GAMETE/QSOdust: High-Redshift Quasar Evolution

• GAMETE/QSOdust is a semi-analytical, hierarchical model that reconstructs dark matter halo merger histories for high-redshift quasar hosts using a binary Monte Carlo algorithm.
• It couples star formation, chemical enrichment, and dust processing, revealing that grain growth in molecular clouds is essential to reproduce observed dust masses.
• The model integrates black-hole seeding, accretion, and AGN feedback to explain rapid SMBH growth and the dynamic interplay between star formation and quasar activity.

Searching arXiv for GAMETE/QSOdust papers to ground the article in the cited literature. GAMETE/QSOdust is a semi-analytical model for the formation and evolution of high-redshift quasars and their host galaxies. In its original applications, it was constructed to reconstruct hierarchical merger histories of a $10^{13}\,M_\odot$ dark matter halo appropriate for SDSS J1148+5251 at $z=6.4$, while self-consistently following the assembly of the central supermassive black hole, the build-up of the host galaxy, the evolution of gas, stars, metals, and dust in the interstellar medium, and the action of stellar and AGN feedback (Valiante et al., 2011). Later versions retained this architecture while extending the treatment of black-hole seed formation, super-Eddington growth, and radiative feedback, including prescriptions for Pop III remnants, disk winds, and dust attenuation of Lyman-Werner radiation (Pezzulli et al., 2017).

1. Model architecture and cosmological setting

GAMETE/QSOdust is a hierarchical, cosmological model implemented through merger trees generated by a binary Monte Carlo algorithm based on Extended Press-Schechter theory. In the applications to SDSS J1148+5251, the model reconstructs the assembly histories of a $10^{13}\,M_\odot$ halo from $z=37$ to $z=6.4$, resolving progenitors down to the atomic-cooling threshold $T_{\rm vir}=10^4$ K. The resulting branches are treated as galaxies that inherit and exchange gas, stars, metals, dust, and black holes during hierarchical growth (Valiante et al., 2011).

The model was designed to follow multiple independent realizations of halo assembly. In the J1148 applications, 50 hierarchical merger histories are generated; in the study of sustainable black-hole growth, $N_r=5$ realizations are evolved for each model. This difference reflects application-specific sampling rather than a change in the underlying framework. In both cases, the final host halo is taken to be $M_h=10^{13}\,M_\odot$ at $z=6.4$, consistent with the expected environment of bright $z\sim 6$ quasars [(Valiante et al., 2012); (Pezzulli et al., 2017)].

A defining feature of the framework is that it combines halo assembly with baryonic evolution rather than treating the quasar host as an isolated system. The model self-consistently tracks star formation in disk and bulge components, Pop III and Pop II star formation and chemical enrichment, black-hole seeding, black-hole mass growth through accretion and mergers, energy-driven AGN and supernova feedback, mergers with dynamical effects and gravitational-wave recoil, and two-phase ISM evolution with hot and cold components (Pezzulli et al., 2017). Parameters are tuned to reproduce observed quasar properties such as black-hole mass, star-formation rate, gas content, and outflow rate.

2. Star formation, chemical enrichment, and dust physics

The baryonic sector of GAMETE/QSOdust evolves gas, stars, metals, and dust through coupled differential equations that include star formation, stellar mass return, gas infall, outflow, and accretion onto the central black hole. A representative star-formation prescription is

$z=6.4$0

while merger-driven formulations use

$z=6.4$1

The burst efficiency is parameterized as a function of merger mass ratio, so the model distinguishes quiescent star formation from merger-induced starbursts [(Valiante et al., 2011); (Valiante et al., 2014)].

The Pop III to Pop II/I transition is controlled by a critical metallicity. When $z=6.4$2, Pop III stars form; otherwise a Pop II/I IMF is adopted, with $z=6.4$3 in the original J1148 implementation. Larson IMFs with different characteristic masses are explored, including a standard form with $z=6.4$4 and a top-heavy form with $z=6.4$5 (Valiante et al., 2011).

Dust evolution is treated explicitly rather than as a by-product of metal enrichment. The model includes dust formation in supernova ejecta and in the atmospheres of AGB stars, dust destruction by interstellar shocks, astration, and grain growth in molecular clouds. A representative dust-mass equation is

$z=6.4$6

The corresponding physical picture is two-phase: dust destruction is associated with the diffuse ISM, while rapid grain growth is associated with molecular clouds. The growth timescale is written as

$z=6.4$7

with $z=6.4$8 yr and typically $z=6.4$9 for the fraction of ISM dust in molecular clouds (Valiante et al., 2011).

A recurrent result across the J1148 studies is that stellar sources alone cannot reproduce the observed dust mass. Supernovae dominate early dust enrichment, AGB stars become important at $10^{13}\,M_\odot$0–10, but the observed dust reservoir requires grain growth in molecular clouds. In the specific J1148 applications, stellar sources yield at most $10^{13}\,M_\odot$1–$10^{13}\,M_\odot$2, whereas grain growth allows dust masses up to $10^{13}\,M_\odot$3, consistent with the inferred $10^{13}\,M_\odot$4–$10^{13}\,M_\odot$5 (Valiante et al., 2011).

3. Black-hole assembly, accretion, and feedback

In the original high-redshift quasar applications, GAMETE/QSOdust plants seed black holes of $10^{13}\,M_\odot$6 in high-$10^{13}\,M_\odot$7 peaks or in the earliest star-forming halos. Subsequent growth proceeds through black-hole mergers and gas accretion. In those versions, the gas accretion rate is the minimum of the Eddington-limited rate and the Bondi-Hoyle-Lyttleton rate, with a normalization parameter $10^{13}\,M_\odot$8 typically in the range 180–200 in best-fit J1148 models [(Valiante et al., 2011); (Valiante et al., 2012)].

AGN feedback is implemented as an energy-driven wind. The injected power is written as

$10^{13}\,M_\odot$9

and the corresponding gas ejection rate as

$z=37$0

In the J1148 studies, $z=37$1 is taken to be $z=37$2 as a fiducial value, $z=37$3, and AGN-driven winds are active when the accretion rate is super-critical, $z=37$4, corresponding to the quasar-mode phase (Valiante et al., 2012).

The model also includes supernova-driven winds, with the total outflow rate written as the sum of SN and AGN contributions. For J1148, the outflow rate predicted by GAMETE/QSOdust is in good agreement with the lower limit of $z=37$5 inferred from observations. In those best-fit models, AGN feedback dominates the outflow budget from $z=37$6 onward by more than two orders of magnitude over supernovae, and supernova explosions give a negligible contribution to the observed wind at $z=37$7 (Valiante et al., 2012).

This coupling between accretion and feedback is central to the model’s co-evolution scenario. Accreting black holes drive winds that deplete the ISM, suppress star formation, and self-regulate further accretion. Within GAMETE/QSOdust, the host enters a bright quasar phase when this feedback becomes dynamically important and the line of sight is progressively un-obscured by the clearing of gas and dust (Valiante et al., 2014).

4. Application to SDSS J1148+5251 and the $z=37$8 quasar population

SDSS J1148+5251 at $z=37$9 is the canonical calibration target for GAMETE/QSOdust. In the model, the host halo mass is $z=6.4$0, and the parameter set is adjusted to reproduce an SMBH mass of $z=6.4$1, an ISM gas mass of $z=6.4$2, a dust mass of $z=6.4$3–$z=6.4$4, and metallicity $z=6.4$5, while allowing a broad stellar-mass range of $z=6.4$6–$z=6.4$7 (Valiante et al., 2011).

The same framework was generalized to a sample of quasars at $z=6.4$8. In that application, the observed properties of the quasars are reproduced by a common formation scenario in which stars form according to a standard IMF, via quiescent star formation and efficient merger-driven bursts, while the central black hole grows via gas accretion and BH-BH mergers. Eventually, a strong AGN-driven wind starts to clear the ISM of dust and gas, damping star formation and un-obscuring the line of sight toward the quasar (Valiante et al., 2014).

A further use of GAMETE/QSOdust is as the physical backbone for radiative-transfer calculations. In the study of the far-infrared continuum of J1148, outputs of the semi-analytical model are passed to PEGASE for stellar SED synthesis and to the 3D Monte Carlo radiative transfer code TRADING. In that combined framework, radiation from the central source provides at least 30% and up to 70% of the observed far-infrared emission at rest-frame wavelengths $z=6.4$9 micron. The remaining fraction is contributed by stellar sources and can only be achieved if the host galaxy sustains a star-formation rate of $T_{\rm vir}=10^4$0 at $T_{\rm vir}=10^4$1 (Schneider et al., 2014).

These results have two direct interpretive consequences. First, the FIR luminosity of high-$T_{\rm vir}=10^4$2 quasars is not a pure star-formation tracer, because AGN-powered dust heating can be substantial. Second, the model favors a co-evolution sequence in which the SMBH and host first grow at comparable pace, then the black hole accelerates once it reaches a mass of $T_{\rm vir}=10^4$3, and the system enters the bright quasar phase while still lying within the scatter of the local scaling relation. In the larger $T_{\rm vir}=10^4$4 sample, about 40% of progenitor galaxies are predicted to be classifiable as Sub Millimeter Galaxies during the transition between the starburst-dominated and active QSO phases [(Schneider et al., 2014); (Valiante et al., 2014)].

5. Super-Eddington growth and the first black holes

A later version of GAMETE/QSOdust was used to study the sustainable growth of the first black holes and the formation of $T_{\rm vir}=10^4$5 SMBHs from light seeds. In that version, seed formation is tied to Pop III stellar remnants, with a minimum $T_{\rm vir}=10^4$6 seed per $T_{\rm vir}=10^4$7 burst, and the model introduces two new physical prescriptions: stellar feedback on Pop III black-hole formation sites and accretion-disk winds that limit the duration of super-Eddington episodes (Pezzulli et al., 2017).

The stellar-feedback prescription assumes that Pop III star formation expels all gas from minihalos, heating and ionizing the ISM and IGM. Cooling and reaccretion are suppressed until the host halo reaches $T_{\rm vir}=10^4$8 K or for 200 Myr after the first burst. This delays seed formation from $T_{\rm vir}=10^4$9 to $N_r=5$0 and shifts the dominant seed-formation sites from minihalos to more massive Ly$N_r=5$1-cooling halos. Within the model, however, this feedback does not prevent the eventual formation of $N_r=5$2 SMBHs (Pezzulli et al., 2017).

Super-Eddington accretion is parameterized as

$N_r=5$3

where $N_r=5$4 depends on mergers, $N_r=5$5 is the bulge gas mass, and $N_r=5$6 is the bulge dynamical time. The luminosity and radiative efficiency during supercritical accretion are taken from the relativistic slim-disk solutions of S$N_r=5$7dowski et al. (2009), as parameterized by Madau et al. (2014). Disk winds are assumed to halt accretion when the disk radius approaches the photon trapping radius, $N_r=5$8, with

$N_r=5$9

where $M_h=10^{13}\,M_\odot$0 Gyr, $M_h=10^{13}\,M_\odot$1 is the fractional angular momentum retained by the accreting gas, and $M_h=10^{13}\,M_\odot$2 is the host bulge velocity dispersion (Pezzulli et al., 2017).

Model	Disk-wind prescription	Outcome
NL	No disk winds	$M_h=10^{13}\,M_\odot$3 at $M_h=10^{13}\,M_\odot$4; unconstrained durations up to Myr
L001	$M_h=10^{13}\,M_\odot$5	$M_h=10^{13}\,M_\odot$6 by $M_h=10^{13}\,M_\odot$7; $M_h=10^{13}\,M_\odot$8–$M_h=10^{13}\,M_\odot$9 yr at $z=6.4$0
L01	$z=6.4$1	$z=6.4$2 at $z=6.4$3; $z=6.4$4–$z=6.4$5 yr at $z=6.4$6

The central conclusion of this extension is that sustainable growth from light seeds critically depends on efficient angular-momentum loss. For $z=6.4$7, super-Eddington duty cycles are so short that growth is strongly suppressed; for $z=6.4$8, early growth is still curtailed relative to no-wind models, but later gas retention enables rapid mass increase to $z=6.4$9 quasar-scale black holes. The same calculations also show that disk winds and short $z\sim 6$0 strongly reduce the active fraction of progenitor SMBHs, especially at $z\sim 6$1, making their direct observation unlikely (Pezzulli et al., 2017).

6. Interpretive tensions, observational implications, and later extensions

One of the main controversies associated with GAMETE/QSOdust is the so-called stellar mass crisis. In the common formation scenario that reproduces the observed properties of quasars at $z\sim 6$2, the final stellar masses of the hosts lie in the range $z\sim 6$3, which is a factor 3–30 larger than the upper limits allowed by the observations. The most likely explanation proposed in that study is that large uncertainties still affect dynamical mass measurements in these high-redshift galaxies; alternatives such as a top-heavy IMF, enhanced stellar dust yields, or very short starburst durations alleviate some aspects of the tension but create difficulties elsewhere in matching FIR luminosities, SMBH growth timescales, or dust masses (Valiante et al., 2014).

A related interpretive issue concerns the use of far-infrared luminosity as a star-formation indicator. Because the central source can power 30%–70% of the observed FIR continuum in J1148, star-formation rates inferred from FIR emission alone can be overestimated if AGN heating is neglected. In the GAMETE/QSOdust framework, FIR-bright quasars correspond to a rapidly evolving phase in which intense star formation and strong AGN activity coexist before feedback-driven quenching (Schneider et al., 2014).

A further extension is the inclusion of dust attenuation of Lyman-Werner feedback in an updated version of the model. In that implementation, the effective LW intensity is reduced according to $z\sim 6$4, with $z\sim 6$5. Even modest dust enrichment, with $z\sim 6$6–$z\sim 6$7, yields non-negligible LW optical depth and modifies the conditions for light, medium-weight, and heavy seed formation. In the reported calculations, dust shielding extends the redshift range for heavy-seed formation from $z\sim 6$8–20 to $z\sim 6$9–25 and increases the number of heavy seeds from $z=6.4$00 to $z=6.4$01 per $z=6.4$02 Mpc$z=6.4$03, while allowing SMBH masses at $z=6.4$04 to exceed $z=6.4$05 rather than plateauing at $z=6.4$06 (Sen, 14 Sep 2025).

Taken together, these applications define GAMETE/QSOdust as a model family rather than a single immutable implementation. Across versions, its stable core is the coupled treatment of hierarchical halo assembly, star formation, black-hole growth, chemical enrichment, dust processing, and feedback. The principal scientific use of the framework has been to test whether the observed combination of large dust masses, rapid SMBH growth, strong outflows, and luminous FIR emission in the early Universe can be reproduced within a unified evolutionary scenario. Within the cited studies, the answer is affirmative, but only under specific conditions: efficient grain growth in molecular clouds, strong AGN-driven winds, and, for light-seed scenarios, very efficient angular-momentum loss in the accreting gas [(Valiante et al., 2011); (Pezzulli et al., 2017)].