COMMON ENVELOPE SIMULATIONS

Papers Relevant for dependence of initial separation on CE outcomes and transition from RLOF phase to CE phase)

​Iaconi, Reichardt, Staff, De Marco, Passy, Price, Wurster & Herwig 2017
​MacLeod, Ostriker & Stone 2018a
​MacLeod, Ostriker & Stone 2018b
​Reichardt, De Marco, Iaconi, Tout & Price 2019
​MacLeod, Vick, Lai & Stone 2019
​MacLeod & Loeb 2019
​MacLeod & Loeb 2020

Goals

  1. Achieve a numerically stable simulation with separation equal to Roche Limit separation
    • Calculated to be 108 Rsun for our fiducial RGB primary (radius 48.1 Rsun, M_1=2 Msun) and secondary (M_2=1 Msun).
    • It would be fine to gradually increase the separation in successive runs, thus working our way out to 108 Rsun, if necessary.
  2. Low ambient density
    • But ambient pressure needs to be high to cut off stellar pressure profile and thus avoid inability to resolve pressure scale height
    • Hence usual problem is that by making density low, temperature gets high and timestep gets very small, making the simulation slow.
    • Is there a way around this problem?
      • One possibility is to just be patient!
      • Another is to include a hydrostatic atmosphere (typically with some constant temperature—see MacLeod+18a, but they use a spherical mesh)
        • This atmosphere cannot extend all the way to the mesh boundary because produces gradients at corners that crash the code (results from 3 years ago)
        • Can try making the atmosphere transition to a uniform ambient medium at some radius (See Run 199, below)
  3. Rotation of the primary
    • The Roche analysis is technically only valid for stars that are in corotation with the orbital angular velocity
    • I have not yet tried to give the star an initial rotation (an important step in the near future)
    • Part of the reason the primary star distorts in some of the simulations below may be physical and related to the results of MacLeod+19a
  4. Stable primary core
    • We have seen with recent models (AGB run) that this can be achieved by putting higher resolution at the core
  5. Larger simulation domain
    • the strategy I am using is to double the mesh size while reducing the base resolution by a factor of two (so 5123 would stay as 5123), and increasing maxlevel by 1. (Alternatively, one could keep the base resolution the same (5123 becomes 10243) and keep maxlevel the same, which would result in ~8 times more base cells but one less AMR level.)

Comments

  • Goal 1 is the main goal.
  • Goal 2 is desirable but not essential in the short term.
  • Goal 3 is also desirable but not essential in the short term, though it may make achieving goal 1 easier (see below).
  • Goal 4 we know how to achieve at a high enough fidelity for the time being.
  • Goal 5 is also desirable but not essential in the short term.

RLOF+CEE Simulations

  1. Continuation of old runs (only the last chombo of each had been kept!)
    1. Run 195: Separation of 1.5 times old separation, restarted from frame 50 of Run 161 with higher density thresholds for refinement and one extra refinement level for very high density—> completed up to frame 201
    2. Run 201: Roche-limit separation (=2.2 times old separation), restarted from frame 53 of Run 160 with higher density thresholds for refinement and one extra refinement level for very high density—> completed up to frame 251
    3. Run 202: Roche-limit separation (=2.2 times old separation) with low density ambient (67 times lower than fiducial), restarted from frame 27 of Run 162 with higher density thresholds for refinement and one extra refinement level for very high density—> completed up to frame 57
  2. New runs
    1. Run 199: Setup with smoothly matched hydrostatic atmosphere matching on to constant ambient at larger radius and density thresholds for refinement. >>Crashes ~immediately due to high pressure gradients just outside primary. Rerunning now with AstroBEAR default refinement. —> get insufficient memory error
    2. Run 196: Setup with low constant ambient, refined on density, eventually slows down a lot. Restarted from frame 29 with AstroBEAR default refinement. Runs fast but chombo sizes are HUGE since much more volume is refined. (First I tried qTolerance 0.2, then later changed to 0.4 to reduce file size. DesiredFillRatios set to 0.7.) Ran but then crashed on a restart when reading chombo (not sure why). Restart from frame 33 with qTolerance 0.2, ran up to frame 40. After that slurm and chombo output stopped, presumably because of insufficient memory. Chombo file sizes >130 GB and increasing. Get problem where no slurm output or chombos produced (in the past this problem was found to be caused by insufficient memory).

Run 195

Movies of gas density slices

  1. ​Face-on, simulation coordinates, frames 0-40 (old movie of Run 161 from ~two years ago))
  2. ​Face-on, simulation coordinates, frames 50-201
  3. ​Face-on, simulation coordinates, frames 50-201, zoomed in

Snapshots of Pressure and Temperature

  1. ​Pressure frame 50
  2. ​Temperature frame 50
  3. ​Pressure frame 70
  4. ​Temperature frame 70
  5. ​Pressure frame 90
  6. ​Temperature frame 90
  7. ​Pressure frame 140
  8. ​Temperature frame 140
  9. ​Pressure frame 201
  10. ​Temperature frame 201

Snapshots showing mesh

  1. ​mesh frame 50 -- for frames 0-50 used more conservative density-based refinement
  2. ​mesh frame 51 -- for frames 51-201 used less conservative density-based refinement with additional AMR level for extra resolution near particles
  3. ​mesh frame 77
  4. ​mesh frame 50, zoomed in
  5. ​mesh frame 51, zoomed in
  6. ​mesh frame 77, zoomed in

Separation vs time

  1. ​Comparison with fiducial run (Run 143)

Run 201

Movies of gas density slices

  1. ​Face-on, simulation coordinates, frames 0-251 (frames 0-53 are from ~2 years ago and frames 54-251 are new)
  2. ​Face-on, simulation coordinates, frames 53-170, now showing mesh

Snapshots with and without mesh

  1. ​Frame 53 with mesh
  2. ​Frame 53 without mesh
  3. ​Frame 53 pressure
  4. ​Frame 53 temperature
  5. ​Frame 54 pressure
  6. ​Frame 54 temperature
  7. ​Frame 116 pressure
  8. ​Frame 116 temperature

Run 202

Movies of gas density slices

  1. ​Face-on, simulation coordinates, frames 0-57 (frames 0-27 are from ~2 years ago and frames 28-57 are new)

Snapshots

  1. ​Frame 27 Density
  2. ​Frame 27 Density extended colorbar
  3. ​Frame 27 Pressure
  4. ​Frame 27 Temperature
  5. ​Frame 27 Temperature with mesh
  6. ​Frame 28 Density
  7. ​Frame 28 Density extended colorbar
  8. ​Frame 28 Pressure
  9. ​Frame 28 Temperature
  10. ​Frame 28 Temperature with mesh
  11. ​Frame 57 Density
  12. ​Frame 57 Density extended colorbar
  13. ​Frame 57 Pressure
  14. ​Frame 57 Temperature
  15. ​Frame 57 Temperature with mesh

Run 196

Snapshots

  1. ​temperature frame 25
  2. ​temperature frame 25, with mesh
  3. ​density frame 29
  4. ​density frame 29, with mesh
  5. ​temperature frame 29
  6. ​temperature frame 29, zoomed colorbar
  7. ​density frame 40
  8. ​density frame 40, with mesh
  9. ​pressure frame 40
  10. ​temperature frame 40
  11. ​temperature frame 40, with mesh

Comparing setup to that of Run 195

  • Recall differences in setup from Run 195:
    1. twice bigger simulation box than Run 195 (and base resolution coarser by factor of two)
    2. extra AMR level not present in Run 195 for refinement around particles
    3. start out with higher density threshold for bulk of envelope so not as much of ambient is refined compared to Run 195, but after frame 29 more of ambient is refined than before frame 29 (switched from density threshold to default gradients-based refinement)
    4. ambient density is 6.7 times lower than for Run 195 (and 6.7 times lower than at the stellar surface)

Results and Interpretation

  • Less stable at stellar surface than Run 195.
    • Is this partly caused by density jump between surface and ambient medium?
      • But surface is also less stable than Run 202, where ambient density was 10 times lower still, so this is probably not the reason.
    • Is it related to refining less of the ambient at the beginning of the simulation than for Run 195?
      • From the snapshots, it does seem like numerical instabilities form where the mesh transitions between AMR levels and propagate inward, "squeezing" the star along the rectangular mesh
      • It seems like more refinement of the ambient medium helps to avert this problem
        • This is consistent with what MacLeod was finding (that lower base resolution exarcebates this effect, also seen in his sims, private communication)

Conclusions

  • Tentatively, it seems necessary to refine very conservatively such that much of the surrounding ambient is refined at the beginning of the simulation before the dynamical plunge-in phase.
  • When the dynamical plunge-in phase begins, this can be relaxed and the refinement can be concentrated on the stellar material only.
  • Fundamental oscillation modes can be excited when the primary does not rotate synchronously with the orbit at the Roche limit separation (MacLeod+19). This instability is likely being seen in our Runs 201 and 202. Initializing the primary in synchronous rotation might drastically improve stability.
  • Seeing as how the numerical instability occurs at the stellar surface, it would probably also be beneficial to increase the resolution there, if possible

Run 199

Snapshots (screenshots for debugging)

  1. ​density frame 0, with mesh
  2. ​density frame 0, with mesh, extended range
  3. ​pressure frame 0, with mesh, extended range
  4. ​pressure frame 0, with mesh, extended range, zoomed in
  5. ​temperature frame 0
  6. ​density frame 1, extended range
  7. ​pressure frame 1
  8. ​pressure frame 1, small range
  9. ​pressure frame 1, small range, zoomed in
  10. ​temperature frame 1

Comments

No comments.