DESY-25-162

When Geometry Radiates Review

Gravitational Waves in Theory, Cosmology, and Observation

Azadeh Maleknejada,b,c
a
Centre for Quantum Fields and Gravity, Swansea University, Swansea SA2 8PP, United Kingdom
b
Deutsches Elektronen-Synchrotron DESY, Notkestraße 85, 22607 Hamburg, Germany
arXiv:2512.21328v1 [gr-qc] 24 Dec 2025

c
Institute of Theoretical Physics, Universität Hamburg, 22761 Hamburg, Germany

E-mail: [Link]@[Link]

Abstract: Gravitational waves provide a unique window into gravity, cosmology, and
high-energy physics, enabling the exploration of fundamental phenomena across a wide
range of scales. This review presents a coherent and pedagogical framework that bridges
foundational theory with observational frontiers. We begin by developing the theory of
gravitational radiation within linearized general relativity, deriving gravitational waves as
solutions to the linearized Einstein equations and clarifying their physical interpretation,
polarization states, and key properties. We then deepen the discussion through a more
rigorous geometric perspective, tracing the connection between gravitational radiation and
the algebraic structure of the Weyl tensor. This naturally leads to the Bondi–Sachs for-
malism, which provides a precise definition of gravitational waves in asymptotically flat
spacetimes and establishes their role as genuine carriers of energy and angular momen-
tum. Extending beyond flat backgrounds, we examine gravitational waves in an expanding
universe, following their evolution across cosmological epochs and their generation during
inflation. Within this setting, we explore adiabatic modes and consistency relations, which
illuminate the infrared structure and universal behavior of long-wavelength perturbations,
and derive the inflationary spectrum of vacuum gravitational waves together with their
contribution to the integrated Sachs–Wolfe effect. The review further surveys the main
observational strategies for detecting gravitational waves across a broad frequency range,
including cosmic microwave background polarization, pulsar timing arrays, laser interfer-
ometers on Earth and in space, and resonant cavity–based detectors. We then discuss the
astrophysical and cosmological mechanisms responsible for generating gravitational radia-
tion. We conclude by summarizing the current status of the field and outlining promising
directions for future theoretical and observational developments.

Keywords: Gravitational radiation, Asymptotically flat space, Cosmology, Geometrical

& algebraic aspects of gravitational waves, Sources of gravitational waves, Detection of
gravitational waves
Contents

1 Introduction and Motivation 1

2 Linearized General Relativity and GWs 5

2.1 The Weak-Field Approximation and Wave Solutions 6
2.1.1 Gravitational Radiation of Compact Objects 8
2.2 Polarization States of GWs and Free Particles 12
2.2.1 Geodesic Distortion by Passing GWs 13

3 Theoretical Aspects of Gravitational Waves 16

3.1 Theoretical Issues of Linear Gravity 16
3.2 The Weyl Tensor: Geometry of Tidal Forces 17
3.3 When Geometry Radiates: Defining Gravitational Waves 22
3.3.1 GWs Carry Energy 27
3.3.2 GWs Carry Angular Momentum 27
3.4 At the Edge of Spacetime: From Poincaré to BMS 29
3.4.1 Asymptotic symmetries of Gravitational Fields 29
3.4.2 Why Radiative Spacetimes Have the Full BMS Group? 31

4 Expanding Universe and Gravitational Waves 33

4.1 Expansion History and Cosmological Eras 33
4.1.1 Cosmic Inflation 37
4.1.2 From the Inflationary Paradigm to Precision Cosmology 40
4.2 Inflation and GWs 41
4.3 IR Structure of Cosmological Perturbations 46
4.3.1 Adiabatic Modes in Cosmology 46
4.3.2 Maldacena’s Consistency Relation & Primordial Non-Gaussianity 50
4.4 Inflationary Gravitational Waves at Late Times 53
4.4.1 The Present-Day Spectrum of Gravitational Waves from Inflation 53
4.4.2 Integrated Sachs–Wolfe effect and Primordial GWs 57

5 Detection Techniques and Observational Frontiers 59

5.1 CMB Polarization: Imprints of Primordial Gravitational Waves 60
5.2 Pulsar Timing Arrays: Precision Cosmic Clocks for GW Detection 67
5.3 Laser Interferometry: Earth- and Space-Based GW Observatories 70
5.4 Photonic Qumodes: a Cutting-Edge Quantum Platform for GW Detection 75

6 Physical Mechanisms Driving Gravitational Radiation 84

6.1 Astrophysical Sources 85
6.1.1 Compact binary coalescences 85
6.1.2 Isolated neutron stars 86
6.1.3 Transient burst sources 87

–i–
 6.1.4 Astrophysical Contributions to the Stochastic GW Background 87
6.2 Cosmological Sources 87
6.2.1 Inflation and Reheating 88
6.2.2 Cosmological Phase Transitions and Hydrodynamical Sources 92
6.2.3 Topological Defects 94
6.2.4 Primordial Black Holes and scalar-induced gravitational waves 95

7 Summary 96

Acknowledgments 97

1 Introduction and Motivation

Gravitational waves (GWs) were first predicted by Einstein in 1916 as ripples in spacetime
propagating at the speed of light, arising from the dynamical degrees of freedom of the
metric in general relativity [1]. They embody a fascinating interplay of physics, mathemat-
ics, and phenomenology, yet their physical reality was debated for decades. In the 1930s,
Einstein himself briefly doubted the physical reality of gravitational waves due to a misin-
terpretation of early solutions. He later revised his view, though questions about their true
physicality remained. It wasn’t until the 1950s, through the work of Bondi, Pirani, and
others, that these issues were resolved. Bondi’s analysis at null infinity demonstrated that
gravitational waves carry energy and can have measurable physical consequences [2]. This
marked the turning point in establishing gravitational waves as real, observable phenomena
in general relativity that carry energy and angular momentum away from isolated systems.
With their physical interpretation rigorously established, attention turned toward their
astrophysical sources and thier detection. Indirect evidence arrived with the Hulse–Taylor
binary pulsar in 1974 [3], whose orbital decay precisely matched the prediction for GW
emission. Early detection efforts began with Weber’s resonant bars and gradually evolved
into large-scale interferometers, propelled by the vision of Weiss, Drever, and Thorne.
These decades of theoretical and experimental progress culminated in the landmark direct
detection by LIGO in 2015 [4], an achievement that inaugurated the era of GW astronomy.
Beyond established physics, gravitational waves provide a powerful probe of high-
energy phenomena beyond the Standard Model (BSM) and a unique window into the
unexplored frontiers of gravity and the early Universe. As messengers from the earliest
moments of cosmic history, they access energy scales far beyond the reach of terrestrial
accelerators, offering potential insights into new physics and even the quantum nature
of spacetime. Stochastic gravitational wave backgrounds may originate from inflationary
fluctuations [5, 6], violent post-inflationary dynamics such as preheating [7], or relics of
high-energy phase transitions and topological defects [8, 9]. Intriguingly, in certain sce-
narios, GWs may not merely carry information about early-universe processes, they could

–1–
Figure 1: Gravitational rainbow. At the lowest frequencies, measurements of the cosmic
microwave background (CMB) provide the most sensitive probe of gravitational waves.
In the nanohertz range, pulsar timing arrays serve as powerful astrophysical detectors. At
intermediate and frequencies as high as a few kiloherz, laser and atomic interferometers are
actively probing the spectrum. Finally, we enter the realm of high-frequency gravitational
waves above 10 kHz, where detection demands innovative techniques.

actively participate in its creation, potentially influencing the matter content of the Uni-
verse [10–21].
Just as electromagnetic radiation, gravitational waves possess their own spectrum (see
Fig. 1). Although their frequency spectrum is, in principle, unbounded from above and
below, only a finite range is observationally accessible. More precisely, their wavelengths
are bounded below by the Hubble radius, corresponding to a minimum frequency of about
10−18 Hz, while detector sensitivity currently limits the upper end to 10 kHz range. Within
this observational window, different frequency bands reveal complementary science: very-
low-frequency GWs imprint the CMB polarization [22, 23]; nanohertz signals are tracked
by pulsar timing arrays [24–26] and astrometric observations [27, 28]; millihertz waves will
be measured by space interferometers such as LISA [29]; and the kilohertz regime is already
accessed by ground-based detectors such as LIGO [4, 30, 31]. Future third-generation facil-
ities, including the Einstein Telescope [32], promise even greater sensitivity. Atom interfer-
ometry offers a complementary pathway to laser interferometers for GW detection [33–35].
Terrestrial and space-based cold-atom concepts under development [36–46] aim to extend
sensitivity into the mid-frequency band (10−2 –1 Hz), thereby bridging the gap between
LISA and ground-based observatories. At the highest frequencies, innovative laboratory-
scale concepts are being developed to extend gravitational wave sensitivity well beyond the
MHz regime [47], including a number of novel approaches based on quantum-sensing tech-
niques [48, 49]. This is not all: there are several further detectors and proposals for probing
GWs across different frequency windows. For an overview of the global network of laser-
interferometry detectors on Earth and in space, see Fig. 32. Together, these approaches
build a multi-frequency gravitational wave astronomy, analogous to the multi-wavelength
program of electromagnetic observations.

–2–
Figure 2: Timeline of GWs observatories from the first detection (Advanced LIGO, 2015)
to planned facilities through the 2040s. Virgo joined in 2017 and KAGRA in 2020, estab-
lishing the first global ground-based network. Forthcoming projects—LIGO–India (early
2030s) [31], LiteBIRD (2032) [50], LISA (2035) [29], and the Einstein Telescope (expected
from ∼2035-2040) [32] —extend coverage across frequency bands and improve sky local-
ization, charting the roadmap for GW astronomy into the next decades.

Figure 3: Hierarchy of astrophysical and cosmological scales. Observations reveal that our
Universe is expanding and becomes statistically isotropic and homogeneous on scales larger
than about 100 Mpc, defining the cosmological regime. Gravitational systems up to these
scales are well described by asymptotically flat geometries, whereas on larger scales the
spacetime must be modeled by the FLRW metric to capture the expansion of the Universe.
The cosmic horizon marks the limit of observation, not the limit of existence.

Since gravitational waves are both generated and propagate through the gravitational
backgrounds of the Universe, it is essential to describe them within the appropriate space-
time geometry relevant to the physical scale under consideration. The behavior of gravity
manifests differently across scales. On relatively small scales, up to roughly 100 Mpc, the
expansion of the Universe is negligible, and gravitational dynamics can be accurately de-
scribed within asymptotically flat geometries. On larger, cosmological scales, however,
the expansion of space itself becomes dynamically relevant, and spacetime must instead
be modeled by an FLRW background (see Fig. 3). In addition to this geometric distinc-
tion, there is also a conceptual difference between the sources of gravitational waves across
these regimes. Individual astrophysical sources, such as binary mergers or supernovae, are
typically deterministic, producing coherent signals with non-vanishing mean amplitude,
⟨hij ⟩ ̸= 0. In contrast, cosmological sources, arising from processes in the early Universe
such as inflation and phase transitions, tend to generate a stochastic gravitational wave

–3–
Figure 4: Comparison between deterministic and stochastic classical gravitational waves.
Left: A smooth, deterministic waveform from an astrophysical binary merger. Right: A
zero-mean Gaussian stochastic background, as expected from cosmological origins.

background (GWB) characterized by vanishing mean but nonzero two-point correlations,

⟨hij hij ⟩ ̸= 0 (see Fig. 4). This distinction underlines why the statistical treatment of
gravitational waves becomes crucial in cosmological contexts. This fundamental separa-
tion, both geometric and statistical, motivates the structure of this review. It is therefore
divided into two parts: the first devoted to gravitational physics in asymptotically flat
spacetimes, and the second to that in expanding cosmological geometries.
All figures presented in this work are original creations of the author, except where
explicit attribution or citation is provided.

This review is structured as follows. Sec. 2 develops the framework of linearized general
relativity, deriving GWs as solutions to the linearized Einstein equations and examining
their physical meaning and properties. Sec. 3 deepens the discussion with a more rigorous
treatment, tracing how gravitational radiation are linked to the Weyl tensor’s algebraic
structure. This naturally leads to the Bondi–Sachs formalism, which clarifies the physical
nature of GWs as real carriers of energy and angular momentum. Sec. 4 examines GWs in
an expanding Universe, following their evolution across different cosmological epochs and
their generation during inflation. We then explore Weinberg’s adiabatic modes and Mal-
dacena’s consistency relations, which shed light on the infrared structure and the universal
behavior of long-wavelength perturbations. We wrap up the section by deriving the infla-
tionary spectrum of vacuum gravitational waves and discussing how they contribute to the
integrated Sachs–Wolfe effect. In Sec. 5 we survey the principal observational techniques
used to detect gravitational waves, including CMB polarization measurements, pulsar tim-
ing arrays, ground- and space-based laser interferometry, and photon-regeneration cavities
as gravitational wave detectors. Sec. 6 is devoted to the physical mechanisms that generate
gravitational radiation, beginning with astrophysical sources and followed by cosmological
sources. Finally, in Sec. 7 we summarize and outline possible future directions.

Notation and Conventions: Throughout this review we set c = 1 while keeping G

explicit, and we work within four-dimensional General Relativity unless stated otherwise.

Quantity Definition / Convention

Reduced Planck mass 2 ≡ (8πG)−1
MPl
Metric signature ηµν = diag(−, +, +, +)

–4–
 Quantity Definition / Convention
Spacetime indices Greek letters µ, ν, ρ, σ = 0, . . . , 3
Spatial indices Latin letters i, j, k = 1, 2, 3
Partial derivative ∂µ ≡ ∂/∂xµ
Lie derivative Lξ X µ ν = ξ ρ ∇ρ X µ ν − (∇ρ ξ µ )X ρ ν + (∇ν ξ ρ )X µ ρ
Riemann tensor Rρ σµν = ∂µ Γρνσ − ∂ν Γρµσ + Γρµλ Γλνσ − Γρνλ Γλµσ
Ricci tensor Rµν = Rρ µρν
Ricci scalar R = g µν Rµν
Gravitational waves in TT gauge γij
Cosmic scale factor a(t)
Hubble parameter H ≡ ȧ/a
Cosmic time t
Bondi time u (u = t − r)
Conformal time τ (dτ ≡ dt/a(t))

Throughout the text, shaded boxes are used to introduce definitions essential for un-
derstanding the material in the current or subsequent sections. Finally, we will frequently
use the following acronyms.

BBN Big Bang Nucleosynthesis LSS Large-Scale Structure

BH Black Hole PND Principal Null Directions
BMS Bondi–Misner–Sachs group PTA Pulsar Timing Array
BSM Beyond the Standard Model GWB Gravitational Wave Background
CMB Cosmic Microwave Background SLS Surface of Last Scattering
GR General Relativity SM Standard Model
GW Gravitational Wave SNR Signal-to-Noise Ratio
HFGW High-Frequency Gravitational Waves TT Transverse-Traceless gauge

2 Linearized General Relativity and GWs

Shortly after formulating the theory of general relativity, Albert Einstein linearized his field
equations and discovered that the theory admits wave-like solutions [1], small perturbations
of the Minkowski space-time, that propagate at the speed of light. These solutions, known
today as gravitational waves, appear in the weak-field approximation, where the space-time
metric is only slightly perturbed from flat space. Although this linearized approach may
initially appear as a mere simplification, it has proven to be remarkably powerful in both
theoretical and observational contexts within astrophysics and cosmology. It enables the

–5–
analysis of gravitational radiation from compact astrophysical sources such as binary black
holes and neutron stars, and forms the theoretical foundation for modern gravitational wave
detectors. We begin this section by considering the weak-field limit of general relativity
and deriving the general solution to the linearized Einstein equations. We next investigate
the properties of gravitational waves generated by localized sources, and analyze their
polarization states and observable effects on freely falling test particles.

2.1 The Weak-Field Approximation and Wave Solutions

A weak gravitational field corresponds to a region of spacetime that is only mildly curved.
In such regions, there exist coordinate systems in which the spacetime metric can be
expressed as

gµν = ηµν + h̃µν , (2.1)

where ηµν is the Minkowski metric

ηµν = diag(−1, 1, 1, 1), (2.2)

and h̃µν represents a small perturbation such that both it and its partial derivatives remain
small

h̃µν ≪ 1 and ∂λn h̃µν ≪ 1 (for all n > 1). (2.3)

This intuitive characterization will be given a mathematically rigorous formulation in Sec. 3.

The formulation Eq. (2.3) sets the stage for linearizing Einstein’s equations
1
Rµν − R gµν = 8πG Tµν , (2.4)
2
and studying the dynamics of gravitational waves as perturbations propagating in flat
spacetime. Note that we can well consider small perturbations about some other back-
ground metric, such that gµν = ḡµν + h̃µν . In particular, later in Sec. 4, we will use a
similar weak field approach for cosmological perturbations around the FLRW metric.
In the weak-field general relativity, we expand the field equations in powers of h̃µν
and keep the linear terms. In fact, from the Einstein equation in Eq. (2.4), we find the
linearized gravitational field equation as

∂α ∂ α h̃µν + ∂µ ∂ν h̃ − ∂ν ∂λ h̃λµ − ∂µ ∂λ h̃λν − ηµν (∂α ∂ α h̃ − ∂σ ∂λ h̃λσ ) = −16πGTµν , (2.5)

where h̃ is the trace of the field, h̃ = h̃µµ . Although linearized, the equation above does
not yet resemble a wave equation. However, it can be significantly simplified through an
appropriate field redefinition
1
hµν ≡ h̃µν − ηµν h̃, (2.6)
2
which is the trace-reverse cousin of h̃µν , i.e. h = −h̃. The metric perturbation hµν possesses
a gauge freedom that originates from the diffeomorphism invariance of general relativity.

–6–
 Under an infinitesimal coordinate transformation

xµ → xµ + ξ µ (x), (2.7)

where ξ µ is a smooth vector field, the background metric ηµν changes according to the Lie
derivative along ξ µ , Lξ ηµν . Consequently, the perturbation transforms as

hµν → hµν + Lξ ηµν = hµν + ∇µ ξν + ∇ν ξµ . (2.8)

This transformation expresses the gauge freedom: the physical content of hµν remains
unchanged under such a shift. To fix this gauge ambiguity, it is conventional to impose the
Lorenz gauge condition
∂ µ hµν = 0 (Lorenz gauge). (2.9)
In this gauge, the linearized Einstein field equations take the particularly simple and co-
variant form
□hµν = −16πG Tµν , (2.10)
where □ = ∂ α ∂α is the d’Alembertian operator in flat spacetime.
We now turn to the vacuum case of the linearized Einstein field equations, obtained
by setting Tµν = 0 in Eq. (2.10). This equation admits plane wave solutions of the form

hµν (x) = eµν (k) eik·x , (2.11)

where eµν (k) is a constant, symmetric tensor representing the amplitude and polarization
of the wave, and kµ is the wave 4-vector. Substituting into the wave equation shows that
the wave vector must satisfy the null condition k µ kµ = 0, which confirms that gravitational
waves propagate at the speed of light.

In the presence of matter or energy, the stress-energy tensor Tµν acts as a source for
the gravitational field, and the linearized Einstein equation becomes inhomogeneous. To
solve it, we employ the Green’s function method for the wave operator

□G(xµ − y µ ) = δ (4) (xµ − y µ ), (2.12)

where G(xµ − y µ ) is the Green’s function of the flat-space d’Alembertian. The physically
relevant solution is the retarded Green’s function, which ensures causal propagation and
takes the form
1
G(xµ − y µ ) = δ x0 − y 0 − |x − y| , for x0 > y 0 .


(2.13)
4π|x − y|
Using this, the solution to the sourced linearized gravitational field equations is given
by
Tµν (y, t − |x − y|)
Z
hµν (x, t) = −4G d3 y . (2.14)
|x − y|
This expression reveals that the gravitational field at the spacetime point (t, x) is
determined entirely by the values of the source Tµν along the past light cone of that
point. In other words, gravitational effects propagate causally at the speed of light, just as
expected for massless fields (see Fig. 5).

–7–
Figure 5: The variation of the gravitational field at a spacetime point (t, x) arises from
the cumulative influence of the source energy–momentum tensor Tµν evaluated at points
(tr , y) lying on its past light cone, where the retarded time is given by tr = t − |x − y|.

2.1.1 Gravitational Radiation of Compact Objects

In many astrophysical scenarios, the gravitational source is spatially compact compared to
the distance from the source to the observation point. In such cases, one can expand the
integrand of the Green’s function solution (2.14) in a Taylor series about the field point,
retaining only the leading-order contribution

1 y i xi 3xi xj − r2 δij
 
1 i j
= + 3 +y y + ..., (2.15)
|x − y| r r r5

where r = |x| is the distance from the origin to the observation point. This is the multipole
expansion of the Newtonian potential. Applying this expansion to the linearized solution
(2.14) and keeping only the leading term yields the far-field approximation
Z
4G
hµν (x, t) = − d3 y Tµν (y, t − r) + O(r−2 ), (2.16)
r

which describes radiation that decays as 1/r, a characteristic feature of outgoing spherical
waves.
Z Each component of the integral above has a direct physical interpretation:
• T 00 d3 y = M c2 represents the total energy (mass) of the source,
Z
• T 0i d3 y = P i c represents the total momentum of the source along the xi direction,
Z
• T ij d3 y = Πij represents the spatially integrated internal stresses of the source, en-
compassing pressure and shear.
For an isolated system, conservation of the energy-momentum tensor implies ∇µ T µν =
0. From this, it follows that M and P i are constants of motion in the linear theory.
Moreover, for compact sources, the total momentum P i can be expressed as
Z 
i ∂ 00 0 i 3
P = T (y , y) y d y . (2.17)
∂y 0 y 0 =t−r

–8–
The total four-momentum P µ = (E, P i ) transforms as a Lorentz vector under changes of
inertial frame. By performing a Lorentz boost with velocity v = P /E, we arrive in the
center-of-momentum frame where

Pi = 0 (center-of-momentum frame). (2.18)

While the stress-energy tensor T µν (x) transforms covariantly under coordinate changes,
one cannot, in general, find a global inertial frame in which T 0i (x) = 0 at all points, only
the integrated momentum P i can be set to zero by a suitable choice of frame.
In the center-of-momentum frame, the gravitational field reveals its most transparent
form
4GM
h00 = − , (2.19)
r
h0i = 0, (2.20)
Z
4G
hij = − Tij (t − r, y) d3 y. (2.21)
r
We can simplify these expression for compact sources. Using integration by parts and
assuming that the stress-energy tensor vanishes outside the source, we write1
Z Z
ij 3 1 h i
T d y=− ∂k (T ik )y j + ∂k (T jk )y i d3 y. (2.22)
2
Now, applying energy-momentum conservation, ∂µ T µν = 0, we obtain
Z Z 
ik j 3 d i0 0 j 3
∂k (T )y d y = − 0 T (y , y)y d y . (2.23)
dy

Using conservation again and integrating by parts, we find

Z Z 
d
T i0 y j + T j0 y i d3 y = 0 T 00 (y 0 , y)y i y j d3 y .


(2.24)
dy

This motivates introducing the (mass) quadrupole moment tensor of the source
Z
Iij (y 0 ) = T00 (y 0 , y) y i y j d3 y. (2.25)

Thus, in the far-field limit, the spatial components of the metric perturbation become
2G ¨ 0
hij (t, x) = − Iij (y ), (2.26)
r
where dot denotes derivative with respect to y 0 = t−r. It is often convenient to decompose
this expression into its trace and traceless parts

hij (t, x) = h(t, r) δij + γij (t, x), (2.27)

1
Since the integration is performed outside the support of T µν , total derivative terms such as
∂k (T ik y j ) d3 y vanish.
R

–9–
where the traceless component is given by
2G
γij (t, x) = − Q̈ij (y 0 ), (2.28)
r
and Qij is the reduced quadrupole moment
1
Qij (y 0 ) = Iij (y 0 ) − δij Ikk (y 0 ). (2.29)
3
The perturbation γij is both transverse and traceless.
Up to now, we found that γij is a symmetric, transverse-traceless (TT) tensor. These
conditions ensure that gravitational waves are purely spatial, propagate perpendicular to
their direction of travel, and carry no longitudinal or scalar modes. This defines the so-
called TT gauge, characterized by the conditions

hTT
0i = 0 and hTT = 0, (2.30)

where hTT
ij = γij . We remain within the Lorenz gauge, which imposes additional constraints
in the TT gauge ∂ 0 hTT i TT
00 = 0, and ∂ hij = 0. For non-stationary (i.e., time-dependent)
sources, these conditions imply that hTT 00 also vanishes. Consequently, the gravitational
wave is fully described by the spatial components of a symmetric, transverse, and traceless
tensor, leaving exactly two dynamical degrees of freedom. We return to this discussion in
Sec. 2.2, where a more rigorous treatment is provided.

Radiation in Electromagnetism vs Gravity

Just as Maxwell’s equations predict the existence of electromagnetic waves, the lin-
earized Einstein equations reveal that gravity, too, supports wave-like solutions. How-
ever, unlike electromagnetic radiation, which begins at the dipole level, GWs first
arise at the quadrupole order, a distinction that may seem counterintuitive at first
glance. Let us pause for a moment and explore why. This fundamental difference
stems from the nature of the sources and the symmetries of the underlying field equa-
tions. In electromagnetism, dipole radiation is ubiquitous because charges, positive
and negative, can accelerate independently, even within an isolated system. For a
localized charge distribution ρ(t, y), the radiative electric field in the far zone can
be expressed in terms of the electric dipole moment pi (t) ≡ d3 y ρ(t, y) y i and the
R

radiative electric field takes the form

1 1  
Erad (t, x) = n̂ × n̂ × p̈(t − r) , (2.31)
4πε0 c2 r
where n̂ = x/r denotes the direction of propagation. Gravity, however, couples to
energy-momentum, which is always positive in mass-dominated systems. As a result,
momentum conservation guarantees a fixed mass dipole in the center-of-momentum
frame as Eq. (2.18), forbidding both monopole and dipole gravitational radiation.
To see how this constraint might be circumvented, consider a thought experiment
involving a gravitational dipole: a pair of positive and negative masses accelerating
apart. Such a system would self-accelerate, violating momentum conservation and

– 10 –
 eliminating the possibility of a global center-of-momentum frame. Though unphysi-
cal, this setup illustrates how gravitational dipole radiation could, in principle, arise
in a theory that allows negative mass.

Feature EM Linearized Gravity

Source jµ T µν
Wave equation □Aµ = j µ □hµν = −16πG Tµν
Conservation law ∇µ jµ =0 ∇µ T µν = 0
Leading radiative Dipole Quadrupole
Monopole × (charge conservation) × (energy conservation)
Dipole ✓ (leading order) × (momentum conservation)
Quadrupole ✓ ✓ (leading order)
Radiated field Eirad ∝ p̈i hTT TT
ij ∝ Q̈ij

Spin of carrier 1 (photon) 2 (graviton)

Table 2: Comparison between electromagnetic and gravitational radiation in the

linearized (weak-field) regime.

Figure 6: Left panel: the spherical coordinate frame along with its associated unit normal
vectors. Right panel: illustrates linearly and circularly polarized waves: in linear polariza-
tion, the deformation pattern in the transverse plane oscillates along fixed axes, while in
circular polarization, the deformation rotates uniformly within the transverse plane.

– 11 –
2.2 Polarization States of GWs and Free Particles
For a gravitational wave propagating in an arbitrary direction k = kn̂, the wave’s spatial
polarization tensor in the TT gauge can be extracted using the projection operator
 
TT k l 1 kl
eij (k̂) = Pi Pj − P Pij ekl (k) (polarization tensor), (2.32)
2
where the spatial projection tensor Pij is defined as Pij = δij − n̂i n̂j , and serves to project
any tensor onto the 2d plane orthogonal to the wave propagation direction n̂. Much like
in electromagnetism, the two independent dynamical modes of the gravitational wave can
be decomposed into distinct polarization states. Expressing n̂ = −r̂ in terms of polar
and azimuthal angles (θ, ϕ) as r̂ = (sin θ cos ϕ, sin θ sin ϕ, cos θ), we define the complex
polarization basis vectors
1  
e±
i (n̂) = √ θ̂ i ± i ϕ̂ i (polarization vector), (2.33)
2
where θ̂ and ϕ̂ are orthonormal unit vectors spanning the transverse plane, aligned re-
spectively with increasing θ and ϕ. Since r̂ points opposite to the direction of photon
propagation, the helicity (defined as the projection of the spin angular momentum along
the direction of motion h = S.n̂) is the negative of the projection along r̂. For this rea-
son, the labels e±
i correspond to photon states of helicity ∓1. These two helicity states
correspond to circular polarizations as
L,R
e±
i (n̂) = ei (n̂). (2.34)
From these, one can construct two equivalent bases for the polarization tensors (see Fig. 6):
• Linear polarization tensors, i.e. the familiar plus and cross modes
1   1  
e+
ij (n̂) = √ θ̂i θ̂j − ϕ̂i ϕ̂j , e×
ij (n̂) = √ θ̂i ϕ̂j + ϕ̂i θ̂j , (2.35)
2 2
• Circular polarization tensors, i.e. the right- and left-handed helicity states
eR R R
ij (n̂) = ei ⊗ ej , and eL L L
ij (n̂) = ei ⊗ ej . (2.36)

These two bases are related through the transformation:

R,L 1  + ×

eij (n̂) = √ e ± i eij , (2.37)
2 ij
showing that circular polarizations are simply complex combinations of the linear ones.
In the special case where the wave propagates along the z-axis, i.e., n̂ = ẑ, the linear
polarization tensors simplify to the following non-zero components:
1 1
e+ +
11 = −e22 = √ , e× ×
12 = e21 = √ , (2.38)
2 2
with all other components. For circular polarization, the non-zero components are:
1 i
eR,L R,L
11 = −e22 = √ , eR,L R,L
12 = e21 = ± √ , (2.39)
2 2
with vanishing longitudinal components.

– 12 –
 GW Polarizations in d Dimensions and Extended Frameworks
In d-dimensional General Relativity (GR), a massless spin-2 graviton carries

d(d − 3)
Npol = , (2.40)
2
independent polarization (radiative) degrees of freedom. This immediately implies
that GR admits propagating gravitational wave modes only for d ≥ 4. In particular,
for d = 3 one obtains Npol = 0, meaning that three-dimensional GR has no local grav-
itational radiation and no graviton polarization states. In contrast, many modified
theories of gravity introduce additional dynamical degrees of freedom, which generi-
cally lead to extra gravitational wave polarization modes beyond the two transverse
(+, ×) modes of four-dimensional General Relativity (GR). In particular, in four di-
mensions the metric begins with ten degrees of freedom; fixing the coordinate system
removes four, leaving six physical degrees of freedom corresponding to the six possible
gravitational wave polarizations. Thus, up to four supplementary modes (two scalar
and two vector polarizations) may arise, although a given theory of gravity need
not include all of them simultaneously. For a thorough review of how additional de-
grees of freedom arise and propagate in modified gravity frameworks, see [51]. More
general discussions of gravitational wave polarizations in alternative theories can be
found in [52, 53].

2.2.1 Geodesic Distortion by Passing GWs

To gain a clearer understanding of the physical manifestation of each GW polarization
state, it is instructive to examine their effect on the geodesic deviation between freely
falling test particles. Consider two nearby test particles that interact only through gravity,
initially at rest and separated by the spacelike vector

X µ (0) = (0, ξ0i ). (2.41)

The passage of a gravitational wave perturbs their geodesic motion, inducing a nonzero
contribution to the geodesic–deviation equation. Recalling that the evolution of the separa-
tion four–vector X µ between neighboring geodesics with tangent four–vector uµ is governed
by
D2 X µ
= Rµ νλσ uν uλ X σ , (2.42)
Dτ 2
where
D
≡ uµ ∇µ (2.43)
Dτ
is the covariant derivative along the particle’s worldline. In the local proper (detector)
frame associated with the reference geodesic, the motion is static to zeroth order and
receives corrections only at O(hµν ), so that

uµ = (1, 0, 0, 0) + O(hµν ). (2.44)

– 13 –
Therefore, the geodesic deviation of the nearly particles is
D2 X i
= −Ri 0j0 X j . (2.45)
Dτ 2

Figure 7: Effect of GW polarization states on a ring of freely falling particles. Solid

lines and red dots show particle positions over time; dashed lines mark their unperturbed
locations. Top: plus and cross polarizations. Bottom: right- and left-handed circular
polarizations. Image credit: Adapted and modified from [54].

In the rest frame of the particle, we have dxi = 0, and the line element reduces to

ds2 = −dτ 2 = − 1 + h00 dt2 .

(2.46)

In the TT gauge, hTT

00 = 0 at leading order, so the coordinate time and the particle’s proper
time coincide up to second order in the perturbation,

τ = t + O hTT 2


ij . (2.47)

Using this, the geodesic deviation simplifies to

d2 X i (t) 1
2
= ∂t2 hTijT ξ0j , (2.48)
dt 2

– 14 –
which integrates to the displacement of the separation vector
1
X i (t) = ξ0j δij + hTijT (t, x) .


(2.49)
2
That shows a passing gravitational wave induces a deformation of the separation vector
between freely falling test particles. Fig. 7 illustrates this tidal effect for each of the
four possible polarization states of gravitational waves, showing how the ring of freely
falling particles is periodically stretched and compressed as the wave propagates through
spacetime.

Gravitational Memory Effect2

Before proceeding to the next part, let us pause to discuss a fascinating phenomenon;
the remarkable way in which spacetime itself remembers the passage of GWs, leav-
ing behind a permanent trace of the universe’s most violent events. When a burst
of gravitational radiation carries a non-oscillatory component, i.e. the waveform ex-
hibits a net offset ∆hTTij ̸= 0, the spacetime curvature leaves a lasting imprint on
freely falling test particles. After the wave has passed, the geodesic deviation in
Eq. (2.49) no longer returns to zero but retains a permanent displacement,

∆X i ≃ 1
2 ∆hTT i j ξ j (0) ̸= 0. (2.50)

This enduring change in the relative separation of particles, known as the gravi-
tational memory effect [55–57], reveals that spacetime itself retains memory of the
passing gravitational wave. Although one may formally remove this step in the metric
by performing a large diffeomorphism, such as a late-time spatial shear,

xi −→ xi + 21 σ i j xj , σ i i = 0, σ i j = − ∆hTT i j , (2.51)

this transformation is not a mere gauge redundancy. Because it does not vanish at
null infinity, it maps the geometry between inequivalent asymptotic vacua [58, 59].
Consequently, the memory effect stands as a genuine observable: a permanent, mea-
surable shift in relative distance or interferometric phase, encoding the asymptotic
difference between the early and late gravitational wave configurations. The gravi-
tational wave memory effect, first explored in gravitational physics by Zel’dovich &
Polnarev [55], was later developed in depth by many others [56, 57]. Observational
prospects for detecting the memory effect have been proposed using ground-based
interferometers such as LIGO [60], space-based detectors including LISA [61], as well
as through observations with Pulsar Timing Array experiments [62]. We will re-
turn to this connection when we examine the infrared structure of asymptotically
flat spacetimes in Sec. 3.4.2, and again in the context of cosmology and Weinberg’s
adiabatic modes in Sec. 4.3.

– 15 –
3 Theoretical Aspects of Gravitational Waves

In the previous section, we employed the weak-field approximation to derive gravitational

waves. This approach dates back to Einstein himself, who was the first to identify grav-
itational waves as solutions within the linearized GR theory. Yet, due to the conceptual
and mathematical ambiguities of the time, he later questioned their physical reality within
the full nonlinear theory. This uncertainty slowed progress for decades, until the pioneer-
ing works of Bondi, Pirani, Robinson, and Trautman finally resolved it, showing, through
a fully consistent physical and mathematical treatment of asymptotically flat spacetimes,
that gravitational waves are real, radiative solutions that carry energy and angular momen-
tum. In this section, we will first discuss the theoretical issues and ambiguities of linearized
gravity. Next, we will examine the algebraic structure of the Weyl tensor and its geometric
interpretation as the carrier of tidal forces. Finally, we will conclude by exploring how
spacetime geometry can radiate, leading to a precise definition of GWs.

3.1 Theoretical Issues of Linear Gravity

Einstein’s field equations are intrinsically nonlinear

Gµν [g] = Rµν − 12 gµν R = 8πG Tµν . (3.1)

The Einstein tensor Gµν [g] is a nonlinear functional of the metric gµν and its derivatives.
Hence, spacetime curvature couples to itself through the very structure of the equations.
Let us now take a more rigorous look at the weak-field approximation introduced in Sec. 2.
We consider small perturbations hµν around flat spacetime

gµν = ηµν + hµν , |hµν | ≪ 1. (3.2)

Expanding the Einstein tensor in powers of hµν gives

Gµν [g] = G(1) (2) 3

µν [h] + Gµν [h] + O(h ), (3.3)
(1) (2)
where Gµν is linear and Gµν is quadratic in hµν . Substituting this into Einstein’s equations
yields

G(1) (2) 3
µν [h] = 8πG Tµν − Gµν [h] + O(h ). (3.4)
(2)
The quadratic term Gµν acts as an additional source, and one may define the effective
stress-energy tensor of the gravitational field as
1
t(grav)
µν =− G(2) [h]. (3.5)
8πG µν
The field equations can then be written in the form
 
G(1)
µν [h] = 8πG Tµν + t (grav)
µν . (3.6)

This relation makes manifest that the gravitational perturbation itself contributes to the
total source of curvature, encapsulating the idea that gravity gravitates. Owing to its

– 16 –
universal coupling to all forms of energy, including its own, the gravitational field renders
general relativity inherently nonlinear. A further subtlety concerns the condition |hµν | ≪
1 (and similarly for its derivatives). The metric, however, is not a directly observable
quantity; its components depend on the chosen coordinates. By an appropriate coordinate
transformation, the same physical spacetime can appear with metric components that are
either large or small. Thus, the weak-field limit must be defined in terms of coordinate-
invariant physical quantities.
These considerations lead to several key conceptual questions:

i) Do the full, nonlinear Einstein equations admit propagating solutions that can mean-
ingfully be interpreted as gravitational waves?

ii) In the weak-field limit, do these solutions reduce to the familiar linearized waves?

iii) How can one consistently define gravitational energy and momentum in a generally
covariant theory where energy cannot be localized?

iv) Given that the metric is not itself an observable, how should one formulate a physi-
cally meaningful and coordinate-invariant definition of the weak-field limit?

For decades, the physical reality of gravitational waves remained clouded due to the
above conceptual ambiguities. A decisive turning point came with F. Pirani’s insightful
analysis [63], which linked the Weyl tensor to measurable tidal effects and inspired a new
understanding of radiation in general relativity. Building on these ideas, H. Bondi in
1957 introduced the first mathematically precise definition of gravitational waves [2]. His
pioneering work, together with subsequent contributions by I. Robinson and A. Trautman,
established that the full, nonlinear Einstein equations admit genuine radiative solutions
that carry energy and angular momentum to infinity. In what follows, we briefly review
these seminal developments.

3.2 The Weyl Tensor: Geometry of Tidal Forces

Even before Bondi’s seminal Nature paper in 1957 [2], Felix Pirani made a brilliant early
attempt to define gravitational waves in purely geometric terms [63]. His insight was to
analyze the algebraic structure of the Weyl tensor through the Petrov classification as
a means to identify radiative regions of spacetime. In doing so, Pirani proposed that
the physicality of gravitational waves should be anchored not in coordinate-dependent
metrics, but in the measurable tidal effects encoded in the curvature tensor, i.e. in the
relative acceleration of nearby geodesics governed by the Weyl tensor in vacuum. This
idea elegantly shifted the focus from perturbative metrics to intrinsic geometry, offering a
criterion that applies even in exact solutions. The approach is well motivated: far from
the source, the Ricci tensor vanishes, and the full Riemann tensor reduces to its traceless
part which is the Weyl tensor (see Fig. 8). In what follows, we briefly outline Pirani’s idea
and the resulting geometric characterization of radiative spacetimes.
In 4d, the Riemann curvature tensor possesses 20 independent components. Of these,
10 correspond to the Ricci tensor, while the remaining 10 are captured by the traceless

– 17 –
Figure 8: The Weyl tensor encodes the free gravitational field, tidal shear and distortion,
independent of matter. To highlight the behavior of the Weyl tensor in different regimes,
we consider two illustrative examples: the Schwarzschild solution, which is non-radiative
and exhibits a leading-order Weyl tensor decay of O(1/r3 ), and a radiative binary black
hole merger, where the leading-order Weyl tensor component falls off as O(1/r) in the
far-field (null infinity) limit.

Weyl tensor, which encodes the conformal structure of spacetime.

1
Cµνλσ = Rµνλσ − (gµ[λ Rσ]ν − gν[λ Rσ]µ ) + R gµ[λ gσ]ν . (3.7)
3
More precisely, it is unchanged under a conformal transformation of the metric
′
gµν 7→ gµν = Ω2 gµν , (3.8)
′
Cµνλσ 7→ Cµνλσ = Cµνλσ . (3.9)
′
Note that Cµνλσ = Ω2 Cµνλσ . More intuitively, the Weyl tensor expresses the tidal forces
that a free-falling body feels along a geodesic. However, unlike the Ricci tensor, it does
not have information about the change of the volume, but only how the shape of the body
is distorted by the tidal forces.
In four dimensions, the Weyl tensor can be viewed as a linear operator acting on the
six-dimensional space of antisymmetric rank-two tensors (bivectors),

(C ·B)µν = 1
2 Cµν αβ Bαβ , Bµν = −Bνµ . (3.10)

A bivector represents an oriented two-plane in the tangent space and can be expressed as

Bαβ = v[α wβ] = vα wβ − vβ wα . (3.11)

Formally, one considers Eq. (3.10) and seeks its eigenbivectors and corresponding eigenval-
ues,
1 µν
2 C αβ Xµν = λ Xαβ . (3.12)
Among them, we are interested in eigenbivectors, constructed from null vectors play a
privileged role: they are invariant under rescaling of lµ and encode the geometry of lightlike
directions. At any given point in spacetime, the Weyl tensor can possess at most four

– 18 –
linearly independent eigenbivectors, each corresponding to a null direction in the underlying
geometry known as the principal null directions (PNDs). Physically, their importance stems
from the fact that the causal structure of a Lorentzian geometry is determined entirely by
its null cones: null vectors specify the propagation of signals, gravitational radiation, and
the horizons of black-hole spacetimes. In particular, by examining how the Weyl tensor
maps null bivectors into themselves, one probes directly how the free gravitational field
influences the focusing, twisting, and shearing of null congruences.
We now show how a single null direction lµ can be completed into a full null tetrad.
First choose a second null vector nµ such that lµ nµ = −1. The 2-dimensional subspace
orthogonal to both k µ and nµ , is spacelike and admits an orthonormal basis {eµ1 , eµ2 }. From
these one constructs two complex null vectors
1 1
mµ = √ (eµ1 + i eµ2 ) , m̄µ = √ (eµ1 − i eµ2 ) , (3.13)
2 2
satisfying mµ mµ = m̄µ m̄µ = 0, mµ m̄µ = 1, and lµ mµ = nµ mµ = 0. The four null vectors
{lµ , nµ , mµ , m̄µ } form a null tetrad. 3 Once the tetrad is constructed, we can introduce
bivectors built from lµ as

Xµν = l[µ mν] , (3.14)

and its square satisfies Xµν X µν = 0. Similarly, its complex conjugate

X̄µν = l[µ m̄ν] , (3.15)

is also a simple null bivector.

The principal null directions and their degeneracies give rise to the Petrov classification,
thereby linking the algebraic structure of curvature to the causal geometry of spacetime.
Originally introduced by A. Petrov in 1954 [65] and later reinterpreted by F. Pirani [63],
the Petrov classification characterizes the possible algebraic symmetries of the Weyl tensor
at each spacetime point. Now you may ask why do these special null directions matter
physically? The answer lies in how the curvature acts on the relative motion of neighboring
geodesics—particularly null geodesics that represent bundles of light rays or gravitational
waves. Consider a smooth one-parameter family of null geodesics xµ (λ, s), where λ is an
affine parameter along each geodesic and s labels the distinct geodesics in the congruence.
The tangent vector is obtained by differentiating with respect to λ at fixed s, while the
geodesic–deviation vector arises from differentiating with respect to s at fixed λ, i.e.
∂xµ ∂xµ
lµ ≡ , ξµ ≡ . (3.16)
∂λ s ∂s λ
These vectors satisfy the commutation relation

[l, ξ]µ = lν ∇ν ξ µ − ξ ν ∇ν lµ = 0, (3.17)

3
The null tetrad formalism is the central geometric structure of the Newman–Penrose approach [64] and
underlies its spin-coefficient representation of curvature and gravitational radiation.

– 19 –
Figure 9: A congruence of null geodesics xµ (λ, s) labelled by the affine parameter λ and
the transverse parameter s. The tangent vector along each geodesic is lµ . The separation
between neighbouring geodesics is described by the geodesic–deviation vector ξ µ . The
screen space is the two-dimensional subspace orthogonal to both lµ and the auxiliary null
vector nµ , which at each fixed value of λ is spanned by an orthonormal basis {eµ1 , eµ2 }.
Within this screen the cross-section of the light bundle is defined, along with the expansion,
shear, and twist.

which holds for any smooth congruence. The screen space (or optical plane) is the 2-
dimensional subspace orthogonal to both lµ and nµ . It is spanned by a pair of orthonormal
vectors {eµ1 , eµ2 }, which serve as a basis for projecting tensors and defining the transverse
degrees of freedom. The screen projector onto this 2D subspace is defined by

hµν = gµν + lµ nν + nµ lν , (3.18)

which projects any vector onto the space orthogonal to both lµ and nµ (see Fig. 9). Next
we define the optical tensor as the doubly projected derivative

Θµν ≡ hµ α hν β ∇β lα . (3.19)

This is the object that decomposes into

Θµν = 21 θ hµν + σµν + ωµν , (3.20)

giving expansion, shear, and twist of the null congruence as 4

θ ≡ hµν ∇ν lµ , σµν ≡ B(µν) − 12 θ hµν , ωµν ≡ B[µν] . (3.21)

4
Frobenius’ theorem states that a vector field v µ is hypersurface orthogonal iff it satisfies the integrability
condition v[µ ∇ν vρ] = 0. Consequently, hypersurface–orthogonal geodesic congruences possess vanishing
vorticity, ωµν = 0 [66].

– 20 –
In vacuum, the Riemann tensor reduces to the Weyl tensor, so the relative acceleration
between two nearby null geodesics is governed by
D2 ξ µ
= − C µ νρσ lν ξ ρ lσ . (3.22)
dλ2
The right-hand side encodes how spacetime curvature focuses, shears, or twists the light
bundle. When projected onto the two-dimensional screen space orthogonal to lµ (the
subspace spanned by the complex null vectors mµ and m̄µ ), this equation splits into an
isotropic focusing (the expansion) and an anisotropic distortion (the shear). If lµ is a PND
with eigenvalue λ, then inserting the eigenbivector relation Eq. (3.12) together with the
explicit form of the null bivector Eq. (3.14) yields
D2 ξ µ
= λ(ξ.l) lµ . (3.23)
dλ2
Geometrically, this shows that along a PND the Weyl curvature produces no transverse
tidal distortion: the optical tensor has vanishing shear, so the cross-section of a narrow
bundle of null geodesics remains circular to first order. Thus a PND generates a locally
shear–free null congruence. If the PND is repeated (double, triple, or quadruple), this
shear–free property extends to an entire null congruence whose integral curves form a
geodesic, shear–free null vector field throughout the spacetime [67, 68].

Petrov Classification
It organizes the possible algebraic symmetries of the Weyl tensor into precisely six
distinct categories, known as the Petrov types presented in Table 3 and summarized
in the left panel of Fig. 10.

Petrov Type Conditions on PNDs Spacetime

I 4 simple General vacuum
II 2 simple + 2 coincide
D 2 pairs coincide Stationary, asymptotically flat
III 1 simple + 3 coincide
N 4 coincide Gravitational radiation
O Weyl tensor vanishes Conformally flat

Table 3: Petrov Classification of Spacetime, ordered from the most general to the
most symmetric.

Near localized massive sources, the spacetime is typically of Petrov type I, reflecting
its most general algebraic structure. As one moves outward, higher-order terms in
the Weyl tensor successively peel off, and the spacetime transitions through types II
and III, ultimately approaching type N at future null infinity where only the radiative

– 21 –
 component of the gravitational field remains. Type D regions are associated with the
gravitational fields of isolated massive objects, e.g., stars and black holes, which is
entirely characterized by its mass and angular momentum. The two coincided PNDs
present radially ingoing and outgoing null congruences near the object. Type O
regions are conformally flat places with zero Weyl tensor, e.g., exact Minkowski and
FRW. In this case, any gravitational effects must be due to the immediate presence
of matter or the field energy of some non-gravitational field. Type N regions are
those regions with transverse gravitational radiation. A spacetime region is type N,
if and only if there exists a null vector, kµ , such that

Cµνλσ k σ = 0. (3.24)

The four coinciding PNDs are given by this null vector which is the wave vector of
the propagating gravitational wave.

Peeling Theorem
The Petrov type of a spacetime need not be uniform: it may vary from region to
region. The right panel of Fig. 10 illustrates how the algebraic type can change along
outgoing null directions. This effect is due to the peeling theorem in general relativity
[69], which describes the asymptotic behavior of the Weyl tensor as one goes to null
infinity. Let γ be a null geodesic from a point p to null infinity, with affine parameter
r. The peeling theorem states that, as r approaches infinity
(1) (2) (3) (4)
Cabcd Cabcd Cabcd Cabcd
Cabcd = + 2 + 3 + 4 + ..., (3.25)
r r r r
(1) (2) (3) (4)
where Cabcd is type N, Cabcd is type III, Cabcd is type II and Cabcd is type I [66]. To
summarize, the Weyl tensor of a radiative spacetime must be of type N very far from
the sources, i.e. in the asymptotic future.

3.3 When Geometry Radiates: Defining Gravitational Waves

Herman Bondi in his Nature paper [2] followed by a subsequent paper by Bondi, Pirani, and
Robinson [70], provided the first mathematically precise definition of gravitational waves in
the full Einstein equation. Moreover, he proved that gravitational radiation carries energy.
In what follows, we review these results in the context of asymptotically flat spacetimes.
Throughout this discussion in Secs. 3.3 and 3.4, we work in Bondi retarded coordinates
(u, r, z, z̄), which are well suited for describing radiation propagating toward null infinity.
These coordinates are related to the standard spherical coordinates (t, r, θ, ϕ) by

θ iϕ θ −iϕ
u = t − r, z = cot e , z̄ = cot e . (3.26)
2 2

– 22 –
Figure 10: Left panel: The schematic form of the Petrov classification of the Weyl tensor.
The blue arrows correspond to each of the principal null directions and the number on
arrows represent the number of coinciding PNDs. Right panel: The Petrov type of the
gravitational field varies along outgoing null directions: it is type I near localized sources,
successively transitions through types II and III, and approaches type N at future null in-
finity, where only the radiative component remains. The blue arrow indicates the evolution
of the Weyl tensor’s Petrov type as one moves toward null infinity in an asymptotically
flat spacetime.

The coordinate u denotes the Bondi (retarded) time and labels outgoing null hypersurfaces.
For comparison, the advanced time in Minkowski spacetime is given by

v = t + r. (3.27)

The angular coordinates on the sphere are collectively denoted by ΘA = (z, z̄) and provide
a complex parametrization of the unit two-sphere. In these coordinates, the metric on the
sphere takes the form
2
γz z̄ = . (3.28)
(1 + z z̄)2

In this coordinate system the Minkowski spacetime metric is

ds2 = −du2 − 2dudr + 2r2 ηz z̄ dzdz̄. (3.29)

Before proceeding further, we need to introduce three essential geometric notions that
underlie the analysis of gravitational radiation in general relativity.

• Asymptotic flatness: An asymptotically flat spacetime is a Lorentzian manifold in

which the curvature vanishes at large distances from some region so that the geometry
becomes indistinguishable from Minkowski. Outside the source, the Ricci tensor is
zero, and therefore, the asymptotic flatness imposes asymptotic falloff conditions on
the Weyl tensor as Eq. (3.25).

– 23 –
Figure 11: Penrose diagram of an asymptotically flat spacetime. The future and past
null infinities, I ± , are parametrized by the retarded and advanced Bondi times u and v.
The blue ring, i0 , denotes spatial infinity. The gray cone in the upper half represent null
surface at u = u0 . The orange wavy lines indicate outgoing gravitational waves (GWs)
propagating toward future null infinity.

• Future null infinity: I + is defined as the endpoints of all future-directed null geodesics
along which r → ∞. This null surface is the product of S 2 with a null line u taking
values in R. Each null hypersurface, σu0 , intersects I + in a 2-sphere with u = u0 .
(See Fig. 11)

• The topology of I + , is that of a null hypersurface with the structure R × S 2 , where

the R factor corresponds to the retarded time and the S 2 factor represents angular
directions. At each fixed retarded time, the cross-sections of I + are 2-spheres, forming
the so-called celestial sphere CS2 .

Now, we want to study gravitational theories in which the metric is asymptotic, but
not exactly equal to, flat metric, and we abbreviate ΘA = (z, z̄). A natural starting point
is to impose the Bondi gauge,
g 
AB
grr = 0, grA = 0, ∂r det = 0, (3.30)
r2
which fixes the four local diffeomorphism degrees of freedom associated with (r, u, ΘA ).
Under these conditions, the most general four-dimensional asymptotically flat metric takes
the form
  
2 2 2β A 1 A B 1 B
ds = −U du − 2e dudr + gAB dΘ + U du dΘ + U du . (3.31)
2 2
where r is the luminosity distance and the metric coefficients appear in the six functions.
Here U A contributes two angular components and gAB , constrained by det(gAB /r2 ) =

– 24 –
det ηAB , carries two independent degrees of freedom. Thus the Bondi gauge fixes four dif-
feomorphisms and leaves precisely six independent metric functions, the expected number
for a general spacetime metric modulo coordinate freedom. Any geometry can be described
locally by the metric Eq. (3.31). In Bondi gauge the hypersurface Einstein equation Gur = 0
fixes the function β rather than leaving it as free data. Using the determinant condition
det gAB = det ηAB and the asymptotic expansion, we have
γAB
gAB = ηAB + + O(r−2 ), (3.32)
r
and the constraint equation Gur = 0 implies the standard hypersurface relation
r AC BD
∂r β = g g (∂r gAB )(∂r gCD ), (3.33)
16
as derived in [59, 71, 72] which integrates to

1 γAB γ AB
β(u, r, xA ) = β∞ (u, xA ) − + O(r−3 ). (3.34)
32 r2
Asymptotic flatness fixes the integration function β∞ (u, xA ): since gur = −1 as r → ∞,
hence β∞ = 0. Thus β is not an independent degree of freedom but is completely deter-
mined by the shear γAB , vanishing at null infinity and beginning at order 1/r2 . Therefore,
in Bondi gauge and after imposing the constraint equation, one degree of freedom is elim-
inated and we are left with five independent degrees of freedom.
Imposing the asymptotic flatness condition at large r with fixed (u, z, z̄) leads to falloff
conditions on the metric components. For the natural choice made by Bondi, van der Burg,
Metzner, and Sachs (BMS) [58], the large-r structure of the metric is constrained to be 5

ds2 = −du2 − 2dudr + 2r2 ηz z̄ dzdz̄

2mB 2
+ du + rγzz dz 2 + rγz̄ z̄ dz̄ 2 + Dz γzz dudz + Dz̄ γz̄ z̄ dudz̄ (3.35)
r 
1 4 1 zz
+ (Nz + u∂z mB ) − ∂z (γzz γ ) dudz + c.c. + . . . ,
r 3 4

where Dz is the covariant derivative with respect to ηz z̄ while γzz , mB and Nz are r
independent and functions of (u, z, z̄). The first three terms in Eq. (3.35) are simply the
flat Minkowski metric, and the remaining terms are the leading corrections:

• The first quantity mB (u, z, z̄) is the Bondi mass aspect (for Kerr BH, mB = GM ),

• The next one is Nz (u, z, z̄), which is the Bondi angular momentum aspect (for a Kerr
BH Nz = 2GmL),
5
Compared with the notation adopted in Strominger’s book [73], we make a slight adjustment for
consistency with the conventions used throughout this review. In particular, what is written as Czz in [73]
is denoted here by γzz for gravitational waves, and their γzz̄ is written here as ηzz̄ . The conventions of
[73] follow those standard in the infrared-gravity and BMS literature, whereas in this work we adopt the
notation commonly used in cosmology, expressed in terms of complex spherical–angle coordinates.

– 25 –
 • The last term is γzz which describes the gravitational waves. This quantity is trans-
verse to I + and r−1 -suppressed comparing to the dominant orders. The Bondi news
tensor is defined as

Nzz = ∂u γzz , (3.36)

which is the gravitational analogue of the field strength in gauge field theories, i.e.
Fuz = ∂u Az .

For later convenience, let us recall a simple but useful identity. On each cut of future null
infinity Su2 , endowed with the standard measure d2 z ηz z̄ , any covariant total derivative on
the sphere integrates to zero. Explicitly, for any smooth vector field V A on S 2 ,
Z
d2 z ηz z̄ DA V A = 0. (3.37)
S2

Post-Newtonian Weak-Field, Slow-Rotation Geometry (Lense–Thirring

Metric)
To build intuitive insight into the structure of the Bondi metric, it is useful to begin
with the line element describing the weak-field, slow-rotation geometry of an isolated
gravitating system [74] In this post-Newtonian regime the metric takes the form

2GM
2 4G 2GM
ds2 = − 1 − )δij dxi dxj .
 
dt + 2 (r̂ × J ) · d⃗r dt + (1 + (3.38)
r r r
The above metric contains the the following terms

• M : the total ADM/post-Newtonian mass (monopole),

• J : the total ADM angular momentum (dipole).

Unlike the Bondi metric introduced in Eq. (3.35), the spacetime above is stationary:
neither M nor J varies in time. In this sense, M represents the static, isotropic limit
of the Bondi mass aspect mB (u, z, z̄). The gravitomagnetic interaction responsible
for frame dragging
4G

g0i = 2
r̂ × J i , (3.39)
r
encodes the characteristic dipolar 1/r2 falloff produced by the angular momentum
J . In the Bondi description, the corresponding contribution appears in
Nz
guz = + ··· , (3.40)
r
where Nz is the Bondi angular-momentum aspect, carrying precisely the dipole mo-
ment associated with J .

– 26 –
3.3.1 GWs Carry Energy
The Bondi mass at a Bondi time, u1 , is defined as the integral over Su21 (the sphere with
u1 at I + ), as
Z
1
MB (u1 ) = d2 z ηz z̄ mB (u1 , z, z̄), (3.41)
4πG S 2
which is positive and time-dependent, such that it is always non-increasing with time.
Moreover, in the limit u → −∞, Su2 asymptotically approaches the spatial infinity, i0 , and
the Bondi mass is equal to the (conserved) ADM mass [75, 76]

MADM = lim MB (u). (3.42)

u→−∞

The time evolution of mB is given by the Einstein equation component Guu at I + as

1  2 zz  1
Dz N + Dz̄2 N z̄ z̄ − Nzz N zz − 4πG lim r2 Tuu
M
 
∂u mB = , (3.43)
4 4 r→∞
M is the matter field’s energy-momentum tensor. Using the Einstein equation in
where Tuu
M ∼ O(r −1 ) at future null infinity, we
(3.41) and considering a compact source with r2 Tuu
have
1 u2
Z Z
MB (u2 ) − MB (u1 ) = − du d2 z ηz z̄ Nzz N zz , (u2 > u1 ), (3.44)
4 u1
in which the Dz2 N zz terms vanishes under the S 2 integral. This is the famous Bondi mass-
loss formula, which measures the amount of mass-loss after some radiation through I + (see
the Penrose diagram in Fig. 11). That is zero in case the Bondi news vanishes. Otherwise,
the Bondi mass is decreasing with time in the form of gravitational radiation. 6

3.3.2 GWs Carry Angular Momentum

The Bondi angular momentum associated with the i-th rotation generator Y(i) A at a Bondi

time u1 is defined by
Z
1
Ji (u1 ) = d2 z ηz z̄ Y(i)
A
(z, z̄) NA (u, z, z̄), (3.45)
8πG S 2
A (z, z̄) denotes the three rotation Killing vectors of the round sphere can be written
where Y(i)
in the geometric form
A
Y(i) (z, z̄) = εijk r̂ j DA r̂ k , (3.46)

where r̂ i is the unit radial vector on S 2 . They generate the SO(3) isometries of S 2 and each
A corresponds to a rotation about one of the three Cartesian axes in R3 . The evolution
Y(i)
of the angular-momentum aspect is governed by the Bondi constraint equation
1 1 1
∂u Nz = ∂z (Dz2 γ zz − Dz̄2 γ z̄ z̄ ) − u∂u ∂z mB + ∂z (γzz N zz ) + γzz Dz N zz
4 4 2
2 M
− 8πG lim [r Tuz ]. (3.47)
r→∞
6
The Bondi mass and the mass loss formula were generalized for spacetimes with non-zero cosmological
constant in [77].

– 27 –
Integrating the above, similar to the mass aspect, one can find the amount of angular
momentum carried by the GWs as
Z
1
∂u Ji (u1 ) = d2 z ηz z̄ γzz (u, z, z̄) Y(i)
A
(z, z̄) DA N zz (u, z, z̄). (3.48)
16πG S 2
Integrating this relation over retarded time yields the total change in Bondi angular mo-
mentum between two cuts u1 and u2 of future null infinity
Z u2 Z
1
Ji (u2 ) − Ji (u1 ) = − du d2 z ηz z̄ Y(i)
A
(z, z̄) γzz (u, z, z̄) DA N zz (u, z, z̄).(3.49)
16πG u1 S2

This quantity is known as the Bondi angular-momentum flux and represents the net angular
momentum carried away by GWs through future null infinity during the interval [u1 , u2 ]. In
particular, a nonzero flux implies that the outgoing radiation transports angular momentum
from the source to I + , thereby reducing the Bondi angular momentum of the system. Thus
gravitational waves, in addition to carrying energy and momentum, also carry angular
momentum to null infinity.

Energy–Momentum Tensor and Flux of GWs

To get a better understanding of the results we found above, let us define the ef-
fective energy–momentum tensor of gravitational waves and the associated energy
and angular-momentum fluxes. In the short-wavelength (Isaacson) approximation,
gravitational waves behave like an effective radiative field whose energy–momentum
tensor is
1 ¯ ¯ ν γαβ .
Tµν
GW
= ∇µ γαβ ∇ (3.50)
32πG
For an observer with four-velocity uµ , the energy density of gravitational waves is
defined as
GW µ ν
ρGW = Tµν u u . (3.51)

In nearly flat spacetime, choosing uµ = (1, 0, 0, 0) gives the standard expression

1
ρGW = γ̇ij γ̇ij . (3.52)
32πG
That is related to the right-hand side of Eq. (3.44). The angular momentum carried
by gravitational waves is characterized through the associated flux. For rotations
about the i-axis, the angular-momentum flux is
Z
dJi 1
=− ϵijk dΩ r2 γ̇jℓ γkℓ . (3.53)
dt 32πG
The above flux is related to the right-hand side of Eq. (3.48). These relations encode
the essential energetic and dynamical content of gravitational radiation in the wave
zone and provide the foundation for computing the energy and angular momentum
carried away from compact astrophysical systems by the gravitational field.

– 28 –
3.4 At the Edge of Spacetime: From Poincaré to BMS
In their pioneering analysis of gravitational waves, Bondi, van der Burg, Metzner [58]
and Sachs [78] set out to characterize the asymptotic symmetry group of spacetimes ap-
proaching Minkowski space at future null infinity I + . Rather than recovering the expected
Poincaré symmetry, they uncovered an infinite-dimensional symmetry, the BMS group.
This surprising result revealed that the gravitational field possesses a far richer asymp-
totic structure than previously anticipated, with far-reaching implications for radiation,
gravitational memory, and the infrared behavior of gravity [79]. Crucially, this richness
is intimately connected to the presence of gravitational waves and their imprints at null
infinity—a connection we will return to and explore in more detail below. In what follows,
we present a brief summary of this historical development. For a more in-depth exposition,
see [73, 80].

3.4.1 Asymptotic symmetries of Gravitational Fields

The asymptotic symmetry group of gravitational fields in asymptotically flat spacetimes
is defined as the set of vector fields whose Lie derivatives preserve both the Bondi gauge
Eq. (3.30) and asymptotic flatness conditions. More concretely, for a vector field ξ µ of the
form

ξ = ξ u ∂u + ξ r ∂r + ξ A ∂A , (3.54)

the Bondi gauge and fall-off preserving conditions require the following asymptotic expan-
sions
u
ξ u = f (z, z̄) + DA Y A (z, z̄) + O(r−1 ), (3.55)
2
1
ξ A = Y A (z, z̄) − DA f (z, z̄) + O(r−2 ), (3.56)
r
r 1
ξ r = − DA Y A (z, z̄) + D2 f (z, z̄) + O(r−1 ), (3.57)
2 2
where f (z, z̄) is an arbitrary function on CS 2 and the vector field Y A (z, z̄) is a conformal
Killing vector on the 2-sphere, obeying

DA YB + DB YA = γAB DC Y C . (3.58)

This equation implies that the Lie derivative of the metric takes the conformal form

LY γAB = Ω γAB , Ω ≡ DC Y C , (3.59)

so that the trace gives the conformal factor explicitly. On the round sphere S 2 , Eq. (3.58)
admits exactly six independent solutions. This follows from the fact that

Conf(S 2 ) ∼
= SL(2, C), (3.60)

whose Lie algebra sl(2, C) is isomorphic to so(3, 1), the six-dimensional Lorentz algebra
consisting of three rotations and three boosts. The three rotational generators appear as

– 29 –
true Killing vectors on the celestial sphere: they preserve the round metric and therefore
satisfy

LY γAB = 0 (rotations). (3.61)

The remaining three generators correspond to Lorentz boosts. Boosts do not preserve the
spatial metric, since they mix time and space, but they induce Möbius transformations on
S 2 and therefore are conformal Killing vectors. Hence the six generators of the Lorentz
group, forming so(3, 1), act on the celestial sphere as the six global conformal symmetries
of CS 2 .
Having identified the the Lorentz algebra so(3, 1) on CS 2 , one may naturally ask:
where are the translations of 4d Minkowski space? Remarkably, translations from the
scalar function f (z, z̄) that parametrizes supertranslations. Unlike the Lorentz transfor-
mations, which are fixed by the conformal geometry of the sphere, the function f (z, z̄) is
completely arbitrary. It assigns to each null direction on the celestial sphere an independent
shift along the null generators of I + . In this sense, translations correspond to the lowest
spherical-harmonic modes of an infinite-dimensional space of functions on S 2 . Since f (z, z̄)
is an arbitrary function of angle on the celestial sphere, it can be expanded in spherical
harmonics,
∞ X
X ℓ
f (z, z̄) = fℓm Yℓm (z, z̄). (3.62)
ℓ=0 m=−ℓ

The ordinary spacetime translations arise from the four lowest modes with ℓ = 0 and ℓ = 1.
All higher multipoles ℓ ≥ 2 generate genuine supertranslations, which deform the structure
of I + in an angle-dependent manner (see Fig. 12). More precisely, a supertranslation is
defined by a smooth function f (n̂) on the celestial sphere that shifts retarded time according
to

u 7→ u + f (n̂), (3.63)

linking the geometry of CS 2 to the infinite-dimensional BMS group.

Structure of the BMS4 Algebra

The four-dimensional BMS algebra is a semidirect product between the Lorentz alge-
bra and the infinite-dimensional abelian algebra of supertranslations. Let Tf denote a
supertranslation with parameter f (θ, ϕ) and let Y denote a Lorentz generator. Their
commutation relations are

[Tf , Tf ′ ] = 0, [Y, Tf ] = TY·f , (3.64)

showing that Lorentz transformations act nontrivially on the space of supertransla-

tions by mixing the spherical-harmonic components of f . This yields the semidirect
product7

BMS4 = SL(2, C) ⋉ T , (3.65)

– 30 –
Figure 12: Illustration of the supertranslation action f (z, z̄) ∂u on the celestial sphere at
fixed Bondi time u0 . Each direction on I + undergoes an angle-dependent shift along the
null generators, highlighting the geometric meaning of BMS4 supertranslations.

where T is the group of smooth functions on the celestial sphere CS 2 under ad-
dition. Thus BMS4 extends the Poincaré group by an infinite-dimensional set of
transformations acting nontrivially under Lorentz rotations and boosts.

3.4.2 Why Radiative Spacetimes Have the Full BMS Group?

Up to now we showed that the asymptotic symmetry group of asymptotically flat space-
times in Bondi gauge is the BMS group. Yet a natural question remains: why is the
asymptotic symmetry structure of gravity so vastly richer than that of Minkowski space-
time? The answer lies in the presence of gravitational radiation. Unlike in flat space,
where no dynamical degrees of freedom propagate, an asymptotically flat spacetime can
carry gravitational waves, and these waves leave imprints at null infinity that enlarge the
space of physically distinguishable configurations. It is precisely this radiative freedom
that opens the door to the infinite-dimensional structure of the BMS group.
In exact Minkowski spacetime the asymptotic metric at I + is globally Lorentz invari-
ant, and the Bondi shear vanishes,
Mink
γAB = 0. (3.67)
7
Definition (Semidirect product): Let N and H be groups, and let ρ : H → Aut(N ) be a homomorphism
describing an action of H on N by automorphisms. The semidirect product of N by H, denoted N ⋊ρ H,
is defined as the set N × H equipped with the group law


(n1 , h1 )(n2 , h2 ) = n1 ρ(h1 )(n2 ), h1 h2 . (3.66)

With this multiplication, N is a normal subgroup of N ⋊ρ H, while H embeds as a (not necessarily normal)
subgroup. The product is semidirect because the group law in N is twisted by the action of H, encoding
the nontrivial way in which elements of H act on elements of N .

– 31 –
Under a supertranslation with parameter f (z, z̄), the shear transforms as

δf γAB = −2DA DB f + γAB D2 f. (3.68)

Demanding that Minkowski spacetime remain invariant forces this variation to vanish. Us-
ing Eq. (3.62), the only solutions are the ℓ = 0, 1 spherical harmonics—corresponding to
the usual time and space translations. Thus the complete invariance group of Minkowski
spacetime is precisely the Poincaré group. The picture changes dramatically once gravi-
tational wave is present. If gravitational radiation passes through future null infinity, the
shear changes between early and late retarded times. Integrating the news gives the net
change
Z +∞
∆γAB = γAB (u = +∞) − γAB (u = −∞) = NAB du. (3.69)
−∞

This nonzero shift is the gravitational memory effect: freely falling observers experience
a permanent relative displacement even after the wave has passed. What is striking is
that the mathematical form of this shift is highly constrained. On the two–sphere at null
infinity, any symmetric tracefree tensor constructed from a scalar function f (z, z̄) takes
the form shown in Eq. (3.68). Thus, if the memory-induced shift ∆γAB matches this
expression for some f , then the initial and final vacuum configurations are related by
a supertranslation. Radiative processes therefore move the system between inequivalent
vacua connected by supertranslations. These transformations cease to act trivially: instead,
they map an infinite continuum of physically distinct gravitational vacua into one another.
The enlargement of the asymptotic symmetry group from Poincaré to the full BMS group is,
in this sense, a direct imprint of gravitational waves—the memory they leave, the resulting
infinite vacuum degeneracy, and the refined geometric structure of spacetime at future null
infinity I + .

Infrared Triangle

Infrared triangle expresses the equivalence between asymptotic symmetries, soft the-
orems, and memory effects in gauge theory and [Link] first corner is formed by
the infinite-dimensional Bondi–Metzner–Sachs (BMS) asymptotic symmetry group,
discovered by Bondi, van der Burg, and Metzner [58], and independently by Sachs
[59]. The second corner contains the universal soft theorems, beginning with Wein-
berg’s 1965 soft photon and graviton theorems [81]. It was later revealed by Stro-
minger that gravitational memory is equivalent to the soft graviton theorem [82]. The
third corner comprises the family of gravitational wave memory effects, encompassing

– 32 –
 both linear and nonlinear memory, as originally described by Zel’dovich and Polnarev
[55], Christodoulou [57], and further developed in [56, 83]. Strominger & Zhiboedov
showed in [84] that gravitational memory corresponds to a transition between BMS
vacua and that the Fourier transform of the memory formula yields Weinberg’s soft
theorem. The structure was later explicitly named the infrared triangle in [73].
Analogous IR effects also arise in QED, where an infrared triangle relates large gauge
symmetries, soft photon theorems, and electromagnetic memory effects. Large gauge
transformations at null infinity give rise to conserved charges whose Ward identities
reproduce the universal soft photon theorem [79, 85–88]. These symmetries have a
direct physical imprint in the form of a permanent electromagnetic memory effect.
For the parity-violating (chiral) electromagnetism memory effect see [89].

4 Expanding Universe and Gravitational Waves

In Secs. 2 and 3, we explored gravitational radiation in Minkowski and asymptotically

flat spacetimes. In this part, we shift our attention to the generation and evolution of
gravitational waves in a cosmological background. The gravitational dynamics can be
accurately described by asymptotically flat geometries only on scales smaller than about
100 Mpc, where the expansion of the Universe is negligible. On larger, cosmological scales,
the expansion becomes dynamically relevant and the spacetime must instead be modeled
by the FLRW metric (see Fig. 3). We begin this section with the expansion history of the
Universe and the sequence of cosmological eras, then examine gravitational waves produced
during inflation and their behavior in (quasi)–de Sitter geometry. Finally, we turn to the
infrared structure of inflation, discuss Weinberg’s adiabatic theorem, and conclude with
Maldacena’s consistency relation for cosmological correlators.

4.1 Expansion History and Cosmological Eras

Modern cosmology is based on two key observations: i) the universe is expanding, and
ii) on large scales (> 100 Mpc) the matter distribution is homogeneous and isotropic.
The average spacetime is then described by the Friedmann-Lemaı̂tre-Robertson-Walker
(FLRW) metric

dr2
ds2FLRW = −dt2 + a2 (t)( + r2 dΩ2 ), (4.1)
1 − Kr2
where t is the cosmic time, a(t) is the scale factor, and K = 0, +1, −1 describe flat, posi-
tively curved and negatively curved spacelike 3-hypersurfaces, respectively. This spacetime
is conformally flat, with a vanishing Weyl tensor (Petrov type D). From now on, we re-
strict our discussion to the case of the flat universe with K = 0, which is favored by present

– 33 –
observations. 8 Now, we define the conformal time τ in terms of the physical time t as
1
dτ ≡ dt, (4.2)
a(t)

We also define the Hubble parameter, H, which measures the expansion rate of the Universe
ȧ
H≡ . (4.3)
a
Using this relation, we find that the flat FLRW geometry is conformally equivalent to
Minkowski spacetime
gµν (τ, x) = a2 (τ ) ηµν . (4.4)
The factor a2 uniformly rescales the Minkowski geometry, stretching both space and time
by the scale factor a(τ ). Light rays (null geodesics) remain unaffected by this conformal
rescaling, since for ds2 = 0 the scale factor cancels out. Consequently, the causal structure
of the Universe in conformal coordinates is identical to that of flat spacetime, making the
conformal time τ a natural variable for describing wave propagation, such as GWs, CMB
photons, and cosmological perturbations, in an expanding universe.
Let us now consider a perfect fluid

Tµν = (ρ + P ) uµ uν + P gµν , (4.5)

where ρ is the energy density measured in the rest frame of the fluid, P is the isotropic
pressure, and uµ is the 4–velocity of the fluid (normalized as uµ uµ = −1). For an equation
of state

P = wρ, (4.6)

the Einstein field equations reduce to the Friedmann equations, i.e.

2 ȧ 2 2 ä 1
3MPl ( ) =ρ and MPl = − (1 + 3w)ρ, (4.7)
a a 6
where MPl2 ≡ (8πG)−1 is the reduced Planck mass. Solving these equations for a constant

equation-of-state parameter w, we have

  2   2
t (1+3w)+2 τ (1+3w)
a(t) = and a(τ ) = , (4.8)
tI τI

where tI and τI are some positive constants. The behavior of the scale factor during
different cosmological eras is summarized in Table 4.
8 K
The curvature density parameter, ΩK ≡ − , quantifying the fractional contribution of spatial
a2 H 2
curvature to the total energy density. Current observations constrain it to |ΩK | < 0.002 98% C.L. [90].
9
Stiff era is a hypothetical phase between inflation and the radiation era, appearing in models where a
scalar field becomes kinetic-energy dominated [91–93]. Although unconfirmed, it is consistent with current
bounds if it ends before BBN and is phenomenologically notable for enhancing high-frequency primordial
gravitational waves [94].

– 34 –
 Cosmological era Eq. of state Scale factor Hubble SEC
a(t) = exp(Ht) H(t) = Hinf
Cosmic inflation w = −1 + 32 ϵ No
1
a(τ ) = − Hτ H(τ ) = − τ1
 1
3 1
9
a(t) = ttS H(t) = 3t
Stiff era w=1  1 Yes
2 1
a(τ ) = ττS H(τ ) = 2τ
 1
2 1
1
a(t) = ttI H(t) = 2t
Radiation era w= 3 Yes
τ 1
a(τ ) = τI H(τ ) = τ
 2
t 3 2
a(t) = tJ
H(t) = 3t
Matter era w=0  2 Yes
τ 2
a(τ ) = τJ H(τ ) = τ

Table 4: Equation of state, scale factor, and Hubble parameter for each cosmological
era. The last column shows validity of the Strong Energy Condition (SEC), i.e., whether
ρ + 3P > 0. Here, H ≡ aH.

To gain insight into the causal structure of cosmological spacetime, let us consider the
propagation of light, characterized by the null condition ds2 = 0. For a radial null geodesic,
this condition leads to
d ln a′
Z τ Z a
′
r(τ ) − r(τI ) = dτ = ′ ′
, (4.9)
τI aI a H(a )

where τ is the conformal time. This integral defines the comoving distance that light can
travel between τI and τ , and hence determines the size of the observable region of the
Universe at a given epoch. In terms of the comoving (particle) horizon, (aH)−1 , we can
write
(1 + 3w)
(aH)−1 = τI a(1+3w)/2 , (4.10)
2
which depends on the scale factor a and the equation-of-state parameter w. The charac-
teristic physical distance scale of the expanding Universe is set by the Hubble radius,
1
RH = , (4.11)
H
which is the boundary between regions that are causally connected and those that are
not. A perturbation mode with comoving wavenumber k has a corresponding physical
wavelength λphys (t) = a(t)
k , which grows with the cosmic expansion and can be compared
with RH to determine whether the mode lies inside or outside the Hubble horizon. Modes
satisfying λphys ≤ RH (or equivalently, k ≥ aH) are said to be sub-horizon and evolve
causally, while those with λphys > RH (super-horizon) remain effectively frozen until re-
(0)
entry [95]. Today, the Hubble radius is approximately RH ≃ 4.4 Gpc, corresponding to

– 35 –
the present Hubble parameter H0 ≃ 67.7 km s−1 Mpc−1 [90]. Hence, the largest scales
inside the horizon today correspond to
H0
k0 = ≃ 2.25 × 10−4 Mpc−1 , λ0 ≃ 4.4 Gpc. (4.12)
c
The observed statistical homogeneity and isotropy of the cosmic microwave background
(CMB) across our entire cosmic horizon imply that perturbation modes with wavelengths
as large as λ0 must have been in causal contact, or in thermal equilibrium, at some stage
of cosmic evolution.

Figure 13: Left panel: The (comoving) causal past of an observer today at τ0 (redshift
z0 = 0), in FLRW spacetime made of ONLY ordinary matter, i.e. 1 + 3w > 0. The
orange line shows the last scattering surface, τCMB (at redshift zrec ≃ 1090). A and B
are two causally disconnected points at the last scattering surface. In fact, the angle
spanned by the shaded area at the surface of last scattering is the comoving horizon at
recombination. Right panel: The causal past of an observer today, including the epoch of
cosmic inflation. Inflation extends conformal time to τinf < 0, enlarging the causal domain
of the Universe. Consequently, regions that are outside each other’s horizons at τ = 0 were
in causal contact during inflation, thereby resolving the horizon problem and explaining
the observed uniformity of the CMB.

Horizon Problem 10

Ordinary forms of matter, with positive pressure, satisfy the strong energy condition
(SEC), i.e. ρ + 3P > 0 (1 + 3w > 0). Thus, from Eq. (4.10), the comoving horizon
increases as the universe expands. Now, let us compute the angle spanned by the
comoving horizon at recombination, θ. As we see in Fig. 13, θ is given as
2(τCMB − τi )
sin θ = . (4.13)
τ0 − τCMB
On the other hand, we can read the τ integral in Eq. (4.9) as
Z z2
dz
τ − τI = − , (4.14)
z1 H(z)

– 36 –
 1
where z is the redshift parameter 1 + z = a(z) , in which we set the scale factor today
to unity, a0 = 1. Moreover, the Hubble parameter is
q
H(z) = Ωm (1 + z)3 + Ωγ (1 + z)4 + ΩΛ , (4.15)

Ωm
where Ωm = 0.3, Ωγ = 1+z eq
, and ΩΛ = 1 − Ωm are the matter, radiation, and (late
time) dark energy fraction today. Using Eqs. (4.14) and (4.15) in Eq. (4.13) and
solving the integral, we find

θc ≃ 2.3◦ . (4.16)

At recombination, the causal wavenumber is

kcausal ≃ a∗ H∗ ≈ 4 × 10−3 Mpc−1 , (4.17)

corresponding to a comoving wavelength λcausal ≃ 250 Mpc. Only perturbation

modes with k ≳ kcausal could have been in causal contact and reached thermal equi-
librium by the time of recombination (Fig. 13, left panel). In contrast, the mode
corresponding to our present cosmic horizon, with λ0 ≃ 4.4 Gpc given by Eq. (4.12),
gives
kcausal
≃ 17. (4.18)
k0
That implies the region encompassing our observable Universe today would have
consisted of roughly (17)3 ∼ 104 causally disconnected domains at recombination.
On the other hand, the observed homogeneity of the CMB implies that even modes
as large as λ0 were once in causal contact, an apparent contradiction known as
the horizon problem. As we see in the following, cosmic inflation, an early period
of accelerated expansion in which the SEC is violated, solves the horizon problem
dynamically and allows our universe to arise from generic initial conditions.

4.1.1 Cosmic Inflation

Our discussion of the horizon problem was based on the validity of the SEC (1 + 3w > 0)
and, therefore, the growing Hubble sphere of the standard Big Bang cosmology. A simple
solution, therefore, is a phase of decreasing Hubble radius in the early history of the
universe,

d(aH)−1
<0 where 1 + 3w < 0. (4.19)
dt
If this lasts long enough the horizon problem can be solved (see the right panel of Fig. 13
and Fig. 14).
10
The covariant form of the strong energy condition (SEC) is Tµν − 12 gµν T uµ uν ≥ 0 for all timelike

vectors uµ .

– 37 –
Figure 14: Top panel: The Hubble radius, H1 , and the physical wavelength, ka , as functions
of the scale factor. Notice that X is the physical coordinate, X = ax. The black line shows
1/H, the red and blue lines denote the physical wavelength of two modes that re-entered
the cosmic Horizon during radiation domination (RD) and matter domination (MD) eras,
respectively. The shaded gray region shows causally connected points. Bottom panel:
close-up view of the mode relative to the cosmic horizon.

To quantify this idea, let us compute the total amount of expansion that took place
between the moment a comoving mode k crossed the Hubble radius during inflation and
the endof inflation,
 a quantity known as the number of e-folds, Nk . It is defined as
aend
Nk ≡ ln ak , where ak and aend denote the scale factor at horizon exit of mode k and at
the end of inflation, respectively. The required number of e-folds for the present horizon
scale k0 = 2.25 × 10−4 Mpc−1 , this gives
 
1 Hinf
N ≃ 61.4 + ln , (4.20)
2 1014 GeV
which sets the minimal duration of inflation required to solve the horizon problem and
to ensure that our entire observable Universe originated from a single, causally connected
patch before inflation began. For typical inflationary energy scales Hinf ∼ 1013 −1014 GeV,
the Universe must have expanded by roughly 60 e-folds during inflation.
The inflation paradigm postulates a brief period (within tinf ≈ HNinf
0
) of quasi-exponential
accelerated expansion during which the scale factor increased by about 60 e-folds [96–99].

– 38 –
This considerable expansion is sourced by a negative pressure component in the energy-
momentum of the matter contents and drives the universe towards almost perfect homo-
geneity, isotropy, and flatness that we have observed. During inflation, the spacetime is
very close to a de Sitter space w ≃ −1, and the Hubble parameter, H, is almost constant.
The scale factor grows exponentially, a(t) ∝ eHt , and the deviation from a perfect de Sitter
space is quantified in terms of two slow-roll parameters 11
Ḣ Ḧ
ϵ≡− and ,
η= (4.21)
H2 2ḢH
which should be small during the slow-roll inflation. This causes the comoving Hubble
radius (aH)−1 to shrink rapidly, so physical scales that were once much smaller than the
Hubble radius (k = aH) are stretched to super-horizon sizes. Modes that were initially
causally connected (inside the horizon) are driven far apart, and after inflation these same
comoving scales re-enter the horizon during the radiation and matter eras (see Fig. 14).

Beyond Cosmic No-Hair Theorem

At cosmological scales, the Universe exhibits an extraordinary degree of homogeneity
and isotropy. Since cosmic evolution may begin from generic initial conditions beyond
our control, this striking simplicity naturally invites a dynamical explanation—one
in which a homogeneous and isotropic state arises as an attractor of the cosmic
evolution. The first such attempt was made by Hawking, Gibbons, and Moss in
[100, 101] arguing that the late-time behavior of any accelerating Universe is an
isotropic Universe. This statement was dubbed as cosmic no-hair conjecture. The
first rigorous attempt to establish this conjecture was provided in Wald’s seminal
work [75]. Wald’s cosmic no-hair theorem asserts that a broad class of initially
expanding Bianchi-type cosmological models (with the exception of Bianchi type IX),
whose total energy–momentum tensor consists of a positive cosmological constant
supplemented by matter satisfying the strong and dominant energy conditions, evolve
exponentially rapidly toward de Sitter spacetime, within only a few Hubble times.
Inflation was originally proposed to explain the observed flatness and large-scale ho-
mogeneity of the Universe, offering a dynamical mechanism through which horizon-
scale correlations are generated. This motivation led to a reassessment of Wald’s the-
orem in inflationary contexts, culminating in a generalized cosmic no-hair theorem for
inflation presented in [102]. This result demonstrates that, although anisotropies may
grow during inflation in the presence of spinning fields, i.e. evading the traditional
cosmic no-hair conjecture, their amplitude remains parametrically small and is sub-
ject to a universal upper bound. An example of inflationary models with anistripic
hair is the dilaton–Maxwell theory [103]. The authors of [104] extended the discus-
sion of the cosmic no-hair conjecture to include initial anisotropies and demonstrated,

11
Besides the definitions introduced in Eq. (4.21), two other slow–roll conventions are widely used in the
 ′ 2
M2 2 V ′′
literature. The potential slow–roll parameters are ϵV ≡ 2Pl VV and ηV ≡ MPl V
; the Hubble slow–roll
ϵ̇H
parameters are ϵH ≡ − HḢ2 and ηH ≡ ϵH H
.

– 39 –
 through fully nonlinear numerical simulations, that inflation can emerge from highly
inhomogeneous, gradient-dominated initial conditions, provided the inflaton field re-
mains within a sufficiently flat region of its potential. Their analysis, however, is
restricted to scalar-field models, and further investigations are required to assess
the impact of spinning fields, e.g. gauge fields interacting with the inflaton, on the
persistence or decay of cosmic inhomogeneities during inflation.

4.1.2 From the Inflationary Paradigm to Precision Cosmology

The inflationary paradigm was pioneered by Starobinsky (1980) [96], Sato (1981) [97],
Guth (1981) [98], and Linde (1982) [99], who independently proposed that a brief pe-
riod of accelerated expansion in the early Universe could resolve the flatness, horizon, and
monopole problems of standard Big Bang cosmology. A profound realization soon followed:
if inflation stretches all preexisting irregularities, it must also amplify the inevitable quan-
tum fluctuations of the inflaton field and of spacetime itself. This idea was first developed
in pioneering works by Mukhanov and Chibisov [105], Hawking [106], Starobinsky [107],
Guth and Pi [108], and Bardeen, Steinhardt and Turner [109]. They formulated the theory
of cosmological perturbations in an inflating background and demonstrated that inflation
generically produces a nearly scale-invariant, Gaussian, and adiabatic spectrum of cur-
vature perturbations, a strikingly specific and testable prediction. As their wavelengths
exceed the Hubble radius, these quantum fluctuations lose their quantum coherence and
become effectively classical, but inherently stochastic fields. These perturbations serve as
the seeds of the large-scale structure of the Universe, eventually growing under gravity into
the cosmic web of galaxies and clusters we observe today. In addition to tensor modes,
inflation also predicts a stochastic background of scalar (curvature) perturbations. These
fluctuations seed the temperature anisotropies of the cosmic microwave background and
eventually grow into the large-scale structure of the Universe. Their statistical properties
are described by the scalar power spectrum ∆2ζ (k), which quantifies the variance of cur-
vature perturbations per logarithmic interval in wavenumber. For slow-roll inflation, the
scalar power spectrum takes the form
ns −1
1 H2

k
∆2ζ (k) = 2 , (4.22)
8π 2 ϵ MPl aH k=aH

where H is the Hubble parameter during inflation and ns = 1−4ϵ+2η is the scalar spectral
index.
More than a decade later, the inflationary predictions were spectacularly confirmed
by the detection of temperature anisotropies in the cosmic microwave background (CMB)
by the COBE satellite [110], providing the first direct evidence for primordial perturba-
tions. Subsequent missions refined this picture with ever-increasing precision. The WMAP
satellite [111] mapped the CMB sky with degree-scale resolution, measured the spectral
index ns < 1, and confirmed the predicted slight red tilt of the primordial spectrum. The
Planck mission [90] further improved the precision, establishing that the fluctuations are
Gaussian and adiabatic to remarkable accuracy. The observations precisely determine the

– 40 –
amplitude of the scalar power spectrum at the pivot scale as
∆2ζ (k∗ ) ≃ 2.1 × 10−9 , (4.23)
providing one of the most important observational anchors for inflationary cosmology. The
latest Planck results [90] give
ns = 0.9649 ± 0.0042, (4.24)
confirming, with high statistical significance, the deviation from exact scale invariance
predicted by slow-roll inflation. This remarkable progression represents one of the cen-
tral achievements of modern cosmology, establishing a direct link between the primor-
dial fluctuations of the early Universe—most plausibly arising from quantum fluctuations
during inflation—and the large-scale structure we observe today. Continued advances
from ground-based and suborbital experiments, including ACT (2011–2023) [112], BI-
CEP/Keck (2016–2021) [113], and the BICEP Array [114], have substantially improved
sensitivity to the CMB B-mode polarization signal. Upcoming observations from the Si-
mons Observatory, with a first major data release expected around 2026 [115], together
with future space missions such as LiteBIRD [22] and the proposed CMB-S4 experiment
[23], aim to detect the subtle imprint of primordial gravitational waves—an observation
that would provide the next decisive test of inflation. Taken together, these measurements
have transformed inflation from a bold theoretical proposal into the leading framework for
early-universe cosmology, offering compelling empirical support for the view that the cos-
mic structures we see today originated from quantum fluctuations stretched to macroscopic
scales during an epoch of accelerated expansion.

4.2 Inflation and GWs

A key prediction of the inflationary paradigm is the existence of a stochastic background of
primordial gravitational waves (PGWs) generated by tensor fluctuations of the spacetime
metric in the very early Universe. Just like the temperature anisotropies of the CMB, this
relic GW background forms a random field with no distinct features in either the time
or frequency domain. However, GWs carry a crucial advantage over photons: while the
CMB decoupled roughly 4 × 105 years after the Big Bang, primordial gravitational waves
have been free-streaming since inflation, potentially tracing back to energy scales near the
Planck regime (see Fig. 15). For detailed discussions of the thermal history of the Universe,
including the evolution from inflation and reheating through Big Bang nucleosynthesis and
recombination, see [116–122].
We are interested in the tensor perturbations of the metric that correspond to gravita-
tional waves observed today. At linear order, the perturbed FLRW metric can be written
as
ds2 = a2 (τ ) −dτ 2 + δij + γij (x, τ ) dxi dxj ,


(4.25)
where the symmetric, traceless, and transverse tensor γij represents the two physical po-
larization states of gravitational waves. The quadratic action for γij is
2 Z
MPl  ′ 2
(2)
dτ d3 x a2 (τ ) (γij ) − (∂ℓ γij )2 ,

Sγ = (4.26)
8

– 41 –
Figure 15: Schematic illustration of the thermal and structural history of the Universe,
from cosmic inflation and Big Bang nucleosynthesis (BBN) to the formation of the cosmic
microwave background (CMB), the emergence of large-scale structure, and the modern
Universe. After inflation ended, the Universe remained for a long period filled with a hot,
charged plasma. In this phase, photons were repeatedly scattered by free charged particles
and could not propagate freely. Only when protons and electrons combined to form neu-
tral atoms did the Universe become transparent, allowing CMB photons to decouple and
free-stream across space, carrying a snapshot of the Universe at that epoch. In contrast,
gravitational waves, though universally coupled to all forms of energy, interact so weakly
that they were never in thermal equilibrium. The cosmic medium has always been trans-
parent to them, meaning that once detected, these primordial gravitational waves could
reveal direct information about the Universe’s highest energy scales and earliest moments,
far beyond the reach of any electromagnetic probe. The horizontal axis in this illustration
represents the energy scale of the Universe, defined as ρ1/4 , which decreases as the Universe
expands and cools.

where primes denote derivatives with respect to conformal time τ . The evolution of GW
modes is governed by the linearized Einstein equations, which in conformal time for the
FLRW background reduce to

′′ a′ ′
γij +2 γ − ∂i2 γλ = 16πG a2 πij
T
. (4.27)
a ij
T is the transverse and traceless part of the anisotropic stress tensor.12 In what
Here πij
12
The anisotropic stress tensor can be decomposed as
1 1
πij ≡ δTij − δij δT kk = ∂ij
2 S
π − ∂ 2 π S δij + 2∂(i πj)
V T
+ πij , (4.28)
3 3
with ∂i πiV = ∂i πij
T
= 0. Here, π S , πiV , and πij
T
denote the scalar, vector, and tensor components, respectively,

– 42 –
follows, we study the evolution of the initial quantum fluctuations of these tensor modes
in the expanding background during inflation. Since inflation is dominated by a nearly
homogeneous scalar field, any anisotropic stress from matter fields can be neglected, so we
T = 0.
set πij
The field operator can be expanded in Fourier modes as
X Z d3 k h †
i
ik·x λ ∗ −ik·x λ∗
γij (τ, x) = γ λ (τ, k) b̂λ (k) e ϵ ij (k̂) + γλ (τ, k) b̂λ (k) e ϵij ( k̂) ,(4.29)
(2π)3
λ

where ϵλij (k̂) denotes the polarization tensor, and b̂λ (k) and b̂†λ (k) are the annihilation and
creation operators, respectively, satisfying

[b̂λ (k), b̂†λ′ (k′ )] = (2π)3 δλλ′ δ (3) (k − k′ ). (4.30)

The tensor perturbations can be treated as a stochastic field, whose statistical properties
are captured by the two-point correlation function of the corresponding quantum operators,

⟨ γ̂λ (k) γ̂λ† ′ (k′ ) ⟩ = (2π)3 δ (3) (k − k′ ) δλλ′ Pγ (k), (4.31)

where Pγ (k) denotes the tensor power spectrum. For convenience, one often defines the
dimensionless tensor power spectrum as
k3
∆2T (k) ≡ Pγ (k), (4.32)
2π 2
which directly characterizes the amplitude of the stochastic gravitational wave background
generated during inflation. In the scale-invariant limit, ∆2T (k) becomes nearly constant,
reflecting the uniform excitation of tensor modes across different wavelengths.
As is evident from Eq. (4.26), the field γij is not yet canonically normalized. To
introduce a canonically normalized variable, we define
aMPl
vij ≡ γij . (4.33)
2
In terms of vij , the action becomes
a′′ 2
Z  
(2) 1 3 ′ 2 2
S = dτ d x (vij ) − (∂ℓ vij ) + v , (4.34)
2 a ij
and leads to the field equation
a′′
 
vλ′′ 2
+ k − vλ = 0. (4.35)
a
This equation describes the propagation of quantum tensor modes in a time-dependent
′′
background, where the term aa ≃ 2+3ϵ τ2
acts as an effective mass induced by cosmic expan-
sion. Its general solution can be expressed in terms of Hankel functions as
p
π|τ | i(1+2νT )π/4 h i
vλ (τ, k) = e c1 Hν(1)
T
(−kτ ) + c 2 Hν
(2)
T
(−kτ ) , (4.36)
2
T
with πij capturing the transverse and traceless part of the anisotropic stress [122].

– 43 –
Figure 16: Constraints on the scalar spectral index and the tensor-to-scalar ratio (ns , r)
at k = 0.002 Mpc−1 from Planck, with BK18 & BAO data, compared with predictions
of representative inflationary models. Image credit: BICEP/Keck Collaborations [124].

where c1 and c2 are constants determined by the choice of initial conditions and νT is 13

3
νT ≃ + ϵ. (4.38)
2
Assuming seeds of perturbations to be from the quantum fluctuations, so-called Bunch-
Davies vacuum, imposes the initial value in the asymptotic past as
1
lim vλ (τ, k) = √ e−ikτ . (4.39)
τ →−∞ 2k
Consequently, the corresponding mode function γλ (up to an overall phase) takes the form
p
π |τ |3 H
γλ (τ, k) = H (1) (−kτ ). (4.40)
2 MPl νT
Now, let us compute the tensor power spectrum for modes after horizon crossing. Dur-
ing inflation, the quantum fluctuations of the metric are stretched to super-horizon scales
(k ≪ aH), where they effectively freeze and behave as classical perturbations. The ampli-
tude of each mode at horizon exit (k = aH) determines the primordial tensor spectrum.
The dimensionless tensor power spectrum is given by

2 H2 k nT
 
2
∆T (k) = 2 2 , (4.41)
π MPl aH k=aH

13
The asymptotic behavior of the Hankel function is as [123]
r
(1) 2 −i(2ν+1)π/4 iz i 2
lim Hν (z) = e e , lim Hν(1) (z) = − ( )ν Γ[ν]. (4.37)
z→∞ πz z→0 π z

– 44 –
where H is the Hubble parameter during inflation. Evaluating the spectrum at horizon
exit (k = aH) ensures that each mode is characterized by the value of H when it left the
horizon. The quantity nT in Eq. (4.41) is the tensor spectral index, defined as
d ln ∆2T
nT ≡ = −2ϵ. (4.42)
d ln k
A strictly scale-invariant spectrum corresponds to nT = 0, in which case the power ∆2T is
independent of the wavenumber k. Inflation predicts a slight red tilt (nT < 0), meaning
that longer-wavelength modes have slightly larger amplitudes. The overall amplitude of
the tensor spectrum is directly related to the energy scale of inflation, while its nearly
scale-invariant shape reflects the quasi-de Sitter nature of the inflationary expansion.
A key observable quantity in inflationary cosmology is the tensor-to-scalar ratio
∆2T (k∗ )
r∗ = , (4.43)
∆2ζ (k∗ )

which measures the relative amplitude of primordial GW perturbations to scalar (curva-

ture) perturbations. For single-field slow-roll (single-clock) inflation, the tensor and scalar
amplitudes are directly linked through the slow-roll parameter, r∗ = 16 ϵ∗ , providing a
direct probe of the energy scale and dynamics of inflation. The most recent joint analysis
by the Planck and BICEP/Keck collaborations places a stringent upper bound on the
tensor-to-scalar ratio [124]

r0.05 < 0.036 (95% C.L.). (4.44)

The Energy Scale of Inflation

Combining the upper limit on r with the measured scalar amplitude yields an upper
bound on the Hubble scale during inflation,
s
r ∆2ζ (k∗ )
Hinf = πMPl ≲ 4.7 × 1013 GeV, (4.45)
2
1/4
corresponding to an inflationary energy scale of V∗ ≲ 10−2 MPl . Thus, even
a small value of r may probe physics near the grand-unified scale and provides a
direct window into the energy scale of inflation. On the other hand, Big Bang
Nucleosynthesis (BBN) requires the Universe to be thermalized before light-element
formation, implying a minimum reheating temperature Treh ≳ 5 MeV. Assuming
instantaneous reheating, this sets a lower bound on the inflationary Hubble scale
r
2
π 2 g∗ Treh
Hinf = σreh ≳ 1.1 × 10−23 GeV, (4.46)
90 Mpl
where g∗ is the effective number of relativistic degrees of freedom, which at Treh ≃
5 MeV takes the value g∗ = 10.75. The parameter σreh is a phenomenological factor
that encodes the efficiency of the reheating process and equals unity in the case of
instantaneous reheating (see Fig. 17).

– 45 –
 p
Figure 17: Current allowed window for the inflationary energy scale, Mpl Hinf .
The upper bound is determined by the observational limits on the tensor-to-scalar
ratio, as given in Eq. (4.45), while the lower bound is set by the requirement that re-
heating occurs before Big Bang Nucleosynthesis (BBN), as shown in Eq. (4.46). This
corresponds to an enormous uncertainty of about 18 orders of magnitude, spanning
from the Grand Unified Theory (GUT) scale at 1016 GeV down to around 10 MeV.

4.3 IR Structure of Cosmological Perturbations

Here, we briefly review Weinberg’s adiabatic modes theorem and Maldacena’s inflationary
consistency relations. The adiabatic mode theorem plays a central role in connecting
the primordial perturbations generated in the very early Universe to the cosmological
perturbations observed today, providing a bridge between inflationary initial conditions and
late–time cosmological observables. Inflationary consistency relation, which fundamentally
relies on the presence of these adiabatic modes, encapsulates the symmetry constraints that
govern correlations between long-wavelength (soft) and short-wavelength perturbations in
single-field inflation.

4.3.1 Adiabatic Modes in Cosmology

To extract information about inflation from observations, we must connect the primor-
dial fluctuations produced during inflation to the structures observed in the later uni-
verse (see Fig. 15). This task is complicated by our limited understanding of the post-
inflationary epoch—when the energy stored in scalar fields was converted into matter and
radiation—and of any subsequent, poorly constrained eras. Although these stages occurred
while the relevant fluctuations were stretched far beyond the horizon (see Fig. 14), they
may still have affected their amplitude or scale dependence. Therefore, to relate the in-
flationary perturbations to those observed in the CMB or large-scale structure, we need a
conservation law that remains valid on super-horizon scales and is independent of the de-
tailed dynamics of the universe’s evolution. Establishing the existence of such a conserved
quantity is precisely the goal of adiabatic mode theorem [95], which demonstrates that cer-
tain curvature perturbations remain constant outside the horizon, regardless of the details
of the cosmic evolution. The need for a conserved, superhorizon quantity has motivated a
long line of foundational work in cosmological perturbation theory. Early seminal studies
established the gauge-invariant formulation of cosmological fluctuations and identified the
conservation of adiabatic modes on superhorizon scales [125, 126] (see also [127]). Building
on these key works, Weinberg demonstrated the full generality of such conservation laws
by proving the existence of adiabatic modes under fairly general conditions [95].

– 46 –
 We start with the construction of Weinberg’s adiabatic modes in the Newtonian gauge
which is essential for the derivation of the consistency relations. Fixing the gauge, uniquely
specifies all the modes with finite momentum, k ̸= 0. However, there are residual gauge
symmetries for the very long-wavelength modes, which remain a symmetry of the gauge-
fixed action. Starting with the flat FLRW metric as the unperturbed metric

ds2 = −dt2 + a2 dx2 . (4.47)

Here we focus on the implications of the diffeomorphism invariance on the nature of the very
long-wavelength modes with k/a ≪ H. Therefore, only the homogeneous diffeomorphisms
are relevant. The general spatially homogeneous perturbed metric in the Newtonian gauge
is
 
ds2 = −(1 + 2Φ(t))dt2 + a2 (1 − 2Ψ(t))δij + γij (t) dxi dxj , (4.48)

where Φ and Ψ are the Bardeen potentials and γij is the traceless tensor perturbation,
gravity wave. 14 One can decompose the general spatially homogeneous perturbed energy-
momentum in the Newtonian gauge as

T00 = −ρ̄g00 + δρ and T0i = −(ρ̄ + P̄ )∂i δu, (4.49a)

 
2 2 S T
Tij = P̄ (t)gij + a δij δP + ∂ij π + πij , (4.49b)

where a bar denotes an unperturbed quantity, δρ, δP and δu are the perturbed density,
pressure and velocity potential respectively. Moreover, π S and πij T are the scalar and tensor

anisotropic inertia which characterize departures of Tµν from the perfect fluid form. 15 In
the scalar sector, the perturbed energy conservation equation, ∇µ T µ0 = 0, is
 
2 1 S
δ ρ̇ + 3H(δρ + δp) − 3(ρ̄ + P̄ )Ψ̇ + ∇ (ρ̄ + P̄ )δu + Hπ = 0. (4.50)
a2
In terms of the curvature perturbation on uniform-density slices
δρ
ζ = −Ψ + . (4.51)
3(ρ + p)
and the non-adiabatic pressure perturbation
p̄˙
δPnad ≡ δP − δρ, (4.52)
ρ̄˙
we can write the continuity equation as

k2 H πS
 
H δu
ζ̇ = − δPnad + + . (4.53)
ρ̄ + P̄ 3 a2 ρ̄ + P̄
14
Here we neglect the vector perturbations due to their damping nature in most of the inflationary models
including our axion-gauge field model.
15
It is interesting to note that in the decomposition (4.49b), the effects of bulk viscosity are included in
δp [95].

– 47 –
In the super-horizon limit, k ≪ aH, the gradient term is suppressed as a−2 , and for
adiabatic perturbations with δPnad = 0, the curvature perturbation is conserved

lim ζ̇ = 0. (4.54)
k
aH
→0

Consequently, the long-wavelength dynamics of ζ admits two independent adiabatic modes:

lim ζ1 (t) = const, lim ζ2 (t) = 0. (4.55)

k k
aH
→0 aH
→0

In the absence of anisotropic stress, π S = 0, one has Φ = Ψ, and these two adiabatic modes
correspond to the following super-horizon behaviors of the Bardeen potential
H(t) t ′ ′
 Z 
H(t)
lim Ψ1 (t) = ζ1 −1 + a(t ) dt , lim Ψ2 (t) = C , (4.56)
k
aH
→0 a(t) t0 k
aH
→0 a(t)

where C is an integration constant.

For tensor perturbations, in the regime where the physical wavelength is much larger
than the Hubble radius, k/a ≪ H, we have
k2
γ̈ + 3H γ̇ = O( ), (4.57)
a2 H 2
and the gravitational wave equation admits two independent super-horizon homogeneous
solutions
dt′
Z t
γ1 (t) = const, γ2 (t) = C 3 ′
. (4.58)
t0 a (t )
The second solution decays with time for any expanding universe. Since tensor pertur-
bations satisfy a second-order differential equation, these two modes exhaust the space of
possible long-wavelength solutions. Consequently, for generic initial conditions, the tensor
perturbations at late times are dominated by the constant mode γ1 . These are precisely
the tensor-sector adiabatic modes guaranteed by Weinberg’s adiabatic mode theorem [122].
Note that in the presence of a non-vanishing πij T , gravitational waves acquire inhomoge-

neous solutions sourced by this anisotropic stress, in addition to the adiabatic modes. For
explicit examples, see [5, 128–132].
Having completed the derivation of adiabatic modes in cosmology, we now comment
on the role of gauge transformations. In our analysis we worked in the Newtonian gauge
k
and solved the Einstein equations in the infrared limit, aH ≪ 1, where Weinberg’s theo-
rem guarantees the existence of two adiabatic solutions (a constant mode and a decaying
one) for both the scalar perturbations and GWs. However, an important subtlety remains:
even after fixing the local gauge completely to Newtonian gauge, there exist residual diffeo-
morphisms that preserve the Newtonian-gauge form of the metric. These transformations
generate solutions that are physically adiabatic modes but arise purely from large coor-
dinate transformations. Thus, adiabatic modes can be understood as originating from
these residual large gauge symmetries, rather than from independent dynamical degrees of
freedom. Under the action of diffeomorphism transformations

xµ 7→ x̃µ = xµ + ϵµ (t, x), (4.59)

– 48 –
there is a ϵµ which generates spatially homogeneous transformations on the metric and
preserve the Newtonian gauge [122]

ϵ0 (t, x) = −f (t) − χ(x), (4.60a)

Z
dt
ϵi (t, x) = (θδij + σij )xj − ∂i χ(x) . (4.60b)
a2 (t)

Here θ is a constant scalar and σij is a constant, traceless and symmetric matrix16 , trσij =
0. Although infinitesimal locally, these transformations become large as one approaches
infinity and are therefore called large gauge transformations. The scalar functions f (t),
χ(t) and θ act only on the scalar perturbations

Φ(t) 7→ Φ(t) + f˙(t) and Ψ(t) 7→ Ψ(t) + θ − Hf (t), (4.61)

while keep the tensor perturbations untouched. On the other hand, σij acts only on the
gravitational waves as

γij (t) 7→ γij (t) − 2σij . (4.62)

If Φ(t), Ψ(t), and γij (t) solve the spatially homogeneous Einstein equations, then any
large gauge transformation acting on these quantities produces new, spatially homogeneous
solutions as well. In particular, consider a redefinition of the background metric
A
ḡµν (t) 7→ ḡµν (t) + δgµν (t), (4.63)

where the perturbation is generated by a diffeomorphism. For a scalar-type residual dif-

feomorphism the induced metric perturbations in Newtonian gauge are

ΦA (t) = f˙(t), ΨA (t) = θ − Hf (t). (4.64)

The corresponding transformations of the background energy–momentum tensor are

δρA (t) = ρ̄˙ f (t), δPA (t) = P̄˙ f (t), δuA (t) = −f (t), (4.65)
S (t) = 0. This leads to a constant curvature
with a vanishing scalar anisotropic stress, πA
perturbation,
ζA (t) = θ. (4.66)
There is likewise a spatially homogeneous solution in the tensor sector,
A
γij (t) = 2σij , (4.67)
T A (t) = 0. At this stage it appears that the
with vanishing tensor anisotropic stress, πij
adiabatic modes we derived earlier are nothing more than gauge artifacts associated with
the k = 0 limit.
16
In general, ϵi (t, x) can have a constant term ϵi0 as well as a term like ωij xj where ωij = −ωji . Here, how-
ever, we ignored them because (due to the spatial translational and rotational symmetry of the background
metric) they do not have any effects on the linear perturbed metric.

– 49 –
 This raises the question: how does Weinberg’s theorem elevate these seemingly pure-
gauge solutions to physically meaningful modes? The essential step is the following. When
both the anisotropic stress, πij (t, k), and the non-adiabatic pressure perturbation, δPnad ,
vanish on super-horizon scales, the spatially homogeneous (k = 0) solutions can be contin-
uously extended to solutions with k ̸= 0. Since modes with nonzero wavenumber possess
no residual gauge freedom in Newtonian gauge, these extended solutions represent genuine
physical perturbations. This seemingly simple observation has a profound physical conse-
quence. During inflation, many Fourier modes exit the Hubble horizon and re-enter during
the radiation or matter eras. Weinberg’s argument shows that once an adiabatic mode
lies outside the horizon, it becomes conserved and remains constant, entirely insensitive to
whatever complicated microphysics may operate on sub-horizon scales. This remarkable
robustness allows us to relate the very early Universe to cosmological observables in the
late Universe. Therefore, in the absence of super-horizon entropy and anisotropic-stress
perturbations, inflation predicts that the primordial fluctuations are adiabatic. Note that
these scalar modes satisfy δρα /ρ̄˙ α equally for all individual constituents α of the universe,
regardless of whether energy is separately conserved for each component. Perturbations
with this property are referred to as adiabatic. Any solutions that do not satisfy this con-
dition are correspondingly called entropic. Present cosmological data agree strikingly well
with the adiabatic initial conditions implied by inflation and Weinberg’s theorem. CMB
temperature and polarization anisotropies constrain the fractional entropy contribution to
the primordial power spectrum,
Piso (k∗ )
≲ 0.03, (4.68)
Pad (k∗ ) + Piso (k∗ )
with k∗ = 0.05 Mpc−1 [133].

4.3.2 Maldacena’s Consistency Relation & Primordial Non-Gaussianity

Now, we turn to the consistency relations, which are a powerful probe of the early universe
physics and hold under very general conditions, i.e., in the presence of a long-wavelength
adiabatic mode. Maldacena’s consistency relation is, at its core, a statement about the
structure of cosmological n-point functions and the pattern of primordial non-Gaussianity
generated during inflation. At the level of linear approximation, vacuum inflationary fluc-
tuations have a Gaussian probability.

Primordial Non-Gaussianty
Primordial non-Gaussianities are most generally characterized by the connected n-
point correlation functions of the primordial scalar perturbation φ,
n
!
X
⟨φk1 · · · φkn ⟩c = (2π)3 δ 3 ki Fn (k1 , . . . , kn ), (4.69)
i=1

where statistical homogeneity enforces momentum conservation and the functions Fn

encode deviations from Gaussian statistics. The statistical properties of an isotropic

– 50 –
 Gaussian field are completely determined by the 2-point function while any odd N -
point functions with N ≥ 3 are exactly zero, and any even N -point functions with
N ≥ 4 are given in terms of the 2-point functions (called disconnected parts of
N -point functions).
The leading contribution to primordial non-Gaussianity arises from the three-point
function, or bispectrum. The deviation from Gaussianity can be formulated in terms
of the bispectrum as

⟨φk1 φk2 φk3 ⟩ = (2π)3 δ 3 (k1 + k2 + k3 ) Bφ (k1 , k2 , k3 ). (4.70)

If the power spectrum is scale invariant, dimensional analysis implies that the bis-
pectrum can be expressed in terms of two independent ratios,

Bφ (k1 , k2 , k3 ) = k1−6 Bφ (1, x2 , x3 ), (4.71)

where x2 = k2 /k1 and x3 = k3 /k1 . In Fig. 18, we present the possible momentum
configurations of the bispectrum. Different inflationary scenarios produce bispectra
that peak in different configurations [134]. The seminal work of Komatsu & Spergel
[135] established the bispectrum as the standard tool for quantifying primordial non-
Gaussianity in the CMB [136].

Figure 18: The shaded area shows the possible momentum configurations of the
bispectrum. Taking k1 to be the smallest of the three momenta, the relevant limiting
configurations are: (i) the squeezed limit k1 ≪ k2 ≃ k3 , (ii) the equilateral limit
k1 = k2 = k3 , and (iii) the folded (or flattened) limit k2 = k3 = k1 /2.

While the non-Gaussianity must be computed separately for each specific inflationary
model, an important universal result emerges: for all single-clock inflationary scenarios
(where there is only one relevant degree of freedom during cosmic inflation), the bispectrum
in the squeezed limit satisfies Maldacena’s consistency relation.
Since the tensor modes are the main focus of this work, here we present the consistency
relation for the gravitational waves. The scalar consistency relation is the same, and one
A with the (adiabatic) curvature perturbation ζ . The key physical
only needs to replace γij A

– 51 –
point behind the consistency relations is the observation that the adiabatic long-wavelength
A (t), can be removed by large coordinate transformations, i.e.
modes, δgµν
A
ḡµν (t) + δgµν (t) 7→ g̃µν (t̃) = ḡµν (t). (4.72)

Hence, adiabatic long-wavelength modes act as a classical background for the short-wavelength
modes, which freeze out much later than the long-wavelength modes. In particular, an n-
point correlation function of the short modes can be written as

⟨ζ(x1 )ζ(x2 ) · · · ζ(xn )⟩γ A (x) = ⟨ζ(x̃1 )ζ(x̃2 ) · · · ζ(x̃n )⟩, (4.73)
ij

where xi and x̃i are related as

1 A j
x̃i = xi + γij x . (4.74)
2
Now, Taylor expanding RHS around xi , we find the change of the short-distance n-point
correlation function as
n
X
δ⟨ζ(x̃1 )ζ(x̃2 ) · · · ζ(x̃n )⟩ = ⃗ I ⟨ζ(x1 )ζ(x2 ) · · · ζ(xn )⟩ + · · · .
δxI .∇ (4.75)
I=1

As a result, the (n+1)-point correlation function including the long-wavelength mode is

given as
 n 
A 1 A A
X
k
⟨γij (x)ζ(x1 )ζ(x2 ) · · · ζ(xn )⟩ ≃ γ (x)γkl (x) xI ∂l ⟨ζ(x1 )ζ(x2 ) · · · ζ(xn )⟩ , (4.76)
2 ij
I=1

in which we only keep the dominant term that has the relevant contribution. The above
equality is the consistency relation in real space. Going to the Fourier space, we arrive at
Maldacena’s consistency relation
n
1 X
⟨γλ (q)ζk1 ζk2 · · · ζkn ⟩′ ≃ − Pγvac (q) eλij (q̂)kIi ∂kIj ⟨ζk1 ζk2 · · · ζkn ⟩′ for q → 0, (4.77)
2
I=1

where the prime in ⟨· · · ⟩ indicates that we extracted the prefactor (2π)3 δ (3) (q + nI kI )
P

associated to momentum conservation. Note that the above result follows directly from
the fact that adiabatic long-wavelength gravitational waves are equivalent to a change
of coordinate for the short-wavelength mode, regardless of the super-horizon behavior of
the short modes. Therefore, as far as our inflationary system generates adiabatic tensor
perturbations, the above consistency relation holds. It is important to stress that Mal-
dacena’s squeezed-limit non-Gaussianity is not a directly observable quantity. As shown
rigorously in [137], the primordial squeezed bispectrum predicted by single-field inflation
does not imprint an observable modulation of small-scale power in late-time cosmological
data. Maldacena’s original relation was elevated to a conformal consistency relation in
[138] and to an infinite hierarchy of Ward identities in [139].

– 52 –
 Cosmological Collider Physics
In inflationary settings, heavy particles can be spontaneously produced by the ex-
pansion of the Universe [140–142]. Their interactions with curvature perturbations
generate characteristic signatures which encode the masses, spins, and interaction
structures of particles. Cosmological collider physics views the inflationary universe
as a natural high-energy probe in which massive fields present during inflation im-
print characteristic oscillatory and angular structures in primordial non-Gaussianity.
Building on early studies of inflationary correlators and quasi-single-field models by
Chen & Wang [143, 144], as well as subsequent developments by Noumi, Yamaguchi,
& Yokoyama [145], the idea was formulated explicitly by Arkani-Hamed & Mal-
dacena [146]. In this picture, heavy particles act as virtual states, encoding their
masses, spins, chemical potential in specific shapes of correlation functions accessible
to observation, thereby probing physics far beyond the reach of terrestrial acceler-
ators [147–153], among many others in a rapidly developing literature. Motivated
by these theoretical developments, there have been important observational searches
for cosmological collider signals in both CMB [154] and Large Scale Structure (LSS)
[155]. This program is naturally linked to the cosmological bootstrap, which seeks
to determine primordial correlation functions from symmetry, analyticity, and con-
sistency conditions alone [156]. Bootstrap methods classify all allowed inflationary
signatures, while the cosmological collider provides their physical interpretation in
terms of massive spinning particles in the early universe.

4.4 Inflationary Gravitational Waves at Late Times

The primordial gravitational waves produced during inflation leave observable imprints only
after going through the universe’s subsequent expansion history. Following our derivation of
the inflationary tensor spectrum and its infrared structure, we now translate these primor-
dial fluctuations into their present-day form. We begin this part by deriving the spectrum
of primordial gravitational waves sourced by vacuum fluctuations during inflation. This
spectrum spans an enormous range of frequencies—from as low as 10−17 Hz, corresponding
to wavelengths comparable to the size of the observable Universe, to as high as 108 Hz,
associated with microscopic scales near the end of inflation and the reheating epoch. We
then explore the contribution of gravitational waves to the integrated Sachs–Wolfe effect
in the cosmic microwave background.

4.4.1 The Present-Day Spectrum of Gravitational Waves from Inflation

Earlier in Sec. 4.1, we introduced the concept of horizon crossing and re-entry, and outlined
the major cosmological epochs. Next we discussed in Sec. 4.2, gravitational waves generated
during inflation arise from the amplification of quantum fluctuations in the space-time
itself. These tensor perturbations are nearly scale-invariant at horizon exit, with a power
spectrum
2 H2
Ph (k) ≃ 2 inf 2 , (4.78)
π MPl

– 53 –
where Hinf is the Hubble parameter during inflation, and MPl is the reduced Planck mass.
We are now ready to explore how the primordial GWs generated during inflation evolve
across these eras, and how they shape the gravitational waves spectrum today. The field
equation of vacuum γλ (τ, k) in the expanding background (regardless of the cosmological
era) is

γλ′′ (τ, k) + 2Hγλ′ (τ, k) + k 2 γλ (τ, k) = 0, (4.79)

where again λ = +, ×, a prime denotes a derivative with respect to the conformal time
and H = aH. We can extract analytically some general features of the solution by using
the field redefinition

h(τ, k) ≡ aγλ (τ, k). (4.80)

In the vacuum case the two polarization states evolve identically and the redefinition above
applies uniformly to both polarization modes. The field equation takes the simple form

a′′
h′′ (τ, k) + (k 2 − )h(τ, k) = 0, (4.81)
a
where ′ means derivative with respect to the conformal time and the effective mass term
a′′ 1 2
a = 2 (aH) (1 − 3w) is

2H2 (1 − ϵ) ≃ τ22 (1 − ϵ) inflation
′′

a 
= 0 radiation-era . (4.82)
a 
 1 H2 ≃ 2

matter-era
2 τ2

When the mode functions are outside the horizon ( ka > 1), the gravitational wave is a
constant while when it is well inside the horizon ( ka ≫ 1), it is a simple oscillating function
scales like 1/a
(
γ0 ( ka ≪ H)
γ(τ, k) ∝ , (4.83)
1
a sin( ka + α) ( ka ≫ H)

where γ0 and α (a phase) are both given by the initial conditions; see Fig. 19. The energy
density of the gravitational waves is given as
1 X
ρGW = ⟨γ̇λ2 ⟩, (4.84)
32πG
λ=+,×

where the expectation value denotes the statistical averages. For gravitational waves inside
the horizon, we have γ̇ ∝ k cos( ka + α)/a2 which gives

ρGW ∝ a−4 ,

as it is expected from any form of radiation.

– 54 –
Figure 19: The time evolution of γ+,× (τ, k), normalized to its initial value as a function
of ln a, for three different values of the wavelengths. Image credit: Gravitational Waves.
Vol. 2: Astrophysics and Cosmology [157].

After inflation, each mode re-enters the horizon at different times, and the energy
density in GWs evolves as radiation, redshifting as a−4 . The present-day gravitational
waves energy density per logarithmic interval in wavenumber is given by
1 dρGW
ΩGW (k) = ∝ k 2 Ph (k)T 2 (k), (4.85)
ρc d ln k

where T (k) is the transfer function encoding the evolution of each mode through the
radiation- and matter-dominated eras. The late-time gravitational-wave spectrum depends
on the horizon re-entry history of inflationary tensor modes as follows.

• Modes re-entering during the radiation era: the transfer function is nearly constant,
so the spectrum remains approximately scale-invariant ΩGW (k) ∝ const.

• Modes re-entering during the matter era: the energy density is less suppressed, scaling
as ΩGW (k) ∝ k −2 .

• Modes never left the horizon: At very high frequencies, the gravitational wave spec-
trum exhibits an exponential cutoff. These modes never exited the horizon during
inflation and therefore were not stretched or amplified. This sets a natural upper
bound on the frequency of primordial GWs, kmax = aend Hend , where aend and Hend
denotes these quantities at the end of inflation. Assuming instantaneous reheating,
the highest frequency of primordial gravitational waves is given by
r 1
kmax ∼ 108 ( ) 4 Hz. (4.86)
0.01

Modes re-entering during radiation domination start oscillating (and redshifting) ear-
lier, leading to greater suppression in ΩGW . Modes that re-enter after matter domination
experience less redshift and thus appear less suppressed today. The spectrum of gravita-
tional waves today is typically expressed in terms of frequency (in Hz), which is related

– 55 –
Figure 20: The primordial GW spectrum predicted by inflation (from quantum fluctua-
tions) for r = 10−2 (solid red line) and r = 10−3 (dashed blue line).

to the comoving wavenumber k = 2πf . Hence, the overall shape of the primordial GW
spectrum today in terms of r and f is a broken power law:

−2
r f , for f < feq ,


ΩGW (f ) ∝ r, for feq < f < fmax , (4.87)


exponential cut-off for f > fmax ,

where feq ≈ 3 × 10−17 Hz corresponds to the mode that re-entered the horizon at the time
of matter-radiation equality. This spectral break encodes important information about
the thermal history of the Universe and is shown in Fig. 20. Note that the primordial
gravitational wave spectrum from inflation can exhibit a richer structure than discussed
here once thermalization effects in the early universe are taken into account. In particular,
interactions with the thermal plasma and free-streaming SM particles can induce additional
damping features and smooth transitions in the spectrum at high frequencies, as shown in
Refs. [158, 159].

BBN Bound on the Cosmological GW Background

Big Bang Nucleosynthesis provides a robust upper bound on the total energy density
stored in a stochastic background of cosmological gravitational waves. An excess
radiation component increases the Hubble expansion rate during nucleosynthesis,
thereby altering the neutron–proton freeze-out ratio and the primordial abundances
of light elements. This requirement can be expressed as a constraint on the present-
day gravitational wave energy density fraction [157]
Z ∞
d ln q ΩGW ,0 (q) ≲ 3.5 × 10−6 , (4.88)
fCMB

– 56 –
 where ΩGW ,0 (q) denotes the spectral GW energy density today and fCMB ≃ 3 ×
10−17 Hz is the frequency associated with the comoving Hubble radius at the time of
photon decoupling. This bound limits the total gravitational wave radiation present
prior to BBN and constrains a wide range of cosmological GW sources, including
high-frequency backgrounds inaccessible to current detectors.

4.4.2 Integrated Sachs–Wolfe effect and Primordial GWs

After decoupling, the photon mean free path becomes comparable to the size of the observ-
able Universe, allowing photons to reach our detectors without further scattering (Fig. 15).
Along their journey, photon geodesics are subtly perturbed by spacetime inhomogeneities
(see Fig. 21). These perturbations give rise to temperature fluctuations, an effect known
as the Sachs–Wolfe effect, which arises from scalar and tensor modes of the underlying
geometry. In general, the total Sachs-Wolfe effect consists of two components: the original
Sachs-Wolfe effect, which arises at the surface of last scattering τdec by scalar perturba-
tion, and the integrated Sachs-Wolfe (ISW) effect, which accumulates along the photon’s
trajectory, τ ∈ (τdec , τ0 ), due to the time variation of metric perturbations (both scalar
perturbations and GWs). 17 In the following, we only focus on GWs and neglect the
contribution of scalar perturbations in the above discussion.

Figure 21: The photon geodesics are perturbed by gravitational fields, e.g., cosmic per-
turbations. That affects both the direction and energy of photons, contributing to the
Integrated Sachs-Wolfe effect. The CMB anisotropies imprinted at the scattering of last
surface (SLS) are subsequently altered by cosmic perturbations and large-scale structures
encountered along their path from the decoupling time τdec to the present epoch τ0 , giving
rise to the anisotropies we observe today.

17
Note that τrec denotes the epoch of recombination, when protons and electrons form neutral hydrogen
(z ∼ 1100–1400, Tγ ∼ 3000 K). On the other hand, τdec refers to the epoch of photon decoupling, when the
Thomson scattering rate falls below the Hubble expansion rate and CMB photons begin to free–stream;
this occurs slightly later, at z ≃ 1050 and Tγ ≃ 2800 K, and defines the true last–scattering surface relevant
for primary anisotropies. For the Integrated Sachs–Wolfe (ISW) effect, the line–of–sight integral extends
from τdec to today, since the ISW contribution arises from the evolution of gravitational potentials after
decoupling.

– 57 –
 The observed anisotropies in the CMB are related to how photons are seen from the
cosmic fluid rest frame, i.e., baryon–photon fluid. The photon’s energy measured with
respect to the rest frame of the baryon-photon fluid is

E(xν ) ≡ −P µ (xν )uµ (xν ), (4.89)

µ
where P µ = dx µ
dλ is the photon’s 4-momentum and u is the baryon-photon fluid bulk
µ
velocity. The photon’s four-momentum, P Pµ = 0, can be expressed as

1
P µ = E(1, n̂), (4.90)
a
where n̂ denotes the unit vector along its direction of propagation. Note that n̂ = −r̂
where r̂ is a radial direction in the celestial sphere (line-of-sight direction), as illustrated
in the left panel of Fig. 6. Suppose that a photon emitted with energy EE at a point, xµE ,
and is observed at xµO with EO . Then, the temperature at xµE and at xµO in a rest frame of
baryon-photon fluid are given as
EO TO
= . (4.91)
EE TE
To understand how evolving gravitational potentials leave imprints on the CMB tem-
perature through the Integrated Sachs–Wolfe (ISW) effect, we must examine how these
perturbations influence the propagation of photons. For that purpose, we need to compute
the effect of the perturbations on the photon trajectories
dP µ
+ Γµνσ (xα (η))P ν P σ = 0, (4.92)
dη

where Γµνσ (xα (η)) is the Christoffel symbols up to the first order of perturbations. Since
our focus is on temperature perturbations, we are particularly interested in the projection
of the above equation along the four-velocity uµ . For the perturbed metric Eq. (4.48), we
find the first order perturbed uµ as uµ = (1 − Φ, 0, 0, 0).
Here, we only consider the contribution of the GWs to the perturbed geodesic trajec-
tory. The perturbed geodesic is

dP 0 dt
+ Γ0ij (xα (η))P i P j = 0, (4.93)
dt dη
dt
in which dη = P 0 and Γ0µν = a2 δνj δµi ( ȧa δij + 12 γ̇ij ). At leading order, we find a(t)T̄ (t) = a0 T0
where T̄ (t) is the mean CMB temperature at time t. At first order in perturbation, we

– 58 –
arrive at 18

1 dδT (t, n̂) 1

= − γ̇ij (t, x)n̂i n̂j , (4.96)
T̄ (t) dt 2

where δT (t, n̂) denote the perturbed CMB temperature at time t. Integrating the above
equation, we find the temperature perturbation today in terms of the perturbations at the
time of the photon decoupling, τdec , as (see Fig. 21)
 Z τ0 
∆T (n̂) 1 i j ′
=− dτ n n hij (τ, (τ0 − τ )n̂) , (4.97)
T 2 τdec

where rL = τ0 − τdec . For an intuitive and detailed explanation of the gravitational effects
on temperature anisotropy, see here. This discussion is taken up again in Sec. 5.1, where
we analyze the role of inflationary gravitational waves in shaping the cosmic microwave
background and their manifestation in temperature and polarization anisotropies.

5 Detection Techniques and Observational Frontiers

Gravitational waves span an enormous range of frequencies, the gravitational rainbow il-
lustrated in Fig. 1, from wavelengths comparable to the size of the observable Universe to
frequencies of order 10 kHz and beyond, where both compact astrophysical systems and
early-Universe cosmology can act as sources. Each portion of this spectrum demands a dis-
tinct observational strategy, and taken together these complementary approaches enable
a coherent, multi-band view of the gravitational-wave Universe. In this section, we be-
gin by examining how cosmic microwave background polarization, particularly the search
for primordial B-modes, provides access to gravitational waves on the largest observable
scales, opening a unique window onto the physics of the earliest moments of cosmic history.
We then turn to pulsar timing arrays, which exploit the remarkable rotational stability of
millisecond pulsars to probe gravitational waves in the nanohertz regime. We summarize
both the current observational status and the prospects for next-generation PTA experi-
ments. Next, we outline the physical principles underlying laser interferometric detection
of spacetime strain and introduce the current global network of ground-based and space-
based interferometers. Finally, we discuss resonant and cavity-based detectors as potential
probes of gravitational waves at much higher frequencies.
18
The change in photon energy along its trajectory, including the effects of scalar perturbations, is given
by

1 dP 0 ȧ dΦ 1
=− − + Ψ̇ + Φ̇ − γ̇ij n̂i n̂j , (4.94)
P 0 dt a dt 2
1 dΦ
in which we used a
n̂.∇Φ = dt
− Φ̇ [160]. The full Sacks-Wolfe effect assuming adiabatic initial conditions,
is
Z τ0
δT 1 1
= Ψ(τdec ) + dτ (Φ̇ + Ψ̇ − γ̇ij n̂i n̂j ), (4.95)
T̄ 3 τdec 2

where the first term is the Sachs-Wolfe effect and the last three terms are the ISW effect.

– 59 –
5.1 CMB Polarization: Imprints of Primordial Gravitational Waves
In addition to the temperature anisotropies, there is more information to be gained by
measuring CMB. More precisely, CMB photons are expected to be polarized due to the
Thomson scattering by free electrons before decoupling. The polarization of CMB photons
provides the cleanest and a promising method to detect primordial gravitational waves.
Before turning to CMB polarization, let us briefly pause to introduce the systematic de-
scription of light polarization in optics.
The polarization of an electromagnetic wave is described in terms of the Stokes param-
eters. Consider a monochromatic electromagnetic plane wave that propagates along the
direction x̂3 . Its electric field can be written as the following complex 2-component vector
 
E1 eiθ1
E = e−i(wt−k.x) 
 
E e iθ2  . (5.1)
 1 
0

The coherency matrix is defined as a 2 × 2 Hermitian operator

 
⟨|E | 2 ⟩ ⟨E E ∗ ⟩
x x y 
C = ⟨EE † ⟩ =  , (5.2)
∗
⟨Ex Ey ⟩ ⟨|Ey | ⟩2

where the ⟨...⟩ denotes the ensemble average. The coherency matrix can be expanded in
the basis of Pauli matrices 19 as
 
1 1 I + Q U − iV 
C = (Iσ0 + Qσ3 + U σ1 + V σ2 ) =  . (5.3)
2 2 U + iV I − Q

Here, I, Q, U , and V are the Stocks parameters, i.e.,

I ≡ ⟨E12 ⟩ + ⟨E22 ⟩, (5.4)

Q ≡ ⟨E12 ⟩ − ⟨E22 ⟩, (5.5)
U ≡ 2⟨E1 E2 cos(θ1 − θ2 )⟩, (5.6)
V ≡ 2⟨E1 E2 sin(θ1 − θ2 )⟩. (5.7)

See Fig. 22 for the polarization corresponding to each of the Stokes parameters.

• The quantity I is the total intensity of the wave, here the temperature anisotropy.

• Q describes the difference between the linear polarization in ê1 and ê2 directions.

• The U and V parameters give information on the phases. In particular, expanding

√
the wave in the ±1 helicity polarization states, êR,L
i = 1/ 2 (ê1 ± iê2 )i , we have
2
V = ⟨ER ⟩ − ⟨EL2 ⟩, (5.8)
       
19 1 0 0 1 0 −i 1 0
The Pauli matrices σµ are σ0 =  , σ1 =  , σ2 =  , and σ3 =  .
0 1 1 0 i 0 0 −1

– 60 –
Figure 22: The Poincaré sphere representation of the polarization state of light. Each
point on the sphere corresponds to a unique polarization state. The three axes represent the
Stokes parameters S1 , S2 , S3 . Specifically, S1 = U characterizes linear polarization at ±45◦ ,
S2 = Q corresponds to linear polarization along horizontal and vertical directions, and S3
describes circular polarization, distinguishing between right- and left-handed helicities. The
radial vector from the center to a point on the sphere represents the full polarization state,
and its length gives Ipol .

which represents the difference between the positive and negative helicity intensities.
The U can be written as

U = 2⟨ER EL sin(θR − θL )⟩. (5.9)

• Here Q and U describe the linear polarization, while V quantifies the circular polar-
ization.

• The square of the total polarized intensity of light is

2
Ipol = Q2 + U 2 + V 2 , (5.10)

which quantifies how much of the light is polarized, regardless of whether it is linearly
or circularly polarized. From that, we can define the degree of polarization as
Ipol
P = , (5.11)
I
which is a quantity equal to or less than one. This leads to the constraint

I 2 ≥ Q2 + U 2 + V 2 , (5.12)

with equality satisfied in the case of fully polarized light.

– 61 –
• The most generic shape of a polarized light is elliptical, which is described by two
parameters; polarization angle θ
1 U
θ= tan−1 ( ), (5.13)
2 Q
and the ellipticity angle χ
V
sin(2χ) = . (5.14)
Ipol

These two parameters fully describe the shape of the ellipse.

Spin-Weight of the Stokes parameters

A spin-s function on the sphere under a rotation of angle φ around the n̂ transforms
as

s f (n̂) → e−isϕ s f (n̂). (5.15)

Analogous to ordinary spherical harmonics Yℓm (θ, ϕ) (basis for s = 0), there exist
spin-weighted spherical harmonics s Yℓm (θ, ϕ) which form a complete orthonormal
basis for spin-s functions
∞ X
ℓ
(s)
X
s f (n̂) = aℓm s Yℓm (θ, ϕ). (5.16)
ℓ=|s| m=−ℓ

For s ̸= 0, they are obtained by acting with the spin-raising ð+ and spin-lowering ð−
operators on the ordinary Yℓm . The operators ð± has the explicit forms
 
± ±s ∂ i ∂ 
(sin θ)∓s ,

ð = −(sin θ) ± (5.17)
∂θ sin θ ∂ϕ

which shift the spin weight as follows

ð± s f (n̂) → s±1 f (n̂). (5.18)

The spin-weighted spherical harmonics are related to the scalar (spin-0) spherical
harmonics as
q
(l−s)! + s
s Ylm (n̂) ≡ (l+s)! (∂ ) Ylm (n̂) for 0 ≤ s ≤ l (5.19)
q
(l+s)! s − −s
s Ylm (n̂) ≡ (l−s)! (−1) (∂ ) Ylm (n̂) for − l ≤ s ≤ 0. (5.20)

This implies that only the spherical harmonics with l ≥ s can give rise to s Ylm . Under
the action of parity, n̂ → −n̂, the spin-weighted harmonics transform as

s Ylm (−n̂) = (−1)l −s Ylm (n̂). (5.21)

– 62 –
Figure 23: E-mode and B-mode patterns of polarization. The cold and hot spots are
shown in blue and red, respectively.

Let us now take a second look at the Stokes parameter. Under a rotation of the
polarization basis in the transverse plane by angle ϕ, the Pauli matrices transform as

R(ϕ) σ0 R−1 (ϕ) = σ0 , R(ϕ) σ2 R−1 (ϕ) = σ2 , (5.22)

and
R(ϕ) (σ3 ± iσ1 ) R−1 (ϕ) = e∓2iϕ (σ3 ± iσ1 ). (5.23)
The coherency matrix can be written as
 
1 1 1
C= Iσ0 + V σ2 + (Q + iU )(σ3 − iσ1 ) + (Q − iU )(σ3 + iσ1 ) . (5.24)
2 2 2

As a result, the intensity, I, and the helical (circular) polarization, V , remain invariant
under rotation, i.e. spin-0. However, the linear polarizations, Q, and U , transforms like
spin-2 fields Q ± iU → e∓2iφ Q ± iU , which implies

X
(Q ± iU )(n̂) = a±2
lm ±2 Yℓm (n̂). (5.25)
ℓm

From these, we define a scalar quantity (E) and a pseudo-scalar quantity (B) (see Fig. 23),
i.e.
X
aE B


(Q ± iU )(n̂) = − ℓm ± i aℓm +2 Yℓm (n̂). (5.26)
ℓm

– 63 –
Figure 24: Left panel: Schematic of the celestial sphere showing the observable Universe,
where the radial coordinate r corresponds to cosmological redshift, ranging from the surface
of last scattering (z ≈ 1100) to the observer today (z = 0), with the reionization era marked
in the interval 6 ≲ z ≲ 10. The shaded regions in gray denote the redshifts at which the
Universe was ionized. Right panel: Thomson scattering of an electromagnetic wave by a
free electron. The incident photon with wavevector k is scattered into direction k′ at polar
angle θ.

Thomson Scattering in an Anisotropic Background

After decoupling at the surface of last scattering (z ≃ 1100), CMB photons free-
stream toward us. Along the way, mainly during reionization, they can rescatter
off free charges via Thomson scattering. In the Thomson limit (hν ≪ mc2 ), the
differential cross section is
dσT rq2 q2
= (1 + cos2 θ), rq ≡ . (5.27)
dΩ 2 4πε0 mc2
Here θ is the scattering angle, given by cos θ = ϵout · ϵ, and q, m, and rq are the
particle’s charge, mass, and classical radius. Given that (me /mp )2 ≈ 10−7 and
σT ∝ 1/m2 , electrons are far more efficient scatterers than protons.
A single Thomson scattering converts a local quadrupole anisotropy of the incident
radiation into linear polarization of the CMB, producing the observed large-scale
E-mode signal. In the nonrelativistic limit, the electron’s equation of motion is

me ẍ(t) = −e Einc (t), a(t) = −(e/me ) Einc (t),

so the acceleration is parallel to ϵ̂. The far-zone (radiation) electric field from an
accelerating charge is
e 1
re
Esc (r, t) = 2
n̂ × n̂ × a(tr ) = E0 sin θ, (5.28)
4πε0 c r r

– 64 –
 where n̂ points from the electron to the observer and tr is the retarded time (see
Fig. 24).

Figure 25: Illustration of Thomson scattering of initially unpolarized CMB photons.

For an isotropic radiation field or one with only dipole anisotropy, the scattered
radiation remains unpolarized. Polarization arises only when a quadrupole anisotropy
is present, making the quadrupole the lowest-order multipole that generates CMB
polarization.

In the last equality, we used a ∥ ϵ̂, while θ is the angle between ϵ̂ (the acceleration
direction) and n̂, and re is the classical electron radius. For fixed incident polariza-
tion, the corresponding intensity is Isc (θ) ∝ |Esc |2 ∝ sin2 θ, the characteristic dipole
radiation pattern. Now, let n̂ be the observed line of sight and (θ, ϕ) the angles of
the incident direction n̂′ in the scattering frame. For a generic unpolarized incident
intensity I(n̂′ ), the scattered linear polarization is
Z
3σT
(Q ± iU )(n̂, τ ) = ne (τ ) dΩ′ sin2 θ e±2iϕ I(n̂′ , τ ), (5.29)
16π

where e±2iϕ is the spin-±2 phase from the scattering geometry, σT is the Thomson
cross section, and ne (τ ) is the electron number density at time. Expanding the in-
cident field in spherical harmonics, I(n̂′ ) = ℓm aℓm Yℓm (n̂′ ), and using Eq. (5.25),
P

the angular integral in Eq. (5.29) picks out l ≥ 2. More precisely, the linear po-
larization, Q ± iU , transforms as a spin-±2 field on S 2 , so it admits an expansion
in ±2 Yℓm and therefore vanishes unless ℓ ≥ |s| = 2. Thus, neither the monopole
(l = 0) nor the dipole (l = 1) components of the incident radiation generate linear
polarization. The leading contribution arises from a nonzero quadrupole anisotropy,
as illustrated in Fig. 25. In other words, Thomson scattering converts a quadrupole
intensity anisotropy of the incident radiation into linear polarization on the sky.

Along the line of sight from last scattering to today, polarization is sourced wherever
free electrons see a local temperature quadrupole. In essence, the observed polarization
anisotropies can be expressed as an integral over conformal time
Z τ0
(Q ± iU )(n̂) = dτ g(τ ) S±2 (τ, n̂), (5.30)
τSLS

– 65 –
where g(τ ) = a(τ )ne (τ )σT e−τe (τ ) is the visibility function and S±2 (τ, n̂) is the polarization
source term. Given that the polarization is related to the electron number density in
the Universe, there are two key epochs: (i) recombination, when the mean free path grows
rapidly and a finite quadrupole develops, creating the acoustic E-mode pattern; and (ii) late
reionization (see left panel of Fig. 24). To compute CℓEE and CℓBB , cosmologists first solve
the linear cosmic perturbation equations to obtain the polarization source from primordial
anisotropies. This source feeds into the line-of-sight integral, which is then projected onto
spin-2 spherical harmonics to extract E- and B-mode components. Tools like CAMB and
CLASS perform this numerically [161]. A detailed computation is beyond the scope of the
current review, and the final result of these impressive calculations is shown in Fig. 26,
where the T T , EE, and BB spectra (both lensing and primordial) are presented. In the
following, we summarize the main features of the CMB polarization spectra.

Figure 26: CMB power spectra for temperature anisotropy (CMB-TT), E-mode polar-
ization (CMB-EE), and B-mode polarization (CMB-BB) are shown. The solid curves
represent the predictions of the best-fit ΛCDM model, including a scale-invariant tensor
perturbation with a tensor-to-scalar ratio r = 0.004. For comparison, the thin dashed line
indicates the contribution of tensor modes to the B-mode spectrum alone. Also displayed
are current measurements of the CMB power spectra (colored points) and the projected
polarization sensitivity of LiteBIRD (black points). Image credit: LiteBIRD mission [50].

• The scalar quantities E and B, completely determine the linear polarization fields.

• The E-mode is curl-free and even under the action of parity. Its polarization vectors

– 66 –
 are radial around cold spots (under dense) and tangential around hot spots (over
dense).

• The B-mode is the divergence-free field and odd under parity. Its polarization vectors
have vorticity around the under- and over-dense areas.

• The angular power spectra are defined as

1 X X Y
ClXY ≡ ⟨a a ⟩ where X, Y = T, E, B. (5.31)
2l + 1 m lm lm

The autocorrelations of E- and B-modes, denoted by EE and BB, are presented in

Fig. 26.

• At linear order, scalar (density) perturbations source only E-modes, while tensor
perturbations (gravitational waves) generate both E- and B-modes.

• At nonlinear order, gravitational lensing by large-scale structure distorts E-modes

into B-modes, producing a lensing-induced B-mode signal at small angular scales.

Primordial gravitational waves and B-modes have not yet been detected. Recalling
that scalar perturbations do not generate B-modes, while tensor (gravitational wave) per-
turbations do, a detection of primordial B-modes would be a distinctive fingerprint of
primordial gravitational waves and thus a powerful confirmation of inflation. This is the
ambitious goal of the upcoming LiteBIRD satellite mission: to measure the CMB polariza-
tion with unprecedented sensitivity and probe tensor-to-scalar ratios as low as r = 10−3 ,
pushing the frontier of our understanding of the early Universe. For additional information
on LiteBIRD, see [50]. CMB B-modes also offer a unique window into a broad range of
new physics including parity violation, relic particle species. We refer the interested reader
to the recent reviews article [5] for a comprehensive overview. Finally, CMB spectral dis-
tortions act as an intermediate probe of primordial gravitational waves, bridging B-mode
polarization measurements and pulsar timing array observations [162].

5.2 Pulsar Timing Arrays: Precision Cosmic Clocks for GW Detection

A population of exceptionally stable millisecond pulsars, timed with precisions approaching
100 ns over many years, acts as an array of cosmic clocks. Their remarkably regular pulses
are subtly and coherently disturbed by gravitational waves propagating along the Earth–
pulsar line of sight: the passing waves perturb the null geodesics along the photon’s path
to Earth. A global collaboration now monitors dozens of these pulsars with ever-increasing
precision, effectively transforming the Milky Way into a galaxy-scale detector for nanohertz
gravitational waves through pulsar timing arrays (PTAs).
By tracking the arrival times of pulses from a network of pulsars, PTAs are sensitive
to GWs in the nanohertz band. Crucially, PTAs are sensitive not only to individual
deterministic sources, but also to stochastic gravitational wave backgrounds (SGWBs).
Indeed, the strongest existing bounds on the primordial nanohertz background arise from

– 67 –
Figure 27: Pulsar timing arrays detect nanohertz gravitational waves by measuring timing
residuals across widely separated millisecond pulsars.

PTA measurements, providing leading constraints on the gravitational wave energy density
in this frequency band.
Pulsar timing arrays operate by monitoring the times of arrival (TOAs) of radio pulses
from their ensemble of millisecond pulsars, whose exceptional rotational stability makes
them superb astrophysical clocks [163, 164]. A passing GW propagating in direction q̂
perturbs the spacetime metric along the photon trajectory, inducing a small but coherent
shift in the observed pulse frequency. The measured fractional frequency shift at Earth
can be expressed as

δν(t) 1 n̂i n̂j  

= γij (te , xe ) − γij (tp , xp ) , (5.32)
ν 2 1 − q̂· n̂
where n̂ points from the pulsar to Earth, te is the reception time at Earth, and tp =
te − L(1 + q̂ · n̂) is the retarded emission time at the pulsar, with L the pulsar distance.
Here γij (te , xe ) denotes the gravitational wave metric perturbation evaluated at the Earth
when the pulse is received, the Earth term, while γij (tp , xp ) denotes the perturbation at
the pulsar at the time of emission, constituting the pulsar term. The corresponding timing
residual is obtained by integrating the frequency shift as

δν(t′ )
Z t
R(t) = dt′ . (5.33)
ν

Gravitational waves in the nanohertz band (f ∼ 10−9 –10−7 Hz) induce correlated timing
residuals across the pulsars in a PTA. The essential observable for a stochastic background
is the two-point function
Cab ≡ ⟨Ra (t) Rb (t)⟩, (5.34)
which captures the GW-induced covariance between pulsars a and b. This geometry is
illustrated schematically in Fig. 27. The pulsar–pair angular separation ζab is defined as

– 68 –
Figure 28: Left panel: Angular correlation of pulsar timing residuals measured by
NANOGrav (blue points with uncertainties) [24], shown as a function of the separation
angle between pulsar pairs. The dashed black curve is the Hellings–Downs prediction for an
isotropic GWB with spectral index γ = 13/3. The data exhibit the characteristic quadrupo-
lar trend expected from gravitational waves, whereas the absence of a signal would yield
correlations consistent with zero across all angles. Image credit: The NANOGrav Collab-
oration [24]. Right panel: Characteristic strain sensitivity curves for current pulsar timing
array (PTA) collaborations. The blue, orange, and green curves show the sensitivities of
EPTA+InPTA, NANOGrav, and PPTA, respectively. The dashed black line indicates the
joint-median sensitivity obtained by combining information from all PTA datasets. Image
credit: [165].

the angle between the sky directions of pulsars a and b

cos ζab = n̂a · n̂b . (5.35)
For an isotropic and stationary gravitational wave background, this covariance depends
only on the angular separation between the pulsars, and may be written as
2
Cab = Γab (ζab ) σGW , (5.36)
2
where σGW encodes the overall amplitude of the background and Γab (ζab ) is the geometrical
overlap reduction function which depends only on the angular separation between the
pulsars when the background is isotropic. In general relativity, only the two transverse–
traceless tensor modes are present. In this special case, ξab reduces to a single, characteristic
quadrupolar correlation function,
     
HD 3 1 − cos ζab 1 − cos ζab 1 1 − cos ζab 1
Γab = ln − + δab , (5.37)
2 2 2 4 2 2
known as the Hellings–Downs (HD) curve. Notice that the term Hellings–Downs correlation
therefore refers specifically to the ORF produced by an isotropic background of tensor
gravitational waves in general relativity, and does not generally apply when additional
polarizations or modified propagation effects appear in modified–gravity theories.
Recent PTA measurements provide compelling evidence for a nanohertz gravitational
wave background (see left panel of Fig. 28). The NANOGrav 15-year data set [24] identi-
fies a common-spectrum process exhibiting the expected Hellings–Downs correlations [24].

– 69 –
Figure 29: The International Pulsar Timing Array (IPTA), a worldwide network of radio
telescopes that times an ensemble of millisecond pulsars to search for nanohertz gravita-
tional waves. Image credit: H Thankful Cromartie.

Independent analyses by the European Pulsar Timing Array (EPTA) together with the
Indian Pulsar Timing Array (InPTA) [166–168], the Parkes Pulsar Timing Array (PPTA)
[169, 170], and the Chinese Pulsar Timing Array (CPTA) [171] report consistent signals.
Together, these regional collaborations form the International Pulsar Timing Array (IPTA),
and collectively these results mark the emergence of gravitational-wave astronomy in the
nanohertz regime (see Fig. 29), likely dominated by the cosmic population of slowly in-
spiraling supermassive black-hole binaries. The characteristic strain sensitivity curves of
current PTA collaborations are shown in the right panel of Fig. 28. Future observations
with the Square Kilometre Array (SKA) are expected to significantly improve timing pre-
cision and greatly enhance PTA sensitivity [172, 173]. Astrometric observations provide
an alternative to pulsar timing arrays for probing the stochastic GWB and its angular
properties [27, 28]. Together with LISA and ground-based interferometers, PTAs anchor a
coordinated multi-band gravitational wave network spanning nearly nine orders of magni-
tude in frequency, enabling a unified and comprehensive view of compact-object populations
and their evolution across cosmic time.

5.3 Laser Interferometry: Earth- and Space-Based GW Observatories

The idea of detecting gravitational waves using laser interferometry dates back to the early
1960s, ultimately culminating in the construction of the LIGO and Virgo observatories,
whose landmark detections opened an entirely new observational window onto the Universe
[4] (see Fig. 30). These achievements established laser interferometers as the premier tools
for probing gravitational waves in the audio-band and demonstrated, for the first time, the
direct measurability of spacetime dynamics. To set the stage for the methods reviewed in
this section, let us begin with a brief overview of Michelson interferometers and the physical
principles that govern their response to passing gravitational waves. This foundational
framework will allow us to understand how modern detectors extract minute strain signals

– 70 –
from noise and how their design has evolved into the sophisticated global network operating
today.

Figure 30: Time-domain strain data for the first directly detected gravitational wave
event, GW150914, as measured by the LIGO Hanford (H1) and Livingston (L1) detectors.
The top panels show the observed detector strains, with the Livingston trace time-shifted
and inverted to account for the light-travel delay and detector orientation. The middle
panels compare the data with numerical-relativity waveforms and reconstructed signals
obtained via wavelet and template methods, demonstrating the consistency of the inspi-
ral–merger–ringdown signal with a binary black-hole coalescence. The bottom panels show
the residuals after subtraction of the best-fit waveform. Image credit: LIGO Scientific
Collaboration and Virgo Collaboration [4].

Michelson Interferometer
Consider a laser emitting monochromatic light in direction x̂ with an electric field
 
Ein,1
E⃗ in,1 (t, x) =   with Ein,1 = E0 e−i(ωL t−kL ·x) , (5.38)
0

where ωL is the laser frequency, kL its wavevector, and E0 the amplitude. The laser
beam is directed onto a beam splitter, which divides it into two components of equal
probability amplitude: one propagating along the x̂-axis and the other along the
orthogonal ŷ-axis. A lossless 50:50 beam splitter (half-reflecting, half-transmitting
mirror) can be described by a 2 × 2 unitary matrix, beam-splitter transfer matrix, as
 
1 1 i
U=√  . (5.39)
2 i 1

– 71 –
 The factor of i corresponds to a relative π/2 phase shift for the reflected beams. This
convention ensures unitarity, i.e. total intensity is conserved, and the beams interfere
correctly. It thus splits the initial laser electric field into E spl,1 as

E spl,1 = U E in,1 . (5.40)

Next, each beam travels down its respective arm, reflects off a mirror at the far end,
and returns to the beam splitter. We place the origin of the coordinate system at the
beam splitter. The mirror terminating arm 1 is located at (Lx , 0), while the mirror
at the end of arm 2 is positioned at (0, Ly ). At this point we have

E0 E0
Eref,1 = − √ e−i(ωL t−2kL Lx ) , Eref,2 = −i √ e−i(ωL t−2kL Ly ) . (5.41)
2 2
The beam splitter then recombine and re-split the two lights. In the recombination
stage, the returning beams impinge on opposite faces of the beam splitter compared
to the forward pass. This geometric reversal flips the sign of the reflection coefficient
for one arm, represented compactly by the diagonal surface matrix
 
1 0
S = σ3 =  . (5.42)
0 −1

Hence, the total transformation for the returning fields is given by U S, i.e.
 
e 2ikL Ly +e 2ik L Lx
E0
E out = U S E ref = − e−iωL t  . (5.43)
2 i(e L x − e2ikL Ly )
2ik L

Consequently, when the arm lengths are equal (Lx = Ly ), one output port remains
dark while the other is bright. We put the potodetector at the dark port. The total
power recorded at the photodetector is proportional to

|Edet |2 = E02 sin2 kL (Lx − Ly ) .

(5.44)

Thus, any variation in the relative arm lengths induces a corresponding modulation
in the detected power (see Fig. 31).

Interaction of Interferometer with Gravitational Waves

We now turn to how this fundamental principle enables the detection of GWs. For clarity,
let us consider a wave with only the plus polarization propagating along the z-axis, i.e.
ds2 = −c2 dt2 + [1 + h+ (t, z)]dx2 + [1 − h+ (t, z)]dy 2 + dz 2 , (5.45)
where h+ (t, z) = h0 cos(ωgw t − kgw z). In the z = 0 plane of the interferometer, we therefore
have
h+ (t) = h0 cos ωgw t. (5.46)

– 72 –
Figure 31: The LIGO detector operates as a Michelson interferometer: passing gravita-
tional waves alternately stretch one arm and compress the other, inducing a phase shift in
the light at the beam splitter that produces an interference pattern at the output.

Photons propagate along null geodesics, ds2 = 0. For light traveling in arm 1, the path is,
to first order in h0 ,

dx = ±c dt (1 − 21 h+ (t)), (5.47)

where the plus sign corresponds to propagation from the beam splitter to the mirror, and
the minus sign to the return trip.
In the TT gauge description, the coordinates of the mirrors and of the beam-splitter
are not affected by the passage of the wave and the physical effect of the GW is manifested
in the fact that it affects the propagation of light between these fixed points. Consider a
photon that leaves the beam-splitter at a time tspl,1 . It reaches the mirror, at the fixed
coordinate x = Lx , at a time tref,1 obtained
Z tref,1 Z tref,1
1 c
Lx = c dt[1 − h+ (t)] = c(tref,1 − tspl,1 ) − h+ (t)dt. (5.48)
tspl,1 2 2 tspl,1

Then the photon is reflected and reaches again the beam-splitter at a time tout,1 obtained
integrating with the minus sign

c tout,1
Z
Lx = c(tout,1 − tref,1 ) − h+ (t)dt. (5.49)
2 tref,1

As a result, the arrival time tout,1 after a round trip in the arm-1 is
Z tout,1
1 2Lx
tout,1 − tspl,1 = h+ (t)dt + . (5.50)
2 tspl,1 c

– 73 –
Given that the GWs is small, we can solve it perturbative for tout,1 as
Z tspl,1 + 2Lc x
1 2Lx
tout,1 − tspl,1 ≃ h+ (t)dt + , (5.51)
2 tspl,1 c

which gives

Lx sin(ωgw Lx /c) 2Lx

tout,1 − tspl,1 ≃ h0 cos(tspl,1 + Lx /c) + . (5.52)
2c ωgw Lx /c c

Since LIGO is fundamentally a time-delay interferometer, it is necessary to determine the

corresponding travel time for the second arm. Doing the same analysis for arm-2, we find

Ly sin(ωgw Ly /c) 2Ly

tout,2 − tspl,2 ≃ − h0 cos(tspl,1 + Ly /c) + . (5.53)
2c ωgw Ly /c c

Now, we set

L = Lx = Ly , t = tout,1 = tout,2 , (5.54)

and find the total phase difference induced by GWs in the Michelson interferometer as

ωL L sin(ωgw L/c)
∆ϕ = h(t − L/c). (5.55)
c ωgw L/c

Therefore, the total power observed at the photodetector, P ∼ |Edet |2 , is modulated by

the GW signal.
For a simple Michelson interferometer with arm length L = 4 km the GW-induced
single-pass response contains sin(ωgw L/c); the condition for maximal response, sin(ωgw L/c) =
1, occurs at ωgw L/c = π/2, i.e.
πc c
ωgw = , fgw = ≃ 1.9 × 104 Hz. (5.56)
2L 4L
In contrast, LIGO is designed to detect signals at frequencies that are orders of magnitude
lower, down to ∼ 10 Hz, which would correspond to an optimal arm length
c
Lopt = ≃ 7500 km, (5.57)
4fgw

far exceeding any feasible terrestrial baseline. LIGO instead operates in the long-wavelength
regime (ωgw L/c ≪ 1), where the interferometer response is nearly flat, and achieves high
sensitivity through Fabry–Pérot arm cavities, which increase the photon storage time by
a factor of O(102 ) (see Fig. 32). The effective optical path length is thus enhanced to
∼ 103 km while keeping the physical arms at 4 km, allowing the detector to probe frequen-
cies across the 10–103 Hz observational band [4, 174].
Advanced LIGO can detect compact binaries with total masses up to O(102 ) M⊙ ; how-
ever, more massive systems, such as supermassive black hole binaries, radiate at lower
frequencies, beyond the reach of ground-based detectors. To extend sensitivity, next-
generation observatories will employ advanced technologies to form a global network of

– 74 –
Figure 32: Network of laser interferometer gravitational wave detectors on Earth (left)
and in space (right). The terrestrial detectors are L-shaped Michelson interferometers with
kilometre-scale arms, including LIGO, which began its first observing run in 2015, Virgo,
which joined in 2017 [4], and KAGRA, operational since 2020 [30], together establishing the
first global ground-based gravitational wave network. Forthcoming facilities, LIGO–India,
expected to begin observations in the early 2030s [31], and the Einstein Telescope, planned
for the mid-2030s to 2040s [32], will further extend this network’s reach. In contrast,
proposed space-based interferometers will form triangular constellations with arm lengths
a million times larger, of order 106 km, dramatically enhancing sensitivity to low-frequency
gravitational waves. The LISA mission is scheduled for launch around 2030 [29], while
several complementary proposals, Taiji [175, 176], TianQin [177], and DECIGO [178], aim
to explore similar frequency bands with distinct orbital configurations and desigs. Image
credit: CERN Courier Feature (Aug 2023) [179].

terrestrial interferometers, probing gravitational waves across the ground-accessible fre-

quency band (∼ 1 Hz–10 kHz). At still lower frequencies, they will be complemented by
space-based missions and completing a multi-band gravitational wave detection network
spanning from 10−4 Hz to the kilohertz regime. Fig. 32 illustrates the global network
of ground-based laser-interferometer detectors, while their corresponding sky sensitivities
are shown in Fig. 33. Atom interferometry offers a complementary route to established
laser-interferometric techniques for GW detection [33–35]. On Earth, several long-baseline
cold-atom experiments are under construction or commissioning [36–38], while large-scale
facilities such as ELGAR [39] and AION [40] are being developed. In space, ambitious
mission concepts such as AEDGE [41], STE-QUEST [42], and CAL/MAIUS [43] aim to
extend atom-interferometric sensitivity toward even lower GW frequencies. Together, these
terrestrial and orbital instruments target the mid-frequency GW band, approximately 10−2
to 1 Hz.

5.4 Photonic Qumodes: a Cutting-Edge Quantum Platform for GW Detection

Gravitational waves have opened an unprecedented observational window onto the uni-
verse, yet this window currently extends only up to frequencies of order 10 kHz. The
detection of higher-frequency gravitational waves (HFGWs) remains an open frontier and

– 75 –
Figure 33: Sensitivity curves of current and future laser-interferometer gravitational wave
detectors (solid colored lines), including ET-D, Advanced Virgo, Advanced LIGO, LISA,
Cosmic Explorer (CE), and DECIGO. The dashed black curves correspond to the charac-
teristic strain amplitude and associated frequencies of intermediate-mass black-hole bina-
ries with total masses between 2 × 102 to 2 × 104 M⊙ , assuming a luminosity distance of
dL = 1 Gpc. These trajectories illustrate how binaries of different masses sweep through
the sensitivity bands of detectors across multiple frequency decades. Image credit: The
Science of the Einstein Telescope [32].

an area of active research and development. Such HFGWs may originate from a wide range
of cosmological and astrophysical processes, and could even be generated in laboratory set-
tings through the conversion of high-intensity laser fields (see [180] and the references
therein). Their observation would provide access to rich new information about the early
universe, particle physics, and potentially the quantum nature of gravity itself. Detect-
ing gravitational waves at frequencies above that regime, however, requires experimental
strategies fundamentally different from those employed by large-scale laser interferometers.
As the GW frequency increases, the corresponding wavelength becomes much shorter than
any practical detector baseline, leading to a rapid degradation of the strain sensitivity of
conventional interferometric techniques. This limitation has motivated the exploration of
alternative detection concepts that rely not on large spatial separations, but on resonant
enhancement, electromagnetic interactions, and quantum-coherent effects.
Joseph Weber pioneered the first experimental searches for gravitational waves using
resonant bar detectors operating in the kilohertz regime [181], thereby initiating the explo-
ration of gravitational waves at the highest frequencies accessible at the time and laying the
conceptual foundations for modern high-frequency GW detection. The main experimen-
tal approaches currently pursued for HFGW detection include high-frequency extensions of
laser interferometers [182, 183], resonant mass detectors [184], modern resonant mechanical
sensors [48, 185, 186], electromagnetic oscillators such as microwave cavities [187, 188], pho-

– 76 –
ton (re-)generation experiments [49, 189–194], and other electromagnetic conversion–based
schemes. Comprehensive reviews, sensitivity estimates, and discussions of experimental
implementations can be found in Ref. [180] and the references therein. In the following,
we focus on a brief overview of photon (re-)generation experiments and highlight recent
advances in this direction.

Graviton–Photon Transitions: The Gertsenshtein Effect

In a seminal 1962 paper, Gertsenshtein pointed out that electromagnetic and grav-
itational waves can interconvert in the presence of a static magnetic field, i.e. the
Gertsenshtein effect [195]. This observation provides an early and concrete example
of the coupling between electromagnetism and gravity at the level of field theory
and forms the basis of many modern proposals for high-frequency gravitational wave
detection. This effect can be understood from the Maxwell–Einstein action by con-
sidering electromagnetic fields propagating in a weakly curved spacetime. We write
the spacetime metric as a small perturbation around flat space,

gµν = ηµν + γµν , |γµν | ≪ 1,

where γµν represents the gravitational perturbation. To leading order in γµν , the
interaction between the gravitational field and any field in nature is
1
Hint = γµν T µν , (5.58)
2
where T µν is the energy–momentum tensor of a generic field. For the electromagnetic
field, this interaction Hamiltonian density can be written explicitly as
 
1 λσ µν 1 µν
Hint = − η γ − γη Fµλ Fνσ , (5.59)
2 2
where Fµν is the electromagnetic field-strength tensor and γ ≡ γ µ µ is the trace
of the metric perturbation. Note here we adopt natural units and set c = ℏ =
ϵ0 = 1. The gravitational interaction Eq. (5.59) reveals how spacetime perturbations
directly influence electromagnetic dynamics, opening a channel for graviton–photon
conversion in a background field. The transition amplitude from an initial state |i⟩
to a different final state |f ⟩ is given by the S-matrix element
 Z 
4
Af i = ⟨f |T exp −i d x Hint |i⟩, (5.60)

where T is the time-ordering operator.

– 77 –
 Photon regeneration is a leading approach to probe HFGWs, which is based on this
mechanism [196]. In such experiments, GWs traverse a photon-shielded cavity permeated
by a strong magnetic field, allowing for resonant graviton-to-photon conversion. Consider
a cylindrical electromagnetic cavity of length L and cross-sectional area A. We assume
that the cavity walls are made of a perfect conductor. A uniform, static magnetic field is
turned on inside the cavity, oriented along the ẑ direction, i.e.

B(t, x) = B0 ẑ. (5.61)

At this stage, it is necessary to fix the gauge. The electromagnetic fields measured in the
laboratory are most naturally described in the Detector Proper (DP) frame, i.e., the local
inertial frame in which the detector is in free fall. By contrast, theoretical descriptions of
gravitational waves often employ the TT gauge, in which freely falling test masses remain
at fixed coordinate positions while the effect of the gravitational wave is entirely encoded
in oscillations of the metric. In this work, we adopt the TT gauge, defined by h00 = h0i =
∂i hij = 0, where i, j = 1, 2, 3. We further assume that the background magnetic field B0 is
static in the TT frame. Although, in general, the transformation between the DP and TT
frames in the presence of gravitational waves requires careful treatment, this approximation
is well justified in the high-frequency regime ωL ≫ vs [197]. Here vs denotes the sound
velocity in the detector material, which is typically small, vs ≲ 10−4 .

Figure 34: Illustration of a resonant cavity prepared in the electromagnetic state |i⟩, which
transitions to |f ⟩ as a passing gravitational wave induces graviton–photon conversion in a
strong magnetic field.

We begin by considering the interaction of the cavity with an unpolarized, monochro-

matic gravitational wave of frequency ω, propagating in the direction q̂ (with q = ωq̂)
as
X
hij (t, x) = eσij (q̂) ĥω eiqx + h.c., (5.62)
σ

the interaction Hamiltonian density can be written as

X
Ĥint = −B0 sin θq δ B̂i (x) eσi (q̂) ĥω eiqx , (5.63)
σ

– 78 –
where δ B̂i (x) denotes the magnetic-field operator associated with fluctuations of the elec-
tromagnetic field, and
1
e±
j (q̂) = √ (cos θq cos ϕq ∓ i sin ϕq , cos θq sin ϕq ± i cos ϕq , − sin θq ), (5.64)
2
in cartesian coordinate. A cosmological stochastic background of unpolarized gravitational
waves can be modeled as a superposition of plane waves with different frequencies and
propagation directions as
XZ Z 2
σ d q̂
hij (x) = d ln ω eij (q̂) ĥω eiqx + h.c., (5.65)
σ=±
2π

where q = ωq̂, hω is the mode function, and ⟨hij (t, x)hij (t, x)⟩ = 2 d ln ω|hω |2 . In this
R

stochastic setting, the interaction Hamiltonian is obtained by superposing the monochro-

matic contributions Eq. (5.63) over all frequencies and directions, with the cavity response
effectively integrating over the accessible bandwidth.
Let us focus on electromagnetic transitions occurring inside the cavity, of the form

|i⟩ 7−→ |f ⟩ = b̂†ω |i⟩ . (5.66)

where b̂†ω denotes the creation operator of a photon with frequency ω. More precisely, we
consider transitions in which a single photon is created within the cavity, thereby increasing
the occupation number of the mode at frequency ω by one. The transition rate for the
process |i⟩ → |f ⟩, evaluated to first order in perturbation theory with respect to the GW
amplitude hω , is given by
Z
1 2
Γf i = lim d4 x ⟨i| b̂ω Hint (x) |i⟩ . (5.67)
T →∞ T

To proceed further and evaluate this expression explicitly, it is now necessary to specify
the form of the initial state |i⟩. The current active Photon regeneration experiments are
OSQAR [189] and ALPS [190] at optical frequencies, and CAST [191] at X-ray frequen-
cies. In addition, BabyIAXO and IAXO [192, 193] are under development, and JURA [194]
proposed. All of these experiments are based on assuming initial vacuum state inside the
cavity, i.e. |i⟩ = |0⟩. Recently, a qualitatively new approach has been put forward under the
name QuGrav [49]. This proposal exploits highly occupied n-photon states (qumodes) in-
side the cavity, allowing the graviton-induced transition rate to be parametrically enhanced
by Bose–Einstein statistics. By coherently populating the cavity with a large number of
photons, the conversion process benefits from stimulated emission, leading to a substan-
tial enhancement relative to vacuum-based schemes. These experiments detect the photon
count rate Γ within an energy band ∆f = (ff − fi ), with a single photon detection effi-
ciency ε, and over a cross-sectional area A. The background noise is quantified in terms of
the dark count rate in that frequency bin, i.e., ΓD (f ). It is the rate at which the detector
falsely registers a photon even when no real photon is present. In Table 5 we summarized
these experiments and proposals.

– 79 –
 f [Hz] B0 [T] L [m] A [m2 ] ΓD [mHz] ε Cavity state Ref.
15 −4
OSQAR II 10 9 14.3 5 × 10 1.14 0.9 [189]
15 −3 −3
ALPS II 10 5.3 106 10 10 0.75 [190]
15 −3 −6 vacuum |0⟩
JURA 10 13 960 8 × 10 10 1 [194]
18 −3
CAST 10 9 9.26 2.9 × 10 0.15 0.7 [191]
18
IAXO 10 2.5 20 3.08 0.1 1 [192, 193]
10 −3
QuGrav 10 10 1 1 10 1 Qumode |n⟩ [49]

Table 5: Experimental parameters of photon-regeneration setups. All experiments assume

a vacuum cavity state |0⟩ except QuGrav, which considers a general Fock state |n⟩. Here
ΓD (f ) is the dark count rate in a frequency bin of width ∆f , and ε is the single-photon
detection efficiency. Italicized entries indicate proposals.

From Qumodes to QuGrav

In the context of quantum information science, qubits constitute the canonical build-
ing blocks for storing and processing quantum information, encoding logical states
in arbitrary superpositions of the basis states |0⟩ and |1⟩. While this abstraction
underlies much of quantum computation, many advanced applications benefit from
moving beyond strictly two-level systems. Within this broader framework, quantum
harmonic oscillators, commonly referred to as qumodes, provide a natural extension
of the qubit paradigm. By replacing or supplementing qubits with bosonic modes,
qumodes offer a countably infinite-dimensional Hilbert space spanned by Fock states
|n⟩, with n = 0, 1, 2, . . . [198–200]. This expanded state space enables richer encod-
ing strategies and greater flexibility in the manipulation of quantum information,
positioning qumodes as a powerful platform for continuous-variable quantum com-
putation and cavity-based quantum technologies.
Recent technological breakthroughs have significantly expanded our ability to ma-
nipulate and measure complex quantum states of light. In particular, experimen-
tal platforms have begun to realize true n-photon qumodes with high fidelity and
control [201, 202], enabling exploration of rich, high-dimensional Hilbert spaces be-
yond the qubit paradigm. A key milestone in this direction is the demonstration of
photon-number-resolving measurements capable of distinguishing photon counts up
to n = 100 [203], far exceeding the few-photon regime of traditional detectors.
Building on these developments, Kharzeev, Shalamberidze, and the present author

– 80 –
 highlighted the transformative potential of this major advance in quantum measure-
ment capability for the detection of HFGWs [49]. In particular, they proposed ini-
tializing the cavity in an n-photon Fock state rather than the vacuum, demonstrating
that for large occupation numbers, n ∼ 102 , the sensitivity of photon-regeneration
cavities can be enhanced by up to two orders of magnitude. More preciously, given
that |⟨n + 1| b† |n⟩|2 ∝ (n + 1), we have Γf i ∝ (n + 1). This ambitious proposal was
termed QuGrav, reflecting its quantum-enhanced approach to gravitational wave de-
tection.

Following the QuGrav proposal, we consider an n-photon Fock state as the initial
state of the cavity in order to explore genuine Bose enhancements associated with finite
photon occupation. The formalism, however, remains completely general: by setting n = 0,
one straightforwardly recovers the standard results corresponding to vacuum-initialized
cavities. As an initial configuration, we consider an n-photon state at a single resonant
frequency ω,

|i⟩ = |Qn ⟩ = |n(ω)⟩γ , (5.68)

where |n(ω)⟩γ = (b†ω )n |0⟩γ . Now using the above initial state in Eq. (5.67), and considering
a monochromatic gravitational wave with the same frequency in Eq. (5.63), we find the
transition rate to |f ⟩ = |Qn+1 ⟩ as

Γf i = 2(n + 1) QV B02 h2ωg sin2 θq , (5.69)

where Q is the quality factor of the cavity. Remarkably, due to Bose–Einstein statistics,
once the frequency of the GW matches the frequency of the n-photon mode inside the
cavity, the probability of graviton-to-photon conversion is enhanced by a factor of n + 1.
This process is the EM analogue to the stimulated absorption of gravitons in the massive
quantum acoustic resonators, with the process being most effective at frequencies that
match the resonator’s natural frequency [48].
To complement this minimal setup and to enhance sensitivity to the more challenging
gravitational wave background, we now introduce a more elaborate qumode configuration
that illustrates the theoretical reach and long-term potential of this approach for broadband
detection. Specifically, we consider a multimode qumode state in which each resonant cavity
mode is prepared in an n-photon Fock state, with mode frequencies given by ωl = πl/L,
namely
⌊2∆f L/c⌉
O
|Qn ⟩ = |n(ωl )⟩γ . (5.70)
l=1

Here |n(ωl )⟩γ = (b†ωl )n |0⟩γ , and ⌊2∆f L/c⌉ denotes the number of resonant modes supported
by the cavity within the finite frequency bandwidth ∆f (with f = ω/2π).
Considering the stochastic gravitational wave background described in Eq. (5.65), and
assuming for simplicity that the spectrum is approximately flat across this frequency in-
terval, the total transition rate is obtained by summing the contributions from all resonant

– 81 –
modes, as

Γtot = 2 (n + 1) ∆f L V B02 hω2 sin2 θq . (5.71)

Here ∆f characterizes the effective detection bandwidth of the cavity. While a bright
coherent drive also enhances the generation rate by the factor n + 1, it simultaneously
introduces amplitude and phase noise. For a quantitative comparison see [49].
For completeness, let us also consider a single graviton propagating in direction q̂ with
frequency ωg , i.e. |1ωg ⟩g , as
X
hij (t, x) = eσij (q̂) hσωg âωg ,σ e−iωg t+iq.x + h.c., (5.72)
σ=±

where â†ω,σ and âω,σ are the graviton creation and annihilation operators, [âω,σ , â†ω′ ,σ′ ] =
δσσ′ δ(ω − ω ′ ). In this case, we set the initial and final states as

|i⟩ = |Qn ⟩ |1ω ⟩g , |f ⟩ = b̂†ω |Qn ⟩ |0⟩g , (5.73)

which is the conversion of a single graviton to a new photon inside cavity. The transition
rate in this case is equal to Eq. (5.69).

Single Gravitons

Modern detector architectures can detect single photons and even count them up
to n = 100 [201, 202], but building a detector sensitive to single gravitons remains
an extraordinary challenge. In fact, Dyson [204], and independently Rothman &
Boughn [205], have argued that such detection is fundamentally impossible within
certain experimental architectures, while remaining feasible in others. Recently, this
question has been revisited in light of recent technological advances [48, 206].
If gravity is quantized in the same manner as the other fundamental interactions, the
number of gravitons contained within a single de Broglie volume is

π h2ω MPl
2
Ngrav = ngrav λ3dB = , (5.74)
2f 2

where λdB = f −1 . As an example, a signal of LIGO detection with hω ∼ 10−25 at

f ∼ 102 Hz, includes Ngrav ∼ 1029 gravitons. Requiring less than a graviton per de
Broglie volume, we arrive at
 f 
hω ≤ 2 × 10−31 , (5.75)
103 GHz

– 82 –
 where below that limit, the gravitational field is more accurately described as a highly
dilute graviton gas rather than a classical gravitational wave.
Naively, one might expect that observing a gravitational wave with an amplitude
below this particle-like threshold would amount to the detection of a single gravi-
ton. However, establishing the genuinely quantum nature of gravitational radiation
requires measurements that are far more demanding than the mere detection of indi-
vidual gravitons. In QED, the photoelectric effect alone does not suffice to establish
the photon as a quantum particle, since it probes only discrete energy exchange.
Definitive evidence requires a scattering process, such as Compton scattering, that
demonstrates momentum transfer and particle, like kinematics. By analogy, propos-
als for single–graviton detection must go beyond quantized energy absorption and
instead identify processes that reveal graviton-mediated exchange or quantum gravi-
tational states. To obtain a measurement that unambiguously discriminates between
classical and quantum descriptions, two major challenges must be addressed: the
generation of genuinely nonclassical gravitational radiation, such as squeezed or en-
tangled states, and its subsequent detection with a gravitational sensor operating at
high efficiency [206].

We are now in a position to derive upper limits on the dimensionless gravitational

wave characteristic strain, h(f ), using the parameters of current cavity experiments and
proposed setups employing Qumode technology. For each configuration, the sensitivity
can be quantified through the signal-to-noise ratio (SNR), obtained by requiring that the
effective gravitationally induced transition rate satisfy

ε Γf i > ΓD . (5.76)

More precisely, for a gravitational wave signal to be detectable, the rate of real photon
production via gravitational transitions must exceed the detector’s dark count rate, which
accounts for false detection events. For resonant detectors with narrow bandwidth sensi-
tivity, a monochromatic gravitational field, or a GW background with a sharply peaked
spectrum, we find
 s   5 s 
1.4 × 10−27 1 m 1 m2
   
10 T 10 1 ΓD 1 GHz
hnoise (f ) ≈ √ .
n+1 L B0 A F ε 1 µHz f
(5.77)

For the broadband case, using the theoretical benchmark Qumode state Eq. (5.70) with
Eq. (5.71) for detectors, we have
 s s 
4.5 × 10−25 1m 1 m2
   
10T 1 GHz 1 ΓD
hnoise (f ) ≈ p . (5.78)
(n + 1) L B0 ∆f ε 1 µHz A

The performance of a given detector operating in the Qumode configuration can be directly
contrasted with that of its standard single-photon counterpart. In this case, the noise-

– 83 –
equivalent gravitational wave strains satisfy

hsingle 1
noise
= (n + 1)− 2 , (5.79)
hQu
noise

which demonstrates that the Qumode implementation achieves a parametric reduction in

the effective strain noise as the photon occupation number n increases.

CAST
OSQAR II
ALPS II

IAXO
QuGrav nb

CAST
QuGrav
Characteristic Strain h(f)

QuGrav Qu-bb
OSQAR Qu-bb

QuGrav II nb
QuALPS II
JURA
QuJURA

Qu-detector (w/ |𝑄100 〉 )

Future prospect
Qu-bb detector (w/ |𝑄100 〉 )

Frequency of GW f [Hz]

Figure 35: Strain sensitivity of photon-regeneration–based HFGW detectors. The re-

gion below the orange curve corresponds to gravitational fields that exhibit particle-like
behavior, i.e., are describable in terms of gravitons. The dashed gray curve indicates the
cosmological constraint from ∆Neff . Hatched regions show the projected sensitivity gains
of existing experiments and proposed setups employing a 100-photon Qumode (|Q100 ⟩,
see Eq. (5.68)) within the cavity. The checkered (black–blue) region represents additional
improvements achievable through near-term advances in cavity technology. Finally, the
crossed regions illustrate the enhanced broadband sensitivity attainable with the futuristic
broadband 100-photon Qumode |Q100 ⟩ introduced in Eq. (5.70). Image credit: [49].

In Fig. 35, we present the strain sensitivity of photon-regeneration–based HFGW

detectors, both in their conventional configuration and with enhancements provided by
qumode technology. Remarkably, at optical frequencies, the implementation of a 100-
photon qumode enables ALPS II to probe well below the single-graviton threshold, crossing
the wave–particle boundary. At microwave frequencies, the same approach has the poten-
tial to reach sensitivities surpassing the Big Bang Nucleosynthesis (BBN) bound, opening
a new window on cosmologically motivated high-frequency gravitational wave signals.

6 Physical Mechanisms Driving Gravitational Radiation

As discussed earlier, gravitational waves arise as tensor perturbations of the spacetime

metric, satisfying the linearized Einstein equation in the TT gauge. Their evolution is
governed by
□γij = 16πG ΠTTij , (6.1)

– 84 –
where ΠTTij denotes the transverse–traceless component of the anisotropic stress tensor.
20

In Secs. 2 and 3, we examined the propagation of gravitational waves in asymptotically flat

spacetimes, in which local astrophysical systems provide the anisotropic stresses appearing
on the right-hand side of the equation above. As shown in Eq. (2.28), the transverse–
traceless stress ΠTT
ij is directly related to the second time derivative of the mass quadrupole
moment, establishing the familiar connection between gravitational wave generation and
the non-spherical, accelerated motion of matter within such sources, schematically Q̈ij /r.
In contrast, in Sec. 4 we turned to cosmological backgrounds and focused on vacuum
fluctuations for which the source term vanishes, ΠTT ij = 0. The present section focuses on
the various physical mechanisms capable of generating a nonzero ΠTT ij and thus sourcing
gravitational wave emission in both astrophysical and cosmological settings.

6.1 Astrophysical Sources

A wide variety of astrophysical objects and dynamical phenomena can generate gravi-
tational waves, each imprinting characteristic spectral features and signal morphologies.
These sources span an enormous range of masses, environments, and timescales—from
compact binaries to supernovae, isolated neutron stars, and stochastic backgrounds. The
main classes of gravitational wave sources are naturally grouped by the duration of the sig-
nals they produce. Short-duration events include compact binary inspirals, mergers, and
burst-like transients, whereas long-duration signals arise from continuous-wave emitters
and the stochastic gravitational wave background. Together, these categories capture the
broad spectrum of astrophysical processes capable of radiating gravitational waves across
cosmic time. Fig. 36 illustrates the main categories of gravitational wave sources and their
characteristic signal durations. In the following, we provide a brief summary of each.

6.1.1 Compact binary coalescences

Compact binary coalescences (CBCs) are the most mature and well-understood astrophys-
ical sources of gravitational waves. Binary neutron stars (M ∼ 1–2M⊙ ), stellar-mass black
holes (M ∼ 10–100M⊙ ), and mixed neutron-star–black-hole systems lose orbital energy
through quadrupole emission and ultimately merge in the strong-field regime. Their signals
follow the characteristic inspiral–merger–ringdown sequence and populate the ∼ 10–103 Hz
band of ground-based interferometers. Neutron-star mergers peak at a few hundred hertz,
while stellar-mass black-hole binaries radiate at tens to hundreds of hertz depending on
total mass. The first LIGO–Virgo detections [4, 213] firmly established CBCs as the dom-
inant sources (see Fig. 30), enabling detailed population studies and stringent tests of
general relativity [214].
At lower frequencies, supermassive black-hole (SMBH) binaries represent the most
luminous gravitational wave emitters known. Formed through hierarchical galaxy merg-
20
Beyond Einstein general relativity, a wide
 class of modified-gravity
 theories can alter the propagation
2
of tensor modes as γ̈ij + (3H + δ1 ) γ̇ij + c2T ka2 + m2T + δ2 γij = 16πGeff ΠTT ij Π, where cT is the tensor
propagation speed, mT an effective graviton mass, and δ1 , δ2 encode possible modifications to friction
and dispersion arising in modified gravity (e.g., [51, 53, 207–211]). These modified-gravity frameworks,
however, lies beyond the scope of the present review and we refer the interested reader to [212].

– 85 –
Figure 36: Illustration of the main astrophysical classes of gravitational wave sources,
grouped by the duration of the signals they produce. Short-duration events (left) include
compact binary inspirals and burst-like transients, while long-duration signals (right) arise
from continuous-wave emitters and the stochastic gravitational wave background, together
representing the broad range of astrophysical processes that generate gravitational waves.

ers [215, 216], they radiate across the millihertz to nanohertz spectrum. Their mergers
fall within the 10−4 –10−1 Hz band accessible to LISA [217], providing a prime target for
space-based detectors. LISA will also observe extreme mass-ratio inspirals (EMRIs), the
gradual inspiral of a stellar-mass compact object into a 105 –107 M⊙ SMBH. EMRIs emit
in the 10−4 –1 Hz range and encode high-fidelity information about the spacetime geome-
try near massive black holes, offering precision tests of the Kerr hypothesis [218, 219]. In
addition, LISA will detect ultra-compact Galactic binaries, early inspirals of stellar-mass
black-hole binaries, and potentially signals of primordial or exotic origin [220]. The earliest
inspiral stages of massive binaries produce nanohertz gravitational waves (10−9 –10−7 Hz),
which lie in the domain of pulsar timing arrays (PTAs) [221]. These systems are expected
to dominate the low-frequency gravitational wave sky and offer insight into galaxy growth,
nuclear stellar dynamics, and black-hole coevolution [222]. As a window to fundamental
physics, gravitational wave signals from compact binary coalescences may provide an ideal
testing ground for potential deviations from classical horizon dynamics. In this context,
J. Abedi, H. Dykaar, and N. Afshordi proposed that GW echoes could serve as observa-
tional signatures of new physics near black-hole horizons. In their framework, a partially
reflective quantum structure replacing the classical horizon would generate a sequence of
delayed post-merger echoes potentially detectable by LIGO/Virgo [223–225].

6.1.2 Isolated neutron stars

Isolated neutron stars furnish an additional class of promising sources through long-lived,
nearly monochromatic continuous-wave emission [226, 227]. Such radiation, typically in
the ∼ 102 –103 Hz range and set by the stellar rotation rate and deformation, remains un-
detected but is actively sought by all-sky searches and targeted pulsar analyses. Achieving
astrophysically significant sensitivities would provide valuable constraints on neutron-star

– 86 –
microphysics, crustal rigidity, and internal magnetic-field structure. Neutron stars can
serve as natural astrophysical laboratories to search for new physics [228–233].

6.1.3 Transient burst sources

Highly dynamical and asymmetric stellar explosions also contribute to the gravitational
wave landscape. Core-collapse supernovae produce short-duration, broadband bursts through
turbulent convection, proto–neutron-star oscillations, rotation-driven instabilities, and stand-
ing accretion shock instabilities. Their gravitational wave spectra typically span ∼ 102 –103 Hz,
reflecting the characteristic dynamical timescales of proto–neutron-star cores. Although
modeling remains challenging due to the nonlinear interplay of multidimensional hydrody-
namics, neutrino transport, and magnetic fields, simulations indicate that next-generation
detectors could observe signals from supernovae in the Milky Way or nearby galaxies [234,
235]. Additional transient phenomena, such as magnetar flares and gamma-ray-burst en-
gines, may likewise produce gravitational radiation through violent magnetic or accretion-
driven processes [236].

6.1.4 Astrophysical Contributions to the Stochastic GW Background

Astrophysical sources, most notably compact-object mergers, give rise to the individually
resolved GW events detected by current observatories. Taken collectively, these same
populations also generate a stochastic gravitational wave background, produced by the
superposition of numerous weak or distant events across cosmic history, each individually
undetectable yet together forming a persistent, measurable signal. PTAs uniquely access
the nanohertz band, where the background from the cosmic population of slowly inspiraling
SMBH binaries is expected to dominate [237]. Recent observations by NANOGrav, the
EPTA, the PPTA, and the IPTA provide compelling evidence for a spatially correlated
common-spectrum process consistent with the emergence of this background [24, 238, 239]
(see Fig. 28).
At higher frequencies, space-based laser interferometers such as LISA [29] will probe
the millihertz band, where the astrophysical stochastic background is expected to be domi-
nated by compact binaries in early inspiral, including galactic white-dwarf binaries and
massive black-hole systems. At still higher frequencies, next-generation ground-based
detectors such as the Einstein Telescope [240] will access the audio band with unprece-
dented sensitivity, enabling precise measurements of both individually resolved mergers
and the astrophysical stochastic background generated by stellar-mass compact binaries
across cosmological distances. Together, PTAs, LISA, and the Einstein Telescope provide
complementary coverage of the gravitational-wave spectrum, enabling a truly multi-band
characterization of the astrophysical gravitational-wave background.

6.2 Cosmological Sources

A rich variety of processes in the early Universe can generate stochastic backgrounds of
gravitational waves, offering a unique window into physics far beyond the reach of ter-
restrial experiments. Among the most extensively studied cosmological sources are in-
flation/reheating, first-order phase transitions, topological defects, primordial black holes,

– 87 –
scalar-induced tensor modes, and magnetohydrodynamical phenomena in the radiation-era
plasma. Each mechanism imprints characteristic spectral features that encode fundamen-
tal information about high-energy symmetries, inflationary dynamics, or hidden sectors.
For comprehensive overviews, see for example the recent Einstein telescope Science Books
[32, 240], as well as the LISA Cosmology Reports [6, 29]. In the following, we discuss the
main classes of cosmological GW sources and the physics that underlies their distinctive
signatures.

6.2.1 Inflation and Reheating

In addition to the standard vacuum tensor modes from slow-roll inflation discussed in
Sec. 4.4, signals associated with spectator fields, i.e. additional scalar or gauge fields present
during inflation but not driving it, may source a stochastic gravitational wave background
with distinctive spectral features. For a recent comprehensive review of inflationary sce-
narios capable of enhancing the stochastic GWB to amplitudes observable by LISA, see
[241]. Following inflation, similar dynamics can occur during reheating. In scenarios of
gauge preheating, the coherent oscillations of the inflaton transfer energy efficiently into
Abelian or non-Abelian gauge sectors, sourcing gravitational waves through highly non-
linear, anisotropic field configurations [242–245]. The resulting GW backgrounds can be
sufficiently large that the requirement of avoiding excess radiation energy density at BBN
imposes important constraints on the viable parameter space of such models [246–248].
Inflation also sources gravitational waves indirectly through scalar fluctuations. Large
or resonantly enhanced curvature perturbations can generate second-order tensor modes
already during inflation or shortly after horizon re-entry, producing scalar-induced contri-
butions with sharply peaked spectra. The underlying mechanism was first established in
the seminal second-order calculations of [249, 250], and later applied to scenarios featuring
large small-scale scalar enhancements, such as those associated with primordial black-hole
production [251, 252]. Resonant or non-adiabatic dynamics during inflation can further
amplify scalar fluctuations and source tensor modes directly, as shown in models of par-
ticle production and inflationary resonance e.g. [253, 254]. Together, these mechanisms
highlight inflation and reheating as fertile grounds for generating non-vacuum gravita-
tional waves with characteristic signatures—sharp spectral features, non-Gaussianity, and
potential chirality—not shared by the standard slow-roll vacuum tensor background. For
comprehensive reviews of scalar-induced gravitational waves, see [5, 255, 256].

Gauge Fields in Inflation and GWs

Gauge fields are a foundational component of the SM and of nearly all well-motivated
extensions of high-energy physics. Because cosmic inflation probes energy scales far beyond
those accessible to terrestrial experiments, it is natural to view the early Universe as a
particle collider capable of revealing new interactions—an idea formalized in the framework
of the Cosmological Collider [143, 146]. This perspective motivates the question of whether
gauge fields played a dynamical role during inflation and, if so, whether they left observable
imprints in primordial correlators. Axion-like fields, protected by an approximate shift
symmetry, provide compelling inflaton candidates and couple universally to gauge sectors

– 88 –
through the CP-violating interaction ϕFµν F̃ µν . Originally introduced in the context of
the Peccei–Quinn mechanism [257–260] and later explored in cosmology [261, 262] and
inflation [263–266], this coupling naturally mediates interactions between inflaton dynamics
and gauge degrees of freedom. In addition, axions may be realized as composite states
originating from confining gauge dynamics [267, 268]. Axions and axion-like particles play
a central role in both cosmology and particle physics, and their theoretical foundations
and detection prospects have been thoroughly reviewed in [269–275]. During inflation, the
axion-gauge field interaction can lead to the exponential amplification of Abelian gauge-field
fluctuations [276, 277]. Inflationary scenarios involving gauge fields therefore provide some
of the earliest and most well-developed mechanisms for producing non-vacuum gravitational
waves. These effects offer a rich arena for probing gauge sectors through primordial non-
Gaussianity and gravitational-wave signatures [5, 278].
In 2011 it was shown in [279, 280] that non-Abelian gauge fields can contribute nontriv-
ially to the dynamics of inflation while—unlike Abelian U (1) background fields—preserving
the observed cosmological isotropy within an SU (2) subsector. In particular, the back-
ground gauge field can consistently acquire a homogeneous and isotropic vacuum expecta-
tion value of the form
(
0, µ = 0,
Āµ (t) ≡ Āaµ (t)Ta = (6.2)
a(t) ψ(t) δia Ta , µ = i,

where {Ta } are the generators of the SU (2) gauge algebra. This homogeneous and isotropic
field configuration is possible due to the isomorphism between the SU (2) gauge algebra and
the spatial rotation algebra SO(3), which allows gauge and spatial indices to be aligned
(see Fig. 37). Perturbations of the gauge fields around this isotropic background contain an
induced spin-2 mode that couples linearly to gravitational waves [279, 280]. In particular,

Figure 37: Coherent homogeneous SU (2) background gauge field and the restoration of
spatial isotropy. A homogeneous SU (2) gauge-field configuration transforms as a vector
under spatial rotations. At the same time, the SU (2) gauge symmetry contains residual
transformations with the same algebraic structure, reflecting the Lie-algebra isomorphism
su(2) ∼
= so(3). Thus, any spatial rotation can be compensated by an appropriate SU (2)
gauge transformation, corresponding to the diagonal subgroup of SO(3)rot × SU (2)gauge .
Since physical observables are gauge invariant, they depend only on this diagonal action
and therefore remain fully rotationally symmetric.

– 89 –
the spatial gauge-field perturbations include a tensorial component of the form

δAai , ∋, tij , δ aj , (6.3)

which transforms as a spin-2 degree of freedom. A rigorous mathematical proof of the

spin-2 nature of this induced tensor mode can be found in Appendix B.1 of [132]. This
mode undergoes a brief phase of tachyonic instability near horizon crossing, leading to the
formation of a chiral cloud of gauge-field tensor fluctuations localized around the cosmic
horizon. Owing to its linear coupling to the metric tensor perturbations, this induced
spin-2 mode acts as a chiral source for gravitational waves, corresponding to a nonvan-
ishing transverse–traceless anisotropic stress, ΠTT
ij ̸= 0. As a result, this coupling enables
the efficient production of chiral, blue-tilted gravitational waves [128, 130–132, 281–284],
accompanied by characteristic non-Gaussian tensor perturbations [285, 286] (see Fig. 38).
Consequently, inflation in this class of models is intrinsically parity violating (see Fig. 39),
with a pronounced asymmetry between left- and right-handed gravitational-wave modes
imprinted at the level of primordial correlations.

Figure 38: Characteristic imprints of SU(2)–axion inflation on the gravitational wave

background and CMB B-mode polarization. Image credit: [5].

This framework admits several distinct realizations within inflationary model building,
which have been explored extensively in the literature [14, 131, 280–282, 284, 287–295]. In
most cases, the isotropic configuration corresponds to an attractor solution [296–299], ex-
cept in scenarios involving Higgsed non-Abelian gauge fields with unequal masses for at
least two gauge components [300]. Variants of this setup have also been applied in the
context of warm inflation [301, 302] and have found further applications in dark-energy
model building [303–307]. Beyond their distinctive tensor signatures, these models exhibit
a rich phenomenology extending beyond standard scalar inflation, including the Schwinger
effect during inflation [308–311], as well as mechanisms for baryogenesis and leptogenesis
mediated by the chiral anomaly [293, 312, 313]. For broader context and a systematic clas-
sification of Abelian and non-Abelian mechanisms, we refer the reader to the comprehensive
review [128], together with the recent perspectives presented in [5] and [241].

– 90 –
Figure 39: Schematic illustration of slow-roll field trajectories related by parity. In the
axion–SU(2) gauge-field model, slow-roll solutions satisfy sign(ϕ̇) = sign(Aai δai ), selecting
one of the two parity-related branches.

Cosmic Chirality: Signatures of Broken Fundamental Symmetry in CMB,

LSS, and GWs
Parity (P ), defined as invariance under spatial inversion (x → −x), is one of the fun-
damental discrete symmetries of nature. While parity is conserved by gravitational,
electromagnetic, and strong interactions in the Standard Model of particle physics,
it is known to be maximally violated by weak interactions, establishing P violation
as a fundamental feature of particle physics. This fact motivates the possibility
that new physics beyond the SM, may also violate parity. An open and profound
question is therefore whether parity was ever an exact symmetry of the Universe
at earlier stages of its evolution, or whether parity-violating interactions played a
role throughout cosmic history. Cosmology provides a unique laboratory to address
this question, offering observational access to physical processes across vastly dif-
ferent epochs. Parity violation in cosmology has consequently been explored across
multiple epochs using CMB observations and LSS measurements, and more recently
through proposed searches for stochastic GWBs.
Parity violation in the CMB can manifest as cosmic birefringence, a rotation of
the linear polarization that induces nonvanishing T B and EB correlations. For
decades, this effect was thought to be unobservable because it is degenerate with
instrumental polarization angle miscalibration and is therefore removed by standard
self-calibration procedures. A recent breakthrough realization by Minami & Komatsu
[314] showed how this degeneracy can be broken using correlations with Galactic fore-
ground emission, leading to the first statistically significant hint of a nonzero cosmic
birefringence signal in Planck data. This result has revitalized efforts to probe fun-
damental parity-violating physics using the polarized sky. Subsequent studies using
polarization data from WMAP, Planck (2018 and DR4), and the Atacama Cosmol-

– 91 –
 ogy Telescope have progressively tightened constraints on cosmic birefringence, with
some reporting evidence for a nonzero uniform rotation angle while emphasizing the
critical role of instrumental systematics [314–317]. Parity violation may also orig-
inate during inflation, where chiral gravitational waves sourced by parity-violating
interactions generate intrinsically parity-odd signatures in CMB polarization, most
prominently at large angular scales in the B-mode spectrum through nonzero T B
and EB correlations [318–321]. These low-ℓ B modes provide a unique probe of
primordial parity violation, complementary to post-recombination birefringence. See
[5] for a clear and comprehensive recent review.
In addition, parity-violating physics during inflation or phase transitions can produce
circular polarization of gravitational waves, which can be detected by gravitational
wave detector networks, providing an independent and complementary probe of par-
ity violation beyond electromagnetic observations [321–326].
At late times, parity violation has further been investigated in LSS through parity-
odd components of higher-order galaxy correlation functions. Recent work has pro-
posed galaxy intrinsic alignment as an additional probe [327]. Measurements from
BOSS and SDSS find yet no statistically significant evidence for such signals [328–
330], while early results from DESI extend these tests with improved sensitivity
[331]. Together, low-ℓ CMB B modes, birefringence measurements, gravitational
wave backgrounds, and LSS correlations form a complementary framework for test-
ing parity-violating physics across cosmic history.

6.2.2 Cosmological Phase Transitions and Hydrodynamical Sources

First-order cosmological phase transitions can act as powerful factories of gravitational
waves: as bubbles of the true vacuum nucleate, expand, and collide, they generate ten-
sor perturbations through the violent dynamics of bubble walls [332, 333]. In realis-
tic transitions, however, most of the released vacuum energy is transferred to the sur-
rounding plasma, making hydrodynamical processes—the acoustic waves generated by ex-
panding bubbles and subsequent magnetohydrodynamic (MHD) turbulence—the dominant
and longest-lasting GW sources [334–336]. The resulting spectra are typically broad and
peaked, with characteristic frequencies set by the bubble size and transition temperature,
making electroweak-scale transitions a prime target for space-based interferometers such
as LISA [6] (see Fig. 40). Importantly, within the Standard Model the electroweak phase
transition is a smooth crossover rather than first order, and therefore does not generate
significant GW production. Therefore, detectable signals require extensions of the Stan-
dard Model capable of strengthening the transition. Many extensions involving additional
scalar degrees of freedom, hidden sectors, or supersymmetric frameworks predict strongly
first-order transitions capable of producing observable gravitational waves. Consequently,
cosmological phase transitions offer a unique observational window into BSM physics at
energy scales ranging from the electroweak regime up to grand-unified theories [337–341].
Primordial magnetic fields provide an additional MHD source of gravitational waves:
their anisotropic stresses, together with the MHD turbulence and Alfvén waves they induce

– 92 –
in the radiation-era plasma, generate a stochastic background whose amplitude and spectral
shape depend on the field strength, coherence scale, and helicity [342–344]. Such magnetic
fields may themselves originate from early-Universe phase transitions or other high-energy
processes, making them a natural complement to hydrodynamical GW sources. Modern
quantitative treatments combine bubble dynamics, plasma interactions, and numerical sim-
ulations to model these signals more reliably [345, 346]. Comprehensive recent reviews of
phase-transition and MHD-sourced gravitational waves, and their implications for particle
physics and cosmology, can be found in Refs. [248, 347, 348].

Temperature of the Universe

10-9 10-6 10-3 1 103 106 109 1012 1015 GeV

| | | | | | | | | λinf
GW =H
-1

H -1
| | | | | | | | | λGW =
102
λGW ≃ H -1 /106 | | | | | | | |

Figure 40: Sensitivity curves of current and future gravitational wave observatories across
a broad frequency range, shown together with representative cosmological GW sources. The
thick black curve denotes the nearly scale-invariant (vacuum) GW spectrum from inflation,
while the blue and red curves represent GW backgrounds from cosmic strings and a first-
order electroweak phase transition, respectively. The GAIA and THEIA curves are based
on [27] assuming SNR = 10 with observation times of 10 and 20 years, respectively. The
colored lines at the top of the figure indicate the present-day GW frequencies corresponding
to the production time for different source sizes λGW (colors match those of the spectra).
For illustration, the cosmic string spectrum is shown for Gµ = 10−11 (µ is the energy per
unit length of a string and G is Newton’s constant), and the first-order phase transition
signal for T∗ = 150, GeV, β/H∗ = 20, and α = 0.5. Image credit: Peera Simakachorn and
[349].

– 93 –
6.2.3 Topological Defects
Topological defects arise when a system with gauge (or global) symmetry undergoes spon-
taneous symmetry breaking during a phase transition. Let a theory have symmetry group
G, acting on scalar fields Φ, with a symmetry-breaking potential V (Φ). When the vacuum
expectation value ⟨Φ⟩ selects a ground state, the symmetry is broken to a subgroup H ⊂ G.
The degenerate vacua form the vacuum manifold M = G/H, which is a coset space whose
global topology determines the allowed defect structures. The topological defects corre-
spond to topologically non-trivial mappings Φ : S n −→ M, where S n is an n-sphere
enclosing the defect in physical space. If this map is non-contractible, the configuration
cannot be continuously deformed to the vacuum. The relevant homotopy group πn (M)
therefore classifies the defect. As shown by Tom Kibble [350], causality in an expanding
universe implies that regions separated by more than the horizon scale at a phase transition
choose vacua independently. Mismatches between these choices generically produce defects
whenever the topology of M is nontrivial [351].

Defect type Vacuum topology Homotopy Cosmology

Domain walls Disconnected vacua π0 (M) Broken discrete symmetry
Strings Loops in vacuum manifold S1 π1 (M) Cosmic strings
Monopoles Non-contractible spheres S 2 π2 (M) GUT magnetic monopoles
Textures Higher-order winding S 3 π3 (M) Global textures

Table 6: Cosmologically relevant topological defects from gauge symmetry breaking. De-
fects are classified by the homotopy groups of the vacuum manifold M = G/H, a coset
space determined by the symmetry-breaking pattern.

Topological defects can naturally arise during symmetry-breaking phase transitions

in the early Universe, taking the form of domain walls, monopoles, cosmic strings, or
textures depending on the topology of the vacuum manifold. Among these, cosmic strings
are by far the most robust and observationally relevant gravitational wave sources. They
may be produced in field-theoretic transitions or in a variety of beyond-the-Standard-
Model scenarios. Early work by Vilenkin [352] and Brandenberger, Albrecht & Turok [353],
together with later reviews of defect cosmology [354], place cosmic strings within the wider
context of early-Universe physics. Once formed, cosmic strings rapidly organize into a
long-lived scaling network that continuously produces string loops [350, 351]. These loops
oscillate relativistically under their tension and lose energy predominantly through GW
emission, as established in the foundational works of Vachaspati and Vilenkin [355] and
Allen and Shellard [356].
GW production is enhanced at relativistic features on the loop—cusps, where segments
briefly approach the speed of light, and kinks, which arise from intercommutation events.
These features generate strong bursts whose superposition over cosmic history leads to
a nearly scale-invariant stochastic background (see Fig. 40), with characteristic spectral
breaks encoding the expansion history and loop-size distribution [357, 358]. Extensions such

– 94 –
as superconducting or current-carrying cosmic strings, originally proposed by Witten [359],
further enrich the GW phenomenology by introducing current-induced energy-loss channels
and modified loop dynamics, with detailed modern treatments presented in [360]. Several
recent comprehensive reviews specifically address the gravitational wave phenomenology
of cosmic-string networks [9, 94, 361–363]. Other topological defects can also source GWs.
Global textures generate smooth, nearly scale-invariant tensor spectra, while monopole–
string composite systems radiate during annihilation or collapse [364–367]. Domain walls,
when present, undergo dynamical evolution that can drive powerful defect-induced phase
transitions and corresponding GW signals [368]. Although typically more model-dependent
than cosmic strings, such defects provide complementary probes of high-energy symmetry
breaking.

Figure 41: Comparison of detector sensitivities with GW spectra associated with primor-
dial black hole PBH scenarios and scalar induced tensor modes. The thick black curve
on the left shows the second-order curvature-induced GWs generated during PBH forma-
tion, while the thick black curve on the right shows the stochastic GW background from
PBH binaries. Dashed black curves correspond to predicted backgrounds for different PBH
abundance fractions. Image credit: Adapted from [369].

6.2.4 Primordial Black Holes and scalar-induced gravitational waves

Primordial black holes (PBHs) are black holes formed in the early Universe, typically from
the collapse of large overdensities shortly after horizon re-entry, rather than from stellar
evolution [370]. Their formation is most naturally described in terms of enhanced pri-
mordial curvature perturbations, which collapse if they exceed a critical threshold. The
associated GW signatures arise through two main channels. First, PBHs can form bi-
nary systems whose inspirals and mergers generate a population of resolvable GW events
and a stochastic background, potentially contributing to the binary black-hole signals ob-
served by ground-based interferometers [371]. Second, the large scalar perturbations re-

– 95 –
quired for PBH formation inevitably source tensor perturbations at second order, leading
to scalar-induced GWs with peaked spectra that can fall in the PTA, LISA or ground-based
frequency bands [251]. This connects PBH dark-matter scenarios directly with GW obser-
vations. Scalar-induced gravitational waves (SIGWs) arise at second order in cosmological
perturbation theory: large primordial curvature fluctuations, though not sourcing tensor
modes at linear order, inevitably generate gravitational waves as they re-enter the horizon
in the radiation era. This mechanism was first established in [249, 250] and later applied
to scenarios with enhanced small-scale perturbations, such as those leading to primordial
black-hole formation [251]. For the gravitational wave features in relation to the sensitivi-
ties of current probes, see Fig. 41. The resulting GW spectra are typically sharply peaked,
with frequencies determined by the horizon scale at re-entry, making SIGWs a powerful
probe of inflationary physics on scales far smaller than those constrained by the CMB. For
comprehensive discussions of PBHs and their GW signatures, including end-to-end treat-
ments of inflationary production, binary formation, and induced backgrounds, see recent
reviews [255, 372, 373].

7 Summary

In this review, we present a pedagogical study of gravitational waves, integrating theory,

phenomenology, and observation. Gravitational waves offer a powerful probe of high-
energy BSM physics and a unique window into the unexplored frontiers of gravity and the
early Universe. Unlike photons, which could only free-stream after the formation of the
cosmic microwave background (T = 1 eV), and neutrinos, which decoupled at T = 1 MeV,
gravitational waves have always propagated freely across cosmic history. As a result, once
detected, they carry valuable and unique information about the Universe at energy scales
far beyond those accessible by terrestrial particle colliders (see Fig. 15).
Spanning a gravitational wave rainbow with wavelengths ranging from the size of the
observable Universe down to kilometer scales (10−17 -104 Hz), these signals enable a truly
multi-scale exploration of fundamental physics, providing insights impossible to obtain from
any other messenger. Gravitational waves provide a unique probe of fundamental physics,
connecting gravity and high-energy phenomena across multiple scales. We develop a coher-
ent framework linking the mathematical foundations of gravitational radiation with the sig-
nals accessible to current and future detectors, aiming to bridge the gap between theoretical
predictions and experimental measurements. Since gravitational waves are both generated
and propagate through the Universe’s gravitational backgrounds, their description depends
on the relevant spacetime geometry. On small scales (up to 100 Mpc), the expansion of the
Universe is negligible, and asymptotically flat spacetimes accurately describe gravitational
dynamics. On larger, cosmological scales, expansion becomes significant, and the FLRW
geometry must be used. This distinction also reflects the nature of gravitational wave
sources: deterministic astrophysical events produce coherent signals, whereas cosmological
processes generate stochastic backgrounds, requiring a statistical treatment (see Fig. 3).
The review begins with the linearized theory of general relativity, deriving gravitational
waves as solutions to the Einstein equations and analyzing their physical properties. Next,

– 96 –
we provide a more mathematically rigorous definition of gravitational waves by examining
their geometrical nature through the Weyl tensor, including its principal null directions
and Petrov classification. Within the Bondi–Sachs framework, this approach clarifies how
gravitational waves propagate as true distortions of spacetime, demonstrating that they are
genuine carriers of energy and angular momentum. By combining algebraic and geometri-
cal perspectives, this framework establishes a solid theoretical foundation for interpreting
observational data and connecting theory with experiment.
We then explore cosmological gravitational waves, tracing their evolution across dif-
ferent cosmic epochs and their generation during inflation, phase transitions, and other
early-Universe processes. Building on this foundation, we survey the principal detection
strategies across a broad range of frequencies, including cosmic microwave background
measurements, pulsar timing arrays, ground- and space-based laser interferometers, and
resonant cavity experiments. We also highlight several ongoing developments and proposed
concepts that, in the near future, are expected to substantially enhance both the sensitiv-
ity and frequency reach of gravitational-wave observations. We then examine the physical
mechanisms responsible for generating gravitational radiation, clearly distinguishing be-
tween astrophysical sources and cosmological origins.
The study of gravitational waves is a vibrant and rapidly advancing frontier, marked by
open theoretical, numerical, and observational challenges and by rapid progress on multiple
fronts. By integrating theory, phenomenology, and observation, this review emphasizes the
extraordinary power of gravitational waves to probe the Universe across all scales, grant-
ing access to energy regimes and physical processes far beyond the reach of conventional
experiments. We hope this work will guide and inspire both newcomers and experienced
researchers as gravitational-wave physics continues to open a transformative window onto
the cosmos.

Acknowledgments

First, I would like to express my gratitude to the Astronomy and Astrophysics Review for
the invitation to write this article. I owe particular thanks to Eiichiro Komatsu, whose
encouragement and intellectual guidance not only sparked my enduring interest in the
subject, but also greatly deepened my understanding of it. I am also grateful to Robert
Brandenberger, Sadra Jazayeri, Eiichiro Komatsu, and Geraldine Servant for their valu-
able feedback and thoughtful suggestions, which significantly improved the clarity of this
work. I thank Peera Simakachorn for helpful discussion and also providing the plot shown
in Fig. 40. I would further like to thank Dionysios Anninos, Daniel Baumann, Dmitri
Kharzeev, Joachim Kopp, Nicholas Rodd, Andrew Strominger, and Alexander Zhiboedov,
whose insights and discussions over the years have profoundly shaped my perspective on
many aspects of this broad and evolving field. I am grateful to the Max Planck Institutes
for Physics and for Astrophysics in Garching, and to Johannes Henn and Eiichiro Komatsu,
for their kind hospitality during the initial stage of this work, as well as to DESY and Geral-
dine Servant for their warm hospitality during its final stage; both institutes provided an
excellent environment for the completion of this manuscript. This work was supported by

– 97 –
the Royal Society through a University Research Fellowship (Grant No. RE22432) and
by the Deutsche Forschungsgemeinschaft under Germany’s Excellence Strategy (EXC 2121
Quantum Universe – 390833306).

References

[1] A. Einstein, Approximative Integration of the Field Equations of Gravitation, Sitzungsber.

Preuss. Akad. Wiss. Berlin (Math. Phys. ) 1916 (1916) 688.
[2] H. Bondi, Plane gravitational waves in general relativity, Nature 179 (1957) 1072.
[3] R.A. Hulse and J.H. Taylor, Discovery of a pulsar in a binary system, Astrophysical Journal
Letters 195 (1975) L51.
[4] LIGO Scientific and Virgo collaboration, Observation of gravitational waves from a
binary black hole merger, Phys. Rev. Lett. 116 (2016) 061102 [1602.03837].
[5] E. Komatsu, New physics from the polarized light of the cosmic microwave background,
Nature Rev. Phys. 4 (2022) 452 [2202.13919].
[6] LISA Cosmology Working Group collaboration, Cosmology with the Laser
Interferometer Space Antenna, Living Rev. Rel. 26 (2023) 5 [2204.05434].
[7] C. Caprini and D.G. Figueroa, Cosmological backgrounds of gravitational waves, Class.
Quant. Grav. 35 (2018) 163001 [1801.04268].
[8] M. Hindmarsh, S.J. Huber, K. Rummukainen and D.J. Weir, Gravitational waves from the
sound of a first order phase transition, Phys. Rev. Lett. 112 (2014) 041301 [1304.2433].
[9] A. Dimitriou, D.G. Figueroa, P. Simakachorn and B. Zaldivar, Cosmic string gravitational
wave backgrounds at LISA: I. Signal survey, template reconstruction, and model
comparison, 2508.05395.
[10] S.H.-S. Alexander, M.E. Peskin and M.M. Sheikh-Jabbari, Leptogenesis from gravity waves
in models of inflation, Phys. Rev. Lett. 96 (2006) 081301 [hep-th/0403069].
[11] A. Maleknejad, Chiral Gravity Waves and Leptogenesis in Inflationary Models with
non-Abelian Gauge Fields, Phys. Rev. D 90 (2014) 023542 [1401.7628].
[12] M. Garny, M. Sandora and M.S. Sloth, Planckian Interacting Massive Particles as Dark
Matter, Phys. Rev. Lett. 116 (2016) 101302 [1511.03278].
[13] A. Maleknejad, Gravitational leptogenesis in axion inflation with SU(2) gauge field, JCAP
1612 (2016) 027 [1604.06520].
[14] R.R. Caldwell and C. Devulder, Axion Gauge Field Inflation and Gravitational
Leptogenesis: A Lower Bound on B Modes from the Matter-Antimatter Asymmetry of the
Universe, Phys. Rev. D97 (2018) 023532 [1706.03765].
[15] S. Alexander, E. McDonough and D.N. Spergel, Chiral Gravitational Waves and Baryon
Superfluid Dark Matter, JCAP 05 (2018) 003 [1801.07255].
[16] N. Bernal, M. Dutra, Y. Mambrini, K. Olive, M. Peloso and M. Pierre, Spin-2 Portal Dark
Matter, Phys. Rev. D 97 (2018) 115020 [1803.01866].
[17] S. Clery, Y. Mambrini, K.A. Olive and S. Verner, Gravitational portals in the early
Universe, Phys. Rev. D 105 (2022) 075005 [2112.15214].

– 98 –
[18] A. Maleknejad and J. Kopp, Weyl fermion creation by cosmological gravitational wave
background at 1-loop, JHEP 01 (2025) 023 [2406.01534].
[19] A. Maleknejad and J. Kopp, Gravitational Wave-Induced Freeze-In of Fermionic Dark
Matter, 2405.09723.
[20] R. Garani, M. Redi and A. Tesi, Particle production from inhomogeneities: general metric
perturbations, JHEP 08 (2025) 037 [2502.12249].
[21] A. Maleknejad, Gravitational ABJ Anomaly, Stochastic Matter Production, and
Leptogenesis, 2412.09490.
[22] P. Campeti and others (LiteBIRD Collaboration), LiteBIRD Science Goals and Forecasts:
A Case Study of the Origin of Primordial Gravitational Waves using Large-Scale CMB
Polarization, Journal of Cosmology and Astroparticle Physics (2024) [2312.00717].
[23] CMB-S4 Collaboration, CMB-S4 Science Book: Probing Fundamental Physics with
Next-Generation CMB Polarization Measurements, Proceedings of the National Academy of
Sciences (2019) [20190032159].
[24] NANOGrav Collaboration, The NANOGrav 15-year Data Set: Evidence for a
Gravitational-Wave Background, Astrophys. J. Lett. 951 (2023) L8 [2306.16213].
[25] J. Antoniadis and others (EPTA Collaboration), The second data release from the European
Pulsar Timing Array III. Search for gravitational wave signals, Astron. Astrophys. 678
(2023) A50 [2306.16214].
[26] D.J. Reardon and others (PPTA Collaboration), Search for an isotropic gravitational-wave
background with the Parkes Pulsar Timing Array, Astrophys. J. Lett. 951 (2023) L6
[2306.16215].
[27] M. Çalışkan, Y. Chen, L. Dai, N. Anil Kumar, I. Stomberg and X. Xue, Dissecting the
stochastic gravitational wave background with astrometry, JCAP 05 (2024) 030
[2312.03069].
[28] J. Garcia-Bellido, H. Murayama and G. White, Exploring the early Universe with Gaia and
Theia, JCAP 12 (2021) 023 [2104.04778].
[29] M. Colpi and et al., LISA Definition Study Report (Red Book), 2402.07571.
[30] T. Akutsu and others (KAGRA Collaboration), KAGRA: 2.5 generation interferometric
gravitational wave detector, Nature Astron. 3 (2019) 35 [1811.08079].
[31] S. Priyadarshini, India to begin construction of gravitational wave project, Nature 632
(2025) 85.
[32] C.F. Sintes et al., The Science of the Einstein Telescope, 2503.12263.
[33] P.W. Graham, J.M. Hogan, M.A. Kasevich and S. Rajendran, New method for gravitational
wave detection with atomic sensors, Physical Review Letters 110 (2013) 171102.
[34] S. Dimopoulos, P.W. Graham, J.M. Hogan, M.A. Kasevich and S. Rajendran, Atomic
gravitational wave interferometric sensor, Physical Review D 78 (2008) 122002.
[35] J.M. Hogan, D.M.S. Johnson, S. Dickerson, T. Kovachy, A. Sugarbaker, S.-w. Chiow et al.,
An atomic gravitational wave interferometric sensor in low earth orbit (agis-leo), General
Relativity and Gravitation 43 (2011) 1953.
[36] B. Canuel et al., Exploring gravity with the miga large scale atom interferometer, Nature
Communications 8 (2017) 1572.

– 99 –
[37] M. Zhan et al., ZAIGA: Zhaoshan long-baseline atom interferometer gravitation antenna,
International Journal of Modern Physics D 29 (2020) 1940005.
[38] P.W. Graham et al., Mid-band gravitational wave detection with precision atomic sensors,
arXiv e-prints (2017) [1711.02225].
[39] B. Canuel et al., ELGAR: A european laboratory for gravitation and atom-interferometric
research, Classical and Quantum Gravity 37 (2020) 225017.
[40] L. Badurina et al., AION: An atom interferometer observatory and network, Journal of
Cosmology and Astroparticle Physics 05 (2020) 011.
[41] Y.A. El-Neaj et al., AEDGE: Atomic experiment for dark matter and gravity exploration,
EPJ Quantum Technology 7 (2020) 6.
[42] S. Schiller et al., The space-time explorer and quantum test of the equivalence principle
(ste-quest), arXiv e-prints (2015) [1501.03989].
[43] F. Dünkel et al., The maius-1 mission: Space atom interferometry on a sounding rocket, npj
Microgravity 4 (2018) 16.
[44] I. Alonso et al., Cold atoms in space: community workshop summary and proposed
road-map, EPJ Quant. Technol. 9 (2022) 30 [2201.07789].
[45] S. Abend et al., Terrestrial Very-Long-Baseline Atom Interferometry: Workshop Summary,
AVS Quantum Sci. 6 (2024) 024701 [2310.08183].
[46] A. Abdalla et al., Terrestrial Very-Long-Baseline Atom Interferometry: summary of the
second workshop, EPJ Quant. Technol. 12 (2025) 42 [2412.14960].
[47] N. Aggarwal et al., Challenges and Opportunities of Gravitational Wave Searches at
Frequencies Above the LIGO/Virgo Band, 2501.11723.
[48] G. Tobar, S.K. Manikandan, T. Beitel and I. Pikovski, Detecting single gravitons with
quantum sensing, Nature Commun. 15 (2024) 7229 [2308.15440].
[49] D.E. Kharzeev, A. Maleknejad and S. Shalamberidze, QuGrav: Bringing gravitational
waves to light with Qumodes, 2506.09459.
[50] LiteBIRD collaboration, Probing Cosmic Inflation with the LiteBIRD Cosmic Microwave
Background Polarization Survey, PTEP 2023 (2023) 042F01 [2202.02773].
[51] C. de Rham, Massive Gravity, Living Rev. Rel. 17 (2014) 7 [1401.4173].
[52] C.M. Will, The Confrontation between General Relativity and Experiment, Living Rev. Rel.
17 (2014) 4 [1403.7377].
[53] J.M. Ezquiaga and M. Zumalacárregui, Dark Energy After GW170817: Dead Ends and the
Road Ahead, Phys. Rev. Lett. 119 (2017) 251304 [1710.05901].
[54] N.T. Bishop and L. Rezzolla, Extraction of Gravitational Waves in Numerical Relativity,
Living Rev. Rel. 19 (2016) 2 [1606.02532].
[55] Y.B. Zel’dovich and A.G. Polnarev, Radiation of gravitational waves by a cluster of
superdense stars, Sov. Astron. 18 (1974) 17.
[56] V.B. Braginsky and L.P. Grishchuk, Kinematic Resonance and Memory Effect in Free Mass
Gravitational Antennas, Sov. Phys. JETP 62 (1985) 427.
[57] D. Christodoulou, Nonlinear nature of gravitation and gravitational-wave experiments,
Phys. Rev. Lett. 67 (1991) 1486–1489.

– 100 –
[58] M.G.J.v.d.B. H. Bondi and A.W.K. Metzner, Gravitational waves in general relativity. vii.
waves from axi-symmetric isolated systems, Proc. Roy. Soc. Lond. A 269 (1962) 21.
[59] R.K. Sachs, Gravitational waves in general relativity. viii. waves in asymptotically flat
space-times, Proc. Roy. Soc. Lond. A 270 (1962) 103.
[60] P.D. Lasky, E. Thrane, Y. Levin, J. Blackman and Y. Chen, Detecting gravitational-wave
memory with LIGO: implications of GW150914, Phys. Rev. Lett. 117 (2016) 061102
[1605.01415].
[61] H. Inchauspé, S. Gasparotto, D. Blas, L. Heisenberg, J. Zosso and S. Tiwari, Measuring
gravitational wave memory with LISA, Phys. Rev. D 111 (2025) 044044 [2406.09228].
[62] J.B. Wang et al., Searching for gravitational wave memory bursts with the Parkes Pulsar
Timing Array, Mon. Not. Roy. Astron. Soc. 446 (2015) 1657 [1410.3323].
[63] F.A.E. Pirani, Invariant formulation of gravitational radiation theory, Phys. Rev. 105
(1957) 1089.
[64] E.T. Newman and R. Penrose, An approach to gravitational radiation by a method of spin
coefficients, Journal of Mathematical Physics 3 (1962) 566.
[65] A.Z. Petrov, Klassifikatsiya prostranstv, opredelyayushchikh polya tyagotenia, Uchenye
Zapiski Kazanskogo Gosudarstvennogo Universiteta im. V. I. Ulyanova-Lenina 114 (1954)
55.
[66] R.M. Wald, General Relativity, Chicago Univ. Pr., Chicago, USA (1984),
10.7208/chicago/9780226870373.001.0001.
[67] J.N. Goldberg and R.K. Sachs, A theorem on Petrov types, Acta Physica Polonica 22
(1962) 13.
[68] H. Stephani, D. Kramer, M.A.H. MacCallum, C. Hoenselaers and E. Herlt, Exact Solutions
of Einstein’s Field Equations, Cambridge University Press, 2 ed. (2003).
[69] R.K. Sachs, Gravitational waves in general relativity. vi. the outgoing radiation condition,
Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences
264 (1961) 309.
[70] H. Bondi, F.A.E. Pirani and I. Robinson, Gravitational waves in general relativity. 3. Exact
plane waves, Proc. Roy. Soc. Lond. A251 (1959) 519.
[71] T. Mädler and J. Winicour, Bondi-Sachs Formalism, Scholarpedia 11 (2016) 33528
[1609.01731].
[72] J. Winicour, Characteristic Evolution and Matching, Living Rev. Rel. 12 (2009) 3
[0810.1903].
[73] A. Strominger, Lectures on the Infrared Structure of Gravity and Gauge Theory,
1703.05448.
[74] J. Lense and H. Thirring, Ueber den Einfluss der Eigenrotation der Zentralkoerper auf die
Bewegung der Planeten und Monde nach der Einsteinschen Gravitationstheorie, Phys. Z.
19 (1918) 156.
[75] R.M. Wald, Asymptotic behavior of homogeneous cosmological models in the presence of a
positive cosmological constant, Phys. Rev. D28 (1983) 2118.
[76] L. Bieri and P.T. Chruściel, Future-complete null hypersurfaces, interior gluings, and the
Trautman-Bondi mass, 1612.04359.

– 101 –
[77] V.-L. Saw, Mass-loss of an isolated gravitating system due to energy carried away by
gravitational waves with a cosmological constant, Phys. Rev. D 94 (2016) 104004
[1605.05151].
[78] R. Sachs, Asymptotic symmetries in gravitational theory, Phys. Rev. 128 (1962) 2851.
[79] A. Strominger and A. Zhiboedov, Gravitational Memory, BMS Supertranslations and Soft
Theorems, JHEP 01 (2016) 086 [1411.5745].
[80] T. Mädler and J. Winicour, Bondi-Sachs Formalism, Scholarpedia 11 (2016) 33528
[1609.01731].
[81] S. Weinberg, Infrared photons and gravitons, Phys. Rev. 140 (1965) B516.
[82] A. Strominger, On BMS Invariance of Gravitational Scattering, JHEP 07 (2014) 152
[1312.2229].
[83] K.S. Thorne, Gravitational-wave bursts with memory: The Christodoulou effect, Phys. Rev.
D 45 (1992) 520.
[84] A. Strominger and A. Zhiboedov, Gravitational memory, bms supertranslations and soft
theorems, JHEP (2016) [1411.5745].
[85] L. Bieri and D. Garfinkle, An electromagnetic analogue of gravitational wave memory,
Classical and Quantum Gravity 30 (2013) 195009 [1307.5098].
[86] A. Strominger, Asymptotic Symmetries of Yang-Mills Theory, JHEP 07 (2014) 151
[1308.0589].
[87] T. He, P. Mitra, A.P. Porfyriadis and A. Strominger, New Symmetries of Massless QED,
JHEP 10 (2014) 112 [1407.3789].
[88] D. Kapec, M. Pate and A. Strominger, New Symmetries of QED, Adv. Theor. Math. Phys.
21 (2017) 1769 [1506.02906].
[89] A. Maleknejad, Chiral photon memory effect and parity violation in electromagnetic
radiation, Journal of High Energy Physics 07 (2020) 154 [2003.05396].
[90] Planck collaboration, Planck 2018 results. VI. Cosmological parameters, Astron.
Astrophys. 641 (2020) A6 [1807.06209].
[91] B. Spokoiny, Deflationary universe scenario, Phys. Lett. B 315 (1993) 40 [gr-qc/9306008].
[92] M. Joyce, Electroweak baryogenesis and the expansion rate of the universe, Phys. Rev. D 55
(1997) 1875 [hep-ph/9606223].
[93] P.G. Ferreira and M. Joyce, Cosmology with a primordial scaling field, Phys. Rev. D 58
(1998) 023503 [astro-ph/9711102].
[94] Y. Gouttenoire, G. Servant and P. Simakachorn, Kination cosmology from scalar fields and
gravitational-wave signatures, 2111.01150.
[95] S. Weinberg, Adiabatic modes in cosmology, Phys. Rev. D67 (2003) 123504
[astro-ph/0302326].
[96] A.A. Starobinsky, A new type of isotropic cosmological models without singularity, Physics
Letters B 91 (1980) 99.
[97] K. Sato, First-order phase transition of a vacuum and the expansion of the universe,
Monthly Notices of the Royal Astronomical Society 195 (1981) 467.

– 102 –
 [98] A.H. Guth, Inflationary universe: A possible solution to the horizon and flatness problems,
Physical Review D 23 (1981) 347.
[99] A.D. Linde, A new inflationary universe scenario: A possible solution of the horizon,
flatness, homogeneity, isotropy and primordial monopole problems, Physics Letters B 108
(1982) 389.
[100] G.W. Gibbons and S.W. Hawking, Cosmological Event Horizons, Thermodynamics, and
Particle Creation, Phys. Rev. D 15 (1977) 2738.
[101] S.W. Hawking and I.G. Moss, Supercooled Phase Transitions in the Very Early Universe,
Phys. Lett. B 110 (1982) 35.
[102] A. Maleknejad and M.M. Sheikh-Jabbari, Revisiting Cosmic No-Hair Theorem for
Inflationary Settings, Phys. Rev. D85 (2012) 123508 [1203.0219].
[103] M.-a. Watanabe, S. Kanno and J. Soda, Inflationary Universe with Anisotropic Hair, Phys.
Rev. Lett. 102 (2009) 191302 [0902.2833].
[104] W.E. East, M. Kleban, A. Linde and L. Senatore, Beginning inflation in an inhomogeneous
universe, JCAP 09 (2016) 010 [1511.05143].
[105] V.F. Mukhanov and G.V. Chibisov, Quantum fluctuations and a nonsingular universe,
JETP Lett. 33 (1981) 532.
[106] S.W. Hawking, The development of irregularities in a single bubble inflationary universe,
Phys. Lett. B 115 (1982) 295.
[107] A.A. Starobinsky, Dynamics of phase transition in the new inflationary universe scenario
and generation of perturbations, Phys. Lett. B 117 (1982) 175.
[108] A.H. Guth and S.-Y. Pi, Fluctuations in the new inflationary universe, Phys. Rev. Lett. 49
(1982) 1110.
[109] J.M. Bardeen, P.J. Steinhardt and M.S. Turner, Spontaneous creation of almost scale-free
density perturbations in an inflationary universe, Phys. Rev. D 28 (1983) 679.
[110] G.F. Smoot, C.L. Bennett, A. Kogut, E.L. Wright, J. Aymon, N.W. Boggess et al.,
Structure in the COBE differential microwave radiometer first-year maps, Astrophys. J.
Lett. 396 (1992) L1.
[111] G. Hinshaw, D. Larson, E. Komatsu, D.N. Spergel, C.L. Bennett, J. Dunkley et al.,
Nine-year wilkinson microwave anisotropy probe (WMAP) observations: Cosmological
parameter results, Astrophys. J. Suppl. 208 (2013) 19 [1212.5226].
[112] R. Kallosh, A. Linde and D. Roest, ACT, SPT, and chaotic inflation, arXiv e-print (2025)
[2503.21030].
[113] P.A.R.e.a.B.C. Ade, Improved constraints on primordial gravitational waves using Planck,
WMAP, and BICEP/Keck observations through the 2018 observing season, Physical Review
Letters 127 (2021) 151301.
[114] H.e.a.B.A.C. Hui, BICEP Array: a multi-frequency degree-scale CMB polarimeter, arXiv
e-print (2018) [1808.00568].
[115] Simons Observatory collaboration, The simons observatory: Science goals and forecasts,
JCAP 02 (2019) 056 [1808.07445].
[116] A.A. Starobinsky, Spectrum of relict gravitational radiation and the early state of the
universe, JETP Lett. 30 (1979) 682.

– 103 –
[117] E.W. Kolb and M.S. Turner, The Early Universe, vol. 69 of Frontiers in Physics,
Addison-Wesley, Redwood City, USA (1990).
[118] S. Dodelson and F. Schmidt, Modern Cosmology, Academic Press, Cambridge, MA, 2 ed.
(2020).
[119] R.H. Brandenberger, Introduction to Early Universe Cosmology, PoS ICFI2010 (2010) 001
[1103.2271].
[120] J.M. Cline, TASI Lectures on Early Universe Cosmology: Inflation, Baryogenesis and Dark
Matter, PoS TASI2018 (2019) 001 [1807.08749].
[121] V.F. Mukhanov, Physical Foundations of Cosmology, Cambridge University Press,
Cambridge, UK (2005), 10.1017/CBO9780511790553.
[122] S. Weinberg, Cosmology (2008).
[123] F.W.J. Olver, A.B.O. Daalhuis, D.W. Lozier, B.I. Schneider, R.F. Boisvert, C.W. Clark
et al., eds., NIST Digital Library of Mathematical Functions, National Institute of
Standards and Technology, Gaithersburg, Maryland (2023).
[124] BICEP, Keck collaboration, Improved Constraints on Primordial Gravitational Waves
using Planck, WMAP, and BICEP/Keck Observations through the 2018 Observing Season,
Phys. Rev. Lett. 127 (2021) 151301 [2110.00483].
[125] J.M. Bardeen, Gauge Invariant Cosmological Perturbations, Phys. Rev. D 22 (1980) 1882.
[126] R.H. Brandenberger and R. Kahn, COSMOLOGICAL PERTURBATIONS IN
INFLATIONARY UNIVERSE MODELS, Phys. Rev. D 29 (1984) 2172.
[127] V.F. Mukhanov, L.R.W. Abramo and R.H. Brandenberger, On the Back reaction problem
for gravitational perturbations, Phys. Rev. Lett. 78 (1997) 1624 [gr-qc/9609026].
[128] A. Maleknejad, M.M. Sheikh-Jabbari and J. Soda, Gauge Fields and Inflation, Phys. Rept.
528 (2013) 161 [1212.2921].
[129] P. Adshead, E. Martinec and M. Wyman, Gauge fields and inflation: Chiral gravitational
waves, fluctuations, and the Lyth bound, Phys. Rev. D88 (2013) 021302 [1301.2598].
[130] E. Dimastrogiovanni and M. Peloso, Stability analysis of chromo-natural inflation and
possible evasion of Lyth’s bound, Phys. Rev. D87 (2013) 103501 [1212.5184].
[131] A. Maleknejad, Axion Inflation with an SU(2) Gauge Field: Detectable Chiral Gravity
Waves, JHEP 07 (2016) 104 [1604.03327].
[132] A. Maleknejad and E. Komatsu, Production and Backreaction of Spin-2 Particles of SU (2)
Gauge Field during Inflation, JHEP 05 (2019) 174 [1808.09076].
[133] Planck collaboration, Planck 2018 results. X. Constraints on inflation, 1807.06211.
[134] E. Komatsu et al., Non-Gaussianity as a Probe of the Physics of the Primordial Universe
and the Astrophysics of the Low Redshift Universe, 0902.4759.
[135] E. Komatsu and D.N. Spergel, Acoustic signatures in the primary microwave background
bispectrum, Phys. Rev. D 63 (2001) 063002 [astro-ph/0005036].
[136] E. Komatsu, Hunting for Primordial Non-Gaussianity in the Cosmic Microwave
Background, Class. Quant. Grav. 27 (2010) 124010 [1003.6097].
[137] E. Pajer, F. Schmidt and M. Zaldarriaga, The observed squeezed limit of cosmological
three-point functions, Phys. Rev. D 88 (2013) 083502 [1305.0824].

– 104 –
[138] P. Creminelli, J. Noreña and M. Simonović, Conformal consistency relations for single-field
inflation, JCAP 1207 (2012) 052 [1203.4595].
[139] K. Hinterbichler, L. Hui and J. Khoury, An Infinite Set of Ward Identities for Adiabatic
Modes in Cosmology, JCAP 1401 (2014) 039 [1304.5527].
[140] E. Schrodinger, The proper vibrations of the expanding universe, Physica 6 (1939) 899.
[141] L. Parker, Particle creation in expanding universes, Phys. Rev. Lett. 21 (1968) 562.
[142] E.W. Kolb and A.J. Long, Cosmological gravitational particle production and its
implications for cosmological relics, Rev. Mod. Phys. 96 (2024) 045005 [2312.09042].
[143] X. Chen and Y. Wang, Quasi-Single Field Inflation and Non-Gaussianities, JHEP 08
(2010) 044 [0911.3380].
[144] X. Chen, Primordial Non-Gaussianities from Inflation Models, Adv. Astron. 2010 (2010)
157080 [1002.1416].
[145] T. Noumi, M. Yamaguchi and D. Yokoyama, Effective field theory approach to quasi-single
field inflation and effects of heavy fields, JHEP 06 (2013) 051 [1211.1624].
[146] N. Arkani-Hamed and J. Maldacena, Cosmological Collider Physics, 1503.08043.
[147] X. Chen, M.H. Namjoo and Y. Wang, Quantum Primordial Standard Clocks, JCAP 02
(2016) 013 [1509.03930].
[148] H. Lee, D. Baumann and G.L. Pimentel, Non-Gaussianity as a Particle Detector, JHEP 12
(2016) 040 [1607.03735].
[149] N. Arkani-Hamed, D. Baumann, H. Lee and G.L. Pimentel, The Cosmological Bootstrap:
Inflationary Correlators from Symmetries and Singularities, JHEP 04 (2020) 105
[1811.00024].
[150] S. Jazayeri and S. Renaux-Petel, Cosmological bootstrap in slow motion, JHEP 12 (2022)
137 [2205.10340].
[151] A. Bodas, S. Kumar and R. Sundrum, The Scalar Chemical Potential in Cosmological
Collider Physics, JHEP 02 (2021) 079 [2010.04727].
[152] S. Jazayeri, S. Renaux-Petel, X. Tong, D. Werth and Y. Zhu, Parity violation from
emergent nonlocality during inflation, Phys. Rev. D 108 (2023) 123523 [2308.11315].
[153] C. de Rham, S. Jazayeri and A.J. Tolley, Bispectrum islands: Bootstrap bounds on
cosmological correlators, Phys. Rev. D 112 (2025) 083531 [2506.19198].
[154] W. Sohn, D.-G. Wang, J.R. Fergusson and E.P.S. Shellard, Searching for cosmological
collider in the Planck CMB data, JCAP 09 (2024) 016 [2404.07203].
[155] G. Cabass, M.M. Ivanov, O.H.E. Philcox, M. Simonović and M. Zaldarriaga, Constraints on
Single-Field Inflation from the BOSS Galaxy Survey, Phys. Rev. Lett. 129 (2022) 021301
[2201.07238].
[156] D. Baumann, D. Green, A. Joyce, E. Pajer, G.L. Pimentel, C. Sleight et al., Snowmass
White Paper: The Cosmological Bootstrap, SciPost Phys. Comm. Rep. 2024 (2024) 1
[2203.08121].
[157] M. Maggiore, Gravitational Waves. Vol. 2: Astrophysics and Cosmology, Oxford University
Press (2018).

– 105 –
[158] Y. Watanabe and E. Komatsu, Improved Calculation of the Primordial Gravitational Wave
Spectrum in the Standard Model, Phys. Rev. D 73 (2006) 123515 [astro-ph/0604176].
[159] K. Saikawa and S. Shirai, Primordial gravitational waves, precisely: The role of
thermodynamics in the Standard Model, JCAP 05 (2018) 035 [1803.01038].
[160] R.K. Sachs and A.M. Wolfe, Perturbations of a cosmological model and angular variations
of the microwave background, Astrophys. J. 147 (1967) 73.
[161] J. Lesgourgues, The cosmic linear anisotropy solving system (CLASS) III: Comparison with
CAMB for ΛCDM, JCAP 09 (2011) 032 [1104.2934].
[162] T. Kite, A. Ravenni, S.P. Patil and J. Chluba, Bridging the gap: spectral distortions meet
gravitational waves, Mon. Not. Roy. Astron. Soc. 505 (2021) 4396 [2010.00040].
[163] R.S. Foster and D.C. Backer, Constructing a pulsar timing array, Astrophys. J. 361 (1990)
300.
[164] R.N. Manchester et al., The parkes pulsar timing array project, Publ. Astron. Soc. Australia
30 (2013) e017 [1309.7392].
[165] J.P.W. Verbiest, S.J. Vigeland, N.K. Porayko, S. Chen and D.J. Reardon, Status report on
global pulsar-timing-array efforts to detect gravitational waves, Results Phys. 61 (2024)
107719 [2404.19529].
[166] EPTA collaboration, The second data release from the European Pulsar Timing Array - I.
The dataset and timing analysis, Astron. Astrophys. 678 (2023) A48 [2306.16224].
[167] EPTA, InPTA: collaboration, The second data release from the European Pulsar Timing
Array - III. Search for gravitational wave signals, Astron. Astrophys. 678 (2023) A50
[2306.16214].
[168] P. Rana et al., The Indian Pulsar Timing Array Data Release 2: I. Dataset and Timing
Analysis, 2506.16769.
[169] D.J. Reardon et al., Search for an Isotropic Gravitational-wave Background with the Parkes
Pulsar Timing Array, Astrophys. J. Lett. 951 (2023) L6 [2306.16215].
[170] A. Zic et al., The Parkes Pulsar Timing Array third data release, Publ. Astron. Soc.
Austral. 40 (2023) e049 [2306.16230].
[171] H. Xu et al., Searching for the Nano-Hertz Stochastic Gravitational Wave Background with
the Chinese Pulsar Timing Array Data Release I, Res. Astron. Astrophys. 23 (2023) 075024
[2306.16216].
[172] A. Weltman et al., Fundamental physics with the Square Kilometre Array, Publ. Astron.
Soc. Austral. 37 (2020) e002 [1810.02680].
[173] G. Janssen et al., Gravitational wave astronomy with the SKA, PoS AASKA14 (2015) 037
[1501.00127].
[174] P.R. Saulson, Fundamentals of Interferometric Gravitational Wave Detectors, World
Scientific, Singapore (1994).
[175] W.-R. Hu and Y.-L. Wu, The Taiji Program in Space for gravitational wave physics and the
nature of gravity, Natl. Sci. Rev. 4 (2017) 685.
[176] W.-H. Ruan, C. Liu, Z.-K. Guo, Y.-L. Wu and R.-G. Cai, The LISA-Taiji network, Nature
Astron. 4 (2020) 108 [2002.03603].

– 106 –
[177] TianQin collaboration, The TianQin project: current progress on science and technology,
PTEP 2021 (2021) 05A107 [2008.10332].
[178] S. Kawamura et al., Current status of space gravitational wave antenna DECIGO and
B-DECIGO, PTEP 2021 (2021) 05A105 [2006.13545].
[179] A. Maleknejad and F. Rompineve, Gravitational waves: a golden era, CERN Courier
(2023) .
[180] N. Aggarwal et al., Challenges and opportunities of gravitational-wave detection at high
frequencies, Living Rev. Relativ. (2025) .
[181] J. Weber, Detection and Generation of Gravitational Waves, Phys. Rev. 117 (1960) 306.
[182] A.S. Chou et al., MHz gravitational wave constraints with decameter Michelson
interferometers, Phys. Rev. D 95 (2017) 063002 [1611.05560].
[183] A. Patra et al., Broadband Limits on Stochastic Length Fluctuations from a Pair of
Table-Top Interferometers, Phys. Rev. Lett. 135 (2025) 101402 [2410.09175].
[184] AURIGA collaboration, Present performance and future upgrades of the AURIGA
capacitive readout, Class. Quant. Grav. 23 (2006) S103.
[185] N. Aggarwal, G.P. Winstone, M. Teo, M. Baryakhtar, S.L. Larson, V. Kalogera et al.,
Searching for New Physics with a Levitated-Sensor-Based Gravitational-Wave Detector,
Phys. Rev. Lett. 128 (2022) 111101 [2010.13157].
[186] V. Domcke, S.A.R. Ellis and N.L. Rodd, Magnets are Weber Bar Gravitational Wave
Detectors, Phys. Rev. Lett. 134 (2025) 231401 [2408.01483].
[187] A. Berlin, D. Blas, R. Tito D’Agnolo, S.A.R. Ellis, R. Harnik, Y. Kahn et al., Detecting
high-frequency gravitational waves with microwave cavities, Phys. Rev. D 105 (2022)
116011 [2112.11465].
[188] G. Marconato, J. Branlard, C. Dokuyucu, B. Giaccone, W. Hillert, T. Krokotsch et al.,
High frequency gravitational wave sensing with superconducting microwave cavities, .
[189] R. Ballou et al., Latest results of the OSQAR photon regeneration experiment, Phys. Rev. D
92 (2015) 092002 [1506.08082].
[190] O. Albrecht et al., The ALPS II experiment: a laser-based search for new physics, JINST
15 (2020) P10003 [2003.05835].
[191] V. Anastassopoulos et al., New CAST limit on the axion–photon interaction, Nature Phys.
13 (2017) 584 [1705.02290].
[192] A. Abeln et al., Conceptual design of the international axion observatory (IAXO), Eur.
Phys. J. C 80 (2020) 497 [2010.12076].
[193] E. Armengaud et al., Physics potential of the international axion observatory (IAXO),
JCAP 06 (2019) 047 [1904.09155].
[194] J. Beacham et al., Physics beyond the standard model at CERN: beyond the LHC, Phys.
Rept. 873 (2020) 1 [1901.09966].
[195] M.E. Gertsenshtein, Wave resonance of light and gravitational waves, Sov. Phys. JETP 14
(1962) 84.
[196] D. Ejlli et al., Probing high-frequency gravitational waves with photon regeneration, Eur.
Phys. J. C 79 (2019) 1032 [1908.00232].

– 107 –
[197] W. Ratzinger, S. Schenk and P. Schwaller, A coordinate-independent formalism for
detecting high-frequency gravitational waves, JHEP 08 (2024) 195 [2404.08572].
[198] A. Blais, A.L. Grimsmo, S.M. Girvin and A. Wallraff, Circuit quantum electrodynamics,
Reviews of Modern Physics 93 (2021) 025005.
[199] T.J. Stavenger, E. Crane, K.C. Smith, C.T. Kang, S.M. Girvin and N. Wiebe, Bosonic
qiskit, arXiv e-prints (2022) [2209.11153].
[200] J.Y. Araz, M. Grau, J. Montgomery and F. Ringer, Toward hybrid quantum simulations
with qubits and qumodes on trapped-ion platforms, arXiv e-prints (2024) [2410.07346].
[201] R.W. Heeres, B. Vlastakis, E. Holland, S. Krastanov, V.V. Albert, L. Frunzio et al., Cavity
state manipulation using photon-number selective phase gates, Physical Review Letters 115
(2015) 137002.
[202] A. Eickbusch, V. Sivak, A.Z. Ding, S.S. Elder, S.R. Jha, J. Venkatraman et al., Fast
universal control of an oscillator with weak dispersive coupling to a qubit, Nature Physics
18 (2022) 1464.
[203] R. Cheng, Y. Zhou, S. Wang, M. Shen, T. Taher and H.X. Tang, A 100-pixel
photon-number-resolving detector unveiling photon statistics, Nature Photonics 17 (2023)
112.
[204] F. Dyson, Is a graviton detectable?, International Journal of Modern Physics A 28 (2013)
1330041.
[205] T. Rothman and S. Boughn, Can gravitons be detected?, Foundations of Physics 36 (2006)
1801 [gr-qc/0601043].
[206] D. Carney, V. Domcke and N.L. Rodd, Graviton detection and the quantization of gravity,
Physical Review D 109 (2024) 044009 [2308.12988].
[207] T. Clifton, P.G. Ferreira, A. Padilla and C. Skordis, Modified Gravity and Cosmology, Phys.
Rept. 513 (2012) 1 [1106.2476].
[208] M. Lewandowski, A. Maleknejad and L. Senatore, An effective description of dark matter
and dark energy in the mildly non-linear regime, JCAP 05 (2017) 038 [1611.07966].
[209] S. Capozziello, M. De Laurentis, S. Nojiri and S.D. Odintsov, Evolution of gravitons in
accelerating cosmologies: The case of extended gravity, Phys. Rev. D 95 (2017) 083524
[1702.05517].
[210] P. Creminelli and F. Vernizzi, Dark Energy after GW170817 and GRB170817A, Phys. Rev.
Lett. 119 (2017) 251302 [1710.05877].
[211] T. Baker, E. Bellini, P.G. Ferreira, M. Lagos, J. Noller and I. Sawicki, Strong constraints on
cosmological gravity from GW170817 and GRB 170817A, Phys. Rev. Lett. 119 (2017)
251301 [1710.06394].
[212] N. Frusciante and L. Perenon, Effective field theory of dark energy: A review, Phys. Rept.
857 (2020) 1 [1907.03150].
[213] LIGO Scientific and Virgo collaboration, GW170817: Observation of gravitational
waves from a binary neutron star inspiral, Phys. Rev. Lett. 119 (2017) 161101
[1710.05832].
[214] L. Barack et al., Black holes, gravitational waves and fundamental physics: A roadmap,
Class. Quant. Grav. 36 (2019) 143001 [1806.05195].

– 108 –
[215] M.C. Begelman, R.D. Blandford and M.J. Rees, Massive black hole binaries in active
galactic nuclei, Nature 287 (1980) 307.
[216] M. Volonteri, Formation of supermassive black holes, Astron. Astrophys. Rev. 18 (2010) 279.
[217] P. Amaro-Seoane et al., Low-frequency gravitational-wave science with eLISA/NGO, Class.
Quant. Grav. 29 (2012) 124016 [1201.3621].
[218] L. Barack and C. Cutler, LISA capture sources: Approximate waveforms, signal-to-noise
ratios and parameter estimation accuracy, Phys. Rev. D 75 (2007) 042003 [gr-qc/0612029].
[219] S. Babak et al., Science with the space-based interferometer LISA. V: Extreme mass-ratio
inspirals, Phys. Rev. D 95 (2017) 103012 [1703.09722].
[220] P. Amaro-Seoane, Gravitational waves from compact binaries and their detection with
LISA, Living Rev. Rel. 25 (2022) 4 [2203.12508].
[221] S. Burke-Spolaor et al., The astrophysics of nanohertz gravitational waves, Astron.
Astrophys. Rev. 27 (2019) 5 [1811.08826].
[222] L. Mayer, Massive black hole binaries in gas-rich galaxy mergers, Class. Quant. Grav. 30
(2013) 244008.
[223] J. Abedi, H. Dykaar and N. Afshordi, Echoes from the Abyss: Evidence for Planck-scale
structure at black hole horizons, Phys. Rev. D 96 (2017) 082004 [1612.00266].
[224] J. Abedi, H. Dykaar and N. Afshordi, Echoes from the Abyss: The Holiday Edition!, arXiv
e-prints (2017) [1701.03485].
[225] N. Afshordi, J. Abedi and H. Dykaar, Echoes from the Abyss: A highly spinning black hole
remnant for GW150914, Phys. Rev. D 98 (2018) 104046 [1803.10454].
[226] N. Andersson and K.D. Kokkotas, The r-mode instability in rotating neutron stars, Int. J.
Mod. Phys. D 10 (2001) 381 [gr-qc/0010102].
[227] K. Riles, Recent searches for continuous gravitational waves, Living Rev. Rel. 25 (2022) 3
[2206.06447].
[228] J. Kopp, S. Reddy, L. Roberts and X.-N. Zhou, Neutron stars and the dark matter problem,
Phys. Rev. Lett. 120 (2018) 201102 [1801.09738].
[229] A.E. Nelson, S. Reddy and D. Zhou, Dark halos around neutron stars and gravitational
waves, JCAP 07 (2019) 012 [1803.03266].
[230] J. Ellis, A. Hektor, G. Hütsi, K. Kannike, L. Marzola, M. Raidal et al., Search for dark
matter effects on gravitational signals from neutron star mergers, Phys. Lett. B 781 (2018)
607 [1710.05540].
[231] M. Baryakhtar, R. Lasenby and M.S. Teo, Dark matter in extreme astrophysical
environments, Phys. Rev. D 96 (2017) 035019 [1704.05081].
[232] L. Shao, Tests of gravity with pulsars and neutron stars, Universe 6 (2020) 173
[2012.10233].
[233] G. Grippa, G. Montani, M. Orsaria and I.F. Ranea-Sandoval, Dark matter admixed neutron
stars: A review, Universe 8 (2022) 326 [2203.08684].
[234] K. Kotake, Multiple physical elements governing core-collapse supernova explosions,
Comptes Rendus Physique 14 (2013) 318 [1110.5107].

– 109 –
[235] C.D. Ott, The gravitational wave signature of core-collapse supernovae, Class. Quant. Grav.
26 (2009) 063001 [0809.0695].
[236] A. Corsi, Gravitational waves from gamma-ray burst progenitors, Front. Astron. Space Sci.
6 (2019) 29 [1902.01335].
[237] A. Sesana, Systematic investigation of the expected gravitational wave signal from
supermassive black hole binaries in the pulsar timing band, Mon. Not. Roy. Astron. Soc.
433 (2013) L1.
[238] J. Antoniadis and others (EPTA Collaboration), The european pulsar timing array:
Evidence for a gravitational-wave background, Astron. Astrophys. 678 (2023) A50
[2306.16214].
[239] D.J. Reardon and others (PPTA Collaboration), Search for an astrophysical
gravitational-wave background in the second data release of the parkes pulsar timing array,
Astrophys. J. Lett. 951 (2023) L6 [2306.16215].
[240] ET collaboration, The Science of the Einstein Telescope, 2503.12263.
[241] LISA Cosmology Working Group collaboration, Gravitational waves from inflation in
LISA: reconstruction pipeline and physics interpretation, JCAP 11 (2024) 032
[2407.04356].
[242] J.-F. Dufaux, J. Bergman, G.N. Felder, L. Kofman and J.-P. Uzan, Theory and numerics of
gravitational waves from preheating after inflation, Phys. Rev. D 76 (2007) 123517.
[243] D.G. Figueroa and F. Torrenti, Gravitational waves from nonlinear dynamics of gauge fields
during preheating, JCAP 2017 (2017) 057.
[244] P. Adshead, J.T. Giblin and M. Pieroni, Gravitational waves from gauge preheating, JCAP
2019 (2019) 057.
[245] K.D. Lozanov and M.A. Amin, Gravitational perturbations from resonant non-abelian
gauge-field production after inflation, Phys. Rev. D 99 (2019) 123504.
[246] T. Clarke, R. Caldwell and K. Hinterbichler, Primordial gravitational waves and cosmology,
Phys. Rev. D 103 (2021) 043512.
[247] L. Pagano, L. Salvati and A. Melchiorri, New constraints on primordial gravitational waves
from planck data, Phys. Lett. B 760 (2016) 823.
[248] C. Caprini and D.G. Figueroa, Cosmological backgrounds of gravitational waves, Class.
Quant. Grav. 37 (2020) 163001.
[249] K.N. Ananda, C. Clarkson and D. Wands, The cosmological gravitational wave background
from primordial density perturbations, Physical Review D 75 (2007) 123518.
[250] D. Baumann, P.J. Steinhardt, K. Takahashi and K. Ichiki, Gravitational wave spectrum
induced by primordial scalar perturbations, Physical Review D 76 (2007) 084019.
[251] R. Saito and J. Yokoyama, Gravitational-wave background as a probe of the primordial
black-hole abundance, Physical Review Letters 102 (2009) 161101.
[252] K. Kohri and T. Terada, Semianalytic calculation of gravitational wave spectrum nonlinearly
induced from primordial curvature perturbations, Phys. Rev. D 97 (2018) 123532.
[253] M. Biagetti, M. Fasiello and A. Riotto, Enhancing the tensor-to-scalar ratio in simple
inflationary models, Phys. Rev. D 88 (2013) 103518.

– 110 –
[254] J. Garcı́a-Bellido, M. Peloso, L. Sorbo and C. Ünal, Gravitational wave signatures from
inflationary particle production, JCAP 2021 (2021) 029.
[255] G. Domènech, Scalar induced gravitational waves review, Universe 7 (2021) 398.
[256] C. Yuan and Q.-G. Huang, A topic review on probing primordial black hole dark matter
with scalar induced gravitational waves, iScience 24 (2021) 102860.
[257] R.D. Peccei and H.R. Quinn, CP conservation in the presence of instantons, Phys. Rev.
Lett. 38 (1977) 1440.
[258] R.D. Peccei and H.R. Quinn, Constraints imposed by CP conservation in the presence of
instantons, Phys. Rev. D 16 (1977) 1791.
[259] S. Weinberg, A new light boson?, Phys. Rev. Lett. 40 (1978) 223.
[260] F. Wilczek, Problem of strong P and T invariance in the presence of instantons, Phys. Rev.
Lett. 40 (1978) 279.
[261] M. Dine, W. Fischler and M. Srednicki, A simple solution to the strong CP problem with a
harmless axion, Phys. Lett. B 104 (1981) 199.
[262] L.F. Abbott and M.B. Wise, Constraints on generalized axion models, Nucl. Phys. B 244
(1984) 541.
[263] M.S. Turner and L.M. Widrow, Inflation-produced, large-scale magnetic fields, Phys. Rev. D
37 (1988) 2743.
[264] K. Freese, J.A. Frieman and A.V. Olinto, Natural inflation with pseudo nambu-goldstone
bosons, Phys. Rev. Lett. 65 (1990) 3233.
[265] F.C. Adams, J.R. Bond, K. Freese, J.A. Frieman and A.V. Olinto, Natural inflation:
Particle physics models, power-law spectra for large-scale structure, and constraints from
cobe, Phys. Rev. D 47 (1993) 426.
[266] W.D. Garretson, G.B. Field and S.M. Carroll, Primordial magnetic fields from
pseudo-goldstone bosons, Phys. Rev. D 46 (1992) 5346.
[267] R. Contino, A. Podo and F. Revello, Chiral models of composite axions and accidental
Peccei-Quinn symmetry, JHEP 04 (2022) 180 [2112.09635].
[268] A. Maleknejad and E. McDonough, Ultralight pion and superheavy baryon dark matter,
Phys. Rev. D 106 (2022) 095011 [2205.12983].
[269] D.J.E. Marsh, Axion Cosmology, Phys. Rept. 643 (2016) 1 [1510.07633].
[270] P.W. Graham, I.G. Irastorza, S.K. Lamoreaux, A. Lindner and K.A. van Bibber,
Experimental Searches for the Axion and Axion-Like Particles, Ann. Rev. Nucl. Part. Sci.
65 (2015) 485 [1602.00039].
[271] I.G. Irastorza and J. Redondo, New experimental approaches in the search for axion-like
particles, Prog. Part. Nucl. Phys. 102 (2018) 89 [1801.08127].
[272] E.G.M. Ferreira, Ultra-light dark matter, Astron. Astrophys. Rev. 29 (2021) 7 [2005.03254].
[273] C.B. Adams et al., Axion Dark Matter, in Snowmass 2021, 3, 2022 [2203.14923].
[274] A. Ringwald, Axion dark matter (theory & experiment), PoS TAUP2023 (2024) 015
[2311.11660].
[275] A. Ringwald, Review on Axions, J. Phys. Conf. Ser. 3162 (2025) 012006 [2404.09036].

– 111 –
[276] M.M. Anber and L. Sorbo, N-flationary magnetic fields, JCAP 10 (2006) 018
[astro-ph/0606534].
[277] M.M. Anber and L. Sorbo, Naturally inflating on steep potentials through electromagnetic
dissipation, Phys. Rev. D 81 (2010) 043534 [0908.4089].
[278] N. Barnaby, E. Pajer and M. Peloso, Gauge field production in axion inflation:
Consequences for monodromy, non-gaussianity and gravitational waves, Phys. Rev. D 85
(2012) 023525 [1110.3327].
[279] A. Maleknejad and M.M. Sheikh-Jabbari, Gauge-flation: Inflation From Non-Abelian
Gauge Fields, Phys. Lett. B723 (2013) 224 [1102.1513].
[280] A. Maleknejad and M.M. Sheikh-Jabbari, Non-Abelian Gauge Field Inflation, Phys. Rev.
D84 (2011) 043515 [1102.1932].
[281] P. Adshead and M. Wyman, Chromo-Natural Inflation: Natural inflation on a steep
potential with classical non-Abelian gauge fields, Phys. Rev. Lett. 108 (2012) 261302
[1202.2366].
[282] E. Dimastrogiovanni, M. Fasiello and T. Fujita, Primordial Gravitational Waves from
Axion-Gauge Fields Dynamics, JCAP 1701 (2017) 019 [1608.04216].
[283] P. Adshead, E. Martinec and M. Wyman, Perturbations in Chromo-Natural Inflation,
JHEP 09 (2013) 087 [1305.2930].
[284] P. Adshead and E.I. Sfakianakis, Higgsed Gauge-flation, JHEP 08 (2017) 130 [1705.03024].
[285] A. Agrawal, T. Fujita and E. Komatsu, Large tensor non-Gaussianity from axion-gauge
field dynamics, Phys. Rev. D 97 (2018) 103526 [1707.03023].
[286] A. Agrawal, T. Fujita and E. Komatsu, Tensor Non-Gaussianity from Axion-Gauge-Fields
Dynamics : Parameter Search, JCAP 06 (2018) 027 [1802.09284].
[287] A. Maleknejad and M. Zarei, Slow-roll trajectories in Chromo-Natural and Gauge-flation
Models, an exhaustive analysis, Phys. Rev. D88 (2013) 043509 [1212.6760].
[288] C.M. Nieto and Y. Rodriguez, Massive Gauge-flation, Mod. Phys. Lett. A 31 (2016)
1640005 [1602.07197].
[289] P. Adshead, E. Martinec, E.I. Sfakianakis and M. Wyman, Higgsed Chromo-Natural
Inflation, JHEP 12 (2016) 137 [1609.04025].
[290] E. McDonough and S. Alexander, Observable Chiral Gravitational Waves from Inflation in
String Theory, JCAP 11 (2018) 030 [1806.05684].
[291] J. Holland, I. Zavala and G. Tasinato, On chromonatural inflation in string theory, JCAP
12 (2020) 026 [2009.00653].
[292] Y. Watanabe and E. Komatsu, Gravitational Wave from Axion-SU(2) Gauge Fields:
Effective Field Theory for Kinetically Driven Inflation, 2004.04350.
[293] A. Maleknejad, SU(2)R and its axion in cosmology: A common origin for inflation, cold
sterile neutrinos, and baryogenesis, Phys. Rev. D 104 (2021) 083518 [2012.11516].
[294] O. Iarygina and E.I. Sfakianakis, Gravitational waves from spectator Gauge-flation, JCAP
11 (2021) 023 [2105.06972].
[295] E. Dimastrogiovanni, M. Fasiello, J.M. Leedom, M. Putti and A. Westphal, Gravitational
axiverse spectroscopy: seeing the forest for the axions, JHEP 08 (2024) 072 [2312.13431].

– 112 –
[296] A. Maleknejad, M.M. Sheikh-Jabbari and J. Soda, Gauge-flation and Cosmic No-Hair
Conjecture, JCAP 1201 (2012) 016 [1109.5573].
[297] A. Maleknejad and E. Erfani, Chromo-Natural Model in Anisotropic Background, JCAP
1403 (2014) 016 [1311.3361].
[298] I. Wolfson, A. Maleknejad and E. Komatsu, How attractive is the isotropic attractor
solution of axion-SU(2) inflation?, JCAP 09 (2020) 047 [2003.01617].
[299] I. Wolfson, A. Maleknejad, T. Murata, E. Komatsu and T. Kobayashi, The isotropic
attractor solution of axion-SU(2) inflation: universal isotropization in Bianchi type-I
geometry, JCAP 09 (2021) 031 [2105.06259].
[300] P. Adshead and A. Liu, Anisotropic Massive Gauge-flation, JCAP 07 (2018) 052
[1803.07168].
[301] V. Kamali, Warm pseudoscalar inflation, Phys. Rev. D 100 (2019) 043520 [1901.01897].
[302] K.V. Berghaus, P.W. Graham and D.E. Kaplan, Minimal Warm Inflation, JCAP 03 (2020)
034 [1910.07525].
[303] A. Mehrabi, A. Maleknejad and V. Kamali, Gaugessence: a dark energy model with early
time radiation-like equation of state, Astrophys. Space Sci. 362 (2017) 53 [1510.00838].
[304] L. Heisenberg, R. Kase and S. Tsujikawa, Anisotropic cosmological solutions in massive
vector theories, JCAP 11 (2016) 008 [1607.03175].
[305] R.R. Caldwell and C. Devulder, Gravitational Wave Opacity from Gauge Field Dark
Energy, Phys. Rev. D 100 (2019) 103510 [1802.07371].
[306] A. Guarnizo, J.B. Orjuela-Quintana and C.A. Valenzuela-Toledo, Dynamical analysis of
cosmological models with non-Abelian gauge vector fields, Phys. Rev. D 102 (2020) 083507
[2007.12964].
[307] K. Kaneta, H.-S. Lee, J. Lee and J. Yi, Dark energy under a gauge symmetry: A review of
gauged quintessence and its implications, Int. J. Mod. Phys. A 40 (2025) 2530009
[2504.04820].
[308] L. Mirzagholi, A. Maleknejad and K.D. Lozanov, Production and backreaction of fermions
from axion-SU (2) gauge fields during inflation, Phys. Rev. D 101 (2020) 083528
[1905.09258].
[309] K.D. Lozanov, A. Maleknejad and E. Komatsu, Schwinger Effect by an SU (2) Gauge Field
during Inflation, JHEP 02 (2019) 041 [1805.09318].
[310] A. Maleknejad, Dark Fermions and Spontaneous CP violation in SU (2)-axion Inflation,
JHEP 07 (2020) 154 [1909.11545].
[311] L. Vicente Garcı́a-Consuegra and A. Maleknejad, The Stochastic Schwinger Effect,
2510.14468.
[312] A. Maleknejad, Chiral anomaly in SU(2)R -axion inflation and the new prediction for
particle cosmology, JHEP 21 (2020) 113 [2103.14611].
[313] G. Elor et al., New Ideas in Baryogenesis: A Snowmass White Paper, in Snowmass 2021, 3,
2022 [2203.05010].
[314] Y. Minami and E. Komatsu, New extraction of the cosmic birefringence from the planck
2018 polarization data, Phys. Rev. Lett. 125 (2020) 221301 [2011.11254].

– 113 –
[315] P. Diego-Palazuelos et al., Cosmic birefringence from the planck data release 4, Phys. Rev.
Lett. 128 (2022) 091302 [2201.07682].
[316] J.R. Eskilt and E. Komatsu, Improved constraints on cosmic birefringence from the WMAP
and Planck cosmic microwave background polarization data, Phys. Rev. D 106 (2022)
063503 [2205.13962].
[317] P. Diego-Palazuelos and E. Komatsu, Cosmic birefringence from the atacama cosmology
telescope data release 6, arXiv e-prints (2025) [2509.13654].
[318] A. Lue, L. Wang and M. Kamionkowski, Cosmological signature of new parity-violating
interactions, Phys. Rev. Lett. 83 (1999) 1506 [astro-ph/9812088].
[319] S. Saito, K. Ichiki and A. Taruya, Probing polarization states of primordial gravitational
waves with CMB anisotropies, JCAP 09 (2007) 002 [0705.3701].
[320] L. Sorbo, Parity violation in the Cosmic Microwave Background from a pseudoscalar
inflaton, JCAP 06 (2011) 003 [1101.1525].
[321] B. Thorne, T. Fujita, M. Hazumi, N. Katayama, E. Komatsu and M. Shiraishi, Finding the
chiral gravitational wave background of an axion-SU(2) inflationary model using CMB
observations and laser interferometers, Phys. Rev. D 97 (2018) 043506 [1707.03240].
[322] N. Seto and A. Taruya, Measuring a parity violation signature in the early universe via
ground-based laser interferometers, Phys. Rev. Lett. 99 (2007) 121101 [0707.0535].
[323] J.L. Cook and L. Sorbo, Particle production during inflation and gravitational waves
detectable by ground-based interferometers, Phys. Rev. D 85 (2012) 023534 [1109.0022].
[324] J. Crowder and N.J. Cornish, Beyond LISA: Exploring future gravitational wave missions,
Phys. Rev. D 88 (2013) 083005 [1307.5330].
[325] J. Chen, C. Liu, Y.-L. Zhang and G. Wang, Alternative LISA-TAIJI networks: Detectability
of parity violation in stochastic gravitational wave background, Phys. Rev. D 111 (2025)
084026 [2412.18420].
[326] G. Sato-Polito and M. Kamionkowski, Pulsar-timing measurement of the circular
polarization of the stochastic gravitational-wave background, Phys. Rev. D 106 (2022)
023004 [2111.05867].
[327] T. Kurita, D. Jamieson, E. Komatsu and F. Schmidt, Parity Violation in Galaxy Shapes:
Primordial Non-Gaussianity, 2509.08787.
[328] O.H.E. Philcox, Probing parity violation with the four-point correlation function of BOSS
galaxies, Phys. Rev. D 106 (2022) 063501 [2206.04227].
[329] J. Hou, Z. Slepian and R.N. Cahn, Measurement of parity-odd modes in the large-scale
four-point correlation function of SDSS BOSS galaxies, Mon. Not. Roy. Astron. Soc. 522
(2023) 5701 [2206.03625].
[330] A. Krolewski, S. May, K. Smith and P. Hopkins, No evidence for parity violation in BOSS,
JCAP 08 (2024) 044 [2407.03397].
[331] Z. Slepian et al., Measurement of parity-violating modes of the DESI year 1 luminous red
galaxies four-point correlation function, arXiv e-prints (2025) [2508.09133].
[332] E. Witten, Cosmic separation of phases, Phys. Rev. D 30 (1984) 272.
[333] C.J. Hogan, Gravitational radiation from cosmological phase transitions, Mon. Not. R.
Astron. Soc. 218 (1986) 629.

– 114 –
[334] M. Kamionkowski, A. Kosowsky and M.S. Turner, Gravitational radiation from first-order
phase transitions, Physical Review D 49 (1994) 2837.
[335] R. Apreda, M. Maggiore, A. Nicolis and A. Riotto, Gravitational waves from electroweak
phase transitions, Nuclear Physics B 631 (2002) 342.
[336] C. Caprini, R. Durrer and G. Servant, Gravitational wave generation from bubble collisions
in first-order phase transitions, Physical Review D 77 (2008) 124015.
[337] P. Schwaller, Gravitational waves from a dark phase transition, Phys. Rev. Lett. 115 (2015)
181101 [1504.07263].
[338] M. Breitbach, J. Kopp, E. Madge, T. Opferkuch and P. Schwaller, Dark, cold, and noisy:
Constraining secluded hidden sectors with gravitational waves, JCAP 07 (2019) 007
[1811.11175].
[339] D. Croon et al., Classifying dark sectors that give rise to a first-order phase transition,
JHEP 08 (2018) 203 [1806.02332].
[340] Y. Nakai, M. Suzuki, F. Takahashi and M. Yamada, Gravitational waves and dark radiation
from dark phase transition, Phys. Rev. D 102 (2020) 083521 [2009.09754].
[341] R. Pasechnik, M. Reichert, F. Sannino and Z.-W. Wang, Gravitational waves from
composite dark sectors, JHEP 02 (2024) 159 [2309.16755].
[342] C. Caprini and R. Durrer, Gravitational waves from stochastic relativistic sources:
Primordial magnetic fields, Phys. Rev. D 65 (2001) 023517.
[343] C. Caprini and R. Durrer, Gravitational wave production from primordial helical magnetic
fields, Phys. Rev. D 74 (2006) 063521.
[344] C. Caprini, R. Durrer and G. Servant, The stochastic gravitational wave background from
turbulence and magnetic fields generated by a first-order phase transition, Journal of
Cosmology and Astroparticle Physics 2009 (2009) 024 [0909.0622].
[345] M. Hindmarsh, S.J. Huber, K. Rummukainen and D.J. Weir, Numerical simulations of
acoustically generated gravitational waves at a first-order phase transition, Physical Review
D 92 (2015) 123009.
[346] M. Hindmarsh, S.J. Huber, K. Rummukainen and D.J. Weir, Shape of the acoustic
gravitational wave power spectrum from a first-order phase transition, Physical Review D
96 (2017) 103520.
[347] C. Caprini et al., Science with the space-based interferometer eLISA. II: Gravitational
waves from cosmological phase transitions, Journal of Cosmology and Astroparticle Physics
2016 (2016) 001.
[348] A. Mazumdar and G. White, Review article: Cosmic phase transitions and their
gravitational-wave signatures, Reports on Progress in Physics 86 (2023) 122901.
[349] P. Simakachorn, Charting Cosmological History and New Particle Physics with Primordial
Gravitational Waves, Ph.D. thesis, U. Hamburg (main), Hamburg U., 2022.
[350] T.W.B. Kibble, Topology of cosmic domains and strings, Journal of Physics A 9 (1976)
1387.
[351] A. Vilenkin and E.P.S. Shellard, Cosmic Strings and Other Topological Defects, Cambridge
University Press, Cambridge, UK (1994).
[352] A. Vilenkin, Gravitational radiation from cosmic strings, Phys. Lett. B 107 (1981) 47.

– 115 –
[353] R.H. Brandenberger, A. Albrecht and N. Turok, Gravitational Radiation From Cosmic
Strings and the Microwave Background, Nucl. Phys. B 277 (1986) 605.
[354] R. Brandenberger, Topological defects and structure formation, Int. J. Mod. Phys. A 9
(1994) 2117.
[355] T. Vachaspati and A. Vilenkin, Gravitational radiation from cosmic strings, Phys. Rev. D
31 (1985) 3052.
[356] B. Allen and E.P.S. Shellard, Gravitational radiation from cosmic strings, Phys. Rev. D 45
(1992) 1898.
[357] T. Damour and A. Vilenkin, Gravitational wave bursts from cosmic strings, Phys. Rev. Lett.
85 (2000) 3761.
[358] J.J. Blanco-Pillado, K.D. Olum and X. Siemens, New limits on cosmic strings from
gravitational wave observation, Phys. Lett. B 778 (2018) 392 [1709.02434].
[359] E. Witten, Superconducting strings, Nucl. Phys. B 249 (1985) 557.
[360] P. Auclair, S. Blasi, V. Brdar and K. Schmitz, Gravitational waves from current-carrying
cosmic strings, JCAP 2023 (2023) 009 [2207.03510].
[361] P. Auclair et al., Probing the gravitational wave background from cosmic strings with LISA,
JCAP 04 (2020) 034 [1909.00819].
[362] Y. Gouttenoire, G. Servant and P. Simakachorn, Beyond the Standard Models with Cosmic
Strings, JCAP 07 (2020) 032 [1912.02569].
[363] K. Schmitz and T. Schröder, Gravitational waves from cosmic strings for pedestrians,
2412.20907.
[364] N. Turok, Global texture as the origin of cosmic structure, Physical Review Letters 63
(1989) 2625.
[365] R. Durrer, A. Gangui and M. Sakellariadou, The CMB anisotropies from scaling seeds,
Physical Review D 59 (1999) 103501.
[366] J. Preskill, Semilocal defects, Physical Review D 46 (1992) 4218.
[367] T. Vachaspati and A. Vilenkin, Evolution of cosmic string loops, Physical Review D 35
(1987) 1131.
[368] S. Blasi, A. Mariotti, A. Sevrin, T. Van Proeyen and A. Van Steenhuyse, Domain-wall
induced cosmological phase transitions and gravitational waves, JCAP 2023 (2023) 051
[2210.06432].
[369] K. Kohri and T. Terada, Solar-Mass Primordial Black Holes Explain NANOGrav Hint of
Gravitational Waves, Phys. Lett. B 813 (2021) 136040 [2009.11853].
[370] S. Hawking, Gravitationally collapsed objects of very low mass, Monthly Notices of the
Royal Astronomical Society 152 (1971) 75.
[371] S. Bird, I. Cholis, J.B. Muñoz, Y. Ali-Haı̈moud, M. Kamionkowski, E.D. Kovetz et al., Did
LIGO detect dark matter?, Physical Review Letters 116 (2016) 201301.
[372] LISA Cosmology Working Group collaboration, Primordial black holes and their
gravitational-wave signatures, Living Rev. Rel. 28 (2025) 1 [2310.19857].
[373] C. Yuan and Q.-G. Huang, A topic review on probing primordial black hole dark matter
with scalar induced gravitational waves, iScience 24 (2021) 102860.

– 116 –