Accepted by ApJ

Preprint typeset using LATEX style emulateapj v. 01/23/15

EVOLUTION OF GALAXY STAR FORMATION AND METALLICITY:

IMPACT ON DOUBLE COMPACT OBJECTS MERGERS
L. Boco1,2,3 , A. Lapi1,2,3,4 , M. Chruslinska5 , D. Donevski1 , A. Sicilia1 , L. Danese1,2
Accepted by ApJ

ABSTRACT
We study the impact of different galaxy statistics and empirical metallicity scaling relations on the
merging rates and on the properties of compact objects binaries. First, we analyze the similarities
and differences of using the star formation rate functions or the stellar mass functions as galaxy
arXiv:2012.02800v1 [astro-ph.GA] 4 Dec 2020

statistics for the computation of the cosmic star formation rate density. Then we investigate the
effects of adopting the Fundamental Metallicity Relation or a classic Mass Metallicity Relation to
assign metallicity to galaxies with given properties. We find that when the Fundamental Metallicity
Relation is exploited, the bulk of the star formation occurs at relatively high metallicities even at high
redshift; the opposite holds when the Mass Metallicity Relation is employed, since in this case the
metallicity at which most of the star formation takes place strongly decreases with redshift. We discuss
the various reasons and possible biases originating this discrepancy. Finally, we show the impact that
these different astrophysical prescriptions have on the merging rates and on the properties of compact
objects binaries; specifically, we present results for the redshift dependent merging rates and for the
chirp mass and time delay distributions of the merging binaries.
Subject headings: galaxies: statistics - galaxies: formation - galaxies: evolution - gravitational waves
- stars: black holes - stars: neutron

1. INTRODUCTION ume and chirp mass M6 as a function of the cosmic time

The discovery of gravitational waves (GWs) by the can be computed as (see Barrett et al. 2018 and Neijssel
LIGO/Virgo team (Abbott et al. 2016a, 2016b, 2016c, et al. 2019):
2017a, 2017b, 2017c, 2017d, 2017e, 2019; 2020a; 2020b; dṄ
Z Z
dN dṀSFR
also https://www.ligo.org/) has opened an observational (t) = dtd dZ (Z) (t−td )
window on the Universe with a new messenger. On the dVdM dMSFR dMdtd dVdZ
one hand, even few GW events with detected electro- (1)
magnetic counterparts can be of enormous importance where t is the cosmic time, equivalent to redshift, MSFR
for cosmology and fundamental physics (Creminelli et is the star formed mass, td is the delay time between the
al. 2017; Radice et al. 2018; Yang et al. 2019); on formation of the progenitor binary and the merging of
the other hand, large statistics of GWs can yield many the compact objects binary, Z is the metallicity and V
astrophysical information on stellar and binary evolution the comoving cosmological volume.
(e.g., Belczynski et al. 2016; Dvorkin et al. 2018; Mapelli The first term in the integral dN/dMSFR dM dtd is re-
& Giacobbo 2018), on the properties of the host galaxies lated to stellar and binary evolution and represents the
such as chemical evolution, star formation histories, ini- number of merging double compact objects (DCOs) per
tial mass function (IMF; e.g., O’Shaughnessy et al. 2010; unit of star forming mass per bin of chirp mass and time
de Mink & Belczynski 2015), and even on cosmology at delay. It can be evaluated via stellar and binary evo-
large (e.g., Taylor & Gair 2012; Nissanke et al. 2013; Liao lution simulations (see e.g. Dominik et al. 2012, 2015;
et al. 2017; Fishbach et al. 2019). Moreover, the pos- de Mink et al. 2013; de Mink & Belczynski 2015; Bel-
sibility to cross correlate the GW detected signals with czynskiet al. 2016; Spera & Mapelli 2017; Giacobbo &
some other tracers of the Large Scale Structures can help Mapelli 2018; Mapelli & Giacobbo 2018; Chruslinska et
to improve cosmological constraints or to test competing al. 2018; Spera et al. 2019; Santoliquido et al. 2020).
astrophysical frameworks (e.g. Oguri 2016; Raccanelli et Various processes involved in stellar and binary evolution
al. 2016; Scelfo et al. 2018, 2020; Calore et al. 2020). depend on metallicity (e.g. radiation-driven stellar wind
Given these numerous applications, it is important mass loss rates, core-collapse physics, mass transfer char-
to well characterize the population of compact objects acteristics and stability) and so the number of merging
(COs) merging binaries, to compute the related merging BH/NS binaries that form per unit mass formed in stars
rates and to understand their dependence on different also varies with this quantity.
astrophysical scenarios. The merging rates per unit vol- The second term dṀSFR /dV dZ is instead related to
galaxy evolution: it represents the star forming mass per
1 SISSA, Via Bonomea 265, 34136 Trieste, Italy units of time, comoving volume and metallicity, i.e. it is
2 IFPU-Institute for Fundamental Physics of the Universe, Via the star formation rate (SFR) density per metallicity bin.
Beirut 2, 34014 Trieste, Italy There are two main ways to estimate it: exploiting the
3 INFN-Sezione di Trieste, via Valerio 2, 34127 Trieste, Italy
4 INAF-Osservatorio Astronomico di Trieste, via Tiepolo 11, results of cosmological simulations (e.g., Mapelli et al.
34131 Trieste, Italy
5 Department of Astrophysics/IMAPP, Radboud University, 6 The chirp mass is defined as: M ≡ (m m )3/5 /(m + m )1/5 ,
1 2 1 2
PO Box 9010, NL-6500 GL Nijmegen, The Netherlands where m1 and m2 are the masses of the two merging objects.
2 L. BOCO ET AL.

2017; O’Shaughnessy et al. 2017; Lamberts et al. 2018; converted into SFR (e.g., Kennicutt 1998; Kennicutt &
Mapelli & Giacobbo 2018; Artale et al. 2019) or using Evans 2012).
empirical recipes concerning the cosmic SFR density and In fact, the SFR of a galaxy can be related to its
metallicity distributions inferred from observations (e.g., UV luminosity which mainly comes from young, blue
Belczynski et al. 2016; Lamberts et al. 2016; Cao et stars. However, since dust absorbs UV radiation and re-
al. 2018; Elbert et al. 2018; Li et al. 2018; Boco et al. emits it in the mid and far-IR band, the SFR estimated
2019; Chruslinska & Nelemans 2019; Neijssel et al. 2019; only from UV luminosities can be substantially under-
Santoliquido et al. 2020). estimated. Nevertheless it is still possible, for galaxies
The main focus of this work is to revise the differ- with a relatively low SFR (ψ . 30 − 50 M /yr) and rel-
ent empirical approaches pursued to compute the galac- atively small dust content, to estimate it from UV data
tic term, trying to quantify the impact of different alone using standard UV slope corrections (see Meurer
choices and to understand their advantages and draw- et al. 1999; Calzetti et al. 2000; Bouwens et al. 2015).
backs; moreover, we shall propose new ways to compute Therefore, the SFRF for ψ . 30 − 50 M /yr can be
it. We stress that the methods discussed in the present well constrained using data from deep UV surveys (see
paper to compute the galactic term are purely based on Wyder et al. 2005; Oesch et al. 2010; van der Burg et al.
observations and on empirically derived scaling relations, 2010; Cucciati et al. 2012; Finkelstein et al. 2015; Alavi
and do not rely on semi analytical models or simulations. et al. 2016; Bouwens et al. 2016, 2017; Bhatawdekar
Finally, we study the effects of the different prescriptions et al. 2018). Contrariwise, for what concerns highly
on the merging rates and on the properties of merging star-forming galaxies with ψ & 30 − 50 M /yr that are
binaries. much more rich in dust, UV corrections tend to fail (see
The plan of the paper is as follows. In section 2 we com- Silva et al. 1998; Efstathiou et al. 2000; Coppin et al.
pare the two main empirical ways to compute the cosmic 2015; Reddy et al. 2015; Fudamoto et al. 2017); as a
SFR density: via the luminosity/SFR functions (SFRF) consequence, the estimates of the SFR must be based
or via the galactic stellar mass functions (GSMF) and the on far-IR/(sub)mm wide-area surveys (see Lapi et al.
main sequence (MS) of star forming galaxies. In section 3 2011; Gruppioni et al. 2013, 2015; Gruppioni & Pozzi
we present the two main scaling relations to empirically 2019; Magnelli et al. 2013). However, given the sen-
assign metallicity to galaxies (the Fundamental Metal- sitivity limit of far-IR surveys, the shape of the SFR
licity Relation (FMR) or a Mass Metallicity Relation functions at the bright end becomes progressively un-
(MZR)) and we compute the galactic term dṀSFR /dV dZ certain at z & 3. Still relevant constraints in this regime
combining the two possible metallicity scaling relations have been obtained from deep radio surveys (Novak et al.
with the two galaxy statistics. In section 4, we compute 2017), from far-IR/(sub)millimeter stacking (see Rowan-
the merging rates and some properties of the compact bi- Robinson et al. 2016; Dunlop et al. 2017) and super-
naries for these different prescriptions, basing on the out- deblending techniques (see Liu et al. 2018), and from tar-
comes of the STARTRACK binary evolution simulations geted far-IR/(sub)millimeter observations of significant
as to compute the stellar factor dN/dMSFR dM•• dtd . Fi- yet not complete samples of star-forming galaxies (e.g.,
nally, in section 5 we summarize our main findings. Riechers et al. 2017; Marrone et al. 2018; Zavala et al.
Throughout this work, we rely on the standard flat 2018) and quasar hosts (e.g., Venemans et al. 2017, 2018;
ΛCDM cosmology (Planck Collaboration 2019) with cos- Stacey et al. 2018). Moreover, very recently, Gruppioni
mological parameters: ΩM = 0.32, Ωb = 0.05, H0 = et al. (2020) estimated the total IR luminosity func-
67 km s−1 Mpc−1 . The Chabrier IMF (2003, 2005; see tions up to redshift z . 6 from a sample of 56 galaxies
also Mo et al. 2010) is adopted, with mass range serendipitously detected by ALMA in the COSMOS and
0.08 − 150 M . A value Z = 0.0153 for the solar metal- ECDFS fields, finding a pleasant agreement with previ-
licity and 12 + log(O/H) = 8.76 for the solar oxygen ous far-IR/(sub)mm Hershel data, within the still large
abundance is adopted (Caffau et al. 2011). observational uncertainties.
The aforementioned set of data can be fitted via a sim-
2. COSMIC STAR FORMATION RATE AND GALAXY ple Schechter function:
STATISTICS
 1−α(z)
The first important ingredient in the computation of dN ψ
(ψ, t) = N (z) e−ψ/ψc (z) (2)
the factor dṀSFR /dVdZ is constituted by the cosmic d log ψ dV ψc (z)
SFR density dṀSFR /dV, representing the average rate at in terms of three fitting parameters: N (z), ψc (z) and
which new stars are formed in the Universe at different α(z) (see Table 1 of Mancuso et al. 2016a); in Fig. 1,
redshifts per unit comoving volume. There are mainly left panel, we show the datasets mentioned above and
two different ways to compute and exploit it, that are the fitted SFRF. At z & 1 heavily obscured, strongly
recalled and discussed below. star forming galaxies populate the bright end of the SFR
functions; these galaxies are the progenitors of local mas-
2.1. SFR/Luminosity functions sive ellipticals (ETGs) with final stellar mass M? & a few
The most direct approach to compute the cosmic SFR ×1010 M . Mildly star forming objects, instead, popu-
density relies on the galaxies star formation rate func- late the faint end and will end up in spheroid-like ob-
tions (SFRF) dN/dV d log ψ at different redshifts, repre- jects with rather low stellar mass (. 1010 M ). Finally,
senting the number density of galaxies per logarithmic late type disk galaxies (LTGs), with SFR of a few so-
bin of SFR (ψ). The SFRF can be computed from the lar masses per year, are well traced by the UV-inferred
UV and IR luminosity functions of galaxies (see Mancuso SFR function at z . 1. This is confirmed also by the
et al. 2016a; Boco et al. 2019), since luminosity can be link between the SFRF and the stellar mass function of
 Evolution of galaxy star formation and metallicity: impact on double compact objects mergers 3

different morphological types, obtained via a continuity Note that the MS is only an average relation between
Eq.approach by Lapi et al. (2017). mass and SFR; actually, star forming galaxies with fixed
From the SFRF, the cosmic SFR density can be easily mass at a given redshift tend to be distributed in SFR fol-
estimated as: lowing a double gaussian shape (see Sargent et al. 2012;
Z Béthermin et al. 2012; Ilbert et al. 2015; Schreiber et al.
dṀ dN 2015). This bimodal distribution highlights the existence
(t) = d log ψ ψ (ψ, t) (3)
dV d log ψ dV of two kind of galaxy populations: the dominant popula-
tion of main sequence galaxies (MSG), whose Gaussian
The resulting determination of the cosmic SFR den- distribution in SFR is centered around the MS value and
sity is shown as a dot-dashed black line in Fig. 2. It can the subdominant population of starburst galaxies (SBG),
be noticed that the cosmic SFR computed in this way whose Gaussian distribution is centered around a SFR
tends to be appreciably higher with respect to most of typically ∼ 3 − 4σ above the MS value. In the aforemen-
the previous determinations. This is due to the recent tioned works it is empirically found that the shape of
discovery via IR and far-IR/submm observations with the distribution is almost independent of the galaxy stel-
Hershel and ALMA, of a significant number of dusty star lar mass and redshift. On the other hand, other recent
forming galaxies very attenuated or even invisible in the studies, probing the SFR distribution of galaxies around
optical/UV bands. Such dusty galaxies, featuring an ex- the MS in a more extended range of masses and redshifts,
tremely high level of star formation (∼ 50−3000 M /yr), found an increase of the starbursts fraction at low masses
seem to have a significant impact on the total star forma- M? ≤ 109 M or at high redshifts z ≥ 2 − 3 (see Caputi
tion at 2 ≤ z ≤ 6 (see e.g. Wang et al. 2019; Gruppioni et al. 2017; Bisigello et al. 2018) .
et al. 2020; Smail et al. 2020). Therefore, since our For the sake of simplicity, in the present work we de-
SFRF fit is based also on data coming from recent far- scribe the galaxy distribution in SFR at fixed mass and
IR/(sub)mm surveys, the resulting cosmic SFR density redshift via a double Gaussian shape with the same pa-
is larger and more in agreement with the recent IR data rameters indicated by Sargent et al. (2012), and a fixed
(see e.g. Casey et al. 2018; Rowan-Robinson et al. 2016). starbursts fraction in each redshift and mass bins:
(log ψ − hlog ψiMS )2
 
2.2. Stellar mass functions + main sequence dp
(ψ|z, M? ) = AMS exp − 2 +
Another method to estimate the cosmic SFR density d log ψ 2σMS
is convolving the stellar mass functions of star forming (log ψ − hlog ψiSB )2
 
galaxies at different redshifts with a distribution around + ASB exp − 2
2σSB
the main sequence. This approach has been adopted in
(4)
Chruslinska & Nelemans (2019); here we recall it and im-
prove it by adding a simple treatment of starburst galax- where AMS = 0.97 is the fraction of MSG, ASB = 0.03
ies. the fraction of SBG, hlog ψiMS the value given by the MS
The galaxy stellar mass is routinely estimated via near- and representing the central value for the first Gaussian,
IR data and broadband SED fitting (e.g., da Cunha et hlog ψiSB = hlog ψiMS + 0.59 the central value of the
al. 2008; Boquien et al. 2019). The star forming galaxies second Gaussian, σMS = 0.188 the one-sigma dispersion
stellar mass function (GSMF) dN/dV d log M? has been of the first Gaussian and σSB = 0.243 the dispersion of
determined at different redshifts by several authors (e.g. the starburst population.
Ilbert et al. 2013; Muzzin et al. 2013; Tomczak et al. Once the distribution in Eq. (4) is convolved with the
2014; Davidzon et al. 2017). A good review of the differ- GSMF, one can reconstruct the SFRF of galaxies as
ent determinations can be found in Chruslinska & Nele- Z
mans (2019), where the authors provide an average fit dNGSMF+MS dN
(z, log ψ) = d log M? (z, log M? )×
between many different works. Here we adopt their fit, d log ψdV d log M? dV
and in particular their prescription for a redshift inde- dp
pendent slope at the faint end (see Fig. 1, right panel × (log ψ|z, M? )
d log ψ
and Fig. 3 in Chruslinska & Nelemans 2019, solid lines). (5)
In order to compute the cosmic SFR density, the
GSMF must be convolved with a distribution of SFR In Sargent et al. (2012) and Ilbert et al. (2015) it is
around the main sequence (MS) of star forming galaxies. also demonstrated that such a convolution yields a good
The MS is a well known (approximately powerlaw) rela- reconstruction of the luminosity functions.
tion between the stellar mass of the galaxy and its SFR Integrating the reconstructed SFRF in the Eq. above
at a given redshift. It has been determined both obser- over the whole range of star formation, as in Eq.(3), we
vationally and theoretically in different works (see e.g. obtain the cosmic SFR density as a function of the cos-
Daddi et al 2007; Rodighiero et al. 2011, 2015; Whitekar mic time. In Fig. 2 we show the cosmic SFR density
et al. 2014; Speagle et al. 2014; Schreiber et al. 2015; computed by integrating the SFRF directly fitted from
Mancuso et al. 2016b; Dunlop et al. 2017; Bisigello et the data (as in Eq.(2)) as a dot-dashed line, and the one
al. 2018; Pantoni et al. 2019; Lapi et al. 2020). How- derived from the GSMF as in Eq.(5) as a solid line. We
ever, we point out that the MS shape and evolution with find that the two determinations of the cosmic SFR den-
redshift is still debated, with relevant differences among sity are in rather good agreement up to redshift z ∼ 2. At
various works; in particular, its behaviour at large masses z > 2 the integration of the SFRF directly fitted from the
is very uncertain, with some authors advocating a pos- data yields a larger cosmic SFR density, with the maxi-
sible flattening (although it may be effectively due to mum differences being a factor ∼ 2.5 at z ∼ 4.5. These
contamination from passive galaxies). discrepancy, even if rather small, can be due to biases
4 L. BOCO ET AL.

Fig. 1.— Left panel: SFR functions at redshifts z = 0 (blue), 1 (red), 3 (green), 6 (magenta), and 8 (cyan). Solid lines show the rendition
from UV plus far-IR/(sub)millimeter/radio data, referring to the overall population of galaxies; dotted lines (only plotted at z ' 0 and 1)
show the rendition from (dust- corrected) UV data, referring to disk galaxies. UV data (open symbols) are from van der Burg et al. (2010;
triangles), Bouwens et al. (2016, 2017; pentagons), Finkelstein et al. (2015; hexagons), Cucciati et al. (2012; inverse triangles), Wyder
et al. (2005; circles), Oesch et al. (2010; squares), Alavi et al. (2016; stars), Bhatawdekar et al. (2018; rhombus); far-IR/(sub)millimeter
data (filled symbols) are from Gruppioni et al. (2020; filled plus),Gruppioni et al. (2015; circles), Magnelli et al. (2013; inverse triangles),
Gruppioni et al. (2013; triangles), Lapi et al. (2011; stars), and Cooray et al. (2014; pentagons); radio data are from Novak et al. (2017;
squares). Right panel: Stellar mass functions for star forming galaxies at redshifts z = 0 (blue), 1 (red), 3 (green), 6 (magenta), and 8
(cyan). Data points are taken from Davidzon et al. 2017.

and selection effects arising respectively in the chosen de- redshift is well established in literature. Therefore, once
termination of the SFR and stellar mass functions. For the stellar mass is known, it is easy to associate a SFR
example, the shape of the faint end of the GSMF at high and to use a scaling relation to infer its metallicity. Con-
redshift is highly uncertain and, using a mass function trariwise, fixing the SFR and redshift, the association of
whose shape steepens toward higher redshifts (see e.g. a stellar mass is not straightforward from an empirical
Fig. 3, dashed lines in Chruslinska & Nelemans 2019), point of view, and some assumptions about the galaxy
drastically reduce the differences. Other factors that can star formation history (SFH) should be made, as we will
produce these discrepancies are possible biases in the de- see in subsection 3.1.2. It is therefore trickier to use a
termination of the SFR from the UV+IR luminosity, in scaling relation to assign metallicity. Still, starting from
the shape of the main sequence or in the relative contri- the SFRF, it is possible to follow the chemical enrich-
butions of the main sequence and starburst populations. ment of a galaxy using a model of galaxy evolution, as
As for the latter, in Caputi et al. (2017) and Bisigello et done e.g. in Boco et al. (2019).
al. (2018), it is pointed out that the population of star- A final comparison concerns the possibility of disentan-
burst galaxies tends to increase at z & 2; keeping into gling different galactic populations using the two statis-
account this trend can reduce the differences between the tics mentioned. From the SFRF it can be determined
two cosmic SFR densities (Chruslinska et al. in prepa- the contribution to the total cosmic SFR density com-
ration). All in all, from Fig. 2 we have shown that the ing from late type disk galaxies (mainly traced by UV
two approaches yield a rather good agreement especially data) and progenitors of local early type galaxies (mainly
at z . 2, and we have quantitatively characterized the traced by far-IR/(sub)mm data). From the GSMF, in-
differences toward higher redshifts. For reference, in Fig. stead, it is possible to separate between main sequence
2 the classic determination of the cosmic SFR density by galaxies and starburts. Understanding the contribution
Madau & Dickinson (2014) is also reported, which is seen to the total SFR and the metallicities of different galactic
to be a factor ∼ 2 lower than some more recent IR data. populations can be important also for the association of
The main advantage of an approach based on the SFRF a host galaxy to a GW event.
is that it is rather direct. Indeed the SFR is the main Another method commonly used in the literature to de-
quantity we are interested in, since, provided an IMF, scribe the SFR density per metallicity bin as a function of
it gives the effective number of stars formed and so it redshift is to combine one of the cosmic SFR determina-
provides a normalization for the DCOs merging rates. tions with a standalone metallicity distribution (see e.g.
Starting from the SFRF we directly have a measure of the Belczynski et al. 2016; Cao et al. 2018; Li et al. 2018).
number density of galaxies with given SFR at a certain In this approach, however, the link with the properties
redshift. Instead, starting from the GSMF, the compu- of star forming galaxies and their evolution is lost and
tation of the SFR requires a step more, since it involves it is not easy to retrieve an accurate cosmic metallicity
the convolution with the main sequence and a correct distribution without passing through a galaxy statistics.
modelization of the relative abundance of MS galaxies
3. METALLICITY DISTRIBUTION
and starbursts.
On the other hand the GSMF provide a direct statistics Along with the stellar mass and star formation rate,
of the star forming galaxies stellar masses and, as shown the metal content of the gas-phase of the ISM (i.e. the
above, the distribution of SFRs at fixed stellar mass and gas-phase metallicity, Zgas ) is one of the key physical
quantities that has to be considered in statistical galaxy
 Evolution of galaxy star formation and metallicity: impact on double compact objects mergers 5

Fig. 2.— Cosmic SFR density as a function of redshift. The black solid line shows the result obtained integrating the SFR functions
reconstructed from the stellar mass functions plus the main sequence (Eq.(5)). The black dot dashed line shows, instead, the result of
the integration of the SFR functions (Eq.(2)). For reference, the dotted line illustrates the determination by Madau & Dickinson (2014).
Data are from (dust-corrected) UV observations by Schiminovich et al. (2005; cyan shaded area) and Bouwens et al. (2015; cyan squares);
ALMA submillimeter observations of UV-selected galaxies on the HUDF by Dunlop et al. (2017); VLA radio observations on the COSMOS
field by Novak et al. (2017); Herschel far-IR observations by Gruppioni et al. (2013 red shaded area) and Casey (2018; red pentagons);
Herschel far-IR stacking by Rowan-Robinson (2016; magenta circles); far-IR/(sub)millimeter observations from super-deblended data on
the GOODS field by Liu et al. (2018); and estimates from long GRB rates by Kistler et al. (2009, 2013; green reversed triangles).

evolution studies (for a review, see Maiolino & Mannucci sions with the modern cosmological simulations of mas-
2019). As it can be seen in Eq. (1), it is a crucial ingre- sive galaxy formation (e.g. Davé et al. 2017; Torrey et
dient also to compute the merging rates of DCOs, since al. 2018). The problem of accurately determining Zgas
many aspects of stellar and binary evolution depend on become strongly pronounced for high-z (z > 3) massive,
it. On global galaxy scales, the interplay between stel- dusty galaxies (see e.g. discussions in Tan et al. 2014;
lar mass, SFR and metallicity is naturally reflected by Liu et al. 2019; Tacconi et al. 2020), where the MZR
different scaling relations which encode informations on should be extrapolated. Indeed, if a linear extrapolation
the galaxy evolutionary stage. There are different ways of a sharply declining MZR is performed, very low values
to parametrize Zgas as a function of M? , redshift and/or of metallicities (12 + log(O/H) < 8.0) are found at z > 3
SFR, either through a Mass Metallicity Relation (MZR, even for massive systems.
e.g. Kewley & Ellison 2008; Maiolino et al. 2008; Man- The FMR, instead, is a three parameter relation among
nucci et al. 2009; Magnelli et al. 2012; Zahid et al. 2014; M? , SFR, and Zgas . The inclusion of SFR is to account
Genzel et al. 2015; Sanders et al. 2020a), or a Funda- for the secondary dependence of metallicity on SFR ini-
mental Metallicity Relation (FMR; e.g. Mannucci et al. tially observed in local SDSS galaxies (Mannucci et al.
2010; Mannucci et al. 2011; Hunt et al. 2016; Curti et 2010) where Zgas decreases increasing SFR at fixed M? .
al. 2020). This dependence has been also confirmed over larger data
The MZR is a correlation between Zgas (typically mea- sets: galaxies with the same stellar mass at the same red-
sured from strong optical oxygen nebular emission lines shift can have different metallicities due to their different
as 12+log(O/H)) and M? , and it is observationally found SFR, showing a clear anti-correlation between Zgas and
to be valid for objects with an M? spanning over 5 orders sSFR (see e.g. Hunt et al. 2016). The FMR is thought to
of magnitude. In general, at fixed M? the MZR predicts be almost redshift independent and this is confirmed by
a decline in Zgas towards higher redshifts and the level observations out to z ∼ 3.5 (Mannucci et al. 2010; Hunt
of redshift evolution is actively debated (Onodera et al. et al. 2016). Indeed, z is not a parameter directly en-
2016; Sanders et al. 2020a). Some earlier works (e.g. tering in the relation, and the metallicity evolution with
Maiolino et al. 2008; Mannucci et al. 2009; Magnelli redshift at fixed stellar mass can be traced back to the
et al. 2012) found a slow evolution of the MZR out to redshift evolution of the SFR (or sSFR), described by
z ∼ 2, but a very sharp decline in Zgas of about 0.4-0.5 the main sequence. Therefore the extrapolation of the
dex between z = 2.5 and z = 3.5, suggesting a huge drop FMR at z > 3.5 can be done following the redshift evo-
in Zgas in the early universe and creating somewhat ten- lution of the main sequence, which is determined out to
6 L. BOCO ET AL.

Fig. 3.— The average MZR relation hZMZR i computed convolving the FMR of Mannucci et al (2011) with the MS at different redshifts
(solid lines), compared with the MZR determination of Mannucci et al. (2009) at different redshifts (dashed lines).

z ∼ 6. This originates a rather shallow decline of metal- Nonetheless, very recent studies obtained a great
licity with redshift. agreement between the FMR and smoothly evolving
Thus, while the level of redshift evolution for the FMR MZR. Sanders et al. (2020) retrieved MOSDEF spec-
and MZR at z . 2 is somewhat comparable, the evolu- troscopy of a large sample of massive galaxies at z > 3,
tion of the two relations becomes completely different at and significantly improve the statistics upon past studies
z & 3. To explicitly show these differences we put the over the same redshift range. By employing a novel dust-
two relations on the same ground computing a kind of correction method, they found a much slower evolution of
averaged MZR from the FMR. We do this, at fixed stel- MZR observing a very shallow metallicity decline of only
lar mass and redshift, averaging the metallicity given by 0.11 dex between 2.5 < z < 3.5. This result greatly sup-
the FMR over the distribution of SFRs around the MS ports FMR and slowly evolving MZRs calibrated from
in Eq.(4): UV+IR data (Genzel et al. 2015).
Z On top of that, an important evidence of a significant
dp metal enrichment in the early Universe came from the
hZMZR i(z, M? ) = d log ψ (ψ|z, M? )ZFMR (M? , ψ) novel dust mass (Mdust ) estimates in distant galaxies at
d log ψ
(6) z > 3 − 6 (e.g. da Cunha et al. 2015, Donevski et al.
In Fig. 3 we show this averaged MZR (solid lines), com- 2020; Ginolfi et al. 2020). For instance, by analysing
puted from the Mannucci et al (2011) FMR which pro- a large sample of 300 massive (M? > 1010 M ) dusty
vides an updated version of the FMR presented in Man- galaxies in the COSMOS field observed with ALMA
nucci et al (2010) for lower mass galaxies. In the Fig- over a wide redshift range (0.5 < z < 5.25), Donevski
ure, for comparison, it is also reported the MZR of Man- et al. 2020 show that, in order to explain the ob-
nucci et al. (2009) (dashed lines), linearly extrapolated served Mdust , their Zgas are, on average, close to so-
at z > 3.5. It can be noted that, while at low redshifts lar (12+log(O/H)=8.64 and 12+log(O/H)=8.52 for MS
(z . 2) the two relations give comparable results, at high and SB galaxies, respectively). These values are in great
redshifts (z & 3) the MZR evolution is very rapid yield- agreement with recent direct Zgas measurements through
ing values of metallicity much lower than those obtained [NII]λ6584/Hα ratio by Shapley et al. (2020) for dusty
from the FMR. galaxies within the same mass range at z ∼ 2.
Trying to solve this tension is crucial, since, as seen All of that complements classical arguments from stel-
in section 2, the amount of SFR is not negligible at lar archaeology suggesting a fast metal enrichment of
z > 3. However, the quest is rather challenging, since on galaxies even at high z. Indeed the study of stellar emis-
the one hand, optical/near-IR spectroscopy suffers from sion in local massive early type galaxies can place very
large dust attenuation, on the other hand, the statis- good constraints on their metallicity evolution: stars ob-
tics of sources that have been spectroscopically studied served in these galaxies, formed typically at high red-
through fine structure lines with ALMA is still limited shifts, are found to be almost coeval and α-enhanced,
(Boogard et al. 2019). indicating a short (< 1 Gyr) burst of high star formation
 Evolution of galaxy star formation and metallicity: impact on double compact objects mergers 7

stopped by some form of energetic feedback (e.g., Ro- dṀSFR /dV dZ can be computed as:
mano et al. 2002; Thomas et al. 2005, 2010; Gallazzi et Z
al. 2006; Johansson et al. 2012). Their stellar metallic- dṀSFR dN
(Z|z) = d log M? (log M? |z)×
ity ranges from 0.5 Z − 2 Z (see Thomas et al. 2010, dVdZ dVd log M?
Gallazzi et al. 2014; Maiolino & Mannucci 2019) im- Z
dp
plying that their chemical enrichment should have been × d log ψ ψ (ψ|z, M? )×
rather rapid. In Morishita et al. 2019 a sample of 24 d log ψ

quiescent galaxies at z ∼ 2 has been studied finding av- dp 
× (Z|ZFMR (M? , ψ))
erage stellar metallicities of Z ∼ 1.5 − 2 Z (see also dZ FMR
Saracco et al. 2020); in particular, the authors find that (7)
the relation between stellar mass and stellar metallic-
ity of their sample shows no evolution with respect to where dp/d log Z|FMR (Z|ZFMR (M? , ψ))
 ∝
the same relation for z ∼ 0 galaxies in Gallazzi et al. exp −(log Z − log ZFMR (M? , ψ))2 /2σFMR 2
is just a
(2014). Finally, a direct measure of metallicity through log-normal distribution around the logarithmic metal-
[OIII]88µm /[NII]122µm line ratio in high redshift quasar licity value setted by the FMR at fixed stellar mass
hosts (up to z ∼ 7.5) has been performed by several and SFR, the factor dN/dV d log M? represents the
authors, showing solar and supersolar metallicity values GSMF, and the factor dp/d log ψ is the distribution in
with no signs of redshift evolution (see e.g. Juarez et al. SFR around the MS computed as in Eq.(4). Notice
2009; Novak et al. 2019; Onoue et al. 2020; Li et al. that, using the FMR, there is not an explicit redshift
2020). dependence on the value of the metallicity; the redshift
These numerous findings point towards the need of a dependence enters only indirectly through the GSMF
rapid metal enrichment in the distant Universe, a sce- and the distribution of SFR dp/d log ψ around the MS
nario that has recently been proposed theoretically by value.
several authors (Asano et al. 2013; Béthermin et al. In Fig. 4 (top panel), we show the result of Eq.(7),
2015; Popping et al. 2017; Vijayan et al. 2019; Pantoni using the Mannucci et al. (2011) FMR and the GSMF
et al. 2019; Lapi et al. 2020). We have also explored from Chruslinska & Nelemans (2019). The redshift de-
predictions on the metallicity evolution from the state- pendence of the cosmic SFR density reflects the black
of-the-art cosmological simulations (Davè et al. 2019) solid line in Fig. 2, as expected since the GSMF is used
that self-consistently model gas and dust under stan- as starting point. As for the metallicity dependence, we
dard IMF. By looking at different snapshots over the notice that its redshift evolution is very mild: there is not
redshift range 0 < z < 5 for the most massive objects a net evidence of a strong decrease with redshift of the
(1010 M < M? < 1011 M ) we found very shallow Zgas metallicity at which star formation occurs, as expected
evolution of only 0.3 dex drop from z ∼ 0 to z ∼ 5. This looking at Fig. 3. In fact, while at z . 2 most of the star
further strengthen the above cited observational findings formation takes place at around solar values, at z ∼ 4−5
that can suffer of selection biases. the typical values of metallicity at which star formation
Evidence for substantial metal content is also found occurs are around Z ∼ 0.4 − 0.5 Z . The bottom left and
for less massive galaxies (109 < M? < 109.5 M ) in the bottom right panels show, respectively, the contribution
epoch of re-ionization (6 < z < 9, Jones et al. 2020; of main sequence galaxies and starbursts. It can be no-
Strait et al. 2020). These studies claimed that the ob- ticed that the metallicity at which starbursts form stars
served Zgas can be achieved by extrapolating FMR or tends to be slightly lower. This is natural, since, at fixed
slowly evolving MZRs. mass, the FMR predicts lower metallicities increasing the
All these reasons motivate us to apply prescriptions SFR.
based on FMR as a main scaling relation to infer the
metal properties of galaxies at high z’s (see the next 3.1.2. SFRF + FMR
Section). However, given the substantial uncertainties, The FMR can be used to assign metallicities even if the
in Section 3.2 we also show the case in which a MZR SFRF are chosen as galaxy statistics. The main difficulty
with a rapid decrease in Zgas with redshift (e.g., Man- is that, while at fixed redshift and stellar mass we are
nucci et al. 2009) is assumed as representative for the able to construct a distribution of SFRs, it is not clear
whole population of galaxies at z > 3. how to derive a distribution of galaxies stellar masses at
fixed redshift and SFR; there are no works in literature
3.1. The galactic term computed through a FMR facing the issue of deriving a stellar mass distribution
Given all the arguments above we compute the galac- from empirical data. This is why, in order to roughly
estimate such stellar mass distribution, we must assume
tic term dṀSFR /dV dZ using the FMR presented in Man- a star formation history for our galaxies.
nucci et al. (2011), assuming that we can extrapolate the Since we are considering only star forming galaxies, the
FMR in the same form even at z > 3.5, as said in section value of masses that they can assume at fixed redshift and
3. Since the FMR is a relation between stellar mass, SFR SFR is less or equal than the value of mass given by the
and metallicity (ZFMR = ZFMR (M? , ψ)) we can use both main sequence M?,MS (z, ψ); all the values of mass larger
the GSMF (subsection 3.1.1) and the SFRF (subsection than that represent quenched galaxies which are no more
3.1.2) as galaxy statistics to perform the computation. forming stars. Actually we will not sharply cut all the
stellar masses above M?,MS , rather we put a Gaussian tail
3.1.1. GSMF + FMR
for masses M? ≥ M?,MS . As for the mass distribution for
Fixing redshift and stellar mass, we can derive a stellar masses smaller than the MS mass (M? < M?,MS )
distribution in SFR as in Eq.(4); therefore the factor we should make some assumptions on the galaxies SFH.
8 L. BOCO ET AL.

Fig. 4.— Top panel: The factor log (dṀSFR /dV d log Z) computed convolving the GSMF with the FMR of Mannucci et al. (2011) (color
coded) as a function of redshift on the x axis and gas phase metallicity 12 + log (O/H) on the left y axis; on the right y axis it is plotted
the logarithm of the metallicity log Z and the solar value is plotted as a black solid line. Bottom left panel: The contribution to the cosmic
SFR density coming from main sequence galaxies. Bottom right panel: The contribution to the cosmic SFR density coming from starburst
galaxies.
For ETG progenitor galaxies, SED-modeling studies tion of what has been done in Mancuso et al. (2016b),
(e.g., Papovich et al. 2011; Smit et al. 2012; Moustakas where the authors shows how to reproduce the MS with
et al. 2013; Steinhardt et al. 2014; Cassará et al. 2016; similar prescriptions.
Citro et al. 2016) suggest that the SFH can be described Using the SFRF as a starting point and the distribu-
with a truncated power-law shape rising with a shallow tion in Eq.(8) as stellar mass distribution at given z and
slope ≤ 0.5 for a quite short star formation timescale ψ, we can assign a metallicity to galaxies with the FMR
≤ 1Gyr. On the other hand, late type disk dominated and compute the factor dṀSFR /dVdZ as:
galaxies tend to have, on average, a SFH exponentially
declining over rather long star formation timescales τψ ∼
Z
dṀSFR dN
several Gyr (see Chiappini et al. 1997; Courteau et al. (Z|z) = d log ψ ψ ×
dV dZ dV d log ψ
2014; Pezzulli & Fraternali 2016; Grisoni et al. 2017; Z
Lapi et al. 2020). Even if the star formation timescales dp
× dM? (M? |z, ψ)
are very different, the SFRs in both cases are nearly con- dM?
stant with time: for the ETG progenitors the SFR range

dp 
is not larger than a factor ∼ 1.5 for most of their lifetime, × (Z|ZF M R (M? , ψ))
while for LTGs the SFR changes only of a factor ∼ 2.5 dZ FMR
over ∼ 8−9 Gyr. For this reasons, for the sake of simplic- (9)
ity, we assume a constant SFH for the galaxies considered
in this work. Under this assumption, the stellar mass of where
 dp/d log Z|FMR (Z|ZFMR (M? , ψ))
 ∝
a galaxy increases linearly with time, with a slope set by exp −(log Z − log ZFMR (M? , ψ))2 /2σFMR2
is the same
its SFR. The logarithmic distribution of masses, at fixed log-normal distribution around the central logarithmic
redshift and SFR, is therefore proportional to the mass value of metallicity set by the FMR appearing in Eq.(7).
itself and can be written as: The result of the computation in Eq.(9) is shown in Fig.
5. It can be noticed that the redshift dependence of the
dp SFR density reflects the shape of the cosmic SFR den-
(M? |z, ψ) ∝
d log M? sity derived by the integration of the SFRF (dot-dashed
(
M? M? < M?,MS lines in Fig. 2), with a broader peak slightly shifted to-
∝ 
(log M? −log M?,MS )2
 wards z ∼ 2.5−3, as expected, since the employed galaxy
M?,MS × exp − 2σM? M? ≥ M?,MS statistics is the same. The metallicity dependence on red-
(8) shift is similar to Fig. 4, since they share the same pre-
scription to assign metallicity (the FMR). In particular
normalized to unity. Actually, this is a crude approxima- the redshift decrease is mild also in this case, with most
 Evolution of galaxy star formation and metallicity: impact on double compact objects mergers 9

Fig. 5.— Top panel: The factor log (dṀSFR /dV d log Z) computed convolving the SFRF with the FMR of Mannucci et al. (2011) (color
coded) as a function of redshift on the x axis and gas phase metallicity 12 + log (O/H) on the left y axis; on the right y axis it is plotted
the logarithm of the metallicity log Z and the solar value is plotted as a black solid line. Bottom left panel: Contribution to the cosmic
SFR density coming from LTGs. Bottom right panel: Contribution to the cosmic SFR density coming from ETG progenitors.
of the star formation occurring at solar metallicities for a SFR to a galaxy with given stellar mass and redshift:
z . 2 and at Z ∼ 0.4 − 0.5 Z for z ∼ 4 − 5. In case the Z
SFRFs are used as a starting point, it is more difficult dṀSFR dN
(Z|z) = d log M? (log M? |z)×
to disentangle the contribution of main sequence galax- dVdZ dV d log M?
ies with respect to starbursts. On the other hand, it is 
dp 
easier to look at the contribution to the total SFR den- × (Z|ZMZR (z, M? ))×
sity given by LTGs and ETG progenitors, as explained in dZ MZR
section 2. These contributions are shown, respectively,
Z
dp
in the bottom left and bottom right panels of Fig. 5. × d log ψ (ψ|z, M? ) ψ
d log ψ
It can be noticed that LTGs, having on average a lower (10)
stellar mass, tend to produce a tail of lower metallicity
star formation even at low redshift. However, the bulk of where dp/d log Z|MZR (Z|ZMZR (z, M? )) ∝
the SFR density occurs at z & 1 and it is given by ETG 
exp −(log Z − log ZMZR (z, M? ))2 /2σMZR
2
is a log-
progenitors. normal distribution around the logarithmic value given
by the MZR (log ZMZR ).
In Fig. 6 we show the resulting dṀSFR /dV dZ (color
code) as a function of redshift and metallicity. As for
the redshift dependence of the cosmic SFR, it reflects
the shape presented in Fig. 2 (solid lines) obtained using
the GSMF as galaxy statistic, with a peak of star forma-
3.2. The galactic term computed through a MZR
tion around z ∼ 2. As for the metallicity dependence, at
lower redshifts z . 2 the metallicity stays rather high,
In subsection 3.1 we computed the galactic term using similarly to the FMR cases, with most of the star for-
a FMR which, as shown in our results, imply a shallow mation occurring at slightly supersolar values, while at
descrease of metallicity with redshift. We now compute higher z the metallicity starts to decline rapidly, with
the same factor dṀSFR /dV dZ using instead a sharply most of the star formation occurring at Z ≤ 0.1 Z at
declining MZR (Mannucci et al. 2009), linearly extrapo- z & 4, in contrast to the FMR cases in which the metal-
lated at z > 3.5 to see how much this choice will impact licity stays around ∼ 0.4 Z , reaching values of ∼ 0.1 Z
on the final results. only in the less massive systems. We stress that the dif-
The factor dṀSFR /dV dZ can be computed convolving ferences between the two approaches are rather small at
the MZR with the stellar GSMF at a given redshift and low redshifts z ≤ 2.5 and start to be significant at higher
using the distribution around the main sequence to assign redshifts, mainly in the regions where both the relations
10 L. BOCO ET AL.

Fig. 6.— The factor log (dMSFR˙ /dV d log Z) computed convolving the GSMF with the MZR of Mannucci et al. (2009) (color coded) as a
function of redshift on the x axis and gas phase metallicity 12 + log (O/H) on the left y axis; on the right y axis it is plotted the logarithm
of the metallicity log Z and the solar value is plotted as a black solid line.
have been extrapolated. The main message here is that tic terms dṀSFR /dV dZ, the merging rates could even
extrapolating the FMR, which is a redshift independent be strongly affected by the modelization of the factor
relation, can yield higher metallicity values with respect dN/dMSFR dMdtd , as we will see in subsection 4.1. So,
to a sharply declining MZR, more in agreement with the the presented merging rates should not be intended as an
arguments discussed at the beginning of this section. exact determination, but just as the results of the differ-
ent galactic prescriptions applied to a specific reference
4. MERGING RATES OF COMPACT BINARIES case for the stellar term dN/dMSFR dMdtd .
In this Section we show how the different galactic terms
dṀSFR /dV dZ computed above impact on the merging
rates and on the properties of merging compact binaries.
The three cases for which the galactic term has been com-
puted are: the stellar mass functions as galaxy statis-
tics and the FMR to assign metallicity (GSMF+FMR 4.1. The stellar term
case, Eq.(7)), the SFRF as statistics and the FMR to as- Within the isolated binary evolution scenario leading
sign metallicity (SFRF+FMR, Eq.(9)), the stellar mass to the formation of merging DCOs (such as considered in
function as statistics and the MZR to assign metallicity this paper), the stellar term appearing in Eq. (1) is com-
(GSMF+MZR, Eq.(10)). monly obtained from binary population synthesis simu-
However, to compute the merging rates (Eq.(1)) lations (e.g. Belczynski et al. 2016, Eldridge et al. 2017,
we need not only the galactic term dṀSFR /dV dZ Mapelli et al. 2017, Stevenson et al. 2017). The outcome
discussed throughout the paper, but also the term of those simulations (and therefore also the stellar term)
dN/dMSFR dMdtd , depending on stellar and binary evo- depends on a number of assumptions made in order to
lution, counting the number of merging events per units describe the evolution of massive stars and binary inter-
of star formed mass, chirp mass and time delay. In sub- actions. Many of those are highly uncertain (e.g. com-
section 4.1 we describe the choices done to compute this mon envelope evolution, core-collapse physics and the
term and then, in subsection 4.2, we derive the merging related natal kicks) and are known to strongly affect the
rates for the three types of compact binaries: BH-BH, properties of the simulated populations of merging DCO
NS-NS and BH-NS. Finally, in subsection 4.3, we discuss (e.g. Portegies Zwart & Yungelson 1998, Dominik et al.
about the time delay between the formation of the bi- 2012, Chruslinska et al. 2018). Stellar evolution depends
nary and the merger which will give us information on on metallicity - it affects, for instance, stellar winds and
the typical ages of the stellar population of the galaxy radii, also impacting the nature and outcome of binary
hosting the merging event. interactions (e.g. Maeder 1992; Hurley et al. 2000; Vink
A general caveat for this Section is that, while we com- et al. 2001; Belczynski et al. 2010a). As a consequence,
pare the results arising from the usage of different galac- the resulting stellar term also depends on metallicity.
 Evolution of galaxy star formation and metallicity: impact on double compact objects mergers 11

Fig. 7.— Top panel: merger rate density of double compact objects binaries as a function of redshift, computed using the GSMF as
galaxy statistics and the FMR, following Eq.(7). Blue lines refers to BH-BH, red lines to NS-NS, green lines to BH-NS events. Solid lines
represents the merging happening in all the galaxies (main sequence and starbursts), while the dashed lines highlight the contribution of
starbursts. The red and blue patches and the green arrow at z ∼ 0 represents the LIGO/Virgo 90% confidence intervals on the local rates
for NS-NS and BH-BH and the upper limit for BH-NS after the O1, O2 and first half of O3 runs (Abbott et al. 2019; Abbott et al. 2020).
Bottom panels: differential merging rates log (dṄ/dV d log M) for the BH-BH case (color code) as a function of redshift and chirp mass.
Contribution coming from all the galaxies (left panel), from main sequence galaxies (top right panel) and starbursts (bottom right panel).

This term can be separated into three main factors: erence B’ from Chruslinska et al. (2018) 7 . We note
that this is just an example. The exact results concern-
dN dN dp dp ing the populations of merging DCO presented further in
(Z) = (Z) × (Z) × this section would generally be affected by the choice of
dMSFR dM dtd dMSFR dM dtd the population synthesis model (e.g. Chruslinska et al.
(11) 2019). However, the main focus of this work is on the
where dN/dMSFR is the number of merging DCOs galactic term and the more in-depth discussion of the
formed per unit of mass formed in stars (formation uncertainties related to binary evolution is beyond the
efficiency) at metallicity Z, dp/dM is the metallicity- scope of this study.
dependent chirp mass distribution and dp/dtd is the dis- The formation efficiency as a function of metallicity for
tribution of delay times between the formation of the the chosen evolutionary model is shown in Chruslinska et
progenitor binary and the DCO merger. al. 2019 (Fig. 1, thin lines). The formation efficiency of
The delay time distribution resulting from binary pop- merging BH-BH is typically found to show a strong low
ulation synthesis is commonly found to be well described metallicity preference (e.g. Belczynski et al. 2010b; Do-
with a simple inverse proportionality dp/dtd ∝ t−1 d , in- minik et al. 2012; Eldridge & Stanway 2016; Stevenson et
dependent of the DCO type or metallicity. We assume al. 2017; Klencki et al. 2018; Giacobbo et al. 2018). This
dp/dtd ∝ t−1d with the minimum td,min = 50 Myr. The dependence is also present in the chosen model, with BH-
distribution is normalized to unity between td,min and BH formation efficiency dropping by almost two orders
the age of the Universe.
7 We use the simulation data publicly available under this url:
We base the remaining two factors on the results of
population synthesis calculations, using the model ’ref- https://www.syntheticuniverse.org/
12 L. BOCO ET AL.

Fig. 8.— Top panel: merger rate density of double compact objects binaries as a function of redshift, computed using the SFRF as
galaxy statistics and the FMR, following Eq.(9). Blue lines refers to BH-BH, red lines to NS-NS, green lines to BH-NS events. Solid lines
represents the merging happening in all the galaxies (LTGs, ETGs and their progenitors), while the dashed lines highlight the contribution
of LTGs. The red and blue patches and the green arrow at z ∼ 0 represents the LIGO/Virgo 90% confidence intervals on the local rates
for NS-NS and BH-BH and the upper limit for BH-NS after the O1, O2 and first half of O3 runs (Abbott et al. 2019; Abbott et al. 2020).
Bottom panels: differential merging rates log (dṄ/dV d log M) for the BH-BH case (color code) as a function of redshift and chirp mass.
Contribution coming from all the galaxies (left panel), from LTGs (top right panel) and ETGs and their progenitors (bottom right panel).

of magnitude between 0.2 Z and 0.4 Z . This depen- the chirp mass, we get the redshift distribution of the
dence is generally weaker for other DCO types and in the merging rate density for the three types of merging bi-
adopted model shows a factor of .10 increase/decrease naries: BH-BH, NS-NS and BH-NS. In Figs. 7, 8 and 9
towards high metallicity for NS-NS/BH-NS. we show, respectively, the results for the three different
Metallicity dependence of the chirp mass distribution ways to compute the galactic term described in this pa-
is to large extent a consequence of the metallicity de- per: GSMF+FMR (Eq.(7)), SFRF+FMR (Eq.(9)) and
pendence of the maximum mass of the stellar remnant GSMF+MZR (Eq.(10)).
resulting from single stellar evolution. Due to metallic- In the top panels are plotted the merging rates redshift
ity dependent line-driven wind mass loss rates of mas- distributions, highlighting the contribution of starburst
sive stars (e.g. Vink et al. 2001, Vink & de Koter 2005, galaxies in the GSMF+FMR case and the contribution
Sundqvist et al. 2019, Sander, Vink & Hartman 2020), of LTGs in the SFRF+FMR case. The local (z ∼ 0)
star with the same initial mass leaves a more massive merging rates determinations by LIGO/Virgo for BH-BH
stellar remnant at lower metallicity. As a result, a pop- (15.3 − 38.8 Gpc−3 yr−1 ), NS-NS (80 − 810 Gpc−3 yr−1 )
ulation of DCO containing a higher fraction of objects and BH-NS (≤ 610 Gpc−3 yr−1 ) are also reported (see
originating from low metallicity progenitors will result Abbott et al. 2019; Abbot et al. 2020b). The NS-NS
in a chirp mass distribution with a more extended high and BH-NS merging rates fall inside the LIGO/Virgo in-
mass tail (e.g. Fig. 4 in Chruslinska et al. 2019). terval for all the 3 cases considered, the BH-BH merging
4.2. Merging rates computation rates are slightly above the LIGO/Virgo interval for the
GSMF+FMR and GSMF+MZR cases, while they fall in-
The merging rates per chirp mass units are computed
as in Eq.(1) and, performing a further integration over
 Evolution of galaxy star formation and metallicity: impact on double compact objects mergers 13

Fig. 9.— Top panel: merger rate density of double compact objects binaries as a function of redshift, computed using the GSMF as
galaxy statistics and the MZR, following Eq.(10). Blue lines refers to BH-BH, red lines to NS-NS, green lines to BH-NS events. The red
and blue patches and the green arrow at z ∼ 0 represents the LIGO/Virgo 90% confidence intervals on the local rates for NS-NS and
BH-BH and the upper limit for BH-NS after the O1, O2 and first half of O3 runs (Abbott et al. 2019; Abbott et al. 2020). Bottom panel:
differential merging rate log (dṄ/dV d log M) for the BH-BH case (color code) as a function of redshift and chirp mass. Note the change
in the color code scale due to the larger number of BH-BH mergers occurring in this case.
side for the SFRF+FMR case8 . Comparing the top pan- lies in between. In fact, even if the metallicity, assigned
els of Fig. 7 and 9, referring to the GSMF+FMR and through the FMR, stays rather high suppressing BH-BH
GSMF+MZR cases, which uses the same galaxy statis- mergers, this fact is partially compensated by the higher
tics as a starting point (GSMF) and differ only in the cosmic SFR density at high redshift obtained when the
metallicity prescriptions, we can notice that the merging SFRF are employed as galaxy statistics (see Fig. 2). This
rates of NS-NS and BH-NS are similar, since they are less is also reflected in the larger merging rates for NS-NS and
dependent on metallicity. On the other hand, the BH-BH BH-NS in the SFRF+FMR case.
merging rates, which are strongly dependent on metallic- The bottom panels of Figs. 7, 8 and 9 are also
ity, are substantially different: at low redshift (z < 1.5) rather informative. They show the redshift and chirp
they are similar due to the similar behaviour of the metal- mass distribution of the BH-BH merging rates. In the
licity distribution, while at high redshift (z ≥ 1.5) they GSMF+MZR case (Fig. 9) the chirp mass distribution
are larger for the MZR case (by a maximum factor of extends up to M & 30 M at high redshift (z ≥ 2) where
∼ 10) due to the strong decrease of metallicity at high the metallicity tends to drop at subsolar values. This
redshift in the MZR case (see Figs. 4 and 6). As for high chirp mass tail is reduced for the cases in which
the SFRF+FMR case (Fig. 8) the BH-BH merging rates the FMR is used (Fig. 7 and 8), since metallicity never
drops too much, even at high redshifts, producing rem-
8 We stress again that the agreement/disagreement with the nants with lower masses on average. We note here that
LIGO/Virgo determinations can be due to the modelization of the none of the three cases analyzed is able to reproduce
stellar term. Moreover this is only one of the many constraint that the high chirp mass events (M & 30 M ) recently ob-
a galactic or stellar model should be able to satisfy. So, the local served at z < 1 by the LIGO/Virgo collaboration (see
rate alone does not represent a proof of the goodness of a model Abbott et al. 2020a; Abbott et al. 2020b). This is not
with respect to the others.
14 L. BOCO ET AL.

Fig. 10.— Top panel: differential merging rate log(dṄ/dV dtd ) at z ∼ 0 for BH-BH as a function of the time delay between the formation
of the binary and the merger, computed using the GSMF as galaxy statistics and the FMR, following Eq.(7). Bottom panels: differential
merging rates log(dṄ/dV d log M dtd ) at z ∼ 0 for BH-BH as a function of the chirp mass and time delay. On the x axis there is the time
delay, on the y axis the chirp mass and the color code represents the logarithmic number density of merging events. Contribution coming
from all the galaxies (left panel), from main sequence galaxies (top right panel) and starbursts (bottom right panel).

an issue for us, since even the chirp mass distribution, and starbursts, for the GSMF+FMR case (Fig. 7), and
as well as the total rates of DCOs mergers, are strongly the contribution of LTGs, ETGs and their star forming
dependent on the selected model of stellar and binary progenitors, for the SFRF+FMR case (Fig. 8). Between
evolution, whose discussion is out of the scope of the main sequence galaxies and starbursts, no evident dif-
current work. The comparisons shown here are useful ference can be found, it is clear just that main sequence
just to understand the general trend of DCOs mergers galaxies are the main contributors to the BH-BH merging
for different galactic prescriptions, but are not meant to rates. This is clearly dependent on the way we choose to
reproduce the real chirp mass distribution. However, we model starbursts: we fixed their fraction to be ∼ 3% for
stress that, in the three cases shown here, events with all the stellar masses at all redshifts; it would be interest-
M & 30 M are still produced, simply their rate is much ing to see how this would change treating the starburst
less than the rate of M < 30 M events. A GW de- fraction in a more detailed way (see Chruslinska et al.
tector as AdvLIGO/Virgo would tend to detect mainly in prep.). Instead, the contribution of LTGs and ETGs
high chirp mass events since they produce stronger GW is substantially different: LTGs contribute to the merg-
signals, so that the chirp mass distribution of merging ing rates only at low redshift (z ≤ 2) while only ETG
DCOs may substantially be altered by selection effects. progenitors are present at higher redshifts. At z ≤ 2 the
In addition, other channels of GW emission should not relatively longer tail towards larger chirp masses of LTGs
be excluded: dynamical formation and merger of com- can be explained by the fact that they have, on average,
pact object binaries (see e.g. Boco et al. 2020) as well lower metallicities (see Fig. 5).
as primordial black holes mergers (see e.g. Scelfo et al.
2018) could somewhat contribute to the GW detections 4.3. Chirp mass and time delay
and change the detected chirp mass distribution. The differential merging rates as a function of chirp
In the small plots on the bottom right of Figs. 7 and mass and time delay tell us how the time delays and chirp
8 it is shown the contribution of main sequence galaxies masses are distributed for the merging events. They can
 Evolution of galaxy star formation and metallicity: impact on double compact objects mergers 15

Fig. 11.— Top panel: differential merging rate log(dṄ/dV dtd ) at z ∼ 0 for BH-BH as a function of the time delay between the formation
of the binary and the merger, computed using the SFRF as galaxy statistics and the FMR, following Eq.(9). Bottom panels: differential
merging rate log(dṄ/dV d log M dtd ) at z ∼ 0 for BH-BH as a function of the chirp mass and time delay. On the x axis there is the time
delay, on the y axis the chirp mass and the color code represents the logarithmic number density of merging events. Contribution coming
from all the galaxies (left panel), from LTGs (top right panel) and ETG progenitors (bottom right panel).

be very helpful even for host galaxy association, since, t−1

d favouring short time delays, while the peak at td ∼
given the chirp mass of the signal, they give informations 10 − 12 Gyr is due to the huge amount of star formation
on the average age of the stellar population producing the happening at redshift z ∼ 2 − 3 that compensates for
merger. the time delay distribution favouring short time delays:
They can be computed by just avoiding the first inte- a small fraction of the many objects formed at z ∼ 2 − 3
gration in Eq.(1): can be seen through GW emission at z ∼ 0.
Z Apart from this shape shared by all the three cases,
dṄ dN dṀSFR there are some differences among them that we explain
(t, td ) = dZ (Z) (t−td )
dV dM dtd dMSFR dM dtd dV dZ in the following. In the cases where the GSMF is used
(12) as statistics (Figs. 10 and 12) the NS-NS time de-
We presents results for BH-BH mergers at z ∼ 0. The lay distribution is similar since NS-NS mergers are al-
results are shown in Figs. 10, 11 and 12 for the three dif- most independent on metallicity. The BH-BH time de-
ferent ways to compute the galactic term described in this lay distribution is instead rather different: while for the
paper: GSMF+FMR (Eq.(7)), SFRF+FMR (Eq.(9)) GSMF+FMR case (Fig. 10) the time delay distribu-
and GSMF+MZR (Eq.(10)). tion is flatter, with ∼ 48% of the BH-BH merging with
The top panels illustrate the merging rates per units td ≤ 6 Gyr, for the GSMF+MZR case the second peak is
of time delay dṄ/dV/dtd , meaning that Eq. (12) has more pronounced, with only ∼ 20% of BH-BH merging
been integrated over the chirp mass. In the GSMF+FMR with td ≤ 6 Gyr and many events with td ≥ 9 − 10 Gyr.
case we show the contribution of starburst galaxies, while This is due to the fact that, as already seen, in the MZR
in the SFRF+FMR case we show the contribution of case the metallicity is much lower at high redshift, in-
LTGs. In all the Figs. it is clearly visible a double peak creasing the contribution to the z ∼ 0 merging events
distribution: the peak at low delay times is due to the from BHs formed at high redshift. For the BH-NS merg-
shape of the intrinsic time delay distribution dp/dtd ∝ ers the same effect, even if milder, can be noted. The
16 L. BOCO ET AL.

Fig. 12.— Top panel: differential merging rate log(dṄ/dV dtd ) at z ∼ 0 for BH-BH as a function of the time delay between the formation
of the binary and the merger, computed using the GSMF as galaxy statistics and the MZR, following Eq.(10). Bottom panel: differential
merging rate log(dṄ/dV d log M dtd ) at z ∼ 0 for BH-BH as a function of the chirp mass and time delay. On the x axis there is the time
delay, on the y axis the chirp mass and the color code represents the logarithmic number density of merging events.
starbursts contribution, shown only in the GSMF+FMR is due to the larger amount of cosmic SFR, even if the
case, is subdominant. Comparing the NS-NS and BH- metallicity remains pretty high. The contribution at low
BH mergers in starbursts it can be noted a slight differ- time delays comes almost exclusively from LTGs, which
ence with respect to the all galaxies case. In fact, while are less metallic, as seen in Fig. 5, while events with large
the NS-NS contribution is always larger than the BH-BH time delays come from ETGs that formed stars at higher
one in the all galaxies case, if we restrict to starbursts redshifts, producing the second peak at td > 10 Gyr.
the two contribution are roughly comparable, with BH- The bottom panels show also the dependence on the
BH events being even dominant with respect to NS-NS chirp mass. It can be seen that in the GSMF+MZR
for td > 11 Gyr; this is due to the average lower metallic- case high chirp mass events tend to have huge time de-
ities of starbursts that slightly enhances the occurrence lays, while the distribution for the GSMF+FMR case is
of BH-BH mergers. smoother. This means that in the GSMF+MZR case
In the SFRF+FMR (Fig. 11) case NS-NS have a sim- the GW events at z ∼ 0 with M ≥ 20 M can be clearly
ilar shape to the other cases for td ≤ 9 − 10 Gyr, while linked to long delay times (& 10 Gyr) and so to an older
there is an enhancement at larger time delays, due to the stellar population, while in the GSMF+FMR case the
higher cosmic SFR at z ≥ 2. The contribution of LTGs association between chirp mass and time delay is much
to the NS-NS merging rates follows the relative abun- less clear. The SFRF+FMR case lies in between. In the
dance of LTGs with respect to ETGs with the cosmic bottom right small panels of Figs. 10 and 11 it is shown,
time. For the BH-BH mergers the shape is in between the respectively, the contribution of main sequence galaxies
GSMF+FMR and GSMF+MZR case: it can be seen a and starbursts and of LTGs and ETGs. Between main
decrease at 1 ≤ td ≤ 6 Gyr and a moderate enhancement sequence and starburst galaxies differences are not so ev-
at td ≥ 10 Gyr, with a resulting ∼ 37% of the BH-BH ident, due to our treatment of the starburst population,
mergers having td < 6 Gyr. The behaviour at small time while between LTGs and ETGs the difference is huge:
delays can be explained by the rather high metallicity at as already seen the ETGs clearly contribute mostly to
low redshift, and the enhancement at high time delays events with large time delays (td > 9 Gyr) and LTGs to
 Evolution of galaxy star formation and metallicity: impact on double compact objects mergers 17

the events with td < 9 Gyr. High chirp mass events can • the merging rates computed with the different
come from both the populations. galactic terms are roughly consistent with the lo-
cal merging rates determined by the LIGO/Virgo
5. CONCLUSIONS team (see Section 4);
Throughout the paper we focused on the computa-
tion of the cosmic SFR density per unit metallicity • differences in the merging rate shape are present
(dṀSFR /dV d log Z) with different prescriptions for the especially at high redshift where the two galaxy
galaxy statistics and for the metallicity scaling relations. statistics and the two metallicity scaling relations
In particular: are more different (see Section 4 and Figs. 9, 7 and
8). In short, using the SFRF enhances the com-
• we have shown the similarities and differences of pact remnants production at z > 2 with respect
using the stellar mass functions and the SFR func- to the GSMF case. The metallicity relation used
tions as galaxy statistics, finding a good agreement affects mainly the BH-BH merging rates and, in
between the two methods up to z ∼ 2 and a larger particular, an higher metallicity in the early uni-
cosmic SFR density at z > 2 if the SFR functions verse (obtained through the extrapolation of the
are employed (by a maximum factor of ∼ 2.5) (see FMR) hampers the BH-BH merging events even
Section 2 and Fig. 2). We have also discussed by a factor ∼ 10;
the main advantages and drawbacks of the two ap-
proaches: on the one hand the SFR functions pro-
vide a more direct statistics of the SFR of galaxies, • differences are present also in the chirp mass and
on the other hand the GSMF can be more useful time delay distributions (see Section 4 and Figs.
in the estimation of the galaxies’ metallicities (see 7, 8, 9, 10, 11 and 12). In short if there is little
discussion at the end of Section 2 and Section 3). metallicity evolution with redshift, the association
between the chirp mass of the GW event and the
• we have presented the two main empirical scaling redshift at which the merging occurs is less clear
relations to associate metallicities to galaxies: the than in the case in which there is a strong metal-
Mass-Metallicity Relation and the Fundamental licity evolution.
Metallicity Relation. We analyzed the similarities
and differences between the two relations, showing However, we remark again that the results on the merg-
that the extrapolation of the FMR yields rather ing rates are also dependent on the selected stellar model,
large average metallicity values (Z ∼ 0.4 − 0.5 Z ) so they should be intended just as a case study to com-
even at z > 2, while the MZR usually implies very pare the effects of the different galactic terms.
low metallicities (Z < 0.1 Z ) at z > 2. We have We hope that this work on different prescriptions for
brought theoretical arguments and recent observa- the galaxy statistics and metallicity can help in under-
tional evidences testifying that the metallicity of standing the main properties of the merging binaries that
high redshift dusty obscured star forming galaxies will be detected through GW, especially with the future
is rather large, arguing that, in order to reproduce third generation detectors as Einstein Telescope. Know-
those metallicities we should rely on extrapolations ing the effect of different galactic properties on the fea-
of the FMR or of a slowly evolving MZR (see Sec- tures of the merging binaries can be extremely helpful in
tion 3). order to better understand the star formation and galaxy
evolution across cosmic time, when a large statistics of
• we have combined our fiducial scaling relation (the GW events will become available. From the observa-
FMR) with both the two aforementioned galaxy tional point of view, a huge boost in the characteriza-
statistics to compute the cosmic SFR density per tion of galaxies star formation rate and metallicity at
units of metallicity (dṀSFR /dV d log Z) in the two increasingly high redshifts will come with the advent of
cases (see Section 3 and Figs. 4 and 5). We JWST, which is suitable for measuring line diagnostics
have also considered an alternative case in which from galaxies across a broad range of redshifts, eventu-
a sharply evolving MZR is used and combined it ally in synergy with (sub)mm instruments like ALMA.
with the GSMF (see Fig. 6)). We find that the We thank the anonymous referee for useful comments.
differences in the employed galaxy statistics and We acknowledge financial support from the EU H2020-
metallicity evolution are clearly reflected in the fac- MSCAITN-2019 Project 860744 ‘BiD4BEST: Big Data
tor dṀSFR /dV d log Z. applications for Black hole Evolution STudies’ and from
the PRIN MIUR 2017 prot. 20173ML3WW 002, ‘Open-
Finally, in the last section, we have chosen a stellar ing the ALMA window on the cosmic evolution of gas,
and binary evolution model as an example to show the stars and massive black holes’. L.B. warmly thanks Luigi
effect of the different galactic terms on the merging rates Bassini, for many helpful discussions. MC acknowledges
and on the properties of the merging binaries. We find support from the Netherlands Organisation for Scientific
that: Research (NWO).
REFERENCES
Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2016a, PhRvL, Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2016c, PhRvX, 6,
116, 241103 041015
Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2016b, PhRvL, Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2017a, ApJL, 851,
116, 061102 L35
18 L. BOCO ET AL.

Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2017b, ApJL, 848, Daddi, E., Alexander, D. M., Dickinson, M., et al. 2007, ApJ, 670,
L12 173
Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2017c, PhRvL, Davé, R., Anglés-Alcázar, D., Narayanan, D., et al. 2019, MNRAS,
119, 161101 486, 2827
Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2017d, PhRvL, Davé, R., Rafieferantsoa, M. H., Thompson, R. J., & Hopkins, P.
119, 141101 F. 2017, MNRAS, 467, 115
Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2017e, PhRvL, Davidzon, I., Ilbert, O., Laigle, C., et al. 2017, A&A, 605, A70
118, 221101 de Mink, S. E., & Belczynski, K. 2015, ApJ, 814, 58
Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2019, PhRvX, 9, de Mink, S. E., Langer, N., Izzard, R. G., Sana, H., & de Koter,
031040 A. 2013, ApJ, 764, 166
Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2020a, Dominik, M., Belczynski, K., Fryer, C., et al. 2012, ApJ, 759, 52
arXiv:2010.14527 Dominik, M., Berti, E., O’Shaughnessy, R., et al. 2015, ApJ, 806,
Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2020b, 263
arXiv:2010.14533 Donevski, D., Lapi, A., Malek, K., et al. 2020, arXiv:2008.09995
Alavi, A., Siana, B., Richard, J., et al. 2016, ApJ, 832, 56 Dunlop, J. S., McLure, R. J., Biggs, A. D., et al. 2017, MNRAS,
Andrews, B. H., & Martini, P. 2013, ApJ, 765, 140 466, 861
Artale M.C., Mapelli, M., Giacobbo, N., et al. 2019, MNRAS 487, Dvorkin, I., Uzan, J.-P., Vangioni, E., & Silk, J. 2018, MNRAS,
1675 479, 121
Asano, R. S., Takeuchi, T. T., Hirashita, H., & Inoue, A. K. 2013, Efstathiou, A., Rowan-Robinson, M., & Siebenmorgen, R. 2000,
Earth Planets and Space, 65, 213 MNRAS, 313, 734
Barrett, J.W., Gaebel, S.M., Neijssel, C.J., et al. 2018, MNRAS, Elbert, O. D., Bullock, J. S., & Kaplinghat, M. 2018, MNRAS, 473,
477, 4685 1186
Belczynski, K., Dominik, M., Bulik, T., et al. 2010, ApJL, 715, Eldridge, J.J., Stanway, E.R., Xiao, L., et al. 2017, PASA, 34, e058
L138 Eldridge, J.J. ,& Stanway, E.R. 2016, MNRAS, 462, 3302
Belczynski, K., Bulik, T., Fryer, C.L., et al. 2010, ApJ, 714, 1217 Finkelstein, S. L., Ryan, R. E., Jr., Papovich, C., et al. 2015, ApJ,
Belczynski, K., Holz, D. E., Bulik, T., & O’Shaughnessy, R. 2016, 810, 71
Natur, 534, 512 Fishbach, M., Gray, R., Magana Hernandez, I., et al. 2019, ApJ,
Béthermin, Matthieu; Daddi, Emanuele; Magdis, Georgios, et al. 871, L13
2015, A&A, 573, A113 Fudamoto, Y., Oesch, P. A., Schinnerer, E., et al. 2017, MNRAS,
Béthermin, M., Daddi, E., Magdis, G., et al. 2012, ApJL, 757, L23 472, 483
Bhatawdekar, R., Conselice, C., Margalef-Bentabol, B., & Duncan, Gallazzi, A., Charlot, S., Brinchmann, J., & White, S. D. M. 2006,
K. 2019, MNRAS, 486, 3805 MNRAS, 370, 1106
Bisigello, L., Caputi, K.I., Grogin, N., & Koekemoer, A. 2018, Gallazzi, A., Bell, E. F., Zibetti, S., Brinchmann, J., & Kelson, D.
A&A, 609, 82 D. 2014, ApJ, 788, 72
Boco, L., Lapi, & Danese, L. 2020, ApJ, 891, 94 Genzel, R., Tacconi, L. J., Lutz, D., et al. 2015, ApJ, 800, 20
Boco, L., Lapi, A., Goswami, S., et al. 2019, ApJ, 881, 157 Giacobbo, N., & Mapelli, M. 2018, MNRAS, 480, 2011
Boogaard, L. A., Decarli, R., González-López, J., et al. 2019, ApJ, Giacobbo, N., Mapelli, M., & Spera, M. 2018, MNRAS, 474, 2959
882, 140 Ginolfi, M., Jones, G. C., Béthermin, M., et al. 2020, A&A, 633,
Boquien, M., Burgarella, D., Roehlly, Y., et al. 2019, A&A, 622, A90
103 Grisoni, V., Spitoni, E., Matteucci, F., et al. 2017, MNRAS, 472,
Bouwens, R. J., Oesch, P. A., Illingworth, G. D., Ellis, R. S., & 3637
Stefanon, M. 2017, ApJ, 843, 129 Gruppioni C., Bethermin, M., Loiacono, F., et al. 2020, A&A, 643,
Bouwens, R. J., Aravena, M., De Carli, R., et al. 2016, ApJ, 833, A8
72 Gruppioni, C., & Pozzi, F. 2019, MNRAS, 483, 1993
Bouwens, R. J., Illingworth, G. D., Oesch, P. A., et al. 2015, ApJ, Gruppioni, C., Calura, F., Pozzi, F., et al. 2015, MNRAS, 451, 3419
803, 34 Gruppioni, C., Pozzi, F., Rodighiero, G., et al. 2013, MNRAS, 432,
Caffau, E., Ludwig, H. G., Steffen, M., Freytag, B., & Bonifacio, 23
P. 2011, solphys, 268, 255 Hunt, L., Dayal, P., Magrini, L., & Ferrara, A. 2016, MNRAS, 463,
Calore, F., Cuoco, A., Regimbau, T., Sachdev, S., Serpico, P. D., 2002
PhRvR, 2, 023314 Hurley, J.R., Pols, O.R., & Tout, C.A. 2000, MNRAS, 315, 543
Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682 Ilbert, O., Arnouts, S., Le Floc’h, E., et al. 2015, A&A, 579, 2
Cao, L., Lu, Y., & Zhao, Y. 2018, MNRAS, 474, 4997 Ilbert, O., McCracken, H. J., Le Févre, O., et al. 2013, A&A, 556,
Caputi, K. I., Deshmukh, S., Ashby, M. L. N., et al. 2017, ApJ, 55
849, 45 Johansson, J., Thomas, D., & Maraston, C. 2012, MNRAS, 421,
Casey, M.C., Zavala, J.A., Spilker, J., et al. 2018, ApJ, 862, 67 1908
Cassará, L. P., Maccagni, D., Garilli, B., et al. 2016, A&A, 593, Jones, T., Sanders, R., Roberts-Borsani, G., et al 2020, ApJ, 903,
A9 150
Chabrier, G. 2003, ApJL, 586, L133 Juarez, Y., Maiolino, R., Mujica, R., et al. 2009, A&A, 494, L25
Chabrier, G. 2005, in The Initial Mass Function 50 Years Later, Kennicutt, R. C., & Evans, N. J. 2012, ARA&A, 50, 531
Vol. 327, ed. E. Corbelli & F. Palle (Dordrecht: Springer), 41 Kennicutt, R. C. 1998, ApJ, 498, 541
Chiappini, C., Matteucci, F., & Gratton, R. 1997, ApJ, 477, 765 Kewley, L. J., & Ellison, S. L. 2008, ApJ, 681, 1183
Chruslinska, M., Nelemans, G., & Belczynski, K., 2019, MNRAS, Klencki, J., Moe, M., Gladysz, W., et al. 2018, A&A, 619, A77
482, 5012 Lamberts, A., Garrison-Kimmel, S., Clausen, D. R., & Hopkins, P.
Chruslinska, M., & Nelemans, G. 2019, MNRAS, 488, 5300 F. 2016, MNRAS, 463, L31
Chruslinska, M., Belczynski, K., Klencki, J., & Benacquista, M. Lamberts, A., Garrison-Kimmel, S., Hopkins, P. F., et al. 2018,
2018, MNRAS, 474, 2937 MNRAS, 480, 2704
Citro, A., Pozzetti, L., Moresco, M., & Cimatti, A. 2016, A&A, Langer, N., & Norman, C. A. 2006, ApJL, 638, L36
592, A19 Lapi, A., Pantoni, L., Boco, L., & Danese, L., ApJ, 897, 81
Coppin, K. E. K., Geach, J. E., Almaini, O., et al. 2015, MNRAS, Lapi, A., Gonzalez-Nuevo, J., Fan, L., et al. 2011, ApJ, 742, 24
446, 1293 Li, J., Wang, R., Cox, P., et al. 2020, ApJ, 900, 131
Courteau, S., Cappellari, M., de Jong, R. S., et al. 2014, RvMP, Li, S. S., Mao, S., Zhao, Y., & Lu, Y. 2018, MNRAS, 476, 2220
86, 47 Liao, K., Fan, X.-L., Ding, X., Biesiada, M., & Zhu, Z.-H. 2017,
Creminelli, P., & Vernizzi, F., PhRvL, 25, 251302 NatCo, 8, 1148
Cucciati, O., Tresse, L., Ilbert, O., et al. 2012, A&A, 539, A31 Liu, D., Schinnerer, E., Groves, B., et al. 2019, ApJ, 887, 235
Curti, M., Mannucci, F., Cresci, G., Maiolino, R. 2020, MNRAS, Liu, D., Daddi, E., Dickinson, M., et al. 2018, ApJ, 853, 172
491, 944 Ma, X., Hopkins, P. F., Faucher-Giguére, C. A., et al. 2016,
da Cunha, E., Walter, F., Smail, I. R., et al. 2015, ApJ, 806, 110 MNRAS, 456, 2140
da Cunha, E., Charlot, S., & Elbaz, D. 2008, MNRAS, 388, 1595 Madau, P., & Fragos, T. 2017, ApJ, 840, 39
 Evolution of galaxy star formation and metallicity: impact on double compact objects mergers 19

Madau, P., & Dickinson, M. 2014, ARA&A, 52, 415 Sanders, R. L., Shapley, A. E., Reddy, N. A., et al. 2020, MNRAS,
Maeder, A. 1992, A&A, 264, 105 491, 1427
Magnelli, B., Popesso, P., Berta, S., et al. 2013, A&A, 553, A132 Santoliquido, F., Mapelli, M., Giacobbo, N., Bouffanais, Y., &
Magnelli, B., Lutz, D., Berta, S., et al. 2012, A&A, 548, A22 Artale, M. C. 2020, arXiv:2009.03911
Maiolino, R., Mannucci, F. 2019, A&AR, 27, 3 Saracco, P., Marchesini, D., La Barbera, F., et al. 2020,
Maiolino, R., Nagao, T., Grazian, A., et al. 2008, A&A, 488, 463 arXiv:2011.04657
Mancuso, C., Lapi, A., Shi, J., et al. 2016a, ApJ, 823, 128 Sargent, M. T., Béthermin, M., Daddi, E., & Elbaz, D. 2012, ApJL,
Mancuso, C., Lapi, A., Shi, J., et al. 2016b, ApJ, 833, 152 747, L31
Mannucci, F., Salvaterra, R., Campisi, M. A. 2011, MNRAS, 414, Savaglio, S., Glazebrook, K., Le Borgne, D., et al. 2005, ApJ, 635
1263 260
Mannucci, F., Cresci, G., Maiolino, R., Marconi, A., & Gnerucci, Scelfo, G., Boco, L., Lapi, A., Viel, M. 2020, JCAP, 2020, 10, 045
A. 2010, MNRAS, 408, 2115
Scelfo, G., Bellomo, N., Raccanelli, A., Matarrese, S., & Verde, L.
Mannucci, F., Cresci, G., Maiolino, R., et al. 2009, MNRAS, 398,
2018, JCAP, 2018, 9, 039
1915
Schreiber, C., Pannella, M., Elbaz, D., et al. 2015, A&A, 575, 74
Maoz, D., Mannucci, F., & Nelemans, G. 2014, ARA&A, 52, 107
Mapelli, M., & Giacobbo, N. 2018, MNRAS, 479, 4391 Silva, L., Granato, G. L., Bressan, A., & Danese, L. 1998, ApJ,
Mapelli, M., Giacobbo, N., Ripamonti, E., & Spera, M. 2017, 509, 103
MNRAS, 472, 2422 Shapley, A. E., Cullen, F., Dunlop, J. S., et al. 2020, ApJL, 903,
Marrone, D. P., Spilker, J. S., Hayward, C. C., et al. 2018, Natur, L16
553, 51 Smail, I., Dudzevičiūtė, U., Stach, S. M., et al. 2020,
Meurer, G. R., Heckman, T. M., & Calzetti, D. 1999, ApJ, 521, 64 arXiv:2010.02250
Mo, H., van den Bosch, F., & White, S. D. M. 2010, Galaxy Smit, R., Bouwens, R. J., Franx, M., et al. 2012, ApJ, 756, 14
Formation and Evolution (Cambridge: Cambridge Univ. Press) Speagle, J. S., Steinhardt, C. L., Capak, P. L., & Silverman, J.
Morishita, T., Abramson, L. E., Treu, T., et al. 2019, ApJ, 877, 2014, ApJS, 214, 15
141 Spera, M., Mapelli, M., Giacobbo, N., et al. 2019, MNRAS, 485,
Moustakas, J., Coil, A. L., Aird, J., et al. 2013, ApJ, 767, 50 889
Muzzin, A., Marchesini, D., Stefanon, M. 2013, ApJ, 777, 18 Spera, M., & Mapelli, M. 2017, MNRAS, 470, 4739
Neijssel, C.J., Vigna-Gomez, A., Stevenson S., et al. 2019, MNRAS, Stacey, H. R., McKean, J. P., Robertson, N. C., et al. 2018,
490, 3740 MNRAS, 476, 5075
Nissanke, S., Holz, D. E., Dalal, N., et al. 2013, arXiv:1307.2638 Steinhardt, C. L., Speagle, J. S., & Capak, P. 2014, ApJL, 791,
Novak, M., Bañados, E., Decarli, R., et al. 2019, ApJ, 881, 63 L25
Novak, M., Smolcic, V., Delhaize, J., et al. 2017, A&A, 602, 5 Stevenson, S., Vigna-Gómez, A., Mandel, I., et al. 2017, NatCo, 8,
O’Shaughnessy, R., Bellovary, J. M., Brooks, A., et al. 2017, 14906
MNRAS, 464, 2831 Strait, V., Bradac, M., Coe, D., et al. 2020, arXiv:2009.00020
O’Shaughnessy, R., Kalogera, V., & Belczynski, K. 2010, ApJ, 716, Sundqvist, J.O., Björklund, R., Puls, J., & Najarro, F. 2019, A&A,
615 632, A126
Oesch, P. A., Bouwens, R. J., Carollo, C. M., et al. 2010, ApJL, Tacconi, L. J., Genzel, R., Sternberg, A., ARA&A, 58, 157
725, L150 Tan, Q., Daddi, E., Magdis, G., et al. 2014, A&A, 569, A98
Oguri, M. 2016, PhRvD, 93, 083511 Taylor, S. R., & Gair, J. R. 2012, PhRvD, 86, 023502
Onodera, M., Carollo, C. M., Lilly, S., et al. 2016, ApJ, 822, 42 Thomas, D., Maraston, C., Bender, R., & Mendes de Oliveira, C.
Onoue, M., Bañados, E., Mazzucchelli, C., et al. 2020, ApJ, 898, 2005, ApJ, 621, 673
105 Thomas, D., Maraston, C., Schawinski, K., Sarzi, M., & Silk, J.
Pantoni, L., Lapi, A., Massardi, M., Goswami, S., & Danese, L. 2010, MNRAS, 404, 1775
2019, ApJ, 880, 129 Tomczak, A. R., Quadri, R. F., Tran, K. H., et al. 2014, ApJ, 783,
Papovich, C., Finkelstein, S. L., Ferguson, H. C., Lotz, J. M., & 85
Giavalisco, M. 2011, MNRAS, 412, 1123 Torrey, P., Vogelsberger, M., Hernquist, L., et al. 2018, MNRAS,
Pezzulli, G., & Fraternali, F. 2016, MNRAS, 455, 2308 477, L16
Planck Collaboration 2019, A&A, 641, A6 Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al. 2004, ApJ,
Popping, G., Somerville, R. S., & Galametz, M. 2017, MNRAS, 613, 898
471, 3152
van der Burg, R. F. J., Hildebrandt, H., & Erben, T. 2010, A&A,
Portegies Zwart, S.F. ,& Yungelson, L.R. 1998, A&A, 332, 173
523, A74
Raccanelli, A., Kovetz, E. D., Bird, S., Cholis, I., & Muñoz, J. B.
Venemans, B. P., Decarli, R., Walter, F., et al. 2018, ApJ, 866, 159
2016, PhRvD, 94, 023516
Radice, David, Perego, Albino, Zappa, Francesco, & Bernuzzi, Venemans, B. P., Walter, F., Decarli, R., et al. 2017, ApJL, 851,
Sebastiano 2018, ApJL, 852, L29 L8
Reddy, N. A., Kriek, M., Shapley, A. E., et al. 2015, ApJ, 806, 259 Vijayan, A. P., Clay, S. J., Thomas, P. A., et al. 2019, MNRAS,
Regimbau, T., Evans, M., Christensen, N., et al. 2017, PhRvL, 118, 489, 4072
151105 Vink, J.S. ,& de Koter, A. 2005, A&A, 442, 587
Riechers, D. A., Leung, T. K. D., Ivison, R. J., et al. 2017, ApJ, Vink, J.S., de Koter, A., & Lamers H.J.G.L.M. 2001, A&A, 369,
850, 1 574
Rodighiero, G., Brusa, M, Daddi, E., et al. 2015, ApJL, 800, L10 Wang, T., Schreiber, C., Elbaz, D., et al. 2019, Natur, 572, 211
Rodighiero, G., Daddi, E., Baronchelli, I., et al. 2011, ApJL, 739, Weinberg, D. H., Andrews, B. H., & Freudenburg, J. 2017, ApJ,
L40 837, 183
Romano, D., Silva, L., Matteucci, F., & Danese, L. 2002, MNRAS, Whitaker, K. E., Franx, M., Leja, J., et al. 2014, ApJ, 795, 104
334, 444 Wyder, T. K., Treyer, M. A., Milliard, B., et al. 2005, ApJL, 619,
Rowan-Robinson, M., Oliver, S., Wang, L., et al. 2016, MNRAS, L15
461, 1100 Yang, T., Hu, B., Cai, R., & Wang, B. 2019, ApJ, 880, 50
Sander, A.A.C., Vink, J.S., & Hamann, W.-R. 2020, MNRAS, 491, Zahid, H. J., Kashino, D., Silverman, J. D., et al. 2014, ApJ, 792,
4406 75
Sanders, R. L., Shapley, A. E., Jones, T., et al. 2020, Zavala, J. A., Montana, A., Hughes, D. H., et al. 2018, NatAs, 2,
arXiv:2009.07292 56