Want to know what my research is about? Follow this thread 🧵 based on a 10min talk I've drawn for a meeting.
The talk was aimed at non-specialist space science colleagues (not the general public!). The slides were built up step by step, but I'm omitting this here & showing only the final graphs, less this becomes a 34-part thread. 11 is plenty enough!
So: "Understanding Winds of Massive Stars Using High Mass X-ray Binaries"
@vicgrinberg thanks so much for posting such a cool thread! Am curious if clumps with similar radius could have different masses? If so would an HXMB star affect the movement of each clump differently? Am guessing it doesn’t work that way — but if it did it would be fun to try and tease out what these data might show.
@stepheneb we approximate the clumps as spheres, in reality they are much more complex - it's overdensities in the wind that have typical scales, but are not necessarily all the same but have a certain mass and density distribution.
And we totally want to measure the distribution - so yes, it would be fun to tease this from the data! The data isn't good enough at the moment (the data is amazing, but we are looking for small things very far away moving fast), possibly with future instruments.
@vicgrinberg I wish I could follow this more closely, but I think I only grasped the gist of what your work involves. What I can confidently say is that your slides are beautiful as well as informative! I love them! Thanks for your work, looking forward to more informative posts about absorption spectra and binary high mass systems!
@done Apologies - the slides were made originally for a specialist audience. Pretty broad folks working in space science and not even necessarily in astro, but still ...
And thank you! The slides were quiet a bit of work but I also enjoyed making them. And was able to re-use quiet a few in other talks (though I've never done another one that was fully hand-drawn).
@vicgrinberg nothing to apologize about, take it as a lack of fundamental understanding on my part. I still grasped the general concept, and I think that the simplified diagrams were easier to understand conceptually than an actual graph with error bars etc. which may be more accurate, but overwhelming for someone naive to what the data represents. Thank you again for sharing!
O/B supergiants are the "superstars" of the Universe: they have much higher temperature & luminosity than our sun (& higher mass).
They live short & bright lives during which they can loose up to one hundred-thousandth of mass per year through winds. That's a VERY high mass loss and strongly influences how these stars develop and interact with their environment.
Yet, we struggle understanding these winds because their clumpy structure is hard to study in single stars.
Luckily, nature gave us High Mass X-ray Binaries (HMXB), where a black hole or neutron star (short: compact object) is close to an O/B star & accretes matter from its wind. The accretion process leads to production of high energy radiation.
These X-rays then pass through the remaining wind and can be used as a "backlight" to probe individual wind clumps & overall wind structure. To do so, we need to observe the X-rays with space-based X-ray telescopes, such as XMM-Newton.
There are two main ways how we can observe the effects of the wind on the X-rays.
looking at broadband energy spectra (say in the 1-100 keV range), we see increased overall absorption, seen as additional decline in soft X-rays in HMXBs vs. source where we can more directly look onto the compact object, without the intervening wind material.
This gives us information about the total amount of wind material along the line of sight.
@radioastronomer haha yeah, it's just scaling on the X axis 😂 Plus you folks also often work in frequency not those pesky wavelengths that make one invert the axies.
Looking at high resolution spectra at low X-ray energies where there are many line transitions from different ions, we can see narrow features (lines, RRC) imprinted in absorption or emission onto the continuum.
This gives us information about properties of the wind material such as ionization level and sometimes process, temperature, composition etc.
Why study HMXBs? They are key sources that touch all kind of astro areas (high energy, stellar, accretion physics) + we can often use data taken for other purposes to study the winds (absorption is noise for accretion physics studies).
They are systems where one supernova has happened & another will and are progenitors for gravitational wave events. We also want to know about accretion history of HMXBs (their total energy output) & use them as unique labs for complex physics.
That's why a few years ago - well, more like 5 or 6 by now - a few of us (including @fuerst and @pkretsch) started the X-Wind collaboration, aiming to bring together X-ray astronomers and experts in winds from massive stars from other wavelength & from theory in order to better understand winds in HMXBs: https://www.sternwarte.uni-erlangen.de/~grinberg/x-wind/
The problem is that the winds are complex and clumpiness and with them short-term variability we observe depends on e.g. line of sight, clump properties such as mass and radius, clump movement, wind ionisiation ... Enough parameters to feel totally lost and drowning!
So what I've done with my amazing colleague I. El Mellah is to try to disentangle a part of this complexity via simulations!
We simulate "absorption light curves", i.e. measurement of how much material is along the line of sight. Such lightcurves have two main properties: a typical timescale and a typical absorption variability.
We can get the typical timescales from the autocorrelation function and it's a good estimate for the clump radius. If we can measure absorption variability, we can then also get clump muss. But how measure that?
@vicgrinberg color-color diagrams, when you get down into the Poisson weeds, have some "fun" characteristics which I know @vlk has spent time thinking (and writing) about
@dburke@vlk -- oh, any paper you (any of you two!) want to point me towards? I know the work from Saeqa Vrtilek's group, but nothing that @vlk was involved in to my knowledge ...
And it's definitely something I keep exploring myself (and hope to get a bit more interesting stuff from).
@dburke@vlk Aaaah, yes, this one I know :) Sorry, I thought there was something more current about the color-color diagrams that I missed :) My case tends to be somewhat different - very bright sources, not faint ones, though comparatively short exposures. But still: thanks for the reminder! I may actually send a student down this rabbit hole :)
@vicgrinberg@vlk We are also thinking about this for the Chandra Source Catalog (as we worry about "low count" data there), but I don't think we've done much to explore the actual data/errors.
@dburke@vlk you ah, it's definitely a good one for catalogue work! I think there was some old work by Carpano et al., would need to dig into the lit (if you are interested) when not on the phone Sunday morning...
@vicgrinberg@dburke Yeah, no substantial updates to the BEHR paper since then (but stay tuned -- Andreas has Plans™), though we did do something similar to Saeqa's and your 2020 work with Stampoulis et al. 2019 (https://ui.adsabs.harvard.edu/abs/2019MNRAS.485.1085S/abstract) -- which is strictly speaking neither hardness ratios nor X-rays, but seeks to classify volumes in multi-flux space in astrophysically interesting manners.
@vlk@dburke definitely keep me in the loop! What I've been most interested in as the next step is finding a good way to compare to model predictions for tracks caused by changes in absorption + a characterization for how fast source moves along these tracks. Future music, though, a bit too many things on my desk + high functional load...
@vicgrinberg@vlk for CIAO we recently released the color_color https://cxc.cfa.harvard.edu/ciao/ahelp/color_color.html script that calculates this for a particular range of model parameters, but this is more on the "generating data" side rather than the "what does the data tell me" side
@dburke Nice! I use the "hardnessratio_simulate_grid" function from the isisscripts, but it's good to know that there is a python module since python is the programming language of choice for a number of folks I work with :) Could actually be really useful for my incoming master students, thanks! (May send him with questions to you ...)
Doug, an #RFE if I may! support also for log(S/M) v log(M/H) por favor? They are better behaved than (S-M)/(S+M) v (M-H)/(M+H) both because there are no hard boundaries at ±1 and it is easier to propagate errors and because you can subdivide passbands trivially (log(S/H) = log(S/M)+log(M/H) = log(S1/M)+log(S2/M)+log(M/H), etc)
@vlk@dburke and if we are at it, then also S/M; this is what we will typically use - there are reasons not to go to log space and not to use the sum/difference definition :)
@vicgrinberg@dburke But if there’s no arf/rmf you cannot compare to data. Best one can do is ratios of ph/s/cm^2, which is what I assumed Doug meant by “flux”
@vlk@dburke I was mainly afraid that by having fluxes as default users may first try to turn their measured count rates into fluxes (and we know how well that works for highly absorbed spectra; in my case with highly variable ionised absorption), so counts as default would be my way to go (possibly throwing an error or a warning that fluxes are calculated instead if no arf/rmf supplied) and a switch to fluxes.
@vicgrinberg@dburke Indeed, this is a hard problem to solve in the general case because the grid folds over and becomes multi-valued at the "corners". Dealing with edge cases consumes all your effort. Not to mention when counts fluctuations and background and RMFs and pileup and model misspecification sends data points outside the theoretical domain! Andreas Z is trying to solve it with ML. The trick as always is to solve what we can and defer the rest to the yung'uns!
@vlk@dburke I'm actually exclusively interested in the cases where it naturally folds over because of an underlying partial covering model... 😅 That's part of my problem.
@vicgrinberg@dburke And there is also the effort to sidestep Differential Emission Measures by looking at thermal segmentations in the solar corona based on ratios of pixel-wise intensities, Stein et al. 2016 SII 9, 535 (https://ui.adsabs.harvard.edu/abs/2015arXiv151204273S/abstract) but that's very far from unresolved compact objects emitting in X-ray!
@vlk@dburke I'll need to sit down and read this one in more detail, thanks! And seriously, more hints towards things you think may be relevant are welcome, you obviously have a somewhat different perspective than I and my collaborators on this!
Add comment