a super-weak portal.
Assuming that the inflaton dominantly decays into the
SM particles, the initial DM abundance after reheating
can be neglected, so that the FIMP relic density is pro-
duced via an accumulation process. In our simplified
model, the massive graviton is a perfect FIMP candidate
for DM due to the smallness of the gravitational cou-
pling CH . As we will see, however, as the graviton can
decay via the same coupling, CH is forced to be too small
to produce a sizeable density of gravitons. Here we will
point out a mechanism in place below the EWSB scale
that invalidates this conclusion. Firstly, we introduce the
Boltzmann Equation (BE) describing the production of
the FIMP graviton in the early Universe. For a generic
scattering process B1 + B2 → B3 + G, where B1,2,3 are
the SM particles, the evolution of the graviton number
 2

density nG follows the BE [4]: (the bottom and charm, as the top is too heavy to be in
Z ∞ thermal equilibrium below the EWSB) and can overcome
dnG T
+ 3HnG ≈ ds PB1 B2 PB3 G the Planck suppression for MG is in the keV–MeV range.
dt 512π 6 s0 For the process qg → qG, we can use the cross symmetry
√ √
×AB1 B2 →B3 G K1 ( s/T )/ s ; (2) to get the amplitude squared:
2 2 2

2

where H is the Hubble expansion parameter defined as 256πCH gs mq s − m2q s + m2q
the time derivative of the logarithmic scale factor ln a(t), Aqg = Aq̄g = 4 . (6)
3sMG
K1 is the first modified Bessel function of the second kind
and Note that when inserting the amplitudes squared in the
1 1
BE (2), we need to consider the renormalization group
(s − (mBi + mBj )2 ) 2 (s − (mBi − mBj )2 ) 2 evolution
g2
of the coupling constant, αs (µ) = 4πs where µ =
PB i B j = √ , √
2 s s. In the numerical results, we will use the following
Z
one-loop running coupling
AB1 B2 →B3 G = dΩ|M|2B1 B2 →B3 G ,  
1 7 µ
s0 = max {(mB1 + mB2 )2 , (mB3 + mG )2 } . (3) = 8.47 + log . (7)
αs (µ) 2π MZ
with AB1 B2 →B3 G standing for the amplitude squared af- We are now ready to compute the graviton relic den-
ter the solid angle integration and s = (pB1 + pB2 )2 . sity. After reheating, the universe entered into the
The right hand side of Eq. (2) is the interaction rate radiation-dominated era, with the reheating temperature
density γ(T ) for DM production where the thermal av- TRH representing the maximum temperature reached by
erage is performed by the technique developed in [7]. the thermal bath [10, 11]. The freeze-in production of the
By solving the BE with the proper initial condition, one graviton can be divided into two phases: the UV phase
finds that the FIMP relic density is directly proportional above the EWSB scale and the Infra-Red (IR) one after
to the 2 → 2 cross section, in contrast to the inverse EWSB, i.e. for temperatures above and below the critical
proportionality in freeze-out models. At high tempera- temperature TC ' 160 GeV [12]. The SM particles are
tures, above the EWSB scale, we find that all amplitudes in thermal equilibrium after reheating, while the massive
2
squared scale like AB1 B2 →B3 G(1) ∼ CH gi2 s, where gi is graviton is not. This is because the super-weak portal
an appropriate SM coupling. The only exception is the ensures γ(T ) < Hneq γ both in the UV and IR phases. As
process hh → hG, for which the amplitude is a constant a simple estimation, after EWSB, for the massive gravi-
2 mh
4
−1 M4
Ahh→hG ∝ CH v 2 . As CH ∼ MPl is suppressed by the
ton to stay in non-equilibrium requires T 3 . π 6 MPl mG2 .
q
Planck mass to ensure metastability, the cross-sections Provided the FIMP mass is MG ∼ 1 MeV, this trans-
are too small to provide a DM-like relic density. lates to an upper bound TC . 1 TeV. Rewriting the BE
In this letter we discovered a chiral enhancement of a in terms of the yield YG = nG /S, with S being the en-
class of processes that is active below the EWSB scale, tropy density, the IR contribution to the freeze-in density
and in the limit of light graviton. Thus, in the following can be written as:
we will assume that MG is much smaller than any other 1
Z TC
dT
Z ∞
scale in the process. The chirally enhanced processes YIR ' ds(s − 4m2q )1/2 Aq̄q
2048π 6 TQCD SH 4m2q
involve SM fermions and a massless gauge boson. The √  √  
Z ∞
most dominant ones, therefore, involve quarks and glu- s (s − m2q )2 s
K1 +2 ds A qg K1 , (8)
ons: q q̄ → gG, qg → qG, etc. For the process q̄q → gG, T m2q s3/2 T
the total amplitude squared before EWSB is calculated
(see Appendix ) using the Feynman Rules in [6, 8]: with
r
128π 2 2 2π 2 g∗s T 3 g∗ρ π 2 T 2
A0q̄q = CH gs s , (4) S= , H= (9)
3 45 90 Mpl
where gs is the chromodynamic coupling and the result is where Mpl is the reduced Planck mass for the Hubble
consistent with the results in [9] in the limit of MG → 0. parameter in the radiation-dominated era, TQCD ' 150
In contrast, after EWSB, the leading term in the small MeV and g∗s ' g∗ρ ' 102 . As typically TRH > TC , the IR
MG expansion is given by: contribution is not sensitive to the reheating temperature

TRH . By evaluating Eq.(8) numerically and taking into
2 2 2
256πCH gs mq s s + 2m2q account the bottom and charm quarks, we obtain the
Aq̄q = 4 . (5)
9MG following result:
This term comes from the contribution of the longitudi- MG
ΩIR h2 = YIR
nal polarization of the graviton, which contributes only 3.6 × 10−9 GeV
via a chirality flip of the fermion line via a mass inser- C2
' 3.0 × 1031 GeV5 H3 . (10)
tion. The enhancement is more effective for heavy quarks MG
 3

Since the couplings of G to the SM particles are model 1016

independent and uniquely dictated by symmetries, we
k r1 = 11.0
can ignore the self interactions after including a radion- 1012
like field r. The decay width of a graviton of a few MeV

τ (× 10 17 Sec)
mass is governed by the Lagrangian in Eq. (1): 108
2 3
9CH MG
Γ(G → e+ e− + νi ν̄i + γγ) ' (11)
320π 104
Now we can estimate the upper bound of the IR freeze-
in contribution by combining Eq. (10) with the lifetime 1
τG = 6.58 × 10−25 /ΓG from Eq. (11),
10-4
 6 27
2 1.6 MeV 10 Sec
ΩIR h . 0.12 × . (12) 14.5 15.0 15.5 16.0 16.5 17.0
MG τG
k r2
Given the larger branching ratio of graviton into γγ
than into e+ e− , we impose an appropriate lifetime limit FIG. 1. The life time for the lightest KK graviton (red lines)
τG & 1027 Sec, deriving from the stringent bounds from and the IR brane radion (blue line) as a function of kr2 , with
(1)
indirect detection and CMB [13, 14]. Hence, in Eq.(12), kr1 = 11.0. The solid line corresponds to mr > 2mG while
for each fixed value of MG , the maximum relic den- the dashed line is for mr = 3.45ΛIR . The region with lifetime
sity is given by the largest value of CH that satisfies below the age of the Universe (∼ H0−1 ) is shaded in green.
the lifetime bound. We found, therefore, that the DM
relic density can be saturated by a single graviton with
MG . 1.6 MeV, with increasing lifetimes (decreasing where G0 is the massless graviton (zero mode). The last
CH ) for smaller masses. term in Eq.(14) is composed of cubic self-interactions.
The UV contribution can be computed with a similar The effective couplings for G(n) are derived by the inte-
formula to Eq. (8), taking into account all the quarks in gration of 5D wave-function overlaps [20, 21]:
the SM (which are massless in this regime). Numerically,
1 x2
we obtain CH = √ n , (15)
ΛH 4 2 J2 (xn )
2
ΩUV h = 1.2 × 10 29
GeV CH2
MG √
1 6(1 − J0 (xn ))

MG
4 Cr = , (16)
' ΩIR h2 . (13) Mpl x2n J2 (xn )
4.0 GeV
where we applied an approximation in the limit of
Note that we used TRH = 106 GeV as a template value e−k(r2 −r1 )π  1 and defined ΛH = Mpl e−k(4r1 −3r2 )π .
[10] in the above estimate. Due to the linear divergence The xn is the root of J1 (xn ) = 0. The self interaction
in s of Eq. (4), the integration of BE gives ΩUV h2 ∝ TRH . strength is characterized by CQ · p2 ∼ O(1) and the bulk
One can anticipate, for MG ∼ O(1) MeV, that the UV integration predicts CQ ∼ 1/ΛIR with ΛIR = Mpl e−kr2 π .
result is at most 10−14 times the IR one and much smaller The coupling of radion to di-photon originates from the
in case of a lower TRH input. trace anomaly at the loop level:
We now connect this result to a realistic model for the
!
massive graviton. The model setup is an extension of αEM X
Randal Sundrum (RS) model [15–17] with 3 branes [18, dγ = √ bEM − Fi , (17)
8π 6Λr
19], where all the SM particles are put on the middle i

brane y = r1 π. We take the curvature k ∼ Mpl and the

with the cut off scale Λr = Mpl e−k(2r1 −r2 )π and we
heavy radion associated with the SM brane decouples.
used
P the electromagnetic beta function bEM = − 11 3 , with
Thus, besides a tower of Kaluza-Klein (KK) gravitons 1
F i ' − related to the loop functions of W and those
G(n) , the low energy theory contains a potentially light i 9
heavy fermions including t, b, c, s,τ, µ [22, 23].
radion r peaked around the IR brane y = r2 π. The
effective Lagrangian in Eq. (1) needs to be extended with The mass and coupling orders are of crucial impor-
the interactions between the massive gravitons and the tance for the cosmological stability of the lightest KK
radion r by expanding the RS metric: graviton as DM. Note that in this 5D model the coupling
X G(1) -G0 -G0 is highly suppressed by order of Mpl ekr2 π ,
Lef f = CH G(n)µν Tµν
i
+ dV rVµν V µν unlike the bigravity model where this coupling is ab-
i sent [24]. The decay width of a massive graviton into
two zero gravitons is negligible. Also the hadronic decays
 
+ Cr 2G(n)µν G0µν − G(n)µ
µ G 0ν
ν r
are kinematically forbidden for MG in MeV range. For
+ CQ Q(G3 , G2 r, Gr2 , r3 ) , (14) the lightest KK graviton, due to the O(1) self-coupling,

we have to require mr > MG(1) /2 to shut down the de-
cay of G(1) → 2r. Hence the relevant decay channels are
G(1) → e+ e− , ν ν̄ and γγ with their decay width given by
Eq. (11). The other decay patterns will be determined
by the mass of radion. Note that from NDA a radion
mass (or dilaton in the 4D CFT dual) is expected to be
around the KK scale [25–27], without the necessity for a
fine-tuning. First of all, we calculated the decay width
for G(n) → G0 + r:
C 2 m4 m2
 
Γ(G(n) → G0 r) = r r 1 − r2 , (18)
4πMG MG
with the other relevant decay widths of G(n) → 2G(1) , 2r
and r → 2G(1) , 2γ given in [28, 29].
Here we consider two scenarios: in one case, MG(1) /2 <
mr . MG(2) /2 allows for a Planck-suppressed decay of
G(1) → G0 + r and a prompt decay of G(2) → 2r; in
the second, the radion is slightly heavier mr & 2MG(1) >
MG(2) so that the radion and G(n) (n ≥ 3) quickly decay
into 2 G(1) in less than 10−14 Sec for MG(1) ∼ 2 MeV.
As shown in Figure 1, for kr1 = 11.0, we roughly need
kr2 & 14.8 to make the lowest KK graviton to be stable
beyond the Hubble time level. However for the IR brane
radion, a much larger value of kr2 & 16.7 is required
to barely ensure the same property. This reflects the
fact that the cut off scale ΛH of the KK graviton is of
order Planck scale, while the one for the radion is smaller.
For the lighter radion case, the dashed red line bends
at a certain point, indicating that the dominant decay
becomes G(1) → G0 + r, as the coupling Cr does not
decrease with the radius r2 of the IR brane. A small mr
will render the graviton G(1) less stable. Thus the second
scenario seems to be a more natural option, where the
lightest two gravitons are both long-lived enough to play
the role of DM.
In the following, we will focus on the heavier radion
scenario. The heavier KK gravitons will also undergo a
non-thermal IR freeze-in and afterwards cascade deposit
their density into the stable gravitons. Since the energy
is conserved during the subsequent freeze-in, the relic
density approximately is:
10
X Sundrum model with 3 branes. Depending on the radion
ΩIR h2 ' 1.75 × 10−62 ek(8r1 −3r2 )π x2n , (19)
n=1
where we traded (CH , MG ) in Eq. (10) with the two ra-
dius parameters in the 5D realization and summed up all
the contribution till the n = 10 KK graviton. Note that
the self interactions of the radion receive a very large (fi-
nite) 1-loop correction from the KK gravitons, hence the
5D description will no longer be reliable beyond 10MKK
scale. The prediction of Figure 2 is consistent with the
upper limit derived in Eq. (12). Since G(1) and G(2)
mainly decay into di-photon and neutrinos, we impose a
proper Gamma-Ray bound τG(1) ∼ 10 τG(2) > 1028 Sec
for the viable relic density. In Figure 2, the contours for
4
17.0
10-6
10-4
16.5
10-2
Ωh2 > 0.12
k r2
16.0
τG(1) < 1028 Sec
15.5
11.0 11.2 11.4 11.6 11.8
k r1
FIG. 2. The contour of relic density in the plane of (kr1 , kr2 ),
where the blank region is viable for the graviton as long-lived
DM. The boundary of the pink region satisfies Ωh2 = 0.12 by
summing the direct and subsequent freeze-in. The light blue
region is excluded by the indirect direction.
relic density (Ωh2 = 10−2 , 10−3 , 10−4 , 10−5 , 10−6 , 10−7 )
are parallel to each other and equally spaced. Further-
more, the ratio of the lines is determined by the expo-
∂r2
nential factor in Eq. (19) ∂r1 = 8/3. For kr1 ∼ 11.2 and
kr2 ∼ 15.8, the lowest two KK gravitons can achieve a
large fraction of the observed DM relic density.
In summary, in this letter we presented the effect of a
chiral enhancement for the single graviton production in
the early Universe. This enhancement is active below the
EWSB scale and insensitive to the UV physics. Thanks
to this novel effect, a generic massive graviton can play
the role of a FIMP DM candidate for masses MG ∼ a few
MeV. Note that the mass range makes the spin-2 graviton
an ideal warm DM particle, which is currently favoured
by the small scale galaxy structure [30, 31]. We have
applied this new result to an extension of the Randall-
mass, only the first two KK gravitons can be long-lived at
a cosmological scale, while the higher modes and radion
decay very effectively into DM. Note that the radion was
considered as a DM candidate in [32], but being difficult
to account for a heavy relic density. This is the first time
that a light spin-2 mode, or a class of “glueball” interpo-
lated by the conserved energy-momentum tensor of the
strong dynamics from AdS/CFT correspondence [33–35],
is shown to be a feasible DM candidate. We illustrated
that there exists a sizeable parameter space where the
light gravitons can saturate the DM relic density, while
escaping the stringent bounds on decaying DM from in-
direct detection and the CMB.

ACKNOWLEDGEMENTS
H.C. and S.L. were supported by the National Research
Foundation of Korea (NRF) grant funded by the Korea
government (MEST) (No. NRF-2021R1A2C1005615).
H.C. acknowledges the support of Tsung-Dao Lee Insti-
tute, SJTU where this project is initiated. G.C. and
S.L also acknowledge support from the Campus-France
STAR project “Higgs and Dark Matter connections”.
∗
hcai@korea.ac.kr
†
g.cacciapaglia@ipnl.in2p3.fr
‡
sjjlee@korea.ac.kr
[1] D. Y. Cheong, S. M. Lee, and S. C. Park, JCAP 01, 032
(2021), arXiv:1912.12032 [hep-ph].
[2] N. Aghanim et al. (Planck), Astron. Astrophys. 641, A6
(2020), arXiv:1807.06209 [astro-ph.CO].
[3] G. Bertone and M. P. T. Tait, Nature 562, 51 (2018),
arXiv:1810.01668 [astro-ph.CO].
[4] L. J. Hall, K. Jedamzik, J. March-Russell, and S. M.
West, JHEP 03, 080 (2010), arXiv:0911.1120 [hep-ph].
[5] F. Elahi, C. Kolda, and J. Unwin, JHEP 03, 048 (2015),
arXiv:1410.6157 [hep-ph].
[6] T. Han, J. D. Lykken, and R.-J. Zhang, Phys. Rev. D
59, 105006 (1999), arXiv:hep-ph/9811350.
[7] P. Gondolo and G. Gelmini, Nucl. Phys. B 360, 145
(1991).
[8] G. F. Giudice, R. Rattazzi, and J. D. Wells, Nucl. Phys.
B 544, 3 (1999), arXiv:hep-ph/9811291.
[9] E. A. Mirabelli, M. Perelstein, and M. E. Peskin, Phys.
Rev. Lett. 82, 2236 (1999), arXiv:hep-ph/9811337.
[10] C. Cheung, G. Elor, and L. J. Hall, Phys. Rev. D 85,
015008 (2012), arXiv:1104.0692 [hep-ph].
[11] S. Hannestad, Phys. Rev. D 70, 043506 (2004),
arXiv:astro-ph/0403291.
[12] M. Quiros, in ICTP Summer School in High-Energy
Physics and Cosmology (1999) arXiv:hep-ph/9901312.
[13] R. Essig, E. Kuflik, S. D. McDermott, T. Volansky, and
K. M. Zurek, JHEP 11, 193 (2013), arXiv:1309.4091
[hep-ph].
5
[14] V. Poulin, J. Lesgourgues, and P. D. Serpico, JCAP 03,
043 (2017), arXiv:1610.10051 [astro-ph.CO].
[15] L. Randall and R. Sundrum, Phys. Rev. Lett. 83, 4690
(1999), arXiv:hep-th/9906064.
[16] L. Randall and R. Sundrum, Phys. Rev. Lett. 83, 3370
(1999), arXiv:hep-ph/9905221.
[17] W. D. Goldberger and M. B. Wise, Phys. Rev. Lett. 83,
4922 (1999), arXiv:hep-ph/9907447.
[18] I. I. Kogan, S. Mouslopoulos, A. Papazoglou, and L. Pilo,
Nucl. Phys. B 625, 179 (2002), arXiv:hep-th/0105255.
[19] K. Agashe, P. Du, S. Hong, and R. Sundrum, JHEP 01,
016 (2017), arXiv:1608.00526 [hep-ph].
[20] H. Davoudiasl, J. L. Hewett, and T. G. Rizzo, Phys.
Rev. Lett. 84, 2080 (2000), arXiv:hep-ph/9909255.
[21] H. Davoudiasl, J. L. Hewett, and T. G. Rizzo, Phys.
Rev. D 63, 075004 (2001), arXiv:hep-ph/0006041.
[22] G. F. Giudice, R. Rattazzi, and J. D. Wells, Nucl. Phys.
B 595, 250 (2001), arXiv:hep-ph/0002178.
[23] C. Csaki, J. Hubisz, and S. J. Lee, Phys. Rev. D 76,
125015 (2007), arXiv:0705.3844 [hep-ph].
[24] E. Babichev, L. Marzola, M. Raidal, A. Schmidt-May,
F. Urban, H. Veermäe, and M. von Strauss, JCAP 09,
016 (2016), arXiv:1607.03497 [hep-th].
[25] B. Bellazzini, C. Csaki, J. Hubisz, J. Serra, and J. Tern-
ing, Eur. Phys. J. C 73, 2333 (2013), arXiv:1209.3299
[hep-ph].
[26] Z. Chacko and R. K. Mishra, Phys. Rev. D 87, 115006
(2013), arXiv:1209.3022 [hep-ph].
[27] Z. Chacko, R. K. Mishra, and D. Stolarski, JHEP 09,
121 (2013), arXiv:1304.1795 [hep-ph].
[28] H. M. Lee, M. Park, and V. Sanz, Eur. Phys. J. C 74,
2715 (2014), arXiv:1306.4107 [hep-ph].
[29] B. M. Dillon, C. Han, H. M. Lee, and M. Park, Int.
J. Mod. Phys. A 32, 1745006 (2017), arXiv:1606.07171
[hep-ph].
[30] M. Boylan-Kolchin, J. S. Bullock, and M. Kaplinghat,
Monthly Notices of the Royal Astronomical Society: Let-
ters 415, L40–L44 (2011).
[31] K. A. Oman et al., Mon. Not. Roy. Astron. Soc. 452,
3650 (2015), arXiv:1504.01437 [astro-ph.GA].
[32] E. W. Kolb, G. Servant, and T. M. P. Tait, JCAP 07,
008 (2003), arXiv:hep-ph/0306159.
[33] J. M. Maldacena, Adv. Theor. Math. Phys. 2, 231 (1998),
arXiv:hep-th/9711200.
[34] E. Witten, Adv. Theor. Math. Phys. 2, 253 (1998),
arXiv:hep-th/9802150.
[35] S. S. Gubser, I. R. Klebanov, and A. M. Polyakov, Phys.
Lett. B 428, 105 (1998), arXiv:hep-th/9802109.
 6

Amplitude Squared for Single Graviton Production

For the process q̄(p1 )+q(p2 ) → g(k1 )+Gµν (k2 ), there are a contact term contribution plus the S, T and U channels,
listed in order as following:
i
Ma1 = − CH gs v̄(p2 ) (ηµρ γν + ηρν γµ ) T a u(p1 )ρ (k1 )µν (k2 ) (20)
2
Wµν,σρ (−p1 − p2 , k1 ) σ a
Ma2 = −iCH gs v̄(p2 ) γ T u(p1 )ρ (k1 )µν (k2 ) (21)
(p1 + p2 )2
i γµ (p1 − k1 − p2 )ν + γν (p1 − k1 − p2 )µ
Ma3 = CH gs v̄(p2 ) (6 p1 − 6 k1 − mq )γρ T a u(p1 )ρ (k1 )µν (k2 ) (22)
4 (p1 − k1 )2 − m2q
i γµ (p1 − k2 + p1 )ν + γν (p1 − k2 + p1 )µ
Ma4 = CH gs v̄(p2 )γρ (6 p1 − 6 k2 − mq ) γρ T a u(p1 )ρ (k1 )µν (k2 ) (23)
4 (p1 − k1 )2 − m2q
where T a is the SU (3) gauge generator, ρ (k1 ) and µν (k2 ) are the gluon and graviton polarization tensors respectively.
When we work in the ξ = 1 gauge, the symbol Wµν,σρ (k1 , k2 ) including the gauge fixing term is defined as [6]:
Wµν,σρ (k1 , k2 ) = k1 · k2 Cµν,σρ + Dµν,σρ (k1 , k2 ) + ξ −1 Eµν,σρ (k1 , k2 ) (24)
with
Cµν,σρ = ηµσ ηνρ + ηµρ ηνσ − ηµν ησρ ,
 
Dµν,σρ (k1 , k2 ) = ηµν k1ρ k2σ − ηµρ k1ν k2σ + ηµσ k1ρ k2ν − ησρ k1µ k2ν + (µ ↔ ν) ,
 
Eµν,σρ (k1 , k2 ) = ηµν (k1σ k1ρ + k2σ k2ρ + k1σ k2ρ ) − ηνρ k1µ k1σ + ηνσ k2µ k2ρ + (µ ↔ ν) . (25)

In the limit of mq → 0, the amplitude squared can be derived to be:

|Matot |2 = |Ma1 + Ma2 + Ma3 + Ma4 |2
2 2
4CH gs
4 s2 + 2st + 2t2 (s + t) t − MG 6


= 2 (s + 4t)
st (s + t − MG )
4 2
s3 + 6s2 t + 18st2 + 16t3



+6MG (s + 2t) t − MG (26)
where s = (p1 + p2 )2 and t = (p1 − k1 )2 are the Mandelstam variables.
We can show how the longitude modes enter into effect.The sum of polarization for a massive graviton is:
 
1 2
Pµν,αβ = Gµα Gνβ + Gνα Gµβ − Gµν Gαβ (27)
2 3
kµ kν
For the tensor Gµν = ηµν − MG 2 , the last piece comes from the longitude polarisation. We can peel off this pure
k2,µ k2,ν
longitude part MG 2 so that the respective amplitude is :

a,L iCH gs ρ (k1 )
v̄(p2 ) m2q − (p1 − k1 )2 γ ρ + 2mq (γ ρ 6 k2 − 6 k2 γ ρ )


Mtot = 2
2MG

ρ ρ ρ ρ
− 2 6 k2 (p2 − k1 ) − 2 6 k1 k2 − k1 MG /s − 6 k2 6 k1 γ T a u(p1 )
2 ρ


(28)

where we keep the terms proportional to k1ρ . For mq = 0, because the longitude mode will not contribute to the cross
a,L 2
section due to helicity conservation, we will get |Mtot | = 0. However after the quark obtains mass, the amplitude
a a
squared with the color factor Tr T T = 4 counted is:
2 2 2



a,L 2 128CH gs mq m2q − t s + t − m2q
|Mtot | = 4
3MG
1
+ O( 2 ) (29)
MG
As we can see there is an enhancement factor m4q /MG
4
in the amplitude squared provided the mass of graviton is in
the KeV-MeV order. This type of enhancement also applies to the process of q q̄ → hG(1) . However for the latter, its
2 q m4 s
amplitude squared is proportional to CH v2 M 2
, thus contributes less to the freeze-in relic density.
G