Fly-by galaxy encounters with multiple black holes produce star-forming linear wakes

ASTRID is a cosmological hydrodynamical simulation performed using a new version of the Smoothed Particle Hydrodynamics code MP-Gadget. The simulation evolves a cube of $250h^{-1}{\rm Mpc}$ per side with $2\times 5500^{3}$ initial tracer particles comprising dark matter and baryons, and has currently reached $z=1.3$. ASTRID has a dark matter particle mass resolution of $M_{\rm DM}=6.7\times 10^{6}{h^{-1}M_{\odot}}$ and $M_{\rm gas}=1.3\times 10^{6}{h^{-1}M_{\odot}}$ in the initial conditions. The gravitational softening length is $\epsilon_{\rm g}=1.5h^{-1}\,{\rm kpc}$ for both DM and gas particles. The simulation includes a full-physics sub-grid treatment for modeling galaxy formation, SMBHs and their associated supernova and AGN (Active Galactic Nuclei) feedback, as well as inhomogeneous hydrogen and helium reionization. We refer the readers to the introductory papers Bird et al. (2022); Ni et al. (2022) for detailed descriptions of physical models deployed in the simulation.

Here we briefly summarize BH modeling, which is most relevant to our current investigation of runaway MBHs. BHs are seeded in haloes with $M_{\rm halo,FOF}>5\times 10^{9}{h^{-1}M_{\odot}}$ and $M_{\rm*,FOF}>2\times 10^{6}{h^{-1}M_{\odot}}$, with seed masses stochastically drawn between $3\times 10^{4}{h^{-1}M_{\odot}}$ and $3\times 10^{5}{h^{-1}M_{\odot}}$, motivated by the direct collapse scenario proposed in Lodato & Natarajan (2007). The gas accretion rate onto the black hole is estimated via a Bondi-Hoyle-Lyttleton-like prescription (Di Matteo et al., 2005). The black hole radiates with a bolometric luminosity $L_{\rm bol}$ proportional to the accretion rate $\dot{M}_{\bullet}$, with a mass-to-energy conversion efficiency $\eta=0.1$ in an accretion disk according to Shakura & Sunyaev (1973). 5% of the radiated energy is coupled to the surrounding gas as the AGN feedback. Dynamics of the black holes are modeled with a sub-grid dynamical friction model (Tremmel et al., 2015; Chen et al., 2022) to replace the original implementation that directly repositioned BHs to the local minimum of the potential. This gives well-defined black hole trajectories and velocities. Per this implementation, two black holes can merge if their separation is within two times the gravitational softening length $2\epsilon_{g}$, once their kinetic energy is dissipated by dynamical friction and they are gravitationally bound.

Galaxies in the simulation are identified with the SUBFIND algorithm (Springel et al., 2001). The large volume of ASTRID provides a rich galaxy population matching the host galaxy of the observed systems, from which we can conduct a comprehensive search for runaway BHs and star-forming wakes. We perform the search at two redshifts, $z=2$ and $z=1.3$ (the current lowest redshift of the simulation), with the latter being closer in time to the observed streak in VD23 at $z=0.964$.

Target galaxies: In Figure 1, we show the properties of $z=2$ galaxies in ASTRID and our selection criteria. To match with the properties of the main galaxy in van Dokkum et al. (2023), we focus on the star-forming (SFR$>1\,M_{\odot}/{\rm yr}$), compact ($r_{\rm half}<2{\rm kpc}$), and non-satellite galaxies with masses $5\times 10^{9}\,\,M_{\odot}<M_{\rm gal}<5\times 10^{10}\,\,M_{\odot}$ that contain at least 3 BHs with a total BH mass above $10^{7}\,\,M_{\odot}$. We apply the 3-BH criterion because it is necessary to produce a runaway BH. After applying these cuts, we narrow down our search to $29888$ (11399) galaxies at $z=1.3$ ($z=2$), and hereafter we refer to these selected galaxies as our “target galaxies”.

Runaway BH selection: We identify runaway BHs within the target galaxies above (i.e. the orange population in Figure 1). To quantitatively define a “runaway” MBH in these galaxies, we compute the ratio between consecutive apocentric radii between each satellite BH in the galaxy and the central MBH (similar to Hoffman et al., 2023). If we see a consistent and sudden increase in orbital size by a factor of more than three in at least three full orbits, then we categorize the MBH as a “runaway” MBH. Among the $\sim 11399$ target galaxies at $z=2$, we found $\sim 200$ potential runaway BHs that satisfy these criteria.

Linear star-forming wake selection: We look for radial linear star-forming features with the following method: we divide all the young stars (stellar age $<100\,{\rm Myrs}$) in each halo into 45 radial bins [10, 100] kpc from the central galaxy center. To match a $\sim 50\,{\rm kpc}$ observable linear feature, we select systems with at least $2\times 10^{6}\,M_{\odot}$ stars in each bin for 25 continuous radial bins. This gives us 514 (435) candidates with potentially observable stellar streaks at $z=1.3$ ($z=2$) respectively, which makes up $2-4\%$ of the target galaxy population (we will refer to these as “linear feature candidates”). We then visually inspect these potential linear features to select the ones with large signal-to-noise, resulting in $\sim 30$ visually-confirmed linear features at each redshift (note that this number of viable candidates is likely a lower limit as we only look for linearity along the $x$-$y$ and $x$-$z$ planes).

van Dokkum et al. (2023) proposed that one of the most likely mechanisms for producing the observed star-forming wake is a runaway BH induced by a three-body encounter. Following this proposal and the works by Saslaw & De Young (1972) and de la Fuente Marcos & de la Fuente Marcos (2008), we look for possible associations between linear star-forming features and runaway BHs. However, upon an initial search through the target galaxies (orange points in Figure 1), we do not find any visible stellar features along the wakes of the runaway BHs.

As the BH-induced star formation may be more prominent within a denser circum-galactic medium (CGM) than that in our target galaxies, we relax the upper limit on the galaxy mass and did another search among MBH triplets in larger halos (the sample of MBH triplets from Hoffman et al. (2023)). In these halos, the runaway BHs also tend to be more massive, and the formation of a stellar wake is less affected by the resolution limit. Figure 2 shows an example runaway BH candidate within a MBH triplet, found in a galaxy with mass $1.3\times 10^{11}\,M_{\odot}$. In this system, a BH with mass $1.3\times 10^{7}\,M_{\odot}$ is ejected from the host galaxy $\sim 100\,{\rm Myrs}$ after the initial formation of the MBH triplet at a speed of $\sim 700\,{\rm km/s}$, and it traverses a distance of $\sim 45\,{\rm kpc}$ from the galaxy center in another $\sim 100\,{\rm Myrs}$. Even in this case, we still do not see a star-forming signature associated with the wake of the runaway BH: we found at most $<10^{6}M_{\odot}$ total newly formed stars within $10\,{\rm kpc}$ of the ejected BH.

We note a few limitations of the simulation potentially responsible for this null result: the resolution of ASTRID cannot fully resolve the ejection due to three-body encounters, and BHs that “run away” are mostly disrupted by a blob of incoming stars/gas. This implies that the maximum speed reached by any of our runaway BH candidates is $\sim 800\,{\rm km/s}$ and decreases to $\sim 100\,{\rm km/s}$ at the target separation of $\sim 50\,{\rm kpc}$, barely reaching the speed to shock the surrounding gas (e.g. Dopita & Sutherland, 1996; Allen et al., 2008). Also, the simulation may lack the mass resolution to resolve gas shocked by a BH: the target runaway BH ($M_{\rm BH}\sim 10^{7}M_{\odot}$) is a few times larger than the gas particle mass, and so the gas over-density drawn by the BH may not be directly resolvable. A follow-up study would need to deploy higher-resolution simulations, to detect BHs ejected at a higher velocity in environments similar to our target galaxies to evaluate whether a similar star-forming wake can be produced.

Although we do not find visible star-forming wakes associated with the passage of runaway BHs, we do find a few tens of star-forming wakes in the simulation following the method detailed in Section 2. The natural question to ask is then: what mechanisms, if not runaway BHs, produce these wakes? By inspecting the simulation counterparts to the observed systems selected in Section 2.2, we find that most star-forming streaks originate from the fly-by encounter of a recently merged, dual-BH young galaxy with a more massive compact galaxy. We present some representative cases of such systems and their evolution in this section.

Figure 3 shows star and gas visualizations of two systems with linear star-forming features, and the mock images of the systems seen through the different filters of HST and JWST. In both cases, the young stars extend linearly to more than 50 kpc away from the main, compact galaxy (Galaxy 1, on the left end of each image). The young galaxy (Galaxy 2, on the right) is embedded in a stream of cold, dense gas with $T\sim 10^{4}-10^{5}\,{\rm K}$ clearly distinguishable from the background hot gas in the surrounding medium (second row), along which we see ongoing star formation and young stars formed within a few tens Myrs (fourth row).

Figure 3 also shows the mock HST and JWST observations, created by assigning spectral energy distributions (SEDs) to each star particle according to the Binary Population and Spectral Population Synthesis model (BPASS v2.2.1; Stanway & Eldridge, 2018). These SEDs are convolved with the filter transmission functions associated with the chosen filters. We smooth the stars with an SPH kernel and make 2D projections with pixel sizes matching the sensor for HST and JWST (0.049” for HST ACS, and 0.031”/0.063” for JWST NIRCAM short/long wavelength).

The stellar age along the wake follows a bi-modal distribution: half of the stars are relatively old with ages $\sim 1\,{\rm Gyr}$, and half are formed within $<100\,{\rm Myrs}$ during the emergence of the streak. The young stars form the linear feature when captured by the HST F606W and F814W bands (third row), while the older stars are better seen with JWST at longer wavelengths (fifth row). The redder stars may be a distinctive signature of the stripped galaxy scenario, which are unlikely to be found in the runaway-BH case.

To better understand the formation of the star-forming streak, we trace System 1 across a Gyr before and after the time when the linear feature is most prominent. Figure 4 shows the time evolution of the two galaxies involved in the linear stellar feature in a face-on view. The feature is a by-product of a fly-by encounter between two galaxies at a speed of $580\,{\rm km/s}$. The galaxies barely touch each other during their first passage (frame 1), after which Galaxy 1 remains unperturbed while Galaxy 2 develops elongated tidal arms on both sides. The feature formed around $150\,{\rm Myrs}$ after the initial passage between the two galaxies, and lasts for $\sim 200\,{\rm Myrs}$ (frame 2-4), after which the stars along the streak become old (frame 5), and the two galaxies experience a head-on collision (frame 6-7).

Our simulated linear features are extreme cases of the tidal arms/tails found in previous numerical studies of galaxy mergers (e.g. Toomre & Toomre, 1972; Barnes & Hernquist, 1992; Ploeckinger et al., 2018). Idealized simulations have shown that within these tails accumulations of gas and stars can be found (e.g. Barnes & Hernquist, 1992; Baumschlager et al., 2019; Shin et al., 2020). These clumps harbor molecular gas which provides a reservoir for new stars (e.g. Mirabel et al., 1991, 1992), and could be the origin of the blue knots along the VD23 system. Upon the close inspection of the $\sim 50$ visually-confirmed linear features, we find that the long tails are usually associated with dual-BH galaxies. $70\%$ of these tails contain close BH pairs, indicating a recent galaxy merger within the tail. Isolated galaxy simulations may not produce features like System 1 and System 2 if it takes two consecutive galaxy mergers to produce very long star-forming tails.

We also show in Figure 4 the trajectories of the BHs along the tail. BH3 has come within Galaxy 2 in a previous galaxy merger a few hundred Myrs before the wake formation. Then it keeps orbiting around BH2 during and after the production of the stellar wake, until the final merger between Galaxy 1 and Galaxy 2 (frame 7), after which both BH2 and BH3 ended up in wide orbits around BH1. We will examine the detailed evolution of the triple-BH system in the next section.

A common feature shared by our linear feature candidates is a dual-BH system along the streak merging into a central BH. Such MBH triplets are indicators of consecutive galaxy mergers producing the stellar streak, and if active, they may be also responsible for triggering the OIII emission along the streak. Here we examine the evolution of BHs along the linear features and their AGN activity.

Figure 5 shows the time evolution of the X-ray luminosity, BH mass, and the orbits of the three BHs associated with the merging galaxies in System 1 and System 2. In both cases, the BH in the compact galaxy is fairly luminous, maintaining an X-ray luminosity of nearly $10^{43}\,{\rm erg/s}$ for more than a Gyr. The second brightest BH (BH2) embedded along the linear feature also has a relatively high X-ray emission between $10^{42}-10^{43}\,{\rm erg/s}$. If the observed wake is produced by the scenario we show here, then BH2 is responsible for triggering the OIII emission along the wake. As mentioned in VD23, the amount of OIII emission seen would correspond to an AGN with $L_{X}>10^{43}\,{\rm erg/s}$ according to the scaling relation in Ueda et al. (2015). Our BH2s are slightly fainter for a $1.9\times 10^{41}\,{\rm erg/s}$ OIII luminosity, but still fall well within the Ueda et al. (2015) relation with the scatter and are capable of triggering the observed OIII signature.

The detectability of the third BH may depend on its orbits within Galaxy 2, but it is possible that BH3 was captured at the pericenter of the orbit and at it is brightest when the linear feature is produced (e.g. the case in System 1). In this (optimistic) case, follow-up observations may simultaneously see an MBH triplet associated with the stellar wake.