New Work

1. Did force plots for runs 149 (half m2), 151 (fourth m2), 132 (subgrid accretion).
2. Found paper with analytic formula for drag force in presence of density gradient.
3. Continued AGB run on stampede 2: problem with code.
4. CM relative motion and energy plots for runs 149 & 151 (with Yisheng, ongoing).
5. Hubble proposal.

Force

Here are the updated notes on drag force which now include the same figures for the other runs: see ​notes, especially Fig. 9-12.

Here are the relevant papers with the analytic formula:

• ​Taam & Bodenheimer 1989, which quotes the formula from ​Dodd & McCrea 1952

AGB Run

See ​en_fig_run164.pdf.

Hubble proposal

See ​ce.pdf.

Next steps

1. Finish Hubble Theory proposal and submit by Friday.
2. Analysis for drag force work.
   1. Plot new analytic formula that takes into account density gradient.
   2. New 2D slices/movie.
      1. Color as force/volume and contours as density (before I plotted the reverse).
      2. Subtract out axisymmetric contribution to force/volume at each radius (radius measured from particle 2? particle 1?).
      3. Put force and velocity vectors on particle 2 in VisIt.
      4. Do all the same graphs for runs 149, 151 and 132.
3. AGB simulation.
   1. Fix bug that is slowing down code and eventually making it crash.
   2. Continue to plot a(t) and 2D slices showing bound/unbound gas.
4. Runs 149, 151 figures as in Paper II with Yisheng.
5. Posters for Baltimore conference.

Links to a working draft of the paper:

​figures

​paper draft

​references

on overleaf:

​https://www.overleaf.com/read/tcqgypntptny

AMR Line Transfer Tests

Running the same test seen ​here, which is IonizationFront module. Parameters are given in Mathematica notebook attached to that page, but for thoroughness:

Dependent on test (initial test parameters listed)

Working my way up the test hierarchy.

* Single ray (1D), single processor. Ran for very long simulation time, to give solid point of comparison.

* Multiple rays (2D, GmX(2) = 32, GxBounds(5) = 2.5d-1), single processor (final_time = 1d1, final_frame = 100), refined from x = 0.25 to 0.75.

* Multiple rays (2D), 2 processors (neighbor patch).

* Single ray (1D), single processor, 2 levels in middle-ish (child patch, MaxLevel = 1).

Result of refinement for optical depth

* Multiple rays (2D), single processor, 2 levels (child patch).

* Multiple rays (2D), multiple processors, 2 levels (neighbor(? - may need to force somehow) and child patches).

* Multiple rays (2D), multiple processors, 2 levels (neighbor and child patches), with domain boundary refined.

* Single ray (1D), single processor, for radiation pressure.

* Single ray (1D), single processor, for radiation pressure.

* Compare runs from ​combination test.

Comments: y and z directions are treated identically, so it's unnecessary to test both "2D" and "3D" cases (note that nDim is always set to 3 in global.data, resolution just set to 1 and size very small in "1D" and "2D" cases). Believe the above set should test all cases in each if statement.

Some working changes have been pushed.

New Work

1. More analysis on force in fiducial run 143.
2. Continued AGB run on stampede 2.
3. Finished postprocessing for forces for runs 149, 151 and postprocessing still underway for energy terms.

Force

Here are the new notes on drag force for run 143: ​notes.

Next steps

1. Write Hubble Theory proposal.
2. Analysis for drag force work.
   1. Put force and velocity vectors on particle 2 in 2D plots from previous post.
   2. Calculate contribution of drag force out to different distances from particle 2.
   3. Do all the same graphs for runs 149, 151 and 132.
3. Continue to run AGB sim and plot results.
4. Continue postprocessing work, also postprocessing to get forces in run 132 (sub-grid accretion run).

PN paper

• Prepared list of references and graphs to show. Writing sections for simulation setup and results. Share on overleaf?
• Forwarded last blog post to Orsola, waiting for her comments.
• High momentum runs are getting slow on Bluehive: 4 frames / 5 days using 120 cores. Maybe end the simulation at different times?
• Currently we have 220 frames (4400 days) for high momentum runs, and 300 frames (6000 days) for low momentum (need another five days for the non-cooling case to get there).
• Movies and graphs from the production runs are now password protected on my PAS server, what I learnt from Yisheng's instruction, same setup as he did.

CEJet

• Made scripts to calculate jet mass loss and boundary outflow. It seems the jet is not turned on, but keep in mind that the recent test runs have only a few frames each, so actually can't make any conclusion from this calculation.
• to-do: add tracer onto the jet.

AMR line transfer

Jonathan and I worked out the remaining logic last week. Short summary of what we discussed ​here. Also includes some thoughts on spherical line transfer.

Current to-do list is on AMR_line_transfer_2 branch in linetransfer_control.f90. Will need to clean up git when it's ready.

Other projects

• Radiation pressure
  • Needs comments, punching-up of how we address Cherenkov paper
• Charge exchange
  • On debugging to-do list, for weird recombination(?) in bottom right hand corner

New Work

1. Referee report
   1. Ready to re-submit?
2. CEJet module
   1. Reran tests on bluehive with much low ambient and larger initial separation by factor of two.
   2. Jet does not turn on! We don't know why.
   3. Could have something to do with density protections but lowering threshold doesn't solve it.
3. AGB run
   1. Continuing to run the simulation, now up to 185 days. Material is leaving the box so analysis will have to consider fluxes through the boundary.
   2. See new separation vs time plot below
4. Forces
   1. Wrote post-processing script, tested it against VisIt results (including a frame with the full-resolution data that I managed to obtain before VisIt crashed)
   2. Now running post-processing script (with reservation on bluehive) on data from fiducial run (143) as well as runs with ½ and ¼ the secondary mass (149 and 151, respectively)
   3. Before writing the script I tried using parallel visit on stampede as Gabe has done, but still there was insufficient memory, so gave up!
5. Alternative energy formalism
   1. Will write a draft when paper is accepted
6. Hubble Theory proposal
   1. How could it be relevant to Hubble data?
      1. Explaining post-CE binary separations in single and double degenerate systems?
      2. Explaining transient objects believed to be CE events?
7. XSEDE allocation
   1. Need a plan for using the roughly 185,000 node-hours that remain!
      1. AGB simulation in a very large box leading to PN initial condition?

AGB run

Next steps

1. Continue to run AGB CE simulation on stampede (3-4 day queue time) and plot results, including movies.
2. Analysis of force in runs 143, 149 and 151 using new force data from post-processing.
3. Continue to test CEJet module and get rid of the bugs! (with Amy).
4. Do post-processing for energy terms for runs 143, 149 and 151, and also unbound mass
   1. Need to redo fiducial run 143 with new script, which shows slight differences at 1% level to that used in Paper II—this is a small concern, hopefully will help to account for the 5% apparent violation in energy conservation.

A quick glance at the four models — density evolution

panels are not time synchronized

• Top left: Model 1, non-cooling, high-momentum
• Top right: Model 2, non-cooling, low-momentum
• Bottom left: Model 3, DM-cooling, high-momentum
• Bottom right: Model 4, DM-cooling, low-momentum (notice the scaling difference)

Numerical setup: ​link to Overleaf write-up

Some take-away points

1. Outflow is highly supersonic, and creates layers of shock.

​Example: Mach number in model 1

2. Outflow is highly collimated. Most of the outflow materials are redirected into the top and bottom lobes.

​Example: Outflow tracer and mesh in model 3

3. Top-bottom asymmetry is most dramatically shown in model 4, see the bottom right panel above (or click to show in full-screen).

Movies and plots from the four models:

​everything

Table of index

Naming rule decoded:

• "ADB" = "adiabatic", no extra cooling in the code
• "DMC" = "DM-cooling", cooling curve applied to T > 30000 K
• "_High"/"_Low" = "high-momentum outflow" or "low-momentum outflow"
• For each model, I plotted the following variables:
  1. Mach number with contours for selected Mach number
  2. Mesh and outflow tracer density
  3. Temperature and contour at 10000 K
  4. Density, labeled as "rho"
  5. Density at the orbital disk, labeled as "zrho"
• Numerals indicate certain zoom-in options:
  0. the entire simulation domain
  1. core area, focused on the CEE ejecta
  2. center, slightly zoomed out from the core
  3. lobes in the early stage
  4. tall, extended lobes in later stage

Current work on PN and CEJet

1. PN runs: restarting non-cooling this week, cooling low-momentum finished at 6000 days.
2. Analyse run time as a function of jets extend (TBD)
3. Write up a first draft paper (TBD)
4. (With Yisheng) build modulated code doing all the visit plots and movies. Can be use for both PN and CE simulations (in progress)
5. Analyse boundary mass loss rate in the jet runs (TBD)

Discuss the CEJet paper, Shilber et al. 2019. (see attachment)

Updates are posted on the accretion project page, here:

​https://astrobear.pas.rochester.edu/trac/wiki/u/erica/SteadyStateAccretionProblemPage

Radiation Pressure

Final measurements made. Here are related plots:

During bubble-out portion

During blowback portion

Charge Exchange

Compiles, runs, and does something. Need to test it — possibly determine steady-state mixing ratio for some density? Also still seeing that excess neutral square behind the ablated planet material. It almost looks like the recombination shadow we see in the other simulations, but since we don't have any ionizing radiation I don't see what would maintain the completely ionized state outside that region (perhaps it's not actually complete?). Could also be from some mixing, or more time spent there by the stellar wind… thoughts?

Uses HD209458b parameters, as given in ​this post.

Initial state

Final state

AMR line transfer

Made good progress last Wednesday. Latest commit just has a short to-do list, mostly cleanup. Primary functional code left is implementing an MPI derived type for sends and receives of rays, and a custom scatter and gather (because of differing array sizes).

Updated Plots from earlier post.

​https://guidarellig.github.io/ShellMasses.png

​https://guidarellig.github.io/TotalMassInside0p5.png

​https://guidarellig.github.io/AccretionRateFrom0p5.png

New Work

1. CE Jet (with Amy)
2. Continued AGB run on stampede 2

CE Jet Movies

Movies are in the frame of reference of secondary, with secondary at the center.

Here are the old ​notes on CE Jets.

Here is the Xsede ​proposal.

Below we compare two test runs, identical except that the jet is turned on in Run 014, whereas the jet is turned off in Run 015. The jet model is adapted from Federrath et al. 2014.

For both Run 014 and Run 015, parameters are as in the fiducial run 143 (Model A of Paper 1) except:
Base resolution = 64³;
Max AMR level = 4;
Region of maxlevel refinement = sphere of radius 57 Rsun around particle 1 (primary radius is 48 Rsun);
Size of smallest resolution element = 1.1 Rsun.

For Run 014, the jet parameters are:
Jet mass loss rate = 0.02 Msun/yr;
Jet velocity along jet axis = 10³ km/s;
Jet temperature = 3000 K;
Jet collimation half opening angle = pi/12;
Jet radius = 64 grid cells.

Other jet runs are similar to Run 014 except with certain differences, mentioned below.

Comparison of Run 015 with Run 014
​Density, face-on (left = no jet, right = jet)
​Density, edge-on through particles (left = no jet, right = jet)
​Density, edge-on view from particle 1 (left = no jet, right = jet)

Comparison of Run 015 with Run 016 (like Run 014 but 100 times the jet mass loss rate)
​Density, face-on (left = no jet, right = superstrong jet)
​Density, edge-on through particles (left = no jet, right = superstrong jet)
​Density, edge-on view from particle 1 (left = no jet, right = superstrong jet)

Comparison of Run 015 with Run 017 (like Run 016 but with ½ the jet radius)
​Density, face-on (left = no jet, right = superstrong jet with small radius)
​Density, edge-on through particles (left = no jet, right = superstong jet with small radius)
​Density, edge-on view from particle 1 (left = no jet, right = superstong jet with small radius)

Comparison of Run 014 with Run 016
​Density, face-on (left = jet, right = superstrong jet)
​Density, edge-on through particles (left = jet, right = superstrong jet)
​Density, edge-on view from particle 1 (left = jet, right = superstrong jet)

Comparison of Run 016 with Run 017
​Density, face-on (left = superstrong jet, right = superstrong jet with small radius)
​Density, edge-on through particles (left = superstrong jet, right = supersrong jet with small radius)
​Density, edge-on view from particle 1 (left = superstrong jet, right = superstrong jet with small radius)

Comparison of Run 014 with Run 018 (like Run 014 but with 1/100 of jet radial velocity
​Density, face-on (left = superstrong jet, right = superstrong jet with low velocity)
​Density, edge-on through particles (left = superstrong jet, right = supersrong jet with low velocity)
​Density, edge-on view from particle 1 (left = superstrong jet, right = superstrong jet with low velocity)

Comparison of Run 018 with Run 019 (like Run 018 but with ½ the jet radius, like Run 017 but with 1/100 of jet radial velocity)
​Density, face-on (left = superstrong jet with low velocity, right = superstrong jet with small radius and low velocity)
​Density, edge-on through particles (left = superstrong jet with low velocity, right = superstrong jet with small radius and low velocity)
​Density, edge-on view from particle 1 (left = superstrong jet with low velocity, right = superstrong jet with small radius and low velocity)

Comparison of Run 017 with Run 019
​Density, face-on (left = superstrong jet with small radius, right = superstrong jet with small radius and low velocity)
​Density, edge-on through particles (left = superstrong jet with small radius, right = superstrong jet with small radius and low velocity)
​Density, edge-on view from particle 1 (left = superstrong jet with small radius, right = superstrong jet with small radius and low velocity)

Comparison of Run 015 with Run 019
​Density, face-on (left = no jet, right = superstrong jet with small radius and low velocity)
​Density, edge-on through particles (left = no jet, right = superstrong jet with small radius and low velocity)
​Density, edge-on view from particle 1 (left = no jet, right = superstrong jet with small radius and low velocity)

Notes on CEJet runs

1. Notes on the results
   1. The late emergence of the jets with lower radius is hard to comprehend. It seems like one quadrant emerges first and then spreads to cover all 4 quadrants.
   2. It appears that the mass loss rate needs to be quite large or the radial velocity quite small to get an obvious effect on the morphology, but that's also because the radius is too large. We need to understand the dependence on jet radius and whether the effects we're seeing are physical or numerical.
2. Notes on the simulations
   1. Number of resolution elements per softening radius is only 2.2, leading to numerical instability of primary core region.
   2. Particle orbits are not reliable as their separation a increases even in the no jet case, due to insufficient resolution.
   3. Partly for these reasons, stopped movie around t=7 days.
   4. The simulations take a few hours to run up to this time with 115 cores on bluehive. Increasing the max AMR level by 1 or 2 should be possible.
   5. Also, we should start to move away from uniform resolution inside a sphere and toward full AMR.
   6. It would be good to introduce a tracer into astrobear for the jet material
   7. The ambient density and pressure could be lowered considerably. (I realized that the whole reason for the high ambient, namely to stabilize the outer layers, is not really relevant because that numerical instability that we saw when the scale height at the surface was not highly resolved was likely mostly due to grid effects when the star was fixed on the grid (e.g. during the damping run). Now we no longer do damping runs anyway. Amy and I have seen that the surface of the primary appears stable, even when the initial separation is increased and for much lower resolution than the fidicial run 143. The lesson seems to be that these pesky grid effects mostly go away when the star moves over the grid since errors accumulate randomly rather than uni-directionally. Bottom line is we should be able to get away with much lower ambient pressure and density, alleviating various headaches…)
   8. Currently, it is not possible to do restarts from the fiducial run since they used a different version of astrobear and the chombos are not quite compatible (the jet feedback module contains new outputs to the chombo that were not present in the previous version of astrobear used to run the fiducial CE run 143).
   9. The above item is not really a problem. Anyway, we should do a new fiducial run without jets, with low ambient, more efficient refinement strategy and bigger box.
   10. We should also try starting the secondary from farther out, as this case is likely to be interesting.
3. Notes on the analysis
   1. It would be good to plot planes parallel to the orbital plane (say 10 Rsun above it) to see the jet cross-section.
   2. Velocity vectors would be useful.
   3. Plots of normalized energy density (blue/red for bound/unbound) would be helpful.
   4. Put circular contour with radius=jet_radius around point particle 2 for reference.

Next steps

1. Test CEJet module, e.g. is it adding mass to the grid at the correct rate (Amy).
2. Continue to experiment with CEJet module on bluehive.
3. Continue to run AGB CE simulation on stampede and plot results.
4. Begin writing post-processing to obtain forces from full resolution data sets of Run 143 (fiducial), 149 (half secondary mass of fiducial) and 151 (fourth secondary mass of fiducial), as serial VisIt cannot handle the data sets (parallel VisIt is another option but we weren't successful when we tried to get it to work—anyway postprocessing means that in later simulations these quantities could be calculated on the fly, which would be useful).