rss
Binary wind capture and accretion diks formation 2012
There was a meeting on Fri the 27th Jan '12. There we though about the relevant parameters for the binary simulation again.
Characteristic parameters:
- a, the separation of the stars, which we want to keep 20AU.
- AGB gravity
- AGB wind temperature, Tw
- AGB wind speed, vw, should be ~ 20km/s
- q=Mprimary/Msecondary
- gamma, which for AGB winds should be close to 1
- sigmas= gravity softening radius of the secondary / grid resolution = rsoft/ dx
- sigmaB= gravity softening radius of the secondary / Bondi accretion radius = rsoft/rB , where rB=2GMsecondary/(vw 2 + cs 2 + vsecondary 2), cs is the AGB wind's sound speed and vsecondary is the orbital velocity of the secondary with respect to the center of mass, located at the origin.
Test 1: Binaries
- Isothermal solver. It's the first time I use it for this problem
- q= 1.5, as before
- a=20AU
- Tw=1000K
- sigmas=4cells
- sigmaB=10, i.e. 40 cells per rB. This is accomplished with a grid of 83 computational units with a grid of 323+5particle refinement levels.
- vw=20km/s
- tfinal=5 orbits.
Progress:
| 29 Jan, 14:38. Test 1 started running on Sat the 28th morning in bluehive. I've seen some high cfl number reports after the secondary's gravity is turned on, which significantly decrease the timestep and slow the run. I've been (i) restarting with smaller cfl numbers, (ii) trying to progressively increase the secondary's gravity. The secondary has completed about .5 orbits so far. At this point, the gas that has been captured by the secondary star does not look like a disk: | 
|
I'm monitoring the progress and have 2 instances of the problem running.
30 Jan 8:15. Test1 is still running (see image), t=.8orbit. I've been finding high cfl reports which sometimes freeze the code and sometimes significantly reduce the timestep. I'm looking into the part of the code dealing with this, but in the mean time I've been running with cfls .1-.3, so the simulation is going slow. Another instance of the problem will start running in bluegene today (at some point).
| The upper right panel shows a zoom of the left panel. The bottom right panel show a zoom with velocity vectors. The flow does not look like a keplerian one at this point. Seems to be early to judge whether this is correct or not. | 
|
| 31jan12 9:17am. InterpOrder=2. The flow does not look like a disk but we're far from 5 orbits. Compare the image to the right with the one of the 29th Jan (two above) to see the differences with the interpOrder=3 case. It's going very slow for the reasons reported in ticker167 (https://clover.pas.rochester.edu/trac/astrobear/ticket/167) which I'm working on. In the meantime I've been running (in bluehive and have another instance of the sim waiting in bluegen's queue) with small cfls ~.09-.1. | 
|
| 1 Feb '12 7:55am. InterpOrder=2 continues, time=.8orbit. The flow about the secondary looks more uniform than before (see the zoomed in image →) | 
|
7 feb Running well in bluehive, 64 afrank procs. I've reduced one amr level so we can get data asap. This run includes some fixes and solver parameters that we've been discussing.
If test 1 fails (i.e. it produces a tilted or an amorphous disk) then in test 1.1 I will increase sigmaB for a fixed sigmas. If test 1.1 fails (ditto) then in test 1.2 I will increase sigmas and will keep everything else fixed.
Test 2: Disk at the centre of the grid
- Isothermal solver. Its the first time I will use it for this problem.
- Disk to ambient density contrast of 106
- sigmas=4cells
- extrapolated BC.
Progress.
30 Jan 8:22. Test 2 has completed 4 orbits (see images). The central part of the disk shows an inclination with respect to the orbital plane of ~ 30o, from t=1orbit on. I do not yet understand why this happens. The outer parts of the disk remain fairly axisymmetric though. These conditions do not vary too much in time. The grid is 323 +3amr, with a disk radius of 1 comp units.
| ~1 orbits | 
|
| ~2 orbits | 
|
| ~3 orbits | 
|
| ~4 orbits | 
|
Below is the early velocity field evolution, superimposed on the logarithmic gray scale of the density. Zoom in:
| edge-on view | pole-on view |
|
|
|
|
| The grid seems small for the problem. | 
|
31jan12 9:17am. Here's a movie of test 2 (amr+interpOrder=3) showing the disk plane velocity field superimposed on the log(density). cs=1. Arrows have a fixed length and are color coded.
http://www.pas.rochester.edu/~martinhe/2011/binary/31jan12a.gif
I see: a keplerian-like vel distribution at t=0. Then there's a fast radial expansion which, I think, results from the initial pressure gradient between the disk and the amb. Would this, or a similar process, happen for IC with a disk-to-amb density contrast=1 and a super Keplerian disk vel distribution (in that case there should be no poloidal expansion though)? Then the flow still shows a velocity that scales down with r and the density seems to show some spiral-like pattern. The right panel (zoom) shows gas located well within the Bondi radius. The initial 'red' central vels seem to quickly disappear and some of the green ones do too -the central disk seems to decelerate? After such vel change, the distribution of the central disk seems to change modestly.
31jan12 10:43am. Here's a movie of test 2.1 (fixed grid+interpOrder=2). Everything is as in the movie of test 2 (above).
http://www.pas.rochester.edu/~martinhe/2011/binary/31jan12b.gif
Group research meeting. After discussing the results of tests 2 and 2.1 we agreed I'll try:
- a colder disk, t=.1K, instead of 1000K
- taller cylindrical (not flared) disk, such that the disk height is ~ 2rsoft (it was ~.8 rsoft)
- more time resolution
- timefinal=2orbits, for the warping should happen in this time.
If the above still produce warping:
- phi=pi/2, i.e. make the disk rotate about the y axis, instead of the z axis to catch potential grid related bugs
- hydrostatic disk, with no ambient medium.
Find our previous research at:https://clover.pas.rochester.edu/trac/astrobear/blog/binary
Comments
Martin,
It looks as if the disk is expanding quite a bit? What is the thermal radius? And can you run a flared setup?
The thermal radius in test2 (central disk) is 10.7cu which is 10.7 times larger than the disk radius. I will run a flared disk version of it, but first I'll do the fixed grid test that we discussed on Friday.
I did notice one possible bug in the code that might cause assymetry but should only occur with InterpOrder = 3. I would recommend using InterpOrder = 2 for now…
I see. Tests 1 and 2, and all my previous sims, have used InterpOrder=3. I'll try InterpOrder=2 from now on.
If anybody wants to play with the data from test2 it is located (and has open permissions) at:
bgene:/scratch/mhuartee/astrobear2/26jan3/out
@Jonathan - what is nature of big for interporder = 3? @martiin - movies of V and rho in the mid plane would be great. @all - some weird things here
1) Why does disk expand like what is seen in that last image. The disk almost seems to be exploding. Does this make sense with our ICs? We are over pressured but gravity + rotation should be balanced at low latitudes.
Martin you might want to check your balance of forces in different directions.
2) Why is the only the inner portion of the disk warping? How many zones is the warp? Looks like 7 or 8? Does this tell us anything about effect of softening?
The interpolated edge states in z were second order in space instead of 3rd order when using PPM.
Inside the softening radius, stuff shouldn't be rotating very quickly - so not much angular momentum to resist warping…
@Adam. I do expect expansion form the IC because of the high disk-to-ambient density contrast of 106. Poloidal expansion should be stronger far from the center, hence the butterfly morphology perpendicular to the disk plane. Given that we see warping for a such a disk-to-ambient density contrast, I think we can rule out poloidal inflow as a cause of the problem. Don't we?
The central inclination seems to show a 'knee' of length ~ rsoft (zoom in at http://www.pas.rochester.edu/~martinhe/30jan1103.png). I think that this feature and the shape of the 'polar cones' depend on rsoft. The warped region is resolved by 12 cells from the symmetry axis. Any thoughts on why is the warping plane symmetric with respect to x=0?
Following Jonathan's comment. The vel field snapshots that I posted earlier do show the central material (<~rsoft) has lower angular momentum than the outer one. I will post disk plane velocity field movies soon.
Test2.1: interpOrder=2+fix grid, still shows the inclination of the central flow. Qualitatively, the warping seems to be less pronounced than in test 2 (amr+interpOrder=3), but the problem does remain. In both cases, the disk hight from the origin is 3cells < soft radius. Soft radius=4cells; we need more cells to resolve the height. Also, the grid volume (which is the one I've used for the other disk sims that worked ok) seems to be small for the problem (at least for densdisk=106). So I'd like to try:
test2.2:
thoughts?
If the disk is flared then disk height is no longer a constant - and goes to zero?
Right.
The BC in tests 2 and 2.1 (central disk) are actually extrapolated, not periodic.