Paper Summary
End-to-end study of the home and genealogy of the first binary neutron star merger
ABSTRACT
Binary neutron star mergers are one of the ultimate events of massive binary star evolution, and our understanding of their parent system is still in its infancy. Upcoming gravitational wave detections, coupled with multi-wavelength follow-up observations, will allow us to study an increasing number of these events by characterising their neighbouring stellar populations and searching for their progenitors. Stellar evolution simulations are essential to this work but they are also based on numerous assumptions. Additionally, the models used to study the host galaxies differ from those used to characterise the progenitors and are typically based on single star populations. Here we introduce a framework to perform an end-to-end analysis and deploy it to the first binary neutron star merger – GW170817. With the Binary Population And Spectral Synthesis (BPASS) codes we are able to retrieve the physical properties of the host galaxy NGC 4993 as well as infer progenitor candidates. In our simulations there is a >98% chance that GW170817 originated from a stellar population with Z=0.010 born between 5 and 12.5 Gyrs ago. By carefully weighing the stellar genealogies we find that GW170817 most likely came from a binary system born with a 13-24 M primary and 10-12 M secondary which underwent two or three common envelope events over their lifetime
Methods Take Home Points
Why does it matter if we use the same suite of predictions to infer the home galaxy properties and the progenitor routes?
=> Not all codes are built the same and different hidden assumptions can (will) creep up whether you know it or not. Also some of the most commonly used template SED libraries are pretty dated. In the supplementary information we compare the difference between the SED fitting results obtained with the same MUSE data when using classic BC03 templates Vs what we get with BPASS (see Supp. Info Section 1.4). BPASS isn’t “right” and the others “wrong”, but with this type of analysis the stellar models used to predict the SEDs are the same as the ones where we look for transient progenitors - it adds a level of consistency that was not possible before and will make comparison and interpretation with future results easier.
I used ppxf for our SED fitting. New hoki pipelines create BPASS SED templates (ppxf friendly) and automate tedious tasks
=> Although this paper took nearly 3 years to come together (including some COVID shenanigans), it should take only a few days to do this type of analysis in the future. I have released hoki v1.7 which contains the helper classes and functions I created.
=> The SED fitting is discussed in the main text and in Supp. Info. Section 1. We also compare to previous results obtained for NGC 4993 (summarised in Supp. Table 1).
I have created exhaustive tutorials on how to do SED fitting with BPASS -hoki. You can now independently to this work for your transients!
If you have any issues with BPASS-hoki SED fitting, don’t hesitate to get in touch! I will be more than happy to help you
Stellar Evolution Take Home Points
The main stellar population has metallicity Z=0.010 (half-solar in BPASS) with log(age) between 9.7 and 10.0.
=> So most of the stellar mass of NGC 4993 was born during Cosmic Noon. Beautifully average.
A variety of possible evolutionary channels, all with two episodes of Common Envelope (CE)
=> CE tends to be more common in BPASS than in other simulations. This is down to the fact that the BPASS CE is more efficient at removing the envelope and therefore systems can undergo CE with a lesser risk of merging (which would remove them from the pool of acceptable candidates). This has implications for final periods distributions too, although it is somewhat degenerate with the effects of supernova kicks. For an extensive discussion of Common Envelope in BPASS-STARS see Supp. Info. Section 2.4.
Different evolutionary pathways occupy distinct ZAMS parameter space regions but overlap in final parameter space (e.g. neutron star masses, Period-eccentricity post supernova 2).
=> This is not news, but Figure 2 in the main text and Figures 10, 11, 12 in the Supp. Info. illustrate our problem very well: figuring out the evolutionary channel of a system from the moment it died is ‘tricky’ and highly degenerate.
Ultra-Stripped Supernova kicks need-not-apply?
=> The numerical simulations of the supernova kicks used a Hobbs 2005 distribution and I also included a condition for Ultra-Stripped Supernovae (see Supp. Info. Section 2.1). For this particular analysis none of the suitable models went through an Ultra-Stripped Supernova type of explosion at the end of the life of the secondary. This is not to say that Ultra-Stripped Supernova are never necessary or uncommon, but they may not be ubiquitous for binary neutron star merger progenitors.
Triple Common Envelope Systems and “lucky kicks”
=> We found systems where the secondary undergoes two phases of CE (a case B and a case C later on). We also find that the matching progenitors with that type of channel have a different supernova kick distribution (same prior) than the systems where the secondary only goes through CE once (see Figure 3 main text).
=> It turns out that initially these systems seem like they would not be adequate neutron star merger progenitors: After the first CE they retain separations of >100 Rsol; the second CE phase further reduces the orbit but it is still quite wide (>1 Rsol). A kick that is too strong or in the wrong direction will therefore unbind them - they need a lucky kick to be viable. This results in the skewed kick velocity distribution (compared to the more typical routes) and the need for a high eccentricity after supernova 2 (they are the only models that stand out from the crowd) in order to merge quickly enough (high eccentricity reduces in-spiral time).
IMPORTANT DEFINITIONS
How is Common Envelope defined in BPASS-STARS
The common envelope phase is a very tricky one. It is hard to model and it is even more hazardous when you are modeling an inherently 3D hydrodynamic process in a 1D hydrostatic code.
In the code the CE starts when the radius of the star == or > the separation of the binary. Anytime I mention CE, that is what is implied. Further progress in CE modeling could affect some of this. An interesting case is that of the secondary stars that go through two CE events (a case B and a case C). It is likely that these two events behave differently given the vast differences in the state of the stellar envelopes at those times. Keep this definition in mind when you compare our BPASS results to the literature or to future, maybe more nuanced, computational definitions of the CE. Not all “CE” - as defined here - will be created equal and have the same impact.
Genealogy Vs Evolutionary Channel
In the paper I use both the terms “genealogy” and “evolutionary channel”: they are not synonyms. I use the word “genealogy” to refer to an individual, unique (in our simulations), stellar evolutionary pathway. Several genealogies can fall under the same “evolutionary channel”.
For example a system with M1 = 14, M2 = 13, P = 10 days, and another with M1 = 15, M2 = 13, P = 10 days can follow the same evolutionary route (e.g. case B CE) but they are different genealogies: their initial masses are different, the mass loss and orbital loss are different, the lifetimes are different.
In short, “genealogy” is quantitative and “channel” is qualitative.