Modeling gravitational wave sources in the MillenniumTNG simulations

Abstract

(Edited) We introduce a flexible framework for building gravitational wave (GW) event catalogs in hydrodynamic simulations of galaxy formation. Our framework couples the state-of-the-art binary population synthesis code SEVN with Arepo-GW -- a module fully integrated into the moving-mesh code Arepo -- to assign merger events of binary compact objects to stellar particles in simulations by stochastically sampling merger tables generated with SEVN. Arepo-GW supports both on-the-fly operation, producing event catalogs during simulations, and post-processing, using snapshots from existing runs. The algorithm is fully parallel and can be readily adapted to outputs from other simulation codes. To demonstrate the capabilities of our new framework, we applied Arepo-GW in post-processing to simulations from the MillenniumTNG suite, including its flagship box. We investigate key properties of the resulting GW event catalog, built on SEVN predictions, focusing on comoving merger rates, formation efficiencies, delay-time distributions, and progenitor mass and metallicity distributions. We also examine how these properties vary with simulated volume. We find that GW progenitor rates closely track simulated star formation histories and are generally consistent with current observational constraints at low redshift, aside from a factor of $\sim 4.5$ excess in binary black hole mergers. Moreover, our binary black hole merger rates decline more slowly with redshift than current observational estimates for $z \lesssim 1$. Finally, the analysis of progenitor mass functions across different formation channels reveals only mild redshift evolution, while the binary black hole mass function displays features compatible with current observational determinations. These findings highlight the potential of our novel framework to enable detailed predictions for upcoming GW surveys within a full cosmological context.

Figures (19)

• Figure 1: Map of projected stellar surface brightness and GW progenitors merger surface density events for the BBH, BHNS, and BNS channels (from right in a clockwise direction, as indicated in the figure). The maps are centered on the most massive FoF halo of the MTNG740 run and have a side length of 5 Mpc. The stellar map has been obtained by mapping the $g$, $r$ and $z$ photometric bands to the blue, green and red color channels of the figure, respectively, with intensities linearly scaled over the interval $[20, 30]\,{\rm mag\,arcsec^{-2}}$. For GW progenitors merger rate events only compact binary mergers occurring at $z < 0.1$ are considered. Values are logarithmically scaled over the interval $[10^{-8},10^{-2}]\,{\rm pc}^{-2}$.
• Figure 2: Comoving SFRD as a function of time/redshift in four different boxes of the MTNG simulations, as indicated in the legend. The determinations of the cosmic SFRD as reported in 2014ARAA..52..415M and 2018MNRAS.475.2891D are also shown for comparison. Owing to the constant mass resolution of the analysed MTNG boxes, the SFRD is well-converged, with negligble differences among the runs. The MTNG SFRD peaks at earlier redshifts ($z \sim 3$ rather than $z \approx 2$) and shows a less steep decline towards lower redshift compared to the results presented in 2014ARAA..52..415M. Overall, the SFRD values in MTNG are broadly consistent with the determinations of 2014ARAA..52..415M and 2018MNRAS.475.2891D, although they tend to be lower by up to a factor of $\lesssim 2$ in the redshift range $z \sim 1$–$2$.
• Figure 3: Rate of GW progenitor events per comoving volume as a function of lookback time/redshift in the simulation MTNG740. The total progenitor merger rate and the contribution of different compact binary merger channels are displayed as indicated in the legend. Colored bands represent constraints on the low-redshift ($z\approx0$) merger rate of the different channels, as inferred from GWTC-4 high-confidence events GWTC4.
• Figure 4: Merger efficiency of compact binary objects per unit of stellar mass formed as a function of time/redshift in the simulation MTNG740. The total merger efficiency, along with the contributions from different merger channels, is displayed as indicated in the legend. The shaded region represents constraints on the evolution of BBH merger efficiency at low redshift ($z \lesssim 1$) derived by extrapolating the LVK collaboration limits shown in Fig. \ref{['fig3']}. This extrapolation assumes a redshift evolution of the form $(1+z)^{3.2}$GWTC4, and adopts the SFRD evolution given by 2014ARAA..52..415M reported in Fig. \ref{['fig2']}.
• Figure 5: Delay time distribution of GW progenitor events in the simulation MTNG740. The delay time distribution of the total merger events, along with the contributions from different compact binary merger channels, is displayed as indicated in the legend. All distributions are normalized to the total number of merger events (i.e., N$_{\rm tot} =$ BBH + BHNS + BNS), so that each delay time bin represents the fractional contribution of that specifc delay time to the overall number of events.