A New Type of Stellar Feedback

Title: Stellar feedback from high-mass X-ray binaries in cosmological hydrodynamical simulations

Authors: M. C. Artale, P. B. Tissera, & L. J. Pellizza

First Author’s Institution: Instituto de Astronomia y Fisica del Espacio, Ciudad de Buenos Aires, Argentina

Paper Status: Accepted for publication in MNRAS

The commonly accepted theory of dark matter (called Lambda-CDM) provides us with good general understanding of how structure (filaments, galaxy clusters, galaxies, etc.) forms in our Universe. This is shown again and again in simulations of our Universe in a box, made by following the motion of dark matter particles and their clustering as controlled by gravity. However, problems arise in trying to reproduce the detailed baryonic properties (namely stars and gas) of real galaxies. One of the biggest problems is faithfully reproducing the star formation history of real galaxies, both in terms of when they form (early in the Universe or more recently) and how many form.  As it turns out, “feedbackprocesses are essential in reproducing properties of galaxies, including supernova explosions which can heat, ionize, and remove gas from galaxies (shutting off star formation). In many cases supernovae may be the dominant form of feedback, but, as the authors of today’s paper explore, it may not be the only important form of feedback, especially in the early Universe.

High-mass X-ray binaries (HMXB’s) are binary star systems consisting of a massive star orbiting either a neutron star or a black hole. These systems produce a significant amount of X-ray emission, originating from the accretion of the massive star onto the more compact companion; this is shown in an artist’s drawing in Fig. 1. In some cases, these systems can produce fast moving jets, dumping kinetic energy into the surrounding gas. Both of these can heat, ionize, and blow away gas within a galaxy, and in turn play a significant role in how stars form within the galaxy. The authors indicate that recent work has shown that these systems are more powerful in the early Universe, and may play a role in controlling star formation in early galaxies that will have a lasting impact for their entire evolution. For the first time, the authors implement a HMXB feedback method into a cosmological hydrodynamics code, and explore how it affects galaxy evolution in the early Universe.

Fig. 1:

Fig. 1: An artist’s conception of a black hole binary, where a massive star (right) is orbiting a black hole (left). The massive star is loosing gas to the black hole, forming an accretion disk and jets. (Image Credit: NASA)

Modelling the Feedback from HMXB’s

The authors use a version of the GADGET-3 smooth particle hydrodynamics (SPH) code that includes a chemical evolution model, radiative heating and cooling (allowing the gas to heat and cool by radiating away or absorbing energy from photons) , and methods for star formation and supernova feedback from Type II and Type Ia supernova. New in this work, however, is a model to simulate the feedback from HMXB’s. Due to the finite resolution of the simulations, they construct a model that accounts for a population of HMXB’s in a galaxy, rather than individual HMXB’s. Using known estimate for the number distribution of stars (known as the IMF), the authors estimate that about 20% of massive stars that form black holes in a given galaxy form in binary systems. In their simulations, these systems deposit 1052 erg of kinetic energy into their surroundings (about 10 times that of a single supernova explosion) , and spend about 3 million years radiating X-rays at a luminosity of about 1038 erg s-1. These numbers are not well constrained, however, but the authors tested a few different values and found these to produce the most realistic amount of feedback (as compared to observations).

The authors produce two primary cosmological simulations, evolving from high redshift down to a redshift of about 2 (about 5 billion years from present day), using only supernova feedback (designated S230-SN) and supernova feedback + HMXB feedback (designated S230-BHX). The authors use these two simulations to compare how a HMXB feedback model affects the total rate of star formation in their simulations, and the properties of star formation of individual galaxies in their simulations. Some of these results are discussed below.

Controlling Star Formation

Fig. 2:

Fig. 2: The star formation rate density in the entire simulation (i.e. how much mass is converted into stars per year per unit volume, computed for the entire simulation box) as a function of redshift for the SN only (blue, dashed) and SN + HMXB (red, solid) simulations. The horizontal axis is redshift, with lower redshift (z=2) closer to present day than higher redshift (z=10). The black data points are observations from Behroozi et. al. 2013. (Source: Fig. 1 of Artale et. al.)

Fig. 2 compares the effect of the SN (blue) vs. SN + HMXB (red) feedback models on the star formation of the entire simulation box as a function of redshift, from the early Universe (right) to z = 2 (left). The vertical axis gives the star formation rate (SFR) density in the entire simualtion. The SFR density is computed as the total mass of gas converted into stars per unit time per unit volume. At high redshift (early time), the HMXB feedback suppresses the total SFR density, and delays star formation towards lower redshift (later time). The higher SFR density in the HMXB model at lower redshifts is due to the fact that there is much more gas leftover at lower redshifts to be transformed into stars, since it wasn’t used up as dramatically at high redshifts as the SN feedback only model.

Fig. 3: Ratio of galaxy gas mass to dark matter halo mass for galaxies in the simulations as a function of their dark matter halo mass. This is plotted for three redshifts, z = 7, z = 6, and z = 4, for the SN only model (dashed) and the SN + HMXB model (solid). (Source: )

Fig. 3: Ratio of galaxy gas mass to dark matter halo mass for galaxies in the simulations as a function of their dark matter halo mass. This is plotted for three redshifts, z = 7, z = 6, and z = 4, for the SN only model (dashed) and the SN + HMXB model (solid). (Source: Artale et. al. 2015 )

The three most significant (in terms of mass) components in a galaxy are its dark matter halo, gas, and stars. The two feedback models have a significantly different effect on the amount of gas contained within galaxies over time. Fig. 3 shows the ratio of galaxy gas mass to galaxy dark matter mass as a function of the galaxy’s dark matter halo mass over three redshifts, z =7 (green triangles), z = 6 (blue squares), and z = 4 (violet circles). The SN only simulated is shown with the dashed lines, and SN + HMXB with solid lines. As shown, the SN only simulations have less gas at all redshifts and at almost all halo masses. In the SN only model, the higher SFR at early times uses up more gas, and drives bigger outflows, causing overall less gas to remain within the galaxy. The HMXB + SN, however, reduces SFR at early times without driving large gas outflows, allowing for gassier galaxies that have gradually increasing SFR (Fig. 2) towards low redshifts, as the effectiveness of HMXB’s decreases.

Finding the Right Balance

Getting feedback and star formation right in simulations (in part) means reproducing star formation history over cosmic time. The authors used HMXB with SN feedback to reduce star formation in galaxies at high redshift (in better agreement with observations than SN feedback alone), while maintaining it at low redshift (i.e. delaying star formation). This is a promising step in developing a complete model of feedback in galaxies.

About Andrew Emerick

After braving the harsh long winters of the great white north as an undergraduate at the University of Minnesota, I am working through my first year as a graduate student at Columbia University, in the balmy NYC. My research as an undergraduate and now has been mainly focused on galaxy clusters. I am currently working with Dr. Greg Bryan and Dr. Mary Putman using simulation and observation to map and better understand warm gas in galaxy clusters, as a possible probe of cluster formation and evolution. Whenever I’m not in the midst of solving the world’s problems, you can find me rock climbing… which is a surprisingly easy hobby to maintain despite being right in Manhattan.


  1. Interesting! It is amazing that binaries might play a major role in stellar feedback.

  2. How well studied/understood are HMXBs? It seems like we haven’t got everything figured out, since some of the values for their properties have to be estimated here, but is this sort of system frequently observed or more rare?

  3. Here I thought supernova explosions themselves constituted the most energetic events in the universe, but these binary systems are thought to expel several times that amount. I can see how they may have played a bigger role than once thought in the early cosmic structure.

  4. Have HMXBs been observed or are they just theoretical predictions? If so, how common are they?

  5. What kinds of computers are needed to run these types of simulations? Do they require a lot of processing power and/or special systems?

    • Simulations like these are generally run on national level high performance computers that usually fall under the category of supercomputers. Using these systems often means applying for “time” on them, in a similar fashion to a researcher applying for grant money or an observer applying for time on a telescope. The advantage of using a supercomputer is that you can use many many many processors at once on your simulation. As an example from my own research, a recent simulation I ran took about a full week to compute, running on 512 processors. This would take roughly 86,000 hours to run on a single processor, or about 2.5 years on a modern 4 processor laptop!!!! Not all simulations are this computationally expensive, many can be run in much much less time, yet there are also some that can take months to run, even on supercomputer!

  6. I know HMXBs are much more energetic than supernovas, but are they also brighter? If so, for how long?

    • Supernova explosions, at their brightest, have luminosities that are a factor of about 1000 brighter than the HMXB sources , but the latter last quite a bit longer. Supernova explosions drop in brightness by a factor of 100-1000 after a few hundred days. The lifetime of HMXB’s is somewhere around million years. In terms of total energy dumped into their surroundings, a single supernova deposits roughly 1051 erg, and these systems about 1052 erg. However, HMXB’s are much rarer.

  7. It is interesting how different the results for different models actually are. Is it possible to further constrain these models with more improved observations?

    • Yes it is! There is still a lot to be learned observationally and theoretically about HMXB’s. Many of the “free parameters” used in this work for the HMXB feedback were motivated by our current best understanding of HMXB’s, but this could be greatly improved (especially during the early Universe).

  8. When the authors say that they “estimate that about 20% of massive stars that form black holes in a given galaxy form in binary systems” does this then mean that there can be more than one black hole in a given galaxy? I don’t think I’ve ever heard about multiple black holes in a galaxy before!

    • That is correct! Black holes (like those discussed here) fall under the category of “stellar mass black holes”, as their masses are on order of a few to ten times that of our Sun. These are formed in the deaths of massive stars (greater than about 20 solar masses). There can be many of these in a given galaxy. In fact, there is estimated to be up to 100 million stellar mass black holes in our galaxy alone. However, what there is only about one of in every galaxy is a super massive black hole (SMBH), which can have a mass of hundreds of thousands to billions of solar masses. These are suspected to inhabit the center of (nearly) every galaxy, including our own Milky Way.

  9. Is the phenomenon where HMXB feedback suppresses the total SFR density at all related to the theory that jets from black holes suppress star formation? Is there a special property of the binary that makes the two phenomena different?



  1. How black hole and merger can kill a galaxy | astrobites - […] supernovae (SNe) or massive stars, collectively known as stellar feedback? Various research (see this bite, for instance) has shown that…

Leave a Reply