
2
astrophysical contribution decreases at
z.
30 [
69
,
71
,
72
]. The proposed next-generation detectors, such as the
Cosmic Explorer (CE) [
73
–
75
] and the Einstein Telescope
(ET) [
76
,
77
], whose horizons are up to
z∼
100 for stellar-
mass BBHs [
37
,
78
], may provide a unique opportunity
to test and shed light on the primordial origin of BH
mergers at high redshifts. A key question is therefore to
understand the uncertainties related to the measurements
of the source parameters, such as the redshift, masses and
spins.
In Ref. [
1
], we established the possibility of identify-
ing the PBH mergers with masses of 20 and 40
M
at
z≥
40 using single-event redshift measurements. We also
discussed how the prior knowledge of relative abundance
between Pop III and PBH mergers affects the statisti-
cal significance, assuming that there is a critical red-
shift,
zcrit
= 30, above which no astrophysical BBHs
are expected to merge. The results were based on full
Bayesian parameter estimation with a waveform model,
IMRPhenomXPHM
, which includes the effects of spin pre-
cession and higher-order modes (HoMs) [
79
–
81
]. In this
paper, we show the importance of HoMs to the param-
eter estimation of the high redshift BBHs at
z≥
10
in the context of PBH detections. We compare the
Bayesian posteriors of the relevant parameters obtained by
IMRPhenomXPHM
and the similar waveform family without
HoMs,
IMRPhenomPv2
[
82
–
84
] to systematically study the
improvement on measurements due to the HoM modeling
in the waveform.
We first recap the details of our simulations and the
settings of the parameter estimation in Sec. II. Then, we
show whether and how
IMRPhenomXPHM
performs better
when measuring redshift (Sec. III), as well as masses and
spins (Sec. IV), for BBHs with different sets of the source-
frame total mass, mass ratio, orbital inclination, and
redshift. Finally, in Sec. V, we re-examine the estimation
of the probability that a single source originated from
PBHs using redshift measurements under different choices
of
zcrit
, and discuss the possible implications of the mass
and spin measurements for PBH detections.
II. SIMULATION DETAILS
As in Ref. [
1
], we simulate BBHs at five different red-
shifts,
ˆz
= 10
,
20, 30, 40 and 50. The hat symbol denotes
the true value of a parameter here and throughout the pa-
per. To encompass the detectable mass range, we choose
the total masses in the source frame to be
ˆ
Mtot
= 5, 10,
20, 40, and 80
M
, with mass ratios
ˆq
= 1, 2, 3, 4 and 5.
Here, we define q≡m1/m2for m1> m2, where m1and
m2
are the primary and secondary mass, respectively. For
each mass pair, we further choose four orbital inclination
angles,
ˆι
= 0 (face-on),
π/
6,
π/
3, and
π/
2 (edge-on). All
simulated BBHs are non-spinning, as we expect that PBHs
are born with negligible spins [
16
,
17
,
85
] and may be spun-
up by accreting materials at later times [
25
,
26
,
85
,
86
].
However, we do not assume zero spins when performing
parameter estimation of the source parameters and in-
stead allow for generic spin-precession. For each of these
500 sources, the sky location and polarization angle are
chosen to maximize the signal-to-noise ratio (SNR) for
each source. The reference orbital phase and GPS time
are fixed at 0 and 1577491218, respectively. The baseline
detector network is a 40-km CE in the United States, 20-
km CE in Australia, and ET in Europe. We only analyze
simulated sources whose network SNRs are larger than
12. We use Planck 2018 Cosmology when calculating the
luminosity distance dLat a given redshift [87].
We employ a nested sampling algorithm [
88
,
89
] pack-
aged in Bilby [
90
] to obtain posterior probability densi-
ties. As we are only interested in the uncertainty caused
by the loudness of the signal, we use a zero-noise real-
ization [
91
] for the Bayesian inference and mitigate the
offsets potentially caused by Gaussian fluctuations [
92
].
To ensure our results are free from the systematics due
to the difference in the two waveform families, we use
the same waveform family for both simulating the wave-
forms and calculating the likelihood. That is, we use the
IMRPhenomXPHM
(
IMRPhenomPv2
) waveform template to
analyze the
IMRPhenomXPHM
(
IMRPhenomPv2
) simulated
waveforms
1
. The low-frequency cut off in the likelihood
calculations is 5 Hz for all sources.
As in Ref. [
1
], we first sample the parameter space
with uniform priors on the detector-frame total mass,
MD
tot
=
Mtot
(1+
z
), between [0
.
5
,
1
.
5]
ˆ
MD
tot
, and
q
between
[1
,
10]. The prior on redshift is uniform in the comoving
rate density,
∝dVc
dz
1
1+z
, between [
z
(
ˆ
dL/
10)
, z
(5
ˆ
dL
)]. In
Sec. V, we will revisit the physically motivated prior on
redshift. We use uniform priors for other parameters: the
sky position, the polarization angle, the orbital inclination,
the spin orientations, the spin magnitudes, the arrival
time and the phase of the signal at the time of arrival.
Then, we reweigh the posteriors into uniform prior on
the source-frame primary mass,
m1
, and the inverse mass
ratio 1
/q
(which is between [0
.
1
,
1]). Strictly speaking,
the marginalized one-dimensional priors on
m1
and 1
/q
are not exactly uniform after the reweighing because the
boundary of the square domain of (
MD
tot, q
) transforms
into a different shape according to the Jacobian. For ex-
ample, the marginalized prior on the redshift and that on
the inverse mass ratio have additional factors of 1
/
(1 +
z
)
and
q/
(
q
+ 1), respectively, upon the coordinate trans-
formation. However, we find that such boundary effect
has negligible effect on the posteriors. As we will discuss
below, the degeneracy among different parameters and
the scaling in p0(z) is more significant.
1
See Ref. [
93
] for the analysis of waveform systematics for the
next-generation GW detectors.