Raymond vs. AstroBEAR - Hydrogen Ionization and Recombination
Ionization (H —> HII)
I went through the Raymond shock code and figured out how it calculates the collisional ionization rate for H. I then plotted this (CION, black) as a function of temperature alongside the table that we use in AstroBEAR (AB Rate, red). In AstroBEAR, the units we assume on these values are cm3/s.
The two curves match pretty well which is great. Unfortunately, this also means that the discrepancy between the codes is elsewhere and still unknown. If we have the units wrong in AstroBEAR, then we are doing the scaling wrong which occurs after this ionization rate calculation. Also, the Raymond shock code has photoionization, but AstroBEAR does not.
I also made a plot of the ratio of the two rates. Even though they are on the same order of magnitude, you can see that the Raymond rate is often 10% - 60% higher than the AstroBEAR rate within the temperature range that is relevant to my simulations. I'm not sure if this difference is significant or not.
Recombination (HII —> H)
I did the same thing for HII recombination. This time, I put the y-axis on a log scale to make the difference more visible.
The two rates only differ at very high temperatures. My simulations do not use such high temperatures so this is fine.
Annual Research Review Meeting, 1/29/14
This blog post has been moved to a wiki page
Meeting Update 01/27/2014 - Eddie
- Worked mostly on paper edits. I should be finished with them tomorrow if not later today.
- Completed some different mach stem runs: ehansen01232014. I am going to put together a complete set of runs for different gammas and separation distances. I will have to rerun some of my old set ups, but other runs are already completed. I will also start exploring simulations in which I give one of the clumps a velocity.
- Another thing on my to do list this week is to dive into the Raymond shock code. More specifically, I will be looking for subroutines dealing with ionization and recombination and trying to compare these to what we have in astrobear.
- Lastly, I am going to see how far I can get with my 3D pulsed jet simulations. I want to make emission maps for a hydro run and for a beta = 1 run. Then, these could perhaps be used for Pat's HST proposal.
Meeting Update 01/27/2014 --Baowei
- Tickets
- new: #334(Help running on bluestreak)
- closed: none
- Resources
- grass: One disk is dead. Rich is wiping out the disks and rebuilding the array with the left 7. One spare disk (1TB) might be needed in the future.
- microphone of the laptop: Lost the plastic cover (outside the chip of USB) two weeks ago. The chip seems working OK. Mike Culver is helping us to wrap it again.
- Worked on
- #331 (Ablative RT): the failure (dE/dt !=0) of test with both hydro and diffusion seems due to hypre: http://astrobear.pas.rochester.edu/trac/astrobear/ticket/331#comment:5
- #317 (Quasi Periodic boundaries in a quarter plane): There are a bug related to the number of cores (exchanging of data?). Suggested by Jonathan, I added Divergence as diagnostic variable. Haven't been able to track the bug yet: http://astrobear.pas.rochester.edu/trac/astrobear/ticket/317#comment:8
CND with cooling and B-field - Marvin
This simulation of the CND was done with a vertical initial magnetic field with 100 microgauss. The first Animation shows the 3D mass density, after approx. half of the time the disk (at least the inner part) reaches a more or less settled state. However, there is a warping of the disk, but we have already seen this in the simulations without magnetic field. The second animation shows a face-on view of the mass density, a spiral pattern develops, but this feature is also present in simulations without magnetic field, as is the feature in the third animation, which shows the edge-on view of the gas velocity and the development of an outflow. The fourth and fifth Animations show edge-on and face-on views of the magnetic energy density with streamlines. The initially vertical field gets transformed into a toroidal field, the ambient medium hosts a chaotic magnetic field. The field strength in the disk reaches values of about 1 mG, which is indeed observed in the CND. So I think an initial field strength of 100 microgauss is a good choice. The magnetic field does not have a large effect on the gas density, as we see in the last Animation, which shows a face-on view of the beta parameter. It is almost always greater than one.
Animation of the 3D mass density
Animation of the mass density (face-on)
Animation of the gas velocity (edge-on)
Animation of the magnetic field (face-on)
Animation of the magnetic field (edge-on)
Binary simulation
Wind density
Wind speed
Gravity
Arctan(t)These are very zoomed in simulations. Actual simulation box is far beyond these movies.
- Density and velocity vector plot. There is fall back as well as outflow process at the same time.
- Velocity magnitude and velocity vector plot. The stationary region (characterized with blue color) is rotating. As the time goes on, stationary region also emerges in outer spiral.
The nature of this rotating stationary region is due to retardation.
some new 2D mach stem runs
I have started looking at some different cases for my 2D mach stems with 2 clumps. These all have the lower boundary extended to avoid the boundary effects we saw before.
So far, I have run the
case with clump separations of 20, 15, 10, and 5.I will do runs with gamma = 1.4 as well to have a total of 16 runs that fill a parameter space for different gammas and separations.
I also tried a run with the clumps offset. This is with gamma = 5/3, a separation of 15, and an offset of 4.
I won't do any more runs with offsets like this. Rather, I am going to start studying simulations in which one of the clumps has a velocity.
Meeting Update 01/21/2013 -- Baowei
- Tickets
- worked on
- Ablative RT module: fixed a bug in the open top boundary (ThermalConduction). by lowering the hypre tolerance to , hydro off results match the analytic value (#331). Working on double-check the hydro boundaries.
- Compiling Erica's code on BlueStreak.
Meeting Update 01/21/2014 - Eddie
- 3D pulsed jets seem to be working better with the density and velocity smoothing ehansen01162014
- the cell position is now in the output with the 'restart due to nan in flux' error
- ran the 2 clump, mach stem simulations for an intermediate gamma = 1.20. I did clump separations of 5, 10, 15, and 20 rclump. Images and movies will be posted soon.
Meeting update - Erica
Colliding Flows
Adjusted Christina's set-up (15-degree shear angle) slightly and sent to Bluehive where I was able to run it on 64 cores to about 30/200 frames. The amount of memory for the simulation is growing with the filling fractions, and the time it is going to take to complete there is > 1 month. I asked her to check the output and verify it looks good on Friday. Will ping her again today. Am ultimately going to have to move this to a bigger machine to cut the time it will take to finish significantly, but am currently running into some issues in compiling on bluestreak, will sit down with Baowei to work on it.
Also, as I have seen in the past, the memory being used by info arrays in the simulation is different than the actual memory usage on the nodes. I am seeing a difference of again ~30 GB over the 8 nodes I am using. When I ran a baseline test for the memory usage of the code (by running very small sim. with no multiphysics), I was getting the base usage as about 1.6 GB per node (or 200 mb per processor >> 17 mb astrobear is on disk), which should account for ~12GB over the nodes. Not sure about the additional memory here being used..
Paper submitted to ApJ and Astroph
Still playing with optimizing refinement
Can do geometric refinement, but when use with refinement object, get wonky behavior. Need to look into this a bit more.
Some documentation updates to the wiki
Updated standard out page
The job sizes page needs a bit more work still, Question - how many aux variables in 2.5 D and 2.5 D with angular momentum?
Also want to update the debugging page to include stack memory issues (definition of stack and the quiet seg fault you see with this type of limit, and also how fortran can run out of stack fast). Is default stack 10 mb on most clusters?, and this is per node right?
Updated cooling simulation - Marvin
I repeated the cooling simulation, but I changed some parameters: Gamma is now 1.4 (instead of 5/3) as it is expected for a diatomic molecule and the initial temperature is 200 K (instead of 300 K). The 3D mass density looks the same as last week, so I don't show it here. The Animation below shows the face-on view of the Temperature. It is roughly between 200 and 300 K, as it is expected for the CND. So I think our treatment of cooling is appropriate.
Animation of edge-on view of the Temperature
Face-on view of the Temperature:
Binary simulation
Our next intention is to consider the radiation effect of the AGB (primary) star in forming the fall back disk.
Roughly speaking, this will modify the hydrodynamical equations:
\[\frac{\partial \rho\mathbf{u}}{\partial t}+\nabla\cdot(\rho\mathbf{u}\mathbf{u})=-\nabla P-\nabla E-\nabla \Psi\]
Where
stands for radiation energy density. Radiation driven wind is actually radiation-driven dusty wind. A portion of Radiation energy is transferred to dust's kinetic energy then transferred to gas kinetic or thermal energy. According to Poynting Flux theorem, we know that radiation energy density drop by . Neglect the inhomogeneity and non-linearity in energy transfer between radiation and dust, dust and wind, we can combine the radiation effect with gravity.Assume the total radiation power of the AGB will fall in binary evolution. I use the following prescribed time dependent curve to illustrate this problem.
- Gravity Arctan(t)
- Wind speed, wind density or
The goal is to explore the parameters to get fall back phenomenon. Unfortunately, after two weeks of trial, the result is still not good.
Test A: gas is gravitationally bounded, no falling back process.
Test B: gas run away supersonically, no falling back.
The AGB seems to blow out all its envelop at some point, which looks as if I suddenly turned off the wind, but I did not.
pulsed jet working better with smoothed profile
I implemented a tanh function to smooth the density and velocity radial profiles at the injection boundary. The code runs much better now since there have not been any nans in the flux calculations. Also, there have only been a couple of high CFL restarts which is not a big deal.
I started with smoothing the profiles out to the edge of the domain. Clearly, this makes a very wide jet that is not what we want. But it was a good test to make sure what I had written was correct, and to see how the code handled it.
I then ran with smoothing out to twice the jet radius. This looks like a more realistic jet, but the head of the jet looks weird. You can see what looks like secondary bow shocks that protrude beyond the head of the jet. These are located at the original jet radius (aka where the smoothing begins).
I am waiting on a new run with smoothing again out to twice the jet radius. I fixed an error with the species densities. If it looks the same, I may try adjusting the pressure a little bit to see if the protrusions go away.
UPDATE
Fixing the species densities did not change much. Not sure if changing the pressure is the right way to go, but I am trying it anyways. I did another run with the smoothing out to 1.1 Rjet. This is the closest to what I actually want, and it ran without any issues.
None of the movies are uploaded yet because I have an issue with visit. For some reason when I try to use the 'save' options in visit, the images get yellow-ed out. This seems to only be an issue when running visit remotely from clover with 3D data. Running locally on clover with 3D data is fine, and running remotely from clover with 2D data is fine. Not sure what the fix is; could be a driver issue.
Summary for the current status of the Ablative RT project
- Jonathan suggested to test the flux and with hydro-off. Ideally we can extend the test to 3-cases
- currently we still have problem with the hydro-off test result will be posted in the third part.
- When doing the test we find a mismatch of
- Putting everything in computational units, the bottom flux calculated from Kappa1 and Temperature is 2.32E-4 and the bottom flux converted from Betti's data is 3.48E-4 (5.876e18 W/m2 → 5.876e21 erg/s/cm2 and divided by fluxScale=pScale*velScale). The difference is a factor of 1.5 which is , or gamma7 in the code, or as in Jonathan's blog post.
for the bottom flux and the equation:
2.One way of understanding this is that Kappa1~K0/Cv where
from Betti, but in AstroBEAR we have
and
- One easy fix for this is to multiply both size of the equation by gamma7: Since in the code there was already a gamma7 in the left side of the equation (which is a bug), we just include gamma7 in the Kappa1 as Jonathan mentioned in his blog. So now the Kappa1 becomes
which is 1.5 time larger than before.
- Since we still have problem with the hydro-off test, I can only show the effect of this 1.5 factor with bug-buried results
without gamma7 | ![]() ![]() |
with gamma7 | ![]() ![]() |
- Results for flux test: the way I ran the temperature limiting case tests is multiplying a small factor to the Temperature) with hydro on and off.
- limiting case with hydro off, multiply by 1e-17
- limiting case with hydro on, multiply by 1e-13 — limited by the BCs in BeforeStep, cannot go too small
![]() | ![]() |
- normal hydro-off test
paper reading
Skimmed through:
1.A fast, robust, and simple implicit method for adaptive time-stepping on adaptive mesh-refinement gridslink
2.Numerical Simulations of Radiatively-Driven Dusty Windslink
I attended MS graduation test on Friday last week (tested by Professor John Thomas and Professor Chuang Ren). It was good.
I attended NSF workshop today.
Meeting update 01/13/2014 -- Baowei
- Tickets
- new: #330(too many restarts for 3D pulsed jets)
- closed: none
- Resources
- grass: new controller card needed to rebuild the raid http://www.newegg.com/Product/Product.aspx?Item=N82E16816124070
- clover: got a wiki backup problem. Working on it with Rich
- Ablative RT
- there's an inconsistence of gamma7 (1/(gamma-1)) for the flux part between the equations and bottom flux. A gamma7 has to be included when calculating energy and flux to match the values in cgs with Betti's data. Jonathan posted a blog explaining this part here: http://astrobear.pas.rochester.edu/trac/astrobear/wiki/ThermalConduction . The limit case test passed but the non-hydro test for the energy increase ratio still won't match the heat flux.
Meeting update
- I looked at some 2D shear effects with the CF's (previous post)
- Working on getting the BE paper to adhere to all of the ApJ conventions
- Thinking about the computational difficulty I might run into with this shear sim in 3D. In thinking about the evolution of the mesh/filling fractions, I just ran this out in 2D for ~1.4 Myr. Looks like the mesh is going to grow a lot along the flow axis, and we see some grids forming in the splashed regions, that Christina says is not important to resolve. I think I should come up with refinement criteria that will prevent this from happening..
Cooling in the CND - Marvin
I am using the H2-cooling function that is part of the NEQ-cooling module. This cooling function considers rotational and vibrational lines for several wavelength. I assume that the disk consists to 100% of H2 throughout the simulation.
The first two Animations show the 3D mass density of the CND, once with and once without cooling enabled. With cooling the CND is much thinner and the central densities are larger by two orders of magnitude.
The next two Animations show an edge-on view of the disk temperature. Without cooling the disk heats up to about 80000 K (initial disk temperature: 300 K) and thus is indistinguishable from the ambient medium. With cooling the disk reaches a temperature of about 800 K.
Animation of 3D gas density, without cooling
Animation of 3D gas density, with cooling
Animation of edge-on view of the Temperature, without cooling
Animation of edge-on view of the Temperature, with cooling
3D gas density, with cooling:
Edge-on view of the Temperature, with cooling:
Hydro Colliding Flows with shear - Tests with fixed grid
Checking staircase effect
Running the 2D shear flow, fixed grid, shows the stair casing effect is not a function of the AMR grid placement. The effect seemed most prominent in the 30-degree shear angle case from my previous post, so am using that for comparison here. Fixed grid, 5122 is on bottom. Movie attached.
I also checked whether the AMR speed-up factor predicted by astrobear.log for the non-shear and 45-degree shear cases were valid. These were run on bamboo, 8 processors, machine not busy.
Checking AMR speed-up factor
Non-shear case
To make the table, I took the data from astrobear.log at the time-step right before the last frame of the simulation.
Simulation | Filling fractions | Info allocations | Efficiency | Speed-up | Actual Speed-up |
---|---|---|---|---|---|
642 + 3 | 0.135 0.614 0.762 | 18.3 mb 3.6 mb | 54% | 3.3 | ~5 |
5122 | na | 41.4 mb 5.2 mb | 82% | 5 | na |
45-shear case
Simulation | Filling fractions | Info allocations | Efficiency | Speed-up | Actual Speed-up |
---|---|---|---|---|---|
642 + 3 | 0.207 0.643 0.789 | 25 mb 5.4 mb | 44% | 1.6 | ~4 |
5122 | na | 41.4 mb 5.2 mb | 82% | 5.3 | na |
- Between the simulations of varying shear, I notice an increase in the filling fraction on level 0. This is because of the increase in the collision region. So, shear runs take longer.
- In both cases here, the amr speed-up factor as reported in the .log file for the AMR cases is an UNDER-estimate. AMR speeds up the simulation by ~5 times, even though in the high shear run it estimated the speed-up would be closer to 1.5.
- Current refinement criteria has the collision region nicely refined, but the splashing effects are not captured at high resolution. We see some features in these areas being lost by the refinement when compared to fixed grid. Is this something we would like to improve?
Resolution
Here is the fixed grid, 5122 45-degree resolution, and the mesh for the AMR (64+3) for comparison,
A possible concern here might be some loss of detail in the shear flows being splashed away from collision region with current refinement criteria (arrows in left plot). This might be a region that would otherwise form sink particles, but since the highest level of mesh isn't there, no sinks can be generated in these regions.
To get an idea of how large the flow region is, and the velocity field, here is a movie of vx:
- We see entrenched gas with high, positive speed in the x-direction in the splashing region. This likely would encounter similar gas from the other direction, which would collide and form dense regions of possible core-formation. This seems to indicate we might want to increase resolution into these regions. Other than that, should I try to reduce refinement in the collision region and check for loss of features in an attempt to speed of computation time?
- There is an apparent asymmetry in the high density forming regions, at least at early times, in the lower half of the collision region. This is curious.
- There appears to be some improper nesting occurring due to the tilted cross-section. This might be problematic should there be sharp gradients across this boundary, possibly resulting in nans.
Cone-shape
Looking at these simulations again today, I am curious about this cone-shape feature that appears at the splashing region of the collision area. I am interested in why this shape is so smooth, and remains so as it expands over time? By first impression, it looks too smooth. Would expect some turbulent features to appear along the boundary as it is pushed into the surrounding medium. Am left thinking it must have to do with the thermodynamics/EOS, which in this case is ideal with gamma = 5/3 + cooling.
- The clump-crushing time came to mind in thinking about this; the crushing time might be > the time of this sim. Further, the density of the ambient is lower than the material in the cone. It just doesn't seem to provide enough drag on the incoming material to disrupt the flow..
Self gravitating equilibrium
Self-gravitating equilibrium profile using density profile and external pressure
Mathematically, we are solving the equation of hydrostatic equilibrium
subject to the constraint that
.Since
we can rewrite the equation for HSE asand we can express the enclosed mass as
Now we can discretize equation 2 as
where
and equation 1 as
but a taylor series expansion shows that the error at each step in the integration goes as
so we need to limit out point spacing - or improve the accuracy of our discretization.
If we apply linear interpolation to our density profile, we can discretize equation 2 as
and equation 4 becomes a bit of a monster - and we have to worry about limits as r → 0.
if (r0 == 0d0) then dp=drho*((pi*drho*dr**4)/4d0 + (4d0*pi*rho*dr**3)/9d0) + (rho*(pi*drho*dr**3 + 2d0*pi*rho*dr**2))/3d0 else epsilon=log(r0/(r0+dr)) dp=((-36d0*drho*r0)*epsilon*M0*dr + (-36d0*drho*r0**2)*epsilon*M0 + (- 12d0*pi*drho**2*r0**5 + 48d0*pi*rho*drho*r0**4)*epsilon*dr + (-12d0*pi*drho**2*r0**6 + 48d0*pi*rho*drho*r0**5)*epsilon + (36d0*rho - 36d0*drho*r0)*M0*dr + (9d0*pi*drho**2*r0)*dr**5 + (17d0*pi*drho**2*r0**2 + 28d0*pi*rho*drho*r0)*dr**4 + (2d0*pi*drho**2*r0**3 + 64d0*pi*drho*r0**2*rho + 24d0*pi*r0*rho**2)*dr**3 + (- 6d0*pi*drho**2*r0**4 + 24d0*pi*drho*r0**3*rho + 72d0*pi*r0**2*rho**2)*dr**2 +(- 12d0*pi*drho**2*r0**5 + 48d0*pi*rho*drho*r0**4)*dr)/(36d0*r0**2 + 36d0*dr*r0) end if data(i,i_Press)=data(i+1,i_Press)+ScaleGrav*dp
Meeting Update 01/09/2014 - Eddie
Mach Stems
Ran the d = 20 simulation with an extended lower y boundary. The size of the y-domain was doubled from 40 to 80. The distance from the clump center to the bottom boundary was increased from 30 to 70.
Here is a comparison with the previous run. The run with the extended boundary is on the right, but I am not plotting the extended region.
So the bottom boundary does appear to be playing a role in regards to this nonphysical upwards motion. Now we just have to decide what to do about it. The easiest thing would be to extend the boundary (perhaps even farther than I already have), and then ignore the bottom portion of the domain. This would be ok for these 2D runs, but the 3D runs might be different. This also might not work for the runs where I put one or both clumps in motion.
3D Pulsed Jet
Jonathan implemented a new CFL and maxspeed check for the astrobear standard output. When there is a "high CFL restart", it now reports the maxspeed across all processors instead of just processor 0, and it gives the position. Here is an example:
High hydro CFL ( 0.21E+01 > 0.80E+00) found at position ( 0.1875E+00, 0.8929E-02, 0.9911E+00) on processor 3175, due to a maxspeed of 0.27E+03 - Restarting Step
In regards to my pulsed jet problem, this is great for figuring out where in the domain the maxspeed is getting too high. This is most likely due to low densities and thus high alfven velocities.
However, my simulations are also showing restarts "due to nan in flux". We thought these might be caused by the high CFLs and maxspeeds, but these restarts are not preceded by high CFLs. Attached to this post is a copy of the standard out from my most recent run on kraken.
The issue has probably evolved since we first started this problem, but the original ticket created for the 3D pulsed jet runs is ticket #289. The problem could also be related to tickets #301 and #302. For simplicity, I closed #289 and #302, and started a new ticket #330.
Unit converts for Ablative RT problem
Equation solved in Betti's code (SI units and Temperature in Joules)
Betti's document and is the normal specific heat capacity. And the flux is
where is Boltzmann constant and as inTo convert this to cgs units we write
That is
In AstroBEAR we define
andComparing the definition of
and we haveIn Betti's data, and so
Meeting update
- Revised BE paper
- Played with 2D Shear flows. Getting an error when trying to post a movie to this ticket:
Error
The server encountered an internal error or misconfiguration and was unable to complete your request.
Please contact the server administrator, webmaster@localhost and inform them of the time the error occurred, and anything you might have done that may have caused the error.
More information about this error may be available in the server error log.
Apache/2.2.22 (Ubuntu) Server at astrobear.pas.rochester.edu Port 80
Finding the no shear runs to run twice as fast, producing 2x the frames in the same amount of time. The AMR speed up is 2x bigger in the no shear run (~2-3) than in the shear (~1-1.5). The efficiency is 50% in no shear, and 30% in shear. The filling fractions 0.148 0.617 0.670 in no shear, and 0.155 0.641 0.742 in the 15deg shear, at the same frame. So I guess the filling fraction maybe the culprit here.
Movie is here, http://www.pas.rochester.edu/~erica/CompareShearAngles.gif
- Waiting to hear back from Christina for CF parameters
Binary Progress
Read following 3 papers:
- "Collimated Outflow Formation Via Binary Stars: Three-Dimensional Simulations of Asymptotic Giant Branch Wind and Disk Wind Interaction" by F. Garcia-Arredondo and Adam Frank link
- "Bipolar Preplanetary Nebular: Hydrodyamics of Dusty Winds in Binary Systems. I. Formation of Accretion Disks" by Nikos Mastrodemos and Mark Morris link
- "Bipolar Preplanetary Nebular: Hydrodyamics of Dusty Winds in Binary Systems. II. Morphology of Circumstellar Envelops" by Nikos Mastrodemos and Mark Morris link
They are all very interesting. I am inspired by these three papers.
Meeting Update 01/08/2014 -- Baowei
- Resources
- grass: Got problem with the array card, the raid is degraded, Dave is backuping the data (will take several more days). Very likely we need a new array card (300~500$) which will support bigger size of disks(10~20TB VS. the current 3.3 TB ) or a new machine.
- XSEDE: 0.64M SUs on Stampede for Martin's renewed allocation and 2.0M SUs left on Kraken for Adam's allocation. The renewal of Adam's allocation will need to be submitted before March 30th.
- Skype account: will need Adam's credit card since it's a recursive charge.
- Worked on
- installed AstroBEAR3.0 code on Rice's machines: DaVinCI and STIC (for Andy Liao) —schedule for Andy to come over?
- Ablative RT: http://astrobear.pas.rochester.edu/trac/astrobear/ticket/309#comment:50
CND with a very strong initial magnetic field - Marvin
These simulations have a strong (10-4 G) vertical initial magnetic field. Unlike my previous simulations, where the properties of the gas (e.g. gas density) were not that much affected by the magnetic field, here the development of the accretion disk seems to be dominated by the magnetic field.
The first animation shows the 3D mass density, after the first 20% of the simulation the disk starts to show strong spatial and temporal variations. Note also the density scale: The disk seems to be affected by a high mass loss.
The second animation shows the face-on view of the disk's plasma beta. The disk has regions with a high beta of 1000 or higher and regions with a beta near equipartition, but beta almost never falls below 1.
The last animation shows an edge-on view of the radial velocity (positive values only). The blue areas indicate inflowing material. We see that a conical outflow develops with speeds of about 700 km/s (these simulations were done without an outflow object). A very rough estimation gives outflow rates of 10-2 M_sun/yr. The total simulation time is 106 yr, so the huge mass loss is probably due to the conical outflow (total initial disk mass is 4x104 M_sun).
Animation of face-on view of the plasma beta
Animation of edge-on view of the radial velocity
3D gas density:
Face-on view of the plasma beta:
Edge-on view of the radial velocity: