Predictions on SMBH mergers for LISA

Author: Pau Bosch Cabot

Facultat de Fı́sica, Universitat de Barcelona, Martı́ Franquès 1, 08028 Barcelona, Spain.

Advisor: Eduard Salvador-Solé

Abstract: We compute the number of Super-Massive Black Hole (SMBH) mergers that are
expected, according to the AMIGA model, in hierarchical galaxy formation within the standard
ΛCDM cosmology. We show how our results can be converted to the number and properties of
Gravitational Waves (GWs) expected to be detected with LISA, a new generation of GW detector
currently in development. The comparison of those predictions with observations will allow one to
check the validity of current models of SMBH growth and galaxy evolution.

I. INTRODUCTION tectors such as LISA (Laser Interferometer Space An-

tenna)[3] will be able to do. Indeed, LISA will cover the
GW frequency range (0,03-100 mHz) which will open the
The detection of gravitational waves (GWs) [1] has possibility to directly observe SMBH mergers and to re-
been one of the major scientific achievements of the cen- construct their growth process since the very early stages
tury. Not only has it confirmed the most amazing predic- of galaxy evolution.
tions of General Relativity (GR), but it has also opened In this work we use the Analytic Model of Intergalactic
a fully new powerful observational window in astronomy. Medium and GAlaxy evolution(AMIGA) model, devel-
The detected GWs arise from the formation of Black oped by the team led by Prof. Salvador [4], to compute
Holes (BHs) from the merger of two very dense objects, the expected number of SMBH mergers of various char-
neutron stars or BHs. The observation with GW of neu- acteristics. This will allow us to derive, in the future, the
tron star mergers and their electromagnetic counterparts properties of the GWs to be detected by LISA. The only
has had notable astrophysical consequences. However, attempts made so far to predict SMBH mergers have been
the only BH mergers (with no electromagnetic counter- carried out using galaxy formation semi-analytic models
part) possible to observe with current GW detectors cor- [5] and N-body simulations [6] which include important
respond to BH of a few tens of solar masses, i.e. in- approximations so that the results are not fully reliable.
between the stellar and super-massive black hole (SMBH) The AMIGA model is the most complete and detailed
mass ranges, where there is a lack of natural candi- galaxy formation model built so far. Thus, its predic-
dates from current astrophysical or cosmological theories. tions, which have been shown to be very accurate [7],
Thus, apart from demonstrating the reality and interest should really serve as a guide for the search and inter-
of GWs, the only impact of those observations has been pretation of the GWs detected by LISA and to check the
showing the existence of those objects. If, instead, we current ideas on SMBH and galaxy formation and evolu-
could observe the merger of SMBHs, this would provide tion.
very relevant information on the theory of galaxy forma-
tion and evolution.
SMBHs grow indeed in parallel with galaxies at their cen-
ter and play an important role in the evolution of their II. SMBH MERGERS AND GWS
hosts. However, SMBHs are often surrounded by dense
gas regions which absorb the light produced in their ac- The prediction of the existence of GWs was made in
cretion disks near the Schwarszchild radius, causing them 1916 by A. Einstein and published in 1937 in collabo-
to go undetected in the electromagnetic window. This ration with N. Rosen [8]. Nonetheless, it would not be
is the reason that SMBHs are so elusive to direct ob- confirmed until 2016, one hundred years later [1].
servation, particularly at high redshift. Certainly, even GWs are first order linear perturbations, hαβ , of the
though we cannot see all SMBHs, we could still detect Minkowsky metric, ηαβ , in empty space,
their mergers, which would be enough to monitor their
cosmological growth and probe galaxy formation mod- gαβ = ηαβ + hαβ , (1)
els. However, so far the only approach to studying these
events is by means of the so-called Pulsar-Timing Arrays caused by some energetic event, such as the in-spiraling
(PTAs) technique [2], which is extremely hard to carry and merger of a compact binary system giving rise to a
out and delivers ambiguous results. Indeed, it only de- BH, which propagate in space. As the motion they in-
tects SMBHs in the phase prior to the merger itself, so duce on any intervening mass is friction-less, GWs are
it informs on the existence of close SMBH pairs which not damped and they can travel very long distances be-
will sooner or later merge, but not on SMBH mergers fore they hit the detector.
themselves. More specifically, in the more practical transformed form
This is precisely what the next generation of GW de- hαβ = h̄αβ + 21 ηαβ h, the perturbed part of the field equa-
Predictions on SMBH mergers for LISA Pau Bosch

tions of GR satisfies the general relation to the relation

16πG GM
h̄αβ = − Tαβ ; ∂ β h̄αβ = 0. (2) λmin = 25/2 π , (5)
c4 c2
At the source, the non-null energy-momentum tensor which allows one to calculate the mass M = M1 + M2
Tαβ due to the binary system rotating around the cen- from the maximum pulsation or the minimum wavelength
ter of mass, generates the GW strain hαβ . Far from the of the oscillatory signal. Finally, from the values of Mc
source, Tαβ vanishes, and one is led to the wave equation, and M one can obtain the masses M1 and M2 of the two
h̄µν = 0, identical to that for electromagnetic waves in merging bodies.
relativistic electrodynamics. Finally, at reception, the On the other hand, assuming the two masses in the detec-
strain h̄αβ of the GW induces, through the non-null en- tor moving along the x1 -axis, the separation l(t) between
ergy momentum tensor, the oscillation of two free-falling them at the time t, equal to the integral of the element
masses separated by a distance l at rest. of separation dx1 , takes the form
The signal is thus characterized by: i) the wavelength λ Z q  
of that oscillation, set by the angular velocity or pulsa- 1
l(t) = gαβ dxα dxβ ' 1 + h11 (t) l. (6)
tion ω of the binary system, and ii) its amplitude ∆ll , set
2
by the power radiated by the binary system. As shown
next, quantity i) is related to the masses M1 and M2 of Equation (6) shows that the amplitude ∆l/l of the os-
the merging objects, while quantity ii) is related to their cillation is proportional to the maximum instantaneous
luminosity distance d (or redshift z) and location on the strain h11 of the GW, which is determined by the prop-
sky. Thus, from the detected signal one can infer the erties of the binary system through the equation (in any
properties of the rotating binary system and the result- reference frame)
ing BH and, conversely, from the latter one can predict
4 5/3 5/3
the characteristics of the former. hαβ (t) = G Mc ω(t)2/3 Aαβ , (7)
Specifically, the pulsation of a rotating binary system c4 d
moving in quasi-circular orbits around the center of mass where Aαβ is the GW polarisation (of modulus one) and d
approximately satisfies the Newtonian relations is the luminosity distance to the source. From equations
(6) and (7) we have the relation
GM c2 Rg
ω2 = = , (3)
R3 2R3 ∆l 12G5/3 5/3 2/3
= Mc ω , (8)
where G is the gravitational constant, c is the vacuum l c4 d
light-speed, R is the separation between the two bodies
and Rg = 2GM/c2 is the Schwarszschild radius of the from which we can infer the value of d.
BH with mass M = M1 + M2 that will result from the Strictly speaking, the previous relation presumes the de-
merger. tector along the plane orthogonal to the propagation of
As the binary system rotates, it loses energy and angular the GW where Aαβ is known to be circular. In any other
momentum through gravitational radiation, causing the plane, it is elliptical and the measured amplitude ∆l/l is
rotational velocity to slow down and the two bodies to smaller in general. The direction of propagation of the
approach each other in an in-spiraling motion. Given the GW is in principle unknown. However, provided we can
power radiated in the form of GWs, the pulsation ω of simultaneously detect the GW in more than one detector
the orbit varies according to with different orientations, one can infer the orthogonal
plane and the amplitude ∆l/l the signal would have in
96 5/3 5/3 11/3 that plane [9]. Thus, we can unambiguously determine in
ω̇ = G Mc ω , (4) that case not only the location in the sky of the source,
5c5
but also its distance d.
where Mc = (M1 M2 )3/5 /(M1 + M2 )1/5 is the chirp mass This is the theory. In practice, the amplitudes ∆l/l pro-
of the binary system. Equation (4) shows that, by mea- duced in mergers of compact binary systems are so tiny
suring the pulsation of the signal and its time variation, (∼ 10−21 ) that the detection of GWs is only possible
one can derive the value of Mc . with very sensitive instruments such as laser interferom-
When the distance R between the rotating bodies equals eters with extremely large values of l [1]. These instru-
the Schwarzschild radius Rg of the final BH, the two bod- ments consist of two free-falling mirrors in Fabry-Pérot
ies coalesce. This causes the pulsation to reach a maxi- resonant cavities which multiplies by a great factor their
mum ωmax (or the wavelength to reach a minimum λmin ) effective separation. The mirrors reflect coherent laser
after which it rapidly vanishes as the system relaxes into beams, which, at rest, arrive to the detector in counter-
a new rotating (Kerr) BH. The maximum pulsation of phase so that no light is detected. When the GW reaches
the signal thus satisfies the relations (3) with R = Rg . the detector and the mirrors oscillate, the optical path of
The second member of that equation leads to the value the beams varies and the phase of the laser beam changes,
of R in terms of ωmax and then the first member leads causing some light to be detected at the crossing point

Treball de Fi de Grau 2 Barcelona, June 2020

Predictions on SMBH mergers for LISA Pau Bosch

between the beams. and, in fact, has a large influence on their evolution.
The maximum sensitivity of such detectors is reached for What it is not clear is the ultimate origin of those
long arms so as to have large variations ∆l, although not SMBHs, i.e. wat are their seeds. In the most natural sce-
longer than λ/2 for the change in the phase of the beam nario, these seeds are the remnants of massive Population
due to the variation ∆l not to exceed one wavenumber. III (Pop III) stars which form from pristine gas through
In other words, the optimum configuration is for l = λ/2 H2 cooling. Although those stars have not been detected
[10]. But the wavelength of GWs depends on the source, yet, they must have existed because ordinary stars can-
so the size of the detector must be chosen according to not form from the primeval intergalactic medium with no
the kind of events intended to be observed. From ex- metals. Pop III stars, which can be extremely massive
pression (5) we see that the more massive the final BH, (hundreds of solar masses), would be the actual respon-
the greater the wavelength (and the larger the ampli- sible of the metal-enrichment of the cosmic gas enabling
tude, as well; eq.[7]). Consequently, in order to detect the formation of ordinary galaxies. And the most mas-
SMBH mergers, with strong signals, very long interfer- sive Pop III stars would also leave massive BH remnants,
ometer arms are needed. The arms of LIGO and VIRGO, with masses > 270 M , which would act as the seeds of
the most modern ground based GW detectors built so far, SMBHs.
have effective light paths ∼ 1.3 × 103 Km long, which is However, since the first galaxies form at redshifts of at
still too short to detect SMBH mergers. This will only most z = 30, some authors believe that such seeds can-
be possible with LISA, an interferometer with mirrors not grow fast enough to explain the existence of SMBHs
on board of three spacecrafts in formation flying in equi- with 109 M at z ∼ 6 as observed. For this reason an
lateral triangle, with effective light paths of 2.5 Gm [3]. alternate scenario has been proposed [12], according to
This means that it will be suited to monitor SMBHs with which SMBHs would form at z ∼ 10 through the direct
∼ 5 × 107 M , with a tolerance factor of ∼ 100 around collapse of huge gas clouds of ∼ 106 M which could not
that value. form stars because of being kept ionized by other nearby
galaxies.
The origin of SMBHs is thus an important issue of mod-
III. SMBH EVOLUTION ern cosmology as SMBHs play a crucial role in galaxy
evolution itself. Being able to directly determine, by ob-
servational means, the growth of SMBHs and, in par-
In the current paradigm of galaxy formation, structure
ticular, checking whether or not they already existed at
develops hierarchically [11]. The primordial tiny density
z >> 10 is thus an important goal.
fluctuations of dark matter and baryons emerging from
inflation develop by gravitational instability until they
become non-linear and collapse giving rise to a relaxed
structure. As dark matter is not supported by radiation IV. CALCULATIONS
pressure, it is the first component to form halos, which
continuously merge with each other and grow. After re- As mentioned, the AMIGA model [4] is a very complete
combination, baryons begin to move freely and fall into and detailed hierarchical galaxy formation model in the
dark matter halos. Inside them, the gas radiates and ΛCDM cosmology, which has been shown to make accu-
cools, so it concentrates at their center, where it gives rate predictions on galaxy properties in general [7]. In
rise to a new galaxy or is accreted onto any previous one. this model SMBHs evolve though accretion and mergers
As halos merge, their central galaxies orbit inside the new from the BH remnants of Pop III stars, so our predictions
halo, suffer orbital decay by dynamical friction, and fall must be interpreted within that particular framework.
to the center where they also merge. The technical characteristics of the AMIGA code are not
At least all nearby bright galaxies harbour SMBHs at relevant for our purposes here. What is important is that
their center as it has been detected through the emission AMIGA monitors the growth of every dark matter halo
of active galactic nuclei (AGN). Their mass (in the range of mass MH at redshift z = 0, i.e. the past history of all
of 106 − 109 M ), makes it hard to conceive a stellar ori- their central and satellite galaxies and, in particular, of
gin for these objects. In fact, their mass is always about the respective inner SMBHs. However, the mergers un-
5 × 10−3 times that of the host galaxy, which suggests dergone by those SMBHs are not stored. Thus, to achieve
SMBHs grow in parallel to their hosts. This is the reason our predictions we must classify and store all those merg-
that, in all current models of galaxy formation, SMBHs ers and, in the end, extract the resulting numbers.
are supposed to grow at the center of galaxies through We have thus proceeded to classify and accumulate in
the accretion of part of the gas fallen into the galaxy and the quantity N (M1 , M2 , MH , z) all SMBH mergers ac-
from the merger of the SMBHs originally at the center cording to the masses M1 > M2 of the progenitors and
of merging galaxies which suffer dynamical friction inside the redshift z where the merger takes place found over
the final galaxy and fall to its center, where they form the past history of every halo of mass MH at z = 0. Once
an in-spiraling compact binary system and coalesce. the cumulative process is finished we calculate the total
According to this commonly admitted scenario, the comoving density of SMBH mergers with M1 and M2 at
SMBH growth thus traces that of galaxies themselves z, ρ(M1 , M2 , z), by simply weighting the previous num-

Treball de Fi de Grau 3 Barcelona, June 2020

Predictions on SMBH mergers for LISA Pau Bosch

FIG. 1: Predicted number of SMBH mergers as a function of progenitor masses in various redshift intervals:
z = (24, 25), z = (8, 9) and z = (0, 1) (panels 1 to 3) and in the whole redshift range, z = (0, 30) (panel 4).

bers N according to the abundance of halos with MH at forming in our model as BH remnants of massive Pop III
z = 0, φH (MH ), stars to the most massive ones observed through AGN
Z (a few times 109 M ), in redshift intervals of ∆z = 1 in
ρ(M1 , M2 , z) = N (M1 , M2 , MH , z)φH (MH )dMH . (9) the range of z (0, 30). This has allowed us to monitor the
SMBH growth since the birth of ordinary galaxies.
To obtain the number of mergers within every redshift Our results are as expected in the assumed hierarchical
bin of width ∆z we must thus integrate that comoving galaxy formation paradigm. Indeed, from Figure 2 we
density over the redshift bin see that the total number of SMBH mergers N (z) in-
creases with decreasing redshift, reaching a maximum at
Z r(z+ ∆z
2 ) the last redshift bin z = (0, 1). (The secondary maxi-
N (M1 , M2 , z) = 4π ρ(M1 , M2 , z)r2 dr mum at z ∼ 2 is due to the fall of small galaxies with
r(z− ∆z
2 )
mini SMBHs into groups and clusters of galaxies, which
4π
 
∆z
 
∆z
 temporarily inhibits their mergers until satellite galax-
3 3
= ρ(M1 , M2 , z) r z + −r z− . ies become so numerous that they begin to be captured
3 2 2
by central ones.) More specifically, at z ∼ 24, the only
(10)
mergers observed are between SMBHs with 250−104 M
Finally, we can sum those numbers for all progenitor
(see panel 1 of Fig. 1), while, as SMBHs grow, mergers
masses to obtain the total number of mergers at differ-
involve increasingly massive SMBHs (panels 2 and 3).
ent redshift bins, N (z), or for all redshifts to obtain the
Nonetheless, SMBHs with 109 M begin to be present
total number of mergers per bin of progenitor masses,
at z ∼ 6, as observed. Thus the assumed SMBH seeds
N (M1 , M2 ).
fulfill that condition without problem. In fact, similarly
to what happens in the direct collapse scenario [12], at
V. RESULTS AND DISCUSSION z ∼ 10 we begin to find SMBHs more massive than 106
M . However, our predictions differ from those in the al-
ternate SMBH seed scenario in that: 1) there are SMBH
We have studied mergers between SMBHs in the mass mergers at z >> 10, and 2) there continue to be mergers
range 250 M −1010 M , that is from the smallest objects

Treball de Fi de Grau 4 Barcelona, June 2020

Predictions on SMBH mergers for LISA Pau Bosch

can be detected or not also depends on its location

in the sky relative to the orientation of the detector.
However, the geometrical configuration of LISA (three
interferometers in an equilateral triangle) guarantees the
full detection of all sources in the sky regardless of their
specific location. Finally, it may be preferable to have
the predicted numbers of SMBH mergers in terms of the
properties of the signal they will produce, i.e. in terms of
variables ω̇/ω = −λ̇/λ, λmin and ∆l/l, rather than of the
properties of the sources, M1 , M2 and z. This transfor-
mation is straightforward from equations (4), (5) and (7).

FIG. 2: Predicted total number of mergers as a VI. CONCLUSIONS

function of redshift for different M1 ranges.
We have shown how SMBH mergers produce GWs able
to be detected with LISA, a new space GW detector cur-
with low mass objects at z < 10. rently under development. We have calculated the num-
Mergers involving small SMBHs are in fact the most com- ber of SMBH mergers for different progenitor masses over
mon over all redshifts (see the green line in Fig. 2), and their full mass range and different redshift intervals over
are essentially all mergers seen (purple line). Nonethe- their whole evolution that are predicted by AMIGA, the
less, mergers between very massive SMBHs are still quite most complete and detailed galaxy formation model so
abundant (orange line), both at low z and globally (panel far built. Our results show that assuming that SMBHs
4 of Fig. 1), which is good news for LISA. The large fluc- arise as BH remnants of massive Pop III stars which grow
tuations seen at high redshifts in some lines are statistical through accretion and mergers is not contradictory with
(due to very small numbers). current observations. Nonetheless, a full check of that
But the previous absolute predictions are to be com- scenario must wait for the comparison with the detections
pared to the observations carried out with LISA. Thus found by LISA. This should indeed be possible because,
we must take into account the actual limited capabilities even though the big majority of mergers at all redshifts
of the instrument. First, the effective arm length of the correspond to low mass objects hard to detect with LISA,
interferometer, l ∼ 2.5 Gm, will make it hard for the less mergers of very massive SMBHs are still common enough
massive objects included in the previous analysis to be at z < 6 to be easily detected. Finally, we have explained
detected. Thus, we should apply a high pass-band filter how these predictions are to be adapted to the particular
of about 5 × 105 M to such predictions. Second, the arm length and sensitivity of LISA. I personally expect
limited sensitivity ∆l/l, still to be determined, of LISA to continue working in this subject in my Master Thesis.
will cause the signal-to-noise (S/N) from very distant
objects in principle detectable with LISA not to be high
enough. We must thus apply a cut-off in the mass of
objects that can be seen according to their distance Acknowledgments
using equation (8) which relates the minimum chirp
mass Mc of a system at redshift z with the minimum I would like to thank my advisor Eduard Salvador for
amplitude ∆l/l able to be distinguished from the noise. his constant dedication in transmitting his knowledge to
The two previous conditions will fix the real values me. I also thank Alberto Manrique for his help. This
of M1 and M2 which can be detected with LISA at degree would not have been the same without my parents,
every z. It could also be argued that whether a source who I deeply love and respect.

[1] Abbott, B.P. et al., 2016, PRL 116, 61. tivitätstheorie, 304:316.
[2] Sesana, A., 2013, MNRAS, 433, L1-L5. [9] Abbott, B.P. et al., 2017, PRL 119, 141101.
[3] Amaro-Seoane, P. et al., 2017, LISA, Mission Project. [10] Santamaria Lara, L., 2010, Doctoral dissertation, As-
[4] Manrique, A. et al., 2015, ApJ, 216, 13. trophysical Relativity, AEI-Golm, MPI for Gravitational
[5] Menou, K. et al., 2002, ApJ, 558, 535. Physics, Max Planck Society.
[6] Sesana, A. et al., 2004, ApJ, 611, 623. [11] Rees, M.J., Ostriker, J.P., 1977, MNRAS, 179, 541.
[7] Salvador-Solé, E. et al., 2017, ApJ, 834, 49. [12] Lodato, G., Natarajan, P. , 2006, MNRAS, 371, 1813.
[8] Einstein, A. & Rosen, N. , 1937, Albert Einstein’s Rela-

Treball de Fi de Grau 5 Barcelona, June 2020